1

Substation Fault Analysis Requirements

M. Kezunovic, Fellow, IEEE

with a goal of improving fault analysis. The requirements are

Abstract—Substation automation has critical role in power aimed at expanding the substation automation role in
systems. Substations are responsible for protection, control and automated fault analysis towards better serving many utility
monitoring functions that allow robust routing of power from groups: operations, protection and asset management.
generators to loads through a complex network of transmission
lines. With the latest technology development, many intelligent II. BACKGROUND
electronic devices (IEDs) available in substations today are
capable of performing enhanced functionalities beyond what To better understand the requirements for automated fault
their basic function is. This brings an opportunity for adding new analysis in substations, several research efforts were
functionalities that go well beyond what the traditional undertaken, many of them still on-going:
substation automation solutions have provided.
• A solution for automated merging of data captured by
digital fault recorders (DFRs), digital protective relays
Index Terms—substation automation, expert system, fault
location, genetic algorithm, intelligent electronic device, neural (DPRs), and circuit breaker monitors (CBMs) is
network, power system monitoring, substation measurement, developed [1-10]
sampling synchronization, asset management, alarm processing. • A concept for detecting, classifying and mitigating
cascading events based on local and system-wide
I. INTRODUCTION monitoring data is demonstrated [11-14].

V arious types of users in the utilities may benefit from

substation automation: operations, protection, asset
management, market operations, etc. In order to achieve such
• An optimal fault location algorithm that uses data from
substation IEDs, as well as data from SCADA PI
Historian and simulation data from short circuit
benefits, new concepts of substation automation have to meet program is implemented [15-25]
the following expectations: • A risk-based asset management methodology for
• Local protection concept needs to be enhanced to maintenance scheduling taking into account condition-
simultaneously improve dependability and security based data captured by substation IEDs is being
• Substations need to be interfaced directly and developed [26-29]
through coordinating centers for better system-wide • An intelligent alarm processor approach to take
protection advantage of enhanced protective relay data in
• Data from IEDs needs to be merged with data from explaining cause-effect relationships between alarms is
SCADA to enhance monitoring capabilities proposed [30-32]
• Condition based data from substation IEDs needs to • A Neural network based protective relaying scheme
be utilized for failure rate assessment of assets that enables simultaneous enhancements in
• Local automated analysis needs to be coordinated dependability and security of transmission line
with centralized analysis to cope with N-m cascades protection is envisioned [33-37]
• Control actions need to be defined hierarchically
from substations to control centers for best outcome III. AUTOMATED FAULT ANALYSIS REQUIREMENTS
• Monitoring and reporting of disturbances needs to be The conducted research resulted in several potential and
automated to meet recent NERC standards demonstrated requirements brought forward by advanced
substation automation and fault analysis solutions:
This paper summarizes requirements for automated fault • Operator efficiency: automated fault analysis of events,
analysis functions that may be performed in substations in the which increases speed of drawing conclusions leading
future. Particular focus is on requirements for implementing a to better staff decisions, reduces time of restoration
new concept of merging operational and non-operational data • Assessment accuracy: fault location and asset
management functions will increase quality of
This work was jointly supported by DOE, EPRI, ERCOT, PSerc, and several conclusions related to maintenance and restoration
utilities (AEP, CenterPoint Energy, FirstEnergy. Oncor, and Hydro One) • Operator awareness: intelligent alarm processor
M. Kezunovic is with the Department of Electrical and Computer combined with optimized fault location gives more
Engineering, Texas A&M University, College Station, TX 77843-3128, USA
(e-mail: [email protected]).
comprehensive view of the faults and consequences
 2

• Equipment operations: automated analysis of faults and system protection, circuit breaker operations, fault location,
disturbances as well as circuit breaker operation using etc [6, 7]. The software modules are shown in Fig. 2.
IED data, can offer better monitoring than SCADA
• Protective relaying: new approaches based on pattern
recognition and accurate fault location are more
dependable and secure than distance protection
• Regulatory compliance: as regulatory requirements
from NERC standards are increasingly becoming more
stringent, new solutions offer much more details about
disturbances than ever before

IV. SMART GRID AUTOMATED ANALYSIS

A. Automated merging of IED data

Fig. 2. Digital Fault Recorder Analysis
Proper substation data integration and information
exchange are the main requirements to enable full IED data
The Circuit Breaker Monitor Analysis (CBMA) performs
utilization. The Fig.1 shows substation data integration
analysis of data in COMTRADE format taken from the circuit
conceptual diagram. The benefits of the IED data integration
breaker control circuit using a Circuit Breaker Monitor (CBM)
can be found in [1, 2]. Assuming that all devices are
and generates report in ASCII format that describes circuit
synchronized using Global Positioning System (GPS) of
breaker operating conditions and suggests repair actions [8,
satellites or some other time synchronization technique (IEEE
9]. It enables protection engineers, maintenance crews and
1588) the first steps towards substation data integration were
operators to quickly and consistently evaluate circuit breaker
introduction of standardized file format, COMTRADE [3],
performance, identify performance deficiencies, and trace
and standardized IEEE naming convention [4]. By converting
possible reasons for malfunctioning. CBMA software modules
all data to the same format and using file naming standard it
are shown in Fig. 3.
makes data transparent to adding new analysis functions.
Other requirements to make this concept possible are
assumption that substation data are collected automatically
and integrated into substation database with system
configuration data. The developed applications for processing
data from DFR (Digital Fault Recorder Analysis-DFRA),
CBM (Circuit Breaker Monitor Analysis- CBMA) and DPR
(Digital Protective Relay Analysis-DPRA) use the integrated
data from the database so that the vendor and product model
specifics do not affects their implementation and further
extensions. The detailed system solution is described in [5]. Fig.3. Circuit Breaker Monitor Analysis

The Digital Protective Relay Analysis (DPRA) is an expert

system which automates validation and diagnosis of relay
operation [5, 10]. It takes various relay reports and files as
inputs and using embedded expert system generates a report
on the results of analysis. Validation and diagnosis of relay
operation is based on comparison of expected and actual relay
behavior in terms of the status and timing of logic operands.
DPRA software modules are shown in Fig. 4.

Fig.1. Substation data integration diagram

The following are examples of the automated analysis

applications:
Digital Fault Recorder Analysis (DFRA) performs signal
processing to identify pre- and post-fault analog values,
statuses of the digital channels (corresponding to relay trip,
breaker auxiliary, communication signals), fault type, fault
location and faulted phases. It also checks and evaluates Fig.4. Digital Protective Relay Analysis
 3

B. Detecting and mitigating cascading events sparse as recording devices (DFRs or IEDs with DFR
Power systems are exposed to many kinds of disturbances. capabilities) are generally not installed at the ends of each
Among those disturbances, cascading events draw special transmission line (instead they are installed in critical
attention since they can cause great economic loss to utility substations only) and sometimes not all the DFRs installed
companies and other businesses and devastating impact on may not be triggered by a fault. Although protective relays
people’s life. Considering the large number of individual exist on every transmission line, most of them may still be
components it involves, the wide range of time scales of the electromechanical and they do not have capability to record
event dynamics, and the different mechanisms of how the measurements. Scarcity of measurements at the line ends
components interact, the detecting and mitigating of cascading require system-wide sparse measurement based fault location
events is extremely complex. method [16], which may come from only some substations in
A new monitoring and control scheme for detection and the region where the fault has occurred.
mitigation of cascading events that coordinates the system- An optimal fault location method which can obtain accurate
wide and local substation algorithms has been proposed [11]. fault location by selecting proper fault location algorithm from
The overall interaction is conceptually shown in Fig. 5. the following algorithms depending on the availability and
location of the measurements is proposed: (1) Two or
multiple-ended time-domain synchronized sampling [17], (2)
Two-ended phasor-based asynchronized sampling [18], (3)
Single-ended phasor-based [19] and (4) System-wide sparse
measurement based [16] method.
1) Implementation of the OFLA method
The architecture of the OFLA scheme is shown in Fig. 6
[20]. Several commercial packages are used to implement this
solution. The required data [21] are: (1) System level data
which include power system model data (modeled using
PSS/ETM 31 [22]) and data reflecting real time changes in
power system (SCADA PI-Historian data [23] at pre-fault)
and (2) Field data which include event data recorded by
different IEDs (DFRs) after occurrence of any abnormality.
Fig. 5. Overall interactive scheme for protection

The system-wide monitoring and control tool can find the

vulnerable elements and send request to the local tool for
detailed monitoring. The vulnerability and security margin
information can be obtained by Vulnerability Index (VI) and
Margin Index (MI) [12]. Emergency control approaches for
expected events can be found by the routine security analysis
and activated when such events occur. Emergency control
approaches for unexpected events can be found by event-
based security analysis and activated to mitigate the
disturbance and keep the system secure.
The local monitoring and control tool can find the exact
disturbance information and make a correction if there is relay Fig. 6. Architecture of Optimal Fault Location Algorithm
failure or unintended operation [13, 14]. Further information
can be sent to the system-wide monitoring and control tool for The implementation of the software (using data captured
better security control. The substation-based solution can by DFRs converted into information and integrated with
provide the system-wide tool with local disturbance system level data) should be done in the following step by
information and diagnostic support so that the system-wide step process [24]: (1) Extraction of phasors, (2)
tool can utilize local information to take better control action Synchronization of phasors, (3) Tuning the power system
to ensure the secure operation. model with real-time power system conditions and (4)
C. Optimal Fault Location Application (OFLA) Choosing the appropriate algorithm to find accurate location
of the fault.
Accurate location of transmission line faults is highly
2) Evaluation of the OFLA method
desirable to help find and repair the faulted line quickly.
The software is implemented using Java programming
Traditional fault location techniques [15] require very specific
language. To interact between PSS/E activities and Java
measurements from one or both ends of the line to produce
programming language, IPLAN [25] language (part of PSS/E
results with desired accuracy. In reality, measurements may be
package) is used, which is able to modify the system topology,
 4

control the load flow and short circuit studies, and control the parameters such that if new value of ‘ti’ falls in this range,
reporting of the results of the PSS/E activities. then those part of breaker which cause the occurrence of time
The method was tested for real life test case provided by a instant ‘ti’, operate properly. These limits are the expert
utility where DFR data of two subsequent (5ms gap) phase to settings used in developing automated analysis of CB
ground faults occurred in one circuit between two substations operation as reported earlier [29]. Fig. 7 shows the probability
and PI Historian data (extracted in *.xls format) was provided distribution function of timing parameter t2, result of Bayesian
for both of the substations for 10s time interval for a duration updating approach. The shaded area between the lower and
from pre-fault to post-fault. DFR triggered for only one upper limits is the probability that the breaker will operate
substation for both of the faults. The DFR-resident algorithm properly.
could estimate the fault location for the 1st fault only while the
new method estimated that accurately for both. The method is 0.35
Probability Between Limits is 0.94016
also robust as the same measurements may be obtained by
multiple IEDs, which allows redundancy that can be explored 0.3

to account for bad measurements. The method can be totally 0.25

automated to perform quickly.

Density
0.2
D. Risk-based asset management
For cost-effective maintenance scheduling of power system 0.15

equipment, the problem may be formulated as follows: if it is 0.1

the same availability of labor crews, and labor hours, and the
given budget is constrained, how the maintenance decision 0.05

needs to be optimized. A risk-based decision approach is 0

8 10 12 14 16 18 20
proposed which suits best this kind of situation [26]. In this Lower
Critical Value
Upper (msec)
approach, the classic definition of ‘risk’ definition is adopted, Fig. 7. Probability distribution function of timing parameter t2
which is the product of event probability and event
consequence. Condition monitoring devices can be used to get For the component to operate properly, all of the timings
informed about the equipment condition up to date, which should fall into the specified interval.
Pf (B ) = 1 − ∏ P(ti )
plays major role in this approach. The ‘Event probability’ is
(1)
updated after a specified maintenance action and risk is
recalculated, and the difference is the direct result of that Note that this failure probability is different from the failure
maintenance activity. Following are the definitions used to rate or failure frequency often used in reliability index. Now,
designate the risk in this section. the event probability is computed as the product of failure
Event, E: ‘Failure of a component or a group of probabilities of components involved in that event.
components to operate properly’. Components can be line, To illustrate the impact of the event on the system, two
breaker, transformer or bus bar. different scenarios are considered: single and multiple
Event probability, p(E): ‘Probability that a component or a contingencies. Single contingency involves fault on bus bar,
group of components fail to operate properly’. line, or breaker. Multiple contingency involves failure of more
Event consequence, Con(E): ‘Impact of failure of a than one component. In conclusion, for both scenarios, the
component of a group of the component on the system’. event consequence term can be divided into four parts: loss of
Event risk, R(E): ‘The product of event probability and load, loss of line, loss of generator and repair cost. The total
consequence is the summation of their costs.
event consequence’.
In order to evaluate the event probability, it is necessary to Con(E ) = Conload + Conline + Congen + RepairCost (2)
estimate the failure probability of all components involved in Based on the risk concept associated with the product of
that event. The focus in this study is the circuit breaker but event probability and event consequence, a maintenance
other components may be viewed the same way. The breaker scheduling is proposed. According to the risk number of
data is basically a record of wave forms taken from the circuit components, a maintenance schedule for equipment can be
breaker control circuit by using a portable or on-line recorder defined.
[27] while manually or automatically forcing a breaker E. Advanced alarm processor approach
operation. Signal processing and expert system modules
A major power system disturbance could trigger hundreds
developed in [28] can be used to extract the various features
and sometimes thousands of individual alarms and events
of the waveforms. Each of these extracted timing parameters [30]. For the case in this paper, there were 2125 alarm
is fitted with probability distribution and failure probability messages pop up within only 45minutes.
index is defined based on the probability distributions. The The alarm diagnosis algorithm and model have been
distributions and hence the failure probability index is updated illustrated in [31, 32]. The protection system configurations
using Bayesian approach as the new data comes. for sample test cases are shown in Fig.8 and Fig. 9
First of all, define upper and lower limits for each timing respectively.
 5

Fig.10. Model analysis procedure for Line

BBSES_60A_with_assumed_relay_data
Fig.8. Protection System Configuration Diagram
F. Neural network based protective relaying
The system consists of 9 sections, including 3 buses, 2 The neural network based algorithm classifier is used to
generators and 4 transmission lines. detect and classify the disturbances that require protective
The logic reasoning method uses the relay status obtained
relay action [33, 34]. Comparing with traditional method, this
from the online-database to validate each candidate fault
neural network based fault diagnosis algorithms uses the time-
section. The strategy is to build one diagnosis model for each
domain voltage and current signals directly as patterns instead
section of the power system. Each model establishes reasoning
starting from a set of SCADA data to the conclusion of fault of calculating phasors. This neural network based protective
occurrence on its associated section with certain truth degree relaying scheme compares the input voltage and current
value. signals with well-trained prototypes instead of predetermined
settings, which enables simultaneous enhancements in
dependability and security of transmission line protection.
A self-organized, fuzzy ART neural network based fault
detection and classification algorithm has been developed,
which is conceptually shown in Fig 11. [35, 36]

Fig.9. Model Analysis Procedure for Line BBSES_60A

CASE 1: No protective relay signals. Circuit breaker

CB4210, CB4220, CB4160, CB4920 status changes are Fig. 11. Fuzzy ART neural network algorithm
detected.
Diagnosis result: Line BBSES_60A is faulted, and its truth The neural-network training uses a mechanism of clustering
value is 0.5130. technique with combined unsupervised and supervised
CASE 2: The operation of the circuit breaker is caused by learning. Voltage and current signals from the local
tripping by the associated relays, thus allowing the relay status measurement are formed as patterns by certain data processing
to be obtained to validate the fault. The assumption is that the method. Thousands of such patterns obtained from power
relay trip signals related to this case are available. All the system simulation or substation database of field recordings
devices worked correctly with no false signals. Circuit breaker are used to train the neural network offline and then the
CB4210, CB4220, CB4160, and CB4920 are detected. pattern prototypes are used to analyze faults on-line by using
Diagnosis result: Line BBSES_60A is faulted, and its truth the Fuzzy K-NN classifier. The more details about this can be
value is 0.8550. With the input of the related relay signals, the found in [36, 37].
fault certainness has been increased dramatically.
 6

V. ACKNOWLEDGMENT [22] PSS™E 31.0, Siemens Power Transmission & Distribution, Inc., PTI,
Schenectady, NY, USA.
My graduate students at TAMU contributed to the solutions [23] OSISoft Tech Support, http://techsupport.osisoft.com/
discussed in this paper: Z. Djekic, P. Dutta, M. Knezev, Y. [24] M. Kezunovic, P. Dutta, “Fault Location Using Sparse Wide Area
Measurements”, to be published on Cigré SC B5 colloquium, 2009
Liao, X. Luo, J. Mrkic, S. Natti, C. Pang, Z. Ren, H. Song, S. [25] IPLAN Program Manual, PSS™E 31.0, Release 17.
Vasilic, Y. Wu, and N. Zhang. [26] S. Natti, M. Kezunovic, "A Risk-based Decision Approach for
Maintenance Scheduling Strategies for Transmission System
VI. REFERENCES Equipment," The 10th Intl. Conf. on Probabilistic Methods Applied to
Power Systems – PMAPS 08, Rincon, Puerto Rico, May 2008.
[1] M. Kezunovic, A. Abur, "Merging the Temporal and Spatial Aspects of [27] M. Kezunovic, G. Latisko, N. Ved, "Real-Time Monitoring and
Data and Information for Improved Power System Monitoring Assessment of Circuit Breaker Operations for Diagnostics and Control
Applications," IEEE Proceedings, Vol. 9, no. 11, pp. 1909-1919, 2005. Applications," EPRI Substation Equipment Diagnostics Conference
[2] M. Kezunovic, "Future c of Substation Data," 7th International XIII, New Orleans, Louisiana, March, 2005.
Conference on Advances in Power System Control, Operation and [28] M. Kezunovic, Z. Ren, G. Latisko, D.R. Sevcik, J. Lucey, W. Cook, E.
Management - APSCOM 2006, Hong Kong, October/November 2006. Koch, "Automated Monitoring and Analysis of Circuit Breaker
[3] IEEE Inc., "IEEE Standard Common Format for Transient Data Operation," IEEE Transactions on Power Delivery, Vol. 20, No. 3, pp.
Exchange(COMTRADE) for Power Systems", IEEE Std. C37.111-1999 1910-1918, July 2005.
[4] IEEE Inc., “IEEE Recommended Practice for Naming Time Sequence [29] Christopher Donald Nail, "Automated Circuit Breaker Analysis", M.Sc.
Data Files”, IEEE Std C37.232-2007 Thesis, Dept. of ECE., Texas A&M University, College Station, TX,
[5] M. Kezunovic, T. Popovic, "Substation Data Integration for Automated 2002.
Data Analysis Systems," IEEE PES General Meeting, Tampa, Florida, [30] W. Prince, B. Wollenberg and D. Bertagnolli, “Survey on excessive
June 2007. alarms,” IEEE Trans. on Power Systems, pp. 188-194, August 1989.
[6] M. Kezunovic, I. Rikalo, "Automating the Analysis of Faults and Power [31] X. Luo, M. Kezunovic, "Implementing Fuzzy Reasoning Petri-nets for
Quality," IEEE Computer Applications in Power, Vol. 12, No. 1, pp. 46- Fault Section Estimation," IEEE Transactions on Power Delivery, Vol.
50, January 1999 23, No. 2, pp. 676-685, April 2008.
[7] M. Kezunovic, "Automation of Fault Analysis Using DFR Data," in [32] X. Luo, M. Kezunovic, "An Expert System for Diagnosis of Digital
Proceedings of the 60th American power Conference, Chicago, IL, pp. Relay Operation," 13th Conference on Intelligent Systems Application to
91-98, Apr 1998. Power Systems, Washington DC, USA, November 2005.
[8] M. Kezunovic, Z. Ren, G. Latisko, D.R. Sevcik, J. Lucey, W. Cook, and [33] M. Kezunović, S. Vasilić, F. Gul-Bagriyanik, “Advanced Approaches
E. Koch, "Automated monitoring and analysis of circuit breaker for Detecting and Diagnosing Transients and Faults,” Med Power 2002,
operation,” IEEE Transactions on Power Delivery, vol. 20, no. 3, pp. Athens, Greece, November 2002.
1910-1918, Jul 2005. [34] M. Kezunović, S. Vasilić, D. Ristanović, “Interfacing Protective Relays
[9] M. Knezev, Z. Djekic, and M. Kezunovic, "Automated Circuit Breaker and Relay Models to Power System Modeling Software and Data Files,”
Monitoring," IEEE PES General Meeting, Tampa, FL, Jun 2007, pp. 1-6. PowerCon 2002, Kunming, China, October 2002.
[10] X. Luo, M. Kezunovic, "Automated Analysis of Digital Relay Data [35] M Kezunović, S. Vasilić, “Analysis of Protective Relaying Operation
Based on Expert System," PowerTech 2005, St. Petersburg, Russia, June and Related Power System Interaction,” IFAC Symposium on Power
2005. Plants and Power System Control, Seoul, Korea, September, 2003.
[11] H. Song, M. Kezunovic, "A New Analysis Method for Early Detection [36] S. Vasilić, M. Kezunović, “Fuzzy ART Neural Network Algorithm for
and Prevention of Cascading Events," Electric Power Systems Research, Classifying the Power System Faults,” IEEE Transactions on Power
Vol. 77, no. 8, pp. 1132-1142, June 2007. Delivery, Vol. 20, No. 2, pp 1306-1314, April 2005
[12] H. Song, M. Kezunovic, "Static Analysis of Vulnerability and Security [37] PSerc Projcet S29 Final Report - Part I, “Detection, Prevention and
Margin of the Power System," IEEE 2005 PES Transmission & Mitigation of Cascading Events,” PSerc Publication 08-18. [Online]
Distribution Conference & Exposition, Dallas, Texas, May 2006 Available: http://www.pserc.org
[13] M. Kezunovic, C. Pang, "Improved Transmission Line Protection
During Cascading Events," CIGRE B5 Colloquium, Madrid, Spain,
October 2007 VII. BIOGRAPHIES
[14] N. Zhang, Kezunovic, "Improving Real-time Fault Analysis and
Validating Relay Operations to Prevent or Mitigate Cascading Mladen Kezunovic (S’77-M’80–SM’85–F’99)
Blackouts," IEEE 2005 PES Transmission & Distribution Conference & received the Dipl. Ing., M.S. and Ph.D. degrees in
Exposition, Dallas, Texas, May 2006 electrical engineering in 1974, 1977 and 1980,
[15] M. Kezunovic, B. Perunicic, “Fault Location”, Wiley Encyclopedia of respectively. Currently, he is the Eugene E. Webb
Electrical and Electronics Terminology, Vol.7, pp. 276-285, John Wiley Professor and Site Director of Power Engineering
1999. Research Center (PSerc), an NSF I/UCRC.at Texas
[16] M. Kezunovic, Y. Liao, “Fault Location Estimation Based on Matching A&M University He worked for Westinghouse
the Simulated and Recorded Waveforms Using Genetic Algorithms”, Electric Corp., Pittsburgh, PA, 1979-1980 and the
Development in Power System Protection, Amsterdam, The Netherlands, Energoinvest Company, in Europe 1980-1986. He
April 2001. was also a Visiting Associate Professor at
[17] M. Kezunovic, B. Perunicic, and J. Mrkic, “An Accurate Fault Location Washington State University, Pullman, 1986-1987 and Visiting Professor at
Algorithm Using Synchronized Sampling”, Electric Power Systems the University of Hong Kong in fall 2007. His main research interests are
Research Journal, Vol. 29, No. 3, pp. 161-169, May 1994. digital simulators and simulation methods for relay testing as well as
[18] D. Novosel, D.G. Hart, E. Udren, and J. Garitty, “Unsynchronized two- application of intelligent methods to power system monitoring, control, and
terminal fault location estimation”, IEEE Trans. Power Delivery, Vol. protection. Dr. Kezunovic is a member of CIGRE, Registered Professional
11, No. 1, pp. 130 – 138, Jan. 1996. Engineer in Texas, and a Fellow of the IEEE.
[19] T. Takagi, Y. Yamakoshi, M. Yamaura, R. Kondow, and T. Matsushima,
“Development of a New Type Fault Locator Using the One-Terminal
Voltage and Current Data”, IEEE Trans. on Power App. and System,
Vol. 101, No. 8, pp 2892-2898, Aug. 1982.
[20] M. Kezunovic, E. Akleman, M. Knezev, O. Gonen, S. Natti, “Optimized
Fault Location”, IREP Symposium 2007, Charleston, South Carolina,
Aug 2007.
[21] PSerc Projcet T32 Final Report “ Optimized Fault Location,” PSerc
Publication 08-07. [Online] Available: http://www.pserc.org