Electric Vehicle Integration

in a Smart Microgrid
Environment
Electric Vehicle Integration
in a Smart Microgrid
Environment

Edited by
Mohammad Saad Alam
Mahesh Krishnamurthy
First edition published 2021
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

© 2021 selection and editorial matter, Mohammad Saad Alam and Mahesh Krishnamurthy
individual chapters, the contributors

CRC Press is an imprint of Taylor & Francis Group, LLC

MATLAB® and Simulink® are trademarks of The MathWorks, Inc. and are used with permission.
The MathWorks does not warrant the accuracy of the text or exercises in this book. This book’s
use or discussion of MATLAB® and Simulink® software or related products does not constitute
endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular
use of the MATLAB® and Simulink® software.

The right of Mohammad Saad Alam and Mahesh Krishnamurthy to be identified as the author[/s]
of the editorial material, and of the authors for their individual chapters, has been asserted in
accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988.

Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all materials
reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged, please write and
let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known
or hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.
com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. For works that are not available on CCC, please contact mpkbookspermissions@
tandf.co.uk

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are
used only for identification and explanation without intent to infringe.

ISBN: 978-0-367-42391-9 (hbk)

ISBN: 978-1-032-01050-2 (pbk)
ISBN: 978-0-367-42392-6 (ebk)

Typeset in Times
by codeMantra
Contents
Foreword ..................................................................................................................vii
Preface.......................................................................................................................ix
Editors .................................................................................................................... xiii
Contributors ............................................................................................................. xv

Chapter 1 Trends in Electric Vehicles, Distribution Systems, EV Charging

Infrastructure, and Microgrids .............................................................1
April Bolduc

Chapter 2 Fog Computing for Smart Grids: Challenges and Solutions ................7
Linna Ruan, Shaoyong Guo, Xuesong Qiu, and Rajkumar Buyya

Chapter 3 Opportunities and Challenges in Electric Vehicle Fleet

Charging Management ....................................................................... 33
Chu Sun, Syed Qaseem Ali, and Geza Joos

Chapter 4 Challenges to Build a EV Friendly Ecosystem: Brazilian

Benchmark ......................................................................................... 73
Ana Carolina Rodrigues Teixeira

Chapter 5 Coordinated Operation of Electric Vehicle Charging and

Renewable Power Generation Integrated in a Microgrid ...................97
Alberto Borghetti, Fabio Napolitano, Camilo Orozco Corredor,
and Fabio Tossani

Chapter 6 Energy Storage Sizing for Plug-in Electric Vehicle Charging

Stations ............................................................................................. 119
I. Safak Bayram, Ryan Sims, Edward Corr, Stuart Galloway,
and Graeme Burt

Chapter 7 Innovative Methods for State of the Charge Estimation for EV

Battery Management Systems .......................................................... 143
Zeeshan Ahmad Khan and Franz Kreupl

v
vi Contents

Chapter 8 High-Voltage Battery Life Cycle Analysis with Repurposing in

Energy Storage Systems (ESS) for Electric Vehicles........................ 181
Mamdouh Ahmed Ezzeldin, Ahmed Alaa-eldin Hafez, Mohamed
Adel Kohif, Marim Salah Faroun, and Hossam Hassan Ammar

Chapter 9 Charging Infrastructure for Electric Taxi Fleets.............................. 213

Chandana Sasidharan, Anirudh Ray, and Shyamasis Das

Chapter 10 Machine Learning-Based Day-Ahead Market Energy Usage

Bidding for Smart Microgrids........................................................... 249
Mohd Saqib, Sanjeev Anand Sahu, Mohd Sakib, and
Esaam A. Al Ammar

Chapter 11 Smart Microgrid-Integrated EV Wireless Charging Station ........... 267

Aqueel Ahmad, Yasser Rafat, Samir M. Shariff, and Rakan Chabaan

Chapter 12 Shielding Techniques of IPT System for Electric Vehicles’

Stationary Charging.......................................................................... 279
Ahmed A. S. Mohamed and Ahmed A. Shaier

Chapter 13 Economic Placement of EV Charging Stations within

Urban Areas.................................................................................... 295
Ahmed Ibrahim AbdelAzim

Chapter 14 Environmental Impact of the Recycling and Disposal of EV

Batteries ............................................................................................ 313
Zeeshan Ahmad Arfeen, Rabia Hassan, Mehreen Kausar Azam,
and Md Pauzi Abdullah

Chapter 15 Design and Operation of a Low-Cost Microgrid-Integrated EV

for Developing Countries: A Case Study.......................................... 335
Syed Muhammad Amrr, Mahdi Shafaati Shemami,
Hanan K. M. Irfan, and M. S. Jamil Asghar

Index....................................................................................................................... 359
Foreword
It is an honor for me to add my comments to a very important book by Dr. Mohammad
Saad Alam and Dr. Mahesh Krishnamurthy, Electric Vehicle Integration in Smart
Microgrid Environment.
Transport electrification, as a response to the global decarbonization goal, has
advanced significantly over the decade. The IEA, in its Global Electric Vehicle (EV)
Outlook 2020, projects an increase in the global EV stock (excluding two/three
wheelers) from around 8 million in 2019 to 50 million by 2025 and close to 140
million vehicles by 2030, constituting about 7% of the global vehicle fleet by 2030.
The smooth and successful adoption of projected EVs will require an ecosystem that
facilitates the efficient operation of EVs. However, this large-scale integration of EVs
with the grid comes with its set of technical and economic challenges. The high pen-
etration of EVs can affect the power quality of the distribution network. Challenges
facing the utilities such as power quality issues, proper planning of charging stations,
interoperability of charging infrastructure, peak power demand, battery manage-
ment, etc. are important issues to be addressed.
In this edited book, the authors have examined the current energy scenario for
microgrids and have discussed the challenges and opportunities due to the increas-
ing penetration of distributed power generation systems and EVs into the microgrids.
Through the course of 15 chapters, the various aspects of EV integration with the
grid are presented – current trends in EV and their charging stations, the opportuni-
ties and challenges, EV-connected microgrid planning, power market operation and
planning, innovative methods for EV battery management, and economic and envi-
ronmental impacts. This book is a timely endeavor as it has captured the state-of-the-
art technologies and trends in smart microgrid management with EV integration, EV
charging infrastructure technologies, smart charging, and deployment.
I believe, the contents of this book will expose the readers to subjects that could
potentially alter the paradigm for integration of transportation electrification to
microgrids and emerging EV charging infrastructure technologies and would serve
as a valuable tool and guide to a wide spectrum of stakeholders such as power system
architects, practitioners, developers, new researchers, and graduate-level students,
especially for developing countries of the Middle East and South East Asia, to pro-
duce an even more capable and diverse insight in various domains from smart home,
smart cities, industry, business, and consumer applications. I would like to congratu-
late Dr. Mohammad Saad Alam and Dr. Mahesh Krishnamurthy for their keen inter-
est in the emerging area of integration of EVs with the grid of the future and thank
them for their efforts on introducing state-of-the-art topics to all of us.

Reji Kumar Pillai

President, India Smart Grid Forum (ISGF)
Chairman, Global Smart Energy Federation (GSEF)

vii
Preface
Electric vehicles (EVs) offer numerous benefits in comparison with traditional
gasoline-powered vehicles, including lower operating costs, zero tail-pipe emissions
and the potential to run on locally generated renewable energy. Wider adoption of
EVs could take us closer to reducing the environmental impacts of the transpor-
tation infrastructure and achieving energy independence. Although these are very
important goals, several challenges still exist in the wider adoption of EVs that need
careful consideration. This leaves considerable scope for research and innovation.
Among the current issues being discussed, one of the biggest challenges for automo-
tive manufacturers, utilities, and customers is the issue of charging battery for EVs
to recreate the "filling a tank" experience without overwhelming an ageing grid.
A solution to this problem has been presented in the form of a microgrid concept.
The ability to generate, control, and expend energy locally can reduce computational
complexity, shorten response times, and create solutions that are cost-effective, while
being easier to deploy and maintain.
A cursory review of existing literature in the area of grid integration of EVs shows
that there are several important aspects that need to be carefully studied. This book
provides a technological insight into several of these topics and includes specific case
studies that will help the reader identify the context of the present application.
With growing acceptance of EVs, it is very important to consider the overall eco-
system in context. Chapter 1 deals with the fundamentals of the EVs, microgrid, and
charging infrastructure, starting with energy consumption in charging stations and
laying groundwork for integration of renewable energy sources. Chapter 2 explores
numerical methods in smart grids. Chapter 2 evaluates fog computing-enabled smart
grid applications and identifies key challenges and the possible approaches.
In order to evaluate the impact of electrification, it is important to consider fleet
operation of EVs. This topic is tackled in Chapter 3. Starting with fundamentals of
fleet charging management for EVs, it introduces charging technologies and aggre-
gated chargers towards providing V2G services and identifies challenges associated
with their deployment. A specific application example is considered in Chapter 9,
which presents a holistic overview of electric taxi operations with case studies focus-
ing on charging infrastructure. The chapter also provides multi-criterion decision-
making tools for the selection of charging technology and sites for fleet charging.
A case study for an EV ecosystem in the Brazilian system is studied in Chapter 4,
where specific challenges are presented in adoption of EVs in the country.
Chapter 5 focuses on the coordinated operation of EVs and renewable energy in a
microgrid by studying parking lots equipped with bidirectional charging stations and
renewable generation such as photovoltaic (PV) panels and stationary battery storage
units. In order to accommodate peak power demand in a microgrid system, stationary
energy storage systems are being seen as very strong candidates, which is discussed
in Chapters 6–8. Chapter 6 takes a practical outlook and presents criteria for sizing
the energy storage system in an EV charging station. It presents case studies to show
relationships between energy storage size, grid power, and PEV demand and also

ix
x Preface

explores methods to reduce peak electricity consumption and the station’s monthly
electricity bill. Chapter 7 provides an innovative approach for reducing the computa-
tional effort for SoC estimation by adaptively resampling the current measurements
running at a fixed sample rate. The algorithm is not limited to new developments but
can also be implemented in existing commercial systems. For validating the algo-
rithm, a worldwide harmonized light vehicle test procedure is applied with charging
and resting phases. To estimate SoC, a modified 2-RC model in combination with
an extended Kalman filter is implemented. Chapter 8 and 14 takes a deep dive into
the environmental impacts of batteries and explores the environmental impacts and
challenges in repurposing and disposal of high-voltage batteries.
In order to evaluate the impact of electrification, it is important to consider fleet
operation of EVs. This topic is tackled in Chapters 3 and 9. Starting with funda-
mentals of fleet charging management for EVs, it introduces charging technologies
and aggregated chargers towards providing V2G services and identifies challenges
associated with their deployment. A specific application example is considered in
Chapter 9, which presents a holistic overview of electric taxi operations with case
studies focusing on charging infrastructure. The chapter also provides multi-criterion
decision-making tools for the selection of charging technology and sites for fleet
charging.
Chapter 10 deals with various aspects of grid integration of EVs. Chapter 10 pres-
ents an expert system of the bidding process using cloud-based artificial intelligence.
The next level of the charging infrastructure deals with emerging charging technolo-
gies. Chapter 11 takes a wireless approach to the charging concept and demonstrates
a PV-based charging approach. Chapter 12 explores shielding types, impact, and
design in inductive power transfer by exploring three main shielding techniques, pas-
sive, active, and reactive.
To address practical challenges in implementing EV charging in different envi-
ronments, a case study for an EV ecosystem in the Brazilian system is studied in
Chapter 4, where specific challenges are presented in adoption of EVs in the country.
Chapter 13 discusses technical and economic challenges and practical constraints
of choosing the locations of EV charging stations within a large metropolitan area
highlighting their merits and shortcomings of various approaches citing examples
from the city of Cairo, Egypt. For the Indian market, Chapter 15 explores the design
and realization of a low-cost solar PV-based microgrid system for providing alternate
mode of EV charging while mitigating the load-shedding scenarios.
It is expected that this book will serve as a reference for a larger audience such
as power system architects, practitioners, developers, researchers, and graduate-level
students, especially for developing countries of the Middle East and Southeast Asia.

Mohammad Saad Alam

Mahesh Krishnamurthy
Preface xi

MATLAB® is a registered trademark of The MathWorks, Inc. For product infor-

mation, please contact:
The MathWorks, Inc.
3 Apple Hill Drive
Natick, MA 01760-2098 USA
Tel: 508-647-7000
Fax: 508-647-7001
E-mail: [email protected]
Web: www.mathworks.com
Editors
Mohammad Saad Alam is a professor at the Department of Electrical Engineering,
Aligarh Muslim University and Founding Director of the Center of Advanced
Research in Electrified Transportation (CARET). Before joining Aligarh Muslim
University, he has worked in the North American Automotive Industry in the
Electric Vehicle research and product development. His current research inter-
ests include electric mobility and connectivity, xEV charging infrastructure,
smart Microgrid n, large-scale new and renewable energy integration, high-volt-
age electric energy storage systems, transactive energy, block chain application
to Transportation electrification and big data analytics in the sustainable energy
industry. He has coauthored over 150 publications. He is an associate editor of the
IEEE Transactions on Transportation Electrification and the Journal of Modern
Power System and Clean Energy.

Mahesh Krishnamurthy is a professor of Electrical Engineering and the director

of the Electric Drives and Energy Conversion Lab and Grainger Power Electronics
and Motor Drives Laboratory at the Illinois Institute of Technology. Before joining
Illinois Tech, he worked as a design engineer at EF technologies in Arlington, Texas.
His research primarily focuses on design, analysis, and control of power electron-
ics, electric machines, motor drives and energy storage for electrified transportation,
renewable energy, and industrial applications.
Dr. Krishnamurthy was the recipient of the 2006–2007 IEEE VTS-Transportation
Electronics Fellowship Award for his contributions. Since 2015, he has been serving
as a Distinguished Speaker with the IEEE-Vehicular Technology Society after serv-
ing as a Distinguished Lecturer from 2011–2013 to 2013–2015. He has co-authored
over 125 scientific articles, book chapters, and technical reports. He is currently the
advisor for the Formula Electric racecar team at Illinois Tech, which won the pres-
tigious Fiat Chrysler Innovation award at the SAE Formula Hybrid Competition and
the NASA Robotic Mining Competition team. He has received several teaching and
research awards, including the 2019 Thomas Jacobius Excellence in Inter-professional
Education Award (single award), 2017 Bauer Family Teaching Excellence Award
(single award), and the 2017 Armour Excellence in Education Award (single award)
at Illinois Tech.
Dr. Krishnamurthy was the General Chair for the 2014 IEEE Transportation
Electrification Conference and Exposition. In the past, he has served as the Technical
Program Chair for the 2011 Vehicle Power and Propulsion Conference and 2013
IEEE- Transportation Electrification Conference. Dr. Krishnamurthy has served as
the guest Editor or associate Editor for several IEEE journals including the Special
Section of IEEE Transactions on Vehicular Technology on Sustainable Transportation
Systems, Special Issue of IEEE Transactions on Power Electronics on Transportation
Electrification and Vehicle Systems, and Special Issue of IEEE Journal of Emerging

xiii
xiv Editors

and Selected Topics in Power Electronics on Transportation Electrification. He is

currently serving as an editor for IEEE Transportation Electrification Magazine,
Chair for the IEEE Power Electronics Society Technical Committee on Vehicle and
Transportation Systems, and editor-in-chief for IEEE Transactions on Transportation
Electrification.
Contributors
Ahmed Ibrahim AbdelAzim M. S. Jamil Asghar
Ethos Esco Consultancy Center of Advanced Research in
Cairo, Egypt Electrified Transporation
Aligarh Muslim University
Md Pauzi Abdullah Aligarh, India
School of Electrical Engineering
University Technology Malaysia Mehreen Kausar Azam
Johor Bahru, Malaysia College of Engineering and Sciences
Institute of Business Management Sindh
Aqueel Ahmad
Karachi, Pakistan
Center of Advanced Research in
and
Electrified Transportation
N.E.D University of Engineering &
Aligarh Muslim University
Technology
Aligarh, India
Karachi, Pakistan
Hossam Hassan Ammar
I. Safak Bayram
School of Engineering and Applied
Department of Electronic and Electrical
Science
Engineering, Faculty of Engineering
Nile University
University of Strathclyde
Giza, Egypt
Glasgow, United Kingdom
and
Smart Engineering Systems Research
April Bolduc
Center (SESC)
S Curve Strategies
Nile University
San Diego, USA
Giza, Egypt
Alberto Borghetti
Essam A. Al-Ammar
Department of Electrical Engineering
King Saud University
University of Bologna
Riyadh, Saudi Arabia
Bologna, Italy
Syed Muhammad Amrr
Graeme Burt
Indian Institute of Technology Delhi
Department of Electronic and Electrical
New Delhi, India
Engineering, Faculty of Engineering
University of Strathclyde
Zeeshan Ahmad Arfeen
Glasgow, United Kingdom
School of Electrical Engineering
University Technology Malaysia
Rajkumar Buyya
Johor Bahru, Malaysia
School of Computing and Information
and
Systems
Electrical Engineering Department
The University of Melbourne
The Islamia University of Bahawalpur
Melbourne, Australia
(IUB)
Bahawalpur, Pakistan
xv
xvi Contributors

Rakan Chabaan Shaoyong Guo

Hyundai Kia America Technical State Key Laboratory of Networking &
Center Inc Switching Technology
Superior Township, Beijing University of Posts and
Michigan, USA Telecommunications
Beijing, China
Edward Corr Ahmed Alaa-eldin Hafez
Power Networks Demonstration Centre School of Engineering and Applied
University of Strathclyde Science
Glasgow, United Kingdom Nile University
Giza, Egypt
Camilo Orozco Corredor Rabia Hassan
Department of Electrical Engineering College of Engineering and Sciences
University of Bologna Institute of Business Management Sindh
Bologna, Italy Karachi, Pakistan

Hanan K. M. Irfan
Shyamasis Das
Abul Kalam Azad University of
Alliance for an Energy Efficient
Technology
Economy (AEEE)
Kolkata, India
India
Geza Joos
Mamdouh Ahmed Ezzeldin McGill University
School of Engineering and Applied Montreal, Canada
Science
Nile University Zeeshan Ahmad Khan
Giza, Egypt Development Engineer
and TKI Automotive GmbH
Smart Engineering Systems Research Germany
Center (SESC)
Nile University Mohamed Adel Kohif
Giza, Egypt School of Engineering and Applied
Science
Nile University
Marim Salah Faroun
Giza, Egypt
School of Engineering and Applied
Science Franz Kreupl
Nile University Chair of Hybrid Electronic Systems
Giza, Egypt Technical University Munich
Munich, Germany
Stuart Galloway
Department of Electronic and Electrical Ahmed A. S. Mohamed
Engineering Center for Integrated Mobility Sciences
Faculty of Engineering National Renewable Energy Laboratory
University of Strathclyde (NREL)
Glasgow, United Kingdom Golden, Colorado
Contributors xvii

Fabio Napolitano Mohd Saqib

Department of Electrical Engineering Indian Institute of Technology (Indian
University of Bologna School of Mines)
Bologna, Italy Dhanbad, India

Syed Qaseem Ali Chandana Sasidharan

Opal-RT Technologies Alliance for an Energy Efficient
Canada Economy (AEEE)
India
Xuesong Qiu
State Key Laboratory of Networking & Ahmed A. Shaier
Switching Technology Zagazig University
Beijing University of Posts and Zagazig, Egypt
Telecommunications
Beijing, China Samir M. Shariff
Department of Electrical Engineering
Yasser Rafat Taibah University
Center of Advanced Research in Medina, Saudi Arabia
Electrified Transportation
Aligarh Muslim University Mahdi Shafaati Shemami
Aligarh, India Center of Advanced Research in
Electrified Transporation
Anirudh Ray Aligarh Muslim University
School of Planning and Architecture Aligarh, India
(SPA)
India Ryan Sims
Power Networks Demonstration Centre
Linna Ruan
University of Strathclyde
School of Computing and Information
Glasgow, United Kingdom
Systems
The University of Melbourne
Chu Sun
Melbourne, Australia
McGill University
and
Montreal, Canada
State Key Laboratory of Networking &
Switching Technology
Ana Carolina Rodrigues Teixeira
Beijing University of Posts and
University of São Paulo
Telecommunications
São Paulo, Brazil
Beijing, China
Fabio Tossani
Sanjeev Anand Sahu
Department of Electrical Engineering
Indian Institute of Technology (Indian
University of Bologna
School of Mines)
Bologna, Italy
Dhanbad, India

Mohd Sakib
Aligarh Muslim University
Aligarh, India
 1 Trends in Electric
Vehicles, Distribution
Systems, EV Charging
Infrastructure, and
Microgrids
April Bolduc
S Curve Strategies

CONTENTS
1.1 Introduction: Transportation Electrification Trends .........................................1
1.2 Distribution System Trends ..............................................................................2
1.3 Charging Technology Trends............................................................................4

1.1 
INTRODUCTION: TRANSPORTATION
ELECTRIFICATION TRENDS
With the rapid growth of transportation electrification, efficient electric vehicle (EV)
integration with the grid is becoming exponentially more important. Geographically,
China is leading the EV and electric bus market, followed by Europe and then the
United States. Automakers continue to accelerate their EV manufacturing efforts to
comply with increasingly stringent regulations in these countries. While pandemics
like COVID-19 can demonstrate initial delays in manufacturing, the overall impact
of such world events is low. By 2022, more than 500 models of EVs will be available
globally due to competitive pricing and consumer choice, making EVs attractive to
new buyers in the market.1
Passenger EV sales has grown from 450,000 in 2015 to 2.1 million in 2019 as
battery prices decrease, battery capacity improves for a longer driving range, the
installation of charging infrastructure continues, and EV sales move into new mar-
kets. Globally, sales will increase to 8.5 million by 2025, 26 million by 2030, and 54
million by 2040 when over half of all passenger vehicles sold are electric.2
As for the electricity consumption required by this growing technology grows,
the rise in EV sales increases the demand for more fast charging stations. If the U.S.

1 BNEF https://about.bnef.com/electric-vehicle-outlook/.
2 BNEF https://about.bnef.com/electric-vehicle-outlook/.
1
 2 Electric Vehicle Integration in a Smart Microgrid Environment

reaches its forecasted growth of more than 20 million EVs by 2030, the vehicles
could require annual energy consumption of 93 terawatt-hours (TWh).3 If these vehi-
cles demonstrate larger battery capacities and rates of charge as current automakers
are demonstrating, the collective electricity consumption could reach between 58 and
336 TWh annually.4 By 2040, passenger electric cars could consume 1,290 TWh,
while commercial EVs consume 389 TWh and electric buses consume 216 TWh.5
To prepare the electric grid to support such a need, there is much being done
across the globe. Key drivers for such support include policy requirements for regions
to reduce pollution and meet air quality goals and recognition of EV electricity con-
sumption as an opportunity by the electric power industry to sustain electric load
growth reduced by energy efficiency. Additionally, the demonstrated ability of grid-
integrated technologies such as smart microgrids and managed charging is needed to
smooth the grid transition to accommodate this load – even for the most congested
grids with intermittent power supply across the globe.
The grid must be able to integrate this technology while meeting both the capacity
needs of transportation electrification and the need for increased renewable energy to
reduce greenhouse gas emissions.

1.2 DISTRIBUTION SYSTEM TRENDS

Many utilities are taking a leading role in facilitating transportation electrification.
Trends in increased infrastructure investment, collaboration across utilities, and grid
modernization are apparent. Atlas EV Hub tracks the number of U.S. investor-owned
utility transportation electrification programs being implemented. By April 2020,
almost $3 billion in utility investments were approved or pending approval to sup-
port this growth.6 Increasingly, programs have moved from a focus on light-duty EVs
to medium- and heavy-duty transportation electrification due to the benefits these
vehicles can provide the grid, while at the same time, heavy-duty vehicle charging
could require 1 megawatt per charge.
California utilities have made the majority of this investment and are now cre-
ating a collaborative 10-year plan across the state’s different utilities that looks to
minimize transportation electrification grid impacts and accelerate EV adoption.
The state’s climate, air quality, and economic development goals require broad elec-
trification of both passenger and fleet vehicles and require support for the widespread
adoption of transportation electrification.7 Over the past decade, numerous utility
transportation electrification programs have been filed with their regulating body, the
California Public Utilities Commission, in number and scale. During this time, the
regulator assessed the utility programs that did not contain transportation infrastruc-
ture deployment planning strategies or projections on how to include incremental
transportation electrification load into their distribution and transmission systems.

3 EEI/IEI, November 2018, EV Sales Forecast and the Charging Infrastructure Required through 2030.
4 National Renewable Energy Laboratory, 2018, Electrification Futures Study: Scenarios of Electric
Technology Adoption and Power Consumption for the United States.
5 BNEF https://about.bnef.com/electric-vehicle-outlook/.
6 Atlas EV Hub, 2020. Utility Filings Dashboard. www.atlastevhub.com.
7 California Senate Bill 350, DeLeon, 2015.
Trends in Electric Vehicles 3

Therefore, they proposed a “transportation electrification framework” requiring the

utilities to develop an overarching 10-year plan that details investments in transporta-
tion electrification infrastructure.8
The goal of this framework is to create a process that best harnesses lessons
learned from past regulator proceedings, research, and transportation electrifica-
tion efforts taking place in the state, as well as create a competitive market. Such
a 10-year plan can provide guidance and standardize the key components of trans-
portation electrification programs, such as charging vendor criteria, open access,
cybersecurity, safety, and the length of time a utility should take to interconnect EV
charging infrastructure. Most importantly, a plan like this can encourage utilities
to collaborate across their distribution planning departments to assess the research
from EV charging pilots from within their territories and across the globe to more
fully understand the possible impacts of increasing the load from EVs and how to
best use technology to integrate these efforts with the grid.
An example of such a collaboration is the West Coast Clean Transit Corridor
Initiative in the U.S. made up of nine electric utilities and two agencies represent-
ing more than two dozen municipal utilities that worked together to develop a study
to electrify 1,300 miles of interstate from the Mexican to the Canadian border for
freight haulers and delivery trucks.9 The study proposes a phased approach that
could lead to significant reductions of pollution from freight transportation along the
Pacific Coast providing a roadmap for electric utilities to electrify transportation in a
coordinated fashion. The first phase would involve installing 27 charging sites along
Interstate-5 at 50-mile intervals for medium-duty EVs, such as delivery vans, by
2025. A second phase would expand 14 of the 27 charging sites to also accommodate
charging for electric big rigs by 2030 when it is estimated that 8% of all trucks on the
road in California could be electric. Of the 27 proposed sites, 16 are in California, 5
in Oregon, and 6 in Washington. The study also demonstrated that an additional 41
sites on highways connecting to Interstate-5 should be considered for electrification.
Near- and long-term distribution planning such as this can help determine the
number of shovel-ready charging infrastructure locations vs. those that will trigger
expensive distribution upgrades. For example, a transit agency converting its fleets
to electric buses over time could trigger the need for a new substation upgrade.10
For a majority of grids, improving the modeling and transparency into a distribu-
tion system’s hosting capacity can provide visibility of gaps in grid infrastructure
when aligned with possible charging site locations. This visibility supports charging
infrastructure deployment in regions where the incremental load would not trigger
distribution system upgrades, and where load management technology could defer
otherwise necessary upgrades.
While these gaps are identified and modeled by grid modernization planning
departments, parallel efforts can be performed to design charging infrastructure pro-
grams in distribution system locations where the grid currently has the capacity and

 
 

10 https://ww2.arb.ca.gov/rulemaking/2018/innovative-clean-transit-2018.
4 Electric Vehicle Integration in a Smart Microgrid Environment

where costly upgrades can be avoided. The advancement of smart charging technol-
ogy and the implementation of these efforts in EV charging infrastructure is one of
the best ways to reduce distribution impacts.

1.3 CHARGING TECHNOLOGY TRENDS

The global EV charging infrastructure market is projected to reach $140 billion
by 2030 and grow at an estimated annual rate of 31%.11 Germany, home to major
automakers such as Volkswagen, BMW, and Mercedes, significantly propelled their
demand for EV charging infrastructure by passing a policy to ban internal combus-
tion engines by 2030. Such a rapid pace of adoption will be assisted by charging
innovation and the ability to both manage charging loads and reduce on-peak charg-
ing by incentivizing drivers to shift their charging time when there is the most capac-
ity on the grid. For drivers to participate in any such advancement technology, the
ease of use for the driver or commercial fleet operator must not be hindered.
Standardization of charging technology accessibility and interoperability is a
growing trend. The way the first EV charging technologies in China and the U.S.
evolved are broadly similar, but fast charging in China has one standard, known as
China GB/T, while the U.S. has three EV fast charging standards: CHAdeMO, SAE
Combo, and Tesla.12 Considerations around charging options for EV owners include
the ease in accessibility at the place and time it is needed and that it is competitively
priced.
EVs are often compared to the phenomenon of rooftop solar installations and fre-
quently cluster in a particular neighborhood as awareness grows about the benefits of
the technology, leading to increased adoption. Utility EV time-of-use rates are made
available only to those with an electric car. Until recently, they have been the only
mechanism to encourage off-peak charging. Incentivizing drivers with lower rates to
charge at times of the day when there is more capacity on the grid has proven effec-
tive. Utilities should consider being mindful that unintended consequences can arise
if large numbers of vehicles start to shift to the same time causing new distribution
load spikes. Managed charging is a solution that helps to intelligently stagger vehicle
charging and avoid grid spikes. Transportation electrification programs across the
globe are increasingly including managed charging in their efforts. Managed charg-
ing can be implemented by one-directional load control of the vehicle telematics
or charging station. The goal of managed charging is to avoid costly grid upgrades
and effectively integrate EVs into the grid to help accelerate adoption and advance
a clean, smart, and affordable energy system. Wind and solar energies are now the
cheapest sources of electricity across more than two-thirds of the world, and by 2030,
they undercut commissioned coal and gas almost everywhere, further incentivizing
the transition to transportation electrification.13

11 Electric Vehicle Charging Infrastructure Market: Global Opportunity and Trend Analysis, 2019–
2030, Research and Markets.
12 https://energypolicy.columbia.edu/sites/default/files/file-uploads/EV_ChargingChina-CGEP_
Report_Final.pdf.
13 https://about.bnef.com/new-energy-outlook/.
Trends in Electric Vehicles 5

Managed charging intelligence can be found within the charger or the vehicle itself
and ideally respond to a signal sent from the utility or entity requesting the load shift
and can effectively manage charging efforts to benefit grid needs. Implementing such
charging across fleets of commercial EVs is another way to maximize grid benefits.
EV charging station technology is also evolving with new standards to improve open
access across different EV charging vendors and networks to simplify the charging
experience for drivers. A key to this success is interoperability or the capability of
drivers to use other vendor’s charging networks without having to sign up for each
one separately as has been the requirement in the past. This is made possible through
software that provides the exchange of driver payment data across platforms.
In most cases, a grid modernized for transportation electrification must meet the
capacity needs of EVs as well as the need for clean renewable energy to fuel cars
for consumers and fleets, especially when it includes grid integrated technologies
such as managed charging, open source, and interoperability to smooth grid peaks.
Developing partnerships among grid planning departments to share smart charging
pilot results and leverage EV program data from around the world will improve grid
impacts with seamless managed charging programs that are invisible to the customer
and ensure both drivers and electric fleet owners have a positive experience and con-
tinue to grow the adoption of EVs.
https://www.researchandmarkets.com/reports/5023828/electric-vehicle-
charging-infrastructure-market?utm_source=dynamic&utm_medium=BW&utm_
code=v8g9wg&utm_campaign=1389168+-+Global+Electric+Vehicle+Charging+
Infrastructure+Market+(2019+to+2030)+-+Opportunity+and+Trend+Analysis&
utm_exec=jamu273bwd.
 2 Fog Computing
for Smart Grids
Challenges and Solutions
Linna Ruan
Beijing University of Posts and Telecommunications

Shaoyong Guo and Xuesong Qiu

Beijing University of Posts and Telecommunications

Rajkumar Buyya
The University of Melbourne

CONTENTS
2.1 Introduction.......................................................................................................8
2.2 SGs.....................................................................................................................9
2.2.1 Architecture...........................................................................................9
2.2.2 Current and Upcoming Problems........................................................ 12
2.3 Fog Computing-Driven SG Architecture........................................................ 15
2.3.1 Features................................................................................................ 15
2.3.2 Fog Computing Complements the Cloud............................................ 16
2.3.3 Fog Computing Helps Address SG Problems...................................... 17
2.4 Current Solutions for Applying Fog Computing to SGs.................................. 17
2.4.1 Fog-based SG Architecture................................................................. 17
2.4.2 Mainly Discussed Applications........................................................... 17
2.4.3 Key Problems Focused in Strategy Design.........................................20
2.4.4 Fog+..................................................................................................... 23
2.5 Research Challenges and Future Directions....................................................26
2.5.1 Security and Privacy............................................................................26
2.5.2 Huge Amounts of Data Processing......................................................26
2.5.3 Fog and Cloud Combination................................................................28
2.5.4 Fog Device Deployment......................................................................28
2.6 Summary and Conclusions..............................................................................28
References................................................................................................................. 29

7
8 Electric Vehicle Integration in a Smart Microgrid Environment

2.1 INTRODUCTION
In recent years, significant climate change, such as global warming and air quality
deterioration, threatens all the lives on Earth and attracts worldwide concern about
harmful gas emission as well as energy issues. Given this background, the traditional
power grids are transformed to smart grids (SGs) to enhance energy efficiency and
system reliability, providing a promising solution to address environmental prob-
lems. Following this trend, microgrids, as small-scale local power systems, are also
proposed to optimize energy management individually or through collaboration with
main grid. The two types of grids are implemented at the utility level and facility
level respectively, while both contribute to energy system and environment. This
chapter mainly focuses on SGs, the large-scale conception. Due to many commonali-
ties, most of the discussion also fits for microgrids.
SGs enable two-way communication and integrate renewable resources for
power generation, being used to support smart cities and other energy required
scenarios. Apart from these benefits, it is subject to some problems during imple-
mentation, mainly reflected in four aspects. (1) A huge amount of data generated
by SG devices requires robust processing capability. (2) The emerging delay-
sensitive applications propose instant response requirements. (3) Transmission of
all data over the uplink increases the burden on the communication channels. (4)
Uploading data to cloud through the open Internet increases the risk of privacy
violations.
The traditional mode of processing data in Cloud shows its limitations in this
background, mainly due to the limited transmission resources and long response
delay and, in particular, to the data privacy risk. Moving the processing of emer-
gency data to the edge side is regarded as an efficient way to address these prob-
lems, which is also the main intention of fog computing. Therefore, as one of the
advanced technologies and the vertical downward extension of cloud computing,
fog computing is discussed to be employed in SGs to enhance edge-side processing
capability, reduce response time, relieve the burden of core network, and protect
user privacy. During the application of fog computing, two problems are deputed
most. First, how to deal with the relationship between the two computing modes;
should fog computing replace or complement cloud computing. Second, fog com-
puting benefits SGs on multiple aspects, while some new problems also emerged,
how to cope with that.
In this chapter, we aim to conduct a comprehensive analysis of fog computing in
SGs. We begin with a brief introduction of SGs and fog computing, focusing on their
features, components, advantages, and challenges. Then, we discuss the application
of fog computing in SGs by analyzing the existing research and the current solutions,
so as to illustrate the application scenarios and summarize the frequently used meth-
ods. Further, we outline the challenges of fog-enabled SGs and the future research
directions. Finally, we conclude the chapter.
In general, we hope to clarify three problems through this chapter: Why is fog
computing suitable for SGs? What are the current solutions? And what challenges
may exist for future applications?
Fog Computing for Smart Grids 9

2.2 SGs
SGs are defined in various ways by different organizations. In the United States, SGs
are viewed as a large-scale solution to realize energy transformation from global net-
work to the localized. While in China, SGs are defined as an approach that ensures
energy supply based on physical network. For Europe, SGs mean a broader RE
(renewable energy)-based system with society participation and countries’ integra-
tion [1]. Although there are differences in the definition of SGs, consensus has been
reached on three aspects. (1) SGs are envisioned as the next-generation electrical
energy distribution network and an important part of smart cities. With the reliable
communication system, SGs can manage energy more intelligently and effectively;
(2) SGs allow two-way both electrical flow and information flow interaction between
demand side and supply side, which makes energy consumption and pricing strategy
easier to be monitored. In addition, supply-demand match, efficient energy utiliza-
tion, and energy cost reduction can be realized; (3) SGs allow devices to interact
information and are suitable for Internet of Things scenarios. In a nutshell, SGs inte-
grate advanced information and communication technologies into the physical power
system to

• enhance the level of system automation and hence contribute to operation

efficiency;
• improve system security and reliability;
• fit the requirements of sustainable development better by using cleaner elec-
tricity resources and storage devices;
• enhance energy efficiency by facilitating two-way information interaction;
• allow customers to monitor their energy consumption and schedule electric-
ity usage plan, which would benefit both themselves and the systems.

2.2.1 Architecture
A SG can be described from three perspectives. First, from the perspective of func-
tions, SGs can be viewed as the combination of physical infrastructures and infor-
mation technologies. Second, considering the core processes and participants, a SG
system consists of seven domains. Third, in view of the coverage scale, small-scale
microgrid as an important component of SGs has been hotly discussed in recent
years. The details of these three ways are introduced as below.

10 Electric Vehicle Integration in a Smart Microgrid Environment

• Bulk generation domain

Generate electricity for later transmission and distribution and finally
for residential, commercial, or industrial use. The generation sources
include traditional sources (such as fossil fuel and coal) and distributed
energy sources (such as solar and wind power).
• Transmission domain
It is commonly defined as the carrier for long distance power trans-
mission. In some specific scenarios, it also has the capability of electric-
ity storage and generation.
• Distribution domain
Distribute electricity to or from (when the surplus power generated
by distributed resources needs to be sent back to the market) customers.
Similar to power transmission, the distribution domain has the capabil-
ity of electricity storage and generation in some cases.
• Customers

FIGURE 2.1 Conceptual model of SGs defined by NIST.

Fog Computing for Smart Grids 11

They are the end users of electricity. According to the consumption

habits and levels, they are divided into three types: residential, commer-
cial, and industrial. Besides consuming electricity, customers may also
generate and store electricity by embedding distributed resource infra-
structures and batteries. In addition, demand-side management allows
them to monitor and manage their energy usage.
• Service providers
Organizations that provide services for electricity users and utilities.
• Operations
The managers of the electricity movement.
• Markets
A trading place for operators and customers. In power grid systems,
the markets are divided into wholesale markets and retail markets,
depending on the transaction mode.

Similar to SGs, microgrids are defined by different organizations. For example, the
Microgrid Exchange Group has defined microgrids as “a group of interconnected
loads and distributed energy resources within clearly defined electrical boundaries
that act as a single controllable entity with respect to the grid”, while the consensus
on microgrids lies in four aspects. (1) They are an important component of SGs [4].
(2) Compared with SGs, microgrids refer to a smaller distributed local power system
[4–5]. (3) Distributed generators and power-storage units are included. (4) They can
operate either in conjunction with the main grid (excessive power that cannot be
totally consumed locally can be sold to utilities through electricity market) or in an
isolated mode (only provide services for end customers, which differentiates it from
the centralized power generation form). Such obvious benefits listed below make
microgrids a hot topic for SG researchers.

FIGURE 2.2 A microgrid architecture.

12 Electric Vehicle Integration in a Smart Microgrid Environment

• Localized mode eases the integration of distributed and renewable energy

sources (such as solar and wind power), relieving the burden of generators in
peak load periods and therefore reducing harmful gas emission.
• Locating close to demand side makes microgrids easier to get the knowl-
edge of users’ needs, resulting in efficiency increase and transmission cost
reduction [6]. It provides better service by ensuring the energy supply of
critical loads (loads generated by devices or organizations, which require
continuous power supply, such as military equipment, hospitals, and data
centers [2]), system quality, resilience, and reliability and allows customers
to join demand-side management in an easier way.
• Provide strong support to the main grid by handling local urgent issues,
such as the variability of renewables and the sensitive loads, which
require local generation, and provide auxiliary services to the bulk power
system.
• Local-global controlling can be realized. With both regional require-
ments and overall performance taken into account, microgrids offer bet-
ter performance insurance and more opportunities for multi-technology
integration (electric vehicle, residential energy storage, rooftop photovol-
taic systems and smart flexible appliances [2]), which is beneficial for SG
development.

Many researchers have investigated the relationship between SGs and microgrids.
They are widely regarded as the implementation of new era grids at utility level and
facility level, respectively [4–5], [7]. Though with some differences in construction,
both contribute significantly to the energy system and environment.

2.2.2 current And upcoming problems

1. Latency requirements
Emerging mission-critical and delay-sensitive SG applications, such as
demand response, emergency restoration process, and substation monitor-
ing, require low round trip latency [8]. Based on characteristics, these appli-
cations can be categorized into two types, the flexible real-time ones and
the fault-tolerant but continuity-required ones. A typical application of the
first type is demand-side management, which allows customers to monitor
their electricity consumption in almost real time. However, cloud comput-
ing cannot meet the requirements, due to long distance data transmission,
possible channel congestions, and server failures. The second type cannot
be satisfied by cloud computing either, because possible connection failures
and processing delay can cause interruption, making it difficult for cloud to
support a continuity-required service.

Fog Computing for Smart Grids 13

• Device level—security measure upgrade

Resource-constrained devices. SGs have many resource-constrained
devices that hold limit capability to upgrade the security corresponding
hardware and software in their lifespans. Moreover, a SG environment
includes more participants, more technologies, and frequently informa-
tion interaction compared to traditional power grid, improving energy
efficiency while being more vulnerable to security attacks. Hence,
there is an urgent need to figure out an effective protection method for
resource-constrained devices [9].
Large number of devices. As mentioned above, resource constraints
pose a challenge to the security protection of devices. In some cases, con-
necting to cloud seems a proper way to upgrade the security credentials
and software. But with the exponentially growing number of devices, it is
impractical to allow all devices to do that, which is a resource-, energy-,
and time-consuming process. Therefore, in the face of a large number of
devices, how to ensure system security is still a big challenge.
• Communication and system level
Cope with security problems while ensuring operation. When faced
with security issues, shut down-then-fix is the common way currently.
However, shutdown also means interruption, which is intolerable to
mission-critical or delay-sensitive services. Like the example proposed
in Ref. [9], if an electric power generator chooses to shut down when
met malware attack, severe disruption will be caused, leading to power
outages. Therefore, fix-while-operate is a prospect mode and still a
challenge for SGs.
Keep private data private. In SGs, the collected data are usually
stored in cloud-based data centers for further processing or future use.
Since the private information contained in the dataset is valuable for
making energy strategy and then making benefits, it is preferred by
intruders, and even service providers or cloud operators. Therefore, it
is important to ensure transmission security during the way from smart
meters to the central cloud to prevent data leakage. However, the data
are transmitted through open Internet, and the number of connected
users continues to increase, making it harder to figure out a strong pri-
vacy protection solution [10].
Robustness. Communication network is one of the adding com-
ponents that differentiate SGs from traditional power grids. Hence,
its robustness directly impacts how a SG system would be judged. A
robustness communication network means that it can keep normal or
recover quickly even in such terrible situations, like natural disasters
or human intervention. In this case, we care about whether there are
advanced technologies that can be included to deal with emergencies
intelligently in addition to existing resource-consuming solutions (pro-
viding redundant links or power backup facilities).
Reliability. Reliability has always been viewed as a challenge for one
system especially for SGs due to the high outage cost. According to a
14 Electric Vehicle Integration in a Smart Microgrid Environment

Sun Microsystems analysis, blackouts cost approximately US$1 million

every minute to electric companies [8]. The main reasons leading to
outages lie in three aspects. (1) Lack of accurate knowledge of system
status in real time; (2) Lack of prediction and analysis capabilities; (3)
Lack of timely and effective response measures. SGs offer better com-
munication, autonomous control, and management methods to relieve
these problems. However, how to extend the current framework to han-
dle diverse and sophisticated issues in the future still seems a problem.

4. Prediction responsiveness
Prediction is the basis of making predecisions, and its accuracy directly
impacts whether a strategy proposed is appropriate. In a SG environment,
demand and generation prediction are studied most. Demand prediction is
divided into long-term prediction and short-term prediction according to
the time interval. It is important for demand-side management, while gen-
eration prediction is usually used for renewable energy resources, such as
solar panels and wind turbines. Its accuracy mainly depends on the weather
prediction, and it is an important component of microgrids. In the past sev-
eral years, cloud platform is the carrier for the two types of predictions.
However, instant decision-making is required for some specific applications
recently and nearly real-time prediction is expected, which cannot be satis-
fied by cloud computing. In this case, how to enhance prediction responsive-
ness while meeting the requirements of computing capability is a challenge.
5. Supply-demand match
Communication network enables information interaction between power
providers and consumers. It enhances energy efficiency by supporting system
balancing and provides incentives to customers to optimize their electricity
usage by cutting or shifting peak period demand. Supply-demand match
intends to realize less energy waste and higher energy efficiency, beneficial
for both the system and environment. However, the demand and renewable
resource generation is always changing dynamically, which requires real-
time information flow to support customer’s immediate participation. The
requirement cannot be satisfied by centralized cloud processing due to large
response delay. Given this background, how to implement real-time infor-
mation interaction between providers and consumers is a problem.

Fog Computing for Smart Grids 15

guaranteeing their requirements, and balancing the interests of all partici-

pants are really a challenge.

2.3 FOG COMPUTING-DRIVEN SG ARCHITECTURE

Fog computing was first proposed by Cisco as the vertical downward extension of
cloud computing. By providing computing, communication, controlling, and net-
work storage capability at the proximity of data source, fog computing contributes
to response time reduction and complements edge-side processing capability. It is
mainly used to handle mission-critical and delay-sensitive applications. From the
tech giants to manufactures, fog computing is discussed to be used in many sce-
narios, especially the SGs, which have such challenges mentioned above and view
fog computing as a proper technology to break the barriers. Before delving into the
application of fog computing in SGs, we give a brief introduction about the main
features of fog and discuss how to deal with the relationship between fog computing
and cloud computing.

2.3.1 FeAtures
The features of fog can be simplified as AESR (Awareness, Efficiency, Scalability,
Responsiveness) illustrated as follows. They also reflect the advantages of fog com-
puting on different aspects.

• Awareness
The awareness refers to two aspects: objective awareness and location
awareness. In a SG, users’ preferences are various, such as profit, quality
of experience (QoE), and energy efficiency. Since fog nodes are geographi-
cally distributed, each masters a relatively small area, making the nodes
easier to get users’ expectations and preferences and providing suitable and
even customized strategies. It is like the general saying “the right is the
best”, awareness makes fog computing a good service provider.
• Efficiency
In a broader perspective, fog computing is regarded as the added com-
puting nodes between the end devices and the cloud. Moreover, the capa-
bilities of fog computing are not limited to computing, communication, and
storage, and these basic functions make it also a good resource manager and
task scheduler. It integrates all the edge-side resources, such as smart appli-
ances and computers, and finds the best place for task processing with the
combination of resource scheduling. In this way, fog computing attains high
efficiency in terms of both resource and system operation.
• Scalability
Fog platform locates close to users and is small in size, making it easier
to adjust according to environment requirements, supporting infrastructures
update and scaling with less cost. In addition, fog permits even small groups
to access public programming interfaces and copes with new emerging ser-
vices well with good scalability.
16 Electric Vehicle Integration in a Smart Microgrid Environment

• Responsiveness
Quick response is one of the main advantages of fog computing and also
the motivation for its proposal. Fog platforms implement data processing
close to users, significantly shortening the transmission link, which makes
actuators obtain data analysis results and operation suggestions almost in
real time, meeting the requirements of mission-critical and delay-sensitive
applications. This is essential for not only the SG stable operation but also
for enabling millisecond reaction times of embedded AI to support emerg-
ing artificial applications, which is mentioned as one solution for applying
fog computing to SGs in the next section.

2.3.2 Fog computing complements the cloud

After introducing the features of fog computing, we should explain why it is still
proposed in the context of ‘cloud computing everything’ and then clarify what kind
of relationship exists between them. We will analyze from the following aspects.

• Latency
As mentioned before, geo-distributed fog computing enables quick
response due to locate proximity to end users. It offers users an opportu-
nity to obtain analysis results timely and then go on operation or cope with
urgent issues. As a comparison, centralized cloud computing is a time-
consuming process, caused by long distance data transmission both uplink
and downlink. Moreover, huge amounts of data transmitted to cloud put
great pressure on network channels, which may lead to congestion or even
interruption. In a word, fog computing can complement cloud on real-time
performance and reduce the possibility of channel congestions.
• Accessibility
Fog platforms locate at the edge side, enhancing the possibility for end
devices to get served, especially for resource-constrained ones. Besides, fog
computing costs less either on time or energy compared to cloud computing
and processes data locally, protecting the system from channel congestions.
In a conclusion, fog complements cloud on high accessibility, providing a
more general and affordable solution for devices and services.
• Privacy
Data privacy protection, which means the protection of sharing confi-
dential data with the third parties, is important for the reliability of a SG
system. Similarly, we analyze the data processing mode of fog and cloud
computing, so as to show their performance differences on data privacy.
Fog computing enables data to be processed separately, indicating that pri-
vate data can only be accessed by the fog while public data also can be
transmitted to the cloud. While cloud works on shared background [12] and
data is transmitted through Internet, each link has a risk of data leakage,
which endangers the safe and stable operation of SG systems. Therefore, fog
complements cloud by providing another location for data, protecting data
privacy while ensuring efficiency.
Fog Computing for Smart Grids 17

After the above analysis, it is easy to get a conclusion: cloud has powerful comput-
ing capability, while fog outperforms on latency, availability, and privacy. Now we
are able to answer the question “what relationship exists between cloud computing
and fog computing?”. An appropriate answer is that fog computing is a complement
of cloud computing. They have their own advantages, and no one can replace the
other. In a specific scenario, which paradigm to apply depends on the requirements
of services and pursuing of users. Therefore, combining the centralized cloud and the
distributed fog nodes to create a hybrid fog-cloud platform is the best way currently
to address SG problems.

2.3.3 Fog computing helps Address sg problems

After clarifying the relationship between fog and cloud, let us discuss one of the main
application scenarios—SGs. The specific solutions will be introduced in the next
section, while before that, we want to illustrate why fog computing is suitable for SGs
by comparing the earlier mentioned challenges faced by SGs (under traditional cloud
computing form) with the performance supplement that fog can provide. The analysis
can be summarized in Table 2.1.

2.4 CURRENT SOLUTIONS FOR APPLYING

FOG COMPUTING TO SGs
A lot of research has discussed how to strengthen SGs with fog computing. Since
edge computing is interchangeably defined as fog computing in most of the cases,
both of the two computing paradigms in SG environments are discussed in this
chapter and are uniformly called fog computing. In this section, we depict a generic
architecture for fog-enabled SGs (fog-SGs) and review the mainly discussed services,
key problems in strategy design, and other technologies that may provide further
performance improvement.

2.4.1 Fog-bAsed sg Architecture

A fog-based SG architecture is proposed in Figure 2.3. It contains three layers,
which are the infrastructure layer, constructed with residential, commercial, and
industrial buildings, acting as power demand side; the access layer, with fog and
cloud computing, providing computing, communication, and storage capabili-
ties. Fog servers are deployed with base stations and the supply layer, which is
mainly responsible for power generation, transmission, and distribution. The main
services supported by fog computing are also listed in the architecture, and the
details are shown as below.

2.4.2 mAinly discussed ApplicAtions

From the perspective of core links of SGs, we introduce how can fog computing
benefit the applications.
18 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 2.1
Fog Provides Effective Ways to Address Smart Grid Problems
Classifications Smart Grid Challenges How Fog Can Help
Latency constraints The flexible real-time applications Fog locates at the proximity of end
and the fault-tolerant but continuity- devices, strengthens edge computation,
required applications have real-time communication, controlling, and storage
response or continuous operation capabilities, provides delay-reduced
requirements, which cannot be services, avoids the risk of channel
satisfied by cloud. congestions, and ensures consistent
operation.
New security On device level, continuous Edge resource is empowered with fog and
challenges upgrading of security measures is able to support security infrastructure
hard to realize. upgrade.
On communication and system level, Fog computing provides service with
service continuity cannot be reduced delay, which can ensure the
guaranteed. Private data face the continuous operation. Private data are
risk of leakage. Robustness and processed at the edge; only public data
reliability are also not that satisfied. can be further transmitted to the cloud.
Distributed control The basic components of SGs are Fog computing follows distributed form
geo-distributed. Centralized cloud and is able to provide location-based
is high-cost and not that suitable. A services and analytics and location-free
distributed paradigm is preferred. billing and charging.
Prediction Demand and generation prediction Fog nodes have the computing capability
are preferred in future SGs. Cloud to do basic prediction and can send back
can meet the requirement of the results with short delay. It can catch
computation while fail to update the dynamic changes of information and
information in real time. update within latency limit.
Supply-demand Demand and renewable resource Fog computing interacts the demand and
match generation is always changing, pricing strategies between customers
while cloud cannot process this and providers timely, which facilitates
frequently changing status. the demand response process.
Complexity With the rapid increase of data Fog computing can undertake data
volume and rate, and the need to analysis, support delay-sensitive
support various technologies and services, and relieve the burden of end
services, the SG system becomes devices, network, and cloud. Distributed
more and more complex. mode means tasks can be split for
Complexity complicates the control processing, which decreases the
of SGs. complexity of SGs.

• Power supply
For power supply, fog computing is mainly used to set price strategy,
balance supply-demand, and identify abnormal fluctuations. Demand-
side management is one of the important applications in SGs, which has
a potential of cutting off or shifting electric demand in peak periods. In
this application, real-time pricing is the main motivation for customers,
and supply-demand balancing is one of the aims. Due to the suitable
Fog Computing for Smart Grids 19

FIGURE 2.3 An architecture of fog-cloud-enabled SGs.

computing and communication capabilities and providing delay-reduced

services, fog computing is envisioned as a well-suited technology to
facilitate demand-side management. Besides, in Ref. [13], fog nodes are
considered to be embedded in charging points and detect the abnormal.
It mentions that fog computing can identify the abnormal status by ana-
lyzing sudden power fluctuations and then report to cloud for further
management.
• Power transmission
For power transmission, line status monitoring is a widely mentioned
service that applies fog computing [13–15]. Transmission line monitoring is
important for obtaining full knowledge of the equipment condition (espe-
cially during bad weathers, such as high temperature, heavy rain, strong
wind, or snowstorm), supporting safe and stable operation of power sys-
tems. In this application, graph, video, and data information is collected by
unmanned aerial vehicles [13] or video sensors [15], which are controlled
by edge network, and then the information is sent to fog nodes to filter
and process. The analysis results reflect whether there is a possibility of
failure and effectively alleviate the bad effects as a result. In this process,
quick response, data privacy, and system reliability are really important,
which can be better satisfied by fog computing. Of course, if there is a large
amount of data waiting for processing or the status is too complicated for
fog, hybrid fog-cloud will be adopted, in which preprocessing and further
processing are considered to be executed, respectively.
20 Electric Vehicle Integration in a Smart Microgrid Environment

• Power distribution
For power distribution, fog computing is considered to be used in smart
low voltage (LV) district management and outside force (manual or natural)
risk monitoring [16]. For LV district management, the transformer terminal
unit is enhanced with fog computing, which is used as low voltage side
agent. LV topology identification, distribution fault diagnosis, and line loss
analysis therefore can be realized with less delay and have light pressure on
storage and processing. For monitoring, the role of fog is similar to when it
is used for power transmission. In traditional mode, the collected data are
stored in a local recorder and then sent to cloud for processing. By facilitat-
ing fog computing, lightweight data can be processed locally, and warning
information can be sent almost in real time.
• Substation
For substation, the function of fog computing is to monitor the opera-
tion status and equipment environment and analyze data, similar to the
power transmission process, while the difference is the source of data. In
this scenario, the data mainly come from a large number of various sen-
sors. With fog computing, most of the information and data processing
can be implemented locally and respond timely to ensure the warning
before events, the suppression during events, and the reviews after events
[13].
• Power consumption
For microgrid systems, fog computing is used to collect the information
of power generation and users’ electricity consumption in real time and then
abstract their behavior mode based on power information in the time dimen-
sion. During system operation, the behavior mode is used to judge the power
balance level and identify abnormal status [13].

For advanced metering systems, fog devices are embedded in power meter concen-
trator to support the storage and analysis of data as well as services. The enhanced
real-time interaction and price prediction capabilities empower the demand-side
management application.
For smart home, most smart appliances require initial processing of power
information. With time and cost-saving characteristics, fog computing would be
preferred to obtain better performance. In addition, fog nodes can act as user
agents, interacting information between users and cloud, providing local data col-
lection, operation status monitoring, and small-scale controlling functions, and
contributing to emerging applications, such as demand response and fault diag-
nosis [16]. Table 2.2 summarizes the mentioned SG applications supported by fog
computing.

2.4.3 Key problems Focused in strAtegy design

Specific problems that are often discussed in fog-SGs include resource management,
task scheduling, security and privacy protection, and comprehensive ones. The solu-
tions provided are illustrated as follows.
Fog Computing for Smart Grids 21

TABLE 2.2
SG Applications Supported by Fog Computing
Applications Functions Realized Benefits Provided References
Power supply Price strategy formulation Supports demand-side management. [13]
(demand prediction)
Supply-demand match Enhances energy efficiency.
Abnormal fluctuation Identifies abnormal situations and
identification sends warning information timely.
Enhances system reliability.
Power Transmission line Conceives the status of the equipment [13–15]
transmission monitoring timely.
Data collection and Ensures the warning before events,
analysis the suppression during events, and
the reviews after events. Supports
safe and stable operation of the
power system.
Power distribution Smart low voltage (LV) Fast topology identification, [16]
district management distribution fault diagnosis, and line
loss analysis.
Outside force (manual or Lightweight data localized
natural) risk monitoring processing. Finds abnormal
situations and sends warning
information timely.
Enhances system reliability.
Power substation Operation status and Gets the condition of the operation [13]
equipment environment status and equipment environment
monitoring timely.
Data collection and Ensures the warning before events,
analysis the suppression during events, and
the reviews after events.
Supports safe and stable operation of
the power system.
Power Power generation and Supports demand-side management. [13,16]
consumption users’ electricity
consumption
information collection.
Price prediction Helps users optimize electricity usage
plan.
Abstracts the behavior Judges the power balance level and
mode identifies abnormal status.

• Resource management
A residential scenario is considered in Ref. [17]. It proposes a cloud-fog-
based SG architecture, in which multiple buildings are formulated as end
user layer, and each building corresponds to one fog device. By applying
the cloud-fog mode, quick response can be realized. In addition, particle
22 Electric Vehicle Integration in a Smart Microgrid Environment

swarm optimization (PSO), round robin, and throttled algorithms are used
to implement electric load balancing. It is verified that PSO outperforms the
other two algorithms in terms of response time and total cost.
The work [18] focuses on a residential scenario and aims to optimize
energy consumption, which is important for demand-side management
application. A cloud-fog architecture is proposed, and each fog node man-
ages energy demand scheduling of several buildings. It aims to reduce total
energy cost and formulates the problem as a distributed cooperative demand
scheduling game.
• Task scheduling
A fog computing system is considered to provide strong storage and
computing resources in SG communication network. The work [19] aims
to minimize the total cost for the system running with subject to the tasks’
requirements. A green greedy algorithm is designed to provide a solution
for the optimization problem.
Another work [20] takes SG communication network into consideration
and designs a service caching and task offloading mechanism, to realize
network load balancing and accelerate response. A computing migration
model is also proposed to support task offloading from central cloud to the
edge network.
• Security and privacy protection
By facilitating information communication, benefits such as energy sav-
ing and customers’ satisfaction improvement can be realized. However, it
also means SG is more vulnerable to attacks, which can be summarized as
device attack, data attack, privacy attack, and network attack [21]. Therefore,
confidentiality, integrity, and availability (CIA), as the key judgments of a
security policy, should be highly guaranteed.
In Ref. [21], a fog computing-based strategy is proposed to detect
cybersecurity incidents in SGs. It models the anomaly detection process
on the basis of generally used Open-Fog Reference Architecture. Sensors,
actuators, smart meters, and a central monitoring unit are included in the
model, and fog is embedded in phasor measurement units to do device mea-
surement. The main process is data acquisition, data preparation, feature
extraction (using k means algorithm) and behavior modeling, anomaly
detection, and score provision. Hence, unusual power usage can be detected,
which is probable for an attack on a SG system.
For failure recovery, a fog computing-based dispatching model is pro-
posed in Ref. [22]. In that paper, each regional power grid is in charge
of a fog node, which is responsible for fault information storage, repair-
ing resource allocation, and making dispatching plan with short delay.
A dispatching algorithm based on genetic algorithm is carried out in fog
nodes, realizing cost reduction and satisfaction increase during the repair
process.
The work [23] presents a security fog-SG model, in which access authen-
tication, data security, and real time protection are described as expected
functions. Then, for physical layer authentication, it adopts k-nearest
Fog Computing for Smart Grids 23

neighbors as a solution. A differential privacy data distortion technique

consisting of Laplace and Gaussian mechanisms is provided in Ref. [13].
• Comprehensive problems
Due to the complex environment of SGs, some research also tried to figure
out solutions for comprehensive problems to balance multiple performance
metrics, such as security, efficiency, and functionality. Reference [24] builds
a fog computing-based SG model and proposes a concrete solution for both
aggregation communication and data availability. By encrypting under a
double trapdoor cryptosystem, security data aggregation is implemented.
The solution is designed for service providers to realize dynamic control
and electricity distribution. Similarly, a privacy-preserving authentication
and data aggregation scheme for fog-based SGs is proposed in Ref. [25].
In that scheme, short randomizable signature and blind signature are used
for anonymous authentication, and then fog nodes are used to solve billing
problems. Table 2.3 summarizes the mentioned problems and solutions.

2.4.4 Fog+
Fog complements cloud in multiple aspects, while for a fog-SGs system, it is not
enough. By integrating other advanced technologies, the performances are possible
to be further improved. We select three hot discussed technologies, which are block-
chain, AI, and SDN. They are expected to empower energy security and privacy
protection, prediction, and resource flexibility for Fog-SGs, respectively.

• Fog + Blockchain
As mentioned before, fog computing assists SG security on transmission
and service continuous operation. Nonetheless, the security and privacy
protection of fog-SG systems are still a challenge, since fog server itself
is less secure than cloud, and there exists interaction and service migra-
tion between heterogeneous fog nodes [26]. Hence, this problem has been
stressed in recent years, and some security solutions are proposed.
However, since SGs have many resource-constrained devices, most of
the conventional methods (certificate authority-based, ring signature, blind
signature, and group signature) are not suitable, due to bad traceability
and participation flexibility, high computation, and communication over-
head. Given this background, blockchain is envisioned as a new chance for
SGs. Blockchain allows network participants to record the system in a dis-
tributed shared ledger. In a blockchain enabled system, fog servers need
to join, while end uses are not required to, which prevents users’ identity
leakage. At the same time, the change in the participation status of one
user will not influence others, as the registration is identity-based. The
smart contract included also supports traceability and revocability [27].
Therefore, blockchain is chosen as fitting all the security requirements of
SGs, and its scalability requirements can be complemented by fog com-
puting. Moreover, both blockchain and fog computing follow decentralized
network form, which makes them possible to integrate.
24 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 2.3
Key Problems and Solutions
Classification Problems Objectives Solutions References
Resource Electric load Reduces response Builds a cloud-fog-based [17]
management balancing time and total SG architecture.
cost Proposes an algorithm
based on PSO.
Energy demand Reduces total Builds a cloud-fog-based [18]
scheduling energy cost SG architecture.
Formulates a distributed
cooperative demand
scheduling game.
Task scheduling Task scheduling Minimizes the Proposes a green greedy [19]
total cost for algorithm.
system running
Service caching Balances network Proposes a computing [20]
and task load and reduces migration model. A
offloading communication load-balancing algorithm
delay based on popularity and
centrality.
Security and Cybersecurity Detects anomalies Proposes an anomaly [21]
privacy incident and reduces detection process model
protection detection communication and a detection method
delay based on fog computing.
Failure recovery Reduces cost and Proposes a dispatching [22]
increases model based on fog
satisfaction computing and a
during the repair dispatching algorithm
process based on genetic
algorithm.
Accesses Realizes physical Proposes a secure fog-SG [23]
authentication, security and model, a physical security
data security, improves energy approach, and an
and real-time efficiency electricity forecasting
protection method.
Comprehensive Aggregation Realizes dynamic Proposes a fog computing- [24]
problems communication control and based SG model. A
and data electricity concrete solution for both
availability distribution aggregation
communication and data
availability.
Security and Anonymous Proposes a privacy- [25]
privacy issues in authentication preserving authentication
fog-based SG and billing and data aggregation
communication scheme for fog-based
SGs.
Fog Computing for Smart Grids 25

In Ref. [27], a blockchain-based mutual authentication and key agree-

ment protocol is proposed for fog-SG systems. In addition to providing
basic security properties, it also offers an efficient method for key update
and revocation as well as conditional identity anonymity with less costs.
From another perspective, [28] proposed a permissioned blockchain edge
model for SG network. It focuses on privacy protections and energy secu-
rity issues and formulates an optimal security-aware strategy by smart
contracts.
• Fog + AI
SGs require accurate prediction of demand, pricing, and generation
capability to support resource preparing or policy setting and contribute to
SG applications, such as demand-side response and monitoring. The basic
elements of prediction are huge amounts of data, computation capability,
and intelligent algorithms. Evaluation metrics are speed, accuracy, memory,
and energy [29]. The amount of data generated by the SG infrastructures
increases sharply both in terms of quantity and rate, which could be the first
element of prediction. Fog enhances edge-side computation capability and
can provide delay-reduced services and match the second element. Then,
for the third element, artificial intelligence algorithms seem to be the best
choice currently. The following two cases illustrate how AI is supported by
fog computing and how it is applied in SGs.
The work [16] focuses on distribution outside force damage monitor-
ing application. For the processing platform, the fast and accurate vehicles’
identification with AI is the most critical segment. Since the AI algorithm
is a resource-consuming process, the hybrid fog-cloud is envisioned as the
most suitable technology to enhance its operation performance and save
system resources.
Demand and dynamic pricing predictions are considered in Ref. [13].
Demand prediction is important to let providers understand the needs of
users, guide the generation plan, and contribute to electric resource balanc-
ing. Dynamic pricing acts as the main incentive for customers to respond to
demand shifting or cutting off in peak period and therefore directly relates
to users’ benefits. The suitable AI approaches are listed as Auto Regressive
Integrated Moving Average models, Auto Regressive models, ANN, fuzzy
logic, and (long short-term memory) LSTM in that paper.
• Fog + SDN
Fog computing offers SGs benefits on multiple aspects. However, for the
increasing scale of SG networks, how to transmit data to the fog servers or
sometimes to the cloud could bring significant effect on resource efficiency.
To face the challenge of data routing, involvement of software-defined net-
working (SDN) is viewed as a potential solution of fog-SGs, mainly due to
driving a more manageable and flexible network with a global view brought
by decoupling the network control plane and data plane. Reference [30] pro-
posed an SDN-based data forwarding scheme for fog-enabled SGs. With the
global information provided by SDN, the shortest path is calculated, and a
26 Electric Vehicle Integration in a Smart Microgrid Environment

path recovery mechanism is designed to avoid link failures. SDN is mainly

used to facilitate multicasting and routing schemes for SGs [30].

2.5 RESEARCH CHALLENGES AND FUTURE DIRECTIONS

2.5.1 security And privAcy
As mentioned before, the security of fog computing is still a challenge, also for fog-
SGs. The reasons are threefold: (1) heterogeneous. As the emerging of Internet and
SGs, fog computing will be combined with various technologies and access to other
systems. Interacting information and migrating services between heterogeneous
devices in large or small scales face the danger of malicious attacks. (2) Dynamism
and openness. Fog computing is a small edge-side computing platform and runs in
an open and dynamic environment, which is vulnerable to attacks. (3) Data disper-
sion. Since each fog node has limited capability, data are usually transmitted to geo-
distributed nodes to store or process. It brings a risk of data leakage, packet loss or
incorrect organization, and breaking data integrity.
In conclusion, security issues of fog-SGs, such as edge control, data dispose,
computation, and communication still need new ideas to match the distributed, het-
erogeneous, and complex environment. Blockchain is a probable method to address
identity risks, while the research is still at the initial stage. Scalable enhancement and
layered mechanisms are the research directions, and how to deal with outsourcing
services as well as off-chain status need further discussion.

2.5.2 huge Amounts oF dAtA processing

With the emerging of SG infrastructures, various system data increase both in rate
and amount, which brings high complexity and aggravates the burden for data pro-
cessing. The challenges brought by huge amounts of data are mainly reflected in the
following aspects.

• Data control, security, and privacy

The quantity, diversity, and rate of data in SGs increase
dramatically.  Based on broad consensus, fog computing strengthens the
computation, communication, controlling, and storage capability of edge
side and is envisioned as a good choice to cope with the data explosion of
SGs. However, a more accurate statement is that fog can relieve the prob-
lems, not solve. Due to the characteristics, fog computing platform means
a set of capacity limited nodes. Indeed, the nodes can interact to support
powerful data processing, but either for tech giants or manufacturers, it
is really challenging. We will analyze the reasons from different stages
of data processing. First, data split. When the resource of one fog node
is not sufficient to hold the data collected by sensors, advanced metering
infrastructure, or other devices, data will be split into several parts to pro-
cess. However, what split granularity should be selected. Large granularity
is beneficial for organization and data integrity, while small granularity
Fog Computing for Smart Grids 27

can use resources with high efficiency and is more likely to be processed
quickly. In this case, the choice seems dependent on what performance we
pursue. Is it easy? Don’t get the conclusion too early, let’s continue. Then,
data distribution. This stage contains three problems, which node to send,
how much to send, and which path to select. For the first problem, the node
with the richest available resources, or closest distance, or with the best
comprehensive performance in combination with the degree of credibility
is always selected as the destination. For the second problem, it is usually
decided along with the first one and depends on what we are pursuing for,
the system balance, the short delay, the energy efficiency, or anything else.
Now for the third problem, two choices usually are mentioned, the shortest
one or the most secure one. Finally, data storage. Where to store and how
much to store. Should the data be sent back to end device, retained at the
fog node, or sent to the cloud. We also have to take a decision.
Now, it is easier to reach a conclusion. During data processing, there
are several problems to be considered in each stage, and each problem has
several choices. It seems that these choices all depend on our goals, but the
goals of different stages may not be the same. To achieve the best compre-
hensive performance, we must make various considerations and trade-offs
between different performances with subject to the requirements of specific
services. In the existing research, the solutions proposed are usually based
on a specific application with such assumptions, which has a limited value
for a real industrial scenario. Therefore, for the data processing of fog-SGs,
how to deal with more and more complicated situations is very challenging
and needs more discussion.
• Artificial intelligence integration
Prediction is one of the important aspects to show the smartness of SGs.
As we discussed, the large amount of data collected from SG devices is
the basis of prediction, and the artificial intelligence (AI) algorithm is an
essential tool to guarantee the accuracy. However, since the AI algorithms
are resource-consuming, in fog-enabled SG environments, some aspects are
still challenging: (1) limited capability. Considering the complexity of pre-
diction, it is usually carried out in cloud traditionally. However, due to the
emerging applications and higher requirements proposed, real-time predic-
tion is expected in SGs. Fog computing can realize quick response for most
of the services, but it is not sure for prediction, depending on the complex-
ity of the algorithm. Therefore, the edge side still needs other technologies
to facilitate capability of fog-SGs. (2) Follow the dynamics of algorithms.
AI algorithms are not that stable, which means that they are more like an
art, designed for the specific problem, and innovation always comes out,
including more parameters, more layers, or new structures. Enabling fog-
SG systems keeping up with new AI designs is obviously another challenge.
28 Electric Vehicle Integration in a Smart Microgrid Environment

2.5.3 Fog And cloud combinAtion

The hybrid fog-cloud computing is envisioned as the most promising mode to be inte-
grated in SGs. But how to collaborate the two kinds of paradigms to show their own
best and also realize the complement is really a challenge. Resource management
problems are discussed most, which contain resource aggregation, task offloading,
caching, and storage. Due to virtualization technologies, the resource can be custom-
ized as containers or virtual machines, which enhances the resource flexibility but
makes it more complex to formulate resource management strategies. Besides, the
participation process of data and applications complicates the problem further. In
the future, how to make generic strategies and customized strategies for specific SG
scenarios with given optimization objectives still needs more discussion.

2.5.4 Fog device deployment

How to integrate fog computing into SGs has been discussed in recent years.
However, most of the solutions have such assumptions and not suitable for the real
status. Consider the deployment of fog computing; there are usually two choices. (1)
Embedding fog computing functions in the existing SG equipment. (2) Designing
new fog computing equipment. The first option costs less but requires manufacturers’
permission for the equipment transforming. It is necessary to ensure the performance
of the embedded part, especially the safety performance, while not bringing negative
impact on the existing performances. The second option costs higher, but it does not
need to negotiate with the device manufacturers but just needs to consider the access
with other devices. But both of the options require the support of the government and
power suppliers, which control device accessing and ensure the stable operation of
SGs. As discussed above, the deployment and realization of fog functions in a real
SG are still a challenge.

2.6 SUMMARY AND CONCLUSIONS

In this chapter, we investigated the application of fog computing in SGs from
three perspectives: characteristics, solutions, and challenges. Through the brief
introduction to the components and features of SGs and fog computing, two prob-
lems are analyzed. (1) Why fog computing is still proposed in the context of ‘cloud
computing everything’ and how to deal with the relationship between fog and
cloud computing? A proper answer is: fog computing is proposed as a comple-
ment of cloud computing. They have their own advantages, and no one can replace
the other. Under a specific scenario, which paradigm to apply depends on the
constraints of devices, requirements of services, and users. The hybrid fog-cloud
computing is a generic and the most promising mode to be integrated in future
SGs. (2) Why fog computing is suitable for SG systems? The answer is: the chal-
lenges faced by SGs (under traditional cloud computing form), such as latency
constraints, new security issues, distributed control, prediction, demand-supply
match, and complexity problems, all can be relieved with the performance supple-
ment that fog provides.
Fog Computing for Smart Grids 29

Through the analysis of the existing research, we identified the state-of-the-art

of fog-SGs. A generic architecture is proposed, and the main applications of fog
computing are discussed from the perspective of core links of SGs. In summary,
fog computing is mainly used for delay-sensitive or mission-critical applications, to
support prediction, monitoring, and information interaction. These functions are the
basis of microgrids, demand-side management, and communication network, which
are the emerging areas in SGs.
From the perspective of theoretical research, hot issues in policy design are also
introduced. They are divided into resource-related, task-related, security-related, and
comprehensive issues, which usually are summed up as an optimization problem with
such constraints. The optimization objectives mainly include delay, energy consump-
tion, satisfaction, system balancing, resource efficiency, safety, credibility, etc. Since
applications in SGs usually involve multiple participants and their benefits need to be
balanced, game theories are commonly used to solve these problems. In addition, opti-
mization algorithms such as genetic evolution algorithms are widely used as solutions.
In response to the evolvement of industrial applications and technologies, we also
proposed research challenges and future directions for fog-enabled SGs from the
aspects of security and privacy, huge amounts of data processing, fog and cloud com-
bination, and fog device deployment.
In summary, this chapter has analyzed the advantages and potential value of
applying fog computing in SGs. Considering many commonalities, many concepts
presented in this chapter also apply to microgrids. Obviously, fog computing acts
as a strong tool for designing optimized SG systems that can fulfill the emerging
requirements of applications with its outstanding “compute, communicate, storage,
and control” capability.

REFERENCES

1. X. Yu, C. Cecati, T. Dillon and M. G. Simões, “The new frontier of smart grids,” IEEE
Industrial Electronics Magazine, vol. 5, no. 3, pp. 49–63, Sept. 2011.

3. M. Krarti, “Utility Rate Structures and Grid Integration,” Optimal Design and Retrofit
of Energy Efficient Buildings, Communities, and Urban Centers, 2018, pp. 189–245.

6. F. Jalali, A. Vishwanath, J. de Hoog and F. Suits, “Interconnecting Fog computing and

microgrids for greening IoT,” IEEE Innovative Smart Grid Technologies - Asia (ISGT-
Asia), Melbourne, 2016, pp. 693–698.
7. S. Parhizi, H. Lotfi, A. Khodaei and S. Bahramirad, “State of the art in research on
microgrids: A review,” IEEE Access, vol. 3, pp. 890–925, 2015.
8. V. C. Gungor et al., “Smart grid and smart homes: Key players and pilot projects,” IEEE
Industrial Electronics Magazine, vol. 6, no. 4, pp. 18–34, Dec. 2012.
30 Electric Vehicle Integration in a Smart Microgrid Environment

9. M. Chiang and T. Zhang, “Fog and IoT: An overview of research opportunities,” IEEE
Internet of Things Journal, vol. 3, no. 6, pp. 854–864, Dec. 2016.
10. F. Y. Okay and S. Ozdemir, “A fog computing based smart grid model,” International
Symposium on Networks, Computers and Communications (ISNCC), Yasmine
Hammamet, 2016, pp. 1–6.
11. D. N. Palanichamy and K. I. Wong, “Fog computing for smart grid development and
implementation,” IEEE International Conference on Intelligent Techniques in Control,
Optimization and Signal Processing (INCOS), Tamilnadu, 2019, pp. 1–6.

13. S. Chen et al., “Internet of things based smart grids supported by intelligent edge com-
puting,” IEEE Access, vol. 7, pp. 74089–74102, 2019.
14. Y. Huang, Y. Lu, F. Wang, X. Fan, J. Liu and V. C. M. Leung, “An edge computing
framework for real-time monitoring in smart grid,” IEEE International Conference on
Industrial Internet (ICII), Seattle, WA, 2018, pp. 99–108.
15. Y. Zhang, K. Liang, S. Zhang and Y. He, “Applications of edge computing in PIoT,”
IEEE Conference on Energy Internet and Energy System Integration (EI2), Beijing,
2017, pp. 1–4.

18. S. Chouikhi, L. Merghem-Boulahia and M. Esseghir, “A fog computing architec-

ture for energy demand scheduling in smart grid,” 15th International Wireless
Communications  & Mobile Computing Conference (IWCMC), Tangier, 2019, pp.
1815–1821.
19. J. Yao, Z. Li, Y. Li, J. Bai, J. Wang and P. Lin, “Cost-efficient tasks scheduling for
smart grid communication network with edge computing system,” 15th International
Wireless Communications & Mobile Computing Conference (IWCMC), Tangier, 2019,
pp. 272–277.

21. R. El-Awadi, A. Fernández-Vilas and R. P. Díaz Redondo, “Fog computing solution for
distributed anomaly detection in smart grids,” International Conference on Wireless
and Mobile Computing, Networking and Communications (WiMob), Barcelona, 2019,
pp. 348–353.

23. A. Xu et al., “Efficiency and security for edge computing assisted smart grids,” IEEE
Globecom Workshops (GC Wkshps), Waikoloa, HI, 2019, pp. 1–5.

25. L. Zhu et al., “Privacy-preserving authentication and data aggregation for fog-based
smart grid,” IEEE Communications Magazine, vol. 57, no. 6, pp. 80–85, Jun. 2019.
Fog Computing for Smart Grids 31

26. R. Yang, F. R. Yu, P. Si, Z. Yang and Y. Zhang, “Integrated blockchain and edge com-
puting systems: A survey, some research issues and challenges,” IEEE Communications
Surveys & Tutorials, vol. 21, no. 2, pp. 1508–1532, Secondquarter 2019.
27. J. Wang, L. Wu, K. R. Choo and D. He, “Blockchain-based anonymous authentica-
tion with key management for smart grid edge computing infrastructure,” IEEE
Transactions on Industrial Informatics, vol. 16, no. 3, pp. 1984–1992, Mar. 2020.

30. F. Y. Okay, S. Ozdemir and M. Demirci, “SDN-based data forwarding in fog-enabled

smart grids,” 1st Global Power, Energy and Communication Conference (GPECOM),
Nevsehir, 2019, pp. 62–67.
 3 Opportunities and
Challenges in Electric
Vehicle Fleet Charging
Management
Chu Sun
McGill University

Syed Qaseem Ali

Opal-RT Technologies

Geza Joos
McGill University

CONTENTS
3.1 Introduction ....................................................................................................34
3.2 EV Chargers ................................................................................................... 35
3.2.1 Interfaces and Standards..................................................................... 37
3.2.2 Features and Topologies .....................................................................40
3.2.3 Controls...............................................................................................40
3.2.4 Capabilities ......................................................................................... 41
3.3 EV Aggregation .............................................................................................. 42
3.4 Available Ancillary Grid Services with Aggregated EVs .............................. 45
3.4.1 Frequency Response and Regulation .................................................. 45
3.4.2 Power Smoothing ................................................................................ 47
3.4.3 Load/Generation Following ................................................................ 49
3.4.4 Spinning Reserve ................................................................................ 50
3.4.5 Reactive Power Support ...................................................................... 51
3.4.6 Voltage Support .................................................................................. 51
3.4.7 Discussion ........................................................................................... 53
3.5 Case Studies .................................................................................................... 53
3.5.1 Frequency Regulation ......................................................................... 53
3.5.2 Power Smoothing ................................................................................ 55
3.5.3 Load/Generation Following ................................................................ 55
3.5.4 Spinning Reserve ................................................................................ 55
3.5.5 Voltage and Reactive Power Support .................................................. 56

33
34 Electric Vehicle Integration in a Smart Microgrid Environment

3.6 Challenges and Future Research Directions................................................... 56

3.6.1 Technology Initiatives......................................................................... 56
3.6.2 Economical Aspects ........................................................................... 58
3.6.3 Environmental Aspects....................................................................... 59
3.6.4 Safety and Security ............................................................................. 59
3.6.5 Future Directions ................................................................................60
3.7 Conclusion ...................................................................................................... 61
References ................................................................................................................ 61

3.1 INTRODUCTION
Sales of plugin electric vehicles (EVs) have been continuously increasing around the
world over the last few years. Numerous manufacturers, new and old, are getting into
the EV market [1,2]. The increased number of EV chargers is expected to impact the
grid in terms of load increase, requiring infrastructure upgrade and capacity expan-
sion [3]. Due to the schedulable nature of the EV loads, they can be leveraged to the
benefit of the grid by proper coordination [4]. Since the batteries in the EVs can serve
for energy storage as well, they can be used to provide vehicle-to-grid (V2G) services
[5]. They can also be used in conjunction with various renewable energy resources
to increase reliability and hosting capacity of the power system while reducing the
effects of load increase on the grid [6]. Since individual vehicle charger interfaces
and batteries are not capable of making a significant impact on providing these ser-
vices, an aggregating entity is necessary to assemble enough power to be considered
as a player in the market [7]. Some of the V2G services have been shown to provide
economic benefits to the vehicle owners as well as the aggregating entities [8] and
increased reliability and stability for the system operator [9]. The economic benefits,
however, can be significantly reduced by battery degradation due to cycling [10].
The economic impact is highly dependent on the market operation in which the EVs
participate and depends on the coordination and scheduling framework. The technol-
ogy and infrastructure required are common to all V2G service platforms. Several
aspects of the V2G technology are considered. Various battery charger topologies are
reviewed in Ref. [11–13], and the impact of V2G technologies on the distribution grid
are considered in Ref. [14,15]. A review of the various frameworks of the V2G and
their extension to vehicle to home and vehicle to vehicle are presented in Ref. [16].
This chapter provides a review of charger topologies and their associated control
principles along with the functions that need to be implemented by the EV and char-
ger manufacturers to provide V2G services. It also reviews the literature on V2G
service frameworks and technologies proposed for the vehicles and the associated
control functions that need to be implemented in the chargers by their manufacturers
to enable such use. Challenges that impede a wider acceptance and implementation
of V2G are also presented. This review will help the reader to understand the rela-
tions between the various power electronic interfaces of EV battery chargers and
their capabilities in providing V2G services.
The chapter is organized as follows: Section 3.2 gives a brief overview and clas-
sification of the typical topologies of EV battery chargers; Section 3.3 introduces the
idea of EV aggregation, while Section 3.4 presents the ancillary services that can
EV Fleet Charging Management 35

be provided using EV aggregation. Section 3.5 presents case studies of EV aggrega-

tion providing ancillary services from experiences around the world. Various chal-
lenges and future trends of EV fleet aggregation are presented in Section 3.6 with
Section 3.7 concluding the chapter.

3.2 EV CHARGERS

Chargers are typically classified according to their interfaces and power levels as
shown in Table 3.1. The table also shows the types of power interfaces for each level
and their placement, i.e., on-board or off-board. The charging time of a given vehicle
battery varies with the power of the charger used, e.g., high-power chargers will
result in reduced charging times.
Many power electronic interfaces for EV chargers have been proposed in the lit-
erature. They can be broadly classified into chargers that transfer power to the vehicle
by induction [17] or by conduction [18]. Inductive power transfer (IPT) chargers are
in a development stage and require wider deployment and greater maturity to com-
pete with the conductive charging systems. Conductive chargers, however, are widely
deployed with various features and capacities.
Chargers can completely reside on-board of the EV with only the power supply
outside, it can partially reside on and off-board of the EV, or it can completely
reside off-board of the vehicle with a direct DC interface to the battery. The
equipment that resides off-board of the vehicle is called the EV supply equipment
(EVSE). The EVSE can be connected to the battery via several paths including
directly, via a DC/DC converter, via a dedicated charger, or via another com-
ponent or set of components on the electric power train for the EV as shown in
Figure 3.1.
The batteries used in ful EVs are rated at 400–600 V. Since the battery voltage
may vary by up to 150 V from fully charged to deeply discharged, an additional DC/
DC converter can be used to keep the output DC voltage constant. This output volt-
age creates the high voltage (HV) bus. The traction motor is then connected to this
bus via a DC/AC inverter. The DC/DC converter connected to the battery and the
traction inverter is designed to handle bidirectional power flow. Power flows in the
opposite direction, from the motor to the battery, during regenerative braking and
charges the battery. Another DC/DC converter then steps down the HV bus volt-
age to the low voltage (LV) bus level for the auxiliary loads of the vehicle including
air conditioning, power steering, and lighting. The LV bus voltage is conventionally
selected to be between 12 and 48 V for safety reasons [19]. The charger output can be
connected directly to the battery terminals or to the HV bus having an input that is
single phase AC, three-phase AC, or DC depending on the power level and the design
of the charger.
The charging system of the EV electric power train shown in Figure 3.1 requires inde-
pendent power electronic converters for traction drive and charger, which increases the
overall volume, weight, and cost. To address this, an integrated charging system is pro-
posed, which uses the winding of the traction motor as filter inductor for the EV charger,
while the same power electronic interface can be employed for traction [12]. The same
principle can be applied to the integrated EV charger with six-phase permanent magnet
36 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 3.1
Charger Classification
Charging Charging Vehicle Charger
Level Power Interface Placement Power Interface
AC Level 1 1.4 kW (12 A) Single phase 120 On-board Any single phase for
1.92kW (16 A) VAC opportunity charging
– residential
AC Level 2 ≤19.2 kW (80 A) Single phase On or off-board Dedicated EVSE –
208–240 VAC commercial and residential
AC Level 3 >19.2 kW (80 A) Three phase On or off-board Any three-phase outlet or
208–480 VAC dedicated EVSE –
commercial and industrial
DC Level 1 50 kW (80 A) 200–500 VDC Off-board Dedicated EVSE –
commercial and industrial
DC Level 2 100 kW (200 A) 200–500 VDC Off-board Dedicated EVSE –
commercial and industrial
DC Level 3 >100 kW (200 A) 200–500 VDC Off-board Dedicated EVSE –
commercial and industrial

Traction

Regen. Braking
Battery DC/DC DC/AC
HV τ

M ω

LV

Charger Aux.
Charging

To aux loads
V2G

(AC, steering etc. )

Grid Essential
Optional

FIGURE 3.1 A typical electric power train of an EV.

synchronous motor drive. However, the torque generated during charging must be can-
celed to avoid motor rotation during charging [20].
The following subsections discuss further possible classifications of chargers
according to the typical standards applicable to EV chargers along with their inter-
faces, their topologies, their controls, and finally according to their capabilities in
terms of sourcing and sinking active and reactive power.
EV Fleet Charging Management 37

3.2.1 interFAces And stAndArds

The input voltage and the power ratings for various chargers are defined as differ-
ent levels by the SAE J1772 standard (for North America area) [18]. The classifica-
tions are summarized in Table 3.1. The table also presents the operating voltage
levels, the charger placement, and the conventionally used interfaces. Injected
harmonics and other power quality requirements for systems connected to the grid
are standardized by the IEEE and IEC [21–23] and by the SAE specifically for
EV chargers in J2894-1 [24]. Vehicles operating in the V2G mode are viewed as
a distributed resource (storage), for which the IEEE 1547 standard [25] could be
applied. Electromagnetic emission limits are defined by IEEE in Ref. [26], and the
limits for electromagnetic immunity are defined by the SAE in Ref. [27]. Galvanic
isolation from the grid is a preferred option to meet the safety requirements of the
EVSE handler. However, for high-power chargers using transformers to isolate
from the grid increases the weight, size, and cost of the charging system and can
be avoided by using residual current monitoring devices and disconnecting the
system if the residual current increases beyond the limits set by various countries
(5–30 mA).
The standards set a minimum charging efficiency of 90%, the charging THD to
10%, the discharging THD to 5%, and the minimum power factor while charging to
0.95. Individual limits for various harmonic groups are also defined in applicable
standards. Various other standards exist for charge couplers and communications.
They are outside the scope of this review.
The commonly used interfaces and connectors for the chargers are shown in
Table 3.2, along with their diagram, ratings, regions where they are most commonly
used, and the standards that define them. The first column of the table shows the
interfaces that the chargers are plugged into. There are more variants of such inter-
faces, but only the most common ones are listed.
The standards related to V2G services involve grid integration requirement,
EV connector, EV charging system, and communication links. The standards of
EV charging connectors are summarized in Table 3.2. The integration of EVs and
charging stations as distributed energy resources (DERs) with utility grid is speci-
fied in IEEE Std.1547-2018 [25]. For EV charging systems, four charging modes
are defined, and their major parameters are summarized in Table 3.3 [28,29]. The
relevant standards for each mode in different regions of the world are listed in
Table 3.4 [30], ranging from charging station and electromagnetic compatibility
(EMC) to the protection system. For the communication between EV and EVSE,
generally three standards can be found [31]: (1) ISO 15118, which is based on
power line communication (PLC) and applies to AC or DC charger. It lacks suf-
ficient security measures and suffers from poor response times (60 seconds). (2)
SAE J2847, also using PLC and suitable for AC or DC charging application. The
communication is based on the SEP 2.0 protocol and has small response time (less
than 1 seconds), but it is mainly designed for aggregation with nonvehicle smart
home technologies. The use cases of SAE J2847 standards are defined in SAE
J2836 series standards. (3) CHAdeMO, which is based on controller area network
and applies to DC charging only. Its response time is also less than 1 s, but there is
38 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 3.2
Charging Interfaces for EVs
Connector Supported
Plug Type Name Ratings Regions Charging Type Standards
CEE 7/7, 7/16 110–240 VAC All over 1φ AC: CEE 7;
NEMA 1-15, 5-15 2.5–15 AAC the world Level 1 & 2 NEMA WD-6;
Type 1 IEC TR
Type G 60083:2015
SAE J1772-2009 120–240 VAC North 1φ AC: SAE J1772
16–80 AAC America; Level 1 & 2
Japan
SAE J1772 CCS 120–240 VAC North 1φ AC: SAE J1772
Type 1 16–80 AAC America; Level 1 & 2;
200–600 VDC Japan DC Level 3
200 ADC
Mennekes Type 2 250–400 VAC Europe 1φ & 3φ AC: IEC 62196
63 AAC Level 1, 2 & 3

EU DC CCS 250–400 VAC Europe 1φ & 3φ AC: IEC 62196-3

Combo Type 2 63 AAC Level 1, 2 & 3
200–850 VDC DC Level 3
65–200 ADC
Chademo Yazaki 500 VDC Europe, DC Level 1 & 2 IEEE
120 ADC Japan, 2030.1.1TM
and North 2015
America
Tesla 110–250 VAC Europe and 1φ AC: Level 1, Tesla proprietary
12–80 AAC North 2
America
GB/T 20234.3 750/1,000 VDC China DC: GB/T
80–250 ADC Level 1, 2 & 3 20234.3-2015

no vehicle identification or guaranteed security, and it only works with CHAdeMO

chargers (high-capacity DC). For V2G services to be available, communication
should also be established among EV, EVSE, and the grid side. The information
model of EV or EVSE can be defined based on common information model stan-
dards such as IEC 61850, IEC 61970, IEC 61968, and IEC 62325. The relevant
communication protocols include Open Charge Point Protocol (OCPP), Hubject’s
Open Interchange Protocol (OICP), and other proprietary protocols. Some meter-
ing communication standards on the EV side and EVSE side are also defined. A
summary of the communication standards is given in Table 3.5 [32–36].
EV Fleet Charging Management 39

TABLE 3.3
Different Charging Modes and Their Major Parameters
Mode Maximum Current and Voltage Description
Mode 1 16 A and 250 V (AC, 1-phase) Conductive connection between EV and a standard
16 A and 480 V (AC, 3-phase) socket-outlet of an AC supply network without
communication or additional safety features
Mode 2 32 A and 250 V (AC, 1-phase) Conductive connection between EV and a standard
32 A and 480 V (AC, 3-phase) socket-outlet of an AC supply network with
communication and additional protection features
(in the cable)
Mode 3 (1): 32 A and 250 V (AC, 1-phase) Conductive connection between EV and an AC
(2): 70 A and 250 V (AC, 1-phase) EVSE permanently connected to an AC supply
63 A and 480 V (AC, 3-phase) network with communication and additional
(3): 16/32 A and 250 V (AC, 1-phase) protection features (permanent on the wall).
63 A and 480 V (AC, 3-phase)
Mode 4 (AA): 200 A and 600 V (DC) Conductive connection between EV and an AC or
(BB): 250 A and 600 V (DC) DC supply network utilizing a DC EVSE, with
(EE): 200 A and 600 V (DC) (high-level) communication and additional safety
(FF): 200 A and 1,000 V (DC) features (permanent in the installation)

TABLE 3.4
Standards of EV Charging System in Different Regions
Mode USA/Canada EU China Japan
Mode 1 Not allowed CE; LVD 2014/35/EU Not allowed Restricted
(In some countries),
EN 61851-1, EN
61851-22
Mode 2 NRTL; UL 2594, CE; LVD 2014/35/EU, GB/T PSE; IEC 61851,
UL 2231-1, EMC 2014/30/EU, 18487.1-2015, JIS C 8221,
UL 2231-2, UL 991, UL 1998, EN 61851-1, GB JARI A 0101
CSA C22.2 No. 280, CSA EN 61851-22, 22794-2008,
C22.2 No. 281-1, CSA C22.2 EN 62752 NB/T
No. 281.2, CSA C22.2 No. 0.8 42077-2016
Mode 3 NRTL; UL 2594, CE; LVD 2014/35/EU, GB/T PSE; IEC 61851,
UL 2231-1, EMC 2014/30/EU, 18487.1-2015, JIS C 8221,
UL 2231-2, UL 991, UL 1998, RED 2014/53/EU; EN NB/T 33002, JARI A 0101
CSA C22.2 No. 280, CSA 61851-1, EN NB/T 33008.2
C22.2 No. 281-1, CSA C22.2 61851-22
No. 281.2, CSA C22.2 No. 0.8
Mode 4 NRTL; UL 2202, UL 2231-1, CE; LVD 2014/35/EU, GB/T CHAdeMO; JIS D
UL 2231-2, UL 991, UL 1998, EMC 2014/30/EU, 18487.1-2015, 61851-23:2014,
CSA-C22.2 No. 107.1, CSA RED 2014/53/EU; EN NB/T 33001, JIS D 61851-
C22.2 No. 281-1, 61851-1, EN NB/T 33008.1 24:2014, JIS TS
CSA C22.2 No. 281.2, CSA 61851-23, EN D 0007
C22.2 No. 0.8 61851-24
40 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 3.5
Communication Standards for V2G Services
EV-EVSE EVSE-Grid
Metering Communication: ANSI C12.22 and IEEE Automation of DER: IEC 61850, Distributed
1703 Network Protocol (DNP3), and IEEE 1815
PEV as a DER: SAE J2847/3 and ISO 15118 Metering Model: ANSI C12.19, IEEE 1377
AC PEV smart charging standards: SAE J2847/1 and Multiple information models: CIM IEC 61970,
ISO 15118, DIN SPEC 70121:2014, IEEE 2030.5. IEC 61968, IEC 62325, and IEC 61850 data
DC PEV charging standards: SAE J2847/2, ISO models
15118, DIN SPEC 70122:2018, IEC 61851-24, and Communication protocols exist for charging
CHAdeMO (2030.1.1), GB/T 27930-2015 station networks: OCPP, eMI3, OICP and
Wireless PEV charging standard: SAE J2847/6 proprietary protocols such as ChargePoint,
PEVs and customers: SAE J2847/5 Blink Network, SemaConnect, etc.

3.2.2 FeAtures And topologies

The chargers can be classified as unidirectional and bidirectional on the basis of the
power flow, as single phase and three-phase chargers on the basis of the number of
phases used, and as isolated and nonisolated chargers on the basis of the galvanic
isolation provided. All the classification choices impact their capability to participate
in V2G schemes and services. Each type is shown in Figure 3.2a–e. Typical single
and two-stage unidirectional battery chargers with power factor correction (PFC) are
shown in Figure 3.2a and b, respectively. The dashed line represents the third phase
for three-phase chargers. Chargers can also be categorized as being single-stage or
dual-stage based on their topology. The operations performed by the charger must
meet the grid interface standards on the AC input side and control the DC current
and/or voltage at the output to charge the battery.

3.2.3 controls
The chargers shown in Figure 3.1 first have a PFC control loop that ensures the
current drawn from the grid is at a controlled power factor, which, for unidirec-
tional chargers, should be unity (or within acceptable limits as defined by the
standards). There are typically two control objectives for any battery charger: (1)
control the input current to be sinusoidal in shape preferably in phase with the
input voltage and (2) control the charging current or output voltage as required
by the charging algorithm. Single stage chargers perform both functions together,
while dual stage chargers use the first stage to perform the first function while
maintaining the DC link, and the second stage performs the second function inde-
pendently. Typical control loops used for these functions differ for unidirectional
and bidirectional chargers and are shown in Figure 3.3a–d. In the figures, the
block labeled “Mod” represents the modulator with pulse width modulation being
the most common modulator used.
EV Fleet Charging Management 41

Ig PFC VDC
Ib
Iin
Vg EMI Vb

(a)

Ig PFC VDC DC/DC Ib

Iin
Vg EMI Vb

(b)
Ig AFE VDC I Ig AFE VDC DC/DC
b Ib

Vg EMI Vb Vg EMI Vb

(c) (d)
DC/AC HFT AC/DC

(e)

FIGURE 3.2 Battery charger topologies: (a) unidirectional single stage; (b) unidirectional
dual stage; (c) bidirectional single stage; (d) bidirectional dual stage; and (e) isolated DC/DC
converter.

3.2.4 cApAbilities
The power electronic interfaces used in chargers can operate only in certain quad-
rants of the power plane. As per their capabilities, the chargers can be divided into
the following three types (defined as Type I, II, and III for the sake of this chapter)
as shown in Figure 3.4.

1. Type I Chargers
Chargers that are fully bidirectional are either single stage with an active
front end (AFE) or dual stage with a bidirectional DC/DC stage cascaded
with an AFE. In theory, they can provide ±1 pu reactive power and ±1
pu active power independently. When providing both active and reactive
power, the apparent power is limited to the charger kVA rating, Smax. The
operating area of Type I chargers in the power plane is shown in Figure 3.4a.
2. Type II Chargers
Unidirectional chargers with an AFE can provide ±1 pu reactive power,
but since they cannot handle reverse power flow, the absorbed active power
can only be varied from 0 to 1 pu. The operating point must remain within
the kVA rating, the circle defined by Smax. The operating area of Type I char-
gers is shown in Figure 3.4b.
42 Electric Vehicle Integration in a Smart Microgrid Environment

inref
g g To
I
V |1/V | x in Inner PFC
Mod
DCref I Control Gate
V Outer Drive
DC
V Control

(a)
PLL

Phase θg
Ibref Vg Filter Oscillator
DC To DC/DC Detect
Mod
Ib Control Gate Drive

(b) (c)

Vd*
I d* Current dq To Gate
- Mod
I q* Control abc Drive
- Vq*
θg
Id
dq Iag
Iq Ibg
abc Icg

(d)

FIGURE 3.3 Battery charger control loops: (a) front end control; (b) DC/DC converter con-
trol; (c) angle detection phase locked loop (PLL); and (d) d-q control for front end three phase
converters.

I II III
Q Smax Q Smax Q

Pmax

P P

(a) (b) (c )

FIGURE 3.4 Charger active and reactive power capabilities for (a) bidirectional charger;
(b) unidirectional charger with AFE; and (c) unidirectional charger with passive rectifier.

3. Type III Chargers

Unidirectional chargers that have a passive rectifier with a PFC preregu-
lator on the front end will only be able to absorb 0–1 pu active power and are
only capable of operating at unity power factor, i.e., Q = 0 even if the second
stage is a bi-directional DC/DC converter. The operating area of type III
chargers is shown in Figure 3.4c.

3.3 EV
 AGGREGATION
Since EV chargers are directly interfaced with the batteries of EVs, they can be used
to provide certain grid services, as long as they are capable of receiving external
power setpoints in addition to the setpoint required to charge the EV battery. To be
EV Fleet Charging Management 43

able to provide ancillary services to the grid, for example, to contribute to the grid
primary frequency response, the available power should be at least 1 MW [37] (the
number may be different depending upon the strength of the grid at the point of con-
nection). A 1 MW interface is not realizable with a single charger; however, an entity
responsible for aggregation of multiple chargers can negotiate a contract with the
utility and provide a significant (e.g. 1 MW) controllable capacity while maintaining
individual contracts with the vehicle owners [4]. This however requires predictable
charging habits of the vehicle owners and large number of individual vehicle owners
to minimize uncertainty caused by the variation in the participation of the individual
EV owners [38]. Commercial fleet owners are more suitable for such applications,
as the fleet availability can be predicted based on the fleet schedule. The charger
interfaces for such fleets are also higher in power to ensure timely charging of high
capacity batteries, when compared to passenger vehicles [39], thus requiring lesser
number of EVs to aggregate large amounts of power.
An analysis was done to evaluate the number of vehicles required to aggregate 1
MW of power based on different types of EVs. Passenger EVs can be broadly clas-
sified into regular, extended range, and high-performance vehicles. Regular vehicles
are either EVs or PHEVs, with a battery capacity ranging from 4 to 40 kWh. To
completely charge the vehicle in 1 hour, the charger has to be rated at 4–40 kW and
with chargers of these power ratings, the aggregator has to contract 25–250 vehicles
depending on the power ratings of the individual chargers aggregated to accumulate
1 MW of available power. If the charging is allowed to be completed in 8 hours
instead, the charger ratings change to 0.5–5 kW, and with these charger ratings, the
aggregator will need to contract 200–2,000 vehicles depending on the power ratings
of the individual chargers aggregated to accumulate 1 MW of available power. A
similar analysis was carried out for extended range and high-performance passenger
vehicles, light-duty, medium-duty, and heavy-duty vehicles. The results are shown in
Table 3.6. It shows that the nature of aggregation changes with the type of vehicles
used. For example, passenger vehicles usually have a defined availability pattern, i.e.,
they are expected to spend most of the time either in their garage (at night) or in a
parking lot at work (during the day) for most urban and suburban settlements. When
parked in their garages, the aggregator needs to ensure communication with various
vehicles (assuming a contract exists between its owner and the aggregating entity)
that may not be as physically close as in a parking lot or in a fleet garage.
Figure 3.5 shows the approximate power levels that can be achieved by aggregat-
ing vehicles with similar chargers and the grid services that they can be offered with
such aggregation [40]. If only Level 2 chargers were aggregated, around 1 MW could
be achieved with 89–160 vehicles which is a realistic number for fleet or parking lot
operators. Depending on how many vehicles are aggregated, these vehicles can offer
the available capacity for frequency regulation services or they can offer connected
reserve service that can respond within minutes. To achieve higher power levels,
the number of vehicles required will be unrealistic. DC fast chargers do provide
an opportunity to accumulate a higher available power with fewer vehicles, but the
purpose of installing fast chargers is to save time while charging, so managing the
number of connected vehicles to obtain the required total power may be a challenge.
However, DC fast stations with multiple DC fast chargers are integrated with energy
44 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 3.6
Vehicle Types and the Typical Specifications
Chargers Total Chargers
Battery Required Vehicles Required Total
Capacity for 1 hour for 1 for 8 hour Vehicles for Aggregating
Vehicle Type (kWh) charging MW charging 1 MW Entity
Passenger EV/ 4–40 AC Level 2 25–250 AC Level 1 200–2,000 Parking lot
PHEV AC Level 3 AC Level 2 owners
Commercial
aggregators
Extended 15–50 15-50 kW 20–67 AC Level 1 160–530 -
Range AC Level 2 AC Level
Passenger EV/ AC Level 3 2
PHEV
High 50–90 AC Level 2 12–20 AC Level 2 89–160 -
Performance AC Level 3
Passenger EV
Light Duty EV 11–90 AC Level 2 12–91 AC Level 1 89–728 Parking lot
AC Level 3 AC Level 2 owners
DC Fast Fleet owners
Charger
Medium Duty 100–150 AC Level 3 7–10 AC Level 2 54–80 Fleet owners
EV DC Level 2
Heavy Duty EV 200–400 AC Level 3 3–5 AC Level 3 20–40
DC Level 2 DC Level 1

Time
Scale

Energy
Hours

DC Fast Chargers
Management

Level 3
Minutes

Reserves
Level 2
Seconds

Frequency Home Level 1

Response

0.1 1 10 MW

FIGURE 3.5 Charger levels and grid service capabilities.

EV Fleet Charging Management 45

Electric Aggregate Bidding

Service
Vehicles Aggregator
Allocate Equilibrium Market
Price

FIGURE 3.6 EV aggregation, bidding, and disaggregation process.

storage and renewable energy resources form an aggregating entity, which can be
treated as virtual power plant or microgrid, thus providing usable grid services [41].
There are generally two options for the relationship between the aggregator and EV
owner/operator: a noncontractual and a contractual form [42]. The former involves a
free participation and “pay-as-you-go” type of remuneration and the second a regular
payment for the obligations of the service. Three types of contracts are defined based
on the manner and degree that EV can be controlled: price-based, volume-based,
and control-based [43]. Price-based contracts are associated with a price signal for
the activation of V2G, by which the driver defines a minimum price to provide V2G
service. Volume-based contracts involve the commitment of a predefined volume of
energy within a certain time interval, while the driver cedes control to the aggregator
as soon as the EV is plugged in with control-based contracts. The aggregated EVs
can then bid in the service market based on the price and reward opportunities [44].
After market clearing, the aggregator is also in charge of the disaggregation process
based on the service request considering the state of the EV and the expectation of
EV owners, as illustrated in Figure 3.6.

3.4 AVAILABLE ANCILLARY GRID SERVICES

WITH AGGREGATED EVs
Aggregated EVs can be used to offer various ancillary grid services. Ancillary ser-
vices are defined by the US Federal Energy Regulatory Commission as the support
services provided to the main electric power transmission function to maintain reli-
able operation of the electric power system. Various ancillary services have been
defined and identified for third-party storage specifically [40] and for DER operators
generally [45]. New ancillary services are also defined at the distribution grid level
due to the deeper penetration of distributed renewable energy resources such as bat-
tery storage in EVs [46]. The services that can be provided, along with the examples
found in the literature, by aggregated EVs are introduced and explained next.

3.4.1 Frequency response And regulAtion

Frequency response and regulation services are offered when a generation unit (or a
group of units), aggregated EVs in this case, can change its output in response to the
load fluctuations on a second to minute basis. EVs make a good candidate for this
service as they are interfaced with power electronic converters, and the response
rate of such converters can be very fast (only limited by the communication delays)
as there is no inertia involved, unlike rotating generators. Typically, a 1%–2% of
the peak load is required to perform regulation [45]. Since there is zero net energy
46 Electric Vehicle Integration in a Smart Microgrid Environment

exchanged for this service, it does not impact the overall battery charging process.
Providing this service, however, limits the available charging power to the vehicle
at a given time. This can be resolved by prioritizing the battery charging function at
a compromised charging power and only participates with the remaining capacity
resulting in a slow charging process or by participating in regulation only when the
battery has reached 80%~90% of its charge. If regulation is performed using aggre-
gated EVs, this can help offset more expensive fast response generators.
Various aggregator frameworks have been proposed in the literature that achieve
frequency regulation. A framework that optimizes for aggregator revenue is pro-
posed in Ref. [38] such that the available regulation capacity is modeled as a function
of the battery state of charge of each vehicle. The work in Ref. [47] maximizes the
available regulation capacity while scheduling the chargers for valley filling. The
aggregator framework in Ref. [48] shows a method of aggregating unidirectional and
bidirectional power chargers while considering their respective charging modes (i.e.,
constant current and constant voltage) and [49] presents an optimization approach for
a fractional order controller used for load frequency control using aggregated EVs.
The EVs in these papers are modeled as a large static battery, which has a collec-
tive SoC and power limits based on the aggregated individual EVs. This modeling
approach neglects the dynamics of the effective available regulation capacity due to
EV mobility.
A hierarchical model predictive control based aggregation framework with the
EV mobility modeling to follow a regulation signal is proposed in Ref. [50]. The
proposed framework works for all three levels of control, i.e., primary, secondary,
and tertiary [51] and proposes an economic analysis strategy for the available regula-
tion reserve (up and down) from EVs modeled with their dynamic mobility behavior,
while [52] presenting a combined aggregation framework for plugin EVs, thermal
loads, and CHPs for load frequency control with their individual uncertainties. The
authors in Ref. [53] present a framework to follow a regulation signal that results in
increased regulation capacity for the aggregator and therefore increased fleet uti-
lization while meeting the consumer demand. The research in Ref. [54] presents a
frequency regulation framework that either keeps the battery state of charge static or
charges it while providing frequency support to the grid and for an interconnected
area in Ref. [55]. Reference [56] proposes a regulation (up and down) capacity esti-
mation algorithm for the aggregator by modeling it as an EV queuing network. The
authors in Ref. [57] present contracting terms to maximize revenue for the aggrega-
tors with real regulation capacity estimation and [58] studies the effects of bidding in
frequency regulation market that provides more compensation for a higher quality of
regulation, i.e., a higher rate of ramping up and down.
EVs can be coordinated with other DERs for frequency regulation. An EV fleet
which is coordinated with a high penetration of wind generators is presented in Refs.
[59,60], while an economic analysis is presented in Ref. [61] for a system with high
penetration of solar and wind generators. Integrated control of data centers and PEVs
is demonstrated in Ref. [62].
A fair distribution of the regulation power among the participants is presented in
Ref. [63] by introducing a fairness evaluation index and approaches for fair alloca-
tion. The results differ as the fairness basis is changed. Reference [64] proposes a
EV Fleet Charging Management 47

fair allocation charge scheduling for regulation for an aggregator that models the EV
with its mobility uncertainties along with external energy sources (i.e. the vehicle
getting charged from another entity).
The impact of communication delays and uncertainties on the provision of regula-
tion service to the grid is assessed in Ref. [65]. As opposed to the central controller
approach for frequency regulation, the controller can also be of a distributed nature
such that communication requirements are minimized, and the control is performed
with local droop control that is calibrated to meet the global objectives as shown in
Ref. [39].

3.4.2 power smoothing

Though not formally included in the existing ancillary service scheme, power
smoothing services, especially high-frequency power smoothing, are proposed by
many researchers to smooth out highly variable loads [60] or generators connected to
the grid [68]. Due to the variable nature of the wind and solar energy resources, the
output profile of these generators contains large peaks, and therefore, these sources

Pmax
Charging Power

Pch Frequency

-Pmax
f0

FIGURE 3.7 Charger control function for frequency–watt control.

48 Electric Vehicle Integration in a Smart Microgrid Environment

need to be curtailed to ensure the stability of the network. Smoothing the output
profile not only decreases the curtailed wind or solar power but also enhances the
hosting capacity for such variable generation [69]. The service requires the provider
to react in a seconds-to-minutes time range that is overlapping with or is slightly
slower than the frequency regulation service.
Ramp rate-limiting services are introduced to compensate wind power fluc-
tuations in the grid by power smoothing control at the point of common coupling
(PCC) through demand response of EVs [70]. Power fluctuation smoothing of a wind
farm located near a smart parking lot is achieved by an ADAptive Linear NEuron
(ADALINE)-based method, which is more accurate than the conventional first-order
low-pass filter-based approach [71]. A moving averaging-based wind power smooth-
ing approach considering the constraints of wind ramping rate and the SoC of EV
battery is proposed in Ref. [72]. A load power smoothing control using the EV stor-
age capacity is also achieved by evaluating the load ramping rate and then allocating
power output requirements to each EV based on an SoC-adaptive method [73]. When
using EV batteries to smooth the voltage fluctuation due to renewable energy in the
distribution system, a hull moving average (MMA) can eliminate the lag problem of
the simple average method [74].
For real-time implementation and aggregation, a two-stage control strategy to
smooth the wind power fluctuation is proposed in Ref. [75], with the first-order filter
method adopted at the first stage to calculate the total power demand from aggre-
gated EVs, while at the second stage, an aggregator management center is introduced
to disseminate the power command to each load-management control center.

Pgen Pload
Padd_max
Charging Power
Additional

0 ΔP

-Padd_max

FIGURE 3.8 Charger control function for ΔP–watt control.

EV Fleet Charging Management 49

3.4.3 loAd/generAtion Following

The following service, unlike frequency regulation, is an energy service that is
required to make the profile of load or generation trackable by the dispatchable gen-
erators with limited ramping rate [78,79]. More general following services may also
involve peak shaving and valley filling in the profiles, for which the following service
can be treated as low-frequency power smoothing [80,81]. The following service is
also termed as energy imbalance service in some literature [82]. Since this service
requires a large amount of energy exchange, it depends on the state of charge of the
individual EV batteries along with its total power capacity. Another issue with pro-
viding this service using aggregated EVs as a storage unit is that it requires the stor-
age unit to be always connected to the grid. This is a difficult task to achieve for the
aggregator as most EVs will have a pattern of availability. To realize this service, the
aggregator will require significant coordination and optimization effort. The service
requires reaction in a minute-to-hour time.
Various aggregator frameworks have been proposed in the literature that perform
load and/or generation following. The probability of the location and charging state
of EVs is calculated to determine the zonal available capacity to provide flexible
ramping products for reliable power system operation [83]. A smart charge schedul-
ing algorithm for the EVs was proposed in Ref. [84] that also coordinates with a wind
generator such that the fleet discharges when the wind generator underproduces and
charges when the wind generator overproduces. Network hosting capacity for wind
was shown to be increased by optimal scheduling of vehicle charging in Ref. [85]. A
coordination strategy between EV fleets and building energy management systems
was presented showing that with optimal coordination, the peak demand on grid can
be reduced in Ref. [86].
A scenario-based robust approach to tackle the uncertainty of EVs and wind
power is described in Ref. [87] where the interdependency between the day-ahead
market prices and the aggregator’s bidding decisions is addressed using complemen-
tarity models. Vehicle uncertainty was modeled in Ref. [88], while a realistic battery
model was included in the optimization formulation in Ref. [89]. A stochastic match-
ing method for wind generation and EV charging load was proposed considering a
sizeable fleet with a mobility pattern that complements the wind generation in Ref.
[90] while [91] achieves the same using a three-level optimization problem for the
energy management system without an aggregator. A decentralized scheduling algo-
rithm for peak shaving and valley filling of the demand profile was proposed in Ref.
[92]. Real-time charging control with distributed optimization to reduce communica-
tion overhead was presented in Ref. [93] and to reduce the computation burden on the
central controller in Ref. [94].
A bidding strategy for the EVs to minimize the aggregator-owned renewable
energy resource (RES) generation variation was proposed in Ref. [95], while [96]
presents a bidding framework for aggregated EVs to bid for a downward reserve (i.e.
the charging load is reserved for low price period). The authors in Ref. [97] present
an aggregated control of wind power with EVs ensuring a match between the two by
scheduling the EVs being charged and curtailing the wind. A distributed aggregation
and coordination scheme was proposed [98] to increase renewable energy sources
50 Electric Vehicle Integration in a Smart Microgrid Environment

and EV hosting capacity that also improves the voltage profile of the network. Joint
optimization for the aggregated EVs, carbon capture plant, and a fossil fuel plant was
presented in Ref. [99] with an additional objective to limit emissions. Asset-aware
aggregation frameworks that limit feeder capacity violations [100], transformer con-
gestion [101], and feeder losses [102] and that limit the total load, voltage profile of
the network, and the phase imbalances in the network due to single phase chargers
were also proposed [103].

3.4.4 spinning reserve

Spinning reserves are required by the grid as online generation (or load) that can be
ramped up (or ramped down) within 10 minutes of generation outage and sustain
the operating point for typically up to 2 hours, the time required for the secondary
reserves to takeover. It is demonstrated that EV providing spinning reserve service is
more profitable than providing energy service only [107]. This service like the load/
generation following function also requires net energy exchange during the time the
service is provided. Therefore, it is also dependent on the state of charge of the indi-
vidual EV batteries along with its total power capacity. To realize this service, the
aggregator will also require significant coordination and optimization effort.
Various aggregator frameworks have been proposed in the literature that allow
estimation of the available spinning reserve to bid in the reserve market. An autono-
mous distributed spinning reserve based on the expected EVs connected to the grid
for charging is described in Ref. [67], while a distributed control strategy for the avail-
able reserve is proposed in Ref. [108]. A combined bidding framework for spinning

Pmax
Charging Power

Pgen 0 Pload

-Pmax

FIGURE 3.9 Charger control function for watt–watt control.

EV Fleet Charging Management 51

reserve and regulation for bidirectional chargers is proposed in Ref. [109] and for
unidirectional chargers in Ref. [110], while a combined aggregation platform for reg-
ulation and spinning reserve is proposed in Ref. [111]. An estimation methodology
for the true reserve available to the aggregator, factoring in the uncertainty in the
mobility patterns of the vehicles, is proposed in Ref. [112]. Combined optimization of
the cost of spinning reserve and the expected energy not supplied is performed [113].
A reserve provision for a network with a high wind penetration scenario using aggre-
gated EVs is proposed in Ref. [114], while a synchronous machine emulating control
for individual chargers and charging stations is proposed in Refs. [115,116] such that
the chargers are able to implement the droop control with an emulated spinning rotor
inertia reserve. The charger function already presented in Section 3.4.3 should also
suffice for the implementation of spinning reserve applications as well.

3.4.5 reActive power support

Reactive power support is required to increase the power quality and reliability of the
grid [117,118]. It can be provided by the power electronic interfaces of the charger.
The benefits of providing reactive power using aggregated EVs are that they provide
it where it is needed, i.e., near the loads, reducing the power losses due to transmis-
sion of reactive power. It also replaces or reduces the requirements for capacitor
banks on the distribution system. Since there is no energy exchange required, this
service can be provided even when the batteries are fully charged; however, when
the battery is being charged, the available reactive power is limited by the rating of
the charger interface [119]. Reactive power is mostly required for voltage support
and harmonic compensation applications. Since all these functions depend on the
local operating variables, the communication burden is not increased significantly
by offering this service.
Various aggregator frameworks have been proposed in the literature that allow
reactive power support from the EV chargers. EV chargers capable of injecting reac-
tive power to the grid as well as of bi-directional power flow are proposed in Ref.
[120], while an integrated PV and battery charger system is proposed in Ref. [121]
with a common bidirectional interface to the grid. The charging operation is con-
trolled to compensate for the dips in the PV system power output while providing
reactive power support to the grid which can be used to provide voltage support or
other power quality services.
If the charger controls are designed to follow reactive power setpoints, provision
of reactive power support can be supported by the charger.

3.4.6 voltAge support

Typically, voltage support is provided by reactive power injection; however,
in distribution grids (due to their resistive nature) voltage profiles rise due to
the power injected by the distributed resources. Therefore, by curtailing the active
power injected or by increasing the active power absorbed by the aggregated EVs, the
voltage rise can be alleviated. Offering this service increases the hosting capacity of
the grid for distributed generation.
52 Electric Vehicle Integration in a Smart Microgrid Environment

Various aggregator frameworks have been proposed in the literature that provide
voltage support service to the grid. A framework that improves the voltage profile of
the network with the chargers to increase the EV hosting capacity of the network was
presented in Ref. [122] with an assumption that EVs were always available for V2G
service. A stochastic distributed optimization of reactive power with conditional
ensemble predictions of V2G capability was introduced in Ref. [123]. A combined
formulation for both voltage and frequency support was proposed in Ref. [124] for
generalized reactive end-user devices including PEVs. A local and distributed volt-
age droop control was proposed in Ref. [125] that de-rates the charging power if the
voltage goes down resulting in a better voltage profile, albeit at the cost of increased
charging time for the vehicles. Another local voltage controller was proposed in Ref.
[126] to reduce losses and communication requirements. An integrated EV and PV
system was proposed in Ref. [127] that allows charging of the EV to support the PV
output by charging when PV overproduces. Demonstration projects were presented
in Ref. [128] for a fleet of EVs with unidirectional chargers that is able to provide
V2G services like voltage support and frequency support.

1. Charger Functions for Voltage Support: Voltage–Watt Control

To provide an active power voltage support service, a control loop that is
capable of changing the charger maximum active power output in response
to a change in the voltage of the node to which it is connected is discussed
in Refs. [125,129]. The required voltage vs. watt control function is shown in
Figure 3.10. It changes the maximum power limit in reaction to the voltage
at the node it is connected to [130].

Pmax
Power Limit

Voltage (pu)
1

-Pmax

FIGURE 3.10 Charger control function for voltage–watt control.

EV Fleet Charging Management 53

% Reactive power available

Qmax

0 Voltage (pu)

-Qmax
1

FIGURE 3.11 Charger control function for voltage–VAR control.

3.4.7 discussion
Each service mentioned in Sections 3.4.1–3.4.6 requires a control loop that can be
customized to provide a service based on local measurements of each EV [133]. The
control loops can be implemented at the aggregator control center that applies the
loop for each vehicle and then dispatches the set points [134] or they could be indi-
vidually implemented in the EVs themselves. A purely centralized approach would
require high processing power with the aggregator and a communication link with
low latency, high bandwidth, and reliability. A purely decentralized approach, on the
other hand, reduces the processing requirement and can operate with low bandwidth,
high latency communication links [135]. A good compromise is when a hybrid hier-
archical approach is used [136]. The aggregator may serve as a gateway to the rest of
the vehicles, while only performing coordination and optimization functions, while
each EV manages its own output.
The V2G services that can be offered using aggregated EVs, the functions required
to be implemented on the chargers to offer the services, the reaction time required for
the services, and their value propositions are summarized in Table 3.7. In the table,
all except the last row are active power support functions and can be provided using
type I chargers with maximum limits of 1 to −1 pu. Types II and III can provide
the same function with limits of 0 and 1 pu. Similarly, the reactive power support
function can be provided using a type I and II charger with limits equal to the rated
charger power; however, type III charger does not have the capability of providing
reactive power support.

3.5 CASE STUDIES

In this section, several case studies for the various types of ancillary services dis-
cussed in Section 3.4 are presented. The typical results selected are straightforward
and illustrative to showcase different V2G services, while studies based on more
advanced approaches can be found in the literature.

3.5.1 Frequency regulAtion

In Ref. [137], the control strategy for EVs participating in voltage and frequency
regulation is investigated. A frequency droop control is adopted, while the
54 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 3.7
Types of V2G Services and the Typical Specifications
Control
Grid Function Function Reaction Time Value Proposition
Regulation/fast Frequency–watt Seconds–minutes • Offset need for fast-response or
regulation control very fast-response generation
Power smoothing ΔP–Watt control Seconds–minutes • Increase the grid hosting
capacity for highly variable
generation
• Increase the network reliability
• Deferral of infrastructure
upgrades to cater peaking loads
Load following Watt–watt control Minutes–hours • Reduce generation variability
• Extra generation sold elsewhere
Generation following Watt–watt control Minutes–hours • Enhanced hosting capacity for
highly variable renewable
generation
• Reduce generation curtailment
for variable renewable
generation
Energy arbitrage/load Direct P control Hourly • Allow peak power generation to
leveling/peak shaving be used else where
Spinning reserve Frequency–watt Seconds–10 • Free up the generation reserve to
control minutes be used elsewhere
Voltage active power V–watt control Seconds • Increase renewable sources
support hosting capacity at the
distribution level
Voltage reactive power V–var control Seconds • Provide reactive power support
support near the loads
• Increased power quality and
system stability

available V2G regulation power is also constrained by the EV owners’ transpor-

tation demand, initial and real-time battery SOC, and departure time. In case of
voltage and frequency deviation beyond a threshold, the EVs will participate in
regulation by fast adjustment of charging or discharging power without violating
the constraints.
A case study of frequency response after 0.02 pu load increase was presented in
Ref. [137]. Before EVs participate in frequency regulation, the maximum frequency
deviation can reach up to 0.03 pu, which is reduced to 0.005 pu with the EVs partici-
pating in frequency regulation. The turbine output power of the conventional syn-
chronous generators will also be affected. The overshoot of the mechanical power
output can reach 0.025 pu without EVs’ frequency regulation, which was almost
completely eliminated due to EVs’ control effect.
EV Fleet Charging Management 55

3.5.2 power smoothing

A real-time control strategy for aggregated EVs to smooth the fluctuation of wind-
power is presented in Ref. [75]. The energy boundary model of EV battery is estab-
lished to determine the charging/discharging margin of an EV at each moment. To
smooth the power fluctuation mainly caused by wind variation, a first-order low-pass
filter is used to determine the total power command of aggregated EVs, which is then
allocated among individual EVs.
The effect of power smoothing control is demonstrated in the showcased results.
In the disorderly charging mode, the maximum and minimum total load power is
4,440 kW and 2,050 kW, respectively, and the average total load power fluctuation
rate is 0.0179. With the proposed control, the maximum and minimum are changed
to 4,464 and 2,513 kW, respectively, and the average total load power fluctuation rate
is reduced to 0.0078. The power fluctuation rate and the peak-valley difference are
reduced by 56.4% and 18.4%, respectively. Therefore, not only the fluctuation rate of
the total load is greatly decreased, but the peak-valley difference is reduced as well.

3.5.3 loAd/generAtion Following

A case study with EV used for load leveling (a general type of load following) is
presented in Ref. [138]. Assuming that the target power curve with peak-shaving and
valley-filling function is known, for example, determined based on grid efficiency,
EV operation and its power, and the time-of-use power, a minimization function is
designed to match the target power curve, and the V2G plan curve is proposed. The
corresponding constraints include connected EV numbers, EV battery characteris-
tics, and the EV user customized parameters such as charging period.
For a case study presented in Ref. [138], the valley-filling period approximately
covers 23h00 to 7h00, and peak shaving occurs between 8h00 to 12h00 and 14h00
to 16h00. The maximum peak-shaving value is about 10% of the general power load
at 10 am. The forecast load value is less than the target load value at 1 am when the
grid is at the valley time, and its spare power can charge the EVs. The absolute value
of the difference at this time is equal to the maximum power value of the grid supply
and smaller than the total maximum charging demand of EVs. The charging power
will then be limited to this value and distributed among individual EVs. On a daily
scale, the V2G plan curve is close to the target curve except in the period between
19h00 and 21h00 because of the small number of connected EVs. As demonstrated,
the load profile is greatly flattened, with the peak-shaving and valley filling control
of EVs.

3.5.4 spinning reserve

The authors in Ref. [139] present a market-clearing mechanism with a probabilistic
spinning reserve considering the uncertainty of wind and EVs. The system operator
minimizes payments to generators and EV aggregators for providing energy and
reserve, including energy cost, the start-up cost of generation units, reserve cost from
conventional units, the cost of expected energy-not-supplied (EENS), and the cost
56 Electric Vehicle Integration in a Smart Microgrid Environment

of services from EV, with the presence of uncertain wind and EV participation. The
proposed model is formulated and linearized so that it can be solved by means of
available mixed-integer linear programming (MILP) solvers. It is shown that the
EVs’ participation in energy and reserve markets with competitive offer prices can
counterbalance the increasing EENS and reserve costs of higher levels of uncertain
wind profile.
Assuming that the price of V2G service is lower than the spinning reserve offered by
generation units, the required amount of spinning reserve can be offset when EVs are
used for this service. The results indicate the system operator schedules V2G only in the
energy clearing process because of its low price. The spinning reserve and EENS are
reduced by increasing V2G injected power. The operating cost is reduced by purchasing
more V2G service and committing fewer conventional generation units.

3.5.5 voltAge And reActive power support

In Ref. [131], the sole effect of P(V) voltage-watt droop charging (lower charging
power at LV level), conventional voltage-var droop support Q(V), and the combina-
tion of the two methods [P Q](V) is examined for controlled EV charging with weak
distribution grids. The control performance of the voltage profile due to a voltage dip
at PCC is examined. Due to the high R/X ratio of the distribution feeder, the volt-
var droop control from EV chargers alone is unable to meet the minimum voltage
threshold, though the result is better than for the base case without any voltage sup-
port. In contrast, the P(V) droop is more effective in improving the voltage, making
voltage lie within the threshold. The combined voltage support method has the best
performance regarding the effect of voltage regulation. The combined method can
also partially relieve the adverse effect of the P(V) droop charging, given that for the
same voltage improvement, increased charging power and thus shorter charging time
can be achieved by the additional voltage-var support.

3.6 CHALLENGES AND FUTURE RESEARCH DIRECTIONS

Despite the various frameworks proposed for the implementation of V2G tech-
nologies, the concept faces impediments in its wide acceptance leaving room for
further research to be performed to develop practically implementable solutions.
This section will elaborate on the challenges faced by the concept in the follow-
ing subsections.

3.6.1 technology initiAtives

1. Battery and Charger Technology
One of the major drawbacks of providing V2G services is the additional
degradation of the battery and its associated electronics due to additional
cycling. Conventionally, EVs use high power lithium-ion batteries. Failure
modes that reduce the battery lifetime include (1) overcharging; (2) over-
heating; (3) deep discharge; and (4) cycling. The battery lifetime is charac-
terized by its residual usable capacity. By offering V2G services, the battery
EV Fleet Charging Management 57

and its charger are cycled more often, which results in additional aging of
the system reducing its lifetime, i.e., the usable capacity fades earlier than
it would have if the battery was only used for vehicle propulsion. The pro-
vision of reactive power service does not cycle the battery. The charger,
however, may be over-cycled in this case.
One way of addressing the issue of battery degradation is by online mon-
itoring of the battery state of health [140] and proactively acting to avoid
undue degradation of the battery. Another approach can be to include the
battery degradation cost in the contract negotiated with the aggregator [141]
or by adopting a framework that limits the services provided to the grid
based on the cost of battery degradation [142,143]. A few solutions based
on a second use of battery are also proposed [144]; however, the economic
justification of such a use is unclear [145,146].

The communication infrastructure should ensure scalability, security,

and privacy-preservation to allow integration of a large number of EVs by
the aggregator. Currently, the concept of V2G has only been implemented
in small scale demonstration projects using short-range communication net-
works. To be widely accepted, the concept entails a robust communication
network that meets performance requirements [148,149]. Furthermore, the
impact of aggregator–EV communication link uncertainties on the opera-
tion of the aggregator is yet to be ascertained.
3. Impact on Power Infrastructure
A high penetration of EVs may overload the distribution system when
EVs are charged without any coordination measures implemented. Even
with coordination, as the penetration of EV loads increases, an eventual
overload on the network may result. This overload may not require immedi-
ate network upgrades as most of the distribution transformers are designed
for cold pickup, i.e., for 200% load. On the other hand, if the growth of EV
loads accelerate, a network upgrade will be required. Furthermore, not all
EV users will participate in coordinated charging and providing V2G ser-
vices. Integrating EVs with the grid will not only require a power network
upgrade but will also require investment in additional energy generation.
This additional energy can be provided by RESs; however, the stochastic
nature of the generation must be matched with the stochastic nature of the
EV load.
58 Electric Vehicle Integration in a Smart Microgrid Environment

Additionally, since most V2G services require bidirectional power flow capability
from the charger, a comprehensive recalibration and re-coordination of protection
devices will be required to ensure adequate protection of the grid. With the trend
toward fast charging and higher efficiency requirements, high-power chargers based
on multilevel power electronic converters, wide-band-gap power switches, and solid-
state transformers replacing the conventional bulky AC transformers can enhance
the performance of EV charging stations and better enable V2G services to the grid
[150,151].

3.6.2 economicAl Aspects

1. Fair play
There needs to be a strong economic incentive for the individual EV
owner to allow the time-share use of their EV battery as a load leveling
and load management device. For each ancillary service, the mechanism
of market participation and payment should be designed to benefit both the
grid and the EV owner without incurring economical losses to EV owner.
This will mainly involve battery and charger life reduction and may result
in disrupting travel plans and increased range anxiety. A complete business
model of a V2G market design should take into account the critical cost and
revenue elements of each player [152]. The cost elements include battery
wear and replacement, charger over cycling, EVSE network and updates,
communication hardware and software, and aggregator devices and soft-
ware. The revenue elements, however, will involve a pool of aggregated
vehicles, the EV plugged-in hours, and the available power and energy for
V2G services. At the present stage, EV customers are motivated to partici-
pate in the most profitable service market such as frequency regulation. In
the long term, they can also be encouraged to the services that are more
urgent. For example, during an emergency with grid disconnection due to
a major storm, the EVs within a community can provide emergency power
for a limited period to selected devices, such as smartphones, lights, and
refrigerators. In that case, the back-up power services will be given more
importance and offered higher prices than services under normal circum-
stances. Furthermore, the issue of fair distribution of the service load to
individual EVs is not easy to capture in contractual and operational terms.
The concept of fairness varies depending on user, region, culture, and other
socioeconomic factors. A step in this direction was regulation ensuring
higher compensation for higher quality of reserve in the frequency regula-
tion market [153].
2. Capacity estimation
One of the issues that impedes the wide acceptance of the V2G con-
cept is the ability of the aggregator to accurately estimate (or forecast
if necessary) the total capacity available to provide specific V2G ser-
vices. On the grid side, the estimation of the required capacity for given
services becomes more challenging for the markets that have a lower
time resolution, for example, the day ahead and hour ahead reserves
EV Fleet Charging Management 59

and regulation market [154]. Furthermore, the structure of the regional

energy and ancillary services market, along with the energy and power
tariff scheme, may play an important role in the operation and control
of the aggregated EVs. The market may allow for a viable business case
(assuming the aggregation of adequate number of EVs) for the EV owners
to participate in V2G services, as shown in Ref. [155] for California and
in Ref. [61] for Florida. However, if the market price for such services is
too low, due to various socio-political reasons, the economics for V2G
services no longer hold [156]. The rates and prices of the grid services are
not always set by economic factors. Socio-political issues significantly
influence the electricity rates. For this reason, the solutions for removing
such impediments are not straightforward.

3.6.3 environmentAl Aspects

Climate change and air pollution are the two potential environmental concerns
that V2G technology can help address. EVs can lower total emissions, compared
to conventional gasoline-powered vehicles. By replacing high carbon electricity
sources with renewable energy resources and replacing gasoline-powered vehi-
cles with EVs, emission reduction and climate change benefits can be achieved.
Moreover, the ancillary service markets can also be decarbonized (to an extent)
by employing V2G technologies. However, research indicates that the energy mix
in a utility power procurement, which might vary temporally and spatially, among
other aspects, will also affect the decarbonization effect of V2G. It is pointed out
that the penetration of EV participating in V2G services should exceed a certain
ratio for the emission savings to exceed the additional electricity-based emissions
as a result of charging [157]. By optimized charging schedules, V2G can achieve
significant carbon emission reductions compared to conventional gasoline vehi-
cles. However, for a grid with higher emission intensity, using V2G for load shift-
ing may even increase the carbon emission [158]. In Ref. [159], environmental
factors such as carbon reduction are incorporated in the impact of EV integra-
tion except for the conventional techno-economic benefits. Besides carbon reduc-
tion, the air pollution issues from gasoline oil such as NOx, SOx, etc. can also be
relieved by the introduction of renewable energy and EVs.
However, EV battery manufacturing and disposal may also introduce some water
and soil pollution from electrodes and electrolytes, especially heavy metals such as
Pb, Cd, Ni, Co, and Mn and harmful acid and alkaline liquids. If EV battery life is
significantly reduced in V2G services and more battery manufacturing is needed,
such pollution issues will be more prominent.

3.6.4 sAFety And security

There are generally two types of safety issues in V2G services: (1) electrical safety
when EV is used by the customer or is connected to EVSE. (2) Communication
safety and cybersecurity, namely the privacy information shared by individual EV
users with EV aggregator.
60 Electric Vehicle Integration in a Smart Microgrid Environment

1. Electrical safety
For AC charging, the AC arc can self-extinguish at the zero-crossing
point, so the hazard of electrical shock and difficulty of protection will be
lower. In contrast, DC-charging will not self-extinguish, and there is inad-
equate fault-sensing equipment and switchgear for protection. A DC arc
does not have a zero-crossing point, rather a constant potential, if constant
power is available. There are many causes for DC arcing, among which two
types of arcing are most common for EV: a series arcing and a parallel arc-
ing. Both types of DC arcs are difficult to detect but can be dangerous if not
mitigated. Some of the standards on electrical safety and protection related
to EV charging include IEC 61140, IEC 62040, IEC 60529, IEC 60529, ISO
6469-3, etc. [160].
2. Cybersecurity
Grid power control signals and billing information will be transmitted
over the communication channel. EV users will be concerned about the
loss of data privacy such as the charging behavior or transaction details
when using EV or providing V2G service [161]. Security for communica-
tion, or cybersecurity, must be guaranteed in the communication process.
The need for security considerations can be seen in 802.11b and 802.1X
security failures. It shows that poor implementations of an encryption algo-
rithm can lead to security issues. The communication medium, either wired
or wireless, will be the critical factor determining cybersecurity. Wireless
networks, compared with wired networks, are more susceptible to secu-
rity problems because the medium can be “tapped” from anywhere in the
proximity of the network. Session hijacking, Denial of Service (DoS), and
“rogue” access points are three common forms of attack on wireless net-
works. Wired networks can also be “tapped”, but it requires a physical con-
nection to the network. The standards on cyber-security include IEC 62351
and SAE J2931/7 [33].

3.6.5 Future directions

1. Advanced EV charging networks with better placement and covering wider

areas will be built [162]. These include slow EV chargers in home and urban
environment and fast EV charging station integrating renewable energy
resources and onsite energy storage for long-distance travel. Though V2G is
mainly targeted at slow EV chargers, by aggregating fast chargers or using
charging stations, V2G service may be also feasible, with more advanced
coordination. Besides, new technologies such as wireless charging and
unmanned vehicle driving will be matured to reduce the siting limitation of
charging and make the V2G process smoother and less impactful to the EV
users.
2. Internet and communication technology aided new business and operation
modes such as blockchain and digital twin will boost V2G development.
The blockchain technology adopting point-to-point (P2P) transaction mode
EV Fleet Charging Management 61

will digitalize and secure V2G business [163]. The digital twin or cloud-
based platform can monitor the real-time status of EV and grid and ensure
precise information procurement to control EV charging and V2G service
provision [164]. The high-speed 5G communication technology will be a
key driver in making these applications implementable. The merge of the
Internet of Things (IoT), artificial intelligence (AI), sharing economy, smart
grid, smart city, and cloud computing will make the integration of V2G
service into the grid market more natural and the electrified transportation
more intelligent [165].
3. More interdisciplinary studies will be carried out. As reviewed, the V2G
integration will involve various factors, including technical, financial, socio-
environmental, and individual/behavioral [166]. The previous studies mainly
work on a single type of modeling or discipline, especially techno-economic
assessment of V2G, but sophisticated behavioral and environmental mod-
els are not fully taken into account [167]. Consumers’ notion, willingness,
and expectation will be as important as technology in determining whether
they will participate in V2G services. For this, more research based on
multiple-method modeling approaches should be conducted.

3.7 CONCLUSION
The requirements, general charger topologies, and their control loops are reviewed
in this chapter. It also presents the potential of EV batteries and their associated
chargers to be used as distributed storage to provide ancillary grid services when
aggregated. The services that can be provided depend on the power electronic grid
interfaces of the chargers, the cumulative power of the chargers, and the cumula-
tive energy of the batteries. The functions include fast frequency regulation, energy
arbitrage and generation following in addition to the conventional ancillary support
functions performed by synchronous generators such as reactive power compensa-
tion. The control functions required in the chargers to implement the services are
also reviewed, and the impact of the charger topologies on their capability to provide
the services is described. Challenges that impede wide adoption of the V2G services
and the future research directions are also discussed.

REFERENCES
1. Statista. 2020. Worldwide number of electric vehicles in use from 2012 to 2019 (in 1,000s).
Available: http://www.statista.com/statistics/270603/worldwide-number-of-hybrid-and-
electric-vehicles-since-2009/.
2. E.D.T. Association. 2019. Electric drive sales dashboard. Available: http://electricdrive.
org/index.php?ht=d/sp/i/20952/pid/20952.

4. M. A. Ortega-Vazquez, F. Bouffard, and V. Silva, “Electric vehicle aggregator/sys-

tem operator coordination for charging scheduling and services procurement,” IEEE
Transactions on Power Systems, vol. 28, no. 2, pp. 1806–1815, 2013.
62 Electric Vehicle Integration in a Smart Microgrid Environment

5. W. Kempton, J. Tomic, S. Letendre, A. Brooks, and T. Lipman, “Vehicle-to-grid power:
Battery, hybrid, and fuel cell vehicles as resources for distributed electric power in
California,” 2001. University of California at Davis, Institute of Transportation Studies
Report (UCD-ITS-RR-01-03), Davis, CA(2001), Available: https://itspubs.ucdavis.edu/
publication_detail.php?id=360.

7. R. J. Bessa and M. A. Matos, “The role of an aggregator agent for EV in the electric-
ity market,” in 7th Mediterranean Conference and Exhibition on Power Generation,
Transmission, Distribution and Energy Conversion (MedPower 2010), Agia Napa,
Cyprus, 2010, pp. 1–9.

12. S. Haghbin, S. Lundmark, M. Alakula, and O. Carlson, “Grid-connected integrated

battery chargers in vehicle applications: Review and new solution,” IEEE Transactions
on Industrial Electronics, vol. 60, no. 2, pp. 459–473, 2013.

17. SAE International Surface Vehicle Recommended Practice. “J1773 - SAE electric vehi-
cle inductively coupled charging,” SAE International J1773, Rev. Jun 2014.

EV Fleet Charging Management 63

64 Electric Vehicle Integration in a Smart Microgrid Environment

52. M. D. Galus, S. Koch, and G. Andersson, “Provision of load frequency control by

PHEVs, controllable loads, and a cogeneration unit,” IEEE Transactions on Industrial
Electronics, vol. 58, no. 10, pp. 4568–4582, 2011.

EV Fleet Charging Management 65

60. C. Mu, W. Liu, and W. Xu, “Hierarchically adaptive frequency control for an
EV-integrated smart grid with renewable energy,” IEEE Transactions on Industrial
Informatics, vol. 14, no. 9, pp. 4254–4263, 2018.

72. S. Ziqi, P. Wei, D. Wei, and Q. Hui, “Capacity configuration of electric vehicle charging
station for wind power smoothing,” in 2014 IEEE Conference and Expo Transportation
Electrification Asia-Pacific (ITEC Asia-Pacific), Beijing, China 2014, pp. 1–5, IEEE.

66 Electric Vehicle Integration in a Smart Microgrid Environment

78. N. Navid and G. Rosenwald, “Market solutions for managing ramp flexibility with high
penetration of renewable resource,” IEEE Transactions on Sustainable Energy, vol. 3,
no. 4, pp. 784–790, 2012.

81. W. Zhang, K. Spence, R. Shao, and L. Chang, “Grid power-smoothing performance

improvement for PV and electric vehicle (EV) systems,” in 2018 IEEE Energy
Conversion Congress and Exposition (ECCE), Portland, OR, USA 2018, pp. 1051–1057.
82. M. G. Vayá and G. Andersson, “Self Scheduling of plug-in electric vehicle aggregator to
provide balancing services for wind power,” IEEE Transactions on Sustainable Energy,
vol. 7, no. 2, pp. 886–899, 2016.

85. C. Shao, X. Wang, X. Wang, C. Du, C. Dang, and S. Liu, “Cooperative dispatch of wind
generation and electric vehicles with battery storage capacity constraints in SCUC,”
IEEE Transactions on Smart Grid, vol. 5, no. 5, pp. 2219–2226, 2014.

91. C. T. Li, C. Ahn, H. Peng, and J. Sun, “Synergistic control of plug-in vehicle charging
and wind power scheduling,” IEEE Transactions on Power Systems, vol. 28, no. 2,
pp. 1113–1121, 2013.

EV Fleet Charging Management 67

68 Electric Vehicle Integration in a Smart Microgrid Environment

117. H. Tanaka, T. Tanaka, T. Wakimoto, E. Hiraki, and M. Okamoto, “Reduced-capacity

smart charger for electric vehicles on single-phase three-wire distribution feeders with
reactive power control,” IEEE Transactions on Industry Applications, vol. 51, no. 1, pp.
315–324, 2015.

EV Fleet Charging Management 69

128. K. Knezovic, S. Martinenas, P. B. Andersen, A. Zecchino, and M. Marinelli, “Enhancing

the role of electric vehicles in the power grid: Field validation of multiple ancillary ser-
vices,” IEEE Transactions on Transportation Electrification, vol. 3, no. 1, pp. 201–209,
2016.

135. P. Richardson, D. Flynn, and A. Keane, “Local versus centralized charging strategies
for electric vehicles in low voltage distribution systems,” IEEE Transactions on Smart
Grid, vol. 3, no. 2, pp. 1020–1028, 2012.

70 Electric Vehicle Integration in a Smart Microgrid Environment

EV Fleet Charging Management 71

 4 Challenges to Build a
EV Friendly Ecosystem
Brazilian Benchmark
Ana Carolina Rodrigues Teixeira
University of São Paulo

CONTENTS
4.1 Introduction .................................................................................................... 73
4.2 Context and Brazilian Portrait ........................................................................ 74
4.3 Challenges and Opportunities through the Brazilian Initiatives .................... 76
4.3.1 Economy and Production.................................................................... 76
4.3.2 Public Policies..................................................................................... 78
4.3.3 Customer Acceptance ......................................................................... 81
4.3.4 Market, Logistics, Energy Matrix, and Environment ........................ 82
4.3.5 Smart Grid ..........................................................................................84
4.4 Case Study ...................................................................................................... 85
4.4.1 Public Perception ................................................................................ 85
4.4.2 Numeric Model ................................................................................... 88
4.5 Summary and Conclusions ............................................................................. 91
Acknowledgments....................................................................................................92
References ................................................................................................................92

4.1 INTRODUCTION
Discussions about electric vehicle (EV) cannot be based just on automotive technol-
ogies. Many countries have been investing in fleet replacement using EV to reduce
greenhouse gas (GHG) emissions and local air pollution. Different goals were pro-
posed over the last years to replace the conventional fleet all around the globe, as
the Norwegian case in which they had planned to change their entire fleet. This
kind of target, however, cannot be equally used, and the whole ecosystem must be
taken into consideration to stimulate and disseminate such technology. Hence, the
introduction of EVs depends on many factors that should be considered, such as
geography, public and private subsidies, energy, customer acceptance, investments,
and infrastructure.
Many pilot projects have been coming to fruition to evaluate efficiency, costs,
maintenance, and customer acceptance, although the extension and different char-
acteristics of each region contribute to hinder the process. Therefore, this chapter

73
74 Electric Vehicle Integration in a Smart Microgrid Environment

covers these different characteristics of the Brazilian reality, since the country pres-
ents several economic problems in adittion to where the EVs started to be part of
the country’s reality over the last five years. Furthermore, this chapter also provides
some examples to clarify how the country has been dealing with these issues.
Challenges of the EV’s introduction using the Brazilian benchmark and the impor-
tance to develop the market before the actual introduction and possible fleet replacement
will be discussed. Then, the chapter identifies the existing opportunities in Brazil that
can be used as a differential to boost this kind of technology. This benchmark will be
based on previous studies made by the author where the likelihood of replacement of the
conventional taxi fleet to EV in different Brazilian cities was analyzed.
This chapter is divided into four parts. Section 4.2 brings the Brazilian context
and its current situation, demonstrating how EVs can be influenced by different
issues. Section 4.3 presents the challenges the country has been facing in this area,
the opportunities to expand the market, and in what shape or form other characteris-
tics such as energy matrix and environment can help us to reach a better ecosystem
to introduce EV. Section 4.4 shows a case study applied to a municipality in Brazil
about the replacement of conventional vehicles to EV in the taxi fleet. The last sec-
tion concludes such chapter with lessons learned and final remarks.

4.2 CONTEXT AND BRAZILIAN PORTRAIT

Firstly, there is a chance that one would table the question: “why are you talking
about ecosystem while EVs are being explained?”, and the answer is: because the
word “ecosystem” is commonly related to biological issues. Nevertheless, according
to the Cambridge dictionary [1], an ecosystem is “all the living things in an area and
the way they affect each other and the environment”. This word, however, can have
another meaning related to the business scope such as “a group of business activi-
ties that affect each other and work well together” [1] or “all the different activities,
companies, systems, etc. that are involved in a particular area of business, especially
new technology” [2]. These definitions explain why it is important to consider differ-
ent aspects when we talk about the introduction of EVs in a specific place, wherein
geography, subsidies, energy, customer acceptance, investment, and infrastructure
will always influence the success of this new product. Thus, to build a friendly eco-
system for the EV’s introduction, it is necessary to consider all these multiplicities.
Different countries require distinct strategies to insert a new product in the mar-
ket; it is important, though, and it adds knowledge to observe what other subjects
are doing over the world. It is helpful not only to build actions to deal with future
problems but also to learn with flawed solutions. Geography and country extension,
for example, will impact, directly the use of EVs, especially when the battery electric
vehicles (BEVs) are considered. In Brazil’s case, for example, the country depends on
different modal types (i.e. road, rail, waterway, airway, and pipeline), but more than
60% stands for the road modal. Brazil has around 1.7 million km of road networks
and a total fleet of 100 million vehicles, which makes it common to travel more
than 300 km during holidays and vacation periods. Consequently, there is a need for
higher-autonomy batteries and a strong recharging-complex infrastructure to supply
the electricity demand.
Challenges to Build a EV Friendly Ecosystem 75

Another issue to be taken into consideration is the electric matrix, which is differ-
ent in each country, although extremely important to the following discussion: (1) it
is the base of the supply chain related to the demand required for EVs, and its man-
agement is a crucial key to reach intermittent energy quality; (2) the lack of a strong
electric matrix contributes to failures and energy outages.
Furthermore, EV has been considered a way to reduce GHG emissions and local
air pollutants. Hence, a clean electric matrix is important to reach this goal, since,
in 2018, 83% of the electricity produced in Brazil came from renewable sources,
which include hydropower, wind energy, solar energy, and biomass [3]. Nonetheless,
it is necessary to manage supply and demand during drought periods to avoid cases
of energy outage in Brazil. Lastly, even with an electric matrix based on fossil
fuel sources, EVs can contribute to reducing local air pollution, directly improv-
ing the air quality in cities and contributing to better public health by reducing the
attributable deaths to air pollution. Figure 4.1 shows the Brazilian electric matrix
composition.
As mentioned above, the infrastructure is a key for the introduction of EV in the
market since without recharging infrastructure, there is no possibility to maintain
BEVs in any place. However, this issue might increase in developing countries, such
as Brazil, due to high EV purchase costs, which may lead to a fewer number of such
vehicles on streets and consequently minor need for public infrastructure.
In this way, some companies have been building recharging points in specific
areas and highways in Brazil, but there is still a minimal number of fast recharge
stations. Taking the opportunity for this issue, government subsidies are necessary to
push and incentivize the market, both from the customers’ and companies’ point of
view. Public policies are another topic discussed in this chapter, and it is presented
as a challenge faced in Brazil since EVs are still a new product, and there are a few
subsidies to push the market and stimulate costumers to the conventional vehicles.

FIGURE 4.1 Brazilian electric matrix considering the data for 2018. Data from EPE 2019 [3].
76 Electric Vehicle Integration in a Smart Microgrid Environment

Brazil accounts with 210 million people divided into five regions with different
needs in terms of energy demand, consumption and availability, economy, education,
and environmental problems; besides that, another issue that must be considered is
customers’ acceptance since they surely would be the key for success. Each place has
a specific culture that affects the way people buy and deal with innovation. Moreover,
a new product in the market always comes with many doubts, in which the case of EV
is no different, such as purchasing price, maintenance costs plus its availability, and
energy price. Owing to economic struggles, especially, the EV acquisition represents
several uncertainties for Brazilians.
Indeed, in several countries, EV’s history started in the 19th century with ups
and downs over the centuries and just by the end of the 20th century that it was
stimulated. Some initiatives started to appear in Brazil in this period as the case of
Itaipu Electric and Itaipu e400, which were developed by Gurgel S.A. Nonetheless,
due to high costs and low autonomy, EV came back to the Brazilian market only
over the last years. Aside from many advantages of EVs, there are uncertainties
related to costs, maintenance, and lack of knowledge about the new product, which
is responsible for restricting its introduction in the Brazilian market. Throughout
this chapter, this affair will be discussed and applied for a case study with taxi
drivers in Brazil.

4.3 CHALLENGES AND OPPORTUNITIES

THROUGH THE BRAZILIAN INITIATIVES
Considering the Brazilian portrait, which was briefly presented, it is possible to intro-
duce some challenges and opportunities faced in Brazil about the EV insertion. It
is worth mentioning that to consider all these issues together as an ecosystem, it is
necessary that each aspect, in this case, illustrated as a gear (Figure 4.2), converses
with each other. Any changes made in the production area, for example, could affect
all the chain. The same occurs with the others (Figure 4.2).

4.3.1 economy And production

Considering the EV’s insertion in the Brazilian market, the main challenge faced by
the country is the economic issue. According to Rubens [4], the price is the major
key to reach the customers, followed by maintenance costs and energy prices (fuel/
electricity) [5]. Based on research conducted by Teixeira & Sodré [6], the purchase cost
would influence the purchase decision the most. From this perspective, the Brazilian
context is not favorable to the EV introduction. In 2020, the national average income
is USD 430, and the basic salary is USD 191 (with the currency of 1 USD/5.433 BRL
– 31 April 2020). On the other hand, as we can see in Figure 4.3, the best-selling
vehicle in the country can be found by USD 9,000 (with the currency of 1 USD/5.433
BRL – 31 April 2020), and the cheaper EV on the market is being sold by the amount
of USD 22,000 (with the currency of 1 USD/5.433 BRL – 31 April 2020). Therefore,
the only way people can afford any kind of vehicle is through credit loans or using the
vehicle as a way to generate income, the same way taxi drivers do, aiming to reduce
the payback period [7]. However, even with this possibility, taxi drivers struggle to
Challenges to Build a EV Friendly Ecosystem 77

FIGURE 4.2 Challenges and opportunities approached in this chapter.

maintain an EV in Brazil (Figure 4.3). Additionally, the gasoline price is high in the
country when compared to the basic and average incomes – along with other countries
with a similar economy, which is estimated at about US$ 0.72/liter [8].
This purchase cost – despite being high – is only possible (and it is not higher)
because the country invests in massive production in different states. For example,
companies like FIAT Chrysler, Volkswagen, Chevrolet, Ford, and Honda have

FIGURE 4.3 EVs’ purchase price in Brazil. Data from Angelo 2019 [67].
78 Electric Vehicle Integration in a Smart Microgrid Environment

industrial plants in the country, and they produced, altogether, almost 3 million
vehicles in 2019 [9]. It is worth noting that Brazil does not have any kind of pure EV
production nowadays, which can impact the purchase costs.
Nevertheless, this is only possible because of the incentives they receive to stay in
the country, the labor costs, and the supply chain that is developed around those manu-
facturers. It is worth mentioning that despite the production of vehicles, the government
also invests in fuel production through its state-owned company Petrobras, which is also
responsible for the distribution and part of the gas stations in the country. Understanding
such a structure is a paramount issue because talking about the production is not a strict
conversation on how to manufacture vehicles but also about the whole structure that
needs to be created to guarantee the vehicle’s full operation through the year, i.e.,  it
is important to discuss the production of components, fuel production, and even the
production of asphalt that will interfere with the useful life of the vehicles.

4.3.2 public policies

Since we are talking about production and public incentives, it is crucial to mention
public policies. Over the last decade, the use of EVs was considered a way to hold
the low-carbon technologies and reach sustainability [10]. Many countries have been
promoting the EV market through investments. Due to the quest to reach emission
targets and standards, distinct countries determined goals about the EV introduc-
tion in the market and have encouraged people to use this technology. In the USA,
incentives vary according to the state that can include a subsidy for purchase, home
charge aid, preferential route, free or reduced-price parking, and discounts on licens-
ing [11,12]. The Netherlands [13] and Norway [14] have national policies such as fee
waivers, reduction on the energy prices, and road priority. China has several targeted
policies that affect less than 50% of inhabitants, which cover fee waivers, reduction
on the energy prices, road priority, and access to restricted traffic zones [15,16]. In
another direction, Brazilian public policies and incentives are not enough to boost
the EV market. Figure 4.4 shows a timeline with public policies carried out in Brazil
regarding different types of fuels considering the automotive area.
In Brazil, the government usually offers tax reduction and financial aids to those
brands willing to open factories – which is also usual all over the world. Additionally,
it is usual to offer a tax reduction in the final prices to ensure an increase in the demand
based on price reduction. As can be seen in Figure 4.4, there were different types
of policies in the automotive area to boost low-carbon fuel technologies. However,
comparing EV with other fuel technologies, a small number of public policies cover
legislation and programs to incentivize EVs in Brazil. In this context, public policies
can be presented as a challenge to be faced in the country.
The history with fuel incentives in Brazil started with ethanol, and the addition
of this fuel to gasoline became mandatory, one of the government measures to guar-
antee its competitiveness in the market. In 1975, the National Alcohol Fuel Program
PROALCOOL was launched, which had different phases, and it was responsible for
increasing the number of ethanol plants and establishing the obligation of the etha-
nol-gasoline mixture. Another consequence of this program was the launch of vehi-
cles powered by 100% ethanol to maintain this biofuel in the Brazilian market [17].
Challenges to Build a EV Friendly Ecosystem 79

FIGURE 4.4 Timeline with Brazilian policies considering the automotive approach.

However, the natural gas was introduced in the country from 1970 to 1980 with
National Anticipation Planning for the Gas Production (PLANGAS), which was the
project to increase investment in the fuel supply and develop the market [18]. These
initiatives reduced the investments on the PROALCOOL program and sales of vehi-
cles powered by ethanol fuel. In the same period, the government launched the Air
Pollution Control Program by Motor Vehicles (PROCONVE) to reduce and control
air pollutant emissions [11], and percentage of 22% of ethanol added to gasoline
became mandatory throughout the national territory. Besides that, the PROCONVE
program suggested new technologies to reduce especially nitrogen oxides, sulfur
oxides, and particulate matter emissions [19–21].
Other public policies over the years through the legislation implemented help to
increase the natural gas for automotive purposes, the addition to biodiesel instead
of diesel, and regulation of biomethane as an automotive fuel. The INOVAR-AUTO
program was established in 2012 and lasted until 2017 by the Brazilian government
to stimulate competition and increase efficiency and productivity in the automotive
chain, considering the whole manufacturing, technological, and commercial service
network [22]. The results can be summarized as investments increasing, larger pro-
duction capacity, better automotive energy efficiency through new technologies, and
the development of sustainability through new indicators [23].
The sequence of this program was followed by the 2030 Route Program, which
was created to spread the global insertion of the Brazilian automotive sector, through
the export of vehicles and auto parts. Besides, the program aims to increase com-
petitiveness through technology improvement, reach new standards related to sus-
tainability, and stimulate the research and development (R&D) to help the industry
to reach its goals [24]. This program was the first to look for EVs in Brazil and has
some measures such as the prediction of tax reduction over industrialized products
as the case of EVs, which may reduce the acquisition cost and, thus, encourage their
purchase. Also, the government has regulated the commercialization of energy in
recharging infrastructures [25], which facilitates the provision of this service in dis-
tributors, gas stations, and shopping centers.
80 Electric Vehicle Integration in a Smart Microgrid Environment

Instead of the 2030 Route established over the last years in Brazil, there are still
few policies related to EVs, which represents a challenge for its insertion in the
country. There is a need for other public policies strictly for this type of vehicle
in the same way that PROALCOOL was for ethanol. According to Consoni et al
[26], national efforts to increase the market and improve the possibilities of the EV
implementation are represented by recent actions, mainly in the academic area. At
the same time, some actions toward EV implementation represent an opportunity
for the market. Considering the research scope, the 13,243/2016 National Law [27]
establishes measures to encourage innovation, scientific and technological research
focusing on training, the achievement of technological autonomy, and the develop-
ment of national/regional productive systems in the country. An example of an R&D
project, related to EV, from 2010 is the Smart Grid Brazilian Program funded by
Brazilian Electric Energy Agency (ANEEL). The project proposed, for the Smart
Grid Brazilian Plan, the transition of the current national electricity sector to full
adoption of the smart grid concept throughout the country. Several areas as gover-
nance, energy measurement, distribution automation, and the insertion of EVs were
taken into consideration [6,28,29].
Regarding the production scope, the first large-scale initiatives started in the
1980s, when Gurgel S.A., in partnership with Furnas Centrais Elétricas S.A., devel-
oped two models of purely electric cars – Itaipu Elétrico and Itaipu e400. However,
these models did not obtain market space due to their high costs and low autonomy,
as well as nationalization and oil substitution measures, such as PROALCOOL [30].
Another initiative took place in 2004, when the company Itaipu Binacional and
Kraftwerke Oberhasli AG (KWO), the Swiss hydro controller, signed a technology
cooperation agreement for the design and construction of electric cars. The initiative
was also supported by FIAT, as well as other institutions in Brazil, Paraguay, and
Switzerland [31]. Renault is also a partner of the Itaipu Binacional plant, which facili-
tates and creates favorable conditions for the acquisition of the models. Additionally,
Renault has also delivered several units of the Zoe model 100% electric compact
hatch, which are used in impact studies of the new technology in connection to the
power grid.
Considering the infrastructure scope, the partnership between ANEEL, National
Bank for Economic Development (BNDES), and Technology’s Research and Projects
Financing institution (FINEP) called “Inova Energia” was launched from 2013 to
2016 to hold the smart grid implementation and technological development of solar
and wind energy devices, promote the production of EV components, and improve
its energy efficiency [32,33]. In 2018, ANEEL approved the legislation NR 819/2018
[25], which settled the conditions for public recharging infrastructure to EV, repre-
senting an opportunity to the area, and other companies can be attracted to invest in
Brazil. Through this measure, companies responsible for supply stations can charge
vehicle owners for the electricity, the same way gas stations can charge for the fuel
provided and negotiate energy prices. It is noteworthy that the electricity from EVs
cannot be transferred to the distribution grid according to the Normative Resolution
482/2012 [34], which means that the bi-directional flow of energy as expected in a
smart grid is not allowed yet. Some companies such as EDP and BMW are investing
in EV public infrastructure along highways between the cities of São Paulo and Rio
Challenges to Build a EV Friendly Ecosystem 81

de Janeiro, creating the possibility to use EVs for long distances and solving prob-
lems with autonomy.
In the consumption scope, some municipal regulations contribute to the use of low
carbon vehicles. The 6,545/2017 Act [19] for the city of São Bernardo do Campo (SP)
provides municipal policy to encourage the use of electric, hybrid, and hydrogen-
powered vehicles and other future alternative technologies that do not pollute the
environment according to the municipal 6,163/2011 Act [20]. The incentive corre-
sponds to a return of 25% (twenty-five percent) of the value paid for motor vehicle
property tax (IPVA) [21] considering the limits: first five years of taxation; vehicles
licensed in the city; and vehicles with a value equal or less than an equivalent of USD
31,120 (Exchange Rate – USD/5.433 BRL – 31 April 2020). The city of São Paulo
also has the 15,997/2014 Act, which establishes the municipal policy to encourage
the use of electric/hybrid vehicles and hydrogen-powered vehicles. The incentive
corresponds to a partial refund of the IPVA considering the limits: first five years of
taxation; vehicles licensed in the city; and vehicles with a value equal or less than
an equivalent of USD 27,700 (Exchange Rate – USD/5.433 BRL – 31 April 2020).
Regarding the motor vehicle property tax, seven Brazilian states (Rio Grande do Sul,
Maranhão, Piauí, Ceará, Rio Grande do Norte, Pernambuco, and Sergipe) have its
exemption.
To improve the benefits for users/buyers, it is important to highlight some part-
nerships in Brazilian cities. From 2012 through 2016, Rio de Janeiro and São Paulo
received the Nissan Leaf model as a pilot project to be used in the taxi fleet. In gen-
eral, taxis circulated at airports and could be recharged on quick recharging points at
some specific points. The results of the project showed a reduction of about 13 tons
of CO2 and saving around USD 1,800 (USD/5.433 BRL – 31 April 2020) per year for
each taxi. In Curitiba, the project called “Eco-Elétrico” started in 2014, as a result
of partnerships between Itaipu Binacional, Renault, Nissan, and Portugal’s Center
for Excellence and Innovation in the Automotive Industry, aimed to integrate EVs
into the city public service fleet. In Belo Horizonte – MG, for example, the Toyota
Prius hybrid vehicle has been inserted in the taxi fleet in 2017. This is the result
of a partnership between Belo Horizonte City Council, BHTrans (Belo Horizonte
Transportations and Traffic Company), and the Development Bank of Minas Gerais
(BDMG), which intends to incorporate 600 vehicles into the city fleet. It is worth
mentioning that, in 2017, the government presented the BYD e6 to also be used as a
clean technology alternative for taxi drivers. These initiatives agree with the mobility
plan and are part of the project designed to implement a low carbon mobility system
in Minas Gerais.
Despite the initiatives that arose in Brazil during the last years, there is still much
to be done to ensure that EVs can be used on a large scale. Besides that, the govern-
ment ought to have goals and priorities to increase the possibilities, investments, and
development of public policies regarding EVs to push the sector.

4.3.3 customer AcceptAnce

Another issue to be discussed when we consider a new product is the customer accep-
tance, which shows if the new product could be absorbed by the market and could
82 Electric Vehicle Integration in a Smart Microgrid Environment

contribute to improving the technologies [35], especially in a country with differ-

ent types of regions and cultures, which represent a challenge for the EV adoption.
As previously mentioned, Brazil always faces periods with economic difficulties,
directly impacting the way products are consumed, especially EVs – due to many
issues already discussed, such as the lack of national EV production, aside from the
different types of uncertainties that come with EVs. First and foremost, the lack of
knowledge about EVs – considering the costs, maintenance, life cycle, efficiency,
and autonomy – is an important aspect that justifies low adherence. Additionally,
this issue is connected to another one aforementioned: if the government does not
have any goals set to develop and promote EV technology, the dissemination of the
information becomes compromised. Second, EV prices are extremely high when
compared to conventional vehicle prices as shown in Figure 4.3. Therefore, the popu-
lation has doubts about high investments in an innovative product. Then, autonomy
is always a concern in Brazil since the country is extended and highly dependable of
highways. Last but not least, the lack of public recharging infrastructure is a concern
once it is common to travel for more than 300 km many times per year.
Considering the high EV purchase costs, an option in Brazil is investing in this
type of vehicle for the taxi’s fleet, as presented by some pilot projects. Owing to the
revenue related to this job, the short distances generally traveled per trip, the high
mileage over the years, and intermediate waiting times, this makes taxis the ideal
candidates to be replaced by the EVs. This adoption has been observed in countries
such as China, Korea, and the USA [36–38]. Besides that, the increase in EVs in the
taxi’s fleet contributes to improving the diffusion of electric mobility and encourage
other customers to purchase EVs [39]. Some studies such as the one introduced by
Ozaki and Sevastyanova [40] tell us that the main motivational factors that influence
the purchase of an EV are the government subsidies for these vehicles. According to
Rudolph [41], the likelihood for a driver to exchange his internal combustion engine
vehicle to an EV is 73.4% when we have an increase in the availability of charging
infrastructure and 61.7% when a discount is offered to reduce the EV’s purchase
price. In the study presented by Krause et al. [42], subsidies for the purchase of the
EV would help 82% of the interviewees to consider the purchase of the EV. A study
with taxi drivers developed by Teixeira [43] in Belo Horizonte, Brazil, showed that
70% of the respondents would change the conventional vehicle by an EV, and regard-
ing around 33% of them, it would depend on the purchase cost. Some of these results
will be presented in the case study section.

4.3.4 mArKet, logistics, energy mAtrix, And environment

The first three topics presented in this section cover economic aspects, production,
public policies, and customer acceptance. Although these topics represent many of
the challenges faced in Brazil, the country had a different type of development in
those areas, also representing opportunities to the industry. Considering the topics
listed as opportunities in Figure 4.2, another important issue to be discussed is the
Brazilian market and logistics. As mentioned, the country has almost 2 million km
of road networks mainly used by the logistics sector, which is extremely dependable
of this structure to supply to different regions. It means that if there is a technology
Challenges to Build a EV Friendly Ecosystem 83

with good efficiency and autonomy, low cost, and minimal emissions, this could be
thoroughly applied in Brazil, especially considering ground vehicles.
Brazilian road freight is responsible for about 60% of the food and fuel goods that
circulate throughout the country and is especially important due to the small rail
network in Brazil. According to data from the CNT Annual Report [44], the percent-
age of cargo vehicles comprising trucks, vans, trucks, trailers, and semi-trailers cor-
responds to about 17% of the Brazilian fleet. In 2018, the Brazilian population faced
a strike of the cargo service, having a great impact on the distribution of essential
goods and especially on the fuel’s distribution. This fact was also a watershed which
led to a change in the thought about road transport, promoting research and develop-
ing technologies capable of diversifying the Brazilian logistics and energy matrix.
In a context where freight transport is mostly performed by diesel-powered cargo
vehicles, any service shutdown will always have major impacts. Consequently, the
use of other mobility technologies needs to be analyzed and considered.
Studies regarding buses and electric trucks have been developed in many coun-
tries [5,45–47] as the case of the United States and in the European Union. However,
other studies [29,48] show that impacts on GHG emissions need to be evaluated when
the introduction of EV since the energy to supply this technology is dependent on the
country’s electrical matrix and the life-cycle boundary analyzed. The energy matrix
diversification represents a step to reduce the dependence of any type of source and
increase the possibility to develop new strategies for energy supply, using differ-
ent technologies. As it was mentioned, Brazil has a cleaner energy matrix when
compared to other countries, since the country has more than 80% of energy from
renewable sources. Thus, the low CO2 emission factor from electricity production is
a favorable component for the introduction and use of EVs in the country.
Another point of view considering the use of EV is not only the benefits from
GHG reduction but also the reduction of local air pollution in the cities, which con-
tributes to improving air quality and public health. It is interesting to note that in the
São Paulo State [49] and the city of Rio de Janeiro [50] (Brazil), air pollution had a
significant reduction during the partial lockdown period due to COVID-19 pandemic –
according to the World Health Organization (WHO) [51], premature deaths around
the world can be associated with outdoor air pollution. In city centers, intensive road
traffic contributes to increasing the concentration of pollutants and the number of
deaths due to cardiovascular, respiratory, and lung cancer diseases. Several stud-
ies show the relationship between particulate matter emissions and those diseases
[52–54]. Therefore, the possibility to reduce pollutant concentration can contribute to
improving human health. Hence, the synergy between all the factors already quoted
and the impacts that each one has on the globe should be highlighted.
Another type of business that has been gaining ground in the Brazilian market
and has a high potential of dissemination is car sharing with EVs as the core initia-
tive for this type of service. Some factors as the case of the reduction of the number
of vehicles in the streets and the parking stress [55], the possibility to have a vehicle
for a short period and when it is really necessary [56], and the lack of all fixed costs
(insurance and maintenance, i.e.) associated to the vehicle [57] lead to the success
of this business model. Some government actions and private companies have been
investing in this kind of service in different cities in Brazil such as the city of São
84 Electric Vehicle Integration in a Smart Microgrid Environment

Paulo with the Beep-beep startup, the city of Belo Horizonte with car sharing for taxi
fleet, and in the city of Brasília with car sharing for government approach.

4.3.5 smArt grid

Considering the EV scenario in Brazil, it is also important to highlight the concept
of a smart grid that is being developed in the country with some pilot projects, seek-
ing to incorporate technologies for sensing and monitoring network performance.
The current Brazilian electric grid has a unidirectional flow, which means that the
energy is produced from different sources and then is distributed to the customers,
with no possibility to insert energy back in the system from a distributed genera-
tion, for example. Differently, in a smart grid, the flow of information is bidirec-
tional, which means that there is communication with all areas considering the smart
grid system, helping to increase the economy with loss reduction, improving the
system efficiency, and increasing the credibility and safety for all customers [58]. In
this case, the energy efficiency can be increased due to this whole communication,
and customers can manage the consumption in real-time, being able to trade energy
through microgeneration. In general, the first step to reach the smart grid system is
to change the measurement equipment to monitor consumption, allowing customers
to understand energy spending according to each device.
Many advantages can be observed with smart grid implementation. Considering
the customer’s point of view, there is a cost reduction with different types of rates,
energy efficiency improvement, access to new energy sources, and the possibility of
participating in the market in a different role – as a producer. Considering the deal-
er’s position, there is an increase in operational efficiency, narrowing the relationship
between company and customer, in addition to the optimization of such investments.
These improvements can promote business recovery and increase partnership. From
the government’s perspective, the smart grid system can contribute to cost reduction,
promote new taxes according to the demand, and improve the energy quality and its
respective quality index. For other areas, the new system may increase the research,
develop the productive chain, and create new jobs.
A smart grid system promotes an easier EV introduction in the market consider-
ing its evolution, interaction, and monitoring since it provides capable management
of charging processes. It increases the project’s feasibility by providing a better dis-
tribution of electricity and preventing energy outages. Furthermore, EVs can be used
as energy sources to provide electricity for the grid in periods of the day with high
demand, and this procedure can be managed by the consumer. Particularly in Brazil,
this energy market provided by the smart grid can contribute to enticing the popula-
tion to invest in EVs, reducing the total costs associated with this product. This is the
intersection between both themes.
Despite the issues faced by many countries with smart grid implementation and
this new system, they can be settled as a challenge and an opportunity in Brazil. The
first one is related to its geographic extension, which makes the smart grid imple-
mentation in the whole country an issue and hard to come to fruition equally for all
regions. On the other hand, since there are some pilot projects in Brazil, it could help
the EV insertion in the market. In 2010, as previously mentioned, ANEEL launched
Challenges to Build a EV Friendly Ecosystem 85

an R&D project to carry out the technological migration of the Brazilian current
electric for the full adoption of the concept of smart grids nationally [59]. Some
pilot projects can be highlighted such as the case of the “Cities of the Future” devel-
oped by CEMIG (Minas Gerais State Energy Company) in Sete Lagoas (MG/Brazil),
“Smart Grid Light” developed by Light in Rio de Janeiro (RJ/Brazil), and “InovCity”
developed by EDP Bandeirante, which has been implemented through the installa-
tion of smart meters, cybersecurity, distributed generation from solar energy, EV’s
consideration, and customer awareness [60]. Even though these pilot projects have
been developed over the years, there is still a lack of public policies to hold and regu-
late all these changes in the system [61,62].

4.4 CASE
 STUDY
In order to illustrate the topics previously discussed and display the synergy in an
EV ecosystem, a case study related to the Brazilian scenario will be presented – the
results [43] are gathered in this section, which has two parts: public perception from
the taxi drivers and numeric model. Figure 4.5 shows a general flowchart about the
case study steps.

4.4.1 public perception

The case study was developed in the municipality of Belo Horizonte – the capital of
the second largest Brazilian state (Minas Gerais) [63] – aiming to analyze the possi-
bility to introduce EVs in the taxi fleet in 2017. To begin with, the taxi driver’s profile
was evaluated through an interview process since understanding taxi drivers’ needs
and issues was the starting point of the study.
The interview was performed with 238 respondents, considering all regions in the
municipality. One of the concerns in this research was to evaluate basic knowledge

FIGURE 4.5 Case study flowchart.

86 Electric Vehicle Integration in a Smart Microgrid Environment

about EVs, such as advantages and disadvantages regarding this technology since
it is not totally disseminated in the country. Figure 4.6 shows the number of EVs
licensed in Brazil from 2011 to 2020. As can be seen, the number of EVs is still lower
than conventional, but it has been increasing in the last years, especially considering
2019, in which the number of EVs being licensed has tripled. This fact shows the
potential of EV’s introduction in the country even with a slow process when com-
pared to the rest of the world.
To evaluate the customer acceptance, taxi drivers were invited to answer some
questions regarding the advantages of EVs instead of conventional vehicles. The
level of instruction of the interviewees was requested, enabling to infer the relation
between the level of education and basic knowledge about EVs.
Figure 4.7 shows the relationship between the respondents’ education level and
knowledge about the advantages of the VE. As can be seen, the lack of information
decreases with the increase in schooling. For those with elementary education, about
40% of the respondents do not know any EV advantage. However, these percent-
ages are reduced for those with high school education (about 27%) and even lower
for those with higher education (about 4%). Besides, for those with only primary
education, the savings that EV could bring would be the main advantage. For the
respondents with high school education, the first advantage would be the reduction
of pollutants and air quality improvement, and for those who have higher education,
both pollutant reduction and savings would be the major advantages.
Public policies to promote basic education and increase insertion in higher educa-
tion could expand the access to information, contributing to improving knowledge
about alternative transportation. Therefore, enhancing the information diffusion

FIGURE 4.6 Brazilian EV licensed from 2011 to 2020 (*Data until July 2020). Data from
ANFAVEA 2020 [9].
Challenges to Build a EV Friendly Ecosystem 87

FIGURE 4.7 EVs’ advantages according to the taxi drivers.

about collective and alternative transport models could help the population to have
the opportunity to understand the advantages, whether social, economic, and envi-
ronmental, of using these types of vehicles. Another concern evaluated with the
interviews was the efforts to replace the conventional vehicle by an EV, the results of
which are represented in Figure 4.8. The main concern reported by the taxi drivers
was the purchase cost, which is one factor responsible for preventing vehicle replace-
ment. As previously mentioned, these values are way higher when compared to the
conventional vehicle prices, and the income provided by the taxi service is not suf-
ficient to afford the EV initial costs. Besides, the lack of government incentives and
knowledge was also the most cited factor, by the taxi drivers, as a barrier to change
their vehicles since there are many uncertainties related to infrastructure, electricity
prices, politics, and others.
Figure 4.9 represents the data reported by the taxi drivers as their monthly income
and the costs due to operating and maintaining their vehicles. Considering these
data, 11% of the respondents have a lower income than the costs to maintain the
conventional car, which makes it difficult to change to an electric option. In this
case, it means that taxi drivers are facing losses with the service. Furthermore, 41%
of the respondents have the operation and maintenance costs representing 50% of
their monthly income, which means that the profit from the taxi service is minimal,
restricting the possibility of changing the vehicle. In this context, the importance of
government incentives is mentioned again to support and facilitate EV access, espe-
cially in public transportion (taxis and buses).
88 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 4.8 Challenges about replacement of conventional vehicles by EVs.

FIGURE 4.9 Income and costs associated with operation and maintenance reported by the
taxi drivers.

4.4.2 numeric model

Based on the answers from the interviews, a numeric model was developed using
Matlab software to analyze several aspects that might influence the EV insertion in the
Brazilian market. Different alternative vehicles were compared to the conventional
Challenges to Build a EV Friendly Ecosystem 89

one, the Toyota Etios – which was considered as the basis for the economic analysis
due to taxi drivers’ preferences and its efficiency. For the electric/hybrid vehicles,
three different models were compared: the Nissan Leaf, Toyota Prius, and BYD e6.
These alternative vehicles were chosen since Prius started to be tested in the Belo
Horizonte’s taxi fleet; Leaf is the bestselling EV in the world; and e6 is very com-
mon within taxi drivers in China. The maintenance items were considered according
to the Preventive Maintenance Plan of each type of vehicle, and costs associated
with each one were considered through a market research. Energy costs, whether
with gasoline, ethanol, or electricity, were also considered depending on the type of
vehicle evaluated.
Regarding the environment analysis, it was developed considering the CO2e
(CO2 equivalent) emissions from the fuel/electricity production – Well-to-Tank
(WTT) – and from the vehicle’s operation phase – Tank-to-Wheel (TTW). Both
together bring the Well-to-Wheels analysis (WTW). Emission factors for the
WTT phase were used based on data from SEEG [64], and CO2e emissions from
the ethanol, gasoline, and electricity production in the country were considered.
Figure 4.10 shows the values from the period from 1970 to 2018. It can be seen
that the emission factor from the electricity production has increased in last years
due to the use of energy from thermoelectric power system to support the dry
period, which brings the need for a diversification of the energy/electric matrix.
Brazil has 67% of the electric matrix from hydropower, as mentioned before,
which is directly impacted by the weather. Thus, there is a need for diffusion of
other renewable sources such as wind and solar, since both of them have a huge
potential for expansion in the country.
The inventory data of licensed taxis were provided by Belo Horizonte’s
Transportation and Traffic Company (BHTRANS) [65] and are grouped according to
brand, model, type of fuel, and year of manufacture, which can interfere in the emis-
sion calculation. In Brazil, there is the possibility to choose the fuel between gasoline

FIGURE 4.10 CO2e emission factor from the fuel and electricity production. Data from
SEEG 2018 [64].
90 Electric Vehicle Integration in a Smart Microgrid Environment

and hydrous ethanol, and to consider it in the model, there is a function developed
by Goldemberg [66], which relates the fuel choice and the ratio between hydrated
ethanol and gasoline prices. For the State of Minas Gerais, this ratio is around 70%,
which was considered in the model. Fuel consumptions from the vehicles were used
to calculate the emissions for the vehicles analyzed.
Figure 4.11 shows the costs associated with each vehicle under consideration
for fuel prices, maintenance, purchase price, and taxes. Due to higher gasoline
and ethanol prices compared to the electricity in Brazil, the Etios model showed
the highest percentage (18.79%) of operating costs compared to other vehicles
analyzed. The purchase cost was the most representative (55.19%), and taxes rep-
resent about 18% of total costs, and maintenance accounts for around 7%. For
electric and hybrid vehicles, the cost of purchasing is still the most representative
(around 70% of total costs). Interestingly, for Leaf, e6, and Prius, tax-related costs
(industrialized products tax and licensing fees) are higher than the operating costs
of these vehicles, and they can vary from 4% to 6% of the total costs. In general,
EVs and the hybrid ones have maintenance costs of approximately 1%−3% against
7% of conventional vehicles.
Figure 4.12 shows the emissions for each vehicle per kilometer traveled.
Considering WTT phase, pure EVs (Leaf and e6) have higher CO2e emissions com-
pared to Etios and Prius, due to the higher emission factor from electricity produc-
tion. However, the absolute values are quite similar. On the other hand, in the TTW
phase, Toyota Etios has the higher CO2e emission factor compared to the other vehi-
cles due to the automotive technology and fuel, representing 90% of the total emis-
sions. For the EV side, during the operation phase, there are no emissions. In this
context, it is important to show the importance of the possibility to use EVs to reduce
local emissions, helping to improve the air quality.

FIGURE 4.11 Percentage costs regarding different vehicles.

Challenges to Build a EV Friendly Ecosystem 91

FIGURE 4.12 CO2e emission factor.

4.5 SUMMARY AND CONCLUSIONS

This chapter brought the main topics considering challenges and opportunities to
create a friendly EV ecosystem to facilitate the introduction of EVs in the market and
presented a case study to show this synergy. Some Brazilian actions were shown as
challenges that the country has been facing over the years, which are related to bar-
riers to market diffusion, technologies, economic, environmental, and energy issues
related to this area. Among these, it is possible to quote high purchase costs, lack
of recharging infrastructure, the need for specific public policies to support the EV
introduction, and customer acceptance. From the opportunities side, which are asso-
ciated with the EV introduction, it was mentioned that the market and logistics are
key components and the potential areas, since there is a country dependency on road
networks; energy matrix and the environment, once it is based on renewable sources
and can contribute to reducing local air pollution; and also smart grid system, which
can be considered as both: challenge and opportunity, while Brazil has some pilot
projects, which can help the EV diffusion, although its continental size can make it
difficult for such project to be implemented.
92 Electric Vehicle Integration in a Smart Microgrid Environment

Some measures that help a new product boost in the market are public policies
through specific programs such as the experience mentioned about the ethanol topic
in Brazil. From the production perspective, the national industry can promote EV and
its components with low costs instead of importing them, and it can also increase its
feasibility and the competition with conventional vehicles. From the customer accep-
tance side, another challenge to be faced is the necessity to disseminate correct and
helpful information about EVs since the lack of data represents a barrier to change
the conventional vehicle fleet. Therefore, the idea of starting this dissemination with
heavy-duty drivers is a way to implement EVs from a different perspective, since
the Brazilian road network is extended and have a great potential for investments.
Finally, as long as all aspects of the ecosystem are not taken into consideration simul-
taneously, introducing it in the country might be a harsh reality to be faced on a large
scale, and the same goes for the rest of the world, in which successful initiatives in
different countries can show the synergy between these aspects.

ACKNOWLEDGMENTS
The author gratefully acknowledges the Pontifical Catholic University of Minas Gerais,
Higher Education Improvement Coordination (Coordenação de Aperfeiçoamento de
Pessoal de Nível Superior - CAPES), the University of São Paulo, and support from
SHELL Brazil and São Paulo Research Foundation (Fundação de Amparo à Pesquisa
do Estado de São Paulo – FAPESP), through the Research Centre for Gas Innovation
(RCGI) hosted by the University of São Paulo (FAPESP Grant Proc. 2014/50279-4
and 2019/09242-3).

REFERENCES

Challenges to Build a EV Friendly Ecosystem 93

94 Electric Vehicle Integration in a Smart Microgrid Environment

Challenges to Build a EV Friendly Ecosystem 95

59. ANEEL, “Redes Inteligentes do Brasil,” 2020. [Online]. Available: http://redesinteli-

gentesbrasil.org.br/.

61. G. G. Dranka and P. Ferreira, “Towards a smart grid power system in Brazil: Challenges
and opportunities,” Energy Policy, vol. 136, no. September 2019, 2020.

 5 Coordinated Operation
of Electric Vehicle
Charging and Renewable
Power Generation
Integrated in a Microgrid
Alberto Borghetti, Fabio Napolitano, Camilo
Orozco Corredor, and Fabio Tossani
University of Bologna

CONTENTS
5.1 Introduction ....................................................................................................97
5.2 The Stochastic Optimization Model ............................................................. 100
5.2.1 Model of the Microgrid .................................................................... 100
5.3 Scenarios and Tree Generation Procedure ................................................... 102
5.3.1 Scenario Generation for the V2G Parking Lot ................................. 102
5.3.2 Scenario Generation for the PV Unit and Local Load ..................... 103
5.3.3 Tree Generation by Using k-Means .................................................. 103
5.4 Microgrid Simulation Results ....................................................................... 105
5.4.1 Description of the Case Study .......................................................... 105
5.4.2 Scenario-Based Tree Generation ...................................................... 105
5.4.3 Solution of the Multistage Stochastic Model .................................... 108
5.5 Conclusions ................................................................................................... 114
Acknowledgments.................................................................................................. 116
Nomenclature ......................................................................................................... 116
References .............................................................................................................. 117

5.1 INTRODUCTION
The literature on charging load modelling is becoming quite large, as shown in the
recent survey presented in Ref. [1]. An analysis of the advantages and drawbacks
of different approaches to the integration of electric vehicles (EVs) is presented in
Ref. [2]. Additionally, a study of the state of art of fast-charging stations, including
experimental test, has been introduced in Ref. [3].

97
98 Electric Vehicle Integration in a Smart Microgrid Environment

The current transition to electric mobility and the integration of distributed energy
generation into microgrids both support the employment of bidirectional public
charging stations of plug-in electric vehicles (PEVs). Moreover, the appearance of
clusters of rapid charging stations in parking lots has propitiated the development of
mainly two types of interactions between PEVs and the power distribution network,
according to their dispatchability: grid-to-vehicle (G2V) charging stations, in which
PEVs arrive in the stations to be charged as fast as possible, and vehicle-to-grid
(V2G) scenarios, which represent typical parking lots where PEVs stay for a signifi-
cantly long time. For the first scenario (G2V), one of the key aspects is the queuing
model that can represent the arrival and departure times of EVs, as illustrated, e.g.,
in Ref. [4] and reference therein. For the second scenario, the V2G parking lot can
be operated so as to achieve either or both of the two typical objectives of storage
systems: load flattening and balancing of renewable generation [5].
Typically, in the V2G scenario, the owner of the charging facility aims to obtain
economic benefits for the services, while offering to the PEV owner the option to
charge the vehicle at the lowest possible cost. To achieve these objectives, smart
charging approaches are implemented to align the charging and dispatching pro-
cesses of storage systems with the optimization objectives. For example, Ref. [6]
propose an optimization model for the assessment of the contribution of V2G sys-
tems. The method described in Ref. [7] addresses the integrated operational planning
of a distribution system, by using an aggregator conceived as an intermediate agent
between end-users and distribution system operators (DSOs).
The integrated operation of parking facilities with renewable energy resources
(RESs) has been extensively studied, as shown in Ref., e.g., [8]. An evaluation of the
integration of PEVs with photovoltaic (PV) systems, in order to cope with the fluctua-
tion of solar irradiance, has been performed in Ref. [9]. An additional approach that
takes into account the uncertainties of PEV’s arrival and grid power price has been
presented in Ref. [10]. A generation scheduling method for the coordinated operation
of an industrial microgrid, which considers electricity and heat generation, electrical
loads, PV units, and PEVs, is presented in Ref. [11].
Among the various schemes typically adopted for regulating the participation of
distributed energy resources (DERs), fixed or time-of-use (TOU) tariffs, designed
by retailers or DSOs, are often adopted. A study of the feasibility of premium tariff
rates for V2G services similar to feed-in-tariff (FIT) programs for RES has been
presented in Ref. [12].
This chapter focuses on the operation of the parking lot equipped with bidirec-
tional charging stations inside the energy management system of a microgrid. A
typical scenario corresponding to the integration of G2V and V2G services, bat-
tery energy storage (BES) systems, and DERs in a microgrid site is illustrated in
Figure 5.1. In the represented case, a central dispatching system solves the optimiza-
tion problem that minimizes the energy procurement costs of the considered site.
As an example, the model described in this chapter represents the day-ahead opti-
mization of the global charging and discharging of the batteries of the EVs connected
to the charging stations in order to reduce the energy procurement cost. The optimi-
zation problem is solved by the dispatching centre of the microgrid. The optimiza-
tion model is based on the application of stochastic linear programming, where the
EV Charging and Renewable Power Generation 99

FIGURE 5.1 Scheme of the microgrid with both local generation and EV charging stations.

uncertainty associated with the operation of the system during the day is modelled
by stochastic processes.
The contributions of this chapter are:

• The presentation of an approach for the day-ahead scheduling of a microgrid,

including a parking lot, that facilitates the integration of the production
from renewable resources;
• The presentation of a scenario-based approach to describe the uncertainties
associated with the departures and arrivals of the EVs, as well as with the
power production from renewable energy and load profile;
• The development of a clustering procedure that allows the solution of the
scenario-based multistage optimization model with a reasonable computa-
tional effort.

The following sections describe the tree generation procedure for the case of a
microgrid that includes both distributed generation (e.g., PV units) and a parking
lot equipped with charging stations for PEVs with V2G technology. For this pur-
pose, we make reference to the linear programming model presented in Ref. [13],
although the procedure can be suitably adapted to be used with different stochastic
models for the optimal operation of EV’s parking lots in microgrids (e.g., those
100 Electric Vehicle Integration in a Smart Microgrid Environment

presented in Refs. [6] and [14]) and in power distribution systems (e.g., Refs. [15]
and [16]).
A peculiar characteristic of the linear programming model presented in Ref. [13]
is the representation of specific operating rules relevant to the initial energy avail-
able in the cars entering the parking lot. While in Ref. [13] a two-stage stochastic
programming model has been adopted, this chapter describes the extension of the
model to the multistage case, following the approach of Ref. [17]. For such a purpose,
a scenario tree is needed to define each scenario as a root-leaf path through all the
stages, representing the conditional probabilities with links between nodes. Each
path corresponds to a set of values of stochastic parameters with a specific history up
to the relevant node.
The structure of this chapter is as follows: Section 5.2 reviews the main
characteristics of the stochastic optimization model of the parking lot within the
microgrid. Section 5.3 describes the generation of the scenarios and the relevant
tree implemented in the AIMMS modelling environment. Section 5.4 presents some
optimization results. Section 5.5 concludes this chapter.

5.2 THE STOCHASTIC OPTIMIZATION MODEL

This section describes the specific linear programming model applied to the park-
ing lot with EV’s charging stations. In order to perform the analysis of the model,
we consider the day-ahead optimization problem of the dispatching centre of a
microgrid that is aimed at reducing the total cost for electric energy procurement.
The microgrid is connected to the external utility grid and includes the EV’s parking
lot with V2G technology, as well as a PV unit and local loads.

5.2.1 model oF the microgrid

In the multistage stochastic optimization model, each scenario ω is represented by
the path that connects the consecutive nodes from the root of the tree to the relative
leaf node. All the variables associated with the scenarios that pass through the same
intermediate node are identical. In general, a stage consists of multiple time periods
t; therefore, all the variables associated with the periods of the same stage coincide.
The optimization problem considered by the dispatching centre is represented by
(5.1), where we consider a day-ahead optimization horizon of 24 hours:

min ∑π C
ω
ω ω (5.1)

with

Cω = ∑  ρ
t
T
t ( E ω − rE ω ) + C ω .
−
t
+
t t
S
(5.2)

For the V2G parking lot, the decision variables Pt ω and µt ω are defined as follows:
Pt ω is the power exchange of the V2G parking lot calculated for each node of the
scenario tree (positive if provided to the grid), and µt ω is the nonnegative utilization
EV Charging and Renewable Power Generation 101

coefficient of the total energy EtS+

ω initially stored in the EVs that arrive at time t in
the parking lot.
The constraints that represent the behaviour of the V2G parking lot are:

Pt min
ω ≤ Pt ω ≤ Pt max
ω (5.3)

EtSω = (1 − δ ) E(St −1)ω − Pt ω ∆t + EtSω+ − EtSω− (5.4)

S max
EtSω ≤ Et ω (5.5)

(E S+
− emin N tinω Enom
EV
) ⋅u
0 ≤ µt ω ≤
tω

EtSω+
(E S+
tω − emin N tinω Enom
EV
) (5.6)

EtV2G
ω = (1 − δ ) E(V2G S+
t −1) ω − Pt ω ∆t + µt ω Et ω + ∑ (1 − µ
j=0
tω ) E(µj ,t )ω − EtSω− (5.7)

(1 − η r ) Pt ω

lt ω ≥  1  (5.8)
 1 − P
s  tω
 1 − η 

cr Pt ω ∆t + c µ µt ω EtSω+

C S
tω ≥ . (5.9)
−c s Pt ω ∆t + c µ µt ω EtSω+

Constraint (5.3) limits the maximum and minimum power exchange of the V2G
parking lot.
The energy stored by each battery at the end of period t is determined by (5.4), and
constraint (5.5) limits the maximum value of the storage capability.
When EVs reach the parking lot, their energy adds to the total energy of the
V2G parking lot EtSω . Constraint (5.6) limits the maximum value of µt ω so the ini-
tial charge of the battery may be used only for the amount exceeding a predefined
EV
minimum fraction (emin) of the rated energy size Enom . This constrain avoids deep
discharge conditions. In (5.6), function u(·) represents the step function with value 0
for negative arguments and 1 for positive arguments.
Constraint (5.7) is aimed to calculate the available energy for dispatching services
EtV2G
ω . For such purpose, the dispatching centre determines the exploitation factor
µt ω of the total energy EtSω . This scheme permits the implementation of the battery-
to-battery charging strategy in a V2G system. The total cost CtSω considers the cost
of this retrieval according to specific price (i.e., c µ ). Such a cost is considered at the
102 Electric Vehicle Integration in a Smart Microgrid Environment

arriving time of the EVs, while the associated energy can be retrieved in all of the
periods during the parking time.
Disregarding network power losses, the energy balance in each period of the
microgrid is

∑P
u
ut ω ∆t = Et+ω − Et−ω (5.10)

The net power of the V2G parking lot, calculated considering the losses, is given by

Put ω = Pt ω − lt ω (5.11)

Both lt ω and CtSω (nonnegative variables according to equations 5.8 and 5.9, respec-
tively) are minimized as a result of the minimization of the objective function in
equation 5.1.
According to this model, the power values associated with the PV unit and the
local loads cannot be dispatched.

5.3 SCENARIOS AND TREE GENERATION PROCEDURE

To build the scenario tree described in the Introduction, a set of equiprobable sce-
narios is at first generated. This section deals with the procedure for the scenario
generation of the V2G parking lot. For completeness, this section of this chapter will
also describe the scenario generation of the PV unit output power and the local loads.
Once the scenarios have been generated, the scenario tree is obtained by a routine
based on a recursive k-means clustering procedure applied to the total set of initial
scenarios, with the scenario reduction technique described in Ref. [18]. The relevant
parameters considered by the procedure of aggregation are the number of parked
EVs, PV generation, and the non-dispatchable load in the microgrid.

5.3.1 scenArio generAtion For the v2g pArKing lot

The scenarios for the V2G parking lot are generated as follows. For each scenario ω,
a population of Ntot EVs willing to enter in the parking lot is generated according to
chosen statistical distributions (e.g., a normal distribution with the mean value equal
to the forecasted number of potential users of the parking lot). In general, not all the
Ntot EVs can be connected to a charging station if all the charging stations available
in the parking lot are engaged.
Each i-th EV that is allowed to be recharged is characterized by the parameters ti+ ,
0
Ei , and si . In the simulation example reported in this chapter for illustrative purposes,
also these parameters are generated following normal distributions with the mean
value equal to the forecasted value. Although the rated capacity of the batteries is
considered constant and equal for all the EVs, it could also be assumed as a variable.
The energy stored Ei− is equal to 1 pu if the time of stay si is long enough to get
a full recharge; otherwise, it is set to ratio between si and the time needed for a full
recharge.
EV Charging and Renewable Power Generation 103

For each time t and scenario ω, the procedure builds two sets: St+ω that consider the
EVs incoming at time t and St−ω that consider the EVs leaving at time t.
On the basis of these sets, the procedure calculates the parameters needed in
equations 5.3–5.9: the increase in the energy stored in the parking lot due to arrivals
( )
ω = ∑ i∈St+ω Ei , the decrease in the energy stored in the parking lot due
at time t EtS+ 0

to departures at time t E ( S−
tω )
= ∑i∈St−ω Ei− , and the decrease in the energy stored in the
parking lot at time t due to the initial charge of the EVs entered at time j and leaving
(
at t E(µj ,t )ω = ∑i∈ S + ∩S − Ei0 .
{ jω tω } )
5.3.2 scenArio generAtion For the pv unit And locAl loAd
The scenarios of the PV output and the local loads are generated on the basis of the
day-ahead forecasts and the probability distributions that characterize the expected
deviations with respect to the forecasts.
For the PV units, the scenarios are generated by using the procedure described in
Ref. [19] based on a first-order autoregressive Markov process that provides a time
series xt ω that represents the one-lag autocorrelation

xt ω = φ x(t −1)ω + ε , (5.12)

where ε is the white noise with zero mean and standard deviation σ = 1 − φ 2 . In
order to obtain the generation of the PV outputs Pt PVω , xt ω is added to the normalized
profile of the forecast, and then the probability transformation, which includes the
cumulative density functions of both the normalized time series and the forecasted
profile, is applied.
In order to avoid unrealistic scenarios, all the Pt PV
ω profiles do not differ by more
than 10% from the forecasted profile for at least 90% of the time.
In the numerical test presented in this chapter, the profiles of the local loads,
besides the EVs at the unidirectional charging stations, are obtained by multiplying
the forecasted profile by 1 + kt , where kt is generated by using a normal distribution
with a mean value equal to 0 and standard deviation σ = 1 − ψ ( t )2 , where ψ ( t ) is a
decreasing function of t in order to represent the increase of the incertitude with time.

5.3.3 tree generAtion by using k-meAns

As described in Ref. [18], after the generation of the set of initial scenarios, the appli-
cation of the k-means clustering method allows to build the scenario tree. For such
a purpose, the 24-hour optimization horizon is divided into four stages of six hours
each. Figure 5.2 illustrates the main steps of the implemented method.
Each scenario ξt ω corresponds to a realization of the stochastic parameters of
number of EVs in the parking lot, PV generation, and total load (i.e., considering
local load and charge of EVs at the charge station) in the form:

ξt ω =  N tEV PV Load
ω , Pt ω , Pt ω  (5.13)
104 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 5.2 Implemented k-means algorithm.

At stage s = 1 (i.e., the very first six hours of the day), all scenarios have the same
value of parameter, i.e., ξt ω = ξt ω .
At stage s = 2 (i.e., from 7 am to 12 pm), the set of scenarios is aggregated in
a number K of desired groups represented by centroid Cks ; for the numerical tests
included in this chapter, a number of three clusters have been predefined (K = 3).
Firstly, the initial K cluster centres are selected (i.e., centroids); the first cluster cen-
tre C1s is randomly chosen from the relevant set of scenarios in the stage. Then, the
following two initial cluster centres, i.e., C2s and C3s , correspond to the farthest two
scenarios ξt ω obtained, on the basis of the Euclidean distance.
Then, in order to identify the closest centroid Cks to the relevant scenario ξt ω , dis-
tance d is calculated as

( )
d ξt ω , Cks = ∑ξ
t ∈Ts
tω − Cks ∀k = 1,..., K , (5.14)

where 2 indicates the Euclidean distance, and the individual scenario is assigned to
the cluster with the corresponding minimal distance.
Next, the centroid of each cluster Cks is updated as the mean value of all the sce-
narios aggregated in each cluster. This process is iteratively carried on, until there are
not changes in the centroid values.
EV Charging and Renewable Power Generation 105

Starting from an original set of equiprobable scenarios at the considered stage s,

the probability of each cluster π sk corresponds to the total sum of the probabilities
of the scenarios aggregated in the same cluster.
The profiles of all the scenarios belonging to the same cluster Cks are replaced by
the profile of the relevant centroid, i.e., ξt ω = ξt ω ∀t ∈ Ts if ξtsω ∈Ωsk .
At the subsequent stages, the k-means clustering routine is carried on for each
centroid k obtained in the previous stage.
Finally, the implemented k-means clustering algorithm delivers the scenario tree
that represents the realization of the considered stochastic model, while considering
the relevant non-anticipativity constraints.
At each stage defined, each scenario is grouped to one of the centroids of the
algorithm on the basis of the average value of the number of parked EVs, PV output,
and total non-dispatchable load in the six hours.
All the values of the parameters of the scenarios grouped together are averaged
in order to assign a unique value to each of them for each hour of the stage. At
each period j, the values of E(µj ,t )ω of all the scenarios in the same group are aver-
aged for all t > j. Moreover, a matrix that contains the number of EVs that arrive in
period j and leave in period t is defined and averaged as done for E(µj ,t )ω . In order
to guarantee the existence of a feasible solution of the optimization problem of the
V2G parking lot, this matrix is used to define the average values of the number of
parked EVs, and therefore of Et ω , while E(µj ,t )ω is used to define the average values
S max

of EtS+ S−
ω and Et ω .

5.4 MICROGRID SIMULATION RESULTS

5.4.1 description oF the cAse study
The case study refers to a microgrid composed by a parking lot with 100 bidirec-
tional charging points - each with 7kW rated power -, by a 3.5MWp-PV system, local
loads up to 3 MW and a connection point to the external utility grid.
In the presented simulation results, the parking lot is empty at time t = 0, the
energy capacity of the EV’s batteries is 24 kWh, efficiencies η r and η s are equal to
0.96, c µ is 1.8 €/MWh, the minimum energy value emin is equal to 0.2 pu, price ρtT
is equal to 72.39 €/MWh from 7 am to 11 pm and is equal to 51.62 €/MWh in the
other hours, ratio r between selling price and buying price of electric energy is equal
to 0.5, and δ is neglected.
The stochastic linear programming model has been implemented in the AIMMS
development environment and tested by using the Cplex solver version 12.9 on
a PC equipped with a 2-GHz Intel-i7vPro with 8 GB of RAM, running 64-bit
Windows 10.

5.4.2 scenArio-bAsed tree generAtion

The number of generated scenarios, before the grouping procedure is applied, is 60
for all the stochastic variables of the microgrid.
 106 Electric Vehicle Integration in a Smart Microgrid Environment

As mentioned, the stochastic events for the parking lot are obtained starting from
the random generation of N Tot, ti+ , Ei0, and si according to normal distribution func-
tions with the mean values and the standard deviations as reported in Table 5.1.
Figure 5.3 shows the 60 profiles of N tEV ω .
The profiles of the PV output are generated assuming ϕ = 0.999 and are shown
in Figure 5.4. The load profiles are generated by assuming ψ linearly decreasing
from 1 at t =1 to 0.99 at t = 24. Figure 5.5 shows the profiles of the total load of the
microgrid.
Figure 5.6 shows the scenario tree with 26 leaf nodes obtained by the implemented
k-means, with the indication of the probability associated with each leaf node.

TABLE 5.1
Parameters of the Normal Distribution Functions for the V2G Scenario
Generation
Population Parameters Mean Value Standard Deviation
Time of entrance t 9 2
Initial charge (pu) Ei0 0.3 0.3
Time of stay si 8 2
Number of entrances N Tot 100 10

100

80
number of vehicles

60

40

20

0
0 6 12 18 24
time (hour)

FIGURE 5.3 Number of parked EVs in the parking lot.

EV Charging and Renewable Power Generation 107

3.5

2.5
power (MW)

1.5

0.5

0
0 6 12 18 24
time (hour)

FIGURE 5.4 PV output profiles

2.4

2
power (MW)

1.6

1.2

0.8

0.4
0 6 12 18 24
time (hour)

FIGURE 5.5 Total load profiles

Figures 5.7 to 5.11 illustrate the 26 aggregated scenarios of Figure 5.6: Figure 5.7
shows the number of parked EV’s N tEV S+ S−
ω , Figure 5.8 shows Et ω , Figure 5.9 shows Et ω ,
Figure 5.10 shows the PV power outputs, and Figure 5.11 shows the local load
profiles.
108 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 5.6 Multistage scenario tree with the indication of the probabilities associated
with each leaf node.

5.4.3 solution oF the multistAge stochAstic model

The optimization procedures have been implemented in AIMMS Developer model-
ling environment and solved in around 15 seconds.
Figure 5.12 shows the dispatched power of the V2G parking lot. The results of
Figure 5.12 can be compared with the power profiles at the EV’s unidirectional
charging station as shown in Figure 5.13, which correspond to an uncontrolled charge
of the EVs for each of the 60 scenarios before grouping. In particular, Figure 5.12
illustrates the effect of the energy stored during the hours of maximum production
of the PV unit and the power contribution of the parking lot during the evening load-
peak hours.
Figure 5.14 shows the profile of µt ω for the scenario in the tree of Figure 5.6. The
initial energy of the incoming EVs is used in the early morning in order to increase
EV Charging and Renewable Power Generation 109

100

80
number of vehicles

60

40

20

0
0 6 12 18 24
time (hour)

FIGURE 5.7 Selected scenarios of number of parked EV’s N tEV

ω .

0.35

0.3

0.25
energy (MWh)

0.2

0.15

0.1

0.05

0
0 6 12 18 24
time (hour)

FIGURE 5.8 Selected scenarios of total energy EtS+

ω entering the parking lot due to EV’s
arrivals.

the storage of the PV output in the parking lot during the following hours (taking
into account the low value of the ratio between sale and purchase utility tariffs) and
in late afternoon although the effect is limited due to the few EVs entering in those
hours.
110 Electric Vehicle Integration in a Smart Microgrid Environment

0.5

0.4
energy (MWh)

0.3

0.2

0.1

0
0 6 12 18 24
time (hour)

FIGURE 5.9 Selected scenarios of total energy EtS−

ω leaving the parking lot due to EV’s
departures.

3.5

2.5
power (MW)

1.5

0.5

0
0 6 12 18 24
time (hour)

FIGURE 5.10 Selected scenarios of power delivered by the photovoltaic system.

Figure 5.15 shows the total energy stored in the parking lot, and Figure 5.16 shows
the power exchange between the microgrid and the external utility grid.
The value of stochastic solution (VSS) and the expected value of perfect informa-
tion (EVPI) are typical performance metrics. According to, e.g., Ref. [20], the defini-
tions of VSS and EVPI are as follows:
EV Charging and Renewable Power Generation 111

power (MW) 2

1.5

0.5
0 6 12 18 24
time (hour)

FIGURE 5.11 Selected scenarios of power consumptions of the local loads of the microgrid.

0.6

0.4

0.2
power (MW)

-0.2

-0.4

-0.6

-0.8
0 6 12 18 24
time (hour)

FIGURE 5.12 Power delivered by the parking lot calculated by the stochastic optimization
model in the selected scenarios.

• VSS is the difference between EEV and RP, where EEV is the expected
value solution and RP is the solution of the recourse problem, i.e., the cost
value of the multistage stochastic problem (5.1). In order to calculate EEV, at
first, the values of the decision variables for each time period t are obtained
by the solution of the deterministic model in which all random variables are
112 Electric Vehicle Integration in a Smart Microgrid Environment

0.7

0.6

0.5
power (MW)

0.4

0.3

0.2

0.1

0
0 6 12 18 24
time (hour)

FIGURE 5.13 Uncontrolled charge at the EV’s charging station for the 60 initial scenarios.

0.6
utilization coefficient

0.4

0.2

0
0 6 12 18 24
time (hour)

FIGURE 5.14 Utilization coefficient calculated by the stochastic optimization model in the
selected scenarios (emin = 0.2 pu).

replaced by their expected values; then, EEV is the solution of the stochastic
problem in which the decision variables are fixed parameters.
• EVPI is the difference between RP and the wait-and-see (WS) solution given
by the calculation of the expected value of the set of deterministic solutions,
each relevant to one of the tree scenarios.
EV Charging and Renewable Power Generation 113

2.5

2
energy (MWh)

1.5

0.5

0
0 6 12 18 24
time (hour)

FIGURE 5.15  Energy stored in the parking lot.

2.00
power (MW)

0.50

-1.00

-2.50
0 6 12 18 24
time (hour)

FIGURE 5.16  Power exchanged between the microgrid and the external grid (positive if
exported from the microgrid).

The VSS and EVPI calculated for the considered case studies are presented in Table 5.2.
For a further comparison, the value of the objective function obtained for the
deterministic problem that uses as input the average values of the stochastic param-
eters N tEV PV Load
ω , Pt ω , and Pt ω is € 907.55. The average value of the deterministic solution
114 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 5.2
Performance Evaluation for the Multistage Stochastic Solution
Solution Cost Value (€) VSS (€) EVPI (€)
EEV 870.79 6.04 0.05
RP 864.74
WS 864.69

TABLE 5.3 
Objective Function of the Stochastic Solution for Different Values of emin
emin 0.1 pu 0.15 pu 0.2 pu 0.3 pu 0.4 pu 0.5 pu
RP (€) 864.38 864.52 864.74 865.46 867.29 867.30

for an initial set of 60 scenarios (which does not provide a unique profile of charge,
discharge decision, and utilization coefficient µt ω ) is € 851.62.
In view of these results, the adoption of the multistage stochastic model is expected
to allow a significant improved flexibility and cost reduction for many scenarios with
respect to the adoption of the solution provided by the deterministic model.
We compare now the RP cost values obtained for different values of the minimum
energy value emin, which is used in equation 5.6 in order to limit utilization coefficient
µt ω of the initial charge of the vehicle arriving in the parking lot. As expected, Table
5.3 shows that larger values of emin cause the increase in the objective function values.
Furthermore, Figure 5.17 shows the effect of the different emin values on the profile of
utilization coefficient µt ω . For emin values equal and exceeding 0.5 pu, the resulting
µt ω is always 0.

5.5 CONCLUSIONS
This chapter deals with the operation of a microgrid with a parking lot that allows
bidirectional charging services. The optimization problem minimizes the daily cost
of the microgrid by means of a multistage stochastic problem. For such a purpose,
the uncertainties associated with the number of parked EVs, PV generation, and non-
dispatchable loads in the microgrid are represented by the relevant scenario tree.
The employed k-means clustering procedure for the scenario tree generation pro-
vides appropriate results even with a limited number of cluster centres. In order to
evaluate the performance of the multistage stochastic optimization, VSS and EVPI
have been calculated, confirming the advantage of the multistage scenario-based
approach over the solution given by a deterministic model based only on the forecast
of the stochastic parameters. In addition, the computational effort is reasonable for
the considered four-stage stochastic problem.
The use of the optimization model has been illustrated for the day-ahead dispatch
of an industrial microgrid with a significant production from renewable resources,
but is expected to be useful also to study the services of V2G parking lots in other
EV Charging and Renewable Power Generation 115

(a) 0.8

0.6

utilization coefficient
0.4

0.2

0
0 6 12 18 24
time (hour)

(b) 0.8

0.6
utilization coefficient

0.4

0.2

0
0 6 12 18 24
time (hour)

(c) 0.4

0.3
utilization coefficient

0.2

0.1

0
0 6 12 18 24
time (hour)

(d) 0.2
utilization coefficient

0.1

0
0 6 12 18 24
time (hour)

FIGURE 5.17 Profiles of utilization coefficient µt ω calculated by the stochastic optimiza-

tion model for different values of emin: (a) 0.1 pu; (b) 0.15 pu; (c) 0.3 pu; (d) 0.4 pu.
116 Electric Vehicle Integration in a Smart Microgrid Environment

context (e.g., the minimization of the procurement energy costs of a single or multiple
EV parking lots and the congestion reduction analysis in urban distribution networks).

ACKNOWLEDGMENTS
This work was funded by the Italian Ministry for Education, University and Research
under the grant PRIN-2017K4JZEE “Planning and flexible operation of micro-grids
with generation, storage and demand control as a support to sustainable and efficient
electrical power systems: regulatory aspects, modelling and experimental validation”
and by Regione Emilia-Romagna through the European Regional Development Fund
2014–2020 - “Energynius - Energy networks integration for urban systems” project (CUP
E31F18001040007). These activities have been carried out in collaboration with Saeed
Rahmani Dabbagh and Mohammad Kazem Sheikh-El-Eslami (co-authors of the papers
that present some of the obtained results). The authors are indebted to Sara Chiriatti and
Paolo Prevedello for their help in performing the simulations. The comments and sugges-
tions provided by Prof. Carlo Alberto Nucci are gratefully acknowledged.

NOMENCLATURE
Symbols relevant to the optimization model of the microgrid:

πω : Probability associated with scenario ω

Cω : Cost associated with scenario ω
ρtT : Time-of-use tariff for purchasing from the grid in period t
r:  Ratio between sale and purchase tariffs
Et+ω and Et−ω : Nonnegative variables that correspond to the hourly energy sold
and purchased from the utility grid
Pt ω : power exchange of the parking with the external provider
Pt max min
ω and Pt ω : available total maximum power for discharging and charging pro-
cesses of the parking lot, respectively, which depend on the total set
of parked EVs.
∆t: duration of period t
EtSω : energy stored in the parking lot
S max
Et ω : maximum storage capability that depends on the number of parked EVs
EV
Enom : rated energy size of the EV batteries
N tinω : number of EV’s arrivals
EtV2G
ω : energy available for vehicle-to-grid (V2G) services
S−
Et ω : decrease in the energy stored in the parking lot due to EV’s departures
EtSω+: increase in the energy stored in the parking lot due to EV’s arrivals
µt ω : is the nonnegative utilization coefficient of the total energy EtSω+ ini-
tially stored in the EVs that arrive at time t in the parking lot
E(µj ,t )ω : the initial energy of the EVs arriving in period j and leaving in time
period t (j = 0 indicates the EVs already parked when the optimiza-
tion horizon begins)
lt ω : charging/discharging power losses
η r and η s :  average efficiencies of the EV batteries and charging stations dur-
ing discharging and charging processes, respectively
EV Charging and Renewable Power Generation 117

δ : self-discharging rate
CtSω : total cost associated with the use of the V2G parking lot accord-
ing to prices cr and c s of the discharging and charging processes,
respectively, and to the price c µ of the utilization of EtSω+
Put ω : net active power provided or absorbed by unit u (positive if pro-
vided to the microgrid), i.e., in our case, the parking lot, the PV
unit, and the local loads

Symbols relevant to the scenario generation for parking lot:

Ntot:  population of EVs willing to enter in the parking lot

ti+ : time of entrance of the ith EV
Ei0: initial charge of the ith EV
si : time of stay of the ith EV
Ei− : charge at departure of the ith EV
St+ω : set of EVs incoming at time t in scenario ω
St−ω : set of EVs leaving at time t in scenario ω

Symbols relevant to the scenario generation for the PV unit and local load:

xt ω : time series that represents the one-lag autocorrelation

φ : autocorrelation parameter
ε : additive white noise employed for the scenario generation technique

Symbols relevant to the scenario tree generation:

Ω: set of scenarios ξt ω
ω : number of EVs in the parking lot in period t and scenario ω
N tEV
Pt ω : photovoltaic generation in period t and scenario ω
PV

Pt Load
ω :  total load (i.e., considering local load and charge of EVs at the
charge station) in period t and scenario ω
Cks : set of K centroid clusters at stage s
Ts:  set of periods in stage s
ξt ω : mean value of all the scenarios ξt ω assigned to the same cluster for
time t.

REFERENCES
1. Y. Xiang, S. Hu, Y. Liu, X. Zhang, and J. Liu, “Electric vehicles in smart grid: A sur-
vey on charging load modelling,” IET Smart Grid, vol. 2, no. 1, pp. 25–33, 2019, doi:
10.1049/iet-stg.2018.0053.
2. J. García-Villalobos, I. Zamora, J. I. San Martín, F. J. Asensio, and V. Aperribay, “Plug-in
electric vehicles in electric distribution networks: A review of smart charging approaches,”
Renew. Sustain. Energy Rev., vol. 38, pp. 717–731, Oct. 2014, doi: 10.1016/j.rser.2014.07.040.
3. D. Sbordone, I. Bertini, B. Di Pietra, M. C. Falvo, A. Genovese, and L. Martirano, “EV
fast charging stations and energy storage technologies: A real implementation in the
smart micro grid paradigm,” Electr. Power Syst. Res., vol. 120, pp. 96–108, Mar. 2015,
doi: 10.1016/j.epsr.2014.07.033.
118 Electric Vehicle Integration in a Smart Microgrid Environment

4. F. Varshosaz, M. Moazzami, B. Fani, and P. Siano, “Day-ahead capacity estimation and

power management of a charging station based on queuing theory,” IEEE Trans. Ind.
Informatics, vol. 15, no. 10, pp. 5561–5574, 2019, doi: 10.1109/TII.2019.2906650.
5. C. Develder, M. Strobbe, K. De Craemer, and G. Deconinck, “Charging electric vehi-
cles in the smart grid,” Smart Grids from a Global Perspective, pp. 235–248, 2016.
6. C. Battistelli, L. Baringo, and A. J. Conejo, “Optimal energy management of small
electric energy systems including V2G facilities and renewable energy sources,” Electr.
Power Syst. Res., vol. 92, pp. 50–59, 2012, doi: 10.1016/j.epsr.2012.06.002.
7. A. Zakariazadeh, S. Jadid, and P. Siano, “Integrated operation of electric vehicles and
renewable generation in a smart distribution system,” Energy Convers. Manag., vol. 89,
pp. 99–110, Jan. 2015, doi: 10.1016/j.enconman.2014.09.062.
8. F. Mwasilu, J. J. Justo, E.-K. Kim, T. D. Do, and J.-W. Jung, “Electric vehicles and smart grid
interaction: A review on vehicle to grid and renewable energy sources integration,” Renew.
Sustain. Energy Rev., vol. 34, pp. 501–516, Jun. 2014, doi: 10.1016/j.rser.2014.03.031.
9. J. Traube et al., “Mitigation of solar irradiance intermittency in photovoltaic power
systems with integrated electric-vehicle charging functionality,” IEEE Trans. Power
Electron., vol. 28, no. 6, pp. 3058–3067, 2013, doi: 10.1109/TPEL.2012.2217354.
10. T. Zhang, W. Chen, Z. Han, and Z. Cao, “Charging scheduling of electric vehicles with
local renewable energy under uncertain electric vehicle arrival and grid power price,” IEEE
Trans. Veh. Technol., vol. 63, no. 6, pp. 2600–2612, 2014, doi: 10.1109/TVT.2013.2295591.
11. S. Y. Derakhshandeh, A. S. Masoum, S. Deilami, M. A. S. Masoum, and M. E.
Hamedani Golshan, “Coordination of generation scheduling with PEVs charging in
industrial microgrids,” IEEE Trans. Power Syst., vol. 28, no. 3, pp. 3451–3461, Aug.
2013, doi: 10.1109/TPWRS.2013.2257184.
12. D. B. Richardson, “Encouraging vehicle-to-grid (V2G) participation through premium tariff
rates,” J. Power Sources, vol. 243, pp. 219–224, 2013, doi: 10.1016/j.jpowsour.2013.06.024.
13. S. R. Dabbagh, M. K. Sheikh-El-Eslami and A. Borghetti, “Optimal operation of vehicle-to-grid
and grid-to-vehicle systems integrated with renewables,” 2016 Power Systems Computation
Conference (PSCC), Genoa, Italy, 2016, pp. 1–7, doi: 10.1109/PSCC.2016.7540933.
14. M. Honarmand, A. Zakariazadeh, and S. Jadid, “Integrated scheduling of renewable
generation and electric vehicles parking lot in a smart microgrid,” Energy Convers.
Manag., vol. 86, pp. 745–755, Oct. 2014, doi: 10.1016/j.enconman.2014.06.044.
15. M. E. Khodayar, L. Wu, and M. Shahidehpour, “Hourly coordination of electric vehicle
operation and volatile wind power generation in SCUC,” IEEE Trans. Smart Grid, vol.
3, no. 3, pp. 1271–1279, 2012, doi: 10.1109/TSG.2012.2186642.
16. M. Shafie-khah et al., “Optimal behavior of electric vehicle parking lots as demand
response aggregation agents,” IEEE Trans. Smart Grid, vol. 7, no. 6, pp. 2654–2665,
Nov. 2016, doi: 10.1109/TSG.2015.2496796.
17. A. Borghetti, F. Napolitano, S. Rahmani-Dabbagh, and F. Tossani, “Scenario tree gen-
eration for the optimization model of a parking lot for electric vehicles,” in 2017 AEIT
International Annual Conference: Infrastructures for Energy and ICT: Opportunities for
Fostering Innovation, AEIT 2017, 2017, vol. 2017-Jan, doi: 10.23919/AEIT.2017.8240519.
18. C. Orozco, A. Borghetti, S. Lilla, G. Pulazza and F. Tossani, “Comparison Between
Multistage Stochastic Optimization Programming and Monte Carlo Simulations
for the Operation of Local Energy Systems,” 2018 IEEE International Conference
on Environment and Electrical Engineering (EEEIC Europe), Palermo, Italy, 2018,
pp. 1–6, doi: 10.1109/EEEIC.2018.8494563.
19. G. J. Osório, J. M. Lujano-Rojas, J. C. O. Matias, and J. P. S. Catalão, “A new scenario
generation-based method to solve the unit commitment problem with high penetration
of renewable energies,” Int. J. Electr. Power Energy Syst., vol. 64, pp. 1063–1072, 2015,
doi: 10.1016/j.ijepes.2014.09.010.
20. L. F. Escudero, A. Garín, M. Merino, and G. Pérez, “The value of the stochastic solution in
multistage problems,” Top, vol. 15, no. 1, pp. 48–64, 2007, doi: 10.1007/s11750-007-0005-4.
 6 Energy Storage Sizing for
Plug-in Electric Vehicle
Charging Stations
I. Safak Bayram, Ryan Sims, Edward Corr,
Stuart Galloway, and Graeme Burt
University of Strathclyde

CONTENTS
6.1 Introduction���������������������������������������������������������������������������������������������������119
6.2  Literature Review������������������������������������������������������������������������������������������121
6.2.1 Literature on Smart Charging and Impacts of PEV Charging����������121
6.2.2 Literature on Charging Station Design����������������������������������������������122
6.2.3  Literature on Probabilistic Modelling of PEV
Charging Infrastructures�������������������������������������������������������������������122
6.2.4 Contributions�������������������������������������������������������������������������������������123
6.3 Demonstration and Testing Platform of a PEV Charging Infrastructure����� 124
6.3.1 Overview of PEV Research and Testing Projects at PNDC������������� 124
6.3.2 Summary of Results��������������������������������������������������������������������������126
6.4 System Model������������������������������������������������������������������������������������������������129
6.4.1 Markov-Modulated Poisson Process�������������������������������������������������130
6.4.2 Matrix Geometric Approach�������������������������������������������������������������132
6.4.3 Algorithmic Solution Technique��������������������������������������������������������133
6.5 Numerical Evaluations�����������������������������������������������������������������������������������134
6.5.1 Computation of Station Parameters���������������������������������������������������134
6.5.2 Charging Station Economic Analysis������������������������������������������������135
6.6 Conclusions����������������������������������������������������������������������������������������������������139
Acknowledgement������������������������������������������������������������������������������������������������� 139
Bibliography���������������������������������������������������������������������������������������������������������� 139

6.1 
INTRODUCTION
The future of electric power grids is currently shaped by two major advancements,
namely higher use of renewables on the supply side and increasing adoption of plug-in
electric vehicles (PEVs) on the demand side. These advancements aim to decarbon-
ize electricity and transportation networks since more than half of the global energy-
related carbon emissions are attributed to these two sectors. The push towards PEVs
is supported by legislations and regulations to encourage PEV uptake. For instance, a

119
120 Electric Vehicle Integration in a Smart Microgrid Environment

number of countries, including the UK, France, and Norway, plan to phase out fossil
fuel cars by introducing a ban on the sale of such vehicles and increase the cover-
age of charging network within the next two decades (Bayram and Tajer, 2017). On
the other hand, reaching net zero goals would require an exponential adoption of
PEVs; for instance, in the UK, there are currently two hundred thousand PEVs on the
road, and this number needs to be around four million by 2030 to meet government
policies (Haslett, 2019). Similarly, the State of California has a mandate to acquire
one-and-half million PEVs by 2025 and generate half of its electricity with renew-
ables by 2030. Net zero policy impacts are visible in France as there is a spring-back
effect on the year-on-year (January 2019–2020) PEV market share jump from 2.7%
to 11%. To that end, after being considered as a fringe technology, PEV market is get-
ting closed to a tipping point (Sperling, 2018). This can be viewed in PEV sales and
forecasts, as shown in Figure 6.1.
To support electrification of transportation, there is a need to deploy charging
nodes to meet various customer needs shaped by time, location, and duration
of service. In this chapter, we present a large-scale PEV charging station archi-
tecture equipped with an on-site storage unit. The primary goal is to develop a
probabilistic method to optimally size storage unit and show how on-site storage
can be effective in reducing peak demand and operational costs. Furthermore,
we present an overview of demonstration studies conducted at Power Networks
Demonstration Centre (PNDC is a research and testing hub founded by govern-
ment, industry, and academia and is part of the University of Strathclyde) charg-
ing infrastructures.
At the moment, there are three typical charging options for PEVs (Falvo, 2014).
First option is level 1 charging, which takes place in customer’s premises. Level
1 charging uses the existing, typically single-phase, electrical circuit at residential
units (2–3 kW) and fills the PEV battery during the night. Second charging option
uses AC level 2 chargers, which are typically located at public parking lots (e.g.,

150
Global BEV Sales (in Millions)

100

50

0
2014 2018 2022 2026 2030
Years

FIGURE 6.1 Battery electric vehicle (BEV) sales and forecasts (International Energy
Agency).
Energy Storage Sizing for PEVs 121

workplaces, shopping malls). This type of chargers typically supplies 6–7 kW power
to stationary vehicle. Third option is called fast or ultrafast charging, which can
transfer DC power at a rate of 50–350 kW (Srdic and Lukic 2019). Note that a typical
fast model (50 kW) can deliver enough charge for a 100-mile trip with 30 minutes
of charging, while ultrafast models are more preferred by high-end PEVs with large
batteries. In this chapter, the proposed model employs level 2 chargers to serve cars
parked at a large-scale lot. The related literature can be classified into three groups.
First group of studies is related to smart charging (see Section 6.2.1), which is aimed
at mitigating disruptive impacts of PEV demand on the power grid by exploiting the
demand flexibility of PEVs. Second group of studies is related to design of charging
stations and is categorized according to technology and economic operation domains
(see Section 6.2.2). Third group is related to probabilistic modelling of charging sta-
tions (see Section 6.2.3).

6.2 LITERATURE REVIEW

6.2.1 literAture on smArt chArging And impActs oF pev chArging
Smart charging of PEVs is critical in transition towards electric transportation.
Existing electric power grids are not designed to serve large PEV loads, and concur-
rent charging of PEVs will lead to major technical challenges on the distribution,
transmission, and generation components. At the distribution level, clusters of PEV
load during peak hours can lead to premature aging of transformers, increase distri-
bution system losses, and deteriorate power quality (García-Villalobos et al., 2014).
Increased stress and voltage fluctuations will further risk the consistency and safety of
the network. According to a field study conducted in the UK (Cross, 2016), one-third
of the low-voltage (LV) feeders will require intervention when 40%–70% of residents
have PEVs. At the transmission level, PEV load increases transmission congestion
level, which is a major challenge as the investments towards new transmission lines
have been declining. To support increasing electrification demand, there is an urgent
need to expand transmission network capabilities (Jurgen Weiss, 2019). Finally, at
the generation level, uncontrolled PEV load could lead to an increase in peak system
load, which requires a system additional deployment of new system upgrades (Wang
and Tehrani, 2015). To overcome the aforementioned disruptive impacts, design and
operation of charging facilities play a key role in transition to electric transportation.
PEVs are considered as a new kind of electric loads that have both temporal and
charging power flexibility. When a PEV is connected to a charger, charging session
starts at a constant power. In the case of multiple PEVs connected simultaneously,
they collectively increase peak loading and potentially trigger the aforementioned
disruptions. Smart charging is the optimization of charging power by exploiting PEV
flexibility to maximize one or more benefits such as reducing peak load, increasing
renewable energy utilization, lowering the cost of PEV charging, or deferring infra-
structural upgrades (García-Villalobos et al., 2014). Smart charging can be imple-
mented through standards such as IEC 61851 and ISO 15110 that enable control and
communication between the charger and the vehicle. In addition, a group of PEV
owners, coordinated by an aggregator, can participate in ancillary energy markets
122 Electric Vehicle Integration in a Smart Microgrid Environment

to stabilize electricity grids and, in return, receive payments for services rendered
(Han et al., 2010). Vehicle-to-grid (V2G) applications are particularly important to
smoothen what is known as solar “duck curves”, which are used to define the net
electricity generation curve when there is a significant solar generation. In this case,
PEVs’ charging rates are adjusted in a way to minimize the ramping-up requirements
of traditional power generators and lower financial losses (Lee et al., 2019).

6.2.2 literAture on chArging stAtion design

The approach described in this chapter focuses on economic operation of charging
stations and energy storage sizing (Negarestani et al., 2016; Sarker, 2018). In this
type of works, a critical component is local storage unit, which is typically employed
to shave peak load, reduce demand charges, and provide an additional income via
energy market participation. In the study by Negarestani et al., (2016), an optimal
sizing approach is proposed for energy storage systems (ESSs) in fast charging sta-
tions. In this work, PEV demand is calculated based on driving patterns and optimal
storage size is determined based on cost minimization. In the study by Sarker et al.,
(2018), an optimization framework is presented for an optimal bidding strategy in
day-ahead electricity markets for a PEV charging station with an on-site storage. In
the current charging station applications, one of the main issues is related to expen-
sive demand charges that constitute a sizable portion of monthly electricity bills and
reduce station profits (The 50 States of Electric Vehicles, 2018). Demand charges are
pricing tools to limit peak consumption of large customers by inducing a fee com-
mensurate to the peak consumption during any fifteen minutes during each month.
In June 2016, using a charging facility with two fast chargers, the following bill was
issued (Bayram and Ismail, 2019). The energy charge was 284 USD, and the demand
charge for the peak power was more than 2900 USD, representing 91% of the total
operational cost. High demand charges both compromise business models and nega-
tively impact PEV sales if prices are reflected to customers.

6.2.3 literAture on probAbilistic modelling oF

pev chArging inFrAstructures
Since the experimentation of capital-intensive PEV charging stations is not possible,
analytical modelling of PEV charging demand and infrastructure is used to provide
insights to system planners in how different system components interact with each
other. In line with the previous discussion, economic operation of charging infra-
structures has been the topic of several mathematical modelling and optimization
research works. Stochastic modelling and queuing systems have been widely used as
such methods capture the probabilistic nature of problem related to different battery
packs, technologies, weather parameters, and customer arrival and departure pro-
cesses (Hu et al., 2016). Moreover, station may have uncertainties related to renew-
able energy output, and storage unit can be modelled as a linear for simplification or
nonlinear “buffer” if battery’s chemical dynamics are taken into account. Some of
the related studies can be enumerated as follows. In the study by Aveklouris et al.,
(2017), fluid approximation of queueing models is adopted to calculate charging
Energy Storage Sizing for PEVs 123

station overloading probabilities. In the study by Ucer et al., (2019), a queueing

model is employed to calculate waiting times and service quality for a number of
charging stations located in Ohio by using actual traffic traces. In the study by Fan
et al., (2015), a charging station is modelled using a queuing model and captured the
effect of constant-current and constant-voltage charging on customer waiting times
in the station. Customer arrival and charging demand statistics are the important
system parameters in charging stations. In the study by Fotouhi et al., (2019), using
actual PEV charging data (level 2 chargers) from a major North American University
Campus between 2010 and 2015, a Markovian model for representing the charging
behaviour of PEV owners is presented. The results show that PEV owners not neces-
sarily fully charge their batteries; hence, service duration is shorter than expected.
In the study by Bayram et al., (2014), shared-based ESS located at residential units
is modelled using fluid dynamic approach and storage sizing problem is solved by
computing outage probability of the system, which is defined as the event when the
load is higher than the supply.

6.2.4 contributions
The contributions of this chapter can be enumerated as below:

• Firstly, an actual demonstration and testing platform of a PEV charging

infrastructure is introduced to show a detailed overview of a PEV charging
infrastructure and sample measurement results.
• Secondly, a probabilistic system model for the large-scale PEV charg-
ing station equipped with an on-site energy storage is presented (shown
in Figure 6.2). By considering the probabilistic nature of the customer
demand, the proposed architecture is modelled by a Markov-modulated
Poisson process (MMPP).
• Thirdly, a matrix geometric-based algorithm is presented in detail and used
to solve the associated capacity planning problem to find optimal energy
storage size and station capacity with respect to customer demand statistics.
• Fourthly, practical case studies are developed to show that (1) by accounting
for the statistical variations in customer demand, the power required for the
station is significantly less than the sum of chargers’ rated power and (2)
on-site storage units can help station operators to significantly reduce their
electricity bills.

FIGURE 6.2 System overview.

124 Electric Vehicle Integration in a Smart Microgrid Environment

6.3 DEMONSTRATION AND TESTING PLATFORM

OF A PEV CHARGING INFRASTRUCTURE
Before presenting an analytical model for a large-scale charging station, an over-
view of an actual small-scale PEV charging station with associated hardware and
software components is presented to provide a better understanding of actual system
components. The PNDC has completed several research projects relating to the elec-
trification of transport and charging infrastructure. This includes electrical impact
assessments of wireless inductive charging, on-street pop-up charger performance
testing, and active power quality compensation of single-phase harmonic and load
imbalance impacts of electric vehicle (EV) charge points. The PNDC was founded
with the goal of accelerating the penetration of disruptive technologies from early-
stage research into business as usual adoption by the electricity industry. The facility
comprises a fully representative distribution network, including the capabilities sum-
marized in Table 6.1. This enables the research, test, and demonstration of hardware,
software, and integrated systems solutions in a safe, controlled environment.

6.3.1 overview oF pev reseArch And testing projects At pndc

In 2019/2020, the PNDC supported Power Line Technologies Ltd. (PTL), Chronos
Technology Ltd, and the University of Strathclyde with the development of the
ENERSYN platform, which enables the hosting of partner-developed applications.
The platform monitors the LV network via high-fidelity voltage and current mea-
surements, making this data available to hosted “apps”. Two apps developed were a
micro-phasor measurement unit (PMU) and a non-intrusive load monitoring (NILM)
algorithm, to detect the connection of EVs to their chargers. The NILM algorithm
uses machine-learning techniques applied to LV network data to detect unique fea-
tures related to EV charger operation. The test setup was varied in two ways. The
first variation involved the Enersyn platform monitoring a rapid EV charging load
isolated from background noise. The second variation involved monitoring a rapid

TABLE 6.1
Selected Hardware and Software Assets of PNDC
Asset Rating/Comments
11-kV overhead/underground distribution Up to 60 kM of representative 11-kV network
400-V low-voltage distribution Up to 6 kM of representative LV network
Controllable motor-generator (MG) set 1-MW Motor/5-MVA generator
Controllable load banks 600-kVA controllable resistive/inductive
Real-time digital simulator (RTDS) 6 racks of RTDS execution hardware
Power hardware in the loop (PHIL) 540-kVA bidirectional power converter
Distributed energy resources (DERs) E.g., EV charge points, PV inverters, loads
Distribution management system Operational GE PowerOn SCADA
Data acquisition system Fluke & Beckhoff monitoring and logging
Energy Storage Sizing for PEVs 125

EV charging load with other background loads supplied from the same distribution
circuit. A high-level representation of the setup is illustrated in Figure 6.3. Rogowski
coil current transducers are installed on the incoming cables to the distribution
board, and voltage transducer measurements taken off terminals inside the building’s
distribution board. Conducting testing with and without background load permitted
the assessment of the NILM algorithm and its ability to disaggregate system noise
from loads of interest.
The ENERSYN platform and an off-the-shelf data acquisition system were
deployed in parallel to ensure the accuracy of data capture. Data was recorded
using a National Instruments CompactRIO (NI CRIO) data acquisition system and
a LabVIEW Virtual Instrument (VI) hosted on a PC. Three analogue input mod-
ules were deployed, with 4 channels per module. The other module employed in this
setup was a global positioning system (GPS) time synchronization module. Data was
sampled at a rate of 100 kHz to provide a high-resolution data set for the development
of machine-learning features that underpin the NILM algorithm.
Concurrently, PTL, Chronos, and University of Strathclyde deployed the devel-
oped ENERSYN monitoring platform. The developed system incorporates current
and voltage measurements, and records high-speed waveform events (100 kHz sam-
pling rate) which are time-stamped using GPS. The ENERSYN platform uses a long-
range (LoRa) GPS timing module supplied by Chronos. Captured waveform events
are then analysed by the on-board PMU and passed to the NILM algorithm (depicted
in Figure 6.4). Noteworthy waveform events are flagged and forwarded onto the
ENERSYN server for further analysis.

FIGURE 6.3 Iteration of test equipment and facility setup at the PNDC for NILM testing.
126 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 6.4 ENERSYN platform block diagram.

6.3.2 summAry oF results

High-resolution load signatures for a range of PEV charging profiles were analysed
using data gathered by the National Instruments CRIO monitoring system. Previous
studies (Zhang et al., 2011) have established a set of general usage patterns for charg-
ing PEVs, enabling four prescribed charging schedules to be derived, as illustrated in
Figures 6.5–6.8. These prescribed profiles were recreated, using PNDC-owned EVs,
and used as inputs to the development of the NILM algorithm.
PEV charger behaviour for non-prescribed charging profiles was also investigated
by logging the electrical parameters on a public rapid 50-kW charger. The logged

Charging profile power over time - Profile b)

4000
3500
3000
2500
Power (W)

2000
1500
1000
500
0
500
2500
4500
6500
8500
10500
12500
14500
16500
18500
20500
22500
24500
26500
28500
30500
32500
34500
36500
38500

-500

Time (s)

FIGURE 6.5 Profile a) empty to full (0%–100%).

Energy Storage Sizing for PEVs 127

Charging profile power over time - Profile b)

4000
3500
3000
2500
Power (W)

2000
1500
1000
500
0
500
2500
4500
6500
8500
10500
12500
14500
16500
18500
20500
22500
24500
26500
28500
30500
32500
34500
36500
38500
-500

Time (s)

FIGURE 6.6 Profile b) empty to part full (0%–75%)

Charging profile power over time - Profile c)

4000
3500
3000
2500
Power (W)

2000
1500
1000
500
0
500
2500
4500
6500
8500
10500
12500
14500
16500
18500
20500
22500
24500
26500
28500
30500
32500
34500
36500
38500

-500

Time (s)

FIGURE 6.7 Profile c) part full to full (25%–75%);

data was correlated with voluntary questionnaire responses by charge point users
about the start and end state of charge of their charging session. Figure 6.9 out-
lines the responses received, each point corresponding to a charge start and stop
percentage.
Based on public responses, the majority of charging sessions started at 10%–30%
state of charge (SOC) and stopped in the range of 70%–100% SOC. This additional
step provided inputs to ensure that the data for training the NILM algorithm was
128 Electric Vehicle Integration in a Smart Microgrid Environment

Charging profile power over time - Profile c)

4000
3500
3000
2500
Power (W)

2000
1500
1000
500
0
500
2500
4500
6500
8500
10500
12500
14500
16500
18500
20500
22500
24500
26500
28500
30500
32500
34500
36500
38500
-500

Time (s)

FIGURE 6.8 Profile d) part full to another higher capacity (25%–75%).

100
90
80
State of charge at end (%)

70
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70
State of charge at start (%)

FIGURE 6.9 Public responses for usage of the PNDC rapid charger.

representative of how typical PEV drivers use public rapid chargers, in addition to
the prescribed profiles found in the literature.
Training data for the NILM algorithm was logged at a sampling frequency of
100 kHz, which was exported in a technical data management system (TDMS) for-
mat, compressed, and shared on a data storage platform with the project team. In
terms of file size, a TDMS data file corresponding to 1-hour monitoring equated
to 10 GB of data. This learning emphasized the need for on-board edge-processing
Energy Storage Sizing for PEVs 129

analysis via the “apps” on the ENERSYN platform, to avoid unnecessary data trans-
mission to a server for centralized analysis.
The PEV charger data generated at the PNDC was critical to the NILM algorithm
development. The PNDC is exploring further avenues of research, which could make
use of this high-fidelity data. The developed NILM algorithm is now operating at
a success rate of over 90% in detecting PEV charger events after being trained and
tested by the data gathered at PNDC. The next stage of testing for the project will be
the deployment of the ENERSYN platform on the PNDC test network and monitor-
ing the public PEV chargers at PNDC over a longer period of time.

6.4 SYSTEM MODEL

In this section, we consider a large-scale charging station with N chargers serving
PEV demand. Charging station draws grid power and employs an on-site ESS shared
by all users. We denote the total charging power at time t by ν t and the ESS charge
level by i(t ) fir t ∈ + . It is worth noting that grid power is used to charge vehicles and
storage unit whenever possible. When the total PEV demand is higher than ν t , ESS is
used to support PEV demand unless it is fully empty. Customer statistics are as fol-
lows. We assume that PEVs’ arrivals at the parking station is a Poisson process with
rate λ . The average parking duration follows an exponential distribution with rate µ
(Fan et al., 2015). Furthermore, when a vehicle is parked, its power demand follows
the Poisson process with rate β . Finally, when an arriving customer finds all system
resources in use, an outage event occurs. In our model, we use the outage probability
as the natural performance metric.
Since PEV arrivals are independent of each other, the system state space
{0,1,, N } is represented with a birth–death process. The composite model for N
slot charging station is depicted in Figure 6.10.
For the given system description, it is natural to assume that the system operates in
a stable region. For this, the average demand should be strictly less than the available
station capacity. To that end, we have

 λ 
Nβ  < νt , (6.1)
 λ + µ 

or (6.1) can be rewritten as

 λ  1
ρ ≡ Nβ  < 1, (6.2)
 λ + µ  ν t

FIGURE 6.10 Birth–death process for N slot charging station.

130 Electric Vehicle Integration in a Smart Microgrid Environment

where ρ is the utilization parameter. Furthermore, the following assumptions are

made for the storage unit. Firstly, energy rating (in kWh) or the size of the energy
storage is denoted by B. Secondly, energy storage efficiency (charge–discharge) is
denoted by η , which takes values between 0 and 1. Note that this parameter reflects
the percentage of energy transfer after losses are excluded. Thirdly, in actual ESSs,
small percentage of energy is lost due to leakage. To simplify matters, dissipation
losses are ignored.
During charging station operations, on-site storage unit’s energy level changes in
one of the following cases:

• When the storage unit is entirely discharged, i.e., i ( t ) = 0, the total demand
is more than ν t . In this case, the rate of change in storage charge level would
be zero.
• ESS is fully charged, i.e., i ( t ) = 1, and the total demand is less than ν t .
Similar to the previous case, ESS charge level would not change.
• ESS is partially discharged, i.e., 0 < i ( t ) < 1, with any level of system
demand. In this case, ESS charge level would change commensurate
to the difference between charging power and the system demand, i.e.,
di ( t )
dt ( )
= η ν t − ∑ Ln ( t ) , where Ln ( t ) is the total demand when n ∈ {0,…, N }
n

chargers are on at time t.

It is noteworthy that due to the probabilistic nature of the system, by choosing storage
size B, only probabilistic guarantees can be provided to system reliability. Therefore,
let us define ε as the outage storage capacity, i.e., B ( ε ), as the minimum B satisfies to
serve (100 −  ε )% per cent of the total load, i.e.,

 min B  
B (ε ) =  . (6.3)

subject to  P ( i ( t ) ≥ B ) ≤ ε  


Note that our main goal is to calculate ε − outage storage capacity B ( ε ) based on grid
power, the number of PEVs, and other system parameters. To simplify mathemati-
cal notations, ESS size is scaled, and instead of B/η , we redefine B as the storage
size. Furthermore, power systems planning is typically done for “peak hour” period.
Therefore, in the rest of the paper, time index t is dropped and calculations for the
peak statistics are made.

6.4.1 mArKov-modulAted poisson process

Recall from the preceding discussion that at each state (see Figure 6.10), the aggre-
gate demand generates a state-dependent Poisson process (e.g., β , 2β , ). Therefore,
the entire charging station can be modelled with an MMPP, and energy storage sizing
option will be coupled with the computation of steady-state distribution probabilities.
Let pin denote the joint probability that the storage charge level is i and there are n
active PEVs, i.e.,
Energy Storage Sizing for PEVs 131

pin = P ( ESS  charge  level = i, n   active  customers ) . (6.4)

Then, the probability that ESS charge level is at level i can be written as
N

pi = ∑ p . 
n= 0
in (6.5)

Given the aforementioned assumptions, the system is modelled with a two-

dimensional birth–death process as depicted in Figure 6.11. Note that system states of
the Markov chain are represented by a doublet ( i, n ), where the first dimension reflects
storage charge level and varies from 0 to B, and the second dimension represents the
number of PEVs in the system and varies from 0 to  N. Alternatively, when customers
arrive at or depart from the station, the system state moves in the horizontal direction.
Similarly, when the on-site storage is charged or discharged, the system state moves
in the vertical direction. Moreover, transition rates, e.g., λ ,  µ , are determined based

FIGURE 6.11 An illustrative MMPP model for the charging station with storage size B.
132 Electric Vehicle Integration in a Smart Microgrid Environment

on the Poisson assumptions made earlier. It is worth emphasizing that storage sizing
calculations are made based on the assumption that the storage size has an infinite
capacity and the overflow probability is calculated as given in equation 6.3.

6.4.2 mAtrix geometric ApproAch

To compute steady-state probabilities of the MMPP model, an algorithmic solution
technique called matrix geometric approach has been employed (Neuts, 1994). As a
first step, balance equations for the Markov chain are written in the form of (6.4). For
instance, the last raw in the Markov chain (where i = 0), there are three intervals: (1)
i = 0, n = 0, (2) i = 0,  1 ≤ n ≤ N − 1, and (3) i = 0, n = N.
Vertical state transitions represent the energy storage charge–discharge events,
while horizontal ones depict PEV arrival and departure.
For the first interval, the balance equation is

Nλ p00 = µ p01 + ν p10 . (6.6)

This equation can be rewritten as

p00 = (1 − Nλ ) µ p01 + ν p10 . (6.7)

For the second interval, the balance equation can be rewritten as

  

The third case includes the rightmost boundary states and the balance equation can
be rewritten as

p0 N = λ p0 N −1 + (1 − Nλ − Nβ ) p0 N + ν p1N . (6.9)

Balance equations for other rows. e.g., i > 0, can be written similar to equations 6.7–6.9
by further incorporating additional vertical state transitions. Let pi denote ( N + 1)
-element row vector consisting of probabilities defined by balance equations, i.e.,
pi ≡ [ pi 0 , pi1 ,…, piN ]. Then, balance equations, such as the ones defined in equations
6.7–6.9, can be written in a compact matrix-vector equation. For p0 ,

p0 = p0 B0 + p1 B1 , (6.10)

where ( N + 1) x ( N + 1) matrices B0 and B1 are readily given by

 (1 − Nλ ) Nλ 0 … 0 
 
 µ (1 − µ − ( N − 1)λ − β ) ( N − 1)λ  0 
B0 =  0 2µ (1 − 2 µ − ( N − 2)λ − 2β )  0 
 
      
 0 0 0 … (1 − Nµ − Nβ ) 
 
(6.11)
Energy Storage Sizing for PEVs 133

and B1 is a diagonal matrix of the elements that contain vertical transition rate ν .
Remaining rows, e.g., i > 0, can be written similar to (6.11) by including transitions
between vertically adjacent states. Hence, a complete set of balance equations for the
remaining rows can be constructed from the matrix recurrences relation with three
matrices such as A0 ,  A1 ,  and A3, i.e.,

pi = pi −1 A0 + pi A1 + pi +1 A2  ,  for i > 0 (6.12)

Recall the assumption that the energy storage is initially assumed to have an infinite
size. Then, an infinite-dimension equation p is introduced as p = [ p0 , p1 ,…, pi ,…].
Then from (6.12), it is easy to see that

p = p P, (6.13)

where matrix P is a stochastic matrix of infinite size and called as the transition
probability matrix with each row summing to one. It is trivial that matrix P is con-
catenated from previously constructed submatrices, namely A0 ,  A1 ,  A2 , B0 , B1, in the
following repetitive form:

 B0 A0 0 0 … 
 
 B1 A1 A0 0 … 
 0 A2 A1 A0 … 
P=  (6.14)
 0 0 A2 A1 … 
 0 0 0 A2 … 
 
 … … … …  

In the next section, we present the solution methodology to compute minimum i that
satisfies ε = 1 − ∑ pi .
i

6.4.3 Algorithmic solution technique

In this section, the algorithmic probability solution developed by Neuts (1994) is
adopted. The solution to pi is written as

pi +1 = pi R,  i ≥ 0, (6.15)

where R is a ( N + 1) x ( N + 1) matrix that has a non-negative solution to the following

matrix equation
∞

R= ∑R A .
k =0
k
k (6.16)

Note that the calculation of pi is equivalent to finding the minimal non-negative solu-
tion to R matrix. A recursive calculation method is used to compute the matrix R. As
a first step, (6.16) is rewritten as
134 Electric Vehicle Integration in a Smart Microgrid Environment

R [ I − A1 ] = ∑RA k =0
k
k (6.17)

k ≠1

where I is an identity matrix of ( N + 1) x ( N + 1) size. Multiplying both sides of (6.17)

would yield
∞

R= ∑ R A [I − A ]
k =0
k
k 1
−1
. (6.18)

k ≠1

R can be iteratively solved for an initial solution of R = 0,

R =  A0 + R 2 A2  [ I − A1 ] .
−1
(6.19)

Once R is found, p0 can be calculated as follows:

p0 = p0 B ( R ) , (6.20)
∞
where B ( R ) = ∑ R k Bk . Recall that for the charging station model, B ( R ) = B0 RB1 .
k=0
It is important to note that the sum of probabilities should add up to 1. Once the prob-
abilities are found (e.g., p0 ), the computed results need to be normalized by dividing
each probability to the sum of all probabilities. To that end, the matrix geometric
solution can be summarized as below:

1. Construct matrix R by solving the equation in (6.16) and iteration in (6.19)

2. Compute p0 by solving the eigenvector equation in (6.20)

4. Normalize p0 and pi by dividing by the sum of all probabilities

Case studies are presented in the next section to provide more insights.

6.5 NUMERICAL EVALUATIONS

6.5.1 computAtion oF stAtion pArAmeters
Next, a number of case studies are presented to show how the proposed methodology
can be used to size ESS sizing in a charging point. It is assumed that charging station
employs typical level 2 chargers (6 kW), average parking duration is set as one hour
(µ = 1), and the charge request is set to β = 0.9. Three levels of customer arrival rate
per charger are chosen (from λ = 0.25 to λ = 0.75) to reflect different traffic regimes,
while station size is varied from N = 50 to N = 150. For system’s stability (see equa-
tion (6.1)), the power drawn from the grid is chosen as
Energy Storage Sizing for PEVs 135

 λ 
ν = Nβ  + ∆, (6.21)
 λ + µ 

where ∆ is a small constant set to 0.02. Computations for the size of on-site ESS with
respect to different station sizes and traffic regimes are presented in Figure 6.12. As
an example, for a charging station with 150 chargers and a peak traffic regime of
λ = 0.50, the size of the energy storage to provide 2% outage performances would
be 128 kWh.
From the results presented in Figure 6.12, two key observations are made. Firstly,
as the arrival rate increases, there is a need to a bigger energy storage to provide the
same level of outage performance. Secondly, as the station size increases, the need
for storage size per charger decreases due to “statistical gains”. For example, consider
the following two charging stations with N1 = 50 and N 2 = 250. Both stations oper-
ate under λ = 0.5N and employ a storage size of B = 112 kWh, while station 2 draws
three times more power than station 1, i.e., ν 2 = 3ν1. For these two stations, outage
probabilities are calculated as ε1 = 0.1093 and ε 2 = 0.0275. Then, it is easy to see

ε 
that gains in system performance  1 = 0.0275 are higher than the corresponding
 ε2 
 ν1 
capacity increase  = 3.97 .
 ν2 
As a second evaluation, the case in which storage size and the number of char-
gers are known and the computation of the amount of power that is needed are
investigated. In Figure 6.13, the results for arrival rate λ = 0.5 and varying station
sizes ( N = 100,  200,  and 300 ) are presented. These findings help system opera-
tors to decide on appropriate amounts of power for the station. Similar to the previ-
ous case, due to statistical gains, as the station size increases, per-charger resource
requirement decreases. For a target outage probability of 0.05 and on-site storage
size of B = 30 kWh, per-charger power requirement for a N = 100 charging station is
4.69 kW, while this value is only 2.3 kW for a charging station with 300 slots.

6.5.2 chArging stAtion economic AnAlysis

Recall the principal motivation to acquire on-site ESSs at charging stations to lower
running cost and defer major system upgrades. A typical electricity bill of a charging
station is composed of three parts, namely (1) a fixed fee, (2) energy charges (USD
per kWh), and (3) demand charges (USD per kW). In this case study, actual bill-
ing tariff of a utility company in San Diego is employed with the following details.
Monthly fee is $140, demand charge is $35 per highest kW, and energy tariffs are
time of use based on details in Table 6.2.
It is assumed that the charging station operates between 6 am and 10 pm, and
hourly traffic demand per charger is listed in Table 6.3. Moreover, the target outage
probability is set to 0.005, and the maximum grid power is limited to 610 kW.
As a first step, energy storage size according to peak consumption hour is cal-
culated using the methodology described in the previous section and found as
136 Electric Vehicle Integration in a Smart Microgrid Environment

(a) 1
N=50
N=100
0.8

Outage Probability ( )
N=150

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140 160
On-site Energy Storage Size (kWh)
(b) 1
N=50
N=100
0.8
Outage Probability ( )

N=150

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140 160
Energy Storage Size (kWh)
(c) 1
N=50
N=100
0.8
Outage Probability ( )

N=150

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140 160
On-site Energy Storage Size (kWh)

FIGURE 6.12 Computation of on-site energy storage size for different traffic regimes and
station sizes.
Energy Storage Sizing for PEVs 137

(a) 1
B=30 kWh
B=60 kWh
0:8

Outage Probability ( )
B=90 kWh

0:6

0:4

0:2

0
216 218 220 222 224 226 228 230
Power (kW)
(b) 1
B=30 kWh
B=60 kWh
0:8
Outage Probability ( )

B=90 kWh

0: 6

0: 4

0: 2

0
425 430 435 440 445 450 455 460
Power (kW)
(c) 1
B=30 kWh
B=60 kWh
0:8
Outage Probability ( )

B=90 kWh

0:6

0:4

0:2

0
640 650 660 670 680
Power (kW)

FIGURE 6.13 Energy storage size for varying grid power.

138 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 6.2
Electric Vehicle Charging Tariffs (in US Cents per
kWh) Adopted from San Diego Gas and Electric
Company
Time of Day Winter Summer
4 pm–9 am 26 54
12 am–6 am 9 25
10 am–2 pm 9 25
Other 25 30

TABLE 6.3
Hourly PEV Demand Per Charger for the Case Study
Hour 6 am 7 am 8 am 9 am 10 am 11 am 12 pm 1 pm 2 pm
Demand 0.2 0.3 0.4 0.5 0.6 0.7 0.7 0.8 0.9
Hour 3 pm 4 pm 5 pm 6 pm 7 pm 8 pm 9 pm 10 pm
Demand 0.8 0.7 0.7 0.6 0.5 0.4 0.3 0.2

1,200

1,000

800
PowerinkW

600

400

200
With Storage Without Storage

5 10 15 20
Hour of the Day

FIGURE 6.14 Hourly demand profile of the charging station.

B = 108 kWh. Next, the amount of power needed to provide the target outage proba-
bility is calculated and shown in Figure 6.14. The differences between the two curves
are related to peak demand reduction enabled by the employed energy storage unit.
It can be observed from the presented results that station’s peak demand is reduced
by more than one-third.
Energy Storage Sizing for PEVs 139

TABLE 6.4
Comparison of Monthly Electricity Bills (in Thousands USD)
Summer Tariff Winter Tariff
With storage 106.64 66.03
Without storage 159 100.3

Now that necessary parameters are calculated, monthly electricity bill of charg-
ing station with and without an on-site energy storage is calculated for summer and
winter tariffs.
As presented in Table 6.4, employing an on-site storage reduces typically opera-
tional cost by 34% in winter and 33% in summer. To make a fair comparison, level-
ized cost of electricity (LCOE), which includes acquisition, operational expenses,
and financial costs of storage unit, needs to be included. According to a recent report
(Henze, 2019), LCOE for lithium ion batteries has dropped to $187 per MWh. In this
case study, the employed storage unit has a size of 106 kWh; hence, the LCOE would
be close to $18.7. By further incorporating this cost in the presented results, it can be
concluded that employing storage units makes economic sense for charging station
operators.

6.6 CONCLUSIONS
In this chapter, we have presented a probabilistic capacity planning approach for PEV
charging stations equipped with an on-site ESS. The system is modelled with MMPP
where each system state is represented by the number of customers in the station and
energy storage charge level. To solve the steady-state probability distributions, an
algorithmic solution technique (matrix-geometric) was adopted. The principal goal
was to compute the minimum energy storage size that can provide a good level of
QoS measured by probability of outage events. In the last part, a number of case
studies were presented to provide insights on how the model can be used in capacity
planning. The results also showed that on-site storage systems can significantly lower
station’s peak demand and associated demand charges.

ACKNOWLEDGEMENT
The ENERSYN Project was funded by the European Space Agency (ESA) via the
Navigation Innovation and Support Programme (NAVISP) with Programme Code
NAVISP-EL2-012 (ENERSYN project and funding, n.d.).

BIBLIOGRAPHY
Aveklouris, A. N., Nakahira, Y., Vlasiou, M. and Zwart, B. (2017). Electric vehicle charging:
A queueing approach. ACM SIGMETRICS Performance Evaluation Review, 45(2), pp.
33–35.
140 Electric Vehicle Integration in a Smart Microgrid Environment

Bayram, I. S. and Ismail, M. (2019). A stochastic model for fast charging stations with energy
storage systems. IEEE Transportation Electrification Conference and Expo. Novi, MI.
Bayram, I. S., Michailidis, G. and Devetsikiotis, M. (2014). Unsplittable load balancing in a
network of charging stations under QoS guarantees. IEEE Transactions on Smart Grid,
6(3), 1292–1302.
Bayram I. S., Michailidis, G., Devetsikiotis, M. and Granelli, F. (2013). Electric power
allocation in a network of fast charging stations. IEEE Journal on Selected Areas in
Communications, 31(7), 1235–46.
Bayram, I.S. and Tajer, I. S. (2017). Plug-in Electric Vehicle Grid Integration. London: Artech
House.
Cross, J. a. (2016). My Electric Avenue: Integrating electric vehicles into the electrical net-
works. 6th Hybrid and Electric Vehicles Conference, London, UK.
ENERSYN project and funding. (n.d.). Retrieved from https://navisp.esa.int/project/details/26/
show
Falvo, M. S. (2014). EV charging stations and modes: International standards. International
Symposium on Power Electronics, Electrical Drives, Automation and Motion,
pp. 1134–1139. Ischia, Italy.
Fan, P., Sainbayar, B. and Ren, S. (2015). Operation analysis of fast charging stations with
energy demand control of electric vehicles. IEEE Transactions on Smart Grid, 6(4),
1819–1826.
Fitzgerald, G. and Nelder C. (2017). Evgo Fleet and Tariff Analysis. Rocky Mountain Institute.
Fotouhi, Z., Hashemi, M. R., Narimani, H. and Bayram, I. S. (2019). A general model for
EV drivers’ charging behavior. IEEE Transactions on Vehicular Technology, 68(8),
7368–7382, Boulder, Colarado.
García-Villalobos, J., Zamora, I., San Martín, J.I., Asensio, F.J. and Aperribay, V. (2014).
Plug-in electric vehicles in electric distribution networks: A review of smart charging
approaches. Renew. Sustain. Energy Rev., 38, 717–731.
Global EV Outlook. (2019). International Energy Agency. https://www.iea.org/reports/
global-ev-outlook-2019.
Han, S. H., Han, S. and Sezaki, K. (2010). Development of an optimal vehicle-to-grid aggrega-
tor for frequency regulation. IEEE Transactions on Smart Grid, 1(1), 65–72.
Haslett, A. (2019). Smarter Charging-A UK Transition to Low Carbon Vehicles: Summary
Report. Loughborough: Energy Technologies Institute.
Henze, V. (2019). Battery Power’s Latest Plunge in Costs Threatens Coal, Gas. New York:
Bloomberg.
Hu, J., Morais, H., Sousa, T. and Lind, M. (2016). Electric vehicle fleet management in smart
grids: A review of services, optimization and control aspects. Renewable and Sustainable
Energy Reviews, 56, 1207–1226.
Jurgen Weiss, M. H. (2019). Increased Transmission Investment to Support Growing
Electrification Demand and Access Low-Cost Resources Can Reduce Customer Rates.
Brattle Group, San Francisco, California.
Khaligh, A. and Dusmez, S. (2012). Comprehensive topological analysis of conductive
and inductive charging solutions for plug-in electric vehicles. IEEE Transactions on
Vehicular Technology, 61(8), 3475–3489.
Krein, M. and Yilmaz, P. T. (2012). Review of battery charger topologies, charging power
levels, and infrastructure for plug-in electric and hybrid vehicles. IEEE Transactions on
Power Electronics, 28(5), 2151–2169.
Lee, Z. L., Li, T. and Low, S.H., (2019). ACN-Data: Analysis and applications of an open EV
charging dataset. Proceedings of the Tenth ACM International Conference on Future
Energy Systems, pp. 139–149.
Energy Storage Sizing for PEVs 141

Liang, X., Srdic, S., Won, J., Aponte, E., Booth, K. and Lukic, S. (2019). A 12.47 kv medium
voltage input 350 kw ev fast charger using 10 kv sic mosfet. IEEE Applied Power
Electronics Conference and Exposition (APEC), Anaheim, CA.
Liu, Y., Zhu, Y. and Cui, Y. (2019). Challenges and opportunities towards fast-charging battery
materials. Nature Energy, 4(7), 540–550.
Negarestani, S., Fotuhi-Firuzabad, M., Rastegar, M. and Rajabi-Ghahnavieh, A. (2016).
Optimal sizing of storage system in a fast charging station for plug-in hybrid electric
vehicles. IEEE Transactions on Transportation Electrification, 2(4), 443–453.
Neuts, M. F. (1994). Matrix-Geometric Solutions in Stochastic Models: An Algorithmic
Approach. Courier Corporation, Baltimore, Maryland, USA.
Power Networks Demonstration Centre. (n.d.). Retrieved February 2020, from https://pndc.
co.uk/
Ronanki, D. and Williamson, S. S. (2018). Modular multilevel converters for transportation
electrification: challenges and opportunities. IEEE Transactions on Transportation
Electrification, 4(2), 399–407.
Sarker, M. R., Pandžić, H., Sun, K. and Ortega-Vazquez, M.A. (2018). Optimal operation
of aggregated electric vehicle charging stations coupled with energy storage. IET
Generation, Transmission Distribution, 12(5), 1127–1136.
Sperling, D. (2018). Electric vehicles: Approaching the tipping point. In Three Revolutions
(pp. 21–54). Washington DC: Island Press.
Srdic, S. and Lukic, S. (2019). Toward extreme fast charging: Challenges and opportunities in
directly connecting to medium-voltage line. IEEE Electrification Magazine, 7(1), 22–31.
Tehrani, N. H. and Wang, P. (2015). Probabilistic estimation of plug-in electric vehicles charg-
ing load profile. Electric Power Systems Research, 124, 133–143.
Ucer, E., Koyuncu, I., Kisacikoglu, M.C., Yavuz, M., Meintz, A. and Rames, C. (2019).
Modeling and analysis of a fast charging station and evaluation of service quality for
electric vehicles. IEEE Transactions on Transportation Electrification, 5(1), 215–225.
Williamson, S. S., Rathore, A.K. and Musavi, F. (2015). Industrial electronics for electric
transportation: Current state-of-the-art and future challenges. IEEE Transactions on
Industrial Electronics, 62(5), 3021–3032.
Zhang, P., Zhou, C., Stewart, B.G., Hepburn, D.M., Zhou, W. and Yu, J. (2011). An improved
non-intrusive load monitoring method for recognition of electric vehicle battery load.
Energy Procedia, 12, 104–112.
Autumn P. (2018). The 50 States of Electric Vehicles. Raleigh, NC: NC Clean Energy
Technology Center.
 7 Innovative Methods
for State of the Charge
Estimation for EV Battery
Management Systems
Zeeshan Ahmad Khan
Car.SW Org, Volkswagen AG

Franz Kreupl
Technical University Munich

CONTENTS
7.1 Introduction .................................................................................................. 143
7.2 Literature Review ......................................................................................... 145
7.2.1 Battery Management System ............................................................ 146
7.3 State-of-Charge Estimation .......................................................................... 149
7.3.1 Kalman Filter Algorithm .................................................................. 153
7.3.1.1 Extended Kalman Filter ..................................................... 154
7.3.1.2 Central Difference Kalman Filter ...................................... 156
7.3.1.3 Adaptive Extended Kalman Filter ..................................... 162
7.3.2 Sliding Mode Observer ..................................................................... 165
7.3.3 Backpropagation Neural Network .................................................... 170
7.3.3.1 Forward Propagation.......................................................... 173
7.3.3.2 Backward Propagation ....................................................... 174
7.4 Conclusion .................................................................................................... 176
7.5 Framework for Integrating EV Energy Storage Systems ............................. 176
Nomenclature ......................................................................................................... 178
References .............................................................................................................. 178

7.1 INTRODUCTION
Batteries have evolved as one of the most credible energy storage sources in the past
few decades or so. A majority of the electronic devices and tools that we use daily from
cell phones and laptops all the way to residential and commercial energy storage sys-
tems use batteries as the medium to store energy. In the automotive industry as well,
battery-powered vehicles are being considered as an alternative for the traditional

143
144 Electric Vehicle Integration in a Smart Microgrid Environment

internal combustion engine (ICE) vehicles. As batteries are getting employed in more
demanding applications like electric and hybrid vehicles, it becomes imperative to
exactly determine the dynamics of battery during its operation. The battery forms
the heart of electric powertrain in a car, and therefore, it needs a precise real-time
monitoring and control system. The battery terminal voltage, current, temperature,
internal resistance, state of charge (SOC) and state of health (SOH) are some of
the parameters that need to be evaluated and estimated while in operation and in
standby mode. The battery management system (BMS) is the part responsible for
monitoring the necessary battery parameters. Along with monitoring the parameters,
BMS also performs communication with other controllers; performs cell balancing;
and ensures that operating limits for voltage, current and temperature both at the
cell and at the pack level are not exceeded. The SOC is one of the most important
parameters that always need to be known. However, it turns out that SOC is not a
physically measurable quantity; rather, it is a quantity which can be estimated from
other physically measurable quantities. This thesis work is dedicated to developing
algorithms for SOC estimation using the lithium-ion cell modelling and comparing
their performance.
In the past two decades, increasing environmental concern and heavy reliance
of transportation sector on nonrenewable fossil fuels have forced the automotive
industry to look for sustainable and environment-friendly alternatives. Therefore,
the automotive industry has turned its attention towards electricity as the source to
power the coming age vehicles. At the moment, the automotive industry is experi-
encing a paradigm shift from the conventional ICE vehicles to the battery electric
vehicles (BEVs), hybrid electric vehicles (HEVs) and plug-in hybrid electric vehicles
(PHEVs). BEVs and PHEVs have batteries as their main source of propulsion, while
for HEVs lower rating batteries are present which are mostly used during vehicle
acceleration and braking. Higher energy efficiency, lower maintenance of the elec-
tric motors, simple powertrain design and zero tailpipe emissions must be a few
reasons to root for a future with electric vehicles (EVs) forming the heart of trans-
portation sector.
EVs of all types require a reliable, robust and dynamic energy storage during their
operation. Currently, batteries, supercapacitors, fuel cells and hydrogen-based stor-
age are being considered for this purpose. The choice of an energy storage depends
on the requirements of the application. For example, in case of HEVs, the main
requirement for energy storage device is to be able to quickly provide power dur-
ing vehicle acceleration and absorb power during braking. Therefore, energy storage
devices which have higher power density, such as supercapacitors, lithium-polymer
and nickel metal hydride (NiMH) batteries, would be more suitable for HEVs.
Mercedes-Benz B250e (28-kWh battery), Toyota Prius (fourth generation) (1.31-kWh
battery) and Honda Accord Hybrid (1.3-kWh battery) are some examples of HEVs.
For BEVs and PHEVs, the main goal is to select a storage device that can provide
a long range of operation. Therefore, lithium-ion batteries would be more suitable
for operation. BMW i3 (33-kWh battery), Tesla Model S (60- to 100-kWh battery),
Audi A3 Sportback e-Tron (8.3-kWh battery), Volvo XC90 T8 (9.0-kWh battery) and
Nissan Leaf (30-kWh battery) are some of the BEVs and PHEVs using the lithium-
ion batteries.
EV Battery Management Systems 145

The lithium-based batteries have gained prominence over their lead-acid and
NiMH counterparts. This could be attributed to the exceptionally high energy den-
sity, greater operating voltage range, higher capacities, possible combination with
other elements to form different stable chemistries and low maintenance require-
ments. Additionally, what makes them attractive for EV applications is the fact that
the self-discharge rate for lithium-ion battery is lower than half the self-discharge
rate for lead-acid and NiMH batteries. However, there are also several challenges that
are faced with lithium-ion batteries. First major issue is with protection. The batteries
require a protection circuit or device to prevent the cell voltage to go beyond upper
voltage limit (4.2 V in most cases) and below lower voltage limit (3.0 V in most cases
and 2.4 V in some cases). A battery management circuit with a resistor for bypass
can be the simplest circuit that comes in operation when limits are crossed. Ageing
is another major issue with lithium batteries. Ageing can be calendric (depending on
time) or cyclic (depending upon the number of charging or discharging cycles battery
has undergone). Other major issues include the cost and problems in transportation.
However, significant advantages offered by lithium-ion batteries outweigh the com-
plications, and therefore, the utility and scope of lithium battery-powered EVs is on
the rise and is likely to be at the centre stage in the automotive sector in the coming
future.
The increasing utility of lithium-ion batteries in dynamic applications poses a
challenging task of battery monitoring and control under all conditions. SOC appears
to be one of the most important parameters to be determined for any battery. The
concept of SOC can be simply understood by the concept of fuel gauge in IC engine
car. The battery in an EV can be analogous to the fuel tank. The fuel tank requires
fuel for vehicle operation, while the battery requires charge for operation. The charge
is simply the product of current and instantaneous time. Higher the current applied
to the battery, greater will be the charge flowing in the battery, and vice versa. The
fuel level in the tank can be seen with the fuel meter on the car dashboard to have
an idea of the remaining distance the car can drive. Since SOC is not a physically
measurable quantity, it is estimated by algorithms that use battery voltage, current
and temperature. These algorithms are implemented inside the controller in the BMS
and provide estimations of SOC. In order for BMS to accurately estimate the battery
SOC, BMS requires an accurate, high-fidelity, robust mathematical model of the
battery. Since in an EV the battery is subjected to varying currents during operation,
which causes changes in voltage and temperature, it becomes an exhaustive task to
estimate SOC accurately.

7.2 LITERATURE REVIEW

This chapter performs the analysis of the advanced mathematical methods that are
being widely used for SOC estimation in EVs. This chapter begins with an introduc-
tion into the BMS, which is the electronic device on which the SOC estimation algo-
rithms are run. The BMS is the controller, which monitors the battery voltage, current
and temperature, and uses these real-time values in the estimation algorithms. This
is followed by the analysis of the most commonly used SOC estimation algorithms,
namely the Kalman filters (KFs), sliding mode observer, and the backpropagation
146 Electric Vehicle Integration in a Smart Microgrid Environment

neural network (BPNN). The mathematical equations formulating these methods

are explained in detail so that the readers understand the underlying mathematics
behind these complex estimation algorithms. This understanding of the underlying
mathematics will enable the readers to implement these algorithms on their own and
estimate the SOC for any given battery system.

7.2.1 bAttery mAnAgement system

According to Ref. [1], BMS is the central component of the battery pack and guar-
antees safety and efficiency of cells and the battery pack by monitoring variables
such as voltage, current and temperature of each cell. Still according to Ref. [1], the
BMS uses the monitored variables to estimate SOC, SOH, state of energy (SOE),
and state of function (SOF) of each cell and the battery pack. In terms of safety, the
BMS ensures that the cells are protected from degradation, from extremely high and
low temperature conditions, from over-voltage and under-voltage conditions, from
high currents, from over- and under-charging conditions, and from deep discharge
condition.
The battery SOC depends on the residual capacity of cell, cell electrochemistry as
a function of the anode and cathode materials, cell internal structure, charging and
discharging rates, operating temperatures and the chemical reactions inside the cell.
Therefore, the estimation of SOC depends on several complex factors and is a secret
proprietary information of companies manufacturing and designing BMS.
Figure 7.1 illustrates a hypothetical example of the concept of battery SOC [2]. The
battery SOC is a direct indication of lithium concentrations at cathode and anode.
Figure 7.1a shows a charged cell, whereas Figure 7.1b shows a discharged cell. Figure
7.1c shows the variation in the open-circuit potential at the negative electrode with
the quantity of lithium in LiyC6, where y indicates the number of lithium atoms in the
chemical compound LiyC6. For a fully charged cell (orange colour), the concentra-
tion of lithium in LiyC6 (material which forms the negative electrode of cell) is very
high (0.8 in the figure). Figure 7.1d shows the variation in open-circuit potential at the
positive electrode with the quantity of lithium in LixMO2. For a fully discharged cell
(blue colour), the concentration of lithium in LixMO2 (material which forms the posi-
tive electrode of cell) is very high (0.8 in the figure). Hence, we can say that the bat-
tery residual capacity at any instant depends on lithium concentrations at anode and
cathode, thereby demonstrating the reliance of SOC on the internal chemical reaction
of the cell. This is just one of the examples of the complex relationships of SOC with
battery parameters, consequently making direct measurement of SOC a nonviable
alternative. The accurate and precise knowledge of SOC is important for estimating
the remaining charge in the battery and hence the remaining drivable distance; how-
ever, sampling, sensor noise and errors in parameter determination pose a significant
challenge in this area [2]. V. Prajapati et al. have conducted case studies to establish
examples of poor accuracy and low reliability in SOC prediction. The battery pack
in an EV is composed of large number of cells, thereby increasing the complexity of
the entire system and making control and management at cell level a daunting task.
Therefore, this compels the use of a BMS to make sure that the utilization of each
cell is maximized, and cell operates within limits. Among the multitude of tasks
EV Battery Management Systems 147

FIGURE 7.1 Schematics representing the correlation between battery SOC and residual
lithium concentration in anode and cathode at [Gray portion in electrodes in Fig. a] 80% SOC
and [Gray portion in electrodes in Fig. b] 20% SOC. The image is taken from Ref. [2].

performed by BMS during its operation, the most significant ones include the estima-
tion of battery states (SOC, SOH, SOE, and SOF) at each sampling interval during its
operation and measurement of voltage, current and temperature continuously at both
the cell and pack levels. Other major tasks include performing cell balancing, follow-
ing the specified operational limits (safety and diagnosis) and communicating back
and forth with the main vehicle computer. One commonly used example of commu-
nication technology is the controller area network (CAN) bus. Figure 7.2 presents the
block diagram displaying the major tasks performed by a BMS.
The BMS is a complex electronic device consisting of several components to per-
form the tasks listed in Figure 7.2. Figure 7.3 shows the simplified block diagram of
the components of a BMS. Battery pack, as the name suggests, represents the battery
under test and/or operation. Measurement unit consists of number of measurement
electronic components and sensors for current measurements, voltage measurements
and temperature measurements at the cell and battery pack levels. Battery controller
has an algorithm implemented for estimation, which exploits a mathematical model
of battery. The BMS controller adjusts the operating point based on the SOC val-
ues provided by the algorithm. Thus, the fundamental step towards developing an
accurate SOC estimation is dependent on the design and selection of actual battery
model implemented in algorithm. Cell balancing and equalization block comprises
a comparator that compares the cell voltages and finds the difference between the
148 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 7.2 Block diagram listing the major tasks of a BMS. State estimation and predic-
tion, cell measurement and balancing, safety and diagnosis, and communication are the four
major tasks that are executed by a BMS when operating within a battery pack.

highest and lowest cell voltage. If this voltage difference between two cells is larger
than the threshold, then the charging is stopped, and the charge is transferred from
cell with the highest voltage to the cell with the lowest voltage. The cell balancing
is achieved by means of charge shuttling through capacitors and MOSFETs (active
balancing) or simply draining the charge via a shunt resistor (passive balancing).
The CAN bus is responsible for communication between modules and the battery
controller. Moreover, it is also responsible for communication between battery con-
troller and other components of the battery pack. A battery pack during its opera-
tion is continuously receiving and sending information via CAN bus. A battery is
generally housed inside a battery junction box. The battery junction box has digital
communication ports that allow the transfer of information to and from the battery
via the CAN bus. User interface and display is the interactive unit (mostly a computer
screen), which shows the measured quantities from the BMS and permits the user to
provide an input from the available selection. Based on the measurements performed
by the measurement unit as shown in Figure 7.3, the battery controller estimates the
SOC in real time. However, due to the complex factors involving the SOC, estimation
errors can occur. Shen et al mentioned the most important sources of errors. These
include capacity errors, current measurement errors, voltage measurement errors,
and model-based errors. The capacity of a battery reduces as it ages; thus, error
in capacity estimation can occur which can propagate to SOC estimation through
ampere-hour method; these errors are referred to as capacity errors. The measure-
ment of the electric current is performed through shunt resistors or through Hall
effect sensors. Both approaches are prone to errors. These are the current measure-
ment errors. Since the measurement of electric currents is the basis of SOC estima-
tion, current measurement errors caused by these sensors can drastically degrade
the SOC estimation. Voltage measurement errors are also possible, and in this case,
EV Battery Management Systems 149

FIGURE 7.3 Simplified block diagram of a BMS. Measurement unit, battery pack, user
interface and display, BMS controller, CAN bus and cell balancing, and equalization blocks
are the major components of any BMS.

due to the voltage-based correction being used in the SOC estimation, the SOC esti-
mation can be significantly degraded. Finally, model-based errors, caused by the
incorrect value of parameters in a battery model, can lead to highly inaccurate SOC
estimation. Generally, the models are voltage-based, and incorrect computation of
terminal voltage due to an incompatible battery model can cause significant devia-
tions in SOC from the actual value.

7.3 STATE-OF-CHARGE ESTIMATION

Historically, SOC is calculated as the ratio between the remaining capacity of cell to
nominal capacity of cell. The nominal capacity can be obtained from the manufac-
turer’s data sheet.

Qa
SOC(%) = *100 (7.1)
Qn

where Qa represents the instantaneous available capacity in ampere hours and Qn

outlines the nominal capacity of the battery in ampere hours specified by the manu-
facturer in data sheet.
Figure 7.4 provides a pictorial representation of correlation between SOC and
stored energy in battery [5]. 100% SOC implies that battery has a rated capacity,
which is the sum of the stored energy, energy in empty region, and energy in inactive
region which will not be used under any condition. Fifty percent of SOC indicates
that the battery is half empty and half with stored energy, whereas 0% SOC implies
150 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 7.4 Representation of SOC in battery. Light Gray indicates the stored energy in the
battery, dark grey indicates the empty region, which is the energy used by the battery, and the
inactive region which will not be used. The image is taken from Ref. [5].

battery is completely drained out of energy (actual capacity); however, some energy
remains unused in the inactive region. The SOC estimation is one of the most impor-
tant tasks to be performed by a BMS periodically during battery operation. The
sampling rate of the SOC in BMS depends on the application and the type of sensors
and logging devices available in the circuit. From the battery module, voltage, cur-
rent, and temperature are continually measured at both the cell and pack levels. The
BMS utilizes these measurements in an algorithm to provide a robust estimate for the
SOC. However, the estimation of SOC is not a straightforward process; several fac-
tors influence the SOC, including applied charging/discharging rates, usable capac-
ity, temperature, aging, internal resistance, and self-discharge. The basic method to
estimate the SOC is obtained by integrating the current over the time and dividing
the value with the nominal capacity of battery; this technique is known as Coulomb
or Ampere-hour counting. Equation 7.2 gives the expression for SOC estimation
using Coulomb counting in discrete time.

η∆tI ( k )
SOC ( k + 1) = SOC ( k ) + (7.2)
Cn

where SOC(k + 1) and SOC(k) are the SOC values at (k + 1)th and kth sampling time,
respectively; ƞ is the Coulombic efficiency assumed as 1 for charging and as 0.98 for
discharging when a current pulse I(k) is applied; Cn is the nominal capacity from the
manufacturer’s data sheet; and Δt represents the sampling interval [6].
The significant advantage of coulomb counting is that it can be applied irrespective
of the cell chemistry. In addition, it requires low computing power when implemented
in real time and can be implemented in conjunction with other techniques. Although
coulomb counting is simple to implement, it has some serious disadvantages. As it
includes the integration of current over time, the error in estimation is high due to
the integral term. Also from equation 7.2, it is clear that estimation requires knowl-
edge of previous SOC value, which makes it dependent on the accuracy on initial
measurements.
Ng et al. used a smart estimation approach (termed ‘enhanced coulomb count-
ing’) based on conventional coulomb counting to perform SOC and SOH estimations
EV Battery Management Systems 151

in lithium-ion batteries. The research reports that SOC estimation is accurate with
conventional coulomb counting. However, as the cell ages, the error in SOH esti-
mation increases. The research reports a reduction in estimation error of about 8%
after 28th cycle (1.08% error), thus providing SOH estimation with a minimal error.
Zheng et  al. [8] performed coestimation of SOC, capacity, and resistance of lith-
ium-ion batteries. Their study used proportional integral (PI) observers to realize
the coestimation of battery SOC, capacity, and resistance. Further, the moving win-
dow ampere-hour counting method is used to improve the accuracy of estimation.
High-fidelity electrochemical model is adopted to capture the battery dynamics. The
research reports SOC root mean square error (RMSE) drops from 3.97% to 1.01%,
while the maximum SOC error declines from 5.86% to 1.58% after the application
of moving window ampere-hour counting method. When ageing effect is considered,
SOC RMSE can be decreased to an error band of 1% of actual SOC, and the maxi-
mum SOC can be limited in an error band of 2% for all ageing levels. In order to
overcome the dependency on current integration and initial SOC measurement, it is
necessary that methods be used that estimate SOC independent of these quantities.
KFs, observer methods, and artificial neural network can be used to estimate SOC
regardless of the dependencies shown by coulomb counting. Table 7.1 shows a sum-
mary of the techniques used in SOC estimation, and lists the techniques, their pos-
sible application areas, their advantages, and their disadvantages [9–11].

TABLE 7.1
Commonly Used SOC Estimation Techniques with Their Applications,
Advantages, and Drawbacks [9–11]
Techniques Application Areas Advantages Drawbacks
Look-up tables Nickel Cadmium Comparison between Offine1, sensitive to
(NiCd) two batteries (one with battery and operating
known SOC) conditions
Current sharing Lithium, nickel metal Easy implementation, Sensitive to battery and
method hydride (NiMH), low computation time operating conditions
lead acid
Discharge test Used for capacity Easy and accurate, Offine1, time intensive,
determination at the independent of SOH modifies the battery
beginning of life state, loss of energy
Physical properties Lead acid, zinc Online2, information Sensitive to temperature
of electrolyte bromide, vanadium about SOH and impurities, error due
(density, to acid stratification
concentration,
colour etc.)
Coulomb counting All battery chemistries Accurate if enough Sensitive to parasite
recalibration points are reactions, needs regular
available and with good recalibration points
current measurements
Open-circuit voltage Lead acid, lithium, Online2, computationally Needs long rest phases
(OCV) zinc bromide inexpensive, OCV (current = 0)
prediction
(Continued)
152 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 7.1 (Continued)

Commonly Used SOC Estimation Techniques with Their Applications,
Advantages, and Drawbacks [9–11]
Techniques Application Areas Advantages Drawbacks
Coup de Fouet Lead acid’ Estimating battery Sensitive to battery and
capacity operating conditions
Linear model Lead acid, Online2, easy Needs reference data for
photovoltaic batteries Implementation fitting parameters
Impedance All battery chemistries Information about SOH Temperature sensitive,
spectroscopy and quality expensive
Internal resistance Lead acid, Information about SOH, Good accuracy but only
lithium, nickel online2 for shorter intervals
cadmium
Artificial neural All battery chemistries Offine1, online2 Needs training data of a
networks similar battery,
backpropagation
algorithms are
computationally
expensive and
sometimes require
advanced techniques for
optimization
Fuzzy logic All battery chemistries Offine1, online2 Definition of membership
functions are highly
subjective, does not
scale to large models
Observers (sliding All battery chemistries Offine1, online2 Require accurate battery
mode observer and model in state pace
Luenberger format, determination of
observer) initial parameters by
trial and error
Kalman filters All battery Offine1, online2 Large computing
chemistries, capacity, need accurate
photovoltaic battery model, problem
batteries, dynamic in the determination of
applications such as initial parameters
HEVs
Hybrid techniques All battery chemistries Online2, dynamic Difficult implementation
e.g., artificial neural and stabilization as two
networks + Kalman separate approaches are
filters, fuzzy implemented in
logic + Kalman combination
filters

Offline1 implies entire dataset available for algorithm, whereas

Online2 means data is processed piece-by-piece in a sequential manner without having it available from
start.
EV Battery Management Systems 153

7.3.1 K AlmAn Filter Algorithm

KF is essentially an optimal recursive data processing algorithm. KF algorithms
are among the most widely used adaptive estimation techniques for SOC estima-
tion. Generally, KF can be applied for estimating the inner states of any dynamic
system based on system model and sensor measurements. The states of the system
are not directly measurable, and therefore by defining system model and obtaining
the measurements, KF performs state estimation. One important point to note is that
filter requires previous values of states, inputs and outputs to update the state in next
iteration. KF is usually quite robust when subjected to noises and uncertainties. The
filter can be applied to both discrete time and continuous time domain. However, for
accurate estimations, it is necessary that the filter be supplied with the exact initial
value of state and the measurements.
There exist different versions of KF. The filter can be applied to both linear and
nonlinear systems, and the filter choice depends on the nature of problem. In case of
SOC estimation, battery serves as the dynamic system and SOC as the inner state.
Figure 7.5 provides a classification diagram for KF techniques. Based on system type,
the KF is divided into two categories: linear and nonlinear Kalman filter. The non-
linear filter can be further divided into four groups: extended Kalman filter (EKF),
sigma point Kalman filter (SPKF), cubature Kalman filter, and particle filters. SPKF
is further divided into two categories based on the number of tuning parameters:
unscented filter (2 tuning parameters) and central difference filter (1 tuning param-
eter). In this chapter, we discuss the mathematics behind the EKF, adaptive extended
Kalman filter (AEKF), and central difference Kalman filter (CDKF).

FIGURE 7.5 Kalman filter classification diagram.

154 Electric Vehicle Integration in a Smart Microgrid Environment

7.3.1.1 Extended Kalman Filter

For nonlinear battery systems, a robust and more precise estimation can be obtained
using the EKF. The EKF utilizes the same prediction and updating steps as the linear
KF; however, the distinction lies in linearizing the system each point in time using
the Taylor-series expansion, which renders the system as a linear time-varying sys-
tem. The EKF has a very high accuracy when applied to systems with low to medium
degree of nonlinearity. This can be attributed to two reasons: Firstly, when comput-
ing the output estimates for the nonlinear system, EKF assumes that the expected
value of nonlinear function of unknown state is equal to the same nonlinear function
evaluated at the expected value of state. That is, it approximates

E [ fn ( x )] = fn [ E ( x )] (7.3)

where fn is a nonlinear function which represents the relationship between the sys-
tem input, state and output. This is true only when fn(x) is linear. Therefore, the EKF
works best for system with low to mild nonlinearities [6]. Secondly, when computing
covariance estimates, EKF uses a truncated Taylor-series expansion to linearize the
system equations around the present operating point; hence, higher-order terms are
rejected [6].
Consider a nonlinear system in discrete time in state space format as below:

x k +1 = f ( x k , uk ) + wk = A * x k + B * uk + wk (7.4)

yk = g ( x k , uk ) + vk = C * x k +1 + D * uk + vk (7.5)

a nonlinear function representing the system output, A, B, C, D are the system matrix,
input matrix, output matrix and feedforward matrix, respectively. xk is the system
state vector with a covariance Pk , uk is the input vector, yk is the output vector, wk is
the system noise vector, and vk is the measurement/sensor noise vector. wk is taken as
a white noise and is assumed to be zero mean Gaussian with a covariance Qk, and vk
is also assumed to be a zero mean Gaussian with covariance Rk .
The EKF involves mainly two steps: prediction step and update step. The predic-
tion step is used to calculate the priori estimate of the state and the covariance of
the system, while the update step provides the posteriori estimate of the same. The
relevant equations describing the EKF formulation are provided below.

7.3.1.1.1 Prediction Step

State estimate time update


X k +1| k   = A * X k + B * uk (7.6)

Error covariance time update

Pk +1|k == A * Pk * AT + Q   (7.7)
EV Battery Management Systems 155

Measurement error calculation


ek +1|k = yk +1|k − y X (
k +1| k   , uk ) (7.8)

Kalman gain calculation

K k +1|k =  Pk +1|k *C T * ( C* Pk +1|k *C T + R ) .

−1
(7.9)

7.3.1.1.2 Measurement Update Step

State estimate measurement update


X 
k +1| k +1 = X k +1| k + K k +1| k * ek +1| k (7.10)

Error covariance measurement update

Pk +1|k +1   =  ( I   −  ( K k +1|k * C )) * Pk +1|k (7.11)

where X̂ is the state vector of the order n × 1 with n being the number of states, A is
the system matrix of order n × n, B is the input matrix with order n × 1, C is the output
matrix with order m × n with m being the number of outputs, uk   is the input matrix,
P is the covariance corresponding to system state, Q is the covariance corresponding
to system noise, R is the covariance corresponding to measurement noise, y is the
system output, and e represents the error between the real system output and model
output, respectively. Covariance P, which is a square matrix, represents the uncer-
tainty in the system state. Depending on the dimensions of the system, this can be a
scalar or a vector. As can be observed from equations 7.7 and 7.11, covariance P is
updated twice during one iteration: one for a priori estimate at the start and the other
for a posteriori estimate after the measurements are added to the filter. In equation
7.7, term APA T uses the system matrix A to compute the covariance between differ-
ent system states and then adds it to the covariance Q. Equation 7.7 outputs a matrix
which depicts the variance of each state along its main diagonal elements and dis-
plays the correlation between the different states along its off-diagonal elements. The
choice of the covariance matrix strongly affects the filter performance. A high value
for covariance resembles a fast convergence of the filter to the actual state, whereas a
low value of covariance indicates relatively slower convergence. The initial value of
covariance P can either be selected by trial and error or be based on the sensor accu-
racy, if known. Covariance Q accounts for the unmodelled states and the random
variations in model parameters, which contribute to system noise. Higher value of Q
indicates model is accurate, while a lower value of Q indicates model is inaccurate
and system heavily relies on measurements. Matrix R models the noise in sensor
measurements as covariances. This matrix represents the correlation between differ-
ent sensors in the system. R is a diagonal matrix as the diagonal elements represent
the variance in each sensor measurement, while the off-diagonal elements represent
the correlation between different sensors, which is essentially zero.
156 Electric Vehicle Integration in a Smart Microgrid Environment

Equation 7.8 calculates the error or the residual, which is essentially obtained
by subtracting the model output from the sensor output or measurement. The term
(
y X  )
k +1| k   ,uk uses the predicted state to calculate the output as the state vector is
transformed into the measurement through the nonlinear function that represents the
output in terms of state vector and the control input. Once we know the nonlinear
relation between system output and state, residual can be easily obtained. Equation
7.9 outputs the Kalman gain, which is a real number between 0 and 1. Kalman gain
tells how much the estimate should change given a measurement. In equation 7.9, if
R = 0, Kalman gain, K = C −1. This tells us that if the measurements are very accu-
rate, then Kalman gain is simply obtained as the inverse of C matrix. However, if
P = 0, K = 0, this means that there is no change in state, so the estimates will not
change at each instant and gain will be zero. Equation 10 provides the new state
estimate based on the Kalman gain and the residual term. The Kalman gain has the
term C T , which when multiplied by the residual coming from the measurements pro-
vides an output in state space which is then simply added to the priori state estimate,

X k +1| k . In equation 11, I represents an identity matrix, a square matrix whose diagonal
elements are 1, off-diagonal elements are 0 and determinant = 1. In this equation,
product K × C is subtracted from I. In this product, output matrix C is being scaled
by Kalman gain. Therefore, if the Kalman gain K is large, (I – K × C) is small and
the resulting posteriori value of P will be smaller. However, if K is small, (I – K × C)
is large and the resulting posteriori value of P will be larger. This equation gives us
the ability to adjust the magnitude of P, which represents the uncertainty in state by
tuning via Kalman gain, K.
The EKF is an iterative process and estimates output and state at each iteration.
For the first iteration, the initial values of system state X, state covariance P, system
noise covariance Q, and measurement noise covariance R are provided. The initial
values of system state are either determined or taken as the first value of the experi-
mental data. Since P and Q are related to the system state, it is difficult to provide
an exact initial value. The initial values of P and Q are largely selected by trial and
error. On the other hand, covariance R is related to the noise in measurement. The
initial value of R can be taken as the square of the difference between the first value
of measurement output and that of model output. After having determined the values
of initial values of X, P, Q, R and the relevant battery parameters, these values can be
utilized in EKF for SOC estimation.

7.3.1.2 Central Difference Kalman Filter

SPKF is an alternative approach to KF generalization for nonlinear systems. Due to
its strong computations and resulting estimations, SPKF can be applicable to sys-
tem with mild to high degree of nonlinearity. Unlike EKF, which uses Taylor-series
expansion to estimate covariance matrix, SPKF performs a number of function eval-
uations to compute the estimated covariance matrices. This has several advantages:
(1) Derivatives do not need to be computed (which is one of the most error-prone
steps when implementing EKF), also implying (2) the original functions do not need
to be differentiable, and (3) better covariance approximations are usually achieved,
relative to EKF, allowing for better state estimation, (4) all with comparable compu-
tational complexity to EKF [12].
EV Battery Management Systems 157

In SPKF, a set of special points, known as sigma points, are computed, with
the condition that the mean and covariance of these points match exactly with the
weighted mean and covariance of the a priori random variable being modelled.
These points are then transformed after being passed through the nonlinear func-
tions. A posteriori mean and covariance are then approximated by mean and covari-
ance of these points. The SPKF can further be divided into two classes based on the
calculation of weighting factors: unscented Kalman filter (UKF) and CDKF.
The CDKF uses the Sterling’s polynomial interpolation method, avoids calculat-
ing the partial derivative, and is suitable for arbitrary functions. CDKF is capable of
providing state estimates even if nonlinear function is discontinuous or has singular
points, and has better accuracy than EKF [13]. Unlike UKF, which uses the three tuning
parameters for the calculation of sigma points, CDKF uses only one tuning parameter
‘h’ for the same. The CDKF formulation and stabilization are a little bit less complex
as compared to UKF; however, there is no difference in accuracy of estimation [12].
Consider a nonlinear system in discrete time as described by the following
equations:

x k +1 = f ( x k , uk ) + wk = A * x k + B * uk + wk (7.12)

yk = g ( x k , uk ) + vk = C * x k +1 + D * uk + vk (7.13)

The description of the above equations 7.12 and 7.13 has been provided in the section
EKF and will not be provided here. The CDKF also utilizes the state space matrices
A, B, C, and D for estimation. The state vector x k, control input uk , output yk , process
noise wk , and associated covariance Q, sensor noise vk , and associated covariance R
will all have the same definitions in the CDKF. However, some additional variables
will be added and the equation for CDKF will be different from EKF. The CDKF
also involves two steps: prediction step and update step. The difference lies in the fact
that instead of linearizing the system through the first-order Taylor-series approxi-
mation at the mean point as in EKF, CDKF uses a set of special points called the
Sigma points to approximate the mean and covariances between the states. These
sigma points are chosen deterministically to completely reflect the dynamics of the
states. They completely obtain the true mean and covariance of the state, and when
propagated through a nonlinear system, they capture the posterior mean and covari-
ance accurately to second order for a nonlinear function with any degree of nonlin-
earity [14]. The number of chosen sigma points generally depends on the number of
states/variables to be estimated. The general rule is for n number of states, we require
a minimum of 2n + 1 sigma points to completely capture the associated mean and
covariance. The first sigma point is generally the mean of the state and other two
sigma points are chosen to be equidistant from the mean on either side (Figure 7.6).
Behind the operation of the CDKF lies the algorithm called unscented trans-
form. The unscented transform states that the set of chosen sigma points   χ when
passed through a nonlinear function yield a set of transformed sigma points in a new
dimension.

γ = f (χ ) (7.14)
158 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 7.6 Example of sigma point selection for a random variable. For single variable
under normal distribution, three sigma points are chosen. First sigma point is the mean (red
square in the figure), while other two sigma points are chosen on either side of mean (black
squares in the figure). The size of the sigma points is proportional to the associated weights,
which indicates the influence of the particular sigma point over the output. In this figure,
highest weight is given to the mean and equal weights are assigned to the neighbouring sigma
points.

FIGURE 7.7 Pictorial representation of the sigma points. The points in dark gray ellipse
denote the original sigma points, while the points in light gray ellipse are transformed sigma
points after passing through the nonlinear function f (x ). This picture represents how the
sigma points are transformed from one space to another via a nonlinear function. The picture
is taken from Ref. [14].

where f is the nonlinear function,   χ denotes the sigma points in the original dimen-
sion, and γ represents the set of transformed sigma points. Later, the algorithm
computes the mean and covariance of the transformed sigma points. The mean and
covariance become the new estimate [14]. Figure 7.7 gives a pictorial representation
of sigma point taken from Ref. [14].
The formulae for sigma point calculation, weights calculation, and weighted mean
and covariance calculation along with the associated constraints for sigma points are
explained below.

χ =  x , x + h Px , x − h Px  (7.15)
EV Battery Management Systems 159

h2 − n
w0m =
h2

1
wim = (7.16)
2h 2

h2 − n
w0c =
h2

1
wic =
2h 2
2 n +1

µ= ∑w γ
i=0
m
i i (7.17)

2 n +1

Σ= ∑ w (γ
i=0
c
i i − µ )(γ i − µ )
T
(7.18)

2 n +1

1= ∑w i=0
m
i (7.19)

2 n +1

1= ∑w  
i=0
c
i (7.20)

where x is the average of the state variable x, Px is the associated covariance, h is the
scaling factor in CDKF, µ is the weighted mean of the generated sigma points, ∑ is
the weighted covariance of the generated sigma points, wim represents ith   weight cor-
responding to ith   sigma point, wic represents the ith   covariance weight corresponding
to ith   sigma point, γ i is the transformed output of the ith   sigma point, and n is the
number of states.
Equation 7.15 provides the formula for calculating the sigma points for a variable
x using its mean x and covariance P{ x}. Equation 7.16 gives the formula for calcu-
lating the weights for sigma points, which calculate the mean in equation 7.17 and
covariance in equation 7.18 of the transformed sigma points. The weighted mean
and covariance provide a strong approximation of the actual mean and covariance of
the system output. Each sigma point has an associated weight, and equation 7.7 pro-
vides the constraint that the sum of all the mean weights associated with the sigma
points should be equal to 1. Equation 7.20 provides the same constraint on the sum
of all covariances weights associated with sigma points. The weights for mean and
covariances can have different values or even the same value, which are calculated
empirically using a set of equations given in equation 7.16. Now with this background
knowledge, the CDKF can be developed. The prediction step provides the priori
estimate of the states and covariance, while the measurement update step provides
160 Electric Vehicle Integration in a Smart Microgrid Environment

the posteriori estimate of the same. These estimates are calculated by incorporating
the concept of sigma points, which form the backbone of the CDKF algorithm. The
relevant equations associated with CDKF in discrete time are provided below.
Nonlinear state space model

x k +1 =   fk ( x k , uk , wk )

yk +1 =   gk  ( x k +1 , uk +1 , vk +1 )  (7.21)

Variable definition
T
x ka =  x kT , wkT , vkT 

χ ka = ( χ kx ) , ( χ kw ) , ( χ kv ) 
T T T T
(7.22)

 
Initial values (k = 0)

( )
T
 T

x
a ,+
0 =  x0+ , w, v 
 

Pxa,0,+ = diag  Px+,0 , Q, R  (7.23)

7.3.1.2.1 Prediction Step

State estimate time update

χ ka−,+1 = x
a ,+  a ,+ a ,+  a ,+ a ,+
k −1 , x k −1 + h Px , k −1 , x k −1 − h Px , k −1

χ kx,,i− = fk −1 ( χ kx−,+1 , uk −1 , χ kw−,+1 ) (7.24)

2 n +1

xk− = ∑w *χ
i=0
m
i
x ,−
k ,i

Error covariance update

∑w * ( χ )( )
2 n +1
T
Px−,k = c
i
x ,−
k ,i − xk− χ kx,,i− − xk− (7.25)
i=0

Output estimate

Yk ,i = gk ( χ kx,,i− , uk , χ kv−,+1 )
2 n +1

  yk = ∑w *Y
i=0
m
i k ,i (7.26)
EV Battery Management Systems 161

Output covariance
2 n +1

∑w * (Y )( )
T
σ y ,k = c
i k ,i − yk Yk ,i − yk (7.27)
i=0

Cross-covariance

∑w * ( χ )(
2 n +1

− xk− Yk ,i − yk )
T
x ,−
σ xyˆˆ ,k = c
i k ,i (7.28)
i=0

Kalman gain

K k = σ xyˆˆ ,k * (σ y ,k )
−1
(7.29)

7.3.1.2.2 Measurement Update Step

State estimate measurement update

(
xk+ = xk− + K k * yk − yk ) (7.30)

Error covariance measurement update

Px+,k = Px−,k + K k *σ y ,k *K kT (7.31)

where k is the notation for discrete time, x k +1 is the state vector at instant k+1, f _ {k }
is the nonlinear input function, and gk is the nonlinear output function forming equa-
tion 7.21. In equation 7.22, augmented state vector x ka and augmented sigma points
vector x ka are defined. x ka consists of state vector  x k , process noise wk , and measure-
ment noise vk . χ ka is the vector for augmented sigma points consisting of sigma points
x kx related to state variables, sigma points χ kw related to process noise, and sigma
points χ kv related to measurement noise. Since CDKF is an iterative algorithm, after
defining the variables it is necessary to provide initial values at k = 0 for the state
and covariance vector. Equation 7.23 provides an expression for initial values of state
vector x0
a,+
and covariance P a,+ . x
x,0
a,+
consists of initial values of state variables in
0

mean value of x0+ ,  process noise w, and mean value of measurement noise v . Since
both the noises are white noises (zero mean and unit standard deviation), both the
terms w and v are taken as zero in expression. The second expression in equation
7.23 computes the covariance matrix for the system consisting of covariance cor-
+
responding to state variables Px,0 , covariance for process noise Q and covariance for
measurement noise R. This is a diagonal matrix with the different elements placed
along the main diagonal.
After the variables are defined and initial values have been provided, the iterative
filter can be started. At first, we compute the priori state estimate in equation 7.24.
This equation consists of three expressions. In the first expression, we calculate
the sigma points for the estimate. Recall equation 7.15 provides the expression for
162 Electric Vehicle Integration in a Smart Microgrid Environment

calculating sigma points, this expression is essentially like equation 7.15. This expres-
sion uses the posteriori state estimate x a,+ a,+
0 , posteriori covariance estimate Px,0 from
the previous time instant and scaling factor for CDKF h for calculating augmented
sigma points χ ka,+
−1. In the second expression, the sigma points χ k −1 and χ k −1 are passed
x,+ w,+

through the nonlinear input function fk −1 to provide the priori sigma points χ kx,− ,i .
Finally, these sigma points are multiplied by the corresponding mean weights wim
and summed together to generate priori state estimate xk− . Equation 7.25 computes
the priori covariance estimate Px−,k on the transformed sigma points using the covari-
ance weights wic and difference between sigma points χ kx,− −
,i and state estimate x k . This
is calculated for each sigma point, multiplied with the respective covariance weight
and summed together.
After the priori estimates have been calculated using the estimates from previ-
ous iteration, it is now the time to correct them using the available measurements.
In equation 7.26, output estimate yk is obtained by first calculating the sigma points
for output Yk ,i using the transformed sigma point for state χ kx,− ,i , sigma points related
to measurement noise χ kv,+ −1 and control input u k . These values are passed through the
nonlinear output function gk to yield the sigma points in output space. These sigma
points are then multiplied with corresponding mean weights to provide the output
estimate yk . This output estimate will serve as a feedback in updating the state at
the second step of this algorithm after obtaining Kalman gain. The relevant output
covariance σ y ,k is provided by equation 7.27 using the covariance weights wic, multi-
plying them with difference between output sigma points Yk ,i and output estimate yk .
A peculiar feature of SPKF algorithms is the calculation of cross-covariance. The
output of equation 7.28 is a cross-covariance matrix σ xyˆˆ ,k , in which each matrix ele-
ment shows the correlation between the state and measurements at each time instant.
This matrix shows how the output gets affected as the internal state variables change
over the different time instants.
The output covariance σ y ,k and cross-covariance σ xyˆˆ ,k are used to calculate the
Kalman gain K k in equation 7.29. This Kalman gain serves as the scaling factor
when providing the posteriori state estimate xk+ and posteriori covariance estimate
Px+,k in equations 7.30 and 7.31. respectively. These posteriori estimates are again used
in the next iteration, and the algorithm runs continuously to provide the desired state
and output estimates.

7.3.1.3 Adaptive Extended Kalman Filter

AEKF is an extension of the EKF algorithm with the incorporation of an updating
step for process noise covariance matrix (Q) and/or measurement noise covariance
matrix (R). The matrix Q represents the covariance of the system noise, and it is very
difficult to determine it precisely. The matrix R is the measurement noise covari-
ance and can be determined if the accuracy of the measuring instrument is known.
One big issue in designing a KF is the selection of the values of matrices Q and R.
Improper choice of Q and R can considerably deteriorate the performance of KF
algorithm and in some extreme cases, can even cause it to diverge. Presently, a major-
ity of research works adopt a makeshift approach for determining the values of Q and
EV Battery Management Systems 163

R matrices. In this method, these matrices are kept constant throughout the iterative
state estimation procedure and are manually adjusted by trial-and-error approach
outside the loop to obtain a suitable value [15]. Mehra in 1970 [16] published a paper
suggesting the suitability of Bayesian approach, correlation matrix approach, covari-
ance matching approach, and/r maximum-likelihood approach for adaptive estima-
tion. Any of these approaches can be used for performing the adaptive Q and/or R
estimation. Akhlaghi et al. [15] use the covariance matching approach to include an
update step for Q and R matrices in the recursive estimation loop. They propose an
innovation-based approach to adjust Q matrix and a residual-based approach to
iteratively adjust R matrix. The algorithm is similar in functioning to EKF along
with two additional steps to update Q and R. The innovation is defined as the differ-
ence between the measured value and the predicted value before the state is updated,
whereas the residual is defined as the difference between the measured value and
the predicted value after the state has been updated inside the loop. With the help
of an example, they find that AEKF is more accurate than conventional EKF and
simultaneously more robust to improper initial values of Q and R. Since this method
recursively updates Q and R in a loop, the designer must no longer worry about using
trial and error to select an appropriate initial value for Q and R.
In this chapter, the explained AEKF uses only the adaptive estimation of Q
matrix. The matrix R is taken as a constant throughout the run time of the loop. The
study has implemented only the adaptive estimation of Q because of the filter stabil-
ity. During the implementation of the adaptive steps of Q and R simultaneously, they
had problems trying to stabilize the filter and therefore opted to implement only the
adaptive step for process noise covariance matrix, Q, in this study for battery SOC
estimation. The adaptive step for Q is implemented using the maximum-likelihood
estimation. This approach defines some adaptive parameters that are used with the
conventional EKF estimation [17]. The filter stores the last N values of Q and pro-
cesses the data at each time step k ≤ N. Furthermore, the Q matrix is considered
completely stationary over any given interval of N sample periods, and tuning is
required for finding the suitable size of the moving window to achieve closed loop
convergence [18]. The relevant equations describing the AEKF in discrete time are
provided below.

7.3.1.3.1 Prediction Step

State estimate time update


X k +1| k = A * X k + B * uk (7.32)

Error covariance time update

Pk +1|k == A*Pk |k *AT + Q (7.33)

Measurement error calculation


ek +1|k = yk +1|k − y X(k +1| k   , uk ) (7.34)
164 Electric Vehicle Integration in a Smart Microgrid Environment

Kalman gain

 K k +1|k =  Pk +1|k *C T * ( C* Pk +1|k *C T + R )

−1
(7.35)

7.3.1.3.2 Measurement Update Step

State estimate measurement update


X 
k +1| k +1 = X k +1| k + K k +1| k * ek +1| k (7.36)

Error covariance measurement update

  Pk +1|k +1   =   ( I  −   ( K k +1|k *C )) *Pk +1|k (7.37)

Residual calculation


Res k +1|k +1 = yk +1|k +1 − y X (
k +1| k +1  , u k ) (7.38)

Moving window estimation

Res k +1|k +1  * Res  ek +1|k +1

H k +1|k +1    =   (7.39)
N
Process noise covariance update

Qk +1|k +1 == K k +1|k *H k +1|k +1*K k +1|k (7.40)

where X̂ is the state vector of the order n × 1 with n being the number of states, uk is
the input matrix, P is the covariance corresponding to system state, Q is the covari-
ance corresponding to system noise, R is the covariance corresponding to measure-
ment noise, y is the system output, and e represents the error between the real system
output and model output, respectively. Covariance P, which is a square matrix, repre-
sents the uncertainty in the system state. Depending on the dimensions of the system,
this can be a scalar or a vector. N is the size of the moving window, Res is the residual
which is the difference between the real system output and the model output after the
state has been updated through Kalman gain, and H is the output of the moving win-
dow estimation. Equations 7.32–7.37 are the same as the equations 7.6–7.11 used in
EKF. The description of these equations has been provided in the section of EKF and
therefore, will not be provided here again. The additional equations from equation
7.38 to 7.40 describe the implementation of the adaptive step for process noise cova-
riance Q. Equation 7.38 calculates the residual, which gives the difference between
the real measurement and the estimate value from the filter at that particular time
step. The residual helps to analyse the offset in the filter output at the given time step
and is used to update Q later. In equation 7.39, the moving average filter is applied to
the stored values of Q to obtain a smoothed output. The moving average filter pro-
vides an average single value over the N stored samples. Equation 7.40 provides the
EV Battery Management Systems 165

updated value of Q using the output of the moving window and the Kalman gain. The
multiplication of Kalman gain with the moving window output scales the value of Q
to bring it closer to its theoretical value, thus increasing the accuracy of the estimator.

7.3.2 sliding mode observer

The structure of the SMO is like any other state observer; however, in its estimations,
it is highly robust against errors induced in the system during modelling and the
system parameter variations. The robustness of the SMO can be attributed to the fact
that it generates a sliding motion on the error between the measured plant output and
the output of the observer, ensuring that the produced state estimates are precisely
commensurate with the real plant output [19]. This sliding motion is obtained by
inserting a nonlinear discontinuous term (switching term) into the observing system,
depending upon the output error. This discontinuous injection is designed in such a
way that the trajectories of the system are forced to remain on some sliding surface
in the error space. This resulting motion is referred to as sliding mode [20]. The
presence of the discontinuous term allows the system to reject disturbances, and the
classes of mismatch between the actual system and the model can be used for design
[20]. However, this discontinuous term that produces the sliding mode can cause fast
oscillations in the modelled system dynamics. These fast oscillations are termed as
‘chattering effect’ [21]. The chattering effect is observed in the estimated states as the
fast oscillations that cause the estimations to deviate from the actual values. Higher
the magnitude of the switching term, greater are the oscillations in the modelled
output. Therefore, the methods of chattering suppression are developed so that the
magnitude is decreased properly holding the establishment of sliding mode. First
option is to decrease the magnitude along with the system states [22]. The second
one implies that the magnitude is the function of an equivalent control derived by
a low-pass filter. The method can be applied for the plants subjected to unknown
disturbances [22].
Figure 7.8 represents the structure of a SMO in discrete time. The upper part of
the diagram is the discrete time state space representation of the real system. The
lower part of the diagram represents the modelled system in real time. The gain of
the observer, H, is obtained using the difference between the real system output and
the modelled system output. Furthermore, the block representing the term ‘sliding
motion’ represents the discontinuous switching term that provides the sliding motion
for the model. This switching term is a combination of the switching gain multiplied
by the uncertainty input matrix and the Lyapunov function. These terms in combina-
tion with the system matrix, A, are used to obtain the state estimate, xhat(k). The state
estimate is then applied to the output matrix, C, to get the estimated output, yhat(k).
The difference between the real and estimated output is calculated and the algorithm
operates iteratively.
The switching term which introduces a sliding motion is basically a signum func-
tion, which is discontinuous in nature and switches in zero time (infinite frequency).
In case of presence of noise in signals, the sliding mode observer can reduce the out-
put estimation error to zero in finite time. This robustness of the observer is due to the
strong mathematical design, which allows it to perform strongly even in the presence
166 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 7.8 Sliding mode observer implementation structure in discrete time. The model
system (system at the bottom) and the real system (system at the top) are shown in parallel.
The observer design requires an accurately developed and parametrized state model.

of any disturbances. Furthermore, the strict condition of observability is not manda-

tory for the SMO. The discontinuous switching function is defined in equation 7.41.

ν = ρ sign ( ey ) (7.41)

where ν is the discontinuous switching term, ρ is a positive scalar, and ey is the error
between the observer output and the real system output. The signum function takes
in ey as its input argument and depending on the sign of ey gives the output. When
ey is positive, output is 1; output is zero when ey is zero; and output is −1 when ey is
negative (Figure 7.9).
The signum function introduces a discontinuity in the system, which leads to a
phenomenon known as chattering. The problem of chattering is quite common in
SMO, and it appears as noise in the estimated states. In practical control systems,
this induced noise can come across as a serious problem that could undermine the
accuracy of the designed controller or observer. The solution for this problem can be
provided by a sigmoid function. The switching term ν is now represented by equa-
tion 7.42.

ey
ν=ρ (7.42)
ey + δ

where δ is a small positive scalar. The switching term in equation 7.42 when used in
the SMO can solve the problem of unwanted noise; however, the performance of the
system is unaffected by this modification to the switching term. The SMO gets its
name from the sliding motion that occurs on the sliding surface when the error in the
output estimation ey goes to zero. The ey is referred to as the sliding variable. When
EV Battery Management Systems 167

FIGURE 7.9 Example showing the output of a signum function. In this example, the value
of input varies between −5 and +5. When the input is −5, the output is 1. If the input is 0, the
output is 0; when the input is +5, the output is 1.

the value of ey becomes zero, the sliding motion begins. The error ey slides along the
surface, which is defined by the equation of ey. When the sliding motion occurs, the
error ey continues its movement on the surface and stays there. The switching term
ν oscillates along the vertical line defined by the signum function. With the error in
output estimation is reduced to zero, the observer output converges to the real system
output and the error in state estimation also reaches a very small value. The size of
the sliding surface can be adjusted by the value of ρ . Higher the value of ρ , larger
the area of sliding surface. However, this also increases the problem of chattering.
Therefore, a trade-off between the accuracy of observer and the severity of chattering
needs to be made when selecting the value of ρ .
Equations 7.41 and 7.42 will be used in the observer that has been designed for
performing SOC estimation iteratively. Additionally, it is required that starting values
of the states and control input should be provided during the observer initialization.
These starting values will be utilized in the first iteration of SOC estimation. The
relevant equations in discrete time that are used to design SMO for SOC estimation
are described below.
Nonlinear state space model

x k +1 = fk ( x k , uk , wk )

yk +1 = gk ( x k +1 , uk +1 , vk +1 ) (7.43)

Initial values (k = 0)

( )
x0 =  x0 
T
(7.44)
 
Parameters for solving Riccati equation
168 Electric Vehicle Integration in a Smart Microgrid Environment

R   =  1

 1 0 0 
Q   =    0 1 0 
 (7.45)
 0 0 1 

Parameters for solving Lyapunov equation

W   =  1

 1 0 0 
Q f   =    0 1 0 
 (7.46)
 0 0 1 

Observer output

k +1 = gk ( x k , uk )
y (7.47)

Measurement error calculation

ek +1 = yk +1 − y
k +1 (7.48)

Solution to Riccati equation

(
Pk = −Q / A + AT − C T *R −1 *C ) (7.49)

Observer gain matrix

( )
T
H k = R −1 *C*Pk (7.50)

Solution to Lyapunov stability equation

(
Pf ,k = −Q f * ( A − H *C ) + ( A − H *C )
T
) (7.51)

Model uncertainty factor

γ k = (W *C*Pf−1 )
T
(7.52)

State estimate measurement update

x   (  ) 
k +1 = A* x k + B*uk − H k * yk +1 − yk +1 + ρ*γ *sign yk +1 − yk +1( ) (7.53)

Equation 7.43 describes the structure of the nonlinear system in discrete time.
Equation 7.44 provides the initial values of the states for the first iteration.
EV Battery Management Systems 169

These values should be known and provided accurately as a model-based estima-

tion process is affected strongly by the starting values. In order to obtain the gain
matrices for the SMO state estimate equation in 7.53, we use the Riccati equation.
The constants for Riccati equation are R and Q that are provided default values in
equation 7.45. The asymptotic stability of the designed SMO in state estimation is
proved by Lyapunov stability. Therefore, it is necessary to define the constant for
solving the Lyapunov equation after a suitable candidate function has been selected.
The constants W and Q f for Lyapunov stability equation are provided in equation
7.46. Equations 7.44–7.46 are defined outside the iterative loop for SOC estimation.
The first step inside the recursive loop is the calculation of the observer output. The
observer output is dependent on the control input and the present value of the system
states. These are passed through the nonlinear function gk to get the observer out-
put for the current iteration given by equation 7.47. After observer output has been
calculated, the error in output estimation needs to be found. Equation 7.48 gives the
formula for output estimation error, which will be fed back to the observer at the end
of each iteration. Equation 7.49 provides the solution to the Riccati equation. The
solution involves the use of system matrix A and the output matrix C. The solution
of Riccati equation is a n × m matrix, where n is the number of rows of A and m is
the number of columns of C. If the denominator is large, then matrix P has a small
determinant, and when the denominator is small, the matrix P has a large deter-
minant. The value strongly affects the value of observer gain matrix H in equation
7.50, and when P is small, the resulting gain is small. However, with larger values of
P, observer gain matrix H has a higher value. After having calculated the observer
gain matrix H, it is important to check the stability of the designed observer by the
Lyapunov stability equation. The selected candidate Lyapunov function is provided
in equation 7.54.

V = eT * P * e (7.54)

This candidate function is differentiated with respect to time to find its derivative
and check whether the observer error ey slides along the surface. If the condition is
satisfied, then the sliding motion occurs when ey goes to zero; however, when this
is not satisfied, other candidate functions can be chosen to make ey slide along the
surface. After having checked for the sliding condition, the solution to the Lyapunov
stability equation can be found using equation 7.51. This solution is used to consider
the uncertainties associated with the model using the matrix γ calculated in equa-
tion 7.52. Since the observer considers all these uncertainties within the model itself,
it is more robust to the noise/disturbance present within the system states and the
control input. After having calculated both the gain matrices, H and γ , to account
for observer gain and model uncertainty, these can now be used to provide the state
estimate measurement update in equation 7.53. The third term in equation 7.53 scales
the error with the observer gain and provides a linear effect on the system state.
However, the fourth term multiplies the uncertainty with the discontinuous switching
function to show the sliding effect when the error converges to zero. After the state
has been updated, the recursive loop starts all over again from equation 7.47 and
continues until the maximum number of iterations are reached.
170 Electric Vehicle Integration in a Smart Microgrid Environment

7.3.3 bAcKpropAgAtion neurAl networK

Backpropagation Neural Networks (BPNN) are one of the most widely used neural
network architectures; with the intuition that the error between actual output and net-
work output is propagated backwards from output towards input. BPNNs are capable
of accurately mapping any nonlinear input–output relationship and can be used for
both regressions and classification problems. BPNNs use the learning principle of
error backpropagation.
Figure 7.10 shows the signal propagation in a BPNN. The forward signal (blue)
refers to the signal that propagates in the forward direction from input layer and
generates an output by passing thorough neurons and undergoing transformation by
means of weights and activation functions. The error signal (red) represents the offset
between the real output and the network output. Since we have the knowledge of real
output and not the intermediate output from different layers in a BPNN, the error
needs to be propagated in a backward direction in order to adjust the weights and
biases. The weights and biases are adjusted by means of a learning technique called
gradient descent. The network can be designed to operate for offline as well as online
applications.
BPNN presents a novel data-based approach to SOC estimation problem. The
aforementioned methods require the preparation of a model in state space in order to
perform SOC estimation. However, the BPNN does not require any model and uses
the measurement data for SOC estimation. The data from the measurement is filtered
and normalized before it is fed into the BPNN. This normalization is necessary as
different inputs to the network have different range of values. These different ranges

FIGURE 7.10 Forward signal and error signal flow in a BPNN. The circles represent the
neurons in different layers. Arrows in forward direction indicate the movement of the forward
signal from the input layer to the output layer, while arrows in backward direction indicate
the movement of error signal from the output back towards the input. The movement of error
signal from output towards input is termed as ‘backpropagation’.
EV Battery Management Systems 171

mean that inputs have different scales. Due to different scales of inputs, the network
performance will be slow and will take a longer time to converge. After normaliza-
tion has been done, all the different inputs to the network lie within the range −1
to +1 for the whole dataset. The normalization is performed outside of the iterative
estimation loop. In addition to normalization, the network configuration needs to be
defined before the training begins. The number of layers in the network, the number
of neurons in each layer, the activation function for each neuron in these layers, the
learning rate for the network, and the error threshold for training need to be pre-
defined. This network configuration is stored and will be subsequently used when we
test the network in the absence of a real system output.
The training is performed using an optimization algorithm, which allows the
network to reach a global minimum solution. This optimization algorithm corrects
the weights and output associated with each neuron through a set of mathematical
equations. The most commonly used optimization algorithms include the gradient
descent, scaled conjugate gradient descent, and Levenberg–Marquardt, among oth-
ers. In this study, stochastic gradient descent (SGD) algorithm is implemented for
training the network. Gradient descent aims to achieve the global minimum value
of the cost function by iteratively updating its parameters and checking the function
value at those parameters. The cost function in neural networks is given by calculat-
ing the squared difference between the real value and the predicted network output
over the total number of used samples. Gradient descent is the algorithm that is
applied to this cost function in order to minimize it by obtaining the optimal values
of the variables that constitute the expression of this cost function. The gradient
descent operates by calculating the partial derivative (gradient) of the cost function
with respect to each of the parameters. If the gradient is negative implying that the
cost function is being minimized, then we continue in this same direction to arrive
at optimal parameter values. The partial derivative is calculated iteratively for each
sample, and after reaching an error threshold defined by the user, the algorithm stops
and does not minimize any further. This point will be the global minimum of the
function, and the optimal parameter values will be obtained. The three forms of gra-
dient descent algorithm are batch gradient descent, mini-batch gradient descent, and
SGD. The fundamental equations are the same for each of three methods; however,
the main difference lies in the quantity of data samples used to compute the gradi-
ents in each step. Here, only the description of SGD will be provided. In SGD, the
gradient is calculated, and the parameters are updated for each sample. Therefore,
the training of weights and neurons happen at each iteration in the network. The
samples for training are usually selected at random from the whole dataset and fed
into the network. The cost function is calculated for each sample in every iteration;
however, under such conditions, the time to converge to a global minimum can be
longer and the path can contain more deviations before reaching the minima.
The rate at which the parameters are updated in each iteration of gradient descent
algorithm is defined by the learning rate. In this study, the learning rate for the net-
work is defined as a constant parameter with a value between 0 and 1. The learning
rate updates the weights of the network in each iteration so as to achieve the optimal
value of weights that minimize the loss function. Smaller values of learning rates
provide small updates to the weights, leading to slower decay of the loss function
172 Electric Vehicle Integration in a Smart Microgrid Environment

towards the global minimum. This implies that a network with a very small learning
rate will require more updates to the weights during training before weights reach an
optimal value; on the other hand, when the learning rate is too high, the updates to
the weights are quite large. However, under these conditions, it could happen that the
weights might skip the optimal value and the loss network instead of decaying starts
to rise. Therefore, an optimal value of learning rate needs to be chosen when perform-
ing training of the network. This learning rate and the network weights are saved so
that they can be used when the network is being tested against an unknown input.
The BPNN gets its name from the implemented backpropagation algorithm,
which propagates the error in backward direction from the output layer towards the
input layer. After the network output has been obtained the difference between the
network output and the real output (reference), the error is calculated. This error is
propagated backwards to update the weights and output of neurons. The weights of
the final layer are updated first, and the weights of the first layer are updated last. In
this network, a concept of backpropagation is implemented with the SGD algorithm.
This means that the gradients corresponding to the weights and neurons in the last
layer will be calculated first, and then, this will be propagated towards the first layer.
The gradients of one layer are used in computing the gradients of the previous layer.
This connected approach in gradient computation allows the efficient exchange of
information between the different layers of the network. The set of relevant equations
used to implement BPNN in this study are provided below.
Network parameter initialization

I = Number of inputs to the network

O = Number of outputs from the network

HN = Number of hidden layer neurons

 η = Learning rate

λ   = Regularization parameter

Error threshold = User defined

Total layers =  I HN O

Network weights and biases initialization

WH1  = 0.01* randn (HN, I)

WO = 0.01* randn (O, HN)

B1 = 0.01* randn(HN,1)

BO = 0.01* randn ( O,1) (7.55)

EV Battery Management Systems 173

Delta WH1 = 0.01* randn ( HN, I )

Delta WO = 0.01* randn ( O, HN )

Delta B1 = 0.01* randn ( HN,1)

Delta BO = 0.01* randn ( O,1)

where WH1 is the matrix of weights for hidden-layer neurons, WO is the matrix of
weights for output-layer neurons, B1 is the matrix of biases for the hidden-layer neu-
rons, BO is the matrix of bias for the output-layer neuron, Delta WH1 is the matrix of
regularization coefficients for hidden-layer neurons, Delta WO is the matrix of regu-
larization coefficients for weights of the output-layer neurons, Delta B1 is the matrix
of regularization coefficients for the hidden-layer neurons, and Delta BO is the matrix
of regularization coefficients for the output-layer neurons. The values initialized in
equation 7.55 will be used outside the recursive loop to calculate the gradients for
the first sample. randn is the command in MATLAB to generate random numbers
between 0 and 1. This is multiplied by 0.01 to make the coefficients even smaller as
higher initial values may affect the values, which will be calculated subsequently
inside the loop.

7.3.3.1 Forward Propagation

Output of each neuron of the hidden-layer unit
HN

zHN,1 = ∑WH1
i =1
i ,I *aI + Bi ,1 *1.0

(
yHN,1 = −1 + 2/ 1 + exp ( −2*zHN,1 ) ) (7.56)

Network output
O

zo = ∑WH1
i =1
i ,I *aHN + Bi ,1 *1.0

yo = −1 + 2/ (1 + exp ( −2*zO )) (7.57)

Estimation error calculation

 E = ym − yo (7.58)

Cost function calculation

M

J = 1/M ∑( y
k =1
m − yo )
2
(7.59)
174 Electric Vehicle Integration in a Smart Microgrid Environment

7.3.3.2 Backward Propagation

Regularization for weights of hidden and output layers

RegWH1 = λ*DeltaWH1

RegWO = λ*DeltaWO (7.60)

Output-layer error calculation

∂J
δ out = * (1 + yo ) * (1 − yo ) (7.61)
∂ yo

Hidden-layer error calculation for each neuron

δ H = WoT,i *δ out * (1 + yHN,i ) * (1 − yHN,i ) (7.62)

Gradient of a single output-layer weight

Grad o = yHN,i *δ out (7.63)

Gradient of a single hidden-layer weight

Grad H = aI ,i *δ H (7.64)

Gradient of output-layer bias

Grad B ,O = 1.0*δ out (7.65)

Gradient of hidden-layer bias

Grad B , H = 1.0*δ H (7.66)

Updating weights and bias matrices

WO = WO − eta*Grad o − RegWO

WH1 = WH1 − eta*Grad H − RegWH1

B1 = B1 − eta*Grad B , H (7.67)

BO = BO − eta*Grad B ,O

Equations 7.56–7.67 represent the set of equations used to implement the BPNN. The
first step is the forward propagation, which is then followed by the backward propaga-
tion. Equation 7.56 consists of two equations. The first equation is simply the sum of
EV Battery Management Systems 175

the product of the neuron output and the weight that originates from that neuron. This
equation gives the sum of all the weights that are connected to the first neuron in the
hidden layer from each neuron in the first layer. A bias term Bi ,1 is also added as each
layer has an associated bias neuron except the first layer. This bias has a constant value
of 1 and has an associated weight. After getting the sum of weights and biases, the
output of the first equation is fed into the second equation. The second equation is
the activation part of a neuron where the summation is fed into an activation function
(the activation function used here is the hyperbolic tangent sigmoid function). This
step gives the output of each neuron from the hidden layer. This small loop is run for
each hidden-layer neuron, and the results are stored in an array. Equation 7.57 is simi-
lar to the previous equation. This equation gives the predicted output from the neural
network. After we have an output from neural network yo, it is then subtracted from
the reference value from the measurement dataset ym . Furthermore, in equation 7.59,
we calculate the cost function of this problem, which needs to be optimized. After
calculating the cost function, the forward propagation step is completed. Now we
start with backpropagation. In this step, there are fundamentally four equations. The
first equation is for calculating the error associated with each neuron output, which is
essentially the deviation of a neuron output from the actual value that it should have.
This is first done for the output layer and then subsequently backpropagated for the
remaining layers except the input layer. The second equation is the calculation of the
gradients with respect to weights of the network. Gradient is the rate of change of cost
function with respect to a weight in the network. Third equation is the gradient for
the biases in the network. Lastly, after calculating the gradients, the existing weights
and biases are updated and algorithm runs all over again with these updated values
in the next iteration. Equation 7.60 calculates the matrices for applying regularization
to the updates of weights and biases. Regularization is performed to avoid overfitting
of the network to the output. Regularization reduces the values in the weight matrix,
which cause a reduced summation part in the neuron output, leading to a reduction in
the impact of the activation function on the neuron output in the forward propagation
step. This will tackle the problem of overfitting. The regularization is controlled by
the parameter λ , which is constant and tunable. Higher values of λ imply a stronger
effect of regulation, while smaller values of λ correspond to a smaller regularization
effect. Equation 7.61 calculates the error δ out associated with the output neuron. The
first term in the equation shows the change of the cost function with respect to the
output neuron. If the cost function does not strongly depend on the output neuron,
then the resulting δ out will be smaller. The second term is simply the derivative of the
activation function at the neuron output. Equation 7.62 shows the implementation of
the backpropagation principle. The δ out from the output neuron is used to calculate
the error associated with each neuron in hidden layer. The formulae for calculating the
error associated with neurons in the output layer and hidden layer are different. In this
step, the error is propagated backwards in the network, which gives us the intuition
about error in the output of neurons in a particular layer. After the errors associated
with neurons have been calculated, we move towards calculating the gradients for
weights in the network. Equation 7.63 shows the formula for calculating gradient for
one weight of the output layer. This is simply the product of the output of the neuron
in the previous layer from which the weight originates and the delta of the neuron to
176 Electric Vehicle Integration in a Smart Microgrid Environment

which it connects. This formula is repeated to obtain the gradient corresponding to all
the weights connected to the output-layer neuron. If the neuron output is small, then
the gradient will also be small. Smaller gradient implies that the weights learn slowly
and do not change too much in one step. Equation 7.64 shows the gradient for the one
weight of hidden layer. Similarly, equations 7.65 and 7.66 provide the formulae for
calculating gradients associated with biases in the network. After the gradients have
been calculated, the weights and biases are updated using the gradients multiplied by
the learning rate and regularization term. These two terms are subtracted from the
weights to bring the weights closer to the optimal value with each iteration. These
values are then used as the initial values in the next iteration, and the network operates
until a global minimum solution is reached.

7.4 CONCLUSION
This chapter explains the underlying mathematics behind the five most widely used
methods for battery SOC estimation. The relevant mathematical equations for each
model are explained in detail and serve as a base for developing these algorithms.
These methods after being developed can be implemented on a BMS with ease to
perform the desired estimation. All these methods are quite robust in their design
and have high accuracies. However, when implementing on a BMS, the computation
time, complexity, and memory requirements must also be kept in mind. The imple-
mentation of a BMS would require that the code of the selected algorithm based
on user requirements be written in C language. This can be done in two ways: One
would be to rewrite the whole code in C and the other would be to write the code in
MATLAB and then use the inbuilt MATLAB C-coder. Both methods would work
fine and will give the desired results when implemented correctly.
One possibility that arises from this work is the estimation of further states in the
battery. With the results of SOC estimation and a chosen algorithm, techniques to
evaluate SOH, state of power (SOP), and SOE can be developed. The algorithms to
further the problem of state estimation in a battery can also be implemented in real
time.
Another possibility would be to further analyse the implemented algorithms, and
attempts should be made to dissect their structure. The in-depth analysis of complex
mathematical equations forming the algorithm could be done to understand the effect
of altering a single step or a single variable on the result. This would be an interest-
ing research topic for enthusiasts, engineers, and researchers with a background in
mathematics and electrical engineering.

7.5 FRAMEWORK FOR INTEGRATING EV

ENERGY STORAGE SYSTEMS
The energy storage systems form an integral part of the electrification process. The
widespread adoption of EVs to replace the ICE vehicles depends to a large extent
on the availability of the electrical energy storage systems. The electrical energy
storage systems in the EVs make it possible to transfer electrical energy bidirec-
tionally between vehicle and grid, vehicle and home etc. The available energy in
EV Battery Management Systems 177

any electrical energy storage system is generally estimated by means of some math-
ematical algorithms. In case of batteries, the present energy in battery is interpreted
in terms of the remaining charge. The SOC of the battery is estimated in the back-
ground by means of different algorithms whose complexity depends on the level of
precision demanded and the feasibility of implementation in the BMS. The BMS has
its own engine control unit (ECU) inside the vehicle where it communicates back and
forth with the different ECUs over the CAN bus.
The algorithms for SOC estimation have evolved over the period. Earlier simpler
methods like the internal resistance method were used to get an idea of the SOC,
where the value of internal resistance was measured and SOC was estimated based
on it. Nowadays, a wide range of mathematical algorithms similar and even more
advanced to the ones mentioned in this chapter are being implemented as well. This
can also be attributed to the increased capabilities of the processors and control-
lers on which these complex algorithms are implemented. This would also imply
increased costs of the whole estimation process. Although the algorithms are costly
and complex, the accuracy and precision of estimation make them worthy of imple-
mentation. Researchers all over the world are also looking to implement hybrid
methods wherein they combine two different methods and try to make them work
together. This approach is even more complex but is not applicable in all the cases,
as the increment in accuracy of estimation is not very much when either of the two
methods is implemented independently. Therefore, for the industry standard, only
one mathematical method is developed in software and then implemented on the
BMS.
Along with the implementation of SOC algorithm, BMS also monitors the tem-
perature, voltage level, and current level across each cell and the whole module as
well. Since a battery is made up of several different modules, each module has its
own BMS and these BMSs exchange information with a central BMS in a battery
pack to monitor the essential variables at all levels. Since a battery pack consists
of thousands of cells arranged in series and/or parallel combinations, the task of
these controllers becomes more complex. There have been incidents when the cells
got overheated and went into thermal runaway state causing the whole battery to
catch fire. Incidents have also occurred where the battery caught fire due to faulty
wirings.
Since in an EV the battery forms the heart of electrical energy transfer, a lot of
precautions need to be taken while designing the battery module and integrating
it into the whole vehicle. Measures also need to be taken to ensure that there is no
fault on the software level and that the implemented software is working as per the
requirements. Also, as the battery is integrated into the EV, test and trial runs need
to be conducted thoroughly taking into account every possible test case to avoid these
incidents from happening in the near future.
As the EVs become more and more popular, the whole infrastructure surrounding
the EVs also needs to be developed. Proper quality and safety assurances need to be
met according to the specified standards to ensure a safe and reliable operation of the
whole EV ecosystem. In the near future, EVs would see the implementation of more
sophisticated and advanced technologies, which would raise the safety and quality
standards to make the whole process more secure and steadfast.
178 Electric Vehicle Integration in a Smart Microgrid Environment

NOMENCLATURE
AEKF: Adaptive extended Kalman filter
Ah: Ampere hour
ANN: Artificial neural network
BEVs: Battery electric vehicles
BPNN: Backpropagation neural network
CAN: Controller area network
CDKF: Central difference Kalman filter
EKF: Extended Kalman filter
EVs: Electric vehicles
HEVs: Hybrid electric vehicles
ICE: Internal combustion engine
LKF: Linear Kalman filter
LO: Luenberger observer
MOSFET: Metal oxide semiconductor field effect transistor
OCV: Open-circuit voltage
PHEVs: Plug-in hybrid electric vehicles
SMO: Sliding mode observer
SOC: State of charge
SOE: State of energy
SOF: State of function
SOH: State of health
SOP: State of power
SPKF: Sigma point Kalman filter
UKF: Unscented Kalman filter

REFERENCES
1. S. Dhameja. Electric vehicle battery systems. In Electric Vehicle Battery Systems,
Newnes, Woburn, 2002.
2. Marie-Therese von Srbik, Advanced Lithium-Ion Battery Modelling for Automotive
Applications, PhD thesis, Imperial College London, 2015.
3. V. Prajapati, H. Hess, E. J. William, V. Gupta, M. Hu, M. Manic, F. Rufus, A. Thakker,
and J. Govar, A literature review of state of-charge estimation techniques applicable to
lithium poly-carbon monofluoride (li/cfx) battery, In India International Conference on
Power Electronics 2010 (IICPE2010), pp. 1–8, Jan 2011.
4. P. Shen, M. Ouyang, X. Han, X. Feng, L. Lu, and J. Li, Error analysis of the model-based
state-of-charge observer for lithium-ion batteries, IEEE Transactions on Vehicular
Technology, 67(9), 8055–8064, Sep. 2018.

6. G. L. Plett, Extended Kalman filtering for battery management systems of LiPB-based

HEV battery packs part 1. Background, In Extended Kalman Filtering for Battery
Management Systems of LiPB-Based HEV Battery Packs Part 1, 134, 252–261, 2004.
https://doi.org/10.1016/j.jpowsour.2004.02.031.
EV Battery Management Systems 179

7. K. S. Ng, C.S. Moo, Y. p. Chen, and Y.-C. Hsieh, Enhanced coulomb counting
method for estimating state-of-charge and state-of-health of lithium-ion batteries, In
Enhanced Coulomb Counting Method, 86, 1506–1511, 2009. https://doi.org/10.1016/
j.apenergy.2008.11.021.
8. L. Zheng, L. Zhang, J. Zhu, G. Wang, and J. Jiang, Co- estimation of state-of-charge,
capacity and resistance for lithium-ion batteries based on a high-fidelity electrochemi-
cal model, In Co-Estimation of State-of-Charge, 180, 424–434, 2016. https://doi.org/
10.1016/j.apenergy.2016.08.016.
9. B. R. Pattipati, C. Sankavaram, and K. R. Pattipati, System identification and estimation
framework for pivotal automotive battery management system characteristics, IEEE
Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews),
41, pp. 869–884, 2011.

11. V. Pop, H. J. Bergveld, P. H. Notten, and P. P. Regtien, State-of-the-art of battery state-

of-charge determination, In State-of-the-Art of battery, 16(12), R93–R110, 2005. https://
doi.org/10.1088/0957-0233/16/12/R01.
12. G. L. Plett, Sigma-point Kalman filtering for battery management systems of LiPB-
based HEV battery packs: Part 1: Introduction and state estimation, In Sigma Point
Kalman Filtering for Battery Management Systems of LiPB-Based HEV Battery Packs
Part 1, 161, 1356–1368, 2006. https://doi.org/10.1016/j.jpowsour.2006.06.003.
13. E. A. Wan, Sigma-point Fillters: An overview with applications to integrated naviga-
tion and vision assisted control, 2006 IEEE Nonlinear Statistical Signal Processing
Workshop, pp. 201–202, 2006, Cambridge, UK. doi: 10.1109/NSSPW.2006.4378854.

16. R. Mehra, On the identification of variances and adaptive Kalman filtering, IEEE
Transactions on Automatic Control, 15(2), pp. 175–184, April 1970.

21. I. Boiko, L. M. Fridman, and M. I. Castellanos, Analysis of second - order sliding-mode

algorithms in the frequency domain, IEEE Transactions on Automatic Control, 49:
946–950, 2004.

 8 High-Voltage Battery
Life Cycle Analysis with
Repurposing in Energy
Storage Systems (ESS)
for Electric Vehicles
Mamdouh Ahmed Ezzeldin, Ahmed Alaa-
eldin Hafez, Mohamed Adel Kohif, Marim
Salah Faroun, Hossam Hassan Ammar
Nile University

CONTENTS
8.1 Introduction .................................................................................................. 182
8.2 Literature Review ......................................................................................... 182
8.2.1 Conventional Cars and Electrical Vehicles ...................................... 183
8.2.2 Life Cycle.......................................................................................... 184
8.2.3 Manufacture ...................................................................................... 184
8.2.3.1 Battery Cell ........................................................................ 185
8.2.3.2 Packaging ........................................................................... 187
8.2.3.3 Battery Management System ............................................. 187
8.2.3.4 Battery Pack Assembly ...................................................... 188
8.2.3.5 Solutions to Minimize the Impact Due to Manufacturing .....188
8.2.4 Battery Life cycle Analysis with Repurposing in Energy
Storage Systems (ESS) ..................................................................... 189
8.2.4.1 First Use in Electric Vehicles............................................. 189
8.2.4.2 Second Use in Energy Storage Systems ............................ 191
8.2.5 Environmental Approaches for Battery Disposal ............................. 192
8.2.5.1 Introduction and Background Information ........................ 192
8.2.5.2 Currently Applied Recycling Techniques .......................... 192
8.3 Methodology ................................................................................................. 195
8.3.1 Power Peak Shaving.......................................................................... 195
8.3.1.1 A Sample of Current Simple Comparative Algorithms ..... 196
8.3.2 The Proposed Simple Comparative Algorithm ................................ 197
8.3.2.1 Definition of Variables ....................................................... 197
8.3.2.2 Solution Flow ..................................................................... 198
181
182 Electric Vehicle Integration in a Smart Microgrid Environment

8.4 The Methodology Study Design....................................................................200

8.5 Factors of the Methodology’s Ideal Environment......................................... 201
8.5.1 Drivers............................................................................................... 201
8.5.2 Barriers.............................................................................................. 203
8.6 Case Study.....................................................................................................205
8.6.1 System Briefing..................................................................................205
8.6.2 System Parameters & Assumptions...................................................205
8.7 Results............................................................................................................207
8.8 Conclusion.....................................................................................................209
Acknowledgements.................................................................................................209
Bibliography...........................................................................................................209

8.1 
INTRODUCTION
Since the world has discovered electricity as an energy form, to be used in a mas-
sive and exponentially growing rate, it has been recognized as a clean energy
source and form, until it was discovered that the damage that could be evolved by
using electricity as a clean form of energy was not prevented, but the methods used
to generate electricity are the problem. Burning of coal and oil causes a damage to
the environment in every aspect – one of the causes is emission process that occur
during the burning. However, there are efforts to implement other cleaner ways of
generating electricity such as solar panels and wind turbines to meet the increas-
ing need for electricity. To narrow down our vision upon the expected definite
environmental disasters, we have to analyse the pattern of the market needs and
its consequences on the short term to verify the expected negative impacts that we
assume on the long term, so by looking at the late twenty years, it could be obvious
to see how the rate of disposed lithium-ion batteries (LIBS) has increased since
the demand on electrical energy storage is escalating. It should be elaborated that
the chemicals of LIBS have their own consequences on the environment such as
damaging the soil and water supply to kill the biological form of any exposed area.
In addition, as the industry of transportation is witnessing a leap and transforma-
tion into electrical vehicles with a slow development rate in the field of recycling
LIBS into the bargain. Expectations should be confirmed that if there is neither a
solution to fully restore heavy metals from LIBS with a commercial applicability,
nor a new technology for electricity storage, there will be fields of LIBS dispos-
als and closed loop of environmental disaster. In this chapter, we discussed how
the currently conducted methods are not efficient on the long term, and suggested
a strategy to hold down the speed of reaching peak point as it cannot be ended or
stopped its consequences, by studying the need for LIBS in our life and reposition-
ing it to serve longer before disposed.

8.2 
LITERATURE REVIEW
The first section compares the differences between the electrical vehicles and the
conventional cars, while the second section talks about the life cycle of the battery
and its components.
High Voltage Battery Life Cycle Analysis 183

8.2.1 Conventional Cars and Electrical Vehicles

Since the invention of electrical vehicles, they are always in comparison with the con-
ventional cars, and nowadays there is a great controversy between both conventional
cars and electrical vehicles, which leads to some confusions of which one is the per-
fect to use. In order to have a considerable judgement between the two vehicles, some
points should be clarified. Firstly, the conventional vehicles are those that depend on
the internal combustion for propulsion without any electric motors or mechanism.
An internal combustion engine exploits the fuel/air mixture’s repeated explosion
to move a collection of pistons down a corresponding set of cylinders (Granovskii
et al. 2006). Those engines, in effect, provide a mechanical rotational energy via an
attached crankshaft. It is observed throughout the years that the conventional cars
have better power, their maintenance cost is low, and they are at economically fea-
sible price. Secondly, the electrical vehicles (EVs) are the vehicles that run with one
or more electric motors which are responsible for propulsion. They consist of four
essential elements: an on-board converter that converts the household AC energy to
DC, batteries that stock charged electricity, an inverter that regulates the electricity
flow from the batteries to the motor, and motors that convert the electricity to propul-
sion power (Mebarki 2017). Although conventional vehicles have better power and
are of lower cost than the electrical ones, they have a huge negative impact on the
environment. They deploy harmful CO2 emissions, which affect the global warm-
ing (Granovskii et al. 2006). They are also responsible for tangential environmental
impacts such as human toxicity potential, terrestrial toxicity potential, freshwater
toxicity potential, mineral depletion potential, and fossil fuel depletion potential
(Mebarki 2017). In the last decade, electrical vehicles have witnessed an increase
in research and development. As more technological enhancements are processed
to extend EV battery life and improve its performance, a great demand for EV has
risen. Electrical vehicles have lots of advantages over the conventional vehicles due
to their low environmental impacts since no carbon dioxide-containing exhaust gases
are emitted and 97% reduction in CO2 emissions are observed in contradistinction to
conventional vehicles; in addition, they have lower charging cost as compared to the
fuel cost (Mebarki 2017). But contrarily, the lack of the infrastructure of the charging
stations and the charging time needed to recharge the battery are considered as the
drawbacks of the EVs in addition to the secondary environmental impacts caused by
the manufacturing of the batteries and their disposal (Li et al. 2014). However, there
are some methods to make these batteries environmentally friendly, which will be
discussed later.
Although conventional car batteries are similar to the electrical vehicle’s ones, there
are major differences between both types of batteries. As the batteries used in the con-
ventional cars can’t replace the ones used in the EV, and vice versa. Each battery has
its capabilities and abilities to withstand the conditions of the car in which it is imple-
mented, such as the heat and the power. Moreover, the batteries in the EV are used to
move the whole car, while in the conventional ones, they are used just to start the car.
The battery of the EV is made up of two subpacks which are connected in parallel.
Each subpack consists of 6 battery modules and 12 totals; each module is made up
184 Electric Vehicle Integration in a Smart Microgrid Environment

of equal number of battery cells, which are different in each type of car (Granovskii
et al. 2006). The weight of the battery is almost 60% of the total weight as it can
weigh 253 kg and can have an energy capacity of 26.6 kWh. In addition, the batteries
in the EV are made of thick lead plates and can be recharged without causing any
damage (Ellingsen et al. 2014). Generally, the most commonly used batteries are the
LIBS. On the other side, conventionally, a standard battery is made up of six cells
that are connected in series and provide two volts, thus constituting a total of twelve
volts. The weight of the battery (18 kg) – which is almost 1% of the total mass of a
car – compared to the whole mass of the conventional cars is negligible. Moreover,
the most commonly used batteries are the lead acid batteries that are made of thin
lead plates which allow the car to start fast, and these batteries are available for single
use only as they can’t be recharged.
LIBS are the mostly used type of batteries in the EV because they are of low
weight and have a great energy density. EV may seem that it has no environmental
impact, as it is going to reduce the greenhouse gas (GHG) emissions in 2030 com-
pared to those emissions in 2005 , but studies showed the opposite as only the manu-
facturing of the cells includes toxic chemicals, which increase the global warming
potential (GWP) (Ellingsen et al. 2014). Although manufacturing of the LIBS has a
lesser impact on the environment than manufacturing traditional ones, it still sig-
nificantly affects the environment and its effects on the environment will increase as
the estimated number of EV in 2050 is predicted to reach almost 2 billion and more.
With this large demand, the impact on the environment (GHG and the climate) will
dramatically increase (Ellingsen et al. 2014).

8.2.2 liFe cycle

The life cycle of a battery has a lifespan, and after a certain period of time, it needs to
be replaced with a new one. Several studies showed that a 26.6-kWh battery cell can
perform in an efficient way till reaching 80% of its initial capacity under the normal
use. And after reaching 80% of its initial capacity, the battery isn’t recommended to
be used in the car as it will drain fast. On the other hand, it can be reused in other
applications till attaining 65% of the initial condition, and then it can be recycled or
disposed away. Studies showed that the GWP for a driven distance is calculated by
dividing the impact of the total battery manufacture over the total distance covered
by the battery’s life while operating in the vehicle. So by assuming the battery cycle
number to be 3,000 cycles and supplying 0.5 MJ km−1 power, the assumption will
result in the emission of 11 gCo2 − eq kg1 (Ellingsen et al. 2014).
The life cycle of the battery depends on several stages such as material extraction,
processing of the material, manufacturing of the components, usage in car, battery
reuse, and end of life. The whole process, including the manufacturing, usage, recy-
cling, and disposal of the materials, shapes the life cycle of the battery, as shown in
Figure 8.1.

8.2.3 mAnuFActure
EVs are called the green vehicles as they have a zero emission impact on the environ-
ment, but studies showed the opposite (Ellingsen et al. 2014). Several studies showed
High Voltage Battery Life Cycle Analysis 185

FIGURE 8.1 The life cycle of the battery.

that the batteries used in the manufacturing of EV have minor effects on the envi-
ronment; on the contrary, the process of producing the battery itself is considered
the main burden to the environment and has a large impact on it (Notter et al. 2010).
An average of 4.6 tons of carbon dioxide is emitted into the environment, which is
caused only by the manufacturing of 26.6-kWh, 253-kg battery pack. Other studies
showed that not only the manufacturing of the battery but also the mining and refin-
ing processes of the material used are also the reasons for its impact on environment
(Romare & Dahll¨of 2017a).
Lithium ion is the mostly used in manufacturing of battery in the EV cars. Several
types of LIBS are used, such as lithium manganese oxide (LMO), lithium iron phos-
phate (LFP), and lithium manganese cobalt oxide (NMC). The different types of
batteries are attributable to the different materials of the cathode while using the
graphite as an anode material (Notter et al. 2010).
The manufacturing of the cell pack is made by assembling four components: cool-
ing system, battery cell, packaging, and battery management system (BMS). Each
component of the cell pack has several subcomponents, as shown in Figure 8.2.

8.2.3.1 Battery Cell

The manufacturing of a battery cell consists of two stages. The first is mining and
refining of the materials, and the second is assembling the cells altogether.

8.2.3.1.1 Mining of the Material

Mining of the materials used in the manufacturing of the battery cell is the first stage
of its life cycle since several materials are extracted from the ground. Various details
regarding the impact of these materials on the environment are found at different
databases. In this chapter, three databases are compared together: GaBi (Romare
& Dahll¨of 2017b), Ecoinvent (Dones et al. 2003), and GREET (Li et al. 2017). The
primary cobalt emits 90 and 8.8 kg Co2 – eq kg−1 in GaBi and Ecoinvent equiva-
lent. Moreover, the primary nickel emits 19.1 and 11.2 kg Co2 − eq kg−1 in GaBi and
Ecoinvent equivalent. The primary copper emits 2.7, 3–5, and 3–4 Co2 – eq kg−1 in
GREET, Ecoinvent, and GaBi.

8.2.3.1.2 Assembling the Battery Cell

While dissociating the battery cell, five subcomponents are found: anode, cathode,
separator, cell container, and electrolyte. The anode is made from a mixture of cop-
per current collector and a layer of negative electrode paste, which mainly consists
of graphite and a small amount of binders. The cathode is made from a mixture
of aluminium current and a layer of positive electrode paste (PEP), which mainly
186 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 8.2 Manufacturing and assembling a battery pack.

consists of positive material such as lithium manganese or NMC (Li et al. 2014). The
electrolyte is a mixture of solvents and a salt, mainly lithium hexafluorophosphate
(LiPF6). The separator, which acts as a filter, is made of porous polyolefin film to
achieve its function (Ellingsen et al. 2014). The container is made of many tabs and
pouch which surround the cell component. After adding all the components, the
welding of both copper tab and aluminium tab to the negative and positive current
collector, respectively, occurs.
Studies showed an industrial process of manufacturing a LMO which is used in
several cars such as Chevrolet Volt and Nissan Leaf. The study showed the manufac-
turing of a 24-kWh battery pack that consists of 192 prismatic cells, and the manu-
facturing processes include coating, baking, welding, mixing, cutting, and finally
assembling. A total of 88.6 GJ energy is used in the manufacturing process: 29.9
GJ is consumed by the material, and 58.7 GJ is utilized in the fabrication of battery
(Yuan et al. 2017).
Another study showed the percentage of the components used in the generation
of electricity that is needed for the manufacture of a battery cell: 46% coal, 33%
nuclear, 15% gas, 4.4% oil, and the remaining 1.6% (which is gained from renew-
able resources such as wind, hydraulic, and solar energy) (Notter et al. 2010). The
High Voltage Battery Life Cycle Analysis 187

manufacturing of the cell, PEPs, and negative current collector (NCC) has the most
significant effect on the environment with 56%–87% of the total effect caused by the
batteries. For the manufacturing of the battery cells, coal, natural gas, and uranium
are used to supply the energy needed, which is found to cause 51% of the battery’s
total GWP impact, 32% of the battery’s total fossil deficiency positional (FDP), and
31% of the battery’s total ozone depletion potential (ODP). After some analyses have
been conducted, the study showed the total impact of the manufacturing on 13 differ-
ent impact categories and the percentage of production of the cell (MOC), PEPs, and
NCC in each of these impact categories. All values are at the lower boundary value
(LBV) of the battery and listed in Table 6.1 (Ellingsen et al. 2014).
Some materials used in the manufacturing of the battery components are respon-
sible for most of the negative environmental impacts, such as magnesium and cop-
per, which are used in manufacturing of the PEP and NCC battery, respectively.
The copper, for example, results in the disposal of sulfidic tailings, which is the
reason behind the pollution of the most waterbodies, such as freshwater eutrophica-
tion potential (FEP), freshwater ecotoxicity potential (FETP), and marine ecotoxicity
potential (METP) with a percentage of 62%, 65%, and 54%, respectively (Ellingsen
et al. 2014).
Another study showed three different scenarios of generating energy for manufac-
turing the cells of the battery. In the first scenario, coal is used to generate electric-
ity for the manufacture of the battery, which increases the GWP of the battery by
exceeding 40%. In the second scenario, natural gas is used to generate electricity,
which has found to have corresponding carbon intensity as the current uses mixed
electricity. In the third scenario, hydroelectric power is used to manufacture the cell.
In comparison with the current GWP, the usage of the hydroelectric power decreases
the GWP by 60% and more, as shown in Table 8.1 (Notter et al. 2010).

8.2.3.2 Packaging
The process of packaging is divided into three processes. The first process is the
module packaging composed of inner and outer frames, connectors, and a cover. The
inner and outer frames are made of nylon cassettes surrounding the cells, which are
used to protect and support the battery cell. Besides every frame, an aluminium heat
radiator is found to make sure that the optimal conditions are achieved for the cells.
The battery retention system is used to keep the modules in place using foams, con-
strains, and leach to keep it inside the battery tray. After making all the modules, the
components are placed inside the tray made of steal and closed with a lid (Ellingsen
et al. 2014).

8.2.3.3 Battery Management System

The BMS has five subcomponents: battery module boards (BMBs), integrated bat-
tery interface system (IBIS), fasteners, high-voltage (HV) system, and low-voltage
system. For every module in the battery, one BMB is found and is used for monitor-
ing the voltage and the temperature of the battery cells. While the BMBs monitor
the cells, the IBIS controls the BMBs and the percentage of the HV. The HV system
contains cables, fuse, aluminium lid, and connectors. The low-voltage system con-
tains clips and straps (Romare & Dahllöf 2017a).
188 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 8.1
Total Impact on 13 Different Impact Categories
Impact Percentage Impact
Impact Units LBV Impact PEP NCC
GWP kg CO2-eq 4,580 62 9 2
FDP kg oil-eq 1,320 59.5 9 2
ODP kg CFC-11-eq 2.8E-04 43 12 3
POFD kg NMVOC 18 37 21 16
PMFP kg PM10-eq 16 22 29 29
TAP kg SO2-eq 51 22 36 26
FEP kg P-eq 8 13 7 64
MEP kg N-eq 6.4 7 42.5 6.5
FETP kg 1,4-DCB-eq 256 6.5 10.5 67.5
METP kg 1,4-DCB-eq 276 6 10 69
TETP kg 1,4-DCB-eq 1.3 6 25 50
HTP kg 1,4-DCB-eq 15,900 4 7 76
MDP kg Fe-eq 4,100 1 40 43

Impact categories: GWP, global warming potential; FDP, fossil depletion potential; ODP, ozone depletion
potential; POFP, photooxidation formation potential; PMFP, particulate matter formation potential; TAP,
terrestrial acidification potential; FEP, freshwater eutrophication potential; MEP, marine eutrophication
potential; FETP, freshwater toxicity potential; METP, marine toxicity potential; TETP, terrestrial eutrophi-
cation potential; HTP, human toxicity potential; and MDP, metal depletion potential.

8.2.3.4 Battery Pack Assembly

By adding all the systems and the modules together, the process of assembling the
pack is done. Besides all the energy needed in the manufacturing of the battery cells,
the systems, and the components inside it, additional energy is consumed in the
assembly process itself. Studies show that 0.3 GJ energy is used for assembling the
battery pack (Yuan et al. 2017).

8.2.3.5 Solutions to Minimize the Impact Due to Manufacturing

The studies showed that the manufacturing of the batteries consumes a lots of
energy, which is gained from burning the fossil fuel, which in turn affects the
environment. In contrast, if the manufacturing of the batteries is done with clean
energy, such as the hydroelectric, the impact of the manufacturing on the envi-
ronment will reduce. Another study showed that the impact done by manufacture
can be reduced by more than 60% if the energy needed is gained by the hydro-
electric power instead of the current electricity mix. Besides changing the way of
energy used to reduce the GWP, reducing the energy needed for manufacturing the
cell, recycling the materials, and increasing the lifetime of the battery are other
solutions to decrease the GWP. Impact can be reduced by these suggestions till it
reaches a point where EVs are very cleaner than the internal combustion engine
vehicles (ICEVs) (Ellingsen et al. 2014).
High Voltage Battery Life Cycle Analysis 189

8.2.4 bAttery liFe cycle AnAlysis with repurposing

in energy storAge systems (ess)

In this section, the life cycle analysis of the LIBS will be given. This will be done
taking into consideration the use of the batteries firstly in mobility solutions such as
their use in electric vehicles (EVs) and secondly in the smart grid solutions repre-
sented here as their use in the energy storage system (ESS).

8.2.4.1 First Use in Electric Vehicles

The first stage of the scenario proposed for the life cycle analysis will be discussed
as the mobility application of the batteries is meant for use in EVs after the manu-
facturing process.

8.2.4.1.1 Background
As people shift from petroleum-derivative vehicles towards EVs, they demand EVs
due to their cost-effectiveness, cleaner ways of generating electricity, and inexhaust-
ible nature, and also sustainable power sources can be met. Be that as it may, the exis-
tence of environmental impact cycle profile shifts from use-related emission weights
as the case is with internal combustion engines, to material mining, manufacturing,
and assembling in EVs progressively. In this light, commercial transportation and
personal vehicles account for around 10% of worldwide energy use and – conse-
quently – leading to harming emission discharges. However, as far as the full cycle
of supply and demand is concerned, the natural assets depletion and waste disposal
into the environment reach their utmost contribution at the production stage of the
EVs. This can be – to a great extent – the counterbalance contrast with fossil fuel-
powered internal combustion powertrain systems at the utilization stage (Granovskii
et al. 2006). However, the main disadvantages of the batteries are their high initial
costs and high material depletion potential considering the chemical and physical
compositions of all the metals used in the manufacturing of batteries. On top of
that, batteries fail to hold their efficiencies at an operable level for the EVs beyond
the 8-year mark. The efficacy of batteries drops to be as far as 80% of their original
effectiveness when they were primarily manufactured during those first 8 years of
their life cycle. That is why, those batteries with 80% efficiency should be reused in
a less-demanding environment than those used in mobility services. Here, the focus
will be on repurposing in energy storage utilities such as a stationary ESS. With this
in hand, spent EV batteries can be repurposed for reusing them. Although repur-
posing demands a limited attempt for dismantlement; analysis for deterioration and
malfunction; wrapping the batteries for a reuse; and attaching control systems, safety
systems, and electrical equipment to the repurposed packs, it is proven that this will
make the investment of energy and money in batteries of a better and greener return
of investment (ROI) (Ahmadi et al. 2017a, 2014).

8.2.4.1.2 Types of Capacities

Since batteries are the main source of power in EV given the electric power train,
the batteries need to be of matching capacity to the operational loads imposed on the
vehicle to be able to meet the mobility needs of the operators. This means that for the
190 Electric Vehicle Integration in a Smart Microgrid Environment

diverse mobility solutions offered in the markets, there is a meeting diversity in terms
of the types and capacities of the ESSs here referred to as batteries (Kushnir 2015).
As heterogeneous as this market is, there is something in common with the majority
of those types and capacities. This common facet between most of the battery types
is their reliance on (Li-ion) in the chemistry of battery. In this light, the previously
mentioned batteries can be manufactured using diverse amalgamations of materials
comprising both the anode and the cathode. Nevertheless, as far as it is a mobility
and transportation-related application in which the batteries will be used, the batter-
ies are evaluated according to certain criteria, the most important of which is power
and energy per kg (Kushnir 2015). This is why, some of those materials are preferred
by battery manufacturers more than others, as shown in Table 8.2.
On the other hand, anode materials are not as diverse as those of the cathode
materials. Most of those cathodes rely on graphite as their anode material.
Graphite is preferred as an anode material as it is cheap and relatively easy to
manufacture. Moreover, some synthetic compounds of graphite can be used as well.
Overall, graphite has a decent lifetime in the battery and is considered to have a
great value for its cost of excavating and manufacturing. Contrarily, lithium titanium
oxide is an alternative material for anode manufacturing that is found on a much
smaller scale than graphite. This alternative offers excellent power and life cycle.
Nonetheless, it operates at a lower voltage, and thus, it has lower energy content and
is more expensive than graphite indeed (Kushnir 2015).

TABLE 8.2
Overview of the Most Common Battery Chemistries Used in Manufacturing
Cathodes Along with their Advantages and Disadvantages (Kushnir 2015)
Cathode material Abbr. Use Advantages Disadvantages
LiCoO2 LCO Mainly in Performance, well Safety, uses nickel and
Lithium cobalt oxide small-scale understood cobalt
electronics
LiNi0.33 NMC Common in EVs Better safety and Cost, nickel and cobalt
Mn0.33 Co0.33 O2 (333) performance than
Lithium manganese LCO
cobalt oxide
LiFePO4 LFP High-power Excellent power, Low energy density
Lithium iron option, potential lifetime and safety,
phosphate choice for EVs abundant materials
LiMn2O4 LMO Historically used Cheap, abundant, Lifetime, low capacity
Lithium manganese in EVS, now less high power means low energy
oxide common density
LiNi0.8 NCA Used in some EVs High capacity and Safety, cost, uses nickel
Co0.15 voltage, high power and cobalt
Al0.05 O2
Lithium nickel
cobalt aluminium
oxide
High Voltage Battery Life Cycle Analysis 191

8.2.4.2 Second Use in Energy Storage Systems

For achieving the high efficiency, the batteries should be retained after being used
in EVs; research shows the implication for a second use of the batteries in order to
increase their utilization factor and account for the emissions and materials used in
their manufacturing. This repurposing use in ESS is proposed in the scenario of the
life cycle analysis.

8.2.4.2.1 Definition of ESS

ESSs can be defined as a complementary system added to a power generation source
(mainly intermittent) to serve as a substitute power supplier when the source fails to
meet the demands or loads imposed on it by the grid connected to it. Since there are
several ways to store this energy, ESS can be divided according to the form of the
stored energy inside the system: mechanical, electrical, and chemical (Ahmadi et al.
2017a, Aghajani & Ghadimi 2018). Here, the discussion will be about the repurpos-
ing of the spent LIBS as an electrochemical ESS. The main idea of electrochemical
ESSs involves converting electrical energy into chemical energy stored in the bonds
between reactants and the products formed with respect to the cathode, anode, and
electrolyte of the batteries. Thus, the chemical part of the system requires the reac-
tion of the – at least two – reactants existing in the system; then, the output of this
chemical reaction is electrons that are obligated to flow through a circuitry forming
a potential difference between the poles of the battery, which is referred to as voltage
difference or simply voltage (Kushnir 2015).

8.2.4.2.2 Power Peak Sheaving

Battery charge/discharge profiles are greatly connected with what is happening on
the subject of the primary power sources that in times when the supply is surpassing
the demand, the batteries are toggled to start charging off the excessive power found
on the system. On the other side, the batteries serve to discharge when the primary
power sources are outcompeted by the loads imposed on the system. In the scope of
the studies proposed, the depth of discharge of the batteries, which is the amount
of the total charge contained in the battery when it is fully charged with the amount
of current charge subtracted from it, is 75% (Ahmadi et al. 2017a, 2014). Power peak
shaving is the process of meeting the peak hour demands on the grid by supplying
energy through the ESSs. This can come in handy because of the price variations
imposed by electricity companies during the peak demand hours. This extra pric-
ing enables the utility company to increase the power capacity needed to supply for
the peak demand. Without the usage of ESS, this extra supply is generated using
older, more expensive, and less environmentally friendly power generation equip-
ment. The peak pricing also encourages customers to reduce demand in order to cut
down electricity costs. According to economists, the strategy of pricing is what gives
peak shaving an attractive outlook for organizations with large electrical consump-
tion during peak times (Olsson et al. 2018). In the scenario when this equipment is
replaced by battery-powered ESSs, the consumption of those organizations will be
moderated as they paid for power during low-utility pricing yet used it during the
peak pricing times. This can save a lot of money for the third-party organizations and
192 Electric Vehicle Integration in a Smart Microgrid Environment

also save the environment from the emissions produced by the obsolete equipment
used to meet the peak demands (Ahmadi et al. 2014, 2017b).

8.2.4.2.3 Use with Intermittent Power Sources and in Smart Grids

In today’s world, the reliance on intermittent power sources of renewable energy
such as solar and wind is on a hype. This can be contributed to many reasons; among
them are the current environmental status and the pressing demands for clean energy.
However, the reliability of those sources always comes to question as they have pre-
determined power cycles related to occasional phenomena such as the winds or the
sunshine. That is why, a complementary battery energy storage systems will increase
the reliability of those sources. In the same vein, an increased interest arises in the
use of microgrid systems. In the microgrids, a small network of fossil and renew-
able energy sources, battery ESSs, and loads can be connected. As mobile as this is,
microgrids can be utilized in inaccessible locations or when the pricing of the utility
provider is not feasible (Aghajani & Ghadimi 2018, Azaza & Wallin 2017).

8.2.5 environmentAl ApproAches For bAttery disposAl

In this section, the dimension and the disposal and recycling of the LIBS will be
discussed.

8.2.5.1 Introduction and Background Information

Based on the aforementioned environmental impacts of LIB disposal, it is a neces-
sity to search for an appropriate method for disposal. Spent LIBS contents of heavy
metals and toxic electrolyte exceed 50%, which varies from one manufacturer to
another; hence, land filling or burning is not considered as safe disposal method
since toxic substances find their way to underground waterbodies and atmosphere,
respectively. Environmental pollution as well as the loss of such significant amounts
of active metals is the trigger behind recycling research. 4,000 tons of spent LIBS
contains more than 1,100 tons of heavy metals. Thus, research has been conducted
to find the optimum recycling of spent LIBS. Some techniques showed a great poten-
tial to be effective, while some other techniques have already made their way to the
execution phase.

8.2.5.2 Currently Applied Recycling Techniques

The main purpose of the recycling processes is the extraction of heavy metals from
spent LIBS since they hold a great economic value. Procedures are mainly catego-
rized into three stages: pretreatment processes, metal extraction process, and product
preparation process. Pretreatment processes are mainly conducted to avoid short cir-
cuiting or spontaneous combustion. Meanwhile, the extraction process is considered
to be the most fundamental stage where metals are separated to ease their recovery
in the next step. After acid leaching, the obtained leachate contains large amounts
of metal ions, so solvent extraction, crystallization, and chemical precipitation are
conducted in order to recover these metals (Zheng et al. 2018).
High Voltage Battery Life Cycle Analysis 193

8.2.5.2.1 Pretreatment Processes

The target of pretreatment processes is not only avoidance of short circuits but also
to reduce waste: the better the separation, the more optimum the extraction will be.
At this phase, either mechanical separation or manual dismantling is used to disas-
semble the spent battery components after it is immersed in a salt solution although
mechanical methods are more common. Firstly, the plastic shell is removed so that
harmful substances are deactivated when the battery is immersed in liquid nitrogen.
Then, the end of the battery shell is removed, and anode, cathode, and separator
are separated. At the next step, they are dried in an oven, but further separation
is required before metal extraction; thus, several methods were developed for this
advanced separation, which are as follows:

194 Electric Vehicle Integration in a Smart Microgrid Environment

8.2.5.2.2 Extraction Processes

Metal extraction is the main process in battery recycling. Metals are transformed to
their metal alloys; the solution is formed as a step towards their recovery in the final
step. Several methods are presented to conduct this separation, but the three main
methods are found to be the most common and effective, which are as follows:

8.2.5.2.3 Product Preparation Processes

A. Recovery of Metals from Leachate
This stage starts with a variety of metal ions all dissolved in solution.
Several methods are there with the purpose of recovery and separation of
these metals, such as chemical precipitation, crystallization, and solvent
High Voltage Battery Life Cycle Analysis 195

extraction; usually, a combination of these methods is used in this process.

The main drawback of this stage is the large amount of reagents used. This
loss is considered to be unavoidable since impurities in the reagents signifi-
cantly affect the product purity (Zheng et al. 2018).

8.3 METHODOLOGY
In this section, real-life examples reflecting the importance of the usage of ESSs
are given in continuation to what was mentioned before. The methodology will be
divided into two sections: the use in power peak shaving applications and the usage
in developing smart grids.

8.3.1 power peAK shAving

Due to the variant consumer behaviour imposed by both individuals and entities
regarding the electrical power consumption, the rise in demand and fluctuations
in consumption caused a noteworthy variation in power peaks and valleys as seen
from the substation side. Factually, this has induced difficulties in the operation of
the grid, and this causes problems in controlling the stability and reliability of the
grid. Various techniques have been suggested to determine charging and discharging
quantities and their respective time periods in the quest to remove the peaks and val-
leys from the load configuration. These techniques depend on analysis methods such
as swarm particle optimization, dynamic programming, and nonlinear programming
(Rahimi et al. 2013). On the other hand, some of these techniques rely on resiz-
ing of the BESS to deliver optimal efficiency and power for boosting the financial
advantage by cutting the energy bill. In this section, a simple comparative approach
is proposed as an algorithm for power peak load shaving. The algorithm compares
the combined load profile with the mean loading profile over a specific time span;
the next step is then triggering a charge or discharge response from the batteries
according to the corresponding accompanying load profile. For the algorithm to be
functional, several parameters are needed to be defined, among which are “energy
bars” and “weighting factors”.
196 Electric Vehicle Integration in a Smart Microgrid Environment

8.3.1.1 A Sample of Current Simple Comparative Algorithms

Although the literature is rich with many algorithms solving the optimization
problems related to the power peak shaving using BESS, this algorithm proposed
in Dones et al. (2003) was chosen for its simplicity and straightforwardness. This
peak shaving algorithm is independent of both the system topology and the imposed
impedances of its wiring. Here, an assumption of a variable time window is taken
into consideration. Through a sequence of intervals of (τ) for their following utiliza-
tion periods (Up), the method iterates the calculation of the combined load profile
(Pav). Consequently, the contribution of each battery module in both charging and
discharging processes will depend on their individual capacity compared to the total
capacity of the system. In this light, the charging and discharging are estimated for
the next suitable time period (τ), which is calculated by the utilization factor (Uf).
Utilization factor (0Uf1) can be defined as the area utilized by energy bars (the tar-
geted area to be shaved) over the total area above (or below) the average load line, as
shown in the figure. The variable window approach mentioned earlier enables the
system to act in an up-winding approach for load peak shaving. This means that the
periodic calculation of (Pav) minimizes the effect of inaccurate load forecasting on
load peak shaving. In Figure 8.3, a schematic illustration of the method is pictured,
in which the usage time span (Up), energy bars, and time intervals of (τ) are depicted.
In the figure, (P) is the combined load drew from the power station. In the time when
(Uf = 1), the batteries will be toggled to charge or discharge. Contrarily when (Uf = 0),
the batteries will not be used at all. Meanwhile, (0Uf 1), the energy bars represented
will be the only times when charge or discharge takes place. The amount of charged
(discharged) energy from the arbitrary battery (m) can be calculated by:

  E1
Ebat,m = (  SOCmax,m −  SOCmin,m ) ×   , (8.1)
∑ i ∈S
Ei

where (S) is the set of all energy bars determined by (Uf).

From 8.1, the absolute charged or discharged power from the arbitrary battery (m)
at the grid side is as follows:

Ebat,m
Pbat,m = ηm ⋅  . (8.2)
τ
As previously stated, the power charged/discharged from (m) is calculated in equa-
tion 8.2. However, the power needs to be confined regarding two conditions to accu-
rately represent the real operating conditions. In equation 8.3, the first condition,
where the charged/discharged power of the BESS is limited to the maximum rate of
charge/discharge, is explained.

Ebat,m
Pbat,m ≥ . (8.3)
τ
The second case, as shown in equation 8.4, confines the state of charge of the bat-
tery (m) to its minimum (during discharge) or maximum (during charge) intervals,
respectively.
High Voltage Battery Life Cycle Analysis 197

FIGURE 8.3 Sample power consumption (load) distribution over time for the proposed
algorithm (Ahmadi et al. 2017a, 2014).

SOCmax,m ≥ SOCm ≥ SOCmin,m . (8.4)

Figure 8.3 depicts an immediate t of the intersection of the dotted line with the load
profile in order to calculate the energy bars used for calculations in equation 8.1. The
dotted line is formalized according to the utilization factor (Uf).

8.3.2 the proposed simple compArAtive Algorithm

Following the previous section, here the researchers will propose a novel compara-
tive algorithm for the determination of charge–discharge cycles of the batteries in the
ESS. An arbitrary power–load curve is going to be used to clarify the analysis and
determine the parametric equations of the important parameters.

8.3.2.1 Definition of Variables

where M(t) is the characteristic load–time curve.
(PH): The limit at which the load levels reach the limit that the obsolete equipment
pieces – of low efficiencies – are used. Thus, the cost of the energy unit – usually
KWH – increases after the (PH) mark.
(PL): The lower limit of the lowest valley in the load curve indicating the best time
to charge the batteries.
198 Electric Vehicle Integration in a Smart Microgrid Environment

C(t): The cumulative batteries charge corresponding to time (t).

(δ): The rush–peak time over charge factor.
(αgen): The GHG emission per unit energy generated from a specific power source.
(β Ghg): The values of the GHG emissions released into the atmosphere on gener-
ating power from a specific source.
(ηP): Energy generation efficiency in the peak.
(ηv): Energy generation efficiency in the valley.
(ηEss): The efficiency of the batteries.
We know from Ahmadi et al. (2017a, 2014) that the efficiency of the batteries
through their repurposing life 10 years in stationary ESSs ranges from:

65% ≤ ηEss ≤ 80% (8.5)

Time marks: (T1) start time of the peak, (T2) ultimate load time, (T3) end time of the
peak, and (T4) weighted average load correspondence time to the actual load.

∴  M ( T4 ) = Pavg  . (8.6)

Then (T4) can be determined by

T3 T7

T4   =
 
∫ T1
M ( t ) ⋅dt × T2 +
T7
∫ T5
M ( t ) ⋅dt × T6
. (8.7)
∫ T1
M ( t ) . dt

(T5) start time of the valley, (T6) minimum utilization time, and z(T7) end time of the
valley are shown in Figure 8.4.

8.3.2.2 Solution Flow

Although the previously stated algorithm is one of the easiest battery ESS’s control
algorithms in the literature, the algorithm nomenclature is targeting the research seg-
ment of the audience, which may make it hard to be understandable by the operators
or the general audience (Figure 8.5). This is why, a new simple comparative control
algorithm will be proposed here to further simplify the analysis and to highlight the
value proposition of BESS to the public. Inspired from Rahimi et al. (2013), (τ) will
be the proposed time step in order to iterate the calculations of the curve points to
be defined later.
Consequently, this will decrease the error due to inaccurate load forecasting.
Moreover, the algorithm works to be in a state of continuous comparison of param-
eters that the iterations will continuously check the value of the battery charge per-
centage and correspondingly determine the state of charge or discharge. This will
be further illustrated in the flowchart of the algorithm. Finally, the amount of money
and emissions saved by the utilization of the BESS will be calculated.
Algorithm since peak area to be shaved is defined by (T1), (T3),
High Voltage Battery Life Cycle Analysis 199

FIGURE 8.4 Arbitrary load–time curve.

FIGURE 8.5 Flowchart for the control algorithm.

T3
∴ Ap =
∫
T1
M ( t ) ⋅ dt . (8.8)

To make up for the excessive consumption at this peak, the batteries should charge
utilizing the area of the valley,
200 Electric Vehicle Integration in a Smart Microgrid Environment

T7
∴ Av =
∫ T5
M ( t ) ⋅ dt , (8.9)

where

T3 T7

∫T1
M ( t ) ⋅ dt =  ηEss ×
∫
T5
M ( t ) ⋅ dt. (8.10)

To calculate the environmental impact (GHGs) of generation of each energy area,

Peak area,

A
( βGhg ) p =  α gen ×   η p    (8.11)
P

Valley area,

( βGhg )v =  α gen ×   ηAv   

v (8.12)

∴ Environmental benefit = ( β Ghg ) p − ( β Ghg )v. (8.13)

8.4 
THE METHODOLOGY STUDY DESIGN
It is against the flow to have such a plan and aim to change the current system direc-
tion so, by having a good strategy and tools, it is achievable. Ahlborg and Hammar
(2014) and Painuly Painuly (2001) proposed a qualitative study design to collect data
in Tanzania and Mozambique, and made a literature review, interviews with power
sector representatives, and at-site monitoring. The literature review consists of six
categories negotiating the rural electrification (RE): institutions and representative’s
work; economy and finance; social dimensions; technical system and its manage-
ment; technology diffusion and adaption; and rural infrastructure. The represen-
tatives interviewing part covered seven subjects: technical, economic, and social
aspects of RE, off-grid, and renewable energy technologies (RETs): current state
of the electricity infrastructure in rural areas; RE strategies (including capacities of
both own and other organizations), roles, and relations between actors; institutional,
social, and economic drivers and barriers to RE; productive uses of electricity; poten-
tial for off-grid and renewable energy technologies; domestic involvement in electri-
fication processes; and effect of electricity on people’s lives (Painuly 2001).
Seventeen interviews were followed through with government officials, interna-
tional funding agencies, expert advisers, and the civil society organization (TaTEDO).
Depending on their impact in and expertise of RE processes. The accomplished tech-
nique relied on the open-ended questions, without neglecting the professional expe-
rience while constructing the questions. Site visiting was applied to the operating
off-grid systems using solar and diesel generators, and pico-hydro was in the favour
of discussions. The limited data was not included in the analysis. There were some
High Voltage Battery Life Cycle Analysis 201

weak spots in this study design, which are needed to be considered while it is applied
to other countries, such as the limited scope of analysis, the number of interviewees,
and time assigned to each interview.

8.5 FACTORS OF THE METHODOLOGY’S IDEAL ENVIRONMENT

The previously mentioned methodology has its own ideal conditions to be fully ben-
eficial and customized, which is explained in the study design. However, there are no
neglectable drivers, which form a base of support for both the study design and meth-
odology, and barriers that stand opposing the flow of the methodology sequence,
which plays a major factor in this whole process.

8.5.1 drivers
The power sector is mainly centralized and controlled by the government; the power
sector structure is expecting new actors and a change in the sector dynamics and
structure. The market actors are playing a major role in strongly affecting the driv-
ers and barriers in both of their organizational capacity. Describing the contribution
of the actors of power sector to RE development. Local initiatives, local demand,
and other drivers are followed by the policy at national level compiled, as shown in
Figures 8.6 and 8.7.
“According to interviews, the core drivers category for RE in both Tanzania and
Mozambique is political priorities” (Ahlborg and Hammar 2014). When it comes to
RE, it is an expected result since the government has a significant role in that field
in most of the countries. Also, the politics has its priority over many other things
on the frequently mentioned list of important aims such as better healthcare and
education. Electrification for some actors might seem to hold back the urbaniza-
tion and decrease the rate of birth in the rural areas. Regarding the use of financial
aids and the pro-poor policies, the connection rate in the rural areas increases by
the drivers, according to Akesson and Nhate (2006). As well, Akesson mentioned
the affordability for the targeted customers as it is under the required tariffs in the
case of no subsidized privet actors’ off-grid systems. As the focus in both coun-
tries was directed towards grid extension, Mozambican donor and consultant B is
claiming that the political interests are inflating in the fields of RETs and off-grid
electrification. It is considered as a driver when the donor (international and local
donors) support for RE is directed. However, a major part of the financial support
in the energy budget is provided by the international donors aiming for reforms. It
is noteworthy to mention that influential politicians advocate electrification for their
districts during the elections time. The study was divided into categories under the
drivers and barriers sections DS & BS. As Figures 8.6 and 8.7 are negotiating seven
major aspects and themes, from how currently the electricity infrastructure state is
through RE strategies, organizational, social and economic electrical power drivers
and barriers in rural areas, passing by the beneficial uses of electricity, the future
strength of technology regarding off-grid systems and renewable energy, how deep
is the local role in the electrification process and finally the alter of electricity on
people.
202 Electric Vehicle Integration in a Smart Microgrid Environment

The first component of the DS is policy and poverty mitigation ambitions drivers.
Figure 8.6 shows the part which the interviewed actors picked from there point of
view the effective actors are related to the government and policies drivers and the
mean of the answers points at the governmental policies as the main affecting driver.
In the second part of the DS, local initiatives drivers are being illustrated in Figure
8.6 as the interviewees showed generally an equal importance to all the listed driv-
ers. Except for the Churches installing or promoting RETs. That was not as expected
as the churches are a major implementor of RE from there projects to support rural
hospitals and communities. Following up with the DS, Figure 8.6 shows the local
demand DS as the collected data from the participants shows how effectively the
increasing demand drives the RE developments. The last part in Figure 8.6 demon-
strates other drivers which are not under a specific category: high cost of grid exten-
sion drives off-grid and required rising in sustainability in grid.

FIGURE 8.6 Mega diagram for the RE drivers.

High Voltage Battery Life Cycle Analysis 203

8.5.2 bArriers
Tanzania and Mozambique encountered similar obstacles towards RE, which
occurred because of many factors such as geographical, political, financial, and
social aspects. Figure 8.7 shows the barriers to RE. The “economic issues, funder

FIGURE 8.7 Mega diagram for the RE barriers.

204 Electric Vehicle Integration in a Smart Microgrid Environment

subordination, lack of private sector involvement and institutional and organizational

barriers” with a combination of other factors make up the barriers to grid exten-
sion, which is the reason why RE is difficult for both countries. Electrifying rural
areas is considered one of the sustainability plans as it has a beneficial impact on the
national economy, which is restricted by the grid extension of the RE in most of the
cases, according to Bugaje (2006). RE is dependent on international financing due
to the shortage of capital. Actors in Tanzania and Mozambique preview RE as being
held back due to poverty because it is donor dependent and there is a shortage of
fund as well as being rural poverty. The grid can be extended; however, this means
the payback interval will increase, while the population density in other progressing
countries within the vicinity is low and settlements are dispersed. Many users have
reported the conveyance over long distances, which has resulted in high cost and has
become a geographical barrier. Households are not able to afford the cost of trans-
mission as it equals to several months of income due to the undeveloped economy of
the rural area – this barrier is stated by Barnes (2010) and Haanyika (2008). Also,
a problem considering the traditional building methods raises as these houses are
unfit for electrification. Residences made out of clay and grass in Tanzania are not
recognized for linkage. Therefore, only 10% of the rural population can pay for the
connections and install electricity in their houses.
Customers with low income face problems in covering fees for connection, which
is an important barrier as stated by Kankam (2009) and Murphy (2001). As a result,
the government and donor projects provide help by using subsides and tariffs for the
certain users so as to increase the rate of connections. Although the tariffs are low, a
majority of people are not able to pay it. Also, rural users that are connected to the grid
can pay for more expensive commercial tariffs; on the other hand, Hindman Persson
asserts that the main trouble is rather the flow of cash, which needs to be solved using
the applicable payment strategy. The financial outlook for the power sector can be
developed by steadily increasing the tariffs; the only obstacle is that increasing tariffs is
a tricky political constrain. Furthermore, there are specimens of grid and off-grid proj-
ects where users pay the fees without financial aid, which is said by Akesson (2006)
and Ilskog et al. (2005). Another disadvantage for using variant tariffs while reducing
the fees for low-income groups is that the threshold value remains the expenditure rate
at the lowest margin. Power generation becomes less appealing to private properties
when subsidized tariffs are used. In Tanzania, the cost of grid electricity generation
has increased to double the present tariff; thereafter, its circulation has become a con-
tinuous loss, which caused TANESCO to not be able to afford its operation costs. As
a result, financial aid works as a barrier and a driver as stated by Kankam (2009) and
Thomalla et al. (2006). The costs of the supported linkage have been donor-driven
projects intrusive to projects paid by the government where people are demanded to
pay full cost of the connection fees. In addition, the objections were raised due to the
low payoff for property loss (such as farmland) alongside power lines, since donor-
driven projects dispense greater payoffs than governmental RE projects.
The upcoming diagram is handling the BS to RE by grid extension. Firstly, it dis-
cusses the barriers under the weak institutions and organizations category, and the
interviewees stated that the lake of organizational capacity has the main role among
the rest of the barriers; however, they did not neglect the management techniques
High Voltage Battery Life Cycle Analysis 205

that have been used in the energy sector, since it came in the second place. Economy
could be one of the highest affecting barriers if it is not the centre of them.
As it is known, economic interests can drive or stop a development process
in most of the fields, by putting the financials in the priority. As an output for
these practices, the rollers of the RE are witnessing many barriers under this
category. These barriers lead to the low productive and little industry coefficient,
followed by the high cost of diesel as Figure 8.7 shows. The study did show an
equal importance to the social dimension’s barriers as shown in Figure 8.7, as
it is a cornerstone in the study, and they picked the poverty and low household
affordability as major drivers in the drivers’ category. The category in Figure 8.7
has consensus on low maintenance culture and the truncate generation volume as
the highest affecting drivers. The rest are the average priority from there point
of view.
The research did include the spread of the technology in the energy industry, and
how it is getting usual for the new adapting communities, and the results revealed the
cultural mindset as the major barrier, but with fifty percent of the results for the lake
of entrepreneurial mindset as Figure 8.7 shows.
If it is not the reason, which can be related to cultural mindset driver? At the end,
the investment in rural areas infrastructure has been known for its lake of priority
from the governmental developments and investors. However, they cannot take the
blame for the scattered population as it will not serve as much people as the invest-
ment can cover, so it came at the first place to concord the results, as Figure 8.7
demonstrates.

8.6 
CASE STUDY
8.6.1 System Briefing
After developing the operational methodology for the smart grid in general, the BMS
is specific, and the next step is to test the system. To test the hypothesis and try to
quantify the efficiency of the system, a model scenario of a future city’s smart grid
during its daily cycle was studied. In this scenario, the residential and industrial
power demands are met through various renewable and nonrenewable power sources.
In addition, mobility load was taken into consideration in order to model various
mobility loads as well. This was done by forecasting the behaviour of different age
groups and customer segment, which was in turn modelled by battery charging loads
deployed at different times of the day to match their respective scenarios. In future,
the reliance on electric cars is projected to increase. Consequently, this added load
on the power grid can be used to alter the behaviour of the load curves as a form of
“demand-side management” (DSM) depending on the forecasted scenarios of the
users’ usage cycles.

8.6.2 System Parameters & Assumptions

In this exemplary grid, the subsystems are modelled to be either active (generating
power) or passive (grid power loads). However, the problem here is to optimize the
206 Electric Vehicle Integration in a Smart Microgrid Environment

operation of the grid to meet the power needs in the greenest way relying on the
renewable energy sources rather than the base load plant that runs on diesel.

• The residential metropolitan loads are modelled as simple resistive
loads for the household supplies and lighting.
• On the other hand, the industrial loads are modelled as electric motors,
which portray the driving force for the industrial machinery.
• Vehicle-to-grid systems for charging mobility solutions (V2G) is com-
prised of 125 vehicles each of (40 kW) rated power and a rated capacity
of (85 kWh) (Figure 8.8). The overall system efficiency is set to be 90%.
This mobility consumption is encompassed of 5 different user profiles.
For each user profile, there are a certain amount of vehicles assigned.
This is what will be used later to manage the demand side and help
make the overall consumption more dependent on renewable sources
rather than diesel fuels. The profiles are divided as follows:
1. Working professionals commuting to work and charging their cars
on the premises of their workplace. (35 cars)
2. Working professionals commuting to work but for longer distances
and charging their cars on the premises of their workplace. This
means a longer delay before recharging in the morning. (25 cars)
3. Working professionals commuting to work but with no access to
charging stations on the premises of their workplace. (25 cars)
4. Working professionals working from home, i.e., no morning com-
mutes. (20 cars)
5. Working professionals working on night shift. Morning commutes
are replaced by night commutes. (20 cars)
• Overall, the power loads sum up to a 15 MW representing the loads of
a medium consumption day during spring or fall seasons (Figure 8.9).

FIGURE 8.8 Exemplary smart grid with V2G capacity.

High Voltage Battery Life Cycle Analysis 207

FIGURE 8.9 Power loads.

B. Power Generation (Maximum Capacity)

• The base load of the grid is met by a typical steam power plant that runs
on diesel to produce a nominal power of 15 MW. However, the goal of
the grid management algorithm and the DSM of the mobility loads is to
minimize the reliance on the fossil fuel in order to produce less emis-
sions and greener electricity.
• In addition, a photovoltaic farm with an area of 8,000 m2 and an effi-
ciency of 10% is capable of delivering 8 MW of nominal power in day-
time. The dependency of the solar power production on the exemplary
solar irradiance and cloud shading among other properties were taken
into consideration while modelling the real conditions of an arbitrary
PV farm.
• Finally, a wind farm represents the other part of the renewable energy
sources. The wind turbine is capable of producing a nominal power of
4.5 MW. The wind profile is modelled by varying data points through-
out the day ranging from a maximum of 15 m/sec and a nominal value
of 13.5 m\sec.
• Collectively, the power output is modelled to be stepped down using a
transformer station 25 kV/600 V before it is stepped down once more
before getting to the usable form 220 V/60 Hz at the respective con-
sumption areas (Figure 8.10).

8.7 RESULTS
After simulating the performance of the arbitrary grid, comparison curves between
various power generation sources’ performance were plotted. As shown in Figure 8.11,
the load fluctuates according to the variation in the activities of the users during the
day. Collectively, those varying demands of varying peaks and valleys (at their cor-
responding times of the day) were met using all the power generation stations and
farms, respectively. The demand management system behaviour was apparent in peak
shaving using the renewable sources. This was done by maximizing the utilization of
208 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 8.10 Power generation (maximum capacity).

FIGURE 8.11 Load curve vs production curves through 24-hour duration.

wind and solar power (when available) to decrease the reliance on the diesel power
plant in order to decrease emissions and lessen the environmental impact of charg-
ing the batteries used for the mobility solutions. Consequently, it was proven that the
High Voltage Battery Life Cycle Analysis 209

system can be utilized to decrease the overall environmental footprint of reliance on

EVs for mobility rather than relying on fuel-powered vehicles. Although the analysis
was not discussed quantitatively – as a result for the arbitrary assumptions that will
render the results useless in specific real-life scenarios – the trends of the graphs
simulated should be reassuring enough to promote the conversion to electric mobil-
ity solutions.

8.8 CONCLUSION
In this chapter, the environmental impacts of the EV batteries’ repurposing and
disposal were meant to be examined. Firstly, this was done by framing the differ-
ences between the conventional cars and EVs in terms of the main energy source
and propulsion mechanisms. In addition, the current environmental status was
discussed to compare between the impacts of the different mobility solutions.
Secondly, a life cycle analysis of the batteries was made referring to research
from the literature about the environmental impact metrics and their quantifica-
tion for the EV batteries’ product life cycle. This evaluation was conducted on all
processes from the mining for the material till the disposal of the spent batteries
in addition to their complementary parts as the BMSs. Afterwards, repurpos-
ing applications for the Li-ion-spent batteries were discussed mentioning their
use in power peak shaving and smart grids. Moreover, a conclusive overview of
the recycling process of the batteries was given, passing through all the steps
and procedures involved. To wrap up, operational methodologies of the energy
management systems were discussed reflecting on models from the literature and
proposing a novel one. Finally, a study was conducted on the drivers and barri-
ers to RE in the sub-Saharan region using smart grids. This study design can be
transformed into a general formula to be applied on other cases taking most of
the aspects which are similar to many cases without even editing, and in other
cases, it will have to adapt as it is ready for any market or society or country to
be applied depending on percentages and ratios between the usage of different
energy sources.

ACKNOWLEDGEMENTS
We would like to thank Prof. Dr. Ahmed G. Radwan (Vice President for Research,
Nile University, Egypt) for his helpful efforts and continued supervision throughout
this chapter. Also, we express our sincere gratitude to Nile University for providing
us with the all the facilities and equipment used in the case study.

BIBLIOGRAPHY
Aghajani, G. & Ghadimi, N. (2018), ‘Multi-objective energy management in a micro-grid’,
Energy Reports 4, 218–225.
Ahlborg, H. & Hammar, L. (2014), ‘Drivers and barriers to rural electrification in Tanzania and
Mozambique–grid-extension, off-grid, and renewable energy technologies’, Renewable
Energy 61, 117–124.
210 Electric Vehicle Integration in a Smart Microgrid Environment

Ahmadi, L., Fowler, M., Young, S. B., Fraser, R. A., Gaffney, B. & Walker, S. B. (2014),
‘Energy efficiency of li-ion battery packs re-used in stationary power applications’,
Sustainable Energy Technologies and Assessments 8, 9– 17.
Ahmadi, L., Young, S. B., Fowler, M., Fraser, R. A. & Achachlouei, M. A. (2017a), ‘A cas-
caded life cycle: reuse of electric vehicle lithium-ion battery packs in energy storage
systems’, The International Journal of Life Cycle Assessment 22(1), 111–124.
Ahmadi, L., Young, S. B., Fowler, M., Fraser, R. A. & Achachlouei, M. A. (2017b), ‘A cas-
caded life cycle: reuse of electric vehicle lithium-ion battery packs in energy storage
systems’, The International Journal of Life Cycle Assessment 22(1), 111–124.
Akesson, G. & Nhate, V. (2006), ‘Study on the socio-economic and poverty impact of the
rural electrification project rib´au`e/iapala, nampula, mozambique’, Maputo: Swedish
Embassy in Mozambique.
Azaza, M. & Wallin, F. (2017), ‘Multi objective particle swarm optimization of hybrid micro-
grid system: a case study in Sweden’, Energy 123, 108–118.
Barnes, D. F. (2010), The Challenge of Rural Electrification: Strategies for Developing
Countries, Earthscan.
Bugaje, I. M. (2006), ‘Renewable energy for sustainable development in Africa: a review’,
Renewable and Sustainable Energy Reviews 10(6), 603–612.
Dones, R., Bauer, C., Bolliger, R., Burger, B., Faist Emmenegger, M., Frischknecht, R., Heck,
T., Jungbluth, N., R¨oder, A. & Tuchschmid, M. (2003), ‘Sachbilanzen von energiesyste-
men’, Final Report Ecoinvent 2000 6.
Ellingsen, L. A.-W., Majeau-Bettez, G., Singh, B., Srivastava, A. K., Valøen, L. O. & Strømman,
A. H. (2014), ‘Life cycle assessment of a lithium-ion battery vehicle pack’, Journal of
Industrial Ecology 18(1), 113–124.
Granovskii, M., Dincer, I. & Rosen, M. A. (2006), ‘Economic and environmental compari-
son of conventional, hybrid, electric and hydrogen fuel cell vehicles’, Journal of Power
Sources 159(2), 1186–1193.
Haanyika, C. M. (2008), ‘Rural electrification in Zambia: a policy and institutional analysis’,
Energy Policy 36(3), 1044–1058.
Ilskog, E., Kjellstr¨om, B., Gullberg, M., Katyega, M. & Chambala, W. (2005), ‘Electrification
co-operatives bring new light to rural Tanzania’, Energy Policy 33(10), 1299–1307.
Kankam, S. & Boon, E. K. (2009), ‘Energy delivery and utilization for rural development:
lessons from northern Ghana’, Energy for Sustainable Development 13(3), 212–218.
Kushnir, D. (2015), ‘lithium ion battery recycling technology 2015’, Current State and Future
Prospects; ESAReport 18.
Li, B., Gao, X., Li, J. & Yuan, C. (2014), ‘Life cycle environmental impact of high-capacity
lithium ion battery with silicon nanowires anode for electric vehicles’, Environmental
Science & Technology 48(5), 3047–3055.
Li, Q., Cai, H. & Han, J. (2017), Life-Cycle Analysis of Greenhouse Gas Emissions and Water
Consumption–Effects of Coal and Biomass Conversion to Liquid Fuels as Analyzed with
the Greet Model, Technical report, Argonne National Lab.(ANL), Argonne, IL.
Mebarki, A. (2017), ‘Civil engineering and urban planning’.
Murphy, J. T. (2001), ‘Making the energy transition in rural east Africa: is leapfrogging an
alternative?’ Technological Forecasting and Social Change 68(2), 173–193.
Notter, D. A., Gauch, M., Widmer, R., Wager, P., Stamp, A., Zah, R. & Althaus, H.-J. (2010),
‘Contribution of li-ion batteries to the environmental impact of electric vehicles’.
Olsson, L., Fallahi, S., Schnurr, M., Diener, D. & Van Loon, P. (2018), ‘Circular business mod-
els for extended EV battery life’, Batteries 4(4), 57.
Ordon˜ez, J., Gago, E. & Girard, A. (2016), ‘Processes and technologies for the recycling and
recovery of spent lithium-ion batteries’, Renewable and Sustainable Energy Reviews 60,
195–205.
High Voltage Battery Life Cycle Analysis 211

Painuly, J. P. (2001), ‘Barriers to renewable energy penetration; a framework for analysis’,

Renewable Energy 24(1), 73–89.
Rahimi, A., Zarghami, M., Vaziri, M. & Vadhva, S. (2013), A simple and effective approach
for peak load shaving using battery storage systems, in 2013 North American Power
Symposium (NAPS), IEEE, pp. 1–5.
Romare, M. & Dahll¨of, L. (2017a), ‘The life cycle energy consumption and greenhouse gas
emissions from lithium-ion batteries’, Stockholm. Zugriff am 23, 2017.
Romare, M. & Dahll¨of, L. (2017b), ‘The life cycle energy consumption and greenhouse gas
emissions from lithium-ion batteries’, Stockholm. Zugriff am 23, 2017.
Thomalla, F., Downing, T., Spanger-Siegfried, E., Han, G. & Rockstr¨om, J. (2006), ‘Reducing
hazard vulnerability: towards a common approach between disaster risk reduction and
climate adaptation’, Disasters 30(1), 39–48.
Yuan, C., Deng, Y., Li, T. & Yang, F. (2017), ‘Manufacturing energy analysis of lithium ion
battery pack for electric vehicles’, CIRP Annals 66(1), 53–56.
Zheng, X., Zhu, Z., Lin, X., Zhang, Y., He, Y., Cao, H. & Sun, Z. (2018), ‘A mini-review on
metal recycling from spent lithium ion batteries’, Engineering 4(3), 361–370.
 9 Charging Infrastructure
for Electric Taxi Fleets
Chandana Sasidharan
Alliance for an Energy Efficient Economy (AEEE)

Anirudh Ray
School of Planning and Architecture (SPA)

Shyamasis Das
Alliance for an Energy Efficient Economy (AEEE)

CONTENTS
9.1 Introduction: Background and Driving Forces ............................................. 214
9.2 Commercial Electric Taxi Fleets .................................................................. 216
9.2.1 Case Study: Uber Electric Vehicle Trial in London ......................... 216
9.2.2 Case Study: Ola Electric Mobility Pilot ........................................... 217
9.2.3 Key Findings from the Study of Fleet Operations............................ 217
9.3 Important Charging-Related Aspects of Electric Cars................................. 218
9.3.1 Battery Capacity and Range ............................................................. 218
9.3.2 Charger Capacity and Charging Time .............................................. 218
9.3.3 Factors Affecting Charging Time..................................................... 220
9.4 Charging Technologies for Electric Cars ..................................................... 222
9.4.1 EV Charging Standards .................................................................... 223
9.4.2 Charger Classifications Worldwide................................................... 225
9.4.3 Charging Technologies ..................................................................... 226
9.4.3.1 AC Charging ...................................................................... 227
9.4.3.2 DC Charging ...................................................................... 228
9.4.3.3 Wireless Charging.............................................................. 229
9.4.3.4 Battery Swapping ............................................................... 230
9.4.4 Charging Technology Trends............................................................ 231
9.4.4.1 Mobile EV Charging .......................................................... 231
9.4.4.2 Solar EV Charging ............................................................. 231
9.4.5 Charging Station Safety .................................................................... 232
9.5 Categories of Commercial Four-Wheeler Passenger Fleet ........................... 232
9.5.1 Ride-Hailing Fleet ............................................................................ 232
9.5.2 Corporate Fleet ................................................................................. 233

213
214 Electric Vehicle Integration in a Smart Microgrid Environment

9.6 Plausible Locations for Charging Electric Taxi Fleet................................... 233

9.6.1 Charging Facilities for Taxi Fleet ..................................................... 234
9.6.1.1 Public Charging Hubs for En route Charging.................... 234
9.6.1.2 Charging Facility at Public Parking Spaces ...................... 234
9.6.1.3 Captive Charging Facilities................................................ 234
9.6.2 Critical Factors for Siting Charging Facilities .................................. 235
9.7 Techniques for Locating Charging Facilities ............................................... 235
9.7.1 Analytic Hierarchy Process (AHP) .................................................. 235
9.7.2 Technique for Order of Preference by Similarity to Ideal
Solution (TOPSIS) ............................................................................ 237
9.7.3 Rationale for Using AHP and TOPSIS ............................................. 238
9.8 Configuring a Charging Facility for an Electric Taxi Fleet ......................... 238
9.8.1 Selection of Charging Technology ................................................... 238
9.8.1.1 Identification of Charging Technologies for Evaluation .... 238
9.8.1.2 Selection of Parameters for Decision-Making ................... 238
9.8.1.3 Deciding Relative Weights of Parameters ......................... 238
9.8.1.4 Ranking of Parameters ...................................................... 239
9.8.1.5 Preparing Decision Matrix ................................................ 239
9.9 Recommendations for Fleet Charging ..........................................................240
9.9.1 Public Chargers Are Required to Support Fleet...............................240
9.9.2 Role of AC Charging for Fleet..........................................................240
9.10 Grid Interaction and Integration of Fleet...................................................... 242
Nomenclature ......................................................................................................... 243
Works Cited............................................................................................................ 243

9.1 INTRODUCTION: BACKGROUND AND DRIVING FORCES

Electrification of commercial ride-hailing operations is identified as an effective way
to realize sustainable regional transportation (Rokhadiya, et al., 2019). Many cit-
ies, including London, California, Boston, New York, San Francisco, and Shenzhen,
have rolled out policies to support the replacement of internal combustion engine
(ICE) taxi vehicles (Nicholas, Slovik & Lutsey, 2020). The prime drivers for the
adoption of electric vehicles (EVs) are mitigation of greenhouse gas emissions and
decreased local air pollution. A study in New York found that the replacement of one-
third of city taxi fleets would lead to 18% reduction in CO2 emissions (Kettles, 2016).
In Beijing, electric taxi fleets were introduced in order to reduce vehicular emissions
(Merkisz-Guranowska & Maciejewski, 2015).
The transition to electric mobility is also an economical option for fleet operators
as it can lower their operating costs. The accrued savings becomes an important fac-
tor to support commercial EV adoption, as the average travel distances of commer-
cial fleets are higher than those of private vehicles. A study on ride-hailing fleets in
the United States showed that the payback period for commercial EVs is half of that
of private vehicles (Pavlenko, Slowik & Lutsey, 2019). This fact was also corrobo-
rated by an electric taxi pilot programme in India (Arora & Raman, 2019). A case
study on electric taxi operation in Stockholm showed that electrification led to lower
total cost of ownership (TCO) and slightly higher profitability than the investigated
ICE taxis (Hagman & Langbroek, 2019).
Electric Taxi Fleets 215

Charging infrastructure for electric taxi fleets is an area that requires special
attention to planning and strategy. As the average annual mileage of a commercial
EV is higher than that of a private vehicle, the requirement for charging would be
higher in the case of the former (Clairand et al., 2019). Jager et al. (2017) found that
250-km range is ideal for electric taxi operation in Munich and that an average of
22% of the time is used for charging. But there has not been a deep investigation on
the range of existing EV models and their suitability for electric taxi fleets.
Research has highlighted the importance of adequate charging infrastructure to min-
imize the number of detours needed for charging and revenue loss to drivers (Hagman &
Langbroek, 2019). Wang and Cheu (2013) studied the Singapore taxi problem and
proved that there is a direct correlation between the decline in visits to charging facilities
and increase in revenue per taxi. Wang, Cheu, and Balal (2016) continued the research
and found that though increasing charging facilities was beneficial, the utilization of
charging facilities was not the same. Their study found that utilization of a charging
station depends on location and geographical distribution of customer demand. Hence,
it is important to select appropriate locations for fleet charging. Jia et al. (2017) trans-
formed the charging station allocation problem to a location problem and proposed a
data-driven model for the allocation of charging stations, which was tested against the
data of commercial taxis in Beijing. As data asymmetry is a practical challenge, there is
a need to develop simpler yet robust tools for the selection of locations.
Moreover, taxi fleets can be divided into two segments: ride-hailing fleets and cor-
porate fleets. Demand for clean employee transport is leading to the electrification of
corporate fleets, but there is hardly any research on the charging needs of these fleets
(WBCSD, 2019). The operations of these two types of electric taxi fleets differ based
on the purpose of use/ mobility, origin and destination points of the trips, trip attraction/
generation models, service/business catchment area, etc. As the characteristic features
of both the fleets are different, there is a need to assess the requirements for corporate
fleet electrification. Both electric taxi fleets have increased public charging infrastruc-
ture utilization in some markets (Jenn, 2019). The charging needs will be dependent on
their operational patterns, and consequently, charging preferences of fleets is a key fac-
tor while planning of charging infrastructure (Das, Sasidharan & Ray, 2020).
The multiple dimensions make fleet charging a complex riddle to solve. The
existing research on fleet charging options has primarily focused on fleet optimi-
zation and allocation of charging stations. The need of the hour is a comprehen-
sive understanding of the charging-related aspects of EVs, and linking them with
the operational characteristics of fleets. This chapter is dedicated to answering
the key questions related to the shift to electric technology, particularly related to
charging.

1. Where are EVs currently being used as taxis? What are the learnings from
fleet operation with EVs?
2. What are the important charging-related aspects of EVs? What are the fac-
tors that affect the charging time?

216 Electric Vehicle Integration in a Smart Microgrid Environment

Though there is no dearth of literature on EV charging and taxi fleets, there are no
studies that can comprehensively answer all of the above questions. Most of the exist-
ing works are focused on solving different aspects of the fleet charging conundrum
in isolation. This chapter presents a holistic overview of the multiple dimensions of
the charging puzzle, including transport characteristics, charging technologies, and
practical tools that can be effectively used. Firstly, a few instances of commercial
electrical fleet operations are studied.

9.2 COMMERCIAL ELECTRIC TAXI FLEETS

Europe has been the focal point for the introduction of many taxi fleets. The com-
mon denominator for taxi fleet electrification in all places has been the establish-
ment of charging facilities. One of the first taxi electric fleets for Europe was set
up in Amsterdam in 2011, supported by the subsidy for the procurement of EVs.
Special taxi charging facilities were set up for the taxis apart from the normal public
charging points. Subsequent to that, the city of Rotterdam introduced electric taxis
fleets along with 500 charging facilities. In 2014, electric taxis fleets were introduced
in Zurich along with a network of fast charging stations (Merkisz-Guranowska &
Maciejewski, 2015; Nicholas, Slovik & Lutsey, 2020).
London has launched a charging facility plan to support the fleet electrification. The
city has adopted a two-pronged approach with dedicated taxi charging. It has already
dedicated 25% of existing public charging facilities for taxi fleets and is keen on install-
ing more on-street fast charging infrastructure in priority locations. The locations are
identified in commercial hubs based on a mapping of taxi driving patterns and the
capacity of the electricity grid. London is also planning to launch ultrafast charging sta-
tions for electric fleets (Nicholas, Slovik & Lutsey, 2020; Hall & Lutsey, 2020).
Many of the Chinese cities are currently the front runners in electric taxi fleets.
Electric taxis was initially introduced in Beijing and Shenzhen, but it has spread over
to multiple cities across the nation. Most of the Chinese fleets comprise home-grown
electric BYD e6 models. In China, the EVs in taxi fleets are charged twice a day and
travel 360–500 km daily. Pilot projects are launched in many North American and
South American cities, including New York, Mexico, Rio de Janeiro, and Sao Paulo
(Hirschfeld, 2019; Merkisz-Guranowska & Maciejewski, 2015; Teixeira & Sodre, 2017).
The two case studies on the EV taxi fleet in the UK and India, and their findings
are presented below.

9.2.1 cAse study: uber electric vehicle triAl in london

Uber conducted an electric taxi trial in London with 50 EVs between August 2016
and January 2017. Three different car models, namely Nissan Leaf, Tesla Model
S, and BYD E6, were used in the trial, and typically, drivers would recharge
with 10%–40% remaining charge. The major challenges throughout the trial were
range anxiety and lack of adequate fast charging points. The drivers found the
Electric Taxi Fleets 217

charging network inadequate with respect to the numbers and its distribution.
They found that the time taken for waiting at charging stations had a compound-
ing effect on charging time. Ninety-nine percent of drivers were dependent on
public charging facilities for charging, whereas home charging accounted for only
26%. Still, one of the important aspects of the trial highlighted is that home charg-
ing was a supportive factor for taxi fleet adoption. The ability to charge at home
allowed 40% of drivers to charge at domestic electricity rates, thus minimizing
the cost of charging. Most of the charging happened at on-street parking as the
majority of partner drivers lacked access to off-street parking. In summary, driv-
ers preferred to use public fast charging facilities supported with slow charging at
their homes (Lewis-Jones & Roberts, 2017).

9.2.2 cAse study: olA electric mobility pilot

In May 2017, Ola Electric started an EV taxi trial in the city of Nagpur, India. The
EV used in the trial was Mahindra e2o with 15-kWh battery and a travelling dis-
tance range of 100 km. The average daily travel distance for the cars was 142 km,
and typically, the cars underwent a fast charge and slow charge cycle in a day. They
found that EVs usually have 20%–25% less running time compared to ICE vehicles
because of the time taken for charging. In the hot summer season in India, the situ-
ation worsened as the time taken for fast charging was impacted by ambient tem-
perature. When temperatures increased over 45°C, the charging time increased from
1.5 to 3 hours. Ola established a network of charging stations at times partnering with
conventional fuel pump operators. The power demand at charging stations peaked
twice a day, which was coincident with a lean traffic demand period. Between the
faster and slow charging options offered in public charging facilities, drivers pre-
ferred the fast charging option. Similar to the London trial, home charging was also
a strategy used in this trial. Ola assisted in setting up of charging points at driver
partner residences for overnight charging (Arora & Raman, 2019).

9.2.3 Key Findings From the study oF Fleet operAtions

The major observations from studying the fleet operations are as follows:

• In all cases of electrification, public charging infrastructure is a key compo-

nent to successful fleet operations.
• For fleet operation, drivers preferred the faster charging options at public
charging facilities.
• Dedicating a percentage of charging facilities for taxi fleet is a good tactic.
• The best strategy for fleet operations includes the provision of home charg-
ing facilities at driver homes.
• Charging demand might peak during a lean transport demand time and
could lead to an increase in waiting time at charging facilities.
• The choice of EV has an impact on the range and the charging time. The
driving range of EV and the travel distance determine the charging patterns.
In the next section, this interrelationship is explored.
218 Electric Vehicle Integration in a Smart Microgrid Environment

9.3 IMPORTANT CHARGING-RELATED

ASPECTS OF ELECTRIC CARS
The choice of an ideal electric car for a taxi fleet will depend on the techno-economic
feasibility of using an EV. While operating an electric fleet, charging time of the elec-
tric car represents a loss of revenue for the taxi fleet operator (Arora & Raman, 2019).
Hence, the charging-related aspects are of prime importance for choosing a suitable
electric car. In this section, a brief synopsis of the factors that impact the charging-
related aspects of the electric car models in the market is presented. The battery capac-
ity and energy efficiency together determine the range and charging need of an electric
car. To plan charging, other parameters, including the battery chemistry, voltage, and
maximum charging rate, are also relevant. Additionally, the factors including the capac-
ity of the on-board chargers are an essential consideration to plan charging. Mapping
of the important charging-related aspects of major electric car models available in dif-
ferent geographies is undertaken. The section summarizes the charging-related aspects
by studying the specifications of over 45 electric car models available in major geogra-
phies of the world. It should be noted that there are disparities in the global adoption of
EVs, and not all EV models are available in all geographies.

9.3.1 bAttery cApAcity And rAnge

One of the key parameters considered to select an electric car for a taxi fleet is the
driving range of the vehicle. The range of an EV represents the distance it can travel
in a single charge. The range of a car is dependent on the battery capacity. The range
of electric cars available in the world is anywhere between 75 and 525 km, and bat-
tery capacities range between 11 and 100 kWh, as shown in Figure 9.1. The ratio
between battery capacity and range is a proxy indicator of the energy efficiency for
the electric car. For the cars studied in this research, the energy efficiency is esti-
mated to be between 14.7 and 25.7 kWh/100 km. It should be noted that the actual
range of an EV will depend on the operating conditions such as ambient temperature
and traffic conditions. A study on EV adoption showed that the EVs with higher bat-
tery capacity (80–100 kWh) are preferred for taxis in markets where these models
are available (Clairand et al., 2019). In cost-sensitive and nascent markets like India,
the choice of EVs is limited, and the selected models have a battery capacity under
20 kWh (Das, Sasidharan & Ray, 2020).

9.3.2 chArger cApAcity And chArging time

Charging time is an important consideration with respect to operation of an elec-
tric taxi fleet. Almost all electric cars have an on-board charger that facilitates the
basic charging requirement of the vehicles. However, the power of these chargers is
typically very low, and this on-board charging power corresponds to the maximum
charging time reported by the manufacturers. Manufacturers also specify the maxi-
mum charging power for the EV models, which corresponds to the minimum charg-
ing time. Maximum charging power associated with the on-board charger of EVs
studied is shown in Figure 9.2.
Electric Taxi Fleets 219

FIGURE 9.1 Mapping of battery capacity, efficiency, and range of electric cars.

FIGURE 9.2 Mapping of charging capacity of electric cars.

It is observed that as the size of the battery pack in an EV increases, both on-board
charger and maximum charging power increase as seen in Figures 9.3 and 9.4. Most
of the on-board chargers have capacities under 11 kW. In the case of the maximum
charging power, a key observation is that there are a few car models currently with
a charging power over 150 kW. Typically, the maximum charging power needed to
charge the existing electric cars is less than 100 kW. With the advancement in tech-
nology, this trend may change.
The minimum and maximum charging times of the cars provided by manufactur-
ers are presented in Figure 9.5. The maximum charging time can be correlated with
the on-board charger output. It can be seen that 75% of cars take 3–9 hours for a full
220 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 9.3 Scatter plot of battery capacity against on-board charging power.

FIGURE 9.4 Scatter plot of battery capacity against the maximum charging power.

charge using on-board chargers. The minimum charging time for 80% of electric
cars is less than 60 minutes. This implies that a driver might need to wait for an
hour to charge an electric taxi. As charging time is a vital consideration, the factors
impacting the charging time are investigated in the next section.

9.3.3 FActors AFFecting chArging time

Estimation of charging time of EV batteries is a challenging exercise as the charg-
ing time depends on the charger output power as well as the battery characteris-
tics. Charging is often described as ‘slow’ or ‘fast’. Typically, the fast chargers have
Electric Taxi Fleets 221

FIGURE 9.5 Snapshot of charging time of electric cars.

higher power output rating and are capable of charging EVs under an hour. On the
other hand, slow chargers with lower power output often take six hours or more to
charge an EV. Still, it is not possible to demarcate fast and slow chargers based on
the level of output power of chargers alone. A 50-kW rated charger could be a fast
charger for an electric car with a battery of 40 kWh as it would be able to charge the
car within an hour. At the same time, the same charger would take more than two
hours to charge an electric car with a 100-kWh battery. Hence, it is better to under-
stand charging from the perspective of the battery. Batteries themselves are complex
energy storage devices wherein the available energy, the chemistry, configuration
within the battery pack, and the safe operating limits would determine the charging
rate and time (Sasidharan, Ray & Das, 2019).

a. Battery chemistry. Electric car batteries can have different material chem-
istries even within lithium-ion type of battery. The common battery chem-
istries that are found in electric four-wheeler batteries are lithium nickel
manganese cobalt oxide (NMC), lithium nickel cobalt aluminium oxide
(NCA), lithium manganese oxide (LMO), and lithium iron phosphate (LFP)
(Coffin & Horowitz, 2018). Each battery chemistry has its own set of advan-
tages when it comes to charging.
b. Battery pack configuration. The three most common design variants of lith-
ium-ion batteries are cylindrical, pouch, and prismatic cells. Each design
has an implication on the thermal management of the battery. The electric
car batteries are generally designed as high-voltage packs, with the voltage
rating above 110 V. However, in India, there are electric cars in the mar-
ket, which have low-voltage battery packs below 110 V. The battery voltage
determines the voltage of the battery charger and hence is an important
aspect for planning charging.
222 Electric Vehicle Integration in a Smart Microgrid Environment

To summarize the discussion, the factors affecting charging time are battery chemistry,
configuration, C-rate, and state of charge. The next piece of the puzzle is identifying
the charging technologies for electric taxi fleet, which is dealt with in the next section.

9.4 CHARGING TECHNOLOGIES FOR ELECTRIC CARS

One of the most important questions that need answering is about the most ideal
charging technology for the fleet. There is a need to study charging technologies
in detail. EV chargers currently deployed worldwide for charging electric cars are
quite diverse in their method of electricity transfer, power output levels, control and
communication capabilities, etc. (IEA, 2018). A mapping of the important param-
eters of commonly available chargers for over 40 EVs in the market is presented in
Figure 9.6. It is evident from the analysis that most of the chargers available in the
market are designed considering the maximum charging rate possible for the battery
packs used in the available electric car models (Refer to Section 9.3.2). Most of the
chargers for EVs have an output power between 3.3 and 200 kW. Typically, the output
current for most of the chargers in the market is between 32 and 200 A, though the
maximum current is 500 A. Most of the chargers are designed to have low-voltage
(<1,000 V) alternating current (AC) inputs and provide either direct current (DC) or
AC output. In the case of AC output chargers, there is not much difference between
input and output voltage levels. However, the most DC output chargers offer output
voltages between 130 and 720 V DC. In order to make an objective assessment of
charging technologies, the existing standardization practices for charger needs are
examined in the next section (Sasidharan, Ray & Das, 2019).
Electric Taxi Fleets 223

FIGURE 9.6 Mapping of major parameters of electric car chargers in market.

9.4.1 ev chArging stAndArds

There is no universally accepted standard for EV charging – the reason for which
could be attributed to the global variations in EV and electricity grid design. One
can observe standardization in prominent EV markets, including the United States,
Europe, China, and Japan. Multiple international organizations are invested in the
development of standards for EV charging these markets. One of the premier insti-
tutions associated with standards development is the International Electrotechnical
Commission (IEC). The IEC 61851 standard is a popular standard associated with
the conductive charging of EVs. IEC is also in the process of making standards for
battery swapping and wireless charging.
China is one of the major EV markets, which has developed its own standards for
EV charging, which are commonly called GB/T. The country also has a pivotal role
in many of the recently announced EV standards for wireless charging (CEC, 2020)
and battery swapping (Bloomberg, 2020). China also plays a key role in the develop-
ment of a high-power DC charging standard dubbed ‘ChaoJi’, which is an updated
version of the popular DC charging protocol CHAdeMO (2020). The CHAdeMO
protocol developed in Japan, the first fast charging standard, is already codified as
part of two international standards: IEC 61851 and IEEE Standard 2030 (CHAdeMO,
2015). There are also standards covering other important aspects for EV charging,
including communication, security, and interoperability. For example, ISO 15118 is
an international standard that is associated with the digital communication protocol
important for vehicle grid integration.
The Society of Automotive Engineers (SAE) is the major EV charging standards
organization in North America. All major vehicle and charging system manufactur-
ers in the United States follow their standards. SAE J1772 and J1373 standards are
associated with conductive and inductive charging, respectively. SAE J2931 covers
and establishes the security requirements for digital communication between EV,
EVSE (electric vehicle supply equipment), and the metering equipment. Details of
the common standards associated with EV charging are given in Table 9.1.
224 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 9.1
Summary of EV Charging Standards
Standard Description
IEC 61851 Electric vehicle conductive charging system
Part 1: General requirements
Part 21: Electric vehicle requirements for conductive connection to an AC/DC
supply
Part 22: AC electric vehicle charging station
Part 23: DC electric vehicle charging station
Part 24: Digital communication between a DC EV charging station and an electric
vehicle for control of DC charging
IEC 61980 Electric vehicle wireless power transfer (WPT) systems
Part 1: General requirements
Part 2: Specific requirements for communication between electric road vehicle
(EV) and infrastructure
Part 3: Specific requirements for the magnetic field wireless power transfer systems
IEC 62196 Plugs, socket outlets, vehicle connectors, and vehicle inlets – conductive charging
of electric vehicles
IEC 62840 Electric vehicle battery swap system
Part 1: General and guidance
Part 2: Safety requirements
GB/T 20234 Conductive charging of electric vehicles
Part 1: General requirements
Part 2: AC charging coupler
Part 3: DC charging coupler
GB/T 38775 Wireless charging system for electric vehicles
Part 1: General requirements
Part 2: Communication between on-board chargers and charging equipment
Part 3: Special requirements
Part 4: Electromagnetic environment limits and test methods
GB/T 29317 Terminology of electric vehicle charging/battery swap infrastructure
GB/T 27930 Communication protocols between off-board conductive charger and battery
management system for electric vehicle
SAE J1772 Electric vehicle and plug-in hybrid electric vehicle conductive charge coupler
SAE J1773 Electric vehicle inductively coupled charging
SAE J2847 Communication for smart charging of plug-in electric vehicles
SAE J2836 Wireless charging communication for plug-in electric vehicles
SAE J2954 Wireless power transfer for light-duty plug-in/electric vehicles and alignment
methodology
SAE J2931 Security for plug-in electric vehicle communications
SAE J3068 Electric vehicle power transfer system using a three-phase capable coupler
SAE J3105 Electric vehicle power transfer system using conductive automated connection
devices
ISO 15118 Road vehicles – vehicle-to-grid communication interface
ISO 19363 Electrically propelled road vehicles – magnetic field wireless power transfer
– safety and interoperability requirements
Electric Taxi Fleets 225

9.4.2 chArger clAssiFicAtions worldwide

One of the commonly recognized characterizations worldwide for EVSEs is catego-
rized into levels, models, and types.

TABLE 9.2
EVSE Levels
EVSE Levels Maximum Power (kW)
Level 1 2.5
Level 2 24
Level 3 50
Level 4 150
Level 5 350

1 As defined by the Infrastructure Working Council formed by Electric Power Research Institute
(EPRI) and subsequently codified in the National Electric Code (NEC) under article 625
2 The modes are defined in the international industry norm DIN/ IEC 61851
3 In US the Levels were defined first based on output power and the associated installation and safety
requirements were specified later. But in Europe, the output power, installation, communication and
protection are used in defining the Modes of charging.
226 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 9.3
Charging Modes
Modes Description
Mode 1 AC charging in households at normal mains outlets without no protection devices in the
charging cable
Mode 2 AC charging from normal mains with semi-active connection between EV and EVSE.
Charger cable with integrated safety devices in an in-cable control box comprising residual
current device, control pilot, and proximity detection.
Mode 3 AC charging at charging stations with active connection between EV and EVSE. No in-cable
control as the safety equipment is a permanent part of the charging station required in the
cable.
Mode 4 DC charging at charging stations with an active connection between EV and EVSE.
Charging system can adopt the charging currents and voltages to suit battery requirements.

proximity detector, and control signalling between the in-cable control box
and the EV. However, under this mode, the charger does not receive energy
feedback from the vehicle. Mode 3, on the other hand, is an advanced charg-
ing mode for an EV as there is energy feedback. A Mode 3 charger can con-
trol the rate of the charging and hence enables smart charging functionality.
The charging is done from a fixed outlet or a tethered cable that is capable
of continuous control and communication between the EV and the charger.
All DC charging, where the charger is located outside the EV, is classified
under Mode 4. It is seen that with the progressive increase in charger power
output from Mode 1 to Mode 3, the associated communication and protec-
tion protocols get more complex (Vesa, 2019; CCS, 2019).
d. Types. Chargers are classified into ‘types’ based on the types of the connec-
tors used. In the United States and Japan, Type 1 connectors that conform
to SAE J1772 standards are used for AC charging. In Europe and other
geographies, Type 2 connectors that conform to IEC 62196 standard are
commonly used for AC charging. In addition, there are four types of con-
nectors associated with the most popular DC charging protocols: combined
charging system (CCS), CHAdeMO, GB/T, and Tesla.4 In the case of CCS
and Tesla connectors, the same type of connectors can be used for AC and
DC charging. For all the other protocols, AC charging and DC charging
have two different sets of connectors. All types of connectors have power
pins, earthing pin, and control pins (CHAdeMO, 2015; CCS, 2019).

9.4.3 chArging technologies

There are primarily three technologies for charging of EVs. Charging of EV batteries
can be performed through a wired connection, i.e., by conduction, or wirelessly. In
the case of conductive charging, the vehicle is supplied with either AC or DC. Battery
swapping is the third charging technology where a fully charged battery replaces a
4 These protocols are described in detail in DC charging.
Electric Taxi Fleets 227

FIGURE 9.7 Categories of charging technologies.

depleted battery. These charging technologies are presented in Figure 9.7, and more
details on the categories are captured in the subsequent sections.

9.4.3.1 AC Charging

An EV is charged by conduction by plugging it into EVSE. If the EV has an on-board
charger, it is charged by plugging in into an AC power socket. The charging tech-
nology associated with this type of charging is called AC charging, as AC power is
supplied to the vehicles. AC charging is commonly categorized into two levels: Level
1 and Level 2 (refer to Table 9.4). For a taxi fleet, better classification types are home
and public charging.

9.4.3.1.1 Home
 Charging
Level 1 charging at 120 V and Level 2 charging at 230 V are commonly used for
charging EVs at residences. The wall-mounted domestic socket provides a small
amount of power to charge the EV (refer to Table 9.4) and results in prolonged charg-
ing times. However, this charging technology is relevant for a taxi fleet, as it enables

TABLE 9.4
Techno-Economic Specifications of AC Chargers
Specification Level 1 Level 2
Input voltage AC (V) 120 230 380–480
Output voltage AC (V) 120 230 380–480
Maximum output current (A) 20 32 63
Output power range (kW) 1.4–2.4 3.3–7.4 11–43
Typical charger cost ($) 225–790 1,100–4,500
Installation and grid connection cost ($) 0–500 5,000
228 Electric Vehicle Integration in a Smart Microgrid Environment

the cheapest charging for EVs using the existing electricity connection at residences
of drivers. Case studies on EV charging also show that home charging is encouraged
by most fleet operators. The cost estimates of the charger and installation are also
shown in Table 9.4 (Spöttle, et al., 2018; Hall & Lutsey, 2020).

9.4.3.1.2 Public Charging

AC public charging facilities operate at Level 2 charging as shown in Table 9.4.
Typically, all types of AC charging above 11 kW require the three-phase AC connec-
tions in most geographies. Though it is possible to charge vehicles till 43 kW using
this method, the limiting factor for this kind of charging is the capacity of the on-
board charger. For example, the on-board charger power of EVs in Europe is 22 kW,
and the most common AC EVSEs in Europe are 480-V, 22-kW, three-phase chargers.
High-power AC chargers may not be built in future for electric cars as there are no
new electric car models with on-board charging power more than 22 kW. In case of
India, most EVs have on-board chargers of 7.4 kW or less. Hence, most of the AC
chargers in India are powered from single-phase 230-V supply. In China also, most of
the public charging piles offer Level 2 charging at 220 V, 32 A. This type of charging
is generally Mode 2 or Mode 3, and charging is performed with a protection and pilot
control function (Spöttle, et al., 2018). The cost estimates of the EVSE and installa-
tion are shown in Table 9.4 (Spöttle, et al., 2018; Hall & Lutsey, 2020; SGCC, 2013).

9.4.3.2 DC Charging

DC charging is performed by supplying DC power directly to the battery as chargers
are not on board in the EVs. The power output level of DC chargers is only restricted
by the safe power limit the battery can accept. Thus, with the improvement in battery
technology, the power output levels of DC chargers are on the rise. The existing DC
charging facilities offer 50-kW power, whereas the upcoming facilities are 150 kW,
350 kW, or more. The costs of charging facilities also increase significantly with the
increase in charging power. 150-kW and 350-kW charging facilities are typically 2
times and 5 times costlier than a 50-kW charging facility. The technical specifica-
tions and cost estimates for DC chargers currently available in the market suitable
for charging are presented in Table 9.5 (Spöttle, et al., 2018; Hall & Lutsey, 2020).
As the charger is not colocated with the battery, the communication between the
charger and the battery is of prime importance in DC charging. DC charging comes
with standardized protocols governing communication and connectors. One of the

TABLE 9.5
Techno-Economic Specifications of DC Chargers
Specification Level 3 Level 4 Level 5
Output voltage range DC(V) 200–500 200–500 200–1,000
Maximum output current (A) 100 300 500
Maximum output power (kW) 50 150 350
Typical charger cost ($) 8,500–11,000 17,000–22,000 45,000–67,000
Installation and grid connection cost ($) 5,000 11,000 >11,000
Electric Taxi Fleets 229

first DC charging protocols is the Charge-de-Move or CHAdeMO protocol, which

supports the Nissan Leaf EV. Currently, this protocol offers charging to EVs from
6 to 400 kW power with the help of CAN-based communication between the char-
ger and EV. The latest version of CHAdeMO protocol called ‘ChaoJi’ enables high-
power DC charging up to 900 kW. The maximum charging current possible with this
technology is increased to 600 A (CHAdeMO, 2020).
The most common DC charging protocol is perhaps the CCS protocol that origi-
nated in Europe. As the moniker denotes, the CCS connector is designed such that
DC charging and AC charging are possible with the same plug, minimizing the cost.
This protocol is used by popular EV manufacturers, including Volkswagen, General
Motors, BMW, Daimler, Ford, Morris Garages, and Hyundai, that support CCS.
Tesla also provides CCS charging ports for its EVs in Europe. The charging network
established under the Electrify America programme provides CCS charging at 350
kW (CCS, 2019).
Another charging is used in the proprietary Supercharger network operated by
Tesla. These Superchargers currently offer a maximum charging power of 120 kW
and are expected to reach power levels as high as 350 kW in the near future (Hall &
Lutsey, 2020; Hove & Sandalow, 2019).
China has established its own national DC charging standard protocol (GB/T)
mandatory for all EVs. The maximum charging power offered by GB/T is currently
237.5 kW (950 V, 250 A). India adopted the GB/T standard to develop the Bharat DC
charging protocol for its home-grown electric car models. The battery banks of these
cars operate at low voltages that cannot be provided by the established charging pro-
tocols (Hove & Sandalow, 2019; CHAdeMO, 2020; Das, Sasidharan & Ray, 2020).

9.4.3.3 Wireless Charging

Wireless charging makes EV charging easier without a physical wire connection
between the power supply and EV battery. Wireless transfer of power is possible by
induction (short-range WPT) or magnetic resonance (mid-range WPT). There are
three different modes of charging EVs wirelessly: stationary, quasi-dynamic, and
dynamic. Wireless charging when the vehicle is parked or idle is called stationary
charging. In the case of quasi-dynamic wireless charging, charging happens when
the vehicle is moving slowly or in stop-and-go mode. It is also possible to transfer
power dynamically when the vehicle is in motion. Among the three, dynamic charg-
ing is the most advanced and costliest option as it requires power transmitters embed-
ded in the road to charge batteries as the vehicles move over it (Ahmad, Alam &
Chabaan, 2017; Lukic & Pantic, 2013; Jang, Jeong & Lee, 2016).
To achieve wireless charging, power-transmitting and receiving equipment pieces
are needed, apart from the chargers. Typically, the stationary transmitting equipment
has an induction coil supplied from a high-frequency power supply, which gets mag-
netically coupled with the receiver. The EVs need to be equipped with receiver units
with induction coils, and power conditioning equipment. One of the major concerns
associated with wireless charging is the health impact from exposure to electromag-
netic fields. There are a few standards for wireless charging that sets limits for the
maximum power and distance at which energy can be transferred wirelessly (Qiu,
Ching, & Liu, 2014; Lu et al., 2015; Musavi, Edington & Eberle, 2012).
230 Electric Vehicle Integration in a Smart Microgrid Environment

The wireless charging standards that have been published or expected to be

released soon are SAE J2954, ISO 19363, IEC 61980, and GB/T 38775. The Standard
SAE J2954 establishes the specification of wireless charging between 3.7 and 11 kW
(WPT 1-3) for unidirectional power transfer (SAE, 2019). This standard is designed
for interoperability and supports both home and public charging installations. ISO
19363 standard specified the details of on-board vehicle equipment for unidirectional
power transfer via magnetic field for passenger cars and light-duty vehicles (ISO,
2020). IEC 61980 standard is for magnetic field wireless power transfer to electric
road vehicles when the EV is stationary. This standard is also applicable for wire-
less power transfer from on-site storage systems. China recently published a set of
standards for EV wireless charging, based on the magnetic resonance technology
(CEC, 2020). The standards for bidirectional power transfer or charging while EV is
in motion are under development.
Wireless charging technologies have lower efficiencies in comparison with con-
ductive charging technology. However, inductive charging is gaining popularity as a
convenient charging option for taxis as it requires a minimal action from the driver.
This technology is apt for opportunity charging at any predetermined location such
as taxi bays. Norway is planning to install such wireless charging for electric taxis
for the city of Oslo. WPT units of 75 kW power are expected to minimize waiting
time and charging time for electric taxis. Charging will start when the driver parks
the EV over the WPT units embedded in roads. Additionally, by eliminating cables
and connectors, wireless charging improves the safety of the charging processes
(Navigant, 2018; Statt, 2019; Reuters, 2019; Lukic & Pantic, 2013).

9.4.3.4 Battery Swapping

Battery swapping is often hyped as the future of EV charging, even though it is not
a popular technology in practice for charging EVs. Battery swapping is considered
as the best option to tackle the trilemma of long charging time, limited range, and
high battery replacement cost. In swapping, depleted vehicle batteries are swapped
with fully charged batteries in under 5 minutes. The battery swapping system has
two components: the battery charging system and the battery swapping mechanism.
A critical consideration in planning swapping operations is the battery inventory
requirements. The number of spare batteries that the station requires depends on
the demand rate for swapping at a station, which determines the power connection
needed at the facility. Battery swapping is grid-friendly, with the ease in control of
charging schedules and the ability to offer grid support. With respect to swapping
technology, battery technological advancements that lower production costs will only
have significant impacts if charging speed is fast enough (Rao et al., 2015; Sarker,
Pandžić, & Ortega-Vazquez, 2014; Mak, Rong & Shen, 2013).
Only selected car models from Renault, NIO, and Tesla are designed with swappa-
ble batteries. Battery swapping emerged as a solution in a project for converting ICE
vehicles to EVs. The retrofitted EV model Renault was used by Better Place in their
battery swapping project. The company built a network of swapping facilities, but
unfavourable economics for swapping operations led to project failure. The cost esti-
mate for each swapping facility was $2 million, and EV owners had to incur a cost
of $3,000 annually for subscribing to swapping services. A battery swapping-based
Electric Taxi Fleets 231

taxi fleet pilot project in Tokyo in April 2010 also closed down with the downfall of
the charging technology partner, Better Place. In the United States, Tesla tested bat-
tery swapping, but the solution was not commercially launched. Nevertheless, China
is a keen believer that swapping technology is ideal for electric taxis. It is reported
that around 200 battery swapping stations are operational to support electric taxis in
15 Chinese cities. Battery swapping is used for taxi fleet in the city of Hangzhou in
China (Feldman, 2017; Chafkin, 2014; Kang et al., 2015; Bloomberg, 2020; Hove &
Sandalow, 2019).
IEC 62840 and GB/T 29317 are the standards associated with battery swapping
technology. A battery swapping station can promote charging at the low energy
price periods of the day to in grid-to-battery mode (G2B), and enjoy the benefit
of price arbitrage during the high-energy price periods by discharging in battery-
to-grid mode (B2G). A battery swapping station is in effect an aggregator and
with enough capacity can participate in capacity and ancillary markets (Wang
et al., 2017).

9.4.4 chArging technology trends

There is hardly any consensus on the appropriate charging technology for fleet
charging. Apart from charging technologies referred to in the section above, there
are many upcoming options in the pipeline. Two of those options are selected and
explained in brief in this section considering their suitability for commercial fleet
applications.

9.4.4.1 Mobile EV Charging

Many charging service providers are exploring mobile charging services for park-
ing lots and roadside assistance. These chargers have batteries that can provide AC
or DC charging. These units may be beneficial for taxi fleet in ensuring access to
charging and help avoid stranding of vehicles. For large fleet, it can help manage
any spikes in demand. This technology can be a solution for the second life of
lithium-ion batteries. However, the cost of supply will be higher with this charging
technology as it entails additional losses in the storage. The standard governing
moveable charging equipment is SAE J3068 (Hove & Sandalow, 2019; Freewire,
2020; Lee & Clark, 2018).

9.4.4.2 Solar EV Charging

Sustainable electrification of transportation goes hand in hand with renewable gen-
eration. Globally, there has been a rapid progress in the growth of solar photovoltaic
(PV) power generation. With economies of scale, solar power generation has emerged
as one of the cheapest sources of electricity in many markets. Fleet operators have
benefited from coupling renewable energy to EV charging to reduce energy costs
(Arora & Raman, 2019). The latest trend in the fleet electrification is the deployment
of solar EV charging facilities, wherein power from the distributed solar plants is
directly fed to EVs or stored in batteries for later use. Such solar EV charging solu-
tions costing $60,000 are also being deployed in parking lots as part of the Electrify
America scheme (Forbes, 2020; GT, 2020).
232 Electric Vehicle Integration in a Smart Microgrid Environment

9.4.5 chArging stAtion sAFety

One of the major concerns associated with charging facilities is its safety. Most EV
markets have safety guidelines for EVSE installation and EV charging. In almost all of
the countries, Mode 1 charging is not permitted as it is considered as a safety hazard.
This is because there are no isolators or fault interrupters between the EV and EVSE
in Mode 1 charging. A few nations, including the New Zealand, do not permit Mode 2
charging for public charging facilities. In other regions, safety restrictions for Mode 2
charging are applicable as a maximum current limit. All the standardized EV connec-
tors have a proximity pilot wire, which ensures that the EV and EVSE are connected.
When the proximity is not detected, the charging stops automatically. One of the most
common safety precautions is with respect to protective earthing or an earth continuity
conductor during charging. One of the key safety recommendations while charging
is to maintain earthing. Most EVSEs and charging protocols have systems to monitor
health of earthing connection continuously and disconnect the charging in the event of
failure. Safety guidelines have provisions to prohibit the use of any connectors that is
not recommended by the vehicle or EVSE manufacturer, which is considered as a safety
violation. As expected, the safety norms for domestic charging include provision for a
dedicated circuit with residual current protection for the EVSE. These residual current
devices offer shock protection. For the public chargers, the safety norms are stricter,
and it includes mandatory provisions for earthing monitoring system, i.e., residual cur-
rent devices. Minimum ground clearance is specified for EVSEs that are charged under
Mode 3 or Mode 4. After installation, procedures are developed to do the testing and
inspection before the commissioning of the charging station. Periodic maintenance is
another mandatory provision found in most of the safety guidelines. For public charg-
ing of EVSE or its installation in a damp location, all equipment pieces should have an
adequate ingress protection. Any public chargers installed in locations such as park-
ing sites should also have protection from mechanical damage (NEC, 2017; Worksafe,
2019; CEA, 2013; Glowacki, 2016).

9.5 CATEGORIES OF COMMERCIAL FOUR-

WHEELER PASSENGER FLEET
Across the globe, four-wheeler passenger vehicles are employed for multiple private
and commercial purposes. The commercial four-wheeler passenger (electric taxi)
fleet can be categorized into two broad classes: ride-hailing fleet and corporate fleet.
The ride-hailing fleet are the taxi fleet operated by taxi aggregators or car rental
platforms for passenger transport from one location in an urban area to another.
Corporate fleets are ride-sharing taxis that caters to the mobility needs of employees
of an organization. The operations of these two types of taxi fleet differ based on the
purpose of use/ mobility, origin and destination points of the trips, trip attraction/
generation models, service/ business catchment area, etc.

9.5.1 ride-hAiling Fleet

Ride hailing is a service wherein commercial cars can be called upon for transport-
ing passengers from one place to another in a city, which is the predominant form
Electric Taxi Fleets 233

of intermediate between passenger and public transport in cities across the globe.
This is due to the route flexibility offered by service, in addition to door-to-door
mobility and other convenience factors. The service is normally backed by a mobile
application equipped with matchmaking algorithms based on passengers’ demand.
However, simple roadside hailing by passengers is also common in the urban context.
They operate in two ways primarily:

a. Single node operation. In this case, the starting point (node) for a taxi is fixed,
which is also the point to which it will return after making one or multiple
trips. The taxis may start their trip at the origin node or travel to a different
customer-requested starting point (Rodrigue, Comtois & Slack, 2013).
b. Double node-buffer zone operation. Here, a taxi’s movement generally cen-
tres around two nodes, i.e., the starting point (taxi’s initial location) and the
ultimate destination point (often, it is the resting place as preset by the taxi
driver). Matchmaking is done between the taxi and the rider(s) based on the
preference of the driver and the requirement of the riders. For instance, if
a taxi driver is travelling from point A to point B and a customer seeks a
ride from point C to point D on the mobile application, then a matchmak-
ing algorithm shall evaluate whether the travel route from C to D (along a
reasonable measure of road lengths) falls within a predefined buffer zone
of the route from A to B. In case the routes are in tandem with the require-
ments, the driver travels from point A to B, picks the passenger and goes
to C, drops the passenger, and continues to D (Riejos, 2019; Schiller &
Kenworthy, 2017).

9.5.2 corporAte Fleet

Such fleet ferries the employees of an organization between their workplaces (offices)
and residences or to and from other prefixed drop points. The trips are mostly pre-
planned and the drivers are generally aware of the travel demand that may arise dur-
ing a day. Such an operational model closely conforms to the single node operation
model of cab (taxi)-on-demand services, as office fleet operates out of a single node
and travels to other points on a road network based on passenger demand (Iles, 2005;
Yaghoubi, 2017).

9.6 PLAUSIBLE LOCATIONS FOR CHARGING ELECTRIC TAXI FLEET

Public charging infrastructure is already identified as the key enabler for promoting
the adoption of EVs. It has been estimated that there are almost 600,000 charging
points in the world at the end of 2018. Both EV uptake and public charging infra-
structure have experienced a growth rate of 60% in the 2013–2018 time period. It
is not surprising that most of the public charging infrastructure is spread across 25
cities in the major EV markets in Europe, the United States, and China. The typi-
cal locations for installation of charging infrastructure other than homes and work
places are public locations such as highway exits, fuelling stations, parking lots, and
curbside (Hall & Lutsey, 2020).
234 Electric Vehicle Integration in a Smart Microgrid Environment

9.6.1 chArging FAcilities For tAxi Fleet

Facilitating zero-emission taxi fleet is understood as critical for sustainable urban
transport. The recent trends in the charging infrastructure deployment showcase
dedicated facilities to cater to the demand of electric taxi fleet. Taxi fleet often makes
trips from certain popular hubs and requires charging facilities to reduce. It is envis-
aged that the identified commercial electric passenger fleet would depend on a host
of charging options. Some of these charging facilities would also cater to the oppor-
tunity for charging requirement of private fleet. The charging facilities primarily
differ on the locational aspect of the charging stations. Upon careful examination of
the different possibilities of charging, two types of public charging facilities along
with captive charging are important to support the mobility of commercial passenger
vehicles.

9.6.1.1 Public Charging Hubs for En route Charging

Such facilities may be called the electrical counterparts of modern-day fuel refill-
ing stations. As the name suggests, these stations are located adjacent to roads
and are intended to serve ‘en route’ charging demand; i.e., they cater to charging
demands of those EVs, which require charge mid-way in order to complete a trip
between two locations in a city. In addition, such stations may be deployed for
intercity travel. Erstwhile fuel refuelling stations equipped with chargers for EV
can also fall in this category. Many cities, including New York and Amsterdam, are
planning installation charging hubs for taxi cars. Stockholm is mapping priority
areas for public charging facilities to suit the needs of commercial and taxi fleet its
Charging Master Plan. In Vienna, the electricity distribution utility setup is done for
charging infrastructure for the electric taxi fleet (Hall, Cui & Lutsey, 2018; Hall &
Lutsey, 2020).

9.6.1.2 Charging Facility at Public Parking Spaces

Charging facilities are needed locations where vehicles are parked for an extended
period of time to support fleet electrification. Parking access is already an established
policy stance to promote EV adoption. It is logical to establish charging facilities at
public parking spaces; clubbing parking access can be clubbed with charging facili-
ties. In these locations as parking space is already available, space is only required
for the placement of charging infrastructure. At public places like shopping malls,
theatres, hospitals, parks, playgrounds, community centres, and transit nodes, charg-
ing facilities for EV charging can be set up. Charging piles can also be installed
on curb areas. London is planning an installation of charging facilities for taxis in
public hubs, commercial hubs, and semi-public depots. San Francisco and Oslo have
mandated the installation of chargers in parking spaces (Hall, Cui & Lutsey, 2018;
Hall & Lutsey, 2020).

9.6.1.3 Captive Charging Facilities

These are dedicated charging facilities that are set up for captive use by commercial-
purpose fleet. Many charging service providers like Innogy, New Motion, Fortum,
and ChargePoint have adopted business models supporting charging of fleet. For
Electric Taxi Fleets 235

corporate fleet, the captive facilities are often set up at workplaces as they are the
optimal locations for charging of fleet (Navigant, 2018).

9.6.2 criticAl FActors For siting chArging FAcilities

The physical siting fleet charging facilities may pose a major challenge to the charg-
ing service providers and fleet operators. Finding an ideal site to establish a charging
facility could be a challenging puzzle to solve. The effectiveness and the business
viability of running a charging facility depend on its utilization, which, in turn, is
contingent on the alignment of the preferences of fleet operators and charging service
providers in terms of site selection. These preferences are believed to be tied with a
set of factors or criteria for site selection for a public charging facility, and these are
expected to vary between different types of charging facilities. Just taking note of
these siting criteria may not help select a site for setting up a charging facility. One
should be conscious of the fact that seldom one can find a site that would fully sat-
isfy each criterion. Selection of a site based on the set criteria would involve certain
trade-offs. Some criteria may take precedence over others during decision-making.
In this regard, understanding the relative importance of the siting criteria would be
very useful. The possible siting criteria are categorized under physical planning and
grid connections, as listed in Table 9.6.

9.7 TECHNIQUES FOR LOCATING CHARGING FACILITIES

In order to maximize the return on investments in charging infrastructure, it is
essential to select optimal locations for the installation of chargers. Taxis and other
commercial fleet can provide initial high-usage and relatively predictable anchors to
ensure a viable business case. Proper siting of charging facilities can also maximize
the usage of chargers and minimize stress on the electricity grid simultaneously. The
common strategy adopted to identify locations by a decision-making exercise consid-
ers multiple factors, including availability of land, taxi driving and parking patterns,
and electricity grid capacity.
Multicriteria decision analysis, often called multicriteria decision-making
(MCDM), deals with the process of making decisions in the presence of multiple
objectives. MCDM is a versatile technique that is employed in EV charging problems
(Liu, et al., 2018). Selection of suitable locations for charging stations is viewed as a
MCDM problem. MCDM techniques, including analytical hierarchy process (AHP),
technique for order preference by similarity to ideal solution (TOPSIS), are used
for selecting locations for charging stations (Karaşan, Kaya, & Erdoğan, 2018; Wu,
et al., 2016; Stojčić, et al., 2019).

9.7.1 AnAlytic hierArchy process (Ahp)

AHP is a tool widely employed by the researchers in the MCDM problem. It helps quan-
tify the relative priorities of a given set of criteria on a ratio scale, based on the judge-
ment of decision-maker (Saaty, 1980, 2008). AHP relies on the judgement of experts
to derive the priority of criteria influencing the decision. AHP has been widely used to
236 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 9.6
Important Criteria for Selection of Sites for Setting up Different Types of
Charging Facilities (Sasidharan, Ray & Das, 2019)
Types of Charging Facility Criteria for Physical Planning Criteria for Grid Connection
Charging hub Road hierarchy 1 Proximity to distribution
Land availability and cost transformer, feeder or electric
Proximity to intersections or traffic substations5
signals Capacity of existing connection
Traffic volume on the road2 at the site
Charging facility at public Parking turnover ratio at the site Loading in the existing
parking space Hierarchy of space at the site3 distribution network 5)
Proximity to transit node4 Power quality at the site 5)
Captive charging facility for Land availability and cost
ride-hailing fleet Proximity to transit node
Captive charging facility at Number of employees availing the
office premise taxi services

1 Roads follow a hierarchy as they are composed of classes like arterial roads, subarterial roads, col-
lector roads, and local roads, wherein each class of roads has a separate function, design, and usage
rules. Road hierarchy determines the purpose, capacity, and functions of different types of roads
in a city’s/ town’s road network. Different types of roads could impact the charging demand. It is
understood that higher hierarchies of roads cater to higher volumes of traffic.
2 The volume of traffic that flows on a particular road could impact the charging demand and thereby
the usage of a charging facility. Traffic volumes on these roads should be compared for two sites.
This is because an axiomatic linkage exists between traffic volume and potential charging demand.
Post this comparison, it is suggested to gauge the proximity of the shortlisted sites to near-by traffic
intersections, and the visibility of the sites from the adjoining roads. Closeness of a site to a traffic
intersection and high visibility favour higher potential usage.
3 A city has spaces where people agglomerate. The size, type, purpose, density, length of stay,
etc. of such an agglomeration are governed by the hierarchy of space, where space could range
from being a commercial centre to an educational hotspot, from being a recreational facility to
a healthcare hub. It is common knowledge that higher the placement of space in that hierarchy,
higher the volume of citizen and traffic it invites, which makes it an important parameter to
consider from a business perspective when setting up a charging facility. For example, it would
be more lucrative to invest on a large-scale charging facility in the central business district of an
urban centre, instead of a convenience shopping centre which houses a milk booth, a stationery
shop, and a grocery shop.
4 Transit nodes are the locations in the transport network of a city where one mode of
transport meets another, such as railway stations, where rail-based transport meets road-based
transport, or metro stations, where rail-based transport meets the road- and walk-based trans-
port. Parking turnover ratio determines the capacity utilization of a parking space. High traffic
attraction of a parking area is desirable as it may potentially translate to high charging demand
at the site.
5 In case of electrical connection, three important aspects to examine would be the ease of getting a
connection, network capacity, and the reliability of electricity supply. In this regard, the approach
considers proximity of a site to a distribution equipment and the loading in the distribution network
as the relevant indicators. The closer the site is to a distribution equipment, the lower would be the
cost and time for obtaining a new electricity connection. The loading in the distribution network at
the site and power quality are indicators for available network capacity and supply reliability at the
site.
Electric Taxi Fleets 237

solve the real-world planning problems (Løken, 2007; Zyoud & Fuchs-Hanuch, 2017).
This tool helps to determine the optimal alternatives when applied to the problems in
planning. This is a convenient tool as it provides a methodology to calibrate the numeric
scale for intangibles. It also includes a consistency index testing mechanism, which
imposes discipline on the group decision-making (Sasidharan, Chandra & Das, 2019).
AHP helps to determine the important criteria by constructing a hierarchic structure
and conducting pairwise comparisons. A complex decision-making problem is broken
down into hierarchic levels descending from the overall goal, listing out the criteria
and subcriteria in the successive levels. At each level, a pairwise comparison of com-
ponents is to be made. An analytical process uses the results from pairwise comparison
and determines the priority of components at each level by assigning weights through
matrix operation. The individual scores from the experts would be aggregated and then
used to identify the criterion weights (Sasidharan, Chandra & Das, 2019).
The criterion weights are calculated using the following formula:
1
Wr =
k
{Wr1 + Wr2 +  + Wrk ,} (9.1)

where
wrk is the weight of rth criterion for k decision-makers (r = 1, 2,…,n; k = 1,2,…,k).
The classic AHP algorithm necessitates experts to make pairwise comparisons
and results in a complex questionnaire in the matrix format. Making pairwise com-
parison at times results in spurious scores, which could reduce the overall consis-
tency index of AHP (Sasidharan, Chandra & Das, 2019).

9.7.2 technique For order oF preFerence by similArity

to ideAl solution (topsis)

TOPSIS is an MCDM matrix with intuitive and clear logic that represents the ratio-
nale of human choice. It is a technique with the ability to measure the relative per-
formance for each alternative in a simple mathematical form. TOPSIS is a method
that accounts for both the best and worst alternatives in scalar values. The process
for the TOPSIS algorithm starts with the preparation of the decision matrix repre-
senting the relative importance of each criterion. Next, the matrix is normalized, and
the values are multiplied by the criteria weights derived from AHP. Subsequently,
the best (positive-ideal) and worst (negative-ideal) solutions are calculated. Later, the
distance of each alternative to these solutions is calculated with a distance measure.
Finally, the alternatives are ranked based on their relative closeness to the ideal solu-
tion (Lai, 1994; Guo & Zhao, 2015).
The TOPSIS technique is helpful for decision-makers to take informed deci-
sion based on analysis, comparisons, and ranking of the alternatives. The classi-
cal TOPSIS method solves problems only when there is no ambiguity in the data.
However, most real-world problems, including planning, are complex and subject to
face challenges in data availability. As a result, researchers have developed variants
of TOPSIS method, which can deal with the lack of information and uncertainty. If
the present study faces data limitations or uncertainty, a suitable variant of TOPSIS
method would be applied for ranking.
238 Electric Vehicle Integration in a Smart Microgrid Environment

9.7.3 rAtionAle For using Ahp And topsis

AHP and TOPSIS, two techniques proposed to be employed, are complementary
in nature. While TOPSIS is a purely numerical approach, AHP is a human-centric
assessment technique. AHP helps to overcome the limitations of TOPSIS with provi-
sions for weight elicitation, and consistency in checking of judgements. On the other
hand, TOPSIS alleviates the weaknesses of AHP such as the requirement of pairwise
comparisons and restrains imposed by the human capacity.

9.8 CONFIGURING A CHARGING FACILITY

FOR AN ELECTRIC TAXI FLEET
9.8.1 selection oF chArging technology
An important decision would be choosing the appropriate charging technology. The
effectiveness and feasibility of deployment and the use of a charging technology
hinge on a range of factors, both technical and economic. The selection of best-fit
charging technology can also be treated as a decision-making problem, and suitable
MCDM tools can be applied for optimizing diverse techno-economic objectives. A
decision matrix is a simple, but sturdy MCDM tool that can be used for the demon-
stration of the ranking of alternatives in charging technology selection problem. An
example of an application of the decision matrix to identify the ‘best-fit’ charging
technology for any public charging facility is illustrated below.
9.8.1.1 Identification of Charging Technologies for Evaluation
The first step in a decision-making problem identification of plausible charging tech-
nologies is suitable for installation in public charging. This decision could depend on
the maturity of charging technology and the availability of suitable EV models. As
an example, battery swapping and inductive charging are not mature technologies
for electric car charging in most EV markets. Level 1 charging is not applicable in
regions where 120-V AC distribution is not available. There are hardly any cars suit-
able for Level 6 charging power.

9.8.1.2 Selection of Parameters for Decision-Making

The decision matrix consists of a set of techno-economic parameters which receive a
rank and a weight. Some of the suggested decision-making parameters for the selec-
tion of charging technology are charging time, easiness of power connection, cost
of charger, cost of installation and grid connection, and cost of electricity. While
selecting the decision-making parameters, one should check the interdependencies
of contesting parameters.

9.8.1.3 Deciding Relative Weights of Parameters

Each parameter is assigned a weight based on the assessed degree of importance
using the interval scale in Figure 9.8. The relative weights of the parameters are a
subjective call of the decision-maker, and could vary from location to location. For
example, the importance of charging time will be different for charging at different
locations. This parameter could be most important for a charging station operating
Electric Taxi Fleets 239

FIGURE 9.8 Scale for assessing the importance of a parameter.

like a regular refuelling station, whereas it could be fairly important for a charging
facility located at a public parking space. It is possible to use other MCDM tech-
niques such as AHP and TOPSIS to develop the weights for the ranking.

9.8.1.4 Ranking of Parameters

Each charging technology is ranked against individual parameters whereby the
technology, which is closest to the ideal value for a parameter, is ranked highest.
The ideal values of the parameters could be the best in the industry given the con-
straints. It is also important to ensure while ranking that all of the parameters are
selected, such an ideal condition is represented either by maxima or by minima for
all the parameters. For example, low charging time, low cost of charger, connection,
and electricity cost are ideal conditions for charging technology selection. The ideal
value of charging time is the minimum charging time seen, i.e., half an hour (refer
to Figure 9.5). The relative ranks of charging times for the most common charging
technologies used in public charging facilities are presented in Table 9.7 for easy
reference for the decision-maker.
Similar rankings can be developed for other parameters for different charging
technologies. It is essential to remember that the technical and financial aspects
would vary in different geographical regions. Costs are dependent on other socio-
economic factors such as maturity of the EV market and policy support for charging
infrastructure. The easiness of grid connection is dependent on the design of the
distribution network and the loading patterns. Indicative rankings are developed for
the parameters using the details available in Tables 9.4 and 9.5.

9.8.1.5 Preparing Decision Matrix

Once the ranking is done and weights are decided, both are applied to a decision
matrix. Justification for the weights used in the example is presented in Table 9.8.
The individual weights are multiplied by the ranks for every parameter for all charg-
ing technologies. The total score of any ranking technology is obtained by normal-
izing and obtaining the weighted rank for all parameters. The charging technology

TABLE 9.7
Sample Ranks for Charging Time
Parameter Level 2 Level 3 Level 4 Level 5
Charging power considered (kW) 7.4 22 50 150
Charging time 1 2 3 4
240 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 9.8
Sample Weights Used in MCDM Example
S. No. Parameter Weight Justification of Weight
1. Charging time 10 This parameter is given the highest weight for fleet
operation as time taken for charging is loss of revenue
2. Ease of getting 8 This parameter is only secondary in importance to
connection charging time as fleet charging energy and power
demand for fleet charging are high
3. Cost of charger 5 This parameter is given the lowest weightage as cost of
charger is minimal in comparison with the cost of EV
4 Cost of installation and 6 Cost of installation is also given a lower weightage as
electricity connection the cost of installation is minimal in comparison with
the cost of EV

which notches up the highest normalized weighted rank would qualify as the most
preferred option. The least normalized weighted rank would determine the least pre-
ferred option. According to the weights and ranks used in the example presented in
Table 9.9, Level 3 charging technology is the most preferred option.

9.9 RECOMMENDATIONS FOR FLEET CHARGING

9.9.1 public chArgers Are required to support Fleet
One of the key recommendations from the examination of electric taxi fleet opera-
tions is that public charging facilities are essential for fleet charging. MCDM tools
can assist in the selection of suitable locations and best-fit charging technology
for fleets as described in the section above. As an example, for charging hubs,
the best possible chargers identified based on the selection criteria are Level 3
chargers. Once the suitable charging technology is identified, the next step is the
estimation of the number of chargers at each location. Typically, the conserva-
tive estimates are required for DC fast charging, as they are relatively expensive
than AC charging options. The decision-making factor is the number of vehicles
that can be supported by one EV charger, and this factor is inherently dependent
on the vehicle models. The number of vehicles that can be charged from Level 3
charger is around 10, 16, and 22 based on a utilization of 4, 6, and 8 hours in a day,
respectively (Slowik, Wappelhorst & Lutsey, 2019). The most common car-to-fast
charger ratio maintained by fleet operators in India is 5:1 (Das, Sasidharan &
Ray, 2020).

9.9.2 role oF Ac chArging For Fleet

When the City of Los Angeles initiated a ride-sharing programme in 2015 for 80
EVs, the supporting charging facilities were all AC Level 2 (Kettles, 2016). With the
 Electric Taxi Fleets

TABLE 9.9
Example of Decision-Making Matrix for Charging Technology Selection
Level 2 Level 3 Level 4 Level 5
Decision-Making
Criteria Weights Rank Rank × Weight Rank Rank × Weight Rank Rank × Weight Rank Rank × Weight
Charging time 10 1 10 2 20 3 30 4 40
Ease of getting 8 2 16 2 16 2 16 1 8
connection
Cost of charger 5 4 20 3 15 2 10 1 5
Cost of installation and 6 3 18 2 12 2 12 1 6
electricity connection
Sum 29 64 63 68 59
Normalized weighted ranks 2.20 2.17 2.34 2.03
241
242 Electric Vehicle Integration in a Smart Microgrid Environment

advancement in charging technology and improvement in battery chemistries, fast

charging infrastructure becomes more of a consideration when EV taxis is deployed.
Yet, Level 2 AC charging is a viable option for overnight charging of commercial
fleet. Research has also shown that charging electric ride-hailing vehicles at homes
can lower the total cost of operation by 25% compared to using public rapid DC
charging (Pavlenko, Slowik, & Lutsey, 2019).

9.10 GRID INTERACTION AND INTEGRATION OF FLEET

Fleet charging requires more energy and power in comparison with individual
vehicle charging. Most of the chargers are connected to medium- and low-voltage
distribution networks. These networks were not designed considering these kinds
of loads. The grid impact from fleets will be more because of two factors: the num-
ber of vehicles and preference for high-powered charging. Though higher-powered
charge points can reduce charge times, they cause a strain on electricity networks.
The impact will be compounded by an increase in the number of vehicles charging
at the same time and eventually lead to overloading and power quality issue parts of
the network (Hu et al., 2016; Wu et al., 2011).
On the other hand, fleet charging also presents more opportunities for manag-
ing the charging in comparison with that of individual cars. The uncertainties
in the charging behaviour is the biggest challenge to grid operation with EVs.
There is a possibility to reduce the uncertainty in the charging of fleet vehicles,
as it is easier to predict their charging needs based on operational patterns. Fleet
charging infrastructure will have predictable energy requirements, typically high
power, but at specific times of the day. Electric power utilities are critical partners
in the roll-out of charging infrastructure as they have the best understanding of
locations where there is available grid capacity for installing new stations and
where upgrades would be needed (Sundström & Binding, 2010; Khodayar, Wu &
Li, 2013).
The charging technology market has matured with controllable chargers and
even bidirectional chargers. Fleet electrification thus opens a new market segment
for grid services, which can help fleet operators and electricity distribution compa-
nies reduce their costs by managing energy in efficient ways. EVs have the capabil-
ity of acting as demand response resources for the grid using grid-to-vehicle (V1G)
functionality. EV fleets are potentially easier to bring in fleet charging manage-
ment strategies. Fleet charging facilities can be configured to refill EV batteries
with grid power when prices dip or when renewable energy generation is in excess.
Fleet operators can benefit from charging at low energy rates (Nyugen et al., 2014;
Heymann et al., 2017).
EVs, by virtue of being energy storage units, can potentially provide bidirectional
flow of power using vehicle-to-grid (V2G) functionality. It is expected that the aggre-
gation of EV fleets will help to promote V2G. Aggregated EV fleet can operate as
virtual power plants and provide real and reactive power support to the local grid.
Apart from that, by the versatile nature of being a distributed energy resource, they
can assist in voltage and frequency support (DeForest, MacDonald & Black, 2018;
Yi, Smart & Shirk, 2018).
Electric Taxi Fleets 243

NOMENCLATURE
AC:  Alternating current
AHP:  Analytical hierarchy process
B2G:  Battery-to-grid
CCS:  Combined charging system
CHAdeMO: 
Charge de Move
DC:  Direct current
DoD:  Depth of discharge
EV:  Electric vehicle
EVSE:  Electric vehicle supply equipment
G2B:  Grid-to-battery
ICE:  Internal combustion engine
IEC:  International Electrotechnical Commission
LFP:  Lithium iron phosphate
LMO:  Lithium manganese oxide
MCDM:  Multicriteria decision-making
NCA:  Lithium nickel cobalt aluminium oxide
NMC:  Lithium nickel manganese cobalt oxide
PV:  Photovoltaic
SAE:  Society of Automotive Engineers
SoC:  State of charge
TCO:  Total cost of ownership
TOPSIS: Technique for order preference by similarity to ideal solution
V1G:  Grid-to-vehicle
V2G:  Vehicle-to-grid
WPT:  Wireless power transfer

WORKS CITED
Ahmad, A., Alam, M. S., & Chabaan, R. (2017). A comprehensive review of wireless charging
technologies for electric vehicles. IEEE Transactions on Transportation Electrification,
4(1), pp. 38–63.
Arora, N., & Raman, A. (2019). Beyond Nagpur: The Promise of Electric Mobility. New
Delhi: Ola Mobility Institute.
Bloomberg. (2020). China embraces battery-swapping system for electric vehicles. Bloomberg
News.
CCS. (2019). Combined Charging System 1.0 Specification. CHARIN. https://www.charinev.
org/ccs-at-a-glance/ccs-specification/
CEA. (2013). Draft Safety Provisions for Electric Vehicle (EVs) Charging Stations. Central
Electricity Authority. https://cea.nic.in/old/reports/others/enc/legal/cea_safety_regula-
tions.pdf.
CEC. (2020). GB / T 38775, Wireless Charging System for Electric Vehicles. Chinese
Electricity Council. https://www.chinesestandard.net/PDF.aspx/GBT38775.1-2020.
CHAdeMO. (2015). CHAdeMO is published as IEEE standard. Retrieved from: https://www.
chademo.com/chademo-is-published-as-ieee-standard/.
CHAdeMO. (2020). CHAdeMO 3.0 released: The first publication of ChaoJi, the new
plug harmonised with China’s GB/T. Retrieved from: https://www.chademo.com/
chademo-3-0-released/.
244 Electric Vehicle Integration in a Smart Microgrid Environment

Chafkin, M. (2014). A broken place: The spectacular failure of the startup that was going
to change the world. Retrieved from Fast Company: https://www.fastcompany.
com/3028159/a-broken-place-better-place.
Clairand, J. M., Guerra-Terán, P., Serrano-Guerrero, X., González-Rodríguez, M., & Escrivá-
Escrivá, G. (2019). Electric vehicles for public transportation in power systems: A
review of methodologies. Energies, 12(16), p. 3114.
Coffin, D., & Horowitz, a. J. (2018, December). The supply chain for electric vehicle batteries.
The Journal of International Commerce and Economics. Retrieved from https://www.
usitc.gov/publications/332/journals/the_supply_chain_for_electric_vehicle_batteries.pdf.
Das, S., Sasidharan, C., & Ray, A. (2020). Charging India’s Four-Wheeler Transport. New
Delhi: Alliance for an Energy Efficient Economy.
DeForest, N., MacDonald, J. S., & Black, D. R. (2018). Day ahead optimization of an electric
vehicle fleet providing ancillary services in the Los Angeles air force base vehicle-to-
grid demonstration. Applied Energy, 210, pp. 987–1001.
Feldman, Y. (2017, August 31). What Happened to Better Place’s Electric Dreams? Retrieved from
The Jerusalem Post: https://www.jpost.com/Metro/A-story-that-should-be-told-503905.
Freewire. (2020). Mobi EV charger. Retrieved from: https://freewiretech.com/products/
mobi-ev/.
Forbes. (2020). Solar-powered electric vehicle charging stations are just around the corner.
Retrieved from: https://www.forbes.com/sites/kensilverstein/2020/02/10/solar-pow-
ered-electric-vehicle-charging-stations-are-just-around-the-corner/#5ff42f30320f.
GT. (2020). Solar EV chargers simplify installation and adoption. Retrieved from https://
www.govtech.com/fs/infrastructure/Solar-EV-Chargers-Simplify-Installation-and-
Adoption-.html.
Glowacki, P. (2016). Developing Electrical Safety Standards to Introduce Electric Vehicles
into Canada. CSA Group, Canada.
Guo, S., & Zhao, H. (2015). Optimal site selection of electric vehicle charging station by using
fuzzy TOPSIS based on sustainability perspective. Applied Energy, 158, pp. 390–402.
Hagman, J., & Langbroek, J. H. (2019). Conditions for electric vehicle taxi: A case study in
the Greater Stockholm region. International Journal of Sustainable Transportation,
13(6), pp. 450–459.
Hall, D., & Lutsey, a. N. (2020). Electric Vehicle Charging Guide for Cities. Washington DC:
International Council for Clean Transport.
Hall, D., Cui, H., & Lutsey, N. (2018). Electric Vehicle Capitals: Accelerating the Global
Transition to Electric Drive. Washington DC: International Council for Clean Transport.
Heymann, F., Miranda, V., Neyestani, N., & Soares, F. J. (2017). December. Mapping the
impact of daytime and overnight electric vehicle charging on distribution grids. In 2017
IEEE Vehicle Power and Propulsion Conference (VPPC) (pp. 1–6). IEEE.
Hirschfeld, A. (2019) Taxis go electric: Tesla joins cab fleet in New York city. Retrieved from
Observer https://observer.com/2019/11/tesla-taxi-electric-vehicles-cab-fleet/.
Hove, A., & Sandalow, D. (2019). Electric Vehicle Charging in China and the United States.
Columbia, School of International and Public Affairs, Center on Global Energy Policy.
Available online: https://energypolicy.columbia.edu/sites/default/files/file-uploads/EV_
ChargingChina-CGEP_Report_Final.pdf.
Hu, J., Morais, H., Sousa, T., & Lind, M. (2016). Electric vehicle fleet management in
smart grids: A review of services, optimization and control aspects. Renewable and
Sustainable Energy Reviews, 56, pp. 1207–1226.
IEA. (2018). Global EV Outlook 2018. International Energy Agency, France.
Iles, R. (2005). Regulation of public transport services. Public Transport in Developing
Countries, pp. 403–443. doi: 10.1108/9780080456812-019.
Electric Taxi Fleets 245

ISO. (2020). ISO 19363:2020 Electrically Propelled Road Vehicles- Magnetic Field Wireless
Power Transfer — Safety and Interoperability Requirement. International Standards
Organisation.
Jäger, B., Wittmann, M., & Lienkamp, M. (2017). Agent-based modeling and simulation of
electric taxi fleets. In 6. Conference on Future Automotive Technology.
Jang, Y. J., Jeong, S., & Lee, M. S. (2016). Initial energy logistics cost analysis for station-
ary, quasi-dynamic, and dynamic wireless charging public transportation systems.
Energies, 9(7), p. 483.
Jenn, A. (2019). Electrifying ride-sharing: Transitioning to a cleaner future. Retrieved from
the University of California, Davis, https://escholarship.org/uc/item/12s554kd.
Jia, Y., Zhao, Y., Guo, Z., Xin, Y., & Chen, H. (2017). October. Optimizing electric taxi charg-
ing system: A data-driven approach from transport energy supply chain perspective. In
2017 IEEE Electrical Power and Energy Conference (EPEC) (pp. 1–6). IEEE.
Kang, Q., Wang, J., Zhou, M., & Ammari, A. C. (2015). Centralized charging strategy and
scheduling algorithm for electric vehicles under a battery swapping scenario. IEEE
Transactions on Intelligent Transportation Systems, 17(3), pp. 659–669.
Karaşan, A., Kaya, İ., & Erdoğan, M. (2018). Location selection of electric vehicles charging
stations by using a fuzzy MCDM method: A case study in Turkey. Neural Computing
and Applications, 32(9), pp. 1–22.
Kettles, D. (2016). Electric vehicle fleet implications and analysis (No. FSEC-CR-2031-16).
Khodayar, M. E., Wu, L., & Li, Z. (2013). Electric vehicle mobility in transmission-con-
strained hourly power generation scheduling. IEEE Transactions on Smart Grid, 4(2),
pp. 779–788.
Lai, Y. J. (1994). Topsis for MODM. European Journal of Operational Research, 76, pp.
486–500.
Lee, H., & Clark, A. (2018). Charging the Future: Challenges and Opportunities for Electric
Vehicle Adoption, Harvard Kennedy School, Cambridge.
Lewis-Jones, A., & Roberts, J. (2017). Electric private hire vehicles in London on the road,
here and now, energy savings trust. Retrieved from: https://energysavingtrust.org.
uk/sites/default/files/reports/Uber%20EV%20Trial%20-%20Electric%20Private%20
Hire%20Vehicles%20in%20London_1.pdf.
Liu, H.-C., Yang, M., Zhou, M., & Tian, G. (2018). An integrated multi-criteria decision
making approach to location planning of electric vehicle charging stations. IEEE
Transactions on Intelligent Transportation Systems, 20(1), pp. 362–373.
Løken, E. (2007). Use of multicriteria decision analysis methods for energy planning prob-
lems. Renewable and Sustainable Energy Reviews, 11(7), pp. 1584–1595.
Lu, X., Wang, P., Niyato, D., Kim, D. I., & Han, Z. (2015). Wireless charging technologies:
Fundamentals, standards, and network applications. IEEE Communications Surveys &
Tutorials, 18(2), pp. 1413-1452.
Lukic, S., & Pantic, Z. (2013). Cutting the cord: Static and dynamic inductive wireless charg-
ing of electric vehicles. IEEE Electrification Magazine, 1(1), pp. 57-64.
Mak, H. Y., Rong, Y., & Shen, Z. J. M. (2013). Infrastructure planning for electric vehicles
with battery swapping. Management Science, 59(7), pp. 1557–1575.
Merkisz-Guranowska, A. and Maciejewski, M. (2015). The Implementation of the Electric
Taxi Fleet in the City of Poznan, Poland. València: WIT Transactions on the Built
Environment, pp. 243–254.
MIT, 2008, A Guide to Understanding Battery Specifications, MIT electric vehicle team
http://web.mit.edu/evt/summary_battery_specifications.pdf.
Morrow, K., Karner, D., & Frankfort, J. (2008). Plug-in hybrid electric vehicle charging
infrastructure review. US Department of Energy-Vehicle Technologies Program.
246 Electric Vehicle Integration in a Smart Microgrid Environment

Musavi, F., Edington, M. and Eberle, W., (2012). Wireless power transfer: A survey of EV bat-
tery charging technologies. In 2012 IEEE Energy Conversion Congress and Exposition
(ECCE) (pp. 1804–1810). IEEE.
Navigant. (2018). EV Charging Equipment Market Overview. Boulder: Navigant Consulting, Inc.
NEC. (2017). Article 625 Electric Vehicle Charging and Supply Equipment Systems. National
Electricity Code.
Nicholas, M., Slowik, P., & Lutsey, N. (2020). Charging infrastructure requirements to sup-
port electric ride-hailing in US cities, Washington DC: Publisher International Council
for Clean Transport.
Pavlenko, N., Slowik, P., & Lutsey, N. (2019). When Does Electrifying Shared Mobility Make
Economic Sense? Washington DC: The International Council on Clean Transportation.
Qiu, C., KT, C., Ching, T. W. & Liu, C. (2014). Overview of wireless charging technologies for
electric vehicles. Journal of Asian Electric Vehicles, 12(1), pp. 1679–1685.
Rao, R., Zhang, X., Xie, J., & Ju, L. (2015). Optimizing electric vehicle users’ charging behav-
ior in battery swapping mode. Applied Energy, 155, pp. 547–559.
Reuters. (2019). Oslo becomes the first city in the world to get wireless charging systems for
EVs. Retrieved from Auto NDTV: https://auto.ndtv.com/news/oslo-becomes-the-first-
city-in-the-world-to-get-wireless-charging-systems-for-evs-2012267.
Riejos, F. A. (2019). Urban Transport XXIV. WIT Press, Spain.
Rodrigue, J.-P., Comtois, C., & Slack, B. (2013). The Geography of Transport Systems.
Routledge, New York.
Rokhadiya, S., Wappelhorst, S., Pavlenko, N., & Bandivadekar, a. A. (2019). Near-Term
Incentives for Electrifying Electrifying Ride Sharing Fleets in India. International
Council of Clean Transportation, India.
Sasidharan, C., Chandra, D., & Das, S. (2019). Application of analytic hierarchy process for the
siting of electric vehicle charging infrastructure in India. In 2019 IEEE Transportation
Electrification Conference (ITEC-India) (pp. 1–6). IEEE.
Sasidharan, C., Ray, A., & Das, S. (2019, October). Selection of charging technology for elec-
tric bus fleets in intra-city public transport in India. In 2019 Global Conference for
Advancement in Technology (GCAT) (pp. 1–8). IEEE.
Saaty, T. L. (1980). The Analytic Hierarchy Process. New York: McGraw-Hill.
Saaty, T. L. (2008). ` with the analytic hierarchy process. International Journal of Services
Sciences, 1(1), 83–98.
SAE. (2019). Wireless Power Transfer for Light-Duty Plug-in/Electric Vehicles and Alignment
Methodology J2954, Society for Automotive Engineers.
Sarker, M.R., Pandžić, H., & Ortega-Vazquez, M.A., (2014). Optimal operation and services
scheduling for an electric vehicle battery swapping station. IEEE Transactions on
Power Systems, 30(2), pp. 901–910.
Schiller, P. L., & Kenworthy, J. (2017). An Introduction to Sustainable Transportation: Policy,
Planning and Implementation. Routledge, UK.
Slowik, P., Wappelhorst, S., & Lutsey, N. (2019). How Can Taxes and Fees on Ride Hailing
Fleets Steer them to Electrify?, Washington DC.
Spöttle, M., Jörling, K., Schimmel, M., Staats, M., Grizzel, L., Jerram, L., & Gartner, J.
(2018). Research for TRAN Committee – Charging infrastructure for electric road
vehicles. European Parliament, Policy Department for Structural and Cohesion Policies.
Brussels: European Parliament. Retrieved from http://bit.ly/2JBVvHq.
SGCC. (2013). EV Infrastructure and Standardization in China, State Grid Corporation of
China, Washington DC.
Statt, N. (2019). Norway will install the world’s first wireless electric car charging stations for
Oslo taxis. Retrieved from The Verge: https://www.theverge.com/2019/3/21/18276541/
norway-oslo-wireless-charging-electric-taxis-car-zero-emissions-induction.
Electric Taxi Fleets 247

Stojčić, M., Zavadskas, E. K., Pamučar, D., Stević, Ž., & Mardani, A. (2019). Application
of MCDM methods in sustainability engineering: A literature review 2008–2018.
Symmetry, 11(3), p. 350.
Sundström, O., & Binding, C. (2010) October. Planning electric-drive vehicle charging
under constrained grid conditions. In 2010 International Conference on Power System
Technology (pp. 1–6). IEEE.
Teixeira, A. C. R, & Sodré, J. R. (2017). A discussion on the application of electric vehicles to
taxi cab fleets. Retrieved from The European Energy Centre: https://www.euenergycen-
tre.org/news/a-discussion-on-the-application-of-electric-vehicles-to-taxi-cab-fleets/.
Vesa, J. (2019). SESKO. Retrieved from https://www.sesko.fi/files/671/EV-charging_stan-
dards_may2016_Compatibility_Mode_.pdf.
Wang, H., & Cheu, R. L. (2013). Operations of a taxi fleet for advance reservations using elec-
tric vehicles and charging stations. Transportation Research Record, 2352(1), pp. 1–10.
Wang, H., Cheu, R. L., & Balal, E. (2016). Operations of electric taxis to serve advance
reservations by trip chaining: Sensitivity analysis on network size, customer demand
and number of charging stations. International Journal of Transportation Science and
Technology, 5(2), pp. 47–59.
Wang, Y., Ding, W., Huang, L., Wei, Z., Liu, H., & Stankovic, J. A. (2017). Toward urban elec-
tric taxi systems in smart cities: The battery swapping challenge. IEEE Transactions on
Vehicular Technology, 67(3), pp. 1946–1960.
WBCSD. (2019). India Business Guide to EV Adoption. World Business Council for
Sustainable Development, India.
Worksafe. (2019). Guidelines for Safe Electric Vehicle Charging. New Zealand: Government
of New Zealand.
Wu, Y., Yang, M., Zhang, H., Chen, K., & Wang, Y. (2016). Optimal site selection of elec-
tric vehicle charging stations based on a cloud model and the PROMETHEE method.
Energies, 9(3), p. 157.
Wu, Q., Nielsen, A. H., Østergaard, J., Cha, S. T., & Ding, Y. (2011) December. Impact study
of electric vehicle (EV) integration on medium voltage (MV) grids. In 2011 2nd IEEE
PES International Conference and Exhibition on Innovative Smart Grid Technologies
(pp. 1–7). IEEE.
Yaghoubi, H. (2017). Urban Transport Systems. Croatia: BoD - Books on Demand.
Yi, Z., Smart, J., & Shirk, M. (2018). Energy impact evaluation for eco-routing and charging of
autonomous electric vehicle fleet: Ambient temperature consideration. Transportation
Research Part C: Emerging Technologies, 89, pp. 344–363.
Zyoud, S. H., & Fuchs-Hanuch, D. (2017). A bibliometric-based survey on AHP and TOPSIS
techniques. Expert Systems with Applications, 78, pp. 158–181.
 10 Machine Learning-Based
Day-Ahead Market
Energy Usage Bidding
for Smart Microgrids
Mohd Saqib
Indian Institute of Technology (Indian School of Mines)

Sanjeev Anand Sahu

Indian Institute of Technology (Indian School of Mines)

Mohd Sakib
Aligarh Muslim University

Esaam A. Al-Ammar
King Saud University

CONTENTS
10.1 Introduction................................................................................................... 250
10.2 Different Aspects of EVs............................................................................... 251
10.3 Description of Power Market Stakeholder Interaction Model....................... 254
10.3.1 Activity Diagram............................................................................... 255
10.3.2 Data Flow Diagram........................................................................... 255
10.3.3 ER Diagram....................................................................................... 256
10.4 AI Strategies.................................................................................................. 258
10.4.1 Artificial Neural Networks................................................................ 258
10.4.2 Autoregressive Moving Average........................................................ 258
10.4.3 Support Vector Machine ................................................................... 259
10.5 Overall Demonstration..................................................................................260
10.6 Case Studies................................................................................................... 262
10.6.1 Forecasting of Energy Price.............................................................. 262
10.6.2 Aggregate Demand-Supply System................................................... 262
10.6.3 xEV Market Analysis and Forecast................................................... 263
10.7 Result............................................................................................................. 263
10.8 Conclusion..................................................................................................... 263
Nomenclature..........................................................................................................264
References...............................................................................................................264
249
250 Electric Vehicle Integration in a Smart Microgrid Environment

10.1 INTRODUCTION
Increment in the pollution has bound the government to take some strong decisions to
implement zero-emission due to which many more xEVs come on the road. The high
energy demand will be required due to the rapid growth of xEV and maybe become
the reason of system failures or give rise to mismanagement. So it is the reason, in
this era, we need an automated charging infrastructure that can operate by various
entities related to the power grid. Furthermore, we know that the most paranoid part
of the smart grid is the bidding process of the charging/electricity because like other
resources of the energy (e.g., coal, petroleum, etc.), it cannot be stored. That is why
we need to automate the bidding process for the system. As we are living in an era of
artificial intelligence (AI), it can bring revolutionary changes in the energy market.
To achieve these benefits, proposed work came into existence. In the proposed sys-
tem, we introduce an AI-based cyberinfrastructure that automates and optimizes the
bidding process in the day-ahead market.
The industries are now ready to establish a new system and move on from our
antiquated, human-oriented, traditional gasoline engine vehicles because of lots of
problems created by such vehicles [1]. To reduce emission, many gasoline engines
are transferring to electric vehicles [2]. This [3] study shows different government
programs to implement zero-emission and needed to achieve CO2 targets soon. xEVs
can be performed as a dependable, sustainable, and finest powertrain alternative for
transportation soon, but its adoption in the present infrastructure can pose difficulty
for management unless it is smartly integrated. The increasing number of electric
vehicles caused new challenges like tracking the nearest charging station, bidding
price, queuing delay, etc. The massive demand for electric vehicles in the present trans-
port sector requires a conventional system and coordination mechanisms to operate.
As we are aware, BigData is time-demanding and complex to handle, which needs
a powerful system to maintain good accuracy. The future demand for EV is rapidly
growing, but it is hard to estimate the upcoming load, so it is impractical to develop
an optimal scheduling scheme for EVs [4]. Also in Ref. [5], the authors explain dif-
ficulties with the new computer-oriented power plants and discussed proper solutions
for it as well. However, in Ref. [6], researchers discuss to formulate the real-time
charging and pricing system of xEVs based on the Internet of Things (IoT) and also
conclude with a global optimal scheduling method. A synchronized approximate
dynamic programming model for coordinated charging control is introduced in Ref.
[7], and similarly, a scheduling paradigm has been represented in Ref. [8] to mini-
mize electricity depletion. Also, Information and Communication Technology (ICT)
opens new doors for improvement in the scenario of power grid management. In
Refs [7] and [9], the authors have presented a smart infrastructure model to handle
xEV market in India using ICT. Refs [7] and [10] present the need from xEV users
to xEV charging control mechanism based on hierarchical aggregators, which can
provide coordination between demand and distribution of power through real-time
charging infrastructure. In Ref. [8], a genetic algorithm-based charging controller
has been demonstrated to maintain the system threshold voltage within permissible
limits, and to perform this tedious job, various inputs like SOC, minimum voltage,
and price of energy are loaded. A mathematical model has been employed by the
Day-Ahead Market Energy Usage Bidding 251

researcher in this Ref. [11] study to manage energy demand near a highway entry-exit
port. This study also overcomes the further issues of peak energy demand for xEVs
on the highway. A sliding window optimization algorithm has been presented in this
Ref. [12] to achieve low-cost charging in an automated grid. In Ref. [13], the dynamic
programming paradigm is adopted to manage multidimensional energy flow.
Although ICT brings revolutionary changes in the energy market and tackled many
engineering problems, we are living in the era of AI, machine learning (ML), and deep
learning (DL), which may take the energy market to the next level. This [14] article
explains all the existing problems with ICT-based systems, for example, the complexity
of solutions, demand of quality programming and computer experts, threats of cyber
tools, etc. and also speculates on the regulation of various xEV-related forecasting.
And this [15] research presented how AI can transform the energy market and also
introduce the platform architectural logic that circumscribes technical and financial
architecture of AI-based smart grid platforms for the upcoming energy market. Same
as in Ref. [16], a framework to forecast electricity prices is proposed using DL. Also,
these [17–20] studies proposed an artificial neural network (ANN)-based algorithm
to forecast energy price and load as well.
The manuscript is organized as follows. First of all, the manuscript starts with
a detailed section of “Different aspects of EVs” followed by the “Description of
Power Market Stakeholder Interaction Model” which explains the software-work-
ing model for the stakeholder entities. The section “AI Strategies” describes the AI
techniques that can be formed by automated bidding system followed by “Overall
Demonstration”, which explains the overall system and different connectivity. In
the next section, three case studies have been given to justify results namely “Case
Studies”. And results are described in the last section “Result” followed by the con-
clusion in “Conclusion”.

10.2 DIFFERENT ASPECTS OF EVs

As we have understood from the literature survey, vehicle to grid (V2G) and grid to
grid (G2G) infrastructure has been executed and analyzed in an inadequate number
of pilots, and various problems need to be solved before wide adoption. While newly
developed charging infrastructures also have some issues, solutions are under pro-
cess. In Feb 2017, around 580 thousand were sold in the U.S.A. due to a lack of proper
regulation [20]. To overcome all these issues, various other technologies need to be
discussed apart from AI and ML as follows [21]:

• Faster charging stations. Slow charging stations are the major inconve-
nience, and it is one of the reasons most users do not feel free to take out
EVs for long-distance roots. Various industries have already head-started on
deploying global supercharging stations, for example, Tesla.
• Battery quality improvement. Using lithium is not a better way to make
batteries because it is hard to find. As in this [22], the authors discussed a
system to enhance battery life as well as battery performance.
• Autonomous driving. Autonomous driving is the need of today to make
roads safer. Tesla and many others are leading in this technology.
252 Electric Vehicle Integration in a Smart Microgrid Environment

• Software and mobile apps. There are so many apps developed to manage
and monitor real-time data related to charging stations and EVs. For exam-
ple, in Ref. [23], the author discussed cloud-based monitoring software to
access real-time information.

EVs have several environmental benefits compared to gasoline engine cars. Carbon
dioxide (CO2) is produced by burning gasoline, which causes greenhouse gas and
leads to climate change issues [3]. Using EVs instead of gasoline vehicles reduces
the amount of CO2 about 4,096 pounds per year [24]. It is also estimated by Stanford
that the amounts of carbon emissions will reduce to roughly $4,506 over 10 years.
Furthermore, EVs also show significant benefits from the methods used for elec-
tricity generation. The problem to obtain gasoline is carbon-intensive production in
pumping, mining, and refining tasks. So the conclusion is EVs also will be beneficial
even for those countries that produce electricity by burning fossil fuels.
So many studies represented the economic benefits of EVs [25]. Surely, the
arrival of EV infrastructure will result in job losses in the conventional gasoline and
mechanical industries, but many other new doors will open for research and develop-
ment fields, automation industries, AI companies, and battery manufacturers. Also,
EVs cost less to maintenance and operational tasks as summarized in Table 10.1 [25].
Apart from the economic and environmental aspects of EVs, they also have vari-
ous safety aspects as discussed in Ref. [26]. The first thing that should be kept in
mind is the safety of the electrical system. Different ranges of voltage used in EVs
(small vans vary from 48 to 120 V, large vans from 90 to 250 V, buses from 250
to 500 V, and AC vehicles more than 250 V) are considered dangerous for human
safety. Therefore, there is a need to prevent electric shocks in direct and indirect
touches. Second, the functioning of EV systems is also very sensitive, e.g., range
anxiety, real-time information regarding CS locations, nearby areas, queuing delay,
etc. Apart from all these regulations and monitoring safety, cyber-security is also
an obstacle. Automated EVs need special cyber securities, otherwise it may cause
very serious accidents on the roads. Third, battery safety is the most critical part of
EVs. It may cause several electrical, chemical, and mechanical complications and an
explosion [26]. There is a need for strong locking connections when multiple batter-
ies are implemented to prevent short circuits. There should be surveillance measures

TABLE 10.1
Comparisons of Traditional Vehicle and EV
Factors Gasoline Vehicle (Price in $) EV (Price in $)
Brakes 400.00 200.00
Oil 600.00 0
Spark plugs and wires 200.00 0
Transmission fluid 60.00 0
Tires 700.00 700.00
Muffler 180.00 0
Total 2,140.00 900.00
Day-Ahead Market Energy Usage Bidding 253

to avoid explosion, which especially occurs at the end of charging. At the time of
charging, EVs are connected with the main distributed network, so there should be
precautions to avoid electric shock.
After all these discussions, let’s come on major government programs empow-
ering EV infrastructures in various markets and highlight some emerging best
practices that increase the stock of EVs, subsidies, public-private partnerships,
etc. as discussed in this [27] detailed report of the International Council on Clean
Transportation, U.S.A.

China. Energy stockholders in China, including government sectors, have been

fast-moving toward developing EV infrastructures in the country. In the
upcoming years, it will enhance its market position. In the time frame of
2020, China wants to set new goals to sell EVs that are in the embryonic
stage to achieve the target of 300,000 EVs [27]. Also, it has announced EVs
having 20% increment in production by 2025. Furthermore, the numbers of
charging stations are also rapidly growing in China especially in the 88 des-
ignated pilot cities led by Shanghai, Beijing, and Shenzhen. There is a plan
to build around 100,000 smart charging stations and more than 400,000
total public stations by 2020. The goal of the program is to provide 8 EVs
per charging stations and farther than 1 km any point within the center area
of the city.
Japan. Since the introduction of modern EVs in Japan in 2011, at a very early
stage, energy stockholders and government sensed that the key requirement
for EV sales is charging infrastructure, so they supported [27]. In 2013,
the government sector worked on advance charging architecture projects
to developed CSs around the cities, and also the Bank of Japan cooperated
with Nissan, Honda, and Toyota to develop Nippon Charge Service (NCS).
Now, 7,500 CSs are connected to this project with regular funding.
United States. The U.S. EV sales are rapidly growing due to government sup-
port, zero-emission vehicle regulatory policy, public-private relationship, etc.
In California, a zero-emission policy is expected to increase electric vehicles
in the market from more than sixty thousand to several million in decay.
To complete the requirement of guiding the charging deployment process,
California developed a model that projects needs of CSs by 2025 [28].
France. In France, the government has initiated a plan to move all vehicle sales
to electric by 2040. Also, the target is to operate 12,000 charging stations in
the city. To achieve this goal, many funds are releasing, and the recipients
must ensure to develop at least 20 char points and also make slots for free
parking. The federal government’s strong role is evident in empowering the
charging infrastructure in France.
Germany. Germany has a goal to reach 1 million by 2020 and 5 million by
2030. At the start of 2009, In Hamburg and Saxony, the country started to
support more than 200 projects in eight “model regions” with €130 mil-
lion, boosting charging infrastructure. Furthermore, the country planned to
promote EVs, including €300 million earmarked for public charging infra-
structure through 2020.
254 Electric Vehicle Integration in a Smart Microgrid Environment

Europe. xEV charging infrastructure has been regularized by private as well

as government bodies. To empower the sales of EVs and charging infra-
structure, many countries within the European Union have created funding
schemes. Some of them are using charging infrastructure for several years
like Norway and Netherland and some are recently launched like Germany.
Under the initiative for increasing the number of CSs in the city, a program,
TEN-T/CEF-T, was launched [27]. And under this program, the European
Commission is ensuring that charging stations are in public reach without
any suffering of range anxiety.

So, not only in China, Japan, and the U.S.A that the EVs became the major trans-
port, but also in Europe, more than 70% of transport responsibility depends on them
[29,30]. Besides, it has the following issues:

• Range anxiety. Although technically it sounds good, but still users of

EVs have the problem of range anxiety due to technical limitations of the
battery.
• Lack of understanding. People still have less awareness and understanding
of e-mobility.
• Charging system. Although government is working so far to build a reach-
able charging infrastructure, there is still a lack of reachability, and users
are facing monitoring problems.

The government is still struggling to build up sufficient infrastructure and implement

policies favoring e-mobility systems in urban areas [30]. In 2018, a meeting was
conducted between energy experts in Stockholm to discuss on different initiatives
and solution to operating and adoption of e-mobility, and they have concluded the
following seven points [30]:

• Increment in charging points/stations

• User-friendly systems
• Smart solutions based on ICT and AI
• Enhancement in the number of the training-testing scenario of the projects
related to smart charging infrastructure
• Capacity enhancement of power grid
• More focus and regularization in urban areas
• Spread awareness of e-mobility among urban areas

10.3 DESCRIPTION OF POWER MARKET

STAKEHOLDER INTERACTION MODEL
This section used different diagrams of Unified Modeling Language like activity
diagram used to represent different entities of the system and its activities, entity-
relation (ER) diagram constructed to demonstrate the relationship among the entities
that later used to create database schema of the software, and data flow diagram that
explains the flow of data during different tasks of the system.
Day-Ahead Market Energy Usage Bidding 255

10.3.1 Activity diAgrAm

Activity diagrams also known as use case diagrams illustrate to show an association
between different entities and their actions (use cases). In other words, it illustrates
to which functionality (s) the entities/actors in the system are involved. Each use case
diagrams manifest some relevant, beneficial, and observable results to the software
entities or other stakeholders of the software. It shows the individual work of each
entity and manifests very useful information to the system developer. For example,
in our case, Figure 10.1 represents the task of xEV users like registration, login, bid-
ing for the energy, information updates, changing bid, etc. and in the same way also
shows the task of charging stations and power grids like registration, login, view all
bidding, etc. This diagram also helps the programmer to identify coupling among
methods of entities.

10.3.2 dAtA Flow diAgrAm

A very essential need of a developer to construct a stable design of the software is to
understand the flow of information and distribution of the data among entities. Here,
data flow diagram (DFD) comes into the picture. DFD is a diagrammatical repre-
sentation of database modeling and relationship with the methods of entities (see
Figure 10.2). In the proposed system, mainly three tables of different energy stake-
holders are developed – xEVs_Users_DB (to keep track of all the information of the
xEV User), CSP_DB (storing charging service provider details), and EP_DB (hold

Registration

LogIn

Update
Information

Bid
xEVs User

View Biding

Electricity Provider
Change Bid

Finalize Bid

Charging Service Logout

Provider

FIGURE 10.1 Activity diagram.

256 Electric Vehicle Integration in a Smart Microgrid Environment

xEVs_User_DB
Registered
xEVs User

LogIn
EP_DB

Charging Service
Provider User_Bid_DB

Bidding

EP_Bid_DB
Finalize Bid
Electricity Provider
CSP_DB

FIGURE 10.2 DFD.

information regarding different power plants associated with the system). Another
two tables are User_Bid_DB and EP_Bid_DB. These two tables store all the infor-
mation regarding the bidding by users and power plants and provide insight for the
charging station admin to finalize the bidding.

10.3.3 er diAgrAm
Developers use ER diagram to store real-world objects into the database using its
parameters and attributes; e.g., in this proposed system, xEV users, charging service
provider, and power plants are the entities of the system. In the system, an entity can
be defined and distinguished by its attributes (Figure 10.3). All the entities and their
attributes are the following:

• xEV users. xEV user is the consumer of energy who bids for it by using an
online process, but it has to be authenticated on the system before accessing.
xEV user needs to be registered by filling the online form and providing the
details, e.g., xEV vehicle identification number (VIN), name, contact num-
ber, email, address, type of xEV battery, etc. After providing all the neces-
sary details, xEV users will receive USER_ID and self-generated password
to login to the system.
• Charging service providers. Charging service provider is the main entity
of the system. It is also registered in the same way as xEV users. It has dif-
ferent attributes, e.g., name, contact detail (address, contact number, fax,
email, etc.), price-quantity of energy, capacity, the places where xEV user
can relax during charging. It can log in on the system by system/self-gener-
ated CSP_ID and password.
Day-Ahead Market Energy Usage Bidding 257

Password Date

Contact Quantity Price

USER_ID
No.

M
xEV User canBid
Zip

Address
Nation

VIN
House No. State
/Street etc.

District V_Type

Battery
Capacity

Price

CSP_ID
Date Energy
Transformation
Password
M

Price Charging Service Provider

Contact
No.
M
Quantity Address
Nation
Zip
canBid
House No. State
EP_ID /Street etc.

District
Password M

Electricity Provider
Contact
No.
Address Energy
Nation Transformation
Zip
Price
House No. State
/Street etc.
Unit
District

FIGURE 10.3 ER diagram.

258 Electric Vehicle Integration in a Smart Microgrid Environment

• Electricity provider. It is the main entity of the system, which is responsible

to originate the energy and having attribute similar to the charging service
provider.

10.4 AI STRATEGIES

Energy bidding is a very sensitive process because of the nature of electricity that
cannot be stored like other sources of energy, coal, petroleum, etc. There is a need
for optimal bidding to be chosen for both (Energy Provider and xEV Users), and this
process is nothing but a time series forecasting, which can achieve any AI strategy.
Some of them are the following:

10.4.1 ArtiFiciAl neurAl networKs

Since energy demand in day-ahead market depends on previous demand, an ANN
model is implemented for prediction of the future energy price. In this algorithm
model train by using previous year data, and after that we can predict the upcom-
ing rate for bidding [31]. Previous years’ bidding data need to be split into training
and testing datasets. Training data can be used for model construction, and testing
data are for validating the results. Many metadata like weights, learning rates, the
numbers of hidden layers need to be initialized after splitting the datasets. Minimum
error and epochs are initialized. Many activation functions can be used for regression
time series forecasting. As Figure 10.4 demonstrates, vector X = [ x1 ,   x 2 ,  x3 , …  x n ]
defines an input layer which can be previous energy prices. And y will be the pre-
dicted price of the upcoming day. In the last step, we need to get the accuracy of the
model, which can be achieved using either the mean square error method (MSEM)
or the mean absolute error method (MAEM).

10.4.2 Autoregressive moving AverAge

In this [32] paper, the researcher has used to predict a load of power system using
autoregressive moving average (ARMA). In the same way, we can also work on fore-
casting bidding price quantity for the electricity. The predictions take place based on
a data sequence over time. Once model parameters are obtained from training set,
bidding price can be predicted for futuristic demand. For the time series, observa-
tions are recorded based on time in the following format:

…, …, …, Zt +l +1  ,  Zt +l  , Zt +l −1  , … , … , … (10.1)

The present term can be represented in the form of previous terms as follows:

Zt +l +1 =   ∅1 Zt +l +   ∅ 2 Zt +l −1  +  +∅ p Zt +l − p − θ1at +l    − θ q at +l − q +  at +l +1   (10.2)

where   ∅1 , ∅ 2 ∅ p , and θ1,… θ q are the constants. The model task is to find the val-
ues of these constants to reduce prediction error up to a certain limit.
Day-Ahead Market Energy Usage Bidding 259

Weight 2
Weight 1

a a
11 21

x1 a a
12 22

x2 a a
13 23

x3 … … y

… … … Output Layer

xn … ..

Input Layer a
a
1 2n
n

Hidden Layer 1 Hidden Layer 2

FIGURE 10.4 Demonstration of ANN.

10.4.3 support vector mAchine

This ML algorithm is used for classification and regression problems, but its model
can be extended feasibly to the time series prediction [33,34]. The following function
is used to solve regression problem:

f ( xi ) = w.∅ ( xi ) + b (10.3)

Where xi is the previous bidding price as input data, b is the bias, and w is the weight
vector. And ∅ ( xi ) is a Kernel function (KF), which has many types, but the most
used KF is radial basis function kernel:

 x −x2
( )
∅ xi ,  xi , σ = exp  − i 2 i 
 2σ 
(10.4)
260 Electric Vehicle Integration in a Smart Microgrid Environment

where xi denotes the mean, and σ represents the standard deviation of the input val-
ues. The regression line is best defined to that line which minimizes the following
cost function (CF):

CF =
1
2
w
2 
∑L (x , y , f )
+c ε
i i (10.5)

Using constraints,

yi − wxi − b ≤ ε + ξ i (10.6)

wxi + b − yi ≤ ε + ξ i* (10.7)

ξ i  ξ i* ≥ 0 (10.8)

where c is the positive constant, ε   is the deviation, and ξ i ,  ξ i* the slack variables. In
Ref. [33], the author used support vector machine (SVM) to predict the short-term
load on the grid.

10.5 OVERALL DEMONSTRATION

The proposed system is developed using many programming languages and environ-
ments. The entire server-related processing tasks and database interaction code are
written in PHP. On the client-side, interactive web pages are designed using HTML,
CSS, and Ajax. An android app is also designed to reduce the load on the server,
making it more user friendly, which is written in JAVA. The main task of the system
is to predict both, electricity provider and xEV users. Python of version 3.x is used
to implement AI predictions. The proposed expert system for the bidding process of
day-ahead market for energy using ML strategies illustrates the important ICT-based
paradigm focused to deploy a conventional system and coordination mechanisms to
operate a virtual energy market. It demonstrates the Bid and auction mechanism that
undertook between xEV users and charging stations stack holders working as energy
providers as well as power system stack holders with charging stations working as
energy consumers. The real-time information regarding predicted energy tariff, near-
est charging station, ensuring least queuing delay, etc. is easy and can be monitored
using the proposed system. A prototype of the model is represented in Figure 10.5.
The whole scenario is divided into four layers as represented in Figure 10.6. All the
layers have some significant tasks.

• Layer 1 is the first layer of the system, which contains all the end-users of the
system and has access to Layer 2 via client-side web pages or android app.
• Layer 2 has server-side codes and APIs which interact with databases on
the cloud.
Day-Ahead Market Energy Usage Bidding 261

FIGURE 10.5 Prototype of the model.

Layer 4 Artificial Intelligence Strategy

Layer 3

Layer 2

Layer 1

Charging Service Provider xEVs Users Electricity Provider

FIGURE 10.6 Overall demonstration.

• Layer 3 is the layer where a centralized database comes into existence,

which is available on the cloud. This layer provides all the data needed to
process AI activity, which happens on Layer 4.
• Layer 4 has AI algorithms and libraries that work with available data and
predict the information.
262 Electric Vehicle Integration in a Smart Microgrid Environment

10.6 CASE STUDIES

To justify the results and importance of the system, we have taken three case studies –
Forecasting of energy price, Aggregate Demand-Supply System, and xEV Market
Analysis and Forecast.

10.6.1 ForecAsting oF energy price

According to Ref. [15], AI is rapidly growing in the fields related to smart grid and
energy markets. In this study, two methods are introduced to predict price and price
spike, ARMA, and ANNs. In this case study, the model applied real data from the
Finnish Nord Pool Spot day-ahead energy market and compared both original price
and predicted price.

10.6.2 AggregAte demAnd-supply system

Figure 10.7 presents a chart showing one-day (29th March 2020) market scenario
(Buy & Sell). The total no. of buy bids on 29th March 2020 is 221 MW, and the total
no. of sell bids is 107 MW. All the informative data such as the no. of bidder, seller,
cleared volume, market-clearing price, and locations to withdraw and inject the bided
power and the details of the customer behavior based on previous traded transactions
will be stored in the cloud system to execute the proper day-ahead as well as term-
ahead energy trading mechanism [35].

FIGURE 10.7 Aggregate demand-supply system.

Day-Ahead Market Energy Usage Bidding 263

10.6.3 xev mArKet AnAlysis And ForecAst

According to Ref. [36], xEV market was evaluated at $162.34 billion in 2019 and
predicted to be $802.81 billion by 2027. In the world, Asia-Pacific was on the top of
revenue contribution of $84.84 billion in 2019, which is predicted to reach $357.81
billion by 2027 with a compound annual growth rate (CAGR) of 20.1%. Even though
EV infrastructure is in a nascent phase in India, EV market is expected to grow with
a CAGR of 43.13% and installation of charging Infrastructure to grow 42.38% during
the time from 2019 to 2030. The various aspects such as increasing demand for fuel,
high performance, and zero-emission government rule supplement the demand for
xEVs. We need to be ready to adopt this rapid growth of EVs and have to develop an
intelligent management system as proposed in this study.
Now from analysis of these case studies, we conclude that the applications of AI
may help a lot in the energy market to solve various engineering problems, and we
also observed that bidding for the energy is so frequent, which may cause a system
failure. That is why in this era of ICT and AI, we need to bring revolutionary changes
in the energy market via using the proposed expert system.

10.7 RESULT

The proposed work demonstrated an AI-based smart grid interface for the bidding
process of day-ahead market. The framework describes different parameters and cir-
cumstances to predict the future price for energy tariff and detailed ML methods
suitable for the task, e.g., ANN, ARMA, and SVM. Also, the system successfully
describes the various objectives of different stakeholder-specific infrastructures
that handle massive demand for electric vehicles into the present transport sector
through implementing conventional and coordination mechanisms to operate. The
model provides an extremely good analytical and computational power to operate
BigData because of many powerful tools and libraries introduced to compile ML
algorithms, e.g., Python, Sklearn, and cloud-based 3-tier architect database show-
ing that after the prediction, we can implement an optimal scheduling scheme for
xEV users. Moreover, the software also does dedicated jobs for xEV users, charging
providers, and other entities to evolve to the charging-bidding process. The proposed
system keeps track of previous data of biddings and predicts for the current situation.
The work also ensures the cyber security to maintain personal records of energy
stockholders.

10.8 
CONCLUSION
The proposed work presented an AI-based automated bidding expert system for the
day-ahead market of energy. The application provides stakeholder-specific interface
and module for optimal bidding for electricity. At the xEV user side, it provides an
efficient and user-friendly GUI (Graphical User Interface) for bidding, and for power
providers, it provides a suggestion for optimal service quotes to enhance the chance
for their selection. All these suggestions are based on AI strategies using and ana-
lyzing previous years’ bidding data. Moreover, it can reduce a load of processing on
264 Electric Vehicle Integration in a Smart Microgrid Environment

charging service providers by implementing some AI algorithms and will start work-
ing more and more accurately, as well as data grow on the cloud.

NOMENCLATURE
b: Bias
C: Positive constant
o(tk): Output training attribute at time tk
W: Weight
x(tk): Input training attribute at time tk
Xi: ith input vector
xi : Mean
Zt +l : l observations recorded based on time t
∅ p: pth Constant of time series
∅ ( xi ): Kernel value for Xi input vector
σ: Standard deviation
ξ i ,  ξ i*: Slack variables
Lε : Regularization term
ε: Deviation

REFERENCES

1. A. Holms and R. Argueta, “A Technical Research Report: The Electric Vehicle,” in
Business, 2010, pp. 1–12.

4. I. Abaker et al., “The rise of ‘big data’ on cloud computing : Review and open research
issues,” Inf. Syst., vol. 47, pp. 98–115, 2015.

7. M. Saqib, M. M. Hussain, M. S. Alam, M. M. S. Beg, and A. Sawant, “Public Opinion
on Viability of xEVs in India,” in ISGW 2017: Compendium of Technical Papers, 2018,
pp. 139–150.

9. A. K. Digalwar and G. Giridhar, “Interpretive Structural Modeling Approach for
Development of Electric Vehicle Market in India,” Procedia CIRP, vol. 26, no.
December, pp. 40–45, 2015.

Day-Ahead Market Energy Usage Bidding 265

12. I. S. Bayram, G. Michailidis, I. Papapanagiotou, and M. Devetsikiotis, “Decentralized

Control of Electric Vehicles in a Network of Fast Charging Stations,” Globecom
2013-Symposium Selected Areas in Communications (GC13 SAC), Atlanta, GA, pp.
2785–2790, 2013.

14. R. Weron, “Electricity Price Forecasting: A Review of the State-of-the-Art with a Look
into the Future,” Int. J. Forecast., vol. 30, no. 4, pp. 1030–1081, 2014.

18. T. Y. Kim and S. B. Cho, “Predicting Residential Energy Consumption Using CNN-
LSTM Neural Networks,” Energy, vol. 182, no. 1, pp. 72–81, 2019.
19. M. A. R. Biswas, M. D. Robinson, and N. Fumo, “Prediction of Residential Building
Energy Consumption: A Neural Network Approach,” Energy, vol. 117, no. 1, pp. 84–92,
2016.

23. M. Saqib, M. M. Hussain, M. S. Alam, M. M. Sufyan Beg, and Amol Sawant. “Smart
Electric Vehicle Charging through Cloud Monitoring and Management,” Technol. Econ.
Smart Grids Sustain. Energy, vol. 2, p. 18, 2017. https://doi.org/10.1007/s40866-017-0035-4.

25. I. Malmgren, “EVS29 symposium quantifying the societal benefits of electric vehicles,”
World Electr. Veh. J., vol. 8, no. June, pp. 996–1007, 2016.

266 Electric Vehicle Integration in a Smart Microgrid Environment

30. I. Strategic, M. Conference, O. F. Demirel, and S. Zaim, “Forecasting Electricity

Consumption with Neural Networks and Support Vector Regression,” 8th International
Strategic Management Conference, vol. 58, pp. 1576–1585, 2012.

35. A. Singh, “Electric Vehicle Market” [Online], Allied Market Research, Available: https://
www.alliedmarketresearch.com/electric-vehicle-market. [Accessed: 05-May-2020].
 11 Smart Microgrid-
Integrated EV Wireless
Charging Station
Aqueel Ahmad and Yasser Rafat
Aligarh Muslim University

Samir M. Shariff
Taibah University

Rakan Chabaan
Hyundai Kia America Technical Center Inc

CONTENTS
11.1 Introduction .................................................................................................. 267
11.2 Solar PV Module Configuration with a Wireless Charging System ............ 268
11.3 Solar to EV Battery Feasibility Analysis...................................................... 269
11.4 Wireless Charging System for EVs............................................................... 269
11.5 Finite Element Analysis Modeling and Simulation of the WPT Coils
for Magnetic Analysis................................................................................... 272
11.6 Results and Discussion ................................................................................. 273
11.7 Conclusion .................................................................................................... 277
Acknowledgment ................................................................................................... 278
References .............................................................................................................. 278

11.1 INTRODUCTION
The growing public concern on environmental problems and rising fuel costs has
led to growing emission-free, eco-friendly means of transportation [1]. Hence,
EVs and PHEVs are emerging as an alternative to ICEVs [2]. Various interna-
tional and national government such as (China, Japan, France, Germany, USA,
UK, Netherland, Norway, etc.) have been passed resolutions, owing to these
environmental concerns such as air quality, global warming, etc. and allocated
significant funds to promote PEV and EV implementation and deployment [3].
Enduring scenario planning specifies the global vehicle fleet capturing by 2050
especially attracted by renewable energy sources, to overcome the worst case of
global climate change scenario [4].

267
268 Electric Vehicle Integration in a Smart Microgrid Environment

The EV deployment has been directed toward challenges such as limited fuel
storage (i.e., small battery size or lower battery capacity) and range anxiety, which
restricts EV deployment due to a limited number of commercial EV charging sta-
tions in the long driving range of EVs. The conductive charging faces challenges
such as charger connectivity, manual connectivity, and limited life of contact char-
ger. However, wireless charging technology can address most of the challenges
confronted by the EV charging infrastructure. The EV wireless charging includes
the convenience of charging during rest as well as motion, safe due to noncontact
operation. Hence, EV charging using renewable energy will lead to pollution-free,
economic, and efficient transportation [5].
The solar energy received from the sun around the earth in 1 hour can provide
sufficient energy to fulfill a full year of utilization [6]. Solar energy is the cleanest
and most sustainable form of energy received by the earth. It will take estimated
another 40 years for the solar photovoltaic (SPV)-based energy generation to reach
its maximum potential [7]. The measurement of sunlight intensity on any plane sur-
face can be analyzed as energy per unit time per unit area, i.e., irradiance or power
per unit area. Approximately the planet earth receives about four million exajoules
(1 EJ = 1,018 J) of annual solar energy, out of which practically harvestable amount
of energy has been claimed as 5.104 EJ [8].
In this manuscript, a solar-based wireless charging system has been proposed,
where the power is transferred using the principle of magnetic induction, which is
based on Faraday’s law and Lenz’s law. Where the magnetic field can be generated by
a time-variant current-carrying conductor, the generated magnetic field is linked with
a secondary coil conductor to generate a time-varying magnetic field [9]. The receiver
is connected to the load to close the circuit; hence, the power is transferred wirelessly.
In this manuscript, SPV has been utilized to receive solar energy as a DC power
supply. This power is wirelessly transferred to charge the EV.

11.2 SOLAR PV MODULE CONFIGURATION WITH

A WIRELESS CHARGING SYSTEM
The block diagram of the solar PV system, grid, and EV wireless charging system is
shown in Figure 11.1. During daytime, the solar energy can be utilized in charging
EVs, and extrasolar power can be utilized to store the energy for off-day time utiliza-
tion. However, if during daytime, the SPV generation is lower than the demand of
the EV wireless charging system, then it is recommended to use the grid supply for
charging the EV. The wireless charging makes charging very convenient, and the
automation with this system may lead to a human-free system.
The major components of the solar-based wireless charging system are as follows:

1. Solar PV system. Depending on the demand, the capacity of the solar panel
can be decided; however, this manuscript presents the 3.3 kW solar-based
wireless charging system.
2. Solar inverter and Maximum Power Point Tracking (MPPT) controller. The
solar power generated from PV varies with the intensity of the solar radia-
tion. Hence, an inverter embedded with the MPPT controller is required.
Integrated EV Wireless Charging Station 269

FIGURE 11.1 Block diagram of the solar, energy storage, and EV system.

11.3 SOLAR TO EV BATTERY FEASIBILITY ANALYSIS

The solar energy utilized for charging EV is very efficient, sustainable, and environ-
mentally friendly. Many of the researchers have performed the solar-based analysis
to charge EVs such as in Refs [1–6]. However, there is none that discussed the solar-
based wireless charging system. Most of them have already determined the efficient
solar energy utilization, calculation of SPV requirement based on demand to charge
EVs [10], optimization of solar panel, and energy demand utilization.
In this manuscript, the wireless charging system has been discussed in detail,
which receives the input from the SPV system. Figure 11.2 presents the overall
concept of the proposed system.

11.4 WIRELESS CHARGING SYSTEM FOR EVs

The fundamental principle of application of wireless power transfer (WPT) is similar
to an air-cored transformer where the transmitter and receiver of the WPT system
are considered as primary and secondary coil transformers [12]. Figure 11.3 shows
270 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 11.2 Block diagram of the overall wireless charging system.

FIGURE 11.3 Simplified block diagram of two-coil WPT system.

a separate compensation topology of WPT. A simple WPT analysis has been per-
formed neglecting coil resistances and magnetic losses. Hence, the exchanged com-
plex power from LP and Ls can be evaluated as follows.
For the two-coil system, the primary to the secondary exchange of apparent power
is given by (11.1)
˙
SPS = −U PS IS* = − jω MIP IS* = ω MI P I S sin ϕ PS − jω cos ϕ PS (11.1)

Further, for the two-coil system, the secondary to the primary exchange of apparent
power is given by (11.2)
Integrated EV Wireless Charging Station 271

SSP = U SP IP* = − jω MIS IP* = −ω MI P I S sin ϕ PS − jω MI P I S cos ϕ PS (11.2)

Furthermore, the active power transfer from the primary side to the secondary side
can be expressed by (11.3)

PPS = ω MI P I S sin ϕ PS (11.3)

Hence, for the overall WPT system, the total reactive power (Q) developed between
the primary and secondary coils is (11.4)

(
Q = ω LP I P2 + LS I S2 + 2 MI P I S cos ϕ PS ) (11.4)

For the maximum efficiency of the WPT system, the ratio between the active power
( PPS ) and reactive power (Q ) should be maximized as shown in Equation 11.5:

PPS ω MI P I S sin ϕ PS
= (11.5)
Q ω LP I P + ω LS I S2 + 2ω MI P I S cos ϕ PS  
2

To achieve the maximum value of f (ϕ PS ) , Equation 11.6 can be solved as follows:

∂ ∂2
f (ϕ PS ) = 0 2 f (ϕ PS ) < 0 (11.6)
∂ϕ PS ∂ ϕ PS

Since we have neglected magnetic losses, it is considered as a traditional transformer,

and hence, the value of coupling coefficient (k) is close to 1 if the current induced by
Ip to the secondary side is Is , and the value of k is near 1; hence, the cosϕ PS is also
close to 1. However, there is approximately 180o phased difference between Ip and Is .
Further, in the case of the WPT system, the coupling coefficient is near 0. Hence, the
maximum value of f (ϕ PS ) is at sin ϕ PS = 1. Thus, WPT is maximized, and the phase
difference between the Ip and  Is is 90° instead of 180°. Hence, the difference between
the loosely coupled system and a tightly coupled system is observed. Furthermore,
the designing of a compensation network depends on the coupling degree. For the
series-series coupling, there are two ways to design a resonant capacitor. Firstly, for
tight coupling such as WPT, where k > 0.5, the f (ϕ PS ) has to be increased to achieve
high efficiency. Further, for the self-coil inductance, the resonance can be achieved
if ϕ PS = π/2 and f (ϕ PS ) is low. Due to an increase in magnetic loss, the above con-
dition is not recommended. The traditional transformer can work with resonance
compensation with leakage inductance if there is an increase in f (ϕ PS ); however,
the overall WPT system would not perform at the resonant condition. For wireless
power system, i.e., at loosely couple transformer system where k < 0.5 for the reso-
nance condition, the tuning of the capacitor with self-inductance is required for the
maximum WPT. For this system, electrical energy is stored between the primary and
secondary windings in the form of a magnetic field. The copper loss is proportional
to the square of the conducting current. To achieve maximum efficiency, the current
272 Electric Vehicle Integration in a Smart Microgrid Environment

induced in secondary coil current Is must lag to the current induced to the primary
current Ip by 90o. Is and U PS should be in phase since U PS and Ip lag by 90o on the
receiving coil. The pure resistive characteristics are found on the secondary side.
For maximum power transfer, the induced current in the secondary coil Is must
lag to primary induced current Ip by 90o. U PS and Is should be in phase since U PS
and Ip lag by 90o on the receiving coil. On the secondary side, the pure resistive
characteristic is seen. Meanwhile, at the primary side, the apparent power S3 must be

()
reduced. When cos ϕ PS   = 0, the complex power S1 ± is
˙

S1 = jω LP I P2 + ω MLP I S (11.7)

k 2 QP QS
Maximum efficiency is ηmax = , which is achieved at
( )
2
1 + 1 + k 2 QP QS

α η  max = 1 + k 2 QP QS (11.8)

The maximum efficiency for various types of compensation topologies has been
derived by many authors. For the efficiency evaluation of static wireless charging, if
the primary and secondary coils have a quality factor of 300 and coupling coefficient
between coils around 0.1–0.25, the theoretical calculation states that WPT efficiency
will be approximately 96.7%.

11.5 FINITE ELEMENT ANALYSIS MODELING AND SIMULATION

OF THE WPT COILS FOR MAGNETIC ANALYSIS
The analysis performed in Section 11.4 presents the factors affecting the WPT sys-
tem such as the coupling coefficient, transmitter side quality factor, and receiver side
quality factor. The coupling coefficient is related to the designing of the primary
and secondary coils, and quality factors are related to the power electronics circuit
design. In this section, we are focusing on the coupling coefficient. Hence, to analyze
the coupling coefficient, finite element analysis (FEA) has been performed for the
transmitter and receiver coil. The modeling of the FEA model has been designed in
the Ansys Maxwell® as shown in Figure 11.4.
The circular coils for the transmitter and receiver have been taken for the analy-
sis. For the performance analysis, the airgap and horizontal misalignment have
also been varied to present the robustness of the system. The vertical air gap has
been varied from 100 to 150 mm, whereas the horizontal misalignment from 0 to
50 mm in a single direction. The overall parameters for the FEA model have been
listed in Table 11.1. The frequency of analysis is 85 kHz. The maximum number
of passes is 10, with a percentage error of 1%. The convergence parameters are
refinement per pass 30%, minimum converged passes 1, and the minimum number
of passes 2. Nonlinear solver with residual on 0.01 has been taken. The solution
type is magnetostatic, which shows the finite element model for the circular coil
WPT system.
Integrated EV Wireless Charging Station 273

FIGURE 11.4 Circular coil model for a wireless charging system.

TABLE 11.1
Simulation Parameters for the Analysis
Parameters Symbol Values
Number of turns in the transmitter coil NT 10
Number of turns in the receiver coil NR 10
The outer radius of the coil ro 75 mm
Internal radius of the coil ri 50 mm
Conductor diameter D 2 mm
The air gap distance between the coils d 100 mm
Current in the receiver coil IR 8A
Current in the transmitter coil IT 8A

11.6 RESULTS AND DISCUSSION

The FEA results present a very small coupling coefficient of about 0.05. However,
this can be improved by further improvement in the power pad design, impedance
matching, and performing the WPT at resonance condition. The coupling coeffi-
cient for horizontal misalignment and vertical height variation has been presented
in Figure 11.5. The graph shows the reduction in the coupling coefficient while
increasing the air gap between the transmitter and receiver, vertically as well as
horizontally.
A very important factor that determines the WPT efficiency is the coupling coef-
ficient. Figure 11.5 shows the coupling coefficient of the presented WPT model. The
results show the variation of the coupling coefficient when there is an increase in the
air gap and when the horizontal alignment between the power pads has been varied.
Figures 11.6 and 11.7 present the magnetic field plot around the transmitter coil,
which verifies the reduction in the magnetic field with an increase in the distance
from the coil.
274 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 11.5 A 3D graphical presentation of the results from FEA analysis.

FIGURE 11.6 Plot presenting the magnetic field around the transmitter coil.
Integrated EV Wireless Charging Station 275

FIGURE 11.7 Stepwise magnetic flux pattern around the transmitter coil.

The results from the above analysis have been presented in Table 11.2 where the
coupling coefficient is 0.53645. The value of the coupling coefficient is very small as
compared to the transformer application, due to the air as medium and larger air gap
and since no core has been used.
The results present the feasibility of the WPT system for the implementation of
the solar-based wireless charging system. Figure 11.8 presents the overall wireless
charging system magnetic field plot in the air as a medium. The power transfer effi-
ciency of the available WPT system can be more than 85%, which is less than the
conductive wireless charging.

TABLE 11.2
Result for the FEA Analysis at Zero Misalignments and
100 mm Height
Parameters Symbol Values
Mutual inductance M 1.024816 μH
Self-inductance of the receiver coil LR 19.130200 μH
Self-inductance of the transmitter coil LT 19.077310 μH
Coupling coefficient k 0.053645
Magnetic flux of transmitter ϕT 0.000145 Wb
Magnetic flux of receiver ϕR 0.000144 Wb
276 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 11.8 Overall plot of the WPT system.

The Ansys Maxwell model has been exported to Ansys Simplorer and through
magnetostatic analysis. Using series-series compensation topology, the input and
output power has been analyzed and recorded. Figure 11.9 shows the Simplorer
model, and Figure 11.10 shows the output waveform.
The FEA model has been fed to the Ansys Simplorer to analyze circuit analy-
sis of the wireless charging system. The Simplorer circuit parameters are listed in
Table 11.3. Figure 11.10 presents the input and output power waveforms.

FIGURE 11.9 Simplorer model for transient simulation of a solar-powered wireless charging
system with LCC topology.
Integrated EV Wireless Charging Station 277

FIGURE 11.10 Input and output power across the wireless charging system.

TABLE 11.3
Electrical and Circuit Parameter for the
Solar-powered Wireless Charging System
S. No Parameter Values
1. VS   220 V
2. RL 4.61 Ω
3. LP 62 µH
4. LS 43 µH
5. Lf 100 µH
6. Crp 54.6 nF
7. CP 9.02 µF
8. CS 42.67 nF
9. Crs 84.2 µH
10. Cf 500 µF
11. RL 18 Ω

11.7 CONCLUSION
The manuscript presents a sustainable, environmentally friendly, and economic idea
of charging EVs using solar energy. The circuit analysis for the wireless charging
system has been performed to present the factors that affect the efficiency of the
WPT system where the coupling coefficient is found to be a major factor that influ-
ences the efficiency of the WPT system. Further, the coupling coefficient for the
WPT system has been analyzed using FEA modeling and simulation. The results
present the variation in the coupling coefficient by varying the alignment between
the transmitter and receiver coil. However, the perfect alignment may lead to more
278 Electric Vehicle Integration in a Smart Microgrid Environment

than 85% efficiency of the WPT. Hence, the EV charging using solar energy is a very
efficient and sustainable solution for the clean environment. However, for the coun-
tries having solar energy available throughout the year, the presented idea is a very
efficient and viable solution for the deployment of EV technology.

ACKNOWLEDGMENT
This work was supported by the Center of Advanced Research in Electrified
Transportation, Aligarh Muslim University, through the Science and Engineering
Research Board (SERB)-approved IMPRINT-2 Project (IMP/2018/001267).

REFERENCES

 12 Shielding Techniques
of IPT System for
Electric Vehicles’
Stationary Charging
Ahmed A. S. Mohamed
National Renewable Energy Laboratory (NREL)

Ahmed A. Shaier
Zagazig University

CONTENTS
12.1 Introduction .................................................................................................. 279
12.2 Components of Transmitter and Receiver Pad ............................................. 281
12.2.1 Conductive Wires ............................................................................. 281
12.2.2 Flux Concentrator ............................................................................. 281
12.2.3 EMF Shielding .................................................................................. 283
12.2.3.1 Passive Shielding................................................................ 283
12.2.3.2 Active Shielding ................................................................. 285
12.2.3.3 Reactive Shielding ............................................................. 287
12.3 Conclusion .................................................................................................... 288
Acknowledgment ................................................................................................... 289
References .............................................................................................................. 289

12.1 INTRODUCTION
The transportation sector is one of the primary consumers of fossil fuels in the world,
which makes it the biggest contributor to the greenhouse gases (GHGs) [1]. Clean
transportation technologies are crucial to reduce the dependency of fossil fuels and
emission of GHGs. Electric vehicles (EVs) are one of the primary players in this
space, due to the associated advantages related to performance, emission, and safety.
Charging infrastructure of EVs is one of the main challenges that slows down the
EV penetration market. Among the charging technologies, inductive charging shows
promising features for EV because of being automatic, convenient, reliable in harsh
environment, durable against vandalism, and flexible (can be implemented on the
road, public parking, and private parking) [2,3]. Inductive charging methods present a
new revolution in the EV industry. Unlike conventional plug-in tethered to a charger,
279
280 Electric Vehicle Integration in a Smart Microgrid Environment

no power connection is needed and, instead, an EV can charge its battery remotely,
either during long-term parking (stationary), driving (dynamic), or transient stops
(quasi-dynamic). The conventional configuration of an EV inductive charger is
depicted in Figure 12.1. The system includes two galvanically isolated sides: primary
(grid) and secondary (vehicle). The former includes a transmitter coil, which is con-
nected to the power source through a grid-rectifier, high-frequency (HF) inverter, and
compensation network. The latter consists of a receiver coil, which is coupled to the
EV battery though a compensation network and rectifier. The transmitter is embedded
in the road and is responsible for generating HF electromagnetic fields (EMFs) that are
coupled to the receiver when the vehicle exists above the system. The linked EMFs
induce power in the secondary circuit, which is rectified and stored into the battery.
The system operates at high frequency (79–90 kHz) [4], which helps to reduce the sys-
tem components’ size (transmitter, receiver, power converters, etc.) and enhance the
power transfer [5,6]. For high-power and high-efficiency operation, resonance capaci-
tors are connected to both the transmitter and receiver coils. These capacitors help
to compensate for the large leakage inductances due to the large airgap and provide
the required reactive power for magnetizing this airgap. These capacitors can be con-
nected in series, or in parallel, and can be a combination of LC circuits [7–9].
In indictive power transfer (IPT) systems, the power transfers from the transmit-
ter to a receiver by transmitting a significant amount of magnetic field through a
large airgap. Part of these fields is coupled with the receiver coil to represent the
useful power, and the other part leaks in the air around the system. If these leakage
fields exceed the safe limits, they have the potential to present safety concern to the
living objects in the proximity of the system [10]. These fields may result in high
induced currents inside human organs, which lead to exposure of the body tissues
to heat stress and create serious risks to human health [3,11]. In addition, these leak-
age EMFs may have an adverse impact on the portable medical devices, such as
pacemakers, by interrupting their operation [12,13]. Several international entities
have defined safe limits for the leakage EMFs at different operating frequencies. The
J2954 committee recommended considering the 2010 International Commission on
Non-Ionizing Radiation Protection (ICNIRP 2010) limits for IPT systems [14]. The
ICNIRP 2010 recommends limits of external magnetic field density (B) of 27 µT for

FIGURE 12.1 Inductive charger system components. (Generated by authors).

Shielding Techniques of IPT System 281

humans and 15 µT for pacemakers [15,16]. In order to meet these limits, the transmit-
ter and receiver need to be carefully designed, optimized, and shielded.

12.2 COMPONENTS OF TRANSMITTER AND RECEIVER PAD

The inductive pad is the most sensitive part of IPT system, as it is responsible for
the energy transfer from the source to the vehicle. An IPT system includes two pads
(transmitter and receiver), and each pad incorporates three main parts: conductive
wires, flux concentrator, and EMF shield.

12.2.1 conductive wires

The conductive wires carry HF currents that are responsible for generating magnetic
fields. The high operating frequency of IPT system results in high eddy current losses
in the windings due to skin and proximate effects, which leads to high coil ac resis-
tance. High coil resistance means less quality factor and efficiency [17]. Therefore,
there is an urgent need for using special types and structures of wires in IPT systems
to minimize the conductive losses and enhance the system efficiency. Several classes
of wires have been tested, demonstrated, and reported in the literature for IPT systems
that show low ac resistance, such as litz wire [18,19], magneto-plate wire [20,21], mag-
neto-coated wire [22,23], tubular conductor [24,25], REBCO wire [26], and Cu-clad-Al
wire [18,27], as indicated in Figure 12.2. Analyzing the different wires shows a trad-
eoff between performance and cost, as summarized in Table 12.1 [28].

12.2.2 Flux concentrAtor

As HF currents flow in the wires, very fast pulsating magnetic fields are generated in
form of closed loops around the source coil. The strength of these loops diminishes

FIGURE 12.2 Types of conductive wires; (a) litz wire, (b), magneto-plate wire (c) magneto-
coated wire, (d) tubular conductor, (e) REBCO wire, and (f) Cu-clad-Al wire. (Generated by
authors).
282 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 12.1
Characteristics of Different Types of Wire
Tubular
Characteristic Litz Wire LMPW LMCW Copper REBCO CCA
Rskin Medium Very low Very low Low Low Medium
Rprox Medium Very low Very low Low Low Medium
Rhys - Low Low - - -
Rac Medium Very low Very low Low Medium Slightly high
Cost High Very high Slightly high Medium High Low
Complexity Medium High Medium Low High Low
Density (m3/kW) Medium High High Slightly high Medium small
Flexibility High Low Low Medium Slightly high High
References [18,19] [20,21] [22,23] [24,29] [26,30] [18,27]

with distance, which limits the charging distance (< 1 m). Flux concentrators are
typically used in IPT systems to direct the flux lines from the transmitter toward
the receiver [31]. These concentrators help to enhance the coupling performance and
system efficiency and reduce the leakage EMFs around the system. Flux concentra-
tors are typically made of a magnetic material that is highly conductive for magnetic
fields. Ferrite cores are the most used as flux concentrator due to their high magnetic
permeability and low electrical conductivity [32]. In Refs [33,34], a ferromagnetic
nanoparticle material was proposed for the IPT system, which offers higher power
transfer capability, improves the shielding performance, and reduces the system
weight [35]. Utilization of flexible magnetic cores for IPT system was discussed in
Refs [36,37], which leads to less core losses and better-quality factor and system per-
formance. Ferrite, magnetic nanoparticle, and flexible core are more convenient for
vehicle pad; however, they are not ideal for transmitter pad as they are incompatible
with the road. Any cracks in the road will damage the transmitter pad. Therefore, a
magnetizable concrete was developed and proposed for a transmitter design in Refs
[38,39]. It is very cheap since the magnetic particles can be made of recycled materi-
als [40]. A comparison among the different magnetic materials for flux concentrator
is presented in Table 12.2.

TABLE 12.2
Characteristics of Magnetic Materials
Material Cost Flexibility Weight References
Ferrite High Hard and fragile High [33,41]
Magnetizable concrete Very low High Very high [39,40]
Flexible magnetic Low Very high Low [36,37]
Nanoparticles Medium Medium Low [33,34]
Shielding Techniques of IPT System 283

12.2.3 emF shielding

IPT-based EV charger involves high power (up to few hundreds of kilowatts) to be
transferred through a relatively large distance based on magnetic fields. Strong EMFs
are typically generated around the system during charging. These fields may surpass
the safe limits reported in the international standards and guidelines [10,42]. EMF
shielding is typically used in IPT systems to minimize these leakage EMFs, thus
improving the coupling performance leading to better efficiency and quality factor
[43]. Different types of shield have been reported in the literature, such as passive
(magnetic, conductive, or both), active, and reactive, as indicated in Figure 12.3.

12.2.3.1 Passive
 Shielding
Passive shielding is achieved by adding a passive component (conductor or/and mag-
netic) that helps to block and/or direct the EMFs and reduce the leakage part. Magnetic
passive shielding is manufactured from nonconductive materials with high magnetic
permeability to direct magnetic flux lines, enhance self and mutual inductance,
improve the system performance, and reduce the leakage flux [31,42]. Ferrite flux
concentrators discussed in Section 12.2.2 provide this shielding functionality [43].
These cores are installed in the pad as a plate, multiple bars, or multiple tiles, as in
Refs [44–46]. The design of an inductive pad with the use of ferrite only as a magnetic
shield is illustrated in Figure 12.4. Although using magnetic materials for shielding
shows good performance, it increases the system weight and cost. Therefore, another
direction was taken to minimize the magnetic materials in the system for flux con-
centration and add a lighter and less expensive conductor for shielding. Therefore, the
trend of IPT system design is to combine a magnetic material with another type of
shield, whether conductive passive, active, or reactive [47,48].
The conductive passive shield combined with magnetic shield is depicted in
Figure 12.5. It is made of copper or aluminum [49,50] and installed under the mag-
netic core in the case of transmitter or over the core in the case of receiver. It also can
be attached directly to the coil in case of air-core coil, as indicated in Figure 12.6. It
is a conductive plate that acts as another coil, in which eddy currents are generated
due to the HF EMFs [51]. These eddy currents produce magnetic fields that oppose
the original fields and reduce them. This helps to minimize the leakage flux around
the system but adversely affects the system performance by reducing the coupling
factor and efficiency [31,47,52,53]. Several studies are presented in the literature to

FIGURE 12.3 Types of EMF shielding in the IPT system. (Generated by authors).
284 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 12.4 Inductive pad with the use of ferrite only as a magnetic shield. (Generated
by authors).

FIGURE 12.5 Inductive pad with the use conductive passive shield. (Generated by authors).

optimize the shield design for reducing its negative impact on the system efficiency,
considering different materials, such as copper [54], aluminum [55,56], different
dimensions [57,58], and different positions [49,59,60]. In Refs [31,48], the impact of
magnetic and conductive passive shield on the system performance (self-inductance,
coupling factor, efficiency, and EMF emissions) was investigated. It was concluded
that the use of ferrite core increases the power transfer, transmission efficiency, and
coupling coefficient between the two coils. In addition, the consideration of the vehi-
cle chassis as an additional passive conductive shielding and investigating its impact
on the IPT system performance were presented in Refs [41,49,61].
In Ref. [42], a 100 kW IPT system with matched DD coils has been laboratory-
tested at 85 kHz frequency with an aluminum plate only as shielding, which shows
a 25% increase in EMF emissions. The study proposed an extra magnetic loop of
Shielding Techniques of IPT System 285

FIGURE 12.6 Inductive pad with the use of conductive shield. (Generated by authors).

ferrite bars to be added around the coil for minimizing the leakage flux, which was
able to reduce the emission by 60%. In Ref. [10], a comparison was made between
two types of passive shields for circular pad: the traditional aluminum plate (Figure
12.7a) and a thin aluminum plate surrounded by a copper shield-ring (Figure 12.7b).
It was concluded that the use of a copper ring with aluminum plate as a shield reduces
the losses by 21% compared to the traditional aluminum.

12.2.3.2 Active
 Shielding
For high-power IPT systems (>100 kW), it is a challenge to manage the leakage
EMFs around the vehicle and keep them within the safe limits using the conven-
tional passive shielding [42]. Therefore, active shielding, as a more effective shield-
ing technique, was investigated for the IPT system [62]. In this case, extra turns are
added to each coil and wounded with reverse polarities, as depicted in Figure 12.8.
When current passes through the original coils, the same current will pass in the
shielding turns that generate intentional EMFs that have the same frequency and
value of the original fields but oppose them [47,52] acting as cancelling fields to
minimize the  leakage EMFs. This type shows an effective shielding performance

FIGURE 12.7 (a) traditional aluminum plate and (b) copper ring-based shield. (Generated
by authors).
286 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 12.8 Principle of active shield coil system. (Generated by authors).

compared to the passive one; however, the negative impact on the original field and
the system performance is worse. In addition, adding extra turns increases the system
cost, weight, and coil losses. Both passive conductor and active shield are used with
magnetic shield [47].
In Ref. [63], an active coil, placed on the primary side only, was used for shield-
ing of the planar coils, which resulted in a significant reduction in the leaked EMF
around the system. Two active coil structures were explored and compared: tradi-
tional active coil (Figure 12.9a) and adopted active coil (Figure 12.9b). The system
was tested at 85 kHz frequency and 7.7 kW output power. The impact of the two
shields on the self-inductance and mutual inductance and coupling coefficient was
studied. The adopted active coil shows a significant reduction in the leakage EMF
around the system.
In Ref. [64], planar circular coils were tested by applying active shielding coil on
the primary and secondary side. Two structures were presented based on the connec-
tion of the active coils: the inductive-based structure, in which the shield coils are
electrically isolated, and the shield currents are induced using magnetic induction,

FIGURE 12.9 Active coil system, (a) traditional active shielding, and (b) proposed active
shielding system. (Generated by authors).
Shielding Techniques of IPT System 287

as depicted in Figure 12.10a. The other structure involves shield coils connected in
series with the main coils at opposite polarities, as described in Figure 12.10b. The
impact of different number of turns at the shield coil on the transmission efficiency,
coupling coefficient, and EMF distribution was analyzed. The study showed that
with a small number of turns in the inductive structure, the system was able to pro-
vide a better shielding effect than the case of using a large number of turns with the
series-connected structure.

12.2.3.3 Reactive
 Shielding
For reducing the negative impact of the active shield on the system performance, the
resonant reactive shield was proposed in Refs [47,52]. It depends on the use of a pas-
sive compensation loop coil with a resonant capacitor, as indicated in Figure 12.11.
This type doesn’t need to be powered to produce an opposite intentional field.
However, an extra shield coil is added near the original one and wound in a way such
that when the original magnetic field passes through the shielding coil, an induced
voltage is generated, which results in a HF current in the shield coil that generates
the opposite magnetic field to cancel the original one. This structure shows a better
system efficiency than the active shield case [52].
In Ref. [47], reactive shield was considered in an inductively charged electric bus
to reduce the EMF around the system which consists of two circular pads. The impact
of the shield on the self-inductance and mutual inductance and magnetic field was
investigated. It was found that the reactive shielding was able to reduce the mag-
netic field by 64%, with minimal impact of the system parameters. Furthermore, the
study concluded that reactive shielding is more effective and efficient than the pas-
sive conductive shielding of the same size. In Ref. [60], a different configuration for
the shield capacitor was presented and compared with the conventional one. Four
phase shifter capacitors are used in the shield and called double reactive shield, as
depicted in Figure 12.12. The use of phase shifter capacitors leads to the generation
of a magnetic field that opposes the original field and is more effective and efficient
in reducing leakage magnetic field around the system. In Ref. [65], various types of
shielding (passive, active, and reactive) were analyzed and compared for circular coils.

FIGURE 12.10 Shielding coils: (a) inductive structure and (b) series-opposing structure.
(Generated by authors)
288 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 12.11 Principle of reactive shield system. (Generated by authors).

FIGURE 12.12 Double reactive shield. (Generated by authors).

Their impact on the system losses, EMF, and self-inductance and mutual inductance
was investigated. Different operating cases were explored, including no shielding
case, the use of the vehicle’s body as shield, using passive shield, applying a combina-
tion of active and passive shield, and finally when passive, active, and reactive shields
are applied together.

12.3 
CONCLUSION
This chapter presents an inclusive study and review for the current state-of-the-art
of the shielding techniques in inductive power transfer systems for EV charging.
Different types of shielding (passive, active, and reactive) are presented and com-
pared in detail. As a conclusion, passive shields are appropriate for low- and medium-
power IPT systems (<100 kW), and they should be able to bring the EMF levels to be
below the safe limits. In addition, they are cheaper, simpler in implementation and
design, and more robust. For high-power IPT systems, active shielding and reactive
Shielding Techniques of IPT System 289

shielding are more promising for the system to comply with the standard limits of
EMFs. This chapter provides comprehensive guidelines for researchers, students,
and engineers who are interested in designing an appropriate shield for inductive
charging technology.

ACKNOWLEDGMENT
The National Renewable Energy Laboratory (NREL) is the current address for the
first author only. NREL and the U.S. Department of Energy (DOE) did not contribute
to this work.

REFERENCES
1. US Environmental Protection Agency, “Sources of Greenhouse Gas Emissions,” US
EPA, Dec. 29, 2015. https://www.epa.gov/ghgemissions/sources-greenhouse-gas-emis-
sions (accessed Oct. 28, 2019).
2. M. Budhia, G. A. Covic, and J. T. Boys, “Design and optimization of circular magnetic
structures for lumped inductive power transfer systems,” IEEE Trans. Power Electron.,
vol. 26, no. 11, pp. 3096–3108, 2011.
3. A. A. S. Mohamed, A. Meintz, P. Schrafel, and A. Calabro, “In-Vehicle Assessment of
Human Exposure to EMFs from 25-kW WPT System Based on Near-Field Analysis,”
2018 IEEE Vehicle Power and Propulsion Conference (VPPC), Chicago, IL, 2018, pp.
1–6, doi: 10.1109/VPPC.2018.8605011.
4. X. Lu, P. Wang, D. Niyato, D. I. Kim, and Z. Han, “Wireless charging technologies:
Fundamentals, standards, and network applications,” IEEE Commun. Surv. Tutor., vol.
18, no. 2, pp. 1413–1452, 2015.
5. A. A. S. Mohamed, D. Allen, T. Youssef, and O. Mohammed, “Optimal Design of High
Frequency H-Bridge Inverter for Wireless Power Transfer Systems in EV Applications,”
2016 IEEE 16th International Conference on Environment and Electrical Engineering
(EEEIC), Florence, Italy, 2016, pp. 1–6, doi: 10.1109/EEEIC.2016.7555646.
6. Shuo Wang and D. Dorrell, “Review of Wireless Charging Coupler for Electric
Vehicles,” IECON 2013 - 39th Annual Conference of the IEEE Industrial Electronics
Society, Vienna, Austria, 2013, pp. 7274–7279, doi: 10.1109/IECON.2013.6700342.
7. A. A. S. Mohamed, A. Berzoy, F. G. N. de Almeida, and O. Mohammed, “Modeling and
assessment analysis of various compensation topologies in bidirectional IWPT system
for EV applications,” IEEE Trans. Ind. Appl., vol. 53, no. 5, pp. 4973–4984, Sep. 2017,
doi: 10.1109/TIA.2017.2700281.

10. A. A. S. Mohamed, A. Meintz, P. Schrafel, and A. Calabro, “Testing and assessment of

EMFs and touch currents from 25-kW IPT system for medium-duty EVs,” IEEE Trans.
Veh. Technol., vol. 68, no. 8, pp. 7477–7487, Aug. 2019, doi: 10.1109/TVT.2019.2920827.

290 Electric Vehicle Integration in a Smart Microgrid Environment

12. C. Xiao, K. Wei, D. Cheng, and Y. Liu, “Wireless charging system considering eddy
current in cardiac pacemaker shell: theoretical modeling, experiments, and safety sim-
ulations,” IEEE Trans. Ind. Electron., vol. 64, no. 5, pp. 3978–3988, May 2017, doi:
10.1109/TIE.2016.2645142.

16. S. Cruciani, T. Campi, F. Maradei, and M. Feliziani, “Wireless Charging in Electric

Vehicles: EMI/EMC Risk Mitigation in Pacemakers by Active Coils,” 2019 IEEE PELS
Workshop on Emerging Technologies: Wireless Power Transfer (WoW), London, Jun.
2019, pp. 173–176, doi: 10.1109/WoW45936.2019.9030634.

Shielding Techniques of IPT System 291

292 Electric Vehicle Integration in a Smart Microgrid Environment

51. L. Tan, J. Li, C. Chen, C. Yan, J. Guo, and X. Huang, “Analysis and performance
improvement of WPT systems in the environment of single non-ferromagnetic metal
plates,” Energies, vol. 9, no. 8, p. 576, Aug. 2016, doi: 10.3390/en9080576.

Shielding Techniques of IPT System 293

64. T. Campi, S. Cruciani, F. Maradei, and M. Feliziani, “Active Coil System for
Magnetic Field Reduction in an Automotive Wireless Power Transfer System,” 2019
IEEE International Symposium on Electromagnetic Compatibility, Signal & Power
Integrity (EMC+SIPI), New Orleans, LA, July 2019, pp. 189–192, doi: 10.1109/
ISEMC.2019.8825202.

 13 Economic Placement of
EV Charging Stations
within Urban Areas
Ahmed Ibrahim AbdelAzim
Ethos Esco Consultancy

CONTENTS
13.1 Introduction .................................................................................................. 295
13.2 The Problem of Choosing Charging Stations’ Locations ............................. 297
13.3 Methodologies for Placing Charging Stations .............................................. 299
13.4 Economics of Charging Station Placement ..................................................302
13.5 Case Study: Applying an Agent-Based Network Graph Placement
Method on Cairo, Egypt ............................................................................... 305
References .............................................................................................................. 311

13.1 INTRODUCTION
The environmental footprint of the transportation sector is huge; it consumes around
19% of the world total energy use, with light-duty vehicles (including passenger cars)
being the most significant contributor [1]. In 2015, governments around the world
endorsed 17 sustainable development goals (SDGs), including SDG 7, which is to
secure affordable and clean energy. The promotion of leaner transport technologies,
most importantly electric vehicles (EVs), contributes to achieving SDG 7 and is a
cornerstone of the efforts to reduce transportation sector-related emissions [2]. The
United Nations Environment Programme estimates that 20% of all road vehicles
must be electric-powered by 2030 to stay within the global 2°C climate scenario
[3]. EVs reduce fossil fuel consumption and GHG emissions, especially when using
electricity generated from renewable resources.
The increasingly stringent regulations, the growing public awareness of environ-
mental challenges, and the volatility of oil prices have encouraged car makers not
only to reduce their conventional vehicles’ emissions but also to venture into produc-
ing hybrid and fully battery-powered EVs [4]. The number of EVs reached 5.1 million
globally in 2018, with China and Norway leading the market in sales and market
share. Forecast scenarios, which account for the impact of announced policy ambi-
tions and pledges, estimate that global EV sales will be between 23 and 43 million

295
296 Electric Vehicle Integration in a Smart Microgrid Environment

in 2030, while the total EV stock is expected to be between 130 and 250 million (not
including two and three wheelers) [5].
Potential adopters of any new product (or service) usually inquire about two main
aspects: practicality and pricing. People compare new products to the ones they are
currently using based mainly on these two aspects. Any marketing efforts to con-
vince the users to adopt the new product or service instead of what they are currently
used to (which is called conversion in marketing terminology) must typically address
practicality and price among other things.
Like any new technology, the potential adopters are often skeptic about practical-
ity, as compared to the established technology. In the case of EVs, people simply look
around and observe that there are more gasoline fueling stations than EV charging
stations on the streets. As such, they might be worried that the EV’s battery could run
out of charge before reaching either his or her destination or a charging point. This
perception is known as “range anxiety”, and it is a psychological barrier to large-
scale adoption of EVs. Solving the lack of conveniently placed EV charging infra-
structure is therefore an important issue, which can significantly encourage people to
adopt EV cars. If people see more and more EV charging stations, their perception
will change, and range anxiety might ease off.
In terms of pricing, and with battery technology advancement being a signifi-
cant driver, EV prices are dropping. The price of lithium-ion (Li-ion) battery storage
dropped around 87% between 2010 and 2019 (see Figure 13.1). Consequently, pas-
senger EV car initial prices, although still generally higher, became comparable with
conventional fuel-powered cars.
In terms of operating costs, EVs are significantly cheaper; maintenance is mainly
related to changing brake fluid, coolant, or tires, as opposed to the more costly
internal combustion engine maintenance. The American Automobile Association

1,400

1,183
1,200

1,000
917
Price ($/kWh)

800
721
663
588
600

400 381
293
219
200 180
156

-
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
Year

FIGURE 13.1 Li-ion battery price trend (2010–2019).

Economic Placement of EV Charging Stations 297

estimates the annual maintenance and repair costs of gasoline cars to be around
1,200 $ per annum, versus around 450 $ for EVs. Figure 13.2 below shows a life
cycle cost (LCC) comparison of a gasoline and electric compact passenger car, based
on a gasoline price of 2.6 $/gal, an electricity tariff of 13.2 $cents/kWh, and around
10,200 miles of driving per annum.
Another important aspect is the environmental aspect. Conventional vehicles,
despite the efficiency improvement imposed by laws, still inevitably emit exhaust
gases. For illustration, the gasoline-fueled Volkswagen Golf in the above LCC analy-
sis emits around 2.9 tons of CO2 equivalent per annum.
It can be predicted that EV growth will first happen in urban centers, where daily
trip distances are not very long, and consequently, large EV battery ranges are not
needed. Early EV adopters living in cities may therefore be encouraged to purchase
and use EVs if the charging infrastructure is well established.

13.2 THE PROBLEM OF CHOOSING CHARGING

STATIONS’ LOCATIONS
Charging stations have to be placed at convenient locations that are frequent enough
to cover all common routes of EV users. Yet, increasing the number of charging sta-
tions also increases their capital and operational expenditures. Therefore, while ide-
ally a charging station could be placed at each and every block within a large city, the
investment required would be practically and economically prohibitive. Moreover,
policies and subsidies play an important role in supporting the deployment of charg-
ing infrastructure, as the public and governmental sector will most likely not be able
to keep up with the rapid pace of EV expansion alone. The financial burden of the
EV charging infrastructure development on national budgets may not be attractive to

50,000

45,000

40,000

35,000

30,000

25,000

20,000

15,000

10,000

5,000

0
Year 0 (Initial Cost) Year 1 Year 5 Year 10
Volkswagen e-Golf (with tax credit) Volkswagen e-Golf (no tax credit)
Volkswagen Golf

FIGURE 13.2 Total LCC comparison of electric and gasoline-fueled Volkswagen Golf.
298 Electric Vehicle Integration in a Smart Microgrid Environment

policy and decision-makers. Indeed, the amounts required would better go to other
sectors, which are already in dire need such as education or health. Private sector
participation in EV infrastructure development can then be encouraged through pub-
lic policies, subsidies, and incentives. For example, laws can be put in place to allow
private sector participation through various concessionary or public–private partner-
ship schemes, including [6] the following:

• Build, Own, and Operate (BOO), where a private sector investor finances,
owns, builds, and then operates a project for a specific concession period
(e.g. 10–20 years) while selling the project’s output to the government at
predetermined contract rates (and sometimes to other customers at differ-
ent rates). The investor recovers investment, operation, and maintenance
(O&M) costs from the project’s revenues.
• Build, Operate, and Transfer (BOT), where the private sector investor enters
into a concession with a governmental entity to finance build then operates
a project for a specific period, recovering costs through the project revenue
minus a predetermined percentage to the governmental entity. At the end of
the concession period, the project assets’ ownership and operational respon-
sibility are wholly transferred from the concession entity (special purpose
company) to the governmental entity.
• Build, Own, Operate, and Transfer (BOOT), same as BOT, except that the
private sector investor owns the project during the concession period and
therefore pays no annual percentage to the concessionary governmental
entity until the ownership transfer.
• Build, Lease, and Transfer (BLT), where the private sector investor finances,
owns, and builds the project and then leases it to the governmental entity for
a specific period (in years) in exchange for a periodic lease amount allowing
the investor to partly or fully recover the project investment. At the end of
the leasing period, the project ownership is transferred to the governmental
entity at a predetermined price, which may be zero.

Such schemes have proven to be generally more effective in developing public infra-
structure by ensuring higher quality while reducing the necessary public sector
investment. As private sector players are mainly profit-driven, they tend to seek opti-
mum utilization of their capital investment for profit maximization. Consequently,
they would target to minimize the required capital expenditure and maintain an
acceptable level of service to users while still selling at a profitable rate. In EV charg-
ing station context, this simply means that private sector investors will seek to deploy
the minimum number of charging stations required to effectively cover a given target
service area.
Another hurdle is the profitability of EV charging stations themselves. With the
number of EV users still growing, it might not be financially attractive to deploy and
operate charging stations to serve such a niche market. This hurdle can be largely
overcome by public sector incentives such as targeted subsidies or tax credits which,
when factored into the project’s financials, might turn the investment profitable to the
private sector investors.
Economic Placement of EV Charging Stations 299

Technically, the EV charging stations’ deployment problem can be broken down

into three subproblems, namely:

1. Placement. Selecting the locations of the charging stations in order to mini-

mize their number and optimize their accessibility for EV users.
2. Power supply. Choosing the source of electricity for the charging stations,
be it from the grid or from on-site generation (renewable or conventional),
and whether energy storage should be used.

Clearly, the first subproblem (placement of stations) is the most influential; it affects
in whole or part the other two. The location of a charging station may mandate the
electricity source to be used and, depending on the number of possible users, may
suggest increasing the number of charging points. Based on this, the first subproblem
is tackled in this chapter: the placement problem.

13.3 METHODOLOGIES FOR PLACING CHARGING STATIONS

To illustrate the basic strategy for EV charging station placement, consider the fol-
lowing simple case: an EV user only drives on a single road between two points:
from home to work in the morning and then vice versa in the evening. Where would
a charging station be placed so as to serve her or him best? To answer this question,
we break it down into few simpler questions:

A trivial case would be that the road distance (L) is equal to or less than half of the
range of the fully charged EV (REV ). In this scenario, if the EV user starts the day
with a fully charged EV battery, she or he would be able to drive to work in the morn-
ing then back home in the evening before depleting the EV battery. Consequently, the
suggested solution for this scenario is to place an EV charger at the user’s home. The
EV user would plug the EV into the charger upon arriving in the evening until it fully
charges the next day. We can express this scenario as follows:

2 * L ≤ REV (13.1)

Suppose the road distance is longer than half the EV range but still less than or equal
to the full EV range. In that case, starting the day with a full charge, the EV user
would then drive to work normally in the morning but run out of charge while driv-
ing back in the evening before reaching her or his home. For this scenario, a solution
would be to place one charger at the user’s home and another at work premises. The
EV user would then plug her or his EV into the charger at work upon arriving to
300 Electric Vehicle Integration in a Smart Microgrid Environment

fully charge during the workday and then drive home and plug it again into his home
charger to charge for the next day. This scenario can be expressed as follows:

REV < 2 * L ≤ 2 * REV (13.2)

Interestingly, if we know for a fact that the user can always maintain the EV state
of charge at 50% or more, we could replace the two chargers in this scenario with a
single public fast charger in the middle of the road. That way, the user can charge the
EV halfway at least once throughout the morning and evening trips.
Now consider one last scenario where the road distance is longer than the whole
EV range but less than or equal to its double:

2 * REV < 2 * L ≤ 4 * REV (13.3)

In that case, even if the user maintains a starting state of charge of 100% by using
a home charger, at least two additional chargers would still be needed: one halfway
through the road and another at work premises. Adding even more charging stations
along the road, e.g., two stations on the road (plus those at home and work premises)
would reduce the required starting state of charge to 66.7%. This can be linked to the
concept of range anxiety, which was discussed earlier; adding more charging stations
reduces the starting state of charge to the level required to reach only the first charg-
ing station on the driving route. This would give more confidence to the EV user that
she or he can recharge during his trip and therefore would alleviate range anxiety.
Although considerable research efforts have been directed toward the charging sta-
tion placement problem, only few seem to be practically applicable to real-life situa-
tions, especially in metropolitan settings. Shi, Pan, Wang, and Cai [7] investigated the
minimum EV battery size required to satisfy 100% of the travel distance demands of
taxis and private vehicles. This is done by analyzing the historical travel data of both

FIGURE 13.3 Illustration of the discussed simple case scenarios.

Economic Placement of EV Charging Stations 301

taxi fleets and private vehicles in Beijing, China. The dataset included trip start and end
locations and times and travel distance, as well as total trip travel time. The paper con-
cluded that the battery technology is unlikely to be a major bottleneck to EV adoption.
The dataset analysis in the paper shows that current EV battery ranges (which the paper
estimates to a maximum of 480 km at 0.217 kWh km−1 consumption) are sufficient for
the majority of travel patterns of taxis, as well as private vehicles, which were found to
require even lower battery ranges compared to taxis.
Predicting that EV growth will mainly happen in urban areas where battery range
is not the biggest concern, Cui, Weng, and Tan [9] attempted a cost optimization
approach to the EV charging station placement problem. The paper formulated the
placement problem as a mixed-integer objective function with constraints added one
by one to preserve convexity. The effect of these cost constraints, which included
the cost of the charging station, distribution network expansion, protective device
upgrade, and voltage regulation costs, is presented, and a global minimum of this
objective function is investigated. Finally, the paper conducted a sensitivity analysis
and demonstrated, through numerical results, the trade-off relationships between dif-
ferent factors affecting this optimization problem.
ElBanhawy and Nassar [10] discussed the influence of early adaptors’ perceptions
about EVs (including range anxiety) on the market penetration level and proposed a
movable charging unit (MCU) as a solution to charging points’ placement problem
in metropolitan residential areas. The proposed MCU would be capable of both fast
charging and battery swapping and would move through a hypothetical neighbor-
hood using a shortest-path algorithm to “home-deliver” charging to EV users, espe-
cially those with no domestic charging facilities.
Xiong et al. [12] addressed the charging station placement problem using an
approach that mixes game theory and algorithmic analysis. In their paper, the authors
formulated the placement problem as a charging optimization game, where EV drivers’
individual strategies are to minimize their own charging cost, while their collective
behaviors affect the road network traffic conditions. The paper used a suggested algo-
rithm (called OCEAN-C) to compute the optimal solution to the placement problem
and validated the approach using real dataset from Singapore. Wagner, Götzinger, and
Neumann [13] proposed a placement approach based on proximity to points of interest
such as parks, restaurants, and banks. An “attractiveness” metric is defined for charg-
ing stations, based on the usage frequency of a charging station (as the more a charging
station is frequented, the higher importance it has) as well as how close it is to points
of interest. The paper performs a regression analysis on data collected over a period of
several months for approximately 230 charging stations in Amsterdam, Netherlands.
A location model is then derived, which provides infrastructure planners with opti-
mal locations for new charging stations based on point of interest locations. Beside
Amsterdam, the paper also presented a case study for Brussels, Belgium.
The paper by Franke et al. [14] thoroughly examined the perception of comfort in
EV driving ranges among EV users. A 6-month field study was conducted where forty
EV users were surveyed, with an emphasis on psychological stress-buffering personal-
ity traits and coping mechanisms with regard to range anxiety. The study introduces
the term “safety buffer”, which quantifies the increase in the range needs perceived
by the EV user. More importantly, the paper highlights that the perceived EV range
302 Electric Vehicle Integration in a Smart Microgrid Environment

barriers, including suboptimal range utilization, may be overcome using psychological

interventions including user training and EV interface design. The study also suggests
that a reliable usable EV range may be more important than enhancing the maximum
possible range. In the paper by Yuan et al. [13], the notion that EV drivers tend to have
a higher range anxiety compared to conventional vehicle drivers is highlighted. The
paper conducts a survey on EV drivers in China to assess and quantify their actual
range anxiety level, as well as their resulting behavioral changes. The paper also made
use of the concept of a “safety buffer” introduced by Franke et al. [12], which is defined
as the minimum difference between the EV’s remaining mileage and the trip range.
Accessibility to charging stations, driver experience, and emotional confidence were
found to be significant factors to the perceived range anxiety level. The paper analyzes
the data for Chinese EV drivers based on three trip scenarios: short (10 km), medium
(30 km), and long (60 km) and concludes that the average safety buffer for the three trip
scenarios is approximately 47%, which is quite high. Finally, the authors suggest that
EV human machine interface design should address range anxiety by incorporating
driving style and charging strategy detection and recommendation algorithms.

13.4 ECONOMICS OF CHARGING STATION PLACEMENT

Establishment of public EV charging infrastructure is a costly investment, and the
costs are broadly categorized into capital costs, comprising hardware and instal-
lation costs and operational costs comprising electricity and maintenance costs.
Capital expenditure is usually referred to as CAPEX, while operational expenses are
referred to as OPEX. To discuss the economics of public charging stations’ place-
ment, we shall use a LCC approach, which aims to quantify the expenses throughout
the expected lifetime of operating the charging station. Table 13.1 below lists typical
CAPEX and OPEX cost components related to public charging stations.
IEC 61851 defines four modes of charging, namely [14]

• Mode 1. Domestic slow AC charging, from a regular electric socket cord.

• Mode 2. Domestic AC charging with EV-specific protective device.
• Mode 3. Slow or fast AC charging on a dedicated circuit using specific EV
multipin sockets and EV-specific protective features.
• Mode 4. Fast direct current (DC) charging using a special connector and
special charger technology.

TABLE 13.1
Typical EV Charging Station CAPEX and OPEX Cost
Components
CAPEX OPEX
Charging hardware equipment Electricity supply costs
Material (cabling, mounting, etc.) Maintenance and repair
Installation labor Taxes
Permits and licenses
Economic Placement of EV Charging Stations 303

Similarly, the US Society of Automotive Engineers (now known as SAE International)

defines three levels of charging: level 1 (slow household charging on 110–120 VAC),
level 2 (upgraded household or commercial charging on 208–240 VAC), and fast DC
charging, which is sometimes referred to as level 3, although not fully standardized
(charging on 208–480 VDC). The IEC 62196 also standardizes the EV charging
multipin connectors into three configurations, [15] with type 2 being the most widely
used. We will mainly talk about public charging infrastructure serving users in met-
ropolitan areas. Therefore, the charging stations that we will discuss are mainly level
2 in the US (or IEC 61851 Mode 3 or 4 in Europe).
In order to simplify discussing the economics of charging infrastructure deploy-
ment, few assumptions have to be made. First, it is assumed that only 1 charger is
installed (i.e. single charging port), and therefore, economies of scale are ignored.
Second, the charger’s capacity (i.e. kW rating) is ignored and, for the sake of practi-
cality, shall be assumed not to be a deciding factor in the placement decision. More
conveniently, this could be thought of as if the charger’s rating is large enough that
the charging time required by an EV to fully charge is consequently reasonably short
enough to be done during morning and evening trips in the previously discussed
one-road simple scenario. Another important simplifying assumption is that power
source is readily available at the charging station’s location, and therefore, no utility
upgrades or special power connection provisions are required. One final assumption
is to ignore land ownership or leasing arrangements, that is to assume the charging
station’s installation location is already owned by the infrastructure developer, and
therefore, no leasing, renting, or usufruct fees are paid. In practice, many factors
affect the installation cost of EV charging station, including the following:

• Distance from the power connection point

• Number of charging ports to be installed
• Whether the installation is indoor or outdoor
• Labor cost at the installation site

The analysis presented here uses CAPEX figures estimated by Electric Power
Research Institute (EPRI) for the US in 2013, but the purpose is to illustrate the gen-
eral framework of analyzing the economics of EV charging station deployment [16].
The EPRI study is particularly useful because it provides the total average costs per
charging station as well as a breakdown by labor, materials (including the charger),
permits, and taxes. Figure 13.4 below shows the split of these cost categories in per-
centages of the total.
According to EPRI, the average total CAPEX per 1-port charging station in the
US is approximately 4,484 $. We will therefore take the rounded-up figure of 4,500 $
as the investment required to install one charging station. The OPEX consists mainly
of the cost of electricity consumed by the charging station (i.e., to charge EVs of cus-
tomers) plus any amounts spent on maintenance and repairs. The electricity tariff is
assumed to be 13.2 $cents/kWh, and we will estimate the maintenance and repair to
consume 30% of the required CAPEX each year, i.e., around 1,350 $/year.
One of the first questions the investor would tackle is: How much should we
charge customers for using the EV charging station? The answer to this question
304 Electric Vehicle Integration in a Smart Microgrid Environment

Tax
Permits 3%
5%

Materials
34%
Labor
58%

FIGURE 13.4 Installation cost split of EV charging station installation in the US.

requires a market study to estimate the number of customers that may use the
charging station, and the more thorough this study is conducted, the more accu-
rately the rate can be determined. Estimating the potential number of customers
allows the selling rate to be calculated so as to cover the costs (or “break-even”),
recover the CAPEX amount over a certain number of years, and achieve a certain
profit margin to the investor(s).
This exercise also usually begins with calculations for a single unit, i.e., one
charging station, and once the unit economics are settled upon, the numbers are
simply applied to the required total number of charging stations to be deployed.
Consequently, we shall consider a single charging station, and the calculations are in
Table 13.2 below:
So the charging station revenue needs to equal at least 10,850 $/year to cover
its estimated yearly expenses. The investor would also require recovering the
CAPEX over, say, three years. This means that the revenue needs to increase by
an additional 1,500 $/year (4,500 $/3 years) to become 12,350 $/year. This revenue

TABLE 13.2
Example Economic Calculations for a Single Port Charging Station
Estimated number of users (from 300 users (monthly), with 40 kWh EV battery at an average
market study): charge state of 50%
CAPEX: 4,500 $
Yearly cost of electricity = Yearly electricity demand (kWh) × 13.2 $cents/kWh
= 300 users/month × 12 months × 40 kWh × 50% ×
13.2 $cents/kWh ≈ 9,500 $/year
Maintenance and repair = 30% × 4,500 $ = 1,350 $/year
Total yearly OPEX = 9,500 + 1,350 = 10,850 $/year
Economic Placement of EV Charging Stations 305

figure would then be divided by the estimated yearly demand to yield a selling rate of
17.2 $cents/kWh. Figure 13.5 below shows the cash flow for the first 8 years of opera-
tion (note the break-even point at year 3). After the break-even point, the difference
between revenue and costs is net profit (assuming zero tax).
At this point, it is useful to consider multiple scenarios for the number of custom-
ers, the required CAPEX payback period, or both. Table 13.3 below shows the varia-
tion of the selling rate with varying estimates for the number of customers and the
CAPEX recovery period.

13.5 CASE STUDY: APPLYING AN AGENT-BASED NETWORK

GRAPH PLACEMENT METHOD ON CAIRO, EGYPT
This section will describe an approach to solving the placement problem of EV
charging stations in a large metropolitan area. The method that will be used com-
bines graph theory (GT) with agent-based modeling (ABM) and will be applied to
Greater Cairo area (in Egypt) as a case study.
GT is concerned with the study of the relations and properties of complex sys-
tems by abstracting them into graphs. A graph is a mathematical network structure
comprising a set of nodes (or “vertices”) interconnected by a set of links (also called
“edges”). The links of a graph can be either directed, which signifies a one-way, asym-
metrical relation between the interconnected nodes, or undirected, signifying a two-
way, symmetrical relation. The graph links can also be assigned numerical weights
(or costs) so as to represent the modeled system or network more meaningfully.
ABM is the computational approach of simulating entities (i.e., the agents)
and studying their interactions with each other as well as with the environment.

FIGURE 13.5 Example 8-year cash flow of EV charging station operation.

306 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 13.3
Variation of Selling Rate (in $cents/kWh)
with Users and Recovery Period
CAPEX Recovery Period
(in years)
3 4 5 6
No. of 300 17.2 16.6 16.3 16.1
monthly 450 15.8 15.5 15.3 15.1
users 600 15.2 14.9 14.8 14.7
750 14.8 14.6 14.5 14.4

An individual agent is an autonomous, programmatic entity defined using specified

attributes, which influence its behavior. The agent attributes typically incorporate
random elements to simulate real-life complexity.
Although each of the two methods was previously used separately to tackle the
placement problem of EV charging stations, both approaches are combined here into
a hybrid distance- and agent-based one. A research paper by Sun, Zhao, He, and Li
[17] identifies the lack of charging stations as the main obstacle facing EV adoption
and presents a methodology to deploy charging stations within a road network that is
modeled as a directed graph. The paper proposed a statistical model to describe the
spatiotemporal characteristics of the charging demand of EVs. Charging stations are
then placed so as to maximize the number of EV trips passing at least one charging
station, while not exceeding the electricity supply grid capacity.
Sweda and Klabjan [18] proposed an agent-based decision support system for
deploying new charging infrastructure. The proposed model used road network data
to identify EV ownership and driving patterns in Chicago metropolitan area. One
main aim of the paper is to quantify the market demand for and transition to EV
usage in order to attract investors to build charging stations. Agents are assigned
attributes including income, preferred vehicle class, and range anxiety level, and
their EV purchasing behavior is simulated based on their interaction. The results
of the model compare a base case (20 existing charging stations) to two proposed
charging station deployment scenarios, each with 70 additional charging stations.
The paper also illustrates how projected EV adoption is affected by gasoline price
adjustment, with higher gasoline prices accelerating EV adoption trend.
For the proposed method, the Python programming language was used to implement
the models of both the graph and the agents. Python is a high-level, object-oriented
programming language widely used in scientific computing. The road network of the
Greater Cairo area in Egypt is modeled using Python programming language as an
undirected network graph, with its link weights representing the approximate driving
distances between different network nodes. As such, nodes simply represent crossroads
in the road networks, and we can imagine users navigating between starting and ending
nodes through this road network. The road network graph extends between downtown
Cairo and eight of its surrounding edge cities (see Figure 13.6 below).
Economic Placement of EV Charging Stations 307

FIGURE 13.6 Network graph representation of Greater Cairo metropolitan area (including
edge cities).

The graph includes 30 nodes and 53 links, and each link is assigned a cost attri-
bute equal to its approximate real driving distance. Most graph nodes have more than
one link to surrounding nodes, which results in multiple alternate routes between any
two given nodes. Table 13.4 below lists the eight edge cities included in the graph
model.
The drivers of EVs are modeled as agents with defining attributes including
home node, destination node, EV charge state, and range anxiety level. The driver
agents are implemented as a Python class objects, and the object’s home and desti-
nation attributes are randomly selected from the graph network model. Similarly,
charge state and anxiety level are initialized to a random percentage between
0 and 100%. Figure 13.7 below presents the attributes and actions of the driver
agent.

TABLE 13.4
The 8 Edge Cities Surrounding Cairo
Approximate Distance to
City Downtown Cairo (km)
10th of Ramadan city 63
5th Settlement 27
6th of October city 46
Badr city 59
New Administrative Capital city 75
Obour city 40
Sheikh Zayed city 32
Shorouk city 47
308 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 13.7 Class representation of the driver agent.

The agent object (named DriverAgent) has the following attributes and methods:

hom e: The graph node from where the driver agent starts the trip.
destination : The graph node to which the driver agent makes the trip.
curr _ location : The current location of the driver agent within the graph network.
ch arg e _ level: The level of charge of the EV, defined as a percentage between 0% and 100%.
anxiety _ level : The range anxiety level of the driver agent and which affects the driving range
perceived as possible. It is defined as a percentage between 0% and 100%.
range _ attainable: The actual attainable driving range based on the charge level.
range _ perceived : The perceived driving range based on the actual attainable range and the range
anxiety level of the agent
range _ covered : The actual range driven by the agent so far.
panicking : A flag triggered when range _ covered exceeds range _ perceived.
arrived : A flag triggered when the agent reaches destination.
drive(): This is the main action of the driver agent, which advances him through the path
between home and destination.

The range _ attainable is calculated as follows:

range _ attainable = charge _ level  (%) * ev _ battery _ capacity ( kWh ) *

(13.4)
ev _ mileage ( km / kWh )
Economic Placement of EV Charging Stations 309

And the range _ perceived is calculated as follows:

range _ perceived = range _ attainable ( km ) * (1 − anxiety _ level (%)) (13.5)

The main concept of the proposed approach is to randomly instantiate a large number
of EV driver agents, let them drive within the network graph through the shortest path
while tracking their charge state and anxiety level, and then register the locations where
they actually run out of charge before reaching their destinations. These locations are
then surveyed to identify and rank them based on their frequency of “out-of-charge”
incidences (OCIs). The rationale is that the more EVs run out of charge at or around a
specific location, the more suitable it is for placing an EV charging station. There are few
assumptions made in order to simplify the model, and they are as follows:
1. The EV drivers use compact to subcompact cars, which are less expensive
and more suitable for urban environments. Consequently, the battery capac-
ity of the EVs is assumed to be 24 kWh.
2. The mileage of the EVs is assumed at an average of 6.0 km/kWh, which is
obtained from Environmental Protection Agency MPG data based on city
cycle1.
3. The only criterion for routing through the graph network is the driving dis-
tance, and therefore, traffic congestion and any associated time impact are
ignored.
4. Driver agents can start their trip from any node on the graph network,
including nodes that represent real-life road intersections.

The driver agents’ trip routes are determined using a custom implementation of the
Dijkstra shortest path algorithm. Dijkstra’s algorithm is an established algorithm
used to find the shortest path between two given graph nodes by performing a “best-
first” search. For each driver agent, the shortest path of the current trip is computed
using the algorithm, and then the agent is iteratively simulated to drive through it
until the agent either reaches the trip destination or runs out of charge. If the agent
runs out of charge en route to the destination, the location is saved to an array. The
array of possible locations is then rearranged so as to remove duplicates and count
their frequency of occurrences, and finally, the locations and their frequencies are
exported to a comma-separated file (CSV) for analysis.
For the simulation, 5,000 driver agents were instantiated. The exact number of driver
agents did seem to affect the statistical distribution of the OCIs over the graph links, but
it was found that this effect diminishes around 2,000–2,500 agents. Consequently, the
links with the most frequent OCIs tend to remain the same above this range, and the
5,000 figure was chosen as both large enough to minimize the statistical distortion and
small enough to be simulated within a practically short code running time.
The output CSV file lists the graph links and the number of OCIs, and the list was
reordered in a descending no. of OCI frequencies to obtain a “priority” list of links
where an EV charging station was recommended. Table 13.5 below lists the top 10
links prioritized by the algorithm for EV charging station placement.

1 https://www.fueleconomy.gov/feg/download.shtml
310 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 13.5
Links Prioritized for EV Charging Station Deployment (Results of Simulation)
Link
OCI Frequency Cumulative OCI
From To (%) Frequency (%)
Intersection – Ring Road Intersection – Suez 9.5% 9.5%
– Suez Road Road – Shorouk City
Entrance
Intersection – Rode El Zamalek 6.6% 16.1%
Farag Axis – Orouba Road
Intersection – Suez Road Intersection – Suez 6.1% 22.2%
– Shorouk City Entrance Road – Badr City Entrance
Intersection – Ring Road Intersection – Ring 5.5% 27.7%
– Mosheer Axis Road – Nile Corniche
(Maadi)
Intersection – 26th July Intersection – Ring 5.3% 33.0%
Axis – Cairo Alex Desert Road – 26th July Axis
Road
Intersection – Ring Road Intersection – Rod El Farag 4.7% 37.7%
– Ismailiya Road Axis – Orouba Road
Intersection – Suez Road New Administrative Capital 4.6% 42.2%
– Badr City Entrance City
Intersection – Ring Road Intersection – Ring 4.1% 46.3%
– Al-Bahr Al-Aazam Road – Wahat Link
10th of Ramadan City Intersection – Ring 4.0% 50.4%
Road – Ismailiya Road
Helwan Maadi 3.3% 53.7%

Those familiar with Cairo will quickly note that the first graph network link pri-
oritized by the algorithm for charging station placement is in a suburban area: the
link between the ring road/Suez road intersection and the entrance of Shorouk City.
This result is not intuitive, as it was expected that the most frequented graph links
would be more central to Greater Cairo area that is within downtown area. On the
other hand, the last link in the list (the link between Maadi and Helwan) is in line
with the intuition as the main road connection between Helwan and Greater Cairo
area passes through Maadi (south of Nile corniche road).
The total cumulative frequency of OCIs for the 10 listed links is 53.7%. In other
words, by locating the charging stations at only 10 locations out of 53 possible ones,
more than 50% of the whole intended area was covered. Therefore, this algorithm
can be used to maximize the charging infrastructure coverage while optimizing
the number of stations required. This can result in significant saving in the initial
CAPEX for the investor.
While the proposed method was applied to Greater Cairo area, it can nonetheless
be adapted to any metropolitan area. Furthermore, the algorithm could definitely be
further elaborated to account for more real-life effects such as
Economic Placement of EV Charging Stations 311

• Road traffic congestion-related route choice

• Topographical profile to exploit EV regenerative braking
• Prioritizing locations based on power supply availability
• Exploiting distributed renewable energy generation sources for powering

REFERENCES

10. Y. Xiong, J. Gan, B. An, C. Miao, and A. L. C. Bazzan, “Optimal electric vehicle charg-
ing station placement,” in Proceedings of the 24th International Joint Conference on
Artificial Intelligence (IJCAI), 2015.
11. S. Wagner, M. Götzinger, and D. Neumann, “Optimal location of charging stations in
smart cities: A point of interest based approach,” in 34th International Conference on
Information Systems, Milan, 2013.

14. International Electrotechnical Commission (IEC), “IEC 61851-1: Electric vehicle con-
ductive charging system - Part 1: General requirements,” 2017.

19. D. Hall and N. Lutsey, “Emerging best practices for electric vehicle charging infrastruc-
ture,” The International Council on Clean Transportation, 2017.
 14 Environmental Impact
of the Recycling and
Disposal of EV Batteries
Zeeshan Ahmad Arfeen
University Technology Malaysia
The Islamia University of Bahawalpur (IUB)

Rabia Hassan
Institute of Business Management Sindh

Mehreen Kausar Azam

Institute of Business Management Sindh
N.E.D University of Engineering & Technology

Md Pauzi Abdullah
University Technology Malaysia

CONTENTS
14.1 Introduction .................................................................................................. 314
14.1.1 Battery Repurposing and Clearance for Sustainable Society ........... 314
14.2 Delaying Recycling through Repurposing ................................................... 316
14.2.1 Repurposing ...................................................................................... 317
14.3 Economic Aspects ........................................................................................ 319
14.3.1 Identifying Domestic Demand ......................................................... 319
14.3.2 Identifying Industrial Demand ......................................................... 319
14.4 Standards for Reusing EV Batteries ............................................................. 320
14.5 Environmental Impacts of EV Batteries | EVBs .......................................... 320
14.5.1 Raw Material Manufacturing Effects ............................................... 321
14.5.2 Battery Manufacturing Effects ......................................................... 321
14.5.3 Thermal Gas Emission ..................................................................... 321
14.5.4 Chemical Hazards............................................................................. 321
14.6 Battery Dismantling and Handling Health Hazards .................................... 322
14.6.1 Lithium-Ion Battery Landfill ............................................................ 323
14.6.2 Impact of Recycling on the Environment ......................................... 323
14.6.3 Recycling of EV Batteries ................................................................ 324

313
314 Electric Vehicle Integration in a Smart Microgrid Environment

14.7 Environmental Aspects of Reuse ............................................................... 324

14.8 Environmental Aspects of Recycling ......................................................... 325
14.9 Recycling .................................................................................................... 326
14.9.1 Recycling Methods ....................................................................... 326
14.9.2 Mechanical Procedure | MP ......................................................... 327
14.9.3 Pyro Metallurgical Procedure | PM .............................................. 327
14.9.4 Hydrometallurgical Procedure | HP ............................................. 327
14.9.5 Direct Recycling Procedure | DRP ............................................... 328
14.10 Best Practices of Lithium-Ion Battery Recycling ...................................... 329
14.10.1 Umicore Company ........................................................................ 329
14.10.2 Retrieve Technologies ................................................................... 330
14.10.3 Onto Technology........................................................................... 330
14.11 Safety Indicators......................................................................................... 330
14.12 Dismantling and Storage ............................................................................ 330
14.12.1 Reorganizing and Screening......................................................... 330
14.13 Technological Initiatives ............................................................................ 331
14.14 Conclusion .................................................................................................. 331
14.15 Recommendations and Future Directions .................................................. 332
References .............................................................................................................. 332

14.1 INTRODUCTION
Electric vehicles (EVs) have substantial advantages contrasted with internal combus-
tion-based engine vehicles due to their eco-friendly nature. EVs are preferred as they are
more economical to maintain and run as well as they add more value for life because of
their negligible air contamination and less reliance on oil as well as since no poisonous
substances emerge which can destroy the ozone layer [1]. Furthermore, electric engines
and batteries add to the vitality of electric vehicle. Many studies observed that the mate-
rials utilized in batteries are perilous for Earth. The mining and preparing of items
such as lithium, copper, and nickel require much exuberance, and this can lead to the
emission of harmful substances. In the chapter, it is comprehensively discussed how the
second-life batteries, which already completed their useful life, could be reconditioned.
Furthermore, if they need to be disposed of what are the best ways of doing so to opt.

14.1.1 bAttery repurposing And cleArAnce For sustAinAble society

We live in a period of sustainability, which underpins the crucial endeavoring to
improve the natural wellbeing and personal satisfaction for the network. As reusing
the paper, cardboard boxes, and utilize containers of spaghetti sauce, however, what
is demanded by EV battery after it has lost its dynamism? So analysts of the world
are dealing with the territories where they can create ways on which these EV bat-
teries which are disposed of can be reused. EV sales are skyrocketing. Figure 14.1
below shows the best five EV automotive organizations for the year 2019.
As EV sales increase, the number of batteries that will get access for a second life
outside of the vehicle increases. EV batteries are a gauge to surpass what might be
compared to about 3.4 million packs by 2025, contrasted and around 55,000 of every
Recycling and Disposal of EV Batteries 315

FIGURE 14.1 Sale of top 5 EV manufacturing companies in 2019 [2].

2018 as expressed in Bloomberg Businessweek. Does the inquiry rise that where will
these batteries land? Various roads have been inquired about for the repurposing
of the EV battery, and a portion of the major is examined in detail in this chapter.
Figure 14.2 signifies that the battery’s production volume increases with a fall of net
cost yearly.
There are several manufacturers of EVs. A compact summary of the types of plug-
in EVs is given in Table 14.1.

FIGURE 14.2 Sales of EV increases with falling battery cost [3].

316 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 14.1
Types of Plug-in EVs [3–5]
Types of EVs EV Models Prominent Features
Plugin EV Toyota Prius, Ford It is powered by dual-fuel similar to Hybrid EV with
(gasoline + battery) Fusion, Audi A3 another option of plug-in charging with any electrical
E-Tron; BMW outlet. The energy stored in battery packs supplied power
530Le to an electric motor or by an internal combustion engine
that runs on conventional fuel or regenerative braking.
It employs both in the charge sustaining mode as well as
a charge depleting mode of operation.
Hybrid, EV Honda Civic Hybrid, HEVs cannot be charged from outside outlets like the
(gasoline + battery) Ford Fusion power grid. The battery bank charged by IC engine
Hybrid, Toyota default built-in and by regenerative braking, i.e., by
Prius Hybrid, transforming its kinematic energy into chemical energy
LEXUX NX that deposited in the banks of the battery.
Extended range Chevrolet, Cadillac This is an example of a merger between PHEV and BEV,
EV/PEV which has a more efficient system for fuel and fewer
emissions of toxic nature
Battery, EV Nissan Leaf, BMW Zero-emission is achieved in this type by the merger of
(Battery) i-3, Mitsubishi I batteries, which can be recharged by an electrical outlet.
MiEV, Tesla It works in the charge depleting states of operation and
Roadster Model S, requires enough power battery packs and high power. As
Ford Focus Electric, the fixed cost is high which does not break even quickly
Benz EQC because of short mileage, the charging duration also long
makes this option less feasible.
Fuel cell EV, Nissan Motor It accumulates hydrogen as a fuel that is produced from
(FC +SC/battery) natural gas. The energy, in this case hydrogen, provisions
of a fuel cell within combustions with the oxygen from
around produces a flow of electrons, which is due to
electrolysis in the reverse direction, thus resulting in heat
and water as byproducts of this chemical reaction

14.2 DELAYING RECYCLING THROUGH REPURPOSING

Three kinds of procedures are right now accessible for reusing EV batteries. In
the first place, the batteries can experience a refining procedure, accessible on
an enormous scale for different kinds of batteries, including lithium-particle and
nickel-metal batteries. These batteries are encouraged into the smelter to recu-
perate significant metals. Remaining materials, for example, lithium, are lost to
slag, which can be used for the manufacturing, not accessible for a wide range of
batteries, including isolating parts through different physical and synthetic proce-
dures, including hydrometallurgical advances, and afterward recouping any bat-
tery-grade materials directly. Finally, transitional procedures include extricating
exclusively perilous battery segments toward the end of the battery life, which is
Recycling and Disposal of EV Batteries 317

TABLE 14.2
Pros and Cons of Reconditioned EV Batteries
Sr. No Pros Cons
1 Making their repurposing for a second and even The reusing lithium-particle batteries involve
third life much progressively significant to expense and potential waste.
expand their monetary and ecological
incentive before reusing.
2 The batteries can be used for stationary storage, New batteries are used for stationary storage
which is monetary favorable, thus saving on at the moment which has a high density and
the initial amount of the product. This battery has a bigger life than the batteries that have
property is another factor that will be making served their first life. At the moment,
EVs popular. companies are not issuing warranty to
batteries that are used.
3 The stability of the grid and the ability of the The size and shape, as well as the
grid to integrate renewable energy are aided performance of the EV batteries, differ. The
by the secondary storage. batteries perform under different climatic
conditions and stressors and for different
car models.
4 As per the costing, this is the most economical When the battery packs are of the
system where the battery packs are of the nonuniform constitution, the output has to
uniform constitution. be regulated by using software that is
noneconomical. Research is being
conducted to attain performance by
incorporating old batteries with new ones,
but it will take time and resources to reach
the desired results.

useful in limiting the measure of dangerous substances that advance into the earth.
Table  14.2 gives a very precise view of the merits and demerits, which can be
attained by reusing the EV batteries [6].
Figure 14.3 sheds light on the EV batteries currently used and how they can be
further utilized in second life.

14.2.1 repurposing
Synthetic batteries in EVs give a quick and enormous force supply. Most flow module
EVs – characterized as vehicles that attach into the lattice for a few or the entirety
of their capacity – use lithium-particle batteries [7]. These batteries, though in vari-
ous configurations, normally utilized inconvenient customer hardware, for example,
phones and PCs. In contrast, their higher vitality per unit mass is comparative with
other electrical vitality stockpiling frameworks. Nevertheless, the specific science of
vehicle batteries regularly contrasts from buyer hardware, batteries, just as from one
another relying upon the automaker. In general, batteries in vehicles have a higher
all-out force limit and size.
318 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 14.3 Possible vehicle battery life [2].

Automakers lean toward lithium-particle batteries since they convey predominant

execution in both force and vitality thickness, permitting them to accomplish a high
weight-to-execution proportion. Moreover, most segments of lithium-particle batter-
ies can be recycled. Most of the industries are using LiBs because of their economic
and ecological perspectives; re-using the LiB batteries which are no longer to be used
in their prime application field is preferred [8]. Participants at the assembling noticed
that expanded assembling and economies of scale have discounted costs and broad-
ened their helpful life cycle, with an anticipated value diminishing around seven-to-
eight percent for each year. Table 14.3 depicts the mechanical design of the battery
for different brands of EVs.
Rather than reusing them promptly, the huge number of batteries that will be
leaving EVs in the coming years could be repurposed, prompting a surge of reason-
able batteries that can give vitality stockpiling administrations to clients, utilities,
and matrix administrators. Since second-life batteries will hold critical limits, they
might be appropriate for different clients and lattice applications, especially when-
ever amassed for mass vitality stockpiling [10,11].

TABLE 14.3
Common Zero-Emission Vehicle Battery Packs [9]
Battery Types Cell Design Company Cells/Modules Energy in kWh
Tesla model S Cylindrical Panasonic 7,104 /16 85-large pack
Nissan leaf Pouch NEC 192/48 24
Chevy volt Prismatic LG Chem 288/9 16.5
Recycling and Disposal of EV Batteries 319

14.3 ECONOMIC ASPECTS
Electricity is usually generated by nuclear or hydroelectric means in developing coun-
tries, whereas developing and underdeveloped countries are still stuck with hydro-
electric resources. A country’s demand for electricity is never-ending and increases
every day as population and industry are on the rise, and most of the appliances use
electrical energy to operate. As we enter the digital age, the electricity demand has
increased many folds. A nuclear supply can only be operated on a level load, as it
is not feasible to turn down or increase the level easily. To cater to these demands,
nonrenewable energy sources need to be used. During the peak energy situation, the
produced energy could be stored somewhere else. Thus, at this point, reusing the EV
batteries is not a choice but a necessity.
Reused batteries for EVs, HEVs, and PHEVs present a superb, financially per-
ceptive alternative for vitality stockpiling applications. The organizations involved
reusing lithium-particle batteries to form a circular economy are currently working
with battery creators to embrace effectively disassembled item structures and are
preparing them to take up the new green science forms for the green recuperation of
all esteemed battery segments.
From a business stance, the risk is marginalized for investments by putting the
resources into an enormous number of little applications than few huge applications
following the age-old saying of “Don’t put all your eggs in one basket”. Smaller
applications can incorporate various private customers and a wide assortment of
business clients, including telecommunication organizations requiring 5–10 EV bat-
tery packs each, light business structures requiring 10–15 packs each, and nourish-
ment dissemination focuses requiring 30–40 packs each. The batteries can be used
in several applications. A few of them are discussed below.

14.3.1 identiFying domestic demAnd

Energy storage systems (ESSs) are capturing the market by storm. It works by reus-
ing the stationary Li-ion EV batteries as energy storage devices. It helps to counter
the effects of supply and demand gaps. It can be used to power households and indus-
tries by improving the economy and decreases greenhouse gas (GHG) emissions.
Now ESS is an established system that is controlled centrally. The demand side
used it for deployment at the microgrid side. The system is currently catering to the
efficient usage of renewable energy resources; other resources such as power demand
shortfalls and standby operator as the weather conditions change, and the usage of
peak demand rate factors and climate changes. However, the efficiencies are not up
to the standard, and the high cost of installations makes the system lagging and gives
room for further research.

14.3.2 identiFying industriAl demAnd

The impact of solutions provided by ESS in the areas of the smart grid is quite visible.
The utility grid problems can be stabilized by using ESSs. The ESS optimizes the
quality of the grid as it maintains the power as constant. Renewable energy sources
320 Electric Vehicle Integration in a Smart Microgrid Environment

are used to balance the load curve, and the frequency is controlled. The capability of
the transmitting lines can be enhanced, as well as the voltage fluctuations are miti-
gated by using them [4].
The industry needs systems that provide capacity with a high rating, and the elec-
tric wires that are used should be of high grade as well to handle the rise in voltage
and current. This is provided by centralizing power generation systems.

14.4 STANDARDS FOR REUSING EV BATTERIES

The organization that is currently working for the EV is known as EVs and the
Environment Informal Working Group (EVE IWG) which was formed in March 2012.
This organization is currently working on environmental issues related to EVs [12].
A new guideline in China presently considers EV creators answerable for the
recuperation of batteries, expecting them to set up reusing channels and adminis-
tration outlets where used batteries can be gathered, put away, and moved to reuse
organizations. Before the finish of Feb 2019, 393 carmakers, 44 rejected vehicle
destroying ventures, 37-course use endeavors, and 42 reusing undertakings had just
joined the new discernibility stage to follow starting point and proprietors of dis-
posed of batteries [13].
Worldwide battery reusing prerequisites are either missing totally or where they
exist; they vary a lot, as along with the profundity of inclusion. This presents another
interesting area of research on a worldwide premise because of the unpredictable
idea of practices and mentalities toward reusing EV batteries around the world. What
is prescribed is that assets be assigned to assess the benefit of creating producing for
recyclability necessities [12].

14.5 ENVIRONMENTAL IMPACTS OF EV BATTERIES | EVBs

The design of LiBs depends on the selection of active elements to be used for cath-
ode, whereas these materials have a substantial impact on the health of Homo sapi-
ens in the surrounding environment. These impacts are due to the design procedure
of materials, the use of energy, and the way they are extracted from the surface of the
earth. The elements which are needed for the manufacturing of LiBs contribute more
toward metal degradation and GHGs. An Environmental Protection Agency (EPA)
study found that the batteries which used aluminum have higher rates of contributing
to ozone depletion [14]. Environmental risks arising from lithium-ion batteries must
be addressed during execution of recycling and waste strategies. Lithium-ion batter-
ies contain a high proportion of hazardous heavy metals that are harmful to the envi-
ronment as well as for the human body [15]. In 2005, the used LiBs were gathered,
and the result showed that from 4,000 tons of LiBs, 1,100 tons of heavy metals and
around and above 200 tons of toxic materials were produced; toxic materials in lith-
ium-ion batteries include nickel, copper, and organic and lead chemicals [16]. LiBs
are categorized as class-nine dangerous materials. As LiBs are hazardous, especially
the possibility of rapid discharge, in this case, the batteries generate avalanche cur-
rents, which cause overheating and risk of explosion. Therefore, lithium-ion batteries
might be very hazardous when transported, particularly by air flights [17].
Recycling and Disposal of EV Batteries 321

The energy demand for the manufacturing of LiBs has estimated at approximately
90 MJ kg−1 and 12.5 kg of carbon dioxide per kg. The main contribution toward
the health and environmental effects starts with the development of raw materials
accompanied by battery manufacturing, supply, uses, services, recycling, and waste
management [18].

14.5.1 rAw mAteriAl mAnuFActuring eFFects

Extraction, casting procedure, smelting, leaching, and filtering are the main steps
used to convert the metal and its elements in the particular material used for the
batteries. In the production of LiBs for EVs, the aluminum alone produces 2–3 kg of
CO2 equivalent/kg of battery. The manufacturing of electrode like lithium manga-
nese oxide (LiMn2O4) generates about 800–1,000 kg of carbon dioxide equivalent/
kg of battery. However, the manufacturing of electrolytes such as dimethyl carbonate
and lithium hexafluoro phosphate (LiPF6) generates 100–500 kg of CO2 equivalent/
kg of battery. The manufacturing of such raw materials has high threats to the envi-
ronment [19].

14.5.2 bAttery mAnuFActuring eFFects

Hazardous air emissions, solid waste, and water pollution can generate damage
to health and the environment during the manufacturing of LiBs. The automotive
industry has surged the demand for LiB, with the dire negligence of GHG (GHG)
emissions. It is reported that three categories of cathode materials are mostly used
in the manufacturing of LiBs that are mainly responsible for GHG emission. It is
reported in Ref. [20] that the manufacturing of 28 kWh battery, the cathode materi-
als of cobalt oxide, lithium nickel manganese, lithium-ion manganese oxide, and
lithium iron phosphate produced GHG emissions approximately 2,912 Kg.CO 2 .eq,
2,705 Kg.CO 2 .eq and 3,061 Kg.CO 2 .eq, respectively.

14.5.3 thermAl gAs emission

The major thermal run-away effects are gas and heat emissions, which are highly
flammable. The structure of the batteries and cells typically involve protections to
release the gas without generating a risk of cells or batteries exploding. Likewise,
nonflammable materials like plastics are also avoided to prevent the additional par-
ticipation of plastic combustion to the generated heat. Table 14.4 shows the emission
of gases throughout the lithium-ion battery thermal lifecycle.

14.5.4 chemicAl hAzArds

Lithium-ion battery leaking is equal to hydrofluoric acid production. Since the elec-
trolyte is fluid, there is a possibility that this fluid may be leak inside the battery and
make contact with water and air. Two chemical reactions may cause the hydrofluoric
acid production, the electrolyte ions of PF6− in water presence and that PF6− ion com-
bustion. PF6− ion hydrolysis takes place in the existence of water in basic and acidic
322 Electric Vehicle Integration in a Smart Microgrid Environment

TABLE 14.4
Lithium-Ion Battery Thermal Run-Away Emission of Gases [10]
Compound Concentration (Percentage)
Carbon monoxide 40
Hydrogen 30
Carbon dioxide 20
Hydrocarbons 7
Hydrogen fluoride Less than 3

(PH level between 1 and 12) medium, whereas hydrolysis production is not favor-
able. Through this reaction, hydrofluoric acid is produced. The hydrofluoric acid may
cause serious chemical accidents when it interacts with the eyes or skin.

14.6 BATTERY DISMANTLING AND

HANDLING HEALTH HAZARDS
The batteries of EVs are available in a wide variety in terms of chemistry and forms.
To recycle batteries of an EV, the first important thing is sorting and classifying the
chemistry of the battery through inspection, and after that forwarding the batteries
for dismantling safely toward workbenches. Trained technicians dismantle the packs
of batteries to the module or cell, circuitry, wiring, and pieces of assembly segregated
from individual cells. Disassembling through the manual process is performed by
manual and power equipment as cells and packs of the battery are constructed with
complexity since it is arduous to devise a cost-effective end-of-life (EOL) process
plan for EV batteries. Safety training is vital for the personnel who are responsible
for the dismantling of these items, and the instructions for dismantling should be
given by the manufacturers.
Lithium-ion battery disassembly is a complicated, costly, and lengthy process
because of the extensive range of battery composition, components, and potential
hazards due to high voltage and chemical products in cells of the battery. All of these
elements must be taken into consideration while planning the dismantling process
and appropriate workstations. A lithium-ion battery includes toxic chemicals and
by-products, which are either destructive, or by the reaction with other materials,
they produce toxic, heat-generating, or flammable chemicals. The details of the main
components, material incompatibility, and their health effects are summarized in
Table 14.5 below.
The component of the cathode has a significant risk of toxicity. Therefore, the
probability of health effects is termed as low as the elements are solids, and a high
temperature is used to melt them. If these elements are disclosed, they will involve
the breaking of a battery with the chemical component separation or vaporization.
In a battery of lithium-ion, the cathode and anode are covered by toxic substances
such as the most hazardous chemicals like solid electrolyte interface layer, which is
coated on anode and cathode, electrolyte salt, carbon anode, and electrolyte solvent.
Recycling and Disposal of EV Batteries 323

TABLE 14.5
Component Details of Lithium-Ion Battery [21]
Effects on Noninflammable/
Components Examples Health Incompatibility Flammable
Cathode Lithium cobalt Respiratory, skin, Nil, but elude of Noninflammable
(negative) (LiCoO2); eye, and extreme
Lithium manganate gastrointestinal temperature and
(LiMn2O4); irritant; possible fire
Lithium phosphate carcinogens
(LiFePO4)
Anode Graphite (C) In solid form no Nil, but elude of Noninflammable
(Positive) effect (evade extreme
dust) temperature and
fire
Salt electrolyte Lithium hexafluoro Causes Agents of oxidizing; Noninflammable
phosphate (LiPF6) respiratory, eye, Water; strong acids
skin and
gastrointestinal
Solvent EC; DMC; PC; Respiratory, skin, Agents of oxidizing; Flammable
electrolyte DEC eye, and alkalis and acids;
gastrointestinal
irritant;
Separator Polyethylene; Nil Nil, but elude of Noncombustible
polypropylene extreme
temperature and
fire

14.6.1 lithium-ion bAttery lAndFill

The reuse of LiBs is done in fewer countries, while in other countries, LiBs were
landfill or thrown away. Discarded batteries have eventually reached landfills and
saturated, and this may also harm the environment [22]. GHG emissions from
organic components like electrolyte solvent, plastic, and paper are produced during
landfill processing. Besides, heavy metals can be leached to the atmosphere and can
enter water or soil based on the landfill state. Landfilling possesses more danger to
the environment in comparison to the recycling of the batteries.

14.6.2 impAct oF recycling on the environment

Growing waste required more land space for disposal, as these batteries contain harmful
chemicals, so ultimately, these chemicals return to the environment. However, with the
comparison to other batteries, the LiB contains small toxic materials, so recycling of
batteries may have a less impact on the environment. Particularly by recycling lithium-
ion batteries, the considerable reduction in energy consumption, GHG emission, and
the use of natural resources are found in comparison to the manufacturing of new LiBs.
324 Electric Vehicle Integration in a Smart Microgrid Environment

Additionally, by recycling, we can retrieve lithium and other materials. In particu-

lar, the use of lithium is on the rise due to the increasing need for LiBs, which are
used in EVs. Around 25% of the total amount of lithium is used by battery manu-
facturing industries for EVs, and this number will surge exponentially shortly [23].

14.6.3 recycling oF ev bAtteries

The demand for EVs is rising and will be increased in the coming years. It is esti-
mated that post-vehicle battery pack application will be increased from 1.4 to 6.8
million by 2035 [24]. LiBs are a suitable choice for EVs; secondly, the disposal of
LIBs is vital to conserve the environment and health of Homo sapiens. Regrettably,
the LiBs’ recycling percentage is only 3%, whereas its revival is negligible. It is
estimated that if the rate of recycling will not increase, the requirement of lithium
transcends by 2023 [25]. Figure 14.4 summarizes that after the manufacturing of
lithium-ion battery, its life is approximately 8–10 years. When the battery completes
its life, it should be sent toward service and retail centers where these batteries can be
repaired. If the product has the end of life, then it should be sent toward the recycling
process from where raw materials can be obtained to manufacturing new ones [26].

14.7 ENVIRONMENTAL ASPECTS OF REUSE

Assuming that the second life of batteries of EV may reduce the strain of producing
a new battery or eventually reduce the usage of resources, to adequately evaluate the
environmental aspects of using secondary batteries, the environmental impacts of
EV batteries should be empirically measured over the entire life cycle. Contemplating

FIGURE 14.4 The lifecycle of a lithium-ion battery.

Recycling and Disposal of EV Batteries 325

the reuse of nearing retirement of EV batteries, the complete environmental effect

of the production, primary reuse, and recycling of EV batteries should be dealt with
consideration in the analysis of the life of the batteries. Reusing an EV retired lith-
ium-ion battery in a stationary system might reduce global warming potential (GWP)
by 15% when compared to nonstationary use. It has been reported that LiBs’ refur-
bishment may reduce GWP effects by 12%–46% compared with the similar battery
functionality of Pb acid [27]. The production phase of the battery pack produced the
majority of greenhouse gas emissions by approximately 40% when compared with
both phases of usage and re-usage (31% and 26%), thus contributing to greenhouse
gas emissions significantly [28].

14.8 ENVIRONMENTAL ASPECTS OF RECYCLING

When EV batteries are decommissioned, the overall impact on the environment
must be evaluated while considering the processes of battery recycling, which
include collection, transport, dismantling, and recycling. Transporting lithium-ion
batteries to recyclers has a slight contribution to the environment as compared to
manufacturing and usage phases. However, there is a need for adequate transport
to minimize or exclude the environmental effects of the process. Research has
shown that truck and rail transport is capable of reducing greenhouse gas emis-
sions related to transportation by about 23%–45%. Besides, they combined geo-
spatial models and life cycle analysis to recognize the anticipated infrastructure
needs, categorize emissions, and material flows and evaluate human health risk
from the lithium-ion battery recycling system in California [29]. These findings
indicate that developing of facilities more than two for dismantling would prob-
ably reduce investment returns. Optimum recycling technologies and multimodal
supply chain collection and transport will help in dealing with the economic and
environmental challenges of LiB production.
Pyrometallurgical methods of recycling can minimize greenhouse gas emis-
sions and primary energy utilization by 23% and 6%–56%, respectively, com-
pared to virgin production [29]. Hydrometallurgy recycling provides enormous
electrical energy savings and slight air pollution; furthermore, pyrometallurgi-
cal recycling has significant benefits concerning water utilization. The significant
environmental challenges regarding the processes of recycling are waste mate-
rial landfills, plastics incineration, and electricity consumption, specifically in the
procedures of smelting, which releases a lot of energy. The hydrometallurgical
processes’ major impact on climate change is landfill, while plastics incineration
is the major impact of pyrometallurgical processes. Mechanical and hydrometal-
lurgical techniques may retrieve more materials consuming lesser energy than the
pyrometallurgical technique.
The resources required for battery recycling is just 387.4 MJ for the manufactur-
ing of 1 kg of cathode material, whereas the manufacturing of new battery production
required 795.4 MJ of 1 kg of cathode material. The recycling of batteries may also
recuperate 51% of the valuable resources used during the production of new batter-
ies, while nuclear energy consumption and fossil fuel were decreased by 57.25% and
45.3%, respectively [30].
326 Electric Vehicle Integration in a Smart Microgrid Environment

14.9 RECYCLING
Recycling of the EV batteries will be the major concern in upcoming decades,
whereas in the manufacturing of EV batteries, different chemical materials are used,
which makes the recycling process more complex. Lithium-ion battery packs have a
complex cell, module, and structure due to cylindrical or prismatic cells associated
with a series-parallel configuration such as wire bonding, welding, and mechanical
appending [31].

14.9.1 recycling methods

In each cell of lithium-ion, batteries require a broad spectrum of materials, which
results in a more complex recycling process, including

• Lithium, copper, manganese, aluminum, zinc, cobalt, and steel

• Components of plastic: polyethylene terephthalate and polypropylene
• Graphitic carbon
• Sulphuric acid electrolyte and solvent
• Fiberglass
• A management system of battery/coolant

The flow process of a lithium-ion battery recycling is depicted in Figure 14.5.

During recycling, all of the above materials must be separated from one another.
The following are the various recycling methods/processes currently used for lith-
ium-ion batteries.

FIGURE 14.5 Lithium-ion battery-recycling procedures.

Recycling and Disposal of EV Batteries 327

14.9.2 mechAnicAl procedure | mp

A mechanical method is used to split the components of EV batteries where elec-
trolyte securely separates to cut down the metals so that the process becomes easier.
Batteries can be fed into equipment through a conveyer, and the components of the
battery are shredded into small pieces and then moved toward recycling.

14.9.3 pyro metAllurgicAl procedure | pm

This is the process of extreme temperature, which includes two operations. Lithium-
ion batteries firstly kindle in the smelter at which substances are smashed down, and
other materials like separators and plastic are burned away. Then by carbon curtail-
ment, new alloys are produced, and metal alloys are further extracted to retrieve
materials [32]. During this process, costly metals like copper (Cu), nickel (Ni), and
cobalt (Co) are retrieved with maximum competence. The electrolyte, plastic, and
anode are oxidized and provide the energy for the procedure. The major advantages
of this process are as follows: (1) a mature and simple process, (2) reduction in size
and sorting out is not mandatory – a composition of LiB and nickel-metal hydride
batteries, which can be recovered, and (3) production is comprised of basic building
blocks, which might be used to synthesize materials of the cathode by various chem-
istries. On the other hand, the main drawbacks are as follows: (1) carbon dioxide
formation and higher power consumption during the procedure of smelting, (2) alloy
needs more refining due to which the cost of recycling will increase, (3) many prod-
ucts in lithium-ion batteries like graphite, plastic, and aluminium were not retrieved;
the procedure recuperates cobalt and nickel from the material of cathode and cop-
per from anode current collector, which accounts for only 30% w.r.t. lithium-ion
batteries for electronics, (4) the business strategy might not work effectively for EV
batteries, because of the lower concentration of cobalt. Furthermore, the aim of the
industries is cobalt reduction or eventually to produce cobalt-free cathode materi-
als. Recently, the pyro-metallurgical process exists extensively in the industry due
to its naivety and higher efficiency. There is no standardized procedure for pyro-
metallurgical recycling of the batteries, and every current method is peculiar to the
concerned company. Given Figure 14.6 portrays the whole recycling processes.

14.9.4 hydrometAllurgicAl procedure | hp

In the hydrometallurgical procedure, material retrieval is obtained through aqueous
chemistry, by leaching method, and later, concentration and filtration are done. In
lithium-ion batteries, solution ions are segregated using different technologies like
solvent extraction, ion exchange, electrolysis, and chemical precipitation [33]. In this
process, various steps are involved in recycling. The following steps are for disman-
tling, crushing, and sieving and after that, the leaching process takes place where
metals in current EOL of lithium-ion batteries are dissolved by the leaching process
and the latter leachate goes through by further process to isolate the metal ions. The
last step is the separation of different solvents and production of the cathode from
where one can develop a new battery. The major benefits of the hydrometallurgical
328 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 14.6 Recycling methods.

process are as follows: (1) generation of high purity materials, (2) most constitu-
ents of a LiB can be retrieved, (3) the operation of the hydrometallurgical process
is at a lower temperature, (4) the emission of CO2 is lesser in the comparison of
a pyro-metallurgical process [34]. The major disadvantages of this process are as
follows: (1) sorting is required, which needs more storage space due to which cost
increases and process complexity arises, (2) difficulty in the segregation of elements
like nickel, iron, cobalt, aluminum, and copper from the solution as they have similar
properties, and this again results in higher costs, and (3) water waste treatment and
associated costs that are associated with them.

14.9.5 direct recycling procedure | drp

DRP method of recovery is suggested for the direct processing and recovery of active
lithium-ion battery material, while maintaining the structure of its origin [35]. In this
procedure, battery components are separated with the use of methods such as mag-
netic separation, physical separation, and thermal process to evade active materials’
chemical breakdown. Active materials are then refined, and defects are retrieved
by the hydrothermal or re-lithiation process. Cathodes might be a combination of
two or more active materials, and the segregation of these materials is difficult. The
major advantages of this process involve the following: (1) relatively easy procedure,
(2) after regeneration of active materials, they can be used directly, and (3) in the
comparability of hydrometallurgy and pyro-metallurgy process, the recycling pro-
cess generates fewer emissions and secondary pollutants [20]. The major disadvan-
tages of this process are as follows: (1) involves preprocessing/comprehensive sorting
Recycling and Disposal of EV Batteries 329

for active chemical substances, (2) untested technology, which to date only exists at
the laboratory scale, (3) significant susceptibility to changes in the data stream, and
(4) this process is not flexible because the material is not purified, and therefore, the
cycle might not be acceptable to the cathode chemistry.

14.10 BEST PRACTICES OF LITHIUM-ION BATTERY RECYCLING

The best practices for recycling LIBs in the industries are discussed below

14.10.1 umicore compAny

Umicore company exploits the pyro-metallurgical procedure for the recycling of LIBs
by using the method of ultra-high temperature. This company is able to handle 7,000
metric tons/year, and the process of ultra-high temperature creates an alloy, copper
(Cu), nickel (Ni), and cobalt (Co), and generates slag for construction materials.
However, the further steps of segregation or purification involved the technique
of hydrometallurgy, which includes leaching, extraction of solvents, and precipita-
tion for the production of new cathode substances like nickel-manganese-cobalt
(NMC) and lithium cobalt oxide (LCO). The lithium had existed previously in the
slag, which was mainly used for building [36], but Umicore recently illustrates
that by doing the further process, the lithium-ion battery slag can be converted
into Li-recovery flow sheets with external collaborators [37]. Figure 14.7 showed
the systematic process.

FIGURE 14.7 Umicore recycling process [38].

330 Electric Vehicle Integration in a Smart Microgrid Environment

14.10.2 retrieve technologies

In North America, the company Retrieve Technologies utilizes a hydrometallurgical
procedure to recycle lithium-ion batteries. They manually dismantled the lithium-ion
battery pack to the cell level or module. Then, it is fed into a lithium brine flooded
hammer mill for crushing to lower the battery reactivity, neutralization of electro-
lyte, and avoid the emissions of gas [39]. This company follows a three-stream pro-
cess of materials that include plastic fluff, metal solids, and metal-enriched liquid.
The metal-enriched liquid contains mainly lithium-ion, precipitated and purified,
and solid metals might contain copper (Cu), aluminium (Al), and black mass. The
solid metals are provided to downstream smelters who are interested in nickel (Ni)
and cobalt (Co) contents. On the other hand, plastic may be disposed of or recycled.
However, lithium is recovered as lithium carbonate [40,41].

14.10.3 onto technology

Onto Technology recycled lithium-ion batteries by using the direct recycling method
where they obtain the electrode from lithium-ion batteries. After that cathode, the mate-
rial is obtained by combining and detaching the electrode from the current collector in
an aqueous alkaline solution. The devalued cathode content might revive and is possible
to use in the development of cells by using hydrothermal and additional heat treatment.

14.11 SAFETY INDICATORS
Safety precautions are the most important aspect of lithium-ion batteries while dis-
mantling, sorting, reorganizing, and recycling the retired batteries and must require
proper management to avoid risks.

14.12 DISMANTLING AND STORAGE

The lithium-ion batteries must be reassigned for reuse, and after cessation, the mod-
ule and battery pack must be dismantled with the elimination of the exterior circuit
and removal of battery scarp and data sensors. The remaining ability, immensely
high voltage, an exterior circuit, and reliability of connectivity need the dismantling
of demolished batteries to be carried out in an open area with safety precautions.
Disassembling processes might also prevent permanent impairment to the battery.
To achieve nondestructive and secure dismantling, it is suggested to dismantle the
module or pack level instead of cell level soft connection modules. Dismantling at
the cell level is not viable both economically and technically. To minimize safety
risks, transport and storage of demolished batteries should evade transportation of
long-distance, long-term storage, fire sources, rain exposure, and other collisions.

14.12.1 reorgAnizing And screening

Decommissioned batteries cause several risks like flatulence, maligned casing, leakage,
insulation failure, and short circuit. Batteries’ screening out with long service life and
Recycling and Disposal of EV Batteries 331

good performance is a difficult task while maintaining the protection of decommis-

sioned batteries. Traditional screening techniques are based primarily on cells that
are singled out, comprising assessment and testing for performance and appearance
through nano-CT to evaluate the three-dimensional structure qualitatively; to create
an electrochemical model, and to evaluate the electro-chemical external parameters’
performance. Some technologies have recently been developed, like contact-type ultra-
sonic, nondestructive screening technology that can quickly and correctly determine the
cell’s state of health and state of charge by correlating ultrasonic signals with a perfor-
mance of electrochemical and nondestructive and noncontact screening methodology.
Detection procedures are fairly successful for battery modules and might make visible
module, hence lowering the risk of short circuits and “dead Li”. Furthermore, at this
point, research has not come forward with any technology or equipment that can detect
and view battery modules.
The modules of batteries have to be rearranged after the screening, depending
upon database compiling demand details for secondary usage, functional details, and
test requirements comprising of the capacity, residual cycle life, internal resistance,
and material system. The battery management system of decommissioned batteries
seems more systematic when concerning about safety. Thus, new packs of the bat-
tery must consider the following factors while reorganization to minimize the safety
risk of separate integration of decommissioned batteries: optimizing the module
model connection to enhance flexibility; implementing intelligent technology of
time-division; equipping an effective temperature monitoring system and current
equalizers and the module protection through high-voltage and thermal monitoring.

14.13 TECHNOLOGICAL INITIATIVES
Research is simultaneously being done in three major avenues to curb the issue of
used batteries in the current scenario.
Composition: Different compositions such as lithium-sulphur batteries and lith-
ium-air batteries are being tested for performance and cost in comparison to LiBs.
Available resources: The reserves of lithium will become scarce as they are being
consumed at a very high rate since its demand increases exponentially with the
increase in EVs. Presently, we are expecting an exceed in the lithium cost if the bat-
tery recycling is not made efficient and economical.
Recycling methods: It has been observed that the methods of hydrometallurgical
and mechanical procedures are giving better results in the recollection of materials,
as well as it is a low energy-seeking procedure in comparison to the pyro-metallur-
gical procedure. The Lithium-ion Battery Recycling Initiative (LiBRi) is teaming
with different companies that are working on ways to recollect lithium from the slag
content. Research is underway to recover the maximum amount of lithium during the
dismantling of the batteries

14.14 CONCLUSION
Researchers are experimenting with different combinations of chemicals to replace
lithium and cobalt. The alternatives can be the materials that are found in abundance
332 Electric Vehicle Integration in a Smart Microgrid Environment

and are less toxic. Nevertheless, if new batteries are not efficient or more costly than
lithium, they could wind up negatively affecting the earth severely in general. If
recycling of EV batteries is taken as a business practice, key areas to be considered
can be a system to collect expired batteries, storing them, making logistical arrange-
ments, etc.

14.15 RECOMMENDATIONS AND FUTURE DIRECTIONS

Further research can be done on developing a supply chain model of closed loop
for the batteries which can have properly marked and labeled pieces right at the
manufacturing unit. This will help in the retrieval of these batteries after their end
life. Also, sustainability can be applied to the whole manufacturing procedure for
LiBs, which help in reusing the batteries to achieve zero carbon emissions.

REFERENCES

2. R. Irle. (2019). Global BEV & PHEV Sales for 2019. Available: www. EV-volumes.com.

3. P. Slowik, N. Pavlenko, and N. Lutsey, “Assessment of next-generation electric vehicle
technologies,” White paper, International Council on Clean Transportation (October
2016), 2016.

6. (2019). Better World Solutions. Available: https://www.betterworldsolutions.eu/

repurpose-ev-batteries/.
7. Z. A. Arfeen, M. P. Abdullah, R. Hassan, B. M. Othman, A. Siddique, A. U. Rehman,
and U. U. Sheikh “Energy storage usages: Engineering reactions, economic-technolog-
ical values for electric vehicles – a technological outlook” International Transactions
on Electrical Energy Systems, 2020.

9. L. Canals Casals and B. Amante García, “Assessing electric vehicles battery second life
remanufacture and management,” Journal of Green Engineering, vol. 6, pp. 77–98, 2016.
10. k. Environmental, “Research Study on Reuse and Recycling of Batteries Employed in
Electric Vehicles,” energy API, 2019.

12. (2019). Electric Vehicle Regulatory Reference Guide. Available: https://www.unece.

org/.


Recycling and Disposal of EV Batteries 333

19. P. Meshram, A. Mishra, and R. Sahu, “Environmental impact of spent lithium ION
batteries and green recycling perspectives by organic acids–A review,” Chemosphere,
p. 125291, 2019.

25. A. Sonoc, J. Jeswiet, and V. K. Soo, “Opportunities to improve recycling of automotive

lithium ion batteries,” Procedia cIRP, vol. 29, pp. 752–757, 2015.
26. J. Weisman. (2019). Tesla LIB and PV Recycling Programs. Available: www2.calre-
cycle.ca.gov.

32. L. Gaines, “The future of automotive lithium-ion battery recycling: Charting a sustain-
able course,” Sustainable Materials and Technologies, vol. 1, pp. 2–7, 2014.
334 Electric Vehicle Integration in a Smart Microgrid Environment

35. Y. Shi, G. Chen, F. Liu, X. Yue, and Z. Chen, “Resolving the compositional and struc-
tural defects of degraded LiNixCoy MnzO2 particles to directly regenerate high-perfor-
mance lithium-ion battery cathodes,” ACS Energy Letters, vol. 3, pp. 1683–1692, 2018.

37. (2020). “Recycling Process,” Available: [Online]. Available: https://csm.umicore.com/

en/recycling/battery-recycling/our-recycling-process.

 15 Design and Operation of
a Low-Cost Microgrid-
Integrated EV for
Developing Countries
A Case Study
Syed Muhammad Amrr
Indian Institute of Technology Delhi

Mahdi Shafaati Shemami

Aligarh Muslim University Aligarh

Hanan K. M. Irfan
Abul Kalam Azad University of Technology

M. S. Jamil Asghar
Aligarh Muslim University Aligarh

CONTENTS
15.1 Introduction .................................................................................................. 336
15.1.1 Central Power Station System ......................................................... 337
15.1.2 Distributed Generation System ........................................................ 337
15.2 The Design Scheme of Proposed Microgrid System ....................................340
15.2.1 Modifications in the Proposed Grid-Connected PV System ............ 341
15.2.2 Layout of the Proposed Control Strategy ......................................... 341
15.3 Detailed Controller Design and Its Working ................................................ 343
15.3.1 Mode Selector Controller.................................................................. 343
15.3.2 Source Selector Controller ................................................................346
15.4 Hardware Implementation of the Designed Controllers ............................... 347
15.4.1 The Experimental Setup and Results ............................................... 347
15.5 Hardware in the Loop Testing of Proposed Strategy ................................... 351
15.5.1 Hardware in Loop Results ................................................................ 351
15.6 Conclusion .................................................................................................... 356
References .............................................................................................................. 357

335
336 Electric Vehicle Integration in a Smart Microgrid Environment

15.1 INTRODUCTION
The overall prosperity of any country can be measured by electrical energy
production and its accessibility to its citizens. The socioeconomic development of
a country cannot be improved without an adequate and uninterrupted source of
energy. Conventional sources of energy viz., coal, gas, oil, etc. had been exten-
sively exploited to meet the ever-growing demand for energy. The power demand
was continuously increasing linearly in the past, and the same trend is predicted for
the future. Relying on coal, gas, oil, etc. is increasingly becoming unpopular due to
their adverse climatic and health impact so much so that 40% of CO2 production is
attributed to power and energy industries [1]. Further, with the increase in demand
for power, the continuous supply of these nonrenewable resources is not possible
due to their limited stock. As the awareness of the harmful effects of coal and fuel
has been promulgated, scientists and researchers have shifted their focus toward
cleaner sources of energy [2,3].
The last few decades have seen a significant jump in the process of energy har-
nessing and storage from renewable energy sources. As of today, several developed
as well as developing countries have significantly shifted their focus toward the har-
nessing of “green energy”, which is not only beneficial to the environment and popu-
lation but is economical too. Out of the many available options of generating power,
i.e., hydro energy, wind energy, geothermal energy, etc., it is solar energy that has
attracted main attention and investment globally. Due to its global availability and
ease of harnessing and storage, it has turned out to be quite popular in the under-
developed countries where nonrenewable resources are limited [4]. Solar photovol-
taic (PV) system is the most preferred means of tapping solar energy. Moreover,
the wide commercialization of the PV system is made possible because of power
electronic converters, which helps in energy conversion and grid integration of dis-
tributed generating system [5].
In today’s world, electricity is no longer a luxury, and it is of utmost importance
that even the most remote regions receive enough energy to power their homes.
Unfortunately, in the developing and under-developed countries, the demand for
power is more than the power generation. In countries like India, owing to factors
like the population size, the demand for electric power is significantly higher than
the power generating capacity of conventional power plants [6]. Thus, a major per-
centage, up to 70%, of power generation in India, is still done in thermal power
plants, whereas renewable energy constitutes only 22% of the total generated power.
The Indian government, like several other developing countries, has recognized
the potential of power generation through solar PV and has included incentives to
encourage investment in the same. The Indian government hopes to reach 175 GW
of renewable energy power generation, which includes 100 GW from solar, 60 GW
from wind, 10 GW from biomass, and 5 GW from small hydropower by 2022 with a
huge investment of around US$100 billion [7].
Considering the ambitious goal of achieving 100 GW solar energy, the Government
of India has introduced many subsidized schemes for the public to generate solar
power on their own. The power generation from the solar PV systems can be broadly
classified as [8] follows:
Design and Operation of a Low-Cost EV 337

15.1.1 centrAl power stAtion system

A power station or a generating station is the conventional power generation unit
from where power is transmitted to the grid only. No mechanism is used for energy
storage. Moreover, power generation entirely depends upon whether condition (solar
insolation), and no power is generated at night. From morning to noon and from noon
to evening, the power generation does not remain the same, and it varies.

15.1.2 distributed generAtion system

Instead of one centralized location of power generation, smaller units of different
types of generating plants are clubbed together, which are generally closer to the end
consumers. In general, there is a great saving of transmission and distribution losses
in the distributed generation system, which varies between 10% and 30%. In many
under-developed countries, it is about 30%. Distributed generation systems can be
further divided as follows:

The proposed chapter is based on the grid-connected PV system or grid assisted PV

system, where power is generated from the solar PV modules for the home loads
only. The surplus power is not fed to the grid at any point in time. The system is
338 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 15.1 A general solar grid-connected system.

connected to the grid to receive power from it when solar or battery storage energy
is not sufficient for the load. The basic diagram of the grid-connected PV system is
shown in Figure 15.1.
These systems assist in reducing the stress on the utility grid during the daytime
peak hours by isolating the home load from the grid to the PV system [9]. In develop-
ing countries, such a system (solar inverter) is commercially available in the market
at a reasonable price (100 US$ for 1 kVA system). The batteries in these systems get
charged from solar PV through the power conditioning unit (DC-DC converter), and
the home load demands are fulfilled through the utility grid. However, in the absence
of sunshine hour, the utility grid charges the batteries. During the blackout or load
shedding hours (nonavailability of grid power), the home load demands are met from
the batteries through the DC-to-AC inverter of the system. These commercial solar
inverters have some drawbacks in their control schemes, which makes inefficient
utilization of solar power. First, when the batteries are fully charged, the solar PV
system is underutilized, and only the trickle current is supplied from the solar PV
to the batteries. Because the battery cannot be charged further, stored energy can be
used only during the load shedding hours. Second, even in the presence of sufficient
Design and Operation of a Low-Cost EV 339

solar power, the home loads receive power from the utility grid, and solar power is
kept as a backup supply for the load-shedding period only. This results in the con-
sumption of a costly tariff-based grid power instead of a freely available solar power.
The charging preference scheme of the commercially available solar grid inverter
is based on the voltage level of the battery. Therefore, solar PV power charges the
batteries when the voltage level of the battery is greater than a preset voltage value.
Otherwise, the utility grid supplies power for charging the batteries. Moreover, in
the evening or night (in the absence of solar irradiance) and if the battery is in a
discharged condition, the utility grid will charge the batteries throughout the night.
Consequently, in the next morning, the batteries become fully charged (under the
assumption of no blackout or load shedding for longer hours at night). As a result of
this, the solar PV power, which will be available in the morning, will be unutilized
or underutilized since the batteries are already fully charged or partially charged
in the night from the utility grid. This implies that although such inverters will
ensure the reliable backup power supply, underutilization of solar PV power still
exists in these control schemes. Thus, solar grid inverters require a new control
algorithm to address the drawbacks of underutilization and inefficient harnessing
of solar energy.
This chapter establishes a very simple, low-cost, and effective control strategy to
illustrate a complete or total harnessing of solar energy, while reducing the stress on
the utility grid and shaving the peak loads. The proposed microgrid is established
with the use of a conventional (low cost) single-phase DC-to-AC inverter with the
integration of a solar charge controller to operate as a grid-connected solar inverter.
The microgrid control strategy is governed by the proposed low-cost controller cir-
cuits, which works without any microcontroller or digital signal processor (DSP).
The contributions of this chapter are as follows:

• The problems associated with commercially available solar grid-connected

inverter are identified.
• The existing backup UPS is modified into a solar grid-connected inverter.
• Two simple analog controllers are developed without microprocessors for
the proposed strategy for total harnessing and utilization of solar energy.
• Hardware lab testing of controllers and hardware in the loop testing of the
proposed scheme has been successfully implemented for the household
loads.

This chapter is organized as follows:

• Section 2: discussion about the solar grid-connected inverters and modifica-

tion in the proposed control strategy.
• Section 3: the details and working principle of proposed controllers.
• Section 4: hardware lab testing of controllers.
• Section 5: closed-loop application of the proposed strategy with compara-
tive analysis for a case study.
• Section 6: concluding remarks.
340 Electric Vehicle Integration in a Smart Microgrid Environment

15.2 THE DESIGN SCHEME OF PROPOSED MICROGRID SYSTEM

For a poorly maintained power distribution system, blackout and load shedding for
hours are a common practice to meet the power demands. Normally, people have
some backup power supply for emergency loads, e.g., emergency lights and fans.
The most common solution for the problem is the installation of conventional low-
cost DC-to-AC inverters with a lead-acid battery as power storage for emergencies.
An AC-to-DC charger circuit is also integrated with this inverter. Moreover, the
aspect of energy requirement in rural and urban areas is significantly different.
The demand in urban areas is for economic electric power, and in rural areas, it
is reliable electric power. With the availability of solar power and the low-cost
PV modules, a commercially available grid-connected inverter could solve these
problems, i.e., offering both economical and reliable sources of energy. The com-
mercial name of this inverter is the solar inverter. The grid-connected inverters
have a conventional DC-to-AC inverter, an AC-to-DC charger circuit, and an addi-
tional DC-to-DC buck-boost converter, which charges the batteries from solar PV
modules.
The proposed grid-connected PV system has the load classification in three dif-
ferent categories as shown in Figure 15.2. The high rating appliances are clubbed in
a heavy load category, which is powered by the utility grid only. The heavy loads
cannot be run during the load shedding hours by the proposed inverter scheme. The
moderate and low rating appliances are categorized in normal loads, which derive
the power from the proposed grid-connected PV system. The essential and emer-
gency loads are mainly powered by the proposed system. However, in the worst sce-
nario, when both grid and the proposed scheme gets exhausted, then a standby power
supply will be used for the emergency loads.

FIGURE 15.2 Block diagram of the proposed hierarchical control with load classification.
Design and Operation of a Low-Cost EV 341

15.2.1 modiFicAtions in the proposed grid-connected pv system

In a commercially available solar inverter, the batteries are more likely to be charged
from the grid than from the solar PV modules, which causes the problem of underuti-
lization and inefficient storage of solar energy. This problem can be resolved by inte-
grating two controllers in the system. One at the supply side and another on the load
side. Moreover, the existing backup inverter (which is normally available in every
house in the developing countries) can be transformed into a solar grid-connected
inverter by incorporating a solar MPPT charge controller with it. Moreover, the pro-
posed controller circuits are an addendum to the system. Therefore, the customer
need not buy a new inverter, and it can easily be implemented in the existing inverter.
Thus, the new system is based on grid assistance, which is a combination of stand-
alone and grid-connected systems. The proposed system will make sure that the
proper and complete utilization of solar energy takes place whenever solar irradiance
is available without drawing power from the grid.
Furthermore, the daytime peak load demand matches the maximum solar
irradiance hours. Therefore, the total and effective utilization of PV power by letting
off the household loads from the utility grid will also assist in reducing the stress on
the grid. Thus, the proposed system also acts as a peak power plant and called home-
to-grid (H2G) system [10].

15.2.2 lAyout oF the proposed control strAtegy

The proposed grid-connected solar inverter system consists of a solar charge
controller, a single-phase DC-to-AC inverter, and two proposed controller circuits as
shown in Figure 15.3. There are two input power sources, i.e., utility grid and solar
PV modules. The charging of batteries and the home load demands can be fulfilled
by both input sources. However, charging from the solar PV modules is given the
priority whenever solar irradiance is available. The solar MPPT charge controller

FIGURE 15.3 Proposed model of the solar inverter with two additional controllers.
342 Electric Vehicle Integration in a Smart Microgrid Environment

optimizes the power output from the solar PV module, which is then stored in the
battery bank. The battery bank of the electric vehicle can also be used as an energy
storage element and thus can work as a vehicle-to-home (V2H) system [11,12].
On the other hand, the source of power supply to charge the battery is determined
by the input controller called mode selector. This controller works on two modes:
(1) economy mode and (2) reliable mode. In economy mode, the objective is to trap
maximum solar power so that no or minimum energy gets unutilized from the PV
modules. This scheme is more suitable for the urban localities where load shedding
does not happen for a long duration. Whereas in rural areas, long hours of scheduled
and unscheduled load shedding are very common, and therefore, reliable mode oper-
ation is more suitable. In reliable mode, batteries are charged either through solar PV
module or grid depending upon their availability to fulfill the load demands during
the long hours of load shedding. The block diagram of the mode selector is shown
in Figure 15.4. The capacitor in the RC filter circuit 2 imitates the charging and
discharging behavior of the battery voltage. The filter circuits are used to tune the
charging and discharging rate of the capacitor to match the battery condition. Based
on the capacitor voltage, the decision of selecting the economy and reliable mode is
made with the help of level detectors and the SR flip-flop. The detail of its working is
demonstrated in the next section.
The second controller is placed at the load side, and it is termed as source selector.
This controller decides whether the household loads should meet their requirement
through solar energy stored in the battery or utility grid. The controller diagram of
the source selector is illustrated in Figure 15.5. This controller is set up with the help

FIGURE 15.4 Block diagram of mode selector controller.

FIGURE 15.5 Circuit diagram of the source selector controller.

Design and Operation of a Low-Cost EV 343

FIGURE 15.6 Schematic diagram of the read-type relay operation.

of two comparators with an adjustable predefined input reference voltage, SR flip-

flop, MOSFET, and static relay to switch the load. The basic schematic of load and
source connection using a read-type relay is demonstrated in Figure 15.6. So, once
the output of the SR flip-flop is high (shown in Figure 15.5), the read-type relay will
get actuated. The read-type relay is used for galvanic isolation of the load from either
of the power supply, i.e., utility grid or solar inverter. The detailed working of the
source selector is explained in the next section.
Using this controller, the consumer will be able to customize how they want to
utilize the stored solar energy. Even if there is no load shedding, the consumer can
choose to use free solar power instead of drawing power from the grid supply. This
makes the entire system more reliable and economical and solves the problem of
underutilization of solar energy.

15.3 DETAILED CONTROLLER DESIGN AND ITS WORKING

In this section, the working details of two analog controller circuits have been
explained individually along with their flowcharts.

15.3.1 mode selector controller

The circuit of a mode selector, as shown in Figure 15.7, has a rectifier, which is
used as an indicator for grid supply availability and the load shedding. The output of
LM7805 gives the logical output, i.e., logic 1 during the availability of grid supply
and logic 0 for load shedding period. The RC filters, which have large time constants,
adjust the charging and discharging time of capacitor C2. The voltage VC2 is the refer-
ence voltage for two-level detectors (LD) LM324. The other input port of these LD
has a predefined set and reset values that are related to the modes, i.e., reliability and
economy.
344 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 15.7 Detailed circuit diagram of mode selector controller circuit.

FIGURE 15.8 Control operation of mode selector controller.

During a prolonged power cut of grid supply, the voltage VC2 keeps on decreasing,
and when VC2 reaches the predefined voltage VLD1 of LD1, the output of S-R flip-flop
becomes set (Q = 1) as shown in Figure 15.8. This means the reliable mode opera-
tion gets activated, and the battery is going to be charged from the grid whenever the
mains supply gets restored. On the other hand, during the availability of grid supply
for the long hours, VC2 keeps on rising, and when it crosses the VLD2 level, the output
of LD2 will reset the flip-flop (Q = 0). The relay will change the charging connection
between the grid and battery to the solar PV module and battery, and now, the system
Design and Operation of a Low-Cost EV 345

TABLE 15.1
Operation of Reliable and Economy Modes Using Reference and LD Voltages
VC2 ≤ VLD2 VC2 > VLD1 Mode of
Output of LD2 Set Output of LD1 Reset QFF(t + 1) Operation
0 0 No change -
1 0 Charging by mains + PV Reliable
0 1 Charging by PV only Economy

will operate on the economy mode. Afterward, VC2 will remain at a high level, and
SR flip-flop will remain at a low level. One thing to note is that both modes of opera-
tions can be tuned using VLD1 and VLD2 according to the needs of the customer and
load shedding pattern of the area. The control operations of the mode selector can
be summarized using Table 15.1. Moreover, the working of this controller is also
described by the flowchart in Figure 15.9.

FIGURE 15.9 Flowchart of mode selector controller.

346 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 15.10 Detailed circuit diagram of source selector controller circuit.

15.3.2 source selector controller

The second controller is attached to the load side to switch the load from inverter
supply (battery) to mains supply and vice versa according to the battery voltage con-
dition. The detailed circuit diagram source selector is shown in Figure 15.10.
The source selector controller also has two level detectors where the inputs are
attenuated battery voltage and tunable voltage levels of set and reset as shown in
Figure 15.10. The battery starts to get charged either from the PV module or from
the mains depending upon the mode selector operation, i.e., reliable or economy
modes. The working principle of this controller is similar to the mode selector.
For instance, the output of LD1 is low and LD2 is high. Then, S will be 1 and
R will be 0, and hence, the flip-flop output will get set, i.e., Q = 1. The actuation
of SR flip-flop will make load connected to the inverter supply. The output will
remain under this state until battery voltage VB becomes equal to V’LD1. Then,
S = 0 and R = 1, and flip-flop will become reset, i.e., Q = 0. Once the output of
flip-flop becomes reset, the load will get disconnected from the inverter and will
get connected to the grid supply. Now, the load will meet the power requirement
from the utility grid.
Under this scheme, the battery can exist in two conditions. In condition 1, the bat-
tery will get charged and utilized for home load power demand, simultaneously. This
condition will exist during the daytime. The input of SR flip-flop will remain as S = 1
and R = 0. Thus, the output of SR flip-flop will remain high (Q = 1), and the load will
receive the power via inverter supply (battery).
Whereas in condition 2, the battery gets only discharged when solar power would
not be there, and the mode selector controller could not have actuated yet to switch
over to the mains. The battery voltage will start to decline (with Q being high) till
VB = V’LD1. At this condition, S = 0 and R = 1, which will reset the flip-flop, and thus,
the load will get connected to mains. The working of this scheme is also demon-
strated through a flowchart as shown in Figure 15.11.
Design and Operation of a Low-Cost EV 347

FIGURE 15.11 Flowchart of source selector controller.

15.4 HARDWARE IMPLEMENTATION OF THE

DESIGNED CONTROLLERS
The hardware circuits of the proposed controllers are shown in Figure 15.12. The
working of these circuits is first experimentally verified, and then, they are imple-
mented for the hardware in the loop analysis, which is discussed in the next section.
For brevity, the hardware analysis of the mode selector controller is only shown in
this section. The performance of the controller is tested according to the load shed-
ding condition, and its input-output results are recorded using a data logger as shown
in Figure 15.13. The data logger has been used for the continuous monitoring of
controller performance. The data has been stored at the sampling time of 20 seconds.

15.4.1 the experimentAl setup And results

The experimental setup comprises of mode selector controller, data logger, 5 V DC
power supply for saturation voltage (VCC) of flip-flops, multimeter, relay, and 230 V
AC power supply. The hardware testing of the mode selector controller along with the
relay operation is shown in Figure 15.14. The relay operation using the controller has
been accomplished with the proposed charging and discharging capacitor voltage-
based control strategy.
348 Electric Vehicle Integration in a Smart Microgrid Environment

(a)

(b)

FIGURE 15.12 Hardware control circuit of (a) mode selector and (b) source selector.
Design and Operation of a Low-Cost EV 349

(a)

(b)

FIGURE 15.13 (a) Data logger and (b) screenshot of the running status.

The recorded performance of the mode selector is shown in Figure 15.15. The
sky-blue line in Figure 15.15a represents the availability of grid power supply. In
Figure 15.15b, the grid supply voltage level falls to zero, implying the occurrence of
load shedding. The red line represents the voltage across the capacitor C2 in the mode
selector controller. The set and reset voltage levels of this experiment are VLD1 = 0.5 V
and VLD2 = 1.5 V, respectively.
350 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 15.14 Hardware testing setup of mode selector controller in the lab.

(a)

(b)

FIGURE 15.15 Hardware performance analysis of mode selector controller. during

(a) charging phase, and (b) discharging phase.
Design and Operation of a Low-Cost EV 351

Initially, the capacitor voltage VC2 is zero. But when the grid power supply comes
in, the VC2 starts to rise, and after around 2.45 hours, it becomes equal to 1.5 V. Then,
the output of S-R flip-flop becomes low (Q = 0) as shown in Figure 15.15a. This
implies, by the action of relay operation, that the battery charging will get discon-
nected from the mains and will now be charged only through the solar PV mod-
ules. On the other hand, during load shedding, C2 is in the discharging condition.
As VC2 decreases to 0.5 V, then the S-R output becomes high (Q = 1) as shown in
Figure 15.15b. The relay actuation will change the charging status, and now the bat-
tery will be charged through the mains, whenever available. Thus, the experimental
results are found to be consistent with the proposed theoretical control strategy.

15.5 HARDWARE IN THE LOOP TESTING

OF PROPOSED STRATEGY
In this section, the working of the proposed control scheme integrated with the
hybrid solar grid-connected inverter system has been demonstrated. The system has
been tested for the variable home loads for a period of 24 hours. These loads consist
of five ceiling fans of 100 W, four tube lights of 40 W, and five CFL bulbs of 12 W.
The hardware in the loop setup of the proposed scheme is shown in Figure 15.16,
where four rooftop solar PV panels of 250 WP each (Figure 15.16a) with two lead-acid
batteries of 12 V, 200 Ah for solar energy storage are used. Moreover, the comparison
between the proposed hybrid inverter control system and commercially available
solar inverter is also illustrated in this section. In the proposed hybrid inverter, a
separate MPPT solar charge controller is used with a rating of 24 V and 40 A. The
AC-to-DC and DC-to-DC converters in both the system are taken the same for better
comparison. The complete experimental setup is shown in Figure 15.16b. The per-
formance of a 1.44 kVA commercial solar grid-connected inverter and the proposed
system is demonstrated in the next subsection.

15.5.1 hArdwAre in loop results

First, the experimental results of the commercial solar inverter are demonstrated on
a typical day of the month of October from 6:00 to 18:00 hours, and the results are
recorded in Table 15.2. The three important parameters, i.e., power demand, solar
power, and grid supply are plotted from Table 15.2 in Figure 15.17 for better visu-
alization. Moreover, the power demand requirement of home load met by different
supply sources is also illustrated in Figure 15.18. It is evident from Figures 15.17,
15.18, and Table 15.2 that the solar PV modules come into usage only during the load
shedding hours (10 am–1 pm), and the rest of the times, the mains supply is used to
feed the power to the home loads. Similarly, when the mains power is restored, the
battery is getting charged by the grid even when there is enough solar power avail-
able from the solar PV modules. Thus, this demonstrates that the commercially avail-
able solar inverter is ineffective in trapping and utilizing solar energy, as identified
in the previous section.
On the other hand, the proposed system controls the effective charging of bat-
tery as well as the load allocation to ensure the complete utilization of solar energy.
352 Electric Vehicle Integration in a Smart Microgrid Environment

(a)

(b)

FIGURE 15.16 (a) 1 kWP rooftop PV panels and (b) hardware setup of the proposed system.

Under the economy mode during the sunshine hours, the charging of the battery is
done by the solar PV modules, and the home load requirements are met by the stored
energy using the DC-to-AC inverter. As the solar irradiance decreases to zero when
the sun sets, the stored battery power feeds the power to the load. When the battery
voltage VB decreases to the value V’LD1 of the source selector, the flip-flop will get
reset (Q = 0), and the load will get connected to the mains supply. In this way, both
the proper harnessing of solar irradiance (using mode selector) and the complete uti-
lization of solar energy (using source selector) have been achieved. The results of the
proposed scheme on a different day for a similar load demand pattern is recorded for
24 hours in Table 15.3. The power demand and supply from different sources under
TABLE 15.2
Performance of Commercially Available Solar Grid-Connected Inverter
PV Inverter I/ Power
Temp (in) Current DC input PV Volt Battery AC Load Status Energy Meter Solar Power Mains Power Demand
Time (Hr) I (W/m2) (°C) (A) (A) (V) Volt (V) (A) (Mains/PV) Reading (kWh) Output (W) (W) (W)
6 2 25 0.02 7.3 26 28.8 1.62 M 0.4 0.52 400 372.6
7 32 25.2 0.05 6.9 31.1 28.6 1.77 M 0.45 1.555 450 407.1
8 290 28.6 0.03 6.1 32.5 28.5 1.56 M 0.4 0.975 400 358.8
9 522 34.8 0 5.1 32.8 28.5 0.95 M 0.3 0 300 218.5
Design and Operation of a Low-Cost EV

9.8 866 40 0 5.4 32.6 28.5 0.9 M 0.2 0 200 207

10 866 40.2 9.79 1.25 27.6 26.8 0.9 PV 0 270.204 0 207
11 978 42.2 9.8 1.52 27.4 26.8 0.86 PV 0 268.52 0 197.8
12 1,024 42 5.86 5.23 29 28.6 0.96 PV 0 169.94 0 220.8
12.8 955 42 5.9 5.5 30.2 28.5 0.92 PV 0 178.18 0 211.6
13 948 42.2 0.03 1.15 31 26.4 0.81 M 0.2 0.93 200 186.3
14 920 44.1 0.04 1.03 30.9 26.4 0.83 M 0.2 1.236 200 190.9
15 796 43.2 0.03 0.9 31.3 26.4 1.22 M 0.3 0.939 300 280.6
16 670 41.9 0 0.47 31.6 26.5 1.52 M 0.4 0 400 349.6
17 160 34.8 0.03 0.72 31 26.5 1.98 M 0.5 0.93 500 455.4
18 5 31.7 0.04 0.71 25.3 26.5 1.03 M 0.2 1.012 200 236.9
353
354 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 15.17 Performance analysis of commercially available solar PV inverter.

FIGURE 15.18 The commercial solar grid-connected inverter fulfils the home load power
requirement on a typical day.

the proposed system is also illustrated in Figure 15.19. As can be seen in Table 15.3
and Figure 15.19, once enough solar irradiance is available to the system (at around
7 am), the battery gets charged through the solar PV module irrespective of the load
shedding. The mains supply gets cut off from the home load at around 9 AM, and till
6:25 pm, solar energy feeds the power to the home load. It is important to note that
solar irradiance becomes negligible after 5 pm, and even then, the stored solar energy
continues to fulfill the power demand of the load.
The power demand pattern under the proposed scheme is depicted in Figure
15.20. It can be seen from Figure 15.20 that the PV power gets utilized as soon as
the voltage level of the battery reaches the predefined set value, and when the battery
voltage level comes down to the reset value, then the load gets connected to the grid
supply. Thus, the proposed scheme ensures the complete utilization and harnessing
of solar energy.
TABLE 15.3
Performance of the Proposed Solar Grid-Connected Inverter
Time Ins Temp PV Current Inverter I/DC PV Voltage Battery AC Load Status Energy Meter Solar Power Mains Power battery
(Hr) (W/m2) (in) (°C) (A) input (A) (V) Volt (V) (A) (Mains/PV) Reading (kWh) Output (W) Power (W) Demand (W) backup (W)
0 0 23 0 0 0 24.3 2.84 M 0.7 0 700 653.2 0
1 0 23 0 0 0 24.3 2.84 M 0.7 0 700 653.2 0
2 0 23 0 0 0 24.3 2.84 M 0.7 0 700 653.2 0
3 0 23 0 0 0 24.3 2.84 M 0.7 0 700 653.2 0
4 0 23 0 0 0 24.3 2.84 M 0.7 0 700 653.2 0
5 0 24 0 0 0 24.3 2.84 M 0.7 0 700 653.2 0
6 1.8 26 0.03 0.5 24.8 24.3 2.0 M 0.6 0.7 600 460 12.15
7 130 25.3 0.85 0.72 24.5 24.4 2.45 M 0.5 20.8 500 563.5 17.57
8 353 28.9 1.73 0.71 24.8 24.8 2.11 M 0.6 42.9 600 485.3 17.6
Design and Operation of a Low-Cost EV

9 670 33 12.42 9.55 26.2 26 0.98 PV 0.3 325.4 300 225.4 248.3
10 835 40.2 13.85 8.46 25.9 25.8 0.84 PV 0 358.7 0 193.2 218.2
11 1038 42.3 14.12 9.45 25.7 25.6 0.95 PV 0 362.8 0 218.5 241.9
12 1,064 42.8 13.76 8.7 25.7 25.7 0.86 PV 0 353.6 0 197.8 223.6
13 1,030 42.4 12.9 7.6 26 25.8 0.78 PV 0 335.4 0 179.4 196.08
14 928 42.1 12.5 7.9 25.9 25.8 0.83 PV 0 323.7 0 190.9 203.8
15 703 42.8 12.02 14.36 25.2 25.1 1.53 PV 0 302.9 0 351.9 360.4
16 535 41.5 7.85 18.46 24.1 24 2.0 PV 0 189.2 0 460 443.04
17 165 35.7 2.45 16.35 23.4 23.4 1.62 PV 0 57.3 0 372.6 382.6
18 3 31.9 0.06 36.67 21.6 21 3.46 PV 0 1.3 0 795.8 770
18.25 0 31 0 37.9 0 19.8 3.52 PV 0 0 0 809.6 750.4
19 0 29 0 0 0 23.2 3.35 M 0.78 0 780 770.5 0
20 0 28 0 0 0 23.2 3.28 M 0.8 0 800 754.4 0
21 0 26 0 0 0 23.2 3.34 M 0.8 0 800 768.2 0
22 0 24 0 0 0 23.2 2.75 M 0.7 0 700 632.5 0
23 0 23 0 0 0 23.2 2.1 M 0.75 0 750 713 0
355
356 Electric Vehicle Integration in a Smart Microgrid Environment

FIGURE 15.19 Performance of the proposed system for home load.

FIGURE 15.20 The proposed solar grid-connected inverter control scheme fulfils the home
load power requirement on a different day.

15.6 CONCLUSION
The proposed work develops a solar grid-connected inverter using a conventional
single-phase convention backup inverter along with the solar charge controller and
two proposed controllers. The control strategy can be used in economical or reliable
modes, depending upon the grid supply availability. The proposed technique enables
the utilization of solar power even when there is no load shedding. The system also
indirectly works as a home-to-grid mode and thus reduces the electricity bill by let-
ting off the home load from the grid during the hours of high tariff rate (i.e., peak
hours). The proposed control strategy illustrates that the modified grid-connected
solar inverter is harnessing the solar energy more efficiently with better utilization of
stored solar energy. Moreover, the control scheme is built with low cost and does not
Design and Operation of a Low-Cost EV 357

require any expensive equipment. Thus, the design can be implemented in existing
home-based inverters; therefore, the customer does not have to invest a lot for the
complete system setup. The experimental results of the proposed strategy and their
comparison with the commercially available solar inverters illustrate that the devel-
oped scheme ensures the complete utilization of solar energy in contrast to the com-
mercially available system. Furthermore, the proposed system has surpassed months
of testing and has proved to be effective in its practical applicability. The future
extension of this work could involve more intelligent control features with a greater
number of input scenarios by using microcontrollers or DSPs.

REFERENCES

2. F. Ahmad and M. S. Alam, “Feasibility study, design and implementation of smart
poly-generation microgrid at AMU”, Sustain. Cities Soc., vol. 35, pp. 309–322, 2017.

7. https://pib.gov.in/PressReleaseIframePage.aspx?PRID=1514215 (Accessed: April 27,

2020).
8. B. H. Khan, “Non-Conventional Energy Sources”, 2nd Edition, Tata McGraw Hill
Education, New Delhi, 2009.

Index
Note: Bold page numbers refer to tables; italic page numbers refer to figures.

AC Level 1 36, 38, 44 charging mode 37, 39, 46, 55, 226
AC Level 2 36, 38, 44, 120, 240 charging standard 4, 40, 223, 224, 225, 229, 230
AC Level 3 36, 38, 44 charging station 1, 4, 5, 40, 51, 119–139, 267–278,
active power 41, 42, 51–53, 54, 117, 124, 271 295–311
active shielding 285–287 cloud-based 13, 252, 263
activity diagram 254, 255, 255 cobalt (Co) 185, 190, 194, 195, 221, 321, 323,
agent based modeling (ABM) 305 326–331
aggregated electric vehicle 34, 35, 42–45, 45, 50, cobalt oxide 185, 190, 194, 221, 321, 329
51, 55, 59, 242 communication delay 24, 45, 47
ancillary service 34, 35, 43, 45, 47, 53, 58, 59 communication infrastructure 57
artificial intelligence (AI) 16, 23, 25, 27, 61, communication standards 38, 40
250–252, 254, 258, 260–264 comparative algorithms 196–201
artificial neural networks (ANN) 25, 151, 152, conductive wires 275, 281–282
251, 258, 259, 263 control function 34, 47, 48, 50–53, 52, 54,
autoregressive moving average (ARMA) 258, 61, 228
262, 263 control loop 40, 42, 48, 50, 52, 53, 61
copper (Cu) 185–187, 271, 282–285, 314, 320,
backup power supply 339, 340 326–328, 330
battery capacity 1, 2, 43, 44, 152, 218, 219, 220,
222, 268, 308, 309 data flow diagram (DFD) 254, 255, 256
battery cell day-ahead-market 49, 249–264
assembling 185–187 DC/AC inverter 35
mining for the material 209 DC charging 37, 60, 223, 224, 226, 228–229, 231,
battery charger 34, 40, 41, 42, 51, 221 242, 302, 303
battery degradation 34, 57 DC/DC converter 35, 41, 42
battery disposal see battery recycling techniques DC Level 1 36, 38, 44
battery lifecycle 321, 324 DC Level 2 36, 38, 44
battery management system 143–177, 185, 187, DC Level 3 36, 38, 44
224, 331 deep learning 251
battery packaging 185, 187 dijkstra algorithm 309
battery pack assembly 188 direct recycling procedure (DRP) 328–329
battery purposing see energy storage systems distributed energy resource 11, 37, 98, 124, 242
(ESS) distributed optimization 49, 52
battery recycling techniques 192–195 distributed resource 10, 11, 37, 51
battery types capacities 318 distribution grid 34, 45, 51, 56, 80
bidding distribution system 1–5, 48, 51, 57, 98, 100,
framework 49, 50 121, 340
process 250, 260, 263 dynamic programming 195, 250, 251
bidirectional charger 40, 42, 51, 242
bidirectional power flow 35, 58 electric vehicle (EV)
blockchain 23, 25, 26, 60 batteries manufacture 184–185, 252
fleet 33–61
carbon dioxide (CO2) 183, 185, 252, 321, electromagnetic field shielding 283–288
322, 327 end-of-life (EOL) 184, 322, 324
carbon monoxide (CO) 322 energy management system 49, 209
challenge 7–29, 33–61, 73–92, 121, 145, 146, 215, energy service 49, 50
216, 235, 237, 242, 250, 268, 279, 285, energy storage 12, 34, 60, 98, 119–139, 143, 144,
295, 325 176–177, 181–209, 221, 242, 269, 299,
charging interface 38 319, 337, 342, 351

359
360 Index

energy storage systems (ESS) 122, 123, 129–131, machine learning 124, 125, 249–264
134, 135, 139, 143, 176–177, 181–209, mean absolute error 258
319 mechanical procedure (MP) 327, 331
entity-relation (ER) diagram 254, 256–258, 342 microgrid 1–7, 97–116, 249–264, 267–278,
environmental protection agency (EPA) 309, 320 335–357
expert system 260, 263
nickel (Ni) 144, 151, 152, 185, 190, 194, 195, 221,
finite element analysis 272 314, 316, 320, 321, 327–330
flammable chemicals 322 nickel manganese cobalt (NMC) 185, 186, 190,
flux concentrator 281–283 221, 329
forecasting 24, 196, 198, 205, 251, 258, 262
frequency control 46 passive shielding 283–285
frequency regulation 43, 46, 47–49, 53–54, 58, 61 placement problem 299–301, 305, 306
frequency response 43, 45–47, 54 power capacity 49, 50, 191
frequency support 46, 52, 242 power electronic converter 35, 45, 58, 336
power electronic interfaces 34, 35, 41, 51
generation following 49–50, 54, 55, 61 power factor 37, 40, 42
global warming potential (GWP) 184, 187, power factor correction (PFC) 40, 42
188, 325 power flow 35, 40, 41, 51, 58
graphite (C) 185, 190, 323, 327 power level 35, 43, 229
graph theory 305 power peak sheaving 191–192
greenhouse gasses (GHGs) 2, 73, 75, 83, 184, power quality 37, 51, 54, 121, 124, 236, 242
198, 200, 214, 252, 279, 295, 319–321, power smoothing 47–49, 54, 55
323, 325 pyro-metallurgical procedure (PM) 327, 329
grid service 42–53, 59, 61, 242
reactive power 36, 41, 42, 51–53, 54, 56, 57, 61,
hosting capacity 3, 34, 48–52, 54 242, 271, 280
hydrocarbons 322 reactive power support 51, 53, 54, 56, 242
hydrofluoric acid 321, 322 reactive shielding 287–288
hydrogen 81, 144, 316, 322 renewable energy resource 14, 34, 45, 49, 59, 60,
hydrogen fluoride (HF) 322 98, 319
hydrometallurgical procedure (HP) 327–328, 330 residential microgrid 10–12, 17, 164, 206

integrated charging system 35 smart grids 7–29, 61, 80, 84, 91, 189, 192,
195, 205, 206, 209, 250, 251, 262,
lithium cobalt (LiCoO2) 190, 323 263, 319
lithium cobalt oxide (LCO) 190, 329 software-defined networking (SDN) 23, 25, 26
lithium hexafluoro phosphate (LiPF6) 186, solar photovoltaic (SPV) system 231, 268–269,
321, 323 336–342, 344, 351, 352, 354
lithium-ion battery (li-battery) 145, 321–326, 328, solution 4, 7–29, 56, 57, 59, 74, 80, 99, 105,
329–331 110–112, 114, 124, 132–134,
lithium-ion battery recycling initiative 139, 166, 168, 169, 171, 176, 182,
(LiBRi) 331 188–190, 193–195, 198, 206, 208,
lithium-ion manganese oxide 321 209, 230, 231, 235, 237, 250, 254,
lithium iron phosphate (LFP) 185, 190, 221, 321 301, 327, 340
lithium manganate (LiMn2O4) 323 spatial allocation
lithium manganese oxide (LMO) 185, 186, 190, spinning reserve 50–51, 54, 55–56
221, 321 supervisory charging scheme 339
lithium nickel manganese 221, 321 support vector machine (SVM) 260 , 263
lithium phosphate (LiFePO4) 190, 323
load following 50, 54, 55 time series 103, 117, 258, 259
load frequency control 46
load levelling 54, 55, 58, 197 unidirectional charger 40–42, 51, 52
load shedding 338–340, 342, 343, 345, 347, 349, unified modeling language (UML) 254
351, 354, 356 urban mobility 234, 254
Index 361

vehicle interface 36 vehicle type 44

vehicle owner 34, 43, 80 V2G see vehicle to grid (V2G)
vehicle to grid (V2G) 34, 36–38, 40, voltage profile 50–52, 56
45, 52–61, 98–102, 105, 106, voltage support 51–53, 56
108, 116, 206, 212, 224,
242, 251 wireless charging 60, 223, 224, 229–230,
vehicle to home (V2H) 34, 342 267–278
 Taylor & Francis eBooks
www.taylorfrancis.com

A single destination for eBooks from Taylor & Francis

with increased functionality and an improved user
experience to meet the needs of our customers.

90,000+ eBooks of award-winning academic content in

Humanities, Social Science, Science, Technology, Engineering,
and Medical written by a global network of editors and authors.

TAYLOR & F RANCIS EBOOKS OF FERS:

Improved
A streamlined A single point search and
experience for of discovery discovery of
our library for all of our content at both
customers eBook content book and
chapter level

REQUEST A FREE TRIAL

[email protected]