Intelligent Electrical Systems A Step Towards Smarter Earth 2021

INTELLIGENT ELECTRICAL
SYSTEMS
A STEP TOWARDS SMARTER EARTH
Edited by
Satyajit Chakrabarti, Ayan Kumar Panja,
Amartya Mukherjee, and Arun Kr. Bar
Intelligent Electrical Systems:
A Step towards Smarter Earth
Satyajit Chakrabarti is Pro-Vice Chancellor, UEM a premier organization in education

sector in India, started many entrepreneurial ventures. He is also Director of Institute of
Engineering & Management (IEM) and University of Engineering & Management (UEM),
Kolkata, India.
Ayan Kumar Panja is Assistant Professor at IEM, Kolkata, India.
Amartya Mukherjee is Assistant Professor at IEM, Kolkata, India. He has authored several
books on embedded systems, IoT and sensor networks.
Arun Kumar Bar is PhD from Jadavpur University, He has more than 20 years of teaching
and research experience in the field of Mechanical Engineering.
Conference Proceedings Series on Information and Communications
Technology
Intelligent computing, data analysis and smart communication are currently among the
paramount research domains in IT. This series consisting of conference proceedings
mainly focuses on the recent advancement and research techniques in the field of intelligent
information processing, data analysis, computing and smart communication. The series
brings together latest research done in the area that addresses key challenges and issues
in the domain of intelligent sensing, ubiquitous computing, cyber-physical systems, signal
processing, data analysis, smart energy management, mechatronics, biomedical systems,
smart city infrastructure and artificial intelligence.
Series Editor
Nilanjan Dey is an Assistant Professor in the Department of Information Technology at
Techno International Newtown, Kolkata, India. He is a visiting fellow of the University of
Reading, UK. He was awarded his PhD. from Jadavpur University in 2015. He has authored/
edited more than 75 books with Elsevier, Wiley, CRC Press and Springer, and published more
than 300 papers. He is the Editor-in-Chief of International Journal of Ambient Computing
and Intelligence, IGI Global, Associated Editor of IEEE Access and International Journal of
Information Technology, Springer. He is the Series Co-Editor of Springer Tracts in Nature-
Inspired Computing, Springer Nature, and Series Co-Editor of Advances in Ubiquitous
Sensing Applications for Healthcare, Elsevier, Series Editor of Computational Intelligence
in Engineering Problem Solving and Intelligent Signal processing and data analysis, CRC.
His main research interests include Medical Imaging, Machine learning, Computer-Aided
Diagnosis, Data Mining etc. He is the Indian Ambassador of International Federation for
Information Processing (IFIP) – Young ICT Group.
Other titles in the Series

ICT for Competitive Strategies Proceedings of 4th International Conference on Information
and Communication Technology for Competitive Strategies (ICTCS 2019), December
13th–14th, 2019, Udaipur, India.
Edited by Durgesh Kumar Mishra, Nilanjan Dey, Bharat Singh Deora, Amit Joshi.
Computer-Aided Developments: Electronics and Communication Proceeding of the First
Annual Conference on Computer-Aided Developments in Electronics and Communication
(CADEC-2019), Vellore Institute of Technology, Amaravati, India, 2–3 March 2019.
Edited by Arun Kumar Sinha, John Pradeep Darsy.
Intelligent Electrical Systems:
A Step towards Smarter Earth
Proceedings of the International Conference on Industrial Electronics Mechatronics
Electrical & Mechanical Power (IEMPOWER 2019), November 21–23, 2019, Kolkata,
India.
Edited by
Satyajit Chakrabarti, Ayan Kumar Panja, Amartya Mukherjee, and Arun Kr. Bar
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 by CRC Press, Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, LLC
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 material
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.
Library of Congress Cataloging in Publication Data

A catalog record has been requested for this book
ISBN: 978-0-429-35599-8 (eBook)
Typeset in Minion Pro font

Typeset by Ozone Publishing Services, Puducherry, India.
Contents
Organizing Committee xix

Acknowledgements xxi
Preface xxiii
Foreword xxv
About the Editors xxvii
Chapter 01  Design of One Adaptive Controller for a

Dynamic Framework with Non-linear
Parameterisation 1
Samiran maiti and aChintya daS
1.1 INTRODUCTION 1
1.2 NON-LINEAR PARAMETERIZED FRAMEWORK 2
1.3 CONTROLLER DESIGN 3
1.2.1 Lyapunov Stability Analysis 4
1.4 RESULTS 5
1.5 CONCLUSION 8
REFERENCES 8
Chapter 02  A Comparative Study of Phasor Estimation

Techniques 11
p. r. pattanaik, BaSanta. k. panigrahi, S. pati, S. k. Sanyal, and nakul Sahu
2.1 INTRODUCTION 11
2.2 PHASOR ESTIMATION TECHNIQUES 12
2.2.1 Discrete Fourier Transform Based Phasor Estimation 12
2.2.2 Discrete Wavelet Transform Based Phasor Estimation 13
2.2.3 Least Square Method Based Phasor Estimation 14
v
vi  Contents
2.3 RESULTS AND DISCUSSION 14

2.3.1 Signal Reconstruction 14
2.3.2 Total Vector Error 15
2.3.3 Transient Monitor 16
2.3.4 Effect of Noise 16
2.3.5 Effect of Harmonics 18
2.3.6 Effect of Sampling Frequency 18
2.4 CONCLUSION 19
REFERENCE 19
Chapter 03  Adaptive Fuzzy PID Controller for a Current

Controlled Multilevel Inverter Fed Matrix
Converter 21
g. mohapatra, m. k. deBnath, and k. k. mohapatra
3.1 INTRODUCTION 21
3.2 MULTILEVEL INVERTER 22
3.3 SYSTEM MODELLING 23
3.4 MATRIX CONVERTER 24
3.5 DIFFERENTIAL EVOLUTION ALGORITHIM (DEA) 24
3.6 PROPOSED CONTROLLER 24
3.7 SYSTEM BLOCK DIAGRAM 26
3.8 RESULT AND DISCUSSION 27
3.9 RESULT ANALYSIS 29
3.10 CONCLUSION 29
APPENDIX 29
REFERENCES 29
Chapter 04  Multi-Objective Optimization for Enhanced Surface

Roughness and Microhardness of AA 6063/TiO2
Cast Aluminium Metal Matrix Nanocomposites
Fabricated by Friction Stir Processing 31
laxmanaraju, lingaraju dumpala, and pavankumar r
4.1 INTRODUCTION 32
4.2 EXPERIMENTATION 33
4.3 RESULTS AND DISCUSSIONS 35
4.4 CONCLUSION 38
REFERENCES 39
Contents  vii
Chapter 05  Self-Adaptive Multivariate q-Gaussian-based

Capacitor Placement for Reactive Power
Compensation in Distribution Feeders 41
deBaShiS jana and ranjita Chowdhury
5.1 INTRODUCTION 41
5.2 PROBLEM FORMULATION 42
5.3 DISCUSSION 46
5.4 CONCLUSION 47
REFERENCES 48
Chapter 06  Differential Evolution Coupled Neural-based

Prediction of Soil Water Content 49
Sayan Saha, kauShik majumder, Bitan ChakraBorty, and Sankhadeep Chatterjee
6.1 INTRODUCTION 49
6.2 BACKGROUND 50
6.3 METHODOLOGY 52
6.4 RESULTS & DISCUSSION 54
6.5 CONCLUSION 56
REFERENCES 56
Chapter 07  Numerical Investigation of Heat Conduction

in a Rectangular Composite Plate with Sinusoidal
Heat Input from Top Wall 59
deBarghya kar and Sumanta Banerjee
7.1 INTRODUCTION 59
7.2 METHODOLOGY 60
7.3 DISCUSSION 62
REFERENCES 64
Chapter 08  Analytical Modelling and Simulation of a

Hot-Wall Condenser Applied to Domestic
Refrigerator 65
S. n. Sapali and n. d. Shikalgar
8.1 INTRODUCTION 66
8.2 ANALYTICAL HEAT TRANSFER MODEL OF THE HOT-WALL
CONDENSER 67
8.2.2 Variation of outside heat flux with contact angle in the hot-wall
condenser 68
viii  Contents
8.3 NUMERICAL HEAT TRANSFER SIMULATION OF THE HOT-WALL

CONDENSER WITH ANGULAR CONTACT 69
8.4 CONCLUSION 69
REFERENCES 70
Chapter 09  Design and Analysis of a Dual-Axis Solar Tracker

with an In-built Low Cost Self-Cleaning
Mechanism 71
moumita pramanik, tuhin kumar Barui, pankaj kumar giri, hiranmay Samanta,
nirmal kumar deB, and hiranmay Saha
9.1 INTRODUCTION 71
9.2 DESIGN 72
9.3 OPERATION 74
9.5 CONCLUSION AND FUTURE SCOPE 77
REFERENCES 78
Chapter 10  Indices based Comparative Performance Analysis

of Standard Test Systems for Reliable Power System
Operation 79
r. Chowdhury and d. jana
10.1 INTRODUCTION 79
10.2 METHODOLOGY 80
10.3 DISCUSSION 81
10.4 CONCLUSION 84
REFERENCES 85
Chapter 11  Natural Ground-Water Recharge Scenario in Kolkata

and a Review of Artificial Recharging 87
j. Begam, S. Sarkar, S. Sarkar, and S. BhattaCharya
10.2 METHODOLOGY 88
11.3 DISCUSSION 90
11.4 CONCLUSION 90
REFERENCES 91
Contents  ix
Chapter 12  Performance Analysis of Grid Connected PV system

under Varying Irradiance and Temperature 93
B. tudu, a. BiSwaS, a. maji, k. Sardar, and k. k. mandal
12.2 METHODOLOGY 95
12.3 DISCUSSION 96
12.4 CONCLUSION 103
ACKNOWLEDGMENT 103
REFERENCES 104
Chapter 13  Development of a Low Cost Autonomous

Car Parking System: Towards Smart City 105
arjun dutta, ankur BhattaCharjee, and aBhijit kar gupta
13.2 METHODOLOGY 107
13.2.1 Explanation 108
13.2.2 Explanation 108
13.3 DISCUSSION 109
13.4 CONCLUSION 110
REFERENCES 111
Chapter 14  Monitoring Static Security Assessment in Its Full

Scope Using Common Artificial Neural Network 113
S. k. tiwary, j. pal, and C. k. Chanda
14.2 ANN DEVELOPMENT AND APPLICATION TO
TEST-BUS-NETWORK 114
14.4 CONCLUSION 119
ACKNOWLEDGMENT 119
REFERENCES 120
Chapter 15  Modelling of Solar Cell Considering One Diode

Model in MATLAB/Simulink Environment 123
SnehaShiS ghoShal, Sumit Banerjee, and Chandan kumar Chanda
15.2 MODELLING OF SOLAR CELL 124
x  Contents
15.3 SIMULATION MODEL OF A SOLAR CELL 126

15.5 CONCLUSION 128
REFERENCES 128
Chapter 16  Modelling of Intelligent Cooling of a Building in

MATLAB/Simulink Environment 131
SnehaShiS ghoShal, Sumit Banerjee, and Chandan kumar Chanda
16.2 THERMAL MODELLING OF A HOUSE REQUIRING
COOLING LOAD 133
16.2.1 Thermostat subsystem 133
16.2.2 Chiller subsystem 133
16.2.3 House subsystem 134
16.2.4 Environment subsystem 135
16.3 SIMULATION MODEL OF A HOUSE REQUIRING
COOLING LOAD 135
16.4 RESULT S AND DISCUSSION 136
16.5 CONCLUSION 137
REFERENCES 137
Chapter 17  Control and Monitor of Non-linear Gas Flow by

Developing a Self-Adaptive Fuzzy Controller 139
a. k. pal, Sampa paul, indrajit naSkar, and a. mondal
17.2.1 Mathematical Modelling of the System 140
17.2.1.1 Modelling of Control Valve 140
17.2.1.2 Modelling of Gas Sensor 141
17.2.1.3 Modelling of Process 141
17.2.1.4 Overall Representation 142
17.2.2 Controller Design 143
17.3 RESULTS AND COMPARATIVE STUDY 145
17.4 CONCLUSION 148
REFERENCES 148
Contents  xi
Chapter 18  Transient Stability Controlling and Assessment

of a Congested Power System in a Deregulated
Environment 151
d. Sarkar, r. BrahmaChary, and S. d. Barma
18.2 EIGEN VALUE ANALYSIS 153
18.3 STUDY OF THE IEEE 14 BUS TEST MODEL 155
18.4 CONGESTION STUDY IN THE IEEE14 BUS TEST SYSTEM 156
18.5.1 Analysis without using Facts Devices 156
18.5.2 Analysis with using Statcom Model 158
18.5.3 Analysis Using SVC Model 160
18.6 CONCLUSION 161
REFERENCES 162
Chapter 19  Bagasse Drying by Waste Heat Recovery from

Surrounding Walls of the Combustion Chamber
of a Jaggery Plant 163
a. B. Shinde, S. n. Sapali, and y. v. Sonone
19.2.1 Experimentation 165
19.3 DISCUSSION 166
19.3.1 Numerical Analysis 166
19.4 EXPERIMENTAL RESULTS 169
19.5 CONCLUSION 172
REFERENCES 172
Chapter 20  Study of Score Coefficient Functions for

DBSCAN-based Photometric Invariant Approach
in Vision Sensor-Based Shoe Detection for Human
Tracking 173
p. paral, a. Chatterjee, and anjan rakShit
20.2 PROPOSED RESEARCH FRAMEWORK 175
20.3 EXPERIMENTAL RESULTS 177
xii  Contents
20.3.1 Experimental Setup 177

20.3.2 Impacts of score coefficient functions on ACSAD distances 177
20.3.3 Performance Evaluations 178
20.4 CONCLUSION 179
ACKNOWLEDGMENT 180
REFERENCES 180
Chapter 21  Auto-adaptive Fuzzy Controllers with Automatic

Setpoint Adjustment 181
a. k. pal, indrajit naSkar, Sampa paul, and j. ChakraBorty
21.2.1 Fuzzy Setpoint weighting (FSW) scheme 184
21.3 RESULT 186
21.4 CONCLUSION 188
REFERENCES 188
Chapter 22  An Application of Multivariate Control Chart for

Online Process Monitoring in SMEs 191
S. Samanta and S. mondal
22.3 DISCUSSION 194
22.4 CONCLUSION 197
REFERENCES 198
Chapter 23  Analysis and Performance Indexing of Khagaria

132/33 kV Sub-station Using ETAP Load
Flow Solutions 199
arChiSman kar, deBaShiS jana, Chiranjit dey, and ranjita Chowdhury
23.2 METHODOLOGY & NETWORK FORMULATION 201
23.2.1 About ETAP 203
23.2.2 Khagaria Substation 204
23.3 RESULT ANALYSIS & DISCUSSION 205
23.4 CONCLUSION 208
REFERENCES 209
Contents  xiii
Chapter 24  Fractional Order PID controller for Setpoint

Tracking and Load Rejection 211
Sudipta ghoSh, arijit Bhowmik, dipaBala Sarkar, anirBan Bhatta, and
BiSwajit ChakraBorty
24.3 RESULTS 214
24.4 CONCLUSION 215
REFERENCES 216
Chapter 25  Analysis and Simulation of PWM-Based Half-Bridge

and Full-Bridge Switch Mode Amplifier for Active
Magnetic Bearing (AMB) 217
S. deBnath, p. k. BiSwaS, S. gupta, and j. laldingliana
25.2 ACTIVE MAGNETIC BEARING 218
25.3 CLOSED LOOP SYSTEM OF AMB 219
25.4 POWER AMPLIFIER 220
25.4.1 Full bridge switch mode power amplifier 221
25.5 PULSE WITH MODULATION 221
25.6 SIMULATION RESULTS 222
25.7 CONCLUSION 225
ACKNOWLEDGEMENT 225
REFERENCES 225
Chapter 26  Design and Analysis of High-Speed Phase Locked

Loop in 180 nm Technology 227
a. r. aBdul rajak, azeem gafoor mohammed, and deep Sehgal
26.2 MATHEMATICAL MODEL FOR PLL 229
26.3 TRANSISTOR LEVEL IMPLEMENTATION AND RESULTS 231
26.3.1 Phase Frequency Detector (PFD) 231
26.4 CHARGE PUMP (CP) 233
26.5 LOOP FILTER DESIGN (LF) 235
26.6 VOLTAGE CONTROLLED OSCILLATOR (VCO) 235
26.7 DIVIDER (DIV)\ 238
26.8 CONCLUSION 239
xiv  Contents
ACKNOWLEDGEMENTS 239
REFERENCES 239
Chapter 27  Design of APOD-PWM Based Multi-level Inverter

using Cuk DC–DC Converter with MPPT 241
rajarShi dhar and SarBojit mukherjee
27.2 SYSTEM DESIGN AND ANALYSIS 242
27.2.1 Tracking of MPPT 244
27.2.2 DC-DC Circuit 245
27.2.3 Proposed Multi-Level Inverter 247
27.2.4 Proposed Pwm Technique 248
27.3 SIMULATION RESULT AND ANALYSIS 249
27.4 CONCLUSION 253
REFERENCES 253
Chapter 28  Battery Storage Technologies: A Review 257

Bikram kumar Samanta, hiranmay Samanta, nirmal kr. deB, and hiranmay Saha
28.2.1 Lead Acid Battery 258
28.3 SODIUM SULFUR BATTERY 259
28.4 NICKEL CADMIUM BATTERY 260
28.5 LITHIUM-ION BATTERY 261
28.6 VANADIUM REDOX FLOW BATTERY 262
28.7 ZINC BROMIDE BATTERY 263
28.8 POLYSULPHIDE BROMINE 264
28.9 SODIUM-NICKEL CHLORIDE 264
28.10 DISCUSSION 264
28.11 CONCLUSION 267
REFERENCES 268
Chapter 29  Experimental Investigation of Solar Energy-Assisted

DC Refrigerator 269
n. d. Shikalgar and S. n. Sapali
29.2 DESIGN OF SOLAR PHOTOVOLTAIC PLANT 270
Contents  xv
29.3 EXPERIMENTAL STUDIES 271

29.5 CONCLUSION 274
REFERENCES 275
Chapter 30  Automatic Speed Control of a Three Phase

Induction Motor by Variable Frequency Drive 277
Sayantan Chanda, Chayan Bandopadhyay, maitreyee Banerjee,
Shalini mondol, and ankur BhattaCharjee
30.1 MOTIVATION 277
30.2 V/F SPEED CONTROL METHOD 278
30.3 VARIABLE FREQUENCY DRIVE 278
30.4 CLOSED LOOP SIMULATION OF V/F SPEED CONTROL
OF THREE PHASE INDUCTION MOTOR 279
30.4.1 Rectifier 280
30.4.2 L-C Filter 281
30.4.3 DC-DC Booster 281
30.4.4 PWM Generator 282
30.4.5 Inverter 283
30.4.6 Simulation Result 284
30.4.6.1 Inverter Output Voltage Curve 284
30.4.6.2 Rotor Speed Characteristics 284
30.4.7 FFT Analysis and Speed Analysis of Three Phase
Induction Motor 284
30.5 CONCLUSION 285
REFERENCES 286
Chapter 31  OWEC: Optimised Weighted Ensemble

Classifier Using Particle Swarm Optimisation 287
Syed fahim karim and ayan kumar panja
31.2 BACKGROUND 288
31.3 OPTIMIZED WEIGHTED ENSEMBLE CLASSIFIER (OWEC) 289
31.3.1 Approach 290
31.3.2 Advantages 292
31.3.3 Results 292
31.4 FUTURE SCOPE 293
xvi  Contents
31.5 CONCLUSION 294

REFERENCES 294
Chapter 32  Detection of Disaster Affected Regions Using

Drone-assisted Deep Learning Model 295
nikhil Singh, amitoSh Banerjee, Swadhin BhowmiCk, Sneha ChakraBorty,
and amartya mukherjee
32.2 METHODOLOGY AND ALGORITHM 297
32.3 DISCUSSION 300
32.4 CONCLUSION 301
REFERENCES 302
Chapter 33  Enhancing Massive Open Online Courses with

Brain Computing Interface and Artificial Intelligence 305
Sayantan pal, SaptarShi neogi, Sayanti dutta, Sumalyo datta,
nilanjana Bandyopadhyay, and yaShaSwi raj
33.3 CONCLUSION 310
REFERENCES 310
Chapter 34  Application of Multi-Objective Particle Swarm

Optimization Technique for Analytical Solution
of Economic and Environmental Dispatch 313
S. roy, d. laha, a. daS, S. Chatterjee, m. BiSwaS, r. k. mandal, and B. k. ghoSh
34.2 PROBLEM STATEMENT 314
34.5 CONCLUSION AND FUTURESCOPE 318
REFERENCES 318
Chapter 35  Generation of Electricity From Harmful Gases

using Fuel Cell 321
diptadip daS, Sayak Samanta, and praBal Saha
Contents  xvii
35.2.1 Materials Required 323

35.2.2 Fuel Cells 324
35.3 EXTRACTION OF OXYGEN 324
35.3.1 Extraction of Hydrogen 325
35.4 DISCUSSION 326
35.4.1 Parameters of fuel cell 326
35.4.2 Results 328
35.5 CONCLUSION 329
REFERENCES 329
Chapter 36  Exploring Potential Building Energy Saving

through Daylight Harvesting through Light Pipe 331
v. aditya, BidiSha BhattaCharya, and gunjan kumar
36.1.1 Aims of the Current Research 332
36.1.2 Problems and Factors that are to be Dealt 332
36.1.3 Research Problem 332
36.2 BACKGROUND INFORMATION 332
36.3.1 Window-wall ratio 333
36.3.2 Energy Modelling and Simulation Result 333
36.3.3 Building Occupancy Schedule 334
36.3.4 Building Energy Simulation 335
36.3.5 Daylight calculation 336
36.3.5.1 Daylight analysis report Zone wise 337
36.4 DISCUSSIONS 338
36.4.1 Expected Cost for Light Pipe 339
36.5 CONCLUSION 339
REFERENCES 339
Chapter 37  Greenhouse Gas Mitigation for College Building 341

amulayan Bhargow, vivek kumar, aman kumar, ruStam alam, and nikeSh kumar
37.3 CONCLUSION 345
REFERENCES 345
Organizing Committee
Chief Patron
Dr. Satyajit Chakrabarti, President, IEM Kolkata.
Patron
Dr. Satyajit Chakrabarti, Director, IEM Kolkata.
Dr. Amlan Kusum Nayak, Principal, IEM Kolkata.
Conference Chair Dr. Arun Kumar Bar, Dean (Engg.), IEM Kolkata.
Prof. Tapas Kr. Datta, Head, Dept. of EE, IEM Kolkata.
Convenor
Prof. Nikesh Kumar Singh, IEM Kolkata.
Prof. Swetadri Ghosh, IEM Kolkata.
Co-convenor
Prof. Ayan Kumar Panja, IEM Kolkata.
Prof. Amartya Mukherjee, IEM Kolkata.
Prof. K. Yashaswi Raj, IEM Kolkata.
Coordinator
Prof. Somnath Hazra, IEM Kolkata.
Prof. Prabir Kumar Das, IEM Kolkata.
Organizing Committee Members
Prof. Ranjita Chowdhury, IEM Kolkata. | Prof. Rajat Subhra Pal, IEM Kolkata. | Prof.
Debashis Jana, IEM Kolkata. | Prof. Ankit Ray Ghatak, IEM Kolkata. | Prof. Subhendu
Banerjee, IEM Kolkata | Prof. Nirban Kr. Saha, IEM Kolkata. | Prof. Debashish Ghosh,
IEM Kolkata. | Prof. Aramita De, IEM Kolkata. | Prof. Avishek Ray, IEM Kolkata. |
xix
xx  Organizing Committee
Prof. Aramita De, IEM Kolkata. | Prof. Arijita Das, IEM Kolkata. | Prof. Dipanwita
Bhattacharya, IEM Kolkata. | Prof. Gunjan Kumar, IEM Kolkata. | Prof. Joydip Ray, IEM
Kolkata. | Prof. Sudipta Swain, IEM Kolkata. | Prof. Mayurakshi Dev, IEM Kolkata. |
Prof. Dwaipayan De, IEM Kolkata. | Prof. Sandip Kr. Chaudhury, IEM Kolkata.
Acknowledgements
This book is itself an acknowledgement of cutting edge technological application in

various aspects of our daily life. We are thankful to the researchers, the contributors and
their enormous research effort towards the modern industrial applications. We are also
thankful to our co-researchers, chief patron sir, patron sir, conveners, co-coordinators and
all committee members.
Editors
xxi
Preface
In the modernized world, the intelligent systems are everywhere, starting from smart
city, society office, home and, of course, in our day-to-day life. The introduction of the
Industry 4.0 and the Society 4.0, however, increases the need for automated systems
while also opening doors for advanced research on intelligent automated systems. In a
modern industry, we often see highly smart pieces of equipment such as robotics arm
in an assembly line, automatic spraying and welding systems, software control actuators
to control the functionality of the deferent components of the industry without human
intervention. In addition, the automatic navigation systems have become so popular
and have eventually started replacing the manual driving systems. Furthermore,
the drones have evolved to be the greatest technological advancement of navigation
systems, in addition to advanced processing units and intelligent algorithms, which
has made drone research go an extra mile. Currently, the introduction of the MEMS
devices makes it possible to embed numerous pieces of sensing equipment in a credit
card size board. Another big revolution is the open-source technology revolution. This
ensures the advancement of hardware and software development to the next level by
exploiting the power of community-driven project development. This book primarily
emphasizes the current technological advancements in multidisciplinary fields such as
control systems, power system design, internet of things and many more. The book also
accumulates the recent research trends of various cutting-edge domain line machine
learning, robotics and mechatronics applications, advanced control algorithm designs
which are quite relevant to the contemporary researches on smart city, smart society
and industry 4.0.
Editors
xxiii
Foreword
On behalf of the local organizing committee and Institute of Engineering &

Management, the hosting institution, I take immense pleasure in welcoming you to
IEMPOWER 2019, Kolkata, for the Industrial Electronics Electrical & Mechanical
Power (IEMPOWER-2019).
Institute of Engineering & Management is one of the oldest premier technical
institutes in Eastern India. It has been imparting quality engineering and management
education since 1989. Being a NAAC-A grade institution, we have conducted numerous
international conferences and workshops across India and abroad. IEEE IEMCON, IEEE
UEMCON, IEEE CCWC, UEM Congress are some of the distinguished conferences
organized by IEM in University of British Columbia, Vancouver Canada, Las Vegas, New
York, Bangkok and Thailand every year.
In the present era of clean energy and mitigation of carbon emission, renewable
energy sources (Solar, Wind, Biogas, Micro and Mini Hydel etc.) have become
need of the hour. In addition to this, power system and demand-side management
can be addressed by the extensive use of energy storage devices whose efficient
integration with renewable energy sources is a challenge. To make the integrated
system smart in operation, we need an intelligent communication platform and
various smart technologies (IoT-based, wireless communication). The entire system
monitoring and control has to be executed by automation and supervisory control
schemes. International Conference on Industrial Electronics Electrical & Mechanical
Power (IEMPOWER-2019) is being organized by the Department of Electrical &
Mechanical Engineering, Institute of Engineering & Management, Kolkata, during
21st–23rd November 2019 under the mentorship of the Institution of Engineers
India (IEI) and ASME-IEM chapter. IEMPOWER-2019 aimed to provide a premier
platform for Electrical, Electronics and Computer Engineers, researchers, scientists
and academician from universities, research organizations and industries across the
globe to present their work and research experiences in the emerging areas, such as
Renewable Energy, Energy storage, Power Electronics & drives, Smart devices and
communication systems, Artificial Intelligence, Robotics, Control and automation etc.
We have invited speakers from Japan, Taiwan and premium universities of India, who
xxv
xxvi  Foreword
have kindly agreed to become part of this event. We are thankful to all IEM and UEM
family members, participants and keynote speakers for helping us to build this exciting
and highly technical conference program. We are very ambitious with this conference
as in coming years we may plan to extend this conference at an international level with
indexing in web of conferences.
Editors
About the Editors
Dr. Satyajit Chakrabarti, who is the Pro-Vice Chancellor, UEM, a premier organization
in education sector in India, has started many entrepreneurial ventures. He is also
Director of Institute of Engineering & Management (IEM) and University of Engineering
& Management (UEM), Kolkata. He received his PhD in computer science from
NIT, Agartala. He has more than 12 years of experience in teaching and research. His
primary research area includes Expert systems, Data mining, Nano Electronics and
Nanotechnology. Apart from academia, he also has a huge industry experience. He
was a Project Manager with extensive experience in planning, execution, monitoring,
communicating with stakeholders, negotiating with vendors and cross-functional teams
and motivating members.
Ayan Kumar Panja has received his B.Sc. degree in Computer Science (Honors) in
the year 2013 from Asutosh College, University of Calcutta, M.Sc. in Computer
Science from St.Xavier’s College in the year 2015, Kolkata and M.Tech from University
of Calcutta, Technology Campus. He is currently holding the position of Assistant
Professor at Institute of Engineering and Management, Kolkata. His main research
domain includes Machine Learning, Pattern Recognition, Audio Signal Processing,
Wireless Communication and Sensor Networks.
Amartya Mukherjee is an Assistant Professor at the Institute of Engineering &

Management, Salt Lake, Kolkata, India. He holds a bachelor’s degree in computer
science and engineering from West Bengal University of Technology and a master’s
degree in computer science and engineering from the National Institute of Technology,
Durgapur, West Bengal, India. His primary research interest is in embedded application
development, including mobile ad-hoc networks and FANET, Internet of Things,
mobile edge computing and machine learning. He has written several research
articles and books in the field of wireless networking and embedded systems from
Taylor & Francis, Elsevier, Springer and IEEE. His book Embedded Systems and
Robotics With Open Source Tools is one of the bestselling books in the field of
embedded application development.
xxvii
xxviii  About the Editors
Dr. Arun Kumar Bar is a PhD holder from Jadavpur University. He has more than
20 years of teaching and research experience in the field of Mechanical Engineering,
His Primary research area includes Mechatronics, Electrical Machines. He has written
many research articles in well-known Journals and Conferences.
CHAPTER 1
Design of One Adaptive

Controller for a Dynamic
Framework with Non-linear
Parameterisation
Samiran Maiti1 and Achintya Das2
1Department of ECE, Maulana Abul Kalam Azad University of Technology, West Bengal,
Haringhata-741249, India.
2Department of ECE, Kalyani Government Engineering College, Kalyani, Nadia, West Bengal,
Kalyani-741235, India.
1samiran.cemk@gmail.com, 2achintya.das123@gmail.com
ABSTRACT: This chapter represents a technique to design an adaptive controller for

a non-linearly parameterised framework with obscure parameters. Contingent upon
the variation of parameters such as lump inertia, rubbing coefficients, and a technique
has been utilised to devise a versatile control for a non-linear system. Further, stability
of the framework is guaranteed by reasonably choosing the Lyapunov capacity to develop
the controller. The adequacy of the proposed plan is illustrated with a cart pendulum
framework. Simulation results for a dynamic framework with non-linear parameterisation
show the effectiveness of the present work.
Keywords: Adaptive control, Non-linear system, Lyapunov stability theory, Non-linear

parameterisation.
1.1 INTRODUCTION
Adaptive control of non-linear framework has been a topic of broad research area. Most
accessible adaptive controllers manage control issue of frameworks with non-linearities
being linear in the obscure parameters. Non-linear parameterisation is exceptionally
normal in numerous physical plants. Versatile control for non-linearly parameterised
1
2  Intelligent Electrical Systems: A Step towards Smarter Earth
frameworks is a fascinating and challenging issue in control network. The Cart Pendulum
(Ge, Hang, and Zhang, 1999) is an intrinsically flimsy framework with exceptionally non-
linear dynamics. This framework has a place with the class of under-incited mechanical
frameworks having fewer control contributions than degrees of freedom. This renders
the control task even more challenging making the Cart Pendulum framework a classical
benchmark for the structure, testing, assessing, and looking at of changed control
procedures. The aim of this contextual investigation is to balance out the Cart Pendulum to
such an extent that the situation of the truck on the track is controlled rapidly and precisely
with the goal that the pendulum is constantly raised in its reversed position during such
developments. Reasonably, this straightforward mechanical framework is illustrative
of a class of elevation control issues whose objective is to keep up the desired vertically
arranged position consistently. The performance of the controlled frameworks is desired
to be optimal which ought to be substantial likewise when connected in the genuine
circumstance. Recently, a couple of specialists have endeavoured to research the keen
computational frameworks with adaptable and perfect control design by applying certain
methodologies for explicit applications. Ge, Hang, and Zhang (1999) presented controller
design for a non-linear system. In 2006, Muskinja and Tovornik proposed counteraction for
pendulum framework. Man and Lin (2010) also investigated control design for non-linear
system. Huang, Guan, Fukuda, and Sekiyuma (2010) presented sliding mode speed control
of a portable wheeled framework. In 2011, Razzaghi and Jalali discussed controller design
for an inverted pendulum system. Swarnkar, Kumar, and Nema (2011) proposed analysis
of versatile control system. Ghosh and Subudhi (2012) proposed controller design for cart
pendulum framework. Ranjan and Rai (2012) discussed performance analysis of a system
using versatile technique. Adhikary Mahanta (2013) proposed controller design for cart
pendulum system using sliding mode technique. Yang et al. (2014) also proposed controller
design under-actuated system. Pawar and Parvat (2015) presented versatile control design
for inverted pendulum system. Orostica, Duarte-Mermoud, and Salinas (2016) discussed
control of pendulum framework.
In the present work, the distinction is observed in terms of modification as the
performance of transient response being improved for dynamic framework with non-linear
parameterisation using the adaptive technique in spite of parametric uncertainties. The
remainder of the chapter is sorted out as follows: Section 1.2 – Non-linear parameterised
framework, Section 1.3 – Controller design, Section 1.4 – Simulation results, finally, Section
1.5 – concludes the work.
1.2 NON-LINEAR PARAMETERISED FRAMEWORK

The non-linear parameterised framework is considered as in Eq. (1.1):
(dxi (t ) / dt ) = x (t ) , i = 1,...n - 1
i +1
(dxn (t ) / dt ) = (1 / g ( x(t ),q (t )))[ f ( x(t ),q (t )) + g ( x(t ))u (t )] (1.1)
y (t ) = xi (t )
Design of One Adaptive Controller for a Dynamic Framework  3
where x(t ) = [ x1 (t ), x2 (t ),...., xn (t )] ÎÂn : state vector, u (t): control input and y(t): output.
Functions f ( x(t ),q (t )) and g ( x(t ),q (t )) are characterised as in Eq. (1.2):
k
f ( x(t ),q (t )) = å ql (t ) f l ( x(t )) = w Tf ( x(t ))q (t )
l =1
k
g ( x(t ),q (t )) = å ql (t )g l ( x(t )) = w gT ( x(t ))q (t ) (1.2)
l =1
where θ and k: number of unknown parameters.

The following supposition that is made about the framework in Eq. (1.1)
i. g ( x(t )) / g ( x(t ),q (t )) ¹ 0, "x(t ) ÎÂn . Without losing generality, it is assumed that
g ( x(t )) > 0 and g ( x(t ),q (t )) > 0 "x(t ) ÎÂn .
ii. q (t ) Î Sq , where Sq Ì Âk : compact set.
iii. Framework has full relative degree n.
iv. Framework (1): asymptotically steady.
1.3 CONTROLLER DESIGN

Consider a virtual input v(t) as in Eq. (1.3):
v(t ) = (1 / g ( x(t ),q (t ))[ f ( x(t ),q (t )) + g ( x(t ))u (t )] (1.3)

Then control input u(t) is written as in Eq. (1.4):
u (t ) = (1 / g ( x(t ))[ - f ( x(t ),q (t )) + g ( x(t ),q (t ))v(t )] (1.4)

v(t) can be designed (Narendra, & Annaswamy, 1989; Slotine, & Li, 1991) as in Eq. (1.5):
v(t ) = yd( n ) (t ) + cn ( yd( n -1) (t ) - y ( n -1) (t )) + .... + c1 ( yd (t ) - y (t )) (1.5)

where c1,…,cn: constant coefficient and yd(t): desired trajectory.
Hence, a stable estimator model for the framework is written as in Eq. (1.6):
Ù Ù
(d xi (t ) / dt ) = li ( xi (t ) - xi (t )) + xi +1 (t ) , i = 1,...n - 1
Ù Ù Ù Ù Ù Ù
(d xn (t ) / dt ) = l n ( xn (t ) - xn (t )) + (1 / g ( x(t ),q (t )))[ f ( x(t ),q (t )) + g ( x(t ))u (t )] (1.6)
Ù
q (t ) : estimate of θ(t) and li , l n : negative constants.
Subsequently, the estimation error dynamics of system is given as in Eq. (1.7):
(dei (t ) / dt ) = li ei (t )
Ù Ù
(den (t ) / dt ) = ln en (t ) + (1 / wgT ( x(t ))q (t ))[w Tf ( x(t ))q (t ) + g ( x(t ))u (t )] -
Ù Ù
T
g ( x(t )) q (t ))[w Tf ( x(t )) q (t ) + g ( x(t ))u (t )] - (1 / w gT ( x(t ))q (t ))[w Tf ( x(t ))q (t ) + g ( x(t ))u (t )]
Ù Ù
(w gT ( x(t ))q (t ))[w Tf ( x(t )) q (t ) + g ( x(t ))u (t )] - (w gT ( x(t )) q (t ))[w Tf ( x(t ))q (t ) + g ( x(
(den (t ) / dt ) = l n en (t ) + Ù
Ù (Ùw gT ( x(tT)) q (t ))(w gT ( x(t ))q (t )))
(w g ( x(t ))q (t ))[w f ( x(t )) q (t ) + g ( x(t ))u (t )] - (w g ( x(t )) q (t ))[w f ( x(t ))q (t ) + g ( x(t ))u (t )]
T T T
) = l n en (t ) + Ù
(1.7)
(w g ( x(t )) q (t ))(w g ( x(t ))q (t )))
T T
Ù ∼ Ù
where e(t ) = x(t ) - x(t ) : estimation error and q (t ) = q (t ) - q (t ) : parameter error.
Eq. (7) can be rewritten as in Eq. (1.8): Ù ∼ ∼ Ù ∼
(w gT`` ( x(t )) q (t ))(w Tf ( x(t )) q (t )) - (w gT`` ( x(t )) q (t ))(w Tf ( x(t )) q (t )) - (w gT`` ( x(t )) q (t )) g
(den (t ) / dt ) = l n en (t ) + Ù
Ù(w g ( x (t )) q (t ))(w∼ g ( x (t ))q (t )))
T T
Ù ∼ ∼
(w g ` ( x(t )) q (t ))(w f ( x(t )) q (t )) - (w g ` ( x(t )) q (t ))(w f ( x(t )) q (t )) - (w g ` ( x(t )) q (t )) g ( x(t ))u (t )
T` T T` T T`
t ) = l n en (t ) + Ù (1.8)
(w gT ( x(t )) q (t ))(w gT ( x(t ))q (t )))
1.3.1 Lyapunov Stability Analysis

To establish the stability properties of the non-linear framework, consider the Lyapunov
function as in Eq. (1.9):
æ ∼T ∼ ö
æ ö
( )
∼
V ç e(t ),q (t )÷ = t1 (eT (t )e(t )) / 2 + t 2 ç (q (t ) q (t )) / 2÷ (1.9)
è ø è ø
where t1 and t 2 : positive constants.
Using adaptive law in Eq. (1.10–1.11): Ù Ù
t1 (w g q (t ))w f - w g (w f q (t )) - w g gu (t )
T` T T` T T
∼
(d q (t ) / dt ) = - ( ) en (t ) (1.10)
t2 Ù
(w gT q (t ))(w gT q (t ))
Ù Ù
t (w g q (t ))w f - w g (w f q (t )) - w g gu (t )
T` T T` T T
Ù
(d q (t ) / dt ) = - ( 1 ) en (t ) (1.11)
t2 Ù
(w q (t ))(w q (t ))
T T
g g
Ù
Using estimate q (t ) in place of θ(t), the adaption law written as in Eq. (1.12)
Ù Ù
t (w g q (t ))w f - w g (w f q (t )) - w g gu (t )
T` T T` T T
Ù
(d q (t ) / dt ) = - ( 1 ) en (t ) (1.12)
t2 Ù
(w g q (t ))
T 2
Derivative of Eq. (1.9) can be written as in Eq. (1.13)

∼ n -1
(dV (e(t ),q (t )) / dt ) = t1 å li ei2 (t ) + t1l n en2 (t ) £ 0 (1.13)
i =1
∼
Since (dV (e(t ),q (t )) / dt ) in Eq. (1.13) is negative. Thus, the adaptive control system is
an asymptotics table. Utilising Hypothesis 2.1 in (Kristic, Kanellakopoulos, & Kokotovic,
1995), it tends to be ensured that eI (t ) ® 0 as t ® ¥ .
1.4 RESULTS
To investigate the planned control technique for accomplishing the ideal transient reaction,
the proposed control schemes have been actualised in simulation environment utilising
MATLAB software. This discussed technique is applied to a cart pendulum framework.
The structure goal is to manipulate the vertical edge of the cart pendulum system by
controlling the balanced force. The cart pendulum framework (Kristic, Kanellakopoulos,
and Kokotovic, 1995) described as in Eq. (1.14):
(dx1 (t ) / dt ) = x2 (t )
( g sin x1 (t ) - (mlx2 2 (t )sin x1 (t ) cos x1 (t )) / ( M + m)) cos x1 (t ) / ( M + m)
(dx2 (t ) / dt ) = 2
+ u
l ((4 / 3) - (m cos x1 (t ) / ( M + m))) l ((4 / 3) - (m cos 2 x1 (t ) / ( M + m)))
( g sin x1 (t ) - (mlx2 2 (t )sin x1 (t ) cos x1 (t )) / ( M + m)) cos x1 (t ) / ( M + m)
dx2 (t ) / dt ) = 2
+ u (t )
l ((4 / 3) - (m cos x1 (t ) / ( M + m))) l ((4 / 3) - (m cos 2 x1 (t ) / ( M + m)))
y (t ) = x1 (t ) (1.14)
where x1(t) denotes the angular displacement, x2(t) the angular velocity, M denotes the
lump inertia of cart, m: mass of pendulum, l signifying the half-length of pendulum,
g: gravity constant.
Consider suitable state variable as x1(t) and x2(t). The physical parameters belonging to the
system taken as, M = 1 kg, m = 0.2 kg, l = 0.5 m, and g = 9.81 kg–m/s2.
Therefore, the state space structure (Marc & Sastry, 1989; Wittenmark & Astrom, 2008) can
be described as in Eq. (1.15):
(dx(t ) / dt ) = f ( x(t ),q (t )) + g ( x(t ))u (t ) (1.15)
where x(t) denotes state vector, f(x(t)) and g(x(t)): smooth function.
Equation (1.14) can be represented in the form of Eqs (1.1–1.2) with q k as:
é ù
éq1 ù ê ( M + m) ú
ê ú
q k = êêq 2 úú = ê - ml ú
(1.16)
êëq3 úû ê 4 ú
ê l ( M + m) ú
ë3 û
w f ( x(t )) and w g ( x(t )) are given as in Eq. (1.17):
é g sin x1 (t ) ù
w f ( x(t )) = ê x2 (t )sin x1 (t ) cos x1 (t ) úú
ê 2
êë 0 úû ,
é 0 ù
w g ( x(t )) = êcos x1 (t )úú and g ( x(t )) = cos( x1 (t ))
ê 2
(1.17)
ëê 1 ûú
Control input u(t) considered as in Eq. (1.18)
ÙT ÙT
u (t ) = (1 / g ( x(t ))[ - q (t )w f + q (t )w g v(t )] (1.18)
The adaption laws given as in Eq. (1.19):

Ù Ù Ù Ù Ù
(d q1 (t ) / dt ) = -G1 (q 2 (t ) g sin x1 (t ) cos 2 x1 (t ) + q3 (t ) g sin x1 (t ))e2 (t ) / (q 2 (t ) cos 2 x1 (t ) + q3 (t )) 2
Ù Ù Ù Ù Ù Ù Ù Ù
qqÙ2 (2(ttx))1//(tdt
q 2 (t ) g sin x1 (t(()ddcos ) )+) ==
dt q -G
3 (t )2g((qq
-G
Ù
3(
sin (ttx))1xx(2t22))((ett )sin
(t ) /xx(1q((tt2))(cos
)sin
2 t ) cos
cos (tq) 1+(tq)3g(tsin
xx1 (2(ttx))1-
- )) 2 x1 (t ) cos x1 (t ) - u (t ) cos x1 (t ))e2 (t ) / (q 2 (t ) co
2 3
2 2 3 2 1 1
Ù Ù Ù Ù Ù Ù Ù Ù
qx((tt))x+ (d 1q(3t()tcos
) / dt ) = -G Ù -q sin (t ) g2x1sin x1 (t 2)Ù- qt2)(2-
t ) ux2(2t ()tcos
)sin3 xx11((tt)) e2 (tx)1/(t()q-
) cos 2 x (t ))e (t ) / 2(q (t ) cos 2 x (t ) +
u (t ) cos
x (t ))xe1 (t()t )-/ q(q13 (t()tg)1cos 1t ) + q23 (t ))
2 Ù
2 3 (t )sin
2 q (t ) gxsin x(t()tcos
) + q x(1t()) 2 (t ) cos x1 ( 2 1
1 3 1 2 2 1 3
Ù Ù Ù Ù
q1 (tx) g(tsin
)-cos ) - xÙ1 ((tt)) g-
q ( d q
Ù
qt2)(/tx)dtx(t2))2=cos
sin
( -G 2 x1 (t ) cos x1 (t )3- u (t ) cos x1 (2t ))e2 (t ) / (2q 2 (t ) cos x1 (t 2) + q 3 (t ))
(t )sin
x( (
-t
Ù
q) -(t u
) (
gt )
sincos
x (x
t ) (
-t q
Ù
)) e( (
t t
) )
x / (
(qt
Ù
(t
)sin ) cos
x ( t )x (
cost )x+(qt
Ù
2
) (
-t ))
2
1 1 3 1 31 1 1 1 22 2 2 1 1 1 3
Ù Ù Ù
q 2 (t ) x2 2 (t )sin x1 (t ) cos x1 (t ) - u (t ) cos x1 (t ))e2 (t ) / (q 2 (t ) cos 2 x1 (t ) + q3 (t )) 2 (1.19)
where, G1 , G 2 and G1 are positive gain. Figure 1.1. shows the output response of the
framework. The cart velocity response is represented in Figure 1.2. The control input and
output responses are delineated in Figure 1.3. The error response is delineated in Figure 1.4.
FIGURE 1.1 Output response.

FIGURE 1.2 State response (x2).
FIGURE 1.3 Control input and output.
FIGURE 1.4 Error response.

1.5 CONCLUSION
In this paper, designing an adaptive controller for a non-linear parameterised framework
is considered. The performance of the system with known and obscure dynamic within the
sight of disturbance is examined as well. The controller guaranteed the strength of the cart
pendulum framework in the upstanding position; likewise, it guaranteed the solidness of
the inner elements of the framework. Theoretical analysis demonstrates that the controller
can exhibit performance in the vicinity of parametric vulnerability. The simulation studies
are conducted on cart pendulum framework to evaluate the proficiency of the work.
Configuration using Lyapunov steadiness gives a stable versatile controller.
REFERENCES
1. Ge, S. S., Hang, C. C., and Zhang, T., 1999. A direct adaptive controller for dynamic systems
with a class of nonlinear parameterizations. Automatica, 35(4), 741–747.
2. Muskinja, N., and Tovornik, B., 2006. Swinging up and stabilization of a real inverted
pendulum. IEEE Transactions on Industrial electronics, 53(2), 2004–2016.
3. Man, S. W., and Lin, S. J., 2010. Nonlinear control design for a class of under actuated
system. IEEE International conference on Control Applications, 1439–1444.
4. Huang, J., Guan, H. Z., Matsuno, T., Fukuda, T., and Sekiyama, K., 2010. Sliding-mode
velocity control of mobile-wheeled inverted-pendulum systems. IEEE Transactions on
Robotics, 26(4), 241–247.
5. Razzaghi, K., and Jalali, A. A., 2011. A New Approach on Stabilization Control of an
Inverted Pendulum Using PID Controller. International Conference on Control, Robotics
and Cybernetics, 81–85.
6. Pankaj, S., Kumar, S. J., and Nema R. K., 2011. Comparative Analysis of MIT Rule and
Lyapunov Rule in Model Reference Adaptive Control Scheme. Innovative Systems Design
and Engineering, 2(4), 154–162.
7. Ghosh, A., Krishnan, R. T., and Subudhi, B., 2012. Robust proportional-integral derivative
compensation of an inverted cart-pendulum system: an experimental study. IET Control
Theory and Applications, 6(8), 1145–1152.
8. Ranjan, R., and Rai, P., 2012. Performance Analysis of a Second Order System Using MRAC.
International Journal of Electrical Engineering and Technology, 3(3), 110–120.
9. Adhikary, N., and Mahanta, C., 2013. Integral backstepping sliding mode control for
underactuated systems: Swing-up and stabilization of the cart-pendulum system. ISA
Transactions, 52(6), 870–880.
10. Yang, C., Li., Z., Cui, R., and Xu., B., 2014. Neural network-based motion control of
underactuated wheeled inverted pendulum models. IEEE Transactions on Neural Networks
and Learning Systems, 25(11), 2004–2016.
11. Pawar, J. R., and Parvat, J. B., 2015. Design and Implementation of MRAC and Modifed
MRAC technique for Inverted Pendulum. International Conference on Pervasive
Computing, 1–6.
12. Oróstica, R., Duarte-Mermoud, A. M., and Salinas, J. C., 2016. Stabilization of inverted
pendulum using LQR, PID and fractional order PID controllers: A simulated study.
IEEE International Conference on Automatica, 1–7.
13. Narendra, K. S., and Annaswamy, A., 1989. Stable Adaptive Systems. NJ, Prentice Hall,
705–752.
14. Slotine, J., and Li, W., 1991. Applied nonlinear control. NJ, Prentice Hall, 311–391.
15. Kristic, M., Kanellakopoulos, I., and Kokotovic, P. V., 1995. Nonlinear and Adaptive Control
Design. New York, Jhon Wiley and Sons, 1st edition, 87–122.
16. Marc, B., and Sastry, S., 1989. Adaptive control stability, convergence, and robustness.
NJ, Prentice-Hall, 1st edition, 294–323.
17. Wittenmark, B., and Astrom, K. J., 2008. Adaptive Control. Dover Publications, 2nd edition,
1–40.
CHAPTER 2
A Comparative Study of Phasor

Estimation Techniques
P. R. Pattanaik, Basanta. K. Panigrahi, S. Pati, S. K. Sanyal, and
N. Sahu
SOA University, Bhubaneswar, India.
ABSTRACT: Phasor measurement units (PMU) are an essential part of modern power
system. Phasor of alternating signals is widely used in protection and control schemes of
power network. The sampled data of voltage and current signal are processed continuously
through phasor estimation algorithms to obtain phasor values at each instant. Digital
relays and controllers utilize these phasor values for system operation and control. Phasor
estimation techniques based on discrete Fourier transform, discrete Wavelet transform, and
least square method are briefly presented and a comparative assessment has been carried
out. The performances of phasor estimators have been evaluated based on ‘Total vector
error’ and ‘Transient monitor’ criterion. The effects of noises, harmonics, and sampling
frequency have been studied and results are shown and discussed. A two-terminal system
has been considered for simulation under Matlab®/Simulink™ environment.
Keywords: Phasor estimation, Discrete Fourier transform, Discrete wavelet transform,

Least square method, Total vector error, Transient monitor.
2.1 INTRODUCTION
Phasor measurement units (PMU) are assigned for providing phasor values of voltage
and current signal at critical buses. PMU used phasor estimation algorithm to extract
phasor by processing real-time sampled data of analog signals provided by instruments
transformer. PMU-based relaying was originally developed by American Electric Power
Corporation. Signals of voltage and current are sampled using appropriate Nyquist filters,
and phasors of sequential components are calculated using discrete Fourier transform
(DFT) (Phadke et al., 2018). Phasor also plays an important role in wide area management
system (WAMS) and in controlling Flexible AC transmission system (FACTS) devices.
Relays need to be fast, accurate, and adaptive as operating conditions of modern power
11
network are usually varying. An adaptive relay requires phasor data to adjust their setting
according to prevailing conditions. Therefore, accurate phasor estimation becomes a
necessity for the development of modern protective schemes (Xia and David, 1994; Sarangi,
2016). Technical literature survey suggests several phasor estimation techniques developed
by researches. A DFT-based algorithm of fault detection and location in a transmission
line is proposed in Jiang et al. (2000a) and (2000b). A computer measurement-based time
synchronizing technique is used to calculate real-time phasors in Ree et al. (2010). A
synchronized phasor measurement concerning research also includes frequency estimation,
state estimation, instability prediction, and adaptive relaying. An impedance relaying
scheme based on modal Kalman filtering (KF) is proposed in Si et al. (1990). Estimating
phasor is modelled as a linear filtering problem and an adaptive algorithm for calculating
real-time phasor values by quasi-second-order optimization technique is proposed in Abbas
and Member (2012). Prony method is discussed as a dynamic estimator, and a comparative
analysis with DFT-based method has been made in Serna and Member (2013). An artificial
neural network-based algorithm is presented to estimate voltage magnitude of critical buses
under normal and contingent states (Aravindhababu and Balamurugan, 2012). The concept
of estimating phasors through discrete wavelet transforms (DWT) is given in Wong et al.
(2001). A detailed analysis on the performance of DWT-based phasor estimation has been
carried out and limitations are discussed in Brahma et al. (2008). Least square method
(LSM) has been used to estimate phasors and frequency of signals in Nayak et al. (2014)
and Pradhan et al. (2005). An investigation on the performance of LSM and KF-based
phasor estimating algorithm is presented in Khodaparast and Khederzadeh (2017).
In this chapter, a comparative assessment has been carried out between three phasor
estimators based on DFT, DWT, and LSM. These techniques are tested under steady-state
conditions by processing current signal obtained from a simulation model. Total vector
error (TVE), a criterion set by IEEE C37. 118 standards for evaluating phasor estimation
quality have been used while comparing the estimators. Another important index ‘Transient
Monitor’ (TM) for evaluating phasor techniques performances has also been discussed.
The effect of noises, harmonic, and sampling frequency is studied and discussed.
2.2 PHASOR ESTIMATION TECHNIQUES

Phasors are rotating vectors used for the representation of alternating voltage and
current signals. The decision regarding protection and control mostly depends upon the
magnitude and phase angle of voltage and current. Therefore, phasor extraction method
accuracy needs to be very high. The phasor estimation techniques based on DFT, DWT,
and LSM are briefly discussed in the following sections.
2.2.1 Discrete Fourier Transform-Based Phasor Estimation

A sinusoidal signal can be expressed as
xk = X m cos( Wk + f ) (2.1)
A Comparative Study of Phasor Estimation Techniques  13
where Xm is the amplitude of signal, k is the sample number, φ is the phase angle in radians,
and Ω= 2π/N (N is one cycle window length).
DFT of the sinusoidal signal corresponding to fundamental frequency is
N -1 2P
2 -j k
X1 =
N
å
k =0
xk e N
where X1 is the phasor estimated in terms of peak and can also be represented as
X 1 = X m e jf (2.3)
The signal can be reconstructed by using phasor values at each instant as
Ù
x k = [ X m e jf ]e jw0 kTs (2.4)
where Ts is the sampling time and ω0 is the fundamental angular frequency
2.2.2 Discrete Wavelet Transform-Based Phasor Estimation

DWT-based phasor extraction method requires one referenced sinusoidal signal (Wong
et al., 2001). The referenced and the measured signals must be defined over a common
clock pulse provided by GPS satellites. The signals then decomposed by the same wavelet
function into low-frequency and high-frequency components. Only low-frequency
component of the signals are considered for extraction as power system signals are of low
frequency (50 Hz) with high magnitude. Let Aref and Ames be the vectors representing the
approximation coefficient of referenced signal and measured signal, respectively. Then
the phase angle φ can be calculated as
æ Aref , Ames ö
j = cos - 1ç ÷ (2.5)
ç | Aref | Ames | ÷
è ø
where <> operator is for dot product between two vectors.
Now another sinusoidal signal is constructed with a phase angle value of φ, and a
low frequency component is obtained using DWT. Let A1 be the vector representing the
approximate coefficient of the new sinusoid. Then magnitude (Xm) of the required phasor
is calculated as
Ames
Xm = (2.6)
A1
X1 phasor estimated in terms of peak and can be represented as
X 1 = X m e jj (2.7)
The signal can be reconstructed by using phasor values at each instant as
Ù
x k = [ X m e jf ]e jw0 kTs (2.8)
2.2.3 Least Square Method-Based Phasor Estimation

A sinusoidal signal with a decaying DC component can be expressed as
- kTs
xk = X m sin(k w 0Ts + f ) + c0 e t (2.9)
where c0 is value of decaying DC at t = 0 sec.
The signal in state space representation can be expressed as
[ P ][ x ] = [Q ] (2.10)
where
é sin ( w0Ts ) cos ( w0Ts ) 1 -Ts (Ts ) ùú
2
ê
ê sin ( 2w0Ts ) cos ( 2w0Ts ) 1 -2Ts ( 2Ts ) ú
2
[ P ] = êê · · · · ·
ú
ú (2.11)
ê · · · · · ú
ê ú
êsin ( N w0Ts ) cos ( N w0Ts ) 1 - NTs ( NTs ) úû
2
ë
The state vector
T
c0 c0 ù
[ X ] = éê X m cos f X m sin f c0 (2.12)
ë t 2t2 úû
and the measurement vector
T
[Q ] = éë x ( t0 + Ts ) x ( t0 + 2Ts ) · · · · · x ( t0 + NTs ) ùû (2.13)
The required vector containing phasor information can be obtained as
[X ] = ( éë P Pùû P ) *[Q]
T -1 T
(2.14)
The vector of reconstructed signal samples can be obtained as
éÙù
êQ ú = [ P ][ X ] (2.15)
ë û
2.3 RESULTS AND DISCUSSION

A comparative assessment has been made between DFT-, DWT-, and LSM-based phasor
estimation techniques. Highly accurate and fast phasor estimation is the requirement of
modern relaying. Performances of each estimator have been evaluated based on parameters
like TVE and TM. A normalised current signal of 50 Hz is sampled at 20 kHz used for
steady-state testing. The test results are plotted and discussed in the following sections.
2.3.1 Signal Reconstruction

The phasor values can be estimated through sampled data of the signal and any signal
can be reconstructed through its phasor value as discussed in the phasor estimation
techniques section.
Under steady-state conditions, the signal and the reconstructed signal should be equal
in phase and magnitude. Accurate phasor estimation will result in overlapping of the real
signal and estimated signal while plotting them in the time domain. A normalised steady-
state current signal is obtained from simulation and its sampled data is processed through
the phasor estimation techniques. The phasor values obtained are used to reconstruct the
signal. The real signal and estimated signal are plotted in Figure 2.1. It can be seen that all
the estimated signals are overlapping the real signal. The techniques satisfy the matching
criterion of accurate phasor estimation. However, this visual criterion may fail to detect
very small errors in phasor estimation and therefore cannot be set as an absolute standard
to determine the quality of phasor estimation technique.
2.3.2 Total Vector Error

Total vector error (TVE) is a parameter defined in IEEE C37.118 standard to evaluate the
performance of the phasor estimation technique. Inaccurate phasor estimation produces
considerable error in calculating magnitude and phase of signal. The TVE value of a phasor
measurement technique should be below 1% under steady-state conditions. TVE can be
mathematically defined as
® ®
Xe - X r
TVE ( % ) = ®
´ 100 (2.16)
Xr
It can be observed from Figure 2.2 that under steady-state condition, the values of DFT and
LSM TVEs are approximately zero, whereas TVE value of DWT oscillates between 6.5%
and 7.5% and therefore fails to satisfy the IEEE criterion of accurate estimation.
Real Estimate
2
Current (A)
–2
(a) 0.4 0.45 0.5 0.55 0.6
2
Current (A)
–2
(b) 0.4 0.45 0.5 0.55 0.6
2
Current (A)
(c) –2
0.4 0.45 0.5 0.55 0.6
Time (s)
FIGURE 2.1 Current signal reconstruction under prefault conditions (a) DFT (b) DWT (c) LSM.
× 10−7
4
TVE (%)
2
(a) 0
0.48 0.49 0.5 0.51 0.52 0.53 0.54 0.55
8
TVE (%)
(b) 6
0.48 0.49 0.5 0.51 0.52 0.53 0.54 0.55
× 10 −6
TVE (%)
(c) 0
0.48 0.49 0.5 0.51 0.52 0.53 0.54 0.55
Time (s)
FIGURE 2.2 TVE values in steady-state conditions prefault conditions (a) DFT (b) DWT (c) LSM.
2.3.3 Transient Monitor

Transient monitor is another parameter used as a quality measure of phasor estimation
techniques. TM index registers sudden change between actual and reconstructed signal.
Mathematically, TM index can be expressed as follows:
é Ù
ù (2.17)
tn = ê xn - xn ú
ë û
n=k
TM = å
n=k - N
tn (2.18)
where tn is the estimation error at nth instant.

For accurate phasor estimation, the value of TM should be zero. Estimation error and
transient monitor index value have been calculated and results are shown in Figure 2.3. It
can be observed that for steady-state condition tn and TM index of all the three techniques
maintaining almost zero value.
2.3.4 Effect of Noise

In this section, the effect of noise concentration on phasor estimation technique is
studied. The current signal is contaminated with random white Gaussian noise. The level
of noise contamination is increased from 0 dB to 100 dB SNR with a step size of 1 dB
SNR. The TVE value of all three techniques is obtained at each step and plotted against
the SNR level. It can be observed from Figure 2.4 that TVE value of phasor estimation
techniques vary inversely with SNR values. This means phasor estimation error increases
with increase in noise level.
DFT DWT LSM

0.2
0.1
TM (Amp)
−0.1
(a)
−0.2
0.4 0.45 0.5 0.55 0.6
Time (s)
× 10−6
3
TM (Amp)
1
(b)
0
0.4 0.45 0.5 0.55 0.6
Time (s)
FIGURE 2.3 (a) Estimation error (b) Transient monitor.
150
TVE (%)
100
50
(a) 0
0 20 40 60 80 100
SNR (dB)
400
TVE (%)
200
(b)
0
0 20 40 60 80 100
SNR (dB)
4000
TVE (%)
2000
(c)
0
0 20 40 60 80 100
SNR (dB)
FIGURE 2.4 Effect of noise (a) DFT, (b) DWT, (c) LSM.
2.3.5 Effect of Harmonics

The current signal is contaminated with third and fifth harmonics and their relative
amplitude is successively increased from 1 to 50%. TVE of all the three-phasor estimation
techniques is calculated for different total harmonic distortion (THD) values and results are
plotted in Figure 2.5. The THD can be expressed as
X 32rms + X 52rms
THD ( % ) = ´ 100 (2.19)
X 1rms
It can be seen that DFT technique shows better performance than DWT- and LSM-based
techniques against high THD values.
2.3.6 Effect of Sampling Frequency

The number of samples per cycle increases with increment in sampling frequency. All the
phasor estimation techniques compared in this chapter uses one cycle data of current samples.
Increasing sampling frequency means the data processed at an instant also increases resulting
in increment of computation time. The effect of sampling frequencies on phasor estimation
accuracy of the techniques is studied in this section. The current signal is sampled with
frequency from 1 kHz to 30 kHz. For each sampling frequency, TVE of different techniques is
calculated and plotted in Figure 2.12. It is to be noticed that increasing sampling frequencies
results in high estimation accuracy along with burden of high computational time. Therefore
an optimal sampling frequency needs to be selected. The change in TVE value of DFT
technique with varying sampling frequencies is almost negligible.
100
DFT
90
DWT
80
LSM
70
60
TVE (%)
50
40
30
20
10
0
0 20 40 60 80 100
THD (%)
FIGURE 2.5 Effect of harmonics.

0.4
TVE (%)
0.2
(a) 0
0 5 10 15 20 25 30
Sampling Frequency (kHz)
20
TVE (%)
10
(b)
0
0 5 10 15 20 25 30
10
TVE (%)
(c) 0
0 5 10 15 20 25 30
FIGURE 2.6 Effect of sampling frequencies (a) DFT (b) DWT (c) LSM.
2.4 CONCLUSION
In this chapter, three-phasor estimation techniques based on DFT, DWT, and LSM are
discussed and compared. The criterion based on total vector error and transient monitor
have been considered to evaluate the performance of the techniques. The effects of
noise content, harmonics, and sampling frequencies are studied. The DFT-based phasor
estimation techniques is found to be more accurate and robust as compared to DWT and
LSM.
REFERENCE
1. Abbas, S. A., and Member, S., 2012. A New Fast Algorithm to Estimate Real-Time Phasors
Using Adaptive Signal Processing, 1–9.
2. Aravindhababu, P., and Balamurugan, G., 2012. ANN based online voltage estimation.
Applied Soft Computing Journal, 12, 313–319. https://doi.org/10.1016/j.asoc.2011.08.041
3. Brahma, S. M., Member, S., Kavasseri, R. G., and Member, S., 2008. Investigating the
Performance of Discrete Wavelet Transform for Phasor Estimation in Digital Relays, PES
General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century, July
2008, 1–6.
4. Jiang, J., Lin, Y., Yang, J., Too, T., and Liu, C., 2000a. An Adaptive PMU Based Fault
Detection / Location Technique for Transmission Lines — Part II: PMU Implementation
and Performance Evaluation, 15, 1136–1146.
5. Jiang, J., Yang, J., Lin, Y., Liu, C., and Ma, J., 2000b. An Adaptive PMU Based Fault Detection/
Location Technique for Transmission Lines Part I: Theory and Algorithms, 15, 486–493.
6. Khodaparast, J., and Khederzadeh, M., 2017. Least square and Kalman based methods for
dynamic phasor estimation: a review. Protection and Control of Modern Power Systems,
1–18. https://doi.org/10.1186/s41601-016-0032-y
7. Nayak, P. K., Pradhan, A. K., Member, S., and Bajpai, P., 2014. Secured Zone 3 Protection
During Stressed Condition, 1–8.
8. Phadke, A. G., Bi, T., Hlibka, T., Adamiak, M., and Jauch, J., 2018. Phasor measurement
units, WAMS, and their applications in protection and control of power systems. Journal
of Modern Power Systems and Clean Energy, 6, 619–629. https://doi.org/10.1007/s40565-
018-0423-3
9. Pradhan, A. K., Routray, A., and Basak, A., 2005. Using Least Mean Square Technique,
20, 1812–1816.
10. Ree, J. D. La, Member, S., Centeno, V., Member, S., Thorp, J. S., Fellow, L., Phadke, A. G.,
and Fellow, L., 2010. Synchronized Phasor Measurement Applications in Power Systems,
1, 20–27.
11. Sarangi, S., 2016. Enhanced Alpha Plane Line Protection.
12. Serna, D. O., Member, S., 2013. Synchrophasor Estimation Using Prony’s Method, 62,
2119–2128.
13. Si, J. L. P. De, IEEE, Senior Member, Pedro, L., IEEE, Student Member, 1990. Modal Kalman
Filtering Based Impedance Relaying, 6, 78–84.
14. Wong, C., Leong, I., Lei, C., Wu, J., and Han, Y., 2001. A Novel Algorithm for Phasor
Calculation Based on Wavelet Analysis, 00, 1500–1503.
15. Xia, Y. Q., and David, A. K., 1994. Adaptive Relay Selting For Stand-Alone Digital Distance
Protection. IEEE Transactions on Power Delivery, 9, 480–491.
CHAPTER 3
Adaptive Fuzzy PID Controller

for a Current Controlled
Multilevel Inverter Fed Matrix
Converter
G. Mohapatra, M. K. Debnath, and K. K. Mohapatra
Siksha ‘O’ Anusandhan Deemed to be University, Bhubaneswar, Odisha, India.
ABSTRACT: Multilevel inverters (MLI), when used in voltage source inverter, have
acquired wide utilization in industries. They are used, to obtain higher-output voltage
with less total harmonic reduction (THD) with an increased number of levels. The matrix
converter is used for ac to AC converter for variable frequency drives basically giving a
good power factor improvement. This chapter aims in developing a simulation model of
the matrix converter, established in Matlab Simulink with proper modulation technique.
This also deals in controlling a PV-integrated three-phase Cascaded H Bridge (CHB)
inverter, loaded with a matrix converter using the proportional resonant (PR) controller
and adaptive fuzzy PID controller. The PR controller is helping to remove the steady-state
error, and the fuzzy has the advantage of updated PI parameters. The superiority of fuzzy is
emphasised based on THD reduction as compared to PR regulator.
Keywords: Multilevel inverter, Matrix converter, Adaptive fuzzy PID.
3.1 INTRODUCTION
In the context of large power handling and controlling the grid disturbance optimized
control parameters and the use of a suitable controller is predominant in the recent era.
The enormous amount of power requirement is fulfilled by the multilevel inverter (MLI)
with an eye on the evaluation of the system performance, the fidelity of the device, and
system prudential constraints (Rodriguez, 2002). In the field of cyclo converters, the
21
matrix converters play a significant role in developing the desired output irrespective
of the input for different frequency. By employing the proper modulation technique
(Forzani, 2009), the input side power factor can be improved, which can eliminate the
requirement of power factor correction devices at the cost of an increased number of
switches. A trade-off is always desirable when selecting control between the converter
components and the conventional system (Lahlou, 2017; Escobar, 2018; Zammit, 2017;
Sefa, 2015; Teodorescu, 2006; Teodorescu, 2003). A new controlling method with
variation of the fuzzy parameter is proposed for the inverter-fed matrix converter with
the additional gain factor which can be optimised with differential evolution algorithm
(DEA) considering the THD as the objective function. Tuning of the controller by
modelling the same is customised to control the current with the PR controller, and the
fuzzy controllers maintain the load power constant. Despite conventional PI regulator, the
advantages of the PR controller and fuzzy controller for current control are discussed for
different conditions by generating disturbances from the input as well as output side and
the performances of the controllers are compared in the scale of Fast Fourier Transform
(FFT), assuming a constant PV system.
3.2 MULTILEVEL INVERTER

The ac to dc converter with distinct steps contains a cluster of IGBT and capacitor voltage
sources as in Figure 3.1 to frame a MLI.
Sa1 Sa3
Outputa V1
Sa4 Sa2
Outputn
Sn1 Sn3
Vn
Sn4 Sn2
FIGURE 3.1 Represents the cascaded H bridge inverter.

Adaptive Fuzzy PID Controller for a Current Controlled  23
U an ia daa dab dac
4-quadrant Switch
Input Filter
3-Phase Input
U bn ib dba dbb dbc
ic dca dcb dcc

U cn
UA UB UC
Output filter
Clamp Circuit
RL Load
iA iB iC
Three -phase output N
FIGURE 3.2 Represents schematic diagram of matrix converter.
3.3 SYSTEM MODELLING

The level, k, in the voltage per phase across the load of an MLI is explained by Eq. (3.1)
(Escobar, 2018; Gholizade-Narm, 2014; Alesina, 1981; Iyer, 2018). The stretch of the load
voltage U an on the basis of Fourier’s expansion is explained in Eq. (3.2).
k=2s-1 (3.1)
¥ 4Udc é cos ( hb1 ) + ..........ù
Uan ( wt ) = å ê ú sin ( hwt ) (3.2)
x=1,3,5
xp ë + cos ( hbs ) û
cos (b1 ) + ......................... + cos (bS ) = S´ M m

cos ( 5b1 ) + ............................. + cos ( 5bS ) = 0
cos ( 7b1 ) + ............................. + cos ( 7bS ) = 0 (3.3)
..............................................................................
..............................................................................
cos ( xb1 ) + cos ( xb3 ) + ....... + cos ( xbS ) = 0
Inverter’s level can be evaluated from Eq. (3.2) to achieve the suitable angle of delay (b) as
in Eq. 3.3 of the desired load potential using any iterative technique (Mohapatra, 2018;
Mohapatra, 2019). Here is the number of H bridge of the inverter, Uaz is the individual
source voltage for S number of sources, Uan is expressed as the inverter phase voltage, Udc
is the voltage of the DC capacitor, h is the number of order of harmonic, and Mm gives the
index required for modulation.
3.4 MATRIX CONVERTER

This converter, shown in Figure 3.2, is a set of four-quadrant switches connected to three
input legs to synthesize one output leg voltage. To maintain constant input current, three-
input phase connection is mandatory (Dey, 2019).
3.5 DIFFERENTIAL EVOLUTION ALGORITHM (DEA)

This is an evolutionary algorithm (Mohapatra, 2018) which is based upon the
optimization of the objective function of the harmonic content and has the following
steps:
• Initialize the number of operating parameters such as population, iteration.
• Select the target vector as the minimum THD with the objective function as
4 2 2 2
ObjF = ( V1 - SM1 ) + ( V5 ) + ( V7 ) + ...... + ( Vz ) + THD
• Calculate the values of the voltages from Eq. (3.2) for required harmonic
percentages.
• Evaluate the optimised value; if below the target, stop the operation, or else increase
the iteration by 1 with generating the trial vector by cross over and mutant vector
by mutant.
• Selection of the best vector is done if the target is reached.
• The PID parameters are updated.
3.6 PROPOSED CONTROLLER

Proportional Resonant controller A proportional resonant (PR) is a controlling system
with Hc(S) in Eq. (3.4) as the transfer function (Mohapatra, 2018), which is equivalent
to a PI controller in structure as per Eq. (3.5) and can identify the frequencies near the
resonance without disturbing the spread of the controller. The overall transfer function of
the PR controller can be given in Eq. (3.6).
s (3.4)
Hc ( s ) =K P1+K I1
s 2 +w2
s
Gh = å
h =3,5,7
K th
s 2 +( wh )
2
(3.5)
2k p1s+c p1 2k p2s+c p2 2k p3s+cp3

G PR (s)= + + +c p4 (3.6)
s 2 +w12 s 2 +w22 s 2 +w32
s 2 Lf1Cf +1 (3.7)
G F (s)=R f +
s3Lf1Lf 2Cf +s(Lf1+Lf 2 )
U dc 15%(MVAR) L
Lf di =U dc , Lf = ,Cf = 2
, R f =2x f (3.8)
dt DI*mi *fsw 2*p*fsw *U Cf
The filter parameters of a LC filter can be designed as Lf, Cf, Rf and are the filter inductance,
capacitance, and resistance with the lumped values as rL, rcf; R and L are the load components.
GF(S) is the transfer function of the filter, and Fsw is the switching frequency and the filter
coefficient as given in Eqs. (3.7) and (3.8). The complete PR controller, as given in Figure 3.3,
takes the controller parameters as set with suitable tuning done by linearizing the system. The
final block diagram of a PV-integrated three-phase MLI with a matrix converter employed
with PR controller is given in Figure 3.5 and that with fuzzy controller in Figure 3.6.
I∗ REF +
G PR (s) G F (s) G INV (s)

I ACTUAL −
FIGURE 3.3 Represents schematic diagram of PR controller.

Knowledge based if
and then
Reference Error
Output
X Fuzzifier
Inference
engine
D-Fuzzification
Process to be
controlled
FIGURE 3.4 Represents schematic diagram of fuzzy controller.
TABLE 3.1 Rule matrix table for KP and KI

Output r1 r2 r3 r4 r5 Output r1 r2 r3 r4 r5
r1 n2 n2 m1 m1 s0 s0 s0 m1 n2 n2
r2 CHANGE n2 m1 m1 s0 s0 CHANGE s0 s0 m1 n2 n2
r3 IN ERROR m1 n2 m1 m1 m1 IN ERROR m1 m1 m1 m1 m1
r4 KP s1 m1 m1 n2 n2 KI n2 n2 m1 s0 s0
r5 s1 s1 m1 m1 n2 n2 n2 m1 s0 s0
r1 Negative Large n2 Large
r2 Negative Small m1 Medium
r3 Zero s0 Small
r4 Positive Small
r5 Positive Large
Adaptive Fuzzy PID Controller The fuzzy controller, shown in Figure 3.4 (Mohapatra,
2018), has input as error and the rate of change of error. Here KP and KI are taken as
outputs. The parameters are digitalized as defined in Eqs (3.6) and (3.7).
Eri ( j )=Ila ( j )-Ilb ( j ) (3.9)
¶Eri ( j )=Eri ( j )-Eri ( j-1) (3.10)
Here Eri(j), ðEri(j), Il(a)j, Il(b)j, j are the error, change of error, reference current, actual
current, and the parameter of discretization, respectively. A rule matrix table is developed
as in Table 3.1.
3.7 SYSTEM BLOCK DIAGRAM

A five-level, three-phase photovoltaic-operated, three-phase cascaded H bridge (CHB)
inverter is developed in Matlab Simulink which is connected to a three-phase load
with a matrix converter. The overall block diagrams with the controllers are given in
Figures 3.5 and 3.6.
3φ 5 LEVEL
PV System
LCL Filter
Converter
3φ Load/
Inverter
Matrix
Grid
3φ
Ɵ
PLL
PWM
Generator Id
−
PR Ʃ abc to
Controller + dq Ɵ
dq to Iq
abc
− 3φ REF
PR Id* abc to
Ɵ Ʃ Current
Controller + dq
Iq*
FIGURE 3.5 Explains the overall system arrangement to control the load current using PR
regulator.
PV System
3 φ _ Load/
Converter
LCL Filter
Inverter
3φ _ 5L
Matrix
Grid
Ɵ
Fuzzy Logic PLL
Controller
PWM
d/dt
Generator
Id
PI − Ɵ
Controller Ʃ abc to
dq
dq to Iq
abc
CONTROLLER +
Id*
−
PI Ʃ abc 3 φ REF
Controller to dq
Ɵ Current
d/dt
+ Iq*
Fuzzy Logic
Controller
FIGURE 3.6 Explains the diagram to control the load current using adaptive fuzzy PID regulator.
3.8 RESULT AND DISCUSSION
200 Ua Ub Uc
Load Voltage (V)
100
–100
–200
0.04 0.06 0.08 0.1 0.12 0.14 0.16
Time (s)
FIGURE 3.7 Explains the load voltage of the system.
Load Current in PID Controller
Ia Ib Ic
5
Load Current (A)
–5
0.04 0.06 0.08 0.1 0.12 0.14 0.16

Time (s)
FIGURE 3.8 Explains the load current with PID controller in loading of the system.
Load Current in Fuzzy Controller

5
Ic Ib Ia
Load Current (A)
–5
0.04 0.06 0.08 0.1 0.12 0.14 0.16
Time (S)
FIGURE 3.9 Gives the load current with fuzzy PID controller in loading of the system.
Load voltage in volts with PR controller

Ua Ub Uc
100
–100
0.04 0.06 0.08 0.1 0.12 0.14
Load voltage in volts with fuzzy controller

Ua Ub Uc
100
–100
0.04 0.06 0.08 0.1 0.12 0.14

Time (s)
FIGURE 3.10 Explains the load voltage of the system with PR and fuzzy PID controller with
voltage sag.
Load current in Amp with PR controller

10
Ia Ib Ic
Load Current (A)
–5
–10
0.04 0.06 0.08 0.1 0.12 0.14
Time (s)
FIGURE 3.11 Explains the load current of the system with PR controller with voltage sag.
Load current in Amp with fuzzy controller
10
Ia Ib Ic
5
Load Current (A)
–5
–10
0.04 0.06 0.08 0.1 0.12 0.14
FIGURE 3.12 Explains the load current of the system with fuzzy controller with voltage sag.
18
with PR CONTROLLER
16
with FUZZY PID CONTROLLER
14
12
10
PERCENTAGE OF THD
(expressed in pu)
0
0 5 10 15 20 25 30
HARMONIC ORDER
FIGURE 3.13 Explains the THD variation of different controllers.
3.9 RESULT ANALYSIS

A PV-integrated five-level system is developed with a voltage and load variation. Different
controllers are analysed based on the harmonic content when the system is integrated with
a matrix converter. The load current is forced to be stabilized with a set of controllers in
Figures 3.8, 3.9, 3.11 and 3.12. A voltage variation of 0.8 pu is introduced for 0.06s–0.12s
and is analysed in Figure 3.10 for voltage stability.
3.10 CONCLUSION
A Matlab-simulated system of a five-level CHB is developed with a PV input. A matrix
converter is connected and coupled with the load. The load current is stabilized with fuzzy
controller in a better manner as compared to PR controller which is observed for different
loading conditions. Voltage sag of 0.8 pu is inserted from the input side for a time interval
of 0.06 s to 0.12 s, and the response of the system is analysed with PR and fuzzy controller is
observed. The comparison of THD as described in the percentages of fundamental explains
that the fuzzy PID is giving a better result as compared to PR.
APPENDIX
PV voltage 70 V, Matrix output voltage 120 V, Load voltage 100 V, Reference current set 5A,
Load current <5A, Km 0.5773, Km 10–0.2222, Input Filter 6 mH, 1 micro F, Output Filter
3mH, 10 micro F, Kvgain 0.866, Km 10–0.2222.
REFERENCES
1. Rodriguez, J., Lai, J. S., and Peng, F. Z., 2002. Multilevel inverters: a survey of topologies,
controls, and applications. IEEE Transactions on industrial electronics, 49(4), 724–738.
2. Farzaneh, A., Si, J. L. P. De, IEEE, Senior Member, Pedro, L., IEEE, Student Member
Nazarzadeh, J., 2009. Precise loss calculation in cascaded multilevel inverters. In 2009
Second International Conference on Computer and Electrical Engineering, 2, 563–568.
3. Lahlou, T., Malik, S. N., and Herzog, H. G., 2017. Simulation and analysis of control
and modulation schemes for a multilevel cascaded H-bridge inverter for use in a battery
energy storage system. In 2017 18th International Scientific Conference on Electric Power
Engineering (EPE), 1–6.
4. Escobar, G., Martinez‐Rodriguez, P. R., Ho, C. N., and Sosa, J. M., 2018. Design of an
inverter‐side current reference and controller for a single‐phase LCL‐based grid‐connected
inverter. International Transactions on Electrical Energy Systems, 28(1), e2476.
5. Zammit, D., Staines, C. S., Apap, M., and Licari, J., 2017. Design of PR current control
with selective harmonic compensators using Matlab. Journal of Electrical Systems and
Information Technology, 4(3), 347–358.
6. Sefa, I., Altin, N., Ozdemir, S., and Kaplan, O., 2015. Fuzzy PI controlled inverter for grid
interactive renewable energy systems. IET Renewable Power Generation, 9(7), 729–738.
7. Teodorescu, R., Blaabjerg, F., and Liserre, M., 2004. Proportional-resonant controllers. A
new breed of controllers suitable for grid-connected voltage-source converters. Proceedings
Optim, 3, 9–14.
8. Teodorescu, R., Blaabjerg, F., Liserre, M., and Dell’Aquila, A., 2003. A stable three-
phase LCL-filter based active rectifier without damping. In 38th IAS Annual Meeting on
Conference Record of the Industry Applications Conference, 3, 1552–1557.
9. Escobar, G., Martinez‐Rodriguez, P. R., Ho, C. N., and Sosa, J. M., 2018. Design of an
inverter‐side current reference and controller for a single‐phase LCL‐based grid‐connected
inverter. International Transactions on Electrical Energy Systems, 28(1), e2476.
10. Gholizade-Narm, H., 2014. A novel control strategy for a single-phase grid-connected
power injection system. International Journal of Engineering, 27(12), 1841–1849.
11. Alesina, A., and Venturini, M., 1981. Solid-state power conversion: A Fourier analysis
approach to generalized transformer synthesis. IEEE transactions on circuits and systems,
28(4), 319–330.
12. Iyer, N. P., 2018. A dual programmable AC to DC rectifier using three-phase matrix
converter topology—analysis aspects. Electrical Engineering, 100(2), 1183–1194.
13. Mohapatra, G., and Nayak, M. R., 2018. Switching angle and power loss calculation for
THD minimization in CHB-multilevel inverter using DEA. In Information and Decision
Sciences, Springer, 491–502.
14. Mohapatra, G., 2018. Current control of a PV integrated CHB-multilevel inverter using PR
controller. In 2018 Technologies for Smart-City Energy Security and Power (ICSESP), 1–6.
15. Mohapatra, G., Dey, A. K., and Sharma, R., 2019. Comparison of controller in a three phase
CHB Inverter. In 2019 IEEE International Conference on Sustainable Energy Technologies
(ICSET), 1–6.
16. Dey, A. K., Mohapatra, G., Mohapatra, T. K., and Sharma, R., 2019. A Modified Venturini
PWM scheme for Matrix converters. In 2019 IEEE International Conference on Sustainable
Energy Technologies (ICSET), 013–018.
CHAPTER 4
Multi-Objective Optimization
for Enhanced Surface Roughness
and Microhardness of AA 6063/
TiO2 Cast Aluminium Metal Matrix
Nanocomposites Fabricated by
Friction Stir Processing
S. Laxmanaraju,1 Lingaraju Dumpala,2 and R. Pavankumar3
1
Department of Mechanical Engineering, Sri Sivani College of Engineering, Srikakulam, Andhra Pradesh,
India.
2
Department of Mechanical Engineering JNTUK, Kakinada, Andhra Pradesh, India.
3
Department of Mechanical Engineering, Sri Sivani College of Engineering, Srikakulam, Andhra Pradesh,
India.
1
laxman.raju@yahoo.co.in, 2dlraju@gmail.com, 3pavan5103@gmail.com
ABSTRACT: In this chapter, friction stir processing (FSP) is used as an effective technique
to improve the strength and hardness of various AA6063 aluminium alloys so as to alter the
nanocomposite by using TiO2. This FSP process has also been used to impart superplasticity
for eliminating casting defects. The FSP process parameters are tools with different rotation
speeds of 1050, 1200, 1450 rpm; tool transverse feed of 30, 45, 60 mm/min; tool pin profile
shapes of conical, rectangular, and threaded forms of the nano-reinforced particles that are
strategically varied during the FSP process. The FSP experiments are conducted by using
L9 Taguchi-based design of experiment to fabricate cast nanocomposite. Subsequently,
the nanocomposites fabricated through FSP undergo few mechanical tests to evaluate
mechanical metallurgical properties such as microhardness, tensile strength, and surface
roughness. Multi-objective model using Grey relational analysis is proposed to solve the
problem. Analysis of variance is used to determine the percentage of contribution of
the given process parameters, and the tool rotation speed is high influenced parameter,
31
threaded tool profile is second influenced parameter, and finally, tool feed rate is the low
influenced parameter.
Keywords: Friction stir processing, Mechanical properties, Grey relational analysis,

ANOVA.
4.1 INTRODUCTION
Friction stir processing (FSP) is a technique of changing and upgrading the properties of
a metal through intense, localized plastic deformation. This deformation is produced by
forcibly inserting a non-consumable tool into the workpiece (Figure 4.1). This process mixes
the material without changing the phase and creates a microstructure with fine grains. It is
a solid-state process in which the material is in the solid state at the time of processing. The
contact between the rotating tool and the material generates heat which softens the material
below the melting point of the material. FSP offers an excellent choice for surface modification
and development of surface as well as bulk composites of metal alloys (Mishra and Ma, 2005).
Manufacturing industries such as aircraft, automobile, marine impose the need for
optimizing product designs using lightweight materials such as alloys of aluminium,
magnesium, titanium. Because of increasing lightweight requirements especially these alloys
are quite propitious for various applications owing to their reduced weight concentration,
enhanced strength-to-weight ratio, and higher corrosive resistance. However, strength of
these alloys is not adequate for some structural purposes thereby necessitating requirement
of reinforcement. After reinforcement, these metal matrix composites (MMCs) are
found to exhibit improved mechanical, metallurgical, and tribological characteristics
(Puviyarasan and Senthil Kumar, 2016). The effect of tool pin and shoulder profiles
on particle distribution and microstructure of 6063 aluminium alloy reinforced with
SiC particles was investigated. There are eight tool pin profiles used: a tapered cylindrical
pin, plain cylindrical pin, triangle pin, square pin, clockwise scroll and anti-clockwise
scroll with cylindrical pin, clockwise scroll with square pin, and anti-clockwise scroll
with square pin. Anti-clockwise scroll with square pin and cylindrical pin will give high
reinforced zone area compared to clockwise. Hence, anti-clockwise scrolled shoulder
with cylindrical pin tool geometry was found best among the other tool pin with
uniform distribution of particles. There are no defects that can be found in this process
(Gangil, Maheshwari, and Siddiquee, 2018). The fabrication of aluminide-reinforced
nanocomposites such as Al-Al2Cu, Al-Al3Ti, and Al-Al3Fe2 by this method has recently been
shown. In situ formation of Al3Ti particles during FSP of Al–Ti and Al–TiO2 systems are
very attractive because the nanocomposites demonstrate high specific strength and modulus
at ambient and elevated temperatures (Zhang, Xiao, Wang, and Ma, 2011; 2012; Cavaliere,
2005). An investigation to fabricate a composite by FSP of AA6061/TiB2 aluminum alloy
was carried out. It was observed that tensile strength and hardness were augmented with
increase of rotational speed. The analysis of variance (ANOVA) results indicate that rotational
speed was found to be the most influential process parameter. The optimal speed for ultimate
Multi-Objective Optimization for Enhanced Surface Roughness  33
tensile strength (UTS) and microhardness is 1400 rpm (Venkateswarlu, Davidson, and
Sammaiah, 2014). The effects of the process parameters such as tilt angle, rotation speed,
and traverse speed on the mechanical properties such as tensile strength, microhardness, and
impact energy of the AZ31B Mg alloy were analysed. Experiments have been carried out
with Taguchi L9 orthogonal Array that consists of nine combinations of rotational speeds,
traverse speed, and tilt angle. Traverse speed has the strongest influence on the impact energy
(Abraham, Isaac, Selvam, and Akinlabi, 2019). Experiments are conducted on the three-level
process parameters – welding speed, tool speed, and tilt angle – as per Taguchi’s L9 orthogonal
array. The mechanical properties of the weld specimens studied are % elongation and Vickers
hardness (Prasad, Durga, and Namala, 2018).
The current study uses TiO2 nanoparticles as reinforcement particles with the AA 6063
aluminium alloy in order to fabricate a surface composite metal matrix using single-pass
FSP. Moreover, this work focuses on the nanocomposite matrix fabricated by using three
parameters, and three levels are selected and generated by L9 matrix using MINITAB-17
statistical software. The process parameters were optimized by using Taguchi-based
Grey relational analysis (GRA) for the multi-objective response characteristics that are
microhardness, tensile strength, surface roughness. The optimal process parameters are
predicted and validated for the model by conducting confirmation experiments.
4.2 EXPERIMENTATION
Aluminium alloy of AA6063 plates prepared have a 6-mm thickness, 60-mm width, and
250-mm length. Table 4.1 represents the chemical composition of the aluminium alloy
plates.
TABLE 4.1 Chemical composition of AA6063
Elements Mg Si Fe Mn Cu Cr Zn Zr Ti Al
Wt % 0.45 0.35 0.15 0.02 0.010 0.09 0.04 0.02 0.02 balance
Figure 4.2 shows 2 mm diameter and 2 mm depth circular holes along the plate surface
which are drilled and take the 5 mm equal interval gaps throughout the length of the plates.
The circular holes are stuffed with TiO2 nanoparticles. FSP is conducted by using a FSW-3T
machine as shown in Figure 4.3.
In the FSP, various process parameters are effected mainly tool pin profiles,
tool rotation speeds, and tool transverse speeds. The process parameters are arranged in
Table 4.2. Experimentation is conducted by using those process parameters. Before going
to the experimentation, design of experimentation is prepared by using Taguchi method in
MINITAB-17 statistical software.
The process is initiated by drilling a hole in the workpiece, as it allows the pin to easily
penetrate the workpiece as there is not enough heat generated at the beginning of the process
to make the material soft. The forces generated during the entire process are recorded using
the data acquisition system. Tensile test is conducted by using a computerized tensile tester
(Instron 1195).
The weld pieces are cut according to ASTM E8M-04 standard code on a wire-
EDM cutting machine. The tensile specimens are shown in Figure 4.4. Before going to
the specimen preparations, the surface roughness test is conducted by using Taly-surf.
Microhardness values are measured by using Vickers hardness tester at the load of 300
g and dwell time 10 sec. The UTS, microhardness, and surface roughness test results are
shown in Table 4.3.
FIGURE 4.1 Workpiece dimensioning. FIGURE 4.2 Workpiece AA6063 stuffed with TiO2.
FIGURE 4.3 Experimental setup.

TABLE 4.2 Process parameters
Parameter LEVELS
Process parameter
designation Level 1 Level 2 Level 3
N Tool rotation speed N rpm 1050 1200 1450
F Tool transverse feed F mm/min 30 45 60
T Type of pin profile T Conical Rectangular Threaded
TABLE 4.3 UTS, microhardness, surface roughness results
Feed F mm/ Tool pin profile

S. No Speed N Rpm UTS Mpa HV Ra µm
min T
1 1050 30 Cone 272.56 96.33 0.37
2 1050 45 Rec 273.89 92.00 0.73
3 1050 60 Thread 282.56 96.67 0.27
4 1200 30 Rec 285.56 99.67 0.34
5 1200 45 Thread 278.56 105.33 0.21
6 1200 60 Cone 275.00 93.00 0.63
7 1450 30 Thread 296.00 113.00 0.16
8 1450 45 Cone 285.56 104.67 0.33
9 1450 60 Rec 289.56 102.33 0.36
4.3 RESULTS AND DISCUSSIONS

Grey relational analysis (GRA) is used to find out the multi-response characteristics that are
analysed into a single response. In GRA, first thing to do is that the results are normalized
in between zero and one. The second step is calculating the grey relational coefficient, and
finally, the grey relational grade is calculated in which maximum value gets optimum value
and gives the rankings. Based on the grey grade, the highest value of grade was ranked as 1
for experiment 10 and the mean response of the parameters was calculated and tabulated
in Table 4.5. The highest value of mean response of parameters and levels was the optimal
parameter.
The data sequence for tensile strength and microhardness, which is larger-the-
better performance characteristics are UTS, and microhardness is normalized as follows
(Rao and Ramanaiah, 2017):
xio (k ) - minxio ( k )
x*i (k ) =
max xio ( k ) - minxio ( k )
where k = 1–n, i = 1–9, n is the performance characteristic, and i is the trial number.
The input power, which is lower-the-better performance characteristic in the surface
roughness conditions, is normalized as follows:
max xio (k ) - xio (k )

x*i ( k ) =
max xio ( k ) - minxio ( k )
TABLE 4.4 Normalized values and deviational sequence of response characteristics

Normalized Values of Response Deviation Sequence
S. No UTS HV Ra UTS HV Ra
1 0.00 0.21 0.63 1.00 0.79 0.37

2 0.06 0.00 0.00 0.94 1.00 1.00
3 0.43 0.22 0.82 0.57 0.78 0.18
4 0.55 0.37 0.68 0.45 0.63 0.32
5 0.26 0.63 0.91 0.74 0.37 0.09
6 0.10 0.05 0.18 0.90 0.95 0.82
7 1.00 1.00 1.00 0.00 0.00 0.00
8 0.55 0.60 0.70 0.45 0.40 0.30
9 0.73 0.49 0.65 0.27 0.51 0.35
where x*i (k) is the value after grey relational generation, min x*i (k) is the smallest value of
x*i (k), and max x0i (k) is the largest value of x0i (k). The grey relation coefficient xi (k ) can be
calculated as follows:
D min + zD max
xi ( k ) =
D oi (k ) + zD max
Note that higher is better and is achieved when x*i (k) = x0i (k), that is when X= reference.
Here x*0 (k) denotes the reference sequence, x*j (k) denotes the comparability sequence,
z [0–1] is the distinguishing coefficient; 0.5 is widely accepted, D 0i = x0 ( k ) - xi ( k ) is the
* 0
difference in absolute value between x*0 (k) and x*j (k), D min = minÑjÎk , minÑjÎk x0 ( k ) - xi ( k )
* 0
is the smallest value of D 0i , D max = maxÑjÎk , maxÑjÎk x0* ( k ) - xi0 ( k ) is the largest value
of D 0i .
n
1
Grey relational grade:
n åx (k ) =n
i =1
i i
TABLE 4.5 The grey relational grade

Grey Relational Coefficient
S. No UTS HV Ra Rank
1 0.33 0.39 0.57 7
2 0.35 0.33 0.33 9
3 0.47 0.39 0.74 5
4 0.53 0.44 0.61 6
5 0.40 0.58 0.85 2
6 0.36 0.34 0.38 8
7 1.00 1.00 1.00 1
8 0.53 0.56 0.62 4
9 0.65 0.50 0.59 3
TABLE 4.6 Response table for means

Speed N Feed F
Level Toolpin Profile T
RPM mm/min
1 0.4334 0.6527 0.4539
2 0.4992 0.5062 0.4801
3 0.7154 0.4890 0.7139
Delta 0.2819 0.1637 0.2600
Rank 1 3 2
Main Effects Plot for Means

Data Means
Speed N Feed F Tool pin profile T
0.75
0.70
0.65
Mean of Means
0.60
0.55
0.50
0.45
0.40
1050 1200 1450 30 45 60 Cone Rec Thread
FIGURE 4.4 Graph mean values of GRG.
Based on the grey grade, the highest value of grade was ranked 1 for Experiment 7, and
mean response of the parameters was calculated and tabulated in Table 4.6. The highest
value of mean response of parameters and levels was the optimal parameter. The optimal
process parameters for UTS, microhardness, and surface roughness are tool speed N3 (1450
rpm), tool transverse feed F1 (30 m/min), and tool pin profile (3T-threaded). It has been
found from Table 4.6, and the graph shows the tool rotation speed ranks first, high speed in
the contribution of responses, the tool pin profile ranks second, and the threaded tool pin
profile in the contribution of response, tool transverse speed ranks third, and the low tool
transverse feed rate.
The confirmation tests are conducted, predicted GRG optimum = N3 + F1 + T3 – 2T =
0.7154 + 0.6527 + 0.7139 – 2*0.54931 = 0.9833 The experimental value of GRG is 1 and the
variation from the optimum result is 0.0166.
TABLE 4.7 ANOVA process
Source DF Adj SS Adj–MS F- value P- value Contribution %
Speed N rpm 2 0.130555 0.065277 103.50 0.010 43.03

Feed F mm/min 2 0.048558 0.024279 38.49 0.025 16
Tool pin Profile T 2 0.122976 0.061488 97.49 0.010 40.5
Error 2 0.001261 0.000631 0.41
Total 8 0.303350
TABLE 4.8 Model summary

S R-sq R-sq(adj) R-sq(pred)
0.0251141 99.58% 98.34% 91.58%
ANOVA has been determined to investigate the importance of input process parameters.
The contribution percentage of each process parameter on the total variation was estimated
with ratio of adj-SS of corresponding parameter to total adj-SS. The F-test value and
P-test value were used to determine which process has a significant effect on output
responses. Usually, the change of the process parameter has a significant effect on the
quality characteristics, when P value is smaller and F value is large. The results of ANOVA
were presented in Table 4.7. The speed has the highest contribution 43.03% in the output
response for UTS, microhardness, and surface roughness, whereas tool transverse feed
has 16 percentage of contribution and the second highest is the tool profile, that is, 40.5
percentage of contribution, respectively.
4.4 CONCLUSION
The AA 6063/TiO2 metal nanocomposite was successfully processed by using FSP.
The mechanical properties UTS, microhardness, and surface-roughness are evaluated.
The process parameters were optimized towards tensile strength, microhardness, and
surface roughness using Taguchi technique and GRA. The optimal process parameters
were estimated and confirmation experiments were conducted and lies within the 10%
predicted value. The ANOVA results show that the tool rotational speed was found
to be the most influential process parameter with 43.03% followed by tool pin profile
(40.5%), tool transverse speed (16%), which are highly important. The mean value
shows higher tool rotational speed, lower tool transverse speeds, and threaded tool pin
profile gives the better mechanical properties.
REFERENCES
1. Mishra, R. S., and Ma, Z. Y., 2005. Friction stir welding and Processing. Material science
and Engineering. 250, 1–78.
2. Puviyarasan, M., and Senthil Kumar, V. S., 2016. ‘Microstructural Evolution and Mechanical
Behaviour of AA6063/SiC p Bulk Composites Fabricated Using Friction Stir Processing
Microstructural Evolution and Mechanical Behaviour of AA6063/SiC p Bulk Composites
Fabricated Using Friction Stir Processing*’., 4846(February)
3. Gangil, Namrata, Sachin Maheshwari, and Arshad Noor Siddiquee., 2018. ‘Influence of
Tool Pin and Shoulder Geometries on Microstructure of Friction Stir Processed AA6063/
SiC Composites’. Mechanics & Industry 19(2), 211.
4. Rao, Dakarapu Santha, and Nallu Ramanaiah., 2017. ‘Process Parameters Optimization for
Producing AA6061/TiB2 Composites by Friction Stir Processing’. Strojnicky Casopis 67(1),
101–17.
5. Venkateswarlu, Ganta, Davidson, M. Joseph, and Sammaiah, Pulla., 2014. ‘Effect of Friction
Stir Processing Process Parameters on the Mechanical Properties of AZ31B Mg Alloy’.
Manufacturing and Industrial Engineering 13(1–2), 1–5.
6. Abraham, Sahayam Joyson, Isaac Dinaharan, Jebaraj David Raja Selvam, and Esther Titilayo
Akinlabi., 2019. ‘Microstructural Characterization and Tensile Behavior of Rutile (TiO2)-
Reinforced AA6063 Aluminum Matrix Composites Prepared by Friction Stir Processing’.
Acta Metallurgica Sinica (English Letters) 32(1), 52–62.
7. Shivananda Murthy, K. V. et al., 2017. ‘Mechanical and Thermal Properties of AA7075/
TiO2/Fly Ash Hybrid Composites Obtained by Hot Forging’. Progress in Natural Science:
Materials International 27(4), 474–81.
8. Qian, J., Li, J., Xiong, J., Zhang, F., and Lin, X. In situ synthesizing Al3Ni for fabrication of
intermetallic reinforced aluminium alloy composites by friction stir processing. Materials
Science and Engineering: A, 2012, A550, 279–285.
9. Zhang, Q., Xiao, B. L., Wang, W. G., and Ma, Z. Y., 2012. Reactive mechanism and
mechanical properties of in situ compo-sites fabricated from an Al-TiO2 system by friction
stir processing, Acta Mater. 60, 7090–7103
10. Zhang, Q., Xiao, B. L., Wang, Q. Z., and Ma, Z. Y., 2011. ‘In Situ Al3Ti and Al2O3Nanoparticles
Reinforced Al Composites Produced by Friction Stir Processing in an Al-TiO2 System’.
Materials Letters 65(13), 2070–72.
11. Cavaliere, P., 2005, ‘Mechanical Properties of Friction Stir Processed 2618/Al2O3 /20p
Metal Matrix Composite’, Composites Part A: Applied Science and Manufacturing, 36,
1657–1665.
12. Prasad, M. V. R. Durga, and Namala, Kiran Kumar, 2018. ‘Process Parameters Optimization
in Friction Stir Welding by ANOVA’. Materials Today: Proceedings 5(2), 4824–31.
CHAPTER 5
Self-Adaptive Multivariate
q-Gaussian-based Capacitor
Placement for Reactive Power
Compensation in Distribution
Feeders
Debashis Jana and Ranjita Chowdhury
Institute of Engineering & Management, Kolkata, India.
ABSTRACT: Reactive power compensation at per level is one of the fundamental

operational issues of power systems operation and control. To solve the reactive power
compensation problem, the conventional real coded genetic algorithm (RCGA) is
customized by the concept of self-adaptation of multivariate q-Gaussian distribution. This is
followed by a polynomial mutation approach with arithmetic crossover for optimal control
of reactive power in distribution feeders to evade premature convergence with fast rate of
convergence. The improved technique has been added to the conventional RCGA. A new
combination of crossover and mutation technique has been adopted to get a better route
for escaping the local minimum solution. The performance of the algorithm is verified on a
sample test system and the better simulation results turn out improved solutions compared
to different methods.
Keywords: Reactive power compensation, Power loss, Capacitor placement, Modified

real coded genetic algorithm (MRCGA), Self-adaptation, Multivariate q-Gaussian
distribution.
5.1 INTRODUCTION
Optimal reactive power compensation has important utility functions for economically
operated distribution systems. For reactive power compensation in the distribution
41
system, it is been widespread practice to use capacitors. They are used to trim down
power losses, to improve power factor, and to sustain voltage profile inside tolerable
limits. The voltage profiles throughout the electric power system network have to be
maintained at acceptable levels to make sure network reliability. The optimal location,
optimal size of the capacitors, and associated economic cost are the most vital governing
segments of reactive power compensation. In this study main intention lies in optimal
reactive power planning which leads to minimize system losses while gratifying various
operating constraints under an assured load model.
The finest reactive power compensation is a multifaceted combinatorial optimization
problem, and more than a few optimization methods and algorithms have been
implemented over the years. To solve the problem, some of the methods included were
heuristic numerical algorithm (Baghzouz and Ertem, 1990), mixed integer programming
technique (Baran and Wu, 1989), fuzzy-reasoning method (Su and Tasi, 1996), nonlinear
evolutionary programming approach (Lai and Ma, 1997), and an algorithm based on
particle swarm optimization technique that was proposed by Yu et al. (2004) for capacitor
placement considering harmonic distortion, reactive power compensation using the self-
adaptive particle swarm optimization (Mandal and Jana, 2012). A new technique based
on plant growth algorithm was proposed by Wang et al. (2008). Chiou et al. (2006) used
variable scale differential evolution method to discover optimal solution for capacitor
placement problem in bulky distribution systems.
Realizing the adaptability of real coded genetic algorithm (RCGA), an attempt has
been made in this chapter to solve reactive power problems based on modified real coded
genetic algorithm (MRCGA). MRCGA enhances the multivariate q-Gaussian self-adaptive
searching of most fitted chromosome for improved solution which will not be fascinated
by local minima. The anticipated metaheuristic method is implemented on a sample test
system to reveal its effectiveness.
5.2 PROBLEM FORMULATION

The key point of optimal reactive power compensation is to minimize the total annual
cost of the system while fulfilling some operating constraints under a certain load
outline. The mathematical model of optimal reactive power compensation can be
produced as follows:
min F = min (Cost) (5.1)
where Cost comprises of power loss and capacitor placement. The voltage level at each bus
must be retained within its limits and is expressed as
Vmin £ Vi £ Vmax (5.2)
where, |Vi| is the voltage magnitude of ith bus, Vmin and Vmax are the minimum and maximum
bus voltage limits, respectively.
Self-Adaptive Multivariate q-Gaussian-based Capacitor Placement for Reactive  43
A number of simplified feeder-line flow equations are assumed for ease. Considering
the one-line diagram shown in Figure 5.1, the following set of equations may be used for
power flow calculation (Su and Tasi, 1996).
(
é Pi 2 + Qi2
Pi +1 = Pi - PLi +1 - Ri ,i +1 ê
) ùú (5.3)
ê Vi 2 ú
ë û
(
é Pi 2 + Qi2
Qi +1 = Qi - QLi +1 - X i ,i +1. ê
) ùú (5.4)
ê Vi 2 ú
ë û
(P 2
+ Qi2 )
2 2
(
Vi +1 = Vi - 2 ( Ri ,i +1.Pi + X i ,i +1.Qi ) + Ri2,i +1 + X i2,i +1 ) i
Vi
2
(5.5)
where Pi and Qi are the real and reactive powers flowing out of ith bus, respectively. PLi, QLi
are the real and reactive load powers at the ith bus, respectively. The resistance and reactance
of the line section between buses i and i+1 are denoted by Ri, i+1 and Xi, i+1, respectively.
FIGURE 5.1 Single-line diagram of a main feeder.
The power loss of the line section connecting buses i and i + 1 can be calculated as follows:
Pi 2 + Qi2
PLoss ( i, i + 1) = Ri ,i +1. 2
(5.6)
Vi
The total power loss of the feeder PT, Loss may then be determined by summing up the
losses of all line sections of the feeder. The loss is given by
n -1
PT , Loss = å P ( i, i + 1)
i =0
Loss (5.7)
The objective of insertion of compensating capacitor along distribution feeders is to limit

the total power loss and keep the bus voltages within their specified limits while minimizing
the total operating cost. In view of the practical capacitors, it may be found as a finite
number of regular sizes which are integer multiples of the smallest size Q0c . In common
cases the capacitors of larger size generally have lower unit prices. The available capacitor
size is usually limited to
Qcmax = LQ0c (5.8)

where L is an integer. Therefore, for each installation location, there are L capacitor
{ } { }
sizes Q0c , 2Q0c ,........., LQ0c available. Let K1c , K 2c ,............, K Lc be their corresponding
equivalent annual cost per kVAr. Therefore, the function for annual cost due to capacitor
sitting and power loss may be established as
n
Cost = K P PT , Loss + åK Q
i =1
i
c
i
c
(5.9)
where KP is the equivalent annual cost per unit of power loss in $/ (kW–year) and
i = 1, 2,..., n are the indices of the preferred buses for compensation. The bus reactive
compensation power is limited to
n
Qic £ åQ
i =1
Li
(5.10)
RCGA works directly with continuous variables instead of discrete variables. Initially,
the population of random sets of points has been created and a reproduction operator,
that is, roulette wheel or other selection operators, may be used to pick good strings in
the population. A number of new and efficient crossover operators have been intended
so that search along variables is also promising. A number of combinations of crossover
and mutation techniques are used in RCGA. In this study to optimize objective function,
RCGA has been implemented using the new improved blend of arithmetic crossover and
polynomial mutation and it is discussed elaborately afterward.
The application of real-valued illustration in the GA recommends a number of
advantages in numerical function optimization over binary encoding. Besides competent
floating-point representations or losslessness in precision by discretisation to binary
or other values, there is greater freedom to use of different genetic operators efficiently
also. Thus the crossover operator has remarkable impact on genetic algorithm recital. It
is obvious that the production of new offspring is done by the mating of selected pair of
parents for crossover
Child 1 = v * father + (1 - v ) * mother
Child 2 = (1 - v ) * father + v * mother (5.11)
where ‘v’ is a weighting factor which is selected by an user randomly in the range of
[0, 1]. The two chromosome vectors are chosen randomly for crossover. The Cinew_genx
and Cjnew_genx will breed two offspring, Ci and Cj which is considered to be a linear
combination of their parents.
Ci = v * Cinew _ genx + (1 - v ) * C new
j
_ genx
C j = (1 - v ) * Cinew _ genx + v * C new _ genx (5.12)

j
where, u = (1 + 2 * a ) * rand - a , here α is chosen as 0.5.

The mutation operator is engaged to impart new genetic properties into the
population. Though, the mutation is typically known as a secondary operator which may
not be trusted upon for reaching the objective in the search space. But the proposed
mutation operation is being aimed to put in a particular combination to find the effective
output for a complete search space. Currently in the real coded GA, the polynomial
mutation (Hogan, 2004) has been used satisfactorily. A solution is generally produced
in the m-dimensional real-valued search space by the mutation operator from the
®
individual x i , where i = 1,., μ, as follows:
® ® ®
xi = xi + c z
where c is the matrix that explains the mutation potency in each coordinate j = 1,., m and
®
z is an m-dimensional random vector created from a provided multivariate distribution.
At this point, an m-dimensional random vector is formed from the multivariate Gaussian
distribution (Su, Chan and Chiou, 2005). We proposed®to model®
the random
® mutation
®
vector z from an isotropic q-Gaussian distribution as z ~ r u n , where r is a uniform
random vector configured by sampling ®
a random vector with Gaussian distribution. The
random vector may be denoted by z ~ c q m , while a random vector of m-dimension
can be generated by sampling m self-determining q-Gaussian arbitrary variables xq(0,1).
In q-Gaussian distribution self-adaption (Su, Chan and Chiou, 2005) is made of by the
parameter q which illuminates the shape of the distribution. The computation has been
done for muted child is as follows:
(
Pgi = Pgi + d Pgimax - Pgimin ) (5.13)
A random number un is generated between 0 and 1 and a calculation is done for the variable, δ
ì éêë2un +(1- 2un )(1-j )hm+1 ùúû 1 (hm+1) -1 if un £ 0.5

ï
d =í 1 (h m+1)
(5.14)
ï1- éëê2(1-un ) + 2( un -0.5)(1-j )(hm+1) ùûú , otherwise
î
where computation is made for f as below
(
j = min é Pgi - Pgi
ë
min
) , ( Pgimax - Pgi )ùû ( Pgimax - Pgimin )
and the parameter ηm indicates the distribution index for the proposed mutation and
obtains any nonnegative value. The perturbance may be attuned by changing ηm and pm
with generations in the expression mentioned earlier.
pm = 1 N + ( gen genmax ) (1 - 1 N )
h m = h mmin + gen,
where pm is recognized as the probability of mutation, ηmmin is the minimum value for
ηm and N = NG–1 is the maximum number of decision variables in the solution vector.
The proposed result also clarifies that no solution would be shaped outside the range of
Pgimin and Pgimax.
5.3 DISCUSSION
The proposed algorithm was implemented using Matlab 7.8.0 (R2009a) on a PC
(Core 2 duo, RAM 2 GB, 2.10 GHz). In the computation, population size and maximum
number of generation have been selected as 100 and 300, respectively, with crossover
probability of 0.8 and the value of q in the interval of 1 < q < 3.
To exhibit the effectiveness and viability of the proposed algorithm, it was applied
on a model test system. The test system (Lai and Ma, 1997; Su, Chan and Chiou, 2005)
under consideration consists of a 23 kV, 9 section feeder. The equivalent cost per unit
of power loss estimated for the present problem is $168/ (kW-year) (Lai and Ma, 1997;
Su, Chan and Chiou, 2005). The three-phase load information, feeder impedance data,
available capacitor sizes, and other data are from (Lai and Ma, 1997; Su, Chan and
Chiou, 2005), and they are not reproduced here. The limits which are selected for bus
voltages are as follows:
Vmin = 0.90 p.u.
Vmax = 1.10 p.u.
It is primarily considered that all the network buses were accessible for compensation. The
annual costs for proposed planning, system power loss for both the cases, that is, before
and after compensation, capacitor introduction at the desired place are shown in Table 5.1.
TABLE 5.1 Results for Voltage Profile, Capacitors, Net Power Loss and Annual Cost
Bus No. Uncompensated Voltage (p.u.) Placed (Qc) (kVar) Compensated Voltage (p.u.)
0 1 0 1.0000
1 0.9929 0 0.9998
2 0.9874 3600 1.0044
3 0.9634 1350 0.9932
4 0.9619 1500 0.9830
5 0.9480 600 0.9617
6 0.9072 600 0.9550
7 0.8890 300 0.9545
8 0.8587 450 0.9177
9 0.8375 450 0.9010
Total cap. size (Mvar) 8.850
Total loss (MW) 0.7857 0.6731
Annual cost in ($/year) 131,684 114,820
CPU time (sec) 97.18
Table 5.3 reveals that the voltage profile for the referred buses are within the system limits
after compensation. The annual cost for this arrangement is $114,808, while the system
power loss has been calculated as 0.6729 MW, but in comparison with uncompensated
situation, the annual expenditure is $131,675 and the loss is 0.7836 MW. The computation
takes time of 90.12 sec.
FIGURE 5.2 Convergence characteristics for annual cost.
Figure 5.2 shows the optimal annual cost convergence characteristics curve. It also
compares the convergence characteristics of other population-based metaheuristic
algorithm which are listed in Table 5.2.
TABLE 5.2 Comparison of Results with Different Methods
Fuzzy DE SGA ACSA Proposed
Reasoning Method
Total Loss (MW) 0.7048 0.6763 0.6766 0.6753 0.6731
Annual cost in ($/year) 119,420 115,471 115,572 115,395 114,820
The result is also compared with other methods like fuzzy reasoning (Lai and Ma, 1997),
differential evolution (DE), simple genetic algorithm (SGA), ant colony search algorithm
(ACSA) (Su, Chan and Chiou, 2005) and is shown in Table 5.2. From Table 5.2, it is seen
that proposed method can produce better results with effective annual cost of $1,14,820
producing reduced amount of 0.6731 MW total power loss after placing capacitors of
different sizes of total 8.850 Mvar.
5.4 CONCLUSION
Optimal reactive power compensation is one of the significant tasks in the operation of
distribution systems. The basic intention is to diminish power losses as well as to improve
voltage profile. In this chapter, an algorithm based on self-adaptive multivariate q-Gaussian-
based genetic algorithm technique (Mandal and Jana, 2012) has been successfully applied
to avoid premature convergence. To evaluate the effectiveness of the proposed algorithm,
it has been implemented on a sample test system mentioned earlier. The results obtained
by the proposed scheme have been compared with other population-based algorithms like
fuzzy reasoning, DE, SGA, and ACSA. The results reflect that the designed algorithm is
certainly capable of obtaining superior quality solutions.
REFERENCES
1. Baghzouz, Y., and Ertem, S., 1990. Shunt capacitor sizing for radial distribution feeders with
distorted substation voltages, IEEE Trans. on Power Delivery, 5, 650–657.
2. Baran, M. E., and Wu, F. F., 1989. Optimal Sizing of Capacitors Placed on a Radial
Distribution System, IEEE Transactions Power Delivery, 1,1105–1117.
3. Su, C. T., and Tasi, C. C., 1996. A new fuzzy reasoning approach to optimum capacitor
allocation for primary distribution systems, Proc. IEEE on Industrial Technology
Conference., 237–241.
4. Lai, L. L., and Ma, J. T., 1997. Application of evolutionary programming to receive power
planning-comparison with nonlinear programming approach, IEEE Transactions on Power
Systems, 12,198–204.
5. Yu, X., Xiong, X., and Wu, Y., 2004. A PSO based approach to optimal capacitor placement
with harmonic distortion consideration, Electric Power System Research, 71, 27–33.
6. Mandal, K. K., Jana, D., 2012. Reactive Power Planning for Distribution Feeders Using a
Novel Improved Self Adaptive Particle Swarm Optimization Technique, CIIT International
Journal of Artificial Intelligent Systems and Machine Learning, 4, 6, 409–415.
7. Chun Wang and Hao Zhong Cheng, 2008. Reactive power optimization by plant growth
simulation algorithm, IEEE Transactions on Power Systems,23, 1,119–126.
8. Ji-Pyng Chiou, Chung-Fu Chang and Ching-Tzong Su, 2006. Capacitor placement in large
scale distribution system using variable scaling hybrid differential evolution, Electric Power
and Energy Systems,28, 739–745.
9. Kennedy, J., and Eberhart, R., 1995. Particle swarm optimization, In Proceedings IEEE
Conference Neural Networks (ICNN’95), Perth, Australia, IV, 1942–1948.
10. Hogan, J., 2004. Hawking cracks black hole paradox, New Scientists (Online); www.
newscientists.com.
11. Su, C. T., Chan, C. F., and Chiou, J. P., 2005. Capacitor placement in distribution system
employing Ant Colony Search Algorithm, Electric Components and Systems, 33,931–946.
12. Deb, K., Goyal, M.: A combined genetic adaptive search (GeneAS) for engineering design,
Comput. Sci. Inform. 26 (4), 30–35.
CHAPTER 6
Differential Evolution Coupled

Neural-based Prediction of Soil
Water Content
Sayan Saha, Kaushik Majumder, Bitan Chakraborty, and
Sankhadeep Chatterjee
Department of Computer Science & Engineering, University of Engineering & Management, Kolkata, India.
ABSTRACT: In soil science, the water content in the soil plays a major role in agriculture
and soil chemistry. As the formation of water content depends on various factors, such as
field air temperature, field soil temperature, and relative humidity (Rh), accurate prediction
of water content was hardly possible without the introduction of any advanced technology.
With the advancement of machine learning by various researchers, accurate prediction of
water content in the soil became possible. The emphasis of this paper is on the prediction of
water content in the soil by utilizing differential evolution (DE) to train in artificial neural
network (ANN) to make this prediction possible. A comparison of average root mean
square error (RMSE) values and test phase predictions of DE-supported ANN method
with other models, such as support vector machine (SVM), multi-layered perceptron
(MLP) feed-forward network, Cuckoo Search-supported ANN model, and particle swarm
optimization (PSO)-supported ANN model has been reported. Results provided by all the
machine-learning models used in this study show that the proposed DE-supported ANN
method provides a better solution by reflecting lesser RMSE values than other models in
predicting soil water content.
Keywords: Soil moisture, Relative humidity, Differential evolution, Artificial neural

network, Support vector machine, Multilayer perceptron, Cuckoo search, Particle swarm
optimization, Root mean square error.
6.1 INTRODUCTION
In soil science, the water content in the soil plays a major role mainly in agriculture and
soil chemistry. The water content in the soil depends on various factors such as field
49
air temperature, field soil temperature, and relative humidity. Relative humidity (Rh)
(Wang, 2019; Johnston, 2019) is a combination of barometric pressure, temperature,
and moisture contained in the air and thus Rh plays an important role in determining
the water content in the soil. Along with the RH factor, air and soil temperatures also
help to predict the same. The better the conditions mentioned above, higher is the water
content, and the higher the water content, the better is the soil fertility (Mäder, 2002;
Muhammed, 2018; Clivot, 2019). Thus, with higher soil fertility, proper cultivation
could be done on that soil. Hence, by taking some factors into account, such as Rh and
temperatures of field soil and air around it, soil water content can be predicted using
suitable machine-learning methods.
6.2 BACKGROUND
Artificial Neural Networks (ANN) (Blanco, 2018 and Morse, 2016) contains artificial
neurons or perceptron units that take inputs and their respective weights and multiply it
to find out the solutions. There are two types of perceptrons: single-layered perceptrons
and multi-layered perceptrons (MLPs). Single-layer perceptrons contain input and output
layers, but in case of MLPs (Storm, 2019 and Murata, 1994), it contains one or more extra
layer known as the hidden layer(s) which is in-between the input layer and the output
layer. Hidden layer hides all the inter-processes from the outside world and also hides
information thus stopping any direct access from outside. Though MLP (Riedmiller,
1994) is complex in nature, it can achieve non-linear separations provided that we enter
corrected inputs.
FIGURE 6.1 Layers of multilayer perceptron.
As depicted in Figure 6.1, input layer provides information to the network, hidden layer
performs computations and transfers the information to the output layer, and the output
Differential Evolution Coupled Neural-based Prediction of Soil Water Content  51
layer transforms the information to the outside world. Support vector machines (SVM)
(Suykens, 1999; Basak, 2007) are a type of supervised machine learning model, which
is also known as Support Vector Networks and are a more developed machine learning
model that analyses data mainly for classification analysis. Unlike ANN, SVM does not
depend upon any dimensions of the inputs space, and the solution of this is global and
more optimum.
FIGURE 6.2 Flowchart diagram showing the working principle of SVM model.
Following the flowchart (Figure 6.2), datasets are prepared and then normalized.
Then an activation function is selected, and the parameters are optimized after cross-
validation. Training and testing of SVM network are then done to evaluate model
performance.
Particle swarm optimization (PSO) (Chatterjee, 2017) is an optimization method
that iteratively solves a problem by improving a candidate solution taking its quality into
account. In this optimization method, particles in a solution space which moves around
the solution space by some velocity (calculated by using various mathematical formulae
over particle velocity) are considered. Particle movements are then influenced by their best-
known position and then guided towards the best positions. In this way, the collection
of particles moves towards the best solution. Cuckoo Search (CS) (Yang, 2010) is an
optimization algorithm inspired by a bird family known as CUCKOO where they lay their
eggs in other nests of other breed birds. Host birds finding that the eggs are foreign may
either throw it away or build a new nest and transfer there. Using this concept, Yan and Deb
(Yang, 2010) designed an algorithm known as Cuckoo Search algorithm, where cuckoo
eggs are considered as a solution and hence, this algorithm finds the optimized solutions
by visiting random nets. In the current study, ANN has been trained with a well-known
optimization algorithm called differential evolution (DE) (Price, 2013; Kaelo, 2006). The
performance of DE in stochastic optimizations has been proved to be better than other
algorithms. Motivated by this, the current study used DE to find the optimal weights of
ANNs. In addition, a comparison with other two well-known optimization algorithms PSO
and CS has been reported (Das, 2010).
FIGURE 6.3 Flowchart diagram showing DE model working principle.
Following the flowchart as depicted in Figure 6.3, DE generates the initial population by
using a population size provided in the algorithm. Following the generation stage is the
evaluation stage, where fitness values get evaluated which determines how close solutions
can come so that it can achieve the objectives required to generate optimized solutions.
After the evaluation stage gets over, evolution or generation of new population takes place
until and unless total range or iterations are completely covered. In this way, DE provides us
the best optimum solution by iteratively selecting the past generation’s best solutions taking
fitness value factors into account.
6.3 METHODOLOGY
It has been already discussed that predicting water content in soil could certainly help
farmers to know which soil provides maximum fertility. This can be done perfectly
with the help of machine-learning models that can correctly predict the water content
in the soil by taking some major factors into account. Some problems which are non-
continuous, flat, noisy, and non-differentiable are difficult to solve analytically. Hence,
DE was introduced that takes ANN’s initial weights and finds the optimal set of weights.
Then, with the help of those optimized values, the NN-DE (ANN trained with DE) model
is tested.
FIGURE 6.4 Flowchart showing the complete working cycle of DE model using ANN.
As depicted in Figure 6.4, population size is initialized to 40 to obtain optimized accurate

solutions. However, population size can also be initialized to a value (less or more than
40), but in that case, solution accuracy could have been degraded and solutions had been
less optimized as compared to the optimized solution observed taking population size
as 40. When the initialization procedure gets over, generation of initial population takes
place from where evolution starts to occur. Taking mutation scale factor as 0.8, mutation
takes place to maintain genetic diversity from older generations to newer generations
of a population. As soon as mutation (Mallipeddi, 2011) is completed, crossover or
recombination takes place taking the probability of crossover as 0.70. It results in the
generation of a new population by recombining the genetic information of two solutions
(or parents) using genetic crossover operators (Poon, 1995).
In this study, mutation scale factor and probability of crossover are taken as 0.8 and
0.7, respectively, because after implementing trial and error method 10 times each for
finding best mutation scale factor and probability of crossover, it was found that taking
mutation scale factor as 0.8 and probability of crossover as 0.7 provided more optimized
solutions than taking any other values. After crossover is completed, selection of solution
from the generated population is done by choosing optimized solutions using fitness
function. In this study, fitness function (Baresel, 2002; Eggermont, 2019) is used to
determine how close a solution could come in order to match the aim in this study which
is to predict the perfect optimized solution. This whole cycle repeats for 500 times, as
iteration was initially declared to 500. In this study, iteration of this cycle was required
in order to get the final optimized solution and also to show proper evolution of fitness
values. The above-mentioned parametric setup has been engaged for DE algorithm which
is the training phase of NN-DE. Here, root mean square error (RMSE) (Abraham, 2009;
Chai, 2014) has been used to measure differences between values predicted by the model
used in this study.

In this study, a dataset from the University of Toronto, Mississauga, has been used
for predicting the soil water content. This dataset is publicly available in (http://geog.
utm. utoronto.ca/UTMMS/env_data/). Here, soil temperature, air temperature, and Rh
(in percentage) at the field site have been taken as inputs, and with the help of DE
technique in ANNs, it predicts the soil water content at the field site and also the best
soil fertility using RMSE function. In this study, soil temperature, air temperature, and
Rh (in percentage) at the field site has been taken as the inputs and with the help of five
different machine-learning models, different RMSE values have been observed in the
comparison as shown in Table 6.1.
TABLE 6.1 Comparison of Different Models in Terms of RMSE
TEST NN-DE SVM MLP NN-CS NN-PSO
1 0.0139 1.7512 0.0611 0.0152 0.0153

2 0.0141 0.7104 0.0997 0.0141 0.0143
3 0.0142 0.2440 0.0936 0.0154 0.0175
4 0.0138 0.2372 0.0242 0.0157 0.0165
5 0.0140 0.2372 0.0159 0.0154 0.0162
6 0.0137 0.3838 0.0146 0.0149 0.0172
7 0.0143 0.0706 0.0539 0.0146 0.0160
8 0.0140 0.3673 0.0339 0.0154 0.0145
9 0.0135 0.0638 0.0160 0.0151 0.0153
10 0.0145 4.1350 0.0202 0.0151 0.0164
AVG. 0.0140 0.8200 0.0433 0.0151 0.0159
Taking the average RMSE values of NN-DE, support vector machine (SVM) model,
multi-layered perceptron (MLP), Cuckoo Search-trained ANN (NN-CS), and PSO-
trained ANN (NN-PSO) from Table 6.1, it has been observed that NN-DE model
which was implemented in this study provided the minimum average RMSE value
among all the other models used in this study. In this study, test phase predictions of
MLP, SVM, and NN-DE models have been made by using regression plots provided
in Figure 6.5.
FIGURE 6.5 Test phase predictions by MLP model.
FIGURE 6.6 Test phase predictions by SVM model.
FIGURE 6.7 Test phase predictions by DE model.

FIGURE 6.8 Graph showing evolution of fitness values on 500 iterations.
Analysing all the regression plots (Figures 6.5, 6.6, and 6.7), it has been observed that in the
regression plot of NN-DE model (Figure 6.7), the output generated is more accurate than
the other generated outputs in the regression plots of MLP and SVM model (Figures 6.5
and 6.6), as the output generated in NN-DE model is almost equal to the expected output.
It revealed that the accuracy of prediction of NN-DE is better than all the other studied
models used in this study. The convergence plot of training phase of NN-DE has been
shown in Figure 6.8. It depicts that the RMSE value has been saturated at the use of 500
iterations.
6.5 CONCLUSION
Predicting soil water content has direct impact on agricultural sectors. The current study
proposed a hybrid NN-DE model to predict soil water content. It has been observed that
by training in ANN using DE, minimum RMSE can be achieved compared to MLP, SVM,
NN-CS, and NN-PSO. Experimental results revealed that the performance improvement
of the ANN trained with Cuckoo Search over NN-PSO is nominal in predicting soil water
content. However, NN-DE has been able to improve performance to a great extent.
REFERENCES
1. Wang, X., Hu, Y., Gu, Y., Zhao, X., and Chen, X., 2019. Effects of relative humidity on the
broadband extinction performance of bioaerosol. Optics Express, 27(17), 23801–23813.
2. Johnston, A. N., Bruggeman, J. E., Beers, A. T., Beever, E. A., Christophersen, R. G., and
Ransom, J. I., 2019. Ecological consequences of anomalies in atmospheric moisture and
snowpack. Ecology, 100(4), e02638.
3. Mäder, P., Fliessbach, A., Dubois, D., Gunst, L., Fried, P., and Niggli, U., 2002. Soil fertility
and biodiversity in organic farming. Science, 296(5573), 1694–1697..
4. Muhammed, S. E., Coleman, K., Wu, L., Bell, V. A., Davies, J. A., Quinton, J. N.,..., and
Naden, P. S., 2018. Impact of two centuries of intensive agriculture on soil carbon, nitrogen
and phosphorus cycling in the UK. Science of the Total Environment, 634, 1486–1504.
5. Clivot, H., Mouny, J. C., Duparque, A., Dinh, J. L., Denoroy, P., Houot, S.,..., and Mary, B.,
2019. Modelling soil organic carbon evolution in long-term arable experiments with AMG
model. Environmental Modelling & Software, 118, 99–113.
6. Blanco, C. M. G., Gomez, V. M. B., Crespo, P., and Ließ, M., 2018. Spatial prediction of soil
water retention in a Páramo landscape: Methodological insight into machine learning using
random forest. Geoderma, 316, 100–114.
7. Morse, G., and Stanley, K. O. (2016, July). Simple evolutionary optimization can rival
stochastic gradient descent in neural networks. In Proceedings of the Genetic and
Evolutionary Computation Conference 2016 (pp. 477–484). ACM.
8. Storm, H., Baylis, K., and Heckelei, T., 2019. Machine learning in agricultural and applied
economics. European Review of Agricultural Economics.
9. Murata, N., Yoshizawa, S., and Amari, S. I., 1994. Network information criterion-
determining the number of hidden units for an artificial neural network model. IEEE
transactions on neural networks, 5(6), 865–872.
10. Riedmiller, M., 1994. Advanced supervised learning in multi-layer perceptrons—from
backpropagation to adaptive learning algorithms. Computer Standards & Interfaces, 16(3),
265–278.
11. Suykens, J. A., and Vandewalle, J., 1999. Least squares support vector machine classifiers.
Neural processing letters, 9(3), 293–300.
12. Basak, D., Pal, S., and Patranabis, D. C., 2007. Support vector regression. Neural Information
Processing-Letters and Reviews, 11(10), 203–224.
13. Chatterjee, S., Hore, S., Dey, N., Chakraborty, S., and Ashour, A. S., 2017. Dengue fever
classification using gene expression data: a PSO based artificial neural network approach.
In Proceedings of the 5th international conference on frontiers in intelligent computing:
theory and applications (pp. 331–341). Springer, Singapore.
14. Yang, X. S., and Deb, S., 2010. Engineering optimisation by cuckoo search. arXiv preprint
arXiv: 1005.2908.
15. Price, K. V., 2013. Differential evolution. In Handbook of Optimization (pp. 187–214).
Springer, Berlin, Heidelberg..
16. Kaelo, P and Ali, M. M., 2006. A numerical study of some modified differential evolution
algorithms. European journal of operational research,169(3), 1176–1184.
17. Mallipeddi, R., Suganthan, P. N., Pan, Q. K., and Tasgetiren, M. F., 2011. Differential
evolution algorithm with ensemble of parameters and mutation strategies. Applied soft
computing, 11(2), 1679–1696.
18. Poon, P. W., and Carter, J. N., 1995. Genetic algorithm crossover operators for ordering
applications. Computers & Operations Research, 22(1), 135–147.
19. Baresel, A., Sthamer, H., and Schmidt, M. (2002, July). Fitness function design to improve
evolutionary structural testing. In Proceedings of the 4th Annual Conference on Genetic
and Evolutionary Computation (pp. 1329–1336). Morgan Kaufmann Publishers Inc.
20. Eggermont, J., Eiben, A. E., and van Hemert, J. I. (1999, May). Adapting the fitness function
in GP for data mining. In European Conference on Genetic Programming (pp. 193–202).
Springer, Berlin, Heidelberg.
21. Abraham, B., and Ledolter, J., 2009. Statistical methods for forecasting (vol. 234). John
Wiley & Sons.
22. Chai, T., and Draxler, R. R., 2014. Root mean square error (RMSE) or mean absolute
error (MAE)?–Arguments against avoiding RMSE in the literature. Geoscientific model
development, 7(3), 1247–1250.
23. Das, S., and Suganthan, P. N., 2010. Differential evolution: A survey of the state-of-the-art.
IEEE transactions on evolutionary computation, 15(1), 4–31.
CHAPTER 7
Numerical Investigation of Heat

Conduction in a Rectangular
Composite Plate with Sinusoidal
Heat Input from Top Wall
Debarghya Kar and Sumanta Banerjee
Mechanical Engineering Department, Heritage Institute of Technology, Kolkata 700107, India.
ABSTRACT: The results of steady-state numerical simulation of heat conduction in

a rectangular composite plate are presented in this chapter. Sources of heat generation
within the plate are not considered, and heat input is assumed to take place only through
the top edge. A three-dimensional solid of rectangular parallelepiped geometry of depth(z)
substantially greater than the transverse dimensions(x,y) can be thermally treated and
¶T ¶T ¶T
analysed as a 2–D plate, with ∼ . The top edge is subjected to non-uniform heat
¶x ¶y ¶z
input with sinusoidal temperature distribution. This kind of Dirichlet boundary condition
is relevant from the viewpoint of practical applications. The side edges and the bottom edge
are maintained isothermally cold at the ambient temperature. The plate is composed of two
sections, each with a different value of thermal conductivity. Contact resistance is neglected.
The heat transfer through the plate is studied numerically by employing the Finite Volume
Method. The temperature field is visualised through isotherm plots. The corridors of enthalpy
transport are diagrammatically represented through the plots of heatlines. The fraction of
thermal energy dissipated through each of the side and bottom edges are calculated.
Keywords: Numerical simulation, Composite plate, Sinusoidal temperature distribution,

Finite volume method, Isotherm, Heatline.
7.1 INTRODUCTION
In many heat-transfer applications, the heat-conduction medium is composed of multiple
sections of materials with different thermal properties (Holman, 2010). The individual
59
sections, composed of different materials, are integrated as a composite structure to suit

specific technological applications. Many of these composites, structurally fabricated
such that one (of the three) dimension is considerably larger than the other two, can be
thermally modelled as quasi–2D composite plates, provided that the field-gradients along
the transverse directions are only significant. Practical examples are found in furnaces and
ovens, electronic components, thermal insulation systems, and composite fins (Sami, 1994).
The cooling and heating loads of buildings also depend on the thermal behaviour in the
composite constructions of their walls and roofs. Roads and highways are also constructed
using an asphalt-concrete composite, with the soil layer underneath. The temperature
variations and thermal stresses developed within these composite laminas are of central
interest to the engineer, as this information has a direct bearing on the formation of cracks
under different traffic-load conditions. These practical applications, therefore, warrant the
necessity to calculate the temperature distribution and heat transfer rates in composite
structures under different thermal loading conditions and material properties (Ozisik, 1980).
Moreover, in many engineering applications, the thermal conductivity of materials varies with
temperature. When small temperature differences are involved, the conductivity is assumed
to vary linearly with temperature. In such situations, the heat transfer analysis is conducted by
taking a temperature-averaged value of conductivity for mathematical simplicity. In general,
however, the relation is non-linear. The heat conduction is rendered more complex when the
thermal conductivity depends upon the direction of heat flow in anisotropic materials.
A number of analytical and numerical works have been reviewed for the problem
of heat conduction in a plate. Various analytical methods are available for solving one-
dimensional, time-dependent heat conduction in composite media, with temperature-
independent properties and different stationery and/or spatially varying (periodic/non-
periodic) boundary conditions (Poulikakos, 1994; Kar, Chan, and Mazumder, 1992; Zedan
and Mujahid, 1993). The temperature and heat flux distribution for steady-state heat
conduction in 3D rectangular parallelepiped geometry have been presented by Crittenden
and Cole (2002). In this work, the authors have presented an algorithm to obtain the results
at different boundary conditions. Haji-Sheikh, Beck, and Agonafer (2003) have presented a
computational study of temperature distribution under steady state in 3D multilayer bodies.
Sun and Wichman (2004) have solved the problem of transient 1D heat conduction in a
composite slab analytically by the eigenfunction expansion method. Other published works
(Kidawa-Kukla, 2008, Beck and Cole, 2008) specify the influence of different boundary
conditions, temperature-dependent thermal conductivity, and heat flux pathways in 2D
rectangular plates, based on analytical and/or numerical analyses. The imposed thermal
loads are due to stationary and/or moving heat sources.
7.2 METHODOLOGY
The primary objective of the present work is to conduct a simulation study of 2D steady-
state heat conduction in a rectangular plate composed of two sections, each with a different
(constant) value of thermal conductivity k. The individual sections are assumed to be
homogeneous and isotropic. There is no internal heat generation. Perfect thermal contact
Numerical Investigation of Heat Conduction in a Rectangular Composite  61
is assumed at the interface so that the contact resistance is neglected. The schematic of
the rectangular plate, with appropriate thermal boundary conditions, is depicted in Figure
7.1(a). The lengths of the horizontal edge and the vertical edge of the plate are W and H,
respectively. The thermal boundary conditions (Dirichlet-type) are also specified. The two
sections are visually demarcated by distinct hatchings.
The governing differential equation for the chosen system is
d æ dT ö d æ dT ö
çk ÷+ k =0 (7.1)
dx è dx ø dy çè dy ÷ø
As the plate has a composite structure, the interfacial conductivity ki is interpolated as
a harmonic mean of the component thermal conductivities of the upper and lower
sections(kU, kL):
2 æ 1 1ö (7.2)
=ç + ÷
ki è kU k L ø
x y T - To
Defining the non-dimensional variables as X = , Y = , q = , the governing
equation in dimensionless form is W W Tm
d æ dqö d æ dqö
çk ÷ + k =0 (7.3)
dx è dx ø dy çè dy ÷ø
The specified-temperature boundary conditions in non-dimensional form are shown in
Figure 7.1(b). In addition, the forms of Eq. (7.3) valid for the two sections are (AR=H/W is
the plate aspect ratio):
d æ dq ö d æ dq ö æ H1 ö (7.4a)
çè ÷ø + çè ÷ø = 0;0 £ X £ 1, çè AR - ÷ø £ Y £ AR
dX dX dY dY W
d æ kL d q ö d æ kL d q ö æ H ö
ç ÷ + ç ÷ = 0;0 £ X £ 1,0 £ Y £ ç AR - 1 ÷ (7.4b)
dX è kU dX ø dY è kU dY ø è Wø
(a) (b)
FIGURE 7.1
The finite volume method (Patankar, 980) is used to solve the steady-state 2D differential
equation of heat conduction. A set of algebraic equations are formulated and solved
for the interior and the boundary nodes of the plate. The formulation of the interfacial
thermal conductivity as a harmonic mean of the neighbouring conductivities ensures the
conservation of heat flux across the interface (Versteeg and Malalasekera, 2007). Literature
review shows that analytical results exist for the case of steady-state 2D heat conduction
through a homogeneous and isotropic plate subjected to sinusoidal heat input at the top
edge and cooling from the sides as well as the bottom. The analytical expression governing
the temperature distribution is (Holman, 2010):
py
sinh
T = Tm W sin æ px ö + T (7.5)
pH çè ÷ø o
sinh W
W
7.3 DISCUSSION
The results of simulation study have been validated against the analytical solution of the
temperature field obtained from Eq. (7.4). The isotherm plots for the two (‘green’ continuous
line for the analytical solution and ‘black’ broken line for the numerical solution), as
depicted in Figure 7.2(a), are seen to overlap and match excellently. The corresponding
heatline plots (see Figure 7.2(b)) show the symmetrical distribution of heat load. The heat
functions (Bejan, 1983) are defined as follows:
¶H ¶T ¶H ¶T
=k ; = -k (7.6)
¶x ¶y ¶y ¶x
(a) (b)
FIGURE 7.2
Parametric studies have been carried out to investigate the effects of grid size and
conductivity-ratio of the materials in the (composite) plate. Figure 7.3(a–b) depicts the
Numerical Investigation of Heat Conduction in a Rectangular Composite  63
isotherm plots for two representative cases of kL/kU = 0.25 and kL/kU = 0.30 (L: Lower
plate; U: Upper plate). The isotherm plots depict a dense crowding of isotherms near the
vicinity of the top edge and are indicative of high values of temperature gradients. The
case kL/kU = 0.25 implies that the lower plate has a value of thermal conductivity one
quarter that of the upper plate. While in kL/kU = 0.30, the thermal conductivity of the lower
plate is one-third of upper plate. Physically, one must expect that the concentration of
the isotherms shall be practically concentrated over the extent of the top section
(for kL/kU = 0.25); for the latter, it is expected that the isotherms shall spread out (permeate)
more into the lower plate. This fact is corroborated well by the isotherm plots. It can be
observed that the horizontal dashed line marks the reach of the bottom tip of q = 0.06
isotherm. The bottom tip permeates somewhere between 0.35 £ Y £ 0.4 for kL/kU = 0.25;
for kL/kU = 0.30 the bottom tip reaches down between 0.35 £ Y £ 0.35. This corresponds to
increased participation of the bottom section in heat transfer.
Frame 001 ½ 14 Sep 2019 ½ Frame 001 ½ 14 Sep 2019 ½
0. 0.
1 74 78
0. 4 1 74
0.74
8
0.26 .18
0.02
0.26 .18
0.02
0.86 0.9 0.86 0.9
0.7 0.82 0.7 0.7 0.82 0.7
54
0.3
4
0
4
0
0.
0.3
54
0.3
0.3
0.
0.
4
0.02
0.66
0.
0.66
46
46
4
0.5 0.62 0.5 0.62
42 42
0.02
0.2
0.2
0. 4
38 0.58 0. 0.58 0.
0.
2
2
0.14
0.14
38
0.4
0.8 0.5 6
0.8 0.4 6
2
0.4 0.5
2
2 0.4
0. 0.3
2
0.2
0.06
0.2
0.06
8
26
0.3
0. .3
0.06
26
0.3
0.06
0.1
0.1
0.3 0.34 0.3 0.34

14
0.26 18 0.2
0. 6
0.1
14
18
0.
0.
0.
0.18 0.22 0.1
0.6
0.1
0.6 8 0.22
0.02
0.02
0.14
0.14
2
0.1
Y
Y
0.0
0. 0.1
0.02
06 0.
6
0.0 06
0.4 0.4 6
0.0
0.0
2 0.
0.2 0.2 02
0.02
0.02
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
X X
Figure 3(a) Figure 3(b)

FIGURE 7.3
Figure 7.4(a–b) depict the heatline plots for two representative cases of kL/kU = 0.25
and kL/kU = 0.30. The plots depict a symmetrical layout of the enthalpy corridors for
both kL/kU = 0.25 and kL/kU = 0.30. This is because the central vertical line represents
an axis of symmetry. The enthalpy transport (heat flux) is locally normal to the
isotherms. In addition, the heatlines terminate normally over the side and the bottom
edges of the plate, as the latter are maintained isothermally cold. Physically, as one
compares the situations of kL/kU = 0.25 and kL/kU = 0.30, one should expect a greater
participation of heat transfer by the side edges as the ratio of kL/kU improves. This
fact is corroborated well by the heatline plots. It can be observed that the heatline
H = 0.1 (H = H/kTm) is pushed marginally from the vertical centre line towards the
side edges, and all the heatlines that terminate at the side edges are (likewise) lifted
up. This is due to the decreased value of local curvature of the isotherms as they
are permeated more into the domain of the lower section.
(a) (b)
FIGURE 7.4
REFERENCES
1. Holman, J. P., 2010. Heat Transfer. New York: Mc Graw Hill Inc.
2. Al-Sanea, Sami A., 1994. Two-dimensional heat conduction in a composite slab with
temperature-dependent conductivity. Journal of Engineering Science, 7, 219–246.
3. Ozisik, M. N., 1980. Heat Conduction. New York: John Wiley Inc.
4. Poulikakos, D., 1994. Conduction Heat Transfer. Prentice Hall Inc.
5. Kar, A., Chan, C. L., and Mazumder, J., 1992. Comparative studies on nonlinear hyperbolic
and parabolic heat conduction for various boundary conditions: analytic and numerical
solutions. Journal of Heat Transfer. 114, 14–20
6. Zedan, M. F., and Mujahid, A. M., 1993. Laplace transform solution for heat transfer in
composite walls with periodic boundary conditions. Journal of Heat Transfer. 114, 14–20.
7. Crittenden, P. E., and Cole, K. D., 2002. Fast-converging steady-state heat conduction in a
rectangular parallelepiped. International Journal of Heat and Mass Transfer, 45, 3585–3596.
8. Haji-Sheikh, A., Beck J. V., and Agonafer D., 2003. Steady-state heat conduction in multi-
layer bodies. International Journal of Heat and Mass Transfer, 46, 2363–2379.
9. Sun, Y., and Wichman I. S., 2004. On transient heat conduction in a one-dimensional
composite slab. International Journal of Heat and Mass Transfer, 47, 1555–1559.
10. Kidawa-Kukla, J., 2008. Temperature distribution in a rectangular plate heated by a moving
heat source. International Journal of Heat and Mass Transfer, 51, 865–872.
11. Beck, J. V., and Cole, K. D., 2008. Conduction in rectangular plates with boundary
temperatures specified. International Journal of Heat and Mass Transfer, 51, 4676–4690.
12. Patankar, S. V., 1980. Numerical Heat Transfer and Fluid Flow. Washington: Hemisphere
Inc.
13. Versteeg, H. K., and Malalasekera, W., 2007. An Introduction to Computational Fluid
Dynamics: The Finite Volume Method. Pearson Education.
14. Kimura, S., and Bejan, A., 1983. The ‘Heatline’ Visualization of Convective Heat Transfer.
Transactions ASME, 105, 916–919.
CHAPTER 8
Analytical Modelling and

Simulation of a Hot-Wall
Condenser Applied to Domestic
Refrigerator
S. N. Sapali1 and N. D. Shikalgar2
Department of Mechanical Engineering, College of Engineering Pune, Maharashtra, India.
1,2
1
sns.mech@coep.ac.in, 2nds.mech@coep.ac.in
ABSTRACT: Domestic refrigerator works on vapour compression refrigeration principle

with isobutane as a refrigerant. The heat from the refrigerated space is carried away by
the refrigerant and is rejected into the atmosphere by the condenser. In earlier domestic
refrigerators the condenser coil should be placed outside the refrigerator wall. To avoid
any damage to the condenser coils while in transportations, it is embedded in the sidewalls
of insulation, which is referred to as a hot-wall condenser. In hot-wall condenser, the tube
surface is supported with a galvanized plate and an aluminium foil is used as an adhesive
tape, which holds the tube in place. The heat rejected from the hot-wall condenser to the
atmosphere is very small. In such a refrigerator, the average heat flux is within the range
of 135 W/m2 to 142 W/m2 and the temperature gradient along the surface of the hot-wall
condenser is in the range of10 K/m to the 43 K/m. The hot-wall condenser additionally
increases the heat load of about 28 W. In the present research, the model is presented in
order to enhance the rate of heat transfer from condenser tube to the metallic wall of the
refrigerator; the condenser tube is supported with a hemispherical cross-sectioned plate to
increase the contact surface. To carry out the heat transfer analysis with such a modified
arrangement the steady-state, analytical modelling and a two-dimensional simulation of a
hot-wall condenser is carried out. The simulation results show 9.3% improvement in heat
transfer rate through walls and 50% drop in polystyrene surface temperature.
Keywords: Heat transfer, Household refrigerator, Numerical imulation, Hot-wall

condenser.
65
8.1 INTRODUCTION
In modern refrigerators, a condenser called a hot-wall condenser has been introduced
instead of wire tube condenser. The geometry of skin condenser is a simple heat exchanger
composed of a condenser tube, in which condensation of the refrigerant takes place; a
galvanized plate, which serves as the support and fin, to increase the heat transfer area,
an aluminium foil as an adhesive tape, which holds the tube in place, and like the plate,
increases the heat transfer area that is similar to hot-wall or skin condensers (Rebora and
Tagliafico, 1997; Bansal and Chin, 2002; Gupta and Gopal, 2008; Colombo et al., 2016 ) and
polyurethane foam as an insulation. The polyurethane foam prevents the amount of heat
transfer from the condenser tube to the refrigerator compartment. Their main advantages
are low production costs, compared to other types of the condenser and a slim profile. This
makes them suitable for use in household refrigerators, where they can be positioned at the
back and the sides of the refrigerated compartment. The cross-section of actual hot-wall
condenser used in the domestic refrigerator is shown in Figure 8.1. This condenser consists
of a steel or aluminium tube that was in direct contact with the inner surface of the iron
plate on the sidewalls of a refrigerator.
FIGURE 8.1 Cut sectional model hot-wall condenser used in refrigerator.
One of the first attempts to analyse a hot-wall condenser and evaporator in a chest freezer for
household applications was made by Rebora and Tagliafico (1997). Bansal and Chin (2002)
conducted a study in which they compared their experimental data with the results of an
analytical model of a hot-wall condenser. The simulation model employed a finite-element
and variable-conductance approach. The model considered an aluminium foil as a fin. It
was also found that the aluminium foil has a significant impact on the heat transfer from
the condenser to the ambient. A study that compared the numerically and experimentally
determined performance of a skin condenser was performed by Colombo et al. (2016).
Both the aluminium foil and the steel plate were treated as independent fins. To validate the
model, they performed a series of experiments under different operating conditions. The
purpose of this study is to present a simplified methodology for the analytical modelling of
hot-wall condenser, based on the heat diffusion equation, which describes the temperature
distributions and heat flow in the entire structure of the condenser. The obtained results
from the numerical simulation in ANSYS Fluent and analytical results are presented. The
Analytical Modelling and Simulation of a Hot-Wall Condenser  67
focus of the study is limited to the composite structure of the condenser wall and the effect
of the parameter on the heat transfer.
8.2 ANALYTICAL HEAT TRANSFER MODEL OF THE HOT-WALL CONDENSER

The steady-state energy conservation equation is used to find temperature distribution
along with aluminium foil. Aluminium foil is divided into two sections as shown in
Figure 8.2. Section 1 consists of aluminium tape attached to the condenser tube on one
side and insulation on the other side. The path of heat transfer to section 1–2 is from the
refrigerant to condensing tube by convection and then condenser tube to aluminium foil
by conduction. Section 2 consists of an aluminium tube in contact with outer metal sheet
on one side and insulation on the other side. The heat is carried out from section 1 to 2 by
conduction in aluminium foil and passed to ambient through outer metal sheet.
FIGURE 8.2 Detailed fin model of one-half of a single element.
The governing equation for each section of aluminium foil is given by the differential
equation
d 2Ti
2
- a i 2T + bi dz = 0 (8.1)
dx
TABLE 8.1 Values of ai and bi for Different Sections
Section ai bi
ìï é dz dz ù üï ìï é T Tic ù üï
1–2 íê + ú / [ K al Aal ]ý íê
r
+ ú / [ K al Aal ]ý
R + Rc R f1 + Ric úû R + Rc R f1 + Ric úû
îï êë r þï îï êë r þï
ìï é dz dz ù üï ìï é T T¥ ù üï
ic
+ ú / [ K al Aal ]ý
ïî ëê R f 2 + Ric Ros + R¥ ûú ïþ ïî ëê R f 2 + Ric Ros + R¥ ûú ïþ
ìï é dz dz ù üï ìï é T T ù üï
ic
+ ic ú / [ K al Aal ]ý
ïî êë R f3 + Ric R¥ úû ïþ ïî êë R f3 + Ric R¥ úû ïþ
The general solution of the differential equation is

bi
Ti ( x ) = K j e(a1 ) x + K k e - (a1 ) x +
a2
where Kj and Kk are constants; which are different for different sections. The values of
constants are calculated by applying boundary conditions given in Table 8.2.
TABLE 8.2 Boundary Conditions for Different Sections
Section Inlet Exit
dT
1–2 =0 T = Tj1
dx
3 T = Tj1 T = Tj2
dT
3–4 T = Tj2 =0
dx
8.2.2 Variation of outside heat flux with contact angle in the hot-wall condenser
The variation of contact angle on outside heat flux (Qo) within hot wall condenser is
presented in Figure 8.3; from the figure it is clear that outside heat flux is minimum
(143.38 W/m2) at HWAC with point contact. It is increasing as the contact angle is
increased, and a maximum of 155.292 W/m2 at 180° contact angle. The increase in
heat flux due to the increase in surface area in contact with condenser tube and outside
plate of the condenser tube length is decreased from 23.39 m for 0° to 21.41 m for
180° contact angle. It is clear that as contact between tube plate interfaces is changed
from point contact to line (angular) contact, inside heat flux decreases with increase
in contact angle. The net heat ingress in the refrigerator compartment is reduced from
22.00 W/m2 to 17.73 W/m2.
FIGURE 8.3 Variation of outside and inside heat flux with a contact angle.
Analytical Modelling and Simulation of a Hot-Wall Condenser  69
8.3 NUMERICAL HEAT TRANSFER SIMULATION OF THE HOT-WALL

CONDENSER WITH ANGULAR CONTACT
Steady-state numerical simulation of the two-dimensional heat transfers inside a hot-wall
condenser is performed using ANSYS fluent. In the hot-wall condenser, the outer steel plate
acts as a fin. The plate is provided for an increase in the contact area in convective heat
transfer. By the using of fins, the contact area of the tube to air is increased therefore the
heat transfers rate increase. In hot-wall condenser composite structure, heat is dissipated
from refrigerant to aluminium tube, aluminium foil, air cavity, PUF insulation, steel plate
and atmosphere by conduction and convection. The thermal analysis as shown in Figure
8.7 is carried out at the following boundary conditions.
Condenser refrigerant temperature: 46°C
Atmospheric temperature: 25°C
FIGURE 8.4 Meshing of a single element FIGURE 8.5 Heat flux of a single element
with foil. with foil.
Total heat flux without aluminium foil with h = 6 W/m2K. The result of heat flux is almost
the same throughout the whole tube plate configuration. The value of average heat flux
is 0.7 W/m2 for a small element, which is within the range of 42.5 to 68.5 W/m2 for the
backplate.
8.4 CONCLUSION
This study presents an analytical model for describing the heat transfer in a hot-wall
condenser structure, using basic heat transfer equations, and numerical simulation to serve
as a reference for the analytical model. The results between the mathematical model and
the simulation were found to be in agreement. The following conclusions are made based
on this observation,
1. The small temperature difference can be attributed to the heat conduction along
with the aluminium foil, which shows the importance of including it in the
calculation.
2. The heat flux of hot-wall condenser with the atmosphere is increased by 8–10% as
compared to existing hot wall condenser.
3. The amount of heat ingress to refrigerator compartment due to the hot-wall
condenser is reduced in the range of 2–4%.
4. The required condenser cooling length for same condenser cooling is reduced by
2.08 m for hot wall condenser with angular contact.
REFERENCES
1. Bansal, P. K., and Chin, T. C., 2002. ‘’Design and modelling of hot-wall condensers in
domestic refrigerators’’, Applied Thermal Engineering, 22(14), 1601–1617.
2. Bansal, P. K., and Chin, T. C., 2003. ‘’Heat Transfer Characteristics of Wire-and-Tube and
Hot-Wall Condensers’’, HVAC&R Research, 9(3), 277–290.
3. Gupta, J. K., and Gopal, M. R., 2008. ‘’Modelling of hot-wall condensers for domestic
refrigerators’’, International Journal of Refrigeration, 31(6), 979–988.
4. Raiyani, Akash D., Sheth, N. R., Mehta, Niraj C., 2012. Thermal Analysis of Hot Wall
Condenser for Domestic Refrigerator’’, International Journal of Science and Research
(IJSR), ISSN: 2319–7064
5. Labigalini, Vinícius Brandani and Seixlack, Andre Luiz, 2013. Modelling of a hot-wall
condenser used in domestic refrigerators.22nd International Congress of Mechanical
Engineering ISSN 2176–5480
6. Ghule, Nitin and Mahajan, S. D., 2015, Mathematical Modelling of Skin Condensers for
Domestic Refrigerator. International Journal of Mechanical and Industrial Technology,
ISSN 2348–7593.
7. Elias Gava Colombo, Rodolfo da Silva Espindola, Fernando TestoniKnabben, Claudio Melo,
2016. A Numerical and Experimental Study on Skin Condensers Applied to Household
Refrigerators. International Refrigeration and Air Conditioning Conference.
8. Luka Lorbek, PrimozPoredos, Andrej Kitanovski, AlojzPoredos, 2018. Analytical modelling
and numerical simulation of heat transfer in a skin evaporator. International Journal of
Refrigeration 88(2018), 195–203.
9. Incropera, Frank, P., and DeWitt, David P., 1990. Fundamentals of Heat and Mass transfer,
3rd edition. New York, John Wiley & Sons, 1990.
CHAPTER 9
Design and Analysis of a Dual-Axis

Solar Tracker with an In-built Low
Cost Self-Cleaning Mechanism
Moumita Pramanik, Tuhin Kumar Barui, Pankaj Kumar Giri,
Hiranmay Samanta, Nirmal Kumar Deb, and Hiranmay Saha
Centre of Excellence for Green Energy and Sensor Systems (CEGESS), IIEST Shibpur, India.
ABSTRACT: The impending energy crisis due to rapid exhaustion of fossil fuels has
necessitated the exploitation of solar energy in more efficient ways. Solar tracking systems
track the maximum insolation by mechanically orienting the photovoltaic panel conforming
to the actual position of the sun and electrically through MPPT which is inbuilt in solar
inverters. Dual axis tracking system tracks both the altitude angle of the sun corresponding
to seasonal variations and azimuth angle of the sun corresponding to diurnal changes to
harness the maximum efficiency. However, with the passage of time, unwanted particles
such as dust, sand, twigs, and dry leaves tend to accumulate on the solar panels thereby
decreasing the maximum output power and hence the efficiency. The novelty in this work
is that we have designed and analysed a low-cost self-cleaning mechanism inbuilt in a
designed dual-axis solar tracker. The cleaning mechanism system has been implemented in
the same tracking system by a slight change in the fabrication technology. The design and
operating principle, efficiency with and without the in-built self-cleaning mechanism, the
future scope have been discussed and analysed in detail.
Keywords: Dual axis solar tracker, Efficiency, Self-cleaning mechanism.
9.1 INTRODUCTION
Solar energy is globally being used as a potential inexhaustible and clean energy source to
cater to our ever-increasing energy requirements. Solar panels convert solar energy directly
into electricity. Dust, twigs, dry leaves, and sand tend to accumulate on the solar panels as
time elapses, thereby, creating a hindrance for the maximum insolation on the solar cells.
71
Hence, the efficiency of the solar panels greatly reduces. Topology and seasonal variations
plays the role in deciding the type of accumulations on the solar panels.
Research is, therefore, being greatly done on cost-effective cleaning systems of solar
panels. The availability of maintenance facility, the generated power, the cost per unit of the
generated power, and the type and amount of accumulations on the solar panels determine
the type of cleaning mechanism to be installed in a solar power plant.
This work addresses this major problem by implementing an in-built low-cost self-
cleaning mechanism by a slight change in the fabrication technology in a designed automatic
dual-axis solar tracker. A detailed study and analysis of the design has been done.
9.2 DESIGN
The self-cleaning mechanism is in-built in the tracking system. It is therefore essential to
understand first the working and design of the designed tracker.
The mechanical design of the automatic programmable dual-axis tracking system has
been done on Solid Works 2016. Figure 9.1 shows the drawings of the designed solar tracker.
The tracker has been designed such that the weight of the solar panel is balanced
proportionately by a hollow cylindrical pipe filled with sand that balances the torsional
moment produced by the solar panel. Less moment means less torque which means less
electrical energy to be drawn to drive the dc motors. The design has been made in such a way
that the two dc series motors (12V, 100W, 1500 rpm) require energy only for overcoming
the bearing frictional losses. Figure 9.3 shows the designed robust dual-axis solar tracker
installed at IIEST, Shibpur.
FIGURE 9.1 The manufacturing drawing of the designed dual-axis solar tracker in SolidWorks 2016.
Design and Analysis of a Dual-Axis Solar Tracker with an In-built Low Cost  73
The smart control is designed using a combination of micro-controller (PIC 18F46K22), dc

series motor and light-dependent resistors (LDRs). The azimuth angle and the altitude angle
are controlled by two different sets of control systems. The micro-controller is programmed
to detect the sunlight with the help of LDRs and then actuate the dc motor to position
the solar panel in such a way so that it gets the maximum sunlight. The microcontroller
program has been done on mikroC PRO for PIC and fed to the microcontroller through
MPLAB IDE v8.92. Figure 9.2 shows the schematic of the PCB (Printed Circuit Board).
FIGURE 9.2 The schematic of the PCB for the control of the designed dual-axis solar tracker.
FIGURE 9.3 The designed and fabricated robust dual-axis solar tracker installed at IIEST, Shibpur.
9.3 OPERATION
The sun’s position in the sky varies with both the change in seasons as well as the time of the
day. The designed dual-axis tracking system tracks both the corresponding changes in the
altitude angle and azimuth angle by two separate control mechanisms.
When both the LDRs in any PCB are perpendicular to the incident radiation, the
voltage difference between them is below the error voltage. At this point, the motor stops
rotating for a pre-set value of time after which the microcontroller again reads the voltage
values of both the LDRs. They are again compared and necessary action is taken and the
cycle continues. Figure 9.5 shows the flowchart of east to west tracking for diurnal changes.
To implement the cleaning mechanism, the solar cell panel is turned to a vertical
position (perpendicular to the ground) by the control system maintaining the altitude
angle of the sun (when the voltage reading of its LDRs goes below certain pre-set value
tested and determined experimentally). The accumulated dry leaves or twigs or sand is
removed easily. This feature also saves the solar panel from damage during a hailstorm
or a sand storm. Figure 9.4 shows the flowchart of self-cleaning and tracking for seasonal
variations.
FIGURE 9.4 The flowchart for self-cleaning and tracking of maximum insolation for seasonal
variations.
FIGURE 9.5 The flowchart for east –west tracking for maximum insolation during diurnal
changes.

The solar panel used has dimensions = (1960 × 987) mm, Open Circuit Voltage = 84.621V,
Short Circuit Current = 4.345A, Power = 300Wp, Maximum Power Voltage = 72.834V,
Maximum Power Current = 4.136A, Fill Factor (FF) = 81.94%.
The designed solar tracker is kept at IIEST, Shibpur (22°35’04”N, 88°15’59”E). Under
the environmental conditions data has been recorded on
• 13 October 2019 for a fixed angle of 22.3°N and south facing in Table 9.1.
• 12 October 2019 for dual-axis tracking without any cleaning mechanism in Table
9.2 and dual-axis tracking with the in-built self-cleaning mechanism in Table 9.3.
Dry leaves covering around 20% of the panel area had been distributed on the solar module
surface for data in Table 9.2. Resistance of 19.5Ω had been connected to the output terminals.
TABLE 9.1 Experimental Data at a Fixed Angle without any Tracking and Cleaning Mechanism
Intensity of
Sl. Time of the Input Power Voltage Current Output Power Efficiency
Incident Light
No Day (Hours) (W) (V) (A) (W) (%) (ɳ)
(W/m2)
1. 07:40 36 67.64 1.3 0.05 0.065
2. 10:10 201 388.8 15.8 0.78 12.32
3. 12:15 665 1286.11 56.12 2.09 117.3
4. 13:20 645 1247.43 56.19 2.07 116.31 9.12
5. 14:46 346 669.3 22.3 0.9 20.07
6. 15:35 109 210.86 6.86 0.24 1.64
7. 17:00 10 19.32 0.5 0.02 0.01
TABLE 9.2 Experimental Data of Solar Tracking by the Designed Tracker without Any Cleaning Mechanism
Intensity of
Incident Light
(W/m2)
1. 07:30 35 67.69 1.3 0.06 0.078
2. 10:00 200 386.8 18.2 0.81 14.74
3. 12:00 665 1286.11 60.82 2.65 161.17
4. 13:00 645 1247.43 60.29 2.59 156.15 12.53
5. 14:30 350 676.9 24.8 1.1 27.28
6. 15:30 108 208.87 7.32 0.26 1.903
7. 17:00 10 19.32 0.5 0.02 0.01
TABLE 9.3 Experimental Data of Solar Tracking by the Designed Tracker with the Self-Cleaning Mechanism
Intensity of
Incident Light
(W/m2)
1. 07:32 35 67.69 1.5 0.08 0.12
2. 10:03 200 386.8 19.56 0.89 17.41
3. 12:02 665 1286.11 62.27 2.85 177.47
4. 13:03 645 1247.43 61.86 2.78 171.97 13.8
5. 14:32 350 676.9 27.8 1.3 36.14
6. 15:31 108 208.87 8.68 0.32 2.78
7. 17:02 10 19.32 0.5 0.03 0.015
As evident from the data in Tables 9.2 and 9.3, the efficiency with the inbuilt self-cleaning
mechanism increases by 1.3% over the efficiency without the cleaning mechanism. This
means that more solar energy can be harnessed efficiently if proper cleaning is done.
The efficiency of the solar panel fixed at a constant angle increases by 4.5% when solar
tracking is done by the designed tracker with the self-cleaning mechanism as evident from
Tables 9.1 and 9.3.
The energy used of the motors for dual-axis tracking and cleaning is negligible to the
energy generated by the 300 Wp solar module. The power consumed by the two motors
for tracking and cleaning purpose is around 72 W. Also, the designed tracking system
which additionally removes the preliminary accumulations on the solar panel is quite
cost-effective. The total cost of the designed tracking system is around 36000/– INR for a
single module of 300 Wp. By increasing the gear ratio, more panels may be operated at a
particular axis by a single motor. This may decrease geometrically the effective cost of the
solar tracking system in a solar plant.
9.5 CONCLUSION AND FUTURE SCOPE

For more precise tracking, better cleaning and increase in robustness of the designed model,
the following measures may be adopted in the future:
• The inbuilt cleaning mechanism removes any preliminary dust, twigs, dry leaves,
and sand. Additionally, a sprinkler or an air-jet depending on the type of dust
(dry or moist) may be operated for a few minutes by a reed-relay when the panel
stands perpendicular to the ground. This will help in cleaning off the remaining
accumulation present on the solar panel.
• For coastal areas that experience heavy rainfall and deserts that experience high
velocity of wind, the solar module may be kept at a particular angle from time to
time to remove any accumulation by the rain and the wind.
• Automatic wipers too may be used for better cleaning purposes.
• The design has been done as such that the motors require energy for overcoming
the frictional losses. For large solar plants, the gear ratio used may be further
decreased to increase the torque. A single motor may then be used for tracking of
many solar panels together at a particular axis.
• Ag braided connecting wires may be used to increase the robustness of the tracking
system. Hence, regular maintenance may not be required.
• For more accurate tracking, stepper motors may be used in place of the dc series
motors.
National Solar Mission of India has set a target of achieving 100 GW of installed solar
capacity by 2022. So far as on December 2018, India has installed about 25 GW of solar
power plants. This means more solar power plants and more maintenance of the old and
the newly installed plants.
The designed tracking system with the in-built cleaning system has much higher
efficiency than a normal dual-axis tracker as evident from the results obtained. The
designed system may find use at small-scale solar power generating stations in the future.
Modifications in the existing model may be done as suggested to increase the overall
efficiency and longevity of the system.
REFERENCES
1. Visconti, P., Ventura, V., Tempesta, F., Romanello, D., and Cavalera, G., 2011. Electronic
System for Improvement of Solar Plant Efficiency by Optimized Algorithm Implemented
in Biaxial Solar Trackers. 10th International Conference on Environment and Electrical
Engineering (EEEIC).
2. Boicea, A. V., Di Leo, P., Graditi, G., and Spertino, F., 2010. Comparison of Operating
Parameters in Grid Connected Photovoltaic Systems with Single/Double Sun-Trackers
at Different Latitudes. International Symposium on Power Electronics, Electrical Drives,
Automation and Motion, 130–133
3. Gaofa He, Chuande Zhou, Zelun Li, Review of Self-Cleaning Method for Solar Cell Array,
Elsevier Procedia Engineering 16 ( 2011 ) 640 – 645.
4. Mehdia, Ghazanfar, Alia, Naveed, Hussaina, Shafquat, Zaidib, Asad A., Shahd, Ahmer
Hussain, Mustafa Azeeme, M., 2019. Design and Implementation of the Dual-axis Solar
Tracking System., 2019 International Conference on Computing, Mathematics and
Engineering Technologies – iCoMET 2019.
5. Aliyu, Salihu O., Okwori, Michael, and Onwuka, Elizabeth N., 2016. A Prototype Automatic
Solar Panel Controller (ASPC) with Night-time Hibernation, I. J. Intelligent Systems and
Applications, 8, 18–25.
6. Chakrabory, Shreyasi, Mukherjee, Nilanjana, Biswas, Rashmi, Saha, Tanushree, Mohinta,
Astika, Modi, Neha Kumari, Samajdar, Dip Prakash., 2016. Microcontroller based Solar
Tracker system using LDRs and Stepper Motor – International Journal of Computer
Applications (0975 – 8887) International Conference on Microelectronic Circuit and
System (Micro-2015).
7. 2019 International Conference on Computing, Mathematics and Engineering Technologies
– iCoMET 2019
8. Mehdia, Ghazanfar, Alia, Naveed, Hussaina, Shafquat, Zaidib, Asad, A., Shahd, Ahmer
Hussain, and Mustafa Azeeme, M., 2019. Design and Fabrication of Automatic Single Axis
Solar Tracker for Solar Panel.
9. Moacir, Fábio, 2018. Monthly profile analysis based on a two-axis solar tracker proposal for
photovoltaic panels.
10. Hoffmann, Moacir, Fábio Molz, Rolf Fredi, Kothe, João Victor, Nara, Elpidio Oscar Benitez,
Tedesco, Leonel Pablo Carvalho, 2017. Universidade de Santa Cruz Do Sul. Published by
Elsevier Ltd.
11. Kulkarni, Aneesh, Kshirsagar, Tushar, Laturia, Akash, Ghare, P. H., 2013. An Intelligent
Solar Tracker for Photovoltaic Panels. Texas Instruments India Educators’ Conference.
12. Arlikar, Pratik, Bhowmik, Abhijit, Patil, Manoj., and Deshpande, Amruta, 2018. Three
Dimensional Solar Tracker with Unique Sensor Arrangement, 2015 International
Conference on Smart Technologies and Management.
CHAPTER 10
Indices based Comparative

Performance Analysis of Standard
Test Systems for Reliable Power
System Operation
R. Chowdhury and D. Jana
Institute of Engineering & Management, India.
ABSTRACT: For a reliable power system operation, selection of a Distributed Generator

(DG) unit, its suitable placement techniques, and its optimal location are a necessity. To
cater with the above requirement, this chapter focuses its attention on identifying the
strategic points of DG allocation for a set of distribution systems, from a smaller to a larger
one. It also targets to achieve optimal loading, voltage stability, profile improvement, and
several other primary objectives. This chapter not only focuses on the exact location of
DG units, but the other benefits associated with its placement are also an important point
of concern. It has been observed that depending on the size of the distribution system, the
performance of the different methods varies. By a suitable placement of a DG unit, the
primary requirements such as improvement in loading, voltage stability, power transfer
capacity, to name a few, are achieved.
Keywords: Distribution System, Allocation, Distributed Generation (DG), Loadability,

Performance Indices.
10.1 INTRODUCTION
Power generation by conventional sources (e.g. coal, water, nuclear materials) and its
distribution has been largely overpowered by the use of non-conventional resources
(solar, wind, natural gas, diesel, biomass etc.) due to their abundance in availability and
ease of use. Also protection and control of them can be easily done. Such a generation and
corresponding transmission and distribution are popularly known as distributed generation
79
(DG) (Ackermann, 2001). DG technology allows power to be generated in small size close
to load centres. So the power supplied by the distribution channel to the end consumers
not only comprises of the power received from the generation side but also the DG side.
As a result, by-directional flow of power occurs and production of electricity occurs in
close proximity to the consumers. It converts a passive distribution system to an active
one (Akorede, 2010) DG has a lot of advantages such as increase in reliability, reduction
in losses, reducing environmental impact of power generation (i.e. pollution), improving
voltage profile, voltage stability, and so many, to name a few (Chiradeja, 2004) Also
protection of a radial distribution system is greatly improved by placement of a DG unit in
that network. (Butler, 2009) It helps to open a new front of alternative energy when energy
crisis is at its peak. So the supply of backup power is obtained at minimum investment. But
proper location of a DG unit is one of the most important points of concern.
Various types of DG placement techniques are obtained when surveyed through the
research fraternity. Some are based on classical methods, and others on evolutionary
techniques. A mixed integer non-linear programming (MINLP) method is also proposed
by (Kaur, 2014). It points out the exact location of a DG unit with the objective of loss
minimisation. It uses combined loss sensitivity (CLS) as parameter to select the target
bus. For minimizing the power loss of medium voltage lines, a heuristic method based on
“uniform voltage distribution based constructive reconfiguration algorithm” (UVDA) is
proposed elsewhere by Bayat, Bagheri, and Noroozian (2016). Ettehadi (2012) takes into
consideration the compensation of reactive power by providing modal analysis method to
create the list. In order to visualize the effect of DG placed at a particular bus, an improved
multi-objective harmony search algorithm (IMOHS) is used.
Here, an objective function is created using power loss and voltage profile as factors
and then it is optimized (Nekooei, 2013). Particle swarm optimization (PSO) technique has
been used by Kansal, Kumar, Tyagi (2013) to solve the basic problem of exact DG allocation.
This method also minimizes the power losses. Intersect mutation differential method is
also used to select the optimal location and size of a DG unit with the ultimate objective of
cost reduction (Khodabakhshian, 2016).
There are a large number of methods available to find the optimum location of a DG
unit. As an additional benefit, they also affect the voltage profile, losses, power transfer
capacity and so on. But all these methods, whether classical or evolutionary, either suffers
from the problem of complexity, lack of convergence, or is lengthy and inconsistent. Some
lack robustness as size of the system increases. As a result, a particular method, found suitable
for a particular system, often fails to give better results when the size and complexity of
the system increases. So the problem now narrows down to shortlist some methods which
identify the exact position of DG allocation and is suitable for all types of systems.
10.2 METHODOLOGY
In this chapter, four different DG allocation techniques namely voltage stability index
(VSI), (Kayal, 2013). Voltage profile index (VPI) (Ravikumar, 2012), active power loss
Indices based Comparative Performance Analysis of Standard Test Systems  81
sensitivity factor (APLSF) (Elsaiah, 2014) and performance index (PI) (Elmitwally, 2013)
are shortlisted and selected for analysis. These methods are tested on four different systems
of variable size (i.e. 15 bus, 33 bus, 85 bus, and 123 bus). Suitability of these methods are
analysed based on their effect on voltage stability limit. As increase in voltage stability is
indicated by the increase in loadability, so these methods of DG allocation are used to find
the optimum position of a DG unit in all the four systems where loadability limit is increased
to its maximum limit. For identifying loadability limit, CPF method is used (Ajjarapu,
1992). To find the effectiveness of these methods, certain other analyses such as voltage
profile improvement analyses, voltage stability analyses, and degree of improvement of
power transfer capacity are performed. These methods give satisfactory results for systems
of variable size, ranging from a smaller to a larger one. Voltage profile improvement index
(VPII), voltage stability margin index (VSM), and power transfer capacity index (PTC) are
also found to be improved. For the above analyses, MATLAB-based power system analysis
toolbox (PSAT) is used. Two types of DG units, namely Solid Oxide Fuel Cell (SOFC) and
Photovoltaic PV cell are used.
10.3 DISCUSSION
Voltage stability analysis and DG performance analysis of 15 bus, IEEE 33 bus, IEEE 85
bus, and IEEE 123 bus distribution system is performed with the help of power system
analysis toolbox (PSAT). This toolbox works in MATLAB environment. First, using the
four DG techniques, the exact location of the DG units is obtained. Continuation power
flow technique is applied to all the four systems to find out the change in maximum loading
and hence its reflection on the stability of the system. The results are summarized in Table
10.1. Finally, three performance indices are calculated for all four systems to validate the
result.
Table 10.1 shows the value of maximum loading parameter for all the four types of
systems considered in this study. It is observed that for smaller system (i.e. 15 bus) and
the larger system (i.e. 123 buses) VSI show better results. This is because comparatively
smaller and the larger systems are sensitive to small change in voltage. On the other hand,
for medium-size systems (i.e. 33 bus and 85 bus), either PI or APLSF method emerges to
be the best in terms of increase in maximum loading. This is also reflected in Figure 10.1.
TABLE 10.1 Maximum Loading Parameter (SOFC and Photovoltaic PV Unit)
Systems No DG VSI VPI PI APLSF
15 bus 2.67 7.67 2.76 3.46 3.82
33 bus 4.01 4.46 4.03 6.56 6.45
Fuel Cell
85 bus 2.55 1.7 2.63 5.44 2.58
123 bus 5.54 5.62 5.58 5.6 5.57
15 bus 2.67 8.48 2.76 3.59 4.09
33 bus 4.01 4.53 4.03 6.36 6.45 Photovoltaic PV
85 bus 2.55 1.7 2.64 5.52 2.58 cell
123 bus 5.54 6 5.7 5.6 5.57
The figures below are the pictorial representation which depicts the change in value of the
maximum loading parameter for different methods considered. All the four systems chosen
are shown here. For the different types of DG units, namely Photovoltaic PV type and Fuel
Cell type, VSI method stands out from the lot of four for the smaller (i.e. 15 bus system) and
larger (i.e. IEEE 123 bus system) test systems.
(a) (b)
FIGURE 10.1 Comparative study of loading parameter variation for (a) SOFC and
(b) Photovoltaic cell.
Table 10.2 shows the variation of power transfer capacity for two types of DG units.
Here the percentage increase in power transfer capacity for Fuel Cell and Photovoltaic
PV unit are the same for a particular method. In this case, all the methods enhance
the power transfer capacity. This is because increase in loading of a system in turn
will increase its power transfer capacity. Maximum loading of all the systems are
improved and hence PTC. For most of the cases, PTC improvement occurs by at least
70%, and for some, it is more than 90%. Though improvement in PTC for all the
methods occur, but VSI method gives better results in all types of systems, ranging
from the smaller to the larger one. By using VSI, PTC is increased more than 90% for
all types of systems. Figure 10.2 illustrates the percentage change in power transfer
capacity for different systems. VSI acts as a pioneer in PTC improvement as compared
to the other methods.
TABLE 10.2 Power Transfer Capacity (PTC) for SOFC and Photovoltaic PV Unit
VSI VPI PI APLSF Systems
97% 90% 97% 98% 15 bus
88% 75% 83% 58% 33 bus
95% 92w% 93% 90% 85 bus
104% 98.8% 98.9% 103% 123 bus

FIGURE 10.2 Comparative study of variation in power transfer capacity for all types of DG units.
Table 10.3 depicts a comprehensive analysis of change in voltage stability margin for all
three types of DG units for all four systems. VSM mostly shows a positive change for
all the systems but variations do occur. Unlike PTC, VSM for Fuel Cell is different from
Photovoltaic PQ and Photovoltaic PV unit. VSI again dominates for 13 and 123 bus system
as evident from Table 10.3. On the contrary, for middle-sized systems, VSI shows a negative
result. Bus systems of 85 and 123 buses show more positive change when PI or APLSF
method is used. All these are pictorially represented in Figures 10.5, 10.6, and 10.7 shown.
TABLE 10.3 Voltage Stability Margin (VSM) for SOFC and Photovoltaic PV Unit
Systems VSI VPI PI APLSF
15 bus 69% 3% 26% 35%
33 bus 11% 0.5% 37% 38%
Photovoltaic PV cell
85 bus –50 3.77% 53.7% 1.18%
123 bus 1.60% 1.59% 1.08% 0.54%
15 bus 65% 30% 3% 23%
33 bus 10% 0.5% 39% 38%
Fuel Cell
85 bus –49.1% 3.04% 53.1% 1.16%
123 bus 1.42% 0.71% 0.53% 0.54%
(a) (b)
FIGURE 10.3 Comparative study of variation in voltage stability margin for (a) SOFC and
From Table 10.3 and Figure 10.3 it is evident that for all types of DG units, VSI method
shows its superiority for 15 bus and 123 bus system but for the other two, APLSF and PI
give improved results.
Variation in voltage profile is indicated in Table 10.4 and graphically portrayed in Figure
10.4. Voltage profile greater than 1 indicates that the placement of DG by that method is
fruitful else that method is not so useful. Here, though APLSF shows an improvement in
the voltage profile to a large extent (as much as four times), but improvement in the value
to such an extent is not possible. Also, this method shows variation in the 15 bus system,
where VPII less than 1. On the contrary, VSI shows an improvement in the voltage profile
value similar trend, and for all the test systems, it is greater than 1. So we conclude that VSI
is better. It uniformly improves the profile for all the test systems considered in this chapter.
TABLE 10.4 Voltage Profile Improvement Index (VPII) for SOFC and Photovoltaic PV Unit
Systems VSI VPI PI APLSF
15 bus 2.5 1.43 2 2.06
33 bus 1.25 1.12 1.81 4.36
Photovoltaic PV cell
85 bus 1.1 1.1 1.43 1.5
123 bus 1.001 1.001 1.001 1.001
15 bus 2.28 1.43 0.7 0.79
33 bus 1.12 1.12 1.63 4.36
Fuel Cell
85 bus 1.23 1.22 1 1
123 bus 1.002 1.002 1.002 1.001
(a) (b)
FIGURE 10.4 Comparative study of variation in voltage stability margin for (a) SOFC and
10.4 CONCLUSION
This chapter analyses the effect of DG allocation at different strategic points of a few
distribution systems. Initially, three different types of DG units are allotted at different
positions of distribution systems of variable sizes. Four techniques are taken into
consideration. Then the effect of DG placement is analysed by calculating certain
stability indices. It is observed that the suitability of a DG allocation technique depends
on the size of the system chosen. For a smaller and larger system, VSI method gives
better result whereas for medium-size system, PI or APLSF method shows positive
result. VSI method shows uniformity in the betterment of the performance of a system
for variable sizes.
REFERENCES
1. Ackermann, T., Anderson, G., and Lennart, S., 2001. Distributed generation: A definition.
Electric Power Systems Research 57, 195–204.
2. Akorede, M. F., Hizam, H., and Pouresmaei, E., 2010. Distributed energy resources and
benefits to the environment. Renewable and Sustainable Energy Reviews, 14, 724–734.
3. Chiradeja, P., and Ramakumar, R., 2004. An approach to quantify the technical benefits of
distributed generation. IEEE Transactions on energy conversion, 19(4).
4. Butler, P. K. L., and Funmilayo, H. B., 2009. Over current protection issues for radial
distribution systems with distributed generators. IEEE Power & Energy Society General
Meeting.
5. Kaur, S., Kumbhar, G., and Sharma, J., 2014. A MINLP technique for optimal placement
of multiple DG units in distribution systems. International Journal of Electrical Power &
Energy Systems, 63, 609–617.
6. Bayat, A., Bagheri, A., and Noroozian, R., 2016. Optimal siting and sizing of distributed
generation accompanied by reconfiguration of distribution networks for maximum loss
reduction by using a new UVDA-based heuristic method. International Journal of Electrical
Power & Energy Systems, 77, 360–371.
7. Ettehadi, M., Ghasemi, H., and Vaez-Zadeh, S., 2012. Voltage Stability-Based DG Placement
in DistributionNetworks. IEEE Transactions on Power Delivery, 28(1), 171–178.
8. Nekooei, K., Farsangi, M. M., Nezamabadi, H. P., and Lee. K. Y., 2013. An Improved Multi-
Objective Harmony Search for Optimal Placement of DGs in Distribution Systems. IEEE
Transactions on Smart Grid, 4(1), 557–567.
9. Kansal, S., Kumar, V., and Tyagi, B., 2013. Optimal placement of different type of DG
sources in distribution networks. International Journal of Electrical Power & Energy
Systems, 53, 752–760.
10. Khodabakhshian, A., Andishgar, M. A., 2016. Simultaneous placement and sizing of DGs
and shunt capacitors in distribution systems by using IMDE algorithm. International
Journal of Electrical Power & Energy Systems, 82, 599–607.
11. Kayal, P., and Chanda, C. K., 2013. A simple and fast approach for allocation and size
evaluation of distributed generation. International Journal of Energy and Environmental
Engineering, 4(7).
12. Ravikumar, V. P., Zeineldin, H. H., and Xiao, W., 2012. Determining Optimal Location
and Size of Distributed Generation Resources Considering Harmonic and Protection
Coordination Limits. IEEE transactions on power systems.
13. Elsaiah, S., Benidris, M., and Mitra, J., 2014. An Analytical Method for Placement and
Sizing of Distributed Generation on Distribution Systems. Power Systems Conference.
14. Elmitwally, A., 2013. A new algorithm for allocating multiple distributed generation units
based on load centroid concept. Alexandria Engineering Journal, 52, 655–663.
15. Ajjarapu, V., and Christy, C., 1992. The continuation power flow: A tool for steady state
voltage stability analysis. Transactions on Power Systems, 7(1), 416–423.
CHAPTER 11
Natural Ground-Water Recharge

Scenario in Kolkata and a
Review of Artificial Recharging
J. Begam, S. Sarkar, S. Sarkar, and S. Bhattacharya
Civil Engineering Department, Techno International New Town, Kolkata, India.
ABSTRACT: This abstract summarizes the preliminary results from groundwater

recharge estimates at Kolkata, West Bengal, which is under the Kolkata Metropolitan
Corporation (KMC) covering an area of 187.33 sq.km. Water is indispensable to all life on
earth. However, freshwater is constantly formed newly through a phenomenon known as
hydrological cycle. Groundwater is the source of about 33% of the water that county and
city water departments supply to households and businesses. Groundwater recharge is the
process by which water percolates down the soil and reaches the water table, either by natural
or artificial methods. Here, natural groundwater recharge is dealt with. Quantification of the
rate of natural groundwater recharge is a pre-requisite for efficient groundwater resource
management. For a region like Kolkata, there is large demand for groundwater supplies,
where such resources are the keys to economic development. However, the rate of aquifer
recharge is one of the most difficult factors to measure in the evaluation of groundwater
resources. Estimation of recharge, by any method, is normally subject to large uncertainties
and errors. In this study, various methods of estimating natural groundwater recharge in
Kolkata region (2016) are outlined taking one year precipitation and evapotranspiration
data, calculating the runoff of this region and getting the net storage capacity of soil, which
is mostly alluvial. This study estimates the amount of recharge taking place in Kolkata and
it critically reviewed with regard to their limitations and associated uncertainties leading to
the need for artificial groundwater recharge by roof-top rainwater harvesting. And finally, a
detailed design for a rooftop rain water harvesting system is provided.
Keywords: Latent heat, Hydrogeology, Rainfall infiltration, Specific yield, Gross draft,
Recharge pit, Artificial recharge.
87
11.1 INTRODUCTION
Groundwater is the underground water that occurs in the saturated zone of variable thickness
and depth, below the earth’s surface. Groundwater is water located beneath the ground
surface in soil pore spaces and the fractures of lithological formations. A unit of rock or an
unconsolidated deposit is called an aquifer when it can yield a usable quantity of water. The
depth at which soil pore spaces or fractures and voids in rock become completely saturated
with water is called the water table. Groundwater is recharged from, and eventually flows
to, the surface naturally; natural discharge often occurs at springs and seeps, and can form
oases or wetlands. Groundwater is also often withdrawn for agricultural, municipal, and
industrial use by constructing and operating extraction wells. The study of the distribution
and movement of groundwater is hydrogeology, also called groundwater hydrology. The
endless circulation of water as it moves in its various phases through the atmosphere, to
the Earth, over and through the land, to the ocean, and back to the atmosphere is known
as the hydrologic cycle. This cycle is powered by the Sun; through phase changes of water
(i.e. evaporation and condensation) involving storage and release of latent heat, it affects the
global circulation of both atmosphere and oceans, process for the fluxes of water, energy,
and the chemical elements. This cycle is the foundation of hydrological science and occurs
over a wide range of space and time scales.
11.2 METHODOLOGY
There are mainly two methods of recharge volume against rainfall data namely Water Level
Fluctuation (WLF) and Rainfall Infiltration Method (RIM).
A water budget reflects the relationship between input and output of water through
a region. The water balance graph shows precipitation and potential evapo-transpiration
both as line graphs. Thus, we have a direct comparison of supply of water and the natural
demand for water.
Water Budget equation: A general Water Budget Equation is-
P = Q + E + ΔS
TABLE 11.1 Bar Diagram of Precipitation and Evapo-transpiration
Natural Ground-Water Recharge Scenario in Kolkata and a Review  89
Calculation of Surface Run-off: For January,

For January,
Q = CIA
Q = 0.5 ´ 185 ´ 38.39 ´ 1012 [A = 185 km (Kolkata)
Q = 3.59 ´ 1015 mm3/month C = 0.5
Q = 3.59 ´ 106 m3/month I = 38.39 mm]
In a similar way, we calculate the run-off for other months
Calculation of Change in Storage:
For January,
ΔS = (P ´ A) – (E ´ A) – R
ΔS = (38.39 ´ 185 ´ 1012) – (4.88 ´ 187 ´ 1012) – (3.59 ´ 1015)
ΔS = 2.67 ´ 1015 mm3/month = 2.67 ´ 106 m3/month
Similarly, the change in storage and run-off was calculated per month.
According to water budget equation total change in storage in 2016 is
ΔS = 1.59*1017 mm3/month = 1.59*108 m3
TABLE 11.2 Total Ground Water Recharge in Kolkata 2016
Method Equation Remarks
Recharge from rainfall RRF = (A ´ WLF ´ Sy) + D – Rother A = area suitable for recharge
using water level WLF = water level fluctuation
RRF = (187.33*10002*15*7) +1200
fluctuation method
100 Sy = specific yield
RRF = 1.97*10 m
8 3
D = gross draft
RRF = 1.97*1017 mm3
RRF = 1.97*104 he–m
Recharge from RIF = NMR*A*RFIF NMR = normal monsoon rainfall
rainfall using rainfall RIF = 992.38*187.33*0.15*10 12
RFIF = rainfall infiltration factor
infiltration method
RIF = 2.79*1016 mm3 A = total area
RIF = 2.79*107 m3
RIF = 2.79*103 ha–m
Recharge from tanks RT = ASWS*days*RFact ASWA = avg. water spread area
and ponds RT = 11984353*365*1.4 RFact = a recharge factor in
RT = 6.12*106 m3 mm/d(here RFact = 1.4mm/d)
RT = 6.12*102 ha–m
Total recharge of water = 1.97 × 104 + 2.79 × 103 + 6.12 × 102

= 2.31*104 ha–m
11.3 DISCUSSION
Considering the hydrogeological condition in KMC area vis-à-vis the withdrawal of
groundwater, it is very much necessary to reduce the stress on groundwater resources.
Net rainwater available annually in KMC area is 247mil m3. In KMC area, the presence
of a thick clay layer at the top of the sedimentary sequence and the metalled and
concrete pavement in the surface restrict the rainfall recharge to groundwater in the
area. Considering hydrogeological setup of KMC area gravity head recharge tube well
using the RTR is the only recharge structure feasible. The exact depth, dimension of the
tube well and position of slots will be decided on the basis of the depth of aquifer to be
recharged and the quantum of roof top rainwater available. The recharge shafts can be
constructed in two different ways namely vertical and lateral. Vertical recharge shafts
can be further improvised with or without injection well. Recharge pits overcome the
difficulty of artificial recharge of phreatic aquifer from surface water sources. Recharge
pit is excavated sufficiently deep to penetrate less permeable strata. In case aquifers are
located below the land surface and overlain by poorly permeable strata, a recharge shaft
similar to a recharge pit but much smaller in cross section is constructed. In KMC area,
gravity head recharge tube well may be used where a huge groundwater trough has been
formed in central Kolkata around Park Street, Camac Street, Fort William, Kalighat,
Ballygungj etc., and has been demarcated in this figure.
11.4 CONCLUSION
The total annual natural Groundwater Recharge in KMC is 2.31×104 ha-m for the year
2016. Utilizing rainwater harvesting provides certain advantages to the community. First
of all, harvesting rainwater allows us to better utilize an energy resource. It is important to
do so since drinking water is not easily renewable and it helps in reducing wastage. Water
collected in the rainwater harvesting system can be put to use for several non-drinking
functions as well on an industrial scale, harvesting rainwater can provide the needed
amounts of water for many operations to take place smoothly without having to deplete
the nearby water sources. Rainwater is free from many chemicals found in groundwater,
making it suitable for irrigation and watering gardens. In fact, storing large reservoirs of
harvested water is a great idea for areas where forest fires and bush fires are common during
summer months.
Rainwater, when collected, can be used for several non-drinking functions including
flushing toilets, washing clothes, watering the garden, washing cars, etc. It is unnecessary to
use pure drinking water if all we need to use it for some other purpose rather than drinking.
During the rainy season, rainwater is collected in large storage tanks, which also helps
in reducing floods in some low-lying areas. Apart from this, it also helps in reducing soil
Natural Ground-Water Recharge Scenario in Kolkata and a Review  91
erosion and contamination of surface water with pesticides and fertilizers from rainwater
run-off which results in cleaner lakes and ponds.
But they do call for the community effort and create the spirit of co-operation needed
to subsequently manage sustainably groundwater as a community resource.
Hence, for implementation of a fruitful artificial recharge technique and for sustainable
recharge of groundwater rainwater, harvesting technique can be implemented initially from
small scale leading to a massive revolution.
REFERENCES
1. Bhattacharya, A. K., 2010. Artificial ground water recharge with a special reference to India,
Academic Research Publishing Agency Press, 4(2).
2. IOSR Journal of Humanities and Social Science (JHSS), ISSN: 2279–0837, ISBN: 2279–
0845. 5(1) (Nov.–Dec., 2012), pp, 1–5. www.Iosrjournals.Org.
3. Central Ground Water Board Ministry of Water Resources, Guide on Artificial Recharge to
Ground Water, New Delhi, May, 2000.
4. Ground Water Information Booklet Kolkata Municipal Corporation, West Bengal
5. New Methods of Artificial Recharge of Aquifers: A Review, Mahati Kavuri1, Manasa Boddu1
and Venu Gopal Madhav Annamdas. https://globaljournals.org/GJRE_Volume13/3-
Rooftop-Rain-Water-Harvesting.pdf
6. Sophocleous, M., Groundwater Recharge, University of Kansas, Lawrence, United States
7. Ground Water Scenario In India,Premonsoon, 2017 (Central Ground Water Board, Ministry
of Water Resources, Govt of India) Report of The Ground Water Resource Estimation
Committee, New Delhi, 2009.
CHAPTER 12
Performance Analysis of
Grid Connected PV system
under Varying Irradiance and
Temperature
B. Tudu, A. Biswas, A. Maji, K. Sardar, and K. K. Mandal
Department of Power Engineering, Jadavpur University, Kolkata, India.
ABSTRACT: The present study focuses on design and analysis of grid-connected PV

system under varying irradiance and temperature. The modelling and analysis of the system
are carried out in MATLAB/Simulink. The simulated model considers an array of PV panels,
with a maximum power of 100.7 kW, connected to a 25 kV medium tension (MT) power
utility grid via a DC-DC voltage booster, and a three-level, three-phase DC-AC power
converter with natural point clamped (NPC). The maximum power point tracking (MPPT)
regulator is implemented in the DC-DC converter using a sub-system block of incremental
conductance (InC) algorithm. An efficient control strategy has been considered in inverter for
stable operation of the system. An LC filter is also used to remove the harmonics in inverter.
The active and reactive powers injected to grid along with grid voltage have been observed.
Other parameters such as output voltage and current of PV, DC link voltage, inverter output
current, and voltage without filter and after filtering, grid injected voltage and current at
point of common coupling (PCC). have been observed. However, system characteristics
change when it is under varying irradiance and temperature. Thus, performance of the
system is also reported under varying weather conditions. The performance of the system
under steady-state condition is found to be satisfactory. It has been observed that the power
output from PV system solely depends on the solar irradiance and temperature. The DC-
link voltage of PV system remains fixed with the help of controllers. Furthermore, it can be
concluded that the MPPT technique plays a significant role to improve power extraction.
Keywords: DC-DC boost converter, Grid connected system, MPPT, PV, Incremental
conductance method.
93
12.1 INTRODUCTION
The transition from coal-based power plants to renewables-based power plants is aggravated
due to various factors such as depleted coal reserve, vulnerable environment, energy
security, sustainability. PV and wind--based energy systems have become the major players
in the world of power generation. The PV and wind are currently the most developed
renewable energy sources (RES) being scalable to big commercial power plants (Tudu,
2019). The PV is more reliable and its installation does not involve complexity. Moreover,
the system has very low operating costs and limited maintenance requirements. The wind
power system has also very low operation costs, limited maintenance requirement, and is
reliable. However, wind power system has some drawbacks over PV system such as the
installation is not as easy as the PV system, and it is not silent too. Nevertheless, the wind
turbine is capable of producing huge power (MW range) with moderate cost. Hence, wind
and solar system can be integrated to supply power to local load or can be transferred to the
grid. With proper control, hybrid renewable systems are superior in terms of quality power
delivery compared to single renewable source.
The production cost of the PV cells and arrays are decreasing day by day due to
revolution in semi-conductor and panel fabrication technology (Mancilla-David, 2012). The
power output from the PV cells solely depends on the temperature and solar irradiation. A
maximum power point tracking (MPPT) circuit is integrated to maximize power extraction
(Zhou, 2015). The output voltage of PV array is relatively low, but grid integration requires
higher voltage output at the PV terminals. Hence, a DC-DC boost converter is used to
achieve the required voltage level (Amir, 2018). The DC output voltage is converted to
AC by the use of inverter. The DC link voltage is kept constant to its reference value by
controlling the grid injected current. This is required for regulating the active and reactive
power injection to the grid (Ravi, 2011). An optimized and state-of-the-art control strategy
is required to get the best performance from the system.
Various researchers have reported several works on the design, modelling, and control
of grid-connected solar PV systems. A review on solar PV emulators has been presented by
Ram et al. (2018). The technical requirements for grid connection of PV system with inverters
have been discussed by Mundada et al. (2016). The authors further analysed the grid codes,
grid coupling applications, and presented the solutions to the barriers for plug and play
the PV systems. The parameters specification and standards of photovoltaic inverter and
its topologies have been reported by Jana et al. (2017). Moreover, the authors emphasized
on the analysis of solid-state switching inverters, multi-level inverter, transformers, and
different types of interconnections, and the cost of implementation. A similar study has been
performed by Eltawil et al. (2010) focusing on the connected load ratio, grid-connected PV
inverters and possible difficulties in grid synchronization. A review on the categorization of
islanded and grid-connected PV system, various energy management approaches, MPPTs,
and economic and environmental issues have been presented by Lupangu et al. (2017). Wu
et al. (2017) investigated the guidelines for PV integration to grid, effects of high penetration
of solar PV to grid along with protection methodologies.
Performance Analysis of Grid Connected PV system under Varying  95
The power generation form renewable-based energy systems and their integration
to grid have huge potential that comes with lot of challenges. Hence, different aspects of
these sources must be studied categorically. The design, control, and issues of integration to
grid of solar PV system urge immediate attention. Accurate characterization and optimal
control of solar PV system can be achieved by analysing the systems under different
operating conditions. Hence, this work focuses on the performance of grid-connected solar
PV system under different operating conditions.
12.2 METHODOLOGY
The system performance of a grid-connected solar PV system is considered. The system
is simulated in MATLAB/Simulink environment. The system has an array of PV panels
of 100.7 kW at irradiance of 1,000 W/m2 and a temperature of 25°C, and it is connected
to a 25 kV medium tension (MT) power grid through a DC-DC voltage converter, and
a three-level, three-phase DC-AC power converter with natural point clamped (NPC).
The DC-DC voltage converter operating at modulation frequency of 5 kHz increases the
voltage level from 273 V to 500 V DC. The Incremental Conductance (InC)-based MPPT
algorithm is implemented in the DC-DC converter. The MPPT regulator automatically
varies and optimizes the filling factor of the DC-DC voltage booster to maximize the power
output at the desired voltage level. The three-phase voltage source converter (VSC) having
modulation frequency of 1.98 kHz converts 500 V continuous voltage to voltage of 260V at
unity power factor.
FIGURE 12.1 Simulink model of grid-connected solar PV system.
However, the alternating output from inverter contains harmonics and an LC filter is
connected at the output terminals of inverter to remove the harmonics. The filter contains
a 250 μH inductor, and a capacitor of 10 kVAr. The output voltage is then boosted up by
using a three-phase 100 kVA, 260V/25kV transformer before finally feeding the power to
grid. A medium-tension distribution network of 25 kV is considered for grid feeding. The
network of 60 Hz is considered for the Simulink implementation. The detailed Simulink
model is shown in Figure 12.1.
12.3 DISCUSSION
Initially, the system performance is evaluated at fixed value of solar irradiance and
temperature. The model is run in standard condition at the solar irradiance of 1 kW/m2
and the temperature of 25°C represented by signal builder block of Simulink library. The
output voltage and current from PV array are shown in Figures 12.2 and 12.3, respectively.
PV Output Voltage
400
350 PV Voltage
300
250
200 280
Vpv
330
150 280 260
320
260 240
100 310
0.01 0.02 240 0.2 0.25 0.3
50 Open-circuit With MPPT

220
0.1 0.14 0.18
0 Without MPPT
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.2 PV array output voltage.
PV Output Current
500
PV Current
400
300
Current (A)
200
100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.3 PV array output current.

The pulses to DC-DC boost converter and VSC converter are fed at the time of t = 0.02 sec.
Hence, during the simulation time period from t = 0 sec to t = 0.02 sec, the PV output voltage
is open-circuit voltage and corresponding current is zero as shown in Figures 12.2 and 12.3.
During this time duration, the power extraction is without applying MPPT algorithm. At
the time of t = 0.2 sec, the MPPT controller starts its action and controls the output voltage
by varying the duty cycle to extract the maximum power.
The power output of PV array is 96 kW until MPPT is enabled, that is, during
t = 0 sec to t = 0.2 sec. Once the MPPT controller is applied, the output power reaches to
100.2 kW, which is approximately the specified (100.7 kW) maximum power as depicted in
Figure 12.4. The output voltage from DC-DC boost converter is regulated by the DC-linked
voltage regulator as illustrated in Figure 12.5. The output voltage is 500 V which is equal
to the input reference voltage Vref = 500 V of the regulator. After MPPT is applied at time
t = 0.2 sec, the regulated voltage observes little fluctuation and deviates from its reference
point mainly due to the switching transient.
PV Output Power
140
PV Power
120
100
80
Power (kW)
60
40
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.4 PV array output power.
DC Linked Volatge
700
Boosted Voltage
600
500
400 510
Voltage (V)
300
500
200
100 490
0.18 0.2 0.22 0.24
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.5 DC-DC converter output voltage.

The output voltage of inverter has been observed before inserting the filter circuit and shown
in Figure 12.6. The conversion from dc to ac has been done with the help of three-level,
three-phase voltage source converter (VSC). However, VSC contributes some harmonics,
and these harmonics are filtered with an LC filter. The series connected inductance is of
250 µH and capacitor bank has capacitive reactive power of 10 kVAr along with its internal
resistance of 0.01885 Ω. Figures 12.7 and 12.8, respectively, depict the output three phase
voltages and currents of the inverter after filtering with LC filter.
Inverter Output Voltage
600
400
200
Voltage (V)
-200
-400
-600
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.6 Inverter output voltage without filtering.

Inverter Output Voltage
200
300
200 0
100
-200
Voltage (V)
0 0.1 0.11 0.12 0.13 0.14
-100
-200
Phase A
Phase B
-300 Phase C
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (s)
FIGURE 12.7 Inverter output voltage after filtering.

Inverter Output Current
0.4
0.2
0
0.2
-0.2
Current (kA)
-0.4
0 0.18 0.2 0.22 0.24
-0.2
Phase A
Phase B
Phase C
-0.4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (s)
FIGURE 12.8 Inverter output current after filtering.

The output line voltage from inverter of 254.6 V must be boosted up to synchronize with
the grid of voltage of 24.49 kV. Hence, a step-up delta-star transformer is employed to
increase the voltage magnitude. Figures 12.9 and 12.10 represent the grid current and grid
voltage curve, respectively. At the time of starting the simulation, it has been observed that
some switching transients as shown in Figures 12.9 and 12.10. However, these transients
are eliminated by adjusting the values of LC filter parameters. The injected grid current
and voltage reach their steady-state values approximately after 0.06 sec and the steady-state
current of 4.05 A is injected into the grid.
Grid Current
5 4
-4
0.2 0.21 0.22 0.23
Current (A)
Phase A
Phase B
Phase C
-5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (s)
FIGURE 12.9 Grid injected current at PCC.
Grid Current
20
20
10
-20
Voltage (kV)
0.1 0.105 0.11 0.115 0.12

0
-10
Phase A
Phase B
-20
Phase C
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (s)
FIGURE 12.10 Grid voltage at PCC.
The active power injected into the grid is 98.75 kW. Hence, a loss of approximately 1.5 kW
power is observed in various system components. The losses are taking place mainly in
inverter, transformer, and filter. The active power injected into the grid is shown in Figure
12.11, and it is seen that at 0.2 sec, maximum power is being extracted from solar PV system.
Figure 12.12 shows the reactive power injected to grid for the duration of simulation.
The reactive power is zero at steady state. This is due to a consideration of reference q-axis
current Iq of the current regulator of zero value. However, during switching transient period,
the PV system is drawing some reactive power.
Active Power
150
100
Power (kW)
50
-50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.11 Grid injected active power at PCC.

Reactive Power at PCC
15
Reactive Power Flow
Reactive Power (kVAR)
-15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
FIGURE 12.12 Grid injected reactive power at PCC.
The analysis of grid-connected solar PV system until now has been done for a fixed irradiance
and temperature value. However, the output from PV system depends on solar irradiance
and cell temperature. The PV panel output linearly varies with the irradiance and inversely
proportional to the temperature. A variation in solar irradiance for the simulation period of
2.5 seconds has been considered as shown in Figure 12.13 keeping the temperature at 25°C.
The corresponding power output from the PV panel is shown in Figure 12.14. This figure
demonstrates the linear variation of power output from PV with solar irradiance.
FIGURE 12.13 Solar irradiance in different time durations.

FIGURE 12.14 Power output for irradiance variation.
A similar change in output current from PV is seen with respect to irradiance as shown
in Figure 12.15. Figure 12.16 shows the output voltage of the PV array and the DC-link
voltage supplied by the boost converter. Figure 12.17 shows the grid injected active power
and reactive power. It has been observed that the active power changes in accordance to
irradiance; however, the reactive power remains fixed at zero.
FIGURE 12.15 PV output current for irradiance variation.
FIGURE 12.16 PV and DC link voltage for irradiance variation.

FIGURE 12.17 Grid injected active and reactive power.
Figure 12.18 shows the variation in temperature with respect to time. Here, the irradiance
is kept constant at 1000 W/m2 and analysis is carried out to realize the impact of the
temperature on the PV panel.
FIGURE 12.18 Temperature variation with time.
The power output gets affected by the change in temperature, and it can be observed in
Figure 12.19. However, the change in power output is not that much prominent as in case
of power output for variation in solar irradiance. Likewise, the temperature change directly
affects the output voltage. Here, the change in current output is less affected unlike voltage
changes as shown in Figure 12.20.
FIGURE 12.19 Power output for temperature variation.

FIGURE 12.20 PV output current and voltage with temperature variation.
12.4 CONCLUSION
A grid-connected solar PV system has been designed and the impacts on different
performance parameters for change in solar irradiance and temperature have been analysed.
The system is designed and simulated in MATLAB/Simulink environment. Initially, the
performance of the system has been analysed under normal conditions for fixed values
of solar irradiance and temperature. The changes in various parameters such as output
voltage, current, and power have been highlighted for change in irradiance and temperature
values. The major findings of this work can be summarized as follows:
• The MPPT plays crucial role in extracting power.
• The optimal design of filter can smooth out the harmonics contained in voltage
and currents.
• Change in irradiance largely and linearly affects the output current and power.
However, the change in voltage is small.
• The variation in temperature changes the output voltage and power keeping current
almost fixed.
The performance of the system under steady-state condition is found to be satisfactory.
However, the dynamic performance of the system under different abnormalities like under
fault conditions needs to be assessed and is considered in future work scope.
ACKNOWLEDGEMENT
Authors would like to acknowledge and thank Jadavpur University for extending the
required support and facility to carry out this particular research work. It is also to be
mentioned that the present work is financially supported by the Major Research Support
of JU RUSA-2.0 to faculty members for project titled ‘Dynamic response analysis of grid
connected and stand-alone PV based renewable energy system’.
REFERENCES
1. Amir, A., Amir, A., Che, H. S., El Khateb, A., and Rahim, N. A., 2018. Comparative analysis
of high voltage gains DC-DC converter topologies for photovoltaic systems. Renewable
Energy.
2. Eltawil, M. A., and Zhao, Z., 2010. Grid-connected photovoltaic power systems: Technical
and potential problems—A review. Renewable and sustainable energy reviews, 14(1),
112–129.
3. Jana, J., Saha, H., and Bhattacharya, K. D., 2017. A review of inverter topologies for single-
phase grid-connected photovoltaic systems. Renewable and Sustainable Energy Reviews,
72, 1256–1270.
4. Lupangu, C., and Bansal, R. C., 2017. A review of technical issues on the development of
solar photovoltaic systems. Renewable and Sustainable Energy Reviews, 73, 950–965.
5. Mancilla-David, F., Arancibia, A., Riganti-Fulginei, F., Muljadi, E., and Cerroni, M., 2012.
A maximum power point tracker variable-dc-link three-phase inverter for grid- connected
PV panels. In 2012 3rd IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe)
(pp. 1–7). IEEE.
6. Mundada, A. S., Nilsiam, Y., and Pearce, J. M., 2016. A review of technical requirements for
plug-and-play solar photovoltaic microinverter systems in the United States. Solar Energy,
135, 455–470.
7. Ram, J. P., Manghani, H., Pillai, D. S., Babu, T. S., Miyatake, M., and Rajasekar, N., 2018.
Analysis on solar PV emulators: A review. Renewable and Sustainable Energy Reviews,
81, 149–160.
8. Ravi, A., Manoharan, P. S., and Anand, J. V., 2011. Modelling and simulation of three
phase multilevel inverter for grid connected photovoltaic systems. Solar Energy, 85(11),
2811–2818.
9. Wu, Y. K., Lin, J. H., and Lin, H. J., 2017. Standards and guidelines for grid-connected
photovoltaic generation systems: A review and comparison. IEEE Transactions on Industry
Applications, 53(4), 3205–3216.
10. Zhou, T., and Sun, W., 2015. Study on maximum power point tracking of photovoltaic
array in irregular shadow. International Journal of Electrical Power & Energy Systems, 66,
227–234.
CHAPTER 13
Development of a Low Cost

Autonomous Car Parking
System: Towards Smart City
Arjun Dutta,1 Ankur Bhattacharjee,2 and Abhijit Kar Gupta3
1
Department of Computer Science Engineering, I.E.M Kolkata, India.
2
Department of Electrical and Electronics Engineering, BITS Pilani- Hyderabad Campus, India.
3
Department of Basic Science and Humanities, I.E.M Kolkata, India.
1
arjuncode47@gmail.com, 2a.bhattacharjee@hyderabad.bits-pilani.ac.in, 3abhijit.kargupta@iemcal.com
ABSTRACT: In this modern world, the paradigm of smart cities have gathered major
popularity. Thanks to the evolution of microcontrollers, the notion of smart city seems
more convincing. This study proposes a novel algorithm that helps users to find a free
parking space for cars automatically. Microcontrollers like Arduino has already made
a huge impact on learning. The widespread affirmation gave a new life to this open
source hardware component which potentially became a big challenge to many industrial
products and new interests in electronic system and prototyping. The automatic parking
system aims to enhance the comfort and safety of driving in constrained environments
where much attention and experience are required to steer the car. This paper deals with
two major objectives: anautomated parking system and automated vehicle. The first
objective is to design a low-cost microcontroller-based model of a car whose working
is to park the car by itself when free space is obtained. The second objective is related to
automated car parking garage setup. The speciality of the system is the use of efficient
low-cost sensors such that the two systems are embedded together working with single
power source. The sensors are fixed in the track through which the cars move into the
garage or when moving out from the garage. Our proposed system, however, forecasts
upon developing an intelligent, user friendly automated car parking system that not only
reduces manpower but also reduces traffic congestion, facilitating secure parking system
within a limited area. A comparative study of the performance of the ultrasonic sensor
upon different atmospheric conditions has also been provided.
Keywords: Arduino, Autonomous, Congestion, Parking, Smart city, Ultrasonic.

105
13.1 INTRODUCTION
With the growth of robotics and intelligent systems, automation has gained importance
in the modern era. Initially, the concept started with the identification of communication
devices where the devices are connected via Internet, controlled by remote computers and
independently. Rather than the traditional use of manpower, controlling and monitoring
in/out of the car inside the car parking lot and to check whether there is a safe place inside
the garage or not is driven by automatic means as referred in Figure 13.1. This study aims to
solve the above-mentioned problem along with the architecture of car performing automatic
parking. Moreover, the proposed system is controlled by microcontroller Arduino IDE and
is faster, flexible, and can meet market needs.
Similarly (Pham, 2015) proposed an algorithm which reduces the average waiting
time taken by the users for parking their car where the architecture is based upon
Internet of things (IoT) technology. Wei (2014) proposed an automatic parking concept
using 89c52 microcontroller, IR sensors for counting the number of cars in/out of the
parking zone. Wei configured the system in a way such that the number of cars present
in the parking lot is displayed with the help of display unit, and the lock gate through
which car moves in/out becomes restricted if more than adequate amount of cars
reaches the lane of the parking lot. With the enhancement of sensors and IoT, curiosity
towards ‘smart city’ turned into possibility. Changes are taking place with respect to
traffic management system as well as in car parking manoeuvring systems (Zhou, 2014).
Sensing and computing are two important features with which electronic devices grow
smarter in this modern trend. The availability and robustness of Cloud technology
allow developers to create and explore in this domain. Cloud can be considered as the
best friend of IoT through which sensor data can be stored and accessed from distant
locations. However, this gives rise to the combination of two technologies which gave
birth to a new technology known as Cloud of Things (CoT) which can be accessed,
controlled, and monitored from distant locations through cloud. Due to its high
accuracy and robustness, nodes can be added or deleted from IoT upon real-time basis.
Thus IoT can be defined as the combination of embedded object, controllers, sensors,
and Internet (Fox, 2012, Khanna, 2016).
Basavaraju (2015) implemented a smart parking system which consists of the
following components centralized system, raspberry pi, pi-cam, display device, and user
tool which uses access to the map status of parking slots from distant areas. This saves
wastage of time, avoids travel across filled car parking lot, and is a cost-effective measure.
Albagul (2013) proposed a multi-level advanced car parking module with a well-
fabricated control system responsible for the entrance and exit of cars along the passage
of the garage system. The system is driven by a PLC controller and some sensors
which consist of an elevator system for the passage of cars to available free spots in
different floors. Papacostas devoted a design of parking system based on control system
paradigm. The proposed control system paradigm acts as an important aspect for the
passage of cars in/out of the parking spot. However, it also depicts a low-cost multi-
Development of a Low Cost Autonomous Car Parking System  107
level parking lot which allows the provision for more cars in a small area (Papacostas,
1993). As we know, an ancient proverb – ‘necessity is the mother of invention’, hence
technologies nowadays have been man’s best friend. In the modern era, parking space
for vehicles especially cars has been the cause of big trouble. It is quite important to
reduce the wastage of space and use technologies for man’s better living with ease.
Hence, the architecture of multi-level automatic car parking lot plays a major role for
sustainable growth in the future (Reza, 2012).
13.2 METHODOLOGY
The developed prototype has two parts combined together as referred in Figure 13.1.
First objective depicts the function of automatic car parking system. Second objective
illustrates the development of a low-cost autonomous vehicle. The speciality of the system
is such that the two systems are embedded together which can work with only one power
source. For the accurate measurement of the control system, voltage regulators are used.
This study depicts the use of low-cost sensors which also adds a positive fact towards our
objective.
The sensors are fixed in the track through which the cars will move into the garage or
when going out from the garage. When the sensors attached to the cardboard senses the car,
the lock gate opens and allows the passage for car parking. This study illustrates the use of
seven-segment display, to display ‘F’, i.e. full, if the parking lot consists of more than nine
cars. The system allows total allotment of nine cars in the parking lot.
Initially, we can show the parking of cars in the parking lot manually, but one car is
also built in such a way that it uses ultrasonic sensors to detect its obstacles and park in
the parking lot according to its adequate space. Finally, when the car will find adequate
space for parking, it will automatically park by itself. Block diagrams 1 and 2, illustrates the
systematic design of the developed architecture.
FIGURE 13.1 Developed prototype of a low-cost automated car parking system.

Block diagram of the developed automatic car parking garage
FIGURE 13.2 Block diagram 1.
13.2.1 Explanation
Arduino is an open hardware development board which depends on a microcontroller that
makes installed programming a lot simpler than conventional strategies. The regulator
present in the system is used to lower the voltage from 12V to 5V. The seven-segment
display is used to display the digits from 0 to 9 and ‘F’. The digits being displayed signifies
the number of cars parked in the garage. The alphabet ‘F’ signifies full condition which
states no space available in the parking lot. Here, the servo motor is used to pull up and
down the gate when cars pass through the track. Two switches act like a pressure sensor,
when the cars pass through the track, the switches sense the pressure and the lock gate
opens and closes simultaneously as referred in Figure 13.1.
Block Diagram of the developed automated car.
FIGURE 13.3 Block diagram 2.
13.2.2 Explanation
To estimate the minimum error in the system, a regulator is used to convert 12V DC to
6V. This voltage is gained by H-Bridge to run the forward and back motors present in the
base of the car. The forward motor is used to steer the front wheels and the back motor is
responsible for generating power to the car system for its movement. And the ultrasonic
sensor is used to detect obstacles. The Arduino board used is well programmed such that
when the car finds free space, it parks automatically.
13.3 DISCUSSION
As referred in Figure 13.1, the ultrasonic sensor is present in the Arduino-based automated
vehicle. The speed of the values obtained from the ultrasonic sensor differs with respect
to the temperature of the surroundings according to the equation stated below, shown in
Table 13.1.
(Youngtae Jo, 2014) ultrasonic speed = 331.5 m/s + (0.61 × temperature)
TABLE 13.1 Shows the Day Time Variation of Speed of Ultrasonic Sound with Respect to Recorded
Air-Temperature during the Month of September till January 2018 in Kolkata Region, India
Month Temperature (·C) Ultrasonic Speed (m/s)
15 September
th
34 352.24
15st October 32 351.02
15th November 29 349.20
15th December 27 347.98
15th January 26 347.36
25th January 24 346.14
The prototype was built at a low cost; Bhattacharjee and Gupta have worked hard to meet
the probable results, but due to systemic error, the automated vehicle parks itself at proper
estimated place, but at some distance forward which can cause collision with other vehicles.
The error is calculated and stated accordingly in Table 13.2.
TABLE 13.2 Error Analysis of Over Traversed Distance
Actual Distance Observed Distance Error %
1.5 1.8 20.0
1.8 2.0 11.1
2.1 2.4 12.5
2.4 2.7 12.5
2.7 3.0 12.5
Actual Distance = ld, Distance travelled = lt

Error % = ((ld – lt)/lt)*100
Average Error % = (20.0 + 11.1 + 12.5 + 12.5 + 12.5)/5 = 13.72
FIGURE 13.4
Figure 13.4 depicts the graphical representation of the error that occurred due to more
displacement of the car model after detecting adequate free space.
Therefore, the proposed model facilitates the following:
A. This system facilitates easy inlet and outlet of cars from the garage.
B. In the developed setup, the total number of cars which can fit in the parking lot is
nine.
C. The developed car parking lot setup is suitable in offices, malls, and in congested
residential areas.
D. Much lower maintenance is required for the developed car parking lot system.
E. Facility towards maximum automation, with cheaper price.
F. No necessity of line of sight operation.
13.4 CONCLUSION
The concept of smart city has been a dream for humanity. Since the past decade, researchers
have been in a quest for building smart cities as a reality. The growth of robotics and
cognitive systems have introduced new possibilities in terms of smart cities.
Our proposed study of parking system improves performance by reducing the users
who fail to park at a proper parking place. Our proposed model has been successfully tested
and designed for use in real-life situations. Our proposed approach yields fast results such
that anyone can find suitable place for parking and saves time in doing so. As Arduino is
among the trending technologies, using it makes our proposed work unique.
The electronic model has been designed, and the software as well as the control circuit
has been implemented successfully. The main advantages are optimization of space, cost,
and security.
The overall system is designed at low cost, hence, adds a positive factor towards the
developed prototype. The working and simulation of our developed system achieved
an optimal solution when the vehicles find a free space for parking. Hence, the average
waiting time as well as the total time taken by each car for performing parallel parking is
reduced. Overall, the designed architecture is simple, economic, and commences towards
a sustainable solution to reduce carbon footprints to atmosphere. We will consider the
security parameters as well as implement our proposed system in the real world as our
further study.
REFERENCES
1. A Cloud-Based Smart-Parking System Based on Internet-of-Things Technologies
2. Pham, Thanh Nam and Tsai, Ming-Fong, 2015. Digital Object Identifier 10.1109/
ACCESS.2015.2477299, 3, 2015, Received July 24, 2015, accepted August 16, 2015, date of
publication September 9, 2015, date of current version September 23, 2015.
3. Wei, Wang Guang, Study on Automated Car Parking System Based on Microcontroller,
International Journal of Engineering Research & Technology (IJERT), ISSN: 2278–0181,
3(1), January 2014.
4. Zhou, F., and Li, Q. (2014, November). Parking Guidance System Based on ZigBee and
Geomagnetic Sensor Technology. In Distributed Computing and Applications to Business,
Engineering and Science (DCABES), 2014 13th International Symposium on (pp. 268–271).
IEEE.
5. Fox, G. C., Kamburugamuve, S., and Hartman, R. D. (2012, May). Architecture and
measured characteristics of a cloud based internet of things. InCollaboration Technologies
and Systems (CTS), 2012 International Conference on (pp. 6–12). IEEE
6. Khanna, Abhirup, IoT based Smart Parking System, 2016 International Conference on
Internet of Things and Applications (IOTA) Maharashtra Institute of Technology, Pune,
India, 22 Jan–24 Jan, 2016.
7. Basavaraju, S. R., 2015. Automatic Smart Parking System using Internet of Things (IOT),
International Journal of Scientific and Research Publications, 5(12), December 2015, ISSN:
2250–3153.
8. Albagul, A., Design and Fabrication of an Automated Multi-level Car Parking System,
Manufacturing Engineering, Automatic Control and Robotics, Published June 2013,
ISBN: 978-960-474-371-1.
9. Papacostas, C. S., and Prevedouros, P. D., 1993. Transportation Engineering and Planning,
2nd Edition, Prentice Hall, Englewood Cliffs, New Jersey.
10. Reza, M. O., Ismail, M. F., Rokoni, A. A., Sarkar, M. A. R., 2012. Smart Parking System with
Image Processing Facility, I. J. Intelligent Systems and Applications, 3, 41–47.
11. Jo, Youngtae, and Jung, Inbum, 2014. Analysis of Vehicle Detection with WSN-Based
Ultrasonic Sensors, Sensors, 2014, 14, 14050–14069. doi: 10.3390/s140814050I.
CHAPTER 14
Monitoring Static Security

Assessment in Its Full Scope
Using Common Artificial Neural
Network
S. K. Tiwary, J. Pal, and C. K. Chanda
Electrical Engineering Department, Indian Institute of Engineering Science and Technology, Shibpur, India.
ABSTRACT: Monitoring power system in real-time is a complicated and paramount task.

The power system in real-time behaves very spontaneously, whose behaviour is difficult to
understand for a person who is not acquainted with a given power network. The traditional
methods of using load-flow studies to ascertain the state of the power network sometimes
seems obsolete to power system operators in the current scenario. This work presents an
alternate method for real-time power system security-assessment and monitoring using a
network of common artificial neural network. A common ANN, as elaborated in this work,
is a neural network that monitors several power system parameters with similar units. In
simple terms, it is a multiple input, multiple layer ANN. In this proposed method, all the
important load-flow parameters of the power network, such as active-power, reactive-
power, bus voltage-magnitude, and bus voltage-angles are monitored using a common
ANN. To test the feasibility of the work, the proposed method was simulated in the Simulink
environment of MATLAB R2016b (academic license) with Hardware-in-the-Loop device
of RT Lab OP5600 Simulator. To test the effectiveness and response time of the common
artificial neural networks (ANNs), a critical and a non-critical contingency, with respect
to the active-power and bus voltage-angles was induced in the IEEE-14-bus test system.
Similarly, a critical and a non-critical contingency, with respect to the voltage security
and reactive power security, was also induced at different time-interval. The results are
presented in this study for the readers to peruse.
Keywords: Artificial neural network, On-line monitoring, Power system security, Power
system protection, Power system stability.
113
14.1 INTRODUCTION
With the rapid improvements in the field of computers, software, machine intelligence, and
learning in the last three decades, engineers and researchers everywhere are looking for a
smarter and astute heuristic solution in every field of engineering and technology. Machine
intelligence and artificial neural networks have been extensively applied to the field of
power system, especially in power system security since the last three decades (Tiwary,
2017). There are many literatures available that show the feasibility of the real-time/on-line
applicability as well as off-line/non-real-time applicability of the artificial neural networks
(ANNs), specifically to the field of power system security (Panciatici, 2012).
In power systems all over the world, performing load-flow studies are routine work that
are a pivotal task of paramount emphasis (Tiwary, 2017). Load-flow studies are customarily
performed in all power network utility all over the world, at frequent intervals to determine
the state of the power network at the given instant (Tiwary, 2017). There are various load-
flow programs that can be run to determine the system-state, such as Newton–Raphson
load flow, Gauss–Seidel load flow, Fast–Decoupled load flow, DC load flow (Kumar, 2018).
But all these load-flow programs have their drawbacks that may prove to be a catastrophic
problem in the real-time monitoring of the power networks. Newton–Raphson method
is the most accurate load-flow program, and the total number of iterations required to
arrive at an acceptable result using it is very less, but it takes a lot of time for each iteration,
making it non-feasible for real-time application purposes. Gauss–Seidel method may prove
to be problematic in real-time contingencies because of the time taken for finding out
the appropriate acceleration factor for faster convergence and results. In Fast–Decoupled
method, more numbers of iterations are needed to arrive at acceptable results. And DC
method is also called an approximate method as it does not provide very accurate results
and provides only the results corresponding to active power and voltage angles, but it
can provide them faster. All in all, there is no perfect solution to monitoring the state of
the power network using the load-flow studies (Saeh, 2008). To surmount these above-
mentioned drawbacks of the load-flow methods, artificial neural networks can be applied,
as substantiated in this study.
In this work, it is demonstrated how an ANN can be utilized to monitor all the aspects of
the power network’s load-flow parameters and provide results much faster. Here, a common
(multiple input multiple layer) ANN is used to monitor each parameter, namely active-
power (MW), reactive-power (MVAR), voltage magnitude, and voltage angles, separately.
The methodologies, application, advantages, and drawbacks are elaborated in the following
sections.
14.2 ANN DEVELOPMENT AND APPLICATION TO TEST-BUS-NETWORK

As explained in a previous section, a network of multiple inputs, multiple layer neural
networks, referred in here as a common ANN, is used to monitor the four most crucial
parameters of a power-system. These parameters are active-power (MW), reactive-
Monitoring Static Security Assessment in Its Full Scope Using Common  115
power, (MVAR), voltage magnitude, and voltage angles, as mentioned previously. The
whole set up of the common ANN monitor, as applied to an IEEE-14 bus system, can be
demonstrated as in Figure 14.1.
FIGURE 14.1 Common ANN monitoring the power system load-flow parameters.
The ANNs are developed using extensive data for training and development purposes (Jain,
2005). These extensive training datasets can be calculated utilizing the day ahead load forecast,
historical data records, seasonal temperature effects on the transmission grids, the maximum
and minimum variation of the respective parameters in the past years, and other such crucial
factors which may play a detrimental role in the event of an unanticipated contingency
scenario, in a real-time power system. Using the above constraints, an ample amount of
training datasets is calculated and tabulated in separate Microsoft Excel sheets, for the active-
power and reactive-power of each line, as well as voltage magnitudes and voltage angles of
each bus, respectively (Tiwary, 2017). In all, there will be four training datasets/excel-sheets,
as we need to develop four different ANN models: one excel sheet each for training datasets
of active-power flowing on each line, reactive-power flowing on each line, voltage magnitude
at each bus, and voltage angles at each bus, respectively. Since the ANN used here is a feed-
forward-backpropagation ANN, which is a supervised learning type ANN, we need to set a
target classification for the ANNs and its training data. Here the target or output classification
will be of only two types, namely secure state (numerically classified as ‘0’ [zero]) and insecure
state (numerically classified as ‘1’ [one]). It means that if the common ANN monitoring the
active power flowing on the whole power network gives an output of ‘0’ in real-time, then
the active power flow on the whole power network is in a secure or stable state, and if the
common ANN gives an output of ‘1’, it means the active-power flow on the whole power
network is in an insecure or unstable state. To test the contingencies in the power system,
‘null’ was also given as an input within the training data. After all the datasets and its target
classifications are set up, the ANN is initialized with as many numbers of inputs as are the
parameters to be monitored, a certain number of hidden layers and neurons therein, and the
training algorithm (Nakawiro, 2008). Then the ANN is trained with the training dataset and
its convergence is verified using some testing data. Once the ANN converges, its Simulink
(MATLAB) models can be developed, which can then be used to monitor the power network
in real-time as demonstrated in Figure 14.1. The diagram of the Simulink model is shown in
Figure 14.2.
FIGURE 14.2 Common ANN monitoring the power system load-flow parameters.
The model shown in Figure14.2 was simulated in SIMULINK environment of MATLAB

R2016b in a networked mode, with an academic license. To verify the applicability
and responsiveness of the common ANNs, four different outages were induced on the
network and its respective effects on the whole power network were monitored and
recorded for analysis. To assess the performance of the common ANN monitor, four
contingencies were performed. One contingency which is severe with respect to MW
security and another which was non-severe for the MW security was induced at different
times. Similarly, a third contingency, which was severe with respect to voltage security,
as well as the fourth contingency, which was non-severe with respect to voltage security,
was also induced. These contingencies were induced at different times. The results are
discussed in the following section.

As explained in Section 14.2, the power system was simulated, and different line outages
were induced. As per the MW security assessment results of power system using the
Newton–Raphson power flow algorithm (Ding, 2015; Bahmanyar, 2014), one of the
most severe contingencies, with respect to MW security for IEEE-14-Bus system, is
for line 17, and the least severe contingency is for line 10. Similarly, one of the most
severe contingencies, with respect to voltage security assessment using the Newton–
Raphson (NR) power flow algorithm, for IEEE-14-Bus system, is line 3, and the least
severe contingency is line 10. To test and verify the performance of the common ANN
monitoring, the previously mentioned four parameters, outages were induced in the
above mentioned, three lines one after the other. Since line number 10 is an isolated line
connecting an isolated bus, it has the least severe repercussions on the MW security or
Voltage security of the network under study.
The simulation of the model was run for 30 minutes or 1,800 seconds (Khazaei, 2014;
Wand, 2014). The contingency of line 17, the most severe contingency with respect to MW
security using the NR method, was induced at 10 minutes or 600 seconds, and the response
of the common ANN is shown in Figure 14.3.
FIGURE 14.3 Response of common ANN monitoring the MW-power flow on all lines.
Since, the MW power and the voltage angles are linked, also MVAR power and voltage
magnitude are linked to each other, the responses of the common ANN’s monitoring
them are also similar. It can be seen from the above response that the common ANN
correctly responds to the contingency and gives an output of 1. This output shows that
the respective contingency is harmful for the secure operation for the whole power-
network (Sulaiman, 2016; Yalcin, 2016). The response of the common ANN monitoring
the voltage angles is also similar, hence is represented by the same diagram shown in
Figure 14.3.
The contingency of line 10, the least severe contingency with respect to MW security as
well as voltage security using the NR method, was induced at 18 minutes or 1,080 seconds,
and the response of the common ANN is shown in Figure 14.4.
FIGURE 14.4 Response of common ANN monitoring the MW-power flow on all lines during
line 10 outage.
It can be seen from the above response that the common ANN monitor shows no
significant effect of line 10 outage. The power flow on the remaining lines remains within
their specified limits, and there is no threat to the security of the power system either with
respect to MW security or voltage security (Voumvoulakis, 2009; Boyen, 1999). All the
crucial parameters of the power system remain within their specified limits because line
10 is isolated and does not form any critical flowgate. Even the reactive power flows and
voltages angles stay within limits.
Next, to test the voltage security and reactive power security, the outage of line 3 was
performed at 1,500 seconds, i.e., 25 minutes, and the response of the common ANN is as
shown in Figure 14.5.
FIGURE 14.5 Response of common ANN monitoring the voltage magnitude on all buses during
line 3 outage.
Figure 14.5 shows the response of the common ANN monitoring the voltage magnitudes
on all the buses of the network (Sun, 2007; Xu, 2012). At exactly 1,500 seconds the
response of the common ANN spikes to 1, signifying that a contingency affecting the
voltage security of the whole power-network has occurred (Konstantelos, 2016). The
response of the common ANN monitoring the reactive power on the whole system is also
more or less similar, as shown in Figure 14.6.
FIGURE 14.6 Response of common ANN monitoring the reactive power on all lines during line
3 outage.
As explained previously, the outage of line 10 does not have any severe repercussions with
respect to voltage security; hence, the response of the common monitoring is same as
14.4 CONCLUSION
From the work elaborated in the preceding section, it has been shown that a common ANN
can serve as an ancillary service for online monitoring and security assessment. It has also
been shown that the network of common ANN can provide the changes in the dynamics
of the power system parameters instantly, which may forewarn the system operator in any
eventuality to take necessary remedial actions. The common ANN monitoring system can
also provide results in the event of a contingency. This system of common ANNs may also
be used to trigger protective devices to take precautionary measures when needed. For
future work, deep learning methods may be employed in conjunction with faster learning
algorithms and training error compensation.
ACKNOWLEDGEMENT
The authors would like to thank the Technical Education Quality Improvement
Programme Phase II for their financial support. The authors would also like to thank
their colleagues in the Electrical Engineering Department at Indian Institute of
Engineering Science and Technology, Shibpur.
REFERENCES
1. Tiwary, S. K., and Pal, J. (September 15, 2017). ANN application for MW security assessment
of a large test bus system. In 2017 3rd International Conference on Advances in Computing,
Communication & Automation (ICACCA) (Fall) (pp. 1–4). IEEE.
2. Tiwary, S. K., and Pal, J., October 2017. ANN application for voltage security assessment
of a large test bus system: A case study on IEEE 57 bus system. In 2017 6th International
Conference on Computer Applications InElectrical Engineering-Recent Advances (CERA)
(pp. 332–334). IEEE.
3. Tiwary, S. K., Pal, J., and Chanda, C. K., 2017, December. Mimicking on-line monitoring
and security estimation of power system using ANN on RT lab. In 2017 IEEE Calcutta
Conference (CALCON) (pp. 100–104). IEEE.
4. Tiwary, S. K., Pal, J., and Chanda, C. K., 2019. Application of Common ANN for Similar
Datatypes in On-line Monitoring and Security Estimation of Power System. In Emerging
Technologies in Data Mining and Information Security (pp. 3–11). Springer, Singapore.
5. Panciatici, P., Bareux, G., and Wehenkel, L., 2012. Operating in the fog: Security management
under uncertainty. IEEE Power and Energy Magazine, 10(5), 40–49.
6. Jain, T., Srivastava, L., Singh, S. N., and Erlich, I., 2005, November. New parallel radial basis
function neural network for voltage security analysis. In Proceedings of the 13th International
Conference on, Intelligent Systems Application to Power Systems (7pp.). IEEE.
7. Nakawiro, W., and Erlich, I., 2008, April. Online voltage stability monitoring using artificial
neural network. In 2008 Third International Conference on Electric Utility Deregulation
and Restructuring and Power Technologies (pp. 941–947). IEEE.
8. Saeh, I. S., and Khairuddin, A., 2008, December. Static security assessment using
artificial neural network. In 2008 IEEE 2nd International Power and Energy Conference
(pp. 1172–1178). IEEE.
9. Khazaei, J., Piyasinghe, L., Miao, Z., and Fan, L., 2014, July. Real-time digital simulation
modelling of single-phase PV in RT-LAB. In 2014 IEEE PES General Meeting| Conference
& Exposition (pp. 1–5). IEEE.
10. Wang, J., Tang, X., and Yin, Z., 2014, September. Research on medium voltage battery energy
storage system based on RT-LAB. In 2014 IEEE Conference and Expo Transportation
Electrification Asia-Pacific (ITEC Asia-Pacific) (pp. 1–5). IEEE.
11. Ding, N., Benoit, C., Foggia, G., Bésanger, Y., and Wurtz, F., 2015. Neural network-based
model design for short-term load forecast in distribution systems. IEEE transactions on
power systems, 31(1), 72–81.
12. Bahmanyar, A. R., and Karami, A., 2014. Power system voltage stability monitoring using
artificial neural networks with a reduced set of inputs. International Journal of Electrical
Power & Energy Systems, 58, 246–256.
13. Sulaiman, S. M., Jeyanthy, P. A., and Devaraj, D., 2016, January. Artificial neural network
based day ahead load forecasting using smart meter data. In 2016 Biennial International
Conference on Power and Energy Systems: Towards Sustainable Energy (PESTSE)
(pp. 1–6). IEEE.
14. Yalcin, T., and Ozdemir, M., 2016, October. Pattern recognition method for identifying
smart grid power quality disturbance. In 2016 17th International Conference on Harmonics
and Quality of Power (ICHQP) (pp. 903–907). IEEE.
15. Voumvoulakis, E. M., and Hatziargyriou, N. D., 2009. A particle swarm optimization
method for power system dynamic security control. IEEE Transactions on Power Systems,
25(2), 1032–1041.
16. Boyen, X., and Wehenkel, L., 1999. Automatic induction of fuzzy decision trees and its
application to power system security assessment. Fuzzy sets and Systems, 102(1), 3–19.
17. Sun, K., Likhate, S., Vittal, V., Kolluri, V. S., and Mandal, S., 2007. An online dynamic security
assessment scheme using phasor measurements and decision trees. IEEE transactions on
power systems, 22(4), 1935–1943.
18. Xu, Y., Dong, Z. Y., Zhao, J. H., Zhang, P., and Wong, K. P., 2012. A reliable intelligent system
for real-time dynamic security assessment of power systems. IEEE Transactions on Power
Systems, 27(3), 1253–1263.
19. Konstantelos, I., Jamgotchian, G., Tindemans, S. H., Duchesne, P., Cole, S., Merckx, C.,
Strbac, G., and Panciatici, P., 2016. Implementation of a massively parallel dynamic
security assessment platform for large-scale grids. IEEE Transactions on Smart Grid, 8(3),
1417–1426.
20. Kumar, N., and Mahajan, V., 2018, December. Reconfiguration of Distribution Network
For Power Loss Minimization & Reliability Improvement using Binary Particle Swarm
Optimization. In 2018 IEEE 8th Power India International Conference (PIICON) (pp. 1–6).
IEEE.
CHAPTER 15
Modelling of Solar Cell

Considering One Diode Model in
MATLAB/Simulink Environment
Snehashis Ghoshal,1 Sumit Banerjee,2 and Chandan Kumar Chanda3
1,2Department of Electrical Engineering, Dr. B.C. Roy Engineering College, Durgapur, India.
3Department of Electrical Engineering, IIEST Shibpur, India.
ABSTRACT: This study investigates simulation of a photovoltaic cell in MATLAB/

Simulink environment keeping in view one diode model of the same. Performance analysis
of a photovoltaic (PV) module/array is carried with the help of simulation results. Based
on an equivalent circuit model and existing mathematical model validation is carried out
in this study. Output characteristics of solar cell represents maximum power point under
different environmental conditions.
Keywords: Equivalent circuit, One diode model, Output characteristic, PV cell.
15.1 INTRODUCTION
Application of renewable energy such as solar (photovoltaic (PV) as well as solar thermal)
and wind is gaining importance day by day. They are more significant among other
renewable resources from electricity generation point of view. Generation of electricity
through PV cell is becoming popular in both small-scale and large-scale basis. PV plant
possesses lower operating cost as well as maintenance cost. Moreover, a PV system can
act as in grid connected or in an islanded mode (Bellia, Youcef, and Fatima, 2014; Iov
et al., 2007). The use of renewable energy resources like solar energy and wind energy
is increasing rapidly for electricity generation purposes. Solar energy can be converted
in to electricity by using photovoltaic cell. A PV cell can convert photon energy into
the form of electrical signals. This method of power generation is highly eco-friendly
and, hence, PV power generation systems have gained popularity for generation in
small-scale as well as in large-scale production. It is observed that a building-integrated
123
photovoltaic (BIPV) involves advanced local energy management rather than PV power
permanent grid injection. Their study was focused on building integrated micro-grid
(BIMG) design and implementation applied to BIPV system with energy storage and
smart grid communication. The goal was to formulate a power balancing strategy
with smart grid interaction, aiming at reducing grid peak consumption, avoiding
undesirable power injection at grid, and making full utilization of local PV production.
Energy management was carried out considering grid time-of-use tariffs, grid access
limits storage capacity, load, and PV power shedding was necessary. Based on the
experimental platform, the results showed that the system maintained stability and
confirmed the relevance of the proposed local energy management giving perspectives
on better integration of small PV plant in power grid.
Facing high-penetration level of future BIPV, grid issues can be improved by a proper
local energy management strategy instead of permanent grid injection. Experimental
results showed that proposed strategy was simple but effective in power balancing, reducing
peak consumption and avoided undesired grid injection. The energy management was easy
to implement giving better integration of small urban PV plant in grid.
In this study a mathematical model of solar cell is developed and simulated in
MATLAB/Simulink environment based on existing one diode model and its mathematical
equations.
A solar cell is a photodiode which emits electrons when subjected to radiation of a
certain wavelength. Flow of this electron causes current flow when attached with an external
closed circuit.
This study validates the one diode model (Krishan, Sood, and Kumar, 2013) of solar
cell in MATLAB/Simulink environment. This study helps in understanding the behaviour
of solar cell under different environmental conditions.
15.2 MODELLING OF SOLAR CELL

There are several software packages which are utilized to analyse solar cells such as
MATLAB, PSPICE. In this study, MATLAB/Simulink environment was considered
for the analysis regarding solar cell. Even in MATLAB, solar cells can be modelled
using several methods such as mathematical block modelling, embedded MATLAB
programming, and physical block modelling. In present study, mathematical block
modelling of solar cells is used for the modelling purpose. As it is already mentioned,
this study focuses on one diode model of solar cell for modelling of PV module. The
behaviour and characteristic of solar cell can be well explained by existing established
expressions (Krishan, Sood, and Kumar, 2013; Mohamed, 2013; Garg, Singh, and
Gupta, 2014). The practical model of one diode based solar cell is shown in Figure 15.1
(Paul, 2012).
Modelling of Solar Cell Considering One Diode Model in MATLAB/Simulink  125
FIGURE 15.1 One diode model of solar cell.
In the given one diode model, Rs represents series resistance of PN junction cell and Rsh
represents shunt resistance which is inversely proportional to the leakage current to ground.
Id and Ish are diode current and shunt leakage current.
Abbreviations used in solar PV Modelling
Iph = photo current
I0 = saturation current
Irs = reverse saturation current
Isc = short-circuit current = 7.34 A
ki = diode factor considering short-circuit situation = 0.0032
T = atmospheric temperature
Tn = nominal value of temperature which is 298 in the present case
G = solar irradiance in W/m2
q = charge of an electron = 1.6 × 10–19 C
Voc = open-circuit voltage from the module = 0.6 V
n = diode ideality factor = 1.3 in present case
k = Boltzman’s constant = 1.38 × 10–23
E0 = band gap energy = 1.1 eV
Ns = number of solar cells connected in series = 1

Rs = series resistance of the solar cell = 0.2 Ω
Rsh = shunt resistance of the solar cell = 420 Ω
It is evident that the solar cell is prone to different atmospheric conditions such as
different operating temperature. To be specific, solar radiation can be classified as
direct, indirect, and global radiation. Performance of solar modules depends mostly on
direct solar radiation. Solar radiation does not remain constant throughout the day. It
varies unpredictably. Cloud and other climatic disturbances affect the amount of solar
radiation (mainly direct solar radiation) which in turn affects the performance of solar
cell.
Similarly, temperature also does not remain constant throughout the day. Change in
temperature affects the output of a solar cell significantly.
So if the performance or specific output of a solar cell is known at any particular
temperature and at particular radiation, then its performance can be predicted at another
temperature with suitable accuracy.
Keeping the above-mentioned requirement regarding the standardization of solar
cell, its performance specifications are mentioned at 25°C temperature and 1000 W/m2
insolation level.
15.3 SIMULATION MODEL OF A SOLAR CELL

One diode model of solar cell was developed in MATLAB/Simulink environment.
Table 15.1 (Erdem and Erdem, 2013) shows the parameter specification of solar cell for the
said mathematical block modelling.
TABLE 15.1 Parameter Specification of Solar Cell
Parameters Value
Short-circuit current (Isc) 7.34 A

Open-circuit voltage (Voc) 0.6 V
Irradiance (G) 1000 W/m2
P-V and V-I characteristics of a solar cell at nominal temperature and insolation were
simulated using the Simulink model shown in Figure 15.2. However, output for different
environmental conditions such as different insolation level and temperature can be
simulated by changing the concerned input to the particular model is in Figure 15.2. Output
waveforms indicate the non-linear nature of the solar cell.
FIGURE 15.2 Simulink model of a solar cell.

The behaviour of the Simulink model was validated and implemented in MATLAB
environment.
Basically, PV and V-I characteristics are key features to describe the performance of
a solar cell. Open-circuit voltage, short-circuit current, and maximum power point are
significant (Krismadinata, 2013; Bouraiou, 2014) to yield the maximum power point of a
solar cell at any insolation for a certain operating temperature.
FIGURE 15.3 IV characteristics of solar cell.

FIGURE 15.4 Output (PV) characteristics of solar cell.
Output power from a solar cell depends on variation in different weather conditions such
as temperature and shading. If the temperature increases above the nominal value, then a
voltage loss in output is observed (Khanna, 2013).
15.5 CONCLUSION
During this study, a model of a solar cell based on one diode model has been simulated
and validated in MATLAB/Simulink environment. Effect of different temperatures and
insolation levels are also analysed. Since a typical PV cell produces less than 3.5 W at
0.6V approximately, then high power can achieve by connecting many solar cells in series-
parallel configurations. So the performance of a solar array can be analysed using several
series-parallel combination of solar cells.
REFERENCES
1. Bellia, H., Youcef, Fatima, M., 2014. A detailed modelling of photovoltaic module using
MATLAB, NRIAG Journal of Astronomy and Geophysics, April 2014, 1–9.
2. Iov, F., Ciobotaru, M., Sera, D., Teodorescu, R., Blaabjerg, F., 2007. Power electronics and
control of renewable energy systems, In Proceedings PEDS 2007, 7–28.
3. Krishan, R., Sood, Y. R., and Kumar, B. U., 2013. The Simulation and design for analysis
of photovoltaic system based on MATLAB, In Proceedings ICEETS, April 2013, 647–671
4. Mohamed, N., Shannan, A. A., Yahaya, N. Z., and Singh, Balbir. Single-Diode Model and
TwoDiode Model of PV Modules: A Comparison, In Proceedings ICCSCE, Nov., 2013,
210–214.
5. Garg, R., Singh, A., and Gupta, S., PV cell models and dynamic simulation of MPPT trackers
in MATLAB, In Proceedings INDIACOM, March 2014, 6–12.
6. Paul, P. S., Mondal, S., Akter, N., and Mominuzzaman, S. M., 2012. Modelling combined
effect of temperature and irradiance on solar cell parameters by MATLAB/Simulink, In
Proceedings ICECE, Dec., 2012, 1–6.
7. Erdem, Z., Erdem, M. B., A proposed model of photovoltaic module in matalab/Simulink
for distance education, 13th International Education Technology Conference, Elsevier, 103,
2013, 55–62.
8. Krismadinata, S., Rahim, N. A., Ping, H. W., and Selvaraj, J.. Photovoltaic module modelling
using Simulink/matlab, The 3rd International conference on sustainable future for human
security, Elsevier, 17, 2013, 537–546.
9. Bouraiou, A., Hamouda, M., Chaker, A., Sadok, M., Mostefaoui, M., and Lachtar, S.,
Modelling and simulation of photovoltaic module and array based on one and two diode
model using Matlab/Simulink, International conference on Technology and materials for
renewable energy, environmental and sustainability, Elsevier, 74, 2014, 864–877.
10. Khanna, A., Mueller, T., Stangl, R. A., Hoex, B., Basu, P. K., and Aberele, A. G., 2013. A fill
factor loss analysis method for silicon wafer solar cells, IEEE Journals of Photovoltaics, 3(4),
1170–1177.
CHAPTER 16
Modelling of Intelligent Cooling

of a Building in MATLAB/
Simulink Environment
Snehashis Ghoshal,1 Sumit Banerjee,2 and Chandan Kumar Chanda3
1,2
Department of Electrical Engineering, Dr. B. C. Roy Engineering College, Durgapur, India.
3
Department of Electrical Engineering, IIEST Shibpur, India.
ABSTRACT: This study analyses thermal behaviour of a house through simulation in

MATLAB/Simulink environment considering the effect of outdoor environment over the
same. For a given house geometry (including dimension, number of windows, thermal
properties of materials used, etc.), cooler characteristics (temperature, flow rate etc.),
and Setpoint, this model predicts the cost-effectiveness in view of energy efficiency. The
model is developed for a room located in a hot climate zone, where in general the outside
temperature is high and the room temperature needed to be set at a lower value for occupant
comfort.
Keywords: Building energy simulation, Demand side management, Energy efficiency,

Sustainable development.
16.1 INTRODUCTION
Currently, the contribution of energy towards development is indefeasible. The rate
of economical growth and utilization of energy are very closely related to each other.
Unfortunately, the major resources of energy extraction are finite by nature. Because most
of these are fossil fuel-based such as coal, oil, and natural gas, concern has been developing
for utilization of renewable resource extraction, such as solar (PV and thermal), wind
biomass, etc. Besides, harnessing of energy has a direct impact on the environment. Fossil
fuel-based resources emit detrimental greenhouse gases during harnessing. Hence, these
resources are responsible for global warming. On the other hand, renewable resources are
eco-friendly.
131
It is evident that fossil fuel will not sustain in distant future, and renewable resources
alone cannot meet the increased demand of energy. But per capita, energy consumption
has to be maintained at a reasonable level to enhance the economical growth (Per capita
energy consumption is a strong indicator of the standard of living of the people of a country.
High consumption of energy has been associated with better quality of life, which in turn
is related to Gross National Product). Hence, an optimized mix of renewable and non-
renewable energy is necessary which is achieved by co-generation, energy conservation,
energy efficiency, and energy management.
Advance in economy and social development causes more and more investment in
the real estate sector. Hence, gradually, it has become more energy-intensive. Building
consumes about 40% of overall energy consumption worldwide. Hence, this has become
a challenge for the researchers to implement energy management in buildings and achieve
energy efficiency.
In a study, US department of energy has reported that buildings were responsible for
energy consumption of about 39% in USA (2009) (Agrawal and Tiwari, 2010). As energy
consumption is closely related to occupant behaviour, a thorough study has been carried
out regarding occupant energy conservation (Chow, Ji, and He, 2005; Dubey, Solanki,
Tiwari, 2009). Automated building systems can be designed to save about 5–10% energy
in residential buildings (Huang et al., 2001; Nayak and Tiwari, 2009; Abdalla and Wilson,
2001).
Earlier, most of the studies were carried out in cold climatic zones. But recently,
researchers paid attention to develop passive buildings in hot climatic zones (Nayak and
Tiwari, 2009). Efficient shading is necessary to reduce energy consumption (Sarhaddi
et al., 2010). Proper simulation (Solar Energy Laboratory, 2011; Sharan and Kandpal,
1992) for assessing becomes very helpful in this regard. Sohel et al. (2014) illustrated
the difficulties encountered in predicting thermal behaviour and energy consumption in
buildings.
Besides the development of well-insulated thermal envelopes, availing highly efficient
cooling technologies coupled with renewable will be the primary focus to achieve
sustainable development. Primary interest in designing energy efficient buildings is to
consider quantifying and qualifying factors to maintain occupant comfort (Sopian et al.,
1996; Tonui and Tripanagnostopoulos, 2007).
Moreover, it is well understood that a significant amount of energy can be saved if a
certain portion of energy can be conserved within the building thermal mass itself. In such
a case, the building acts as energy storage. Generally, thermal energy can be preserved in
this case.
In a country like India, cooling of houses to achieve occupant comfort consumes huge
energy. Hence, thermodynamic modelling of a house becomes necessary to carry out heat
balance equations. In this study, a Simulink model was developed in MATLAB/Simulink
environment. This modelling is based on certain specifications such as thermal properties
Modelling of Intelligent Cooling of a Building in MATLAB/Simulink Environment  133
of house material, thermal resistance, mass airflow rate, chiller temperature, and initial set
temperature of the room.
16.2 THERMAL MODELLING OF A HOUSE REQUIRING COOLING LOAD

As stated earlier, the proposed model is based on hot climatic zones where cooling of room
is required for occupant control. Here, the desired temperature of the room is assumed at
20°C and average outside temperature is referenced as 35°C. However, keeping the practical
point of view, the atmospheric temperature is assumed to vary sinusoidally.
Thermodynamically, a model for a house can be divided into four sub-components.
• Thermostat
• Chiller
• House
• Environment
Every component mentioned here is represented by a sub-system while modelling and are
discussed in detail in the following section.
16.2.1 Thermostat Subsystem

The function of thermostat is to maintain the temperature of a certain place within a
given limit. In the present case, the room temperature is set at 20°C, and the thermostat
is required to maintain this temperature with a fluctuation of 5°C throughout the day.
Thermostat is activated by the error signal obtained from the difference between the
set-point temperature and instantaneous temperature. When temperature rose above the
upper limiting value, the thermostat activates the flow of coolant from chiller, and when
it reaches below the lower limiting value, it cuts off the same. The thermostat sub-system
is shown in Figure 16.1.
Basically, it consists of a relay block within itself which does the monitoring of
temperature variation limit within 5⁰C as mentioned earlier.
FIGURE 16.1 Simulink model for thermostat sub-system.
16.2.2 Chiller Sub-system

Chiller is assumed to yield a constant mass flow rate (3600 kg/hr) with the help of a coolant
material which is maintained at a fixed temperature say at –20⁰C as in the present case.
Thermostat signal turns the flow rate from chiller on or off depending on the indoor and
outdoor temperature difference.
Heat extraction rate from the chiller is expressed by Eq. 16.1.
dQ
= (Troom - Tchiller ) ´ M dot ´ c (16.1)
dt
• dQ/dt = heat extraction rate from the room
• c = heat capacity of air at constant pressure
• Mdot = air mass flow rate through heater (kg/hr)
• Tchiller = temperature of the chiller
• Troom = instantaneous room temperature
Sub-system model for chiller is given in Figure 16.2.
FIGURE 16.2 Simulink model for chiller sub-system.
16.2.3 House Sub-system

Basically, house sub-system should calculate the room temperature variation at every instant
and considers the heat extraction from the house and heat losses to the environment.
Figure 16.3 represents the Simulink model of house sub-system.
Equations (16.2) and (16.3) are utilized while modelling house sub-system
dQ T - Troom (16.2)
lloss = out
dt Req
dTroom 1 æ dQchiller dQlosses ö (16.3)

= ç - ÷
dt M .c è dt dt ø
• M is mass flow rate inside the house
• Req thermal resistance of the house
FIGURE 16.3 Simulink model of house sub-system.
16.2.4 Environment sub-system

In this modelling, outside environment is treated as a heat source with infinite capacity and
a function of temperature Tout. The average temperature is treated to vary sinusoidally over
a day.
16.3 SIMULATION MODEL OF A HOUSE REQUIRING COOLING LOAD

Figure 16.4 represents the schematic representation of thermal model of a house having
cooling load demand. This model was simulated considering 35°C ambient temperature
(which varies in sinusoidal manner over the day). For occupant comfort, the temperature
was assumed as 20°C
FIGURE 16.4 Simulink implementation of a house having cooling load demand.


The behaviour of the Simulink model was validated and implemented in MATLAB
environment. A plot is obtained against indoor temperature to outdoor temperature
in Figure 16.15. From the figure, it is observed that the outdoor temperature is varying
sinusoidally throughout the day, whereas indoor temperature is maintained within
5 degrees of the Setpoint.
FIGURE 16.5 Simulation result of building envelope.

16.5 CONCLUSION
During this study, thermal model of a house requiring cooling load has been simulated
and validated in MATLAB/Simulink environment. The ambient average temperature was
assumed to be 35°C, whereas the surrounding temperature was assumed to be varied by
10°C. For optimal occupant control, the room temperature was taken as 20°C. Future work
may include modelling of a building with adequate building thermal insulation including
the effect of windows so that the thermal mass of the building can be utilised as a storage
medium for energy
REFERENCES
1. Agrawal, B., Tiwari, G. N., 2010. Optimizing the energy and exergy of building integrated
photovoltaic thermal (BIPV/T) systems under cold climatic conditions. Appl. Energy 87,
417–426.
2. Chow, T. T., Ji, J., He, W., 2005. Photovoltaic-thermal collector system for domestic
application. In Proceedings of the ISEC 2005, the 2005 Solar World Congress, Orlando,
USA.
3. Dubey, S., Solanki, S. C., Tiwari, A., 2009. Energy and exergy analysis of PV/T air collectors
connected in series. Energy Build. 41 (8), 863–870.
4. Dubey, S., Solanki, S. C., Tiwari, A., 2009. Energy and exergy analysis of PV/T air collectors
connected in series. Energy Build. 41 (8), 863–870.
5. Huang, B. J., Lin, T. H., Hung, W. C., Sun, F. S., 2001. Performance evaluation of solar
photovoltaic/thermal systems. Sol. Energy 70 (5), 443–448.
6. Nayak, S., Tiwari, G. N., 2009. Theoretical performance assessment of an integrated
photovoltaic and earth air heat exchanger greenhouse using energy and exergy analysis
methods. Energy Build. 41 (8), 888–896.
7. Abdalla, F. K., Wilson, P., 2001. Analysis of a roof-top combined photovoltaic/solar thermal
plant at Christchurch. In ISES Solar World Congress, Adelaide.
8. Nayak, S., Tiwari, G. N., 2009. Theoretical performance assessment of an integrated
photovoltaic and earth air heat exchanger greenhouse using energy and exergy analysis
methods. Energy Build. 41 (8), 888–896.
9. Sarhaddi, F., Farahat, S., Ajam, H., Behzadmehr, A., 2010. Exergetic performance assessment
of a solar photovoltaic thermal (PV/T) air collector. Energy Build. 42, 2184–2199.
10. Solar Energy Laboratory, 2011. TRNSYS 17, A transient systems simulation program.
11. Sharan, S. N., Kandpal, T. C., 1992. Testing of a prototype combined photovoltaicthermal
linear Fresnel concentrator-receiver system. Energy Convers. Manage. 33, 37–39.
12. Sohel, M. I., Maa, Z., Cooper, P., Adams, J., Scott, R., 2014. A dynamic model for airbased
photovoltaic thermal systems working under real operating conditions. Appl. Energy 132,
216–225.
13. Sopian, K., Yigit, K. S., Liu, H. T., Kakac, S., Veziroglu, T. N., 1996. Performance analysis of
photovoltaic thermal air heaters. Energy Convers. Manage. 37 (11), 1657– 1670.
14. Tonui, J. K., Tripanagnostopoulos, Y., 2007. Air-cooled PV/T solar collectors with low cost
performance improvements. Sol. Energy 81 (4), 498–511.
CHAPTER 17
Control and Monitor of Non-

linear Gas Flow by Developing
a Self-Adaptive Fuzzy Controller
A. K. Pal, Sampa Paul, Indrajıt Naskar, and A. Mondal
Department of AEIE, Heritage Institute of Technology, Kolkata, India.
ABSTRACT: This study gives a detailed account about the adaptation of an intelligent
gas flow controller for non-linear application. Before elaborating the design of the
process-specific adaptive scheme for fuzzy gas flow controller, it describes how to
develop the mathematical model of gaseous system and study of non-linearity. In the
latter part, the proposed controller is developed and its performance is compared to PID
and conventional fuzzy controllers. In the present work, the performance of the proposed
controller is tested with the help of MATLAB simulations. Instead of using a fixed gain
multiplier, finding an appropriate multiplying factor for the fine tuning of Fuzzy Logic
Controller becomes a great challenge, which is directly related to the process dynamics.
In view of this, a non-fuzzy adaptive scheme is implemented for the proposed fuzzy gas
flow controller. Effectiveness of the proposed self-adaptive fuzzy controller is reviewed in
non-linear gas flow control.
Keywords: Mathematical modelling, Gas flow, Fuzzy controller, Tuning, Adaptive control.
17.1 INTRODUCTION
The main objective of this study is to design a suitable controller to control and monitor the
gas flow. The fluid flow equations are derived from the physical principles of conservation of
mass, momentum, and energy (Feistauer, 1993). To design a perfect controller, knowledge
of mathematical model of the system is very important. Natural gas is an inviscid fluid as
the viscosity of the gas is equal to zero, and it is compressible also (Chorin and Marsden,
1993). If the gas flows in along a horizontal pipe and then it can be considered as one-
dimensional flow. Also in gas flow modelling, it is assumed no heat source occurs inside
139
the pipe and transfer of heat due to the heat conduction is much less than the heat exchange
with the surrounding (Herrań-Gonza´lez et al., 2008).
Soft computing is an innovative approach of any intelligent system according to the
human requirement, which can make decisions and take action keeping in mind the
importance of requirement (Tong, 1980, Sugeno, 1985). Unlike Boolean or crisp logic,
fuzzy logic deals with uncertainty in the field of engineering along with degrees of
uncertainty to the answer to a logical question. Its main advantage is to reduce the design
complexity as-well-as hardware cost and improves control performances (Pedrycz, 1984,
Takagi, and Sugeno, 1985). It had been investigated that while considering any non-linear
system, fuzzy logic control (FLC) system is more convenient than PID controllers.
17.2 METHODOLOGY
17.2.1 Mathematical Modelling of the System
FIGURE 17.1 Block diagram of proposed control scheme for gas flow.
The proposed control scheme for gas flow control by a fuzzy gas flow controller is depicted
in Figure 17.1. The whole system is consisting of gas cylinder, final control element, process,
controller, and gas sensor. Here, the measured value of gas concentration by a gas sensor
is compared with a reference value, to generate an error signal and according to that error
signal, an intelligent controller provides a control action signal to the final control element
for proper opening of the valve, connected with the source (gas cylinder). The system can
be utilized for oxygen supply to the patient during respiratory problems.
In this section, the mathematical modelling is carried out for each part of the closed-
loop flow diagram shown in Figure 17.1.
17.2.1.1 Modelling of Control Valve

æ Kv ö
Control valve is a first-order system and its transfer function ç GV = depends on
è t s + 1 ÷ø
valve capacity, time constant (τ), and valve input pressure range (X).
Control and Monitor of Non-linear Gas Flow by Developing a Self-Adaptive  141
FIGURE 17.2 Block diagram of control valve.
In case of pneumatic control valve,

valve capacity
gain of control valve ( Kv ) = (17.1)
input span of pressure (3 - 15 psi)
17.2.1.2 Modelling of Gas Sensor

In gas detection, response time is generally defined as the time taken to achieve 90% of
the ultimate change in conductance subsequent to a stepchange in gas concentration at
the sensor. In gas response determination, the sensor output initially changes very rapidly
and may observed overshoot in some cases; however, it reaches steady state after that. The
shape of curves and response time of the sensors are very important parameters as they
are directly related with the applicability of the sensor. The other important parameters are
the concentration of the gas sample and the operating temperature of the testing sample
(Moshayedi, Ata Jahangir et al., 2013).
The characteristic response of a typical gas sensor can be represented by a second-order
model:
kwn2
H (s) = (17.2)
s 2 + 2twn s + wn2
where, k is the gain, t and represents the time constant and natural frequency respectively.
These parameters can be determined from the step response of sensor by giving a step
excitation.
kwn2
H (s) = (17.3)
s 2 + 2twn s + wn2
17.2.1.3 Modelling of Process

The state of the gas volume is represented by differential variables at the internal node of
the block. As differential variables, they require initial conditions to be specified prior to
the start of simulation. The dialogue box of each block modelled with finite gas volume has
a Variables tab, which lists three variables:
• Pressure of gas volume
• Temperature of gas volume
• Density of gas volume
By default, Pressure of gas volume and Temperature of gas volume have high priority,
with target values equal to the standard condition (0.101325 MPa and 293.15 K). One
can adjust the target values to represent the appropriate initial state of the gas volume
for the block.
Density of gas volume has the default priority because only the initial conditions of two
of the three variables are needed to completely determine the initial state of the gas volume.
If desire is the proportionality constant]
(17.4)
dP
Thus, Eq (17.3) can be modelled as A + a P = Qi (17.5)
dt
No approximation is needed for this linear dynamic system.
Non-linear Mathematical Model.
In case, the output flow is represented by the square root of pressure differential, then
Q0 = b P (17.6)
Q0 = a P. The resulting equation yields a non-linear dynamic model Q0 = a P as presented
in Eq. (17.7)
dP
A + b P = Qi (17.7)
dt
17.2.1.4 Overall Representation
FIGURE 17.3 Overall block diagram of the system with individual mathematical model.
The gas flow diagram for overall control and sensing mechanism is presented in Figure 17.3
with respective mathematical models. The output gas flow rate of the process is measured
by a gas sensor and the measured value is compared with a reference flow rate as shown in
Figure 17.3. The generated error signal is then fed to the proposed intelligent controller that
in turn gives a control action signal to the final control element for the required opening of
the valve, connected with the gas cylinder source.
17.2.2 Controller Design

PID controller fails to deliver accurate control in non-linear systems but fuzzy logic controller
(FLC) ensures accurate control (Sugeno and Yasukawa, 1993, Sugeno and Tanaka, 1991).
Fuzzy logic is a powerful way to put engineering expertise. Due to their simple formulas
and computational efficiency, triangular MFs have been extensively used in FLCs (Palm,
1992). Seven triangular MFs with equal base-width and 49 if-then rules are used in design
of FLC as shown in Figure 17.4 and Table 17.1.
FIGURE 17.4 Membership functions of e, ∆e, u for FLC.
TABLE 17.1 Fuzzy Rules for Computation of U

∆e/e NB NM NS ZE PS PM PB
NB NB NB NB NM NS NS ZE
NM NB NM NM NM NS ZE PS
NS NB NM NS NS ZE PS PM
ZE NB NM NS ZE PS PM PB
PS NM NS ZE PS PS PM PB
PM NS ZE PS PM PM PM PB
PB ZE PS PS PM PB PB PB
Tuning of FLC is a critical issue as it involves too many parameters (Pal, Mudi, and Maity,
2013). Among the different tuning strategy, rule-based tuning offers good results. But
the limitation of rule-based tuning scheme is that it increases the design complexity
all the more by augmenting the number of if-then rules required for the particular
application (Pal, and Mudi, 2013). Apart from the need of determining the input/output
SFs for the conventional fuzzy controller; there is also necessary to determine the design
parameters for the tuning of controller (Pal, and Chakraborty, 2015). This increases the
computational time and thereby leading to delay in system operation. Thus, the idea here
is to implement a certain non-fuzzy adaptive module to alter the output scaling factor
on-line, to achieve better performance. Such a scheme will reduce the design complexity
by eliminating the need of multiple rules and the number of MFs while being able to
exercise a similar control on the system (Pal et al., 2018). Therefore, to eliminate the
design complexity, output SF of the fuzzy controller is updated according to the process
trend by again modification factor ε, which is determined by the normalized error of the
system. Figure 17.5 shows the feedback closed-loop control system with proposed self-
adaptive fuzzy PI controller (SAFPIC).
FIGURE 17.5 Feedback closed-loop control system with SAFPIC in MATLAB Simulink (V15b).
In the developed model of SAFPIC, constant scaling factors 0.9 and 11 are used for the
inputs e and ∆e, respectively.
eN = 0.9 *e (17.8)
∆eN = 11 *∆e (17.9)
The output scaling factor (Gu = 0.02) is constant for a particular application of FLC, but it
does not remain constant for the proposed adaptive fuzzy controller while in operation.
The output SF of the adaptive fuzzy controller is modified in each sampling instant, which
depends on the instantaneous process condition. The gain-updating factor as shown in
Figure 17.6 is calculated using the relation:
γ = K(1 + eN) (17.10)
In the above relation, ‘K’ is a positive constant, used to provide the appropriate range of
variation of and eN is the normalized value of input error.
From Figure 17.6, the following equations can be established:
u ' = g .u (17.11)
& u N = Gu .u '
= Gu .uk (1+ eN ) (17.12)

The adaptive controller output derived by Eq. 17.12, provides the following important
information in a typical second-order under-damped system.
• If input error (e) is +ve, then the system output resides below the Setpoint, thus the
system needs an extra control action to match the Setpoint which will be provided
by the proposed adaptive action g = K (1 + e ).
• Similarly, when e is negative, that indicates the system is moving away from the
Setpoint then to counteract, it is required to reduce the control action by applying
the equation g = K (1 + (-e) ). (17.13)
17.3 RESULTS AND COMPARATIVE STUDY
dP
1. Non-linear Model: A + b P = Qi
dt
Where, A = 0.7 and = 0.5
FIGURE 17.6 Response of non-linear system with SAFPIC, FLC, and PID for delay = 0.2 and
load = 0.5.
load = 0.5.
TABLE 17.2 Performance of Non-linear System with SAFPIC, FLC, and PID for Delay = 0.2 and Load = 0.5
Types of Controller Rise Time (Sec) Settling Time(Sec) Peak Overshoot (%)
SAFPIC 2.5 6.0 15
FLC 1.9 7.7 37
PID 1.2 - -
SAFPIC 2.8 4.7 4
FLC 1.8 5.5 19
PID 1.0 - 38
load = 0.8.
load = 0.8.
SAFPIC 2.8 4.8 4
FLC 1.8 5.5 19
PID 1.0 - 38
SAFPIC 2.5 6.0 15
FLC 1.9 17.5 37
PID 1.2 - -
Study of Figures 17.6–17.9 and analysis of Tables 17.2–17.5 reveal that proposed self-
adaptive fuzzy PI controller can improve the system performance by incorporating the
proposed adaptive scheme that varies with system parametere.
dP
Linear Model: A + a P = Qi
dt
where A= 0.7 and a = 0.5.
FIGURE 17.10 Response of linear system with SAFPIC, FLC, and PID (P = 0.5, I = 3, D = 0.02).
From Figure 17.10, it is observed that PID controller is working satisfactorily for linear
system and therefore, it is desirable to select PID controller for such system.
Study found that the proposed SAFPI controller out-performs PID as well as
conventional fuzzy controller. In case of SAFPIC, a significant improvement of settling time
is recorded and only a small overshoot is observed even with delay and load variations. It is
sure that the recorded values of settling time and overshoot also take an account to improve
the long-term process performance indices such as ISE, ITAE and IAE.
17.4 CONCLUSION
The study suggested a new self-adaptive tuning technique for fuzzy PI controller for
non-linear gas flow control. This non-fuzzy tuning scheme tuned the FLCs by an online
gain updating parameter. By implementing the proposed adaptive algorithm, system
non-linearity along with the impact of variation of delay and load are studied. It is
observed that the slightest variation of both the parameters (delay and load) affects the
controller performance in large in case of PID and FLC, which are not so prominent in
case of SAFPIC. The proposed controller provides a useful alternative for controlling
the non-linear gas flow and hope it can be implemented to any other complex processes
and also in real-time hardware processes, which are difficult to control using FLC and
conventional PID controller.
REFERENCES
1. Feistauer, M., 1993. Mathematical Methods in Fluid Dynamics, Longman Scientific &
Technical, New York.
2. Chorin, A. J., and Marsden, J. E., 1993. A Mathematical Introduction to Fluid Mechanics.
Springer, New York, http://dx.doi.org/10.1007/978-1-4612-0883-9
3. Herrań-Gonza´lez, Cruz A. J. M. De La, Andre´s-Toro, Risco-MartıńB. De J. L.,
2008. Modelling and simulation of a gas distribution pipeline network doi: 10.1016/j.
apm.2008.02.012, 1584–1600.
4. Moshayedi, Ata Jahangir, Toudeshki, Arash, Gharpure, Dayamanti C., 2013. Mathematical
Modelling for SnO2 Gas Sensor, IEEE Symposium on Industrial Electronics & Applications
(ISIEA 2013), pp 33–38.
5. Tong, R. M., 1980. The evaluation of fuzzy models derived from experimental data, Fuzzy
Sets and Systems, 4(1), 1–12.
6. Sugeno, M., 1985. Industrial applications of fuzzy control, Elsevier Science Inc.
7. Pedrycz, W., 1984. An identification algorithm in fuzzy relational systems, Fuzzy Sets and
Systems, 13, 153–167.
8. Takagi, T., and Sugeno, M., 1985. Fuzzy identification of systems and its applications to
modelling and control, IEEE Transactions on Systems, Man, Cybernetics, 15(1), 116–132.
9. Sugeno, M., and Yasukawa, T., 1993. A fuzzy logic based approach to qualitative modelling,
IEEE Transactions on Fuzzy System, 1(1), 7–31.
10. Sugeno, M., and Tanaka, K., 1991. Successive identification of a fuzzy model and its
application to prediction of a complex system, Fuzzy Sets and Systems, 42, 315–334.
11. Palm, R., 1992. Sliding mode fuzzy control, Proceedings of Fuzz IEEE, 519–526.
12. Pal, A. K., Mudi, R. K., and Maity, R. R. De 2013. A Non-fuzzy self-tuning scheme of PD-
type FLC for overhead crane control. Advances in Intelligent Systems and Computing,
Publisher: Springer-Verlag Berlin Heidelberg, 199, 35–42.
13. Pal A. K., and Mudi, R. K., 2013. Anadaptive PD-type FLCanditsrealtimeimplementationto
overhead crane control, International Journal of Emerging Technologies in Computational
and Applied Sciences, 6(2), 178–183.
14. Pal A. K., and Chakraborty, J., 2015. Design a fuzzy logic controller with a non-fuzzy
tuning scheme for swing up and stabilization of inverted pendulum. Advances in Intelligent
Systems and Computing, Publisher: Springer, 308, 221–230.
15. Pal, A. K., NaskarIndrajıt and Paul Sampa 2018. Fuzzy-based Gain Adaptive Scheme for
Setpoint Modulated Model Reference Adaptive Controller. International Journal of Natural
Computing Research (IJNCR), 7 (4), 1–19.
CHAPTER 18
Transient Stability Controlling

and Assessment of a Congested
Power System in a Deregulated
Environment
D. Sarkar,1 R. Brahmachary,1 and S. D. Barma2
1
Electrical and Electronics Engineering, National Institute of Technology Nagaland, Chumukedima,
Dimapur-797103, India.
2
Electrical Engineering, Tripura Institute of Technology, Tripura, India.
dipusarkar5@rediffmail.com
ABSTRACT: Congestion controlling is a fundamental errand in a de-regulated power

framework condition. The congestion might be calmed by methods for FACTS devices.
The other motivation behind FACTS devices is to improve transient stability. In the present
study, the presentation and examination of FACT devices, for example, STATCOM, SVC
has appeared for the mentioned purposes and results are analysed in the IEEE 14 Bus test
system in PSAT MATLAB environment. The transient stability analysis has been performed
utilising affectability based eigenvalue investigation. The results of the reproduction
demonstrate that the proposed philosophy has the competency to improve transient
stability in the de-regulated condition control environment.
Keywords: Congestion management, FACTS devices, Eigenvalue analysis, De-regulation,

Transient stability.
18.1 INTRODUCTION
The procedure wherein guidelines and guidelines which control the electric business are
changed to permit rivalry so the clients are given decision concerning power providers is
known as de-regulation. The monetary effectiveness of creation and utilization of power
can be improved by de-regulation. If challenge exists in the power business, the cost of
intensity is probably going to descend and along these lines, the clients can be profited.
151
De-regulated power market, as suggested in by Sambasivarao (2013), aims to provide

power at a reasonable demand, it encourages competition, continuity, and the quality of
supply is maintained and also promotes efficiency and economy of power systems. Since
one of the important issues in the de-regulated electricity market for reliable and secure
system operation is congestion management (Karthikeyan, 2013), to fix this issue flexible
alternating current transmission system (FACTS) devices are used for many years, as these
devices can be easily adopted in controlling power and increasing the usable capacity of
the present system. Congestion usually occurs where there is a violation of the constraints
of the grid from its normal state. For optimizing the location of FACTS devices to control
congestion management in de-regulated power system, Singh and David have suggested a
simple and efficient model, by controlling parameters, as in ref. (Singh, 2001). The most
efficient way to find the optimal location of FACTS devices in the case of a de-regulated
power system is to determine the performance index sensitivity factor. In case there is no
congestion, then from a static viewpoint, the placement of FACTS devices are done. Ejebe
(1979) considers the security and stability of the de-regulated power system, thus uses a
method based on the real power flow performance index (PI), for suitable placement of
FACTS devices in the power system, the choice is based on how much congestion cost is
reduced as explained by Singh (2001). Since more benefit is offered to customers by a de-
regulated power system, so it is quite popular in the modern world. The ever-increasing
power demand has resulted in operating the power system close to its stability limits
for which problems such as voltage instability, power loss, and transmission congestion
arises. By congestion management, we aim that the limits of the transmission system are
not violated thus concluding that the system security and reliability are in the acceptable
range. If there is any violation of transmission constraints, then that gets reflected by the
variation of the cost of electrical energy but in the de-regulated environment, the main aim
is to ensure effective control over both producers and consumers to maintain the power
in an acceptable limit in which there are generation and transfer in real-time keeping
in mind the market efficiency. We can treat congestion management by ways like load
shifting, generation re-dispatch, and contracts curtailment. For controlling congestion in a
transmission line of the power system, two methods are discussed by Sharma (2019). These
are (1) OPF-based method (2). ATC-based method. Yousefi (2012) suggests congestion
management for transmission lines in a restructured market surrounding using FACTS
devices and demand response. Congestion can also be relieved using cost free-methods
as discussed in paper (Reddy, 2006). FACTS can be used for reducing congestion without
disturbing the monetary matter. Gawande (2018) proposes to relieve congestion using
FACTS device and location is found by using the reduction of total system reactive power
loss on electrical IEEE-14 bus and MATLAB software is used here for study purposes. The
study of power flow, power system stability, and power system security using sensitivity
indices are done here. In this chapter, considering the line loading limits of each line,
means in which line power flow is exceeding than the prescribed limit. The analysis is
done on IEEE-14 bus system. Somasundaram (2018) proposes that the cost-free method
using SVC gives the remedy for congestion by enhancing the voltage profile. The WSCC-9
bus system was taken as test system and it was built using MATLAB-PSAT software. The
Transient Stability Controlling and Assessment of a Congested Power System  153
power flow has run for base case without SVC. The SVC was connected to each load bus
in the WSCC-9 bus system, and the load flow has run for each case to find the optimal
location of SVC. Purnapatra (2016) gives the system voltage deviation reduction that has
been performed by allocating the STATCOM device.
In this chapter, the IEEE 14 bus network is considered on which congestion management
is performed utilizing FACTS devices like STATCOM, SVC. The position of eigenvalue
gives the proper placement for the arrangement of the FACTS devices. The damping of the
system has been increased so that the system reaches stability within a short period with the
help of time domain simulation in PSAT application software.
18.2 EIGEN VALUE ANALYSIS

In a power system, each node or bus is associated with the following parameters or variables:
The magnitude of the voltage (|V|).
The phase angle of the voltage (d).
Active power (P).
Reactive power (Q).
The power flow problem solves the complex matrix equation as follows:
I = YV
I = S * /V * (18.1)
Where-
I = Nodal current injection matrix.
Y = System nodal admittance matrix.
V = Unknown complex node voltage vector.
S = Apparent power nodal injection vector representing the specified load and
generation of nodes.
Also-
S = P + JQ (18.2)
The Newton–Raphson method solves the portioned matrix equation:
é DP ù é Dq ù
ê DQ ú = J ê DV ú (18.3)
ë û ë û
Here-
ΔP and ΔQ = Mismatch active and reactive power vectors.
ΔV and Δθ = Unknown voltage value and angle rectification vectors.

J = Jacobian matrix of partial derivative terms.
The analysis can be expressed as follows:
Equation (18.3) can be written in more matrix form as
é DP ù é J11 J12 ù é Dq ù
ê DQ ú = ê J J ú ê DV ú (18.4)
ë û ë 21 22 û ë û
Now,
DQ = J R DV (18.5)
where-
J R = éë J 22 - J 21 J11-1 J12 ùû
JR is the reduced Jacobian matrix of the system.

Equation (18.5) can be written as-
DV = J R-1DQ (18.6)
The matrix JR represents the linearized relationship between the incremental changes in bus
voltage (ΔV) and bus power injection (ΔQ).
The real power (ΔP = 0) and angle part in Eq. (18.4) are reduced so that computational
efforts can be minimized by reducing dimensions of the Jacobian matrix JR and for focusing
on the study of the supply problem and reactive demand of the system.
Voltage instability can be spotted by recognizing modes of the eigenvalues matrix
ΦJR. Its values give a relative measure of proximity to instability. The eigenvectors give
information related to the loss of voltage instability.
Modal analysis of JR results in the following:
J R = lF x (18.7)
where
Φ = Right Eigenvector matrix of JR.
ξ = Left Eigenvector matrix of JR.
ʎ = diagonal Eigenvalue matrix of JR.
Equation (18.18) can be written as:
J R-1 = Fl -1x (18.8)
Eigenvalue investigation depicts the small-signal behaviour of the system, for example,
the behaviour around one point and doesn’t consider the account of the behaviour of the
entire non-linear system. In this way, we can say the eigenvalue examination is utilized to
investigate the dynamic behaviour of the operating system under different characteristic
frequencies. Since in power system it is necessitated that all modes should be steady, hence,
it is necessitated that every single electromagnetic oscillation damps out as quickly as
possible. In fact, modal analysis not only helps us in determining eigenvalue but also the
eigenvectors of the system. The right eigenvector gives us information about the observability
of oscillation. The left eigenvector gives us information about the controllability. The
combination of the left and right eigenvector indicates the setting of controllers.
FIGURE 18.1 IEEE 14 bus test model.
18.3 STUDY OF THE IEEE 14 BUS TEST MODEL

The modelling of the IEEE 14 bus test system using PSAT is shown in Figure 18.1. The
system consists of 16 transmission lines, 4 transformers, 5 generators, 1 P-V generator bus
and 11 static loads; the base MVA is taken as 100 MVA and the system is having 69 KV as
the base voltage.
This IEEE 14 bus test system is congestion-free, and to get the required result, we need
to create a three-phase to ground fault in it. We are considering a three-phase to a ground
fault occurring on bus 5. This is done so that abnormal condition occurs in the system and
we can place the different FACTS devices like STATCOM, SVC and check their performance
and efficiency in the faulted condition. The proper position of placing these FACTS devices
are found by using eigenvalue analysis.
18.4 CONGESTION STUDY IN THE IEEE14 BUS TEST SYSTEM

Improving the transient stability in the IEEE 14 bus test system is the main aim of this
chapter. Analysis of eigenvalue is done using PSAT to assess the stability of this system
and also to find out the best placement of STATCOM and SVC. The fault has occurred
at bus 5 and the foretold FACTS devices are placed at different locations and eigenvalue
analysis is performed to check the best location for placing then. The location where we
got the maximum stability and most efficient performance of the system that location is
considered optimum for placing FACTS devices. The entire analysis is done using PSAT
application software. The result of eigenvalue analysis with and without using FACTS
devices is shown in Table 18.1. The analysis clearly shows that dynamic order and negative
eigenvalue increases after the use of FACTS devices. Thus, dynamic stability of the system
also increases. Since placing the FACTS device at bus 5 increases the damping of the system
more than placing at other bus locations, so bus 5 is taken as the optimal location for placing
the FACTS devices. Along with these FACTS devices, a PV bus is added to bus 5 to supply
the required reactive power deficiency in the system. Thus, the Overall Transient Stability
of the system has increased.
TABLE 18.1 Eigenvalue Analysis in an IEEE 14 Bus System
When FACTS Devices are
With STATCOM With SVC model
not used
Dynamic order 58 59 59
Buses 14 14 14
Positive eigens 0 0 0
Negative eigens 57 58 58
Complex pairs 13 12 11
Zero eigens 1 1 1

After doing the eigenvalue analysis, time domain simulation is done by creating a fault in
the IEEE 14 bus system, and the result is analysed by using FACTS devices and without
using FACTS devices. The results thus obtained are discussed herein.
18.5.1 Analysis without using Facts Devices

Figures 18.2–18.4 show the time domain analysis done at the IEEE 14 bus system having a
three-phase to ground fault at bus 5, without using any FACTS devices. Figure 18.2 shows
the plot of relative rotor angles with respect to time. Figure 18.3 shows the plot of angular
speed of generator with respect to time. Figure 18.4 shows the plot of highest three voltages
with respect to time. Figure 18.5 points out the position of eigenvalue when no FACTS
devices are used.
FIGURE 18.2 Relative rotor angle plot with respect to time (when no FACTS device is used).
FIGURE 18.3 Angular speed of generator with respect to time (when no FACTS device is used).
FIGURE 18.4 Highest three voltages with respect to time (when no FACTS device is used).
FIGURE 18.5 Eigenvalue representation (when no FACTS devices are used).
18.5.2 Analysis with using Statcom Model

By the eigenvalue examination performed utilizing PSAT, the ideal placement of STATCOM
along with the PV generator system is discovered to be Bus-5. The time domain analysis
demonstrates the plot of relative rotor angle, angular speed of generator, and highest three
voltage, as appeared in Figures 18.6–18.8 individually. From the figures, after using the
STATCOM device, the transient peaks died out much before 5 seconds, and the system
reaches stability within the minimum time possible. Figure 18.9 demonstrates the position
of eigenvalue when STATCOM is utilised.
FIGURE 18.6 Relative rotor angle plot with respect to time (when STATCOM is used).
FIGURE 18.7 Angular speed of generator with respect to time (when STATCOM is used).
FIGURE 18.8 Highest three voltages with respect to time (when STATCOM is used).
FIGURE 18.9 Eigenvalue representation (when STATCOM is used).

18.5.3 Analysis Using SVC Model

With the assistance of eigenvalue investigation in PSAT, bus 5 is viewed as the best location
for putting the PV generator bus and SVC devices. Figures 18.10–18.12 demonstrates the
plot of relative rotor angle, angular speed of generator, and most elevated three voltage,
separately done through time domain analysis, utilising PSAT. As can be assessed from the
figures that after using the SVC, the transient peaks died out much before 5 seconds, and
the system reaches stability within the minimum time possible. Figure 18.13 demonstrates
the situation of eigenvalue when the SVC model is utilised.
FIGURE 18.10 Relative rotor angle plot with respect to time (when SVC MODEL is used).
FIGURE 18.11 Angular speed of generator with respect to time (when SVC MODEL is used).
FIGURE 18.12 Highest three voltages with respect to time (when SVC MODEL is used).
FIGURE 18.13 Eigenvalue representation (when SVC MODEL is used).
18.6 CONCLUSION
From this study, it very well may be expressed that the SVC model gives the best execution
with respect to improving transient dependability, as after using it, the system response
takes minimum time to give the result and furthermore the eigenvalue examination
demonstrates an impressively better outcome.
REFERENCES
1. Ejebe, G. C., and Wollenberg, B. F., 1979. Automatic contingency selection. IEEE
Transactions on Power Apparatus and Systems, 1, 97–109.
2. Gawande, A. V., and Jadhao, C. W., 2018. Placement of FACTS Device using Reduction
of Total System Reactive Power Loss Sensitivity Indices Analysis Method. International
Journal of Engineering Technology Science and Research, 925–931
3. Karthikeyan, S. P., Raglend, I. J., and Kothari, D. P., 2013. Impact of FACTS devices on
exercising market power in deregulated electricity market. Frontiers in Energy, 7(4),
448–455.
4. Purnapatra, S., Kuanr, B. R., Haldar, V., Ghosh, A., and Chakraborty, N., 2016, December.
Voltage profile improvement and congestion management using STATCOM and UPFC
device. In 2016 IEEE Uttar Pradesh Section International Conference on Electrical,
Computer and Electronics Engineering (UPCON), 146–150.
5. Reddy, K. R. S., Padhy, N. P., and Patel, R. N., 2006, April. Congestion management in
deregulated power system using FACTS devices. In 2006 IEEE Power India Conference.
6. Sambasivarao, N., Amarnath, J., and Purnachandrarao, V., 2013. Congestion management
using FACTS devices in deregulated power system. IJRET: International Journal of Research
in Engineering and Technology, 2, 90–95.
7. Sharma, A., and Jain, S. K., 2019. Gravitational search assisted algorithm for TCSC
placement for congestion control in deregulated power system. Electric Power Systems
Research, 174, 105874.
8. Singh, S. N., and David, A. K., 2001. Optimal location of FACTS devices for congestion
management. Electric Power Systems Research, 58(2), 71–79.
9. Somasundaram, P. L., Jayakumar, V., and Sundararaju, K., 2018. Congestion management
using SVC under deregulated powersystem. International Journal of Pure and Applied
Mathematics, 118(20), 2307–2317.
10. Yousefi, A., Nguyen, T. T., Zareipour, H., and Malik, O. P., 2012. Congestion management
using demand response and FACTS devices. International Journal of Electrical Power &
Energy Systems, 37(1), 78–85.
CHAPTER 19
Bagasse Drying by Waste Heat

Recovery from Surrounding
Walls of the Combustion
Chamber of a Jaggery Plant
A. B. Shinde, S. N. Sapali, and Y. V. Sonone
ABSTRCT: In conventional jaggery plants, the extracted sugarcane juice is used for
preparing jaggery, while the residue bagasse is used as fuel. Normally bagasse contains 48
to 52% moisture which affects the calorific value of bagasse, hence drying of bagasse is
necessary. In conventional jaggery plants, wet bagasse is spread over the ground and dried
under sunlight, which is time-consuming, space-consuming, and laborious work. In the
rainy season, conventional bagasse drying is impracticable due to unavailability of intense
sunlight. In conventional jaggery plant, a large amount of heat is wasted inside the wall of
the combustion chamber. This chapter deals with the experimental performance analysis of
newly designed jaggery plant with bagasse drying mechanism. The walls of the combustion
chamber are designed and manufactured to recover waste heat to dry the wet bagasse. The
experimental outcome reveals that around 5–6% of heat is recovered from walls of the
combustion chamber to dry the wet bagasse. It is also observed that around 23–25% of heat
is utilised for jaggery preparation and around 69–71% of heat is lost. The moisture removed
during the experimentation is ranged between 68% and 70%. Thus, a quick and non-
laborious bagasse drying process alternative to conventional sundrying process is developed.
Keywords: Jaggery, Bagasse drying, Waste heat recovery.
19.1 INTRODUCTION
At present, 115 countries of the world cultivate sugarcane for sugar production and produce
about 133 million tonnes of sugar and 10 million tonnes of jaggery. The conventional
163
jaggery plants consist of a combustion chamber, which is underground and the pan
containing sugar juice is placed over the combustion chamber on the earth surface. During
the combustion process of dry bagasse, large amount of heat is lost to the surrounding
area. Dry bagasse is used as a fuel during jaggery production. For drying the wet bagasse, it
requires large space and time and it also depends on the availability of solar radiation. The
reduction in moisture content increases the calorific value of bagasse. Many researchers
have attempted various methods and techniques to remove moisture content from wet
bagasse to improve the gross calorific value. Sharon, Abirami, and Alagusundaram (2013)
have done the research work on energy losses in traditional jaggery plant. The efficiency of
the crushing and concentration process is 60% and 14.75%, respectively. He concluded that
by using steam-jacketed vessels for concentration, circulation of cooled water for cooling
the hot jaggery and recirculating it for steam production and during crushing will make the
jaggery-processing energy efficient. Sudhakar and Vijay (2006) have developed the bagasse
dryer to reduce the moisture content from wet bagasse. The average moisture content in
the bagasse is 51.5%. They observed that after the installation of bagasse dryer, about 45%
of wet bagasse was routed through the dryer. Panchal and Shinde (2006) have developed
the charts for calorific values and bagasse composition. Several researchers worked on
the jaggery-processing industry to improve the performance and to recover the heat from
exhaust gases. The heat content in the exhaust gas can be utilised for drying of bagasse
and pre-heating of sugarcane juice in the pre-heater. Manjare and Hole (2016) have shown
that the thermal efficiency can be improved from 16% to 24%, and bagasse consumption
is reduced by 1.2 kg per kg jaggery production by utilising heat from exhaust gas for pre-
heating the sugarcane juice in the pre-heater. Jakkamputi and Jagadeesh Kumar Mandapathi
(2016) worked on the performance improvement of jaggery plant using solar energy. The
sensible heat required for jaggery preparation is utilised from solar panels. Anwar (2010)
has worked on the determination of moisture content from bagasse using a microwave
oven. It took about 20 to 25 minutes for the determination as compared to 8–10 hours in a
conventional hot air-drying method.
TABLE 19.1 The Test Data for Case 5 using Steel Plate Walls with Cage Rotation
S. No. Contents
1 Total sugarcane used 109 kg
2 Amount of sugarcane juice 60 kg
3 Amount of wet bagasse 49 kg
4 Time required for jaggery preparation 2hr 15min
5 The weight of bagasse after the experiment 41 kg
6 Amount of moisture removed 8 kg
7 Average moisture removed. 33.33 %
19.2 METHODOLOGY
The experimentation is performed within walls of combustion chambers made up of
steel plates and also brickwalls. Dry bagasse is supplied to combustion chamber from
Bagasse Drying by Waste Heat Recovery from Surrounding Walls of the Combustion  165
the feeder located at the front face. Wet bagasse is packed in the iron cage, surrounding
the walls of the combustion chamber. There is a gap of around 50 mm between steel
plate and wet bagasse to avoid self-ignition of bagasse. During the combustion, the
heat reaches sugarcane juice kept in juice pan above the combustion chamber. Hot flue
gasses are flowing through the exhaust pipe. Ash is collected in a tray at the bottom
of the combustion chamber. The steel plate and the brick wall gets heated during
the combustion. The air trapped between the hot steel plate and wet bagasse passes
through the wet bagasse, which helps to remove the moisture from the wet bagasse. The
experimental setup is shown in Figure 19.1.
FIGURE 19.1 Experimental set up for developed jaggery plant.
19.2.1 Experimentation
Case 1: The experimentation is carried out within the walls of the combustion chamber
made of porous steel plate as shown in Figure 19.2, which results in burning of wet bagasse
due to direct contact of burning flame with the wet bagasse. Case 2 is carried out with a
porous steel plate with a flame arrester. Even after providing the flame arrester, the flame
travels to the wet bagasse through the porous part of the plate. Case 3 relates to plain steel
plate without porosity as walls of combustion chamber. The direct contact of burning
flame and the wet bagasse is avoided, but the temperature of steel plate is more than the
self-ignition temperature of bagasse, hence, the pyrolysis of bagasse takes place. Case 4
is to avoid self-ignition of bagasse, and the gap of 50 mm is maintained between steel
plate and wet bagasse. It is observed that only the bagasse facing the combustion wall gets
dried quickly. Hence, in the next trail, as case 5, the cage rotation by 180 degrees is done
to bring the bagasse facing the atmosphere that it now gets to face walls of combustion
chamber. Now the bagasse gets dried from both sides of the cage in order to achieve
quick drying of bagasse without pyrolysis. Trails of cases 6, 7, and 8 are carried out by
using a one-inch thick plain brick, two-inch thick plain brick, and porous brick as walls
of combustion chamber, respectively. Out of all these tested cases, experimental results
of case 5 is derived using plain steel plate as walls of combustion chamber, wherein a gap
of 50 mm is maintained inbetween the walls of combustion chamber and cage containing

wet bagasse.
The experimental data for case 5 using steel plate walls with cage rotation is given below
in Table 19.1.
Figure 19.2 shows the experimentation that is carried out for case 1 using porous brick
wall for combustion chamber. It shows that the burning flame comes out from porous wall
and gets in direct contact with wet bagasse resulting in the burning of wet bagasse. Figure
19.3 shows the experimentation carried out for case 5 using steel plate walls with cage
rotation.
19.3 DISCUSSION
19.3.1 Numerical Analysis
The temperature of the steel plate is an essential variable of the project as the quality of
bagasse drying depends on the plate temperature. Bagasse near the plate surface may
adhere to it and may reach the self-ignition temperature of bagasse as in case 3. To avoid
such a condition, there is gap of 50 mm between the steel plate and bagasse. The partial
arrangement of the combustion chamber and the bagasse packed surrounding it is shown
in Figure 19.4, which is applicable to cases 4 and 5. There is a gap of 50 mm between steel
plate and wet bagasse so that the dry bagasse is burnt in the combustion chamber to get the
heat flux at a rate of 4800 W/m2.
FIGURE 19.2 Experimental trail on porous steel plate (case 1).

FIGURE 19.3 Experimental trail on plain steel plate wall with cage rotation (case 5).
FIGURE 19.4 Combustion chamber and cage for wet bagasse, case 4 and case 5.
The geometry of the plate is having the dimensions of 0.917 m × 0.917 m× 0.003 m. The
number of nodes and elements used for meshing are 4,802 and 2,304, respectively. The
boundary conditions for the simulations are heat flux 4,800 W/m2 on the combustion
side. This is required to remove the moisture from bagasse as a part of the cycle time.
The convective heat transfer coefficient is assumed to be 2 W/m2K on the ambient side of
the bagasse packed based on environmental conditions. Energy scheme of second-order
upwind is used for the simulation.
FIGURE 19.5 Simulation results of steel plates of 3 mm thickness.
The temperature distributions obtained through simulation in ANSYS Fluent are presented
in Figure 19.5. The values of the temperatures recorded are in the range of 976 K to 300 K.
The temperature of steel plate is recorded as 976 K, while the temperature of air between
the gap of the plate and wet bagasse is varied from 800 K to 300 K. The ambient air comes
in contact with the bottom part of the steel plate and gets heated. As the temperature of air
increases, it moves in an upward direction and passes through the wet bagasse depending
upon the porosity. The similar simulation is done for brick wall of one and two inch
thickness. The temperature distributions obtained through simulation in ANSYS Fluent
are presented in Figure 19.6, as in case 6 and Figure 19.7, as in case 7. The values of the
temperatures recorded are in the range of 600–300 K. The temperature of brick wall is
recorded as 600 K, while the temperature of air between the gap of the wall and wet bagasse
is varied from 550 K to 300 K.
FIGURE 19.6 One-inch brick wall simulation.

FIGURE 19.7 Two inch brick wall simulation.
A plain steel plate of 3 mm thickness attains the temperature about 850 K to 950 K which is
much more than the pyrolysis temperature of bagasse. In case 5, to avoid pyrolysis, one has
to maintain the gap between bagasse and plate. The air gap could provide the conductive
resistance to heat flow, but at the same time, it enhances the convective heat transfer to
the bagasse. The air in the gap between bagasse and plate get heated and pass through the
bagasse that helps to remove the moisture from wet bagasse. The moisture removed from
the combustion side is much more than the ambient side, hence, provision is required to be
made to change the direction of cage in order to dry the bagasse from both the sides. The
brick wall of one inch thickness reaches the temperature above 500 K which is more than
the pyrolysis temperature of bagasse, as in case 6 and brick wall of two inch thickness could
attain the temperature below 500 K which is below the pyrolysis temperature of bagasse, as
in case 7. Increase in brick wall thickness up to two inches, reduces the wall temperature, but
sufficient amount of heat is not transferred to bagasse because of higher thermal resistance
of brick wall.
19.4 EXPERIMENTAL RESULTS

It is observed in the simulation results that the steel plate temperature is 976 K which is
nearly equal to the temperature record by thermocouple during experiments that are from
800 to 900 K. Similarly, the brick wall temperature obtained through simulation is 600 K,
and during the experiment, it is measured around 510 to 580 K. The heat utilisation for
different cases, observed during experimentation of case 1 to case 6 is presented in Figures
19.8 to 19.11.
FIGURE 19.8 Avg. moisture removed during experiment.
FIGURE 19.9 Heat utilised for jaggery preparation.
Figure 19.8 shows average moisture removed during the experiments. Bagasse contains 50%
moisture in it before drying. During the experiment, heat from the combustion chamber
is transferred to wet bagasse and moisture is removed. It is observed that case 5 gives the
maximum moisture removal during the experiment that is 69.23%.
Figure 19.9 shows per cent heat utilised for jaggery preparation. During the experiment,
heat is transferred to the juice pan placed at top of the combustion chamber. The heat from
combustion chamber is first utilised for sensible heating of juice then for latent heat of
juice. Figure 19.9 shows how much heat out of total heat supplied for jaggery preparation is
utilised that is indirectly the efficiency of the jaggery plant.
FIGURE 19.10 Total heat loss during experiment.
FIGURE 19.11 Heat utilised for moisture removal.
Figure 19.10 shows total heat loss during the experiment. Mass of bagasse burn multiplied
by calorific value of bagasse gives the total amount of heat supplied in the combustion
chamber. Out of the total heat that is distributed, heat is required for jaggery and for
moisture removal, and the rest of the heat is lost into the environment. Maximum heat
loss observed in case 1 is 74.36%, whereas the minimum heat loss observed in case 5 is
68.53%.
Figure 19.11 shows heat utilised for moisture removal during the experiments.
Heat utilised for moisture removal is based on the amount of water removed during the
experiment multiplied by latent heat of water. It is observed that the utilization of heat is
increasing from case 1 to case 5 and suddenly dropped down in case 6, that is because of
brick wall. Brick wall added extra resistance to heat transfer to wet bagasse.
19.5 CONCLUSION
In this research work, the analysis of heat recovery from the walls of a newly design
combustion chamber of a jaggery plant is presented. The following conclusions are drawn
from the research work:
1. Heat is recovered from the walls of the combustion chamber and utilised to remove
moisture content of wet bagasse.
2. The system for quick drying of wet bagasse is developed and implemented
successfully with compare to the conventional sun drying process.
3. The thermal analysis of the designed jaggery plant indicates 69–71% total
losses accompanied by the process. The minimum energy required for jaggery
processing is estimated to be only 23–25% of the total energy supplied.
4. The amount of heat recovered by the wet bagasse sample is observed to be
around 6%.
5. Percentage moisture removal from the wet bagasse sample is observed to be
around 69%.
REFERENCES
1. Esther Magdalene Sharon, M., Kavitha Abirami, C. V., and Algusundaram, K., 2013. Energy
Losses in Traditional Jaggery Processing. Indian Food Industry Mag. 32, 134–138.
2. Manjare, Appasaheb., and Hole, Jitendra, 2016. Exhaust Heat Recovery and Performance
Improvement of Jaggery Making Furnace, International Journal of Current Engineering
and Technology.
3. Jakkamputi, Lakshmi Pathi, Jagadeesh Kumar Mandapathi, Mohan, 2016. Improving the
Performance of Jaggery Making Unit Using Solar Energy. Perspectives in Science, 23–24.
4. Anwar, S. I., 2010. Determination of Moisture Content of Bagasse of Jaggery Unit Using
Microwave Oven. Journal of Engineering Science and Technology, 5(4), 472–478.
5. Shiralkar, Kiran Y., Kancharla, Sravan K., Shah, Narendra G., Mahajani, Sanjay, M., 2014.
Energy Improvements in Jaggery Making Process, Energy for Sustainable Development,
18, 36–48.
6. Sardeshpande, Vishal R., Shendage, D. J., Pillai, Indu R., 2010. Thermal Performance
Evaluation of a Four Pan Jaggery Processing Furnace for Improvement in Energy Utilization,
Energy, 35, 4740–4747.
7. Pattnayak, P. K., and Misra, M. K., 2004. Energetic and economics of traditional gur
preparation: a case study in Ganjam district of Orissa, India, Biomass and Bioenergy,
26, 79–88.
8. Techno-Economic Analysis of Jaggery Production in Maharashtra, Chapter 1, 2012. Indian
Institute of Technology, Bombay.
9. Sudhakar, J., and Vijay, P., 2006. Control of Moisture Content in Bagasse by Using Bagasse
Dryer.
10. Panchal, R. J., and Shinde, S. M., 2006. Effect of Bagasse Moisture on Boiler Performance.
2, 3–4.
CHAPTER 20
Study of Score Coefficient

Functions for DBSCAN-based
Photometric Invariant Approach
in Vision Sensor-Based Shoe
Detection for Human Tracking
P. Paral, A. Chatterjee, and Anjan Rakshit
Department of Electrical Engineering, Jadavpur University, Kolkata, India.
ABSTRACT: The problems of accurately tracking people in human-robot co-existing

environments are considered as one of the significant problems in present times. A
state-of-the-art randomized template matching algorithm, called photometric- invariant
CFAsT-Match (PICFAsT-Match), can be implemented in real robot to solve an active
problem within this genre, where vision sensing is used for detecting human shoes in
challenging real-life environments, during pursuit. The performance of the PICFAsT-
match depends on suitable choices of a parameter, named score coefficient. In this
chapter, a detailed study on possible choices of the variations of the score-coefficient
parameter is performed, and the overall impacts of the chosen variations on the shoe
detection outcomes are evaluated.
Keywords: Mobile robotics, Shoe detection, Vision sensing, DBSCAN algorithm,

Photometric-invariant CFAsT-Match, Score coefficient.
20.1 INTRODUCTION
In recent years, research in the domain of human-centred robotics technology is
becoming more and more prevalent. An active research topic in the relevant domain is
the accurate tracking of people under motion (Bellotto and Hu, 2009; Yuan et al., 2015)
with mobile robots, in which cases the robots are scheduled to follow and perform tasks
173
in a specific manner, as supervised by humans. Several contemporary, state-of-the-art

approaches (Paral, Chatterjee, and Rakshit, 2019a, 2019b, in press-b; Yuan et al., 2015)
addresses a very crucial sub-problem within this genre, where a monocular camera-
based vision sensing system is employed for shoe detection purposes, by estimating the
target person’s positions and orientations in consecutive camera frames, during people
following.
Our earlier works in Paral et al. (2019a, 2019b, in press-b) have considered that during
the tracking process, the shoe poses undergo various general transformations, such as 2D
translations, scales, rotations, or 2D affine transformations, and aptly demonstrated how
the problem of shoe detection can be solved as an image template matching problem under
general conditions, where the conventional 2D translation-based template matching is
extended to take care of more general transformations, for example, scales, rotations, or 2D
affine transformations.
In Paral et al. (2019a), we have first presented the effective implementations of a
general-purpose template matching algorithm, called FAsT-Match (Korman et al., 2013),
and its variant in colour space, called CFAsT-Match (Jia et al., (2016), for detecting shoes
in subsequent frames. But as we have discussed in Paral, Chatterjee, and Rakshit, (2019a),
CFAsT-Match does not produce reliable results in real-life shoe detection problem where
the images undergo photometric changes, such as shadows, illumination, specularities, and
suffer from inferior brightnesses and contrasts. To overcome the shortcomings of CFAsT-
Match, we have proposed a novel variant of CFAsT-Match, named photometric-invariant
CFAsT-Match (PICFAsT-Match) in Paral, Chatterjee, and Rakshit (2019a). The research
work in Paral, Chatterjee, and Rakshit (2019a) has implemented both CFAsT-match and
PICFAsT-match for the real-life shoe detection purposes, using DBSCAN (Ester et al.,
1996), a density-based clustering approach. Also, we have proposed another novel visual
object detection algorithm in Paral, Chatterjee, and Rakshit (2019b, in press-b), called the
OPTICS-based visual target detection (OBVTD) algorithm which is implemented utilizing
OPTICS (Ankerst, 1999), another density-based clustering algorithm. As reported in
Paral, Chatterjee, and Rakshit (2019a, 2019b, in press-b), the implementations based on
PICFAsT-Match and OBVTD, clearly outperform the CFAsT-match based implementation
in challenging real-life scenarios of shoe detection.
In case of PICFAsT-Match (Paral, Chatterjee, and Rakshit, 2019a), DBSCAN is
employed to group the pixels of the template image into various arbitrarily shaped clusters.
With this cluster information at hand, we compute the parameters, named the score-
coefficient and the RGB vector centroid (Jia et al., 2016; Paral, Chatterjee, and Rakshit,
2019a). It has to be mentioned that the performance of the PICFAsT-Match algorithm for
shoe detection purpose depends on suitable choices of the parameter score coefficient.
In Paral, Chatterjee, and Rakshit (2019a), it was proposed that each score coefficient be
computed as a direct reciprocal of the size of the corresponding cluster. In this work, an
intensive study on possible choices of the variations of the score-coefficient parameter
with variation in the corresponding cluster size is analysed, and the overall effects of
Study of Score Coefficient Functions for DBSCAN-based Photometric  175
these chosen variations of score coefficients on the final shoe detection performances are
evaluated.
The rest of this study is organized as follows: Section 2 presents our proposed research
framework. Section 3 describes the experimental results and real-life performance
evaluations of our proposed research work in detail. Then, Section 4 concludes the chapter.
20.2 PROPOSED RESEARCH FRAMEWORK

Our proposed research framework primarily consists of four major phases. The schematic
diagram in Figure 20.1 depicts those major phases. It should be noted that, during pursuit,
we consider the initial frame captured as the reference frame in our earlier proposed
PICFAsT-Match approach, and the corresponding frame is also taken into account as the
matching image in the same algorithm (Paral, Chatterjee, and Rakshit, 2019a).
Now, in the first phase, the target shoes are detected in the matching image with
an intelligent algorithm and a shoe template which is a rectangular image block mainly
accommodating the shoes, is extracted from that matching image (Paral, Chatterjee, and
Rakshit, 2019a). The shoe template, denoted by, is then pre-processed to fit it to the needs
of the proposed scheme (Paral, Chatterjee, and Rakshit, 2019a).
In the second phase, the DBSCAN algorithm is utilized to form the arbitrarily shaped
clusters of the pixels of the pre-processed shoe template. Once this cluster information is
obtained, we calculate the parameter score-coefficient using different choices of functions
and the parameter RGB vector centroid (Paral, Chatterjee, and Rakshit, 2019a), which are
required to evaluate the competencies of the affine transformations (Paral, Chatterjee, and
Rakshit, 2019a). Algorithm 1 presents the method of computing the score-coefficient and
the RGB vector centroid (Paral, Chatterjee, and Rakshit, 2019a).
In the third phase, we consider a succeeding frame, denoted by, for matching purpose
and pre-process the frame in the same way as the shoe template, and then the optimal 2D
affine transformation providing the best matching result for that frame is found out (Paral,
Chatterjee, and Rakshit, 2019a). It is worthy to mention that, in order to determine the
optimal transformation corresponding to a frame, we evaluate every single transformation
belonging to a specified grid of transformations. A single transformation is assessed with
respect to a sublinear approximation (Korman et al., 2013; Paral, Chatterjee, and Rakshit,
2019a) of the similarity measure function, i.e. the ACSAD distance function (Paral,
Chatterjee, and Rakshit, 2019a). This approximation dramatically brings down the runtime
of the algorithm maintaining the accuracy almost unchanged and is very crucial for real-
life implementation. The sub-linear approximation is based on a handful number of pixels
sampled randomly from the shoe template and the target frame (Raskhodnikova, 2003).
Besides, we have formulated a branch-and-bound scheme (Paral, Chatterjee, and Rakshit,
2019a) in our implementation, which, though utilises the grids of transformations with
finer resolutions, restricts the number of transformations needs to be evaluated in the fast-
evolving nets. The runtime of the algorithm is reduced much further with this formulation.
Finally, we investigate how various choices of the score coefficients govern the
computations of the ACSAD distances and make impacts on the shoe detection results.
Now, it should be kept in mind that, with the shoe template available, the shoe detection
algorithm can be then repeatedly implemented for the consecutive frames captured, and in
this way, we make the algorithm fit for following people in real-life (Paral, Chatterjee, and
Rakshit, 2019a).
FIGURE 20.1 Overall schematic diagram of our proposed research framework.
Algorithm1: Computing Score Coefficient and RGB Vector Centroid (Paral, P., et al., 2019a)
Input: Colour Shoe Template Image
Output: (1) Score Coefficient, (2) RGB value of each cluster centroid of.
STEP 1: Apply DBSCAN on to generate arbitrarily shaped clusters of the pixels.
STEP 2: Calculate the total number of clusters.
STEP 3: Compute the length of each cluster (). Let the element of the cluster be denoted by.
STEP 4: Cumulate the R, G, and B values of the pixels of each cluster individually, for each channel.
i
The aggregate R, G, and B values corresponding to each cluster x (x = 1, 2,..., Tn) are saved in s x
(for i = R, G, B) as follows:
for x = 1: Tn
sxi = 0;
for y = 1: lx
( )
sxi + = I Ai k x , y ;
end
end
STEP 5: Compute nj x and ( m ( I A ) ) x (i = R, G , B ) , for each cluster x ( x = 1, 2,...,  n ) as

i
follows:
for x = 1:  n
( m ( I A ) ) x = sxi / lx ;
i
nj x = f ( lx ) ; nj x =1 / lx ( x = 1, 2,¼,  n ) F 1: nj x = 1 / lx
F 2 : nj x = 1 / lx F 3: nj x = 1 / lx2 F 4 : nj x = e -lx
end
STEP 6: Return nj x and m ( I A ) .
20.3 EXPERIMENTAL RESULTS

20.3.1 Experimental Setup
In this chapter, the experimental platform is a wheeled mobile robot (Paral, Chatterjee, and
Rakshit, 2019a) whose onboard computer is Raspberry Pi 1 model B+. The robot is furnished
with a stationary monocular camera and an ultrasonic range sensor for the sensing purposes.
20.3.2 Impacts of Score Coefficient Functions on ACSAD Distances

As we can see in Paral, Chatterjee, and Rakshit (2019a), the computations of the ACSAD
distances are governed by the score coefficient function. Mathematically, the lower values
of the score coefficients diminish the ACSAD distances and vice versa (Paral, Chatterjee,
and Rakshit, 2019a). However, for a specific score coefficient function, the ACSAD
distances converge at different values if we vary the parameter likeness threshold (Paral,
Chatterjee, and Rakshit, 2019a). So the computations of ACSAD distances heavily rely on
score-coefficients and likeness thresholds. Also, in Paral, Chatterjee, and Rakshit (2019a),
we observe that the convergences of the ACSAD distances lead to the determination of the
optimal transformations, and hence the shoe detection results. Thus, proper choice of score
coefficient functions influences the final matching results. In order to analyse the impact
of the various score-coefficient functions on the computations of ACSAD distances, we
take into account two image sequences, namely S1 and S2, whose frames are captured by
the monocular camera mounted on the robot, during the pursuit in different laboratory
environments significantly affected by various photometric variations. The sequences pose
various challenges, for example, photometric changes, improper brightnesses, and contrasts,
rotations, scale changes, background clutters. In Figure 20.2, the mean ACSAD distances
calculated considering the frames of the sequences S1 and S2 are plotted as a function of
likeness thresholds for various choices of the score-coefficient functions.
20.3.3 Performance Evaluations

We evaluate the performance of PICFAsT-Match for different choices of score coefficient
functions, qualitatively as well as quantitatively. For this purpose, we consider experimental
works on the sequences S1 and S2. Figure 20.3(a) depicts the matching image considered
for S1 and S2, as well as the shoe template determined from the image.
The qualitative evaluation can be done with Figure 20.3(b) which shows the tracking
outcomes of some representative frames of S1 and S2. As we can see in Figure 20.4, for
a likeness threshold, which was considered in Paral, Chatterjee, and Rakshit (2019a), the
results vary notably with the variations of the score coefficient functions. PICFAsT-Match
based on F2 improves the overall results obtained with the existing F1 based approach,
although the results deteriorated with F3 and F4.
To evaluate the quantitative effects of different choices of score coefficient functions
on the shoe detection results, we use four metrics, namely center location error (CLE),
overlap ratio (OR), distance precision (DP), and overlap precision (OP) (Yao et al., 2016).
In addition, we quantitatively compare the PICFAsT-match based on different choices of
score-coefficient functions with various contemporary detection and tracking algorithms,
namely CFAsT-Match, OBVTD, KCF (Henriques et al., 2015), and ECO (Danelljan et al.,
2017). Table 20.1 reports the average CLE, OR, DP, and OP values of the tracking algorithms
on the image sequences S1 and S2. From the experimental results, we get a clear view on
how the PICFAsT-Match approach with various choices of score-coefficient functions
performs in comparison to the other four algorithms.
FIGURE 20.2 The mean ACSAD distances vs likeness thresholds plot for various choices of the
score-coefficient functions.
(I) (II) (III) (IV) (V)

(a)
(VI) (VII) (VIII) (IX)

(b)
FIGURE 20.3 (a) The reference frame (matching image) and the shoe template
(magenta bounding box) considered for the sequences S1 and S2, (b) Qualitative comparisons
among the PICFAsT-Match approaches with different choices of score-coefficient functions in
terms of the visual target tracking outputs (bounding quadrilaterals) of some representative frames
of S1 (I-IV) and S2 (V-VIII).
TABLE 20.1 Average Comparison of Performances among the Considered Algorithms

Algorithms CLE OR DP (%) OP (%)
10.47 0.662 79.67 68.34
CFAsT-match 3.75 0.825 92.66 84.49
OBVTD 8.31 0.617 83.85 62.33
KCF 11.58 0.643 79.06 65.60
ECO F1 2.25 0.887 96.32 87.06
F2 2.34 0.882 96.14 86.79
PICFAsT-match F3 5.15 0.711 87.08 70.73
F4 6.09 0.708 86.29 69.41
20.4 CONCLUSION
In this chapter, a detailed study on possible choices of the variations of a crucial parameter
associated with PICFAsT-Match, namely score-coefficient is made to analyse the selected
variations’ sensitivities with respect to other system parameters, such as likeness threshold.
Also, the overall impacts of the selected variations on the ultimate shoe detection results
are assessed. These intensive studies along with various experimental results aptly
demonstrate how different choices of the score-coefficient functions can influence the
detection of shoes during the pursuit in a human-robot coexisting environment.
ACKNOWLEDGMENT
This work is supported by the Visvesvaraya Ph.D. Scheme for Electronics and IT, Ministry
of Electronics and IT, Government of India.
REFERENCES
1. Ankerst, M., Breunig, M. M., Kriegel, H. P., and Sander, J., 1999. OPTICS: ordering points
to identify the clustering structure. In Proceedings of the ACM SIGMOD International
Conference on Management of Data, 31 May – 03 Jun 1999, Philadelphia, PA. New York
(NY): ACM Press, 49–60.
2. Bellotto, N., and Hu, H., 2009. Multisensor-based human detection and tracking for mobile
service robots. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics),
39(1), 167–181.
3. Danelljan, M., Bhat, G., Khan, F. S., and Felsberg, M., 2017. ECO: efficient convolution
operators for tracking. In Proceedings of the IEEE Conference on Computer Vision and
Pattern Recognition, 21–26 July 2017, Honolulu, HI. Piscataway (NJ): IEEE, 6931–6939.
4. Ester, M., Kriegel, H. P., Sander, J., and Xue, X., 1996. A density-based algorithm for
discovering clusters in large spatial databases with noise. In Proceedings of the 2nd
International Conference on Knowledge Discovery and Data Mining, 02–04 August 1996
Portland, OR. Menlo Park (CA): AAAI Press, 226–231.
5. Henriques, J. F., Caseiro, R., Martins, P., and Batista, J., 2015. High-speed tracking with
kernelized correlation filters, IEEE Transactions on Pattern Analysis and Machine
Intelligence, 37(3), 583–596.
6. Jia, D., Cao, J., Song, W. D., Tang, X. L., and Zhu, H., 2016. Colour FAST (CFAST) match:
fast affine template matching for color images. Electronics Letters, 52(14), 1220–1221.
7. Korman, S., Reichman, D., Tsur, G., and Avidan, S., 2013. FAsT-match: fast affine template
matching. In Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition, 23–28 June 2013 Portland, OR. Piscataway (NJ): IEEE, 2331–2338.
8. Paral, P., Chatterjee, A., and Rakshit, A., 2019. Vision sensor-based shoe detection for
human tracking in a human-robot coexisting environment: a photometric invariant
approach using DBSCAN algorithm. IEEE Sensors Journal, 19(12), 4549–4559.
9. Paral, P., Chatterjee, A., and Rakshit, A., 2019. OPTICS-based template-matching for vision
sensor-based shoe detection in human-robot coexisting environments. IEEE Transactions
on Instrumentation and Measurement. Advance online publication. doi: 10.1109/
TIM.2018.2890400.
10. Raskhodnikova, S., 2003. Approximate testing of visual properties. In Proceedings of the 7th
International Workshop on Randomization and Approximation Techniques in Computer
Science, 24–26 August 2003 Princeton, NJ. Berlin, Heidelberg (Germany): Springer,
370–381.
11. Yao, R., Xia, S., Shen, F., Zhou, Y., and Niu, Q., 2016. Exploiting spatial structure from parts
for adaptive kernelized correlation filter tracker. IEEE Signal Processing Letters, 23(5),
658–662.
12. Yuan, J., Chen, H., Sun, F., and Huang, Y., 2015. Multisensor information fusion for
people tracking with a mobile robot: a particle filtering approach. IEEE Transactions on
Instrumentation and Measurement, 64(9), 2427–2442..
CHAPTER 21
Auto-adaptive Fuzzy Controllers

with Automatic Setpoint
Adjustment
A. K. Pal, Indrajıt Naskar, Sampa Paul, and J. Chakraborty
Department of AEIE, Heritage Institute of Technology, Kolkata, India.
ABSTRACT: In this chapter, an auto-adaptive scheme is presented for fuzzy PD

controllers. The proposed scheme is a novel approach based on an online modifying
parameter, which is updated by normalised change of error (Δe) of the process. Further, a
fuzzy-based dynamic Setpoint weighting (FSW) technique is incorporated. Implementation
of the proposed adaptive technique along with setpoint weighting scheme on fuzzy PD
controllers helped to control any system very effectively. Controlling the swing angle and
cart position of the pendulum arm with load in its inverted position is a very difficult job.
The effectiveness of the proposed setpoint weighting auto-adaptive fuzzy PD controller
(FSW-AAFPDC) is tested to control the dual pendulum arm at its inverted position on a
laboratory-based inverted pendulum.
Keywords: Adaptive tuning, Fuzzy logic control, Inverted pendulum, Setpoint weighting.
21.1 INTRODUCTION
Quite often the characteristics of a system are specified in terms of the system’s dynamic
response to a step input. Systems that store energy cannot instantaneously respond when
subjected to input or load variations. The transient response characteristic and its different
performance indices are considered as the most important aspects in system design. In any
effective system design, the dynamic system should not only be stable but also the system
response must reach and remain close to the desired reference value in minimum time
possible. To obtain an improved transient response, setpoint weighting (SPW) method is
widely accepted for conventional PID controllers. In this method, the idea is to multiply the
181
existing setpoint value with a fixed parameter (Visioli, 1999), and also for different control
action of different SPW factors (Prashanti and Chidambaram, 2000).
The methodology of SPW is extended here, for optimal design of fuzzy PD controller
based on the proposed adaptive tuning scheme. Here instead of operating with a fixed
weighing factor, a scheme is proposed very similar to the one in Sugeno (1985), where
the varying weighting factor is calculated online using instantaneous normalised error of
the controlled variable to get the dynamically weighted setpoint for tuning of the FLC.
Incorporation of dynamic SPW scheme for tuning of the controller greatly reduces the
process overshoot at the cost of a slight increase in rise time. In order to achieve the twin
objectives of low rise time as well as process overshoot, the aforementioned weighing
scheme is implemented through fuzzy logic both of whose inputs are computed based
on dynamic weighted set value. Thus, the response to the reference signal is expected to
improve because the reference is now itself being guided by the input values. The input
and output SFs are the chosen trial basis and are religiously maintained throughout for
performance comparison. The effectiveness of the proposed tuning scheme is tested on
a laboratory-based inverted pendulum, which is a non-linear and highly unstable system
(Zhang et al., 2012; Mudi and Pal, 1999).

The output scaling factor (SF) is similar in function to that of the controller gain and
should be considered a very important parameter (Mudi and Pal, 1999) due to its impact
on stability of the process. The output SF should therefore be evaluated very cautiously.
In the proposed scheme, the output SF of the fuzzy controller is updated according to the
process trend by a gain modification factor ε, which is dependent on the normalised Δe of
the process. The block diagram (Figure 21.1) shows the working model of auto-adaptive
fuzzy PD controller (AAFPDC).
FIGURE 21.1 Block diagram of AAFPDC.

Auto-adaptive Fuzzy Controllers with Automatic Setpoint Adjustment  183
As demonstrated in the diagram, the non-fuzzy adaptive parameter (ε) is determined from
the relation (Pal and Chakraborty, 2015),
|DeN |
e = d. e (21.1)
At, De = 0 ; e = d (21.2)
Here, ε is an adjustable parameter for the output SF Gu, and δ is a positive constant that
will bring require change in ε. The exponential factor in Eq. (21.2) is so chosen that it
takes care of the response stability, even when the normalised change of error is zero.
Thus, the output SF is regularly updated by the proposed deterministic rule defined
by Eq. 21.1. The study of any dynamic response reveals that the change of error is
maximum at the initial portion of the response, that is, before the first overshoot. The
increase of the controller output at that time by implementing ε, decreases the time
delay and makes the system faster. However, increases of controller output near the
Setpoint sometimes face the problem of large overshoot. In this chapter, that part is
well taken care of by the second component of the design, that is, setpoint weighting.
As normalised change of error gradually reduces to zero, its effect on the gain modifier
vanishes as presented by Eq. 21.2. This online adjusted output SF is expected to
improve the process performance since it assimilates the dynamics of the process (Pal
and Chakraborty, 2015, Pal, Mudi, and Maity, 2013). FLC uses SF Gu, while proposed
controller uses effective SF of εGu. All MFs for FLCs are defined on the common interval
[–1, 1] (Figure 21.2). The set of five symmetrical MFs generate an arrangement of
52 = 25 fuzzy if-then rules that would dictate the evaluation of the controller output (u)
in case of FPDC as given by Table 21.1.
FIGURE 21.2 MFs of e, ∆e, u for FLC.

TABLE 21.1 Fuzzy Rules for Computation of U

∆e/e NB NM ZE PM PB
NB NB NB NB NM ZE
NM NB NB NM ZE PM
ZE NB NM ZE PM PB
PM NM ZE PM PB PB
PB ZE PM PB PB PB
21.2.1 Fuzzy Setpoint Weighting (FSW) Scheme

In order to fine-tune the system transient response, dynamic SPW scheme is applied
to the adaptive controller where a fuzzy controller generated SPW factor β, which
is multiplied to the existing Setpoint (r) as shown in Figure 21.3. The input to the
controller are error (e’) generated from the deviation of the system response (y) from
the actual setpoint (r) and change of error (∆e’) generated from difference of two such
consecutive error signals separated by a time instant. The controller in discussion
then generates a control output (u’). This gives the dynamic weighing factor (β)
which engenders the variable Setpoint (r’). r’ now becomes the new reference to the
system, and setpoint following becomes more precise because the e and Δe to the fuzzy
controller gets revised. The input variables e and ∆e to the fuzzy controller driving the
system also adapts to the changing setpoint event. The set of governing mathematical
expressions are given below by Eqs. 21.3 to 21.8.
r'(t) = b(t) ´ r (21.3)
b(t) = r(t) - u'(t) (21.4)
e'(t) = r(t) - y(t) (21.5)

De'(t) = e'(t) - e'(t-1) (21.6)
e(t) = r'(t) - y(t) (21.7)
De(t) = e(t) - e(t-1) (21.8)
The input and output MFs for the setpoint weighting controller are defined over the
interval (–1, 1), by a set of very simple and least possible number of unbiased MFs. The
tuning scheme operates based on three symmetrical triangular MFs as demonstrated in
Figure 21.4. The three symmetrical MFs generate a set of 32 = 9 if-then rules given in Table
212. These rules help to determine the value of the SPW factor β to be multiplied with the
existing Setpoint.
The upper part of Figure 21.5 describes the FSW scheme. The self-adaptive non-fuzzy
approach for tuning of PD type fuzzy controller is shown in the lower part of Figure 21.5.
The methodology is extended towards obtaining a better and superior control response by
incorporating the idea of dynamic setpoint weighting with the proposed adaptive tuning
methodology (FSW-AAFPDC). This tuning scheme is expected to deliver most superior
performance in comparison to its conventional counterparts. The MFs, SFs, the rule bases,
and the adaptive scheme are not separately discussed here as they have already been dealt
with in the previous sections.
FIGURE 21.3 Block diagram of FSW-FPDC.
FIGURE 21.4 MFs of e’, ∆e’, u’ for the setpoint weighting controller.
TABLE 21.2 Fuzzy Rules for the Computation of Setpoint Weighting Controller Output
∆e/e NB ZE PB
NB PB PB ZE
ZE PB ZE ZE
PB ZE ZE ZE
FIGURE 21.5 Block diagram of FSW-AAFPDC.
21.3 RESULT
The basic structure of laboratory-based inverted pendulum, which is a non-linear and highly
unstable system consists of two flexible pendulum arms mounted to a cart on a moving rail.
In the absence of a stabilising controller, the pendulum arms, which have their centre of
mass above their pivot point, are unable to maintain their upright position. Control of an
inverted pendulum is a very common but difficult control engineering problem, based on
flight simulation of rockets and missiles during the initial stages of flight. The main aim
of the proposed design is to stabilise the inverted pendulum such that the position of the
carriage on the track is controlled quickly and accurately. The pendulum should remain
erected in its inverted position during such movements.
FIGURE 21.6 Schematic of the inverted pendulum system (Make: Feedback, UK).
The equations of motion for the inverted pendulum system are given by Eqs 21.9 and 21.10
(Pal and Chakraborty, 2015).
2
d 2x dx dq d2q
( M + m) 2 + K = Fv + ml sin q æçè ö÷ø - ml cos q 2 (21.9)
dt dt dt dt
2 2
( I + ml ) ddtq = mgl sin q - ml cosq ddt x
2
2 2
(21.10)
The performance of the proposed FSW-AAFPDC is evaluated in a laboratory-based inverted

pendulum system as shown in Figure 21.6. Both the cart position and pendulations of the
inverted pendulum are studied using FPDC, FSW-FPDC, and FSW-AAFPDC. From Figure
21.7 and Table 21.3, it is observed that the pendulum became stable in its inverted position
within 4 seconds. The control of swing of any inverted pendulum arm is very difficult.
However, the proposed FSW-AAFPDC is not only smoothly controlling the oscillation of
pendulum arm (Figure 21.8) but also minimising the other time response criteria as shown
in Table 21.3.
0.5
0.45
0.4
0.35
CART POSITION (METER)
0.3
0.25
0.2
0.15
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90 100
TIME (SECONDS)
FIGURE 21.7 Plot of cart position against time (for m = 0.2 Kg and l = 0.3 m) FPDC _____,
FSW-FPDC -------, FSW-AAFPDC ……).
3.3
3.25
INVERTED PENDULUM ANGLE (RADIAN)
3.2
3.15
3.1
3.05
3
0 10 20 30 40 50 60 70 80 90 100
TIME (SECONDS)
FIGURE 21.8 Plot of inverted pendulum angle against time (for m = 0.2 Kg and l = 0.3m)
(FPDC____, FSW-FPDC……, FSW-AAFPDC -------).
TABLE 21.3 Performance Comparison of Different Controller on Inverted Pendulum Control
Type of Controller %OS tr (sec) ts (sec) IAE ISE ITAE
FPDC 87.17 1.1 28.12 0.13 0.15 9.83
FSW-FPDC 31.17 1.4 14.41 0.51 0.05 2.13
FSW-AAFPDC 8.61 2.0 3.99 0.28 0.04 0.32
Short-term evaluation parameters like % over-shoot and settling time improved a lot in case
of FSW-AAFPDC. The improvement of long-term parameters like IAE and ISE are also
visualised in Table 21.3.
21.4 CONCLUSION
In this chapter, a new tuning technique for PD-type FLCs is suggested. This non-fuzzy
tuning scheme is designed to tune the FLCs by an online gain-updating parameter ε. For
further improvement of the system performance, a fuzzy-based dynamic setpoint weighting
scheme is incorporated with the designed controller. By implementing the proposed adaptive
algorithm along with setpoint weighting scheme, the objectives of optimisation of process
parameters are fulfilled. The proposed controller is well suited both for simulation-based
systems as well as real-time systems. In this chapter, the auto-tuning scheme is applied to
stabilise the laboratory-based inverted pendulum and that exhibits effective and improved
performance compared to its conventional fuzzy and Setpoint weighted fuzzy controllers.
Also, the proposed FSW-AAFPDC brings stability faster in an inverted pendulum and
reduces the processing time by decreasing the number of fuzzy if-then rules. The successful
implementation of the proposed scheme in the highly non-linear inverted pendulum
indicates the suitability of the controller on other complex processes.
REFERENCES
1. Visioli, A., 1999. Fuzzy logic based Setpoint weight tuning of PID controllers. IEEE trans.
on sys, man, and cybernetics, 29(6), 587–592.
2. Prashanti, G., Chidambaram, M., 2000. Setpoint weighted PID controllers for unstable
systems. J. Franklin Institute 337(2–3), 201–215.
3. Sugeno, M., 1985. Industrial Applications of Fuzzy Control. Amsterdam, The Netherlands,
Elsevier.
4. Zhang, Y., Shuang, M. B., Chen, X., and Qi, W., 2012. Stability Control of Inverted Pendulum
using Fuzzy Logic and Genetic Neural Networks. IEEE trans. on computer sc., and service
sys, 1495–1498.
5. Mudi, R. K., and Pal, N. R., 1999. A Robust Self-Tuning Scheme for PI- and PD-Type Fuzzy
Controllers. IEEE trans. on fuzzy systems, 7(1), 2–16.
6. Mitra, P., Dey, C., and Mudi, R. K., 2013. Fuzzy PI controller with dynamic Setpoint
weighting. Proc. Int. conf. on Frontiers of Intelligent Computing: Theory and applications,
51–58.
7. Manual no. 33–936S Ed01 122006, 2011. Manual on digital pendulum control experiments.
Feedback Instruments Ltd.
8. Ahmad, A., 2009. Active sway suppression techniques of a granty crane system. European
Journal of Scientific Research, 27(3), 322–333.
9. Pal, A. K., and Chakraborty, J., 2015. Design a fuzzy logic controller with a non-fuzzy
tuning scheme for swing up and stabilization of inverted pendulum. Advances in Intelligent
Systems and Computing, Publisher: Springer, 308, 221–230.
10. Pal, A. K., Mudi, R. K., and Maity, R. R. De 2013. A Non-fuzzy self-tuning scheme of PD-
type FLC for overhead crane control. Advances in Intelligent Systems and Computing,
Publisher: Springer-Verlag Berlin Heidelberg, 199, 35–42.
11. Hong, K. S., and Ngo, Q. H., 2009. Port automation: modelling and control of container
cranes. International Conference on Instrumentation, Control and Automation, 19–26.
12. Q. Xiong, W. J. Cai and M. He., 2006. A practical decentralized PID auto-tuning method for
TITO systems under closed-loop control. International Journal of Innovative Computing,
Information and Control, 2(2), 305–322.
13. Park, M. S., Chwa, D., and Hong, S. K., 2008. Antisway tracking control of overhead cranes
with system uncertainty and actuator nonlinearity using an adaptive fuzzy sliding mode
control, IEEE Transactions on Industrial Electronics, 55(11).
14. Pal, A. K., and Mudi, R. K., 2013. An adaptive PD-type FLC and its real time implementation
tooverheadcranecontrol, International Journalof Emerging Technologiesin Computational
and Applied Sciences, 6(2), 178–183.
15. Pal, A. K., Naskar Indrajıt and Paul Sampa 2018. Fuzzy-based Gain Adaptive Scheme for
Setpoint Modulated Model Reference Adaptive Controller. International Journal of Natural
Computing Research (IJNCR), 7 (4), 1–19.
CHAPTER 22
An Application of Multivariate
Control Chart for Online Process
Monitoring in SMEs
S. Samanta and S. Mondal
Department of Mechanical Engineering, Jalpaiguri Govt. Engineering College, India.
ABSTRACT: The main goal in this research is to control the process parameter of CNC
machine by using statistical process control tools such as multivariate control chart (i.e.
Hotelling T2 control chart, MEWMA control chart, and MCUSUM control chart) for small
manufacturing industry. This work is focused on the application of multivariate control
chart for online process monitoring in the industrial process control, quality control of
the product, and business. But when more than one quality characteristics come in one
product or process then, univariate control chart becomes less efficient as the multivariate
control chart is used instead of the univariate control chart. The result of this research was
compared, application of the univariate control chart (X̅ chart and R chart) and multivariate
control chart (Hotelling T2 control chart, MEWMA control chart, and MCUSUM control
chart) for online process monitoring. This study was essential to determine one of the best
options of this control chart for the industrial process and for online monitoring the quality
of the product.
Keywords: CNC machine, Control chart, Multivariate control chart, Statistical process
control, Univariate control chart, Statistical quality control.
22.1 INTRODUCTION
In statistical (quantitative) quality control (SQC), control charts are the most powerful
statistic tools which is used to monitor the process parameter and improve the quality of
the product by reducing variable in the manufacturing process. SPC aims to achieve higher
product quality and lower the production cost due to the minimisation of the defected
product.
191
Nowadays, control charts are becoming extensively accepted in the manufacturing

industry. In this research work, two types of control charts have been used. They are
univariate control chart and multivariate control chart. Univariate control chart analysis
is a method for analysing data on a single variable at a time, where we are observing only
one aspect of phenomenon at a time. Multivariate control chart analysis is a method for
analysing data on a multiple variables at a time, where we are observing more than one
aspect of the phenomenon at a time.
In this project, normal probability distribution used to check all data are normally
distributed or not. There are two types of data sets used, i.e. univariate data set and
multivariate data set. While the univariate version of normality is pretty simple to think
about, multivariate normality paints a little more of a complex picture. This is because
it moves from a graph that can be imagined in two dimensions to higher and higher
dimensional graphs. In this project, Anderson-Darling normality test used to verify all
univariate data are normally distributed or not. And also, Royston’s H test used among all
of them is that all multivariate data are being checked normally distributed or not.
Control chart is a graphical representation of the collected information. A control chart
always has a central line for the average, an upper line for the upper control limit, and a
lower line for the lower control limit. Univariate control charts are used to monitor a single
independent process characteristic and multivariate control charts are used to monitor two
or more interrelated process variables, and multivariate control charts are necessary when
process variables are correlated.
There are several types of univariate control chat, but among them, X̅ chart and R chat
are very necessary to measure the central tendency of a variable. X̅ chart is used to evaluate
the consistency of process averages by plotting the average of each sub-group. R chart is
used to evaluate the consistency of process variation and also for reduction of the range
of measurable quantitative design dimensions of variable data and also for different types
of multivariate control chart (i.e. Hotelling T2 control chart, multivariate exponentially
weighted moving average (MEWMA) control chart, multivariate cumulative sum control
chart (MCUSUM), and, etc.), but among them Hotelling T2 control chart, MEWMA control
chart, and MCUSUM control chart used in this research work.
22.2 METHODOLOGY
A small medium enterprises (SMEs) has made some final sample component through
different machining operation (i.e., CNC Milling, grooving, grinding, etc.). Important
quality parameters were identified of this final product. As a quality parameter, the
diameter (the length) of the product was measured with the help of vernier caliper, after
that, normality test to determine with the help of experimental measured value, and then all
data is justified and analysed by univariate control chart and multivariate control chart. For
batch production, various industries used online process monitoring in an efficient manner
to improve the quality of product or process.
An Application of Multivariate Control Chart for Online Process  193
FIGURE 22.1 Schematic diagram of the flow chart of working process.
In this study, the outer diameter and inner diameter of the product (synchro ring) is a major
quality characteristic of this product (synchro ring). According to the study, the dimension
of the outer diameter and the inner diameter is defined as [162]^(±0.1) and [102]^(±0.05).
In this work, n = 60 random samples have been taken, and each variable (outer diameter
and inner diameter) of the sample following a normal distribution with means x̅ and a
standard deviation σ, Np ( x̅ , σ).
FIGURE 22.2 Product details and design diagram.
22.3 DISCUSSION
As per the quality of the product, two quality parameters are taken such as outer diameter
and inner diameter. In this work, 60 samples are taken and analysed and their quality is
based on those two quality parameters. All the statistical processes can be controlled online.
At the end of the production line, there will be a dimension measuring machine that will
automatically measure the dimension of the product. Then all the measuring data send
to computer and computer analysis all data using different statistical software (MATLAB,
Minitab, Statistica and R Project). This statistical software creates control chart which helps
to understand easily.
FIGURE 22.3 Normal probability plot of product (outer diameter and inner diameter).
The graphical representation shows (Figure 22.3) the normal distribution of data of outer
diameter and inner diameter of product. So all data are normally distributed and SPC tool
can be used for further process. Here, p values are more than 0.05, it can be said that all data
are normally distributed. If p values are less than 0.05, then reject null hypothesis (H0) and
also all data are not normally distributed.
FIGURE 22.4. x̅ chart of outer and inner diameter of the product.
Only x̅ chart indicates that 15th sample is rejected due to outer diameter quality parameter.
But other control charts have accepted the quality of 15th sample based on quality specified.
So it is clear that x̅ chart (univariate control chart) gave false signal for 15th sample.
FIGURE 22.5 R chart of outer and inner diameter of the product.
This happens because the univariate control charts cannot take into account the
relationship between variables, while multivariate control charts do this. Multivariate
control charts take into account the possible correlation between the variables.
FIGURE 22.6 Chi-square plot and hotelling T2 control chart.

Multivariate normality test is done for checking whether multivariate data are normally
distributed or not. Determining whether data is multivariate normally distributed is usually
done by Royston’s H test which depends on chi-square statistic. As the p values of all factors
are more than 0.05, it can be said that all the three factors are statistically significant. It also
said that all data are normally distributed. According to these three multivariate control
chart, all points fall within the control region and this indicates that the sampled data was
drawn from a statistically monitored and controlled production process.
FIGURE 22.7 MEWMA control chart and MCUSUM control chart.
22.4 CONCLUSION
The results obtained in this project suggest that the multivariate control chart is an excellent
statistical tool for the monitoring of machining process with multiple different qualities
characteristic. When there is more than one quality parameter present and also monitor all
quality at a time, this work proposed to use multivariate control chart and avoid the false
signals which are associated with separate univariate control chart.
REFERENCES
1. Waterhouse, M., Smith, I., Assareh, H., et al., 2010. ‘Implementation of Multivariate Control
Charts in a Clinical Setting’, International Journal for Quality in Health Care., 22(5),
408–414.
2. Moraes, D. A. O., Oliveira, F. L. P., Duczmal L. H., 2015, ‘On the Hotelling’s T2, MCUSUM
and MEWMA Control Charts’ Performance with Different Variability Sources: A Simulation
Study’, Brazilian Journal of Operations & Production Management., 12(2), 196–212.
3. Alves, C. C., Samohyl, R. W., Henning, E., 2013. ‘Application of Control Charts For
Monitoring an Industrial Process’, TECNO-LÓGICA, Santa Cruz do Sul., 17(2), 101–107.
4. Bersimis, S., Panaretos, J., Psarakis, S., 2005. Multivariate Statistical Process Control Charts
and the Problem of Interpretation: A Short Overview and Some Applications in Industry,
Research Gate, 1−7, https://www.researchgate.net/publication/24114966.
5. Lowry, C. A., Montgomery, D. C., 2007. A Review of Multivariate Control Charts, 27,
800–810, IIE Transactions, Taylor & Francis, England (London) and Wales, http://dx.doi.
org/10.1080/07408179508936797.
6. Tôrres, A. R., Grangeiro S. J., Fragoso W. D., 2015. Multivariate Control Charts For
Monitoring Captopril Stability, Micro chemical Journal, 118, 18, 259–265. Science Direct,
www.elsevier.com/locate/microc.
CHAPTER 23
Analysis and Performance

Indexing of Khagaria 132/33 kV
Sub-station Using ETAP
Load Flow Solutions
Archisman Kar,1 Debashis Jana,1 Chiranjit Dey,2 and
Ranjita Chowdhury1
1
Department of Electrical Engineering, Institute of Engineering & Management, Kolkata 700091, India.
2
Department of Electrical Engineering, CITM, Bainchi, Hooghly-712134, India.
1
archisman_kar@outlook.com, 2debashis.jana@iemcal.com
ABSTRACT: The load flow analysis is performed to present the characteristic of a power
system network under stable or unstable conditions for evaluating the various operating
states of an existing power system. Power system engineers always rely on effective load
flow solutions to plan efficient and reliable power networks with flexibility in power
exchange activity. After all the generated power has to be transmitted and distributed to
the consumers with specified quality and reliability as and when consumers need. The
primary importance is given to this chapter on load flow simulator for collecting various
data obtained after precise solutions. Electrical Transient Analyser Program (ETAP) allows
in choosing from several different methods in order to accomplish the best calculation
efficiency and accuracy. ETAP analyser computes bus voltages, currents, branch power
factors, and power flows throughout the entire electrical network. This chapter presents
detailed power flow monitoring of the Khagaria 132/33 kV sub-station for further analysis
using built in intelligent graphics of ETAP. Hence, in connection to operation and control
of the said sub-station, the load flow solution has been analysed for loss minimisation,
maintaining stable voltage profile within prescribed tolerance, economic system operation,
interconnection of power networks, meeting increase load demand, etc. The analysis of the
proposed simulation result is used to address the problem of voltage fluctuation and poor
power quality. The performance indexing of voltage and power factors of this sub-station
199
puts effective compensation implication, optimal operation identification of capacitors/

inductors sizing and sitting.
Keywords: Power system network, Effective load flow solutions, ETAP analyzer, Flow
monitoring, Stable voltage profile, Economic system operation, Distribution management
system, Gauss–Seidel method.
23.1 INTRODUCTION
In power system engineering, one of the important part is load flow analysis. Day by day the
demand of power is increasing by nature so the study about power requirement, nature of
load, demand forecasting is necessary. In A.C. power system, there is three-part generation,
transmission, and distribution, power flows from generation to load through different bus
transmission line and distribution line. This flow of active power and reactive power is
called power flow process.
In the recent past, it has been noted that most of the grid collapses have been attributed
to protection system failure or malfunctioning. The reports on the North American
blackout (August 2003) and the more recent Indian grid collapse (July 2012) have all
emphasised the need for “Protection Management System.” As a commendation of the
Enquiry Committee headed by chairperson CEA on grid disturbances in NEW grid on 30
and 31 July 2012, Ministry of Power constituted a task force on power system analysis under
contingencies in December 2012. The task force recommends creation and maintenance
of protection database under RPCs. Accordingly, Secretary (Power), Govt. of India, in a
meeting with Ministry of Power held on 11March 2014 directed all RPCs to implement
the recommendations of their port submitted by ‘Task Force’ in a time-bound manner.
Following the decision of Ministry of Power, ERPC as a pioneering venture has taken up the
project ‘Creation and Maintaining a Web-based Protection Database and a Desktop based
Protection Setting Calculation Tool for Eastern Regional Grid’.
M/s PRDC has been awarded with the order by ERPC to implement the project in
its entirety including creation of database and supply of software and hardware along
with necessary power system analysis relevant for the project. Eastern Regional (ER)
Grid comprises of the electrical system of the states of Bihar, Jharkhand, West Bengal,
Odisha, Sikkim, and area under DVC. The major constituents of ER grid are the State/
UT Transmission and Distribution Utilities, Central Transmission Utility, State and Central
sector Generating Companies, DVC, CESC, DPL, IPPs, and Private Sector Transmission
and Distribution Utilities. The ER covers a geographical area of 4,25,432 square km with
an installed capacity of 36,575 MW. In addition to this, the region has an installed capacity
of 7,840 MW in the form of CPPs. Eastern Regional Power Committee (ERPC) formed
by Ministry of Power is entrusted for facilitating the integrated operation of the power
system in the region. As a pre-requisite of building the protection management system, the
first step that was envisaged is the electrical modelling of the entire network data under
ER system from 765 kV to 132 kV and 66 kV for Sikkim and carrying out the base case
Analysis and Performance Indexing of Khagaria 132/33 kV Sub-station  201
operational load flow analysis and short-circuit studies. Creation and maintaining of a web-
based Protection Database and Desktop-based Protection setting calculation tool for ER
Grid is necessary which is done by load flow studies of the different incoming and outgoing
Generator Bus, Load Bus, etc. Buses via Electrical Transient Analyser Program (ETAP)
Computer-Based Software (Operational Load Flow Studies, ERPC).
There are so many research articles available about load flow analysis which deals with
simulated 132 kV sub-station based on actual data from sub-station by ETAP where load
flow is performed and under voltage problems also overcome.
Load Flow Analysis of 132kV sub-station using ETAP Software: Power is essentially
required for the development of any country. To maintain the generation of electric
power at adequate level, the power has to be transmitted in proper form and quality to the
consumer. This research paper deals with the simulation of 132 kV sub-station in ETAP
with detailed load flow analysis and also to overcome the problem of under voltage. The
results are based on actual data received from 132 kV sub-station. Patil and Namekar (2018)
enlighten on load flow and short-circuit analysis of 132/33/11 kV sub-station using ETAP
(Kapahi, 2013).
Load Flow and Short-Circuit Analysis of 132/33/11KV Sub-station using ETAP:
Electrical power system provides a vital service to the society. For the healthy operation of
electrical power generation, transmission, and distribution, it is important that the system
should be balanced. Load flow is a basic requirement to conduct power system analysis of
any system. The load flow gives us information about voltages, real, and reactive power
generated and absorbed and line losses across the entire system. This research paper deals
with the simulation of 132/33/11kV sub-station. The analysis is done by using advance
software ETAP with detailed load flow analysis. Also, we have carried short-circuit study of
132/33/11 kV sub-station system using ETAP software. From the ETAP generated load flow
details and the short-circuit details are studied. (Patil and Namekar, 2018).
In this chapter, we discuss about analyse and performance indexing of a 132/33 kV
BSPTCL sub-station named Khagaria using ETAP.
23.2 METHODOLOGY AND NETWORK FORMULATION

Power flow method is a fundamental tool in application software for distribution
management system. In the past decades, a mass of methods to solve the distribution power
flow problem have been developed and well documented. These methods can be roughly
categorised as node-based methods and branch-based methods. The first category used
node voltages or currents injection as state variables to solve power flow problem. In this
category, the most notable methods include network equivalence method, Z-bus method,
Newton–Raphson‘s algorithm, Gauss–Seidel Method, Fast Decoupled algorithm. The
second category utilised branch currents or branch powers as state variables to solve power
flow problem. The backward/forward sweep based methods and loop impedance methods
can be categorised in this group.
The complex power produced by the source into ith bus of a power system is
Si = Pi + jQi = Vi I i* , i = 1, 2,¼, n
Since its convenient to work with Ii instead of I i* , we take complex conjugate in the
above equation,
Pi - jQi = Vi* I i , i = 1, 2,3,¼, n
æ n ö
å
Substituting Ii= ç (YikVk ) ÷
è k =1 ø
æ n ö
å
Pi - jQi = Vi* ç (YikVk ) ÷ , i = 1, 2,¼, n
è k =1 ø
Equating real and imaginary parts, we get
æ æ n öö
Pi ( Real power ) = Real ç Vi* ç (YikVk ) ÷ ÷
ç ÷å
è è k=1 øø
æ æ n öö
Qi ( Reactive power ) - Imaginary ç Vi* ç (YikVk ) ÷ ÷ .
ç ÷ å
è è k=1 øø
e jdi e jdi
Let, Vi = Vi Vk = Vk
e jqik
Yik = Yik
n
Pi ( Real power ) = Vi åV
k =1
k Yik cos (qik + d k - d i )
n
Pi ( Reactive power ) = - Vi åV
k =1
k Yik cos (qik + d k - d i )
(i=1,2,3,…,n)
These two equations are called power flow equations. There are n real and n reactive power
flow equations giving a total of 2n power flow equations.
• Load Flow Analysis using ETAP Software for Network Simulation
In a bid for – planning and coordination of relay in distribution system using ETAP,
carried out a load flow analysis to investigate the performance of the electrical system
during normal and abnormal operating conditions, providing information needed to
optimise circuit usage; develop practical voltage profiles; minimise MW and MVAR losses;
develop equipment specification guidelines; and identifies transformer tap settings. ETAP
is computer-based software that simulates real-time steady-state power system operation,
enabling the computation of system bus voltage profiles, real and reactive power flow and
line losses, etc.
• Newton–Raphson Method
Newton–Raphson method is an iterative technique for solving a set of various non-
linear equations with an equal number of unknowns. There are two methods of
solutions for the load flow using Newton–Raphson method. The first method uses
rectangular coordinates for the variables while the second method uses the polar
coordinate form. Out of these two methods, the polar coordinate form is used widely.
The various advantages of Newton–Raphson method are as follows: It possesses
quadratic convergence characteristics. Therefore, the convergence is very fast; the
number of iterations is independent of the size of the system. Solutions to high accuracy
is obtained nearly always in two to three iterations for both small and large systems;
the Newton–Raphson method convergence is not sensitive to the choice of slack bus;
overall, there is a saving in computation time since fewer number of iterations are
required. (circuitglobe.com › Newton–Raphson method)
• Accelerated Gauss–Seidel Method
An accelerated Gauss–Seidel method is used to solve the linear system equations. This
method is named after the German scientist Carl Friedrich Gauss and Philipp Ludwig
Siedel. It is a method of iteration for solving n linear equation with the unknown variables.
This method is very simple and used in digital computers for computing. The Gauss–
Seidel method is the modification of the Gauss-iteration method. This modification
reduces the number of iteration. In this method, the value of unknown immediately
reduces the number of iterations, the calculated value replace the earlier value only at
the end of the iteration. Because of it, the Gauss–Seidel methods converges much faster
than the Gauss methods. In Gauss–Seidel methods, the number of iteration method
required for obtaining the solution is much less as compared to Gauss method. In the
Gauss–Seidel method, a large number of the iteration is required to arrive at the specified
convergence. The rate of convergence can be increased by the use of the acceleration
factor to the solution obtained after each iteration. The acceleration factor is a multiplier
that enhances correction between the values of voltage in two successive iterations. After
calculating the required number of iterations, we calculate the value of new estimated
bus voltage and is finalised, and this new value replaces the previously calculated value.
For real and imaginary components of the voltage, different accelerating factors are used
(circuitglobe.com/gauss-seidel-method).
23.2.1 About ETAP

ETAP is the most efficient place to design and simulate power system. ETAP is the most
comprehensive analysis platform for the design, simulation, operation, and automation
of generation, distribution, and industrial power systems. ETAP is developed under an
established quality assurance program and is used worldwide as a high-impact software.
ETAP is completely localised in four languages with translated output reports in six
languages.
As a fully integrated enterprise solution, ETAP extends to a Real-Time Intelligent
Power Management System to monitor, control, automate, simulate, and optimise the
operation of power systems. It has very big capability to perform load flow study, short-
circuit study, arc flash, protection coordination studies, cable pulling, cable ampacity
study, and much more.
23.2.2 Khagaria Sub-station

• The Khagaria Sub-station under the Bihar State Power Transmission Company
Limited (BSPTCL) involves 2 ´ 160 + 2 ´ 50 MVA, 220/132/33 kV grid sub-station
with sub-station automation system in Khagaria district of Bihar.
• The approximate cost of the project was Rs. 64.06
Crores. Distance from main cities
• From Patna – 170.6 km
• From Gaya – 231.4 km
• From Bhagalpur – 87 km
(Khagariya Sub-station, BSPTCL, GoI).
FIGURE 23.1 Khagariya substation in single line diagram in ETAP.

23.3 RESULT ANALYSIS AND DISCUSSION

In this project, we study a 33/11kV sub-station and find the load flow study of the same
using the ETAP software. We have collected the details of the different components of the
sub-station circuit and completed the load flow study accordingly.
The OBJECTIVE consists of determining the complete datasets of the system under
study like magnitudes and phase angle of voltages at each bus and active and reactive
power flow in each line, operating quality of each bus, load level of transformers, critical
conditions, etc. In solving a power flow problem, the system is assumed to be operating
under balanced conditions and a single-phase model is used in ETAP software.
After the simulation work was completed, the computer runs the simulation and
generated a document file with all the data profiles and results of the simulations in it.
TABLE 23.1 ETAP Generated Summary of Power Flow Simulation of Khagariya Sub-station
Summary of Total Generation, Loading and Demand
MW Mvar MVA %PF
Source (Swing Buses) 107.757 68.499 127.686 84.39 Lagging
Source (Non-Swing Buses) 0.000 0.000 0.000
Total demand 107.757 68.499 127.686 84.39 Lagging
Total motor load 82.000 21.071 84.664 96.85 Lagging
Total static load 11.986 3.027 12.362 96.96 Lagging
Total constant I load 0.000 0.000 0.000
Apparent losses 13.771 44.401
System mismatch 0.000 0.000

TABLE 23.2 ETAP-Generated Bus Loading Summary of the Sub-station Simulation
206
Bus Loading Summary Report


Directly Connected Load Total Bus Load

Constant kVa Constant Z Constant I Genetic
Bus KV Rated Amp MW Mvar MW Mvar MW Mvar MW Mvar MVA %PF Amp Percent Loading
32 KV Main 32.000 13.600 8.429 2.060 1.277 37.740 91.3 866.0
32 KV Transfer 32.000 8.160 5.057 1.251 0.776 37.740 91.3 866.0
132 KV Transfer 132.000 24.000 7.586 3.131 54.494 84.9 279.4
132 KV Main bus 132.000 9.037 85.5 46.3
BTPS Bus 132.000 12.240 1.572 0.974 127.686 84.4 558.5
Bus 3 132.000 24.000 3.972 81.906 92.0 420.0
Bus 4 132.000 18.075 85.5 92.7
Bus 8 132.000 22.593 85.5 115.9
Bus 9 132.000 45.736 84.0 234.5
Bus 12 132.000 55.090 94.9 282.5
Bus 13 132.000 108.906 90.4 558.5
Bus 23 32.000 16.663 91.9 382.3
Bus 24 32.000 29.532 98.7 677.7
Bus 25 32.000 42.163 90.7 967.5
Bus 26 32.000 18.611 84.6 427.0
Bus 27 32.000 16.663 91.9 382.3
Bus 29 32.000 11.116 85.0 255.1
Intelligent Electrical Systems: A Step towards Smarter Earth
Bus 30 32.000 17.825 81.6 409.0

Khagariya New 32.000 18.423 85.0 427.0
Khagariya Old 32.000 11.072 85.0 255.1
Mansi New 32.000 27.131 100.0 677.7
Mansi Old 32.000 16.250 85.0 409.0
Pumea Bus 132.000 27.972 100.0 150.4
* Indicates operating load of a bus exceeds the bus critical limit (100.0% of the continuous ampere rating).
* Indicates operating load of a bus exceeds the bus marginal limit (95.0% of the continuous ampere rating).
TABLE 23.3 ETAP-Generated Critical Summary Report of Fault Sections in Sub-station Simulation
Critical Report
Deviced ID Type Condition Rating/Limit Unit Operating % Operating Phase Type

132 KV Transfer Bus Under Voltage 132.000 kV 112.586 85.3 3 Phase
132 KV Main bus Bus Under Voltage 132.000 kV 112.59 85.3 3 Phase
32 KV Main Bus Under Voltage 32.000 kV 25.16 78.6 3 Phase
32 KV Transfer Bus Under Voltage 32.000 kV 25.16 78.6 3 Phase
Bus 12 Bus Under Voltage 132.000 kV 112.59 85.3 3 Phase
HEL 2 Transformer Overload 20.000 MVA 45.74 228.7 3 Phase
Khagariya New Bus Under Voltage 32.000 kV 24.91 77.8 3 Phase
Khagariya Old Bus Under Voltage 32.000 kV 25.06 78.3 3 Phase
Mansi New Bus Under Voltage 32.000 kV 23.11 72.2 3 Phase
Analysis and Performance Indexing of Khagaria 132/33 kV Sub-station
Mansi Old Bus Under Voltage 32.000 kV 22.94 71.7 3 Phase


Pumea Bus Bus Under Voltage 132.000 kV 107.39 81.4 3 Phase

207
23.4 CONCLUSION
Here, the study of the Khagariya sub-station and simulation of this specific scenario in
Etap software has given us a comprehensive report on the load flow in this system scenario,
has pointed out the problems and critical situations that arose out of the system, and it
corroborates the real-life turn of events on the ground, where HEL2 transformer burnt
out, as shown in Table 23.3 and Figure23. 2, during a fault under this exact schematics and
scenario.
FIGURE 23.2 Percentage operating load on the various buses in the simulation result in ETAP.
It goes to show that this software simulation-driven approach can be deployed on power
grid systems of any load or size, can be retrofitted to virtually any control segments
irrespective of its modern or not, and can help us save huge amount of resources in
preventive measures and maintenance suggestions.
FIGURE 23.3 Percentage power factor values of various buses in simulation Result in ETAP.
REFERENCES
1. Khagariya Substation, Bihar State Power Transmission Corporation Limited, State
Government of Bihar, India
2. Patil, B., and Namekar, S., 2018. Load Flow & Short Circuit Analysis of 132/33/11KV
Substation using ETAP. International Journal of Applied Engineering Research, 13(11),
9943–9952. http://www.ripublication.com 9943
3. Hussain, Z., Hussain, W., Ullah, R., and ud Din, Z., 2017. Load Flow Analysis of 132/11kv
Substation Using Etap: A Case Study. Sarhad University International Journal of Basic and
Applied Sciences, 5(1), 40–48.
4. Kapahi, R., 2013. Load flow analysis of 132kV substation using etap software. International
Journal of Scientific & Engineering Research, 4(2), p. 5. http://www.ijser.org
5. Idoniboyeobu, D. C., and Ibeni, C., Analysis for Electical Load Flow Studies in Port
Harcourt, Nigeria, Using Newton Raphson Fast Decoupled Techniques.-American Journal
of Engineering Research (AJER), e-ISSN: 2320–0847, p-ISSN: 2320–0936, 6(12), 230–240
6. Creation and maintaining a Web based Protection Database and Desktop based Protection
setting calculation tool for Eastern Regional Grid: Power Research and Development
Consultants Private Limited (Eastern Regional Power Committee)
7. # 5, 11th cross, 2nd stage, West of chord road, Bangalore - 560 086, Karnataka Tel: +91-
80- 2319215/23192168 - Web site: www.prdcinfotech.com, http://erpc.gov.in/wp-content/
uploads/2016/12/OLFS-Operational_LFA-Final_Report.pdf, https://etap.com › docs ›
default-source › whats-new › etap-16-readme, https://circuitglobe.com/gauss-seidel-
method.html, https://circuitglobe.com › newton-raphson-method
CHAPTER 24
Fractional Order PID controller

for Setpoint Tracking and Load
Rejection
Sudipta Ghosh,1 Arijit Bhowmik,1 DipaBala Sarkar,1 Anirban Bhatta,2
and Biswajit Chakraborty3
1
Department of Electrical and Electronics Engineering, St. Mary’s Technical Campus, Kolkata, India.
2
Department of Electrical Engineering, University of Engineering and Management, Kolkata, India.
3
Department of Electrical Engineering, Jadavpur University, Kolkata, India.
1
sudiptaghosh0591@gmail.com, bhowmik11arijit@gmail.com, sarkardipa555@gmail.com
2
anirban.bhatta91@gmail.com
3
biswajit.chakraborty1991@gmail.com
ABSTRACT: This chapter deals with the minimisation of the problem of overshoot
in transient phase and load rejection phase. Zeigler–Nichols tuned PID controller can
be regarded as a popular controller among process industries. But it cannot restrict the
overshoot to the satisfactory limit. Hence, fixed setpoint weighting and variable setpoint
weighting provide improvement to some extent only in the transient phase. Moreover, the
enhancement in both phases can be observed by applying dynamic setpoint weighting.
On the other hand, it has to be remarked that the enhancement in transient phase in
case of dynamic setpoint weighting is not within the satisfactory limit. So, for the further
enhancement, fractional order PID (FOPID) (denoted by PIλDµ) has been improvised
here. The simple design and easy tuning scheme of the parameters λ and µ have added
more advantage. The superiority of the proposed scheme has been evaluated through the
comparison of the performance indices along with the responses applying to the two basic
models and their perturbed models.
Keywords: Setpoint tracking, Load rejection, Ziegler–Nichols, Fixed setpoint weighting,

Dynamic setpoint weighting, Fractional order proportional integral derivative.
211
24.1 INTRODUCTION
PID controller has been considered a very popular and useful controller in most of the
process industries for its easily tuneable scheme. Profuse researches have been focused
on the tuning method of PID controllers amongst them Ziegler–Nichols (ZN) tuning
rules are mostly acceptable. It has been revealed that Ziegler–Nichols (ZN) tuned PID
controller cannot restrict the response within the acceptable limits throughout the
process. To overcome the high overshoot in setpoint kick as well as in load disturbance,
several techniques have been used. For this purpose, fixed setpoint weighting (FSPW),
where a weighting factor is multiplied with setpoint and applied to the proportional
gain, is used. Hence, it is observed that the usage of VSPW tuned PID controller is
beneficial for decreasing high overshoot but failed to reduce the rise time. Therefore, for
reducing the process rise time, variable setpoint (multi valued) weighting method is quite
satisfactory. But this method shows extremum (maximum or minimum) value of Integral
Time Absolute Error (ITAE), Integral Square Error (ISE), Integral Absolute Error (IAE).
It is necessary to mention that both fixed and variable setpoint weighted PID controller
cannot influence any change in load disturbance. For improving the performance under
the load disturbance as well as at setpoint changes, the dynamic setpoint weighting
method is introduced. But it has certain limitations for improving the transient response.
For enhancing the performance in transient period and also in load rejection behaviour,
fractional order PID (FOPID) (denoted by PIλDµ) has been improvised here. The idea
of fractional order algorithms for the control of dynamical systems was first introduced
by Podlubny over the classical PID controller. In this proposed strategy, two extra
parameters have been implemented (λ > 0, µ >0) for the improvement of response in
load disturbance and also at the setpoint kick. The values of kp, ki, kd is predetermined
by Ziegler–Nichols (ZN) tuning method. Moreover, the steady-state error can also be
minimised by tuning these two parameters (λ and µ). On the other hand, from it can be
noticed that the flexibility of fractional order Proportional Integral Derivative (FOPID)
controller is more than the other existing techniques and provides more prospects for
improving the dynamic properties of fractional order control system. The superiority of
the proposed controller has been established by determining the performance indices
such as percentage overshoot (% Os), rise time (tr), peak time (tp), delay time (td), Integral
Absolute Error (IAE), Integral Square Error (ISE). The values of λ and µ are selected
through trial and error procedure.

The block diagram representation of FOPID controller is shown in Figure 24.1. Reason
behind the use of proposed FOPID controller than the other existing methods is easy
circuitry and tuning method, moreover, there is no need of weighted setpoint. Here only
the integral order (λ) and derivative order (µ) are used in addition with the normal P, I, D
controllers. The integral order λ and the derivative order µ are associated with the integral
gain ki and the derivative gain kd , respectively.
Fractional Order PID controller for Setpoint Tracking and Load Rejection  213
FIGURE 24.1 Block diagram representation of FOPID controller.
Basically, FOPID is the modified method of conventional PID controller, where two
parameters that are integral order (λ) and derivative order (µ) are added which are fractional
in nature. Usually, FOPID is denoted by PIλDµ.
The differential equation of FOPID in time domain is given below:
u ( t ) = k p e ( t ) + ki J tl e ( t ) + kd Dtµ e(t )
The Laplace Transformation of the expression is given below:

k
u ( s ) = k p + li + kd sµ
s
TABLE 24.1 Variation of the Performance Indices with the Change in Main Tuning Parameter
Steady State Load
OS% tr ts Damping
Error Disturbance
λ (increased) Increased Minor change No effect Decreased No effect Decreased
λ (decreased) Decreased Minor change Increased Increased No effect Increased
µ (increased) Decreased No effect Decreased Decreased Increased Minor change

µ (decreased) Increased No effect Increased No effect Decreased Minor change
Here, it is seen that in the case of FOPID, there are five parameters available for tuning.
The three gain parameters (i.e. Kp, Ki, Kd) of the PID controller are tuned by ZN tuning
method. Then the other two parameters λ and µ of FOPID vary to get the desired output
by trial and error method since there is no certain thumb rule for the adjustment of λ
and µ. As in the denotation of FOPID, it is seen that λ and µ are placed to the power
of integral controller and derivative controller, respectively. It is observed through
simulation that the performance indices are influenced with changing the values of λ and
µ. A pattern of the variation of the performance indices with change in λ and µ are shown
in Table 24.1. From Table 24.1, it is seen that an adjustment is needed to get a better
response in transient, setpoint changing as well as in load disturbance. Small values of
λ and µ ensure the beneficial characteristics at the high-frequency specification and the
system gain will not change drastically. This approach can apply the transient and steady-
state error Compensation, but with a modification of the value of λ, µ(≠0). Since the final
value theorem states that fractional system provides negligible steady-state error.
24.3 RESULTS
Performance of the proposed FOPID controller is compared with fixed setpoint weighting
(FSPW), variable setpoint weighting (VSPW), and dynamic setpoint weighting (DSPW) for
two types of process one is second-order linear process
æ e -0.3s ö
G
ç p ( s ) = ÷ , and the other one is second order marginally stable process
è s2 + s + 1 ø
æ e -0.3s ö
G
ç p ( s ) = ÷ and also the results will be observed for the perturbed processes of
è s 2 + 2s ø
those two processes.
2 2
FSPW FSPW
1/(1.2s 2+1.2s+1.2)
VSPW VSPW
DSPW
1.5 1/(S2+S+1) DSPW
FOPID
1.5 FOPID
Response
Response
1 1
0.5 0.5
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Time(s) Time(s)
FIGURE 24.2 Time-response plot for (a) second-order linear process (b) 20% perturbed model.
TABLE 24.2 Performance Analysis of (a) Second-order linear process (b) 20% Perturbed Model
FSPW (0.7) VSPW DSPW FOPID FSPW (0.7) VSPW DSPW FOPID
OS% 41.31 36.54 61.63 -- OS% 22.71 31.61 55.30 --

TP (S) 2.81 2.65 2.60 2.67 TP (S) 2.83 1.22 2.21 2.51
Tr(S) 1.85 1.48 1.34 1.87 Tr(S) 1.85 1.45 1.3 2.05
Ts(S) 30.20 22.21 10.32 5.52 Ts(S) 18.62 14.31 8.52 5.54
Td(s) 1.39 1.09 0.99 1.01 Td(s) 1.39 1.02 0.98 0.02
IAE 10.39 7.09 4.69 2.84 IAE 7.22 5.05 3.87 2.84
ITAE 231.45 133.62 69.42 31.21 ITAE 145.5 99.24 51.91 27.69
The time response of second-order linear process and its 20% perturbed model are
given in Figures 24.2(a) and 24.2(b), and their performance has been analysed by the
Fractional Order PID controller for Setpoint Tracking and Load Rejection  215
performance indices %OS (percentage overshoot), tp (peak time), tr (rise time), ts (settling
time), IAE (Integral Absolute Error), and ITAE (Integral Time Absolute Error) for
different PID controllers provided in Table 24.1 (a, b). The time response of second-order
marginally stable process and its 20% perturbed model are given in Figure 24.3(a) and
24.3(b) and their performance has been analysed by the aforementioned performance
indices provided in Table 24.2 (a, b). The main tuning parameters λ and µ provide better
results for 0.55 to 0.6 and 0.77 to 0.82, respectively. It has been slightly varied with the
change of process models. From the graphs for the above four models as well as from
the performance table, it is observed that FOPID can provide better response for both
setpoint tracking and load rejection than the other existing weighting techniques.
2 2
FSPW FSPW
1/(S2+2S) VSPW 1/(1.2S2+2.4S) VSPW
DSPW
1.5 DSPW 1.5
FOPID FOPID
Response
Response
1 1
0.5 0.5
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Time(s) Time(s)
FIGURE 24.3 Time-response plot for (a) second-order marginally stable process
(b) 20% perturbed model.
TABLE 24.3 Performance Analysis of (a) Second-Order Marginally Stable Process (b) 20% Perturbed Model
FSPW (0.7) VSPW DSPW FOPID FSPW (0.7) VSPW DSPW FOPID
OS% 31.80 32.90 63.10 -- OS% 20.42 21.33 57.41 --

TP (S) 2.91 2.40 2.31 2.37 TP (S) 2.92 2.52 2.34 2.75
Tr(S) 1.82 1.28 1.18 2.12 Tr(S) 1.22 0.77 0.66 0.71
Ts(S) 5.92 5.48 5.71 2.71 Ts(S) 5.62 5 3.91 2.75
Td(s) 1.44 1.00 0.89 1.37 Td(s) 1.94 1.32 1.19 2.12
IAE 3.34 3.42 2.93 2.10 IAE 2.77 3.15 2.52 2.23
ITAE 44.32 68.35 28.94 22.46 ITAE 32.41 69.04 21.61 20.44
24.4 CONCLUSION
The present work is focussed on the minimization of the overshoot within the satisfactory
limit without compromising the other parameters in setpoint tracking along with load
rejection period. For fulfilling the aforementioned purpose, fractional order PID controller
has been used. The main parameters λ and µ are tuned by trial and error method. Moreover,
it can be observed that for two basic models and their perturbed model and the tuned
values of the parameters are almost same. From the results, it can also be observed that the
responses are improved satisfactorily for the above four models using FOPID controller
than other existing controllers.
REFERENCES
1. Khan, B. Z., and Lehman, B., 1996. Setpoint PI controllers for systems with large normalized
dead-time. IEEE Transactions on Control Systems Technology, 4(4), 459–466.
2. Hang, C. C., Astrom, K. J., and Ho, W. K., 1991. Refinements of Zeigler-Nichols tuning
formula IEE Proceedings D - Control Theory and Applications, 138(2), 111–118.
3. Prashanti, G., and Chidambaram, M., 2000. Setpoint weighted PID controllers for unstable
systems. Journal of the Franklin Institute, 337(2–3), 201–215.
4. Hang, C. C., and Cao, L., 1996. Improvement of transient response by means of variable
Setpoint weighting. IEEE Transactions on Industrial Electronics, 43(4), 477–484.
5. Dey, C., Mudi, R. K., and Lee, T. T., 2010. Dynamic Setpoint weighted PID controller.
Control and Intelligent Systems, 37(4), 212–219.
6. Podlubny, I., 1994. Fractional order system and fractional order controller. Slovak Academy
of Sciences Institute of Experimental Physics, UEF – 03–94.
7. Tajjudin, M., Rahiman, M., Arshad, N. M., and Adnan, R., 2013. Robust Fractional-Order
PI Controller with Ziegler-Nichols Rule. International Journal of Electrical and Computer
Engineering, 7(7), 1034–1041.
8. Pandey, S., and Pandey, R. K., 2013. Performance evaluation of anti-windup FOC with
Setpoint tracking. Proceedings of Students Conference on Engineering and Systems (SCES).
9. Kumar, K., and Smitha, V., 2014. Design of Tuning for Fractional order PIλDμ Controller
using PSO Algorithm. International Journal for Research in Applied Science & Engineering
Technology (IJRASET), 2(12), 436–442.
CHAPTER 25
Analysis and Simulation of PWM-

Based Half-Bridge and Full-
Bridge Switch Mode Amplifier for
Active Magnetic Bearing (AMB)
S. Debnath,1 P. K. Biswas,2 S. Gupta,1 and J. Laldingliana1
Department of Electrical and Electronics Engineering, NIT Mizoram, Aizawl, India.
1,2
sukanta.eee@nitmz.ac.in, pabitra.eee@nitmz.ac.in, suraj.mtech.eee@nitmz.ac.in,

jonathan.phd.eee@nitmz.ac.in
ABSTRACT: Active Magnetic Bearing (AMB) is naturally non-linear and shaky

framework. Current in the magnet coil ought to be kept up legitimately to meet the
alluring power demand. So a power amplifier is expected to control in the closed-loop
design. In this chapter, distinctive sort of power amplifier is described for AMB. The yield
of which is variable DC, whereas input is constant DC. To guarantee the best working
for these power amplifiers in any AMB framework, switching and controlling of the
switches is important, for this Pulse Width Modulation (PWM) system is utilized. PWM
is extremely well known for their dependability, straightforwardness, and performance.
A PWM circuit utilizing 555 timer and operational amplifier are imitated in MULTISIM
and PSIM which is additionally used to control the power amplifier.
Keywords: Power amplifier, MOSFET, Active Magnetic Bearing (AMB), PSIM,

MULTISIM.
25.1 INTRODUCTION
In magnetic bearing, the rotor is hovered by an attractive power with no mechanical contact
and ready to do relative movement either in radial and axial direction. Magnetic bearing
is differentiated into two types: (i) active Magnetic bearing (AMB) which comprises of
all electromagnets (ii) passive magnetic bearing which comprises of electromagnets and
permanent magnets. However, magnetic bearing based on permanent magnets only cannot
217
guarantee for stable contact-less levitation in all degree of freedom (DOF) and furthermore
cannot control rotor vibration actively; so AMB is broadly utilized for active magnetic
bearing framework. In spite of the fact that these frameworks normally have instability
and nonlinearity due to electromagnetic forces and interference. Because of their inherent
instability, these AMB systems are effectively controlled by different classical and modern
controllers and the controller used for this AMB system are itself complex in nature and
designing. The electromagnetic force depends on the current flowing through the coil due
to this power amplifier that is required in AMB setup (Debnath, Biswas, and Laldingliana,
2017). An amplifier receives a signal from some input source and provides a larger version
of the signal to some output device or to another amplifier stage. In cased of the closed-loop
system of an AMB, the input signal provided by the controller is generally small and needs
to be amplified sufficiently to meet the power demand by the system. Thus, various power
amplifiers are proposed for AMB framework, which incorporates solid state and linear
type solid-state power amplifiers and in addition magnetic amplifiers. Switch-mode power
amplifier is utilized for low-power applications. The efficiency of switch-mode DC to DC
amplifier is around 70 to 90 per cent. A MOSFET based power amplifier has considerably
bigger power dealing with limit when works in ON/OFF mode.
25.2 ACTIVE MAGNETIC BEARING

The schematic diagram of single axis active magnetic bearing is displayed in Figure 25.1.
AMB works with no mechanical contact. In this way, the bearing will have low losses
and reduced maintenance with long life span and can be utilized in vacuum and harsh
environments. The speeds of AMB is infinite, it can be boost-up till the ultimate strength
of the rotor. The non-appearance of lubrication increases the working proficiency and
halt contamination of the process. Along these lines, it can be used in numerous fields, for
example-turbo molecular pumps, variable speed spindles, power generating station, high-
speed vehicle system et cetera (Debnath, Biswas, and Laldingliana, 2017; Liu, Chen, and Xu,
2005). The equivalent electrical circuit of AMB is presented in Figure 25.2.
FIGURE 25.1 Single axis AMB. FIGURE 25.2 Electrical equivalent circuit of AMB.
Analysis and Simulation of PWM-Based Half-Bridge and Full-Bridge  219
The magnetic flux of an AMB system is given as (Ghosh and Datta, 2015; Schulz,
Wassermann, and Schneeberger, 2004) [14]
mmf
f= (25.1)
Rtotal
Rt Where, is the total reluctance and is expressed as
Rtotal = R1 + R2 + R3 (25.2)
The reluctance value of air-gap, actuator iron path and rotor iron path are respectively.
Also, the total reluctance value can be expressed as
2L
Rtotal = (25.3)
m0 Agap
L Where, is the length of air-gap, m0 is the permeability of free space and Agap is the area of
the actuator pole. Magnetic flux shown in equation 25.1 can be rewritten as
nI m0 Agap
f= (25.4)
2L
n Here, is the number of turns in the coil and I is the actuating current. The flux density of
the system is shown in equation 25.5.
nI m0
B= (25.5)
2L
The force generates by the actuator is expressed in equation 25.6 which is a function of
flux density.
B 2 Agap
F=
m0
I 2m0 n 2 Agap
F= (25.7)
2 L2
25.3 CLOSED LOOP SYSTEM OF AMB

Directly days AMB is used as a piece of numerous mechanical and modern application [8].
There points of interest lead over their detriments. A fundamental closed loop diagram of
an AMB shows up in Figure 25.3. In this figure, there are two closed loops [1]. The internal
loop circle is for current control reason and the outer loop circle is for position control
reason. In some unique applications, single close circle AMB is used. The piece outline of
this figure can be portrayed as-
1. Actuator - The actuator for an AMB is U shaped, I shaped etc. is indicated which
is made of protected overlay press sheet which works as core and on which copper
winding is drawn. The power of attractive power creates by it relies on the current
flowing through the winding.
2. Rotor - A sphere-shaped rotor utilized as a part of this paper, which is equipped to
bear the high stress due to the centrifugal and centripetal force generated during
the rotation.
3. Controllers - AMB is exceedingly unstable and nonlinear system. Along these
lines, controllers are required to influence them to work. In this paper, one current
controller and one position controllers are used. Controller can be classical or,
modern or both.
4. Sensors - The Current sensor is utilized for detecting the current in the magnet
loop and the position sensor is utilized to detect the hole between the actuator and
rotor ball. Sensors can be of any sort like – IR sensor, a laser sensor, optical sensor,
capacitive sensor, Hall impact sensor and soon.
5. Power Amplifier - as its name infers it amplifies the current which is heading
off to the magnet loop. The output from the current controller is input to it and
the control signal coming from the current controller is amplified and fed to the
magnet coil to maintain the required magnetic force and field.
FIGURE 25.3 Closed loop system of AMB.
There are different kinds of a power enhancer is accessible like–single switch, half-bridge,
full-bridge etc. (Ghosh and Datta, 2015; Bhimra, 2004; Schweitzer, 2009). Here a detail
study is done on half-bridge and full-bridge power amplifier.
25.4 POWER AMPLIFIER

Current sensor output is compared with the reference current and this is fed to the current
controller. The output of the current controller is then going to the power amplifier
which essentially increases the present which will also give magnet coil to keep up the
accommodating appealing force for levitation reason.
DC to DC converters for power amplification reason like-buck converter can’t be

used in AMB because the output voltage of buck converter is just having positive esteems.
For levitating an object, the voltage supply across the AMB should be ought to be in both
positive and additionally negative esteem. Power amplifiers for AMB system are-
Half-bridge switch mode power amplifier
Half-bridge switch mode control enhancer comprises of two switches, two capacitors
and magnet loop as a load (Schulz, Wassermann, and Schneeberger, 2004; Schulz,
Schneeberger, and Wassermann, 2004).
It is supplied by a dc voltage source as appeared in Figure 25.4. In perspective of two
switches, its efficiency is improved when appeared differently in relation to single switch
amplifier, however, the cost gets extended.
FIGURE 25.4 Half bridge amplifier. FIGURE 25.5 Full bridge amplifier.
25.4.1 Full bridge switch mode power amplifier

It comprises of four switches, protection diode, capacitor and magnet coil as a load. In
view of four switches viability get improved yet the cost and multifaceted nature of this
power intensifier are high. The circuit diagram of full bridge amplifier is appeared in
Figure 25.5.
25.5 PULSE WITH MODULATION

Pulse width modulation works on modulation technique in which a carrier signal is
modulated over reference signal and pulses are generated (Guo, Ling, and Zhang, 2016;
Lam et al., 2013; Schulz, Wassermann, and Schneeberger, 2004). The generated pulses
are further feed to switches to control and operate them. Using PWM some of the lower
order harmonics can be easily eliminated and higher order harmonics can be easily filtered,
through which output gets improved.
PWM control strategy is broadly utilized as a part of mechanical and household
application in light of their lower cost, better execution and straightforwardness in
controllability.
FIGURE 25.6 Pulse with modulation output waveform.
A triangular signal is utilized as carrier signal and sinusoidal signal is used as a reference
signal when modulated pulses of different width is obtained in each half cycle. The carried
and reference signals are mixed in comparator for modulation when carrier signal has
magnitude higher than reference, comparator output is high and output pulses are obtained
so by varying the frequency of the carrier and reference number of pulse in each half cycle
can be varied shown in Figure 25.6.
25.6 SIMULATION RESULTS

In this investigation work, a power amplifier is intended for U-type actuator alongside
a circular kind rotor. The pickup current and the pickup voltage for various air-holes
(2 mm to 20 mm) are shown up in Figure 25.7, here, most extreme current is 5A and greatest
voltage is 8V. Accordingly, for the proposed AMB framework, control enhancer should
deal with up to greatest get voltage and current. Reproduction is complete utilizing
MULTISIM and PSIM. PWM circuit is planned using 555 clock IC and an operational
amplifier IC. The simulation circuit of PWM using MULTISIM base displaying
condition appears in Figure 25.8, Figure 25.9 demonstrates the output pulse for various
obligation cycle.
FIGURE 25.7 Characteristic of current and voltage for airgap 2 mm to 20 mm.

FIGURE 25.8 555 timer based PWM circuit.
The pulse creates by PWM is utilized for driving the MOSFET for both the half-bridge and
full-bridge amplifiers.
Half-bridge amplifier is composed and simulated in PSIM software and the MOSFET
is gated with pulse generator as displayed in Figure 25.9 and the waveform half-bridge
amplifier is appeared in Figure 25.10.
FIGURE 25.9 Half-bridge amplifier FIGURE 25.10 Output waveform for half-bridge
without PWM circuit. amplifier.
FIGURE 25.11 Full-bridge amplifier without PWM circuit.

FIGURE 25.12 Output waveform for full switch amplifier.
Half-bridge amplifier with PWM is composed and simulated in PSIM software and the
MOSFET is gated with PWM as displayed in Figure 25.13 and the waveform half-bridge
amplifier is appeared in Figure 25.14.
FIGURE 25.13 Half-bridge amplifier FIGURE 25.14 Output waveform for half-bridge
with PWM circuit. amplifier with PWM.
The simulation diagram of full-bridge alongside the output waveform is given in

Figure 25.15 and 25.16.
FIGURE 25.15 Full-bridge amplifier FIGURE 25.16 Output waveform for full switch
with PWM circuit. amplifier.
For both of the cases, positive and negative output voltage is observed which is required for
an AMB framework to hover a rotor in the magnetic field.
25.7 CONCLUSION
In this work Power amplifier is intended for the purposed U type single-axis AMB framework,
the pickup voltage and current for different air gaps is shown in this examination work,
likewise rely upon the maximum pickup voltage and current two sorts of power amplifier
alongside pulse width modulation circuit is composed and imitated and the yield voltage
waveform is seen from the recreation comes about. It is demonstrated that both the half-
bridge switch-mode power amplifier and full-bridge switch-mode power amplifier gives
positive and negative voltage at the yield, which is important for drifting the rotor in the
space. utilizing this examination result it is presumed that half- bridge switch-mode power
amplifier requires lesss witching devices and consequently cost is less when contrasted with
the full-bridge switch mode power amplifier.
ACKNOWLEDGEMENT
The authors of this paper wish to thank Science and Engineering Research Board (SERB),
Department of Science and Technology, Govt. of India for sponsoring the project.
REFERENCES
1. Debnath, S., Biswas, P. K., and Laldingliana, J., Analysis and simulation of pwm based power
amplifier for single axis active magnetic bearing (amb), In Transportation Electrification
Conference (ITEC-India), 13–15 Dec., 2017, Pune, India. IEEE, december 2017.
2. Liu, S., Chen, D., and Xu, F., 2004. Study of switching power amplifier for active magnetic
bearing, In 4th International Power Electronics and Motion Control Conference, 14–16
Aug., 2004, Xi’an, China. IEEE, may 2005.
3. Sun, P., Li, L., and Zhang, C., 2014. Effects of displacement sensor noise on power amplifiers
of active magnetic bearings, In 13th International Conference on Control Automation
Robotics & Vision (ICARCV), 10–12 Dec., 2014, Singapore, Singapore. IEEE, March 2015.
4. Gong, J., Zhang, G., Zhang, J., Wu, H., and Cheng, X., 2013. Design of digital switching
power amplifier for magnetic suspended bearing, In Proceedings of 2013 Chinese Intelligent
Automation Conference, China, 255. Berlin, Heidelberg: Springer, July 2013, pp. 733–739.
5. Guo, Y., Ling, Z., and Zhang, X. A novel pwm power amplifier of magnetic suspension spindle
control system for micro edm, In The International Journal of Advanced Manufacturing
Technology, 83. London: Springer, March 2016, pp. 961–973.
6. Ghosh, S. P., and Datta, L.. Electromagnetic field theory. India: McGraw Hill education
private limited, 2015.
7. Bhimra, P. S., 2004. Power electronics. NaiSarak, Delhi: RomesChander Khanna,
pp. 349–314.
8. Schweitzer, G., 2009. Magnetic bearing theory design and application to rotating
machinery, E. H. Maslen, Ed. Berlin: Springer, January 2009, pp. 69–107.
9. Saith, S. P., Radhakrishnan, K., and Krishnakumar, P., 2013. Switch mode power amplifier
with feedback control for electro acoustic projectors, International Journal of Engineering
and Innovative Technology (IJEIT), 3(3), pp. 330–336.
10. Parveg, D. R., 2008. A study of different switched mode power amplifiers for the burst mode
operation, September 2008.
11. Lam, C. K., Tan, M. T., Cox, S. M., and Yeo, K. S., 2013. Class d amplifier power stage
with pwm feedback loop, IEEE Transactions on Power Electronics, 28(3), pp., 3870–3881.
[Online]. doi: 10.1109/TPEL.2012.2230027
12. Zhang, J. and Karrer, N., 1995. IGBT power amplifiers for active magnetic bearings of high
speed milling spindles, In 21st International Conference on Industrial Electronics, Control,
and Instrumentation, 6–10 Nov., 1995. Orlando, FL, USA: IEEE, August 2002.
13. Carabelli, S., Maddaleno, F., and Muzzarelli, M. High-efficiency linear power amplifier for
active magnetic bearings, IEEE Transactions on Industrial Electronics, 47(1), pp. 17–24,
2000. [Online]. doi: 10.1109/41.824021
14. Schulz, A., Wassermann, J., and Schneeberger, M., 2004, A reliable switching amplifier for
active magnetic bearings, In IEEE International Conference on Industrial Technology, 10–
12 Dec., 2003, Maribor, Slovenia, Slovenia. IEEE, May 2004.
15. Schulz, A., Schneeberger, M., and Wassermann, J., 2004. A reliable switching amplifier
for active magnetic bearings error detection strategies and measurement results, In IEEE
International Conference on Industrial Technology, 8–10 Dec., 2004, Hammamet, Tunisia,
Tunisia. IEEE, August 2005.
CHAPTER 26
Design and Analysis of High-

Speed Phase Locked Loop in
180 nm Technology
A R Abdul Rajak,1 Azeem Gafoor Mohammed,2 and Deep Sehgal3
Dept. of Electrical and Electronics Engineering, Birla Institute of Technology and Science Pilani, Dubai
1, 2
Campus, Dubai, UAE.

3
Semiconductor Laboratory-ISRO, Mohali, Punjab, India.
1
abdulrazak@dubai.bits-pilani.ac.in, 2azeemmuhammad34@gmail.com, 3sehgal@scl.gov.in
ABSTRACT: The phase locked loop (PLL) is a control system that compares the phase
of the output signal with that of the input signal and proportionally varies its output
based on the phase difference between them. PLL most prolifically finds its applications
in frequency generation, clock and data recovery, carrier frequency generation in
modulation and demodulation, frequency synthesizers, jitter and skew reduction, and
many more. Mixed-signal and digital circuits employ PLL for clock generation to get
the synchronous transactions with the clock. So it is mandatory to properly design the
PLL to obtain minimum jitter on the clock edges. The PLL designed in this work was
based on the famous charge-pump (CP) architecture and was implemented using the
semiconductor laboratory (SCL) 180 nm technology in Cadence Virtuoso. This work
uses NOR-based PFD using TSPC latches, Transmission gate CP, Current Starved VCO,
and TSPC-based divider circuit. The proposed model of PLL has phase noise of the PLL
as –113 dBc/Hz, the power consumption of the total PLL as 3.2 mW at 1.8 V power
supply with deterministic jitter of about 5.62 ps and the settling time of the control
voltage as 850 ns.
Keywords: Charge pump, Jitter, Phase frequency detector, Phase noise, Phase-locked
loop.
227
26.1 INTRODUCTION
In the past five decades, development in IC Fabrication technology along with
advancement in the scaling of the devices has led to unprecedented growth in the
field of the semi-conductor. MOSFET scaling has revolutionized the rate at which the
circuits operate. These types of MOSFET leads to operating at higher data rates and
increased computing capabilities. It gives rise to clock frequencies hitting the Giga-
Hertz (GHz) range and data rates reaching multi-Gbps range in the daily life devices
such as mobiles, laptops, PDAs. The most important driving factor that is pushing the
boundaries of IC technology to access the information quickly and robustly in terms
of lower power dissipation, size, and ease of portability. This increase in speed is of no
value at all if the designed system is not robust. For example, today’s communication
transceivers require a bit error rate (BER) of 10–12, which is one error bit in every
1012 bits. It is necessary to design the system carefully to achieve such high accuracy
(Abdul Rajak et al. 2013). Hence, there is a greater need for low power and highly
efficient system on chips (SoCs) which should also have high integrability to be used
along with other communication interfaces such as serializer/deserializer, memory
operations, processor, interfacing devices, and many more. Since the demand for these
high-frequency devices has increased drastically, higher clock frequencies are needed
for faster read/write operations, thus making it more explicit for the use of phase-
locked loops for on-chip frequency generation.
Figure 26.1 represents the advances in the scaling of data rates in high-speed I/O
links published in the annual semi-conductor road-mapping report of 2011 (Friedman
et al., 2013; Alon, 2011) by International Solid-State Circuits Conference (ISSCC). It
stated an essential point that the data-rates are increasing by a scale of 2× for every four
years, but the bandwidth of the channel remains the same. While the designing of the
PLL (Levatino et al., 2015; Venerus et al., 2015) seems an easy task at low frequencies, it
is a bit more of a daunting task to design a robust PLL with low jitter at high frequencies
due to many non-idealities such as dead-zone, ringing, cross-talk, reflections involving
the circuits. This work discusses the underlying architecture of the famous Charge
Pump, CP-PLL, various types of PLL, the mathematics involved in an integer-N PLL
which mainly has
Phase frequency detector (PFD)
charge pump (CP)
loop filter/low pass filter (LPF)
voltage controlled oscillator (VCO)
divider (DIV)
Design and Analysis of High-Speed Phase Locked Loop in 180 nm Technology  229
FIGURE 26.1 Data-rate trends in signalling links (Ratan, 2014).
26.2 MATHEMATICAL MODEL FOR PLL

Figure 26.2 shows the basic architecture of the charge pump (CP), switches UP and DN.
I
Pulse width goes UP by ÑT . 1 when UP switches close. Similarly, a pulse on DN yields
C1
a drop in VOUT. If UP and DN are inserted simultaneously, it created no change in output.
FIGUR 26.2 Block diagram of integer-N PLL (Ratan 2014).
For the linear model analysis of the CP-PLL (Ratan, 2014), the s-domain model is shown in
Figure 26.3 below. The open-loop transfer function of the PLL is defined in Venerus et al.
(2015) and Ahidi (2006) as follows:
FIGURE 26.3 CP-PLL s-domain model (Ratan, 2014).
1
s+
KVCO RC1
LG ( s ) = K PD F ( s ) = K PD KVCO (26.1)
s c +c
c2 s 2 ( s + 1 2 )
Rc1c2
1 c +c
From the above expression, it concludes that w z = ; w p1 = 0; w p 3 = 1 2 (26.2)
Rc1 Rc1c2
The phase margin can be obtained from expression as
æ w0 ö æ w0 ö
Æ M = arctan ç ÷ - arctan çç ÷÷ (26.3)
è wZ ø è w p3 ø
where ω0 is the unity gain bandwidth of the open loop and ωz< ω0
The value of ωz and ωp3 depends upon C1 and C2, so as to achieve maximum value of
the phase margin, find out its derivative with respect to ω0, and equate it to zero, hence,
c1
wo = w z 1 + (26.4)
c2
Thus, maximum phase margin is obtained by substituting in the equation:
æ ö
ç ÷
æ c1 ö ç 1 ÷
Æ M _ max = arctan çç 1 + ÷÷ - arctan (26.5)
c2 ø ç c ÷
è çç 1 + 1 ÷÷
è c2 ø
The guidelines for the design loop filter are as follows:

According to the specification provided, select the desired ω0, ΦM, R. After choosing
these values, calculate Mc from Eq. (26.6).
Mc =
c1
c2 ( (
= 2 tan 2 ( Æ M ) + tan Æ M tan 2 ( Æ M ) + 1 )) (26.6)
From Eq. (26.6), it leads to

wo 1 C
wz = c1 = ; C2 = 1 (26.7)
c wz R MC
1+ 1
c2
Then calculate the value of ICP

2 2
2p C2 2 w0 + w p 3 (26.8)
I CP = wo
K vco w02 + wZ2
Each block of the PLL contributes to some noise in the PLL. To model these noises,
the following noise transfer functions were used
Æ out ( s ) N .LG ( s )
NTFIN ( s ) = = (26.9)
Æ in ( s ) 1 + LG ( s )
NTFDIV ( s ) = NTFIN ( s ) (26.10)
Æ out ( s ) 2p
NTFCP ( s ) = = NTFIN ( s ) (26.11)
I CP ( s ) I CP
KVCO
Æ (s) s
NTFR ( s ) = out = (26.12)
VCTRL ( s ) 1 + LG ( s )
2
SÆÆOUT
IN
= S I IN NTFIN ( s ) (26.13)
2
SÆICPOUT = S ICP NTFCP ( s ) (26.14)
2
SÆVCTRL
OUT
= SÆ IN NTFR ( s ) (26.15)
2
SÆVVCO
OUT
= SÆ VCO NTFVCO ( s ) (26.16)
26.3 TRANSISTOR LEVEL IMPLEMENTATION AND RESULTS

In this section, the transistor-level implementation of the CP-PLL has been explained.
Cadence Virtuoso Analog Design Environment (ADE) is used to implement the design of
the PLL in SCL’s 180 nm technology. Spectre simulator of cadence is used to simulate and
analyse various performance parameters of the PLL. Different analyses, like the transient
simulation, dc sweep, p noise, and pss simulation, are required in this design.
26.3.1 Phase Frequency Detector (PFD)

Conventional NAND-based PFD as described in Gines et al. (2014) and Jianzheng et al.
(2008) suffers from critical drawbacks such as dead zone, higher power consumption,
larger area, and NAND-based PFD is used in almost all the PLLs today. A new type of
PFD using TSPC based design reduces all the drawbacks stated. This technique makes use
of a total of 24 transistors for the entire design of the PFD, which is useful in terms of
area. Also, it makes use of Q’ output of the D Flip-Flop and also employs a NOR gate in
the feedback path. In conventional NAND-based PFD, the D-input of the Flip-Flops is to
connect to VDD, as shown in Gines et al. (2014). In this circuit, the D-input of both the
Flip-Flops connects with the output of NOR gate in the feedback path. Figure 26.6 provides
a schematic of the design.
FIGURE 26.4 Design of PFD.
Since TSPC-based design forms a ratioed inverter, the sizing of the devices is decided based
on the beta ratio given:
bn
=4 (17)
bp
Figure 26.7 shows the output waveforms of the implemented PFD block:
FIGURE 26.5 Output of PFD.
26.4 CHARGE PUMP (CP)

To achieve this, employ a self-biased cascode current mirrors, as shown in Figure 26.8.
The loop filter is designed using the equations given below [28]
FIGURE 26.6 Transmission gate charge pump.

The architecture proposed in Behazad (1996), Behazad (2002), Floyd (1999), and Arakali et
al. (2010) employs an op-amp in their charge pump to maintain the constant output current
as that of input. The technique ‘bootstrapping’ is used to reduce the mismatches between
the UP and DN currents and to reduce the charge sharing problem. But designing the op-
amp itself increases the area, power consumption, and complexity, so avoid the use of the
op-amp in this topology. Therefore, the careful design of the CP is necessary to ensure that
there is no mismatch between the UP and DN currents. To reduce the current mismatch,
use a single reference current and derive both UP and DN, currents. As stated earlier, both
UP and DN currents are extracted from the same reference current source IREF that leads
to the reduction of the mismatches between the UP and DN currents. Moreover, many non-
ideal effects of the op-amp like offset voltage, PSRR, CMRR, etc., might hamper the proper
operation of the CP. The CP is responsible for converting the output of the PFD, that is
voltage signals into current and injects it into the loop filter. The widths of the transmission
gates are adjusted depending upon the beta ratio obtained. T1–3 are introduced into the
circuit to eradicate non-ideal effects introduced by the insertion of TUP and TDN in the
circuit. Whenever switching occurs in UP, UPb, and DN, DNb, the glitches at these nodes
will not pass to the out node because of M6 and M9 act as shields. Proper selection of W/L
ratios of the transistors M5 and M7 is very much essential because these transistors are
responsible for the replication of current in the output node. The control voltage at the
output of the CP is shown in Figure 26.9.
FIGURE 26.7 Simulation results of PLL using TG CP.

26.5 LOOP FILTER DESIGN (LF)

The loop filter is designed using following equation (Wolaver 2015)
1
2.5 wn = win (26.17)
10
R1 I P C1 KVCO
z = (26.18)
2 2p N
R1 I P KVCO
wn = (26.19)
2 2p C1 N
where ωin = input frequency, N = divider circuit, and KVCO = Gain of the VCO. From
Eqs (26.17)–(26.19), IP = 100 uA, z = 1 and obtained that R1 = 6.5 MΩ, C1 = 12.2 pF,
C2 = 0.2 × C1 = 2.44 Pf.
26.6 VOLTAGE-CONTROLLED OSCILLATOR (VCO)

The oscillator is a circuit which generates periodic output waveform without taking any
input. This work uses a ring oscillator. A voltage-controlled oscillator is a device which
generates the periodic output depending on the input control voltage. The CSVCO offers
an extensive range of advantages such as wide tuning range, low-power consumption, and
linear characteristics between the inputs and the output. Three stage VCO is designed at a
supply voltage of 1.8V with the centre frequency of 1.2 GHz. The purity of the output signal
obtained depends on designing VCO.
The following steps describe the methodology for the design of CSVCO:
Step 1
Calculate the delay through each inverter stage in the VCO by using the equation
below:
1
tp = (26.20)
Nf
1
where t p = ´ (Propagation Delay of each inverter stage) (26.21)
2
N = number of stages of inverter

Step 2
Calculate the aspect ratios of MOSFET’s in the inverter stage by using Eqs (26.22) to (22.23):
æ 2VT ,n æ 4(Vdd - VT ,n ) ö ö
(W L ) n
=
Cload
çç
t plh mn Cox (Vdd - VT ,n ) è Vdd - VT ,n
+ ln ç
è Vdd
- 1÷ ÷
øø
÷
(26.22)
æ 2VT , p æ 4(Vdd - VT , p ) ö ö
(W L ) =
Cload
( )
çç + ln ç - 1÷ ÷
ø ø÷
(26.23)
t plh m p Cox Vdd - VT , p è Vdd - VT , p Vdd
p
è
Step 3
Now, calculate the aspect ratios of the MOSFETs acting as current sour.
Step 4
Calculate the value of total capacitance at the output node of each inverter stage using
Ctot = 2.5 Cox (WP LP + WN LN ) (26.24)
where Cox = oxide capacitance of transistors,
WP , LP , WN , LN , are the respective widths and lengths of the PMOS and NMOS
transistors in the inverter stages.
Step 5
Subsequently, calculate the corresponding drain current at the required centre frequency
by using Eq. (26.25).
I DCENTRE = NCtotVDD f osc (26.25)
where fOSC = centre frequency.
Step 6
After calculating IDCENTRE, determine the aspect ratios of the transistors acting as current
sources by using the equation:
(W L ) n
=
2 XI Dcenter
mn Cox ( (Vgs - VT ,n ) 2 )
(26.26)
(W L ) p
= 2.5 X W ( L) n
(26.27)
The gain of the VCO is given by

f max - f min
KVCO = (26.28)
Vmax - Vmin
Figure 26.10 represents the obtained specifications of the VCO and its block diagram.
VCO Architecture: Three Stage Current Starved Ring VCO
Frequency range: 140 MHz to 1.7 GHz
Central frequency: 1.2 MHz
The gain of the VCO: 865 MHz/V

IDCENTRE: 120 μA.
FIGURE 26.8 CSVCO design.
Table 26.1 shows the variation of frequency as a function of the control voltage.
TABLE 26.1 Input Control Voltage vs. Output Frequency of VCO
Control Voltage (V) Frequency (MHz)
0.4 140
0.5 230
0.6 410
0.7 580
0.8 723
0.9 900
1.0 1100
1.1 1114
1.2 1194
1.3 1325
1.4 1430
1.5 1591
1.6 1591
1.7 1671
1.8 1712
FIGURE 26.9 VCO characteristics.
26.7 DIVIDER (DIV)

The divider circuit is used to divide the input frequency by a certain amount. The entire
divider circuit comprises of three D Flip-Flops to realize a divider with eight circuits as
shown in Figure 26.13. The D Flip-Flop was implemented using the TSPC architecture
as discussed in section A. Again, careful sizing of the MOS transistors should be done as
n
the input to the divider circuit is of very high-frequency signal. Q of the D Flip-Flop is
connected to the D input as feedback and Q is fed to the clock input of the next D Flip-
Flop to ensure the left shift operation which is equivalent to division in binary logic.
Figure 26.14 represents the schematic of the TSPC D Flip-Flop.
For all the digital applications, it is mandatory to measure the jitter, that is, the spectral
purity of the clock. Clock generator in digital applications uses PLL. Designing an analogue
PLL with the pure clock is possible, but the loading effect on the clock is what introduces
the jitter in the clock waveform. So the figure of merit of the PLL is to determine the jitter
to reduce the clock timing errors during the read/write operations. The deterministic and
the random jitter in PLL are calculated using the eye-diagram obtained from the calculator
which is available in the Cadence Virtuoso software. The deterministic jitter of the PLL
obtained is 5.62 ps. This value is a reasonable range for the PLL designed.
TABLE 26.2 All Parameters
Parameter Value
Input frequency 200MHz
Output frequency 1.2GHZ
Divider 8
Settling time 850ns
Phase noise –113.31dBc/Hz@ 1MHz offset
ICP 100uA
R1 6.5M
C1 12.2pF
C2 2.44pF
VDD 1.8V
26.8 CONCLUSION
A fast-locking phase-locked loop (PLL) was designed with a lock frequency of 1.2 GHz,
phase noise of 113.31dBc/Hz @ 1 MHz offset. The deterministic jitter obtained was 5.62 ps
and edge-to-edge random jitter of 5.27 ps. In this work, an integer-N PLL was designed, and
reasonable phase noise and jitter characteristics were obtained. A low noise LC-tank VCO
at the cost of higher power consumption with a larger area will improve their values still. It
is the dependency upon the application that chooses the topology of the circuit. Table 26.3
shows a comparison of work done here with some of the other significant works.
TABLE 26.3 Comparision of This Work with Other Works
(Arakali et (Janadhan (yutaka et (Liangge et (Huang et
(Gines et
al. 2010) * et al.2006) * al 2008) * al. 2009)* al. 2014)* This Work
al. 2014)
CICC’09 ISSCC’14 APCCAS’08 CICC’09 ISSCC’14
Power supply
1.2 1.1 1.2 1/1.2 1.2 1.4/1.8 1.8
in V
Output in GHz 2.29– 2.92 2.4 2.4 3.5 1.6–1.9 1.7 – 2.5 1.2 –1.6
Power 274.346
3.2 mW @
Consumption 12 6.4 5.1 21 2.4–2.6 μW @
1.2GHz
mW 2GHz
Settling time 30μs n/a 48.7μs n/a n/a 14.1 μs 850ns
Jitter in ps n/a 3.29 n/a n/a 0.5 35.26 5.62
UMC SCL
Technology 65nm 40nm 90nm 65nm 65nm
180nm 180nm
ACKNOWLEDGEMENTS
Semiconductor Laboratory, Mohali, Punjab, completely supported this work. The authors
would like to thank Deep Sehgal, Head, VLSI Design Division and M. Bharath Reddy,
Scientist ‘SF’, SCL, for their constant support, guidance, and encouragement by providing
the necessary stimuli and environment for accomplishing the work.
REFERENCES
1. Abdul Rajak, A. R., et al., 2013. Efficient way of identification of mobile location, Springer
Lecture series in Electrical Engineering, 1 July 2013.
2. Abidi, Phase noise and jitter in CMOS ring oscillators, IEEE Journal Solid-State Circuits,
41(8), pp., 1803–1816, Aug., 2006.
3. Alon, E., High-speed electrical interface circuit design: Lecture11, 2011. Available online:
http://bwrcs.eecs.berkeley.edu/Classes/icdesign/ee290c_s11/lectures/Lecture11_PLL_
DLL_Components_2up.pdf
4. Arakali, Gondi S., and Hanumolu P. K., 2010. Analysis and Design Techniques for Supply-
Noise Mitigation in Phase-Locked Loops, IEEE Transactions on Circuits and Systems—I:
Regular Papers, 57(11), Nov., 2010.
5. Razavi, Behzad, 2002. Design of Analog CMOS Integrated Circuits, Tata-McGraw Hill
2002, Ch. 5, pp. 135–165.
6. Razavi, Behzad, 1996. Monolithic Phase-Locked Loops and Clock Recovery Circuits,
Chapter 1. Piscataway, NJ: IEEE Press, 1996.
7. Wolaver, Dan H. Phase-Locked Loop Circuit Design, Prentice Hall, Ch. 4, pp. 47–80, 2015
8. Gardner, Floyd M. Charge Pump Phase Lock Loops, IEEE Transactions on Communication,
28(1), pp., 1849–1858, Nov., 1999.
9. Friedman, 2013. International solid-state circuits conference trends 2013, Available online:
http://isscc.org/doc/2013/2013 Trends.pdf2008. APCCAS 2008, pp., 1252–1255, Nov. 30
2008-Dec. 3 2008
10. Gines, A. J., Doldan, R., Villegas, A., Acosta, A. J., Jalon, M. A., Vazquez, D., Rueda, A.,
and Peralias, E., 2010. A 1.2V 5.14mW quadrature frequency synthesizer in 90nm CMOS
technology for 2.4GHz ZigBee applications, IEEE Asia-Pacific Conference on Circuits and
Systems.
11. Huang, Y. C., Liang C. F., Huang H. S., and Wang P. Y., 2014. A 2.4 GHz ADPLL with digital-
regulated supply-noise- insensitive and temperature self- compensated ring DCO, IEEE
ISSCC Digest Technical Papers, 2014, pp. 270–272.
12. Janardhan, H., and Wagdy M. F., 2006. Design of a 1GHz Digital PLL Using 0.18 μm CMOS
Technology, IEEE Proc. of the Third International Conference on Information Technology.
13. Zhou, Jianzheng, Wang, Zhigan, 2008. High-Performance CMOS Charge Pump for Phase-
Locked Loops, International Conference on Microwave and Millimetre Wave Technology
(ICMMT), 2, pp. 839–842.
14. Levantino, S., Marucci G., Marzin, G., Fenaroli, A., Samori C., and Lacaita A. L., A 1.7 GHz
Fractional-N Frequency Synthesizer Based on a Multiplying Delay- Locked Loop, IEEE
Journal of Solid-State Circuits, 50(11), 2678–2691, Nov., 2015.
15. Xu, Liangge, Lindfors, S., Stadius, K., and Ryynanen, J., 2009. A 2.4-GHz low-power all-
digital phase-locked loop, IEEE Custom Integrated Circuits Conference, 2009. CICC ‘09,
pp. 331–334, 13–16 Sept., 2009.
16. Ratan, Rishi. http://emlab.uiuc.edu/jose/Theses/Ratan.pdf, Master thesis, Available
online: https://www.semanticscholar.org/author/Rishi-Ratan/120340333, https://
pdfs.semanticscholar.org/0751/aa84a51eef6098332594cab8db3348a312e6.pdf ?_
ga=2.217829205.274856050.1588768083-1458125171.1588768083
17. Kaipu, Sreedhar Vineel R., Vaish, Kriti, Komati reddy, Sneha, Sood, Akshat, Goswami,
Manish, 2016. Design of a Low Power Wide Range Phase Locked Loop using 180nm CMOS
Technology, 2016 IEEE International Conference on Signal Processing and Communication,
26 December 2016, JIIT-Noida.
18. Venerus, C., and Galton, I., 2015. A TDC-Free Mostly-Digital FDC-PLL Frequency Synthesizer
with a 2.8–3.5 GHz DCO, IEEE Journal of Solid-State Circuits, 50(2), pp. 450–463.
19. Nakanishi, Yutaka, Kobayashi, Fuminori, Kondoh, Hitoshi. Low-jitter PLL by interpolate
compensation, APCCAS 2008, 1078–108.
CHAPTER 27
Design of APOD-PWM Based

Multi-level Inverter using Cuk
DC–DC Converter with MPPT
Rajarshi Dhar1 and Sarbojit Mukherjee2
1
Dept. of Electronics and Communication Engineering, RCCIIT, Kolkata, India.
2
Dept. of Electrical Engineering, RCCIIT, Kolkata, India.
1
rajarshidhar1997@gmail.com, 2sarbo.1234@gmail.com
ABSTRACT: This chapter focuses on the generation of AC power from renewable

sources which in our case happens to be a solar photovoltaic cell or solar PV cell. The
power from the solar PV panel is fed into a Cuk DC-DC converter whose function is to
boost the solar output to a certain level as well as hold the voltage steady at that level so
that the next stage, that is the inverter stage, can have a fixed input. In this chapter, we
have discussed the AC generation using a multi-level inverter (MLI) with novel topology
whose gate pulses are controlled by a method called the asymmetric phase opposition
disposition pulse width modulation (APOD-PWM). Our work in this chapter also
focuses on the hard and soft switching problems in a converter. It is taken care that the
circuit which we have tried to implement in this work are soft switching in nature and
thus the stress on the switches is greatly reduced. Moreover, the project does not involve
any transformers for the step-up and step-down purposes and thus the leakage occurring
due to them are also reduced. The whole circuit is simulated in MATLAB 2017a and the
results are given henceforth.
Keywords: APOD-PWM, MLI, MPPT, Step up DC-DC converter, Solar PV.
27.1 INTRODUCTION
In recent years, concern towards green energy and the energy demand as well has been
increased significantly. Moreover, the demand for power is increasing day by day but
there is unavailability of resources to meet the power demands. Therefore, extensive e
241
researches are going on to shift our focus to renewable forms of energy as the primary
energy source for the planet. The solar power is the most common and easily usable
and accessible, and thus many works are done on this domain to extract the maximum
power from the sun. Solar PV panels which are a series-parallel combination of many
solar cells are used for the extraction of power from the sun. Now the output of a solar
cell is not always constant; it depends on the temperature and the irradiance of the
sun which are continuously fluctuating in nature. Thus, for a particular combination
of temperature and irradiance, there is maximum power output. Different algorithms
are implemented for extracting this maximum power at that combination, out of which
the maximum power point tracking or the MPPT algorithm is the most effective. Many
works on this algorithm have been done over the years, some of which are presented in
literatures (Safari and Mekhilef, 2011; Kachhiya, Patel, and Lokhande, 2011; Femia et
al., 2010; Hohm and Ropp, 2003; Tse et al., 2002; Balata and Vitelli, 2013; Raveendhra
and Thakur, 2013). These studies mainly concern themselves with different maximum
power point tracking algorithms to make the MPPT more efficient. In this study, we
have discussed the P&O MPPT algorithm which is the most commonly used. The output
of the algorithm is to output a continuous PWM wave whose duty cycle varies with the
power extracted from the PV cell at that instant. The next stage of our project deals with
the DC-DC conversion. As discussed earlier, this part of the circuit deals with two things
mainly, the first one being the boosting of the PV panel output and second, making the
coupling capacitor as a constant voltage source for the inverter stage. There are many
types of normal converters and also many special types of converters which are specially
designed for solar purposes or hybrid power purposes, which have been discussed in the
following literatures (Jha and Singh, 2006; Singh and Bist, 2015; Hwu and Peng, 2012;
Wu et al., 2016; Murthy-Bellur and Kazimierczuk, 2011; Khatab, Marei, and Elhelw, 2018;
Huang and Shi, 2014; Chen and Fayed, 2015; Raghamadhuri and Subramanian, 2017;
Tidke et al., 2018; Ranjana, Reddy, and Kumar, 2014; Kaouane, Boukhelifa, and Cheriti,
2015). The next and the final stage of our project is the inverter part whose function is
to convert DC power to AC power. There are numerous types of inverters starting from
the basic full H-bridge inverters. Research is done extensively so as to get the output of
the inverter to as nearly sinusoidal as possible and the most common method for doing
that is the multi-level inverters or MLIs. There are numerous methods for MLI operation
which are presented in Lu et al. (2009), Chithra and Dasan (2011), Marquez et al. (2014),
Manoharan et al. (2015), Patil and Prasad (2015), and Xiao et al. (2015), but the one
which we have implemented here is a novel kind of MLI called the APOD-PWM MLI
which will be discussed further into the chapter.
27.2 SYSTEM DESIGN AND ANALYSIS

The system that we have presented in this chapter consists of a solar panel whose output
generates a pulse which controls the functioning of the next part of the circuit that is the
DC-DC converter and finally the converter drives the MLI at the end.
Design of APOD-PWM Based Multi-level Inverter using Cuk DC–DC Converter  243
FIGURE 27.1 DC-DC circuit diagram.
FIGURE 27.2 Inverter schematic.

As it can be seen from Figure 27.1, that the output of the solar PV panel is fed into a
controller, which produces a pulse that further runs the switch for the dc-dc converter
circuit. The output of the converter is then fed into the MLI which is shown in Figure 27.2,
and the final output is obtained at the filtering end which gives a modified sinusoidal wave.
Each section of this whole system is described below.
27.2.1 Tracking of MPPT

To improve the efficiency of the system and the power extraction from the PV-panel, the
maximum power point tracking algorithm is being used as shown in Figure 27.3. Among
the different methods that can be used, perturb and observe algorithm is being implemented
here to take care of reliability and also for enhancing the efficiency.
FIGURE 27.3 Perturb and observe algorithm.
In this method, as the method suggests, the controller changes the voltage of the panel
each time and measures power and if the power increases, further adjustments are made
in the controlling procedure till the power becomes constant. This is called the perturb
and observe method as the observation of the change in power is observed each time.
This is the most commonly used method, but this can result in oscillations in the power
output.
27.2.2 DC-DC Circuit

The DC-DC converter that we have implemented in our system is a basic Cuk converter
which comprises of two inductors and two capacitors. The motivation behind using this
converter is that it comprises of a coupling capacitor which proves to be very essential in
solar circuits as it helps to maintain a constant voltage, thus acts as a constant source to
the next stage of our circuit which is the inverter. The canonical cell forms the basis of
analyzing switching circuits, but the energy transport mechanism forms the foundation
of the building blocks of such converters. The buck converter may consequently be seen
as a voltage to current converter, the boost as a current to voltage converter, the buck-
boost as a voltage-current-voltage and the CUK as a current-voltage-current converter.
Figure 27.4 shows the schematic of a basic buck-boost converter.
FIGURE 27.4 Circuit diagram of buck-boost converter.
Cuk converter is the cascade combination of a boost and a buck converter with output
voltage lesser or greater than the input voltage. The main advantage of this type of
converter is that they have a coupling capacitor between the two cascaded parts which
helps in providing a constant voltage to the second part and thus provides a constant
output. Figures 27.5 and 27.6 show the basic schematic diagram and the circuit diagram
of a Cuk converter.
FIGURE 27.5 Schematic diagram of cuk converter.

FIGURE 27.6 Circuit diagram of cuk converter.
FIGURE 27.7 Circuit diagram of cuk converter for different modes.
The calculation of the average output voltage and inductor currents is shown as follows.
Applying the volt-sec law across L1,
VinDT + (Vin – VC1)(1 – D)T = 0 (27.1)
or, Vin(1 – D)VC1 = 0
or, VC1 = Vin/(1 – D) (27.2)
Applying the volt-sec law across L2,

(VO + VC1)DT + VO (1 – D)T = 0 (27.3)
or, VO + DVC1 = 0
or, VO = –DVC1 = –DVin/(1 – D) (27.4)
Expression for average inductor current can be obtained from charge balance of C2,
IL2 + IO = 0
or, IL2 = –IO = –VO/R = –VODVin/R(1-D) (27.5)
and from power balance, we can get that
VinIL1 = VOIO = VO2/R = D2Vin2/R(1–D)2
or, IL1 = D2Vin/(1–D)2R (27.6)
27.2.3 Proposed Multi-Level Inverter

In this study, we have developed a novel single-phase 13-level H-Bridge multi-level
inverter that has three front-end controls switches S1, S2, and S3 and a single H-Bridge
inverter. The advantages of the proposed multilevel inverter topology are that the
number of switching devices and the total harmonic distortion are reduced. This reduces
the switching loss hence increasing the efficiency of the output. This proposed multi-
level inverter overcomes all the disadvantages while being more compact. Here we have
implemented a novel PWM technique, which makes use of 13 carrier signals compared
to a reference signal which is fed to a priority encoder to control the switching. The
proposed topology is evaluated by simulation and hardware implementation.
This proposed MLI generates 13-level output without using bi-directional switches and
capacitors. It consists of three sources (connected to the outputs of the DC-DC converters)
and diodes connected in between the switches S1, S2, and S3 as shown in Figure. The four
sources are in series as the voltage source to the H-Bridge inverter. Different voltage levels
are produced by connecting and disconnecting the sources V1–V3 using the switches S1–
S3. Switches S4–S7 controls direction of the current flow, hence, producing alternating
output at the load. So this circuit uses only seven switching devices and three diodes to
generate 13 level output.
Considering the maximum input DC voltage as Vdc, the inverter produces thirteen
output voltage levels (Vdc/13, 2Vdc/13, 3Vdc/13, 4Vdc/13, 5Vdc/13, 6Vdc/13, 7Vdc/13, 8Vdc/13,
9Vdc/13, 10Vdc/13, 11Vdc/13, 12Vdc/13, Vdc) from the DC supply voltage. The operation is
divided into thirteen modes having different voltage levels. If we consider the variable
‘m’ mode m will have a voltage of (m–1)Vdc/13. The switching states are illustrated in
the following figures. The same voltage levels are obtained in the negative as well. The
modified H-bridge topology is significantly advantageous over other topologies, that is

less power switches, power diodes, and less capacitor for inverters of the same number
of level.
FIGURE 27.8 Schematic diagram of MLI.
The switching sequence of the switches S1, S2, and S3 are shown by the “Binary Logic”
in which the digits are formed which is between 0000 and 1111. The ‘0’ indicates that the
switch is in OFF condition and ‘1’ indicates that the switch is in ON condition. The logic
is explained in the next section. The switching sequence in the front-end of the IGBTs is
fired based on this sequence, hence, generating multi-level waveforms. Positive voltages are
obtained when switches S4 and S7 are ON and S5 and S6 are OFF. Similarly, negative voltage
is obtained when S4 and S7 is OFF and when S5 and S6 are ON. The switching pulses
are controlled by normal sinusoidal pulse width modulation technique for the H-bridge
IGBTs. Here, pair of IGBTs are fired simultaneously to obtain positive and negative cycle
waveforms.
27.2.4 Proposed PWM Technique

In Figure 27.3, the pulse generation for our inverter section is shown. Here we have used the
multi-level inverter system, thus to generate the pulses for the switches, we have compared
a triangular wave with a sinusoidal wave. The number of comparison levels is equal to
seven since we wish to have a 13-level output. Thus, for a complete sinusoidal cycle, there
are a total of 14 comparisons. The comparison is such that when the sinusoidal function is
greater than the triangular, then the output of the comparator is high else low. This method
of PWM generation is known as the alternate phase opposition disposition (APOD) PWM
control method.
FIGURE 27.9 Simulation result of APOD-PWM.
On each level, a symmetric pulse is produced with the on-time for the pulses reducing in
each step. Thus, clubbing all the levels together gives rise to the multi-level sinusoidal wave.
Figure (put the output pulse figure) shows the final pulse output of the inverter switches.
The switching pulses given to the H-bridge IGBTs are controlled by normal complementary
gate pulses such that when a pair of switches is ON, the other switches are OFF and vice-
versa. When S4 and S7 are ON, positive half of the waveform is generated and when S5 and
S6 are ON, the negative half of waveform is formed.
27.3 SIMULATION RESULT AND ANALYSIS

Figures 27.10 and 27.11 are the simulation diagram of our proposed topology.

The following results were obtained while simulating the proposed system on MATLAB.
FIGURE 27.12 Pulse output from the MPPT controller.

FIGURE 27.13 Voltage curve of capacitor C1.
FIGURE 27.14 Voltage curve of inductor L1.
FIGURE 27.15 Voltage curve of inductor L2.

FIGURE 27.16 Output DC voltage curve.
FIGURE 27.17 APOD PWM pulse for inverter.
FIGURE 27.18 Inverter output voltage curve.

From the simulation results, we have seen that the output voltage at the inverter output is
a steady 230 Vrms voltage in the form of a 13-level modified sinusoidal wave which was
out target of this system. With further modifications, the DC output levels can be further
stabilised.
27.4 CONCLUSION
In normal solar converter systems, the power is first stored in a battery and then
converted into AC power through an inverter. Our system, on the other hand, provides
a topology for the direct conversion of solar power to AC using a moderate gain DC-
DC converter and a MLI at the end to get a modified sine wave as the final output.
The novelty of our circuit lies in the fact that the MLI that we have implemented uses
same level inputs unlike conventional MLIs and the level shifting is done using the
APOD method. Thus, converters with the same parameters can be used to run this
MLI. Moreover, the use of MPPT algorithm ensured the fact that we are always getting
the maximum power from the solar panels and the DC converters make sure that the
fixed input is always present at the inverter input terminals by the use of a coupling
capacitor.
REFERENCES
1. Safari, Azadeh, and Mekhilef, Saad, 2011. Simulation and Hardware Implementation of
Incremental Conductance MPPT With Direct ControlMethod Using Cuk Converter, IEEE
Transactions on Industrial Electronics, 58(4), April 2011.
2. Kachhiya, Kinal, Patel, Mukesh, and Lokhande, Makarand, 2011. MATLAB/Simulink
Model of Solar PV Module and MPPT Algorithm, National Conference on Recent Trends
in Engineering & Technology, 13–14 May 2011.
3. Femia, N., Petrone, G., Spagnuolo, G., Vitelli, M.. A new analog MPPT technique TEODI,
Progress in Photovoltaics: Research and applications, Wiley-Blackwell, 2010, 18, pp. 28–41.
4. Hohm, D. P., and Ropp, M. E. Comparative study of maximum power point tracking
algorithms, Progress in Photovoltaics: Research and Applications, l1(l), 2003, pp. 47–62.
5. Tse, K. K., Ho, M. T., Chung, H. S.-H., and Hui, S. Y., 2002. A novel maximum power point
tracker for PV panels using switching frequency modulation, IEEE Transactions Power
Electron, 17(6), Nov., 2002, pp. 980–989.
6. Balata, M., and Vitelli, M. A Hybrid MPPT technique based on the Fast Estimate of the
Maximum Power Voltages in PV applications Department of Industrial and Information
Engineering-Second University of Naples DIII-SU N Aversa (CE)-Italy marco.balato@
unina2.it; massimo.vitelli@unina2.it
7. Stability Analysis of FPGA Based Perturb and Observe Method MPPT Charge Controller
for Solar PV System Tripti Saini, Dogga Raveendhra and Padmanabh Thakur.
8. Jha, A., and Singh, B., 2006. Power quality improvement using CSC converter for high
power LED driver, In Proceedings of the 6th IEEE International conference on power
systems (ICPS), New Delhi, India, 1–6.
9. Singh, B., and Bist, V., 2015. A BL-CSC converter-fed bldc motor drive with power factor
correction, IEEE transactions on industrial electronics, 62(1), 172–183.
10. Hwu, K. I., and Peng, T. J., 2012. A novel buck–boost converter combining KY and buck
converters, IEEE transactions on power electronics, 27(5), 2236–2241.
11. Wu, H., Mu, T., Ge, H., and Xing, Y., 2016. Full-range soft-switching-isolated buck-boost
converters with integrated interleaved boost converter and phase-shifted control, IEEE
transactions on power electronics, 31(2), 987–999.
12. Murthy-Bellur, D., and Kazimierczuk, M. K., 2011. Isolated two-transistor zeta converter
with reduced transistor voltage stress, IEEE transactions on circuits and systems II: Express
briefs, 58(1), 41–45.
13. Khatab, A. M., Marei, M. I., and Elhelw, H. M., 2018. An electric vehicle battery charger
based on zeta converter fed from a PV array, In Proc. IEEE International conference on
Environment and Electrical Engineering and IEEE Industrial and Commercial Power
systems Europe (EEEIC/I&CPS Europe), Palermo, Italy, 1–5.
14. Huang, J., and Shi, H., 2014. Reducing the common-mode voltage through carrier peak
position modulation in an SPWM three-phase inverter, IEEE transactions on power
electronics, 29(9), 4490–4495.
15. Chen, C., and Fayed, A., 2015. A low-power dual-frequency SIMO buck converter topology
with fully-integrated outputs and fast dynamic operation in 45 nm CMOS. IEEE journals
& magazines, 50, 2161–2173.
16. Raghamadhuri, S. V., and Subramanian, K., 2017. SIMO based inverter with MPPT
technique, In Proc. International conference on Computation of Power, Energy Information
and Communication (ICCPEIC), Melmaruvathur, India, 506–509.
17. Tidke, A., Savakhande, V., Chewale, M., Sonawane, P., and Wanjari, R., 2018. MPPT based
sepic converter with voltage doubler cell for PV systems, In Proceedings International
conference on Control, Power, Communication and Computing Technologies (ICCPCCT),
Kannur, India, 195–199.
18. Ranjana, M. S. B., Reddy, N. S., and Kumar, R. K. P., 2014. A novel sepic based dual output
dc- dc converter for solar applications, In Proc. Power and Energy systems: towards
sustainable energy, Bangalore, India, 1–5.
19. Kaouane, M., Boukhelifa, A., and Cheriti, A., 2015. Design of a synchronous sepic dc-dc
converter for a standalone photovoltaic system, In Proceedings IEEE Canadian conference
on Electrical and Computer Engineering, Halifax, Canada, 870–874.
20. Lu, X., Sun, K., Ma, Y., Huang, L., and Igarashi, S., 2009. High efficiency hybrid cascade
inverter for photovoltaic generation, In Proceedings IEEE Region 10 conference TENCON.
doi: 10.1109/TENCON.2009.5396210
21. Chithra, M., and Dasan, B., 2011. Analysis of H-bridge multilevel inverters with photo
voltaic arrays, In Proceedings International conference on emerging trends in Electrical
and Computer technology, Nagercoil, India, pp 441–447.
22. Marquez, A., Leon, J. I., Vazquez, S., and Franquelo, L. G., 2014. Advanced control of
multilevel h-bridge converter for pv applications, In Proceedings of the 40th annual
conference of the IEEE industrial electronics society, Dallas, TX, USA, 4548–4553.
23. Manoharan, M. S., Ahmed, A., Kim, H. W., and Park, J. H., 2015. A single source photo
voltaic power conditioning system using asymmetric cascaded multi-level inverter, In
Proceedings of the 9th International conference on Power Electronics and ECCE Asia
(ICPE-ECCE Asia), Seoul, South Korea, 21–26.
24. Patil, S. N., and Prasad, R. C., 2015. Design and development of mppt algorithm for high
efficient dc-dc converter for solar energy system connected to the grid, In Proc. International
conference on energy systems and applications, Indore, India, 228–233, 2015.
25. Xiao, B., Hang, L., Mei, J., Riley, C., Tolbert, L. M., and Ozpineci, B., 2015. Modular cascaded
h-bridge multilevel PV inverter with distributed MPPT for grid connected applications,
IEEE transactions on industrial applications, 51(2), 1722–1731.
CHAPTER 28
Battery Storage Technologies:

A Review
Bikram Kumar Samanta,1 Hiranmay Samanta,2 Nirmal Kr. Deb,3
and Hiranmay Saha4
*Centre of Excellence for Green Energy and Sensor Systems, Iiest Shibpur, Howrah, India.
1
bksamanta9@gmail.com, 2hiranmaysamanta@gmail.com, 3debnirmal97@gmail.com,
4
sahahiran@gmail.com
ABSTRACT: Batteries are promising energy storage technologies for various power
system applications because of their maturity and the ease with which they are designed
and installed compared to other conventional energy storage technologies. The various
requirements for battery power and the different environmental and electrical conditions
under which they must operate necessitate the use of a number of different types of
batteries and designs, each having the most advantageous performance under specific
operational conditions. Although many advances have been made in battery technology
in recent years, both through continued improvement of a specific electrochemical system
and the development and introduction of new battery chemistries, there is still no one
‘ideal’ battery that performs optimally under all operating conditions. As a result, over
time, many different electrochemical systems and battery types have been and are still
being investigated and promoted. However, a relatively small number have achieved wide
popularity and significant production and sales volumes. A number of factors must be
considered in selecting the best battery for a particular application. It is important that the
selection of the battery be considered at the beginning of equipment development rather
than at the end when the hardware is fixed. In this way, the most effective compromises can
be made between battery capabilities and equipment requirements. The chapter discusses
selection criteria of batteries with improved cycling capacity, higher lifespan, and lower cost
that can achieve lower environmental impacts for various future applications.
Keywords: Cycle life, Lifetime, Power capital cost, Round trip efficiency, Specific energy,
Working temperature.
257
28.1 INTRODUCTION
The batteries are selected on the basis of types of battery (primary, secondary, or reserve
system), electrochemical system (matching of the advantages and disadvantages and of
the battery characteristics with major equipment requirements), voltage (nominal or
operating voltage, maximum and minimum permissible voltages, voltage regulation,
profile of discharge curve, start-up time), load current and profile (constant current,
constant resistance or constant power, value of load current or profile), duty cycle
(continuous or intermittent), temperature requirement (temperature range at which
operation is required), service life (length of time operation is required), physical
requirement (size, shape, weight, and terminals), charge discharge cycle (cycling service,
cycle requirement, availability and characteristics of charging source, and charging
efficiency), safety and reliability (permissible variability, failure rates, freedom from
leakage, use of potentially hazardous or toxic components, type of effluent or signature
gases or liquids, high temperature, operation under severe or potentially hazardous
conditions, and environmentally friendly), maintenance and resupply (ease of battery
acquisition, accessible distribution, ease of battery replacement, available charging
facilities, special transportation, recovery or disposal procedures required), cost (initial
cost, operating, or life-cycle cost). The cycle life is the number of complete charge-
discharge cycles that the battery can support before its capacity falls under 80% of its
original capacity. Battery lifetime is a measure of battery performance and longevity.
Energy storage typically consumes electricity, and it also saves and backs it to grid. The
ratio of energy into energy retrieved from storage is known as the round trip efficiency.
It expressed in percentages.
The power density is the rated output power divided by the volume of the storage device.
The energy density is calculated as stored energy divided by the volume. The volume of
the storage device is the volume of the whole energy storage system including the energy
storing element, accessories and supporting structures, and the inverter system. Working
temperature is the range of temperature in which the battery works with good efficiency.
Cell voltage of a battery is the potential difference between the working electrode and
counter electrode. Capital cost is one of the most important factors for industrial purpose.
They can be expressed in the form of cost per kWh, per kW, and per kWh per cycle (Chen
et al., 2009, Akinyele et al., 2017).
28.2 METHODOLOGY
28.2.1 Lead Acid Battery
The lead-acid technologies are the oldest form of battery energy storage system
developed by a French physicist Gaston Planté in 1859 (Chen et al., 2009, Zakeri et
al., 2015). They have been a common storage option for mini or micro-grids or the
grid-independent electrical power systems, uninterrupted power supply, and spinning
reserve application. The electrochemistry of lead-acid technologies in the charge state
consists of a lead dioxide (PbO2) and lead (Pb) in a concentrated tetraoxosulphate
Battery Storage Technologies: A Review  259
(VI) acid electrolyte. The PbO2 and the Pb are the positive and negative electrodes,
respectively. However, in the discharge state, the electrodes – lead dioxide and the lead
are converted to lead sulphate (PbSO4); thus, they consumed the sulphate ions. This
development reduces the specific gravity of the electrolyte to a level similar that of water,
meaning that the electrolyte loses its dissolved tetraoxosulphate (VI) acid and turns to
water (Chen et al., 2009). The PbO2 (positive electrode) is the key factor that influences
the performance and cycle life of lead-acid technologies, while Pb (negative electrode)
determines the cold-temperature performance of the systems. There are different types
of lead-acid technologies including the flooded type that requires regular topping up
with distilled water and the sealed maintenance free type that has a gelled or absorbed
electrolyte, and the valve regulated type.
Lead-acid batteries have a low cycle life ranging from 2000 to 2500, a round trip-
efficiency (RTE) of 70–90%, and a lifespan of 5–15 years and low-specific energy of
40–50 Wh/kg (Zakeri et al., 2015). Specific power is 170–200 W/kg. Cell voltage is
2.1 volt. They are identified as low-cost secondary battery technologies, which is one of the
reasons for their widespread application of electrical and renewable energy applications.
They possess a moderately good operating temperature ranging from –40°C to 60°C.
However, they have a potential for generating a negative environmental influence because
of the toxic remnants they produce. Their grids also contain antimony and arsenic, which
constitutes health hazards.
28.3 SODIUM SULFUR BATTERY

The NaS technologies were developed by NGK Insulators and Tokyo Electric Power in
1987 (Zakeri et al., 2015). They are one of the most proven battery storage technologies
for the mega-watt scale electrical applications. They NaS batteries have been applied
for power quality and power time shift purposes, because of their relatively higher RTE
ranging from 75 to 90% (Chen et al., 2009, Zakeri et al., 2015). The NaS battery consists
of molten sulphur at the anode and molten sodium at the cathode, which are separated by
a solid beta alumina ceramic electrolyte. The electrolyte allows the positive sodium (Na+)
ions to flow through it and then combines with sulphur to form sodium polysulphides.
In the discharge state, the Na+ ions flow through the electrolyte, and this causes electrons
to flow in the battery’s external circuit, thus, delivering 2 V. The electrochemistry is
a reversible process, as charging causes the Na2Sx to release the Na+ back through the
electrolyte, to recombine as elemental sodium (Na). They have a cycle life, lifespan, and
discharge time of 2500–4500, 10–15 years, and up to 7 hours, respectively (Díaz-González
et al., 2012). The power rating of the technology is scalable, and this promises widespread
utility-scale applications in the future. The typical energy and power densities of the NaS
technology range from 150–240 Wh/kg to 150–230 W/kg, respectively. However, the
main drawback of this technology is that a heat source is usually required that uses a part
of the battery’s energy; thus, reducing the battery performance, as it needs to operate at a
high temperature of about 300–350°C.
FIGURE 28.1 Physical diagram of sodium sulphur battery (Akinyele et al., 2017).
28.4 NICKEL CADMIUM BATTERY

This technology is one of the oldest battery storage technologies (Chen et al., 2009;
Zakeri et al., 2015; EPRI and the US Department of Energy, 2003), for instance, the
nickel-cadmium (Ni–Cd) battery was invented by Waldemar Jungner in 1899. However,
the application of the Ni–Cd was limited because of material cost and the difficulty to
manufacture. There exist five different battery technologies that use the nickel-electrode
in their design, manufacture, and operation. These include the nickel-iron (NiFe), Ni–Cd,
nickel hydrogen (Ni–H2), nickel metal hydride (Ni–MH), and nickel-zinc (Ni–Zn). Ni–
Cd and Ni–MH are the widely used technologies compared to others; however, Ni–Cd is
currently the most utilised nickel-electrode in modern utilities around the world (EPRI
and the US Department of Energy, 2003). They have energy and power density of 50–75
Wh/kg and 150–300 W/kg, respectively, cycle life of 2000–2500, a lifespan of 10–20 years
and RTE of 70–75% (Díaz-González et al., 2012). The Ni–Cd battery is employed for
different applications such as power quality and for forestalling unscheduled outage, that
is, emergency reserve for communication services, for power tools, portable devices and
emergency lighting, UPS, and generator starting. The anode of the Ni–Cd battery is nickel
hydroxide (Ni(OH)2) while the cathode is a cadmium hydroxide plate. The battery also
has a separator and an alkaline electrolyte. The battery usually designed with a metal case
and a sealing plate, equipped with a safety valve. During the charging process, the Ni(OH)2
is converted to nickel oxyhydroxide–NiOOH, while the anode is converted to cadmium
hydroxide Cd(OH)2 (Hadjipaschalis et al., 2009). The conversion process of the anode
is possible through oxidation when it is charged in the presence of aqueous potassium

hydroxide KOH. In the discharge state, NiOOH reacts with H2O to produce the Ni(OH)2
and hydroxide ion at the anode. Cell voltage for this battery is 1.2 volt.
FIGURE 28.2 Physical diagram of nickel cadmium battery (Akinyele et al., 2017).
28.5 LITHIUM-ION BATTERY

Traditionally, lithium-based batteries are widely used in smaller appliances, such as mobiles
and laptops, but not employed for electric power supply purposes. Lithium batteries are fast
developing since their invention in the 1960s and are categorised into lithium polymer cells
and lithium-ion technologies (Chen et al., 2009; Whittingham, 2012; Battery University).
In lithium-ion battery technology, the organic carbonates of lithium (LiPF6) are used as
electrolyte. The negative electrode is usually a lithium metal oxide such as LiMO2, LiCoO2,
or LiNiO2, while the positive electrode is made of graphitic carbon. During charging, the
lithium atoms in the negative electrode become ions and migrate to the positive electrode
(i.e. carbon) where they recombine with external electrons to form the lithium atoms.
During discharging, the process is reversed. Cell voltage is 3.6–3.8 volt. As research and
development in the technology advances, the energy density has increased from 75 to 200
Wh/kg, with an increased cycle life of 10,000 cycles (Chen et al., 2009). The RTE of the
Li-ion batteries is about 90%. This is an important of this technology over other battery
technologies. The lithium polymer batteries, on the other hand, have relatively lower
efficiency and lifespan. Although Li-ion batteries take over 50% of the small portable
devices market, there are some challenges for making large-scale Li-ion batteries. The main
hurdle is the high cost due to special packaging and internal overcharge protection circuits.
Charge–discharge phenomena is shown in the figure.
FIGURE 28.3 Physical diagram of lithium-ion battery (Akinyele et al., 2017).
28.6 VANADIUM REDOX FLOW BATTERY

These battery technologies are designed to store energy in the electrolyte solutions – a
feature that is opposite to the conventional battery technologies in which the electrodes are
used for such a task (Chen et al., 2009; Zakeri et al., 2015). The battery arrangement also
includes additional electrolyte that is stored externally in tanks and is pumped through the
cell (or cells) of the reactor. The mode of operation of flow battery technologies is based
on reduction–oxidation (redox) reactions of the electrolytes. In the charging state, one
electrolyte is oxidized at the positive electrode, while the other is reduced at the negative
electrode (Luo et al., 2015). This way, the electrical energy is converted to chemical energy.
The chemical reaction is reversible, and it allows the battery to be charged, discharged and
recharged as desired. The power and energy capacities of these battery technologies could
be designed independently, that is, the energy capacity is determined by the amount of
electrolyte that is stored in external tanks but the power rating is ascertained through the
active area of the cell compartment. These technologies have the potential for continuous
release of energy at a high rate, with a discharge of up to 10 hours. The vanadium redox flow
(VRFB) is one among the mature flow battery technologies, and it stores energy by using
the vanadium redox ions V2+/V3+ and V4+/V5+ in two different electrolytic tanks (Chen et al.,
2009; Luo et al., 2015; Linden et al., 2002; Divya et al., 2009) as shown in the figure. It uses
the vanadium in the four oxidation states, thus, making it to have only one active element
in the anolyte and catholyte. In the charge/discharge states, the process involves the
exchange of H+ ions through the selective membrane of the ion, and a cell voltage of
1.15 to 1.55 volt is developed in the process. Vanadium redox batteries possess quick
responses, even faster than 0.001 s and they have cycle life in the range of 10,000 to over
16,000 cycles (Luo et al., 2015; Barton et al., 2004). They have a RTE of 85% and can guarantee
continuous power with discharge duration of over 24 hours. They have energy density of
10–20 Wh/kg and lifespan of 15–20 years. They are suitable for enhancing power quality,
uninterruptible power supply UPS, forestalling unscheduled power outages, and balancing
the intermittent characteristics of renewable energies. However, the major challenges with
VRBs are low electrolyte stability and solubility, which results in low energy density, and
relatively higher operating cost (Chen et al., 2009; Luo et al., 2015).
FIGURE 28.4 Physical diagram of vanadium rdox flow battery. (Akinyele, D. et al 2017)
28.7 ZINC BROMIDE BATTERY

In these battery technologies, two aqueous electrolyte solutions contain the reactive
materials made of zinc (Zn) and bromine (Br) elements, which are stored in two external
tanks (Luo et al., 2015). During the discharge process, the reactive materials combine to
form zinc-bromide and then generate 1.8 V across each of the battery cells. This then
increases the Zn2+ and Br– ion densities in the two electrolyte tanks (Chen et al., 2009).
However, in the charge state, zinc is deposited as a thin layer on a side of the composite
electrode, while bromine is developed as a dilute solution on the other side of the membrane.
It then reacts with other agents, that is organic amines to develop thick bromine oil,
which sinks to the bottom of the electrolytic tank. This is mixed with the remaining
electrolyte during discharge. The chemical reactions at the anode and the cathode are
shown in the figure. The zinc-bromine batteries have a lower RTE and lifespan of
65–75% and over 2000, respectively, compared to the conventional lead-acid batteries.
They have an energy density ranging from 35 to 55 Wh/kg (Chen et al., 2009) However,
the major problems with ZnBr battery are material corrosion and dendrite formation
(Luo et al., 2015).
28.8 POLYSULPHIDE BROMINE

In these batteries, a reversible electrochemical reaction is allowed between two
electrolytes (salt solution) viz. sodium bromide and sodium polysulfide. The electrolytes
are separated in the cells by a polymer membrane that only allows the migration of
positive sodium ions, making the battery to generate 1.5 V across the membrane.
The main advantages of PSB batteries are the abundance and cost-effectiveness of the
materials for producing the two electrolytes and their solubility in aqueous electrolyte.
They also have a fast response time of around 0.02 s (Chen et al., 2009; Luo et al.,
2015) and this makes them a suitable technology for power system frequency control
and voltage control. They have a lifespan ranging from 10–15 years, and a RTE
efficiency of 75%, and they also operate at room temperature [1, 9]. However, during
the electrochemical reactions, bromine and sodium sulphate crystals are produced and
this may result in negative environmental impact (Luo et al., 2015).
28.9 SODIUM-NICKEL CHLORIDE

This technology has been commercially available since about two decades ago. It is
otherwise known as Zero Emission Battery Research (ZEBRA) and is also recognised as
a high-temperature battery with temperature ranging from 270–350°C (Chen et al., 2009;
Zakeri et al., 2015). In the case of NaNiCl2, the nickel chloride is used as the anode instead
of sulphur as is the case with NaS. ZEBRA battery technologies can withstand limited
overcharge and discharge; they possess better safety characteristics, a higher cell voltage
of 2.58 V, compared to the NaS technologies (Chen et al., 2009). They have a cycle life of
2500 and a lifespan ranging from 10–15 years, with a discharge time of seconds to hours
and RTE of 85–90%. However, they possess energy and power densities of 100–120 Wh/
kg and 150–200 W/kg, respectively, which are better than the values reported for lead-acid
technologies.
28.10 DISCUSSION
All the comparison data are summarised in Tables 28.1 and 28.2
TABLE 28.1 Comparison of the Characteristics of Different Types of Battery Technologies (Chen et al 2009; Akinyele et al., 2017; Luo et al., 2015)
Round Trip Specific Nominal Energy

Lifetime Specific Power Working Power Capital Cost ($/
Technologies Cycle Life Efficiency Energy Cell Voltage Capital Cost
(years) (W/kg) Temperature (C) kwh)
(RTE) (%) (Wh/kg) (V) ($/kwh)
Lead acid 2000 – 2500 70 – 90 5 – 15 170 – 200 40 – 50 -40 – 60 2.1 200 - 400 300 – 600
Sodium sulphur 2500 – 4500 75 – 90 10 – 15 150 –230 150– 240 300 – 350 2 300 - 500 1000 – 3000
Nickel cadmium 2000 – 2500 70 – 75 10 – 20 150 – 300 50 – 75 -20 – 65 1.2 500 - 1500 500 – 1500
3.6 /3.7
Lithium-ion 8000 90 3–5 250 – 340 100 – 265 -20 – 60 900 - 1300 600 – 2500
/3.8
VRFB 10000 – 16000 80 – 85 15 – 20 - 10 – 20 15 – 55 1.15 – 1.55 600 - 1500 150 – 1000
Zinc bromide 2000 – 3000 65 – 75 10 – 15 - 35 – 55 20 – 50 1.8 700 - 2500 150 – 1000
Poly-sulphide
- 70 - 80 10 – 15 - - 300 – 350 1.5 700 - 2500 150 – 1000
Bromine
Sodium nickel
2000 – 3000 85 – 90 10 – 15 150 – 200 100 – 120 270 – 350 2.58 100 - 200 150 – 300
Chloride
Battery Storage Technologies: A Review
265
TABLE 28.2 Comparison of Development Status, Advantages and Shortcomings of Battery Technologies (Akinyele et al., 2017; Electropedia)
266
Technologies Development Status Advantages Shortcomings


Lead acid Mature Cost is low, easily available, reliable, and easily replaced. It requires high maintenance, short cycling capability
It is suitable for power quality, UPS, and spinning reserve compared to others, low power and energy density,
applications. charging is slow, low weight to energy ratio, thermal
management required, it has an environmental hazard
(toxic component).
Sodium sulphur Developed / It has relatively high power and energy density. It is very For this heat source is required. Cost is also high.
Commercialized much economical for power quality and peak shaving
purposes.
Nickel cadmium Developed / It has relatively high-energy density, high mechanical Cost is high for this battery. It has an environmental hazard
Commercialized resistance, low maintenance requirements. It is very much (e.g. toxic heavy metal ‘cadmium’), memory effect in which
useful for power tools, emergency lighting, generator case charge of battery becomes full after a couple of full
starting, telecoms, and portable devices. discharge.
Lithium-ion Developed / Early It has relatively high power and energy density, 90% Cost is high in this case and it also degrades at high
stage technology efficient, higher cycling capacity, it responses fast to temperature.
charge and discharge operations.
VRFB Demonstration / Early It has high round trip efficiency (RTE). It is suitable for One major drawback of the VRFB is the complicated
commercialization improved power quality, UPS, peak shaving. It is very electrolyte circulation system, which leads to low efficiency
stage much useful for integration of renewable resources. when it is operated at less than 20% of its rated power.
Zinc bromide Developed It has higher energy density compared to lead acid It has low round trip efficiency (RTE) and lifespan
Intelligent Electrical Systems: A Step towards Smarter Earth
battery. It has 100% depth of discharge capability on a compared to the conventional Lead Acid battery. It also
daily basis. suffers for material corrosion and dendrite formation.
Poly-sulphide Developing stage It is very much useful as it operates at room temperature. It has relatively low dc output voltage (1.5 volts).
Bromine
Sodium nickel Commercialized stage It has the ability to withstand limited overcharge and It has lower power and energy density compared to NaS
Chloride discharge with better safety features and a relatively battery. It is suitable for mainly over 20kWh. Very few
high cell voltage of 2.58 volts, it is suitable for load- manufacturers produce this battery technology.
levelling applications in the industry.
From the comparison tables, we can learn that Lead Acid battery is inexpensive compared
to newer technologies, so they are widely used even when surge current is not important
and other designs could provide higher energy densities. It is also mature and readily
available, reliable and easily replaced. It is suitable for power quality, UPS, and spinning
reserve application, but it has short cycling capability, low weight-to-energy ratio, and it has
an environmental hazard. Sodium sulphur battery has high power and energy density but
cost is too high for this battery. Nickel cadmium battery has relatively high energy density,
high mechanical resistance, low maintenance requirements, but it has an environmental
hazard because of toxic heavy metal cadmium. Lithium-ion batteries are commonly used
for portable electronics and electric vehicles and are growing in popularity for military
and aerospace application. It degrades at high temperature. Vanadium redox flow battery
has high round trip efficiency (RTE), it has higher lifetime so it is very much useful for
integration of renewable resources. Zinc bromide battery has higher energy density
compared to lead-acid battery. It has 100% depth of discharge capability on a daily basis.
Poly-sulphide bromine is very much useful as it operates at room temperature. Sodium
nickel chloride battery has the ability to withstand limited overcharge and discharge with
better safety features and a relatively high cell voltage of 2.58 volts, which is suitable for load
levelling applications in the industry. It has lower power and energy density compared to
sodium sulphur battery. It is suitable for large capacity applications mainly over 20 kWh
capacity (Chen et al., 2009; Akinyele et al., 2017; Electropedia).
28.11 CONCLUSION
While smartphones, smart homes, and even smart wearable are growing ever more advanced,
they are still limited by power. The battery has not advanced in decades. But we are on the
verge of a power revolution. So a few new technologies is discussed which can be useful
in future. This alternative type of lithium-ion battery uses silicon to achieve three times
better performance than current graphite Li-ion batteries. The battery is still lithium-ion
like the one found in the smartphone, but it uses silicon instead of graphite in the anodes.
Graphene batteries have the potential to be one of the most superior available. Grabat has
developed graphene batteries that could offer electric cars a driving range of up to 500
miles on a charge. Prieto believes the future of batteries is 3D. The company has managed
to crack this with its battery that uses a copper foam substrate. This means these batteries
will not only be safer, thanks to no flammable electrolyte, but they will also offer longer life,
faster charging, five times higher density, be cheaper to make, and be smaller than current
offerings. Scientists at WMG at the University of Warwick have developed a new technology
that allows current lithium-ion batteries to be charged up to five times faster than current
recommended limits. The technology constantly measures a battery’s temperature far more
precisely than current methods. Solid-state batteries represent a paradigm shift in terms of
technology. The first huge advantage is a marked improvement in safety at cell and battery
levels; solid electrolytes are non-flammable when heated, unlike their liquid counterparts.
Second, it permits the use of innovative, high-voltage high-capacity materials, enabling
denser, lighter batteries with better shelf-life as a result of reduced self-discharge. Moreover,
at system level, it will bring additional advantages such as simplified mechanics as well
as thermal and safety management. As the batteries can exhibit a high power-to-weight
ratio, they may be ideal for use in electric vehicles. In this way, the battery technologies are
developing day by day.
REFERENCES
1. Chen, H., Cong, T. N., Yang, W., Tan, C., Li, Y., and Ding, Y., 20009. Progress in electrical
energy storage system: A critical review. Progress in Natural Science, 19, 291–312.
2. Akinyele, D., Belikov, J., and Levron, Y., 2017. Battery Storage Technologies for Electrical
Applications: Impact in Stand-Alone Photovoltaic Systems, Energies.
3. Zakeri, B., and Syri, S., 2015. Electrical energy storage systems: A comparative life cycle cost
analysis. Renewable and Sustainable Energy Reviews, 2015, 42, 569–596.
4. Díaz-González, F., Sumper, A., Gomis-Bellmunt, O., Villafáfifila-Robles, R., 2012. A review
of energy storage technologies for wind power applications. Renewable and Sustainable
Energy Reviews, 16, 2154–2171.
5. EPRI and the U. S. Department of Energy. EPRI-DOE Handbook of Energy Storage for
Transmission and Distribution Applications, Technical Report, Electric Power Research
Institute (EPRI): Palo Alto, CA, USA, 2003.
6. Hadjipaschalis, I., Poullikkas, A., Efthimiou, V., 2009. Overview of current and future
energy storage technologies for electric power applications. Renewable and Sustainable
Energy Reviews, 2009, 13, 1513–1522.
7. Whittingham, M. S., 2012. History, evolution, and future status of energy storage.
Proceedings IEEE, 100, 1518–1534.
8. Battery University. History of Battery Advancements. Available online: http://
batteryuniversity.com/learn/article/battery_developments
9. Luo, X., Wang, J., Dooner, M., and Clarke, J., 2015. Overview of current development
in electrical energy storage technologies and the application potential in power system
operation. Applied Energy 2015, 137, 511–536.
10. Linden, D., and Reddy, T. B., 2002. Handbook of Batteries, 3rd edition, McGraw-Hill: New
York, NY, USA.
11. Divya, K. C., and Østergaard, J., 2009 Battery energy storage technology for power systems—
An overview. Electric Power Systems Research, 79, 511–520.
12. Barton, J. P., Infield, D. G., 2004. Energy storage and its use with intermittent renewable
energy. IEEE Trans. Energy Convers, 19, 441–448.
13. Electropedia. Available online: http://www.mpoweruk.com
CHAPTER 29
Experimental Investigation
of Solar Energy-Assisted
DC Refrigerator
N. D. Shikalgar and S. N. Sapali
ABSTRACT: Protecting the environment and saving energy have always been a great
concern for human beings. The domestic refrigerator is no more a luxurious item but is an
essential commodity. Therefore, looking at a large number of refrigerators being installed
every year, it is essential to account the energy consumption of the refrigerator. The focus of
this chapter is to convert a conventional refrigerator to self-sustainable refrigerator operated
with solar energy input. Special attention is given to experimental performance analysis of
solar refrigerator operated with direct current (DC) compressor and alternating current
(AC) compressor. The outcomes of the experimental analysis revealed that the per day
energy consumption of a DC refrigerator is 0.87 kWh, which is indirectly saved using solar
energy. The coefficient of performance of a solar-assisted refrigerator is found to be in the
range of 1.98–2.49. The energy loss in AC system components is found to be 15% more than
the solar energy-operated direct drive DC compressor system. The research work signifies
that the solar-powered DC refrigerator has significant scope for research and development
in order to operate a self-sustainable refrigerator.
Keywords: Direct current compressor, Energy consumption, Refrigerator, Solar energy.
29.1 INTRODUCTION
In developing nations like India, there are few places where the power supply is
sporadic, in such territories, for the preservation of food, medicines, and vegetables, the
refrigeration systems operated on solar energy can be considered as the most favourable
solution. The photovoltaic (PV) power systems are getting to be essential in the power
generation showcase in the succeeding year (Enibe, 1997). The design and development
in a solar-powered refrigeration system are approaching towards the application base
269
optimisation of systems. Ouali et al. (2017) have conferred about the better innovative
technique to minimise the energy utilisation of solar-powered refrigerators, which helps
in reducing greenhouse gas emission. They have evaluated the solar energy losses and
energy deficiency in the refrigerator. Kaplanis and Papanastasiou (2016) have presented
the report on design methodology for converting a refrigerator to a solar-powered
refrigerator. They have investigated factors which affect the sizing of the solar refrigeration
system with minimum heat losses. Sidney and Mohan Lal (2016) have experimentally
studied the energetic and exergetic performance analysis of PV powered DC refrigerator.
The average exergy loss observed in the panel is 86.23%. Ekren and Celik (2013) analysed
the pull-down, steady-state performance of a solar panel driven DC refrigerator with
a battery bank. They observed that the exergy efficiency of the refrigeration system is
very low due to low energy. In the works of Opoku, Anane, and Seidu (2016) and Modi
et al. (2009), different sizes of conventional refrigerators have been converted to solar-
powered systems. However, there is a research gap in the comparative techno-economic
assessment of powering an AC refrigerator with solar energy. From literature, there is
no remarkable work on DC refrigerator assisted with solar energy without an inverter
and battery bank. This research is focused on the correlation between AC refrigerators
powered by a solar PV system and DC refrigerator assisted with hybrid solar PV system
with phase change material.
The performance of a solar-powered refrigerator is examined by a pull-down test,
energy consumption test. The losses in the inverter and charge controller are calculated for
records. This study is performed to evaluate the possibility of saving in energy consumption
of conventional refrigerator when powered with solar energy.
29.2 DESIGN OF SOLAR PHOTOVOLTAIC PLANT

Prior to experimentation work, the conventional refrigerator is first tested using grid
electricity during normal operating conditions to measure the energy consumption and
losses using energy meter with accuracy ± 0.01kWh (a product of Selec, model number
EM368). The output data from the study is utilised to determine the size (specification) of
solar panels, inverter, and battery bank requirement to operate the AC refrigerator with solar
power. To know the average power consumption of the AC refrigerators and compressor
run-time (etc) per day of the refrigerators during normal operation is measured. The daily
average energy consumption of each refrigerator is determined by using Eq. 29.1. The sizes
of the components of the solar PV system are found with Eq. (29.1):
é hours ù
Eavg, daily = Avg. power consumption Xtc ê ú (29.1)
ë day û
The intensity of solar radiation arriving on the earth’s surface is roughly 1361 W/m2. The
Sun’s ray leaving maximum irradiance normal to the surface is about 1000 W/m2. The
annual average solar direct irradiance of the Pune city is 5.44 kWh/m2/day. The solar panel
power (Wpeak) is determined using Eq. 29.2.
Experimental Investigation of Solar Energy-Assisted DC Refrigerator  271
E daily
Size of Solar PV Panel = (29.2)
hbat ´ hch ´ PSH ´ Fmfg ´ Tc ´ FDirt
The capacity of the inverter is taken as three times more than the capacity of a motor or
compressor to handle surge current during starting. The size of a solar charge controller
and the size of the battery bank are determined with Eqs (29.3) and (29.4).
E daily × DA × dt
Deep cycle battery = (29.3)
Vbat × DoD
Wpeak
Charge controller = (29.4)
Vbat
29.3 EXPERIMENTAL STUDIES

The experimental set-up is developed as shown in Figure 29.1. The solar-operated
experiment set up consists of a refrigerator with an AC compressor, solar PV panels, solar
inverter, DC compressor, charge controller, and battery bank.
FIGURE 29.1 The configuration of the experimental setup of a solar-powered refrigerator.
The methodology suggested by Opoku in 2015 is used for the conversion of AC

refrigerator to DC refrigerator. The temperatures are monitored inside the freezer, food
compartment, crisper compartments, compressor outlet, and condenser outlet of the solar-
powered refrigerator as per ASHRAE refrigerator standards. The energy consumption
of the compressor is measured with the help of digital energy meter. The solar radiation
is measured using solar pyranometer attached at the top of the solar panel. The inverter
with inbuilt charge controller is used to convert DC to AC power and regulate the current
and voltage. A charge regulator used to maintain the power supply within the range and
prevents overcharge of the battery. Experimentation is carried out as per ISO 15502:2005
household refrigerating appliances test guidelines.

The experimental investigation of fabricated solar-powered refrigeration system with AC and
DC compressor at the same speed is carried out. The DC and AC refrigerators performance
characteristics are evaluated in terms of their cabinet and evaporator temperatures.
The test data and results are used to analyse the variation of performance parameters
such as COP, energy consumption, pull-down speed, and solar power loss. Experimentation
is carried out to analyse the performance of the fabricated solar-powered refrigeration
system by measuring various related parameters under different operating conditions.
During the normal running test, the variation of solar radiation intensity is plotted with
respect to time. It is observed that during a day variation of solar radiation intensity with
time is a sinusoidal function. Figure 29.2(a) shows an increasing trend for the current with
solar radiation. During study time, SPV panels generated maximum charging current 10.54
A at 807.85 W/m2 solar radiation.
(a) (b)
FIGURE 29.2 The variation of (a) panel output current with intensity (b) compressor input
current with time.
Figure 29.2(b) shows the variation of compressor current requirement with time. To
overcome the starting torque and friction compressor take 866 mA starting current, and
compressor attains the stable condition with time and consumes constant current about
600 mA. The variation in current shows the efforts made by the compressor to maintain the
same temperature after the long, steady operation. The two valleys indicate defrost cycle.
The solar-power generated by PV panels directly depends on the solar irradiation. During
the experimentation test, the variation of solar panel conversion efficiency with respect to
time is as shown in Figure 29.3. The average solar panel efficiency of solar panel is 11.70 %.
FIGURE 29.3 The solar panel conversion efficiency.
The DC and AC refrigerators performance characteristics were evaluated in terms of their

cabinet and evaporator temperatures. During the pool down test, the refrigerator is kept
open to establish the thermal equilibrium condition with ambient temperature. Solar-
powered refrigerators temperature profiles for AC and DC compressor are shown in Figure
29.3. Very small variations are observed in temperature profiles for both refrigerators.
It is observed that during the pull-down test with DC compressor, −12°C temperature
of the freezer compartment is achieved in 127 minutes, whereas AC compressor achieved
this in 117 minutes. The energy consumed by the compressor during an energy test is 127
W, while in steady-state operation the energy consumption drops to 95 W at compressor
speed 3000 rpm. The compressor is operated for 8 hours continuously. The variation in the
energy consumption is 0.9 kWh with AC compressor and 1.1 kWh with DC compressor as
FIGURE 29.4 Energy consumption of the refrigerator.
The distribution of energy generated and supplied in a solar-powered DC refrigeration

system is shown in Figure 29.5. The rate of energy generated by solar panels is in the
range of 88 W to 294 W. The DC compressor consumes the energy at the rate of 110 W
to 135 W. The energy losses in solar charge controller are less as compared to the solar
inverter system.
FIGURE 29.5 Distribution of energy generated and supplied in a solar-powered DC refrigeration

system.
The energy consumption during experimentation for conventional solar-operated AC

system and solar-operated DC systems are 0.87 kWh and 1.1 kWh, respectively. The energy
consumption with a loss in solar-operated AC compressor refrigeration system is 1.16 kWh.
29.5 CONCLUSION
In this experimental study, the thermal performance of a household refrigerator with
AC and DC compressor driven by solar energy is presented. There is no degradation in
the thermal performance of a solar energy-assisted refrigerator with DC compressor.
The major facts observed during the study are discussed as follows:
1. Using solar energy one can run the refrigerator without grid electricity by employing
DC compressor. It is practically and technically viable to convert a conventional AC
refrigerator to a DC refrigerator.
2. The average solar panel efficiency during experimentation is 11.70%.
3. Based on different tests and results obtained in this study, solar energy-assisted DC
refrigerator is more economical than AC refrigerator for standalone refrigeration
applications.
4. The coefficient of performance of a solar-assisted refrigerator is found to be in the
range of 1.98 to 2.49.
5. The energy loss in system components of solar energy-operated AC Compressor is
15–17% more as compared to DC compressor.
6. The 0.87 kWh per day electricity consumption is saved with solar-operated DC
compressor refrigeration systems.
REFERENCES
1. Enibe, S. O., 1997. Solar refrigeration for rural applications, Journal of Renewable Energy,
12, 157–67.
2. Oualim, M., Djebiret, M. A., Ouali, R., Mokrane, M., Merzouk, N. K., Bouabdallah, A.,
2017. Thermal control influence on energy efficiency in domestic refrigerator powered by
photovoltaic International Journal of Hydrogen Energy, 42 (13), 8955–8961.
3. Kaplanis S., and Papanastasiou, N., 2006. The study and performance of a modified
conventional refrigerator to serve as a PV powered one. Renewable Energy 31, 771–780.
4. Kattakayam, Thomachan A., and Srinivasan, K., 2004. Lead acid batteries in solar
refrigeration systems. Renewable Energy, 29, 1243–1250
5. Cherif, A., and Dhouib, A., 2002. Dynamic modelling and simulation of a photovoltaic
plant. Journal of Renewable Energy. 26, 143–53.
6. Ewert, M. K., Agrella, M., De Monbrun, D., Frahm, J., Bergeron, D. J., Berchowitz, D., 1998.
Experimental evaluation of a solar PV refrigerator with thermoelectric, sterling, and vapour
compression heatpumps. In Proceedings of ASES Solar Conference, Albuquerque, USA.
7. Sidney, Shaji., Mohan Lal, D., 2016. Exergy Analysis of A Solar PV Driven DC Refrigerator
For Different Ambient Conditions» IEEE, International Conference on Energy-Efficient
Technologies for Sustainability (ICEETS), 280–284.
8. Ekren, Orhan and Celik, Serdar., 2013. Energetic and Exergetic Performance Evaluation of
AC and Solar Powered DC Compressor sustainability in Energy and Buildings, 22, 357–365.
9. Opoku, R., Anane, S., Seidu, R., 2016. Comparative techno-economic assessment of a
converted DC refrigerator and a conventional AC refrigerator both powered by solar PV
International Journal of Refrigeration, 72, 1–11.
10. Modi, Anish, Chaudhuri, Anirban, Vijay, Bhavesh, Mathur, Jyotirmay, 2009. Performance
analysis of a solar photovoltai coperated domestic refrigerator. Applied Energy, 86,
2583–2591
11. Eltawil, M. A., and Samuel, D. V. K., 2007. Performance and economic evaluation of a
solar photovoltaic powered cooling system for potatostorage. Agricultural Engineering
International the CIGR E-Journal. Manuscript EE, 7(8), vol. IX.
CHAPTER 30
Automatic Speed Control of

a Three Phase Induction Motor
by Variable Frequency Drive
Sayantan Chanda,1 Chayan Bandopadhyay,2 Maitreyee Banerjee,3
Shalini Mondol,4 and Ankur Bhattacharjee5
1
B. Tech Electrical Engineering 3rd Year, IEM Kolkata, India.
2
B. Tech Electrical Engineering, IEM Kolkata, India.
3
4
5
Assisstant Professor, Electrical & Electronics Dept., BITS Pilani, Hyderabad
ABSTRACT: This study demonstrates speed control of three-phase induction motor

using variable frequency drive (VFD). Out of the various methods of controlling Induction
motors, V/f Control has proven to be the most versatile. The overall scheme of implementing
V/f control has been presented. One of the basic requirements of this scheme is the PWM
inverter. In this, PWM inverters have been modelled and their outputs fed to the induction
motor drives. The uncontrolled transient and steady-state response of the induction motor
has been obtained and analysed. A MATLAB code was developed to successfully implement
open loop V/f control on a PWM-inverter fed three-phase induction motor, and the torque
was found to be constant for various rotor speeds. This was followed by a MATLAB model for
closed-loop V/f control on a PWM-inverter fed three-phase induction motor. It was observed
that using a closed-loop scheme with a PID controller gave a very superior way of controlling
the speed of an induction motor while maintaining a constant maximum torque and V/f ratio.
Keywords: VFD, PWM inverter, Bridge diode rectifiers, Voltage to Frequency ratio.
30.1 MOTIVATION
Motivation for designing a variable frequency drive (VFD) for speed control of induction
motor comes from the fact that there are so many VFDs are available in market which are
manufactured by different companies and are very expensive.
277
One of the main application of VFDs is in irrigation purpose. In irrigation application,

VFD helps farmers to control or vary the pressure of the flow according to their requirements.
So for farmers, it is very difficult to afford those expensive VFDs. If it is possible to design a
VFD in the affordable price for the farmers, it can be sold in the market as a product which
will be cheaper than the other VFDs available and will reduce the investment cost of the
farmers as well (Garud et al., 2016).
30.2 V/F SPEED CONTROL METHOD

Whenever three-phase supply is given to three-phase induction motor rotating magnetic
field is produced which rotates at synchronous speed given by
Ns=120.f / P (30.1)
In three-phase induction motor, emf is induced by induction similar to that of transformer
which is given by
V=4.44 K ǾTf (30.2)
where K is the winding constant, T is the number of turns per phase, and f is frequency.
Now if we change frequency, synchronous speed changes, but with decrease in frequency
flux will increase, and this change in value of flux causes saturation of rotor and stator cores
which will further cause increase in no load current of the motor. So it is important to
maintain flux, φ constant, and it is only possible if we change voltage, that is if we decrease
frequency flux increases, but at the same time, if we decrease, voltage flux will also decease
causing no change in flux and, hence, it remains constant. So here we are keeping the ratio
of V/f as constant. Hence, its name is V/f method. For controlling the speed of three-phase
induction motor by V/f method, we have to supply variable voltage and frequency which is
easily obtained by using converter and inverter set.
The various advantages of V/f control are as follows:
1. Large energy savings at lower speed.
2. Increased life of rotating components due to lower operating speed.
3. Reduced noise and vibration level.
30.3 VARIABLE FREQUENCY DRIVE

VFD helps to save energy in many applications. It is a power conversion device. It
converts a basic fixed frequency to a variable frequency and variable voltage output
(Garud et al., 2016).
The main parts of VFD are the converter, inverter, and controller. The converter
rectifies the AC input to DC, and the inverter converts the DC to an adjustable frequency
and adjustable voltage AC signal. Both must be adjustable to
Automatic Speed Control of a Three Phase Induction Motor  279
provide a constant volt to hertz ratio. A circuit filters the DC before it is sent to the
inverter. The controller regulates the output voltage and frequency based upon feed-back
signal from process. If the load is a pump, this is usually a pressure sensor. The modern
strategy for controlling the AC output of such a power electronic converters is the technique
known as pulse-width modulation (PWM), which varies the duty cycle of the converter
switches at a high-switching frequency to achieve a target average low-frequency output
voltage or current (Win, Sabai, and Nandar Maung, 2008).
FIGURE 30.1 Closed loop Simulink model of V/f speed control using VFD.
30.4 CLOSED LOOP SIMULATION OF V/F SPEED CONTROL OF THREE

PHASE INDUCTION MOTOR
The system is designed and simulated in discrete time domain with the sampling time of
the system being 0.00001 seconds. The motor in this model is a 1 HP, 415 V, 50 Hz three
phase squirrel cage induction motor with the parameters listed below:
TABLE 30.1 Three Phase Induction Motor Parameters
Parameters Parameter Values
Stator resistance 0.087
Stator inductance 0.8 e–3
Rotor resistance 0.228
Rotor inductance 0.8 e–3
Mutual inductance 34.7 e–3
Inertia 1.662
Frictional factor 0.1
No. of pole pairs 2
Torque constant calculation:

We know [Speed * Torque = Power] where speed = 1500, rpm = 157 rad/sec,
power = 1HP = 746 watt
Or [Torque] = [Power/Speed] =[746/157.07]=4.749N-m
Now, [Torque = k * speed2] where k = torque constant
Or [k] = [Torque/speed2] = [4.749/157.072] = 1.89*e-4
The VFD is made up of a rectifier block, an L-C filter block, DC-DC booster, inverter,
and a PWM generator which acts as a controller and controls the frequency by generating
gate pulses for inverter block for different system requirements. The error signal which is
generated by subtracting actual rotor speed from required rotor speed is passed through a
PID controller and then fed into the PWM generator. The PID controller is used to reduce
the noise and distortion in the error signal thus improving the rotor speed curve as shown
below.
30.4.1 Rectifier
TABLE 30.2 Diode Parameters
Parameters of Diode Parameter Values
Resistance 0.001
Inductance 0
Forward voltage 0.8
Snubber resistance 1000000
Snubber capacitance Infinite
FIGURE 30.2 Rectifier sub-block model.

30.4.2 L-C Filter

TABLE 30.3 L-C Filter Parameters
Inductance 0.1 mH
Capacitance 680 uF
FIGURE 30.3 L-C filter sub-block model.
30.4.3 DC-DC Booster

TABLE 30.4 Diode Parameters
Resistance 0.01
Inductance 0
Forward voltage 0.8
Snubber resistance 500
Snubber inductance 250e–9
FIGURE 30.4 DC-DC booster sub-block model.

TABLE 30.5 Pulse Generator Parameters

FET Resistance 0.001
Internal diode inductance 0
Internal diode resistance 0.01
Internal diode forward voltage 0
Snubber resistance 1e5
Snubber capacitance Infinite
TABLE 30.6 MOSFET Parameters

Pulse type Time Based
Time Simulation Time
Amplitude 1
Period 1/2000
Pulse width(%) 10
Phase delay 0
Due to a large number of inductive and resistive loads in series, the supply voltage drops. To
raise the voltage, a DC-DC booster is added to the circuit. The DC-DC booster raises the
inverter input voltage to 591 volts.
30.4.4 PWM Generator
FIGURE 30.5 Pulse width modulator sub-block model.

FIGURE 30.6 PWM generator block diagram.
The error signal from the PID controller is fed to the PWM generator, where the error
signal is produced with three sinusoidal waves of amplitude 0.5 and frequency of 50 Hz.
Then the resultant wave is superimposed with a Sawtooth wave of 20,000 Hz with the
sampling time being the same as that of the system. The PWM generator checks the points
of zero crossings, that is the points where the two waves cut and generate pulses which act
as the gate pulses of the IGBTs in inverter module.
30.4.5 Inverter
TABLE 30.7 IGBT Parameters
Internal resistance 1e-3
Snubber resistance 1e5
Snubber capacitance Inf
FIGURE 30.7 Inverter block diagram.

30.4.6 Simulation Result

30.4.6.1 Inverter Output Voltage Curve
FIGURE 30.8 Inverter output voltage waveform.
30.4.6.2 Rotor Speed Characteristics
FIGURE 3.9 Speed characteristics.
30.4.7 FFT Analysis and Speed Analysis of Three-Phase Induction Motor

TABLE 30.8 THD Analysis Values at Different Frequency
Frequency (Hz) % Distortion
0 (DC) 3269556.31%
25 8.86
50 (Fundamental) 100
75 5.25
100 (h2) 2.57

125 2.76
150 (h3) 3.32
175 1.92
200 (h4) 1.51
225 1.47
250 (h5) 2.54
275 1.20
300 (h6) 8.91
325 1.02
We have used the Fast Fourier transform tool in Simulink and done the total harmonic
distortion analysis at 1400 rpm. The total harmonic distortion of the system was 9.43%. The
parameters of the FFT analysis toolbox are given below.
TABLE 30.9 FFT Analysis Toolbox Parameters
Start Time 1.056
Number of cycles 2
Fundamental frequency 50
Maximum frequency 1000
Max frequency for THD computation Nyquist frequency
Display style List (relative to fundamental)
We notice the total harmonic distortion value is slightly above desired value. This can be
corrected by using more advanced and efficient filters than a normal L-C Filter which we
have used for analysis purposes. However, this is beyond the scope of this paper and can be
pursued in further papers.
30.5 CONCLUSION
The main purpose of this chapter is to verify that closed loop speed control of a three-phase
induction motor is indeed possible by implementing a PWM generator. The inverter output
voltage curve shows successful generation of three phase of voltage variable frequency. The
PWM generator can be realised by integrating the system with a microcontroller preferably
of low cost with a high PWM signal frequency.
The THD can be brought down to optimum value by designing and incorporating
more efficient and sophisticated filters instead of a regular L-C filter which has been used
in this case.
However, it can be concluded from the resulting curves and THD analysis that closed
loop speed control of three-phase induction motor is indeed possible by this system.
REFERENCES
1. IJIRSET 5(3), March 2016
2. Garud, Yogita Y., Gole, Sayali R., Jadhav, Rutuja T., Deoghare, Seema U. A Study on Variable
Frequency Drive and its Applications. International Journal of Innovative Research in
Science, Engineering and Technology
3. World Academy of Science, Engineering & Technology, 42, 2008.
4. Win, Thida, Sabai, Nang, and NandarMaung, Hnin, 2015. Analysis of Variable Frequency
Three Phase Induction Motor Drive. IJIERE, 2(3), 2015.
5. Julaha, Junaid and Tyagi, Priti., 2014. Speed Control of AC Motor using VFD. IJERD, 10(4),
April 2014, 94–103
6. Gandhi, Krupa, Mokariya, K. L., Karvat, Deepa., 2013. Simulation of PWM Inverter for
VFD application using MATLAB. International Journal of Science and Research (IJSR),
2(6), June 2013.
7. Tamal Aditya, Research to study Variable Frequency Drive and its Energy Savings.
CHAPTER 31
OWEC: Optimised Weighted

Ensemble Classifier Using
Particle Swarm Optimisation
Syed Fahim Karim1 and Ayan Kumar Panja2
1
fkarim6@gmail.com, 2ayanpanja1992@gmail.com
ABSTRACT: Machine learning is the process of teaching computers to learn from

data and think like humans. Ensemble learning is a process by which multiple machine
learning models can be combined strategically to solve a problem. In this subject, a new
approach is proposed to find the optimal weights for a weighted ensemble classifier.
The optimised weights can be found using an optimiser. Normally, grid search is used
to find the optimal weights for the ensemble model. This subject proposes a new
approach, Optimised weighted ensemble classifier (OWEC), to find the optimal weights
for a weighted ensemble classifier using particle swarm optimisation as an optimiser.
This approach is shown to be able to generalise to different datasets with different base
learners, reduce the bias and variance of the overall model and is computationally more
efficient to train compared to the base learners.
Keywords: Ensemble learning, Particle swarm optimisation, Optimised ensemble

learning.
31.1 INTRODUCTION
Particle swarm optimisation (PSO) is a population-based stochastic optimisation
technique that falls under the category of swarm intelligence. It was originally proposed
by Kennedy in 1995 and is inspired by social behaviour of bird flocking or fish schooling.
PSO can be used as a computationally inexpensive optimisation algorithm for various
problems. PSO has the advantage over other optimisation techniques under the swarm
intelligence category by converging to the global optima and not getting trapped in a
local optima. In PSO algorithm, a population of particles are generated randomly in the
287
search space. Each particle in the swarm has three components – position (P), velocity
component (V), and personal best (Pbest). Position keeps track of position of the particle.
Velocity component is the speed by which a particle moves to a new position. Personal
best is the best position reached by the particle at any point in time. Additionally, the
swarm maintains a component called global best which keeps track of the best position
attained by any particle at any point of time. At each iteration of the algorithm, we do the
following updates:
Vnew = W*Vold + c1 * r1 * ( Pbest - Pold ) + c2 * r2 * ( Gbest - Pold )
Pnew = Pold +Vnew
W ® Damping Constant
c1 ® Cognitive Coefficient
r1 ® Random Number
c2 ® Social Coefficient
r2 ® Random Number
Ensemble learning is the process of combining various machine learning models (called
base learners) to solve a particular task more efficiently than that was possible individually
by each base learner. Using ensemble learning, we can reduce the bias and variance of the
base models thereby generalising the overall model which gives better performance on
unseen data.
Weighted ensemble learning is an averaging based ensemble learning that combines the
base learners and applies some weights to the output of these models, then the maximum
of the weighted output is chosen. With the right weights applied, the ensemble model can
perform better than the base learners and is more generalised.
To find the optimal weights of the ensemble model, particle swarm optimisation
is used. PSO is primarily used to reduce an objective function. In this model, the
objective function is to reduce the number of misclassifications for a classification task.
PSO tries to find the optimal weights for the ensemble model such that the number of
misclassifications in the test dataset is reduced. The weights must sum up to 1 in case of
a classification problem.
31.2 BACKGROUND
Weighted ensemble learning models have been applied and extensively researched in
recent years. Weighted ensembles have also been studied extensively and a lot of literature
exists on this topic. Opitz et al. (1999) presented an empirical study on popular ensemble
methods, where they discuss about various literature on ensemble classifiers and how
ensemble methods with classifiers that disagree on their predictions can produce better
OWEC: Optimised Weighted Ensemble Classifier Using Particle  289
results and generalise better. Shen and Kong (2004) proposed a weighted ensemble of
neural networks for regression tasks where more weight is given to neural network with
higher training accuracy. Kim et al. (2019) proposed a weighted ensemble model using
the least squares method for regression task to estimate weights. In Shahhosseini et al.
(2019), a new approach is proposed called cross-validated optimal weighted ensembles
with internally tuned hyper-parameters (COWE-ITH) to find the optimal weights and
hyper-parameters of the model. This approach finds the optimal weights to combine the
base learners and tunes the hyper-parameters of the base learners to produce a better
performing ensemble model.
Ensemble models are placed first in many prestigious machine learning competitions
such as the Netflix Competition, KDD 2009, and CrowdFlower competition on Kaggle.
Ensemble methods have been successfully used in various areas of application. Martelli et
al. (2003) proposed an approach where an ensemble of neural networks was used to predict
membrane protein, achieving better results than previous models. In Alobaidi et al. (2019)
an ensemble framework is proposed that predicts seismic-induced liquefaction of soil. One
form of weighted ensemble learning is average ensemble learning. In average ensemble
learning, equal weights are given to the models and the average of base learners outputs is
taken. This approach may increase accuracy, but it is not a good approach as it disregards
the individual model’s accuracy. In many proposed weighted ensemble models, the optimal
weights are either calculated manually or through some searching technique. Manually
calculating the weights involves a lot of trial and error. In case of a classification task, all the
weights must sum up to 1 and manual calculation becomes tedious. Another way to find
the weights is to employ a searching technique like grid search or random search. These
search techniques are time-consuming, and in many cases cannot find the optimal weights,
settling for a lower-quality ensemble model. The time complexity for grid search increases
polynomially with the increase in the number of base learners. Thus, it becomes vital that
we find the optimal weights for the ensemble model and find these weights in the least
possible time.
This chapter proposes a new approach to find the optimal weights of the ensemble
model. In this approach, PSO is used to find the weights which increases the overall model
accuracy. This approach has been designed to be fast to minimise bias and variance error,
thereby improving overall model accuracy. PSO has proven to be a robust optimisation
technique that can find the optimal weights quickly for any number of base learners.
31.3 OPTIMISED WEIGHTED ENSEMBLE CLASSIFIER (OWEC)

Optimised weighted ensemble classifier (OWEC) is a new approach to find the optimal
weights for an ensemble classifier quickly. We test the ensemble model on two different
datasets and measure the performance, both in terms of accuracy and computation time.
We use two datasets – CIFAR 10 and Fashion MNIST. The base learners for CIFAR 10 are
convolutional neural networks and for Fashion MNIST are neural networks. We use three
base learners for both datasets.
31.3.1 Approach
First, the data is prepared for training the base learners. Both datasets (CIFAR 10 &
Fashion MNIST) have 50,000 training samples and 10,000 testing samples. Three models
are prepared for each dataset as shown in Table 31.1. For Fashion MNIST dataset, we use
three simple deep neural networks as base learners. For CIFAR-10 dataset, we use three
convolutional neural networks as base learners. We train the models on a single-core hyper-
threaded Intel Xeon processor.
Start
Prepare Data
Define Base Learners
Train Base Learners
Initialize PSO
Run PSO and

Find optimal weights
Apply weights and

Measure accuracy
End
FIGURE 31.1 Optimised weighted ensemble classifier flow chart.

The models are trained for 10 epochs with a batch size of 100. We use Adam as an
optimiser for the base learners and sparse categorical cross-entropy as a cost function.
The models are trained for 10 epochs for fast training which produces low-quality
base learners, thereby exemplifying the impact of the ensemble model on the overall
accuracy. The models are then evaluated on the test data, and we observe the accuracy
of the models.
TABLE 31.1 Model Architectures of the base Learners for Fashion MNIST and CIFAR 10
Models Fashion MNIST CIFAR 10
Conv, layer 1–32, 3x3

Conv. layer 2–64, 3x3
Model 1 Hidden layer 1–30 units
Dense layer–64 units

Hidden layer 1–30 units Conv. layer 2–64, 3x3
Model 2
Hidden layer 2–30 units Dense layer 1–64 units
Dense layer 2–64 units

Hidden layer 1–20 units
Model 3 Hidden layer 2–30 units
Hidden layer 3–35 units
Dense layer–50 units
A class is defined for each particle of the swarm in PSO. This class defines the variable
properties of each particle such as the position, velocity, cost, best position, and best cost.
Then the PSO is initialised with some parameters. The damping constant (W) is set at 0.5.
Both cognitive coefficient and social coefficient are set at 2. The number of iterations of
the PSO algorithm is set to 50, and the total population is set at 40. The minimum value
is set to 1, and the maximum value is set to 50. These values define the search space of the
problem. We keep the number of iterations low so that the algorithm takes less time, which
is a desired characteristic of the algorithm.
The swarm of particles is then initialised using the class defined before. Each particle of
the swarm is initialised randomly in the search space. The cost of each particle is calculated
using randomly sampled data from the test set. The PSO algorithm is the run, at each
iteration, the Eqs (31.1) and (31.2) are repeated for each particle. At each iteration, we
sample data randomly from the test set and calculate the cost of each particle, the global
cost and global position are updated accordingly. Finally, the global position is applied
as weights to the ensemble model and it is evaluated on the test set. The accuracy of the
ensemble model is measured and compared with that of each base learner.
31.3.2 Advantages
OWEC can find the optimal weights for the ensemble model in a reasonable amount of
time. The computation cost for the algorithm is low. Using this approach is better than
training a more complex model. More complex models are computationally expensive to
train, and there is always a possibility of over-fitting the data, thereby leading to a loss of
generalisation and poor performance on the test set. This approach is better than manual
estimation and average ensemble learning. Manually estimating weights is tedious especially
for classification task. Average ensemble learning disregards the accuracy of each model
leading to a relatively less accurate ensemble model. This approach also has an advantage
over grid search as it is faster and the increase in computation time with increase in number
of base learners is not polynomial.
31.3.3 Results
The accuracy of the base learners and ensemble model is compared along with the
computation time.
FIGURE 31.2 Training accuracy of the base learners on CIFAR-10 dataset.

TABLE 31.2 Training & Test Accuracy of PSO and the Ensemble Model on both the Datasets
Trial 1 Trial 2 Trial 3
Fashion Fashion Fashion
Cifar 10 Cifar 10 Cifar 10
MNIST MNIST MNIST
PSO accuracy 95.1% 84.4% 94.4% 82% 94.8% 81.6%
Model accuracy 87.68% 73.661% 87.78% 73.72% 87.723% 73.69%
Weights 0.33790 0.34016 0.32780 0.32698 0.3379 0.34722
0.32907 0.32238 0.31952
0.31122 0.31122 0.30022
0.33076 0.35062 0.32393
0.35086 0.36075 0.36186
Execution time 45.6 sec 456.70 sec 45.1 sec 467.28 sec 44.5 sec 461.49 sec
For CIFAR-10 dataset, Model 1 achieves 70.14%, Model 2 achieves 66.3%, and Model 3
achieves 69.35%. For training, Model 1 takes 560 seconds, Model 2 takes 578 seconds, and
Model 3 takes 568 seconds. Average accuracy of OWEC is 73.68% and average execution
time is 461.82 seconds.
For Fashion MNIST dataset, Model 1 achieves 85.37%, Model 2 achieves 85.73%,
and Model 3 achieves 85.91%. For training, Model 1 takes 49 seconds, Model 2 takes 51
seconds, and Model 3 takes 50 seconds. Average accuracy of OWEC is 87.727% and average
execution time is 45.06 seconds.
We can observe from Table 31.2 that OWEC achieves 3.54% more than the most
accurate base learner (Model 1) for CIFAR-10 dataset and 1.817% (Model 3) for Fashion
MNIST. OWEC also takes less time to find the optimal weights.
From the above results, we can conclude that using OWEC leads to better accuracy.
OWEC also takes less time compared to the training time of the base learners, giving it an
advantage over more complex models which will take more time to train. The ensemble
model can achieve a better trade-off between bias and variance while helps it generalise
more, thus it achieves better accuracy on the test set. Comparing our results with those
achieved in Ju, Bibaut, and van der Laan (2017), we have achieved a respectable increase in
accuracy with ensemble learning with low-quality base learners for CIFAR-10.
31.4 FUTURE SCOPE

We have carried out our computation on a CPU. If we use a GPU, the base learners benefit
a lot and the training time decrease drastically. For CIFAR-10 dataset, the average training
time of the base learners is 62 seconds, while PSO takes on an average, 83 seconds to
find the optimal weights. This is because that PSO is unable to take advantage of the
multiple cores present on a GPU unlike the base learners. However, the PSO algorithm
can be improved to take advantage of a GPU which will surely improve the performance
of OWEC on the whole.
The swarm, after initialisation, can be divided into sub-swarms. Each sub-swarm is
trained on a different part of the dataset, parallelly running on multiple cores. After a
number of iterations, each sub-swarm can exchange information (global cost and position)
among the particles. This will lead to a better performance of the ensemble model, and it
will be able to generalise more thereby improving the accuracy.
31.5 CONCLUSION
In this chapter, we have discussed a new approach to find the optimal weights of a weighted
ensemble model called OWEC. We have used PSO as an optimiser to find the optimal
weights. OWEC has been designed to achieve a respectable increase in accuracy in minimum
execution time. We have used two datasets for measuring the performance – CIFAR-10 and
Fashion MNIST. We used three models for each dataset.
We have achieved a respectable increase in accuracy with OWEC, 3.54% for CIFAR-10,
and 1.817% for Fashion MNIST. The execution time of OWEC was less than the training
time of the base learners. OWEC has proven to be a fast algorithm that can be used to
develop a more accurate model using a number of base learners. OWEC takes less time
to find the optimal weights than the training of the base learners which is advantageous
overtraining a more complex model. OWEC is more accurate than some other ensemble
learning techniques like average ensemble learning. Using PSO to optimise the weights
leads to better performance than using grid or random search. PSO also scales well with
an increase in number of base learners compared to grid search whose execution time
increases polynomially. Thus, OWEC is a computationally efficient algorithm that can find
the optimal weights for the ensemble model.
REFERENCES
1. Ju, Cheng, Bibaut, Aurélien, and van der Laan, Mark J., 2017. The Relative Performance of
Ensemble Methods with Deep Convolutional Neural Networks for Image Classification.
arXiv preprint arXiv: 1704.01664v1.
2. Kim, D., Yu, H., Lee, H., Beighley, E., Durand, M., Alsdorf, D. E., et al., 2019. Ensemble
learning regression for estimating river discharges using satellite altimetry data: Central
Congo River as a Test-bed. Remote sensing of environment, 221, 741–755
3. Martelli, P. L., Fariselli, P., and Casadio, R., 2003. An ENSEMBLE machine learning approach
for the prediction of all-alpha membrane proteins. Bioinformatics, 19(suppl_1), i205-i211
4. Alobaidi, Mohammad H., Meguid, Mohamed A., and Chebana, Fateh, 2019. Predicting
seismic-induced liquefaction through ensemble learning frameworks. Springer Nature, 9,
11786. URL https://www.nature.com/articles/s41598-019-48044-0.pdf?origin=ppub
5. Shahhosseini, Mohsen, Hu, Guiping, Pham, Hieu, 2019. Optimizing Ensemble Weights and
Hyperparameters of Machine Learning Models for Regression Problems. arXiv preprint
arXiv: 1908.05287
6. Shen, Z.-Q., and Kong, F.-S., 2004. Dynamically weighted ensemble neural networks for
regression problems. Paper presented at the Proceedings of 2004 International Conference
on Machine Learning and Cybernetics (IEEE Cat. No. 04EX826)
CHAPTER 32
Detection of Disaster Affected

Regions Using Drone-assisted
Deep Learning Model
Nikhil Singh, Amitosh Banerjee, Swadhin Bhowmick,
Sneha Chakraborty, and Amartya Mukherjee
Institute of Engineering and Management, Kolkata, India.
ABSTRACT: Monitoring of disasters is crucial for mitigating their effects on the

environment and human population, and can be facilitated by the use of unmanned
aerial vehicles (UAV), equipped with camera sensors that produce aerial photos of the
areas of interest. A modern technique for recognition of events based on aerial photos is
deep learning. In this chapter, we present the state of the art work related to the use of
deep learning techniques for disaster identification. We demonstrate the potential of this
technique in identifying disasters with high accuracy, by means of a relatively simple deep
learning model. Based on a dataset of 1988 images (containing disaster images such as fires,
earthquakes, collapsed buildings, tsunami, and flooding, as well as ‘non-disaster’ scenes),
our results show an accuracy of 89% achieved, indicating that deep learning, combined
with UAV equipped with camera sensors, have the potential to predict disasters with high
accuracy.
Keywords: Deep learning, Disaster monitoring, Unmanned aerial vehicles, Drones,

Image processing.
32.1 INTRODUCTION
Climate change and global warming lead to an increase of hydro-meteorological
disasters around the world. According to the International Disaster Database
295
(EM-DAT) (Centre for Research on the Epidemiology of Disasters, 2017), there is a

large increase of natural and/or technological disasters in the last two decades, and this
increase is expected to continue. Although preparedness and prevention of disasters
are processes far more important than dealing with them after happening, some are
inevitable and cannot be easily predicted. Monitoring of possible disasters and fast
identification of their occurrence are crucial for mitigating their effects on the physical
environment or to humans. Fast alerting and immediate response are important aspects
for dealing effectively with them by means of appropriate disaster management. Disaster
monitoring can be facilitated by the use of unmanned aerial vehicles (UAV) such as
drones (Figure 32.1), which are becoming more and more popular, having various
benefits such as small size, low cost of operation, exposure to dangerous environments,
and high probability of mission success without the risk of loss of aircrew resource.
These and other benefits are explained in related work, either independently (Petrides
et al., 2017; Zhang and Kovacs, 2012), or in relation to other remote-sensing techniques
such as airplanes and satellites (Matese et al., 2015). UAV have competed monitoring
with traditional acquisition platforms (i.e. satellite and aircraft) due to low operational
costs, high operational flexibility, and high spatial resolution of imagery. UAV has been
successfully used in scenarios of detecting earthquake-triggered roof holes (Li et al.,
2015), oil spills and flooding (Luo et al., 2015), victims (Andriluka et al., 2010), etc.
Drones are usually equipped with camera sensors, which could be optical cameras
operating over the visible light spectrum, infrared to operate in low-light conditions
(by detecting emitted heat) or hyper-spectral ones, considering particular spectra for
events of interest. Images collected by the sensors can then be processed locally by an
on-board computer system or sent to control centres for remote processing. In both
cases of local and remote processing, automatic identification of events of interest
(i.e. potential disasters) are important for sensing the disaster while it is happening,
taking immediate life-saving actions, either by governmental organizations or the
citizens themselves (Kamilaris and Pitsillides, 2014). A modern, promising technique
for image recognition is deep learning or convolutional neural networks (CNN).
Deep learning allows computational models composed of multiple processing layers
to learn representations of data with multiple abstraction levels. Krizhevsky et al.
demonstrated the potential of CNN in the ImageNet large-scale visual recognition
challenge (ILSVRC) in 2012 (Krizhevsky, Sutskever, and Hinton, 2012). Deep learning
seems to have better performance in image recognition tasks (Liu and Wu, 2016), in
comparison to common techniques used for analysing images such as scalable vector
machines (SVM) and artificial neural networks (ANN) (Saxena and Armstrong, 2014).
Thus, drones equipped with camera sensors using deep learning models to identify
various (possibly disastrous) events constitute a recent practice with high potential,
as the identification is fast (i.e. in seconds after the model has learned the problem),
flexible (i.e. many events can be identified at once), and scalable (i.e. learning process
is continuous and adaptable).
Detection of Disaster Affected Regions Using Drone-assisted Deep Learning Model  297
FIGURE 32.1 Layered ecosystem of the drone-assisted disaster monitoring system.
32.2 METHODOLOGY AND ALGORITHM

As a deep learning model, we used the VGG16 architecture (Simonyan and Zisserman,
2014), pre-trained with the ImageNet2 image dataset (Krizhevsky, Sutskever, and Hinton,
2012). VGG constitutes one of many successful architectures, which researchers may use
to start building their models instead of starting from scratch. Other popular ones include
AlexNet (Krizhevsky, Sutskever, and Hinton, 2012), GoogleNet (Szegedy et al., 2015), and
Inception-ResNet (Szegedy et al., 2017). Each architecture has different advantages and
scenarios, where it is more appropriate to be used (Canziani, Paszke, and Culurciello, 2016).
We selected VGG as it is easy to use, being automatically integrated into the TensorFlow
open-source software library for machine intelligence as a Python class (TensorFlow, 2017).
FIGURE 32.2 The architecture of VGG.

The pre-processing step is very crucial for making the model efficient so that model has
good detection rate (Figure 32.2). In the pre-processing step, we need to cut the 1988
images into smaller patches each of size 128 × 128. After the preparation of patches is over,
we need to label those patches for efficient training. Labelling of the image patches is one
of the important steps before we proceed to the process of training our model. Here based
upon the ground truth images, we label the image patches as disaster hit or not hit. This is
important as CNN needs to learn the features of the disaster image patches. We have used
the concept of finding out the percentage of the white pixels from the ground truth patches,
and if the percentage is more than 10%, then we label the corresponding true image as 1
(disaster hit) or else if it less than 10%, we label it as 0 (not disaster hit). The 10% which we
used here is internationally acceptable standard. We have used the openCV and NumPy
packages for the labelling of the images.
We aimed to ‘wire’ the VGG-16 architecture (i.e. the pre-trained convolutional layers)
with a dense layer, trained with a dataset of aerial photos containing various natural
disasters. As such dataset is not publicly available (to our knowledge), we decided to
develop our own, by searching Google Images for relevant images and downloaded
necessary datasets which contained parts required for our dataset. We decided to map
together images from earthquakes and hurricanes showing collapsed buildings, as well
as images from hurricanes, earthquakes, or tsunami showing flood. For fire, we selected
images showing active flames or smoke. Finally, we selected normal images of various
[Landscapes] showing aerial views of cities, villages, forests, and rivers, to examine if
and how the deep learning model can predict disastrous events. The different groups of
images are listed in Table 32.1.
TABLE 32.1 Image Class and its Possible Outcome
No. Image Class No. of Images Possible Outcome
1 Buildings collapsed 397 Earthquake
2 Flames and smoke 490 Fire
3 Flood 402 Flood
4 Villages 415 No disaster
5 Forests, rivers, and Water bodies 284 No disaster
Based on this dataset, we trained our deep learning architecture to classify automatically
aerial photos according to the image group (i.e. disaster category or landscape) where they
belonged. To increase our dataset, we employed data augmentation techniques (Krizhevsky,
Sutskever, and Hinton, 2012) to enlarge artificially the number of training images using
label-preserving transformations, such as translations, transposing, and reflections, and
altering the intensities of the RGB channels. In this way, at every run of our training
procedure (i.e. epoch), each image from the training dataset was randomly transformed
before used as input to the model.
In this section, we describe our findings by experimenting with our VGG architecture.
We used 82% (1630 images) of our dataset as training data and 18% (358 images) as testing
data, dividing our dataset randomly in training and testing. We note that the combination
80–20 (i.e. for splitting training and testing data) showed the best results in our experiments.
Besides, it is the combination highly preferred in the literature too (Deng and Yu, 2014),
(Wan et al., 2014). We also note that 89% precision was obtained after a total of 20 epochs,
as shown in Figure 32.3. After this number of epochs, the precision was slightly decreased
up to 85% at the 30th epoch. Hence, we stopped the learning procedure at the 20th epoch
of the training. By epochs here, we refer to each complete run of the algorithm during
training, using the whole of the training dataset each time. Training vs. validation can be
shown corresponding to model loss has been depicted in Figure 32.4.
FIGURE 32.3 Accuracy vs epochs relation.
FIGURE 32.4 Loss vs epochs relation.

The training procedure required 20 minutes on a Windows OS (Intel Core i7, 8 GB RAM),
while the testing procedure needed less than 5 minutes for all the 100 images used for
testing. It is reported that the deep learning architecture managed to learn the problem
relatively correctly, although it confused flooding events with urban landscapes that
involved lakes and rivers or collapsed buildings (3% of the 9% total error) as well as buildings
collapsed with urban landscapes (4% of the 9% total error), and vice versa. Examples of this
‘confusion’ are listed in Figure 32.5. An intuitive approach to reduce the error could be to
avoid during training images of cities with rivers or lakes, or surfaces that reflect the sun,
especially during sunset.
FIGURE 32.5 Examples of wrong predictions. Buildings collapsed instead of urban landscape
(top-left). Fire instead of urban landscape (top-right). Buildings collapsed instead of flooding
(bottom-left). Urban landscape instead of flooding (bottom-right).
32.3 DISCUSSION
Analysis results in Section 2 have indicated the high accuracy obtained in disaster
recognition by means of aerial photos, which could have been collected by means of
cameras embedded in drones or other UAV. Except from the relatively high accuracy,
there are some other important advantages of using deep learning in image processing.
Previously, traditional approaches for image classification tasks had been based on hand-
engineered features, whose performance affected heavily the overall results. Feature
engineering (FE) is a complex, time-consuming process which needs to be altered
whenever the problem or the dataset changes. Thus, FE constitutes an expensive effort
that depends on experts’ knowledge and does not generalize well (Amara, Bouaziz,
and Algergawy, 2017). On the other hand, deep learning does not require FE, locating
the important features itself through training. It generalizes well (Pan & Yang, 2010),
and it is quite robust even under challenging conditions such as illumination, complex
background, different resolution, size, and orientation of the images (Amara, Bouaziz,
and Algergawy, 2017). Even though it takes sometimes considerable time to learn the
problem, after it does, it’s testing time efficiency is much faster than other methods like
SVM or KNN, even when considering hyper-spectral cameras/images (Chen et al., 2014).
Its main disadvantages are that it takes (sometimes much) longer time to train, and it
requires the preparation and pre-labelling of a large dataset, which would serve as the
input during the training procedure. In spite of data augmentation techniques which
augment some dataset with label-preserving transformations, in reality, at least some
hundreds of images are required, depending on the complexity of the problem under
study (i.e. number of classes, precision required, etc.). A problem with some datasets
is the low variation among different classes (e.g. urban rivers/lakes vs. flooding) or the
existence of noise in the form of low resolution, inaccuracy of sensory equipment, high
sun/lights reflection in the image, and others. Finally, although deep learning does not
require FE since data annotation is a necessary operation, some tasks are more complex
and there is a need for experts (who might be difficult to involve) in order to annotate
input images. In the case of disasters, this is easier to do even by non-experts, but in
general, this is an important consideration.
32.4 CONCLUSION
In this chapter, we have presented the state of the artwork related to the use of deep learning
techniques for disaster monitoring and identification, based on aerial photos captured
by UAV. We have also created a small dataset of 1988 images, creating a deep learning
model based on VGG, in order to show the potential of this technique in identifying
disasters automatically and with high accuracy. Based on our small dataset, by employing
data augmentation techniques, accuracy has reached 89%. However, the authors expect
that a larger dataset would effectively reduce the error (possibly reaching more than
95% accuracy) and that these results indicate that the CNN architecture used (and deep
learning in general) has the potential to predict disasters with high accuracy in the near
future. Challenges in the use of UAV for disaster monitoring still exist, such as their limited
computational capability, low energy resources, and regulation issues for flight allowance,
which hinders real-time data processing, area coverage and flexible use. Nevertheless, the
opportunities combining UAV with deep learning techniques are large, as deep learning can
provide high-accuracy event identification in real-time without requiring much processing
capacity. Finally, UAV with deep learning could enable better disaster modelling, especially
when combined with geo-tagging of the events identified and geospatial applications. This
would facilitate the integration of relevant actors (i.e. action forces/authorities, citizens/
volunteers, other stakeholders) in disaster.
REFERENCES
1. Amara, J., Bouaziz, B., and Algergawy, A., 2017. A Deep Learning-based Approach for
Banana Leaf Diseases Classification. Stuttgart: BTW workshop.
2. Amit, S., Shiraishi, S., Inoshita, T., and Aoki, Y., 2016. Analysis of satellite images for disaster
detection. Geoscience and Remote Sensing Symposium (IGARSS). IEEE.
3. Andriluka, M., Schnitzspan, P., Meyer, J., Kohlbrecher, S., Petersen, K., Von Stryk, O.,
Schiele, B., 2010. Vision based victim detection from unmanned aerial vehicles. IEEE/RSJ
International Conference on Intelligent Robots and Systems (IROS).
4. Bejiga, M., Zeggada, A., Nouffidj, A., and Melgani, F., 2017. A convolutional neural network
approach for assisting avalanche search and rescue operations with UAV imagery.
5. Canziani, A., Paszke, A., and Culurciello, E., 2016. An Analysis of Deep Neural Network
Models for Practical Applications. Centre for Research on the Epidemiology of Disasters.
April 2017.
6. Chen, Y., Lin, Z., Zhao, X., Wang, G., and Gu, Y., 2014. Deep Learning-Based Classification
of Hyperspectral Data. IEEE Journal of Selected topics in applied earth observations and
remote sensing.
7. Deng, L., and Yu, D., 2014. Deep learning: methods and applications. Foundations and
Trends in Signal Processing.
8. Kamilaris, A., and Pitsillides, A., 2014. The impact of remote sensing on the everyday lives
of mobile users in urban areas. Seventh International Conference on Mobile Computing
and Ubiquitous Networking (ICMU). Singapore: IEEE.
9. Kang, B., and Choo, H., 2016. A deep-learning-based emergency alert system. ICT Express.
10. Kim, S., Lee, W., Park, Y., Lee, H., and Lee, Y., 2016. Forest fire monitoring system based
on aerial image. 3rd International Conference on Information and Communication
Technologies for Disaster Management (ICT-DM). IEEE.
11. Krizhevsky, A., Sutskever, I., and Hinton, G. E., 2012. Imagenet classification with deep
convolutional neural networks. Advances in neural information processing systems.
12. LeCun, Y., Bengio, Y., and Hinton, G., 2015. Deep learning. Nature, 521(7553).
13. Liu, Y., and Wu, L., 2016. Geological Disaster Recognition on Optical Remote Sensing
Images Using Deep Learning. Procedia Computer Science.
14. Pan, S. J., and Yang, Q., 2010. A survey on transfer learning. IEEE Transactions on knowledge
and data engineering.
15. Paton, D., 2003. Disaster preparedness: a social-cognitive perspective. Disaster Prevention
and Management: An International Journal.
16. Saxena, L., and Armstrong, L., 2014. A survey of image processing techniques for agriculture.
Perth, Australia: Proceedings of Asian Federation for Information
17. Kamilaris, Andreas, and Prenafeta-Boldú, Francesc X., 2018. Disaster Monitoring using
Unmanned Aerial Vehicles and Deep Learning, Institute for Food and Agricultural Research
and Technology (IRTA) Barcelona, Spain.
18. Lettieri, E., Masella, C., and Radaelli, G., 2009. Disaster management: findings from
asystematic review. Disaster Prevention and Management: An InternationalJournal, 18(2),
117–136.
19. Li, S., Tang, H., He, S., Shu, Y., Mao, T., Li, J., and Xu, Z., 2015. Unsupervised detectionof
earthquake-triggered roof-holes from UAV images using joint color andshape features.
IEEE Geoscience and Remote Sensing Letters, 12(9), 1823–1827.
20. Luo, C., Nightingale, J., Asemota, E., and Grecos, C., 2015. A UAV-cloud system fordisaster
sensing applications. IEEE Vehicular Technology Conference (VTCSpring) (pp. 1–5).
21. Matese, A., Toscano, P., Di Gennaro, S., Genesio, L., Vaccari, F., Primicerio, J., and Gioli, B.,
2015. Intercomparison of UAV, aircraft and satellite remote sensingplatforms for precision
viticulture. Remote Sensing, 7(3), 2971–2990.
22. Petrides, P., Kolios, P., Kyrkou, C., Theocharides, T., and Panayiotou, C., 2017.
DisasterPrevention and Emergency Response Using Unmanned Aerial Systems. In Smart
Cities in the Mediterranean (pp. 379–403). Springer International Publishing.
23. Shaluf, I. M., Ahmadun, F. L., and Mat Said, A., 2003. A review of disaster and crisis. Disaster
Prevention and Management: An International Journal, 12(1), 24–32.
24. Simonyan, K., and Zisserman, A., 2014. Very deep convolutional networks for large-
scaleimage recognition. arXiv preprint arXiv: 1409.1556.
25. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., and Rabinovich, A., 2015.
Going deeper with convolutions (pp. 1–9). Boston, MA, USA: IEEEconference on computer
vision and pattern recognition.
26. TensorFlow, 2017. An open-source software library for Machine Intelligence. Retrieved
from https://www.tensorflow.org/
27. Wan, J., Wang, D., Hoi, S. C., Wu, P., Zhu, J., Zhang, Y., and Li, J., 2014. Deep learningfor
content-based image retrieval: A comprehensive study (pp. 157–166).
28. Orlando, F. L., Proceedings of the 22nd ACM international conference onMultimedia, ACM.
29. Yalcin, H., 2017. Plant phenology recognition using deep learning: Deep-Pheno. 6th
International Conference on Agro-Geoinformatics. Fairfax VA, USA.
30. Zeggada, A., Melgani, F., and Bazi, Y., 2017. A Deep Learning Approach to UAV Image
Multilabeling. IEEE Geoscience and Remote Sensing Letters.
CHAPTER 33
Enhancing Massive Open Online

Courses with Brain Computing
Interface and Artificial
Intelligence
Sayantan Pal,1 Saptarshi Neogi,2 Sayanti Dutta,3 Sumalyo Datta,4
Nilanjana Bandyopadhyay5, and Yashaswi Raj6
Electrical Engineering, Institute of Engineering & Management, India.
1,6
Electronics and Communication Engineering, Institute of Engineering & Management, India.

2,3,4,5
1
sayantan639@gmail.com, 2saptarshineogi123@gmail.com, 3sayantidutta.dgp@gmail.com,
4
datudutta@gmail.com, 5nilanjanabandyopadhyay13@gmail.com, 6k.yashaswi.raj@iemcal.com
ABSTRACT: The chapter quantitatively measurers the effects of Massive Open Online
Courses (MOOCs) on Indian students, analyses the results, and suggests a possible solution
to the problem using Brain Computing Interface and Artificial Intelligence Predictive
Models to adapt the course of study for the students and identify the key problems in this
field. This is inspired from various studies on the effectiveness of online education and
public datasets available in this domain.
Keywords: Brain Computing Interface (BCI), Artificial Intelligence (AI), Massive Open
Online Education (MOOE), Confusion, Feedback, Electroencephalography (EEG).
33.1 INTRODUCTION
Massive Open Online Courses (MOOCs) today form the backbone of any technical
educational structure and is an integral part of both industry-specific areas as well as
modern-day research domains. It is quite natural that the best talent in the world can’t be
found in every institution, while the world today demands a standard of education that
is both universal in academics and skill. Thus, online education has become a necessity
305
in these areas. But many educators have argued against this unorthodox method stating
that this method is responsible for creating achievement gaps and racial boundaries in
education, and this one-way video feed system without the teacher having any feedback
from students during the lesson does not recreate the classroom ambience which
the student is most suited for (Mansureh Kebritchi, Angie Lipschuetz et al). Studies
conducted in Language Technologies Institute School of Computer Science-Carnegie
Mellon University using EEG (Gordillo, Galindo, Medina, 2015) on students of various
ethnic groups and intellectual categories show that there might be some truth to this
statement; and that is what inspired us to carry out the same experiment in India among
its indigenous ethnic groups to analyse the results and suggest a possible solution to
the problem using EEG in closed-loop feedback system with an AI Backend and thus
adaptively model the course for the student or a group of students in real-time and also
improve standards holistically in the future. The EEG signal is a voltage signal that can
be measured on the surface of the scalp, arising from large areas of coordinated neural
activity manifested as synchronization (group of neurons firing at the same rate). This
neural activity varies as a function of development, mental state, and cognitive activity,
and the EEG signal can measurably detect such variation. Rhythmic fluctuations in the
EEG signal occur within several particular frequency bands, and the relative level of
activity within each frequency band has been associated with brain states such as focused
attentional processing, engagement, and frustration (Wang et al., 2009), which in turn are
important for prediction of learning (Thompson, 1990). The recent availability of simple,
low-cost, portable EEG monitoring devices now makes it feasible to take this technology
from the lab into schools. The NeuroSky “MindSet,” for example, is an audio headset
equipped with a single-channel EEG sensor (NeuroSky, Brain wave signal (EEG), 2009,
Neurosky, Inc.). It measures the voltage between an electrode that rests on the forehead
and electrodes in contact with the ear. This Single Channel recording of data is precisely
able to indicate various mind-states of the subject.
The course has multiple video feed designed instead of a single video feed for a lesson for
different learners which vary in pace, difficulty of concepts discussed, and the sophistication
of language used for explanation, etc. These parameters would be best decided by the
MOOC providers, and the scope used in the system would broaden as more and more data
is collected in these areas and experimental results are analysed by the algorithm. It would
also contain questions of various levels of difficulty which would appear to the students
as per their analysis. All this while it is verified that the tougher questions are given more
weightage score and merit does not get ignored in the system. This approach is best suited
for a recommendation basis for the student but in a way surely that learning of the subject
is not hampered.
33.2 METHODOLOGY
Mr. Haohan Wang performed a test on ten students across different regions. He and his
team collected EEG signals of those students while watching MOOCs. A single channel
Enhancing Massive Open Online Courses with Brain Computing Interface  307
wireless Mindset was worn by the students for measuring the activity signals of the brain.
The videos were of different difficulty levels so that the confusion of the student can be
detected. On that datasheet, many machine learning algorithms were applied.
Similarly, we had performed a test on a few numbers of students. For the video reference
some SWAYAM (or Study Webs of Active Learning for Young Aspiring Minds) video
lectures were used. During that test, Neurosky helmet was used for collecting the data.
Students from various ethnic backgrounds speaking various dialects and languages all of
whom were versatile in English participated in it. Videos on various subjects of engineering
from SWAYAM were shown to them on several new concepts most of which none of the
candidates had encountered before. After each video, the student was told to give manual
feedback of the lesson grading various parameters on a scale of 1 to 10. For each student, the
single channel feed was deconstructed for confusion level interfacing (Gordillo, Galindo,
Medina, 2015) analysed using the following techniques.
FIGURE 33.1 Raw output.
The recorded sample (Figure 33.1) was first cleaned using a kernel from Kaggle using Erik
Walzacks filter for EEG Signals by sample mean filtering and scaling the components of the
signals correctly to be used for further analysis. Interestingly the average of simple sampled
data reveals quite a lot about brain activity during a time. Hence, it could be assumed that
this is some realizable linear function of the input which would help us investigate the
dataset totally with the given real-time feedback of the students to prove effective in cross-
validation.
FIGURE 33.2 Filtered output used for analysis.
Figure 33.2 shows how the signals would look after cleansing. Now the data is loaded in
Python using NumPy and pandas library, and relevant data structures are used to make
training and testing sets. These are trained against the scales of confusion provided by
each student at the end of each video. Clearly, there was a need for a suitable supervised
machine learning model to handle the data and simple gradient based classifiers could
not get better than an accuracy of about 0.598. Thus, came the need of using XGBoostan
open-source software library which provides a gradient boosting framework for Python
for analysis in higher-order gradient cost functions to codify a Linear in Time Polynomial
Approximation Model for the Academic as well as the Demographic datasets. For each of
the videos a predefined label was initialized by us based on the arguments of the chapter
which claimed Online Education to be biased towards a few; which thereby indicated
whether the student was confused or not during the course based on their ethnicity and
academic exposure just in Wang’s Experiments. The results were found in alignment with
Wang’s arguments making it clear that the problem was both deep-rooted within the
system and also widespread.
FIGURE 33.3 Data collection and feedback process.
However as seen in Figure 33.3, for the deployment of this model, we propose the use
of Cloud-based solutions for handling the dynamic data in real-time using a storage-as-
service node in the cloud for storage of data and analysis results and a software-as-service
node for organization of data for analysis using custom Big Data Management tools like
Big Query in Google Cloud Console. Also note that while researching due to limited
resources at our disposal during prototyping, we only conducted primary analysis using
Python-based open-source modules during the experiment, fully understanding that
during a real-time deployment of the model in cloud, these may prove to be redundant or
processor extensive as compared to custom Ml Data Handling and Analysis tools custom-
designed for the cloud, and we fully encourage the use of these, say Amazon SageMaker
(in AWS) whenever they are deemed fit for analysis of BCI data in accordance with our
proposed algorithm or any other which aims to measure the parameters we require to
calculate the level of understanding the student has of the current academic material he
or she is subjected to.
A possible solution to this problem could be to use study materials in course which
can be recommended to each one in accordance to their region, ethnicity, and comfort
in the subject. We should in no way discourage students to challenge their abilities, but
machine learning algorithms can surely identify the inequality of merit among students
and aid them to grow accordingly. As these courses mostly have subjects with practical
applications in the future and with direct ramifications of the society and industry, they
would be a part of, the student must not miss out on learning the subject in the zeal of
competition. Thus, with the help of Brain Computing using EEG Signals, we can aid the
system to adapt to the child in closed loop, where it would with the help of predictive
modelling (as shown in Figure 33.4) of the learning curve provide or recommend
appropriate teaching materials and exposure to the student analysing its subconscious.
Teachers also for a long term can assess their course to improve the method of explanation
used in that lecture.
FIGURE 33.4 Analysis of data and deployment of the proposed solution.
33.3 CONCLUSION
We can make the MOOCS platform uniform among the students of different demographic
regions by analysing their understanding ability through the use of BCI technology. It will
be quite beneficial since the toughness level of the assignment given to the students will
be determined by brain activity. If a student cannot understand a video properly, then that
particular video will be recommended to the student to watch once more. Beside this easier
question will be shown to the student first for that particular video so that they can build
up confidence as well as interest. Most importantly feedback can be provided to the course
instructor with the help of Machine Learning algorithm for further improvisation of the
course. Thus, it will help both the learner and instructor to understand each other in a
better way and improve themselves.
REFERENCES
1. Kebritchi, Mansureh, Lipschuetz, Angie, Santiague, Lilia. Issues and Challenges for Teaching
Successful Online Courses in Higher Education: A Literature Review.
2. Gordillo, L. F. M. A. R. M., Galindo, M. C. M. V., and Medina, M. A. E., 2015. Design and
construction of a brain-computer interface for application in neuro-robotics. In Jaffray, D.
(eds) World Congress on Medical Physics and Biomedical Engineering, June 7–12, 2015,
Toronto Canada. IFMBE Proceedings, vol. 51. Springer, Cham. Available from: https://
www.kaggle. com/wanghaohan/confused-eeg/version/6
3. Wang, Haohan, Li, Yiwei, Hu, Xiaobo, Meng, Zhu, Chang, Kai-min. Using EEG to Improve
Massive Open Online CoursesFeedback Interaction NeuroSky, Brain wave signal (EEG),
2009, Neurosky, Inc.
4. Thompson, G., 1990. How Can Correspondence-Based Distance Education be Improved?

A Survey of Attitudes of Students Who Are Not Well Disposed toward Correspondence
Study. The Journal of Distance Education, 5(1), 53–65
5. Niedermeyer, E., Lopes da Silva, Fernando H., 2005. Electroencephalography: basic
principles, clinical applications, and related fields, Lippincott Williams & Wilkins.
6. Marosi, E., et al., 2002. Narrow-band spectral measurements of EEG during emotional
tasks. International Journal of Neuroscience, 112(7), 871–891. Available from: https://www.
mathworks.com
7. Lutsyuk, N.V., Éismont, E.V., and Pavlenko, V. B., 2006. Correlation of the characteristics of
EEG potentials with the indices of attention in 12- to 13-year-old children. Neurophysiology,
38(3), p. 209–216.
8. Berka, C., et al., 2007. EEG correlates of task engagement and mental workload in vigilance,
learning, and memory tasks. Aviation, Space, and Environmental Medicine, 2007. 78
(Suppl. 1): p. B231–244.
9. Baker, R., et al., 2010. Better to be frustrated than bored: The incidence, persistence and
impact of learners’ cognitive -affective states during interactions with three different
computer-based learning environments. International Journal of Human-Computer
Studies. 68(4), 223–241.
CHAPTER 34
Application of Multi-Objective
Particle Swarm Optimization
Technique for Analytical
Solution of Economic and
Environmental Dispatch
S. Roy,1 D. Laha,2 A. Das,3 S. Chatterjee,4 M. Biswas,5 R. K. Mandal,6
and B. K. Ghosh7
1, 2, 3
Department of Electrical Engineering, Camellia Institute of Technology and Management, Boinchi,
West Bengal, India.
4
Cognizant Technology Solutions India Pvt. Ltd., Kolkata, India.
5, 6, 7
Department of Electrical Engineering, Techno International New Town, Kolkata, India.
ABSTRACT: A multi-objective particle swarm optimization (MOPSO) technique has

been employed in this chapter to solve economic scheduling, environmental scheduling,
and both economic and environmental dispatch (EED) problem considering both
economic and environmental issues. Different types of particle swarm optimization (PSO)
techniques have been employed, and a comparison study has been done. This chapter also
proposes an analytical strategy using PSO techniques which gives better outcome than
other techniques. A four-bus system with six generators of three plants has been used in
this study considering transmission power loss. The results show that among different PSO
techniques, hierarchical particle swarm optimisation (HPSO) gives the best result.
Keywords: Multi-Objective Particle Swarm Optimization (MOPSO), Economic and

Environmental Dispatch (EED), Hierarchical Particle Swarm Optimization (HPSO)
Technique, Analytical Solution, Economic and Environmental Dispatch.
313
34.1 INTRODUCTION
Operating at optimal cost was the only criterion for economic dispatch (ED) of thermal
power generation when the environmental pollution was not significant matter of concern.
Releasing of CO2, SO2, NOx , etc., into the atmosphere is increasing the rate of emission
which becomes significant in the present scenario of atmosphere. As a result pollution
minimisation becomes another important criteria for the load dispatch of thermal power
generation to maintain clean air. Hence, the environmental economic dispatch (EED)
problem becomes a large-scale, non-linear problem which has multi-objective, that is, to
optimise both the emission and the cost which are conflicting in nature and could not
be minimised simultaneously. There are various kinds of soft computing techniques that
use inexact solutions to solve non-linear problems for which an exact solution could not
be derived in polynomial time. These algorithms are based on biological phenomenon
of Mother Nature to optimised specific constraints of multi-objective function. The
contributions of this study are given below:
• Proposing short-term economic, environmental, and combined economic and
environmental scheduling of a four-bus test system including three plants and six
thermal generating units and considering power loss using hierarchical particle
swarm optimisation technique.
• Comparing the results obtained from basic PSO and hierarchical PSO and showing
the superiority of result obtained from hierarchical PSO than basic PSO.
• Comparing the result of proposed method with previous method.
34.2 PROBLEM STATEMENT

Cost minimisation and emission minimisation are two objectives in this problem, where
both are conflicting in nature. The aim is to determine optimal load dispatch with optimal
cost and emission. There are three-case analyses which are given in below:
Case 1: Cost is the only objective function to minimise, that is weightage factor is only
given to cost and corresponding emission is being calculated.
Case 2: Emission is the only objective function to minimise, that is, weightage factor is only
given to emission and corresponding cost is being calculated.
Case 3: Here both cost and emission have been minimised equally.
F1 = å(a p
i =G
i
2
i + bi pi + ci ) (34.1)
F2 = å(d p
i =G
i
2
i + ei pi + fi ) (34.2)
Application of Multi-Objective Particle Swarm Optimization Technique  315
In this case study, F1 is the cost function, and F2 is the emission function, and both objective
functions are quadratic functions of generated power of thermal generating units, where
the G denotes thermal generation unit. The test system is given below in Figure 34.1.
FIGURE 34.1 A four-bus test system for case study.
According Figure 34.1, there are three plants – Plant 1, Plant 2, and Plant 3. Plant 1
comprises of three generators G1, G2, and G3. Plant 2 comprises of two generators G4
and G5, and Plant 3 has only generator G6. These are interconnected with four-bus
system. The transmission power loss has been considered here which is denoted by Ploss.
It is a function of power output by generating units and B loss coefficients, and it can be
stated as follows:
NTP NTP
PLoss = åå P B
a=1 b=1
a ab Pb (34.3)
where Pa NTP, and Pb are the power generation of thermal power plants. is the number
of thermal power plants and there are three plants in this case. Hence, the total power
generation must be equal to the sum of power demand and power loss. Hence, the total
power demand of this test system is 900 MW.
Fuel cost coefficients, greenhouse gas coefficients, and maximum and minimum range
of power of six generators, B loss coefficient are given in Table 34.1.
TABLE 34.1 Thermal Unit Data
Greenhouse Gas Emission

Fuel Cost Coefficients
Coefficients
Plant Unit PMin PMax
AI BI CI DI EI FI
1 G1 0.15274 38.54 756.8 0.00419 0.32767 13.8593 10 125
G2 0.10578 46.159 451.33 0.00419 0.32767 13.8593 10 150
G3 0.02803 40.397 1049.3 0.00683 – 0.54551 40.2669 40 250
2 G4 0.03546 38.306 1243.5 0.00683 – 0.54551 40.2669 35 210
G5 0.02111 36.328 1658.6 0.00461 – 0.51116 42.8955 130 325
3 G6 0.01799 38.27 1356.7 0.00461 – 0.51116 42.8955 125 315
Table for B loss coefficient
0.000091 0.000031 0.000029

Bij = 0.000031 0.000062 0.000028
0.000029 0.000028 0.000072
PSO technique has been used to solve this problem. It gives superior results than other
methods. Among the different types of PSO, hierarchical PSO gives the optimal solution.
34.3 METHODOLOGY
• Particle Swarm Optimization (PSO) Technique
Among the meta-heuristic techniques, PSO is the one of most efficient technique which
can solve multi-objective optimisation problem for its robustness, high convergence speed,
and simplicity.
• Basic PSO
The basic PSO was first proposed by Kennedy and Eberhart. The concept of fish schooling
and bird flocking is the base of PSO. Swarm is the number of population employed for
searching food, and particles are the variables of objective functions. A velocity vector
based on the personal and global best particle is used in PSO to address the flying of the
particle. Improving particle’s positions and velocities, the solution can reach the optimal
position. Position and velocity is updated according to Eqs (34.5) and (35.6):
(
Vi k +1 = Vi k + c1.rand1. Pbest
k
2
) (
- X ik + c2 .rand 2 . Gbest
k
- X ik ) (34.5)
X ik +1 = X ik + Vi k +1 (34.6)
Vi k denotes the velocity of individual i at iteration k and Pbestk

denotes the position of
individuali c1c2rand2 at iteration k are the acceleration coefficients which are generally
k
set to 2. rand1 and rand2 are the random numbers between 0 and 1. Pbest is the best
k
position of individual i until iteration k and Gbest is the global best position until iteration
k. Different types of PSOs are listed below.
Canonical particle swarm optimisation
Time-varying inertia weight particle swarm optimisation stochastic inertia weight
particle swarm optimisation
Time-varying acceleration coefficients particle swarm optimisation fully informed
particle swarm optimisation
Hierarchical particle swarm optimisation
• Hierarchical PSO
Hierarchical PSO is an improved version of PSO, where it includes only the social and
cognitive parts, but the velocity update equation is not included with initial velocity.. So the
velocity updated equation will be
(
Vi k +1 = c1.rand1. Pbest
k
2
) (
- X ik + c2 .rand 2 . Gbest
k
- X ik ) (34.7)
Weight aggregation method has been employed for solving the multi-objective problems.
W1 is the weightage factor for cost function, and W2 is the weightage factor for emission
function.
F is the multi-objective function:
F = W1*F1 + W2*F2

In first phase, only economic scheduling is done, that is, only cost is optimised and
corresponding emissions have been calculated. Then environmental scheduling is done
that is, only emission is optimised and corresponding cost has been calculated. Basic PSO
has been employed here. The obtained pay off table is given below.
46642.04 793.9329
f=
48678.04 685.6718
For this calculation, the number of population taken is 50 and number of iteration taken is
1000. Here optimised cost is $46642.04 and corresponding emission is 793.9329 kg. Again,
optimised emission is 685 kg and corresponding cost is $48678.04.
Now, a comparison study of different PSO and lexicographic optimisation methods
have been given below to solve the multi-objective problem, that is, environmental and
economic. Dispatch of load where the number of population is 50 and the number of
iteration is 1000. Calculation of CPU time also be attached here. It also shows the superiority
of HPSO. The HPSO has also given a better result than lexicographic optimization method.
Figure 34.2 shows the cost and estimation values for different iterations using HPSO.
FIGURE 34.2 Variation of cost and emission with number of iteration using HPSO.
34.5 CONCLUSION AND FUTURESCOPE

This study has proposed economic scheduling, environmental scheduling, and also
combined economic-environmental scheduling of thermal power generation. This study
also shows the superiority of proposed method than the previous method and the conclusion
is that HPSO is better than other PSOs. Different kinds of optimization technique can be
used for better result. It also can be extended by increasing the number of buses and power
plants.
REFERENCES
1. Rezvani, A., Gandomkar, M., Izadbakhsh, M., and Ahmadi, A., 2015. Environmental/
economic scheduling of a micro-grid with renewable energy resources. Journal of cleaner
production, 87, 216–226.
2. Pappula, L., and Ghosh, D., 2017. July. Unequally spaced linear antenna array synthesis
using multi- objective Cauchy mutated cat swarm optimization. In 2017 IEEE International
Symposium on Antennas and Propagation & USNC/URSI National Radio Science Meeting
IEEE. 313–314
3. Xu, L., Jiawei, D., and Ming, H., 2017. October. Research on hybrid cloud particle swarm
optimization for multi-objective flexible job shop scheduling problem. In 2017 6th
International Conference on Computer Science and Network Technology (ICCSNT). IEEE.
274–278
4. Shao, X., Liu, W., Liu, Q., and Zhang, C., 2013. Hybrid discrete particle swarm optimization
for multi- objective flexible job-shop scheduling problem. The International Journal of
Advanced Manufacturing Technology, 67(9–12), 2885–2901.
5. Xu, W., Wang, W., He, Q., Liu, C., and Zhuang, J., 2017. October. An improved multi-
objective particle swarm optimization algorithm and its application in vehicle scheduling.
In 2017 Chinese Automation Congress (CAC) IEEE., 4230–4235
6. Izadbakhsh, M., Gandomkar, M., Rezvani, A., and Ahmadi, A., 2015. Short-term resource
scheduling of a renewable energy based micro grid. Renewable Energy, 75, 598–606.
7. Nazir, R., Laksono, H. D., Waldi, E. P., Ekaputra, E., and Coveria, P., 2014. Renewable energy
sources optimization: A micro-grid model design. Energy Procedia, 52, 316–327.
8. Zakariazadeh, A., Jadid, S., and Siano, P., 2014. Smart microgrid energy and reserve
scheduling with demand response using stochastic optimization. International Journal of
Electrical Power & Energy Systems, 63, 523–533.
9. Zakariazadeh, A., Jadid, S., and Siano, P., 2014. Stochastic multi-objective operational
planning of smart distribution systems considering demand response programs. Electric
Power Systems Research, 111, 156–168.
10. Mohammadi, S., Soleymani, S., and Mozafari, B., 2014. Scenario-based stochastic operation
management of microgrid including wind, photovoltaic, micro-turbine, fuel cell and energy
storage devices. International Journal of Electrical Power & Energy Systems, 54, 525–535.
11. Mazidi, M., Zakariazadeh, A., Jadid, S., and Siano, P., 2014. Integrated scheduling of
renewable generation and demand response programs in a microgrid. Energy Conversion
and Management, 86, 1118–1127.
12. Sah, S. K., Kumar, R., Biswas, S., Ghosh, S., Mandal, R. K., Ghosh, B. K., Biswas, M., and Maji,
D., 2018, June. Reactive Power Control of Modified IEEE 14 Bus System Using STATCOM.
In 2018 Second International Conference on Intelligent Computing and Control Systems
(ICICCS. IEEE. 335–339.
13. Sinha, A. A., Arya, M. K., Kumar, A., Mandal, R. K., Biswas, M., Ghosh, B. K., and Roy, S.,
2018, May. Linear Sensitivity Based Congestion Management of IEEE 30 Bus System Using
Distributed Generator. In 2018 2nd International Conference on Trends in Electronics and
Informatics (ICOEI) IEEE. 691–696.
CHAPTER 35
Generation of Electricity From

Harmful Gases using Fuel Cell
Diptadip Das, Sayak Samanta, and Prabal Saha
Department of Electrical Engineering, Institute of Engineering & Management, India.
ABSTRACT: The prominence and sustainability of energy generation is a global

target which binds together the techno-industrial and socio-economic welfare of
any developing countries. Among the renewable energy technologies for hydrogen
production, photocatalytic water splitting and photoelectrochemical (PEC) approaches
have been widely studied as energy resources without additional environmental
pollution. Studies carried out is a test on the proficiency of renewable energy by passing
air through zeolites which separates air into constituent gases (such as H2, NOx, SOx,
CO2) and breaking them into constituent particles and oxygen then, passing this oxygen
and hydrogen through fuel cell to generate useful power supply. Obtained results at
the test period are displayed on a characteristic curve which shows the relationship of
the performance parameters such as electrical energy, power, voltage, electrical charge.
Therefore, it is recommended that employing this innovative system in the energy
sector will save the environment from decaying, and also, as an alternative means to
the conventional system of energy generation. Thus, sustainability in electrical power
generation will drastically improve.
Keywords: Renewable energy, Clean energy, Purification of air, New source of energy,
Fuel cell, Electrical energy.
321
35.1 INTRODUCTION
Hydrogen is one of the promising, renewable energy sources beyond fossil fuel era
because of its clean and green nature, also having high-energy density characteristics.
Nanostructured materials have been studied extensively for solar hydrogen production
from water because they promise to offer superior photocatalytic performance.
Therefore, the main requirement of the photocatalytic and PEC water splitting is to
seek for nanostructured materials with the capability of efficient and cost-effective
conversion to hydrogen in presence of sunlight. In summary, focusing on water splitting
to molecular hydrogen and oxygen and re-iterating the conditions, all being known in the
community and some already mentioned by others, but the following two points seem to
be important. (i) A molar ratio of two, and (ii) direct measurements (not current based)
of molecular hydrogen and oxygen. One advantageous fact is that the end product of the
reaction (water) is environmentally friendly, produces no emissions capable of depleting
the ozone layer and also, can be converted to home use for economical purposes.
The production of electric power by means of a synchronous machine operating at
a constant synchronous speed at a fixed frequency is most often driven by a shaft power
system referred as gas turbine engines. These are natural gas-fuelled machineries, where
chemical combustion takes place in its combustion chamber for the efficient production
of thermal power (electricity) mainly used for crude oil drills and gas transportation
purposes. Also, creates useful thrust for aero gas turbine applications. Despite the benefits
of these natural gases as the means of powering this equipment, it also creates negative
environmental hazard. Its result depletes the ozone layer with significant increase of
atmospheric greenhouse gases (CO, CO2) by the production of consistent by-products of
NOx, SOx, etc. Therefore, remedying these constrains has led to the search of sustainable
alternative energy resources that are readily accessible, environmentally friendly, and cost-
competitive with fossil fuels. Meanwhile, the key point of research in this study is focused
on the production of electric power from harmful gases using fuel cells which operate
quietly and efficiently (Fontes and Eva, 2001).
Studies unveil that fuel cells are electrochemical devices which is capable of
transforming the chemical energy in a fuel to electrical energy or power. The developmental
effort of making fuel cells relevant in the scientific industrial market is the target of every
researcher because of its low exhaust emissions, high energy efficiency and variability,
energy density, economy friendly, unique operating conditions, planning flexibility, and
future development potential in comparison with the conventional fuel used in internal
combustion engines (Fontes and Eva, 2001; Florian, Yixin, and Laura, 2006; Xiaofeng,
Demba, and Claude, 2008; Alhassan and Garba, 2006). Another advantageous feature
of fuel cells is the lower susceptibility to fuel composition changes at high temperature
operating range. This is a suitable factor for distributed power generation and integration
with other types of bottoming cycles such as gas turbine cycles and cogeneration
(Florian, Yixin, and Laura, 2006).
Generation of Electricity From Harmful Gases using Fuel Cell  323
35.2 METHODOLOGY
Proposed Model
FIGURE 35.1
The setup facilitates the proposed extraction of oxygen from harmful gases like CO2, SO2,
and NO2 by the help of a semi-permeable membrane and passing it through the fuel cell as
an oxidant. The membrane acts like molecular sieve only allows oxygen molecules to pass
through. The oxygen extracted is then channelled through a one-way valve to maintain the
flow rate and pressure constraints. It also uses hydrogen as the fuel for the fuel cell and is
generated in a pressurised steel vessel through the reaction of water and metal, preferably
aluminium. The reaction is given by
2/3 AI + 2H2O = 2/3 AI(OH)3 + H2(g)
The pressure for the vessel is maintained at – and the temperature at – to gain maximum
yield of H2.
35.2.1 Materials Required

Fuel cell – alkali fuel cell
• Fuel required: Hydrogen
• Oxidant required: Oxygen
• Electrolyte used: Alkali Metal Hydroxide
• Electrodes used: Graphite
Setup for generating hydrogen
• Reactants: Aluminium, Water
• Vessel Used: Three-way flask
• End product: Hydrogen gas, Aluminium Hydroxide
setup for extracting oxygen
• Source: Flue gases extracted from combustion of fossil fuels
• Semi-permeable membrane
• End products: Oxygen.
35.2.2 Fuel Cells

The hydrogen–oxygen fuel cell rig is a small test-bed which is used to experiment and
determine the electric power output from hydrogen and oxygen gases.
Alkali fuel cells operate on compressed hydrogen and oxygen. They generally use a
solution of potassium hydroxide (chemically, KOH) in water as their electrolyte. Efficiency
is about 70%, and operating temperature is 150°C to 200°C, (about 300°F to 400°F). Cell
output ranges from 300 watts (W) to 5 kilowatts (kW). Alkali cells were used in Apollo
spacecraft to provide both electricity and drinking water.
Proton exchange membrane (PEM) fuel cells work with a polymer electrolyte in
the form of a thin, permeable sheet. Efficiency is about 40 to 50%, and operating
temperature is about 80°C (about 175°F). Cell outputs generally range from 50 to 250 kW.
These cells operate at a low temperature to make them suitable for homes and cars. But
their fuels must be purified, and a platinum catalyst is used on both sides of the membrane,
raising costs.
Solid oxide fuel cells (SOFC) use a hard, ceramic compound of metaloxides (like
calcium or zirconium) as electrolyte. Efficiency is about 60%, and operating temperatures
are about 1,000°C (about 1,800°F). Cells outputs generally range around 100 kW. However,
the high-temperature limits applications of SOFC units.
Although the problem of leaking is resolved with solid electrolytes but chances of
cracking is there. Hence, hydrogen–oxygen fuel cell is used to design our experiment.
35.3 EXTRACTION OF OXYGEN

Oxygen is separated from atmosphere using microporous, aluminosilicate minerals
called zeolites. Zeolites have a porous structure that can accommodate various ions such
as Na+, K+, Ca2+. It has the potential to separate gases such as O2, H2O. The polluted air
is sucked in using exhaust which is then passed over zeolite bed where extraction of O2
takes place. This oxygen is supplied to the fuel cell through a pipe. Below is the table of
composition of flue gases.
Pulv. Coal Waste Coal Coal-Fired
Gas-Fired CC
Combustion Incinerate Gasification IGCC Natural Gas
Flue Gas
Flue Gas Flue Gas Flue Gas Flue Gas
O2 % –v 6 7–14 12 14
N2 % –v 76 Balance 4 66 14 76
CO2 % –v 11 6–12 4 7 1 3
Therefore, in pulverised coal combustion flue gas,

Mol wt of CO2 is 44
C = 12
O = 2*16 = 32
% of oxygen in CO2 is 32/44 * 100 = 71%.
Therefore,
% of oxygen in 11% –V CO2 is 7.99%
Using same calculation, we find oxygen concentration in waste incinerate, coal
gasification, coal-fired flue gas, natural gas, and in gas-fired cc flue gas as 8.72, 2.90, 5.08,
0.72, and 2.18, respectively.
Flue Gases Extracted Oxygen %-v from co2 %-v of these Flue Gases
Pulv. Coal combustion flue gas 7.99
Waste incinerate flue gas 8.72
Coal gasification flue gas 2.90
Coal-fired IGCC flue gas 5.08
Natural Gas 0.72
Gas-fired CC flue gas 2.18
Hence, total oxygen extracted from our source (waste incinerate flue gas) is
8.72 + 14 = 22.72%
35.3.1 Extraction of Hydrogen

• The production of hydrogen via an aluminium–water reaction is explored at
temperatures and pressures ranging from 273.15 to 600 K and 0.1–10 MPa,
respectively. Across this range, aluminium and water can react to form different
aluminium oxide and hydroxide species, resulting in differences in the release of
thermal energy, as well as the amount of water required stoichiometrically for the
reaction to proceed.
• Aluminium and water react to produce H2 and AlOOH (boehmite) at 1 bar above
294K.
• Al(OH)3 (gibbsite) is produced for temperatures below 294 K at 1 bar, or high
pressure.
• In this experiment, we are using the latter, by the use of a pressurised steel container.
FIGURE 35.2
Thus, we are using the reaction as 2/3AI + 2H2O = 2/3AI(OH)3 + H2(g).

As it is the most feasible reaction under room temperature and no extra constraints
are required for this reaction. As we can see, the ease of reaction increases at high pressure,
hence, the use of a pressurised vessel is considered.
35.4 DISCUSSION
35.4.1 Parameters of fuel cell
FIGURE 35.3 Parameters of fuel cell.

FIGURE 35.4 Detailed parameters.
FIGURE 35.5 MATLAB simulation.

Nominal composition of hydrogen supplied is 99.95% and is supplied from the hydrogen
tank. Nominal composition of oxygen supplied is 27.59% and is extracted from polluted gas
present in atmosphere such as COx, NOx, SOx, etc.
From the above calculations, we supply 22.72% extracted oxygen and 99.95% hydrogen
from hydrogen tank to the fuel cell. With a constant time of 1 second, the converter is
loaded with RL load of 6 kW. During the first 10 seconds, the utilisation of the hydrogen
is constant to the nominal value (Uf_H2 = 99.56%) using a fuel flow rate regulator. After
10 seconds, the flow rate regulator is avoided, and the rate of fuel is expanded to the most
extreme estimation of 85 lpm and the change in the stack voltage is observed. This will
result in affecting the stack efficiency, the air, and the fuel consumption. The outcome of the
simulation is as follows: At t = 0 s, the DC/DC converter applies 100 Vdc to the RL load (the
initial current of the load is 0 A). The fuel utilisation is set to the nominal value of 99.56%.
The current increases until the value of 110.5 A. To maintain the nominal fuel utilisation,
the flow rate is automatically set. At t = 10 s, the fuel flow rate is increased from 50 litres
per minute (lpm) to 85 lpm during a time period of 3.5 s, reducing by using hydrogen. This
results in an increase of the Nernst voltage and decrease in fuel cell current. Therefore, the
stack consumption and efficiency is decreased.
35.4.2 Results
FIGURE 35.6 Current variations.

FIGURE 35.7 Voltage variations.
35.5 CONCLUSION
From this experiment, we capture polluted gases from atmosphere which is responsible
for global warming and other hazardous effects thereby cleaning the atmosphere. After
passing these gases through semi-permeable membrane, we extract oxygen which is
used as a fuel for our fuel cell. From the results and graph obtained, we are capable of
generating electricity. Thus, we obtain a new source of renewable energy which is free
from any kind of pollutants and is, therefore, an efficient way to replace already existing
conventional fuels. Thus, a huge construction of this kind of systems will provide a new
source of sustainable energy.
REFERENCES
1. Fontes, E., and Eva, N., 2001. Modelling the Fuel Cell, American Institute of Physics, 14–17.
2. Florian, Z., Yixin, L., and Laura, S., 2006. A solid oxide fuel cell system for buildings, Energy
Conversion and Management, 809–818.
3. Xiaofeng, L., Demba, D., and Claude, M., 2008. Design Methodology of Fuel Cell Electric
Vehicle Power System, Proceedings of the International Conference on Electrical Machines
and Systems, ID., 1270.
4. Alhassan, M., and Umar Garba, M., 2006. Design of an Alkaline Fuel Cell Leonardo
Electronic, Journal of Practices and Technologies, ISSN: 1583–1078, 99–106.
5. Luna-Sandoval, G., Urriolagoitia, C. G., Hernández, L. H., Urriolagoitia, S. G., and
Jiménez, E., 2006. Hydrogen Fuel Cell Design and Manufacturing Process Used for Public
Transportation in Mexico City. Proc. of the World Congress on Engineering, London, U.K.,
vol. III
6. DOE Energy Efficiency & Renewable Energy Information Centre, 2006, http://www.
hydrogen.energy.gov, Accessed on 19 August 2014.
7. Scott, S., 2004. Fuel Cell/Gas Turbine Hybrid Systems, ASME International Gas Turbine
Institute, 2004.
8. Cifrain, M., and Kordesch, K., 2003. Hydrogen/oxygen (Air) fuel cells with alkaline
electrolytes, Handbook of Fuel Cells– Fundamentals, Technology and Applications,
(Chichester, 2003), 267 – 280, ISBN: 0-471-49926-9.
CHAPTER 36
Exploring Potential Building

Energy Saving through Daylight
Harvesting through Light Pipe
V. Aditya, Bidisha Bhattacharya, and Gunjan Kumar
ABSTRACT: Areas far away or isolated from the windows, like rooms in centralised
air conditioned buildings, can be illuminated with natural light during the daytime by
using light pipes. However, depending on the weather condition, sometimes daylight
coming from the light-pipe is not sufficient for illumination. During these times,
when natural lighting is not adequate or more lighting is needed, artificial lighting is
used. However, generally, all the artificial lighting components are operated together
without considering the intensity of the illumination. This in turn, causes efficient
renewable energy use. Prior to the savings in energy, energy simulation is to be done
in the sample building taken, using desired software and considering the weather
conditions for determining the effective availability of daylight. Control systems can be
used for illumination systems. Savings in energy consumption for lighting or daylight
harvesting will be achieved through daylight integration by controlling the system with
methods such as switching on–off by user, manual or automatic dimmers with timers or
demand control, etc. Actually, the main concern is the maximum optimisation in usage
of daylight keeping in mind cost-effectiveness, i.e. maximum utility of solar energy at
minimum cost possible.
Keywords: Windows, Light pipes, Daylight harvesting, Energy simulation, Cost-

effectiveness.
36.1 INTRODUCTION
Energy is the most fundamental part for all human activities like domestic, industrial,
medical, transport, etc., the world is facing a stern situation such as gradual depletion
331
of fossil fuels, increase in global warming, depletion in ozone layer, and other
environmental concern factors, and continuing rise of price in fuel if there is continued
usage of fossil fuels (Freewan, Shao, and Riffat, 2009; Carlosa and Corvachob, 2015).
Future smart buildings are needed to be constructed by utilising all the opportunities
brought about by the environment, societal, and changes in technology. Energy can
be saved without lowering the standard of living, but with the effective use of smart
technology which will produce the same comfort and convenience (Granqvist,
2014). The increase in government and sustainable development issues has led to the
development of measures for energy-saving technologies in building and other sectors.
The evaluation of energy saving through designing a building has to undergo a lot
of complexity and uncertainty. The saving of energy through daylight opening in the
buildings is not only useful for quantitative analysis but also qualitative description
(Tiwari, 2016).
Daylighting is considered as the efficient substitute of electrical energy for artificial
lighting. It will not only reduce the energy consumption through lighting but also help in
effective reduction in the electrical peak loads (Energy and Buildings, Elsevier).
36.1.1 Aims of the Current Research

The full consideration of the energy savings by using daylighting is taken into account.
The designs and solutions are needed to maintain a balance between provision of
daylighting and performance of energy saving. This paper aims is to find whether
there is a potential saving of daylighting energy or the solar energy, along with the
achievement of effective thermal and visual comfort.
36.1.2 Problems and Factors that are to be Dealt

The major problem that is to be dealt with is the energy consumption of the whole
building. Solutions are being provided in the study in order to reduce the consumption of
daily artificial energy by 20–30% using daylighting (Shin, Yun, and Kim, 2011). The other
problems that are to be checked are increasing the cardinal cycles and diurnal cycles of the
human well-being where the activities and health conditions in the working environment.
To increase the performance and productivity from 0.45% to 15% by using the daylighting
energy instead of artificial lighting energy. The human factors that are to be dealt with
are Thermal comfort, Indoor Air quality, Visual Comfort, Glare, Daylight penetration
factor.
36.1.3 Research Problem

Energy analysis of typical building of Kolkata climate condition for lighting energy and
daylight estimation through simulation and further integration of light pipe for daylight
harvesting.
Exploring Potential Building Energy Saving through Daylight Harvesting  333
36.2 BACKGROUND INFORMATION

The building is located at Kolkata, West Bengal, India. The climate is warm and humid. The
area occupied by the building is 13,426 sq. metres. The occupancy type is daytime and the
load that is occupied is 400 KVA.
36.3 METHODOLOGY
It is evident that daylight has been used for centuries as a primary means of lighting. Now in
today’s era it has been an essential responsibility to ensure the proper use of daylighting in
the non-residential buildings which aids in the improvement of energy efficiency. Daylight
is important for its quality, spectral composition, and variability. It also provides a high
illuminance and an excellent distinction between the colour discrimination and rendering.
So the performance of daylight on the tasks is being dependent on the utilisation of it.
To utilise the maximum of the daylight, a building is considered as a reference and
calculations are being done on it. Window-wall ratio, occupancy schedule, daylight affected
area, and other things have been calculated in order to know the improvement areas. These
calculations are being modelled in a software for simulation and thereby calculating the
energy consumption. After simulation, the solution for improving the amount of daylight
is being provided in this paper.
36.3.1 Window-Wall Ratio

Windows are the conventional means of penetrating daylight into the interiors. The
effective use of them provides the allowance of daylight deep into the interior space and
provides visual comfort without the creation of any glare. The architectural design also
plays a major role in determining the utilization of daylight in the interior of buildings.
The window-to-wall ratio is the measure of the percentage area determined by dividing
the building’s total glazed area by its exterior envelope wall area. The window-to-wall
ratio helps in determining the amount of daylight that is obstructed due to the walls and
gets transpired through the windows. The performance of daylight is determined on the
amount of daylight availability on the building envelope which makes it a potential to use
it on that space. Table 36.1 shows the window-to-wall ratio (WWR) of the building that
is being taken into consideration in this study.
TABLE 36.1 Window-to-Wall Ratio
FACADE AREA, m2 Opaque wall area, m2 Glazed area, m2 WWR
North 1848.87 1813.81 35.06 1.90%

East 1120.69 997.97 122.72 10.95%
South 1849.85 1674.54 175.31 9.48%
West 1147.70 1021.14 126.55 11.03%
TOTAL 5967.10 5507.5 459.64 7.70%
36.3.2 Energy Modelling and Simulation Result

Energy conservation is one of the main accents of today’s era. Energy efficient measures
are being taken in every building in order to meet with the demands of the day-to-day life.
The building is being modelled with its constructional features along with other important
parameters such as glazing, cooling, heating, reflective value of the roof, window shading.
This helps in obtaining the amount of energy that is being consumed up. This also gives an
idea of about to what extent the solar heat gain and the formation of glare can take place
through the windows.
The building is an eight-storeyed office type which runs during the daytime and is an
ECBC compliance one. The building is modelled according to the baseline design and the
proposed design by Energy Conservation Building Code (ECBC). The project is located
at Kolkata and hence Kolkata weather file of ISHRAE has been used. The building model
description is being shown in Table 36.2.
TABLE 36.2 Building Model Description
Description Baseline Unit
ECBC 2017
Wall material
Table 4.7
Wall U-value 0.0704 Btu/hr-sqft F
ECBC 2017
Roof material
Table 4.4
Roof U-value 0.0581 Btu/hr-sqft F
Roof reflectivity 0.3
Glazing U-value 0.528 Btu/hr-sqft F
Glazing SHGC 0.27
Glazing VLT 0.27 %
Window shading Yes
Cooling COP (excluding fan power) 3.02
Fan power As per ECBC 2017 In WG
Lighting power density 0.88 W/ft2
Equipment power density 0.5 W/ft2
Process loads 26 kW
Occupancy Office - 200 ft /Person
2
Zone cooling setpoint 78 deg F

Zone heating setpoint 68 deg F
36.3.3 Building Occupancy Schedule

The building schedule is the business day-time schedule as per Energy Conservation
Building Code has been considered. Accordingly, the varying occupancy schedule is taken
into consideration along with the utility of lighting components and other equipment as
well as of the elevators are demonstrated here, to show the dependency and variation in all
kinds of load for a 24-hour span. Table 36.3 shows the building occupancy schedule.
TABLE 36.3 Building Occupancy Schedule
Time Period Occupancy Schedule Lighting Schedule Equipment Schedule Elevator Schedule
00:00–01:00 0.00 0.05 0.00 0.05

01:00–02:00 0.00 0.05 0.00 0.05
02:00–03:00 0.00 0.05 0.00 0.05
03:00–04:00 0.00 0.05 0.00 0.05
04:00–05:00 0.00 0.05 0.00 0.05
05:00–06:00 0.00 0.05 0.00 0.05
06:00–07:00 0.00 0.10 0.00 0.20
07:00–08:00 0.10 0.30 0.00 0.40
08:00–09:00 0.20 0.90 0.10 0.80
09:00–10:00 0.95 0.90 0.90 0.80
10:00–11:00 0.95 0.90 0.90 0.55
11:00–12:00 0.95 0.90 0.90 0.35
12:00–13:00 0.95 0.90 0.90 0.25
13:00–14:00 0.50 0.50 0.80 0.95
14:00–15:00 0.95 0.90 0.90 0.95
15:00–16:00 0.95 0.90 0.90 0.35
16:00–17:00 0.95 0.90 0.90 0.15
17:00–18:00 0.95 0.95 0.90 0.75
18:00–19:00 0.30 0.50 0.50 0.95
19:00–20:00 0.10 0.30 0.10 0.50
20:00–21:00 0.10 0.30 0.10 0.30
21:00–22:00 0.10 0.20 0.00 0.20
22:00–23:00 0.00 0.10 0.00 0.05
23:00–24:00 0.00 0.05 0.00 0.05
36.3.4 Building Energy Simulation

Energy simulation is done through the modelling of building as per the ECBC Guidelines
using a software e-Quest. This software not only provides the building 3-D model, envelope,
lighting, process, and HVAC loads to accurately represent the energy consumption of the
building. Energy is taken in kWh. Figure 36.1 shows the consumption of energy by different
utilities in the form of pie chart. Figure 36.2 shows the bar graph of the consumption of
energy by different utilities in the building.
The simulation output for the baseline design in compilation with the ECBC 2017 is
shown below:
FIGURE 36.1 Baseline design.
FIGURE 36.2 Baseline design.
36.3.5 Daylight Calculation

In the computer analysis method, the daylight analysis is done by creating a model in
ECOTECT and then simulating it further, and the result is given in Table 36.4 as well as
the daylight analysis zone wise report is provided. As per the baseline design demonstrated
above, the effective daylight area is calculated. The effective daylight area is very vital
to be obtained for ensuring potentiality of solar illumination and thus having effective
energy conservation. Considering the same floor area, same visible light transmittance,
same recommended Lux level, or range in each floor of our sample building, the optimum
daylight affected area is obtained to be around 41% which is fair enough to be taken into
considerations and implementation as well. Table 36.4 shows the effective daylighted area
in the building.
TABLE 36.4 Calculation of Effective Daylight Area
Floor Area Recommended % of Effective Effective

Floors VLT of Glazing
(SQM) Lux Level Daylight Area Daylight Area
Ground Floor 1664.87 0.7 100–2000 41.94% 698.41
First Floor 1470.16 0.7 100–2000 41.08% 603.94
Second Floor 1470.16 0.7 100–2000 41.08% 603.94
Fourth Floor 1470.16 0.7 100–2000 41.08% 603.94
Fifth Floor 1470.16 0.7 100–2000 41.08% 603.94
Sixth Floor 1470.16 0.7 100–2000 41.08% 603.94
Seventh Floor 1470.16 0.7 100–2000 41.08% 603.94
Eighth Floor 1470.16 0.7 100–2000 41.08% 603.94
Total 13,426.20 5529.93
% of Effective Daylighted Area 41.11%
36.3.5.1 Daylight Analysis Report Zone Wise

The proposed model has been simulated in ECOTECT and after simulation, we could find
the results of the daylight variation in different floors. Figure 36.3 shows the availability of
daylight in the ground floor (100–2000 lux). Figure 36.4 shows the availability of daylight
in different floors starting from first to eighth (100–2000 lux).
FIGURE 36.3 Simulation result of ground floor.

FIGURE 36.4 Simulation of first–eight floor (100–2000 lux).
36.4 DISCUSSIONS
Estimation and integration of light pipe
Light pipe is an innovative design to direct daylight into deep areas where the daylight
cannot reach. It is gathered via a polycarbonate dome at rooftop and then transmitted to
different interior areas within the building (Zhang, 2000).
Constructional Features
• Light Collector: Collects and redirects sunlight at low angle for maximum capture,
all day long.
• Reflective System: delivers 99% specular reflectivity for maximum sunlight transfer
with durability and long life.
• Light diffuser: Can be flat or dome shaped and is used for high light transmission
in industrial roofs of buildings. This is of two types: soft and bright light diffusers.
Principle of operation
Light pipes guide light from the source to the object that is to be illuminated by total
internal reflection. If a ray of light inside the material hits the side of the light pipe less than
the critical angle part of the light is reflected according to Snell’s Law and Fresnel Reflection
and transmission Relationships.
36.4.1 Expected Cost for Light Pipe

A sample of typical 760 mm diameter Light Pipe with lumen output is 21,000 lumens is
taken. The electrical lighting wattage equivalent to 760 mm diameter light pipe is 465 W.
Using the light pipe the following were calculated:
1. Electrical Savings per Year:
1 light pipe × 0.465 kW × 10 hours per day × 365 days × Rs. 7.00/kWh = Rs. 11,880.75
2. Maintenance Cost Savings:
Considering bulb replacement and other costs,
Total W × Rs/W = 465 × 5 = Rs. 2,325.00
3. Total cost of each light pipe: Rs. 13,500.00.
Finally, it has been found that the approximate cost for a light pipe is Rs 13,500
which is lesser than the artificial electrical wattage lamp.
36.5 CONCLUSION
Daylighting is essential for every human. It maintains the cardiac rhythm of the body. By
utilising the maximum daylight, we could end up in saving the non-renewable sources
of energy and would last longer in use. From the above, we can observe the total annual
lighting energy consumption for ECBC baseline compliance simulated office building
is 343240 kWh (41.1% of total load) with overall EPI 62.3. The percentage of floor
daylight area meeting UDI requirement during 90% of year is 41% which is above the
40% requirement for ECBC compliance. With the utilisation of light pipes, the payback
period for one light pipe for solar daylight harvesting is 1.8–2 years.
REFERENCES
1. Zhang, X., 2000. Light pipes for daylight Penetration indoors, In Proceedings of renewable
energy for housing conference, The UK Solar Energy, 20 October 2000, Perth.
2. Muneer, T., and Zhang, X. Cost and Value Analysis of Piped Daylight.
3. Helms, Ronald, N., 1980. Illumination engineering for energy efficient luminous
environments, Prentice. Hall, Inc.
4. Fasi, Mohammed Abdul, and Budaiwi, Ismail Mohammad, 2015. Energy performance of
windows in office buildings considering daylight integration and visual comfort in hot
climates, Elsevier, 15 September 2015.
5. Shin, J. Y., Yun, G. Y., and Kim, J. T., 2011. Evaluation of daylighting effectiveness and
energy savings potential of light pipe systems in buildings, Indoor and Built Environment,
21, 129–136.
6. Mayhoub, M. S., and Carter, D. J., 2014. The Costs and Benefits of Using Daylight guidance
to light office buildings, Building and Environment, 46, 698–710.
7. Energy efficient lighting strategies in buildings, Elsevier.
8. Sudan, Madhu and Tiwari, G. N., 2016. Daylighting and energy performance of a building
for composite climate: An experimental study, Alexandria Engineering Journal, 55(4).
9. Granqvist, C. G., 2014. Electrochromics for smart windows: oxide-based thin films and
devices, Thin Solid Films 564 1–38.
10. Freewan, A., Shao, L., and Riffat, S., 2009. Interactions between louvers and ceiling geometry
for maximum daylighting performance, Renewable Energy, 34, 223–232.
11. Carlosa, J. S., and Corvachob, H., 2015. Evaluation of the performance indices of a ventilated
double window through experimental analytical procedures: SHGC-values, Energy Build,
86, 886–897.
CHAPTER 37
Greenhouse Gas Mitigation for

College Building
Amulayan Bhargow, Vivek Kumar, Aman Kumar, Rustam Alam,
and Nikesh Kumar
Institute of Engineering & Management, Kolkata, West Bengal, India.
ABSTRACT: In this era of globalisation, and with the increasing demand of natural
resources, it is our responsibility to reduce the harmful effects of global warming,
and this technique will be a green solution through which we can reduce the energy
consumption to the extreme level in every college of our country. This study gives an
idea about the effect of repetitive audits by measuring different heat load calculation
on different sites of building. Our country is the third largest greenhouse gas (GHG)
emitter in the world and two-third of this energy is from electrical power consumption.
So this technique is a green and sustainable concept to meet our future demand which
is the need of the hour. Mitigation of GHG emissions will reduce the extent of climatic
change and future needs for different aspects of livelihood. About 80% of the technical
mitigation potential could be accomplished only by cost and energy saving. This idea
can be extended to whole city and our country subsequently in order to save energy on
higher platform.
Keywords: Green solution, GHG (Greenhouse gas), Global warming, Mitigation

potential.
37.1 INTRODUCTION
Decrease of greenhouse gas (GHG) emissions could be used as our first priority to reduce
the greenhouse effect. This paper deals with major factor responsible for increasing trends
of GHG emissions and solution to reduce this effect in college building of our country.
When sun rays falls into earth, some rays reflect while some part of heat is captured by
GHG and sky in atmosphere. Energy that is captured is called GHG. It has been estimated
that average temperature of the Earth is 15° Celsius but without the GHG it would be –25°
Celsius, due to increase in GHG average temperature of Earth is increasing day by day
341
leading to adverse condition for living. This warming of Earth is called global warming.
Most of GHG emissions are due to electrical power consumption and building construction
in our country. So use of renewable energy source is a green solution leading to sustainable
development. Different analysis of the effect of the carbon content on the economic
development of different countries with different economic structures indicates that the
GHG emissions can reflect their contribution to climate change for developing countries,
but for developed countries, a single index reflecting the carbon emissions is not enough
and other factors must be considered to determine whether the carbon emission intensity
reduction is faster than the GDP growth.
The present scenario of GHG emissions in India is that India’s total GHG emissions
in 2014 were 3,202 million metric tons of carbon dioxide equivalent, totalling 6.55% of
global GHG emissions. India’s GDP increased by 357% from 1990 to 2014, while GHG
emissions increased by 180%. Relative to GDP, India emits twice as many GHGs as the
world average.
37.2 METHODOLOGY
Reducing energy demand is the global emerging trends and alarming topic to be focused for
sustainable growth and development. GHG emissions play a major role for ability to boost
climate change public policies and strategies, economic growth, GDP rate development, etc.
Energy use in building design and operation is integral to these efforts. In 2010, buildings
accounted for 32% of total global final energy use and 19% of energy-related GHG
emissions, with projections for further increase. Till date, the global promise of building-
level interventions remains mixed, at best. Alternatives are yet to be rigorously adapted
for application in specific local circumstances, making country-, city- and building-level
research necessary components for sustainable growth and development. Under the Paris
agreement, governments have agreed to hold global mean warming to well below 2 ° Celsius
and pursue efforts to limit warming to 1.5° Celsius.
FIGURE 37.1 Graph indicating the percentage consumption of GHG while building
construction.
Greenhouse Gas Mitigation for College Building  343
Under the future scope and innovation steps taken to reduce GHG emissions for
College building are Better Design considerations under which comes the initial stage,
alternative layouts, and materials can be considered and details can be refined to use
less material for better enhancement. As the design progresses, these opportunities
diminish. Part of this early planning involves a life cycle assessment on which carbon
footprint is based. Next up is HVAC (heating and ventilation system for air conditioning):
HVAC comprises 40% of all carbon emissions, incorporating the most efficient heating,
ventilation, and air conditioning systems, along with efficient operations and scheduled
maintenance of such systems, reduces carbon footprint and hence, GHG emissions can
be reduced to a certain extent. Installing a low-energy humidifier instead of a typical
electric steam humidifier will be green by reducing a building’s carbon footprint. Also,
equipping a building with sensors can measure indoor air quality (IAQ) and determine
how much ventilation is needed. This means less electricity and natural gas will be
needed for the HVAC system, which lowers energy bills and reduces the building’s
carbon footprint.
Continuous insulation is installed on the interior, exterior, or is integral to any opaque
surface of the building. In a steel-framed building, thermal shorts reduce the assembly
R-value of cavity-insulated wall systems by more than 50%. So insulation of material
plays a vital role for sustainability and durability. Continuous insulation saves energy and
reduces the carbon footprint. It provides thermal air, water, and vapour control layers
in one system and simplifies the construction process. Lighting is done to install large
number of window panes in order to provide better lighting system. Lighting accounts
for approximately 40% of the energy used in a typical commercial building. Solar thermal
gain lowers carbon footprints, but too much solar gain in summer causes overheating
and increases the need for cooling. Too little solar gain in winter increases the need for
heating. Solar control window films can reduce carbon contents by reducing energy
expenditures by up to 30%. Open plan offices and light-coloured interior finishes help
distribute daylight more deeply within the building and hence provide better ventilation
system. High-performance curtain wall systems with integrated sunshades and light
shelves can increase natural light which is a better alternative. Properly executed day
lighting strategies can reduce HVAC peak loads with corresponding reductions in
mechanical equipment capacity and carbon content.
Choose recyclable building materials that have less negative effect on the
environment and better efficient features. Support green suppliers and vendors that
embrace green practices. Recycled steel reduces mining waste by 97%, air pollution by
86%, and water pollution by 76% which is green solution for college building. Using
efficient fixtures and appliances can reduce water usage up to a large extent. Specify
high-efficiency toilets with reduced average flush volumes compared to conventional
toilets. Use rainwater as a valuable water source. Rooftop rainwater harvesting and
xeriscaping can result in 50% savings in outdoor water use which is a boon for
sustainable growth development.
FIGURE 37.2 Commercial building emissions intensity.
Renewable energy: Installing the walls or roof of a building for solar air heating, solar electric
photovoltaic (PV) systems, or solar water heating can permanently eliminate part of the
demand for conventional energy. This is the next phase where large-scale CO2 reductions
from the building sector will be realised because we can only reduce so much energy usage,
without generating it at some point and hopefully from carbon-free sources like solar. Metal
roofs are a good host for solar PV because they are very durable and have lifetimes that meet
or exceed the 25–30 year expected life of the solar PV system.
‘Highly efficient buildings should be oriented along an east-west axis, maximising
north-and south-facing glazing.’ Also, siting the building near public transportation reduces
its carbon footprint.
FIGURE 37.3 Per capita building emissions intensity.
Autoclaved aerated concrete (AAC) is a lightweight, precast, foam concrete building material
suitable for producing concrete masonry unit (CMU) like blocks. Improve thermal efficiency
reduces the heating and cooling loads in building. It is a green solution for sustainable
development as it produces at least 30% less solid waste than traditional concrete and hence
decreases GHG emissions by 50%.
The calculation is done for GHG mitigation System Selection and CO2 emissions for our
College
Greenhouse Gas Mitigation for College Building  345
Total power consumption annually in kWh = 2,16,025 kWh

Carbon footprint due to consumption of electricity: 2,16,025*0.8 = 172.82 tonnes of
CO2.
Green Solar plant: Off grid system Inverter – 6kw, battery – 150 Ah of total 8 in nos,
solar panels – 5kwp, i.e. 350 kw* 15 which means 5 kw off-grid system is required. On grid
system 10 kw power supply (Inverter – 10 kw, solar panel – 10 kwp which means 325 watt*
(30s in no).
37.3 CONCLUSION
In this report, we have discussed about carbon footprint and its emission. We have also
determined the approximate number of tonnes of carbon produced due to electricity in
our campus with the help of heat-load calculation and global warming potential. From
this report, it is transparent that the reduction of carbon footprint is possible using two
ways – the usage of renewable sources for electricity production and consumption in
part-load.
Not only this, but we can also apply the same methodology to other campuses as well
and further to the whole city of Kolkata. This would help in determining data which can
further be analysed for better control of carbon footprint in the respective areas.
REFERENCES
1. Technologies for Climate Change Mitigation - UN CC: Learn https://www.uncclearn.org/
sites/default/files/inventory/unep223.pdf
2. Desjardins, R. L., et al., 2001. Canadian greenhouse gas mitigation options in agriculture.
Nutrient Cycling in Agroecosystems 60.1–3, 317–326.
3. Desjardins, R. L., Kulshreshtha, S. N., Junkins, B., Smith, W., Grant, B., and Boehm, M.,
2001. Canadian greenhouse gas mitigation options in agriculture. Nutrient Cycling in
Agroecosystems, 60 (1–3), 317–326.
4. Michaelis, Laurie, and Davidson, Ogunlade, 1996. GHG mitigation in the transport sector.
Energy Policy, 24(10–11), 969–984.
5. Dhakal, Shobhakar, 2010. GHG emissions from urbanization and opportunities for urban
carbon mitigation. Current Opinion in Environmental Sustainability, 2(4), 277–283.
6. Jianyi, Lin, Cao, Bin, Cui, Shenghui, Wang, Wei, and Bai, Xuemei, 2010. Evaluating the
effectiveness of urban energy conservation and GHG mitigation measures: The case of
Xiamen city, China. Energy Policy, 38(9), 5123–5132.
7. Brown, S., Cannell, M., Heuveldop, J., Kauppi, P., Sathaye, J., Singh, N., Weyers, S., Dixon,
R., Grainger, A., Leemans, R., Moura-Costa, P.H., Nilsson, S., Pinard, M., Schopfhauser, W.,
Sedjo, R., and Trexler, M., 1996. Chapter III.F. Establishment and management of forests
for mitigation of greenhouse gas emissions. In Working group II, Intergovernmental Panel
on Climate Change, 1995 Assessment for the Framework Convention On Climate Change,
773–799
8. Yedla, Sudhakar, Shrestha, Ram M., and Anandarajah, Gabrial, 2005. Environmentally
sustainable urban transportation—comparative analysis of local emission mitigation
strategies vis-a-vis GHG mitigation strategies. Transport Policy, 12(3), 245–254.
“BuildingEmissionIntensity” |LawandtheEnvironment https://www.lawandenvironment.
com/tag/building-emission-intensity/ Andrea Ramirez: International Journal of
Greenhouse Gas Control https://www.journals.elsevier.com/international-journal-of-
greenhouse-gas-control
9. Smith, Pete, Martino, Daniel, Cai, Zucong, Gwary, Daniel, Janzen, Henry. Greenhouse gas
mitigationinagriculturehttps://royalsocietypublishing.org/doi/abs/10.1098/rstb.2007.2184
10. Lucon, Oswaldoand Ürge-Vorsatz, Diana. https://www.ipcc.ch/site/assets/uploads/2018/02/
ipcc_wg3_ar5_chapter9.pdf Ministry of environment: Methodology – Greenhouse
Gas Emissions https://www.mfe.govt.nz/publications/climate-change/quantification-air-
quality-co%E2%80%93benefits-climate-change-mitigation-15

Intelligent Electrical Systems A Step Towards Smarter Earth 2021

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Intelligent Electrical Systems A Step Towards Smarter Earth 2021

Uploaded by

Copyright:

Available Formats

INTELLIGENT ELECTRICAL

Satyajit Chakrabarti is Pro-Vice Chancellor, UEM a premier organization in education

Ayan Kumar Panja is Assistant Professor at IEM, Kolkata, India.

Other titles in the Series

and by CRC Press

© 2021 by CRC Press, Taylor & Francis Group, LLC

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

Library of Congress Cataloging in Publication Data

ISBN: 978-0-429-35599-8 (eBook)

Typeset in Minion Pro font

Organizing Committee xix

Chapter 01  Design of One Adaptive Controller for a

Chapter 02  A Comparative Study of Phasor Estimation

2.3 RESULTS AND DISCUSSION 14

Chapter 03  Adaptive Fuzzy PID Controller for a Current

Chapter 04  Multi-Objective Optimization for Enhanced Surface

Chapter 05  Self-Adaptive Multivariate q-Gaussian-based

Chapter 06  Differential Evolution Coupled Neural-based

Chapter 07  Numerical Investigation of Heat Conduction

Chapter 08  Analytical Modelling and Simulation of a

8.3 NUMERICAL HEAT TRANSFER SIMULATION OF THE HOT-WALL

Chapter 09  Design and Analysis of a Dual-Axis Solar Tracker

Chapter 10  Indices based Comparative Performance Analysis

Chapter 11  Natural Ground-Water Recharge Scenario in Kolkata

Chapter 12  Performance Analysis of Grid Connected PV system

Chapter 13  Development of a Low Cost Autonomous

Chapter 14  Monitoring Static Security Assessment in Its Full

Chapter 15  Modelling of Solar Cell Considering One Diode

15.3 SIMULATION MODEL OF A SOLAR CELL 126

Chapter 16  Modelling of Intelligent Cooling of a Building in

Chapter 17  Control and Monitor of Non-linear Gas Flow by

Chapter 18  Transient Stability Controlling and Assessment

Chapter 19  Bagasse Drying by Waste Heat Recovery from

Chapter 20  Study of Score Coefficient Functions for

20.3.1 Experimental Setup 177

Chapter 21  Auto-adaptive Fuzzy Controllers with Automatic

Chapter 22  An Application of Multivariate Control Chart for

Chapter 23  Analysis and Performance Indexing of Khagaria

Chapter 24  Fractional Order PID controller for Setpoint

Chapter 25  Analysis and Simulation of PWM-Based Half-Bridge

Chapter 26  Design and Analysis of High-Speed Phase Locked

Chapter 27  Design of APOD-PWM Based Multi-level Inverter

Chapter 28  Battery Storage Technologies: A Review 257

Chapter 29  Experimental Investigation of Solar Energy-Assisted

29.3 EXPERIMENTAL STUDIES 271

Chapter 30  Automatic Speed Control of a Three Phase

Chapter 31  OWEC: Optimised Weighted Ensemble

31.5 CONCLUSION 294

Chapter 32  Detection of Disaster Affected Regions Using

Chapter 33  Enhancing Massive Open Online Courses with

Chapter 34  Application of Multi-Objective Particle Swarm

Chapter 35  Generation of Electricity From Harmful Gases

35.2.1 Materials Required 323

Chapter 36  Exploring Potential Building Energy Saving

Chapter 37  Greenhouse Gas Mitigation for College Building 341

This book is itself an acknowledgement of cutting edge technological application in

On behalf of the local organizing committee and Institute of Engineering &

Amartya Mukherjee is an Assistant Professor at the Institute of Engineering &

Design of One Adaptive

ABSTRACT: This chapter represents a technique to design an adaptive controller for

Keywords: Adaptive control, Non-linear system, Lyapunov stability theory, Non-linear

1.2 NON-LINEAR PARAMETERISED FRAMEWORK

where θ and k: number of unknown parameters.