Innovations in Electrical and Electronic Engineering: Proceedings of ICEEE 2022, Volume 1 (Lecture Notes in Electrical Engineering, 893) 9811917418, 9789811917417

Citation preview

Lecture Notes in Electrical Engineering 893

Saad Mekhilef Rabindra Nath Shaw Pierluigi Siano   Editors

Innovations in Electrical and Electronic Engineering Proceedings of ICEEE 2022, Volume 1

Lecture Notes in Electrical Engineering Volume 893

Series Editors Leopoldo Angrisani, Department of Electrical and Information Technologies Engineering, University of Napoli Federico II, Naples, Italy Marco Arteaga, Departament de Control y Robótica, Universidad Nacional Autónoma de México, Coyoacán, Mexico Bijaya Ketan Panigrahi, Electrical Engineering, Indian Institute of Technology Delhi, New Delhi, Delhi, India Samarjit Chakraborty, Fakultät für Elektrotechnik und Informationstechnik, TU München, Munich, Germany Jiming Chen, Zhejiang University, Hangzhou, Zhejiang, China Shanben Chen, Materials Science and Engineering, Shanghai Jiao Tong University, Shanghai, China Tan Kay Chen, Department of Electrical and Computer Engineering, National University of Singapore, Singapore, Singapore Rüdiger Dillmann, Humanoids and Intelligent Systems Laboratory, Karlsruhe Institute for Technology, Karlsruhe, Germany Haibin Duan, Beijing University of Aeronautics and Astronautics, Beijing, China Gianluigi Ferrari, Università di Parma, Parma, Italy Manuel Ferre, Centre for Automation and Robotics CAR (UPM-CSIC), Universidad Politécnica de Madrid, Madrid, Spain Sandra Hirche, Department of Electrical Engineering and Information Science, Technische Universität München, Munich, Germany Faryar Jabbari, Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA, USA Limin Jia, State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Alaa Khamis, German University in Egypt El Tagamoa El Khames, New Cairo City, Egypt Torsten Kroeger, Stanford University, Stanford, CA, USA Yong Li, Hunan University, Changsha, Hunan, China Qilian Liang, Department of Electrical Engineering, University of Texas at Arlington, Arlington, TX, USA Ferran Martín, Departament d’Enginyeria Electrònica, Universitat Autònoma de Barcelona, Bellaterra, Barcelona, Spain Tan Cher Ming, College of Engineering, Nanyang Technological University, Singapore, Singapore Wolfgang Minker, Institute of Information Technology, University of Ulm, Ulm, Germany Pradeep Misra, Department of Electrical Engineering, Wright State University, Dayton, OH, USA Sebastian Möller, Quality and Usability Laboratory, TU Berlin, Berlin, Germany Subhas Mukhopadhyay, School of Engineering & Advanced Technology, Massey University, Palmerston North, Manawatu-Wanganui, New Zealand Cun-Zheng Ning, Electrical Engineering, Arizona State University, Tempe, AZ, USA Toyoaki Nishida, Graduate School of Informatics, Kyoto University, Kyoto, Japan Luca Oneto, Dept. of Informatics, Bioengg., Robotics, University of Genova, Genova, Genova, Italy Federica Pascucci, Dipartimento di Ingegneria, Università degli Studi “Roma Tre”, Rome, Italy Yong Qin, State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing, China Gan Woon Seng, School of Electrical & Electronic Engineering, Nanyang Technological University, Singapore, Singapore Joachim Speidel, Institute of Telecommunications, Universität Stuttgart, Stuttgart, Germany Germano Veiga, Campus da FEUP, INESC Porto, Porto, Portugal Haitao Wu, Academy of Opto-electronics, Chinese Academy of Sciences, Beijing, China Walter Zamboni, DIEM - Università degli studi di Salerno, Fisciano, Salerno, Italy Junjie James Zhang, Charlotte, NC, USA

The book series Lecture Notes in Electrical Engineering (LNEE) publishes the latest developments in Electrical Engineering - quickly, informally and in high quality. While original research reported in proceedings and monographs has traditionally formed the core of LNEE, we also encourage authors to submit books devoted to supporting student education and professional training in the various fields and applications areas of electrical engineering. The series cover classical and emerging topics concerning:

• • • • • • • • • • • •

Communication Engineering, Information Theory and Networks Electronics Engineering and Microelectronics Signal, Image and Speech Processing Wireless and Mobile Communication Circuits and Systems Energy Systems, Power Electronics and Electrical Machines Electro-optical Engineering Instrumentation Engineering Avionics Engineering Control Systems Internet-of-Things and Cybersecurity Biomedical Devices, MEMS and NEMS

For general information about this book series, comments or suggestions, please contact leontina. [email protected]. To submit a proposal or request further information, please contact the Publishing Editor in your country: China Jasmine Dou, Editor ([email protected]) India, Japan, Rest of Asia Swati Meherishi, Editorial Director ([email protected]) Southeast Asia, Australia, New Zealand Ramesh Nath Premnath, Editor ([email protected]) USA, Canada: Michael Luby, Senior Editor ([email protected]) All other Countries: Leontina Di Cecco, Senior Editor ([email protected]) ** This series is indexed by EI Compendex and Scopus databases. **

More information about this series at https://link.springer.com/bookseries/7818

Saad Mekhilef Rabindra Nath Shaw Pierluigi Siano •

•

Editors

Innovations in Electrical and Electronic Engineering Proceedings of ICEEE 2022, Volume 1

123

Editors Saad Mekhilef School of Science, Computing and Engineering Technologies Swinburne University of Technology Hawthorn, VIC, Australia

Rabindra Nath Shaw Office of the International Relations Bharath Institute of Higher Education and Research Chennai, India

Pierluigi Siano Department of Management and Innovation Systems University of Salerno Fisciano, Salerno, Italy

ISSN 1876-1100 ISSN 1876-1119 (electronic) Lecture Notes in Electrical Engineering ISBN 978-981-19-1741-7 ISBN 978-981-19-1742-4 (eBook) https://doi.org/10.1007/978-981-19-1742-4 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

The book features selected high-quality papers presented at the International Conference on Electrical and Electronics Engineering (ICEEE 2022), jointly organized by University of Malaya and Bharath Institute of Higher Education and Research, India, during January 8–9, 2022, through Webex. The book focuses on current development in the fields of electrical and electronics engineering. The volume one covers electrical engineering topics—power and energy including renewable energy, power electronics and applications, control, and automation and instrumentation, and volume two covers the areas of robotics, artificial intelligence and IoT, electronics devices, circuits and systems, wireless and optical communication, RF and microwaves, VLSI, and signal processing. The book is beneficial for readers from both academia and industry. We are thankful to all the authors that have submitted papers for keeping the quality of ICEEE 2022 at high levels. The editors of this book would like to acknowledge all the authors for their contributions and the reviewers. We have received invaluable help from the members of the International Program Committee and the chairs responsible for different aspects of the workshop. We also appreciate the role of Special Sessions Organizers. Thanks to all of them, we had been able to collect many papers on interesting topics, and during the conference, we had very interesting presentations and stimulating discussions. We hope that the volume will provide useful information to professors, researchers, and graduated students in the area of electrical and electronics engineering and technologies along with AI and IoT applications, and all will find this collection of papers inspiring, informative, and useful. We also hope to see you at a future ICEEE event. Saad Mekhilef Rabindra Nath Shaw Pierluigi Siano

v

ICEEE 2022 Organization

Patron Bhim Singh R. Venkatesh Babu (Pro Vice Chancellor)

Indian Institute of Technology Delhi, India BIHER, India

General Chair Saad Mekhilef

Swinburne University of Technology, School of Science, Computing and Engineering Technologies, Australia

Conference Chair and Chairman, Oversight Committee Rabindra Nath Shaw

Director International Relations, BIHER, India

Conference Secretary Saravanan D.

Galgotias University, India

Technical Chairs Nishad Mendis Ankush Ghosh

Det Norske Veritas, Australia The Neotia University, India

Publication Chairs Valentina E. Balas Sanjoy Das

Aurel Vlaicu University of Arad, Romania IGNTU, Manipur, India

vii

viii

ICEEE 2022 Organization

Springer/ICEEE Liaison Aninda Bose (Senior Editor)

Springer Nature, India

International Advisory Board Vincenzo Piuri Georges Zissis (President) Lakhmi C. Jain Tamas Ruzsanyi Valentina Balas N. R. Pal (President) George T. Yen-Wei Chen Milan Simic M. Paprzycki Maria Virvou D. P. Kothari (Ex Director) B. K. Panigrahi C. Boccaletti Akshay Kumar

University of Milan, Italy IEEE IAS University of Technology, Sydney Ganz-Skoda Electric Ltd., Hungary University of Arad, Romania IEEE CIS University of Piraeus, Greece Ritsumeikan University, Japan RMIT University, Australia Polish Academy of Sciences University of Piraeus, Greece IIT Delhi, India IIT Delhi, India Sapienza University, Italy Concordia University, Canada

Contents

Estimating the Levelized Cost of Electricity of the First Nuclear Power Plant in Bangladesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Abid Hossain Khan, Mehedi Hasan, M. Mizanur Rahman, and S. M. Sayeed Anowar Dynamic and Transient Stability Enhancement of Multi Area Power Systems Using Fuzzy Logic Control Against Load Disturbances . . . . . . Venu Yarlagadda, Rashmi Kapoor, Chava Sunil Kumar, and Giriprasad Ambati FFT Analysis and Power Quality Improvement in Grid Connected Solar Power Plant with MPPT Algorithm . . . . . . . . . . . . . . . . . . . . . . . Venu Yarlagadda, G. Lakshminarayana, Giriprasad Ambati, and Garikapati Annapurna Karthika

1

12

30

Transient Response and Stability Analysis of Weak Power System Using Hybrid Compensator, STATCOM and SVC: A Comprehensive Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Venu Yarlagadda, T. Haripriya, N. Amarnath Reddy, and Giriprasad Ambati

44

Comparative Analysis of GCSC and Hybrid Compensators Influence on Power Transfer Enhancement of UHV Transmission Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Venu Yarlagadda, Rashmi Kapoor, G. Lakshminarayana, and N. Amarnadh Reddy

60

Wind Energy System Using Self Excited Induction Generator with Hybrid FACTS Device for Load Voltage Control . . . . . . . . . . . . . Venu Yarlagadda, Garikapati Annapurna Karthika, Giriprasad Ambati, and Chava Suneel Kumar

77

ix

x

Contents

Microgrid Islanding Detection Using Travelling Wave Based Hybrid Protection Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shashank Gupta and Suryanarayana Gangolu

92

Power Flow and Stability Improvement in Distribution Systems Using Phase Angle Regulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Venu Yarlagadda, M. Naga Jyothi, G. Lakshminarayana, and T. Hari Priya Optimization and Comparison of High Performance and Low Power NOR Gate Circuit Using Hybrid Model of Dynamic Voltage Scaling and MTCMOS Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 Kamlesh Kumar, Mohit Dahiya, Manoj Kumar, and Priyanshu Lakra Design of UWB Antenna with Frequency Interference Mitigation Technique for Wireless Communication Applications . . . . . . . . . . . . . . 132 Koduri Sreelakshmi and Gottapu Sasibhushana Rao Design of Power-Efficient Operational Transconductance Amplifier in the Application of Low Pass Filter Using 180 nm CMOS Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 Nupur Mittal, Imran Ullah Khan, and Piyush Charan First Order Control System Using Python Technology . . . . . . . . . . . . . 152 Palash Jain and Jay Kumar Jain Water Cleaning Bot with Waste Segregation Using Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 M. M. Anas, M. Athiram, Anugraha Suresh, K. Archana, and Maneesha Shaji Design and Analysis of a Photovoltaic P&O-Based MPPT Lead-Acid Battery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 Shiv Prakash Bihari, Aruvand Gupta, Vikalp Gupta, and Abhinav Kumar Babul Comparative Study of PID and Model Predictive Control with Boost Converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Akash Verma, Sarthak Kamta Prasad, and Amritesh Kumar A Bidirectional DC-DC Converter Fed DC Motor for Electric Vehicle Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Mohd Faizan, Mohd Mehroz, Qazi Ramish, and Hina Nasir PV-WIND Hybrid System Based Cuckoo Search Maximum Power Point Tracking Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 V. Dhanunjaya, K. Vijaya Bhaskar Reddy, S. Vijaya Kumar, and P. Venkata Kishore

Contents

xi

TriboOnto: A Strategic Domain Ontology Model for Conceptualization of Tribology as a Principal Domain . . . . . . . . . . . 215 S. Palvannan and Gerard Deepak ANN-CF-PSO Algorithm Based Selective Harmonic Elimination in Cascaded Multilevel Inverter for PV Applications . . . . . . . . . . . . . . . 224 B. Ganesh Babu and M. Surya Kalavathi Performance Evaluation of Field Aged Polymer Insulators . . . . . . . . . . 237 G. Nithin Reddy, B. Subba Reddy, M. Ramez Halloum, Jangamreddy Rajasekhar Reddy, Febin Francis, and B. Kiran Kumar Reddy Particle Swarm Optimization Based Self-tuned PID Controller for Digital Excitation Control System . . . . . . . . . . . . . . . . . . . . . . . . . . . 248 Surendrasinh K. Solanki, S. S. Kanojia, Shanker Godwal, Akhilesh Nimje, and Vinod Patel Stability and Time Response Analysis of Integrated Renewable Energy System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 Vikalp Gupta, Shiv Prakash Bihari, Aruvand Gupta, and Abhinav Kumar Babul Design a Robust Control Platform for Grid Connected Synchronous Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 Fariya Tabassum, Md. Shajid Hussain, and M. S. Rana Dynamic Power Flow Control Using Dual-Input Single-Output Non-Isolated DC-DC Converter for Renewable Energy Applications . . . 280 Saikat Das, Nirakar Nayak, and Amritesh Kumar Performance Analysis of Machine Learning, Deep Learning and Ensemble Techniques for Breast Cancer Diagnosis . . . . . . . . . . . . . 292 Piyush Sharma, Pradeep Laxkar, and Anuj Kumar Decarbonizing Indian Electricity Grid . . . . . . . . . . . . . . . . . . . . . . . . . . 312 Parvathy Sobha Future Mobility with eVTOL Personal Air Vehicle (PAV): Urban Air Mobility (UAM) Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . 323 M. Naga Praveen Babu, Sidvik Basa, Prasanth Kumar Duba, and P. Rajalakshmi Dam Water Discharge and Flood Prediction Cum Warning System . . . 338 V. Anantha Narayanan and Prashant R. Nair Virtual Inertia Control Strategy for High Renewable Energy-Integrated Interconnected Power Systems . . . . . . . . . . . . . . . . . 346 Anuoluwapo Aluko, Rudiren Pillay Carpanen, David Dorrell, and Evans Ojo

xii

Contents

Multimodal Visual Question Answering Using VizWiz Data; A Visual Assistant for the Blind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 B. Sreedha and Prashant R. Nair Design and Neural Control for Insect-Copter for Smooth Perching on Outdoor Vertical Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373 Sandeep Gupta and Laxmidhar Behera A 12T SRAM of 16 nm CMOS Technology Using One Sided Schmitt Trigger Inverter and Read Port . . . . . . . . . . . . . . . . . . . . . . . . 383 Komaladitya Challa and Vinay Kumar Pamula Multi-Stage Edge Detection for Generative Spatial Robotic Artwork . . . 392 Sukanya Nag, Deepsikha Bhattacharjee, Archisman Bhaumik, and Suman Deb Design and Analysis of Micro-grid Stability with Various DGs . . . . . . . 406 B. Devulal, M. Siva, D. Ravi Kumar, and V. Rajashekar Risk-Seeker Information Gap Decision Theory Based Smart Grid Operation Encompassing Demand Response . . . . . . . . . . . . . . . . . . . . . 415 Tanuj Rawat, K. R. Niazi, Sachin Sharma, and Jyotsna Singh Performance Analysis of Multi-user Diversity Schemes on Interference Limited D-GG Atmospheric Turbulence Channels . . . . . . . . . . . . . . . . . 426 Anu Goel and Richa Bhatia A Sparse-Dense HOG Window Sampling Technique for Fast Pedestrian Detection in Aerial Images . . . . . . . . . . . . . . . . . . . . . . . . . . 437 Ranjeet Kumar and Alok Kanti Deb Experimental Analysis for Distance Estimation Using RSSI in Industry 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451 Robin Singh Chouhan, Advait Kale, Anand Singh Rajawat, Rabindra Nath Shaw, and Ankush Ghosh Agriculture Field Security System Using Faster R-CNN . . . . . . . . . . . . 464 Vishesh Kumar Mishra, Sourov Bhowmick, and Sharzeel Saleem Automated Relational Triple Extraction from Unstructured Text Using Transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472 Akshay Hari and Priyanka Kumar A Modified SVPWM Strategy to Improve the Performance of Variable Frequency Induction Motor Drive . . . . . . . . . . . . . . . . . . . . 481 Nazmul Islam Nahin, Shuvra Prokash Biswas, and Md. Rabiul Islam A High Performance Multi-pulse AC-DC Converter for Adjustable Speed Motor Drives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494 Sharmin Shila, Shuvra Prokash Biswas, and Md. Rabiul Islam

Contents

xiii

Passivity-Based Control of Tidal Turbine Based PMSG Using Interconnection and Damping Assignment Approach . . . . . . . . . . . . . . . 505 Youcef Belkhier, Younes Sahri, Thiziri Makhlouf, Rabindra Nath Shaw, Mohit Bajaj, and Ankush Ghosh Recommendation System Based on EEG Emotion Recognition . . . . . . . 515 R. Vasanthradevi, R. Priyadharshini, P. Jai Rajesh, R. Reena, and R. Kalpana Design and Implementation of a Self-charging System to Improve the Operating Range in Quad-Motor EV . . . . . . . . . . . . . . . . . . . . . . . . 533 M. Chandra Mohan, A. Bright Selva Kumaran, V. S. Rohith, C. Sanjay, K. Prathap Reddy, and J. Jayashankari Design and Implementation of a Weed Removal Agriculture Robot . . . . 541 J. Dhanasekar, B. Sathish Kumar, S. Akash, P. Balamurugan, G. Vasanth, and B. Umamaheswari Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551

About the Editors

Dr. Saad Mekhilef is IET Fellow and IEEE Senior Member. He is Associate Editor of IEEE Transaction on Power Electronics and Journal of Power Electronics. He is Professor in the Department of Electrical Engineering, University of Malaya, since June 1999. He is currently Dean of the Faculty of Engineering and Director of Power Electronics and Renewable Energy Research Laboratory-PEARL. He is Author and Co-author of more than 400 publications in international journals and proceedings (253 ISI journal papers) and five books with more than 21000 citations and 73 H-index, and 116 Ph.D. and master students have graduated under his supervision. He has six patents. He frequently invited to give keynote lectures at international conferences. He was listed by Thomson Reuters (Clarivate Analytics) as one of the Highly Cited (Hi Ci) engineering researchers in the world and included in the Thomson Reuters’ The World’s Most Influential Scientific Minds: 2018. He is actively involved in industrial consultancy for major corporations in the power electronics projects. His research interests include power conversion techniques, control of power converters, renewable energy, and energy efficiency. Dr. Rabindra Nath Shaw is currently working as Director, International Relations, Bharath Institute of Higher Education & Research (Deemed to be University), Chennai, India. Before joining BIHER, he has served also Galgotias University as Director, IR&C. He is an alumnus of the Applied Physics Department, University of Calcutta, India. He is Senior Member of IEEE Industry Application Society, USA, and Fellow of Nikhil Bharat Shiksha Parishad, India. He is a global leader in organizing international conferences. His brand of world leading conference series includes IEEE International Conference on Computing, Power and Communication Technologies (GUCON), IEEE International Conference on Computing, Communication and Automation (ICCCA), IEEE IAS Global Conference on Emerging Technologies (GlobConET), International Conference on Electronics & Electrical Engineering (ICEEE), International Conference on Advances in Computing and Information Technology (ICACIT), etc. He holds the position of Conference Chair, Publication Chair, and Editor for these conferences. These conferences are held in collaboration with various international universities xv

xvi

About the Editors

like Aurel Vlaicu University of Arad, University of Malaya, University of Siena. Many world leaders are working with Dr. Shaw in these conferences. Most of these conferences are fully sponsored by IEEE Industry Applications Society, USA. He is also an expert in organizing international seminars/webinars/faculty development program in collaboration with leading institutes across the world. His research interests include optimization algorithms and machine learning techniques for power systems, IoT applications, renewable energy, and power electronics converters. He has published more than hundred twenty-five papers in Scopus/Web of Science/SCI indexed international journal/proceedings including high-impact factor journals in the field of renewable energy. He has also published several edited books from Springer and Elsevier publishing houses. He has worked on many national/international patents. He has successfully executed his duty in various positions like Governing Body Member, Centre in Charge, NBA Coordinator, University Examination Coordinator, University MOOC’s Coordinator, University Conference Coordinator, and Faculty-in-Charge, Centre of Excellence for Power Engineering and Clean Energy Integration. He has more than eleven years of teaching, research, and administrative experience in leading institutes like Motilal Nehru National Institute of Technology Allahabad, Jadavpur University, Gargi Memorial Institute of Technology, Brainware Group of Institution, Dream Institute of Technology. Prof. Pierluigi Siano received the M.Sc. degree in electronic engineering and the Ph.D. degree in information and electrical engineering from the University of Salerno, Salerno, Italy, in 2001 and 2006, respectively. He received the award as 2019 Highly Cited Researcher by ISI Web of Science Group. According to the study conducted by some Stanford University researchers, it has been included in the ranking of 100 thousand international scientists, belonging to different disciplinary fields and with greater scientific impact. The investigation was conducted by some Stanford University researchers who, starting from the SCOPUS database containing the list of scholars of the world classified by scientific production, developed a composite indicator that takes into account all possible bibliometric parameters (H-index, number publications, citations, journal relevance, sector relevance, etc.), in order to generate a wider ranking, focused on the scientific relevance of scientists, in relation to the activity of the last 23 years. He is Professor and Scientific Director of the Smart Grids and Smart Cities Laboratory with the Department of Management & Innovation Systems, University of Salerno. His research activities are centered on demand response, on the integration of distributed energy resources in smart grids and on planning and management of power systems. He has co-authored more than 450 papers including more than 200 international journal papers that received more than 7600 citations with an H-index equal to 44. He has been Chair of the IES TC on Smart Grids. He is Editor for the Power & Energy Society Section of IEEE Access, IEEE Transactions on Industrial Informatics, IEEE Transactions on Industrial Electronics, Open Journal of the IEEE IES and of IET Renewable Power Generation, Editor of the journal Smart Cities, MDPI Publisher since 2018, Editor of Intelligent Industrial Systems, Springer, 2014–2017.

Estimating the Levelized Cost of Electricity of the First Nuclear Power Plant in Bangladesh Abid Hossain Khan1(B) , Mehedi Hasan2 , M. Mizanur Rahman3 , and S. M. Sayeed Anowar4 1 Institute of Nuclear Power Engineering, Bangladesh University of Engineering and

Technology, 1000 Dhaka, Bangladesh [email protected] 2 Chemistry Division, Atomic Energy Center, Bangladesh Atomic Energy Commission, GPO Box No. 164, 1000 Dhaka, Bangladesh 3 Institute of Energy Science, Atomic Energy Research Establishment, Ganakbari, Savar, GPO Box No. 3787, 1000 Dhaka, Bangladesh 4 Department of Industrial and Production Engineering, Jashore University of Science and Technology, 7408 Jashore, Bangladesh

Abstract. In this work, an economic study of Rooppur Nuclear Power Plant is conducted. It is the first nuclear power plant being constructed in Bangladesh. Correlation from available literature is used to predict the net power output of VVER-1200 type nuclear power plant in the weather condition of Bangladesh. This predicted value is used to estimate the Levelized Cost of Electricity (LCOE) of the plant. Results reveal that the estimated LCOE for the plant is 91.19 $/MWh for 50 years operating life, lower than most of its conventional and renewable competitors. However, LCOE of the nuclear power plant becomes higher than that of coal-based power plants in the country if the plant lifetime is 20 years. This is, however, very unlikely since the average lifetime of a nuclear power plant is much longer. Therefore, Rooppur nuclear power plant is found to be a feasible option from geo-economic point of view. Keywords: Rooppur nuclear power plant · Levelized cost of electricity · Economic feasibility

1 Introduction Nuclear power plants (NPPs) are thought to be among the most realistic solutions to greenhouse effect due to their low carbon footprints [1]. Yet, the sector experienced a stagnation period in late twentieth and early twenty first century owing to accidents like Chernobyl disaster and Fukushima disaster, which have shaped a negative public opinion towards nuclear power [2, 3]. Still, the enhanced safety features of Gen III+ reactors have succeeded in convincing a portion of the world population that nuclear power can also be safe if design-in-depth is ensured [4]. Many countries are now going for nuclear to meet their ever-growing need of energy. Bangladesh, a middle-income © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 1–11, 2022. https://doi.org/10.1007/978-981-19-1742-4_1

2

A. H. Khan et al.

country, is in the same path to sustain its economic growth which is directly related to energy consumption, and thus, to energy production [5]. Rooppur NPP is the first nuclear power plant being constructed in Bangladesh. Russian VVER-1200 technology is to be adopted in this NPP. There are two units of this NPP and the total rated capacity of this power plant is 2400 MWe [6]. The first unit of this power plant is expected to be operational by 2023 [6]. The estimated cost of the plant is approximately 12.65 billion USD, 90% of which is to be financed by the Russian government on a 1.75% interest rate capped at 4% for 28 years with 10 years grace period [7]. The government of India is also expected to finance multiple projects at an interest rate of 1% for 20 years with 5 years grace period [8], Rooppur NPP being one of those projects. Nevertheless, the investment required for this power plant is much higher than most other NPPs being built across the world [7], although the low-interest rate on debt is quite favorable for the project. Therefore, an economic feasibility study of the nuclear power plant is very crucial. There are many approaches to studying the economic feasibility of a power plant. However, one of the most common parameters used to describe the feasibility of a power plant is the Levelized Cost of Electricity (LCOE), also called the Levelized Cost of Energy or Levelized Unit Electricity Cost (LUEC). It is a useful parameter for comparing different power generation options such as conventional, nuclear, renewable, etc., and predicts the possible cost of electricity per unit consumed [9]. There have been numerous studies focusing on LCOE to understand the prospect of various power generation technologies in the socio-economic scenario of Bangladesh [10–12]. However, these studies didn’t consider the effect of weather of Bangladesh on the performance of the power plants, which is actually a very important factor [13]. Thus, an economic feasibility study of the first NPP in Bangladesh considering the effects of both weather and existing socio-economic parameters of the country is yet to be done. In this work, the economic feasibility of Rooppur NPP is studied. At first, the important performance parameters of the NPP such as efficiency, output power etc. are calculated the weather condition of Bangladesh using the correlations proposed by Khan and Islam [13]. The optimum performance parameters obtained from the analyses are used to calculate the LCOE of Rooppur NPP for various plant operational lifetimes. Finally, a comparative study is conducted among Rooppur NPP and other conventional and renewable power plants in context to the socio-economic scenario of Bangladesh to determine the feasibility of the Rooppur project.

2 Methodology 2.1 Predicted Energy Generation from Rooppur NPP VVER-1200 is a Gen III+ Pressurized Water Reactor (PWR) developed by ROSATOM, Russia with increased safety features such as passive safety systems, inherent safety systems, double containment, core catcher, etc. [14]. PWR are the most common reactor type used worldwide [15]. The enhanced safety features of VVER-1200 are the key reasons behind choosing this reactor technology by the Government of Bangladesh for the Rooppur NPP. However, these enhancements came with an increased cost of installation and commissioning. While the investment cost for starting up a single reactor

Estimating the LCOE of the First Nuclear Power Plant in Bangladesh

3

unit of an NPP with output capacity in the range 700–1200 MWe was around 2–3.5 billion USD in the past for many countries [7], Rooppur NPP is expected to cost around 12.65 billion USD for two reactor units of 2400 MWe cumulative power output [7]. That’s why it is very important to study the feasibility of the first NPP in Bangladesh. Novovoronezh NPP-2 in Russia is the reference plant for Rooppur NPP. This power plant is also based on the VVER-1200 nuclear reactor [16]. Although they are identical in design, the weather conditions of the surroundings of the two plants are completely different. Novovoronezh NPP-2 can operate with 4–5 kPa condenser pressure as it is located in a cold weathered country. On the other hand, the condenser pressure is expected to be at least 7 kPa for Rooppur NPP since it is located in a tropical region country [13]. Khan and Islam [13] proposed a simplified Rankin cycle model for analyzing the secondary coolant loop of a VVER-1200-based NPP. The authors also developed correlations for multiple plant performance parameters such as plant efficiency, power output, condenser thermal load, etc. The developed correlation for power output is [13], W = 991.59 +

2203.97(2524.03 − hcond ,in ) 2524.03 − hcond ,out − ϑ(50 − P)

(1)

Here, hcond,in and hcond,out are the specific enthalpies of condenser inlet steam and water entering low-pressure pump respectively. Also, υ and P are specific volume of liquid water and condenser pressure respectively. Using the correlation, it is estimated that the net power output of Rooppur NPP should be around 1152 MWe , which is much below its rated output of 1198 MWe . This limitation is also acknowledged by the manufacturer. Thus, Rooppur NPP will have decreased power production and subsequent increase in its investment cost per unit energy produced due to the weather condition of Bangladesh. 2.2 Levelized Cost of Electricity for Rooppur NPP Although NPPs and renewable power plants have much lower thermal efficiency than other thermal power plants like Combined Cycle Power Plants (CCPP), Coal-based power plants, etc. [13], they are preferred on many occasions depending on their economic feasibility. One way to compare the economic competitiveness of different power generation options is to compare their Levelized Cost of Electricity (LCOE). It denotes the ratio of the sum of all the levelized costs of a power plant over its lifetime to the sum of the levelized energy production during the same period [9]. LCOE for a NPP may be calculated using Eq. (2), TL LCOE =

(It +OM t +FC t +DC t ) (1+rAI )t TL Et t=1 (1+rAI )t

t=1

(2)

Here, I t , OM t , FC t , DC t and E t , denote investment expenditures (in USD), operation and maintenance costs (in USD), fuel costs (in USD), average annual decommissioning cost (in USD) and electricity generated (in MWh) at year t respectively. r AI denotes the annual discount rate of money due to inflation. Also, TL is the total life of the plant from

4

A. H. Khan et al.

the beginning of installation to the end of decommissioning. Ignoring the cooling down time of a NPP after shutdown, TL is given by, TL = IT + OLT + DT

(3)

Here, IT is installation time, OLT is operational lifetime and DT is decommissioning time. In this work, the LCOE of Rooppur NPP has been calculated for different OLT s within the range of 10–50 years to predict the minimum required years of operation for being economically competitive to other power options. The investment expenditure may account for multiple areas of expenses such as installation and commissioning costs, interest on loan, annual allowance for investment return, decommissioning costs, income taxes, etc. One or more of these expenses may be omitted depending on the type and location of the power plant. Thus, there is no universal equation of investment expenditure. The investment expenditure considered in this work at a specific year t is given by, It = IC t + YIOLt + ROCEAt + ITX t ; t = 1, . . . TL

(4)

Here, IC t is the average annual installation and commissioning cost (in USD), YIOL t is the annual payable interest amount on loan (in USD), ROCEAt is the annual return on common equity allowance, and ITX t is income tax (in USD) at year t. Average annual installation and commissioning cost may be obtained from Eq. (5), IC t =

TNI ; t = 1, . . . IT IT

(5)

Here, TNI is the total net investment. For Rooppur NPP, the loan amount from Russia is approximately 11.38 billion USD on a 1.75% annual interest rate and 28 years duration [7]. Equation (5) considers the worst-case scenario where the plant takes the total allocated time limit to repay the debt. The usual repay time is 7–8 years [12]. In this study, this longer than usual loan repay time has been considered, keeping in mind the socio-economic instability of the country. The annual decommissioning cost may be obtained from Eq. (6), DC t =

TDC ; t = (TL − DT + 1), . . . TL DT

(6)

Here, TDC is the total decommissioning cost. To simplify calculations, it has been assumed that annual decommissioning cost is constant. Finally, the income tax rate is assumed 25% [19]. The annual operation and maintenance (O&M) cost of an NPP may be calculated using Eq. (7), OM t = OM 0 (1 + rOM )t ; t = (IT + 1), . . . (TL − DT )

(7)

Here, OM 0 is the operation and maintenance cost at the beginning year of operation and r OM is the operation and maintenance discount rate. OM 0 may be calculated using Eq. (8), OM 0 = COM × NU × 1200 × 365 × 24 × PUF

(8)

Estimating the LCOE of the First Nuclear Power Plant in Bangladesh

5

Here COM, NU, and PUF are the O&M cost per MWh, number of plant units (02 for Rooppur NPP), and plant utilization factor respectively. Similarly, the annual fuel cost may be calculated using Eq. (9), FC t = FC 0 (1 + rF )t ; t = (IT + 1), . . . (TL − DT )

(9)

Here FC 0 is the fuel cost at the beginning year of operation and r F is the fuel discount rate. FC 0 may be calculated from Eq. (10), FC 0 = CNF × NU × 1200 × 365 × 24 × PUF

(10)

Here CNF is the cost of nuclear fuel. In most cases, a country has to keep an allowance in the LCOE for radioactive waste disposal. This allowance is called a waste fee. Including the waste fee, the corrected LCOE for the NPP becomes, LCOE IWF = LCOE EWF + WF

(11)

Here, LCOE IWF and LCOE EWF denote LCOE including and excluding waste fees respectively. WF denotes the waste fee, taken as 1.0 $/MWh, following the regulations of the US government [18]. All the economic constants and assumptions considered in this study are summarized in Table 1. Table 1. Economic constants and assumptions [17–19] Economic parameter

Value

Installation and commissioning time, IT (years)

8

Plant operational lifetime, OLT (years)

10–50

Decommissioning time, DT (years)

5

Debt terms, LT (years)

28

Annual discount rate of money, r AI (%)

5.0

Fuel discount rate, r F (%)

−5.0

O&M discount rate, r OM (%)

5.0

Income tax rate, r TX (%)

25.0

Debt fraction, r D (%)

90.0

Return on common equity allowance, r CE (%)

12.5

Cost of fuel, CNF ($/MWh)

4.1

Cost of O&M, COM ($/MWh)

10.0

Total net investment, TNI (billion $)

12.65

Total decommissioning cost, TDC (million $)

700

Radioactive waste fee, WF ($/MWh)

1.0

6

A. H. Khan et al.

3 Results and Discussion Figure 1 shows the LCOEs of Rooppur NPP for different plant operating lifetimes. The Figure also presents LCOEs for a hypothetical VVER-1200 NPP that has the same economic parameters as Rooppur NPP but is located in a cold-weather country. This was done to compare these two NPPs and identify the economic disadvantage that Rooppur NPP may face due to the weather condition of Bangladesh. From Fig. 1, it may be observed that the LCOEs of Rooppur NPP were higher than those of the hypothetical NPP, as one would expect due to the lower power output. The difference between the LCOEs decreased with the increase in the operating lifetime of the plant. The minimum difference was observed to be 2.52 $/MWh and the maximum difference was observed to be 7.02 $/MWh. Thus, it may be opined that the economic disadvantage of Rooppur NPP may be minimized by increasing plant operating lifetime. It is noteworthy that after 30 years of operational lifetime, the LCOE became less responsive to further increase in plant operational lifetime. The LCOE of Rooppur NPP was estimated to be within the range of 91.19–179.25 $/MWh. The predicted values were somewhat lower than the predicted range of LCOE for an NPP in the USA, typically within 118.0–192.0 $/MWh [20]. This may be due to the comparatively lower interest rate on debt for Rooppur NPP (only 1.75% compared to 4–5% in the USA [18]) and a lower income tax rate in Bangladesh (25% compared to around 38% in many developed countries [18]). Further subsidy on this sector may make the power plant even more feasible from an economic point of view. The global average value in 2015 was found to be 95.9 $/MWh in the available literature [12], very close to but slightly higher than the predicted value of 91.19 $/MWh for 50 years plant operational lifetime. The results of a study conducted by Sieed et al. [21] found the LCOE for Rooppur NPP aound 94.8 $/MWh for a plant lifetime of 60 years using INPRO method, again very similar to the findings of this study. An exceptional contradiction is observed between the predicted values of LCOEs in this study and the ones in a study conducted by Islam and Bhuiyan [12]. The later one predicted the LUEC for Rooppur NPP in the range 43.8–82.5 $/MWh for 60 years plant lifetime using FINPLAN model. Although the upper limit of LUEC obtained in the study is realistically close to the world average, the lower limit of 43.8 $/MWh is too low, even for the leaders of nuclear power sector like USA, RUSSIA, France, etc. Bangladesh is a newcomer country and lacks necessary technology as well as technical knowledge in this field. The fuel and O&M costs assumed in the study of Islam and Bhuiyan [13] for obtaining this lower limit of LUEC are, therefore, very difficult to achieve considering the absence of local source of nuclear fuel and the technological dependence of the country. In contrast, this work presents a somewhat conservative range for the LCOE of Rooppur NPP [22, 23].

Estimating the LCOE of the First Nuclear Power Plant in Bangladesh

7

A comparative study among different power options for Bangladesh is presented in Fig. 2. From Fig. 2, it may be observed that for 50 years operating lifetime, Rooppur NPP is predicted to have lower LCOE (91.19 $/MWh) than almost all of its conventional and renewable competitors, except for solar PV (91.00 $/MWh), hydroelectric (14.00 $/MWh), and gas-fired (28.00 $/MWh) power plants. However, both solar PV and hydroelectric power plants require a huge land area (31.35 and 237.55 acres/MW respectively), as shown in Table 2, which is not feasible for a densely populated country like Bangladesh; although this is not the case for gas-fired power plant (0.34 acres/MW). A typical NPP requires 0.93 acres of land per MW capacity. Thus, gas-fired is the only power option that is better than Rooppur NPP from a geo-economic point of view, which is why it is the most common type observed in Bangladesh till date. The only reason behind choosing an NPP over a gas-fired plant in Bangladesh for future projects is the rapid depletion of natural gas from the gas-fields in the country. Bangladesh doesn’t have enough gas reserved in its gas-fields to support its growing energy demand in the future. Another reason behind preferring nuclear power over gas and coal, another close competitor, is that both gas and coal power plants emit significantly higher greenhouse gas (GHG) (499 and 888 t-CO2/GWh respectively) compared to an NPP (29 t-CO2/GWh). Even solar PVs have higher GHG emissions than a typical NPP, as seen in Table 10. Only hydro has less GHG emission than NPP, but their land requirements make them unrealistic in the context of Bangladesh. Therefore, it may be stated that Rooppur NPP is geo-economically feasible for Bangladesh and, perhaps, one of the best power generation options for fighting climate change besides keeping the economy of the country stable. Finally, it may be observed from Fig. 2 that the LCOE of Rooppur NPP is slightly higher than that of a coal-fired power plant in Bangladesh if the operating lifetime is 20 years. This difference may be neglected considering the huge difference in the GHG emissions for the two plants. However, a further decrease in the operational lifetime of Rooppur NPP will increase the LCOE significantly, taking the NPP almost out of the competition against coal-fired power plants. Thus, it is suggested that Rooppur NPP has to operate for 20 years or higher to be economically feasible for Bangladesh. Table 2. Comparison of GHG-emission and land requirement of different power generation options for Bangladesh. Type of power source GHG emission (t CO2 /GWh) Typical land requirement (acres/MW) Proposed Rooppur NPP

29 [1]

Plant: 0.90 [23] Storage: 0.03 [23]

Solar PV

85 [1]

Total: 31.35 [23]

Rooftop solar PV

85 [1]

–

CSP

–

Total: 45.25 [23]

Wind

26 [1]

Total: 60.00 [23] (continued)

8

A. H. Khan et al. Table 2. (continued)

Type of power source GHG emission (t CO2 /GWh) Typical land requirement (acres/MW) Existing (report of 2016) Hydro

26 [1]

Excluding Catchment Area: 237.55 [23]

Gas

499 [1]

Plant: 0.34 [23]

Coal

888 [1]

Plant: 0.70 [23] Mine and Storage: 0.72 [23]

HFO

733 [1]

–

Diesel

733 [1]

–

Fig. 1. Comparison of LCOEs for different NPP operational lifetimes

Estimating the LCOE of the First Nuclear Power Plant in Bangladesh

9

Fig. 2. Comparison of LCOE of different power generation options for Bangladesh

4 Conclusion This work attempts to estimate the LCOE of Rooppur Nuclear Power Plant. While doing so, the effect of weather condition of Bangladesh on the net power output of the plant is considered. The correlations proposed by Khan and Islam are utilized to predict the reduction in net power output. After that, LCOE is calculated for different plant operational lifetimes. Results reveal that the LCOE of Rooppur NPP is within the range of 91.19–179.25 $/MWh. This value is comparable to the LCOEs of other NPPs in different countries. Results also suggest that Rooppur NPP is expected to have lower LCOE than most of its conventional and renewable competitors if the plant operational life is 50 years, except for solar PV, gas-fired power plants, and hydroelectric power plants. However, considering GHG emission and land requirement, Rooppur NPP is found as the most feasible and promising power generation option for Bangladesh. Another important finding is that Rooppur NPP should operate for 20 years or longer to remain economically competitive with the coal-fired power plants.

10

A. H. Khan et al.

This study is based on a specific nuclear reactor technology, which is VVER-1200. Further studies may be conducted to evaluate the effect of weather on the performance of NPPs with other reactor designs. A comparative study with different reactor technologies may also be carried out to find the best performing technology in the weather condition of Bangladesh, both from a thermodynamic and economic standpoint.

References 1. World Nuclear Association: Comparison of lifecycle greenhouse gas emissions of various electricity generation sources. WNA Report, London (2011) 2. Iimura, A., Cross, J.S.: Influence of safety risk perception on post-Fukushima generation mix and its policy implications in Japan. Asia Pac. Policy Stud. 3(3), 518–532 (2016) 3. Misnon, F.A., et al.: Malaysian public perception towards nuclear power energy-related issues. In: AIP Conference Proceedings, vol. 1799, no. 1. AIP Publishing LLC (2017) 4. Sun, C., Zhu, X., Meng, X.: Post-Fukushima public acceptance on resuming the nuclear power program in China. Renew. Sustain. Energy Rev. 62, 685–694 (2016) 5. Alam, F., Sarkar, R., Chowdhury, H.: Nuclear power plants in emerging economies and human resource development: a review. Energy Procedia 160, 3–10 (2019) 6. Sabhasachi, S., et al.: Rooppur nuclear power plant: current status & feasibility. Strojníckyˇcasopis J. Mech. Eng. 68(3), 167–182 (2018) 7. Sakib, K.N.: Nuclear power plant in bangladesh and the much talked about rooppur project. Glob. J. Res. Eng. 15(2), 17–22 (2015) 8. India commits $4.5 billion loan for projects in Bangladesh, including Rooppur Nuclear Power Plant. https://www.nuclearasia.com/news/india-commits-4-5-billion-loan-projectsbangladesh-including-rooppur-nuclear-power-plant/1199/ 9. De Roo, G., Parsons, J.E.: A methodology for calculating the levelized cost of electricity in nuclear power systems with fuel recycling. Energy Econ. 33(5), 826–839 (2011) 10. Khan, E.U., et al.: Techno-economic analysis of small scale biogas based polygeneration systems: Bangladesh case study. Sustain. Energy Technol. Assess. 7, 68–78 (2014) 11. Shaw, R.N., Walde, P., Ghosh, A.: Review and analysis of photovoltaic arrays with different configuration system in partial shadowing condition. Int. J. Adv. Sci. Technol. 29(9s), 2945– 2956 (2020) 12. Shafiqul, I.M., Bhuiyan, T.H.: Assessment of costs of nuclear power in Bangladesh. Nucl. Energy Technol. 6(3), 181–194 (2020) 13. Khan, A.H., Islam, M.S.: Prediction of thermal efficiency loss in nuclear power plants due to weather conditions in tropical region. Energy Procedia 160, 84–91 (2019) 14. The State Atomic Energy Corporation ROSATOM. The VVER today: Evolution, Design, Safety. http://www.rosatom.ru/upload/iblock/0be/0be1220af25741375138ecd1afb18743.pdf 15. Anglart, H.: Applied Reactor Technology (2011) 16. Asmolov, V.G., et al.: New generation first-of-the kind unit–VVER-1200 design features. Nucl. Energy Technol. 3(4), 260–269 (2017) 17. Rothwell, G.: Nuclear power economics (2004) 18. Varro, L., Ha, J.: Projected costs of generating electricity–2015 Edition, Paris, France (2015) 19. Income Tax Nirdeshika (2020–2021). http://nbr.gov.bd/publications/income-tax/eng 20. Lazard’s Levelized Cost of Energy Analysis-Version 13.0. https://www.lazard.com/media/ 451086/lazards-levelized-cost-of-energy-version-130-vf.pdf 21. Sieed, J., Hossain, S., Kabir, K.A.: Application of INPRO methodology to assess economic feasibility of proposed rooppur nuclear power plant. In: International Conference on Materials, Electronics and Information Engineering, ICMEIE, Rajshahi, Bangladesh (2015)

Estimating the LCOE of the First Nuclear Power Plant in Bangladesh

11

22. Shiraishi, K., Shirley, R.G., Kammen, D.M.: Geospatial multi-criteria analysis for identifying high priority clean energy investment opportunities: a case study on land-use conflict in Bangladesh. Appl. Energy 235, 1457–1467 (2019) 23. Stevens, L., et al.: The footprint of energy: land use of US electricity production. STRATA: Logan, UT, USA (2017)

Dynamic and Transient Stability Enhancement of Multi Area Power Systems Using Fuzzy Logic Control Against Load Disturbances Venu Yarlagadda1(B) , Rashmi Kapoor1 , Chava Sunil Kumar2 , and Giriprasad Ambati1 1 VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, India 2 BVRIT Hyderabad College of Engineering for Women, Hyderabad, India

Abstract. Modern Power systems are designed for the fine tuning of frequency and less tolerance for system frequency deviation from nominal value. The Power System is dynamically subjected to the small perturbations of load leading to non-oscillatory Instability due to insufficient damping. The article entente the multi area load frequency control and dynamic and transient stability analysis. It dispenses the simulation of three area power systems with medium and large perturbations of load, with three cases for both kinds of power systems. Case1 without any controller, case2 with PI controller and case3 with Fuzzy Controllers. The simulation is carried out for three area power system with all three cases. In this article the simulation results of three area power system for all three cases have been presented. The simulation results demonstrate the effectiveness of Fuzzy Control perpetuates the frequency with in the endurable range of frequency and subsequently it ensures the dynamic as well as transient stability of both the Power systems against load disturbances. Keywords: Fuzzy logic · Load frequency control · Small signal stability · Stability improvement · Single area power system · Two area power systems

1 Introduction The management of modern power systems became most difficult task to ensure the system security and reliability with good quality. India is aiming at forming a single national grid with fine tuning of frequency all over the system, constituted with many of the power system components and distinct loads. The load perturbation on all over the grid is dynamically varying, leads to frequency deviations over each of the control areas. The focus of this article is to maintain the constant or tolerable range of frequency in all control areas of power grid. The single area power system is modelled as a single transfer function block diagram, comprised of governor, turbine and generator including load transfer function models. The two-area model is composed of two single areas with different individual block models of different parameters connected together with a tie line model. The Simulink models have been developed for both single area and two area power systems to carryout simulation study and analysis. These simulation model of three area power system with three different case studies viz. case I: with 5% load © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 12–29, 2022. https://doi.org/10.1007/978-981-19-1742-4_2

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

13

disturbance with all three sub cases of (a), (b) and (c) without any controller, with PI controller and with Fuzzy controller respectively. Case II: with 5% load disturbance with all three sub cases of (a), (b) and (c) without any controller, with PI controller and with Fuzzy controller respectively. Case III: with 5% load disturbance with all three sub cases of (a), (b) and (c) without any controller, with PI controller and with Fuzzy controller respectively [1–6].

2 Single Area Power System The power system network comprehends of many distinct loads with huge transmission and distribution networks. The complete network is categorized into different control areas based on frequency deviation over the part of power network. The small signal model is derived with individual power system components viz. Speed Governor, Turbine and Generator including load models. 2.1 Speed Governor Model The load on power system is continuously varying may lead to speed deviation subsequently it leads to frequency fluctuations. The speed governing system is used to control these fluctuations. The speed governing system composed of fly ball speed governor, hydraulic amplifier and speed changer with linkage mechanism. The governor has two basic inputs, one is changes in reference power and the second is frequency variations are modelled into an equation, which describes both of these inputs as portray in the below equation and Fig. 1. Time constant (Tg ) (as expressed in terms of a constant, depends on orifice, cylindrical geometries and fluid pressure of Hydraulic Amplifier and its typical value is in the range of 0.1 to 0.6 s [3–7]. Pv (s) = [Pref − 1R F(s)]

Kg 1 + Tg

(1)

2.2 Turbine Model Turbine model has been derived with a single stage turbine with a single time constant of Tt and typical value is in the range of 0.3 to 0.7 s. The transfer function model of a single time constant turbine model is depicted in below equations (Fig. 2). Pm (s) = Pt Tt =

Kt 1 + Tt

1 Kh

(2) (3)

14

V. Yarlagadda et al.

Fig. 1. Governing system model

Fig. 2. Small signal model of turbine

2.3 Generator Load Model The incremental power input to the system can fractionated towards stored kinetic energy in the generator being proportional to square of the frequency and load changes of sensitive loads due to frequency deviations. These two events have been modelled and derived the transfer function of generator load model as depicted in the following Eq. (4) and (5) below. The inertia constant is designated as H and its typical value is about 2 to 10 s, hence the power system time constant i.e. Tps lies in the range of 4 to 20 s. The following equations illustrate expressions for Tps and Kps in terms of inertia constant (H) and load damping factor (D) [5–12] (Fig. 3). TPS =

2H DF O

(4)

1 D

(5)

KPS =

[Pm (s)−[Pm (s) =

2HP dF F dt

F(s) = [Pm (s) − Pe (s)

Kps ] 1 + Tps

(6) (7)

Fig. 3. Generator load model

The model of LFC as illustrated in Fig. 4, LFC with PI Controller is illustrated by the Figs. 5 and 6 shows the Single Area Load Frequency Control with Fuzzy Controller [4–12].

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

15

Fig. 4. LFC without Controller

Fig. 5. LFC with PI Controller

Fig. 6. LFC with Fuzzy Controller

3 Single Area Power System Power Systems can be sub divided into number of areas through tie lines, without out loss of generality, consider a simple two area power system with a single tie line as depicted

16

V. Yarlagadda et al.

by Fig. 7 as shown below. The power transfer of the tie line is P12 is given in Eq. (10), the synchronizing torque coefficient T12 as illustrated in Eq. (11), subsequently expressions have been derived for Area Control Error (ACE1 and ACE2 ) in terms of P12 and P21, bias factors B1 and B2 respectively as shown in equations from (12) to (15). Figure 8 illustrates the Block diagram model of Fuzzy controller based two area power system, with fuzzy based controller in each of the areas respectively [1–8] (Figs. 9 and 10).

Fig. 7. Two area power system with tie line

P12 =

V1 V2 sin(δ) X12

(8)

T12 =

V1 V2 cos(δ) X12

(9)



P12 = 2π P12

V1 V2 cos(δ) X12  = 2π FT12 F

P12 (s) = P21 (s) = −

2πT12 F(s) s

2πT12 F(s) = −P12 (s) s

(10) (11) (12) (13)

ACE 1 = P12 + B1 F1

(14)

ACE 2 = P21 + B2 F2

(15)

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

17

Fig. 8. Block diagram model of Fuzzy controller based two area power system

Fig. 9. Three Area Power System with tie lines

Fig. 10. Block diagram model of Fuzzy controller based multi area power system showing only one area signals

3.1 Tie Line Bias Control of Multi Area Power Systems Each of the control area is connected with as many areas as possible in interconnected systems, with tie lines. Consider one area in the interconnected system, the net interchange with other control areas equal to the sum of tie line powers through outgoing tie lines and is given by the following equation below Eq. (16) similarly other area

18

V. Yarlagadda et al.

controllers can be obtained. ACE i =



Pij + Bi Fi

(16)

4 Power System Stability Power system stability is defined as a system ability to regain its initial equilibrium state after being subjected to a disturbance and its classification are given in Fig. 11.

Fig. 11. Power system stability stratification

Fig. 12. Power angle curve

4.1 Rotor Angle Stability The system should remain in synchronism even after being subjected to the disturbance, which involves output power oscillates reflected in rotor oscillations. 4.2 Rotor Angle Stability Relation between the power and angular position of a rotor in synchronous machine is nonlinear relation, when the synchronous generator is feeding a synchronous motor through transmission line. The power transferred to the motor from the generator is depends on the function of angular displacement between rotors of the generator and motor this is because of Motor internal angle, generator internal angle, and the angular displacement between motor and generator terminal voltage and power flow is given by the Eq. 17. P=

V1 V2 sinδ XT

(17)

This equation say that the power transferred to a motor from generator is maximum when the angle is 90°, if the angle is further increased beyond 90°, power transferred starts decreasing is shown in Fig. 12. The maximum power transferred is directly proportional to machine internal voltage [8–12]. Pmax =

V1 V2 XT

(18)

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

19

4.3 Dynamic Stability Whenever the synchronous machine is subjected to small and medium load perturbations then the machine is subjected to rotor oscillations. These oscillations will die out very soon due to damping provided in the machine or external damping provided by the controller or compensator. The swing curves of the machine will illustrate the dynamic instability mechanism of the machine for such disturbances as illustrated in the below Fig. 13(a) and (b). The Fig. 13(a) depicts that the system is stable for small and medium disturbances with quiet effective damping and is unstable for medium disturbances with insufficient damping. 4.4 Transient Stability Whenever the synchronous machine is subjected to medium and large load variations or contingencies then the machine is subjected to wide rotor oscillations. These oscillations will die out very soon due to damping provided in the machine or external damping provided by the controller or compensator. The swing curves of the machine will illustrate the transient instability mechanism of the machine for such disturbances as illustrated in the below Fig. 13(a) and (c). The Fig. 13(a) depicts that the system is stable for medium disturbances with suitable controller and is unstable for large disturbances without any controller [1–12].

Fig. 13. Power system stability swing curves (a) Stable System (b) Oscillatory Unstable System or dynamic instability and (c) Non-oscillatory Instability or Transient Instability

5 Fuzzy Logic Control Fuzzy control is used when vaguness in the decission making is present or when non linearities are involved in the system dynamics. Figure 14 shows Mamdani Rule based Fuzzy Logic Controller input and output functional relationship in simulink model and Fig. 15 3-d surface of the Fuzzy rules. The Table 1 shows the Fuzzy rules table for the single and two area power systems. There are two input membership functions of Fuzzy logic controller, one is error i.e. the reference voltage minus actual voltage of seven input triangular membership functions varying range of + or −16%. The second is derivative of error seven input triangular membership functions varying range of + or −16% as

20

V. Yarlagadda et al.

depicted in Fig. 14 and seven output triangular membership functions varying range of + or −18%. The Fuzzy rules [6–12]. Table 1 shows the 49 rules framed with input1 variables viz. nb1, nm1, ns1, zo1, ps1, pm1 and pb1, similarly input2 variables viz. nb2, nm2, ns2, zo2, ps2, pm2 and pb2 and output variables viz. nb, nm, ns, zo, ps, pm and pb [4–11].

Fig. 14. Triangular input and output membership functions of fuzzy controller

Fig. 15. Fuzzy 3-D rule surface

Table 1. Fuzzy rule table de/dt

e nb1

nm1

ns1

zo1

ps1

pm1

pb1

nb2

n1

n1

n1

n1

n2

n3

zo

nm2

n1

n1

n1

n2

n2

Zo

p3

ns2

n1

n1

n2

n3

zo

p3

p2

zo2

n1

n2

n3

zo

p3

p2

p2

ps2

n2

ns

zo

p3

p2

p1

p1

pm2

n3

zo

p3

p2

p1

p1

p1

pb2

zo

p3

p1

p1

p1

p1

p1

6 Case Study and Simulation Results The single area and two area power system Simulink models have been derived from the small signal transfer function models of both areas. Both of these areas have been simulated for three cases, one is without controller, second is with PI controller and third is with Fuzzy logic based controller. In subsequent part of the article single area system has been presented and succeeding part describes the simulation results of two area system.

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

21

6.1 Simulation Results of Three Area Power System The Simulink models have been developed for three area power system for all three cases, case I (a): without any controller as depicted by the Fig. 16, case I (b): Three area system with PI Controller as shown in Fig. 17 and case I (c): Single area system with Fuzzy Controller as illustrated in Fig. 18.

Fig. 16. Three area system Simulink model without Controller

Fig. 17. Three area system Simulink model with PI Controller

Fig. 18. Three area system with Fuzzy Controller

22

V. Yarlagadda et al.

Case I: 5% disturbance results of three area system Area1 frequency deviations with 5% disturbance In Case I results have been presented with 5% disturbance only with sub cases of Case I-1 for area1, Case I-2 for area2 and Case I-3 for area3 respectively. Case I-(a) Area 1 without controller error is −15% which is not acceptable and unstable, Case I-(b) with PI Controller the peak undershoot is −22% and settling time of 25 s and Case I-(c) with fuzzy controller the peak undershoot is −12.5% and settling time of 14 s with zero steady state error which is most accepted one among all systems as illustrated by the below Fig. 19.

Fig. 19. Area1 frequency deviations with 5% disturbance

Case I-2 (a) Without controller error is −19% which is not acceptable and unstable, Case I-3 (a) with PI Controller the peak undershoot is −29% and settling time of 25 s and Case I-3 (a) with fuzzy controller the peak undershoot is −18% and settling time of 14 s with zero steady state error as depicted by Fig. 20 below. Case I-(a) without controller error is −22% which is not acceptable and unstable, Case I-3 (b) with PI Controller the peak undershoot is −43% and settling time of 25 s and Case I-3 (c) with fuzzy controller the peak undershoot is −23% and settling time of 14 s with zero steady state error and Fig. 21 shows the Area3 frequency deviations with 5% disturbance. Case I-4 the power and delta deviation without controller it is unstable and with controllers the system is stable for small load disturbances in Area1 and the deviation is very small for fuzzy controller as depicted in Figs. 22 and 23 below. Case II: 10% disturbance results of three area system Case II-1 (a) without controller error is −29% which is not acceptable and unstable, Case II-1 (b) with PI Controller the peak undershoot is −49% and settling time of 15 s and Case II-1 (c) with fuzzy controller the peak undershoot is −11.5% and settling time of 6 s with zero steady state error as illustrated in the below Fig. 24.

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

Fig. 20. Area2 frequency deviations with 5% disturbance

Fig. 21. Area3 frequency deviations with 5% disturbance

Fig. 22. Tie line Power in area1 without, with PI and with Fuzzy Controllers

23

24

V. Yarlagadda et al.

Fig. 23. Area1 deviation in delta1, delta2 and delta3 without, with PI and with Fuzzy Controllers

Fig. 24. Area1 frequency deviations with 10% disturbance

Case II-1 (a) Without controller error is −39% which is not acceptable and unstable, Case II-2 (b) with PI Controller the peak undershoot is −64% and settling time of 16 s and Case II-2 (c) with fuzzy controller the peak undershoot is −11.5% and settling time of 12 s with zero steady state error as Fig. 25 depicts Area2 frequency deviations with 10% disturbance. Case II-3 (a) Without controller error is −49% which is not acceptable and unstable, Case II-3 (b) with PI Controller the peak undershoot is −83% and settling time of 16 s and Case II-3 (c) with fuzzy controller the peak undershoot is −40% and settling time of 12 s with zero steady state error as Fig. 26 shows Area3 frequency deviations with 10% disturbance. Case II-(d) the power and delta deviation without controller it is unstable and with controllers the system is stable for small load disturbances in Area1 and the deviation is very small for fuzzy controller as depicted in Figs. 27 and 28 respectively.

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

Fig. 25. Area2 frequency deviations with 10% disturbance

Fig. 26. Area3 frequency deviations with 10% disturbance

Fig. 27. Tie line Power in area2 without, with PI and with Fuzzy Controllers

25

26

V. Yarlagadda et al.

Fig. 28. Area2 deviation in delta4, delta5 and delta6 without, with PI and with Fuzzy Controllers, without controller it is unstable and with controllers the system is stable for small load disturbances in Area2

Case III: 30% disturbance results of three area system Transient Stability Analysis Case III-1 (a) Without controller error is −88% which is not acceptable and unstable, Case III-1 (b) with PI Controller the peak undershoot is −130% and settling time of 25 s and Case III-1 (c) with fuzzy controller the peak undershoot is −29% and settling time of 5 s with 15% steady state error as Fig. 29 depicts Area1 frequency deviations with 30% disturbance.

Fig. 29. Area1 frequency deviations with 30% disturbance

Case III-2 (a) Without controller error is −148% which is not acceptable and unstable, Case III-3 (b) with PI Controller the peak undershoot is −252% and settling time of 23 s and Case III-3 (c) with fuzzy controller the peak undershoot is −15% and settling time of 5.7 s with a small acceptable steady state error as Figs. 30 and 31 illustrates Area2 and Area3 frequency deviations with 30% disturbance respectively.

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

27

Fig. 30. Area2 frequency deviations with 30% disturbance

Fig. 31. Area3 frequency deviations with 30% disturbance

Case III-4 the power and delta deviation without controller it is unstable and with controllers the system is stable for small load disturbances in Area1 and the deviation is very small for fuzzy controller as depicted in Figs. 32 and 33 respectively.

Fig. 32. Tie line Power in area3 without, with PI and with Fuzzy Controllers

28

V. Yarlagadda et al.

Fig. 33. Area3 deviation in delta4, delta5 and delta6 without, with PI and with Fuzzy Controllers

7 Conclusions The article entente the three area load frequency control and dynamic and transient stability analysis. It dispenses the simulation of three area systems with small, medium and large load disturbances, with three cases. Case I with 5% load disturbances, Case II with 10% and case III with 30% load disturbances. In each of the three cases, there are three sub cases (a), (b) and (c) sub case (a) without any controller, sub case (b) with PI controller and sub case (c) with Fuzzy Controllers. For all three cases. The simulation is carried out for all three cases of three area power system with all three sub cases of (a), (b) and (c). In the first part of the case study, the simulation results of three area power system for all three cases have been presented. The frequency deviation for the three area system for all three cases shows that the system is working well with less settling time and zero steady state error, when it concerned with PI controller, the response is showing that the steady state error is there but still it growing and is not completely accepted as far as performance is concerned and without any controller the system performance is not accepted and it is unstable. In the second part, the simulation results of three area power system for last part of all three cases have been presented. The power and swing curves being illustrated that for small, medium and also for large disturbances the system is completely stable with fuzzy logic controller and where as it is not accepted with PI controller and completely unstable without controller.

References 1. Yarlagadda, V., Ambati, G.P., Prasad, E.S., Veeresham, K., Radhika, G.: Synchronous and voltage stability improvement using SVC and TCSC and its coordination control. Des. Eng. 2021(6), 3624–3635 (2021). ISSN 0011-9342 2. Yarlagadda, V., Ravi Kumar, D., Shiva Prasad, E., Ambati, G.P., Gadupudi, L.: Prototype models of FACTS controllers and its optimal sizing and placements in large scale power systems using voltage stability indices. Des. Eng. 2021(6), 3636–3659 (2021). ISSN 00119342

Dynamic and Transient Stability Enhancement of Multi Area Power Systems

29

3. Yarlagadda, V., Turaka, N., Ambati, G.P., Poornima, S.: Optimization of voltage stability based shunt and series compensation using PSO. Des. Eng. 2021(7), 8679–8694 (2021) 4. Nayak, J.R., Pati, T.K., Sahu, B.K., Kar, S.K.: Fuzzy-PID controller optimized TLBO algorithm on automatic generation control of a two-area interconnected power system. In: 2015 International Conference on Circuits, Power and Computing Technologies [ICCPCT-2015], pp. 1–4 (2015). https://doi.org/10.1109/ICCPCT.2015.7159427 5. Doley, R., Ghosh, S.: Application of PID and fuzzy based controllers for load frequency control of a single – area and double – area power systems. In: 2019 5th International Conference on Advances in Electrical Engineering (ICAEE), pp. 479–484 (2019). https://doi.org/ 10.1109/ICAEE48663.2019.8975568 6. Pati, T.K., Nayak, J.R., Sahu, B.K.: Application of TLBO algorithm to study the performance of automatic generation control of a two-area multi-units interconnected power system. In: 2015 IEEE International Conference on Signal Processing, Informatics, Communication and Energy Systems (SPICES), pp. 1–5 (2015). https://doi.org/10.1109/SPICES.2015.7091560 7. Garg, C., Namdeo, A., Singhal, A., Singh, P., Shaw, R.N., Ghosh, A.: Adaptive fuzzy logic models for the prediction of compressive strength of sustainable concrete. In: Bianchini, M., Piuri, V., Das, S., Shaw, R.N. (eds.) Advanced Computing and Intelligent Technologies. LNNS, vol. 218, pp. 593–605. Springer, Singapore (2022). https://doi.org/10.1007/978-98116-2164-2_47 8. Mitra, P., Chowdhury, S., Chowdhury, S.P., Pal, S.K., Song, Y.H., Taylor, G.A.: Performance of a fuzzy logic based automatic voltage regulator in single and multi-machine environment. In: Proceedings of the 41st International Universities Power Engineering Conference, pp. 1082– 1086 (2006). https://doi.org/10.1109/UPEC.2006.367644 9. El Yakine Kouba, N., Menaa, M., Hasni, M., Boudour, M.: Optimal load frequency control based on artificial bee colony optimization applied to single, two and multi-area interconnected power systems. In: 2015 3rd International Conference on Control, Engineering & Information Technology (CEIT), pp. 1–6 (2015). https://doi.org/10.1109/CEIT.2015.7233027 10. Sonker, B., Kumar, D., Samuel, P.: Differential evolution based TDF-IMC scheme for load frequency control of single-area power systems. In: TENCON 2019 - 2019 IEEE Region 10 Conference (TENCON), pp. 1416–1420 (2019). https://doi.org/10.1109/TENCON.2019.892 9572 11. Nakayama, K., Fujita, G., Yokoyama, R.: Load frequency control for utility interaction of wide-area power system interconnection. In: 2009 Transmission & Distribution Conference & Exposition: Asia and Pacific, pp. 1–4 (2009). https://doi.org/10.1109/TD-ASIA.2009.535 6942 12. Udhayashankar, C., Thottungal, R., Yuvaraj, M.: Transient stability improvement in transmission system using SVC with fuzzy logic control. In: 2014 International Conference on Advances in Electrical Engineering (ICAEE), pp. 1–4 (2014). https://doi.org/10.1109/ICAEE. 2014.6838505

FFT Analysis and Power Quality Improvement in Grid Connected Solar Power Plant with MPPT Algorithm Venu Yarlagadda, G. Lakshminarayana, Giriprasad Ambati, and Garikapati Annapurna Karthika(B) VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, India [email protected]

Abstract. Modern Power systems are equipped with green and renewable energy resources, one among them is Photovoltaic solar generating plants.This article is focused on developing the Solar Power plant model equipped with dc-dc boost converter controlled with MPPT algorithm with Matlab code inbuilt in the Simulink Model and the Power Quality analysis of Solar plant connected to the AC Grid with DSTATCOM. MPPT algorithm is developed in order to obtain optimal power from PV cell at all loading conditions. Fast Fourier Transformation (FFT) analysis has been done for the entire system at each stage of the conversion. This article focused on the implementation of the DSTATCOM to minimize the harmonics in the proposed system.Pulse Width modulation is implemented for controlling three phase DSTATCOM and Simulink models have been developed for the case study with Solar generating plant connected to boot converter to increase the dc voltage to a higher level, and is given to DSTATCOM to obtained AC Power and it is connected to the grid through coupling transformer for feeding the loads. The FFT analysis have been performed and presented in the article, the results have been proven the fact that the power quality of the solar plant output voltage and current waveforms, improved enormously. Keywords: Solar power plant · Power quality improvement · Harmonics mitigation · DSTATCOM · Boot converter · MPPT algorithm · Renewable energy generation

1 Introduction Renewable energy generation draws the attention of power researchers in the modern age in order to minimize greenhouse emissions and pollution. Photo voltaic (PV) generation used to convert the solar radiation into DC Electric Power. The Maximum Power Point Tracking (MPPT), is developed with the usage of Perturb and Observe algorithm. The closed loop control of the boost converter is achieved with the help of the Matlab code embedded in the Simulink Model and is connected to the DSTATCOM. The boost converter uses MOSFET for controlling its dc terminal voltage, and the duty ratio is controlled using closed loop control with MPPT Algorithm. The PWM pulses have been © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 30–43, 2022. https://doi.org/10.1007/978-981-19-1742-4_3

FFT Analysis and Power Quality Improvement

31

generated with the use of sinusoidal pulse width modulation technique. The developed solar power plant is connected to the AC Power Grid through a coupling transformer to get match with the system voltage magnitude. In the subsequent parts of the article, the total FFT analysis at each stage of the convention, before and after Power Quality (PQ) compensating equipment have been done and presented in this article. The overall system power quality being improved with the DSTATCOM Control.

2 Solar Power Plant The single line diagram of the developed solar power plant used to integrate with AC Distribution Power Grid, Solar power plant comprises of the Photo voltaic cells. The two basic inputs to the solar PV cell are irradiance and temperature by which it converts solar energy into DC Electrical Power. The MPPT Matlab coded algorithm is used to have the closed loop control of the boost converter. The output of this converter is used to feed the DSTATCOM to get the AC Power conversion and used to supply ac loads as well as to the AC Power Grid through a coupling transformer as shown in the Fig. 1 below.

Fig. 1. Single line diagram of Grid tied solar power plant with MPPT based Boost Converter and DSTATCOM

2.1 Solar Photovoltaic (PV) Cell The equivalent circuit of PV cell is consisting of series and shunt resistors and parallel diode is shown in Fig. 2 and its modelling equations for the current as well as constants A and B as mentioned in the below Eqs. (1), (2) and (3) respectively. The Voltage and current, voltage and power characteristics of the PV cell has been illustrated in the Fig. 3 below. The Solar plant equations   (1) I = IL − I 0 e A − 1 − B V + IRS nVt V + IRS B= Rsh A=

(2) (3)

32

V. Yarlagadda et al.

Fig. 2. PV cell equivalent

Fig. 3. VI and Power Characteristics of Solar power plant

2.2 MPPT Algorithm Maximum Power Point Tracking (MPPT) Algorithm is most commonly used P & O or Hill climbing technique as illustrated in Fig. 4. The MPPT algorithm for extracting the maximum power output from the solar power plant is clearly illustrated in the Fig. 5.

Fig. 4. MPPT Extraction from VI Characteristics

FFT Analysis and Power Quality Improvement

33

Fig. 5. MPPT Algorithm of Solar power plant

2.3 Boost Converter The DC-DC boost converter is utilized to improve the DC Voltage magnitude output of the solar power plant to match with the system requirements and the output voltage equation of the converter as specified in the Eq. (1) below. The schematic diagram of the boost converter has clearly illustrated in the Fig. 6 shown below. Vo =

Vi 1−D

(4)

34

V. Yarlagadda et al.

Fig. 6. Boost converter schematic circuit

3 DSTATCOM Flexible AC Transmission System (FACTS) devices have been emerged for enhancement in power system stability. The conditions like voltage stability, transient stability, reactive power flow control, power quality etc. are enhanced with power electronic based FACTS devices. DSTATCOM is a shunt connected device which employs power electronic devices such as IGBT, GTO, MOSFET etc. as depicted in Fig. 5, which are basically fast acting switching devices in order to improve stability of power system and control of reactive power flow i.e. by absorbing reactive power from the system or by generation of reactive power meeting the demand to maintain the voltage at specific limits [5–10]. The working principle of DSTATCOM, consider Inverter output voltage as V1 and system output voltage as V2. The exchange of reactive power in between the DSTATCOM and system is based on the voltages V1 and V2 i.e. if the demand in reactive power in the system increases, then the output voltage of the DSTATCOM gets increases and vice versa with no flow of active power in between the system and DSTATCOM by keeping the angle zero [4]. STATCOM comprises mainly Voltage Source Converter (VSC), is generally GTO type and IGBT based converters which converts DC voltage into AC voltage. In GTO type converter, AC output voltage can be varied by DC capacitor input voltage and similarly in IGBT based converter, Pulse Width Modulation (PWM) technique is used to generate a sinusoidal wave form with frequency of kHz from DC voltage. DC capacitor supply constant voltage to VSC and a transformer is coupled in between power system and DSTATCOM and also reduces the harmonics in square wave generated by VSC [8–10] (Fig. 7).

FFT Analysis and Power Quality Improvement

35

Fig. 7. D-STATCOM Schematic Diagram

3.1 Mathematical Modelling of STATCOM The equations for active power, reactive power of STATCOM are as follows; Consider Vt = system terminal voltage Vst Vst = STATCOM output voltage XL = Inductive reactance VC = DC capacitor input voltage Vt VC sin α XL

(5)

Vt Vt Vt VC − cos α XL XL

(6)

P= Q=

The equation of STATCOM DC side can be given as; The mathematical equations of STATCOM can be expressed as; L = series inductance R = series resistance iac , ibc , icc are output currents of STATCOM Vac , Vbc , Vcc are output voltages of STATCOM Vta , Vtb , Vtc are terminal voltages L

diac = R + Vac − Vat dt

(7)

L

dibc = R + Vbc − Vbt dt

(8)

L

dicc = R + Vcc − Vct dt

(9)

36

V. Yarlagadda et al.

4 Case Study and Results The Solar power plant connected to the grid system for presenting case study and results have been developed with the usage of photovoltaic solar plant connected to boost converter and DSTATCOM coupled to the transformer is integrated to the grid is described in subsequent topics. Figure 8 shows the Solar power plant with MPPT Algorithm based Boost Converter, Fig. 9 depicts the Power Circuit of DSTATCOM including LCL Passive Filter, Fig. 10 illustrates the PWM signals generating Circuit of DSTATCOM and Fig. 11 shows the Grid Connected Solar System with DSTATCOM. Figure 12 illustrates the complete setup of Grid Connected Solar System with MPPT based Boost Converter and DSTATCOM complete system Simulink Model.

Fig. 8. Solar power plant with MPPT Algorithm based Boost Converter

Fig. 9. Power Circuit of DSTATCOM including LCL Passive Filter

FFT Analysis and Power Quality Improvement

37

Fig. 10. PWM signals generating Circuit of DSTATCOM

Fig. 11. Grid connected solar system with DSTATCOM

The simulation results of the developed system for the case study have been presented herewith, as mentioned in the following FFT analysis and waveforms. Figure 13 shows the Solar power plant Output Voltage FFT Analysis with THD is of 77.04% in the dc output voltage, Fig. 14 depicts the Solar power plant Boost Converter Output Voltage before filter FFT Analysis with THD is of 229.67%, which is very high. These harmonics have been decreased with the compensating equipment.

38

V. Yarlagadda et al.

Fig. 12. Grid connected solar system with MPPT based boost converter and DSTATCOM complete system Simulink model

Fig. 13. Solar power plant Output Voltage FFT Analysis

The boost converter output with compensating equipment is decreased to 99.71% from 229.67% and the Fig. 15 illustrating the same for Solar power plant Boost Converter Output Voltage FFT Analysis. Figure 16 depicts the Solar power plant Inverter Output Voltage before Filter FFT Analysis with THD is of 69.85% and Fig. 17 illustrates the solar power plant Inverter Output Current FFT Analysis with THD is of 17.33% before compensating equipment.

FFT Analysis and Power Quality Improvement

39

Fig. 14. Solar power plant Boost Converter Output Voltage before filter FFT Analysis

Fig. 15. Solar power plant Boost Converter Output Voltage FFT Analysis

The last part of the FFT analysis being presented in this section, the Power Quality is poor without any compensating equipment at each stage of the converting equipment which was proved in the simulation results presented earlier part of the article. DSTATCOM and compensating equipment have been used to improve the power quality of the system with very low THD values have been obtained and are presented herewith. Figure 18 illustrates the Solar power plant Inverter Output Voltage Waveform which is pure sinusoidal being proved in the following FFT analysis. The FFT analysis being

40

V. Yarlagadda et al.

Fig. 16. Solar power plant Inverter Output Voltage before Filter FFT Analysis

Fig. 17. Solar power plant Inverter Output Current FFT Analysis

FFT Analysis and Power Quality Improvement

41

done for the said system with different loadings and Fig. 19 depicts the solar power plant Inverter Output Voltage FFT Analysis with THD is of negligible value with 0.88% only. The current wave form FFT analysis of grid connected Solar system with DSTATCOM and compensating equipment is being reduced to a very low value of THD is about 3.8% and Fig. 20 proves the same with Solar power plant Inverter Output Current FFT Analysis.

Fig. 18. Solar power plant Inverter Output Voltage Waveform

Fig. 19. Solar power plant Inverter Output Voltage FFT Analysis

42

V. Yarlagadda et al.

Fig. 20. Solar power plant Inverter Output Current FFT Analysis

5 Conclusions This article is focused on developing the Solar Power plant model equipped with dc-dc boost converter controlled with MPPT algorithm with Matlab code inbuilt in the Simulink Model and the Power Quality analysis of Solar plant connected to the AC Grid with DSTATCOM.Further this article fascinated on the implementation of the DSTATCOM to minimize the harmonics in the proposed system. Pulse Width modulation is implemented for controlling three phase DSTATCOM and Simulink models have been developed for the case study with Solar generating plant connected to boot converter to increase the dc voltage to a higher level, and is given to DSTATCOM to obtained AC Power and it is connected to the grid through coupling transformer for feeding the loads. The last part of the FFT analysis being presented in this section, the Power Quality is poor without any compensating equipment at each stage of the converting equipment which was proved in the simulation results. DSTATCOM and compensating equipment have been used to improve the power quality of the system with very low THD values have been obtained and are presented. The FFT analysis being done for the said system with different loadings and presented which proves that the solar power plant Inverter Output Voltage has negligible harmonics with THD is of insignificant value and similarly the FFT analysis current waveform shows the THD is of insignificant value, Hence the DSTATCOM along with other power quality mitigating equipment is capable of minimizing harmonics and subsequently the power quality of the system is significantly improved.

FFT Analysis and Power Quality Improvement

43

References 1. Swetha, K., Sivachidambaranathan, V.: Reactive power control and harmonic mitigation by using DSTATCOM with different controllers. In: 2019 3rd International conference on Electronics, Communication and Aerospace Technology (ICECA), pp. 490–496 (2019). https:// doi.org/10.1109/ICECA.2019.8822160 2. Gidd, A.R., Gore, A.D., Jondhale, S.B., Kadekar, O.V., Thakre, M.P.: Modelling, analysis and performance of a DSTATCOM for voltage sag mitigation in distribution network. In: 2019 3rd International Conference on Trends in Electronics and Informatics (ICOEI), pp. 366–371 (2019). https://doi.org/10.1109/ICOEI.2019.8862554 3. Srikakolapu, J., Arya, S.R., Maurya, R.: Algorithm for DSTATCOM using cascaded delayed signal cancellation effect in three wire system. In: 2019 IEEE 1st International Conference on Energy, Systems and Information Processing (ICESIP), pp. 1–7 (2019). https://doi.org/10. 1109/ICESIP46348.2019.8938307 4. Chawda, G.S., Shaik, A.G.: Power quality mitigation in weak AC grid with low X/R ratios using distribution static compensator controlled by LMF algorithm. In: 2020 IEEE Region 10 Symposium (TENSYMP), pp. 44–47 (2020). https://doi.org/10.1109/TENSYMP50017. 2020.9230727 5. Pandey, S.K., Singh, B.: Hybrid DSC with compensation capability based control for grid integrated SPV system. In: 2020 IEEE 9th Power India International Conference (PIICON), pp. 1–6 (2020). https://doi.org/10.1109/PIICON49524.2020.9113069 6. Chaithanya, S., Jayaprakash, P.: Power quality improvement strategy for DSTATCOM using adaptive sign regressor LMF control. In: 2020 IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES), pp. 1–6 (2020). https://doi.org/10.1109/ PEDES49360.2020.9379684 7. Kumar, N., Singh, B., Panigrahi, B.K., Chakraborty, C., Suryawanshi, H.M., Verma, V.: Integration of solar PV with low-voltage weak grid system: using normalized Laplacian kernel adaptive Kalman filter and learning based InC algorithm. IEEE Trans. Power Electron. 34(11), 10746–10758 (2019). https://doi.org/10.1109/TPEL.2019.2898319 8. Singh, B., Shukl, P.: Control of grid fed PV generation using infinite impulse response peak filter in distribution network. IEEE Trans. Ind. Appl. 56(3), 3079–3089 (2020). https://doi. org/10.1109/TIA.2020.2968287 9. Shaw, R.N., Walde, P., Ghosh, A.: Review and analysis of photovoltaic arrays with different configuration system in partial shadowing condition. Int. J. Adv. Sci. Technol. 29(9s), 2945– 2956 (2020) 10. Shah, P., Singh, B.: Robust EnKF with improved RCGA-based control for solar energy conversion systems. IEEE Trans. Industr. Electron. 66(10), 7728–7740 (2019). https://doi.org/ 10.1109/TIE.2018.2885727 11. Sinha, S., Arora, A.: Comparison of IRPT and ANN based control algorithm for shunt compensation in grid connected systems. In: 2021 International Conference on Intelligent Technologies (CONIT), pp. 1–5 (2021). https://doi.org/10.1109/CONIT51480.2021.9498281 12. Arora, A., Singh, A.: Design and implementation of biquad filter for shunt compensation under normal and distorted grid conditions. In: 2020 IEEE 9th Power India International Conference (PIICON), pp. 1–6 (2020). https://doi.org/10.1109/PIICON49524.2020.9112949 13. Shah, P., Singh, B.: LVRT capabilities of solar energy conversion system enabling power quality improvement. In: 2019 IEEE International Electric Machines & Drives Conference (IEMDC), pp. 2083–2088 (2019). https://doi.org/10.1109/IEMDC.2019.8785096

Transient Response and Stability Analysis of Weak Power System Using Hybrid Compensator, STATCOM and SVC: A Comprehensive Comparison Venu Yarlagadda(B) , T. Haripriya, N. Amarnath Reddy, and Giriprasad Ambati VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, India [email protected]

Abstract. Modern Power Systems are equipped with many distinct loads with huge transmission network. To ensure dynamic and transient generator and voltage stability in order to maintain the power system security is a major task of the power system Engineer. FACTS Controllers are most effective devices to ensure system security by enhancing the stability margins with reactive power support all over power system network. The major shunt compensation devices of FACTS are SVC and STATCOM, in this article the modelling of Hybrid compensator comprehended with SVC and GCSC as well and comparison among all of these devices have been made for weak power system. This article dispenses the modelling and simulation of these shunts devices such as Static Synchronous Compensator (STATCOM), Static Var Compensator (SVC) and Hybrid Compensator. The transfer function models of these devices have been derived from the first principles and obtained the transfer function models of weak power system. The dynamic response is obtained with the exact models of all these controllers for weak system, subsequently the root locus plots as well as bode plots have been obtained with MATLAB Programs and evaluated the performance of these devices and comparison is made. The Stability margins of the power system with all three devices have been obtained from the bode plots. The transient response of these devices have been assessed with time responses. The power system transient response as well as stability analysis using root locus and bode plots have been obtained and critically evaluated the merits and demerits of all these controllers. The power system performance has been improved with STATCOM as well as Hybrid Controllers. Keywords: SVC · STATCOM · Hybrid compensator · GCSC and SVC · Stability margins · Modelling of SVG · Bode plots · Transient response of STATCOM · Comaritive analysis of FACTS devices

1 Introduction Modern power systems are closely operating at its critical points due to exponenticial groth of power demand, due to this reason the power system performance playing a © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 44–59, 2022. https://doi.org/10.1007/978-981-19-1742-4_4

Transient Response and Stability Analysis of Weak Power System

45

crucial role in power system security. The power system transient response as well as the stability and performance can be greatly effected by the use of FACTS devices viz. SVC, STATCOM and Hybrid Compensators. This article engrossed on the modelling and simulation ofthese devices, it’s comprehensive coparitive analysis, focussing on it’s transient response, root locus plots and bode plots for weak power system. The transfer function models have been derived for all these controlles along with it’s time constants associated with all parts of the Compensators. These transfer function models have been used for the simulation studies, for the comparision of these three devices. The transient response is evaluated for weak system with three devices using transfer function models with the performance indices viz. peak overshoot and settling time. The subsequent part of the article deals with the root locus and bode plots and stability margins of these compensators for two kinds of systems. The gain and phase margins have been derived from the Matlab progrms and analysis of comparition have been made for all cases, [4–10].

2 STATCOM The single line diagram of power system with STATCOM, generating station is feeding load through a transmission system with two buses, bus1 and bus2 at sending and receiving end respectively. The STATCOM schematic is connected to the receiving end bus as shown in Fig. 5, it may comprise of IGBT’s and DC link capacitance, which is used to enhance the performance of the power system as illustrated in Fig. 1 below. The system should remain in synchronism even after being subjected to the disturbance, which involves output power oscillates reflected in rotor oscillations [3–11].

Fig. 1. Schematic diagram of STATCOM

46

V. Yarlagadda et al.

2.1 Mathematical Modelling of DSTATCOM The equations for active power, reactive power of STATCOM are as follows; Consider Vt = system terminal voltage Vst = STATCOM output voltage XL = Inductive reactance VC = DC capacitor voltage

Vt VC sinα XL

(1)

Vt Vt Vt VC − cosα XL XL

(2)

P= Q=

The equation of DSTATCOM on DC side can be given as; The mathematical equations of DSTATCOM can be expressed as; L = series inductance R = series resistance iac , ibc , icc are output currents of DSTATCOM Vac , Vbc , Vcc are output voltages of DSTATCOM Vta , Vtb , Vtc are terminal voltages

L

diac = Riac + Vac − Vat dt

(3)

L

dibc = Ribc + Vbc − Vbt dt

(4)

L

dicc = Ricc + Vcc − Vct dt

(5)

3 Static Var Compensator (SVC) 3.1 Description of Working and Power Circuit of SVC The single line diagram of power system with SVC, generating station is feeding load through a transmission system with two buses, bus1 and bus2 at sending and receiving end respectively. The SVC schematic is connected to the reciving end bus, it may comprise of SCR’s, which is used to enhance the performance of the power system as depected in Fig. 3 below [11] and [12].

Transient Response and Stability Analysis of Weak Power System

47

Fig. 2. Schematic diagram of SVC

3.2 Mathematical Modelling of SVC The susceptance of the device can be controlled with the control of the firing angle of the TCR and Fig. 2. Shows the Variable susceptance of SVC [21–28]. BSVC = BTSC − BTCR

(6)

BTCR = BL ((π − 2α − sin α)/π) + BC

(7)

QSVC = ((XC [2π − α + sin 2α]−πXL ))/((π XC XL ))

(8)

In steady state an SVC can be treated as a reactive power injection source, which can be presented as the following mathematical expression: QSVC = VT (VT − Vref )XSL

(9)

where XSL is the slope of voltage control characteristic, Vt is the terminal voltage of SVC and Vref is the reference voltage, the above Equation can be rewritten as: QSVC = BSVC × Vref2

(10)

The value of BSVC can be varied between minimum and maximum susceptance and the reactive power generated by SVC is given by QSVCmin ≤ QSVC ≤ QSVCmax

(11)

4 Hybrid Compensator 4.1 GTO Controlled Series Capacitor (GCSC) The GCSC schematic is described with the antiparallel combination of GTO Thyristors used to control the series injected voltage with the feeder as depicted in the below figure.

48

V. Yarlagadda et al.

The reactance and voltage variation of GCSC with the variation of the conduction angle as depicted by the following Eqs. (1) and (2) respectively. The harmonics injected by the device is illustrated in the Eq. (3) and the voltage wave form of complete control of the device is illustrated in Fig. 4 comprising of parts a, b and c as (a) GCSC Schematic circuit, (b) One complete cycle, (c) total current of GCSC. The total current shown in the waveform comprises of harmonics since the conduction angle is rapidly adjusted for controlling series voltage of the system [1–3].

Fig. 3. Schematic diagram of GCSC

  2γ sin(2γ ) 1 1− − wC π π   2γ sin(2γ ) I VCF (γ) = 1− − wC π π    sin(γ )cos(nγ ) − nsin(nγ )cos(γ ) sin(2γ ) I 4 VCn (γ) = − wC π n(n2 − 1) π Xc(γ) =

(12) (13) (14)

4.2 Hybrid Compensator Hybrid Compensator is consists of one variable impedance series compensator i.e. GTO Controlled Series Capacitor (GCSC) and Static Var Compensator (SVC) as illustrated in Fig. 5. His hybrid compensator is used in the weak power system to improve its transient response and system stability which is compatible to that of the SATCOM [1–3].

Transient Response and Stability Analysis of Weak Power System

49

Fig. 4. (a) GCSC Schematic circuit, (b) One complete cycle, (c) total current of GCSC

Fig. 5. Schematic diagram of Hybrid Compensator

5 Transfer Function Model of Static VAR Generator (SVC/STATCOM) The transfer function model of Static Var Generator for both SVC and STATCOM have been shown in Fig. 3. It comprising of the regulator transfer function G1 with PI controller time constant T1, droop k value in typical range of 1 to 5%, Controller transfer function

50

V. Yarlagadda et al.

with transport dealy Td, which is different for both the controllers and feed back transfer function of measuring circuit with time constant of T2 respectively as depicted in Fig. 3. The transport delay is very low for the STATCOM and it very significant in infunecing the power system performance. 5.1 Transfer Function Model of Voltage Regulator The transfer function model of the Voltage regulator is obtained with slope of the VI Characteristics of the STATCOM and SVC, which is in the range of 1 to 5% as illustrated below in the equation, where the T1 is the time constant of the PI Controller, typically it is in range of 10 to 50 ms as depicted in Fig. 6 below.  1 k G1 = 1 + sT1

Fig. 6. Voltage regulator model

5.2 Transfer Function Model of Static Var Generator The transfer function model of the Static Var Generator viz. STATCOM and SVC is obtained with the transport delay time Td of the STATCOM and SVC as illustrated below in the equation, where the Td is transport delay of the STATCOM and SVC Controller, typically it is in range of 0.5 ms for STATCOM and 5.55 ms for the SVC. This transport delay makes the distinction between both of the controllers as illustrated in the following equation and block diagram as shown in Fig. 7 below. G2 =

1 1 + sTd

Fig. 7. SVG model

(15)

Transient Response and Stability Analysis of Weak Power System

51

5.3 Transfer Function Model of Feedback Circuit The transfer function model of the feedback measurement circuit is obtained with the delay time in the measurement as T2 as illustrated below in the equation, where the T2 is the delay in the measuring circuit, typically it is in range of 8 to 16ms as illustrated in the following equation and block diagram as depicted in Fig. 8 below. H=

1 1 + sT2

(16)

Fig. 8. Measuring feedback model

5.4 Complete Transfer Function Model of STATCOM and SVC The complete transfer function model of an Static Var Generator either STATCOM or SVC, since same model is equally valid for both the controllers with different transport delay time is depicted in the following Fig. 8. This transfer function model is comprising of two input signals, one is the reference voltage i.e. Vref and the second one is SVG output voltage as disignated Vo and one output as terminal voltage of the power system. The slope of the VI charectoristics of SVG is represented as k and is explained in the previous part of the article, furter the power system reactance, these two parameters will discriminate between weak and strong power systems. The typical values of reactance for strong system is about 4 to 5, and weak system in per unit values and for weak system it is about 9 to 10 p.u as taken in simulation study [5–10] (Fig. 9). P=

V1 V2 sin δ XT

Fig. 9. Transfer Function model of SVG (SVC and STATCOM)

(17)

52

V. Yarlagadda et al.

6 Case Study and Results The Simulink models have been designed and developed for weak power systems along with the SVC, STATCOM and Hybrid Compensators as illustrated in Fig. 10 below, to get the comarion of both Transient responces on same graph, a mux with three inputs, one is from STATCOM, the second is from SVC and the third is from Hybrid Compensator are connected to one scope, in which the comparitive plot is achieved.

Fig. 10. Simulink Model weak system with SVC, STATCOM and Hybrid device with two distinct firings of GCSC and SVC

6.1 Comparative Analysis of SVC, STATCOM and Hybrid Compensator Transient Responces for Weak Power System The Transient responses of the SVC, STATCOM and Hybrid Compensator for the weak power sytem with two simultanious inputs of reference voltage and output voltage of the Compensators as depicted in following Fig. 11. The weak power system with SVC is completely unstable, The peak overshoot of STATCOM is 94.56% is more when compared to the Hybrid Compensator, which is 70.4%. The settling time is concerned, there is no settling time for SVC hence the system is unstable. The STATCOM settling time is 0.06S and is very low comared to Hybrid Compensators settling time 0.107S and both systems are stable as depicted in the Fig. 10 below and the steady state error is more for Hybrid Compensator. The steady state error of Hybrid Compensator is decreased with different set of conduction, firing angles of GCSC and SVC respectively and the settling time is increased to a higher value for it as illustrated in the Fig. 12 below.

Transient Response and Stability Analysis of Weak Power System

53

Fig. 11. Dynamic Responses of STATCOM and SVC for strong system

Fig. 12. Transient Response for two different firing angles of Hybrid Compensator

6.2 Comparative Analysis of SVC, STATCOM and Hybrid Compensator Root Locus Plots for Weak Power Systems The root locus plots of all three compensators for the weak power sytem with reference voltage as depicted in following Fig. 13. The root locus plots of SVC indicates that for weak system and it indicating small relative stability margins. Figure 14 below, which indicates that that the system is stable with STATCOM and relative margin of stability is high. Figure 15 shows the Root locus plot of Hybrid Compensator with conduction angle 1 for weak power system and Fig. 16. depicts the Root locus plot of Hybrid Compensator

54

V. Yarlagadda et al.

with conduction angle 2 for weak power system but for both relative stability margin is higher than SVC and lower than STATCOM.

Fig. 13. Root locus plot of SVC for weak power system

Fig. 14. Root locus plot of STATCOM for weak power system

Transient Response and Stability Analysis of Weak Power System

55

Fig. 15. Root locus plot of Hybrid Compensator with conduction angle 1 for weak power system

Fig. 16. Root locus plot of Hybrid Compensator with conduction angle 2 for weak power system

56

V. Yarlagadda et al.

6.3 Comparative Analysis of SVC, STATCOM and Hybrid Compensator Using Bode Plots for Weak Power Systems The bode plots of all three compensators for the weak power sytem with reference voltage as depicted in following Fig. 17. The bode plots of SVC shows that the system is unstable with negative phase margin. The STATCOM response shows that the system is completely stable with phase margin of 13.9976° and with infinity gain margin. The bode plots of the Hybrid Compensator also showing negative margins and hence among all compensators SATATCOM performance is better, whereas the system can be made stable with Hybrid Compensator and the system is completely unstable with SVC alone (Figs. 18, 19 and 20).

Fig. 17. Bode plot result of SVC for Strong system

Fig. 18. Bode plot result of STATCOM for weak system

Transient Response and Stability Analysis of Weak Power System

57

Fig. 19. Bode plot result of Hybrid device with conduction angle 2 for weak system

Fig. 20. Bode plot result of Hybrid device with conduction angle 1 for weak system

7 Conclusions This article dispense the modelling and simulation of three devices viz. one is Static Synchronous Compensator (STATCOM) and the second is Static Var Compensator (SVC) and the last is Hybrid Compensator comprised of GCSC and SVC. The transfer function models of these devices have been derived from the first principles. The transient response is obtained with the exat model weak system. The transient performance of all

58

V. Yarlagadda et al.

devices have been simulated and results have proven that the STATCOM is relatively more stable compared to other two devices. The system is completely unstable with SVC and it can be made stable with hybrid compensator comprising of GCSC and SVC, which proved by simulation resopnces. Subsequently the root locus plots as well as bode plots have been obtained with MATLAB Programs and evaluated the performance of these devices and comparision is made. The root locus plots of all three compensators for the weak power sytem with reference voltage have been presented and proves that the STATCOM is stronger device compared to other two and Hybride compensator has the better margin compared to SVC. The bode plots of all three compensators for the weak power sytem with reference voltage have been presented and proves that the system is Stable only for STATCOM and is unstable for other two devices. The Hybrid device transient response as well as root locus plots shows that the system is stable and can be accepted, where as only root locus plots of SVC are accepted and other two plots of SVC are not accepted.

References 1. Oshnoei, A., Kheradmandi, M., Khezri, R., Mahmoudi, A.: Robust model predictive control of gate-controlled series capacitor for LFC of power systems. IEEE Trans. Industr. Inf. 17(7), 4766–4776 (2021). https://doi.org/10.1109/TII.2020.3016992 2. Morsali, J., Zare, K., Hagh, M.T.: MGSO optimised TID-based GCSC damping controller in coordination with AGC for diverse-GENCOs multi-DISCOs power system with considering GDB and GRC non-linearity effects. IET Gener. Transmiss. Distrib. 11(1), 193–208 (2017) 3. Maza-Ortega, J.M., Acha, E., García, S., et al.: Overview of power electronics technology and applications in power generation transmission and distribution. J. Mod. Power Syst. Clean Energy 5, 499–514 (2017). https://doi.org/10.1007/s40565-017-0308 4. Shaw, R.N., et al.: Effects of solar irradiance on load sharing of integrated photovoltaic system with IEEE standard bus network. Int. J. Eng. Adv. Technol. 9(1), 424–429 (2019) 5. Liu, J., Yao, W., Wen, J., et al.: Impact of power grid strength and PLL parameters on stability of grid-connected DFIG wind farm. IEEE Trans. Sustain. Energy 11(1), 545–557 (2020) 6. Xu, Y.: A review of cyber security risks of power systems: from static to dynamic false data attacks. Prot. Control Mod. Power Syst. 5(1), 1–12 (2020). https://doi.org/10.1186/s41601020-00164-w 7. Tu, G., Li, Y., Xiang, J.: Analysis and control of energy storage systems for power system stability enhancement. In: 2019 Chinese Control Conference (CCC), pp. 560–565 (2019). https://doi.org/10.23919/ChiCC.2019.8865814 8. Paital, S.R., Ray, P.K., Mohanty, A.: A review on stability enhancement in SMIB system using artificial intelligence based techniques. In: 2018 IEEMA Engineer Infinite Conference (eTechNxT), pp. 1–6 (2018). https://doi.org/10.1109/ETECHNXT.2018.8385324 9. Baadji, B., Bentarzi, H., Bouaoud, A.: SMIB power system model with PSS for transient stability studies. In: 2017 5th International Conference on Electrical Engineering - Boumerdes (ICEE-B), pp. 1–5 (2017). https://doi.org/10.1109/ICEE-B.2017.8191996 10. Kalyani, S., Prakash, M., Ezhilarasi, G.A.: Transient stability studies in SMIB system with detailed machine models. In: 2011 International Conference on Recent Advancements in Electrical, Electronics and Control Engineering, pp. 459–464 (2011). https://doi.org/10.1109/ iconraeece.2011.6129781

Transient Response and Stability Analysis of Weak Power System

59

11. Cherkaoui, N., Haidi, T., Belfqih, A., El Mariami, F., Boukherouaa, J.: A comparison study of reactive power control strategies in wind farms with SVC and STATCOM. Int. J. Electr. Comput. Eng. 8, 4836 (2018) 12. Pilotto, L.A.S., Bianco, A., Long, F.W., Edris, A.A.: Impact of TCSC control methodologies on subsynchronous oscillations. IEEE Trans. Power Delivery 18(1), 243–252 (2003)

Comparative Analysis of GCSC and Hybrid Compensators Influence on Power Transfer Enhancement of UHV Transmission Systems Venu Yarlagadda(B) , Rashmi Kapoor, G. Lakshminarayana, and N. Amarnadh Reddy VNR Vignana Jyothi Institute of Engineering and Technology, Hyderabad, India

Abstract. The exponential growth of everlasting power demand, the management of power grid is easy with the increased usage of EHV and UHV Transmission systems. The performance of UHV Power Transmission can be improved with the usage of FACTS Devices. The GTO Thyristor Turn off Series Capacitor (GCSC) is a powerful device among all FACTS Devices in order to control power flow and stability improvement. This article realized the fact of effectiveness of both the FACTS devices such as GCSC and Hybrid Compensator comprises of GCSC and Static Var Compensator (SVC) in improving the power Transmission Capability and reactive power margins of UHV Transmission systems. The UHV Transmission systems have been simulated with operating voltages of 800 KV and 1500 KV. All of these systems have been simulated without, with GCSC and Hybrid Compensators and are presented in this article. The simulation results prove the power transfer capability and reactive power margins have been improved substantially with both of the controllers and especially the hybrid compensator showing the superior performance in 1500 KV UHV Transmission system. Keywords: GCSC · Hybrid compensator · Power flow control · Reactive power margin · UHV transmission · FACTS devices · GCSC and SVC · SVC

1 Introduction The Modern Power System is growing with the use of Extra High Voltage A.C Transmission (EHV Transmission) and Ultra High Voltage A.C Transmission Systems to meet the everlasting power demands. The UHV Transmission system performance can be magnifying with the use of FACTS and Controllers. GTO Thyristor controlled Series Capacitor (GCSC) is a powerful device, which can be used as GCSC in Transmission Sector and Hybrid Compensator is designed with the use of both GCSC and Static Var Compensator (SVC). These devices being used to enhance the power transfer Capability and reactive power margins in UHV Transmission systems with the voltage range from 800 KV to 1500 KV respectively [3–6]. In this article, UHV Transmission systems ranging from 800 KV to 1500 KV have been simulated in Simulink and results have been presented in two sections. One is 800 KV UHV Power Transmission system and the other is 1500 KV UHV system. The case study illustrates the effectiveness of each of these controllers and the comparison between them [1–4]. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 60–76, 2022. https://doi.org/10.1007/978-981-19-1742-4_5

Comparative Analysis of GCSC and Hybrid Compensators Influence

61

2 Power Transmission The Transmission lines have been categorized into three types such as short, medium and long lines as per line length, are classified into Transmission (220 KV), EHV A.C Transmission (400 KV), UHV A.C Transmission (800 KV, 1200 KV and 1500 KV Systems). The long lines is of more than 250 km line length, where the line parameters are distributed over entire transmission line and the equivalent circuit of long transmission line represented by equivalent-T and equivalent- networks [6–10] (Fig. 1).

Fig. 1. Equivalent-pi network

The equivalent impedance and admittances of the long line rigorous solution leads to the following expressions expressed in Eq. 1 and 2 respectively. Z  = Zsinh(γ l)/γ l

Y = 2

  Ytanh γ2l γl

Fig. 2. Equivalent-T network

(1)

(2)

62

V. Yarlagadda et al.

Y  = Ysinh(γ l)/γ l

Z = 2

  Ztanh γ2l γl

(3)

(4)

3 Power System Stability Power system stability is defined as a system ability to regain its initial equilibrium state after being subjected to a disturbance and its classification are given in Fig. 3 below.

Fig. 3. Power system stability stratification

3.1 Relation Between Power and Angle The power transfer through a transmission line or distribution feeder can be expressed in terms of the bus voltages at sending and receiving ends, designated as V1 and V2 respectively as depicted in the following Eq. (1). It is maximum when the load angle is equal to the 90°; corresponding sine of the angle is unity, leads to maximum power as expressed in the following Eq. (2) below. P=

V1 V2 sin δ XT

Pmax =

V1 V2 XT

(5) (6)

Comparative Analysis of GCSC and Hybrid Compensators Influence

63

Fig. 4. Power angle curve

4 GTO Controlled Series Capacitor (GCSC) The GCSC schematic is described with the antiparallel combination of GTO Thyristors used to control the series injected voltage with the feeder as depicted in the below Fig. 5. The reactance and voltage variation of GCSC with the variation of the conduction angle as depicted by the following Eqs. (1) and (2) respectively. The harmonics injected by the device is illustrated in the Eq. (3) and the voltage wave form of complete control of the device is illustrated in Fig. 4 comprising of parts a, b and c as (a) GCSC Schematic circuit, (b) One complete cycle, (c) total current of GCSC. The total current shown in the waveform comprises of harmonics since the conduction angle is rapidly adjusted for controlling series voltage of the system [4–8] (Fig. 6).

Fig. 5. Schematic diagram of GCSC

  2γ sin(2γ ) 1 1− − Xc (γ) = wC π π   2γ sin(2γ ) I VCF (γ) = 1− − wC π π

(7) (8)

64

V. Yarlagadda et al.

I 4 VCn (γ) = wC π



sin(γ )cos(nγ ) − nsin(nγ )cos(γ )   n n2 − 1



sin(2γ ) − π

(9)

Fig. 6. (a) GCSC schematic circuit, (b) One complete cycle, (c) total current of GCSC

5 Static Var Compensator (SVC) 5.1 Description of Working and Power Circuit of SVC The single line diagram of power system with SVC, Generating station is feeding load through a transmission system with two buses, bus1 and bus2 at sending and receiving end respectively. The SVC schematic is connected to the reciving end bus, it may comprising of SCR’s, which is used to enhance the performance of the power system as depected in Fig. 3 below [7–10]. 5.2 Mathematical Modelling of SVC The susceptance of the device can be controlled with the control of the firing angle of the TCR and Fig. 2 shows the Variable susceptance of SVC [21–28]. BSVC = BTSC − BTCR

(10)

BTCR = BL ((π − 2α − sinα)/π) + BC

(11)

QSVC = ( (XC [2π − α + sin2α] − πX L))/((π XCX L))

(12)

Comparative Analysis of GCSC and Hybrid Compensators Influence

65

Fig. 7. Schematic diagram of SVC

In steady state an SVC can be treated as a reactive power injection source, which can be presented as the following mathematical expression: QSVC = VT (VT − Vref)XSL

(13)

Where XSL is the slope of voltage control characteristic, Vt is the terminal voltage of SVC and Vref is the reference voltage, the above Equation can be rewritten as: QSVC = BSVC × Vref2

(14)

The value of BSVC can be varied between minimum and maximum susceptance and the reactive power generated by SVC is given by QSVCmin ≤ QSVC ≤ QSVCmax

(15)

6 Hybrid Compensator Power Systems can be sub divided into number of areas through tie lines, without out loss of generality, consider a simple two area power system with a single tie line as depicted by Fig. 7 as shown below. 6.1 GTO Controlled Series Capacitor (GCSC) The GCSC schematic is described with the antiparallel combination of GTO Thyristors used to control the series injected voltage with the feeder as explained in the earlier section, it is connected in series with the transmission line in the hybrid compensator configuration.

66

V. Yarlagadda et al.

6.2 Static Var Compensator (SVC) The single line diagram of power system with SVC, as explained in the erlier section, is used to connect in parallel with the system and it is a shunt part of the hybrid compensator. It rapidly adjustes the shunt susuptance and it further enhances the hybrid controller performance ratherthan the single compensator. 6.3 Hybrid Compensator (GCSC-SVC) Hybrid Compensator is consists of one variable impedance series compensator i.e. GTO Controlled Series Capacitor (GCSC) and Static Var Compensator (SVC) as illustrated in the below Fig….This hybrid compensator is used in the weak power system to improve its transient response and system stability which is compatible to that of the SATCOM [1–3] (Fig. 8).

Fig. 8. Schematic diagram of hybrid compensator

7 Case Study and Results UHV Transmission systems of 800 KV to 1500 KV have been developed and simulated in Simulink and the results have been presented in this article. Figure 9 depicts the Simulink model of UHV Transmission System without any controller, Fig. 10 shows the UHV Transmission System with GCSC. Figure 11 encapsulate the UHV Transmission System with Hybrid Compensator and Fig. 12 instantiated per phase equivalent Simulink diagram of UHV Transmission System with Hybrid Compensator. The following sections of this article encapsulate the simulation results of 88 KV and 15000 KV UHV Transmission systems without any compensator, with GCSC and with Hybrid Compensators [1–5].

Comparative Analysis of GCSC and Hybrid Compensators Influence

Fig. 9. UHV Transmission system without GCSC

Fig. 10. UHV Transmission system with GCSC

Fig. 11. UHV Transmission system with hybrid compensator

67

68

V. Yarlagadda et al.

Fig. 12. Per phase equivalent Simulink diagram of UHV Transmission system with hybrid compensator

7.1 Power Transfer and Reactive Power Margins in 800 KV UHV Transmission System with Different Loads The following table encapsulate the simulation results of 800 KV UHV Transmission systems without any Compensator, with GCSC and Hybrid Compensators. Table 1 portrays the Power Transfer and reactive power flow in 800 KV UHV Systems for five different loads of RL type. Figure 13 shows the Power Transfer in 800 KV UHV Transmission system with Load 3, Fig. 14 depicts the Reactive Power in 800 KV UHV Transmission system with Load 3. Figure 15 illustrates the Power Transfer in 800 KV UHV Transmission system with Load 4, Fig. 16 shows the Reactive Power in 800 KV UHV Transmission system with Load 4. Figure 17 portrays Power Transfer in 800 KV UHV Transmission system with Load 5 and Fig. 18 represents the Reactive Power in 800 KV UHV Transmission system with Load 5. All these results encapsulate the relative merits and demerits of both of the controllers, the Hybrid compensator is having superior benefits compared to GCSC in power transfer capability and reactive power margins. 7.2 Power Transfer and Reactive Power Margins in 1500 KV UHV Transmission System with Different Loads The following table encapsulate the simulation results of 1500 KV UHV Transmission systems without any Compensator, with GCSC and Hybrid Compensators. Table 2 portrays the Power Transfer and reactive power flow in 1500 KV UHV Systems for five different loads of RL type. Figure 19 shows the Power Transfer in 1500 KV UHV Transmission system with Load 1, Fig. 20 depicts the Reactive Power in 1500 KV UHV Transmission system with Load 3. Figure 21 illustrates the Power Transfer in 1500 KV UHV Transmission system with Load 2, Fig. 15 shows the Reactive Power in 1500 KV UHV Transmission system with Load 3. Figure 22 portrays Power Transfer in 1500 KV UHV Transmission system with Load 4 and Fig. 23 represents the Reactive Power in 1500 KV UHV Transmission system with Load 5. Figure 24 shows the Reactive Power margin in 1500 KV UHV Transmission system with Load 3, Fig. 25 depicts the Power

Comparative Analysis of GCSC and Hybrid Compensators Influence

69

Table 1. Power transfer and reactive power flow in 800 KV UHV system with all compensators S.no Load

Parameter

800 KV UHV Transmission system Without compensator With GCSC With hybrid compensator

1

Load 1 P in MW

15.27 18.15

26.28

26.52

Q in MVAr 91.74

Load 2 P in MW

3

Load 3 P in MW

5

26.52 −10.64

2

4

26.5 −10.63

Q in MVAr 92.1

−10.55

−10.64

19.73

26.16

26.52

Q in MVAr 91.91

−10.55

−10.64

Load 4 P in MW

22.83

25.92

26.52

Q in MVAr 91.26

−10.41

−10.64

24.35

25.81

26.52

Q in MVAr 90.96

−10.36

−10.64

Load 5 P in MW

P in MW 30 20 10 0

without Compensation n MW with GCSC in Capacitive -MW with Hybrid Device in Capacitive -MW

Fig. 13. Power transfer in 800 KV UHV Transmission system with Load 3

Q in MVAr 100

50

0

without Compensation n MVAr with GCSC in Capacitive -MVAr with Hybrid Device in Capacitive MVAr

Fig. 14. Reactive power in 800 KV UHV Transmission system with Load 3

70

V. Yarlagadda et al.

P in MW 28

without Compensation n MW with GCSC in Capacitive -MW

26 24

with Hybrid Device in Capacitive -MW

22 20

Fig. 15. Power transfer in 800 KV UHV Transmission system with Load 4

Q in MW 100

without Compensation n MVAr with GCSC in Capacitive -MVAr

50 0

with Hybrid Device in Capacitive -MVAr

Fig. 16. Reactive power in 800 KV UHV Transmission system with Load 4

P in MW 28 26 24 22

without Compensation n MW with GCSC in Capacitive -MW with Hybrid Device in Capacitive -MW

Fig. 17. Power transfer in 800 KV UHV Transmission system with Load 5

Comparative Analysis of GCSC and Hybrid Compensators Influence

71

without Compensation n MVAr

Q in MVAr 100

with GCSC in Capacitive -MVAr

80 60

with Hybrid Device in Capacitive -MVAr

40 20 0

Fig. 18. Reactive power in 800 KV UHV Transmission system with Load 5

Transfer in 1500 KV UHV Transmission system with Load 4, Fig. 26 illustrates the Reactive Power margin in 1500 KV UHV Transmission system with Load 4. Figure 27 demonstrates the Power Transfer in 1500 KV UHV Transmission system with Load 5 and Fig. 28 shows the Reactive Power margin in 1500 KV UHV Transmission system with Load 5. All these results encapsulate the relative merits and demerits of both of the controllers, the Hybrid compensator is having superior benefits compared to GCSC in power transfer capability and reactive power margins. Table 2. Power transfer in 1500 KV UHV system with all compensators S. no

Load

Parameter

800 KV UHV Transmission system Without D-GCSC

1

Load 1

P in MW Q in MVAr

2

Load 2

P in MW Q in MVAr

3

Load 3

P in MW

4

Load 4

P in MW

Q in MVAr Q in MVAr 5

Load 5

P in MW Q in MVAr

39.62 106.6

With D-GCSC 64.68 110.3

With hybrid compensator 84.34 −40.9

36.9

62.3

98.3

161.0

110.3

−40.86

34.15 161.4 31.36 161.7 29.96 161.7

59.73 116.2 57.01 119.0 55.59 119.0

105.4 −47.82 98.19 −45.92 93.08 −45.92

72

V. Yarlagadda et al.

P in MW 100

without Compensation n MW with GCSC in Capacitive -MW

80 60

with Hybrid Device in Capacitive -MW

40 20 0

Fig. 19. Shows the power transfer in 1500 KV UHV Transmission system with Load 1

Q in MVAr 200 150

without Compensation n MVAr with GCSC in Capacitive -MVAr

100 50 0

with Hybrid Device in Capacitive MVAr

-50 Fig. 20. Shows the reactive power margin in 1500 KV UHV Transmission system with Load 1

P in MW 100 50 0

without Compensation n MW with GCSC in Capacitive -MW with Hybrid Device in Capacitive -MW

Fig. 21. Shows the power transfer in 1500 KV UHV Transmission system with Load 2

Comparative Analysis of GCSC and Hybrid Compensators Influence

Q in MVAr 200 100 0

73

without Compensation n MVAr with GCSC in Capacitive -MVAr with Hybrid Device in Capacitive -MVAr

-100 Fig. 22. Shows the reactive power margin in 1500 KV UHV Transmission system with Load 3

P in MW 150 100

without Compensation n MW with GCSC in Capacitive -MW with Hybrid Device in Capacitive -MW

50 0

Fig. 23. Shows the power transfer in 1500 KV UHV Transmission system with Load 3

Q in MVAr 200 100 0 -100

without Compensation n MVAr with GCSC in Capacitive -MVAr with Hybrid Device in Capacitive MVAr

Fig. 24. Shows the reactive power margin in 1500 KV UHV Transmission system with Load 3

74

V. Yarlagadda et al.

P in MW 100 50

without Compensation n MW with GCSC in Capacitive -MW with Hybrid Device in Capacitive -MW

0 Fig. 25. Shows the power transfer in 1500 KV UHV Transmission system with Load 4

Q in MVAr 200 100 0

without Compensation n MVAr with GCSC in Capacitive -MVAr with Hybrid Device in Capacitive -MVAr

-100 Fig. 26. Shows the reactive power margin in 1500 KV UHV Transmission system with Load 4

P in MW 100

50

without Compensation n MW with GCSC in Capacitive -MW with Hybrid Device in Capacitive -MW

0 Fig. 27. Shows the power transfer in 1500 KV UHV Transmission system with Load 5

Comparative Analysis of GCSC and Hybrid Compensators Influence

Q in MVAr 200 150

75

without Compensation n MVAr with GCSC in Capacitive -MVAr

100 50 0

with Hybrid Device in Capacitive -MVAr

-50 Fig. 28. Shows the reactive power margin in 1500 KV UHV Transmission system with Load 5

8 Conclusions The performance of UHV Power Transmission can be improved with the usage of FACTS Devices. The GTO Thyristor Turn off Series Capacitor (GCSC) is a powerful device among all FACTS Devices in order to control power flow and stability improvement. This article realized the fact of effectiveness of both the FACTS devices such as GCSC and Hybrid Compensator comprises of GCSC and Static Var Compensator (SVC) in improving the power Transmission Capability and reactive power margins of UHV Transmission systems. The UHV Transmission systems have been simulated with operating voltages of 800 KV and 1500 KV. Both of these systems have been simulated without, with GCSC and Hybrid Compensators and are presented in this article. The simulation results of 800 KV and 1500 KV UHV Transmission systems without any Compensator, with GCSC and Hybrid Compensators have been proving that the hybrid compensator is much more effective especially with 1500 KV UHV Transmission system with very high value of Power Transfer for the same impedance load and with surplus reactive power margins. The GCSC is also more effective in improving the Power Transfer Capability compared to without any controller in the system but it also consuming reactive power in 1500 KV EHV System.

References 1. Oshnoei, A., Kheradmandi, M., Khezri, R., Mahmoudi, A.: Robust model predictive control of gate-controlled series capacitor for LFC of power systems. IEEE Trans. Ind. Inf. 17(7), 4766–4776 (2021). https://doi.org/10.1109/TII.2020.3016992 2. Morsali, J., Zare, K., Hagh, M.T.: MGSO optimised TID-based GCSC damping controller in coordination with AGC for diverse-GENCOs multi-DISCOs power system with considering GDB and GRC non-linearity effects. IET Gener. Transmiss. Distrib. 11(1), 193–208 (2017) 3. Maza-Ortega, J.M., Acha, E., García, S., et al.: Overview of power electronics technology and applications in power generation transmission and distribution. J. Mod. Power Syst. Clean Energy 5, 499–514 (2017). https://doi.org/10.1007/s40565-017-0308

76

V. Yarlagadda et al.

4. de Souza, L.F.W., Watanabe, E.H., Alves, J.E.R., Pilotto, L.A.S.: Thyristor and gate controlled series capacitors: comparison of components rating. In: Proceedings of IEEE PES General Meeting, Toronto, July 2003 5. Watanabe, E.H., Aredes, M., de Souza, L.F.W., Bellar, M.D.: Series connection of power switches for very high power applications and zero voltage switching. IEEE Trans. Power Electron. 15(1), 44–50 (2000) 6. Jowder, F.A.R.A., Ooi, B.T.: Series compensation of radial power system by a combination of SSSC and dielectric capacitors. IEEE Trans. Power Deliv. 20(1), 458–465 (2005) 7. Umre, B.S., Helonde, J.B., Modak, J.P., Renkey, S.: Application of gate-controlled series capacitors (GCSC) for reducing stresses due to sub-synchronous resonance in turbinegenerator shaft. In: 2010 IEEE Energy Conversion Congress and Exposition, pp. 2300–2305 (2010). https://doi.org/10.1109/ECCE.2010.5617863 8. Pandey, S.K., Singh, B.: Hybrid DSC with compensation capability based control for grid integrated SPV system. In: 2020 IEEE 9th Power India International Conference (PIICON), pp. 1–6 (2020). https://doi.org/10.1109/PIICON49524.2020.9113069 9. Kakimoto, N., Phongphanphanee, A.: Sunsynchronous resonance damping control of thyristor-controlled series capacitor. IEEE Trans. Power Deliv. 18(3), 1051–1059 (2003) 10. Pilotto, L.A.S., Bianco, A., Long, F.W., Edris, A.A.: Impact of TCSC control methodologies on subsynchronous oscillations. IEEE Trans. Power Deliv. 18(1), 243–252 (2003)

Wind Energy System Using Self Excited Induction Generator with Hybrid FACTS Device for Load Voltage Control Venu Yarlagadda1(B) , Garikapati Annapurna Karthika1 , Giriprasad Ambati1 , and Chava Suneel Kumar2 1 Department of Electrical and Electronics Engineering, VNR Vignana Jyothi Institute of

Engineering and Technology, Hyderabad, India [email protected], [email protected] 2 Department of Electrical and Electronics Engineering, BVRIT Hyderabad, Hyderabad, India [email protected]

Abstract. The Modern Power systems are incorporated with renewable energy generation resources such as solar and wind power plants. The everlastingDemand for electrical power leadsto dynamic load variations on the electric grid subsequently results in frequency and voltages fluctuations. This article engrossedon development of a Hybrid FACTS Device to control load voltages over wide range of load variations on wind power plant equipped with Hybrid FACTS device with Self Excited Induction Generator (SEIG). The distribution generationsystem is developed with Hybrid FACTS Device fed SEIG based wind power plant feeding an isolated load. The wind plant is subjected to speed variations as well as load perturbations leading to voltage fluctuations; hence SEIG is best suited generator for the wind plant. The SEIG is equipped with shunt compensator or any Hybrid FACTS Device as used in the present case. This article deals with wind energy generation unit used to supply local loads.The terminal voltage has been maintained at constant nominal value for different load scenarios by the use of Hybrid FACTS Device and these results are presented in this article, which proves the effectiveness of the proposed device on maintaining constant voltage over wide range of load variations. Keywords: Series compensation with DSTATCOM · SEIG · Wind energy generation · Voltage control in wind plant · Hybrid FACTS device · Voltage stabilization · Reactive power control · Series and shunt compensation

1 Introduction The energy demand is rising day by day as the population is increasing at a very rapid rate. So to meet the energy needs we have to generate more electricity. In conventional methods to produce electricity the fossil fuels are used that pollute the environment and have a very negative impact on the environment. So to have sustainable growth © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 77–91, 2022. https://doi.org/10.1007/978-981-19-1742-4_6

78

V. Yarlagadda et al.

we need to generate more electricity by the use of renewable energy. The renewable energy includes wind energy, solar energy, tidal energy etc. These types of energy are clean energy sources and produce no pollution and do not harm the environment. The power generated by the wind must be integrated into the grid. In recent years there has been extensive growth in the use of wind energy. The wind plant is subjected to speed variations as well as load perturbations leading to voltage fluctuations. The closed loop control of the DSTATCOM ensures the voltage stabilization of load for various load scenarios presented in this article. DSTATCOM injects the reactive power into the system and hence voltage fluctuations have been reduced by the injection of reactive power. The flow of the reactive power is controlled by the control of the bus voltage, the magnitude of the bus voltage decides the direction of reactive power flow.

2 Wind Power Plant Wind Energy in Generating Electrical Power through wind was started in the nineteenth century, due to consumption of fossil fuels at a high rate that led to innovation of new ideas related to wind energy.Wind generation with large commercial scale Advancements, wind energy systems become more cost efficient and it is widely used electrical source in modern age, which minimizes the greenhouse effect and subsequently global warming [1–6]. 2.1 Wind Turbine Wind Turbine is a device that converts the wind’s kinetic energy into electrical energy. Wind turbines are manufactured in a wide range of sizes, with either horizontal or vertical axes as depicted in Fig. 1. The available range of power generation by the horizontal wind turbine is low and is 50 W to 4.5 KW, whereas vertical type is high power range is about 2.5 to 3 MW [1–6]. The wind power generation Pw is determined by the power coefficient CP as specified in the following Eqs. (1) and (2) respectively. PW is the wind power, ρ is the air density, CP is the coefficient of performance, A if frontal area and v is the velocity of wind. The power coefficient is a nonlinear function of the blade pitch angle and the blade tip speed ratio λ and is the ratio of the angular rotor speed of the wind turbine to the linear wind speed at the tip of the blades given by the Eq. (4) and 1/x in terms of λ and β as depicted in the Eq. (3) [1–6].  1 CP (λ, β)ρAV 3 (1) PW = 2   −12.5 116 CP = 0.22 − 0.4β − 5 e x (2) x 1 1 0.035 = − x (λ − 0.08β) (1 + β 3 ) λ=

wR v

(3) (4)

Wind Energy System Using Self Excited Induction Generator

79

The Wind Power Plant with Self Excited Induction Generator with DSTATCOM as illustrated in the Fig. 1 below. The wind turbine input parameters, which controls the wind power generation are wind speed (v) and pitch angle (β). These wind turbine characteristics have been illustrated in Fig. 2, which shows the power generation curves against per unit turbine speed with constant wind speed of 10 m/s, 12 m/s to maximum 24 m/s respectively. The wind turbine output is given to the SEIG excited and controlled with Hybrid FACTS Device [3, 4, 8].

Fig. 1. Wind power plant with self excited induction generator with DSTATCOM

Fig. 2. Wind turbine characteristics

80

V. Yarlagadda et al.

3 Self Excited Induction Generator Induction Machines when run at more than synchronous speed, the slip is negative and power flow reversal will be the end result, which means it is running as Generator. It has three different modes of operation, one is grid connected: in which it consumes reactive power from the grid and supplies active power to grid and is not possible for isolated loads. The second is in self excited mode of operation in which capacitor bank or any shunt reactive power compensator is used to get the self excitation process with suitable amount of reactive power and is suitable for isolated loads. The third is Doubly Fed Induction Generator, it also involves the grid in its study, and hence for isolated loads we may not prefer it. Hence SEIG is chosen in this article to present the case study of different load scenarios in isolated mode without grid connection and have been presented the results. This article deals with the SEIG equipped with closed loop control of the DSTATCOM as depicted in the Fig. 3, which consists of SEIG, DSTATCOM and loads [7, 10].

Fig. 3. Schematic diagram of 3-ph self excited induction generator with hybrid FACTS device

3.1 Mathematical Modelling of SEIG The modelling is equations of Induction Machine as described by the stator and rotor side voltage and flux linkage equations as below, Eq. (5) shows the stator voltage equation, Eq. (6) describes the stator flux linkages and similarly Eq. (7) illustrates the rotor voltage equation and Eq. (8) shows the flux linkage equation of rotor. The machine will work as motor when the speed is less than synchronous speed i.e. the slip is positive and it is around less than +5%, whereas the machine will work as Generator when the speed is greater than synchronous speed i.e. the slip is negative and it is around less than −5%, corresponding torque slip characteristics as shown in Fig. 4 [5–10].

Wind Energy System Using Self Excited Induction Generator

81

Stator side equations   d λa abcs V a abcs = rs ia abcs + dt     λa abcs = (Ls + M ) ia abcs + Lsr ia abcr

(5) (6)

Rotor side equations

λa abcr

  d λa abcr 0 = rr ia abcr + dt     = (Lr + M ) ia abcs + Lsr ia abcs

(7) (8)

Fig. 4. Torque slip characteristics of induction machine

4 Hybrid Compensator The Hybrid D-FACTS Controller is composed of a series compensator of Distributed GTO Controlled Series Capacitor (D-GCSC) and Distributed Static Synchronous Compensator (D-STATCOM). This device is used to get the superior performance of Wind Energy System (WES) against the load disturbances. 4.1 DSTATCOM The DSTATCOM schematic is connected to the receiving end bus as shown in Fig. 4, it utilizes the GTOs as switching devices and d.c. link capacitance, which is used to enhance the performance of the wind plant as illustrated in Fig. 5 below [3–6].

82

V. Yarlagadda et al.

Fig. 5. Distribution system with D-STATCOM

4.1.1 Operation of Distributed Static Synchronous Compensator (DSTATCOM) Flexible AC Transmission System (FACTS) devices have been emerged for enhancement in power system stability. The conditions like voltage stability, transient stability, reactive power flow control, power quality etc. are enhanced with power electronic based FACTS devices. STATCOM is a shunt connected device which employs power electronic devices such as IGBT, GTO, MOSFET etc. as depicted in Fig. 4, which are basically fast acting switching devices in order to improve stability of power system and control of reactive power flow i.e. by absorbing reactive power from the system or by generation of reactive power meeting the demand to maintain the voltage at specific limits [3–5]. The working principle of DSTATCOM, consider Inverter output voltage as V1 and system output voltage as V2. The exchange of reactive power in between the DSTATCOM and system is based on the voltages V1 and V2 i.e. if the demand in reactive power in the system increases, then the output voltage of the DSTATCOM gets increases and vice versa with no flow of active power in between the system and DSTATCOM by keeping the angle zero [4]. DSTATCOM comprises mainly Voltage Source Converter (VSC), is generally GTO type and IGBT based converters which converts DC voltage into AC voltage. In GTO type converter, AC output voltage can be varied by DC capacitor input voltage and similarly in IGBT based converter, Pulse Width Modulation (PWM) technique is used to generate a sinusoidal wave form with frequency of kHz from DC voltage. DC capacitor supply constant voltage to VSC and a transformer is coupled in between power system and DSTATCOM and also reduces the harmonics in square wave generated by VSC [1–10].

Wind Energy System Using Self Excited Induction Generator

83

4.1.2 Mathematical Modelling of DSTATCOM The equations for active power, reactive power of STATCOM are as follows; Consider Vt = system terminal voltage Vst = STATCOM output voltage XL = Inductive reactance VC = DC capacitor voltage Vt VC sinα XL

(9)

Vt Vt Vt VC − cosα XL XL

(10)

P= Q=

The equation of DSTATCOM DC side can be given as; The mathematical equations of DSTATCOM can be expressed as; L = series inductance R = series resistance iac , ibc , icc are output currents of DSTATCOM Vac , Vbc , Vcc are output voltages of DSTATCOM Vta , Vtb , Vtc are terminal voltages

L

diac = Riac + Vac − Vat dt

(11)

L

dibc = Ribc + Vbc − Vbt dt

(12)

L

dicc = Ricc + Vcc − Vct dt

(13)

4.2 D-GCSC Distributed system to carry out the simulation study of our interest is designed with a feeder connected in series with D-GCSC used to feed both resistive and RL loads as shown in the Fig. 6 below. The same system with DSTATCOM for mitigating current harmonics subsequently voltage harmonics in the distributed system [6–8].

84

V. Yarlagadda et al.

Fig. 6. Single line diagram of D-GCSC and DSTATCOM

The distribution feeder with feeder impedance connected in series with the D-GCSC feeding both resistive and RL loads as shown in Fig. 7. The D-GCSC schematic is described with the antiparallel combination of GTO Thyristors used to control the series injected voltage with the feeder as depicted in the below figure. The reactance and voltage variation of D-GCSC with the variation of the conduction angle as depicted by the following Eqs. (1) and (2) respectively. The harmonics injected by the device is illustrated in the Eq. (3) and the voltage wave form of complete control of the device is illustrated in Fig. 8 comprising of parts a, b and c as (a) D-GCSC Schematic circuit, (b) One complete cycle, (c) total current of D-GCSC. The total current shown in the waveform comprises of harmonics since the conduction angle is rapidly adjusted for controlling series voltage of the system [4–10].

Fig. 7. Schematic diagram of D-GCSC

Wind Energy System Using Self Excited Induction Generator

  2γ sin(2γ ) 1 1− − wC π π   2γ I sin(2γ ) 1− VCF (γ) = − wC π π    sin(γ )cos(nγ ) − nsin(nγ )cos(γ ) sin(2γ ) I 4 − VCn (γ) = wC π n(n2 − 1) π Xc(γ) =

85

(14) (15) (16)

Fig. 8. (a) D-GCSC Schematic circuit, (b) One complete cycle, (c) total current of D-GCSC

5 Case Study and Conclusions The Simulink diagram of the wind energy generation plant furnished with the SEIG controlled with the Hybrid D-FACTS device built with D-GCSC and DSTATCOM along with the different loads in isolated mode of operation as illustrated in Fig. 9. The wind speed is taken as 15 m/s and pitch angle as 30° as shown in the Fig. 9 below. The simulation is carried out with widely variable loadsincluding Hybrid Compensator and results have been presented in this article. Figure 10 depicts the DSTATCOM Simulink and Fig. 11 shows the wind energy generation including D-GCSCand Fig. 9 illustrates the Wind Energy Generation with Hybrid Compensator and SEIG. Figure 12 SEIG shows the currents, voltages and torque waveforms with small loads, Fig. 13 depicts the SEIG currents, voltages and torque waveforms with small loads and voltage is at nominal value of 400V and per unit voltage of 1 p.u. as illustrated by the both the waveforms. Figure 14 SEIG shows the currents, voltages and torque waveforms with medium loads, Fig. 15 depicts the SEIG currents, voltages and torque waveforms with medium

86

V. Yarlagadda et al.

loads and voltage is at nominal value of 400V and per unit voltage of 1 p.u. as illustrated by the both the waveforms. Figure 16 SEIG shows the currents, voltages and torque waveforms with heavy loads, Fig. 17 depicts the SEIG currents, voltages and torque waveforms with heavy loads and voltage is at nominal value of 400V and per unit voltage of 1 p.u. as illustrated by the both the waveforms. In all load scenarios, the terminal voltage is maintained at strictly its nominal value of 400V and in per unit it is 1p.u, which is possible with the Hybrid Compensator.

Fig. 9. Wind energy generation with hybrid compensator and SEIG

Fig. 10. DSTATCOM simulink power circuit

Wind Energy System Using Self Excited Induction Generator

Fig. 11. DSTATCOM simulink closed loop control circuit

Fig. 12. Load voltage and current waveforms with small loads

87

88

V. Yarlagadda et al.

Fig. 13. SEIG currents, voltages and torque waveforms with small loads

Fig. 14. SEIG currents, voltages and torque waveforms with medium loads

Wind Energy System Using Self Excited Induction Generator

Fig. 15. SEIG currents, voltages and torque waveforms with medium loads

Fig. 16. SEIG currents, voltages and torque waveforms with heavy loads

89

90

V. Yarlagadda et al.

Fig. 17. SEIG currents, voltages and torque waveforms with heavy loads

6 Conclusions This article engrossedon development of a Hybrid FACTS Device to control load voltages over wide range of load variations on wind power plant equipped with Hybrid FACTS device with Self Excited Induction Generator (SEIG). The distribution generation system is developed with Hybrid FACTS Device fed SEIG based wind power plant feeding an isolated load. The SEIG is equipped with shunt compensator or any Hybrid FACTS Device as used in the present case. This article deals with wind energy generation unit used to supply local loads. The terminal voltage has been maintained at constant nominal value for different load scenarios by the use of Hybrid FACTS Device and these results are presented in this article. The Simulink diagram of the wind energy generation plant furnished with the SEIG controlled with the Hybrid D-FACTS device built with DGCSC and DSTATCOM along with the different loads in isolated mode of operation. The simulation is carried out with widely variable loads including Hybrid Compensator and results have been presented herewith. The load on the wind energy generating system is varied from small load to heavy loads with Compensator, the results are illustrating the robust performance of the compensator against wide variations of loads and terminal voltage is maintained at rated voltage at all loads.

References 1. Hook, K.S., Liu, Y., Atcitty, S.: Mitigation of the wind generation integration related power quality issues by energy storage. EPQU J. 12(2), 77–82 (2006) 2. Carrasco, J.M., et al.: Power-electronic systems for the grid integration of renewable energy sources: a survey. IEEE Trans. Ind. Electron. 53(4), 1002–1016 (2006) 3. Han, C., Huang, A.Q., Baran, M., Bhattacharya, S., Litzenberger, W.: STATCOM impact study on the integration of a large wind farm into a weak loop power system. IEEE Trans. Energy Conv. 23(1), 226–232 (2008)

Wind Energy System Using Self Excited Induction Generator

91

4. Bhatia, R.S., Jain, S.P., Jain, D.K., Singh, B.: Battery energy storage system for power conditioning of renewable energy sources. In: Proceedings of International Conference on Power Electron Drives System, vol. 1, pp. 501–506, January 2006 5. Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based backstepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 6. Mohod, S.W., Aware, M.V.: Power quality issues & it’s mitigation technique in wind energy conversion. In: Proceedings of IEEE International Conference on Quality Power & Harmonic, Wollongong, Australia, 2122 (2008) 7. Dyanamina, G., Kakodia, S.K.: SEIG voltage regulation with STATCOM regulator using fuzzy logic controller. In: 2021 International Conference on Sustainable Energy and Future Electric Transportation (SEFET), pp. 1–6 (2021) 8. Ezzeddine, T.: Reactive power analysis and frequency control of autonomous wind induction generator using particle swarm optimization and fuzzy logic. Energy Explor. Exploit. 38(3), 755–782 (2020) 9. Meena, R.S., Lodha, M.K., Parira, A.S., Gupta, N.: Hybrid micro-generation scheme suitable for wide speed range and grid isolated remote applications. In: Pant, M., Ray, K., Sharma, T.K., Rawat, S., Bandyopadhyay, A. (eds.) Soft Computing: Theories and Applications. AISC, vol. 583, pp. 185–193. Springer, Singapore (2018). https://doi.org/10.1007/978-981-10-56871_17 10. Chilipi, R.R., Singh, B., Murthy, S.S.: Performance of a self-excited induction generator with DSTATCOM-DTC drive-based voltage and frequency controller. IEEE Trans. Energy Convers. 29, 545–557 (2014)

Microgrid Islanding Detection Using Travelling Wave Based Hybrid Protection Scheme Shashank Gupta(B) and Suryanarayana Gangolu Department of Electrical and Electronics Engineering, National Institute of Technology, Uttarakhand, Srinagar (Garhwal), India [email protected]

Abstract. Travelling wave protection scheme is based on propagated forward and backward electromagnetic waves. During islanded mode of operation, microgrid gets completely isolated from main grid and this disturbance initiates the forward and backward electromagnetic waves. The frequency of transmission line voltage is monitored and when islanding occurs, it is used to signal the travelling wave protection scheme for detection. In this chapter, travelling wave reflection coefficient and refraction coefficient principles are used to detect un-intentional islanding. The techniques are implemented in MATLAB-SIMULINK. Keywords: Microgrid · Travelling wave · Reflection coefficient · Refraction coefficient · IDTs · Voltage frequency monitoring · Tripping signal · Signal perturbation

1 Introduction 1.1 Motivation The Microgrid (MG) is capable to solve the present energy crisis and comprises of distributed generation sources (also known as Distributed Energy Resources (DER)), storage system, etc. The MG can operate either in grid-connected mode or in case of a islanding it operates in the isolated mode of operation. MG is integrated with distributed energy resource units, supply different loads in a distribution system and maintain to run in synchronism with the main grid, islanded mode from the main grid, and also maintain healthy transitions between the main grid-connected and islanded mode [1]. The increasing contribution of distribution generating units in Microgrid has many protection issues, like both transient and steady-state over-voltage and under-voltage at point of common coupling, unnecessary tripping due to maloperation, increment in short circuit levels, and degrading power quality [2]. The upcoming protection problem with the DER units like Photovoltaic systems, wind turbines, the gas turbine in the microgrid is very significant [3]. Whenever, unintentional islanding occurs, DER units still feed power to the remaining part of the Microgrid, isolated from the main grid [4]. For the maintenance of the transmission line, one has to disconnect both sides of the transmission line. An important problem of power systems with integration of DGs, is the scope of © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 92–105, 2022. https://doi.org/10.1007/978-981-19-1742-4_7

Microgrid Islanding Detection Using Travelling Wave

93

this chapter, is sudden unintentional disconnection of Microgrid from the main grid. If islanding is not detected within a specific predefined protection time limit, then it results in synchronizing problems in reclosing, life and safety threatening to the lineman working near the line, unwanted abrupt voltage change, and decrement in system inertia which further make undesirable fluctuation in line frequency [4]. Therefore, it is required to pause the functioning of DG immediately after detection of unintentional islanding in the Microgrid system. Islanding detection can be categorized mainly of two types, local end data-based and remote end data access-based [4]. In the Remote end, data-based technique communication channels between micro-grid and protecting relays are used for detection of unintentional islanding [5]. The local end data-based IDT is divided into three subgroups of active methods, passive method, and combining these two called hybrid methods [6]. In Passive detection technique, the transmission line grid variable parameters like frequency of line, the voltage of the line, rate of change of frequency (ROCOF) of line, rate of change of phase angle difference (ROCOPAD) between two instances, impedance variation before and after islanding, etc. are tapped directly for detection of islanding in the Microgrid. However, zero power mismatch drawback of passive detection techniques can be eliminated by using active based and communication-based islanding detection method [4]. In the Active practice based islanding detection technique, islanded mode is detected by injecting a signal or making a signal distortion in the grid and analyze the line parameter response at the point of common coupling for microgrid and the main grid [7]. Communication-based techniques use communication channels to communicate for data sharing and therefore it does not have any Non-detection zone problem, but if somehow communication channel failed to respond then it would be a disadvantage for this technique for detection of islanding. Therefore, active detection methods which have negligible Non-detection zone problems are more feasible than passive methods and communication-based islanding detection techniques. To prevent continuous monitoring operation of active method which reduces voltage quality of grid, a hybrid technique based on both active and passive method is proposed [8]. 1.2 Literature Review Some hybrid IDTs presented in various literature surveys are: In [10] ROCOV and ROCOAP based algorithm is implemented, in [11] combination of passive with active ROCOF based methods are used, in [12] simultaneous drop of the active power output and the PCC voltage are used for islanding detection, in [13] Voltage based Stockwell transform (ST) algorithm is used, in [14] combinedly three passive criteria (VV), (VU) and (ROCOF) with newly designed (CF) between RPD and FV is used, in [15] VPA and VU based hybrid algorithm is used, in [16] VU with THD based passive detection and BRPV based active detection algorithm is used, in [21] (ROCOAP) and (ROCORP) based hybrid scheme is used, in [26] IDS and VUF based scheme is used, in [27] AIP method with combinedly reactive power-frequency droop and (ROCOF) based algorithm is used, in [28] dual active with reactive control power loop and signal processing based scheme is used, in [29] cloud-based IoT and machine learning (ML) based scheme is used, in [30] Q-factor WT with ANN based algorithm is used, in [31] Discrete fractional

94

S. Gupta and S. Gangolu

FT based algorithm is used, in [32] Pattern-recognition with cuckoo optimal adjustment based algorithm is used. 1.3 Contribution and Organization of the Chapter This chapter aims, a hybrid based method for the detection of an islanded mode of Microgrid, based on the theory of traveling waves and frequency variations of line parameters. To perform this technique, simulation is performed in MATLAB/Simulink software [9]. Moreover, in this chapter, a prototype model of the transmission line is modelled in Simulink which shows reflections and refraction principles of traveling wave theory which is used to detect islanding in Microgrid. This chapter is organized in following manner: Sect. 2, illustrates the islanding detection schemes. In Sect. 3, Proposed islanding scheme is presented. Simulation results & discussion are presented in Sect. 4. Finally, conclusions of this chapter given in Sect. 5.

2 Microgrid Islanding Detection Techniques A Microgrid network is a highly complex network and it is the combination of various renewable energy resources, in this chapter, solar photovoltaic renewable grid-connected systems are modeled to test the unintentional islanding. Proper integration of these renewable energy units with microgrid depends on factors like DG inverter layouts, technology installed for anti-islanding, maintain a balance between net generated output power from DG and the remaining load connected after islanding, continuously unpredictable load change behavior in the distribution network, etc. [23]. A novel IDTs are categorized into, Passive methods, Active technique, Hybrid methods, communication techniques. 2.1 Passive Detection Technique In this method, the changes in transient conditions are monitored at the coupling point which connect the main grid with microgrid, these transient disturbances are compared with the threshold parameters, based on this comparison islanding condition is confirmed. For comparison with threshold, we compare with system parameters such as line frequency, line voltages, distortion in harmonics, etc. during islanded mode these parameters show evident disturbance for confirmation of islanded situation [23]. There are certain advantages of this method over other methods like the requirement of hardware circuit is very less and this let the overall cost of the circuit to be very less [23]. Non-detection zone drawback in passive technique occurs when power generation is equal to power demand after microgrid gets isolated due to unintentional islanding. While setting up the parameter comparison limit, we must ensure that the lower limit should not have to be very low because in that that the relay may encounter false tripping, whereas we also cannot set the upper limit to be very high because in such case again relay may not give trip signal even islanding occurred. To overcome this Non-detection zone problem, active detection methods are used which can accurately work even if generated power is equal to demand after the microgrid gets isolated [1].

Microgrid Islanding Detection Using Travelling Wave

95

2.2 Active Detection Technique In this method, signal injection and then continuously monitoring of system parameters process is used to detect the un intentional islanding. Basic monitoring parameters which can be analyzed after signal injection can be line voltage of network, frequency change analysis, variation impedance after perturbation, etc. In this method, we do not have the problem of a Non-detection zone [23]. This signal injection technique is required to inject and monitor it continuously, due to this the overall system power quality is decreased and, in this fashion, most of the time signal is unnecessarily continuously present in the line Another limitation of this technique is that it will not give valid results when DGs are operated at a power factor equal to one. For inverter fed DGs mostly shifting in phase are compared to get detection alert. Active Frequency Drift technique is used for mainly pure resistive nature load but this method has chances of failure for some other non-resistive load. Whenever there is a change in phase or shift is encountered in the system by methods like Slip mode frequency shift, Active frequency drift with positive feedback, Automatic phase shift, and Adaptive logic phase shift, will degrade the quality of output power of the system [1]. 2.3 Hybrid Detection Technique In this method, both passive methods and active methods work simultaneously which identify islanding. The main advantage of this combination is that one of the two technologies will remove the drawback of the other, in this manner this hybrid combination offer almost negligible Non-detection zone problem and signal injections are also not always injected into line, it will come into the picture only when there is an abrupt disturbance is confirmed. Therefore, the quality of power output will also not deteriorate [23]. Due to the hybrid combination of passive with active technique, there is an improvement in power output quality but detection also increased due to confirmation lag by passive hybrid with active method [22]. 2.4 Communication-Based Technique This method is the most stable method among all other anti-islanding techniques and for implementation of this anti-islanding technique, mostly Programmable logic controllers and supervisory control data acquisition methods are used [22]. Forgiving signal of present status to such acquisition system, the communication signals which are used to communicate between DG inverter panel and main grid panel, are analyzed and at the moment islanding occur there will be a disturbance occur in these signals and with the help of these disturbance signals islanding detection is confirmed [23]. In the Power Line Communication Signaling method, one signal is directed to the DG units directly from the line to detect islanded mode. With the installation of this signaling, any disturbance noise caused by active detection methods is removed [22]. This type of technique is generally installed in those areas where the quality of output power is much more concerned over the installation cost.

96

S. Gupta and S. Gangolu

3 Proposed Islanding Detection Scheme During islanded mode of operation microgrid still feeds the remaining connected load by the DGs which are still part of an isolated microgrid system. Our main focus is to design a system in such a manner that the moment islanding is confirmed by the anti-islanding mechanisms, the remaining DGs in the islanded microgrid should be shut down immediately. According to standards of IEEE 929-1988, DGs must disconnect immediately after islanding is confirmed. According to standards of IEEE 1547-2003, the maximum allowable time for un intentional islanding detection is within 2 s [25]. Therefore, in this chapter voltage frequency monitoring and Traveling wave theory-based hybrid protection scheme is introduced, and results simulated in MATLAB/SIMULINK software. 3.1 Passive Method: Frequency Monitoring For the Passive technique, Voltage frequency is used for continuously monitoring the unbalance condition. Therefore, any islanding disturbance that occurred in the line frequency can be direct used for the conformation of islanded mode. According to CERC Standards for power frequency in India, the nominal frequency of operation in Indian grid is 50 Hz and the permissible frequency band specified by IEGC is 49.5 Hz to 50.2 Hz w.e.f. 3rd May 2010. Therefore in this paper, frequency monitoring algorithm continuously monitor frequency at PCC and if it remains under the threshold value of 50.2 Hz then the output of frequency monitoring algorithm maintained at 50 Hz but when islanding occur at t = 1 s. this will make the monitored frequency crosses the IEGC permissible limit of 50.2 Hz then immediately the output of frequency monitoring algorithm will set at 50.1 Hz and this output frequency signal will trigger voltage perturbance mechanism for active detection as shown in Fig. 5. Change in load and change in generation by extra integration of renewable energy resources in a microgrid may lead to some mal-operation due to noise introduced in the line. Therefore, to remove these deviations in frequency, a delay in the system signal input is provided so that it can make a clear distinction between islanded mode and normal load discrepancy as given in Fig. 1. The flow chart for frequency monitoring is shown in Fig. 2. 3.2 Active Method: Travelling Wave The active method is comprehensively based on traveling wave theory. Travelling waves in the transmission system are composed of a combination of two waves, forward wave and backward wave [20]. This wave originates in line when there is an abrupt change occur in the impedance point, which leads these waves to propagate with a multiplication factor of reflection and refraction coefficients. Therefore, in this paper, traveling wave-based hybrid mechanism is proposed which is more reliable and the detection of islanding is based on the time required to receive the first reflected wave from the islanded disturbance point [24]. In this paper, this reflection coefficient factor is calculated for both the receiving (utility grid) end and sending (at PCC) end side for islanding detection considering

Microgrid Islanding Detection Using Travelling Wave

97

Fig. 1. Extracting the islanding detection signal using system frequency

long transmission lines modeled as distributed electrical networks [18]. The reflection coefficient factor at both ends are: At Receiving end (utility grid): ρR = At Sending end (at PCC): ρS =

ZR −Zc ZR +ZC

ZS −ZC ZS +ZC

Where, ZS = impedance offered by sending end side (at PCC) ZR = impedance offered by receiving end side (utility grid) ZC = characteristic impedance of transmission line connecting utility grid and PCC. Case1: (Normal Condition) In normal conditions, the main grid (or utility grid) is connected to the microgrid without any abnormal situation. Both the Utility grid and source-side offer very little impedance in normal working condition as compared to the characteristic impedance of the transmission line (ZC ) which connect utility grid (main grid) and point of common coupling, shown in Fig. 7. Therefore, we can say that: reflection coefficient factor at both the grid side and source side is approximately equal to (−1), i.e.: [ρS = −1, ρR = −1]. Case2: (Islanded Mode) In case islanding occurs at the main grid side then the circuit breaker at the main grid side opens and the microgrid will be islanded. So, in this case due to circuit breaker operation microgrid will be completely isolated from the main grid. Therefore, in this case, impedance offered by the main grid (ZR ) would infinite so the reflection coefficient factor at the grid side now will be (+1) whereas the reflection coefficient factor at the source side will remain the same equal to (−1), i.e.: [ρS = −1, ρR = +1]. From the above cases, it is evident that in normal working conditions change in the reflected wave received at the PCC will be significant and this first reflected wave shows a positive peak at the transmitter end. In islanded mode reflection coefficient factor at the utility grid is unity, therefore, the complete transmitted wave will reflect without any change and this first reflected wave shows a negative peak at the transmitter end. In an islanded mode of operation, the frequency variation will be seen to a significant amount and this will trigger the voltage perturbation in the transmission line which later after its first reflection, shows a negative peak at the transmitted end and in this way, we can detect the islanded mode of operation based upon this hybrid technique of frequency monitoring and the first reflected wave received at PCC. The flow chart of proposed hybrid detection technique is shown in Fig. 3.

98

S. Gupta and S. Gangolu

Fig. 2. Flow chart for frequency monitoring

Microgrid Islanding Detection Using Travelling Wave

Fig. 3. Flow chart of proposed hybrid detection technique

99

100

S. Gupta and S. Gangolu

4 Simulation Results and Discussions In MATLAB, the provision for simulation of travelling wave is directly not possible because only π model of transmission line is available. Therefore, a prototype model is used, which can give the results for traveling wave after reflection at different impedance variations point due to islanding disturbance [17]. Line parameters for MATLAB simulation are provided in Table 1. Simulation parameters for the power system are defined in Table 2. Table 1. Specifications for simulation S. no.

Line variables

Values

1

RS ()

0.05

2

Rr ()

19.87

3

Line capacitance (F/Km)

7.5 × 10−9

4

Line inductance (H/Km)

5.7 × 10−7

5

Line length (Km)

20

6

ZC ()

8.72

7

TL (sec)

10.2 × 10−5

8

ρS

−0.7

9

ρr

0.7

Table 2. Simulation parameter of power grid system S. no.

Parameter

Values

1

Generator type

Synchronous

2

Output power (kW)

150

3

L-L RMS voltage (V)

230 ×

4

Local load (kW)

75

5

Grid frequency (Hz)

50

√ 3

4.1 Simulink Results For simulation, islanding is shown at t = 1 s. by opening grid side CB 1 as shown in Fig. 4. The results are observed as follows: From Fig. 5, we can observe that islanding incepted at t = 1 s, the frequency monitoring mechanism detects this abnormal change to be greater than the threshold value and this will trigger a signal perturbation mechanism. In this hybrid mechanism, after detection of the undesirable situation by the frequency monitoring system, the voltage signal perturbation mechanism gets triggered immediately and will inject the signal at the common point of coupling (PCC), shown in Fig. 6.

Microgrid Islanding Detection Using Travelling Wave

101

After signal injection, islanding will be detected based on the first reflecting wave received at PCC. The resistance of send side is always constant and lower than ZC , the ρS has always negative value. When the main grid relay gives the trip signal for islanding the Microgrid, the resulting terminated resistance will be infinite and consequently ρR becomes equal to 1, thus the first reflected voltage has a negative value at the PCC as shown in Fig. 7. This Negative Peak confirms the occurrence of islanding and later this negative peak signal the shutdown mechanism of DG connected to the Microgrid. In this signal perturbation, the effect of this traveling wave is hardly observable and hence voltage perturbation is very low in the proposed method. The net system voltage will also not be very much affected by the signal injection and offer very little total harmonic distortion (THD), as shown in Fig. 8. Proposed method confirms and give trip signal in 22 ms after CB 1 open for islanding.

Fig. 4. a) Normal grid connected mode b) Islanded mode

102

S. Gupta and S. Gangolu

Fig. 5. Frequency monitoring at PCC

Fig. 6. Voltage perturbation at PCC

Microgrid Islanding Detection Using Travelling Wave

Fig. 7. Reflected wave received at PCC

Fig. 8. System voltage with signal perturbation

103

104

S. Gupta and S. Gangolu

5 Conclusion In this proposed scheme, a hybrid technique is introduced in which both frequency monitoring and traveling wave-based principles are used to detect islanding. But now signal injection will be done only when islanding disturbance signal is accurately distinguished between overload unbalance condition and islanding condition whether linear or nonlinear load are present, this will improve the output quality of system generated power [19]. The quick response with low perturbation of this method makes it more reliable than other hybrid methods. To verify the proposed method in this study, comprehensive simulations in MATLAB/Simulink are carried out.

References 1. Hooshyar, A., Iravani, R.: Microgrid protection. Proc. IEEE 105, 1332–1353 (2017) 2. Laghari, J.A., Mokhlis, H., Karimi, M., et al.: An islanding detection strategy for distribution network connected with hybrid DG resources. Renew. Sustain. Energy Rev. 45, 662–676 (2015) 3. Aftab, M.A., Hussain, S.M.S., Ali, I., et al.: Dynamic protection of power systems with high penetration of renewables: a review of the traveling wave based fault location techniques. Int. J. Electr. Power Energy Syst. 114, 105410 (2020) 4. Ganivada, P.K., Jena, P.: Frequency disturbance triggered D-axis current injection scheme for islanding detection. IEEE Trans. Smart Grid 11, 4587–4603 (2020) 5. Sareen, K., Bhalja, B.R., Maheshwari, R.P.: Evaluation of superimposed sequence components of currents based islanding detection scheme during DG interconnections. Int. J. Emerg. Electric Power Syst. 17, 1–14 (2016) 6. Samet, H., Hashemi, F., Ghanbari, T.: Minimum non detection zone for islanding detection using an optimal Artificial Neural Network algorithm based on PSO. Renew. Sustain. Energy Rev. 52, 1–18 (2015) 7. Laghari, J.A., Mokhlis, H., Karimi, M., et al.: Computational Intelligence based techniques for islanding detection of distributed generation in distribution network: a review. Energy Convers. Manag. 88, 139–152 (2014) 8. Ohrstrom, M., Geidl, M., Soder, L., et al.: Evaluation of travelling wave based protection schemes for implementation in medium voltage distribution systems. In: 18th International Conference and Exhibition on Electricity Distribution (CIRED 2005). IET (2005) 9. Nusinovich, G.S., Li, H.: Theory of gyro-travelling-wave tubes at cyclotron harmonics. Int. J. Electron. 72, 895–907 (1992) 10. Seyedi, M., Taher, S.A., Ganji, B., et al.: A hybrid islanding detection method based on the rates of changes in voltage and active power for the multi-inverter systems. IEEE Trans. Smart Grid 12, 2800–2811 (2021) 11. Rami Reddy, C., et al.: Hybrid ROCOF relay for islanding detection. J. Electr. Eng. Technol. 17, 51–60 (2022). https://doi.org/10.1007/s42835-021-00856-9 12. Bakhshi-Jafarabadi, R., Popov, M.: Hybrid islanding detection method of photovoltaic-based microgrid using reference current disturbance. Energies 14, 1390 (2021) 13. Mahela, O.P., Sharma, Y., Ali, S., et al.: Voltage-based hybrid algorithm using parameter variations and Stockwell transform for islanding detection in utility grids. Informatics 8, 21 (2021) 14. Chen, X., Li, Y., Crossley, P.: A novel hybrid islanding detection method for grid-connected microgrids with multiple inverter-based distributed generators based on adaptive reactive power disturbance and passive criteria. IEEE Trans. Power Electron. 34, 9342–9356 (2019)

Microgrid Islanding Detection Using Travelling Wave

105

15. Kapoor, G., Walde, P., Shaw, R.N., Ghosh, A.: HWT-DCDI-based approach for fault identification in six-phase power transmission network. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 395–407. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_29 16. Wang, G.: Design consideration and performance analysis of a hybrid islanding detection method combining voltage unbalance/total harmonic distortion and bilateral reactive power variation. CPSS TPEA 5, 86–100 (2020) 17. Chapter 4: Traveling waves. In: Basic Wave Analysis, pp. 121–137. Society of Exploration Geophysicists and the European Association of Geoscientists and Engineers (2020) 18. Xu, W., Zhang, G., Li, C., et al.: A power line signaling based technique for anti-islanding protection of distributed generators—part I: scheme and analysis. IEEE Trans. Power Deliv. 22, 1758–1766 (2007) 19. Sbordone, D.A., Huq, K.M.M., Baran, M.: An experimental microgrid for laboratory activities. In: 2015 IEEE 15th International Conference on Environment and Electrical Engineering (EEEIC). IEEE (2015) 20. Dolatabadi, S., Tohidi, S., Ghasemzadeh, S.: A new active method for islanding detection based on traveling wave theory. IJEEE 14(4), 382–391 (2018) 21. Jhuma, U.K., Mekhilef, S., Mubin, M., et al.: Hybrid islanding detection technique for malaysian power distribution system. In: 2020 IEEE 5th International Conference on Computing Communication and Automation (ICCCA). IEEE (2020) 22. Pouryekta, A., Ramachandaramurthy, V.K.: A hybrid islanding detection method for distribution systems. Distrib. Gener. Altern. Energy J. 33, 44–67 (2018) 23. Llonch-Masachs, M., Heredero-Peris, D., Chillón-Antón, C., et al.: Impedance measurement and detection frequency bandwidth, a valid island detection proposal for voltage controlled inverters. Appl. Sci. 9, 1146 (2019) 24. Kapoor, G., Mishra, V.K., Shaw, R.N., Ghosh, A.: Fault detection in power transmission system using reverse biorthogonal wavelet. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 381–393. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_28 25. Pinto, S.J., Panda, G.: Performance assessment of islanding detection using complex wavelet in a three-phase utility interactive inverter system. In: 2017 IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC). IEEE (2017) 26. Nayak, A.M., Mishra, M., Pati, B.B.: A hybrid islanding detection method considering voltage unbalance factor. In: 2020 IEEE International Symposium on Sustainable Energy, Signal Processing and Cyber Security (iSSSC). IEEE (2020) 27. Raipala, O., Makinen, A., Repo, S., et al.: An anti-islanding protection method based on reactive power injection and ROCOF. IEEE Trans. Power Deliv. 32, 401–410 (2017) 28. Zamani, R., Golshan, M.E.H., Alhelou, H.H., et al.: A novel hybrid islanding detection method using dynamic characteristics of synchronous generator and signal processing technique. Electric Power Syst. Res. 175, 105911 (2019) 29. Ali, W., Ulasyar, A., Mehmood, M.U., et al.: Hierarchical control of microgrid using IoT and machine learning based islanding detection. IEEE Access 9, 103019–103031 (2021) 30. Kumar, S.A., Subathra, M.S.P., Kumar, N.M., et al.: A novel islanding detection technique for a resilient photovoltaic-based distributed power generation system using a tunable-Q wavelet transform and an artificial neural network. Energies 13, 4238 (2020) 31. Dutta, S., Olla, S., Sadhu, P.K.: A secured, reliable and accurate unplanned island detection method in a renewable energy based microgrid. Eng. Sci. Technol. Int. J. 24, 1102–1115 (2021) 32. Marín-Quintero, J., Orozco-Henao, C., Velez, J.C., et al.: Micro grids decentralized hybrid data-driven cuckoo search based adaptive protection model. Int. J. Electr. Power Energy Syst. 130, 106960 (2021)

Power Flow and Stability Improvement in Distribution Systems Using Phase Angle Regulator Venu Yarlagadda, M. Naga Jyothi, G. Lakshminarayana, and T. Hari Priya(B) VNRVJIET, Hyderabad, India {nagajyothi_m,lakshminarayana_g,haripriya_t}@vnrvjiet.in

Abstract. The increased penetration of renewable power resources into power grid leading to more stability and security issues in the grid. The dynamic power flow control as per system requirements is highly desirable in order to ensure the system security. The congested management and power flow control can be achieved with FACTS Controllers. One among the FACTS devices is the Phase Angle Regulator (PAR), which is used in the article for controlling the power flow and system stability. The power and control circuits of the PAR has been developed in Maltlab Simulink environment and has been simulated with domestic, commercial and distribution systems with phase angle control and results have been presented in the article. The control logic has been developed to improve the power flow with automatic switching action of PAR for the desired power flow in a feeder. Finally laboratory based testing have been performed without and with PAR with different phase angles and loads such as small, medium and heavy loads and results have been presented herewith and which proves the effectiveness of PAR in improving power transfer capability and stability. Keywords: Power flow control · Phase Angle Regulator · Power system stability · Congested management · Power system security · FACTS devices · Phase shifting transformer · Zigzag transformer

1 Introduction In the present scenario the rise of power demand leads to interconnection of power system for various reasons like reducing economical cost and to enhance reliability of the system. It leads to growing complexity which will cause further system collapse due to major outages. Power flow control and congested management playing a crucial role in power system security and power restoration process. Phase Angle Regulator (PAR) can be widely used in the low marginal control of power especially in commercial and distribution systems. The PAR is used to control the phase angle between two buses, which are interconnected with a transmission line or feeder. This alteration of the phase angle leading to the variation of load angle and subsequently the power flow and congested management in the system. In this work the test systems of domestic, commercial and © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 106–120, 2022. https://doi.org/10.1007/978-981-19-1742-4_8

Power Flow and Stability Improvement in Distribution Systems

107

distribution systems have been developed to carry out the simulation and testing. The developed systems have been simulated without and with PAR and presented the results. The domestic system is experimentally tested in the laboratory and results have been presented and prove the effectiveness of the PAR [1–6].

2 Power System Stability and Congested Management Power system stability is defined as a system ability to regain its initial equilibrium state after being subjected to a disturbance and its classificationare given in Fig. 1.

Fig. 1. Power system stability stratification

2.1 Rotor Angle Stability The system should remain in synchronism even after being subjected to the disturbance, which involves output power oscillates reflected in rotor oscillations. 2.1.1 Relation Between Power and Angle Relation between the power and angular position of a rotor in synchronous machine is nonlinear relation, when the synchronous generator is feeding a synchronous motor through transmission line. The power transferred to the motor from the generator is depends on the function of angular displacement between rotors of the generator and motor this is because of Motor internal angle, generator internal angle, and the angular

108

V. Yarlagadda et al.

displacement between motor and generator terminal voltage and power flow is given by the Eq. 1. P12 =

V1 V2 sin δ X12

(1)

This equation say that the power transferred to a motor from generator is maximum when the angle is 90°, if the angle is further increased beyond 90°, power transferred starts decreasing is shown in Fig. 2. The maximum power transferred is directly proportional to machine internal voltage [5–10]. P12max =

V1 V2 X12

(2)

Fig. 2. Power angle curve

2.2 Congested Management The congested management of the power system is related to the KVA or MVA capacity of the lines or feeders in the ring main distribution system and Transmission Systems. The Fig. 3 illustrating the need of Congested Management since some of the lines are carrying its power to its maximum MVA limit nearly 100%. These situations can be well managed with the use of PAR in both kinds of systems such as distribution as well as transmission systems [11–14].

3 Phase Angle Regulator The Phase Angle Regulator (PAR) or Phase Shifting Transformer (PST) is a power full device in maintaining the congested management and power flow control in electrical systems. Figure 4 is illustrating the Single line diagram of the test system with PAR, it comprised of a generating source, feeder connected to the PAR, which feeds the load.

Power Flow and Stability Improvement in Distribution Systems

109

Fig. 3. System single line diagram without congested management

Figure 5 depicts the equivalent circuit of the system without PAR and Fig. 2 shows the corresponding power angle curve without compensator. Figure 6 illustrating the Phase Angle Regulator (PAR) (a) equivalent circuit (b) Power circuit (c) Phasor diagram. Figure 7 depicts the Phase Angle Regulator (PAR) (a) Single line diagram (b) Phasor diagram (c) Power angle curve. The power transfer without PAR is expressed in the Eq. (1) and maximum allowable power without PAR in a system as expressed in Eq. (2) respectively. The Eq. (3) showing the equivalent bus voltage with PAR in terms of original bus voltage and equivalent injected voltage of PAR, corresponding Eqs. (4), (6) shows the improved power with its expressions and (5) reactive power of the system. Figure 8 illustrates the phasor diagram and power angle characteristics with different phase angles and Fig. 9 depicts the transient stability improvement with PAR [1–5]. V 1eff = V 1 + V σ V 1V 2 sin(δ ± σ ) X 12

(4)

V 1V 2 [1 − cos(δ ± σ )] X 12

(5)

V 1V 2 Vσ [sin δ + cos(δ) X 12 V2

(6)

P 12 = Q12 = P 12 =

(3)

110

V. Yarlagadda et al.

Fig. 4. Single line diagram of the test system with PAR

Fig. 5. Equivalent circuit of the system without PAR

Fig. 6. Phase Angle Regulator (PAR) (a) Equivalent circuit (b) Power circuit (c) Phasor diagram

Power Flow and Stability Improvement in Distribution Systems

111

Fig. 7. Phase Angle Regulator (PAR) (a) Single line diagram (b) Phasor diagram (c) Power angle curve.

Fig. 8. Phasor diagram and power angle characteristics with different phase angles

Fig. 9. Transient stability improvement with PAR

112

V. Yarlagadda et al.

4 Case Study and Results The test system is developed in the Matlab Simulink environment as Fig. 10 illustrates the Simulink Model of PAR including closed loop control circuit being simulated and results have been presented herewith. Table 1 depicts the Power Transfer in 230 V Domestic System without and with PAR simulation results. The results of domestic system with variation in phase angle of PAR showing the credibility of the PAR on power flow control and stability improvement. 4.1 The Test Results of 230 V Domestic System The test results of 230 V Domestic system being presented with, Table 2 depicts the Power Transfer in 230 V Domestic System without PAR experimental results, and Table 3 illustrates the Power Transfer in 230 V Domestic System without PAR experimental results for small loads. Table 4 spectacles the Power Transfer in 230 V Domestic System without PAR experimental results for medium loads and Table 5 prospects the Power Transfer in 230 V Domestic System without PAR experimental results for large loads. Figure 12 shows the 230 V Practical System with PAR with angle variation for small loads, Fig. 13 illustrates the 230 V Practical System with PAR with angle variation for medium loads and Fig. 14 encapsulates the 230 V Practical System with PAR with angle variation for heavy loads (Fig. 11).

Fig. 10. Simulink Model of PAR including closed loop control circuit

Power Flow and Stability Improvement in Distribution Systems

113

Table 1. Power Transfer in 230 V Domestic System without and with PAR simulation results S. No.

Power Transfer without PAR P in W

Power Transfer with PAR P in W

1

530

1340

Table 2. Power Transfer in 230 V Domestic System without PAR experimental results S. No. Load

Voltage in V

Current in A

Power in W

1

228

3.8

750

2

225

6.7

1125

3

224

8.3

1200

Table 3. Power Transfer in 230 V Domestic System without PAR experimental results for small loads S. No. Phase Angle

Voltage in V

Current in A

Power in W

1

196

1

368

2

198

1

380

3

200

1

400

Table 4. Power Transfer in 230 V Domestic System without PAR experimental results for medium loads S. No. Phase Angle

Voltage in V

Current in A

Power in W

1

182

1.38

468

2

186

1.4

488

3

188

1.42

496

Table 5. Power Transfer in 230 V Domestic System without PAR experimental results for large loads S. No. Phase Angle

Voltage in V

Current in A

Power in W

1

170

1.6

520

2

174

1.68

552

3

176

1.7

560

114

V. Yarlagadda et al.

P in W 1500 1000 500 0 without PAR P in W

with PAR P in W

Fig. 11. Power flow without and with Phase Angle Regulator in domestic system

P in W 400 390 380 370 360 350 Phase Angle 1

Phase Angle 1

Phase Angle 1

Fig. 12. 230 V Practical System with PAR with angle variation for small loads

Power Flow and Stability Improvement in Distribution Systems

115

P in W 500 490 480 470 460 450 Phase Angle 1

Phase Angle 1

Phase Angle 1

Fig. 13. 230 V Practical System with PAR with angle variation for medium loads

P in W

560 540 520 500 Phase Angle 1

Phase Angle 1

Phase Angle 1

Fig. 14. 230 V Practical System with PAR with angle variation for heavy loads

4.2 The Test Results of 415 V Commercial System The test results of 415 V Commercial system have been presented herewith as Fig. 15 shows the Power transfer through the feeder in commercial system for load 1 with different phase angles, Fig. 16 illustrates the Power transfer through the feeder in commercial system for load 2 with different phase angles. Table 6 indicating the Power Transfer in 415 V Commercial System Results with PAR, Fig. 15 prospects the Power transfer through the feeder in commercial system for load 1 with different phase angles, Fig. 16 encapsulates Power transfer through the feeder in commercial system for load 2 with different phase angles and Fig. 17 illustrating the Power transfer through the feeder in commercial system for load 3 with different phase angles. All these results proves the

116

V. Yarlagadda et al.

effectiveness of PAR in power flow control and stability enhancement of the commercial system with power angle control. Table 6. Power Transfer in 415 V Commercial System Results with PAR S. No. Phase Angle

Power Transfer without Power Transfer with PAR P in KW PAR angle1 P in KW

Power Transfer with PAR angle2 P in KW

1

0.099

0.202

0.231

2

1.104

2.066

2.146

3

10.83

14.2

–

P in W for load 1 250 200 150 100 50 0 without PAR P in with PAR P in W W with phase shi of 40deg.

with PAR P in W with phase shi of 80deg.

Fig. 15. Power transfer through the feeder in commercial system for load 1 with different phase angles

4.3 The Test Results of 11 KV Distribution System The test results of 11 KV Distribution system have been presented herewith as Table 7 shows the Power Transfer in 11 KV Distribution System results with PAR, Fig. 18 illustrates the Power transfer through the feeder in distribution system for load 1 with different phase angles. Figure 19 prospects the Power transfer through the feeder in distribution system for load 2 with different phase angles and Fig. 20 encapsulates the Power transfer through the feeder in distribution system for load 3, all these results have been indicating that the power angle control made with PAR is most effective in improving the Power Transfer and system stability improvement for all kinds of systems such as domestic, commercial as well as distribution systems.

Power Flow and Stability Improvement in Distribution Systems

117

P in KW for Load 2 2.5 2 1.5 1 0.5 0 without PAR P in with PAR P in KW with PAR P in KW KW with phase shi of with phase shi of 40deg. 80deg. Fig. 16. Power transfer through the feeder in commercial system for load 2 with different phase angles

P in KW for load 3 15 10 5 0 without PAR P in KW with PAR P in KW with phase shi of 40deg. Fig. 17. Power transfer through the feeder in commercial system for load 3 with different phase angles

Table 7. Power Transfer in 11 KV Distribution System Results with PAR S. No. Phase Angle

Power Transfer without Power Transfer with PAR P in KW PAR angle1 P in KW

Power Transfer with PAR angle2 P in KW

1

70.8

143.3

162.2

2

863.2

1480

1530

3

7630

10060

–

118

V. Yarlagadda et al.

P in KW for Load 1 180 160 140 120 100 80 60 40 20 0 without PAR P in with PAR P in KW with PAR P in KW KW with phase shi of with phase shi of 40deg. 80deg. Fig. 18. Power transfer through the feeder in distribution system for load 1 with different phase angles

P in KW for load 2 2000 1500 1000 500 0 without PAR P with PAR P in with PAR P in in KW KW with phase KW with phase shi of 40deg. shi of 80deg. Fig. 19. Power transfer through the feeder in distribution system for load 2 with different phase angles

Power Flow and Stability Improvement in Distribution Systems

119

P in MW for Load 3

12 10 8 6 4 2 0 without PAR P in KW

with PAR P in KW with phase shi of 40deg.

Fig. 20. Power transfer through the feeder in distribution system for load 3

5 Conclusions The power and control circuits of the PAR has been developed in Maltlab Simulink environment and has been simulated with domestic, commercial and distribution systems with phase angle control and results have been presented in the article. The control logic has been developed to improve the power flow with automatic switching action of PAR for the desired power flow in a feeder. Finally laboratory based testing have been performed without and with PAR with different phase angles and loads such as small, medium and heavy loads and results shows the effectiveness of PAR in improving power transfer capability and stability. The test results of all three systems such as domestic, commercial and distribution systems have been presented in the article. The simulation and experimental test results of domestic system with different power angles proves the effectiveness of the PAR on power flow and system stability. The simulation results of both commercial and industrial systems shows the significant improvement in power transfer capability and system stability with the control of power angle.

References 1. Prasai, A., Kandula, R.P., Moghe, R., Heidel, T., Schauder, C., Divan, D.: Compact dynamic phase angle regulator for power flow control. In: 2015 IEEE Energy Conversion Congress and Exposition (ECCE), pp. 4985–4992 (2015). https://doi.org/10.1109/ECCE.2015.7310363 2. Ramamoorty, M., Toma, L.: Phase shifting transformer: mechanical and static devices. In: Eremia, M., Liu, C.-C., Edris, A.-A. (eds.) Advanced Solutions in Power Systems: HVDC, FACTS, and Artificial Intelligence, pp. 409–458. IEEE (2016). https://doi.org/10.1002/978 1119175391.ch7

120

V. Yarlagadda et al.

3. Peterson, N.M., Meyer, W.S.: Automatic adjustment of transformer and phase-shifter taps in the newton power flow. IEEE Trans. Power Apparatus Syst. PAS-90(1), 103–108 (1971). https://doi.org/10.1109/TPAS.1971.292904 4. IEEE Guide for the Application, Specification, and Testing of Phase-Shifting Transformers – Redline. IEEE Std C57.135-2011 (Revision of IEEE Std C57.135-2001) - Redline, pp. 1–71, 19 August 2011 5. Zhang, W., et al.: Conceptual design of the power supply for magnetic island divertor configuration on J-TEXT. IEEE Trans. Plasma Sci. 48(6), 1670–1675 (2020) 6. Galantino, S., Fulaza, E., Nkweendenda, A.: Engineering of power flow control across the Zambia – Zimbabwe interconnector with phase-shifting transformers. In: PES/IAS PowerAfrica 2020, pp. 1–5. IEEE (2020) 7. Sun, K., Xiao, H., Pan, J., Liu, Y.: VSC-HVDC interties for urban power grid enhancement. IEEE Trans. Power Syst. 36(5), 4745–4753 (2021) 8. Siddiqui, A.S., Khan, S., Ahsan, S., Khan, M.I., Annamalai: Application of phase shifting transformer in Indian Network. In: 2012 International Conference on Green Technologies (ICGT), pp. 186–191 (2012). https://doi.org/10.1109/ICGT.2012.6477970 9. Acha, E., Ambriz-Perez, H., Fuerte-Esquivel, C.R.: Advanced transformer control modeling in an optimal power flow using Newton’s method. IEEE Trans. Power Syst. 15(1), 290–298 (2000) 10. Mahavishnu, K.B.P., Kumar, P., Surjith, H.K.: New approaches to solve radial distribution system problem with FACTS controller. In: International Conference on Electrical Electronics and Optimization Techniques (ICEEOT), pp. 4683–4690 (2016) 11. Shaw, R.N., et al.: Review and analysis of photovoltaic arrays with different configuration system in partial shadowing condition. Int. J. Adv. Sci. Technol. 29(9s), 2945–2956 (2020) 12. Xiao, Y., Song, Y.H., Sun, Y.Z.: Power injection method and linear programming for FACTS control. In: Power Engineering Society Winter Meeting 2000, vol. 2, pp. 877–884. IEEE (2000) 13. Jardim, J.L., Neto, C.S., Kwasnicki, W.T.: Design features of a dynamic security assessment system. In: IEEE PES Power Systems Conference and Exposition 2004, vol. 1, pp. 446–452 (2004) 14. Wang, X., Louie, K.-W., Wilson, P., Liu, W.Z.: Fast decoupled power flow solution with automatic adjustment of generator remote voltage control. In: 2005 IEEE/PES Transmission and Distribution Conference and Exhibition: Asia and Pacific, pp. 1–6 (2005)

Optimization and Comparison of High Performance and Low Power NOR Gate Circuit Using Hybrid Model of Dynamic Voltage Scaling and MTCMOS Technique Kamlesh Kumar1 , Mohit Dahiya2(B) , Manoj Kumar3 , and Priyanshu Lakra4 1 Ministry of Electronics and Information Technology (MeitY), Government of India,

New Delhi, India [email protected] 2 CDAC-T, posted at Ministry of Electronics and Information Technology (MeitY), New Delhi, India [email protected] 3 IIT Hyderabad, posted at Ministry of Electronics and Information Technology (MeitY), Hyderabad, India 4 University School of Information and Communication Technology (USICT), GGSIPU, New Delhi, India

Abstract. The strategy joins VS (Voltage Scaling) and MTCMOS procedure that aids in lessening active and passive power dissemination separately deprived of corrupting the circuit’s execution. The anticipated procedure set aside power dispersal by 35% to 85% when contrasted with regular CMOS and other existing procedures and the numbers of transistors is reduced in existing circuit to reduce the overall energy consumption as well as the reduced transistor logic is area efficient and comparison is done with existing design and NMOS structure. A 2-terminal input NOR gate is executed utilizing the VS-MTCMOS procedure in sub-edge district throughout various ranges of temperature at different voltage level. Electronic Design Automation Tool is utilized in the direction of reproduce the planned circuit. As convenience of electronic frameworks requires longer battery life, it is important that they should have instruments in spot to diminish the force utilization. One of the methods used to build power productivity at the framework level is Dynamic Voltage and Frequency Scaling (DVFS). CMOS rationale is broadly utilized in VLSI circuits yet because of scrambling of innovation, limit voltage of the semiconductors utilized in CMOS circuits decline, results in increment in spillage power. Active power utilization, that is relative to source potential difference (VDD)2 auxiliary combined to general power dispersal. This outcomes in short battery lifespan of cell phones. Transitory is a clever strategy in the direction of shorten mutually unique power dissemination furthermore, spillage power. Keywords: VS- MTCMOS · NMOS · CMOS · NOR · Power · Efficient · DVFS

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 121–131, 2022. https://doi.org/10.1007/978-981-19-1742-4_9

122

K. Kumar et al.

1 Introduction Since past years area, cost, unwavering quality and execution were viewed as the essential worries intended for Very Large-Scale Integration originators despite the fact power utilization is as an optional apprehension [6, 8]. Be that as it may, with the presentation and consistently expanding request of portable electronic gadgets and other remote correspondence frameworks like personal digital assistant and individual correspondents, power utilization be located currently specified equivalent significance cutting-edge contrast with region and execution [1]. It involves power productive Very Large-Scale Integration circuits. Sub-limit rationale circuits have been existing as of late for applications that require super low force utilization [2, 3]. In designed circuits necessitate that the occupied potential difference VDD ought to be not exactly the limit voltages of the semiconductors that are available in the VLSI circuit. The general power feasting utilization i.e., decremented for mentioned circuits. We defined two major sources of power intemperance in any CMOS designed circuit. Primary source is the intemperance of active energy through the exchange of circular motion. In this case, the power of the parasite is amplified and released with a focus between the two levels of sensitivity. Due to current trends through our semiconductors, some of the disturbances of today’s scenario after the transformation of electrical energy and heat have now been published [4]. The expansion of this dynamic power corresponds to a doubling of VDD voltage, operating capacity as well as frequency that is provided by PDyn = CV2DD f. Another power loss source occurs when input is given to the logic gate and output of the logic gate remains unchanged [9]. This also known as static energy dissipation and is caused by scattering of energy transitions, side leaks and door leaks. A subcutaneous fissure indicates the rate of flow from the canal to the eye when the semicircle is closed. This leakage increases significantly when we lower the power limit and calculate the device [7]. The sub-threshold leakage relies upon the boundaries of. This power loss is relational to component supply voltage, the voltage changes of the node, and the average power switching cycle. The voltage change is in maximum circumstances equivalent to the supply voltage, as transition, discharge, is usually different from the square of the supply voltage [5]. In this present study, the technology reduces both energy loss and energy loss by combining voltage technology and MTCMOS technology. In this work the intermediate network of push pull network is eliminated by AOI network in MTCMOS technique and existing structure is optimized MTCMOS NOR circuit with AOI network and compared with existing structure and NOR NMOS logic with MTCMOS.

2 Execution and Simulation of 3T NOR Gate Circuit with MTCMOS The MTCMOS system is one of the strategies that prevents power failure. The MTCMOS system has two ways of action, a dynamic mode and a backup mode. Typically, an ordinary CMOS circuit merely connects a single-phase transistor (VT), although the MTCMOS strategy consists of two limiting power transistors. High VT transistors are known as breaker transistors which help in stopping the current flow and uses low VT transistors in legal circuits to aid in the starting of circuit. In this circuit energy efficient

Optimization and Comparison of High Performance and Low Power

123

an area efficient 2 Input NOR gate is executed with MTCMOS logic. Simulated and executed result is as follows (Figs. 1 and 2):

Fig. 1. Schematic of 3T NOR gate with MTCMOS technique

Fig. 2. Output waveform of 3T NOR gate with MTCMOS technique

124

K. Kumar et al.

3 Execution and Simulation of Existing Circuit with VS-MTCMOS Pull Down and Pull Up Network of NOR Gate The given circuit activates in two ways: dynamic (active) mode and passive mode. During dynamic (dynamic) operation, the electric field is reduced by voltage sharing and sharing process to do justice to the active power loss. The high sleep transistors VT H1 and H2 can use logic of 0 and 1 TBAR of T, which allows them to function as standard transistors. When S is high circuit becomes ordinary CMOS circuit, also turns on to M2, gives out low voltage level of VP. It also switches to M3 and gets full circuit performance. Related products are collected. As S decreases, M1 and M3 decrease by reducing the level of voltage of VS. At that time, the power of Ca and CB share the charge by self-regulation, that will reduce the level of voltage of Vp. When this happens, it will make M3 on or active, which will increase the level of voltage of VS and aid to reduce the massive VS types. The active loss of power is then minimized using the Danab ratio method. In a regression model, in which the input(s) given in the input terminal of logic gate does not change its variable output, the upper semiconductors VT H2 and H1 inside the circuit loop are cleared on giving T = 1 and TBAR = 0. That will lead to very high level of resistance among two organizations. These lines are delayed due to discrimination of power drop due to the introduction of MTCMOS system of high transistors VT H1 and H2. In the current control loop, the opposite side predicts the power loss of the active power supply and storage system. In this way, generally speaking force is diminished by joining these two methods. This technique is used with pull down and pull up network of NOR gate and output are observed (Figs. 3, 4 and 5).

Fig. 3. NOR gate by pull down & pull up network with VS-MTCMOS technique

Optimization and Comparison of High Performance and Low Power

125

Fig. 4. Output waveform of NOR gate by pull down & pull up network with VS-MTCMOS technique at various temperature points

Fig. 5. Output waveform of NOR gate by pull down & pull up network with VS-MTCMOS technique

4 Study and Design of AOI CMOS Logic 3T NOR Gate with VS-MTCMOS In this circuit the existing VS-MTCMOS technique is used with AOI network instead of pull down and pull up network of NOR gate which reduce the ripples and noise in

126

K. Kumar et al.

the output waveform of circuit as well as reduce the power consumption of the circuit and give out energy efficient and smooth output of NOR gate. The simulation result of AOI or network nor gate is given below and if compared with the output of pull up and pull-down network giver the better waveform with less ripples and deviations and uses less number of transistors that give out area efficient and power efficient output (Figs. 6, 7, 8 and Table 1).

Fig. 6. 3T NOR gate with VS-MTCMOS technique

Fig. 7. Output waveform of 3T NOR gate with VS-MTCMOS technique at various temperature points

Optimization and Comparison of High Performance and Low Power

127

Fig. 8. Output waveform of 3T NOR gate with VS-MTCMOS technique

Table 1. Power consumption of 3T NOR gate by CMOS with VS-MTCMOS technique Modified design of 2-Input NOR gate using hybrid voltage scaling and MTCMOS in CMOS logic VDD (V)

Average power consumed

Max power consumed

Min power consumed

0.6

1.779E−007

3.057E−005

0.0032E+000

0.5

1.046E−007

2.467E−005

0.0022E+000

0.4

6.0297E−007

0.273E004

0.0020E+000

0.3

4.005E−007

0.010E−004

0.0028E+000

5 Study and Design of VS-MTCMOS NOR Gate with AOI NMOS Logic In this circuit VS-MTCMOS is used on 2 input NOR gate using NMOS and the analysis of output in terms of power consumption output deviation is noted and compared with other circuits which used CMOS AOI logic network and CMOS pull down and pull up network (Figs. 9, 10, 11 and Table 2).

128

K. Kumar et al.

Fig. 9. The 2T NOR gate by NMOS logic with VS-MTCMOS technique

Fig. 10. Output waveform of 2T NOR gate by NMOS logic with VS-MTCMOS technique at various temperature points

Optimization and Comparison of High Performance and Low Power

129

Fig. 11. Output waveform of 2T NOR gate by NMOS LOGIC with VS-MTCMOS technique

Table 2. Power consumption of 2T NOR gate by NMOS LOGOIC with VS-MTCMOS technique Modified design of 2-input nor gate using hybrid voltage scaling and MTCMOS in NMOS logic VDD (V)

Average power consumed

Max power consumed

Min power consumed

0.6

2.510E−001

1.513E−004

0.001E+000

0.5

1.233E−007

2.411E−004

0.0011E+000

0.4

3.670E−007

3.29E−004

0.001E+000

0.3

2.815E−007

3.202E−004

0.023E+000

6 Comparison of Power Consumption of 3 Different Designed Circuit with Using Hybrid VS-MTCMOS Techniques at Different Voltage Ranges Power consumption of 2 input NOR gate of obtained of Hybrid Voltage Scaling and MTCMOS technique is compared with Standard CMOS, Voltage scaling and simple MTCMOS techniques and keeping the value of VDD slightly above the value of threshold value. Testing is done at different values of the voltage from 0.3 V to 0.6 V (Table 3).

130

K. Kumar et al.

Table 3. Comparison of power consumption of various designed NOR gates with MTCMOS AND VS-MTCMOS technique VDD Standard (V) CMOS [10] [10]

2-terminal input NOR circuit using voltage scaling [10]

2-terminal input NOR circuit using MTCMOS [10]

2-terminal input NOR circuit with hybrid VS-MTCMOS [10]

Modified design of 2-input NOR circuit with hybrid VS and MTCMOS in NMOS logic

Modified design of 2-terminal input NOR circuit with hybrid voltage scaling and MTCMOS in CMOS logic

0.6

7.067E−8 6.460E−8

5.895E−8

2.832E−004

2.510E−001 1.779E−007

0.5

4.588E−8 3.281E−9

3.812E−8

1.060E−9

1.233E−007 1.046E−007

0.4

2.734E−8 1.997E−9

2.247E−8

8.047E−10

3.670E−007 6.0297E−007

0.3

1.405E−8 9.382E−10

1.126E−8

6.451E−10

2.815E−007 4.005E−007

7 Conclusion In this paper, MTCMOS electrical technology and techniques are implemented and implemented in various NOR circuits using AOI CMOS, PULL UP and PULL-DOWN networks. The current system combines voltage measurement and multiple CMOS technology (MTCMOS), which minimizes electrodynamic or static losses without compromising circuit performance. By this technique power consumption can be saved by 35% to 85% as compared to traditional CMOS and other current technologies. In the subfield a variety of 2-door NOR gates were developed using VS-MTCMOS technology and using MTCMOS technology to create the circuits intended to simulate, compared to using Tanner’s EDA equipment was used.

References 1. Pedram, M.: Power minimization in IC design: principles and applications. ACM Trans. Des. Autom. Electron. Syst. 1, 3–56 (1996) 2. Soeleman, H., Roy, K.: Ultra-low power digital subthreshold logic circuits. In: International Symposium on Low Power Electronics and Design, pp. 94–96 (1999) 3. Soleleman,H., Roy, K., Paul, B.: Sub-domino logic: ultra-low power dynamic sub-threshold digital logic. In: 14th International Conference on VLSI Design, pp. 211–214 (2001) 4. Roy, K., Prasad, S.C.: Low-power CMOS VLSI circuit design. Wiley, New Delhi (2009) 5. Roy, K., Mukhopadhyay, S., Mahmoodi-Meimand, H.: Leakage current mechanisms and leakage reduction techniques in deep-submicron CMOS circuits. Proc. IEEE 91(2), 305–327 (2003) 6. Tsai, Y.-T., Huang, H.-H., Hsu, S.-W., Chang, C.-H., Guo, J.-I.: A low-power VDDmanagement technique for high- speed domino circuits. In: International Symposium on VLSI Design, Automation and Test (VLSI-DAT), pp. 1–4 (2011) 7. Arun, P., Ramasamy, S.: A low-power dual threshold voltage-voltage scaling technique for domino logic circuits. In: Proceedings of ICCNT, pp. 1–6 (2012)

Optimization and Comparison of High Performance and Low Power

131

8. Calhoun, B.H., Chandrakasan, A.P.: Ultra-dynamic voltage scaling (UDVS) using subthreshold operation and local voltage dithering. IEEE Trans. Solid State Circ. 41(1), 238–245 (2006) 9. Agarwal,K., Deogan, H., Sylvester, D., Nowka, K.: Power gating with multiple sleep modes. In: 7th International Symposium on Quality Electronic Design, pp. 633–637, March 2006 10. Handa, A., Chawla, J., Sharma, G.: A novel high performance low power CMOS NOR gate using voltage scaling and MTCMOS technique. In: 2014 International Conference on Advances in Computing, Communications and Informatics (ICACCI) (2014)

Design of UWB Antenna with Frequency Interference Mitigation Technique for Wireless Communication Applications Koduri Sreelakshmi(B) and Gottapu Sasibhushana Rao Department of Electronics and Communication Engineering, Andhra University College of Engineering (A), Andhra University, Visakhapatnam, Andhra Pradesh, India [email protected]

Abstract. In this research article, an attempt is made to design a compactcoplanar waveguide (CPW) fed ultra-wideband (UWB) microstrip antenna with acceptable WLAN band-rejection performance. The investigated antenna is made up of a rectangular patch with U- shaped aperture and dualrectangular parasitic stubs that allow it to function in the UWB spectrum. The investigated antenna is forged on low-cost Fire Retardant-4 glass epoxy substrate having dimensions of 24 × 28 × 1 mm3 . The C-slot is incorporated at the center of the rectangular patch of the investigated antenna to realize notch band characteristics. The results shows that the investigated antenna provides UWB operation over 2.38 GHz to 12.49 GHz with sharp frequency notched characteristics to mitigate the interference with the existing licensed frequency spectrum WLAN operating over 5.21 to 5.60 GHz. Experimental and simulated results offer a marginally decent agreement with each other. Keywords: UWB · CPW · WLAN · Slot · Stub

1 Introduction UWB communication systems has become massively popular by virtue of its higher bandwidth, offering several advantages such as lower power consumption as compared to the existing system, higher data rate, ability to co-exist with narrowband systems, covert data transmission, high time-resolution, resistance to interference and most importantly low-cost implementation [1] for a wide range of UWB applications in communication, radar, imaging and positioning systems. Ultra-wideband (UWB) refers to a system that occupies either 500 MHz of the spectrum or that has 10 dB bandwidth beyond 20% of its center frequency. The research on UWB dates back in 1893 [1], when Hertz first demonstrated experiment on UWB however, UWB antenna design gained more significance in February 2002 when Federal Communications Commission (FCC) allotted license-free frequency overlay of 3.1–10.6 GHz for UWB usages [2]. Thus, the 7.5 GHz bandwidth of UWB allows lower power consumption as signal power (Ps) can be traded off with bandwidth (W) for a given noise spectral density (N0 )of a channel having the capacity © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 132–139, 2022. https://doi.org/10.1007/978-981-19-1742-4_10

Design of UWB Antenna with Frequency Interference Mitigation Technique

133

(C) which is evident from Shannon’s channel capacity theorem. With the low power level of −41.3 dBm/MHz, UWB’s interference can be disregarded by many existing systems operating over the same frequency band. These attractive features of UWB technology have lured many researchers in developing and designing UWB antennas suitable for UWB communication which have been reported in [3]. However, the newly assigned frequency spectrum for UWB overlaps with existing licensed frequency spectrums allocated for 5.2/5.8 GHz IEEE 802.11a WLAN and 5.5 GHz HiperLAN applications. These licensed frequency spectrums need to be isolated from UWB’s frequency band resulting in need of design UWB antenna that exhibits frequency notched characteristics. Many techniques were proposed to have wide impedance bandwidth with notched function like by employing EBG structure resemblingmushroom [4], EBG structure identical to that of mushroom loaded with L-slots [5], by inserting L-slit which is inverted at the radiator’s edge [6], by engraving aperture of shape U in the circular patch [7], by employing stub of U-shape [8], by employing quarter wave length resonator of hook shape [9], by utilizing tuning stub resembling fractal geometry [10], by embedding U-shaped aperture [11], by engraving λ/2 annular ring slot of C-shape [12], by employing couple of ground stubs at the ground plane edge [13], by using U-shaped and meandered aperture [14], by attaching the ground plane with couple of two stubs [15].The investigated antenna outperforms the antennas reported in [4–15] in terms of size, cost, substrate thickness, wider impedance bandwidth and simplicity of design. In this research report, a UWB antenna having single frequency notched features had been designed, optimized, and examined, that provides measured VSWR (≤2) bandwidth over 2.38–12.49 GHz frequency band (UWB) except at 5.21–5.60 GHz. To achieve the single notch band characteristics, the C-Slot is incorporated at the center of the rectangular patch of the investigated antenna. The antenna structure presented in this article is optimized using HFSS EM simulation software and tested using Agilent FieldFox VNA N9916A having frequency range 30 kHz to 14 GHz. The article is structured like this: Sect. 2 describes the approach for developing the investigated antenna and parametric analysis. Section 3 displays the results and discussions. Table 2 presents a comparison of the presented antenna with other available models, while Sect. 4 concludes the article.

2 Structure of CPW-Fed Antenna Without and with C-Slot Figure 1 represents the investigated antenna structure. The presented antenna is fabricated on single-sided Fire Retardant/Flame Resistant glass epoxy substrate of thickness 1 mm, dielectric constant εr = 4.4 and tan δ = 0.02. The presented antenna has compact dimension of 24 × 28 mm2 . Table 1 provides dimensions of the presented antenna which are optimized. Initially, the UWB antenna consisting of a rectangular patch with U-shaped aperture with dual rectangular parasitic stubs and fed by a CPW is developed as shown in Fig. 1(a). As a result of this CPW feed and its configuration, the antenna exhibits UWB features. The presented antenna provides simulated UWB operation over 2.36–12.54 GHz with VSWR ≤ 2, as shown in Fig. 2. Furthermore, the radiator is embedded with a C-shaped aperture to reject the WLAN frequency range, as illustrated in Fig. 1(b).The electrical length of C-shaped apertureis calculated by employing the following formula:

134

K. Sreelakshmi and G. S. Rao

Fig. 1. Structure of the investigated antenna a) without C-slot b) with C-slot

Table 1. Investigated antenna dimensions Parameter

Value (mm)

Parameter

Value (mm)

L

28

a

5

W

24

b

5

i

11.075

c

5

j

5

d

3.5

m

0.85

e

4

n

1.65

f

3

g

5

L1

10

L2

3

L3

2.8

t

0.2

LSlot = L1 + 2 × L2 + 2 × L3 = λg =

c √ fn εreff

λg 2

(1) (2)

Where εreff = effective dielectric constant = (εr + 1)/2, c = speed of light and fn = notch frequency. The theoretical value of the Lslot = 16.3 mm at fn = 5.5 GHz while the optimized length of the Lslot = 20.4 mm and it determines the notch frequency. The optimized width of C-shaped aperture is 0.2 mm and it determines the bandwidth of rejected band. The investigated UWB band notch antenna realized an impedance bandwidth of 2.47–12.44 GHz and single notched band centered at 5.6 GHz (WLAN) with the impedance bandwidth 5.14–5.68 GHz as illustrated in Fig. 2.

Design of UWB Antenna with Frequency Interference Mitigation Technique

135

Fig. 2. Simulated VSWR characteristics of investigated antenna

2.1 Parametric Analysis A detailed parametric investigation has been performed in HFSS software to examine the functioning of the investigated antenna when the length of the C-shaped aperture is varied. Figure 3 illustrates the impact of length of slot on the VSWR of the presented antenna. The notch band (WLAN) moved in the direction of the lower frequency, When the length of slot (Lslot ) was increased from 20.4 mm to 21.4 mm. The band-notch frequency was noticed at 6 GHz for Lslot = 21.4 mm and reducing the value of Lslot from 21.4 mm to 19.4 mm, shifted the band -notch frequency from 5.6 GHz to 5.25 GHz because the length of the C-shaped aperture is inversely related to the band notch center frequency as illustrated by Eq. (1) and (2).

Fig. 3. Variations in notched band by varying Lslot length

3 Results and Discussion Furthermore, to acquire additional insights about the operation of the UWB antenna the surface current distribution is observed (as illustrated in Fig. 4). The distribution of current at the lower frequency 3.5 GHz is largely accumulated on the ground planetop edgeand in the lower portion of the patch radiator and feed-line as illustrated in Fig. 4(a).

136

K. Sreelakshmi and G. S. Rao

Hence, radiation characteristics at lower frequency thus can be controlled by varying the feed-line physical dimensions, the separation between feed-line and ground plane and radiating patch respectively. As examined in Fig. 4(b), the current distribution at a notch frequency of 5.5 GHz is centered around C-shaped aperture structure, while it is largely concentrated on the ground plane top edge, rectangular stubs junction location in the radiator lower portion and feed-line at the higher edge frequency 8.5 GHz as illustrated in Fig. 4(c).

Fig.4. Surface current disseminations at (a) 3.5 GHz, (b) 5.5 GHz and (c) 8.5 GHz respectively of the investigated antenna

Figure 5 displays the photograph of fabricated antenna. Figure 6 shows the simulated and measured VSWR performance of the investigated antenna, the presented antenna provides UWB operation over the frequency range 2.38–12.49 GHz with notch centered at 5.5 GHz (5.21–5.60 GHz) to avoid interference with WLAN applications.

Fig. 5. Fabricated prototype of the presented antenna

Fig. 6. Measured VSWR performance of the investigated antenna

Design of UWB Antenna with Frequency Interference Mitigation Technique

137

The radiation characteristics of the fabricated model is studied by testing it in an anechoic chamber as illustrated in Fig. 7. The normalized measured far-field radiation pattern along E-plane is directional (dumbbell-shaped) and H-plane is omni-directional and is in decent agreement with the simulated radiation patterns. The investigated UWB band-reject antenna exhibits an average radiation efficiency of about 90% except at band-notch frequency. The proposed UWB single notch antenna exhibits minimal gain variations of 0.5–5 dBi over the complete UWB frequency band except at band-notch frequency.

Fig. 7. Normalized simulated and measured radiation pattern of the presented UWB antenna in E-plane and H-plane at a) 3.5 GHz, and b) 8.5 GHz.

Table 2 presents the analogy of the performance of the investigated UWB antenna with other existing UWB antennas. It is obviously noticed from the table that the investigated structure has lowcost, smaller size, wider impedance bandwidth and simplicity of design.

138

K. Sreelakshmi and G. S. Rao Table 2. Performance comparison of the proposed research with existing research work

Ref./Publication year

Substrate

Antenna size & volume (dimensions in mm3 )

UWB Operational frequency (GHz)

Bandwidth (%)

Notch-band frequency

Notchedband bandwidth

[3] /2013

RT/Duroid 4003

39 × 45 × 0.794 Vol = 1393

3.1–10.6

109.5

5.5 GHz (WLAN)

5.15–5.94

[4] /2013

FR4

40 × 30 × 1.6 Vol = 1920

2.3–11.4

132.8

5.5 (WLAN)

4.9–6

[6] /2018

FR4

50 × 40 × 1.64 Vol = 3280

3.3–12

113.7

3.4 GHz (Wi-MAX)

3.2–3.8

[10] /2013

FR4

32 × 32.6 × 1.6 Vol = 1669

3.1–9.3

100

5.5 GHz (WLAN)

5.12–5.99

[12] /2011

PTFE

30 × 39.3 × 0.8 Vol = 943

2.57–12

129.4

5.5 GHz (WLAN)

5.15–5.825

Proposed work

FR4

24 × 28 × 1 Vol = 672

2.38–12.49

135.9

5.5 GHz (WLAN)

5.21–5.60

4 Conclusions In this research article, the design, analysis, and optimization of UWB antenna with frequency interference mitigation technique that provides operation over 2.38–12.49 GHz with sharp band-reject characteristics at 5.21–5.60 GHz that allows non-interference frequency operation with WLAN is investigated. The investigated antenna is tested to justify the simulated results. The investigated antenna offers a reasonable minimal gain deviation from 0.5–5.0 dBi, average radiation efficiency of about 90% except at band-notch frequency. The radiation pattern is approximately omnidirectional along H-plane while directional (8-shaped or dumbbell-shaped) along the E-plane. Thus, the compact size, stable radiation characteristics, excellent gain, and the existence of band-notch makes the investigated antenna appropriate for using in wireless communication applications.

References 1. Valderas, D., Sancho, J.I., Puente, D., et al.: Ultrawideband Antennas: Design and Applications. Imperial College Press, London (2011) 2. Federal Communications Commission: First report and order, Revision of part 15 of the commission’s rule regarding ultra-wideband transmission system FCC 02-48 (2002) 3. Ammann, M.J., John, M., Ruvio, G.: Ultra-wideband antennas. In: Chen, Z.N., Liu, D., Nakano, H., Qing, X., Zwick, T. (eds.) Handbook of Antenna Technologies, pp. 1657–1695. Springer, Singapore (2016). https://doi.org/10.1007/978-981-4560-44-3_59 4. Yazdi, M., Komjani, N.: Design of a band-notched UWB monopole antenna by means of an EBG structure. IEEE Antennas Wirel. Propag. Lett. 10, 170–173 (2011). https://doi.org/10. 1109/LAWP.2011.2116150 5. Pandey, G.V., Singh, H.S., Bharti, P.K., Meshram, M.K.: Design of WLAN band notched UWB monopole antenna with stepped geometry using modified EBG structure. Prog. Electr. Res. B 50, 201–217 (2013).https://doi.org/10.2528/PIERB13030101

Design of UWB Antenna with Frequency Interference Mitigation Technique

139

6. Banerjee, A., et al.: Construction of effective wireless sensor network for smart communication using modified ant colony optimization technique. In: Bianchini, M., Piuri, V., Das, S., Shaw, R.N. (eds.) Advanced Computing and Intelligent Technologies. LNNS, vol. 218, pp. 269–278. Springer, Singapore (2022). https://doi.org/10.1007/978-981-16-2164-2_22 7. Mohamed, D., Zoubir, M.: Single notched band characteristics UWB antenna using cylindrical dielectric resonator and U-shaped slot. J. Microw. Optoelect. Electr. Appl. 17(3), 340–351 (2018). https://doi.org/10.1590/2179-10742018v17i31190 8. Ibrahim, A., et al.: UWB monopole antenna with band-notched characteristics mitigating interference with WiMAX. Radio Eng. 26(2), 438–443 (2017). https://doi.org/10.13164/re. 2017.0438 9. Sanyal, R., et al.: Miniaturized band-notched UWB antenna with improved fidelity factor and pattern stability. Radio Eng. 27(1), 39–46 (2018). https://doi.org/10.13164/re.2018.0039 10. Lui, W.J., Cheng, C.H., Cheng, Y. Frequency notched ultra-wideband microstrip slot antenna with a fractal tuning stub. Electr. Lett. 41(6), 294–296 (2005). https://doi.org/10.1049/EL: 20058420 11. Kang, X., et al.: A Band-notched uwb printed elliptical-ring monopole antenna. Prog. Electrom. Res. C 35, 23–33 (2013). https://doi.org/10.2528/PIERC12082818 12. Avez, S., Aldhaheri, R.W.: A very compact and low profile UWB planar antenna with WLAN band rejection. Sci. World J. (2016). https://doi.org/10.1155/2016/3560938 13. Weng, Y.F., Cheung, S.W., Yuk, T.I. Compact ultra-wideband antennas with single bandnotched characteristic using simple ground stubs. Microw. Opt. Technol. Lett. 53(3) (2011). https://doi.org/10.1002/mop.25817 14. Sohail, A., Alimgeer, K.S., Iftikhar, A., Ijaz, B, Kim, K.W., Mohyuddin, W.: Dual notch band UWB antenna with improved notch characteristics. Microw. Opt. Technol. Lett. 60(4) (2018). https://doi.org/10.1002/mop.31071 15. Singh, H.S., Kalraiya, S.: Design and analysis of a compact WiMAX and WLAN band notched planar monopole antenna for UWB and bluetooth applications. Int. J. RF Microw. Comput. Aided Eng. 28, e21432 (2018). https://doi.org/10.1002/mmce.21432

Design of Power-Efficient Operational Transconductance Amplifier in the Application of Low Pass Filter Using 180 nm CMOS Technology Nupur Mittal(B)

, Imran Ullah Khan , and Piyush Charan

Department of Electronics and Communication Engineering, Integral University, Lucknow, India {mittal,iukhan,piyush}@iul.ac.in

Abstract. In modern transreceivers, analog base band section is very crucial which deals with channel selectivity, antialiasing and dynamic range. Nowadays OTA become a basic building block of any analog system.For better performance of RF front end a filter which is used in base band section must include many characteristics like high linearity, tunable BW, low noise etc. A second order low pass Gm-C Filter is implemented.The core of this filter is power efficient OTA. The OTA is implemented to operate at a ±1.0 V supply voltage with a power consumption of 0.42 mW. All simulations has been performed using Tanner EDA tool using CMOS technology with parameters TSMC 0.18 μm. The simulation results of this circuit show that it has a high DC gain of 76 dB and a transconductance of 360 μS. Keywords: CMOS · OTA · LPF · DC gain and CMRR

1 Introduction Rapidly growing, mobile and wireless communication market is highly dependant on the receiver architecture for analog baseband signal processing, there fore high frequency and low power, fully integrated filters received considerable worldwide attention. In today’s Scenerio multi-standard transceivers are so much in demand and direct conversion architectures are best suited forthat purpose [1]. The RF front end module of the receiver includes a variable gain amplifier and a LPF. The range of variable gain for UWB analog front ends should be small. Therefore, the realization of voltage gain amplifier is not important. Therefore, the low-pass filter design has become the most crucial design in the analog front end [2, 3]. Analog filter based on operational transconductance amplifier (OTA) Capacitor (so-called OTA-C or gm-C) filters have attracted the attention in recent research. In comparison of traditional active RC filter, OTA-C filter has better performance. OTA-C filters offer simple in design, high-frequency capability, Electronic adjustability, suitable for monolithic integration, reducing the number of components and the potential of design automation. The low-pass filter based on Gm-C © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 140–151, 2022. https://doi.org/10.1007/978-981-19-1742-4_11

Design of Power-Efficient Operational Transconductance Amplifier

141

is the only choice when the cutoff frequency of the filter reaches in the range of GHz [4– 10]. OTA-C filters also have great limitations. Good linearity of filters needed a highly linearized OTA. The increase in the linear range will inevitably reduce the usable range of transconductance of the OTA at a given power supply voltage, and increase the OTA noise. At present, great importance is attached to the implementation of integrated filters using sub-micron CMOS technology. CMOS technology provides favorable features for high-density integration, such as low power consumption, Power supply voltage and low static power consumption of digital circuits.

2 Basics of Filters The function of an ideal low-pass filter is that, it will pass the signals below the cut off and eliminate the signals above the cut off frequency. When any filter is designed for a particular application various trade-offs should be considered between the parameters. So many types of filters provide the trade off between attenuation and phase response which will depend on the application. Butterworth filters provides the good trade off between attenuation and phase response. Some times butterworthe filter is known as maximally flat filter [11]. The realization of a second-order low-pass Butterworth filter is made with the following transfer function H =  2 f fc

K + 1.414 fjfc

(1) +1

Butterworth filter has zero ripples in the pass band and stop band. Frequency domain response and time domain response plays a important role while designing a filter for a particular application which further decides the complexity and cost of the filter. In all pole filters category two filters are generally used Butterworth and Bessel [13–15]. Butterworth filter gives good amplitude with fair trasient behaviour. Chebyshev filters give smaller transition region in comparison of butterworth. For higher order filters butterworth filter is preferred. In designing higher order filter multiple sections of first/second order filters are used. The characteristics of these sections should be aligned for good response of a higher order filter. 2.1 Basics of Operational Transconductance Amplifier Due to the feature of tunability of transconductance with bias current operational transconductance amplifier become very useful in the realization of active filters. OTA provides linear electronic adjustability. OTA circuits consume less power and can be operated at lower voltage. At present high frequency, high linearity, and low power are the important parameters of CMOS OTAs [16]. Filters designed using OTA do not need resistors that are why they are easier to fabricate. Filter performance directly depends on the characteristics of the OTA (Figs. 1 and 2). OTA: I = gm (V+ − V− )

(2)

142

N. Mittal et al.

-

V-

I +

V+

Fig. 1. Basic OTA

-

V-

+

+

V+

-

I+ I-

Fig. 2. Balanced OTA

Balanced OTA: I+ = gm (V+ − V− )

(3)

I− = gm (V+ − V− )

(4)

gm =

 β IR = β+2 

IBIAS

IBIAS 2VT

(5) (6)

3 Results and Discussion 3.1 Conventional OTA With three high performance current mirrors and using a differential pair (M1 , M2 ) the conventional OTA is achieved. This OTA is working on low operating voltage. By using a super cascode transistor high slew rate, a large bandwidth and a high open loop gain have been achieved [17]. Both the transistors M1 and M2 are working in the saturated region. So, the differential input voltage is as follows   I1 I2 − (7) Vid = Vin+ − Vin− = VGS1 − VGS2 = βn βn

Design of Power-Efficient Operational Transconductance Amplifier

143

Expressions of current Iout is as follows  1 2βn Iss × Vid − 2βn Iss (2βn Iss ) × V3id 4  Where βn = μn Cox W L 1,2 and Iss is the bias current and Gm is given by Iout ∼ =

gm =

 2βn Iss

(8)

(9)

Fig. 3. Simulated conventional OTA on tanner EDA

Proposed OTA By using stage attenuation technique linearity can be better than the conventional OTA (Fig. 3). MOS transistor pair (M1a, M2a) forms attenuator and (M3, M4) transistors of the PMOS type will act as load [18, 19]. Both M1a and M2a are operated in the saturated region. So the differential input voltage is defined by   I1a I2a − (10) Vida = Vina+ − Vina− = VGSa1 − VGSa2 = βn βn Output current is as follows 1 1 Iout ∼ 2βn Issa × Vida + 2βn Issa (βn Issa ) × V3ida = Issa /2 − 2 8

(11)

144

N. Mittal et al.

The differential gate voltage of the load transistors is given below: VidA Vg = √ m

(12)

Where m = attenuation factor βp βn

 W βp = μp Cox L 3,4 m=

(13)

Transconductance is given by gma =

√ 2βn Iss √ m

(14)

Fig. 4. Simulated conventional OTA on tanner EDA

The simulation results of the proposed CMOS OTA and Conventional OTA has been done by using Tanner EDA tool using TSMC 180 nm technology (Fig. 4 and Table 1). This circuit is working with ±1.0 V supply voltage with a dc voltage VB1 = 0.1 V and the value of capacitive load is 10 pF. The DC transfer characteristic of the proposed OTA is shown in Fig. 5, the OTA (proposed) provides good linearity in the range [−0.6 V, 0.6 V]. The trascondcutance Gm of the OTA proposed is plotted and it can be seen that it is more stable in the interval of [−0.6 V, 0.6 V] and the maximum value of gm is seen to be equal to 360 μS (Fig. 6) (Table 2).

Design of Power-Efficient Operational Transconductance Amplifier

145

Table 1. DC transfer characteristics S. no.

Differential input voltage

Output current Conventional OTA

Proposed OTA

1

−1

−25

−210

2

−0.8

−25

−200

3

−0.6

−25

−150

4

−0.4

−25

−100

5

−0.2

−13

−50

6

0

0

0

7

0.2

65

60

8

0.4

110

120

9

0.6

125

180

10

0.8

125

200

11

1

125

200

Fig. 5. DC transfer characteristics

The frequency response of the conventional and proposed OTA is shown in Fig. 7. The open loop gain of proposed OTA 76 dB. For OTA conventional, a less gain of 44 dB is acheived. In Table 3, the performance parameters of both OTA (conventional and proposed) along with some of the works are compared (Table 4).

146

N. Mittal et al. Table 2. Simulated transconductance S. no.

Differential input voltage

Transconductance Conventional OTA

Proposed OTA

1

−1.0

0

0

2

−0.8

0

25

3

−0.6

0

125

4

−0.4

18

360

5

−0.2

270

322

6

0.0

300

320

7

0.2

270

300

8

0.4

150

350

9

0.6

25

340

10

0.8

24

175

11

1.0

13

10

Fig. 6. Transconductance Table 3. Performance comparison with previously reported work Performance parameters

Conventional OTA

Proposed OTA

OTA (18)

OTA (20)

Technology CMOS (μm)

180 nm

180 nm

180 nm

180 nm

Supply voltage (V)

±0.8

±1.0

1.8

0.9

Power consumption (mW)

0.39

0.42

0.45

0.0588

Transconductance (μS)

320

360

110

38.8

DC gain (dB)

44

76

–

34.8

Linear range (V)

–

±0.6

±0.5

±0.6

CMRR (dc) (dB)

16.04

88.53

–

139.8

Design of Power-Efficient Operational Transconductance Amplifier

147

Table 4. Open loop frequency response S. no.

Frequency

Open loop frequency response (dB) Conventional OTA

Proposed OTA

47

76

1

10

2

100

47

76

3

1000

47

76

4

10,000

47

70

5

100000

33

54

6

1000000

12

36

7

1E7

−4

14

80 70 60 50 40 30 20 10 0 -10

Gain in dB

Conventional OTA

Frequecy in Hz Fig. 7. Open loop frequency response of OTA

4 Application 4.1 Theoretical Analysis Range of transconductance can be increased for the filter applications. For analog base band circuit low pass filter plays important role. Io = gm (Vin − V− )

(15)

V− = Vout

(16)

On putting value of V− from Eq. (16) into Eq. (15) Io = gm (Vin − Vout )

(17)

148

N. Mittal et al.

On applying KVL Io = sCVout

(18)

sCVout = gm (Vin − Vout )

(19)

Vout gm = Vin sC + gm

(20)

Equating (17), (18) and then solving

g

m Vout =  C gm Vin s+ C

(21)

Comparing with standard transfer function, It is Low Pass Filter. Second order butterworth low pass filter is shown in Fig. 8. Transfer fuction of this filter is given in Eq. (29).This circuit is designed using two proposed OTA.

IBias g gm1 m1 Vin

IBias -

Io1 Vo1

gm2 gm2 V

+

Io Vo

+

C1 C2

GND GND

Fig. 8. Second order LPF

Io1 = gm1 (Vin − Vo )

(22)

Io1 = V01 sC1

(23)

Io = gm2 (Vo1 − Vo )

(24)

Io = Vo sC2

(25)

  Io = gm2 Io1 sC1 − Vo

(26)



 Vo sC2 = gm2 gm1 Vin − Vo sC1 − Vo

(27)

Design of Power-Efficient Operational Transconductance Amplifier

149

Vo sC2 + Vo gm2 = gm1 gm2 sC1 (Vin − Vo )

(28)

 gm1 gm2 sC1 Vo

= Vin sC2 + gm2 + gm1 gm2 sC1

(29)

Simulation Results The proposed filter operates at a low supply voltage of ±1.0 V. Power consumption of this filter is 0.8 mW (Table 5 and Fig. 9). Table 5. Frequency response of LPF S. no.

Frequency (GHz)

Voltage gain (dB)

1

1

0

2

5

0

3

10

0

4

50

0

5

60

0

6

80

−3

7

100

−8

8

200

−13

9

300

−16

10

500

−24

11

1000

−24

12

2000

−24

Frequency vs Voltage Gain(dB) 0 -10 Voltage Gain(dB) -20 -30 0

10

80

200

2000 Frequency (GHz)

Fig. 9. Frequency response of low pass filter

150

N. Mittal et al.

5 Conclusion In this first a low voltage and low power OTA was implemented using CMOS technology using TSMC 0.18 nm parameters. Which improves the linearity of operational transconductance amplifier. In this OTA signal attenuation technique is used. This topology acheives a good differential input range ±0.6 V with a gain of 76 dB. Based on this circuit, a second order voltage mode low pass filter has been implemented and a power consumption 0.7 mW has been obtained. Acknowledgements. This work is an intellectual property of Integral University vides the Manuscript Communication no. IU/R & D/2021-MCN0001257. We would like to acknowledge the Integral University, Lucknow, India for providing an opportunity to carry out this research work.

References 1. Zhou, M., Wang, K.: 0.18 mW/pole inverter-based Gm C bandpass filter with automatic frequency tuning. Electron. Lett. 54(15), 943–945 (2018) 2. Rezaei, F., Azhari, S.J.: A new controllable adaptive biasing linearization technique for a CMOS OTA and its application to tunable Gm-C filter design. Microelectron. J. 46, 810–818 (2015) 3. Sarrafifinazhad, A., Kara, I., Baskaya, F.: Design of a digitally tunable 5th order GM-C filter using linearized OTA in 90 nm CMOS technology. In: IEEE International Symposium on Signals, Circuits and Systems (ISSCS) (2015) 4. Acosta, L., Jiménez, M., Carvajal, R.G., Lopez-Martin, A.J., Ramírez-Angulo, J.: Highly linear tunable CMOS Gm-C lowpass filter. IEEE Trans. Circ. Syst. I Reg. Pap. 56(10), 2145– 2158 (2009) 5. Lo, T.-Y., Hung, C.-C., Ismail, M.: A wide tuning range Gm-C filter for multi-mode CMOS direct conversion wireless receivers. IEEE J. Solid-State Circ. 44(9), 2515–2524 (2009) 6. Crombez, P., Craninckx, J., Wambacq, P., Steyaert, M.: A 100-kHz to 20-MHz reconfigurable power linearity optimized GM-C biquad in 0.13-μm CMOS. IEEE Trans. Circ. Syst. II Exp. Briefs 55(3), 224–228 (2008) 7. Glib, J.P.K.: Bluetooth radio architectures. In: IEEE Radio Frequency Integrated Circuit Symposium, Digest of Papers, 11–13 June 2000, pp. 3–6. IEEE (2000) 8. Elwan, H.O., Younus, M.I., Al-Zaher, H.A., Ismail, M.: A buffer-based baseband analog front end for CMOS Bluetooth receivers. IEEE Trans. Circ. Syst.-II Analog Digit. Process. 49(8), 545–554 (2002) 9. Johns, D., Martin, K.: Analog Integrated Circuit Design, 1st edn. Wiley, New York (1997) 10. Razavi, B.: RF Microelectronics, 1st edn. Prentice Hall, New York (1998) 11. Ishikuro, H., et al.: A single-chip CMOS Bluetooth transceiver with 1.5 MHz IF and direct modulation transmitter. In: 2003 IEEE International Solid-State Circuits Conference, Digest of Technical Papers, ISSCC, vol. 1, pp. 94–480 (2003) 12. Andreani, P., Mattisson, S.: On the use of Nauta’s transconductor in low-frequency CMOS Gm-C bandpass filters. IEEE J. Solid-State Circ. 37(2), 114–124 (2002) 13. Leung, W.Y., Cheng, K.M., Wu, K.-L.: Design and implementation of LTCC filters with enhanced stop-band characteristics for Bluetooth applications. In: 2001 Asia-Pacific Microwave Conference, APMC 2001, 3–6 December 2001, vol. 3, pp. 1008–1011 (2001)

Design of Power-Efficient Operational Transconductance Amplifier

151

14. Son, M.H., Lee, S.S., Kim, Y.J.: Low-cost realization of ISM bandpass filters using integrated stripline structures. In: IEEE Radio and Wireless Conference, RAWCON 2000, 10–13 September 2000, pp. 261–264 (2000) 15. Liu, H., Zhu, X., Lu, M., Sun, Y., Yeo, K.S.: Design of reconfigurable dB-linear variablegain amplifier and switchable-order GM-C filter in 65-nm CMOS technology. IEEE Trans. Microwave Theory Tech. 67(12), 5148–5158 (2019). https://doi.org/10.1109/TMTT.2019. 2947668 16. Abdolmaleki, M., Dousti, M., Tavakoli, M.B.: Design and simulation of tunable low-pass Gm-C filter with 1 GHz cutoff frequency based on CMOS inventers for high speed telecommunication applications. Analog Integr. Circ. Sig. Process. 100(2), 279–286 (2019). https:// doi.org/10.1007/s10470-019-01484-0 17. Taleja, M.K., Kumar, M.: Bias current effect on gain of a CMOS. In: IEEE International Conference on Advanced Computing & Communication Technologies, pp. 396–397 (2011) 18. Das,S.S., et al.: An analytical 2-D model of triple metal double gate graded channel junctionless MOSFET with hetero-dielectric gate oxide stack. Solid State Commun. 340, 114521 (2021). ISSN: 0038-1098. https://doi.org/10.1016/j.ssc.2021.114521 19. Kar, S., Sen, S.: A highly linear CMOS transconductance amplifier in 180 nm process technology. Analog Integr. Circ. Sig. Process. 72, 163–171 (2012). https://doi.org/10.1007/s10 470-011-9796-1 20. Abbasalizadeh, S., Sheikhaei, S., Forouzandeh, B.: A 0.9 V supply OTA in 0.18 nm CMOS technology and its application in realizing a tunable low-pass Gm-C filter for wireless sensor networks. Sci. Res. Circ. Syst. 4, 34–43 (2013)

First Order Control System Using Python Technology Palash Jain1 and Jay Kumar Jain2(B) 1 Department of Electronics and Communication Engineering, SIRT, Bhopal, MP, India 2 Department of Computer Applications, SIRT, Bhopal, MP, India

[email protected]

Abstract. The research work highlights the application of python to practically study some basic concepts of control system engineering. The presented chapter consists of basic operations needed to perform in control system designing in python using libraries like Matplotlib, NumPy etc. Here the work presents computation and calculation of poles and zeros, Nyquist plot, bode plot, sinusoidal response, impulse response, step response on the given transfer function. And it is identified that the python code is very small and output time is very less as compared to MATLAB. Keywords: Python · Control system · Bode plot · Impulse response · Root Locus · Poles-zeros

1 Introduction Due to the technological advancements in the twentieth century, there is a need to fill up the gap between theory and the real world in the future engineering education system in colleges. Control system engineering is an important and specialized course for students of electronics, electrical and mechanical engineering in colleges. In this chapter, authors has introduced python programming in learning and understanding the basic & tedious concepts of control system engineering. With the help of python programming the practical problems of the control system can easily be solved in a few steps. This chapter is structured as follows: In Sect. 2 a brief description of python and its libraries has been discussed. In Sect. 3 basics of first order control system 4 we will discuss how to use python programming in understanding and solving the problems of control system engineering & do some analysis of control system engineering. In Sect. 5, a conclusion is drawn on the basis of above analysis and study.

2 Python Python [1] is a general purpose, dynamic, versatile and powerful programming language. It has high-level data structures and have a simple approach to object-oriented programming. Python is known for its simplicity, as it has very simple syntax and can © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 152–160, 2022. https://doi.org/10.1007/978-981-19-1742-4_12

First Order Control System Using Python Technology

153

work on any platform like windows, Mac , linux, raspberry pi etc. Python’s prototyping is very quick as it runs on an interpreter system. The biggest strength of python is its vast/rich collection of open source libraries which is used in scientific computing, image processing, GUI applications, web frameworks. etc. [1] In this paper the latest version of python i.e python 3 is used along with jupyter notebook. In the study of control systems using python, three libraries or packages are used namely NumPy, matplotlib [2] & control system package. NumPy stands for numerical python. Python numpy package contains n-dimensional array object & it is core library for scientific computing. Numpy array is in the form of rows & columns . Mathematical & logical operations can be performed using numpy. Generation of random numbers and linear algebra operations can also be performed using numpy. The combination of NumPy with SciPy (Scientific python) & matplotlib library is widely used as an alternative to MATLAB. NumPy can be installed by typing “pip install numpy” in command prompt and can be import by typing-Import numpy as np [2]. Matplotlib is a graphic plotting library used in python programming language. Simple & complex plots can be designed using matplotlib package with few commands. It is a multiplatform data visualization tool built on numpy & scipy packages. Matplotlib can be installed by typing “pip install matplotlib” in command prompt and can be import by typing-Import matplotlib.pyplot as plt. [3]. The python control system library performs operations for design and analysis of feedback control systems. Using python control system library, both time and frequency response analysis can easily be performed which includes frequency plots & analysis of linear time invariant systems. The python control system package can easily be installed by typing “!pip install –U Control” and can be import by using- Import control as co [4]. In this paper, Jupyter notebook is used in the analysis and study of control system engineering using python. Jupyter notebook is an open source application that is used for machine learning, modelling simulation, data visualization and much more. Installation of jupyter notebook is done using Anaconda distribution which includes python3, jupyter notebook and various other packages for data science and scientific computing [5].

3 Control System Control system is a system in which the output of the system varies in coordination with the input system. Large varieties of control systems are available. Like, open loop/closed loop system, time independent/time dependent system, deterministic/non deterministic, linear/non linear etc. Here the first order linear time invariant system is taken into consideration to understand the concepts. The physical system is analyzed using its mathematical model. The transfer function is the mathematical representation of the physical system. The transfer function defines the system properties. The analysis of the control systems can be done in both time domain and frequency domain. If the output of the control system varies with respect to time, then it is called the time domain response of the control system. It includes transient response and Steady state response. The standard test signals used for analysis are impulse, step, ramp and parabolic. These signals are used to analyze the performance of the control systems in the time domain. The frequency domain analysis gives the response of a mathematical model with respect to

154

P. Jain and J. K. Jain

frequency. frequency domain elaborates the stability of the system. Here BODE and NYQUIST plot analysis were discussed for frequency domain.

4 Python Application in Control At the education level control systems problems are mainly analyzed using MATLAB. MATLAB software provides Control Engineering Toolbox in which systematic analysis, designing, and simulation of linear control systems are performed. In control system toolbox transfer function can be specified for the system, state-space will be done, analysis of pole zero gain it’s frequency response modeling can be done with ease. It provides features like step response, Bode plot, Nyquist plot etc. for analysis of behavior of system with respect to time and frequency. But the software is not freely available and affordable for individuals. This work provides an insight for the possibility of using python open source platform to perform all above mentioned tasks of MATLAB for control system engineering [6, 7]. Here the Fig. 1 below shows how to make the transfer function in Python. It also shows the calculation of poles and zeros.

Fig. 1. Transfer function in Python

Figure 2 below shows the pole zero plot of the 2nd order transfer function. Here poles are represented by ‘x’ and zeros are represented by ‘o’. Given transfer function has 2 poles and 1 zero. Figure 3 below shows the Step response of the transfer function using Python. When unit step input is applied to the given transfer function the output shows transient state for a few seconds, then shows steady state response. Transfer function, T.F =

s2

2s + 5 + 2s + 3

First Order Control System Using Python Technology

155

Fig. 2. Pole zero plot of the 2nd order transfer function

Fig. 3. Step response of the transfer function using Python

Figure 4 below shows the Impulse response of the transfer function using Python. Impulse response is basically inverse Laplace transform of the transfer function, which is a single command operation in python. Transfer function, T.F =

2s + 7 s2 + 3s + 2

156

P. Jain and J. K. Jain

Fig. 4. Impulse response of the transfer function using Python

Figure 5 below shows the Sinusoidal response of LTI System using Python. The graph is drawn between sinusoidal input applied to the transfer function and its output in the time domain. Transfer function, T.F =

s2

2s + 5 + 2s + 3

Fig. 5. Sinusoidal response of LTI system using Python

First Order Control System Using Python Technology

157

Python also facilitates the conversion of State space representation to Transfer Function & Transfer Function to state space representation. The Fig. 6 and Fig. 7 below depicts the same.

Fig. 6. State space representation to transfer function

Fig. 7. Transfer function to state space representation

Control systems behavior are analyzed using various plots like Root Locus, Nyquist and Bode plots. These plots are easily generated in Python. Figure 8 below shows the Root locus plot, with centroid at −1 and poles below and above the imaginary axis along with asymptotes. Transfer function, T.F =

s3

+ 2s2

s + 3s + 1

158

P. Jain and J. K. Jain

Fig. 8. Root locus plot

Figure 9 below shows the Nyquist plot, To check whether the system is stable or not.

Fig. 9. Nyquist plot

First Order Control System Using Python Technology

159

Figure 10 below shows the Bode plot. The plot displays the magnitude (in dB) and phase (in degrees) of the system response as a function of frequency. Transfer function, T.F

2s2 + 5s2 + 12s2 + 5s + 1 s2 + 2s + 3 s2 + 2s + 3

Fig. 10. Bode plot

5 Conclusion and Future Work In this chapter, an overview study is presented on how Python can perform all the tasks which are required to be done using the Control Toolbox of MATLAB. The chapter discussed calculation of transfer function, pole-zero calculation, step response, impulse response along with various plots like Root Locus, Bode and Nyquist plot for first order and second order LTI systems. All other functions required in control system analysis can also be explored using Python. The open source access of Python and implementation using Jupiter and empowers the researchers and students to work economically and produce the same results in Python compared to MATLAB. Acknowledgement. We would like to thank the Institute Innovation Council, SIRT Bhopal, who have given us time to time support for this prestigious work.

References 1. Hashemian, R.: S-Plane bode plots - identifying poles and zeros in a circuit transfer function, In: 2015 IEEE 6th Latin American Symposium on Circuits & Systems (LASCAS). IEEE (2015)

160

P. Jain and J. K. Jain

2. Ranjani, J., Sheela, A., Pandi Meena, K.: Combination of NumPy, SciPy and Matplotlib/Pylab a good alternative methodology to MATLAB - a comparative analysis. In: 2019 1st International Conference on Innovations in Information and Communication Technology (ICIICT) IEEE (2011) 3. Teixeira, M.C.M.: A method for plotting the complementary root locus using the root-locus (positive gain) rules. IEEE Trans. Educ. 47(3) (2004) 4. Pejovic, P.: Replotting the Nyquist plot—a new visualization proposal. In: 2019 20th International Symposium on Power Electronics (Ee), 23–26 October 2019, Novi Sad, Serbia (2019) 5. www.python.org 6. Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based backstepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 7. Jain, J.K.: Secure and energy-efficient route adjustment model for internet of things. Wireless Pers. Commun. 108(1), 633–657 (2019)

Water Cleaning Bot with Waste Segregation Using Image Processing M. M. Anas, M. Athiram(B) , Anugraha Suresh, K. Archana, and Maneesha Shaji Department of Electronics and Communication Engineering, Government Engineering College Wayanad, Wayanad, India [email protected], [email protected]

Abstract. Dumping of waste into water bodies has become a major threat to the environment. Removal of these floating waste such as plastic, paper, cardboard etc. is a biggest challenge. This paper proposes a ‘Water cleaning bot’ along with a waste segregating mechanism to remove the waste from the water surfaces. It also includes an automated waste classification system using Convolution Neural Network (CNN) algorithm, a Deep Learning based image classification model used to classify objects into bio and non-biodegradable, based on the object recognition accuracy in real-time. The hardware part consist of ESP32 and Raspberry pi as the core module. The waste collection mechanism is established through a belt conveyor mechanism. Keywords: CNN · Waste detection · Waste sorting · Web analysis · Bot control mechanism

1 Introduction The main objective of our project is removal of floating waste from water surface. If we collected the waste from water there is no method to separate without any manual control. So here we propose an efficient method to collect and separate waste from water using image processing. It will definitely create a break through on the society and government organizations. The main concept is that a remote controlled bot which will moves through the water surface collects waste and separates the wastes into biodegradable and non biodegradable by using image processing. Waste is detected using Convolution Neural Network. Once the waste is detected the conveyor start working and the waste will comes to the collection unit. Once waste reaches the conveyor image processing is done with the help of Raspberry pi and the waste is separated in to two categories biodegradable and non biodegradable, for that a wiper is used.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 161–171, 2022. https://doi.org/10.1007/978-981-19-1742-4_13

162

M. M. Anas et al.

2 Bot Design 2.1 Block Diagram of Overall System

Fig. 1. Bot structure

Figure 1 Shows the basic block diagram of the system. The major components are ESP32, Raspberry Pi, Motor drivers, Motors, Conveyor and Wiper. Here ESP32 is used as the main control of the bot, cam on it will capture the images and transmit them to the destination (Remote terminal) for the detection. Manual control and manual triggering is done through the WiFi network. Waste is collected using a conveyor belt and to drive this conveyor a dc geared motor is used and to energize this motor a motor driver (L293D) is also used. Once the waste is detected the conveyor start working and the waste will come to the collection unit. At that time the waste is separated in to two categories biodegradable and non-biodegradable for that a wiper is used. A servo motor is used to control this wiper. The thrust provided to control the bot is given by a geared dc motor and a motor driver is used to control it. Once waste reaches the conveyor image processing is done with the help of Raspberry pi, the pi cam is used to capture these images and these images are used for further classification and decides to which category it belongs to, this decision is transferred to ESP32 and the wiper takes action accordingly. 2.2 Data Set Collection Data set is a collection of data used for training and testing of a network. The data set used for training is collected from trash net. The floating waste classification data set contains 1400 images from 3 categories:- plastic (500), paper (400), cardboard (500) (Fig. 2).

Water Cleaning Bot with Waste Segregation

2.3 Waste Detection Step *1:- Creating the Network head

Fig. 2. .

Step *2:- Classification of Data set collection Step *3:- Creating Regression box

Fig. 3. Regression box

163

164

M. M. Anas et al.

Step *4:- Training: The flowchart shows the steps of training process.

Fig. 4. Training flowchart

• First is to initialize the list of data and also load the input images and bounding box. • After updated the loaded data convert the data to numpy arrays and scaling the input pixel. • Then split the data into training and testing. • Creating a model which is FC layer removal and new one is created. Then train the network and plot the accuracy and loss plot. Step *5:- Testing (Figs. 3 and 4). Determine the input file type, by loading image path in the testing file and load input image from disk (Figs. 5, 6 and 7). • Preprocess the loaded image and predict the object along with the class label. • Then load the input image (in OpenCv format). • Mark the predicted bounding box and class label on the image.

Water Cleaning Bot with Waste Segregation

Fig. 5. Class label accuracy

Fig. 6. (a) Total loss (b) Class label loss (c) Bounding box loss

165

166

M. M. Anas et al.

Fig. 7. Testing flowchart

Fig. 8. Detection output

2.4 Waste Classification *Training Process steps:- Initialize the number of epochs to train for, initial learning rate, batch size, image dimensions, data and labels (Figs. 8, 9 and 10). • Grab the image paths and randomly shuffle them loop over the input images. • Scale the row pixel intensities to range [0,1] and binarize the labels and partition the data and construct image generator for augmentation. • Creating a model then train the network and save the model to disk and plot the training loss and accuracy.

Water Cleaning Bot with Waste Segregation

Fig. 9. Classification training flowchart

Fig. 10. Training loss and accuracy

167

168

M. M. Anas et al.

*Testing Process steps:• Load the image and preprocess the image for classification. • Then load the trained CNN and classify the input image. • Mark the prediction of the input image filename contains the predicted label text and build the label. • Draw the label on the image and show the output image.

Fig. 11. Classification testing

Fig. 12. Classification output

Water Cleaning Bot with Waste Segregation

169

2.5 Web Analysis

Fig. 13. Web analysis flowchart

2.6 Bot Control Flow chart shows the working of motors. Here the data is giving through a virtual terminal for the motion of motors. If data is “f” the 2 dc motors move forward, if it is “b” move backward, “r” and “l” for right and left movements. And “s” for stop. “C” and “c” is given to start and stop of conveyor. The servo motor work according to the input, it rotate 180° (Figs. 11, 12, 13 and 14). The main problem of existing system is when human does the waste sorting, they take lot of time. But the proposed system entire work can be automated so there is no need of any kind of skilled workers. It is user friendly. The second problem is the waste is highly contagious and harmful as it contain bacteria, virus which may lead to bad health of people working in the waste sorting process. The third problem of existing system is very costly. The proposed system initial and maintenance cost is low. Bot can be constructed using cheap materials so the initial cost can be reduced. It is applicable to reduce water pollution in rivers, canals and ponds. It does not harm the aquatic animals (Fig. 15).

170

M. M. Anas et al.

Fig. 14. Web analysis output

Water Cleaning Bot with Waste Segregation

171

Fig. 15. BOT control

3 Conclusion The primary objective of our project is to collect floating waste and to reduce human labour. The framework consists of two parts, one of which is the hardware platform with Raspberry Pi and ESP32 as the core module and the other is the software platform using CNN algorithm. The experimental result demonstrate that our system has 90 percentage average accuracy rate. In the long term, we will evaluate various feature extraction algorithms and classifiers to realize higher image classification accuracy.

References 1. Sinha, A., Bhardwaj, P., Vaibhav, B., Mohammad, N.: Research and development of robot – an autonomous river cleaning robot. Intell. Robots Comput. Vis. Algor. Tech. 9025, 1–8 (2014) 2. Ganesh, U.L., Rampur, V.V.: Semi-automatic drain for sewage water treatment of floating materials. Int. J. Res. Eng. Technol. 5(7), 1–4 (2016) 3. Chinnathurai, BM.: Design and Implementation of a Semi-Autonomous Waste Segregation Robot. In: IEEE Conference 3rd edn., vol. 2, pp. 68–73 (2016) 4. Das, S., et al.: Advance machine learning and artificial intelligence applications in service robot. Artif. Intell. Fut. Gen. Robot. 2021, 83–91 (2021). https://doi.org/10.1016/B978-0-32385498-6.00002-2 5. Ghuge,M., Bhadoria, S.N.: Automatic waste sorting based on image processing. Int. J. Adv. Res. Comput. Eng. Technol. 7(6), (2018), ISSN: 2278-1323

Design and Analysis of a Photovoltaic P&O-Based MPPT Lead-Acid Battery Shiv Prakash Bihari1(B) , Aruvand Gupta2 , Vikalp Gupta3 , and Abhinav Kumar Babul4 1 Department of Electrical and Electronics Engineering, Inderprastha Engineering College,

Ghaziabad, UP, India [email protected] 2 EEE, Inderprastha Engineering College, Ghaziabad, UP, India 3 EEE, Maharaja Agrasen Institute of Technology, Delhi, India 4 Industrial Systems and Drives, Madhav Institute of Technology and Science, Gwalior, M.P., India

Abstract. The solar is one of the leading energy producer when it comes to renewables therefore extending our knowledge on the connection between the solar module and the battery charging process has vital role in today’s sceneries. PV Solar module’s power is affected by changing temperature & insolation, which are mitigated with the help of an MPPT system. To obtain the greatest power out of the solar system, we used a perturb and observe based MPPT technique in this work. A PV system requires an appropriate battery charge and to stabilize the power flow from the solar PV module to the battery and load a controller is utilized, allowing solar electricity to be used efficiently. In this situation, a boost regulator is in charge of the battery’s charging mechanism. For the boost converter, PV module, parameter extraction and evaluation MATLAB/Simulink models are utilized for analysis. Keywords: Solar energy · Photovoltaics (PV) system · Perturb and observe (P&O) · Converter · Li-ion batteries · Maximum Power Point Tracking (MPPT)

1 Introduction Despite supply chain issues and construction delays caused by the pandemic, renewable capacity additions in 2020 increased by more than 45% over 2019 and set a new high. The expansion was fueled by a 90% increase in global wind capacity expansions. The development of new solar PV installations by 23% to nearly 135 GW in 2020 also contributed to this record growth [1]. This is owing to the consistent decline in the cost of solar modules over the last decade. Furthermore, because there are no moving apparatus included in its procedure, it is generally acknowledged by various commerce due to its ease of flexible scalability, inexpensive maintenance and installation. The photovoltaic system generates power using sunlight. As a result, it can only be utilized during the day when there is sufficient sunlight. One way for harnessing solar energy during any interval of the day even when solar radiations isn’t accessible, is battery energy storage. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 172–182, 2022. https://doi.org/10.1007/978-981-19-1742-4_14

Design and Analysis of a Photovoltaic P&O-Based MPPT

173

Solar power is regarded as clean, great reliability, and source of energy renewable with an extended lifespan. The system introduced could be installed around or near where the prevention in transmission losses and contribute to CO2 emission reductions in metropolitan areas can take place. A solar array is built out of numerous PV cells associated in series and parallel. The connections in parallel are accountable for boosting the current in array, while the connections in series are accountable for enhancing the voltage of the modules. The amount of power collected from a PV module is determined by the voltage generated in the solar PV module, the cell’s temperature, and the amount of solar radiation received. The biggest downside of PV is its minute power conversion effectiveness when equated to other substitute energy sources. PV is a non-linear energy source that operates in response to irradiance and temperature. To derive maximum about of power from the solar PV module MPPT is utilized. The P&O technique is now the supreme and commonly utilized method in compared to other methods [4]. There appears to be a limited number of PV models in the Matlab/Simulink Sim Power System programmed to integrate with current electronics modelling technology. As a result, assessing and simulating a PV power system is quite difficult [17]. This study incorporates the solar PV modules, MPPT approach based on P & O technique, boost converter, and charging system of Li-ion battery. The structure of chapterSection 2 delivers a gist literature review. Section 3 talks about the methodology used. And finally, the result and discussion are concluded in the Sect. 4.

2 Literature Review In this section, research on charging Li+ batteries, MPPT, and charge controllers is reviewed and examined, as well as information on MPPT, charging Li+ batteries, and charge controllers that do both is discussed. This part is intended to serve as the primary source of background information while maintaining the research’s reliability [16]. The charging process of battery by PV modules is often very private, and the ideas behind how it function is rarely made public. However, significant research has been conducted on a few of distinct combined topologies for charging batteries with the help of solar PV module. The fundamental notion behind creating a solar PV module charger for a battery is to integrate MPPT concepts with efficient battery charging. The P&O algorithm MPP is used to track a solar panel system in this work [5]. This method is overridden by a CC/CV charging algorithm. Because the maximum power capability of the solar PV module may exceed the capacity of the battery, CC/CV is employed. Despite its resemblance to other chargers, this method utilizes a lead-acid battery rather than a Li+ battery [10]. This work [6] revisits a lead-acid battery based P&O algorithm. They don’t say how they charge, but it’s safe to presume they’re utilizing MPPT. Because of the P&O algorithm’s constant step sizes, their MPPT skills are limited.

174

S. P. Bihari et al.

Under rapid weather change and partial shade circumstances, this [18] work offers a novel hybrid MPPT-algorithm combining the MIWO (Modified Invasive Weed Optimization) and P&O techniques for effective calculation of MPP (maximum power point) from a freestanding PV dependent hybrid module. In order to obtain speedy global peak (GP) and MPP, MIWO carries the earliest phases of MPPT, which is trailed by the utilization of the P&O in the latter phases. The perturb and observe technique of MPPT is implemented in this research [19] in the PV solar module to harvest maximum efficiency. A reliable battery charge controller is essential for a PV system to regulate the energy flow provided by the solar PV module to the battery and load in order to properly utilize photovoltaic power. In this situation, a boost regulator is in charge of the battery’s charging mechanism. The converter, modal evaluation, and parameter extraction are all done using a MATLAB/Simulink model. This paper [20] describes a low-voltage grid-connected PV system that is appropriate for industrial, small-scale, and suburban consumer utilization. MPPT approaches are also executed by means of boost converter topology employing INC (incremental conductance) and P&O. The INC algorithm is more precise than the P&O technique at tracking swiftly changing irradiation environments. The voltage never reaches a precise value in the P&O approach, but it fluctuates about the MPP. As a result, the INC technique gets the MPP enhanced and faster than P&O since it does not suffer from drifting and is the most effective under quickly varying situations. A solar PV boost dc–dc converters based adaptive P&O-fuzzy control MPPT is proposed in this [21] research. Both of these advantages are combined in the proposed technique. It should enhance MPPT performance, particularly in the presence of disturbance. For assessment and comparison analysis, traditional fuzzy logic control and P&O algorithms have also been created. All of the methods were merged and evaluated in MATLAB-Simulink utilizing a solar PV module. It is the simplest algorithm and is easiest to apply. It does, however, have some typical flaws, which are listed below [21]. 1) During rapid fluctuations in irradiance as it moves away from the real MPP, there is insufficient monitoring, intelligence, and efficiency (maximum power point) 2) The ability to determine whether the new higher output power value is due to a duty cycle shift or a change in irradiation has been lost. 3) At low irradiance, constant oscillations around the optimal operating point cause the average power level to depart from the MPP. 4) It swoops back and forth over the MPP, struggling to remain stationary. 5) Sluggish response time. When the power is changed, the P&O control approach considers whether the voltage is improved or reduced. Understanding which direction the voltage altered when the power changed can help you figure out how to change the duty cycle. This technique has several advantages, including low complexity, continual convergence rates to MPP, and the capability to step in the accurate course a set amount depending on which side it is climbing. Limitations include at MPP the difficulty of achieving 0% steady-state inaccuracy, four different step sizes, and dependency on calculating equipment features [7].

Design and Analysis of a Photovoltaic P&O-Based MPPT

175

3 Methodology The solar PV P&O based MPPT battery charge with a boost controller model constructed in the MATLAB/Simulink is shown in Fig. 1.

Fig. 1. Overview of solar PV MPPT charge controller mode.

The PV solar module in this model was rated at 349 W, with three parallel strings and ten series strings, totaling 30 solar panels connected. The solar arrays inputs are 1000 W/m2 irradiance and the temperature around 25 °C. The solar PV module’s output is taken through the measurement port and routed to a block, which is then routed to the MPPT controller. The P&O algorithm was utilized for the MPPT controller. The panel voltage was compared to the reference voltage, which was then supplied to the PI controller, which was then compared to a repeating sequence for accuracy, and maybe a gate pulse was created. The boost converter’s MOSFET was now supplied this gate pulse. A capacitive filter was integrated to the load end of the converter to get a smooth voltage and current waveforms. The battery utilized has a capacity of 40 KWh and is a Li-ion battery. For performance study, the waveforms of its SOC, Voltage, and Current Gate pulse, the PV characteristics and VI characteristics of the solar array are evaluated and simulated in the Simulink.

176

S. P. Bihari et al.

a) Boost converter A resistive load connected to boost converter is coupled to a solar PV module, as shown in Fig. 1. A switching-mode regulator can be utilized in a Dc-Dc converter to convert an uncontrolled dc voltage to a controlled dc output voltage. The voltage is regulated using the PWM (pulse-width modulation) approach and a MOSFET switching device. To increase the dc voltage, a boost converter is employed. When the MPPT algorithm changes and modulates the boost Dc-Dc converter’s PWM duty cycle, maximum power is achieved. The duty cycle D of the power switch modulates the power transmission from the input source to the load [13].   (1) Vpv ton = Vout − Vpv .toff And, Vout =

ton + toff Vpv toff

(2)

Where; T = ton + toff

(3)

The report ton/T is referred to as the duty cycle α, and as a result α=

ton T

(4)

The voltage release can be calculated using Eq. (3): Vout =

1 Vpv 1−α

(5)

where: • Vout : is the output voltage. • Vpv : is then voltage input (solar cell). • ton : is the duration of time when the switch is closed. b) Maximum power point tracking algorithm In this article, the P&O MPP algorithm is used to model a solar panel system. The P&O MPPT is widely employed in medium and small commercial PVgrid-connected inverters and charge controllers, due to its tracking efficiency and ease of utilization. The MPPT algorithm monitors the peak power of the PV generator and communicates the duty cycle of the PV generator to the battery charge controller, which is relevant to the monitored peak power [11]. This method, which is based on trial and error, is used to track the highest performance point (Figs. 2 and 3). The duty cycle of the converter switching device is altered, which changes the boost converter’s actual input resistance, the software detects power fluctuations and disrupts the photovoltaic panel’s working voltage. Then wait until it reaches peak performance,

Design and Analysis of a Photovoltaic P&O-Based MPPT

Fig. 2. Waveforms of VI characteristics and PV characteristics of the solar module.

Fig. 3. Typical waveforms of boost converter.

177

178

S. P. Bihari et al.

then repeat the process forever. Because just voltage is measured in this manner, it is simple to implement. The power output of the system is verified using this method by adjusting the provided voltage. If power rises in tandem with voltage, the “is increased further; otherwise, the” is reduced. Similarly, if there is a reduction in voltage the power is increased, the duty cycle is reduced. These processes are carried out till the time the MPP has been attained. The voltage at which MPP is obtained is referred to as the reference point (Vref). As shown in Fig. 4, If the PV module’s operating voltage changes and power increases, the control system shifts the operating point in that direction; otherwise, it shifts in the other direction. As a result, depending on which side of the hill you’re on, smaller or bigger duty cycle increments may be more favorable. In comparison to the hill-climbing approach, this results in less ripple at MPP. Because of its intuitive nature and strong convergence to MPP, this is the most extensively used approach in the industry. Adding a variable size to the duty cycle increments, on the other hand, improves the process even more. This approach is also vulnerable to rapidly changing weather conditions, though this can be mitigated by updating at a higher frequency [14, 15].

Fig. 4. P&O flowchart

Design and Analysis of a Photovoltaic P&O-Based MPPT

179

Fig. 5. Waveform of voltage and current of battery

Fig. 6. Waveform of SOC of battery

c) PI Controller To charge a lead-acid battery, the PI controller was created. The integral controller which is proportional generates an output, u(t), that is proportional to both the input, vi (t), and the integral of the input, vi (t).  u(t) = Kp vi (t) + ki vi (t)dt The algorithm works according to the equation of the PI controller.

180

S. P. Bihari et al.

Vref is the reference voltage attained from the MPPT. This Vref is compared to the PV voltage, Vpv , to generate an error signal, which is then sent to the PI control. The appropriate response can be attained by carefully setting the proportional gain, Kp , and integral gain, Ki [12]. When the power from the solar PV module is pumped into the boost converter and the PI controller is turned on, the duty cycle is varied, changing the input value perceived by the PI controller.

Fig. 7. Waveform of gate pulse of MOSFET

4 Discussion and Results The simulations of MPPT battery charge controller for an isolated photovoltaic system model was successfully carried out in the MATLAB/Simulink for research‘s performance. This performance investigation verifies the MPPT algorithm’s tracking capability. In this system solar panel array was rated 349 W, which is having 3 parallel and 10 series strings, meaning a total of 30 solar panels are connected in this solar array. The input of the solar array are 25 °C temperature and 1000 W/m2 irradiance. The output of the solar panel is taken through the measurement port and given to go to block which was further given to the MPPT controller. For MPPT controller P and O (Perturb and Observe) algorithm was used. Where we compared the panel voltage with the reference voltage, which was then sent to the PI controller which was then again compared with repeating sequence for accuracy and perhaps a gate pulse was generated. Now this gate pulse was given to the MOSFET of the boost converter. A capacitive filter was added to the load end of the converter to get a smooth voltage and current waveforms. The rating of battery used is of 40 KWh capacity which is a Li-ion battery. Which is charged and the waveforms of its SOC (Fig. 6), Voltage (Fig. 5) and Current (Fig. 5) are also simulated and displayed. And also, Gate pulse (Fig. 7), the PV characteristics and VI characteristics (Fig. 2) of the solar array is simulated and displayed. Any commercialized MPPT charge controller with a comparable architecture can be modified using this Simulink

Design and Analysis of a Photovoltaic P&O-Based MPPT

181

model. An actual commercialized solar PV MPPT charge controller experimental setup was used to test the Simulink model’s performance. This validated model assists in the optimum sizing of solar PV module and battery energy storage for small and medium freestanding PV installations.

References 1. IEA: Renewable Energy Market Update 2021. IEA, Paris (2021) 2. Arrouf, M.: Optimization of inverter, motor and pump connected with a photovoltaic cell, Ph.D. thesis, University of Constantine (2007) 3. Elgendy, M.A., Zahawi, M.A., Atkinson, D.J.: Evaluation of perturb and observe MPPT algorithm implementation techniques. In: 6th IET International Conference on Power Electronics, Machines and Drives (PEMD 2012), pp. 1–6. IEEE Publication, March 2012 4. Chouder, A., Guijoan, F., Silvestre, S.: Simulation of fuzzy-based MPP tracker and performance comparison with perturb & observe method. Revue des Energies Renouvelables 11(4), 577–586 (2008) 5. Jana, J., Bhattacharya, K.D., Saha, H.: Design & implementation of MPPT algorithm for battery charging with photovoltaic panel using FPGA. In: 2014 6th IEEE Power India International Conference (PIICON) (2014) 6. Padhee, S., Pati, U.C., Mahapatra, K.: Design of photovoltaic MPPT based charger for lead-acid batteries. In: 2016 IEEE International Conference on Emerging Technologies and Innovative Business Practices for the Transformation of Societies (EmergiTech) (2016) 7. Elgendy, M.A., Zahawi, B., Atkinson, D.J.: Operating characteristics of the P&O algorithm at high perturbation frequencies for standalone PV systems. IEEE Trans. Energy Conver. 30(1), 189–198 (2015) 8. Esram, T., Chapman, P.L.: Comparison of photovoltaic array maximum power point tracking techniques. IEEE Trans. Energy Conver. 22(2) 439–449 (2007) 9. Subudhi, B., Pradhan, R.: A comparative study on maximum power point tracking techniques for photovoltaic power systems. IEEE Trans. Sustain. Energy 4(1), 89–98 (2013) 10. Mohanty, P., Bhuvaneswari, G., Balasubramanian, R., Dhaliwal, N.K.: MATLAB based modeling to study the performance of different MPPT techniques used for solar PV system under various operating conditions. Renew. Sustain. Energy Rev. 38, 581–593 (2014) 11. Padmagirisan, P., Sankaranarayanan, V.: Powertrain control of a solar photovoltaic-battery powered hybrid electric vehicle. Front. Energy 13(2), 296–306 (2019). https://doi.org/10. 1007/s11708-018-0605-8 12. Shaw, R.N., et al.: IOT based MPPT for performance improvement of solar PV arrays operating under partial shade dispersion. In: 2020 IEEE 9th Power India International Conference (PIICON) held at Deenbandhu Chhotu Ram University of Science and Technology, SONEPAT. February 28–March 1, India (2020) 13. Attou, A., Massoum, A., Saidi, M.: Photovoltaic power control using MPPT and boost converter. Balkan J. Electr. Comput. Eng. 2(1), 23–27 (2014) 14. Bhattacharyya, S., Samanta, S., Mishra, S.: Steady output and fast tracking MPPT (SOFTMPPT) for P&O and InC algorithms. IEEE Trans. Sustain. Energy 12(1), 293–302 (2020) 15. Shukla, S., Singh, B.: Reduced-sensor-based PV array-fed direct torque control induction motor drive for water pumping. IEEE Trans. Power Electr. 34(6), 5400–5415 (2018) 16. Heinen, G.D.: Modeling and charging control of a lithium-ion battery system for solar panels, Ph.D. thesis. Faculty of California Polytechnic State University (2017) 17. Shaw R.N., et al: A new model to enhance the power and performances of 4x4 PV arrays with puzzle shade dispersion. Int. J. Innov. Technol. Explor. Eng. 8(12) (2019)

182

S. P. Bihari et al.

18. Pradhan, C., Senapati, M.K., Malla, S.G., Nayak, P.K., Gjengedal, T.: Coordinated power management and control of standalone PV-hybrid system with modified IWO-based MPPT. IEEE Syst. J. 15, 3585–3596 (2020) 19. Garg, V.K., Garg, V.K.: To perform Matlab simulation of battery charging using solar power with maximum power point tracking (MPPT). Int. J. Electr. Electr. Eng. 7(5), 511–516 (2014) 20. Mohamed, S.A., Abd El Sattar, M.: A comparative study of P&O and INC maximum power point tracking techniques for grid-connected PV systems. SN Appl. Sci. 1(2), 174 (2019) 21. Zainuri, M.A.A.M., Radzi, M.A.M., Che Soh, A., Rahim, N.A.: Development of adaptive perturb and observe-fuzzy control maximum power point tracking for photovoltaic boost dc–dc converter. IET Renew. Power Gener. 8(2), 183–194 (2014)

Comparative Study of PID and Model Predictive Control with Boost Converter Akash Verma(B) , Sarthak Kamta Prasad, and Amritesh Kumar NIT Silchar, Silchar, India [email protected], [email protected]

Abstract. Model Predictive Control is a computationally intensive adaptive control strategy, which was firstly employed in chemical process industries . With the recent developments in computer processors, it is now one of the most effective control techniques for any power converter and can handle multiple constraints, disturbances, inputs, and outputs. On the other hand, the PID control technique which is used conventionally is simple and effective under normal operating conditions of power converters. The main objective of this chapter is to compare the performance of these two control strategies for power converters with mathematical modeling and simulation in MATLAB Simulink. A Boost converter is used as the power converter. Keywords: Boost converter · Duty · MPC · PID

1 Introduction Power electronic converters are the basic need of any controlled power circuit today. These circuits are used almost everywhere from household appliances to industrial environments [1]. The main reason behind it is the high efficiency, high controllability, and high power handling capability of power converters. The main components of these systems are storing elements (Inductor and Capacitor) and switching devices such as MOSFETs, IGBTs etc. All the power electronic systems include a power source and load linked with a power converter controlled by a control circuit [2]. The control circuit is a low power circuit, which drives the electronic switches of the converter. The block diagram of a power electronic converter system is shown in Fig. 1. The control signal for any converter is the switching sequence of the switches used. The output depends upon the pattern of this switching sequence from the control circuit. Parameters of any switching signal are frequency and duty ratio. These two parameters play important roles in designing and controlling a converter circuit [3]. In any power electronic system, switching sequence is modified by a separate control unit which consists of controller, driver, and some sensors. The control unit takes feedback of different states of the circuit and modifies the control signal accordingly. To control any power electronic circuit, there are many control techniques implemented successfully. PID is the most commonly used technique in power electronics. PID controllers are tuned © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 183–195, 2022. https://doi.org/10.1007/978-981-19-1742-4_15

184

A. Verma et al.

Source

DC

DC

Load

Control Signal

Feedback

Control Unit Fig. 1. Block diagram of a power electronic system

according to the parameters of open-loop response of the plant [4]. But it has limitations when it is used for non-linear systems under disturbances and varying operating points. In this chapter, PID control is implemented and compared with a more advanced control technique named Model Predictive Control (MPC). The converter used here is one of the most simple and efficient converters named Boost converter. Boost converter is employed for boosting a dc voltage using two storing elements and two switches. The diagram of a boost converter is shown in Fig. 2 (Table 1). L

Io

IL

+

D

S

C

VS

R IC

VO -

Fig. 2. Boost converter circuit diagram Table 1. Symbols and their descriptions Symbols

Description

VS

Input voltage

R

Load resistance

C

Capacitance

L

Inductance

IL , Iin

Inductor current (continued)

Comparative Study of PID and Model Predictive Control

185

Table 1. (continued) Symbols

Description

IC

Capacitor current

VC , VO

Output voltage

IO

Output current

u

Switching signal

T

Gate signal time period

TS

Sampling time

fS

Sampling frequency

rL

Inductor series resistance

îL

Small variation in input current

d

Small variation in duty

vo

Small variation in input voltage

2 Mathematical Modelling 2.1 State Space Model Model predictive control can also be used for non-linear and MIMO systems [5]. It decides the best control signal to be given to the system, by exploring all possible outputs from the model. State-space small-signal model is preferred for the application of MPC because all the intermediate states are used as the inputs of the controller. The smallsignal state-space model of the system is obtained using all the differential equations in both the switching states of the converter (Fig. 3) [6]. L

L

Io

IL

VS

+ C

R IC

(a)

IL

VO VS

Io ID

+ C

R IC

-

VO -

(b)

Fig. 3. Circuit diagram of boost converter (a) when switch is ON, (b) when switch is OFF

186

A. Verma et al.

Case – 1: When Switch is ON (u = 1, for time period D.T). dIL VS = dt L

(1a)

VO dVO =− dt R.C

(1b)

Case – 2: When Switch is OFF (u = 0, for time (1-D).T) VS − VO dIL = dt L

(2a)

dVC IL VO = − dt C R.C

(2b)

After combining both the cases (Eqs. 1 and 2), we get: dIL VS VO = + (u − 1) dt L L

(3a)

dVO IL VO =u − dt L R.C

(3b)

A digital controller is employed for implementing MPC. So, the state equations of the system are to be discretized. After sampling the system at sampling frequency f S , the state equations can be represented as: TS =

1 fS

TS .VS TS .VO + L L   TS TS .IL (k) VO (k + 1) = u. + 1− VO C R.C

IL (k + 1) = IL (k) + (u − 1)

(4a) (4b)

Let IL = x1 and VO = x2 , then the general continuous time state space model can be represented as x˙ = Ax + BVS

(5a)

y = Cx + DVS

(5b)

x˙ can be written as x˙ =

x(k + 1) − x(k) TS

Comparative Study of PID and Model Predictive Control

187

Discrete time representation: x(k + 1) = A x(k) + B Vs

(6a)

y(k) = C  x(k) + D Vs

(6b)

Where, x(k) = [x1 (k)x2 (k)] By comparing equation sets (4) and (6), we get   1(u−1).Ts  L  A = u.Ts Ts C 1 − R.C   TS  0 B = L C  = [0 1] D = 0 A could further be written as A = A1 + u*A2. 2.2 Linearized Model As the boost converter is a nonlinear system and PID control is always applied to linear systems. Therefore, to drive a boost converter using PID control, the converter has to be linearized near an operating point [7]. Let the converter is operating at duty ratio D, then by linearizing the state equation (3) around duty D, we get:      d IL + ˆiL     = D + dˆ VS + vˆ in + 1 − D + dˆ . VS + vˆ in − V + (VO + vˆ o L dt d ˆiL L = vˆ in + dˆ VO − (1 − D)ˆvo (7) dt C

d vˆ o vˆ o = (1 − D).ˆiL − − dˆ IL dt R

(8)

The duty to voltage transfer function of the converter can be given as:  G(s) =

vˆ c (s) dˆ (s)

 =− Vin =0

VO . 1−D

L s−1 R.(1−D)2 L.C L s2 + s+1 R.(1−D)2 (1−D)2

(9)

188

A. Verma et al.

îL

L

d̂VO +

v̂in

(1-d)î d̂ILL

(1-D)v̂o (1-D)îL

C

R

v̂o -

(a)

(b) L

îL

d̂VO

(1-D):1 + (1-d)î d̂ILL

v̂in

C

R

v̂o -

(c) îL

L/(1-D)2

d̂VO /(1-D) + (1-d)î d̂IL L

v̂in /(1-D)

C

R

v̂o -

(d) Fig. 4. (a) Average current circuit model of boost converter representing Eq. (7), (b) Average voltage circuit model of boost converter representing Eq. (8), (c) Transformer circuit model of boost converter, (d) Average DC circuit model of boost converter

3 Model Predictive Control MPC can be defined as a control technique, which predicts all possible future outputs of the system associated with the present inputs and outputs. The control signal, which leads to the best system response trajectory, will be given to the plant so that it can work optimally. The controller takes the best decision at each sampling instant by minimizing an objective function. Therefore, the technique is called an online control technique [8]. The best thing with MPC is that it can handle multiple constraints and disturbances effectively [9].

Comparative Study of PID and Model Predictive Control

Input States

189

Inputs from the converter

Initialize J, N, λ Initialization Block Store all possible input vectors of length N

End of the loop?

Yes

No Compute Vector cost J1

No

Input States Control Signal output to the converter

Is J1 < J ? Yes J = J1, uout = u1

Computing the objective function and selecting the optimum control signal

Fig. 5. Flowchart representing the working of MPC

For a boost converter, an objective function can be represented as: J = VO + α ∗ u Where, VO = VO (k) − VO (k − 1) α = Weighting factor u = u(k) − u(k − 1) The starting and ending points of the flowchart in Fig. 5 are the inputs and outputs of the controller respectively. As the controller runs on the basis of prediction of the future states of the converter corresponding to the present states and the control signals, the controller takes the inputs of all the states from the converter. Before prediction, it initializes the values of objective function, prediction horizon, and the weighting factor. After initialization, it starts a loop for predicting all the output states corresponding to all possible control signals and then updates the values of the objective function. At the end of the loop, it selects the control signal which results in the minimum value of the objective function and gives it as the output of the controller.

190

A. Verma et al.

4 PID Control PID control is a conventionally used control technique for power electronic systems. PID control uses the control to output transfer function to control a system. For nonlinear systems, the transfer function could not be directly found out. So, in the case of nonlinear systems, the system is first, linearized around an operating point, and then the plant transfer function is determined. The transfer function obtained in Eq. (9) is used for tuning the PID controller.

5 Simulations and Results The values of simulation parameters are: rL = 0.3 , L = 450 µH, C = 220 µF, R = 70 , T = 50 µs, P = 0.01, I = 0.3, D = 0, Input Voltage = 20 V. The MPC parameters are as follows: A1 = [1 −0.0056; 0.0114 1]; A2 = [0 0.0056; −0.0114 0]; B = [0.0056; 0]; Cost function J1 is initialized as J1 = inf; prediction horizon Nc = 5; Control Horizon = 1; Cost function constant lmd = 0.1; The tuning parameters of the PID controller are as follows: Proportional (P) gain = 0.01, Integral (I) gain = 0.3, Derivative (D) gain = 0 and Filter coefficient (N) = 100.

Fig. 6. Simulation of boost converter controlled by PID controller

Comparative Study of PID and Model Predictive Control

191

Fig. 7. Simulation of boost converter controlled by MPC

Implementation of PID and MPC Simultaneously Figure 8 shows the MATLAB simulation for the comparative study of MPC and PID control techniques with Boost converter. The simulation has two circuits, the upper one shows a boost converter controlled by MPC and the lower one is a boost converter with the same parameters, controlled by PID. Results of the Simulations In Fig. 9, the behavior of two similar boost converters is shown. Both the converters are controlled by PID and MPC control techniques separately. The voltage is boosted from 20 V to 40 V. For PID controlled Boost Converter, the voltage settling time is around 0.06 s (Fig. 9(a)) whereas the boost converter controlled by MPC has settling time around 0.004 s (Fig. 9(b)). From the Fig. 10, the inductor current is observed. Here, the final values of the inductor currents are almost the same but the inductor current of the PID controlled boost converter is noisier as compared to that of the MPC controlled boost converter.

192

A. Verma et al.

Fig. 8. Simulation for the comparative analysis of PID and MPC control techniques on boost converter

Figure 11 shows the waveforms of the output voltage waveforms of two similar boost converters controlled by MPC and PID controllers separately. Here, the reference voltages are changed for the two circuits simultaneously, and the performance of the controllers is observed. When the desired voltage reference is changed from 50 V to 40 V, both the controllers have almost similar performances. But at the time instant 0.36 s, when the voltage is varied from 60 V to 35 V, MPC as compared to PID controller.

Comparative Study of PID and Model Predictive Control

193

Fig. 9. (a) Transient response of the boost converter controlled by PID controller, (b) Transient response of the boost converter controlled by MPC

Fig. 10. (a) Input current of the boost converter controlled by PID controller, (b) Input current of the boost converter controlled by MPC

Figure 12(a) shows that the transient period of a MPC controlled circuit is smaller than the PID controlled circuit and thus must be preferred for applications where low response time is required. Figure 12(a) shows that a MPC controlled circuit handles output voltage variations much better than the PID controlled circuit and thus must be preferred for applications where the output voltage required by the converter may vary over a large range.

194

A. Verma et al.

Fig. 11. Comparative analysis of PID and MPC control techniques on boost converter

Fig. 12. Comparison of the transient behaviors of outputs from PID and MPC controlled boost converters (a) under normal conditions, (b) under varied Vref.

6 Conclusion From all the results above, some positives and negative points of both the control techniques are observed. Response from model predictive control settles faster and handles output voltage variations much better but it generates more ripples than a PID controller which could turn out to be a notable drawback. MPC must generally be preferred for the applications where operating conditions are time variable and the system is expected to adapt to heavy changes in parameters over a small period of time. PID control, on

Comparative Study of PID and Model Predictive Control

195

the other hand, must generally be preferred for applications where precision is more important than speed.

References 1. Steigerwald, R.L.: Power electronic converter technology. Proc. IEEE 89(6), 80–105 (2001) 2. Habetler, T.G., Harley, R.G.: Power electronic converter and system control. Proc. IEEE 89(6), 913–925 (2001) 3. Nikolay, L., Hinov, V., Dimitrov, V., Hranov, T.H.: Digitization of control systems for power electronic converters. In: International Scientific Conference Electronics (ET), Sozopol, Bulgaria, September 2020 4. Ang, K.H., Chong, G., Li, Y.: PID control system analysis, design, and technology. IEEE Trans. Control Syst. Technol. 13(4), 559–576 (2005) 5. Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based backstepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 6. Rodney, H.G. Tan, L., Hoo, Y.H.: DC-DC converter modeling and simulation using state space approach. In: IEEE Conference on Energy Conversion, Johor Bahru (2015) 7. Lejeune, R., Rugh, W.: Linearization of nonlinear systems about constant operating points. IEEE Trans. Autom. Control 30(8), 804–808 (1985) 8. Karamanakos, P., Geyer, T., Manis, S.: Direct voltage control of DC-DC boost converters using model predictive control based on enumeration. In: 15th International Power Electronics and Motion Control Conference, EPE-PEMC 2012 ECCE Europe, Novi Sad (2012) 9. Vazquez, S., et al.: Model predictive control: a review of its applications in power electronics. IEEE Ind. Electr. Mag. 8(1), 16−31 (2014) 10. Abbas, G., Samad, M.A., Gu, J., Asad, M.U., Farooq, U.: Set-point tracking of a DC-DC boost converter through optimized PID controllers. In: IEEE Canadian Conference on Electrical and Computer Engineering, Vancouver, BC, Canada (2016)

A Bidirectional DC-DC Converter Fed DC Motor for Electric Vehicle Application Mohd Faizan1(B) , Mohd Mehroz2 , Qazi Ramish1 , and Hina Nasir1 1 Aligarh Muslim University, Aligarh, U.P, India

[email protected] 2 AKTU, Lucknow, U.P, India

Abstract. Electric vehicles with regenerative braking are becoming more common in today’s environment, and this study is intended to push the DC motor to its limits due to the bidirectional DC/DC converter. In the case of electric vehicles DC/DC bidirectional converter are gaining a lot of traction. A battery, as well as traction chain, and individually stimulated motor are required for a DC/DC bidirectional converter. Control and DC/DC converter. The first step is to match the motor rated battery voltage in order to manage the braking scenario of the power flow, and the second is to use a DC/DC converter to perform two duties. The motor is accelerated and in normal mode, and the kinetic energy of the motor is transformed to electrical energy. Keywords: DC/DC converter · Generator mode · Motor mode · Vehicle dynamics

1 Introduction Electric vehicles (EVs) are the best solution to tackle an environmental problem while reducing the use of natural resources that are fossilized [1]. Bidirectional DC/DC converters have recently been created mass-produced for a variety of purposes, including electric automobiles. Electrical energy is used in battery electric vehicles. Between the battery side and the motor (BFEVs). The only way to reach zero emissions is for vehicles should be totally powered by batteries. Batteries having a high energy density are used on a big scale, easy maintenance, and features in terms of vehicle durability were used here on the field. [1, 2] methods for extending the range of electric vehicles’ driving range permits the battery voltage’s motor voltage level. In this mode, the vehicle kinetic energy of the battery is used to step down DC/DC [5]. The load as well as the braking power of battery electric vehicles were provided by this. The use of regenerative braking has been found to extend by up to 15%, the driving range is increased. It is not possible to brake a vehicle in a convectional or convectional mode. In the application of a DC motor DC connection in EVs, a DC/DC bidirectional converter is used. The speed of motor difference of a DC/DC bidirectional converters is utilized to reduce the motor voltage fluctuation. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 196–202, 2022. https://doi.org/10.1007/978-981-19-1742-4_16

A Bidirectional DC-DC Converter Fed DC Motor

197

2 The Structure of the Proposed System Figure 1 DC/DC bidirectional converter fed separately energised motor after this block diagram.

Fig. 1. The suggested method is depicted as a block diagram.

3 Description of the Circuit A DC/DC bidirectional converter is used to power an electric car with a battery circuit. The electric vehicle battery powered car that has been proposed is conceived and executed as a motion control stem with the help of a PI controller. The entire electric vehicle driving system is a net system to system comparison in terms of complexity, cost and size.

Fig. 2. With battery and dc motor, bidirectional DC-DC converter

This bidirectional DC-DC half-bridge provides power to the battery bank. In this configuration, In regenerative mode, the motor operates in both a buck-boost and a boost mode. Q1 and D1 act in buck mode, and current flow to the motor sides. Q2 and D2 act in boost mode, and electricity flows in the opposite direction in the battery [7]. 3.1 Operation of the Converter Figure 2 shows a BDDC motor running the mode of continuous conduction with regenerative braking. Q1 and Q2 of the four sub-converter MOSFETs are not consistent throughout time, as they try to switches the intervals 1 (T0 –T1 ), 2 (T1 –T2 ), 3 (T2 –T3 ), and 4

198

M. Faizan et al.

(T2 –T4 ) (T3 –T4 ). In this setupV1 represents the low battery voltage, and V2 represents the high voltage load side are used. The circuit drivers’ gate of time switches Q1 and Q2 steady state operation is described below [7, 8]. At time T0 , interval1 (T0 –T1 ) is turned on, D1 and D2 are reversed andQ2 is turned off the converter is in boost modes. At 2 (T–T) interval, the passage of time Q and Q have been disabled, while Q1 and D are both in conduction mode. The converters provided power to motor because converters works in boost mode, it does this to drive the motor by increasing the battery voltage ahead [9]. During the period, the buck converter is turned on and switch Q2 diode D1 , D2 reverse biased m1 ode of operation is closed (7). At 3 (T2 –T3 ) interval: At time T3 , the lower switch Q2 is set to low buck converter mode while the upper switch Q1 is turned on, However, a diode D1 , D2 closes the gap. At 4 (T3 –T4 ) interval: Both Q1 and Q2 are closed during this interval. With the low switch Q2 , begin conducting the body diode D2 (Fig. 3).

Fig. 3. Converter operating modes

3.2 The Battery Model There is a block in the Simple Power Systems library that connections a general parameterized dynamic model that represents the most common type of rechargeable battery. Figure 2 [9] depicts the similar circuit battery. We chose the lithium-ion battery for our project because of the benefits listed in the reference [1]. The discharge and charge of a lithium-ion battery are described by equations below: Model of discharge (i.∗>0) Q ∗ Q ← − ← −

∗

f .(i.t, i. , i) = E − K. i. − K. i.t + A.exp .(−B.i.t)

(1)

O−i.t

O−i.t ∗

f . (i.t, i. , i) = E i.t+0.1.Q

Model of discharge (i.∗>0) Q ∗ Q − i. − K.← − i.t +A.exp .(−B.i.t) − K.← Q−i.t

(2)

A Bidirectional DC-DC Converter Fed DC Motor

199

where: EBatt = nonlinear voltages E0 denotes a constant voltages. Exp(s) = the dynamics of an exponential zone (V). K = Resistance polarization or Polarization constant (Ah-1) (Ohms). i* = dynamics current at low frequencies (A) I = Current in the battery (A). it = stands for extracted capacity (Ah). Q = is the maximum battery capacity (Ah). A = stands for exponentially voltage (V). B = stands for exponentially capacity (Ah)−1 . The charge level of a battery (SOC) is 100% in the case of a fully battery charged and 0% for an empty battery. 3.3 DC-DC Converter A DC/DC converters with half bridge is powered by a battery bank. In buck mode, K1 and D1 is turned on, current flow to motor side. The system operates in buck mode for motor operation or boost mode for regenerative braking in boost mode. K2 and D2 are turned on in boost mode, and the power flows are inverted [10]. A smoothing inductor must be connected in series with the motor to reduce residual ripple in the current. 3.4 DC Machine Model There are two modes of operation for the machine block: generating and motoring. The mode of action (positive for moving, negative for braking) is determined by the mechanical torque. Four equations can be used to translate the general functioning of a mechanical load linked to a separately energized dc machine: |j. dd ω.t = T .e− T .L− B.m. ω − T .f |U = E + Ri + L. ddi.t |T .e = K.i |E = K.ω

(3)

ω, Te , J, TL , Tf and Bm represent speed, Machine-produced electromagnetic torque, rotating-part inertia moment, viscous friction torque, external torque and fluid friction coefficient, respectively. The motor’s voltage, counter-electromotive force, absorbed current, internal resistance, and inductance are all factors to consider by U, I, E L, and R, respectively. The number K is a fixed number (the constant flux).There is no field current in the model of a permanent DC machine since the magnets are in charge of determining the excitation flux.

4 Simulation Results The motorization is made more sensitive by the slope to create attempts in order to keep the electric traction chain running smoothly requirement, and the aerodynamic force is

200

M. Faizan et al.

Fig. 4. Battery regeneration mode SOC, voltage, and current

Fig. 5. Battery generation modes voltage, current and SOC

A Bidirectional DC-DC Converter Fed DC Motor

201

Fig. 6. Regeneration modes of motor speed, field current, armature current, electric torque.

exactly proportional to the vehicle’s speed, as shown in Fig. 4. Figures 5 and 6 depict the battery’s current and SOS, respectively. As previously stated, the operating mode determines the battery’s current direction, discharge, and charge (Table 1). Table 1. Table value Parameters

Output values

SOC battery

99.987%

Current battery

4.02 A

Voltage battery

49 V

Voltage armature

41 V

Voltage field

301 V

DC motor speed

277 rad/ s

Current armature DC motor

6.6 A

Current area DC motor

0.34 A

DC, electrical motor torque

1.397 Nm

DC motor load speed

242.8 rad/ s (continued)

202

M. Faizan et al. Table 1. (continued) Parameters

Output values

DC motor load current armature

8.139 A

DC motor electrical load torque

6.297 Nm

5 Conclusion This chapter explains how to use or battery electric vehicles, a bidirectional converter is required, as well as how to use a DC/DC bidirectional converter to operate a DC motor. Simulation in the MATLAB/SIMULINK environment was used to verify system performance. The electric vehicle’s topology is a simple circuit that increases range of motion and is easy to grasp, and work on the DC/DC bidirectional converters systems in the future is recommended, and the EV’s performances may be examined using MATLAB/Simulink.

References 1. Pany, P., Singh, R.K., Tripathi, R.K.: Bidirectional DC-DC converter fed drive for electric vehicle system. Int. J. Eng. Sci. Technol. 3(3), 101–110 (2011). https://doi.org/10.4314/ijest. v3i3.68426 2. Zhang, J., Lai, J.-S., Kim, R.Y., Yu, W.: High power density design of a soft-switching high-power bidirectional DC-DC converter. IEEE Trans. Power Electron. 22(4), 1145–1153 (2007) 3. Zhang J., Lai, J.-S., Yu W.: Bidirectional DC-DC converter modeling and unified controller with digital implementation. In: Applied Power Electronics Conference and Exposition, APEC pp.1747–1753 (2008) 4. Yang, L.S., Liang, T.J.: Analysis and implementation of a novel bidirectional DC-DC converter. IEEE Trans. Ind. Electron. 59(1), 422–434 (2012) 5. Bertoluzzo, M., Buja, G.: Development of electric propulsion system for light electric vehicles. IEEE Trans. Ind. Inf. 7(3), 428–435 (2011) 6. Khaligh, A., Li, Z.: Battery, ultracapacitor, fuel cell, and hybrid energy storage system for electric, hybrid electric, fuel cell and plug-in hybrid electric vehicles: state of the art. IEEE Trans. Veh. Technol. 59(6), 2806–2814 (2010) 7. Chan, C.C.: The state of the art of electric and hybrid vehicles in the state of the art. Proc. IEEE 90(2), 247–275 (2002) 8. Jain, P.K., Kang, W., Soin, H., Yusuf, X.: A large load and line voltage zero through pool DC/DC switching topology in 2002 to analyze and design thinking, power electronics. In: IEEE Transactions on September, vol. 17, no. 5, pp. 649–657 (2022) 9. Priyaratnam, Jain, A., Verma, N., Shaw, R.N., Ghosh, A.: Modeling of electric vehicle charging station using solar photovoltaic system with fuzzy logic controller. In: Applications of AI and IOT in Renewable Energy, Academic Press, pp. 151–167 (2022). https://doi.org/10. 1016/B978-0-323-91699-8.00008-5 10. Caricchi, F., Crescimbini, F., Noia, G., Pirolo, D.: DC link voltage control and electric motor drives vehicles. In: Applied Power Electronics Conference devoted to the Prime Minister of regenerative braking in a bidirectional DC-DC converter experimental study, vol.1, pp. 381– 386 (1994)

PV-WIND Hybrid System Based Cuckoo Search Maximum Power Point Tracking Algorithm V. Dhanunjaya1 , K. Vijaya Bhaskar Reddy1(B) , S. Vijaya Kumar2 , and P. Venkata Kishore3 1 Bvrit, Narsapur, India [email protected] 2 Dr. K V Subba Reddy Institute of Technology, Dupadu, AP, India 3 St. Peters Engineering College, Hyderabad, India

Abstract. The main goal of this project is to find a PV-Wind hybrid system’s Maximum Power Point using Cuckoo search MPPT. The PV system and wind turbine based on a permanent magnet synchronous generator are both part of the proposal. Particle swarm optimization (PSO) and traditional MPPT algorithms (incremental conductance, P and O, and particle swarms) fail to keep up with rapidly changing environmental conditions. For this reason, an evolutionary algorithmic technique known as the Cuckoo Search Algorithm (CSA) is used to track down the optimal amount of power. The DC-DC step-up boost converters are utilized to apply to WIND and PV, with CS artificial intelligence method being used for each. The DC/DC converters increase the voltage of the 2 source after maximum power point tracking method suggested to the DC link or DC loads. The simulation results indicate the PV/WIND with CSA MPPT with constant irradiance and step changes in the irradiance. Keywords: Incremental conductance · Perturb and Observe · PSO · Cuckoo Search Algorithm · Incremental conductance (IC) · PV · Wind · DC-DC boost converter

1 Introduction It’s no longer a possibility to generate all of your electricity from coal or nuclear power plants. There has been a rise in the usage of solar photovoltaic’s (PV) for electricity generation due to depletion of fossil fuel supplies and growing concern about the environmental impact of fossil fuel consumption. The electricity generated by photovoltaic (PV) systems is a low-maintenance, environmentally friendly option. PV modules’ ability to generate electricity is dependent on a variety of variables, including temperature, solar irradiation, and amount of shade [1]. The voltage-current (V-I) characteristics of PV modules are non-linear. PV modules have a voltage-power characteristic curve (V-P) with a single maximum power point (Pmax). MMP fluctuates in response to changes in the surrounding environment. As a result, the power given to the load is reduced to its highest possible level. “In order to match the characteristics of the PV module with the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 203–214, 2022. https://doi.org/10.1007/978-981-19-1742-4_17

204

V. Dhanunjaya et al.

load and minimize power loss, MPPT is utilized” [1, 2]. When a DC-DC converter is used between the load and the PV modules in a PV system, the MPPT controller comes in useful. Several MPPT strategies are documented in the literature for use in various photovoltaic applications to extract the maximum power from the PV modules. There are a number of MPPT approaches that are routinely utilized, including Perturb and Observe, Fractional Short Circuit Current, and Incremental Conductance. Fuzzy logic, Artificial Neural Networks (ANN), and Particle Swarm Optimization (PSO) are some of the more advanced soft computing-based MPPT algorithms. “MPPT algorithms differ in a variety of ways, including: effectiveness, complexity, the number of sensors required, the ease of hardware implementation, and the speed of convergence, among others” [3]. Since it’s simple, easy to apply, and reliable, the P&O MPPT method is the most commonly utilized MPPT method. However, there are two fundamental drawbacks to this approach: To begin with, when one gets closer to the MPP, the output power oscillates endlessly, resulting in a drop in energy yield. It also loses energy because it can’t keep up with rapidly changing irradiance, leading the operating point to move farther from the Maximum power point and lose efficiency Soft computing-based MPPT approaches are becoming prominent as a solution to these issues [4]. It is possible to track MPP using two simple meta-heuristic algorithms, such as Particle Swarm Optimization (PSO) or Cuckoo Search (CS). GMPP can be extracted using MPPT algorithms even when there is some shade (PSC). GMPP location does not have to be identified using these approaches [5]. “Rezk et al.” [6] Looked at PSO and CS data. MPP tracking strategies were evaluated in MATLAB with and without partial shading. In comparison to incremental resistance, the convergence of PSO and CS to MPP was much rapid. Because it required less time to track, the CS-based tracker beat out the PSO. To compare the MPPT technique’s findings with those of other MPPT methods, Mosaad et al. [7] used computer simulations (CS). With different conditions, it was discovered that CS had the most power when compared to IC and ANN. Upon approaching MPP, the output power remained stable. Traditional MPPT approaches (IC, P&O) in PV systems were compared to PSO by Koad and Zobaa [8], who found that the former was superior. Cuk converter in MATLAB is used to implement these MPPT algorithms in order to compare their accurateness, speed of tracking, price. In simulations, it was discovered that PSO has the ability to precisely track MPP regardless of the situation. In comparison to other approaches, it has a higher tracking efficiency. Comparing it to IC and P&O, it also has a faster convergence time and is easier to implement. Cuckoo search for MPP tracking in a photovoltaic scheme were carried out by Ahmed and Salam [9]. The outcomes were compared to those of the standard P&O procedure. Simulated results revealed that even in altering atmospheric conditions with no steady-state oscillations in progress, the CS technique follows the MPP fast and accurately. Despite the fact that renewable energy is a new concept, it has unpredictable results. Its erratic supply necessitates the use of backup power sources like batteries [10]. Renewable energy resources are intermittent, therefore adopting just one result in larger components, additional operational and lifetime costs, and other drawbacks [11]. A hybrid energy system compensates for the shortcomings of each particular energy resource by combining two or more types of energy resources. Consequently, the design objectives for a hybrid power system are to minimize the

PV-WIND Hybrid System

205

cost of power production, minimize the cost of obtaining electrical energy from the grid and to reduce emissions while also lowering total life cycle costs [2, 10, 11]. The “hybrid renewable energy system” (HRES), which combine two or more RESs to produce more stable and higher-quality power, is becoming more popular because the sources complement one another. This study proposes and tests a hybrid PV-WIND system based on Cuckoo Search MPPT in the Simulink environment.

2 Proposed System The suggested system consists PV model with boost converter and wind model with boost converter and these DC/DC converters controlled by use CSA-based maximum power point tracking the overall block diagram shown in Fig. 1. Finally, use artificial intelligence-based approaches like the CSA-based maximum power point tracking control algorithm to determine the maximum power point from all the individual modules. Better results and less variation around the maximum power point are obtained using this method (MPP).

Fig. 1. Block diagram of a suggested hybrid system

206

V. Dhanunjaya et al.

2.1 Modeling of PV Panel The PV (Photo Voltaic) cell equivalent circuit is shown below. The solar cell is treated with the current source and diode in parallel connected.

Fig. 2. Solar PV module equivalent circuit

From Fig. 2, apply KCL (Kirchhoff’s Current Law), then Iph = Id + Ish + I

(1)

I = Iph −(Ish + Id )

(2)

We have [6] following equation for Solar cell current, I = Iph − Io [eq(V+I.Rs)/nkT) −1(V + IRs )/Rsh

(3)

where VT represents Terminal Voltage Iph denotes isolation current V represents the cell voltage I represent cell current Io denotes reverse saturation current Rsh is Shunt Resistance Rs denotes Series Resistance q represents elementary charge n denotes diode ideality factor T represents absolute Temperature K denotes Boltzmann’s constant 2.2 Wind Turbine Modeling Wind generator turbine rotates which is coupled to an alternator generates electrical energy. Electric power magnitude [6] of speed changes with a given turbine is described as: pw =

1 m.vw3 = 2 t

1 3 2 ρ.A.d .vw

t

=

1 ρ.A.vw3 2

(4)

PV-WIND Hybrid System

207

where Vw = wind speed (distance/time) (m/s) Pw = Wind Power (W) ρ = Air density (kg/m3 ) A = Area swept by the turbine blades (m2 ) d = radius of the swept area of blades (m) m = mass of the air m = air density X volume = ρ.A.d (Kg) The generated mechanical power is expressed by [5] pm = pw .cp (λ, β) =

1 ρ.A.vw3 .cp (λ, β) 2

(5)

Here Cp denotes power coefficient λ represents tip speed ratio of the rotor blade tip speed to wind speed β represents the pitch angle The power coefficient of the turbine is expressed by cp = c1 (

−c5 c2 − c3 β − c4 )e λi + c6 λ λi

(6)

where, 0.035 1 1 − 3 = λi λ + 0.08β β +1

(7)

And also Cp =

Pm ; Cp < 1 Pw

Pm = Cp .

ρ.A 3 v 2 w

(8) (9)

The power produced Pm depends on the magnitude of Cp . The Cp is defined as a ratio of electric power generated by wind generator turbine to mathematical wind generator power. TSR denotes relation among wind speed angular speed which is expressed by λ =

ω·d Vw

(10)

Here ω is the rotor speed expressed in rpm. For a gearless wind turbine, the mechanical torque is expressed as Tm = Pm

d R Pm = 1/2 p. A. Cp (λ, β) = λVw Vw W

(11)

208

V. Dhanunjaya et al.

3 Boost Converter Boost converter is also known as step-up DC/DC converter. The boost converter contains inductor, MOSFET, diode and capacitor and circuit of boost converter as indicated in Fig. 3.

Fig. 3. DC/DC step-up (Boost) converter

There are 2 operating condition. When Switch is “ON” Position the current at the inductor can be written as Eq. (12). iL (ON ) =

0300 s

0

Hydro turbine

>6000 s

0%

10000 s ∼ 10 s =

∞

BESS WTG

0 control weighting matrix of p × p As LQG controller is implemented on a reduced order system, there is no direct systematic approach to select the elements of Q and R matrices. In order to finding a suitable controller act, trial and error optimization technique can be adopted for selecting Q and R.Small value of R helps to speed up the dynamic response of the controlled structure.The system’s states are penalized through the elements of Q whereas the control inputs are penalized with R.LQG is consists of linear-quadratic-regulator (LQR) and linear-quadratic-estimator (LQE) or state observer and according to separation principle these two parts of LQG can be designed separately. Here, Kalman filter (KF) is considered as LQE, whose mathematical model is given in Eq. (7). x˙ (t) = Ax(t) + Bu(t) + Gf (y(t) − Cx(t)),









here, Gf is the filter gain and x(t) denotes the estimated states of the plant.

(7)

Design a Robust Control Platform for Grid

273

Plant Disturbances and Uncertainties

Reference Input

Control Input, u

Plant Dynamics without conventional PSS

Proposed LQG

Proposed Controller for Optimal Oscillation Damping

Fig. 4. Schematic arrangement of oscillation damping with the proposed control approach.

For our test system the state vector, control input and disturbance are defined as T

 x = ω δ E q Efd ; u = [ω]T ; Np = Tm ; the elements of the A, B, C, Q and R matrices are dependent on the system operating points i.e., will be updated every time the operating conditions are changed. Figure 4 shows the schematic arrangements of the proposed LQG controlled system for maintaining the system stability.

4 Performance Investigation of Proposed Controller This section verifies the effectiveness of the proposed controller to damp out low frequency oscillation during abrupt demand variation considering system uncertainties and disturbances. For robustness analysis purpose proposed controller is implemented in four different test plants. The numerical values of the test plant upon which the controller implementation is dependent are given in Appendix. After controller implementation the results are compared with conventional PSS, bat algorithm optimized conventional PSS (BA-CPSS), fractional order proportional-integral-derivative PSS (FOPID-PSS) and hybrid Firefly Algorithm-Particle Swarm Optimized (hFAPSO) Interval Type-2 Fractional Order Fuzzy PID based PSS (hFAPSO-IT2FOPID-PSS) based on some important dynamic response characteristics. Finally, a Frequency domain analysis is also carried out to prove the effective damping capability of the proposed controller. 4.1 Test Plant I Here, the operating conditions of the system are taken as real power, P = 0.75 pu and line reactance, Xe = 0.2 pu. The compared response for 5% step change in load is shown in Fig. 5(a).

274

F. Tabassum et al.

4.2 Test Plant II At this operating condition the real power, P; reactive power, Q and line reactance, Xe becomes 1, 0.3 and 0.2 Pu respectively. Figure 5(b) represents the compared responses of different controllers.

Shaft Speed Deviation (pu)

-3

10

6

No PSS Conventional PSS Proposed Controller

4 2 0 -2 -4 0

1

2

-3

Shaft Speed Deviation (pu) Shaft Speed Deviation (pu)

6

7

No PSS Conventional PSS Proposed Controller

4 2 0 -2 -4 -6 0

1

2

-3

3 4 Time (Seconds)

5

6

7

(b)

10

No PSS Conventional PSS Proposed Controller

5

0

-5 0

1

2

-3

Shaft Speed Deviation (pu)

5

(a)

10

6

3 4 Time (Seconds)

8

3 4 Time (Seconds)

5

6

7

(c)

10

No PSS Conventional PSS Proposed Controller

4

0

-4 0

1

2

3 4 Time (Seconds)

5

6

7

(d)

Fig. 5. (a)–(d) Shaft speed response of different controllers for test plant I, II, III and IV respectively.

Design a Robust Control Platform for Grid

275

4.3 Test Plant III For this point it is considered that the real and reactive power, P and Q become 1, 0.3 and due to adding an extra transmission line the line reactance, Xe becomes 0.4 pu. for different control schemes the compared responses are shown by Fig. 5(c). Table 1. Performance parameters of different control techniques.

Where, (–) indicates the absence of data in the respective references. 4.4 Test Plant IV For this operating point the real power, P; reactive power, Q and line reactance, Xe is considered as 1, −0.3 and 0.4 pu respectively. For the test plant the compared responses of different controllers are in Fig. 5(d).

F. Tabassum et al. Phase (deg) Magnitude (dB)

276

-20

No PSS Proposed Controller

-40 -60 90 0 -90

1

Phase (deg) Magnitude (dB)

10

Frequency (rad/s) (a) No PSS Proposed Controller

-20 -40 90 0 -90

1

10 Frequency (rad/s) Phase (deg) Magnitude (dB)

(b) 20 No PSS Proposed Controller

0 -20 -40 90 0 -90 1

10 Frequency (rad/s)

Phase (deg) Magnitude (dB)

(c)

-20

No PSS Proposed Controller

-40 -60 90 -90 -270 0 10

1 10 Frequency (rad/s)

(d)

Fig. 6. (a)–(d) Frequency responses for test plant I, II, III and IV respectively references

With the increasing value of external system reactance (test plant III and IV) the value of constant K5 becomes negative (given in Appendix). This negative value introduces a negative damping torque and facilitates system instability [8]. Some of the important dynamic response characteristics of the aforementioned four test plants for different

Design a Robust Control Platform for Grid

277

control strategies are summarized in Table 1. Only 0.51 and 0.61 s is needed for the proposed controller for operating conditions I and II to compensate the speed deviation in generator shaft without having any steady-state error. This response is 4.47 and 3.97 times faster than the bat algorithm optimized conventional PSS (BA-CPSS). At oscillatory (test plant III) and unstable (test plant IV) conditions the proposed control scheme has proved its superiority by stabilizing the plant within 1.32 and 1.1 s with no offset at shaft speed which is 1.6 and 2.2 times quicker than the BA-CPSS approach. Frequency domain analysis is also carried out to justify the effective damping capability of the proposed controller. The resonant mode of −16.4; −13.6; 12.6 and −23.6 dB at frequency of 6.7; 6.32; 5.75 and 5.64 rad/s for four different plants are fully eliminated using the proposed controller as shown in Fig. 6(a)–(d). It has also increased the system bandwidth. From the graphical and numerical point of view it is evident that proposed controller exhibits the best result ensuring the robust control of the plant.

5 Conclusion In this brief, an LQG based optimal control scheme is proposed to compensate the low frequency oscillations in the power generating system considering the disturbances and uncertainties. The proposed scheme eliminates the shaft speed deviation of synchronous generator within 0.51, 0.61, 1.32 and 1.1 s for the four different operating points without the conventional PSS. The compared responses prove the robustness of the proposed control scheme. In future, the proposed controller can be designed for eliminating high frequency oscillations and for a system having distributed energy resources.

6 Nomenclatures KD = damping torque coefficient in pu speed deviation H = Inertia constant of synchronous machine in MVA ω = pu speed deviation of shaft δ = Rotor angle deviation in electrical radian ω0 = rated sped of rotor in electrical radian/second Tm = Deviation in mechanical input torque Vref = Deviation in reference terminal voltage Efd = Deviation in exciter output voltage Eq = Deviation in field transient voltage of q-axis  = Open circuit transient time constant of d -axis Tdo K1 ,., K6 = K constants of power system dynamics KA = Gain of voltage regulation loop TA = Time constant of voltage regulation loop KPSS = Gain of PSS Tw = Filter Time constant in washout block T1 , T2 = Time constants of phase compensation block VPSS = Output voltage deviation of PSS s = Laplace operator xd = Synchronous reactance of d -axis

278

F. Tabassum et al.

xq = Synchronous reactance of q-axis xd = Transient reactance of d -axis VB = Infinite bus bar voltage

Appendix Value of K constants (K1 ,..,K6 ):

Test plant Test Plant I

Test Plant II

Test Plant III

Test Plant IV

Constants 1 2 3 4 5 6

1.9242 1.2663 0.2304 2.8197 0.0348 0.2185

1.7271 1.5799 0.2304 2.6385 0.0128 0.2844

1.3801 1.1512 0.2954 1.9225 -0.0132 0.4748

1.1108 1.4158 0.2954 2.3643 -0.1454 0.3264

Synchronous machine’s parameters:  = 6.84 s, K = 100, T = 0.02 s, xd = 1.97 pu, xq = 1.9 pu, xd = 0.3 pu, Tdo A A ω0 = 2π × 50 rad/s, H = 6.5, VB = 1 pu. PSS parameters: KPSS = 2, Tw = 10 s, T1 = 0.5 s, T2 = 0.05 s.

References 1. Mohandes, B., Abdelmagid, Y.L., Boiko, I.: Development of PSS tuning rules using multiobjective optimization. Int. J. Electr. Power Energy Syst. 100, 449–462 (2018) 2. Soliman, H.M., Yousef, H.A.: Saturated robust power system stabilizers. Int. J. Electr. Power Energy Syst. 73, 608–614 (2015) 3. Shafiullah, M., Rana, J., Shahriar, M., Zahir, M.: Low-frequency oscillation damping in the electric network through the optimal design of UPFC coordinated PSS employing MGGP. Measurement 138, 118–131 (2019). https://doi.org/10.1016/j.measurement.2019.02.026 4. Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based backstepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 5. Ray, P.K., et al.: A hybrid firefly-swarm optimized fractional order interval type-2 fuzzy PIDPSS for transient stability improvement. IEEE Trans. Ind. Appl. 55(6), 6486–6498 (2019)

Design a Robust Control Platform for Grid

279

6. Chaib, L., Choucha, A., Arif, S.: Optimal design and tuning of novel fractional order PID power system stabilizer using a new metaheuristic Bat algorithm. Ain Shams Eng. J. 8(2), 113–125 (2017) 7. Demello, F.P., Concordia, C.: Concepts of synchronous machine stability as affected by excitation control. IEEE Trans. Power App. Syst. 88(4), 316–329 (1969) 8. Bhadu, M., Senroy, N., Kar, I.N., Sudha, G.N.: Robust linear quadratic Gaussian-based discrete mode wide area power system damping controller. IET Gener. Transm. Dis. 10(6), 1470–14278 (2016)

Dynamic Power Flow Control Using Dual-Input Single-Output Non-Isolated DC-DC Converter for Renewable Energy Applications Saikat Das(B) , Nirakar Nayak, and Amritesh Kumar Department of Electrical Engineering, NIT Silchar, Silchar, Assam, India [email protected], [email protected]

Abstract. This paper presents a non-isolated dual-input single-output (DISO) dc-dc converter capable of integrating input sources of distinct V-I characteristics for dynamic power flow control of renewable energy sources. The structure of the proposed converter is incorporated with the conventional boost and buck-boost converter having two semiconductor switches and three diodes. The proposed converter has a higher voltage gain than the conventional converter. Renewable energy sources like solar PV, fuel cells (FC) can be connected as inputs to the converter. In this paper, a closed-loop control strategy has been presented for maintaining the constant output voltage irrespective of variation in the sources. Four operating modes of this proposed converter are analyzed with switching pulses. Further, the dynamic response of the converter with load fluctuation has been discussed in the paper. The proposed converter under different modes of operation has been validated using MATLAB/Simulink.

Keywords: DC-DC multiport converter (MPC) Sources (RES) · Hybrid Energy Systems (HES)

1

· Renewable Energy

Introduction

Drastically increase in population and the growth of the industrial sector enforce a large amount of demand in the electric sector. However, our traditional conventional method of producing power is insufficient to fulfill our requirement because the raw material of this plant-like coal, natural gas, is not available in ample amount. It is depleting day by day and has an adverse effect on the environment. To meet this enormous demand without harming the environment, renewable energy sources (RES) like solar PV, fuel cell, and wind turbine-based hybrid energy systems (HES) play a crucial role. In HES, more than one RES with the same or distinct V-I characteristics is connected to supply the load. N. Nayak and A. Kumar—Member IEEE. c The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022  S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 280–291, 2022. https://doi.org/10.1007/978-981-19-1742-4_24

Dynamic Power Flow Control

281

A suitable combination of RES such as solar PV/fuel cell, battery/ supercapacitor, wind/solar PV is generally used to build an appropriate HES. HES provides high reliability, improves transient response, high efficiency, high energy density. The use of multiple converters to connect various RES is not suitable. This causes the usage of more number semiconductor devices resulting in high switching loss, high cost, and low efficiency. Therefore suitable interfacing multiport converter has attracts the attention of researchers. HES with solar PV has low output voltage, and it depends on the irradiation of the sun and ambient temperature [1,2]. For high voltage gain, conventional boost and buck-boost converters require a high duty ratio, resulting in high current stress in boost switches [3]. Therefore require an MPC with high voltage conversion ratio that meets the desire output level. MPC is generally two types isolated and non-isolated. In isolated MPC, a coupled inductor or transformer with high frequency is used to gain high ratio. This transformer offers electrical isolation between input side and output side, but the problem is complexity in control and high expenses compared to nonisolated MPC. In non-isolated MPC, traditional single input port converters like boost, buck-boost, SEPIC, Cuk converter with multiple passive elements are combined to construct a suitable new topology. In recent trends, researchers are innovating new topologies to connect multiple RES. In [4], proposed a four-port non-isolated symmetrical bipolar output converter. In this converter, the bipolar output is achieved by combining a threeport SEPIC converter with unipolar output and a three-port Cuk converter with unipolar output. This converter has one unidirectional input port where PV has connected and another bi-directional input port where the battery has connected. It having two switching configurations; one is for charging, and another is for discharging. In [5], proposed a bridge-type non-isolated dc-dc converter for HES integration. This converter uses a supercapacitor and battery bank as input sources. It has four semiconductor switches. This converter has the capability of feeding power to the load and from the load to the source. The main drawback of this converter is that it has two bi-directional input ports. So, RES cannot be used as input, only energy storage devices mounted as input. Different topology-based dual input dc-dc converters have been introduced in [6–8]. An MPC based on switched boost action is presented in [9]. It shows how multiple outputs can be supplied by customized boost topology. A bi-directional threeport dc-dc converter topology is presented in [10], where a multivariable control scheme is implemented to control the converter. In [11,12], a non-isolated dc-dc converter is introduced. In [13], a switched capacitor converter based MPC is proposed. This converter is constructed by incorporating a bi-directional PWM, series resonance converter, and a ladder-type switched capacitor converter. It operates in three operating modes, SIDO, SISO, and MPPT mode. The control of this converter is very complex that is the main drawback of this converter.

282

S. Das et al.

Fig. 1. Circuit model of the dual-input single-output dc-dc converter.

This paper proposes a non-isolated DISO dc-dc converter with a closed-loop control scheme for maintaining constant dc voltage irrespective of changes in the load. The merit of this converter is high gain, fewer number semiconductor switches, and high efficiency. The proposed converter is constructed by combining conventional boost with buck-boost converter and having two unidirectional input ports and one output port. It is preferred to connect solar PV at the buck-boost configuration and the fuel at boost configuration to minimize the current ripple of FC [1]. This converter has the features of modulation switching as it has no restriction on switching. Therefore, if implement solar PV is an input, maximum power point tracking can easily be equipped. The aim of this research is to present a non-isolated DISO dc-dc converter for RES integration. The background of research, current research work, and the contribution of this study are discussed in Sect. 1. In Sect. 2, the circuit configuration model and its operating modes are introduced. Section 3 deals with the control strategy of the proposed converter. Sections 3.1 and 3.2 represents design parameters and Simulink results to validate the proposed operation. At last, the conclusion part is presented in Sect. 4.

2

Circuit Configuration

The suggested converter topology and associated circuit operation modes are described in this section. Figure 1 illustrates the circuit model of the DISO converter. This MPC is made up of two inductors L1 and L2 , two capacitors C1 and C2 , two switches S1 and S2 , three diodes D1 , D2 , and D3 . The input ports of the converter are connected with two sources V1 and V2 . Control operation of this converter is simple compared to other MPCs and generally done by regulating the duty of two switches. Detailed control strategy has been discussed in Sect. 3. This converter can efficiently perform in a single input mode (only V1 is available) without affecting the output voltage level. Four switching modes of this converter are explained in the following section. Different operating modes of the DISO converter with current flow paths are illustrated in Fig. 3.

Dynamic Power Flow Control

283

Fig. 2. Shows the various signal & switching pattern of carrier signal (C1 ); phaseshifted carrier signal (C2 ); gate pulse of switches (GS1 & GS2 ); inductors current (iL1 & iL2 ); capacitor current (iC1 & iC0 ); capacitor voltage (VC1 ).

284

S. Das et al.

Mode 1 [0 < t < t1 ]: In this mode, the switch S1 is turned on, and S2 is turned off, as shown in Fig. 3(a). Buck-boost side inductor L2 gets demagnetized to capacitor C1 through diode D2 . While the boost side inductor L1 gets magnetized by the source V1 and the stored energy of capacitor C1 . Diode D3 reverse biased, and the output side capacitor C0 delivered its stored energy to the load. Mode 2 [t1 < t < t2 ]: In this mode, both switches S1 and S2 are turned on, and diodes D2 and D3 are non-conducting state. Via the switch S2 , the inductor L2 is magnetized by the source V2 . While the capacitor C1 and the source voltage V1 both magnetized the inductor L1 . The output capacitor C0 still feeds the load, as illustrated in Fig. 3(b). Mode 3 [t2 < t < t3 ]: This mode is identical to mode 1 as depicted in Fig. 3(a). Mode 4 [t3 < t < t4 ]: Both switches S1 and S2 are turned off, and the diode D1 , D2 , and D3 are forward bias in this mode, as shown in Fig. 3(c). The inductor L2 delivered its stored energy to the capacitor C1 through diode D2 So, C1 getting charged. At the same time, the input source V1 and charged inductor L1 delivered power to the load and output capacitor C0 . So, the capacitor C0 is getting charged. Figure 2 shows the variation in inductor current and capacitor charging-discharging periods with respect to different operating modes. Apply volt-second balance on inductors L1 and L2 to get the converter’s average output voltage and ampere-second balance on capacitors C1 and C0 to get the average inductors current. Volt-Second balance on inductors L2 and L1 : L2 :

L1 :

V2 d2 T + (−VC1 ) (1 − d2 ) T = 0 d2 V 2 VC1 = (1 − d2 )

(V1 + VC1 )d1 T + (V1 − V0 ) (1 − d1 ) T = 0 V1 + d1 VC1 V0 = (1 − d1 )

(1)

(2)

where d1 and d2 are the duty ratio of two switches S1 and S2 , respectively, V1 and V2 are the input sources, VC1 is the voltage across capacitor C1 and the switching period of the converter is T. By substituting the value of VC1 form Eq. (1) in Eq. (2), the average output voltage of the converter can be obtained as:   d2 V2 V1 + d1 1−d 2 V0 = (3) 1 − d1 Ampere-Second balance on capacitors C1 and C0 : C1 :

iL2 (1 − d2 ) T + (−iL1 ) d1 T = 0 iL2 d1 = iL1 (1 − d2 )

(4)

Dynamic Power Flow Control

285

Fig. 3. Different operating modes of DISO converter. (a) Mode 1 & 3. (b) Mode 2. (c) Mode 4.

286

S. Das et al.

C0 :

(iL1 − i0 ) (1 − d1 ) T − i0 d1 T = 0 i0 iL1 = (1 − d1 )

(5)

From (4) and (5), the average inductor current of L2 obtained as: iL2 =

d1 i0 (1 − d2 ) (1 − d1 )

(6)

Where i0 is the output current across the load terminal, and iC1 is the current flowing through the capacitor C1 . According to Eq. (3), the converter’s output voltage depends on both voltage sources V1 and V2 , and the duty of two switches d1 and d2 . By regulating the duty, higher voltage gain of the converter can be achieved as our requirement.

3

Control Strategy

In this proposed converter, various modes of operation are entirely reliant on the switching strategy. Two switches with different approaches are used to generate desire switching pulse at a constant switching frequency. In switch S2 , a constant pulse is provided, and it is created by comparing a constant reference signal with a phase-shifted carrier signal. The duty of that constant pulse (d2 ) is 0.3. For the closed-loop operation of the controller to maintain constant output, a controlled switching pulse is provided to switch S1 . The output voltage is first compared to the reference voltage to get an error signal (e1 ). This error is now sent to a PI controller. The PI controller’s output is then compared to the inductor current (iL1 ) before being passed on to another controller. After that, the output is compared to the carrier signal to get a controlled pulse for switch S1 . PI controller reduces the steady-state error and also has a good transient response. As a carrier signal, a sawtooth waveform with a switching frequency of 5 kHz is selected. Figure 4 illustrates the block diagram of the closed-loop control scheme of the DISO dc-dc converter.

Fig. 4. Closed-loop control strategy of the proposed DISOC dc-dc converter.

Dynamic Power Flow Control

3.1

287

Design Considerations

Suitable values of inductors and capacitors are required to design the proposed DISO dc-dc converter. For continuous conduction mode (CCM) of operation of the converter, inductors must be satisfied the following inequality: iL ≥

ΔiL 2

(7)

iL is the inductor current, and ΔiL is the ripple in inductor current. The inductor L2 will be charged for the duration of d2 T at that time; the voltage across the inductor L2 will be V2 , as depicted in Fig. 3(b). So, the current ripple in inductor L2 : V 2 d2 (8) ΔiL2 = L2 fS Similarly, the current ripple in inductor L1 :   d1 V1 + d2 VC1 ΔiL1 = L1 fS

(9)

where VC1 is the voltage across capacitor C1 as in Eq. (1). The average value of inductors currents is shown in (5), (6). Therefore, the marginal value of inductors can be evaluated as follows:    2 d2 V2 (1 − d1 ) RL V1 d1 + d1 1−d 2    L1 C = (10) d2 V2 2fS V1 + d1 1−d 2 2

L2 C =

V2 d2 (1 − d1 ) (1 − d2 ) RL   1 (d2 V2 ) 2fS d1 V1 + d(1−d 1)

(11)

From analyzing the operating modes as in Fig. 3, capacitor value can be evaluated as follow: C0 =

d1  RL fS

ΔV0 V0



(12)

So, the marginal value of inductors L1 and L2 required 225 µH and 61 µH, respectively, for CCM operation of the converter with 1% voltage and current ripple. The inductor size depends on the switching frequency and the amount of current that follows via inductor. 3.2

Simulation Results

A MATLAB/Simulink model is developed to validate the converter topology in the laboratory environment. The input sources of this model are kept at 15 V, and the inductor’s value chosen as 510 µH and 649 µH for L1 and L2 , respectively for CCM operation of the converter as discussed in Sect. 4, capacitor’s value

288

S. Das et al.

taken as 110 µF for C1 and 305 µF for C0 that evaluated from (12). Passive elements are designed to maintain the ripple of 1% at a switching frequency of 5 kHz. At the output terminal, the load resistance of 40 Ω is connected. For controlling the converter, the duty ratio d2 is kept constant at 0.3, while the duty ratio d1 varies utilizing a closed-loop control as illustrated in Fig. 4. The output voltage of the converter can be adjusted across a wide range of about 15 V to 80 V. So, the proposed converter can easily provide the desired output voltage level and be integrated with the dc microgrid. Maintain the constant output voltage of 70 V; this converter shows the duty of switch S1 is 0.735. The output voltage, output current, inductors current, and capacitor charging-discharging with switch pulse are shown in Fig. 5 under the normal operating condition when both input sources are connected.

Fig. 5. Simulation results shows the output voltage (V0 ); output current (i0 ); inductors current (iL1 & iL2 ); capacitors current (iC1 & iC0 ); capacitors voltage (VC1 & VC0 ) when both sources are connected as input.

Dynamic Power Flow Control

289

Fig. 6. Simulation results shows the dynamic behavior of the converter to load changes.

Fig. 7. Simulation results shows the output voltage (V0 ); output current (i0 ); inductors current (iL1 & iL2 ); capacitors current (iC1 & iC0 ); capacitors voltage (VC1 & VC0 ) when one source is connected as input.

290

S. Das et al.

Validate the dynamic behavior of the converter with load variation; the load resistance is decreased by 25%. Figure 6 shows the impact of load changes on the output voltage, current, and inductors. It can observe that the converter maintains constant output voltage even if load variation occurs. Furthermore, the proposed converter operates in single input mode with just voltage source V1 is feeding power to the load. The duty ratio of switch S1 varies from 0.735 to 0.807 for maintaining the constant output voltage of 70 V. The resultant waveform in this state is shown in Fig. 7. The above Simulink results validate the successful operation of the proposed converter.

4

Conclusion

This paper conveys the design of the DISO dc-dc converter having two switches, three diodes, two inductors, and two capacitors that comprise a boost and a buck-boost converter. Two switches are used to control the converter operation; four operating modes of the converter have been explained in this paper. A closed-loop control scheme with dynamic power flow control is presented in the control section for controlling the converter operation. Both the input sources are connected and feeding power to the load during the normal mode of operation, while in a single input mode of operation, the voltage source V1 provides power to the load. In both cases, the proposed converter successfully meets the desired output level and has a good dynamic response to the load variation. So, the Simulink results validate the successful operation of the proposed converter topology. Acknowledgment. The authors thankfully acknowledge the financial assistance provided by the SERB-sponsored IMPRINT-IIC (IMP/2019/000234/EN) project titled “Feasible Coordinated Controlled Grid-connected Photovoltaic Sourced DC based Fast Charging Infrastructure for Electric Vehicle: Design, Development and Experimental Validation” to carry out the work successfully.

References 1. Kardan, F., Alizadeh, R., Banaei, M.R.: A new three input DC/DC converter for hybrid PV/FC/battery applications. IEEE J. Emerg. Sel. Topics Power Electron. 5(4), 1771–1778 (2017) 2. Aryan, R., Ranjan, R., Kumar, A.: Primary control strategies for power sharing and voltage regulation in dc microgrid: a review. In: 2020 3rd International Conference on Energy, Power and Environment: Towards Clean Energy Technologies, pp. 1–6 (2021) 3. Prabhala, V.A.K., Fajri, P., Gouribhatla, V.S.P., Baddipadiga, B.P., Ferdowsi, M.: A DC-DC converter with high voltage gain and two input boost stages. IEEE Trans. Power Electron. 31(6), 4206–4215 (2016) 4. Tian, Q., Zhou, G., Leng, M., Xu, G., Fan, X.: A non-isolated symmetric bipolar output four-port converter interfacing pv-battery system. IEEE Trans. Power Electron. 35(11), 11731–11744 (2020)

Dynamic Power Flow Control

291

5. Athikkal, S., Guru Kumar, G., Sundaramoorthy, K., Sankar, A.: A non-isolated bridge-type DC-DC converter for hybrid energy source integration. IEEE Trans. Ind. Appl. 55(4), 4033–4043 (2019) 6. Sivaprasad, A., Joseph, J., Kumaravel, S., Ashok, S.: Design and analysis of a dual input DC-DC converter for hybrid electric vehicle. In: Proceedings IEEE SPICES, February 2015, pp. 1–5 (2015) 7. Hu, R., Zeng, J., Liu, J., Yang, J.: Double-input DC-DC converter for applications with wide-input-voltage-ranges. J. Power Electron. 18(6), 1619–1626 (2018) 8. Kumar, L., Jain, S.: A novel dual input DC/DC converter topology. In: Proceedings of IEEE PEDES, December 2012, pp. 1–6 (2012) 9. Mishra, S.K., Nayak, K.K., Rana, M.S., Dharmarajan, V.: Switched-boost action based multiport converter. IEEE Trans. Ind. Appl. 55(1), 964–975 (2019) 10. Askarian, I., Pahlevani, M., Knight, A.M.: Three-Port Bidirectional DC/DC Converter for DC Nanogrids. IEEE Trans. Power Electron. 36(7), 8000–8011 (2021) 11. Prabhakaran, P., Agarwal, V.: Novel boost-SEPIC type interleaved DC-DC converter for mitigation of voltage imbalance in a low-voltage bipolar DC microgrid. IEEE Trans. Ind. Electron. 67(8), 6494–6504 (2020) 12. Ardi, H., Ajami, A., Kardan, F., Avilagh, S.N.: Analysis and implementation of a non-isolated bidirectional DC-DC converter with high voltage gain. IEEE Trans. Ind. Electron. 63(8), 4878–4888 (2016) 13. Uno, M., Sugiyama, K.: Switched capacitor converter based multiport converter integrating bidirectional PWM and series-resonant converters for standalone photovoltaic systems. IEEE Trans. Power Electron. 34(2), 1394–1406 (2019)

Performance Analysis of Machine Learning, Deep Learning and Ensemble Techniques for Breast Cancer Diagnosis Piyush Sharma1(B) , Pradeep Laxkar1 , and Anuj Kumar2 1 Department of CSE, Mandsaur University, Mandsaur, M.P., India

[email protected], [email protected]

2 Department of Radiotherapy, Sarojini Naidu Medical College, Agra, U.P., India

[email protected]

Abstract. Cancer is always a life-threatening disease. The lives of people can be saved only through proper treatment of cancer. Within the human body, there are different kinds of cancer that can form, one among them is breast cancer. Women are more at risk of breast cancer as compared to men due to the structure of the human body. To reduce the diagnosis time by the radiologists and increase the accuracy a computer aided diagnosis is required. In this study, we present a review on the different machine learning algorithms used in the past by the researchers and compare them. This paper will be beneficial for the newcomers who want to understand machine learning algorithms and develop the base for the deep learning. Keywords: Machine learning · Breast cancer diagnosis · Deep learning · Ensemble techniques

1 Introduction Across the globe, breast cancer is considered as the most deadly type of cancer. It is essential to diagnose at an preliminary stage to save the lives of patients. It is a kind of tumor that invades inside the tissues of the breast. There are different machine learning algorithms, but selecting a significant and appropriate breast cancer diagnosis algorithm is a tedious task. Across the globe, around 50% to 60% of breast cancer cases are diagnosed in the later stages, resulting in the patient’s lower survival rates [1]. An efficient indicator of an early breast cancer diagnosis is the microcalcification [2]. Hence, there is a need to devise numerous factors that can help in diagnose of breast cancer early so that the survival rate can be increased. It is the leading cause of women’s death. Inside the body there is a regeneration process to maintain the cells. It is a natural body mechanism to maintain the cells balanced growth and death rate, but every time this isn’t the case. However an abnormal situation occurs, when anomalous cells start growing. The various factors for breast cancer are hereditary, age, breast density, fatness, drinking alcohol etc. It is classified into two types © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 292–311, 2022. https://doi.org/10.1007/978-981-19-1742-4_25

Performance Analysis of Machine Learning, Deep Learning

293

Benign and Malignant [3]. Benign is non-life-threatening as it is considered as noncancerous. But there is a probability that it could turn into a cancerous state, whereas malignant cancer begins from anomalous cell growth and it may quickly spread or attack surrounding tissues. The various types of breast cancer are Ductal carcinoma in situ (DCIS), Invasive ductal carcinoma (IDC), Lobular breast cancer (LBC), Mucinous breast cancer (MBC) and Inflammatory breast cancer (IBC) [4]. Ductal carcinoma in situ (DCIS) is a non life threatening since the abnormal cells are fully restricted inside the milk ducts only though there are chances that it can develop further to invasive ductal carcinoma [5]. Invasive ductal carcinoma is developed due to the cells’ anomalous growth that propagate out of the lobules or ducts and are spread to different parts of the body. Around 80% of the breast cancers are invasive ductal carcinoma [6]. Lobular breast cancer initiates in the lobules, in this phase the cells are penetrated outside the lobules and widespread to the lymph nodes and other organs of the body [7, 8]. Mucinous breast cancer [9] invades in the milk ducts and spreads to the tissues surrounding the ducts. It is classified into two types first is the pure mucinous breast carcinoma and secondly mixed mucinous breast carcinoma [10]. Inflammatory breast cancer is an aggressive and unusual disease in which the lymph vessels are blocked by the cancerous cells and the breast appears as red, inflamed [11]. Women with IBC had inferior survival rate in comparison with non-IBC [12] (Fig. 1).

Fig. 1. Major types of breast cancer

The imaging of particular area of body is identified and the doctor target that area for diagnosis. The medical photography is classified in to two types noninvasive and invasive. Ultrasound, X-ray (mammogram), computer aided tomography (CAT) in short CT and Magnetic Resonance Imaging (MRI) are the noninvasive techniques whereas Histopathological images (biopsy imaging) is the invasive technique [3]. Researchers have used different algorithms and examination techniques to examine the breast images. Basically there are four stages for breast image classifier. The first is the breast database selection, second is the feature extraction and selection third is the classifier model fourth is the performance measuring parameter and finally the classifier

294

P. Sharma et al.

output. Image classification methods can be classified into three types: Supervised, unsupervised and semi-supervised learning (Fig. 2).

Fig. 2. Different AI algorithms used for breast cancer diagnosis

2 Machine Learning Techniques The common machine learning algorithms for diagnosis of breast cancer are: 2.1 Logistic Regression It is a supervised learning algorithm used for solving binary classification problems using logistic function [13]. The output can be one or more, which is determined by an independent variable [14]. 2.2 Artificial Neural Network Artificial Neural Network consists of an input layer, hidden layer and output layer. Each layer has neurons linked with weights and the communication links. It is applied in various problem domains of pattern recognition, medical diagnosis [15, 16]. 2.3 Naive Bayes Algorithm The Naive Bayesian network [17] consists of one parent and multiple children. The benefit is that the training time for computations is less, dataset can be categorized using single probability distribution and the different classes can be easily discriminated. It requires less memory space for training and classification.

Performance Analysis of Machine Learning, Deep Learning

295

2.4 Support Vector Machine It is a machine learning method which is efficient in pattern recognition, classification, regression, and data mining due to their increased generalization performance [18]. It uses the principle of structural risk minimization for reducing the generalization error [19]. SVM identifies a hypersurface such that the distance of the closest training data on both sides of the surface is maximum [20, 21]. 2.5 Random Forest It is a supervised learning algorithm used to generate decision trees on randomly selected data examples [22]. It is a meta-estimator algorithm in which a combination of decision trees is used [23]. For each data sample, a decision tree is constructed, and a prediction result is obtained for individual decision tree, for prediction of result vote is performed [24, 25]. 2.6 K Mean Algorithm It is an unsupervised learning algorithm to solve the clustering problem [26]. The objects are grouped based on features into k different groups [27]. It assumes that the features from data figure a vector space and a natural clustering between them are estimated [28]. The patterns belonging to the cluster are defined using a membership function [29]. 2.7 Fuzzy C Mean Algorithm It is an unsupervised learning clustering algorithm for solving clustering problems. It is also named soft clustering because there is no hard bounding [30]. It classifies an object based on measures and the characteristic equation [31]. It is used in the image segmentation process, medical image analysis, to diagnose disease [32]. 2.8 K-Nearest Neighbor (KNN) It is a supervised machine learning technique used for classification and regression [33]. For a data point it first accumulates the data points which are nearer to it. These data points are k training samples known as Nearest-Neighbors [13]. Cross validation and heuristics techniques are used for the determining the alternative values of k. It reviews each instances every time therefore it is a slow algorithm [34]. 2.9 Principal Component Analysis It is a statistical processing technique for decreasing the redundancy by projecting data on an appropriate basis [35]. It transforms datasets into smaller uncorrelated derived variables called as principal components [36]. It is useful in identifying patterns of data and understanding the variation and similarities in data. The applications are in the field of image compression and face recognition [13].

296

P. Sharma et al.

2.10 Linear Regression It establishes a linear relationship between the different input variables and output variables. The relationship between the independent and dependent variables can be determined using linear regression [37]. The attributes are identified based on the characteristics of the class of data [38]. 2.11 Decision Trees It is a supervised learning algorithm for solving classification and regression problems [39]. It is a tree-like structure where the internal nodes represent the test on attribute, and the branch denotes the output of the test, which is either true or false [40], and the leaf nodes represent the class distribution. The decision trees can be transformed to frame the classification rules [41].

3 Ensemble Techniques Ensemble learning techniques are a multi-model approach in which multiple models or methods are chosen instead of a single model therefore, it requires more computations [42]. Voting and weighted averaging are the usual methods of aggregating the output results of various decision systems. The performance of ensemble classifiers is better than individual classifiers [23, 43]. 3.1 Bagging It is also known as Bootstrap Aggregation which is a sampling technique [44]. The bagging algorithm’s main aim is to decrease base classifiers’ variance by performing undersampling [45]. The bags of samples are formed, which are tested using different machine learning model, for the prediction of output results voting is used [46]. 3.2 Boosting It is a combination of different algorithms that are used to convert the weak learners to strong learners by adding more weights for training [47]. It is used for solving classification and regression problems. The common boosting techniques are XGBoost, ADABoost, Gradient Boosting [45, 48]. 3.3 Stacking It is a heterogeneous technique that aggregates different classification or regression models. For this, meta-classifier or meta-regressor are used. Complete training sets are used to train the base models and the obtained outputs of base level models for training meta-model [49].

Performance Analysis of Machine Learning, Deep Learning

297

4 Deep Learning Techniques Deep learning is a sub-class of machine learning and artificial intelligence. The traditional machine learning algorithms had limitations in processing data from raw data. Many of the problems using conventional artificial intelligence are now solved easily using deep learning [50]. Multilevel neural networks are used to extract data from raw images [51]. A general-purpose learning strategy is adopted for learning from data [52]. The common deep learning algorithms are deep neural network, convolution neural network, stacked autoencoders, deep Boltzmann machines, and generative adversarial networks. 4.1 Auto Encoder It is an unsupervised learning technique that uses an encoder accompanied by a decoder for training [53]. It consists of an input layer, a hidden layer and an output layer [54]. The encoder converts the input variable x to z using an activation function. The original input is obtained using the decoder. Training of autoencoder is done in unsupervised manner using unlabeled data [55]. 4.2 Sparse Auto Encoders It is an unsupervised learning algorithm that is used for organized representation of nuclei or non-nuclei patches [56]. These are autoencoders where sparsity is initiated in the hidden units, by increasing the number of nodes in hidden layer [54]. To add sparsity in autoencoder regularizer is added to the cost function which is the average output activation value function of a neuron [53]. 4.3 Stacked Sparse Auto Encoder It is a neural network that consists of several layers of sparse autoencoder [56] in which the output of each previous layer is fed as input to the next layer [53]. Features can be increased by combining together multiple stacked sparse autoencoder [57]. 4.4 Convolutional Neural Network Convolutional neural network are widely used in speech recognition, computer vision, image recognition, object detection, image segmentation [58, 59]. It consists a convolutional layer, pooling layer and fully connected layers [60]. It learns from the input and then subsequently increasing the complex layers [61]. 4.5 Generative Adversarial Networks It consists of a generator and discriminator. The generator network captures data distribution and random samples whereas the probability of data sample from actual data is estimated by discriminator. The output of the discriminator network is feedback to the generator network [62].

298

P. Sharma et al.

4.6 Recurrent Neural Networks It is widely used in natural language processing and speech recognition. It is a class of neural network in which the next input states take the input from previous output states. RNN can process the input of any length, and as the input size increases, the model size doesn’t increase. It processes one element at a time, the hidden unit carries information about the sequence of previous elements [52].

5 Statistics in INDIA, US and CHINA India is the second most populated country in the world. According to the GLOBOCAN2018 report of breast cancer statistics for India versus the world [63], in India, 162468 women were diagnosed with breast cancer, and 87090 died due to breast cancer. On average, one in four newly diagnosed cancer among women accounted for breast cancer in India. 23.5% of deaths among all other cancer-related deaths occurred due to breast cancer in India. In the 25 to 49 age group, 32.8% of women are diagnosed, while 27.9% of cases are diagnosed in the 50 to 69 age group. In the age group of 70 or above, 23.4% of cases are diagnosed [64]. In the United States, 234087 women were diagnosed of breast cancer, and 41904 died due to breast cancer, whereas in China, 367900 women were diagnosed of breast cancer and 97972 died due to breast cancer.

6 Related Work Jiao Z designed a framework for breast mass classification. For training the network CNN architecture was divided into convolution, pooling and rectified linear unit operational units. To optimize the network in training phase stochastic gradient descent is used and SVM classifier is employed for predicting whether the image is malignant or benign [73]. Jalalian A rectified the problem faced by radiologists in interpretation of CTLM images by developing a CAD framework. The first stage is segmentation of VOI, secondly feature extraction and finally classification. They extracted 12 harlick’s features which are used for calculation from 3D GLCM and for classification the MLPNN is used [79]. Xu B proposed a deep selective attention approach for classification of original images by selecting only the valuable regions. They developed a decision network to make a decision which portion of the image needs to be cropped and whether the cropped patch is essential for classification or not. The classification network is then trained using the selected patches, the feedback obtained by the decision network is used to update the selection policy [81]. Qiu Y predicted the near tem risk of breast cancer using the deep learning technology. In the conventional CAD schemes the useful features are selected and defined manually. The network is divided into adaptive feature identification and risk predictor module [92]. Raghavendra U applied locality discriminant analysis and gabor wavelet for data reduction and feature extraction respectively. Linear discriminant analysis, naive bayes

Performance Analysis of Machine Learning, Deep Learning

299

classifier, support vector machine, AdaBoost, quadratic discriminant analysis, k-nearest neighbor, probabilistic neural network, and Fuzzy sugeno classifier techniques are used for classification [94]. Tajbakhsh N demonstrated that the performance of a pre-trained CNN can be improved by fine-tuning which are useful for medical imaging. They analysed the performance of different applications such as polyp detection, pulmonary embolism etc. The most favourable choice can neither be deep tuning nor shallow tuning for particular application. The best performance practically can be achieved by fine-tuning each layer systematically [97]. Asri H compared C4.5, KNN, Naive bayes, and SVM. The simulation error is considered and KS, MAE, RMSE, RAE, RRSE values are evaluated. They compared efficiency in terms of accuracy, precision, sensitivity and specificity [98]. Kooi T demonstrated that when CNN is trained on large datasets, it outperforms the state-of-the-art system in CAD. To identify the suspicious locations they applied the candidate detector. Around 45000 images are used for training both the system and the results show that CNN performs better in comparison to conventional CAD system [99].

7 Database Used for Diagnosis In this section a comparative study is summarized based on the different machine learning and deep learning techniques available for daiagnosis of breast cancer. Table 1 provides a summarized comparison based on the different database available, image types, algorithms (classifiers), features extracted and the final obtained results. Table 1. A comparative study of various machine learning & deep learning techniques for breast cancer diagnosis Database

Image type

MIAS

Local clinical Database & DDSM

Algorithm

Feature extraction

Result

Reference

Mammograms size SVM & Weighted of (1024 * 1024) Feature SVM pixels with 256 level grayscale

Statistical, texture & clinical feature

SVM-Linear obtained best results (Accuracy-98%, Sensitivity-100%, Specificity-96%)

[65]

Mammograms (pixel size of 50 × 50 µm and 12-bit gray level resolution 512 × 512 pixels)

Mass, shape, spiculation, constrast, presence of calcifications, texture, isodensity and other morphologic- al features

SVM Obtained AUC = 0.805 ± 0.012, SFFS combined with SVM obtained AUC = 0.794 ± 0.014

[66]

SVM and sequential forward floating selection based feature selection

(continued)

300

P. Sharma et al. Table 1. (continued)

Database

Image type

Algorithm

Feature extraction

Result

Reference

The cancer Genome Histopathological Fast scanning deep Atlas Breast cancer images (image size convolutional neural Data set (TGCA) of 1000 × 1000 network pixels)

LBP feature

Pavg = 0.91, Ravg = 0.82, F1avg = 0.85

[67]

American college of Radiology university Hospital data

Mammograms

Naïve Bayes Classifier

Shape, margin, During training density, age & inclusive model Bi-rads category achieved AUC of 0.959, descriptor model achieved AUC of 0.910. The validation is done by an external data set, on validation AUC of 0.935 and 0.876 are obtained respectively

[68]

DDSM & MIAS

Mammograms (1024 × 1024 pixels)

Random forest compared with SVM, GA-SVM, PSO-SVM and decision trees

Shape, margin, texture and intensity feature

97.73% of acuuracy is achieved and MCC value reaches 0.8668

[69]

MIAS & (BAMC) database

Mammograms

MCPCNN

Mass feature such as density, size, shape, margin

Using MCPCNN the Az values ranged from 0.84 to 0.89

[70]

University of California santa Barbara biosegmentation Benchmark dataset & Prostate Gleason grading Dataset

Histopathology images

Shearlet transform and CNN

Primary feature such as exture feature, magnitude and phase of complex shearlet coefficients

AUC of 0.82 ± 0.01 and accuracy of 0.86 ± 0.03

[71]

http://wiki.cancer MRI, Ultrasound, imagingarchive.net/ mammograms, tomosynthesis images

DLA-EABA

–

Accuracy of 97.2%, sensitivity of 98.3%, specificity of 96.5%

[72]

LSVRC & DDSM

Mammograms

Convolutional Neural Network & SVM

Hierarchical feature

Achieved Classification accuracy of 96.7%

[73]

DDSM & MIAS

Mammograms

MVMDCNN-Loss

–

0.8208 accuracy for DDSM and 0.6306 accuracy for MIAS

[74]

CBIS-DDSM & INbreast Dataset

Mammograms

CNN with (VGG or – Resnet) patch classifier Resnet 50 & VGG-16

AUC to 0.98 (sensitivity-86.7%, specificity-96.1%)

[75]

CBIS-DDSM & mini-MIAS

Mammograms

Multi-view feature fusion model

For mass and calcification AUC of 0.932 is achieved and for malignant and benign 0.84

[76]

–

(continued)

Performance Analysis of Machine Learning, Deep Learning

301

Table 1. (continued) Database

Image type

Algorithm

Feature extraction

Result

Reference

DDSM

Mammograms

MV-NN compared with logistic regression, support vector machine and neural network classifiers

–

Accuracy of 78.7%, AUC of 0.89, sensitivity of 78.8%, and specificity of 78.7% obtained in MV-NN for CC+MLO views

[77]

MIAS & Inbreast data set

Mammograms

Genetic programming and learning classifier system

6 statistical feature & 256 LBP feature

GP on MIAS dataset using statistical feature obtained an accuracy of 66.67%, & using extracted knowledge from statistical features to LBP feature obtained an accuracy of 54.55%

[78]

Breast wellness centre, Malaysia, TATA Hospital Mumbai & Medoc centre in Budapest,Hungary

Ultrasound, mammography & CTLM images

Multilayer perceptron neural network

Harlick’s texture feature

THE proposed CAD system achieves 95.2% accuracy, 92.4% sensitivity, 98.1% specificity and AROC of 0.98%

[79]

New York Mammograms university school of (2600 × 2600 pixels) medicine

Multi-view deep convolutional network (MV-DCN)

–

MV-DCN achieves macAUC of 0.688 whereas ensemble of MV-DCN and radiologists obtained macAUC of 0.735

[80]

BreaKHis Dataset

Histopathological images

Deep selection attention network model

–

Achieved 98% accuracy at four different magnifications

[81]

CBIS-DDSM & IN-Breast

Mammograms images of (800 × 800 pixels)

CNN’s with logistic regression

–

For INBREAST database RGP obtained Acc. Of 0.919 ± 0.0003 and AUC of 0.934 ± 0.0003, GGP obtained Acc. Of 0.922 ± 0.0002 and AUC of 0.924 ± 0.0003 whereas for CBIS-DDSM database RGP obtained Acc. Of 0.762 ± 0.0002 and AUC of 0.838 ± 0.0001, GGP obtained Acc. Of 0.767 ± 0.0002 and AUC of 0.823 ± 0.0002

[82]

(continued)

302

P. Sharma et al. Table 1. (continued)

Database

Image type

Algorithm

Feature extraction

Result

Reference

GEO database

–

LDA and auto encoder neural network

–

Accuracy of 98.27% is achieved for deep learning frameworks

[83]

DDSM

Mammograms

DCNN (ImageNet, – AlexNet, GoogLeNet)

GoogLeNet achieved accuracy of 0.929, precision of 0.924, recall of 0.934

[84]

DDSM and CBIS-DDSM

Mammograms

SVM and DCNN

–

The DDSM dataset (SVM with linear kernel function) achieved accuracy of 80.5% and for CBIS-DDSM dataset (SVM with medium Gaussian kernel function) achieved accuracy of 87.2%

[85]

NYU Langone health dataset

Mammograms

DCNN, ResNet-22 network

–

Achieved AUC of 0.895

[86]

Department of Radiology, University of Michigan

Mammograms

CNN

Texture features (GLDS, SGLDM)

Achieved ROC curve of 0.87

[87]

DDSM

Mammograms

SVM, KNN, ANN

Root mean square roughness, circularity of the breast mass contour, fractal dimension, RMS slope

SVM achieved 99.66% accuracy

[88]

MITOS dataset

Histopathological images

Regenerative random forest

Harlick texture feature

Recall - 0.8113, precision - 0.8350 and F-measure of 0.823

[25]

US breast image database

Ultrasound images

Multilayered perceptron neural network & wavelet transform

Variance constrast, autocorrelation contrast and distribution distortion of wavelet coefficients

ROC is 0.9396 ± 0.0183, sensitivity is 98.77% and specificity is 81.37%

[89]

InBreast and DDSM

Mammograms

Convolutional Neural Network

–

InBreast: AUC of 0.91(±0.05) and DDSM: 0.97(±0.03)

[90]

DDSM

Mammograms

DCNN

–

DCNN achieved 89.9% sensitivity

[91]

(continued)

Performance Analysis of Machine Learning, Deep Learning

303

Table 1. (continued) Database

Image type

Algorithm

Feature extraction

Result

Reference

FFDM Image database

Mammograms

Deep learning technology

–

Accuracy of 71.4%

[92]

Mini-MIAS

Mammograms images of size 1024 × 1024 pixels

WFRFT+PCA+SVM compared with WFRFT+PCA+k-NN

–

Achieved 92.22% ± [93] 4.16% sensitivity, specificity of 92.10 ± 2.75% and accuracy 92.16% ± 3.60%

DDSM

Mammograms images

Linear discriminant analysis, k-Nearest neighbor, Decision tree, Quadratic discriminant analysis, Fuzzy sugeno, support vector machine, Naïve Bayes classifier, probabilistic neural network, AdaBoost

Gabor wavelet filters are used for extracting textural feature

Accuracy-98.69%, sensitivity-99.34%, specificity-98.26% achieved by K-NN classifier

[94]

MIAS

Mammograms Images

Histogram modified local contrast enhancement (HM-LCE)

–

HM-LCE technique gives better performance as compared to CLAHE technique.HM-LCE obtained EME of 177.7860, AMBE of 13.9412 and H 5.5178

[95]

DDSM

Mammogram Images

SVM & XGBoost (Combination of DCNN)

Texture features

Overall accuracy of XGBoost was 92.80% and for malignant tumors was 84%

[96]

8 Performance Analysis In this section, the performance result of different classifiers are analysed based on Table 1. We have taken accuaccy, AUC and accuracy, sensitivity & specificity for analysis. In Fig. 3 the accuracy of different classifier model is compared, SVM achieved highest accuracy i.e. 99.66% [91]. Support vector machine is used in [68, 69] the accuracy of SVM varies for different database. In Fig. 4 the AUC value of the different classifier are compared, CNN achieved highest AUC value of 0.98 [78]. It is also used in [94]. In Fig. 5 the accuuacy, sensitivity & specificity of different classifier models are compared K-NN achieved the highest accuracy,sensitivity and specificity [98].

304

P. Sharma et al.

Fig. 3. Accuracy of different classifier in percentage

Fig. 4. AUC value of different classifier

Fig. 5. Accuracy-sensitivity-specificity of different classifier in percentage

Performance Analysis of Machine Learning, Deep Learning

305

9 Conclusion In this paper, we have presented and compared different machine learning and deep learning algorithms for diagnosis of breast cancer. For this, we have selected 33 papers taken from the standard journals such as IEEE, Elsevier, Springer, etc. and after performing a rigorous literature survey, we have compared the different classifier methods based on the database used, for performing experimental research, the types of images used such as mammograms, histopathological, ultrasound images etc. Feature extraction methods describe the different types of features extracted for obtaining the ROI in the image. The researchers have evaluated different parameters to determine the overall efficiency of the classifier model, the standard parameters used are accuracy, sensitivity, specificity, AUC, ROC, precision, recall etc. The accuracy varies depending on the different database used for performing experimental research. Most of the work has been done using machine learning algorithms, but there is still a scope of further research using the ensemble techniques and deep learning algorithms for diagnosis of breast cancer.

References 1. Wang, D., Khosla, A., Gargeya, R., Irshad, H., Beck, A.H.: Deep learning for identifying metastatic breast cancer. arXiv preprint arXiv:1606.05718 (2016) 2. Wang, J., Yang, X., Cai, H., Tan, W., Jin, C., Li, L.: Discrimination of breast cancer with microcalcifications on mammography by deep learning. Sci. Rep. 6(1), 1–9 (2016) 3. Nahid, A.A., Kong, Y.: Involvement of machine learning for breast cancer image classification: a survey Comput. Math. Methods Med. 2017, (2017). Article ID 3781951. https://doi.org/10. 1155/2017/3781951 4. Fatima, N., Liu, L., Hong, S., Ahmed, H.: Prediction of breast cancer, comparative review of machine learning techniques, and their analysis. IEEE Access 8, 150360–150376 (2020) 5. Hou, R., et al.: Prediction of upstaged ductal carcinoma in situ using forced labeling and domain adaptation. IEEE Trans. Biomed. Eng. 67(6), 1565–1572 (2019) 6. Alghodhaifi, H., Alghodhaifi, A., Alghodhaifi, M.: Predicting invasive ductal carcinoma in breast histology images using convolutional neural network. In: 2019 IEEE National Aerospace and Electronics Conference (NAECON), pp. 374–378. IEEE (2019) 7. Lips, E.H., et al.: Lobular histology and response to neoadjuvant chemotherapy in invasive breast cancer. Breast Cancer Res. Treat. 136(1), 35–43 (2012). https://doi.org/10.1007/s10 549-012-2233-z 8. Gest, R., Cayet, S., Arbion, F., Vildé, A., Body, G., Ouldamer, L.: Predictive factors of major deviation (>20 mm) between lesion sizes measured by magnetic resonance imaging and histology for invasive lobular breast cancer. Eur. J. Obstet. Gynecol. Reprod. Biol. 251, 14–19 (2020) 9. Lu, K., et al.: Clinicopathological and genomic features of breast mucinous carcinoma. The Breast 53, 130–137 (2020) 10. Wang, P.L., et al.: Imaging features of pure mucinous breast carcinoma: correlation with extracellular mucus content. Clin. Radiol. 74(7), 569-e9 (2019)

306

P. Sharma et al.

11. Chainitikun, S., Saleem, S., Lim, B., Valero, V., Ueno, N.T.: Update on systemic treatment for newly diagnosed inflammatory breast cancer. J. Adv. Res. 29, 1–12 (2021) 12. Dawood, S., et al.: Differences in survival among women with stage III inflammatory and noninflammatory locally advanced breast cancer appear early: a large population based study. Cancer 117(9), 1819–1826 (2011) 13. MurtiRawat, R., Panchal, S., Singh, V.K., Panchal, Y.: Breast cancer detection using K-nearest neighbors, logistic regression and ensemble learning. In: 2020 International Conference on Electronics and Sustainable Communication Systems (ICESC), pp. 534–540. IEEE (2020) 14. Graja, O., Azam, M., Bouguila, N.: Breast cancer diagnosis using quality control charts and logistic regression. In: 2018 9th International Symposium on Signal, Image, Video and Communications (ISIVC), pp. 215–220. IEEE (2018) 15. Saritas, I.: Prediction of breast cancer using artificial neural networks. J. Med. Syst. 36(5), 2901–2907 (2012). https://doi.org/10.1007/s10916-011-9768-0 16. He, S., Wu, Q.H., Saunders, J.R.: Breast cancer diagnosis using an artificial neural network trained by group search optimizer. Trans. Inst. Meas. Control 31(6), 517–531 (2009) 17. Kotsiantis, S.B., Zaharakis, I.D., Pintelas, P.E.: Machine learning: a review of classification and combining techniques. Artif. Intell. Rev. 26(3), 159–190 (2006). https://doi.org/10.1007/ s10462-007-9052-3 18. Levman, J., Leung, T., Causer, P., Plewes, D., Martel, A.L.: Classification of dynamic contrastenhanced magnetic resonance breast lesions by support vector machines. IEEE Trans. Med. Imaging 27(5), 688–696 (2008) 19. Dheeba, J., Selvi, S.T.: Classification of malignant and benign microcalcification using SVM classifier. In: 2011 International Conference on Emerging Trends in Electrical and Computer Technology, pp. 686–690. IEEE (2011) 20. Chang, R.F., Wu, W.J., Moon, W.K., Chou, Y.H., Chen, D.R.: Support vector machines for diagnosis of breast tumors on US images. Acad. Radiol. 10(2), 189–197 (2003) 21. Sewak, M., Vaidya, P., Chan, C.C., Duan, Z.H.: SVM approach to breast cancer classification. In: Second International Multi-Symposiums on Computer and Computational Sciences (IMSCCS 2007), pp. 32–37. IEEE (2007) 22. Rohan, T.I., Siddik, A.B., Islam, M., Yusuf, M.S.U.: A precise breast cancer detection approach using ensemble of random forest with AdaBoost. In: 2019 International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (IC4ME2), pp. 1–4. IEEE (2019) 23. Adegoke, V.F., Chen, D., Banissi, E., Barikzai, S.: Prediction of breast cancer survivability using ensemble algorithms. In: 2017 International Conference on Smart Systems and Technologies (SST), pp. 223–231. IEEE (2017) 24. Ahmad, F.K., Yusoff, N.: Classifying breast cancer types based on fine needle aspiration biopsy data using random forest classifier. In: 2013 13th International Conference on Intellient Systems Design and Applications, pp. 121–125. IEEE (2013) 25. Paul, A., Dey, A., Mukherjee, D.P., Sivaswamy, J., Tourani, V.: Regenerative random forest with automatic feature selection to detect mitosis in histopathological breast cancer images. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F. (eds.) MICCAI 2015. LNCS, vol. 9350, pp. 94–102. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-24571-3_12

Performance Analysis of Machine Learning, Deep Learning

307

26. Dalmiya, S., Dasgupta, A., Datta, S.K.: Application of wavelet based K-means algorithm in mammogram segmentation. Int. J. Comput. Appl. 52(15), 15–19 (2012) 27. Moftah, H.M., Azar, A.T., Al-Shammari, E.T., Ghali, N.I., Hassanien, A.E., Shoman, M.: Adaptive k-means clustering algorithm for MR breast image segmentation. Neural Comput. Appl. 24(7–8), 1917–1928 (2013). https://doi.org/10.1007/s00521-013-1437-4 28. Lee, S.H., Kim, J.H., Kim, K.G., Park, J.S., Park, S.J., Moon, W.K.: Optimal clustering of kinetic patterns on malignant breast lesions: comparison between K-means clustering and three-time-points method in dynamic contrast-enhanced MRI. In: 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp. 2089– 2093. IEEE (2007) 29. Elmoufidi, A., El Fahssi, K., Jai-Andaloussi, S., Madrane, N., Sekkaki, A.: Detection of regions of interest’s in mammograms by using local binary pattern, dynamic k-means algorithm and gray level co-occurrence matrix. In: 2014 International Conference on Next Generation Networks and Services (NGNS), pp. 118–123. IEEE (2014) 30. Heriana, O., Soesanti, I.: Tumor size classification of breast thermal image using fuzzy C-Means algorithm. In: 2015 International Conference on Radar, Antenna, Microwave, Electronics and Telecommunications (ICRAMET), pp. 98–103. IEEE (2015) 31. Kumar, M., Thakkar, V.M., Bhatt, U., Soliyal, N.: Detection of suspicious lesions in mammogram using fuzzy C-means algorithm. In: 2016 International Conference on Advances in Computing, Communications and Informatics (ICACCI), pp. 1553–1557. IEEE (2016) 32. Prakash, R.M., Bhuvaneshwari, K., Divya, M., Sri, K.J., Begum, A.S.: Segmentation of thermal infrared breast images using K-means, FCM and EM algorithms for breast cancer detection. In: 2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS), pp. 1–4. IEEE (2017) 33. Islam, M.M., Iqbal, H., Haque, M.R., Hasan, M.K.: Prediction of breast cancer using support vector machine and K-Nearest neighbors. In: 2017 IEEE Region 10 Humanitarian Technology Conference (R10-HTC), pp. 226–229. IEEE (2017) 34. Cherif, W.: Optimization of K-NN algorithm by clustering and reliability coefficients: application to breast-cancer diagnosis. Procedia Comput. Sci. 127, 293–299 (2018) 35. Huang, Y.L., Lin, S.H., Chen, D.R.: Computer-aided diagnosis applied to 3-D US of solid breast nodules by using principal component analysis and image retrieval. In: 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, pp. 1802–1805. IEEE (2006) 36. Hasan, H., Tahir, N.M.: Feature selection of breast cancer based on principal component analysis. In: 2010 6th International Colloquium on Signal Processing & Its Applications, pp. 1–4. IEEE (2010) 37. Murugan, S., Kumar, B.M., Amudha, S.: Classification and prediction of breast cancer using linear regression, decision tree and random forest. In: 2017 International Conference on Current Trends in Computer, Electrical, Electronics and Communication (CTCEEC), pp. 763– 766. IEEE (2017) 38. Durai, S.G., Ganesh, S.H., Christy, A.J.: Novel linear regressive classifier for the diagnosis of breast cancer. In: 2017 World Congress on Computing and Communication Technologies (WCCCT), pp. 136–139. IEEE (2017) 39. Marne, S., Churi, S., Marne, M.: Predicting breast cancer using effective classification with decision tree and K means clustering technique. In: 2020 International Conference on Emerging Smart Computing and Informatics (ESCI), pp. 39–42. IEEE (2020)

308

P. Sharma et al.

40. Sathiyanarayanan, P., Pavithra, S., Saranya, M.S., Makeswari, M.: Identification of breast cancer using the decision tree algorithm. In: 2019 IEEE International Conference on System, Computation, Automation and Networking (ICSCAN), pp. 1–6. IEEE (2019) 41. Sivakumari, S., Praveena Priyadarsini, R., Amudha, P.: Accuracy evaluation of C4. 5 and Naive Bayes classifiers using attribute ranking method. Int. J. Comput. Intell. Syst. 2(1), 60–68 (2009) 42. Cole, C.A., Nesbitt, K.L., Valafar, H.: Application of ensemble learning to the differential gene expression in left-right breast tumors. In: 2018 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), pp. 1–6. IEEE (2018) 43. Khan, A.A., Arora, A.S.: Classification in thermograms for breast cancer detection using texture features with feature selection method and ensemble classifier. In: 2019 International Conference on Issues and Challenges in Intelligent Computing Techniques (ICICT), vol. 1, pp. 1–6. IEEE (2019) 44. Baneriee, C., Paul, S., Ghoshal, M.: A comparative study of different ensemble learning techniques using Wisconsin breast cancer dataset. In: 2017 International Conference on Computer, Electrical & Communication Engineering (ICCECE), pp. 1–6. IEEE (2017) 45. Lu, W., Li, Z., Chu, J.: A novel computer-aided diagnosis system for breast MRI based on feature selection and ensemble learning. Comput. Biol. Med. 83, 157–165 (2017) 46. Sharma, R.K., Nair, A.R.: Efficient breast cancer prediction using ensemble machine learning models. In: 2019 4th International Conference on Recent Trends on Electronics, Information, Communication & Technology (RTEICT), pp. 100–104. IEEE (2019) 47. Osman, A.H., Aljahdali, H.M.A.: An effective of ensemble boosting learning method for breast cancer virtual screening using neural network model. IEEE Access 8, 39165–39174 (2020) 48. Vo, D.M., Nguyen, N.Q., Lee, S.W.: Classification of breast cancer histology images using incremental boosting convolution networks. Inf. Sci. 482, 123–138 (2019) 49. Dhanya, R., Paul, I.R., Akula, S.S., Sivakumar, M., Nair, J.J.: F-test feature selection in Stacking ensemble model for breast cancer prediction. Procedia Comput. Sci. 171, 1561–1570 (2020) 50. Mahmood, T., Li, J., Pei, Y., Akhtar, F., Imran, A., Rehman, K.U.: A brief survey on breast cancer diagnostic with deep learning schemes using multi-image modalities. IEEE Access 8, 165779–165809 (2020) 51. Houssein, E.H., Emam, M.M., Ali, A.A., Suganthan, P.N.: Deep and machine learning techniques for medical imaging-based breast cancer: a comprehensive review. Expert Syst. Appl. 167, 114161 (2020) 52. LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature 521(7553), 436–444 (2015) 53. Selvathi, D., Aarthy Poornila, A.: Deep learning techniques for breast cancer detection using medical image analysis. In: Hemanth, J., Balas, V.E. (eds.) Biologically Rationalized Computing Techniques for Image Processing Applications, pp. 159–186. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-61316-1_8 54. Hu, Z., Tang, J., Wang, Z., Zhang, K., Zhang, L., Sun, Q.: Deep learning for image-based cancer detection and diagnosis−a survey. Pattern Recogn. 83, 134–149 (2018) 55. Vincent, P., Larochelle, H., Bengio, Y., Manzagol, P.A.: Extracting and composing robust features with denoising autoencoders. In: Proceedings of the 25th International Conference on Machine Learning, pp. 1096–1103 (2008) 56. Xu, J., et al.: Stacked sparse autoencoder (SSAE) for nuclei detection on breast cancer histopathology images. IEEE Trans. Med. Imaging 35(1), 119–130 (2015)

Performance Analysis of Machine Learning, Deep Learning

309

57. Abraham, B., Nair, M.S.: Computer-aided classification of prostate cancer grade groups from MRI images using texture features and stacked sparse autoencoder. Comput. Med. Imaging Graph. 69, 60–68 (2018) 58. Bardou, D., Zhang, K., Ahmad, S.M.: Classification of breast cancer based on histology images using convolutional neural networks. IEEE Access 6, 24680–24693 (2018) 59. Liu, K., Kang, G., Zhang, N., Hou, B.: Breast cancer classification based on fully-connected layer first convolutional neural networks. IEEE Access 6, 23722–23732 (2018) 60. Ting, F.F., Tan, Y.J., Sim, K.S.: Convolutional neural network improvement for breast cancer classification. Expert Syst. Appl. 120, 103–115 (2019) 61. Ha, R., et al.: Convolutional neural network based breast cancer risk stratification using a mammographic dataset. Acad. Radiol. 26(4), 544–549 (2019) 62. Man, R., Yang, P., Xu, B.: Classification of breast cancer histopathological images using discriminative patches screened by generative adversarial networks. IEEE Access 8, 155362– 155377 (2020). https://doi.org/10.1109/ACCESS.2020.3019327 63. Breast cancer statistics: India versus the world. https://www.breastcancerindia.net/statistics/ stat_global.html 64. Latest statistics of Breast cancer in India (2020). https://www.breastcancerindia.net/statistics/ latest_statistics_breast_cancer_india.html 65. Kavitha, S., Thyagharajan, K.K.: Features based mammogram image classification using weighted feature support vector machine. In: Venkata Krishna, P., Rajasekhara Babu, M., Ariwa, E. (eds.) ObCom 2011. CCIS, vol. 270, pp. 320–329. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-29216-3_35 66. Tan, M., Pu, J., Zheng, B.: Optimization of breast mass classification using sequential forward floating selection (SFFS) and a support vector machine (SVM) model. Int. J. Comput. Assist. Radiol. Surg. 9(6), 1005–1020 (2014). https://doi.org/10.1007/s11548-014-0992-1 67. Su, H., Liu, F., Xie, Y., Xing, F., Meyyappan, S., Yang, L.: Region segmentation in histopathological breast cancer images using deep convolutional neural network. In: 2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI), pp. 55–58. IEEE (2015) 68. Benndorf, M., Kotter, E., Langer, M., Herda, C., Wu, Y., Burnside, E.S.: Development of an online, publicly accessible naive Bayesian decision support tool for mammographic mass lesions based on the American College of Radiology (ACR) BI-RADS lexicon. Eur. Radiol. 25(6), 1768–1775 (2015). https://doi.org/10.1007/s00330-014-3570-6 69. Dong, M., Xiangyu, L., Ma, Y., Guo, Y., Ma, Y., Wang, K.: An efficient approach for automated mass segmentation and classification in mammograms. J. Digit. Imaging 28(5), 613–625 (2015). https://doi.org/10.1007/s10278-015-9778-4 70. Lo, S.C.B., Li, H., Wang, Y., Kinnard, L., Freedman, M.T.: A multiple circular path convolution neural network system for detection of mammographic masses. IEEE Trans. Med. Imaging 21(2), 150–158 (2002) 71. Rezaeilouyeh, H., Mollahosseini, A., Mahoor, M.H.: Microscopic medical image classification framework via deep learning and shearlet transform. J. Med. Imaging 3(4), 044501 (2016) 72. Zheng, J., Lin, D., Gao, Z., Wang, S., He, M., Fan, J.: Deep learning assisted efficient AdaBoost algorithm for breast cancer detection and early diagnosis. IEEE Access 8, 96946–96954 (2020) 73. Jiao, Z., Gao, X., Wang, Y., Li, J.: A deep feature based framework for breast masses classification. Neurocomputing 197, 221–231 (2016) 74. Sun, L., Wang, J., Hu, Z., Xu, Y., Cui, Z.: Multi-view convolutional neural networks for mammographic image classification. IEEE Access 7, 126273–126282 (2019)

310

P. Sharma et al.

75. Shen, L., Margolies, L.R., Rothstein, J.H., Fluder, E., McBride, R., Sieh, W.: Deep learning to improve breast cancer detection on screening mammography. Sci. Rep. 9(1), 1–12 (2019) 76. Khan, H.N., Shahid, A.R., Raza, B., Dar, A.H., Alquhayz, H.: Multi-view feature fusion based four views model for mammogram classification using convolutional neural network. IEEE Access 7, 165724–165733 (2019) 77. Bekker, A.J., Greenspan, H., Goldberger, J.: A multi-view deep learning architecture for classification of breast microcalcifications. In: 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI), pp. 726–730. IEEE (2016) 78. Burling-Claridge, F., Iqbal, M., Zhang, M.: Evolutionary algorithms for classification of mammographie densities using local binary patterns and statistical features. In: 2016 IEEE Congress on Evolutionary Computation (CEC), pp. 3847–3854. IEEE (2016) 79. Jalalian, A., Mashohor, S., Mahmud, R., Karasfi, B., Saripan, M.I., Ramli, A.R.: Computerassisted diagnosis system for breast cancer in computed tomography laser mammography (CTLM). J. Digit. Imaging 30(6), 796–811 (2017). https://doi.org/10.1007/s10278-0179958-5 80. Geras, K.J., et al.: High-resolution breast cancer screening with multi-view deep convolutional neural networks. arXiv preprint arXiv:1703.07047 (2017) 81. Xu, B., et al.: Attention by selection: a deep selective attention approach to breast cancer classification. IEEE Trans. Med. Imaging 39(6), 1930–1941 (2019) 82. Shu, X., Zhang, L., Wang, Z., Lv, Q., Yi, Z.: Deep neural networks with region-based pooling structures for mammographic image classification. IEEE Trans. Med. Imaging 39(6), 2246– 2255 (2020) 83. Zhang, X., et al.: Deep learning based analysis of breast cancer using advanced ensemble classifier and linear discriminant analysis. IEEE Access 8, 120208–120217 (2020) 84. Lévy, D., Jain, A.: Breast mass classification from mammograms using deep convolutional neural networks. arXiv preprint arXiv:1612.00542 (2016) 85. Ragab, D.A., Sharkas, M., Marshall, S., Ren, J.: Breast cancer detection using deep convolutional neural networks and support vector machines. PeerJ 7, e6201 (2019) 86. Wu, N., et al.: Deep neural networks improve radiologists’ performance in breast cancer screening. IEEE Trans. Med. Imaging 39(4), 1184–1194 (2019) 87. Sahiner, B., et al.: Classification of mass and normal breast tissue: a convolution neural network classifier with spatial domain and texture images. IEEE Trans. Med. Imaging 15(5), 598–610 (1996) 88. Li, H., Meng, X., Wang, T., Tang, Y., Yin, Y.: Breast masses in mammography classification with local contour features. Biomed. Eng. Online 16(1), 1–12 (2017). https://doi.org/10.1186/ s12938-017-0332-0 89. Chen, D.R., Chang, R.F., Kuo, W.J., Chen, M.C., Huang, Y.L.: Diagnosis of breast tumors with sonographic texture analysis using wavelet transform and neural networks. Ultrasound Med. Biol. 28(10), 1301–1310 (2002) 90. Carneiro, G., Nascimento, J., Bradley, A.P.: Unregistered multiview mammogram analysis with pre-trained deep learning models. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F. (eds.) MICCAI 2015. LNCS, vol. 9351, pp. 652–660. Springer, Cham (2015). https:// doi.org/10.1007/978-3-319-24574-4_78 91. Suzuki, S., et al.: Mass detection using deep convolutional neural network for mammographic computer-aided diagnosis. In: 2016 55th Annual Conference of the Society of Instrument and Control Engineers of Japan (SICE), pp. 1382–1386. IEEE (2016) 92. Qiu, Y., et al.: An initial investigation on developing a new method to predict short-term breast cancer risk based on deep learning technology. In: Medical Imaging 2016: Computer-Aided Diagnosis, vol. 9785, p. 978521. International Society for Optics and Photonics (2016)

Performance Analysis of Machine Learning, Deep Learning

311

93. Zhang, Y.D., Wang, S.H., Liu, G., Yang, J.: Computer-aided diagnosis of abnormal breasts in mammogram images by weighted-type fractional Fourier transform. Adv. Mech. Eng. 8(2) (2016). https://doi.org/10.1177/1687814016634243 94. Raghavendra, U., Acharya, U.R., Fujita, H., Gudigar, A., Tan, J.H., Chokkadi, S.: Application of Gabor wavelet and Locality Sensitive Discriminant Analysis for automated identification of breast cancer using digitized mammogram images. Appl. Soft Comput. 46, 151–161 (2016) 95. Sundaram, M., Ramar, K., Arumugam, N., Prabin, G.: Histogram modified local contrast enhancement for mammogram images. Appl. Soft Comput. 11(8), 5809–5816 (2011) 96. Song, R., Li, T., Wang, Y.: Mammographic classification based on XGBoost and DCNN with multi features. IEEE Access 8, 75011–75021 (2020) 97. Tajbakhsh, N., et al.: Convolutional neural networks for medical image analysis: full training or fine tuning? IEEE Trans. Med. Imaging 35(5), 1299–1312 (2016) 98. Asri, H., Mousannif, H., Al Moatassime, H., Noel, T.: Using machine learning algorithms for breast cancer risk prediction and diagnosis. Procedia Comput. Sci. 83, 1064–1069 (2016) 99. Kooi, T., et al.: Large scale deep learning for computer aided detection of mammographic lesions. Med. Image Anal. 35, 303–312 (2017)

Decarbonizing Indian Electricity Grid Parvathy Sobha(B) Luleå Technical University, Luleå, Sweden [email protected]

Abstract. India, being one of the fastest growing economies of the world, must take a sustainable path for development. India is responsible for 7% of global CO2 emissions. The electricity sector accounts for nearly 35% of emissions from the country. The switch from fossil fuels to renewable sources is the key in decarbonizing this sector and is considered as the crucial step for climate mitigation. This research investigates the potential of renewable energy sources (RES) - wind, solar and hydro. The optimization model developed in this study analyzes various scenarios for the transition to a sustainable future. The results show that India’s aim to achieve 450 GW of installed capacity from RES is far from a Net Zero future. Results confirm that India has the potential to meet 100% of electricity demand in 2030 from RES including wind, solar and hydro. Introducing Social Cost of Carbon (SCC) is a viable option to reduce emissions in India. However, due to the low cost of coal, high coal taxes do not lead to reduced emissions. Keywords: Optimization model · Renewable integration · Scenario analysis · Energy transition · Energy mix

1 Introduction Currently, India is witnessing the tragic impacts of Climate change. The problem threatens the population with food insecurity, water scarcity, flooding, infectious diseases, extreme heat, economic losses, and global warming [1, 2]. Developing countries like India face large-scale climate variability and are exposed to enhanced risks from climate change. The government of the country has conveyed that India is committed to mitigating the problem of climate change and is actively engaging in activities under UNFCC for action [3]. India, the third-largest energy consumer in the world, accounted for the emission of 2.62 billion tons of CO2 in 2019 [4]. In 2020, the contribution from the electricity sector alone was 928 million tons of CO2 , i.e., power generation accounts for nearly 35% of emissions in India [5]. Coal power plants are majorly responsible for the high emissions in the power sector. India is the second-largest coal consumer in the world [6]. More than 50% of the power generated in the country is from coal power plants. Other fossilbased plants contribute less than 10% [7]. However, the power sector is witnessing a massive shift towards renewable energy sources (RES). The installed capacity of RES has increased by 10% during the year 2020–2021 and the capacity of thermal power plants © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 312–322, 2022. https://doi.org/10.1007/978-981-19-1742-4_26

Decarbonizing Indian Electricity Grid

313

were reduced by 1% (until August 2021) [8]. For each MWh of electricity produced by RES, nearly 980 kg of CO2 emissions can be avoided. The literature includes various technologies to incorporate more RES [9]. In addition, the government has introduced various policies and programs to support renewable growth in the country [10]. These policies have aided the nation to curtail CO2 emissions. The government has declared that India is committed to achieve social and economic, development through sustainable ways. India has submitted Nationally Determined Contribution (NDC) to UNFCCC in 2015, which includes three main points [3]: 1. Reduce emissions by 33–35% compared to 2005 levels 2. Achieve 40% of installed capacity from RES or nuclear by 2030 3. To have a cumulative carbon sink of 2.5–3 Gigatons of CO2 emissions by forest and tree cover by 2030 India also aims to include 100 GW of solar, 60 GW of wind, 10 GW of biomass, and 5 GW of hydro (small) by in generation capacity by 2022. India should meet this target while meeting increasing power demand. The country is expected to reach a population of 1.5 billion in 2030 and energy demand of 2499 TWh [3]. Being one of the fastest growing economies of the world, it is critical for India to grow sustainably. In this paper, the author tries to identify how far India has succeeded in meeting the 2022 target, and what India should further in this decade to achieve the 2030 target. Furthermore, the possibility of including 100% RES in power generation by estimating the available potential of RES is also analyzed.

2 Indian Electricity Sector Indian electricity grid is divided into five regional grids, northern, eastern, western, southern, and northeastern grid. The regional grids were operating independently before 2006. The synchronization of the regional grids was completed in 2013. Currently, India has a completely synchronized electricity grid. The advantage of the unified grid is that it opens the possibility of energy sharing between regional grids particularly with the advent of more renewable resources. Indian grid is also connected with Bhutan, Bangladesh, Myanmar, and Nepal. 2.1 Energy Mix India holds the 5th largest installed capacity of RES in the world. However, 60% of installed capacity in the country accounts for fossil fuels. Table 1 shows the current installed capacity in India [8]. Figure 1 shows power generation in India for the year 2020, the annual generation is around 1340 TWh. The demand is primarily met by fossil-based power plants. The generation from RES varies widely between 17–30% during the year. This variation is attributed to the supply from hydro plants. The contribution from hydropower increases largely during the monsoon season and decreases during summer. During the past few

314

P. Sobha

years, the southern states in India open shutter(s) of hydropower dams during monsoon when the water reaches Full Reservoir Level (FRL). The generation share from coal varies between 71 to 80% during the year, following variation in hydro supply. Nuclear generation remains steady around 3%–4%. Table 1. Installed capacity in India (31.08.2021) Fuel

Installed capacity (GW)

% share

Coal

203

53

1

0.2

25

6.5

Lignite Gas Diesel

1

0.1

Nuclear

7

1.8

Hydro

51

13.4

Wind

40

10.4

Solar

46

11.9

Other RES

11

2.8

100% 120 80%

100 80 TWh

60%

60

40% 40 20% 20 0%

0 Jan

Feb

Thermal

Mar

Nuclear

Apr

Maj

Hydro

Jun

Jul

Solar

Aug

Sep

Wind

Okt

Nov

Other RES

Dec

Net

Fig. 1. Power generation mix in India, 2020

2.2 Energy Justice In India, around 0.4% of the population doesn’t have access to electricity (2019 data), i.e., nearly 5 million people live without electricity [11]. Moreover, the quality and reliability of electricity need to be improved in many places including both urban and rural areas. One of the 17 sustainable development goals set by UN, SDG 7 is to “ensure access to affordable, reliable, sustainable and modern energy for all”. One target of SDG 7 is,

Decarbonizing Indian Electricity Grid

315

“stepped-up efforts in renewable energy are needed to achieve long-term climate goals” [12]. The transition to renewable energy leads to the achievement of SDG 7. 2.3 Predicting Energy Supply from RES A prediction model is employed to access the available potential of wind, solar and hydro energy for the year 2020 [13]. The average available potential of RES is assumed to remain constant over the decade. Annual generation capacity for the entire country is derived from the model at “1-h” resolution (i.e., for 8760 h in total). District-level data is generated from the model for better spatial resolution (state also called province is divided into different districts). The generated potential for districts is aggregated geographically to obtain renewable potential for states. Finally, the states’ data are aggregated to get data of different regional grids. The determined potential for each grid is combined with the historical regional capacity factors of the grid. The final energy potential determined for each region is depicted in Fig. 2 along with a comparison to the current installed capacity. On average, only 30% of wind potential is explored in India. In the case of solar, 97% of the potential remain unexplored. 70 60

40

Wind Energy Potenal

Solar Energy Producon

Solar Energy Potenal

TWh

50

Wind Energy Producon

30 20 10 0 Northern Grid

Western Grid

Southern Grid

Eastern Grid

North Eastern Grid

Fig. 2. Potential energy supply from RES

3 System Model 3.1 Optimization Model In this section, a techno-economic-environmental model is developed to compare various pathways to attain reduced CO2 emissions in 2030. The model is developed using single objective linear programming (LP) [14]. The model is fed with the details of existing power plants, power demand, and generation potential of RES (from the prediction model). It is fed with techno-economic & environmental related data, including emission data, fuel consumption, fuel efficiency, various plant expenditures, Levelized Cost of

316

P. Sobha

Energy (LCOE) [15] of different plants, etc. The purpose of the model is to determine a cost-optimized energy mix to meet the national power demand. The model can decide whether to meet demand from existing power plants or to invest in new power plants. The model uses cost, emission data, and other constraints defined to make strategic decisions regarding the electricity generation mix. The model framework is shown in Fig. 3, which describes the type of power plants, techno-economic data used, constraints employed, and the scenarios used in the model.

Fig. 3. Optimization model framework

3.2 Mathematical Model The objective function of the model is to minimize the total system cost for generating the power. In the model, similar types of power plants are aggregated for ease of computation (e.g., demand from all coal plants as summed up, generation capacity from onshore wind is aggregated). Equation 1 describes the objective function. min obj. fn =

n 

xk (αk ∗ βk + γk + δk + εk )

(1)

k=1

x: generation from plant type k (MWh) n: number of type of plants α: fuel efficiency ($kg/MWh) β: fuel cost ($/kg) γ : operating & maintenance cost, fixed ($/MW) ε: operating & maintenance cost, variable ($/MWh) δ: LCOE, calculated for new power plants ($/MWh) The constraints limiting the objective function mainly include power demand, installed capacity, and potential generation by RES. Two types of constraints are

Decarbonizing Indian Electricity Grid

317 2000

90%

1800

80%

1600

70%

1400

60%

1200

GtCO2

100%

50%

1000

40%

800

30%

600

20%

400

10%

200

0%

0

Current 2021 Thermal

No Tax Nuclear

Coal Tax

Tax on CO2 Hydro

450 GW RES Wind

40 % RES Solar

60 % RES

80 % RES

Other RES

100 % RES 2030 Pledge

RES Share

CO2 Emissions

Fig. 4. Optimization model results

employed, one for existing plants and the other for new plants added to meet increased demand. Equations 2, 3, and Eq. 4 describe the constraints. (2) : electricity demand (MWh)

(3) : Generation potential of new power plants (MWh) = f (capacity factor, prediction of RES, hours of operation)

∑

≤ Ć

=1

Ć Ć

(4)

: Generation from existing power plants : = f (capacity factor, hours of operation, installed capacity)

3.3 Scenario Development Scenario definition is the most significant part of the model. Scenarios explore potential futures, i.e., rather than predicting one future, scenarios explore multiple pathways to the future [16]. In the model, scenarios are defined based on both the current system and

318

P. Sobha

the future to which the system should evolve to. The scenarios are motivated by SDGs and India’s plan to increase RES share by 2030. Table 2 explains the scenarios defined in the model. Here, targeted, and untargeted scenarios are considered. Targeted scenarios lead to the desired future (e.g., 40% renewables, Pledge 2030) and untargeted scenarios shows how the future will evolve for a particular situation in the present (e.g., BaU, Tax on CO2 ). 3.4 Scenario Results Scenario results are shown in Fig. 4. In the results, generation from fossil fuel-based power plants (coal, lignite, diesel, natural gas, naphtha) are aggregated as “thermal” sources. The model analyses the possibility of meeting demand with thermal, nuclear, solar, wind, and hydro energy sources. Hence the scenarios are designed likewise. As expected, the emission curve follows RES. With high penetration of RES, emission reduces as low as 84 Gtons of CO2 . The share of nuclear remains constant for all the scenarios owing to the high investment cost. The share of hydro reaches a maximum of 15% and no new investments are planned. The high LCOE of hydropower restrains the model from increasing its share in the total energy mix. Table 2. Scenarios analysed Scenario

Name

Description

Scenario I

BAU

Business as Usual, no change in policies

Scenario II

Coal Tax

Tax on coal increased by 200%

Scenario III

40% RES

Increase RES share to 40% of net generation

Scenario IV

60% RES

Increase RES share to 60% of net generation

Scenario V

80% RES

Increase RES share to 80% of net generation

Scenario VI

100% RES

Increase RES share to 100% of net generation

Scenario VII

2030 Pledgea

Decrease emission by 35%

Scenario VIII

SCC

Social Cost of Carbon, taxing CO2 @ 20$/tCO2

Scenario IX

450 GW-RES

Increase installed capacity of RES to 450 GW

a 2030 Pledge is to decrease emissions from the power sector by 33–35% compared to 2005 levels

4 Discussions The model determines different pathways to meet the electricity demand in 2030. Figure 4 represents the scenario results. The following section discusses the scenario results. Current 2021: shows the present electricity generation mix (from January to August 2021).

Decarbonizing Indian Electricity Grid

319

Scenario I: The scenario analyses business as usual case, i.e., no change in current policy measures. The coal tax remains constant at 5.3 $/ton and no emission taxes are levied. No constraints are put on the energy mix. Results show that 68% of the demand is met by fossil fuels. The share of RES increased by only 4% when compared to 2021 levels. Scenario II: The scenario analyses changes in the energy mix when the social cost of carbon (SCC) is included. The carbon tax was initially established in Europe and it varies from 30 cents/tCO2 –137 $/tCO2 in different countries of the continent. In the analysis 20 $/tCO2 is the rate chosen. In the result, fossil share reduced to 5%, and RES share increased to 93%. Levyingtax on CO2 stands as a strong measure for decarbonizing the electricity sector. Scenario III: Here, the tax on coal is increased by 200%. However, the fossil share is still high at 69% and renewable share is closed to 30%. Due to the low coal cost, the doubling of coal tax doesn’t have any positive impact on the energy mix. Scenario IV: Here, the minimum share of RES required in energy mix should be 40%. This reduced the fossil share to 58%. The policy is relevant as a short-term goal but in long term, the share of RES needs to be gradually increased. Scenario V: The minimum share of RES required is 60%. The fossil share reduced to 38%. The target is relevant as a mid-term goal but in long term, the share of RES needs to be gradually increased to reduce the fossil share gradually. Scenario VI: The minimum share of RES is 80% in this scenario. This reduced the fossil share to 18%. The target is highly relevant as a long-term goal to gradually phase out fossils from energy mix. Scenario VII: In this scenario, the electricity demand is completely met by RES. Model results confirms that Hydro, solar, and wind energy can meet the nation’s electricity demand in 2030. Scenario VIII: India has pledged to reduce emissions by 33–35% compared to 2005 levels by 2030. This plan is reflected in the power sector. The scenario identifies the energy mix to reduce emissions from the power sector by 33–35%. This strategy reduces the fossil share to 14%. India will be on right track if the target is achieved by 2030. Scenario IX: India aims at raising the installed capacity of renewables to 450 GWs by 2030. This scenario is analyzed here. Results show that the share of fossils is 57% and that of RES is 41%. This can be a near future goal than a long-term goal, owing to high emissions from fossil fuels. A significant finding from the analyses is the minimal or zero investment in offshore wind and concentrated solar power (CSP) plants owing to their high LCOE. Even in the scenario of 100% renewables, the share from solar PV panels (ground and rooftop mounted) reaches 78% completely excluding CSP power plants. Figure 5 shows the increase in offshore wind and CSP share in energy mix with reduced LCOE. The results

320

P. Sobha 15% 10% 5% 0%

50%

-5%

60% Decrease in LCOE

-10% -15% -20% Increase in Wind_Offshore share

Increase in Solar_CSP share

Decrease in Solar_Rooop share

Fig. 5. Effect of LCOE on energy mix

show that to increase their share, the LCOE must be reduced by at least 50%. In this case, at reduced LCOE, the maximum potential of offshore wind and CSP are utilized by the model. Figure 6 shows the additional capacity that needs to be installed to meet 2030 demand considering the different scenarios. The hydro capacity has to be increased to 278 GW from 46 GW. The capacity of solar PV (rooftop) must be increased between 500–1000 GWs. The solar panels’ (ground-mounted) capacity is to be increased to 405 GWs from 38 GWs. Wind capacity (onshore) should be increased to 70 GWs from 40 GWs. The government has planned to increase RES share from 130 to 450 GWs by 2030. From Fig. 4, it is evident that the target should be far higher than 450 GWs to meet power demand in 2030 with reduced emissions. With three times increase in RES in the energy mix, the share of fossils stays at 57% compared to 41% of RES. Undoubtedly this is a major achievement but the emissions from this 57% would be about 1.4 Gt of CO2 for producing 1429 TWh of electricity. The annual emissions during 2019–2020 were 0,98 GtCO2 for producing 1196 TWh of electricity. In this case, the annual CO2 emissions increased by 30%. To sum up, to meet the increasing energy demand with reduced emissions, India should aim at high levels of renewable integration.

5 Action Plan India has been keen on promoting clean energy sources during the past decades. The renewable share has grown cumulatively by nearly 70% when compared to the beginning of the last decade. The analysis shows that there is still a large potential of RES to explore. The government has launched mitigation programs including, Green Energy Corridor for the incorporation of RES; Green Generation for Clean & Energy Secure India to achieve 175 GW of RES generation by 2022 and Solar-powered toll plazas. Multiple projects have been rolled out to increase energy efficiency including, the National Smart Grid Mission to improve the efficiency of the power grid; Energy Conservation Campaign to reduce energy consumption by 10%; Clean Coal Policies to improve

Decarbonizing Indian Electricity Grid

321

1000 900 800

600

GW

700

BaU

Coal Tax

Tax on CO2

40 % RES

60 % RES

80 % RES

100 % RES

2030 Pledge

Current Capacity

450 GW - RES

500 400 300 200 100 0 Coal

NG

Hydro

Wind _Onshore Wind _Offshore

Solar_GND

Solar_CSP

Solar_RoofTop

Fig. 6. Additional capacity for RES

the efficiency of coal plants; National Mission for Enhanced Energy Efficiency (NMEEE) to improve energy efficiencies through policies and regulations. National Action Plan on Climate Change (NAPCC) aims at reducing per capita CO2 emissions by increasing the share of RES. Despite all the policies and programs, the renewable potential in India remains underutilized. The analysis shows that 70% of wind potential and 97% of solar potential is unused. New policies and strategies have to be designed in power domain, along with updating existing strategies. India should also aim at grid expansion policies to incorporate the growing supply to meet the increasing demand. This enables the sharing of renewable energy across different regional grids, which stands as a permanent solution for not only energy poverty and energy justice but also SDG 7.

6 Conclusion India has come a long way from the last decade but has to go longer from where it stands now. Despite holding the 5th largest installed capacity of renewable energy in the world, 60% of the electricity in the country is still generated from fossil fuels. Effective strategies must be formulated along with the implementation of new technologies [17] for reducing emissions from the electricity sector and increasing the energy efficiency of electrical equipments and appliances connected to the power grid. Above all, the country should meet the goal,” Electricity for All”.

References 1. Dagdeviren, H., Elangovan, A., Parimalavalli, R.: Climate change, monsoon failures and inequality of impacts in South India. J. Environ. Manag. 299, 113555 (2021) 2. Dash, S., Maity, R.: Revealing alarming changes in spatial coverage of joint hot and wet extremes across India. Sci. Rep. 11, 18031 (2021) 3. UNFCC. https://www4.unfccc.int/sites/ndcstaging/PublishedDocuments/India%20First/ INDIA%20INDC%20TO%20UNFCCC.pdf. Accessed 11 Nov 2021 4. Ourworldindata. https://ourworldindata.org/co2/country/india. Accessed 11 Nov 2021 5. CEA. https://cea.nic.in/wp-content/uploads/baseline/2021/06/User_Guide_ver_16_2021-1. pdf. Accessed 11 Nov 2021

322 6. 7. 8. 9.

10. 11. 12. 13.

14. 15. 16. 17.

P. Sobha IEA. https://www.iea.org/reports/coal-2020/demand. Accessed 11 Nov 2021 CEA. https://cea.nic.in/?lang=en. Accessed 11 Nov 2021 Power Ministry. https://powermin.gov.in/en/content/overview. Accessed 11 Nov 2021 Sobha, P., Patne, N.R., Usman, T.S.: Optimal battery charging forecasting algorithms for domestic applications and electric vehicles by comprehending sustainable energy. In: Singh, A.K., Tripathy, M. (eds.) Control Applications in Modern Power System. LNEE, vol. 710, pp. 978–981. Springer, Singapore (2016). https://doi.org/10.1007/978-981-15-8815-0_3 Sawhney, A.: Striving towards a circular economy: climate policy and renewable energy in India. Clean Technol. Environ. Policy 23(2), 491–499 (2021) IEA. https://iea.blob.core.windows.net/assets/1de6d91e-e23f-4e02-b1fb-51fdd6283b22/ India_Energy_Outlook_2021.pdf. Accessed 11 Nov 2021 UN SDG. https://sdgs.un.org/goals/goal7. Accessed 11 Nov 2021 Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based backstepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 Vanderbei, R.J.: Linear Programming, 5th edn. Springer, Cham (2020). https://doi.org/10. 1007/978-3-030-39415-8 Raikar, S., Seabron, A.: Renewable Energy Finance, 1st edn. Science Direct (2020) Romy, G., Javier, P., Huchery, C., Collier, N., Garnett, S.: Scenario modelling to support industry strategic planning and decision making. Environ. Model. Softw. 55, 120–131 (2014) Sobha, P., Patne, N.R.: Optimized operation of available energy resources based on energy consumption. In: IEEE First International Conference on Smart Technologies for Power, Energy and Control (STPEC), vol. 1, pp. 1–6. IEEE (2020)

Future Mobility with eVTOL Personal Air Vehicle (PAV): Urban Air Mobility (UAM) Concept M. Naga Praveen Babu(B) , Sidvik Basa, Prasanth Kumar Duba, and P. Rajalakshmi NM-ICPS TiHAN Foundation, Indian Institute of Technology (IIT) Hyderabad, Hyderabad, India [email protected]

Abstract. Many countries consider Urban Air Mobility (UAM) a new mode of transportation for intra-regional short-distance journeys. The system is one of the upcoming on-demand airborne transportation networks and includes drone taxis and personal air vehicles. The primary purpose of the UAM concept is to use electric vertical take-off and landing (eVTOL) vehicles to identify passenger locations, fly and cruise, load the passengers, and deliver them to their destinations. UAM system is a currently evolving field, and multiple concepts such as multi-copter concepts, Lift and Cruise concepts, and Tilt-Wing concepts are being proposed. The behavior of each concept vehicle, and hence its energy efficiency, varies. Using low-altitude airspace, UAM is intended to provide an innovative transportation model for passengers and goods in metropolitan areas with significantly increased mobility. Ground infrastructure incorporating vertiports, regulations, policies, and other vital components, is required to transform UAM from design to operation. Electric flight is thought to be the next step toward more environmentally sustainable air travel. In the present study, a personal air vehicle (PAV) based eVTOL concept is proposed for UAM. The preliminary design and modelling of PAV are discussed here. The aerodynamic performance of propeller characteristics used in PAV design is compared using numerical and experimental studies. The results show that at different velocities, the normal and side forces generated by the propellers are found to be more stable in PAV cruise mode/ forward flight mode, respectively. Keywords: eVTOL · Personal Air Vehicle (PAV) · Power · Thrust · Torque · UAM · Vertiports

1 Introduction By efficiently utilizing the three-dimensional airspace, eVTOL aircraft can reduce traffic congestion on the ground, increasing worldwide. While flying taxis were envisioned as a means of transportation in the current technological breakthroughs, predictions indicate that they will become a reality in the next few decades. Urban Air Mobility (or UAM) is a current concept that aims to provide an economical alternative to ground transportation in congested urban areas. The concept of urban air mobility with eVTOL vehicles © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 323–337, 2022. https://doi.org/10.1007/978-981-19-1742-4_27

324

M. N. P. Babu et al.

and their vertiports is shown in Fig. 1. Although UAM has several advantages (cost savings and increased transportation capacity), it also has several problems surmounted to ensure safe and efficient operations. UAMs have been in operation since the 1940s, using helipads to transport people but were discontinued after a series of tragic incidents [1]. Current aviation technology has matured enough to conduct on-demand and planned operations employing quiet and efficient crewed or uncrewed vehicles [2, 3]. Consumer demand has been increasing in many cities worldwide to develop an air taxi system in the urban ecosystem [4]. According to the market research, the USA’s airport shuttle and air taxi sectors alone have a market worth of $500 billion [5]. Recent technological advancements enabled the construction and flight testing of various eVTOL aircraft designs. Over a dozen companies, such as Jobby Aviation, EHang Airbus A3, Volocopter, are working hard to make VTOLs a reality. Despite their design differences, they all use a distributed electric propulsion (DEP) system, improving power-to-weight, efficiency, dependability, and operational flexibility in eVTOLs [6, 7] compared to conventional helicopters rotors. Although helicopters are capable of VTOL operations, the noise they produce has prompted communities to take legal action against their use in UAM. In comparison, DEP-powered eVTOLs have a higher downstream velocity, allowing for a faster vertical descent without approaching a vortex ring state. When paired with fuel management errors, engine failure contributes to 18% of general aviation accidents, which can be minimized by eVTOL systems [8]. eVTOLs are divided into three categories by the Vertical Flight Society: wingless, lift+cruise, and vectored thrust. The E-Hang 216 [9] eVTOL is an upgraded and altered version of the ordinary multi-rotor vehicle. It is one of the vehicles effective in a hover but has a restricted range and velocity, making them appropriate for short-distance travel. Wing-borne lift is required when the desired mission is of more extended range, such as in both lift+cruise and vectored thrust categories. Lift+cruise eVTOL uses two propulsion systems: one for vertical motion and another for wing-borne cruising. For example, the Kitty Hawk Cora [10] takes off vertically with twelve propellers and cruises with one pusher propeller. Though the propulsion systems in the lift+cruise configuration can be customized for a specific flight segment, they contribute to unwanted weight and drag to the eVTOL when they are not required. Using the same thrust and cruise mechanism, eVTOLs with vectored propulsion aim to avoid this problem, but their efficiency decreases in each phase. Vectored thrust eVTOLs tilt the propulsion systems, thereby rotating the thrust vector during the transitional period. For example, the Lilium jet [11] tilts its 36 electric ducted fans to an angle to enable hovering and cruising. Other UAM airspace integration challenges, such as safety and efficiency, must be negotiated in addition to these UAM vehicle-related operations. Safety involves the advancement of technological procedures to ensure isolation from terrain, urban barriers, and other aircraft, for example. Furthermore, methodologies for sequencing, scheduling, and spacing UAM aircraft at constrained resources, such as take-off and landing locations (e.g., vertiports), are crucial for proper operations. UAM vehicles and systems must be interoperable with UAM vehicles and networks and other existing aviation entities to eliminate fundamental safety and reliability limitations of aerospace integration. Figure 2 illustrates the challenges associated with the UAM transportation system. The manuscript is structured as follows. Section 2 describes the UAM concept with PAV, describing the airspace requirements and various eVTOL vehicles. The design

Future Mobility with eVTOL Personal Air Vehicle (PAV)

325

and modelling of PAV are illustrated in Sect. 3. Section 4 involves both the computational and experimental study of PAV propellers. The summary and conclusions of the results are presented in Sect. 5.

Fig. 1. Concept of Urban Air Mobility (UAM).

Fig. 2. Challenges involved in UAM transportation.

2 Urban Air Mobility (UAM) Concept with PAV 2.1 Airspace Requirements in the UAM Mode of Transport Various kinds of airspace are depicted in Fig. 3. Class B, C, D, and E, depicted as regulated airspace, and Class G, marked as uncontrolled airspace, are the most common

326

M. N. P. Babu et al.

types of airspace in the vicinity of major cities. When clearance above terrain is considered, airspace boundaries are established geographically and in terms of altitude above mean sea level (MSL) or height AGL. Only high-performance, pressurized turbojet and turboprop aircraft fly in Class A airspace, which runs from 18,000 ft to 60,000 ft MSL. Because the potential closure rates between these high-performance aircraft make seeand-avoid separation dangerous, all operations must be done under IFR and with the assistance of air traffic separation services. Class A airspace has no direct influence on UAM operations. However, it does provide another example of how airspace requirements are utilized to ensure primary aircraft and ATM capabilities support a specific degree of performance. Class B airspace is set up around the busiest airports to restrict IFR operations and accompanying published procedures that serve those airports. Class C airspace is adapted to the specific needs of local operations, but it extends typically up to 4,000 ft MSL above the primary airport. Major airports are located within or near most urban areas, placing them under Class B or C airspace. Class D airspace exists at airports with an air traffic control tower, with lateral bounds customized to cover local airport operations. Class E airspace begins at 700 ft or 1,200 ft AGL. It is regulated airspace that is not Class A, B, C, or D. At more than 700 ft AGL, most metropolitan locations where UAM operations are projected to take place will have underlying Class E airspace. Class G airspace begins just above the ground and extends up to 700 ft or 1,200 ft AGL, depending on the exceptions and the Class E airspace above. UAM activities, on the other hand, may be exclusively conducted within Class B, C, or D airspace in the metropolitan regions.

Fig. 3. Airspace requirements of UAM Passenger Air Vehicles (PAV).

Future Mobility with eVTOL Personal Air Vehicle (PAV)

327

2.2 Classification eVTOL Vehicles eVTOL vehicles have potential in several applications in various sectors such as defense exploration, airborne observations, search and rescue operations, agriculture, and photography. Fixed-cruise wing and multi-rotor UAVs are the two primary types of UAVs. Traditional aerial vehicles have a few drawbacks, such as a lack of airstrips in rural locations and deep forests. UAVs are better suited for such situations where a lack of runway and dense vegetation are essential concerns for the deployment of services. The current transportation industry is producing an airborne vehicle that requires minimal human input with a high flight and vibrational stability. One of the technologies that meet these parameters is VTOL. VTOLs are being built in various configurations, and some of them are expressed in Fig. 4. A rotary-wing cruise is when (many) rotors entirely generate the lift and thrust. Vehicles in this category have thrust vectors that mostly point in the same direction. In contrast to a helicopter with a tail rotor, differential thrust is used to achieve forward flight and flight maneuvers. In cruise mode, the fixed-wing cruise can be distinguished by the way the wings generate the lift. This fixed-wing cruising category can be divided into two groups. The first is the Lift & Cruise concept, in which the vehicle’s wings are rigidly linked to the fuselage. Lift is generated by the rotors only during hovering and ascending. The Lift & Cruising design incorporates an additional pusher propeller for the cruise that creates thrust. In cruising mode, the wings create all the lift, and the rotors are folded to reduce the drag. In a fixed-wing cruise (the second one), tilt-wing rotor VTOL takes off and lands vertically. A tilt wing rotor UAM is a VTOL with a horizontal wing that changes its axis to vertical during vertical take-off and landing operations. In tilt-rotor VTOL, the propeller position remains constant during landing and take-off phases, but as the speed increases, the propellers line up in the forward direction to produce the required forward thrust. In tilting ducted fan VTOL, the rotors are accommodated by annular-shaped ducts helping the VTOL with fast cruise speed and a good hovering efficiency. Tiltjet VTOL, like tilt-rotor, includes jet engines on edges of the horizontal wing that change axis from vertical to horizontal. These aircraft can perform VTOL maneuvers with ease. Aerodyne VTOL, the first wingless vertical take-off and landing aircraft, consists of a ducted rotor that produces thrust and lift forces. In thrust vector-controlled VTOL, vectoring, which is a phenomenon of shifting the thrust direction, creates the vertical lift and landing. This vectoring phenomenon is also being employed in fighter jets lately. The tail sitter, a form of fixed-wing aircraft that can vertically take-off, land on its tail, and transition into level flight, is another way to combine the desirable characteristics. Like a ducted fan aircraft, a tail-sitter does not require additional mechanical components to achieve VTOL capability. It can also look more like a traditional fixed-wing plane and carry a higher payload than a ducted fan plane. However, this mechanical simplicity comes at the expense of higher transition control complexity. The current UAM based eVTOL vehicles in the research and development stage are shown in Fig. 5.

328

M. N. P. Babu et al.

Fig. 4. Classification of UAM eVTOL vehicles.

Fig. 5. UAM eVTOL vehicles (a) Volocopter (b) Neva Airquad one (c) Boeing PAV (d) Joby S2 VTOL (e) Joby S2 Cruise (f) Opener Blackfly.

3 Design of Personal Air Vehicle (PAV) 3.1 Modelling of PAV UAM Personal Air Vehicles (PAV) is designed based on multi-copter design and features a wide range of propellers/thrusters and configurations. The current study examines a Hexa-configuration-based PAV, as shown in Fig. 6, to analyses the fundamental properties of the multi-copter concept. Six thrusters are mounted on an airframe with a radius of 3m to make up the baseline arrangement. The rotor geometry is based on a multicopter propeller U15 II KV80 that has been resized and changed. The three propellers are rotating clockwise, and the remaining three propellers are rotating in anti-clockwise

Future Mobility with eVTOL Personal Air Vehicle (PAV)

329

directions alternatively. The design details of the passenger air vehicle are shown in Table 1.

Fig. 6. Development of an eVTOL vehicle (a) Isometric view (b) side view without the propellers.

Table 1. Personal Air Vehicle (PAV) particulars. S. No.

Parameters

Dimensions

1

Airframe diameter

3m

2

PAV (Length × Breadth × Height)

2815 mm × 1500 mm × 2060 mm

3

Propeller type

G40 × 13.1

4

Propeller diameter

1016 mm

5

Motor type

U15 II KV80

6

Number of motors

6

7

Number of propellers

6 (3 + 3)

8

Number of passengers

1

4 CFD Study of Electrical VTOL Propulsion System for PAV 4.1 Computational Domain A fixed-pitch propeller of approximately 1m diameter is considered for aerodynamic analysis. The eVTOL propeller with hub is as shown in Fig. 7. For the analysis of the propeller, a rotating region and domain size are to be considered for evaluation such that the domain represents the actual environmental operating conditions of the propeller. The mesh is generated in Ansys Fluent 2021. The domain is divided into two zones - the rotor zone near the propeller has the rotational velocity, and the rest is static. The inlet has a velocity of 8 m/s for an RPM of 1000, gradually increasing to 10 m/s for an RPM of 3200. The sizes of the rotor zone and the overall domain, and the mesh element size

330

M. N. P. Babu et al.

of the rotor zone and stator zone and are considered as shown in Table 2 and Fig. 8. The exact geometry of the domain can be viewed in Figs. 8 and 9. The minimum orthogonal quality observed in the stator-zone is 0.162, and the rotor-zone is 0.0235. The aspect ratio of the mesh in the stator-zone is 15.375, and the rotor-zone is 77.939. Numerical simulations of the propeller are performed in Ansys Fluent 2021. A full-scale model of the propeller is considered for simulation. The rotor zone must rotate in the clockwise direction as the propeller is of anti-clockwise configuration. The details of numerical simulations are presented in Table 3. Table 2. Domain size particulars. S. No.

Quantity

Dimensions

a) Rotor Zone: 1

Rotor zone diameter

1050 mm (D)

2

Rotor zone thickness

80 mm

b) Stator Zone: 3

Length (along airflow)

4000 mm (4 × D)

4

Breadth and Height (normal to airflow)

2000 mm (2 × D)

5

Distance of rotor zone from inlet

1500 mm (1.5 × D)

6

Stator element size

64 mm

7

Maximum stator element size

128 mm

8

Rotor element size

15 mm

c) Mess Element Size:

Table 3. Model parameters. S. No.

Parameter

Property

1

Space

3D

2

Time

Unsteady, 2nd order implicit

3

Viscous model

Realizable, k-epsilon turbulence model

4

Wall treatment

Scalable wall function

4.2 Results and Discussions The full-scale model of the propeller is considered for simulation in Ansys Fluent 2021. Different numerical simulations were conducted for the same mesh to compare the accuracy of the result obtained. These were then compared with the experimental values.

Future Mobility with eVTOL Personal Air Vehicle (PAV)

331

Fig. 7. eVTOL propeller

Fig. 8. Computational domain of eVTOL thrusters.

Fig. 9. Mesh generation of eVTOL thrusters.

Aerodynamic Force Measurements The thrust, torque, and power consumed by the propeller to maintain a steady state were plotted as shown in Figs. 10, 11, and 12, respectively. Torque was considered for the analysis instead of drag force as the net drag force at every instant cancels out due to the symmetric rotation of the propeller. The propeller’s torque is a cross product of the distance from a point on the propeller to the center and the drag at the point. A

332

M. N. P. Babu et al.

quadratic thrust and torque and a cubic plot for power consumption have been observed w.r.t. angular velocity of the propeller. The thrust force measured in the CFD study has a higher deviation from the experimental value than the torque. The thrust and torque values of the propeller at angular velocities of 3200 and 2000 are 194.633 N and 14.228 N-m, and 60.363 N and 5.176 N-m, respectively. The thrust and torque values obtained from the experiment for the same angular velocities are 283.22 N and 14 N-m, and 105.84 N and 5 N-m, respectively. This deviation can be attributed to an incomplete flow mapping inside the domain and occurs due to a smaller domain size considered for evaluation. The inlet velocity considered needs to be the same as the velocity at 1.5m from the propeller. The experimental and numerical values obtained for the torque and power consumed are observed to be similar. The contours of static pressure, velocity magnitude, dynamic pressure, and Reynolds number were analysed on the propeller and along the inlet flow to determine the varying properties and the possibilities for reducing thrust values.

Fig. 10. Propeller thrust force vs angular velocity of eVTOL vehicle.

Fig. 11. Propeller torque vs angular velocity of eVTOL vehicle.

Future Mobility with eVTOL Personal Air Vehicle (PAV)

333

Fig. 12. Propeller power vs angular velocity of eVTOL vehicle.

Static Pressure Static pressure is the pressure measured in the fluid when the fluid is expected to be at rest w.r.t. the measurement. Mathematically, static pressure is the total velocity of the fluid less the pressure due to the velocity of the fluid. Static pressure is often considered as the pressure of the fluid. The contours observed for Static Pressure are depicted in Fig. 13. A high static pressure region is expected to occur on the leading edge of the propeller as the flow is expected to be stationary w.r.t. the propeller causing an increase in the pressure. The upper surface of the propeller has the least static pressure as the velocity of the airflow is the highest at the surface w.r.t. the domain. The lower surface has a higher static pressure than the upper surface leading to a lesser airflow velocity. The pressure difference above and below the propeller may lead to a possible vortex at the propeller’s tip generating a reduced thrust. Velocity Magnitude The contours of the velocity magnitude observed on the propeller along the front and side views are described in Fig. 14. As the angular velocity of the propeller varies from 1000 RPM to 3200 RPM, the inlet velocity changes from 8 m/s to 10 m/s. The maximum velocity is observed near the tip and the low velocity near the centre of the propeller, as shown in Fig. 14. The velocity magnitude observed beneath the propeller is expected to be lesser than the upper surface leading to a pressure variation above and below the propeller generating lift. Other observations include a decrease in the divergence flow from the centre as the flow exits the propeller as the angular velocity of the propeller increases.

334

M. N. P. Babu et al.

Fig. 13. Static pressure at (a) 1000 rpm (b) 2000 rpm and (c) 3200 rpm

Reynolds Number and Dynamic Pressure Reynolds number of the flow is the ratio of inertial forces to the viscous forces calculated when the fluid is subjected to an interruption or blockage, leading to a variation in the fluid’s density, velocity, or viscosity. Dynamic pressure is the pressure occurring due to the velocity of the fluid. It is the kinetic energy of the fluid per unit volume. Mathematically, Reynolds number is the product of density, velocity, and diameter of channel per unit viscosity. Dynamic pressure is the product of density and half the square of velocity. The dynamic pressure and the Reynolds number contours are depicted in

Future Mobility with eVTOL Personal Air Vehicle (PAV)

335

Fig. 14. Velocity magnitude of the propeller at (a) 1000 rpm (b) 2000 rpm (c) 3200 rpm.

Fig. 15. As the angular velocity increases, the dynamic pressure and the Reynolds number of the flow are expected to increase, as illustrated. Other contributors to an increase in the Reynolds number include variation in the density and velocity due to vortices and stagnation regions at the tip and leading edges of the propeller. Other observations include a possible increase in the reversal of flow after the airflow exits the propeller as the angular velocity of the propeller decreases. Some of this reversed flow renters the propeller forming a circular helix-like pattern near the propeller’s tip.

336

M. N. P. Babu et al.

Fig. 15. Dynamic pressure at (a) 1000 rpm (b) 2000 rpm and (c) 3200 rpm; Re at (d) 1000 rpm (e) 2000 rpm and (f) 3200 rpm.

5 Summary and Conclusions Ground transportation in large cities has been facing issues for many years (e.g., resilience and congestion). New paradigms, such as the Urban Air Mobility (UAM) idea, being proposed have been offering hope to decongest the transportation and save time spent in traffic. UAM, in comparison, is low-cost and helpful in intra-city transportation. This study involves the study of a new mode of UAM transportation with PAV vehicles. The experimental and numerical simulation results of eVTOL vehicle are presented in this study. From the numerical analysis of the propeller, a deviation of 30% from the experimental value has been observed, which could be due to the smaller domain size and a larger mesh element size considered. A smaller mesh size compromises various aspects such as vorticity and flow reversal. The inlet velocity is expected to be equivalent to the magnitude of the velocity of the fluid at the same distance under actual operating conditions for accurate results. The possibility of a flow reversal behind the propeller and vortices are the other factors needed to be considered extensively to obtain values equivalent to the experimental conditions. Other aspects, such as stressstrain analysis of the propeller and frame, the effect of different configurations such as counter-rotating, in-plane propellers, are to be considered before full-scale production of the UAM system. The propeller’s experimental and numerical simulation results show

Future Mobility with eVTOL Personal Air Vehicle (PAV)

337

that it can produce sufficient aerodynamic forces for hovering and cruise conditions. So, an eVTOL with such a propeller can be helpful in actual UAM operations. Acknowledgement. This work is supported by DST National Mission Interdisciplinary CyberPhysical Systems (NM-ICPS), Technology Innovation Hub on Autonomous Navigation and Data Acquisition Systems: TiHAN Foundation at Indian Institute of Technology (IIT) Hyderabad.

References 1. Thipphavong, D.P., et al.: Urban air mobility airspace integration concepts and considerations. In: 2018 Aviation Technology, Integration, and Operations Conference 2018, p. 3676 (2018) 2. Elevate, U.: Fast-forwarding to the future of on-demand, urban air transportation [R/OL], pp. 10–27 (2016) 3. Airbus-A3, Vahana Aero (2018). https://vahana.aero/. Accessed 14 Nov 2018 4. Urban Air Mobility Takes Off in 64 Towns and Cities Worldwide (2018). https://www.unm annedairspace.info/urban-air-mobility/urban-air-mobility-takes-off-63-towns-cities-worldw ide/. Accessed 18 Jan 2019 5. Urban Air Mobility (UAM) Market Study, Booz Allen Hamilton (2018). https://www. nasa.gov/sites/default/files/atoms/files/bah_uam_executive_briefing_181005_tagged.pdf/. Accessed 18 Jan 2019 6. Snyder, C.A.: Personal rotorcraft design and performance with electric hybridisation (2017) 7. Kim, H.D., Perry, A.T., Ansell, P.J.: A review of distributed electric propulsion concepts for air vehicle technology. In: 2018 AIAA/IEEE Electric Aircraft Technologies Symposium (EATS), 12 July 2018, pp. 1–21. IEEE (2018) 8. Prevot, T., Rios, J., Kopardekar, P., Robinson III, J.E., Johnson, M., Jung, J.: UAS traffic management (UTM) concept of operations to safely enable low altitude flight operations. In: 16th AIAA Aviation Technology, Integration, and Operations Conference 2016, p. 3292 (2016) 9. E.V. News: Ehang 216 (2020). https://evtol.news/aircraft/ehang-216/ 10. E.V. News: Wisk (Kitty Hawk) Cora (2020). https://evtol.news/aircraft/kitty-hawk-cora/ 11. E.V. News: Lilium jet (2020). https://evtol.news/aircraft/lilium/

Dam Water Discharge and Flood Prediction Cum Warning System V. Anantha Narayanan and Prashant R. Nair(B) Department of Computer Science and Engineering, Amrita School of Engineering, Amrita Vishwa Vidyapeetham, Coimbatore, India [email protected], [email protected]

Abstract. To create an autonomous system that warns people during discharge of dam water predicts the possibility of a flood (based on weather forecast) due to overload of water in a dam’s reservoir and warns people in advance. The system uses a NodeMcu module to read data from water level sensors placed in the reservoir of a dam. Change in water level is indicated to a sensor cloud (Ubidots). A website uses this data along with a weather forecast for that area from Yahoo! Weather API to predict the possibility of a flood over a week’s time frame. This website can be used by tourists to plan their trips to the dam. Discharge of water from the dam is detected by a proximity sensor placed in the control gate of the dam which is also connected to the NodeMcu module. When the proximity sensor comes on, it triggers a series of LED lights and buzzers placed along the discharge area of the dam to alert locals about the discharge of water from the dam. The system sends a warning SMS to people living near the dam when the water level in the dam reaches the critical limit. The system triggers the local warning system as and when the control gate is opened in the prototype. Water level changes are intimated to the sensor cloud and a warning SMS is sent when water reaches the critical level in the reservoir of the dam. Under this condition, if the control gate is also opened, an emergency SMS is sent asking people to evacuate immediately. This system can help save the lives of people living near dams by warning them well before the occurrence of a flood. Keywords: Wi-Fi · Cloud · Dam warning system · Weather

1 Introduction Dams are an important source of power generation. Unlike fossil fuels, hydropower is clean and is renewable. Thus, dams play a major role in providing electricity for the functioning of various devices and equipment on the daily basis. Dams are also some of the major places of recreation with high tourist activity. Discharges of water from the dam due to water needs are common. But, such a discharge without proper warning causes a sudden increase in the flow of water along the bank due to which people can get washed away causing severe injuries or even death [1]. Further, when a dam gets full, water needs to be discharged to prevent the dam from breaking, and often there isn’t enough time to provide such a warning. This project aims to develop an autonomous © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 338–345, 2022. https://doi.org/10.1007/978-981-19-1742-4_28

Dam Water Discharge and Flood Prediction Cum Warning System

339

system that can detect the increase in the flow of water and provide a warning [2]. Further, this system also predicts the possibility of a flood and sends out warnings through a short messaging service well in advance [3, 4]. A website is also included in this system which can predict the possibility of a flood that can be caused by heavy rainfall near the dam area. This website can help tourists to plan their trips accordingly [5, 6].

2 Method The proposed solution uses a set of interconnected devices with an internet backbone that works autonomously without human intervention (Internet of Things). The possibility of flooding is estimated based on weather forecast over a week and water level data from sensor cloud based Predictive Analytics [7, 8]. 2.1 Proposed Solution 2.1.1 Architecture See Fig. 1.

Fig. 1. Architecture

1 – Local warning system comprising of LED lights and buzzers connected to ESP8266 to trigger a warning during discharge of dam water. 2 – Wi-Fi router to connect the system to the internet. 3 – Ubidots: to act as central storage (cloud) so that the website can get data about the water level in a dam whenever a user requests it. The proposed solution aims to solve problems arising out of three conditions: a. Discharge of dam water without a warning leading to the sudden increase in outflow of water. This sudden force can wash away people taking a dip in the river.

340

V. Anantha Narayanan and P. R. Nair

Solution: A proximity sensor is placed on the dam to detect the opening of the dam and triggers the local warning system when the control gate is opened. b. Filling of the dam due to excessive rainfall/inflow of water and discharge of water to prevent breakage. Manual warning may not give people enough time to move their valuables and take themselves to safety. Solution: Using a set of water level sensors, the water level in a dam is constantly monitored. Change in water level is notified to the sensor cloud. When the water reaches a threshold level, a ‘warning’ SMS is sent to some pre-registered phone numbers notifying them about the possibility of flooding. Under this condition, if the water is also discharged from the dam, the local warning system also gets activated and a ‘danger’ SMS is sent to the pre-registered phone numbers asking the people to evacuate immediately. c. Prediction of the possibility of a flood can help tourists to plan their trip accordingly. Solution: A website is designed to communicate with Ubidots and Yahoo! Weather API to periodically get data about the existing water level in a dam and weather forecast of the area for 5 days. This data can be used to predict the possibility of a flood (Fig. 2). 2.2 Hardware 2.2.1 Sensor Placement

Fig. 2. Implementation of the system – placement of sensors

• Water level sensors are placed in the reservoir of the dam to detect the existing level of water in the dam. Changes in water level are updated in Ubidots. Prevention of false alarms is taken care of on the software side. • The proximity sensor detects the opening of the dam and switches on the local alarm system to warn people on the shore.

Dam Water Discharge and Flood Prediction Cum Warning System

341

Water level sensors are placed in such a way that one of them is on the danger line and the other water level sensor is a little below that. The proximity sensor is placed on the control gate to detect the opening of the gate [9, 10]. 2.2.2 NodeMcu Connection (ESP8266) The two water level sensors are connected to the NodeMcu Wi-Fi module through GPIO pins. A proximity sensor is also connected through a GPIO pin. The local alarm system is also connected to the NodeMcu module through a GPIO pin [11, 12]. 2.3 Software The water level sensors and proximity sensor reads are done using GPIO pins in the ESP8266 module. Triggering of the proximity sensor switches on the local alarm system. Water level changes are updated to Ubidots through a POST request using an authentication token and the variable ID associated with the dam. SMS is sent through Ubidots’ ‘events’ feature [13]. The website gets water level information from Ubidots through a GET request. The website also gets weather information from Yahoo! weather API. This weather information contains a forecast for 5 days which is useful in predicting the increase in water level and thus the possibility of flooding. When the water level is below the level of the lowest placed water level sensor, the place is considered safe for a week. When the water level is between the levels of the two water level sensors, the weather forecast is used to determine the possibility of a flood in the area. When the water level is above the critical limit, the place is considered unsafe to visit for a week. • Avoid false alarms – Water level changes are updated to Ubidots only when a change in sensor reading is retained constantly for more than 3 s. • Buzzers can be switched off 2–3 min after the triggering of the local alarm system. This saves power and prevents disturbance to the surroundings. 2.4 Working See Fig. 3. The water level in the dam reservoir is constantly monitored by the ESP8266 module. When water level changes with respect to the water level sensors, an update is made in Ubidots’ variable associated with the dam (Table 1). When the water level rises above the highest placed water level sensor (Ubidots value: 2.0), a “warning” SMS is sent via Ubidots to notify people about the possible occurrence of a flood. When the dam is opened for discharge under this condition (detected by proximity sensor), an “emergency” SMS is sent via Ubidots asking people to evacuate immediately (Ubidots value: 3.0) (Fig. 4). Under normal conditions when the dam is opened for water discharge, the proximity sensor detects the opening and triggers the local alarm system comprising of buzzers and LED lights which helps in warning people near the shore.

342

V. Anantha Narayanan and P. R. Nair

Fig. 3. High-level working – 1

Table 1. The value sent to Ubidots based on the water level. Water level

Value sent to Ubidots

Below both the water level sensors – Low

0.0

Between the two water level sensors – Mid

1.0

Above both the water level sensors – High

2.0

High + proximity sensor gets triggered

3.0

Fig. 4. High-level working – 2

The website gets water level data from Ubidots and weather forecast from Yahoo! Weather API. If the water level is below the two water level sensors (Ubidots value: “0.0”), the dam is considered safe to visit for the next 5 days. If the water level is above the two water level sensors (Ubidots value: “2.0”), the dam is considered unsafe to visit for the next 5 days. If the water level is between the two water level sensors (Ubidots value: “1.0”), weather forecast data from Yahoo! Weather is used to determine the safety

Dam Water Discharge and Flood Prediction Cum Warning System

343

of the dam. If the weather forecast for the next 5 days predicts rainfall for 4 days or more, the dam is considered unsafe to visit for the next 5 days. If the forecast predicts rainfall for 3 days or less, the dam is considered safe to visit [14, 15] (Figs. 5, 6, 7 and 8).

Fig. 5. Safety estimation in website

Fig. 6. Data stored in Ubidots

344

V. Anantha Narayanan and P. R. Nair

Fig. 7. Warning SMS

Fig. 8. Website – results

3 Conclusion and Results Based on the prototype developed, the system can provide a warning about the possibility of a flood in an area in advance. Based on the area of the reservoir of a dam, the placement of water level sensors may change to detect the possibility of a flood in advance with high accuracy [16, 17].

References 1. Mioca, D., et al.: Early warning and mapping for flood disasters. Int. Arch. Photogram. Remote. Sens. Spatial. Inf. Sci. XXXVII(B4), 1507–1509 (2008) 2. Hosler dam – Early warning system. http://www.ashland.or.us/Files/hosler%20dam.PDF 3. Vasudevan, N., Ramanathan, K.: Geotechnical characterization of a few landslide-prone sites in India. In: Sixth International Geotechnical Symposium on Disaster Mitigation in Special Geo-environmental Conditions, Indian Institute of Technology (2015) 4. Divya, P., Geethu, T., Ramesh, M.V., Geethalekshmy, V.: Automated statistical data mining of a real world landslide detection system. In: DMIN 2014 -The 10th International Conference on Data Mining, Las Vegas, USA, July 2014 5. Federal Guidelines for Dam Safety: FEMA, April 2004 6. Integrated flood forecasting, warning and response system. http://www.un.org/esa/sustdev/ publications/flood_guidelines_sec03.pdf 7. Harrington, B.W.: Hazard classifications & danger reach studies for dams. Hazard classification guidelines document

Dam Water Discharge and Flood Prediction Cum Warning System

345

8. Chakraborty, R., et. al.: Study and prediction analysis of the employee turnover using machine learning approaches. In: 2021 IEEE 4th International Conference on Computing, Power and Communication Technologies (GUCON), pp. 1– 6 (2021). https://doi.org/10.1109/GUCON5 0781.2021.9573759. 9. Kalamandeen, A., Scannell, A., de Lara, E., Sheth, A., La Marca, A.: Ensemble: cooperative proximity-based authentication. In: MobiSys 2010, 15–18 June 2010, San Francisco, California, USA (2010) 10. Shafagh, H., Hithnawi, A.: Poster: come closer - proximity-based authentication for the Internet of Things. In: MobiCom 2014, 07–11 September 2014, Maui, HI, USA (2014) 11. Berger, R.: Introduction to wireless sensor networks. In: NI Technical Symposium (2009) 12. What the Internet of Things (IoT) Needs to Become a Reality. ARM and Freescale. www. arm.com/freescale 13. Evans, D.: the internet of things how the next evolution of the internet is changing everything. Cisco IBSG. 1, 1–11 (2011) 14. Rajawat, A., Rawat, R., Mahor, V., Shaw, R., Ghosh, A.: Suspicious big text data analysis for prediction—on Darkweb user activity using computational intelligence model. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 735–751. Springer, Singapore (2021). https://doi.org/10. 1007/978-981-16-0749-3_58 15. Parthasarathi, V., Surya, M., Akshay, B., Murali Siva, K., Vasudevan, S.K.: Smart control of traffic signal system using image processing. Indian J. Sci. Technol. 8(21), 1 (2015). (Article No. 70468) 16. Sn, A., Vasudevan, S.K., Nair, P.R., Thangavelu, S., Sundaram, R.M.: A proposal for mitigating fishermen killing in Indian sea borders through technology – maritime boundary identification device. Indonesian J. Elect. Eng. Comput. Sci. 6, 704–710 (2017) 17. Prathilothamai, M., Prashant, R., Nair, R., Singh, A.P., Aditya, P.N.S.: Offline navigation: GPS based location assisting system. Indian J. Sci. Technol. 9(45), 1–6 (2016)

Virtual Inertia Control Strategy for High Renewable Energy-Integrated Interconnected Power Systems Anuoluwapo Aluko1(B) , Rudiren Pillay Carpanen3 , David Dorrell2 , and Evans Ojo3 1

2

Discipline of Electrical Engineering, University of KwaZulu-Natal, Durban 4001, South Africa [email protected], [email protected] School of Electrical and Information Engineering, University of the Witwatersrand, Johannesburg 4041, South Africa [email protected] 3 Department of Electrical Power Engineering, Durban University of Technology, Durban 4000, South Africa [email protected]

Abstract. With the growing penetration level of renewable energy (RE) systems, the overall inertia of the power system is expected to reduce to values that expose the system to inadvertent frequency eventuality that can threaten the stability, security, and resilience of the system. To tackle this issue, this paper proposes an advanced virtual inertia control strategy in an interconnected power system which considers high renewable energy penetration and power system deregulation. The virtual inertia control is capable of providing dynamic inertia support by adjusting the active power reference of the power electronic converter of an energy storage system (ESS). This improves the response and stability of the system during frequency events. The proposed virtual inertia control strategy is based on the type-II fuzzy logic control scheme. To improve the accuracy and performance of the proposed controller, the artificial bee colony algorithm is used for optimal tuning of the input and output weights of the type-II fuzzy-based virtual inertia control. The proposed control strategy is designed to competently perform during load variation, RE fluctuations, and other power system dynamic disturbances. The simulation results show its robustness in minimizing the frequency deviation and maintaining system frequency within specified operating limits.

Keywords: Frequency control Virtual inertia

· Fuzzy logic · Renewable energy ·

This work is supported by the Council for Scientific and Industrial Research through the Smart Networks Initiative that is funded by Department of Science and Innovation under Grant K9DSEIF.11214.05400.054RC.UNI. c The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022  S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 346–364, 2022. https://doi.org/10.1007/978-981-19-1742-4_29

Virtual Inertia Control Strategy

1

347

Introduction

Major changes have been introduced in the structure of electric power utilities around the world to improve the efficiency and reliability of power systems [1]. Power system deregulation and increased renewable energy (RE) integration are the major factors in the restructured power system environment. In this environment, control is greatly decentralized and Independent System Operators (ISOs) are responsible for maintaining the system frequency and tie-line power flow amongst other ancillary services [2]. Generation utilities may or may not take part in automatic generation control (AGC) or load frequency control (LFC); these make the task of frequency control more complicated to achieve. Several works have reported various approaches for LFC design when considering power system deregulation to improve the frequency response of the system [3–5]. In a similar way to power system deregulation, the structure of the emerging power system that incorporates RE systems, such as wind and solar plants and other distributed energy resources, is different from the conventional power system. The key technical concern that affects the increasing penetration level of RE systems is associated with their capability in performing as efficiently and consistently as conventional generation systems. This is because they affect the dynamic behavior and response of large power systems differently from their conventional counterparts [6]. Despite the rapid technological advancements in the RE industry, the task of frequency regulation remains a challenge, and this constitutes the limitation of increasing RE share in modern power systems. The study conducted in [7] predicted that with the current growth rate of RE penetration, the total system inertia would reduce significantly to values that threaten system frequency recovery and stability during power imbalance and other power system contingencies [8]. Therefore, technical methods for providing or emulating inertia need to be developed. The virtual inertia control strategy is a promising solution in emulating the inertia characteristics of a prime mover by deploying an appropriate control algorithm in the power electronics converter of a dedicated energy storage system (ESS) [9]. Several control strategies have been adopted in the design of a virtual inertia control: the authors in [10] proposed a coefficient diagram method; a dynamic equation and adaptive fuzzy technique were proposed in [11], a derivative-controlled solar and ESS was proposed in [12], and a model predictive control (MPC) approach was designed in [13] to improve the frequency regulation capability of low-inertia microgrid systems. The artificial bee colony (ABC) algorithm was used to tune the derivative-virtual inertia control for wind energy systems in [14]. Reference [15] adopted the chicken swarm optimization algorithm in tuning the parameters of an adaptive virtual inertia controller, and the particle swarm optimization (PSO) was used in [16]. Some artificial intelligence techniques have found application in the design of virtual inertia systems; a reinforcement learning-based approach was used in [17] and [18] to design a virtual inertia controller for low-inertia power systems. In [19], a radial basis function neural network was trained to implement a virtual synchronous generator. Author of [20] implemented a virtual inertia controller using the wavelet

348

A. Aluko et al.

fuzzy neural network. These control methods that apply artificial neural networks are largely dependent on the availability of data to sufficiently train the model, therefore, such control methods might result in over-fitting if the training data is insufficient, thereby compromising the performance of the control system. The fuzzy logic control is an intelligent control technique that can been used to improve performance in terms of virtual inertia control because of its high computational efficiency in handling nonlinear complexities that are characterized with modern power systems [21]. In [22], the type-I fuzzy logic was proposed for virtual inertia control of an interconnected microgrid with high RE penetration. Reference [23] augmented a low-inertia wind farm with a type-I fuzzy-based ESS to improve the primary frequency response of the system. While the type-I fuzzy system has been successfully implemented, it is less efficient in handling different power quality threshold requirements and uncertainties. Based on this limitation, the type-II fuzzy offers an extra degree of freedom to model the uncertainties that cannot be quantified by two-dimensional MFs of the type-I fuzzy system [24]. References [25,26] proposed the type-II fuzzy logic system to mitigate LFC problems in RE-integrated power systems. Based on the literature, no previous work has reported the development of advanced control strategies for virtual inertia emulation in large power systems with high penetration of RE systems in the deregulated environment. Therefore, the contributions of this paper are: the design of a new type-II fuzzy-based virtual inertia control strategy to improve the frequency response and stability of power systems; the performance of the proposed type-II fuzzy-based virtual inertia controller is improved by tuning its weights with the ABC optimization algorithm, the cost function used in the optimization process is formulated to be model-dependent such that the proposed virtual inertia system does not behave as a continuous infinite energy source; the proposed optimal type-II fuzzy-based virtual inertia control strategy is implemented in the emerging power system environment which considers high RE penetration and deregulation. The remainder of this paper is structured as follows: Section 2 presents the small-signal modelling of power system for frequency response analysis in the restructured environment, Sect. 3 presents the concept of inertia and the proposed optimized type-II fuzzy logic for inertia emulation, Sect. 4 discusses the simulation results of the proposed control strategy and Sect. 5 presents the conclusion.

2 2.1

System Modeling Dynamics of Power System Deregulation

In the formulation of the frequency response model of the interconnected power system in the restructured environment, it is important to include the dynamics of the open-energy market scenarios that occur between power vendors– generating companies (GENCOs) and vendees–distribution companies (DISCOs). For an interconnected power system with N control areas, the contract

Virtual Inertia Control Strategy

349

participation matrix (CP M ) between GENCOs and DISCOs can be defined as ⎡ ⎤ cpf11 cpf12 · · · cpf1N ⎢ cpf21 cpf22 · · · cpf2N ⎥ ⎢ ⎥ (1) CPM = ⎢ . .. . . .. ⎥ ⎣ .. . . . ⎦ cpfn1 cpfn2 · · · cpfnN where cpfnN is the contract participation factor of the nth GENCO of the N th area. The sum of the elements in each column of the CPM must be equal to 1. Furthermore, there could exist a scenario where there is exchange of power between interconnected areas via tie lines at scheduled values during steadystate operation. In this scenario, the GENCOs in Area i are in a power purchase contract with the DISCOs is Area j; therefore, the non-diagonal elements of the CPM are non-zero. The scheduled tie line power flow to Area i for N interconnected control areas can be written as ⎛ ⎞ n

N n N     ⎜ ⎟ sch = cpfkj ΔPDc,j − cpfjk ⎠ ΔPDc,i (2) ΔPtie,i ⎝ j=1 j=i

k=1

k=1

j=1 j=i

where ΔPDc,i and ΔPDc,j are the total power demand (contracted) of DISCOs in Areas i and j respectively, given that {i, j ∈ N }. 2.2

Small-Signal Modelling of Modern Power System

In this subsection, the frequency response model of a modern power system is presented using small-signal derivations. The system model takes into account the deductions from the previous subsection. The frequency response model shown in Fig. 1 represents the small signal derivation of the ith control area of an interconnected power system in the restructured power system environment. The frequency response of Area i resulting from the net active power generation and demand is Δfi =

1 (ΔPmi + ΔPresi − ΔPtie,i − ΔPDi ) 2Hi s + Di

(3)

where H is the total inertia constant of the synchronized rotating masses, D is the damping parameter, ΔPm is the total change in mechanical power of the GENCOs, ΔPres is the total change in power of the RE systems, ΔPtie is the change in tie line power flow, and ΔPD is the total load change in Area i. The change in output mechanical power of the GENCOs in the ith area is given by n n   1 ΔPgj ΔPmj = (4) ΔPmi = sTtj + 1 j=1 j=1 where Tt is the time constant of the turbine and ΔPg is the change in input setpoint to the turbine, i.e., adjustment of speed governor set point. It is important

350

A. Aluko et al.

Fig. 1. Frequency response model of a power system in the restructured environment.

to note that unlike the traditional power system environment, the dynamics of the speed governor must include a setpoint resulting from power system deregulation (contracted power) [27]. Therefore, the response of the speed governor of the GENCO in the restructured system can be obtained from   −1 1 Δf + σΔPlf c + ΔPc (5) ΔPg = sTg + 1 R where Tg is the time constant of the governor, R is droop parameter for primary  frequency control, σ is the generator participation index (σ ∈ {0, 1} and σ= 1) for the GENCOs in Area i participating in LFC while ΔPlf c is the change in control signal from the secondary frequency controller and ΔPc is the change in the set point due to deregulation. The ΔPc for the nth GENCO in the Area i can be expressed as N  ΔPci,n = cpfn,i ΔPDi (6) i=1

The LFC action for the restructured power system is different from the conventional power system because it takes into formulation the scheduled inter-area sch . The LFC action for Area i can be deduced from power flow, ΔPtie ΔPlf c,i =

−Ki (βi Δfi + ΔPtie,i ) s

(7)

where Ki and βi are the integral gain of the LFC and frequency bias of Area i respectively, and err sch = ΔPtie,i + ΔPtie,i (8) ΔPtie,i

Virtual Inertia Control Strategy

351

For the hybrid tie line interconnection considered in this study, inter area power flow to Area i from Area j is given   2πTij Kdc + (9) (Δfi − Δfj ) ΔPtie,i = s 1 + sTdc where Tij is the synchronizing coefficient of the ac tie line between Areas i and j, and Kdc and Tdc are the gain and time constant of the dc line between Areas i and j respectively. With the growing share of RE systems in the modern energy mix, they are to contribute to frequency control according to the grid code specifications. For the ith area in N interconnected control areas, the change in total output power of the RE system resulting from droop control and fluctuation can be derived using   n  1 1 var ΔPresj = Δfi + ΔPres (10) ΔPres,i = 1 + sTresj Kresj j=1 var is where Tresj is the time constant, Kresj is the droop parameter and ΔPres the variation in output of the RE system. Finally, the change in demand in Area i that impacts the frequency response of the system is the summation of contracted and non-contracted demand can be mathematically expressed as

ΔPDi = ΔPDc,i + ΔPDuc,i

(11)

The equations derived in (3) to (11) can be used to obtain the state space model for small signal stability analysis. For the ith control area, we can write the state-space model as X˙ i = Ai Xi + Bi Ui (12) Yi = Ci Xi where A, B, and C are the system, input, and output matrices respectively with appropriate dimensions while X, U are state and input vectors respectively that are given as Xi = [Δfi ΔPmi ΔPgi ΔPtie,i ΔPres,i ]T (13) Ui = [ΔPlf c,i ΔPDi ]T

3 3.1

Proposed Virtual Inertia Control Strategy Virtual Inertia

The virtual inertia concept is proposed to emulate the damping and inertia characteristics of conventional synchronous generators, therefore maintaining or increasing the overall inertia of the system. This facilitates the increased penetration of RE systems into the energy mix without compromising the frequency stability of the system. In this work, virtual inertia control as a new ancillary service in which inertia emulation is achieved using the combination of an ESS,

352

A. Aluko et al.

Fig. 2. Small-signal model of virtual inertia strategy.

power electronic converter, and an appropriate control strategy is proposed. The ESS acts as an inertia unit that injects instantaneous active power into the system to reduce the rate of change of frequency (RoCoF) and frequency deviation. The derivative control strategy is the kernel of the virtual inertia control, it can modify the output of the ESS depending on the RoCoF during a frequency event or disturbance. The block diagram of the derivative-based virtual inertia control is presented in Fig. 2. The output of the ESS is constricted between the minimum and maximum active power rating of the ESS which gives the practical operating condition of the ESS and ensure it does not behave as an infinite energy source. The equation for the dynamic inertia emulation using the derivative control strategy can be derived as:   1 d(Δf ) (14) · ΔPvi = Hvi 1 + sTess dt where ΔPvi is the virtual power that contributes to the inertia response of the system, Tess is the time constant of the low pass filter acting as ESS converter, and Hvi is the virtual inertia gain. The choice of Hvi is very important because it significantly affects the response of the virtual inertia controller. Large values of Hvi will repress the frequency deviation with smaller overshoot and undershoot but increase the settling time of the frequency while smaller values of Hvi will quickly restore the frequency of the system but with high oscillation and amplitude of deviation. This makes the conventional virtual inertia control strategy with fixed inertia gain perform poorly under dynamic frequency events. Therefore, this work proposes the application of the type-II fuzzy logic system to calculate the appropriate inertia gain, thus developing an adaptive virtual inertia controller that can perform efficiently and optimally under various operating conditions. 3.2

Type-II Fuzzy Logic System

The control structure has three basic stages: input processing (fuzzification) that converts the crisp input to fuzzy input; an inference engine that determines the fuzzy output based on some complex calculations and rule base; and output processing (type-reduction and defuzzification) that reduces and converts the type-II fuzzy output to crisp output [28]. The type-II fuzzy set (FS) F˜ where l ∈ L and can be characterised as F˜ = {((l, m), μF˜ (l, m)) | ∀ l ∈ L, ∀ m ∈ Jl ⊆ [0, 1]}

(15)

Virtual Inertia Control Strategy

353

where μF˜ (l, m) is a type-II membership function (MF), l is a primary variable, m is the secondary variable in which 0 ≤ μF˜ (l, m) ≤ 1 and Jl is the primary membership of l. F˜ can also be expressed as   F˜ = μF˜ (l, m)/(l, m) Jl ⊆ [0, 1] (16) l∈L

m∈Jl



where denotes union over all possible l and m. When there are no uncertainties, the type-II FS reduces to a type-I FS such that the secondary variable becomes μF (l) and 0 ≤ μF (l) ≤ 1. When all μF˜ (l, m) = 1, F˜ is said to be an interval type-II FS [29]. The interval type-II FS reduces (16) to   F˜ = 1/(l, m) Jl ⊆ [0, 1] (17) l∈L

m∈Jl

Due to the uncertainty associated with the interval type-II FS, the boundary of uncertainty in the type-II MF is defined as the footprint of uncertainty (FOU). The outer boundary (μ) of the FOU is the upper membership function (UMF) and the inner boundary (μ) is the lower membership function (LMF) so that FOU(F˜ ) =



Jl

(18)

μF˜ (l) ≡ FOU(F˜ )

∀l ∈ L

(19)

μF˜ (l) ≡ FOU(F˜ )

∀l ∈ L

(20)

l∈L

The rule structure for the type-II fuzzy logic is similar to the type-I fuzzy, for a fuzzy logic structure with a K number of rules, the nth rule can be expressed in the form: Rn : If l1 is F˜1,n and . . . li is F˜i,n . . . and lI is F˜I,n (21) Then y is Y˜n where n = (1, 2, ..., K). This type reduction block maps the type-II FLC into a type-I FLC by computing the centroid of the interval type-II FLC associated with each fired by using (22) YF˜ = 1/ {yp , . . . , yq } where yp and yq can be calculated using Q yp =

¯F˜ i=1 yi μ P ¯F˜ i=1 μ

Q yq =

i=1 yi μF¯ Q i=1 μF¯

(yi ) + (yi ) + (yi ) + (yi ) +

N i=P +1

yi μF˜ (yi )

i=P +1

μF˜ (yi )

N N

i=Q+1

yi μ ¯F˜ (yi )

i=Q+1

μ ¯F¯ (yi )

N

(23)

(24)

354

A. Aluko et al.

Fig. 3. Proposed virtual inertia control strategy.

The switching point P and Q in (23) and (24) are iteratively determined using the Karnik-Mendel (KM) algorithm [30]. The crisp output y is computed by the defuzzifier using 1 y(x) = [yp (l) + yq (l)] (25) 2 3.3

Adaptive Virtual Inertia Based on Type-II Fuzzy Logic Control

In this subsection, the proposed virtual inertia control method based on the type-II fuzzy system is developed as shown in Fig. 3. The inputs to the type-II ˙ . These fuzzy logic system are the frequency deviation, Δf and its derivative, Δf inputs are firstly normalized with the weights W1 and W2 respectively so that the final inputs to the type-II fuzzy logic system are Input 1 = W1 · Δf

(26)

˙ Input 2 = W2 · Δf

(27)

The normalized inputs are then fuzzified using the membership functions. In this work, the triangular membership function is used because of its ease of computation, few design parameters, and less complexity in practical realization when compared to other types of membership functions. Due to these reasons, seven triangular membership functions are chosen to define the universe of discourse of the type-II fuzzy system. They are defined using the linguistic variables: Negative Big (NB), Negative Medium (NM), Negative Small (NS), Zero (Z), Positive Small (PS), Positive Medium (PM), and Positive Big (PB). The corresponding membership parameters of the seven linguistic variables are determined from (23) and (24). The fuzzified inputs which are the embedded fuzzy sets within the FOU are processed in the inference engine to generate appropriate fuzzy

Virtual Inertia Control Strategy

355

Table 1. Fuzzy rule table Δf ˙ Δf PB PM PS Z NS PM NB

PB NB NB NB NB NM NS Z

PM NB NB NB NB NS Z PS

PS NB NB NM NM Z PS PM

Z NB NM NS Z PS PM PB

NS NM NS Z PM PM PB PB

NM NS Z PS PB PB PB PB

NB Z PS PM PB PB PB PB

outputs with the aid of the fuzzy rules given in Table 1. The inference engine uses the min and max methods for the Meet and Join operations respectively. The Meet and Join operations map the fuzzified inputs into fuzzy outputs with the strength of the fired rule(s). The output of the inference engine is a type-II fuzzy set that cannot be directly converted to the crisp output, that is, virtual inertia gain. The type-reducer is used to convert the type-II fuzzy set to type-I fuzzy set using the KM algorithm, the smallest and largest centroids are computed using (23) and (24). The type-I fuzzy output is then converted to the normalized virtual inertia gain using (25). The normalized virtual inertia gain is scaled with weight W3 to generate the actual inertia gain, Hvi . The values of the input and output weights (W1 , W2 , W3 ) are critical in the design and performance of the fuzzy logic system for the proposed application; therefore, the ABC optimization algorithm is employed to determine their values. Due to space limitation, detail of the ABC can be accessed in [14]. The cost function used in the optimization process is a two-fold objective function that computes the derivative of the frequency deviation and absolute frequency deviation. For the ith control area with virtual inertia capability, the cost function can be formulated as  T   |Δfi (n) − Δfi (n − 1)| 2 Cfi = |Δfi (n)| + (28) t n=1 where t is the time between the actual present and previous frequency measurement. It is specified as the sample time of the controller. The values of the weights for the type-II fuzzy controller are presented in Table ??. The proposed optimized type-II fuzzy-based virtual inertia control system can recalculate the inertia gain to track the frequency to a zero steady-state value. Therefore, frequency deviation in Area i given in (3) can be written as Δfi =

1 × (ΔPmi + ΔPresi − ΔPtie,i − ΔPDi + ΔPvi ) 2Hi s + Di

(29)

356

4

A. Aluko et al.

Simulation and Discussion

In this section, the efficiency of the proposed type-II fuzzy-based virtual inertia controller is validated through extensive simulation in MATLAB. The test system for the simulation is the conventional two-area power system model that is used for frequency stability studies [31]. The model is updated to include the dynamics of modern power systems such as hybrid tie line transmission, power system deregulation, and unequal area capacities. In Area 2 of the test system, one of the conventional plants has been replaced with a RE system to simulate high penetration of RE and reduced equivalent inertia; therefore, the proposed virtual inertia strategy is implemented in Area 2 and its robustness is evaluated against the conventional and type-I fuzzy-based virtual inertia methods proposed in [32] and [22] respectively, it is important to mention that the two methods in the literature were redeveloped to satisfactorily fit the system used in this work. 4.1

Eigenvalue Analysis

In this section, the stability of the test system is analyzed by inspecting the properties of its eigenvalues. For a system to be stable, all the real parts of its eigenvalues have to be negative (on the left half of the complex s-plane). Table 2. Eigenvalues of system model Mode Eigenvalue λ1

−17.2118

λ2

−17.7078

λ3

−13.0658

λ4

−1.3063+j7.0956

λ5

−1.3063−j7.0956

λ6

−1.1801+j3.8124

λ7

−1.1801−j3.8124

λ8

−3.9736

λ9

−3.2251

λ10

−0.9941+j0.5724

λ11

−0.9941+j0.5724

λ12

−0.7513

λ13

−0.4281

λ14

−0.1033

Virtual Inertia Control Strategy

357

In Table 2, the eigenvalues of the test system are presented, it can be observed that all the eigenvalues have a negative real part, therefore, indicating an asymptotically stable system. It can be seen that the modes λ1 and λ2 which corresponds to Δf1 and Δf2 respectively are in a better region of stability with high magnitudes in the real parts of their eigenvalues. The complex modes of the eigenvalues (λ4−7 and λ10,11 ) indicate that the system will oscillate before reaching steady-state during small perturbations. 4.2

Load Variation

In this scenario, the effectiveness of the proposed virtual inertia control strategy is demonstrated. A step load change of ΔPDuc = 0.1p.u in Area 2 at 10 s is simulated to illustrate the frequency response of the system during load variation. Figure 4(a) and (b) show the frequency deviation and derivative of frequency deviation in Area 2 under this contingency.

Fig. 4. (a) Frequency deviation in Area 2 (b) rate of change of frequency deviation in Area 2 under step load variation.

Generally, it can be observed that the frequency response of Area 2 is improved when the virtual inertia control is incorporated into the system compared to the absence of virtual inertia in the system. More specifically, the proposed type-II fuzzy-based virtual inertia control method outperforms the typeI fuzzy-based and conventional virtual inertia control methods with the least

358

A. Aluko et al.

Fig. 5. Frequency deviation in Area 1 under step load variation in Area 2.

amplitude of frequency deviation, transient excursions, and fastest recovery of the derivative of frequency deviation with damped oscillations and minimal overshoots. The performance of the proposed type-II fuzzy virtual inertia control strategy can be demonstrated by observing the change in active power of the ESS ΔPvi , as shown in Fig. 6. It can be seen that with the proposed virtual inertia strategy, the ESS is capable of injecting high instantaneous active power needed to suppress the frequency deviation when compared to other virtual inertia control methods. It is important to mention that virtual inertia elements are zero net energy devices during steady-state operation, this criteria must the considered in the design of the intelligent control systems to implement virtual inertia devices such that they do not become an infinite energy source. Due to the interconnection that exists between Areas 1 and 2, it is consequential that the perturbations in Area 2 affect Area 1 and vice versa. Therefore, it is important to observe the frequency response of Area 1 due to the contingency in Area 2 for this scenario as shown in Fig. 5. It can be observed while the presence of virtual inertia in Area 2 helps to reduce the frequency deviation amplitudes and oscillations, the proposed type-II fuzzy-based virtual inertia achieved the least amplitude of frequency deviation with a smooth transition to zero steady-state deviation.

Fig. 6. Output power of ESS under step load variation.

Virtual Inertia Control Strategy

4.3

359

High Renewable Energy Fluctuation

In this scenario, a more severe contingency is considered to show the robustness of the proposed virtual inertia control strategy. With the high penetration of RE systems, it is practical that the overall systems inertia and damping characteristics are affected, and with the high fluctuation of the RE sources, the frequency response of the system can be significantly affected.

Fig. 7. Net fluctuation of RE power in Area 2.

The high RE fluctuation in Area 2 as shown in Fig. 7 is used to replicate this contingency, and Figs. 8 and 9 show the behaviour under this contingency. From Fig. 8(a), it is shown that the frequency deviation in Area 2 fluctuates above and below zero steady-state due to the nature of the disturbance. It shows that the proposed type-II fuzzy-based virtual inertia control strategy has the least amplitude of frequency deviation when compared to the type-I fuzzy-based and conventional virtual inertia control strategies. The derivative of the frequency deviation shown in Fig. 8(b) shows the effectiveness of the proposed type-II fuzzy-based virtual inertia strategy in achieving faster response to keep the rate of change of frequency deviation within zero with the lowest transient excursion. This good frequency response is a result of the rapid absorption/charging (negative ΔPvi ) and injection/discharging (positive ΔPvi ) by the ESS as shown in Fig. 9(a). It shows that the proposed type-II fuzzy-based virtual inertia control strategy can efficiently adjust the virtual inertia gain, Hvi , to varying values as a function of the system contingency. The adjustment of the virtual inertia gain can be seen in Fig. 9(b). While the conventional virtual inertia control strategy has a fixed gain and the type-I fuzzy-based virtual inertia control strategy adjusts the virtual inertia gain around the fixed inertia gain, the proposed type-II fuzzy-based virtual inertia strategy can adjust the virtual inertia gain to values that improve the frequency response and stability of the system.

360

A. Aluko et al.

Fig. 8. (a) Frequency deviation in Area 2 (b) rate of change of frequency deviation in Area 2 under RE fluctuation.

4.4

Parameter Variation

In system design, inaccurate parameter estimations, the variation of system parameters with time, and the removal of some systems components can reduce the efficient performance of the system practically. Therefore, it is important to test the robustness of the control systems during design with parameter variations. In this scenario, the robustness of the proposed type-II fuzzy-based virtual inertia control strategy is tested by simulating the contingencies in Sect. 4.2 and 4.3 under these parameter variations in Area 2: R3 = −30%, D2 = −50%, H2 = −50%, Tt = +25%, Tg = +20%, K2 = −30%, and β2 = +25%. It can be observed from Fig. 10(a) that with the absence of virtual inertia, there is serious frequency deviation with high oscillations. With the conventional and type-I fuzzy-based virtual inertia control strategies, the frequency dip is improved by 32% and 30% respectively with noticeable oscillations. However, the application of the proposed type-II fuzzy-based virtual inertia control strategy significantly improves the frequency dip by approximately 70% with damped oscillations. Furthermore, the rate of change of frequency deviation rapidly settles to zero with the proposed virtual inertia strategy when

Virtual Inertia Control Strategy

361

compared to other virtual inertia control strategies as seen in Fig. 10(b). The improved performance in the frequency response is due to the adaptive property of the type-II fuzzy controller in selecting the virtual inertia gains as shown in Fig. 11(b) which helps to dynamically adjust the output active power of the ESS as shown in Fig. 11(a) when compared to the conventional and typeI fuzzy-based virtual inertia control strategies. It is shown that the proposed type-II fuzzy-based virtual inertia control strategy is robust so that it efficiently performs during parameter variations and maintains the frequency stability of the system in severe scenarios.

Fig. 9. (a) Output power of ESS (b) Virtual inertia gain under RE fluctuation.

From the results presented in the section, it can be appreciated that a properly designed conventional virtual inertia control strategy can perform as relatively efficiently as the type-I fuzzy-based virtual inertia control strategy; however, the proposed type-II fuzzy-based virtual inertia control strategy significantly outperforms both control strategies.

362

A. Aluko et al.

Fig. 10. (a) Frequency deviation in Area 2 (b) rate of change of frequency deviation in Area 2 under parametric variation.

Fig. 11. (a) Output power of ESS (b) Virtual inertia gain under parametric variation.

5

Conclusion

This paper presents a new virtual inertia control system using a type-II fuzzy logic control strategy. The proposed virtual inertia control strategy is used in the control loop of a dedicated ESS to improve the frequency response of a system with a high penetration of renewable energy sources in a restructured power system environment. The ABC optimization algorithm is utilized to tune the weights of the proposed controller to improve its performance. The robustness of the proposed controller is assessed under several operating scenarios and compared with the type-I fuzzy and conventional virtual inertia control strategies. The simulation

Virtual Inertia Control Strategy

363

results show the effectiveness of the proposed strategy to improve the frequency response of the system, thereby enhancing its stability and resilience.

References 1. Tyagi, B., Srivastava, S.: A decentralized automatic generation control scheme for competitive electricity markets. IEEE Trans. Power Syst. 21(1), 312–320 (2006) 2. Bevrani, H.: Robust Power System Frequency Control, vol. 85. Springer, Heidelberg (2014). https://doi.org/10.1007/978-0-387-84878-5 3. Donde, V., Pai, M., Hiskens, I.A.: Simulation and optimization in an AGC system after deregulation. IEEE Trans. Power Syst. 16(3), 481–489 (2001) 4. Sahu, R.K., Sekhar, G.C., Panda, S.: De optimized fuzzy PID controller with derivative filter for LFC of multi source power system in deregulated environment. Ain Shams Eng. J. 6(2), 511–530 (2015) 5. Alhelou, H.H., Hamedani-Golshan, M.-E., Zamani, R., Heydarian-Forushani, E., Siano, P.: Challenges and opportunities of load frequency control in conventional, modern and future smart power systems: a comprehensive review. Energies 11(10), 2497 (2018) 6. Kumar, V., Pandey, A., Sinha, S.: Grid integration and power quality issues of wind and solar energy system: a review. In: International Conference on Emerging Trends in Electrical Electronics and Sustainable Energy Systems (ICETEESES), pp. 71–80. IEEE (2016) 7. Hoke, A.F., Shirazi, M., Chakraborty, S., Muljadi, E., Maksimovic, D.: Rapid active power control of photovoltaic systems for grid frequency support. IEEE J. Emerg. Sel. Top. Power Electron. 5(3), 1154–1163 (2017) 8. Dreidy, M., Mokhlis, H., Mekhilef, S.: Inertia response and frequency control techniques for renewable energy sources: a review. Renew. Sustain. Energy Rev. 69, 144–155 (2017) 9. Tamrakar, U., Shrestha, D., Maharjan, M., Bhattarai, B.P., Hansen, T.M., Tonkoski, R.: Virtual inertia: current trends and future directions. Appl. Sci. 7(7), 654 (2017) 10. Ali, H., et al.: A new frequency control strategy in an islanded microgrid using virtual inertia control-based coefficient diagram method. IEEE Access 7, 16979– 16990 (2019) 11. Pradhan, C., Bhende, C.N., Samanta, A.K.: Adaptive virtual inertia-based frequency regulation in wind power systems. Renew. Energy 115, 558–574 (2018) 12. Saxena, P., Singh, N., Pandey, A.K.: Enhancing the dynamic performance of microgrid using derivative controlled solar and energy storage based virtual inertia system. J. Energy Storage 31, 101613 (2020) 13. Sockeel, N., Gafford, J., Papari, B., Mazzola, M.: Virtual inertia emulator-based model predictive control for grid frequency regulation considering high penetration of inverter-based energy storage system. IEEE Trans. Sustain. Energy 11(4), 2932– 2939 (2020) 14. Aluko, A.O., Dorrell, D.G., Pillay Carpanen, R., Ojo, E.E.: Heuristic optimization of virtual inertia control in grid-connected wind energy conversion systems for frequency support in a restructured environmentd. Energies 13(3), 564 (2020) 15. Othman, A.M., El-Fergany, A.A.: Adaptive virtual-inertia control and chicken swarm optimizer for frequency stability in power-grids penetrated by renewable energy sources. Neura. Comput. Appl. 33(7), 2905–2918 (2020). https://doi.org/ 10.1007/s00521-020-05054-8

364

A. Aluko et al.

16. Kerdphol, T., Rahman, F.S., Watanabe, M., Mitani, Y.: Optimization of virtual inertia control considering system frequency protection scheme. In: Virtual Inertia Synthesis and Control. PS, pp. 227–247. Springer, Cham (2021). https://doi.org/ 10.1007/978-3-030-57961-6 9 17. Saadatmand, S., Shamsi, P., Ferdowsi, M.: Adaptive critic design-based reinforcement learning approach in controlling virtual inertia-based grid-connected inverters. Int. J. Electr. Power Energy Syst. 127, 106657 (2021) 18. Skiparev, V., Belikov, J., Petlenkov, E.: Reinforcement learning based approach for virtual inertia control in microgrids with renewable energy sources. In: 2020 IEEE PES Innovative Smart Grid Technologies Europe (ISGT-Europe), pp. 1020–1024 (2020) 19. Yao, F., Zhao, J., Li, X., Mao, L., Qu, K.: RBF neural network based virtual synchronous generator control with improved frequency stability. IEEE Trans. Ind. Inform. 17(6), 4014–4024 (2020) 20. Lin, F.J., Tan, K.H., Shih, C.M.: Distributed generator with virtual inertia using intelligent controller for grid-connected microgrid. In: 2020 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pp. 1–8 (2020) 21. Datta, M., Senjyu, T.: Fuzzy control of distributed PV inverters/energy storage systems/electric vehicles for frequency regulation in a large power system. IEEE Trans. Smart Grid 4(1), 479–488 (2013) 22. Aluko, A.O., Dorrell, D.G., Pillay Carpanen, R., Ojo, E.E.: Frequency control of modern multi-area power systems using fuzzy logic controller. In: IEEE PES/IAS Power Africa, pp. 645–649 (2019) 23. Zhang, S., Mishra, Y., Shahidehpour, M.: Fuzzy-logic based frequency controller for wind farms augmented with energy storage systems. IEEE Trans. Power Syst. 31(2), 1595–1603 (2016) 24. Khooban, M., Gheisarnejad, M., Vafamand, N., Boudjadar, J.: Electric vehicle power propulsion system control based on time-varying fractional calculus: implementation and experimental results. IEEE Trans. Intell. Veh. 4(2), 255–264 (2019) 25. Yesil, E.: Interval type-2 fuzzy PID load frequency controller using Big Bang-Big Crunch optimization. Appl. Soft Comput. 15, 100–112 (2014) 26. Gheisarnejad, M., Khooban, M.-H., Dragicevic, T.: The future 5G network based secondary load frequency control in maritime microgrids. IEEE J. Emerg. Sel. Top. Power Electron. 8(1), 836–844 (2019) 27. Aluko, A.O., Dorrell, D.G., Carpanen, R.P., Ojo, E.E.: Heuristic secondary frequency control of AC/DC interconnected power system in a deregulated environment. In: International SAUPEC/RobMech/PRASA Conference, pp. 1–6. IEEE (2020) 28. Mendel, J.M., John, R.I.B.: Type-2 fuzzy sets made simple. IEEE Trans. Fuzzy Syst. 10(2), 117–127 (2002) 29. Mendel, J.M., John, R.I., Liu, F.: Interval type-2 fuzzy logic systems made simple. IEEE Trans. Fuzzy Syst. 14(6), 808–821 (2006) 30. Liang, Q., Mendel, J.M.: Interval type-2 fuzzy logic systems: theory and design. IEEE Trans. Fuzzy Syst. 8(5), 535–550 (2000) 31. Kundur, P.: Power System Control and Stability. McGraw, New York (1994) 32. Rakhshani, E., Remon, D., Cantarellas, A.M., Rodriguez, P.: Analysis of derivative control based virtual inertia in multi-area high-voltage direct current interconnected power systems. IET Gener. Transm. Distrib. 10(6), 1458–1469 (2016)

Multimodal Visual Question Answering Using VizWiz Data; A Visual Assistant for the Blind B. Sreedha and Prashant R. Nair(B) Department of Computer Science and Engineering, Amrita School of Engineering, Amrita Vishwa Vidyapeetham, Coimbatore, India [email protected], [email protected]

Abstract. Visual Question Answering (VQA) is a Multimodal Interaction task between two domains, Computer vision and Natural Language Processing. The task is to develop a system that could answer a question related to an image. The task is easy for a human being, but it is difficult for a system as it involves multiple reasoning over the Image and question to get the most accurate answer. the basic step involved is to extract the image features using a Convolutional Neural Network and the question features extracted via a Recurrent Neural Network. features so extracted undergo multiple reasoning to get the most accurate answer. VQA can act as a visual assistant for the blind to know about the beautiful world around them. This work is basically to develop a VQA assistant for the blind using a real-world dataset known as VizWiz dataset originating from the blind. Keywords: Visual Question Answering · VizWiz

1 Introduction Artificial Intelligence (AI) finds its application in every domain across the globe. It is improving the quality of life of people. When it comes to health care, it provides useful insights and innovative solutions to the problems through medical data analysis. VQA is an AI approach to help blind to know what is happening around them. It can serve as the best visual assistant to the blind. VQA is a combination of computer vision, natural language processing, and knowledge representation and reasoning implemented via deep learning [10, 19]. This can act as an eye for the blind to see the beautiful world around them without the need for human assistance. VQA models should be carefully designed with maximum accuracy as any improper prediction may result in misunderstandings especially while dealing with the blind. real world solutions need a model to be trained on a real-world dataset. VizWiz is the first real-world dataset originating from blind people where photos are taken by blind photographers and questions were asked based on the photo. In this paper, an innovative deep learning [11] solution is proposed on the VizWiz dataset that will serve as a visual assistant for improving the quality of life of the blind.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 365–372, 2022. https://doi.org/10.1007/978-981-19-1742-4_30

366

B. Sreedha and P. R. Nair

2 Motivation Blind people always need external help to know about the things happening around them. however, relying on humans will not be suitable in certain situations. The people may not be available at the time of need, the instantaneous reply cannot be obtained. Moreover, they have to compromise their privacy to get things done. Building an Artificial Intelligent System could greatly help these groups of people, enabling them to get instantaneous answers without compromising their privacy. they no longer need to rely on other people to get their things done. According to a World Health Organization report, around 40 million people across the globe are blind and 250 million people have visual Impairment. One of the WHO’s key areas of work and activities in helping the blind is to support the development and implementation of tools that can assist blind people. VQA could serve the purpose of assisting the blind, hence developing a model, that can accurately answer the question can improve the life of billions of people across the globe.

3 Related Works Malinowski and Fritz [8] proposed an image question answering model in which the image is analyzed via a Convolutional Neural Network (CNN) [13] and the question together with the visual representation is fed into a Long Short Term Memory (LSTM) [4] network to predict the answer. One of the earlier attempts in VQA was done by Zhou et al. [14]. They used naive bag-of-words as the text feature and used the deep features from GoogleNet as the visual features. The combined feature is sent to the softmax layer to predict the answer class. Ren et al. [15] introduced a VIS+LSTM model, where the task is considered as a classification task [12]. Antol et al. Introduced the VQA1.0 dataset where questions related to images are asked and is answered by 10 crowdsourced workers. VQA challenge is held every year based on this dataset. At some of the early attempts, textual and images are concatenated together with point-wise multiplication and passed onto a fully-connected classifier. attention mechanism in the model was first introduced by Lu et al. they used a co-attention [17] mechanism that gives equal importance to image and textual features and jointly reasoned the two modalities. Yang et al. introduced the Stacked attention Mechanism where multiple reasoning is done over the visual features based on the question to get the right answer. The first real-world VQA dataset known as the VizWiz dataset for helping blind people were introduced by Gurari et al., where the dataset is created using the VizWiz application available in android and apple phones [21]. Vahid Kazemi and Ali Elqruish proposed a model [2] in which the image features were extracted with pretrained Resnet 152 and the question are embedded with LSTM and it is passed through a stacked attention mechanism to get the output in the VQA1.0 dataset. the reason for the choice of Resnet152 model is that it is pretrained in VQA1.0 dataset. Anderson et al. [18] proposed a model that uses a pre-trained object detection model (Faster R-CNN) to generate arbitrary regions, where attention is to be estimated. In the

Multimodal Visual Question Answering Using VizWiz Data

367

bottom-up phase approach, attended images features obtained via Region of Interest (ROI) pooling are combined with the question features obtained from Gated Recurrent Unit (GRU). In VQA the question may be related to only a certain region of the image, so object detection helps detect the object. but if the images lack quality, it will not be able to detect the Region of Interest (RoI) as the object detection model [7, 20] will be able to recognize the object if the probability of getting detected is greater than the minimum threshold. moreover, it is not able to finetune if the image does not contain tags.

4 Dataset VQA system could serve as a visual assistant [10] for the blind only if it is trained in the images and questions that originate from the blind. VizWiz is the first real-world VQA dataset introduced by Gurari et al. [1] for helping the blind. This dataset originates from the blind photographers using the VizWiz application that is available on iPhone and Android Phones. The user can take photos and can record the question related to the image. the authors have filtered out those images that contain private information or personally identifying information. VizWiz Grand Challenge, A VQA Challenge is held every year based on this dataset. for the challenge, the question of the training images is answered by 10 Amazon Mechanical Turk Workers (AMT) workers. for each question, the most frequent answer from the 10 Answers is considered as the ground truth. VizWiz is different from other artificially created VQA datasets as it originates from a real-world setting. 4.1 Challenges in the Dataset High Uncertainty of Answers: Different from other Artificial VQA Dataset, there is a high disagreement between the annotators, as the images are often blurry or the question may be not be related to the question. since these images are captured by blind photographers, they cannot confirm the quality of the image. Often, they may ask a question, that may not be relevant to the Image. Conversational Question: In other VQA datasets, the question maybe created with the help of annotation workers. VizWiz is a real dataset from the blind. The questions about the image are recorded with the application. The user may use the salutation words such as hello, thank you, etc. so it is necessary to remove salutations and punctuation from questions. Relatively Small Size: VizWiz data originate from real-world settings, hence it is difficult to collect data. some data has to be filtered as some images may contain any privacy issue or reveal the identity of the user. the size of VizWiz data is relatively smaller compared to other VQA datasets created from artificial settings. The Imbalance Between Answerable and Unanswerable Classes: In the VizWiz dataset, some images may lack quality or the questions may not be relevant to the image. So, the number of unanswerable questions is quite high in the VizWiz dataset.

368

B. Sreedha and P. R. Nair

there might also occur high disagreement between the annotators due to this lack of quality issues.

5 Methodology The goal of the work is to extract the useful information from input images and questions. The extracted features are then passed onto a stacked attention network where the multiple reasoning over the image features based on the question. The output attended image features are combined with the question features and passed onto the classifier to get the answer as shown in Fig. 1.

Fig. 1. VQA model

Image Preprocessing: In the preprocessing stage, the images are resized to 240 * 240 dimension, and they are center cropped. input channels of the image are normalized by subtracting the mean and divided by the standard deviation. Image Feature Extraction: Image features are extracted using a pretrained EfficientNet-B1 [5] model. EfficientNet-B1 model is a convolutional network architecture that uniformly scales in all dimension such as depth, width and compound solution using a compound coefficient. The preprocessed images are passed to pretrained EfficientNet b1 model. The output of the model is a feature map of 1280 * 8 * 8 where 1028 represent the dimension and 64(8 * 8) represent the number of Image region that corresponds to the 30 * 30-pixel region of the input image. Question Preprocessing: Questions are made to lower case and punctuation is removed. Questions are tokenized into sequence of words. Word indices are formed from words. If some words are not present in the dictionary, then it is replaced with UNK token. Question Feature Extraction: Question features is extracted with the text encoder module. Module is composed of 2 units. embedding layer and RNN module.

Multimodal Visual Question Answering Using VizWiz Data

369

Embedding Layer: Texts are passed onto the embedding layer. A word embedding is a vector representation of a word that is semantically significant. Instead of representing words in a one-dimensional space per word (one-hot encoding), a dense representation that preserves semantic linkages is used. Word2Vec is a word embedding model introduced by Begino et al. with the goal of learning word representations that can be used to predict the surrounding words in a document. word2vec embedding is used here. input to the Embedding layer module is a list of indices, and the output is the corresponding word embeddings. Recurrent Neural Network (RNN): Word embedding is passed to an RNN layer to get the semantic representation of the question. RNN will capture the fixed embeddings from text sequences of variable length. Long Short-Term Memory (LSTM) is used as the text encoder. LSTM can get the semantic representation of words in the text. the last hidden state output of LSTM is taken as the question feature vector. Stacked Attention Network: In most cases, the question may be related to a smaller or specific part of the region. the attention or focus is to be given to the specific part of the image. attention mechanism can help to focus on a specific part of the image. In this work, a stacked attention mechanism model [3] is employed. In Stacked Attention Mechanism (SAN), the semantic representation of a question, to search the image regions that are related to answers. here the image is queried multiple times based on the question, to infer an answer.    a = ReLU WI xI ⊕ WQ xQ + bQ (1) pi = softmax(WH a + bA ) The image feature matrix and question vector are passed onto a single layer of neural network and then to a softmax activation to compute the attention distribution over the images. where WI , WQ and WH are the learning parameters and pi corresponds to the attention distribution over 64(8 * 8) regions. xI ∈ Rdxn corresponds to the image feature vector where d represent the image dimension and n represent the number of image region.⊕ represent the element wise addition of image and question vector. The weighted sum of image vector is obtained based on the attention distribution from the different region of the image. It is then combined with the question vector to form the final query vector. The final vector f contains both visual and question informations. It is then passed onto the classifier. m pi xi x˜ I = i (2) f = x˜ I + xQ Classifier: In this work, a two-layer classifier is employed where the first layer is a fully-connected layer with 1024 dimensions with relu [6] nonlinearity and in the second fully connected layer with a size of N, where N corresponds to the most frequent N answer classes.

370

B. Sreedha and P. R. Nair

6 Evaluation Metrics Each image is associated with a question. the question is answered by ten Amazon Mechanical Turk (AMT) workers. Therefore, corresponding to each image, question pair there are 10 answers. The most frequent answer from the 10 answers is taken as the ground truth answer for each image, question pair for training the model. the best model is saved during training. the model obtained is used to predict the answer for 8000 test images, question pairs. the JSON file containing the predicted answer is submitted to the EvalAI [9] server to get the prediction accuracy of the model in test data.   10 i predict = answer i  ,1 (3) acc(predict) = min 3 the evaluation metrics is introduced by Antol et al. [16] for the VQA challenge. the predicted answer is considered correct if at least 3 out of the 10 annotators have given the same predicted answer.

7 Results We have evaluated the result in test standard set of 2020 VizWiz grand challenge. Results and comparison of the models are shown in Table 1. Table 1. A comparison of models Model

Overall Other

Show ask attend and answer model 48.43% 35.3% Our model

Unanswerable Yes or No Number 81.2%

49.69% 36.94% 81.57%

59.6%

18.7%

59.5%

23.04%

EfficientNet-B1 is 7.6× smaller and. 5.7× faster than ResNet-152 showed an improvement in the overall accuracy in VizWiz dataset by 1.26%. EfficientNet b1 has 7.8 million parameters while the resent152 model has 60 million Parameters. It is difficult to deploy huge models in edge devices. Smaller models are always preferred over the big models in the deployment stage.

8 Future Scope and Conclusion The main focus of the research was to improve the result of the existing model by bringing about changes in the existing model. Here, we concentrated on improving the image feature extraction part through a robust, fast and small model. This could bring about an improvement in the overall accuracy of the model by 1.26%. We hope the model performance can be further improved by bringing about changes in the text extraction part.

Multimodal Visual Question Answering Using VizWiz Data

371

References 1. Li, G.D., et al.: VizWiz grand challenge: answering visual questions from blind people. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 3608–3617 (2018). https://doi.org/10.1109/cvpr.2018.00380 2. Kazemi, V., Elqursh, A.: Show, ask, attend, and answer: a strong baseline for visual question answering. arXiv preprint: arXiv:1704.03162 (2017) 3. Yang, Z., He, X., Gao, J., Li, D., Smola, A.: Stacked attention networks for image question answering. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 21–29 (2016) 4. Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997) 5. Tan, M., Le, Q.: Efficientnet: rethinking model scaling for convolutional neural networks. In: International Conference on Machine Learning, pp. 6105–6114. PMLR (2019) 6. Agarap, A.F. Deep learning using Rectified Linear Units (RELU). arXivpreprint: arXiv:1803. 08375 (2018) 7. Le, T., Huy, N.T., Le Minh, N.: Integrating transformer into global and residual image feature extractor in visual question answering for blind people. In: 2020 12th International Conference on Knowledge and Systems Engineering (KSE), pp. 31–36. IEEE (2020) 8. Malinowski, M., Rohrbach, M., Fritz, M.: Ask your neurons: a neural-based approach to answering questions about images. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 1–9 (2015) 9. Sarath, S., Amudha, J.: Visual question answering models evaluation. In: 2020 International Conference for Emerging Technology (INCET), pp. 1–5. IEEE (2020) 10. Kandoth, A., Arya, N.R., Mohan, P.R., Priya, T.V., Geetha, M.: Dhrishti: a visual aiding system for outdoor environment. In: 2020 5th International Conference on Communication and Electronics Systems (ICCES), pp. 305–310 (2020). https://doi.org/10.1109/ICCES48766. 2020.9137967 11. Babu, R., Naren, V.S., Soman, K.P: Indian car number plate recognition using deep learning. In: 2019 2nd International Conference on Intelligent Computing, Instrumentation and Control Technologies (ICICICT), vol. 1, pp.1269–1272. IEEE (2019) 12. Saiharsha, B., Diwakar, B., Karthika, R., Ganesan, M.: Evaluating performance of deep learning architectures for image classification. In: 2020 5th International Conference on Communication and Electronics Systems (ICCES), pp. 917–922. IEEE (2020) 13. Aloysius, N., Geetha, M.: A review on deep convolutional neural networks. In: 2017 International Conference on Communication and Signal Processing (ICCSP), pp. 0588–0592. IEEE (2017) 14. Zhou, B., Tian, Y., Sukhbaatar, S., Szlam, A., Fergus, R.: Simple baseline for visual question answering arXivpreprint: arXiv:1512.02167 (2015) 15. Ren, M., Kiros, R., Zemel, R.: Exploring models and data for image question answering. Adv. Neural. Inf. Process. Syst. 28, 2953–2961 (2015) 16. Malsa, N., et al.: CERTbchain: a step by step approach towards building a blockchain based distributed application for certificate verification system. In: 2021 IEEE 6th International Conference on Computing, Communication and Automation (ICCCA), pp. 800–806 (2021). https://doi.org/10.1109/ICCCA52192.2021.9666311 17. Lu, J., Yang, J., Batra, D., Parikh, D.: Hierarchical question-image co-attention for visual question answering. Adv. Neural Inf. Process. Syst. 29, 289–297 (2016) 18. Anderson, P., et al.: Bottom-up and top-down attention for image captioning and visual question answering. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 6077–6086 (2018)

372

B. Sreedha and P. R. Nair

19. Nirmala, R., Thangavel, S.K.: Develop, implement and evaluate a multimodal system for government organization. In: 2017 4th International Conference on Advanced Computing and Communication Systems (ICACCS), pp.1–8. IEEE (2017) 20. Dushi, D.: Using Deep Learning to Answer Visual Questions from Blind People (2019) 21. Vishnu, R., Krishna Prakash, N.: Mobile application-based virtual assistant using deep learning. In: Sivakumar Reddy, V., Kamakshi Prasad, V. (eds.) Soft Computing and Signal Processing. AISC, vol. 1340, pp. 609–617. Springer, Singapore (2022). https://doi.org/10.1007/ 978-981-16-1249-7_57

Design and Neural Control for Insect-Copter for Smooth Perching on Outdoor Vertical Surface Sandeep Gupta(B)

and Laxmidhar Behera

Electrical Engineering Department, Indian Institute of Technology, Kanpur, Kanpur, India {sngupta,lbehera}@iitk.ac.in

Abstract. Perching on vertical surfaces include impact of aerial vehicle with target surface. The impact may cause damageto the installed payloads. Smooth landing on vertical surface is recommended for such aerial vehicles. This chapter proposes a novel insect copter design to perform smooth landing on vertical surface. The insect copter design is explained with its functionalities. Kinematics and dynamic equation governing the flight and perching is discussed. The trajectory for soft landing on vertical surface is formulated using neural network approximation. An adaptive neural controller is designed to follow the desired trajectory. Further to validate its behaviour simulations are performed in multi-body dynamic software MSCAdams and MATLAB. Simulation Results have shown effective controlled perching on vertical surfaces. Keywords: Wall perching quad-rotor · Hovering · Soft landing and perching · Neural control approach

1 Introduction Nano UAVs (unmanned arial vehicles) are gaining more interest in surveillance operations due to their very small size and hover capability [1] and [2]. They can carry a tiny camera and send pictures or video of target area. However, their flight time is quit less as 5–10 min [3]. To perform the long time survillence, perching of quadrotor on rough surface is effective solution by reducing power requirements as in case of continuous flying [4]. Researchers have implemented different perching mechanism with and without flying capabilities. In [5] a wall climbing robot is designed using suction cups.The designed robot in [6] and [7] can climb on wall using electrostatic charge. But these methods are not suitable for implementation of wall perching and climbing drone. In [12] and [13], the sliding mode control algorithms for perching of nano quadrotor are proposed.Most of previous work related to vertical wall perching is based on continuous thrust in opposite direction to wall.This method requires huge amount of energy to keep robot on to the wall. Others have implemented perching on smooth surface using adhesive pads [8]. These gecko adhesive pads can work only on flat and smooth surface like glass. In [9], a drone was developed with the capability of climbing on a wall and can take-off from wall after detaching itself. One quad copterwith soft landing capabilities © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 373–382, 2022. https://doi.org/10.1007/978-981-19-1742-4_31

374

S. Gupta and L. Behera

has developed in [10] but due to its safety guards which are in wheel shape, effect the flight dynamics of quad copter. In Here we propose a claw based mechanism which causes least aerodynamic obstruction to drone movement. These claws act as multipurpose agent. Due its flying capabilities like a quad rotor helicopter and wall perching using claw like a insect, it would be called insect-copter. Initially, the claws are used to maintain a safety region for propellers from vertical surface and after successful landing on wall, it may penetrate into wall. In our design, front and rear claws can keep flying vehicle steady on the wall even in air flow. This design eliminates the need of circular wheel type safety guard [10] and long wire arm [4] to avoid aerodynamic obstructions. The structure of the chapter is organised as follows: Section (II) describes the mechanical design of the proposed nano air vehicle, Section (III) discusses mathematical modeling of proposed insectcopter and control design, Section (IV) describes the detail strategy of perching mechanism, Section (V) shows the implementation of solid model of insect-copter in ADAM simulation software, section (VI) provides simulation results and discussion and Section (VII) concludes the current work and its future scope.

2 Proposed Mechanical Design The design is based on Hybrid H quad frame as this structure suitable for adding actuation in front and rear arm for tilt mechanism.The rotational movement of these two arms is controlled by micro servo motor coupled with spur gear set. This model is designed in 3D CAD design software (Solid Works) and analysed using multibody dynamics simulation software (Adams). The solid model of the proposed insectcopter is shown in Fig. 1.

Fig. 1. Solid model representation of the proposed insect-copter and Front claw and Gear mechanism

The details of tilting arm mechanism is also shown in Fig. 1. Two servo motors can control front and rear arm separately so it is also be named as differential drive tilting arm quadcopter. These two servo motors eliminate the need of encoder as in [10]. The spur gear sets are used to couple arms with servomotors as they have high power transmission efficiency and no slip like belt drives.The gear ratio of 1:2 is used to deliver sufficient torque for arms.

Design and Neural Control for Insect-Copter

375

Two additional micro servo motors are directly coupled with spiny claws named as front and rear claw 1. These two claws are in downward direction to act as landing gears and in parallel to the quad frame acting as safety guard for wall perching. These claws can stuck in tiny holes in the walls and support the vehicle to perch on the wall. The front and back arm also have extra pair of claw for stable perching in case of large wind flow while perching on the rough vertical surface in outdoor environment.

3 Mathematical Model and Control Design 3.1 Fixed Rotor Quad-Copter Model The Quadcopter mathematocal model is taken from [11].   Jy − Jz k d ˙ ˙ ˙ ∅ + u2 ¨ = θΨ +∅ ∅ Jx Jx Jx   kθ J − Jx d ˙ Ψ˙ z + θ˙ + u3 θ¨ = ∅ Jy Jy Jy   J − Jy 1 k ˙  + u4 ˙ θ˙ x Ψ¨ = ∅ + Jz Jy Jz z¨ = −g −

u1 kz z˙ + (cos∅ cosθ ) m m

x¨ = −

u1 kx x˙ + (sinθ cosΨ cos∅ + sinΨ sin∅) m m

y¨ = −

ky y˙ u1 + (sinθ cos∅ sinΨ − cosΨ sin∅) m m

(1)

where, roll, pitch and yaw are represented by (∅, θ, Ψ ) respectively and (x, y, z) are position co-ordinates. m is the mass, Jx , Jy , Jy are the inertial values in respective coordinates. The aerodynamics damping/drag coefficients are given as kx , ky , kz , k∅ , kθ , kΨ and d is moment arm. u1 is force andu2 ,u2 , u3 torques. These four control inputsare given as u1 = F1 + F2 + F3 + F4 , u2 = −F 1 − F2 + F3 + F4 , u3 = −F1 + F2 + F3 − F4 , and u4 = −F 1 + F2 − F3 + F4 where F1 ,F2 ,F3 , and F4 are applied forces on the rotors. 3.2 Differential Drive Titlting Arm Quad-Copter Model In case of the tilt rotor, if the tilt angle to the wall is α then the pitch angle to the wall ∅w becomes ∅w = 90− α. Then the torque required at the contact point cto the wall will be:  mg  sin(∅w ) − Ft sin(∅w ) Tc = l (2) 2 When the quad-copter is stabilized at hinge point c, the front servo moter provides the 45° angle to front rotor. The trajectory of soft landing is learnt using neural network.

376

S. Gupta and L. Behera

This trajectory is similar to the pendulum motion which is to be stabilized at 90°. Hence, the quad-copter frame is equivalent to a simple pendulum and the dynamics is given as follows. ml 2 sinθ¨ + bθ˙ + mglsinθ = Q

(3)

where Q is control torque input and b is a damping torque coefficient (from friction) (Fig. 2).

Fig. 2. Simple pendulum motion

For the quadcopter perching and landing operation, the equation of motion is given by replacing θ with tilt angle α as follow ml 2 sinα¨ + bα˙ + mglsinα = u

(4)

Let as assume sinα¨ ≈ α, ¨ then the equation will become 1 g α¨ = − sinα − bα˙ + 2 u l ml

(5)

α¨ = f (α) + g(α)

(6)

3.3 Problem Statement The control inputs u and an adaptive control law have to be designed for relative degree two type of subsystems assuming that f (α) is unknown funtion. Here, neural controller is designed for following the desired trajectory and the unknown dynamics of the quadrotor model is estimated using adaptive neural network (ChNN). The error dynamics for which the controller has been designed is discussed below. 3.4 Design of ChNN Based Control Algorithm The altitude dynamics are considered, where we take α = z1 and α˙ = z2 then altitude dynamics will become: z˙1 = z2

(7)

z˙2 = f (z) + g(z)u

(8)

Design and Neural Control for Insect-Copter

377

y = z1

(9)

The control objective is to design a control input u1 so that quadrotor follows the desired trajectory zd (t), while the states stay bounded. Using the feedback linearization technique, the control input law is provided as   (10) u = (g(z))−1 −f (z) + Pdd Ez + ρ e˙ z + z˙ 2d where e is the tracking error,z→ (z1 , z2 ), ez = z1 −z1d , a new variable Ez = ρez + e˙z and design parameter terms Pdd and ρ are selected as positive integers.To approximate the unknown function(z), the Chebyshev neural network is adopted. The numerical transformation as a functional expansion (FE) is to expand input pattern. The two initial Chebyshev polynomials are chosen as P0 (e) = 1, P1 (e) = e and [P2 (e),P3 (e),……,PN (e)] will be generated by a recursive function [20]-[21] as PN +1 (e) = 2ePN (e) − PN −1 (e), where PN (e) represent nth order polynomial. The argument is taken as −1 < e < 1 and input pattern is n dimensional such ase = (e1 , e2 , . . . ., en )T ∈ Rn . Then the enhanced pattern will become as β(e) = [P0 (e1 ), P1 (e1 ), . . . . . . , PN (e1 ), . . . ., P0 (en ), P1 (en ), . . . . . . , Pm (en )]T (11) Hence, ChNN neural network output will be expressed as Och = Wmn T β(e), where Chebyshev polynomial basis function is β(e), order of Chebyshev polynomial is m, number of inputs is n in input layer and Wmn is weight vector of network. The Fig. 3 represents the ChNN structure model.

Fig. 3. Model of single layer two-input chebyshev neural networks

Based on ChNN model, unknown function f (z) can be approximated asf (z) = Wmn T β(ein ) + ε, where ein = [e, e˙ ]T is input vector, Wmn = [W11 , W21 , W31 , W41 , W12 , W22 , W32 , W42 ] is weight vector and ε is approximation 



T



error. Now the estimation of f (z) is provided as f (z) = W mn β(ein ) where W mn is estimation ofWmn .Now putting f (z) in place of f (z) in Eq. (10), the control law will become 



u = (g(z))−1 [−f (z) + Pdd Ez + ρ e˙ z + z˙2d ]

(12)

378

S. Gupta and L. Behera

Let us assume ε = 0 and putting control law u1 in system Eq. (8) we get, z˙2 = f (z) + g(z).(g(z))(−1) [−fˆ (z) + Pdd Ez + ρ e˙ z + z˙2d ]

(13)

T

T = Wmn β − W mn β + Pdd Ez + ρ e˙ z + z˙2 d 

T

˜ T = Wmn T − W mn , the equation will become Defining W ˜ T β − Pdd Ez − ρ e˙z z˙2d − z˙2 = −W

(14)

˜ T β − Pdd Ez − ρ e˙z e¨z = −W

(15)

We have Ez = ρez + e˙z and after differentiating this equation, we get E˙z = ρe˙ z + e¨z . Substituting e¨z = E˙z − ρe˙ z in Eq. (15), ˜ T β − Pdd Ez − ρ e˙z E˙z − ρe˙ z = −W

(16)

˜ T β − Pdd Ez E˙z = −W

(17)

This is the linear and stable closed loop error dynamics after putting controllaw u1 in Eq. (8). To obtain the weight update law, lyapunov theory is used. Let us chose a lyapunov function V z in which D is a positive definite matrix. 1 2 1 ˜ T −1 ˜ Ez + W D W 2 2 ˙˜ ˜ T D−1 W V˙z = Ez E˙z + W

Vz =

(18) (19)

From Eq. (17), putting E˙z into (19),

˙˜ ˜ T β − Pdd Ez + W ˜ T D−1 W V˙z = Ez −W

(20)

˙ ˜ T (βEz + D−1 W V˙z = −P dd E 2 z − W mn )

(21)



If we chose the second term as 0 in Eq. (21), V˙z = −P dd E 2 z which satisfy the condition of stability in sense of Lyapunov (Vz > 0 andV˙z ≤ 0). The update law from Eq. (21) is given below as ˙ (βEz + D−1 W mn ) = 0

(22)

˙ W mn = −DβEz

(23)





Finally the control low for the altitude control will become as,   T ˙ u = ml 2 −W mn β(ein ) + Pdd Ez + ρ e˙ z + z¨d , W mn = −DβEz 



(24)

where z¨d = z˙2d . Hence the proposed control input laws will stabilize the quadrotor dynamics as per the the Lyapunov theory in which weights are modified using derived update laws.

Design and Neural Control for Insect-Copter

379

4 Thrust Assisted Perching Strategy Many researchers are working on the thrust assisted perching strategy like S-Mad air vehicle [9]. There are multiple advantages of low impact landing on vertical surface. In soft landing or low impact landing, the vehicle is quit safe when collide with wall unlike high impact landing. The maximumrotational limit of arms are 180° as we have used servo motors. This angular movement is quite sufficient for providing thrust towards wall to perch the wall. Insect-copter takes off from ground and can fly as conventional quad-copters. When it is needed, quad-copter moves in direction of target location for perching. The on board distance sensor measures the distance from wall and move slowly towards it. The flight controller makes it stable after hitting on the vertical surface. The Fig. 4 shows the above stages during soft landing of the insect-copter. It is assumed that wall is rough and vertical, and the same thrust is available as in case of hovering. The friction coefficient of vertical wall should be very low. The neural controller maintains the required thrust towards wall while both the servo motors are tilting the front and rear arm simultaneously. Figure 5 shows the tilting of quad body at 60° and the final perching on the wall. The front and rear claws grasp the wall to provide stable perching for a longer time. To validate the motion stability and perching strategy simulations are performed using Matlab and Multi-body dynamics software MSC Adams.

Fig. 4. Insect-copter in hover mode, approaching towards wall, front rotor tilting and both the rotor tilting.

The assembly of the insect is imported in ADMAS environment and joint constrained are defined in the assembly. After this, the model is improted to MATLAB to design and test the controller. In Admas, the visuals are captured while pitch, x-y position profile data is loggged using Matlab. The quad-rotor has been given four force components on four quad-rotor. These forces are governed with flight controller equations. Figure 1 shows the insect copter

380

S. Gupta and L. Behera

Fig. 5. Quad-copter maintaining pose for soft landing and Soft landing on vertical wall

model inside the software environment. Each component of the insect copter is assigned with its material properties. The ground and wall is also imported surfaces. Wall and the insect-copter is defined as contact surfaces with impact with damping and having friction. Gravity is taken vertically downward 9.8 m/s2 .The perching include few steps like touching the wall surface and using tilt rotors to stabilize and perch with smooth motion. The Snapshots of the flying, hovering and perching sequences of the insectcopter is shown in Fig. 6.

Fig. 6. Tapped images from Adam to show different stages in perching task

5 Simulation Results This work’s primary contribution is to minimize the impact of frame on the vertical wall during perching mode by regulating the velocity of pose change with an airframe based on tilt-rotor. Since the fixed frame strategy uses hard landing scheme with large

Design and Neural Control for Insect-Copter

381

impact to avoid slippage from dropping, the effect may be large enough to harm the quadrotor. Using ADAM-MATLAB co-simulation method, the simulations are carried outto validate the performance of the proposed mechanism and motion control scheme. Figure 7 shows the evolution of pitch angle with respect to time. Pitch angle gradually increases and reaches its maximum value when it started climbing the wall. Figure 8 shows the evolution of x-y co-ordinatse with respect to time.

Fig. 7. Pitch angle profile in soft landing

Fig. 8. X-Y position profile in soft landing

6 Conclusion In this research, smooth perching on vertical wall is demonstrated using tilt rotor based mechanism which reduce the impact of landing force.In addition, the tilt system in lieu of the force-assisted wall-climbing system can support a small quadrotor to adhere to the wall with less thrust by adapting the propeller direction. We used a quadrotor based framework to introduce the notion of the lateral soft landing system with front and back claw for longer mission life as required in surveillance. Even though the proposed system consists one degree of freedom rotation of claw arm, it optimizes the drone frame for lateral climbing only. There could be a system to alter the driving path of the drone frame for horizontal movement and climbing on vertical wall. Hence, designing a tiltrotor wall-perching and climbing base that enables omni directional motion is the future scope of the chapter.

382

S. Gupta and L. Behera

References 1. Preiss, J.A., Honig, W., Sukhatme, G.S., Ayanian, N.: Crazyswarm: a large nano-quadcopter swarm. In: IEEE International Conference on Robotics and Automation (ICRA), pp. 3299– 3304. IEEE (2017) 2. Mulgaonkar, Y., Cross, G., Kumar, V.: Design of small, safe and robust quadrotor swarms. In: IEEE International Conference on Robotics and Automation (ICRA), pp. 2208–2215. IEEE (2015) 3. Faessler, M., Fontana, F., Forster, C., Mueggler, E., Pizzoli, M., Scaramuzza, D.: Autonomous, vision-based flight and live dense 3d mapping with a quadrotor micro aerial vehicle. J. Field Robot. 33(4), 431–450 (2016) 4. Pope, M.T., et al.: A multimodal robot for perching and climbing on vertical outdoor surfaces. IEEE Trans. Robot. 33(1), 38–48 (2016) 5. Yoshida, Y., Ma, S.: Design of a wall-climbing robot with passive suction cups. In: IEEE International Conference on Robotics and Biomimetics, pp. 1513–1518. IEEE (2010) 6. Graule, M., et al.: Perching and takeoff of a robotic insect on overhangs using switchable electrostatic adhesion. Science 352(6288), 978–982 (2016) 7. Koh, K.H., Chetty, R.K., Ponnambalam, S.: Modeling and simulation of electrostatic adhesion for wall climbing robot. In: 2011 IEEE International Conference on Robotics and Biomimetics, pp. 2031–2036. IEEE (2011) 8. Zhou, M., Pesika, N., Zeng, H., Tian, Y., Israelachvili, J.: Recent advances in gecko adhesion and friction mechanisms and development of gecko-inspired dry adhesive surfaces. Friction 1(2), 114–129 (2013) 9. Mehanovic, D., Bass, J., Courteau, T., Rancourt, D., Lussier Desbiens, A.: Autonomous thrustassisted perching of a fixed-wing uav on vertical surfaces. In: Mangan, M., Cutkosky, M., Mura, A., Verschure, P.F.M.J., Prescott, T., Lepora, N. (eds.) Living Machines 2017. LNCS (LNAI), vol. 10384, pp. 302–314. Springer, Cham (2017). https://doi.org/10.1007/978-3-31963537-8_26 10. Myeong, W., Myung, H.: Development of a wall-climbing drone capable of vertical soft landing using a tilt-rotor mechanism. IEEE Access. 7, 4868–4879 (2018) 11. Xu, R., Ozguner, U.: Sliding mode control of a class of underactuated systems. Automatica 44(1), 233–241 (2008) 12. Singh, P., Gupta, S., Behera, L., Verma, N.K., Nahavandi, S.: Perching of nano-quadrotor using self-trigger finite-time second-order continuous control. IEEE Syst. J. 15, 1–11 (2020) 13. Gupta, S, Singh, P., Behera, L., Verma, N.K.: Perching of nano-quadrotor on vertical wall using periodic event-triggered control. In: 12th National Conference and Exhibition on Aerospace and Defence related Mechanism (ARMS 2021), 2–4 December 2021

A 12T SRAM of 16 nm CMOS Technology Using One Sided Schmitt Trigger Inverter and Read Port Komaladitya Challa(B) and Vinay Kumar Pamula Department of ECE, University College of Engineering Kakinada, JNTUK, Kakinada, Andhra Pradesh, India [email protected]

Abstract. This work describes a one-sided Schmitt-triggered with data independent read and write operation from a 9T Static Random-Access Memory (SRAM) cell that consume less power and has excellent read and write stability. The existing approaches are performed with data leakage problem, huge area, expensive energy per access read data bits. To solve this problem, the proposed work will introduce three duplications of Static RAM cells with read ports to arise ST 12T SRAM, with the goal of greatly reducing data based read port leakage to improve the read quality and minimize area and power. The proposed methodology of Schmitttrigger based 12T SRAM memory cell achieves excellent read robustness in a one-sided Schmitt-triggered inverter with a three different single bit arrangement, while the write ability improves by power gating in Schmitt-trigger-inverter with support of control and trip voltage. This article’s proposed approach 12T SRAM memory cell will be designed at the single bit level, utilizing 16 nm and 22 nm CMOS Technology, and demonstrate the area, latency and power consumption using Tanner EDA Tool. Keywords: Low power · Near threshold · Schmitt-trigger · Read port leakage

1 Introduction Memory is significant in data processing application devices for storing and retrieving data, especially when it comes to storing confidential and sensitive information with extremely protective encryption methods, such as AES, DES, SEA, etc. Static RAM is vital in System-on-Chip (SOC) and a notable commitment to the SOC’s maximum force usage and space. Because area is important concern when designing circuits, memory configuration researchers seek to place as many cells as possible per segment for the sharing of edge hardware [1]. Ordinary 6T SRAM, 8T SRAM and other cells are greatly constrained by their lack of ability to perform in longer segments. This happens to owe the negative impacts of information overflow, distorted ION /IOFF ratios, and read bit line swing when more cells are packed on one section. Various approaches have attempted to face this matter in question by working on the ION /IOFF proportion to empower up to thousand cells per segment [4]. Even though these procedures have been beneficial, © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 383–391, 2022. https://doi.org/10.1007/978-981-19-1742-4_32

384

K. Challa and V. K. Pamula

they continue to endure from the negative impact of vast region or moving information execution. Some also fail to stipulate the base energy point in Static RAMs, resulting in vast amount of energy in each access at extremely low voltages [2]. The greatest method to heavily reduce energy consumptions is reducing power by lowering the supply voltage (VDD ). Power decreases in quadratic function with the decrease in VDD [3]. In any event, when VDD decreases, deferral and soft error rates (SER) rises, while functional yields fall. In the sub-threshold voltage (sub-Vth ) region, where VDD is lower than Vth , the delay increments dramatically [2]. In the same way, the energy utilization is explained essentially in expanded static energy utilization, even though exceptionally low force is accomplished. When set side by side to ultra-scale operation, operating in near-Vth zone, where VDD is slightly greater than Vth , can accomplish a vast power reduction and a good improvement in latency over sub-Vth operation. Therefore, by optimizing both delay and power in near-Vth area, energy consumption may be reduced [3]. The Sensitive miscalculation caused by α-particles becomes a problem in near Vth action. This happens when, the SER is increased by the contracted basic charge in near Vth activity. A multi-bit error can appear in single work in the non-bit interleaving structure, where components of a word are constantly stored away. This happens when a sensitive error occurs leading to bit error in cells that are nearby to one another. Multi-bit error correction must be implemented which involves the error correction code (ECC) circuit to attain an extremely large area and energy usage. Since, bits of a word are spatially interleaved in a bit-interleaving structure, single bit mistake occurs in every word [4]. A basic ECC circuit can be casted to rectify these errors. So, to resolve the increasing SER in the near Vth area, the bit-interleaving structure was used [9].

2 Previous SRAM Cell Designs Many SRAM cells are introduced in near Vth operation. Cross coupled standard inverters formed traditional 6T [6] and 8T [7], WRE 9T, MH’s 9T, Chang’s 10T Static RAM cells, whereas cross coupled ST inverters are introduced in ST 10T, 11T SRAM cells [9]. In addition, depending on whether the storage node experiences read disturbance, these cells will be categorized as read disturbance and read disturbance free cells. Because read disturbance from BL or BLB can flip the stored data, the ordinary 6T and ST 10T Static RAM cells can be differentiated as a read disturbance cell. A read buffer is added to the standard 8T and WRE 9T, Chang’s 10T, MH’s 9T Static RAM cells to overcome the problem of read disturbance. These cells have the same read and hold stability since the read buffer decouples the storage node from the read BL. As a result, Static RAM cells provide enough read stability. By adopting power gating, the WRE 9T SRAM cell enhances the read buffer’s leakage current as well as its write capabilities [7]. All the cells use bit interleaving structure, which helps in reducing SER, except the 6T SRAM cell. Since 8T and WRE 9T SRAM cell experience read disturbance from differential BL. Write back scheme is applied along with bit interleaving structure causing high energy consumption and area. So, to apply bit interleaving without write back Chang’s 10T, ST 10T, MH’s 9T and ST 11T SRAM cells are proposed. But they all have area and energy overheads due to number of transistors, control signals and differential BL structure.

A 12T SRAM of 16 nm CMOS Technology

385

So, a new ST 9T SRAM is introduced to reduce area and energy consumption with sufficient read, write, and hold stability [9]. A cross coupled arrangement of a normal inverter with stacked transistor (PUL1, PUL2, PDL1, PDL2), a ST inverter (PUR, PDR1, PDR2, NF), and one nMOS PG makes up the proposed cell. Write word line A (WWLA) and write word line B (WWLB) are column-based signals, whereas word line (WL) is a row-based signal. The gates of PG and PUL2 are connected to WL and WWLA, respectively. PDL1 gate and NF source are both connected to WWLB. The presented Schmitt-trigger based 9T cell employs a bit interleaving structure without a write-back scheme and consumes less energy than current cells while maintaining appropriate read stability and write ability in the near-Vth region by combining the following features: (1) a reduction in energy consumption and area by using a single bit line (BL) structure, (2) an improvement in read stability and write ability by using a selective power gating technique with the ST inverter write assist technique [9].

3 Proposed SRAM Cell Design The proposed SRAM Cell describes three duplications of SRAM bit cells with nMOS only read ports. Figure 1 depicts a schematic of the proposed 12T SRAM cells. Each one has two access transistors and two cross coupled inverters. Each cell’s read port is made up of three nMOS transistors. Figure 1(a) shows a read port with reduced data based read bit line leakage and a high-performance goal. The read ports in Fig. 1(b) and 1(c) have no data based read bit line leakage and are designed for extremely low power and high density, respectively. The ION /IOFF is drastically deteriorated while working in the near and sub threshold area, making it increasingly difficult to implement a larger number of cells on a single column. The cumulative pass gate leakage grown equivalent to the read current as the number of cells increases, making it difficult for the sense amplifier to accurately evaluate the read bit voltage level. Furthermore, the read bit line leakage is affected by the data stored in the cell, causing the off state read bit line leakage current to fluctuate greatly. At ultra-low voltages, this is compounded by the worst-case data pattern, which might cause the RBL voltage level of ‘zero’ to exceed the RBL voltage level of ‘one’. Although, as shown in Fig. 1, the suggested 12T-S1 cell has a lower data reliance than the 12T-S2 cell, it is still essentially incapable of executing a read operation at ultra-low voltages. However, as seen in the proposed work at ultra-low voltages increases the energy per access, where operating at the near threshold point is the most energy efficient. As a result, the ST 12T cell is operated in the near-threshold area for the best energy efficiency and performance. The read bit line swing is concern for the 12T-S1 cell at near threshold and super threshold voltages. The following proposed work goes deep into the RBL swing of each cell in relation to data pattern and supply voltage.

386

K. Challa and V. K. Pamula

Fig. 1. Schematic of the proposed (a) 12T-S1 (b) 12T-S2 (c) 12T-S3 SRAM cells.

A 12T SRAM of 16 nm CMOS Technology

387

3.1 Main Contribution of This Work This paper describes three duplications of SRAM cells with nMOS only read ports that aim to greatly reduce read port leakage based on data, improve read performance, and reduce area and minimum power consumption over ST 9T cell to enable thousand cells per column. The original ST 9T cell has single bit line structure to be pre-charged for both read and write operations causing increased delay between operations of cell resulting in higher static power consumption. The Proposed SRAM has RBL (Read-Bit line) in read port which carries out read operation ensuring to separate the write operation carried out by BL (Bit line). The ST 12T improves the read stability over ST 9T by implementing the path for read operation independent of storage node. Since, RWL is source to fill the RBL and storage node is a conducting voltage rather than storage node being the source to fill BL in ST 9T greatly reduces the possibility of read disturbance. Partially pre-charging bit lines to reduce the strength of access transistor can also help in further reducing read disturbance. The ST 12T also enables thousand cells per column by adding read buffer RBL which is independent of storage node, read disturbance free and avoiding read bit line swing issues for better sense amplifier evaluation of read bit line voltages. The SRAM cell usually perform more read operation than the write operation because write operation is performed only when a change is needed in cache. Therefore, power consumption is based on frequency of operation and switching activity. The ST 9T has more switching activity in both read and write operation due to WWLA, WWLB, and WL constantly changing voltage levels from ‘0’ to ‘1’ and ‘1’ to ‘0’, whereas in ST 12T during read operation only RWL is switching from ‘0’ to ‘1’ making it to consume less power while performing read operation than ST 9T. The sense amplifier can be adjusted with reduced voltage differentiation at RBL for improved read performance as it is independent of BL for read operation. The technology used is a 16 nm CMOS technology over the 22 nm CMOS technology which greatly reduces the area and power of SRAM cell. We achieve increased read access performance, low energy per access and low area using a unique architecture in each of the three read port cells, therefore expanding the design and application gamut for memory designers in low powered devices.

4 Results and CMOS Implementation Utilizing side channel analysis and power control transistors where this work is implemented in Back-End ASCII technique of 16 nm and 22 nm CMOS technology of low power and highly secure data information will be executed with the goal of reducing power consumption and increasing security on memory control operation. The output of comparisons of one-sided Schmitt-Trigger based 9T SRAM cell for near threshold operation given in Table.1, and comparisons analysis chart given in Fig. 2. The output comparison of proposed work in data independent read port with one sided Schmitttrigger based 12T SRAM cell are given in Table 2, and comparisons analysis chart is given in Fig. 3. The Output waveforms are given in Fig. 4(a), (b), (c).

388

K. Challa and V. K. Pamula

Table 1. Comparison of one-sided Schmitt-trigger based 9T SRAM cell for near threshold operation One-sided Schmitt-trigger based 9T SRAM cell for near-threshold operation Single bit 9T SRAM

8-Bit 9T SRAM

22 nm

22 nm

16 nm

16 nm

MOSFETs

9

9

72

72

Area (nm)

192

144

1584

1152

Power (μW)

21.401

0.325

169.93

2.476

Delay (ns)

21.384

20.14

9.085

0.104

Input voltage (V)

0.8

0.6

0.8

0.6

Analysis of one-sided Schmi-Trigger based 9T SRAM Cell for Near Threshold Operaon 2000 1500 1000 500 0 22nm

16nm

Single Bit 9T SRAM MOSFETs

Area(nm)

22nm

16nm 8-Bit 9T SRAM

Power(μW)

Delay(ns)

Input Voltage(V)

Fig. 2. Analysis of one-sided Schmitt-trigger based 9T SRAM cell for near threshold operation

A 12T SRAM of 16 nm CMOS Technology

389

Table 2. Comparison of one-sided Schmitt-trigger based 9T SRAM and 12T SRAM cells. Comparison of ST 12T SRAM cells with ST 9T SRAM cell ST 9T SRAM

ST 12T SRAM-S1

ST 12T SRAM-S2

ST 12T SRAM-S3

22 nm

16 nm

22 nm

16 nm

22 nm

16 nm

22 nm

16 nm

MOSFETs

9

9

12

12

12

12

14

14

Input voltage (V)

0.8

0.6

0.8

0.6

0.8

0.6

0.8

0.6

Area (μm)

0.198

0.144

0.286

0.208

0.286

0.206

0.330

0.240

Min. power (μW)

46.02

0.42

1.13

0.13

0.80

0.12

10.12

0.56

Delay (ns)

9.14

6.17

40.53

40.17

40.52

40.10

40.59

40.15

Analysis of one-sided Schmi-Trigger based 9T SRAM and 12T SRAM Cells. 50 40 30 20 10 0 22nm

16nm

ST 9T SRAM MOSFETS

22nm

16nm

ST 12T SRAM-S1 Input Voltage(V)

22nm

16nm

ST 12T SRAM-S2

Area(μm)

22nm

16nm

ST 12T SRAM-S3

Min. Power(μW)

Delay(ns)

Fig. 3. Analysis of one-sided Schmitt-trigger based 9T SRAM and 12T SRAM cells.

390

K. Challa and V. K. Pamula

Fig. 4. Output waveforms of (a) 12T-S1, (b) 12T-S2, (c) 12T-S3 SRAM cells.

A 12T SRAM of 16 nm CMOS Technology

391

5 Conclusion In this work, we suggest three area of expertise read ports for SRAM cell with improved data independent read port leakage identified as high performance, low power, and low area. Since none of the intended read ports had any PMOS, the NMOS size was reduced, resulting in lower vertical sizing and shorter bit lines in thin form architectures. This decreased the amount of area per cell and the amount of energy used in each access of read and write operations. Each cell with the intended read ports increased the effective read operation with reduced dynamic power per read bit line, and its allowing for significant area savings through the sharing of peripheral hardware. As a result, the proposed ST 12T SRAM cell consumes less power, energy, and delay.

References 1. Calhoun, B., Chandrakasan, A.: Static noise margin variation for sub-threshold SRAM in 65-nm CMOS. IEEE J. Solid-State Circ. 41(7), 1673–1679 (2006) 2. Calhoun, B.H., Chandrakasan, A.P.: A 256-kb 65-nm sub-threshold SRAM design for ultralow-voltage operation. IEEE J. Solid-State Circ. 42(3), 680–688 (2007) 3. Dreslinski, R.G., Wieckowski, M., Blaauw, D., Sylvester, D., Mudge, T.: Near-threshold computing: reclaiming Moore’s law through energy efficient integrated circuits. Proc. IEEE 98(2), 253–266 (2010) 4. Kim, T.-H., Liu, J., Keane, J., Kim, C.H.: A 0.2 V, 480 kb subthreshold SRAM with 1 k cells per bitline for ultra-low-voltage computing. IEEE J. Solid-State Circ. 43(2), 518–529 (2008) 5. Pasandi, G., Fakhraie, S.M.: A 256-kb 9T near-threshold SRAM with 1 k cells per bitline and enhanced write and read operations, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., 23(11), 2438–2446 (2015) 6. Hu, V.P.-H., Fan, M.-L., Su, P., Chuang, C.-T.: Analysis of GeOI FinFET 6T SRAM cells with variation-tolerant WLUD read-assist and TVC write-assist. IEEE Trans. Electr. Devices 62(6), 1710–1715 (2015) 7. Yabuuchi, M., Nii, K., Tsukamoto, Y., Ohbayashi, S., Nakase, Y., Shinohara, H.: A 45 nm 0.6 V cross-point 8T SRAM with negative biased read/write assist. In: Proceedings of the IEEE Symposium on VLSI Circuits, pp. 158–159, June 2009 8. Gupta, S., Gupta, K., Calhoun, B.H., Pandey, N.: Low-power near-threshold 10T SRAM bit cells with enhanced data-independent read port leakage for array augmentation in 32-nm CMOS. IEEE Trans. Circ. Syst.-I Regul. Papers 66(3), 978–988 (2019) 9. Cho, K., Park, J., Oh, T.W., Jung, S.-O.: One-sided Schmitt-trigger-based 9T SRAM cell for near-threshold operation. IEEE Trans. Circ. Syst.-I Regul. Papers 67(5), 1551–1561 (2020)

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork Sukanya Nag, Deepsikha Bhattacharjee, Archisman Bhaumik, and Suman Deb(B) National Institute of Technology Agartala, Agartala, India [email protected]

Abstract. An approach for enhancing the quality of vectorised images for robotic artwork, that act as input for CNC device, is addressed in this paper. Attempt to upgrade the vectorisation method has been made by comparing and combining the available edge detection techniques. Accordingly the scope of improving plotter functionality to achieve a human-like drawing from the plotter has been discussed. In the era of automation, the XY plotter is a powerful tool capable of producing artwork without external assistance. The plotter relies on vectorised images and thus the initial stage of experimentation includes creation of accurate vectors from ordinary images. The resultant vectors are processed by the CNC plotter as substantial artwork on a drawing surface. The conversion mechanism for producing drawing objects from bitmap image is time consuming. The paper proposes a deep learning based model which allows robust vectorisation of these images following a multi-stage approach. Keywords: Vectorisation · Deep learning · Convolutional Neural Networks · Edge detection · Spatial robotic artwork

1

Introduction

Inspiration behind this work lies in vector formation which is a crucial part of artwork creation. The XY plotter [14] is the CNC device used to physically sketch the input image onto the drawing surface. A recurring problem of using ordinary bitmap images as input for the plotter is the blurring of the image. Bitmap images use pixels for representation which when enlarged can result in the distortion of the image. This causes a hindrance in acquiring clean output sketches as outputs from the device. As a result, vectorised images are chosen as inputs since here, the edges and boundaries are represented by individual vectors or Bezier curves. In case of a vectorised image, the image depends upon the equation of curves and not on quantity of pixels. This is why, even on scaling the images infinitely, we get a clear and distortion free image. A vectorised image constructed from a base image which could either be a hand-drawn sketch or photograph from a camera can be used as the input which is to be fed to the XY plotter. c The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022  S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 392–405, 2022. https://doi.org/10.1007/978-981-19-1742-4_33

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

393

Since vector images are a common input source for XY plotters, it becomes significant to produce accurate and efficient vectors while working with similar CNC devices. The underlying logic of vectorisation is extracting the object boundaries from an image and converting them into vectors or mathematical curves having directions. Edges are the sudden changes of discontinuities in an image [17]. Any change in lighting conditions, shaded regions or texture can result in edge formation. A substantial segment of computer vision, image processing and in turn, vectorisation, is the concept of edge detection. In most cases, the initial stage of information retrieval from an image is edge detection. The process notably reduces the amount of data to be processed and filters out unnecessary information while preserving important structural features of the image. The classical edge detection mechanisms [7,9,16,20,23] are categorised into either Gradient or Gaussian. A number of edge detection techniques are available but each have their own drawback. The qualities of edges detected by the classical techniques are greatly dependent on external factors such as lighting conditions, noise, objects having similar intensities and edge density. Edge detection can be a complex and time consuming process especially for inputs involving presence of noise. Analysing the connected components of the image can become challenging using edge detection in situations involving noise. False edges may get detected which interferes with the outcome of the experiments. This is tremendously problematic while dealing with spatial artwork since every small detail adds up to a significant output. The variety of operators available do not produce the desired output after hardware implementation. The steps include filtering, enhancement and detection but the final result is a mediocre localisation of edges which does not meet the required criteria. The approach presented in this paper targets these connected components and attempts to remove undesired noise in order to retain the necessary details. In this paper, we are interested in a comparison between Canny, Sobel and HED edge detection methods and present the suggested CNN based algorithm that properly synchronises with the X-Y plotter hardware set up. The mentioned algorithm attempts to incorporate the required features of the classical edge detectors in a Deep Learning model to favour the production of spatial robotic artwork. As the physical component of the set up, an X-Y plotter is used which is a CNC device that works on two stepper motors, an L293D driver and Arduino Uno microcontroller. The plotter uses a pen to draw across a plane surface. Vectorised images are fed to the plotter and the resultant image is drawn by the device itself. Later in this paper, a comparison is shown between MSED, which is the proposed algorithm, and the previously known techniques when used individually. The comparison should be sufficient to prove the efficiency of the proposed method of edge detection for spatial robotic artwork.

2

Literature Survey

Analysis of digital images can be simplified with the help of edge detection which deals with identifying pixels with high variations in intensity [17]. Edge detection is a primary step in measuring object size, isolating a desired object, object

394

S. Nag et al.

recognition and classification. With the advancement of small devices and robust computation system, edge detection plays a vital role in image intelligence. If we take into account the human brain, it divides objects and its surroundings by colour. When we discuss image intelligence, we refer to the ability of the system to mimic the human brain in distinguishing objects in its surroundings [25]. An ideal system can accurately distinguish among all kinds of edges. There are several edge detection algorithms currently in use. The most popular among these edge detection algorithms, Canny [7] has certain drawbacks. Since Canny, most of the successive methods have been using non-maximum suppression [6] as the ultimate process in edge detection. A significant problem that was failed to be addressed in the classical methods is the ambiguity in edge detection leading to subjective contours [12]. The edges are always challenged by gradient colour, even proper identification of edges is ambiguous due to close colour representation. The parameters reducing amount of the least ambiguous edges of a detection technique could be considered as the significant ones. The concept of removal of the background, which is based on edge detection and contours, from the foreground has also not been touched upon in the classical edge detection techniques. It is proven to be of use in video calls, determining positions of obstacles in autonomous vehicles, and so on. To overcome these barriers in the path of efficient edge detection, various researchers have provided possible solutions aided by Convolutional Neural Network (CNN) and Deep Learning architectures, like HED [26], DeepEdge [4], DexiNed [21], etc. [10,15]. Although the CNN based schemes outperform the classical edge detectors, they possess a higher computational cost. Since edge detection is a rather simple job in comparison to image segmentation or object detection, the complicated architecture and enormous amount of parameters may be unnecessary. To reduce the complexity, few algorithms were proposed which generated thin edge-maps or good edges without previous training or fine tuning process [21] or by implementing traditional method inspired frameworks [24].

3

Edge Detection for Vector Creation

Edges are the locations in a digital image where there is a sharp change of intensity. A unit change from one pixel to the next can be considered as an ideal edge. Edges are classified into the following types: • Horizontal edges • Vertical edges • Diagonal edges Edges are useful for segmentation, registration, and identification of objects [18]. In image analysis, a problem of fundamental importance is the detection of these edges. Edge detection [17] is defined as the technique of image processing based on regions of discontinuity. Classical edge detection methods are categorised into two types: • Gradient

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

395

• Gaussian Gradient is used to determine the first-order derivatives of the intensity in the image whereas Gaussian methods are implemented to find the second-order derivatives. The process in general, is comprised of a series of fundamental steps that are as follows: • Image smoothing: suppresses the noise without destroying true edges as shown in Fig. 1 • Detection of edge point: determining the pixels that should be retained while discarding the ones that contribute to noise • Edge localisation: determines the exact location of the required edge Many factors are taken into consideration while performing edge detection, such as noise, brightness of the image, corners and edge-like features. 3.1

Sobel Edge Detection

The Sobel operator or filter [16], also known as Sobel-Feldman operator, works on the principle of discrete differentiation to compute the gradient approximation of image intensity function. Two 3 × 3 kernels are taken to be convolved with the input image which calculates the horizontal and vertical derivative approximations respectively. Although the Sobel Edge Detector is beneficial for simple and time efficient computation and detection of smooth edges, it is highly sensitive to noise and fails to give appropriate results for thick and rough edges. Conclusively, it is not considered to be very accurate in edge detection. 3.2

Canny Edge Detection

Canny Edge Detector [7] is Gaussian-based and not susceptible to noise. Image features are extracted without affecting the feature. Canny edge detector has an advanced algorithm which is derived from the previous work on Laplacian of Gaussian operator. Plus points held by the Canny operator are mainly good localization, extraction of features without altering them and minimised noise sensitivity; whereas false zero crossing, complex computation and higher time consumption attribute to the limitations possessed by this edge detector. 3.3

Holistically-Nested Edge Detection

Holistically-Nested Edge Detection, [26] or HED, is an end-to-end system that uses a trimmed VGG-like convolutional neural network for an image to image prediction task. HED automatically learns rich hierarchical representations that are important in order to determine the edge or object boundaries. The edge map produced by this algorithm performs better at preserving object boundaries in comparison to the Canny Edge Detector.

396

S. Nag et al.

Canny edge detection can be overtaken by Holistically Nested edge detection in applications where the environment and lighting conditions are unknown or uncontrollable, as the former requires preprocessing steps, manual tuning of parameters, and often does not perform well on images captured using varying lighting conditions. However, on the downside, HED is more expensive than Canny in terms of computational cost. For real-time performance, Canny would just require a CPU but HED would need a GPU for the same purpose.

4

Experiments on Edge Detectors

Here we have taken a standard image representing the desired category of spatial robotic artwork. To achieve expected results, some pre-processing steps could be carried out according to the necessity, before applying the Canny/Sobel/HED operator. Image smoothing is one of the common pre-processing steps in various edge detection algorithms and could be achieved using filters such as Gaussian or Median blur. Traditional Canny algorithm makes use of Gaussian blur which seems to eliminate lot of information along with the noise. The edge details get smoothened significantly. However, application of Median blur does not create the same problem as edge details are retained even after removal of noise, as can be inferred from Fig. 1. Following the preprocessing step, Canny operator is applied, followed by removal of noise and then vectorisation of the output. The process is repeated with Sobel operator and HED as shown in Fig. 3.

Fig. 1. Two of the techniques used for image smoothing

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

397

Fig. 2. Noise and unnecessary details retained in (a) Sobel (b) Canny and (c) HED detected images make them unsuitable for our purpose

Comparisons on Edge Detectors. As it can be inferred from the above experiments in Fig. 2, the edges formed from Sobel edge detection have low intensity and thus some significant edges get removed while vectorisation. The output from HED contains edges having better intensity than the Sobel output but they are not as sharp as the Canny output. Canny edge detection algorithm produces sharper edges in the vectorised output in comparison to Sobel and HED. These vectorised outputs will not produce an efficient artwork as unwanted objects are still retained. The formation of double lines will also add up to production of unwanted edges while drawing with the XY plotter. The presence of noise, as shown encircled in the images, is a hindrance in achieving the desired effect of human-like drawing. The vector images produced from these algorithms do not meet the desired need and hence, we aim to implement an algorithm that would give more robust results.

5

Multi-Stage Edge Detection

The Multi-Stage Edge Detection (MSED) algorithm introduced in this paper attempts to combine the refined aspects of classical edge detection algorithms with a CNN model assuring the production of robust vector images. The approach followed in MSED can be broken down into five steps : 1. Image pre-processing 2. Application of classical algorithm (Sobel, Canny or a combination of both) 3. Noise Removal 4. Application of a CNN model 5. Vectorisation

398

S. Nag et al.

Fig. 3. The algorithm flowchart for MSED and related comparison

5.1

Image Preprocessing

Image pre-processing is an optional step which depends on the quality of the image. This step mostly includes the primary edits which refine the input image before application of edge detection techniques. Contrast and brightness of the image are the targeted settings which, when adjusted properly, enhance the quality of the image before the further techniques can be applied to it.

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

5.2

399

Application of Classical Algorithm

Any of the two mentioned classical techniques can be used as the first step of MSED. However, since Canny has proven to be more efficient in producing sharp edges than Sobel, the former is preferred. The Frei-Chen gradient calculation algorithm [2] replaces the common gradient calculation in Canny edge detection technique, which makes up for the deficiency of the latter. The calculation method is shown below: ⎤ ⎡ −1 0 √1 √ 1 √ ⎣− 2 0 2⎦ rowgradient (1) 2 + 2 −1 0 1 √ ⎤ 2 1 1 1 √ ⎣ 0 0 0 ⎦ columngradient 2 + 2 −1 −√2 −1 ⎡

5.3

(2)

Noise Removal

Filtering out unnecessary image data is a standard process used for almost every image processing purpose. Filters are used for achieving this by removing noise from images while preserving the details of the same. Image denoising [5] is the process of noise removal from an image. Noise originates by several ways such as image acquisition, transmission, and compression. Since presence of noise in an image leads to loss of valuable information, it is essential to remove the noise and recover the original image from the degraded images where getting the original image is important for robust performance or in cases where filling the missing information is very useful. Several filters are available for noise removal, used according to the type of noise and the requirement. Broadly, the filters are classified under two domains - Time and Frequency. Mean, Gaussian and Median, belonging to the Time Domain, are few of the commonly used filters in pre or post processing of images for removal of noise. The original Canny algorithm uses the Gaussian filter, making the edge detail seem blurry whereas our expected results require intense edges. This can also cause creation of false edges during vectorisation. To solve this problem, we pick the Median filter for our algorithm for preserving the required details of the image while removing the noise as evident from Fig. 1. In the median filter, which can be classified as a low-pass filter, a window slides along the image, and the median intensity value of the pixels within the window becomes the output intensity of the pixel being processed outputs the median intensity value of the pixels. This step provides a better result in comparison to the Gaussian counterpart.

400

5.4

S. Nag et al.

Application of a CNN Model

A human brain is intelligent and thus can easily process any form of data that is provided. On seeing an image with multiple components, the human can identify the various objects and their connected parts. When a robot tries to mimic this natural human behaviour, the concepts of image segmentation, object recognition and analysis of connected components are crucial. These concepts are taken into consideration while building the CNN model. The benefit of using a Convolutional Neural Network (CNN) based Deep Learning model [1] in edge detection is that it generates intense edges without the need of any post processing steps. CNN is applied in the later stage of the MSED algorithm to generate the desired image. Even after the detection of edges using multiple edge detectors, most of the times various post processing steps have to be applied in order to generate the finer edges and reduce the quantity of noise, before it could be used in concerned applications. However, incorporation of CNN based models has helped in obtaining the desired results. Usage of convolutional network layers provides two major advantages: • Parameter sharing [22] • Sparsity of connections [8] Parameter sharing is motivated by the observation that a feature detector, like a horizontal edge detector, that’s useful in one part of the image could be probably useful in another section of the image as well. This implies that, if one figures out, say a 3 × 3 filter for detecting horizontal edges, the same can be applied at another region, for example, the lower right-hand corner, and then another position over to the left, and so on. Hence, the feature detectors can use the same parameters in lots of different positions in the image. Application of CNN has proven to be very efficient as it can compute all features automatically, hence reducing the burden of manual feature extraction. Moreover, feature detection using CNN results in more accurate and precise outcomes when compared to manual feature compilation methods. For the proposal in this paper, the connected component analysis takes an important role. The underlying concept is to identify the pixels that belong to a particular object by exploiting the pixel neighbourhood. The input is a binary image in the form of a matrix where pixels representing background can be represented by false(0) values and the components in the foreground have true(1) values. Connected objects in the foreground are identified using two rules followed throughout the process. Firstly, jumps (traversing through pixels to find connected objects) are allowed only along a row or a column. The diagonal jumps are not allowed. In the second rule it is stated that in a sequence of multiple jumps, a jump in row and column direction can be made only once and it has to be orthogonal. The importance of analysis of connected components is realised while vectorising the image. An input consisting of a clean outline of the components in the image is a necessity for a refined output of the Multi-Stage edge detection algorithm.

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

401

Fig. 4. Architecture of the CNN model

Our CNN-based model consists of Feature Extractor, Feature Enhancer and Feature Summarizer, as shown in Fig. 4, which roughly corresponds to gradient, low pass filter and pixel connection respectively in the classical edge detection techniques. The feature extractor is a 3 × 3 convolutional neural network layer. 16 feature channels are used which are initialized with the 16 directional gradient kernels. Rest of CNN layers are initialized with zero-mean, 0.01 standard deviation Gaussian and zero biases. In the Feature Enhancer module, dilated convolutions are used for creating multiple scale filtering. A larger receptive field can be captured with fewer parameters. This module is named Enhancer as it extracts object information along with the edges. The Feature Summarizer module, which is the last one, summarizes the features generated by the Enhancer modules and produces the final edges. Eight 1 × 1 convolutional layers are applied along with a sigmoid activation function. 5.5

Vectorisation

The final step of the entire technique leading up to the finished output is the creation of desired vector images [3]. Since the motivation behind the work stated in the paper was derived from the creation of spatial robotic artwork primarily for the use of computer controlled devices [14], this step serves as evidence to the entire work that has been done so far. In computer graphics, vectorisation, also called image tracing or raster-vector conversion is the conversion technique of raster graphics into vector graphics. Bitmap images are represented through pixels because of which it tends to resize improperly and get pixelated on being enlarged. This problem is solved using vector graphics where edges in the image are represented through equations of Bezier curves or curved paths instead of pixels. Bezier curves display a prominent role in computer graphics, especially in raster to vector conversion as the dynamic nature of these curves make them appropriate for defining boundaries. They also help in refitting deformed shapes in raster images which often become distorted in quality when magnified to a greater extent. During experimentation for this paper, Inkscape has been used for vectorisation. G-code tools are provided with Inkscape to allow G-code generation from vector images. G-code is the most popular programming language which is used for controlling machines such as a 3D-printer or, in this case, an XY plotter. It has many variants yet a large number of them follow some common set of rules.

402

S. Nag et al.

The use of the Arduino for the XY plotter is facilitated through the built-in GRBL [19] firmware. GRBL then uses the G-code for controlling the hardware. Single line g-code blocks are sent to GRBL, which is an open-source software designed for three-axis control of stepper-motor based CNC machines, awaiting for an acknowledgement. GRBL sends an acknowledgement once processing of the block is completed and there is room in the buffer. GRBL receives ASCII commands and G-code and performs smooth linear and circular interpolation for the stepper motors. It operates through the Arduino serial port interface and requires a constant stream of g-code commands sent through a computer. The single g-code blocks are accepted and processed, followed by a carriage return. It returns an ‘ok’ or ‘error:X’ message after the block is processed and is ready for further information. The entire process of getting the desired output on the drawing board after feeding the bitmap image is shown in the Fig. 5 below.

Fig. 5. Input bitmap image to final drawing

6

Experiments on MSED

The implementation of our model is achieved through Pytorch. The Gaussian distribution algorithm is used to initialize all the filter weights except in the first stage of edge detection, which is Feature Extractor 1. The mean is initialized to 0 and standard deviation to 0.01, and the biases are initialized with 0. The hyper-parameters are: mini-batch size = 1, learning rate = (1e−2), weight decay = (5e−4), momentum = 0.9, and training epochs = 120. For every 10 epochs, the learning rate is divided by 10. We utilize the SGD solver for optimization. All the experiments are conducted with the assistance of NVIDIA GeForce 2080Ti GPU with 11 G memory. The steps mentioned in Sect. 5 are followed sequentially till step D to obtain the left image, shown in Fig. 6. Post vectorisation as on Step E, the image on the right is produced which is the ultimate output of the new algorithm. The experimental output has shown significant improvement from the existing algorithms while dealing with reproduction of spatial artwork using XY robotic plotters.

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

403

Fig. 6. The final output after MSED and vectorisation respectively

Finally the image, which has been edge detected followed by vectorisation, on being uploaded to the robotic plotter, produces an artwork on the drawing surface with the assistance of a pen. Considering the fact that the XY plotter is a project still awaiting completion in terms of hardware assembling, we provide sample images, Fig. 7, that depict the expected outcome on the surface.

Fig. 7. Sample vectorised artwork drawn on surface Source: Adapted from [11, 13]

7

Conclusion

The proposed Multi-Stage Edge Detection technique, based on a blend of classical and convolutional neural networks, has given expected outputs which are computationally faster and qualitatively better for vectorisation, in comparison to the existing algorithms. While emphasising on spatial robotic artwork, the images formed by the application of the method we formulated are comparable to an actual humandrawn sketch which usually consists of outlines of the objects and is devoid of any noise and unwanted objects. Our technique helped in obtaining the desired

404

S. Nag et al.

form of sketch we strived to achieve through the plotter device. This outcome will provide the XY robotic plotter with an additional functionality of reproducing the artwork, fed to it as input, in form of a sketch. Hence this approach demonstrates promising results and has the potential to further contribute greatly in overcoming the targeted working limitations of computer numerical controlled devices.

References 1. Ahmed, A., Byun, Y.-C.: Edge detection using CNN for roof images, pp. 75-78 (2019) 2. Apdilah, D., Simargolang, M., Rahim, R.: A study of Frei-Chen approach for edge detection. Int. J. Sci. Res. Sci. Eng. Technol. 3, 59–62 (2017) 3. Bera, A.: Fast vectorization and upscaling images with natural objects using canny edge detection. In: vol. 3, p. 4 (2011) 4. Bertasius, G., Shi, J., Torresani, L.: DeepEdge: a multi-scale bifurcated deep network for top-down contour detection (2015) 5. Buades, A., Coll, B., Morel, J.M.: A review of image denoising algorithms, with a new one. Multiscale Model. Simul. 4(2), 490–530 (2005) 6. Canny, J.: Finding edges and lines in images. Theory of Computing Systems Mathematical Systems Theory, p. 16 (1983) 7. Canny, J.: A computational approach to edge detection. IEEE Trans. Pattern Anal. Mach. Intell. 6, 679–698 (1986) 8. Changpinyo, S., Sandler, M., Zhmoginov, A.: The power of sparsity in convolutional neural networks (2017) 9. Chaple, G.N., Daruwala, R.D., Gofane, M.S.: Comparisons of Robert, Prewitt, Sobel operator based edge detection methods for real time uses on FPGA. In: 2015 International Conference on Technologies for Sustainable Development (ICTSD), pp. 1–4 (2015) 10. Chaudhary, D.K., Lal, R., Kashyap, N., Choudhury, T.: Hybrid edge detection technique for digital images, pp. 1116–1121 (2016) 11. Das, A.K.: How to make an Arduino drawing machine? November 2018 12. Davis, S., Jernigan, E.: Ambiguous edge detection leads to subjective contours. Percep. Psychophys. 31, 93–94 (1982) 13. Good, I.J., Vite, M.: Science in the flesh in cybernetics, arts and ideas by Reichardt, J. 1971 (1968) 14. Jegan, R.R., Gnanasundaram, E., Gowtham, M., Sivanesan, R., Thiyagarajan, D.: Modern design and implementation of XY plotter, pp. 1651–1654 (2018) 15. Yang, M., Shan, Y., Huang, T., He, J., Zhang, S.: Bi-directional cascade network for perceptual edge detection 16. Kanopoulos, N.: Design of an image edge detection filter using the Sobel operator. IEEE J. Solid State Circuits 23(2), 358–367 (1988) 17. Marr, D.C., Hildreth, E.: Theory of edge detection. Proc. Roy. Soc. London 207, 187–217 (1980) 18. Rosenfeld, A., Thurston, M.: Edge and curve detection for visual scene analysis. IEEE Trans. Comput. C-20(5), 562–569 (1971) 19. Sarguroh, S.S., Rane, A.B.: Using GRBL-Arduino-based controller to run a twoaxis computerized numerical control machine, pp. 1–6 (2018)

Multi-Stage Edge Detection for Generative Spatial Robotic Artwork

405

20. Seif, A., Salut, M.M., Marsono, M.N.: A hardware architecture of Prewitt edge detection. In: 2010 IEEE Conference on Sustainable Utilization and Development in Engineering and Technology, pp. 99–101 (2010) 21. Soria, X., Riba, E., Sappa, A.: Dense extreme inception network: towards a robust CNN model for edge detection (2020) 22. Terry, J.K., Grammel, N., Hari, A., Santos, L., Black, B.: Revisiting parameter sharing in multi-agent deep reinforcement learning (2021) 23. Wang, X.: Laplacian operator-based edge detectors. IEEE Trans. Pattern. Anal. Mach. Intell. 29, 886–890 (2007) 24. Wibisono, J.K., Hang, H.-M.: Traditional method inspired deep neural network for edge detection. In: 2020 IEEE International Conference on Image Processing (ICIP) (2020) 25. Wu, Q.X., McGinnity, M., Maguire, L., Belatreche, A., Glackin, B.: Edge detection based on spiking neural network model, pp. 26–34 (2007) 26. Xie, S., Tu, Z.: Holistically-nested edge detection. In: Proceedings of the IEEE International Conference on Computer Vision (ICCV) (2015)

Design and Analysis of Micro-grid Stability with Various DGs B. Devulal1(B) , M. Siva2 , D. Ravi Kumar1 , and V. Rajashekar1 1 Department of EEE, VNR Vignana Jyothi Institute of Engineering and Technology,

Hyderabad 500090, Telangana, India {devulal_b,ravikumar_d}@vnrvjiet.in 2 Department of EEE, Annamalai University, Chidambaram 608002, Tamil Nadu, India

Abstract. A Micro Grid (MG) is an isolated electric grid that comprises several elements which are the same as that of the distributed electric grid. The paper presents a total model for optimization of the hybrid solar and wind energy in isolated Micro Grid (MG) by implementing the MPP tracking, including a spare rechargeable battery. MG typically works in ordinary interfacing mode. At the point where an extreme shortcoming happens in the essential conveyance system, then, at that point, the MicroGrid will tend to island mode. In this paper, the model exhibits a solitary shaft miniature Wind turbine, photovoltaic system and a battery. This load of miniature sources is associated with the Micro Grid through inverters with the exception of the breeze age framework. with two control techniques of inverters are illustrated. Specific dynamic and responsive powers are infused into the system which is of PQ control by the inverter. This type of inverter is utilized to interface miniature turbine, power module, and photovoltaic boards to the Micro Grid. Vf control is a consequent technique. The first case considers the effect of islanding which measures on the frequency, voltage, and dynamic force of all tiny sources when the Micro Grid imports dynamic and responsive force from the key conveyance system. The impact of islanding on previous quantities is also demonstrated in the second analyzed instance, particularly when the MG sends out dynamic and respective capacity to the vital dissemination system. Power simulations are designed on the MATLAB platform and power converter efficiencies, power quality performance variables are demonstrated. Keywords: MG · Dynamic performance · Islanding · Inverter and distributed generators · Power quality

1 Introduction The electric power being the central spine for on ongoing modernization around the world suffers from aging and weak infrastructures, to be more specific the power system even in the today’s generation is prone to more outages and failures. To deal with these events a paradigm shift is necessary in the power sector. In the up coming decades the renewable sources play a key role in generating the green power [1] in contrast to the © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 406–414, 2022. https://doi.org/10.1007/978-981-19-1742-4_34

Design and Analysis of Micro-grid Stability with Various DGs

407

fossil fuels. So many concepts are being introduced into the power engineering to adapt to the upgrading advancements in electric power. Since electric grid is complex distributed system of various interfacing elements that act to drives the various linear and nonlinear loads. This has led to the introduction of the isolated grids by implementing the various renewable sources and smart controllers that can regulate the voltage, current and power quality for loads which are simply coined as microgrids. In this paper, show the recreation of the PV and wind fuel cell fly wheel based microgrid to show the new properties of the interconnected framework with the implementation of various smart controllers which are discussed in the following sections of paper.

2 Literature Survey The non-renewable power generations inject flue gases polluting the atmosphere. By using Renewable energy Sources, One can prevail over these problems to a large extent such as solar, wind, geothermal, etc. [1] that are greener and ecofriendly. The enhanced performance of hybrid power generation is presented in literature. Power quality enhancements in the PV-wind integrated grid are discussed in [2– 4], the nonlinearities, harmonics that are occurring are primarily related to the network’s existence of nonlinear loads. Maximum Power extracts from solar and wind by employing P & O and PSO Algorithms [5–8] and with the following controllers. • • • •

AC/AC converter with electrical component rectifier. Several DC/DC Vehicle battery charger converters. Supercapacitor battery. Distinctive DC/DC capacitor interface converter.

3 Proposed System Configuration The hybrid system includes two additional power sources connected in parallel. The energy assets of this device are photovoltaic solar cells and wind energy. The nonlinear loads in the system are powered by the hybrid energy of solar PV system and wind-powered system [2] as depicted in Fig. 1. The power generated from the PV system is routed through a boost converter. The pulses to the boost converter are applied through the fuzzy logic control system. On the other hand, the wind generation system is synchronized to the alternator for additional power or as backup power to the system. The power from boost converter passes on to the VSC through battery storage system connected across the Dclink capacitor. The various elements of the system design areas follows:

408

B. Devulal et al.

Fig. 1. Proposed system configuration

3.1 Source To power the nonlinear load of 500kW and an SG of 900 kVA, 20 kV, 50Hz coupled to a wind turbine system. By MPPT strategies One can achieve maximum power and the PV array is designed in such a way. 3.2 Buck Boost Converter The boostBuck Boost converter is modelled with the fuzzy logic controller for the effective generation of pulses for the MOSFET switching in operating at the peak efficiency. The inductor value in the converter topology is designed from: L=

(Vo − Vin) ∗ Vin henry Ipv ∗ frequency ∗ Vo

3.3 Dc Link Capacitor and Battery System Capacitor value is calculated from Cdc =

(Vo − Vin) ∗ Vin farad 2w∗Vdc2

Design and Analysis of Micro-grid Stability with Various DGs

409

A battery of higher power rating is chosen to withstand the power generated from the boost converter and to mitigate the disturbances and harmonics by incorporating the required filters. 3.4 VSC Designing For minimize the conduction and switching losses, VSC’s are designed to extract the maximum power from system Control is predictable in conventional converters because of its simple structuring. The modern VSC are complex to evaluate duty cycle ratios and PWM methods are too difficult to implement which are unpredictable. So these VSC’s are costly both in maintenance and operation. Such converters when incorporated with cost optimized functions and predictive control strategy it evades the limitations detailed. The voltage in each power converters leg is calculated from Vxn = Sx Sx Sx Vxn

(1)

V0 = Vxn Req I0 − L

(2)

For detailing the behavior of the circuit model in Fig. 2 the voltage and current quantities in vector form are Vpul = [ VR VY VB ]

T

(3)

VPwm = 3 ph inv o/p to neutral line voltages Ii = [I R IY IB ] Vl = [ VR VY VB ]

(4) T

(5)

Equation from load view: VL = Rf + Iinv + Lf + Vpwm − Rn In −Ln

(6)

Iinv = IL + Cr

(7)

4 Modeling of Wind Turbine Based Synchronous Generator The wind turbine converts the wind power into the electrical power through an arrangement of mechanical gear settings. The wind power is given as Pw =

pAv3 w 2

The power generated from wind depends on the turbine efficiency and given as Pm = Pw ∗ Cp = 0.5ρL2v3 Cp

410

B. Devulal et al.

The parks transformation equations for the alternator are as follows: Stator side equations: ed = pd − q ωr −Ra id

(8)

ed = pq + d ωr −Ra iq

(9)

ed = pq + d ωr −Ra iq

(10)

d = −Ld id + La fd if + La kd ikq

(11)

o = −Lo io

(12)

efd = pfd + Rf difd

(13)

0 = pkd + Rkd ikd

(14)

0 = pkq + Rkd ikd

(15)

kd = Lfkd ifd + Lkkd ikd − 1.5Lafd id

(16)

kd = Lfkd ifd + Lkkd ikd − 1.5Lakd id

(17)

fq = Lffq ifq + Lfkq ikq − 1.5Lafq iq

(18)

kq = Lkkq ifq + Lfkq ikq − 1.5Lakq iq

(19)

Pe = 0.67(ed id + eq iq )

(20)

Rotor side equations:

Power output is

Depending on above equation the simulation model is designed for the system to analyze the performance of the variables.

5 Simulation Results The proposed simulations are verified in MATLAB/Simulink under various loads, it output performance is depicted in various figures as follows: Waveforms with LLLG Fault in Micro Grid Output Voltage of Synchronous Generator is shown in Fig. 3.

Design and Analysis of Micro-grid Stability with Various DGs

Fig. 2. Simulation diagram for proposed system.

Fig. 3. Synchronous generator output voltage

411

412

B. Devulal et al.

Wave form of line current is shown in Fig. 4.

Fig. 4. Line current

Wave form of load current is shown in Fig. 5.

Fig. 5. Load current

Wave form of battery current is shown in Fig. 6.

Fig. 6. Battery output current

Design and Analysis of Micro-grid Stability with Various DGs

413

Wave form of battery voltage is shown in Fig. 7.

Fig. 7. Battery voltage

6 Conclusion This paper considered a comprehensive modeling of micro grid behavior. Each one of MG’s components is meticulously displayed. The first case investigates the MG’s operation following islanding, when the MG imports dynamic and receptive forces from the principal matrix. The subsequent case shows the unique exhibition when the miniature lattice sends out a lot of dynamic and receptive forces to the fundamental matrix. It was demonstrated that the capacity gadgets are significant to execute sufficient control techniques for MG activity in islanded mode. The need of capacity devices is due to the fact that the MG’s micro sources have exceptionally low inactivity and sluggish slam up rates. A combination of hang control mode (applied to Vf inverter) and a fundamental control circle (applied to controlled tiny sources) is effective in reducing recurrence during islanded activity. In any case, MG should include at least one controllable small source (PV power device or miniature turbine) to aid in rebuilding when islanding occurs. Even if there are no controllable tiny sources in the MG, the capacity devices will continue to pump power into the MG until their energy is depleted and dark out occurs. Creator’s subsequent stage research means to consider the dynamic execution of the MG under various aggravating conditions like disturbances in one of miniature sources, faults in MG feeders etc.

References 1. Hongkai, L., Chenghong, X., Jinghui, S., Yuexi, Y.: Green power generation technology for distributed power supply. In: 2008 China International Conference on Electricity Distribution, Guangzhou, pp. 1–4 (2008) 2. Kamel, R.M., Chaouachi, A., Nagasaka, K.: Detailed analysis of micro-grid stability during islanding mode under different load conditions. Engineering 3, 508–516 (2011) 3. Pilawa-Podgurski, R.C., Perreault, D.J.: Submodule integrated distributed MPPT for solar photovoltaic applications. IEEE Trans. Power Electron. 28(6), 2957–2967 (2013) 4. Woei-Luen, C., Tsai, C.: Optimal balancing control for tracking theoretical global MPP of series PV modules subject to partial shading. IEEE Trans. Ind. Electron. 62(8), 4837–4848 (2015)

414

B. Devulal et al.

5. Shaw, R.N., et. al.: A new model to enhance the power and performances of 4 × 4 PV arrays with puzzle shade dispersion. Int. J. Innov. Technol. Explor. Eng. 8(12), 456–465 (2019) 6. Mohanty, S., Subudhi, B., Ray, P.K.: A new MPPT design using grey wolf optimization technique for photovoltaic system under partial shading conditions. IEEE Trans. Sustain. Energy 7(1), 181–188 (2016) 7. Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based backstepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 8. Karatepe, E., Hiyama, T.: Performance enhancement of photovoltaic array through string and central based MPPT system under non-uniform irradiance conditions. Energy Convers. Manage. 62, 131–140 (2012)

Risk-Seeker Information Gap Decision Theory Based Smart Grid Operation Encompassing Demand Response Tanuj Rawat1,2(B) , K. R. Niazi2 , Sachin Sharma3 , and Jyotsna Singh1 1

Swami Keshvanand Institute of Technology, Management and Gramothan, Jaipur, Rajasthan, India [email protected] 2 Malaviya National Institute of Technology, Jaipur, Rajasthan, India 3 Graphic Era (Deemed to be University), Dehradun, Uttarakhand, India

Abstract. This paper address optimal operational problem of smart distribution systems (SDS) encompassing uncertainties and demand response (DR). Information gap decision theory (IGDT) is adpoted in this work to model the uncertainties in grid prices and power from renewables. The SDS operation is analyzed for opportunity (risk-seeker) framework of IGDT. The proposed risk-seeker IGDT based SDS operational management problem is modeled as a multi-objective optimization approach to simultaneously optimize radius of uncertainty of both grid prices and power from renewables. The -constraint method is utilized to solve the multi-objective IGDT problem. Moreover, impact of different participation levels of load shifting type DR on SDS operation is also presented. In order to demonstrate the efficacy of the proposed work a modified 33bus distribution system is adopted. The results illustrate the effectiveness of proposed multi-objective IGDT model for SDS operational problem.

Keywords: Smart grid

1

· IGDT · Distribution system · Multi-objective

Introduction

Smart grid is emerging as a revolutionized transformation of traditional grid owing to greenhouse gas emissions, eroding fossil fuel reserves and surging demand of load. The smart grid facilitates the integration of two-way cyber communication and computer intelligence across the entire spectrum of power systems from generation to consumer end. Subsequently, the envisioned smart grid will be versatile and robust. Another feature of smart grid is efficient deployment of distribution energy resources (DERs) such as demand response (DR) and distributed generation (DGs). Most of these DERs will be located at the distribution end of smart grid. Thus, the integration of these DERs will transform the conventional distribution network into a smart distribution systems (SDS). However, the unpredictability associated with time varying electricity prices and renewable based DGs such as wind turbine (WT) challenges the secure operation c The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022  S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 415–425, 2022. https://doi.org/10.1007/978-981-19-1742-4_35

416

T. Rawat et al.

of SDS. The potential of SDS can, therefore, be leveraged only through optimal management of DERs in presence of uncertain resources. In this regard, various methods such as robust method [2,15], probability method [2,10] and fuzzy set [1] are reported by authors in literature to incorporate the input parameter’s uncertain nature in SDS operation. These methods, however, have some drawbacks. Robust method, for example, necessitate the exact uncertainty set; probability-based methods necessitate the accurate probability functions and fuzzy methods are computationally expensive and require membership functions. Recently, information gap decision theory (IGDT) which is a non-probabilistic method has shown great promise in managing uncertainties. IGDT method differentiates between actual value and forecasted value. Uncertainty in demand for a low-voltage distribution system using IGDT is addressed in [9]. Uncertainty in grid prices of renewable integrated microgrid is modeled using IGDT in [6]. Problem of voltage and congestion management in presence of WT uncertainty is studied using IGDT in [7]. Scheduling of flexible loads using IGDT to manage uncertainty in grid electricity prices is proposed in [16]. Two IGDT based functions, robustness function and opportunity function are presented in [3] to evaluate the risk associated with uncertain parameters while economically scheduling the resources of power system. Comparison of risk-neutral plan with risk-seeker (RS) and risk-averse (RA) IGDT strategies in controlling the uncertainties of market price in a smart microgrid is provided in [14]. AC load flow equations are not considered in [6,11,16]. Moreover, only one uncertain parameter is modeled using IGDT in [3,6,7,9,14,16]. Very few works such as [4,8,12,13] have simultaneously optimized multiple uncertainties using IGDT for operational problem of SDS. Weighted-sum based multi-objective IGDT model for energy management problem of microgrid without inclusion of grid constraints and DR is presented in [8]. Multi-objective RAIGDT based day ahead scheduling of SDS using enhanced -constraint method is illustrated in [4]. Though, the authors have considered DR through flexible loads, but a sensitivity analysis of DR participation levels is not investigated. An IGDT model formulated as mixed-integer non-linear programming (MINLP) for power management of SDS is proposed in [13]. Multi-objective RA-IGDT based operation of SDS using -constraint method to simultaneously tackle uncertian wind and grid prices is studied in [12]. However, [4,8,12,13] have only addressed RA strategy of IGDT. In other words, multi-objective RS-IGDT strategy for optimal operation of SDS has not been attempted. Based on above discussion, this paper presents a multi-objective RS-IGDT based SDS operation encompassing AC power flow constraints and DR. The proposed multi-objective approach simultaneously considers uncertain grid prices and uncertain WT power in operation of SDS. The multi-objective model is investigated for RS decision maker using opportunity function of IGDT. constraint method is developed in this work to solve the proposed multi-objective IGDT problem. Moreover, impact of different DR participation rate on opportunistic operation of SDS is also presented in this work. The proposed multiobjective frameworks are applied and investigated on the modified 33-bus system

Risk-Seeker IGDT

417

radial distribution system under various operating scenarios and performance indices have been analyzed in this work.

2

Problem Formulation

2.1

Deterministic Framework

One of the priorities of smart grid operator is to efficiently operate the network. Towards this goal, the objective function considered from smart grid operator’s perspective is to minimize cost of operating dispatchable DGs (DDGs) and cost of importing power from sub-station. This objective is mathematically expressed as follows: OC o =

NT N DDG  t=1

cDDGj PDDGj,t +

NT N sbs  

csbs,t Psbs,t

(1)

t=1 sbs=1

j=1

Here, OC o is SDS operating cost when uncertainty is not considered; NDDG , NT and Nsbs are number of DDGs, time intervals and sub-stations respectively; PDDGj,t and Psbs,t are power from DDG j and sub-station respectively; cDDGj and csbs,t are cost of power from DDG and price of power from sub-station respectively. OC o defined in above equation is subjected to AC load flow equations (2)–(7), DR constraints (8)–(12) and DDGs power limit constraint (8) respectively.   2 P Lcd,t − (P Lbc,t −Ibc,t Rbc ) ∀t ∈ Ω T , ∀c ∈ Ω N (2) P Gc,t −P Dc,t = cd∈Ω L

QGc,t −QDc,t =

bc∈Ω L

 cd∈Ω L

2 Vd,t

=

2 Vc,t



QLcd,t −

2 (QLbc,t −Ibc,t Xbc ) ∀t ∈ Ω T , ∀c ∈ Ω N (3)

bc∈Ω L

2 2 − 2(P Lcd,t Rcd + QLcd,t Xcd ) + Zcd Icd,t ∀t ∈ Ω T , ∀cd ∈ Ω L 2 2 Icd,t Vc,t

=

QL2cd,t

+

V ≤ Vc,t ≤ V

P L2cd,t

∀t ∈ Ω , ∀cd ∈ Ω T

∀t ∈ Ω T , ∀c ∈ Ω N

P Dc,t = T 

P Dc,t =

t=1

T 

∀t ∈ Ω , ∀c ∈ Ω T

(4) (5) (6)

Icd,t ≤ Icd ∀t ∈ Ω T , ∀cd ∈ Ω L o P Dc,t κc,t

L

(7) N

o P Dc,t ∀c ∈ Ω N

(8) (9)

t=1

1 − κc,t ≤ κc,t ≤ 1 + κc,t ∀t ∈ Ω T , ∀c ∈ Ω N

(10)

o QDc,t = QDc,t κc,t ∀t ∈ Ω T , ∀c ∈ Ω N

(11)

T  t=1

QDc,t =

T  t=1

o QDc,t ∀c ∈ Ω N

(12)

418

T. Rawat et al.

PDDGj ≤ PDDGj,t ≤ PDDGj ∀t ∈ Ω T , ∀j ∈ Ω DDG

(13)

Here, c and d are index for nodes; bc and cd are index for lines; Ω T , Ω N , Ω L and Ω DDG are sets for time, nodes, lines and DDG respectively; Zcd , Rcd and Xcd are impedance, resistance and reactance of line respectively; V and V are maximum and minimum voltage limit respectively; PDDGj and PDDGj are minimum and maximum power limit of DDG respectively; κc,t is maximum flexibility of load; o o Icd is maximum current through line; QDc,t and P Dc,t are reactive and active load before DR; QDc,t and P Dc,t are reactive and active load after DR; QGc,t and P Gc,t are reactive and active power injected; QLcd,t and P Lcd,t are reactive and active power flowing in lines; Icd,t is current flowing in line; Vc,t is voltage magnitude; and κc,t is flexibility of load. 2.2

IGDT Framework

IGDT assists the decision makers to handle systems having uncertain parameters. Unlike other methods, this method does not require any kind of statistical data pertinent to uncertain parameter. IGDT consists of two approaches, namely, risk-seeker (RS) and risk-averse (RA) strategies. In RS-strategy, the decision maker aims to maximize its profit through favorable variations in uncertain parameters whereas in RA-strategy, the aim of decision maker is to minimize the risk due to unfavorable deviation in forecasted values of uncertain parameters. In context of IGDT, the favorable or unfavorable deviations are represented using immunity functions. Robust function defines the immunity against failures whereas minimum horizon of uncertainty while ensuring that a pre-defined profit is met is expressed using opportuneness function. In RS-strategy an opportuneness function is defined which has positive effects on the objective of decision maker. In contrast, a robust function is expressed for RA-strategy. In this paper, the SDS operation intends to be a RS model that brings about positive effects for decision maker. The opportunity function η RS is defined as shown in Eq. (14). (14) η RS = min {η : min cost ≤ critical cost} η

The uncertainty in prices of power from sub-station and power from WT unit expressed as ut = {csbs,t ,Pwt,t } is modeled using IGDT. Here, demand flexibility, power from DDGs and power taken from sub-station are decision variables of the optimization problem. They are represented as qt = {κc,t ,PDDGj,t ,Psbs,t }. It csbs,t ,P˜wt,t } are accessible is assumed that forecasted values, denoted as u ˜t = {˜ to smart grid operator. The RS-IGDT based SDS problem is presented in Eqs. (15)–(19). (15) min αRS and min β RS s.t. sb

min

T

N N   sbs=1 t=1 c

csbs,t Psbs,t +

DDG N NT 

≤ OC = OC o (1 − δ)

j=1

t=1

cDDGj,t PDDGi,t

(16)

Risk-Seeker IGDT

419

c˜sbs,t (1 − β RS ) ≤ csbs,t ≤ c˜sb,t (1 + β RS ),

∀t ∈ Ω T

(17)

P˜wt,t (1 − αRS ) ≤ Pwt,t ≤ P˜wt,t (1 + αRS ),

∀t ∈ Ω T

(18)

constraints (2) − (13)

(19)

Here, α and β are radius of uncertainty for WT power and prices for power from sub-station; P˜wt,t is forecasted power from WT; c˜sbs,t is forecasted price of power from sub-station; OC c is critical/target SDS operating cost; δ is deviation factor; and Pwt,t is power from WT. Equation (16) indicates that the minimum SDS operating cost should be less than equal to SDS target operation cost. The target operation cost of SDS is expressed as some percentange of deterministic SDS operation cost. Constraints (17) and (18) correspond to the range of uncertain grid electricity prices and wind power. Operating cost in Eq. (16) attains the minimum value when availability of wind power is maximum while value of grid electricity prices is minimum. Therefore, bi-level RS-IGDT problem can be translated into a single level optimization problem.

3

Solution Methodology

Numerous methods such as -constraint method and weighted-sum approach are proposed in literature to handle the multi-objective optimization problems. In this repect, the -constraint approach is adopted in this paper because of its wide range of benefits over weighted-sum method. In -constraint method, the objective functions are classified into two categories, that is, main and secondary objective functions. In this method, optimization of main objective function is performed while restricting all the secondary objective functions by some amount. As a result, the multi-objective model is converted into a singleobjective optimization model. The single-objective optimization model evaluated from -constraint approach is expressed as [5]: max β s.t α≥

(20)

inequalities and equalities constraints

4

Simulation Results

The operation problem of SDS encompassing DR and uncertain parameters is demonstrated on a modified 33-bus system. The details of modified system are derived from [12]. The modified distribution system is assumed to be equipped with four DDGs. These controllable DGs are considered at node 25, 16, 13 and 8 respectively. The maximum power limit of DDGs are 300 kW, 300 kW, 500 kW and 500 kW respectively. In addition, WT of 500 kW and 750 kW are assumed to be installed at node 32 and 14 respectively. The input data pertinent to profile

420

T. Rawat et al.

of wind power, load demand, forecasted grid prices and operating cost of DDGs is adopted from [10–12]. Following scenarios are framed in order to evaluate the effectiveness of proposed IGDT based SDS operation. • Scenario 1) In this scenario, it is assumed that only power from wind is uncertain. • Scenario 2) In this scenario, uncertainty of only grid electricity prices is considered. • Scenario 3) In this scenario, uncertainty of both grid electricity prices and power from WT is considered. In the above scenarios, scenario 1 and scenario 2 are based on single objective optimization whereas scenario 3 is based on multi-objective optimization approach. Moreover, each scenario is evaluated for 0%, 10% and 20% participation levels of DR. The SDS operation cost amounts to 3969.55 $, 3696.79 $ and 3431.56 $ respectively when participation level of DR is 0%, 10% and 20%. These prices are obtained under deterministic case without taking into account the effect of uncertainty. Results of Scenario 1. The variation of opportunity radius of uncertainty in wind generation αRS with critical cost is depicted in Fig. 1. This figure shows variation of minimum deviation in wind power as compared to forecasted value for achieving minimum operation cost that is less than certain critical cost. It is observed that operation cost of SDS reduces due to favorable wind power deviations. For instance, the operation cost reduces from 3431.56$ to 2894.50$ when opportuneness value increases from 0 to 0.578 for 20% participation rate of DR. Thus, the risk seeking strategy provided by opportuneness function of IGDT assists in enduring financial benefit through possible increase in wind generation. It is also observed that the increase in participation level of DR gives lower radius of wind power uncertainty. For more clarity, the opportuneness value with critical cost of 3100$ under different DR levels is shown in Table 1. As compared with 0%

Fig. 1. Opportuneness value αRS versus target cost

Risk-Seeker IGDT

421

Table 1. Opportuneness value αRS for varying DR levels DR levels

0%

10%

20%

αRS (OC c = 3100$) 0.949 0.641 0.355

Fig. 2. Ratio value versus opportuneness value αRS

and 10% DR participation rate, the opportuneness value decreases by 62.60% and 44.61% when DR participation is 20%. This implies that certain critical cost can be obtained with less wind power when DR participation is high. Figure 2 shows the ratio value of power from different sources and energy loss for 20% DR participation rate. Ratio value is ratio of energy in presence of uncertainty to without uncertainty. It is observed that the share of power injected from wind is increased as its uncertainty is now converted into opportunity for decision maker. The power input from sub-station continues to decrease on increasing opportuneness value due to large positive forecast errors in wind power. Moreover, there is no significant change in power from DDGs and energy loss. Results of Scenario 2. The variation in opportuneness value β RS with target cost for different DR participation levels is presented in Fig. 3. The figure indicates that lower operating cost of SDS necessitates favorable deviations in grid electricity prices. For example - with DR participation rate of 20%, in order to reduce the operation cost of SDS by 15.65% (3431.56$ to 2894.50$) the uncertain grid prices in each hour should be 19.86% lower in comparison to forecasted grid electricity prices. In fact, this financial gain is attributed to the acceptance of risk linked with risk-taking strategies. Moreover, Table 2 shows that the performance of SDS becomes superior by integrating DR as the opportuneness value decreases with higher DR participation level. In other words, in order to obtain a critical cost of 3100$, 26.60% and 20.10% decrease in electricity prices is needed when DR participation rate is 0% and 10% respectively. On the other hand, only 12.5% reduction of prices is enough with 20% participation level of DR to obtain similar critical cost.

422

T. Rawat et al.

Fig. 3. Opportuneness value β RS verus target cost Table 2. Opportuneness value β RS for varying DR levels DR levels β

RS

0%

10%

20%

c

(OC = 3100$) 0.266 0.201 0.125

The ratio value of from multiple sources and energy loss for 20% DR participation level is presented in Fig. 4. It is found that in comparison to deterministic case (uncertainty neglected), more energy is preferred from grid due to favorable uncertain electricity prices. Moreover, with larger variations of prices the decision maker would raise power input from grid to take more advantage of uncertainties. Similarly, the share of DDGs reduces to avoid increase in operation cost since cheaper power from grid is available.

Fig. 4. Ratio value versus opportuneness value β RS

Risk-Seeker IGDT

423

Table 3. Pareto solutions for varying DR penetration rates DR penetration (%) = 0 DR penetration % = 10 DR penetration % = 20 

e

αRS

β RS

αRS

β RS

αRS

β RS

1

0

0.266

0

0.201

0

0.125

2

0.136 0.237

0.087 0.179

0.045 0.111

3

0.263 0.207

0.169 0.156

0.088 0.097

4

0.38

0.177

0.247 0.134

0.131 0.083

5

0.49

0.148

0.322 0.112

0.171 0.069

6

0.594 0.118

0.392 0.089

0.211 0.055

7

0.691 0.089

0.46

0.067

0.249 0.042

8

0.783 0.059

0.524 0.045

0.286 0.028

9

0.869 0.03

0.584 0.022

0.321 0.014

0.641 0

0.355 0

10 0.949 0

Results of Scenario 3. In this case, the uncertainties in both grid prices and wind power are modeled using RS-IGDT. The pareto-solutions (e ) obtained consist of 10 points and are as shown in Table 3 for different participation levels of DR. It is worth to note that these solutions are relevant only when the operator decides to take risk. These pareto-solutions are determined for a critical cost of 3100$. In addition, the decision maker can easily obtain other strategies/pareto-solutions on varying deviation factor. For DR participation rate of 20%, solution #5 is the most comprising solution as evaluated using fuzzy criteria. The comparative relation between energy of DDGs and grid sub-station corresponding to most compromising solution and deterministic case is depicted in Fig. 5 and Fig. 6 respectively. The decrease in power generation of DDGs is due to increase in wind power injection. Similarly, as compared with deterministic case power scheduled from grid with preferred solution is reduced because of large positive forecast errors in wind power having negligible operation cost.

Fig. 5. Energy from DDGs (RS-strategy)

424

T. Rawat et al.

Fig. 6. Power taken from sub-station (RS-strategy)

5

Conclusions

A multi-objective approach for optimizing radius of uncertainty of both substation electricity prices and power from wind is proposed in this paper. The uncertainty in both the input parameters is modeled using opportunity function of IGDT. SDS operation under different levels of load shifting type DR program is also investigated. The proposed model is tested on a modified 33-bus distribution system which is equipped with DDGs, WT and flexible loads. The results show that through opportunity function decision maker can make more economic profit through reduction in electricity prices and increase in wind power generation. Also, the opportuneness value increases on decreasing the operating cost. In addition, it is observed that economic performance of SDS is improved on increasing DR participation rate.

References 1. Abapour, S., Zare, K., Mohammadi-Ivatloo, B.: Evaluation of technical risks in distribution network along with distributed generation based on active management. IET Gener. Transm. Distrib. 8(4), 609–618 (2013) 2. Baharvandi, A., Aghaei, J., Nikoobakht, A., Niknam, T., Vahidinasab, V., Giaouris, D., Taylor, P.: Linearized hybrid stochastic/robust scheduling of active distribution networks encompassing PVs. IEEE Trans. Smart Grid 11(1), 357–367 (2019) 3. Dai, X., Wang, Y., Yang, S., Zhang, K.: IGDT-based economic dispatch considering the uncertainty of wind and demand response. IET Renew. Power Gener. 13(6), 856–866 (2018) 4. Khajehvand, M., Fakharian, A., Sedighizadeh, M.: A risk-averse decision based on IGDT/stochastic approach for smart distribution network operation under extreme uncertainties. Appl. Soft Comput. 107, 107395 (2021)

Risk-Seeker IGDT

425

5. Mavrotas, G.: Effective implementation of the ε-constraint method in multiobjective mathematical programming problems. Appl. Math. Comput. 213(2), 455–465 (2009) 6. Mehdizadeh, A., Taghizadegan, N., Salehi, J.: Risk-based energy management of renewable-based microgrid using information gap decision theory in the presence of peak load management. Appl. Energy 211, 617–630 (2018) 7. Murphy, C., Soroudi, A., Keane, A.: Information gap decision theory-based congestion and voltage management in the presence of uncertain wind power. IEEE Trans. Sustain. Energy 7(2), 841–849 (2015) 8. Nasr, M.A., Nasr-Azadani, E., Rabiee, A., Hosseinian, S.H.: Risk-averse energy management system for isolated microgrids considering generation and demand uncertainties based on information gap decision theory. IET Renew. Power Gene. 13(6), 940–951 (2019) 9. O’Connell, A., Soroudi, A., Keane, A.: Distribution network operation under uncertainty using information gap decision theory. IEEE Trans. Smart Grid 9(3), 1848– 1858 (2016) 10. Rawat, T., Niazi, K.R., Gupta, N., Sharma, S.: Impact assessment of electric vehicle charging/discharging strategies on the operation management of grid accessible and remote microgrids. Int. J. Energy Res. 43(15), 9034–9048 (2019) 11. Rawat, T., Niazi, K.R.: Risk averse energy management for grid connected microgrid using information gap decision theory. In: Kalam, A., Niazi, K.R., Soni, A., Siddiqui, S.A., Mundra, A. (eds.) Intelligent Computing Techniques for Smart Energy Systems. LNEE, vol. 607, pp. 465–473. Springer, Singapore (2020). https:// doi.org/10.1007/978-981-15-0214-9 50 12. Rawat, T., Niazi, K., Gupta, N., Sharma, S.: Multi-objective information gap decision theory based operation of smart distribution grid integrated with demand response. In: 2020 21st National Power Systems Conference (NPSC), pp. 1–6. IEEE (2020) 13. Samimi, A., Rezaei, N.: Robust optimal energy and reactive power management in smart distribution networks: an info-gap multi-objective approach. Int. Trans. Electr. Energy Syst. 29(11), e12115 (2019) 14. Sriyakul, T., Jermsittiparsert, K.: Economic scheduling of a smart microgrid utilizing the benefits of plug-in electric vehicles contracts with a comprehensive model of information-gap decision theory. J. Energy Stor. 32, 102010 (2020) 15. Zhang, C., Xu, Y., Dong, Z.Y., Ma, J.: Robust operation of microgrids via twostage coordinated energy storage and direct load control. IEEE Trans. Power Syst. 32(4), 2858–2868 (2016) 16. Zhao, J., Wan, C., Xu, Z., Wang, J.: Risk-based day-ahead scheduling of electric vehicle aggregator using information gap decision theory. IEEE Trans. Smart Grid 8(4), 1609–1618 (2015)

Performance Analysis of Multi-user Diversity Schemes on Interference Limited D-GG Atmospheric Turbulence Channels Anu Goel1(B) and Richa Bhatia2 1

2

ECE Department, NSUT EAST CAMPUS (formerly AIACTR), Affiliated to GGSIPU, Delhi, India [email protected] ECE Department, NSUT EAST CAMPUS (formerly AIACTR), Delhi, India

Abstract. With asymmetric channel fading, this research presents the unified analysis of a multiuser radio frequency/free space optics (RF/FSO) amplify-and-forward (AF) relay network. It has been assumed that relay and destination nodes operate in the presence of co-channel interferers (CCIs). The first hop of the AF strategy is in RF domain, which are used for information transmission between users and relay nodes, and an FSO link is utilized for sending information from the relay to destination node. The RF channels have been assumed to undergo the Nakagami-m fading model, whereas the turbulence on the wireless optical channel is expected to follow the Double Generalized Gamma (DGG) fading model, which takes pointing errors into account. Analytical expression for the outage probability has been presented for the aforementioned system model. Monte Carlo simulations are used to verify all of the theoretical results.

Keywords: Mixed RF/FSO

1

· D-GG channel · Co-channel interferers

Introduction

Researchers have paid close attention to the huge unlicensed bandwidth available in free space optical (FSO) communication systems. The atmospheric turbulence and pointing error limit the capabilities of FSO systems, such as high security, quick deployment, and flexibility [1,2]. In addition, mixed RF/FSO relay systems have arisen as a result of the use of FSO systems as an complementary form of radio frequency (RF) equivalents [3–6]. Because of its simplicity, the amplifyand-forward (AF) relaying approach has gained a lot of traction. Multiuser diversity (MUD) techniques have been developed in [7] to boost the data rate supplied by wireless networks across RF fading channels. RF systems can employ the diversity gain provided by MUD systems from several users transmitting at the same time to boost throughput. MUD can take advantage c The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022  S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 426–436, 2022. https://doi.org/10.1007/978-981-19-1742-4_36

Performance Analysis of Multi-user Diversity Schemes

427

of fluctuating RF channels by assigning common resources to the user with the best channel. The user with the highest signal-to-noise ratio (SNR) is chosen for transmission in this approach. Furthermore, interference from other modes of transmission in the immediate neighborhood limits the performance of any realistic wireless communication system. Importantly, relay networks with various co-channel interferences (CCIs) at the receiving nodes impair the desired signal quality. In the literature, there has been a lot of work on mixed RF/FSO relaying systems [3,5,7–9]. The dual hop mixed RF/FSO asymmetric relaying was proposed in [3], where the authors developed an estimate of the entire system’s outage probability (OP). The Rayleigh distribution was used to represent the RF link, whereas Gamma-Gamma atmospheric turbulence was used to model the FSO link. Using pointing faults on the FSO link subject to intensity modulation with direct detection (IM/DD) and heterodyne detection, the authors of [5] completed a unified study of mixed RF/FSO relaying system. Research works [7–9] examined the performance of a dual hop RF/FSO connection using recently proposed generalized M-distribution and Double Generalized Gamma (D-GG) fading models. While the majority of the published results for FSO relay-assisted communications have depended on the absence of CCI, it becomes crucial to emphasize that because RF links are involved, asymmetric RF/FSO relay systems can be intrinsically subject to the impact of CCI. In RF systems, frequency re-use also creates a source of interference, according to [10]. The authors of [11] were inspired by the interference-limited performance of mixed RF/FSO systems to construct the expression of OP and bit error rate (BER) for an AF relaying system. To illustrate accurate and asymptotic performance of the whole system in [11], the channel of interfering signals was modeled as the Nakagami-m distribution fading, while the FSO systems was studied using the D-GG turbulence model. The authors in [12] have derived the closed form formulation of OP by observing the effect of interference at the relay node. In addition, [13] describes an examination of a dual-hop FSO/RF cooperative system in which the FSO and RF links undergo Malaga and generalized-K distributed fadings, respectively. The majority of past research has focused on mixed RF/FSO relay systems with the premise of interference-free transmission. Furthermore, works that address the issue of interference in FSO systems have not addressed the benefits that the MUD scheme can provide. We intend to study asymmetric relaying with generalized fading models in the presence of MUD, which is missing in the literature. To increase the performance of the RF link, we consider an asymmetric mixed RF/FSO fixed gain relay method with sub-carrier intensity modulation (SIM) and MUD at the source in this study. In addition, we evaluate the influence of several CCIs on the communication to make the system model more feasible. Therefore, the system model presented in this paper is different from that given in [3,5,7–9,11–16]. In addition to this, we model the RF link with Nakagami-m fading model [17]. With interference at relay and destination nodes, we study the system subjected highly adverse conditions. Unequal interferers with non-identical statistical characteristics have been considered on both

428

A. Goel and R. Bhatia

the links to observe the effect of simultaneous transmission from different kinds communication systems. D-GG fading distribution is used to model atmospheric turbulence on the FSO link.

2

System and Channel Models

In this section, the system and channel models of the considered relaying strategy is discussed. Table 1. List of symbols. Symbol Meaning

2.1

PSR,j

Power of signal of j th -user on S − R link

hSR,j

Fading coefficient of j th -user on S − R link

dRF

Symbol of j th -user from source node

PI1

Power of interferer at relay node

xr,i

Symbol from interferer at relay node

ρ

Type of optical demodulation

NRF

AWGN on S − R link

η

Optical to electrical conversion factor

I

Irradiance on optical link

hD,i

Power of interferer at destination node

xd,i

Symbol from interferer at relay node

No

AWGN on R − D link

System Model

Consider an asymmetric RF/FSO dual hop cooperative relaying system with K RF mobile users intend to send data to the destination node D through the relay node R. The randomness offered by the RF channel increases as the number of users on the RF link increases, which is utilized by the node R for increasing system throughput. Moreover, the K users communicate with the relay node in the vicinity of L CCIs at the relay node. On source-to-relay (SR) link, the signal received ySR,j from j th -user at node R is given by

ySR,j =



PSR,j hSR,j dRF +



PI1

L 

hSR,i xr,i + NRF

(1)

i=1

where suffix j corresponds to the j th -user and all the symbols are defined in Table 1. In the second phase, the relay node amplifies the signal ySR,j and transmits the same towards the destination node D. In the presence of M CCIs at

Performance Analysis of Multi-user Diversity Schemes

429

the destination node, the received signal is expressed as L    ρ ρ hR,i xr,i yj,RD =(ηI) 2 G PSR,j hSR,j dRF + (ηI) 2 G PI1 i=1 ρ

+ (ηI) 2



PI2

M 

hD,i xd,i + No

(2)

i=1

The various symbols in (2) have been defined in Table 1. As mentioned in [10], the gain G can be formulated as G2 =

1 2

PSR,j |hSR,j | + PI1

L i=1

(3)

2

|hR,i | + σn 2

For the cooperative relay strategy, the signal-to-interference-plus-noise (SINR) can be written using (2) as [10] γe2e =

ef f ef f γSR,j γRD

(4)

ef f ef f γSR,j + γRD +1

where the effective instantaneous SINR on the RF link corresponding to the j th user considering both RF fading as well as the interference statistics is ef f ef f and is given by the relation γSR,j = 1+Lγk γ . On the other denoted by γSR,j i=1

Ir ,i

ef f hand, γRD denotes the effective SINR on the wireless optical link considering atmospheric scintillation, pointing errors, path loss and interference defined as ef f RD = 1+γM . The variables γIr and γId denote interfering signal at relay γRD i γId ,i and destination nodes, respectively. For mathematical tractability, consider an upper bound, γe2e < γup , such that   ef f ef f (5) , γRD γe2e = min γSR,j

2.2

Channel Model

2.2.1 RF Link The k th user on the RF link follows the probability density function (PDF) of Nakagami-m distribution as given below [17] fγk (γ) =

β mSR mSR −1 γ exp(−βγ) Γ (mSR )

(6)

. Furthermore, using the notion where mSR is the shape parameter and β = mγ¯SR k of ordered statistics as follows, the PDF with K users may be produced as fγSR ,j (γ) = K[Fγk (γ)]K−1 fγk (γ)

(7)

430

A. Goel and R. Bhatia

Hence, considering (6) and (7), for integer values of fading parameter mSR , the closed-form expression for the PDF of RF link experiencing Nakagami-m fading can be obtained as [6]   K−1 Kβ mSR  n1 K − 1 (−1) fγSR ,j (γ) = n1 Γ (mSR ) n =0 1

n1 (mSR −1)

×



ζn1 n2 (mSR )γ mSR +n2 −1 β n2 exp(−β(n1 + 1)γ)

(8)

n2 =0

where ζn1 n2 (mSR ) is the multinomial expansion’s coefficient involved in the derivation of (8) which can be recursively obtained using the relation ζn1 n2 = n1 λbn2 −1 b=n1 −x+1 n1 −b I[0,(n2 −1)(x−1)] where I[a,b] is defined in [17, Eq. (9.120)]. MoreL γI ,i can over, the PDF of the total interference-to-noise ratio (INR)   i=1 r

m1 L γ m1 L−1 1L − m where m1 is be expressed as [6] fIr (γ) = ΩmI11 Γ (m1 L) exp ΩI 1 γ Nakagami-m fading parameter and ΩI 1 is the interference to noise ratio (INR) on the RF link. The effective PDF can be simplified as [6] fγefSRf ,j (γ) =

K−1 SR −1)  n1 (m n1 =0

A2 γ mSR +n2 −1 e[−β(n1 +1)γ]

n2 =0

 × U m1 L, m1 L + 1 + mSR + n2 , B0 (γ)

(9)

where U (a, b, z) is the confluent hypergeometric function of second kind as defined in [18, Eq. (9.211.4)], B0 (γ) = β(n1 + 1)γ + ΩmI11 , the constant A1 =  m 1 L  Kβ mSR m1 n2 , A2 = A1 (−1)n1 K−1 n1 Γ (m1 L)ζn1 n2 (mSR )β . The Γ (m1 L)Γ (mSR ) ΩI 1

cumulative distribution function (CDF) on the RF link can be derived as ∞ FγefSRf (γ) = 0 fγefSRf (γ)dγ. After some mathematical manipulations, the CDF ef f of γSR,j can be expressed as follows [6] n2 +m K−1 SR −1) SR −1  n1 (m  K A3 1 − e(−β(n1 +1)γ) Γ mSR n =0 n =0 m=0 1 2  

{(n1 + 1)βγ}m U m1 L, m1 L + 1 + m, B0 (γ) × m!

FγefSRf ,j (γ) =

where A3 is given by A3 = (−1)n1



(10)

 K −1 (mSR + n2 − 1)!ζn1 n2 (mSR ) n1

× (n1 + 1)−n2 −mSR

(11)

On the other hand, it is assumed that R→D link undergoes the turbulenceinduced fading Ia and the pointing errors Ip such that I = Ia Ip . The pointing

Performance Analysis of Multi-user Diversity Schemes

431

error parameters are obtained from the reference [1]. The atmospheric turbulence fading Ia on the FSO link is modeled as double generalized gamma (GG) fading model [9] and the PDF of Ia can be expressed as [9] 1

1

yσ β1 − 2 λβ2 − 2 (2π)1− f (Ia ) = Γ (β1 )Γ (β2 )Ia

σ+λ 2

G0,λ+σ λ+σ,0

   Ω λ λ λ σ σ Ω  2 1  Iaα2 β1σ β2λ 

1−τ0

 (12)

x x+1 x+z−1 where τ0 = [Δ(σ   : β1 ), Δ(λ : β2 )] with Δ(z : x) representing [ z , z , . . . , z ],  . is the Meijer-G function defined in [18, Eq. (9.301)] and y = where Gm,n p,q α2 λ. Moreover, the SNR PDF on FSO link experiencing a double GG fading model with the effect of pointing errors is given by [9]   yr   γ D1 λ+σ+1,0   τ2 D2 z y (13) G fRD (γ) =  τ1  rγ 1,λ+σ+1 μr

with

1

1

ξ 2 σ β1 − 2 σ β2 − 2 (2π)1− D1 = Γ (β1 )Γ (β2 ) D2 = 

ξ2 y , Δ(σ

β1σ β2λ λ λ σ σ Ω1σ Ω2λ

  : β1 ), Δ(λ : β2 ) , τ2 = 1 +

σ+λ 2

(14)

(15) ξ2 y



r

, μr = (ηE(I)) and r denotes N0  σ+λ  1 the kind of demodulation scheme [9]. Considering D3 = g=1 Γ y + τ0,g3.5pt ,

where τ1 =

1/y

D3 2) z can be given as z = D1 (D . (1+ξ 2 ) Furthermore, the following is the unified CDF of the FSO channel over double GG air turbulence with pointing errors [9]   γ y   n,1  1,τ3 (16) FRD (γ) = D4 Gr+1,n+1 D5 μr  τ4 ,0 r(σ+λ) r  β −1 1− 2 β1 − 1 2 σ 2 2 (2π) 2 r (β1 +β2 −2) D2 z y where n = r(λ + σ + 1), D4 = ξ σ , D = , 5 σ+λ yΓ (β1 )Γ (β2 ) r τ3 = [Δ(r : τ2 )] and τ4 = [Δ(r : τ1 )] comprising of r(λ + σ + 1) terms. Moreover, the interference on the destination is assumed to follow the PDF [19]   γ γ M −1 exp − fId (γ) = (17) ΩI 2 (M − 1)! ΩI 2

where ΩI 2 is the INR on the FSO link. Given that Id and γRD are independent, ef f can be derived using the CDF of γRD  ∞ f (γ) = FRD (γ(1 + x))fId (x)dx (18) FγefRD 0

432

A. Goel and R. Bhatia

Substituting (16) and (17) in (18), and assuming 1 + y = z, above integral may be formulated as f FγefRD (γ) = D6 (J1 − J2 )

where D6 =

D4 ΩI 2 M (M −1)!

(19)

exp(1/ΩI 2 ) and

  zγ y    1,τ3 D5 (20) J1 = (z − 1) e τ4 ,0  μr 0 M −1 M −1 Resorting to binomial expansion (z − 1)M −1 = (−1)M −1−i z i as i=0 i per [18, Eq. (1.11)] for integer values of M and with the aid of [20, Eq. (07.34.21.0013.01)], integral J1 can be given as 

∞

J1 =

z M −1 − ΩI 2

M −1 

Gn,1 r+1,n+1

  y D xi D10 Gn,y+1 (γ)  7 r+y+1,n+1

1,τ3 ,τ5 τ4 ,0

i=0

where D7 = D5 (yΩI 2 /μr )y , D10 =



1

y i+ 2 (ΩI 2 )i+1 y−1 (2π) 2

M −1



and τ5 =

(21) (−i) (v−1−i) y ,..., y

 .

On the other hand, assuming xi = (−1)M −1−i , the integral J2 can be i given as  1 M −1   zγ y    − Ωz n,1 i  1,τ3 2 I (22) xi (z) e Gr+1,n+1 D5 J2 = τ4 ,0  μr 0 i=0 The integral J2 can be expressed in complex integral form on using [20, Eq. (07.34.02.0001.01)]. By virtue of [18, Eq. (8.380)] and [21, Eq. (28-30)] and with some mathematical manipulations, the integral J2 shown in (23) where

J2 =

M −1  i=0

=

M −1  i=0

0,1:1,0:n,1 xi H1,1:0,1:r+1,n+1



    y   (1, 1), (τ3 , [1]τ3 )  1 zγ (−i : 1, v)    , D 5 (−i − 1 : 1, v)  (0, 1)  (τ4 , [1]τ4 ), (o, 1)  ΩI 2 μr

  y  γ xi H ΩI 2 , μr

(23)



1 :m2 ,n2 :m3 ,n3 [1]x represents sequence of 1 s of length x and H0,m p1 ,q1 :p2 ,q2 :p3 ,q3 [A1 , A2 ] = H [A1 , A2 ] is the bivariate Fox-H function given in [22, Eq. (28)]. Substituting (21) and (23) into (19), we obtain the closed form equation of CDF on FSO link in the presence of multiple interferers as given below

f FγefRD (γ) = D6

M −1 



i=0

− H ΩI 2 ,

  y xi D10 Gn,y+1 r+y+1,n+1 D7 (γ)  

γ μr

1,τ3 ,τ5 τ4 ,0



y

(24)

Performance Analysis of Multi-user Diversity Schemes

433

Moreover, the PDF of end-to-end SNR can be derived by utilizing the relationship (18). Using the identity [20, Eq. (07.34.21.0013.01)] and [22], the closed form expression for the overall PDF can be written as shown below  M −1      r(λ+σ+1),y ef f xi D12 Gr+y,r(λ+σ+1) D8 γ y  ττ67 fγRD (γ) = D11 

i=0

− H ΩI 2 , D 2 z y



γ μr

y/r   (25)

In (25), the parameters involved in Meijer-G function has  been y−(i+2) τ2 τ2 +r−1 1−(i+2) and τ7 = τr1 , . . . , derived as τ6 = , ,..., r ,..., r y y  τ1 +r−1 . The argument of the Meijer-G equals , . . . , τ1 (λ+σ+1)+r−1 r r  y D1 exp( Ω1 ) I2 I2 D8 = D2 z y {r}(−r(λ+σ)) Ω , D11 = rλΩI 2 M (M and D12 = yμr −1)! 3

y i+ 2 (ΩI 2 )i+2 (r)c∗

(2π){

( (r−1)(λ+σ) )} 2

y−1+

. The bivariate Fox’s H-function is defined in (26).

    y/r    γ (−i − 1 : 1, y/r)   (τ2 , [1]τ2 )  1 y ,D z    (−i − 2 : 1, y/r)  (0, 1)  (τ1 , [1]τ1 )  ΩI 2 2 μr   y/r γ y = H ΩI 2 , D2 z (26) μr 

H0,1:1,0:λ+σ+1,1 1,1:0,1:1,λ+σ+1

3

Performance Analysis

In this section, we investigate various performance metrics of relay system based on the aforementioned SNR bound. 3.1

Exact Outage Probability

The outage probability (OP) of a wireless communication system is an important performance measure. The likelihood that the instantaneous SNR goes below a pre-specified threshold γth is defined as OP. The end-to-end CDF Fe2 e (γ) of the overall system can be written as f f Fe2 e (γ) = FγefSRf ,j (γ) + FγefRD (γ) − FγefSRf ,j (γ)FγefRD (γ)

(27)

Hence, substituting (10) and (24) into (27), and by putting γ = γth we can obtain the overall CDF of the considered system.

434

A. Goel and R. Bhatia

Fig. 1. Outage Probability of dual hop mixed RF/FSO relaying system with varying number of CCIs.

Fig. 2. Outage Probability of dual hop mixed RF/FSO relaying system for varying FSO link parameters.

4

Results

The results of the research study have been discussed in this section. The outage performance of the mixed RF/FSO relaying system is shown in Fig. 1. The outage probability has been plotted against the average SNR per hop in this graph. The comparison was made for a variety of RF users, as well as the number of interferers at the relay and destination nodes. The graphic has actually showed the fluctuating strength of air turbulence. It can be shown that as the number of CCIs grows, so does the outage. The asymptotic representation of the outage probability is also shown in the plot. Monte-Carlo simulations were used to verify the results. The effect of pointing inaccuracy and optical demodulation type is displayed in Fig. 2. It can be seen that the outage probability rises as the pointing inaccuracy rises (lower value of xi). The chance of an outage is also influenced by

Performance Analysis of Multi-user Diversity Schemes

435

optical demodulation. The plot shows that when coherent detection is used r = 1, the system’s outage probability is much higher than when coherent detection is used r = 2. Also shown are the asymptotic and Monte-Carlo results.

5

Conclusion

The analysis of a mixed RF/FSO relaying system in the presence of interferers on the relay and destination nodes is presented in this paper. The fading of the RF link was modeled using the Nakagami-m distribution, whereas the turbulence of the FSO link was studied using the D-GG distribution. We offer the analysis for the OP of the overall system for the considered system model.

References 1. Farid, A.A., Hranilovic, S.: Outage capacity optimization for free-space optical links with pointing errors. J. Lightwave Technol. 25(7), 1702–1710 (2007) 2. Goel, A., Bhatia, R.: Hybrid RF/MIMO-FSO relaying systems over gamma-gamma fading channels. In: Gupta, D., Khanna, A., Bhattacharyya, S., Hassanien, A.E., Anand, S., Jaiswal, A. (eds.) International Conference on Innovative Computing and Communications. AISC, vol. 1165, pp. 607–615. Springer, Singapore (2021). https://doi.org/10.1007/978-981-15-5113-0 49 3. Lee, E., Park, J., Han, D., Yoon, G.: Performance analysis of the asymmetric dualhop relay transmission with mixed RF/FSO links. IEEE Photonics Technol. Lett. 23(21), 1642–1644 (2011) 4. Goel, A., Bhatia, R.: Joint impact of interference and hardware impairments on the performance of mixed RF/FSO cooperative relay networks. Opt. Quant. Electron. 53(9), 1–15 (2021). https://doi.org/10.1007/s11082-021-03064-x 5. Zedini, E., Ansari, I.S., Alouini, M.S.: Performance analysis of mixed Nakagami-m and Gamma Gamma dual-hop FSO transmission systems. IEEE Photonics J. 7(1), 1–20 (2015) 6. Goel, A., Bhatia, R.: On the performance of mixed user diversity-RF/spatial diversity FSO cooperative relaying AF systems. Opt. Commun. 8 (2020) 7. Yang, L., Hasna, M.O., Ansari, I.S.: Unified performance analysis for multiuser mixed η - μ and M - distribution dual-hop RF/FSO systems. IEEE Trans. Commun. 65(8), 3601–3613 (2017) 8. Wang, P., Wang, R., Guo, L., Cao, T., Yang, Y.: On the performances of relayaided FSO system over M distribution with pointing errors in presence of various weather conditions. Opt. Commun. 367, 59–67 (2016) 9. AlQuwaiee, H., Ansari, I.S., Alouini, M.: On the performance of free-space optical communication systems over Double Generalized Gamma channel. IEEE J. Sel. Areas Commun. 33(9), 1829–1840 (2015) 10. Ikki, S.S., Aissa, S.: Performance evaluation and optimization of dual-hop communication over nakagami-m fading channels in the presence of co-channel interferences. IEEE Commun. Lett. 16(8), 1149–1152 (2012) 11. Soleimani-Nasab, E., Uysal, M.: Generalized performance analysis of mixed RF/FSO cooperative systems. IEEE Trans. Wireless Commun. 15(1), 714–727 (2016)

436

A. Goel and R. Bhatia

12. Petkovic, M.I., Cvetkovic, A.M., Djordjevic, G.T., Karagiannidis, G.K.: Outage performance of the mixed RF/FSO relaying channel in the presence of interference. Wireless Pers. Commun. 96, 2999–3014 (2017). https://doi.org/10.1007/s11277017-4336-7 13. Trigui, I., Cherif, N., Affes, S., Wang, X., Leung, V., Stephenne, A.: Interferencelimited mixed m´ alaga-m and generalized-k dual-hop FSO/RF systems. In: 2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), pp. 1–6, October 2017 14. Balti, E., Guizani, M.: Mixed RF/FSO cooperative relaying systems with cochannel interference. IEEE Trans. Commun. 66(9), 4014–4027 (2018) 15. Abouei, J., Plataniotis, K.N.: Multiuser diversity scheduling in free-space optical communications. J. Lightwave Technol. 30(9), 1351–1358 (2012) 16. Chen, L., Wang, W., Zhang, C.: Multiuser diversity over parallel and hybrid FSO/RF links and its performance analysis. IEEE Photonics J. 8(3), 1–9 (2016) 17. Simon, M.K., Alouini, M.: Digital Communication over Fading Channels. Wiley, Hoboken (2000) 18. Gradshteyn, I.S., Ryzhik, I.M.: Table of Integrals, Series and Products. Academic, New York (2000) 19. Hasna, M.O., Alouini, M.S., Bastami, A., Ebbini, E.S.: Performance analysis of cellular mobile systems with successive co-channel interference cancellation. IEEE Trans. Wireless Commun. 2(1), 29–40 (2004) 20. Wolfram, I.: Wolfram, research, mathematica edition: Version 10.0. champaign. Wolfram Research Inc. (2010) 21. Mittal, P., Gupta, K.: An integral involving generalized function of two variables. Proc. Indian Acad. Sci. Section A 75(3), 117–123 (1972). https://doi.org/10.1007/ BF03049732 22. Alhennawi, H.R., Ayadi, M.M.H.E., Ismail, M.H., Mourad, H.A.M.: Closed-form exact and asymptotic expressions for the symbol error rate and capacity of the H-function fading channel. IEEE Trans. Veh. Technol. 65(4), 1957–1974 (2016)

A Sparse-Dense HOG Window Sampling Technique for Fast Pedestrian Detection in Aerial Images Ranjeet Kumar(B) and Alok Kanti Deb Indian Institute of Technology, Kharagpur, India [email protected], [email protected]

Abstract. Pedestrian detection from Unmanned Aerial Vehicle (UAV) has been an important part of surveillance systems. A Two-stage (Sparse-Dense) sliding window technique has been proposed to increase the speed of pedestrian detection using HOG-SVM classifier. Standard techniques follow a sliding window approach with a fixed sliding strides over a multi-resolution image pyramid for detection. The presented technique breaks down the detection task into sparse sampling and dense sampling stages where the first one is region proposal step and second stage scans only the proposed regions for objects. Sparse sampling stage is working as weak classifier whereas the dense sampling stage works as strong classifier for an image patch. Average pedestrian detection speed using the proposed technique gave improvement from 1.95 fps to 15.36 fps for input images of dimension [640, 360] on a system with 3.2 GHz CPU. UAV123 [1] dataset has been chosen to train the classifier. For detection, Average Center Prediction Error has been taken to quantify detection performance with increased speed. Keywords: Sparse-dense sampling detector · UAV123 · Region proposal · Sparse sampling · Dense sampling · Pedestrian detection · HOG · SVM

1 Introduction Fast pedestrian detection on aerial images has been a challenge due to dynamic nature of the images and hardware constraints. Integral Image for fast feature calculation [2] and Histogram of Oriented Gradients (HOG) [3], classification by Support Vector Machine (SVM) based tree-type neural network [4] are some initial work. Some improved methods for feature extraction are Integral channel feature [5], Boosted HOG [6], Channel Feature Extrapolation [7] and Search Region Proposal based on Saliency Map [8]. Pedestrian detection in infrared images has been shown by Zhang et al. [9]. Some techniques exploiting input image properties are Image Orientation Adjustment by Xu et al. [10] and Locally constraint linear coding based detection by Yang et al. [11]. Some techniques trying to enhance detection speed by breaking the task into multiple stages include local binary pattern with HOG-SVM classifier [12], simplified HOG [13], Center Symmetric - Local Binary Pattern (XCS-LBP) [14], Bin-Interleaved HOG [15] and two-stage linear with non-linear SVM [16] but they need improvement for real-time application. Some © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 437–450, 2022. https://doi.org/10.1007/978-981-19-1742-4_37

438

R. Kumar and A. K. Deb

hardware solutions for fast HOG-SVM based classification have been presented in [17, 18] that proposed hardware design suitable for HOG-SVM classification. This work proposes a Two-stage (Sparse-Dense) sliding window technique for pedestrian detection task and is an improvement over standard single stage sliding window techniques used with HOG+SVM [3, 5] based classifiers. Re-searchers have proposed using features other than HOG too for better detection but the proposed work shows how detection process can be modified to speed it up using existing classification method HOG-SVM and achieve real-time or near real-time performance. Section 2 discusses about standard sliding window based pedestrian detection. Section 3 presents the proposed two-stage (sparse-dense sliding) window based detection technique. Section 4 discusses about experimental setup, results and analysis. Section 5 concludes the work presented.

2 HOG-SVM Classification Based Pedestrian Detection Histogram of Oriented Gradients Histogram of Oriented Gradients (HOG) was given by Dalal et al. [3] to extract visual information from an image patch using pixel gradients. The technique has been used widely for classification/detection [5–9]. One can refer to [3] for HOG feature vector calculation for an image. Parameters to calculate HOG feature descriptor for an image patch of dimension [M , N ] has been shown in Table 1 and its length Fl [3] using (1) is 3780. Table 1. HOG feature descriptor parameters Parameter

  Window size Wh , Wv   Cell size Ch , Cv   Block size Bh , Bv

Value [64, 128] pixels [8, 8] pixels [2, 2] cells

Gamma correction (γ)

0.5

Bin size (b)

9

 Fl =

   M N −1 ∗ − 1 ∗ b ∗ Bh ∗ Bv Ch Cv

(1)

HOG feature plots with corresponding RGB images has been shown in Fig. 1. SVM Classification Support Vector Machines (SVM) [19, 20] is a supervised learning based classification algorithm that creates an N-dimensional hyper-plane that divides m number of classes. Input feature vector (p dimensional) and output label for a sample i has been denoted by

A Sparse-Dense HOG Window Sampling Technique

439

hi ∈ Rp and yi ∈ {−1, 1}, where i = 1, 2, 3, ..., n, Person and Background classes have been represented by label 1 and −1 respectively. f (hi ) = wT ∗ φ(hi ) + b

(2)

Here, f (hi ) is distance of a sample from decision boundary and its sign indicate predicted class, w is weight vector, φ(hi ) is a function (kernel) of hi and b is the offset from decision boundary.

Fig. 1. Example images from UAV123 [1] dataset with their HOG feature

Detection in an Image While HOG-SVM classifier classify an image patch of dimension [Wh , Wv ], pedestrian detection on full image is done by extracting image patches of dimension [Wh , Wv ] in sliding window manner from multi-scale image pyramid.. Five level image pyramid formed by original and downscaled versions of an image has been shown in Fig. 2. An important factor that determines the speed of detection is frequency of classificaI and can be obtained using (3), tion step for an image that has been denoted by HOGcount where Wh and Wv are dimensions of HOG window, DSh and DSv are dense sliding strides for HOG window [Wh , Wv ] in horizontal and vertical direction respectively. [M , N ] is input image width and height in pixels respectively. M − Wh N − Wv ∗ (3) DSh DSv   N Down-scaled image shape can be given by shapedn = M where α = 1.5 and , l l α α l ∈ {0, 1, ..., (L − 1)} are downscaling factor and image pyramid level respectively. L = 5 is the number of levels in image pyramid. Classification step frequency for an P image pyramid (HOGcount ) can be obtained using (4). L−1  M − Wh N − Wv  P (4) HOGcount = ∗ l=0 α l ∗ DSh α l ∗ DSv I = HOGcount

P where, HOGcount denoted number of HOG-SVM classification step for an image pyramid P for the parameters given in Table 2 using (4) is 3692. with L levels. Value of HOGcount This work focuses on reducing the required number of classification steps for an image pyramid by introducing a two-stage sliding (sparse-dense sampling) window technique.

440

R. Kumar and A. K. Deb

Fig. 2. Multi-resolution image pyramid formed from downscaled versions of original image

Table 2. Input image shape, HOG window shape, sliding stride for standard detection technique [3] and image pyramid parameters Parameter

Value

Input image shape ([M , N ])

[640, 360]

HOG window shape ([W _h, W _v])

[64, 128]

Image pyramid levels (L)

5

Downscaling factor (α)

1.5

Dense sliding stride ([DS_h, DS_v])

[8, 8]

3 A Two-Stage Sliding Window Conventional classifier based detection techniques follow a dense sampling approach to classify patches from image pyramid into number of classes. [3, 5, 9]. Dense sampling window stride has been denoted by [DSh , DSv ] and is [8, 8]. Here, DSh and DSv are strides in horizontal and vertical direction respectively. The proposed two-stage sliding window technique divides the detection task into sparse and dense sampling stages. Block diagram has been shown in Fig. 3.

Fig. 3. Block diagram of Two-Stage sliding window pedestrian detection

A Sparse-Dense HOG Window Sampling Technique

441

Stage 1: Sparse Sampling In first stage, image patches of size [Wh , Wv ] are extracted from all the levels in the image pyramid at larger window sliding strides denoted by [SSh , SSv ] as compared to [DSh , DSv ] taken by most of the previous techniques [3, 5, 6, 10, 11]. Here, [SSh , SSv ] ∈ {[32, 64], [43, 90], [51, 102]}. HOG-SVM classifier output represents distance from SVM decision boundary and has been taken to determine confidence of classification. Distance threshold for this stage has been denoted by Thsparse and image patches exceeding Thsparse are recorded as regions for proposal to stage 2. The regions proposed in the stage are of dimension [Wh , Wv ] in their corresponding downscaled image from the pyramid. The shape of proposed regions has been transformed back to level 0 (to represent same region in original image) denoted by PRS0 and can be obtained using (5) where l ∈ {0, 1, ..., (L − 1)}. PRS0 = [Wh , Wv ] ∗ α l

(5)

Fig. 4. Sparse sampling for region proposal and dense sampling on proposed region

Window sliding Stride fir sparse sampling stage has been calculated by taking percentage overlap between consecutive sampling windows. Overlap percentage of 50%, 30% and 20% have been taken for experimentation. Window sliding Stride values [SSh , SSv ] in pixels can be obtained using (6) and are [32, 64], [43, 90] and [51, 102] respectively. HOG-SVM classification step frequency in sparse sampling stage has been S and can be obtained using (7). demoted by HOGcount     SSh , SSv = overlap% ∗ Wh , Wv (6) S HOGcount

L−1  M − Wh N − Wv  = ∗ l l=0 α l ∗SSh α ∗SSv

(7)

Stage 2: Dense Sampling In this stage, image regions proposed from first stage are searched for objects by HOGSVM classifier with dense sampling window strides [DSh , DSv ]. Image patches crossing a threshold Thdense are final detections. An example has been shown in Fig. 4. Only the

442

R. Kumar and A. K. Deb

Fig. 5. Two-stage sliding window pedestrian detector process flow

proposed regions from stage 1 are processed in this stage and not the whole image pyramid. This saves significant processing time. Flow-chart for the two-stage process has been shown in Fig. 5. I Classification step frequency for dense sampling stage has been denoted by HOGcount I (3) and depends upon region proposal. I in HOGcount represent an image region Here, S I is combined classification + HOGcount proposed by sparse sampling stage. HOGcount step frequency and has been determined by average value for 1000 pedestrian images from UAV123 [1] dataset. Suppressing Duplicate Detections Dense sampling stage yields multiple detections around the object as classifier output crosses Thdense . These duplicate detections have been suppressed by computing Intersection-over-Union (IoU) among the detection boxes using (8) where RB1 and RB2 denote area of box B1 and B2 respectively. If IoU (B1 , B2 ) crosses a threshold IoUth , then the box with lower confidence (f (hi )(2)) is discarded. An example can be seen in Fig. 6. IoU (B1 , B2 ) =

RB1 ∩ RB2 RB1 ∪ RB2

(8)

A Sparse-Dense HOG Window Sampling Technique

443

Fig. 6. (a) Sparse sampling stage output, (b) Dense sampling stage output, (c) IoU thresholded output

4 Experimental Results and Analysis Dataset Dataset used for training and testing of HOG-SVM classifier is UAV123 [1]. It has 11575 images in its person class, out of which 2620 images with pedestrians are randomly selected and a window of [64, 128] size has been cropped to form person and 2450 windows cropped for background class. Positive and negative class formation for training has been shown in Fig. 7.

Fig. 7. Person and background dataset creation form UAV123 [1] for training and testing of SVM classifier

SVM Training Dataset size is 5070 images (2620 person class and 2450 background class) with traintest split ratio 80:20. 5-fold cross validation scheme has been adopted to split dataset into 5 mutually exclusive parts. Training has been done on 4 parts combined and testing on the remaining part. 5 trials of training/testing has been done. Scikit-learn [21] python library has been used to train Support Vector Classifier (SVC) for binary classification of Person and Background classes. Classifier parameters have been shown in Table 3. Classification Performance Precision, Recall, F1-score and Accuracy have been taken as classification performance metrics [22]. Metrics for 5 trials for test data with mean values has been presented in Table 4.

444

R. Kumar and A. K. Deb Table 3. SVM parameters Parameter

Value

Kernel

Linear

Train/Test Image window size

128 × 64

Person class data size

2620

Background class data size

2450

Train: Test split

80:20

Iteration

5000

Table 4. Classification performance (5-fold cross validation, SD - Standard Deviation) Dataset

Class

Precision

Recall

F1-score

Accuracy

Set1

Person

0.9943

0.9943

0.9943

0.9941

Background

0.9939

0.9939

0.9939

0.9941

Set2

Person

1.0000

0.9356

0.9667

0.9666

Background

0.9351

1.0000

0.9665

0.9666

Set3

Person

1.0000

0.9261

0.9616

0.9617

Background

0.9263

1.0000

0.9617

0.9617

Set4

Person

1.0000

0.8220

0.9023

0.9077

Background

0.8390

1.0000

0.9125

0.9077

Set5 Mean ± SD

Person

0.9715

0.8386

0.9002

0.9314

Background

0.9127

0.9856

0.9478

0.9314

Person

0.9892 ± 0.019

0.9033 ± 0.064

0.9450 ± 0.038

0.9523 ± 0.03

Background

0.9214 ± 0.05

0.9959 ± 0.006

0.9565 ± 0.027

0.9523 ± 0.03

Detection Performance UAV123 [1] provides ground-truth bounding boxes for evaluation of detection result. The prediction bounding box dimension of HOG-SVM classifier is fixed to [64, 128] as the classifier is not designed for bounding box regression task. Centre Prediction Error (CPE) [23] has been taken as performance metric that measures difference between Si denote average CPE for an predicted object centre and ground-truth box centre. CEavg be obtained using (9) where N image sequence S i and can S i denote number of frames



in a sequence Si , xG , yG and xP , yP represent ground-truth box centre and predicted box centre respectively. Chosen dataset has 23 person class image sequences named Si from person1 to person23 captured from an UAV platform. Average of CPEavg for 23

A Sparse-Dense HOG Window Sampling Technique

sequences and shown in (10) where i = {1, 2, 3, ..., K} and K = 23. 

2

1 NSi G Si xt − xtP + ytG − ytP = CEavg t=1 NSi

445

(9)

1 K NS CEavgi i=1 K

(10)

CEavg × 100 CEavg (%) = √ M 2 + N2

(11)

CEavg =

CEavg as percentage of input image diagonal length can be given by (11) to present a metric invariant to input image dimension. CEavg (%) for person sequences in UAV123 [1] has been found out by running detector for all the 23 sequences and comes out to be 1.47%. Centre coordinate plot for X and Y direction for person1 has been shown in Fig. 8. It can clearly be seen from X and Y coordinate detection graphs that detector is following the ground-truth coordinates almost all the time. An example image with ground-truth box, detection box and a line joining their centers has been shown in Fig. 9.

Fig. 8. x and y coordinate detection vs ground-truth

Proposed Technique Evaluation In the proposed technique, Thsparse = 0.1 has been taken for sparse sampling, Thdense = 1.0 for dense sampling and IoUth = 0.5 for IoU thresholding step. Improvement in processing time has been shown in Table 5 for different percentage of overlap between consecutive sampling windows in sparse sampling stage. Testing has been done on a 3.2 GHz CPU machine. Significant reduction in processing time can be seen as originally HOG-SVM classification step frequency has decreased from 3692 to 469, 344 and 294 (Table 5). As shown in Fig. 10, the technique gives different region proposals for different sliding strides and detections are concentrated around object after dense sampling stage. It can be observed that maximum detections are there in single stage dense sampling (Fig. 10-b) method but most of them are redundant and will be removed after IoU thresholding. Sparse sampling with [32, 64] (Fig. 10-c)and [43, 90] (Fig. 10-f) strides

446

R. Kumar and A. K. Deb

Fig. 9. Ground-truth box (pink), detection (white) and line between centers (green)

Table 5. Percentage overlap vs processing time: 2 stage detector (input size: [640, 360]) % overlap

Slide stride

Classification steps per image

Speed (fps)

[84.4, 92.2] (standard/baseline dense sampling detection)

[8, 8]

3692

1.95

[50.0, 50.0]

[32, 64]

469

15.36

[30.0, 30.0]

[43, 90]

344

20.92

[20.0, 20.0]

[51, 102]

294

24.48

provide significant number of region proposals around object and multiple detections after dense sampling stage (Fig. 10-d and 10-g). IoU thresholding has been applied to remove duplicate detections. Comparison with single stage dense sampling technique has been shown in Table 5. The improvement in speed for input image of size [640, 360] with [SSh , SSv ] = [32, 64] is from 1.95 to 15.36 fps (improved by a factor of 7.88) and 1.95 to 24.48 fps for [SSh , SSv ] = [51, 102] (improved by a factor of 10.50). Detection speed on full images has been compared with existing techniques and presented in Tables 6, 7, 8, 9 and 10 for different input image dimension. It is evident from the comparison tables that the proposed technique performs better than similar techniques. Moreover, the detection quality of the technique has been quantified in terms of average center prediction error (CPE). It is a standard metric used to judge the distance of predicted bounding box to that of ground-truth. Average CPE CEavg (%) for UAV123 dataset (person class) is only 1.47% of image diagonal length. Also, it should be noted that using classification based techniques clubbed with window sliding always introduces a quantization error in detection (evident in Fig. 9) which is equal to 0.5 times HOG window sliding stride [DSh , DSv ]. So, a trade-off has to be maintained between detection speed and quality to choose a particular sliding stride.

A Sparse-Dense HOG Window Sampling Technique

447

Fig. 10. (a) Original image (b) Standard dense sampling detection output on image pyramid, (c) Sparse stage output - [32, 64], (d) Sparse sampling output - [43, 90], (e) Sparse sampling output [52, 102], (f) Dense sampling output for c, (g) Dense sampling output for d, (h) Dense sampling output for e Table 6. Comparison with existing work (input image size: [320, 240]) Author

Technique

Detection speed (fps)

Dalal and Triggs [3] (2005)

Original HOG

Son et al. [15] (2010)

Bi-HOG

2.12

Sheng et al. [13] (2012)

Simplified HOG

3.33

Proposed

Sparse-dense sampling

1.07

24.4

Table 7. Comparison with existing work (input image size: [640, 480]) Author

Technique

Detection speed (fps)

P Dollar et al. [5] (2009)

Integral channel feature

0.5

Min et al. [16] (2013)

Two-stage linear+non-linear SVM

3.33

Vasuki et al. [14] (2016)

XCS-LBP with HOG-linear-SVM

4.05

Proposed

Sparse-dense sampling

8.61

Table 8. Comparison with existing work (input image size: [720, 400]) Author

Technique

Cao et al. [6] (2011)

Boosting HOG

Proposed

Sparse-dense sampling

Detection speed (fps) 4.76 10.14

448

R. Kumar and A. K. Deb Table 9. Comparison with existing work (input image size: [1200, 400])

Author

Technique

Detection speed (fps)

Xu et al. [10] (2017)

Orientation adjustment

5.3

Proposed

Sparse-dense sampling

8.65

Table 10. Comparison with existing work (input image size: [640, 320]) Author

Technique

Detection speed (fps)

Yang et al. [11] (2019)

Locally constraint linear coding

9

Proposed

Sparse-dense sampling

16.8

This work has used [8, 8] as sliding stride in dense sampling stage and is a standard used by other researchers. The presented average CPE value includes this quantization error inherently along with the actual detection error.

5 Conclusion The work introduced a Two-stage (Sparse-Dense) sliding window sampling technique for fast pedestrian detector. The first stage was sparse sampling stage to extract relevant regions. In the second stage, the proposed regions were taken to run classifier with smaller strides and larger classification threshold. Thus, a modified version of simple HOG-SVM detector has been presented. Visual information in image was exploited for region proposal and most of the time was spent in processing proposed regions. The proposed technique can be utilized to run real-time detection on low-cost processor on UAV platform and thus eliminate dependency on external system. This eventually opens up scope of more applications using UAV systems. Acknowledgement. We kindly acknowledge IMPRINT I project, MHRD, Govt. of India for supporting with resources from the project “Decentralized target tracking using swarm of aerial robots”.

References 1. Mueller, M., Smith, N., Ghanem, B.: A benchmark and simulator for UAV tracking. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV 2016. LNCS, vol. 9905, pp. 445–461. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46448-0_27 2. Viola, P., Jones, M.: Rapid object detection using a boosted cascade of simple features. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol. 1 (2001)

A Sparse-Dense HOG Window Sampling Technique

449

3. Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol. 1, pp. 886–893 (2005) 4. Jayadeva, Deb, A., Chandra, S.: Binary classification by SVM based tree type neural networks. In: Proceedings of the 2002 International Joint Conference on Neural Networks, vol. 3, pp. 2773–2778 (2002) 5. Dollár, P., Tu, Z., Perona, P., Belongie, S.: Integral channel features. In: The British Machine Vision Conference, pp. 91.1–91.11 (2009) 6. Cao, X., Wu, C., Yan, P., Li, X.: Linear SVM classification using boosting hog features for vehicle detection in low-altitude airborne videos. In: IEEE Conference on Image Processing, pp. 2421–2424 (2011) 7. Dollar, P., Appel, R., Belongie, S., Perona, P.: Fast feature pyramids for object detection. IEEE Trans. Pattern Anal. Mach. Intell. 36(8), 1532–1545 (2014) 8. Blondel, P., Potelle, A., Pegard, C., Lozano, R.: Human detection in uncluttered environments: from ground to UAV view. In: 13th International Conference on Control Automation Robotics and Vision, pp. 76–81 (2014) 9. Zhang, L., Wu, B., Nevatia, R.: Pedestrian detection in infrared images based on local shape features. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1–8 (2007) 10. Xu, Y., Yu, G., Wu, X., Wang, Y., Ma, Y.: An enhanced Viola-Jones vehicle detection method from unmanned aerial vehicles imagery. IEEE Trans. Intell. Transp. Syst. 18(7), 1845–1856 (2017) 11. Yang, Z., Huang, Z., Yang, Y., Yang, F., Yin, Z.: Accurate specified-pedestrian tracking from unmanned aerial vehicles. In: International Conference on Communication Technology Proceedings, ICCT, October 2019, pp. 1256–1260 (2019) 12. Park, W.-J., Kim, D.-H., Suryanto, Lyuh, C.-G., Roh, T.M., Ko, S.-J.: Fast human detection using selective block-based HOG-LBP. In: 19th IEEE International Conference on Image Processing, pp. 601–604 (2012) 13. Sheng, Y., Jiefa, W., Lingling, Z.: A fast pedestrian detection method based on simplified HOG descriptor. Int. J. Digit. Content Technol. Appl. 6(4), 14 (2012) 14. Vasuki, P., Veluchamy, S.: Pedestrian detection for driver assistance systems. In: International Conference on Recent Trends in Information Technology, pp. 1–4 (2016) 15. Son, H., Lee, S., Choi, J., Min, K.: Efficient pedestrian detection by Bin-interleaved Histogram of Oriented Gradients. In: IEEE Region 10 Conference-TENCON, pp. 2322–2325 (2010) 16. Min, K., Son, H., Choe, Y., Kim, Y.-G.: Real-time pedestrian detection based on a hierarchical two-stage Support Vector Machine. In: 8th IEEE Conference on Industrial Electronics and Applications, pp. 114–119 (2013) 17. Yuan, X., Cai-nian, L., Xiao-liang, X., Mei, J., Jian-guo, Z.: A two-stage hog feature extraction processor embedded with SVM for pedestrian detection. In: IEEEInternational Conference on Image Processing, pp. 3452–3455 (2015) 18. Wang, M.-S., Zhang, Z.-R.: FPGA implementation of hog based multi-scale pedestrian detection. In: IEEE International Conference on Applied System Invention, pp. 1099–1102 (2018) 19. Hearst, M., Dumais, S., Osuna, E., Platt, J., Scholkopf, B.: Support vector machines. IEEE Intell. Syst. Appl. 13(4), 740–755 (1998) 20. Vapnik, V., Cortes, C.: Support vector networks. Mach. Learn. 20, 622–628 (1995) 21. Pedregosa, F., et al.: Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011)

450

R. Kumar and A. K. Deb

22. Lin, T.Y., et al.: Microsoft COCO: common objects in context. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) Computer Vision – ECCV 2014. ECCV 2014. LNCS, vol. 8693, pp. 740–755. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-10602-1_48 ˇ 23. Cehovin, L., Leonardis, A., Kristan, M.: Visual object tracking performance measures revisited. IEEE Trans. Image Process. 25(3), 1261–1274 (2016)

Experimental Analysis for Distance Estimation Using RSSI in Industry 4 Robin Singh Chouhan1(B) , Advait Kale1 , Anand Singh Rajawat2 , Rabindra Nath Shaw3 , and Ankush Ghosh4 1 Shri Vaishnav Institute of Information Technology, Shri Vaishnav Vidyapeeth

Vishwavidyalaya, Indore, India [email protected] 2 School of Computer science and Engineering, Sandip University, Nashik, India 3 Department of Electrical, Electronics and Communication Engineering, Galgotias University, Noida, India [email protected] 4 School of Engineering and Applied Sciences, The Neotia University, West Bengal, India

Abstract. With the increase in the number of smart autonomous bots due to industry 4.0, the need for indoor navigation has also increased. This paper deals with experimental analysis for approximating distance between two nodes using their RSSI value. Estimation and calculation are done using the FSPL algorithm. The experiments include wifi radio waves of frequency 2.4 GHz. The proposed work provides an efficient solution to calculate an approximate distance between wireless nodes for indoor distance proximation. The paper also includes a working mobile application to showcase the real-time results of the proposed work. Keywords: RSSI (Received signal strength indication) · Distance estimation · Wi-Fi · IoT · Industry 4.0

1 Introduction Whenever one rides any transportation service outdoors, the system uses GPS in order to provide its exact location on the map which is in form of latitudes and longitudes. This use of GPS is really efficient for a larger radius of distance, as GPS range is usually used to locate buildings really accurately but if we want to navigate to more-closer proximity we need to move to other sources like wifi or Bluetooth. As internet usage has increased exponentially around the globe. The buildings are now constructed keeping the wifi infrastructure in notice as all employees are almost always connected to the wifi whenever they are working so locating the employees otherwise known as nodes is easy to locate and track the movement in the factory or company. Since all Wi-Fi Access Points are known locations and usually transmit radio waves in-between ranges of 2.4 GHz and 5.7 GHz, the receiver can get to know the proximity of its node from the wifi access point [2]. There are many techniques and algorithms © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 451–463, 2022. https://doi.org/10.1007/978-981-19-1742-4_38

452

R. S. Chouhan et al.

available to measure and estimate the distance between two wirelessly connected nodes using RSSI, if the radio signal used to connect these devices can be used for the measurement of its distance then it becomes a cost-effective solution that can work both indoors and outdoors. The work includes repeated measurements of RSSI between two wirelessly connected nodes at varying distances kept both inside and outside. The nodes are tested at 2.4 GHz to get an optimal battery and greater distance than 5.7 GHz. The main objective of this paper is to point out and solve some of the problems that exploit the vulnerabilities of the RSSI. The paper tends to provide a table with all the essential information regarding the various experiments performed regarding this topic. The paper also provides the various specifications about the data sets and the platforms used to perform the experiments.

2 Related Work Some of the related works include gathering the values from real-life experiments by keeping the node in steady-state for one minute at varying distances from 0 cm to 6 m and getting the median of the fetched values at each location and storing it in a .csv file to further process. The values were collected in both indoor and outdoor conditions to achieve better results to check the application in varying conditions. Some related works include developing an android mobile application using flutter (for UI) and java (for core) to fetch and display values for RSSI and distance in real-time in the text as well as graphical form. Statistical mean and median values were taken for each location on 20 values each to fetch the most approximate value and ensure minimal errors. The paper mainly focuses on the wifi access points operating at 2.4 GHz i.e. 2412 MHz to be exact. Also, a small wifi access point was built using node MCU to act as a small portable node for further testing. The node MCU works as a wifi access point operating at 2.4 GHz and the mobile app acts as a receiver to fetch the RSSI and distance to locate and estimate the proximity between the node (here node MCU) and the user (mobile app).

3 Received Signal Strength Indicator RSSI is measured in the interval of 1 s by using the app though it takes around 3 s to process the changes in real-time, so in the research paper around 20 values are being recorded at distances from 0 cm to 600 cm for 1 min each and then the mean and median are calculated and then stored in the.csv file to further process results. The RSSI value represents the signal strength so in ideal scenarios, the RSSI values to the distance is a curve where the curve begins from −20 dB to −40 dB inside 1m distance and then the curve goes almost linear after −40 dB and before −10 dB the distance remains close to 0 m (Fig. 1).

Experimental Analysis for Distance Estimation

453

The acceptable signal strengths are shown below in the Table 1: Table 1. Acceptable signal strengths Signal strength

Comment

Description

−30 dB

Amazing

Max achievable signal strength

−67 dB

Very good

Timely delivery of data packets

−70 dB

Okay

Reliable data delivery

−80 dB

Not good

Basic connectivity

−90 dB

Unusable

Very unlikely functionality

4 Proposed Methodology In theory, a radio signal attenuates with the square of the distance from the source, hence distance could be calculated using the RSSI. While this seems a lot plausible the RSSI has certain issues involved in it as stated in [2] like the following: • RSSI isn’t constant hence it can fluctuate even if the distance between the source and receiver is constant or the devices are steady. Hence before taking the results into account, the data needs to be processed. • The transmission power also plays a major role in determining the distance. • Obstacles present in between could reflect the waves disturbing the actual results. • In a crowded environment, people could also act as obstacles. But, even with all the above obstacles, the RSSI is still preferred as: i) It requires no additional Hardware. ii) Can work in both indoors and outdoors. [5] Free space path loss: It is the depletion of radio wave energy between two nodes that results from the obstacle-free line of sight path known as free space. Free space loss increases due to imperfections and resistance, the loss increases by the square of the distance between nodes and decreases with the square of the wavelength of radio waves (Fig. 2). Free space path loss formula [5]: FSPL(dB) = 10log10 (4πdf/c)2 = 20log10 (4πdf/c) = 20log10 (d) + 20log10 (f) + 20log10 (4π/c) = 20log10 (d) + 20log10 (f) − 147.55 = 20l og10 (d) + 20log10 (f) − 27.55(frequency in MHz)

454

R. S. Chouhan et al.

Fig. 1. Formula using RSSI and frequency.

The results obtained from above formula were converted into code and then used to calculate the distance. The above formula uses both frequency and RSSI value for calculation. A new formula given by [6] was later on taken and used to calculate results and the results made by [6] gave more accurate results. Formula: Distance = 10 ((Measured Power − RSSI)/(10 ∗ N)) where, Measured Power = RSSI upto 1 m N = Environmental Constant (range 2–4)

Fig. 2. Distance at 100 cm.

From the above figure, it is visible that the distance isn’t the exact value but just an approximation as to achieve the exact distance we need to remove the environmental factor which for the current calculation is taken to be 2.5 and the measured power is calculated to be −43 dB for 2.4 GHz. Though upon testing further it was found that the first formula performs better under 1 m distance.

Experimental Analysis for Distance Estimation

455

5 Applications As the need for robots and autonomous bots is increasing so is the requirement for indoor navigation. In industry 4.0, we need a good and precise indoor navigation system to enable the bots in the factories to be able to navigate easily and more precisely. The given method for using RSSI values to estimate distance provides an affordable and small in size solution for the same. It works in real-time and hence can enable the autonomous bots to recognize their approximate distance from a given wifi access point and figure out the most optimal way to find and travel in the desired direction (Fig. 3).

Fig. 3. Indoor navigation by robots in a factory.

6 Experimental Setup The experiment was performed by using two mobiles both connected over 2.4 GHz Wifi. The devices were then kept at varying distances for 1min each and the measurements were then recorded and stored in the form of.csv files. One of the mobiles acted as the hotspot while the other used our custom-coded mobile application to fetch real-time RSSI values and calculate the distance accordingly (Fig. 4). For another experiment, node MCU was used as the wifi access point while the mobile application acted as the receiver to generate reports based on the experiment. The experiments have been performed both indoors and outdoors while noting the real distance and placing both the nodes at the exact distance between them then running the mobile application on the mobile and then stopping exactly after one minute to save all the values (Fig. 5).

456

R. S. Chouhan et al.

Fig. 4. Node MCU acting as Wifi Access Point (2.4 GHz)

Fig. 5. Wifi coverage on various corners of a house.

The above figure shows that the wifi signal strength decreases as we move further from the source but still the signal can be good enough to remain connected and provide internet services. The wifi can often get noise due to the interruption in between like walls, furniture, etc.

Experimental Analysis for Distance Estimation

457

7 Result Analysis The experiments were done using two mobile phones with the app running in one and the other kept at known and measured distance from 0 cm to 600 cm and then the app was started and closed at one minute for each distance to record around 20 measurements for each location. After each successful recording, a .csv file was created with the fields: RSSI, Distance, Distance 2. The app was then closed and then placed at a different distance to again perform the same experiment but with a different distance between the nodes to reduce human errors and other technical glitches. After successfully recording the readings in the form of .csv files mean of all the RSSI values along with distance was taken to achieve more optimal results. When the above experimental setup was carried out the following results were obtained (Fig. 6):

Fig. 6. Tables with calculated distances from both algorithms with values at known distances.

Now in order to get better accuracy and more efficient results we used python’s code to get the unique strength and distance values and then getting their mean as the result. The results are as follows (Table 2):

458

R. S. Chouhan et al. Table 2. Results from Algorithm 2

From the above table we can produce the following graph (Fig. 7).

Fig. 7. RSSI vs Distance graph

The above figure is made using the mean values of all varying distances which ranges from 0 cm to 600 cm. The below graphs showcases all the entries from all the csv files around 163 entries and is then showcased in form of scatter graph (Figs. 8 and 9).

Experimental Analysis for Distance Estimation

459

Fig. 8. Graph with all entries from all csv files.

Fig. 9. Scatter graph with all points.

With the above results in hand, the average accuracy came out to be around 90%. We also made an android mobile application that works in real-time using flutter and java. The screenshots for the same are given below (Fig. 10):

460

R. S. Chouhan et al.

Fig. 10. Screenshots of the mobile application.

Experimental Analysis for Distance Estimation

461

8 Future Work The given solution works admirably at 2.4 GHz as its more power-efficient and better in distance range than 5 GHz but a user can receive quite a fluctuating RSSI value at the same location which could cause minor issues sometimes, this could be solved by using a different frequency of radio waves for detection but that increases the cost of the entire solution. Future work includes working on a more accurate way to fetch the RSSI values faster so that the calculations are less error-prone and could provide faster and more accurate results. The given solution works on an algorithm for the entire distance estimation but machine learning could also be used to get more accurate results by training it using the current and some more datasets that could be generated by using the mobile application, but using machine learning to estimate distance could take more time than the current solution and hence improvements could be done on that parts as well.

9 Conclusion In this paper, we analyzed and reviewed two algorithms then tested them both experimentally in both indoors and outdoors to achieve almost 90% accuracy. The better algorithm was then taken and modified a bit to work in the desired conditions, this improved the accuracy in case of larger distances (here greater than 2 m). The mobile application tracks and receives the RSSI values in real-time using the core modules of android known as wifi manager, and then calculate the distance. The algorithm works admirably at 2.4 GHz which we used during the experiment. The overall result is quite efficient as has been shown with the help of the results. In this work, we recorded the data using a mobile application to start the recording of the values upon the click of a button and stop at exactly one minute and store the values in a.csv file format. This practice might not be sufficient for every real-world scenario but shows good accuracy currently and which can be further improved and optimized with other recordings by use of better algorithms or better wifi modems. Supplementary Materials Data Availability Statement The data set created by the mobile application and then later used to draw graphs, is available at: https://github.com/RobinSinghChouhan/Research_LocateIOT/tree/mai n/CSV. Mobile Application The mobile application used to fetch results is developed by the authors collaboratively and will be live on play store soon. Code Application Code repository for the node mcu for this project: https://github.com/AdvaitKale01/IoTWifi-Manager.

462

R. S. Chouhan et al.

References 1. Bullmann, M., Fetzer, T.: Comparison of 2.4 GHz WiFi FTM- and RSSI-based indoor positioning methods in realistic scenarios. https://www.mdpi.com/1424-8220/20/16/4515/pdf 2. Daiya, V., Ebenezer, J.: Experimental analysis of RSSI for distance and position estimation. In: IEEE-International Conference on Recent Trends in Information Technology, ICRTIT 2011, IEEE-International Conference on Recent Trends in Information Technology, ICRTIT 2011. 978-1-4577-0590-8/11/$26.00 ©2011 IEEE MIT, Anna University, Chennai. June 3–5 2011 (2011) 3. Miroslav BOTTA, Milan SIMEK, Adaptive Distance Estimation Based on RSSI in 802.15.4 Network. https://core.ac.uk/download/pdf/30312076.pdf 4. https://www.metageek.com/training/resources/understanding-rssi/ 5. Free Space Path Loss Calculator. Pasternack. Accessed 16 Oct 2021 6. https://iotandelectronics.wordpress.com/2016/10/07/how-to-calculate-distance-from-therssi-value-of-the-ble-beacon/ 7. Rajawat, A.S., et al.: Fog big data analysis for IoT sensor application using fusion deep learning. Math. Probl. Eng. 2021, 6876688, 16p (2021). https://doi.org/10.1155/2021/687 6688 8. Rajawat, A.S., et al.: Securing 5G-IoT device connectivity and coverage using Boltzmann machine keys generation. Math. Probl. Eng. 2021, 2330049, 10p (2021). https://doi.org/10. 1155/2021/2330049 9. Rajawat, A.S., et al.: Reformist framework for improving human security for mobile robots in industry 4.0. Mob. Inf. Syst. 2021, 4744220, 10p (2021). https://doi.org/10.1155/2021/474 4220 10. Singh, P., Sammanit, D., Krishnan, P., Agarwal, K.M., Shaw, R.N., Ghosh, A.: Combating challenges in the construction industry with blockchain technology. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 707–716. Springer, Singapore (2021). https://doi.org/10. 1007/978-981-16-0749-3_56 11. Stanculeanu, I., Borangiu, T.: Enhanced RSSI localization system for asset tracking services using non expensive IMU. In: Proceedings of the 14th IFAC Symposium on Information Control Problems in Manufacturing Bucharest, Romania (2012).https://doi.org/10.3182/201 20523-3-RO-2023.00430 12. Pratp Singh, A., Pratap Singh, D., Kumar, S.: Distance measurement using RSSI method in wireless sensor networks. https://www.ukessays.com/essays/computer-science/distance-mea surement-using-rssi-method-8607.php?vref=1 13. Bedi, P., et al.: Impact analysis of industry 4.0 on realtime smart production planning and supply chain management. In: 2021 IEEE 4th International Conference on Computing, Power and Communication Technologies (GUCON), pp. 1–6 (2021). https://doi.org/10.1109/GUC ON50781.2021.9573563 14. Kumar, M., Shenbagaraman, V.M., Shaw, R.N., Ghosh, A.: Digital transformation in smart manufacturing with industrial robot through predictive data analysis. In: Bianchini, M., Simic, M., Ghosh, A., Shaw, R.N. (eds.) Machine Learning for Robotics Applications. SCI, vol. 960, pp. 85–105. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0598-7_8 15. Daiya, V., Ebenezer, J., Murty, S.A.V.S., Raj, B.: Experimental analysis of RSSI for distance and position estimation. In: 2011 International Conference on Recent Trends in Information Technology (ICRTIT), pp. 1093–1098 (2011). https://doi.org/10.1109/ICRTIT.2011.5972367 16. Parameswaran, A.T., Husain, M.I., Upadhyaya, S.: Is RSSI a reliable parameter in sensor localization algorithms - an experimental study. In: Field Failure Data Analysis Workshop, pp. 27–30 (2009)

Experimental Analysis for Distance Estimation

463

17. Goyal, S.B., Bedi, P., Rajawat, A.S., Shaw, R.N., Ghosh, A.: Multi-objective fuzzy-swarm optimizer for data partitioning. In: Bianchini, M., Piuri, V., Das, S., Shaw, R.N. (eds.) Advanced Computing and Intelligent Technologies. LNNS, vol. 218, pp. 307–318. Springer, Singapore (2022). https://doi.org/10.1007/978-981-16-2164-2_25 18. Bedi, P., Goyal, S.B., Rajawat, A.S., Shaw, R.N., Ghosh, A.: A framework for personalizing atypical web search sessions with concept-based user profiles using selective machine learning techniques. In: Bianchini, M., Piuri, V., Das, S., Shaw, R.N. (eds.) Advanced Computing and Intelligent Technologies. LNNS, vol. 218, pp. 279–291. Springer, Singapore (2022). https:// doi.org/10.1007/978-981-16-2164-2_23 19. Lau, E.-E.-L., Lee, B.-G., Lee, S.-C., Chung, W.-Y.: Enhanced RSSI-based High accuracy real- time user location tracking system for indoor and outdoor environments. Int. J. Smart Sens. Intell. Syst. 1(2) (2008) 20. Pahtma, R., Preden, J., Agar, R., Pikk, P.: Utilization of received signal strength indication by embedded nodes. Electr. Electr. Eng. 5(93), 1392–1215 (2009)

Agriculture Field Security System Using Faster R-CNN Vishesh Kumar Mishra(B)

, Sourov Bhowmick , and Sharzeel Saleem

Department of Electrical, Electronics and Communications, Galgotias University, Greater Noida, India [email protected]

Abstract. Field security is very much needed in this time as the Agriculture Industry is rapidly changing and the need of more output in continuously rising. The volatility in the market, rising input costs of operation and increase in the crime rate makes operating Agriculture business a challenging task. In order to overcome this issue, authors have come up with the enhanced security system using Raspberry pi and GSM module techniques for Agriculture industry which protect the farm land from various kind of infiltration or trespassing by humans and other animals. Keywords: Field security · Volatility · Input cost · Faster R-CNN · Security system

1 Introduction Security is turning into a significant issue for ranchers. Despite the fact that the danger of an assault on ranch is insignificant. Burglary of ranch hardware or synthetic substances. Pyro-crime, harming of your well, or the conscious opening of a valve on a compound tank. Criminal naughtiness including unstable gear and hardware. Obliteration of restricted creatures, property, or items. Obliteration of bioengineered plants. Purposeful presentation or arrival of an infectious creature or plant illness. These are the threats that are real and faced by many farmers [1]. Farm land and plantation in India are in very large scale running on hundreds of acres and in most cases fencing these large expanses of land can be prohibitively expensive and very stressful. The damage done by wild creatures by intersection the human areas has being expanding on a dramatic rate. These creatures are known to spread different diseases like herpes B infection, rabies, genuine injury disease, Rocky Mountain spotted fever, bone contamination, tularemia, and toxoplasmosis and so on serious toxoplasmosis can even reason harm to the eyes, cerebrum or different organs. Hence, an answer is in reality expected to keep people from such illnesses happening from wild creatures. Numerous techniques have been won to handle this issue including scarecrow, wall, trackers and so forth However, every one of them needs at a state of time relating to their actual constraints, capital prerequisite, accessibility, etc. In this manner, another technique had © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 464–471, 2022. https://doi.org/10.1007/978-981-19-1742-4_39

Agriculture Field Security System Using Faster R-CNN

465

created which used ultrasound to confine creatures from its reach to beat its precursor’s provisos. In this paper, authors have focused after improving the highlights of the agriculture field security systems to make it more productive and solid for the current human requirements. The system has been sub-partitioned into four areas in which working of the system is defined. Area 1, starts light upon the motion sensors used in the system. Infrared camera system is given in area 2. Continuing ahead, segment 3 infers the Automatic light system. The system ends at area 4 defining the work of Ultrasonic Sound Wave System.

2 Literature Survey Ultrasonic sensors, camera, GSM Module and buzzers to safeguard any intrusion. This framework is high on productivity in forestalling the creatures from the agricultural fields and fending them off. The presence of creatures is conceivable utilizing the sound of buzzer. One won’t hear a buzzer sound if an approved individual enters nearby or if there is some irregular movement. The framework covers five meter round the limit where PIR sensor is set. Also, their system doesn’t require much human supervision [2, 3]. Dongxian, Youlu, Yingzhe, Hua has written in their research work that the constant soil and climate information, and crop pictures can be powerfully gathered in far off region by the harvest field observing frameworks in distant. This yield field distant checking framework utilizing web-worker installed innovation and CDMA administration with IPsec-based VPN work as a hub framework is amazing and helpful to develop an omnipresent remote detecting network in minimal effort and high-security for crop creation [4, 5]. Sai Karthik, Naresh, Shivam Neer, Ranjan in their paper have discussed that the security is given to the dairy cattle ranch without the prerequisite of an individual utilizing this proposed framework. Each time an interruption happens, it very well may be identified independent to the presence of an individual. This can be utilized for support of dairy cattle cultivates and giving them security in a minimal effort and in a proficient way. Depending on the point behind this security alert framework, it is likewise reasoned that this security alert can be executed for a wide range of fields just as to guard diverse costly, remarkable and uncommon things or species. Regardless, the real future extent of this paper is to send the proposed framework into the reality cows ranches [6, 7]. Tanmay, Nitika, Pushpendra had created clever security frameworks with capacity to investigate information and communicate data over organization to the far-off area. That can be upgraded by coordinating not many new advancements with present plan. Current IP based CCTV surveillance cameras require network availability for checking from distant area. It doesn’t have the capacity to advise client by breaking down information. In the gadget, fundamental sensors also, electronic gadgets are utilized. The tactile data are breaking down to initiate electronic gadgets and raspberry pi is utilized as a worker to examine information and communicate data to client [8, 9].

466

V. K. Mishra et al.

3 Proposed System Authors have proposed a system that utilises microcontroller, Raspberry Pi that accepts python commands and the CNN model trained. The system is incorporated with different types of sensor as mentioned below which work together to make the field secure. The sensors used in this system are as follow: 1. 2. 3. 4.

Motion sensor Infrared Cameras Automatic light system Ultrasonic Sound Wave System

3.1 Block Diagram See Fig. 1.

Fig. 1. Block diagram

3.2 Explanation of Block Diagram The microcontroller acts as the backbone of the system and controls the working of different sensors incorporated in the system. The working of different sensors used is mentioned below: 3.2.1 Motion Sensor Motion detector/Movement identifier is a sensor which can recognize any sort of movement with higher precision. A motion sensor is a device that distinguishes moving things, particularly people. Such a device is often organized as a piece of a structure that therefore plays out a task or alerts the customer about the development in a zone. They

Agriculture Field Security System Using Faster R-CNN

467

structure a fundamental section of security. An electronic movement locator contains an optical, microwave, or acoustic sensor, and a large part of the time, a transmitter for light. Regardless, a dormant sensor identifies an imprint just from the moving item by methods for release or reflection, i.e., it might be emanated by the article, or by some enveloping maker. For instance, the sun or a radio station of satisfactory strength. Changes in the optical, microwave, or acoustic field in the gadget’s region are unravelled by the equipment subject to one of the advances recorded under. Most simple movement identifier can perceive up to distances of in any occasion 10 m. Explicit structures cost more, anyway goes have any more. Tomographic development disclosure systems can cover much greater zones considering the way that the radio waves are at frequencies which enter most dividers and snags, and are perceived in various territories, not exactly at the territory of the transmitter. 3.2.2 Automatic Light System It will be a microcontroller-based system which will take input directly from the motion sensor i.e., motion sensor will give input to the microcontroller when it senses any motion in its area and then the microcontroller will trigger the circuit of automatic light system and switch on the light. 3.2.3 Infrared Camera Authors have designed system with infrared cameras connected to motion sensor so that it can capture the image of an intruder/animals trying to enter the field. Camera will also have night vision so that it can recognise in the dark as well, also the camera will have 360-degree rotational capability so that it can cover large area of land. The camera works for the high range up to 1 km. The camera will remain connected with the motion sensor all the time and whenever any interruption sensed by the motion sensor, it will give input to the camera and the camera will click the image in that instance. Image captured by the camera will be recognise using Faster R-CNN technique. Detection speed and accuracy of the Faster R-CNN is quite high then other image recognition techniques which will make this system faster in comparison to others. 3.2.4 Ultrasonic Sound Wave System These are high frequency inaudible sound for human ear. The frequency generally exceeds 20 kHz. But it can hear by animals. So, we adopted this system for the animals who try to enter into the farm and damage the crop.

468

V. K. Mishra et al.

This table represent the table of different kinds of animals of their audible capacity in (HZ) so in this process we can identify which animal entre into the farm land [10, 11] (Fig. 2). 3.2.5 Expected Outcome As mentioned above, the system will work as a single entity though its working depends on various sensors and technologies. The motion sensor will identify the any type of intrusion whether it is done by any animal or a human being and triggers the signal to the ultrasonic device and automatic light system. The ultrasonic device will transmit the waves according the identified creatures and in the meantime, automatic light system will also get triggered to switch on the lights of the field. The output of the motion sensor, ultrasonic device and automatic light system will give input to the Raspberry Pi microcontroller which will switch on the infrared camera then the system will try

Agriculture Field Security System Using Faster R-CNN

469

Fig. 2. Ultrasonic Sound transmitter [12]

to identify the images captured using faster R-CNN technology and if the image is recognized to be suspicious, it will turn on the alarm and send a message to the owner using GSM technology.

Intruders Automatic Light System

Motion Sensor Triggers sensor Ultrasonic Device Image Capturing Security Camera Image Recognition Faster R-CNN Message sent to the owner using GSM Alarm, Message

Owner

470

V. K. Mishra et al.

4 Conclusion After much discussion, authors have come up on this conclusion that this system will provide security to the farm fields with higher accuracy and least chances of error. The system will work on the solar energy, making it more efficient and less dependent on the power grid of the area. Recognition of the animals using Faster R-CNN makes it faster and will increase the impulse response of the system. Using of GSM technology for sending the alert to the owner will help the person to save its field by the intruders. The systems can detect the any type of disturbance within the 10 m of range and will send the alert message to the owner in a matter of seconds.

References 1. https://www.extension.purdue.edu/eden/ruralsecurity/threats.html 2. Yadahalli, S., Parmar, A., Deshpande, A.: Smart intrusion detection system for crop protection by using Arduino. In: 2020 Second International Conference on Inventive Research in Computing Applications (ICIRCA), Coimbatore, India, pp. 405–408 (2020). https://doi.org/ 10.1109/ICIRCA48905.2020.9182868 3. Bodapati, S., Bandarupally, H., Shaw, R.N., Ghosh, A.: Comparison and analysis of RNNLSTMs and CNNs for social reviews classification. In: Bansal, J.C., Fung, L.C.C., Simic, M., Ghosh, A. (eds.) Advances in Applications of Data-Driven Computing. AISC, vol. 1319, pp. 49–59. Springer, Singapore (2021). https://doi.org/10.1007/978-981-33-6919-1_4 4. He, D., Bai, Y., Wang, Y., Wu, H.: A crop field remote monitoring system based on web-serverembedded technology and CDMA service. In: 2007 International Symposium on Applications and the Internet Workshops, Hiroshima, Japan, p. 72 (2007). https://doi.org/10.1109/SAINTW.2007.6 5. Mukhopadhyay, M., et al.: Facial emotion recognition based on textural pattern and convolutional neural network. In: 2021 IEEE 4th International Conference on Computing, Power and Communication Technologies (GUCON), pp. 1–6 (2021). https://doi.org/10.1109/GUC ON50781.2021.9573860 6. Das, S., et al.: Advance machine learning and artificial intelligence applications in service robot. In: Artificial Intelligence for Future Generation Robotics, pp. 83–91 (2021). https:// doi.org/10.1016/B978-0-323-85498-6.00002-2 7. Sharma, P., et al.: Vibration-based diagnosis of defect embedded in inner raceway of ball bearing using 1D convolutional neural network. In: Artificial Intelligence for Future Generation Robotics, pp. 25–36 (2021). https://doi.org/10.1016/B978-0-323-85498-6.00011-3 8. Baranwal, T., Nitika, Pateriya, P.K.: Development of IoT based smart security and monitoring devices for agriculture. In: 2016 6th International Conference - Cloud System and Big Data Engineering (Confluence), Noida, India, pp. 597–602 (2016). https://doi.org/10.1109/CON FLUENCE.2016.7508189 9. Tubaishat, M., Madria, S.K.: Sensor networks: an overview. IEEE Potentials 22(2), 20–23 (2003) 10. Rajawat, A.S., et al.: Depression detection for elderly people using AI robotic systems leveraging the Nelder–Mead method. In: Artificial Intelligence for Future Generation Robotics, pp. 55–70 (2021). https://doi.org/10.1016/B978-0-323-85498-6.00006-X

Agriculture Field Security System Using Faster R-CNN

471

11. Amritha Ashok, K., Savy, A., Shijoh, V., Shaw, R.N., Ghosh, A.: Hospital assistance robots control strategy and machine learning technology. In: Bianchini, M., Simic, M., Ghosh, A., Shaw, R.N. (eds.) Machine Learning for Robotics Applications. SCI, vol. 960, pp. 35–46. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0598-7_3 12. Rajawat, A.S., Rawat, R., Shaw, R.N., Ghosh, A.: Cyber physical system fraud analysis by mobile robot. In: Bianchini, M., Simic, M., Ghosh, A., Shaw, R.N. (eds.) Machine Learning for Robotics Applications. SCI, vol. 960, pp. 47–61. Springer, Singapore (2021). https://doi. org/10.1007/978-981-16-0598-7_4

Automated Relational Triple Extraction from Unstructured Text Using Transformer Akshay Hari(B) and Priyanka Kumar Department of Computer Science and Engineering, Amrita School of Engineering, Amrita Vishwa Vidyapeetham, Coimbatore, India [email protected], [email protected]

Abstract. In modern times, large amount of textual data is generated. Quick comprehension of knowledge from this massive amount of data is difficult for human beings as well as machines. In this paper, we propose a deep learning based framework for joint extraction of entities and relations from unstructured text. This will be implemented with state-of-the-art Transformer based language model. Our model is a light version of the existing state-of-the-art models for the same task with only half of their trainable parameter while maintaining good evaluation scores. The model is trained and tested on NYT and WebNLG dataset and evaluation is done using metrics such as Precision, Recall and F1 scores. Keywords: Deep learning · NLP · Relation extraction · Transformers

1 Introduction In the modern age, there has been an increase in data. These data are mostly stored in the electronic form. The most common is the textual form in which information is stored in an unstructured manner. In order to this data to be useful, we should be able to retrieve most important information from this text seamlessly. In this work, we propose a deep-learning based approach to extract relational triples in text. Relational triples are entities in a sentence which are in the form subject predicate - objects. These relational triples can then be used for knowledge engineering applications. The relational triples are extracted from unstructured text using a DistilBERT [1, 2] based transformer language model. One of the major highlights of our transformer-based model is that it will be able to capture dependencies in long sentences. Our model is also capable of extracting sentences with overlapping entities. This is a case where triples share same entities and relations. This scenario is explained in detail in the coming section. The final important aspect of our model is the joint entity-relation extraction. In the earlier models, entities and relations were separately learned in a pipe-lined manner, which resulted in error propagation from one stage to another.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 472–480, 2022. https://doi.org/10.1007/978-981-19-1742-4_40

Relational Triple Extraction

473

1.1 Overlapping Entities Earlier models for this task were not able to handle sentences with overlapping entities. This is a scenario where entities are shared by multiple triples in same sentence. This can be categorized into mainly two types: Single Entity Overlapping (SEO) and Entity Pair Overlapping (EPO). Single Entity Overlapping occurs when multiple triples have same entity shared as subject or object. Entity Pair Overlap occurs when multiple triples have same entity pairs. A visual representation of these scenarios is given in Fig. 1.

Fig. 1. Categories of triples in a sentence. Subject, Predicate and Object suffixes are added to the entities

1.2 Transformers Most of the state-of-the-art models in the Natural Language Processing domain currently uses Transformer based language models. Transformer models are a new type of deep learning model [3] which uses attention mechanism to find global dependencies in input and output. The transformer model processes sentences in a non-sequential manner which helps in processing sentence as whole rather than word by word. All the above features were not prevalent in earlier deep neural network-based models for same task. However, most of the state-of the-art transformers have very high number of layers and parameters. In our work, we are mainly using encoder mechanism of the transformer for language modelling. Our model can be summarized as follows. First raw textual inputs are converted into tokens using tokenizer. Then these tokens along with masks are fed into the encoder module to get the embedding. Since we’re using DistilBERT based model, the output embedding is contextual in nature. This embedding along with the embedding of triplet labels are fed into the model for training. The loss is calculated in propagation with subject’s and object’s head and tail position in the sentence. The final output is the subject’s and object’s head and tail position in the sentence along with the relation.

474

A. Hari and P. Kumar

2 Related Works In the Information Extraction or Relation Extraction domain, one of the earlier notable work is [4] extracting features using Support Vector Machines. Later [5] approached the problem with a two-step solution, first is finding all entities using Named Entity Recognition (NER) and then classifying all the extracted entity pairs using relation classification (RC). These pipeline-based approaches however suffered from error propagation problem. To address this issue, joint models [6] have been proposed which learns entities and relations together. The earlier works, however did not addressed the problem of overlapping entities encountered in a sentence i.e. multiple triples in same sentence sharing same entities. This problem was only recently addressed using deep neural network based models in the works of [7], which is based on sequence-to-sequence learning with copy mechanism using Bi-directional LSTM. Later the evaluation scores were improved by [8] using Graph Convolutional Networks and Bi-LSTMs. The recent works by [9] and [10] further improves the evaluation scores using BERT based transformer language model. Other recent works involving the usage of transformers in knowledge extractions include [11–13].

3 Dataset For training and testing of our relation extraction framework, we are using two public dataset, New York Times dataset and WebNLG dataset. The original NYT dataset [14] was created with distant supervision approach and WebNLG dataset [15] for Natural Language Generation. These datasets have been modified as per the requirement [7]. The resulting NYT dataset consists of 24 classes, 56195 training data, 5000 validation data and 5000 test data. The WebNLG dataset consists of total 5019 training data, 500 validation data and 703 test data. Detailed information is given in Table 1. Total train data in each dataset used for training, however testing is done on individual component. The testing data can be classified into three types Normal, Entity Pair Overlap and Single Entity Overlap. The testing data can be further classified on basis of number of relational triples exists on a single sentence. All the testing data without categorized is marked as main. Tabulated information of the dataset is given in Table 2. In the Table 2, for the rows ‘Triple-i’, i denotes number of triples in a single sentence. Table 1. Dataset information

Relational Triple Extraction

475

Table 2. Categorization of testing data

4 Model Architecture For the relational extraction model, we followed the work of [9, 10] which was implemented using BERT based encoder and Graph Neural network. We have optimized the size of the same using DistilBERT based transformer framework without using Graph Network Layer from baseline as accuracy gains from Graph Neural Network was negligible in our experiments when considering number of trainable parameters it added. This allowed us to significantly reduce the trainable parameters without compromising much of accuracy as well as lowering the model training time. For relation extraction framework (Fig. 2), our work consists of two parts: encoding words from input sentence into vector embeddings and encoding each relation into vectors and then subject and object tagger based relational triple extraction. The problem can be formulated as mentioned. Given a sentence x, and set of all triplets (s,r,o) in training set T, our goal is to maximize the data-likelihood in the training set. This can be mathematically defined as mention in Eq. 1: 

p((s, r, o)|x)

(s,r,o)∈T

=



p(s|x)



s∈T

p((r, o)|x, s)

(r,o)∈T |s

s∈T

=



p(s|x)



r∈T |s

p(o|x, s, r)



p(o∅ |x, s, r)

(1)

r∈R\T |s

where T | s is the triplet set with s as subject in T. Similarly, (r,o) ∈ T | s is the set of all relation-object pair in T. R is the set of all relations and R\T | s means all the relations except subject s in T. o∅ represents all relations except those in triplet T |s will have no corresponding objects.

476

A. Hari and P. Kumar

Fig. 2. Architecture of our Relation Extraction model

First, for a given input sentence, a pre-trained DistilBERT encoder is used for extracting tokens for each word and for each predefined relation, an embedding is created as shown in the Eq. 2.  [h1 , h2 , · · ·hn ] = ED ([w1 , w2 . . . wn ]) p1 , p2 . . . pm = Wr E([r1 , r2 . . . rm ]) + br

(2)

where wi is word from input sentence and hi is the output token from DistilBERT encoder E D Similarly, pi is the output after relation embedding matrix E embeds predefined relations r i . W r and br are trainable parameters. For relation extraction, subject taggers and object taggers are used. The subject tagger defined in Eq. 4 will identify all possible subjects in the word nodes. More specifically, it will tag the head and tail of the subject using sigmoid function, defined in Eq. 3. σ (x) =

1 1 + e−x

(3)

The sigmoid function maps the values between 0 and 1. ss− head

Pi

Pist tail

    = σ Ws− head Tanh hoi + bs− head     = σ Ws− tail Tanh hoi + bs− tail

(4)

where Pis_head , Pis_tail are the probabilities of identifying the ith word as head and tail position of the subject respectively which is calculated by the sigmoid function σ. The values Ws− head , Ws− tail , bs− hhead , bs− tail are trainable weights. hoi is the encoded representation of the word from previous stage.

Relational Triple Extraction

477

Similarly, the object tagger, defined in Eq. 5 uses encoded word token which is different from token used by subject tagger.   Pio_head = σ Wo− head hijk + bo− head   (5) o tail Pi − = σ Wo− tail hijk + bo− tail o tail

where Pio_head , Pi − are the probabilities of identifying the ith word as the head and tail position of the object respectively which is calculated by the sigmoid function σ. The values Wo_head , Wo_tail , bo_head , bo_tail are trainable weights. The term hijk is encoded word token representation which can be defined as

  (6) hijk = Tanh Wh sk ; pj0 ; h0i + bh where sk is the subject representation of the kth candidate subject, p0 j and ho i are the encoded representation of the pre-defined relation and word token respectively. Therefore, in line with Eq. 1, we can define subject tagger and object tagger as Eq. 7 and 8 respectively: 

Pθs (s|x) =

N   t I{yt =1}  I{yt =0} i Pi 1 − Pit i

(7)

t∈{s− head ,s− tail} i=1



Pθo (o|x, s, r) =

N   t I{yt =1}  I{yt =0} i Pi 1 − Pit i

(8)

t∈{o− head ,o_tail} i=1

where θ s and θ o are the parameters of the subject tagger and object tagger respectively. I{z} = 1 if z is true otherwise it is 0 yis_head , yis_tail and yio_head , yio_tail are binary tags of subject’s and object’s heads and tails respectively for the ith word in x, For the null o object o∅ in Eq. 1, yi ∅_head = yio∅_tail = 0 for all i. Taking the logarithm of 1, we get the objective function which is defined in Eq. 9  p((s, r, o)|x) L = log =



(s,r,o)∈T

logpθs (s|x) +

s∈Tj

 r∈Tj |s

logpθo (o|x, s, r) +



log pθo (o∅|x, s, r)

(9)

r∈R\Tj |s

The log-likelihood function is then maximized by using Stochastic Gradient Descent during training. The learning rate is set as 0.1 for both datasets.

5 Evaluation Metrics We used precision, recall and F1-scores as evaluation metrics following the baseline approach. A triplet is considered correct only if its predicate and its corresponding subject and object is correct. Additionally, we also used number of trainable parameter in transformer model for comparison as it will help us to identify the efficiency of the model with respect to neural network size as well as gives us an idea of model training time.

478

A. Hari and P. Kumar

6 Implementation Details and Results The model is implemented with PyTorch library along with CUDA 11. Base DistilBERT model is used from Huggingface [16] with transformer library version 4.12. For both datasets, the models are set to run on maximum of 60 epochs with an early stopping mechanism. The early stopping mechanism will be triggered if there is no improvement in the score for 15 consecutive epochs. Both of the datasets used Stochastic Gradient Boost optimizer with a learning rate of 0.1. The training data is further split into training and validation data. The hyperparameters are determined from this validation data. We were able to significantly reduce the number of trainable parameters. A comparison of trainable parameters with other transformer-based model is given in Table 3. Table 3. Comparison of trainable parameters

The detailed result of our model from testing of different categories of testing data is tabulated and given in Table 4. It is observable that our model performed fairly good in all triple category scenarios. A slight drop in the score in WebNLG dataset when compared with NYT dataset maybe attributed to the fact that WebNLG main category has most of the data in SEO and EPO form. For the NYT dataset, our model performed the best when there were 4 triples in the sentence and for the WebNLG dataset, the model performed well when there were 3 triples in the sentence. Therefore, from these results, it is evident that our transformer based model is perfectly capable of handling complex scenarios in relational triple extraction. Table 4. Evaluation results on NYT and WebNLG dataset

Relational Triple Extraction

479

7 Conclusion and Future Scope In this paper, we proposed a light version of transformer-based model for Relation Extraction based on joint entity-relation extraction framework. Our model performed well in all triplet overlapping scenarios such as Entity Pair Overlapping (EPO) and Single Entity Overlapping (SEO) and can extract multiple triplets from same sentence while reducing the number of trainable parameters in the transformer. In the future, we aim to reduce the number of trainable parameters further while improving the performance.

References 1. Sanh, V., Debut, L., Chaumond, J., Wolf, T.: DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108 (2019) 2. Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: BERT: pre-training of deep bidirectional transformers for language understanding. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, vol. 1 (Long and Short Papers), pp. 4171–4186 (2019) 3. Vaswani, A., et al.: Attention is all you need. In: Advances in Neural Information Processing Systems, pp. 5998–6008 (2017) 4. Zhou, G., Su, J., Zhang, J., Zhang, M.: Exploring various knowledge in relation extraction. In: Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL 2005), pp. 427–434 (2005) 5. Mintz, M., Bills, S., Snow, R., Jurafsky, D.: Distant supervision for relation extraction without labeled data. In: Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP, pp. 1003–1011 (2009) 6. Miwa, M., Sasaki, Y.: Modeling joint entity and relation extraction with table representation. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 1858–1869 (2014) 7. Zeng, X., Zeng, D., He, S., Liu, K., Zhao, J.: Extracting relational facts by an end-to-end neural model with copy mechanism. In: Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (vol. 1: Long Papers), pp. 506–514 (2018) 8. Fu, T.-J., Li, P.-H., Ma, W.-Y.: Graphrel: modeling text as relational graphs for joint entity and relation extraction. In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 1409–1418 (2019) 9. Wei, Z., Su, J., Wang, Y., Tian, Y., Chang, Y.: A novel cascade binary tagging framework for relational triple extraction. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 1476–1488 (2020) 10. Zhao, K., Xu, H., Cheng, Y., Li, X., Gao, K.: Representation iterative fusion based on heterogeneous graph neural network for joint entity and relation extraction. Knowl.-Based Syst. 219, 106888 (2021) 11. Veena, G., Athulya, S., Shaji, S., Gupta, D.: A graph-based relation extraction method for question answering system. In: 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI), pp. 944–949. IEEE (2017) 12. Nair, A.M., Bindu, K.R.: Semantic role labelling using transfer learning model. In: Journal of Physics: Conference Series, vol. 1767, p. 012024. IOP Publishing (2021) 13. Gangadharan, V., Gupta, D., Amritha, L., Athira, T.A.: Paraphrase detection using deep neural network based word embedding techniques. In: 2020 4th International Conference on Trends in Electronics and Informatics (ICOEI) (48184), pp. 517–521. IEEE (2020)

480

A. Hari and P. Kumar

14. Riedel, S., Yao, L., McCallum, A.: Modeling relations and their mentions without labeled text. In: Balcázar, J.L., Bonchi, F., Gionis, A., Sebag, M. (eds.) ECML PKDD 2010. LNCS (LNAI), vol. 6323, pp. 148–163. Springer, Heidelberg (2010). https://doi.org/10.1007/9783-642-15939-8_10 15. Gardent, C., Shimorina, A., Narayan, S., Perez-Beltrachini, L.: Creating training corpora for NLG micro-planning. In: 55th Annual Meeting of the Association for Computational Linguistics (ACL) (2017) 16. Wolf, T., et al.: Huggingface’s transformers: state-of-the-art natural language processing. arXiv preprint arXiv:1910.03771 (2019)

A Modified SVPWM Strategy to Improve the Performance of Variable Frequency Induction Motor Drive Nazmul Islam Nahin1 , Shuvra Prokash Biswas1(B) , and Md. Rabiul Islam2 1 Rajshahi University of Engineering and Technology, Rajshahi 6204, Bangladesh

[email protected]

2 University of Wollongong, Wollongong, NSW 2522, Australia

[email protected]

Abstract. Variable frequency induction motors are extensively used in industry. The performance of variable frequency induction motor drive (VFIMD) significantly depends on the power quality of the motor driven inverter. Moreover, the power quality of the motor driven inverter is greatly influenced by the pulse width modulation (PWM) technique. Space vector pulse width modulation (SVPWM) is very well known for VFIMD. Traditional SVPWM suffers from harmonics, power loss, and torque ripple problem. In this work, a modified SVPWM strategy is proposed to mitigate the shortcomings of traditional SVPWM. The proposed technique not only improves total harmonic distortion (THD) of inverter output voltage and power loss but also reduces the torque ripple of the VFIMD. All the simulation work are carried out in MATLAB/Simulation environment to verify the effectiveness of the proposed technique. Keywords: Induction motor drive · Pulse width modulation · Total harmonic distortion · Space vector · Power loss

1 Introduction In power electronics, three-phase voltage source inverters (VSIs) are used in a variety of applications such as industrial motor drives, electric vehicles, and alternate energy grid interfacing [1]. In both industrial and commercial applications, pulse width modulated (PWM) VSIs are considered as the driving force [2]. The performances of variable frequency induction motor drive (VFIMD) are greatly influenced by the power quality of motor driven VSI. On the contrary, the PWM technique employed in switching of VSI determines the power quality of the inverter. PWM techniques can improve the power quality of inverter by controlling the harmonic spectrum of the output voltage. It can also help to lessen torque pulsation at lower frequencies [3]. The most famous method is sinusoidal pulse width modulation (SPWM) which is very easy to implement but has number of drawbacks including high total harmonic distortion (THD), power loss, and pulsating torque problems [4]. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 481–493, 2022. https://doi.org/10.1007/978-981-19-1742-4_41

482

N. I. Nahin et al.

In three-phase VSIs, the conventional space vector pulse width modulation (SVPWM) technique is extensively used to produce good static and dynamic performance of VSI fed VFIMD. It is superior to sinusoidal pulse width modulation (SPWM) in terms of better utilization of dc bus voltage and less THD. It also offers smooth dynamic response of VFIMD compared to SPWM. Conventional SVPWM synthesizes the output voltage vector with an average value equal to the reference signal using a set of space vectors. Control of induction motor using conventional SVPWM was presented in [5]. In [6–8], details comparison between carrier based PWM and SVPWM were discussed. In [9], an advance modulation technique is used to improve power quality in induction motor drives fed by 12-pulse rectifiers. Although conventional SVPWM outperforms form SPWM but it has some drawbacks which are still a research challenge. The main drawbacks of conventional SVPWM are high THD, power loss, and torque ripple. In conventional SVPWM, the switching states are given as 000 → 100 → 110 → 111 → 000 → 100 → 110 → 111. For 111 → 000 switching state, three switches simultaneously are being switched off which introduces high switching loss of the VSI [10, 11]. Moreover, conventional SVPWM produces a huge amount of lower order harmonics in inverter output voltage while operating in the variable frequency mode. So, it is always an industrial concern for further improvement of conventional SVPWM technique to improve the performance of VSI based VFIMD using a modified SVPWM technique. To mitigate the abovementioned problems of conventional SVPWM based VFIMD, a modified SVPWM technique is presented in this work. The proposed technique utilizes the zero state vectors by modifying the time period of the switching signals. As a result, a modified SVPWM based modulating signal is produced which produces balanced and symmetric gate pulses for the VSI. The presented modified SVPWM technique not only reduces THD but also improves the power loss and torque ripple profiles of the VSI based VFIMD compared to its conventional counterpart.

2 System Description The system structure of the VFIMD is shown in Fig. 1. A three level VSI is used to feed power to a 7.5 kW induction motor. Second order LC filter (L = 2.3 mH, C = 3.6 µF) is utilized to suppress harmonics from the three level output voltage of the VSI. The presented modified SVPWM technique is utilized to control the power switches of the VSI.

A Modified SVPWM Strategy to Improve the Performance

S1

3-phase 415V,7.5kW Induction Motor

S3

S2

483

Lm

Vdc

Lm

IM

Lm

S5

S4

S6

Cm

Cm

Cm

Three level VSI Fig. 1. Three-phase VSI based VFIMD.

3 Proposed Modified SVPWM Strategy 3.1 Switching Control Structure The switching control structure of the proposed SVPWM technique for the VFIMD of Fig. 1 is depicted in Fig. 2.

Va Vb Vc

abc

αβ

Vref θ

Cartesian to Polar

Sector detection

n

Ts Vdc

Multiport Switch

Sector S1 S2 S3 S4 S5 S6

IM

Induction Motor

Time periods Ta, Tb & T0 Estimation

Three level VSI

Pulse Generator

High frequency carrier signal

Modified SVPWM signals

Fig. 2. Switching control structure of the VFIMD with proposed SVPWM technique.

At first, the three dimensional abc plane is converted into a two dimensional αβ plane using Clarke’s transformation. The αβ plane has a reference voltage with magnitude, V ref and an angle, θ. The αβ complex plane is divided into six sectors each having 60° separation. The angle, θ determines the arrival of the reference voltage in each sector. Then the reference voltage (V ref ), angle (θ ), corresponding sector (n), dc bus voltage (V dc ) and frequency (f ) are used to determine the timing period of the eight active vectors which give the three reference signals. These three reference signals are the modulating signals of the proposed SVPWM technique. After that, the modulating signals are compared to high frequency carrier wave to produce the PWM pulses for the converter.

484

N. I. Nahin et al.

3.2 Mathematical Development of the Proposed SVPWM Strategy The modified SVPWM strategy is proposed based on the utilization of zero state vectors V 0 and V 7 as indicated in Fig. 3. The time instant of the application of zero state vectors V 0 and V 7 are modified to obtain the modified SVPWM technique compared to existing SVPWM. In conventional SVPWM, the two zero state vectors are operated for 0.5T s time span against each sector. Whereas the modified SVPWM technique imposes 0.866T s time span (for V o ) for one of the zero vector. The remaining time spans are occupied by the other vector. The switching sequence of sector 1 for the modified SVPWM technique is depicted in Fig. 3. V0

V1

V2

.433 T0

.5Ta

.5Tb

V7

V7

V1

V2

.5Ta

.5Tb

V0

S1

S2

S3 .134T0

.433T0

Fig. 3. Switching sequence of sector 1.

Three reference balanced voltages V a , V b and V c are given as, Va = Vm sin(ωt)

(1)

  Vb = Vm sin ωt + 120◦

(2)

  Vc = Vm sin ωt + 240◦

(3)

Three phase model can be converted to two phase model using Clark’s transformation by the following equations, ⎡ V ⎤    a 1 1 2 − 1 − Vx 2 √2 √ ⎣ Vb ⎦ = (4) Vy 3 0 23 − 23 Vc

Vref | = V 2 + V 2 (5) x y θ = tan

−1



Vy Vx

 (6)

A Modified SVPWM Strategy to Improve the Performance

485

The corresponding line to neutral voltage and space vectors are shown in Table 1. These eight active vectors divide the complex plane into six sectors shown in Fig. 4. The reference vector for each sector is generated by using the two adjacent vectors and one zero vector (V0 or V7 ) with their corresponding time period T a , T b , and T 0 respectively. The duty cycle for each sector are indicated in Table 2 which are found by the following equations. Vref Tc = V1 Ta + V2 Tb + (V7 /V0 ) T0

(7)

Vref = Da V1 + Db V2 + D3 (V7 /V0 )

(8)

where, Tc = Ta + Tb + T0 .   sin θ− (n−1)π sin( nπ 3 3 −θ ) and Db = a Da = a sin π sin π 3 √ |Vref | 3 where a = modulation index = 3 V s . Sector 2

V2(1,1,0) Vy

V3(0,1,0)

Sector 3

V1(1,0,0) Sector 1

Vref

V0(0,0,0) V7(1,1,1)

θ

Vx

V6(1,0,1)

V4(0,1,1)

Sector 6

Sector 4

Sector 5

V5(0,0,1)

Fig. 4. Voltage space vectors for modified SVPWM technique

The null vectors V 0 and V 7 are operated for the time interval m0 T 0 and (1 − m0 )T 0, respectively where 0 ≤ m0 ≤ 1. The switching interval for the upper switches are S 1 , S 2 and S 3 are, S1 = m0 T0 + Ta + Tb

(9)

S2 = m0 T0 + Tb

(10)

S3 = m0 T0

(11)

486

N. I. Nahin et al. Table 1. Line to neutral voltages for eight vectors.

Vectors

Va

Vb

Vc

Space vectors

V0 (0, 0, 0)

0

0

0

0 < 0°

V1 (1, 0, 0)

V6 (1, 0, 1)

2Vs 3 Vs 3 − V3s s − 2V 3 − V3s Vs 3

− V3s Vs 3 2Vs 3 Vs 3 − V3s s − 2V 3

− V3s s − 2V 3 Vs −3 Vs 3 2Vs 3 Vs 3

2 3 2 3 2 3 2 3 2 3 2 3

V7 (1, 1, 1)

0

0

0

0 < 0°

V2 (1, 1, 0) V3 (0, 1, 0) V4 (0, 1, 1) V5 (0, 0, 1)

< 0◦ < 60◦ < 120◦ < 180◦ < 240◦ < 300◦

Table 2. Duty cycle of modified SVPWM for each sector for Inverter upper switches. Sector

S1

S2

S3

1

Tb + .866T0

.866T0

2

Ta + Tb + .866T0 Tb + .866T0

Ta + Tb + .866T0

.866T0

3

.866T0

Tb + .866T0

4

.866T0

Ta + Tb + .866T0 Tb + .866T0

5

Tb + .866T0

.866T0

Ta + Tb + .866T0 Ta + Tb + .866T0

6

Ta + Tb + .866T0

.866T0

Tb + .866T0

The parameter m0 can be varied in such a way that the THD and power losses of the VSI are optimized. The modulating signal generation algorithm for the modified SVPWM technique is depicted in Fig. 5. The generated gate pulses utilizing modified SVPWM technique are shown in Fig. 6. Figure 7 depicts the modulating signals of conventional and proposed SVPWM technique to clearly observe the major difference between them. The zoomed view of Fig. 7 shows that the modulating signal of modified SVPWM is top flatted compared to conventional SVPWM. This flatted top modulating signal of the modified SVPWM technique generates more symmetric gate pulses compared to conventional SVPWM which aids to reduce THD from the inverter output voltage. It also reduces the conduction time of the switches and number of switching which also aid to reduce the power loss of the inverter. Thus, the proposed SVPWM technique outperforms from conventional SVPWM technique in term of THD and power loss.

A Modified SVPWM Strategy to Improve the Performance

Va

Vc

Vb

abc

αβ

Vref , θ generation

Ta, Tb and T0

Sector

1

2

3

Ta+Tb+.866T0 Tb + .866T0 Modified SVPWM signals

.866T0

Fig. 5. Algorithm for implementing modified SVPWM technique. Proposed modulating signal

Carrier signal

Amplitude (p.u)

1 0 1 0 1 0 1 0

0

0.01 Time (s)

0.02

Fig. 6. Generated gate pulses for the VSI with modified SVPWM technique.

487

N. I. Nahin et al.

Amplitude

488

Modified SVPWM

Conventional SVPWM

1

0 0

10 Time (ms)

20

30

Fig. 7. Modified SVPWM signal in a comparison with conventional SVPWM.

4 Performance Analysis and Comparison The represented VFIMD with modified SVPWM strategy is designed and developed in MATLAB/Simulink environment and corresponding results with the proposed SVPWM technique are described here. 4.1 Dynamic Response of VFIMD

Voltage (V)

Figure 8 shows the dynamic response of the VFIMD with modified SVPWM strategy at different load conditions. The impact of variable voltage, frequency, and torque disturbances are taken into account in a dynamic simulation of an induction motor. At t = 0.2 s, 0.4 s, 0.6 s and 0.8 s, 60%, 80% and 100% of load torque of rated torque are applied and the load is disengaged at other time intervals to obtain the dynamic response. From Fig. 8, we can conclude that the VFIMD with modified SVPWM strategy performs better in terms of all parameter of induction motor.

500 0

Imotor (A)

-500 100

Increase of current

0

Torque (N-m)

Speed (rad/s)

-100 150 Speed changed 0 150 0 -50 0

Torque applied 60%

0.1

0.2

80%

0.3

0.4

60%

100%

0.5

0.6

0.7

0.8

Time (s)

Fig. 8. Dynamic response of VFIMD with modified SVPWM strategy.

0.9

1

A Modified SVPWM Strategy to Improve the Performance

489

4.2 Dynamic Frequency Change Response of VFIMD

Inverter output voltage(V)

The supply frequency is directly coupled with the speed of the induction motor. The speed of the motor (N s ) is proportional with the supply frequency (f s ). Figure 9 shows the dynamic frequency change response of the output voltage variation of the motor driving inverter with the modified SVPWM strategy. 250 200 150 100 50 0 -50 -100 -150

f=10Hz

-200

f=20Hz

f=30Hz

f=50Hz

-250 0

0.2

0.4

0.6

0.8

1

1.2

Time (s)

Fig. 9. Dynamic frequency change response of VFIMD

4.3 Analysis of Power Loss Switching loss (switch turn off loss, switch turn on loss, diode turn off loss) and conduction loss are losses that occur in inverters. Because of the quick conductivity of the forward biased diode, turn-on loss is ignored. This proposed strategy bring down the losses to a reasonable amount utilizing the zero state vectors. The power losses for the conventional SVPWM and modified SVPWM are depicted in Table 3. 4.4 Analysis of Torque Ripple Torque ripple is a phenomena that occurs as the motor starts to rotate and causes a periodic rise or reduction in torque at the output. The torque ripple is defined as, tripple =

tmax − tmin tavg

(12)

where tmax , tmin and tavg represents the maximum, minimum and average torque ripple.

490

N. I. Nahin et al.

Torque ripple in an induction motor is generally unwanted since it creates vibrations and noise, as well as reducing the motor’s lifetime.This modified SVPWM method produces less torque ripple compared to the conventional SVPWM, as a result better performance of induction motor is ensured. The torque ripple for both conventional and modified are depicted in Table 3. Table 3. Comparison of losses for different techniques Switching technique

Conduction loss (% of input power)

Switching loss (% of input power)

Total loss (% of input power)

Torque ripple

Total harmonic distortion (THD)

Conventional SVPWM

0.291

0.282

0.574

11.07%

11.55% (with filter) 52.55% (without filter)

Modified SVPWM

0.197

0.290

0.487

5.38%

7.46% (with filter) 42.46% (without filter)

4.5 Comparative THD Analysis The performance of three phase VFIMD is greatly hampered due to the total harmonic distortion (THD). Figure 10 shows the comparison of Conventional and modified SVPWM in terms of inverter output voltage with and without filter connected. It can be realized that both of the figure are alike and no significance change is observed but the realization of the superiority of the modified SVPWM can be realized in terms of harmonics and losses. Figure 11 and Fig. 12 shows the THD of inverter output voltage using modified and conventional SVPWM with and without using filter.

A Modified SVPWM Strategy to Improve the Performance 500

491

Conventional SVPWM

Amplitude (V)

0 -500 500 0 -500 500 0 -500 0

0.02

0.06

0.04

Time (s) (a)

500

Modified SVPWM

Amplitude (V)

0 -500 500 0 -500 500 0 -500

0

0.02

0.04

Time (s) (b)

0.06

Fig. 10. With and without filter inverter output voltage waveforms with (a) conventional SVPWM and (b) modified SVPWM.

Normalized Mag.

1

Conventional SVPWM

Modified SVPWM

THD= 11.55%

THD= 7.46%

(a)

(b)

.5

0

0

1

2

3

4

5

6

7

Frequency (kHz)

8

9

10

0

1

2

3

4

5

6

7

8

9

10

Frequency (kHz)

Fig. 11. Filtered inverter output voltage THD: (a) conventional SVPWM and (b) modified SVPWM.

492

N. I. Nahin et al.

Fig. 12. Inverter output voltage THD (without filter): (a) conventional SVPWM and (b) modified SVPWM.

5 Conclusion In this work, a modified SVPWM strategy is proposed to improve the performanve of VSI based VFIMD. The presented technique reduces the THD and power loss of the VSI significantly compared to conventional SVPWM. Apart from these, it also reduces the torque ripple of the motor which ensures better performance and smooth dynamic response of the VFIMD. 42.46% voltage THD was recorded at without filter condition with modified SVPWM technique wheareas conventional SVPWM offered 52.55% for the same condition. Moreover, modified SVPWM offered 7.46% voltage THD at filtered condition wheareas conventional SVPWM offered 11.55% for the same condition. 5.38% torque ripple was recorded for the modified SVPWM technique which is also lower than that of convetional SVPWM. The total power loss of the inverter was also reduced for the proposed technique. Thus, the modified SVPWM technique can be considered as a benchmark for VSI based VFIMD. It can also be employed for other power converter applications.

References 1. Huang, Y., Xu, Y., Zhang, W., Zou, J.: Modified single-edge SVPWM technique to reduce the switching losses and increase PWM harmonics frequency for three-phase VSIs. IEEE Trans. Power Electron. 35(10), 10643–10653 (2020) 2. Huang, Y., Xu, Y., Zhang, W., Zou, J.: Hybrid RPWM technique based on modified SVPWM to reduce the PWM acoustic noise. IEEE Trans. Power Electron. 34(6), 5667–5674 (2018) 3. Chinmaya, K.A., Singh, G.K.: Experimental analysis of various space vector pulse width modulation (SVPWM) techniques for dual three-phase induction motor drive. Int. Trans. Electr. Energy Syst. 29(1), 2678 (2019) 4. Ibrahim, Z.B., Hossain, M.L., Bugis, I.B., Mahadi, N.M.N., Hasim, A.S.A.: Simulation investigation of SPWM, THIPWM and SVPWM techniques for three phase voltage source inverter. Int. J. Power Electron. Drive Syst. 4(2), 223 (2014) 5. Zhai, L., Li, H.: Modeling and simulating of SVPWM control system of induction motor in electric vehicle. In: 2008 IEEE International Conference on Automation and Logistics, Qingdao, China, pp. 2026–2030. IEEE (2008) 6. Da Silva, E.R.C., dos Santos, E.C., Jacobina, B.: Pulsewidth modulation strategies. IEEE Ind. Electron. Mag. 5(2), 37–45 (2011) 7. Blasko, V.: A hybrid PWM strategy combining modified space vector and triangle comparison methods. In: 27th Annual IEEE Power Electronics Specialists Conference, Baveno, Italy, pp. 1872–1878. IEEE (1996)

A Modified SVPWM Strategy to Improve the Performance

493

8. Biswas, S.P., Anower, M.S., Sheikh, M.R.I., Islam, M.R., Muttaqi, K.M.: Investigation of the impact of different PWM techniques on rectifier-inverter fed induction motor drive. In: 2020 Australasian Universities Power Engineering Conference (AUPEC), Hobart, Australia, pp. 1–6. IEEE (2020) 9. Islam, M., Biswas, S.P., Anower, M.S., Islam, M.R., Abu-Siada, A.: An advanced modulation technique for power quality improvement in 12-pulse rectifier-inverter fed induction motor drive. In: 2020 Australasian Universities Power Engineering Conference (AUPEC), Hobart, Australia, pp. 1–6. IEEE (2020) 10. Shila, S., Biswas, S.P., Islam, M.R., Rahman, M.M., Shafiullah, G., Sadaba, O.A.: A new PWM scheme to improve the input power quality of 18-pulse rectifier fed 3-level NPC inverter based induction motor drive. In: 31st Australasian Universities Power Engineering Conference (AUPEC), Perth, Australia, pp. 1–6. IEEE (2021) 11. Rony, Z.R., Das, S.C., Khan, M.Z.R.: Space vector modulated PWM generation for motor control systems. In: 10th International Conference on Electrical and Computer Engineering (ICECE), Dhaka, Bangladesh, pp. 149–152. IEEE (2018)

A High Performance Multi-pulse AC-DC Converter for Adjustable Speed Motor Drives Sharmin Shila1 , Shuvra Prokash Biswas1(B) , and Md. Rabiul Islam2 1 Rajshahi University of Engineering and Technology, Rajshahi 6204, Bangladesh

[email protected]

2 University of Wollongong, Wollongong, NSW 2522, Australia

[email protected]

Abstract. This work includes a thorough examination in order to design a novel hexagon-connected autotransformer for a 44-pulse AC-DC converter fed vector controlled induction motor drive (VCIMD). Two paralleled twenty two-pulse AC-DC converters, each with an eleven-phase diode bridge rectifier (DBR), constitute the proposed converter. The system’s performance is evaluated in MATLAB/Simulink environment under various operating conditions of the VCIMD. Traditional 6-pulse converter fed VCIMD suffers from power quality problems. This work intends to alter the design of the suggested autotransformer to make it acceptable for applications that employ a 6-pulse AC-DC converter, in order to minimize the drawbacks of 6-pulse DBR supplied VCIMD. The simulation results show that the power quality characteristics of input AC line current have been improved and are in compliance with the IEEE-519 standard. In order to justify the proposed converter’s efficiency, a comparative study is also presented in comparison to existing configurations. Keywords: Hexagon transformer · AC-DC converter · Power quality · 44-pulse rectifier · Vector controlled induction motor drive

1 Introduction Variable frequency induction motor drives, which are utilized in a variety of applications such as compressors, pumps, air conditioning, rolling mills, and so on, have ramped up as a result of recent improvements in power-electronic converters. Vector control is the most practical means of achieving high-performance control in these induction motor drives when compared to other control techniques. This method is used in voltage source inverters, which are typically fed by six-pulse diode bridge rectifiers (DBR). The six-pulse DBR introduces harmonics in grid current, causing undesirable supply voltage circumstances and lowering power quality. The international standard IEEE519, established in 1981, was reissued in 1992 in order to limit the harmonics fed into the grid by non-linear loads such as vector controlled induction motor drive (VCIMD) [1, 2]. Multi-pulse rectifiers have received much interest because of their ruggedness, reliability, and simplicity to implement. These methods employ two or more converters, © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 494–504, 2022. https://doi.org/10.1007/978-981-19-1742-4_42

A High Performance Multi-pulse AC-DC Converter

495

with harmonics produced by one converter being suppressed by the other converters due to proper phase displacement. When running at light load, the total harmonic distortion (THD) of line current for up to 18-pulse AC-DC converter topologies is more than 5%, which does not fulfill the requirements of IEEE standard. The high rating of the magnetics is also a downside of the aforementioned works. Topologies based on autotransformers, on the other hand, are capable of lowering magnetic ratings. Various designs of autotransformer-connected multi-pulse rectifiers based on pulse doubling, phase shifting, phase multiplication, or a combination of these methods were presented in the literature [3–5]. A polygon-connected 24-pulse autotransformer with an input current THD variation of 4.15% at full load to 7.75% at light load was proposed in [3]. A hexagon-connected 20-pulse autotransformer with an input current THD variation of 4.48% at full load to 5.65% at light load was designed in [4]. In [5], another 36pulse autotransformer with THDs of less than 5% was presented for VCIMD. However, because the voltage of the DC-link is larger than that of a 6-pulse DBR, the topology is unsuitable for retrofit applications. For power quality enhancement, a 44-pulse AC-DC converter using a new hexagonconnected autotransformer is proposed in this work. Two 11-legged DBRs are paralleled via two interphase transformers in this configuration. As a result, the AC line current contains 44 pulses, yielding a 44-pulse output voltage. This work presents a detailed design of the entire IMD system, as well as a model and simulation of the proposed AC-DC converter in the MATLAB/Simulink environment. Different power quality characteristics are obtained and compared to a 6-pulse DBR fed system, including THD of supply voltage and current, distortion factor (DF), power factor (PF), and displacement factor (DPF). is1

a b c

Ld S a1

Sa2

Sa3

Sa5

Sa6

iin

VCA VC

Cd

is2

VAB

VCIMD

VA

VB

VBC 50 Hz, 415 V AC-supply

IPTs Proposed 44-pulse AC-DC converter

Sa4

VSI

Fig. 1. Delta/Hexagon connected autotransformer configuration for 44-pulse AC-DC conversion.

2 Proposed 44-Pulse AC-DC Converter The phase shift angle for a 44-pulse converter system to suppress harmonics can be evaluated as [6]: θ=

360◦ (NC × NP )

(1)

496

S. Shila et al.

where N C is the number of converters, N P is the number of pulse, and θ is the phaseshifted angle. The minimum phase shift in a 44-pulse rectifier made up of two 22-pulse converters is 8.18◦ . The delta/hexagon autotransformer configured 44-pulse AC-DC converter requires a parallel connection between two 22-pulse DBRs with 11-phase shifted voltages each. The phase difference between these two sets of 11-phase voltages is 8.18◦ between the groups and 32.725◦ between the voltages of the same set. Figure 1 depicts the proposed hexagon-connected autotransformer for 44-pulse AC-DC conversion. The proposed autotransformer’s phasor diagram is shown in Fig. 2. Figure 3 illustrates the details of the hexagon transformer’s winding configuration for 44-pulse AC-DC conversion. VA

VA VK2

VA1 VA2

VK1

VJ2

VCA

VJ1

VAB

VB1 VB2

VCA

VC1 VC2

VAB

VD1

VI2 VI1

VC

VBC

VB

VD2

VC VH2 VH1

VB

VBC

VE1 VE2

VG2 V G1

VF2 VF1

Fig. 2. Phasor diagram of delta/hexagon connected transformer for 44-pulse AC-DC conversion.

2.1 Design of the Proposed Transformer The two voltage sets fed to DBRs are referred to as (V A1 , V B1 , V C1 , V D1 , V E1 , V F1 , V G1 , V H1 , V I1 , V J1 , V K1 ) and (V A2 , V B2 , V C2 , V D2 , V E2 , V F2 , V G2 , V H2 , V I2 , V J2 , V K2 ). The phase shift between V A1 and V A2 is 8.18◦ . The phase displacements of V A1 and V A2 from the phase A of input voltages are +4.09◦ and −4.09◦ , respectively. The eleven-phase voltages are generated from AC main line and phase voltages in relation to the turns of the primary winding, as illustrated in the phasor representation in Fig. 2 and the winding arrangement in Fig. 3, which can be expressed as the following equations: The primary winding voltages of the transformer are considered as follows: VA = VS  0◦ , VB = VS  − 120◦ , VC = VS  = +120◦ The eleven-phase voltages applied to the windings are: VA1 = VS  + 4.09◦ , VB1 = VS  − 28.66◦ , VC1 = VS  − 61.385◦ , VD1 = VS  − 94.11◦ , VE1 = VS  − 126.835◦ , VF1 = VS  − 159.56◦ ,

(2)

A High Performance Multi-pulse AC-DC Converter

VG1 = VS  + 167.715◦ , VH1 = VS  + 134.99◦ , VI1 = VS  + 102.265◦ , VJ1 = VS  + 69.54◦ , VK1 = VS  + 36.815◦

497

(3)

VA2 = VS  − 4.09◦ , VB2 = VS  − 36.84◦ , VC2 = VS  − 69.565◦ , VD2 = VS  − 102.29◦ , VE2 = VS  − 135.015◦ , VF2 = VS  − 167.74◦ , VG2 = VS  + 159.535◦ , VH2 = VS  + 126.81◦ , VI2 = VS  + 94.805◦ , VJ2 = VS  + 61.36◦ , VK2 = VS  + 28.635◦

(4)

Input voltages for 22-pulse DBR-I are: VA1 = VA + K1 VC − K2 VB VB1 = VA + K3 VB − K4 VC VC1 = VB + K7 VA − K8 VC VD1 = VB + K11 VA − K12 VC VE1 = VB + K15 VC − K16 VA VF1 = VB + K19 VC − K20 VA VG1 = VC + K21 VB − K22 VA VH1 = VC + K17 VB − K18 VA VI1 = VC + K13 VA − K14 VB VJ1 = VC + K9 VA − K10 VB VK1 = VA + K5 VC − K6 VB

(5)

Input voltages for 22-pulse DBR-II are: VA2 = VA + K1 VB − K2 VC VB2 = VA + K5 VB − K6 VC VC2 = VB + K9 VA − K10 VC VD2 = VB + K13 VA − K14 VC VE2 = VB + K17 VC − K18 VA VF2 = VB + K21 VC − K22 VA VG2 = VC + K19 VB − K20 VA VH2 = VC + K15 VB − K16 VA VI2 = VC + K11 VA − K12 VB VJ2 = VC + K7 VA − K8 VB VK2 = VA + K3 VC − K4 VB VAB =

√ √ √ 3VA  30◦ , VBC = 3VB  30◦ , VCA = 3VC  30◦

(6) (7)

498

S. Shila et al.

VA VA1K2 K1 K2V

VK2 VK1

A2

K4

K4

K6 K5

K3

K5 VC1 K8 K7 VC2 K10 K9

VAB

VCA

VJ1 K7 K10 K

K6 VB2

K3

VJ2 K8

VB1

9

VI2

K11

K12 K11 VI1 K K14 13

K13 K14

VC K16 K15 K17 VH2 K

VBC

18

VH1

K12

K19 K20

K21 K22

VG2 V G1

K21

K19

K22 VF2

K15 K17 K16 K18VE1

VD1

VD2

VB

VE2

K20 VF1

Fig. 3. Winding arrangement of hexagon connected transformer for 44-pulse AC-DC conversion.

2.2 Design of the Transformer for Retrofit Applications The output voltage of a multi-pulse rectifier is much more DC compared to a conventional six-pulse DBR, making it unfit for retrofit applications. The output voltage of the presented 44-pulse rectifier is 20% high compared to a conventional 6-pulse DBR. Modifications to the autotransformer design result in a 20% reduction in output voltage for the proposed transformer. Modifying the tapping places as illustrated in Fig. 4, which results in the desired number of windings. The required phase shift remains the same. The following equations can be used to calculate the new tapping positions, just as they were in Sect. 2, Part 1. Input voltages for 22-pulse DBR-I are: VA1 = VA + K1 VC + K2 VB VB1 = VA + K3 VB − K4 VC VC1 = VB + K7 VA + K8 VC VD1 = VB + K11 VA − K12 VC

A High Performance Multi-pulse AC-DC Converter

499

VE1 = VB + K15 VC + K16 VA VF1 = VB + K19 VC + K20 VA VG1 = VC + K21 VB + K22 VA VH1 = VC + K17 VB + K18 VA VI1 = VC + K13 VA + K14 VB VJ1 = VC + K9 VA + K10 VB VK1 = VA + K5 VC + K6 VB

VA1

VA

VK2 VK1

VA2 K1

VB1 VB2 K4 K3 K6 K5

K2

VJ2

VCA

VJ1

(8)

VAB

K8 VC1 K K10 VC2 7

K9

VI2

K11 K12 VD1 K13 K14 VD2

VI1

K15

VC

VBC

K17 K19 K21

VH2 VH1

K20 VG2

VG1

K22 VF2

VB K16

K18 VE1 VE2

VF1

Fig. 4. Phasor diagram of proposed transformer for 44-pulse AC-DC conversion with modifications for retrofit applications.

Input voltages for 22-pulse DBR-II are: VA2 = VA + K1 VB + K2 VC VB2 = VA + K5 VB + K6 VC VC2 = VB + K9 VA + K10 VC VD2 = VB + K13 VA + K14 VC

500

S. Shila et al.

Fig. 5. MATLAB/Simulink model of the proposed 44-pulse AC-DC converter.

VE2 = VB + K17 VC + K18 VA VF2 = VB + K21 VC + K22 VA VG2 = VC + K19 VB + K20 VA VH2 = VC + K15 VB + K16 VA VI2 = VC + K11 VA − K12 VB VJ2 = VC + K7 VA + K8 VB VK2 = VA + K3 VC − K4 VB

(9)

To achieve the requisite phase shifts and output voltages, the constants K 1 to K 22 specify the secondary winding lengths as a proportion of the primary turns. As we have used the autotransformer in this work, two interphase transformers (IPTs) are connected to the DBRs to ensure that the two 22-pulse DBRs operate independently. The transformer’s kVA rating is calculated as:   Vwinding Iwinding kVA = 0.5 (10) where V winding is the transformer’s winding voltage and I winding is the transformer’s winding current. The apparent power rating of the IPTs is determined in the same method.

A High Performance Multi-pulse AC-DC Converter

501

3 Performance Evaluation of the Presented VCIMD The MATLAB/Simulink environment is used to design the proposed 44-pulse AC-DC converter as depicted in Fig. 5. The 44-pulse converter is supplied from a three-phase 415 V, 50 Hz AC supply. Three multi-winding transformers are utilized to model the designed transformer. The IPT is also modeled using a multi-winding transformer block. A twolevel IGBT-based voltage source inverter (VSI) is connected to the converter output via the DC link consisting of a series inductor (L d ) and capacitor (C d ). Using a sinusoidal pulse width modulation strategy, the firing pulses for the VSI are generated. VSI drives a three-phase squirrel cage winding based induction motor using an indirect vector control technique. The output voltage of the proposed transformer’s simulation model is shown in Fig. 6. 300

Voltage, Vt (V)

200 100 0 -100 -200 -300

0.34

0.344

0.348 Time (s)

0.352

0.356

0.36

Fig. 6. Output voltage of the proposed transformer.

The dynamic response of the presented VCIMD is demonstrated in Fig. 7. A smooth transition from the steady state to a transient state is obtained at around t = 0.16 s. At 0.5 s, when load is varied from 50% of rated torque to 90% of rated torque, there is a change in 3-phase supply current and stator currents of the induction motor. During load variation, it is observed that the IMD responds satisfactorily. The grid current, grid voltage, and motor stator current of the proposed converter are presented in Figs. 8(a), (c), and (e), respectively. The grid current waveform has a THD of 1.43%, as shown in Fig. 8(b). According to harmonic analysis depicted in Fig. 8(d), the grid voltage waveform has 3.02% THD. The stator current comprises only 2.66% THD, according to the harmonic analysis presented in Fig. 8(f). These values are obtained at 100% of the rated load conditions. These waveforms have improved power quality, and THDs are less than 5%, which is in conformity with IEEE-519.

502

S. Shila et al.

Speed (rad/s)

Torque (Nm)

Stator current (A) Grid current (A)

200 100 0 -100 -200 50 0 -50 100 50 0 150 100 50 0 0

0.1

0.2

0.3

0.4 Time (s)

0.5

0.6

0.7

Fig. 7. Dynamic response of the VCIMD.

10

0

-45

THD = 1.43%

8 6 4 2 0

0.49

Stator current (A)

10 THD = 3.02%

8 6 4 2

5 10 15 Harmonic order (b)

20

0

-80

0 0

5 10 15 Harmonic order (d)

20

0.45

0.47 Time (s) (c)

0.49

10

80

Mag (% of 50Hz)

0.47 Time (s) (a)

0

-500

0 0.45

Mag (% of 50Hz)

500 Grid voltage(V)

Mag (% of 50Hz)

Grid current (A)

45

THD = 2.66%

8 6 4 2 0

0.45

0.47 Time (s) (e)

0.49

0

5 10 15 Harmonic order (f)

20

Fig. 8. Analysis of (a) grid current, (b) harmonic spectra of grid current, (c) grid voltage, (d) harmonic spectra of grid voltage, (e) motor stator current, and (f) harmonic spectra of motor stator current.

A High Performance Multi-pulse AC-DC Converter

503

4 Results and Discussion Table 1 depicts the comparison of several performance metrics between the proposed converter and the conventional 6-pulse converter. The THD of supply voltage (V s ) and supply current (I s ) of the proposed converter is less than 5%. At light load (LL), the THD of the 6-pulse converter’s supply current is 61.8%, and at full load (FL), the THD value is 31.3%. As both the supply voltage and supply current of the 6-pulse converter are more than 5%, they do not meet the IEEE requirements. Moreover, it is evident that the power factor using the proposed configuration is closer to unity than the conventional 6-pulse configuration. Therefore, the improved performance metrics of the presented converter make it suitable for retrofit applications. Table 1. Comparative analysis of power quality parameters. Configuration THD of Vs (%)

THD of Is (%)

6-pulse

6.79

61.8

44-pulse

3.02

DF

DPF

PF

LL FL LL FL LL FL LL FL (20%) (100%) (20%) (100%) (20%) (100%) (20%) (100%) 31.3

1.97

1.43

0.843

0.952

0.949

0.973

0.808

0.932

0.997

0.995

0.987

0.979

0.992

0.993

Table 2 presents how large load variations influence the performance metrics of the IMD. With power factors of 0.9924 and 0.9935, the THD of supply voltage (V s ) is 2.43% at light load (LL) and 3.02% at full load (FL). Supply current (I s ) has a THD of 1.97% under light load and 1.43% at full load. THDs of both supply voltage and current satisfy the IEEE-519 standard. Both the displacement factor (DPF) and the distortion factor (DF) are within acceptable limits. The power factor (PF) is close to unity in value. Table 2. Power quality parameters under load variation. Load (%)

THD of Vs (%)

THD of Is (%)

THD of Ist (%)

DF

DPF

PF

20

2.43

1.97

3.94

0.9972

0.9873

0.9924

40

2.76

1.89

3.50

0.9961

0.9854

0.9936

60

2.92

1.73

3.36

0.9965

0.9836

0.9935

80

2.98

1.52

3.05

0.9952

0.9798

0.9932

100

3.02

1.43

2.66

0.9951

0.9792

0.9931

5 Conclusion This work presents a new hexagon-connected transformer with a 44-pulse AC-DC converter fed by VCIMD. The presented converter’s design technique has demonstrated the

504

S. Shila et al.

ability to construct an autotransformer applicable for retrofit applications. The power quality metrics such as THD of the AC main current, THD of the AC main voltage, and power factor have improved employing the proposed converter as compared to the conventional converter. The power factor of the VCIMD is always greater than 0.9924, regardless of the large load variation. As a result, the presented 44-pulse AC-DC converter can efficiently replace the existing 6-pulse converter without requiring major equipment adjustments.

Appendix Specification of IMD: Y-connected, 4-pole, 415 V, 50 Hz, 10 hp (7.5 kW), 3-φ squirrelcage IM. DC-link filter: L d = 2 mH, C d = 2200 μF.

References 1. Islam, M., Biswas, S.P., Anower, M.S., Islam, M.R., Abu-Siada, A.: An advanced modulation technique for power quality improvement in 12-pulse rectifier-inverter fed induction motor drive. In: Australasian Universities Power Engineering Conference (AUPEC), pp. 1–6. IEEE, Hobart (2020) 2. Biswas, S.P., Anower, M.S., Sheikh, M.R.I., Islam, M.R., Muttaqi, K.M.: Investigation of the impact of different PWM techniques on rectifier-inverter fed induction motor drive. In: Australasian Universities Power Engineering Conference (AUPEC), pp. 1–6. IEEE, Hobart (2020) 3. Singh, B., Garg, V., Bhuvaneswari, G.: Polygon-connected autotransformer-based 24-pulse AC–DC converter for vector-controlled induction-motor drives. IEEE Trans Ind Electron 55(1), 197–208 (2008) 4. Abdollahi, R.: Hexagon-connected transformer-based 20-pulse AC–DC converter for power quality improvement. J. Electr. Syst. 8(2), 119–131 (2012) 5. Singh, B., Gairola, S.: Design and development of a 36-pulse AC-DC converter for vector controlled induction motor drive. In: 7th International Conference on Power Electronics and Drive Systems, pp. 694–701. IEEE, Bangkok (2007) 6. Shila, S., Biswas, S.P., Islam, M.R., Rahman, M.M., Shafiullah, G., Sadaba, O.A.: A new PWM scheme to improve the input power quality of 18-pulse rectifier fed 3-level NPC inverter based induction motor drive. In: 31st Australasian Universities Power Engineering Conference (AUPEC), pp. 1–6. IEEE, Perth (2021)

Passivity-Based Control of Tidal Turbine Based PMSG Using Interconnection and Damping Assignment Approach Youcef Belkhier1 , Younes Sahri1 , Thiziri Makhlouf1 , Rabindra Nath Shaw2 , Mohit Bajaj3 , and Ankush Ghosh4(B) 1 Laboratoire de Technologie Industrielle et de l’Information (LTII), Faculté de Technologie,

Université de Bejaia, Bejaia, Algeria {younes.sahri,thiziri.makhlouf}@univ-bejaia.dz 2 Bharath Institute of Higher Education and Research, Chennai, India 3 Department of Electrical and Electronics Engineering, National Institute of Technology Delhi, Narela, India [email protected] 4 The Neotia University, West Bengal, India [email protected]

Abstract. Marine current conversion systems with permanent magnet synchronous generator (PMSG) have several advantages over the renewable energies and is gradually replacing it in the industry. Non-linear equations describe the dynamics of the PMSG. It is subject to unknown external disturbances (load), and its parameters are variable in time. All these constraints make the control task complex. It requires non-linear controls that compensate for non-linearities, external disturbances, and parametric variations. This paper investigates an interconnection and damping assignment passivity-based control (IDA-PBC) h for the PMSG using the model represented in the dq-frame. Inherent advantages of the IDA-PBC method are that the non-linear properties are not canceled but compensated in a damped way. The proposed PBC is responsible for designing the system’s desired dynamic. The efficiency of the suggested technique is investigated numerically using MATLAB/Simulink software. Keywords: Passivity-based control · Tidal turbine · PMSG · Nonlinear control

1 Introduction One of the most promising types of renewable energies is the tidal energy due to its high-power density and high potential of electricity generation. The use of PMSG in tidal turbine system has high potential due to its reliability, increased energy, reduced failure and possibility to eliminate the gearbox which lead to low maintenance and enable to the PMSG to be very favorable in wind applications. However, the controller computation for the PMSG is still challenging work, due to unknown modeling error, external disturbances and parameter uncertainties. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 505–514, 2022. https://doi.org/10.1007/978-981-19-1742-4_43

506

Y. Belkhier et al.

This work provides a novel optimal passivity-based control (PBC). The IDA-PBC technique has the intrinsic advantage of compensating for non-linear features rather than canceling them. The proposed PBC is in charge of creating the system’s intended dynamic, whereas the non-linear observer is in charge of reconstructing the recorded signals. In order to compel the PMSG to track speed [3]. The fundamental goal of this research is to synthesis the controller while taking into consideration the whole dynamic of the PMSG and making the system passive. It is accomplished by reorganizing the suggested strategy’s energy and incorporating a damping component that treats the non-linear parts in a damped rather than deleted manner. The present form organizes the present paper, the detailed about the case study is given in Sect. 2, and along the Sect. 3 we describe the suggested controller. Section 4 is devoted to GSC regulation. Section 5 reports the simulation results and their discussions by the application of IDA-PBC algorithm to model illustrated in Sect. 2. This paper has been completed by some conclusions and perspectives stated in Sect. 6.

2 System Design 2.1 Marine Current Turbine Model The configuration of the investigated conversion system with Matlab/Simulink is presented in Fig. 1. The proposed strategy is applied to the PMSG to regulates the produced power via the generator [4]: Pm =

1 ρCp (β, λ)Av3 2

(1)

Pm ωm

(2)

Tm =

   21  1 116 − − 0.4β − 5 e λi Cp (β, λ) = 2 λi −1  −1 3 λ−1 = (λ + 0.08β) − 0.035 1 + β i

(3) (4)

where, v denotes the tidal speed, β denotes the pitch angle, ωm denotes the generator speed, R denotes the radius of the blades, ρ is fluid’s density, Cp represents the power coefficient, A is blades area, and λ = ωVm R λ denotes the tip-speed ratio.

Passivity-Based Control of Tidal Turbine

507

2.2 Model of the PMSG Model of the PMSG is as below [3–5]:

  vdq = Rdq idq + Ldq ˙idq + pωm  Ldq idq + ψf

(5)

J ω˙ m = Tm − Te − ffv ωm

(6)

Te = 



3 pψdq idq 2

(7)

Rs 0 denotes the stator resistance matrix, ff v represents the coefficient 0 Rs  ϕf is the flux linkages vector, Te ffv represents the of the viscous friction, ψf = 0  Ld 0 denotes the induction matrix of the stator, J electromagnetic torque, Ldq = 0 Lq  vd denotes voltage stator vector, idq = represents the total inertia moment, vdq = vq  id denotes the stator current vector, and ωm denotes the PMSM speed. ϕf are the flux iq  0 −1 linkages due to the permanent magnets, p is the number of pole-pairs, and  = . 1 0 where Rdq =

Wind turbine

PMSG

Back-to-back Converter Filter

D C

A C

Transformer

A C

D C

Grid

Control PBVC

PI

Fig. 1. Tidal conversion system.

3 IDA-PBC Controller Design As stated in Sect. 1, the behavior of the tidal turbine-based PMSG is a nonlinear mathematical issue, and the conversion system needs an optimal improved energy harvesting to increase operating efficiency. The MSC purpose is to communicate the energy generated by the turbine with as little loss as possible.

508

Y. Belkhier et al.

3.1 IDA-PBC Theory As shown below, IDA-PBC was established as a mechanism for controlling physical systems defined by the PCH model [3, 6]:

x˙ = ∂H ∂x (x)[J (x) − R(x)] + gu (8) y = g T (x) ∂H ∂x (x) where, H (x) : n → + is the total stored energy, x ∈ n represent the state vector, u ∈ m , m < n is the controller action, the product of u and y ∈ m has units power, and J (x) = −J T (x), R = RT (x) ≥ 0 are the matrix of interconnection and dissipation, respectively. The PMSG model (5)–(7) can then expressed in PCH form as: ⎡ ⎤ ⎡ ⎤ ⎤ ⎡ id −1 0 0 Rs 0 0 g = ⎣ 0 −1 0 ⎦, R(x) = ⎣ 0 Rs 0 ⎦, x = [x1 x2 x3 ]T = D⎣ iq ⎦ is the ωm 0 0 0 0 0 −1 ⎡ ⎤ v d   state vector, D = diag Ld , Lq, JT , u = ⎣ vq ⎦ is the input vector which repre−Tm   sent the voltage controller vector of the PMSG, ∂H (x) = id iq ωm is the output vector ∂x ⎡ ⎤ 0 0  x2  ⎦. The controller aim is reduced to find the and J (x) = ⎣ 0  0  − x1 − φf 0 −x2 x1 − φf following equation u(x) [6]: 

 −1 ∂Hd ∂H (9) u = gT g gT × (x)[Jd (x) − Rd (x)] − [J (x) − R(x)] (x) ∂x ∂x

3.2 Controller Design The proposed strategy’s objective is to design the desired Hamiltonian energy function based on PCH theory, which is given as follow [3]:  2 2 1 p 1 1 2 Hd (x) = (10) x2 − x2∗ + x2 − x2∗ x1 + 2 Ld Lq JT where, the state vector and the desired state are defined as: ⎤ ⎡ ∗⎤ ⎡ 0 x1 ⎢ Lq Tm∗ ⎥ x∗ = ⎣ x2∗ ⎦ = ⎣ pφ ⎦ f ∗ x3∗ ωm

(11)

∗ , T ∗ are the desired speed and desired torque which are the tidal turbine speed where, ωm m and torque, respectively in our case. We have x1∗ = 0 and we select: ⎡ ⎤ r1 0 0 Rd (x) = ⎣ 0 r2 0 ⎦ 0 0 0

Passivity-Based Control of Tidal Turbine

509

⎤ 0 −J12 J13 Jd (x) = ⎣ J12 0 −J23 ⎦ −J13 J23 0 ⎡

where, the objective is to design the parameters J12 , J13 , J23 , r1 > 0 and r2 > 0 which allows to the controller to guarantee that the speed tracking error is converged. From [6], the controller equation u(x) can be solved with: J13 =

pLd x3 , JT

J13 = −

pLd x2∗ , Lq

J23 = −

pLq x1∗ Ld

Thus, the controller expression become as:   ∗ vd = −id − pLd iq∗ ωm + id∗ (Rs − r1 ) + iq − iq∗ pLd ωm

(12)

  ∗   vq = −r2 iq + pLq id∗ + (Rs − r1 )iq∗ + p Ld id + φf − pLq id∗ ωm

(13)

4 PI Controller for Grid-Side To regulate and transmit to electrical energy produced by the PMSG to the grid through the GSC, a classical method is selected which consists on PI strategy. The GSC mathematical model is expressed as follows:     Lf ˙idf − ωLf iqf Vgd idf Vid = + + R (14) f Viq Lf ˙iqf − ωLf idf Vgq iqf C V˙ dc = idc + idf

3 vgd 2 Vdc

(15)

where, Vgd and Vgq are the grid voltages, idf and iqf are the grid currents, Rf represents the filter resistance Vid and Viq denotes the inverter voltages, CC is the DC-link capacitance, Lf is the filter’s inductance, idc is the line current, ω represents the grid angular frequency, and Vdc is the DC-link voltage. Finally, the mathematical model of the active and reactive powers is formulated as below [9–14]: ⎧ 3 ⎪ ⎨ Pg = vgd idf 2 (16) ⎪ ⎩ Q = 3v i g gd qf 2

510

Y. Belkhier et al.

5 Simulation Results The results of the simulations were obtained using MATLAB/Simulink in order to verify the suggested method’s performance and dependability. Table 1 lists the parameters that were utilized to simulate the conversion system. From the pole placement method, MSC current PI gains are K gp = 100 and K gi = 500. The DC voltage PI gains are K dcp = 5 and K dci = 500. The parameters r 1 = 7Rs and r 2 = 10Rs. The marine current speed dynamic employed for the simulation fluctuation is shown in Fig. 2. The torque is seen in Fig. 3. Once opposed to the DC-link behavior of that in [15], the produced with IDAPBC is extraordinarily well stabilized around the set point value, as illustrated in Fig. 4, with a rapid convergence and low inaccuracy. Figures 5 and 6 indicate that only active power is provided to the electricity network, whilst reactive power generated is severely limited and well-kept at its rated value, and Fig. 7 illustrates that the control operation accomplishes good sinusoidal grid side permeability with minimal overflow. Table 1. System parameters Parameter

Symbol

Value

Grid filter inductance

Lf

0.3 pu

Grid filter resistance

Rf

0.3 pu

Grid voltage

Vg

575 V

DC-link voltage

Vdc

1150 V

DC-link capacitor

C

2.6 F

Viscosity friction

Jfv

0.01 N.m.s

Total inertia

J

35000 kg. m2

Flux linkage

ψ

1.48 wb

Pole pairs

p

48

Stator inductance

Ldq

0.3 mH

Stator resistance

Rs

0.006

Water density

ρ

1024 kg/m2

Tidal turbine radius

R

10 m

Passivity-Based Control of Tidal Turbine

Tidal speed (m/s)

15

10

5

0 0

2

4

6

Time (s)

Fig. 2. Tidal speed.

Fig. 3. Electromagnetic torque.

Fig. 4. DC-link voltage.

8

10

511

512

Y. Belkhier et al.

Fig. 5. Active and reactive power.

Fig. 6. Zoom of reactive power.

Fig. 7. Grid transfer red voltage

Passivity-Based Control of Tidal Turbine

513

6 Conclusion In this paper, a new IDA-PBC using the PCH model based on the concept of passivity has been applied to the PMSG. This controller exploits the PCH model of the motor, which highlights three matrices: the interconnection matrix, which represents the internal energy exchange ports between the states of the PMSG, the damping matrix, which represents all the dissipation elements of the system, and the external interconnection matrix which represents the energy exchanges of the PMSG with its external environment. The particular characteristic of the IDA-PBC is the choice of the PCH structure in CL, then the energy function compatible with this model is determined. The paper’s proposed technique achieves the paper’s aims. The PMSG-based conversion system performs well and efficiently. The structure of the control approach is sensible and easy.

References 1. Dai, Y., Ren, Z., Wang, K., Li, W., Li, Z., Yan, W.: Optimal sizing and arrangement of tidal current farm. IEEE Trans. Sustain. Energy 9(1), 168–177 (2018) 2. Qian, P., Feng, B., Liu, H., Tian, X., Si, Y., Zhang, D.: Review on configuration and control methods of tidal current turbines. Renew. Sust. Energ. Rev. 108, 125–139 (2019) 3. Santos, G.V., Cupertino, A.F., Mendes, A.F., Junior, S.I.S.: Interconnection and damping assignment passivity-based control of a PMSG based wind turbine for maximum power tracking. In: 2015 IEEE 24th International Symposium on Industrial Electronics (ISIE), Buzios, pp. 306–311 (2015) 4. Youness, E., Aziz, D., Abdelaziz, E., Othmane, Z., Najib, E.: Nonlinear back stepping control of variable speed wind turbine based on permanent magnet synchronous generator. In: 2019 IEEE International Conference on Wireless Technologies, Embedded and Intelligent Systems (WITS), Fez, Morocco (2019) 5. Belkhier, Y., Achour, A.Y., Hamoudi, F., Ullah, F., Mendil, B.: Robust energy-based nonlinear observer and voltage control for grid-connected permanent magnet synchronous generator in tidal energy conversion system. Int. J. Energy Res. 1–19 (2021). https://doi.org/10.1002/er. 6650 6. Akrad, A., Hilairet, M., Ortega, R., Diallo, D.: Interconnection and damping assignment approach for reliable PM synchronous motor control. In: 2007 IET Colloquium on Reliability in Electromagnetic Systems, Paris (2007) 7. Yang, B., Yu, H., Zhang, Y., Chen, J., Sang, Y., Jing, L.: Passivity-based sliding-mode control design for optimal power extraction of a PMSG based variable speed wind turbine. Renew. Energy 119, 577–589 (2018) 8. Subramaniam, R., Joo, Y.H.: Passivity-based fuzzy ISMC for wind energy conversion systems with PMSG,” IEEE Trans. Syst., Man, Cyber. Syst. 119, 577–589 (2019). https://doi.org/10. 1109/TSMC.2019.2930743 9. Belkhier, Y., Achour, A., Shaw, R.N., Sahraoui, W., Ghosh, A.: Adaptive linear feedback energy-based back stepping and PID control strategy for PMSG driven by a grid-connected wind turbine. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 177–189. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0749-3_13 10. Belkhier, Y., Achour, A.Y.: Fuzzy passivity-based linear feedback current controller approach for PMSG-based tidal turbine. Ocean Eng. 218, 108156 (2020). https://doi.org/10.1016/j.oce aneng.2020.108156

514

Y. Belkhier et al.

11. Belkhier, Y., Achour, A.Y.: An intelligent passivity-based back stepping approach for optimal control for grid-connecting permanent magnet synchronous generator-based tidal conversion system. Int. J. Energy Res. 45, 5433–5448 (2020). https://doi.org/10.1002/er.6171 12. Belkhier, Y., Achour, A.Y., Shaw, R.N., Ullah, N., Chowdhury, M.D.S., Techato, K.: Energybased combined nonlinear observer and voltage controller for a PMSG using fuzzy supervisor high order sliding mode in a marine current power system. Sustainability 13(7), 3737 (2021). https://doi.org/10.3390/su13073737 13. Belkhier, Y., Achour, A.: Passivity-based voltage controller for tidal energy conversion system with permanent magnet synchronous generator. Int. J. Control Autom. Syst. 19(2), 988–998 (2020). https://doi.org/10.1007/s12555-019-0938-z 14. Belkhier, Y., et al.: Intelligent energy-based modified super twisting algorithm and factional order PID control for performance improvement of PMSG dedicated to tidal power system. IEEE Access 9, 57414–57425 (2021)

Recommendation System Based on EEG Emotion Recognition R. Vasanthradevi1 , R. Priyadharshini1 , P. Jai Rajesh1(B) , R. Reena2 , and R. Kalpana2 1 Bharath Institute of Higher Education and Research, Chennai, India

[email protected] 2 Prince Shri Venkateshwara Padmavathy Engineering College, Chennai, India

Abstract. On the internet, recommendation systems are essential for filtering options from the vast amount of data accessible and recommending the best product to the right customers. The recommendation system got more accurate day by day. This study seeks to extend the selection criteria of a recommendation system by sensing users’ mental states and recommending options accordingly. A system may readily determine our state of mind by monitoring our behaviors and context in a smart and effective method. When we are nervous, this allows the system to recommend films, books, and activities such as deep breathing. This emotion data is received as supplemental data by any existing recommender system. Thus, this information may be used to improve the performance of existing recommendation engines. Therefore, in this study, the emotion detection is seen as predicting arousal and valence based on multi-channel physiological inputs. The outcomes of the experiment were obtained from 28 participants. Electroencephalographic sensors (EEG) are used to track this user’s emotions, and text tags are gathered and suggestions are generated based on their brain activity. Keywords: Electroencephalography · Emotion detection · Time features · Frequency feature · Time-frequency feature · Signal processing · Daubechies4 · KNN classifier

1 Introduction The recommendation engine uses machine learning techniques to filter data and provide the most relevant choices to the user. However, most traditional recommendation algorithms ignore human emotions. It focuses solely on the interests and preferences of the users. The goal of this research is to produce new inputs by combining emotion detection algorithms with wearable technologies. Wearable technologies, also known as “wearables,” are electronic devices that are worn by people and are frequently used to track or monitor their health. We can detect brain waves and identify emotions using built-in EEG sensors with wearables. Human-Computer Interaction (HCI) is important for detecting, analyzing, and providing feedback based on human emotions, and many research have focused on the interaction between people and computers. Speech, sound, brain activity, eye tracking, © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 515–532, 2022. https://doi.org/10.1007/978-981-19-1742-4_44

516

R. Vasanthradevi et al.

and facial expression may be used to track emotions Electroencephalography (EEG) is used to track signals from the brain system in case of brain activity. As a result, in this study, the EEG signal is used to recognize human emotion, and our recommendation system may provide recommendations depending on the emotion.

2 Literature Survey Ekman [1] was the first to propose the idea of universal underlying emotions that cut across civilizations. Russell [2] offered a framework in whicharousal and valence were assigned to emotions (How active or receptive is the person?). Plutchik’s emotion tree [4] and emotions wheel [3] are two more examples. Pleasure, sorrow, anger, fear, trust, contempt, surprise, and anticipation are just a few of the basic emotions that have been studied extensively. The Kolmgorov complexity and associated aspects of intracranial EEG recording were studied by Petrosian and Arthur [5]. H. Altay Guvenir and Aynur Akkus introduce the SFA (Single Feature Accuracy) weight learning method. Because the SFA method isn’t tied to WkNNFP, it may be utilised with a variety of different classification techniques [6]. The survey of word processor usage by blind computer users reported by Garrett et al. found that many blind persons do not utilise word processor facilities including spell checks, grammar checkers, and templates [7]. Robert Horlings et al. suggested a technique for analyzing EEG data and categorizing them into five groups mainly in two emotional aspects: Arousal and valence. Its purpose is to evaluate the quality of emotion identification in practice using EEG data [8]. Subha et al. used modern signal processing techniques to retrieve data for diagnosis and monitoring illnesses [9]. Koelstra et al. created are pository for analyzing spontaneous emotions. The repository contains 32 people (and 22 frontal face videos). Each participant rated 40 music videos on valence, arousal, dominance metrics, along with their satisfaction and acquaintance [10]. Wijeratne and Perera looked examined how EEG signals and facial expressions behaved in various mental states [11]. Duan et al. presented a new effective EEG feature to reflect the properties related with emotional states. According to the findings, combination of DE and its symmetrical electrodes outperform the ES feature [12]. A two-layer approach for identifying emotions in music is presented by Chin et al. Angry, joyful, sad, and tranquil are the goal emotion classifications. According to the authors, the system performs well in music emotion categorization. It trains secondlayer SVMs for categorizing the four target emotions using support vector machines (SVMs) [13]. Zheng et al. have shown that EEG and pupil width are effective indicators for recognizing emotions [14]. Edgar D. Klenske et al. demonstrated that utilizing the maximum a posteriori point estimate, Hyper - parameter evaluation may be conducted online. This GP model’s predictions are then applied to a predictive control architecture [15]. The performance of the KNN classifier to categorize distinct emotions was detailed by Vaishnavi L. Kaundanya, Anita Patil, and Ashish Panat [16]. Isinkaye and colleagues investigate the features and potentials of several prediction approaches in recommendation systems [17]. Sarno et al. offer a parametric, generic, and successful real-time emotion categorization of electroencephalography (EEG) data. The Fourier Transform, Features Extraction, and K-Nearest Neighbors are used in signal processing, evaluation, and categorization [18]. For EEG emotion classification, Ackermann et al. investigated

Recommendation System Based on EEG Emotion Recognition

517

the implementation of state-of-the-art for extracting features, selecting features, and for classification [19]. Ghare and Paithane designed a user interface for an algorithm that recognizes human emotions automatically [20]. To simulate the link between human emotion and brain processes, Liu et al. applied machine learning-based approaches [21]. Iyer et al. present EmoPlayer, an Android application that generates a playlist depending on the user’s current emotions. The technology uses a camera to collect the user’s image and detects their face. It then recognizes the sentiment and compiles a playlist of music that will gradually improve his mood as the songs continue to play [22]. Bazgir et al. used EEG data to construct an emotion detection system using a valence/arousal paradigm [23]. Ayata et al. presenta music recommender system driven by emotion that uses portable physiological detectors like galvanic skin response (GSR) and photo plethysmography to learn about a user’s emotion (PPG) [24]. Using brain signals, Asif et al. investigated the categorization of human stress in reaction to music [25]. Athavipach, Chanavit, and Pan-Ngum research on a single-channel, low-cost, dry contact, in-ear EEG that can be used for non-intrusive monitoring. The results are equivalent to those obtained using older EEG headsets [26]. Sunil R. Hirekhan, et al. looked at the Detrended Fluctuation Analysis (DFA) of EEG signals before and after meditation (mindfulness) therapies, and discovered that the DFA values of EEG data collected from 8 out of 11 patients reduced [27]. Mukalov et al. choose Natural language processing (NLP) as the best choice for solving the article auto-tagging problem because it is less laborious than traditional ways of learning with a teacher and has the benefit of low laboriousness of text corpus preparation [28]. James et al. created a Facial Expression Depending Music Player that scans and interprets facial expression data before producing a playlist based on the parameters specified. This reduces the time-consuming and tiresome effort of manual categorization or dividing music into separate lists and assists in the creation of a suitable playlist depending on a person’s emotional characteristics [29]. The aim of Alhalaseh, Rania, and Alasasfeh is to provide an automated methodology for detecting emotions from EEG data. The extraction techniques employed, the feature selection algorithm, and the classification procedure all have an impact on the system’s efficiency [30]. Li, Zina et al. suggested MLDW-PSO feature selection increases EEG-based emotion identification accuracy [31]. Florence, S. Metilda, and Uma, M. did emotion recognition using facial expressions [32]. Dadebayev, Didar, et al. studied the state of mainstream consumer EEG gadgets during the last five years and reviewed pertinent research that tested the efficiency of such low-cost equipments for emotion identification [33]. Aamir Arsalan and Muhammad Majid gave the state-trait anxiety inventory questionnaire is collecting anxiety ratings. The EEG data of 65 subjects was obtained for two minutes while they were awake. The integrated noise reduction mechanism of the MUSE EEG headband is used to pre-process the collected EEG data [34]. All of these papers have explained how to identify emotion from EEG data and how to use emotion identification to develop recommendation systems. The goal of this research is to develop an EEG-based emotion-detecting recommendation system and to increases the reliability of emotion-based recommender system.

518

R. Vasanthradevi et al.

3 Existing System Emotions have a significant role in people’s daily lives. Computer-assisted teaching, Human-robot interaction, emotion-aware videogames are only few of the uses. For emotion identification, speech analysis and facial expressions have been used. However, in situations where people seek to disguise their feelings, relying just on verbal or facial expression indications to accurately determine emotions may be insufficient. Rather than visible expressions, physiological responses are much more efficient way to track and detect people’s emotion and internal thought abilities. To identify different emotional states in the valence-arousal plane, several physiological data such as EMG, electrocardiogram (ECG), GSR, and breathing changes were obtained. Different temporal and frequency domain characteristics were retrieved, and classification accuracies were used to demonstrate their efficacy. The analytics of physiological parameters which includes GSR and PPG was used to create a music recommendation system that had an accuracy rate of 70.93% for arousal prediction and 70.76% for valence prediction, respectively. The objectives of this article is to use EEG to anticipate emotions and improve the accuracy of emotion recognition in recommendation systems.

4 Proposed System The suggested architecture comprises employing EEG electrodes to read the user’s brainwave through a smart wearable device, then using the K-Nearest Neighbor (KNN) classifier to increase the accuracy of the suggestions by monitoring the user’s instinctive attachment from these signals. The whole technique is visualized in Fig. 1. Because the emotional effects of the same musical track may differ between users, the system saves the emotional impacts of past suggestions in its database and uses them in future recommendations. 4.1 Algorithm for Proposed System 1. 2. 3. 4.

Get signal data from EEG electrodes. Sample and Extract Features from the data. Arousal and Valence levels are used to predict target emotion labels. Combine user profile and emotion label and Feed the Predicted emotion label to the recommendation system. 5. Get recommendation and send to the user.

5 Materials and Methods The proposed architecture makes use of electroencephalography (EEG) sensors to detect brain waves, allowing for synchronization and simultaneous EEG recording. Data collection, pre-processing, extracting features, and categorization are all part of the process. Finally, the user is offered dataset recommendations based on the feedback from the classifier.

Recommendation System Based on EEG Emotion Recognition

519

Fig. 1. Components of recommendation system and data flow

5.1 Electroencephalography The electroencephalogram (EEG) is a technique for measuring and recording information regarding brain electrical activity. Brain cell impulses, both spontaneous and rhythmic, make up EEG signals. In neuroscience and psychology, EEG signals are hypothesized to be able to describe emotional brain states and human behavior. On the other hand, EEG signals are feeble and difficult to capture electrooculography (EOG), Electrocardiography (ECG), and electromyography (EMG) are examples of physiological signals that might easily interfere with them. As a result, EEG signals have a quasi and chaotic structure. Raw EEG data is frequently subjected to denoising and pre-processing. 5.2 Emotiv “Emotiv is one of the most well-known makers of EEG devices for the public. EPOC+ was designed mainly for academic use. The electrical activity of the cerebral cortex is tracked using 14 electrodes. Wireless communication through Bluetooth, improved signal strength, a compact design with spinning headband, movement detectors which monitor head motions, and long-lasting saline-based electrode are just a few of the features.” EMOTIV EPOC+ 14-Channel Wireless EEG Headset - EMOTIV (Fig. 2). 5.3 Emotion Recognition from EEG Signal An EEG device must be worn by all individuals. The test signals (audio/video) are presented to the subjects, and the voltage changes in their brains are recorded. Signals from a group of neurons when they’re active at the same moment are captured by EEG, not the activity of individual neurons. It mostly collects data from the brain’s small sections that surround each electrode. The EEG depicts electrical activity of the brain

520

R. Vasanthradevi et al.

(a)

(b)

Fig. 2. A) Emotiv Epoc+ Head band. B) Electrode positioning

in the form of waves in various frequencies and amplitudes. The collected EEG data will next be pre-processed to reduce external distortions and interferences, such as noise abatement and temporal and spatial screening. After evaluating the data that is cleaned, a feature extraction method is used. 5.4 Feature Extraction The Fast Fourier Transform (FFT) is just a signal processing technique for changing the frequency domain of an EEG signal. EEG data are gathered as frequency bands using FFT and include delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (30–50 Hz). The data for this investigation was acquired from the frequency domain, time domain, Time-frequency domain, and multi electrode. The accuracy of each feature in predicting emotion was calculated using the KNN classifier. 5.4.1 Time Domain Features i) Mean Mean is a statistical feature which is used to characterize EEG time series. For calculating mean NumPy library is used. NumPy. Mean will return Arithmetic mean of the time series which uses the (1) formula. mean : μ =

1 T S(t) t=1 T

(1)

ii) Non-Statistical Features • Detrended Fluctuation Analysis (DFA) DFA is a non-statistical feature that is used to assess each time series’ neural functioning. To begin, compute the average value of the time series (y(n)) and divide the whole time series into w equal windows, removing remainders, Therefore, each window contains s = int(N/w) time points. The total number of points in the time series is N, and the window size is s. Within each window, a straight

Recommendation System Based on EEG Emotion Recognition

521

line does a least square fit to determine the window’s local trend. The intrinsic fluctuation F(s) variance is determined from the trend line in each window. It is the measure of a window’s local detrended fluctuation.  2 1 ws {y(n)} − yw (n) F(s) = (2) (w−1)s+1 s The square root of the average fluctuation of variance of intrinsic fluctuation for overall windows is the RMS fluctuation from the local trend in each window and window size. It’s the outcome of DFA.  1 w FDFA (s) = F(s) (3) w=1 w • Petrosian Fractal Dimension Primarily, the time series should be transformed into a binary sequence, Petrosian’s algorithm, Petrosian may be employed to compute the FD of a signalquickly. PFD in terms of time series is defined as, PFD =

log10 P log10 P + log10 (P/(P + 0.4Psc ))

(4)

where PSc is the number of sign changes, and P is the series length in the signal derivative. PFD is a scalar feature. • Hjorth Parameters Hjorth developed mobility and complexity of a time series which is used in this study. This is a statistical feature computed as, Hjorth mobility in terms of time series(s(x)) [ x1 , x2 ......, xn ], is defined as,  variance(s(x)) (5) Mobility = variance(s(x)) Hjorth complexity is defined as,  Complexity =

mean(s(x)) mean(s(x))

(6)

• Hurst Exponent The Hurst exponent is a metric for signal long-memory qualities. Rescaled range statistics (R/S) is another name for the hurst exponent. The accumulated deviation from the mean of time series within range T is the first step in calculating the hurst exponent for time series., X = [x1 , x2 ......, xn ], X (t, T ) =

t i=1

(xi − x), where x =

1 T

T i=1

xi , t ∈ [1...N ].

(7)

Then, R(T)/S(T) is calculated by, max(X (t, T )) − min(X (t, T )) R(T ) =   S(T ) (1/T ) Tt=1 [x(t) − x]2

(8)

522

R. Vasanthradevi et al.

Calculating the slope of the line created by ln(R(n)/S(n)) versus ln(n) for n ∈ [2 … N]. Hurst exponent has a scalar characteristics. 5.4.2 Frequency Domain Features • Relative Intensity Ratio (RIR) and Power Spectral Intensity (PSI) The EEG signals were described in spectral space using PSI and RIR. For a given data set, PSI determines the intensity of each frequency band. The FFT (Fast Fourier Transform) is a mathematical transformation that (FFT){F1, F2 … Fn} is obtained for the data. And the frequency boundaries is defined {f1, f2, … fm}. So, the PSI of k bands is calculated by, int (nxfk+1 /R) |Fi | PSIK = i=int(nxfk /R)

And RIR is the density of PSI, defined as. RIR = PSI /

k−1 i=1

PSIi

Where k ∈ {1, 2, ..., m − 1} and R is the sampling rate. Figure 3 depicted the power spectrum. • Spectral Entropy The spectral entropy is defined as follows, 1 K RIRi log RIRi , H =− i=1 log(k) Where RIRi and k = 1, 2, …, K − 1. Spectral entropy is a scalar feature. Figure 4 depicted the power spectrum with respect to the frequency. 5.4.3 Time-Frequency Domain Feature • Discrete Wavelet Transform (DWT) The discrete wavelet transform performs signal decomposition and as a result multiple approximation and detail levels that correspond to varying frequencies while keeping the signal’s temporal information is obtained. The root mean square (RMS), abs (log (REE)), log (REE), and its recursive energy efficiency (REE) wavelet functions are used to extract the DWT feature.

j 

Di (n)2 RMS(j) =  i=1 j i=1 ni Where Di are the detail coefficients, ni the total number of Di at the ith decomposition level, and j denotes the number of decomposition levels. REE =

Eband Etotal−3b

Where Eband is the energy of a sub band, and the total energy of sub bands Etotal−3b = Eα+ Eβ+ Eγ .

Recommendation System Based on EEG Emotion Recognition

Frequency (HZ) Fig. 3. Power spectrum analysis using FFT

Frequency (HZ) Fig. 4. Power spectrum analysis using Welch’s periodogram

523

524

R. Vasanthradevi et al.

5.4.4 Multi Electrode Features It is essential to consider the relationship between the various areas of the brain. Left

AF3

F3

F7

FC5

O1

P7

T7

Right

AF4

F4

F8

FC6

O2

P8

T8

7 pairs of asymmetry electrodes

• Differential Asymmetry The most often used are differences in energy spectrum of corresponding pairs of electrodes, calculated as the difference of two attributes. x = x1 − xr , The symmetric pairs of electrodes on the left and right hemispheres of the scalp are denoted by (l, r).The differential asymmetry of 7 pairs of electrodes listed in the table are obtained in this study. • Magnitude Squared Coherence Estimate (MSCE) This property denotes the correspondence of two signals at each frequency, with values ranging from 0 to 1. It is defined as Cij (f ) =

Pij (f ) 2 Pi (f )Pj (f )

,

Where Pij is the cross power spectral density and Pi and Pj are the power spectral density of two signals si and si . In order to reduce the large number of features resulting from all possible combinations of electrodes, Cij is averaged over the frequency bands. Figure 5 depicted the coherence with respect to each frequency bands. 5.4.5 Linear Discriminant Analysis Linear Discriminant Analysis (LDA) is a supervised machine learning technique that can be used to avoid overfitting, which occurs when a machine learning model remembers all the training data, including undesirable or unrelated features, and the model’s performance plummets when a new dataset is used. The excessively complicated model is to reason for this overfitting. Using feature selection methods, the model may be simplified. The features that are most important for categorization should be employed. Dimension reduction is vital for reducing the model’s complexity. The high-dimensional dataset is projected onto a lower-dimensional space using Linear Discriminant Analysis. LDA helps to minimize the dataset’s dimensionality while retaining class differentiation. After determining the mean of each class, LDA creates a new dimension, which is an axis. The axis is generated by considering two parameters: the distance between the means and the variation. The gap between the means of the classes should be the greatest possible, and the variance or dispersion of the data points should be the least possible.

Recommendation System Based on EEG Emotion Recognition

525

Fig. 5. Representation of Magnitude Squared Coherence with respect to Frequency bands.

5.4.6 Signal Classification The K-nearest neighbor method is one of most basic non-parametric supervised machine learning algorithms for solving problems like regression and classification. A classification problem produces a discrete result, but a regression problem produces a real value. The number of training data points that will be analyzed for the categorization of data points to a given class is represented by ‘K’ in K-nearest neighbor. The distance between each data point and the training data points is determined. Different techniques are used to find the closest point, such as: 1. Euclidean Distance



k i=1

(xi − yi )2

Using the formula provided above, the distance between two data points, say x and y, is calculated. 2. Manhattan Distance k |xi − yi | i=1

Using the formula provided above, the distance between two points when measured at right angles along axes is calculated. 3. Minkowski Distance K (|xi − yi )q )1/q ( i=1

It can be considered of as a generalization of both the Euclidean and Manhattan distances.

526

R. Vasanthradevi et al.

These distances between test data points and the K nearest training data points are kept in a list, and each class is subsequently voted on based on the distance measured. A majority vote of its neighbor’s classifies an item, and the object is assigned to the class with the most votes. The Euclidean distance was utilized to calculate the distances between test data points and K closest training data points. 5.4.7 Russell Circumplex Model Different emotions are classified to identify them from one another. James Russell created the circumplex model of emotion. The Russell Circumplex model explains how many emotions are connected to one another. Within a visual framework, this model demonstrates a unique relationship. A two-dimensional circular model was utilized to depict the whole range of emotion. Emotions were separated into quadrants, with arousal and valence acting as a crossing axis. Arousal is depicted on the vertical axis, whereas valence is expressed on the horizontal axis. Emotional states are represented as a mix of arousal and valence in this circular model. Figure 6 visualize the emotions with respect to Arousal and valence in a chart.

Fig. 6. Representation of two-dimensions of-emotions Valence-negative-positive and arousal low-high

Recommendation System Based on EEG Emotion Recognition

527

6 Experiments and Results 6.1 Dataset Experiments were conducted using the Dataset provided by Talha Alakus, Murat Gonen, and Ibrahum Turkoglu, which included EEG data collected from 28 people while playing computer games using the 14 channel Emotivepoc+ wearable and portable EEG equipment [35]. With a total of 20 min of EEG data available for each participant, they played four distinct emotional computer games for five minutes each (boring, calm, scary, and humorous). Subjects rated each computer game on a scale of arousal and valence using the SAM (Self-Assessment Manikin) form. As a result, the arousal/valence label was low if the grade was more than or equal to 3. All the signals were captured at a sample rate of 128 Hz. Figure shows the EEG signal voltage with respect to time while the subject is playing the boring game.

Time (in seconds)

6.2 Feature Extraction Arousal class and valence of EEG data were segregated based on the subjects’ SAM scores. If the SAM score is more than 3 on a scale of 5, the class is considered high, and if it is less than 3 it is considered low. The features extracted from the data are listed in the Table 1 Following that, the full dataset was divided into two sections: test data and train data. This is done to ensure that the model’s performance is generalized to the new test set. Then, to check that the input and output mapping is proper, feature scaling was performed. Feature scaling is done with StandardScaler, which turns the data into a distribution with a mean of zero and a standard deviation of one. After that, Linear Discriminant Analysis, a dimensionality reduction approach, was applied. For classification, a KNN classifier with LDA is utilized. The model’s performance was measured using the cross-validation approach on an unknown dataset. The whole dataset is randomized at random and then divided into k groups in this procedure. The dataset was now divided into test data and train data for each group. The model should next be fitted to the training set and evaluated on the test set.

528

R. Vasanthradevi et al. Table 1. Features extracted in this study

Domain

Features extracted

Time domain

Detrended fluctuation analysis Hjorth mobility Hurst Mean Standard deviation Petrosian fractal dimension

Frequency domain

Relative power intensity

Time-frequency domain

Discrete wavelet transform

Electrodes pair domain

Differential asymmetry

Spectral entropy

Magnitude squared coherence estimate (MSCE)

6.3 Overall Evaluation Maximum accuracy was obtained with MSCE % with Arousal and % with Valence with KNN-LDA classifier. Table 2 listed the Features and its accuracies with the classifier. Table 2. Accuracy (%) of features in arousal and valence with KNN-LDA classifier Features

Arousal

Valence

Detrended fluctuation analysis

67.65

70.59

Hjorth

82.35

52.94

Hurst

76.47

52.94

Mean

70.59

52.94

Standard deviation

79.41

73.53

Petrosian fractal dimension

76.47

52.94

Relative power intensity

73.53

73.53

Spectral entropy

79.41

52.94

Discrete wavelet transform

76.47

52.94

Differential asymmetry

58.82

52.94

Magnitude squared coherence estimate (MSCE)

88.24

79.4

Recommendation System Based on EEG Emotion Recognition

529

After that, the Russell circumplex model was used to classify emotions as Bored, Calm, Fear, and Happy based on their valence and arousal levels. 6.4 Discussion on Use Cases Recommender systems expand the scope of customized information retrieval on the Internet. Using a machine-learning approach in combination with sensing devices, we have given a framework for transforming end-users’ emotional music experiences into arousal and valence evaluations. A mobile device could use the suggested framework to deliver music recommendations to its user. A framework for automatic play list

530

R. Vasanthradevi et al.

creation and next best track recommendation based on physiological indications and demographics can be used by streaming services and/or especially smartphones.

7 Conclusion The purpose of this work is to offer an approach for enhancing the performance of music recommender systems by incorporating EEG data. Based on Russell’s circumplex model, the preprocessed EEG signals from Talha Alakus’ dataset are divided into four different emotional states. For arousal and valence categorization, KNN-LDA was utilized. Classification was done with varied k values; however, the accuracy did not change significantly after (k = 35). The Magnitude Squared Coherence Estimate (MSCE) has a high accuracy of 88.24% in Arousal and 79.4% in Valence. As a result, the suggested approach holds promise for incorporating emotion phenomena to decision logic in recommendation engines. With the progress of wearable sensor technologies, using various types of sensors and collecting data at minimum time interval, performance may be enhanced.

References 1. Ekman, P., et al.: Universals and cultural differences in the judgments of facial expressions of emotion. J. Pers. Soc. Psychol. 53, 712 (1987) 2. Posner, J., Russell, J.A., Peterson, B.S.: The circumplex model of affect: an integrative approach to affective neuroscience, cognitive development, and psychopathology. Dev. Psychopathol. 17, 715–734 (2005) 3. Plutchik, R.: The nature of emotions: Human emotions have deep evolutionary roots, a fact that may explain their complexity and provide tools for clinical practice. Am. Sci. 89, 344–350 (2001) 4. Shaver, P., Schwartz, J., Kirson, D., O’connor, C.: Emotion knowledge: further exploration of a prototype approach. J. Pers. Soc. Psychol. 52, 1061 (1987) 5. Petrosian, A.: Kolmogorov complexity of finite sequences and recognition of different preictal EEG patterns. In: Proceedings Eighth IEEE Symposium on Computer-Based Medical Systems (1995) 6. Guvenir, H.A., Akkus, A.: Weighted k nearest neighbor classification on feature projections. In: Proceedings of the 12-th International Symposium on Computer and Information Sciences (1997) 7. Evans, D.G., Diggle, T., Kurniawan, S.H., Blenkhorn, P.: An investigation into formatting and layout errors produced by blind word-processor users and an evaluation of prototype error prevention and correction techniques. IEEE Trans. Neural Syst. Rehabil. Eng. 11, 257–268 (2003) 8. Horlings, R., Datcu, D., Rothkrantz, L.J.M.: Emotion recognition using brain activity. In: Proceedings of the 9th International Conference on Computer Systems and Technologies and Workshop for PhD Students in Computing (2008) 9. Subha, D.P., Joseph, P.K., Acharya, R., Lim, C.M.: EEG signal analysis: a survey. J. Med. Syst. 34, 195–212 (2010) 10. Koelstra, S., et al.: Deap: a database for emotion analysis; using physiological signals. IEEE Trans. Affect. Comput. 3, 18–31 (2011)

Recommendation System Based on EEG Emotion Recognition

531

11. Wijeratne, U., Perera, U.: Intelligent emotion recognition system using electroencephalography and active shape models. In: 2012 IEEE-EMBS Conference on Biomedical Engineering and Sciences (2012) 12. Duan, R.-N., Zhu, J.-Y., Lu, B.-L.: Differential entropy feature for EEG-based emotion classification. In: 2013 6th International IEEE/EMBS Conference on Neural Engineering (NER) (2013) 13. Rawat, R., Mahor, V., Chirgaiya, S., Shaw, R.N., Ghosh, A.: Sentiment analysis at online social network for cyber-malicious post reviews using machine learning techniques. In: Bansal, J.C., Paprzycki, M., Bianchini, M., Das, S. (eds.) Computationally Intelligent Systems and their Applications. SCI, vol. 950, pp. 113–130. Springer, Singapore (2021). https://doi.org/10. 1007/978-981-16-0407-2_9 14. Zheng, W.-L., Dong, B.-N., Lu, B.-L.: Multimodal emotion recognition using EEG and eye tracking data. In: 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (2014) 15. Bodapati, S., Bandarupally, H., Shaw, R.N., Ghosh, A.: Comparison and analysis of RNNLSTMs and CNNs for social reviews classification. In: Bansal, J.C., Fung, L.C.C., Simic, M., Ghosh, A. (eds.) Advances in Applications of Data-Driven Computing. AISC, vol. 1319, pp. 49–59. Springer, Singapore (2021). https://doi.org/10.1007/978-981-33-6919-1_4 16. Kaundanya, V.L., Patil, A., Panat, A.: Performance of k-NN classifier for emotion detection using EEG signals. In: 2015 International Conference on Communications and Signal Processing (ICCSP) (2015) 17. Rajawat, A.S., Rawat, R., Mahor, V., Shaw, R.N., Ghosh, A.: Suspicious big text data analysis for prediction—on darkweb user activity using computational intelligence model. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 735–751. Springer, Singapore (2021). https:// doi.org/10.1007/978-981-16-0749-3_58 18. Kumar, A., Das, S., Tyagi, V., Shaw, R.N., Ghosh, A.: Analysis of classifier algorithms to detect anti-money laundering. In: Bansal, J.C., Paprzycki, M., Bianchini, M., Das, S. (eds.) Computationally Intelligent Systems and their Applications. SCI, vol. 950, pp. 143–152. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0407-2_11 19. Ackermann, P., Kohlschein, C., Bitsch, J.Á., Wehrle, K., Jeschke, S.: EEG-based automatic emotion recognition: feature extraction, selection and classification methods. In: 2016 IEEE 18th International Conference on e-Health Networking, Applications and Services (Healthcom) (2016) 20. Chakraborty, A., Chatterjee, S., Majumder, K., Shaw, R.N., Ghosh, A.: A comparative study of myocardial infarction detection from ECG data using machine learning. In: Bianchini, M., Piuri, V., Das, S., Shaw, R.N. (eds.) Advanced Computing and Intelligent Technologies. LNNS, vol. 218, pp. 257–267. Springer, Singapore (2022). https://doi.org/10.1007/978-98116-2164-2_21 21. Liu, J., Meng, H., Nandi, A., Li, M.: Emotion detection from EEG recordings. In: 2016 12th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD) (2016) 22. Mukhopadhyay, M., et al.: Facial emotion recognition based on textural pattern and convolutional neural network. In: 2021 IEEE 4th International Conference on Computing, Power and Communication Technologies (GUCON), pp. 1–6 (2021). https://doi.org/10.1109/GUC ON50781.2021.9573860 23. Bazgir, O., Mohammadi, Z., Habibi, S.A.H.: Emotion recognition with machine learning using EEG signals. In: 2018 25th National and 3rd International Iranian Conference on Biomedical Engineering (ICBME) (2018)

532

R. Vasanthradevi et al.

24. Banerjee, A., et al.: Construction of effective wireless sensor network for smart communication using modified ant colony optimization technique. In: Bianchini, M., Piuri, V., Das, S., Shaw, R.N. (eds.) Advanced Computing and Intelligent Technologies. LNNS, vol. 218, pp. 269–278. Springer, Singapore (2022). https://doi.org/10.1007/978-981-16-2164-2_22 25. Asif, A., Majid, M., Anwar, S.M.: Human stress classification using EEG signals in response to music tracks. Comput. Biol. Med. 107, 182–196 (2019) 26. Athavipach, C., Pan-Ngum, S., Israsena, P.: A wearable in-ear EEG device for emotion monitoring. Sensors 19, 4014 (2019) 27. Hirekhan, S.R., Manthalkar, R., Phutke, S.: The detrended fluctuation analysis of EEG signals: a meditation-based study. In: Iyer, B., Nalbalwar, S.L., Pathak, N.P. (eds.) Computing, Communication and Signal Processing. AISC, vol. 810, pp. 771–780. Springer, Singapore (2019). https://doi.org/10.1007/978-981-13-1513-8_78 28. Rajawat, A.S., et al.: Depression detection for elderly people using AI robotic systems leveraging the Nelder–Mead method. In: Artificial Intelligence for Future Generation Robotics, pp. 55–70 (2021). https://doi.org/10.1016/B978-0-323-85498-6.00006-X 29. James, H.I., Arnold, J.J.A., Ruban, J.M.M., Tamilarasan, M., Saranya, R.: Emotion based music recommendation system. Emotion 6 (2019) 30. Rajawat, A.S., Barhanpurkar, K., Shaw, R.N., Ghosh, A.: Risk detection in wireless body sensor networks for health monitoring using hybrid deep learning. In: Mekhilef, S., Favorskaya, M., Pandey, R.K., Shaw, R.N. (eds.) Innovations in Electrical and Electronic Engineering. LNEE, vol. 756, pp. 683–696. Springer, Singapore (2021). https://doi.org/10.1007/978-98116-0749-3_54 31. Li, Z., et al.: Enhancing BCI-based emotion recognition using an improved particle swarm optimization for feature selection. Sensors 20, 3028 (2020) 32. Mridha, K., et al.: Plant disease detection using web application by neural network. In: 2021 IEEE 6th International Conference on Computing, Communication and Automation (ICCCA), pp. 130–136 (2021). https://doi.org/10.1109/ICCCA52192.2021.9666354 33. Dadebayev, D., Wei, G.W., Xion, T.E.: EEG-based emotion recognition: review of commercial EEG devices and machine learning techniques. J. King Saud Univ.-Comput. Inf. Sci. (2021) 34. Arsalan, A., Majid, M.: A study on multi-class anxiety detection using wearable EEG headband. J. Ambient Intell. Humaniz. Comput. 1–11 (2021) 35. Alakus, T.B., Gonen, M., Turkoglu, I.: Database for an emotion recognition system based on eeg signals and various computer games–GAMEEMO. Biomed. Signal Process. Control 60, 101951 (2020)

Design and Implementation of a Self-charging System to Improve the Operating Range in Quad-Motor EV M. Chandra Mohan1(B) , A. Bright Selva Kumaran1 , V. S. Rohith1 , C. Sanjay1,2 , K. Prathap Reddy1 , and J. Jayashankari2 1 Department of Mechatronics, Bharath Institute of Higher Education and Research, Chennai,

India [email protected] 2 Prince Shri Venkateshwara Padmavathy Engineering College, Chennai, India

Abstract. The current technological improvements are playing an important role in transportation industry and making it more advanced at each and every segments. At present the transportation industry is developing rapidly and its growth is increasing every year. Government is focusing on the environment friendly vehicle. For that electric vehicle (EV) is a game changer. For the past few decades electric vehicles and hybrid vehicles are improving rapidly in a vast manner. But also it carries some drawbacks as well. The drawbacks of electric vehicles are improving and changing day by day. There are many factors affecting the performance of the electric vehicles like battery efficiency, charging methods, and price. In our chapter we have developed a self-charging system for improving range of all-wheel drive electric vehicle by applying a method by which the vehicle get charged by itself when it’s in motion. Keywords: Electric vehicles · Limited range · Battery efficiency · Quad motor · Self-charging

1 Introduction Due to the drawbacks in internal combustion engines (ICE’s) like unstable fuel price, causing severe pollution and CO2 emission people are really interested in buying electrical vehicles [5]. And also we are facing oil crisis sometimes, so it pushes everyone to the electric vehicle (EV) side and in upcoming few decades this will replace EV and hybrid vehicles [6]. But electrical vehicles have it’s own disadvantages which are undeniably huge [1]. So we have to make it work effectively because of few charging stations and low range. The charging of electric vehicles needs around 30 min to charge fully that is a time consuming task. Then we are going to talk about range, electric vehicles are only store limited energy according to it’s battery capability, due to limited energy the range is also limited. There is lot of new study shows a prominent growth in energy saving technology [2, 4]. And also comes with a lot of challenges with it. For efficient performance and long range © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 533–540, 2022. https://doi.org/10.1007/978-981-19-1742-4_45

534

M. Chandra Mohan et al.

regenerating the system is necessary. At present a normal EV will travel up-to 100 km to 200 km range in single charge and there is not many charging stations like fuel stations and it takes a lot of time to charge [7]. This makes an inconvenience for users. There is also a lot of drawbacks are in electric vehicles and one of the main problem is it’s range [8]. On the other hand in upcoming years the all-wheel drive electric vehicles will replace rear and front wheel drive electric vehicles due it’s advantages and extraordinary performance, before getting into the topic we have to know about the advantages of all-wheel drive electric cars. The major advantages of all-wheel drive electric cars are reducing stopping distance, and ability to give a smooth ride on low friction surfaces and give better performance with low load and another main advantage of all-wheel drive electric vehicles are their performance and low maintenance [9]. So in future there is lot of advancements are going to come in all-wheel drive electric cars, so it is important to improvise it’s model, working and performance. But like all other electric cars quad-motor electric vehicles also face some serious disadvantages like shorter range, charging time, few charging stations and price [7]. By solving one or two of the issues as mentioned above it will make electric vehicles more viable. In this proposal we are making quad-motor system that is going to produce regenerative power through generators while the motion of each wheels and storing it in the battery. The main concept behind this project is to improve the range in all-wheel drive electric vehicle by charging itself during the trip of the vehicle and our main goal is to achieve this system in a cost efficient way.

2 Methodology The key idea of the concept is to improve the range of all-wheel drive electric car with minimum components. To represent the all-wheel drive concept we are going to use four motors for each wheels and the shaft of the each motors are connected to a pulley system which is connected to the generators that is placed in the center of the car each wheel is connected to each separate generators, so we are going to use four separate generators. When the car is in motion with the help of pulley system the generators will produce electricity and charge the battery. In this concept we are using two sets of battery. The battery 1 and battery 2 are connected to the control system of the model. The control system has battery connections, controller system and battery switching systems. The battery 1 is connected to the motors which are installed at each wheel. When these car batteries are fully charged that can facilitate the movement of the car. Simultaneously with the help of pulley system the generators will powered, which in turn produce electricity. The battery 1 is connected to the motors and battery 2 is charging with the help of generators, when the battery 1 is depleted with the help of switching system the battery 2 will take over the control of motor and battery 1 will be charged by the generators. The proposed system is simple and cost efficient and if we want to implement this concept in real life car we have choose appropriate parts and components that matches

Design and Implementation of a Self-charging System

535

the size of the car. If the manufacturer decides to implement this system according to his car size, he has to choose correct set of batteries, motors and generators. System Layout and Components

Fig. 1. Basic system layout

The above diagram represents the basic representation and gives a understanding about the concept, The A represents the wheel that is going to power the motor like that each wheel got each motors to represent all-wheel drive concept. B represents the wheel of the vehicle then C represents the battery 1 of the vehicle which is connected to the center control system of the vehicle. D represents the control system of the vehicle which consists of vehicle controlling system, charging system for vehicle and switching system for the batteries which are all connected to the control system that placed at the center of the vehicle. E is the pulley system that is connected with the motors that are going to control the vehicles. F is the generators which are connected with the other end of the pulley which is going to power the car in motion. Then the last part G, which is battery 2 that is also connected with the center control system of the vehicles. This is the brief explanation of the parts and their place in the project and their role is clearly represented in this part (Figs. 1 and 2).

Fig. 2. Layout of four wheel hub motor drive system

536

M. Chandra Mohan et al.

The above diagram will give you a example of how each separate motors will be implemented in a car without our system. This will give how the axle mechanism works in a real life car with separate motors for each wheel. The above image will explain the suspension, front-axle movement and suspension of the car with two set of battery packs. Block Diagram Here is the block diagram of the prototype to help you understand the working logic of the prototype. Now we’ll go over the working logic that is involved in the car. On the control system side you have the controller system in which the inputs are passed and the operation is received by the module and with the controller help the data sends to the motor driver which drives the motors. The generator connects with the control system to charge the battery (Fig. 3).

Fig. 3. Functional block diagram

3 Components Description 1. ARDUINO UNO For this work we have used ARDUINO UNO ATmega328P programmable microcontroller and have 14 digital input/output pins, 6 analog inputs and reset button. It is placed in the control system of our prototype which has been interfaced with motor driver, Bluetooth module and battery system for our prototype.

Design and Implementation of a Self-charging System

537

2. Bluetooth Module For this work we have used HC-05 Bluetooth module, because it is well enough to perform the operation when combined with ARDUINO, and we can get our results as expected and another reason to use Bluetooth module instead of ARDUINO BLE is the errors can be easily identified in separate models that’s why we are using ARDUINO and Bluetooth module separately. The major advantage of HC-05 modules communication is via serial which makes an easy way to interface with the controller. And it is able to use neither receiving nor transmitting data. 3. Motor driver The motors that we are using is connected with the motor driver. The motor driver is the interface between the motors and the control circuits. It takes low-current signal and then turns it to a higher-current signal which drives the motor. 4. Motors For this work we have used DC motor, DC is nothing but a type of electrical machine that converts electrical energy into mechanical energy. In our projects we are using 12 V dc motor to operate our prototype, so we are going to use four individual motors which are going to power each wheels. To replicate the all-wheel drive concept we are using each separate motor for each wheel. 5. Generators The generator’s function is to change the mechanical energy into electrical energy. The control gear associated with the generator regulates the current output according to the systems specification. Which are installed in the body of the prototype with connection to the motor.

4 Experiment and Analysis Analisis of Concept The above proposed system in this chapter is applied in a prototype with each wheel is powered by each separate motors to replicate the concept of all-wheel drive electric car and the each motor shaft is connected to the generators using pulley and at center the controlling system and battery switching system is placed. The prototype is just to understand the concept of the model and analyze the system. The applied motor is DC 12 V with rated current of 410 mA with these information. Power of motor = V × I Power of motor = 12 V × 0.410 A = 4.92 W–5 W According to the above equation and comparing it with our concept, we are using four motors for each wheel. so the total power of motors will be 20 W. in this concept we are using a 24 V battery for our prototype the battery have a capacity of 4.5 Ah P battery = 24 V × 4.5 A.h = 54 W.h We can calculate the discharge time of the battery connected to the four motors from the following calculation. Time = (54 W.h)/(20 W)

538

M. Chandra Mohan et al.

= 2.7 h = 162 min From the above calculations we can know about the power created by the motors and power of battery and battery’s discharge time. Features of the System The model and design that we have provided above is different by having some key features that will give more advantage to the electric vehicle users, when compared to the current system cars. The system that we have given here is completely different for all-wheel drive electric vehicles. The main feature of our proposal is the simplicity in concept and design, this is achieved here. The concept here will have a great and huge appeal due to its simplicity and design. In this we did not add any new components or any special complicated design, we just made this with available and existing components. At present, all people have a negative perspective opinion on electric vehicles due to its limited travel range and price. In the current production the electric car which having long range is tesla model 3 with a range of 568 km [11]. But still now experts are claiming that it is not significant enough when compared to its charging time because it takes a long charging time [12]. People don’t want to spend their time in charging the car for hours currently there is lot of new technologies are improvised in self-charging electric vehicles [13], so we focused on implementing a self-charging system for all-wheel drive electric vehicles. Electric cars will able to travel between places and longer distances this will encourage and attract people to use electric cars rather than internal combustion engine cars. The self-charging system for quad-motor electric vehicle system that we developed in this chapter is capable of achieving these goals. Most of the components and systems are currently used in the normal two-wheel drive electric cars, with only couple of new components and along with simple design. This system can be implemented in the real life quad-motor electric vehicle. By adding this system in the current all-wheel drive electric vehicle the consumers for electric vehicle will increase drastically and this will be a major change for internal and two-wheel drive electric vehicles. By applying this concept in real life production line, many changes will occur in transportation field and main environmental issues like air pollution, global warming, fossil fuel usage will be completely reduced. At present there is not lot of electric charging stations and it faces lot of difficulties in deploying charging stations. The deployment of electric charging stations will need a huge sum, we don’t want to spend this much amount when we have a car that charges itself that is the main concept and features of our system.

5 Result and Discussion In the above design and calculations we can know the layout of the prototype and some calculations related to the prototype and after the result we came to know the drawback of the proposed system is speed, when the vehicle moves faster it quickly drains the battery. So we have to maintain the speed or limit the speed of the vehicle to a certain level, that is the main drawback of the concept and even though it improves the range the initial charging of the battery in a real life car will take up to 1–2 h that is roughly 120 min and our discharge value is around 162 min which is more time than refueling

Design and Implementation of a Self-charging System

539

a internal combustion engine cars. But with proper parts and adaptive design the model that we have proposed here which can be implemented in a real life all-wheel drive electric vehicle that will really improve the range the vehicle during the trip.

6 Future Scope and Conclusion The future scope for self-charging system improves in a greater phase because people really don’t like to spend much of their time in charging their vehicles, so the only option is to improve the range is by charging the vehicle during it’s trip, by using some new upcoming researches and implementations will give a huge scope and great improvement in self-charging system for all-wheel drive electric vehicles. In this chapter we have proposed a way to self-charge the all-wheel drive electric vehicle during it’s trip. At present the electric vehicle charging system have lot of issues the current all-wheel drive electric vehicle system will have a huge travel range but it also comes with a lot of issues like price, travel range, and charging time. Through our concept we can eliminate one or two issues that we have mentioned above, if we solve the above problems in future all-wheel drive electric vehicles in future will be more affordable. Our main goal is to keep the design simple and choose cost efficient components to make it work effectively. The concept is applied to a prototype, we can implement this concept in real life all-wheel drive electric vehicles with scalable level of components.

References 1. Riley: The advantage outweigh the disadvantage of electric vehicles. Autowise, 24 June 2011. https://autowise.com/top-7-disadvantages-of-electric-cars/ 2. Liu, J., Dong, Z., Jin, T., Liu, L.: Recent advance of hybrid energy storage systems for electrified vehicles, pp. 2–6. IEEE (2019) 3. Ni,Y., Yang, F., Wang, Q., Zhang, H., Zhou, L., Wang, Y.: A review of charging methodologies to improve QoE for electric vehicles, pp. 1–4. IEEE (2020) 4. Liu, S., Jia, J.: Review of EV’s wireless charging technology, pp. 1–5. IEEE (2019) 5. DeWeerdt, S.: The key to convincing people to buy electric cars in clean energy, 11 April 2017. https://www.googleadservices.com/pagead/aclk?sa=L&ai=DChcSEwjGtsCysoPzA hUImWYCHY3MDcYABAAGgJzbQ&ae=2&ohost=www.google.com&cid=CAESQeD22 2tKOpkMPrsXci_1EyrYprNNqDCK7qdrB6q2uniISNN2dJ8OOouCRN5NTrNwVCNbaXv TMBQ3HswYAYnQvf&sig=AOD64_31LwKF0kXBuWvLSNiNYVZmKPDaA&q&nis= 1&adurl&ved=2ahUKEwi_zriysoPzAhXsxTgGHV_1CV8Q0Qx6BAgDEAE 6. Albatayneh, A., Assaf, M.N., Alterman, D., Jardat, M.: Comparison of the overall energy efficiency for internal combustion engine vehicles and electric vehicles, pp. 2–8. IEEE (2019) 7. Nezamuddin, O.N., Nicholas, C.L., dos Santos, E.C.: The problem of electric vehicle charging: state-of-the-art and an innovative solution, pp. 1–8. IEEE (2019) 8. Kumar, A., Prasad, L.B.: Issues, challenges and future prospects of electric vehicles: a review, pp. 1–6. IEEE (2018) 9. Krishnan, S.: How does all-wheel drive electric vehicle car works, 15 August 2018. https:// getelectricvehicle.com/awd-electric-cars/

540

M. Chandra Mohan et al.

10. Soni, A., Dharmacharya, D., Pal, A., Srivastava, V.K., Shaw, R.N., Ghosh, A.: Design of a machine learning-based self-driving car. In: Bianchini, M., Simic, M., Ghosh, A., Shaw, R.N. (eds.) Machine Learning for Robotics Applications. SCI, vol. 960, pp. 139–151. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0598-7_11 11. How long does it take to charge a Tesla, 11 November 2021. https://evcharging.enelx.com/ eu/about/news/blog/577-how-long-does-it-take-to-charge-a-tesla 12. Pareek, S., Sujil, A., Ratra, S., Rajesh, K.: Electric vehicle charging station challenges and opportunities: a future perspective. In: ICONC3, pp. 1–6, February 2019 13. Mandal, S., et al.: Lyft 3D object detection for autonomous vehicles. In: Artificial Intelligence for Future Generation Robotics, pp. 119–136 (2021). https://doi.org/10.1016/B978-0-32385498-6.00003-4 14. Biswas, S., Bianchini, M., Shaw, R.N., Ghosh, A.: prediction of traffic movement for autonomous vehicles. In: Bianchini, M., Simic, M., Ghosh, A., Shaw, R.N. (eds.) Machine Learning for Robotics Applications. SCI, vol. 960, pp. 153–168. Springer, Singapore (2021). https://doi.org/10.1007/978-981-16-0598-7_12

Design and Implementation of a Weed Removal Agriculture Robot J. Dhanasekar1(B) , B. Sathish Kumar1 , S. Akash1 , P. Balamurugan1 , G. Vasanth1 , and B. Umamaheswari2 1 Department of Mechatronics, Bharath Institute of Higher Education and Research, Chennai,

India [email protected] 2 Prince Shri Venkateshwara Padmavathy Engineering College, Chennai, India

Abstract. Agriculture process is time consuming as well as labour depending, however a demanding sector. Nowadays traditional methods in agriculture fields are leading to huge losses for farmers both in productivity and financially. This chapter is focused on the development of a novel Automated Agriculture Robot. A robot can execute operations like digging the soil, planting seeds, watering, identifying the weed sand removing them with more accuracy and precision with the help of robotics and automation. This will be a solution to the time constraints and labour issues, and it helps to achieve a higher rate of production. Weed control and removing the weeds from the crop fields is an essential part and will make a huge impact in increasing the productivity and quality of the yields. The effective weed management system should be strong and adaptable. A weed robot has been designed and fabricated using image processing technique here. Keywords: Agriculture · Automated Robot · Identity weeds and insects · Higher rate of production

1 Introduction Weed is some kind of unwanted plant that grows in the cultivation lands and the place where the crops grow. (Shanmugam et al. 2020) Weeds are the absolute reason why the agriculture crops is not produced in huge amounts and of a good quality (Shanmugam et al. 2020). The advancement that occurred in the agriculture field particularly managing weeds and control made a huge impact in increasing the quality and productivity of the yields (Eldert van Henten and Gert Kootstra 2020). Achieving the needs of foods and products for the rapid rise of inhabitants of the country is an important one in which weed control will play a massive role. Agriculture is the vital department that has a vital part in a country’s development (Fenninore et al. 2017) . So it is important to improve agriculture fields in a digital manner that is why robots can play a important role in agriculture fields (Slaughter et al. 2017). The number of people in our country increases it is also a major reason to increase the production and quantity of products in agriculture fields (Badkhal et al. 2019). An © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 541–550, 2022. https://doi.org/10.1007/978-981-19-1742-4_46

542

J. Dhanasekar et al.

Agrobot is an autonomous robot used in farming to help improve efficiency and reduce manual labour and time. In most of the farm fields robotic mechanisms like tractors are used to sow the fields. But this Agrobot is used to identify weeds and manage to remove them from the field. Robots operating at low speed can detect weeds using Image sensing and remove them by using a robotic arm (Kesavan et al. 2018). A robotic arm is used to pickup the unwanted plant which identified by the agriobots. Agriculture Robots are not widely used in rural areas and in small scale farming areas. Basically, the usual agriculture works are being done by the farmers or the heavy machines (Kesavan et al. 2018). More traditional methods are used for sowing like using cattle and using labour for identifying weeds and insects and using harmful pesticides. This results in the production of inferior agricultural products. And using heavy machinery results in noise pollution and causes unhealthy conditions to the farmers (McAllister et al. 2019). Agriculture is not only cultivating the plants but also it is necessary to protect the cultivated plants from insects and weeds. If a farmer wants good quality and quantity of agricultural products he/she needs to identify the weeds that occur near the plants and they need to remove them (Steward et al. 2019). Most plant-based weeds are tough to identify on naked eyes so Automated Agriculture Robots can be used to identify the weeds and remove them. There are some effective methods to control and manage the weeds in the agriculture fields (Steward et al. 2019). Those methods should be adaptable and strong. Strong weed management methods are methods in which the weeds in the fields should be managed by any certain circumstances that happen in the agriculture fields (Kujawa et al. 2020). The adaptable weed management methods are methods that manage and control the weeds in conditions like generic, climatic conditions and so on. The main goal is to control the weeds from the fields is to develop the productivity and quality of the agriculture products (Kujawa et al. 2020). However there are some more methods like chemical weed controller and weed controlled by using mechanical methods to control the weeds (Wo et al. 2020). Mechanical methods use larger mechanisms like large machine sprayers which are used to remove them from fields on a large scale and they are quite incompatible with their size in small fields (Wo et al. 2020). This review is depicted on Robots which are used in agriculture fields for various activities like harvesting, sowing and identifying weeds and insects. Agriculture is the major field in any country. To increase the crop productivity, robots can be used in agriculture. The utilization of robots reduces the labour’s field work and lead time of agriculture. The main reason to increase productivity in the agriculture field is the increase in population in our country. Due to increase in population the need for food and its raw material from agricultural fields also increase. Traditional agriculture methods were used in the past and those methods are like ploughing with the use of cows and sowing seeds manually. But nowadays tractors are used to plough the fields. But these traditional methods are used in some village side agriculture fields but they produce small amounts of products. But we are in need of more productivity and food production, so the necessity of development in agriculture fields in artificial ways to improve the quality, quantity and efficiency of agriculture in both rural and developed agriculture. The sudden growth and development of the internet and the online world make it easier for proceed the agriculture process in automated and robotic system for artificial

Design and Implementation of a Weed Removal Agriculture Robot

543

agriculture. Intelligence technologies using machine learning and image processing has evolve not only for sowing, and harvesting, yet for removing or controlling the unwanted plants, whereas weed plant detection still being a difficult challenge (Firthous Begum and Vignesh 2015). Machine learning technologies have been promising technology in processing image and providing the image in perfect condition and that the image is being processed (Ashok Kumar and Tamizharasi 2015). In spite of using Image processing technology, the detection of disease is challenging for microorganism control and it paves a way for the development of mechatronics and robotics solution. Introducing autonomous robot in agriculture may result in many benefits like producing more amount of products, lesser consumption of energy sources, man labour can be decreased and time taken in field will be lesser than manual activities (Caggiano et al. 2019). The increase in outcome of agriculture fields is mainly based on the prevention of the disease and the insects that cause the disease in the plants (Pandey et al. 2020). So, by using pesticides and other harmful products the agriculture products can result in toxic conditions and it is also very difficult to identify these kind of weeds in naked eyes so it is necessary to use the autonomous robots to identify and control these kinds of unwanted plants (Kesavan et al. 2018).

2 Related Works In this particular section, the components and the previous works based on weed control robots like Robotic weed control and weed controlled by chemicals are discussed in support of our topic. 2.1 Weeding Robot Nowadays weeds are basically controlled by automated methods and basically, they work under cameras, GPS systems, Sonar methods and lasers (Wo et al. 2019). These kinds of methods give upto 80% of positive results. The method is based on image processing in which the camera is programmed with some inputs and the perception of unwanted plants grown and field plants is easily done by the camera while the robotic arm is used after the weed is identified and then the arm picks up the weed from the crop fields (Wo et al. 2019). There is another method in which the lasers are used to control weeds from crop fields. The main and most important is to identify which is the crop plant and the weed. Therefore Image processing is used to identify the difference. By using laser technology it results in up to 93% of better positive results. The most vital feature in using a laser is it removes the growth of weeds from its root and doesn’t allow the plant to grow further in any position. (Wo et al. 2019). Figure 1 and Fig. 2 shows the model of the weed robot. 2.2 Weed Controlled by Chemicals In the traditional way of agriculture the weeds are controlled by using pesticides and herbicides which are manually sprayed by the farmers. In these kinds of methods, the weeds cannot be removed from the crops completely because it is difficult for farmers to

544

J. Dhanasekar et al.

Fig. 1. Model weed robot

Fig. 2. These kinds of robots is used for killing the weed by using a sensor which is placed at the green position where it identifies weed and crop plan

identify weeds and crop plants. These methods work based on spot spraying techniques in which the chemicals are directly sprayed in the roots of weeds and control the growth of weeds. These ways are less efficient than the modern ways. Weed tracking in a crop plant is very difficult when they are in an uncertain position so the cameras which are placed in the robots must identify the weeds despite the conditions and the soil colour. The images are processed and then the robotic arm with the chemicals like herbicides sprays at the root of the weeds directly. Then the weeds are gradually controlled by the further process. The Fig. 3 shows design of proposed weed robot. 2.3 Challenges in Robotic Weed Control There can be many challenges while controlling the growth of weeds but there are some main challenges and concerns while controlling weeds using mechanical methods. The first and most main concern is the ability to notice whether it is a crop plant or weed and unwanted plant. The second one is the control mechanism of weeds. 2.3.1 Ability to Notice After many years of research some information about the particular plants in the sensors. These can be carried by the vehicle platforms which can perform with the sensors.

Design and Implementation of a Weed Removal Agriculture Robot

545

Fig. 3. Design of weed robot

Satellites, aerial and ground based machines can be used widely to detect the plants where they can be operated in the fields and based machines which can carry sensors can be used for detecting the plants in the fields. However ground can work based on the crop size and able to be accurately related to the distance that is covered by those weeds in crop fields. There can be two categories in which the ground machine works are Spectral Reflectance and the Biological morphology of the plants. The biological morphology of the plants includes the size and colour and its characteristic of plants that are related to the crop fields. Whereas spectral reflectance is the process in which the rays are introduced there are two types: field/soil reflectance and crop reflectance. The soil reflectance are the one which is very minute and varies which the rays get deflected by the soils and the fields and the other one crop reflectance is the one in which the rays get deflected by the particular plants that need to be grown in the field. There are definitely amounts of reflection for the definite plants that are grown in fields.

3 Hardware and Software In this chapter, it is a agriculture robot which control weed in agriculture which is controlled by micro controller. The weed control is based on row type crops cultivation. The main part of the robot is colour sensor which senses and colour and DC motor which run the wheel of the robots and a servo motor which helps in cutting the weeds from the cultivated crops. An IR sensor is used to detect the location of the stem and root of the weed plants and these sensors are also used to track the obstacles present in front of the robots and allows the robots to proceed further after finishing the first row of the column in the crops. An vision system like image sensing and machine vision is used to detect the position of the crop plant and weed plant in agriculture field.

546

J. Dhanasekar et al.

4 Methodology While there are many methods and ways of controlling the weed grown in the agriculture fields there are some ways which can be accurate and beneficial. These ways are employed in either mechanical ways or chemical ways. Mechanical ways are where the weeds are removed by using the machines and while in chemical ways the solutions are used such that they are sprayed in the unwanted plants and then the growth of those plants stops and it results in controlling them from crop plants. 4.1 Perception and Detection The identification of the center position of crop plant and cannabis plants and position of stem and root of the weed plants is an important process in robotic weeding system. To remove the weeds from crops physically it requires accurate information and location of the weed plants where it is quite difficult tasks. The main areas to be targeted while weeding systems are interring row, intra row and close to crop areas. The inter row cultivation crops is commonly based on row production crops. The weeding system for row and close to crop areas is different. These inter row crops are highly effective in corn and soybean where the herbicides are directly applied over the crop rows. The intra row crops are based on where the weeds are grown in between the crop plants. If the weed is present in the row crops the colour sensor in the camera gets trigger and it gets information and passes it to microcontroller and if there is no weed present in the crops then the sensor fetch and information to the microcontroller and the weeding controller to proceed further. The weed control is placed vertically to the ground where the image sensor sends the signals to the microcontroller. Fig. 4 shows Weed are detected by using image processing technique where the marked indicates the location of weeds. Fig. 5 shows Image taken under digital camera and processed under machine vision to detect the weed.

Fig. 4. Weeds are detected by using image processing technique where the marked indicates the location of weeds

Design and Implementation of a Weed Removal Agriculture Robot

547

Fig. 5. Images taken under digital camera and processed under machine vision to detect the weeds.

4.2 Robotic Weed Controller The robotic system consists color sensors, two DC motors and two servo motors and where all of them are controlled by the microcontroller placed at middle of the robot. The DC works under simple electromagnetic process and energy are given to the wheels. Depending upon the energy the DC motor controls the wheel of the robots and servo motor controller will be worked for cutting the weeds. When the DC motor start working which controls the wheels and move particular crop rows. The color sensors are place at the front and side of the robot. The sensor placed in front of the robots sense the colour of weeds and send information to the microcontroller whereas the sensor at the side helps in movement of the robot. If there no signal from the sensor then the controller indicates that the process is completed. In that particular position the robots tries to rotate at 270°. And then moves to the another row and follows the above process. If any obstacles present before robot the infra red sensor indicates the information to microcontroller and it stops the robot and change its direction. After detecting the weeds the servo controller comes and uses shaft to control the weed by cutting the weeds from the crop plants. Step 1: If weeds present in the crops then the IR sensor present in the front of the robots gets trigger and fetches information from the camera and send the information to the microcontroller and the microcontroller gets the information. If there is no weed then the sensor sends the information to the microcontroller and then the robot moves further by using the DC motor attached along with the wheels of the robot. Step 2: If there is no weed present in the particular row of crops then the information is passed to the microcontroller and then the robot turns at angle of 270 and then repeat the process in another row of crops. Step 3: If there is any obstacle before the robot which affects the motion of the robot then the obstacle is detected by the sensors placed at the side of the robots and passes the information to the microcontroller and then the robot stops and the obstacle needed to be removed by the farmers.

548

J. Dhanasekar et al.

Figure 4 shows the proposed weed removal robot (Fig. 6).

Fig. 6. The Weed removal robot

5 Result In this chapter we have used weed management with the automated robotic technology. It is resulted in 83% success rate while detecting the weed and removing them from crops. The speed of the robot depends on the surface level of soil. The DC motor controls the movement and the speed of the wheels and servo motor use to pick up the weeds from the crops. The future scope on the weed management using automated robot will have high efficiency and result at high successive rate.

6 Conclusion Agrobot is made to ease the work of poor and uneducated farmers by detecting the weeds that are grown in between plants. By detecting the weeds the production for the farmers can be increased and the quality of the products can be very good. By controlling the weeds in fields the requirements that a plant need can be intake only by plants can grow healthier and the quality and productivity will be higher. This is mainly focused on decreasing the usage of energy sourced and man power which can beneficial to the farmers. The robots are used for a variety purpose like sowing, weeding, harvesting and other agriculture purposes. Robots can be created as per the requirements that are made by the farmers in which in return they will get a huge profit and production in agriculture fields. The usage of the internet can take part in a vital impact on usage and improvisation of these kind of robots. These kinds of robots should be practiced and knowledgeable to the rural area farmers and small scale farmers. So that they can use these kinds of robots manually.

Design and Implementation of a Weed Removal Agriculture Robot

549

References 1. Fawakherji, M., Youssef, A., Bloisi, D., Pretto, A., Nardi, D.: Crop and weeds classification for precision agriculture using context- independent pixel-wise segmentation (2019) 2. Shanmugam, S., Assunção, E., Mesquita, R., Veiros, A., Gaspa, P.D.: Automated weed detection systems: a review (2020) 3. Fenninore, S.A., Slaughter, D., Simmens, M., Lenon, R., Saber, M.: Technology for automation of weed control in specialty crops (2017) 4. Fountas, S., Mylonas, N., Pekkeriet, E.: Agricultural robotics for field operations (2020) 5. Slaughter, D., Glies, D.K., Downey, D.: Autonomous robotic weed control systems: a review (2010) 6. Johnson, F.T., Aker, J.: Artificial Intelligence for weed detection (2020) 7. Ruigrok, T., van Henten, E., Kootstra, G.: Application-specific evaluation of a weed-detection algorithm for plant-specific spraying (2020) 8. Badkhal, T., Sharma, R., Sonu, I., Patil, S., Maapri, R.: Automatic Weed Killing Robot for Agriculture Purpose and Insect Killing (2019) 9. Firthous Begum, N., Vignesh, P.: Design, and implementation of pick and place robot with wireless charging application. Int. J. Sci. Res. (2015) 10. Kulkarni, V.A., Deshmukh, A.G.: Advanced agriculture robotic weed control system (2014) 11. Pandey, P., Dakshinamurthy, H.N., Young, S.: A Literature Review of Non- Herbicide, Robotic Weeding (2020) 12. Ashok Kumar, K., Tamizharasi, K.: Gesture controlled robot using MEMS accelerometer for eradication of weeds. Indian J. Sci. Technol. (2015) 13. Bawden, O., et al.: Robot for weed species plant-specific management (2017) 14. Roy, K.C.: High Tech Agriculture in Bangladesh. The Daily Star. N.p., 2012. Web. 3 March 2017 15. Aravind, K.R., Raja, P., Pérez-Ruiz, M.: Task-based agricultural mobile robots in agricultural farming: a review. Research (2017) 16. Kesavan, K., Monish, R., Murali, S.: Automatic weed removal system (2018) 17. McAllister, W., Osipychev, D., Chowdary, G., Davis, A.: Agbots: weeding a field with a team of autonomous robots (2018) 18. Wo, X., Lottes, P., Pradalier, C.: Robotic weed control using automated weed and crop classification (2020) 19. Koteswara Karthik, P., Ravi Chandra, P.: Overview of Agricultural Robots Posted by Yuva Engineers on 25 January 2014 20. Wu, X., Aravecchia, S., Lottes, P., Stachniss, C., Pradalier, C.: Robotic weed control using automated weed and crop classification 21. Kujawa, S., Niedbała, G., Zaborowicz, M.: Plant and weed identifier robot as an agroecological tool using artificial neural networks for image identification (2020) 22. Rajawat, A.S., et. al.: Robotic process automation with increasing productivity and improving product quality using artificial intelligence and machine learning. In: Artificial Intelligence for Future Generation Robotics, pp. 1–13 (2021). https://doi.org/10.1016/B978-0-323-85498-6. 00007-1 23. Sindhu, B., Asha, K.: Department of IS & EGM Institute of Technology Davanagere, INDIA “Automation in Agriculture using AGROBOT”. Int. J. Eng. Res. Technol. (IJERT) (2021) 24. Chandan, S., et. al.: Inverse kinematics analysis of 7-degree of freedom welding and drilling robot using artificial intelligence techniques. In: Artificial Intelligence for Future Generation Robotics, pp. 15–23 (2021). https://doi.org/10.1016/B978-0-323-85498-6.00004-6 25. Rajawat, A.S., et. al.: Depression detection for elderly people using AI robotic systems leveraging the Nelder–Mead method. In: Artificial Intelligence for Future Generation Robotics, pp. 55–70 (2021). https://doi.org/10.1016/B978-0-323-85498-6.00006-X

550

J. Dhanasekar et al.

26. Fountas, S., Mylonas, N., Malounas, I., Rodias, E., Santos, C.H., Pekkeriet, E.: Agricultural robotics for field operations (2020) 27. Mahmud, M.S.A., Abidin, M.S.Z., Emmanuel, A.A., Hasan, H.S.: Robotics and automation in agriculture: present and future applications (2020) 28. Das, S., et. al.: Advance machine learning and artificial intelligence applications in service robot. In: Artificial Intelligence for Future Generation Robotics, pp. 83–91 (2021). https:// doi.org/10.1016/B978-0-323-85498-6.00002-2 29. Amritha Ashok, K., Savy, A., Shijoh, V., Shaw, R.N., Ghosh, A.: Hospital assistance robots control strategy and machine learning technology. In: Bianchini, M., Simic, M., Ghosh, A., Shaw, R.N. (eds.) Machine Learning for Robotics Applications Studies in Computational Intelligence. SCI, vol. 960, pp. 35–46, Springer, Singapore (2021). https://doi.org/10.1007/ 978-981-16-0598-7_3 30. Caggiano, A., Jianjing Zhang, C., Vittorio Alferi, D., Fabrizia Caiazzo, D., Gao, R., Teti, R.: Machine learning based image processing for online defect recognition in additive manufacturing (2019)

Author Index

A Akash, S., 541 Aluko, Anuoluwapo, 346 Amarnadh Reddy, N., 60 Amarnath Reddy, N., 44 Ambati, Giriprasad, 12, 30, 44, 77 Anantha Narayanan, V., 338 Anas, M. M., 161 Anowar, S. M. Sayeed, 1 Archana, K., 161 Athiram, M., 161 B Babu, M. Naga Praveen, 323 Babul, Abhinav Kumar, 172, 257 Bajaj, Mohit, 505 Balamurugan, P., 541 Basa, Sidvik, 323 Behera, Laxmidhar, 373 Belkhier, Youcef, 505 Bhatia, Richa, 426 Bhattacharjee, Deepsikha, 392 Bhaumik, Archisman, 392 Bhowmick, Sourov, 464 Bihari, Shiv Prakash, 172, 257 Biswas, Shuvra Prokash, 481, 494 Bright Selva Kumaran, A., 533 C Carpanen, Rudiren Pillay, 346 Challa, Komaladitya, 383

Chandra Mohan, M., 533 Charan, Piyush, 140 Chouhan, Robin Singh, 451 D Dahiya, Mohit, 121 Das, Saikat, 280 Deb, Alok Kanti, 437 Deb, Suman, 392 Deepak, Gerard, 215 Devulal, B., 406 Dhanasekar, J., 541 Dhanunjaya, V., 203 Dorrell, David, 346 Duba, Prasanth Kumar, 323 F Faizan, Mohd, 196 Francis, Febin, 237 G Ganesh Babu, B., 224 Gangolu, Suryanarayana, 92 Ghosh, Ankush, 451, 505 Godwal, Shanker, 248 Goel, Anu, 426 Gupta, Aruvand, 172, 257 Gupta, Sandeep, 373 Gupta, Shashank, 92 Gupta, Vikalp, 172, 257

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 S. Mekhilef et al. (Eds.): ICEEE 2022, LNEE 893, pp. 551–553, 2022. https://doi.org/10.1007/978-981-19-1742-4

552

Author Index

H Halloum, M. Ramez, 237 Hari Priya, T., 106 Hari, Akshay, 472 Haripriya, T., 44 Hasan, Mehedi, 1 Hussain, Md. Shajid, 268

Nasir, Hina, 196 Nayak, Nirakar, 280 Niazi, K. R., 415 Nimje, Akhilesh, 248

I Islam, Md. Rabiul, 494

P Palvannan, S., 215 Pamula, Vinay Kumar, 383 Patel, Vinod, 248 Prasad, Sarthak Kamta, 183 Prathap Reddy, K., 533 Priyadharshini, R., 515

J Jai Rajesh, P., 515 Jain, Jay Kumar, 152 Jain, Palash, 152 Jayashankari, J., 533 K Kale, Advait, 451 Kalpana, R., 515 Kanojia, S. S., 248 Kapoor, Rashmi, 12, 60 Karthika, Garikapati Annapurna, 30, 77 Khan, Abid Hossain, 1 Khan, Imran Ullah, 140 Kumar, Amritesh, 183, 280 Kumar, Anuj, 292 Kumar, Chava Suneel, 77 Kumar, Chava Sunil, 12 Kumar, D. Ravi, 406 Kumar, Kamlesh, 121 Kumar, Manoj, 121 Kumar, Priyanka, 472 Kumar, Ranjeet, 437 L Lakra, Priyanshu, 121 Lakshminarayana, G., 30, 60, 106 Laxkar, Pradeep, 292 M Makhlouf, Thiziri, 505 Mehroz, Mohd, 196 Mishra, Vishesh Kumar, 464 Mittal, Nupur, 140 N Nag, Sukanya, 392 Naga Jyothi, M., 106 Nahin, Nazmul Islam, 481 Nair, Prashant R., 338, 365

O Ojo, Evans, 346

R Rabiul Islam, Md., 481 Rahman, M. Mizanur, 1 Rajalakshmi, P., 323 Rajashekar, V., 406 Rajawat, Anand Singh, 451 Ramish, Qazi, 196 Rana, M. S., 268 Rao, Gottapu Sasibhushana, 132 Rawat, Tanuj, 415 Reddy, B. Kiran Kumar, 237 Reddy, B. Subba, 237 Reddy, G. Nithin, 237 Reddy, Jangamreddy Rajasekhar, 237 Reena, R., 515 Rohith, V. S., 533 S Sahri, Younes, 505 Saleem, Sharzeel, 464 Sanjay, C., 533 Sathish Kumar, B., 541 Shaji, Maneesha, 161 Sharma, Piyush, 292 Sharma, Sachin, 415 Shaw, Rabindra Nath, 451, 505 Shila, Sharmin, 494 Singh, Jyotsna, 415 Siva, M., 406 Sobha, Parvathy, 312 Solanki, Surendrasinh K., 248 Sreedha, B., 365 Sreelakshmi, Koduri, 132 Suresh, Anugraha, 161 Surya Kalavathi, M., 224 T Tabassum, Fariya, 268

Author Index U Umamaheswari, B., 541 V Vasanth, G., 541 Vasanthradevi, R., 515 Venkata Kishore, P., 203

553 Verma, Akash, 183 Vijaya Bhaskar Reddy, K., 203 Vijaya Kumar, S., 203 Y Yarlagadda, Venu, 12, 30, 44, 60, 77, 106