June 20 to 23, 2017

Cannon Beach, Oregon

Safe Engineering Services & Technologies, Ltd.
(450) 622-5000 Toll free 1-800-668-3737
www.sestech.com
Copyright © 2017 SES ltd. All rights reserved. No part of any of the content of this document may be reproduced,
distributed, modified, or made available in any form by any photographic, electronic, digital, photostat, microfilm, or
other means, or incorporated into or used in any information storage and retrieval system, electronic or mechanical,
without the prior written permission of the copyright owner (SES) or the authors of the pages of the document to be
reproduced, distributed, modified, made available, incorporated into or used in any information storage and retrieval
system.
 1 PRESIDENT’S ADDRESS

It is with great pleasure that I welcome you to the 27th annual CDEGS Users’ Group Conference
hosted by Black & Veatch, at the Tolovana Inn in Cannon Beach, Oregon.

This year’s conference marks an important milestone: SES software Version 16. The ongoing effort
to modernize the interface of SES Software modules and enhance your user experience, which has
brought you entirely new modules such as RowCAD, SESeBundle, SESConverter, etc., in recent
years, has now come full-circle and touches all aspects of the SES Software workflow, starting with
the biggest change in the main CDEGS interface since 2005. Soil resistivity analysis in RESAP is
more ergonomic, adding flexibility in specifying measurement data or an initial estimate of the
soil structure. The computation can be launched directly from the new interface and the results
are immediately plotted, making the iterative process inherent to soil resistivity analysis much
smoother. The new interface for FCDIST offers more intuitive screens for different terminal types
(various configurations of transmission lines or of cables), access to the new SESLibrary of
conductors and better error-spotting tools. Even SESCAD is benefitting from the interface
improvements with a new soil model specification interface and Advanced Options screen.

In a long-awaited major advance that will benefit almost all SES Software users, a new
configuration and computation plot and report processor called SESResultsViewer is being
launched. SESResultsViewer will replace our venerable legacy Output Toolbox interface. It will
provide a more intuitive and user-friendly environment with additional capabilities that will
expand very rapidly thanks to its flexible architecture.

Efforts are also thoroughly underway to provide an enhanced set of software and modeling
documentation, starting with complete online help (F1), up-to-date “How To” Engineering guides
and quick start guides. SES is now creating the infrastructure to house on our website an expanded
and interactive Frequently Asked Questions area, and incorporate further video tutorials to our
documentation arsenal.

Version 16 will also mark the graduation of the recently introduced CorrCAD module, the new
TRALIN interface, and SESThreshold, from the BETA stage to production versions.

In addition to the new interfaces, SES Software continues to benefit from extensive and aggressive
R&D activities from dc (CorrCAD and RESAP) to high frequencies and transients along with
significant improvement in its core computation capabilities as well. Version 16 will see the
introduction of a new multi-layered hemi-spheroidal soil model in MALZ, and an improved finite
soil volume subdivision algorithm in MALT and MALZ. HIFREQ now has more flexible metallic
plates modeling capabilities, and can now optionally account for the actual position of coaxial
cables located inside pipe-type (multicore) cables in the calculation of the magnetic field produced
by those cables. The transformer model in SPLITS now allows the specification of different zero
and positive sequence impedance values. Similar models have been introduced in HIFREQ as
well.

Given the wealth of new interfaces being unveiled and becoming a bigger part of every user’s
workflow, and given that this will now be the most fertile grounding to continue improving the

Copyright © 2017 SES ltd. All rights reserved. Page i

software with new features, we are is eager to receive your comments and suggestions, as well as
feedback on your experience interacting with these new interfaces and analytical enhancements.
It is how we ensure we can provide you with the tools you need to get your work done.

We are grateful to Black & Veatch for hosting this conference, and to Mr. Doug Gilroy and
Mr. Francisco Muñoz, this year’s Users’ Group chairman and vice-chairman, to
Mr. Giancarlo Leone and Mr. Eric Diamond, this year’s secretary-treasurer and assistant
secretary-treasurer, for their hard work.

In addition to the organizers, I would also like to thank all participants for their contributions
through their questions and comments during the presentations, workshops and Q&A sessions,
or even between sessions in informal conversation. An additional thanks to those who took time
out of their busy schedules to prepare presentations and articles, in order to share their experience
and acquired wisdom. It is this exchange of knowledge and ideas that makes the conference so
valuable.

Farid Paul Dawalibi

Page ii Copyright © 2017 SES ltd. All rights reserved.

 1 CHAIRMAN’S WELCOME

27th annual CDEGS Users’ Group Conference

Cannon Beach, Oregon, USA,

June 20 to 23, 2017

Dear CDEGS Users,

Welcome to the 27th annual CDEGS Users’ Group Conference!

Black and Veatch is hosting this year’s conference at the Tolovana Inn, located in beautiful Cannon
Beach, Oregon (near Portland), June 20 - 23, 2017 (Tuesday – Friday).

We are very fortunate to have in attendance a diverse group of talented professionals, including
the expert staff from SES. This event is truly a learning experience and it will certainly enhance
the professional development of the attendees.

Our agenda includes presentations by Users and SES, as well as a vendor demonstration. On
Friday morning, SES will also provide a complimentary workshop. This service by SES has
historically been a great value to the users and an opportunity to improve your skills using the
CDEGS software.

Our hosts, Mr. Rishikesh (Rishi) Adkar and Mr. Swen Sorvala of Black and Veatch, have done a
great job in making the arrangements at the Tolovana Inn, and we look forward to the fine
accommodations along with the exceptional surroundings.

Be ready for volleyball on Tuesday and Thursday, beginning shortly after the last session of the
day. All attendees and their guests are also invited to the reception that will held on Tuesday
night.

Please join me in extending thanks to Black and Veatch, for their gracious hosting of this year’s
conference, as well as to Dr. Farid Dawalibi and the SES staff for their professional support and
considerable contributions to this event. Thanks are also extended to this year’s conference
executive committee team, Giancarlo (GC) Leone (Secretary – Treasurer), Francisco Munoz (Vice
Chair) and Eric Diamond (Assistant Secretary – Treasurer).

Be safe and enjoy the conference.

Thank you,

Doug Gilroy

Chair, 2017 CDEGS Users’ Group Conference

Copyright © 2017 SES ltd. All rights reserved. Page iii

 Table of Contents
President’s Address ...................................................................................................................... i
Chairman’s Welcome ................................................................................................................. iii

Part I: New Features

1 Improvements in SES Software ................................................................................. 1-1

1 Introduction ................................................................................................................... 1-1
2 New User Interface Programs ........................................................................................ 1-1
3 CorrCAD ........................................................................................................................ 1-6
4 Right-of-Way and RowCAD .......................................................................................... 1-8
5 HIFREQ: Junctions between Metallic Plates and Improvements to Magnetic Field Due
to Cables ............................................................................................................................... 1-9
6 New Leakage Current Criterion for Soil Models with Finite Volumes in MALT and
MALZ ................................................................................................................................... 1-11
7 MALZ: Multilayer Hemi-Spheroidal Soil Model ......................................................... 1-12
8 SPLITS: Zero Sequence Impedances for Transformers ............................................... 1-12
9 RESAP: Simulated Annealing Methodology ................................................................ 1-13
10 SESCAD: Composite Soil Volumes and KML Files ...................................................... 1-14
11 Other Improvements .................................................................................................... 1-15

2 Improvements and New Features in CDEGS ........................................................ 2-1

1 Introduction .................................................................................................................. 2-1
2 New Features................................................................................................................. 2-1
3 Improvements ............................................................................................................... 2-3
4 Conclusion ..................................................................................................................... 2-5
5 References ..................................................................................................................... 2-5

3 Improvements in RESAP ............................................................................................ 3-1

1 Introduction .................................................................................................................. 3-1
2 Datagrids ....................................................................................................................... 3-1
3 Computation Panel........................................................................................................ 3-1
4 Computation Options .................................................................................................... 3-2
5 Template........................................................................................................................ 3-3
6 Examine Session ........................................................................................................... 3-3
7 Shortcut Keys ................................................................................................................ 3-3
 8 Languages...................................................................................................................... 3-4
9 Example......................................................................................................................... 3-4
10 Conclusion ..................................................................................................................... 3-4
11 References ..................................................................................................................... 3-5

4 A New Interface for FCDIST ...................................................................................... 4-1

1 Introduction .................................................................................................................. 4-1
2 Application Tour ........................................................................................................... 4-1
3 Access to SESLibrary..................................................................................................... 4-8
4 Issues List ...................................................................................................................... 4-9
5 Output Session ............................................................................................................ 4-10
6 Example Files ............................................................................................................... 4-11
7 Conclusion .................................................................................................................... 4-11
8 References .................................................................................................................... 4-11

5 SESResultsViewer: New Configuration and Computation Plots and Reports

Processor ............................................................................................................................... 5-1
1 Introduction .................................................................................................................. 5-1
2 All You Need to Know about SESResultsViewer........................................................... 5-1
3 Conclusion ..................................................................................................................... 5-6
4 Acknowledgments ......................................................................................................... 5-7

6 Improvements in SESCAD ......................................................................................... 6-1

1 Introduction .................................................................................................................. 6-1
2 The Advanced Options Screen ...................................................................................... 6-1
3 New Editors and Components ...................................................................................... 6-2
4 The New Create Composite Soil Volume Tool .............................................................. 6-4
5 Importing KML Files in SESCAD.................................................................................. 6-7
6 Applied EMF and User-Defined Currents in Path Operations ..................................... 6-9
7 Warnings for Missing Soil Model ................................................................................ 6-10
8 French Version of SESCAD ......................................................................................... 6-10
9 Conclusion .................................................................................................................... 6-11
10 References .................................................................................................................... 6-11

7 New Features in SESLibrary ......................................................................................7-1

1 Introduction ...................................................................................................................7-1
2 New Databases in SESLibrary....................................................................................... 7-2
3 SESLibrary Used in Import Mode................................................................................. 7-5
 4 SESLibrary Used in Standalone Mode .......................................................................... 7-7
5 SESLibrary Multi-Language Interface .........................................................................7-13
6 SESLibrary Online Help .............................................................................................. 7-14
7 Conclusion and Future Developments ........................................................................ 7-14
8 References .................................................................................................................... 7-15

8 ROWCAD and ROW improvements ........................................................................ 8-1

1 Introduction .................................................................................................................. 8-1
2 KML Plot on Google Earth Map .................................................................................... 8-2
3 Combine/Construct Touch-to-GPR and Coating Stress-to-GPR Percentages
Automatically under Steady State Conditions...................................................................... 8-3
4 Compute Ground Impedance Automatically for a Component Inside of a Cable or
Group Path ........................................................................................................................... 8-4
5 Line Parameters Accounting for the Coating Characteristics of Buried Conductors with
Discrete Grounds at Specific Locations ................................................................................ 8-5
6 Model Terminal Circuit Ground in Total Interference ................................................. 8-6
7 Evaluating the Vulnerability of Arcing to Exposed Lines under Fault Conditions
Accounting for X/R Ratio .....................................................................................................8-8
8 Reference Conductor for Total Interference Computations Offers Various Filters to
Discriminate between Various Choices under Steady State Conditions .............................. 8-9
9 Customizations and Options in the Advanced Plot .................................................... 8-10
10 Miscellaneous ............................................................................................................... 8-11
11 Improvements in RowCAD ......................................................................................... 8-13

9 New Features and Improvements for Transformers in SPLITS ...................... 9-1

1 Introduction .................................................................................................................. 9-1
2 Computing Transformer Impedance Matrix Using Symmetrical Components ........... 9-1
3 New Features for Three-Phase Transformer Characteristics in SPLITS ...................... 9-8
4 Comparing Fault Currents for a Transformer with Equal and Unequal Symmetrical
Components........................................................................................................................ 9-19
5 Symmetrical Components Implementation in Three-Phase Transformer in MultiFields
……………………………………………………………………………………………………………………….9-21
6 Example Files .............................................................................................................. 9-26
7 Conclusion ................................................................................................................... 9-26
8 Acknowledgements ..................................................................................................... 9-26
9 References ................................................................................................................... 9-26

10 Improvement to Finite Volume Subdivision Technique: Leakage Current

Criterion ...............................................................................................................................10-1
1 Introduction .................................................................................................................10-1
 2 Definition of Priority Criteria .......................................................................................10-1
3 How Does the Leakage Current Criterion Work? ....................................................... 10-2
4 Examples ..................................................................................................................... 10-3
5 Conclusion ................................................................................................................... 10-8

11 Improvements for Modeling Arbitrary Networks of Metallic Plates in

HIFREQ ................................................................................................................................. 11-1
1 Introduction ................................................................................................................. 11-1
2 Improvements in Modeling Plate Geometries ............................................................. 11-1
3 Technical Details for Unstructured Meshing ............................................................... 11-7
4 Improvements in Visualization of Plate Computation Results .................................. 11-10
5 Numerical Results ...................................................................................................... 11-14
6 Conclusion and Future Developments ....................................................................... 11-23
7 References .................................................................................................................. 11-24

12 Improving the Users’ Experience with SES Software Documentation .......... 12-1
1 Introduction ................................................................................................................. 12-1
2 A Quality-Oriented Integration of New Means ............................................................ 12-1
3 An Acceleration of Information Delivery .................................................................... 12-3
4 A Sprightly Approach to Documentation .................................................................... 12-5
5 Conclusion ................................................................................................................... 12-6

Part II: User Contributions

13 Practical Advice for Electrical Soil Resistivity Testing via the Wenner Method
…………………………………………………………………………………………………………………………..13-1
1 Introduction ................................................................................................................. 13-1
2 Overview of Electrical Soil Resistivity Testing ............................................................. 13-1
3 Conclusion .................................................................................................................. 13-11
4 Acknowledgments ...................................................................................................... 13-11
5 References .................................................................................................................. 13-11

14 Transferred Voltages on Drilling Equipment and Disconnected Electrode

Leads in Energized Substations ...................................................................................... 14-1
1 Introduction ................................................................................................................. 14-1
2 CDEGS Analysis of Voltage Rise ................................................................................. 14-3
3 Conclusion ................................................................................................................... 14-7
4 Acknowledgement ....................................................................................................... 14-8
5 References ................................................................................................................... 14-8
15 Grounding System Design Challenges for Small Distributing Stations in
Urban Los Angeles ............................................................................................................. 15-1
1 Introduction ................................................................................................................. 15-1
2 Case Study ................................................................................................................... 15-2
3 Design Process ............................................................................................................ 15-2
4 Final Grounding System Results and Plots................................................................ 15-13
5 Project Status ............................................................................................................. 15-18
6 Conclusion .................................................................................................................. 15-18
7 References .................................................................................................................. 15-19

16 The Challenges of Designing Solar Park Earthing Systems in the UK ........... 16-1
1 Introduction ................................................................................................................. 16-1
2 Background ................................................................................................................. 16-2
3 Soil Characteristics ...................................................................................................... 16-2
4 Earthing System Design .............................................................................................. 16-5
5 Conclusion .................................................................................................................. 16-12
6 Postscript.................................................................................................................... 16-12
7 Acknowledgements .................................................................................................... 16-13
8 References .................................................................................................................. 16-13

17 Evaluation of Step Voltages on Horses at the Curragh Race Course ............. 17-1
1 Executive Summary ..................................................................................................... 17-1
2 Background ..................................................................................................................17-2
3 Inputs to Model ............................................................................................................ 17-5
4 Results ......................................................................................................................... 17-8
5 Analysis of IEC 60479-3 using MALZ ........................................................................ 17-12
6 Recommendations ..................................................................................................... 17-17
7 Conclusion .................................................................................................................. 17-18
8 References .................................................................................................................. 17-19

18 Currents Induced by Air-core Reactors in a SVC Substation ....................... 18-1

1 Background .................................................................................................................. 18-1
2 Simulation models ...................................................................................................... 18-4
3 Results and solutions ................................................................................................ 18-10
4 Lessons learned ..........................................................................................................18-14
5 Limitations .................................................................................................................18-16
6 Conclusions ................................................................................................................18-16
7 Acknowledgements .................................................................................................... 18-17
 8 References .................................................................................................................. 18-17

Part III: SES Contributions

19 Short Circuit Computations Using Lumped Impedances in MultiGroundZ

and MultiFields Software Packages ............................................................................... 19-1
1 Introduction ................................................................................................................. 19-1
2 Short Circuit Calculations ............................................................................................ 19-1
3 The Per-Unit System ................................................................................................... 19-2
4 Single Line Diagram and Impedance Diagram ........................................................... 19-4
5 Three-Phase Fault Current Calculations ..................................................................... 19-6
6 Single-phase to ground fault current computations ..................................................19-10
7 Inclusion of Mutual Impedance Values .....................................................................19-18
8 Conclusions ................................................................................................................ 19-21
9 References .................................................................................................................. 19-21

20 Short-Circuit Study in Networks with Multiple Transformers Outside the

Central Site in the SPLITS Computation Module ...................................................... 20-1
1 Introduction ................................................................................................................ 20-1
2 Power System Network and its Sequence Equivalents ............................................... 20-2
3 Modeling of the network in SPLITS ............................................................................ 20-3
4 Short-Circuit Analysis ................................................................................................. 20-5
5 Different scenarios of transformers and network in SPLITS.................................... 20-15
6 Conclusions ............................................................................................................... 20-17
7 Acknowledgment ....................................................................................................... 20-17
8 References ................................................................................................................. 20-17

21 The Electromagnetic Interference (EMI) Model for Electric Railways in

CDEGS ................................................................................................................................... 21-1
1 Introduction ................................................................................................................. 21-1
2 Mechanisms of Electric Railway Operation ................................................................ 21-2
3 Description of a Simple AT Electric Railway Interference Problem ........................... 21-6
4 Modeling the AT Catenary Traction Line in HIFREQ ................................................ 21-8
5 Modeling AT Catenary Traction with Moving Train in ROW/SPLITS ...................... 21-11
6 Conclusion .................................................................................................................. 21-21
7 Acknowledgment ........................................................................................................ 21-21
8 References .................................................................................................................. 21-21
22 Modeling and Mitigation Design Considerations for AC Interference
Studies in Low Soil Resistivity ....................................................................................... 22-1
1 Introduction ................................................................................................................ 22-1
2 General Considerations ............................................................................................... 22-1
3 Modeling Considerations in Right-of-Way Pro .......................................................... 22-4
4 A Worked Example ................................................................................................... 22-10
5 Mitigation Design Considerations ............................................................................ 22-12
6 Conclusion ................................................................................................................. 22-13
7 References ................................................................................................................. 22-13

23 Computation of Geomagnetically Induced Currents in Power Systems

using SES MultiFields Software Package .................................................................... 23-1
1 Introduction ................................................................................................................ 23-1
2 Geomagnetic Data ....................................................................................................... 23-3
3 Earth Resistivity Models ............................................................................................. 23-4
4 Geoelectric Field Computation ................................................................................... 23-5
5 GIC model ................................................................................................................... 23-7
6 Mitigation Measures ................................................................................................. 23-12
7 Conclusion ................................................................................................................. 23-13
8 References ................................................................................................................. 23-14

24 Verification and Computation of Conductor Electrical Properties in the

Context of Incomplete Datasheets ................................................................................ 24-1
1 Introduction ................................................................................................................ 24-1
2 Modeling of Galvanized Conductors ........................................................................... 24-2
3 Modeling Chinese Conductors .................................................................................... 24-9
4 Conclusions ............................................................................................................... 24-10
5 References .................................................................................................................. 24-11

25 Finding Equivalent Conductors Representing a Typical Metallic Wire-Mesh

Fence ..................................................................................................................................... 25-1
1 Introduction ................................................................................................................ 25-1
2 Methodology................................................................................................................ 25-2
3 Examples ................................................................................................................... 25-12
4 Conclusion ................................................................................................................. 25-15
 26 Immersed: Short-Circuit Safety Criteria for Swimmers ............................. 26-1
1 Introduction ................................................................................................................ 26-1
2 Proposed Criteria ........................................................................................................ 26-1
3 Example Application ................................................................................................... 26-3
4 Conclusions and Disclaimer ........................................................................................ 26-9

Part IV: Research Work in Progress

27 Increasing Computation Speed in MALT and MALZ Using the Discrete

Complex Image Method ....................................................................................................27-1
1 Introduction .................................................................................................................27-1
2 Computation of the Soil Potentials in Horizontal Multilayer Soil Structures .............27-1
3 Case Studies ................................................................................................................ 27-5
4 Conclusion ................................................................................................................... 27-9
5 References ................................................................................................................... 27-9

28 Derivative Constrained Polarization Curve Fitting ....................................... 28-1

1 Introduction ................................................................................................................ 28-1
2 Nonlinear Electrical Device Modeling using Polynomials .......................................... 28-4
3 Unconstrained Least-Squares Polarization Curve Fitting .......................................... 28-7
4 Derivative Constrained Least-Squares Polarization Curve Fitting ........................... 28-10
5 Numerical Results ..................................................................................................... 28-19
6 Conclusion ................................................................................................................. 28-21
7 References ................................................................................................................. 28-21

29 The Simulated Annealing Method Applied to the Soil Resistivity Analysis

Inverse Problem ................................................................................................................ 29-1
1 Introduction ................................................................................................................ 29-1
2 The Simulated Annealing Metaheuristic .................................................................... 29-1
3 Illustrative Example .................................................................................................... 29-6
4 Statistical Testing ...................................................................................................... 29-10
5 Conclusion ................................................................................................................. 29-22
6 References ................................................................................................................. 29-23

30 Magnetic Field of Pipe-Type Cables with Consideration of Actual Positions

of Sub-Cables ...................................................................................................................... 30-1
1 Introduction ................................................................................................................ 30-1
2 Determination of the Actual Position of Cable Components ...................................... 30-2
3 Activating this Feature ................................................................................................ 30-4
 4 Validation .................................................................................................................... 30-6
5 Practical Examples .................................................................................................... 30-10
6 Future Developments ................................................................................................ 30-12
7 References ................................................................................................................. 30-12

31 New Multi-Layer Hemi-spheroidal Soil Model .................................................... 31-1

1 Introduction ................................................................................................................. 31-1
2 Oblate Spheroidal Coordinate System ......................................................................... 31-1
3 Confocal Multilayer Oblate Hemi-Spheroidal Soil ..................................................... 31-2
4 Benchmarks................................................................................................................. 31-3
5 Conclusion and Future Work ..................................................................................... 31-12
6 References .................................................................................................................. 31-12

32 Comparison between Plate and Equivalent Wire Models in Grounding

Applications ........................................................................................................................ 32-1
1 Introduction ................................................................................................................ 32-1
2 Comparison between Plates and Equivalent Wire-Mesh Grids .................................. 32-1
3 Comparison of Coated Pipe Modeled by Coated Plates or Wire ................................. 32-5
4 Conclusion ................................................................................................................. 32-10
5 References ................................................................................................................. 32-10

Part V: Recently Published Technical Articles

Part I: New Features
 UGM 2017 – CANNON BEACH, OREGON

1 IMPROVEMENTS IN SES SOFTWARE

Simon Fortin

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
This article describes the improvements made to the SES software program since the 2016 CDEGS User’s Group
Meeting.

1 Introduction
This article briefly describes the main improvements that were made to SES Software since the
2016 CDEGS User’s Group Conference.

2 New User Interface Programs

The development of the new user-interface programs for SES Software, undertaken a few years
ago, has reached a point where some of the new programs are now used by default in CDEGS. This
version includes the first official release of the new CDEGS interface and of the new RESAP,
FCDIST, TRALIN, and FFTSES user-interface modules. It also includes the first version of the
SESResultsViewer program, the successor to the CDEGS – Examine (Output Toolbox) and
GrServer programs.
While the older user-interface programs for those modules are still available (using the CDEGS-
Legacy icon) they are now entering a maintenance phase: no new features will be added to those
legacy programs from this point on.

2.1 SESResultsViewer: New Configuration and Computation Plots and

Report Processor
The SESResultsViewer program is a new configuration and computation plot and report processor
that acts as a replacement for the CDEGS – Examine (Output Toolbox) and GrServer programs.
It offers an open, fluid interface where the most important settings are quickly accessible directly
from the main screen.
This tool is currently available for the MALT, MALZ, and HIFREQ modules and will be extended
to support all other computation modules in the future. In addition to supporting essentially all
features available in Output Toolbox for MALT, MALZ, and HIFREQ, it also integrates the
SESThreshold tool, presented in recent User’s Group Meetings, for the specification of safety
parameters and safety zones.
Consult the article “SESResultsViewer: New Configuration and Computation Plots and Report
Processor” for more details.

Copyright © 2017 SES ltd. All rights reserved. Page 1-1

PART I: NEW FEATURES

Figure 1: SESResultsViewer interface.

2.2 FCDIST
A new user interface for FCDIST is now available. It is based on the more visual interface that
AutoGroundDesign users have become used to and replicates all the inputs traditionally specified
through the CDEGS Input Toolbox, while offering more intuitive screens for different terminal
types (various configurations of transmission lines or of cables).

Page 1-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: FCDIST interface.

The article “A New Interface for FCDIST” gives more detail.

2.3 CDEGS
The new version of the main interface for the CDEGS software package, which was presented for
the first time at last year’s User’s Group Meeting, has been improved for this first official release
of the program. Status Indicators have been added below the icon for each module to indicate at
a glance if the input file or the computation database file is available for the module. The Specify
mode for MALT, MALZ, and HIFREQ is now available and gives quick access to SESCAD to edit
the input data for these modules. Also, you can now run the computation engine for each module
directly from this interface and launch the appropriate tool to visualize the computation results.
The article “Improvements and New Features in CDEGS” gives more details.

Copyright © 2017 SES ltd. All rights reserved. Page 1-3

PART I: NEW FEATURES

Figure 3: Main CDEGS screen.

2.4 RESAP
The RESAP interface was released in Beta form during the last UGM to act as a complete
replacement for the traditional CDEGS – Specify session for RESAP. Since then, it has been
steadily improving so that the Beta flag has now been removed. This year’s improvements include:
 A new layout for the Computations panel, which was relocated under the
Measurements panel, to give each panel more space and make the application easier to
use on lower-resolution screens.
 Commands related to filters are now available and the High-Precision filter is now used
by default for new cases to minimize numerical errors in the calculation of the computed
apparent resistivity curves.
 Once the computations are complete, the program now automatically produces a plot
showing the computed apparent resistivity curve and the soil model and displays it in
GraRep. This speeds up considerably the process of refining the analysis. You can also
generate this plot on demand by clicking on the new Plot button.

Page 1-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 4: The Plot feature in RESAP.

The article “Improvements in RESAP” gives more details.

2.5 SESLibrary
The SESLibrary component was introduced for the first time last year, at the CDEGS User’s
Group Meeting. This library gives access to electrical and geometrical data for a large inventory
of conductors and cables. In this version of the software, this component is used by all SES
programs to import conductor data, and replaces the SESConductorDatabase of previous
versions in this function. This version of SESLibrary also adds new classes of Aluminum
Conductor Steel Supported (ACSS) conductors and data for typical fence posts.

Figure 5: The new SESLibrary component.

Consult the article “New Features in SESLibrary” for more details.

2.6 SESThreshold
SESThreshold is a software tool for computing thresholds for touch and step voltages, as
recommended by industry standards. This tool was initially introduced 2 years ago. Since then
several aspects of the program have been improved.
1- Support for new standards has been added. Now, the complete list of available standards is as
follows:

Copyright © 2017 SES ltd. All rights reserved. Page 1-5

PART I: NEW FEATURES

 IEEE
 IEC/TS 61936-1:2010-CENELEC EN 50522:2010 (New)
 BS EN 50522:2010 (New)
 IEC 60479-1:2005 (New)
 User-Defined (New)

Choosing any of the first four standards will restrict some of the options on the interface
according to the specifications in the standard. However, when User-Defined standard is
chosen, all the selectable options are available.
2- A report listing the computed threshold value for each zone as well as some intermediate
results is produced whenever the thresholds are computed (using the Compute Only or the
Compute & Plot buttons). The report file is named “TH_JobID_ThresholdReport.txt” and
is stored in the working directory for the case. It can be displayed by clicking on Show Report
in SESThreshold.
3- In addition to being usable as a standalone tool, SESThreshold is now used in the newly
developed SESResultsViewer program to specify parameters for the computation of safety
limits for touch and step voltages.

Figure 6: The BS EN 50522:2010 standard in SESThreshold.

3 CorrCAD
A Beta version of the new CorrCAD cathodic protection software package was introduced at the
CDEGS User’s Group Meeting two years ago and has been improving steadily since that time. This

Page 1-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

version of CorrCAD is the first official release of this package. It includes the following
enhancements compared to the version presented at last year’s CDEGS User’s Group Meeting.
 Soil model data is now specified using the Soil Model Editor component, common to
most SES Software programs. Also, the program can now recognize that an invalid soil
definition file is selected.
 Numerous improvements were made to the curve fitting tool (SESCurveFit) that is used
to fit polarization data to smooth curves.
o The parameters specified to control the curve fitting process can be saved and
reloaded in subsequent sessions. Also, the Initial Estimates of the parameters
for the curve fitting process can be defined. By default, the values of these
parameters are now taken from the Polarization Curve Properties table
instead of generic default values, which generally improves the quality of the fit.
o Data sampled graphically (on a background picture) in the Polarization Curve
window or directly in the Samples table are now synchronized.
o Axes have been added in the Numerical Fit window.
 It is now possible to assign different polarization curves to different conductors.
 The Leakage Current Density (accounting for coating efficiency) can now be plotted
along conductors, in addition to the Over-Potential. Also, all computation results can
now be displayed in 2D-Spot format (see Figure 7).
 Several improvements were also made with respect to the specification of the geometry
and characteristics of the system:
o The Polyline Operations & Break Polylines at Regular Locations tools
have been made far more efficient and no longer takes an extended amount of time.
o It is now possible to turn the display of the grid drawn in the background of the
model on or off.
o The different types of entities that can be inserted in a model are now displayed
using different representative icons.
 Also, the folders and files storing the input data and the computation results have been re-
organized to make it easier to find pertinent information about a model. See Figure 8 for
a typical example.

Copyright © 2017 SES ltd. All rights reserved. Page 1-7

PART I: NEW FEATURES

Figure 7: 2D-Spot plot of Leakage Current Density in CorrCAD.

Figure 8: Typical folder structure for input and output files in CorrCAD.

4 Right-of-Way and RowCAD

Several enhancements were made to our AC interference analysis package Right-of-Way and to
its graphical user interface RowCAD.
 In the Advanced plot module, a Keyhole Markup Language (KML) plot is available in this
version. It creates a Google Earth™ file that displays path coordinates combined with
computation results directly on the Google Earth™ world map.
 The Create Circuit process has been improved to automatically account for the coating
characteristics of coated buried conductors. Also, this module is now able to automatically
compute buried (bare or coated) conductor shunt (ground) impedance for a component

Page 1-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

inside of a Cable and Group path and can now automatically obtain the complete Touch-
to-GPR and Coating Stress-to-GPR percentage files when a partial circuit build is required.
 The maximum soil breakdown distance and various fault arcing and flashover distance
options can now account the X/R ratio of the system.
 For steady state conditions, the Reference conductor for Total Interference computations
now offers various filtering options such as phase number, conductor-type, coating-type
and profile number, etc., to precisely control the resulting envelopes.
Also, you can now run ROW concurrently with Output Toolbox (or other CDEGS computation
modules).

Figure 9: Pipeline GPR Intensity plot on Google Earth.

Consult the article “RowCAD and ROW Improvements” for more details.

5 HIFREQ: Junctions between Metallic Plates and

Improvements to Magnetic Field Due to Cables
In HIFREQ, the modeling of metallic plates was made more flexible by allowing contacts between
multiple plates at one common plate edge and also partial contacts of plates along an edge. Also,
computation results related to plates, such as the current flowing in plates or their potential, can
now be displayed in Configuration and Conductor Data plots in the new SESResultsViewer
program, in CDEGS – Examine (Output Toolbox) and in GrServer.

Copyright © 2017 SES ltd. All rights reserved. Page 1-9

PART I: NEW FEATURES

Figure 10: HIFREQ examples involving complex contacts between plates.

Figure 11: Visualization of current flowing in plates.

See “Improvements for Modeling Arbitrary Networks of Metallic Plates in HIFREQ” for more
details.
Also, the program can now optionally account for the actual position of coaxial cables located
inside pipe-type (multicore) cables in the calculation of the magnetic field produced by those
cables.

Page 1-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON
Magnetic Fields/Resultant (Total) Field [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

LEGEND

Maximum Value : 590.563

-102.0
Minimum Value : 116.374

590.56

543.14
-101.0
495.73

Z AXIS (METERS)
448.31

-100.0 400.89

353.47

306.05
-99.0 258.63

211.21

163.79
-98.0
-2 -1 0 1 2

Y AXIS (METERS)
Total Magnetic Field Magn. (Amps/M)

Figure 12: Magnetic field around three cable conductors without pipe-enclosure.
See “Magnetic Field of Pipe-Type Cables with Consideration of Actual Positions of Sub-Cables”
for more details.

6 New Leakage Current Criterion for Soil Models with Finite

Volumes in MALT and MALZ
In order to get accurate results while using soil volumes embedded into uniform or horizontally
layered soils, it is important to subdivide the soil volumes interfaces into small surface elements
or patches. The ideal patch distribution and size should reflect the intensity of the electric field
pattern on the soil volume faces. In order to achieve that, the Adaptive Patch Subdivision
technique has been introduced more than 10 years ago to optimize the patch subdivision. The
Adaptive Patch Subdivision is based on physical criteria (designated as Priority Criteria), such as
conductor proximity, segment length, resistivity contrast ratios, location of patches on the faces
of the volumes, etc. However, so far all the criteria related to conductors assumed that all
conductor segments are equally important for the patch subdivision process, regardless of the
amount of current leaking out of the segments, even though segments with large leakage currents
are expected to produce larger electric fields. A new criterion based on leakage current density
has been introduced in order to correct this flaw. It ensures that the soil volume faces located near
conductors with large leakage current densities receive an adequate portion from the total pool of
patch resources and, conversely, that the soil volume faces located near conductors with small
current densities do not “attract” an excessive patch subdivision.
It is expected that the introduction of this new criterion, which is described in detail in the article
“Improvement to Finite Volume Subdivision Technique: Leakage Current Criterion”, will
significantly improve the effectiveness of the Adaptive Patch Subdivision process.

Copyright © 2017 SES ltd. All rights reserved. Page 1-11

PART I: NEW FEATURES

(a) (b)

Figure 13: Patch distribution on low resistivity (left) and high resistivity (right) soil volumes; (a): With
Leakage Current criterion; (b): Without Leakage Current criterion.

7 MALZ: Multilayer Hemi-Spheroidal Soil Model

The hemi-spheroidal soil model that was introduced two years ago in the MALT and MALZ
modules was extended to allow for the specification of multiple confocal hemi-spheroidal layers
with different resistivities. This can be used as an approximation for any soil with a layered
structure occupying a finite region, and is ideal for modeling lakes.
This is currently available only in command mode, for the MALZ module.

Figure 14: Multi-layer hemi-spheroidal soil.

See “New Multi-Layer Hemi-Spheroidal Soil Model” for more details.

8 SPLITS: Zero Sequence Impedances for Transformers

The three-phase transformer model that has been available in the SPLITS module for several
years was extended to allow for the specification of different values for the positive and zero
sequence impedances of the transformer. Taking this information into account can lead to more
accurate calculations of fault currents in the presence of transformers. As is the case for the
existing positive sequence data for the Transformer Characteristics, the zero sequence data

Page 1-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

can be specified either from manufacturer test data or directly as transformer impedance matrix
elements.

Figure 15: SPLITS transformer screen.

See “New Features and Improvements for Transformers in SPLITS” for more details.

9 RESAP: Simulated Annealing Methodology

A new optimization method, the Simulated Annealing methodology, was added in RESAP to
deduce the soil model that best represents specified soil resistivity measurements. This method,
which is available in command mode only) is useful when it proves difficult to find good starting
values for the initial estimates of solution, or when the other available methodologies (namely,
Steepest-Descent and Levenberg-Marquardt) fail to converge to a satisfactory RMS error. This
new method also integrates the possibility of constraining the range of values for each soil model
parameter.
See “The Simulated Annealing Method Applied to the Soil Resistivity Analysis Inverse Problem”
for more details.

Copyright © 2017 SES ltd. All rights reserved. Page 1-13

PART I: NEW FEATURES

Figure 16: RMS error for three-layer cases as obtained with the three different methodologies.

10 SESCAD: Composite Soil Volumes and KML Files

The following enhancements were introduced in the SESCAD program since the last User’s Group
Meeting.
 The Advanced Options screen that was introduced last year to give access to advanced
settings for the MALT, MALZ, and HIFREQ modules is now complete. As a result, it is no
longer necessary to use Input Toolbox to specify data for those modules.
 By default, SESCAD now uses the new Soil Model Editor to edit the characteristics of
the soil model and the new SESLibrary to retrieve the properties of conductors. The
program also gives access to the new SESResultsViewer tool to view the computation
results.
 A new tool to define soil volumes of cylindrical and other shapes was introduced for the
MALT and MALZ modules.
 Data stored in Google Earth™ KML files can now be imported in the program.

Page 1-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

(a) (b)

(c) (d)

Figure 17: Examples of composite soil volumes created with the Create Composite Soil
Volume tool.
See “Improvements in SESCAD” for more details.

11 Other Improvements
Many other improvements were made, particularly in the results visualization modules.

11.1 CDEGS-Examine (Output Toolbox)

 For Computation plots in MALT, MALZ and HIFREQ, a Zoom area and an explicit
selection of profile and observation point numbers can be specified at the same time.

Copyright © 2017 SES ltd. All rights reserved. Page 1-15

PART I: NEW FEATURES

Figure 18: Simultaneous specification of profile and point ranges and of a zoom area.
 Plots of Coating GPR, Coating Stress Voltage, and Coating Type were added for
MALT, both for Configuration and Computation (Conductor Data) plots. Also, a
new filter on Coating Type was added for Configuration plots in MALT.
 A new way to draw the arrow indicating the direction of current flow in Configuration
plots was introduced in MALZ and HIFREQ. The new method draws this arrow from the
origin to the end of each segment, making it easier to determine the correct direction of
current flow in some scenarios. A new Reference Direction option is now available to
control the direction of the arrow.

Page 1-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 19: The Reference Direction option in CDEGS – Examine (Output Toolbox) for
HIFREQ.

(a) (b)
Figure 20: Magnitude and angle of longitudinal current at the origin of conductor; (a) Drawn
using the method used in previous versions of the program (Positive Value) where the
direction of the arrow is based on the sign of the real part of the current; (b) Drawn using the
new Origin to End option.
 The standard FCDIST report now displays the neutral current at the Central Site along
each terminal and the Current Split Factor at the Central Site. The standard SPLITS report
also displays this last quantity. See Figure 21.
 The plotted data is now saved automatically to a file in the CSV format when creating plots
in CDEGS – Examine (Output Toolbox) and GrServer for SPLITS. The file is named
“SP_JobID.CSV” and is saved in the working directory for the case. Note that this file is
overwritten every time a plot is created.

Copyright © 2017 SES ltd. All rights reserved. Page 1-17

PART I: NEW FEATURES

Figure 21: Additions to the standard report for FCDIST.

Figure 22: Excerpt from CSV report created when producing a plot in SPLITS.

11.2 GrServer: Plots of Electric Field for MALT and MALZ

The capability to compute the electric field was introduced a few years ago in the MALT and MALZ
modules. In this version of GrServer, the computed electric field and current density can be

Page 1-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

plotted, either as magnitude (resultant) or as Instantaneous or Semi-Major Axis values. This

makes it possible to generate high-quality 3D plots of those quantities.

Figure 23: Electric field computed with MALZ, plotted in GrServer.

11.3 Fibrillation Current for IEC Standard

In SESThreshold (the new safety analysis tool integrated in the SESResultsViewer program) and
in the Safety module of CDEGS - Examine, GrServer, AutoGrid Pro, and AutoGroundDesign, a
new method was introduced to evaluate the fibrillation current from the curves provided in the
IEC standards. This new method leads to better agreement with the BS EN 50522:2010 and
CENELEC EN 50522:2010 standards newly implemented in SESThreshold.

11.4 SESPlotViewer for SESEnviroPlus

The SESPlotViewer plotting tool that was briefly introduced at the User’s Group Meeting last year
is now used by default to create plots in SESEnviroPlus. In SESPlotViewer, the computation
results are displayed in a table and drawn as a 2D line plot. The fonts for the various screen
elements can be customized and the plot can be copied to the Windows clipboard for further
processing.
Plot files compatible with SESPlotViewer are also produced automatically when running the
SESTLC tool, and provide an alternative to the built-in plot viewer of SESTLC.

Copyright © 2017 SES ltd. All rights reserved. Page 1-19

PART I: NEW FEATURES

Figure 24: The new SESPlotViewer program, displaying SESEnviroPlus results.

11.5 GRSPLITS-3D: Support for Display of Multiple Neutrals in FCDIST

The capability to specify multiple sets of neutral wires with different characteristics was
introduced two years ago in FCDIST. However, it wasn’t possible to inspect the characteristics of
the individual neutral wires when looking at the FCDIST model in GRSPLITS-3D. As shown in
Figure 25, the new version of GRSPLITS-3D now displays those characteristics when pointing to
the neutral wire along a terminal of the circuit model.

Page 1-20 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 25: Characteristics of multiple neutral bundles for FCDIST,

as displayed in GRSPLITS-3D.

11.6 Software Protection: Performance Improvements

The speed of the software protection system for SES Software was greatly improved, especially for
network-based software protection dongles. This considerably speeds up the operations that
require a verification of the license, such as saving files and starting computations.

Copyright © 2017 SES ltd. All rights reserved. Page 1-21

 UGM 2017 – CANNON BEACH, OREGON

2 IMPROVEMENTS AND NEW FEATURES IN CDEGS

CDEGS Team
Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

The CDEGS software package is currently being redesigned in accordance with the new SES User Interface (UI)
standards. This is the second phase of many in which the original CDEGS features will be migrated to a brand new
UI. This article describes in detail the new features implemented and the major improvements made to the
CDEGS software package since the last User’s Group conference.

1 Introduction
The CDEGS software package is a powerful set of integrated software modules designed to
accurately analyze a wide variety of electrical and electromagnetic phenomena. This application
is now getting redesigned to a new user interface and will now comply with the rest of the newly
updated software packages.
The update will take place over multiple phases. As the first phase proposed last year [1] was
about providing a new product while maintaining the capabilities of the previous program, this
year’s efforts were focused on continuing this work while improving the user experience.
This version of the software includes all the features that were available in the original CDEGS
software and also adds some enhancements. It is now the default user interface for the
MultiGround, MultiGroundZ, MultiFields, and CDEGS software packages. Note that the original
CDEGS program is still available but has been renamed CDEGS-Legacy.
The following section outlines the main improvements and modifications implemented in the
program. We first present the status indicators that have been added below each module button,
indicating the availability of its session modes (Specify, Compute, and Examine) without
having to click on the button itself. Then, modifications and additions made to the programs for
each session mode are discussed.

2 New Features

2.1 Status Indicators of the Computation Module

The main CDEGS interface is presented in Figure 1. As can be noticed, the overall aspect of the
application proposed last year is mainly preserved, but two status indicators are now visible
below each module button. As the session mode is now independent from one module to the
other, those status indicators show the available session modes of the corresponding module in
the following manner:
 If the model has not been created for the selected module, the two status indicators
under the module icon are red, as shown in Figure 1. You can click Specify to build a
new model.
Copyright © 2017 SES ltd. All rights reserved. Page 2-1
PART I: NEW FEATURES

 If the model of the selected module is created but has not been run yet (i.e., a valid F05
file exists, but no output files), the color of the left indicator changes to green (the right
one remains red). You can click Specify to modify the model or Compute to launch the
computations, as shown in Figure 2. Note that for the modules not included in your
license, this Compute option is locked.
 Once the model for the selected module is computed (i.e., a valid F21 file exists), the
second status indicator turns to green. You can click Specify to modify the model,
Compute to relaunch the computations or Examine to view/plot the results. For the
example shown in the Figure 3, the MALZ model is created (F05 file), it has been run
(F21 file), and therefore, the results can be viewed. Note that if for any reason, only the
F21 file is present in the working directory, only the second indicator is turned to green.

Figure 1: Main CDEGS screen, showing that there is no MALZ input or output file for the selected Job ID,
but for the other modules, only an input file is defined.

Figure 2: Same case as in Figure 1, but with an input MALZ file defined.

Page 2-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 3: Same case as in Figures 1 and 2, but the MALZ model has been run, and the computation results
can be examined.

3 Improvements

3.1 Data Input (Specify)

All computation modules available in the original CDEGS have been introduced in this new
version. Most computation modules also have a new interface and can be used as standalone
applications. This is in fact the case for RESAP [2], TRALIN [3], FCDIST [4], and FFTSES [5]. In
future versions, it is also intended to do the same thing for SPLITS, but for now, by clicking this
last module you will see the same interface as that of previous versions.
Regarding MALT, MALZ, and HIFREQ, clicking those buttons no longer opens Input Toolbox.
These screens have actually been removed from CDEGS. Instead, by clicking those buttons in
Specify mode, you will see the screen shown in Figure 4 (this screen is shown for HIFREQ, but
the same screen is used for MALT and MALZ). This screen allows you to input basic information
about the model, such as the Run Identification, the System of Units, and a Case
Description. However, the most important element in this screen is the SESCAD button.
Clicking it directly opens SESCAD with the corresponding MALT/MALZ/HIFREQ input file. In
other words, the input file is no longer defined in Input Toolbox as its features are now all
integrated in SESCAD [6].

3.2 Launching the Computations (Compute)

For each module, when Compute is clicked, the Computations screen shown in Figure 5
provides the following options:
 Resume: Allows to redo a simulation or continue one that was paused;
 Pause: Suspends the simulation (freeing computer resources for other tasks), allowing it
be continued later using Resume;
 Stop: Stops the simulation definitely (clicking Resume will restart from scratch).
Also note the progress bar indicating that the program is working, i.e., that the simulation is in
progress.

Copyright © 2017 SES ltd. All rights reserved. Page 2-3

PART I: NEW FEATURES

Figure 4: Input screen that appears when clicking the HIFREQ (or MALT or MALZ) button in Specify
mode.

Figure 5: Computation Trace screen shown when starting a computation.

Page 2-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3.3 Viewing the Results (Examine)

Once the computations are completed for a given module, the following is done by default when
clicking Examine:
 RESAP, FFTSES, and TRALIN: the corresponding program starts and opens into its own
results screen as those modules are standalone applications;
 FCDIST: opens the program in input mode; the results can be viewed by clicking View
Plots and Reports, which allows the user to view the results using Output Toolbox or
GRServer;
 MALT, MALZ, and HIFREQ: the new SESResultsViewer program is launched [7];
 SPLITS: the GRSPLITS-3D tool is launched [8].

4 Conclusion
This article described the new CDEGS user interface. This interface, which is now used by
default in the MultiGround, MultiGroundZ, MultiFields, and CDEGS software packages,
supports all the features of older versions of the CDEGS program and adds a few new features.
In particular, the status of each module (i.e., the operations that can currently be performed for
a given module), can now be viewed easily without clicking the corresponding button. The way
to handle the different session modes (Specify, Compute, Examine) with the new CDEGS
application was also presented.

5 References
[1] É. Larivière and F. Gougeon, “A New Interface for CDEGS,” in CDEGS Users’ Conference Proceedings,
Boulder, Colorado, USA, 2016.

[2] S. Baron, S. Franiatte, M. Golshayan, Y. Jiang, Z. Luo, M. Siahrang, and L. Valcárcel “SESRESAP: A
New Soil Resistivity Measurement Editor,” in CDEGS Users’ Conference Proceedings, Boulder,
Colorado, USA, 2016.

[3] M. Chami, M. Daigle, E. Dawalibi, S. Franiatte, G. Noel, M. Siahrang, L. Valcárcel, and C. Voyer,
"Improvements in SESTralin," in CDEGS Users' Conference Proceedings, San Diego, California, USA,
2015.

[4] FCDIST Team, “New Interface for FCDIST,” in CDEGS Users’ Conference Proceedings, Cannon Beach,
Oregon, USA, 2017.

[5] FFTSES Team, “A New Interface for FFTSES,” in CDEGS Users’ Conference Proceedings, Boulder,
Colorado, USA, 2016.

[6] S. Fortin, “Improvements in SESCAD,” in CDEGS Users’ Conference Proceedings, Boulder, Colorado,
USA, 2016.

[7] SESResultsViewer Team, “SESResultsViewer: New Configuration and Computation Plots and Reports
Processor ,” in CDEGS Users’ Conference Proceedings, Cannon Beach, Oregon, USA, 2017.

Copyright © 2017 SES ltd. All rights reserved. Page 2-5

PART I: NEW FEATURES

[8] M. Chami, E. Dawalibi, M. Deslongchamps, S. Franiatte, G. Noel, L. Valcárcel, and C. Voyer,

"Improvements in GRSPLITS-3D, ROWCAD and SESCrossSection," in CDEGS Users' Conference
Proceedings, Newport, Rhodes Island, USA, 2014.

Page 2-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3 IMPROVEMENTS IN RESAP

Maryam Golshayan, Yanmei Jiang,

Zhiqiong Luo, Alexandre Mailhot and Luis Valcárcel

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract

This article describes the improvements made to the graphical user interface of the RESAP program, since the 2016
CDEGS User’s Group Meeting.

1 Introduction
The RESAP interface was released in Beta form during the last UGM to act as a complete
replacement for the traditional CDEGS – Specify session for RESAP [1]. Since then, it has been
steadily improving such that the Beta flag can now be removed. It could already load, save and
compute all existing RESAP files, but this year’s improvements, the most notable of which are
listed below, make it considerably friendlier to use.

2 Datagrids
The measurements and initial soil model estimate data grids have been improved to make them
more robust regarding the copy-pasting of data from external sources and the handling of
undoable actions.

3 Computation Panel
The Computations panel was relocated under the Measurements panel to give each panel
more space and make the application easier to use on lower-resolution screens or in non-
maximized state. It is still possible to arrange the panel layout arbitrarily to suit every user’s
preference.
This panel also now benefits from a better placed control to specify the number of soil layers, that
allows typing an entry directly or selecting it from a list, and from updated schematics, as shown
in Figure 1.

Copyright © 2017 SES ltd. All rights reserved. Page 3-1

PART I: NEW FEATURES

Figure 1: Computations panel.

4 Computation Options
The Computed resistivity traverse parameters have now been made available through the
Computation options section of this panel, as shown in Figure 2.

Figure 2: Computed resistivity traverse parameters.

The filter commands for the computations have also been made available and the High-
Precision filter is now an application default. Its use minimizes numerical errors in the
calculation of the computed apparent resistivity curves and causes negligible performance
drawbacks on modern computers. Previous files should not be affected, but new cases will take
advantage of it and could converge to a different solution than they would have if the Automatic
filter had continued being used. This is simply a consequence of the fact that the solution to the
RESAP optimization problem may not be unique as there can be various combinations of the soil
structure parameters that yield similar apparent resistivity curves.

Page 3-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The Filter is set through the Advanced Options window, as shown in Figure 3.

Figure 3: Filters.

5 Template
RESAP now prepares new cases based on the Template file prepared from the
CDEGS | Options | Preferences screen. This means your preferred system of units, once set
in this screen, will be taken into account by the new application too.
Be aware that the High-Precision filter command mentioned above is now also included in the
RESAP template.

6 Examine Session
The Examine session for plotting output results (CDEGS’ Output Toolbox) can still be called
from the relevant button in the application. However, since loading this session can be rather
slow and can also conflict with running the computation anew, an auto-plotting macro was
implemented, that uses the most frequently selected plot settings, and is launched immediately
as the computations reach their end. This speeds up considerably the process of refining the
analysis and prevents any conflict with in-use results files. The macro can also be launched
manually by clicking the Plot button, as illustrated in Figure 4.

Figure 4: Auto-plotting macro.

Copyright © 2017 SES ltd. All rights reserved. Page 3-3

PART I: NEW FEATURES

7 Shortcut Keys
Shortcut keys were implemented to perform some frequently used actions directly from the
keyboard, as shown in Table 1:
Table 1: Shortcut Keys

Shortcut Key Action

F5 Launches the computations.

Shift+F5 Opens the F05 input file in a text editor.

Ctrl+F5 Reloads the currently opened input file.

Shift+F9 Opens the F09 output file in a text editor.

Shift+F12 Opens the F21 output file in an Examine session.

8 Languages
Lastly, for our international users, know that the user interface has now been translated in
Chinese and French, as shown in Figure 5. Spanish will follow shortly.

Figure 5: Available languages.

9 Example
An example file can be found under the UGC 2017\Example Files\Improvements in RESAP
subfolder of the SES Software conference distribution. It specifies a Wenner-type measurement
and includes comments for some of the data points, as well as the specification of initial values
for the soil layer parameters.

10 Conclusion
The new RESAP graphical user interface is now mature enough to use it as the primary point of
entry for creating new analysis cases. The new features described in this paper make it
considerably friendlier to use than previous versions. In particular, the improved layout make the
data more accessible, the High-Precision filter will help with cases involving high contrast ratios

Page 3-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

between the resistivity of adjacent layers, and the automatic plotting feature reduces the time
needed between iterations for refining the analysis.
As for the RESAP computation module itself, it now supports a new analysis method capable of
exploring more thoroughly the parameter space in searching for the global solution to the
optimization problem. This is described in [2].

11 References

[1] S. Baron, S. Franiatte, M. Golshayan, Y. Jiang, Z. Luo, M. Siahrang and L. Valcárcel, "SESResap: A New
Soil Resistivity Measurement Editor," in CDEGS Users' Conference Proceedings, Boulder, Colorado,
USA, 2016.

[2] S. Franiatte, O. Ramos-Leaños, L. Valcárcel, A. Hajiaboli, S. Fortin and F. P. Dawalibi, "The Simulated
Annealing Method Applied to the Soil Resistivity Analysis Inverse Problem," in CDEGS Users'
Conference Proceedings, Cannon Beach, Oregon, USA, 2017.

Copyright © 2017 SES ltd. All rights reserved. Page 3-5

 UGM 2017 – CANNON BEACH, OREGON

4 A NEW INTERFACE FOR FCDIST

Farid P. Dawalibi, Yanmei Jiang, Simon Fortin, Luis Valcárcel, Yixin Yang and Jie Liu
Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

A new user interface for FCDIST is now available. It is based on the more visual interface that AutoGroundDesign
users have become used to and replicates all the inputs traditionally specified through the CDEGS Input Toolbox,
while offering more intuitive screens for different terminal types (various configurations of transmission lines or
of cables), access to the new SESLibrary of conductors and better error-spotting tools. This version is an
intermediate version that will be upgraded to the full Microsoft WPF version similar to what you have now with
RESAP or TRALIN.

1 Introduction
The calculation of the current split factor is an important step for many grounding studies as it
can help reduce significantly the requirements on the grounding system. This is often
accomplished using SES’s well-known FCDIST computation module. This program calculates
extremely quickly the distribution of current flowing through the various metallic return paths,
i.e. the shield or neutral wires or cable sheath/armour of the lines feeding the fault, and through
the earth, which is the basis of a grounding study for a substation. Importantly, the calculation
takes into account not only the longitudinal and shunt impedances of these metallic paths, but
also their mutual coupling to the fault current carrying conductors. As for other computation
modules of the SES software suite, FCDIST is getting a renewed interface. It is meant to become
the successor of the traditional CDEGS Input Toolbox interface for that module.
For this first implementation cycle, the focus was put on unifying the interactions that the user
experiences with this module through the different software packages of the CDEGS suite. The
AutoGroundDesign implementation was chosen as a good candidate, given its more explicit visual
representations of the different network components. The exercise was therefore to extract the
relevant screens from AutoGroundDesign and make them accessible through a standalone
application, while adapting them to handle all of the module’s available inputs, for taking
advantage of its full capabilities, i.e. lifting the simplifying restrictions imposed on the entry level
software that is AutoGroundDesign. Behind the scenes, the code was also adapted to make use of
the more modern access structures to the data. In other words, this is foundation work that will
ensure a smooth transition and compatibility for when the graphical user interface will be given
its final appearance using the same development framework as other recent SES WPF based
applications. The following sections provide a tour of the new interface, highlight the novelties
and give a glimpse of the objectives for the next development phase.

Copyright © 2017 SES ltd. All rights reserved. Page 4-1

PART I: NEW FEATURES

2 Application Tour

2.1 Main Screen

Figure 1 shows the application’s main screen. The top half allows setting the case’s Job ID and
folder location on disk, and general options such as case description, system of units and
frequency. It also allows specifying the average soil characteristics and defining the Central Site
under study. A new diagram illustrates more clearly than before the different components of the
system.
The second half of the main screen is dedicated to terminal management, i.e. creation, edition and
deletion of terminals. This is followed by a row of action buttons that allow launching the
computation, viewing the results and managing the session. Finally, a status bar indicates whether
there are input data issues that should be taken care of for the case to be valid.

Figure 1: New FCDIST’s main screen.

Page 4-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.2 Terminal Screens

The application supports an arbitrary number of terminals feeding current contributions to the
fault, which is assumed to be located at the Central Site. Each time a new terminal is added to the
system, a choice of terminal types is offered to the user, as shown in Figure 2.

Figure 2: Terminal type selection screen.

Once a choice is made, the terminal’s configuration can be edited through a customized screen
illustrating the selected type. Common to all the types is the specification of the length of the
terminal, the average length between its grounding structure and their impedance value, the
location of the current carrying conductor, the amount of current contributed by its source, and
the impedance of the grounding system connected to it. What is particular to each type is the
specification of the return path conductor(s). Figure 3 shows the example of the distribution line
terminal type. In this case, the return path is a single neutral wire.

Copyright © 2017 SES ltd. All rights reserved. Page 4-3

PART I: NEW FEATURES

Figure 3: Terminal specification for a distribution line.

The similar case of the transmission line with a single shield wire is shown in Figure 4.
Furthermore, Figure 5 shows that transmission lines having two shield wires of identical
construction are likewise supported.

Figure 4: Terminal specification for a transmission line with a single shield wire.

Page 4-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Terminal specification for a transmission line with two identical shield wires.

For more exotic configurations, a bundle of return conductors is possible, as shown in Figure 6,
even if this is not common for realistic systems. However, a use for this can be for approximating
hollow cylindrical conductors, like a cable sheath or armour, by an arbitrary number of “strands”
arranged in a circular fashion around a given central point.

Figure 6: Terminal specification for a transmission line with multiple shield wires forming a bundle.
Copyright © 2017 SES ltd. All rights reserved. Page 4-5
PART I: NEW FEATURES

For cables, a dedicated interface was actually designed as shown in Figure 7, which ensures that
the phase conductor, i.e. the cable’s core, shares the same center location as its sheath/armour.
Note that, since FCDIST does not support conductors located below ground, buried cables should
instead be modeled above and close to the surface, which is a reasonable approximation [1]. For
this terminal type, the line structure impedance represents the impedance of the grounding
systems associated to cable vaults, to which the sheath and armour normally connect, instead of
overhead line towers. Since cross bonding cannot be accurately modeled with FCDIST, its use is
however restricted to simple cable configurations; the SES software suite contains other
computation modules that can address such situations in extensive details.

Figure 7: Terminal specification for a power cable.

Extending the concept towards full generality, Figure 8 shows the interface for specifying multiple
bundles of return conductors. Each bundle can have a distinct material for its “strands”, making
possible the modeling of a transmission line having two different shield wires, for example, a
feature first introduced in [2].

Page 4-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 8: Terminal specification for a transmission line with multiple shield wire bundles.

Finally, if the impedance of the return path is already known and does not need to be computed
by the program, then an Impedance Method is available where the series impedance of the
return path can be directly specified, together with its mutual coupling to the phase conductor, as
shown in Figure 9.

Figure 9: Terminal specification based on supplied impedances.

Copyright © 2017 SES ltd. All rights reserved. Page 4-7
PART I: NEW FEATURES

For all the terminal types, two switches are available to control the connection of the neutral or
shield wire(s), or cable sheath/armour to the grounding system of the terminal substation. These
switches are actually clickable controls that will open or close the connection. When closed, it is
possible to specify the impedance of the neutral connection and of the terminal grounding system.
When a switch is opened, the corresponding input fields disappear from the interface, as
illustrated by Figure 10 and Figure 11.

Figure 10: Closed switches allow the specification of the neutral connection and terminal grounding
system impedances.

Figure 11: An opened switch removes the corresponding input field from the interface.

Page 4-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3 Access to SESLibrary
For a correct evaluation of the fault current distribution, the material and dimension of the return
conductors need to be specified carefully. In order to help with this, each of the terminal types
includes an Import from DB button. It will call the new SESLibrary component [3], which is an
improved database for many types of conductors, as shown in Figure 12.

Figure 12: Access to the new SESLibrary.

4 Issues List
The application integrates an Issues List, which has now become a staple of newer SES interfaces
(Figure 13). It guides the user by displaying errors that need to be addressed before the
computations can complete successfully. For example, it will flag data input errors such as
negative values for resistances or resistivities, outer radius smaller than the corresponding inner
radius, etc.

Figure 13: Issues List shown below the application’s status bar.

Copyright © 2017 SES ltd. All rights reserved. Page 4-9

PART I: NEW FEATURES

The Issues List will appear as soon as an input field from the main interface is in error, or on
coming back from an Editing Terminal screen, which is in error, or by double clicking anywhere
in the application’s status bar.
If the Issues List is displayed prior to entering the Editing Terminal, it will remain visible in
the background and will respond dynamically to new errors or error fixes. Once inside that screen,
there is also a red exclamation icon next to the fields that are in error and a pink text zone at the
bottom of the window showing the description of the last error made, in case the complete Issues
List remains hidden behind that window, as shown in Figure 14.

Figure 14: Issues List as seen from the Editing Terminal screen,
together with other error-spotting tools.

5 Output Session
The goal is to eventually have the new SESResultsViewer [4] perform the plotting and reporting
of the FCDIST results. Since this is not available yet, the View Plots and Reports button
currently offers the two traditional possible output sessions: Output Toolbox or GRServer (Figure
15).

Figure 15: Choice of Output Toolbox or GRServer.

Page 4-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

As a novelty, the output report now contains two new quantities that were often requested: the
amount of current flowing back through the neutral system via each terminal, and the overall
current split factor. This is shown in Figure 16 and Figure 17.

Figure 16: The Neutral Current at the Central Site returning through each terminal is now available from
the report.

Figure 17: The Current Split Factor is now available from the report.

6 Example Files
New example files have been created to illustrate realistic cases of each type of terminal. They can
be found in the UGC 2017\Example Files\NewFCDIST subfolder of the SES Software conference
distribution folder of the software distribution.

7 Conclusion
The new FCDIST interface released this year succeeds in making the data input more visual, which
should reduce the risk of mistakes or misinterpretations. Since all the program’s input have been
duplicated, the new interface is ready to effectively replace the previous Input Toolbox interface
and is the one being called from the revamped CDEGS interface [5].
The main objective for the next minor release will be to finalize the look of the graphical user
interface by using the same Microsoft WPF development framework as other recent SES
applications. As the program already integrates together the input and computational aspects, we
will then also bring the output session within the same interface, reusing the plotting and
reporting capabilities of SESResultsViewer. Finally we are planning to integrate this computation
module within all of SES applications/components that may need a simplified fault current
distribution analysis.

Copyright © 2017 SES ltd. All rights reserved. Page 4-11

PART I: NEW FEATURES

8 References

[1] W. Ruan, "Modeling tap and underground cable in FCDIST," in CDEGS Users' Conference,
Proceedings, New Orleans, LA, USA, 2005.

[2] S. Fortin, "Improvements in SES Software," in CDEGS Users' Conference Proceedings, San Diego,
California, USA, 2015.

[3] SESLibrary Team, "New Features in SESLibrary," in CDEGS Users' Conference Proceedings, Cannon
Beach, Oregon, USA, 2017.

[4] N. Mitskevitch and F. P. Dawalibi, "SESResultsViewer: New Configuration and Computation Plots and
Reports Processor," in CDEGS Users' Conference Proceedings, Cannon Beach, Oregon, USA, 2017.

[5] CDEGS Team, "Improvements and New Features in CDEGS," in CDEGS Users' Conference
Proceedings, Cannon Beach, Oregon, USA, 2017.

Page 4-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

5 SESRESULTSVIEWER: NEW CONFIGURATION AND

COMPUTATION PLOTS AND REPORTS PROCESSOR
Nina Mitskevitch and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
A new configuration and computation plot and report processor called SESResultsViewer has been developed.
SESResultsViewer will replace our venerable legacy Output Toolbox interface. It is anticipated that it will provide
a more intuitive and user-friendly environment with additional capabilities that will expand very rapidly thanks to
its flexible architecture.

1 Introduction
Armed with years of experience using our venerable Output Toolbox since its infancy and
supporting users all over the world struggling with its present convoluted interface, we knew that
we needed more than a facelift to meet the growing requirements of displaying and reporting
results produced by our computation modules. Indeed we concluded that a complete
restructuring and redesign of this indispensable tool was necessary.
Our new SESResultsViewer software utility coupled with our already introduced SESThreshold
module represents the foundation on which we started planning our next generation of
computation plot and report processor. Analyzing, plotting and reporting your computation
results will be the responsibility of SESResultsViewer.
The version that we will be describing in this article supports presently the MALT, MALZ and
HIFREQ computation modules only. The other computation modules will be supported
progressively in the near future.

2 All You Need to Know about SESResultsViewer

The current version of SESResultsViewer replicates, to a large extent, all the features and
capabilities of the legacy Output Toolbox while introducing some new ones. Access to the various
features, options and settings is somewhat different. The sections below describe how to navigate
in SESResultsViewer to access its most salient features.

2.1 How to Start SESResultsViewer

SESResultsViewer can be launched from the new main CDEGS interface as shown in Figure 1
simply by clicking the Examine icon. It is similar to the legacy CDEGS interface except that the
“Session Mode” used in the old interface in now implicit rather than explicit.

Copyright © 2017 SES ltd. All rights reserved. Page 5-1

PART I: NEW FEATURES

Figure 1: Accessing SESResultsViewer using the new main CDEGS interface.

SESResultsViewer can be started as a standalone program directly from SES Software XXX |
Tools. In this case a computation database (file with extension “F21”) should be specified in the
Project tab as shown in Figure 2.
SESResultsViewer can also be accessed directly via the SES Batch and SESCAD tools.

Figure 2: Accessing SESResultsViewer in a standalone mode.

2.2 Main SESResultsViewer Screen

A typical SESResultsViewer screen is shown in Figure 3. The following describes the components
of this screen:
1. Data Selection Panel. This is the most important component of the screen because it
represents the control center. The Data Selection panel displays various items that can
be selected for plotting and reporting. The type of plots and reports can be specified in this

Page 5-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

panel as well. Plotting thresholds, rendering options and filtering are also available to
further discriminate the desired computation results to be displayed.
2. Ribbon. Currently the ribbon functionalities are basic but will be expanded progressively.
When all planned functionalities are implemented, significant customizations and
personalization will be possible.
3. Configuration Plot Panel. This panel is used to select a group of options that control
the display of the system configuration along with color coding and numerical results
textual labelling.
4. Computation Plot Panel. This panel is used to select various options that control the
computation quantity to be displayed, e.g., touch and step voltages, etc.
5. Soil Volumes Panel. This panel is used to select various options that control the display
of the patches’ currents or charges distribution on soil volume faces. The panel is presently
enabled when the MALT and MALZ computation modules are used and when soil volumes
are specified in the soil model. The HIFREQ computation module will have this capability
soon once the major SES’ R&D effort is completed.
6. Rendering Button. It allows to access various display options that are available for the
Computation, Configuration and Finite Soil Volumes plots and reports.
7. Action Buttons. Buttons that are used to generate the plots and reports.
8. Plotting Area. Area where the plots and reports are displayed. Presently, this area
displays results produced by utilities that were captured by the GraRep and GRServer
(Regular Enhanced mode) tools. Soon, plots generated by SESSystemViewer and
GRServer (Advanced mode) will be implemented as well. The already created plots and
reports are accessed by clicking the appropriate thumbnails in the panel below the plotting
area.

Figure 3: Typical SESResultsViewer interface.

Copyright © 2017 SES ltd. All rights reserved. Page 5-3

PART I: NEW FEATURES

There is an important difference between the legacy Output Toolbox and SESResultsViewer.
Every generated plot or report is not only displayed in the plotting area but also saved as a hard
copy in “.WMF” format in the specified working directory together with a command macro that
used to generate the plot. In the future SESResultsViewer will be able to load and display the
already generated plots from the hard drive along with the exact format and settings that were
used to generate such plot using the associated macro file.

2.3 Safety Thresholds

The SESThreshold module presented in the previous Users Group Meetings is now part of
SESResultsViewer. When SESResultsViewer opens a computation database (file with extension
“F21”) the safety thresholds for touch and step voltages are computed based on either default
settings (in a similar manner as it is done in the legacy Examine mode), or if the computation
database has already been associated with a SESThreshold file, based on the existing safety
settings of that file. To view or modify the safety settings one should access the SESThreshold
module as shown in Figure 4.

Figure 4: Accessing the SESThreshold module from SESResultsViewer.

2.4 Other Important Screens

The most important setting used to produce plots and reports in SESSystemViewer are shown in
the Data Selection screen. Additional settings can be accessed through the More button
available on each major panel. Various filters that can be useful to produce plots and reports are
combined together for easier access. Thus in order to access the Zoom polygon tool and various
conductor filters for the Configuration plot one should click the Conductor Filters button as
shown in Figure 5. The familiar Zoom polygon tool can be accessed from the Search Area tab as
shown in Figure 6.

Page 5-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Accessing conductor filters for Configuration plots and reports.

Figure 6: Accessing the Zoom polygon tool for Configuration plots and reports.

A similar Zoom polygon tool for Computation plots and reports can be accessed using the
Observation Points and Profiles Filters buttons as shown in Figure 7.

Copyright © 2017 SES ltd. All rights reserved. Page 5-5

PART I: NEW FEATURES

Figure 7: Accessing the Zoom polygon tool for the Computation plots and reports.

To access various display options that are available for the Computation, Configuration and Finite
Soil Volumes plots and reports, click the Rendering button. For example, the display options for
the Spot 2D and 2D plots are shown in Figure 8.

Figure 8: Various rendering options for the Computation plots.

3 Conclusion
The first release version of SESResultsViewer has been described briefly. A new interface for the
configuration and computation plot and report processor has been developed. SESResultsViewer
replaces our venerable legacy Output Toolbox interface. It is anticipated that it will provide a more
intuitive and user-friendly environment with additional capabilities that will expand very rapidly
thanks to its flexible architecture.
Page 5-6 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

As always, we eagerly wait for your valuable feedback in order to bring this new tool to its most
desirable state that will satisfy the most demanding customer.
The next step, which has been started already, would be a development of SES’ enhanced 3D
plotting engine that allows interactive and easily customizable plots, similar to the ones already
available in SESSystemViewer and GRServer.

4 Acknowledgments
This software development effort was carried out by a large group of SES employees that spent
hundreds of hours planning, coding and testing this newest member of SES’ software visualization
packages. We would like to thank Eric Dawalibi, Aditya Choubey, Said Touimer, Eric Lariviere,
Daniel Baygin, Marjan Mehrabi, Amir Hajiaboli, Sylvie Lefebvre, Luis Valcárcel and Jie Liu for
their relentless efforts in completing this important work.

Copyright © 2017 SES ltd. All rights reserved. Page 5-7

 UGM 2017 – CANNON BEACH, OREGON

6 IMPROVEMENTS IN SESCAD
Simon Fortin and Christian Voyer

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

1 Introduction
This article describes the most important improvements that were made to the SESCAD program
since the 2016 CDEGS User’s Group Meeting. These are summarized as follows:
 The Advanced Options screen that was introduced last year to give access to advanced
settings for the MALT, MALZ, and HIFREQ modules is now complete. As a result, it is no
longer necessary to use Input Toolbox to specify data for those modules.
 By default, SESCAD now uses the new Soil Model Editor to edit the characteristics of
the soil model and the new SESLibrary to retrieve the properties of conductors. The
program also gives access to the new SESResultsViewer tool to view the computation
results.
 A new tool to define soil volumes of cylindrical and other shapes was introduced for the
MALT and MALZ modules.
 Data stored in Google Earth KML files can now be imported in the program.
 The EMF applied to conductors (in the MALZ module) and user-defined currents assigned
to conductors (in the HIFREQ module) are now copied to the conductor paths created by
the Shift Path operation of the Path Operations tool.
 New warnings are now issued under various conditions related to missing specifications
for the soil model.
Also, the user-interface of the program is now available in French. The remainder of this
document describes these enhancements in greater detail.

2 The Advanced Options Screen

Since it was first introduced as a graphical editor for MALT, MALZ, and HIFREQ files, there were
always a few advanced settings of those programs that could not be edited in SESCAD. Editing
those settings required using CDEGS – Specify (Input Toolbox), which could be inconvenient.
At the last CDEGS User’s Group Meeting, a beta version of a new component, the SESCAD
Advanced Options screen, was first introduced to give access to those advanced settings from
within SESCAD. In this new version of the program, the implementation of this component was
completed. This feature can be accessed using Define | Advanced Options in SESCAD. By
default, this command now loads the Advanced Options screen, shown in Figure 1 for MALZ.
Consult the online help of SESCAD for a detailed list of available settings and a description of their
function.

Copyright © 2017 SES ltd. All rights reserved. Page 6-1

PART I: NEW FEATURES

Figure 1: The Advanced Options screen used to define advanced options affecting the computations for
MALZ in SESCAD. Similar screens are available for MALT and HIFREQ.

3 New Editors and Components

SESCAD now uses the new Soil Model Editor component to edit the characteristics of the soil
model and the new SESLibrary component to retrieve the geometrical and electrical properties
of conductors.
The Soil Model Editor can be accessed using Define | Soil Model in SESCAD. A beta version
of this component was introduced at last year’s CDEGS User’s Group Meeting. Since then, the
implementation of this component was completed and it is now used by default in SESCAD and
in many other SES Software programs.
This component supports all features of the previous Soil Structure screen of SESCAD and adds
a few capabilities. In particular, for MALT cases you can now select to use the soil model computed
by RESAP and stored in the data Share File (SF_*.F11) for the current JobID and Working
Directory, if one is available (see Figure 2). Also, you can now import soil model data defined in
any SES Software input file (or in a Share File) by clicking Import Soil Data.

Page 6-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: The As per Share File option in the Soil Model Editor screen for MALT.
The new SESLibrary component (see Figure 3), which was also introduced as a beta version at
last year’s CDEGS User’s Group Meeting is now complete and is used as a replacement of the SES
Conductor Database. It is used in two contexts:
 For importing conductor data in the Conductor Types screen for MALZ and HIFREQ
(see Figure 4).
 As a stand-alone tool, using Run/Reports | View SESLibrary.

Figure 3: The SESLibrary component.

Copyright © 2017 SES ltd. All rights reserved. Page 6-3

PART I: NEW FEATURES

Figure 4: The Import from Database command in the Conductor Types screen for HIFREQ.
See [1] for more details on the SESLibrary component.
Finally, you can now visualize the computation results for the current document (if any are
available) using the new SESResultsViewer tool by clicking Run/Reports | View Results
with SESResultsViewer.

4 The New Create Composite Soil Volume Tool

The ability to include finite volumes of soil of different resistivity in either a uniform or
horizontally layered substrate has been available for several years in the MALT and MALZ
modules. This feature has been used over the years to create increasingly sophisticated soil models
comprising several regions with different resistivities.
In SESCAD, soil volumes can be defined in the Soil Model Editor screen and can then be
manipulated graphically by activating the Soil Model layer. As shown in Figure 5, soil volumes
created in this way are in the general shape of a prism. To create soil volumes of a different shape,
(for example, cylindrical soil volumes) several prismoidal volumes must be created and assembled
in an appropriate manner. This step can be rather difficult since it requires sticking several
volumes together in such a way that adjoining volumes share one face exactly.
This version of SESCAD introduces the new Create Composite Soil Volume tool (Figure 6) to
help with this task. This tool can create approximations of cylindrical and conical soil volumes,
which can be useful, for instance, for modeling concrete around tower foundations. It can also
create soil volumes consisting of a central prismoidal core surrounded by volumes with a different
resistivity. This can be used, for example, to model a concrete pad elevated above the ground or
the presence of a ditch around a substation area.

Page 6-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: General shape of finite soil volume.

(a)

Copyright © 2017 SES ltd. All rights reserved. Page 6-5

PART I: NEW FEATURES

(b)
Figure 6: The new Create/Edit Composite Soil Volume tool; (a): Cylindrical volume; (b):
“Rectangular” volume consisting of a core surrounded by volumes of different resistivity.
As shown in Figure 6, the data to be specified in this screen is minimal, and consists mainly of the
geometrical characteristics (position and size) of the volumes and of the soil resistivity.
This tool, which is available only for MALT and MALZ, can be invoked in two different ways:
 By activating the Soil Model layer and then using Insert | Composite Soil Volume.
- or -
 By selecting some objects and then using Tools | Surround with Soil Volumes.
When used in this last manner, the size and position attributes in the screen are initially defined
so that the resulting composite soil volume completely surrounds the selected objects. These
attributes can then be modified, if required, before completing the creation of the composite soil
volumes.
Note that composite soil volumes created with this tool can be edited as a unit (by selecting a
composite soil volume and clicking Edit | Edit Object), but only during the session in which
they were initially created. Once the file is saved and reloaded, only the individual soil volumes
are restored; the high-level description of the composite soil volumes is lost.

Page 6-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

(a) (b)

(c) (d)

Figure 7: Examples of composite soil volumes created with the Create Composite Soil Volume tool;
(a): Cylindrical volume with six faces; (b): Cylindrical volume with 12 faces; (c): Cylindrical volume with
12 faces, with the option Allow Volume Edges along Axis turned off; (d): Rectangular volume.

5 Importing KML Files in SESCAD

SESCAD can now directly import keyhole markup language (KML) files. The conversion method
is described in detail in [2], and basically consists of a standard Universal Transverse Mercator
(UTM) using standard UTM zones.
Path and Placemark objects are recognized, whether part of a multigeometry or not. Once
imported into SESCAD, paths appear as conductors following the original path nodes. Placemarks
are imported as vertical conductors indicating the locations of interest. In order to better
manipulate the created conductors, all those belonging to a path are assigned a new conductor
type having the name corresponding to that of the path, with the prefix [Path]. Conductors created
from placemarks are also assigned a new conductor type having the name of the placemark, with
the prefix [Placemark].
Naturally, if the file type is set to MALT, the conductor types can no longer be used to differentiate
between different source objects, but the conductors are still created in the correct locations.

Copyright © 2017 SES ltd. All rights reserved. Page 6-7

PART I: NEW FEATURES

Moreover, when importing into SESCAD, the tool is of most use when the Append option is
selected for conductor types in the File Import Options window.
In order to avoid unnecessarily large coordinates, having a single Placemark named ‘origin’ (not
case sensitive) in the KML file allows SESCAD to re-center the system such that this location is
(0, 0). Note as well that elevations are currently ignored, and all imports will be set to z = 0 ft
(or m).
For example, we can import the system shown in Figure 8 into a blank HIFREQ or MALZ file.

Figure 8: Example system as seen in Google Earth™ to be imported into SESCAD.

Upon importing, the program provides a report of the objects successfully recognized and the
coordinates of the user-defined origin (if one was defined), as well as the UTM zone and
hemisphere used for the conversion, as shown in Figure 9.

Figure 9: Message detailing what KML objects were imported into SESCAD, the coordinates of the user-
defined origin (if one was defined), as well as the UTM zone and hemisphere used for the conversion.
Figure 10 shows the system as it appears in SESCAD once imported.

Page 6-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 10: Conductor network in SESCAD resulting from importing the kml file shown in Figure 8.
SESCAD’s KML import feature also recognizes Structure Locations folders in KML files, i.e.
ordered lists of placemarks that actually represent a path (described in more detail in [2]).
You can add as many cumulative KML imports to this model as you wish, the only caveat being
that the same origin placemark exist in each. Note that the placemarks imported are intended as
location markers for future designs or specifications. The conductors themselves are unlikely to
be kept. You can use custom filters to select the range of [Placemark] conductor types (they will
be in sequence by design) in order to remove them more quickly.

6 Applied EMF and User-Defined Currents in Path Operations

The Shift Path operation in the Path Operations tool (Tools | Path Operations) is used to
duplicate a connected conductor path parallel to itself. This operation can be used, for instance,
to create models of rails or to add mitigation wires running parallel to a pipeline.
In previous versions of this tool, however, any Applied EMF (in MALZ) or User-Defined
Currents (in HIFREQ) specified for the conductors defining the path to be duplicated were not
transferred to the duplicate. Now, these values are copied to the duplicated paths. This is
particularly useful for the Applied EMF, which is often very similar on parallel paths separated
by a small distance.

Copyright © 2017 SES ltd. All rights reserved. Page 6-9

PART I: NEW FEATURES

7 Warnings for Missing Soil Model

When running a MALT, MALZ or HIFREQ case for which no soil model was defined, the program
will use either a uniform soil of 100 ohm-m or (in the case of MALT) a soil model computed by
RESAP and stored in the Share File (SF_*.F11) for the current JobID, if one is available.
This version of SESCAD now issues reminders regarding this behavior whenever saving a file that
does not define a soil model. It also displays a reminder when saving a MALT file that does define
a soil model even though a soil model computed by RESAP is available in the Share File (SF_*.F11)
for the current JobID, since in this case it is possible that the latter soil model was the one that
was intended to be used.
These warnings can be turned off (with a Do not show me again checkbox) if you feel you do
not need this protection. You can turn them back on, if necessary, on the General page of the
General Settings screen (Options | General Options).

Figure 11: Warnings for undefined soil model.

8 French Version of SESCAD

The user-interface of SESCAD is now available in French. The French user-interface of the
program is used by default when run on a French version of Windows. If you prefer, you can select
to use the English version using Options | Langues | Anglais.

Page 6-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

9 Conclusion
This article described the new features and enhancements introduced in the SESCAD program
since the 2016 CDEGS User’s Group Meeting.

10 References
[1] SESLibrary Team, "New Features in SESLibrary," in User's Group Meeting, Portland, Oregon, 2017.
[2] RowCAD Team, "New Features and Improvements in RowCAD," in CDEGS Users' Conference
Proceedings, Boulder, Colorado, 2016.

Copyright © 2017 SES ltd. All rights reserved. Page 6-11

 UGM 2017 – CANNON BEACH, OREGON

7 NEW FEATURES IN SESLIBRARY

Francis Gougeon, Sheng Wei, Natela Botchorichvili, Hongbing Yin,

Sofiane Chabane, Peter Zhao, Jie Liu, and Nariman Tavana
Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

This article introduces the new features implemented in SESLibrary regarding the conductor, the cable and the
fence post databases. SESLibrary has replaced the original SES Conductor Database in all SES applications to
import conductor data of interest. All of the conductor classes in the original SES Conductor Database have been
transferred to SESLibrary, and reorganized with new conductor classes for convenience. In standalone mode,
SESLibrary allows users to create, edit or categorize their own conductors and save them in a conductor database.
The previous user-defined conductor classes can also be imported as a whole into SESLibrary. In the cable
database, a large amount of single-core coaxial cables consisting of various materials at different nominal voltage
levels have been added. Some typical fence posts have also been added in the new database. For rails stored in the
conductor database, the actual rail dimensions are now part of the available data and can be displayed on rail
schematics. SESLibrary not only provides a quick link to the SESImpedance tool but also brings in various
SESImpedance files of SES conductors in order to view and check their geometrical and physical characteristics. In
addition, online help can be easily accessed by pressing F1 in the interface.

1 Introduction
SESLibrary was first revealed during last year’s User Group Meeting (UGM) [1], and is an
enhancement to the existing SES Conductor Database. SESLibrary contains all the
conductors of SES Conductor Database and introduces an interesting classification
mechanism based on four different groups: Industry, Country, Application and Conductor
Class. This allows filtering the conductors based on any desired criterion.
Since SESLibrary was introduced, many features have been added and optimized. So far, three
types of databases have been implemented, namely, a conductor database, a cable database and a
database dedicated to fence posts. The new SESLibrary will progressively include several other
component databases, such as a GIS database, a transformer database, etc.
This year, three new conductor classes (ACSS, ACSS/AW, and ACSS/TW) have been added and
typical fence posts have also been created as an initial work of the fence database and the actual
rail dimensions have been included with an illustration picture. All the existing conductor classes
have been well organized in SESLibrary for a better user experience. Apart from the conductor
database, a lot of single-core coaxial cables at different nominal voltage levels and with various
materials have been added to the cable database.
SESLibrary is now used in other SES applications to import conductor data, replacing the SES
Conductor Database in this role. It can also be used as a standalone application to view
conductor properties or to add your own conductor data to the library. When used in this way,
you can also import user-defined conductor classes that were created using the old SES
Conductor Database to SESLibrary.
Various conductors stored in SESLibrary are now provided with their own SESImpedance
input files, giving some insight about the exact geometrical and electrical properties used to
compute the equivalent relative resistivity and permeability of the conductor. The

Copyright © 2017 SES ltd. All rights reserved. Page 7-1

PART I: NEW FEATURES

SESImpedance tool can be accessed easily from SESLibrary to open the SESImpedance
input file.
Details about these new features are introduced in the following sections.

2 New Databases in SESLibrary

2.1 Conductor Database
The conductor database of SESLibrary contains all the conductors from the old SES
Conductor Database. It stores data for various types of conductors: solid, hollow, stranded and
composite conductors.
This year, three new conductor classes based on Aluminum Conductor Steel Supported (ACSS)
have been added to the database. ACSS conductors are similar to ACSR conductors but sag less
and can operate at higher temperatures.
The conductors under these classes are gathered based on the geometrical and electrical
properties of aluminum and steel strands, as shown in the following:
 Simple ACSS conductors without any additional special feature;
 ACSS conductors whose aluminum strands are trapezoidal (ACSS/TW);
 ACSS conductors whose steel strands are made of aluminum-clad steel.
For rail conductors, a schematic view of the cross-section of the rail is now displayed in the 3D
viewer as shown in Figure 1. It illustrates the geometrical meaning of the data displayed in the
Properties panel regarding the rail dimensions.
The actual rail dimensions are used in order to find an equivalent circular conductor for each rail
section. More details on the procedure used to find the properties of the equivalent circular
conductor can be found in [2].

Figure 1: A rail cross-section.

Page 7-2 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

2.2 Cable Database

The cable database is also enriched by several new cables with various characteristics. The
database now contains data for more than 600 coaxial cables. In the future, other types of cables
will be added, especially multi-core cables. However, for the time being, cables cannot be
imported in any of SES applications.
Table 1: Category options used for the cable database.

Category Type Category Options

5 kV, 6 kV, 8 kV, 15 kV, 20 kV, 25 kV, 30 kV, 35 kV, 66 kV, 69 kV,
Nominal Voltage 77 kV, 110 kV, 115 kV, 132 kV, 138 kV, 154 kV, 161 kV, 220 kV,
230 kV, 345 kV, 380 kV, 400 kV, 500 kV

Core Material Copper, Aluminum

Sheath Material Copper, Aluminum

Insulation XLPE, EPR

Country China, South Korea, United States

Figure 2: Cable database.

Copyright © 2017 SES ltd. All rights reserved. Page 7-3

PART I: NEW FEATURES

2.3 Fence Post Database

This year, a fence post database has been added to SESLibrary, as shown in Figure 3. This
database stores data regarding the geometrical and physical characteristics of fence posts.
The fence posts stored in the database are of various shapes. In addition to posts of circular and
rectangular shapes, some irregular shapes (Omega-shape, T-shape, H-shape, etc.) are also
available, as shown in Figure 4. In the database, fence posts of any shape are represented by an
equivalent conductor of circular shape. Note that, currently, only the equivalent circular
conductor is displayed in the 3D Viewer for a fence post, not the original shape of the conductor.

Figure 3: Fence post database.

Figure 4: Some irregular shapes of the fence posts available in the database.

The fences posts are classified according to four categories, i.e., Industry, Country, Shape, and
Material.

Page 7-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 2: Category options used for the fence post database.

Category Type Category Options

Power Electric, Pipeline, Railway, Telecommunications, Civil

Industry
Engineering

Country United States

Shape Circular, Square, Rectangular, H-shape, Omega-shape, T-shape

Material Steel, Galvanized steel, Stainless steel, Aluminum alloys

3 SESLibrary Used in Import Mode

3.1 Access to SESLibrary from an SES Application
SESLibrary replaces the SES Conductor Database of earlier versions of SES Software, and
is now invoked automatically to import conductor data into all SES applications.
For instance, in the TRALIN application, clicking From Database in the Component Type
column brings the Category Filter screen to help you select the conductor of interest from the
database according to specified filters, as shown in Figure 5. Then, you can go to the SESLibrary
to find and import the conductor you want.

Figure 5: Access SESLibrary from TRALIN.

Copyright © 2017 SES ltd. All rights reserved. Page 7-5

PART I: NEW FEATURES

3.2 Finding and Importing a Conductor or a Fence Post into an SES

Application
In SESLibrary, conductor data can be imported from the conductor database and from the fence
post database. The following example shows how to import a conductor from the conductor
database; similar steps can be used to import a conductor from the fence post database.
After selecting filters in the Categories Filter screen and clicking OK, the filtered conductors
are displayed in the Conductors grid, as shown in Figure 6.
In the Conductors grid, the listed conductors can be sorted according to any column of the table.
This is quite useful to help selecting a conductor when there are a large number of them listed in
the grid.

Figure 6: Filtered conductors displayed in Conductors grid.

After selecting a conductor and clicking the Import button at the bottom of the grid as shown in
Figure 7, the selected conductor will be imported in the calling application and the SESLibrary
window closes automatically.

Page 7-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 7: Export the selected conductor to the calling SES application.

Note that the system of units used to display the data can be changed based on the selected option
under Display Units in the Home tab on the ribbon, as shown in Figure 7. Also, if different
filters need to be selected, the options in the Categories Filter screen can be accessed by clicking
the Library tab, as shown in Figure 8.
In addition, after a conductor is imported to the calling SES application, the previous selections
are kept when SESLibrary is reopened from the same calling SES application. The filtered
conductors based on the filters previously selected in the Category Filter screen are shown in
the Conductor grid as in Figure 6. To search for another conductor not listed in the grid, you can
redefine the filters using the Library tab on the ribbon, as shown in Figure 8. The procedure to
re-select and import a specific conductor is the same as that described above.

Figure 8: Redefine the filters to select conductors from the conductor database.

Copyright © 2017 SES ltd. All rights reserved. Page 7-7

PART I: NEW FEATURES

The Fence Posts database can be easily accessed by clicking the Fence Posts button above the
grid, as shown in Figure 9. Usage of this database is quite similar to that of the conductor database.

Figure 9: Access the fence post database.

4 SESLibrary Used in Standalone Mode

4.1 Adding User-Defined Conductors or Fence Posts
In standalone mode, you can define your own conductors and fence posts and add them to the
conductor and fence post databases. To do this, click the Add | User-Defined Conductor (or
User-Defined Fence Post) button in the Conductors (or Fence Posts) grid, as shown in
Figure 10. A new row is added to the grid. After entering the conductor data in the grid and in the
Properties panel and clicking Save Changes in the Home tab, this conductor is saved in
SESLibrary. User-defined conductors can be categorized using the Edit button in the
Properties panel to open the Specified Categories screen as shown in Figure 11. In this
screen, select the desired categories among all the available options for the conductor.

Page 7-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 10: Adding a user-defined conductor.

Figure 11: Specifying conductor categories.

Copyright © 2017 SES ltd. All rights reserved. Page 7-9
PART I: NEW FEATURES

4.2 Importing User-Defined Conductor Classes into SESLibrary

You can import your own conductor classes into SESLibrary in standalone mode. This feature
can be used in order to import conductor data specified in older versions of the conductor
database. To import a conductor class in the conductor database, follow these steps:

 Access the Databases-Conductors panel.

 In the Tools tab, click the Import from button as shown in Figure 12.
 In the Open screen, browse to the folder containing the user-defined conductor class files
(*.SCC) and select them as shown in Figure 13.
 After clicking OK in the Import Files screen (see Figure 14), the selected user-defined
conductor classes are imported in SESLibrary as shown in Figure 15.
 Once the imported conductors are displayed in the Conductors grid panel, you can select
them one by one to categorize them using the Edit button in the Properties panel.
 The last step is to save the modifications by clicking the Save Changes button in the
Home tab or the Save button in the Project backstage.

Figure 12: Tool to import user-defined classes.

Figure 13: Example of a user-defined conductor class.

Page 7-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 14:Pop-up window confirming the importing success.

Figure 15: Displaying the imported user-defined conductor class.

Copyright © 2017 SES ltd. All rights reserved. Page 7-11

PART I: NEW FEATURES

4.3 SESImpedance Files for SES Conductors

Some SES conductors are linked to a SESImpedance file that defines the geometry and material
characteristics of the conductor.

Table 3: SES conductors with SESImpedance files.

Conductor Class Conductor Name

Stranded Copper --8, 3-2, 4-1, 6-1, 8

Solid Copper 1, 2, 2/0, 6, 7, 8

6 AWG, 4 AWG, 2 AWG, 4 AWG-1, 2 AWG-1, 7 No. 10, 7 No. 8,

Copperweld CCS 40% 2/0, 7 No. 7, 7 No. 6, 7 No. 5, 7 No. 4, 19 No. 10, 4/0, 19 No. 9,
19 No. 8, 19 No. 7, 19 No. 6, 19 No. 5

TW1033-T28, TW1033-T23, Stilt-720, Puffin-780, Peacock-591,

Partridge-257, Ostrich, Lark-400, Goldfinch-634, Flicker-480,
3M ACCR
Eagle575, Condor-TW-795, Condor-824, Cardinal-967, 788-
T26, 680-T19, 558-T73, 427-T13

ASC (A1) Trillium, Pigweed, Jessamine

ACSR Jessamine, Ostrich, Linnet, Ibis

AAC Lily

When an SES conductor is linked to a SESImpedance file, the name of the file is displayed in
the Properties panel under SESImpedance file as shown in Figure 16. This allows visualizing
the geometrical and physical characteristics of the related conductor.

Figure 16: A conductor with its corresponding SESImpedance file.

Clicking on the SESImpedance icon opens the SESImpedance program and displays the
cross-section of the file as shown in Figure 17.

Page 7-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 17: Computation of the equivalent electrical characteristics of conductors in SESImpedance.

5 SESLibrary Multi-Language Interface

Taking into account our worldwide customers, SESLibrary is now available in three different
languages: English, French and Chinese. In the future, Spanish will be added as well.

Figure 18: Language settings in SESLibrary.

Copyright © 2017 SES ltd. All rights reserved. Page 7-13
PART I: NEW FEATURES

6 SESLibrary Online Help

Like the rest of SES applications and tools, SESLibrary also provides an extensive online help
and other functionalities allowing for updates, contacting SES support, looking for applications,
etc., these features can be accessed easily through the interface.
Moreover, the F1 key can be pressed in the interface of SESLibrary to easily access the context
sensitive help.

Figure 19: SESLibrary online help.

7 Conclusion and Future Developments

This article introduced the main new features available in SESLibrary. The introduced version
is still under development and other features and improvements are intended in the future.

Page 7-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

However, the conductor database is already used by default to import conductor data in all SES
applications, as a replacement for the SES Conductor Database of previous versions.
This version contains most of the features planned in last year’s UGM such as:
 Integration between SESLibrary and SES applications such as HIFREQ, ROW, AGP,
etc. to export a selected conductor or fence post to the calling application.
 Context sensitive online help.
 Automatic data interaction between SESImpedance files and the conductor database.
 More conductors and cables are added.
 Creation of a fence post database.
Furthermore, SESLibrary is now available in three different languages and allows for importing
existing user-defined conductor classes.
The following major features are planned for the future:
 Adding more cables and conductors.
 Including GIS and GIL databases.
 Implementing search tools (using keywords or values to locate the items in the database).

8 References

[1] S. Wei, E. Dawalibi, G. Noel, N. Botchorichvili, J. Liu, M. Siahrang, Y. Yang, T. Mukherjee and S.
Chabane, "Introducing the New SES Object Library," in CDEGS Users' Conference Proceedings,
Boulder, Colorado, 2016.

[2] J. Liu, F. P. Dawalibi, Y. Yang, J. Ma and R. D. Southey, "Determination of Rail Conductor

Characteristics for Electromagnetic Interference Studies," in EMC 2002, Beijing, China, 2002.

Copyright © 2017 SES ltd. All rights reserved. Page 7-15

 UGM 2017 – CANNON BEACH, OREGON

8 ROWCAD AND ROW IMPROVEMENTS

Yexu Li, Christian Voyer, Luis Valcárcel, Maxime Daigle,
Eric Dawalibi and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

1 Introduction
Several improvements have been made in the Right-of-Way software package (ROW) since our
last Users’ Group Conference in 2016. The major enhancements are:
1) In addition to the existing standard 2D-Curve plot generated in Excel format in the
Advanced plot module, a Keyhole Markup Language (KML) plot is available in this
version. It creates a Google Earth™ file that displays path coordinates combined with
computation results directly on the Google Earth™ world map.
2) The Create Circuit module can now automatically obtain the complete Touch-to-GPR
and Coating Stress-to-GPR percentage files when a partial circuit build is required.
Furthermore, when a partial build of the circuit is used, the initial SPLITS circuit
SP_ScenarioName_0.f05 will not be constructed, but it will be updated if it already exists.
3) The Create Circuit module is now able to automatically compute buried (bare or coated)
conductor shunt (ground) impedance for a component inside of a Cable and Group path.
4) The Create Circuit process has been improved to automatically account for the coating
characteristics of coated buried conductors.
5) In the Total Interference module, it is now possible to directly export the GPR and earth
current of the terminal grounding system. This can automatically account for the presence
of terminal grounding systems.
6) The existing maximum soil breakdown distance and various fault arcing and flashover
distance options have been improved by considering the X/R ratio of the system.
Therefore, it is possible to estimate critical soil breakdown and flashover distances due to
fault conditions more precisely, whether initiated by lightning or not, based on different
approaches.
7) For steady state conditions, the Reference conductor for Total Interference computations
now offers various filtering options such as phase number, conductor-type, coating-type
and profile number, etc., to precisely control the resulting envelopes.
8) A few customizations and options have been added to the Advanced plot engine:
 You can now request to optionally add the target acceptable Design Objective
threshold to the plot.
 You can define an offset value for the section (span) mileage using the Offset field if
a Terminal/Profile is requested for the plot.
 The plot for two terminals/profiles system is mirrored along the Y-axis to have a
single curve, instead of two curves for each separately.
9) Miscellaneous
 All input string fields, such as names, notes etc., are fully compatible with the target
languages used, particularly in the Chinese language.
Copyright © 2017 SES ltd. All rights reserved. Page 8-1
PART I: NEW FEATURES

 A message window suggests to build the network using RowCAD, when a new ROW
project is created.
 You are now able to run ROW concurrently with Output ToolBox (or other CDEGS
computation modules).
 A column representing the soil resistivity for holiday resistance computation is added
in the file “.\ScenarioFolder\HolidayResis_x.txt”.
10) Improvements in RowCAD
 KML file import feature can now recognize polylines and placemarks (pins) that are
within a multigeometry, thereby avoiding the rather laborious task of editing the text
file to expose the objects.
 The program now uses the polyline name as the default Path name, avoiding the
common issue that all paths have identical names.
 The Paths panel includes a global polyline visibility control.
 Creating cross-sections from the Cross-Section Warehouse has been made
significantly more efficient for cross-sections with a large number of components
(typically cable duct banks).
 A Copy Cross-Section functionality has been added to the Cross-Sections panel.
 The uniform soil resistivity column is now disabled when a valid Soil Definition
File has been specified, in order to avoid displaying seemingly conflicting soil data.
 The name of a cross-section being edited now appears in the cross-section window
title bar.
 The data format (Polar or Cartesian) for the Equivalent Source Impedance,
Voltage or Current in the Terminal Energization screen, the labels are now full
words rather than symbols. Also, the default values have been change in order to have
the Equivalent Source Impedances defined in Cartesian format, but the Phase-to-
Neutral Voltages and Source Currents in polar notation.
This article discusses these new features and briefly explains how to use them.

2 KML Plot on Google Earth Map

A new plot type, KML, is now available in Version 16. It creates a Google Earth™ file that displays
path coordinates in combination with computation results directly on the Google Earth™ world
map. A KML file is created in the Steady State scenario subfolder, in the case of steady state
results, and in the Envelopes scenario subfolder in the case of Fault Conditions results. Three
types of KML plots are produced: Elevation Plots (path with an elevation that represents the
results), Extent Plots (paths with zero elevation that simply show the layout of phases belonging
to the target path) and Shaded Elevation Plots (colored shading shows whether the result is above
or below the design objective). The screen depicting the above process is shown in Figure 1. Figure
2 illustrates an example.
Reference Point (for KML plots only): The program uses the correspondence between a point
in the Right-of-Way system (2D Cartesian coordinates) and a reference latitude and longitude to
convert the entire system coordinates to latitude and longitude.
Display Scaling Factor (for KML plots only): The scaling factor applied to the computation
results and the Design Objective threshold value when determining the elevation of the results
above the earth on the map.

Page 8-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 1: KML plot window.

Figure 2: Pipeline GPR Intensity plot on Google Earth. In order to better display the data, a large plot
scaling factor can be used. The Shaded Elevation Plot includes a shaded green color when the plotted
quantity is below the stated threshold, and a shaded red color when the plotted quantity is above the
stated threshold.

3 Combine/Construct Touch-to-GPR and Coating Stress-to-

GPR Percentages Automatically under Steady State
Conditions
Under steady state conditions, the induced voltage of an exposed line is computed with Process
Circuit Model No. (i.e., run SPLITS computation module) in the Process window of ROW,
CDEGS SPLITS, or SESBatch. Here, the induced voltage is computed with respect to remote earth
(i.e., GPR), not with respect to local soil (i.e. touch or stress voltage), which is the quantity referred
to by most standards concerned with safety. When no mitigation wire is installed and the pipeline
is well coated, the GPR is very close to (typically slightly higher than) the touch voltage. However,
in the presence of mitigation wire(s) (or any grounding grid, deep well, or anode beds), the GPRs
induced on the pipeline result in significant current leaks into the earth that raises the local soil
potential. As a result, the touch voltage can be a fraction of the GPR depending on the

Copyright © 2017 SES ltd. All rights reserved. Page 8-3

PART I: NEW FEATURES

characteristics of the soil. Consequently, touch voltages must be computed instead of GPR in order
to correctly evaluate the effects of the exposed line interference under steady state conditions.
A simple but approximate method to account for the effect of mitigation wires on touch voltages
under steady state conditions was also introduced in 2013. This method applies a reduction factor
to the GPR computed by ROW in order to deduce the corresponding touch and coating stress
voltages. These reduction factors (also referred to as “Touch-to-GPR” for computing touch voltage
percentages and “Coating-Stress-to-GPR” percentages for computing 1 cm2 holiday leakage
current densities for corrosion risk assessment) are computed using a simplified model that
accounts for the local soil conditions and the correct characteristics of the mitigation wires but
includes only a relatively small length of mitigation wire, making the model very fast to run. These
percentages are computed and the percentage files are created when a circuit is built using the
Create Circuit module.
Previously, when a partial build was used to compute the line parameters for a specified range of
Regions in a Terminal (or an entire terminal), a corresponding file containing the required
percentages was created only for the defined range. In order to get the percentage files that
included all regions of all terminals, one had to combine the files into a single file, a manual
operation which can be error prone. With this new version of ROW, the complete percentage files
(“CoatingStressToGPRPercentage.txt” and “TouchToGPRPercentage.txt”) will be automatically
generated when “Construct the Circuit Model” is applied, if all regions of all terminals have
been computed.

Figure 3: Automatically generate the percentage files using the Construct the Circuit Model option.
Furthermore, when a partial build is performed and there is not yet an existing complete build
file, the new version of the program will compute the line parameters, but the initial complete
SPLITS circuit SP_ScenarioName_0.f05 will not be constructed. However, if the initial SPLITS
circuit SP_ScenarioName_0.f05 exists already, it will be updated after the line parameters for
this partial build are computed. This will prevent users from starting a run based on an incomplete
circuit model.

4 Compute Ground Impedance Automatically for a

Component Inside of a Cable or Group Path
Previously, the program could automatically compute the shunt impedance of buried bare or
coated conductors based on the soil characteristics variations along the right-of-way, using the
Computed option in the Unit column of the Phase Leakage, Status… screen (Figure 4).
However, this feature only applied to the Principal path components, and was not available for
the Cable or Group (Pipe-type cable) paths.

Page 8-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 4: Select Computed to automatically calculate the shunt impedance of buried or coated
conductors.
In the new version of ROW, this feature is now available for a Cable or Group path component.
One of the important applications of this feature is modeling a bare conductor installed inside a
cable duct bank, parallel to the cables. In the past, one had to proceed as follows:
 Compute the shunt impedance of the bare conductor by setting up an independent MALZ
model. In the MALZ F05 file, a reasonably long conductor (e.g., 2 km or 6,560 feet) with
the proper conductor characteristics and the appropriate soil characteristics had to be
specified. After running the model in MALZ, the impedance of the wire would be obtained
for a given soil region. This had to be done for each soil region in the right-of-way.
 Define the Shunt Impedance of the conductor in the Phase Leakage, Status… screen
in ROWCAD or ROW with the computed shunt impedances for all regions, or replace the
Shunt Impedance in Modify Circuit.
While it is still possible to use the method described above to define the shunt impedance, in the
new version of ROW, the program can automatically compute the shunt impedances of buried
bare or coated conductors of a Cable or Group path. Select the Computed option in the Unit
column of the Phase Leakage, Status… screen for the target phase, regardless of what path
type the phase belongs to.

5 Line Parameters Accounting for the Coating Characteristics

of Buried Conductors with Discrete Grounds at Specific
Locations
The Create Circuit module builds the circuit model of the network that will be used to carry out
the EMI computations. It does this by running automatically the line parameter module
(TRALIN) for specified Terminals and Regions and by assembling the resulting line parameters
into an electric network definition (SPLITS) file.
Previously, the conductor coating properties, e.g., coating resistivity (conductivity) and
permittivity, were not considered in the line parameter computation for the conductor shunt
impedance, i.e., the conductors are always assumed to be bare in the TRALIN line parameter
computation. This is reasonable because:
 For aboveground conductors, the computed shunt impedance (i.e., capacitance) is not
significantly affected by the coating, if any.
 For buried conductors, the shunt impedances were defined as Computed using MALZ
(for bare conductors such as mitigation wires), Ohms/Section (for metallic structures,
poles or towers connected to the shield or neutral wires), Ohms/UnitArea (for pipelines)
Copyright © 2017 SES ltd. All rights reserved. Page 8-5
PART I: NEW FEATURES

and Ohms/UnitLength (for railways). See Figure 4. The shunt impedance of the
conductors will be replaced with the defined or MALZ computed value after TRALIN has
completed the line parameter computation.
With the new version of ROW, the line parameter computation in TRALIN runs will account for
the conductor coating, if any, no matter what choice was made in the Unit column of Phase
Leakage… initially. The shunt impedance will then be replaced with the actual values specified
in the Unit definition, allowing for the computed TRALIN value (that accounts for any coating)
to be used otherwise.
This may sound a little bit confusing but using an example will make it clearer. Suppose that a
pipeline is installed with anodes connected regularly to it (each anode has a grounding resistance
of 6.3 ohms). We can model the pipeline as a coated pipeline with connected anodes (i.e.,
grounding structures) defined as discrete “towers” in RowCAD or ROW. We therefore specify
both a coating resistivity (or equivalent resistance per unit length or area) and a phase leakage as
ohms/tower (see Figure 5). In Create Circuit, when TRALIN computes the line parameter, the
program will compute the pipe leakage impedance as a coated pipe from the coating specification
for the sections/regions where there are no anodes (towers) defined. Then, in the sections where
the anodes/“towers” are defined the pipeline section will have the shunt impedance with the
specified anode ground resistance (i.e., 6.3 ohms).

Figure 5: Define both coating and shunt impedance.

Note: For bare conductors such as mitigation wires, it is very important that the shunt impedances still have to use the Computed
(Figure 4) option which determines the conductor shunt impedance using the MALZ computation module. The line parameter
TRALIN computation module should not be used for buried bare conductor shunt impedance computations. This is because TRALIN
assumes infinite long conductors and uses an equivalent uniform soil in ROW by default.

6 Model Terminal Circuit Ground in Total Interference

The SPLITS circuit model allows the specification of several independent ground impedances at
each terminal. The different busses of a terminal can be connected to any of the grounds defined
for that terminal. This can greatly simplify the input data specification for complicated circuit
models with complex connection scenarios. Several different earth impedances can be specified
for a terminal. This way a single terminal can be used to represent several substations or
generating stations or exposed line terminating grounding systems. See Figure 6.

Figure 6: Define grounding point for a terminal.

Page 8-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Previously, the terminal ground impedance(s) specified in the Energization screen in Right-of-
Way (or ROWCAD) will only be accounted for in the SPLITS inductive computation but will not
be accounted for in Total Interference. This assumption is valid if the terminals are relatively
distant from the exposed line. If this is not the case, current returning to the source can lead to a
large soil GPR at, and in the area surrounding, the terminal ground and the adjacent towers. It is
important to consider the earth conduction effects of the current return to the source by modeling
the terminal ground explicitly in Total Interference.
In order to have the terminal grounds exported in Total Interference MALZ models, and take
the above-described effects into account previously, we needed to ensure that the last section of a
terminal ended at the terminal ground location, and we had to place the substation impedance as
a shunt impedance on the neutral phase rather than as a SPLITS terminal ground, i.e.:
1) Set the terminal ground impedance in the energization screen to a very high value, say
999999 + j999999 ohms in order not to account for this ground twice (see Step 2).
2) Replace the last section shunt impedance in SPLITS with the actual terminal grounding
system impedance of 0.1 ohm using Modify Circuit as shown in Figure 7.

Figure 7: Define grounding point at the last section in a terminal.

3) Add the grounding grid in the MALZ Total Interference model. Be sure to add a
connection conductor (preferably insulated) such that once the Total Interference file is
produced, the energized tower that actually represents the substation ground is connected
to the MALZ grid.

Figure 8: Define grounding system for a terminal in Total Interference.

In the new version of ROW, an option has been added in Total Interference which instructs
the program which ground to use in order to compute the terminal end GPR or earth current

Copyright © 2017 SES ltd. All rights reserved. Page 8-7

PART I: NEW FEATURES

values. This is achieved using the “Use Terminal Ground” check box shown in Figure 9. If this
option is checked, the program will export the terminal ground GPR (or earth current) otherwise
it will export the GPR (or earth current) of the last section ground.
For example, let’s assume that a terminal includes 20 spans. All tower grounds including the last
one are 20 ohms and the substation ground is 0.1 ohm as shown in Figure 6 (i.e., the terminal
ground point is 2 which is specified as 0.1 ohm). Let’s assume further that the last section ground
is far from the terminal ground and is disconnected from the terminal energization source (i.e.,
terminal equivalent source impedance for that phase is very large). Let’s assume now that after
processing the SPLITS circuit the computation results reveal that the GPR of the last section is
50,000 V while the GPR of the terminal ground point is 1,000 V.
1) The Total Interference process will export the last section GPR (or earth current) as
1,000 V, if “Use Terminal Ground” option is checked (the new method).
2) The Total Interference process will export the last section ground GPR (or earth
current) as 50,000 V, if “Use Terminal Ground” option is not checked (which is
compatible with the previous method).
Consequently, with this option the user can decide which option is more appropriate for his case.
Of course, if the last section was solidly connected to the terminal energization source for that
phase, then the results should be essentially similar because the currents in the proximate
grounds will divide inversely proportional to the ground impedances. In other words, the two
grounds share approximately the same GPR.

Figure 9: Use Terminal Ground option in Total Interference.

7 Evaluating the Vulnerability of Arcing to Exposed Lines

under Fault Conditions Accounting for X/R Ratio
The existing maximum soil breakdown distance and various fault arcing and flashover distance
options have been introduced in ROW since 2015. Therefore, it is possible to estimate critical soil
breakdown and flashover distances due to fault conditions in ROW, whether initiated by lightning
or not, based on different approaches.
This year, we improved the fault arcing and flashover distance computation methods by allowing
the user to consider the X/R ratio (representing the decrement factor) of the network at the fault
location. This factor represents the additional energy that can be present in the fault current due
to a decaying exponential component. Based on sinusoidal currents, it can be applied as a
multiplier to the computed symmetrical earth leakage current (or GPR).

Page 8-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Since the fault current provided by the planning department of utilities is usually an RMS value,
but the arcing and flashover are actually initiated by the peak value at the onset of the fault, we
need to compute the arcing or flashover distance by converting the RMS GPR or current value. To
do so, we multiple the RMS value by √2 to obtain the peak symmetrical sinusoidal value, then we
multiply this number by an additional factor representing the asymmetrical value at the initial
instant of the fault due to the inductance of the circuit (per IEEE Std 367-1996, Section 5.4.3).
The following overall multiplication factor is used to convert the RMS value to the peak value
accounting for the maximum DC offset:

Figure 10: Computing arc or flashover distance accounting for the X/R ratio.

8 Reference Conductor for Total Interference Computations

Offers Various Filters to Discriminate between Various
Choices under Steady State Conditions
When carrying out interference studies, it is convenient to be able to obtain the envelope of the
touch voltage along selected profiles with respect to some specific reference conductors. This
simplifies the analysis of cases where conductor paths are close to each other, or even crossing
one another. By default, the program search algorithm computes stress voltages such as touch
voltages with respect to the nearest conductor segment. This can cause problems when the touch
voltage with respect to a specific set of conductors must be computed. It can end up computing
the touch voltage to a conductor that is not listed in the desired set of conductors and hence will
report incorrect values.
As you know, it is possible to restrict the computation of touch voltages to selected conductors,
thereby avoiding the above problem. Furthermore, you can select observation profiles or a range
of profiles while ignoring the rest of observation profiles. All these options and filters had been
added in the Advanced plot screen. However, until now, these functions were only available to
the computations under fault conditions.
In the new version of ROW, all filter algorithms have also been implemented in the envelope plots
for steady state total interference calculations.
The Use Reference Conductor Group check box allows you to restrict the search to
conductors with certain characteristics. The available characteristics are:

Copyright © 2017 SES ltd. All rights reserved. Page 8-9

PART I: NEW FEATURES

 Conductor Type
 Conductor Coating Type
 Conductor Radius
 Phase Number
 Conductor Depth
The Use Specified Profile Group check box allows you to restrict the profiles to be plotted.
See Figure 11.

Figure 11: Filters in steady state total interference envelope plots.

9 Customizations and Options in the Advanced Plot

The relevant computed values are displayed once the requested computations are completed.
ROW provides various plots in spreadsheet format in the Advanced… plot module. In the new
version of ROW, new options and customization are available:
1) Users can request to optionally add the target acceptable Design Objective threshold to the
plot (Figure 12):

Figure 12: Target Design Objective threshold added to the plot.

Page 8-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2) An offset can be added for the section (span) mileage using the Offset option. This can only
be applied to one Terminal and to one Profile in the Total Interference MALZ results. In
Figure 13, a 10,000 feet offset is added at the beginning of the plot to account for a target
line length that has not been modelled along the right of way.

Figure 13: Offset is added for a 1- terminal/profile plot.

3) You now have the option to mirror the results pertaining to two terminals in SPLITS or
two profiles in MALZ plots along the x-axis to have a single continuous curve, instead of
two curves displayed separately. See Figure 14.

Figure 14: Mirrored plots.

10 Miscellaneous
A few additional items have been included in ROW to improve its performance. They are:
1) SES software includes multi-language support, such as Chinese, in of its interfaces,
documentation, help topics, etc. However, in previous versions of ROW, many input string
fields, such as phase names, terminal names, etc., often had language compatibility
problems. When you would enter a string in other languages, e.g., Chinese, it is possible
that the entered string is not shown correctly when you reopened the project. This problem

Copyright © 2017 SES ltd. All rights reserved. Page 8-11

PART I: NEW FEATURES

has been fully fixed in the new version of ROW. The new version entries are fully
compatible with all languages used in SES software.

Figure 15: Save and Load correctly multi-language screens and documentations.
2) The classical method for building a right-of-way network in the ROW program is to go
through a series of tabular interfaces where all the path coordinates that are present in the
right-of-way must be entered. This can be time-consuming as well as error prone,
especially for a complicated right-of-way network. RowCAD, which can fully describe a
right-of-way network with a user-friendly graphical interface, was introduced in 2011.
A message window has been added to suggest the use of RowCAD whenever a new ROW
project is created. Furthermore, you can turn on or off the message according to your
preference.

Figure 16: Message window recommending to use ROWCAD to define a new network.
3) ROW is a highly integrated software package. It calls and shares many other SES modules
such as Output Toolbox, etc., regularly. In previous versions, some modules of ROW, e.g.,
Create Circuit, Create Envelope Plots etc. cannot run when CDEGS Output Toolbox
was open.
The new version of ROW can continue all its processes while Output Toolbox is running
simultaneously.
4) ROW is able to compute and to report the leakage current density through 1 cm 2 coating
holidays to evaluate the AC corrosion status. The holiday current density is estimated
along the entire pipeline as being the ratio of the coating stress voltage and the holiday
resistance. The resistance of a 1 cm2 holiday generally consists of two contributions, acting

Page 8-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

in series: the resistance of a 1 cm plate located in a semi-infinite medium and the resistance
of the cylindrical volume of local soil contained in the holiday. 2 methods are available in
ROW: NACE Standard (which only includes the plate contribution) and the Advanced
Method (which includes both contributions), using the lowest soil resistivity at the
depth(s) of the pipe, for each soil region.
The new version of ROW, a column that prints out the soil resistivity used for the holiday
resistance computation is added in the file “.\ScenarioFolder\HolidayResis_x.txt”. This
can provide user for useful information.

Figure 17: Print out the soil resistivity used for holiday resistance computation.

11 Improvements in RowCAD
Several new features have been incorporated in RowCAD to increase efficiency in the design
process. Work on RowCAD also focused on improving clarity within the application.
The Keyhole Markup Language (Google Earth™ kml) file import feature can now recognize
polylines and placemarks (pins) that are within a multigeometry, thereby avoiding the rather
laborious task of editing the text file to expose the objects.

When Automatic Path Assignation Mode is selected (which is the default state) in order to
have polylines that have their Central Site Mapping Points and Terminal(s) defined be
automatically assigned to a newly created Path, the program now uses the polyline name as the
default Path name, avoiding the common issue that all paths have identical names.

To simplify the inspection of a model with a large number of polylines, the Paths panel now
includes a global polyline visibility control. You can check or uncheck the Visible box in the
column header line (Figure 18) to make all polylines either visible or not. Note that the checkbox

Copyright © 2017 SES ltd. All rights reserved. Page 8-13

PART I: NEW FEATURES

will also reflect the visibility state of polylines if individual polylines are manually made visible or
invisible.

Figure 18: A global visibility control available in the Paths panel. The checkbox will reflect the visibility
state of polylines, but can also be used to assert all polylines as visible or not with a single click.

Creating cross-sections from the Cross-Section Warehouse has been made significantly more
efficient for cross-sections with a large number of components (typically cable duct banks).
A Copy Cross-Section functionality has been added to the Cross-Sections panel (Figure 19).

Figure 19: A Copy Cross-Section functionality now included in the Cross-Sections panel. Note that
the cross-section is copied, but not the assignment to a given polyline.

The uniform soil resistivity column is now disabled when a valid Soil Definition File has been
specified, in order to avoid displaying seemingly conflicting soil data (Figure 20). The program
also has improved recognition of the soil model type being specified and can more clearly flag
invalid assignments.

Page 8-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 20: The uniform soil resistivity column (with the header 𝜌 (Ω ⋅ 𝑚)) is now disabled when a valid
Soil Definition File has been specified to avoid displaying a resistivity that would not be used in the
computation.
The name of a cross-section being edited now appears in the cross-section window title bar.
Finally, in order to clarify the data format (Polar or Cartesian) for the Equivalent Source
Impedance, Voltage or Current in the Terminal Energization screen, the labels are now
full words rather than symbols (Figure 21). Also, the default values have been change in order to
have the Equivalent Source Impedances defined in Cartesian format, but the Phase-to-Neutral
Voltages and Source Currents in polar notation.

Figure 21: The data format labels for the Equivalent Source Impedance, Voltage or Current in the
Terminal Energization screen are now full words rather than symbols.

Copyright © 2017 SES ltd. All rights reserved. Page 8-15

 UGM 2017 – CANNON BEACH, OREGON

9 NEW FEATURES AND IMPROVEMENTS FOR

TRANSFORMERS IN SPLITS

Marjan Mehrabi and Simon Fortin

Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

An option for specifying different values for the zero sequence components and the positive sequence components
of transformer characteristics for three-phase transformers was introduced in SPLITS. This feature enhances the
capabilities of SPLITS in modeling of unbalanced fault scenarios in power networks. This article describes this new
feature in detail and shows examples of the effect of zero sequence parameters when modeling faults in SPLITS.

1 Introduction
In previous versions of the SPLITS computation module, three-phase power transformers were
modeled using the positive sequence short-circuit and open-circuit tests parameters supplied by
manufacturers under the assumption that the zero sequence test data are identical with the
positive sequence test data. However, depending on how the windings of power transformers are
connected and on the nature of their magnetic circuit, the positive and zero sequence components
of the transformer parameters are not necessarily equal (especially for large power transformers)
and this assumption can lead to underestimating the computed fault currents, in some cases.
This version of SPLITS introduces the possibility to specify different values for the positive and
zero sequence components of transformer parameters. The transformer characteristics can be
defined either by specifying the positive and zero sequence short-circuit and open-circuit tests
parameters or by explicitly specifying the transformer matrix components.
This article is structured as follows. The next section describes the general form of the transformer
impedance matrix and shows how to obtain the components of this matrix from the manufacturer
positive and zero sequence standard short-circuit and open-circuit test data. Next, the enhanced
interface for transformer data entry in CDEGS – Specify for SPLITS is introduced. This is
followed by a comparison between two study cases that model a single-phase-to-ground fault
scenario on two three-winding three-phase transformers, one with equal and the other with
unequal zero and positive sequence components. The final section shows how to use the same
computation method to manually implement transformers with unequal zero and positive
sequence components in the HIFREQ module of MultiFields.

2 Computing Transformer Impedance Matrix Using

Symmetrical Components
The SPLITS module requires the specification of phase-based self and mutual impedances. When
analyzing transformers, however, it is often more convenient to work with sequence components.
This section shows how the general form of the phase-based impedance matrix representing a
transformer is related to the sequence components of the winding impedances. The following sub-
sections then show how to compute those matrix elements from manufacturer short-circuit and
open-circuit test data.
Copyright © 2017 SES ltd. All rights reserved. Page 9-1
PART I: NEW FEATURES

2.1 Relation between Sequence and Phase Components

Consider a three-phase three-winding transformer with a three-limb core and three windings per
phase, as illustrated in Figure 1. This configuration yields the largest transformer matrix of all
transformer models available in SPLITS.

Figure 1: Three-phase three-winding transformer with three-limb core and three windings per phase
(winding connections are not shown) [3].

For such a transformer, the relationship between the current and voltage of each phase, for each
winding is shown by the following phasor equation:
𝑉𝑃𝑟𝑖−𝑃ℎ𝑎 (𝑍𝑠 )𝑎𝑎 (𝑍𝑚 )𝑎𝑏 (𝑍𝑚 )𝑎𝑐 … … … … … … 𝐼𝑃𝑟𝑖−𝑃ℎ𝑎
𝑉𝑃𝑟𝑖−𝑃ℎ𝑏 𝐼𝑃𝑟𝑖−𝑃ℎ𝑏
((𝑍𝑚 )𝑏𝑎 (𝑍𝑠 )𝑏𝑏 (𝑍𝑚 )𝑏𝑐 ) (… … …) (… … …)
𝑉𝑃𝑟𝑖−𝑃ℎ𝑐 … … … 𝑃𝑆 … … … 𝑃𝑇 𝐼𝑃𝑟𝑖−𝑃ℎ𝑐
(𝑍𝑚 )𝑐𝑎 (𝑍𝑚 )𝑐𝑏 (𝑍𝑠 )𝑐𝑐
𝑉𝑆𝑒𝑐−𝑃ℎ𝑎 … … … 𝑃𝑃
… … … … … … 𝐼𝑆𝑒𝑐−𝑃ℎ𝑎
𝑉𝑆𝑒𝑐−𝑃ℎ𝑏 =
(… … …) (… … …) (… … …) 𝐼𝑆𝑒𝑐−𝑃ℎ𝑏 (1)
𝑉𝑆𝑒𝑐−𝑃ℎ𝑐 … … … … … … … … … 𝐼𝑆𝑒𝑐−𝑃ℎ𝑐
𝑉𝑇𝑒𝑟−𝑃ℎ𝑎 … … … 𝑆𝑃 … … … 𝑆𝑆 … … … 𝑆𝑇 𝐼𝑇𝑒𝑟−𝑃ℎ𝑎
𝑉𝑇𝑒𝑟−𝑃ℎ𝑏 (… … …) (… … …) (… … …) 𝐼𝑇𝑒𝑟−𝑃ℎ𝑏
[ 𝑉𝑇𝑒𝑟−𝑃ℎ𝑐 ] [ … … … 𝑇𝑃 … … … 𝑇𝑆 … … … 𝑇𝑇 ] [ 𝐼𝑇𝑒𝑟−𝑃ℎ𝑐 ]

The transformer impedances are represented by a 9 × 9 matrix. The 3 × 3 blocks that are
identified separately in Equation (1) are the self and mutual impedances of the three phases of
each winding or between one winding and another. For example, ( )𝑃𝑃 specifies the phase and
inter-phase impedances of the primary winding, ( )𝑃𝑆 specifies the phase and inter-phase
impedances between the primary and secondary windings, and so on.
In order to compute the elements of this matrix, one can take advantage of the fact that
transformers are electrically symmetrical. According to power system fundamentals, in a
balanced three-phase system the self-impedances of all three phases are equal and the mutual
coupling between all phases are also equal (Figure 2). In this case, the relationship between the
symmetrical components and the phase components of the impedances is as follows:

Page 9-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: Three-phase system with coupling impedances.

𝑍𝑎𝑎 𝑍𝑎𝑏 𝑍𝑎𝑐 𝑍𝑠 𝑍𝑚 𝑍𝑚

𝑍𝑎𝑏𝑐 = [𝑍𝑏𝑎 𝑍𝑏𝑏 𝑍𝑏𝑐 ] = [𝑍𝑚 𝑍𝑠 𝑍𝑚 ] (2)
𝑍𝑐𝑎 𝑍𝑐𝑏 𝑍𝑐𝑐 𝑍𝑚 𝑍𝑚 𝑍𝑠
𝑍0 + 2𝑍1
𝑍𝑠 =
3 (3)
𝑍0 − 𝑍1
𝑍𝑚 =
3

where 𝑎, 𝑏, and 𝑐 indicate the phase number; 0 and 1 identify the zero and positive components
and 𝑍𝑠 and 𝑍𝑚 are the self and mutual impedances. Therefore, in order to obtain the elements of
each individual 3 × 3 block from Equation (1), only two elements (𝑍𝑠 and 𝑍𝑚 ) are needed.
On the other hand, Equation(3) shows that in order to obtain the self and mutual impedances, the
positive and zero sequence components of the impedances have to be separately computed. The
positive and zero sequence components of the impedances of this three-winding transformer for
one phase per winding are given by the following equations:

𝑍1_𝑃𝑃 𝑍1_𝑃𝑆 𝑍1_𝑃𝑇

𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑆𝑒𝑞𝑢𝑒𝑛𝑐𝑒 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠: [𝑍1_𝑆𝑃 𝑍1_𝑆𝑆 𝑍1_𝑆𝑇 ]
𝑍1_𝑇𝑃 𝑍1_𝑇𝑆 𝑍1_𝑇𝑇
(4)
𝑍0_𝑃𝑃 𝑍0_𝑃𝑆 𝑍0_𝑃𝑇
𝑍𝑒𝑟𝑜 𝑆𝑒𝑞𝑢𝑒𝑛𝑐𝑒 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠: [ 𝑍0_𝑆𝑃 𝑍0_𝑆𝑆 𝑍0_𝑆𝑇 ]
𝑍0_𝑇𝑃 𝑍0_𝑇𝑆 𝑍0_𝑇𝑇

In order to find the matrices from Equation (4), assume that 𝑃, 𝑆, and 𝑇 represent the primary,
secondary and tertiary, respectively. Figure 3 illustrates the one-line diagram that corresponds

Copyright © 2017 SES ltd. All rights reserved. Page 9-3

PART I: NEW FEATURES

the configuration indicated in Figure 1. The relation between the voltage and current per-phase is
as follow:
𝑉𝑃𝑟𝑖 = 𝑍𝑃𝑃 𝐼𝑃𝑟𝑖 + 𝑍𝑃𝑆 𝐼𝑆𝑒𝑐 + 𝑍𝑃𝑇 𝐼𝑇𝑒𝑟
{ 𝑉𝑆𝑒𝑐 = 𝑍𝑃𝑆 𝐼𝑃𝑟𝑖 + 𝑍𝑆𝑆 𝐼𝑆𝑒𝑐 + 𝑍𝑆𝑇 𝐼𝑇𝑒𝑟
𝑉𝑇𝑒𝑟 = 𝑍𝑃𝑇 𝐼𝑃𝑟𝑖 + 𝑍𝑆𝑇 𝐼𝑆𝑒𝑐 + 𝑍𝑇𝑇 𝐼𝑇𝑒𝑟 (5)

where 𝑍𝑃𝑃 , 𝑍𝑆𝑆 and 𝑍𝑇𝑇 are the self-impedances of the Primary, Secondary and Tertiary, and
𝑍𝑃𝑆 , 𝑍𝑃𝑇 and 𝑍𝑆𝑇 are the Primary-to-Secondary, Primary-to-Tertiary and Secondary-to-
Tertiary mutual-impedances.

Figure 3: One-line diagram of a three-phase, three-winding Transformer.

The matrix components in Equation (4) can be obtained from the zero sequence and positive
sequence short-circuit and open-circuit test data, as explained in detail in the next section.
Equation (4) shows that a 3 × 3 matrix is required for a three-winding transformer. Each
component of this matrix is a 3 × 3 sub-matrix with the structure shown in Equation (2), which
represents a three-phase system. Therefore, the total transformer impedance matrix, shown in
Equation (1), will be a 9 × 9 matrix.
(𝑍𝑠 )𝑎𝑎 (𝑍𝑚 )𝑎𝑏 (𝑍𝑚 )𝑎𝑐 … … … … … …
((𝑍𝑚 )𝑏𝑎 (𝑍𝑠 )𝑏𝑏 (𝑍𝑚 )𝑏𝑐 ) (… … …) (… … …)
(𝑍𝑚 )𝑐𝑎 (𝑍𝑚 )𝑐𝑏 (𝑍𝑠 )𝑐𝑐 … … … 𝑃𝑆 … … … 𝑃𝑇
𝑃𝑃
… … … … … … … … …
(… … …) (… … …) (… … …) (6)
… … … … … … … … …
… … … 𝑆𝑃 … … … 𝑆𝑆 … … … 𝑆𝑇
(… … …) (… … …) (… … …)
[ … … … 𝑇𝑃 … … … 𝑇𝑆 … … … 𝑇𝑇 ]

In this matrix, the diagonal components (marked by red color) of each submatrix are equal and
the off-diagonal components (marked by black color) of each submatrix are also equal. More

Page 9-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

explicitly, (𝑍𝑠 )𝑎𝑎 , (𝑍𝑠 )𝑏𝑏 and (𝑍𝑠 )𝑐𝑐 have the same value, and (𝑍𝑠 )𝑎𝑏 , (𝑍𝑠 )𝑏𝑎 , (𝑍𝑠 )𝑏𝑐 , (𝑍𝑠 )𝑐𝑏 , (𝑍𝑠 )𝑎𝑐 ,
and (𝑍𝑠 )𝑐𝑎 also all have the same value.

2.2 Transformer Impedance Matrix Reduction when Symmetrical

Components Are Equal
In general, the elements of the total impedance matrix for a three-phase three-winding
transformer are non-zero. However, if the symmetrical components of the transformer
parameters are equal, then, according to Equation (3), the impedances between non-
corresponding phases, which are the off-diagonal elements of each 3 × 3 block in Equation (6),
become zero. Therefore, equation (6) reduces to:
(𝑍𝑠 )𝑎𝑎 0 0 … 0 0 … 0 0
[ 0 (𝑍𝑠 )𝑏𝑏 0 ] [0 … 0] [0 … 0]
0 0 (𝑍𝑠 )𝑐𝑐 𝑃𝑃 0 0 … 𝑃𝑆 0 0 … 𝑃𝑇
… 0 0 … 0 0 … 0 0
[0 … 0] [0 … 0] [0 … 0] (7)
0 0 … 𝑆𝑃 0 0 … 𝑆𝑆 0 0 … 𝑆𝑇
… 0 0 … 0 0 … 0 0
[0 … 0] [0 … 0] [0 … 0]
[ 0 0 … 𝑇𝑃 0 0 … 𝑇𝑆 0 0 … 𝑇𝑇 ]

2.3 Computing the Elements of the Transformer Impedance Matrix from

Short-Circuit and Open-Circuit Test Data
In order to create the total impedance matrix of a three-phase transformer, as explained in the
previous section, the elements of the transformer impedance matrix should be calculated from
the short-circuit and open-circuit test data provided by the transformer manufacturer. The
available parameters from transformer test data are typically as follows:
𝑃𝑅𝑎𝑡𝑒𝑑 : Power rating of transformer.
𝑉𝑃 : Line-to-line primary winding nominal voltage.
𝑉𝑆 : Line-to-line secondary winding nominal voltage.
𝑉𝑇 : Line-to-line tertiary winding nominal voltage.
Open-Circuit Test Parameters:
𝐼𝐸𝑥𝑐−1 , 𝐼𝐸𝑥𝑐−0 : Positive sequence and zero sequence excitation current as a percentage of
primary nominal currents.
𝑃𝑁𝑜𝐿𝑜𝑎𝑑−1 , 𝐼𝑁𝑜𝐿𝑜𝑎𝑑−0 : Positive sequence and zero sequence no-load power losses.
Short-Circuit Test Parameters:
𝑍𝑃𝑟𝑖−𝑆𝑒𝑐−1 , 𝑍𝑃𝑟𝑖−𝑆𝑒𝑐−0 (%) : Positive sequence and zero sequence leakage impedance
measured from primary to secondary winding at the primary to secondary base power,
when secondary is shorted and tertiary is open (in per unit).
𝑃𝑃𝑆 : The base power from primary to secondary winding at which the primary to secondary
impedance was measured. (Assumed to be identical for positive and zero sequence
components.)

Copyright © 2017 SES ltd. All rights reserved. Page 9-5

PART I: NEW FEATURES

𝑍𝑃𝑟𝑖−𝑇𝑒𝑟−1 , 𝑍𝑃𝑟𝑖−𝑇𝑒𝑟−0 (%) : Positive sequence and zero sequence leakage impedance
measured from primary to tertiary winding at primary to tertiary base power, when
tertiary is shorted and secondary is open (in per unit).
𝑃𝑃𝑇 : The base power from primary to tertiary winding. (Assumed to be identical for
positive and zero sequence components.)
𝑍𝑆𝑒𝑐−𝑇𝑒𝑟−1 , 𝑍𝑆𝑒𝑐−𝑇𝑒𝑟−0 (%) : Positive sequence and zero sequence leakage impedances
measured from secondary to tertiary winding at secondary to tertiary base power, when
tertiary is shorted and primary is open (in per unit).
𝑃𝑆𝑇 : The base power from secondary to tertiary winding. (Assumed to be identical for
positive and zero sequence components.)
𝑃𝑆𝐶_𝑃𝑆1 , 𝑃𝑆𝐶_𝑃𝑆0 : Positive sequence and zero sequence short-circuit power losses from
primary to secondary winding.
𝑃𝑆𝐶_𝑃𝑇1 , 𝑃𝑆𝐶_𝑃𝑇0 : Positive sequence and zero sequence short-circuit power losses from
primary to tertiary winding.
𝑃𝑆𝐶_𝑆𝑇1 , 𝑃𝑆𝐶_𝑆𝑇0 : Positive sequence and zero sequence short-circuit power losses from
secondary to tertiary winding.
Using the open-circuit test data, the following six impedances: 𝑍1(0)_𝑃𝑃 , 𝑍1(0)_𝑆𝑆 , 𝑍1(0)_𝑇𝑇 from
Equation (4) can be obtained.
Open-Circuit Test: For the open-circuit test, the rated voltage is applied to the primary while the
secondary and tertiary windings are open:
The primary nominal current is:
𝑃𝑅𝑎𝑡𝑒𝑑
𝐼𝑃 = (8)
√3 𝑉𝑃
Hence:
𝑉𝑃 /√3
𝑍1_𝑃𝑃 =
(𝐼𝐸𝑥𝑐−1 )𝐼𝑃
(9)
𝑉𝑃 /√3
𝑍0_𝑃𝑃 =
(𝐼𝐸𝑥𝑐−0 )𝐼𝑃

From the no-load power losses, the resistive and reactive parts of 𝑍1(0)_𝑃𝑃 are:
𝑃𝑁𝑜𝐿𝑜𝑎𝑑−1
𝑅1_𝑃𝑃 =
3(𝐼𝐸𝑥𝑐−1 × 𝐼𝑃 )2
𝑃𝑁𝑜𝐿𝑜𝑎𝑑−0 (10)
𝑅0_𝑃𝑃 =
3(𝐼𝐸𝑥𝑐−0 × 𝐼𝑃 )2

𝑋1_𝑃𝑃 = √𝑍1_𝑃𝑃 2 − 𝑅1_𝑃𝑃 2

(11)
2 2
𝑋0_𝑃𝑃 = √𝑍0_𝑃𝑃 − 𝑅0_𝑃𝑃

Page 9-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Therefore:

𝑍1_𝑃𝑃 = 𝑅1_𝑃𝑃 + 𝑗𝑍𝑋1_𝑃𝑃 , 𝑍0_𝑃𝑃 = 𝑅0_𝑃𝑃 + 𝑗𝑍𝑋0_𝑃𝑃 (12)

The transformer voltage ratios are:

𝑉𝑆 𝑉𝑇 𝑉𝑇
𝑛𝑆𝑃 = , 𝑛 𝑇𝑃 = , 𝑛 𝑇𝑆 = (13)
𝑉𝑃 𝑉𝑃 𝑉𝑆

where 𝑉𝑃 , 𝑉𝑆 , 𝑉𝑇 are the phase voltage of the primary, secondary, and tertiary windings.
For the secondary and tertiary windings:

𝑍1_𝑆𝑆 = (𝑛𝑆𝑃 )2 𝑍1_𝑃𝑃 , 𝑍0_𝑆𝑆 = (𝑛𝑆𝑃 )2 𝑍0_𝑃𝑃

(14)
𝑍1_𝑇𝑇 = (𝑛 𝑇𝑃 )2 𝑍1_𝑃𝑃 , 𝑍0_𝑇𝑇 = (𝑛 𝑇𝑃 )2 𝑍0_𝑃𝑃

Using the short-circuit test data, the following impedances can be obtained:
𝑍1(0)_𝑃𝑆 , 𝑍1(0)_𝑃𝑇 , 𝑍1(0)_𝑆𝑇 from Equation (3).
Short-Circuit Test: In the short circuit test, the rated current is applied to the primary while the
secondary winding is shorted and the tertiary winding is open:

𝑍𝑃𝑟𝑖𝑚−𝑆𝑒𝑐−1 × (𝑉𝑃 )2
𝑍1_𝑆𝐶 =
𝑃𝑅𝑎𝑡𝑒𝑑
𝑍𝑃𝑟𝑖𝑚−𝑆𝑒𝑐−0 × (𝑉𝑃 )2 (15)
𝑍0_𝑆𝐶 =
𝑃𝑅𝑎𝑡𝑒𝑑

𝑃𝑆𝐶−𝑃𝑆1
𝑅1_𝑆𝐶 =
3(𝐼𝑃 )2
𝑃𝑆𝐶−𝑃𝑆0 (16)
𝑅0_𝑆𝐶 =
3(𝐼𝑃 )2

𝑋1_𝑆𝐶 = √𝑍1_𝑆𝐶 2 − 𝑅1_𝑆𝐶 2

(17)
2 2
𝑋0_𝑆𝐶 = √𝑍0_𝑆𝐶 − 𝑅0_𝑆𝐶

𝑍1_𝑆𝐶 = 𝑅1_𝑆𝐶 + 𝑗𝑍𝑋1_𝑆𝐶 , 𝑍0_𝑆𝐶 = 𝑅0_𝑆𝐶 + 𝑗𝑍𝑋0_𝑆𝐶 (18)

Copyright © 2017 SES ltd. All rights reserved. Page 9-7

PART I: NEW FEATURES

𝑍1_𝑃𝑆 = (𝑛𝑆𝑃 )√(𝑍1_𝑃𝑃 − 𝑍1_𝑆𝐶 )𝑍1_𝑃𝑃 , 𝑍0_𝑃𝑆 = (𝑛𝑆𝑃 )√(𝑍0_𝑃𝑃 − 𝑍0_𝑆𝐶 )𝑍0_𝑃𝑃 (19)

And similarly:

𝑍1_𝑃𝑇 = (𝑛 𝑇𝑃 )√(𝑍1_𝑃𝑃 − 𝑍1_𝑆𝐶 )𝑍1_𝑃𝑃 , 𝑍0_𝑃𝑇 = (𝑛 𝑇𝑃 )√(𝑍0_𝑃𝑃 − 𝑍0_𝑆𝐶 )𝑍0_𝑃𝑃

(20)
𝑍1_𝑆𝑇 = (𝑛 𝑇𝑆 )√(𝑍1_𝑆𝑆 − 𝑍1_𝑆𝐶 )𝑍1_𝑆𝑆 , 𝑍0_𝑆𝑇 = (𝑛 𝑇𝑆 )√(𝑍0_𝑆𝑆 − 𝑍0_𝑆𝐶 )𝑍0_𝑆𝑆

Finally, having the positive and zero sequence components of Equation (5) and using Equation
(3), the diagonal and off-diagonal elements of Equation (6) can be computed.

1.2.1 Effects of Winding Connections

The ability to specify different values for the positive and zero sequence components of
transformer parameters is available for the following types of transformer connections: Star-Star,
Star-Delta, Delta-Star, Delta-Delta, Star-Star-Delta, Auto-Star-Star, and Auto-Star-Star-Delta.
Note that for delta connections, the program assumes that the delta itself is open (so that no
current flows in the delta) during the tests requiring the corresponding winding to be open. If this
is not the case, the following modification should be made for the primary to secondary zero
sequence leakage impedance:

(𝑍𝑃𝑟𝑖−𝑆𝑒𝑐−0 )𝑀𝑜𝑑𝑖𝑓𝑖𝑒𝑑 =
= 𝑍𝑃𝑟𝑖−𝑇𝑒𝑟−0 + 𝑍𝑆𝑒𝑐−𝑇𝑒𝑟−0 (21)
− 2√𝑍𝑆𝑒𝑐−𝑇𝑒𝑟−0 (𝑍𝑃𝑟𝑖−𝑇𝑒𝑟−0 − 𝑍𝑃𝑟𝑖−𝑆𝑒𝑐−0 )

With this modification, Equations (18) are applicable, and the zero and positive sequence short-
circuit tests can be treated similarly. Moreover, in order to take the phase shift for Delta
connections into account, the diagonal and off-diagonal submatrices should be multiplied by 3
and √3, respectively. The reason is that Equation (5) relates the phase voltage and current; hence,
the impedances are also related to the phase impedance of the Star connection which should be
modified for the Delta connection (which is three times that of the Star connection). For example,
in the case of a Star-Star-Delta configuration, the transformer voltage ratios from Equation (13)
will be:

𝑉𝑆 𝑉𝑇 𝑉𝑇
𝑛𝑆𝑃 = , 𝑛 𝑇𝑃 = , 𝑛 𝑇𝑆 = (22)
𝑉𝑃 𝑉𝑃 /√3 𝑉𝑆 /√3

Therefore, it produces a coefficient of 3 for the second line of Equation (14) and of √3 for
Equation (20).

Page 9-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3 New Features for Three-Phase Transformer Characteristics

in SPLITS
The specification of unequal symmetrical components for three-phase transformers is available
in the SPLITS transformer interface. The necessary information can be specified as test data or by
directly providing the independent elements of the transformer matrix. The following sections
explains how to specify those parameters in the Transformer Characteristics screen of
CDEGS – Specify for SPLITS.

3.1 Transformer Characteristics: Test Data

Engineers normally have access to the manufacturer technical data sheet of power transformers
which contains the positive and zero sequence standard test data. This data can be specified in the
Positive Sequence Components and Zero Sequence Components tabs of the Three-
Phase Transformer data editor in SPLITS. The first tab (see Figure 4a) defines the positive
sequence test data. It contains the same test data that was available in previous versions of the
program. The second tab (see Figure 4b) contains the zero sequence parameters.

Copyright © 2017 SES ltd. All rights reserved. Page 9-9

PART I: NEW FEATURES

Figure 4: a) Positive Sequence Components. b) Zero Sequence Components.

By default, as indicated in Figure 4b, the zero sequence parameters are assumed to be equal to the
positive sequence parameters. If zero sequence test data is available, you can uncheck the
Positive and Zero Sequence Components Are Equal checkbox to specify the data. Note
that the program uses the corresponding positive sequence parameter for any parameter that is
left undefined in this tab.
As mentioned in Section 2.3, this new feature is only available for various combinations of Star
and Delta connections and not for the Grounding-Zigzag, Star-Zigzag, and Delta-Zigzag
connections, see Figure 5.

Page 9-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Zero sequence feature is not available for Zigzag winding connections.

3.2 Transformer Characteristics: Impedance

The Impedance option for the specification of transformer characteristics can mostly be utilized
by scientists who are familiar with the foundation of transformer impedance computation -
explained in Section 1 - or by users interested in learning and testing the behavior of transformers.
As indicated in Figures 6 and 8, there are two tabs available when selecting Impedance under
Transformer Characteristics: Corresponding Phases and Non-Corresponding
Phases. The Corresponding Phases tab contains the diagonal components of each submatrix
from Equation (6).

Copyright © 2017 SES ltd. All rights reserved. Page 9-11

PART I: NEW FEATURES

Figure 6: Corresponding Phases of Impedance matrix.

Page 9-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

These components are explicitly identified in Figure 7.

Figure 7: Corresponding Phases of Impedance matrix.

The Non-Corresponding Phases tab contains the off-diagonal components of each submatrix
from Equation (6). By default, the positive and zero sequence components of transformer
parameters are assumed to be equal. As explained in Section 2.2, in this case, the impedance
matrix components for the non-corresponding phases will become zero.

Copyright © 2017 SES ltd. All rights reserved. Page 9-13

PART I: NEW FEATURES

Figure 8: Non-Corresponding Phases of Impedance matrix are zero.

If the zero sequence components are available and not equal to the positive sequence components,
you can uncheck the Positive and Zero Sequence Components Are Equal check box to
specify the data (see Figure 9).

Page 9-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 9: Non-Corresponding Phases of Impedance matrix.

Copyright © 2017 SES ltd. All rights reserved. Page 9-15

PART I: NEW FEATURES

Figure 10 shows how to define the parameters on the Non-Corresponding Phases tab.

Figure 10: Non-Corresponding Phases of Impedance matrix.

3.3 Converting Test Data to Impedances

To get the actual values of the components of the transformer impedance matrix corresponding
to specified test data you can model the transformer with its test data in SPLITS, run the file and
look into the generated *.F72 file. This file lists the SPLITS commands necessary to implement
the transformer and these commands include all the components of the transformer matrix.
For example, consider a transformer with the characteristics indicated in Table 1:

Page 9-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 1: Y-D Transformer Test Data.

Transformer Connection Yg-D

Power Rating (kVA) 100,000

Primary Voltage (kV) 345

Secondary Voltage (kV) 161

Positive Sequence Primary-to-Secondary

10 @ Rated Power: 100,000 kVA
Leakage Impedance (%)

Positive Sequence Primary-to-Secondary

1 @ Rated Power: 100,000 kVA
Power Loss (W)

Positive Sequence Excitation Current (%) 0.1

Positive Sequence No-Load Power Loss (%) 1

Zero Sequence Primary-to-Secondary

5 @ Rated Power: 100,000 kVA
Leakage Impedance (%)

Zero Sequence Primary-to-Secondary

1 @ Rated Power: 100,000 kVA
Power Loss (W)

Zero Sequence Excitation Current (%) 0.1

Zero Sequence No-Load Power Loss (%) 1

This transformer is modeled in SPLITS as shown in Figure 11. The file can be found in the Example
folder and is named “SP_YD_Z0Z1diff_TestData.f05”.

Copyright © 2017 SES ltd. All rights reserved. Page 9-17

PART I: NEW FEATURES

Figure 11: Y-D Transformer.

Running this model generates a *.F72 file that contains the transformer impedance matrix in
Section 2:

Figure 12: The generated *.F72 file for the Y-D Transformer.
Page 9-18 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

The impedance values indicated in Figure 12 can directly be entered in the Corresponding and
Non-Corresponding Phases tabs, selecting Impedance instead of Test Data. However,
since the secondary winding is Delta connected, the self impedances of the Corresponding
phases for the secondary (identified as “Secondary self impedances” in Figure 12) should be
divided by 3 , and the mutual impedances of the Corresponding phases and of the Non-
Corresponding Phases (identified as “Mutual between Primary and Secondary” and “Add
Coupling between phases” in Figure 12) should be divided by √3 as explained in Section 3.
Note also that these transformer commands read from the *.F72 file can be copied to another
location (outside the central site) in order to build a transformer at any desired location.

4 Comparing Fault Currents for a Transformer with Equal

and Unequal Symmetrical Components
As an example of the use of the new capability of SPLITS to model transformer with unequal
symmetrical components, single-phase-to-ground fault currents for two similar power
transformers having equal and unequal symmetrical components for their leakage impedances
were analyzed in SPLITS. The transformer characteristics are summarized in Table 2. For
simplicity, all parameters but the leakage impedances are assumed to have equal symmetrical
components.
Table 2: Test Data for two Auto-YYD transformers with equal and unequal symmetrical
components.

Transformer Test Data Equal Symmetrical Unequal

Components Symmetrical
Components

Transformer Connection Auto-Yg-Yg-D Auto-Yg-Yg-D

Power Rating (kVA) 100,000 100,000

Primary Voltage (kV) 345 345

Secondary Voltage (kV) 161 161

Tertiary Voltage (kV) 13.8 13.8

10 @ Rated
Primary-to- 10 @ Rated Power:
Power: 100,000
Secondary 100,000 kVA
kVA
Positive Sequence
Leakage Impedance Primary-to- 9 @ Rated Power: 9 @ Rated Power:
(%) Tertiary 100,000 kVA 100,000 kVA

Secondary-to- 8 @ Rated Power: 8 @ Rated Power:

Tertiary 100,000 kVA 100,000 kVA

Primary-to- 10 @ Rated Power: 5 @ Rated Power:

Secondary 100,000 kVA 100,000 kVA

Copyright © 2017 SES ltd. All rights reserved. Page 9-19

PART I: NEW FEATURES

Primary-to- 9 @ Rated Power: 4 @ Rated Power:

Zero Sequence Tertiary 100,000 kVA 100,000 kVA
Leakage Impedance
(%) Secondary-to- 8 @ Rated Power: 3 @ Rated Power:
Tertiary 100,000 kVA 100,000 kVA

Positive and Zero Sequence Power Loss 1

1
(W)

Positive and Zero Sequence Excitation 0.1

0.1
Current (%)

Positive Sequence No-Load Power Loss 1

1
(%)

The primary winding is energized at the rated voltage, and phase A of the secondary winding is
solidly grounded. The two other phases of the secondary and tertiary windings are left open. The
fault current on phase A of the secondary winding is indicated in Figure 13 for both cases.

Figure 13: Fault Currents on two Auto-YYD transformers with equal (on the left) and unequal (on the
right) symmetrical components using GRSPLITS-3D.

Page 9-20 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The currents on the primary winding phases and circulating currents in the tertiary are compared
in Figure 14.

Figure 14: Fault currents on each winding of two Auto-YYD transformers with equal (on the left) and
unequal (on the right) symmetrical components from the SPLITS output file.

As indicated in Figure 14, the case with unequal symmetrical components can yield different and
possibly higher fault current values. Hence, the assumption of equal symmetrical components for
transformer parameters is not always conservative.

5 Symmetrical Components Implementation in Three-Phase

Transformer in MultiFields
The technique described in the previous sections can also be used to specify three-phase
transformers with unequal positive and zero sequence parameters in the HIFREQ module of
MultiFields, as shown by the following example. Consider the Auto-YYD transformer of Section
3; the conductor layout shown in Figure 15 can be used to implement this transformer in HIFREQ.
In this figure, conductors # 1, 2 and 3 represent phases A, B, and C of the primary, conductors 4,
5, and 6 represent phases A, B, and C of the secondary and conductors 7, 8, and 9 represent phases
A, B, and C of the tertiary winding. In order to create the connections of the autotransformer, the
end points of conductors # 1, 2, and 3 are connected to the starting points of conductors # 4, 5,
and 6. Also, the starting points of conductors # 4, 5, and 6 are energized with zero GPR

Copyright © 2017 SES ltd. All rights reserved. Page 9-21

PART I: NEW FEATURES

representing the grounding point of the Auto-Star connection. The Delta connection of the tertiary
is made by virtual connections, as shown in Figure 14. Three short conductors for each winding
were created to introduce the winding terminals. In order to model the fault scenario described
in Section 3, conductors # 13, 14, and 15 are energized with the primary phase-to-neutral rated
voltage, and conductor # 16 connected to phase A of the secondary winding is energized with zero
GPR.

Figure 15: Auto-YYD transformer model in HIFREQ.

In order to construct the impedance matrix, we use the MUTUAL command in HIFREQ (available
in command mode only). The components of the impedance matrix can be directly copied from
the SPLITS output file (sp_AutoYYD_LH_Z0Z1diff_TestData.F72) into the HIFREQ input file
knowing the general syntax of the MUTUAL command, see Figure 16.

Page 9-22 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 16: Auto-YYD transformer with unequal symmetrical components model in HIFREQ.

Copyright © 2017 SES ltd. All rights reserved. Page 9-23

PART I: NEW FEATURES

To create the case with equal symmetrical components simply specify zero for all self and mutual
non-corresponding phases (Figure 17).

Figure 17: Auto-YYD transformer with equal symmetrical components model in HIFREQ.

Page 9-24 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The results for those two cases are displayed in Figure 18. The fault current results are in very
good agreement with the results obtained with SPLITS in Section 3.

Figure 18: Fault current results for the Auto-YYD transformer with a) equal and b) unequal symmetrical
components using HIFREQ.

Copyright © 2017 SES ltd. All rights reserved. Page 9-25

PART I: NEW FEATURES

6 Example Files
The examples mentioned in this article, including the SPLITS test cases and the HIFREQ test
cases , are located in the “UGC 2017\Example Files\New Features and Improvements for
Transformers in SPLITS” subfolder of the SES Software conference distribution.

7 Conclusion
The ability to model transformers with unequal symmetrical components was added to the
SPLITS computational module. Zero sequence components options were added to the
transformer data editor to specify the symmetrical components of parameters provided in the test
technical data sheet of the manufacturer. These features make the results more accurate especially
for unbalanced fault scenarios which can be paramount for purpose of grounding studies. The
same technique is planned to be utilized in order to implement three-phase transformers in
HIFREQ in the future.

8 Acknowledgements
The authors thanks, colleagues Mr. Robert D. Southey for the inspiring article referred in [1],
Ms. Cathy Cheng for implementing the interface and Mr. Mark Boisjoli for his help to initiate the
study.

9 References
[1] R. D. Southey, M. Siahrang, “Having Fun with Transformers and Sequence Impedance Networks… The
SPLITS Way” in CDEGS Users' Conference Proceedings, 2014 Newport, Rhode Island, USA, 2014.
[2] V. Brandwajn, H. W. Dommel, and I. I. Dommel, "Matrix representation of three-phase N-winding
transformers for steady-state and transient studies," IEEE Trans. Power App. Syst., vol. PAS-101, no. 6, pp.
1369-1378, Jun. 1982.
[3] MathWorks Library, “Three-Phase Transformer Inductance Matrix Type (Three Windings)”,
www.mathworks.com.

Page 9-26 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

10 IMPROVEMENT TO FINITE VOLUME SUBDIVISION

TECHNIQUE: LEAKAGE CURRENT CRITERION
Farid P. Dawalibi and Nina Mitskevitch

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
The distribution of patches over the faces of soil volumes has a major influence on the computation results when
using finite volumes in the MALT and MALZ modules. The Adaptive Patch Subdivision technique ensures that the
size and distribution of the patches are consistent with the expected concentration of electric charges on the
volume faces, based on a number of physical criteria such as conductor proximity, resistivity contrast ratios,
location of patches on the faces of the volumes, etc. In order to improve further the quality of patch distribution, a
new criterion based on the current density leaking from conductor segments is introduced. This criterion ensures
that the soil volume faces located near conductor segments with large leakage currents receive a greater portion
of patch resources.

1 Introduction
In order to get accurate results while using soil volumes embedded into uniform or horizontally
layered soils, it is important to subdivide the soil volumes interfaces into small surface elements
or patches. The ideal patch distribution and size should reflect the intensity of the electric field
pattern on the soil volume faces. In order to achieve that, the Adaptive Patch Subdivision
technique has been introduced more than 10 years ago to optimize the patch subdivision. The
Adaptive Patch Subdivision is based on physical criteria (designated as Priority Criteria), such as
conductor proximity, segment length, resistivity contrast ratios, location of patches on the faces
of the volumes, etc. However, so far all the criteria related to conductors assumed that all
conductor segments are equally important for the patch subdivision process. In many cases,
however, not all conductor segments represent an equally strong source of electric field which is
related to the current leaking from the conductor segment surface. For example, some
conductors can be insulated or not bonded to energized systems, thereby having zero or very
little leakage current. A new criterion based on leakage current density has been introduced in
order to correct this flaw. It ensures that the soil volume faces located near conductors with
large leakage current densities receive an adequate portion from the total pool of patch
resources and, conversely, that the soil volume faces located near conductors with small current
densities do not “attract” an excessive patch subdivision.
It is expected that the introduction of this new criterion, which is described below in details, will
significantly improve the effectiveness of the Adaptive Patch Subdivision process.

2 Definition of Priority Criteria

The Adaptive Patch Subdivision algorithm currently offers the following priority criteria that
govern the patch distribution:
 Segment Proximity. Criterion based on the ratio (d/2+L) /L, where d is the closest
distance (from any point of a conductor segment) to any point on a neighboring patch
polygon, and L is the length of the conductor segment.
 Patch Area. Criterion based on the ratio d/√S, where d is the minimum distance from
any point on the patch of area S to any point on the closest conductor segment.

Copyright © 2017 SES ltd. All rights reserved. Page 10-1

PART I: NEW FEATURES

 Segment Length. Criterion based on the ratio LNetwork/L, where LNetwork is the sum of
the lengths of all the conductors in the grounding network, and L is the length of the
conductor segment.
 Resistivity Ratio. Criterion based on the resistivity ratios of the media on each side of
the soil volume face ρVolume/ ρNative Soil where ρ represents the resistivity of the indicated
media. This criterion is useful when multiple soil volumes of different resistivities are
specified. Normally, this criterion should always be used along with one of the other
criteria.

The following new criterion has been added to account for the electric field pattern on the soil
volume interfaces:

 Leakage Current. Criterion based on the electric field in the vicinity of the patch
interface. The leakage current (electric field) can be calculated assuming uniform or
layered soil (as a first approximation) depending on which situation appears to be the
most appropriate based on some topological considerations. This criterion can be
considered as based on the conductor segment leakage current intensities since electric
fields are at their maximum at the conductor soil interface and are proportional to these
leakage currents. The expression for this criterion is based on the ratio JMax/Jn, where Jn
is the leakage current (or leakage current density) of n-th conductor segment and JMax is
the maximum leakage current (or leakage current density) among all conductor
segments. The larger the value of the criterion expression is, the more likely the patch
will be subdivided.

3 How Does the Leakage Current Criterion Work?

Before applying the Leakage Current criterion, it is necessary to know the leakage currents on
the conductor segment. But in order to get the currents, one needs to run the program. Since
running the program twice (once to compute the conductor currents, and another time to re-use
the computed current for the final run) could be rather time consuming, the program carries out
the first iteration using a so-called Quasi-Uniform approach that works the following way:
 During the first iteration, the leakage currents of the conductor segments are computed
using multiple localized uniform soil approximations. The computed leakage currents
are stored in a file with extension “.F43”. This first iteration is very fast since it is based
on the uniform soil approach and does not involve the computations at the observation
points. Then the program restarts (second iteration);
 During the second iteration, the program follows the normal execution using the original
soil model. The above-mentioned computed leakage currents are now used to compute
the Leakage Current priority criterion during the Adaptive Patch Subdivision process to
enhance the patch subdivision.

At the end of the second iteration, the program again saves the computed leakage currents into
the file with extension “.F43” which means that this file can be re-used again for the next
computation to reach an even better computation accuracy if required, without any additional
iteration.

In the new Soil Structure Properties screen, the Leakage Current criterion settings
appear in the Adaptive Subdivision Method panel (tab Settings) when the Leakage
Current criterion is checked in the Priority Criteria list as shown in Figure 1.

Page 10-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 1: Screen settings for Leakage Current criterion.

By default, the Quasi-Uniform option (where the quasi-uniform soil approach is used for the
first iteration) is set in the drop-down box Method. If a valid file with extension “.F43” exists
in the working directory, then the Method can be set to the option Existing that allows using
the pre-computed leakage currents of the conductor segments.
The Leakage Current criterion can be built using either conductor segments leakage currents or
leakage current densities. This option is controlled by the setting in the field Quantity.
By default, the Leakage Current criterion is active when a new case is created. Note also that this
criterion is only available through the new Soil Structure Properties interface and is not
implemented in the legacy soil model interface.

4 Examples
Let us consider some simple examples to demonstrate the use of the Leakage Current criterion.

4.1 Identical Grids Located in Low and High Resistivity Soils

In this example, two identical 20imi×i20im four-mesh grids buried 0.5 m deep are immersed
into 30imi×i30imi×i3im soil volumes each as shown in Figure 2. The case is modeled using the
Copyright © 2017 SES ltd. All rights reserved. Page 10-3
PART I: NEW FEATURES

MALT module of the CDEGS software package. Both grids are modeled in the Main electrode
and are energized with 1000 A.
The model is symmetrical except that the grid on the left side is embedded into a 10iΩ-m soil
volume and the grid on the right-side is embedded into a 1000iΩ-m soil volume. The resistivity
of the surrounding soil is 100iΩ-m. This means that the left-side grid is embedded into a low
resistivity soil and is expected to have a significant leakage current, while the other grid is
embedded into a high resistivity soil and is expected to have a negligible leakage current.

Figure 2: Computer model of grids embedded into low resistivity and high resistivity soil volumes.

The leakage current distribution in the two grids is shown in Figure 3 for the case when the
Leakage Current criterion is not used. The figure confirms that the current leaks out mostly from
the grid on the left side which is immersed into the low resistivity volume. This is an indication
that the Leakage Current criterion could be applied to this system to distribute more patch
resources for the low resistivity soil volume.

Figure 3: The leakage current distribution.

Figure 4 shows the patch distribution over the soil volumes as displayed in the
SESSystemViewer tool for the cases with and without the Leakage Current criterion. Figure 4
shows that when the Leakage Current criterion is used, the high-resistivity volume on the right
side does not receive additional patch resources from the Adaptive Patch subdivision process as
expected and is only subdivided by the initial coarse patch distribution.
The soil potential distribution for both scenarios is shown in Figure 5. It demonstrates that
although the resistive volume was only barely subdivided in the case when the Leakage Current
Page 10-4 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

criterion was used, the soil potential distribution for both scenarios is practically identical. This
shows that significant patch resources are simply wasted by subdividing the high resistivity
volume more finely.

Leakage Current Criterion Used Leakage Current Criterion not Used

Figure 4: The patch distribution on the soil volumes.

Leakage Current Criterion Used Leakage Current Criterion not Used

Figure 5: The soil potential distribution.

4.2 Energized Grid and Floating Rebar

Let us consider another example where a grounding grid of dimensions 50im by 50im, with a
6i×i6 mesh is embedded into a 10iΩ-m 54imi×i107imi×i3im backfill layer as shown in Figure 6.
About 40im away from the grid, there is a concrete pad with rebar which is not bonded to the
energized grid. Both the grid and the rebar are embedded into the same low resistivity backfill
layer. The resistivity of the surrounding soil is 100iΩ-m. The grid is energized with 1000iA. The
case is modeled using the MALZ module of the CDEGS software package.

Copyright © 2017 SES ltd. All rights reserved. Page 10-5

PART I: NEW FEATURES

Figure 6: Grounding system embedded into the backfill layer.

Since the rebar is not bonded to the energized grid, one can expect that there is much less
current leaking from the rebar than from the energized grid. This means that if the Leakage
Current criterion is used, then more patch resources should be concentrated in the grid area and
less in the rebar area.
The leakage current distribution in the system for the case when the Leakage Current criterion is
not used is shown in Figure 7. This figure shows that the leakage current from the rebar is a few
times smaller than that from the energized grid.
Figure 8 and Figure 9 show the patch distribution as displayed by the SESSystemViewer tool
for the case when the Leakage Current criterion is used and not used, respectively. Figure 8
shows that, indeed, there are no adaptive patches generated in the vicinity of the rebar. Only the
initial coarse patch subdivision generated prior to the Adaptive Patch Subdivision process is
present. Figure 9 shows the patch distribution for the case when the Leakage Current criterion is
not used. The Adaptive Patch Subdivision Process in this case generates a very fine patch
distribution on the soil volume faces in the vicinity of the rebar.

Figure 7: The leakage current distribution in the case when the Leakage Current criterion is not used.

Page 10-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 8: The patch distribution pattern over the soil volume interface when the Leakage Current criterion
is used.

Figure 9: The patch distribution pattern over the soil volume interface when the Leakage Current criterion
is not used.

Copyright © 2017 SES ltd. All rights reserved. Page 10-7

PART I: NEW FEATURES

Leakage Current Criterion Used Leakage Current Criterion not Used

Figure 10: The soil potential distribution.

The soil potential distribution for both scenarios is shown in Figure 10. As expected, the results
are practically identical in both cases. Again, this example demonstrates that the patch resources
generated by the Adaptive Subdivision process on the soil volume interface over the rebar area
in the absence of the Leakage Current criterion are literally wasted.

5 Conclusion
In most cases, one may not see any dramatic changes in the output results whether the Leakage
Current criterion is applied or not. However, this new feature helps determine the most effective
patch distribution in accordance with the electric field pattern emanating from the conductor
segments, and could be extremely useful for cases when the soil environment is complex and
includes a number of soil volumes and a large number of conductor segments. In such cases, a
lack of patch resources and their inadequate distribution may easily lead to inaccurate results.
The use of the Leakage Current criterion in such cases could be crucial in order to get a more
adequate patch distribution and more reliable results.

Page 10-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

11 IMPROVEMENTS FOR MODELING ARBITRARY

NETWORKS OF METALLIC PLATES IN HIFREQ

Ali Aghabarati, Moein Nazari, Rouzbeh Moini, Peter Zhao,

Simon Fortin, and Farid P. Dawalibi
Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com
Abstract

This paper presents recent improvements in HIFREQ for analyzing arbitrary networks of metallic plates. In this
new version of HIFREQ, modeling of junctions formed by joining two or more metallic plates and the analysis of
geometries defined by arbitrary intersections of plates are allowed. These features provide considerable modeling
enhancements for studying complex problems containing regions with spatially coincident edges between multiple
plates. In addition, improvements are reported on the development of unstructured surface meshing techniques
that are recently added for the discretization of plate networks into quadrilateral patches. Numerical results
demonstrate the applicability of these new features of HIFREQ for analyzing several networks containing multiple
plates with branching of currents and arbitrary intersections of plates. The results also illustrate how the newly
developed features for the visualization of the computation results for plates can be used for examining currents
and voltages along plates.

1 Introduction
The modeling of metallic plates in the HIFREQ program has been supported in recent versions of
SES Software [1-2]. Problems that were traditionally solved using wire-grid models to emulate
surfaces can now be handled more accurately using the plate model. Since the release of version
15.1.4080 of SES Software, the formulation of metallic plates in HIFREQ has been updated
continuously and new features have been added, allowing the analysis of a wider range of
problems for complex networks [3-6]. The last update of the metallic plates feature in HIFREQ
allowed for complex geometries involving plates, including junctions between plates and
junctions between plates and wire conductors [2]. However, several limitations remained with
respect to junctions of plates. In particular, a maximum of two plates could be connected at a
common edge. Also, contacts between plates were allowed only along a complete edge.
This version of the program eliminates these limitations. Several plates can now be joined at a
common edge, with the current flowing in the plates distributing properly between all connected
plates. In addition, the introduction of a new unstructured discretization method for plates makes
it possible to model partial contacts between plates and other complex geometrical scenarios.
Also, the visualization tools for computation results were updated to display results related to
plates directly on a drawing of the network (in Configuration plots). This article describes these
enhancements in detail.
The remainder of this article is structured as follows. In Section 2, the modeling improvements of
plate geometries with arbitrary configurations are discussed. In Section 3, several numerical
examples are given. The first example validates the use of current branching in HIFREQ for the
computation of the scattered electromagnetic fields around a simple conducting network with
junctions. Next, the ground resistance of a grounding system with a special type of rods made out
of plates is presented. In the last example, an electromagnetic interference (EMI) analysis of a
complex model of trees involving wires in contact with numerous metallic plate junctions is
described. Concluding remarks are given in Section 4.

Copyright © 2017 SES ltd. All rights reserved. Page 11-1

PART I: NEW FEATURES

2 Improvements in Modeling Plate Geometries

This section describes improvements and new features introduced since the last User’s Group
Meeting for modeling plate geometries in HIFREQ.

2.1 Plates with Arbitrary Intersections

In the current version of the HIFREQ program, most developments related to metallic plates were
focused on removing geometrical restrictions related to the specification of plates. Examples of
interconnected plates and wire-plate junctions are illustrated in Fig. 1-(a) and 1-(b). The main
difficulty when dealing with networks including connected plates is that the plates must be
subdivided into patches in such a way that any two patch in the network either share a complete
edge or are completely unconnected. This so-called “conformal edge condition” is required to
allow for a correct description of the current flow between patches.
In previous versions of the program, the subdivision of plates into patches was carried out by
dividing each plate into a regular grid. This approach, which is called structured meshing,
provides a convenient way to specify the number of patches to be used over a plate and has the
advantage of simpler numerical computations and easier interpretation of the results. It was used
in the past few years to treat many problems involving complex arrangement of plates [2-6].
However, structured meshing makes it difficult to treat arbitrary intersection of plates, such as
those shown in Fig. 1-(c) and 1-(d)). Previously, the most usable approach for treating such
problems was to specify intersecting plates so that they were completely overlaying along their
common edge, which could be a difficult task. In the latest version of HIFREQ, a new unstructured
meshing scheme was introduced that eliminates most of the difficulties and restrictions around
plate intersections. The unstructured meshing approach, which is explained in more details in
Section 3, allows modeling arbitrary surfaces with intersections similar to those shown in Fig 1-
(c) and 1-(d) using plates. In this new approach, quadrilateral patches are distributed in an
irregular pattern to accurately represent almost any geometrical interface.

(a) (b)

(c) (d)
Figure 1: Illustrative examples on network of plates with arbitrary intersection at plate edges.

Page 11-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Presently, the structured meshing approach is still used by default for plates in HIFREQ. To
activate the unstructured meshing approach for plates, go to the Advanced Options screen in
SESCAD (Define | Advanced Options) and select the Plate Meshing option (see Fig. 2). Note
that this option affects all plates in the model.
You can use the Edit | Edit Attributes of Plates command (Fig. 2) to control the density of the
patches generated over selected metallic plates using Total number of patches on all plates,
Total number of patches on each plate, or Maximum Patch Area options.

(a)

(b)
Figure 2: Illustration of (a)-invoking the unstructured meshing algorithm for plates using the
Advanced Options screen in HIFREQ and (b)-controlling the patch density using the Edit
Attributes of Plates dialog.

Copyright © 2017 SES ltd. All rights reserved. Page 11-3

PART I: NEW FEATURES

2.2 Plates with Multiple Junctions

Another improved aspect of the present version of HIFREQ is that the program now allows
contacts (or junctions) between multiple plates along a common edge. Fig. 3-(a) shows an
example of such a junction formed by four plates. As shown in Fig. 3-(b), the program accounts
for the branching of current that occurs at junctions between plates, using the procedure outlined
below.
First, the program detects all the junction locations, which are defined as interfaces where three
or more plates are connected at an edge. The branching of currents occurs at the junction location
(highlighted in red), and the appropriate approach for modeling current branching first requires
careful discretization of each plates in the junction locations.

(a)

(b)
Figure 3: The configuration of a plate multi-junction and patch junction system.

Page 11-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

In the next step, the amplitude of unknown current on each plate is characterized in terms of those
in a local junction system, which is the collection of all quadrilateral patches and their associated
current coefficients attached to the detected common edge. Fig. 3-(b) shows a patch junction
system where four quadrilateral patches are connected at a common edge. To each expansion of
current and its continuation to its neighbouring path, a unique unknown is assigned by the
program and a mapping between that patch and the global unknown currents is constructed. It
can be seen that each patch can now contribute to more than one current expansion term.
However, these currents associated with the edges are not independent, as current continuity
(Kirchhoff’s law) is employed to determine the net current flow across the edge from one patch to
another. The HIFREQ formulation computes all the involved integral equation for each unknown
in a systematic way and employs current continuity to relate the coefficients of the currents
flowing across all the edges [7-8].
It should be noted that the required number of unknowns (compared to a case without junction)
increases by the detected number of branches. Fig. 4 shows a side view of a general junction
system, where 𝑁 is the number of patches. For such a junction system, 𝑁 − 1 unknowns are
needed.

Figure. 4: The reference direction for N-1 current branches across a junction system with N Patches.

Crucial to the advancement of the formulation to support plate multi-junctions, it was also
necessary to enhance and upgrade a set of ambiguity check routines. Ambiguity routines in
present version are modified considerably to work based on pairs of adjacent plates and global
arrangements of plates in a network. Such ambiguities procedure have been applied to validate
many problems, some of them are shown in Fig. 5-(a) to Fig. 5-(d).

Copyright © 2017 SES ltd. All rights reserved. Page 11-5

PART I: NEW FEATURES

(a) (b)

(c) (d)

(e) (f)
Figure 5: Examples of generalized network of plates examined by developed ambiguity checker.

Page 11-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3 Technical Details for Unstructured Meshing

This section describes an unstructured quadrilateral surface mesh generation technique newly
implemented for modeling complex structures involving plates in HIFREQ. In general, mesh
generation techniques are grouped into structured and unstructured methods. Both methods have
been implemented in HIFREQ, and some technical details are discussed here.

3.1 Structural Discretization of Plates

Since metallic plates were first introduced in HIFREQ, the program has been using a structured
discretization technique to subdivide the plates into quadrilateral patches. This structured
meshing technique subdivides each plate following a regular grid pattern as shown in Fig. 6.

Figure 6: Structured mesh, 16 patches.

Structured meshes can be generated quickly and reliably, but, on the downside, they are only
applicable to relatively simple, regular geometries. For example, the fact that the number of
patches must be the same along opposite edges of a plate makes it difficult to handle junctions
between multiple plates with arbitrary intersections along the common boundary when using
structured meshing. To overcome this problem, a new unstructured quadrilateral mesh
generation scheme was introduced in HIFREQ. This unstructured meshing technique offers far
greater flexibility for handling complex geometries containing regions with spatially coincident
edges (wholly or partially shared) between multiple plates.

3.2 New Unstructured Discretization of Plates

In this section, the process for building an unstructured surface quadrilateral mesh for a 3D object
consisting of networks of plates is briefly discussed. First, the Quadtree-Delaunay mesh
generation technique is introduced for 2D polygons, and then the proposed technique is applied
to a wide class of 2D examples. Then, the approach for generating the quadrilateral mesh for 3D
objects is shown, and finally some examples are presented to demonstrate the capabilities of the
proposed approach in HIFREQ.

3.3 Unstructured Quadrilateral Surface Mesh Generation – 2D

A two-step process is used to generate an unstructured mesh when discretizing a single plate into
patches in HIFREQ:
1. A triangular mesh is created over the plate.
2. The triangular mesh is converted into a quadrilateral mesh by breaking every triangle into
three quadrilaterals.
The triangular mesh is generated using the so-called Quadtree-Delaunay mesh generation
technique [10]:

Copyright © 2017 SES ltd. All rights reserved. Page 11-7

PART I: NEW FEATURES

1. Nodes are distributed along the boundary of the polygon.

2. A quadtree spatial decomposition [11] is performed to help quickly distinguish the internal
and external regions of the polygon.
3. Nodes are added inside the polygon.
4. A Delaunay triangulation of the interior nodes is performed.
5. The resulting triangular mesh is “smoothed” to make the individual triangles as close as
possible to equilateral triangles [12].
This entire procedure is applicable to general polygons. Fig. 7 shows the result of each step of the
procedure for a typical polygon.

Figure 7: Quadtree-Delaunay meshing steps.

A collection of meshing examples obtained from the developed mesh generation module is shown
in Fig. 8, along with the corresponding quality factor, which is a quantity that represents how
closely triangles (or quadrilaterals) in a mesh resemble equilateral triangles (or squares). Meshes
with high quality factors (close to 100%) typically leads to more stable and accurate solutions of
numerical problems. As shown in Fig. 8, the approach implemented in HIFREQ yields very high
quality meshes.
As can be seen, the developed module can be applied to a wide class of problems including general
polygons, non-convex polygons and polygons with holes inside. However for the current version,
only rectangular, parallelogram and general quadrilateral plates are supported.

Page 11-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

(a) 136 triangles (b) 408 patches (c) 119 triangles (d) 357 patches
Quality = 92% Quality = 70% Quality = 92% Quality = 71%

(e) 570 triangles (f) 1710 patches (g) 72 triangles (h) 216 patches
Quality = 96% Quality = 77% Quality = 88% Quality = 66%

(i) 92 triangles (j) 276 patches (k) 200 triangles (l) 600 patches
Quality = 93% Quality = 70% Quality = 92% Quality = 71%

(m) 240 triangles (n) 720 patches (o) 150 triangles (p) 450 patches
Quality = 97% Quality = 78% Quality = 94% Quality = 74%
Figure 8: 2D meshing collection.
3.4 Unstructured Quadrilateral Surface Mesh Generation – 3D
To apply the above procedure to a HIFREQ model consisting of a collection of plates with multiple
partial or complete contacts along their edges, the program first determines the number of
patches to be generated along the edges of the plates based on the specified number of patches
desired for each plate. The number of patches generated along each edge of the plate is taken to
be proportional to the length of the edge and is defined automatically by the program. Also, all of
the common boundaries including wholly shared and partially shared edges are automatically
identified, and the number of patches along each shared edge is adjusted to be the same for all
plates sharing that edge. This way, it is always guaranteed that all produced meshes are conformal
along common edges.

Copyright © 2017 SES ltd. All rights reserved. Page 11-9

PART I: NEW FEATURES

Note that, as a result of these adjustments, the requested number of patches may not be reached
exactly for all the plates in the model, but it will typically be quite close to the requested value.
A collection of 3D examples is shown in Fig. 9.

(a) 576 patches (b) 294 patches (c) 3060 patches

Quality = 60% Quality = 68% Quality = 66%

(d) 117 plates, 2193 Patches, Quality =61%

(e) 119 plates, 13398 Patches, Quality = 63%

Figure 9: 3D meshing collection.

Page 11-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

4 Improvements in Visualization of Plate Computation Results

In previous versions of the software, the calculation results related to plates (for example, currents
flowing in the plates, potential of the plates, etc.) were not reported in the results visualization
tools (CDEGS – Examine, GrServer, SESsystemViewer). This version introduces complete
support for visualization of results related to plates in CDEGS – Examine, in GrServer, and in the
new SESResultsViewer program. Plates can be displayed in Configuration plots and labelled
with the selected quantity (current, GPR, etc…). The presence of plates can also be taken into
account when producing Computation plots of Conductor-Data or Touch-Voltages. The
results can also be exported in Raw-Data Reports for both Configuration and Computation
quantities.
This section describes the features of the results visualization tools that are related to plates and
highlights the differences in the behavior of the programs when dealing with plates and with
wires. Note that the computation results for plates are stored in a new output file for HIFREQ,
with the extension “X21”. Both the “F21” and the “X21” file for a HIFREQ case must be present in
the working directory in order to be able to produce plots or reports of results related to plates.

4.1 Selection of Quantity

Certain quantities that can be selected for plots or reports of results over wires have no direct
correspondence with quantities for plates. In this version of the software, the quantity to be
plotted or reported for plates is determined automatically by the program, based on the quantity
selected for conductor segments. The correspondence between the quantity selected for conductor
wires and that used for plates is shown in Table 1 for Configuration results and in Table 2 for
Computation Results (Conductor Data).
Table 1: Correspondence between Quantities for Wires and for Plates in Configuration Plots and Reports.
Quantity for Wires Behavior for Plates
Leakage Currents of Draws the plate patches and labels the edges of each patch
Conductor Segments with the current leaking out of the patch.
Draws the plate patches and labels the edges of each patch
Leakage Current Density per
with the current leaking out of the patch divided by the
Unit Length
perimeter of the patch.
Draws the plate patches and labels the edges of each patch
Leakage Current Density per
with the current leaking out of the patch divided by its
Unit Area
surface area.
Longitudinal Currents Draws the plate patches and labels the edges of each patch
Flowing in Origin of with the current entering the patch through that edge.
Conductor
Draws the plate patches and labels the edges of each patch
Longitudinal Currents with the current entering the patch through that edge. Note
Leaving End of Conductor that this is the same as the Longitudinal Currents
Flowing in Origin of Conductor.
Draws the plate patches and labels the edges of each patch
GPR of Conductor Metal with the GPR of the patch (i.e. the potential at the surface of
the patch, at its center).
Draws the plate patches and labels the edges of each patch
GPR of Conductor Coating with the potential at the outer surface of the coating of the
patch, at its center. Note that this is the same as the GPR of

Copyright © 2017 SES ltd. All rights reserved. Page 11-11

PART I: NEW FEATURES

Conductor Metal since the specification of a coating on

plates is currently not supported in HIFREQ.
Draws the plate patches and labels the edges of each patch
GPR of Conductor Metal &
with both the GPR of Conductor Metal and the GPR of
Coating
Conductor Coating.
Draws the plate patches and labels the edges of each patch
with the difference between the GPR of Conductor
Coating Stress Voltages
Metal and the GPR of Conductor Coating. For plates,
this is always zero.
Draws the plate patches and labels the edges of each patch
GPD (Ground Potential
with the potential difference between the GPR of
Difference)
Conductor Metal and the GPR of a reference point.
Draws the plate patches and labels the edges of each patch
Apparent Impedance with the ratio of the GPR of Conductor Metal and to the
Leakage Current of the patch.
Draws the plates and labels their edges with the letter ‘P’
Original Conductor Number
followed by the original plate number of the plate.
Draws the plate patches and labels their edges with the
Conductor Segment Number
letter 'P' followed by the global edge number of the edge.
Conductor Node Number Draws the plate patches, unlabeled.
Draws the plate patches and labels their edges as P1/1, P1/2,
Conductor and Segment etc., where the first number represents the global edge
Numbers number of the edge and the second number represents the
original plate number.
Draws the plates and labels their edges with the plate
Radius
thickness.
Draws the plates and labels their edges with the coating type
Coating Type
code of the plate (currently always 0).
Draws the plates and labels their edges with the conductor
Conductor Type
type code of the plate (currently always 0).
Draws the plates and labels their edges with the cable type
Cable Type
code of the plate (currently always 0).
Draws the plate patches and labels their edges with the
Segment Length
length of the edge.

Table 2: Correspondence between Quantities for Wires and for Plates in Computation Plots and Reports
(Conductor Data).
Quantity for Wires Behavior for Plates
Reports the potential at the center of plate patches, at the
Metal GPR
metallic surface of the plate.
Reports the potential at the center of plate patches, at the
outer surface of its coating. Currently, this is the same as the
Coating GPR
Metal GPR option, since the specification of coating on
plates is not available yet.
In-Flowing Current Not reported for plates.
Out-Flowing Current Not reported for plates.
Leakage Current Reports the total current leaking out of plate patches.

Page 11-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Leakage Current Density per Reports the total current leaking out of plate patches
Unit Length divided by the perimeter of the patch.
Leakage Current Density Reports the total current leaking out of plate patches
per Unit Area divided by the area of the patch.
Coating Stress Voltage Reports the stress voltage across the coating of plate
patches. Currently, this is always zero, since the
specification of coating on plates is not available yet.

4.2 Appearance of Configuration Plots

As described in detail in the previous section, Configuration plots that include plates often label
the selected results on the edges of the plate patches. To avoid ambiguous situations that could
occur when a patch edge is shared by two or more patches, the patches are drawn slightly smaller
than they are in reality; this should be taken into account when interpreting plate results displayed
on Configuration plots.
Fig. 10 shows examples of Configuration plots involving plates in HIFREQ.

GPR OF SEGMENTS (VOLTS)

Maximum Value : 6.403
Minimum Value : 2.477

6.40
6.01
5.62
5.23
4.83
4.44
4.05
3.65
3.26
2.87

GPR of Conductor Metal. Magnitude (V) [ID:Free - High @ f=60.0000 Hz ]

(a) (b)
CONDUCTOR RADII (FT)
Maximum Value : 0.220E-01
Minimum Value : 0.305E-02

0.02200
0.01000
0.00610
0.00305

Conductor Radii (FT) [ID:Free - High @ f=60.0000 Hz ]

Conductor & Conductor Segment Numbers [ID:Free - High @ f=60.0000 Hz ]

2/2 6/6

P15/1 P31/1 P47/1 P63/1 P79/2 P95/2 P111/2 P127/2

P112/2

P110/2
P128/2

P126/2
P16/1

P14/1
P32/1

P30/1
P48/1

P46/1
P64/1

P62/1

P80/2

P78/2
P96/2

P94/2

12/8

P13/1
P11/1 P29/1
P27/1 P45/1
P43/1 P61/1
P59/1 P77/2
P75/2 P93/2
P91/2 P109/2
P107/2 P125/2
P123/2
P108/2

P106/2
P124/2

P122/2
P12/1

P10/1
P28/1

P26/1
P44/1

P42/1
P60/1

P58/1

P76/2

P74/2
P92/2

P90/2

P9/1
P7/1 P25/1
P23/1 P41/1
P39/1 P57/1
P55/1 P73/2
P71/2 P89/2
P87/2 P105/2
P103/2 P121/2
P119/2
3/3

4/4

7/7

P104/2

P102/2
P120/2

P118/2
P24/1

P22/1
P40/1

P38/1
P56/1

P54/1

P72/2

P70/2
P88/2

P86/2
P8/1

P6/1

P5/1
P3/1 P21/1
P19/1 P37/1
P35/1 P53/1
P51/1 P69/2
P67/2 P85/2
P83/2 P101/2
P99/2 P117/2
P115/2
P100/2

P116/2

P114/2
P20/1

P18/1
P36/1

P34/1
P52/1

P50/1

P68/2

P66/2
P84/2

P82/2

P98/2
P4/1

P2/1

8/8

P1/1 P17/1 P33/1 P49/1 P65/2 P81/2 P97/2 P113/2

1/1 5/5

(c) (d)
Figure 10: Display of metallic plate results in Configuration plots. (a) The model, with two plates and
two wire loops. The plates are subdivided into 4 by 4 patches. (b) Configuration plots of the GPR. The
plate patches are drawn smaller than their real size. (c) Configuration plot of the radius. The plates are
drawn full size and labeled with their thickness. (d) Configuration of the conductor and segment
numbers, with the patch edges labeled with their edge number and plate number.

Copyright © 2017 SES ltd. All rights reserved. Page 11-13

PART I: NEW FEATURES

4.3 Configuration Raw-Data Reports

All Raw-Data Reports produced by the program include the contribution of plates. In
Configuration Raw-Data Reports, plates are identified by the negative of their plate number in the
Cond. No column. The Seg. No column lists the global edge number of each patch edge.

Figure 11: Excerpt of Raw-Data Report of Metal GPR under Configuration. Metallic plates are indicated
by a negative value for Cond. No.

4.4 Filter and Reference Conductor Group

By default, all plots and reports produced by the program include the contribution of plates. When
specifying a Filter or an explicit range of conductors to be plotted (under Configuration) or a
Reference Conductor Group (under Computation), plate results are always excluded. To
exclude plates from the results while keeping all results from conductors, you can specify a Filter
on Conductor Type ranging from 0 to a very large number, to make sure to include all
conductors.

5 Numerical Results
Several numerical results are now presented that demonstrate the accuracy and versatility of
HIFREQ for analyzing networks of wires and plates. The first example demonstrates the ability of
HIFREQ to treat a benchmark problem of junction between plates. Then, the use of plates with
multiple junctions is illustrated in an analysis of a grounding system including rods built out of
plates. Finally, an example of electromagnetic interference (EMI) analysis involving a complicated
model of plates in contact with wires is presented.

5.1 T-Shaped Network

Fig. 12-(a) depicts a T-shaped network of plates. The system consists of three rectangular metallic
plates meeting at a common edge and forming a right angle. The length and width of each plate
are 𝜆 and 𝜆/2 respectively. The plate network is illuminated by an incident plane wave at 26
observation angles, and the induced currents on the surface of plates generate the scattered
electromagnetic fields examined around the geometry. A schematic of the problem is depicted in
Fig. 13. By varying the incidence angle, the level of scattered field is computed on the observation
points located on a half-circle and in the same direction of incident field. The radius of the half-
circle profile is set to 10 m. Calculations are given for TE polarization, in which the electric field is
tangential to the surface of plates and the magnetic field is normal to x-z plane.

Page 11-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

(a)

(b)

(c)

Figure 12: A simple PEC junction problem modeled by 3 metallic plates in HIFREQ. (a) T-shape
configuration. (b) Thick T-shaped configuration. (c) Gapped T-shaped configuration.
For comparison purposes and in order to validate the performance of the newly developed code
with previous versions, two more right angled T-shaped configuration are also considered. In the
first, shown in Fig. 12-(b), the thin vertical plate is replaced by three plates, one of which having
a small width of 𝜆/50. This geometry does not involve junctions of multiple plates and could be
handled in previous versions of HIFREQ; the numerical results for this case are expected to be
nearly identical to initial geometry. In the third geometry, shown in Fig. 12-(c), the vertical plate
is disjoint from the horizontal plate by a small gap. The small gap eliminates the branching of
induced current at the three-way junction; the comparison in this case demonstrates the
importance of the junction in this problem.

Copyright © 2017 SES ltd. All rights reserved. Page 11-15

PART I: NEW FEATURES

Figure 13: Configuration of T-shape geometries illuminated by plane wave excitation and observation points
located in the x-z plane.
Fig. 14 shows the computed electric field for the three T-shaped configurations and for all
incidence angles. The computation results show a very good agreement between the thin and thick
T-shaped networks, although a noticeable difference can be observed for some excitation angles
due to contribution from the introduced thickness in the thick geometry. The electromagnetic
field around the T-shaped configuration with a gap presents a significant discrepancy with the
other T-shaped configurations, particularly at 70 degrees and 120 degrees. The latter indicates
that the removal of the current branching across a junction can have a significant effect on the
scattered electromagnetic field values.

Figure 14: Computed electric field at all observation angles for the three tri-plate configurations.

Page 11-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The distribution of induced electric current and metal GPR over the quadrilateral patches of the
three networks are shown in Fig. 15 for an incidence angle of 60 degrees. As observed, while the
distribution of the examined quantities for the thin T-Shaped network and the thick T-shaped
network are similar, the variation of the current and potential for the gapped network is different,
particularly close to the junction where no branching of current occurs for that case.

(a) (d)

(b) (e)

(c) (f)
Figure 15: Visualization of HIFREQ computational results for the quadrilateral patches of metallic
plates in T-Shaped configurations. Longitudinal current for (a) Thin T-Shaped Network. (b) Thick T-
Shaped Network. (c) Gapped T-Shaped Network. GPR of conductor for (d) Thin T-Shaped Network.
(e) Thick T-Shaped Network. (f) Gapped T-Shaped Network.

5.2 Grounding Grids with Plate Rods

The next example is related to modeling new grounding electrodes made by metallic plates for
improved performance of grounding systems. Grounding rods made by metallic plate electrodes

Copyright © 2017 SES ltd. All rights reserved. Page 11-17

PART I: NEW FEATURES

can provide lower resistance value and improved potential distribution for the safety of
substations and protection of working personal.

(a)

(b)

(c) (d)
Figure 16: Grounding systems with different rod electrodes. (a) The main grid without any rod electrode.
(b) The grid with conductor rods. (c) The grid with plate rods. (d) Configuration of the plate rod with 4
metallic plate branches.

Page 11-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Fig. 16-(a) shows a 10 m by 10 m square shape grid made by juxtaposition of five conductors along
x and y coordinates having length 10 m and radius of 0. 9525 cm. The grid is placed at a depth of
0.5 m below the earth surface. In order to lower the grid resistance, one approach is the
replacement of the rectangular grid by a big rectangular shape plate electrode. However, as
investigated in [9], similar characteristics will be obtained with a single plate electrode instead of
the grid of same geometrical parameters. In addition, the production cost of conductor electrode
is less than plate ground electrode.
Interestingly, grounding rods can be used to lower the resistance of the main grid. Generally,
characteristics of ground rods are important for better dissipation of fault current in surrounding
soil and improvement in operation and protection of working environment from high voltage
hazard during current fault. Conductor rods are most often used as simple electrodes for
construction of grounding system. Fig. 16-(b) shows the configuration of a HIFREQ model in
which 16 conventional rods with the length of 2.2 m are used around the perimeter of the main
grid in order to reinforce the grounding system. In a third HIFREQ model, plate rods shown in
Fig 16-(c) are used to create an improved grounding system. Plate grounding electrode are made
from juxtaposition of four 0.5 m× 1.2 m copper plate with thickness of 1 mm, and are modeled
with the same depth as of conventional conductor electrodes (see Fig. 16-(d)).

Table 3: Comparison of Grounding Systems without Rod. with Conductor Rod and with Plate Rods.

Grounding Without Grounding With Conductor With Plate

System Rod Electrode Grounding Electrode Grounding Electrode
Soil Model Uniform Two Layer Uniform Two Layer Uniform Two Layer

Grid Resistance (Ω) 4.270 12.02 3.70 11.54 2.99 10.45

Maximum Touch
3.374 7.85 2.805 7.37 2.115 6.43
Voltage (V)
Maximum Step
0.793 1.46 0.642 1.40 0.428 1.17
Voltage (V)

The resistance value of the grounding systems shown in Fig. 16, apart from their configuration,
also depends on soil resistivity which value need to be specified in order to provide accurate
comparison. In this experiment, two soil models are considered: a) a uniform soil with resistivity
of 100 Ω − m. b) A two layer soil model with 100 Ω − m soil and thickness of 1.2 m over a 500 Ω −
m layer.
Table 3 compares the characteristics of all the grounding systems shown in Fig. 16 for two
different soil models. As it is expected, the resistance of the grid with conductor rods is lower than
system without grounding rod electrode. However, by the use of plate ground electrode having
the same dimension and buried at the same depth as a conductor electrode, the resistance of
system is reduced more in each case. For the uniform soil model, the maximum value of step and
touch voltages for third system is about 37% and 46% lower comparing to original design, which
provides additional safety margins in standard engineering design of grounding system.
The potential distributions (step and touch voltages) in the vicinity of all of the grounding systems
are shown in Fig. 17 and Fig. 18. As observed, potential distribution is smoother for the system
with plate rods, as well as maximum voltages has lower value. It is also interesting to observe the

Copyright © 2017 SES ltd. All rights reserved. Page 11-19

PART I: NEW FEATURES

increased zone of safety, which can ultimately leads to enhancement in the design and cost
reduction of new grounding systems that constructed by plate rod electrodes.

(a)

(b)

(c)

Figure 17: Comparison of touch voltages for grounding systems with different rods.
(a) A grid without grounding rods. (b) A grid with conductor rods. (c) A grid with plate rods.

Page 11-20 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

(a)

(b)

(c)

Figure 18: Comparison of step voltages for grounding systems with different rods.
(a) A grid without grounding rods. (b) A grid with conductor rods. (c) A grid with plate rods.

Copyright © 2017 SES ltd. All rights reserved. Page 11-21

PART I: NEW FEATURES

5.3 Modeling of Trees under High Voltage Power Lines

Trees and foliage in close proximity to electrical equipment and power lines cause electrical safety
hazards, power interruptions and potential fires due to electrical arcing. A study of trees and
foliage can be carried out in HIFREQ using a combination of metallic plates and wires as a
simplified model. At power frequencies, since the radii of the trunk and of the branches of the tree
is much smaller than the wavelength, they can all be modeled using wires without any significant
loss of accuracy. Moreover, any arbitrary combination of metallic plates, e.g., simple quadrilateral
plates, with multiple junctions and plates intersecting along the boundaries can be used to
represent the leaves of the tree. Fig. 19 shows a wire-plate tree model that has been used in this
example.
The study model includes two large trees located in the vicinity of a three-phase high voltage
transmission line. The behavior of the electromagnetic fields due to the presence of the trees is
investigated under steady state condition. For this purpose, a three-phase 525 kV transmission
line carrying 430 A is placed at a height of 20 m above the ground. A total length of 1.6 km of the
line is included in the model. The transmission line conductors are in a horizontal configuration,
without shield wires.
The trees are 18 m high and located at distance of 20 m from the center of the transmission line.
The trees are modeled with 566 conductors as trunks and branches, and 290 metallic plates as
leaves. As shown in Fig. 19, the unstructured mesh generation method is used to discretize all the
metallic plates into 3174 quadrilateral patches.

Figure 19: Modeling of trees under transmission lines.

The magnitude of the total electric field across three surface profiles due to the presence of the
trees is shown in Fig. 20 and is compared with the case where there are no trees. Two surface
profiles are located parallel to the soil interface, on the bottom and top of the trees at heights of
1.5 m and 19.5 m, respectively. The third profile is placed behind the trees at a distance of 20 m
away from the center of the transmission line, perpendicular to two other surface profiles. It can
be seen that the presence of the trees can significantly disturb the distribution of the electric field
around the transmission line. As expected, the strength of the electric field is notably reduced

Page 11-22 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

below and behind the trees providing a 30 dB electric shielding at ground level. On the other hand
as shown in Fig. 20(b), the electric field is strongly increased between the power lines and the
trees, and therefore flashovers from the bare conductors to the trees may occur if the distance
between them gets smaller.

(a)

(b)

(c)
Figure 20: The distribution of the total electric field under the transmission line with and without the
trees. (a) Below the trees on the soil surface. (b) Between trees and power lines. (c) Behind the trees.

Copyright © 2017 SES ltd. All rights reserved. Page 11-23

PART I: NEW FEATURES

6 Conclusion and Future Developments

This paper presented recent enhancements in the HIFREQ program for analyzing various
problems involving networks of metallic plates with arbitrary junctions and intersections between
plates. Numerical results showed that by using unstructured discretization of metallic surfaces
and employing current branching across plate multi-junctions, a wider range of problems can be
modeled in HIFREQ. The computation results also demonstrate the applicability of the latest
developments for visualization of the currents and potentials of plates in Configuration plots
of CDEGS – Examine.
The next year will see further developments of the plate feature in HIFREQ, notably acceleration
of the computation engine for large problems.

7 References
[1] A. Aghabarati, R. Moini, S. Fortin, P. Zhao and F. P. Dawalibi, “Modeling of Plate Structures In
HIFREQ”, UGM 2015 –SAN DIEGO, CALIFORNIA.
[2] A. Aghabarati, R. Moini, P. Zhao, S. Fortin, S. Ladan and F. P. Dawalibi, “Modeling Network of
Metallic Plates and Cylindrical Conductors In HIFREQ”, UGM 2016 –BOULDER,COLORADO.
[3] S. Ladan, A. Aghabarati, R. Moini, S. Fortin, Y. Li and F. P. Dawalibi "Induced Transient Currents on
Stationary Vehicles under Power Lines Hit by Lightning", Asia-Pacific Power and Energy Engineering
Conference 2016 (IEEE PES APPEEC 2016), Xian , China, October 25 to 28, 2016.
[4] S. Ladan, A. Aghabarati, R. Moini, S. Fortin, F. Dawalibi "Electromagnetic Shielding Analysis of
Buildings Under Power Lines Hit by Lightning", 33rd International Conference on Lightning
Protection, Estoril, Portugal, September 25 to 30, 2016.
[5] S. Ladan, A. Aghabarati, R. Moini, S. Fortin and F. P. Dawalibi "Induced Disturbances by High
Voltage Transmission Lines on Nearby Stationary Vehicles", International Symposium on
Electromagnetic Compatibility, Ottawa, Canada, July 25 to 29, 2016.
[6] A. Aghabarati, R. Moini, S. Ladan, S. Fortin and F. P. Dawalibi "Electromagnetic Shielding Properties
of Spherical Polyhedral Structures Generated by Conducting Wires and Metallic Surfaces", IEEE
International Symposium on Electromagnetic Compatibility, Ottawa, Canada, July 25 to 29, 2016.
[7] C.A Balanis, “Antenna theory: analysis and design”. John Wiley & Sons. 2016.
[8] Gibson, Walton C, "The method of moments in electromagnetics". CRC press, 2014.
[9] Jovanović, D., Cvetković, N., Raičević, N. and Hederić, Ž., 2016, October. Comparative analysis of
plate and grid ground electrode characteristics as a part of grounding system. In Smart Systems and
Technologies (SST), International Conference on (pp. 35-40). IEEE.
[10] D. Engwirda, "Locally-optimal Delaunay-refinement and optimisation-based mesh generation",
Ph.D. Thesis, School of Mathematics and Statistics, The University of Sydney, September 2014,
(http://hdl.handle.net/2123/13148).
[11] Shaffer, CA, Samet, H., “Optimal Quadtree construction algorithms”, Computer Vision Graphics and
Image Processing, 37, Issue 3, 402–419,1987.
[12] Freitag, L. A., “On combining Laplacian and optimization-based mesh smoothing techniques”,
ASME Applied Mechanics Division -Publications- AMD 220 (1997), 37-44.

Page 11-24 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

12 IMPROVING THE USERS’ EXPERIENCE WITH SES

SOFTWARE DOCUMENTATION

Gaëlle Meslin, Stéphane Baron, Éric Larivière, Peter Zhao and Jie Liu
Safe Engineering Services and Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

With the shift from VB6, VB.Net to Windows Presentation Foundation (WPF) applications, SES is modernizing
most of its products while striving for excellence. The importance of providing documentation that is clear, efficient
and appealing is crucial. A new documentation process has emerged following recent task force committee review
meetings that brought a significant number of novelties and enhancements. This is the subject of this article which
presents how these various changes impact the users’ experience with SES Software.

1 Introduction
The users’ experience with software documentation matters as much as software features and
functionalities. Indeed a positive customer experience determines the overall success of the
software itself. This is the reason why a lot of leading application development providers have
drastically modified their software development approach in order to have the software perform
efficiently, effectively and enjoyably. Software documentation is one of the key elements to a great
users’ experience. For SES Software to be the best product on the market, it is absolutely essential
to let documentation professionals be part of a product team throughout the entire software
development process.
The standardization of software documentation is a significant and necessary step that is not
always straightforward. However, having one clearly-defined process for the creation of any
documentation as well as its translation to other languages positively impacts the quality of the
documentation itself. Coherence, whether it affects the design or content of a product, is the key
to a harmonious design. It directly improves the quality of the documentation, which is our main
objective. By upgrading the quality of the documentation, we enhance the product we sell and the
service we offer. Consequently, the question of how SES can improve the users’ experience with
SES Software is a crucial one. It is not just about perfecting the documentation, but also about
finding the elements that need to be enhanced from a user’s point of view. The integration of new
tools allowed SES to make important changes regarding the design and content of our
documentation. It also accelerated the translation process in order to produce more quality
content and serve our customers better. All of this finally shows a sprightly approach to software
documentation, as we find the best ways to create high-quality products.

2 A Quality-Oriented Integration of New Means

Documentation is an important part of scientific software engineering. Two aspects have recently
been studied in order to standardize SES software documentation (interfaces, Online help, How
To manuals, tutorials, etc.). Indeed, both the form (look, color, use of interfaces), and also the
content (terminology and style) have been improved, resulting in a coherent and friendly software
interface and documentation.

Copyright © 2017 SES ltd. All rights reserved. Page 12-1

PART I: NEW FEATURES

2.1 Design
Look and functionality are key elements that need to be taken into consideration when thinking
about documentation design. Indeed, the concept of design lays between those two ideas and
works only if the duality is respected. This is what we have decided to take into account at SES for
both our software and documentation. To do so, it was necessary to introduce new tools that will
help in the design process. The company started to use what is called “Authoring tools”, which are
programs that help linking objects together to produce attractive and useful documentation. As a
leader in technology, SES has chosen a state-of-the-art and leading technical authoring and
publishing software package.
The main idea is that all our documentation now has the same look, and serves the same purpose.
For example, the first page of all the online help documents is identical and provides a quick
summary of what the software does as shown in Figure 1. The first page also has space devoted to
providing details about SES and the software, and access to “How To” manuals related to the
software. The Table of Content contains the same sections so that you are never lost when
switching from one application to another.

Figure 1: The Home page of all the SES Online help documents is now identical.
We strongly believe that using new technologies minimizes the risk of errors that can be
introduced by humans, hence the importance of knowing those tools. Indeed, the concept of single
source publishing1 reduces the potential for errors, allowing us to focus on a template to manage
our documentation. This has an impact on the quantity and quality of the documentation as we
will see later on, but also gives us more time to spend on the documentation appearance.

1From Wikipedia: Single source publishing, also known as single sourcing, is a content management method which allows the same
source content to be used across different forms of media and more than one time.

Page 12-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.2 Content
New means have also been put in place to help manage documentation content, especially when
it comes to translation. We have introduced Computer-Assisted Translation (CAT) tools which
work with Translation Memories2 (TM) and Termbases3 (TB). The purpose is to create a more
powerful and consistent documentation. Thanks to TM and TB databases which are growing large
on a daily basis, we are able to build up a strong terminology and improve our documentation.
We are aware that by making it efficient and appealing, you will find more satisfaction in your
experience of the software as we try to make it as easy as possible. This understanding also comes
from the collaboration between engineers and linguists, which we think is a true asset.
The CAT tools used at SES enable professionals to share their knowledge, expertise and
competencies to provide content that is technically and linguistically satisfying.

Figure 2: The Termbase helps to create a more powerful and consistent documentation issued from the
collaboration between engineers and linguists.

2.3 Summary
The use of new leading technologies is a very important element at SES. We believe in good-
quality software products which enable us to manage design and content more efficiently. These

2From Wikipedia: A Translation Memory is a database that stores "segments", which can be sentences, paragraphs or sentence-like
units (headings, titles or elements in a list) that have previously been translated, in order to aid human translators.
3From Wikipedia: A Termbase, or Term Base (a contraction of terminology and database), is a database consisting of concept-oriented
terminological entries and related information, usually in multilingual format.

Copyright © 2017 SES ltd. All rights reserved. Page 12-3

PART I: NEW FEATURES

elements both convey an image; they are essential and naturally linked, and are needed to ensure
coherence.
These new means put in place for the creation and translation of our documentation are here to
reinforce the company’s image and provide assistance in the use of the software package.

3 An Acceleration of Information Delivery

The rapidity that comes from using these new tools constitutes a key element that considerably
improves the users’ experience with the software. It enables SES to deliver more pertinent content
that is complementary to our products. Your opinion and feature requests are also quickly
reflected in our software through the accelerated documentation procedure. Standardized
product-related documents are available, making it easier to find appropriate help to get familiar
with our software. The documentation also becomes available in different languages, as the
international dimension is taken into consideration in order to meet the increasing demand for
multilingual content.

3.1 Automation
When we talk about a new process, we mean “automation”. Indeed, we got rid of the manual steps
implied in the creation of our documentation and in the translation process with the objective of
producing quality documentation efficiently.
With the implementation of authoring and CAT tools, the translation projects are always
accessible and contain the latest version of our software. Our experts can now work on the content
whenever they want, bringing more freedom and consequently more up-to-date results.
Software development is in constant evolution: we need to refine the design, add details, provide
new messages, introduce new functionalities; and it is essential that the documentation follows
this constant evolution. Consequently, documentation is part of the entire software development
lifecycle to create a greater experience for customers. A single example is for the translation of
interfaces; the translation projects are linked to the language files and updated on a daily basis.
The English interface is reviewed and approved followed by the translation in Chinese, French
and Spanish, and eventually to other languages as required. This ongoing process allows us to
discuss, improve and bring suggestions to our software and documentation as often as we need.
Figure 3 shows the full cycle for producing a software package in all the languages supported at
SES. This procedure also applies to all our types of documents. As a result, when our software
evolves based on your suggestions, it is directly reflected through this cycle and taken into
consideration during the translation process.

Page 12-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 3: Full cycle for producing a software package in all the languages supported at SES.

3.2 Plurality of Languages

The languages currently supported at SES are English, Chinese, French and Spanish. The
translation of the software is done in accordance with the notions of multiculturalism and cross-
disciplinary approach. SES makes a point of respecting international standards and other
countries’ norms. As a world leader in grounding (earthing), lightning and electromagnetic
interference, it is evident that we understand the importance of localization, as “the process of
translating a product into different languages or adapting a product for a specific country or
region”4. That is why we believe in the plurality of competences to answer this international
demand. Both engineers and translators work together and share their linguistic and technical
knowledge to produce high-quality products. We always take into account our audience and adapt
the documentation accordingly. We recognize the importance of localization which is sometimes
a legal criteria, as well as a cultural one. The fact that all the different languages are supported is
a real plus and also shows the international position that SES wants to maintain.

3.3 Summary
SES has a worldwide reputation built upon years of service to the engineering community. Being
a multicultural company, we also have this capability to adapt to various audiences and to support
our clients in the best possible way. Our documentation reflects this philosophy and that is the
reason why we developed a process to bring not only quality and quantity, but also to facilitate the
use of our products. Thanks to the new procedure, we can easily and quickly maintain as well as
upgrade our software interfaces and documentation.

4
From Wikipedia: Language localization is the process of adapting a product that has been previously translated into multiple languages to a
specific country or region (from Latin locus (place) and the English term locale, "a place where something happens or is set").

Copyright © 2017 SES ltd. All rights reserved. Page 12-5

PART I: NEW FEATURES

4 A Sprightly Approach to Documentation

SES’ priority is to produce a software that corresponds to our customers’ needs. All the
improvements that are brought to the software and its documentation serve this purpose: satisfy
our clients. The quality of our products is not only reflected in the computation power of the
software modules and packages, but also through the software documentation. As a matter of fact,
by rethinking the mechanism used to create content, SES is able to profoundly innovate its
documentation. This approach is used in every aspect of software development, and allows a real
change based upon a well-thought procedure.

4.1 Modern Approach

As mentioned above, there is a change occurring between our old interfaces (VB6, VB.Net) and
the newer state-of-the art WPF applications. The documentation is also here to facilitate this
transition and give you all the elements you need to get used to the new interfaces. By having an
appealing and complete documentation, you will be able to get familiar with the different products
and find answers to your questions in little time. The fact that the interfaces also look similar will
be an asset when you change from one application to another. As illustrated in Figure 4, by taking
advantage of the authoring tools single sourcing publishing, SES now delivers several types of
document, in different formats. These various formats allow you to quickly get started with a
software by using the Quick Start guide, learn how to use a software by following a simple case
study provided by the Mini-How To manual, and get details about a software with the Online help
document.

(a) Quick Start guide (b) Mini-How To (c) Online help

manual

Figure 4: SES now delivers several types of document, in different standard formats.

4.2 Responsiveness
This leads us to a very important criteria: responsiveness. While we can guarantee you will benefit
from a support service that is flawless, it is also an advantage to find all the information you need
without any assistance. We offer various types of documents such as Online help, How To
manuals, FAQ, video tutorials, etc. Whether you are looking for something more interactive or
simply a traditional Acrobat file, you will find the solution adapted to your needs and your project
within our documentation. Going from one application to another becomes easier as you have the
tools necessary to your success. The goal is to use all the media at our disposal to facilitate

Page 12-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

communication and provide a complementary help to our support service. We understand every
customer is different and we create our documentation based on this uniqueness.

4.3 Summary
Our priority is the continuous improvement of our users’ experience with our software packages.
With this objective in mind, SES has been able to offer new resources and to rethink the entire
documentation process to find the best solutions. This sprightly approach is a requirement for a
pioneer and leader in grounding (earthing), lightning and electromagnetic interference such as
SES.

5 Conclusion
With the shift from VB6, VB.Net to Windows Presentation Foundation (WPF) applications, SES
is modernizing most of its products while striving for excellence. The importance of providing
documentation that is clear, efficient and appealing is crucial. A new documentation process has
emerged following recent task force committee review meetings that brought a significant number
of documentation novelties and enhancements. We hope that you will enjoy our new software
documentation look and content.
In order to make our documentation more accessible, more efficient with new services such as
video tutorials, and more powerful by using the most up-to-date technology, the next step will
consist in offering you our documentation directly within your internet browser and this whatever
the device you use; i.e., tablet, smart phone or desktop.

Copyright © 2017 SES ltd. All rights reserved. Page 12-7

Part II: User Contributions
 UGM 2017 – CANNON BEACH, OREGON

13 PRACTICAL ADVICE FOR ELECTRICAL SOIL

RESISTIVITY TESTING VIA THE WENNER METHOD
Mr. Swen Sorvala, M.S., P.E.

Black & Veatch

Email: [email protected], Web Site: www.bv.com

Abstract
This article discusses some practical considerations for electrical soil resistivity testing for the purpose of
grounding/earthing design of substations and electric power generation facilities, including renewable power
projects.

1 Introduction
Each Grounding Safety Analysis requires four common inputs (electrical soil resistivity, short
circuit/fault current data, fault clearing duration, and ground grid geometry) and one of the most
important tasks is to understand the electrical resistivity of the dirt, soil, and/or rock that a
grounding system will be both embedded within and also surrounded by.
It is important to note that the performance of a ground grid is not only influenced by the soil
within a few feet of the ground conductors and ground rods, but many feet deep and many feet
wide/laterally as well.
Obtaining quality electrical soil resistivity data is possible but requires some forethought and
attention to the procedures and methods used.
Electrical Soil Resistivity Testing is not the same as Thermal Soil Resistivity Testing! Thermal
Soil Resistivity is measured for cable ampacity calculations and when both are labeled as “soil
resistivity”, project teams and contractors may confuse the two. While there is a correlation
between the two, there is no direct conversion.
This paper delves into a practical oversight for one that may be considering performing electrical
soil resistivity testing or specifying it for others to be completed.

2 Overview of Electrical Soil Resistivity Testing

2.1 How the Test is Conducted

By far the most common testing methods for electrical soil resistivity testing are the Wenner four-
electrode method and the Schlumberger four-electrode method. There are some newer testing
methods, e.g. – Unipolar Wenner [3], and there are some less common testing methods which
yield results that are useful that a grounding analysis could use, e.g. – cone penetration or soil box
testing.

Copyright © 2017 SES ltd. All rights reserved. Page 13-1

PART II: USER CONTRIBUTIONS

Figure 1: Wenner four-electrode method.

The basics of the Wenner and Schlumberger test methods are that there are 4 electrodes,
essentially metallic rods, inserted along a straight imaginary line. Actually, it is not so imaginary
as there is typically two non-conductive tape measures radiating outward from the center point
of each test location. The tape measures are used to provide accurate electrode placement. Each
test method has its unique relationship between the electrodes.
Both methods call for the outer two electrodes to provide the electrical current source and the
inner two electrodes are to measure the developed potential difference created by the current
passing through the soil. This potential difference is a response, or signature, of the current input
based on the underlying soil’s material, layering of the different soil types, moisture content,
mineral content, temperature, and possibly any nearby very conductive or non-conductive
objects.
A typical response at the surface of the soil due to the current in and current out of the soil results
similar magnitudes, but opposite phase as shown in Figure 2.

Figure 2: Sketch of surface potentials during a typical Wenner four-electrode method.

In the Wenner method, all the electrodes are equidistant apart from each other. Keeping the
center point the same, each electrode is successively moved outwards for each “a” spacing.
Usually a series of logarithmic “a” spacings are completed at each test location, although a linear
distribution of “a” spacings is occasionally seen as well. The author’s recommendation of “a”
spacing would be: 1’, 2’, 3’, 6’, 10’, 20’, 30’, 60’, 100’, 200’, & 300’. Depending upon the size of the
grounding system which will be designed, additional larger “a” spacings may be required to
develop an accurate model of the soil. The Schlumberger method keeps one pair of electrodes
constant (outer current probes) and moves the (inner potential probes) at equal distance from the
center.
The depth of the probes is also important. Per the theory, the depth of the probes should
essentially be a point source, but since this is real life, whenever possible the probe should be
driven to equal or less than 1/20 of the “a” spacing. When this is accomplished, the equation
Page 13-2 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

principally changes from (1) to (2) shown below as “b”, the probe depth, becomes negligible when
compared to the “a” distance and thus can be ignored.
4𝜋𝑅𝑎
𝜌=
2𝑎 𝑎 (1)
1+ 2 2
− 2
√𝑎 + 4𝑏 √𝑎 + 𝑏 2

𝜌 = 2𝜋𝑅𝑎, (2)

where:
𝜌 ∶ 𝐴𝑝𝑝𝑎𝑟𝑒𝑛𝑡 𝑅𝑒𝑠𝑖𝑠𝑡𝑖𝑣𝑖𝑡𝑦 (𝑜ℎ𝑚 − 𝑚),
𝑅 ∶ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 (𝑜ℎ𝑚𝑠),
𝑎 ∶ 𝑃𝑟𝑜𝑏𝑒 𝑠𝑝𝑎𝑐𝑖𝑛𝑔 (𝑚𝑒𝑡𝑒𝑟),
𝑏 ∶ 𝑃𝑟𝑜𝑏𝑒 𝑑𝑒𝑝𝑡ℎ (𝑚𝑒𝑡𝑒𝑟) [2].

2.2 Soil Resistivity Testing Layout

One of the first tasks is determining where the tests locations should be placed, number of test
locations, and the “a” spacings that should be used per each test location. Unless there is prior
knowledge of the soil geology; the larger a site layout, more test locations should be specified. It
is commonly seen that the surface soil material and thickness of near-surface layers varies
significantly more than the layers and their thicknesses of deeper sub-surface layers (There are
always exceptions to this with geological fault lines and other geological anomalies). For this
reason, it is advisable that more shallow tests than deep tests be conducted across a site. It’s the
author’s recommendation that a minimum of 3 shallow tests and 2 deep tests be conducted at
each site. “Shallow” is loosely associated with measurements from 1 foot to 30 feet and “deep”
from 1 foot to measurements greater than 30 feet.
It was a common mistake in the past to start the first “a” spacing measurements at some larger
distance, e.g. – 5 feet or 10 feet, which is greater than the depth of most ground grid installation
depths. This may have been due to the adoption of corrosion or cathodic protection engineer
practices where the very shallow testing results were not commonly taken or needed. Very small
“a” spacings (“a” = 0.5, 1, 1.5 feet) should be taken for each traverse for two reasons. The first is
that it is important to understand the soil resistivity of the soil above the installed grounding
conductor, or that between a person’s feet and the grounding conductor. One caveat to this is that
the very small “a” spacings may not be as important if the soil was brought in as part of site grading
and the top soil is all uniform resistivity. The second reason is that the very small “a” spacings
provide the basis of, or foundation for, interpreting the significance of the larger/deeper soil
readings.
The largest, or deepest, soil resistivity traverse should be approximately that of the diagonal of the
extent of the grounding system. For example, if the site is 200 feet by 200 feet, the diagonal is
282 feet. This is approximately the largest “a” spacing that should be obtained. However, the
percent error of the predicted performance of the grounding system increases with shorter
maximum “a” spacings that are conducted.
Soil resistivity test locations are not required to be centered at future equipment locations as is
typically seen in soil boring locations, but should be spread out across the entire area and if
Copyright © 2017 SES ltd. All rights reserved. Page 13-3
PART II: USER CONTRIBUTIONS

possible, on all sides of the site. The soil model which is developed for the grounding analysis
needs to represent not only the soil in which the grounding system is installed, but also the nearby
soil around the grounding system through which the fault current will travel through as it leaves
the grounding system and the site itself.
To validate each testing profile, it is a common practice to conduct an equal, but perpendicular
test to the first profile. In horizontal layered soil, these tests should be measuring the same
volume of soil and the results should be fairly similar between each other. Significant variations
between the two perpendicular tests may indicate interference from buried conductive materials,
overhead transmission lines, fences (electrified or otherwise), et cetera. In such cases it may be
necessary, depending on the extent of interference, to disregard the test results obtained at that
location. Since resistivity testing is taken at several locations around a project site, the loss of
testing results at one location typically does not result in insufficient data to prepare a soil model.
2.2.1 Greenfield Sites
Greenfield sites refer to areas that have no existing facilities or infrastructure that is being added
onto and thus are often easier to locate suitable locations.
It is good practice to evaluate the existing gradation, elevation changes, across a site and compare
it against the site elevation of the final grade. If possible, it is preferable to test after final
gradation, but we often do not have the luxury of time on our side. If testing is to be completed
before significant earthwork, then it is recommended to conduct a couple test locations after the
earthwork has been completed and has had time to settle or stabilize. This data can then be used
to update the grounding analysis study and make adjustments as needed.
The perpendicular test pairs should be spread across the property and in random orientation to
ensure any soil variations are captured, that usable data is obtained if some test locations are
questionable, and if the final substation footprint moves to a different location within the same
property, retesting would not be necessary.
Sites occasionally have anomalies in one area or another and the more tests that are taken, it’s
less likely that the complete data set would be overly skewed by readings in that one area.
2.2.2 Brownfield Sites
Brownfield sites are those that have an existing substation or generation facility that would be
expanded or re-evaluated. It can be difficult to find suitable locations and usually more difficult
to find locations that accommodate the perpendicular traverse: the perimeter fence is close to the
property line, undeveloped or open areas are limited, metallic fence lines are present, parallel
streets with suspected or known underground utilities, et cetera are some of the more familiar
obstacles encountered at these sites.
The greatest concern are the metallic paths that could alter the test current, e.g. – copper ground
conductors, buried metallic pipe, storm drains, conduit/cable raceway, and even underground
structures such as concrete foundations. It’s best to stay clear of all such paths, and when testing
near them, take perpendicular tests.
Occasionally, it is possible to test within the larger ground grid mesh squares or rectangles. While
testing within the site facility is desirable, the presence of the grounding grid mesh may prevent
meaningful testing or employing a very limited number of “a” spacings, e.g. – 1’, 2’, 3’, 4’, 5’, & 6’,
depending on the size of the grid mesh. Similar to greenfield site and if possible, it is important

Page 13-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

to test on all sides of a substation. Occasionally, one of the traverses will be useless because of the
presence of underground utilities or other conductive paths. These tests will provide insight, or
an average, to what the soil resistivity within the site.
One tool that continues to increase in worth is satellite imaging software, such as Google Earth
and Street View. Attempting to specify locations on an engineered CAD drawing does not capture
the entire picture of current site conditions. It is valuable to see the conditions as they are now;
perhaps the site is still a wetland, fully wooded, a building & parking lot still needs to demolished,
a portion of the “open” land is a steep incline, overall terrain, et cetera. Another perspective these
tools offer is the historical feature which provides satellite images from previous years which can
lend insight into what may be still in the ground that our testing should avoid. In Figure 3, a deep
soil traverse was specified on the left-hand side and only after the traverse was compared to past
imagery was it discovered that this testing location would need to be moved because of the
presence of wind turbine collector cable routing.

Figure 3: Proposed testing layout (left) and review of proposed against historical images (right).

What if I already have soil resistivity at the next substation? Unless that substation is adjacent to
the site in question, it is advisable to perform electrical soil resistivity at every site. Soil conditions
change even in small distances, even when the surface appears very level and flat. Rock ledges
and fissures can cause some quite different results; soil models. The date and quality of the
previous soil resistivity should also be examined. Many of the past soil testing was performed
with insufficient meters or less useful data.
2.2.3 Obstacles
When in the field doing the testing, there are a variety of unforeseen obstacles or challenges to
overcome. Many of them are physical obstacles or below-grade objects and many are political.
More often than not, the deep soil resistivity traverse is unable to be completed adjacent to the
site or even on the plot of land in which the utility owns. The deep soil testing can be completed
in a neighboring farm field, park, or vacant commercial lands. Each of these land access will
require permission to be granted from the land owner. Working from an office far from the actual
plot of land can be difficult in determining whom to contact to gain this permission. Ensure to
include a few weeks or a month to dig into determining the landowner to contact, discussing the
test, and providing documentation to support you to have the utility authorization and in some
cases, insurance to cover damage to crops or incidental damage.

Copyright © 2017 SES ltd. All rights reserved. Page 13-5

PART II: USER CONTRIBUTIONS

Depending on the terrain and geology, there are times when the deep test may be measured up to
a mile from the site! Those areas that have uniform geology or flat terrain, this can be reasonable.
Those sites that are in more hilly country, the deep data should not be taken on great elevation
differences or on other side of a ridge, but rather up or down the direction of the valley.
There are situations where a linear traverse is unable to have a full perpendicular traverse. It is
completely acceptable to shorten one of the perpendicular test lengths or if necessary, not test the
perpendicular line at all.
Avoid testing over and parallel to below-grade objects such as foundations, conduit, ground
conductor, sewer mains, storm drain pipes, other utilities, et cetera. Try to keep parallel testing a
separation distance of at least 1.5 x the traverse’s maximum “a” spacing.
Additionally, avoid testing parallel to above grade objects such as metallic fences, concrete
sidewalks, grounded distribution or transmission lines, et cetera. Try to keep parallel testing a
separation distance of at least 1.5 x the traverse’s maximum “a” spacing.
Testing should not be completed in rainy or storm conditions or immediately after a rain storm.
Water not only mixes with electricity all too well which increases the risk of personnel safety, it
also creates a very thin plane of saturated soil which can render many readings to invalid or “too
good to be true”. It is recommended to delay testing until the soil has had time to drain or absorb
most of the rain water. If there are puddles of water everywhere, give it a day or two to drain or
evaporate.
Those that have or will test in the more northern states and provinces (in the northern
hemisphere), snow is often present during the winter months. While snow is cold and once soil
becomes frozen, it resistivity significantly increases below 32 F (0 °C), however snow is not an
immediate indicator that no soil resistivity testing can occur. The snow acts as an insulator for
the soil and inhibits the frost from becoming thicker. Testing can occur after clearing away the
snow from the soil at the probe locations, determining the depth of the frost and then making sure
the probes are penetrating into the non-frozen soil. However, if the frost depth is 6 inches or
greater, it may prove to be too time consuming.
Low to moderate on-site vegetation is usually not a concern for the electrical soil resistivity results,
but it definitely can slow the process down because of safety concerns as well as difficulty of
maneuvering the equipment and wires.

2.3 Meter Type

One of the most overlooked aspects of electrical soil resistivity testing is the quality of the meter
that is used to conduct the testing. There are many meter types that are labeled as “soil resistivity
meters” and specifically denote the Wenner four-electrode method. However, they are not all
created equal and many of them are not an acceptable testing meter for power system grounding
studies. While they may be sufficient for the needs of residential, commercial, and Department
of Transportation testing of electrical soil resistivity, those meters that only have 10 mA and
50 mA do not provide enough signal power for the diverse soil conditions one is going to
encounter in the power industry.
The author’s recommendation is a meter that has at least 200 W of output power (usually needs
at least a 12 V vehicle battery) and 1,000 mA of output current as well as 400 V peak to obtain
quality and accurate data for large probe spacings.

Page 13-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

For very deep testing (>600 feet spacing or even for low soil resistivity over high resistivity soil
models), a portable generator in conjunction with a signal generator & analyzer may be necessary.
This will provide a much stronger output signal that can be tuned to a non-power frequency and
be easily discerned from any background noise.
The desire is to use a meter that has great enough output current to be able to detect any high-
resistive layers underneath conductive soil. In order to achieve this, high current is needed to be
driven through the soil to obtain meaningful differences of potential on the 2 inner probes (P1 and
P2). In such cases, it is difficult to detect the depth and the resistivity of the lower soil layers,
which may end up being fractions of a milli-volt as the potential difference! On lower quality
meters, the milli-volt difference may be outside the accuracy capability of the meter.
Most soil resistivity meters provide “okay” results for tests with probe spacing of 10 feet and
smaller, but it is those measurements that are greater than 30 feet where the voltage difference
between the inner potential probes can become very to extremely small.
Additionally, the higher quality meters often not only have more output power & more output
current, but they usually also have automatic setup failure detection, provides measurement
quality status, more accurate measurements, a large internal memory, filters out some
background noise, and low power source detection.
The high output power and high output current are much more likely to provide quality
measurements for every soil condition encountered. For example, if the site where the soil
resistivity testing is being performed has top soil with lower resistivity over a higher bottom
resistivity layers then the higher output current is necessary to obtain adequate deep soil
resistivity values.
Figure 4 shows how the surface potential changes for different soil resistivity layering (all plots
have the same color scale) with the identical probe spacing.

Scalar Potentials/Scalar Potentials [ID:UniformSoil @ f=60.0000 Hz ] Scalar Potentials/Scalar Potentials [ID:LowHigh @ f=60.0000 Hz ] Scalar Potentials/Scalar Potentials [ID:HighLow @ f=60.0000 Hz ]
LEGEND LEGEND LEGEND
100 100 100
Maximum Value : 186.538 Maximum Value : 193.459 Maximum Value : 896.331
Minimum Value : 0.328E-08 Minimum Value : 0.635E-03 Minimum Value : 0.146E-03

1 x E+3 1000.00 1000.00

50 50 1 x E+2 50 100.00 100.00
1 x E+1
10.00 10.00
1 x E+0
1.00 1.00
Y AXIS (FEET)

Y AXIS (FEET)

Y AXIS (FEET)

0 0 1 x E-1
0 0.10 0.10
1 x E-2
0.100E-01 0.100E-01
1 x E-3
0.100E-02 0.100E-02
1 x E-4
-50 -50 -50
1 x E-5 0.100E-03 0.100E-03

1 x E-6 0.100E-04 0.100E-04

1 x E-7 0.00 0.00

-100 -100 -100
-100 -50 0 50 100 -100 -50 0 50 100 -100 -50 0 50 100
X AXIS (FEET) X AXIS (FEET) X AXIS (FEET)
Potential Profile Magnitude (Volts) Potential Profile Magnitude (Volts) Potential Profile Magnitude (Volts)

Figure 4: Surface potential for identical probe spacing in different soil models. The left plot is a uniform
soil model, the middle plot is a low soil resistivity layer over high soil resistivity layer, and the right plot is
a high soil resistivity layer over a low resistivity layer.

Copyright © 2017 SES ltd. All rights reserved. Page 13-7

PART II: USER CONTRIBUTIONS

2.4 Obtaining Usable Data

It is important to keep track of the soil resistance data as it is being measured. When starting from
small “a” spacing and increasing the “a” spacing with successive measurements, typical resistance
values trends downward as the “a” spacing increases. This trend is broken when encountering
shallow low resistivity soil over high resistivity soil/rock, interference from parallel metallic
objects, or noise.
The probe-to-soil contact resistance should be kept to a minimum to ensure the resistance
measurements are of the soil and just not bad probe insertion or contact. A typical value is 1,000
ohms per pair of probes (current pair and potential pair). There are soils that will cause this to be
much higher, e.g. – dry clean sand, and it is important to make sure the contact resistance is fairly
similar between the two probes pairs. Water can be placed at the base of each probe to reduce its
contact resistance. In loosely graded soils it is a good habit to pack the soil in around the probe
after pushing or pounding it into the soil. Avoid using the same hole twice because the initial
insertion and subsequent extraction breaks up the soil, a fresh probe insertion is more likely to
result in lower contact resistances.
Ensure the power source for the meter is fully charged as limited input power can lead to
inconsistent output results.
Be in-tune with the measurement data, not testing to just get data. The output current should
vary per each measurement. Typically more current output will be required for low soil resistivity
testing and it may drop into the milli-amperes for high resistivity soil. Be mindful of each value
that is recorded; resistance, standard deviation, output current, battery source voltage, meter
settings, et cetera to prevent having to redoing measurements or entire traverses.
Some of the more advanced testing meters automatically test the soil resistivity in a forward and
backwards current (reversing the current flow). Testing the same soil twice but opposite direction
should result in identical values. If there is some background noise that is at the same testing
frequency, then this should be detected with abnormal results. Additionally the meters often
provide the standard deviation between these repeated measurements. Reversing the current
direction also prevents the probes from becoming polarized if they are iron.
When using any alternating current instead of direct current, the entire length of the wire on the
reels needs to be unrolled to prevent an induction coil. Additionally, when testing very large “a”
spacing, the mutual coupling between the alternating current current-lead and the potential leads
could induce current flow on the potential leads [1], which would cause changes in the measured
potential difference.

2.5 Last Notes

It is commonly portrayed in laymen’s terms to those not involved with grounding that the
apparent resistivity for each “a” spacing is the resistivity of the soil at the depth of the “a” spacing,
see Figure 5.

Page 13-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Soil testing is an approximate half-sphere of soil for uniform soil.

However, it is important to understand that the apparent resistivity is NOT the resistivity of the
soil at the “a” spacing depth. Each measurement is an approximate average resistivity of a half-
sphere of soil. This half-sphere may contain very conductive clay chunks, high resistivity silicon
sand, moisture with dissolved minerals, seams of organics, et cetera. The half-sphere is more
valid in a uniform or heterogenic soil type. When different layers and different resistivity become
involved, the current flows are skewed significantly.
For this reason, we conduct many successive “a” spacing measurements in order to determine
how the apparent resistivity changes with larger and larger half-sphere measurements. An
analogy the author often uses to explain this is similar to determining the path of someone’s cross-
country trip. If we only know that a traveler was in Kansas City, Kansas, we are unable to
determine where he started from or where he ended up. However, if we have more time stamped
locations, we could not only figure out the starting destination, but perhaps could even get a good
feel on the exact route and average speed along the way. The same method is used to understand
soil resistivity values and their thicknesses by collecting more data points per traverse.
Keep your equipment clean and in good working order and it will provide good data and be less
frustrating to work with. Visually check your leads and connections each day for damage that may
compromise the test, as well as electrically check the connections occasionally as well due to the
stresses of pulling wire on leads and connectors. For this reason it is important to bring some
basic repair parts and tools. There is nothing more frustrating than driving hours to a site only
having to do it shortly thereafter to get replacement parts.
Reserve a copy of your meter’s manual for those unusual situations that arise at the least expected
times. Additionally, bring contact information to a knowledgeable source or the meter’s
manufacturer technical support hotline.
It is important that you leave the site as clean or cleaner that when you arrived to do the testing.

2.6 Testing Safety

Safety is the first priority when doing any task, including soil resistivity testing. While it can be
repetitive and often the testers can be the only people onsite, there are many precautions the
testing crew should consider and prepare for. We need to be cautious of both the public and
observers that may come within the testing area as well as the testing personnel.
For each job and for each day, a job hazard assessment as well the plan for the day’s activities
should be conducted with the entire crew and other onsite crews there for other tasks. Discuss

Copyright © 2017 SES ltd. All rights reserved. Page 13-9

PART II: USER CONTRIBUTIONS

and plan for each person’s role, the associated hazards of the job at hand, procedures if an accident
does occur, and how communication will be handled among each crew and crew members.
Proper personal protection gear should be worn, including but not limited to; heavy duty boots,
pants, long-sleeve in certain conditions, high visible covering, properly sized gloves, eye
protection, and hard hats. There are common hazards to all sites; electrical hazards, trips, slips,
& falls, pinched fingers, et cetera. However, there are often very unique hazards to local climates,
such as wind or cold protection or numerous animal ground entrance holes. It is advisable that
some insect repellent is taken along as well. The author’s recommendation is to apply permethrin
spray to the outside of all clothing a couple of days prior to going into the field.
Additionally, the testing crew should be ever wary of the fauna and flora in which they work. There
are many locations that have predatory animals, e.g. – black bear or mountain lion, venomous
animals, e.g. – rattlesnakes and black widow spiders, disease carrying arthropods and insects, e.g.
– deer ticks and mosquitos, or poisonous plants, e.g. – poison ivy and giant hogweed. “An ounce
of prevention is worth a pound of cure” is an applicable saying when doubt of what an animal or
plant is. The author came across the plant on the left in Figure 6 while testing and halted the crew
to ensure it wasn’t poison oak or similar before continuing with testing (It wasn’t, but rather
horsenettle, only toxic if eaten).

Figure 6: Flora and Fauna safety.

Ensure to pause periodically to ensure each crew member is hydrated and taking precautions from
the elements (sun, wind, or cold).
Communication of when the testing meter is on and off is of utmost importance. The operator of
the meter and crew members moving the probes need clear communication signals, either verbal
or non-verbal, to denote when the probes should be considered “hot”. Whether testing is within
a public space; park, school, recreational area, or whether it’s on private land, the probes and
wires should be guarded from the curious and innocent.
Testing wires should not cross small rivers or creeks.
Testing wires should not cross over, through, or under dense-mesh metallic fence lines.
In most cases, road crossings should also be avoided. Public, city, or even lightly traveled roads
are to be avoided for risk of damage to wires, entanglement in axles, and for the safety of the
testing crew. Dirt access roads and roads within the boundary of the site should be examined
carefully. Cable/wire protection systems are available for safe vehicular crossings, although the
vehicles should be limited in crossing speed.

Page 13-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Even when sites have been cleared for testing by the states below-grade authority, treat every
probe insertion with caution should there be some unknown or unmarked cable or pipe below the
surface.
Last but not least, there also could be much said about travel safety as this type of testing is
frequently outside of one’s home area. The list should include at a minimum of: airport security
of personal information and property, rental vehicle and equipment safety, hotel valuables,
personnel and stranger awareness of those in and around the job site, and also local political or
cultural unrest.

3 Conclusion
Test all the way around the site, be aware of parallel conductive objects, take more tests now and
less re-work later, take many “a” spacings per traverse, use a quality meter, check your equipment,
safety first, and get quality data.

4 Acknowledgments
I want to thank my colleagues Alex Silva (B&V) and Lalit Yadav (B&V) for careful reading of the
manuscript and useful suggestions.

5 References

[1] J. Ma & F. P. Dawalibi, "Influence of Inductive Coupling Between Leads on Resistivity Measurements
in Multilayer Soils", IEEE/PES Transactions on PWRD, Vol. 13, No. 4, pp. 999-1004, October 1998.
[2] IEEE 80-2013 IEEE Guide for Safety in AC Substation Grounding, Equations 49&50.
[3] R. D. Southey, S. Tee, W. Ruan and F.P. Dawalibi, "A New Efficient Measurement Techniques Based
on the Wenner Four Electrode Method", The International Conference on Electrical Engineering 2007
(ICEE), Hong Kong, July 8 - 12, 2007.
[4] J. Ma and F. P . Dawalibi, "Study of Influence of Buried Metallic Structures on Soil Resistivity
Measurements", IEEE Transactions on Power Delivery, Vol. 13, No. 2, April 1998.
[5] R. D. Southey and F. P. Dawalibi, "Improving the Reliability of Power Systems with More Accurate
Grounding System Resistance Estimates", 2002 International Conference on Power System
Technology Proceedings, Vol. 1, pp. 98-105, October 2002.

Copyright © 2017 SES ltd. All rights reserved. Page 13-11

 UGM 2017 – CANNON BEACH, OREGON

14 TRANSFERRED VOLTAGES ON DRILLING EQUIPMENT

AND DISCONNECTED ELECTRODE LEADS IN
ENERGIZED SUBSTATIONS
Jesse Rorabaugh PE

Southern California Edison

Email: [email protected], Web Site: www.sce.com
Abstract

Voltages on drilling equipment or leads to deep grounding electrodes in an energized substation can exceed 70% of
the ground potential rise of a substation. This creates hazards for people in the vicinity of drilling equipment in
energized substations which need to be addressed to ensure worker safety. Additional care needs to be taken when
grounding this equipment to ensure an equipotential zone at the worksite. Neither drilling equipment, nor leads to
electrodes should be left ungrounded at any stage of the installation process.

1 Introduction
One technique used at Southern California Edison to mitigate step and touch voltage issues within
existing energized substations is through the drilling of deep grounding electrodes. These deep
grounding electrodes consist of two 4/0 AWG bare copper conductors, lowered to the bottom of
a 10-30 cm diameter bore drilled as deep as 200 meters. They connect to the existing ground grid
below grade so equipment can drive on top of them, and they are installed using standard drilling
equipment used to drill water and cathodic protection wells of a similar depth.
This design significantly reduces the total ground potential rise of the substation with minimal
operational disruption. Minimal trenching is required, so restoration costs are also reduced. Deep
grounding electrodes are also more effective than horizontal grounding conductors in typical
California soil, which is dry in the upper layers but contains groundwater below.

1.1 Challenges in Installation of Deep Grounding Electrodes

During installation there were several reports of sparking on drilling equipment or to deep
grounding electrode leads that installation brought attention to proper bonding and grounding of
this equipment in the worksite. In a typical case shown in Figure 1, sparking to the shaft was
observed after the first 70 meters of the bore were drilled. Measurements showed an open circuit
voltage of 4 volts from the ground grid to the shaft and a short circuit current of 22 amperes were
measured when the shaft was grounded. This current comes from neutral current returning to the
transformer banks in the energized substation. Once the electrode reaches significant depth, it
becomes a very good return path for this neutral current and a significant portion of the return
current takes this path.

Copyright © 2017 SES ltd. All rights reserved. Page 14-1

PART II: USER CONTRIBUTIONS

Next Section of Drill

Shaft

Drill Shaft

Sparking When
Connecting

Figure 1: Sparking to a drill shaft during installation.

The drilling rigs that had sparking were grounded according to the standard substation grounding
procedure used for most vehicles in substations, an example of this grounding procedure is
presented in Figure 2. However, somewhere between the grounding connection to the rig and the
shaft extending into the earth there was a high resistance connection. The grounding failed to
prevent sparking because of this high resistance connection.

Page 14-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: Typical ground attachment to a drilling rig.

2 CDEGS Analysis of Voltage Rise

To better understand what voltages would be expected on ungrounded drilling equipment or
electrode leads, a series of CDEGS models were created in the module HIFREQ. A generic
substation was constructed with dimensions of 30.5 meters and grid spacing of 3 meters in
100 Ωm soil. An electrode was placed in the center of a corner square of the grid as shown in
Figure 3 to represent a drill shaft or a disconnected electrode lead. Enough current was pushed to
bring the Ground Potential Rise (GPR) with no electrode to 100 volts. This was done so all voltages
can be interpreted as a percentage of initial GPR. To know actual voltages for any particular
substation, it would be necessary to model that particular substation in detail; this approach gives
typical values.
To show that grounding the deep grounding electrode provides effective protection from the
hazard, the case was rerun with the deep grounding electrode connected to the grid. This case also
demonstrates the effectiveness of installing deep grounding electrodes as a strategy to reduce step
and touch voltages.

Copyright © 2017 SES ltd. All rights reserved. Page 14-3

PART II: USER CONTRIBUTIONS

2.1 Voltages in the Vicinity of an Ungrounded Deep Grounding

Electrode

As the depth of the electrode increases, remote earth voltage is pulled closer into the substation
by a similar mechanism present in any transferred voltage phenomenon, such as when a fence is
connected to a substation. By the time the electrode reaches 152 meters, a very steep voltage
gradient exists around the electrode, as shown in Figure 4. The soil next to the electrode is nearly
remote earth potential, and full substation GPR is at the ground grid only a few feet away.
These voltage gradients produce high touch voltages for someone touching the electrode or an
ungrounded drilling rig, as in Figure 5. They also produce high step voltages in the vicinity of the
electrodes as presented in Figure 6.

Figure 3: Generic ground grid design used for analysis, disconnected electrode in top left corner.

Figure 4: Scalar potentials in the vicinity of a disconnected 152 meter deep grounding electrode.

Page 14-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Touch voltages in the vicinity of a disconnected 152 meter deep grounding electrode.

Figure 6: Step voltages in the vicinity of a disconnected 152 meter deep grounding electrode.

Copyright © 2017 SES ltd. All rights reserved. Page 14-5

PART II: USER CONTRIBUTIONS

To get a better idea of the depth that these concerns start, the metal to metal voltage as a
percentage of the GPR of the grid without an electrode was recorded for electrodes of various
depths and presented in Figure 7. This shows that metal to metal voltages can exceed 70% of the
total ground potential rise of the substation.

100

90
Metal to Metal Voltage as % of GPR

80

70

60

50

40

30

20

10

0
0 20 40 60 80 100 120 140 160 180
Depth of Electrode (m)

Figure 7: Metal to metal voltage from the electrode to the grid as depth increases.

2.2 Voltages in the Vicinity of a Grounded Deep Grounding Electrode

To confirm the hazard is eliminated by grounding the deep grounding electrode, the analysis was
repeated with the electrode properly grounded to two opposite points of the ground grid. The new
scalar potential plot for a grounded 152 meter electrode is presented in Figure 8.

Figure 8: Scalar potentials in the vicinity of a properly grounded 152 meter deep grounding electrode.

Page 14-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.3 Resistance of Deep Grounding Electrodes

The resistance of a deep grounding electrode as a function of depth is plotted in Figure 9. The
lower the resistance of the grid the better the path for neutral and fault current. Therefore, the
hazard is increased at greater depths, since there will be higher current density in the vicinity of
the electrode.

Electrode Resistance
10
9
8
Electrode Resistance (Ω)

7
6
5
4
3
2
1
0
0 20 40 60 80 100 120 140 160 180
Electrode Depth (m)

Figure 9: Resistance to remote earth of the deep grounding electrode with increasing depth.

3 Conclusion
The step and touch voltages around an ungrounded deep grounding electrode can be in excess of
70% of the total ground potential rise. Since energized substations may have a continuous rise of
several volts from neutral current, this manifests as stray voltages sparking on the drill rig.
The crews making final electrode connection to the ground grid may be different than the drilling
crews installing the electrode conductor into the bore. If the electrode conductor leads are left
ungrounded prior to final connection to the ground grid, then there is a risk of high step and touch
voltages from a fault occurring while someone is in the vicinity of the ungrounded leads. As shown
in Figure 5 and Figure 6, the step and touch voltages in the vicinity of these leads will be hazardous
at most substations. As shown in Figure 8, the problem goes away if the electrode leads are solidly
grounded at all times.
Grounding methods used during the electrode installation process were improved based on these
findings. From the moment the electrodes leads enter the earth they are required to stay
grounded. Depending on the installation practice used, either the end of the conductor is
grounded with a temporary personal ground or a stringing clamp is used, as shown in Figure 10.
It may also be necessary to separately ground the shaft of the drilling rig if it has a poor electrical
connection to the grounded frame.

Copyright © 2017 SES ltd. All rights reserved. Page 14-7

PART II: USER CONTRIBUTIONS

Stringing Clamp

4/0 Copper Electrode

Lead

Temporary Personal
Ground
Substation Location

30cm Bore

Figure 10: Use of a stringing clamp to ground conductor while lowering into bore.

This same phenomena can occur with any large structure not bonded to the main substation
ground grid. For example, if a transmission tower is grounded to ground rods rather than the
substation ground grid, it will provide a similar path for current returning to the substation.

4 Acknowledgement
Dozens of SCE employees and contractors have spent significant time and effort improving safety
for drilling in energized substations. Notable among these people are Ronald Burt, Brian Becker,
and Richard Scott who developed many of the procedures followed today.

5 References

[1] Y. Li, F. Dawalibi, S. Fortin "Transferred Transient Voltages on the Nearby Facilities When a Lightning
Stroke Hit Bridge", in International Conference on Lightning Protection (ICLP), Shanghai, China
2014.

[2] L. Agudelo, R. H. Ortiz, M. Montoya, W. H. Zapata, C. Cardenas, J. C. Echeverri, and G. Moreno

"Transferred Voltages in Multigrounded Systems" in Transmission and Distribution Conference and
Exposition: Latin America, 2008.

Page 14-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

[3] F. Grange, T. Gourdan, P. Blasquez, D. Leschi, F Dawalibi ' A New Methodology of Cranes Modeling
for ITER Grounding Safety Assessment," in CDEGS Users' Conference Proceedings, San Diego
California, USA, 2015.

[4] Y. Li, F. Dawalibi, Z. Jinsong, Q. Feng, G. Bing "Grounding of Urban GIS Substation Connected to
Commercial Buildings and Metallic Infrastructures" in CDEGS Users' Conference Proceedings, San
Diego California, USA, 2015.

[5] C. Moller "Hazards with Temporary Construction Power Substations" in CDEGS Users' Conference
Proceedings, Lake Tahoe California, USA, 2013

Copyright © 2017 SES ltd. All rights reserved. Page 14-9

 UGM 2017 – CANNON BEACH, OREGON

15 GROUNDING SYSTEM DESIGN CHALLENGES

FOR SMALL DISTRIBUTING STATIONS
IN URBAN LOS ANGELES

Travis Betros, Benjamin Hwang, Jason Hills (LADWP) and Than Aung (LADWP)

WorleyParsons Group, Inc. and the Los Angeles Department of Water and Power (LADWP)

Email: [email protected], Web Site: www.worleyparsons.com

Abstract

This project is to evaluate and analyze the grounding system at an existing, decades old distribution station to
determine the sufficiency of the system for operation on the modern bulk electric system. Key input parameters
and information are gathered to create a model of the existing system. Increases in system fault levels overtime
commonly result in original grounding systems not being adequate for the present operating environment.
Grounding augmentations are designed and modeled based upon the input criteria to develop a grounding system
to accommodate present conditions if the existing system is determined to require improvement.

1 Introduction
Grounding systems are an essential component of high voltage electric facilities to ensure the
safety of the public and personnel near and within these installations. Fault events and
lightning strikes can introduce enormous amounts of energy on the bulk electric system where
grounding systems are designed to safely disperse and dissipate these occurrences.
Countless substations throughout the United States have been in-service for more than fifty (50)
years. Equipment and controls within these substations have commonly been upgraded and
replaced over the years; however, grounding systems are ordinarily overlooked as buried and
forgotten components of electric facilities. As fault levels on the U.S. bulk electric system have
increased in recent decades, the capacity and performance of many older grounding systems are
not designed to withstand the conditions as experienced today. Grounding systems are a
necessary component to the bulk electric system for the following reasons:
 Dissipates fault current into the earth
 Provides a return path for current back to the source
 Generates even gradient of ground potentials
 Maintains step and touch voltages to safe threshold limits
 Provides safety for public and utility personnel
This paper is a case study on WorleyParsons’ work with the Los Angeles Department of Water
and Power (LADWP) for the augmentation of a grounding system at an existing, urban
distributing substation, and how WorleyParsons was able to overcome the project design
challenges that were unique to this facility.

Copyright © 2017 SES ltd. All rights reserved. Page 15-1

PART II: USER CONTRIBUTIONS

2 Case Study
LADWP is the largest municipal utility in the United States supplying water and electric services
in Los Angeles, California for over four (4) million people. The power branch of LADWP was
established in 1916 [1]. Many smaller areas of the city are served electrical service through
distributing stations at a 4.8 kV level that provide localized distribution. The Distributing
Station (DS)-102 substation was placed in-service in 1932, and the grounding system today is
believed to be mostly original to the facility. The grounding system at DS-102 was modeled and
analyzed to determine whether the existing grounding system is adequate for the environment
as seen at the station today.
The grounding upgrade project had the following sequence:
1. Consolidate record drawings / details of existing grounding system
2. Obtain relevant design criteria:
a. Utility design standards and practices
b. Fault level values and clearing time
c. Soil resistivity measurements
3. Create new CAD grounding model
4. Determine adequacy of existing grounding system
5. Augment existing grounding layout (as required)
WorleyParsons, a global energy infrastructure engineering firm with over 24,000 employees,
assisted in the detailed engineering and design of the distributing station grounding upgrades.

3 Design Process
The CDEGS software application was the modeling program utilized for the grounding analysis.
1. RESAP module: analyzes soil resistivity measurements and develops soil models to use
in calculation of grid impedance and voltage profiles within the extent of the ground grid
area.
2. MALZ module: analyzes the frequency domain performance of buried conductor
networks and calculates the earth and conductor potentials in order to analyze various
quantities, including the ground potential rise (GPR) and the step and touch voltage
profiles.

3.1 Soil Resistivity Data Measurement

The first step in the design process is the development of the soil model. Soil models are created
from soil resistivity data measurements collected from the following methods:
1. Laboratory measurement of soil borings
2. Wenner four-pin method
After weighing the pros and cons, the Wenner four-pin method was the technique that was
applied for the soil resistivity measurement collection.

Page 15-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 1: Wenner Four-Pin Method (IEEE Std 80-2013) [2].

The Wenner four-pin method is neither suitable nor ideal for existing facilities with grounding
systems or in locations (e.g. urban environments) that have known conductive infrastructure
buried below-grade, as these items will interfere with the measurement results. Due to the
location of the DS-102 facility within Los Angeles, the Wenner method was not recommended to
be performed within the substation fence.
Although the Wenner method is not recommended to be performed local to a facility, the
Wenner method can be and is often performed at a nearby property that is suited for the
application (e.g. recreation parks and fields), as was the practice under this project. The
Wenner four-pin method was performed at several nearby locations for data sampling and
comparison.

Figure 2: Soil Resistivity Measurements.

Copyright © 2017 SES ltd. All rights reserved. Page 15-3

PART II: USER CONTRIBUTIONS

The results need to be evaluated and not blindly applied to a model. Understanding the unique
location, climate and geography of the substation site was key in evaluating the soil resistivity
measurements:
1. The resistivity results shown in Figure 2 for the Hansen Dam Recreational Area show
values significantly lower compared to the other locations. The measurements for the
Hansen Dam Recreational Area were collected in a location near a body of water where
the soil has higher moisture content in comparison to the substation location. This
situation results in lower resistivity values as could be reasonably expected; therefore,
those values were excluded from the soil model development process.
2. The soil resistivity values taken at Ritchie Valens Park appeared suspect; therefore,
WorleyParsons disregarded those measurements in the CDEGS soil model.
The CDEGS RESAP module allows resistivity measurements collected from the Wenner method
to be directly entered into the software for soil model development.

3.2 Split Factor & Return Path

The transmission lines into a substation often feature static wires and the distribution feeder
circuits exiting the station often feature distribution neutrals tied to the main grid. As a result, a
significant portion of the fault current from the electrical system will return (or split) to its
source (wye-connected windings at remote transmission substations) via the static wires and
feeder neutrals that leave a facility As a result, a reduction known as the split factor can be
applied.
The split factor indicates the portion of the ground fault current that will be injected into the soil
surrounding the ground grid to return via the earth; the remainder returns to the source via the
overhead static wires and the distribution neutrals attached to the facility grounding system.
The greater number of terminal connections to a station with static wires and feeder neutrals
that tie into the facility grounding system results in a ground grid not being required to be as
extensive and robust as otherwise be required if there were fewer terminal connections.
Wye-connected distribution systems typically feature a four (4) cable / conductor connection
with three-phases and a neutral. In the case of the LADWP 34.5kV distribution system, the wye-
connected system only features a three-phase configuration without a neutral (and no static
wires); therefore, in the event of a fault condition, no neutral connection is available to be
utilized as a return path to help dissipate fault currents.
The alternate return path for the case of the LADWP DS-102 distributing station is modeled as
earth. The earth path is considered to withstand the full magnitude of any fault event without
the utilization of a split factor. For this situation, the localized grounding system at a specific
distribution station must endure the complete amount of fault current as indicated in the system
model. This directly results in a grounding system being more robust than otherwise would be
required if a split factor could be applied.

3.3 Existing Grounding System

Prior to any analysis work to begin, an updated, consolidated CAD model of the existing system
was created. The existing grounding system was illustrated across three (3) drawing (i.e. raster
images). The existing grounding system was reconstructed in AutoCAD based on available

Page 15-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

record drawings and then imported via a 3D .DXF file into the CDEGS program MALZ module,
along with other relevant engineering data.
The analysis of the soil model and existing grounding system provided a reference point to
determine the adequacy of the grid and identify the required augmentations to be designed.
The following are the key input parameters for the analysis of the existing ground grid:
 Ground fault current value of 7.75 kA with an assumed X/R ratio of 20.
 Existing ground grid 4/0 AWG CU conductor with a radius of 0.261 inches.
 Ground grid depth: eighteen (18”) inches below compacted earth.
 Existing ground wells were modeled as 2” diameter (1” radius), twenty (20’) feet length.
 Foundation wire mesh was modeled as #6 AWG with a 2’ x 2’ spacing.
o Foundation section drawings indicate a 6” x 6” wire mesh of #6 AWG size.
o Foundation section drawings indicate foundation wire mesh to be on a 6” x 6”
spacing; however, the analysis was modeled as 2’ x 2’ spacing due to memory
allocation limitations within CDEGS.
 Conservative two-layer soil model (developed from recent soil resistivity values):
o Upper Layer: 457.8 Ω-m; 4.6 feet thickness depth
o Lower Layer: 142.3 Ω-m; infinite thickness depth
 Surface aggregate: Six (6”) inches, 8,000 Ω-m surface layer (see Subsection 3.5)
 Clearing time of 0.5 s (30 cycles).
 IEEE Std 80-2013, 50 kg ventricular fibrillation limits.

Copyright © 2017 SES ltd. All rights reserved. Page 15-5

PART II: USER CONTRIBUTIONS

3.4 Existing Grounding System Results and Plots

Figure 3: Existing CAD Drawing 2D Top View,

Red lines = existing grid conductor,
Yellow dots = existing ground wells.

Figure 4: Existing Grid Model 2D Top View.

Figure 5: Existing Grid Model 3D View.

Page 15-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 6: Existing GPR Plot.

Figure 7: Existing Step Voltage Plot (without surface rock).

Copyright © 2017 SES ltd. All rights reserved. Page 15-7

PART II: USER CONTRIBUTIONS

Figure 8: Existing Step Voltage Plot (with surface rock).

Figure 9: Existing Touch Voltage Plot (without surface rock).

Page 15-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 10: Existing Touch Voltage Plot (with surface rock).

The analysis of the existing grounding system concluded the grid arrangement is neither
adequate nor safe for limiting step and touch voltages to below minimum threshold values. The
calculation results verified the expected assumption. The next step was to design a grounding
system sufficient to accommodate the operating parameters of the facility.

3.5 Aggregate Rock Selection

During the evaluation of the touch voltages at the DS-102 Distributing Station, the analysis of
the grounding model became problematic. The first method to mitigate touch voltages is to
simply decrease spacing of grid conductor to form a tighter mesh; however, output plots
indicated touch voltage threshold limits were still being exceeded in several locations
(particularly along the fence line) even after modeling the grid with the following parameters:
 Grid spacing at six (6”) inches
 Aggregate surfacing modeled at 3,000 Ω-m, six (6”) inches thick
For all practical purposes, installing grid conductors on six (6”) inch spacing does not provide
for a realistic, cost-effective solution.
The course of action was determined to specify an aggregate surfacing with a higher resistivity
than standard applied aggregates. For the case of the DS-102 Distributing Station, a granite
aggregate modeled with a resistivity of 8,000 Ω-m, at six (6”) inch thickness was specified.

Copyright © 2017 SES ltd. All rights reserved. Page 15-9

PART II: USER CONTRIBUTIONS

Figure 11: Surface Material Resistivity Values (IEEE Std 80-2013) [2].

The 8,000 Ω-m aggregate modeled was based upon a granite aggregate sourced in Florida
following a FDOT course aggregate specification. As the DS-102 station is located in southern
California, the aggregate for the project needed to be an equivalent granite but sourced locally.
A civil / geotechnical counterpart was required to identify an equivalent, locally sourced
aggregate that was viable for the project specifications.

3.6 Foundation Rebar / Mesh Bonding

Touch potentials in areas of native earth are alleviated with the addition of grid conductors and
application of aggregate. Touch voltages are not just an issue for areas of native earth, touch
voltages are a concern over concrete surfaces.
Assumptions are often thought concrete surfaces provide high resistive values. Laboratory
measurements conclude the opposite. The resistivity of concrete surfaces can expect to range
between 300 and 600 Ω-m. Concrete is a material that absorbs and retains moisture resulting
in low resistive values and being semi-conductive [3]. With low resistivity values of concrete,
concrete can feature resistivity levels equal or less than the surrounding native earth surfacing.
Step and touch voltages on concrete surfaces can be mitigated by the following methods:
1. Install network of metal grating over concrete surface
2. Bond foundation rebar / wire mesh to grounding system
Page 15-10 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

Equipotential bonding of foundation reinforcement to a grounding system is a method of

bonding to provide a dense, stabilized voltage profile at the surface of the concrete slabs by
raising the mesh of the reinforcement to the same potential as the grid to which it is bonded. It
is important to note that the reinforcement should not be the ground path for the fault currents;
it should merely provide potential equalization.
All the DS-102 station foundations are slab-on grade type. The detail drawings for the facility
illustrate the foundations are reinforced with a 6” x 6” wire mesh of #6 AWG size. The
grounding system was then analyzed to develop the step and touch profiles with the inclusion of
the foundation wire mesh (non-bonded).
Existing Grounding System (with Wire Mesh non-bonded) Results and Plots:

Figure 12: Existing CAD Drawing 2D Top View (non-bonded wire mesh),
Red lines = existing grid conductor,
Cyan lines = foundation wire mesh,
Green dots = existing ground wells.

Copyright © 2017 SES ltd. All rights reserved. Page 15-11

PART II: USER CONTRIBUTIONS

Figure 13: Existing GPR Plot (non-bonded wire mesh).

Figure 14: Existing Step Voltage Plot (with surface rock and non-bonded wire mesh).

Page 15-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 15: Existing Touch Voltage Plot (with surface rock and non-bonded wire mesh).

For initial assessment, the best case scenario modeled the profile plots with an application of
aggregate 8,000 Ω-m, at six (6”) inch thickness. As shown in Figure 15, touch voltages still
greatly exceed the minimum threshold limit. The DS-102 grounding system is comprised of the
following components:
1. Fence grounding conductor
2. 34.5 kV equipment / structure grounding system
3. 4.8 kV equipment / structure grounding system
These grounding components are originally isolated from each other. As each grounding
component is not connected together, less favorable performance is achieved if otherwise
connected together to form a larger footprint to create a more robust grid layout. As shown in
Figure 15, a fault is simulated on the 34.5 kV main bus, the local grid component takes the full
magnitude of the fault; hence, the color contours are skewed to the right-side of the plot.

4 Final Grounding System Results and Plots

New ground grid members were added to the existing grid to properly route around the facility
foundations and to cover all grounded equipment and structures. The existing and new grid
conductors and ground rods were then imported via a 3D .DXF file into the CDEGS program
MALZ module, along with other relevant engineering data.
The existing grounding conductors for the station are 4/0 AWG copper. This conductor size is
determined to be more than adequate to handle the maximum fault current and fault clearing
time as modeled on the 34.5 kV bus. Due to the situation, the new grounding conductors were
selected to match the existing grounding system.

Copyright © 2017 SES ltd. All rights reserved. Page 15-13

PART II: USER CONTRIBUTIONS

Figure 16: Minimum Conductor Size.

4.1 Output Parameters

 Ground conductor:
o New conductor network length: 2,839 feet of 4/0 AWG CU.
o Existing conductor network length: 1,371 feet of 4/0 AWG CU.
o Total conductor network length: 4,210 feet.
 The grounding network covered an area of approximately 160 feet by 50 feet.
 Existing ground wells: A total quantity of four (4), twenty feet (20’) long, 2” diameter
ground wells were modeled.
 New ground rods: A total quantity of fifteen (15), twenty feet (20’) long, 5/8” diameter
ground rods were modeled.
 Calculated grid resistance
o Existing calculated grid resistance: 10.393 Ω
o New calculated grid resistance: 2.169 Ω
 Ground fault current: 7.75 kA
Page 15-14 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

 This new ground grid will provide a ground potential rise (GPR) of 16,809 volts with
respect to remote earth.
 With six (6”) inches, 8,000 Ω-m surface layer installed (per LADWP), the IEEE Std 80
calculations establish a step voltage of 6,370.8 volts, and a touch voltage of 1,709.7 volts
for the new grid design.

Figure 17: Final CAD Drawing 2D Top View,

Red lines = existing grid conductor,
Yellow lines = new grid conductor,
Cyan lines = foundation wire mesh,
Green dots = existing ground wells,
Magenta dots = new ground rods.

Figure 18: Final Grid Model 2D Top View.

Figure 19: Final Grid Model 3D View.

Copyright © 2017 SES ltd. All rights reserved. Page 15-15

PART II: USER CONTRIBUTIONS

Figure 20: Final GPR Plot.

Figure 21: Final Step Voltage Plot (without surface rock).

Page 15-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 22: Final Step Voltage Plot (with surface rock).

Note that this grid does not exhibit step voltage issues according to the calculation methods of
IEEE Std 80 with a rock surfacing. All areas within the extent of the grounding system are
shown as white coloring to illustrate step voltage values are below the minimum threshold.

Figure 23: Final Touch Voltage Plot (without surface rock).

Copyright © 2017 SES ltd. All rights reserved. Page 15-17

PART II: USER CONTRIBUTIONS

Note the areas within the extent of the foundations with wire mesh do indicate levels of touch
voltages above the minimum threshold. The wire mesh in these areas was modeled on a 2’ x 2’
spacing due to analysis limitations. It is believed the actual spacing of 6” x 6” would be close
enough to mitigate the touch voltages in these locations.

Figure 24: Final Touch Voltage Plot (with surface rock).

Note that this grid does not exhibit touch voltage issues according to the calculation methods of
IEEE Std 80 with a rock surfacing. All areas within the extent of the grounding system are
shown as white coloring to illustrate touch voltage values are below the minimum threshold.

5 Project Status
As of May 2017, the project grounding analysis as prepared by WorleyParsons has been
approved by LADWP. Construction drawings have also been prepared and approved and
awaiting for the commencement of the construction phase.

6 Conclusion
As initially believed, the existing grounding system at the DS-102 facility is not adequate to
properly maintain step and touch voltages to below minimum safety thresholds as seen on
today’s electric system. Grounding systems are an essential component to the bulk electric
system that often gets overlooked and unnoticed. Grounding systems for older electric facilities
should be evaluated for adequacy and safety.

Page 15-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

7 References

[1] LADWP, "Facts & Figures," LADWP, 2013. [Online]. Available:

https://www.ladwp.com/ladwp/faces/ladwp/aboutus/a-power/a-p-factandfigures?_adf.ctrl-
state=yxtz0350z_21&_afrLoop=646128027307261. [Accessed 29 May 2017].
[2] IEEE Power and Energy Society, "IEEE Guide for Safety in AC Substation (IEEE Std 80-2013)," IEEE
Standards Association, New York, 2013.
[3] "Effects of Concrete on Touch and Step Voltages in Substations," in IEEE Substations Committee
Annual Meeting, Chicago, 2011.

Copyright © 2017 SES ltd. All rights reserved. Page 15-19

 UGM 2017 – CANNON BEACH, OREGON

16 THE CHALLENGES OF DESIGNING SOLAR PARK

EARTHING SYSTEMS IN THE UK

Marc Jones - Senior Consultant

RINA Consulting Ltd (formerly Edif ERA)

Ty Canol, Clun Avenue
Pontyclun, Rhondda Cynon Taf, Wales, UK

Email: [email protected], Web Site: www.rinaconsulting.org

Abstract

Soil resistivity within the UK can vary greatly from one area to another, in some instances even within the space
of one or two kilometres. When designing large earthing systems, for example at wind farms and solar (PV)
parks, this variation can pose problems when attempting to prove the safety of a site, both within the earthing
system and outside. This article introduces some of the challenges faced when designing earthing systems for
commercial solar (PV) plant in the UK and how these were overcome. The investigation included a number of
CDEGS MALZ simulations to determine the most cost-effective design solution.

1 Introduction

Circa 2010, financial feed-in-tariff incentives being offered by the UK government lead to a large
increase in solar PV generation schemes, which kick-started a previously non-existent
commercial solar power industry.

As this was an emerging industry, construction companies within the UK did not have sufficient
knowledge or expertise regarding the construction of commercial solar parks, therefore foreign
(mostly European) companies, unfamiliar with the safety documentation and regulations within
the UK began constructing sites with very little regard to earthing and UK electricity company
standards. This quickly became an issue that needed resolving to satisfy distribution network
operators (DNO) and meet the minimum requirements of the relevant safety documentation.

At this time we (along with most other earthing specialists in the UK) had very little experience
in this industry, however using our knowledge of earthing systems and UK safety policies were
able to design suitable earthing systems for these sites, which (for a time) became the standard
that most 33 kV solar parks adopted. This continued for approximately four years until most of
the ‘prime locations’ were utilised and what were previously considered to be less favourable
locations, became viable due to reduced installation costs. The construction of solar parks has
significantly reduced following the withdrawal of subsidies in 2017.

This paper looks at some of the challenges that have been faced and the solutions that were
developed to overcome them.

Copyright © 2017 SES ltd. All rights reserved. Page 16-1

PART II: USER CONTRIBUTIONS

2 Background

2.1 Project Requirements

Design site specific earthing systems that will control any fault current flowing to earth and will
also ensure that the touch and step voltages do not exceed tolerable limits either within the site
boundary or beyond the earthing system at third-party locations, such as private dwellings, farm
buildings, nearby LV supplies and buried utilities, whilst also ensuring the safety (touch and
step voltages) of personnel operating within the site. In addition to this, the earth potential rise
(EPR) must also be considered including its transfer to third-party locations and the effect on
telecommunications equipment.

2.2 Challenges

Most commercial solar parks are constructed on greenfield sites, usually farmland, which are
considered ‘blank canvasses’ for the earthing system, however this does not mean that the
process of designing the earthing system will be straightforward. Each site comes with its own
specific challenges, be it soil resistivity, earth fault current, fault clearance times, nearby third-
party properties/installations, buried structures that prohibit installation (such as archaeology)
beyond ‘hand-digging’ depth (sometimes as shallow as 300 mm) or even areas populated by
protected flora and fauna; some sites may contain a number of these challenges and therefore
may require very specific solutions to achieve a safe, cost-effective earthing system design.

3 Soil Characteristics
Soil resistivity values vary greatly across the UK, which makes predicting the extent/complexity
of an earthing system very difficult. Additionally, depending on the size of a solar park and the
gradient of the land it occupies, there could be vastly differing soil resistivity values across one
site alone.
To overcome these issues and facilitate the earthing system design, multiple soil resistivity tests
are carried out, using the Wenner method, at locations across a proposed site. This then allows
an average soil resistivity model to be created which is usually biased towards higher resistivity
layers to be conservative.
Table 1 below gives an example of the measurement results obtained from a particularly
challenging site in South Wales. In this particular example, the PV arrays are spread across
three different agricultural fields, each at different elevations due to the location of the site (test
locations are shown in Figure 1). The results show that the uppermost array field (represented
by SR1 & 2) has much higher soil resistivity than the other two fields (represented by SR3 and
SR4 respectively). The data was analysed using the RESAP [1] module of CDEGS.

Page 16-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 1: Soil Resistivity Measurement Data.

SR1 SR2 SR3 SR4

Wenner Measured Apparent Measured Apparent Measured Apparent Measured Apparent
Spacing (m) Resistance Resistivity Resistance Resistivity Resistance Resistivity Resistance Resistivity
(Ω) (Ωm) (Ω) (Ωm) (Ω) (Ωm) (Ω) (Ωm)

1.0 284.00 1784.42 294.0 1847.26 246.0 1545.66 56.5 355.00

1.5 241.00 2271.37 232.0 2186.55 146.1 1376.96 44.7 421.29

2.0 191.60 2407.72 222.0 2789.73 123.3 1549.43 41.4 520.25

3.0 173.80 3276.05 187.6 3536.18 86.8 1636.14 32.0 603.19

4.5 131.80 3726.56 135.5 3831.17 54.8 1549.43 23.4 661.62

6.0 116.10 4376.87 123.6 4659.61 39.8 1500.42 19.86 748.70

9.0 78.60 4444.73 87.1 4925.39 25.0 1413.72 15.11 854.45

13.5 39.60 3358.99 55.8 4733.12 14.26 1209.58 10.69 906.76
18.0 28.60 3234.58 - - 9.42 1065.38 8.55 966.98
27.0 12.96 2198.61 16.80 2850.05 5.59 948.32 5.56 943.23

36.0 2.74 619.77 9.16 2071.94 4.06 918.35 4.34 981.68

54.0 - - - - 4.70 1594.67 4.46 1513.24

Figure 1: Soil Resistivity Measurement Locations.

Due to the degree of variation in the test results, the original soil model created for the site was
biased towards the higher resistivity datasets for added conservatism and is shown in Figure 2
below.

Copyright © 2017 SES ltd. All rights reserved. Page 16-3

PART II: USER CONTRIBUTIONS

LEGEND
4
10
Measured Data
Computed Results Curve
Soil Model
Apparent Resistivity (Ohm-meters)

Measurement Method..: Wenner

RMS error...........: 122%

Layer Resistivity Thickness

Number (Ohm-m) (Meters)
3 ====== ============== ==============
10 Air Infinite Infinite
2 1050.000 1.000000
3 4000.000 6.500000
4 1400.000 Infinite

2
10
-1 0 1 2
10 10 10 10 RESAP <Resap Soil Model >

Inter-Electrode Spacing (meters)

Figure 2: RESAP Plot of Initial Soil Resistivity Model.

3.1 Variations in Soil Resistivity

As previously mentioned, the solar park is located on top of a hill and due to the sloping nature
of the ground, the soil resistivity displays a large degree of variation from one array field to the
next. This resulted in a very conservative ‘average’ soil model, which gave very high calculated
impedance values for the earthing system.
To improve the accuracy of the calculations it was recommend that a small amount of earth
electrode would be installed as a trial. This earth electrode formed a perimeter earth loop
around the uppermost PV array field, approximately 700 m in length. This loop was modelled in
MALZ [2] to determine its expected earth impedance in the soil model shown in Table 1, whilst
on site testing (using the Fall-of-Potential method) was used to obtain its actual measured earth
resistance to allow a comparison to be made between the two values. The installed earth
electrode was calculated to have an earth impedance of 7.38 Ω, however interpretation of the
fall-of-potential test results returned a value of 3.66 Ω; which is a significant difference,
indicating that a less conservative soil model may be appropriate.
In response to the large observed difference between the calculations and test result, the soil
model was adjusted, allowing more accurate calculations to be made; the revised soil model is
shown in Figure 3.

Page 16-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

LEGEND
4
10
Measured Data
Computed Results Curve
Soil Model
Apparent Resistivity (Ohm-meters)

Measurement Method..: Wenner

RMS error...........: 83.18%

Layer Resistivity Thickness

Number (Ohm-m) (Meters)
3 ====== ============== ==============
10 Air Infinite Infinite
2 1307.000 1.000000
3 1779.000 5.300000
4 863.0000 Infinite

2
10
-1 0 1 2
10 10 10 10 RESAP <Resap Soil Model >

Inter-Electrode Spacing (meters)

Figure 3: Final Soil Resistivity Model.

The revised soil model resulted in a reduction in calculated value from 7.38 Ω to 5.38 Ω.
Although still quite conservative when compared to the measures value, it was decided that to
reduce the values further (providing a closer match with SR4 values) would result in overly
optimistic calculations that would not account for the higher soil resistivity in the upper array
fields.

4 Earthing System Design

In the UK, solar park earthing systems must comply with various safety standards. Generally the
local DNO will have their own earthing policy document/guidelines which are based on the
requirements of the Energy Networks Association (ENA) document ENA TS 41-24 [3], however
both documents must be taken into account when designing the earthing associated with the
DNO operated intake substation.
The solar parks are often also designed in accordance with ENA TS 41-24, however in some
circumstances the solar park owner will request that a site be designed in accordance
BS EN 50522 [4]. Normally the reason for this request is potential cost saving; BS EN 50522 has
touch and step voltage allowances that are higher than ENA TS 41-24 which can sometimes
result in a smaller earthing system. {ENA TS 41-24 is currently being re-written. Once
complete, touch and step voltage limits will be aligned with those of BS EN 50522.}
For the example site, an initial earthing system design was created based on the soil model in
Figure 1. However, this resulted in a very high EPR (which was above the maximum preferred
limit of 3 kV set by the local DNO) and unsafe touch voltages. Following adjustment of the soil
model (Figure 3), the CDEGS model was revised. This provided a lower EPR, which was below
the 3 kV limit, however was still unsafe for touch voltages. In this case a touch voltage limit of
129 V applied, based on a fault clearance time of 1.25 s.

Copyright © 2017 SES ltd. All rights reserved. Page 16-5

PART II: USER CONTRIBUTIONS

In order to overcome these issues and provide an earthing system design that achieves safety as
well as meeting DNO approval, a variety of different methods were investigated prior to the final
design submission. These methods are described in the following sub-sections.

4.1 Additional Radial Electrodes

Radial earth wires were added to the earthing system design to increase the effective area of the
earthing system in an attempt to reduce the overall impedance and hence touch voltages. Due to
nearby properties to the north of the site, radials were only added to the southernmost array
field.
In the UK, as well as a requirement to achieve tolerable touch and step voltages, electrical
installations are often referred to as either ‘hot’ or ‘cold’ sites. This classification refers to voltage
thresholds used by telecommunications companies to assess the safety of their operators and
equipment in terms of transfer potential. The threshold values are dependent on EPR and fault
clearance time and are displayed in Table 2 below.
Table 2: ENA ER S36 [5] Threshold Values.

Fault Clearance Time Lower Threshold Upper Threshold

< 0.2 seconds 650V 1700V

≥ 0.2 seconds 430V 1150V

If the EPR of a site is calculated to be below 430 V (or 650 V) it will classified as ‘cold’ and no
precautions or mitigation is required; above this threshold a site will be classified ‘hot’. If an
electrical installation is declared a ‘hot’ site, British Telecom (BT) must be informed to ensure
that special precautions are taken and strict working procedures adopted when working on
telecommunications equipment within the ‘hot zone’. Typically BT will be presented with a
report detailing the extent of the ‘hot zone’ contours, i.e. the zone where the ground surface
potential exceeds the value in Table 2.

For telecommunications equipment outside an electrical installation, no special precautions are

required until the upper limit from Table 2 is breached. Where the EPR exceeds 1150 V (or
1700 V) , special mitigation may be required (which would be assessed by BT on a site-by-site
basis). Figure 4 shows the extent of the hot zone contours for the solar park earthing system,
including the additional radials. The fault clearance time at the solar park was stated to be
1.25 s, therefore the 430 V and 1150 V threshold values apply. The 430 V and 1150 V hot zone
contours are shown in Figure 4 below.

Page 16-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

[ID:Alternative (4th) ra @ f=50.0000 Hz ] LEGEND

MAXIMUM VALUE : 2421.241

1500 MINIMUM VALUE : 256.4110

1
1000
2
Y AXIS (METERS)

Level 2 ( 1150.00 )
500

Level 1 ( 430.000 )
-500
-500 0 500 1000
X AXIS (METERS)

Potential Profile Magnitude (Volts)

Figure 4: ‘Hot Zone’ Plot Showing Solar Park Radial Electrodes.

Although a reduction in earth impedance was achieved with the above design it unfortunately
increased the extent of the ‘hot zone’ contours and resulted in an increased transfer potential to
nearby third-party properties; also it had little effect on touch voltages within the solar park
itself.
In addition to the increased transfer potential issues, the land that was to be utilised for laying
the earth wire radials was mostly owned by third-parties and would have required legal access
rights to be sought, which would have resulted in lengthy delays to the construction timetable.

4.2 Internal Solar Park Radials

The second option investigated to reduce touch voltages within the solar park involved
additional earth electrode to form a mesh between the rows of PV arrays and in other areas
within the site where touch voltages were particularly high as shown in Figure 5 below.

Copyright © 2017 SES ltd. All rights reserved. Page 16-7

PART II: USER CONTRIBUTIONS

900

720

540
Y AXIS (METERS)

360

180

0
-90 90 270 450 630 810

X AXIS (METERS)

Top View of Conductors

Figure 5: Conductor Plot Showing the Internal Radial Electrodes.

Although the solution shown above succeeded in reducing touch voltages within the site to
below the tolerable limits, adding electrode inside the perimeter system does not have a
significant effect in reducing the overall impedance for such a large earthing system, therefore
the design still required additional radials outside the site boundary (as shown in Section 4.1) to
reduce the overall EPR. In addition, the amount of additional earth electrode that was required
within the site to mitigate touch voltages was considerable and would have resulted in an
extremely costly earthing system that would have also been difficult to install; in particular in
the north-western corner of the site.

4.3 Use of PV Array Support Piles as Part of the Earth Electrode

The PV arrays are (in most cases) mounted on galvanised steel support piles (see Figure 6). At
large commercial solar parks, these support piles can number in the thousands and are driven
down to 2 m into the ground. When bonded to the earthing system, each support pile not only
acts as an earth rod, which will help to lower the earth impedance, but as they number so greatly
and are equally spaced (a few metres apart) throughout the area occupied by the PV arrays, they
also help to equalise surface potentials, which helps to control touch voltages; occasionally,
bonding the PV array structures is sufficient to reduce touch voltages without the addition of
earth wire radials between the rows (as proposed in Section 4.2).

Page 16-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 6: PV Array Support Structures.

However, despite the number of PV array support piles, due to the high soil resistivity at this
site, their effect on further reducing the overall earth impedance was minimal therefore their
inclusion in the earthing system design is purely to equalise surface potentials and improve
safety. Also, in this example, the local DNO will not accept the use of galvanised steel as an earth
electrode (due to its susceptibility to corrosion), therefore the main elements of the earthing
system must comprise of copper conductor, however they do accept that PV array piles may be
used to provide additional touch voltage control.
The CDEGS plots below in Figure 7 and 8 show touch voltages in the uppermost PV array field
expressed as a percentage of the EPR. In Figure 7, only the perimeter earth loop is included in
the MALZ model, while Figure 7 includes the PV array support piles; the horizontal lines are
insulated conductor which represent the above ground metallic PV array frames that link each
support pile and provide electrical continuity.

[ID:Solar Park Only No P @ f=50.0000 Hz ]

950 LEGEND
Maximum Value : 68.473
Minimum Threshold : 55.000

900 68.47

67.13

65.78
850
Y AXIS (METERS)

64.43

63.08

800 61.74

60.39

59.04
750
57.69

56.35

700
-40 10 60 110 160 210
X AXIS (METERS)
Touch Voltage (% Ref. GPR) [Wors]

Figure 7: Touch Voltages - No Support Piles Included.

Copyright © 2017 SES ltd. All rights reserved. Page 16-9

PART II: USER CONTRIBUTIONS

[ID:Solar Park Only with @ f=50.0000 Hz ]

950 LEGEND
Maximum Value : 56.632
Minimum Threshold : 20.000

900 56.63

52.97

49.31
850
45.64
Y AXIS (METERS)

41.98
800 38.32

34.65

750 30.99

27.33

23.66
700
-40 10 60 110 160 210
X AXIS (METERS)
Touch Voltage (% Ref. GPR) [Wors]

Figure 8: Touch Voltages - Support Piles Included.

In each plot, the minimum threshold has been set to clear all contours from the area occupied by
the PV arrays to determine what the ‘safe’ level would be. Comparing the two plots it can be seen
that simply bonding the PV array support piles reduces the minimum threshold from 55% of the
EPR to 20%, which is a significant improvement. Without bonding the piles, this improvement
would only be possible by installing additional copper conductor, therefore this simple step has
allowed less copper to be utilised in the design; allowing cost reductions to be made.

4.4 Network Interconnection

Due to the issues regarding voltage transfer to nearby properties (as well as safety within the
site), alternative methods of reducing the overall site impedance needed to be considered.
Approximately 2 km from the solar park is an extensive underground cable network associated
with a nearby town and several smaller villages. This cable network consists of 33 kV, 11 kV and
LV cables, a large proportion of which are older lead sheathed cables that provide an earth
electrode contribution (also referred to as a ‘global earthing system’). In addition to the cable
network there are numerous 11 kV to LV distribution substations, a few primary substations and
also a large National Grid 275 kV to 132 kV to 33 kV substation (which also includes the chain
impedance contribution from several tower lines) that each have their own earthing systems
which will provide an additional contribution. Although the solar park is not directly associated
with this cable network, a rather unorthodox suggestion was made to form an interconnection
between the solar park and cable network ‘global earthing system’. Not only would the earth
wires themselves contribute towards lowering the overall impedance of the earthing system, but
the earth contribution from the cable network would further reduce the impedance and improve
safety.

Page 16-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Following numerous suggested connection points, an agreement was reached between the DNO
and Client and the final earth wire interconnection was proposed to be made to an 11 kV
switching station, approximately 2 km away from the solar park. A CDEGS model of the
combined solar park and interconnection earth wires was created in MALZ and is shown below
in Figure 9.

2500

2000

1500
Y AXIS (METERS)

1000

500

0
1000 1500 2000 2500 3000 3500
X AXIS (METERS)

Top View of Conductors

Figure 9: Conductor Plot Showing the Solar Park and Earth Wire Interconnections.

No soil resistivity data was available along the route of the interconnection, therefore to
calculate the earth impedance of the interconnection earth wires, the conductors shown in the
plot above were simulated in the solar park soil model as this was assumed to be a worst case;
the ground along the interconnection route slopes downward and was observed to contain much
more moisture, therefore would generally be at a lower resistivity than at the higher elevation of
the solar park; two earth wires would be installed along this route to lower the series impedance
of the interconnection and improve the contribution that would be ‘seen’ at the solar park.
In order to ascertain the contribution that would be provided by the global earthing system, it
was also necessary to carry out a fall-of-potential test from the 11kV switching site that was to be
utilised for the earth wire interconnection. The results of the test were interpreted using MALZ
which showed the underground cable network to have an earth resistance of 0.31 Ω.

Copyright © 2017 SES ltd. All rights reserved. Page 16-11

PART II: USER CONTRIBUTIONS

5 Conclusion
The final earthing study demonstrated that to overcome the safety issues surrounding high earth
potential rise and high touch voltages rise due to high soil resistivity, most of the options
discussed in Section 4 were needed. The earthing system consisted of earth wire loops that
surround each PV array field, inverter stations and intake substation. In addition to this the PV
array support piles were bonded to provide touch voltage control and earth wire
interconnections were made to a nearby underground cable network global earthing system.
These avoided the need for a copper mesh within the site, thus reducing installation cost.
A final earth resistance test was carried out prior to energization, which returned a value slightly
higher than calculated. Further investigation found that this was mainly due to the network
interconnection earth wires having a higher than anticipated series impedance (potentially due
to the number of joints along their length and the mutual coupling due to their close proximity
to each other). Despite this however, the earthing system at the solar park was still proven to be
safe as it was designed to cater for slight increases in earth resistance.
It is quite often the case at solar parks that, due to the speed of construction and tight deadlines,
at the time of final testing the ground surrounding the earthing system will not have had
sufficient time to consolidate (and occasionally has not been fully backfilled) therefore the final
result is usually slightly higher.
On the 30th of March 2017 (one day prior to the March 31 st tariff deadline) the site earthing
design was accepted by the DNO and successfully energised.

6 Postscript
Throughout construction of this solar park, the local DNO had concerns regarding the effect that
the high soil resistivity in the area would have on the EPR and the knock-on effect it would have
on safety; specifically for third parties not associated with the site both nearby and further afield
(due to the voltage transfer that would occur via the network interconnections). The higher post-
construction measured earth resistance, lead the DNO to include an additional clause in the
energization agreement, which would require a second earth interconnection to be made to a
different part of the nearby ‘global earthing system’. This would ensure that, should an
interconnection fail, the site would remain safe.
The proposal would see an earth connection made to a steel 33 kV tower located within the solar
park (which supports the 33 kV site supply), allowing a fault current reduction to be applied
(accounting for current return via an aerial earth wire) and tower line chain impedance
contribution to be included in the calculations, reducing the EPR further. At present, this circuit
has no aerial earth wire fitted; therefore the DNO would be required to retro-fit an earth wire to
the towers extending back to the source substation, approximately 3 km away.
At the time of writing this article, the second interconnection had not yet been installed,
however an additional study had been carried out to determine the contribution that this second
interconnection would have; no comments have yet been received from the DNO.

Page 16-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

7 Acknowledgements
The author would like to thank his colleagues Mark Davies, Mayur Gaglani and Denis Baudin for
their assistance.

8 References

[1] "CDEGS RESAP", Safe Engineering Services, Analysis of Earth Resistivity Measurements, Version
Number: 15.2.4890.0, Copyright © 1978 - 2015 by SES Ltd.
[2] "CDEGS MALZ", Safe Engineering Services, Frequency Domain Analysis of Conductor Networks,
Version Number: 15.2.4890.0, Copyright © 1978 - 2015 by SES Ltd.
[3] "ENA TS 41-24: Guidelines for the Design, Installation, Testing and Maintenance of Main Earthing
Systems in Substations", Issue1, 1992 (Amended 1999), Published by Energy Networks Association.
[4] "BS EN 50522: Earthing of Power Installations Exceeding 1 kV A.C.", 2010, Published by British
Standards Institute.
[5] "ENA ER S36: Procedure to Identify and Record 'Hot' Substations", 1988, Published by Energy
Networks Association.

Copyright © 2017 SES ltd. All rights reserved. Page 16-13

 UGM 2017 – CANNON BEACH, OREGON

17 EVALUATION OF STEP VOLTAGES ON HORSES AT THE

CURRAGH RACE COURSE
Neil Cowap and Kevin Leavey
BE CEng MIEI
Email: [email protected]; [email protected];
Web Site: www.aecom.com
Abstract

After the deaths of race horses in NEWBURY in 2011 there has been some focus on step voltages around horses.
AECOM Ireland is currently involved in the new grandstand and parade ring in the Curragh racecourse in
Ireland. AECOM Ireland has completed soil resistivity measurements on site, calculated a tolerable step voltage
for a horse (using IEC 60479-3 and MALZ) and identified areas of concern and/or mitigation for these areas.

1 Executive Summary
The new development at the Curragh racecourse is connected to the distribution network via a
three phase supply at 10 kV. This article calculates the expected GPR (Ground Potential Rise)
and step voltages for horses and evaluates if they are above the tolerable limits and recommends
mitigation for areas above tolerable limits.
The technical report IEC 60479-3 is where most of the values in the initial analysis were taken.
One short coming of this is that most of the values are for “cows” not horses.
In Section 5 the “resistances” of a cow were modelled using SESCAD and it was verified that this
analysis resulted in similar results to the spot 2d graph, when a body resistance of 405 Ω was
selected i.e. approximately 2.5 metres clearance was required from earthed steelwork.
After reviewing other publications1 a revised model of a horse was modelled in SESCAD and it
was confirmed that an approximate 2.5 metres clearance was required.
Using a fibrillation current of 0.16 Amps is likely an overly conservative value, but research
yielded no results that could be comfortably used.
In summary a design was completed to calculate the extent of step voltages for horses, using
values from IEC 60479-3. These values were revised to more accurately represent a horse and it
was confirmed that the values in IEC 60479-3 for cows are sufficiently conservative enough to
be used for horses.
CDEGS modules used for this article are SESCAD, RESAP and MALZ.

1“Total body water and ECFV measured using bioelectrical impedance analysis and indicator dilution in horses”, published in the journal of
applied Physiology http://jap.physiology.org/content/89/2/663
Copyright © 2017 SES ltd. All rights reserved. Page 17-1
PART II: USER CONTRIBUTIONS

2 Background
2.1 General

This article outlines the method used to evaluate the GPR and the Step & Touch voltages at the
new development in the Curragh Racecourse in Co. Kildare, Ireland.
The horse racing and breeding industry's economic contribution has been valued at €1 billion
per annum, for Ireland. It employs circa 14,000 people and it is Ireland's most consistently
successful international sport.
Racing is carried out at 26 racecourses, which host over 350 race meetings with prize-money of
over €55 million per annumi.
The value of a racehorse can vary significantly. One of the most expensive horses ever sold was
Fugaichi Pegasus which was sold to Coolmore Stud, Ireland in the year 2000 for a reported $70
millionii. In 2013 the Irish Oaks winner Chicquita was sold for $6.7miii in Goffs Co Kildare.
The stud fees from race horses are also significant private fees that could be between €250,000
and €500,000, depending on the achievements and pedigree of the brood mare. Based on an
average of 150 covers for each of the stud's 24 stallions, Coolmore stud was expected to generate
total fee income of over €160m during the 2013 breeding season. However, just two of the
Coolmore stallions, Sadler's Wells and his son Galileo, could account for up to half of the €160
millioniv.
The new development in the Curragh is the “creation of world class facilities, on a par with the
best anywhere in the world”. Therefore part of AECOM’s design brief was to ensure the safety of
the real stars of the sport.
The power team in Dublin typically design earth grids using CDEGS software to ensure the
calculated step and touch voltages are below the tolerable limits for humans.
IEC technical note 60479-3 1998 “Effects of currents passing through the body of livestock” was
the main standard used in the preparation of this article.

Page 17-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.1 Newbury – Horse Fatality

On the 12th February 2011 two horses, Fenix Two and Marching Song, were killed as a result of
electrocution at Newbury (UK) race course. AECOM had no known involvement in the Newbury
incident, but after reviewing news articles and publications online, the overriding consensus was
that a damaged electrical cable resulted in electrocution.

Photo 1: Horses at Newburyv.

Fenix Two was a six-year-old owned by the Irish financier JP McManus. This horse had never
been raced, but had been well bred and was worth $83,000vi.
Marching Song, aged five, came from a smaller yard and had raced four times before his sudden
death and was considered to have “great potential”.

Copyright © 2017 SES ltd. All rights reserved. Page 17-3

PART II: USER CONTRIBUTIONS

2.2 Curragh Redevelopment

Steeped in history and tradition the Curragh Racecourse is undoubtedly Ireland’s premier flat
racing venue and the spiritual home of the sport in Ireland. Considered to be one of the finest
tracks in the world every year and it is home to the Dubai Duty Free Irish Derby; a major event
for flat racing worldwide.
The new Curragh development will see the creation of world class facilities on a par with the best
anywhere in the world. A new parade ring and stable yard will ensure that the real stars of the
sport will also have first class facilities.

Photo 2: Redevelopment of the Curragh Race Course vii.

2.3 ET 103

ET103 “The National Rules for Electrical Installations exceeding 1 kV a.c. (1.5kV d.c), 1st
Edition”, was published in Ireland in 2015. This document is based on I.S. EN 61936-1, Power
installations exceeding 1 kV a.c. – Part 1: Common rules.
ET103 promotes the use of global earthing for HV developments. This results in all metallic
items in a development being connected to earth. For the Curragh racecourse, there are a
number of metallic lighting columns that are connected to the earth grid. It is therefore prudent
to assess the new development to ensure that there are no step voltages above the tolerable
limits, during an earth fault on the MV system, in areas where it is likely possible for a horse to
be.

Page 17-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3 Inputs to Model
This section details the inputs to the model.

3.1 Soil Resistivity Measurements

Photo 3: Taking soil resistivity measurements.

Using the Wenner method the soil resistivity tests below were completed on site in the Curragh:
Table 1: Soil Resistivity Test Results

Location 1 Location 2 Location 3 Location 4

Electrode Electrode Electrode Electrode

Resistance Resistance Resistance Resistance
Spacing Spacing Spacing Spacing
(Ohm- Ω ) (Ohm- Ω ) (Ohm- Ω ) (Ohm- Ω )
(m) (m) (m) (m)

0.3 40 0.3 40.1 0.3 75.3 0.3 90.7

0.5 26 0.5 27 0.5 49.2 0.5 42.9
1 17.4 1 15.14 1 47.6 1 17.7
2 9.28 2 9.17 2 10.67 2 9.85
5 - 5 5.67 5 5.74 5 5.69
7 4.11 7 4.64 7 4.58 7 4.46
10 3.09 10 3.58 10 3.22 10 3.17
15 2 15 2.33 15 2.84 15 2.81
20 1.71 20 1.36 20 1.25 20 2.27
30 1.3 30 1.11 30 1.5 30 1.32
40 1.06 40 1.1 40 1.05

Copyright © 2017 SES ltd. All rights reserved. Page 17-5

PART II: USER CONTRIBUTIONS

3.2 Fault Levels

ESBN is the local utility and they provided the fault levels for the new development, which is
summarized in the table below:
Table 2: Fault Levels for New Development

Item Value Comment

Connection Voltage 10.7 kV

In order to ensure a conservative
design a 20% safety margin was added
Max Three Phase SCL 1,390 Amps to the max three phase SCL value
supplied by the utility resulting in
1,668 Amps.
R,X Values 1.830, 2.331 Ohms
It is anticipated that the earth fault
detection on the racecourse side will
Io> 2 Amps
Earth Fault Protection operate quicker, but the ESBN backup
DT 1.5 s.
protection operation time was
selected.
Isolated neutral
Neutral Earthing
system.
Single Phase Earth Fault ESBN do not issue earth fault levels as
Level this is subject to change at any given
time due to network constraints and
Not provided
Cross Country Earth operations.
Faultviii The Max Three phase SCL of 1,390
Amps was assumed.

3.3 Horse Step Voltage Calculation

IEC 60479-3 1998 outlines the methodology for determining the equivalent impedances of an
animal and deriving an overall equivalent circuit that can be used in calculations. Various parts
of the body of an animal will have varying impedances, thus affecting flow of current. In IEC
60479-3 1998 the horse hoof impedance values are listed for cattle, since it states that “horse
hoof impedance is somewhat larger than that of cattle”, using the value for cattle will yield
conservative results.”

Page 17-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 1: Current flow and impedances of the relevant parts of the body of a cow for current path from the
nose to the legsix.

Figure 2: Diagrams for an animal, for a current path from the nose to the four legs (path A) and from the
forelegs to the hindlegs (path B)x.

The fact the horses are in contact with the ground at four points and were wearing steel shoes
meant they were more susceptible to fault current. It also requires that any value of hoof
impedance calculated must be divided by two.
The IEC also outlines the fibrillation current for various animals. The standard gives a plot of
ventricular current of sheep for a large range of fault durations. For many species the fibrillation
current is independent of body weight. Due to the long fault duration of 1.5 seconds the lowest
fibrillation current of 0.16 amps was selected.

Copyright © 2017 SES ltd. All rights reserved. Page 17-7

PART II: USER CONTRIBUTIONS

3.4 Earthing System

Figure 3: Earthing system under investigation.

The above earthing system consists of the following:

1. Buried earth conductor is represented by the green lines (50 mm² bare stranded copper),
2. Earth rods along with the Buried earth conductor (2.5 metres),
3. Insulated cable is represented by red lines,
4. Along the insulated cable the foundation for the lighting pole is represented by a
600 mm steel conductor.

4 Results
This section details the results using the input criteria above.

4.1 Step and Touch Voltages Calculations

Using IEC 60479-3 the following values were inserted into MALZ to calculate the tolerable step
voltage.

Page 17-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 3: Inputs to CDEGS model to calculate tolerable Step Voltage.

Item Value Comment

IEC 60479 section 2.3

Foot The hoof impedance is listed, for cattle. Since it is stated that horse hoof
125 Ω
resistance impedance is “somewhat larger” than that of cattle, using the value for
cattle will yield conservative results.

IEC 60479 section 4 – Table 3

Body
450 Ω It was assumed that the body resistance of cattle, to be on the same order
resistance
as that of horses.

IEC 60479 section 5.3

Table 4 details the threshold of ventricular fibrillation for a.c. 50/60 Hz
for different species of livestock, for a shock duration of 3 s.
It states that the threshold of ventricular fibrillation within one species
Fibrillation (horses and ponies are the same species) seems to be independent of the
0.16 A
current weight of the animal.
The minimum value of 0.16 amps was selected, to ensure a conservative
design. This may be further refined following discussions with the client,
and firm indications of actual fault clearance times and back up fault
clearance times, on the network.

The criteria above was entered into CDEGS and the screenshot below displays the output:

Figure 4: Calculation of Step Voltage for a Horse.

Copyright © 2017 SES ltd. All rights reserved. Page 17-9

 PART II: USER CONTRIBUTIONS

4.2 Soil Resistivity Measurements

The values in section 3.1 are entered into RESAP and the resultant soil model is shown below:

Figure 5: Soil model.

4.3 Results from CDEGS Model

Table 4: Results from CDEGS model

Item Value Comment

In order to ensure a conservative design a 20% safety

Current Injected in
1,668 Amps margin was added to the max three phase SCL value
Reference Source Bus
supplied by the utility resulting in 1,668 Amps

Impedance of Calculated by the software

0.805 Ω
Grounding System

GPR of Reference 1668*0.805

1,343 volts
Source Bus

Tolerable Step voltage Assuming a Forelegs to hind legs body resistance of

109.7 V 450 Ω
(Bare hoof to bare soil)

Page 17-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

4.3.1 Results from CDEGS Model – Step Voltages

Figure 6: Extent of Step Voltages for a Horse at the New Curragh.

The above figure illustrates step voltage above 109.7 volts in red; therefore for a distance
extending out to 2.5 metres from an earthed lighting pole there may be a step voltage issue if the
worst case scenario was considered.

Copyright © 2017 SES ltd. All rights reserved. Page 17-11

PART II: USER CONTRIBUTIONS

5 Analysis of IEC 60479-3 using MALZ

5.1 Analysis of Cow

Figure 7: Resistance for a Cow.

In the above diagram and the resistances from IEC 60479-3xi resistance represent the following
 Resistance R1: nose-trunk (75 Ω)
 Resistance R2 = R'2(450 Ω) + R4: trunk – forelegs with hoof (250 Ω)
 Resistance R3 = R'3 (250 Ω + R4: trunk – forelegs with hoof(250 Ω)
 Resistance R'2: trunk – forelegs without hoof (450 Ω)
 Resistance R'3: trunk – hind legs without hoof (250 Ω)
Using the diagram and values above we modelled a “cow” in SESCAD. For this model we
assumed all the items are insulated with the exception of the feet which are represented as a 200
mm length of standard conductor.

Page 17-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 8: Model of cow represented by resistances in SESCAD.

Figure 9: Step Voltages with detailed model of Cow.

Figure 6 above demonstrates that there are potential step voltages extending to about 2.5 metres
from any directly metallic parts. The Figure 9 more accurately models the cow with the result
being the similar to Figure 6.

Copyright © 2017 SES ltd. All rights reserved. Page 17-13

PART II: USER CONTRIBUTIONS

5.2 Analysis of a Horse

After reviewing a paper entitled “Total body water and ECFV measured using bioelectrical
impedance analysis and indicator dilution in horses”xii, in the journal of applied Physiology
there was some interesting observations. A summary of the paper is below
The purposes of the study were:
1. To determine the compartmentation of body water in horses by using indicator dilution
techniques,
2. To simultaneously measure bioelectrical impedance to current flow at impulse current
frequencies of 5 and 200 kHz.
The article mentions that, the resistance to current flow is due to the specific resistivity and the
volume of the conducting fluid or the fat-free mass of the animal. Cell membranes act as
electrical condensers and are a barrier to current flow at low frequencies (≤50 kHz).
Eight horses and ponies were selected with the following ranges in physical properties:
 Body mass spanned 119 kg (406.5–525.5 kg),

 Length ranged from 152 to 160 cm,

 Height ranged from 146.5 to 156.5 cm.

Bioelectrical impedance measured at each frequency was considerably less at the torso site than
at the legs. The variability, as represented by the standard error of triplicate measures on each
horse, was also less with torso impedance measurements than with leg impedance
measurements. There was also less inter-subject variability with torso site measures compared
with leg site measures. The results are summarized below:
 5 kHz, impedance was 44.1 ± 1.8 Ω and 211 ± 16 Ω for torso and leg sites, respectively,
 50 kHz, impedance was 33.4 ± 1.3 and 175 ± 12 Ω,
 200 kHz, 23.0 ± 1.0 and 148 ± 12 Ω ,
 500 kHz, 24.6 ± 2.8 and 132 ± 12 Ω.
Therefore for the analysis 44 Ω for torso and 211 Ω for legs and 250 Ω for the hoof resistance
were used and illustrated in the figure below:

Page 17-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 10: Resistances of Horse.

Equation 1: Calculating Resistance of a Horse.

IEC 60479-3 recommend the use of 450 Ω using a value of 505 Ω does not significantly affect
the results; with the tolerable step voltage for the horse rising from 109 V to 118 V.

Figure 11: Calculation of Step Voltage for a Horse.

Copyright © 2017 SES ltd. All rights reserved. Page 17-15
PART II: USER CONTRIBUTIONS

Plotting the revised safety limits produces a result similar to Figure 6.

Using the resistances in Figure 10 a revised resistance model was used in SESCAD.

Figure 12: Model of Horse represented by resistances in SESCAD.

This results in similar areas that would be of concern for step voltages i.e. within 2.5 metres of
an earthed piece of steelwork.

Figure 13: Step Voltages with detailed model of Horse.

Page 17-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

6 Recommendations
This report uses conservative values when modeling and is therefore a pessimistic evaluation of
the potential step voltages in the Curragh racecourse. Nevertheless, there is a requirement for
some mitigation and some other recommendations.
Possible mitigation would be:
1. Redesign the lighting and move the metallic lighting structures away from the areas of
concern
2. Installation of a high electrical resistivity surface (e.g. Tarmac)
3. Installation of quicker earth fault protection on the customer side
Other recommendations made in the report are also listed below:
1. No MV or LV cable should be routed within 3 metres of an area that will be frequented
by horses,
2. A review of the proposed stables is recommended to ascertain if there are earthed steel
items that could be within 5 metres of an area that will be frequented by horses. If so
mitigation would be required for example
a. The installation of resistive layers,
b. Restriction of access to the area.
3. A risk assessment and LPS design, if required as per IS EN 62305 “Protection against
lightning.
IEC 60479-3 recommends the use of 450 Ω for the resistance of a cow and assuming cows
and horses have similar body impedances this was used to recommend areas of concern for
step voltages for horses. Using the impedances measured in the journal of applied
Physiology document results in body impedance for the horse of 505 Ω. The conclusion
being that the use of 450 Ω is a valid approximation.

Copyright © 2017 SES ltd. All rights reserved. Page 17-17

PART II: USER CONTRIBUTIONS

7 Conclusion
In Section 5 the “resistances” of a cow were modelled using SESCAD and it was verified that this
analysis resulted in similar results to the spot 2d graph, when a body resistance of 405 Ω was
selected i.e. approximately 2.5 metres clearance was required from earthed steelwork.
After reviewing other publications2 a revised model of a horse was modelled in SESCAD and it
was confirmed that an approximate 2.5 metres clearance was required.
Using a fibrillation current of 0.16 Amps is likely an overly conservative value, but research
yielded no results that could be comfortably used.
In summary a design was completed to calculate the extent of step voltages for horses, using
values from IEC 60479-3. These values were revised to more accurately represent a horse and it
was confirmed that the values in IEC 60479-3 for cows are sufficiently conservative enough to
be used for horses.
There is a requirement for mitigation on site and some redesign work. Some of the
recommendations are:
1. Either reconfigure the network to ensure there are no cables buried under areas where
horses have regular access or install an insulating layer of tarmac above the cable.
2. Install a restricted earth fault protection and operate for earth faults within 100 ms.
3. Identify all earthed steelwork in the development and perform a risk assessment taking
into account the following:
a. Likelihood of horse coming into contact with the earthed steelwork,
b. Likelihood of the horse being in close proximity (2.5 metres) to earthed
steelwork,
c. What insulating layers are installed in close proximity (2.5 metres) to earthed
steelwork.

2“Total body water and ECFV measured using bioelectrical impedance analysis and indicator dilution in horses”, published in the journal of
applied Physiology http://jap.physiology.org/content/89/2/663
Page 17-18 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

8 References

i http://www.hri.ie/about-us/mission-statement/
ii http://www.telegraph.co.uk/business/2016/05/20/ten-of-the-most-expensive-animals-in-the-world/fusaichi-pegasus-in-action/
iii
http://www.independent.ie/sport/irish-oaks-winner-chicquita-sold-for-6m-at-goffs-29776770.html
iv http://www.independent.ie/business/irish/super-stallion-to-push-stud-fees-at-coolmore-past-160m-26329187.html

v http://www.dailymail.co.uk/news/article-1356297/2-horses-fall-dead-Newbury-parade-ring-amid-electrocution-fears.html

vi http://www.telegraph.co.uk/news/uknews/8321483/Power-cables-removed-after-horses-die-at-Newbury-racecourse.html

vii http://www.archdaily.com/473868/grimshaw-selected-to-redevelop-ireland-s-most-prestigious-racecourse

viii
The term cross-country fault is used to designate two single phase-to-earth faults that occur at two different locations on the system. A
common cross-country fault starts as a single phase-to-earth fault anywhere on a distribution feeder. The voltage on the two healthy phases
rises to a value close to the phase-to-phase voltage of the system. The increased voltage on the two healthy phases may cause damage to an
already weakened lightning arrester on the same or on another feeder. The resulting fault current is usually higher than the fault current at a
single phase-to-earth fault. The fault current at a cross-country fault is normally lower than the fault current associated with a short circuit with
zero fault resistance at the fault location closest to the feeding substation
ix Taken from IEC 60479 Figure 1
x Taken from IEC 60479 Figure 2
xiIEC 60479-3 Figure 3
xii
http://jap.physiology.org/content/89/2/663

Copyright © 2017 SES ltd. All rights reserved. Page 17-19

 UGM 2017 – CANNON BEACH, OREGON

18 CURRENTS INDUCED BY AIR-CORE REACTORS IN A

SVC SUBSTATION
Yuhan Wu, Eng.

Contributor: Nathalie Grignon, Eng., M.A.Sc.

Hydro-Québec Innovation, équipement et services partagés

Email: [email protected] , Web Site: www.hydroquebec.com

1 Background
A static VAR compensator (SVC) was installed in a 315 kV substation to increase the capacity of
the power transmission system. Its main components are a thyristor controlled reactor (TCR), a
thyristor switched capacitor (TSC) and harmonic filters.
During commissioning, two issues linked to the currents induced by the magnetic fields of TCR
reactors were identified when a hot spot was observed on outside TCR fence grounding conductor
(Figure 1 and Figure 2), and a second hot spot leading to steady electric arcing was observed on
the gate of the TCR thyristor valve room inside the building (Figure 3 and Figure 4).
These issues compromise worker safety and equipment integrity (e.g. risk of burns and fires). To
solve them, three corrective measures were implemented in succession:
1. A section of the outside TCR fence was temporarily replaced with an insulating section
(i.e. an orange plastic snow fence).
2. The building bonding network was interrupted at ten locations (Figure 3).

3. A missing bonding conductor was buried in the concrete slab along the fence of the TCR
and TSC valve rooms (Figure 3 and Figure 4).
The first corrective measure eliminated the first issue, but failed to solve the second one. The
second corrective measure was also unsuccessful in this regard. As expected, the installation of a
bonding conductor over the entire required distance (third corrective measure) proved
immediately successful in eliminating the second hot spot.
Modeling and simulation were carried out using the CDEGS HIFREQ module to study these
complex phenomena, identify any potential unreported hot spots and define permanent
corrective measures to be implemented within the substation.

Copyright © 2017 SES ltd. All rights reserved. Page 18-1

PART II: USER CONTRIBUTIONS

Figure 1: Hot spot on grounding conductor of outside TCR fence.

Figure 2: Thermography of hot spot on grounding conductor of outside TCR fence.

Page 18-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 3: Inside of building and bonding network modifications made during commissioning.

Figure 4: Fence of TCR and TSC valve rooms.

Copyright © 2017 SES ltd. All rights reserved. Page 18-3

PART II: USER CONTRIBUTIONS

2 Simulation models

2.1 Simplified reactor model

Air-core reactors comprise several parallel-connected concentric encapsulated coils consisting of
insulated conductors. They are by far the worst magnetic field generators among the equipment
used in power transmission system.
To properly model them, their basic characteristics (e.g., diameter, height, number of turns and
conductors size) must be known. However, modeling the actual electrical configurations and
mechanical structures of the reactors is inadvisable or even impossible, since manufacturers do
not typically share this type of design data, which is deemed confidential.
For this study, reactors were thus modeled using the basic engineering data provided by the
manufacturer. They were represented by a single coil with a number of turns that was estimated
using the formula below, valid for l > 0.8 R [1]:

𝐿 (9𝑅 + 10𝑙)
𝑛= √ (1)
10𝜋𝜇0 𝑅 2

where:
n = Estimated number of turns
L = Inductance (H)
R = Mean radius (m)
l = Length (m)
𝜇∘ = Magnetic permeability of air (4𝜋·10-7 H/m)

For example, the parameters used to model one of the six TCR reactors are given below:
Table 1: TCR reactor parameters.

Inductance (H) 0.00866

Mean radius (m) 1.74

Length (m) 2.7

Estimated number of turns 56

Nominal current (A) 3,000

We were able to validate the estimated number of turns in comparison with the average number
of turns (57.5) indicated in the TCR technical report provided by the manufacturer. For this study,
however, the reactors were modeled using n = 10 rather than n = 56 in order to shorten duration
of simulation. Incidentally, simulating the magnetic field generated by the six reactors (with n =
56) takes an entire day.

Page 18-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Since the magnetic field generated by the reactor is proportional to the magnetomotive force
(MMF), expressed in ampere-turns, all that required to produce an equivalent magnetic field is to
apply a scaling factor to the reactor energization current. For each TCR reactor, the scaling factor
is 5.6 since n = 10 (Figure 5).

n = 56; I = 3,000 A n = 10; I = 5.6 × 3,000 A

Figure 5: Ampere-turn equivalence of simplified TCR reactor model.

Figure 6 shows the near perfect match between the simulated magnetic flux density for a 56-turn
reactor and that for a 10-turn reactor.

Figure 6: Comparison of magnetic flux density for estimated model, simplified model and approximate
theoretical formula (Θ = 0).

Copyright © 2017 SES ltd. All rights reserved. Page 18-5

PART II: USER CONTRIBUTIONS

Figure 6 also shows that the two curves converge towards the approximate theoretical formula [3]
below:
𝜇0 ∙ 𝑛 ∙ 𝐼 ∙ 𝐷 2
|𝐵|(𝑟, 𝛩) ≈ ∙ 𝑓(𝛩) (2)
8 ∙ 𝑟3

𝑐𝑜𝑠 2 𝛩
𝑓(𝛩) = √𝑠𝑖𝑛2 𝛩 + (3)
4

where:
|B| = Magnitude of magnetic flux density (T)
n = Estimated number of turns
I = Reactor energization current (A)
D = Mean reactor diameter (m)
r = Distance from centre of reactor to calculation point (r > 3R) (m)
R = Mean reactor radius (m)
Θ = Angular coordinate of calculation point

Figure 7: Relation between magnetic field approximation variables.

2.2 Magnetic clearances

The space bounded by the magnetic contours specified by the manufacturer through the RMC1
and HMC1 parameters should contain no metal parts. Components that form closed loops should
be located outside the space defined by the RMC2 and HMC2 parameters. Figure 8 illustrates the
definitions of the magnetic clearance parameters.

Page 18-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 8: Definitions of magnetic clearance parameters.

Simulations of the reactors modeled according to the above method were able to calculate the
magnetic flux densities generated by the reactors. Contours were plotted and compared with the
magnetic clearances specified in the manufacturer’s drawings. This helped validate the reactors
models again.
Figure 9 shows the magnetic flux density contours for a TCR reactor. Table 2 compares the
manufacturer’s recommended magnetic clearances with the magnetic clearances calculated using
HIFREQ.

Figure 9: TCR reactor magnetic clearance contours.

Copyright © 2017 SES ltd. All rights reserved. Page 18-7

PART II: USER CONTRIBUTIONS

Table 2: Comparison of TCR reactor magnetic clearances recommended by manufacturer and those
obtained via simulation.

Relative
Manufacturer Simulation
Magnetic clearance difference
(m) (m)
(%)

RMC1 – 4 mT 4.19 3.3 -21

RMC2 – 2 mT 5.29 4.2 -21

HMC1 – 10 mT 2.14 1.65 -23

HMC2 – 3 mT 4.28 3.4 -21

The TCR reactor magnetic clearances obtained via simulation differ by about 20% from the
manufacturer’s recommendations. According to the manufacturer, the relative error in these
parameters can reach 25%–30% since the manufacturer’s calculations use simplified formulas.
One of the three TSC reactors was also modeled using the same method. Figure 10 shows the
magnetic flux density contours for a TSC reactor. Table 3 compares the manufacturer’s
recommended magnetic clearances with the magnetic clearances calculated using HIFREQ.

Figure 10: TSC reactor magnetic clearance contours.

Page 18-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 3: Comparison of TSC reactor magnetic clearances recommended by manufacturer and those
obtained via simulation.

Relative
Manufacturer Simulation
Magnetic clearance difference
(m) (m)
(%)

RMC1 – 3 mT 3.8 3.8

RMC2 – 2 mT 4.3 4.3

Negligible
HMC1 – 10 mT 1.86 1.88

HMC2 – 2 mT 4.2 4.1

The results thus obtained closely match the parameters provided by the manufacturer of the TSC
reactors. It can therefore be concluded that the latter uses a method allowing for a more accurate
calculation of magnetic clearances.

2.3 Model of SVC substation

The model used to calculate the currents induced by the air-core reactors includes the substation
grounding grid, the outside TCR and TSC fences, the bonding network and the metal structure of
the building, as well as the inside fences of the valve rooms (Figure 11).
The fences and the metal structure of the building are made of low-carbon (0.3%) steel. A relative
resistivity ( r) value of 13 and a relative magnetic permeability ( r) value of 1,000 were used.

Figure 11: Model of SVC substation.

Copyright © 2017 SES ltd. All rights reserved. Page 18-9

PART II: USER CONTRIBUTIONS

3 Results and solutions

The two hot spot issues appeared when the TCR was in operation while the TSC was not. Only
simulation results for this operating mode are presented in this article, though our study covered
all SVC operating modes. Our first priority was to check whether there were any other potential
unreported hot spots during commissioning.

3.1 Currents induced in outside fence and in substation grounding grid

(issue no. 1)
The table below shows the TCR reactors energization currents at the fundamental frequency and
at the third harmonic. Higher-frequency harmonics were ignored in this study. The phasing of the
reactors is shown in Figure 1.

Table 4: TCR reactors energization currents.

IA- rms (A) IB- rms (A) IC- rms (A)

f = 60 Hz 1,500  0° 1,500  -120° 1,500  120°

f = 180 Hz 500  0° 500  0° 500  0°

Figure 12 shows the induced currents circulating in the substation grounding grid at 60 Hz and
180 Hz.

Page 18-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 12: Induced currents circulating in substation grounding grid at 60 Hz and 180 Hz.

This figure shows that:

1. The currents induced by the reactors are clearly present in the grounding grid, far from
the reactors, even though the installation largely complies with the manufacturer’s
magnetic clearance recommendations.
2. The currents induced by the reactors energized at 180 Hz are proportionally much larger
than those induced at 60 Hz, since the current is in phase in all reactors. They mostly
circulate in the loops that circumscribe the reactors rather than transversally in the
conductors that connect these loops. Unsurprisingly, the maximum induced current is
found in the closest of these loops (450 A). As observed during measurements, a current
concentration point is found on a fence grounding conductor.
3. Currents induced by the reactors energized at 60 Hz are found in all the grounding meshes
located around the reactors. They can reach tens of amperes. Their circulating paths are
less predictable than at 180 Hz. The hot spot identified on the TCR fence grounding
conductor during commissioning is however clearly visible (Figure 1, issue no. 1).

Copyright © 2017 SES ltd. All rights reserved. Page 18-11

PART II: USER CONTRIBUTIONS

To permanently eliminate the hot spot no. 1, the outside TCR fence was rebuilt using sections that
are electrically isolated one from the other; each section is grounded using a single grounding
conductor. This eliminated the induced currents circulating in the grounding conductors at both
60 Hz and 180 Hz, and decreased by 50%–100% the currents circulating in the fence itself, as
shown in Figure 13. However, induced currents increased in the buried perimeter conductor of
the fence.

Figure 13: Currents induced at 60 Hz in outside TCR fence built using non-insulated sections, and rebuilt
using insulated sections.

The currents induced in the front portion of the fence did not change, since it comprises three
gates that are intrinsically isolated one from the other.

3.2 Currents induced in inside fence and in building bonding network

(issue no. 2)
To eliminate the hot spot found on the gate of the TCR valve room fence, the building bonding
network was interrupted at ten locations (second corrective measure, shown in Figure 3). The
ineffectiveness of this measure was demonstrated by simulations and measurements of the
currents circulating in the bonding network before and after this implementation.
The integral bonding network helped to decrease currents circulating in the inside fence gate. It
has therefore been proposed to reclose the ten points of interruption, add a conductor missing
along the valve room gate and add extra conductors between various rooms in the building.
Induced currents would then preferentially and more uniformly circulate within the bonding
network.

Page 18-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 14 shows the currents induced at 180 Hz circulating in the building bonding network and
along the gate of the TCR valve room fence.

Figure 14: Currents induced at 180 Hz circulating in building bonding network and along gate of TCR
valve room fence.

The addition of the missing conductor effectively eliminated the hot spot detected at the gate, as
observed through in situ measurements.

3.3 Reactor layout

The reactors can be installed side-by-side or in a triangular configuration. Figure 15 shows the
currents induced in a simple loop circumscribing the reactors when they are energized at 60 Hz,
for both layouts. The same center-to-center (X) distances and the same clearances between the
reactors and the loop (Y) were used in both cases.

Copyright © 2017 SES ltd. All rights reserved. Page 18-13

PART II: USER CONTRIBUTIONS

Figure 15: Currents induced at 60 Hz by reactors installed side-by-side and in a triangular configuration.

The currents induced in the loop are considerably lower when the reactors form a triangle. This
can be explicated by the cancellation of the magnetic fields due to the balanced three phases.
However, the triangular layout does not present any advantage with respect to the currents
induced by the third harmonic (i.e. at 180 Hz).

4 Lessons learned
In light of the project and the study, the following recommendations can be formulated for similar
projects in the future.

Page 18-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

4.1 Reducing induced currents at the source

In section 3.1, it was noted that the TCR reactor fence, the conductor buried on its periphery and
related grounding conductors are subjected to the maximum induced current, even though the
manufacturer’s magnetic clearance requirements are met. In addition, since the metal fence
comprises multiple elements that are mechanically joined, hot spots may occur at connection
points where there is a poor electrical contact.
Since the fence is the facility element that is most exposed to the magnetic fields generated by the
reactors, it has been rebuilt using multiple insulated sections, which required strict attention to
its design and installation. As shown in Figure 13, the currents induced in the buried peripheral
conductor used to ground the fence increased as a result. In future projects of a similar nature, it
would therefore be advisable to require the installation of a nonmetallic fence, which would be
intrinsically immune to induced currents and not require any grounding. The distance between
the first ground conductor loop and the reactors it surrounds would consequently increase. This
ground loop less exposed to the magnetic field would then, in turn, be the first installation
withstanding the maximum induced current, whose intensity would be lower (200 A), as shown
in Figure 16.
Longitudinal Current Flowing in Origin of Conductor. Magnitude (A) [ID:Alstom - Inductance- @ f=180.0000 Hz ]

Figure 16: Currents induced by TCR reactors at 180 Hz with nonmetallic fence.

Another way to decrease currents induced at 60 Hz at their source would consist in specifying a
triangular layout of the reactors, as shown in Figure 15.

4.2 Importance of integrity of building bonding network and grounding

grid
In section 3.2, it was shown that the integrity of the bonding network helps to reduce the induced
currents circulating in the inside fence of the valve room. The third corrective measure
implemented during the commissioning phase proved its importance. In fact, bonding network
conductors short-circuit grounded metal elements, which include not only the fence, but also the
metal structure of the building, control cable sheaths, etc. In this way, undesirable induced
currents are confined to the bonding network, thus reducing the risk of fire due to overheating or

Copyright © 2017 SES ltd. All rights reserved. Page 18-15

PART II: USER CONTRIBUTIONS

a poor contact between two metal components, as well as the risk of equipment failure or
malfunction.
It is much more advisable to maintain the integrity of the building bonding network and
grounding grid than to break them up in order to eliminate all current loops. The latter measure
compromises their effectiveness and integrity, and risks concentrating induced currents in the
remaining loops, where these currents could end up being more harmful [2].

5 Limitations
The simulations presented herein provide a reasonable estimate of the relative induced current in
the installation. However, the results obtained cannot be expected to be very accurate in absolute
terms for several reasons, including those listed below:
1. The model does not include all installation components. For instance, its grounding
system is not limited to the substation grounding grid, fences, bonding network and metal
structure of building. It also includes the neutral conductors of power cables and the
sheaths of control and protection cables. The presence of grounded equipment support
structures could also have an impact on the results due to induction and conduction effects
ignored in this study.
2. The metal structure of the building and fences were represented using simplified models.
3. The relative magnetic permeability value used for the steel fences and the building metal
structure has a significant impact on the induced currents obtained, since it determines
the impedance of the conductors that represent these elements. In fact, it varies according
to the types of steel and the intensity of the magnetic field produced by the reactors.

6 Conclusions
The HIFREQ module was used to study effects of the magnetic fields produced by air-core reactors
in a SVC substation.
Each reactor was modeled using a single coil with an equivalent number of ampere-turns to
shorten duration of simulation.
The simulations demonstrated that induced currents with significant intensity were present in the
substation grounding system well outside the magnetic clearances recommended by the
manufacturers, as observed in situ. It should be noted that these clearances are specified for a
single reactor for a nominal 60 Hz current.
The currents induced by the reactors energized at 180 Hz are proportionally much larger than
those induced at 60 Hz, since the current is in phase in all reactors. Consequently, particular care
must be paid to the harmonic content and overall effect of the reactors.
It is recommended to use a nonmetallic fence to restrict access to the air-core reactors, and to
connect all other metal fences to a parallel buried conductor.
Another way to decrease the currents induced at 60 Hz at the source would consist in specifying
a triangular reactor layout.
Contrary to the design philosophy calling for all loops to be broken, this article highlights the
importance of maintaining the integrity of the building bonding network and grounding grid

Page 18-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

outside the reactors fences. This ensures that any unwanted induced current will be confined to
the grounding grid and bonding network, thus decreasing the risk of fire due to overheating or a
poor contact between two metal components, as well as the risk of equipment failure or
malfunction.
This study is subject to several limitations due to the simplified model of the installation and the
assumptions that were made for the electrical properties of materials. Nonetheless, the results
obtained help to achieve a better understanding of the phenomena arising from induced currents
produced by air-core reactors in their environment, and to better design the measures required
to mitigate their effects.

7 Acknowledgements
The authors wish to thank their colleagues Michaël Germain and Jean-Mathieu Martin for their
constant support and diligence during this study; they are also grateful to Michel Bourdages for
his insightful suggestions and feedback.

8 References
[1] Marc T. Thompson, Inductance Calculation Techniques, Power Control and Intelligent Motion,
Vol. 25, No. 12, December 1999.
[2] Ralph Marrison, Warren H. Lewis, Grounding and bonding, Toronto, John Wiley & Sons, 1990,
p. 190–192.
[3] W.R. Smythe, Static and Dynamic Electricity, McGraw-Hill, 1968.

Copyright © 2017 SES ltd. All rights reserved. Page 18-17

Part III: SES Contributions
 UGM 2017 – CANNON BEACH, OREGON

19 SHORT CIRCUIT COMPUTATIONS USING LUMPED

IMPEDANCES IN MULTIGROUNDZ AND MULTIFIELDS
SOFTWARE PACKAGES
Octavio Ramos-Leaños and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
Handling lumped impedances in MultiGroundZ and MultiFields software packages is a capability that has been
around for a very long time. This capability has not been fully exploited in the past. In this article we will show how
this capability, combined with a few specifications about the soil model and the type of conductors, allows us to use
the software packages as a circuit solver. This introduces interesting possibilities when quick accurate solutions
are needed and conventional power system software packages or general purpose circuit solvers such as SPICE are
not available or are unfamiliar to the user. This could be quite handy, particularly when the network is too large for
approximate hand calculations. To demonstrate these useful complementary capabilities a complete short circuit
study in an industrial network is performed.

1 Introduction
Not all users of SES CDEGS suite of software packages are aware that lumped elements
(resistances, inductances and capacitances) can be inserted at the center of any conductor of the
MultiGroundZ and MultiFields software packages. This is a capability that has been around for a
very long time. This capability has not been fully exploited in the past. In this article we will show
how this capability, combined with a few specifications about the soil model and the type of
conductors, allows us to use the software packages as circuit solvers. This introduces interesting
possibilities when quick accurate solutions are needed and conventional power system software
packages or general purpose circuit solvers such as SPICE are not available or are unfamiliar to
the user. This could be quite handy, particularly when the network is too large for approximate
hand calculations.
To demonstrate these useful complementary capabilities a complete short circuit analysis of a
manufacturing facility whit several voltage levels is presented in this article.

2 Short Circuit Calculations

The calculation of precise asymmetrical short circuit values is a complex task. Consequently,
simplified methods have been developed which provide short circuit currents to match the
assigned ratings of various system protective devices and equipment. The value of the
symmetrical components of the short circuit current are determined through the use of
appropriate impedances in the basic equation I=E/Z where E is the system driving voltage and Z
is the equivalent Thevenin system impedance of the power system as seen from the point where
the short circuit occurs back to the sources of the short circuit current. The value of this equivalent
impedance is in part determined using the basis rating of the device or equipment under
consideration [1].
Knowledge of the available maximum short circuit current is essential in any grounding study.
Much of the short circuit calculation involves the representation of the proper system impedances
from the point of short circuit back to the short circuit sources. After this representation is
accomplished, the actual short circuit computation is straightforward. The steps to carry out a
short circuit calculation can be summarized as follows [1]:
Copyright © 2017 SES ltd. All rights reserved. Page 19-1
PART III: SES CONTRIBUTIONS

1. Build the System One-Line Diagram. Include all significant system components.
2. Decide on short-circuit locations and type of short-circuit current calculations. Consider
the variation of system operating conditions required to display the most severe duties.
Assign bus numbers or suitable identification for the short circuit locations.
3. Prepare a complete impedance diagram.
4. Solve the impedance network and calculate the required symmetrical component currents
for the desired short circuit locations and system conditions.
The most severe fault duty usually occurs when the maximum number of generating units and
motors are in operation and all interconnections are closed [1].

3 The Per-Unit System

The per-unit value of any quantity is the ratio of its actual value in any unit and the reference
base value in the same unit. The per-unit value is dimensionless. The per-unit system is
somewhat equivalent to indirectly scaling the parameters of the network to describe the flow of
energy along a network (instead of currents and voltages), alleviating the need to display
discontinuous values of currents and voltages across transformers. Typically the Per Unit System
provides units for power, voltage, and current. Any two of these are independent and can be
arbitrarily selected as base values, usually power and voltage. All quantities are specified as
multiples of selected base values.
One of the main advantages of using the Per Unit System is the narrow range exhibited by the
parameters of rotating electrical machines and transformers irrespective of their ratings when
expressed in per unit of their ratings.
Short circuit calculations can be made with impedances represented as per-unit values based on
a common voltage or as actual ohm values based on applicable voltages. Both representations will
give identical results. However, the per-unit system is ideal for studying multi-voltage systems.
Most of the components included in high voltage networks are often given in per-unit of reference
base quantities and further conversion is not required [1].
In the per-unit system there are usually four base reference quantities, namely, the base power
(kVA or MVA), the base voltage (V or kV), the base current (A or kA) and the base impedance
(ohms). When any two of the four reference bases have assigned values, the other two bases can
be derived from the first two values. It is common practice to select the power and voltage bases
as the reference ones for most studies. Base current and base impedance are then derived for each
of the voltage levels in the system [1]. The basic per-unit relationship is:

Page 19-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

𝐴𝑐𝑡𝑢𝑎𝑙 𝑃𝑜𝑤𝑒𝑟
𝑃𝑒𝑟 𝑈𝑛𝑖𝑡 𝑃𝑜𝑤𝑒𝑟 =
𝐵𝑎𝑠𝑒 𝑃𝑜𝑤𝑒𝑟
𝐴𝑐𝑡𝑢𝑎𝑙 𝑉𝑜𝑙𝑡𝑎𝑔𝑒
𝑃𝑒𝑟 𝑈𝑛𝑖𝑡 𝑉𝑜𝑙𝑡𝑎𝑔𝑒 =
𝐵𝑎𝑠𝑒 𝑉𝑜𝑙𝑡𝑎𝑔𝑒
𝐴𝑐𝑡𝑢𝑎𝑙 𝐶𝑢𝑟𝑟𝑒𝑛𝑡
𝑃𝑒𝑟 𝑈𝑛𝑖𝑡 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 =
𝐵𝑎𝑠𝑒 𝐶𝑢𝑟𝑟𝑒𝑛𝑡
𝐴𝑐𝑡𝑢𝑎𝑙 𝐼𝑚𝑝𝑒𝑑𝑎𝑛𝑐𝑒
𝑃𝑒𝑟 𝑈𝑛𝑖𝑡 𝐼𝑚𝑝𝑒𝑑𝑎𝑛𝑐𝑒 =
𝐵𝑎𝑠𝑒 𝐼𝑚𝑝𝑒𝑑𝑎𝑛𝑐𝑒
For three-phase systems the reference values are the Line - Line voltage and the three-phase
power. The base values are therefore:
𝐵𝑎𝑠𝑒 𝑃𝑜𝑤𝑒𝑟
𝐵𝑎𝑠𝑒 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 =
√3[𝐵𝑎𝑠𝑒 𝑉𝑜𝑙𝑡𝑎𝑔𝑒]
(𝐵𝑎𝑠𝑒 𝑉𝑜𝑙𝑡𝑎𝑔𝑒)2 𝐵𝑎𝑠𝑒 𝑉𝑜𝑙𝑡𝑎𝑔𝑒
𝐵𝑎𝑠𝑒 𝐼𝑚𝑝𝑒𝑑𝑎𝑛𝑐𝑒 = =
𝐵𝑎𝑠𝑒 𝑃𝑜𝑤𝑒𝑟 √3[𝐵𝑎𝑠𝑒 𝐶𝑢𝑟𝑟𝑒𝑛𝑡]
As an example consider the circuit depicted in Figure 1 as adapted from [1]. This figure contains
the basic information that identifies the various electric components of the system and how they
are interconnected. The diagram also includes the necessary data i.e., voltages, available short-
circuits and X/R ratios from the utility system. Other data include power ratings, connection
impedances and X/R ratios for transformers T1, T2 and T3. Similar data such as reactance, X/R
ratio and horse power are available for motors M1, M2, M3 and M4. Finally, the type, length and
impedance of cables are also provided.

Utility
115 kV
1500 MVA
Bus1 X/R=15

T1,15 MVA
Y SAT

115/13.8 kV
Z=7%, X/R=20 10 ohm
Bus2

100 ft Cu cable1 300 ft Cu cable2 200 ft Cu cable3

LCC

LCC

LCC

1-3/C-2/0 1-3/C-4/0 1-3/C-#4 F1

Z=0.0977+J0.0385 Z=0.0614+J0.0359 Z=0.3114+J0.0472
ohms/1000 ft ohms/1000 ft ohms/1000 ft
Bus4 Bus6
Bus3
T2,3.75 MVA T3,1.5 MVA
Y SAT

Y SAT

Torque 13.8/4.16 kV 13.8/0.48 kV

IM
WI

M1 Z=5.5%, X/R=11 Z=5.75%, X/R=6.5

6 ohm Bus7
Bus5

1.0 pf syn Torque Torque Torque

250 ft Cu cable4
LCC

3-1/C-250 MCM
IM

IM

IM
WI

WI

WI

4000 hp F2
1800 rpm Z=0.0534+J0.0428 F3
X''d = 15% M2 M3 M4 ohms/1000 ft
X/R = 28.9 Bus8
Ind Ind 1500 hp
500 hp 2000 hp 4*100 hp
F4
1800 rpm 3600 rpm 8*50 hp
X''d = 16.7% X''d = 16.7% 28*25 hp
X/R = 19.3 X/R = 30 X''d = 16.7%

Figure 1: A single line and impedance diagram of an industrial power system.

Copyright © 2017 SES ltd. All rights reserved. Page 19-3

PART III: SES CONTRIBUTIONS

The system short circuit currents are to be calculated at buses 2, 5, 7 and 8, i.e., at fault locations
F1, F2, F3 and F4. In this example only the first cycle fault current will be calculated to determine
the momentary fault currents. Maximum short-circuit currents are to be computed. Therefore,
three-phase bolted short-circuit values will be calculated since line-to-ground fault calculations
are limited by the transformer neutral grounding resistors. Note that the most severe fault duty
will occur when all breakers are closed, the power utility is connected and all motors are in
operation.

4 Single Line Diagram and Impedance Diagram

The single line diagram of Figure 1 shows the arrangement of all relevant components of the
system with clear identification of the characteristics of each component. The per-unit system is
an appropriate choice for carrying out short-circuit analysis of this system because of the presence
of several voltage levels. A base power of 15 MVA is chosen as a reference for the entire system.
The assigned base voltages are the nominal system voltages of 13.8, 4.16 and 0.48 kV. The base
current and base impedance for each system voltage levels can then be derived as shown in
Table 1.
Table 1: System reference base quantities.

Assigned Values Derived Values

Base Power Base Voltage Base Current Base Impedance

(MVA) (kV) (A) (ohms)

15 13.8 628 12.7

15 4.16 2,084 1.1539

15 0.48 18,064 0.0153

The impedance representation of each component in the circuit is computed in Table 2. Once the
impedance value is obtained, the resistance (𝑅), reactance (𝑋) and inductance (𝐿) values are
derived using either the computed angle or the provided X/R ratio as follows:
𝑅𝑝𝑢 = 𝑐𝑜𝑠(𝜃)|𝑍𝑝𝑢 |
𝑋𝑝𝑢 = 𝑠𝑖𝑛(𝜃)|𝑍𝑝𝑢 |
𝐿𝑝𝑢 = 𝑋𝑝𝑢 /(2𝜋𝑓)
𝜃° = atan(𝑋/𝑅)

Page 19-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 2: Components per unit impedance.

Component Computations R (pu) X(pu) L(pu)

𝑆𝑏𝑎𝑠𝑒 15
𝑍𝑝𝑢 = = = 0.01
Utility 𝑆 1500 0.0007 0.01 2.65252E-05
𝜃 = atan(15) = 86.19°
𝑍% 𝑆 7(15)
𝑍𝑝𝑢 = = = 0.07
T1 100𝑆𝑏𝑎𝑠𝑒 100(15) 0.0035 0.0699 0.000185411
𝜃 = atan(20) = 87.14°
𝑍𝑝𝑢𝑙 𝑆𝑏𝑎𝑠𝑒
𝑍𝑝𝑢 = 2 𝑙
1000𝑉𝑏𝑎𝑠𝑒
C1 (0.0977 + 𝑗0.0385)(15 × 106 ) 0.0008 0.0003 7.95756E-07
= 100
1000(13.8 × 103 )2
= 0.0008 + 𝑗0003
′′
𝑋𝑑% 𝑆𝑏𝑎𝑠𝑒
𝑋𝑝𝑢 = =
100𝑆
M1 15(15 × 106 ) 0.0251 0.703 0.001864721
= 0.703
100(4000)(746)
𝑅 = 𝑋⁄28 = 0.0251

C2 See C1 above 0.0014 0.0008 2.12202E-06

T2 See T1 above 0.0199 0.2191 0.000581167

M2 Use M1 above times 1.2(R+jX) 0.3331 6.3284 0.016786207

M3 See M1 above 0.0449 1.3917 0.003691512

C3 See C1 above 0.0049 0.0007 1.85676E-06

T3 See T1 above 0.0874 0.5683 0.001507427

M4 (100 hp) Use M1 above then divide by 4 1.0114 8.3947 0.022267109

M4 (50 hp) Use M1 above then divide by 8 1.5263 8.3947 0.022267109

M4 (25 hp) Use M1 above then divide by 28 1.2623 4.797 0.012724138

C4 Use C1 above 0.8691 0.6966 0.001847745

Copyright © 2017 SES ltd. All rights reserved. Page 19-5

PART III: SES CONTRIBUTIONS

Once all per unit impedance values have been computed, an impedance network needs to be
created and solved to obtain the short circuit values. The impedance diagram must take into
account all short circuit contributions and can be created considering that the system is driven by
a unit voltage source connected in parallel to the circuit as in figure 2.

Ut M1 M2 M3 M4-100hp M4-50hp M4-25hp

0.0007 0.0251 0.3331 0.0449 1.0114 1.5263 1.2623
+j0.01 +j0.703 +j6.3284 +j1.3917 +j8.3947 +j8.3947 +j4.797

T1 C1
+ 0.0035 0.0008 T2 C4
Driving F2 T3
+j0.0699 +j0.0003 0.0199 0..8691
Voltage 0.0874
+j0.2191 +j0.6966
F3 +j0.5683
C2 F4
C3
0.0014
+j0.0008 0.0049
+j0.0007

F1

Figure 2: Per unit circuit impedance diagram.

5 Three-Phase Fault Current Calculations

Computations of three-phase fault currents are straightforward, they only require the use of the
per unit impedance network which is based on the positive sequence network since we are dealing
with three-phase faults. If impedance information is provided in symmetrical components
quantities, only the positive sequence values must be used to compute three-phase fault currents
given that they represent the appropriate values. Since a conventional three-phase fault is
considered to be perfectly balanced, no negative or zero-sequence currents are needed to compute
the fault currents. In other words, the three-phase fault current is computed as follows:
𝐸
𝐼3𝑝ℎ = 𝐼1 =
𝑍
Where
𝐼3𝑝ℎ , Three-phase fault current.
𝐼1 , Positive component fault current.
𝐸, Circuit driving voltage.
𝑍, Total positive sequence impedance as seen from the fault location including fault impedance.

5.1 MALZ/HIFREQ as a circuit solver

The per unit impedance diagram that was built can be modeled in the MALZ or HIFREQ
computation modules for each short circuit scenario considered. Giving that MALZ or HIFREQ
are not typical circuit solvers, some leakage currents will exist in all conductor segments used to
represent the network. In order to reduce the effects of current leakage the soil must be modeled
as Uniform with a very high resistivity (1014 ohm-m) in this case. Another option is to build the
circuit in air (for HIFREQ) or to insulate the conductors. Circuit impedance values can be defined
in the Conductor Type window. In this window User-Defined Impedance Specification are
selected for all segments. The Relative Resistivity and Relative Permeability are set to 1e-3
for all segments in order to model quasi-perfect conductors. Impedance values computed in Table
2 are introduced in the Load Resistance and Load Inductance columns respectively.
Meaningful names can be given to all segments in the Identification column. These steps are
shown in figures 3 and 4. Note that it is necessary to use relatively short conductors to keep the
Page 19-6 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

unavoidable computed external impedance value small compared to the impedance inserted
(loaded) in the conductor center.

Figure 3: Set Relative Resistance and Permeability to small values.

Figure 4: Assign computed resistances and inductances in per unit.

Create the per unit circuit impedance in MALZ or HIREQ using small segments, no longer than
0.5m preferably, as in Figure 5.

Copyright © 2017 SES ltd. All rights reserved. Page 19-7

PART III: SES CONTRIBUTIONS

Figure 5: MALZ or HIFREQ circuit.

The driving voltage source is a voltage Energization Type with a magnitude of 1,000 (0 degrees
angle), instead of a 1.0 per unit voltage to increase the accuracy of the displayed results. We simply
create a connection between the faulted point and the voltage source to compute the three-phase
fault currents at any point on the circuit. As an example consider fault location at F3 as shown in
figure 6.

Figure 6: Fault simulation at location F3.

After running MALZ, requesting the plot option of a current-leaving-origin-of-conductor in
the Examine module, the current at the fault location is 1,983 p.u. as shown in Figure 7. This
value should be interpreted as 1.983 p.u. since we multiplied the source magnitude by 1,000.
Recall that F3 is located at the 480 V voltage level of the system where the base current is 18,064 A,
therefore, the short circuit current at this fault location is 18,064*1.983 = 35,821 A.
Note: Simulation Files for all cases presented in this article are available at:
“UGC 2017\Example Files\Short Circuit Computations Using MALZ-HIFREQ”

Page 19-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 7: Fault current (scaled up by 103) at location F3.

If the same procedure is applied at F1, a current of 14.991 p.u.*103 is obtained which translates to
a short circuit current of 9,414.34 A as shown in Figure 8.

Figure 8: Fault current (scaled up by 103) at location F1.

All fault locations can be computed in the same way. Fault current values for fault locations F1,
F2, F3 and F4 obtained with MALZ or HIFREQ using the per-unit approach are compared with
the values obtained from a well-known commercial circuit solver in Table 3. Note that the results
can also be compared to hand calculation results by reducing the circuit into a simple equivalent
impedance value. Table 3 shows that results obtained with MALZ and HIFREQ exhibit a relative
error of less than 1% for all four cases presented here.

Copyright © 2017 SES ltd. All rights reserved. Page 19-9

PART III: SES CONTRIBUTIONS

Table 3: Fault current comparisons with commercial circuit solver.

Fault Current (A)

Fault
Error
Location
MALZ Circuit Solver

F1 9,414.34 9,420.00 0.06%

F2 8,998.71 8,986.20 0.14%

F3 35,821.00 35,911.23 0.25%

F4 11,850.00 11,850.00 0%

6 Single-phase to ground fault current computations

The above sections introduced the procedure to carry out three-phase fault current computations
using MALZ or HIFREQ and the per-unit system. In order to compute single-phase-to-ground
fault currents it is necessary to formally introduce the concept of symmetrical components.

6.1 Symmetrical components

The method of symmetrical components is used to simplify fault analysis by converting a three-
phase unbalanced system into two sets of balanced phasors and a set of single-phase phasors, or
symmetrical components. These sets of phasors are called the positive-, negative-, and zero-
sequence components. These components allow for the simple analysis of power systems under
faulted or other unbalanced conditions. Once the system is solved in the symmetrical component
domain, the results can be transformed back to the phase domain. Symmetrical-domain currents
are related to phase-domain ones by means of the following expressions [2]:
𝐼𝑜 1 1 1 1 𝐼𝐴
[𝐼1 ] = [1 𝛼 𝛼 2 ] [𝐼𝐵 ] (1)
3
𝐼2 1 𝛼2 𝛼 𝐼𝐶
with 𝛼 = 1∠120∘ . Resulting in
1
𝐼𝑜 = (𝐼𝐴 + 𝐼𝐵 + 𝐼𝐶 )
3
1
𝐼1 = (𝐼𝐴 + 𝛼𝐼𝐵 + 𝛼 2 𝐼𝐶 ) (2)
3
1
𝐼2 = (𝐼𝐴 + 𝛼 2 𝐼𝐵 + 𝛼𝐼𝐶 ).
3
Likewise a set of symmetrical components can be converted to phase quantities by means of [2]:
𝐼𝐴 1 1 1 𝐼𝑜
[𝐼𝐵 ] = [1 𝛼 2 𝛼 ] [𝐼1 ]. (3)
𝐼𝐶 1 𝛼 𝛼 2 𝐼2
Resulting in the following equations:
𝐼𝐴 = 𝐼0 + 𝐼1 + 𝐼2
𝐼𝐵 = 𝐼0 + 𝛼 2 𝐼1 + 𝛼𝐼2 (4)
𝐼𝐶 = 𝐼0 + 𝛼𝐼1 + 𝛼 2 𝐼2 .

Page 19-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

These conversions are valid for an A-phase base, which can be used for A-phase-to-ground,
B-phase-to-C-phase, B-phase-to-C-phase-to-ground, and three-phase faults. These conversions
are also only valid for an ABC system phase rotation. If an ACB phase rotation needs to be
considered then the required expression is rewritten as [2]:
𝐼𝑜 1 1 1 1 𝐼𝐴
[𝐼1 ] = [1 𝛼 𝛼 2 ] [𝐼𝐶 ].
3
𝐼2 1 𝛼2 𝛼 𝐼𝐵

6.2 Positive-, Negative-, and Zero-Sequence Impedance Diagrams

In order to compute single-line-to-ground fault currents we need to create impedance circuits for
each of the sequence components. As mentioned in Section 5 the positive-sequence impedance
diagram corresponds to the one created for the three-phase fault case. Seldom electrical
equipment presents different negative-sequence impedance values to those of the positive
sequence. Thus, positive-sequence impedance circuits can be used as negative-sequence values,
unless any of the impedance values really presents a considerable difference, in which case
corresponding values need to be replaced by the correct ones [2].
Zero-sequence impedance circuits are more difficult to produce. In general most power devices
will present different zero-sequence impedance values than their positive-sequence impedance
ones, creating the need to completely replace the impedance values of the network. Also,
transformers need to be represented in a different way and their representation is dependent on
the type of transformer connection. Figure 9 shows some common transformer connections and
their equivalent zero-sequence representations [2].

Figure 9: Zero-sequence circuits for various transformer connections.

For the case at hand we will consider that negative-sequence impedance values are equal to the
positive-sequence ones. The zero-sequence impedance values are shown in Table 4. The
associated values in p.u. are provided in the same table.

Copyright © 2017 SES ltd. All rights reserved. Page 19-11

PART III: SES CONTRIBUTIONS

Table 4: Components zero-sequence impedance values.

Component Ro (ohms) Xo(ohms) Ro (p.u.) Xo(p.u.) Lo(p.u.)

Utility 0.01333 0.1905 0.00105 0.015 3.979E-05

T1 0.04 0.7989 0.00315 0.062906 1.669E-04

ZnT1 10.0 0.0 0.7874 0.0 0.0

C1 0.0305 0.01143 0.0024 0.0009 2.387E-06

M1 0.31877 8.921 0.0251 0.702441 1.863E-03

C2 0.1842 0.03048 0.0042 0.0024 6.366E-06

T2 0.02067 0.2275 0.017913 0.197157 5.230E-04

ZnT2 6.0 0.0 5.2 0.0 0.0

M2 0.3844 7.3023 0.333131 6.328365 1.679E-02

M3 0.0518 1.6059 0.044891 1.391715 3.692E-03

C3 0.1867 0.02667 0.0147 0.0021 5.570E-06

T3 0.0012 0.00783 0.07866 0.5115 1.357E-03

M4 (100 hp) 0.01547 0.1284 1.011111 8.392157 2.226E-02

M4 (50 hp) 0.02335 0.1284 1.526144 8.392157 2.226E-02

M4 (25 hp) 0.01931 0.0734 1.262092 4.797386 1.273E-02

C4 0.0399 0.032 2.607843 2.091503 5.548E-03

The zero-sequence impedance diagram is shown in Figure 10. Notice the inclusion of ground
impedances on the wye sides and open circuits on the delta sides of the transformers.

Page 19-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 10: Zero-sequence circuit impedance diagram.

The MALZ or HIFREQ zero sequence circuit is created in the same way as the positive sequence
circuit that was created in section 5.

Figure 11: MALZ or HIFREQ zero-sequence circuit impedance diagram.

6.3 Sequence Impedance Diagrams Interconnections

Now that the three impedance diagrams are obtained what follows is to connect them in the
correct order to produce single-phase-to-ground short circuit values. To illustrate the connection
principle the following simple circuit is introduced.

L1
G1 T1 T2 G2
BCT LCC BCT

Y Y

Figure 12: Simplified circuit.

Sequence impedance circuits for this network are as shown in Figure 13 [2].

Copyright © 2017 SES ltd. All rights reserved. Page 19-13

PART III: SES CONTRIBUTIONS

Figure 13: Sequence-impedance circuits for the simplified network, a) positive, b) negative and c) zero-
sequence circuits.

If we would like to simulate a phase-to-ground fault at the bus located between T2 and G2 of this
simple network, the required sequence impedance interconnections are shown in Figure 14 [2]:

Figure 14: Sequence-impedance diagrams interconnection for a phase-to-ground fault.

For a three-phase fault the required sequence impedance interconnections are reduced to those
shown in Figure 15 [2]:

Page 19-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 15: Sequence-impedance diagrams interconnection for a three-phase fault.

In the case of a B to C phase fault, the interconnections are as shown in Figure 16 [2]:

Figure 16: Sequence-impedance diagrams interconnection for a phase-to-phase fault.

Finally if a B to C to ground fault is required, then the interconnections need to be as in Figure 17
[2].

Figure 17: Sequence-impedance diagrams interconnection for a phase-to-phase-to-ground fault.

Copyright © 2017 SES ltd. All rights reserved. Page 19-15

PART III: SES CONTRIBUTIONS

6.4 MALZ or HIFREQ circuit impedance interconnections

A MALZ or HIFREQ circuit impedance representing a phase-to-ground fault is presented in
Figure 18. To compute a phase to ground fault you need to proceed as in the three-phase case by
connecting the driving source to the fault point in the zero-sequence network. Figure 18 shows
the results obtained for a fault placed at location F4. The figure shows a value of 0.206*103 p.u.
since the source p.u. voltage is multiplied by a 1,000.
Recall that F4 is located at the 480 V voltage level of the system where the base current is 18,064 A,
therefore, the short circuit current at this faultLongitudinal
location is inIa
Current Flowing =of Conductor.
Origin 3Io = 18,064*0.206*3
Magnitude (A) [ID: @ f=60.0000 Hz ] = 11,163.55 A.
206 68.753.2 51.7 45.2 33.0 20.9

12
1.4

.2

.9
12.1
6.6
137

20
15.6
206

206

206
161
8.0
137

15.6

8.0

161

206
137 153 161
206

137 153 154 161 173 185

12.1
12.
1.4

9
6.6

2
20.
137

15.6

206

161
8.0
206

137

15.6

206
8.0

161

137 153 161

0.0005437.0 132 132 134 153 173 206
19
206

1.0

.4
.1
19.2
4.6
0.00054

33
124
37.0

206
206
134
5.6
124
37.0
206

5.6

134
0
0

206

37.0 132 134134 134

206 206

Figure 18: Phase-to-ground fault current (scaled by 103) for a fault at location F4.

Notice that the simulation of phase-to-ground faults requires the correct interconnection of the
faulted equipment in all three impedance networks. For a fault at F1 the connections shown in

Page 19-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 19 are required resulting in a phase-to-ground fault current of 1.512 p.u. or Ia= 3Io = 1.512*
Longitudinal Current Flowing in Origin of Conductor. Magnitude (A) [ID: @ f=60.0000 Hz ]
3*628 = 2,848.6 A
1512 252 109 95.6 35.2 25.7 16.2

13.3

9.5

.2
60.7

9.4
1260
1512

16
143
0.000031

35.2
74.0
1512

1260

143

74.0

35.2
1260 1403 1477 35.2

1512
12601403 14171477
35.225.7 16.2

1512

13.3

9.5
9.4
2
60.7

16.
1260

143
0.000085

35.2
74.0
1512

1260

143

74.0

35.2
1260 1403 1477 35.2

1512
1512

0.000433401214 1217 1235 205 1300.0013

76
9.2

0
.0
75.4
13
42.0
0.00043

1143
340

0.0016
1512

281
51.2
1143
340

51.2

281
0
0
1512

340 1214 123512351235

151
2
1512 1512 1512

Figure 19: Phase-to-ground fault current (scaled by 103) for a fault at location F1.

6.5 Sequence- to Phase-domain Values Conversion

From the sequence interconnection diagram in Figure 18, we can see that 𝐼0 = 𝐼1 = 𝐼2 , from
equation (4) and considering the case of fault location F4, whose 𝐼0 is equal to 3,721.2 ∠ − 52.5˚
amperes. We have that:
𝐼𝐴 = 𝐼0 + 𝐼1 + 𝐼2 = 3𝐼0 = 11,163.55 ∠ − 52.5˚
𝐼𝐵 = 𝐼0 + 𝛼 2 𝐼1 + 𝛼𝐼2 = 3,721.2 ∠ − 52.5 + 3,721.2 ∠ 187.5 + 3,721.2 ∠ 67.5˚
= 2265.32 − 𝑗2952.22 − 3689.36 − 𝑗485.74 + 1424.04 + 𝑗3437.94 = 0

Copyright © 2017 SES ltd. All rights reserved. Page 19-17

PART III: SES CONTRIBUTIONS

and:
𝐼𝐶 = 𝐼0 + 𝛼𝐼1 + 𝛼 2 𝐼2 = 0.
This result shows a basic relation between the neutral current and the zero sequence current [2],
from Kirchhoff’s law neutral current 𝐼𝑛 can be related to the zero-sequence current 𝐼𝑜 by means
of:

𝐼𝑛 = −(𝐼𝐴 + 𝐼𝐵 + 𝐼𝐶 ) (5)

and:
1
𝐼𝑜 = (𝐼𝐴 + 𝐼𝐵 + 𝐼𝐶 ). (6)
3
Equating Expression (5) to Expression (6) we obtain:
𝐼𝑛 = −3𝐼0

7 Inclusion of Mutual Impedance Values

Assume now that Cable 2 in Figure 1 is replaced by two parallel cables with impedance values as
presented in Table 5, notice there the presence of mutual impedance components. The modified
circuit is presented in Figure 20, its circuit impedance equivalent is presented in Figure 21 and its
MALZ or HIFREQ circuit is presented in Figure 22 where conductor numbers are displayed.
Utility
115 kV
1500 MVA
Bus1 X/R=15

T1,15 MVA
Y SAT

115/13.8 kV
Z=7%, X/R=20 10 ohm
Bus2

300 ft Cu cable2
100 ft Cu cable1 2-3/C-4/0 200 ft Cu cable3
LCC

LCC

LCC

LCC

1-3/C-2/0 1-3/C-#4 F1
Z=0.0977+J0.0385 Z=0.3114+J0.0472
ohms/1000 ft ohms/1000 ft
Bus4 Bus6
Bus3
T2,3.75 MVA T3,1.5 MVA
Y SAT

Y SAT

Torque 13.8/4.16 kV 13.8/0.48 kV

IM
WI

M1 Z=5.5%, X/R=11 Z=5.75%, X/R=6.5

6 ohm Bus7
Bus5

1.0 pf syn Torque Torque Torque

250 ft Cu cable4
LCC

3-1/C-250 MCM
IM

IM

IM
WI

WI

WI

4000 hp F2
1800 rpm Z=0.0534+J0.0428 F3
X''d = 15% M2 M3 M4 ohms/1000 ft
X/R = 28.9 Bus8
Ind Ind 1500 hp
500 hp 2000 hp 4*100 hp
F4
1800 rpm 3600 rpm 8*50 hp
X''d = 16.7% X''d = 16.7% 28*25 hp
X/R = 19.3 X/R = 30 X''d = 16.7%

Figure 20: Modified circuit.

Page 19-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 5: Parallel cables sequence-impedance values.

Sequence Self (ohm) Mutual (ohm) Self (p.u.) Mutual (p.u.)

Impedance R X R X R X R X

Positive 0.02256 0.08415 0.00544 0.7371 0.001776 0.006626 0.000428 0.05804

Zero 0.04025 0.30467 0.00368 0.0122 0.003169 0.02399 0.00029 9.61E-04

Ut M1 M2 M3 M4-100hp M4-50hp M4-25hp

0.0007 0.0251 0.3331 0.0449 1.0114 1.5263 1.2623
+j0.01 +j0.703 +j6.3284 +j1.3917 +j8.3947 +j8.3947 +j4.797

T1 C1
+ 0.0035 0.0008 T2 C4
Driving F2 T3
+j0.0699 +j0.0003 0.0199 0..8691
Voltage 0.0874
+j0.2191 +j0.6966
F3 +j0.5683
F4
C2-1 C2-2 C3
C2-M
0.0018 0.0018 0.0049
+j0.0066 +j0.0066 +j0.0007

F1

Figure 21: Modified circuit impedance diagram.

Figure 22: HIREQ modified circuit impedance diagram.

Copyright © 2017 SES ltd. All rights reserved. Page 19-19

PART III: SES CONTRIBUTIONS

7.1 HIFREQ’s Mutual Command

To compute the three-phase fault current corresponding to the circuit on Figure 22 we need to
take advantage of HIFREQ’s mutual command. The Mutual command specifies a mutual
impedance between two different conductors (10 and 22) as shown in HIFREQ SICL input file
shown in Figure 23:

Figure 23: Mutual Impedance Definition in SICL.

Where in the highlighted command, each parameter separated by a comma is as follows, from left
to right:
 Command Line: Mutual.
 Impedance type: Lumped or on a per unit length basis of the conductor.
 Mutual Resistance: in ohms, ohms/m or ohms/ft.
 Mutual Reactance: in ohms, ohms/m or ohms/ft.
 SourceID: Conductor number acting as the source of the mutual interaction.
 TargetID: Conductor number acting as the target of the mutual interaction.
 IsAsymmetric: Controls whether the mutual interaction is symmetric (0) or
asymmetric (1).
 IsAdditive: Controls whether the specified impedance adds to the mutual
impedance computed by HIFREQ (1), or if it is to be interpreted as the total mutual
impedance between the conductors (0).
If the fault current at location F1 is computed without accounting for the mutual impedance
between cables C2-1 and C2-2, a value of 14.87 p.u. is obtained which is equivalent to 9,336.48 A,
see Figure 24(a). If the mutual impedance is taken into account a value of 14.85 p.u. equivalent to
9,325.8 A is obtained, see Figure 23 (b). Notice that this mutual interaction would have been of
greater importance for coupled conductors of long transmission lines or cables resulting in larger
differences when this mutual interaction parameter is accounted for in the computations.

Page 19-20 Copyright © 2017 SES ltd. All rights reserved.

 CURRENTS ENTERING (AMPS) CURRENTS ENTERING (AMPS)
SCALING FACTOR=1.E+3 SCALING FACTOR=1.E+3

Maximum Value : UGM 14.867 2017 – CANNON BEACH, OREGON Maximum Value : 14.853
Longitudinal Current Flowing in Origin of Conductor. Magnitude (A) [ID:SAMPLE TEST RUN @ f=60.0000 Hz ] Longitudinal Current Flowing in Origin of Minimum
Conductor. Magnitude (A)Value : 0.120E-07
[ID:SAMPLE TEST RUN @ f=60.0000 Hz ] Minimum Value : 0.119E-07

14867 2483 1073 942 347 253 160 0.000012 14853 2469 1059
14.87 930 347 253 1600.000012 14.85
13.38

94.
13.37

94.
131

129
0

160
11.89

93.2

93.2
11.88

0
597

0
586
16

12384
12384

1411
1411
10.41 10.40
8.92 8.91

94.
94.
7.43 7.43

160
93.2

93.2
0

0
5.95

16
5.94

131

129
14853
14867

597

586
0.00017
0.000013 4.46 0.00017
0.000013 4.46
2.97 2.97

12384
12384

1411
1411
1.49 1.49

347

347
728

715
12384

12384
1411

347

1411

347
728

715
373 356
355

347
373

347
359

356
14853
14867

12384

12384
1411

1411
355

347
373

347
12384 13795 14150 14523 14867 148677 12384 13795 14154 14510 14853 14853
6 53
148
6 7 148
148
67 53
14867 14867 14867 148 14853 14853 14853 148

(a) (b)
Figure 24: Three-phase fault current (scaled up by 103) for fault at location F1 when the mutual
impedance is (a) not accounted for, (b) is taken into account.

8 Conclusions
This article has introduced the possibility of conducting circuit based computations by taking
advantage of the lumped impedance capabilities available in MultiGroundZ and MultiFields.
HIFREQ’s Mutual command has been briefly introduced and has been applied using a practical
example. The circuit computation approach has been used alongside the per-unit system and
symmetrical components to carry out short-circuit studies in multi-voltage level systems. The
concept of circuit impedance networks has been introduced and used to perform three-phase and
phase-to-ground fault current analysis.
These techniques can be useful when fast computations are required and dedicated circuit
software packages are not available.
Other circuit based computations involving hybrid networks made of real conductors, cables and
lumped impedances can also be performed as shown in this article.

9 References
[1] General Electric Company, "Application Information, Short Circuit Current Calculations for Industrial
and Commercial Power Systems" in GE white papers, Cincinnati, USA, 1989.
[2] A. A. Amberg and A. Rangel, "Tutorial in Symmetrical Components - Part 2," in SEL white papers,
USA, 2014.

Copyright © 2017 SES ltd. All rights reserved. Page 19-21

 UGM 2017 – PORTLAND, OREGON

20 SHORT-CIRCUIT STUDY IN NETWORKS WITH

MULTIPLE TRANSFORMERS OUTSIDE THE CENTRAL
SITE IN THE SPLITS COMPUTATION MODULE
Nariman R. Tavana, Mark Boisjoli, Yexu Li, Simon Fortin, and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
Presently, the SPLITS computation module has only one automated, user-friendly interface that can model one
transformer at the Central Station. Placing transformers at other locations requires a good knowledge of the
transformer circuit structure and SPLITS input command language. This chapter describes in great details how a
network with multiple transformers can be implemented in SPLITS for more realistic load and fault current
distribution analyses. A typical example illustrating a short-circuit analysis is carried out based on a two-
transformer, two-machine network. The fault currents are computed in SPLITS directly in three-phase coupled
quantities. For the sake of comparison, the uncoupled sequence networks are employed to calculate the same short-
circuit currents. Additionally, the same analysis is also carried out on the same network using ATP-EMTP for
further verifications. A close agreement between three sets of results demonstrates the accuracy of using SPLITS
for short-circuit analysis of a network with multiple transformers. This approach can be useful to users who need
to carry such studies, until a fully automated interface is implemented in SPLITS that will allow the insertion of a
transformer at any location along a terminal.

1 Introduction
Electric transformers are major power system components that allow transfer of energy with high
efficiency between various power system networks operating at different voltage levels. Presently,
the SPLITS and HIFREQ computation modules can be used to model networks with power
transformers. However, SPLITS has only one automated, user-friendly interface that can model
one transformer at the Central Station. Placing transformers at other locations requires a good
knowledge of the transformer circuit structure and SPLITS input command language. This
chapter describes in great details how a network with multiple transformers can be implemented
in SPLITS for more realistic load and fault current distribution analyses.
Short-circuits occur in electric power system networks due mainly to equipment failures,
pollution or mechanical causes. The resulting fault currents in a network depend on the voltage
sources and the system impedances. Fault currents may be several orders of magnitude larger
than normal operating currents and may lead to thermal damage to equipment if they remain in
the system [1]. Therefore, it is necessary to isolate the faulted area in a short time. In this work,
we analyze sub-transient fault currents in a network with multiple transformers using the SPLITS
computation module.
The remainder of this paper is organized as follows. Section 2 gives an overview of the system
under study. Section 3 presents how a network with multiple transformers is created in SPLITS
at specific sections along a terminal. Different fault scenarios are investigated in Section 4. Section
5 illustrates short-circuits in the network for different arrangements of power transformers
outside the Central Site in SPLITS. Section 6 discusses the main conclusions and future work
related to this subject.

Copyright © 2017 SES ltd. All rights reserved. Page 20-1

PART III: SES CONTRIBUTIONS

2 Power System Network and its Sequence Equivalents

Figure 1 shows the single-line diagram of the power system network under study, where positive,
negative, and zero-sequence impedances are given. The neutrals of the generator and
transformers are solidly grounded. The motor neutral is grounded through a reactance 𝑋𝑛 =
0.04 p. u. on the motor base. Transformer 1 has a zero-sequence impedance (𝑋0 = 0.1 p. u.) larger
than its positive sequence (𝑋1 = 0.08 p. u.). During normal operating conditions, the voltage at
Bus 4 is 20 kV and the motor is drawing 50 MVA at 0.8 lagging power factor.
The sequence networks are shown in Figure 2. The negative-sequence network is similar to the
positive-sequence network, except that there are no electric sources. The sub-transient reactance
(𝑋 ′′ ) of the machines are used for determining fault currents. The 30-degree phase shifts of the
delta-wye transformers are omitted in the sequence networks and it is taken into account in the
short-circuit calculations whenever needed. Since the motor is grounded through a reactance 𝑋𝑛 ,
an impedance of 3𝑋𝑛 is included in the zero-sequence motor circuit.

Figure 1: One-line diagram of the network under study.

Based on the assumptions given for Bus 4 where the voltage profile for positive sequence is
1.0∠0𝑜 p. u., the internal phase voltages of motor (𝐸𝑀 ) and generator (𝐸𝐺 ) are calculated as
0.9434∠−4.8645𝑜 p. u. and 1.1528∠9.1845𝑜 p. u., respectively.

Figure 2: Sequence networks for the case under study.

Page 20-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

3 Modeling of the network in SPLITS

SPLITS is a phasor domain circuit software that predicts the system response in steady-state and
fault conditions directly as multi-phase coupled quantities. The three-phase impedances of each
component in the network can be obtained from the sequence components (if not available
otherwise) using the following expression.
𝒁𝑎𝑏𝑐 = 𝑨𝒁012 𝑨−1. (1)
Where 𝑨 is the Fortescue transformation matrix given by

1 1 1 (2)
𝑨 = [1 𝑎 2 𝑎]; 𝑎 = 1∠120𝑜 .
1 𝑎 𝑎2

For three-phase transformers with zero-sequence impedances larger than the positive-sequence
ones, a compensating neutral reactance is added at the transformer neutral point. If the positive-
sequence leakage reactance is 𝑋1 and zero-sequence reactance is 𝑋0 , a compensating neutral
(𝑋0 −𝑋1 )
reactance equal to 𝑋𝑛 = 3
is added at the star point to ground in the network. For
Transformer 1 in Figure 1, the compensating neutral reactance would be equal to 0.006667p. u.
Figure 3 shows the network under study in SPLITS in three-phase domain in its pre-fault
condition. It can be seen that the open-circuit sections are created in the network for inserting
transformers in the corresponding terminal sections. The primary and secondary sides of the
transformers are connected to the appropriate phases using the mutual connection command.
Two virtual nodes are created for the lumped representation of the transformers’ grounding
systems. These nodes are connected to the grounding shunt impedances of the transformers that
can be calculated using MALT, MALZ, or HIFREQ for realistic cases. Since Transformer 1 has a
zero-sequence impedance larger than its positive-sequence impedance, a compensating neutral
reactance is added to the neutral-to-ground impedance. The following step-by-step approach
needs to be done in the sp_XXX.F05 file in order to create a transformer model outside the central
site. It is worth noting that modeling of multiple transformers outside the central station will be
automated in the future in SPLITS.
 Step 1: create the network without transformers.
 Step 2: define open-circuit section (𝑍𝑠𝑒𝑟𝑖𝑒𝑠 ≈ ∞, 𝑍𝑠ℎ𝑢𝑛𝑡 ≈ ∞) in order to patch a
transformer in that section.
 Step 3: Specify the transformer at the central site in such a way that 𝑆𝑗 , 𝑆𝑗 + 1, 𝑆𝑗 + 2 are
primary buses and 𝑆𝑗 + 3, 𝑆𝑗 + 4, 𝑆𝑗 + 5 are secondary buses, where 𝑆𝑗 is the last bus defined
on the network (in this case 𝑆𝑗 is equal to 3).

 Step 4: generate a .F72 file by running the model in order to create the transformer
impedance matrix.
 Step 5: reload the sp_XXX.F05 file and remove the transformer model by deleting the
associated commands, including the TRANSFORMER, CHARACTERIST, PRIMARY and
SECONDARY commands.
 Step 6: open the .F72 file in a text editor and copy the impedance matrix of the
transformer defined in sections 1 and 2.
Copyright © 2017 SES ltd. All rights reserved. Page 20-3
PART III: SES CONTRIBUTIONS

 Step 7: paste these two sections in the content of the appropriate terminal, in the .F05 file
(using a text editor again), where the transformer is located and change their section
numbers to 𝑆𝑡 + 1, and 𝑆𝑡 + 2, where 𝑆𝑡 is the section number of the transformer location
in the terminal.
 Step 8: create extensions at primary and secondary ports of the transformer, with
𝑍𝑠𝑒𝑟𝑖𝑒𝑠 ≈ 0 and 𝑍𝑠ℎ𝑢𝑛𝑡 ≈ ∞ at section 𝑆𝑡 from bus 𝑆𝑗 to 𝑆𝑗 + 5.

 Step 9: connect the primary and secondary of the transformer to the network using
MUTU or MUTUAL command.
 Step 10: create an extra bus 𝑆𝑚 for the neutral-to-ground connection of the transformer
and define a section at 𝑆𝑡 with large series impedance and a shunt impedance equal to the
grounding grid impedance of the transformer.
 Step 11: connect the neutral point of the transformer to the section created in the last step
using the MUTU or MUTUAL command.
 Step 12: create decoupling sections (𝑍𝑠𝑒𝑟𝑖𝑒𝑠 ≈ ∞, 𝑍𝑠ℎ𝑢𝑛𝑡 ≈ ∞) between transformers when
more than one transformer is defined in the network.
 Step 13: define fake energizations from 𝑆𝑗 to 𝑆𝑚 using voltage sources with zero Volt and
large series impedances.

Figure 3: An overview of the system under study in SPLITS.

Page 20-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

For further comparison, a similar network is developed in ATP-EMTP to obtain solutions for
various fault scenarios. A BCTRAN component is employed for modeling of transformer whose
zero sequence impedance can be specified larger than the positive sequence one [2]. A general
overview of the network in ATP is depicted in Figure 4. Shunt reactors with very large impedances
(open-circuits) are inserted in the network at Buses 1 and 4 to improve numerical stability.

Figure 4: An overview of the network in ATP-EMTP.

4 Short-Circuit Analysis
In this paper, the sub-transient fault currents for the following scenarios are investigated;
 Three-phase fault at Bus 1.
 Single line-to-ground fault at Bus 1.
 Line-to-line fault at Bus 1.
 Double line-to-ground fault at Bus 1.
 Single line-to-ground fault at 30% of transmission line seen from Bus 2.

4.1 Three-phase fault at Bus 1.

For the bolted three-phase sub-transient fault current, the positive-sequence network is shorted
at Bus 1 as shown in Figure 5.

Figure 5: Solid three-phase fault at Bus 1.

The contributions of positive-sequence current from the motor (𝐼1𝑇𝑅#1) and generator (𝐼1𝐺 ) during
the three-phase fault condition at Bus 1 have been determined as 1.8498∠ − 94.8645𝑜 p. u., and
7.6852∠ − 80.8155𝑜 p. u., respectively.
Therefore, the motor feeds the faulted bus in phase domain (𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 ) as follows:

0 5.34∠ − 94.86𝑜 (3)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 = 𝑨. 𝑰′′
012_𝑇𝑅#1 . 𝐼𝑏𝑎𝑠𝑒 = 𝑨. [𝐼1𝑇𝑅#1 ] . 𝐼𝑏𝑎𝑠𝑒 = [ 5.34∠ 145.14𝑜 ] kA
0 5.34∠ 25.14𝑜

The contribution of the generator (𝑰′′

𝑎𝑏𝑐_𝐺 ) into the fault current in each phase is calculated as:

Copyright © 2017 SES ltd. All rights reserved. Page 20-5

PART III: SES CONTRIBUTIONS

0 22.185∠ − 80.82𝑜 (4)

𝑰′′
𝑎𝑏𝑐_𝐺 = 𝑨. 𝑰′′
012_𝐺 . 𝐼𝑏𝑎𝑠𝑒 = 𝑨. [𝐼1𝐺 ] . 𝐼𝑏𝑎𝑠𝑒 = [ 22.185∠ 159.18𝑜 ] kA
0 22.185∠ 39.18𝑜

Figure 6 shows the SPLITS fault scenario at Bus 1 in three-phase domain. It can be observed that
the contributions from both sides of the fault at this location are 5.329 kA and 22.19 kA,
respectively.

Figure 6: Bolted three-phase fault at Bus 1.

The maximum value of short-circuit current contributions computed by ATP-EMTP are listed in
Figure 7.

Figure 7: Maximum value of short-circuit currents calculated by ATP-EMTP.

Page 20-6 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – PORTLAND, OREGON

Table 1 lists the RMS values of the short-circuit currents obtained by all three approaches. The
percentage deviation provided in the second column shows the difference of fault current
magnitudes obtained by SPLITS compared to the sequence network calculations. It can be clearly
seen that there is a close agreement between the three sets of results.

Table 1: RMS value of fault currents.

Short-Circuit SPLITS Results and Sequence ATP-EMTP
Currents Deviation in % Network Method Results
𝑰′′
𝒂_𝑻𝑹#𝟏 5.329∠ − 94.87𝑜 kA (0.21%) 5.34∠ − 94.86𝑜 kA 5.34∠ − 94.86𝑜 kA
𝑰′′
𝒃_𝑻𝑹#𝟏 5.329∠145.13𝑜 kA (0.21%) 5.34∠145.14𝑜 kA 5.34∠145.14𝑜 kA
𝑰′′
𝒄_𝑻𝑹#𝟏 5.329∠25.1𝑜 kA (0.21%) 5.34∠25.14𝑜 kA 5.34∠25.14𝑜 kA
𝑰′′
𝒂_𝑮 22.19∠ − 80.82𝑜 kA (0.02%) 22.185∠ − 80.82𝑜 kA 22.185∠ − 80.82𝑜 kA
𝑰′′
𝒃_𝑮 22.19∠159.2𝑜 kA (0.02%) 22.185∠159.18𝑜 kA 22.185∠159.18𝑜 kA
𝑰′′
𝒄_𝑮 22.19∠39.18𝑜 kA (0.02%) 22.185∠39.18𝑜 kA 22.185∠39.18𝑜 kA

4.2 Single line-to-ground (SLG) fault at Bus 1.

A single line-to-ground fault occurs at Bus 1 in this scenario. In this case, the zero, positive, and
negative sequence networks are connected in series at the fault location as shown in Figure 8.

Figure 8: Interconnected sequence networks.

The fault current contributions from the transformer side (𝑰012𝑇𝑅#1 ) and generator side (𝑰012_𝐺 )
to Bus 1 are calculated as follows:

𝐼0𝑇𝑅#1 0 𝐼0𝐺 4.0469∠ − 83.5275𝑜 (5)

[𝐼1𝑇𝑅#1 ] = [0.6817∠ − 115.7667 ] p. u. ; and [𝐼1𝐺 ] = [3.4893∠ − 77.5457𝑜 ] p. u.
𝑜

𝐼2𝑇𝑅#1 0.9197∠ − 83.5275𝑜 𝐼2𝐺 3.1271∠ − 83.5275𝑜

It should be highlighted that since the SLG fault occurs at the delta side of Transformer 1, the
zero-sequence current contribution of Transformer 1 into the fault (𝐼0𝑇𝑅#1 ) is zero because a delta
connection is an open-circuit for ground faults.

Copyright © 2017 SES ltd. All rights reserved. Page 20-7

PART III: SES CONTRIBUTIONS

The sub-transient fault currents in the network for each phase (𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 ) from Transformer 1 are
given by:

4.4452∠ − 97.19𝑜 (6)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 = 𝑨. 𝑰′′
012_𝑇𝑅#1 . 𝐼𝑏𝑎𝑠𝑒 = [ 3.3660∠72.22𝑜 ] kA
1.2942∠111.39𝑜

The contribution of the generator (𝑰′′

𝑎𝑏𝑐_𝐺 ) to the fault current is calculated as:

30.7453∠ − 81.57𝑜 (7)

𝑰′′
𝑎𝑏𝑐_𝐺 = 𝑨. 𝑰′′
012_𝐺 . 𝐼𝑏𝑎𝑠𝑒 = [3.3660∠ − 107.78𝑜 ] kA
1.2942∠ − 68.61𝑜

The following fault currents are computed by SPLITS in phase-domain as shown in Figure 9;

4.436∠ − 97.2𝑜 30.75∠ − 81.57𝑜 (8)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 = [ 3.362∠72.2𝑜 ] kA; 𝑰′′
𝑎𝑏𝑐_𝐺 = [3.362∠ − 107.8𝑜 ] kA
1.291∠111.52𝑜 1.291∠ − 68.48𝑜

It is worth noting that since the SLG fault current at Bus 1 cannot be shared with the grounding
grids of the transformers, the entire SLG fault current returns into the grounding system of the
generator and creates an earth potential rise at the power station.

Figure 9: Bolted single line-to-ground fault at Bus 1.

Page 20-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

Fault currents computed by ATP-EMTP are also presented in Table 2. It can be seen that the short-
circuit currents obtained by the different approaches are very close to each other.

Table 2: RMS value of fault currents.

Short-Circuit SPLITS Results and Sequence ATP-EMTP
Currents Deviation in % Network Method Results
𝑰′′
𝒂_𝑻𝑹𝟏 4.436∠ − 97.2𝑜 kA (0.21%) 4.4452∠ − 97.19𝑜 kA 4.4472∠ − 97.18𝑜 kA
𝑰′′
𝒃_𝑻𝑹#𝟏 3.362∠72.2𝑜 kA (0.12%) 3.3660∠72.22𝑜 kA 3.366∠72.28𝑜 kA
𝑰′′
𝒄_𝑻𝑹#𝟏 1.291∠111.52𝑜 kA (0.25%) 1.2942∠111.39𝑜 kA 1.294∠111.23𝑜 kA
𝑰′′
𝒂_𝑮 30.75∠ − 81.57𝑜 kA (0.02%) 30.7453∠ − 81.57𝑜 kA 30.746∠ − 81.57𝑜 kA
𝑰′′
𝒃_𝑮 3.362∠ − 107.8𝑜 kA (0.12%) 3.3660∠ − 107.78𝑜 kA 3.366∠ − 107.72𝑜 kA
𝑰′′
𝒄_𝑮 1.291∠ − 68.48𝑜 kA (0.25%) 1.2942∠ − 68.61𝑜 kA 1.294∠ − 68.77𝑜 kA
The fault current going into earth due to this event is represented in Table 3. Again a close
agreement is achieved.

Table 3: Earth current in SLG fault event.

Short-Circuit SPLITS Results and Sequence ATP-EMTP
Currents Deviation in % Network Method Results
𝑰′′
𝑺𝑳𝑮 = 𝟑𝑰𝟎 35.04∠−83.52𝑜 kA (0.03%) 35.05∠ − 83.53𝑜 kA 35.05∠−83.53𝑜 kA

4.3 Line-to-line fault at Bus 1.

In the line-to-line fault at Bus 1, the positive- and negative-sequence circuits are connected in
parallel at the fault location. Figure 10 shows the details of sequence network interconnections in
this scenario.

Figure 10: Sequence networks in double phase fault at Bus 1.

The fault current contributions into Bus 1 during a double line fault are as follows:

𝐼0𝐺 0 𝐼0𝑇𝑅#1 0 (9)

[𝐼1𝐺 ] = [4.0263∠−78.3459𝑜 ] p. u. ; and [𝐼1𝑇𝑅#1 ] = [0.8203∠−109.8427𝑜 ] p. u.
𝐼2𝐺 3.6667∠96.4725𝑜 𝐼2𝑇𝑅#1 1.0784∠96.4725𝑜

Copyright © 2017 SES ltd. All rights reserved. Page 20-9

PART III: SES CONTRIBUTIONS

The fault current in kA using the transformation matrix and base current at the generator side
are:

1.4434∠143.13𝑜 1.4434∠−36.87𝑜 (10)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 = [4.0307∠−179.42𝑜 ] kA; 𝑰′′
𝑎𝑏𝑐_𝐺 = [19.7207∠−172.33𝑜 ] kA
5.2504∠−9.04𝑜 18.72∠10.78𝑜

The fault currents for the network with multiple transformers are computed by SPLITS and are
given hereafter.

1.444∠143.09𝑜 1.444∠−36.91𝑜 (11)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1
′′
= [4.021∠−179.403 ] kA; 𝑰𝑎𝑏𝑐_𝐺 = [19.72∠−172.3𝑜 ] kA
𝑜

5.241∠−9.1𝑜 18.72∠10.78𝑜

Figure 11 provides a visual representation of the fault current contributions during a line-to-line
fault using GRSPLITS-3D.

Figure 11: Line-to-line fault scenario in SPLITS.

Table. 4 compares the short-circuit currents obtained by SPLITS, sequence networks, and ATP-
EMTP. It can be observed that there is a close agreement between the results, not only in terms
of fault current magnitudes, bus also phase angles.

Page 20-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

Table 4: RMS value of fault currents.

Short-Circuit SPLITS Results and Sequence Network ATP-EMTP
Currents Deviation in % Method Results
𝑰′′
𝒂_𝑻𝑹#𝟏 1.444∠143.09𝑜 kA (0.04%) 1.4434∠143.13𝑜 kA 1.4436∠143.25𝑜 kA
𝑰′′
𝒃_𝑻𝑹#𝟏 4.021∠−179.403𝑜 kA (0.24%) 4.0307∠−179.42𝑜 kA 4.032∠−179.45𝑜 kA
𝑰′′
𝒄_𝑻𝑹#𝟏 5.241∠−9.1𝑜 kA (0.18%) 5.2504∠−9.04𝑜 kA 5.2514∠−9.02𝑜 kA
𝑰′′
𝒂_𝑮 1.444∠−36.91𝑜 kA (0.04%) 1.4434∠−36.87𝑜 kA 1.4408∠−36.75𝑜 kA
𝑰′′
𝒃_𝑮 19.72∠−172.3𝑜 kA (0.004%) 19.7207∠−172.33𝑜 kA 19.7208∠−172.32𝑜 kA
𝑰′′
𝒄_𝑮 18.72∠10.78𝑜 kA (0.00%) 18.72∠10.78𝑜 kA 18.719∠10.77𝑜 kA

4.4 Double line-to-ground fault at Bus 1.

In this scenario a double line-to-ground fault occurs at Bus 1. The zero, positive, and negative
sequence networks are connected in parallel at the faulted bus. Figure 12 depicts the
interconnections between the sequence networks.

Figure 12: Interconnections of sequence networks in line-to-line-to-ground fault at Bus 1.

The short-circuit currents feeding from both sides into the faulted bus in the sequence domain
are given by:

𝐼0𝐺 5.6149∠96.4725𝑜 𝐼0𝑇𝑅#1 0 (12)

[𝐼1𝐺 ] = [6.1900∠−80.1597 ] p. u. ; and [𝐼1𝑇𝑅#1 ] = [1.4206∠−98.3581𝑜 ] p. u.
𝑜
𝐼2𝐺 1.4973∠96.4725𝑜 𝐼2𝑇𝑅#1 0.4404∠96.4725𝑜

The actual values in kA in the phase domain for the fault current contributions are obtained as:

2.8905∠75.18𝑜 2.8905∠−104.82𝑜 (13)

𝑰′′
𝑎𝑏𝑐_𝐺
′′
= [29.585∠138.26 ] kA ; and 𝑰𝑎𝑏𝑐_𝑇𝑅#1 = [ 4.6004∠157.11𝑜 ] kA
𝑜

30.3067∠58.45𝑜 5.0780∠11.41𝑜

Copyright © 2017 SES ltd. All rights reserved. Page 20-11

PART III: SES CONTRIBUTIONS

The fault currents obtained by SPLITS in the three-phase domain are:

2.885∠75.15𝑜 2.885∠−104.85𝑜 (14)

𝑰′′
𝑎𝑏𝑐_𝐺 = [29.59∠138.3𝑜 ] kA ; 𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 = [ 4.590∠157.1𝑜 ] kA
30.31∠58.46𝑜 5.068∠11.4𝑜

Figure 13 shows a line-to-line-ground fault and displays the fault current contributions using
GRSPLITS-3D.

Figure 13: Double line-to-ground fault scenario in SPLITS.

Table 5 lists the short-circuit currents computed by SPLITS, ATP-EMTP and using the Sequence
Network method. It can be seen that the fault currents obtained by SPLITS are very close to the
results obtained by the other methods.
Table 5: RMS value of fault currents.
Short-Circuit SPLITS Results and Sequence Network ATP-EMTP
Currents Deviation in % Method Results
𝑰′′
𝒂_𝑻𝑹#𝟏 2.885∠−104.85𝑜 kA (0.19%) 2.8905∠−104.82𝑜 kA 2.893∠−104.80𝑜 kA
𝑰′′
𝒃_𝑻𝑹#𝟏 4.590∠157.1𝑜 kA (0.23%) 4.6004∠157.11𝑜 kA 4.602∠157.10𝑜 kA
𝑰′′
𝒄_𝑻𝑹#𝟏 5.068∠11.4𝑜 kA (0.20%) 5.0780∠11.41𝑜 kA 5.0796∠11.43𝑜 kA
𝑰′′
𝒂_𝑮 2.885∠75.15𝑜 kA (0.19%) 2.8905∠75.18𝑜 kA 2.893∠75.19𝑜 kA
𝑰′′
𝒃_𝑮 29.59∠138.3𝑜 kA (0.02%) 29.585∠138.26𝑜 kA 29.586∠138.26𝑜 kA
𝑰′′
𝒄_𝑮 30.31∠58.46𝑜 kA (0.01%) 30.3067∠58.45𝑜 kA 30.3076∠58.46𝑜 kA

Page 20-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

4.5 Single line-to-ground fault (SLG) at 30% of the line seen from Bus 2.
Modeling of this scenario is similar to that of Section 4.2. The only difference is that the sequence
networks are connected together in series at the fault location at 30% of the transmission line as
viewed from Bus 2 (Figure 14).

Figure 14: Interconnections of sequence networks in SLG fault at 30% of transmission line.

The short-circuit current contributions from both sides of the transmission line into the fault
location in the sequence domain are:

𝐼0𝑇𝑅#1 1.3971∠−85.9896𝑜 𝐼0𝑇𝑅#2 0.8122∠−85.9896𝑜 (15)

[𝐼1𝑇𝑅#1 ] = [1.6596∠−72.8230𝑜 ] p. u. ; and [𝐼1𝑇𝑅#2 ] = [0.7035∠−118.4937𝑜 ] p. u.
𝐼2𝑇𝑅#1 1.2888∠−85.9896𝑜 𝐼2𝑇𝑅#2 0.9205∠−85.9896𝑜

Since delta-wye transformers (Yd1) introduce a 30-degree phase shift, the positive-sequence
voltages as viewed from the transmission line are shifted by 30𝑜 . Thus, the contributions to the
fault current leads those in Equation (15) by 30𝑜 as follows:

𝐼0𝑇𝑅#1 1.3971∠−55.9896𝑜 𝐼0𝑇𝑅#2 0.8122∠−55.9896𝑜 (16)

[𝐼1𝑇𝑅#1 ] = [1.6596∠−42.8230𝑜 ] p. u. ; and [𝐼1𝑇𝑅#2 ] = [0.7035∠−88.4937𝑜 ] p. u.
𝐼2𝑇𝑅#1 1.2888∠−55.9896𝑜 𝐼2𝑇𝑅#2 0.9205∠−55.9896𝑜

Transforming to the phase domain, the fault currents are:

722.7∠−50.97𝑜 394.4∠−65.22𝑜 (17)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1
′′
= [91.2∠−116.05𝑜 ] A ; and 𝑰𝑎𝑏𝑐_𝑇𝑅#2 = [ 91.2∠63.95𝑜 ] A
66∠110.16𝑜 66∠−69.84𝑜

It is interesting to note that as the SLG fault occurs at the grounded wye sides of both
transformers, the fault current splits between the grounding grids of the two transformers. In
other words, the fault currents are pulled out from both transformers grounding systems and feed
into the terminal. The main reason for this is that in the zero-sequence network as shown in Figure
14, there is a closed path from the transformer ground connections to the fault location for this
SLG event. Furthermore, the delta sides of the transformers prevent the SLG fault current from
going into the grounding systems of the generator and of the motor.
Copyright © 2017 SES ltd. All rights reserved. Page 20-13
PART III: SES CONTRIBUTIONS

The short-circuit currents computed by SPLITS are as follows:

721.3∠−50.97𝑜 393.9∠−65.2𝑜 (18)

𝑰′′
𝑎𝑏𝑐_𝑇𝑅#1 = [91.13∠−116.1𝑜 ] 𝐴 ; 𝑰′′
𝑎𝑏𝑐_𝑇𝑅#2 = [ 91.12∠63.9𝑜 ] 𝐴
65.94∠110.2𝑜 65.93∠−69.8𝑜

Figure 15 shows the fault currents as obtained directly in three-phase quantities using SPLITS. It
can be seen that the SLG current in this case is shared between the two grounds of the
transformers.

Figure 15: Single line-to-ground fault at 30% of transmission line in SPLITS.

Table 6 shows the fault currents obtained by the different approaches. The fault current
contributions from both sides of the transmission line SLG fault, obtained by SPLITS, are in good
agreement with the Sequence Network and ATP-EMTP results.

Page 20-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

Table 6: RMS value of fault currents.

Short-Circuit SPLITS Results and Sequence Network ATP-EMTP
Currents Deviation in % Method Results
𝑰′′
𝒂_𝑻𝑹#𝟏 721.3∠−50.97𝑜 A (0.19%) 722.7∠−50.97𝑜 A 722.66∠−50.97𝑜 A
𝑰′′
𝒃_𝑻𝑹#𝟏 91.13∠−116.1𝑜 A (0.08%) 91.2∠−116.05𝑜 A 91.13∠−115.96𝑜 A
𝑰′′
𝒄_𝑻𝑹#𝟏 65.94∠110.2𝑜 A (0.09%) 66∠110.16𝑜 A 65.86∠110.27𝑜 A
𝑰′′
𝒂_𝑻𝑹#𝟐 393.9∠−65.2𝑜 A (0.13%) 394.4∠−65.22𝑜 A 394.4∠−65.22𝑜 A
𝑰′′
𝒃_𝑻𝑹#𝟐 91.12∠63.9𝑜 A (0.09%) 91.2∠63.95𝑜 A 91.13∠64.04𝑜 A
𝑰′′
𝒄_𝑻𝑹#𝟐 65.93∠−69.8𝑜 A (0.11%) 66∠−69.84𝑜 A 65.85∠−69.73𝑜 A
Table 7 also shows that the SLG currents as computed by the different approaches are very close.
Table 7: Single-line-to-ground fault current.
Short-Circuit SPLITS Results and Sequence Network ATP-EMTP
Currents Deviation in % Method Results
𝑰′′
𝑺𝑳𝑮 = 𝟑𝑰𝟎 1107∠−55.99𝑜 A (0.20%) 1109.2∠−55.99𝑜 A 1109.2∠−55.99𝑜 A

5 Different scenarios of transformers and network in SPLITS

Based on the step-by-step approach explained in Section 3, various realistic scenarios of power
system network consisting of transformers and overhead lines or underground cables can be
modeled easily in SPLITS. Figure 16 shows how several transformers located along a terminal can
be modelled at any location in SPLITS.

Figure 16: Two transformers along one terminal.

Figure 17 illustrates a feeder connected to the upstream grid (utility) at one end through a
transformer and connected at the other end through the overhead lines and underground cables
that feed downstream grids.
The figure shows that this scenario can be modeled in SPLITS as three terminals where
transformers are defined outside the Central Site.

Copyright © 2017 SES ltd. All rights reserved. Page 20-15

PART III: SES CONTRIBUTIONS

Figure 17: Three transformers modelled outside the Central Site along three different terminals.

Two different parallel transformers that deliver a larger amount of power than a single one can
also be modelled in SPLITS as depicted in Figure 18.

Figure 18: Two parallel transformers outside the Central Site.

It can be observed form Figures 16 to 18 that various realistic configurations of a power system
network consisting of transformers at different locations can be easily studied in SPLITS.

Page 20-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – PORTLAND, OREGON

6 Conclusions
A typical example illustrating a short-circuit analysis is carried out based on a two-transformer,
two-machine network. The fault currents are computed in SPLITS directly in three-phase coupled
quantities.
For the sake of comparison, the uncoupled sequence networks are employed to calculate the same
short-circuit currents. Additionally, the same analysis is carried out on the same network using
ATP-EMTP for further verifications.
A close agreement between the three sets of results confirms the accuracy of using SPLITS for
short-circuit analysis of a network with multiple transformers.
This approach can be useful to users who need to carry such studies, until a fully automated
interface is implemented in SPLITS that will allow the insertion of a transformer at any location
along a terminal.

7 Acknowledgment
Authors would like to thank Dr. Luis Valcárcel for his careful reading of this article and numerous
constructive comments.

References
[1] J. D. Glover, and M. S. Sarma, "Power System Analysis and Design," Third Edition, Thomson-
Engineering, 2002.
[2] V. Brandwajn, H. W. Dommel, and I. I. Dommel, "Matrix representation of three-phase N-winding
transformers for steady-state and transient studies," IEEE Trans. Power App. Syst., vol. PAS-101, no. 6, pp.
1369-1378, Jun. 1982.

Copyright © 2017 SES ltd. All rights reserved. Page 20-17

 UGM 2017 – CANNON BEACH, OREGON

21 THE ELECTROMAGNETIC INTERFERENCE (EMI)

MODEL FOR ELECTRIC RAILWAYS IN CDEGS
Yexu Li, Nariman Tavana and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com
High induced voltages on nearby metallic utilities such as pipelines and rails in joint-use corridors with electric
railways can cause problems to equipment and threaten people safety at exposed locations. This study describes
and demonstrates the applicability of using CDGES to model electric railway tractions. Particularly, the paper
illustrates the approaches of modeling the autotransformers in an AT traction system by using the HIFREQ and
ROW computation modules. An appropriate method in ROW to simulate the moving train is provided and examples
along with numerical results are presented to illustrate the ability of CDEGS for the studies of electromagnetic
interference (EMI) caused by electric railway.

1 Introduction
The first electric passenger train was introduced by Werner von Siemens at Berlin in 1879. Since
that time electric railways became more and more popular and nowadays high-speed railways
(HSR) are a dominant means of transportation in many developed and developing countries,
mainly in Europe and Asia. For example, high-speed rail service in China was introduced on April
18, 2007 and has become immensely popular with an annual ridership of over 1.44 billion in 2016,
making the Chinese HSR network the most heavily used in the world. Notable lines include the
world's longest line, the 2,298 km Beijing–Guangzhou High-Speed Railway and the Shanghai
Maglev line, the world's first high-speed commercial magnetic levitation (maglev) line. See
Figure 1.

Figure 1: Beijing–Guangzhou HSR (Left) and the Shanghai Maglev (Right).

In comparison to the steam and diesel engine locomotives, electric railways offer substantially
better energy efficiency, lower emissions of pollutants and lower operating costs. Electric
locomotives are usually quieter, more powerful, and more responsive and reliable than steam and
diesel ones. Furthermore, while steam and diesel locomotives burn coal and petroleum, electricity
is generated from diverse sources including many that do not produce carbon dioxide such as
nuclear power and various renewable energies sources including hydroelectric, geothermal, wind
and solar.

Copyright © 2017 SES ltd. All rights reserved. Page 21-1

PART III: SES CONTRIBUTIONS

On the other hand, electric traction systems also present their own specific disadvantages, which
include:
a. High capital costs that may be too expensive on lightly trafficked routes.
b. Vulnerability to power adverse effects, because the railway traction power is a single-phase
heavy load, it produces large negative sequence currents and voltages in the power system,
resulting in low power factor, high harmonic content, etc.
c. Single-phase electric traction loads generate electromagnetic interference on communication
lines and nearby metallic structures such as pipelines. See Figure 2.
d. Maintenance of the catenary requires that a power outage be scheduled to ensure a safe
environment for the workers. During this time train operations are stopped along the affected
route.

Figure 2: Electromagnetic interference from electric railways.

2 Mechanisms of Electric Railway Operation

Nowadays, four types of electric railway traction networks are operated worldwide. They are:
a. Direct Power Supply
b. Direct Power Supply with a Return Line
c. Booster Suction Transformer Power Supply
d. Autotransformer Power Supply.
1. Direct power supply: The three-phase power supply from the power utility is converted
to two single phases, to ensure that the current is as balanced as possible on the three
Page 21-2 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

phase source side, one separate phase is fed to each of the overhead up and down the
tracks. Various methods can be used to connect the feeding transformers. Nowadays, the
Scott connected transformer and the modified wood-bridge connected transformer are
typically used. The two single phases are connected directly the catenary system feeding
the electric locomotives with current returning through the tracks and earth back to the
traction substation.
This power supply system is characterized by the simplest configuration of the power
supply circuits, low cost of operation and easy maintenance. However, its electromagnetic
interference effects on adjacent communication lines and metallic pipelines is the most
serious. Furthermore, the rail track remains at high ground potential, which is an unsafe
situation.

Figure 3: Direct power supply electric railway traction.

2. Direct power supply with a return line: In order to retain the advantages of direct
power supply mode but minimize the EMI issues, a return line is added in parallel with
and close to the overhead circuit (contact wire). This has become a direct power supply
with a return line, as shown in Figure 4.

Copyright © 2017 SES ltd. All rights reserved. Page 21-3

PART III: SES CONTRIBUTIONS

Figure 4: Direct power supply with the return Line.

3. Booster transformer power supply: Figure 5 shows the schematic of a booster

transformer (BT) feeding circuit. A BT is installed about every 4 km on the contact wire to
boost the return current circuit on the negative line. This design minimizes the inductive
interference on telecommunication lines and nearby metallic lines because the current is
forced to the return line, cancelling the EMF effects of the contact wire as well as restricting
the rail current only in limited sections. However, the problem with BT system is that,
when an electric car passes a BT section, a large arc is generated in the section, and a large
load current can cause a very large arc that can damage the overhead line. Consequently,
capacitors are often inserted serially in the negative feeder to compensate for the reactance
and reduce the amount of current intercepted by the pantograph, thereby reducing arcing,
and also helping to prevent voltage drop along the catenary.

Page 21-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Booster transformer electric railway traction.

4. AT feeding system: Perhaps the most popular and significant feeding network is known as
the Auto-Transformer (AT) system. In the AT feeding system, the feeding voltage of the
substation is twice the voltage supplied to the electric train. An AT is installed at every 5 km
to 10 km along the track. Interference caused by the load current in the overhead catenary
system and running rails is considerably reduced. Autotransformers inherently attempt to
equalize the current flowing in the two sections of the transformer windings.

This autotransformer effect reduces the current in the running rails between the train and the
remote autotransformer feeding the section, and very much reduced in the adjacent electrical
sections. This is very effective in reducing electromagnetic (especially inductive) interference
in the corridor. Figure 6 shows the schematic of an AT feeding system. In China, the AT is
designed with a turn ratio of 1:1 and the substation feeding voltage is twice the overhead-line
voltage. This system is ideal for high-speed and large-capacity electric cars because there are
no large voltage drops along the catenary and no arcing sections.

Copyright © 2017 SES ltd. All rights reserved. Page 21-5

PART III: SES CONTRIBUTIONS

Figure 6: Autotransformer (AT) feeding electric railway traction.

3 Description of a Simple AT Electric Railway Interference

Problem
The interference caused by a typical AT catenary system located along a nearby pipeline is
analyzed to illustrate how to carry out an electric railway interference study. As shown in Figure
7, an AT electric railway and a 24” diameter (0.61 m) high pressure gas pipeline share a common
corridor for a distance of 15 km. Both the contact and the return wires are made of 100 mm 2
copper conductors. A 55 kV/27.5 kV step-down autotransformer is connected to the contact,
return and rail system as shown in Figure 8. Depending on the studied scenario, the
autotransformer is located every 5 km or 15 km along the track. The contact wire conductor is at
a height of 10 m above the ground. The return wire is at the same height as the contact wire, but
with a 0.5 m horizontal separation offset. The track rail is modeled as an equivalent cylindrical
conductor with a radius of 0.1 m, a relative resistivity of 13.4 (with respect to copper) and a relative
permeability (with respect to free space) of 173. The railway rail is coated with an appropriate
material to represent the ballast contact resistance with the track rails. This equivalent material
coating resistivity is 30,487 ohm-m (for a ballast resistance of 100 ohm-kft and a soil resistivity
of 100 ohm-m). Refer to the [1] for how to accurately model the ballast effects.
The peak load current in the contact wire is on the order of 200 A, resulting in a current of about
400 A feeding the train motor.
A nearby pipeline, about 10 m from the track center runs parallel to the catenary traction line. The
steel pipeline is modeled with a relative resistivity of 12 (with respect to copper) and a relative
permeability (with respect to free space) of 300. It has a coating resistance of 40,000 ohm-m2. It
is buried 1.5 m below ground (soil surface to pipeline center).

Page 21-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 7: Cross-section of the Right-of-Way.

Figure 8: Single phase autotransformer.

Copyright © 2017 SES ltd. All rights reserved. Page 21-7

PART III: SES CONTRIBUTIONS

4 Modeling the AT Catenary Traction Line in HIFREQ

The inductive interference is caused by the combined geometrical and net current imbalance of
the electric system with respect to the rail system while the conductive interference is due to the
earth currents. One of the advantages of the AT catenary system is that it minimizes the
electromagnetic interference on telecommunication and other nearby utilities and restricts the
region affected by the interference to the zone enclosing the location of the train and bounded by
the two nearest autotransformers as shown in Figure 6.
The HIFREQ computer model of the network described in Section 3 is rather straightforward as
illustrated in Figure 8 and Figure 9.
 The primary Point B and the secondary Point C are tied and connected to the rail track.
 The primary Point A is connected to the contact wire and the secondary Point B is
connected to the return line.
 For simplicity, the network will be assumed fed from a single source and current
energizations are used to model it. A 200 amps current is injected at the left end of the
contact wire, where the feeding source is assumed, while a negative 200 amps is
collected at that location on the return line.

Figure 9: Perspective view of HIFREQ model (Y scaled 1:500).

Model 1: This case corresponds to a train located at the center between two ATs separated by
15 km. The rail ballast resistance is 100 ohm-kft. Figure 10 shows the catenary system current
distribution as computed by the HIFREQ computation module. As it can be seen, the current

Page 21-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

distribution obtained from the computer model is as expected for the ideal case. The computer
model is therefore validated.

Figure 10: AT AC electric railway current distribution from HIFREQ.

Model 2: This case corresponds to a train located at a 75% distance between two ATs separated
by 15 km. The rail ballast resistance is 100 ohm-kft. As shown in Figure 11, when the train is at
different locations, the current distribution in the catenary and rails will change compared to the
center location.

Figure 11: Current distribution in the affected region varies with the train location.

Model 3: This case corresponds to a train located at the center between two ATs separated by
5 km. The rail ballast resistance is 100 ohm-kft. The current distribution is very close to what we
should expect in general for a single-end fed autotransformer electric railroad system (Figure 12).
The current imbalance region is restricted to the region between two ATs where the train is
present. Everywhere else outside the affected region, the system network net current is almost
zero.

Copyright © 2017 SES ltd. All rights reserved. Page 21-9

PART III: SES CONTRIBUTIONS

Figure 12: Current distribution with multiple autotransformers.

Model 4: This case corresponds to a train located at the center between two ATs separated by
5 km. The rail ballast resistance is 2 ohm-kft (wet case). Figure 13 shows that the current
distribution can be significantly different from the high ballast resistance case due to the leakage
earth current. Moreover, the currents flowing in the rails vary a lot along the rails when the ballast
has a low resistance. The traction net imbalance current, therefore, can increase dramatically.

Figure 13: Current distribution in the affected region varies with the ballast resistance.

Finally, Figure 14 compares the resulting interference levels on the nearby pipeline between the
four cases modeled above. As can be seen, the level of interference from AT electric railways can
still be serious, despite this feeding scheme designed to reduce it. Induced potentials on
unmitigated exposed lines (pipelines or railways) can reach hundreds of volts at AT locations or
at other locations where the EMF becomes discontinuous, with a current of a couple of hundreds
amps flowing in the catenary line. Note that the catenary current can reach as high as 1,000 A in
a loaded system. This can result in significant pipe-to-earth or rail-to-soil voltages, which
represent a possible direct electrical shock hazard and can threaten safety along pipelines and
railways and jeopardize the integrity of the protection systems.
Consequently, it is essential to model realistically the traction network including transformers,
ballast conditions as well as train locations to evaluate accurately the AT type electric railroad
system EMI compatibility. The automation for moving train to obtain the envelope is difficulty to
do in HIFREQ. ROW, therefore, is the tool required for railway EMI assessment.

Page 21-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 13: Pipeline touch voltages.

5 Modeling AT Catenary Traction with Moving Train in

ROW/SPLITS
The circuit based SPLITS computation module (also available in ROW) allows modelling various
transformers at the central site and along terminals by effectively representing the transfer matrix
of the transformer using the built-in Self and Mutual impedances commands of SPLITS [2]. A set
of dedicated high-level commands (SCLF) and dedicated user-interfaces were introduced to ease
the specification of many types of transformers in SPLITS [3]. Modeling various transformers at
the central site was introduced in ROW a few years ago [4]. However, you probably also know that
you can build many types of single or three-phase transformers in SPLITS using the Self and
Mutual commands and place them anywhere you like in the model (any section, along a terminal)
[5].
We will build here a SPLITS computer model based on the network described in Section 3. The
current distribution in the catenary traction system as calculated by the SPLITS computation
module will be compared to the results obtained using the HIFREQ computation module.
The circuit model shown in Figure 14 contains two terminals corresponding to Electrical
Segmentation (Terminal 1) and Traction Substation (Terminal #2). The central site is selected just
outside of Terminal No. 2. Thus, Terminal 2 is essentially made of a very short length of catenary
line from the Central Station to the terminal and is created simply to define the energization of
the catenary system. There are a total of 7 phases (buses) in this circuit. Phase 4 (Bus 4) represents
the pipeline. Phases 1, 2, 3 (Buses 1, 2 and 3) correspond to the contact wire, the return wire and
the rail track, respectively. Phases 5, 6, 7 (Buses 5, 6, 7) are dedicated to the specification of the
single phase autotransformers. Phase 5 (Bus 5) for the primary winding, Phase 6 (Bus 6) for the

Copyright © 2017 SES ltd. All rights reserved. Page 21-11

PART III: SES CONTRIBUTIONS

secondary winding and Phase 7 (Bus 7) for other connections. A 100 m span length (SPLITS
section) is used, resulting in 300 sections for the 30 km corridor. Three AT transformers will be
located at Sections 1, 150 and 300, respectively. Figure 15 illustrates the circuit based
autotransformer model block for AT railway types.

Figure 14: Circuit model without autotransformers.

Page 21-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 15: Autotransformer SPLITS model block for AT railways.

The following provides the steps to build ROW/SPLITS circuit for the system described in Section
3.
1. Step #1: Build the ROW model without the transformers, and create the initial SPLITS
circuit model (i.e. SP_Scenario1_0.f05 input file) in Create Circuit module of ROW.
2. Step #2: Load the SP_Scenario1_0.f05 in the SYSTEM screen of the SPLITS input
Toolbox, then click the Transformer tab next to the Central Site Busses and
Terminals tabs, click the Add Transformer… button after selecting the Single-
Phase option. The single phase transformer definition screen appears. Define the
autotransformer as shown in Figure 16.
3. Step #3: Run/Process SPLITS. The program will generate a F72 file
(SP_Scenario1_0.F72) that contains the block of commands defining the transformer. See
Figure 17.

Copyright © 2017 SES ltd. All rights reserved. Page 21-13

PART III: SES CONTRIBUTIONS

Figure 16: Defining the autotransformer in SPLITS Input Toolbox.

Page 21-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 17: SPLITS command block for the single phase autotransformer (.F72 file).

4. Step #4: Edit SP_Scenario1_0.f05 created in Step 1 and insert the F72 file commands at
the appropriate sections, i.e., at Sections 1, 150, and 300. Then finalize the model:
 Delete the central site transformer terminal commands, add the commands for
Phases 5 to 8 at the central site and add the energizations for Phases 5 to 8
(transformer phases) in Terminal energization block (see Figure 18a);
 Add the 3 AT blocks at Section 1, 150, 300 (see Figure 18b);
 Insert the train by connecting Phase 1 (contact wire) to Phase 3 (rail) at Section
225.
The final circuit, i.e., Sp_Scenario1_0WithAts.f05, is shown in Figure 19.
Run SPLITS and display the computed results. The current distribution in the AT-feeding system
is as expected. It compares well with the results given by the HIFREQ model.
Consequently, the results from SPLITS and HIFREQ have been cross-validated.

Copyright © 2017 SES ltd. All rights reserved. Page 21-15

PART III: SES CONTRIBUTIONS

Figure 18a: Fix SPLITS model for the autotransformer.

Page 21-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 18b: Fix the SPLITS model by adding the 3 autotransformers.

Figure 19 shows the final SPLITS circuit that represents the complete network. The current
distribution is computed and is shown Figure 20. All results are as expected and match the results
from HIFREQ.
Similarly, another AT electric railway system is modelled in order to further validate the SPLITS
autotransformer computations. In this case, the autotransformers are 5 km apart (i.e., every
50 sections in ROW/SPLITS). As shown in Figure 21, the current distribution is still very close to
what we expect for a single-end fed autotransformer system.

Copyright © 2017 SES ltd. All rights reserved. Page 21-17

PART III: SES CONTRIBUTIONS

Terminal #1: Load

End
Terminal #2:
Energization
Source End

Figure 19: SPLITS circuit model with the 3 autotransformers.

Single Computation LEGEND

Contact : Bus/Line 1.
Return : Bus/Line 2.
300 Rail : Bus/Line 3.
Section Current Magnitude (Amps)

200

100

0
0 100 200 300 RunID:SV10 Term.:Terminal_1

Section Number

Figure 20: AT electric railway current distribution from SPLITS/ROW.

Page 21-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Single Computation LEGEND

Contact : Bus/Line 1.
Return : Bus/Line 2.
300 Rail : Bus/Line 3.

Section Current Magnitude (Amps)

200

100

0
0 100 200 300 RunID:SV10 Term.:Terminal_1

Section Number

Figure 21: Multiple autotransformers network current distribution from SPLITS/ROW.

In electric railway catenary systems, the catenary current distribution varies significantly
depending on the train location. Consequently, all possible train locations have to be examined in
an AC interference analysis. This typically requires that the connection between the contact wire
and the rail be modeled at regular intervals throughout the corridor shared with pipelines or other
metallic paths considered in the study. The monitor fault module of the Right-of-Way software
allows you to automatically create train locations along the catenary at any combination of group
of sections (regions/spans) with the possibility to define arbitrary section increments between
selected train locations as well as the equivalent impedance representing the train. The
computation results that are generated contain useful information such as induced potentials,
voltages and currents along the exposed conductors (pipeline, etc.). The maximum envelope of
the interference levels along the right-of-way are provided.
Define in the Define Faults a “fault” between Phase 1 (Contact Wire) and Phase 3 (Rail),
representing the train location. The “faults” (different train locations) will be simulated along
Terminal 1, from Sections 1 to 300 at every section. This will ensure that the maximum induced
pipeline potentials is accurately determined regardless of the train location (moving train). Click
on the Monitor Data button to define the exposed phase, which is usually the pipeline (Phase 4)
under study (See Figure 22). Figure 23 provides the envelope of the maximum pipeline potential
at all train locations, assuming that the contact wire carries 200 A current and a ballast of
100 ohm-kft resistance.

Copyright © 2017 SES ltd. All rights reserved. Page 21-19

PART III: SES CONTRIBUTIONS

Figure 22: Define a moving train in ROW Monitor Fault module.

Figure 23: Pipeline potentials during steady-state conditions, for all train locations.

Page 21-20 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

6 Conclusion
This paper described the EMI mechanisms of different electric railways. With a typical AT electric
system right-of-way example, the paper demonstrated how to model the AT electric tractions by
using HIFREQ and ROW modules in CDEGS software packages, and the computed results have
been validated. The successful AT transformer models constitute the essential modeling of the AT
electric railway, while the ROW Monitor Fault module provides the possibility to simulate the
moving train. User friendly interfaces for modeling transformers at any sections/spans in
SPLITS/ROW will be implemented in the near future.

7 Acknowledgment
Authors thanks colleague Ms. Marjan Mehrabi for their constructive assistance in the computer
modeling, Dr. Luis Valcarcel for his careful reading of the manuscript and useful suggestions.

8 References
[1]. L. Valcarcel, S. Lefebvre and F. Dawalibi, “A Simplified Method for Computing Equivalent
Parameters for Rail Ballast Resistance in Inductive and Conductive Models” in CDEGS Users'
Conference Proceedings, Montreal, Quebec, Canada, 2010.
[2]. J. Ma, “The New Transformer Modeling Feature in SPLITS” in CDEGS Users' Conference
Proceedings, Gaithersburg, Maryland, USA, 1995.
[3]. J. Ma, “Modeling more Transformer in SPLITS” in CDEGS Users' Conference Proceedings,
Austin, Texas, USA, 1996.
[4]. M. Mehrabi, M. Boisjoli and S. Fortin, “Enhancements to Three-phase Transformer Models
in SPLITS” in CDEGS Users' Conference Proceedings, San Diego, California, USA, 2015.
[5]. R. Southey, M. Siahrang, “Having Fun with Transformers and Sequence Impedance
Networks… The SPLITS Way” in CDEGS Users' Conference Proceedings, 2014 Newport, Rhode
Island, USA, 2014.

Copyright © 2017 SES ltd. All rights reserved. Page 21-21

 UGM 2017 – CANNON BEACH, OREGON

22
MODELING AND MITIGATION DESIGN
CONSIDERATIONS FOR AC INTERFERENCE STUDIES IN
LOW SOIL RESISTIVITY

Christian Voyer, Maxime Daigle and Yexu Li

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract

Soil resistivities in the single 𝛀 ⋅ 𝒎 range have a noticeable impact on the performance of a mitigation system
compared to a soil with a more moderate resistivity, and can pose a substantial design challenge in mitigating for
AC enhanced corrosion. This paper highlights the key differences in mitigation behaviour in moderate and low soil
resistivity, and demonstrates the importance using the most complete computation available, Total Interference in
Right-of-Way Pro, to obtain accurate results.

1 Introduction
AC interference studies have been increasingly focused on reducing the risk of AC enhanced
corrosion. Currently, workers generally target a coating stress voltage value of 15 V or less, but
also have to reach a leakage current density for a 1 𝑐𝑚2 holiday of 30 𝐴/𝑚2 or even 20 𝐴/𝑚2. This
second limit can be particularly daunting to achieve when the pipeline is in contact with a low soil
resistivity soil layer.
As we will demonstrate in the pages that follow, habits developed from experience in more
moderate to high soil resistivities (100 Ω ⋅ 𝑚 or higher) can be misleading. More specifically,
considering inductive coupling only computations for steady state conditions in SPLITS is often
inaccurate, and considering conductive coupling interactions on finely segmented conductors as
well, is necessary.

2 General Considerations

2.1 Coating Stress Voltage Targets are Low

Using the NACE standard for the holiday resistance computation, the target pipeline coating
stress voltage given a holiday of area 𝐴 in contact with a soil of resistivity 𝜌 for a leakage current
density 𝐽 is expressed in Equation (1).

𝐽𝜌√𝜋𝐴
𝑉𝐶𝑜𝑎𝑡𝑖𝑛𝑔𝑆𝑡𝑟𝑒𝑠𝑠 = (1)
4

Substituting 𝐴 for 1 𝑐𝑚2 and consolidating the constants leads to simple equations to calculate
the target coating stress values for a limit of 20 𝐴/𝑚2 (Equation (2)) or 30 𝐴/𝑚^2 (Equation (3)).

Copyright © 2017 SES ltd. All rights reserved. Page 22-1

PART III: SES CONTRIBUTIONS

𝑉𝐶𝑜𝑎𝑡𝑖𝑛𝑔𝑆𝑡𝑟𝑒𝑠𝑠 (20 𝐴/𝑚2 ) = 𝜌 × 0.08862 𝐴/𝑚 (2)

𝑉𝐶𝑜𝑎𝑡𝑖𝑛𝑔𝑆𝑡𝑟𝑒𝑠𝑠 (30 𝐴/𝑚2 ) = 𝜌 × 0.13293 𝐴/𝑚 (3)

For example, in 5 Ω ⋅ 𝑚 soil, the coating stress target for a leakage current density of 20 𝐴/𝑚2 for
a 1 𝑐𝑚2 holiday is 0.44 𝑉 following this standard.

2.2 Accentuated Potential Drop along Mitigation Wire

Gradient control wires are particularly attractive in mitigating AC interference not only because
they reduce the pipe GPR, but mostly because they help equalize the soil and pipeline potentials,
thereby reducing coating stress voltage.
One key feature of low resistivity soil though, is that it can lead to pronounced potential
differences along conductors, even over seemingly short distances. In the case of mitigating
pipeline coating stress voltages, the mitigation wire GPR can vary greatly between connections to
the pipe. The GPR at a given location (especially between connections) can be much lower than a
well-insulated pipe at the same location, reducing the gradient control benefits.
The computation results in Figure 1 demonstrate the effect of reducing soil resistivity on the
mitigation GPR. A coated (1,000,000 Ω ⋅ 𝑓𝑡 2 ) 8 inch pipeline was modeled at a depth of 6 𝑓𝑡 (to
pipe center) with an imposed nominal GPR, and connected every 1,000 𝑓𝑡 to a zinc ribbon
mitigation wire, in 100 Ω ⋅ 𝑚 and 5 Ω ⋅ 𝑚 uniform resistivity soil.
In the case of 100 Ω ⋅ 𝑚 soil, the GPR between connections of the zinc ribbon mitigation wire falls
but only by about 1% compared to the pipeline GPR, whereas in 5 Ω ⋅ 𝑚 soil, it drops by more than
20%. When coupled with a commensurate factor of 20 drop in tolerable coating stress voltage,
mitigating a pipe in 5 Ω ⋅ 𝑚 soil can be quite a challenge.

Figure 1: Mitigation wire GPR as a percentage of the pipeline GPR between connections to a pipe in
100 Ω ⋅ 𝑚 and 5 Ω ⋅ 𝑚 soil.

Page 22-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Using the results shown in Figure 1, we can express the coating stress voltage as a percentage of
the pipe GPR. The results are shown in Figure 2. The mitigation is not only less effective in the
area between connections to the pipeline, but the effect is greatly enhanced in low resistivity soil.

Figure 2: Coating stress voltage on a pipeline as a percentage of the pipeline GPR.

It is worth noting that there is a noticeable end effect starting from the origin. The earth potential
just outside the pipe is lowest near the ends of the mitigation, an effect that can take a rather large
distance to even out, as shown in Figure 3. Note that the plot shows only the first 10,000 ft, but
mitigation in the model continues well past 10,000 ft. The same end effect observed near a
distance of 0 ft is observed at the far end.

Figure 3: Soil potential directly outside the pipeline as a percentage of the pipeline GPR. Note that in this
example, the mitigation ends at a distance of 0 ft, but continues well beyond 10,000 ft. The same end
effect is seen on the other end of the mitigation.

Copyright © 2017 SES ltd. All rights reserved. Page 22-3

PART III: SES CONTRIBUTIONS

3 Modeling Considerations in Right-of-Way Pro

3.1 Soil Model

First and foremost, as soon as the soil resistivity measurements have been interpreted and the soil
models derived, a quick calculation of the target coating stress voltage from Equation (2) or (3)
will help indicate the likely problematic areas.
While it is always important to carefully inspect soil resistivity data and the soil models derived
from them, an additional sticking point arises in the context of leakage current density for 1 𝑐𝑚2
holiday calculations. Consider the soil model fit shown in Figure 4. The soil data used is known to
be deficient in short spacings, though for AC interference studies one might be tempted not to
dwell on this point since the grounding impedance of mitigation wires will not be severely
impacted.

Figure 4: Initial fit to the soil resistivity data. It is clear that the program interprets several layer
boundaries in an area where few measurements exist, exposing a lack of short spacing measurements.

Figure 5 is the side view of the MALZ file automatically created by Right-of-Way (ROW) for the
SPLITS circuit creation process. The pipeline (though not directly modeled here) comfortably sits
in the 143.3 Ω ⋅ m layer. Consequently, for a 20 𝐴/𝑚2 leakage current density on a 1 𝑐𝑚2 holiday
threshold, the target coating stress voltage is 12.7 𝑉. The unit length grounding impedance of the
wire is computed to be 206 Ω ⋅ 𝑚.

Page 22-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Side view of the MALZ file automatically created by ROW for circuit creation, which illustrates
the soil resistivity in contact with the pipe based on the fit shown in Figure 4.
Another individual, following the same procedure on the same data, obtained the fit shown in
Figure 6.

Figure 6: Independently derived fit from the same data as that used in the fit shown in Figure 4. The fit
quality is basically the same, and while the derived structure is different, it is not strikingly so.

Again, given the known short spacings deficiency, it is not surprising that the fit is different. There
is a non-negligible impact on the unit length grounding impedance, which is now 170 Ω ⋅ 𝑚, but
Figure 7 illustrates the most important change.

Copyright © 2017 SES ltd. All rights reserved. Page 22-5

PART III: SES CONTRIBUTIONS

Figure 7: Side view of the MALZ file automatically created by ROW for circuit creation, which illustrates
the soil resistivity in contact with the pipe based on the fit shown in Figure 6.

Looking at the side of the MALZ file created by ROW to determine the mitigation wire grounding
impedance shows that now the lowest soil resistivity in contact with the pipe is now 17.15 Ω ⋅ 𝑚.
Consequently, for a 20 𝐴/𝑚2 leakage current density on 1 𝑐𝑚2 holiday threshold, the target
coating stress voltage is now 1.5 𝑉. A minor fitting difference will lead to wildly different designs.
It is therefore crucial to first ensure that appropriate soil resistivity measurements have been
made in order to properly determine the soil in likely in contact with the pipe. Then, to inspect
the soil resistivity fits in order to ensure that a low soil resistivity layer is not ignored simply
because a refinement happened to place it a few inches away from the pipe.

3.2 Total Interference vs SPLITS Results

To maintain design flexibility, it is common practice in RowCAD to define all likely mitigation
arrangements as satellites to the pipeline conductor in the pipeline path. Adding and removing
mitigation in various areas is then much quicker than recreating the circuit for each mitigation
extent refinement. A consequence of this strategy though, is that a given section in SPLITS does
not contain information about the existence of mitigation in neighbouring sections.
In an effort to provide a preview of coating stress, touch voltage, and leakage current density for
a 1 𝑐𝑚2 holiday results, a feature was added to ROW to estimate these results making use of the
MALZ model that computes the leakage impedance of mitigation wire phases (the feature is
described in [1]). Even an approximate result is attractive if it uses SPLITS computations, given
that it tends to run substantially faster than MALZ, allowing for a good starting mitigation design
which can then be refined by running Total Interference.
A token length of mitigation wire is modeled and energized at its ends and middle, while profile
points are placed 30 ft from the origin around the pipe coordinates (though the pipe is not
explicitly modeled, i.e. assuming that pipe is well coated.). Figure 8 shows a top view of the MALZ
model. The y axis has been stretched to better show the observation points that will be used to

Page 22-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

obtain an estimate of the touch-voltage-to-GPR percentage and the coating-stress-voltage-to-

GPR percentage.
A side view of the arrangement is shown in Figure 9. The program compares the soil potential at
the four points surrounding the pipe (even though the pipe is not present in the model) and
returns the worst case result of coating-stress-to-energization-GPR percentage. Likewise for the
touch voltage using the observation points at the surface of the soil and 1 m away from the center
of the pipe.
The upshot of this strategy is that it is naturally limited by having to select a single location at
which it can make this determination. Given the result shown in Figure 2, it is not possible for the
program to take advantage of the full accuracy typically available from MALZ.

Figure 8: Top view of the MALZ model automatically created by ROW to compute leakage resistance of
bare buried conductors when performing a circuit creation. The model consists of three energizations (all
on the same bus) from which the program can determine the leakage impedance. Observation points are
added to estimate the touch-voltage-to-GPR percentage and the coating-stress-voltage-to-GPR
percentage.

Copyright © 2017 SES ltd. All rights reserved. Page 22-7

PART III: SES CONTRIBUTIONS

Figure 9: Side view of the MALZ model automatically created by ROW to compute leakage resistance of
bare buried conductors when performing a circuit creation.

Figure 10: Coating stress voltage on a pipeline as a percentage of the pipeline GPR, including the
automatically computed percentage from a ROW circuit creation.

Figure 10 overlays the Figure 2 results with the percentage automatically calculated in ROW for
the same configuration. In typical applications, the consequences tend to be overdesign rather
than under-design, but again, low resistivity soil is more likely to cause issues. Moreover, workers
have often expressed surprise that when a design is created based on SPLITS results (using the
percentage files) and then run in Total Interference, the results can be noticeably different.

Page 22-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

It is also worth addressing two habits that are commonly seen in practical ROW that cause
particular issues in low resistivity soil. First, the connections between the pipeline and mitigation
wires are often entered for every x section(s). Unless the model has been carefully designed such
that the section length does indeed correspond to connection locations, the number of
connections will likely be over, or under-represented. At the very least, the option of entering
connections based on distance is a far more reasonable starting point. Of course, the most precise
model will be one the connections are at the correct locations in the Total Interference MALZ
model, usually via the MALZ template.
Second, RowCAD, and in turn ROW, will segment the pipeline path (and therefore the pipeline
and mitigation conductors) at every path node, but also every section. Section cuts can often be
roughly equal to span length, and if a pipeline is fairly straight, there may be few, if any, nodes
between section boundaries. This will not impact SPLITS results. Total Interference will create
additional segmentation around the fault location (under fault conditions), and around the first
section of each soil region (under steady state conditions), and so workers have often not paid
particular attention to this issue. In low soil resistivity under steady state conditions though, since
only one Total Interference MALZ model per soil region will be created, and since the potential
drop along the mitigation conductor will be much more pronounced, it is important to ensure that
the original pipeline path is well segmented all throughout the model (either in the original model,
or by requesting further segmentation in the Advanced tab in Total Interference).
Figure 11 illustrates the impact of segmentation on the computed coating stress voltage (as
percentage of pipeline GPR) for finely segmented (20 ft), reasonably well segmented (100 ft) and
poorly segmented (500 ft) pipelines in MALZ.

Figure 11: Coating Stress Voltage as a percentage of GPR for for finely segmented (20 ft), reasonably well
segmented (100 ft) and poorly segmented (500 ft) pipelines in MALZ, in 5 Ω ⋅ 𝑚 soil.

As illustrated in Figure 12, the effect is also present in 100 Ω ⋅ 𝑚 soil, though it has a far smaller
impact.

Copyright © 2017 SES ltd. All rights reserved. Page 22-9

PART III: SES CONTRIBUTIONS

Figure 12: Coating Stress Voltage as a percentage of GPR for for finely segmented (20 ft), reasonably well
segmented (100 ft) and poorly segmented (500 ft) pipelines in MALZ, in 100 Ω ⋅ 𝑚 soil.

All of the effects demonstrated above cannot be easily considered in SPLITS, though they are
properly accounted for in MALZ, assuming a proper segmentation is applied either in ROWCAD
or in Total Interference. This is indeed precisely why ROW has recently been outfitted with a
Steady State option in Total Interference (the feature is described in detail in [2]). Note as
well that these are often the reason for which SPLITS and Total Interference computation
results can appear to disagree, is because Total Interference (as the name suggests) is quite
simply a more complete computation. As stated above, using SPLITS results to compute touch
voltage, coating stress voltage, or leakage current density of 1 𝑐𝑚2 holiday can be a useful starting
point and provides quick and generally conservative design, and under certain circumstances
provide similar results to Total Interference, but the latter will always provide more accurate
and reliable results.

4 A Worked Example
As a final demonstration of these effects in action, a 10 mile long joint-use corridor was modeled
in 5 Ω ⋅ 𝑚 uniform soil in HIFREQ and ROW. A sketch of the modeled layout is shown in Figure
13.

Page 22-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 13: Sketch of the HIFREQ/ROW validation model.

An 8 inch, well coated (500,000 Ω ⋅ 𝑓𝑡 2 ), pipeline was modeled a relatively short (70 𝑓𝑡) distance
away from an overhead line with a horizontal arrangement carrying a balanced 1 𝑘𝐴 current. The
configuration includes a crossing midway.
The induced GPR on the pipeline without mitigation installed computed in SPLITS under these
conditions actually matches the full HIFREQ computation well, as shown in Figure 14.

Figure 14: Pipeline GPR results with no mitigation installed for the HIFREQ, and ROW (SPLITS)
computations.

The issues considered above though, are most evident with bare conductors. Adding a zinc ribbon
mitigation wire to the pipeline in both models exposes these issues. Focusing on the boundary
between the mitigated and unmitigated portions on the pipeline in Figure 15, there is a substantial
deviation between the HIFREQ and SPLITS results, even though ROW does incorporate
conductive coupling to the SPLITS results using the technique described in section 3.2.

Copyright © 2017 SES ltd. All rights reserved. Page 22-11

PART III: SES CONTRIBUTIONS

Figure 15: Pipeline coating stress voltage results with a single zinc ribbon mitigation wire added. The
results obtained from SPLITS far overestimate the coating stress voltage, while a well-segmented Total
Interference model is a far better (and still conservative) match.

Note as well that special care was given in ensuring that the locations of pipeline-to-mitigation
connections was identical in all ROW and HIFREQ models. A more complete consideration of
conductive coupling in Total Interference fares far better, though, as shown demonstrated
above, a proper segmentation has a noticeable impact on results.

5 Mitigation Design Considerations

The effects described above will also naturally have an impact on mitigation design. Given the
stringent coating stress voltage limits, it is becoming more habitual to see in excess of two
mitigation wires along the pipe. SES has even been involved in a proposed design that consisted
of six parallel mitigation wires with vertical rods in addition, shown in Figure 16.

Figure 16: A project case of extensive mitigation needed to achieve the now typical stringent coating stress
voltage limits.

It is tempting to simply increase the number of mitigation wires to lower the pipeline GPR and
try to improve the gradient control provided by the mitigation system, though doing so does not
help where the issues may be most pronounced: between pipeline-to-mitigation connections.
While often more costly than adding additional horizontal mitigation wires, adding more closely

Page 22-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

spaced PCRs or SSDs is often the most direct way to counter the potential drop effect on the
mitigation wire illustrated in Figure 1.
Adding vertical rods, such as the ones shown in Figure 16, often also provide better gradient
control in the soil around the pipeline than adding additional horizontal mitigation wires. How
much each measure improves the coating voltage though, will depend on the details of the
individual case.
Another concern is the increased importance of tower grounds. While it is already known that
accounting for conductive interference from tower grounds can be important even under steady
state conditions [3], low soil resistivities combined with low coating stress voltage thresholds
enhances the influence of nearby tower grounds.
Horizontal directional drilled sections of pipeline also pose a particular mitigation design
challenge, in that it is often impossible to add parallel mitigation wires, let alone regularly
connected ones. SES recently encountered a case for which an HDD pipeline section was exposed
to a soil resistivity layer measured to be 2 Ω ⋅ 𝑚 (wetlands). An option that has been discussed,
though we have yet to have confirmation that this was put into practice was to incorporate alkaline
earth salts in the drilling mud to offset the low resistivity of the mud.

6 Conclusion
The work presented in this article demonstrates that it is important to appreciate the impact of
low resistivity soil on the behaviour of mitigation wires. It serves to underline the importance of
using the most accurate computation strategies available in order to avoid costly overdesign, or
underdesign if conductive component is important, such as nearby a substation. Even under
steady state conditions, there is undeniable value in running a Total Interference analysis,
rather than settling with results from SPLITS alone. Given that the more accurate computation
methods are already fully integrated within the Right-of-Way Pro software package, workers
should be able to incorporate these effects in their studies.

7 References

[1] Y. Li, "Improvements in Right-of-Way Program," in CDEGS Users' Conference Proceedings, Lake
Tahoe, California, USA, 2013.
[2] Y. Li and F. P. Dawalibi, "New Features in Right-of-Way Pro," in CDEGS Users' Conference
Proceedings, Boulder, Colorado, USA, 2016.
[3] Y. Li and F. P. Dawalibi, "Increasing the Computation Accuracy of AC Interference Level under Power
Line Load Conditions by Accounting for the Conductive Coupling," in CDEGS Users' Conference
Proceedings, Rock Creek Resort, Montana, USA, 2012.

Copyright © 2017 SES ltd. All rights reserved. Page 22-13

 UGM 2017 – CANNON BEACH, OREGON

23 COMPUTATION OF GEOMAGNETICALLY INDUCED

CURRENTS IN POWER SYSTEMS USING SES
MULTIFIELDS SOFTWARE PACKAGE
Rachid El Hani, Simon Fortin and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
Geomagnetically induced current (GIC) in power systems causes an offset of the origin of magnetization curve of
power transformers. This offset leads to large magnetizing currents and distorted wave shapes that produce higher
magnitudes of flux leakages that generate even-order harmonic currents resulting in high eddy and circulating
current losses in the windings and structural parts of the transformer. The net results of these effects is increase of
the transformers temperature, loud audible noise from the transformers and unwanted relay operations tripping
out power lines. In this paper, the computation of GIC in transformer windings using SES’ MultiFields software
package is presented. A hypothetical power network energized by a realistic, measured, one minute variation of
geomagnetic field, is studied. Mitigation scenarios using blocking capacitors are discussed.

1 Introduction
The existence of sunspots was discovered by Galileo in 1610. These sunspots appear as dark spots
on the photosphere of the Sun. they are areas of reduced surface temperature where the magnetic
field flux is concentrated. Figure 1 illustrates how the solar sunspot storms interfere with Earth’s
magnetic field. The solar storm of 1859 was the first recorded geomagnetic disturbance (GMD)
during which telegraph systems over Europe and North-America failed. Such major solar
disturbances occur every 10-11 years sunspot cycle.

Figure 1: A Coronal Mass Ejection (CME) interacting with Earth’s magnetic field
(Image Credit: NASA https://www.nasa.gov/mission_pages/sunearth/news/storms-on-sun.html).

The geomagnetically induced currents (GIC) caused by a GMD have wide range of damaging
effects on the power system, communication systems, railway systems, and contribute to the
corrosion of metallic pipelines. A destructive solar storm in power system history occurred on

Copyright © 2017 SES ltd. All rights reserved. Page 23-1

PART III: SES CONTRIBUTIONS

March 13, 1989 when a K9 level storm resulting from a very large coronal mass ejection (CME) a
few days before, caused a 9-hour blackout of the Hydro Quebec power grid (see [1]).

1.1 GIC in Power Transformers

Figure 2 illustrates how the flow of the induced dc current in transformer windings causes an
offset of the origin of the magnetization curve of the transformers. This offset produces half-cycle
saturation of the transformer cores which leads to a larger excitation current and then to an
increase of the reactive power absorbed by the transformer. Consequently, the induced DC ground
current in transformers generates even-order harmonic currents that causes static compensators
to trip or shut down. Loss of static compensators gives rise to system instability (see [2]).

Figure 2: The saturation of a transformer in the presence of GIC (Figure 1b in [2]).

1.2 Computation process

The typical steps of the computation process of a GIC study involving power networks using
MultiFields are as follows.
1. The target geomagnetic variation in time domain of a realistic GIC event should be
obtained first from appropriate geomagnetic observatories. Next, the corresponding
geomagnetic field in frequency domain is then obtained using a Fast Fourier Transform
(FFT) using the FFTSES computation module.
2. The recommended frequencies computed by FFTSES are used to compute the frequency
response of the power system using the HIFREQ computation module. Using an
equivalent multilayer earth resistivity model of the site of interest, HIFREQ computes the
induced currents in transformers’ windings.
3. Then, using FFTSES, the unmodulated and modulated induced currents in frequency
domain are converted to time domain results using the Inverse Fast Fourier Transform.
The block diagram of the modelling process steps are shown in Figure 3. SES AutoTransient tool
automates this process. The computation is performed for northward and eastward magnetic
fields. The northward GIC is computed using eastward geomagnetic fields and the eastward GIC
is computed using northward geomagnetic fields. The computed Northward and Eastward
components are used to compute the GIC on transformer windings as a function of the direction
of the geomagnetic field:

Page 23-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

GIC(θ) = cos(θ) GICN + sin(θ) GICE (1)

where 𝜃 is the clockwise angle measured from the North and horizontal components of the electric
field.

Forward
Geomagnetic 𝑩(𝝎)
Field B(t)
FFTSES

Frequencies

Inverse
𝑮𝑰𝑪(𝒕)
FFTSES

Earth Structure
Model

HIFREQ 𝑮𝑰𝑪(𝝎)
Power Network
Topology

Figure 3: The process steps of calculating GIC in power systems.

2 Geomagnetic Data
Most North American power utilities use K and A indices to indicate the intensity of solar
disturbances [3]. The K-index is related to the maximum fluctuations (in nT) of the horizontal
components measured at a geographical location relative to a quiet day, during a three-hour
interval. The A-index is a daily average level describing the local geomagnetic activity. Since it is
not meaningful to take the average of a set of K-indices. These indices are converted to the
equivalent three hourly range a-index (note the lower case “a” which is a three hour average level
and the upper case “A” which is a daily average level).
The study in [4] demonstrates a high correlation between GIC and the a-index (3-hour index).
The authors in [4] and [5] point out that the large induced electric filed 𝐸⃗ is related to the time
⃗
𝜕𝐵
derivative of the magnetic field − . This time derivative is designated as the rate-of-change.
𝜕𝑡
⃗
𝜕𝐵
According to Faraday’s law, the area integral of 𝜕𝑡
is proportional to the electromotive force.
Observatories around the word are measuring the geomagnetic field at the earth surface. These
measurements are provided in Canada by Space Weather Canada [7] and in USA by United States
Geological Survey [7]. On March 13, 1989, the geomagnetic fields measured at Yellowknife
observatory in Canada (YKC, located at Latitude 62.480o and Longitude 245.518o), are given in
Figure 4.

Copyright © 2017 SES ltd. All rights reserved. Page 23-3

PART III: SES CONTRIBUTIONS

Figure 4: Measured geomagnetic field at YKC observatory on March 13, 1989

(X is the northward field, Y is the eastward field, Z is the vertical downward field and F is the scalar
magnetic field). Provided by Space Weather Canada [7].

3 Earth Resistivity Models

Since the skin depth of the geomagnetic field can range from a few km to hundreds’ of km, it is
necessary to know how the earth resistivity model varies with depth to estimate the geoelectric
field. At very low frequencies the magnetic field pass through top soil thin layers unaffected.
Therefore, the resistivities of the thin top layers are not very important for GIC studies. However,
when grounding systems and grounded towers are modeled, the resistivities of the top layers play
a major role in determining current flow in power systems conductors and therefore, should be
included in the earth’s resistivity models.
The earth resistivity model used in GIC studies can be computed using robust inversion technique
from magneto-telluric studies [8]. In USA, the earth resistivity models are available for some
regions from the United States Geological Survey website
(https://geomag.usgs.gov/conductivity).
As far as we know, these earth models are not provided by Space Weather Canada. However, the
computed geoelectric fields are available for Canadian geomagnetic observatories based on the
measured one minute variations of the geomagnetic field. Some resistivity models for west

Page 23-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Canada are presented in [9]. Figure 5 gives the resistivity model for the Mackenzie River valley
area (Zone 3, Figure A4.1 and Table A4.3 in [9]).

Figure 5: 1D Resistivity Model for Zone 3 – Mackenzie River valley

(From Figure A4.1 and Table A4.3 in [9]).
The soil layer resistivities and thicknesses at Mackenzie River valley are given in Table 1.
Table 1: Layers Resistivities and thicknesses.
Layer Resistivity (Ω-m) Thickness (m)
1 55 6
2 60 144
3 20 1,450
4 1000 13,400
5 300 25,000
6 3,000 60,000
7 300 300,000
8 10 Infinite

4 Geoelectric Field Computation

The geoelectric field response can be computed using HIFREQ for the recommended frequencies
provided by the Forward FFTSES process performed on the provided time domain geomagnetic
field. This is carried out by specifying the plane-wave energization and the known earth resistivity
model. After completion of HIFREQ computations on this model, the Inverse FFTSES process is
Copyright © 2017 SES ltd. All rights reserved. Page 23-5
PART III: SES CONTRIBUTIONS

used to calculate the geoelectric field in time domain at the surface of the earth. The SES
AutoTransient Tool can be used to automate the computation process. To validate this
computation process, the measured geomagnetic field of Figure 4 is used to compute the
geoelectric field using MultiFields. The computed electric field is compared to the geoelectric field
provided by Space Weather Canada at YKC observatory on March 13, 1989. The northward and
eastward components of this field are given in Figure 6 and Figure 8 respectively. Since the
Mackenzie River valley (zone 3 in [9]) is close to YKC observatory, the resistivity Earth model in
Figure 5 is used in this model. Note that we do not have the earth resistivity model at YKC
observatory and the model in Figure 5 is an approximation. The northward geoelectric field (X
component) is computed based on the eastward geomagnetic field (Y component) in Figure 4 and
eastward geoelectric field (Y component) is computed based on the eastward geomagnetic field
(X component) in Figure 4.

Figure 6: Eastward geoelectric field provided by Space Weather Canada at YKC observatory on March 13,
1989.

Figure 7: Eastward geoelectric field computed by SES Software using the northward geomagnetic field of
Figure 4 and the earth resistivity model of Figure 5.

Page 23-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 8: Northward geoelectric field provided by Space Weather Canada at MEA observatory on March
13, 1989.

Figure 9: Northward geoelectric field computed by SES Software using the eastward geomagnetic field of
Figure 4 and the earth resistivity model of Figure 5.
The components of the computed geoelectric field using SES software are given in Figure 7 and
Figure 9. Figure 6 to Figure 9 show that the computed geoelectric field is very similar to the
geoelectric field provided by Space Weather Canada at YKC observatory.

5 GIC model
Since the frequencies of the geomagnetic field are very low, circuit theory is often used to compute
GIC in power systems. In these studies the induced electric field is modeled as a dc voltage 𝑉 in
series with the transmission line resistances. Assuming that the electric field is uniform over the
entire length of the line, the line voltage is calculated by the following expression:

𝑉 = 𝐸𝑁 𝐿𝑁 + 𝐸𝐸 𝐿𝐸 (2)

Where 𝐿𝑁 and 𝐿𝐸 are the northward and eastward components of the line length respectively and
𝐸𝑁 and 𝐸𝐸 are the northward and eastward components of the computed geoelectric respectively.
To use this circuit approach, the maximum northward and eastward electric fields are estimated
based on the compute variation of the geomagnetic field at the surface of earth as shown in Section
4. This approach can be handled using HIFREQ by specifying the X and Y components of a static
external field energization, 𝐸𝐸 and 𝐸𝑁 respectively. Using this energization, we do not need to
compute the line voltage of each piece of line in Equation (2). To simulate this case using a circuit

Copyright © 2017 SES ltd. All rights reserved. Page 23-7

PART III: SES CONTRIBUTIONS

model, grounding systems can be replaced by lumped resistances. However, HIFREQ is much
more accurate than circuit models. It takes into account many aspects of the interference effects
that circuit models cannot handle:
 Shield wires, buried grounding grids and towers can be included in the model.
Surrounding buried metallic structures, pipes and conductors can be modelled as well.
 Pipelines, railways and transmission lines that share the same right-of-way can be
included in the same model and the conductive interference through earth between all
components, is taken into account.
 The measured geomagnetic field at observatories is directly used as input data in the
model. Transient analysis can be carried out using the AutoTransient tool that automates
the computation process by combining the HIFREQ and FFTSES computation modules
during the automation cycle. This process is illustrated in Figure 3.

5.1 Power System Network

In this section, a fictitious simple power network is studied to compute GIC in transformers’
windings using a transient analysis approach. The power network described in this document is a
network of 725 kV, 500 kV and 345 kV transmission lines and 10 transformers. Figure 10 shows
the single line diagram of the network. The input files are available in the folder
\ExampleFiles\Computation of GIC in Power Systems.

Sub 1 Sub 2 Sub 4

Line 1 Line 4 Line 8
G
T1 T3 T7

Line 2 Line 3
G
Sub 6
T2 T4
G
Sub 5 T8
Sub 3
Line 5 Line 6
G
T5 T9
Sub 7
Line 7
G
T6 T10
Figure 10: Single line diagram of the power system network.

5.1.1 Substations
The substations’ locations and grounding grids are given in Table 2. The grids at all substations
are modelled as an array of 5 horizontal conductors and 5 vertical conductors. The depth of all
grids is 0.5 m. The dimensions of all grounding grids are is 350 × 350 m square areas. The
computed grid resistance, assuming the earth resistivity model of Mackenzie River valley, is
about 0.1 Ω.

Page 23-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 2: X and Y coordinates of the substations’ centers.

X Coordinate Y Coordinate
Substation
(in km) (in km)
1 265 375
2 265 255
3 125 30
4 225 -120
5 -175 0
6 -315 -90
7 -175 -195

5.1.2 Transmission lines

The phase resistances of the transmission lines are provided in Table 3.
Table 3: The resistances of the transmission lines.
Line Resistance (in Ω/km/phase)
725 kV 0.018354
500 kV 0.036547
345 kV 0.073255

The shield wires are assumed to be identical and correspond to the AFL OPGW conductor type.
The relative resistivity of the shield wire (with respect to copper) is 4.08901 and its outer radius
is 8.8519 mm.

5.1.3 Transformers
The resistance of the high-voltage (HV) and low-voltage (LV) windings of the transformers are
given in Table 4.
Table 4: Transformers’ windings resistance per phase.
Transformers Resistance HV (Ω/phase) Resistance LV (Ω/phase)
∆−𝑌
0.1 NA
(T1, T2, T8 to T10)
𝑌−𝑌
0.06 0.04
(T3 to T7)

5.2 Results
The computed GICs in the ground grids at Substations 2, 4, and 6 are plotted in Figure 11. They
show that the maximum GIC reaches 54 A, 76.5 A and 102.5 A at Substations 2, 4, and 6,
respectively. Figure 12 shows that the maximum GIC in Lines 1, 4 and 7 are 24.2 A, 23.8 A and
13.8 A per phase, respectively.
The GIC in the transformer windings are given in Table 5. It reports that the maximum GICs are
between 10.8 A and 24.2 A per phase. The right column in this table, indicates the transmission
lines that are connected to the corresponding transformer winding. These values show how GICs
in transformers’ windings are affected by the direction of the geoelectric field. For example, they
show that Transformers T1 to T4 are more sensitive to a Northward geoelectric field (Eastward
geomagnetic field) and that Transformer T7 is sensitive to both directions of the magnetic field.

Copyright © 2017 SES ltd. All rights reserved. Page 23-9

PART III: SES CONTRIBUTIONS

Note also that the total GIC exceeds 15 A per phase in 7 transformers and exceeds 20 A per phase
in 5 transformers.

(a)

(b)

(c)

Figure 11: GIC (in A) at substations grids 2, 4 and 6.

Page 23-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

(a)

(b)

(c)

Figure 12: GIC (in A/phase) in lines 1, 4 and 7.

Copyright © 2017 SES ltd. All rights reserved. Page 23-11

PART III: SES CONTRIBUTIONS

Table 5: GIC in transformers’ windings.

Maximum GIC in transformer windings

Transformer (Amps/Phase)
Windings
Maximum Eastward Northward
T1, T2, T3 HV & T4
24.2 2.0 24.2 Lines 1 & 2
HV
T3 LV & T4 LV 22.4 10.6 22.3 Lines 3 & 4
T5 HV, T6 HV 10.8 8.7 10.4 Line 3
T5 LV & T6 LV 14.6 14.2 5.5 Lines 5 & 7
T7 HV 23.8 3.4 23.8 Line 4
T7 LV 16.3 16.3 0.4 Line 8
T8 and T9 15.8 15.8 1.5 Lines 6 & 8
T10 13.8 13.1 7.5 Line 7

6 Mitigation Measures
Since the frequencies are very low, the GIC is affected mainly by resistances of the network.
Capacitors block the flow of this quasi-dc current. Consequently, series capacitors are used in
power systems to redirect the GIC flow and improve the stability of the system. The number of
possibilities on how series capacitors can be placed in large power system is significant. Taking
into account that series capacitors are expensive, their number should be optimized by installing
them only on critical transformers. The simple case studied in the previous section includes 8
transmission lines. Therefore, the number of alternative configurations of the installation of series
capacitors along the lines represents 28 = 256 possibilities.
Depending on the given maximum GIC in the transformers’ windings, the goal of the mitigation
measures is to reduce the number of transformers with this maximum GIC using a minimum
number of series capacitors. The blocking capacitors ratings used in the following possible
scenarios are 10 𝜇𝐹. The mitigation objective is to keep the maximum GIC current in a
transformer below 20 A.
 Scenario 1: One capacitor in Line 1.
 Scenario 2: Two capacitors in Lines 1 and 2.
 Scenario 3: Three capacitors in Lines 1, 2 and 4.
 Scenario 4: Four capacitors in Lines 1, 2, 4 and 6.
 Scenario 5: Four capacitors in Lines 1, 2, 4 and 8.
The computed GICs in the transformers’ windings are given in Table 6. These results are
presented in Figure 13. They show that for Scenarios 3, 4 and 5, the maximum induced current is
16.1 A per phase and the GIC exceeds 15 A per phase in 3 transformers. If this value is acceptable,
Scenario 3 is retained because it includes only 3 blocking capacitors instead of 4 capacitors in
scenarios 4 and 5.

Page 23-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 6: GIC in transformer windings before and after mitigation measures.

GIC in transformer windings (A/Phase)

Transformer
Windings No With Mitigation
Mitigation Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5
T1 24.2 0 0 0 0 0
T2 24.2 29.6 0 0 0 0
T3 HV & T4
24.2 15 0 0 0 0
HV
T3 LV & T4 LV 22.4 21.7 20.6 10.5 10.6 10.5
T5 HV & T6
10.8 10.6 10.3 10.6 10.6 10.6
HV
T5 LV & T6 LV 14.6 14.4 14.4 14.4 14.2 14.4
T7 HV 23.8 23.3 22.3 0 0 0
T7 LV 16.3 16.1 16.1 16.1 16.1 0
T8 & T9 15.8 15.7 15.6 15.6 15.6 16.1
T10 13.8 13.7 13.7 13.7 13.7 13.7

Figure 13: GIC in transformer windings after mitigation measures.

Note that switchable resistors can also be added in neutral connections of the Y high-power
transformers. It could be more efficient to add blocking capacitors and switchable resistors to
critical transformers and bypassing others as proposed by the authors in [10]. After critical
transformers are identified, the mitigation scenario to retain should be based on the maximum
GIC that can flow in a transformer windings and the maximum number of blocking capacitors to
use.

7 Conclusion
This paper shows how to compute GIC in power systems using the MultiFields software package.
The measured one minute geomagnetic fields above a 7-layer earth resistivity model are used as
input of the computation process. A fictitious power system network that includes 8 transmission
lines and 10 transformers is studied. The suggested mitigation scenarios show that using 3
blocking capacitors drop the maximum GIC in the transformers windings from 24.2 A per phase
to 16.1 A per phase.

Copyright © 2017 SES ltd. All rights reserved. Page 23-13

PART III: SES CONTRIBUTIONS

8 References
[1] North American Electric Reliability Corporation, "March 13, 1989 Gemagnetic Disturbance," [Online].
Available: http://www.nerc.com/pa/stand/geomagnetic disturbance resources dl/nerc_1989-
quebec-disturbance_report.pdf. [Accessed 09 April 2017].
[2] R. S. Girgis and K. B. Vedante, "Impact of GICs on Power Transformers: Overheating is not the real
issue.," IEEE Electrification Magazine, vol. 3, no. 4, pp. 8-12, 2015.
[3] NOAA / Space Weather Prediction Center,
"www.swpc.noaa.gov/sites/default/files/images/u2/TheK-index.pdf".
[4] L. Trichtchenko and D. H. Boteler,, "Modeling geomagnetically induced currents using geomagnetic
indices and data," Plasma Science, IEEE Transactions on, vol. 32, pp. 1459-1467, 2004.
[5] L. Bolduc, P. Langlois, D. Boteler and R. Pirjola, "A study of geoelectromagnetic disturbances in
Quebec. I. General results," Power Delivery, IEEE Transactions on, vol. 13, pp. 1251-1256, 1998.
[6] Space Weather Canada, [Online]. Available: http://www.spaceweather.gc.ca.
[7] United State Geological Survey, [Online]. Available: https://geomag.usgs.gov.
[8] A. Pulkkinen, R. Pirjola and A. Viljanen, "Deternination of ground conductivity and system
parameters for optimal modeling of geomagnetically induced current flow in technological systems.,"
Earth Planets Space, vol. 59, pp. 999-1006, 2007.
[9] P. A. Fernberg, Earth Resistivity Structures and their Effects on Geomagnetic Induction in Pipelines,
Ottawa, Ontario: PhD in Earth Sciences, 2011.
[10] T. R. Hutchins and T. J. Overbye, "The Effect of Geomagnetic Disturbances on the Electric Grid and
Appropriate Mitigation Strategies.," in Power and Energy Conference at Illinois (PECI), Illinois,
2012.
[11] North American Electric Reliability Corporation, "Geo-Magnetic Disturbance (GMD): A Literature
Review and Summary of the April 2011 NERC GMD Workshop.," October 2011. [Online]. Available:
http://www.nerc.com/files/GMD_Draft_Proceedingst_Nov_10_2011_v3.pdf. [Accessed 10 04
2017].
[12] E. E. Bernabeu, "Modeling Geomagnetically Induced Currents in Dominion Virginia Power Using
Extreme 100-Yeur Geoelectric Field Scenarios - Part 1," IEEE Trans. On Power Delivery, vol. 28, no.
1, pp. 516-523, 2013.

Page 23-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

24 VERIFICATION AND COMPUTATION OF CONDUCTOR

ELECTRICAL PROPERTIES IN THE CONTEXT OF
INCOMPLETE DATASHEETS
Sofiane Chabane, Jie Liu, Nariman Tavana, Peter Zhao, Simon Fortin and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
This paper presents some ongoing R&D work at SES dealing with conductor electrical data computation. IEEE and
IEC standards do not make it mandatory to provide the data related to magnetic effects (inductance) in bare
conductors. This leads to a lack of data (e.g., equivalent permeability, GMR, reactance at one meter/foot spacing)
required to compute the per-unit-length internal parameters of conductors. To tackle this issue, intensive R&D
work is being carried out at SES in order to complete the datasheets provided by manufacturers. This is also useful
to validate the provided data or when data are required for other frequency or temperature conditions.

1 Introduction
Most of CDEGS computation programs, such as AutoGround, MultiGrounnd, MultiGroundZ,
MultiFields, MultiLines, AutoGridPro, AutoGroundDesign, and Right-Of-Way, obtain their
conductor data from SESLibrary [1]- [2], which stores the geometrical and electrical
characteristics of various types of conductors. At power frequencies and for conductors of non-
negligible lengths, both the internal resistance and reactance of conductors are necessary to get
trustworthy results. Most manufacturers provide only the internal DC resistance of conductors
because this is the only quantity that is required in the relevant standards [3]. Besides, the data
are always given at a reference temperature (e.g., 20 °C). As a consequence, it is often necessary
to fill in the missing data, or to compute them under other conditions.
This paper takes as an example the Chinese conductors that are stored in the library, since most
of them miss the information related to the imaginary part of the internal impedance, i.e. the
relative permeability. Some of them, such as conductors in the “Zinc Plated Stranded Steel” class,
even miss the information related to the DC resistance, as shown in Table 1. As the datasheets for
these conductors give only their geometrical characteristics, default values were specified for their
relative resistivity and permeability as a first solution. For the aforementioned class, 10 and 300
were specified, respectively, corresponding to typical electrical characteristics for steel (see Figure
1). However, these values correspond rather to the electrical characteristics of the material for a
solid steel conductor and not to the effective parameters of a stranded galvanized steel conductor.
Is this sufficient to represent a stranded and galvanized steel conductor? Since the galvanized
conductors are also stranded, it can become a demanding task if the investigation is done
manually through analytical calculations. In recent years, a new SES application called
SESImpedance [4] has been developed. This powerful software allows for the computation of the
internal impedance of conductors that can have arbitrary geometry and composition. This
software is used throughout this paper.
The aim of this paper is to describe a method allowing the calculation of more acceptable
resistivity and permeability values that correspond to the actual conductor construction.
However, since no information on the actual galvanized steel characteristics is available in the
pertinent datasheets, we first try to figure out the characteristics of a similar galvanized steel
conductor for which all the necessary data are available (real and imaginary parts of the

Copyright © 2017 SES ltd. All rights reserved. Page 24-1

PART III: SES CONTRIBUTIONS

impedance). Once the actual characteristics of the material and the modeling methodology are
assessed, the conductors for which data is lacking can be modeled easily in the same way.
In the following section, we first show how to model a single strand of a galvanized conductor.
Then, a stranded conductor is modeled and its datasheet data are compared to the results given
by the model. Finally, this method is applied to a class of conductors for which there is a lack of
data.
Table 1: Example of a data sheet that does not contain any electrical data.

Type Nominal diameter Nominal tensile strength (MPa)

Overall steel
Stranded Stranded Reference
cross section 1270 1370 1470 1570
conductor wire weight
area
(Kg/100m)
(mm) Minimum breaking force of the strand
mm
(kN)

16 5.4 7*1.8 17.81 14.83 20.8 22.4 24 25.7

25 6.6 7*2.2 26.61 22.15 31 33.5 35.9 38.4

35 7.8 7*2.6 37.16 30.93 43.4 46.8 50.2 53.6

50 9 7*3 49.48 41.19 57.8 62.3 66.9 71.4

70 11 2.2 72.22 60.12 82.5 89 95.5 102

100 13 2.6 100.88 83.98 115 124 133 142

120 14 2.8 116.99 97.39 133 144 154 165

135 15 3 134.3 118.8 153 165 177 189

150 16 3.2 152.81 127.21 174 188 202 215

Figure 1: Example of conductors extracted from datasheets that do not contain all the necessary data.

2 Modeling of Galvanized Conductors

In SESLibrary, a class of Galvanized Steel Conductors (SC-GZ) whose datasheet is sufficiently
detailed and contains all the necessary data to create and model a galvanized conductor already
exists. It contains the information related to resistance (DC and AC) as well as to inductance (GMR
and reactance at 1 foot spacing) as shown in Figure 2.

Page 24-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Generally, in the case of galvanized steel conductors, only the steel core is considered whereas the
effect of the zinc cladding is ignored [5]. This point will be verified in the following.

Figure 2: SC-GZ datasheet containing all the necessary data.

2.1 Analysis of the SC-GZ 9 conductor

Table 2 shows the equivalent electrical parameters (relative resistivity and permeability) of a
galvanized steel conductor called SC-GZ 9 computed based on different input data that are
available in the library: 𝑅𝑑𝑐 and GMR or 𝑅𝑎𝑐 and 11ft. The data in Table 2 show that the
computation results are slightly different according to the input data used. A first explanation for
this is that, in the datasheets, the 𝑅𝑎𝑐 resistance is given at 75 °C whereas the 𝑅𝑑𝑐 resistance is
given at 20 °C. This obviously can lead to different relative resistivity values. In the case of the
relative permeability, in addition to the effect of the resistance, the difference is also related to the
rounding of GMR values.
For these reasons, in this study we adopt the following values:
 Equivalent relative resistivity: 13.3
 Equivalent relative permeability: 32.7
In the following sections, different configurations of this cladded conductor will be studied using
SESImpedance [4] in order to determine an accurate geometrical model and materials for this
conductor. It is to be emphasized that the current version of SESImpedance does not consider the
effect of stranding leading to an increase in the length of the conductor. As a consequence, the
per-unit-length impedance computed by SESImpedance is always smaller than the reference
values that can be found when the effect of stranding is considered.

Copyright © 2017 SES ltd. All rights reserved. Page 24-3

PART III: SES CONTRIBUTIONS

Table 2: Electrical parameters of SC-GZ 9.

Data extracted from the datasheet

Real part Imaginary part

Impedance computation method

𝑅𝑑𝑐 [Ω/km] 𝑅𝑎𝑐 [Ω/km] 𝐺𝑀𝑅 [m] 1a1ft [Ω/km]

1.1 1.5 3.4e-6 0.8598

Effective 𝜇𝑟 32.6968 32.0482

Computed data

Internal reactance 𝑋𝐿𝑖 [Ω/km] 0.586749 0.586735

Effective 𝜌𝑟 13.3136 17.1644

Internal resistance 𝑅𝑖 [Ω/km] 1.20629 1.5

2.2 Calculation of the Zinc Coating Thickness

A galvanized steel conductor consists of a steel core cladded with zinc, mainly in order to increase
its resistance to corrosion. The Chinese datasheets and standards [6] provide the minimum
average zinc coating mass per square meter in 𝑔/𝑚2 as a function of the diameter of a wire strand,
as shown in Table 3. Generally, and as it is difficult to measure the coating thickness, galvanizers
and standards give rather the coating grade which is the weight of zinc on the surface of a steel
sheet [7]. Using this information, one can calculate the zinc thickness as shown in the following
paragraph. Furthermore, the amount of zinc coating depends on several parameters such as the
intended utilization specification requirements and the service life desired. As this article does
not address galvanization, we will use some simple relations to compute the thickness of the zinc.
This information will then be used in the computation of the per-unit-length parameters of the
conductor.
In order to determine the minimum average coating thickness, we proceed as follows:
 Consider the density of the Zinc: 𝑑𝑧𝑖𝑛𝑐 = 7.14 × 106 𝑔/𝑚3
 Consider the equivalent coating mass per square meter 𝑀𝑧𝑖𝑛𝑐 [𝑔/𝑚2 ]: 𝑀𝑧𝑖𝑛𝑐 = 𝑑𝑧𝑖𝑛𝑐 × 𝑡𝑧𝑖𝑛𝑐
where 𝑡𝑧𝑖𝑛𝑐 [𝑚] is the zinc thickness.
Therefore,
𝑀𝑧𝑖𝑛𝑐
𝑡𝑧𝑖𝑛𝑐 = 𝑑_𝑧𝑖𝑛𝑐 . (1)

Let us compute the coating thickness corresponding to 1 𝑔/𝑚2 coating mass:

1
𝑡𝑧𝑖𝑛𝑐 = = 0.14 𝜇𝑚
7.14 × 106
𝑔
This means that 1 𝑚2 (coating mass) corresponds to 0.14 𝜇𝑚 (coating thickness). This value is
exactly the same as that suggested by galvanizers [7] and allows for converting coating thickness
in microns to equivalent coating mass in grams per square meter.

Page 24-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

As can be noticed from Table 3, the minimum average zinc coating mass depends on the conductor
class as well as on the wire diameter. In order to simplify this study, it has been decided to assess
the following three configurations:
 Minimum zinc coating: this corresponds to 110 𝑔/𝑚2
110+290
 Average zinc coating: this corresponds to 2
= 200 𝑔/𝑚2

 Maximum zinc coating: this corresponds to 290 𝑔/𝑚2

Table 3: Minimum average zinc coating mass as a function of the diameter of strand wire.

Weight of zinc layer (𝒈/𝒎𝟐 )

Wire nominal Winding test mandrel
diameter not less than diameter
several times the
(mm) diameter of steel wire
Special A A B

1.00

1.10 180
110
1.20
160
1.30

1.40 12
200
1.50
130
1.60

1.70 180
220
1.80
160
2.00
230 200
2.20 180

2.40 240
220 200
2.60 250

2.80 270
250 14
3.00 270 230

3.20 280 260

3.50

3.80 290 270 250

4.00

Since no details are given about the radius (1.625 mm) of the SC-GZ 9 conductor (it is not specified
if it is the steel core radius only or the total radius including the steel and zinc coating), we assume
that it corresponds to the steel core radius.

Copyright © 2017 SES ltd. All rights reserved. Page 24-5

PART III: SES CONTRIBUTIONS

Hence, in terms of the coating thickness and using Equation (1), we get:
 110 𝑔/𝑚2 corresponds to a zinc thickness of 0.016 mm. Therefore, the total radius is R =
1.641 mm.
 200 𝑔/𝑚2 corresponds to a zinc thickness of 0.028 mm. Therefore, the total radius is R =
1.653 mm.
 290 𝑔/𝑚2 corresponds to a zinc thickness of 0.041 mm. Therefore, the total radius is R =
1.666 mm.

2.3 Electrical Characteristics of the Used Materials

Since no additional information is provided in the datasheets about the used materials, two
different steel materials are investigated along with the zinc cladding:
1. Steel with 𝜇𝑟 = 75 and 𝜌𝑟 = 8.3
2. Steel with 𝜇𝑟 = 250 and 𝜌𝑟 = 10
The zinc cladding has the following characteristics:
𝜇𝑟 = 0.9999842 and 𝜌𝑟 = 3.572 (𝜌 = 6.158 × 10−8 Ω. m)

2.4 Geometrical Characteristics and Models to Represent a Galvanized

Conductor
In order to develop a method allowing to complete for missing data in datasheets, different
geometries are investigated: zinc cladded steel (with different zinc thicknesses), core steel only,
entire strand (core and cladding) made of steel.
 Zinc cladded steel strands
This is the original conductor and is modeled just as it is. This constitutes the reference data and
considered as the closest model to a zinc cladded steel conductor. The only unknown in this case
is the electrical characteristics of the actual used material. Figure 3(a) represents an overview of
a zinc cladded steel conductor with a steel core radius 𝑟𝑠 and an outer radius (over the cladding)𝑟𝑧 .
 Core steel strands
This conductor is made only of steel and has a radius that corresponds to the core steel radius 𝑟𝑠 .
Figure 3(b) shows an example of a core steel conductor with a steel core radius 𝑟𝑠 .
 Entire steel strands
This conductor is made only of steel and has a radius that corresponds to the zinc cladding radius
𝑟𝑧 . Figure 3(c) depicts an example of a steel conductor with a radius 𝑟𝑧 .

𝑟𝑧 𝑟𝑧

𝑟𝑠 𝑟𝑠

(a) A zinc clad steel strand (b) A core steel strand (c) An entire steel strand

Figure 3: Geometrical characteristics.

Page 24-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.5 Modeling a Single Strand Galvanized Conductor

Before modeling the stranded SC-GZ 9 conductor, and in order to assess the effect of the cladding
on its electrical parameters, a single strand (solid conductor) of this conductor is first studied. All
the above described cases are studied: different cladding thicknesses, different steel
characteristics and different configurations. The results are discussed hereafter.
Table 4 shows that increasing the thickness of the zinc cladding reduces the internal resistance
and inductance for both types of steel. This result is expected and means that the current has a
tendency to pass through the less resistive layer. Nevertheless, since the thickness of the cladding
is still very small, we notice the effect of the steel core through increasing the values of the relative
resistivity and permeability.Table 5 shows that the strand cannot be modeled by steel only even if
the zinc thickness is still very small. It shows also that neglecting the zinc cladding leads to results
that are within 5% and 15% of the reference values (in presence of the zinc cladding) for the
resistivity and permeability respectively.
Table 4: Conductor characteristics as a function of the cladding thickness.

Material
Steel 𝜇𝑟 = 75, 𝜌𝑟 = 8.3 Steel 𝜇𝑟 = 250, 𝜌𝑟 = 10
characteristics

Cladding thickness Cladding thickness

Geometrical
characteristics
16 μm 28 μm 41 μm 16 μm 28 μm 41 μm

Effective 𝜇𝑟 67.705 63.569 59.708 221.04 204.54 189.26

Effective 𝜌𝑟 8.181 8.043 7.887 9.802 9.594 9.361

Internal
resistance 𝑅𝑖 [Ω/ 16.706 16.184 15.622 20.262 19.523 18.737
km]

Internal reactance
1.275 1.197 1.125 4.137 3.830 3.546
𝑋𝐿𝑖 [Ω/km]

Table 5: Conductor characteristics as a function of the configuration.

Material
Steel 𝜇𝑟 = 75, 𝜌𝑟 = 8.3 Steel 𝜇𝑟 = 250, 𝜌𝑟 = 10
characteristics

Zinc clad steel. Zinc clad steel.

Geometrical cladding Entire steel Core steel cladding Entire steel Core steel
characteristics thickness strand strand thickness strand strand
16 μm 16 μm

Effective 𝜇𝑟 67.705 74.884 74.887 221.047 249.444 249.607

Effective 𝜌𝑟 8.181 8.396 8.388 9.802 10.107 10.116

Internal
resistance 16.706 17.149 17.471 20.262 20.950 21.370
𝑅𝑖 [Ω/km]
Internal
reactance 1.275 1.410 1.410 4.137 4.662 4.666
𝑋𝐿𝑖 [Ω/km]

Note that the previous results shown in Table 4 and Table 5 were obtained at 60 Hz but the
conclusions remain valid for 50 Hz.
Copyright © 2017 SES ltd. All rights reserved. Page 24-7
PART III: SES CONTRIBUTIONS

2.6 Modeling a Stranded Galvanized Conductor

The same analysis as in the case of a single strand is made for the stranded conductor SC-GZ 9:
different cladding thicknesses, different steel characteristics and different configurations are
studied. The objective is to assess the effect of all these parameters on the electrical properties of
the stranded conductor.
Table 6 shows that increasing the thickness of the zinc cladding reduces the internal resistance
and inductance for both types of steel. This result is expected (it has the same tendency as in the
case of a single strand) and means that the analysis of a single strand can be generalized to a
stranded conductor. Nevertheless, since the thickness of the cladding is still very small, we notice
the effect of the steel core through increasing the values of the relative resistivity and permeability.
Table 6: Stranded conductor characteristics as a function of the cladding thickness.

Material Data from

Steel 𝜇𝑟 = 75, 𝜌𝑟 = 8.3 Steel 𝜇𝑟 = 250, 𝜌𝑟 = 10
characteristics SESLibrary

Rdc and GMR

Cladding thickness Cladding thickness based
Geometrical computation
characteristics
16 μm 28 μm 41 μm 16 μm 28 μm 41 μm

Effective 𝜇𝑟 15.960 13.323 12.116 26.518 21.708 19.451 32.6968

Effective 𝜌𝑟 10.827 10.625 10.392 12.660 12.503 12.187 13.3136

Internal
resistance 0.909 0.874 0.839 1.10 1.053 1.004 1.20629
𝑅𝑖 [Ω/km]

Internal
reactance 0.295 0.248 0.226 0.481 0.398 0.358 0.586749
𝑋𝐿𝑖 [Ω/km]

Table 7 shows that, independently from the steel material being used, the stranded conductor
cannot be modeled by steel only even if the zinc thickness is still very small. Actually, even if the
relative resistivity results are similar in all the considered cases, the relative permeability depends
highly on the configuration (zinc cladded, core steel only or entire steel strand with diameters
depending on the cladding thickness). This confirms once again that modeling accurately a zinc
cladded conductor requires considering its actual configuration.
Table 7 also shows that the steel with 𝜇𝑟 = 250 , 𝜌𝑟 = 10 leads to results closer to those in
SESLibrary and in the datasheets of SC-GZ 9. Even if, and as expected, the entire steel strand
model leads to results different from those of the core steel strand model, the results for the entire
steel strand model are generally closer to the database and datasheets results than those of the
core steel strand model. The core steel strand radius is the same as that entered in the database
and found in datasheets whereas the optimal thickness added to the radius corresponds to the
minimum zinc thickness calculated as described above.
As the higher resistivity and permeability steel leads to results closer to those provided in the
datasheets, we consider in the following that the Chinese conductors are made of the same
material. Therefore, a steel material with 𝜇𝑟 = 250 , 𝜌𝑟 = 10 is considered. The conductors are
modeled by entire steel strands for which the radius is the sum of the minimum zinc thickness
and of the radius given in the datasheets.
Page 24-8 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

Table 7: Stranded conductor characteristics as a function of the configuration

Material Data from

Steel 𝜇𝑟 = 75, 𝜌𝑟 = 8.3 Steel 𝜇𝑟 = 250, 𝜌𝑟 = 10
characteristics SESLibrary

Entire steel strands comprising the core and an Entire steel strands comprising the core Rdc and
additional thickness accounting for the and an additional thickness accounting for GMR based
Geometrical cladding the cladding computation
characteristics

0 μm 16 μm 28 μm 41 μm 0 μm 16 μm 28 μm 41 μm

Effective 𝜇𝑟 16.71 21.17 22.68 26.28 28.60 36.81 40.30 39.03 32.6968

Effective 𝜌𝑟 11.23 11.14 11.09 11.15 13.17 12.86 12.73 12.69 13.3136

Internal resistance
0.95 0.956 0.952 0.964 1.16 1.179 1.18 1.16 1.20629
𝑅𝑖 [Ω/km]

Internal reactance
0.31 0.387 0.41 0.471 0.52 0.648 0.699 0.678 0.586749
𝑋𝐿𝑖 [Ω/km]

2.7 Discussion
In this section, we presented a method allowing to tackle the lack of data in some conductor
datasheets. The galvanized conductors are modeled in SESImpedance as being made entirely of
steel, with a radius corresponding to the core steel.

3 Modeling Chinese Conductors

3.1 Introduction
In the following, this method will be applied to an example of a class of conductors for which no
reference data could be found (a Chinese conductor class). To demonstrate the method, we use
the same material and procedure as in the case of the SC-GZ 9 conductor. This means that the
galvanized steel conductor is approximated by a conductor made entirely of steel but with a radius
equal to the steel core.

3.2 Results
In this section, the new results for the Chinese conductors are shown. These data are obtained
using SESImpedance and the method described above:
 Consider only the steel core as suggested by current standards.
 Steel characteristics are 𝜇𝑟 = 250, 𝜌𝑟 = 10
These results could be used in the near future to update the data for the Chinese conductors in the
conductor database of SESLibrary. However, more investigations are needed.

Copyright © 2017 SES ltd. All rights reserved. Page 24-9

PART III: SES CONTRIBUTIONS

Table 8: Data for Z_Zinc-Plated-STRNDSTEEL (GB1200-75)

Material
Steel 𝜇𝑟 = 250, 𝜌𝑟 = 10 Data from SESLibrary
characteristics

Internal Internal
Conductor Effective Effective Effective Effective
resistance reactance 𝑅𝑖 [Ω/km] 𝑋𝐿𝑖 [Ω/km]
name 𝜌𝑟 𝜇𝑟 𝜌𝑟 𝜇𝑟
𝑅𝑖 [Ω/km] 𝑋𝐿𝑖 [Ω/km]

GJ-25 13.034 53.479 6.588 1.004 10 300 6.636 4.783

GJ-35 12.980 54.581 4.742 1.021 10 300 5.44 4.29

GJ-50 12.934 52.696 3.585 0.980 10 300 4.63 3.81

GJ-70 13.343 32.810 2.451 0.612 10 300 3.70 3.15

GJ-95 13.316 28.992 1.909 0.539 10 300 3.20 2.78

GJ-120 13.219 31.888 1.543 0.585 10 300 2.82 2.49

3.3 Discussion
As is shown in the above tables, the new relative resistivity and permeability are within the range
of some other equivalent conductors (SC-GZ for instance) and depend upon the stranding, as
expected.

4 Conclusions
This study shows that modeling a stranded conductor made of galvanized strands should take into
account the cladding material as well as the stranding effects on the electrical parameters of the
conductor. Indeed, the equivalent relative resistivity and permeability of a stranded conductor
depend highly on the stranding and are different from the relative resistivity and permeability of
the material of the strands.
Using a stranded conductor whose datasheet provides all the necessary data allowed us to retrieve
approximate properties for the material making up the strands. This conductor class being similar
to a Chinese class that does not have all the required data, its material properties were then used
to compute the electrical characteristics of the Chinese conductors.
It has also been shown that, in the case of a stranded conductor, the “relative permeability” of the
material is different than the “equivalent relative permeability” of the conductor (which considers
that the conductor is solid) whereas in the case of a solid conductor, the “relative permeability” is
closer to the “equivalent relative permeability”.

Page 24-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

5 References
[1] S. Wei, E. Dawalibi, G. Noel, N. Botchorichvili, J. Liu, M. Siahrang, Y. Yang, T. Mukherjee and S.
Chabane, "Introducing the New SES Object Library," in CDEGS Users' Conference Proceedings,
Boulder, Colorado, 2016.
[2] F. Gougeon, S. Wei, N. Botchorichvili, H. Yin, S. Chabane, P. Zhao, J. Liu and N. Tavana, "New Features
in SESLibrary," in User's Group Meeting, Portland, Oregon, 2017.
[3] "Round wire concentric lay overhead electrical stranded conductors," IEC 61089-1991, 20 June, 1991.
[4] M. Golshayan, S. Baron, S. Touimer, M. Deslongchamps and S. Fortin, "Improvements and New
Features in SESImpedance," in User's Group Meeting, San Diego, California, 2015.
[5] W. A. Lewis and P. D. Tuttle, "The Resistance and Reactance of Aluminum Conductors, Steel
Reinforced," AIEE Transactions, vol. 77, no. 3, pp. 1189-1215, 1959.
[6] "Zinc-coated steel wire strands," YB/T 5004-2001, 2001.
[7] "ASTM A653 / A653M-09, Standard Specification for Steel Sheet, Zinc-Coated (Galvanized) or Zinc-
Iron Alloy-Coated (Galvannealed) by the Hot-Dip Process," ASTM International, West Conshohocken,
PA, 2009.

Copyright © 2017 SES ltd. All rights reserved. Page 24-11

 UGM 2017 – CANNON BEACH, OREGON

25 FINDING EQUIVALENT CONDUCTORS REPRESENTING

A TYPICAL METALLIC WIRE-MESH FENCE

Jie Liu, Farid P. Dawalibi, Peter Zhao, and Sofiane Chabane

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
This paper presents an investigation to find equivalent conductors of a metallic wire-mesh fence. The study is
carried out with the HIFREQ and TRALIN computation modules, the SESeBundle tool, and fundamental theoretical
derivations. A typical and realistic substation wire-fence with a dense wire-mesh has been modeled and various
equivalent conductors representing the fence mesh have been determined. Furthermore, the equations used to
calculate the adjustments of the conductor radii and the distance between conductors have been derived to offer
various alternatives to the equivalent set of desirable conductors. Several examples have been studied to illustrate
the steps of the procedure to find equivalent models and the effects of adjustments. This paper mainly proposes
modeling simplifications in order to reduce the run computation time involving modelling fences to a minimum,
while maintaining reasonable accuracy.

1 Introduction
Fences around substations, power plants, and test and valve stations along pipelines are usually
metallic as shown in Figure 1. Metallic Fence grounding in electrical facilities such as substations
is necessary since the fence is usually accessible to the public. The grounding design should insure
that touch voltages on the fence are within the tolerable limit. In grounding studies using
MultiGround and MultiGroundZ, the metallic wire-mesh fence connected to the grid should be
modeled as realistically as possible. Normally, since it is not possible to model aboveground
conductors in MultiGround and MultiGroundZ, only the buried metallic conductors forming the
fence posts and fence conductor loop(s) around the main grid are modeled. If the fence does not
have any loop connected to it then the buried portion of the posts are modeled in MultiGround
and MultiGroundZ only and an insulated conductor loop is added in MultiGroundZ to connect
the fence posts together. Note that the insulated connection wire is not necessary in MultiGround
because all the conductors forming an electrode are assumed to be lossless and equipotential.
The additional insulated conductor representing the aboveground portion of the fence in
MultiGroundZ should preferably be specified with a series impedance as computed hereafter. In
MultiFields, aboveground conductors can be modeled without any restrictions. However, the
complexity and density of wire-mesh fence segments that are generated when a realistic model of
the fence is used may become significant. This may increase computer memory and run time
beyond practical limits. Consequently equivalent models requiring less computer resources are
preferred.

Copyright © 2017 SES ltd. All rights reserved. Page 25-1

PART III: SES CONTRIBUTIONS

Figure 1: Wire-mesh fence with barbed wires around a substation.

Furthermore, in some situations, the substation wire-mesh fence has to be modeled realistically.
For example, in a lightning study, we may need to study the transient electromagnetic fields
response near or across the fence. In grounding studies, we may want to calculate and examine
touch voltages at specific locations on the fence.
To model a fence consisting of a lot of mesh conductor segments, we could in principle model all
of them in MultiFields. However, this might quickly exceed the computer memory capacity, and
if enough memory is available, the computer run time will dramatically increase as a function of
the number of wire-mesh conductor segments. In many cases, modeling all the conductors is an
overkill and is not necessary. It is possible to use only a few conductors, representing a skeleton
of the fence or even one single conductor to simplify the model.
The object of this study is to find equivalent conductors representing the fence, so that the
equivalent fence conductors yield reasonably accurate results with reduced computer memory
requirement and run time. As a compromise alternative, in the area where the computation
results are of interest, we could model the wire-mesh fence realistically if high accuracy is
required.
This paper describes how to obtain an accurate equivalent fence model that reduces computer
memory and run time. Starting from a realistic metallic wire-mesh fence, different scenarios have
been studied using various computation modules and tools, such as HIFREQ, TRALIN and
SESeBundle. Appropriate mathematical equations have been derived in order to properly and
effectively convert the wire-mesh fence to equivalent conductors.
Note that all the example files used to obtain the results presented in this paper are in the folder
UGC 2017\Example Files\Equivalent Conductors for a Fence Mesh.

2 Methodology
Typical fences are shown in Figure 2. The wire-mesh fence or diamond-mesh fence is usually
made of galvanized or steel wire, bare or coated. The wires can run horizontally, vertically, and
obliquely.

Page 25-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: Various patterns of a wire-mesh fence.

The idea is to find one or several equivalent conductors with proper geometrical and electrical
characteristics, such as conductor radius, distance between conductors, relative resistivity, and
relative permeability. The aim is to replace the very dense wire-mesh fence by equivalent
conductors in the model to have the same series impedance as the entire fence.
The study methodology was carried out in accordance with the following steps:
(1) Model a realistic length portion of the wire-mesh fence which could be as short as one
fence section up to a very long fence length compared to the height of the modeled
structure (one mile long or more) in HIFREQ. Then compute the actual series impedance
of the fence between the current injection points. The appropriate length depends on the
injection and collection points of the current path of interest. For example, if the nature
of the problem is induction on a long farm fence from parallel transmission lines, then the
appropriate model would be a long (i.e., one mile) fence.
As a typical illustration of the above possibilities, the results corresponding to one typical
short fence section and a one mile fence section are described and discussed in this article.
The results show no significant differences in the methodology used regardless of the
modeled length. However noticeable differences are observed depending on the fence
section length and current injection model.
(2) Model all horizontal conductors or a subset of all the conductors in SESeBundle to obtain
a single equivalent conductor or a set of equivalent identical conductors having an
equivalent radius, equivalent relative resistivity, and equivalent relative permeability,
which gives the same series impedance of the fence. The internal and external resistance
and reactance of the fence are also available from the computation results of TRALIN
which is invoked by SESeBundle.
(3) Model this single equivalent conductor or the set of equivalent identical conductors in
HIFREQ and compute the equivalent total series impedance of the fence mesh. The series
impedance of the realistic model and the equivalent model are then compared. In most
cases, the results confirm that the equivalent model is accurate enough to be used in the
HIFREQ model.
(4) If necessary, use the analytical equations to adjust the radius or the distance between the
equivalent conductors to improve the accuracy of the equivalent model. If an adjustment
of the conductor radius is required, the internal impedance computed by TRALIN could
be imposed on the conductors instead of continuing the iterative adjustment process.

Copyright © 2017 SES ltd. All rights reserved. Page 25-3

PART III: SES CONTRIBUTIONS

2.1 HIFREQ Model of a Realistic Wire-Mesh Fence

A realistic wire-mesh fence has been modeled in HIFREQ as shown in Figure 3. Its characteristics
are given in Table 1. Note that the length of the HIFREQ model varied between one typical fence
section installed between fence posts and a one mile fence section.

Figure 3: A realistic wire-mesh fence modeled in HIFREQ.

Table 1: Wire-mesh fence characteristics and other parameters used in the models.

Wire-Mesh Fence Characteristics Data

Total length of fence (feet) 5280 15
Fence height (feet) 4.938 4.938
Fence cell height (feet) 0.082 0.082
Wire radius (inch) 0.059 0.059
Wire relative resistivity (p. u.) 12 12
Wire relative permeability (p. u.) 100 100
Total number of horizontal wires 60 60
Total number of vertical wires 3, 19, 54 and 107 3, 19, 54 and 107
Soil resistivity (ohm-m) 100 100
Frequency (Hz) 60 60
A current of 1 A is injected at the left side of the fence mesh and 1 A current is collected at the right
side for the one fence section case using 60 virtual connections to simulate a quasi-equipotential
large radius fence post on either side of the wire-mesh fence section. The quasi-equipotential
nature of the fence posts is demonstrated in the next section. Consequently, they don’t need to be
part of the model if reasonably accurate values are sufficient. The series impedance of the modeled
fence is the difference of potentials at the two energization points divided by the unit current.
Note that in a typical substation fence scenario, the only path for the current to enter the fence
section is through the fence posts, except when the fault involves a broken phase conductor lying
on the fence.
Four different scenarios consisting of 3, 19, 54, and 107 vertical wires are modeled, respectively,
in order to determine the effects of the vertical bonding wires on the computed series impedance.
Table 2 shows the computation results.

Page 25-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 2: Series impedance of the wire-mesh fence.

Number of
Resistance Reactance Run Time
Vertical Wires
(ohms/mile) (ohms/mile) (hours)
Scenarios
One Mile One 15’ One Mile One 15’ One Mile One 15’ One Mile One 15’
Fence Fence Fence Fence Fence Fence Fence Fence
Section Section Section Section Section Section Section Section
Length Length Length Length Length Length Length Length

107 107 0.8538 0.8406 0.9579 0.3557 28.2 38.5

54 54 0.8543 0.8406 0.9582 0.3557 4.6 13.5

19 19 0.8547 0.8406 0.9584 0.3557 0.83 2.1

3 3 0.8549 0.8406 0.9586 0.3557 0.2 1.0

From the above computation results, it is clear that the computed series impedances in the case
of the same fence length are almost the same, regardless of the number of vertical wires in the
model. However, the model with 107 vertical wires (which cuts the horizontal mesh size to 50’
sections for the one mile section fence and 0.14’ for the 15' fence section) requires extremely long
computer run time. Therefore, finding equivalent models with a few conductors is important in
practical studies.

2.2 Obtaining an Equivalent Wire-Mesh Fence Using SESeBundle

The wire-mesh fence can be represented by one or several identical conductors. In SESeBundle,
all the horizontal conductors of the fence or a subset of horizontal conductors can be modeled as
a bundle of linear parallel sub-conductors to obtain the equivalent conductors. Note that in
SESeBundle, although the vertical wires are not specified or mentioned in the input data of a
model, all the conductors modeled are assumed to be regularly interconnected as a bundle.
SESeBundle will find the characteristics of an equivalent single conductor that accurately
represents a bundle of conductors, as far as their series impedance is concerned.
When a single conductor is required to represent the fence, 60 conductors have been specified in
SESeBundle as shown in Figure 4 and the single equivalent conductor is shown in Figure 5. When
two conductors are required to represent the fence, 30 conductors have been specified in
SESeBundle to obtain each one of the two equivalent conductors. The 60 and 30 conductors’
scenarios internal and external impedances are available from TRALIN computation results.

Copyright © 2017 SES ltd. All rights reserved. Page 25-5

PART III: SES CONTRIBUTIONS

Figure 4: A total of 60 conductors representing a fence as specified in SESeBundle.

Figure 5: The single equivalent conductor as computed in SESeBundle.

Page 25-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The results of the computed equivalent conductor characteristics are shown in Table 3.
Table 3: Equivalent conductor characteristics of a fence.

One Single Two

Conductor Characteristics Equivalent Equivalent
Conductor Conductors

Equivalent radius (inch) 12.9662 6.371293

Equivalent relative resistivity (p.u.) 9682.358 4659.885

Equivalent relative permeability (p.u.) 1.66669 3.33329

Internal resistance (ohms/mile) 0.787617 1.569933

Internal reactance (ohms/mile) 0.050525 0.101047

2.3 HIFREQ Model of the Equivalent Fence Mesh

Based on the computation results from SESeBundle, the fence mesh can be modeled by a single
conductor, or a set of two or more conductors in HIFREQ. Figure 6 shows the case of two
equivalent horizontal conductors representing the wire-mesh fence.

Figure 6: Two equivalent horizontal conductors of a wire-mesh fence modeled in HIFREQ.

Note that the two equivalent conductors are placed in the middle of each group containing 30
conductors. In other words, the first equivalent conductor is in the middle of the top 30
conductors and the second is in the middle of the bottom 30 conductors.
In the case of a single equivalent conductor representing the entire fence mesh, the single
equivalent conductor is placed in the middle of the 60 conductors.
The computation results corresponding to the series impedances of a realistic one mile fence, a
single equivalent conductor and two equivalent conductors representing the fence are shown in
Table 4.

Copyright © 2017 SES ltd. All rights reserved. Page 25-7

PART III: SES CONTRIBUTIONS

Table 4: Series impedances of a realistic one mile fence and equivalent fence meshes.

Resistance Reactance Relative Error

Scenarios
(ohms/mile) (ohms/mile) (%)

Realistic one mile fence with 3

0.854942 0.958575
vertical wires

A single equivalent conductor

0.846044 0.959380 0.08
representing the fence

Two equivalent conductors

0.847268 0.952254 0.66
representing the fence

As we can see from the above computation results, the results of the equivalent models are
extremely accurate and more than satisfactory. The largest relative error is 0.66%, which is
considered extremely well for all applications.
However despite these small differences in the studied cases, it is possible that in some scenarios
the differences may not be negligible. Therefore, further adjustments might be needed. In the case
of two equivalent conductors, we can adjust the distance between the two conductors and move
the two conductors to increase or decrease the overall reactance and improve the results.
Furthermore, for the two equivalent conductors, if one is located at the top of the fence and the
other one at the bottom of the fence as a more realistic rendering of the fence, an adjustment to
the conductor radius could be carried out to achieve the same series impedance.

2.4 Theoretical Adjustments of Conductors’ Radius and Spacing

The series impedance of a horizontal conductor is the sum of its internal and external impedances.
As mentioned above, the internal impedance of conductors can be maintained by specifying it
directly in HIFREQ. The external impedance can be split into two parts, one part is the external
reactance (inductance x frequency) relative to the image of the conductor assuming a perfect
conducting earth and the other part is the correction term due to the presence of a non-perfect
earth. The internal reactance for the conductors is always the same. Therefore, the discrepancy
between SESeBundle and HIFREQ results is due to the difference of the computed external
inductance. By adjusting the equivalent conductor radius and the spacing between equivalent
conductors using an iterative method in HIFREQ, the same series impedance can be achieved.
However, the use of theoretical equations will provide a more efficient way to obtain the results
faster. In the following sections, the equations required to adjust the conductor radius or the
spacing between conductors are provided.
2.4.1 Conductor Radius Adjustments to Obtain the Desired External
Reactance
The inductance of any number of conductors interconnected in parallel (bundle), and their images
below perfect earth as illustrated in Figure 7 can be computed easily as shown below.

Page 25-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

d2
d1
d3

Earth’s Surface

H’
d’3
d’1
d’2

Figure 7: Conductors and theirs images.

Assuming that the sub-conductors inside a bundle are identical and that the current is equally
distributed among them, the external inductance of the bundle can be obtained from the following
well known analytical expression:
𝜇0 𝐺𝑀𝐷𝑏 𝜇𝑟
𝐿𝑒𝑥𝑡 = [𝑙𝑛 ( )− ] (1)
2𝜋 𝐺𝑀𝑅𝑏 4𝑛

where:
𝐺𝑀𝐷𝑏 is the geometric mean distance between the ‘go’ bundle and the ‘return’ bundle.
𝐺𝑀𝑅𝑏 is the geometric mean radius of the bundle.
n is the number of sub-conductors in the bundle.
𝜇0 is the permeability of free space or vacuum, equal to 4π⨉10-7 = 1.25664E-06 henries per
meter.
𝜇r is the relative permeability of sub-conductors.
𝐻 is the height of the bundle center.
𝐻′ is the height of the bundle image center and is equal to 𝐻.
𝑑𝑖 is the distance between the sub-conductor center and the height of the bundle center.
𝑑′𝑖 is the distance between the sub-conductor image center and the height of the bundle
image center.

The bundle 𝐺𝑀𝑅𝑏 , assuming that all sub-conductors in the bundle are the same:
𝐺𝑀𝑅𝑏 = 𝑛2 (𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 )𝑛 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗
√ 𝑗=1,𝑛 (2)
𝑖≠𝑗

where 𝐷𝑖𝑗 is the distance between the sub-conductor i and sub-conductor j . 𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 is
the geometric mean radius of a single sub-conductor:
−𝜇𝑟
𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 = 𝑟𝑒 4 (3)

where 𝑟 is the radius of the sub-conductor.

The bundle 𝐺𝑀𝐷𝑏 can be written as:

2
𝑛
⃗ + ⃗⃗⃗
𝐺𝑀𝐷𝑏 = √∏ 𝑖=1,𝑛 |(𝐻 ⃗ + ⃗⃗⃗
𝑑𝑖 ) − (𝐻 𝑑𝑗′ ) | (4)
𝑗=1,𝑛

Copyright © 2017 SES ltd. All rights reserved. Page 25-9

PART III: SES CONTRIBUTIONS

𝑛2
𝐺𝑀𝐷𝑏 = √∏ 𝑖=1,𝑛 √(𝑥𝑖 − 𝑥′𝑗 )2 + (2𝐻 + 𝑦𝑖 − 𝑦′𝑗 )2 (5)
𝑗=1,𝑛

where (𝑥𝑖 , 𝑦𝑖 ) are the local coordinates of the sub-conductors with respect to the bundle center,
(𝑥′𝑗 , 𝑦′𝑗 ) are the local coordinates of the sub-conductor images with respect to the bundle image
center.

We assume that we have two sets of equivalent fence conductors. The first set is the sub-
conductors with radius of 𝑟′ that was determined initially and gave the first approximate external
inductance. The radius of the sub-conductors in the second set needs to be adjusted in order to
produce a more accurate external inductance. The desired external inductance is the one
computed from the actual fence model in HIFREQ.

When the two sets of equivalent conductors are all the same except for the radius of the sub-
conductors, the difference between the computed external inductances is:
𝜇 𝐺𝑀𝐷 𝜇 𝜇 𝐺𝑀𝐷 ′ 𝜇
∆𝐿𝑒𝑥𝑡 = 2𝜋0 [𝑙𝑛 (𝐺𝑀𝑅𝑏 ) − 4𝑛𝑟 ] − 2𝜋0 [𝑙𝑛 (𝐺𝑀𝑅𝑏′ ) − 4𝑛𝑟 ] (6)
𝑏 𝑏

𝜇 𝐺𝑀𝐷 𝐺𝑀𝐷 ′
∆𝐿𝑒𝑥𝑡 = 2𝜋0 [𝑙𝑛 (𝐺𝑀𝑅𝑏 ) − 𝑙𝑛 (𝐺𝑀𝑅𝑏′ )]
𝑏 𝑏

where 𝐺𝑀𝑅𝑏 ′ and 𝐺𝑀𝐷𝑏 ′ are for the sub-conductors with a radius 𝑟′. 𝐺𝑀𝑅𝑏 and 𝐺𝑀𝐷𝑏 are for
the sub-conductors with a radius 𝑟.

Note that in this method, the locations of the sub-conductors or distances between sub-
conductors are the same, i.e., 𝐺𝑀𝐷𝑏 =𝐺𝑀𝐷𝑏 ′. Therefore,
𝜇0 𝐺𝑀𝑅𝑏 ′
∆𝐿𝑒𝑥𝑡 = [𝑙𝑛 ( )]
2𝜋 𝐺𝑀𝑅𝑏
𝜇 𝐺𝑀𝑅 ′
∆𝜔𝐿𝑒𝑥𝑡 = 2𝜋𝑓 2𝜋0 [𝑙𝑛 ( 𝐺𝑀𝑅𝑏 )]
𝑏
∆𝜔𝐿𝑒𝑥𝑡
𝐺𝑀𝑅𝑏 ′
=𝑒 𝑓𝜇0 (7)
𝐺𝑀𝑅𝑏

From Equation (2):

𝐺𝑀𝑅𝑏 ′ = 𝑛2 (𝐺𝑀𝑅 ′ 𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 )𝑛 ∏ i=1,n 𝐷𝑖𝑗 ′
√ 𝑗=1,𝑛
𝑖≠𝑗

𝐺𝑀𝑅𝑏 = 𝑛2(𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 )𝑛 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗

√ 𝑗=1,𝑛
𝑖≠𝑗

Again, 𝐷𝑖𝑗 =𝐷𝑖𝑗 ′, then Equation (7) can be simplified as:

∆𝜔𝐿𝑒𝑥𝑡
𝑛 𝐺𝑀𝑅′
√ 𝐺𝑀𝑅 𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 = 𝑒 𝑓𝜇0 (8)
𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟

From Equation (3):

−𝜇𝑟
𝐺𝑀𝑅′𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 = 𝑟 ′ 𝑒 4

−𝜇𝑟
𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 = 𝑟𝑒 4

Page 25-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Thus Equation (8) can be further simplified as:

∆𝜔𝐿𝑒𝑥𝑡
𝑛 𝑟′
√ =𝑒 𝑓𝜇0
𝑟
∆𝜔𝐿𝑒𝑥𝑡
𝑟 = 𝑟′𝑒
−𝑛
𝑓𝜇0 (9)
Note that if the relative resistivity and relative permeability of the sub-conductors are used to
calculate the internal impedance of the equivalent conductors in HIFREQ, the change of the
radius of the equivalent conductors will also change their internal impedance. Therefore, the pre-
computed internal series impedance of the sub-conductor in ohms/km or ohms/mile (depending
on the specified units) should be used as a User-Defined conductor type in HIFREQ to impose the
internal impedance value.

2.4.2 Conductors’ Distance Adjustments to Obtain the Desired External

Reactance

Similarly, instead of changing the conductor radius, the distance between the conductors can be
adjusted to obtain more accurate results while keeping the radius of all sub-conductors the same:
𝜇0 𝐺𝑀𝐷 𝐺𝑀𝐷𝑏 ′
∆𝐿𝑒𝑥𝑡 = [𝑙𝑛 (𝐺𝑀𝑅𝑏 ) − 𝑙𝑛 ( )] (10)
2𝜋 𝑏 𝐺𝑀𝑅𝑏 ′
𝜇 𝐺𝑀𝐷 ∗𝐺𝑀𝑅 ′
∆𝜔𝐿𝑒𝑥𝑡 = 2𝜋𝑓 2𝜋0 [𝑙𝑛 (𝐺𝑀𝑅 𝑏∗𝐺𝑀𝐷𝑏 ′ )]
𝑏 𝑏
1 𝐺𝑀𝐷 ∗𝐺𝑀𝑅 ′
𝑓𝜇0
∆𝜔𝐿𝑒𝑥𝑡 = 𝑙𝑛 (𝐺𝑀𝑅 𝑏∗𝐺𝑀𝐷𝑏 ′ )
𝑏 𝑏
∆𝜔𝐿𝑒𝑥𝑡
𝐺𝑀𝐷𝑏 ∗𝐺𝑀𝑅𝑏 ′
=𝑒 𝑓𝜇0 (11)
𝐺𝑀𝑅𝑏 ∗𝐺𝑀𝐷𝑏 ′

From Equation (2):

𝐺𝑀𝑅𝑏 ′ = 𝑛2(𝐺𝑀𝑅′𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 )𝑛 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗 ′
√ 𝑗=1,𝑛
𝑖≠𝑗

𝐺𝑀𝑅𝑏 = 𝑛2(𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 )𝑛 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗

√ 𝑗=1,𝑛
𝑖≠𝑗

Equation (4) can be rewritten as:

𝐺𝑀𝐷𝑏 ′ = 𝑛2√∏ 𝑖=1,𝑛 𝐷𝑚𝑖𝑗 ′

𝑗=1,𝑛

𝐺𝑀𝐷𝑏 = 𝑛2√∏ 𝑖=1,𝑛 𝐷𝑚𝑖𝑗

𝑗=1,𝑛

where 𝐷𝑚𝑖𝑗 and 𝐷𝑚𝑖𝑗 ′ are the distance between the sub-conductor i and the image of the sub-
conductor j corresponding to a different distance between sub-conductors.

From Equation (3):

−𝜇𝑟
𝐺𝑀𝑅′𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 = 𝑟 ′ 𝑒 4

−𝜇𝑟
𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 = 𝑟𝑒 4

Copyright © 2017 SES ltd. All rights reserved. Page 25-11

PART III: SES CONTRIBUTIONS

Note that in this approach, the radius of the sub-conductors are kept the same, i.e., 𝑟 = 𝑟 ′ , then
𝐺𝑀𝑅′𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 =𝐺𝑀𝑅𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 and Equation (11) becomes:

𝑛2∏ 𝐷𝑚𝑖𝑗 𝑛2 (𝐺𝑀𝑅′𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 )𝑛 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗 ′

√ 𝑖=1,𝑛
𝑗=1,𝑛 √ 𝑗=1,𝑛 ∆𝜔𝐿𝑒𝑥𝑡
𝑖≠𝑗
𝑛∏ 2 ∏ =𝑒 𝑓𝜇0
2 (𝐺𝑀𝑅 ) 𝐷
𝑖=1,𝑛 𝑖𝑗 𝑛 𝐷𝑚𝑖𝑗 ′
𝑛 𝑠𝑢𝑏𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 √ 𝑖=1,𝑛
√ 𝑗=1,𝑛 𝑗=1,𝑛
𝑖≠𝑗

Then Equation (11) can be simplified further as:

∏ 𝑖=1,𝑛 𝐷𝑚𝑖𝑗 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗 ′
𝑛2 𝑗=1,𝑛 𝑗=1,𝑛 ∆𝜔𝐿𝑒𝑥𝑡
𝑖≠𝑗
√∏𝑖=1,𝑛 𝐷𝑖𝑗 ∏ 𝑖=1,𝑛 𝐷𝑚𝑖𝑗 ′ =𝑒 𝑓𝜇0

𝑗=1,𝑛 𝑗=1,𝑛
𝑖≠𝑗

∏ 𝑖=1,𝑛 𝐷𝑚𝑖𝑗 ∏ 𝑖=1,𝑛 𝐷𝑖𝑗 ′

𝑗=1,𝑛 𝑗=1,𝑛 ∆𝜔𝐿𝑒𝑥𝑡
𝑖≠𝑗 𝑛2
∏ 𝑖=1,𝑛 𝐷𝑖𝑗 ∏ 𝑖=1,𝑛 𝐷𝑚𝑖𝑗 ′ =𝑒 𝑓𝜇0
(12)
𝑗=1,𝑛 𝑗=1,𝑛
𝑖≠𝑗

In the case of two horizontal sub-conductors initially separated vertically by a distance 𝐷 ′ , after
their distance is changed to 𝐷, Equation (12) will become:
2 ∆𝜔𝐿𝑒𝑥𝑡
4𝐻 2 (4𝐻 2 −𝐷2 )∗𝐷 ′ 𝑛2
′2
=𝑒 𝑓𝜇0
𝐷 2 ∗4𝐻 2 (4𝐻 2 −𝐷 )
2𝐻𝐷 ′
𝐷=
√(1−𝑒
𝑛2
∆𝜔𝐿𝑒𝑥𝑡
𝑓𝜇0 )𝐷 ′ 2 +4𝐻 2 𝑒
𝑛2
∆𝜔𝐿𝑒𝑥𝑡
𝑓𝜇0
(13)

In this case, there is no need to worry about a change of the internal impedance since the
conductor radius has not changed.

3 Examples
3.1 Adjustments of Conductor Radius for Two Conductors
The example described in this section demonstrates that the desired series impedance of the
equivalent fence can be achieved by adjusting the radius of the equivalent conductor using
Equation (9).
The two equivalent conductors representing the one mile fence mesh are placed at the top and
bottom of the fence mesh, at 4.938 feet and 0.938 foot, respectively. The series impedance of the
equivalent fence with a radius of 6.3713 inches for the two equivalent conductors is obtained using
HIFREQ. The results are listed in Table 5.

Page 25-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 5: Series impedance of the equivalent fence mesh for the one mile fence section.

Resistance Reactance
Scenarios
(ohms/mile) (ohms/mile)

Realistic fence with 3 vertical wires 0.854942 0.958575

Initial two equivalent conductors with a

radius of 6.3713 inches representing the fence 0.849668 0.922818
as determined using SESeBundle

Series impedance computation difference

0.005274 0.035757
(ohms/mile)

From Table 5, it can be seen that the series impedance of the equivalent fence is smaller than the
desired value. Therefore, the radius of the conductors should be adjusted to get the same
impedance. Note that the resistances are very close and do not need to be improved.
The new adjusted radius of the equivalent conductors can be obtained from Equation (9) and the
computation results are shown in Table 6 for the initial radius and the adjusted radius:
Table 6: Two equivalent conductors with different radii for the one mile fence section.

Initial Adjusted
Parameters
Values Values

Permeability of free space (henries/meter) 1.25664E-06 1.25664E-06

Frequency (Hz) 60 60

Radius (inch) 6.3713 3.5336

Series reactance (ohms/mile) 0.922818 0.957768

External reactance computation difference (ohms/mile)

0.035757 0.000807
(the desired value is 0.958575 ohm/mile)

Relative error (%) 3.73 0.08

0.035757/1609
𝑟 = 6.3713/𝑒 260∗ 1.25664𝐸−06

𝑟 = 3.5336 inches

The computed radius is 3.5336 inches. With this adjusted conductor radius, the two updated
equivalent conductors representing the fence give a series reactance of 0.957768 ohm/mile and
the relative error is reduced from 3.73% to 0.08% with respect to the desired value.
The computation results for the typical fence section of 15 feet are also listed in Table 7. Again,
the relative error is reduced from 6.65% to 2.98% with respect to the desired value when one use
the refinement methodology suggested in this article.

Copyright © 2017 SES ltd. All rights reserved. Page 25-13

PART III: SES CONTRIBUTIONS

Table 7: Two equivalent conductors with different radii for the 15’ fence section.

Parameters Desired Values Initial Values Adjusted Values

Series reactance (ohms/mile) 0.355684 0.332019 0.366270

Radius (inch) 6.3713 3.5336

External reactance
computation difference 0.023666 0.010587
(ohms/mile)

Relative error (%) 6. 65 2.98

3.2 Adjustments of Conductor Distance for Two Conductors

The example described in this section demonstrates that the desired series impedance can be
obtained by adjusting the distance between two equivalent conductors using Equation (13). The
data and computation results are shown in Table 8 and Table 9, respectively.
Table 8: Series impedance of the equivalent fence mesh for the one mile fence section.

Resistance Reactance
Scenarios
(ohms/mile) (ohms/mile)

Realistic Fence with 3 vertical wires 0.854942 0.958575

Two equivalent conductors with a distance

of 2.419 feet representing the fence as 0.847268 0.952254
determined using SESeBundle

Series impedance computation difference

0.007674 0.006321
(ohms/mile)

Table 9: Two equivalent conductors with different distances for the one mile fence section.

Initial Adjusted
Parameters
Values Values

Permeability of free space (henries/meter) 1.25664E-06 1.25664E-06

Frequency (Hz) 60 60

Distance (feet) 2.419 2.215

Series reactance (ohms/mile) 0.952254 0.957532

Radius (inch) 6.3713 6.3713

External reactance computation difference (ohms/mile)

0.006321 0.001043
(the desired value is 0.958575 ohm/mile)

Relative error (%) 0.66 0.11

Page 25-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The computed distance is 2.215 feet. With this adjusted distance, the two equivalent conductors
representing the fence mesh give the series reactance of 0.957532 ohm/mile and the relative error
is reduced from 0.66% to 0.11% with respect to the desired value.
Note that if the difference between the computed resistance values is relatively large, then the
resistivity of the conductors can be adjusted easily to get more accurate results.
The computation results for the typical fence section of 15 feet are also listed in Table 10. However,
in this case, the relative error is already quite low, 0.24% with respect to the desired value.
Reducing the distance will not improve the equivalent result and is a useless step. The main reason
for this situation is believed to be due to the relatively strong edge effects computed in HIFREQ
when you have short fence section models.
Table 10: Two equivalent conductors with different distances for the 15’ fence section

Parameters Actual Values Initial Values Adjusted Values

Reactance (ohms/mile) 0.355684 0.356528 0.364978

Distance (feet) 2.419 2.215

External reactance
computation difference 0.000845 0.009294
(ohms/mile)

Relative error (%) 0.24 2.61

4 Conclusion
In this paper, the methodology aimed at finding equivalent conductors representing a realistic
metallic wire-mesh fence or similar complex structures has been developed and described based
on practical examples using wire-mesh fences.
The proposed method is based on the idea of finding one or several equivalent conductors with
adequate geometrical and electrical characteristics that accurately replace aboveground
structures, such as fences, significantly reducing computer memory and run time in the HIFREQ
computation module. The very dense wire-mesh fence conductor segments are replaced by a small
network of parallel conductors that yield the same series impedance for the entire modeled
structure once the current is injected between the target locations.
Various computer models have been built to illustrate the method and provide meaningful
comparisons between actual realistic fence models and their equivalent models. The equations
used to calculate further adjustments of the conductor radius or the distance between conductors
have been derived to refine and improve the accuracy of the fit further. All the example files used
to obtain the results presented in this paper are provided in the UGC related computer media
folder as a reference template example to carry out similar analysis scenarios yielding equivalent
fence models.

Copyright © 2017 SES ltd. All rights reserved. Page 25-15

 UGM 2017 – CANNON BEACH, OREGON

26 IMMERSED:
SHORT-CIRCUIT SAFETY CRITERIA FOR SWIMMERS
Robert D. Southey and Nina Mitskevitch

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
Electric shock safety criteria are well established for human beings on land. However, the authors are not aware of
any standards that provide explicit electric shock safety criteria for human beings immersed in water. In this paper,
whose objective is to stimulate discussion of this topic in the electrical power industry, electric shock safety criteria
have been proposed for swimmers near AC power facilities, for both fault and load conditions. The use of these
criteria has been illustrated with an example study.

1 Introduction
Electric shock safety criteria are well established for human beings on land. IEEE Standard 80
provides fairly simple guidelines for touch and step voltage limits during 60 Hz faults of limited
duration. IEC Standards 60479-1:2005 and 60479-2:2007 provide quite elaborate guidelines and
data for electric shock scenarios, considering factors such as contact area, wetness and saltiness
of the skin, variation of body resistance with applied voltage for different body parts and paths
through the body, frequency of the applied voltage waveform, various waveform shapes, and so
on. However, the authors are not aware of any standards that provide explicit electric shock safety
criteria for human beings immersed in water.
To be sure, IEC Standard 60479-2:2007 devotes almost 4 pages to the topic and even tabulates
resistivity values for various body tissues and types of water, but never gets past a general
discussion of the matter. So designers looking for a specification of acceptable electric field or
current density criteria for swimmers, resulting from transferred potentials to water from, say, a
nearby faulted substation grounding system, are left high and dry (so to speak).
So what is a lakefront substation grounding system designer to do?
The present article proposes electric shock safety criteria for people immersed in water, based on
data published in the standards cited above and a number of suppositions. The hope is that this
piece will stimulate discussion and further investigation. The authors wish to make it clear that
the electric shock criteria proposed in this paper cannot be considered anything more than
educated speculation and are certainly not based on testing of immersed animals or human
beings. Use them at your own risk.

2 Proposed Criteria
Two approaches are proposed here to establish electric shock safety criteria for swimmers, one
based on a simple application of IEEE Standard 80, the other based on a refinement, making use
of data from IEC Standard 60479-2:2007.
Let us start with a simplistic IEEE Standard 80-based approach. For a person submerged in water,
assume that the body resistance is the internal resistance of 300 Ω specified in IEEE Std. 80-2013,
due to the large, wet contact area, which can include body parts which are not covered with a thick
epidermis, noting that we have not addressed the current path leading to this body resistance so
far. We can then calculate the tolerable touch voltage, using the usual IEEE Standard 80

Copyright © 2017 SES ltd. All rights reserved. Page 26-1

PART III: SES CONTRIBUTIONS

methodology, but with a 300 Ω internal body resistance instead of the 1,000 Ω body resistance
used for standard applications and neglecting the “foot resistance”.
Now for the path. If we divide the tolerable touch voltage obtained as described above by the
maximum possible body length, we obtain one possible limit for the allowable electric field in the
water. One could try to be as conservative as possible by dividing the touch voltage limit by the
length of a human body, from the tips of the toes to the tips of the outstretched fingers above the
head, which is on the order of 3 m for a tall adult. If we assume a homogeneous distribution of
voltage along the human body and a homogeneous medium inside the human thorax (this is really
assuming a lot!), then the current density through the heart should be roughly the same, whether
this electric field is applied longitudinally or transversely through the body.
No matter whether we accept the above methodology to derive a tolerable electric field value, if
the resistivity of the water is greater than that of the human body (on the order of 4 Ω-m for the
trunk and thorax, per IEC Standard 60479-2:2007), then we simply need to study the maximum
electric field in the water and determine whether it exceeds our tolerable electric field limit for
ventricular fibrillation. This is because the low resistivity human body, in parallel with the water,
will tend to reduce the electric field in the water and increase current density, so it is conservative
to consider the electric field in the water, without a swimmer nearby, as this will be higher than
the one that would actually appear in a swimmer’s body. Conversely, if one were to consider the
current density in the water, in the absence of a swimmer, this would tend to underestimate the
current that actually flows through a swimmer’s body, when placed in the water.
On the other hand, if the resistivity of the water is less than that of the human body, then we
should consider instead the current density in the water and determine whether it exceeds our
tolerable current density limit for ventricular fibrillation.
As an example, if we assume an internal resistance of 300 ohms for the human body, a fibrillation
current of 164 mA (for a 0.5 second fault clearing time and decrement factor of 1.0) and a
maximum body length of 3 m (arms outstretched over the head), then we obtain a maximum
allowable electric field of 300 * 0.164 / 3 = 16.4 V/m. On the other hand, as derived below, the
maximum allowable current density would be on the order of 2.56 A/m2.
Now let us try to be a bit cleverer and not make radical assumptions about voltage distribution
along the entirety of the immersed human body. Assume that the cross section of a representative
human adult torso (based on a sample population of one author) to be on the order of 640 cm 2 or
0.064 m2. The average current density flowing through this torso during a conventional hand-to-
foot electrocution scenario would be roughly equal to the fibrillation current divided by the cross
sectional area: i.e., 0.164 A / (0.064 m2) = 2.56 A/m2 (for a 0.5 second fault clearing time and
decrement factor of 1.0). Multiply by a representative tissue resistivity of the human body, say 4
Ω-m, to obtain the resulting electric field. This yields 10 V/m, which is on the same order as the
16.4 V/m value given above, albeit somewhat lower (and more credible, at least for the sample
author population’s body circumference). Use of a greater cross-sectional area of the torso would
reduce the allowable limit (the sample author is a bit on the thin side, but gaining).
The tolerable electric field limit, based on this second approach, would therefore be calculated as
follows:
𝐸𝑚𝑎𝑥 = 𝐼𝑓𝑖𝑏 ∙ 𝜌𝑡𝑜𝑟𝑠𝑜 ⁄𝐴𝑡𝑜𝑟𝑠𝑜
where
Emax is the tolerable electric field limit, in V/m, for people immersed in water,
Ifib is the fibrillation current, in A, computed per IEEE Standard 80 or any other
standard (for voltage applied hand-to-foot),
ρtorso is the average resistivity of the human torso, in Ω-m (4 Ω-m per IEC Standard
60479-2:2007),
Atorso is the cross-sectional area of a human torso, in m2 (0.064 m2, if you are
contemplating ventricular fibrillation of one of this article’s authors).

Page 26-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The corresponding tolerable current density limit, Jmax, in A/m2, would be calculated as follows:
𝐽𝑚𝑎𝑥 = 𝐼𝑓𝑖𝑏 ⁄𝐴𝑡𝑜𝑟𝑠𝑜
Note, however, that this discussion considers only ventricular fibrillation. It does not consider
drowning due to electronarcosis, induced by current flow through the brain, for which we do not
have data. For what it is worth, there is an article at the link below discussing electronarcosis in
animals, in which the current magnitudes recommended for effective stunning of animals are
similar to those recommended for causing ventricular fibrillation:
http://hsa.org.uk/downloads/publications/electricalstunningdownload.pdf.
For the study of steady state conditions, loss of muscle control and therefore potential drowning,
consider that for body currents between 0.5 and 5 mA, IEC Standard 60479-1:2005 predicts
(Table 11 and Figure 20), for AC currents at frequencies of 15 Hz to 100 Hz, “perception and
involuntary muscular contractions likely but usually no harmful electrical physiological effects,”
as compared with “strong involuntary muscular contractions… difficulty in breathing…
immobilization may occur…” for currents in excess of 5 mA. The corresponding electric field,
based on the discussion above, applying the 5 mA limit instead of the fibrillation current limit,
would be on the order of 0.005 x 4 / 0.064 = 0.31 V/m, whereas the current density would be
5xmA / (0.064 m2) = 78 mA/m2. Divide these values by 10 to obtain the perception level for AC,
corresponding to 0.5 mA.
The 5 mA threshold for steady state conditions appears to be consistent with Australian Standard
AS 2832, “Cathodic Protection of Metals,” to which Mr. Doug Gilroy kindly drew our attention
(without endorsing this article in any way). According to Section 5.2.1 of this standard, “electric
field strengths in excess of 3 V/m when applied at chest level can paralyze and cause respiratory
failure…” This voltage limit pertains to DC voltages from impressed current anodes.
IEC Standard 60479-1:2005 indicates that current threshold limits for DC (in general – not
specifically for people in water) are on the order of 4 to 5 times higher, for long duration exposures
(on the order of a few seconds), than those for 60 Hz AC, assuming 10% or less ripple content in
the DC. If we divide the 3 V/m value for DC (from AS 2832) by a factor of 5, we obtain a limit of
0.6 V/m for AC, which is roughly double the 0.31 V/m limit derived based on a 5 mA threshold
for long duration AC and therefore suggests the 0.31 V/m limit to be consistent with experience
obtained with DC anodes.

3 Example Application
Let us consider an example application of these criteria to a lakefront substation grounding
system. Two scenarios will be considered, with very different water resistivity values. The impact
of the presence of the swimmer’s body on the electric field and current density in the water will
also be shown.

3.1 Lakeshore Substation

As Figure 1 shows, a shallow lake has been modeled next to a substation grounding grid, using the
MALZ module of the CDEGS software package. The dimensions of the grounding grid are
30imixi60im. It is made up of 4/0 copper conductors, buried 0.5im deep in a 100iohm-m uniform
soil. The corner of the grid closest to the lake is 7.13im from the shore line. The grid is energized
during a fault to a GPR of 5ikV. The lake is 252im across and gently slopes from the shore, next to
the substation, to its deepest point in the center, where it is roughly 15im deep. The resistivity of
the water in one scenario examined is 100iΩ-m, like the soil around and beneath it; in the other
scenario studied, the resistivity of the water is set to 1iΩ-m. These values are chosen to be,
respectively, larger and smaller than the resistivity of a swimmer’s body.

Copyright © 2017 SES ltd. All rights reserved. Page 26-3

PART III: SES CONTRIBUTIONS

Figure 1: Computer model of lake and substation grid.

Observation points are placed at a depth of 0.3 m below the surface of the soil and water,
beginning in the substation area and extending out across the surface of the lake. Another series
of observation points is placed on a vertical surface that cuts across the substation and the lake,
in order to make it possible to check the electric field and current density as a function of depth
in the water (see Figure 2).

Figure 2: Observation points on vertical surface extending across substation and lake.
Figure 3 shows the electric field distribution along the shore and within the water for the two
different water resistivity values. In the first case, there is no distortion along shoreline, since the
resistivity of the water is the same as that of the soil. On the other hand, when the resistivity of
the water is reduced to 1 Ω-m, there is an abrupt drop in the electric field from land to water. The
10 V/m safety limit occurs at the boundary between the green and yellow colors, located roughly
55 m from the shoreline, for the water resistivity of 100 Ω-m. For the water resistivity of 100 Ω-m,
this is the safety limit that applies, since the water resistivity is greater than the body resistivity of
4 Ω-m. For the 1 Ω-m water, the electric field criterion does not apply: indeed, it can be seen in
Figure 3 that the electric field is less than 10 V/m beginning only 16 m from the shoreline.

Page 26-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 4 shows the current density distribution along the shore and within the water for the two
different water resistivity values. In contrast with the electric field, the current density is
continuous from land to water (if one looks carefully enough!), across the shoreline, near the
corner of the substation, for the 1 Ω-m water. It is notable that the current density is considerably
higher in the 1iΩ-m water than in the 100 Ω-m water. The 2.56iA/m2 safety limit occurs at the
boundary between the green and yellow colors. For the water resistivity of 100iΩ-m, this safety
limit, which occurs on the shore and very near the grid, does not apply, since the water resistivity
is greater than the body resistivity of 4iΩ-m. On the other hand, for the 1iΩ-m water, the current
density safety criterion does apply. It can be seen in Figure 4 that the current density is less than
2.56iA/m2 beginning a full 64im from the shoreline. In comparing Figure 3 with Figure 4, it can
be seen that application of the electric field criterion to the 100iΩ-m water scenario and the
current density criterion to the 1iΩ-m scenario result in the more conservative results.
In both cases, it can be seen that there are hazardous zones in the water.

Lake resistivity is 100 Ω-m Lake resistivity is 1 Ω-m

Figure 3: Electric field distribution in lake for two different water resistivity values.

Copyright © 2017 SES ltd. All rights reserved. Page 26-5

PART III: SES CONTRIBUTIONS

Lake resistivity is 100 Ω-m Lake resistivity is 1 Ω-m

Figure 4: Current density distribution in lake for two different water resistivity values.
So far, we have examined electric field and current density near the surface of the lake. In order
to verify that we have not missed higher values occurring at greater depth, let us plot the electric
field and current density across the lake, along a vertical cross section, defined by the surface
shown in Figure 2. As can be seen in Figure 5 and Figure 6, both electric field and current density
remain constant as a function of depth, except for a very small region close to the shoreline, thus
supporting use of Figure 3 and Figure 4 for the safety evaluation.
Lake resistivity is 100 Ω-m Lake resistivity is 1 Ω-m

Figure 5: Electric field distribution throughout vertical section of lake, for two different water resistivity
values.

Page 26-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Lake resistivity is 100 Ω-m Lake resistivity is 1 Ω-m

Figure 6: Current density distribution throughout vertical section of lake, for two different water
resistivity values.

3.2 Effect of Swimmer Body

As discussed earlier, we have to be careful to use a suitable criterion, which varies according to
water resistivity, in order to judge swimmer safety. This is because the swimmer’s body can distort
both the electric field and the current density occurring in the water, when the water resistivity
differs significantly from that of the swimmer’s body.
To illustrate this effect, we have chosen to model the swimmer illustrated in Figure 7. It is made
up of 6 volumes, each with a resistivity of 4 ohm-m, immersed at a depth of 0.3im, in 100iΩ-m
lake water. Observation points are defined on a horizontal surface bisecting the swimmer’s entire
body. The outline of the swimmer’s body is also modeled by insulated conductors at a depth of
100 m below the lake surface, to serve as a visual reference in the spot 2-D plots of the computation
results.
Figure 8 and Figure 9 show how the swimmer’s body interacts with the water. In Figure 8, it can
be seen that the swimmer’s body increases the electric field in the water at the extremities of the
body that are closest to and furthest away from the substation, i.e., aligned with the electric field
gradient. On the other hand, the electric field in the swimmer’s low resistivity body is significantly
lower than the field to be found in the water. This is an indication of the degree to which applying
a strict electric field safety criterion in high resistivity water is conservative.
In Figure 9, it can be seen that the current density in the water is generally quite low (i.e., less
than 0.1iA/m2). On the other hand, in the swimmer’s low resistivity body, the current that it
attracts from the surrounding water is such that the current density reaches a maximum value of
1.8iA/m2, with roughly 1iA/m2 in the thorax. This is why one cannot use the current density in the
water to judge the level of safety for a swimmer, when the resistivity of the water is higher than
that of the swimmer’s body. This is typically the case for lakes and rivers.

Copyright © 2017 SES ltd. All rights reserved. Page 26-7

PART III: SES CONTRIBUTIONS

Figure 7: Model swimmer – as modeled.

Figure 8: Distortion of electric field by swimmer.

Page 26-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 9: Distortion of current density by swimmer.

4 Conclusions and Disclaimer

In this paper, whose objective is to stimulate discussion of this topic in the electrical power
industry, electric shock safety criteria have been proposed for swimmers near AC power facilities,
for both fault and load conditions. The use of these criteria has been illustrated with an example
study.
Disclaimer:
The authors wish to emphasize that they have seen neither experimental data nor standards
that specify AC electric field or current density limits for human safety in water. This topic, while
discussed in the literature, is not definitively addressed. For example, IEC Standard
60479-2:2007, while devoting almost 4 pages to the topic and tabulating resistivity values for
various body tissues and types of water, never gets past a general discussion of the matter.
Therefore, the values proposed in this paper cannot be considered anything more than educated
speculation and are certainly not based on testing of immersed animals or human beings. Use
them at your own risk.

Copyright © 2017 SES ltd. All rights reserved. Page 26-9

Part IV: Research Work in Progress
 UGM 2017 – CANNON BEACH, OREGON

27 INCREASING COMPUTATION SPEED IN

MALT AND MALZ
USING THE DISCRETE COMPLEX IMAGE METHOD

Marc-André Joyal, Adrian Ngoly, Said Touimer, and Simon Fortin

Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

As the grounding design models encountered in practical studies are becoming larger and more complicated, we
are constantly working to increase the performance of the software, i.e., its computation speed and accuracy. For
this purpose, this article presents a research work in progress that is conducted around the MALT and MALZ
modules of SES Software and that is intended to increase the computation speed when simulating grounding
systems in complex horizontal multilayer soils. More precisely, the method studied herein is the Discrete Complex
Image Method (DCIM), used to compute the soil potentials at low frequency (quasi-static potentials) when a
grounding structure is energized. As a proof of concept, this technique is compared with the method already used
in MALT/MALZ, which serves as a benchmark, and that essentially consists of the Method of Images (MoI) used in
combination with different acceleration techniques.

1 Introduction
In the SES R&D department, a goal we always have in mind is the reduction of the computation
time while maintaining a high degree of accuracy. Although MALT and MALZ can run quite fast
for normal soil structures, the increase in the computation time can be noticeable when the soil
model becomes more complex, i.e., when there are high resistivity contrast ratios between soil
layers and/or when the variation between the thickness of the layers is large. Therefore, the
purpose of this article is first to summarize how the Discrete Complex Image Method (DCIM) can
be applied to compute soil potentials and then find the situations in which it increases the
computational performance without compromising the accuracy.
In the following sections, the Method of Images (MoI) currently used in MALT/MALZ, and the
DCIM are first summarized. Then, a few different models are run with these methods in order to
find their limitations and determine in which situations the DCIM performs better. Finally, we
conclude this article by providing future applications regarding the implementation of the DCIM
in SES Software.

2 Computation of the Soil Potentials in Horizontal Multilayer

Soil Structures
When computing the potentials in multilayer structures, an approach that is often used is the
conversion to the spectral domain using a Fourier (or Hankel) transform. This transform is
actually applied along the plane parallel to the soil layers (r-axis in Figure 1), leaving to solving
for the potentials only along the z-axis, which greatly simplifies the problem.
Let us first assume a point source (i.e., a very short conductor segment) from which a current I is
leaking in an infinite medium. Then, if we assume that this point source is located at the origin
(x = y = z = 0), the potential 𝜙 (in Volts) at a given observation point (x, y, z) is obtained by solving

Copyright © 2017 SES ltd. All rights reserved. Page 27-1

PART IV: RESEARCH WORK IN PROGRESS

Poisson’s equation for quasi-static state and can be expressed by equation (1) in which ρ is the soil
resistivity and 𝑟 = √𝑥 2 + 𝑦 2 .
𝜌𝐼
𝜙(𝑟, 𝑧) = (1)
4𝜋√𝑟 2 + 𝑧 2

Figure 1: Multilayer soil structure showing the definition of the coordinates.

Now, if we take the horizontal multilayer soil structure shown in Figure 1, equation (1) cannot be
used to compute the soil potentials everywhere; it is necessary to consider the effect of the
different thicknesses and resistivity values of the layers. However, since the soil resistivity varies
only along the z direction, it is more convenient to solve the problem in the spectral domain and
restore the expression in the spatial domain by means of the Fourier transform. In fact, the
potential function for horizontal multilayer soils can be computed by the expressions given in
equation (2) where n and m are the source and observation layers, respectively, and J0 is the zeroth
order Bessel function of the first kind [1].
𝜌𝑛 ∞ −𝜆|𝑧|
𝜙𝑛𝑛 (𝑟, 𝑧) = ∫ [𝑒 + 𝐴𝑛𝑛 (𝜆)𝑒 −𝜆𝑧 + 𝐵𝑛𝑛 (𝜆)𝑒 𝜆𝑧 ]𝐽0 (𝜆𝑟)𝑑𝜆
4𝜋 0
𝜌𝑛 ∞ (2)
𝜙𝑚𝑛 (𝑟, 𝑧) = ∫ [𝐴𝑚𝑛 (𝜆)𝑒 −𝜆𝑧 + 𝐵𝑚𝑛 (𝜆)𝑒 𝜆𝑧 ]𝐽0 (𝜆𝑟)𝑑𝜆
4𝜋 0

In the above equations, 𝜙𝑚𝑛 (𝑟, 𝑧) is the potential computed at the observation point in layer m,
when the source is in layer n, whereas 𝜙𝑛𝑛 (𝑟, 𝑧) is the potential when the source and the
observation points are both in layer n. Following this, since the expressions in brackets in the
integrands in equation (2) no longer depend on r, we can solve Ann(λ), Bnn(λ), Amn(λ), and Bmn(λ)
in the spectral domain. In fact, it can be shown that the solution is of the same form as that of a
transmission line terminated at both ends [2]. Therefore, once the functions Ann(λ), Bnn(λ), Amn(λ),
and Bmn(λ) are obtained by solving the well-known transmission line equations, the only thing
that needs to be done to obtain the potential 𝜙(𝑟, 𝑧) in the spatial domain is to compute the
integral given in (2). Unfortunately, this integral includes a Bessel function which is highly
oscillatory and causes the integral to converge very slowly. As a result, techniques to increase the
computational speed are needed, especially when we have to compute the potential at multiple
locations, which is often the case in MALT and MALZ models since there are usually multiple
profile points on the soil surface above substations.
Page 27-2 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

One technique that was found to help reduce the computation time takes advantage of a
remarkable identity known as the Lipschitz formula, which is given by:
∞
1
∫ 𝑒 −𝜆ℎ 𝐽0 (𝜆𝑟)𝑑𝜆 = (3)
0 √ℎ2 + 𝑟 2
If a way can be found to approximate Ann(λ), Bnn(λ), Amn(λ), and Bmn(λ) by a sum of exponential
terms (as in [3], for instance), use of the Lipschitz formula reduces the above integrals to the
calculation of a sum of “point source” terms, which can greatly accelerate the calculation of the
integrals.
Two methods that use this technique are summarized in the following subsections: the MoI
(already implemented in MALT/MALZ) and the DCIM.

2.1 The Method of Images in MALT/MALZ

The Method of Images (MoI) that is implemented in the MALT and MALZ modules uses a Taylor
series expansion to represent the integrands in Equation (2) as an infinite sum of exponential
terms. With the help of the Lipschitz formula, this powerful technique is actually equivalent to
using an infinite number of images of the source reflected on the layer interfaces to compute the
soil potential at a given location. The potential is actually the sum of the contributions of the
source and of all its images [1, 3]. An example of a source and its images in a two-layer soil is
shown in Figure 2.

Figure 2: Example of a point source and its images in a two-layer soil.

In practice, only a finite number of images can be computed. In MALT and MALZ, the number of
images required to reach a specified accuracy is determined automatically by the program, and is
known as the “Number of Terms”. The larger the resistivity ratio is from one layer to the other,
the more significant the contribution of the images will be, and the larger the Number of Terms
will be. Furthermore, as the images are generated by the layer interfaces, more images are needed
when the soil includes many layers.
It is also worth mentioning that MALT and MALZ include different acceleration algorithms that
are used in different situations. The description of those algorithms is out of the scope of this
Copyright © 2017 SES ltd. All rights reserved. Page 27-3
PART IV: RESEARCH WORK IN PROGRESS

article, but let us only mention that when there are multiple observation points close to each other
and far enough from a conductor segment, interpolation tables can be applied to compute the
potentials at the observation points due to the given segment. This results in a reduced
computation time as it is not necessary to use all images for computing the potentials at each
observation point.

2.2 The Discrete Complex Image Method

The basic principle of the DCIM is that each spectral domain functions Ann(λ), Bnn(λ), Amn(λ), and
Bmn(λ) in (2) can be approximated by a sum of complex exponentials (or complex images).
Consequently, even if the DCIM concept still dealing with images, it is defining their strength and
location as complex numbers. From a mathematical point of view, this means that for each image
there is an additional degree of freedom, which corresponds to the phase of the exponential.
Therefore, we could in principle determine the soil potential with fewer images, resulting in fewer
terms to compute, leading to a reduction in the computation time.
Based on this principle, it is thought that the DCIM can help reduce the computation time for
complex soil models, i.e., those leading to a high number of terms (or images). As the number of
images tends to increase with respect to the reflection coefficients at the layer interfaces, we will
say that a soil model is complex when there are high resistivity contrasts between the layers.
2.2.1 Principle
As mentioned above, the DCIM principle is based on the assumption that the potential in
multilayer structures can be approximated by a sum of exponentials [3]. As a result, the first
expression in equation (2) can be expressed as follows using the Lipschitz formula given in
equation (3) (the second expression in equation (2) can be obtained the same way):
𝐾 𝐿
𝜌𝑛 ∞ −𝜆|𝑧|
(𝑟,
𝜙𝑛𝑛 𝑧) = ∫ [𝑒 + ∑ 𝑎𝑖 𝑒 (𝛼𝑖−𝑧) + ∑ 𝑏𝑖 𝑒 (𝛽𝑖+𝑧) ] 𝐽0 (𝜆𝑟)𝑑𝜆 (4)
4𝜋 0
𝑖=0 𝑖=0

In equation (4), 𝑎𝑖 , 𝑏𝑖 , 𝛼𝑖 , and 𝛽𝑖 are the coefficients to be determined whereas 𝐾 and 𝐿 are the
number of terms (i.e., images) needed to reach convergence for the two sums. The coefficients are
called GPOF coefficients as they are obtained from the Generalized Pencil of Function (GPOF)
algorithm. The GPOF is similar to Prony’s method, but it was found to be more stable. These
methods are not described here, but the interested reader may consult [4] for more details.
Therefore, the potential in the spatial domain (4) can be expressed as summations over the total
number of terms 𝐿 and 𝐾 as follows:
𝐾 𝐿
𝜌𝑛 1 𝑎𝑖 𝑏𝑖
𝜙𝑛𝑛 (𝑟, 𝑧) = [ +∑ +∑ ] (5)
4𝜋 √𝑧 2 + 𝑟 2 √(𝑧 − 𝛼𝑖 )2 + 𝑟 2 √(𝑧 + 𝛽𝑖 )2 + 𝑟 2
𝑖=0 𝑖=0

2.2.2 Multilevel DCIM

In order to apply the DCIM algorithm, the spectral domain functions must be sampled correctly.
However, depending on the number of samples used to discretize these functions, the solution
may be inaccurate (if not enough samples are used) or the computations may be too time
consuming (if too many samples are used). In particular, the difficulty in this problem is that the
function to sample may not vary in the same manner over the whole λ range. For instance, the
function may be monotonic for a given range of λ and have sharp variations in other ranges. As it

Page 27-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

is not convenient to use only one sampling rate in such situations, the function may be cut into
multiple regions, or levels. As an illustrative example, Figure 3 shows a spectral function (Amn(λ),
for instance) that is sampled in two levels. In Level 1, a small sampling rate is enough since the
slope of the function is quite small. On the other hand, a higher sampling rate is needed for Level
2 since the slope of the curve is quite large. For functions that have sharp variations, a high
number of levels can be needed, but this comes at the expense of a higher computation time since
GPOF coefficients need to be generated for each level and therefore additional terms need to be
added to the summations in equation (5).

Figure 3: Sampling of a spectral domain function in two-levels.

The challenge in the multilevel DCIM is to correctly determine the number of samples, the
number of levels, and the boundaries λmax1, λmax2, …, λmaxN between the levels to be used in the
algorithm. In the case of soil potential computations, the cases studied so far seem to show that a
two-level approach is acceptable in most situations, as the sharp variations occur for small values
of λ. Therefore, the implementation that has been done to obtain the results presented in the next
section is a two-level approach.
Note that the above multilevel method is very simple as it simply cuts the λ-axis into N regions.
However, at higher frequencies when the propagation constant can be complex valued, one could
decide to sample the function along a complex path, i.e., for complex values of λ. In that case, this
approach is not only used for sampling purposes, but also to handle singularities that may appear.
This is out of the scope of this paper but for more details the references in [5] can be consulted.

3 Case Studies
In order to validate, and study the limitations of, the DCIM with respect to the MoI implemented
in MALT/MALZ, the following grids are computed with both methods and compared:
1. 100’ × 100’ grid, 15 × 15 segments, 101 × 101 profile points;
2. 100’ × 100’ grid, 19 × 19 segments, 101 × 101 profile points;
3. 100’ × 100’ grid, 23 × 23 segments, 101 × 101 profile points;
4. 100’ × 100’ grid, 27 × 27 segments, 101 × 101 profile points;
5. 100’ × 10’ grid, 23 × 23 segments, 101 × 101 profile points.
Copyright © 2017 SES ltd. All rights reserved. Page 27-5
PART IV: RESEARCH WORK IN PROGRESS

For the above cases, the conductor segments and profile points are both located in the top layer.
The segments are at a depth of 1.5 feet, whereas the observation points are on the soil surface.
Also, in addition to the 101 × 101 profile points covering the ground grids, individual observation
points are added from the center of the grid to about 10,000 feet from it. This is done to validate
that the relative error is computed up to points far enough from the grid; this can be an issue in
DCIM problems if this is not carefully addressed. Furthermore, the grids can be computed in one
of the two soil models given in Table 1.
Table 1: Soil Models used in this Study

Simple Soil Model Complex Soil Model

Resistivity Thickness Resistivity Thickness

(ohm-m) (feet) (ohm-m) (feet)

Top Layer 100 4 100 4

Second 300 4 3000 0.03

Layer

Third 100 4 100 4

Layer

Bottom 200 Infinity 200 Infinity

Layer

Table 2: Summary of the Comparison Results for the Tested Scenarios

Grid Soil Model MALT - DCIM - Relative Error1

Execution Execution (%)
Time (minutes) Time (minutes)

1 Simple 0:06 5:51 0.446

1 Complex 14:47 6:44 0.235

2 Complex 15:35 10:53 0.236

3 Complex 16:24 16:21 0.236

4 Complex 16:36 22:39 0.235

5 Complex 31:43 26:14 0.258

1 For a better accuracy, the results in this column have been obtained by disabling the acceleration algorithms.

Page 27-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The results of different scenarios are shown in Table 2. For each scenario, this table shows the
execution time with both methods and the relative error of the DCIM results with respect to the
MoI. It is worth mentioning that for this last parameter, the acceleration algorithms have been
disabled as we want the most accurate benchmark possible.
Regarding the computation process, when Grid 1 is in the Simple soil model, only 15 terms are
generated for the MoI whereas 9,940 are needed when it is in the Complex soil model. In
comparison, the DCIM needed 31 and 57 terms when the grids are in the Simple and Complex soil
models, respectively. From this information, it seems that the proposed DCIM algorithm could
provide a significant reduction in the calculation time for complex soil structures, i.e., those
leading to a high number of images, depending on the performance of the acceleration algorithms
used in the MoI. On the other hand, when the grid is in the Simple soil model, the current DCIM
algorithm needs to be optimized as it clearly generates too many terms, resulting in a MoI that
runs much faster, as can be seen from the execution time shown in the first row of Table 2.
Therefore, in this case, there is no need to use the DCIM algorithm as the MoI is fast enough.
From the execution time obtained for Grids 2 to 4, one might think that starting from a certain
number of conductors, the MoI outperforms the DCIM. However, the results for Grid 5, shown in
the last row of Table 2, contradict this statement, since the latter grid has the same number of
conductors as that of Grid 4, but in this case, the DCIM runs faster than the MoI. This is explained
by the fact that in that scenario, since Grid 5 is compressed along the y-axis (but the same number
of observation points than that of Grid 4 is used), half of those points are closer to long conductor
segments. This can be seen in Figure 4 and Figure 5 showing the MALT models of Grids 4 and 5,
respectively. In this case, the acceleration algorithms can only be applied to a reduced number of
points as most of them are no longer far enough from conductor segments. From this example, it
is thought that in the future, the DCIM algorithm could be used for cases involving arbitrarily
shaped grounding systems that cannot benefit from the acceleration algorithms used in
MALT/MALZ.

Copyright © 2017 SES ltd. All rights reserved. Page 27-7

PART IV: RESEARCH WORK IN PROGRESS

Figure 4: MALT model for Grid 4.

Figure 5: MALT model for Grid 5.

Finally, Figure 6 and Figure 7 show how the relative error of the DCIM with respect to the MoI
varies for Grid 5. Figure 6 shows the relative error for the profile points over the whole grid
whereas Figure 7 shows the relative error for the individual observation points going from the
center of the grid up to 10,000 feet away, where the soil potential has decreased to 0.24% of that
just above the grid (from about 4600 V over the grid to 11 V, 10,000 feet away). Both results show
that the DCIM can lead to a good accuracy, at least for the cases presented here. However, it also
shows that the error tends to increase and become less stable at large distances from the grid,
which will need to be carefully studied and addressed in the future.

Page 27-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 6: Relative error on the computed soil potentials just above the grid for Grid 5.

Figure 7: Relative error on the computed soil potentials up to 10,000 feet from the center of the grid for
Grid 5.

Copyright © 2017 SES ltd. All rights reserved. Page 27-9

PART IV: RESEARCH WORK IN PROGRESS

4 Conclusion
The work reported in this article provides the state of the R&D work at SES regarding the use of
the Discrete Complex Image Method for increasing the computation speed of the soil potentials
in MALT and MALZ. The current progress shows that for complex soil structures, the DCIM can
run faster than the current version of MALT and MALZ, depending on the grounding system,
without even using other acceleration algorithms available in the program. It is therefore thought
that in the near future, this method could become available in MALT and MALZ, not only for
computing soil potentials, but generalized in the whole program for the current distribution
computation, the electric field, etc. In combination with the MoI, the DCIM could fill a gap in the
program such that arbitrarily shaped grids and complex soil structures can be handled more
easily. Furthermore, it is believed that it will be possible to implement it in order to increase the
computation speed for other soil types (cylindrical, finite volumes, etc.) and even for higher
frequencies, i.e., in HIFREQ.

5 References
[1] J. He, R. Zeng, and B. Zhang, Methodology and Technology for Power System Grounding, Singapore:
John Wiley & Sons Singapore Pte. Ltd, 2013.

[2] K. A. Michalski and J. R. Mosig, "Multilayered media Green's functions in integral equation
formulations," IEEE Transactions on Antennas and Propagation, vol. 45, no. 3, March, 1997.

[3] Y. L. Chow, J. J. Yang, and G. E. Howard, "Complex images for electrostatic field computations in
multilayered media," IEEE Transactions on Microwave Theory and Techniques, vol. 39, no. 7, July,
1991.

[4] T. K. Sarkar and O. Pereira, "Using matrix pencil method to estimate the parameters of a sum of
complex exponentials," IEEE Antennas and Propagation Magazine, vol. 37, no. 1, February 1995.

[5] A. Alparslan, M. I. Aksun, and K. A. Michalski, "Closed-form Green's functions in planar layered media
for all ranges and materials," IEEE Transactions on Microwave Theory and Techniques, vol. 39, no.
7, July, 1991.

Page 27-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

28 DERIVATIVE CONSTRAINED
POLARIZATION CURVE FITTING

Adrian Ngoly, Simon Fortin and Farid P. Dawalibi

Safe Engineering Services & technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract

This research article expounds the recent development of an Interior Point Method computational infrastructure,
implemented ab initio, for the solution of industrial scale, inequality constrained, electrical engineering
optimization problems at SES technologies ltd., with a particular emphasis placed on the polarization curve fitting
process contained in the newly developed CorrCAD software package. In this work, the mathematical structure of
derivative constrained least-squares polarization curve fitting is developed in detail, which serves as clear
demonstration of the power of modern convex optimization methods when applied to numerical data fitting in
nonlinear electrical science.

1 Introduction

A polarization curve represents the electrical characteristics of a nonlinear resistor. Specifically,

across the sciences and engineering, the current-voltage characteristic 𝑖(𝑣), or the voltage-current
characteristic 𝑣(𝑖), of a nonlinear, two-terminal, electrical device is often referred to as a
polarization curve. Beyond the (hitherto) traditional interests of the contemporary electrical
scientist, the electrical characterization of nonlinear devices or processes, by means of their
polarization curves, has found important applications in a multitude of (seemingly) disparate
areas of endeavor, including, but not limited to:
1. Materials engineering: In the characterization of the hysteretic properties of the
memristor (that is, the memory resistor), the fourth fundamental electrical circuit
element [1](p.81).
2. Neuromorphic engineering: In the development of the neuristor for synthesizing the
electrical spiking behavior of a firing neuron [2](p.114).
3. Electrochemical engineering: In the state-of-charge (SOC), or depth-of-discharge (DOD),
assessment and management in modern lithium-ion battery systems [3].
4. Corrosion engineering: In the characterization of electrochemical corrosion processes,
with the express aim of predicting, preventing, and protecting, gas pipeline and electrical
utility assets from being jeopardized by corrosion [4].
5. Biomedical Engineering: In the selection and assessment of corrosion resistant materials
for dental implants and extra-oral prostheses and devices [5].
Electrical science is bold and broad. From an electrical systems perspective, what distinguishes
each of these application areas from the other is the nonlinearity of the polarization curve
governing the device or process under consideration. Therefore, given a polarization curve
associated with a nonlinear resistor present in an electrical system, the theory of nonlinear
electrical network analysis has been thoroughly developed in order to ascertain the various branch
voltages and currents in the circuit [6]. However, nonlinear electrical theory is a significant
departure from the more familiar and traditional linear electrical theory; linear theory being a

Copyright © 2017 SES ltd. All rights reserved. Page 28-1

PART IV: RESEARCH WORK IN PROGRESS

subset of nonlinear theory. Specifically, in order to construct a mathematical representation of a

nonlinear electrical device, several mathematical details are presented that are necessary for a
good understanding of the issues involved in numerical polarization curve fitting. Readers
disinterested in the state of the art techniques in computational science and engineering
mathematics developed in §2-§4 can skip directly to §5 to see typical results using the approaches
presented in this article. Others who are analytically minded or those who have occasional
insomnia can delve into the following equations, for a mathematical feast or for temporary relief,
respectively.

Consider the Taylor series expansion [7](p.760) of the current-voltage characteristic 𝑖(𝑣) centered
at some operating voltage 𝑣o :
∞ 𝑖 (𝑘) (𝑣o ) 𝑖 ′′ (𝑣o )
𝑖(𝑣) = ∑ (𝑣 − 𝑣o )𝑘 = 𝑖(𝑣o ) + 𝑖 ′ (𝑣o )(𝑣 − 𝑣o ) + (𝑣 − 𝑣o )2 + ⋯ (1)
𝑘=0 𝑘! 2
Similarly, the Taylor series expansion of the voltage-current characteristic 𝑣(𝑖), centered at an
operating current 𝑖o , is given by
∞𝑣 (𝑘) (𝑖o ) 𝑣 ′′ (𝑖o )
𝑣(𝑖) = ∑ (𝑖 )𝑘 ) ′ (𝑖 )(𝑖
− 𝑖o = 𝑣(𝑖o + 𝑣 o − 𝑖o +) (𝑖 − 𝑖o )2 + ⋯ (2)
𝑘=0 𝑘! 2
In the preceding expressions for 𝑖(𝑣) and 𝑣(𝑖), 𝑖 (𝑘) (𝑣o ) is equal to the 𝑘 th order derivative of 𝑖(𝑣)
with respect to 𝑣, evaluated at the operating point 𝑣 = 𝑣o , and 𝑣 (𝑘) (𝑖o ) is equal to the 𝑘 th order
derivative of 𝑣(𝑖) with respect 𝑖, evaluated at the operating point 𝑖 = 𝑖o , namely

d𝑘 𝑖(𝑣) d𝑘 𝑣(𝑖)
𝑖 (𝑘) (𝑣o ) = | , 𝑣 (𝑘) (𝑖o ) = | . (3)
d𝑣 𝑘 𝑣=𝑣 d𝑖 𝑘 𝑖=𝑖
o o

Importantly, the differential conductance 𝒢(𝑣) and the differential resistance ℛ(𝑖) are defined as
the first order derivatives of 𝑖(𝑣) and 𝑣(𝑖), respectively, with respect to their arguments:

d𝑖(𝑣) d𝑣(𝑖)
𝒢(𝑣) ≝ , ℛ(𝑖) ≝ (4)
d𝑣 d𝑖
As one can see, in stark contrast to the Taylor series expansions in (1) and (2), the voltage-current
characteristic of a linear resistor is a linear function, comprising a single term, namely 𝑣(𝑖) = R𝑖,
where R is the resistance (or, equivalently, the current-voltage characteristic is the linear function
𝑖(𝑣) = G𝑣, where G = 1/R is the conductance).
However, to employ methods of nonlinear circuit analysis, the polarization curve associated with
a nonlinear circuit element must be known. Unfortunately, scientists and engineers usually do not
have an explicit mathematical function that governs the underlying polarization behavior of the
device or process under study. What scientists and engineers do usually have, however, is
experimental electrical measurement data (that is, a set of currents measured at corresponding
operating voltages). It is therefore necessary to find, or construct, a mathematical function that
behaves like the measurement data. This is a difficult and tortuous task. Traditionally, scientists,
particularly corrosion scientists, have used sophisticated nonlinear models to fit experimental
polarization data [8], [9] ,[10], [11]. Besides the fact that nonlinear models are, in general, quite
challenging to deal with mathematically and computationally, the use of nonlinear models
requires deep knowledge of the models themselves if they are to be used effectively. Moreover,
perhaps the most difficult aspect of using nonlinear models is the appropriate selection of an
Page 28-2 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

initial set of parameters (that is, contriving an initial guess) that is required by manual or
automatic curve fitting processes [10](p.617).
In the context of least-squares curve fitting, if the initial set of the parameters of a nonlinear model
is close to the set of parameters that minimizes the least-squares error, then a (nonlinear) least
squares solver can converge quite rapidly to the closest local minimum of the least squares
objective function. However, if the initial set of the parameters of a nonlinear model is not close
enough to the set of parameters that minimizes the least-squares error, the (nonlinear) least
squares solver can fail to converge. This convergence failure often occurs when the nonlinear
model becomes insensitive to changes in its parameters [12]; and it is difficult to predict, a priori,
if and/or when this will occur. Simply put, unless one can prove that the nonlinear least-squares
objective function is strictly convex, thereby possessing one unique global minimum, the
performance of nonlinear least squares solvers, when used with certain nonlinear models, can be
unpredictable and unreliable. This performance unpredictability and unreliability makes the
automation of nonlinear models in commercial software difficult, if not impossible, especially if
users (rightfully) expect to input arbitrary data to the program. If one is to automate the curve
fitting process, the model should be simple enough to get results quickly and reliably, and flexible
enough to handle intricate input data. But most importantly, the model and the fitting method
should provide accurate and consistent results.
This research article is organized as follows. In §2, nonlinear electrical device using polynomials
is described in detail. In §3, the method of unconstrained least squares polarization curve fitting
is developed using the polynomial expansions elucidated in §2. In §4, inequality constrained
convex optimization, by means of the barrier method, is used to constrain the first order derivative
least-squares polynomial curve fits. The motivation of §4 is threefold:
1. The uniqueness theorem for resistive electrical circuits states that any circuit made of
strictly increasing two-terminal resistors and independent sources has at most one
solution [6](p.277). This requirement must be satisfied if an iterative nonlinear circuit
solver is to reliably, and consistently, converge to a unique solution.
2. Moreover, there are to obtain the expression 𝑣(𝑖) from 𝑖(𝑣), or 𝑖(𝑣) from 𝑣(𝑖), which
corresponds to function inversion, it is necessary that the functions 𝑖(𝑣) and 𝑣(𝑖) be
strictly increasing.
3. The method of unconstrained least squares polarization curve fitting detailed in §3 can
lead to unwanted oscillations in the fit. Such oscillatory behavior is eliminated by
constraining the fit to be strictly increasing.
In this work, the strictly increasing condition is enforced by requiring that the first order
derivative of the fitted curve be strictly positive, which corresponds to an inequality constraint.
Detailed numerical results are given in §5. This article has been prepared for SES software
developers and users alike as a reference document for an in-depth understanding of the
methodology behind the newest curve fitting engine of CorrCAD and other future curve fitting
tools.

Copyright © 2017 SES ltd. All rights reserved. Page 28-3

PART IV: RESEARCH WORK IN PROGRESS

2 Nonlinear Electrical Device Modeling using Polynomials

Rather than toiling with the cumbrous infinite Taylor series expansions in (1) and (2), consider
expanding the current-voltage characteristic 𝑖(𝑣) on the voltage interval 𝑎 ≤ 𝑣 ≤ 𝑏 in terms of a
finite linear combination of functions:
N
𝑖(𝑣) = ∑ 𝑥𝑛 𝜙𝑛 (𝑣) , ∀𝑣 ∈ ℝ[𝑎,𝑏] , (5)
𝑛=1

where 𝜙𝑛 (𝑣): ℝ → ℝ, ∀𝑛 ∈ ℤ[1,N] are suitable basis (or trial, or expansion) functions and 𝑥𝑛 , ∀𝑛 ∈
ℤ[1,N] are coefficients that are to be determined. Here, the notation ℤ[1,N] is shorthand for the set
of integers, {1,2, … , N}, namely ℤ[1,N] ≝ {1,2, … , N}; this notation is used throughout to ease various
subtleties in the computational implementation. Also, the notation 𝑣 ∈ ℝ[𝑎,𝑏] is shorthand for 𝑎 ≤
𝑣 ≤ 𝑏. The first order derivative of the current 𝑖(𝑣) in (5) with respect to the voltage 𝑣 gives the
differential conductance of the current-voltage characteristic, namely:
d𝑖(𝑣) N d𝜙𝑛 (𝑣)
𝒢(𝑣) = =∑ 𝑥𝑛 , ∀𝑣 ∈ ℝ[𝑎,𝑏] . (6)
d𝑣 𝑛=1 d𝑣
Importantly, the current-voltage expansion 𝑖(𝑣) in (5), and the corresponding differential
conductance 𝒢(𝑣) in (6), are linear functions of the expansion coefficients 𝑥𝑛 , ∀𝑛 ∈ ℤ[1,N]. The
current-voltage expansion in (5) is therefore linear with respect to the expansion coefficients
𝑥𝑛 , ∀𝑛 ∈ ℤ[1,N], albeit nonlinear with respect to the voltage variable 𝑣. In this work, the expansion
functions in (5) are chosen to be the Chebyshev polynomials of the first kind, which are denoted
by T𝑛 (𝑢): ℝ → ℝ and defined as

T𝑛 (𝑢) ≝ cos(𝑛 cos−1 𝑢) , ∀𝑛 ∈ ℕ0 . (7)

In the above, cos−1 𝑢 is the inverse cosine or arc-cosine function, which is not to be confused with
the reciprocal of the cosine function (which is the secant function), namely
1
cos−1 𝑢 ≠ (cos 𝑢)−1 = = sec 𝑢.
cos 𝑢
Moreover, ℕ0 is the set of nonnegative integers, namely ℕ0 ≝ {0,1,2,3, … }. Observe that T0 (𝑢) and
T1 (𝑢) are easily computed directly from (7):

T0 (𝑢) = cos(0) = 1
(8)
T1 (𝑢) = cos(cos −1 𝑢) = 𝑢
To obtain the Chebyshev polynomials T𝑛 (𝑢) for 𝑛 ≥ 2, consider exploring the mathematical
structure of T𝑛+1 (𝑢), namely
T𝑛+1 (𝑢) = cos((𝑛 + 1) cos −1 𝑢) = cos(𝑛 cos −1 𝑢 + cos−1 𝑢).

Now, from Leonhard Euler’s seminal formula 𝑒 𝑗𝜗 = cos 𝜗 + 𝑗 sin 𝜗, where 𝑗 = √−1, one can write
𝑒 𝑗(A±B) = cos(A ± B) + 𝑗 sin(A ± B). From the basic exponentiation identity 𝑒 𝑥+𝑦 = 𝑒 𝑥 𝑒 𝑦 , one can
also write 𝑒 𝑗(A±B) = 𝑒 𝑗A 𝑒 ±𝑗B = (cos A + 𝑗 sin A)(cos B ± 𝑗 sin B) from which it immediately follows
that
𝑒 𝑗(A±B) = (cos A cos B ∓ sin A sin B) + 𝑗(sin A cos B ± cos A sin B).
Therefore, from the above expression, the following trigonometric identities emerge
Page 28-4 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

cos(A + B) = cos A cos B − sin A sin B,

cos(A − B) = cos A cos B + sin A sin B.
By adding the first line to the second line in the above expression, one obtains
cos(A − B) + cos(A + B) = 2 cos A cos B.
Therefore, by letting A = 𝑛 cos −1 𝑢 and B = cos −1 𝑢 in the above, one obtains
cos((𝑛
⏟ − 1) cos −1 𝑢) + cos((𝑛
⏟ + 1) cos −1 𝑢) = 2 ⏟
cos(𝑛 cos −1 𝑢) ⏟
cos(cos −1 𝑢).
T𝑛−1 (𝑢) T𝑛+1 (𝑢) T𝑛 (𝑢) 𝑢

The above expression corresponds to the recurrence relation for the Chebyshev polynomials of
the first kind, which can be written succinctly as

T𝑛+1 (𝑢) = 2𝑢T𝑛 (𝑥) − T𝑛−1 (𝑢), ∀𝑛 ∈ ℕ, (9)

where ℕ = {1,2,3, … } is the set of natural numbers and, from (8), T0 (𝑢) = 1 and T1 (𝑢) = 𝑢 start
the recurrence relationship. The first seven Chebyshev polynomials T𝑘 (𝑢), ∀𝑘 ∈ ℤ[0,6] are
illustrated in Figure 1(a). For certain practitioners, it is often more convenient to perform the
index substitution 𝑘 = 𝑛 + 1 (with 𝑛 = 𝑘 − 1) in (9), so that the recurrence relation becomes

T𝑘 (𝑢) = 2𝑢T𝑘−1 (𝑢) − T𝑘−2 (𝑢), ∀𝑘 ≥ 2. (10)

Now, by evaluating the first order derivative of (10) with respect to the variable 𝑢, whilst applying
the product rule to 𝑢T𝑘−1 (𝑢), one obtains the recurrence relation for the first order derivative of
the Chebyshev polynomials of the first kind

T𝑘′ (𝑢) = 2𝑢T𝑘−1

′ (𝑢) ′ (𝑢),
+ 2T𝑘−1 (𝑢) − T𝑘−2 ∀𝑘 ≥ 2, (11)

where T0′ (𝑢) = 0 and T1′ (𝑢) = 1. The first seven first order derivatives T𝑘′ (𝑢), ∀𝑘 ∈ ℤ[0,6] of the
Chebyshev polynomials T𝑘 (𝑢), ∀𝑘 ∈ ℤ[0,6] illustrated in Figure 1(b).

Copyright © 2017 SES ltd. All rights reserved. Page 28-5

PART IV: RESEARCH WORK IN PROGRESS

k=0 k=1 k=2 k=3 k=4 k=5 k=6

-1
(a)
-1 0 1

10
k=0 k=1 k=2 k=3 k=4 k=5 k=6

-5

(b)
-10
-1 0 1

Figure 1: Structure of the Chebyshev polynomials of the first kind. (a) The first seven Chebyshev
polynomials T𝑘 (𝑢), ∀𝑘 ∈ ℤ[0,6] obtained using the recurrence relation given inset (and detailed in (9)) with
T0 (𝑢) = 1 and T1 (𝑢) = 𝑢. (b) The first seven first order derivatives of the Chebyshev polynomials T𝑘′ (𝑢), ∀𝑘 ∈
ℤ[0,6] obtained using the recurrence relation given inset (and detailed in (11)) along with T0 (𝑢) = 1, T0′ (𝑢) =
0, T1 (𝑢) = 𝑢 and T1′ (𝑢) = 1.
In this work, the N basis functions 𝜙𝑛 (𝑣): ℝ → ℝ, ∀𝑛 ∈ ℤ[1,N] required by the expansion in (5) are
selected to be the Chebyshev polynomials of the first kind, namely

𝜙𝑛 (𝑣) = T𝑛−1 (𝑢(𝑣)), ∀𝑛 ∈ ℤ[1,N] (12)

Here, it is important to expression the interval −1 ≤ 𝑢(𝑣) ≤ +1 of the Chebyshev polynomials in

terms of the voltage interval 𝑎 ≤ 𝑣 ≤ 𝑏, which is sometimes referred to as interval
parameterization (or interval transplantation), and is accomplished by means of the following
manipulation
Page 28-6 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

𝑣−𝑎 𝑣−𝑎
𝑎 ≤𝑣 ≤𝑏 ⟶0≤𝑣−𝑎 ≤𝑏−𝑎 ⟶0≤ ≤ 1 ⟶ 0 ≤ 2( ) ≤ 2.
𝑏−𝑎 𝑏−𝑎
Therefore, by subtracting +1 from all sides of the last inequality above, one obtains the required
expression for 𝑢(𝑣):
𝑣−𝑎
−1 ≤ 2 ( ) − 1 ≤ +1.
⏟ 𝑏−𝑎
𝑢(𝑣)

By the chain rule, the first order derivative of the basis function 𝜙𝑛 (𝑣), which is required to
determine the differential conductance 𝒢(𝑣) in (6), is written as
d𝜙𝑛 (𝑣) dT𝑛−1 (𝑢(𝑣)) d𝑢(𝑣) dT𝑛−1 (𝜉) 2
= ⋅ = | ⋅ , ∀𝑛 ∈ ℤ[1,N] .
d𝑣 d𝑢(𝑣) d𝑣 d𝜉 𝜉=𝑢(𝑣)
𝑏−𝑎

3 Unconstrained Least-Squares Polarization Curve Fitting

Now let 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M] be the currents at measured at the M voltages 𝑣𝑚 , ∀𝑚 ∈ ℤ[1,M] . By setting
the current-voltage relationship in (5), evaluated at the M measured voltages 𝑣𝑚 , ∀𝑚 ∈ ℤ[1,M], to
the M measured currents 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M], one can write
N
∑ 𝑥𝑛 𝜙𝑛 (𝑣𝑚 ) = 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M] . (13)
𝑛=1

Here the currents 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M] are known; they may come from a known 𝑖(𝑣) curve or from
measurement data. Moreover, it is assumed that the number of measurements M is greater than
or equal to the number expansion functions N, namely M ≥ N, in which case the system in (13)
corresponds to an overdetermined linear system of equations, which can be written compactly as

𝚽𝐱 = 𝐢. (14)

In (14), 𝚽 ∈ ℝM×N is the rectangular matrix with entries given by Φ𝑚𝑛 = 𝜙𝑛 (𝑣𝑚 ) (wherein the
rows index the measurement voltages and the columns index the expansion functions), 𝐱 ∈ ℝN×1
is the unknown vector that contains the N expansion coefficients 𝑥𝑛 , ∀𝑛 ∈ ℤ[1,N], and 𝐢 ∈ ℝM×1 is
the right hand side vector that contains the M measured currents vector that contains the N
expansion coefficients 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M] , namely
𝜙1 (𝑣1 ) ⋯ 𝜙N (𝑣1 ) 𝑥1 𝑖1
𝜙1 (𝑣2 ) ⋯ 𝜙N (𝑣2 ) 𝑥2 𝑖2
𝚽= ∈ ℝM×N , 𝐱 = ∈ ℝN×1 , 𝐢 = ∈ ℝM×1 .
⋮ ⋮ ⋮ ⋮
⏟ [𝜙 1 M)
(𝑣 ⋯ 𝜙N (𝑣M )] ⏟ [𝑥N ] ⏟ [𝑖M ]
tall−and−thin vector of 𝐮𝐧𝐤𝐧𝐨𝐰𝐧 vector of 𝐤𝐧𝐨𝐰𝐧
rectangular coefficient matrix coefficients measured currents

Importantly, since it has been assumed that M ≥ N, the matrix 𝚽 ∈ ℝM×N is a tall-and-thin matrix.
To facilitate the ensuing development, in what follows, it is assumed that the matrix 𝚽 ∈ ℝM×N is
full rank, namely

𝔯𝔞𝔫𝔨(𝚽) = 𝔪𝔦𝔫(M, N) = N (15)

Copyright © 2017 SES ltd. All rights reserved. Page 28-7

PART IV: RESEARCH WORK IN PROGRESS

Put another way, the set of N ∈ ℕ column vectors of length M ∈ ℕ, M ≥ N that form the columns
of 𝚽 ∈ ℝM×N is linearly independent. Specifically, if one writes the matrix 𝚽 ∈ ℝM×N in terms of
its column vectors as 𝚽 = [𝝓1 𝝓2 ⋯ 𝝓N ] ∈ ℝM×N where 𝝓𝑖 ∈ ℝM×1 , ∀𝑖 ∈ ℤ[1,N] , then the linear
independence of the set of column vectors {𝝓𝑖 }N
𝑖=1 is written as

N
∑ 𝝓𝑛 𝑥𝑛 = 𝟎, iff 𝑥𝑛 = 0, ∀𝑛 ∈ ℤ[1,N] . (16)
𝑛=1

The statement in (16) is equivalent to stating that the null space 𝒩(𝚽) of the matrix 𝚽 ∈ ℝM×N
contains only the zero vector:

𝒩(𝚽) = {𝐱 ∈ ℝN×1 |𝚽𝐱 = 𝟎} = {𝟎}. (17)

Indeed, the three statements in (15), (16), and (17) are equivalent. Now, observe that each of the
M equations in (14) can written in expanded form by writing out the matrix-vector product,
namely
N
∑ Φ𝑚𝑛 𝑥𝑛 = 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M] . (18)
𝑛=1

In the above, 𝑚 ∈ ℤ[1,M] indexes the equation number and 𝑛 ∈ ℤ[1,N] indexes the degrees of
freedom (or unknowns). In the special case that M = N, the unique solution to the linear system
in(14) can be computed so long as the determinant of the square matrix 𝚽 ∈ ℝN×N is nonzero,
namely 𝔡𝔢𝔱(𝚽) ≠ 0. In general, however, the matrix 𝚽 ∈ ℝM×N is not square. Therefore, rather
than trying to find a solution to the overdetermined linear system in (14), which generally does
not exist, it is more practical to find a vector 𝐱 ∈ ℝN×1 that minimizes (half of) the sum of the
squares of the errors between 𝑖𝑚 , ∀𝑚 ∈ ℤ[1,M] (the measured currents) and 𝑖(𝑣𝑚 ), ∀𝑚 ∈ ℤ[1,M] (the
current-voltage model evaluated at the measurement voltages 𝑣𝑚 , ∀𝑚 ∈ ℤ[1,M] ). This approach
was first introduced by Carl Friedrich Gauss [13] and Adrien-Marie Legendre [14], in connection
to certain important astronomical problems. To do this, consider defining function 𝑒𝑚 (𝐱) as the
error between 𝑖𝑚 and 𝑖(𝑣𝑚 ), namely
N
𝑒𝑚 (𝐱) ≝ 𝑖𝑚 − 𝑖(𝑣𝑚 ) = 𝑖𝑚 − ∑ Φ𝑚𝑛 𝑥𝑛 , ∀𝑚 ∈ ℤ[1,M] . (19)
𝑛=1

Importantly, it is constructive to assemble the M errors 𝑒𝑚 (𝐱), ∀𝑚 ∈ ℤ[1,M] in the vector 𝐞(𝐱) ∈
ℝM×1, namely

𝐞(𝐱) ≝ 𝐢 − 𝚽𝐱 ∈ ℝM×1 . (20)

Here, the error vector 𝐞(𝐱): ℝN×1 → ℝM×1 is an affine function of the vector 𝐱 ∈ ℝN×1 ; an affine
function is not a linear function. Now, consider defining the multivariable scalar function
𝑓0 (𝐱): ℝN×1 → ℝ as (half of) the sum of the squares of the errors in (19), namely
M
1 1 1
𝑓0 (𝐱) ≝ 2‖𝐞(𝐱)‖22 = 2𝐞T (𝐱)𝐞(𝐱) = 2 ∑ |𝑒𝑚 (𝐱)|2 . (21)
𝑚=1

Hereafter, the function 𝑓0 (𝐱) is referred to as the objective function. In (21), the notation
‖𝐞(𝐱)‖22 = 𝐞T (𝐱)𝐞(𝐱) denotes the square of the 𝕝2 -norm of the error vector 𝐞(𝐱): ℝN×1 → ℝM×1 .
With the function 𝑓0 (𝐱): ℝN×1 → ℝ in place, defined (21) the minimization of (half of) the sum of

Page 28-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

the squares of the errors in (19) can be formally posed as the following unconstrained
minimization problem:

𝔪𝔦𝔫𝔦𝔪𝔦𝔷𝔢 𝑓0 (𝐱). (22)

𝐱∈ℝN×1

Using (20), it is clear that 𝐞T (𝐱)𝐞(𝐱) = [𝐢 − 𝚽𝐱]T [𝐢 − 𝚽𝐱] , in which case, one can write
1 1
𝑓0 (𝐱) = 2[𝐢T − 𝐱 T 𝚽T ][𝐢 − 𝚽𝐱] = 2(+𝐢T 𝐢 − 𝐢T 𝚽𝐱 − 𝐱 T 𝚽T 𝐢 + 𝐱 T 𝚽T 𝚽𝐱).

Since 𝐱 T 𝚽 T 𝐢 ∈ ℝ, it is clear that 𝐱 T 𝚽 T 𝐢 = [𝐱 T 𝚽T 𝐢]T = 𝐢T 𝚽𝐱. Therefore, the objective function 𝑓0 (𝐱)
is a multivariable quadratic function:
1 1
𝑓0 (𝐱) = +2𝐱 T 𝚽 T 𝚽𝐱 − 𝐢T 𝚽𝐱 + 2𝐢T 𝐢. (23)

Since it has been assumed that the tall and thin (M ≥ N) rectangular matrix 𝚽 ∈ ℝM×N is full rank
(see (15), (16), and (17)), the matrix 𝚽 T 𝚽 ∈ ℝN×N is positive definite, which can be ascertained
through its quadratic form:

𝐱 T 𝚽 T 𝚽𝐱 = [𝚽𝐱]T 𝚽𝐱 = ‖𝚽𝐱‖22 > 0, ∀𝐱 ∈ ℝN×1 \{𝟎}, (24)

namely, the quantity 𝐱 T 𝚽T 𝚽𝐱 is strictly positive for all nonzero vectors 𝐱 ∈ ℝN×1. Since the matrix
𝚽 T 𝚽 ∈ ℝN×N is symmetric and positive definite, namely 𝚽 T 𝚽 ∈ 𝕊N >0 , the quadratic objective
N×1
function 𝑓0 (𝐱): ℝ → ℝ in (23) is strictly convex. Therefore, there is one unique stationary point
that locates the global minimum of the quadratic function 𝑓0 (𝐱). Here, and in what follows, it is
convenient to denote the stationary point 𝐱 ⋆ of the quadratic function 𝑓0 (𝐱) in (23) by 𝐱𝔩𝔰 , since
the 𝐱 ⋆ is the optimal solution to the unconstrained least-squares (𝔩𝔰) minimization problem. The
gradient of the quadratic objective function 𝑓0 (𝐱), which is denoted by 𝜕𝐱 𝑓0 (𝐱) ∈ ℝN×1 , is given by:

1 1
𝜕𝐱 𝑓0 (𝐱) = 𝜕𝐱 (+2𝐱 T 𝚽 T 𝚽𝐱 − 𝐢T 𝚽𝐱 + 2𝐢T 𝐢) = +𝚽 T 𝚽𝐱 − 𝚽 T 𝐢. (25)

Now, the stationary point of 𝑓0 (𝐱) is obtained by setting its gradient to zero, namely 𝜕𝐱 𝑓0 (𝐱) = 𝟎 ∈
ℝN×1 , which corresponds to the first order necessary for optimality. In passing, the Hessian matrix
of the quadratic objective function in (23), which is denoted by 𝜕𝐱𝐱 T 𝑓0 (𝐱) ∈ ℝN×N, is given by

𝜕𝐱𝐱 T 𝑓0 (𝐱) = 𝚽 T 𝚽, (26)

and will be required in due course. Setting 𝜕𝐱 𝑓0 (𝐱) in (25) to the zero vector gives rise to the
normal equations whose solution 𝐱𝔩𝔰 ∈ ℝN×1 is the least-squares (approximate) solution to the
overdetermined system of equations in (18):

𝜕𝐱 𝑓0 (𝐱) = 𝟎 → 𝚽 T 𝚽𝐱 = 𝚽 T 𝐢|𝒩(𝚽T 𝚽)={𝟎} → 𝐱𝔩𝔰 = [𝚽

⏟ T 𝚽]−1 𝚽 T 𝐢. (27)
𝚽 † ∈ℝN×M

Specifically, the normal equations 𝚽 T 𝚽𝐱 = 𝚽 T 𝐢, that arise from setting the gradient of 𝑓0 (𝐱) =
0.5‖𝐢 − 𝚽𝐱‖22 to zero, are uniquely soluble so long as 𝒩(𝚽 T 𝚽) = {𝟎} (that is, 𝔡𝔢𝔱(𝚽 T 𝚽) ≠ 0), in
which case the matrix 𝚽 T 𝚽 is invertible. Here, the matrix 𝚽 † = [𝚽 T 𝚽]−1 𝚽 T ∈ ℝN×M is often
referred to the Moore-Penrose pseudo-inverse of the rectangular matrix 𝚽 ∈ ℝM×N . In the special
case that 𝚽 ∈ ℝM×N is a square matrix (M = N), with nonzero determinant (or equivalently
𝒩(𝚽) = {𝟎}), the Moore-Penrose pseudo-inverse becomes the standard matrix inverse, namely

Copyright © 2017 SES ltd. All rights reserved. Page 28-9

PART IV: RESEARCH WORK IN PROGRESS

𝚽 † = 𝚽 −1. Although 𝐱𝔩𝔰 ∈ ℝN×1 is the global minimizer to the unconstrained least-squares (𝔩𝔰)
problem, the error vector 𝐞(𝐱) ∈ ℝM×1 in (20), evaluated at 𝐱 = 𝐱𝔩𝔰 ∈ ℝN×1, is, in general, not zero.
To see this, one can insert the expression for 𝐱𝔩𝔰 ∈ ℝN×1 delineated in (27) into (20), which results
in

𝐞(𝐱𝔩𝔰 ) = 𝐢 − 𝚽𝐱𝔩𝔰 = 𝐢 − 𝚽[𝚽 T 𝚽]−1 𝚽 T 𝐢 = (𝐈M − 𝚽[𝚽 T 𝚽]−1 𝚽T )𝐢, (28)

where 𝐈M ∈ ℝM×M is the M × M identity matrix. In summary, the least-squares (approximate)

solution to the overdetermined system in (14) is given by

𝐱𝔩𝔰 = 𝚽 † 𝐢 = [𝚽 T 𝚽]−1 𝚽T 𝐢. (29)

The present version of SESCurveFit contains an implementation of the solution to the

unconstrained, linear least-squares polynomial curve fitting problem detailed in this section,
which was first prototyped in MATLAB, then implemented in FORTRAN 90, ab initio, by the first
author, then integrated into SES Software under the direction of the second and third authors.

4 Derivative Constrained Least-Squares Polarization Curve

Fitting

By means of rigorous computational experiments with industrial corrosion polarization data

[15](p.190), the authors noticed that the use of higher order polynomials led to highly oscillatory
fits. It was observed that when high order polynomial models were employed, the resulting fitted
curve tended to sharply swing between the measurement data, which is often referred to as
overfitting. These oscillations are not desired. By examining this oscillatory behaviour, the first
author realized that the problem was not with the choice of polynomials used in the expansion of
the current-voltage characteristic, but rather, the desired behavior of the curve fit (that is, the
properties that the fitted curve should satisfy). It is for this reason that computational scientists
often (erroneously) conclude that high order polynomials are ill-suited for curve fitting
applications. The crux of the matter is that if, in addition to minimizing 𝑓0 (𝐱) in (21), one could
constrain the behavior of the first order derivative of the resulting polynomial fit, then unwanted
oscillations could be eliminated altogether. Therefore, by forcing the first order derivative of 𝑖(𝑣)
to be strictly greater than 0, namely 𝑖 ′ (𝑣) = 𝒢(𝑣) > 0, the resulting curve fit will be strictly
increasing (that is, strictly monotonically increasing). This condition on the differential
conductance 𝒢(𝑣) reads as
N
𝑖 ′ (𝑣) = 𝒢(𝑣) = ∑ 𝑥𝑛 𝜙𝑛′ (𝑣) > 0. (30)
𝑛=1

Indeed, a function that does not decrease does not oscillate. The remainder of this article explains
the exact way one can minimize 𝑓0 (𝐱) subject to the monotonicity condition in (30). In this work,
it is more convenient to generalize (30) as follows

𝒢lower (𝑣) ≤ 𝒢(𝑣) ≤ 𝒢upper (𝑣), ∀𝑣 ∈ ℝ[𝑎,𝑏] , (31)

where 𝒢lower (𝑣): ℝ → ℝ and 𝒢upper (𝑣): ℝ → ℝ are the lower and upper bounds on the differential
conductance 𝒢(𝑣), respectively, and it is assumed that 𝒢upper (𝑣) > 𝒢lower (𝑣), ∀𝑣 ∈ ℝ[𝑎,𝑏] .
Moreover, 𝑣 ∈ ℝ[𝑎,𝑏] is the finite interval over which the inequality constraints are to be satisfied.
Page 28-10 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

Importantly, the inequality constraints in (31) are often referred to as semi-infinite inequalities
[16](p.216) (since one must find the finite number of coefficients 𝑥𝑛 , ∀𝑛 ∈ ℤ[1,N] in such a way that
the differential conductance 𝒢(𝑣) = ∑N ′
𝑛=1 𝑥𝑛 𝜙𝑛 (𝑣) satisfies (31) at an infinite number of points on
the interval 𝑣 ∈ ℝ[𝑎,𝑏] ). Note that (30) can be recovered from (31) by setting 𝒢lower (𝑣) = ϵ and
𝒢upper (𝑣) = +∞, where ϵ is some positive factor of machine precision. Note that since one can
write 𝒢lower (𝑣) ≤ 𝒢(𝑣) as −𝒢lower (𝑣) ≥ −𝒢(𝑣), one can write equivalently write (31) as the
following system of semi-infinite inequality constraints

−𝒢(𝑣) ≤ −𝒢lower (𝑣), ∀𝑣 ∈ ℝ[𝑎,𝑏] ,

(32)
+𝒢(𝑣) ≤ +𝒢upper (𝑣), ∀𝑣 ∈ ℝ[𝑎,𝑏] .
In this work, both semi-infinite inequality constraints in (32) are approximated by sampling each
constraint at a finite number of points on the interval 𝑣 ∈ ℝ[𝑎,𝑏]. By sampling the first line in (32)
at the points 𝑣𝑖 , ∀𝑖 ∈ ℤ[1,Klower ] and sampling the second line in (32) at the points 𝑣𝑘 , ∀𝑘 ∈
ℤ[1,Kupper ] (where K lower and K upper corresponds to the number samples associated the lower and
upper bounds on the differential conductance, respectively) the following result is obtained
N
−𝒢(𝑣𝑖 ) = − ∑ 𝑥𝑛 𝜙𝑛′ (𝑣𝑖 ) ≤ −𝒢lower (𝑣𝑖 ), ∀𝑖 ∈ ℤ[1,Klower ] ,
𝑛=1
N (33)
+𝒢(𝑣𝑘 ) = + ∑ 𝑥𝑛 𝜙𝑛′ (𝑣𝑘 ) ≤ +𝒢upper (𝑣𝑘 ), ∀𝑘 ∈ ℤ[1,Kupper ] .
𝑛=1

Note that in this formulation, the samples associated with the lower and upper bounds need not
coincide. In this work, however, the points 𝑣𝑖 , ∀𝑖 ∈ ℤ[1,Klower ] are chosen to coincide with 𝑣𝑘 , ∀𝑘 ∈
ℤ[1,Kupper ] , with K lower = K upper = 20N. Now, in matrix-vector form, the system of K lower + K upper
inequalities in (33) can be written compactly as the following system of linear inequality
constraints:

𝐀𝐱 ≼ 𝐛, (34)

where 𝐀 ∈ ℝ[Klower +Kupper ]×N is a rectangular matrix and 𝐛 ∈ ℝ[Klower +Kupper ]×1 is a vector, each of
which are partitioned as follows:
−𝐂lower − 𝐠 lower
𝐀 ≝ [+𝐂 ] ∈ ℝ[Klower +Kupper ]×N , 𝐛 ≝ [ +𝐠 ] ∈ ℝ[Klower +Kupper ]×1 , (35)
upper upper

where the rectangular matrices 𝐂lower ∈ ℝKlower ×N and 𝐂upper ∈ ℝKupper ×N house the samples of
the first order dervatives of the expansion functions 𝜙𝑛 (𝑣), which are computed using the
recurrence relationship in (11):
[𝐂lower ]𝑖 ,𝑛 = 𝜙𝑛′ (𝑣𝑖 ), ∀(𝑖 , 𝑛) ∈ ℤ[1,Klower ] × ℤ[1,N] ,
[𝐂upper ]𝑘,𝑛 = 𝜙𝑛′ (𝑣𝑘 ), ∀(𝑘, 𝑛) ∈ ℤ[1,Kupper ] × ℤ[1,N] .

Moreover, the vectors 𝐠 lower ∈ ℝKlower ×1 and 𝐠 upper ∈ ℝKupper ×1 contain the samples of the
functions 𝒢lower (𝑣) and 𝒢upper (𝑣), respectively:
[𝐠 lower ]𝑖 = 𝒢lower (𝑣𝑖 ), ∀𝑖 ∈ ℤ[1,Klower ] ,
[𝐠 upper ]𝑘 = 𝒢upper (𝑣𝑘 ), ∀𝑘 ∈ ℤ[1,Kupper ] .

Copyright © 2017 SES ltd. All rights reserved. Page 28-11

PART IV: RESEARCH WORK IN PROGRESS

For convenience, I now define K in ≝ K lower + K upper as the total number of inequality constraints
so that 𝐀 ∈ ℝKin ×N and 𝐛 ∈ ℝKin ×1 . This way, the matrix the rectangular matrix 𝐀 ∈ ℝKin ×N in (34)
can be expressed in terms of its row vectors 𝐚T𝑘 ∈ ℝ1×N , ∀𝑘 ∈ ℤ[1,Kin ] :
𝐚1T
𝐚T2 T
𝐀= = [𝐚1 𝐚2 ⋯ 𝐚Kin ] ∈ ℝKin ×N .
⋮
T
𝐚
[ Kin ]
With this form of the matrix 𝐀 ∈ ℝKin ×N, each of the K in inequality constraints that form 𝐀𝐱 ≼ 𝐛
in (34), can be written succinctly as 𝐚T𝑘 𝐱 ≤ 𝑏𝑘 , ∀𝑘 ∈ ℤ[1,Kin ] or equivalently, by subtracting 𝑏𝑘 from
both sides, one can write

𝐚T𝑘 𝐱 − 𝑏𝑘 ≤ 0, ∀𝑘 ∈ ℤ[1,Kin ] . (36)

Based on this, one can now define the functions 𝑓𝑘 (𝐱): ℝN×1 → ℝ , ∀𝑘 ∈ ℤ[1,Kin ] as the affine scalar
functions appearing on the left-hand side of the inequality in (36), namely

𝑓𝑘 (𝐱) ≝ 𝐚T𝑘 𝐱 − 𝑏𝑘 ∈ ℝ , ∀𝑘 ∈ ℤ[1,Kin ] . (37)

In passing, the gradient vectors of the affine constraint functions defined in (37) are given by

𝜕𝐱 𝑓𝑘 (𝐱) = 𝐚𝑘 ∈ ℝN×1 , ∀𝑘 ∈ ℤ[1,Kin ] . (38)

Moreover, since the constraint functions in (37) are affine, the Hessian matrices of the constraint
functions vanish, namely

𝜕𝐱𝐱 T 𝑓𝑘 (𝐱) = 𝟎 ∈ ℝN×N , ∀𝑘 ∈ ℤ[1,Kin ] . (39)

Now, with the inequality constraints in place, the minimization of the of 𝑓0 (𝐱) subject to the
condition that the constraints in (36) must be satisfied can be formally posed as follows
𝔪𝔦𝔫𝔦𝔪𝔦𝔷𝔢 𝑓0 (𝐱)
𝐱∈ℝN×1 (40)
𝔰𝔲𝔟𝔧𝔢𝔠𝔱 𝔱𝔬 𝑓𝑘 (𝐱) ≤ 0, ∀𝑘 ∈ ℤ[1,Kin ] ,
which corresponds to an inequality constrained minimization problem. Since the objective
function 𝑓0 (𝐱) delineated in (23) is a quadratic function and the constraint functions 𝑓𝑘 (𝐱), ∀𝑘 ∈
ℤ[1,Kin ] given in (37) are affine scalar functions, the optimization problem in (40) is often referred
to as a quadratic program (QP); which makes (22) an unconstrained quadratic program. In
passing, in the optimization literature the words programming and optimization are usually used
interchangeably [17](p.526). Now, with the objective function 𝑓0 (𝐱) and the constraint functions
𝑓𝑘 (𝐱), ∀𝑘 ∈ ℤ[1,Kin ], consider the formation of the Lagrangian function ℒ(𝐱, 𝛌): ℝN×1 × ℝKin ×1 → ℝ,
which is defined as
Kin
ℒ(𝐱, 𝛌) ≝ 𝑓0 (𝐱) + ∑ λ𝑘 𝑓𝑘 (𝐱), (41)
𝑘=1

Where the variables λ𝑘 ∈ ℝ, ∀𝑘 ∈ ℤ[1,Kin ] are the Lagrange multipliers associated with the
inequality constraints 𝑓𝑘 (𝐱) ≤ 0, ∀𝑘 ∈ ℤ[1,Kin ] . For the sake of computational organization, the

Page 28-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Lagrange multipliers are stored in the vector 𝛌 ∈ ℝKin ×1 and the constraint functions are stored
in the vector function 𝐟(𝐱): ℝN×1 → ℝKin ×1 , namely
λ1 𝑓1 (𝐱) 𝐚1T 𝐱 − 𝑏1
λ2 𝑓2 (𝐱) 𝐚T2 𝐱 − 𝑏2
𝛌≝ ∈ ℝKin ×1 , 𝐟(𝐱) ≝ = = 𝐀𝐱 − 𝐛 ∈ ℝKin ×1 .
⋮ ⋮ ⋮
T
[λKin ] [𝑓Kin (𝐱)] [𝐚Kin 𝐱 − 𝑏Kin ]
This way, the Lagrangian function in (41) can be written as ℒ(𝐱, 𝛌) = 𝑓0 (𝐱) + 𝛌T 𝐟(𝐱). Importantly,
the function −𝛌T 𝐟(𝐱) is often referred to as the surrogate duality gap [18](p.612), and denoted
by 𝜂̂ (𝐱, 𝛌): ℝN×1 × ℝKin ×1 → ℝ, namely
Kin
𝜂̂ (𝐱, 𝛌) ≝ −𝛌T 𝐟(𝐱) = − ∑ λ𝑘 𝑓𝑘 (𝐱). (42)
𝑘=1

Importantly, the partial gradient of the Lagrangian function ℒ(𝐱, 𝛌) in (41), with respect to the
variables 𝐱 ∈ ℝN×1 , is given by
Kin
𝜕𝐱 ℒ(𝐱, 𝛌) = 𝜕𝐱 𝑓0 (𝐱) + ∑ λ𝑘 𝜕𝐱 𝑓𝑘 (𝐱) (43)
𝑘=1

Now consider the Karush Kuhn Tucker (KKT) conditions, stated here without proof, that are given
by [19], [20], [18](p.243)

𝑓𝑘 (𝐱 ⋆ ) ≤ 0, ∀𝑘 ∈ ℤ[1,Kin ] ,
λ⋆𝑘 ≥ 0, ∀𝑘 ∈ ℤ[1,Kin ] ,
(44)
λ⋆𝑘 𝑓𝑘 (𝐱 ⋆ ) = 0, ∀𝑘 ∈ ℤ[1,Kin ] ,
𝜕𝐱 ℒ(𝐱 ⋆ , 𝛌⋆ ) = 𝟎 ∈ ℝN×1
The first line in (44) is referred to as primal feasibility, which is the requirement that the optimal
solution 𝐱 ⋆ ∈ ℝN×1 must satisfy the inequality constraints. The second line in (44) is referred to
as dual feasibility, which requires the optimal values of the Lagrange multipliers be nonnegative.
The third line in (44) is referred to as complementary slackness. Finally, the fourth line in (44) is
the stationarity condition which, in expanded form, reads as
Kin
𝜕𝐱 𝑓0 (𝐱 ⋆ ) + ∑ λ⋆𝑘 𝜕𝐱 𝑓𝑘 (𝐱 ⋆ ) = 𝟎. (45)
𝑘=1

Now, by evaluating ℒ(𝐱, 𝛌) at 𝐱 = 𝐱 ⋆ and 𝛌 = 𝛌⋆ , whilst imposing the complementary slackness

condition (that is, the third line in (44)), the optimal value of the Lagrangian function in (41) is
immediately obtained:
Kin
ℒ(𝐱 ⋆ , 𝛌⋆ ) = 𝑓0 (𝐱 ⋆ ) + ∑ λ⋆𝑘 𝑓𝑘 (𝐱⋆ ) = 𝑓0 (𝐱⋆ ),
⏟ (46)
𝑘=1
0

One approach to solving (40) is to restrict the solution search to the region that satisfies the
inequality constraints. To do this, consider the following h-interior function, which is denoted by
𝜙ℎ (𝐱) and defined as [18](p.624), [17](p.541):
Kin
𝜙ℎ (𝐱) ≝ ∑ ℎ(𝑓𝑘 (𝐱)), (47)
𝑘=1

Copyright © 2017 SES ltd. All rights reserved. Page 28-13

PART IV: RESEARCH WORK IN PROGRESS

valid for all points 𝐱 ∈ ℝN×1 that satisfy the inequality constraints 𝑓𝑘 (𝐱) ≤ 0, ∀𝑘 ∈ ℤ[1,Kin ] . Here,
the gradient of the h-interior function is written as
Kin
𝜕𝐱 𝜙ℎ (𝐱) = ∑ 𝜕𝐱 ℎ(𝑓𝑘 (𝐱)). (48)
𝑘=1

Now, in (48), the entries of the gradient vector 𝜕𝐱 𝜙ℎ (𝐱) are given by
𝜕𝜙ℎ (𝐱) Kin 𝜕ℎ(𝑓 (𝐱))
𝑘
[𝜕𝐱 𝜙ℎ (𝐱)]𝑚 = =∑ , ∀𝑚 ∈ ℤ[1,N]
𝜕𝑥𝑚 𝑘=1 𝜕𝑥𝑚
The partial derivatives 𝜕𝑥𝑚 ℎ(𝑓𝑘 (𝐱)) appearing in each term in the above summation are evaluated
by means of the chain rule [7](p.198), namely

𝜕ℎ(𝑓𝑘 (𝐱)) dℎ(𝑓𝑘 (𝐱)) 𝜕𝑓𝑘 (𝐱)

= . (49)
𝜕𝑥𝑚 d𝑓𝑘 (𝐱) 𝜕𝑥𝑚
It is now convenient to adopt the following shorthand notation for the first order derivative of the
function ℎ(𝑓𝑘 (𝐱)) with respect to its argument (that is, the constraint function 𝑓𝑘 (𝐱)):

dℎ(𝜉) dℎ(𝑓𝑘 (𝐱))

ℎ′ (𝑓𝑘 (𝐱)) ≝ | = . (50)
d𝜉 𝜉=𝑓 d𝑓𝑘 (𝐱)
𝑘 (𝐱)

This way, the gradient of ℎ(𝑓𝑘 (𝐱)) can be written succinctly as

𝜕ℎ(𝑓𝑘 (𝐱)) dℎ(𝑓𝑘 (𝐱)) 𝜕𝑓𝑘 (𝐱)
𝜕𝐱 ℎ(𝑓𝑘 (𝐱)) = = = ℎ′ (𝑓𝑘 (𝐱))𝜕𝐱 𝑓𝑘 (𝐱),
𝜕𝐱 d𝑓𝑘 (𝐱) 𝜕𝐱
where 𝜕𝐱 𝑓𝑘 (𝐱) is the gradient of the 𝑘 th constraint function. Therefore, the expression for the
gradient of the h-interior function in (48) becomes
Kin
𝜕𝐱 𝜙ℎ (𝐱) = ∑ ℎ′ (𝑓𝑘 (𝐱))𝜕𝐱 𝑓𝑘 (𝐱). (51)
𝑘=1

The Hessian matrix of the h-interior function, that I denote by is 𝜕𝐱𝐱 T 𝜙ℎ (𝐱) defined as the gradient
of the transpose of the gradient of 𝜙ℎ (𝐱), namely
Min T Min
𝜕𝐱𝐱 T 𝜙ℎ (𝐱) ≝ 𝜕𝐱 [𝜕𝐱 𝜙ℎ (𝐱)]T = ∑ 𝜕𝐱 [𝜕𝐱 ℎ(𝑓𝑘 (𝐱))] = ∑ 𝜕𝐱𝐱 T ℎ(𝑓𝑘 (𝐱)). (52)
𝑘=1 𝑘=1

Now, the entries of the Hessian matrix 𝜕𝐱𝐱 T 𝜙ℎ (𝐱) are given by
Min
[𝜕𝐱𝐱 T 𝜙ℎ (𝐱)]𝑚𝑛 = ∑ [𝜕𝐱𝐱 T ℎ(𝑓𝑘 (𝐱))]𝑚𝑛 , ∀(𝑚, 𝑛) ∈ ℤ2[1,N] ,
𝑘=1

where [𝜕𝐱𝐱 T ℎ(𝑓𝑘 (𝐱))]𝑚𝑛 are the entries of the Hessian matrices of the functions ℎ(𝑓𝑘 (𝐱)), namely

𝜕 2 ℎ(𝑓𝑘 (𝐱)) 𝜕 𝜕ℎ(𝑓𝑘 (𝐱))

[𝜕𝐱𝐱 T ℎ(𝑓𝑘 (𝐱))]𝑚𝑛 = = ( ).
𝜕𝑥𝑚 𝜕𝑥𝑛 𝜕𝑥𝑚 𝜕𝑥𝑛
By inserting the relationship in (49) into the above expression, followed by the application of the
product rule and the chain rule, one can write
𝜕 𝜕ℎ(𝑓𝑘 (𝐱)) 𝜕 dℎ(𝑓𝑘 (𝐱)) 𝜕𝑓𝑘 (𝐱)
( )= ( )
𝜕𝑥𝑛 𝜕𝑥𝑛 𝜕𝑥𝑚 d𝑓𝑘 (𝐱) 𝜕𝑥𝑛
Page 28-14 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

dℎ(𝑓𝑘 (𝐱)) 𝜕 2 𝑓𝑘 (𝐱) 𝜕 dℎ(𝑓𝑘 (𝐱)) 𝜕𝑓𝑘 (𝐱)

= [ ]+[ ( )]
d𝑓𝑘 (𝐱) 𝜕𝑥𝑚 𝜕𝑥𝑛 𝜕𝑥𝑚 d𝑓𝑘 (𝐱) 𝜕𝑥𝑛
(𝐱)) 2 (𝐱) 2 (𝐱)) (𝐱)
dℎ(𝑓𝑘 𝜕 𝑓𝑘 d ℎ(𝑓𝑘 𝜕𝑓𝑘 𝜕𝑓𝑘 (𝐱)
= [ ]+[ 2 ] .
d𝑓𝑘 (𝐱) 𝜕𝑥𝑚 𝜕𝑥𝑛 d 𝑓𝑘 (𝐱) 𝜕𝑥𝑚 𝜕𝑥𝑛
One can therefore write
𝜕 2 ℎ(𝑓𝑘 (𝐱)) dℎ(𝑓𝑘 (𝐱)) 𝜕 2 𝑓𝑘 (𝐱) d2 ℎ(𝑓𝑘 (𝐱)) 𝜕𝑓𝑘 (𝐱) 𝜕𝑓𝑘 (𝐱)
= ( )+ ( ).
𝜕𝑥𝑚 𝜕𝑥𝑛 d𝑓𝑘 (𝐱) 𝜕𝑥𝑚 𝜕𝑥𝑛 d2 𝑓𝑘 (𝐱) 𝜕𝑥𝑚 𝜕𝑥𝑛
Therefore, in addition to the definition of ℎ′ (𝑓𝑘 (𝐱)) given in (50), It is now convenient to adopt
the following shorthand notation for the second order derivative of the function ℎ(𝑓𝑘 (𝐱)) with
respect to its argument as follows

d2 ℎ(𝜉) d2 ℎ(𝑓𝑘 (𝐱))

ℎ′′ (𝑓𝑘 (𝐱)) ≝ | = . (53)
d𝜉 2 𝜉=𝑓 d2 𝑓𝑘 (𝐱)
𝑘 (𝐱)

Therefore, the expression for the Hessian of the function ℎ(𝑓𝑘 (𝐱)) becomes
𝜕𝐱𝐱 T ℎ(𝑓𝑘 (𝐱)) = ℎ′ (𝑓𝑘 (𝐱))𝜕𝐱𝐱 T 𝑓𝑘 (𝐱) + ℎ′′ (𝑓𝑘 (𝐱))[𝜕𝐱 𝑓𝑘 (𝐱)][𝜕𝐱 𝑓𝑘 (𝐱)]T ,
where 𝜕𝐱𝐱 T 𝑓𝑘 (𝐱) ∈ ℝN×N is the Hessian matrix of the constraint function 𝑓𝑘 (𝐱) and
[𝜕𝐱 𝑓𝑘 (𝐱)][𝜕𝐱 𝑓𝑘 (𝐱)]T ∈ ℝN×N is the outer product of the gradient 𝜕𝐱 𝑓𝑘 (𝐱) ∈ ℝN×1 of the constraint
function 𝑓𝑘 (𝐱). Therefore, by inserting the above expression into (52), the Hessian matrix of the
h-interior function is obtained
Min
𝜕𝐱𝐱 T 𝜙ℎ (𝐱) = ∑ {ℎ′ (𝑓𝑘 (𝐱))𝜕𝐱𝐱 T 𝑓𝑘 (𝐱) + ℎ′′ (𝑓𝑘 (𝐱))[𝜕𝐱 𝑓𝑘 (𝐱)][𝜕𝐱 𝑓𝑘 (𝐱)]T }. (54)
𝑘=1

Now, when the function ℎ(𝜉) in (47) is chosen to be ℎ(𝜉) = − ln(−𝜉), where ln(⋅) is the natural
logarithm, the interior function 𝜙log (𝐱) in (47) is said to be the logarithmic interior function,
which is denoted by 𝜙log (𝐱), namely
Kin Kin
𝜙log (𝐱) ≝ + ∑ (− ln(−𝑓𝑘 (𝐱))) = − ∑ ln(𝑏𝑘 − 𝐚T𝑘 𝐱). (55)
𝑘=1 𝑘=1

Figure 2 demonstrates the behavior the logarithmic barrier function for a system of K in = 5
inequality constraints in N = 2 dimensions.

Copyright © 2017 SES ltd. All rights reserved. Page 28-15

PART IV: RESEARCH WORK IN PROGRESS

0.6
Affine Constraint Functions:
0.5

0.4

0.3

0.2 5
𝜙log 𝐱 = − ∑ ln −𝑓𝑘 𝐱
𝑥2

𝑘=1
0.1 logarithmic interior function Linear Inequality Constraints:

-0.1

-0.2

-0.3

-0.6 -0.4 -0.2 0 0.2 0.4 0.6

𝑥1
K
Figure 2: Demonstration of the structure of the logarithmic interior function 𝜙log (𝐱) = − ∑𝑘=1 in
ln(−𝑓𝑘 (𝐱))
T
for the K in = 5 inequality constraints 𝑓𝑘 (𝐱) ≤ 0, ∀𝑘 ∈ ℤ[1,5] where 𝐱 = [𝑥1 𝑥2 ] ∈ ℝ 2×1
and the K in = 5
constraint functions 𝑓𝑘 (𝐱), ∀𝑘 ∈ ℤ[1,5] are given by 𝑓1 (𝐱) = 𝐚1T 𝐱 − 𝑏1 = +2𝑥1 + 𝑥2 − 1, 𝑓2 (𝐱) = 𝐚T2 𝐱 − 𝑏2 =
+2𝑥1 − 𝑥2 − 1, 𝑓3 (𝐱) = 𝐚T3 𝐱 − 𝑏3 = +2𝑥1 + 2𝑥2 − 1, 𝑓4 (𝐱) = 𝐚T4 𝐱 − 𝑏4 = −𝑥1 − 2𝑥2 − 0.25, and 𝑓5 (𝐱) = 𝐚T5 𝐱 −
𝑏5 = −𝑥1 − 0.55.

Here, observe that the first and second order derivatives of the function ℎ(𝜉) = − ln(−𝜉) are given
by
dℎ(𝜉) −1 1 1 d2 ℎ(𝜉) d 1 1
ℎ′ (𝜉) = = −( ) = − = , ℎ′′ (𝜉) = = (−𝜉 −1 )
= + = .
d𝜉 −𝜉 𝜉 (−𝜉) d2 𝜉 d𝜉 𝜉 2 (−𝜉)2
Therefore, by using ℎ′ (𝜉) = 1/(−𝜉) in (51), with 𝜉 = 𝑓𝑘 (𝐱), whilst recalling that 𝑓𝑘 (𝐱) = 𝐚T𝑘 𝐱 − 𝑏𝑘
and 𝜕𝐱 𝑓𝑘 (𝐱) = 𝐚𝑘 , as given in (37) and (38), respectively, one can ascertain that the expression for
the gradient of the logarithmic interior function is given by
Kin 1 Kin 𝐚𝑘
𝜕𝐱 𝜙log (𝐱) = + ∑ ( ) 𝜕𝐱 𝑓𝑘 (𝐱) = + ∑ T
. (56)
𝑘=1 −𝑓𝑘 (𝐱) 𝑘=1 𝑏𝑘 − 𝐚𝑘 𝐱

Moreover, by using ℎ′′ (𝜉) = 1/(−𝜉)2 in (54), with 𝜉 = 𝑓𝑘 (𝐱), whilst recalling that 𝑓𝑘 (𝐱) = 𝐚T𝑘 𝐱 − 𝑏𝑘
and 𝜕𝐱 𝑓𝑘 (𝐱) = 𝐚𝑘 and 𝜕𝐱𝐱 T 𝑓𝑘 (𝐱) = 𝟎, as given in (37), (38) and (39), respectively, one obtains the
expression for the Hessian of the logarithmic interior function:
Kin [𝜕𝐱 𝑓𝑘 (𝐱)][𝜕𝐱 𝑓𝑘 (𝐱)]T Kin 𝐚𝑘 𝐚T𝑘
𝜕𝐱𝐱 T 𝜙log (𝐱) = + ∑ = + ∑ . (57)
𝑘=1 (−𝑓𝑘 (𝐱))2 T 2
𝑘=1 (𝑏𝑘 − 𝐚𝑘 𝐱)

Now, consider adding the function (1/𝑡)𝜙log (𝐱) to the objective function 𝑓0 (𝐱), where 𝑡 > 0 is a
strictly positive parameter. This results in what is known as the logarithmic barrier function
[17](p.541), [21](p.44), which is denoted here by 𝑓log (𝐱, 𝑡):

Page 28-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

𝜙log (𝐱) 1 Kin

𝑓log (𝐱, 𝑡) ≝ 𝑓0 (𝐱) + = 𝑓0 (𝐱) + ∑ (− ln(−𝑓𝑘 (𝐱))) , 𝑡 > 0. (58)
𝑡 𝑡 𝑘=1

Therefore, given a point 𝐱 ∈ ℝN×1 that satisfies the inequality constraints 𝐟(𝐱) = 𝐀𝐱 − 𝐛 ≼ 𝟎, the
minimization of the logarithmic barrier function 𝑓log (𝐱, 𝑡) in (58), for 𝑡 > 0, approximately solves
the inequality constrained minimization problem defined in (40). However, convergence to the
optimal solution is sensitive to the choice of the parameter 𝑡 > 0. Now, using (56), observe that
the gradient of the function 𝑓log (𝐱, 𝑡), that is denoted by 𝐠 log (𝐱, 𝑡), is given by
Kin 1
𝐠 log (𝐱, 𝑡) ≝ 𝜕𝐱 𝑓log (𝐱, 𝑡) = 𝜕𝐱 𝑓0 (𝐱) + ∑ ( ) 𝜕 𝑓 (𝐱). (59)
𝑘=1 −𝑡𝑓𝑘 (𝐱) 𝐱 𝑘
Moreover, using (57), the Hessian of the function 𝑓log (𝐱, 𝑡), that is denoted by 𝐇log (𝐱, 𝑡), is given
by

1 Kin [𝜕𝐱 𝑓𝑘 (𝐱)][𝜕𝐱 𝑓𝑘 (𝐱)]T

𝐇log (𝐱, 𝑡) ≝ 𝜕𝐱𝐱 T 𝑓log (𝐱, 𝑡) = 𝜕𝐱𝐱 T 𝑓0 (𝐱) + ∑ . (60)
𝑡 𝑘=1 (−𝑓𝑘 (𝐱))2
Note that from the symmetry of mixed second order partial derivatives, the Hessian matrix in (60)
T
is symmetric, namely 𝐇log (𝐱, 𝑡) = 𝐇log (𝐱, 𝑡). Now, the comparison of 𝐠 log (𝐱, 𝑡) in (59) to the partial
gradient of the Lagrangian function in (43) suggests the form of the Lagrange multipliers
associated with the inequality constraints:
1
λ𝑘 = .
−𝑡𝑓𝑘 (𝐱)
Therefore, let 𝐱 ⋆ (𝑡): ℝ(0,∞) → ℝN×1 be defined as the optimal point of the logarithmic barrier
function 𝑓log (𝐱, 𝑡) in (58), parametrized with respect to the nonnegative parameter 𝑡 > 0, namely
𝐱 ⋆ (𝑡) ≝ 𝔞𝔯𝔤𝔪𝔦𝔫 𝑓log (𝐱, 𝑡),
𝐱∈ℝN×1

where 𝔞𝔯𝔤𝔪𝔦𝔫 is shorthand for the argument of the minimum. Therefore, by setting
𝐠 log (𝐱 ⋆ (𝑡), 𝑡) = 𝟎, the stationarity condition of the logarithmic barrier function is obtained
Kin 1
𝜕𝐱 𝑓0 (𝐱 ⋆ (𝑡)) + ∑ ( ) 𝜕𝐱 𝑓𝑘 (𝐱 ⋆ (𝑡)) = 𝟎,
𝑘=1 −𝑡𝑓𝑘 (𝐱 ⋆ (𝑡))
Therefore, by comparing the above expression with the Lagrangian stationarity condition in (45),
consider letting λ⋆𝑘 (𝑡) be equal to the bracketed fraction appearing in each term of the summation
in the above expression:
1 1
λ⋆𝑘 (𝑡) = ⟶ λ ⋆ (𝑡)𝑓
𝑘 𝑘 (𝐱 ⋆ (𝑡))
= − . (61)
−𝑡𝑓𝑘 (𝐱 ⋆ (𝑡)) 𝑡

Here, observe that the products λ⋆𝑘 (𝑡)𝑓𝑘 (𝐱 ⋆ (𝑡)), ∀𝑘 ∈ ℤ[1,Kin ] violate the complementary slackness
conditions in (44). However, notice that as 𝑡 → +∞ in (61), complementary slackness is achieved
in the limit

lim λ⋆𝑘 (𝑡)𝑓𝑘 (𝐱 ⋆ (𝑡)) = 0. (62)

𝑡→+∞

Therefore, consider letting 𝐱 ⋆ ∈ ℝN×1 and 𝛌⋆ ∈ ℝKin ×1 be equal to the limits of 𝐱 ⋆ (𝑡) ∈ ℝN×1 and
𝛌⋆ (𝑡) ∈ ℝKin ×1

Copyright © 2017 SES ltd. All rights reserved. Page 28-17

PART IV: RESEARCH WORK IN PROGRESS

𝐱 ⋆ = lim 𝐱 ⋆ (𝑡), 𝛌⋆ = lim 𝛌⋆ (𝑡)

𝑡→+∞ 𝑡→+∞

Moreover, by evaluating the Lagrangian function ℒ(𝐱, 𝛌) in (41) at 𝐱 = 𝐱 ⋆ (𝑡) and 𝛌 = 𝛌⋆ (𝑡), with
the entries λ⋆𝑘 (𝑡) of the vector 𝛌⋆ (𝑡) ∈ ℝKin ×1 given in (61), one finds that the value of the
Lagrangian function ℒ(𝐱 ⋆ (𝑡), 𝛌⋆ (𝑡)) is K in /𝑡 sub-optimal, namely
Kin K in
ℒ(𝐱 ⋆ (𝑡), 𝛌⋆ (𝑡)) = 𝑓0 (𝐱 ⋆ (𝑡)) + ∑ λ⋆𝑘 (𝑡)𝑓𝑘 (𝐱⋆ (𝑡)) = 𝑓0 (𝐱 ⋆ (𝑡)) −
⏟
𝑘=1 𝑡 (63)
1
𝑡

In passing, the quantity −(K in /𝑡) in (63) is equal the surrogate duality gap defined in (42),
evaluated at 𝐱 = 𝐱 ⋆ (𝑡) and 𝛌 = 𝛌⋆ (𝑡), namely
Kin K in
𝜂̂ (𝐱 ⋆ (𝑡), 𝛌⋆ (𝑡)) = −[𝛌⋆ (𝑡)]T 𝐟(𝐱 ⋆ (𝑡)) = − ∑ λ⋆𝑘 (𝑡)𝑓𝑘 (𝐱 ⋆ (𝑡)) = −
⏟
𝑘=1 𝑡
1
𝑡

As one can see from the expression in (63), as 𝑡 → +∞, the quantity (K in /𝑡) approaches zero, and
the Lagrangian function ℒ(𝐱 ⋆ (𝑡), 𝛌⋆ (𝑡)) approaches the optimal value of Lagrangian function
given in (46):
ℒ(𝐱 ⋆ , 𝛌⋆ ) = lim ℒ(𝐱 ⋆ (𝑡), 𝛌⋆ (𝑡)) = lim 𝑓0 (𝐱 ⋆ (𝑡)) = 𝑓0 (𝐱 ⋆ ).
𝑡→+∞ 𝑡→+∞

However, observe that setting 𝑡 = +∞ in (58) results in 𝑓log (𝐱, +∞) = 𝑓0 (𝐱), which ignores the
inequality constraints altogether. Therefore, in practice, starting at say 𝑡 = 1, the parameter 𝑡 is
progressively increased, and each time 𝑡 is increased, the function 𝑓log (𝐱, 𝑡) is minimized using
Newton’s method. This process is known as the barrier method, which is an interior point method
[18](p.569). The application of Newton’s method, which is the workhorse of the barrier method,
for the minimization of 𝑓log (𝐱, 𝑡) is accomplished as follows. Consider the Taylor series expansion
of the logarithmic barrier function 𝑓log (𝐱, 𝑡) centered at 𝐱c ∈ ℝN×1 :
1
𝑓log (𝐱, 𝑡) = 𝑓log (𝐱c , 𝑡) + 𝐠 Tlog (𝐱c , 𝑡)[𝐱 − 𝐱c ] + 2[𝐱 − 𝐱c ]T 𝐇log (𝐱c , 𝑡)[𝐱 − 𝐱c ] + ⋯,

where, 𝐠 log (𝐱c , 𝑡) ∈ ℝN×1 and 𝐇log (𝐱c , 𝑡) ∈ ℝN×N are the gradient and Hessian delineated in (59)
and (60), respectively, evaluated at the expansion point 𝐱 = 𝐱c , namely
𝐠 log (𝐱c , 𝑡) = ∂𝐱 𝑓log (𝐱, 𝑡)| ∈ ℝN×1 , 𝐇log (𝐱c , 𝑡) = ∂𝐱𝐱 T 𝑓log (𝐱, 𝑡)| ∈ ℝN×N .
𝐱=𝐱c 𝐱=𝐱 c

Therefore, by letting 𝐝 = 𝐱 − 𝐱c , so that 𝐱 = 𝐝 + 𝐱c , whilst letting the function 𝑞log (𝐝 + 𝐱c , 𝑡) be

the quadratic approximation of the logarithmic barrier function 𝑓log (𝐝 + 𝐱c , 𝑡), which is obtained
by retaining the first three terms of the Taylor series expansion of 𝑓log (𝐝 + 𝐱c , 𝑡), namely

1
𝑞log (𝐝 + 𝐱c , 𝑡) = 𝑓log (𝐱c , 𝑡) + 𝐠 Tlog (𝐱 c , 𝑡)𝐝 + 2𝐝T 𝐇log (𝐱c , 𝑡)𝐝. (64)

Now, the stationary point of the quadratic approximation in (64), with respect to 𝐝 ∈ ℝN×1 , is
obtained by setting its gradient to zero, namely
𝜕𝐝 𝑞log (𝐝 + 𝐱 c , 𝑡) = 𝐠 log (𝐱c , 𝑡) + 𝐇log (𝐱c , 𝑡)𝐝 = 𝟎 ⟶ 𝐇log (𝐱c , 𝑡)𝐝 = 𝐠 log (𝐱c , 𝑡),
which is a linear system of equations whose solution is given by

−1
𝐝 = −𝐇log (𝐱c , 𝑡)𝐠 log (𝐱c , 𝑡) (65)

Page 28-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Now, Newton’s method produces a sequence of iterates 𝐱0 , 𝐱1 , 𝐱2 , ⋯ , 𝐱𝑘 , ⋯ that progressively

approach the minimum value of the logarithmic barrier function 𝑓log (𝐱, 𝑡). Therefore, if one lets
𝐱c = 𝐱𝑘 ∈ ℝN×1 (that is, the current iterate) and 𝐱 = 𝐱 𝑘+1 ∈ ℝN×1 (that is, the next iterate), then
by solving for 𝐝 ∈ ℝN×1 in (65) (which corresponds to the Newton step, namely 𝐝 = 𝐱𝑘+1 − 𝐱𝑘 ),
then the next iterate is computed straightforwardly by evaluating 𝐱𝑘+1 = 𝐱𝑘 + 𝐝. In this work,
Newton’s method is terminated when the following condition is satisfied:
1 T
2
𝐝 𝐇log (𝐱𝑘 , 𝑡)𝐝 ≤ ϵnewt ,

where 𝐝T 𝐇log (𝐱𝑘 , 𝑡)𝐝 a quadratic form, which corresponds to the square of the Newton decrement
and ϵnewt > 0 is the pre-specified termination criterion associated with the Newton iterations.
The barrier method has two types of iterations, the outer iterations and the inner iterations.
Specifically, the outer iterations update the barrier parameter 𝑡 and the inner iterations
correspond to Newton iterations. The pseudo-code for the barrier method is given below.

given a termination criterion ϵbarr > 0, and

given a parameter 𝜇 > 1, and
given a point 𝐱 ∈ ℝN×1 that satisfies 𝐀𝐱 ≼ 𝐛
set 𝑡 = 1
while (K in /𝑡) ≤ ϵbarr
1. 𝔪𝔦𝔫𝔦𝔪𝔦𝔷𝔢 𝑓log (𝐱, 𝑡) using Newton’s method starting at 𝐱
𝐱∈ℝN×1
2. 𝐱 ∶= 𝔞𝔯𝔤𝔪𝔦𝔫 𝑓log (𝐱, 𝑡)
𝐱∈ℝN×1
3. 𝑡 ∶= 𝜇𝑡
end while

Copyright © 2017 SES ltd. All rights reserved. Page 28-19

PART IV: RESEARCH WORK IN PROGRESS

5 Numerical Results
Unconstrained Least Squares Fit Derivative Constrained Least Squares Fit

0.5

Polarization Data
0

Strictly Increasing Fit Oscillatory Fit

-0.5

(a)
-1

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

4.5

3.5

3 (b)
2.5

1.5

0.5

-0.5

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Figure 3: Comparison of unconstrained (see §3) and derivative constrained (see §4) least squares
polarization curve fitting for M = 200 polarization data points. The number of degrees of freedom is N =
75, corresponding to a 74th order Chebyshev polynomial expansion. Here, the lower and upper bounds on
the differential conductance are set to 𝒢lower (𝑣) = 0.1 and 𝒢upper (𝑣) = 4.0 on the interval −1 ≤ 𝑣 ≤ +1. The
barrier update parameter is set to 𝜇 = 50. The outer iterations are terminated when a tolerance of ϵbarr =
10−3 is reached and the inner iterations are terminated when as tolerance of ϵnewt = 10−2 is reached. (a)
The least squares fits overlay the polarization data. Observe oscillatory behavior and the unconstrained
least squares fit and the monotonicity of the derivative constraint least squares fit. (b) First order derivatives
of the fits. As required, 𝒢(𝑣) of the constrained fit oscillated between 𝒢lower (𝑣) and 𝒢upper (𝑣).

Page 28-20 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Unconstrained Least Squares Fit

-4
Derivative Constrained Least Squares Fit
x 10

3
Oscillatory Fit
(a)
2

Polarization Data

1
Strictly Increasing Fit

-1
-1.1 -1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2
-4
x 10
20

(b)
15

10

-5
-1.1 -1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2

Figure 4: Comparison of unconstrained (see §3) and derivative constrained (see §4) least squares
polarization curve fitting for M = 200 coated steel polarization data points. The number of degrees of
freedom is N = 50, corresponding to a 49th order Chebyshev polynomial expansion. Here, the lower and
upper bounds on the differential conductance are set to 𝒢lower (𝑣) = 0 and 𝒢upper (𝑣) = +∞ on the interval
−1.177 ≤ 𝑣 ≤ −0.180. The barrier update parameter is set to 𝜇 = 2. The outer iterations are terminated
when a tolerance of ϵbarr = 10−8 is reached and the inner iterations are terminated when as tolerance of
ϵnewt = 10−2 is reached. (a) The least squares fits overlay the polarization data. Observe oscillatory behavior
and the unconstrained least squares fit and the monotonicity of the derivative constraint least squares fit.
(b) First order derivatives of the fits. As required, 𝒢(𝑣) of the constrained fit oscillated above 𝒢lower (𝑣).

Copyright © 2017 SES ltd. All rights reserved. Page 28-21

PART IV: RESEARCH WORK IN PROGRESS

6 Conclusion
The effectiveness of derivative constrained polarization curve fitting has been clearly
demonstrated. This mathematical method has been implemented in the SES’ CurveFit software
tool available as a standalone unit or as a component of the CorrCAD software package.

7 References
[1] D. B. Strukov, G. S. Snider, D. R. Stewart, and R. S. Williams, "The missing memristor
found," Nature, vol. 453, pp. 80-83, 2008.
[2] M. D. Pickett, G. Medeiros-Ribeiro, and R. S. Williams, "A scalable neuristor built with
Mott memristors," Nat Mater, vol. 12, pp. 114-117, 2013.
[3] G. L. Plett, Battery Management Systems, Volume I: Battery Modeling: Artech House,
2015.
[4] R. W. Revie, Uhlig's Corrosion Handbook. Hoboken, New Jersey: John Wiley & Sons,
2011.
[5] F. Yu, O. Addison, S. J. Baker, and A. J. Davenport, "Lipopolysaccharide inhibits or
accelerates biomedical titanium corrosion depending on environmental acidity,"
International Journal of Oral Science, vol. 7, pp. 179-186, 2015.
[6] L. O. Chua, C. A. Desoer, and E. S. Kuh, Linear and Nonlinear Circuits. New York:
McGraw-Hill Inc., 1987.
[7] J. Stewart, Calculus: Early Transcendentals: Cengage Learning, 2016.
[8] O. F. Devereux, "Polarization Curve-Fitting by Computer Modelling," CORROSION, vol.
35, pp. 125-129, 1979.
[9] R. S. Munn, Computer Modeling in Corrosion. Philadelphia, PA: American Society for
Testing and Materials, 1992.
[10] R. S. Munn and O. F. Devereux, "Numerical Modeling and Solution of Galvanic Corrosion
Systems: Part I. Governing Differential Equation and Electrodic Boundary Conditions,"
CORROSION, vol. 47, pp. 612-618, 1991.
[11] K. S. Yeum, K. Y. Kim, and O. F. Devereux, "Polarization of the H2, CO, CO2|Ni and H2S,
H2, CO, CO2|Ni electrodes in molten sodium carbonate," Corrosion Science, vol. 25, pp.
1127-1147, 1985.
[12] M. K. Transtrum, B. B. Machta, and J. P. Sethna, "Why are Nonlinear Fits to Data so
Challenging?," Physical Review Letters, vol. 104, p. 060201, 2010.
[13] C. F. Gauss, Theoria Motus Corporum Coelestium in Sectionibus Conicis Solem
Ambientium. Hamburgi: Frid. Perthes et I.H. Besser, 1809.
[14] A. M. Legendre, Nouvelles Méthodes pour la Détermination des Orbites des Comètes.
Paris: F. Didot, 1805.
[15] W. J. Schwerdtfeger, "Corrosion Rates of Binary Alloys of Nickel and Iron Measured by
Polarization Methods," JOURNAL OF RESEARCH of the National Bureau of Standards
- C. Engineering and Instrumentation, vol. 70, pp. 187-194, 1966.
[16] S.-P. Wu, S. Boyd, and L. Vandenberghe, "FIR Filter Design via Spectral Factorization and
Convex Optimization," in Applied and Computational Control, Signals, and Circuits:
Volume 1, B. N. Datta, Ed., ed Boston, MA: Birkhäuser Boston, 1999, pp. 215-245.
[17] A. Forsgren, P. E. Gill, and M. H. Wright, "Interior Methods for Nonlinear Optimization,"
SIAM Review, vol. 44, pp. 525-597, 2002.
[18] S. Boyd and L. Vandenberghe, Convex Optimization. New York: Cambridge University
Press, 2004.
[19] W. Karush, "Minima of Functions of Several Variables with Inequalities as Side
Constraints," M.Sc., Department of Mathematics, University of Chicago, Chicago, Illinois,
1939.

Page 28-22 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

[20] H. W. Kuhn and A. W. Tucker, "Nonlinear Programming," in Proceedings of the Second

Berkeley Symposium on Mathematical Statistics and Probability, California, 1951, pp.
481-492.
[21] M. H. Wright, "The interior-point revolution in optimization: History, recent
developments, and lasting consequences," Bulletin of the American Mathematical
Society, vol. 42, pp. 39-56, 2005.

Copyright © 2017 SES ltd. All rights reserved. Page 28-23

 UGM 2017 – CANNON BEACH, OREGON

29 THE SIMULATED ANNEALING METHOD APPLIED TO

THE SOIL RESISTIVITY ANALYSIS INVERSE PROBLEM

Stéphane Franiatte, Octavio Ramos-Leaños, Luis Valcárcel, Amir Hajiaboli,

Simon Fortin and Farid P. Dawalibi
Safe Engineering Services & Technologies ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

This article introduces a new methodology to fit computed soil models to field measurements data.
This methodology, called Simulated Annealing, is a metaheuristic method capable of finding the global
minimum of complicated functions. Such a feature provides an alternative method to find suitable soil
models when other gradient-based methods fail. The methodology has been implemented in RESAP
with interesting results. Many cases have been tested and the results obtained are reported in this
work.

1 Introduction
The inverse problem in soil resistivity analyses, i.e., finding a soil model which reproduces the
electrical properties of the measurements, can be classified as an optimization problem. In all
optimization methods, an initial model is modified in an iterative manner such that the difference
between the model’s response and the observed data values is reduced. Indeed, soil structure
models are obtained from resistivity measurements using a numerical method that minimizes the
RMS error between the measured and computed resistivity values.
Presently, the Steepest Descent (SD) and the Levenberg-Marquardt (LM) methods are the two
optimization methods dedicated to the inverse problem available in the RESAP computation
module. Although both methods work fine in most cases, they sometimes fail to find a satisfactory
solution, even for noiseless data. A hypothesis to explain those failures is that both methods are
local optimization methods: their convergence process is susceptible to getting trapped in a local
minimum preventing them from finding the true global minimum. Therefore, if the initial
estimate for the parameters to be optimized is not sufficiently close to the parameters of the actual
solution, these may never be found. A solution to this issue would be to vary randomly the initial
parameters’ estimates but this approach could possibly become computationally inefficient.
Global optimization methods, however, do exist and one of them, a metaheuristic called
Simulated Annealing (SA), has been found to be of particular interest and has been applied to
geophysical inversion problems by groups of geophysicists and mathematicians in the past [1].
In order to allow the grounding community to also experiment with this method, and since the
method has proven useful at obtaining complicated soil structure models, it was decided to
implement it in RESAP, as an additional method or simply as a complement to the existing local
optimization algorithms. This article presents the results of the implementation of this global
optimization method.

Copyright © 2017 SES ltd. All rights reserved. Page 29-1

PART IV: RESEARCH WORK IN PROGRESS

2 The Simulated Annealing Metaheuristic

The Simulated Annealing (SA) metaheuristic was first introduced separately by two different
groups in [2] and [3] as a combinatorial optimization method. SA is inspired by the work of
Metropolis et al. [4] in computational statistical mechanics; in [4], a Monte Carlo sampling
method [5] was used to simulate the evolution of an atomic system submitted to a heat bath (hence
the name of the optimization method, which still remains its most common appellation today).
The proposed Monte Carlo method focused on low-energy configurations of the system; this focus
on low-energy, which is akin to the optimization of a cost function, and the combinatorial aspect
of the method inspired the authors of [2] and [3] to adapt and successfully apply the Metropolis
algorithm to combinatorial optimization problems. One of these was the well know, NP-complete
Travelling Salesman problem. SA was later on derived to address functions of continuous
variables [6].
The main feature and advantage of the SA metaheuristic is that it authorizes uphill moves under
the control of a probabilistic criterion, which gives the method the ability to escape local minima
on its path toward an optimal solution. SA is therefore largely independent of the starting values.
For instance, Figure 1 below illustrates two types of well-known test functions exhibiting either
multiple local minima (a), the multiplicity of which represents one of the main challenges in soil
inversion problems, or a unique but hard to localize global minimum, located inside a narrow,
parabolic shaped flat valley (b). For such functions, convergence will be harder for gradient-based
algorithms such as the SD and the LM methods if the initial estimate of the parameters is not
adequately chosen. On the other hand, a well-tuned SA correctly finds the global minimum,
regardless of the initial estimate.

a) b)

Figure 1: Example of functions for which finding the global minimum proves difficult:
a) Second order Griewank function, global minimum at (0, 0);
b) Second order Rosenbrock function, global minimum at (1, 1).

Page 29-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.1 How it Works

The SA version used in RESAP is based on the algorithm of Corana et al. [6], which was designed
for functions of continuous variables. Like the original SA version, the method is essentially an
adequately guided iterative random search procedure with the Metropolis algorithm at its core,
as will be seen below.
2.1.1 The Objective Function
As a derivative-free method, SA can work with almost any form of objective function (the generic
appellation for the function to be minimized, also known as the cost function), in the sense that it
does not need to comply with any restriction of continuity or derivability. In RESAP, the SA
objective function f is the error criterion presented in [7]; which is “the average over the sample
values of the squares of the relative differences between the sample values of the apparent
resistivity obtained from the field, and those obtained from the model” as estimated by the SA
metaheuristic. The error criterion is usually referred to as the RMS error.
2.1.2 The Metropolis Algorithm
The Metropolis algorithm is a stochastic procedure that allows for uphill moves during the SA
optimization process, according to the delta (difference) in RMS error between the current best
solution and the new candidate, and the current value of a temperature parameter 𝑇, which
provides a reference for the magnitude of allowable random fluctuations.
At each iteration, i, if the new solution candidate’s RMS error is higher than the previous one, the
algorithm will either accept or reject the candidate according to the acceptance probability 𝑝:
𝑓 −𝑓′
−( 𝑖 ) (1)
𝑝=𝑒 𝑇𝑘
,
where k denotes the successive temperature reductions, i the current iteration index, 𝑓𝑖 the cost,
or RMS error, of the best solution candidate at iteration i, 𝑓𝑖 (𝒙𝑖 ), and 𝑓′ the cost of the current
candidate. In practice a pseudorandom number 𝑝′ is generated within the range [0,1] and is then
compared with 𝑝. If 𝑝′ < 𝑝 the candidate is accepted as the new current solution, otherwise it is
rejected. This process is repeated until convergence is reached or a termination criterion is met,
e.g. if the RMS error does not significantly change after a given number of iterations or the RMS
error is below a specific reference value.
2.1.3 The Cooling Schedule
The temperature parameter 𝑇 is monotonically decreased during the SA process; at the beginning,
according to the Metropolis criterion (1), a high temperature will allow for more uphill moves
permitting initially a wide exploration of the solution candidates’ space. As SA moves closer to the
actual solution and the temperature slowly decreases less uphill moves will occur.
In the RESAP implementation of SA, the so-called geometric schedule is used to reduce the
temperature, at each iteration:

𝑇𝑘 = 𝛼𝑇𝑘−1 , 𝛼 ∈ ]0,1[, (2)

where 𝛼 is a constant for the whole simulation, named the Cooling Factor.

Copyright © 2017 SES ltd. All rights reserved. Page 29-3

PART IV: RESEARCH WORK IN PROGRESS

2.1.4 Adaptive Neighborhood and Trial Solutions Generation

Corana’s version of SA relies on an adaptive neighboring method, controlling the generation of
the next iteration step size, to maintain a 1:1 ratio between accepted and rejected solution
candidates. This avoids processing waste, considering the fact that an unconstrained continuous
problem has an infinite number of solution candidates, unlike combinatorial problems for which
SA was initially designed.
Adaptive neighboring works with a regularly updated step vector 𝒗 that records the maximum
possible increments along each direction of the solution candidate vector at each iteration. Hence,
starting from an initial solution estimate 𝒙𝑖 , the method generates a random solution candidate
x’ at each iteration:

𝒙′ = 𝒙𝑖 + 𝒗𝒓, (3)

where r is a random number in the range [-1, 1]. The new solution candidate is then either
accepted if it decreases the RMS error, or submitted to the Metropolis criterion as described in
Section 2.1.2.
2.1.5 Constraints
Other than the constraint on the next iterative step implied by the adaptive method above,
constraints limiting the SA method’s search space to realistic physical values for its parameters
were also implemented for RESAP’s inverse problem. These constraints act on each soil model
layer’s thickness and resistivity, as minimum and maximum allowable values for both cases.
2.1.6 Simulated Annealing and Pseudo-Random Numbers
The SA metaheuristic belongs to the family of stochastic global optimization methods. The
“stochastic” part of the optimization refers to the use of random numbers in the process of finding
the best solution to a given problem. Randomness is used to generate new solution candidates
based on the last best solution; as shown in Section 2.1.2, it is also used to determine whether a
candidate solution is to be accepted, based on the Metropolis criterion (1). The optimization
process is therefore non-deterministic: although it converges to probabilistically likely solutions
it doesn't necessarily always give the exact same solution when the initial seed of the random
number generator is different.
In RESAP’s implementation of SA, the possibility to freeze the random numbers sequence using
a fixed seed is offered to the user (and selected by default). However, it is worth noting that while
a fixed seed will always give the exact same soil model, using a random seed on different attempts
may allow the identification of different types, or families, of soil models that could all fit very well
the initial measurements. This can provide additional interpretation possibilities that could help
correlate the results with other sources of information regarding the soil, e.g. from soil boring
analyses.

Page 29-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

2.2 Method’s Parameters

An ANNEALING sub-command was added under the RESAP’s METHODOLOGY command. The
sub-command variables are described in Table 1 below.
Table 1: RESAP's ANNEALING command variables description.

Default
Variable’s name Domain Description
Value

Initial
1 ]0, +∞[ The initial temperature value.
Temperature T0

The geometric cooling factor, as

Cooling Factor 𝜶 0.85 ]0,1[
defined in (2).

Logical value. Set to 1 to choose a

Random Solution 0 {0,1}
random seed.

Limit the algorithm search space to a

Max Resistivity 100,000 ]0, +∞[ maximum resistivity value, for each
layer of the soil model.

Limit the algorithm search space to a

Min Resistivity 0.01 ]0, +∞[ minimum resistivity value, for each
layer of the soil model.

Limit the algorithm search space to a

Max Thickness 10,000 ]0, +∞[ maximum thickness value, for each
layer of the soil model.

Limit the algorithm search space to a

Min Thickness 0.01 ]0, +∞[ minimum thickness value, for each
layer of the soil model.

The maximum number of iterations

Max Iterations 500,000 ]0, +∞[ allowed for one Simulated Annealing
optimization run.

2.3 Command Mode Usage

Unlike the STEEPEST and MARQUARDT commands, the ANNEALING command uses its own
variable for the maximum number of iterations. The only OPTIMIZATION sub-command
ANNEALING needs is the ACCURACY; as for STEEPEST and MARQUARDT, it is used by
ANNEALING as the target RMS error.

Copyright © 2017 SES ltd. All rights reserved. Page 29-5

PART IV: RESEARCH WORK IN PROGRESS

To use the SA method, an RS_.F05 file should have the OPTIMIZATION command and sub-
commands defined as follows:
OPTIMIZATION
METHODOLOGY
ANNEALING

where the default ANNEALING variables values are as shown in Table 1, and the default
ACCURACY value is set at 2.5%.
A more advanced usage of the ANNEALING command could be as follows:
OPTIMIZATION
ACCURACY, 0.0025
METHODOLOGY
ANNEALING, 1, 0.99, 1, 50000, 20, 1000, 1, 250000

In this example, the algorithm will aim at a 0.25% RMS error. The initial temperature is left to its
default value, however it will decrease much slower with a cooling factor value of 0.99, allowing
for a wider exploration of the search space. Maximum and minimum values for both resistivity
and thickness are also defined; doing so may be particularly useful for instance if a previous run
resulted in unrealistic soil models. Finally, the maximum number of iterations was decreased to
250,000: this could be done for instance to quickly identify different types of possible solutions,
when the random-number seed is itself random (see Section 2.1.6), as specified in this example.

3 Illustrative Example
In order to illustrate the advantages of the SA method, consider the inversion of the soil resistivity
measurements presented in Figure 2, which depicts a low resistivity surface soil above higher
resistivity material. Closer examination of the curve however reveals that the equivalent soil
model to be determined, assuming it is a horizontally layered soil, actually necessitates at least
four layers, given the number of inflections in the curve, which is a crucial observation to make in
order to obtain a reasonable interpretation. In the following sections, we therefore try to solve this
example with a 4 layer model with the SD, LM and finally with the SA methods.
In all cases, the target accuracy was the default 2.5% RMS error. SD and LM saw their step size
reduced to 10-6 and their maximum iteration number increased to 105 to augment their chance of
convergence to an acceptable solution. SA parameters assume their default value as described in
Section 2.2. We also gave no initial estimate for the soil model’s layer thickness and resistivity,
giving RESAP the freedom to make that choice.

Page 29-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: Measurements of soil apparent resistivity, for which a soil model yielding such a response is
sought.
3.1.1 Steepest Descent Results
Figure 3 shows that the SD method converges to an unsatisfactory soil model for the example case
yielding a 40.21% RMS error. The two middle layers are barely distinguishable and could have
been combined into a single thicker layer of similar resistivity yielding a 3 layer soil model. This
fact is reflected in the computed apparent resistivity curve, green line, which misses the measured
data everywhere except at the largest spacings.
British/Logarithmic X and Y LEGEND

10 4
Measured Data
Computed Results Curve
Soil Model

Measurement Method..: Wenner

RMS error...........: 40.21%

Layer Resistivity Thickness

10 3 Number (Ohm-m) (Feet )
====== ============== ==============
Air Infinite Infinite
2 16.87609 2.490000
Apparent Resistivity (Ohm-meters)

3 1679.451 4.610080
4 1735.919 15.50075
5 707.0944 Infinite

10 2

10 1
10 -2 10 -1 10 0 10 1 10 2 10 3 RESAP <ANN_Fault500kV_FD >

Inter-Electrode Spacing (feet)

Figure 3: Soil interpretation results obtained with the Steepest Descent method, for a four-layer soil
model and otherwise default parameters.

Copyright © 2017 SES ltd. All rights reserved. Page 29-7

PART IV: RESEARCH WORK IN PROGRESS

3.1.2 Levenberg-Marquardt Results

Similarly, results obtained with the LM method shown in Figure 4 are not better than those
obtained with SD, they even present a larger RMS error of 44.95%. This time, it is still effectively
a 3 layer soil model that was produced if we exclude the unrealistic very high resistivity and very
thin second layer. The computed apparent resistivity curve still misses most measured data points
and is therefore also deemed unsatisfactory.
British/Logarithmic X and Y LEGEND

10 6
Measured Data
Computed Results Curve
Soil Model

10 5 Measurement Method..: Wenner

RMS error...........: 44.95%

Layer Resistivity Thickness

Number (Ohm-m) (Feet )
====== ============== ==============
10 4 Air
2
Infinite
16.85245
Infinite
2.490000
Apparent Resistivity (Ohm-meters)

3 133968.7 0.4854628E-01
4 211.1998 60.36362
5 1747.218 Infinite

10 3

10 2

10 1
10 -2 10 -1 10 0 10 1 10 2 10 3 RESAP <ANN_Fault500kV_FD >

Inter-Electrode Spacing (feet)

Figure 4: Soil interpretation results obtained with the Levenberg-Marquardt method, for a four-layer soil
model and otherwise default parameters.

3.1.3 Simulated Annealing Results

On the other hand, Figure 5 (obtained with the default pseudo-random number seed) and Figure
6 (obtained with a randomized seed) shows that the SA method yields excellent results with 2.81%
and 2.34% RMS errors, respectively. The computed apparent resistivity curves clearly go through
the measured data points in both cases and are both satisfactory. This therefore reveals the
existence of at least two possible families of soil models yielding similar results, an illustration of
the well-known fact that the interpretation of soil measurements may not yield a unique solution.
Here, one family favors a thinner, higher resistivity, second layer compared to the other one,
which compensates with a thinner third layer. Note here that, in both cases, the bottom layer
ended up having a very high resistivity. It is probable that there is a large uncertainty associated
with that layer, which could be lifted by locking its resistivity to a lesser value and observing the
change in the RMS value. In practice, additional measurements at larger spacings could be
requested in such a case, in order to clearly establish the trend of the average deep soil resistivities.

Page 29-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: Soil interpretation results obtained with the SA metaheuristic, for a four-layer soil model and
otherwise default parameters.

Figure 6: Another soil model obtained for the same measurements, with the Random Solution variable of
the ANNEALING command set to 1, i.e. with a different random seed.

Copyright © 2017 SES ltd. All rights reserved. Page 29-9

PART IV: RESEARCH WORK IN PROGRESS

4 Statistical Testing
The section above gives an example of a case where the SA method performs better than the SD
and LM methods. This is so because SD and LM get trapped sometimes in local minima values. It
happens to be often the case when RESAP is having difficulty providing a good estimate for the
initial parameters (it is also the case when this exercise is not necessarily trivial for a human being
as well, especially so for people with little experience with soil measurement interpretation). In
order to better test the usefulness of the SA method, a larger statistical sample of soil resistivity
measurement inversions is in order.

4.1 Test Case Samples

A large collection of quality field measurements is difficult to obtain. Moreover, even if such a
collection happened to be available, it would be a tedious exercise to classify them all by number
of layers to be used in the analysis and this would most likely be complicated by the presence of
noise in the data. A way around this difficulty is to generate synthetic soil measurements instead.
This can readily be achieved with the help of RESAP itself and of SESScript. Indeed, when a soil
model is specified in a RESAP file and all its parameters are locked, then the output of the
computation will simply be a simulation of the measurements that would have been obtained in
the field for that given soil model. SESScript can be used as a means to duplicate such a run, while
varying the soil’s model parameters, one at a time, in a systematic way. The end result is a
collection of noiseless measurement traverses corresponding to known soil structures. The last
step of the exercise is to extract these measurement traverses from the output files and use them
as inputs for new RESAP runs to be analyzed with the different optimization methods without
specifying an initial estimate for the starting values. This can be achieved relatively easily with
additional scripting.
For the purposes of comparing the efficiencies of the different methods, a large number of test
cases were therefore created using the procedure described above. They explore a wide, yet
realistic, part of the parameter space. Layer thicknesses were varied between 0.01 m and 150 m,
but in such a way as to create a tendency for the deeper layers to have values toward the higher
end of the range, because it would be unreasonable, due to the nature of the soil measurement
method, to expect detecting thin layers located deep in the ground. The layers’ resistivity was
varied between 0.1 ∙m (lower range for sea water) and 15,000 ∙m (higher range for sand and
typical for limestone, igneous rock and permafrost, although their resistivity range can still reach
ten times this value). Table 2 lists the parameter values that were given for the different types of
soil. All combinations of parameters were allowed so as to examine systematically all kinds of
layering structures and resistivity contrast ratios. In total, 512 two-layer cases, 576 three-layer
cases, 648 four-layer cases, 768 five-layer case and 1024 six-layer cases were produced. The
resistivity value of adjacent layers was chosen carefully such as to always have them be distinct
and therefore avoid the formation of what would otherwise effectively be a thicker layer.

Page 29-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 2: Soil parameters used for the creation of a large sample of test cases.

The probe spacings used for the creation of the synthetic Wenner-type soil measurements
followed the recommended exponential progression, i.e. 0.1 m, 0.2 m, 0.3 m, 0.5 m, 0.7 m, 1 m,
2 m, 3 m, 5 m, 7 m, 10 m, 20 m, 30 m, 50 m, 70 m, etc., up to 3000 m. The traverses are therefore
sufficiently long to allow probing the deeper layers of all the sample test cases. The High-Precision
filter of RESAP was also used, for best accuracy.

4.2 Parameter Optimization

As explained in Section 2.1.2, the SA method uses a temperature parameter that regulates the size
of the allowable next random step. The initial temperature is set arbitrarily and also decreases at
an arbitrary rate, the cooling factor. The method will perform more or less efficiently depending
on the particular choice of these parameters, and a few other internal meta-parameters not
exposed through the ANNEALING command. Therefore, fine-tuning them is an important part
of the development of this method and becomes an optimization problem of its own, for which
the method could be applied on itself actually. The statistical sample described above can be used
for this purpose and this was done up to a certain extent (optimizing by hand rather than through
an automated process). More work is however required to complete this effort.
The parameters that were settled on for this study have been set as the default parameters for the
method and are described in Table 1.

4.3 Comparisons with Local Optimization Algorithms

In this section, comparisons and analyses of results obtained for all cases introduced in Section 4.1
are presented. Each case was run with the three optimization methods available in RESAP. The
computation options for all three methods were set such that as to maximize the chances of
finding a good candidate solution, as shown below:

Copyright © 2017 SES ltd. All rights reserved. Page 29-11

PART IV: RESEARCH WORK IN PROGRESS

SOIL-TYPE,MULTILAYER
HORIZONTAL,AUTOMATIC

COMPUTATIONS
FILTER,HIGH-PRECISION

OPTIMIZATION
ACCURACY,0.00025
ITERATIONS,100000
STEPSIZE,0.0001

A high goal was set for the target accuracy value and a large number of iterations was allowed for
achieving it. The number of layers and their initial parameters was left to be determined
automatically by RESAP.

4.3.1 Two-Layer Soils

Figure 7 shows the RMS error plot for all the 512 two-layer soil cases. Here and for the other plots
below, the runs have been ordered in decreasing order of RMS error for the SD method, which
explains the trend exhibited by the orange dots. This was done to make it easier to compare the
cases where the RMS error is large for this method, which acts as a reference. Note also that the
y-axis is logarithmically scaled to expose better the wide range of RMS values obtained. The
resolution of the RMS error is of 0.01%, this is why there is no data below 0.01% and why the y-
values increase in 0.01% steps.
The figure shows that, on average, SA yields lower RMS errors than SD, but larger RMS errors
than LM. The close-up view of Figure 8, to the worst performers of the SD method, reveals that
SA dramatically improves the fit of the soil data in this range. It can also be seen that, for cases 6
and 13, SA improves dramatically the fit compared to both SD and LM methods, by reducing the
RMS error from about 50% to 60%, to less than 0.02%. Notice that for these two cases LM and SD
both get stuck in local minima and cannot find a proper solution, even after 100,000 iterations.
The reverse can however also be true and there are several cases where SA failed at finding a
solution below the 10% RMS error mark, say, while the other two methods could. Closer
inspection of these cases could reveal interesting patterns and suggest ways of improving the
different methods.

Page 29-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 7: RMS error for two-layer cases as obtained with the three different methodologies.

Figure 8: RMS error for two-layer cases as obtained with the three different methodologies, cases 0-20.
Table 3 shows the average RMS error for all three methods. It suggests that, for two-layer soils,
they all perform quite well, with more than 91% of the cases yielding an RMS error of 10% or less.
LM presents the largest number of cases having the lowest RMS error of all three methods, with
99.6% of cases resulting in an RMS error of 10% or less, 98.4% of cases with an RMS error of 1%
or less, and 91.2% of cases with an RMS error of 0.1% or less, which is remarkable. SA does not
perform as well, exhibiting less success for the 1% and 0.1% RMS error marks, but does
considerably better than SD for all these cases.

Copyright © 2017 SES ltd. All rights reserved. Page 29-13

PART IV: RESEARCH WORK IN PROGRESS

Table 3: Number of two-layer cases resulting in low RMS error.

Number of Simulations
Average
RMS <= 0.1% <= 1% <= 10%
Method Total RMS RMS RMS
Error
(%) Error Error Error
512 467 504 510
Levenberg-Marquardt 0.26
(100%) (91.2%) (98.4%) (99.6%)
512 189 330 467
Steepest Descent 3.10
(100%) (36.9%) (64.5%) (91.2%)
512 402 449 486
Simulated Annealing 1.34
(100%) (78.5%) (87.7%) (94.9%)

4.3.2 Three-Layer Soils

For the three-layer cases, Figure 9 shows a similar trend to that of the two-layer cases, i.e. LM
produces the largest number of cases with low RMS error values. However, in this case it is
interesting to note that LM is the only method resulting in RMS errors larger than 100%. There
are also cases where SA provides better results than LM, take as an example Case 20 shown in
Figure 10, where the RMS error obtained with LM and SD is around 42% whereas that presented
by SA is of only 0.02%, a dramatic improvement with respect to the other two methods.
In Figure 9, a trend that was slightly noticeable in Figure 7 becomes more apparent: there seems
to be many cases where SA and LM yield the same non-optimal result as SD, at least in terms of
RMS errors (there could be different soil models yielding the same RMS error). This seems to
happen more often for SA and is a trend that becomes still more noticeable for higher numbers of
soil layers as can be seen in the following sections. Further investigation is required to determine
why this is so as it could be an indication that the method needs more optimization of its own
controlling meta-parameters in order to explore the soil structure’s parameter space more
efficiently. For instance, it could be that a greater start temperature for SA is required to improve
its performance. On the other hand it seems that LM has a greater tendency to escape the local
minima and wander off to very good results or to bad ones, which is counter intuitive.

Page 29-14 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 9: RMS error for three-layer cases as obtained with the three different methodologies. The average
for LM is identical to the SD.

Figure 10: RMS error for three-layer cases as obtained with the three different methodologies, cases 15-25.
The average for LM is identical to the SD.

Table 4 shows the average RMS errors for all three methods. It suggests that for three-layer soils,
their ability to provide good results in most cases is still reasonable. LM again yields the greatest
number of cases with the lowest RMS error of all three methods. The percentage of cases below
the 10%, 1% and 0.1% marks are all lower than for the two-layer cases, but the separation between
LM and SA is less significant.

Copyright © 2017 SES ltd. All rights reserved. Page 29-15

PART IV: RESEARCH WORK IN PROGRESS

Table 4: Number of three-layer cases resulting in low RMS error.

Number of Simulations
Average
RMS <= 0.1% <= 1% <= 10%
Method Total RMS RMS RMS
Error
(%) Error Error Error
576 335 397 470
Levenberg-Marquardt 9.54
(100%) (58.2%) (68.9%) (81.6%)
576 75 172 387
Steepest Descent 9.68
(100%) (13.0%) (29.9%) (67.2%)
576 281 356 461
Simulated Annealing 6.14
(100%) (48.8%) (61.8%) (80.0%)

4.3.3 Four-Layer Soils

Figure 11 shows the resulting RMS errors for the four-layer cases. This figure shows again that LM
is providing the largest number of cases with lower RMS errors. However, SA is able to provide a
handful of cases where it provides a lower RMS error than the other two methods. For example,
for case 144, the resulting RMS error for LM and SD is around 14% whereas SA yields a 4% RMS
error. Four-layer cases have proven to be more difficult for SA to distinguish itself from SD than
in the two- and three-layer cases.

Figure 11: RMS error for four-layer cases as obtained with the three different methodologies. The average
for LM is not shown because it would overlap the one for SA.
Table 5 shows that, for four-layer soils, LM was able to solve 80.1% of the cases with good fits
having an RMS error of 10% or less, while SD and SA are pretty much on par at 73.6% and 70.1%,
respectively. This changes the trend observed for the two- and three-layer cases, where SA was
better than SD for the 10% RMS error mark. However, SA solved 38.1% of the cases with very good
fits having an RMS error of 1% or less against 17.6% of the cases for the SD method. A similar
discrepancy in favor of SA is also noticeable for the extremely good fits having an RMS error of
0.1% or less.

Page 29-16 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table 5: Number of four-layer cases resulting in low RMS error.

Number of Simulations
Average
RMS <= 0.1% <= 1% <= 10%
Method Total RMS RMS RMS
Error
(%) Error Error Error
648 241 319 519
Levenberg-Marquardt 9.43
(100%) (37.2%) (49.2%) (80.1%)
648 19 114 477
Steepest Descent 8.42
(100%) (2.9%) (17.6%) (73.6%)
648 155 247 454
Simulated Annealing 9.87
(100%) (23.9%) (38.1%) (70.1%)

4.3.4 Five-Layer Soils

Figure 12 and Table 6 show the RMS errors and their averages, corresponding to the five-layer
cases. It can be seen that the SA method is, overall, performing worse than the other two methods,
given that it yields the largest average RMS error and the minimum number of cases with an error
of 10% or less. However, SA still produces more cases than SD with an RMS error of 1% or less,
and many cases where the error is much lower than that provided by the other two methods.
Evidently, LM still performs better than the other methods for the number of best fits, but is on
par with SD for the 10% RMS error mark.

Figure 12: RMS error for five-layer cases as obtained with the three different methodologies.

Copyright © 2017 SES ltd. All rights reserved. Page 29-17

PART IV: RESEARCH WORK IN PROGRESS

Table 6: Number of five-layer cases resulting in low RMS error.

Number of Simulations
Average
RMS <= 0.1% <= 1% <= 10%
Method Total RMS RMS RMS
Error
(%) Error Error Error
768 138 282 626
Levenberg-Marquardt 5.57
(100%) (18.0%) (36.7%) (81.5%)
768 3 146 619
Steepest Descent 6.02
(100%) (0.4%) (19.0%) (80.6%)
768 63 167 506
Simulated Annealing 10.59
(100%) (8.2%) (21.7%) (65.9%)

4.3.5 Six-Layer Soils

Figure 13 shows the RMS error results obtained for the six-layer cases. This figure shows a similar
trend as that shown by the four- and five-layer cases, i.e. SA yields the largest average RMS error
of 11.56%, while SD yields 7.19% and LM exhibits the lowest average RMS error of 6.66%. As Table
7 shows, LM still produces the largest number of cases with the lowest RMS error. However, there
are some cases for which SA results produced the lowest RMS error of all three methods. Figure
14 shows that in cases 80, 83 and 85, SA results in much lower RMS errors than the other two
methods. It is worth noticing that as the number of layers increases SA yields more cases where
the fit quality is worse than the one obtained with SD. Further investigation on the selection of SA
optimum initial parameters for large number of layer cases is required to avoid this performance
degradation.

Figure 13: RMS error for six-layer cases as obtained with the three different methodologies.

Page 29-18 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 14: RMS error for six-layer cases as obtained with the three different methodologies, cases 75-85.

Table 7: Number of six-layer cases resulting in low RMS error.

Number of Simulations
Average
RMS <= 0.1% <= 1% <= 10%
Method Total RMS RMS RMS
Error
(%) Error Error Error
1024 131 289 746
Levenberg-Marquardt 6.66
(100%) (12.8%) (28.2%) (72.9%)
1024 0 129 738
Steepest Descent 7.19
(100%) (0%) (12.6%) (72.1%)
1024 48 152 624
Simulated Annealing 11.56
(100%) (4.7%) (14.8%) (60.9%)

4.3.6 Further Observations

Table 8 shows the number of cases for which SA resulted in the lowest RMS error of all three
methodologies that have been analyzed here. It also separates the number of cases where the RMS
error difference is at least 5% within the RMS error of the best methodology. Further analysis of
these significantly lower RMS error cases led us to observe that when SA yields a better result, the
soil layers presented high absolute reflection ratios (usually designated as reflection coefficients)
i.e. larger than 0.9. However, the contrary is not true, SA does not necessarily yield lower RMS
errors for soils with high reflection ratios.

Copyright © 2017 SES ltd. All rights reserved. Page 29-19

PART IV: RESEARCH WORK IN PROGRESS

Table 8: Number of cases where SA resulted in lower RMS errors.

No. of Cases Resulting in at Least

No. of
Layers 0.01% Less Error 5% Less Error
Actual Number In % Actual Number In %
2 Layers 46 9.0 2 0.39
3 Layers 161 27.9 48 8.3
4 Layers 253 39.0 10 1.54
5 Layers 214 27.9 14 1.82
6 Layers 293 28.6 47 4.6

4.3.7 Run Times

A comparison of run times is presented in Figure 15 to Figure 19, where results are sorted with
respect to those obtained with SA from smallest time to largest time, which explains the trend
seen for SA results in the plots. Because of this sorting, case numbers shown in the RMS error
figures of preceding sections do not correspond to the ones displayed in the following figures. Run
time results show that SD takes between 10 to 40 seconds for most cases, with the longest time
being around 200 seconds. LM is generally faster and takes on average 5 to 20 seconds to solve
most cases, but also exhibits an alternative trend with some cases reaching times of around 1,000
seconds and the largest run time being around 8,000 seconds. SA presents a wide range of run
times going from 10 seconds to 4,000 seconds, with most cases being in the range of 100 to 1,000
seconds. It is clear from these figures that, for most cases, SA can take several orders of
magnitudes longer in computation time compared to the other two methods. Longer computation
times were however expected, given that a deeper exploration of the parameter space is intrinsic
to the nature of this method. As a side note, the different cases were not all computed on the same
machine, thus presented results are not 100% accurate, but they still serve to show a realistic trend
in run times between the methodologies.

Figure 15: Run time for two-layer cases as obtained with the three different methodologies.

Page 29-20 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 16: Run time for three-layer cases as obtained with the three different methodologies.

Figure 17: Run time for four-layer cases as obtained with the three different methodologies.

Copyright © 2017 SES ltd. All rights reserved. Page 29-21

PART IV: RESEARCH WORK IN PROGRESS

Figure 18: Run time for five-layer cases as obtained with the three different methodologies.

Figure 19: Run time for six-layer cases as obtained with the three different methodologies.

5 Conclusion
Research on the soil resistivity inverse problem continues. The soil’s electrical structure being
such an important aspect of grounding studies, we explore the different avenues to give the best
tools to yield a successful analysis with minimal user intervention. With that perspective in mind,
a new optimization method was added in RESAP and introduced in this work. Although more
work is still required to identify the best set of meta-parameters (i.e., SA’s own parameters, such
as the temperature or the cooling factor), or subsets of meta-parameters suitable for different
configurations, the method is already useful when it proves difficult to find good starting values
Page 29-22 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

for the initial estimates of solution, or when both SD and LM fail to converge to a satisfactory RMS
error. The SA method also integrates the possibility of constraining the range of values for each
soil model parameter.
The results presented suggest that, in general, the LM method yields the lowest RMS error for
most cases, however it is also the only method which exhibits RMS errors larger than 100%. It is
also known that LM can output unrealistic or extreme soil models, so careful examination of its
results is recommended. The results also indicate that there is no single methodology capable of
assuring a good fit for all cases.
Further investigation is required to determine appropriate SA controlling meta-parameters in
order to explore the soil structure’s parameter space more efficiently. For instance, it could be that
a greater start temperature for SA is required to improve its performance. A better meta-
parameter tuning is also likely to improve the method’s runtime.
Future works include the examination of cases with noisy data. In such a setting we think the SA
algorithm may do better than SD and LM, as compared to the results obtained with the noiseless
data used in this work. Indeed, the stochastic and derivative-free aspects of SA makes it far less
sensitive to local bumps or sharp discontinuities in the search space than gradient-based
algorithms such as SD and LM. More robust and advanced variations over the SA algorithms,
implying either fuzzy logic or the simplex method, will also be explored.
From a soil analysis perspective, it is not only the obtained RMS error that should be examined to
assert the different methods’ quality: it will be the purpose of another future work to compare the
soil models obtained by the three methods to evaluate their sensitivity to the initial estimates and
their vulnerability to the soil parameter values, and how close the solutions are to the known soil
model that was generated.
The SA method will soon be introduced in the RESAP User Interface.

6 References
[1] J. Barhen, J. G. Berryman, L. Borcea, J. Dennis, C. de Groot-Redlin, F. Gilbert, et al.,
"Optimization and Geophysical Inverse Problems," Lawrence Berkeley National
Laboratory, 1999.
[2] S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, "Optimization by Simulated Annealing,"
Science, vol. 220, pp. 671-680, 1983.
[3] V. Cerny, "Thermodynamical Approach to the Traveling Salesman Problem: An Efficient
Simulation Algorithm," Journal of Optimization Theory and Applications, vol. 45, pp. 41-
51, 1985.
[4] N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, "Equation
of State Calculations by Fast Computing Machines," The Journal of Chemical Physics, vol.
21, pp. 1087-1092, 1953.
[5] N. Metropolis and S. Ulam, "The Monte Carlo Method," Journal of the American
Statistical Association, vol. 44, pp. 335-341, 1949.
[6] A. Corana, M. Marchesi, C. Martini, and S. Ridella, "Minimizing Multimodal Functions of
Continuous Variables with the "Simulated Annealing" Algorithm," ACM Transactions on
Mathematical Software, vol. 13, pp. 262-280, 1987.
[7] O. Koefoed, H. P. Patra, and K. Mallick, Geosounding Principles: Resistivity sounding
measurements: Elsevier Scientific Publishing Company, 1979.

Copyright © 2017 SES ltd. All rights reserved. Page 29-23

 UGM 2017 – CANNON BEACH, OREGON

30 MAGNETIC FIELD OF PIPE-TYPE CABLES WITH

CONSIDERATION OF
ACTUAL POSITIONS OF SUB-CABLES

Yixin Yang, Yexu Li, Marc-André Joyal and Moein Nazari

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract

In the previous versions of HIFREQ, all cable components inside pipe-type cables were assumed to be located at the
center of the pipe when computing the electromagnetic fields and the induction effects. This article describes a
technique that was introduced in this version of HIFREQ to account for the actual locations of sub-cables within
the pipe-enclosure for the computation of the electromagnetic fields and induction effects due to pipe-type cables.
The program interface and command modifications related to the new implementation are also presented. Three
practical examples are presented to demonstrate the application of the new implementation, along with
comparisons of the computation results with those obtained using the previous version of the program.

1 Introduction
A model of pipe-type (multi-core) cable was introduced in HIFREQ in [2] and found many
interesting applications since that time. The model accounts accurately for all mutual interactions
between the cable components inside the pipe-type cable. It also accounts for the interactions
between the cable components and other conductors (or cables) located outside the pipe-type
cable and for the contribution of the cable components to the electromagnetic field outside the
cable.
In previous versions of HIFREQ, however, all cable components inside pipe-type cables were
assumed to be located at the center of the pipe when computing the electromagnetic fields and
the interactions with conductors located outside the pipe. This approximation is acceptable when
the observation points (or conductors) are far away from the pipe. However, if they are close to
the pipe (as is the case, for example, when computing the electromagnetic fields or induction
effects in a GIS substation), this approximation can introduce considerable inaccuracies. For such
scenarios, the physical location of all cable components within the pipe should be considered for
improved accuracy.
This version of HIFREQ introduces an option to account for the actual locations of cable
components within the pipe-enclosure in the computation of the electromagnetic fields due to
pipe-type cables and of their interactions with other conductors. This new option requires the
definition of a reference direction with respect to which the location of the cable components is
specified. In this version of HIFREQ, this reference direction is determined automatically by the
program, based on a fixed set of rules.
This article introduces this new feature in detail. It describes the algorithm used by the program
to determine the physical location of the cable components inside pipe-type cables. It also shows
how to activate or deactivate this option from the user-interface of the program and describes the

Copyright © 2017 SES ltd. All rights reserved. Page 30-1

PART IV: RESEARCH WORK IN PROGRESS

SICL (SES Input Command Language) commands related to this new feature. Simple examples
demonstrating the validity of the implementation are also presented, along with three practical
examples (the calculation of the electromagnetic field inside a GIS substation, the calculation of
the potential induced on a pipeline running parallel to a pipe-type cable, and the calculation of
the field distribution near the jagged path of a cable).

2 Determination of the Actual Position of Cable Components

In HIFREQ, a pipe-type cable is defined as a conductor with a non-zero cable-type. Figure 1 shows
a cross-section of an example of pipe-type cable that can be modeled in HIFREQ. Generally, a
pipe-type cable encloses one or more coaxial cables, referred to as “sub-cables”. The position of a
sub-cable inside the pipe-type cable is specified by (𝑟, 𝜑), where 𝑟 is the distance between the
center of the sub-cable (point A, B, or C) and the center of the pipe-type cable (point O), while 𝜑
is the angle between the reference line OE and the line segment connecting point O and the center
of the sub-cable. These quantities are defined in the cable cross-section plane, i.e. the infinite
plane whose normal is parallel to the direction (start to end) of the conductor.

Figure 1: Specification of the positions of sub-cables when the cable enclosure is not perpendicular to the
X-Y plane.
In previous versions of HIFREQ, the positions of the sub-cables inside a pipe-type cable were only
used to compute the self and mutual impedances of the cable system; as such, only the relative
separations among the sub-cables were used, and any radial line in the cable cross section could
be taken as the reference line. For magnetic field computations, however, the absolute positions
of the sub-cables are required. Therefore, the actual location of the reference line in three-
dimensional space is important and must be determined.
In this version of HIFREQ, the location of the reference line is determined automatically by the
program, according to the following rules. In general, the reference line OE is defined as the radial
ray lying in the intersection of the cable cross section plane with the horizontal plane (X-Y plane)
which passes through point O, and points to the right starting from O when facing in the start-to-
end direction of the conductor. For the special case where the pipe-type cable is vertical, there is
no line of intersection between the cable’s cross section plane and the horizontal plane. In this

Page 30-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

case, the direction of the reference line OE is taken to be along the positive x axis, as seen in Figure
2.

Figure 2: Specification of the position of sub-cables when the cable is perpendicular to the X-Y plane.
Once the reference line is defined for a cable segment, the coordinates of the sub-cables can be
calculated using simple geometry.
The idea is that, after the node subdivision process of the conductor network, the path of a pipe-
type cable is determined by a series of nodes P1, P2, …, Pn, as shown in Figure 3. In each section,
the location of a sub-cable within the pipe is defined by (𝑟, 𝜑), where 𝑟 is the distance between the
center of the sub-cable and the center of the pipe-enclosure and 𝜑 is the angle that the line joining
the center of the sub-cable with the center of the pipe-enclosure makes with the local X axis. This
data is specified in the CABLE-TYPE command. By geometric calculations, this data can be used
to define shift vectors ⃗⃗⃗⃗
𝑑1 (𝑟1 , 𝜑1 ) , ⃗⃗⃗⃗
𝑑2 (𝑟2 , 𝜑2 ) … ⃗⃗⃗⃗
𝑑𝑛 (𝑟𝑛 , 𝜑𝑛 ) to be applied to all sub-cables in all
sections. Once all the shift vectors are calculated, the coordinates of the sub-cables can be
calculated by adding the shift vectors to the coordinates of the pipe-type cables. Those coordinates
are the actual position of the sub-cables that will be used in the field calculations.

Figure 3: The position of sub-cables in a pipe-type cable path.

Copyright © 2017 SES ltd. All rights reserved. Page 30-3
PART IV: RESEARCH WORK IN PROGRESS

3 Activating this Feature

The possibility to account for the actual positions of the sub-cables in the field calculations was
added as an option, as shown in Figure 4 & 5. Note that, by default, the program still assumes that
the sub-cables are at the center of the cable for the calculation of the magnetic field.
This feature can be activated by checking Account for Actual Sub-Cable Position in Field
Computation, in the Advanced Options screen of SESCAD (see Figure 4 in which the
corresponding option is indicated by the red rectangle). This dialog can be open from Define |
Advanced Options -> Methodology -> Cable Equations.

Figure 4: HIFREQ Advanced Options screen accommodating the option Account for Actual Sub-
Cable Position in Field Computation in SESCAD.
This option can also be activated in CDEGS-Specify (Input Toolbox) for HIFREQ. The checkbox
Account for Actual Sub-Cable Position in Field Computation (indicated by the red
rectangle in Figure 5) was added for that purpose in the HIFREQ-Methodology (Advanced)
screen under the Cable Equations tab.

Page 30-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: HIFREQ Methodology Advanced screen accommodating the option ‘Account for Actual Sub-
Cable Position in Field Computation’ in CDEGS – Specify (Input Toolbox).

The corresponding HIFREQ commands are located under the METHODOLOGY\Cable

Equations command:
METHODOLOGY
CABLE-EQUATIONS,AUTOMATIC,0,1,,1
Sub-Cable-Position, YES/NO

Copyright © 2017 SES ltd. All rights reserved. Page 30-5

PART IV: RESEARCH WORK IN PROGRESS

 Yes: Activates the use of actual sub-cable positions in the calculation of the magnetic
field.
 No: Deactivates the use of actual sub-cable positions in the calculation of the magnetic
field.
Note that the computations could be a little bit slower in the presence of cables when Account
for Actual Sub-Cable Position in Field Computation is checked. More importantly, the
computation of the electric field and potential due to pipe-type cables can become less accurate
when this option is activated.

4 Validation
The following examples verify the validity of the new algorithm and show how the results obtained
when the actual location of the sub-cables is used in the calculations differ from those obtained
when the sub-cables are assumed to be located at the center of the pipe-type cable.

Comparison of Cable Model with Multiple Conductor Model

A three phase pipe-type cable is modeled in three scenarios. In Scenario 1, the cable is modeled
using the new cable model of HIFREQ, which accounts for the actual positions of the three sub-
cables. In Scenario 2, the three cores of the pipe-type cable are modeled as three physical
conductors; the pipe enclosure is ignored. In Scenario 3, the cable model of Scenario 1 is used, but
the option Account for Actual Sub-Cable Position in Field Computation is not selected.
Scenario 1
In this scenario, the new cable model of HIFREQ is used to model a pipe-type cable with an overall
radius of 0.90 m. The sub-cables are located at a distance of 0.72 m from the center of the pipe-
type cable, at angles of 90, 210, and 330 degrees. The pipe enclosure is disconnected (floating) at
both ends to minimize the induced current on its structure. The cores are energized with three
balanced phases: 1000 0° A; 1000 120° A; 1000 240° A. The computations were carried out for
both a copper and a steel pipe-enclosure. In order to eliminate the influence of the earth’s surface,
the Infinite soil model with a soil resistivity 109 Ohm-m is used in the computations. The
computed magnetic field outside the enclosure of the cable, in the plane perpendicular to the cable
axis, is shown in Figure 6.
Magnetic Fields/Resultant (Total) Field [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

LEGEND

Maximum Value : 590.563

-102.0
Minimum Value : 116.373

590.56

543.14
-101.0
495.73
Z AXIS (METERS)

448.31

-100.0 400.89

353.47

306.05
-99.0 258.63

211.21

163.79
-98.0
-2 -1 0 1 2

Y AXIS (METERS)
Total Magnetic Field Magn. (Amps/M)

Figure 6: Magnetic field around a pipe-type cable computed when accounting for the actual location of
sub-cables in the pipe-type cable.

Page 30-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Scenario 2
In this scenario, the model is the same as in Scenario 1 but without the pipe-enclosure modeled
(see Figure 7). The goal here is to compare the magnetic field results with those of Scenario 1
without the effects of eddy currents.

Figure 7: Illustration of the model with three conductors.

As shown in Figure 8, the computation results of the magnetic field are the same as those of
Scenario 1. This validates the new computation algorithm used to Account for Actual Sub-
Cable Position in Field Computation. Magnetic Fields/Resultant (Total) Field [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

LEGEND

Maximum Value : 590.563

-102.0
Minimum Value : 116.374

590.56

543.14
-101.0
495.73
Z AXIS (METERS)

448.31

-100.0 400.89

353.47

306.05
-99.0 258.63

211.21

163.79
-98.0
-2 -1 0 1 2

Y AXIS (METERS)
Total Magnetic Field Magn. (Amps/M)

Figure 8: Magnetic field around three cable conductors without pipe-enclosure.

Scenario 3
In this last scenario, the same model as that of Scenario 1 is used, but the option Account for
Actual Sub-Cable Position in Field Computation is unchecked. This scenario actually
reproduces the results of the previous versions, which assumes that the currents are all located at
the center of the pipe. The computation results in Figure 9 show that the total magnetic field
outside the pipe-enclosure is circularly uniform and almost zero. This is because, in the
computations, the three core conductors are “overlapping” at the center of the cable with the same
magnitude current but 120° phase difference between them.

Copyright © 2017 SES ltd. All rights reserved. Page 30-7

PART IV: RESEARCH WORK IN PROGRESS
Magnetic Fields/Resultant (Total) Field [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

SPOT LEVELS x 1.E-3

Maximum Value : 0.848

-102.0
Minimum Value : 0.848

0.85

0.85
-101.0
0.85

Z AXIS (METERS)
0.85

-100.0 0.85

0.85

0.85
-99.0 0.85

0.85

0.85
-98.0
-2 -1 0 1 2

Y AXIS (METERS)
Total Magnetic Field Magn. (Amps/M)

Figure 9: Magnetic field around a pipe-type cable computed without considering the actual positions of
the sub-cables (previous algorithm).
From the computation results presented in Figure 6 and Figure 8, it can be seen that the new
algorithm accounts for the precise locations of the three individual conductors inside the pipe-
enclosure. In Figure 9, the computed magnetic field is almost zero. This result was expected
because when all the sub-cables inside a pipe-type cable are assumed to be located at the center
of the pipe-type cable, the magnetic fields due to the balanced currents are canceling out.

Magnetic Field along a Profile from the Center of the Cable to the Outside
In this case, an observation profile is placed horizontally along a radial line from the outer surface
of a pipe-type cable to 10 m outside the cable. The outer radius of the pipe-type cable is 0.901 m.
The cable contains three cores in a triangular pattern, at distances of 0.8 m from the center of the
pipe-cable. The energization remains the same as for the previous case: three balanced phase
currents: 1000 0°; 1000 120°; 1000 240° A. A uniform soil model with a soil resistivity of
100 Ohm-m is used.

Figure 10: Observation profile outside a pipe-type cable.

Page 30-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Magnetic Fields/Resultant (Total) Field LEGEND

250

Profile Number 1.

200

150

Total Magnetic Field Magn. (Amps/M)

100

50

0
0 5 10 15 [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

Distance from Origin of Profile (m)

Figure 11: The computation results for the magnetic field along the computation profile.
Figure 11 shows the computation results for the magnetic field along that profile. The field
decreases to a very small value when the observation point is located at a distance greater than 10
times the radius or the shift distance of the pipe-type cable. For observation points located at such
distances from the axis of the pipe-type cable, essentially the same results are obtained with and
without the Account for Actual Sub-Cable Position in Field Computation option
activated.

Effects of Eddy Currents in the Pipe Wall

At this stage of the development, the effects of eddy currents in the wall of the enclosure of the
pipe-type cable are not taken into account when computing the electromagnetic fields. However,
it is still interesting to estimate the importance of these effects. In this section, we use a wire mesh
to simulate the wall of the enclosure of the pipe-type cable. The induced currents in the wires can
be considered as an approximation of the eddy currents in the wall of the enclosure, which can be
used to calculate the magnetic field near the pipe.

Figure 12: Use of a wire mesh to simulate the enclosure of the pipe-type cable.
In Scenario 2 of Section 0, three conductors were used to represent the cores inside the pipe-type
cable, but the enclosure of the cable was not modeled. The maximum magnetic field calculated
for that case was found to be 590.56 A/m. In this section, a wire-mesh structure representing the
wall of the enclosure is added to this model. The radius of the wires representing the enclosure
wall is 1.0 mm. In Table 1, the computation results for two different materials (copper and steel),
are listed in comparison with the model without pipe-enclosure.

Copyright © 2017 SES ltd. All rights reserved. Page 30-9

PART IV: RESEARCH WORK IN PROGRESS

Table 1: Computed maximum magnetic field.

Pipe Type Maximum Magnetic Field (A/m)

No Pipe 590.56

Steel Pipe 590.01

Copper Pipe 582.42

Table 1 shows that eddy currents have a small effect on the magnetic field around the pipe-type
cable (for this scenario) and always tend to reduce the maximum value of the magnetic fields. The
eddy currents in a copper pipe have more effect on the magnetic field than those in a steel pipe.
The reason could be that copper has a smaller resistivity than steel, leading to a higher level of
eddy currents that can be induced in the copper pipe enclosure.

5 Practical Examples
Magnetic Fields along a Jagged Pipe-Type Cable
Input file: hi_Field of Jagged PipeType Cable.f05
The case illustrated in Figure 13 shows the magnetic field along a complicated 3D path followed
by a pipe-type cable. The locations of the peaks of the computed fields correspond to the positions
of the sub-cables, as expected.
Magnetic Fields/Resultant (Total) Field [ID:3Conductors_CirclePr @ f=60.0000 Hz ]

LEGEND

Maximum Value : 3115.931

-3
Minimum Value : 159.505

5 x E+3
3 x E+3
-5
2 x E+3
Relative Distance Along Surface (METERS)

1 x E+3
9 x E+2
-7
8 x E+2
7 x E+2
5 x E+2
-9
5 x E+2
3 x E+2
2 x E+2
-11
-79 -77 -75 -73 -71

Relative Distance Along Profile (METERS)

Total Magnetic Field Magn. (Amps/M)

(a) (b)
Figure 13: Magnetic fields along a 3D jagged pipe-type cable. (a) The jagged path of a pipe-type cable and
the observation profiles in SESCAD. (b) The computed magnetic field along the path.

Magnetic Fields in a GIS Substation

Input file: hi_GIS Substation.f05

Page 30-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

In this case, a GIS substation is modeled and the magnetic fields in the substation area are
computed in two scenarios: with the option Account for Actual Sub-Cable Position in Field
Computation selected, and with this option unselected.
The GIS contains three cores and the pipe-enclosure is made of aluminum. The radius of the pipe
is 0.9 m and the pipe-enclosure is located 1.1 m above the ground. The pipe enclosure is grounded
to the grounding grid at both ends. As was done in the previous scenarios, the GIS is energized
with balanced three phase currents of 1 kA. The observation profile is on the ground surface, inside
the substation. The computed magnetic field obtained by accounting for the actual positions of
the sub-cables is provided in Figure 14 whereas the field without accounting for the actual sub-
cable position is given in Figure 15. By comparing the two figures, it can be seen that the maximum
magnetic field is much larger when the actual positions of sub-cables are accounted for.
Magnetic Fields/Resultant (Total) Field [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

LEGEND

Maximum Value : 301.646

Minimum Value : 0.175

301.65
70 271.50

241.35
Y AXIS (METERS)

211.20

181.06

20 150.91

120.76

90.62

60.47

30.32
-30
-75 -25 25

X AXIS (METERS)
Total Magnetic Field Magn. (Amps/M)

Figure 14: Magnetic field in a GIS substation, accounting for the actual positions of the sub-cables.

Magnetic Fields/Resultant (Total) Field [ID:OneHorizontal_3Condu @ f=60.0000 Hz ]

SPOT LEVELS x 1.E-3

Maximum Value : 0.652

Minimum Value : 0.294E-01

0.65
70 0.59

0.53
Y AXIS (METERS)

0.47

0.40

20 0.34

0.28

0.22

0.15

0.916E-01
-30
-75 -25 25

X AXIS (METERS)
Total Magnetic Field Magn. (Amps/M)

Figure 15: Magnetic field in a GIS substation, without accounting for the actual positions of the
sub-cables.

Induced Potentials on a Pipeline Parallel to a Pipe-Type Cable

Input file: hi_Induction_DuetoPTcable.f05
In this case, a 5 km steel pipeline is parallel to a pipe-type cable at a separation distance of 1.6 m.
The radius of the pipe is 0.5 m and its thickness is 5 mm. The pipe-type cable is energized with
balanced three phase currents of 1 kA. A uniform soil model with a soil resistivity 100 Ohm-m is
Copyright © 2017 SES ltd. All rights reserved. Page 30-11
PART IV: RESEARCH WORK IN PROGRESS

used. Again, the induced potentials on the pipeline are computed using the options: (1) Account
for the actual positions of the sub-cables inside the pipe-type cable; (2) Do not account for the
actual positions of the sub-cables.
Figure 16 shows the induced potentials on the pipeline, which can reach 90 volts when accounting
for the actual positions of sub-cables. The induced potentials are almost zero if the actual positions
of the sub-cables inside the pipe-type cable are not accounted for in the computation of the
magnetic field.

Figure 16: Comparison of the induced potentials computed with and without the Account for Actual
Sub-Cable Position in Field Computation option.

Future Developments
In the next phase of the development, the calculation of the electric field due to pipe-type cable
when the option Account for Actual Sub-Cable Position in Field Computation is selected will be
improved, at which point this option will be selected by default. The contribution of the eddy
currents in the pipe enclosure to the field calculation could also be implemented. The same
approach could also be employed in the calculation of the magnetic fields in the TRALIN module.

References
[1] “Cable Model in HIFREQ,” in CDEGS Users’ Conference Proceedings, San Francisco, California, USA,
2008.

[2] “Pipe-Type Cable Model in HIFREQ,” CDEGS Users’ Conference Proceedings, Myrtle Beach, South
Carolina, USA, 2011.

[3] “Specifying Data and Displaying Results for Pipe-Type Cable Models in HIFREQ,” CDEGS Users’
Conference Proceedings, Rock Creek Resort, Montana, USA, 2012.

Page 30-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

31 NEW MULTI-LAYER HEMI-SPHEROIDAL SOIL MODEL

Amir Hajiaboli, Simon Fortin and Farid P. Dawalibi

Safe Engineering Services and Technologies, ltd.
Email: [email protected], Web Site: www.sestech.com

Abstract

This article presents a newly developed soil model for the numerical analysis of grounding systems in the vicinity
of multilayer hemi-spheroidal soil heterogeneities. A brief theoretical overview of the spheroidal coordinate system
is presented, together with the command structure required to define this soil model in the MALZ module of SES
Software. Several benchmark test cases against existing soil models are presented, along with some analytical
benchmarks. The results obtained from these benchmark tests validate the accuracy of this newly developed model.

1 Introduction
Two years ago a hemi-spheroidal soil model was introduced in the MALT and MALZ modules of
SES Software [1, 2]. This type of soil model, where a single hemi-spheroidal interface separates
an inner region with a different resistivity than that of the outer native soil, can represent local
variations of resistivity that can occur, for instance, around a lake. This article presents a
significantly improved version of this soil model which can accommodate multiple confocal hemi-
spheroidal layers of soil. Using such a multilayer hemi-spheroidal soil can extend the application
of the hemi-spheroidal soil to more realistic and accurate cases such as sedimentary layers
underneath a lake or a locally multilayered soil.
Unlike the previous work [1, 2] where the analytical form of Green’s function for the 2-layer hemi-
spheroidal soil has been found and then numerically solved, in this new implementation no
analytical (closed form) Green’s function formula is sought. The Green’s function is computed
numerically by building and solving the linear system of unknown coefficients to consistently
preserve the boundary conditions (continuity of normal component of current and continuity of
potential) at all the spheroidal interfaces simultaneously. As such, there is a significant difference
in the core of the computational algorithms used in these two methods. Without discussing the
details of the theoretical background, in this paper, we mostly present and validate the integration
of this newly developed soil into MALZ module of SES Software.
The extension to the MALT module and an appropriate user interface for this soil module are also
under development.

2 Oblate Spheroidal Coordinate System

The oblate spheroidal soil model is obtained by rotating an ellipse along its minor axis [3]. Figure 1
shows this coordinate system with its associated parameters. Mathematically, the oblate spheroid
is related to the Cartesian coordinate system using:
𝑥2 + 𝑦2 𝑧2
+ = 1/4
𝑐 2 (𝜁 2 + 1) 𝑐 2 𝜁 2
𝑐 𝑐 𝑐
𝑥 = √(1 + 𝜁 2 )(1 − 𝜂 2 )cos𝜑, 𝑦 = √(1 + 𝜁 2 )(1 − 𝜂 2 )sin𝜑, 𝑧 = 𝜂𝜁
2 2 2

Copyright © 2017 SES ltd. All rights reserved. Page 31-1

PART IV: RESEARCH WORK IN PROGRESS

where, 𝑐 is the focal distance, 𝜁 ∈ [0, ∞), 𝜂 ∈ [−1,1] and 𝜑 ∈ [0, 2𝜋]. Looking at the special case
where 𝑦 = 0, the major axis (a1) and minor axis (a2) of the ellipse formed by the projection of the
spheroidal surface in the XZ plane (i.e., the Radius (R) and Depth (D) of the spheroid) can be
written as:
𝑐 𝑐
𝑎1 = 𝑅 = √𝜁 2 + 1, 𝑎2 = 𝐷 = 𝜁
2 2

Figure 1: Oblate coordinate system and its relevant parameters for defining the coordinate system.

3 Confocal Multilayer Oblate Hemi-Spheroidal Soil

Based on the definition of oblate spheroidal coordinates, a confocal multilayer oblate hemi-
spheroidal soil consists of several oblate hemi-spheroidal interfaces, with constant ζ. As shown in
Figure 2, the confocal points for all those hemi-spheroidal interfaces are the same.

Figure 2: Multi-Layer Hemi-Spheroidal Soil. The confocal points highlighted in red are the same for all
the spheroidal interfaces. The command to specify the layers takes the ratio of the radius of the first layer
to its depth and the difference between the radius of each layer and that of the previous layer.

Page 31-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

3.1 Command Definition

Currently, the multi-layer spheroidal soil can be specified only in the MALZ module, in command
mode. The commands shown in Table I are used to specify a Multi-Layer Spheroidal Soil. Also, in
the SOIL-TYPE command, a new qualifier (Multi-Spheroidal) has been added to specify the Multi-
Layer Spheroidal Soil. Figure 3 shows a sample command structure for defining a 3-layer
spheroidal soil.
Table I: The commands for a Multi-Layer Spheroidal Soil
Command Qualifier Variables Description
SOIL-TYPE Multi- X-Center, Y- Specifies the location of the center of the hemi-spheroidal
Spheroidal Center surfaces for a Multi-Layer Spheroidal Soil.
Multi- Specifies the aspect ratio of the first spheroidal interface,
AspectRatio
Spheroidal i.e. the ratio of the radius of the first layer to its depth.
Specifies the characteristics of one layer of the multilayer
LayerNumber,
spheroidal soil. With the exception of the first layer, layers
Radius-
are defined by specifying the difference between the radius
Difference,
LAYER of the current layer and that of the previous layer. For the
Resistivity,
first layer, this value represents the actual radius of the
Permeability,
layer. The last layer does not have any associated radius
Permittivity
since its radius is infinite.

Sample commands describing a three layer hemi-spheroidal soil are shown below.
SOIL-TYPE, Multi-Spheroidal, 5,0
Multi-Spheroidal, 2
Layer, 1, 100,1000,1,1,
Layer, 2,10,100,1,1,
Layer, 3,,100,1,1,

4 Benchmarks
Figure 4 shows the grid configuration that was considered in most of the benchmarks in this study.
It is a square grid with a side length of 30 m subdivided equally into four meshes in the X and Y
directions. The grid is located 0.5 m below the earth surface.

Figure 3: Grid configuration used in the most cases considered in this section.

Copyright © 2017 SES ltd. All rights reserved. Page 31-3

PART IV: RESEARCH WORK IN PROGRESS

4.1 Comparison with a 2-Layer Spheroidal Soil

In this benchmark, we compare the results obtained using the newly developed multilayer
spheroidal soil (in the limiting case where only two layers are modeled) versus the results obtained
using the previously developed 2-layer spheroidal soil model.
The spheroidal soil has a radius of 50 m and depth of 25 m. The resistivity of the inner layer is
100 Ω-m and that of the outer layer is 500 Ω-m. The commands used in each method are shown
in Table II.
Table II: Command Structure for 2 Layer Spheroidal Soil and Multi-Layer Spheroidal Soil

Soil Type Command Structure

2 Layer Spheroidal Soil SOIL-TYPE,SPHEROIDAL,0,0
SPHEROIDAL-SOIL,
50,25,100,1,1,500,1,1
Multi-Layer Spheroidal Soil SOIL-TYPE,Multi-Spheroidal,0,0
Multi-Spheroidal,2
Layer, 1, 50,100,1,1,
Layer, 2,,500,1,1,

The observation points are located on the surface of the earth along the X-direction. Three
different scenarios are considered: 1) The grid is inside the hemi-spheroidal volume, 2) The grid
crosses the boundary of the hemi-spheroidal volume, 3) The grid is outside the hemi-spheroidal
volume.
Although the numerical implementation of these two methods is significantly different, as it can
be seen in Figures 4 to 6, the results are in very good agreement.

Figure 4: The grid is inside the hemi-spheroidal interface.

Page 31-4 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 5: The grid crosses the hemi-spheroidal interface.

Figure 6: The grid is outside the hemi-spheroidal interface.

4.2 Comparison with Uniform Soil

In this benchmark, a hemi-spheroidal volume where the spheroid and native soils have the same
resistivity of 500 Ω-m has been modeled and the results are compared with the results obtained
from a uniform soil model with resistivity of 500 Ω-m. The hemi-spheroidal soil has a radius of
50 m and a depth of 25 m. The command structure to represent the multi-layer spheroidal soil is
similar to that shown in Table II and used in the previous benchmark. The observation points are
located on the surface of the earth along the X-direction. Similar to the previous benchmark, three
scenarios for the location of the grid relative to the spheroid interface have been modeled. As
shown in Figures 7 to 9, a good agreement has been achieved.

Copyright © 2017 SES ltd. All rights reserved. Page 31-5

PART IV: RESEARCH WORK IN PROGRESS

Figure 7: The grid is inside the hemi-spheroidal interface with equal resistivity inside and outside.

Figure 8: The grid crosses the hemi-spheroidal interface with equal resistivity inside and outside.

Figure 9: The grid is outside the hemi-spheroidal interface with equal resistivity inside and outside.

Page 31-6 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

4.3 Comparison with the Finite Volume Soil Model

Using four trapezoidal soil volumes, we have modeled an approximate hemi-spheroidal soil model
as shown in Figure 10. Although the cross-sectional view in the XY plane at z=0 is a square with a
side length of 40 m, we believe that this is a good approximation of a hemi-spheroidal volume
where the cross-section is a circle with a diameter of 24.14 m. This radius has been calculated as
the average of the radii of the inscribed and circumscribed circles for the square cross-section of
the Finite Volume with the earth interface which are 20 m and 28.28 m respectively. The
resistivity of the soil inside the hemi-spheroidal model is 1000 Ω-m and the native soil resistivity
is 100 Ω-m. The command structure for the hemi-spheroidal soil is similar to the command
structure shown in Table II for multi-layer spheroidal soil, with an inner resistivity of 1000 Ω-m
and an outer resistivity of 100 Ω-m. Again, we consider three different scenarios where the grid is
located inside the hemi-spheroidal volume, on the boundary of the hemi-spheroidal volume, and
outside the hemi-spheroidal volume. The graphs in Figures 11 to 13 show the earth potentials for
the observation points along the x-axis on the surface of the earth for these different scenarios.

Figure 10: Cross section of the finite volumes representing the hemi-spheroidal soil model.

Figure 11: Earth potentials on the soil surface when the grid is located inside the hemi-spheroidal soil.

Copyright © 2017 SES ltd. All rights reserved. Page 31-7

PART IV: RESEARCH WORK IN PROGRESS

Figure 12: Earth potentials on the soil surface when the grid is crossing the hemi-spheroidal soil volume.

Figure 13: Earth potentials on the soil surface when the grid is outside the hemi-spheroidal soil volume.

4.4 Comparison with the Hemi-Spherical Soil Model

We have also modeled a hemi-sphere using the hemi-spheroidal soil model with practically the
same depth and radius. To do this, cases were built using a 3-layer hemi-spherical soil model with
resistivities of 500 Ω-m, 100 Ω-m and 1000 Ω-m and radii of 40 m, 60 m. Similar cases were built
using a hemi-spheroidal soil with an aspect ratio close to 1.005. The command structure for hemi-
spherical and hemi-spheroidal soil models are shown in Table III. As for the previous cases, we
have considered three scenarios where the grid shown in Figure 3 is inside the hemi-spheroid, on
the boundary of the hemi-spheroid and outside the hemi-spheroid. The graphs in Figures 14 to 16
show the earth potentials for different observation points along the x-axis at the surface of the
earth, which indicate a good agreement between the results.

Page 31-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Table III: Command Structure for Different Soil Types.

Soil Type Command Structure

Spherical Soil SOIL-TYPE,SPHERICAL,0,0
HORIZONTAL
LAYER,TOP,500,40,1,1
LAYER,CENTRAL,100,60,1,1
LAYER,BOTTOM,1000,,1,1
Multi-Layer Spheroidal Soil SOIL-TYPE,Multi-Spheroidal,0,0
Multi-Spheroidal,1.005
Layer, 1, 40,500,1,1,
Layer, 2,20,100,1,1,
Layer, 3,,1000,1,1,

Figure 14: Comparison between the hemi-spheroidal and hemi-spherical soil models when the grid is
inside the hemi-spheroidal volume.

Figure 15: Comparison between the hemi-spheroidal and hemi-spherical soil models when the grid is
crossing the boundary of the hemi-spheroidal volume.

Copyright © 2017 SES ltd. All rights reserved. Page 31-9

PART IV: RESEARCH WORK IN PROGRESS

Figure 16: Comparison between the hemi-spheroidal soil and hemi-spherical soil when the grid is outside
the hemi-spheroidal volume.

4.5 Comparison with a Plate

In this benchmark, the resistance of a plate placed inside a uniform soil is calculated using the
newly developed multilayer spheroidal soil model and compared with the formula for the
resistance of a plate, shown in Eq. 1.
𝜌
𝑅𝑝𝑙𝑎𝑡𝑒 = 4𝑎 (1)

Here, ρ is the resistivity of the uniform native soil and a is the radius of the plate. In this example,
a=500 m and ρ =1000 Ω-m which results in Rplate = 0.5 Ω.
The numerical model of the plate was created using a command structure similar to that shown
in Table II. The multilayer hemi-spheroidal soil model has an inner resistivity of 0.001 Ω-m, outer
resistivity of 1000 Ω-m, radius of 500 m and aspect-ratio of 25000 (which results in the first layer
depth of 0.02 m) to represent a circular plate. A small grid located inside the inner spheroid has
been energized.
The computed resistance of this grid is 0.5272 Ω which is in good agreement with the theoretical
value of Rplate = 0.5 Ω.

4.6 Comparison with the Theoretical Resistance of a Hemi-Spheroidal

Soil Model
In this section, we compare the theoretical value of the resistance of a multilayer hemi-spheroidal
soil model with the value calculated based on the newly developed soil model in MALZ. The
resistivity inside the soil volume is set to a very low value (0.02 Ω-m) in order to maintain a quasi-
equipotential surface on the boundary of the hemi-spheroid. Assuming that the multilayer hemi-
spheroid consists of N+1 layer (N spheroid interfaces), Eq. 2 shows the total resistance seen from
the inner layer. The theoretical value is calculated based on the exact value obtained by integrating
the expression of the elemental resistance of a field tube over the hemi-spheroid layer delimited
by two spheroidal interfaces.
1 𝑛 𝜌𝑛+1 𝜋
𝑅 = (C𝜋) ∑𝑘=2 𝜌𝑘 [Arctan(𝜁𝑘 ) − Arctan(𝜁𝑘−1 ) + ( C𝜋
) (2 − Arctan(𝜁𝑛 )] (2)

Page 31-10 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Here,𝜁𝑘 is the value of ζ at interface k, 𝜌𝑘 is the resistivity of layer k, and C is the focal distance. 𝜁𝑘
can be computed using the following expression:
𝐷𝑘
𝜁𝑘 =
√𝑅𝑘 2 − 𝐷𝑘 2

where Rk and Dk are the radius and depth of the spheroidal interface k, respectively.
The results for various depths and radii are given in Table IV.
Table IV: Calculated resistance versus theoretical value

Number Calculated Theoretical

Aspect Layer
of Interface Radius Resistance Resistance
Ratio Resistivities
Layers (Ω) (Ω)
2 2 50 0.02,500 1.9308 1.9245
1000,1100,1150,1260 0.02,50,100,1000,
5 3.5 0.7472 0.7407
5000
500,530,580,594,634, 0.02,500,100,1000,
10 2.5 654,704,714,739 300,5000,0.1,100, 1.2757 1.2689
1000,5000

4.7 Example of a Surface Color Spot Plot of the Earth Potentials

In this section, we show the horizontal and vertical earth potential computed in a multilayer hemi-
spheroidal soil. The multilayer hemi-spheroidal soil is defined using the commands shown below.
SOIL-TYPE,Multi-Spheroidal,0,0
Multi-Spheroidal,2
Layer, 1, 50,100,1,1,
Layer, 2,10,500,1,1,
Layer, 3,5,50,1,1,
Layer, 4,,1000,1,1,
The grid, as shown in Figure 3, has a length of 30 m and is located at 0.5 m below the earth surface.
The color spot plots of the earth potentials in the XY and XZ plane that were obtained using the
SESResultsViewer tool are shown in Figures 17 and 18, respectively.

Copyright © 2017 SES ltd. All rights reserved. Page 31-11

PART IV: RESEARCH WORK IN PROGRESS

Figure 17: Horizontal Earth potential plot in the XY plane.

Figure 18: Vertical earth potential plot in the XZ plane.

5 Conclusion and Future Work

This article introduced a significant enhancement to the capabilities of the MALZ computation
modules to model multilayer hemi-spheroidal soils. In this model, the radii of the layers and the
ratio of the radius of the first layer to its depth can be defined arbitrarily.

Page 31-12 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

The model has been tested against existing soil models in MALZ using several benchmarks
presented in this paper. The computation results confirm the accuracy of the developed soil
model. Furthermore, the resistance of a plate in a uniform soil and the resistance of a multilayer
spheroid have been computed using this model and compared with analytical formulae. Again,
the agreement indicates the accuracy of the developed model.
Future work includes the development of an appropriate interface for defining the multilayer
spheroidal soil, and also the integration of this feature in the MALT computation engine. Also an
investigation related to the acceleration of the computation code and the possibility of using an
automated and adaptive selection of the appropriate number of terms required to achieve the
desired accuracy will be carried out.
The current model was developed for oblate hemi-spheroidal soil volumes. The extension of the
method to prolate hemi-spheroidal soil volumes, where the depth is larger than the radius, can
also be considered using similar mathematical approaches with minor modifications.

6 References

1. A. Hajiaboli, S. Fortin, F. P. Dawalibi, P. Zhao and A. Ngoly, "Analysis of Grounding Systems in the
Vicinity of Hemi-Spheroidal Heterogeneities," IEEE Transactions on Industry Applications, Vol. 51,
No. 6, November/December 2015, p. 5070 - 5077.
2. A. Hajiaboli, S. Fortin and F. Dawalibi, "Numerical Techniques for the Analysis of HVDC Sea
Electrodes," IEEE Transactions on Industry Applications, Vol. 51, No. 6, November/December 2015, p.
5175 - 5181.
3. W. R. Smythe, Static and Dynamic Electricity, McGraw-Hill, 2nd edition 1950.

Copyright © 2017 SES ltd. All rights reserved. Page 31-13

 UGM 2017 – CANNON BEACH, OREGON

32 COMPARISON BETWEEN PLATE AND EQUIVALENT

WIRE MODELS IN GROUNDING APPLICATIONS
Peter Zhao, Jie Liu, Simon Fortin and Farid P. Dawalibi

Safe Engineering Services & Technologies ltd.

Email: [email protected], Web Site: www.sestech.com

Abstract
This article compares the grounding performance of various structures made of metallic plates versus equivalent
structures made of conducting wires. The difference in the ground resistance, the generated scalar potential and
the touch voltage near a square or rectangular metallic plate and an equivalent wire-mesh model are studied, and
the effects of the mesh size and of the area, shape and depth of the electrode are illustrated. The behavior of a coated
pipe modeled either as a single coated wire of large radius or as a collection of coated plates is also investigated. In
one example, the current distribution around a perfectly coated pipe near an energized tower footing is illustrated
for both modeling approaches. The results show that the pattern and magnitude of current flow differ markedly for
the two approaches: when the pipe is modeled with coated plates, the current tends to bypass the coated pipe while
the presence of the pipe has little effect on the current distribution around the pipe when it is modeled as a single
coated wire. The transferred potential between with two conductors located on opposite sides of a perfectly coated
pipe and running parallel to the pipe is also investigated, using both modeling approaches. The results show that
due to the “blocking” effect that occurs when modeling the pipe with coated plates, the transferred potential is
smaller in that scenario than when modeling the pipe as a wire with a large radius, and that the effect gets larger as
the conductors move closer to the coated pipe.

1 Introduction
In recent years, the possibility to model bare or coated plates of arbitrary shape and orientation
has been introduced in the MALT module for the Perfect, Uniform and Horizontal Multilayer soil
models [1-4]. More recently, this feature has been introduced in the HIFREQ module (for bare
metallic plates only) for the Infinite, Uniform and Horizontal 2-Layer soil models [5-6]. This
feature provides greater flexibility for modeling realistic systems (such as those including wells,
storage tanks, cars and trucks, flood gates, building walls or foundations, etc.) and enhances the
capabilities of the MALT and HIFREQ modules [7-9].
In grounding and interference studies, the physical system should be modeled as close as possible
to reality. If some elements of the system consist of bare or coated metallic surfaces, they can be
modeled directly with the plate option. Plates can also be used to approximate collections of wires.
For example, they can be used to represent dense conductor meshes (such as rebar in foundations)
to simplify the model and reduce the computation run time. They can also be used as the limiting
case of a grounding grid to predict the best results that could be achieved in the occupied area.
The reverse situation can also occur: objects that have a large surface area, such as large coated
or bare metallic pipes, are quite often modeled as wires of large radius, with or without a coating
layer. In such cases, the thin wire approximation used for the wire representing the pipe can fail
to account fully for proximity effects when other elements or observation points are located close
to the pipe.
This article compares models built using plates and with wires for some typical applications.

2 Comparison between Plates and Equivalent Wire-Mesh

Grids
The differences in the performance of a metallic plate and of an equivalent wire-mesh grid
occupying the same area is affected by many factors, such as the mesh size of the grid, the shape,
Copyright © 2017 SES ltd. All rights reserved. Page 32-1
PART IV: RESEARCH WORK IN PROGRESS

size and location of the occupied area, etc. The effects of these factors are investigated in this
section.
Unless indicated otherwise, all examples of this section use a 10 m X 10 m horizontal square thin
plate or grid as a model. The thickness of the plate is 0.001 m, which is the same as the diameter
of the conductors in the grid. A large number of patches (100 X 100) is specified in the plate
models, in order to maximize the computation accuracy. The plate (or grid) is centered at (0, 0).
A linear observation profile extends from the center (0, 0) of the system to (10 m, 10 m) with a
step of 0.01 m is specified on the earth surface. The current injected in the plate (or grid) is 1 kA.
The calculated ground impedance and touch voltages are compared, for a typical uniform soil
model with a resistivity of 100 ohm-m.

2.1 Ground Impedance

The relative difference (normalized with respect to the plate results) in the grounding resistance
computed with the plate and the equivalent grid model is shown in Figure 1, as a function of mesh
size for various grid depths.

Figure 1: Difference in the grounding resistance between the plate and grid model, for various grid depths.
The differences are expressed as a percentage relative to the plate results.
From Figure 1, it can be seen that the difference in the ground resistance computed with the plate
model and that computed with the grid model increases significantly with the mesh size. For
instance, at a depth of 0.1 m, the difference increases from 0.52% to 8.73% when the mesh size
changes from 0.1 m to 1 m. Also, the difference in the ground resistance is larger for deeper grids.
This indicates that for shallow depths, a dense rebar mesh can be accurately represented using a
plate. On the other hand, for a typical grounding grid with a large mesh size (up to 10 m, or even
more) buried at a depth of 0.5 m, using a plate to predict the performance of the grid could greatly
overestimate its performance.
Figure 2 illustrates the effect on the difference in the performance of the plate and equivalent grid
models when the area occupied by the plate and grid increases from 10 m X 10 m to 20 m X 20 m.

Page 32-2 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 2: Difference in the grounding resistance between the plate and grid model, for different grid sizes.
The differences are expressed as a percentage relative to the plate results.
From Figure 2, it can be seen that the difference in the ground resistance computed with the two
models decreases with the size of the plate or grid.
The shape of the area occupied by a plate or grid can also affect the discrepancy in the results
computed with the two models. Figure 3 compares the performance of plates and grids of square
(10 m X 10 m) and rectangular (5 m X 20 m) shapes buried at 0.1 m and 0.5 m.

Figure 3: Difference in the grounding resistance between the plate and grid model, for different grid
shapes. The differences are expressed as a percentage relative to the plate results.

Copyright © 2017 SES ltd. All rights reserved. Page 32-3

PART IV: RESEARCH WORK IN PROGRESS

From Figure 3, it can be seen that the difference between the results obtained with the plate and
wire-grid models increases when the shape of the plate or grid is elongated.

2.2 Potential and Touch Voltage

For the square plate (10 m X 10 m) and its equivalent grid, the generated scalar potential and
touch voltage along the specified profile are shown in Figure 4 and Figure 5, respectively.

Figure 4: Potential along the profile on the earth surface.

From Figure 4, it can be seen that the potential generated by a plate is quite flat within area
occupied by the plate along the specified profile on the earth surface. However, the potential along
the profile generated by grids changes with the mesh size and exhibits “ripples” across the meshes.
At a depth of 0.1 m, the ripples become larger when the mesh size increases, especially near the
edge of the grids. This effect is alleviated when the grids are deeper in the soil. At a depth of 0.5
m, the ripples become quite small.

Figure 5: Touch voltage along the profile on the earth surface.

Page 32-4 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

Similar ripple phenomena can also observed for the touch voltage along the profile as computed
using the grid models, as shown in Figure 5.

3 Comparison of Coated Pipe Modeled by Coated Plates or

Wire
In grounding and interference studies, pipelines are quite often being modeled as conductors
(wires) with a large radius, placed along the axis of the pipe. However, as a result of the thin-wire
approximation used in the MALT, MALZ, and HIFREQ modules when modeling wires, the
coating on pipes modeled in this way has no effect on current flowing outside the pipe. To account
for the effect of the coating on currents flowing around the pipe, the pipe can be modeled as a
collection of coated plates representing the pipe wall. The following sections compare some
computation results obtained those two models to represent a coated pipe.

3.1 Current Density around a Coated Pipe near an Energized Tower

Footing
A 50 m long perfectly coated section of pipe with a radius of 0.5 m is modeled in the system. The
pipe is buried at a depth of 1.5 m in a uniform soil with a resistivity of 100 ohm-m. Near the center
of the pipe section, there is a tower footing in the form of a square loop with a side length of 0.5
m, with four legs going from 0.1 m to 2.1 m in depth. The center-to-center distance from the tower
footing to the pipe is 2 m. A 1 kA current is injected in the tower footing. Observation profiles are
defined around the pipe, in the middle of the pipe section.
Figure 6 shows this system where the perfectly coated pipe is modeled using plates perfectly
coated on both sides. The thickness of the plates is 0.005 m. The current density around the pipe
calculated with the MALT module is shown in Figure 7.

Figure 6: A perfectly coated pipe modeled with perfectly coated plates.

Copyright © 2017 SES ltd. All rights reserved. Page 32-5

PART IV: RESEARCH WORK IN PROGRESS

Figure 7: Current density around a perfectly coated pipe modeled with perfectly coated plates.
When the coated pipe is modeled with a perfectly coated wire of 0.5 m radius, the calculated
current density around the pipe is shown in Figure 8.

Figure 8: Current density around a perfectly coated pipe modeled with a perfectly coated conductor
(wire).
From Figure 7 and Figure 8, it can be seen that the magnitude of the current and its pattern of
flow around the pipe are quite different for the two approaches used to model the coated pipe. In
Figure 7, when the coated pipe is modeled by coated plates, the coated pipe shows a strong
Page 32-6 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

“blocking” effect and the current flows around the coated pipe. This effect also forms two zones
with lower current density, one on the side facing the source near the pipe and the second one on
the opposite side of the pipe, as indicated in Figure 7. Within the second zone, the current density
is quite small.
In order to further demonstrate the difference between the results obtained with these two
modeling approaches, the current distribution along the same observation profiles for a case
without the pipe is shown in Figure 9.

Figure 9: Current density along the same profiles for the case without the coated pipe.
From Figure 8 and Figure 9, it can be seen that the current distribution along the same profiles is
essentially the same with and without the presence of the pipe. This means that a perfectly coated
conductor has little effect on the current flowing around the conductor, even if this conductor has
a large radius.

3.2 Transfer Potential on Conductors Parallel to a Perfected Coated Pipe

The system considered in this section consists of two horizontal bare conductors located on
opposite sides of a coated pipe and running parallel to the pipe. The radius of the two conductor
is 0.005 m and the radius of the pipe is 0.5 m. The two conductors and the pipe are buried at a
depth of 1.5 m in a uniform soil with a resistivity of 100 ohm-m. The length of the two conductors
and of the pipe is 50 m. The conductor located to one side of the pipe is held at a potential of 1 kV
while that on the other side of the pipe and the pipe itself are left unenergized. As in the previous
section, the coated pipe is modeled either as coating wire of radius 0.5 m or as a collection of
coated plates. The system for the case where the pipe is modeled by coated plates is shown in
Figure 10.
In the examined cases, the distance between the energized conductor (on the right side of the pipe
in Figure 10) to the center of the pipe is kept constant at 0.6 m while the distance between the
other conductor (on the left side of the pipe in Figure 10) and the pipe varies from 0.6 m to 6 m.
The results for the potential transferred to the unenergized conductor as computed by the MALT
module for those cases are listed in Table 1.

Copyright © 2017 SES ltd. All rights reserved. Page 32-7

PART IV: RESEARCH WORK IN PROGRESS

Table 1: The potential transferred to the unenergized conductor at different distances from the center of
the pipe.

Potential Transferred to Unenergized Conductor (V)

Distance from the
Pipe Center (m) Coated Pipe Modeled Coated Pipe Modeled by a
by Coated Plates Coated Conductor

0.6 386.70 514.18

0.8 399.52 499.01

1 409.74 485.47

1.5 414.62 456.5

2 404.77 432.21

3 376.29 392.38

4 348.93 360.39

5 324.91 333.83

6 304.06 311.35

The results computed using the two models are quite different when the conductor is near the
pipe. With the model using perfectly coated plates, the potential transferred to the unenergized
conductor is smaller than with the model using the perfectly coated conductor and the difference
becomes larger the closer this conductor is to the pipe. The relative difference between those two
values (normalized with respect to the values obtained with the plate model) is plotted in Figure
11 as a function of distance between the unenergized conductor and the center of the pipe.

Page 32-8 Copyright © 2017 SES ltd. All rights reserved.

 UGM 2017 – CANNON BEACH, OREGON

Figure 10: Coated pipe with two parallel conductors on opposite sides of the pipe.

Figure 11: Difference in the potential transferred to the unenergized conductor computed using the coated
plate and coated conductor model. The differences are expressed as a percentage relative to the plate
results.

Copyright © 2017 SES ltd. All rights reserved. Page 32-9

PART IV: RESEARCH WORK IN PROGRESS

4 Conclusion
This article examined the differences in the ground resistances and earth potentials computed
using models including either metallic plates or equivalent structures made of metallic wires. The
study focused in two areas:
 A comparison of models of grounding grids and their approximation as a square or
rectangular plate
 A comparison of models of coated pipes implemented either a coated wires or as a
collection of coated plates.
For the grounding grids and equivalent square or rectangular plates, the results show that the
difference in the ground resistance computed using the two models becomes larger in the
following situations:
 The mesh size in the grid increases.
 The shape of the grid is elongated.
 The grid is buried deeper.
 The size of the grid decreases.
For a plate, the computed scalar potentials and touch voltages are quite flat on the earth surface
whereas for a grid, “ripples” are visible in the computed scalar potentials and touch voltages,
especially when the grid is close to the earth surface.
For a coated pipe modeled using coated plates, current flows around the pipe while when the
coated pipe is modeled using a coated conductor, it has little effect on the current distribution
around the modeled pipe.
The effect of a coated pipe modeled using coated metallic plates on the potential transferred to an
unenergized conductor running parallel to the pipe from an energized conductor located on the
other side of the pipe is also demonstrated. The results show that the potential transferred to the
unenergized conductor is smaller when the coated pipe is modeled using coated plates than when
it is modeled using a coated wire. The difference between the transferred potential in the two
models is larger when the unenergized conductor is located close to the coated pipe.

5 References
[1] Peter Zhao, Simon Fortin and Farid P. Dawalibi," Low Frequency Electric Field and Potential in MALT
for Perfect Soil Models," in CDEGS Users' Conference Proceedings, Rock Creek Resort, Montana, USA,
2009.
[2] Peter Zhao, Simon Fortin and Farid P. Dawalibi, " Electric Field and Potential in Systems including
Plates," in CDEGS Users' Conference Proceedings, Montreal, Quebec, CANADA, 2010.
[3] Peter Zhao, Simon Fortin and Farid P. Dawalibi, " Coated metallic plates in uniform soil model in
MALT," in CDEGS Users' Conference Proceedings, NewPort, Rhode Island, USA, 2014.
[4] Peter Zhao, Simon Fortin and Farid P. Dawalibi, " Coated Metallic Plates in Horizontal Multilayer and
Perfect Soil Models in MALT," in CDEGS Users' Conference Proceedings, San Diego, California, USA,
2015.
[5] Peter Zhao, Simon Fortin and Farid P. Dawalibi, " Modeling of Plates Structures in HIFREQ," in
CDEGS Users' Conference Proceedings, San Diego, California, USA, 2015.
[6] Ali Aghabarati, Rouzbeh Moini, Peter Zhao, Shabnam Ladan, Simon Fortin and Farid P. Dawalibi, "
Modeling Network of Metallic Plates and Cylindrical Conductors in HIFREQ," in CDEGS Users'
Conference Proceedings, Boulder, Colorado, USA, 2016.
[7] Peter (Huiliang) Zhao, Simon Fortin and Farid P. Dawalibi, " Calculation of the ‘Field Intensification
Factors’ using Method of Moments for Lightning Protection," in 7th Asia-Pacific International
Conference on Lightning, Chengdu, Sichuan, CHINA, 2011.
Page 32-10 Copyright © 2017 SES ltd. All rights reserved.
 UGM 2017 – CANNON BEACH, OREGON

[8] Peter (Huiliang) Zhao, Simon Fortin and Farid P. Dawalibi, " Electrostatic Field and Potential
Generated by Arbitrarily-Oriented Line and Surface Sources with Additional External Electric Fields
in Air," in CET Conference Proceedings, Shanghai, CHINA, 2011.
[9] Peter (Huiliang) Zhao, Simon Fortin and Farid P. Dawalibi, "Analysis of Grounding System Including
Freely Oriented Metallic Plates with Arbitrary Shape in Multilayer Soils," in IEEE TIA Special Issue on
Grounding Systems, 2015

Copyright © 2017 SES ltd. All rights reserved. Page 32-11

Part V: Recently Published Technical Articles
 Electromagnetic Shielding Properties of Spherical
Polyhedral Structures Generated by Conducting
Wires and Metallic Surfaces
A. Aghabarati, R. Moini, S. Ladan, S. Fortin and F. P. Dawalibi
Safe Engineering Services and Technologies ltd.
Montreal, CANADA
www.sestech.com

Abstract—The shielding performance of new spherical Spherical periodic structures are mainly important for the
polyhedral structures illuminated by an electromagnetic plane wave design of conformal shield structures for devices whose
is investigated. The spherical shield screens consist of a juxtaposition geometry is not planar. Curved screens can be used on airborne
of wire conductors and metallic screens distributed according to the equipment, such as aircrafts and satellites. They provide both
regular pattern of a spherical polyhedron. It is shown that at low mechanical and electromagnetic protection without disturbing
frequency, the wire frame screen provides high levels of shielding the aerodynamic properties of the protected system. Spherical
effectiveness, while its counterpart of metallic surfaces has a similar FSS can also be used in the realization of multi-frequency
behavior for higher frequency bands. Moreover, the shielding curved reflectors or frequency selective radomes that can
characteristics of a combined wire-surface spherical surface are
provide reflection and transmission in different bands of
reported and compared with the cases of a single screen of wire
conductors or metallic surfaces. The reported numerical results
frequencies [6] [7]. The goal of this paper is to present a new
illustrate the advantage of using the combined configuration. It is type of spherical structure with shielding effectiveness
observed that the combined wire-surface spherical screen can be characteristics that can be used for the protection of transceivers
designed in order to be effective simultaneously against interference working at frequencies up to 80 MHz in both emission and
at the low and high frequency limits for any polarization of the reception mode. Such spherical FSS structures can be also
incident field. implemented as buildings’ dome or roofs to control the
propagation of signals into and out of buildings.
Keywords—Shielding Effectiveness (SE); Electromagnetic
Interference; Spherical Shield Screen, Polyhedral Structures. Accordingly, this article presents an electromagnetic
analysis of spherical screens which are composed of conducting
wires and metallic surfaces. The spherical FSS are constructed
I. INTRODUCTION following the regular pattern of a spherical polyhedron.
Polyhedral structures have recently been investigated for their
With the increase in radio frequency communications and of scattering cross section (RCS) and for the design of enhanced
the related hazards due to intentional and unintentional spherical cavities [8]. Their unique geometrical features can
interfering sources, electromagnetic shielding has become a provide electromagnetic devices that attain multi-band
fundamental concern for problems related to electromagnetic frequency selectivity independent of polarization or angle of
compatibility (EMC) in order to design, assemble and test incidence due to their spherical periodicity.
electrical devices [1]. Sensitive devices are usually placed inside
a metallic box or container to avoid their malfunction when A full wave electromagnetic analysis based on Method of
exposed to undesirable electromagnetic interference [1] [2]. Moment (MoM) is used to investigate the shielding
Ideally, the enclosure should be conformal to the geometry of characteristics of the spherical wire and metallic surface screens.
protected device while providing a level of electromagnetic The methodology takes into account all the electromagnetic
shielding acceptable to the susceptibility of the system. interactions between the wire components and surface parts. The
survey allows EMC designers to interpret the effect of the
The idea of designing shield enclosures that can also control different geometrical parameters involved in the design of
the electromagnetic wave propagation has stimulated many spherical shield enclosures. It can also provide guidance in the
research activities on the development of novel structures that design of low density, low drag and multi-band operating
will allow communication of the protected devices at some spherical shields.
desired frequencies [3]. With the increasing demand for the
design of Frequency Selective Surfaces (FSS), the The organization of the paper is as follow. In Section II,
electromagnetic properties of many planar shield surfaces have spherical polyhedral structures are described together with the
been extensively studied during the past years [3] [4]. Various wire grid and metallic surface shield structures which are
planar periodic structures have been investigated in term of their constructed based on their geometry. In Section III, the full-
shielding effectiveness (SE) to enhance the fabrication of 2-D wave computational methodology employed for the analysis of
shield structures [3] [4] [5]. However, little research has been the shield structures in the spatial domain is presented.
made on the characterization of the shielding properties of Numerical results are presented in Section IV, where the spectral
spherical screens with spherical periodicity. shielding properties of the studied wire and metallic screen
shield enclosures are discussed. The advantages of using a
combined wire-surface spherical surface are reported and the
results are compared with the cases of a single screen of wire or
metallic screen. Finally, in Section V, conclusions are drawn.
II. SPHERICAL POLYHEDRAL STRUCTURES
This paper proposes new shielding enclosures whose
geometries are based on spherical polyhedral structures. As (a)- W240 (b)- S240 (c)- WS240
Fig. 1. Illustration of the spherical polyhedral shield screens.
shown in Fig. 1, three different types of shield enclosures are (a) Wire screen, W240, (b) Surface screen, S240, (c) Combined wire-surface
considered based on approximations to a smooth C240 screen, WS240.
polyhedron spheroid, which has 240 vertices located at a fixed
distance from the center of the spherical surface. As indicated in the CDEGS package, simulation of complex networks of wire
Fig. 1-(a) to Fig. 1-(c), the wire, surface and combined wire- and metallic surfaces can be performed from 0 Hz to several tens
surface shield screens are denoted by W240, S240 and WS240 of MHz for models inside an infinite medium or in a
respectively. It should be noted that the C240 polyhedron is multilayered soil medium. Below is a description of the
composed of 12 pentagons and 110 hexagons that tile the computational modeling theory for the studied networks under a
spherical surface. More details on generating polyhedron plane-wave excitation.
spheroids can be found in [8].
The boundary conditions of Maxwell’s equations dictate that
The W240 wire frame shield enclosure, illustrated in Fig. 1- the tangential components of the total electric field vanishes on
(a), represents the (imperfect) sphere made from 360 conducting the surface of the wire conductors and surface plates as follow:
strips or wires that form the edges of the spherical polyhedron.
The construction shows that at each vertex three conducting (ℒw 𝓙w + ℒs 𝓙s )|tan = 𝐄 inc | (1)
wires come together. From the inherent spherical symmetry of tan
the structure, it can be concluded that the alignment of the
constructed low-drag screen does not depend significantly on its where 𝓙w and 𝓙s represent the total electric current densities
orientation. The radius of the wire conductors can be changed to for the induced unknown currents on the wire conductors and
control the size of the openings over the surface of the structure. metallic surfaces, respectively. The incident electromagnetic
The effects of such parametrization on SE characteristics are field is denoted by 𝐄 inc . The operator ℒ returns the electric field
examined in this study. scattered by the current densities and is defined as:
As depicted in Fig. 1-(b), the S240 is a collection of 132
disjointed perfectly electric conductive (PEC) screens that have ℒΩ 𝓙Ω = ∫ G𝚷 𝓙Ω′ d𝐫 ′ + ∇ ∫ GΦ ∇′ ∙ 𝓙Ω′ d𝐫 ′ , Ω = W, S (2)
an area nearly equal to the area of the pentagons and hexagons Ω′ Ω′

of the underlying polyhedron. The surface shield enclosure is in

fact a complementary spherical screen to the W240 wire grid where G𝚷 and GΦ represent the free-space Green’s functions for
screen in order to create the complete spherical polyhedron. For the vector potential and scalar potential, respectively [10]. In
the surface shield, changing the effective area of the individual addition, Ω stands for the problem’s domains which are the wire
planar surfaces is a controlling parameter on the size of existing conductors or surface plates in general.
gaps between the disjointed surfaces. Such parametrization The numerical approach that has been developed here for
provides a transition between a structures with significant gaps determining the unknown currents is based on a discretization of
to a completely solid polyhedral surface. It is shown in Section surface wire conductors and planar surfaces using tubular
IV that this transition has significant effects on the SE segments and quadrilateral surface patches, respectively. The
performance. current densities on the surface of every individual wire and
Fig. 1-(c) depicts the combined wire and surface spherical patch segments are expanded in terms of their corresponding
shield screen, called WS240, which contains the components of linear basis functions:
both the wire frame and surface screen structures. It should be NΩ
noted that the group of conducting wires are disjointed from the
metallic surfaces. The size of the gap between the components 𝓙Ω = ∑ 𝐼𝑛𝛺 𝓳Ω
n , Ω = W, S (3)
in the new design can be parametrized with both the wire radius n=1
and the area of the surfaces. The numerical results for the SE
behaviour of WS240 illustrate that careful combinations of the After substituting (3) into (1), we use appropriate line test
individual wire and surface components can bring advantageous functions in order to transform the boundary value problem into
shielding characteristics to the WS240 screen. a matrix equation 𝐴𝐼 = 𝑉 of size 𝑁 = 𝑁w + 𝑁S . The structure
of the resulting MoM matrix equation has four blocks, as
follows:
III. COMPUTATIONAL MODELING THEORY
𝐴𝑤𝑤 𝐴𝑤𝑠 𝐼 𝑤 𝑉𝑤
All the introduced shielding structures have been analyzed [ 𝑠𝑠 ] [ 𝑠 ] = [ 𝑠 ] (4)
𝐴𝑠𝑤 𝐴 𝐼 𝑉
using the commercial CDEGS software package that provides
an all-inclusive wire and surface Method of Moment (MoM)
integral equation method [9]. Through the HIFREQ module of
where the unknown coefficients of the wire and surface currents A. Wire Spherical Shield Screen
are denoted by 𝐼 𝑤 and 𝐼 𝑠 , respectively. The excitation vectors The W240 wire grid polyhedron (Fig. 1-(a)), with an inner
resulting from the line integral over the surface and wire regions, radius of 10 m, is placed at the center of the coordinate system
denoted by Υ = W, S, are shown as 𝑉 𝑤 and 𝑉 𝑠 . The elements of in free space. The incident plane wave illuminates the structure,
each block are computed by taking into account the interactions and the plane wave has a polarization angle of 0 degree. In the
between the 𝑚𝑡ℎ test segment 𝓽Υm and the 𝑛𝑡ℎ expansion term following, the analysis is restricted to the zero polarization
𝓳Ω
n , as follows[10]: angle, since the spherical symmetry of the studied screens makes
[𝐴𝛶𝛺 ]mn = [𝑍 𝛶𝛺 ]mn + [𝐶 𝛶𝛺 ]mn them nearly insensitive to the orientation or to the polarization
(5) of the applied excitation. The frequency of the incident wave
for Υ = W, S ; Ω = W, S varies from 1 KHz to 80 MHz in 325 frequency steps. At the
highest frequency of the analysis band, the radius of the screen
where is about 2.667 times the wavelength. At each frequency, the
shielding performance of the screen is evaluated at the center of
[𝑍 𝛶𝛺 ]mn = ∫ 𝓽Υm ∙ ∫ G𝚷 𝓳Ω ′
n d𝐫 ⌋𝐫=𝐫 − d𝐫 the structures, using the formulation described in Section III
Υ Ω with the wire unknowns only.
(6)
+∫ 𝓽Υm ∙ ∫ G𝚷 𝓳Ω ′
n d𝐫 ⌋𝐫=𝐫 + d𝐫 Fig. 2 shows the computed electric and magnetic SEs as a
Υ Ω
function of frequency when the radius of the screen’s wire
conductors is 1cm. As seen in this figure, the electric and
[𝐶 𝛶𝛺 ]mn = ∫ GΦ ∇′ ∙ 𝓳Ω ′
n d𝐫 ⌋𝐫=𝐫𝐜−
magnetic SEs exhibit different oscillatory behavior in the
Ω frequency band of analysis. It can also be observed that the
(7) shielding performance generally degrades as the frequency
− ∫ GΦ ∇ ′ ∙ 𝓳Ω ′
n d𝐫 ⌋𝐫=𝐫𝐜+
Ω
increases, and both SEs exhibit sharp drops at some specific
frequencies. Indeed, at these frequencies the incident power
penetrates into the structure from the openings of the wire screen
[𝑉 𝛶 ]m = ∫ 𝓽Υm ∙ 𝐄 inc d𝐫 (8) and the input power excites the specific resonant modes of the
Υ equivalent spherical cavity. Close to resonant frequencies, the
amplitude of the electric or magnetic field increases in the
For every wire segment, Υ stands for a straight-line interior region, resulting in a significant degradation of the
integration path from the center of the segment (𝐫 − ) to the end- corresponding SEs. The first maximum of the electric SE
point of that segment and then continuing over another straight- occurs at 11.25 MHz while its first minimum takes place at
line to the center of the adjacent segment (𝐫 + ). Similarly, for the 19.25 MHz. The magnetic SE exhibits the opposite trend. The
surface patches, the path of integration is from the center of the ratio of wavelengths to radius of the structure at those
patch to the midpoint of an edge, and then continuing over frequencies is about 2.70 and 1.56. It should be noted that for a
another straight-line path toward the center of the adjacent patch. spherical cavity with the same dimensions, the lowest
Once the induced currents are obtained, the electric and resonances occur at the frequency of 13.10 MHz (TM101 mode)
magnetic fields are computed for the evaluation of the electric and 21.45 MHz (TE101 mode), respectively [12]. Interestingly,
and magnetic SEs. The shielding effectiveness at any interior the ratio of wavelengths to radius of the cavity at these
point is defined based on the amplitude of the EM fields and is frequencies are also 2.32 and 1.40 which are close to their wire
given as follows [11]: enclosure counterpart.

|𝐄𝐚𝐛𝐬𝐞𝐧𝐜𝐞 𝐨𝐟 𝐬𝐜𝐫𝐞𝐞𝐧 | In addition, the electric and magnetic shielding effectiveness

𝐒𝐄𝐝𝐁 (𝐄𝐥𝐞𝐜𝐭𝐫𝐢𝐜) = 𝟐𝟎𝐥𝐨𝐠( ) (9) vanish simultaneously when the frequency increases. This
|𝐄𝐩𝐫𝐞𝐬𝐞𝐧𝐜𝐞 𝐨𝐟 𝐬𝐜𝐫𝐞𝐞𝐧 |
shows that the total electromagnetic field into the structure is
|𝐇𝐚𝐛𝐬𝐞𝐧𝐜𝐞 𝐨𝐟 𝐬𝐜𝐫𝐞𝐞𝐧 | reduced to the incident field at higher frequencies. Finally, the
𝐒𝐄𝐝𝐁 (𝐌𝐚𝐠𝐧𝐞𝐭𝐢𝐜) = 𝟐𝟎𝐥𝐨𝐠( ) (10)
|𝐇𝐩𝐫𝐞𝐬𝐞𝐧𝐜𝐞 𝐨𝐟 𝐬𝐜𝐫𝐞𝐞𝐧 | behaviour of the electric and magnetic SEs for the entire

IV. NUMERICAL RESULTS

In this section, the electromagnetic shielding effectiveness
of the introduced spherical structures and their variation with
respect to frequency are investigated. First the SE properties of
the wire screen structure is illustrated in a selected frequency
band and the results for different conductor radius sizes are
presented. Next the shielding characteristics of the spherical
surface enclosure with various surface areas is investigated and
the observed frequency selective features are discussed in detail.
The SE results for the surface screen are compared with those
for the wire screen shield enclosure. Finally, the combined wire- Fig. 2. Electric SE and Magnetic SE of the W240 wire screen as a function of
surface spherical structures is simulated and the SE frequency of 1cm.
characteristics of this structure are presented.
 (a) (b)
Fig. 3. Variation of the electric field in the vicinity of the W240 wire screen
with conductor radius of 1cm illuminated by a plane wave. (a): 5MHz, (b)- Fig. 5. Electric SE and Magnetic SE of the S240 surface screen as a function of
11MHz frequency with surface are of 80 %.

frequency range implies that the structure performs as a high- over the selected spectrum as the radius increases from 1 cm to
pass filter. 5 cm. It can be observed that the amplitude of the SE peaks and
nulls grows as the size of the apertures decreases by using
The variation of the magnitude of the electric field in the thicker conductors.
vicinity of the structure over an observation plane passing
through the center of the structure and that contains the wave
propagation vector is illustrated in Fig. 3 at two frequencies in B. Surface Spherical Shield Screen
the lower band: 5 MHz and 11 MHz. It can be seen that the The spherical structure used for the analysis of this section
weakest electric field is achieved inside the screen at 5 MHz is shown in Fig. 1-(b). The center of a S240 spherical surface
compared to the exterior. By contrast, at 11 MHz the field is with a radius of 10 m is located at the origin of the coordinate
strong at the center and the largest electric field is found inside system and the structure is illuminated using the plane wave
the spherical wire grid screen. The variation in the amplitude of described in Section IV-A. The electromagnetic field at the
the electric field coupled into the screen is in agreement with the center of the structure is obtained numerically using the
maximum and minimum values of the electric SE computed at approach described in Section III, when the formulation contains
the center of W240 polyhedral screen. only the surface unknowns.
The effect of the radius of the wire on the SE is examined Fig. 5 depicts the variation of the electric and magnetic SEs
next. Three different wire screen structures with different wire with frequency of a spherical structure with 80 % surface area.
conductor radii are simulated. Fig. 4-(a) and 4-(b) present the The computation is performed again for the frequencies varying
variation of the electric and magnetic shielding effectiveness from 1 KHz to 80 MHz.The SE variations exhibit again multiple
resonances. But contrary to the wires structure, the amplitude of
the sharp peaks and troughs is increasing with frequency. In
addition, spectral selectivity of the structure for suppressing the
incident electromagnetic fields appears at higher frequencies.
On the other hand, both electric and magnetic fields can
penetrate through the structure at several resonance frequencies
which results in maximum degradation of their SEs.
The first maximum of the electric SE occurs at 10.25 MHz
while its first minimum takes place at 17.75 MHz. The magnetic
SE exhibits the opposite trend. The ratio of wavelengths to
radius of the structure at those frequencies is about 2.92 and
(a) 1.69. Furthermore, it is now observed that the electric and
magnetic SEs values vanish simultaneously at low frequency.

(b) (a) (b)

Fig. 4. Variations of Electric SE and Magnetic SE for W240 wire screen as a Fig. 6. Variation of the electric field in the vicinity of the S240 surface screen
function of frequency for different conductor radii. with surface area of 80 % illuminated by a plane wave. (a): 77MHz, (b)-
(a)- Electric SE, (b)- Magnetic SE 70MHz
 (a) (b)
(a)

(c) (d)
Fig. 9. Variation of the electric field in the vicinity of the WS240 wire-surface
screen with conductor radius of 1 cm and surface size of 80 % illuminated by a
(b) plane wave. (a): 5 MHz, (b)- 11 MHz, (c): 77 MHz, (d)- 70 MHz
Fig. 7. Variation of Electric SE and Magnetic SE for S240 surface screen as a
function of frequency for different surface areas. electric SE and magnetic SE of the structures with effective
(a)- Electric SE, (b)- Magnetic SE
surfaces varying from 80 percent to 50 percent. A remarkable
From the results, it can be found that this shield configuration is decrease is found in the level of electric and magnetic shield
transparent to low frequency interference and acts as a low pass protection, particularly in the higher frequency band, as the area
filter. of the component surfaces are decreased. This confirms that the
surface screen is more transparent to high frequency waves for
Fig. 6 shows how the surface screen interacts with applied a design with reduced surface area. Moreover, the resonance
high frequency waves at 70 MHz and 77 MHz. The large frequencies corresponding to peaks and drops tends to increase
variations of the electric field inside the metallic shell as shown in the figures.
demonstrate that resonance effects are strong and the strength of
the instantaneous electric field at the center can vary abruptly in
the high frequency band. At 77 MHz the magnitude of the C. Wire-Surface Spherical Shield Screen
electric field at the center of the screen is larger than the field in In our final calculation of the electric and magnetic SE, we
the unscreened scenario, while at 70 MHz the presence of the consider the combined wire and surface structure shown in Fig
polyhedral surface structure greatly reduces the level of the 1-(c), and compare the results with those obtained in section A
penetrated wave at the center point. and B. It will be shown that the shielding performance of the
We now consider decreasing the area of the pentagons and combined wire-surface screen at low and high frequencies is
hexagons for the S240 shield structure, where we can characterized by the attributes of the wire screen and surface
characterize the shield performance with respect to the effective screen, respectively. In Fig. 8, the electric SE and magnetic SE
surface area and the blockage that the metallic pieces provide. curves of the WS240 are shown, when the wire radius is 1 cm
In Fig. 7-(a) and Fig. 7-(b) comparisons are reported for the and surface area is 80 %. It is interesting to note that there is a
frequency (15 MHz) below which the performance of the
combined screen is mostly determined by its wire screen
counterpart. It can also be seen that at sufficiently high
frequencies (more than 35 MHz), the SE curves of the WS240
structure tend to be same as those of the surface screen.
Consequently the advantageous properties of both wire and
surface components can be superimposed in the new design.
However, there is a frequency band (15 MHz to 35MHz) at
which the effects of coupling between the two structures is
producing a different shielding behaviour. The structure can thus
be considered as a band pass filter.
Fig. 9 illustrates the distribution of the electric field of the
Fig. 8. Electric SE and Magnetic SE of the WS240 surface screen as a function combined screen at the four frequencies for which the maximum
of frequency with conductor radius of 1 cm and surface size of 80 %. and minimum values of the SEs for W240 and S240 were
observed in the studied frequency band. Fig.9-(b) demonstrates REFERENCES
the large leakage that happens through the holes of the wire-grid [1] S.Celozzi, R. Araneo, and G. Lovat, Electromagnetic Shielding,
part of WS240 at 11 MHz compared to the blockage provided at Hoboken, NJ, USA: IEEE, 2008.
5 MHz (Fig.9-(a)). As observed, in the low frequency band, the [2] A. Frikha, M. Bensetti, L. Pichon , F. Lafon, F. Duval, and N. Benjelloun,
surface screen has minimum interactions with the applied field “Magnetic shielding effectiveness of enclosures in near field at low
and the distribution of the field is very similar to Fig. 3. On the frequency for automotive applications,” IEEE Trans. Electromagn.
Compat., vol. 57, no. 6, pp.1481-1490, 2015.
other hand, at the two higher frequencies (70 MHz and
77 MHz), maximum and minimum values of the computed [3] D. Sievenpiper, L. Zhang, R. F. J. Broas, N. G Alexopolous and E.
Yablonovitch, “High-impedance electromagnetic surfaces with a
electric field are generated due to the presence of the surface forbidden frequency band,” IEEE Trans. Microw. Theory Tech., vol. 47,
screen as shown in Fig. 9-(c) and Fig. 9-(d). The field no. 11, pp. 2059-2074, 1999.
distribution at those frequencies is in close agreement with the [4] Y. E. Erdemli, K. Sertel, R. Gilbert, D. E. Wright and J. L. Volakis,
results obtained for S240 (Fig. 6). ”Frequency-selective surfaces to enhance performance of broad-band
reconfigurable arrays,” IEEE Trans. Antennas Propag., vol. 50, no. 12,
pp. 1716-1724, 2002.
V. CONCLUSION
[5] E. Unal, A. Gokcen, and Y. Kutlu, “Effective electromagnetic shielding,”
The shielding effectiveness of 3-D polyhedral spherical IEEE Microw Magazine., vol. 7, no. 4, pp. 48–54, Aug. 2006.
screens constituted by wire conductors and metallic surfaces [6] C.Yu and C. C. Lu, “Analysis of finite and curved frequency‐selective
under a plane wave illumination has been investigated through surfaces using the hybrid volume‐surface integral equation approach,”
a full-wave wire and surface formulation in conjunction with the Microw.and Optical Tech. Lett., vol. 45, no. 2, pp. 107-112, 2005.
spatial domain Method of Moment. Three different spherically [7] B. Philips, E. A.Parker, and R. J. Langley, “Influence of a curved FSS on
periodic structures (W240, S240 and WS240) have been studied the radiation patterns of an enclosed source,” Ninth International
in detail and the obtained SE results at their center have been Conference on Antennas and Propagation, pp. 524-527, 1995.
compared. It has been shown that, for the spherical wire screen [8] P. Bernhardt, “Radar backscatter from conducting polyhedral spheres,”
IEEE Antennas Propagat. Mag., vol. 52, no. 5, pp. 52-70, 2010.
under plane-wave excitation, significantly larger values of
[9] “CDEGS 15.1.4080”, Safe Engineering Services & Technologies Ltd.
electric and magnetic SE can be observed at lower frequencies Montreal, Québec Canada, 2015 [Online]. Available:
and the protection level decreases when increasing the http://www.sestech.com/
frequency. In contrast, the surface screen is transparent to low [10] W. C. Gibson, “The method of moments in electromagnetics,” CRC press,
frequency excitations and can provide electromagnetic field 2014.
protection only at selected higher frequencies. The SE [11] C. L. Holloway, D. A. Hill, M. Sandroni, J. M. Ladbury, J. Coder, G.
performance of the composite WS240 structure is characterized Koeple, A. C. Marvin, and Y. He, “Use of revereration chambers to
as the combination of the wire and surface screens, as it attains determine the shielding effectiveness of physical small, electrically large
protection simultaneously at both low and high frequencies. The enclosures and cavities,” IEEE Trans. Electromagn. Compat., vol. 50, no.
4, pp. 770-782, 2008.
influence of various geometrical parameters (the radius of the
[12] K. Zhang and D. Li, Electromagnetic theory for microwaves and
wire conductors and area of metallic surfaces) on the shielding optoelectronics, NY, USA: Springer, 2008.
performance has been assessed to provide general guidance for
the design of effective spherical shield enclosures. Future work
will examine the effects of stratified soil on the shielding
effectiveness of 3-D polyhedral spherical screens and will
provide quantitative information on the distortions caused by
this half space soil region.
 Induced Disturbances by High Voltage Transmission
Lines on Nearby Stationary Vehicles

S. Ladan, A. Aghabarati, R. Moini, S. Fortin and F. P. Dawalibi

Safe Engineering Services and Technologies ltd.
Montreal, CANADA
www.sestech.com

Abstract—This article discusses the safety issues related to Nowadays, due to progress in numerical methods, complex
electrostatic discharge phenomena caused by electric field structures including transmission lines, grounding networks,
induction from high voltage transmission lines. The main objective multilayer soils and metallic enclosures (vehicles) can be
is to evaluate the inducing effects of an overhead line including
analyzed without introducing any significant approximations.
towers and grounding networks on a nearby vehicle. Two different
cases are considered. The first case studies the network under Recently, an approach based on the finite element method [4],
steady-state conditions while the second case treats a tower has been used to compute the short circuit current for a truck.
subjected to a single phase to ground fault. In both cases the However, for the sake of fast computations, the power network
induced current discharged from the vehicle to the ground is modeled in that work was limited to a simple configuration
evaluated in order to define a severity index that is used to consisting of a three-phase transmission line without supportive
determine the location of the unsafe zones around the tower towers and grounding networks. The effect of the position of
structure.
the truck relative to the transmission line was also omitted. In
Keywords—capacitive coupling; inductive coupling; fault all these works, no attention has been paid to the safety issues
condition; electrostatic discharge; steady-state condition. related to the interference effects on a vehicle caused by a power
network under fault conditions. When a fault occurs on a
I. INTRODUCTION transmission line, very large currents will flow through the
Generally, high voltage transmission lines have undesirable network which can result in strong magnetic and electric fields
inducing effects on nearby structures. To reduce these effects, a in the air and soil and in a very large potential rise at and around
careful analysis of electromagnetic disturbances of power the faulted tower [5]-[6].
networks is needed to guarantee a safe design in terms of According to IEC 604791-1 [7], the level of discharge
electrostatic discharge immunity and EMC tests. A common current during a fault that can lead to significant physiological
example is a stationary vehicle located near a high voltage damages depends on the duration of the fault which can be
transmission line. For this case, the electric coupling between maximum to 3 s [8]. A typical value can be taken as 50 mA, for
the transmission line components and the metallic body of the a fault duration of 0.5 ms. Some strong unintentional muscular
vehicle induces voltage on its surfaces. When the body of the contractions, respiratory difficulties, breathing disorders and
vehicle is insulated from the ground, the electrostatic discharge reversible heart disturbances have been reported for this level
due to a contact between a human body and the vehicle may of current. While this current level is larger than the limit value
result in an electric shock to the person. The National Electrical for steady-state conditions, the discharge current can also be
Safety Code (NESC) guidelines in the USA [1] indicates that larger. Therefore, it is important to analyze this case as well.
for a harmless electric discharge the short circuit current The goal of this paper is to investigate the safety issues for
between the vehicle and ground must be limited to 5 mA at 60 equipment located near power networks under steady-state and
Hz under steady-state conditions. This provides a criterion by fault conditions. To this end, the inducing effects of a complex
which the electrical safety of vehicles parked near high voltage power network on a nearby vehicle is expressed in term of the
transmission lines can be assessed. The analysis of safety short circuit current between the vehicle and ground. A large
criteria for a power network operating under steady-state tractor-trailer has been used as vehicle, to maximize its inducing
conditions was the subject of some research studies [2]-[3]. effects. The full wave analysis method of the CDEGS (Current
Analytical methods based on semi-empirical formulas have Distribution, Electromagnetic Fields, Grounding and Soil
been developed to estimate the short circuit current. However, Structure Analysis) software package [9] is used for the
these methods can only be applied to very simple structures [2]. determination of the hazardous zone around energized
 TABLE I.

Phase Currents (A)

Energization
IB IC IA
-j123 j0
7% unbalance 4250e 4250e 4250ej119

Fault 430e-j120 430ej0 12000ej0

(a) TABLE II.

Phase Voltage (kV)

Energization
ΦB ΦC ΦA
7% unbalance 298e-j121 303ej0 297ej120

Fault 303e-j120 303ej0 _____

electromagnetic analysis is performed for two different

scenarios. In the first, steady-state conditions are considered
while in the second a fault is assumed to occur at the tower. Fig.
1 depicts the geometry of the problem. A three-phase
transmission line with a nominal voltage of 525 KV is used as
source of disturbance. The tower is 35 m high and its grounding
system consists of a 16 m by 16 m loop buried at a depth of 0.5
m and four vertical ground rods which are driven to a depth of
10 m. For the sake of simplicity, it is assumed that the tower is
located in a uniform soil area with a soil resistivity of 1000 Ω-
m, relative permittivity of 1, and relative permeability of 1. The
transmission line conductors are in a horizontal configuration,
(b) without shield wires. The conductors are at a height of 27 m
Fig. 1. Geomtetry of the problem: (a): General view, (b): Top view.
above the ground at the tower location while their height drops
to 17 m, at the point of maximum sag. As shown in Fig. 1, the
tower is in the middle of the studied portion of the power
transmission lines. The method is based on a solution of the network. The transmission line is modeled for 400 m on either
electric field integral equation by the Method of Moment side of the structure: this was found to be sufficiently accurate
(MoM), and is applicable to complex structures including for the calculation of electromagnetic fields around the tower.
networks of wires (power network) and metallic surfaces A detailed description of the geometry of the network can be
(truck). The removal of low frequency instabilities, which found in [11]. The nominal values for the current and voltage of
typically affects such computation methods, permits the each phase are shown in Table I and Table II for the two studied
computation of inducing effects from 0 Hz to several MHz [10]. scenarios. The vehicle chosen for this analysis has dimensions
Also, at higher frequencies or for objects with small electrical of 20 m length, 2.5 m width, and 4 m height, representing a
length using finer meshes can guarantee the precision of results. metallic structure with a large effective surface area. It is
The presence of a multilayer soil can also be taken account. The modeled by 63 quadrilateral and rectangular metallic plates.
paper is organized as follows. In Section 2, the geometry of the The bottom of the truck is at distance of 0.6 m from the ground.
model is explained in detail. In Section 3, the inducing effects When the truck is parked under the high-voltage transmission
of the power network on a nearby truck under steady-state line, the electric field coupling produced by the energized
conditions are investigated and the unsafe zone for the location conductors induces voltage on the surface of the truck.
of the truck is identified. In Section 4, the case of a tower However, the rubber tires insulate the metallic body of the truck
subjected to a single phase to ground fault is investigated for from the ground, but any human contact (for example a hand
completeness. touch) provides a path between the truck and the ground. A
small metallic plate connecting the bottom of the truck to the
II. STUDY MODEL
ground models the human body resistance. The current passing
The study model includes a large tractor-trailer located near through the grounded plate represents the truck short circuit
a tower of a three-phase high voltage transmission line. A current and, as will be explained later, this current is used as an
grounding system is also considered for the tower. The goal is index to specify the unsafe regions around the tower in terms of
to compute the electromagnetic fields in the vicinity of the truck electric field induction. After careful modeling of the structure,
in order to predict the hazardous zone around the tower. The the electromagnetic fields around the tower at the truck location
 (a)
(a)

(b)

Fig 2. Short circuit current in mA under steady-state conditions; (a): around (b)
the tower, (b): at the location of maximum sag.
Fig. 3. The electric field; (a): around the tower (D1=5 m, D2=15 m), (b): near the
location of maximum sag (i.e., D1=200 m, D2=0 m with respect to the center of
as well as the level of short circuit current transferred from the the tower).
truck to the ground is studied. In the present study, it is assumed
short circuit current is computed. It should be noted that at
that the truck is parallel to the high voltage transmission lines
ground level, the coordinate (0, 0) represents the location of the
since, as is shown in [2], objects perpendicular to the center of the tower in the (X, Y) plane, while coordinate (D1, D2)
transmission lines have a considerably less induction compared depicts the position of center of truck relative to the origin which
to parallel objects. is at the center of the tower. As shown in Fig. 2(a), the level of
truck to soil short circuit current obtained in the simulation
III. STEADY-STATE CONDITIONS
results is less than 5 mA for D1 and D2 varying from 0 to 20 m.
It is assumed that the network is working with 7% The latter guarantees the safety of the network in terms of
unbalanced phase currents and voltages for the steady-state electrostatic discharge. Another zone that needs to be
conditions. The nominal values for the current and voltage of investigated in terms of electromagnetic field induction is the
each phase are shown in Table I and Table II. The NESC zone where the transmission line conductors reach their
guideline related to clearances of energized electrical supply maximum sag, as shown in Fig. 1(a). Therefore, in order to
conductor [1] states: “For voltages exceeding 98 kV ac to assess the inducing effects of the line in this zone, the analysis is
ground, either the clearances shall be increased or the electric performed for various locations of the truck around the lowest
field, or the effects thereof, shall be reduced by other means as position of the line. At ground level, the location of the lowest
required to limit the steady-state current due to electrostatic center conductor is selected as the origin of the coordinate
effects to 5 mA if the largest anticipated truck, vehicle, or system in the (X, Y) plane, while coordinate (D1, D2) represents
equipment under the line were short-circuited to ground”. the position of center of the truck relative to the origin. Fig. 2(b)
Consequently, to evaluate the safety area around the tower, the depicts the variation of the level of short circuit current for
truck is moved to different locations and for each location the various locations of the truck relative to the position of
 (a)

(a)

(b)

Fig. 4. Short circuit current in mA under fault conditions; (a): around the
tower, (b): under the location of maximum sag.

maximum sag. As the obtained simulation results show that the

level of the truck to soil short circuit current is less than 5 mA
for the studied region, the safety of the network in terms of
electrostatic discharge is still guaranteed. Fig. 3 depicts the
magnitude of the electric field over an observation surface that
(b)
passes through the truck and the transmission line, close to the
tower (Fig. 3(a)) and near the point of maximum sag (Fig. 3(b)). Fig. 5. The electric field; (a): around the tower (D1=5 m, D2=15 m), (b): near
As shown in this figure, the magnitude of the electric field the location of maximum sag (D1=200 m, D2=0 m).
around the truck when it is in the vicinity of the location of
maximum sag is on the range of 103 V/m while the magnitude of
the electric field for the truck near the tower is on the order of phase A conductor is the faulted phase. The fault occurs in an
102 V/m. Since under steady-state conditions the generated area close to the tower where phase A is connected to the tower.
electric field is essentially due to capacitive effects from the As an illustrative example the current applied to Phase A
energized transmission lines, the generated electric field is corresponds to the maximum fault current that could occur at the
stronger in the areas where the energized transmission lines are tower; it was selected to yield a tower GPR close to the nominal
525
closer to the ground. phase potential ( kV). Based on the simulation results, this
√3
turns out to be 12 kA. Note that this maximum fault current
IV. FAULT CONDITIONS depends on the local soil model resistivity values and
The system under study is also examined when a fault characteristics of the shield wires, if any. Therefore, to study the
happens on one of the phase conductors. It is assumed that the system under fault conditions, the network is energized with the
nominal values shown in Table I and Table II. Then the truck ACKNOWLEDGMENT
short circuit current is monitored when it is placed at different The author would like to thank Ms. Yexu Li and Dr. Amir
locations around the tower. Fig. 4(a) depicts the variation of the Hajiaboli from Safe Engineering Services & Technologies ltd.
short circuit current for different positions of the truck relative for all the informative discussion and their invaluable advice and
to the tower. The results shown in Fig. 4(a) indicate that in some comments.
places around the tower the level of truck to soil short circuit
current is quite significant. For example, at a distance of 15 m REFERENCES
from the center of the tower towards the Y direction, the level of [1] National Electrical Safety Code C2-2012, pp. 92-93. Institude of
truck to soil short circuit current transferred to the ground is Electrical and Electronic Engineers, Inc., Ney York. NY 2011.
between 60 mA and 80 mA. According to IEC standard 604791- [2] J. P. Reilly, “Electric field inductiion on long objects – a methodology for
1, this level of body current can cause serious adverse transmission line impact studies,” IEEE Trans Pwr. Apparat. and Sys.,
vol. 98, no. 6, pp. 1841-1849, Nov. 1979.
physiological responses when applied for a fault duration of 0.5
[3] M. Sendaula, D. W. Hilson, R. C. Myers, L. G. Akens, B. J. Woolery,
s. Therefore for the above-mentioned network, under fault “Analysis of electric and magnetic fields measured near TVA’s 500-kV
conditions, the area around the tower could be considered transmission lines,” IEEE Trans Pwr. Apparat. and Sys., vol. 103, no. 6,
unsafe. Fig. 4(b) shows the results of the analysis for the case pp. 328-336, Feb. 1984.
where the truck is located close to the area where the [4] J. T. Leman and R. Schaere, “Quasi-electrostatic induction on stationary
transmission line conductors reach their maximum sag. For this vehicles under high voltage power lines” COMSOL conference in Boston,
2012.
case, the level of short circuit current meets the standard and that
[5] L. Qi, H. Yuan, L. Li and X. Cui, “Calculation of interference voltage on
region can be assumed to be safe. Fig. 5 shows the electric field the nearby underground metal pipeline due to the grounding fault on
in the two studied areas. Referring to Fig. 5(a), when the truck is overhead transmission Llines,” IEEE Trnas. Electromagn. Compat., vol.
located around the tower the maximum level of electric field is 55, no. 5, pp. 965-974, Oct. 2013.
on the order of 104 V/m. This value drops to 103 V/m (ten times [6] G. C. Christoforidis, D. P. Labridis and P. S. Dokopoulos, “Inductive
interference on pipelines buried in multilayer soil due to magnetic fields
less) when the truck is located in the area of maximum sag. from nearby faulted power lines”, IEEE Trnas. Electromagn. Compat.,
Consequently, the short circuit current in this zone is much vol. 47, no. 2, pp. 254-262, May 2005.
smaller than in the area around the tower. So, it can be concluded [7] IEC 604791-1, Effect of current on human beingsand livesstock Part1:
that under fault conditions, the areas near the tower are the only General aspects 2005.
zones that can be considered unsafe. In this case the [8] IEEE Guide for Safety in AC Substation Grounding, IEEE Standard Std.
comparatively noticeable short-circuit current calculated near 80-2013.
the tower under fault conditions is due to the large potential rise [9] “CDEGS 15.1.4080”, Safe Engineering Services & Technologies ltd.
Laval, Québec Canada, 2015 [Online]. Available: www.sestech.com.
of the earth at that location. Indeed, due to the capacitance of the
[10] J. F. Lee, R. Lee, and R. J. Burkholder, “Loop star basis functions and a
truck, which is relatively high, a huge amount of charge must be robust preconditioner for EFIE scattering problems,”. IEEE Trans.
transferred to the truck to equilibrate the potential. The latter Antennas Propag., vol. 51, no. 8, pp.1855-1863, August 2003.
implies, surprisingly, that simply touching a large insulated [11] W. Ruan, F. P. Dawalibi, J. Ma, and S. Fortin, “Study of electromagnetic
metallic object located near a transmission line tower could fields and current distribution near a transmission line tower subjected to
generate hazardous body currents during a fault. Note, however, unbalanced currents and phase to ground faults,” Proceedings of the 57th
Annual Meeting of the American Power Conference, Vol. 57-I, Chicago,
that this large current is partly due to the particular fault scenario April 18-20, 1995, pp. 176-181.
considered in this work, which assumes worst-case conditions.
V. CONCLUSION
The presented study uses the full-wave analysis method of
the CDEGS software package to evaluate the hazardous areas
around energized transmission lines. A large-size truck parked
near a high voltage transmission line tower is studied under two
different scenarios; when the power network operates under
steady-state conditions and when a fault occurs on one of the
phase conductors near the tower. The level of short circuit
current transferred from the truck to the ground is considered as
a severity index to assess the safety of the area in terms of
electrostatic discharge. The simulation results show that under
steady-state conditions the area around the tower is safe since the
level of the short circuit current meets the 5 mA safety rule.
However, under fault conditions the level of short circuit current
can reach a value of 70 mA around the tower. Referring to the
shock hazard categories mentioned in IEC 604791-1, the
reported truck to soil short circuit current can be dangerous and
make the area unsafe.
 Analysis of Grounding Systems Including Freely
Oriented Plates of Arbitrary Shape in Multilayer Soils
Huiliang Zhao Simon Fortin Farid P. Dawalibi
Safe Engineering Services & Safe Engineering Services & Senior Member, IEEE
technologies ltd. technologies ltd. Safe Engineering Services &
3055 Boul. Des Oiseaux 3055 Boul. Des Oiseaux technologies ltd.
Laval, QC H7L 6E8, Canada Laval, QC H7L 6E8, Canada 3055 Boul. Des Oiseaux
[email protected] [email protected] Laval, QC H7L 6E8, Canada
[email protected]

Abstract -- This paper presents the analysis of grounding soil. The latter quantities provide extra information that help
systems including freely oriented conductors and metallic plates better understand the performance of the grounding systems
of arbitrary shapes in horizontal multilayer soils. The analysis and achieve a safe design economically
method yields not only the scalar potential but also the electric
Numerical results are provided for various configurations
field and current density in the soil, in the low frequency limit.
Numerical results are provided for various configurations such such as a single grounding plate, a plate with ground rods,
as a single plate, a plate with rods, parallel plates, a box or multiple plates forming a box or cylinder, parallel plates, and
cylinders formed by multiple plates and verified against coaxial cylinders. The electric field, current density and
available analytical methods. potential calculated with the new approach are verified
against analytical results and compared against the results of
Index Terms--Current density, electric field, grounding,
the approach of [13] for horizontal or vertical rectangular
grounding grid, grounding plate, GPR, ground resistance,
safety, scalar potential plates.

I. INTRODUCTION II. DESCRIPTION OF THE ANALYSIS

The effect of the structure of the soil on the performance The method of analysis used in this paper is based on a
of grounding systems has been extensively studied in the past low frequency or static version of the point-matching
[1]-[10]. Different types of grounding structures have also Moment Method [14]. First, the potential due to a point
been analyzed [11]-[12], most consisting of meshed source of current located in a multilayer soil must be
structures made of cylindrical conductors. obtained. For this work, the method of images was used for
There are scenarios where the presence of metallic plates this purpose. For horizontal multilayer soil models, not only
in grounding systems can play an important role. Several the air-soil interface, but also all the soil-soil interfaces need
years ago, one of the authors developed a method for to be considered in the generated images. This procedure is
analyzing grounding systems including plates [13]. The identical to that used in several low-frequency grounding
approach in [13] was quite successful but was limited to analysis methods for grounding systems consisting of
horizontal and vertical rectangular plates. cylindrical conductors [5]-[12].
The modeling of freely oriented grounding plates of In the analysis, the cylindrical conductors are treated as
arbitrary shape is valuable in practical applications because it thin wires and subdivided into short segments. The potential
can be quite conveniently used to model a number of generated by the leakage current from a segment is calculated
practical electrodes such as flood gates in dams, the by integrating the potential expression of a point current
foundation of buildings, buried pipes and tanks, wells, a source along the segment, assuming that the leakage current
human foot or simply to represent the limiting case of an is uniformly distributed along the segment. Similarly, the
extremely dense grid. In grounding studies, a plate model metallic plates are subdivided into small patches. The
provides the lowest possible ground resistance and touch potential generated by the leakage current from a patch is
voltages for a grounding grid of a given extent. A plate model obtained by an analytical surface integration of the potential
can also be used to approximate steel rebar in concrete which expression of a point current source on the patch (as detailed
often cannot be accurately modeled due to their great density. in Appendix I), again assuming that the leakage current is
In this paper, an analysis of grounding plates of arbitrary uniformly distributed on the patch. This differs from the
shape and orientation in horizontal multilayer soil models is approach used in [13], where the surface integral was
presented. The analysis method yields not only the scalar calculated numerically.
potential but also the electric field and current density in the To determine the current distribution on the segments and
patches, the Method of Moment is applied by selecting a separating them. Likewise, the potential between the plates
matching point on each element and building a matrix increases linearly from one plate to the other.
equation for the whole system. Assuming that Pij is the The situation is a little bit more complex when the parallel
potential at the matching point of element i contributed by the plates are located in different layers of a horizontal multilayer
leakage current Ij from element j, the matrix equation can be soil. Fig. 1 shows an example in a four layer soil model. The
expressed as soil resistivity is 800 Ω-m, 100 Ω-m, 800 Ω-m and 2000 Ω-
𝑃11 𝑃12 … 𝑃1𝑁 𝐼1 𝑉1 m, respectively, from the top to the bottom layer. The
𝑃21 𝑃22 … 𝑃2𝑁 𝐼2 𝑉2 thickness of the soil in the top layer and the successive two
| . . … . || . | | . | central layers is 8 m, 5 m and 5 m, respectively.
. . … . || . | = | . | (1)
|
. . … . . .
𝑃𝑁1 𝑃𝑁2 … 𝑃𝑁𝑁 𝐼𝑁 𝑉𝑁
where N is the total number of elements in the system and Vi
is the potential at the matching point of element i. Vi can be a
known value or unknown. For those elements with unknown
Vi, equation (1) can be modified with the current conservation
condition for a group of elements, which states that the
leakage current (usually a known value) discharged by this
group, equals to the sum of leakage currents from all the
patches and segments in this group. Equation (1) is then
solved to yield the leakage current from each element.
Fig. 1. Horizontal parallel plates in four-layer soil
It is now possible to compute the ground resistance, GPR
(Ground Potential Rise) of the grounding system, the earth In Fig. 1, the horizontal parallel plates are located at Z = 5
potentials, and the touch and step voltages. The electric field m and Z = 20 m, respectively, below the earth surface (Z-axis
due to a point source of current is obtained directly from the is downward from the earth surface, thereafter). The plates
gradient of the corresponding expression for the potential. are square, with a side length of 200 m and a thickness of 2
From this expression, the contribution to the electric field cm, and centered at X = 0, Y=0. The top plate is held at 0 V
from a segment can be obtained by linear integration along and the bottom plate is maintained at 500 kV.
the segment, and from a patch by surface integration through The calculated electric field strength and magnitude of
the patch surface. The final electric field at an observation current density along horizontal profiles running at different
point is calculated by summing the contributions from all the depths (Z = 6 m, 10 m, 15 m, and 19 m) along Y=0 from X =
segments and patches. In the soil, the conductive current -100 to X = +100 are shown in Fig. 2. The potential along
density J can be obtained with the help of Ohm’s law 𝐽 = those same profiles is shown in Fig. 3. As expected, these
𝜎𝐸, where E is the calculated electric field in the soil, and σ is figures show that the electric field strength, potential and
the conductivity of the soil at the location of the calculation current density are constant along each horizontal profile
point. between the parallel plates, as long as the calculation points
are not too close to the edges of the parallel plates system.
III. VALIDATION OF THE APPROACH Fig. 2 shows that the electric field strength changes in
This section presents several examples for the purpose of different layers due to the difference in soil resistivity. The
validating the approach. While some of those examples do electric field strength is essentially the same within the top
not represent realistic grounding designs, they are examined and third layers, due to the fact that these layers have the
here since the results obtained with these models can be same resistivity (800 Ω-m). It is largest in the bottom layer
readily interpreted and verified. and smallest in the second, again following the resistivity of
those layers. This pattern can be easily understood from the
A. Electric Field, Current Density and Potential between
results for the magnitude of the current density along the
Parallel Plates
same profiles, also displayed in Fig. 2, which shows that the
The general behavior of the electric field, current density, current density is the same in all layers, as can be expected
and potential between parallel plates located in multilayer due to the general principle of continuity of current density.
soils can be calculated analytically for some cases, depending The variation of the electric field strength with depth then
on their orientation and location. For instance, if the two follows from Ohm’s law ( 𝐽 = 𝜎𝐸 ), which shows that the
plates are located in the same layer, the electric field and electric field should be proportional to the resistivity of the
current density far from the edges of the plates is constant, soil in each layer for this configuration.
with the electric field pointing in a direction perpendicular to The variation of the electric field and current density with
the plates and having a magnitude equal to the potential depth between the plates is illustrated more clearly in Fig. 4,
difference between the two plates divided by the distance which plots those quantities along a vertical profile running
from the surface of the earth to a depth of 22 m, along the As shown in Fig. 5, due to the variation of electric field
center of the plates (X = 0, Y = 0). The electric field strength strength from one layer to another, the potential along the
is constant within each layer, and changes in different layers vertical profile is piecewise linear. It is linear within each
in proportion to the soil resistivity. The magnitude of the layer, with a slope proportional to the soil resistivity.
current density is constant between the plates and drops to
zero above the top plate, as expected. Below the bottom plate
the current density is non-zero, as required to maintain the
potential of that plate to 500 kV.

Fig. 5. Potential along vertical profile across two horizontal plates

As another example, two vertical parallel plates are buried

in the same soil model (Fig. 6). The corners of one of the
plates (Plate A) are at (0, -100, 1), (0, -100, 31), (0, 100, 31),
Fig. 2. Electric field strength and magnitude of current density between
horizontal plates and (0, 100, 1). The other plate (Plate B) is separated from
the first by 10 m, in the X = 10 plane. The thickness of both
plates is 2 cm. Plate A is held at 0 V and Plate B at 500 kV.

Fig. 3. Potential between horizontal plates

Fig. 6. Vertical parallel plates in four-layer soil

Fig. 7 shows the electric field strength and the magnitude

of current density along four horizontal profiles going from X
= -5, Y = 0 to X = 15, Y = 0 at various depths (6 m, 10 m, 15
m, and 19 m). These profiles are located in the top, second,
third, and bottom layers, respectively. In this case, the electric
field strength is essentially constant between the plates,
independent of the depth of the profile, and is equal to 50
kV/m, as could be expected from a simple theoretical
analysis. The magnitude of the current density along those
same profiles is constant between the plates along each
profile, but varies as a function of depth due to variations of
the soil resistivity. This again follows from Ohm’s law
J  E ; since the electric field strength does not vary with
Fig. 4. Electric field strength and magnitude of current density along depth, the magnitude of the current density is inversely
vertical profile across two horizontal plates proportional to the soil resistivity. As shown in Fig. 8, the
potential along those profiles changes linearly with the Fig. 9 clearly shows that the electric field strength is
displacement between the two plates. independent of depth as long as the calculation points are not
too close to the system boundaries, where it gets lower due to
the edge effect. Accordingly, the magnitude of the current
density changes in a stepwise manner when the observation
point moves from one layer to another with a different soil
resistivity between the parallel plates.
B. Electric Field and Potential from a Closed Metallic Box
In theory, the electric field and potential inside a closed
metallic box are not affected by the soil resistivity and the
layer structure. Inside a closed box, the potential should be a
constant (the same as the potential of the metallic box) and
the electric field should be 0 due to the perfect shielding
effect provided by the metallic plates.
To verify that these results can be reproduced with the
approach described in this paper, this section presents
Fig. 7. Electric field strength and magnitude of current density along
computation results for a simple box made of six plates. As
horizontal profiles across vertical plates shown in Fig. 10, the box is a cube with a side length of
100 m. Horizontally, it is centered at X = 0, Y = 0, and the
top face of the box is buried at a depth of 10 m. The thickness
of all plates is 2 cm. The box is held at 50 kV and buried in a
three-layer soil model. The soil resistivity of the top, central
and bottom layers in this soil model are 100 Ω-m, 500 Ω-m,
and 800 Ω-m, and the thickness of the top and central layers
are 11 m and 30 m, respectively.
The computations are carried out along 19 horizontal
profiles ranging from (-150, 0, Z) to (150, 0, Z) with Z = 15,
20, …, 105 m, respectively. The calculated potential along
these profiles is shown in Fig. 11. It confirms that the
potential inside the box is constant at the energization value
of 50 kV. Outside the box, the potential decreases with the
distance from the box. Although this is not readily visible
Fig. 8. Potential along horizontal profiles across vertical plates from the figure, the potential decreases faster along the lower
profile than along the top ones.
The variation of the electric field and current density with
depth for this configuration is illustrated in Fig. 9, which
plots those quantities along a vertical profile ranging from (5,
0, 0) to (5, 0, 40).

Fig. 10. Metallic box in a three-layer soil

The electric field along these profiles is shown in Fig. 12.

Inside the box, the electric field strength reaches 0 almost
everywhere, except at locations very close to the walls the
Fig. 9. Electric field strength and magnitude of current density along
vertical profile between vertical plates
box. Outside the box, it decreases with the distance from the
box. Close to the wall of the box, the electric field strength at calculated capacitance approaches a stable value which can
the lower locations on the walls is higher than that at higher be compared with the result from the analytical formula (𝐶 =
locations, and it decreases faster along lower profiles than 2𝜋𝐿𝜖/ ln 𝑏/𝑎, where 𝐿 is the length of the cylinder, and 𝑏
that along the higher profiles as the observation point moves and 𝑎 are its outer and inner radii, respectively). As an
away from the box. example in Fig. 13, two vertical coaxial cylinders with radii
of 1 m and 3 m are located in the same soil model as in the
previous section. The relative permittivity of the soil in each
layer is assumed to be 1. The cylinders are 200 m long, and
start at a depth of 31 m below the earth surface. To reduce
edge effects, 30 m long guard cylinders are added at both
ends of the cylinders. When using 20-sided polygons to
represent the cylinders, the calculated capacitance of the
coaxial cylinders is 10139.20 pF, which compares very well
with the results obtained using the analytical formula
(10127.56 pF, with a relative error of 0.11%).

Fig. 11. Potential along horizontal profiles across a metallic box

Fig. 12. Electric field along horizontal profiles across a metallic box

C. Capacitance of Parallel Plates and Coaxial Cylinders

The analytical formulae to calculate the capacitance of
parallel plates and coaxial cylinders are well known and can
be used to verify the approach described in this paper. In
Fig. 13. Calculation of capacitance of vertical coaxial cylinders
general, the current leaking out from plate patches and
conductor segments is related to the charge distributed on To verify the calculation of the capacitance for parallel
their surface by 𝑄 = 𝜌𝜖𝐼 where 𝑄 is the surface charge, 𝐼 is plates, two 200 m by 200 m vertical plates are modeled in the
the leakage current, 𝜌 is the resistivity of the medium and 𝜖 same soil as for the previous example (Fig. 14). Guard plates
is its permittivity. The total charge on the plate can be with a width of 10 m are added at the top and bottom of the
obtained by summing the charges on the individual patches. plates in order to minimize the error caused by the edge
The potential difference Δ𝜙 between the two plates or effect. A third guard plate similar as the right plate is placed
cylinders can then be obtained, and the capacitance between 0.5 m away from it in order to collect the current from
the parallel plates or the coaxial cylinders can be calculated outside. The two plates are separated by 10 m and the
by 𝐶 = 𝑄/Δ𝜙. thickness of the plates is 2 cm. The calculated capacitance of
For the vertical coaxial cylinders, the circular shape of the the two plates is 35480.32 pF, which again compares very
cylinders can be approximated by a regular polygon. When well with the value obtained from the analytical formula 𝐶 =
the number of sides in the polygon is sufficiently large, the 𝜖𝐴/𝑑 (where 𝐶 is the capacitance, 𝐴 is the surface area of
the plates and 𝑑 is their separation), which gives 35416.00 injected into the plate. For comparison purposes, the same
pF for a relative error of 0.18%. system is also modeled with the plate being replaced by a
grid of the same size, as shown in Fig. 17. The potential and
touch voltage at the earth surface along a diagonal profile
going from (0, 0, 0) to (70, 70, 0) are shown in Fig. 18, for
both models. The maximum relative difference is 1.57% for
the potential near the corner.

Fig. 14. Calculation of capacitance of parallel vertical plates

D. Ground Resistance of a Circular Plate in a Uniform Soil Fig. 16. 100 m  100 m grounding plate
For a circular metallic plate on the earth surface in a
uniform soil model, the ground resistance can be calculated
with the equation R=/4a, where  is the soil resistivity and a
is the radius of the plate [15]. For a circular plate with a
radius of 1 m, this formula gives a ground resistance of 25 
for a uniform soil with a resistivity of 100 -m. To verify the
approach described in this paper for this case, this circular
plate can be approximated by a group of plates of various
quadrilateral shapes, as shown in Fig. 15. The computed
ground resistance for this model is 25.434 , resulting in a
difference of 1.74% compared with the theoretical value.
Fig. 17. 100 m  100 m, 400-mesh exponential grid

Fig. 18 also shows the results for the plate model using the
method in [13], which is applicable in this case. The
calculated potential and touch voltage are almost same for the
two approaches. The calculated resistance is 2.8366 ,
2.8364  and 2.8368  for the methods in this paper, in [13]
and the grid model, respectively.

Fig. 15. Approximated circular plate

IV. ANALYSIS OF GROUNDING SYSTEMS INCLUDING

PLATES IN HORIZONTAL MULTILAYER SOILS
A. A Grounding Plate
Consider a 100 m  100 m horizontal grounding plate
buried at a depth of 0.5 m in a three-layer soil (Fig. 16). The
soil resistivity in the top, central and bottom layers are 100
-m, 500 -m and 800 -m, respectively. The thickness of
the top and central layer is 1 m and 2 m, respectively. The
plate has a thickness of 0.01 m, and is subdivided into 400 Fig. 18. Potential and touch voltage along a profile for plate and grid
patches. The subdivision is exponential, in the sense that the
size of the patches along each dimension decreases B. A Grounding Plate with Rods
progressively be a constant factor (set to 0.8) when Rods are used quite often in grounding systems in order to
approaching the edges of the plate. A current of 1,000 A is reduce the grounding resistance. The reason is that rods can
effectively discharge a significant fraction of the fault current
into the deeper soil if the soil resistivity of the bottom layers
is not too high. This effect is more significant for multilayer
soils with lower soil resistivities in the layers reached by the
rods. The variation of the current density in the soil can be
used to demonstrate this effect.
The grounding plate described in the previous section is
used as a starting point. The resistivity of the soil layers is
inverted compared to the previous case, i.e. it is 800 -m,
500 -m and 100 -m in the top, central and bottom layers,
respectively. Different numbers of rods with the length of 3
m are uniformly distributed along the edge of the plate. An
example of such configurations is shown in Fig. 19.

Fig. 21. Potential and touch voltage on the earth surface

V. CONCLUSIONS
An analysis of the low frequency behavior of grounding
systems including freely oriented plates of arbitrary shape
located in horizontal multilayer soil models has been
presented. The method of analysis yields not only the scalar
Fig. 19. Grounding plate with rods potential but also the electric field and current density in the
Fig. 20 shows the magnitude of current density along a soil, which can be used to better understand the performance
profile ranging from (0, 50, 4) to (100, 50, 4), at an elevation of grounding systems in order to achieve an optimized
located below the tips of the rods. The figure shows that the design. The contribution of the potential and electric field
current density near the rods increases sharply compared to from metallic plates is accurately obtained by direct surface
the case without rods, especially at the corners of the plate. integration. The approach was verified against analytical
results for various configurations (parallel plates, coaxial
cylinders, closed box), and compared with existing methods
for horizontal or vertical rectangular plates. In addition to its
intrinsic interest in accurately modeling objects such as gates
or tanks, the grounding plate solution is valuable as a limiting
case of an extremely dense grounding grid. It can also yield a
quick estimate of the lowest possible ground resistance and
touch voltages for a grounding system of a given size.

APPENDIX I. INTEGRATION OVER A PATCH

This section presents the formulas needed to calculate the
surface integrals giving the potential and electric field due to
a patch. Fig. 22 shows a trapeze defined by four coplanar
points 𝐴⃑, 𝐵 ⃑⃑, 𝐶⃑, and 𝐷
⃑⃑. The sides of the trapeze are defined
by the vectors 𝑎⃑ ≡ 𝐷 ⃑⃑ − 𝐴⃑, 𝑏⃑⃑ ≡ 𝐵 ⃑⃑, and 𝑑⃑ ≡
⃑⃑ − 𝐴⃑, 𝑐⃑ ≡ 𝐶⃑ − 𝐵
Fig. 20. Magnitude of the current density below rods ⃑ ⃑⃑
⃑ ⃑⃑
𝐶 − 𝐷 . Assume that side 𝑑 is parallel to 𝑏, but may have a
The potential and the touch voltage along a diagonal different length, i.e. 𝑑⃑ = 𝜆𝑏⃑⃑ , with 𝜆 ≥ 0. The case 𝜆 = 0
profile from (0, 0, 0) to (70, 70, 0) on the earth surface are corresponds to a triangle, while the case 𝜆 = 1 corresponds
shown in Fig. 21. Along the profile, the additional rods to a parallelogram that further reduces to a rectangle when 𝑎⃑ ∙
decreases the earth potential above the plate, but tend to
𝑏⃑⃑ = 0. Other values of 𝜆(> 0) give a general trapeze. An
increase the potential a little bit near the plate. The resultant
arbitrary point 𝑟⃑ located in the same plane as the trapeze can
touch voltage above the plate is near 0 as expected. Outside
be written as:
the plate, the touch voltage decreases with additional rods in
   
the system. Meanwhile, the resistance decreases from 0.5772 r  A  au  b v (2)
, to 0.5742 , 0.5726  and 0.5697 , respectively, when
additional rods are added in the test cases.  with 0 ≤ 𝑢 ≤ 1, 0 ≤ 𝑣 ≤ 1 + (𝜆 − 1)𝑢 for points located in
the trapeze. In 𝑢 − 𝑣 coordinates, the surface element 𝑑𝑆 𝑃⃑⃑ = 𝑅⃑⃑ − 𝑅⃑⃑ ∙ 𝑏⃑⃑𝑏⃑⃑/𝑏 2 − (𝑅⃑⃑ ∙ 𝑎⃑⊥ )𝑎⃑⊥ /𝑎⊥ 2 ; 𝑍 = |𝑃⃑⃑|/𝑎⊥ ;
on the trapeze is |𝑎⃑ × 𝑏⃑⃑|𝑑𝑢𝑑𝑣 and the integral of the leakage 𝑋+ = 1 − 𝑅⃑⃑ ∙ 𝑎⃑⊥ /𝑎⊥ 2 ; 𝑋− = −𝑅⃑⃑ ∙ 𝑎⃑⊥ /𝑎⊥ 2 ;
current over the trapeze giving the potential at a point 𝑟⃑0 is 𝑋 −𝑅⃑⃑⃑⃑∙∙𝑎⃑𝑏⃑⃑⊥+
= (−𝑅 /𝑎𝑎⊥⃑ 2∙ ;𝑏⃑⃑𝑅⃑⃑ ∙ 𝑎⃑ /𝑎 2 )/(𝑏𝑎 );
𝑌− =
− ⊥ ⊥ ⊥
⃑⃑|
𝑉(𝑟⃑0 ) =
𝜌𝐼|𝑎⃑⃑×𝑏 1
∫0 𝑑𝑢 ∫0
1−(1−𝜆)𝑢
𝐺𝑑𝑣 (3) ana
𝑌+ = [𝑏 − 𝑅⃑⃑ ∙ 𝑏⃑⃑/𝑏 − (𝑏 − 𝑏𝜆 − 𝑎⃑ ∙ 𝑏⃑⃑/𝑏)𝑅⃑⃑ ∙ 𝑎⃑⊥ /𝑎⊥ 2 ]/𝑎⊥ .
4𝜋
1
where 𝐺 = , 𝑅⃑⃑ = 𝑟⃑0 − 𝐴⃑ , The electric field can be calculated with the usual formula:
√𝑅 2 −2𝑅 ⃑⃑𝑣+𝑎2 𝑢2 +2𝑎⃑⃑∙𝑏
⃑⃑∙𝑎⃑⃑𝑢−2𝑅⃑⃑∙𝑏 ⃑⃑𝑢𝑣+𝑏 2 𝑣 2

𝑅 ≡ |𝑅⃑⃑ |, 𝑎 ≡ |𝑎⃑| , 𝑏 ≡ |𝑏⃑⃑| , ρ is the resistivity of the soil E  V (6)
surrounding the patch and I is the leakage current density where the gradient is to be taken with respect to 𝑟⃑0 .
from the patch. Equations (5) and (6) can be used to calculate the potential
and electric field generated by the leakage current from a
trapeze, parallelogram, rectangle or triangle by specifying
different values of 𝜆 . A patch of general shape can be
subdivided into two triangles for which (5) and (6) can be
used to calculate the generated potential and electric field.

REFERENCES
[1] R. J. Heppe, "Computation of potential at surface above an energized
grid or other electrode, allowing for non-uniform current distribution,"
IEEE Trans. on PAS, vol. 98, no. 6, pp. 1978-1989, Nov.-Dec. 1979.
[2] A. P. Meliopoulos, R. P. Webb, and E. B. Joy, "Analysis of grounding
systems," IEEE Trans. on PAS, vol. 100, no. 3, pp. 1039-1048, Mar.
1981.
[3] P. J. Lagace, J. L. Houle, Y. Gervais, and D. Mukhedkar, "Evaluation
Fig. 22. Analytical integration over a trapeze
of the voltage distribution around torroidal HVDC ground electrodes in
The integration over 𝜈 in (3) is simple, and gives n-layer soils," IEEE Trans. on PWRD, vol. 3, no. 4, pp. 1573-1577,
Oct. 1988.
⃑⃑|
𝜌𝐼|𝑎⃑⃑×𝑏 1 𝑏𝑣+′ 𝑏𝑣−′ [4] T. Takahashi and T. Kawase, "Calculation of earth resistance for a
𝑉(𝑟⃑0 ) =
4𝜋𝑏
∫0 𝑑𝑢 {sinh−1 ⃑⃑|
|𝑁
− sinh−1 ⃑⃑|
|𝑁
} (4) deep-driven rod in a multi-layer earth structure," IEEE Trans. on
PWRD, vol. 6, no. 2, pp. 608-614, Apr. 1991.
⃑⃑ = 𝑅⃑⃑ − 𝑅⃑⃑ ∙ 𝑏⃑⃑𝑏⃑⃑/𝑏 2 − (𝑎⃑ − 𝑎⃑ ∙ 𝑏⃑⃑𝑏⃑⃑/𝑏 2 )𝑢 [5] F. P. Dawalibi and D. Mukhedkar, "Optimum design of substation
with 𝑁 grounding in two-layer earth structure - Part I, analytical study," IEEE
𝑣−′ = 𝛼− + 𝛽− 𝑢; 𝑣+′ = 𝛼+ + 𝛽+ 𝑢 Trans. on PAS, vol. 94, no. 2, pp. 252-261, Mar.-Apr. 1975.
[6] F. P. Dawalibi and N. Barbeito, "Measurements and computations of
𝛼− = −𝑅⃑⃑ ∙ 𝑏⃑⃑/𝑏 2 ; 𝛼+ = 1 + 𝛼− the performance of grounding systems buried in multilayer soils," IEEE
𝛽− = 𝑎⃑ ∙ 𝑏⃑⃑/𝑏 2 ; 𝛽+ = 𝜆 − 1 + 𝛽− Trans. on PWRD, vol. 6, no. 4, pp. 1483-1490, Oct. 1991.
[7] J. Ma, F. P. Dawalibi, and R. D. Southey, "On the equivalence of
(𝑦+𝑤𝑥) uniform and two-layer soils to multilayer soils in the analysis of
The two terms in (4) are of the form ∫ 𝑑𝑥 sinh−1 ,
√ 𝑥 2 +𝑧 2 grounding systems," IEE Proceedings - Generation, Transmission and
whose solution is 𝐹(𝑤, 𝑥, 𝑦, 𝑧), where Distribution, vol. 143, no. 1, pp. 49-55, Jan. 1996.
[8] J. Ma, F. P. Dawalibi, and W. K. Daily, "Analysis of grounding
𝐹(𝑤, 𝑥, 𝑦, 𝑧) = 𝑥𝐹1 (𝑤, 𝑥, 𝑦, 𝑧) + 𝑦𝐹2 (𝑤, 𝑥, 𝑦, 𝑧) systems in soils with hemispherical layering," IEEE Transactions on
PWRD, vol. 8, no. 4, pp. 1773-1781, Oct. 1993.
+ 𝑧𝐹3 (𝑤, 𝑥, 𝑦, 𝑧) [9] J. Ma and F. P. Dawalibi, “Analysis of grounding systems in soils with
𝐹1 (𝑤, 𝑥, 𝑦, 𝑧) = sinh−1 [(𝑦 + 𝑤𝑥)/√𝑥 2 + 𝑧 2 ] cylindrical soil volumes,” IEEE Transactions on PWRD, vol. 15, no. 3,
pp. 913-918, Jul. 2000.
1 (𝑤𝑦 + 𝑤 2 𝑥 + 𝑥) [10] J. Ma and F. P. Dawalibi, “Analysis of grounding systems in soils with
𝐹2 (𝑤, 𝑥, 𝑦, 𝑧) = sinh−1 finite volumes of different resistivities,” IEEE Transactions on PWRD,
√1 + 𝑤 2 √𝑦 2 + 𝑧 2 + 𝑤 2 𝑧 2 vol. 17, no. 2, pp. 596-602, Apr. 2002.
−1 2 [11] F. P. Dawalibi and D. Mukhedkar, "Optimum design of substation
𝐹3 (𝑤, 𝑥, 𝑦, 𝑧) = tan [(𝑧 𝑤 − 𝑦𝑥)/(𝑧𝑆)]
grounding in a two layer earth structure - Part II, comparison between
𝑆 = √(𝑦 + 𝑤𝑥)2 + 𝑥 2 + 𝑧 2 theoretical and experimental results," IEEE Trans. on PAS, vol. 94, no.
2, pp. 262-266, Mar.-Apr. 1975.
The final expression for the potential is then [12] F. P. Dawalibi, J. Ma, and R. D. Southey, “Behaviour of grounding
systems in multilayer soils: a parametric analysis,” IEEE Transactions
⃑⃑|
𝜌𝐼|𝑎⃑⃑×𝑏
𝑉(𝑟⃑0 ) = {𝐹(𝑊+ , 𝑋+ , 𝑌+ , 𝑍) − 𝐹(𝑊+ , 𝑋− , 𝑌+ , 𝑍) − on PWRD, Vol. 9, No. 1, pp. 334-342, Jan. 1994.
4𝜋𝑏 [13] J. Ma, and F. P. Dawalibi, “Computerized Analysis of Grounding
𝐹(𝑊− , 𝑋+ , 𝑌− , 𝑍) + 𝐹(𝑊− , 𝑋− , 𝑌− , 𝑍)} (5) Plates in Multilayer Soils,” IEEE Transactions on PWRD, vol. 24, no.
2, pp. 650-655, Apr. 2009.
where [14] Roger F. Harrington, Field Computation by Moment Methods, IEEE
Press, May 5 1993.
𝑎⃑⊥ = 𝑎⃑ − 𝑎⃑ ∙ 𝑏⃑⃑𝑏⃑⃑/𝑏 2 ; 𝑎⊥ = |𝑎⃑⊥ |;
[15] IEEE Guide for safety in AC substation grounding, IEEE Std 80-2000.
𝑊
𝑊+− =
=𝑎 ⃑ ∙ 𝑏⃑⃑/(𝑏𝑎
−𝑏(1 − 𝜆⊥− ); 𝑎⃑ ∙ 𝑏⃑⃑/𝑏 2 )/𝑎⊥ ; 𝑊− = 𝑎⃑ ∙ 𝑏⃑⃑/(𝑏𝑎⊥ );
 Electromagnetic Shielding Analysis of Buildings
Under Power Lines Hit by Lightning
S. Ladan, A. Aghabarati, R. Moini, S. Fortin and F.P. Dawalibi
Safe Engineering Services and Technologies ltd.
Montreal, CANADA
www.sestech.com

Abstract—The shielding effectiveness of a metallic enclosure with electromagnetic disturbance radiated by direct or indirect
multiple openings sited close to a transmission line tower hit by lightning is susceptible to damage electronic integrated circuits
lightning is investigated, both in the frequency domain and in the used in communication and control apparatus deployed in the
time domain. Two scenarios are analyzed to define the possible surrounding area. Therefore during the last decades, a great deal
sources of interference. First, the entire power network including
of effort has been devoted to the analysis of the distribution of
the tower hit by lightning, shield wires and phase conductors is
considered. Next, the simplified case of a single tower hit by transient currents in lightning protection systems (LPS) and of
lightning is investigated. The results reveal that for a better the resulting transient electromagnetic interference inside
protection of equipment against lightning, the entire power buildings [8].
network should be considered as source of interference for the
study of shielding effectiveness.
In order to determine the threat level to sensitive equipment,
the electromagnetic field radiated by structures hit by lightning
Keywords-Shielding Effectiveness (SE), Lightning, Electromagnetic
Interference, Transmission Line, Transient Response.

I. INTRODUCTION
Lightning is a transient, high-current electric discharge
whose path length extends over several kilometres. The
consideration of cloud-to-ground (CG) lightning events is
important in protection studies of electric and electronic
equipment used in power systems, information technology
systems, etc. Overhead lines in medium and high voltage
transmission and distribution networks are susceptible to be hit
by CG lightning. Therefore, huge lightning currents can be (a)
induced on structures located near their impact zone.
Particularly, the maximum rate of change of a very small
fraction of these currents circulating in grounding and bonding
wires can cause irreversible damage to nearby electronic printed
circuit boards [1]. Thus, in addition to physical damages and
power interruptions caused by lightning strikes, the problem of
electromagnetic interference caused by these currents has to be
investigated carefully for a better protection of electronic
equipment located in their vicinity. Nowadays, the number of
buildings located under power lines in areas of high ground
flash density is increasing, and the protection levels provided
inside these buildings, with sensitive equipment installed, is of
great interest.
Protection of power components against conducted
interferences, for structures hit by lightning has been widely (b)
investigated in literature [2]-[4]. Usually, the use of optimized
Fig. 1. Sketch of the power network hit by lightning and the shielding
grounding network [5]-[7] and surge arresters limits the level of enclosure close to it: (a): Considering the shield wires and the phase
such threats for power components. However, the conductors, (b): Ignoring the shield wires and the phase conductors.
must be determined through an accurate model and the shielding
properties of the enclosure should be expressed by means of
appropriate shielding effectiveness parameters. The aim of this
paper is to analyze the shielding effectiveness of a typical
building located in the vicinity of a power transmission line
tower, hit by a lightning strike. The paper is organized as
follows. In Section II, the system under study is presented in
detail. In Section III, the computation methodology and the
lightning current waveforms used as a source of disturbance are
presented. Sections IV and V are dedicated to the harmonic and
transient analysis of the shielding effectiveness of the building,
respectively. Fig. 2. Current waveform of the studied lightning signal.

II. SYSTEM NETWORK UNDER STUDY

A schematic of the network, including a tower hit by III. COMPUTATION METHODOLOGY AND LIGHTNING
lightning, a grounding network system, the surrounding WAVEFORM
transmission lines and a building in its proximity, is shown in The computation of the electromagnetic fields is performed
Fig. 1. The tower of the three-phase (525 kV) transmission line in frequency domain, for each frequency component contained
has a height of 35 m and its grounding system consists of a in the surge signal. In order to study the radiating effects of
16 m by 16 m loop buried at a depth of 0.5 m and four vertical lightning and to characterize its interference level, a full-wave
ground rods which are driven to a depth of 10 m. The soil is numerical modeling is conducted using the commercial
assumed to be homogeneous with a 1000 Ω-m resistivity, software package, CDEGS (Current Distribution,
relative permittivity of 1, and relative permeability of 1. As Electromagnetic fields, Grounding and Soil structure analysis)
shown in Fig. 1, the tower is in the middle of the studied portion [9]. The computation results for the described problem were
of the power network. The transmission line is modeled for 400 obtained using the HIFREQ and FFTSES modules of CDEGS.
m on either side of the structure: this was found to be The HIFREQ module provides a solution of the electric field
sufficiently accurate for the calculation of electromagnetic integral equation by Method of Moment (MoM) for complex
fields around the tower. The height of the transmission line structures including networks of wires and metallic surfaces
conductors is about 27 m above the ground at the tower [10]. It should be noted, the conducted methodology takes into
account computation of the vector potential and rise of scalar
location, and the height drops to 17 m at the point of maximum
potential associated with any nearby metallic objects. The
sag. The building, located 15 m away from the center of the
presence of a multilayer soil can also be taken account. The
tower, is modeled by a 7 m × 8 m × 13 m enclosure with metallic
removal of low frequency instabilities, which typically affects
walls. The building is a metallic enclosure with multiple
such computation methods, permits the computation of
openings. It has two 3 m × 3 m rectangular apertures located on
inducing effects from 0 Hz to several MHz [11]. The latter
opposite walls and two more triangular openings under the roof, allows an accurate analysis of lightning phenomena.
as shown in Fig. 1-b. The walls, roof, and floor of the building
are realized by means of metallic sheets and the bottom of the The lightning current is injected into the top of the tower.
building is assigned to electrical ground and its potential is set Its waveform is shown in Fig. 2. The time variation of the
to zero. current used in the computations is represented by the sum of
two double exponential functions as follows:
In order to examine the level of disturbance inside the
building resulting from the transient electromagnetic fields
generated by the tower hit by lightning, observation points are 𝐈(𝐭) = 𝟗𝟎𝟎𝟎(𝐞−𝟒𝟎𝟎𝟎𝟎𝐭 − 𝐞−𝟑𝟎𝟎𝟎𝟎𝟎𝟎𝐭 )
(1)
defined over a straight line passing through the window
+𝟏𝟏𝟓(𝐞−𝟏𝟎𝟎𝟎𝟎𝟎𝐭 − 𝐞−𝟐𝟐𝟐𝟎𝟎𝐭 )
openings and covering the regions inside and outside the
enclosure, as shown in Fig. 1. The building is studied under The transient waveform has a rise time of 1.44 𝜇𝑠 and a
two scenarios: half-value time of 24 𝜇𝑠 . A total number of 512 sample
frequencies ranging from 0 Hz to 5.2 MHz are selected from the
 Scenario 1: The shield wire and phase conductors are discrete Fast Fourier Transform (FFT) to represent the lightning
included in the model and their radiation effects as sources surge in the frequency domain. First, the computation is
of interference are considered. performed at each frequency for a unitary current. Then the
 Scenario 2: The shield wires and the phase conductors are application of the Inverse Fast Fourier Transform (IFFT) to the
removed from the system under analysis.
superposition of the values computed in the frequency domain,
modulated by the amplitude of the lightning current spectrum,
provides the transient response of the system. Once the induced
currents are obtained for the wires and surfaces of the entire
network, the radiated electric and magnetic fields of the
energized system are computed for the evaluation of the
induced effects of the LEMP (lightning electromagnetic pulse)
to the regions around the building.

IV. FREQUENCY-DOMAIN EVALUATION OF THE SHIELDING

EFFECTIVENESS
The Shielding Effectiveness (SE) of a shielding structure is
an index used to assess the adequacy of the structure and to
evaluate the protection level it affords against LEMP.
Electromagnetic shielding problems in many case studies are Fig. 5. The electric field distribution (V/m) at frequency of 3.2 MHz.
analyzed in the frequency domain [12]. To calculate the electric
and magnetic SE at any point in frequency-domain, the
following equations are defined

|𝑬𝒂𝒃𝒔𝒆𝒏𝒄𝒆 |
𝐒𝐄𝐝𝐁 (𝐄𝐥𝐞𝐜𝐭𝐫𝐢𝐜) = 𝟐𝟎𝐥𝐨𝐠 ( ) (2)
|𝑬𝒑𝒓𝒆𝒔𝒆𝒏𝒄𝒆 |
|𝑯𝒂𝒃𝒔𝒆𝒏𝒄𝒆 |
𝐒𝐄𝐝𝐁 (𝐌𝐚𝐠𝐧𝐞𝐭𝐢𝐜) = 𝟐𝟎𝐥𝐨𝐠 ( ) (3)
|𝑯𝒑𝒓𝒆𝒔𝒆𝒏𝒄𝒆 |
in which 𝑬𝑎𝑏𝑠𝑒𝑛𝑐𝑒 and 𝑯𝑎𝑏𝑠𝑒𝑛𝑐𝑒 represent, respectively, the
electric and magnetic fields when there is no shielding (without
presence of the building) while 𝑬𝑝𝑟𝑒𝑠𝑒𝑛𝑐𝑒 and 𝑯𝑝𝑟𝑒𝑠𝑒𝑛𝑐𝑒 denote
to the electric and magnetic fields in presence of the shielding

Fig. 6. The magnetic field distribution (A/m) at frequency of 3.2 MHz.

enclosure (the building). To calculate the SE in the frequency

domain, simulations are performed for Scenario 1 and Scenario
2 described in Section II.
Fig. 3 and Fig. 4 depict the electric and magnetic shielding
effectiveness at a point located in the middle of the observation
profile (as shown in Fig. 1), for Scenarios 1 and 2, respectively.
As shown in Fig. 3, the electric and magnetic shielding
effectiveness over the frequency range from DC to 5.12 MHz
Fig. 3. Electric SE and Magnetic SE for the case considering the shield wires for the observed point inside the building presents a highly
and phase conductors (Scenario 1). oscillatory behavior. Indeed, these oscillations are related to the
electromagnetic fields radiated by a fraction of the surge current
propagating along the shield wires to the adjacent towers and
the subsequent reflections from the termination points. The
maximum magnetic SE occurs at 3.2 MHz while the maximum
electric SE happens at 3.5 MHz.
It can be observed from Fig. 4 that the shielding
effectiveness for the same point and over the same frequency
range has fewer oscillations when removing the shield wires
and phase conductors (second scenario). It is worth mentioning
that, despite the unrealistic behavior in this scenario, the results
can still be used to assess the general level of electromagnetic
Fig. 4. Electric SE and Magnetic SE for the case ignoring the shield wires and
phase conductors (Scenario 2).
fields penetrating inside the enclosure when the network is hit
by lightning.
A surface profile parallel to the ground and passing through
the building is selected for the presentation of the spatial
variation of the electromagnetic fields. Fig. 5 and Fig. 6 show
the distribution of the electric and magnetic fields for the
selected surface profile, respectively (top view of the building).
The computation is performed for the frequency where
magnetic SE is maximum (3.2 MHz). As seen in those two
figures, the minimum level of magnetic field inside the building
is in the range of 10-3 A/m while the level of electric field at the
Fig. 7. Time trend of the electric field (Scenario 1).
center is also reduced to 1 V/m. This shows that due to the
presence of the metallic enclosure, the level of electric and
magnetic field inside the building is decreased significantly
compared to that in the exterior regions.

V. TIME-DOMAIN EVALUATION OF THE SHIELDING

EFFECTIVENESS
Electromagnetic shielding analyses are normally performed
in the frequency domain [12]. This approach is particularly
appropriate when the interfering electromagnetic sources have a
harmonic variation. However, for transient electromagnetic
sources such as lightning strikes where EM protection is
important, a frequency domain analysis cannot provide an all-
around and versatile assessment of the ability of the shielding Fig. 8. Time trend of the magnetic field (Scenario 1).
structure. Consequently, in some recent publications new
parameters have been defined and used to evaluate the level of
shielding structures in the time domain [13]. Since many electric
and electronic devices are susceptible to the maximum level
reached by the electromagnetic fields and their induced effects
caused by time varying magnetic or electric flux, the shielding
parameters can be defined in a way to show the reduction of the
mentioned effects. According to [13], the relevant shielding
parameters are the electric peak value reduction shielding
effectiveness(SEE−PR ) and the magnetic peak value reduction
shielding effectiveness (SEH−PR ) defined as

|𝑬𝑴𝒂𝒙
𝒂𝒃𝒔𝒆𝒏𝒄𝒆 (𝒕, 𝒙, 𝒚, 𝒛)|
𝐒𝐄𝐄−𝐏𝐑 = 𝟐𝟎𝐥𝐨𝐠 ( ) (4)
|𝑬𝑴𝒂𝒙
𝒑𝒓𝒆𝒔𝒆𝒏𝒄𝒆 (𝒕, 𝒙, 𝒚, 𝒛)|
Fig. 9. Time trend of the electric field (Scenario 2).

|𝑯𝑴𝒂𝒙
𝒂𝒃𝒔𝒆𝒏𝒄𝒆 (𝒕, 𝒙, 𝒚, 𝒛)|
𝐒𝐄𝐇−𝐃𝐑 = 𝟐𝟎 𝐥𝐨𝐠 ( ) (5)
|𝑯𝑴𝒂𝒙
𝒑𝒓𝒆𝒔𝒆𝒏𝒄𝒆 (𝒕, 𝒙, 𝒚, 𝒛)|

where 𝑬𝑀𝑎𝑥 𝑀𝑎𝑥

𝑎𝑏𝑠𝑒𝑛𝑐𝑒 and 𝑬𝑝𝑟𝑒𝑠𝑒𝑛𝑐𝑒 denote the maximum level of the
electric field at each observation point, with and without the
shielding over the duration of the transient event, respectively,
while 𝑯𝑀𝑎𝑥 𝑀𝑎𝑥
𝑎𝑏𝑠𝑒𝑛𝑐𝑒 and 𝑯𝑝𝑟𝑒𝑠𝑒𝑛𝑐𝑒 represent the magnetic field in
absence and presence of the enclosure. It is worth mentioning
that SEE−PR (SEH−DR ) shows the peak reduction of the electric
(magnetic) field, which can be helpful in designing structures
employed to protect systems and devices sensitive to a specific
level of EM field. The network under study is analyzed in the Fig. 10. Time trend of the magnetic field (Scenario 2).
time domain.
 by lightning is investigated in both frequency domain and time
domain using the CDEGS software package. Two scenarios are
analyzed: a case where entire power network including the
tower, the grounding structure, the shield wires and the phase
conductors are considered when examining the level of
interference, and a case where only the tower with no shield
wires or phase conductors is considered as the source of
interference in the model. The simulation results show that the
electric and magnetic shielding effectiveness in the frequency
domain for the first scenario have some oscillations caused by
multiple reflection between the towers and the mutual coupling
between the shield wires and phase conductors. Moreover, the
Fig. 11. Electric SE and Magnetic SE for Scenario 1. effects of the shield wires and phase conductors are more
apparent in the time domain analysis since the transient electric
and magnetic fields in the first scenario have more oscillations
compared to those in the second scenario.
ACKNOWLEDGMENT
The authors wish to thank Safe Engineering Services &
Technologies Ltd. for the support and facilities provided during
this research work.

REFERENCES

[1] S. Celozzi, R. Araneo, and G. Lovat, Electromagnetic Shielding.

Hoboken, NJ, USA: IEEE, 2008.
Fig. 12. Electric SE and Magnetic SE for Scenario 2. [2] V. A. Rakov and F. Rachidi, “Overview of recent progress in lightning
research and lightning protection,” IEEE Trans. Electromagn. Compat.,
vol. 51, no. 3, pp. 428–442, August 2009.
The results for the electric and magnetic fields at different [3] V. A. Rakov, M. A. Uman, M. I. Fernandez, C. T. Mata, K. J. Rambo, M.
distances from the EM source (the tower) for both scenarios are V. Stapleton and R. R. Sutil, “Direct lightning strikes to the lightning
shown in Fig. 7 to Fig. 10. For Scenario 1, the fast oscillations protective system of a residential building: triggered-lightning
experiments,” IEEE Trans. Power Del., vol. 17, no. 1, pp. 575–586, 2002.
are due to the multiple reflections of the lightning current along
[4] J. Takami; and S. Okabe, “Characteristics of direct lightning strokes to
the tower, while the abrupt changes that occur every 3.3 phase conductors of UHV transmission lines,” IEEE Trans. Power Del.,
microseconds are related to the lightning surge propagation on vol. 22, no. 1, pp. 537–546, 2007.
the shield wires and their induced effects over the phase wires. [5] F. P. Dawalibi, W. Xiong and J. Ma, “Transient performance of
substation structures and associated grounding systems,” IEEE Trans.
The electric and magnetic peak value reduction shielding Ind. Appl., , vol. 31, no. 3, pp.520-527, 1995.
effectiveness for all points on the observation profile are shown [6] F. P. Dawalibi and W. Xiong “Transient performance of substation
in Fig. 11 and Fig. 12 for the first and second scenarios, grounding systems subjected to lightning and similar surge currents,”
IEEE Trans. Power Del., vol. 9, no. 3, pp. 1412-1420, 1994.
respectively. The maximum level of the electric and magnetic [7] F. Grange, S. Journet, S. Fortin and F. P. Dawalibi, “Analysis of
shielding for both scenarios occurs in the middle of the enclosure grounding grids influence on lightning generated magnetic field,”
at a distance of 26 m from the power line network. It is worth International Symposium on Electromagnetic Compatibility (EMC
EUROPE), pp. 828-832, 2013.
mentioning that the horizontal axes of Fig. 11 and Fig. 12
[8] R. Araneo, S. Celozzi, A. Tatematsu and F. Rachidi, “Time-Domain
(varying from 14 m to 34 m) shows the distance of each analysis of building shielding against lightning electromagnetic fields,”
observation point located on the observation line profile from IEEE Trnas. Electromagn. Compat., vol. 57, no. 3, pp. 397-404, June
the central phase of the transmission line, which is considered at 2015.
x = 0 as shown in Fig. 1-b. Moreover, as is seen in these two [9] “CDEGS 15.1.4080”, Safe Engineering Services & Technologies ltd.
Laval, Québec Canada, 2015 [Online]. Available: www.sestech.com.
figures, some negative values are observed for the electric and
[10] W. C. Gibson, “The method of moments in electromagnetics”, CRC press,
magnetic shielding effectiveness. These negative values (or EM 2014.
field amplification) occur at position of apertures (windows) [11] J. F. Lee, R. Lee, and R. J. Burkholder, “Loop star basis functions and a
where the total amplitude of the field has been intensified due to robust preconditioner for EFIE scattering problems,”. IEEE Trans.
the reflection of the waves from the surface of the metallic walls. Antennas Propag., vol. 51, no. 8, pp.1855-1863, August 2003.
[12] Standard Method for Measuring the Effectiveness of Electromagnetic
VI. CONCLUSION Sheilding Enclosure, IEEE Standard 299, Feb. 2007.
[13] S. Celozzi and R. Araneo, “Alternative definitions for the time-domain
The shielding performance of a metallic enclosure with sheilding effectivemess of enclosures,” IEEE Trnas. Electromagn.
apertures located in the vicinity of a transmission line tower hit Compat., vol. 56, no. 2, pp. 482-485, April 2014.
 2016 IEEE PES Asia-Pacific Power and Energy Conference - Xi'an - China

Induced Transient Currents on Stationary Vehicles

under Power Lines Hit by Lightning

S. Ladan, A. Aghabarati, R. Moini, S. Fortin, Y. Li and F. P. Dawalibi

Safe Engineering Services and Technologies ltd.
Montreal, CANADA
www. sestech.com

Abstract-The objective of this study is to evaluate the inducing electrostatic effects to 5. 0 rnA nns when the power network is
effects of an overhead line hit by lightning on a nearby vehicle. operating under steady-state conditions. On the other hand,
The overhead line includes towers and grounding networks and
less attention has been paid to safety issues related to the
it is assumed that one tower of the transmission line is subjected
inducing effects on a vehicle caused by a transmission line
to a lightning strike. A full wave analysis method based on an
under fault conditions, nor when the transmission line is hit by
integral formulation of Maxwell's equations is utilized for the
determination of the hazardous zone around the energized power
lightning. In both cases, very large currents will flow through
lines. The method uses a solution of the electric field integral the network which can result in strong magnetic and electric
equation by the Method of Moment (MoM), and is applicable to fields in the air and soil and in a very large potential rise at and
complex structures including networks of wires and metallic around the tower [4]-[6].
surfaces. The transient current passing through the body while a
person on the ground touches the vehicle, is evaluated in order to
According to IEC 604791-1[7], the level of discharge
verify the shock hazard level in the area around the tower current during a fault that can lead to significant physiological
structure. damages depends on the duration of the fault [8]-[9]. A typical
value can be taken as 50 rnA, for a fault duration of 0. 5 ms.
Index Terms-- Fault analysis, lightning, touch voltage, transient
current, Step voltage. Some strong unintentional muscular contractions, respiratory
difficulties, breathing disorders and reversible heart
I. INTRODUCTION disturbances have been reported for this level of current. While
this current level is larger than the corresponding limit value
High voltage power lines can have important electric and
magnetic induction effects on nearby equipment. For example for steady-state conditions, the discharge current can also be
when a large tractor-trailer is parked near an overhead larger.
transmission line, electric coupling from the power lines The level of electric shock hazard caused by a lightning strike
induces voltages on the vehicles' surface. Since the metallic depends on some factors such as duration of the lightning
body is insulated from the ground by rubber tires, a contact
strike, its maximum magnitude, the ground resistivity and the
between a human body and the vehicle can provide a path for
grounding network of the structure, which determine the
current and cause an electric shock.
distribution of step and touch voltages around the structure.
The analysis of safety criteria for a power network When a lightning strike hits a tower, huge currents will be
operating under steady-state conditions was the subject of injected into the tower. These currents can be on the order of
many research studies [1]-[2]. The guidelines established by thousands of amperes, and are discharged in less than a few
the National Electrical Safety Code (NESC) in the USA [3] milliseconds. In that brief period of time, they can change the
indicate that for a harmless electric discharge, the clearance soil potential and consequently the GPR of the tower.
shall be increased or the electric field shall be reduced by other Therefore, the behaviour of the struck tower during a lightning
means, as required, to limit the induced current due to
event is similar to that under fault conditions, but for shorter

978-1-5090-5417-6/16/ $ 3l.00 ©2016 IEEE

571
periods of time and larger amplitudes. These lightning strikes
induce some undesirable electromagnetic fields which may
damage the electronic equipment in substations and be
dangerous to personnel working nearby. As the Shield wires
electromagnetic fields caused by lightning are fast transients,
it is necessary to estimate their inducing effects in the time
domain. Phase conductors �����:::::1;--
The goal of this paper is to investigate the safety issues for
people standing and working near a transmission line hit by
lightning. To this end, the inducing effects of a complex power
network on a nearby vehicle are expressed in term of transient
currents passing through the body while a person on the ground
touches the vehicle. A large tractor-trailer has been used as (a)

vehicle, to maximize the induction effects. The full wave

analysis method of the CDEGS (Current Distribution, Phase Phase Phase
Conductor A Conductor C Conductor B
Electromagnetic Fields, Grounding and Soil Structure Human Body

Analysis) software package [10] is used for the determination f !;;�

of the hazardous zone around the energized transmission lines.
Tower
t"- ,
The method is based on a solution of the electric field integral
equation by the Method of Moment (MoM), and is applicable
10
����r Truck

to complex structures including networks of wires (power t--

network) and metallic surfaces (truck). The presence of a
5
D, -- -- ---- - -- -1- !;
I I I

multilayer soil can also be taken account. :§:

" i : i
0
� i : i
The paper is organized as follows. In Section II, the system
>- i : i
i : i
under study is presented in detail. In Sections III and IV, the -5
++
I I
-+I
computation methodology and the lightning current I I I
A..- J_J.
I
waveforms used as a source of disturbance are presented. -10
I
I

Section V is dedicated to the transient analysis of the I

I

electromagnetic fields and current passing through the body. -15 -10 -5 0 5

XAxis(m)
II. SYSTEM STUDIED

A schematic view of the studied model is shown in Figure

(b)
1. It includes a large tractor-trailer located near a tower of a
three-phase high voltage transmission line (at a distance of Figure I. Sketch of the transmission line hit by lightning, (a): General
view, (b): Top view.
20 m from the center of the tower), a grounding network
system and a person who is touching the tractor-trailer. The transient response [11-15]. Nevertheless, in this study, for the
tower has a height of 35 m modeled by l. 5 cm radius steel rods sake of simplicity, it is assumed that the tower is located in a
while its grounding system consists of copper conductors with uniform soil with a soil resistivity of 1000 n-m, and with a
a 7 mm radius. The grounding network is made of 16 m by 16 relative permittivity and a relative penneability both equal to
m loop buried at a depth of 0. 5 m and four vertical ground rods l. The vehicle chosen for this analysis has dimensions of 20 m
which are driven to a depth of 10 m. The height of the length, 2. 5 m width, and 4 m height, representing a metallic
transmission line conductors is 27 m above the ground at the structure with a large effective surface area. It is modeled by
tower location while the height drops to 17 m at the point of 61 quadrilateral and rectangular metallic plates. The bottom of
maximum sag. A detailed description of the geometry of the the truck is at distance of 0.6 m from the ground. The human
network can be found in [6]. As shown in Figure 1, the tower body is modeled using a network of wire conductors with
is located in the middle of the studied portion of the impedances of 500 n for the trunk and a 1000 n for each leg.
transmission line. In the model, the transmission line extends A detailed analysis of the system for the determination of the
for 400 m on either side of the structure since this was found location of the unsafe areas around the tower when a fault
to be sufficiently accurate for the calculation of happens on one of the phase conductors can be found in [16].
electromagnetic fields around the tower. It is worth mentioning In this paper, the same area is used for the evaluation of the
that soil characteristics can have a noticeable effect on the inducing effects of lightning on the vehicle.

572
 truck is computed to verify if it meets the safety limits. It should
TABLE I.
be noted that in Figure 2, the coordinate (0, 0) represents the
01 (m) 02 (m) Position attributes location of the center of the tower in the (X, Y) plane, while
Points refer to (refer to extracted from
Figure I) Figure I) Figure 2 coordinate (D" D2) depicts the position of center of truck
relative to the origin, which is at the center of the tower.
PI 5 15 Unsafe

Border of safe and III. COMPUTATION METHOOOLOGY

P2 0 10
unsafe
The computation of the electromagnetic fields is perfonned
P3 20 20 Safe in the frequency domain, for selected frequency components of
the spectrum of the surge signal. In order to study the radiating
P4 10 0 Safe effects of lightning and to characterize its interference level, a
full-wave numerical modeling is conducted using the
commercial software package, CDEGS. The computation
results for the described problem were obtained using the
• 60-80 rnA 15 HIFREQ and FFTSES modules of CDEGS. The HIFREQ
module provides a solution of the electric field integral equation
• 40-60 rnA
by Method of Moment (MoM) for complex structures including
10 D,(rn)
20-40 rnA networks of wires and metallic surfaces [17]. The presence of a
multilayer soil can also be taken account. The removal of low
• 0-20 rnA 5
frequency instabilities, which typically affects such
computation methods, pennits the computation of inducing
o effects from 0 Hz to several MHz [18]. This range of frequency
allows for an accurate analysis of lightning phenomena. The
D,(rn)
transient analysis of the structure is performed by the use of the
Figure 2. Short circuit current around the tower under fault conditions, in FFTSES Fourier analysis module.
rnA.
IV. LIGHTNING WAVEFORM
When the tower is hit by lightning, the network discharges
a huge level of current into the earth. Therefore, the soil To a good approximation, the wavefonn of a lightning surge
potential is raised. When a truck is parked close to the tower, current can be represented using double exponential functions.
the electric field coupling produced by the lightning strike The surge current injected into the top of the tower in this study
induces a voltage on the surface of the vehicle. However, the is shown in Figure 3. The time variation of the current is defmed
rubber tires insulate the metallic body of the truck from the by the sum of two double exponential functions as follows:
ground, but any human contact (for example a hand touch)
provides a path between the truck and the ground. The objective I(t) = 9000(e-40000t _ e-3000000t)
here is to estimate the current passing through the body of a (1)
person while the person touches the truck and determine human +115(e-l00000t _ e-22200t)
hazard levels in tenns of touch and step voltages generated by
the lightning surge. Figure 2 (taken from [16]) depicts the level in which t is in seconds and 1 is in A. This transient wavefonn
of short circuit current from the truck during a transmission line has a rise time of 1.44 /is and a half-value time of 24 /is. A
fault, for different positions of the truck relative to the tower at
which the fault occurs. The results show that the level of short
circuit current is quite noticeable in some places around the '"
e-7000
tower, under the stated assumptions. Therefore, under fault ::!.6000
conditions, some areas around the tower could be unsafe [16]. �
:€!5000
ii
E4000
As shown in Table I, in this paper, the analysis of the system «
'C30DO
under lightning conditions is restricted to four different �
:;2000
u
locations of the truck around the tower. One of these locations 1000
is inside the zone that was determined to be unsafe under fault ol --�� --�_,-
, �==�==�
o w w � � M M ro M � �

Time [/JS]
conditions, two are inside the zone that was determined to be
safe under fault conditions and one is on the boundary between
Figure 3. Short circuit current around the tower under fault conditions,
those two regions. For each of those locations, the level of in mA.
current passing through the body of the person who touches the

573
total number of 512 sample frequencies ranging from 0 Hz to 20 20
5.2 MHz are used to calculate the discrete Fast Fourier -Pl

Transform (FFT) representing the lightning surge in the 15

-P2
-P3 �15
frequency domain. Next, the computation is performed at each �
-P4
� 10
� 10 U 5
:J

frequency for a unitary current. The electromagnetic fields and

:J
the longitudinal current passing through the network of U 2 4 6 8 10
conductors are then obtained through an Inverse Fast Fourier 5 Time [psI

Transform (lFFT) applied to the product of the frequency

oo������ ===----- ----�
domain unitary response by the spectrum of the lightning 10 20 30 40 50 60 70 80 90 100
Time [1'5]
surge. Once the induced currents are obtained for the wires and
surfaces of the entire network, the radiated electric and Figure 4. Transient current passing through the hand.
magnetic fields of the energized system are computed for the 14 ;
15
evaluation of the induced effects of the tower hit by lightning 12 l -Pl
-P2
-P3 �10
in the regions around the truck. -P4 C
�
V. ANALYSIS AND RESULTS
<3 5

When a structure such as a tower or a building is struck by 6 8 10

Time [J1sJ
lightning, hazardous conditions can arise for persons standing
near the structure. Indeed, the grounding system performance
may deteriorate due to the presence of high frequency 10 20 30 40 50 60 70 80 90 100
Time ["SJ
components in the spectrum of the lightning currents which is
Figure 5. Transient current passing from the feet to the ground.
dissipated in the soil through the grounding system. Therefore,
dangerous values of the step and touch voltages are expected in which ti is the duration of the current in ms and IB is the
around the structure. To reduce the safety hazards to persons, current passing through the body in rnA. Based on the IEC
the level of step and touch voltages should be kept at a tolerable 60479-2, if the current is such that it falls below C" then the
level by installing a good quality grounding system. To evaluate area is considered safe since below C1 there is no fibrillation.
the shock hazard caused by lightning, the level of the transient Between CI and C2 the risk of fibrillation is less than 50%.
current passing through the body of the person who touches the Above C3 is the area in which the risk of fibrillation exceeds
truck can be examined. To do so, the transient currents passing 50%. Based on the obtained results shown in Figure 4 and
through the hand and the feet of the person who touches the Figure 5, all the examined locations can be considered safe
truck is computed for the four locations listed in Table I, to (since the level of current and its duration are very small) except
verify if they meet the safety limits. Figure 4 and Figure 5 show PI where the level of the transient current for the hand of a
the transient behavior of the current passing through the hand human body touching the vehicle can exceed 18 A for a
and the feet of the person standing at those four locations. In the duration of 10 J-lS. In order to determine in which zone the
present case study, two approaches are used to investigate the current occurring at point PI lies and see if this point is unsafe
safety level of a person who is touching the truck. The first or not, the rms value of the current (10 A) which, for a time
approach is based on the IEC standard 60479-2. This standard duration of 10-2 ms, results in a value of 4365as computed with
evaluates the effect of current on human beings and livestock Eq. (2). This is below C1, which is a safe region with no
and determines the threshold level of ventricular fibrillation. In fibrillation. Note that in the present study, the evaluation of the
IEC 60479-2, three zones namely; CI, C2 and C3, have been shock hazards were performed for a lightning surge with an
determined to categorize the threshold of ventricular fibrillation amplitude of 8 kA. For lightning surges with larger amplitudes,
for time durations ranging from 0. 1 ms to 10 ms. The equations the results could change in such a way that PI might be located
describing these zones are approximately: somewhere beyond C3.

The second approach is considering the method proposed

by Dalziel [20]-[21]. In his study, he calculated the minimum
energy criteria shown in TABLE II [20]. He showed that the
energy dissipated in a human body when a person is touching a
(2) conductor energized by a transient current is approximated by

WB = RB i i� (t)dt (3)

574
 [6] W. Ruan, F. P. Dawalibi, J. Ma, and S. Fortin, "Study of electromagnetic
TABLE II
fields and current distribution near a transmission line tower subjected to
Minimum Fibrillation Energy (J) unbalanced currents and phase to ground faults," Proceedings of the 57th
Annual Meeting of the American Power Conference, vol. 57-I, Chicago,
Body Resistance 500 (n) 1000 (n) April 18-20, 1995, pp. 176-181.
[7] lEC 604791-1, Effect of current on human beings and lives stock Partl:
Impulse 27 54 General aspects 2005.
[8] F. P. Dawalibi, S. Fortin and Y. Li, 'Transient response of shielded
substations hit directly by a lightning strike", The 18th Conference of the
where iB is the rms value of the current passing through the Electric Power Supply industry (CEPSI), Taipei, Taiwan, October 25-
body and RB is the body resistance (as mentioned before, in the 28, 2010.
present case the body resistance is assumed to be 1000 Q). [9] S. Visacro, "A comprehensive approach to the grounding response to
lightning currents," iEEE Trans. Power Del., vol. 22, no. I, Jan., 2007.
Referring to P J, the calculated dissipated energy in the body for [10] "CDEGS 15.1.4080", Safe Engineering Services & Technologies ltd.
a period of 10 j1s, is 1.1 J which satisfies the safety level Laval, Quebec Canada, 2015 [Online]. Available: www.sestech.com.
[II] Shunchao Wang, Jinliang He, Bo Zhang, Shuiming Chen, Zhanqing Yu,
mentioned by Dalziel and shows the safety of the region. "Numerical electromagnetic analysis of lightning protection system over
lossy ground", ASia-Pacific Symposium on Electromagnetic
VI. CONCLUSION Compatibility & i 9th international Zurich Symposium on
Electromagnetic Compatibility, 19-22 May 2008, Singapore.
The inducing effects of a transmission line tower hit by [12] Leonid Grcev, "Modeling of grounding electrodes under lightning
lightning on a nearby vehicle were expressed in term of currents", iEEE Trans. Electromagn. Compat., Vol. 51, No. 3, August
2009.
transient currents passing through the body while a person on
[13] F. Menter and 1. Grcev, "EMTP-Based Model for Grounding System
the ground touches the vehicle. A full-wave analysis method Analysis", iEEE Trans. Power Del., Vol. 9, October 1994.
was used to assess the hazardous areas around the tower. To [14] Y. Liu, M. Zitnik, and R. Thottappillil, "An improved transmission-line
model of grounding system," iEEE Trans. Electromagn. Compat., vol.
this end, first the locations of the unsafe areas around the tower 43, no. 3, Aug. 2001.
was determined under fault conditions. Then, various positions [15] F. Rachidi, W. Janischewskyj, A. M. Hussein, C. A. Nucci, S. Guerrieri,
of the truck within the region determined to be unsafe under B. Kordi, and 1. -So Chang, "Current and electromagnetic field associated
with lightning-return strokes to tall towers," iEEE Trans. Electromagn.
fault conditions were used for the lightning study case. It was Compat., vol. 43, no. 3, Aug. 2001.
shown that for a lighting surge with an amplitude of 8 kA, the [16] S. Ladan, A. Aghabarati, R. Moini, S. Fortin and F. P. Dawalibi,
"Induced disturbances by high voltage transmission lines on nearby
level of the transient current passing through the hand of a
stationary vehicles", accepted and presented at Electromagnetic
person touching the vehicle at various locations around the Compatibility Conference (EMC 2016), July 2016.
tower, meets the safety thresholds and all the studied region [17] 1. F. Lee, R. Lee, and R. J. Burkholder, "Loop star basis functions and a
robust preconditioner for EFIE scattering problems," IEEE Trans.
are considered safe with no risk of fibrillation. However if the Antennas Propag., vol. 51, no. 8, pp.1855-1863, August 2003.
amplitude gets 10 times stronger, the studied points will shift [18] W. Ruan, F. P. Dawalibi, J. Ma, and S. Fortin, "Study of electromagnetic
to the region of high risk of fibrillation. fields and current distribution near a transmission line tower subjected to
unbalanced currents and phase to ground faults," Proceedings of the 57th
Annual Meeting of the American Power Conference, vol. 57-I, Chicago,
REFERENCES April 18-20, 1995, pp. 176-181.
[19] F. Grange, S. Journet, F. P. Dawalibi, S. Fortin and Y. Li, "Investigation
of transient voltages in a Tokamak building hit by lightning", 7th Asia­
[I] J. P. Reilly, "Electric field induction on long objects - a Methodology for Pacific International Conference on Lightning (APL 20i 1), Chengdu,
transmission line impact studies", iEEE Trans. Power App. Syst., vol. China, November 1-4, 20II.
PAS-98, No.6 Nov. / Dec. 1979. [20] Li Yexu, Farid Dawalibi, Simon Fortin, Zhigang Wang, Lan Wu and
[2] IEEE Guide for Safety in AC Substation Grounding, IEEE Standard Std. Yuyue Liu, "Transferred transient voltages on the nearby facilities when
80-2013. a lightning stroke hit bridge", Conference on Lightning Protection
[3] National Electrical Safety Code C2-2012, pp.92-93, Institute of (ICLP).
Electrical and Electronic Engineers, Inc., New York, 20II. [21] Charles. F. Dalziel, "A study of the hazards of impulse currents", AlEE
[4] IEEE Std 80-2013: IEEE Guide for Safety in AC Substation Grounding. October, 1953.
[5] Electricity Association Technical Specification 41-24: Guidelines for the
Design, Installation, Testing and Maintenance of Main Earthling
Systems in Substations.

575
 2014 International Conference on Lightning Protection (ICLP), Shanghai, China

Safety of Wind Farm Grounding Systems under Fault

and Lightning Currents
François GRANGE, Sébastien JOURNET Rouzbeh MOINI, Farid Paul DAWALIBI
SES-EUROPE Safe Engineering & Services
Vienne, France Laval, Canada
[email protected] [email protected]

Abstract— This article discusses the safety issues related to wind transfer must be achieved without exceeding the operating
farm grounding systems. Two different cases are considered. The limits of the equipment, electrical, thermal and mechanical
first case treats a wind turbine subjected to a single phase to properties of the materials constituting it, neither affecting the
ground fault while the second case treats a wind turbine hit by continuity of service of the wind power plant. Various works
lightning. In the first case, a detailed analysis of the fault current have been dedicated to safety of wind turbines under fault
distribution for various scenarios, is presented. A comparative conditions without a full analysis of the fault current
study of the touch and step voltages has demonstrated that the distribution [3]-[7]. Nevertheless, the grounding systems of
fault conditions on HV side leads to dangerous touch and step wind turbines hit by lightning are facing transient, high-current
voltages. A focus is made on the human safety regarding
electric discharges. It is also mandatory that the grounding
lightning surges. In this regard, the transient currents flowing
through the legs and body are computed for two different
system quickly disperses lightning currents without dangerous
lightning waveforms. A comparative study of these currents for heating or electrodynamics effects. Therefore, they must avoid
the cases of a single and two turbines reveals the non-validity of damages to the facility for very adverse conditions.
one of the design criterion established in international standards. A number of paper have been devoted to the transient
performance of wind turbine grounding systems hit by
Keywords-component; wind power plant, grounding, fault current, lightning [5]-[7], but few of them has considered the human
lightning, touch voltages.
safety issue. As the main specificity of wind turbines is their
I. INTRODUCTION accessibility to public entities, a safety performance evaluation
of their grounding systems during direct lightning strikes is
The critical need for green energy production has resulted primordial. Contrary to the fault conditions, the determination
to a very rapid growth of wind turbines technology. Due to the of the safety limits for wind turbines hit by lightning is
increasing number and size of wind turbines, the probability of complex. According to related international standards [1], the
lightning strikes to wind farms is rising, while wind turbines grounding system of a wind turbine with a resistance lower
are vulnerable to lightning stroke. As lightning is a high than 10 Ω might be adequate for lightning protection purposes.
transient current, significant voltages induced between any But safety limits for touch and step potentials are not defined
two points of the grounding grid may cause damage to clearly.
equipment and may be dangerous to personnel working nearby
Indeed, we propose in this article a demonstration of non-
[1]-[2]. validity of this design criterion with regard to safety issues. To
When a fault occurs on the grounding network, it is this end, the full wave analysis method of CDEGS (Current
necessary to keep touch and step voltages and ground potential Distribution, Electromagnetic Fields, Grounding and Soil
rises of the installation at a level which does not affect the Structure Analysis) software package [9] is utilized for the
people safety until the outbreak of the protective equipment determination of the induced transient current by lightning. The
and the interruption of the flow of fault current. method uses a solution of the electric field integral equation by
According to IEC 61400-24:2010 [1], the global grounding the Method of Moment (MoM), and is applicable to complex
system shall ensure people safety and equipment integrity structures including networks of wires and metallic surfaces.
regarding two different aggressions: 50 Hz fault conditions, The presence of a multilayer soil can also be taken account.
and lightning strikes. For both cases, safety performance The design methodology introduced aims to integrate the
concerns the capability of the grounding system to maintain various phenomena affecting the performance of earthing
touch and step voltages below the tolerable values as defined systems with appropriate numerical calculations. We address in
per the applicable standards such as IEEE Std 80-2000 [2]. this article two main issues:
Thus, in the general specifications of independent producer  How to take into account the worst fault location?
of renewable energy, it is mentioned that the grounding system
 How to demonstrate that ventricular fibrillation
of the wind power plant shall be able to carry currents to earth
thresholds are met during lightning strikes?
under normal, fault and transient conditions. This current
 Accordingly, the paper is organized as follows. In section II, In this paper, the analysis is restricted to a system
the system under study is presented in detail. In section 3, the including four wind turbines and one collecting substation as
safety issues for different scenarios of fault conditions at low described in Figure 1. In addition, the collecting substation is
frequency, will be examined. Section IV is dedicated to the connected to a high voltage substation located 2 km away. All
safety issues of wind turbine hit by lightning. wind turbines are interconnected through underground cables
to the MV collector without any transformer.
II. DESCRIPTION OF THE SYSTEM
B. Computer modeling
Wind turbines farm grounding system consists in a network
of many small earth-termination systems distributed in a wide The system under study is examined when a fault occurs in
area with various soil types and characteristics. The earth- an area close to the substations or the wind farm collector
termination systems of each wind turbine are generally systems. It is also assumed that the system is subjected to a
interconnected to form a wide network spread across the wind single phase to ground fault. The main purpose of the
power plant. simulation is to estimate the worst ground potential rise at the
locations of wind turbines and the substation.
A. Wind Power Plant characteristics
To this end, the construction of a simplified but realistic 3D
Located closed to Paris in France, the wind farm power model of the entire wind farm including the substations, and a
plant is situated on a land with an approximately area of 5 km2 uniform soil model is needed.
and consists of ten wind turbines. The wind farm delivers its
generated power to a high voltage 63 kV grid that is connected
to the 20 kV collector network through a 75 MVA transformer
located in the substation.
Figure 1. represents a schematic of the system during the
fault conditions scenario at one of the wind turbines.

G E7 HV/MV substation

HV/MV
MV cable with shield
grounded on both sides
G
E8
MV cable with shield
MV grounded on both sides
MV collector + earthing cable
collector
G
E9
Wind turbines

G
E10
Figure 2. Global earthing network model

Accordingly, the HIFREQ module of CDEGS software

package [9] is used as computing engine to simulate the
electromagnetic behavior of topography described by Figure 2.
Figure 1. Schematic of the problem under study The method uses a solution of the electric field integral
equation by the Method of Moment (MoM), and is applicable A. Assessment of wind turbines grounding resistances
to complex structures including networks of wires and metallic The grounding resistance of each wind turbine is
surfaces. determined from the methodology described in previous
For the simulation purpose, a network of underground chapter. TABLE I. depicts the value of computed grounding
conductors is built. The reinforcement steel rebars of resistances for locations E7 to E10, and the MV collector.
foundation are connected to the grounding mesh electrode.
They are modeled as steel conductors coated with a 5 cm thick, TABLE I. GROUNDING RESISTANCES
300 Ω.m resistivity material and bare buried conductors, Location Grounding resistance
respectively. E7 1.7 Ω
Figure 3. presents a perspective view of the model of each E8 1.64 Ω
wind turbine foundation.
E9 1.8 Ω

E10 1.76 Ω
steel rebars in concrete
staircase MV collector 5.1 Ω

copper These values can be used in a global system to take into

account for local soil conditions.
B. Fault current distribution analysis
Previously, we mentioned that the turbines are electrically
interconnected to each other via a buried bare copper
conductor. If a fault occurs on a wind turbine, this conductor
will participate to the fault current distribution which will then
be distributed in all ground systems. In addition, an
interconnection between the HV/MV substation and the MV
collector via the MV cables screens has to be considered.
Figure 3. Perspective view of the wind turbine foundation model Hence, a fault occurring both on the HV side or MV side of
substation can spread to the MV collector and thus probably to
C. Soil resistivity measurements and interpretation each turbine. The fault conditions on HV side will affect the
Indeed, the soil resistivity varies throughout the year, with transferred potentials which depend to the soil resistivity, the
both moisture content and soil temperature. Thus, the soil distance between facilities, connections of the transformers and
resistivity measurements were made during the summer for the nature of the connecting cables.
each wind turbine and the collector in order to consider the
“worst” case for simulation purpose. Five measurements with a A total of 6 different fault cases were examined, one on the
maximum spacing of 20 m and 8 m were carried out using the 63 kV substation side and five on the 20 kV side at each wind
4-pins Wenner method. In principle, soil resistivity turbine site and MV collector. For each scenario, the GPR
measurements should be made up to a spacing (between the (Ground Potential Rises) is observed at all wind turbines sites
current and adjacent potential electrodes) which is at least on and at the MV collector location. In this paper, first a MV fault
the same order as the maximum extent (or diagonal length) of occurring at the MV collector is analyzed.
the grounding system under study. The RESAP module of
TABLE II. 300 A FAULT AT MV COLLECTOR
CDEGS software package is used for the transformation of soil
resistivity measurements into a layered soil model that can be MV collector E7 E8 E9 E10
used by HIFREQ. Maximum
102.2 19.1 37.7 58.8 57.4
GPR (V)
First, an average resistivity value of 2500 Ω-m is obtained Maximum
20 29.7 12.7 8.7 8.1
from measurements. Then a soil model is built for each wind current (A)
turbine in order to assess safety performances of the grounding
systems. Then an HV fault occurring at the MV collector is
considered. Table II and Table III depict the maximum GPR
III. FAULT CONDITIONS and current of each wind turbine when a 300 A fault and a 12
In this section we present a detailed analysis of the fault kA fault occur at MV collector, respectively.
current distribution in order to determine the most conservative
scenario. The main objective is to evaluate the consequences of TABLE III. 12 KA FAULT AT MV COLLECTOR
transferred potentials from HV (High Voltage) side to MV MV collector E7 E8 E9 E10
(Medium Voltage) side. Maximum
594.2 286.3 116.3 85.7 77.0
GPR (V)
Maximum
116.5 168.4 70.9 47.6 43.8
current (A)
 A comparison between TABLE II. and TABLE III. shows Single-Electrode/Reach Touch Voltages/Worst Spherical [ID:pdl_potentiel_HTB]
LEGEND
12.5
that the most restrictive case happens for a HV fault as the Maximum Value : 65.516
Minimum Threshold : 190.500
GPR is about six times higher.
7.5
C. Tolerable body current limit
Touch and step voltages thresholds are determined without 2.5

Y AXIS (METERS)
taking into account the personal protective equipment (gloves,
shoes). These values are determined from the IEC 60479-1 by
-2.5
considering the following assumptions:
 hand to foot current paths, -7.5

 probability of ventricular fibrillation 5%,


-12.5
surface layer of gravel or asphalt not taken into -12.5 -7.5 -2.5 2.5 7.5 12.5
account. X AXIS (METERS)
R-Touch Voltage Magn. (Volts) [Wors]
The fault current duration is set at 0.8 s for a HV fault and
0.2 s for a MV fault. TABLE IV. presents safety criteria for Figure 5. Touch voltage distribution - MV fault
several cases.
Figure 6. and Figure 7. show the distribution of the step
TABLE IV. SAFETY CRITERIA voltage for a one meter inter feet spacing for the same fault
scenarios. As seen in these figures, the maximum step voltage
Threshold values Touch voltage Step voltage for HV fault is about 113.7 V which is above the threshold
MV HV side 46.9 V 57.1 V value of 57.1 V. While, the maximum step voltage is about
collector MV side 190.5 V 246.5 V 19.5 V for MV fault which is below the threshold value of
246.5 V. Single-Electrode/Step Voltages (Spherical)/Worst Spherical [ID:pdl_potentiel]
LEGEND
12.5
D. Safety performances Maximum Value : 113.671
Minimum Threshold : 57.100

In this section, a comparison between computed results and 113.67

7.5
threshold values is performed for both faults scenarios from the 108.01
63 kV and 20 kV substation sides. Safety near the MV 102.36
collector site is evaluated based on the worst case of GPR 2.5
96.70
Y AXIS (METERS)

values. 91.04

-2.5 85.39
Figure 4. and Figure 5. show the distribution of the touch
79.73
voltage one meter away from any metallic structures for the
74.07
HV fault and MV faults, respectively. -7.5
68.41
Single-Electrode/Reach Touch Voltages/Worst Spherical [ID:pdl_potentiel_HTB]

62.76
LEGEND -12.5
Maximum Value : 169.174 -12.5 -7.5 -2.5 2.5 7.5 12.5
Minimum Threshold : 46.900
X AXIS (METERS)
169.17 Step Voltage-Worst Magnitude (Volts)
7.5
156.95 Figure 6. Step voltage distribution - HV fault
144.72 Single-Electrode/Step Voltages (Spherical)/Worst Spherical [ID:pdl_potentiel_HTB]
LEGEND
132.49 12.5
Y AXIS (METERS)

Maximum Value : 19.551

Minimum Threshold : 246.500
120.26

2.5 108.04
7.5
95.81

83.58

71.35 2.5
Y AXIS (METERS)

59.13
-2.5
-6.2 -1.2 3.8 -2.5

X AXIS (METERS)
R-Touch Voltage Magn. (Volts) [Wors]
-7.5
Figure 4. Touch voltage distribution - HV fault
-12.5
For the case of HV fault, the maximum touch voltage is -12.5 -7.5 -2.5 2.5 7.5 12.5
about 169.1 V which is above the threshold value of 46.9 V. X AXIS (METERS)

The maximum touch voltage is about 65.5 V for the case of Step Voltage-W orst Magnitude (Volts)

MV fault which is below the threshold value of 190.5 V. Figure 7. Step voltage distribution - MV fault
 The performed analysis highlights the need to investigate A. Tolerable body current limit
the worst fault scenario. Indeed, it was shown that the HV fault The objective is to estimate the people hazards levels
leads to dangerous touch and step voltages whereas the MV generated by lightning surges. This requires knowledge of the
fault case does not exhibit unsafe issues. effects of surges on human beings and animals. In accordance
IV. WIND TURBINES HIT BY LIGHTNING with IEC 60749-2, to determine the hazard threshold we use, as
parameter, the lower specific energy limits.
Lightning strike to a wind turbine with a large grounding
resistance creates very high ground potential rises (GPR), 1) Determination of specific fibrillating energy Fe
which may lead to intolerable touch and step voltages.
Furthermore, a large differential voltages between power IEC Standard 60479-2 deals with the effects of current on
cables phase conductors and their sheaths, power transformers human beings and livestock due to unidirectional single
phase conductors and their grounded parts can be expected [6]. impulse currents of short durations. The standards of the IEC
60479 series combine two variables for evaluating the effect on
Unlike designing grounding system for a fault at industrial the human body:
frequency, a more complex model of wind turbine that can take
into account all lightning-related phenomena, is needed.  specific energy: Fe in A2.s,
Previous studies have shown that the earth impedance and  charge: Q in A.s
corresponding GPR are much higher at high frequencies when
accounting for the tower structure [8]. Their results also The specific energy is used because it represents the energy
indicate that it is necessary to account for the above-ground dissipated by the lightning current in a unit resistance as shown
structure for more accurate prediction of voltages developed at in Equation 1.
the turbine base. In the following, we are going to use the 𝐸(𝑗) = 𝑍(Ω) ∗ 𝐹𝑒 (𝐴2 𝑠) (1)
capabilities of HIFREQ to perform the full wave analysis of a
wind turbine hit by lightning. Due to transient behavior of Where E is the electric energy in Joule, and Z is the impedance
lightning, the FFTSES module of CDEGS software package is of current path (we consider later in the study a body
also utilized for the transient analysis. To this end, first a impedance value of 500 Ω and a legs impedance value of 1000
frequency decomposition of the transient (lightning Ω)
waveforms) is performed, and then the time-domain response The specific fibrillating energy for rectangular impulses is
of the system under study is reconstituted by a multiple determined by:
frequency analysis of the structure using HIFREQ. Figure 8
𝐼𝐶(𝑝)
depicts the wire grid model of the wind turbine and its 𝐹𝑒 = 𝐼2𝐷𝐶 × 𝑡𝑖 with 𝐼𝐷𝐶 = . (2)
√6
grounding system, which is used in HIFREQ.
Where ti is the shock duration in second and Ic(p) is the peak
current.
A comparison of the current magnitudes for rectangular and
sinusoidal impulses and for a capacitor discharge with the time
❷ constant T having the same specific fibrillating energy Fe and
the same shock-duration ti is performed in Figure 17 of IEC
❶ Standard 60479-2.
2) Threshold of ventricular fibrillation

The threshold of intensities as a function of pulse duration

leading to avoid any risk of ventricular fibrillation is indicated
in the Figure 20 of IEC Standard 60479-2. These are the effects
associated with unidirectional current pulse of short duration.
The curves indicate the probability of fibrillation risk for
current flowing through the body from the left hand to both
feet. For other current paths, see 5.9 in IEC 60479-1.
Insofar as the pulse duration identified in the human body
may be less than 0.1 ms, we have to extrapolate the curve C1
Figure 8. Perspective view of the wind turbine model – zoom on foundation though this pushes us outside the scope of the standard. Indeed,
the IEC 60479-2 standard deals with the biological effects of
The building structure is assumed to have been struck by a
an electric pulse with a duration of 100 µs to 10 µs for a
standard double-exponential lightning waveform. The lightning
unidirectional current [10]. Despite standardization available
strike is modeled by injecting a current surge at the top of the
for the evaluation of allowable values in the lightning field we
highest blade. The direct lightning strike i(t) is simulated by an
use the curve C1. It is therefore more appropriate to rely on the
ideal current source.
geometric mean for a peak pulse. Hence, by considering the
area below C1, we get the following condition to avoid any 350
fibrillation risk due to touch voltage: Single turbine
300

Tansient current (A)

−3 1.4
𝐼1.4
𝐵 × 𝑡𝑖 ≤ 1.5 × 10 𝐴 . 𝑠. (3) Two turbines with crushed stone
250
Where IB is the current flowing through the body.
200
Regarding step voltages, we retain a current factor F of 150
❶
0.04. It leads to the following condition:
1.4
100
𝐼1.4
𝐿 × 𝑡𝑖 ≤ 0.137 𝐴 . 𝑠. (4)
50
Where IL is the current flowing through the legs.
0
B. Safety performances 0 10 Time (µs) 20 30
In the following, assessment of human safety is
investigated for several configurations and two different Figure 10. Current flowing through the body as a function of time – 0,25/100
waveforms. µs waveform
First, we consider a conservative case including only one
wind turbine. Lightning currents are based on LPL I (lightning 200
protection level 1), two different waveforms, have been used
Single turbine

Tansient current (A)

for simulation purpose. They are characterized by their peak
150 Two turbines with crushed stone
values (200 kA and 50 kA), front time (10 μs and 0.25 µs) and
time to half value (350 μs and 100 µs). Figure 9. shows the
influence of the lightning waveforms on the transient current 100
flowing through the body of someone touching the staircases. ❶
50
350
300 0
10/350 waveform
Tansient current (A)

250 0 10 20 30 40
0,25/100 waveform Time (µs)
200
150 Figure 11. Current flowing through the body as a function of time – 10/350 µs
waveform
100
50 25
Single turbine
Tansient current (A)

0 20
0 10 20 30 40 Two turbines
Time (µs) 15
❷
Figure 9. Current flowing through the body as a function of time 10

Due to the grounding system reactance, current peak values 5

are larger for high frequency excitation currents.
Some authors have studied the effect of interconnections 0
[5], [8]. It appears that the contribution of interconnections to 0 10 20 30
the reduction of maximum GPR is limited by their effective Time (µs)
length. Thus, in this paragraph, a comparison between
configurations with one or two turbines is made. Figure 10.
Figure 13. present the currents flowing through the body and Figure 12. Current flowing through the legs as a function of time – 0,25/100
the legs respectively for both lightning current waveforms. A µs waveform
comparison is performed between cases including only one
turbine and two turbines.
 25 V. CONCLUSIONS
Full-wave analysis of wind turbines grounding system has
Single turbine been presented to verify its ability to provide a safe response to
20
Tansient current (A)

Two turbines 50 Hz fault and lightning current.

15 It was established that, for low frequency currents, the
magnitude of GPR and hence touch voltages are significantly
10 higher when the HV fault is taken into account. Depending on
❷ the soil resistivity, distance between facilities and type of
connections, some transferred potentials can occur. This is why
5 specific studies shall be performed for each project.

0 Regarding the lightning, human safety is examined in terms

of energy and current transferred to a person subjected to step
0 20 40 60 80 100 and touch potential mechanisms. The influence of the current
Time (µs)
waveform and interconnections between individual wind
Figure 13. Current flowing through the legs as a function of time – 10/350 µs turbines grounding systems were evaluated. It has been shown
waveform that, even with an individual wind turbine grounding resistance
as low as 1.7 Ω, dangerous touch and step voltages can exist.
C. Results and discussion We strongly recommend this methodology to limit the risks as
A synthesis of previously presented results is made here. A far as possible since no risk at all in lightning protection does
comparison between simulation results and tolerable values is not exist.
performed. As perspective, future work will be performed to show that
lightning energy can be successfully dispersed into the ground
TABLE V. RESULTS FOR A SINGLE TURBINE in spite of individual wind turbines grounding system
Peak RMS resistance larger than 10 Ω.
Case ti IB1,4*ti Safety
current current
❶ 0.25/100 µs 293 119.6 2.3E-5 1,86E-2 NO ACKNOWLEDGMENT
❶ 10/350 µs 180.4 75.1 4.4E-5 1.64E-2 NO SES-EUROPE thanks AKUOENERGY for granting to use
❷ 0.25/100 µs 22.9 9.3 6.2E-5 1.4E-3 YES results from a real project based in France.
❷ 10/350 µs 23.1 9.4 4,35E-4 1E-2 YES REFERENCES
As safety criteria are satisfied for step voltages (❷), a [1] IEC 61400-24, Wind Turbine Generator Systems – Part 24: Lightning
crushed stone layer is added for touch voltages (❶). TABLE protection (2010).
VI. presents touch voltage results for two turbines [2] IEEE Guide for safety in AC substation grounding, IEEE Standard 80-
interconnected. 2000, May 2000.
[3] A. Esmaeilian, A.A. Shayegani Akmal, “ Wind Farm Grounding
TABLE VI. RESULTS FOR TWO TURBINES Systems Design Regarding the Maximum Permissible Touch & Step
Voltage “, IEEE Environment and Electrical Engineering (EEEIC), 2012
Peak RMS 11th International Conference on, pp. 74-79, May 2012;
Case ti IB1,4*ti Safety
current current
[4] H. Kazemi Karegar,”New Wind Turbine Grounding System to Reduce
❶ 0.25/100 µs 145.9 59.5 4.5E-6 1.37E-3 OUI Step & Touch Voltage”, Power and Energy (PECon), 2010 IEEE
❶ 10/350 µs 67.2 27.4 1.30E-6 1.34E-4 OUI International Conference on, pp. 168-171, December 2010;
[5] Lorentzou, M.I.; Hatziargyriou, N.D.; Papadias, B.C., “Analysis of wind
According to Figure 10. Figure 13. one can make the turbines grounding system ,” Electrotechnical Conference, 2000.
following comments: MELECON 2000. 10th Mediterranean;
[6] Hatziargyrious. N, Lorentzou. M, Cotton. I, Jenkins. N: “Windfarm
 interconnections between wind turbines generally Earthing”, The Institution of Electrical Engineers, IEE, Savoy Place,
reduces current peak values and energy passing London, 1997
through the body, [7] Hatziargyrious. N, Lorentzou. M, Cotton. I, Jenkins. N: “Transferred
Overvoltages by Windfarm Grounding Systems” Proceedings of ICHQP
 interconnections do not provide the same effect conference, 14-16 October 1998, Athens, Greece;
according to the lightning current waveform, faster [8] A.Elmghairbi, A. Haddad and H. Griffiths, “Potential rise and safety
rise time waveform are less sensitive to voltages of wind turbine earthing systems under transient conditions,” in
interconnections, this is due to the grounding Electricity Distribution – Part 1, 2009. CIRED 2009. 20th International
system reactance, Conference and Exhibition on, 8-11 June 2009.
[9] CDEGS and MultiFields Software Package, Safe Engineering Services
 safety criteria can be satisfied independently to the & technologies ltd., Montreal, Quebec, Canada, 2009.
wind turbine grounding resistance but depends on [10] IEC 62305-3, Protection against lightning – Part 3: Physical damage to
the resistivity and thickness of the surface layer structures and life hazard, Ed. 2.0, 2010
installed. [11] IEC 60479-2, Effects of current on human beings and livestock – Part 2:
special aspects (2007).
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY 1

Electromagnetic Shielding Analysis of Spherical

Polyhedral Structures Generated by Conducting
Wires and Metallic Surfaces
Ali Aghabarati, Rouzbeh Moini, Senior Member, IEEE, Simon Fortin, Member, IEEE,
and Farid P. Dawalibi, Senior Member, IEEE

Abstract—The shielding performance of novel spherical polyhe- specific frequencies has stimulated many research activities [4]–
dral structures illuminated by an electromagnetic plane wave is in- [7]. These have led to the development of novel frequency se-
vestigated. The spherical shield screens consist of a juxtaposition of
lective structures that allow the use of protected facilities at
wire conductors and planar metallic elements distributed accord-
ing to the regular pattern of a spherical polyhedron. The considered specific frequencies or control the emission of electronic sys-
3-D screens represent four groups of structures with fundamental tems [4]–[6]. The shielding characteristics of enclosures made
elements of design for attaining frequency selectivity. It is shown from wire grid screens have been investigated and research re-
that at low frequency, the wire frame screen provides high levels sults concerning the protection inside grid-like shields against
of shielding effectiveness, while its counterpart of metallic surfaces
electromagnetic fields due to lightning can be found in [4]. With
has a similar behavior for higher frequency bands. Moreover, the
shielding characteristics of a combined wire-surface and wire-loop the increasing demand for the design of frequency selective sur-
spherical surface are reported and compared with the cases of a faces (FSS), the electromagnetic behavior of many planar shield
single screen of wire conductors or metallic surfaces. The reported structures have been extensively studied during the past years
numerical results illustrate the advantage of using each configura- [7], and the frequency response of planar FSS structures with
tion. Detailed physical interpretations of the obtained numerical
resonant unit cell is well understood [5].
results are given based on the operation mechanism of spherical
shield structures according to structural-based and pattern-based For traditional geometries considered as infinite 2-D periodic
resonances. structures, the numerical analysis can be reduced to construct-
ing an integral equation only for the unit cell of the geometry.
Index Terms—Electromagnetic interference, polyhedral struc-
tures, shielding effectiveness (SE), spherical shield screen. Next, by applying Floquet’s theorem to the formulation of the
scattering problem for the whole structure, the continuous con-
volution integral can be converted into an infinite series with
I. INTRODUCTION each term being the product of the induced surface current and
ITH the increase in radio frequency communications the Green’s function in the spectral domain [7]. Using this tech-
W and of the related hazards due to intentional and un-
intentional interfering sources, electromagnetic shielding has
nique, the shielding characteristics of various planar periodic
structure designs based on wire-type elements or patch-type el-
become a fundamental concern for problems related to electro- ements have been investigated to enhance the fabrication of 2-D
magnetic compatibility (EMC) [1], [2]. Sensitive devices are shield structures [5]–[7]. However, little research has been made
usually placed inside a metallic box or container to avoid their on characterizing the shielding properties of screens of spherical
malfunction when exposed to undesirable electromagnetic in- shape, with spherical periodicity.
terference [1]–[4]. Ideally, the enclosure should be conformal Spherical periodic structures are attractive for designing non-
to the geometry of the protected device while providing a level planar electromagnetic devices used as a conformal shield en-
of electromagnetic shielding acceptable to the susceptibility of closure. Particularly, curved screens can be useful for airborne
the system. equipment, such as aircrafts and satellites, as a cover for hid-
The idea of designing enclosures that can selectively con- ing communication facilities or enhancing the level of sig-
trol the electromagnetic wave shielding effectiveness (SE) at nals. They can provide both mechanical and electromagnetic
protection without disturbing the aerodynamic properties of a
Manuscript received December 9, 2016; revised January 26, 2017; accepted
protected system. Spherical FSS can also be used as multifre-
February 6, 2017. This work was supported by Safe Engineering Services & quency curved reflectors that can provide reflection and trans-
Technologies Ltd. mission in different bands of frequencies [8]–[11]. The func-
The authors are with Safe Engineering Services & Technologies
Ltd., Laval, QC H7L6E8, Canada (e-mail: [email protected];
tionality of shield screens with conventional planar design is
[email protected]; [email protected]; farid.dawalibi@ affected by the direction and polarization of incident waves.
sestech.com). However, the design of spherical structures with less sensitiv-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
ity to the characteristics of incident waves can overcome such
Digital Object Identifier 10.1109/TEMC.2017.2670499 limitations. The goal of this paper is to present a new type of

0018-9375 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

spherical structure with unique SE characteristics. The presented

structures can be designed for the protection of transceivers
working in a selected band of frequencies in both emission and
reception mode. Depending on the size of the spherical struc-
tures and on the frequency of operation, the studied spherical
FSS structures can also be implemented in building domes or
roofs as components designed to control the propagation of sig-
nals into and out of the buildings with low emissivity.
Accordingly, this paper presents an electromagnetic analysis
of a family of screens, which are composed of conducting wires
and metallic surfaces distributed on a spherical surface. The
spherical FSS are constructed following the regular pattern of a
spherical polyhedron. Polyhedral structures have recently been
investigated for their scattering cross section and for the design
of enhanced porous spherical cavities [11], [12]. Their unique
geometrical features can provide electromagnetic devices that
attain multiband frequency selectivity independent of polariza-
tion or angle of incidence due to their spherical periodicity. It is
shown here that the studied polyhedral screens produce filtering
characteristics that vary periodically as a function of frequency.
These unique features introduce filtering screens with a fre-
quency response, which is nearly independent of the direction
of the incident wave [1]. Such characteristics can be important Fig. 1. Illustration of the spherical polyhedral shield screens. (a) Wire screen,
W240. (b) Surface screen, S240. (c) Combined wire-surface screen, WS240. (d)
for communication links, where a trace of the incident wave is Complementary wire-loop screen, WL240.
present in all directions due to multipath propagation.
An efficient full wave electromagnetic analysis based on the
method of moments (MoM) is used to investigate the shielding wire-surface screen, and a structure obtained by wire loops,
characteristics of the 3-D spherical screens. This methodology which are denoted by W240, S240, WS240, and WL240, re-
allows us to easily investigate the effect of different geometrical spectively. It should be noted that the underlying polyhedron,
parameters involved in the design of spherical shield enclosures. which belongs to a family of highly symmetrical polyhedrons,
The results presented in this paper can provide guidance to EMC is composed of 110 hexagons and 12 pentagons that tile the
engineers in the design of multiband operating spherical shields. spherical surface. Details on generating polyhedral spheroids of
The organization of the paper is as follows. In Section II, higher indexes (i.e., polyhedrons with 540 vertices, 960 vertices
the spherical polyhedral structures considered in this paper are or more) can be found in [11], where by increasing the index
described. Section III is dedicated to the computational method- only the number of hexagons increases. Consequently, smoother
ology used for the efficient analysis of the shield structures and structures having higher spherical symmetry and lower ratio of
to the definition of measures for evaluating the shielding effec- the number of pentagons to the number of hexagons can be
tiveness. Numerical results are described in Section IV, where achieved, resulting in improved functionality of the FSS under
the shielding properties of these enclosures are presented. The plane wave excitation.
different types of spherical shields are compared with respect The W240 wire frame shield enclosure, illustrated in Fig. 1(a),
to their shielding behaviour, and the advantages and disadvan- represents the (imperfect) sphere made from 360 conducting
tages of each are discussed. Finally, in Section V, conclusions strips or wires that form the edges of the spherical polyhedron.
are drawn that can provide insight into the operation mechanism The construction shows that at each vertex three conducting
of the structures with respect to their frequency response. wires come together. From the inherent spherical symmetry
of the structure, it can be concluded that its electromagnetic
characteristics under a plane wave illumination do not depend
II. SPHERICAL POLYHEDRAL STRUCTURES significantly on the orientation of the structure. On the other
The new shielding enclosures presented in this study are con- hand, the radius of the wire conductors can be changed to control
structed based on spherical polyhedral geometries. As illustrated the size of the openings over the surface of the structure. The
in Fig. 1, four different types of shield enclosures are obtained effects of this parameter on SE characteristics are examined in
based on approximations to a smooth polyhedral spheroid [11], this study.
which has 240 vertices located at a fixed distance from the cen- The S240 structure depicted in Fig. 1(b) is a collection of 132
ter of the spherical surface. The geometries considered here disjointed perfectly electric conductive (PEC) screens that have
represent a complete set of structures that can be obtained by an area nearly equal to the area of the pentagons and hexagons
periodic repetition of several wire-type elements or patch-type of the underlying polyhedron. The surface shield enclosure is
elements over a spherical base. As indicated in Fig. 1(a)–(d), in fact a complementary spherical screen to the W240 wire
these structures are a wire screen, a surface screen, a combined grid screen, obtained by exchanging the conductive parts of
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

AGHABARATI et al.: ELECTROMAGNETIC SHIELDING ANALYSIS OF SPHERICAL POLYHEDRAL STRUCTURES 3

S240 with nonconductive parts over its surface. For this type medium over a wide band of frequencies. Below is a description
of shield, the area of the individual planar surfaces acts as a of the computational modeling theory for the studied networks
controlling parameter for the size of the existing gaps between under a plane-wave excitation.
the disjointed surfaces. This parameter allows for a transition The boundary conditions of Maxwell’s equations dictate that
between a structure with significant gaps and a completely solid the tangential component of the total electric field should van-
polyhedral surface. It is shown in Section IV that this parameter ish on the surface of wire conductors and surface elements as
has important effects on the SE performance. follows:
Fig. 1(c) depicts the combined wire and surface spherical 
(LJ w + LJ s )|tan = − Einc tan (1)
shield screen, called WS240, which contains the components of
both the wire frame and surface screen structures. It should be where J w and J s represent the total electric current densi-
noted that the group of conducting wires are disjointed from the ties for the induced unknown line and surface currents on the
metallic surfaces. The size of the gap between the components wire conductors and metallic surfaces, respectively. The incident
in the new design can be parameterized with both the wire radius electromagnetic field is denoted by Einc . The operator L returns
and the area of the surfaces. The numerical results for the SE the electric field scattered by any kind of current densities and
behavior of WS240 illustrate that careful combinations of the is defined as:
individual wire and surface components can bring advantageous   
1
shielding characteristics. LJ = jkη GJ + 2 ∇G∇ · J dr

(2)
Ω k
Finally, Fig. 1(d) shows the WL240 wire loop screen obtained
by exchanging the conductive parts of WS240 with nonconduc- where k and η denote the free-space wave number and
tive parts over its surface. The WL240 screen can, thus, be con- impedance, and G represent the free-space Green’s functions
sidered as a complementary structure to the WS240 screen, with [15]. In this equation, Ω stands for the problem’s domain which
SE characteristics closely related to those of its complement. For consists of the surface of wire conductors or planar surfaces, in
this conductive wire FSS, the perimeter of the hexagonal and general.
pentagonal loops and the radius of the wires can change the The numerical approach that has been developed here for
frequency response of the structure. determining the unknown currents is based on a discretization
Each spherical screen from the four major groups generates of surface wire conductors and planar surfaces using tubular
a scattered wave with a specific frequency response when il- segments and quadrilateral surface patches, respectively. The
luminated by electromagnetic plane waves. It should be noted current densities on the surface of wire and patch segments are
that an efficient computational approach is needed in order to expanded in terms of their corresponding linear (first order)
to analyze electromagnetic interactions and characterize the ef- basis functions jw and js :
fects of the thickness or surface area of the conductive parts on N w
N s
 
the shielding behavior of the presented FSS structures. The next J = Inw jw + Ins jsn (3)
n
section describes the approach used in this paper. n =1 n =1

where N w and N s stand for the total number of degree of

III. COMPUTATIONAL MODELING THEORY freedom (DOF) introduced over wires and planar surfaces, re-
For the analysis of the proposed structures, the usual assump- spectively. We note that in our analysis, the PEC geometries
tion of a doubly periodic and infinite FSS used for analyzing of W240 and WL240 structure require only the first term of
planar geometries is invalid and conventional methods based on (3), while S240 requires only the second term (i.e., surface cur-
Floquet’s theorem are inadequate. On the other hand, for calcu- rents). WS240 is the only structure in which the induced current
lating near field quantities (especially at the center of screens), includes both sets of currents.
the effects of strong mutual coupling between induced current After substituting (3) into (1), we use appropriate line test
over the entire structures have to be taken into account accu- functions in order to transform the boundary value problem
rately using a full wave methodology [11], [13]. This allows into a matrix equation AI = V of size N = N w + N s . The
accurate evaluation of field quantities close to the screens. structure of the resulting MoM matrix equation has four blocks,
The electromagnetic fields generated by an external source as follows:
    
induce currents on the conducting wires and metallic surfaces Aww Aws Iw Vw
of the structures. An efficient and cost-effective methodology = (4)
Asw Ass Is Vs
should be able to compute the variation of spatial 1-D linear
currents (induced on conducting wires) and 2-D surface cur- where the unknown coefficients of the wire and surface currents
rents (induced on metallic surfaces). Thus, all numerical com- are denoted by I w and I s , respectively. The excitation vectors
putations have been performed using the commercial CDEGS resulting from the line integral over the surface and wire re-
software package. The high-frequency module of this package gions are shown as V w and V s . The elements of each block
is based on a numerical solution of the electric-field integral are computed by taking into account the interactions between
equations using MoMs that provides a combined wire and sur- the test functions and the expansion terms [1], [15]. Once the
face modeling of scatterers [14]. It is suitable for the electro- induced currents are obtained, the electric and magnetic fields
magnetic scattering analysis of complex objects in a stratified are computed for the evaluation of various measures that can
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

quantify the overall efficiency of studied structure due to differ-

ent kinds of shielding mechanisms. The electric, magnetic, and
electromagnetic SEs at any interior point r are defined based on
the amplitude of the electric and magnetic fields in the shielded
and unshielded scenarios and are given as follows [3]:

|Eunshielded
(r ) |2
SE ( r ) (Electric) = 10 log dB (5)
|Eshielded
(r ) |2

|Hunshielded
(r ) |2
SE ( r ) (Magnetic) = 10log dB (6)
|Hshielded
(r ) |2

SE(r ) (Electromagnetic)
⎛ ⎞ Fig. 2. Electric SE, magnetic SE, and electromagnetic SE of W240 as a
function of frequency, exhibiting a behavior similar to that of a low-frequency
⎜ 2 ⎟ shield/resonator.
= 10log ⎜
⎝ |E s (hri e) l d e d |2 |H s( rh )i e l d e d |2
⎟ dB.
⎠ (7)
|E u( rn)s h i e l d e d |2
+ |H u( rn)s h i e l d e d |2
At each frequency, the shielding performance of the screen is
We note that while (5) and (6) are useful measures for evalu- evaluated at the center of the structure, using (5)–(7).
ating shield induced reduction of electric and magnetic fields at Fig. 2 shows the computed electric, magnetic, and electro-
low frequency, the electromagnetic SE is an overall measure for magnetic SEs as a function of frequency when the radius of the
evaluating the attenuation of electromagnetic power delivered screen’s wire conductors is 1 cm. As seen in this figure, the
to an infinitesimal load at point r. Particularly, the electromag- electric and magnetic SEs exhibit different behaviors in the fre-
netic SE is useful at high frequencies where the dimension of the quency band of the analysis. While both shielding performances
shield enclosure is larger than the wavelength, and the spatial generally degrade as the frequency increases, the electric and
distribution of SE is mainly related to the modal behavior of the magnetic SEs exhibit sharp peaks and drops at some specific
enclosure [3]. frequencies. These minima and maxima are anticorrelated. The
magnetic SE is maximal at frequencies where the electric SE is
IV. NUMERICAL RESULTS minimal; conversely, the magnetic SE is minimal at frequencies
where the electric SE is maximal. On the other hand, the elec-
In this section, the electromagnetic SE of the introduced
tromagnetic SE is between the electric and magnetic SEs, and
spherical structures and their variation with respect to frequency
exhibit drops at every minimum of both. Indeed, at some fre-
and geometrical parameters are investigated. In every case, the
quencies the incident power penetrates into the structure from
observation point r is chosen to be the center of the spherical
the openings of the wire screen and the input power excites
structure. First, the SE properties of the wire screen structure
specific TE or TM resonant modes of the equivalent spherical
are illustrated in a selected frequency band and results for dif-
cavity. Close to these resonant frequencies, the amplitude of
ferent conductor radius sizes are presented. Next, the shielding
either the electric or magnetic field increases in the interior re-
characteristics of spherical surface enclosures with various sur-
gion, resulting in a significant degradation of the corresponding
face areas are investigated and the observed frequency selective
SE, and to negative values for the electromagnetic SE. In other
features are discussed in detail. The SE results for the surface
words, negative values of the electromagnetic SE indicate en-
screen are compared with those for the wire screen shield en-
hanced localization of either the electric or the magnetic field
closure. Finally, the combined wire-surface and the wire-loop
and resonating behavior for the structure.
spherical structures are simulated and their SE characteristics
The first minimum of the electric SE occurs at 11.25 MHz,
are presented.
followed by a maximum at 19.25 MHz. The magnetic SE ex-
hibits the opposite trend. The ratios of wavelength to radius
A. Wire Spherical Shield Screen
of the porous structure at those frequencies are about 2.70 and
A W240 wire grid polyhedron [see Fig. 1(a)], with a radius 1.56. It should be noted that for a spherical cavity with the same
of 10 m, is placed at the center of the coordinate system in dimensions, the lowest resonances occur at the frequency of
free space. The length of each wire conductor is about 2 m. An 13.10 MHz (TM101 mode) and 21.45 MHz (TE101 mode), re-
incident plane wave propagating along the z-axis illuminates spectively [16]. Interestingly, the ratios of wavelength to radius
the structure. In the following, the analysis is restricted a po- for the spherical cavity at these frequencies are 2.32 and 1.40,
larization angle of zero, since the spherical symmetry of the which is close to the corresponding values for its wire enclosure
studied screens makes them nearly insensitive to the orientation counterpart.
or polarization of the applied excitation. The frequency of the In addition to the alternating behavior observed in the fre-
incident wave varies from 1 KHz to 80 MHz in 325 frequency quency response of W240, the electric and magnetic SE vanish
steps. At the highest frequency of the analysis band, the radius simultaneously when the frequency increases. This shows that
of the spherical screen is about 2.667 times the wavelength. the total electromagnetic field into the structure is equal to the
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

AGHABARATI et al.: ELECTROMAGNETIC SHIELDING ANALYSIS OF SPHERICAL POLYHEDRAL STRUCTURES 5

Fig. 3. Variation of the electric field (V/m) in the vicinity of the W240 wire
screen with conductor radius of 1 cm illuminated by a unit amplitude plane
wave. (a) 5 MHz. (b) 11 MHz.

Fig. 5. Electric SE, magnetic SE, and electromagnetic SE of the S240 surface
screen as a function of frequency with surface area of 80%, exhibiting a behavior
similar to that of a high-frequency shield/resonator.

The variation of the magnitude of the electric field in the vicin-

ity of the structure over an observation plane passing through
the center of the structure and that contains the wave propa-
gation vector is illustrated in Fig. 3 at two frequencies in the
lower band: 5 and 11 MHz. It can be seen that the weakest
electric field is achieved inside the screen at 5 MHz compared
to the exterior. By contrast, at 11 MHz the field is strong at the
center and the largest electric field is found inside the spherical
wire grid screen. The variation in the amplitude of the electric
field coupled into the screen is in agreement with maximum
and minimum values of the electric SE computed at the center
of polyhedral screen W240. Fig. 3 also demonstrates the ori-
entation of the structures under plane wave illumination, where
pentagons are located on the two sides while looking from the
direction of the propagation vector.
The effect of the radius of the wire on the SE is examined
next. The simulations were performed for three different wire
radii. Fig. 4(a) and (b) present the variation of the electric, mag-
netic, and electromagnetic SE versus frequency as the radius
increases from 1 to 5 cm. It can be observed that the ampli-
tude of the SE peaks and nulls grows as the size of apertures
decreases by using thicker conductors. Consequently, struc-
tures with improved shielding and resonating performance can
be achieved for operation at lower frequencies. Application of
the strongest resonating mode of the wire grid spherical structure
has been reported in [12] for plasma cloud generation, where a
highly optimized wire-grid FSS with very large Q factor is used
to build up the electric field amplitude at localized regions to
values above the breakdown threshold of a neutral gas.

Fig. 4. Variations of electric SE, magnetic SE, and electromagnetic SE for

W240 wire screen as a function of frequency for different conductor radii. B. Surface Spherical Shield Screen
(a) Electric SE. (b) Magnetic SE. (c) Electromagnetic SE.
The spherical structure used for the analysis of this section is
shown in Fig. 1(b). The center of a S240 spherical surface with
incident field at higher frequencies. In fact, the array of con- a radius of 10 m is again located at the origin of the coordinate
nected wires, acts as a low-pass filter due to the inductive be- system and the structure is illuminated using the plane wave
havior of wire meshes. The behavior of the electric and magnetic described in Section IV-A. Electromagnetic fields at the center
SEs for the entire frequency range implies that the structure can of the structure are obtained numerically using the approach
perform well as a low frequency shield or low-frequency res- described in Section III, when the formulation contains only the
onator. surface unknowns.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

Fig. 6. Variation of the electric field (V/m) in the vicinity of the S240 surface
screen with surface area of 80% illuminated by a unit amplitude plane wave. (a)
70 MHz. (b) 77 MHz.
Fig. 8. Electric SE, magnetic SE, and electromagnetic SE of the WS240
surface screen as a function of frequency with conductor radius of 1 cm and
surface size of 80%, exhibiting a behavior similar to that of a low- and high-
frequency shield/resonator.

variations observed in the SE curves (resulting from the reso-

nance of the whole sphere), the increasing trend in the SE values
demonstrates another spectral selectivity at higher frequencies
as a result of the pattern of elements on the surface of screen. On
the other hand, both the electric and magnetic fields can pen-
etrate through the structure at several resonance frequencies,
which results in maximum degradation of their SEs. At each
resonating frequencies, the electromagnetic SE also exhibits a
drop and its negative value is at most 3 dB larger than the lower
of the electric and magnetic SEs.
It can be seen that the first maximum of the electric SE curve
now occurs at 10.25 MHz, while its first minimum takes place at
17.75 MHz. The magnetic SE exhibits the opposite trend. The
ratio of wavelength to radius of the structure at those frequencies
is about 2.92 and 1.69. Furthermore, it is now observed that
all SE values vanish simultaneously at low frequency. Indeed,
the high pass characteristics of S240 are expected due to the
capacitive coupling of the unconnected plates. From the results,
it can be concluded that this shield configuration is transparent
to low-frequency waves.
Fig. 6 shows how the surface screen interacts with applied
high-frequency waves at 70 and 77 MHz. The large variations
of the electric field inside the metallic shell demonstrate that
resonance effects are strong and the strength of the instanta-
neous electric field can vary abruptly in the high-frequency
band within the shielded spherical domain. This is an indication
for the dependence of SE at high frequencies on the location
of the observation point [3]. At 77 MHz, the magnitude of the
electric field at the center of the screen is larger than the field
in the unscreened scenario, while at 70 MHz the presence of
Fig. 7. Variation of electric SE, magnetic SE, and electromagnetic SE for
the polyhedral surface structure greatly reduces the level of the
S240 surface screen as a function of frequency for different surface areas. penetrated wave at the center point.
(a) Electric SE. (b) Magnetic SE. (c) Electromagnetic SE. We now consider decreasing the area of the pentagons and
hexagons over the S240 shield structure, where we can char-
Fig. 5 depicts the variation of the SEs versus frequency for acterize the shield performance with respect to the effective
a S240 structure in which the metallic plates cover 80% of surface area and the blockage that the metallic pieces provide.
the area of each facet. The computation is performed again for In Fig. 7(a)–(c), comparisons are given for the electric, mag-
frequencies varying from 1 KHz to 80 MHz. The SE variations netic, and electromagnetic SEs of the structures with effective
exhibit again multiple attenuations and amplifications. But con- surface areas varying from 80% to 50%. A remarkable decrease
trary to the wires structure, amplitude of the sharp peaks and is found in the level of protection or amplification as the area of
troughs is increasing with frequency. In addition to the periodic the component surfaces are decreased, particularly in the higher
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

AGHABARATI et al.: ELECTROMAGNETIC SHIELDING ANALYSIS OF SPHERICAL POLYHEDRAL STRUCTURES 7

Fig. 9. Variation of the electric field (V/m) in the vicinity of the WS240 wire-surface screen with conductor radius of 1 cm and surface size of 80% illuminated
by a unit amplitude plane wave. (a) 5 MHz. (b) 33 MHz. (c) 70 MHz.

advantageous properties of both wire and surface components

can be superimposed in the new design. However, there is a
frequency band (15–45 MHz) for which the pattern of elements
over the combined screen is producing a different behavior, i.e.,
reduced shield or resonating performance. Outside of this fre-
quency band, the structure can, thus, be considered as a shield
or resonating device.
Fig. 9 illustrates the distribution of the electric field through
the combined screen for three selected frequencies: 5, 33, and
70 MHz. It should be noted that SE exhibits a peak at 5 MHz
for W240, while for S240 this occurs at 70 MHz. Fig. 9(a)
demonstrates the blockage provided by WS240 at 5 MHz. As
observed, in the low-frequency band the distribution of the field
Fig. 10. Electric SE, magnetic SE, and electromagnetic SE of the WL240 is very similar to Fig. 3 for W240. Interestingly in Fig. 9(b),
surface screen as a function of frequency with conductor radius of 1 cm and
loop circumference match to 80% surface area, exhibiting a behavior similar to the structure is almost transparent at 33 MHz, which is the
that of a middle frequency shield/resonator. frequency in the middle of the analysis band. On the other hand,
at the higher frequencies (70 MHz), the computed electric field
at the center of the structure reaches minimum values due to
the presence of the surface screen, as shown in Fig. 9(c). The
frequency band. This confirms that the surface screen is more
field distribution at this frequency is in close agreement with the
transparent to high-frequency waves for a design with reduced
results obtained for S240 [see Fig. 6].
surface area. As illustrated in Fig. 7(c), by an appropriate design
S240 can act well either as a high-frequency shield or a resonator
at a number of frequencies located in the high-frequency band.
D. Wire-Loop Spherical Shield Screen
Moreover, the shield and resonance frequencies corresponding
to peaks and drops tend to increase slightly as shown in the In this last section, the SE characteristics of a WL240 structure
figures. are presented. This structure contains wire loop elements placed
periodically over the spherical surface as shown in Fig 1(d).
The frequency selective properties of the wire loop screen are
C. Wire-Surface Spherical Shield Screen
compared with those of the previous structures, in particular
The combined wire and surface structure used for the analysis with WS240 which is its complementary design.
in this section is shown in Fig. 1(c). The computed SE results Fig. 10 presents the electric, magnetic, and electromagnetic
are compared with those obtained in Sections IV-A and IV-B. It SEs of the structure over the frequency band of operation. For the
is shown that the shielding performance of the combined screen chosen WL240, the wire radius is 1 cm and the wire loops repre-
at low and high frequencies is characterized by the attributes of sent the perimeter of metallic surfaces considered in Section IV
the wire screen and surface screen, respectively. In Fig. 8, the C (i.e., 80%. surface area). The length of each wire conductor
SE curves of the WS240 are shown, when the wire radius is in WL240 is about 1.67 m. Simulation results show clearly dif-
1 cm and the surface area is 80%. It is interesting to note that ferences between the transmission and reflection modes of the
there is a frequency (15 MHz) below which the performance wire loop FSS. Toward the middle of studied frequency band
of the combined screen is mostly determined by its wire screen (15–45 MHz), the interaction between the structure and the inci-
counterpart. It can also be seen that at sufficiently high frequen- dent plane wave leads to either substantial scattered electromag-
cies (more than 45 MHz), the SE curves of the WS240 structure netic fields (high positive values of SEs), or strong impedance
tend to be same as those of the surface screen. Consequently, the match and coupling of the external wave to the screen (high
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

Fig. 11. Variation of the electric field (V/m) in the vicinity of the WL240 wire-loop screen with conductor radius of 1 cm and loop circumference corresponding
to the surface size of 80% illuminated by a unit amplitude plane wave. (a) 11 MHz. (b) 33 MHz. (c) 77 MHz.

TABLE I pentagon and hexagon surfaces is subdivided into 15 and 18

COMPUTATIONAL STATISTICS FOR THE MOM ANALYSIS OF SPHERICAL FSS
GEOMETRIES AT 40 MHZ
patches, respectively. As observed, modest computational effort
(memory and CPU time) is needed in each simulation by the
combined wire and surface modeling of the scatterers.
FSS # Wire DOF # Surface Memory CPU Time
Geometry (N w ) DOF (N s ) (MB) (mm:ss)

W240 1320 – 365 (02:41)

S240 – 5040 1658 (25:52)
V. CONCLUSION
WS240 1320 5040 3251 (54:13) The SE of 3-D polyhedral spherical screens constituted by
WL240 2880 – 831 (10:57)
wire conductors and metallic surfaces under a plane wave illumi-
nation has been investigated through a full-wave combined wire
and surface formulation in conjunction with the spatial domain
negative values of SEs), while weak values of the SEs at lower MoM. The interactions of electromagnetic waves with four dif-
and higher extremities of the studied frequency band indicate an ferent spherically periodic structures (W240, S240, WS240, and
essentially transparent behavior at these frequencies. It should WL240) have been studied in detail and the obtained wideband
be noted that the circumference of each hexagonal loops is about SE results at their center have been compared. It has been shown
10 m, corresponding to a resonance frequency of 30 MHz. In that, for the spherical wire screen under plane-wave excitation,
contrast to the WS240 structure, significant scattering of the significantly larger values of electric and magnetic field atten-
incident wave by the WL240 FSS happens at frequencies close uation or amplification can be observed at lower frequencies
to the middle of the analysis band due to these resonances. In- and the level of disturbance in the field decreases by increasing
deed, the pattern-based resonances happen approximately for the frequency. In contrast, the surface screen is transparent to
frequencies at which the loop elements on the structure resonate low-frequency excitations and can provide electromagnetic field
(30 MHz). On the other hand, the SE characteristics at very high protection and intensification only at selected higher frequen-
and low frequencies tend to be the same as those of the wire cies. The SE performance of the composite WS240 structure is
screen and surface screen, respectively. In fact, each wire loop characterized as combination of the wire and surface screens,
shows an inductive behavior, while the capacitive coupling be- as it attains strong interaction with the plane wave simultane-
tween unconnected loops added to the inductive feature causes ously at both low and high frequencies. The wire loop structure
the screen to have a frequency band selectivity behavior. These was found to transmit both low- and high-frequency content of
indicate that aside from structural-based alternating variations, the wave, but highly reflect or focus the energy in middle band
the WL240 can perform as a middle band shield or resonator frequencies.
for the attenuation or amplification of power delivered to a load The influence of various geometrical parameters (the radius
at the center of screen. of the wire conductors and area of metallic surfaces) on the
For a better understanding of the operation mechanism of shielding performance has been assessed and physical and nu-
WL240, a set of simulations were also performed at 11, 33, and merical interpretations of the behavior of the studied finite-size
77 MHz. Fig. 11 plots the corresponding electric field varia- FSS structures were given. The presented simulation results
tion. As can be observed by comparing the level of transmitted provides insight into the operation mechanism of 3-D spherical
field at the center of screen in Fig. 11(b), the amplitude of the FSS under plane wave illumination based on existing effects of
electric field is decreased significantly at 33 MHz, which is the structural-based resonances and pattern-based resonances. This
opposite behavior when compared to the results for WS240 [see can lead to further improvements in the creation of more com-
Fig. 9(b)]. plex spherical shield screens with desired frequency responses.
Table I compares computational statistics for the simulation Frequency selective structures with superimposed spectral re-
of the studied spherical FSS geometries using the presented sponse can be obtained by centering carefully chosen spherical
full-wave MOM at one of the frequencies (40 MHz), in which FSS with different radii and patterns of elements. Future work
every wire conductor is subdivided into four segments, and each will study the effects of dielectric substrates on the shielding
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

AGHABARATI et al.: ELECTROMAGNETIC SHIELDING ANALYSIS OF SPHERICAL POLYHEDRAL STRUCTURES 9

characteristics of 3-D polyhedral screens and examine spherical Rouzbeh Moini (M’93–SM’05) received the B.S.,
FSS formed by arranging conductive elements over spherically M.S., and Ph.D. degrees in electronics from the Uni-
versity of Limoges, Limoges, France, in 1984, 1985,
layered dielectric shells in order to investigate the effects on the and 1988, respectively.
frequency response. In 1988, he joined Amirkabir University of Tech-
nology, Tehran, Iran, where he was a Professor in
the Department of Electrical Engineering. He was
REFERENCES also a Visiting Professor with the University of
Florida in 1995–1996, and with the University of
[1] A. Aghabarati, R. Moini, S. Ladan, S. Fortin, and F. P. Dawalibi, “Electro-
magnetic shielding properties of spherical polyhedral structures generated Toronto, Toronto, ON, Canada, in 2010–2011. Since
2012, he has been with Safe Engineering Services
by conducting wires and metallic surfaces,” in Proc. IEEE Int. Symp. Elec-
& Technologies Ltd., Laval, QC, Canada, as a Senior Research Manager. His
tromagn. Compat., 2016, pp. 277–282.
research interests include numerical methods in electromagnetics, electromag-
[2] S. Celozzi, R. Araneo, and G. Lovat, Electromagnetic Shielding. Hoboken,
NJ, USA: Wiley, 2008. netic compatibility, and lightning electromagnetics. He is the author or coauthor
of one book, one book chapter, and more than 250 scientific papers published
[3] L. Klinkenbusch, “On the shielding effectiveness of enclosures,” IEEE
in reviewed journals and presented at international conferences.
Trans. Electromagn. Compat, vol. 47, no. 3, pp. 589–601, Aug. 2005.
[4] T. Maksimowicz and K. Aniserowicz, “Investigation of models of grid-
like shields subjected to lightning electromagnetic field: Experiments in
the frequency domain,” IEEE Trans. Electromagn. Compat., vol. 54, no. 4,
pp. 826–836, Aug. 2012.
[5] B. A. Munk, Frequency-Selective Surfaces: Theory and Design.
New York, NY, USA: Wiley, 2000, pp. 63–90.
[6] Y. E. Erdemli, K. Sertel, R. Gilbert, D. E. Wright, and J. L. Volakis,
Simon Fortin (M’06) received the B.Sc. degree in
“Frequency-selective surfaces to enhance performance of broad-band re-
configurable arrays,” IEEE Trans. Antennas Propag., vol. 50, no. 12, physics from Laval University, Quebec, QC, Canada,
in 1985 and the Ph.D. degree in physics from The Uni-
pp. 1716–1724, Dec. 2002.
versity of British Columbia, Vancouver, BC, Canada,
[7] R. Mittra, C. H. Chan, and T. Cwik, “Techniques for analyzing frequency
in 1991.
selective surfaces—a review,” Proc. IEEE, vol. 76, no. 12, pp. 1593–1615,
Dec. 1988. His area of specialization has been in the the-
oretical aspects of high-energy particle physics. In
[8] S. Skokic and Z. Sipus, “Moment method analysis of a spherical FSS in
1992–1993, he was a Research Assistant in the De-
free space,” in Proc. IEEE 18th Int. Conf. Appl. Electromagnetics Com-
partment of Nuclear Physics, University of Montreal,
mun., Oct. 2005, pp. 1–4.
[9] C. Yu and C. C. Lu, “Analysis of finite and curved frequency-selective Montreal, QC, again specializing in particle physics.
Since 1994, he has been with Safe Engineering Ser-
surfaces using the hybrid volume-surface integral equation approach,”
vices & Technologies Ltd., Laval, QC, which he joined as a Research Scientist.
Microw. Opt. Technol. Lett., vol. 45, no. 2, pp. 107–112, 2005.
[10] B. Philips, E. A. Parker, and R. J. Langley, “Influence of a curved FSS He has authored or coauthored about 40 papers on lightning, grounding, and
electromagnetic fields compatibility-related problems. His research interests
on the radiation patterns of an enclosed source,” in Proc. 9th Int. Conf.
include the computation of electromagnetic fields at various frequencies and
Antennas Propag., 1995, pp. 524–527.
transient phenomena.
[11] P. Bernhardt, “Radar backscatter from conducting polyhedral spheres,”
IEEE Antennas Propag. Mag., vol. 52, no. 5, pp. 52–70, Oct. 2010.
[12] P. A. Bernhardt et al., “Visible plasma clouds with an externally excited
spherical porous cavity resonator,” IEEE Trans. Plasma Sci., vol. 43, no
6, pp. 1911–1918, Jun. 2015.
[13] A. Aghabarati and J. P. Webb, “Multilevel methods for p-adaptive finite
element analysis of electromagnetic scattering,” IEEE Trans. Antennas
Propag., vol. 61, no. 11, pp. 5597–5606, Nov. 2013.
[14] “CDEGS 15.1.4080,” Safe Engineering Services & Technologies Farid P. Dawalibi (M’72–SM’82) was born in 1947.
Ltd., Montreal, QC, Canada, 2015 [Online]. Available: http:// He received the B.Eng. degree from St. Joseph’s Uni-
www.sestech.com/ versity, Beirut, Lebanon, affiliated with the Univer-
[15] W. C. Gibson, The Method of Moments in Electromagnetics. Boca Raton, sity of Lyon, Lyon, France, and the M.Sc. and Ph.D.
FL, USA: CRC Press, 2014, pp. 25–26. degrees from the École Polytechnique de Montréal,
[16] J. M. Jin, Theory and Computation of Electromagnetic Fields. Hoboken, Montreal, QC, Canada.
NJ, USA: Wiley, 2011, pp. 255–258. From 1971 to 1976, he was a Consulting Engi-
neer with Shawinigan Engineering Company, Mon-
treal. He has worked on numerous projects involving
power system analysis and design, railway electrifica-
Ali Aghabarati received the B.Sc. and M.Sc. degrees tion studies, and specialized computer software code
in electrical engineering from the Iran University of development. In 1976, he joined Montel-Sprecher & Schuh, Montreal, a man-
Science and Technology and Amirkabir University ufacturer of high-voltage equipment, as a Manager of Technical Services and
of Technology, Tehran, Iran, in 2008 and and 2010, was involved in power system design, equipment selection, and testing for sys-
respectively. He received the Ph.D. degree in com- tems ranging from a few to several hundred kilovolts. In 1978, he founded Safe
putational electromagnetics from McGill University, Engineering Services & Technologies Ltd., Laval, QC, Canada, a company that
Montreal, QC, Canada, in 2014. specializes in soil effects on power networks. Since then, he has been responsi-
Since 2014, he has been with Safe Engineering ble for the engineering activities of the company, including the development of
Services & technologies Ltd., Laval, QC, as a Re- computer software related to power system applications. He has authored over
search Scientist. His research interests include com- 300 papers on power system grounding, lightning, inductive interference, and
putational electromagnetics for wave propagation, electromagnetic field analysis. He has also written several research reports for
scattering, and electromagnetic compatibility. the Canadian Electricity Association and the Electric Power Research Institute.