Efficient Utilization of Energy Employing Meta

Efficient Utilization of Energy Employing Meta-heuristic Techniques with the Incorporation of Green Energy Resources in Smart Cities

By

Asif Khan CIIT/FA15-PCS-004/ISB

PhD Thesis In Computer Science

COMSATS University Islamabad, Islamabad, Pakistan

Spring, 2018

COMSATS University Islamabad


A Thesis Presented to

COMSATS University Islamabad, Islamabad, Pakistan

In partial fulfillment of the requirement for the degree of

PhD (Computer Science)

By


Spring, 2018 ii


A Post Graduate Thesis submitted to the Department of Computer Science as partial fulfilment of the requirement for the award of Degree of PhD (Computer Science).

Name

Registration Number


Supervisor:

Dr. Nadeem Javaid, Associate Professor, Department of Computer Science, COMSATS University, Islamabad, Islamabad Campus.

Co-Supervisor:

Dr. Mariam Akbar, Assistant Professor, Department of Computer Science, COMSATS University, Islamabad, Islamabad Campus. iii

Certificate of Approval This is to certify that the research work presented in this thesis, entitled “Efficient Utilization of Energy Employing Meta-heuristic Techniques with the Incorporation of Green Energy Resources in Smart Cities” was conducted by Mr. Asif Khan under the supervision of Dr. Nadeem Javaid. No part of this thesis has been submitted anywhere else for any other degree. This thesis is submitted to the Department of Computer Science, COMSATS University Islamabad, Islamabad, in the partial fulfillment of the requirement for the degree of Doctor of Philosophy in the field of Computer Science.

Asif Khan

Signature::

Examinations Committe:

................................................... External Examiner 1:

................................................... External Examinar 2:

(Designation and Office Address)

(Designation and Office Address)

................................................... Dr. Nadeem Javaid Supervisor, Department of Computer Science CUI, Islamabad.

................................................... Dr. Mariam Akbar Co-Supervisor, Department of Computer Science CUI, Islamabad.

................................................... Dr. Majid Iqbal Khan Chairperson/HoD, Department of Computer Science CUI, Islamabad.

................................................... Prof. Dr. Syed Asad Hussain Dean, Faculty of information science and technology, CUI.

iv

Author’s Declaration I Asif Khan, CIIT/FA15-PCS-004/ISB hereby state that my PhD thesis titled “Efficient Utilization of Energy Employing Meta-heuristic Techniques with the Incorporation of Green Energy Resources in Smart Cities” is my own work and has not been submitted previously by me for taking any degree from this University i.e., COMSATS University Islamabad or anywhere else in the country/ world. At any time if my statement is found to be incorrect even after I graduate the University has the right to withdraw my PhD degree.

Date:

Signature of the student:


v

Plagiarism Undertaking I solemnly declare that research work presented in this thesis titled, “Efficient Utilization of Energy Employing Meta-heuristic Techniques with the Incorporation of Green Energy Resources in Smart Cities” is solely my research work with no significant contribution from any other person. Small contribution/ help wherever taken has been duly acknowledged and that complete thesis has been written by me. I understand the zero tolarnace policy of HEC and COMSATS University Islamabad towards plagiarism. Therefore, I as an author of the above titled thesis declare that no portion of my thesis has been plagiarized and any material used as reference is properly referred/ cited. I undertake if I am found guilty of any formal plagiarism in the above titled thesis even after award of PhD degree, the University reserves the right to withdraw/ revoke my PhD degree and that HEC and the university has the right to publish my name on the HEC/ university website on which names of students are placed who submitted plagiarized thesis.

Signature of the student:

Date:


vi

Certificate It is certified that Asif Khan, CIIT/FA15-PCS-004/ISB has carried out all the work related to this thesis under my supervision at the Department of Computer Science, COMSATS University Islamabad, Islamabad and the work fulfills the requirement for award of PhD degree.

Supervisor:

Date:

Dr. Nadeem Javaid, Associate Professor Department of Computer Science, COMSATS University Islamabad, Islamabad. Co-Supervisor:

Dr. Mariam Akbar, Associate Professor Department of Computer Science, COMSATS University Islamabad, Islamabad.

Head of Department:

Dr. Majid Iqbal Khan, Associate Professor Department of Computer Science, COMSATS University Islamabad, Islamabad. vii

DEDICATION

Dedicated to my beloved family, teachers, and friends

viii

ACKNOWLEDGEMENT Firstly, all praises to Allah Almighty, the most merciful, beneficial and most gracious for blessing me with courage, confidence, and strength to complete my dissertation. After that, I would like to express my profound appreciation to many people who helped and supported me during my Ph.D. duration to successfully complete my thesis on time. This research work would not be possible without their generous support. I would like to express my sincere gratitude to my mentor and supervisor Dr. Nadeem Javaid for his continuous support, kindness, efforts, and involvement in my research work. I will always be thankful to him for his guidance and motivation which helped me in completing this thesis. I could not have imagined having a better mentor and advisor for my Ph.D. study and truly indebted to him for his immense knowledge, useful thoughts, and fatherly advice sharing. I am also grateful to my co-supervisor Dr. Mariam Akbar for her kind support. I would like to also pay a bundle of thanks to Dr. Muaz A. Niazi, Dr. Yasir Faheem and Dr. Zara Hamid for paying special attention to me during my first-year of coursework. I gratefully acknowledge the indigenous HEC scholarship fundings under Aghaz-eHaqooq-e-Balochistan Project (AHBP) which made my Ph.D. work possible. I am also thankful to Science and Information Technology, Department, Government of Balochistan (GoB) for sanctioning and approving study leave. I would like to thank my family for their continuous support, understanding, and assistance whenever I needed them throughout my studies and research work. I am always grateful to them for their encouragement and support. Last but not the least, I am greatly thankful to each of my ComSens Research Lab fellows for being with me through all thick and thins and for providing me a friendly atmosphere and sharing the best memories throughout this duration.

ix

ABSTRACT Efficient Utilization of Energy Employing Meta-heuristic Techniques with the Incorporation of Green Energy Resources in Smart Cities A smart city is an efficient, reliable, and sustainable urban center that facilitates its inhabitants with a high quality of life standards via optimal management of its resources. Energy management of smart homes (SHs) is one of the most challenging and demanding issues which needs significant effort and attention. Demand side management (DSM) in smart grid (SG) authorizes consumers to make informed decisions regarding their energy consumption pattern and helps the utility in reducing the peak load demand during an energy stress time. In DSM, scheduling of appliances based on consumer-defined priorities is an important task performed by a home energy management controller (HEMC). However, user discomfort is caused by the scheduling of home appliances based on the demand response or limiting its time of use. Further, rebound peaks that are regenerated in the off-peak hours is also a major challenge in DSM. In addition, an increase in the world population is resulting in high energy demand; thus, causing a huge consumption of fossil fuels. This ultimately results in severe environmental problems for mankind and nature. Renewable energy sources (RESs) emerge as an alternative to the fossil sources. These RESs have the advantages of environmental friendliness and sustainability, which are incorporated in SHs via two modes: grid-connected (GC) or stand-alone (SA). The reliability concerns in RESs are usually met with the usage of hybrid RESs along with the integration of energy storage systems (ESS). The efficient usage of these components in the hybrid RESs requires an optimum unit sizing that achieves the objectives pertaining to cost minimization and reliability in SA mode. These are some of the main concerns of a decision maker. This thesis focuses on employing meta-heuristic techniques for efficient utilization of energy and RESs in SH. At first, an evolutionary accretive comfort algorithm (EACA) is developed based on four postulations which allows the time-varying priorities to be quantified in time and device-based features. Based on the input data, considering the appliances’ power ratings, its time of use, and absolute comfort derived from priorities, the EACA is able to generate an optimal energy consumption pattern which would give maximum satisfaction at a predetermined user budget. A cost per unit comfort index (χ) which relates the consumer expenditure to the achievable comfort is also demonstrated. To test the applicability of the proposed EACA, three budget scenarios of 1.5 $/day, 2.0 $/day, and 2.5 x

$/day are performed. Secondly, a priority-induced DSM strategy based on the load shifting technique considering various energy cycles of an appliance is presented. The day-ahead load shifting technique is mathematically formulated and mapped with multiple knapsack problem (MKP) to mitigate the rebound peaks. The proposed autonomous HEMC embeds three meta-heuristic optimization techniques: genetic algorithm (GA), enhanced differential evolution (EDE), and binary particle swarm optimization (BPSO) along with the optimal stopping rule, which is used for solving the load shifting problem. Next, we integrate the RESs and ESS in a residential sector considering GC mode. The proposed optimized home energy management system minimizes the electricity bill by scheduling the household appliances and ESS in response to the dynamic pricing of the electricity market. Here the appliances are classified into shiftable and non-shiftable categories, and a hybrid GA-BPSO (HGPO) scheme outperforms to other algorithms in terms of cost and a peak-to-average ratio (PAR). Finally, meta-heuristic schemes that do not depend on algorithmic-specific parameters are focused for RESs and ESS integration in a SA system. Preliminary, Jaya algorithm is used for finding an optimal unit sizing of RESs components, including photovoltaic (PV) panels, wind turbines (WTs), and fuel cell (FC) with an objective to reduce the consumer total annual cost. The methodology is applied to real solar irradiation and wind speed data taken for Hawksbay, Pakistan. Next, an improved Jaya and the learning phase as depicted in teaching learningbased optimization (TLBO), named JLBO algorithm for optimal unit sizing of a PV-WT-Battery hybrid system is also demonstrated for another site located in Rafsanjan, Iran. The system reliability is considered using the maximum allowable loss of power supply probability (LP SP max ) provided by the consumer. Thus, the thesis objectives achieved are to have a green, reliable, economical, and sustainable power supply in the SH.

xi

Journal Publications Dr. Nadeem Javaid, Associate Professor (Supervisor) Asif Khan 1 Khan, Asif , Nadeem Javaid, and Majid Iqbal Khan. “Time and device based priority induced comfort management in smart home within the consumer budget limitation.” Sustainable Cities and Society (2018): 41, 538555. (IF=3.073). Download. 2 Khan, Asif , Nadeem Javaid, Adnan Ahmad, Mariam Akbar, Zahoor Ali Khan, and Manzoor Ilahi. “A priority-induced demand side management system to mitigate rebound peaks using multiple knapsack.” Journal of Ambient Intelligence and Humanized Computing (2018): 1-24. (IF=1.423). Download. 3 Ahmad, Adnan, Asif Khan, Nadeem Javaid, Hafiz Majid Hussain, Wadood Abdul, Ahmad Almogren, Atif Alamri, and Iftikhar Azim Niaz. “An Optimized Home Energy Management System with Integrated Renewable Energy and Storage Resources.” Energies 10, no. 4 (2017): 549. (IF=2.676). Download.

xii

Conference Proceedings 1 Khan, Asif , Nadeem Javaid, Muhammad Nadeem Iqbal, Naveed Anwar, Inzimam-ul-Haq, and Faraz Ahmad. “Time and device based priority induced demand side load management in smart home with consumer budget limit.” In IEEE 32nd International Conference on Advanced Information Networking and Applications (AINA), pp. 874-881. IEEE, 2018. Download. 2 Zahoor, Saman, Nadeem Javaid, Anila Yasmeen, Isra Shafi, Asif Khan, and Zahoor Ali Khan. “Optimized Energy Management Strategy for Home and Office.” In International Conference on Emerging Internetworking, Data & Web Technologies, pp. 237-246. Springer, Cham, 2018. Download. 3 Yasmeen, Anila, Nadeem Javaid, Itrat Fatima, Zunaira Nadeem, Asif Khan, and Zahoor Ali Khan. “A Metaheuristic Scheduling of Home Energy Management System.” In International Conference on Emerging Internetworking, Data & Web Technologies, pp. 214-224. Springer, Cham, 2018. Download. 4 Saman Zahoor, Nadeem Javaid, Asif Khan, Fatima J. Muhammad, Maida Zahid, and Mohsen Guizani. “A Cloud-Fog-Based Smart Grid Model for Efficient Resource Utilization.” Submitted in International Wireless Communications and Mobile Computing Conference, IWCMC 2018 - Big Data Net 2018. 5 Khan, Asif , Nadeem Javaid, and Sakeena Javaid. “Optimum unit sizing of stand-alone PV-WT-Battery hybrid system components using Jaya.” Submitted in 21st IEEE International Multi Topic Conference (INMIC), Hamdard University, Karachi, Pakistan, 2018. 6 Khan, Asif , Nadeem Javaid, Mahnoor Khan, and Asma Rafique. “Optimum unit sizing of a stand-alone hybrid PV-WT-FC system using Jaya algorithm.” Submitted in International Conference on Cyber Security and Computer Science (ICONCS), Karabuk University (KBU), Turkey, 2018. 7 Khan, Asif , Nadeem Javaid, Adnan Ahmed, Saqib Kazmi, Hafiz Majid Hussain, and Zahoor Ali Khan. “Efficient Utilization of HEM Controller Using Heuristic Optimization Techniques.” In International Conference on Emerging Internetworking, Data & Web Technologies, pp. 573-584. Springer, xiii

Cham, 2017. Download. 8 Ahmed, Adnan, Awais Manzoor, Asif Khan, Adnan Zeb, Hussain Ahmad Madni, Umar Qasim, Zahoor Ali Khan, and Nadeem Javaid. “Performance Measurement of Energy Management Controller Using Heuristic Techniques.” In Conference on Complex, Intelligent, and Software Intensive Systems, pp. 181-188. Springer, Cham, 2017. Download. 9 Kazmi, Saqib, Hafiz Majid Hussain, Asif Khan, Manzoor Ahmad, Umar Qasim, Zahoor Ali Khan, and Nadeem Javaid. “Balancing Demand and Supply of Energy for Smart Homes.” In Conference on Complex, Intelligent, and Software Intensive Systems, pp. 1000-1008. Springer, Cham, 2017. Download. 10 Hafeez, Ghulam, Rabiya Khalid, Abdul Wahab Khan, Malik Ali Judge, Zafar Iqbal, Rasool Bukhsh, Asif Khan, and Nadeem Javaid. “Optimal Residential Load Scheduling Under Utility and Rooftop Photovoltaic Units.” In International Conference on P2P, Parallel, Grid, Cloud and Internet Computing, pp. 142-153. Springer, Cham, 2017. Download.

xiv

Contents Dedication

viii

Acknowledgements

ix

Abstract

x

Journal Publications

xii

Conference Proceedings

xiii

List of Abbreviations

xxvii

List of Symbols

xxxii

1 Introduction 1.1 Energy management in smart cities . . . . . . . . . . . . . . . . . . 1.2 Smart grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Demand side management . . . . . . . . . . . . . . . . . . . 1.2.2 Demand response . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Energy flexibility and buildings . . . . . . . . . . . . . . . . . . . . 1.4 Smart home . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 Smart appliances . . . . . . . . . . . . . . . . . . . . . . . . 1.4.2 Sensors and actuators . . . . . . . . . . . . . . . . . . . . . 1.4.3 Advance metering infrastructure . . . . . . . . . . . . . . . . 1.4.4 Home communication network . . . . . . . . . . . . . . . . . 1.4.5 Home energy management controller . . . . . . . . . . . . . 1.5 Research challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.1 Increased user comfort and priority-induced demand response strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.2 Mitigation of rebound peaks . . . . . . . . . . . . . . . . . . 1.5.3 Environmental issues and towards green energy . . . . . . . 1.5.3.1 Integration of renewable energy sources and energy storage system in a grid-connected system . . . . . xv

1 2 3 5 6 8 9 10 10 11 11 11 13 14 16 17 17

1.5.3.2

1.6 1.7 1.8

Integration of renewable energy sources and energy storage system in a stand-alone system . . . . . . Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . Research contributions . . . . . . . . . . . . . . . . . . . . . . . . Thesis organization . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Meta-heuristic and other optimization schemes along with their application 2.1 Applications of formal and other optimization techniques used for energy management . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1 Limitations of formal approaches . . . . . . . . . . . . . . 2.2 Meta-heuristic approaches . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Meta-heuristic approaches with algorithmic-specific parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1.1 Genetic algorithm . . . . . . . . . . . . . . . . . 2.2.1.2 Enhanced differential evolution algorithm . . . . 2.2.1.3 Binary foraging optimization algorithm . . . . . . 2.2.1.4 Binary particle swarm optimization algorithm . . 2.2.1.5 Wind driven optimization algorithm . . . . . . . 2.2.1.6 Backtracking search algorithm . . . . . . . . . . . 2.2.2 Applications of meta-heuristic approaches with algorithmicspecific parameters . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Meta-heuristic approaches without algorithmic-specific parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3.1 Jaya algorithm . . . . . . . . . . . . . . . . . . . 2.2.3.2 Teaching learning-based optimization . . . . . . . 3 Literature review: subproblem statements and contributions 3.1 Demand side management challenges . . . . . . . . . . . . . . . 3.2 Priority management of appliances . . . . . . . . . . . . . . . . 3.3 Renewable energy sources and energy storage system utilization 3.4 Optimization methods proposed in the literature . . . . . . . . . 3.5 Analysis of literature . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Subproblem statements and contributions . . . . . . . . . . . . 3.6.1 Subproblem-1 and contributions . . . . . . . . . . . . . . 3.6.2 Subproblem-2 and contributions . . . . . . . . . . . . . . 3.6.3 Subproblem-3 and contributions . . . . . . . . . . . . . . 3.6.4 Subproblem-4 and contributions . . . . . . . . . . . . . . 3.6.5 Subproblem-5 and contributions . . . . . . . . . . . . . . xvi

. . . . . . . . . . .

. . . .

18 19 20 23

25 . 26 . 29 . 31 . . . . . . .

31 32 33 33 34 34 35

. 35 . 39 . 40 . 42

. . . . . . . . . . .

44 45 50 52 60 62 63 63 64 66 67 68

3.6.5.1 3.6.5.2 3.6.5.3

HOMER software limitations . . . . . . . . . . . . 68 Formal techniques limitation . . . . . . . . . . . . 69 Algorithmic-specific meta-heuristic limitations . . . 69

4 System models, problem formulations, and proposed solutions 4.1 Priority concept and postulates . . . . . . . . . . . . . . . . . . . 4.1.1 User’s priority and comfort enabled system model . . . . . 4.1.1.1 Transmission and communication infrastructure . 4.1.1.2 Absolute comfort derivation using time and device based priorities . . . . . . . . . . . . . . . . . . . 4.1.1.3 Time based priority table . . . . . . . . . . . . . 4.1.1.4 Device based priority table . . . . . . . . . . . . 4.1.1.5 Absolute comfort table . . . . . . . . . . . . . . . 4.1.1.6 Schematic of the process . . . . . . . . . . . . . . 4.1.2 Proposed solution . . . . . . . . . . . . . . . . . . . . . . . 4.1.2.1 Objective of evolutionary accretive comfort algorithm . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2.2 Consumer’s budget and energy constraints . . . . 4.1.3 Mapping the priority induced scheduling problem with genetic algorithm . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Appliance scheduling problem formulation and proposed solution . 4.2.1 Load categorization . . . . . . . . . . . . . . . . . . . . . . 4.2.1.1 Clothes dryer . . . . . . . . . . . . . . . . . . . . 4.2.1.2 Dishwasher . . . . . . . . . . . . . . . . . . . . . 4.2.1.3 Refrigerator . . . . . . . . . . . . . . . . . . . . . 4.2.2 Peak to average ratio . . . . . . . . . . . . . . . . . . . . . 4.2.3 Threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.4 Knapsack problem formulation . . . . . . . . . . . . . . . 4.2.5 Objective function . . . . . . . . . . . . . . . . . . . . . . 4.2.6 Proposed solution . . . . . . . . . . . . . . . . . . . . . . . 4.2.6.1 Enhanced differential evolution . . . . . . . . . . 4.2.6.2 Genetic algorithm . . . . . . . . . . . . . . . . . 4.2.6.3 Binary particle swarm optimization . . . . . . . . 4.2.6.4 Optimal stopping rule . . . . . . . . . . . . . . . 4.2.7 Proposed system model for appliance schedulling . . . . . 4.2.7.1 Home area network . . . . . . . . . . . . . . . . . 4.2.7.2 Advance metering infrastructure . . . . . . . . . 4.2.7.3 Home energy management controller . . . . . . . xvii

72 . 73 . 75 . 76 . . . . . .

77 77 80 80 82 83

. 83 . 85 . . . . . . . . . . . . . . . . . . .

86 88 88 88 89 90 91 91 92 93 94 94 95 96 98 98 99 100 100

4.3

4.4

4.5

4.2.7.4 Appliances energy consumption pattern . . . . . . 100 Grid-connected photovoltaic, storage and appliances problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 4.3.1 Energy generation model of photovoltaic system . . . . . . . 102 4.3.2 Energy storage model . . . . . . . . . . . . . . . . . . . . . . 103 4.3.3 Energy consumption modeling . . . . . . . . . . . . . . . . . 104 4.3.4 Peak to average ratio for aggregate appliances . . . . . . . . 104 4.3.5 Energy pricing model . . . . . . . . . . . . . . . . . . . . . . 105 4.3.6 Appliance scheduling problem . . . . . . . . . . . . . . . . . 106 4.3.7 System architecture . . . . . . . . . . . . . . . . . . . . . . . 107 4.3.8 Meta-heuristic scheduling algorithms for grid connected system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 4.3.8.1 Genetic algorithm for aggregate appliances . . . . . 109 4.3.8.2 Binary particle swarm optimization for aggregate appliances . . . . . . . . . . . . . . . . . . . . . . . 111 4.3.8.3 Wind driven optimization for aggregate appliances 113 4.3.8.4 Binary foraging optimization for aggregate appliances . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.3.8.5 Hybrid genetic particle optimization . . . . . . . . 117 Stand-alone HRES configuration and sizing formulation for PVWT-FC system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 4.4.1 Photovoltaic power . . . . . . . . . . . . . . . . . . . . . . . 120 4.4.2 Wind turbine power . . . . . . . . . . . . . . . . . . . . . . 121 4.4.3 Accumulative generation and consumer’s load formulation . 122 4.4.4 Storage capacity of hydrogen fuel tanks . . . . . . . . . . . . 122 4.4.5 Calculation of hydrogen fuel tanks . . . . . . . . . . . . . . 123 4.4.6 Cost formulation . . . . . . . . . . . . . . . . . . . . . . . . 124 4.4.6.1 Objective function . . . . . . . . . . . . . . . . . . 124 4.4.6.2 Constraints . . . . . . . . . . . . . . . . . . . . . . 126 4.4.7 Proposed methodology . . . . . . . . . . . . . . . . . . . . . 128 4.4.7.1 Jaya scheme applied for an optimum unit sizing . . 128 4.4.7.2 Genetic algorithm applied for an optimum unit sizing129 4.4.7.3 Backtracking search algorithm applied for an optimum unit sizing . . . . . . . . . . . . . . . . . . . 130 Stand-alone HRES configuration and sizing formulation for PVWT-Battery system . . . . . . . . . . . . . . . . . . . . . . . . . . 132 4.5.1 PV-WT-Battery system model . . . . . . . . . . . . . . . . . 132 4.5.2 Sizing formulation of proposed model . . . . . . . . . . . . . 133 xviii

4.5.2.1 4.5.2.2 4.5.2.3

Sizing formulation of photovoltaic power system . Sizing formulation of wind turbine power system Accumulative power generation by RESs and consumer’s load formulation . . . . . . . . . . . . . . 4.5.2.4 Sizing formulation of battery bank . . . . . . . . Calculation of batteries for battery bank . . . . . . . . . . System reliability . . . . . . . . . . . . . . . . . . . . . . . Total annual cost formulation and constraints . . . . . . . 4.5.5.1 Objective function . . . . . . . . . . . . . . . . . 4.5.5.2 Constraints . . . . . . . . . . . . . . . . . . . . . Proposed methodology . . . . . . . . . . . . . . . . . . . . 4.5.6.1 Hybrid Jaya learning-based optimization . . . . .

. 133 . 134 . . . . . . . . .

135 135 136 137 138 138 140 141 143

5 Simulation results and discussion 5.1 Results on three budget scenarios . . . . . . . . . . . . . . . . . . . 5.1.1 Scenario-1: budget = 1.5 $/day . . . . . . . . . . . . . . . . 5.1.2 Scenario-2: budget = 2.0 $/day . . . . . . . . . . . . . . . . 5.1.3 Scenario-3: budget = 2.5 $/day . . . . . . . . . . . . . . . . 5.1.4 Convergence of the algorithm . . . . . . . . . . . . . . . . . 5.2 Priority-induced results for atomic and aggregate appliances . . . . 5.2.1 Performance parameter definition . . . . . . . . . . . . . . . 5.2.1.1 Cost . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1.2 Delay . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1.3 Energy consumption . . . . . . . . . . . . . . . . . 5.2.1.4 Peak-to-average ratio . . . . . . . . . . . . . . . . . 5.2.2 Scenario-1 atomic appliances . . . . . . . . . . . . . . . . . . 5.2.2.1 Clothes dryer . . . . . . . . . . . . . . . . . . . . . 5.2.2.2 Dishwasher . . . . . . . . . . . . . . . . . . . . . . 5.2.2.3 Refrigerator . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Scenario-2 aggregate appliances . . . . . . . . . . . . . . . . 5.2.3.1 Aggregate appliances without knapsack capacity limit . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3.2 Aggregate appliances with knapsack capacity limit 5.2.4 Feasible region . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.5 Performance parameters trade-off . . . . . . . . . . . . . . . 5.3 Grid-connected renewable energy and storage results . . . . . . . . 5.3.1 Case 1: Integration of renewable and storage system without meta-heuristic algorithms . . . . . . . . . . . . . . . . . . .

147 148 149 154 157 160 162 163 163 163 164 164 164 164 168 170 173

4.5.3 4.5.4 4.5.5

4.5.6

xix

173 176 180 182 183 186

5.3.1.1

5.4

5.5

Energy consumption without meta-heuristic algorithms . . . . . . . . . . . . . . . . . . . . . . . . . 187 5.3.1.2 Electricity cost without meta-heuristic algorithms . 189 5.3.1.3 Total cost without meta-heuristic algorithms . . . . 190 5.3.1.4 Peak to average ratio without meta-heuristic algorithms . . . . . . . . . . . . . . . . . . . . . . . . . 190 5.3.2 Case 2: Integration of renewable and storage system using meta-heuristic algorithms . . . . . . . . . . . . . . . . . . . 191 5.3.2.1 Energy consumption using meta-heuristic algorithms192 5.3.2.2 Electricity cost using meta-heuristic algorithms . . 194 5.3.2.3 Total cost using meta-heuristic algorithms . . . . . 194 5.3.2.4 Peak to average ratio using meta-heuristic algorithms196 5.3.3 Feasible region of the objective function . . . . . . . . . . . 197 Stand-alone renewable and storage results for PV-WT-FC system . 199 5.4.1 Consumer’s low load profile scenario . . . . . . . . . . . . . 201 5.4.2 Consumer’s high load profile scenario . . . . . . . . . . . . . 207 5.4.3 Comparison of Jaya with genetic algorithm and backtracking search algorithm . . . . . . . . . . . . . . . . . . . . . . . . 211 5.4.4 Trade-off among performance parameters . . . . . . . . . . . 211 Stand-alone renewable and storage results for PV-WT-Battery system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 5.5.1 Comparison of Jaya and JLBO algorithms . . . . . . . . . . 218 5.5.2 Algorithms convergence results . . . . . . . . . . . . . . . . 226

6 Conclusion and Future Work 229 6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230 6.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

xx

List of Figures 1.1 1.2 1.3

Key components enabling SH . . . . . . . . . . . . . . . . . . . . . 12 Smart home as integral part of SG . . . . . . . . . . . . . . . . . . 13 Relationship diagram of research challenges, problem statements, contributions, methods, results, and publications . . . . . . . . . . . 23

2.1

Flowchart of Jaya algorithm . . . . . . . . . . . . . . . . . . . . . . 42

3.1

Energy sources for hybrid system . . . . . . . . . . . . . . . . . . . 60

4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 4.13 4.14 4.15 4.16 4.17

Consumer priority enabled system model for a home . . . . . . . . . Flow chart of time based appliances priority . . . . . . . . . . . . . Absolute user comfort derived from time and device based priorities Schematic of the process . . . . . . . . . . . . . . . . . . . . . . . . Proposed system model . . . . . . . . . . . . . . . . . . . . . . . . . Appliance energy profile data . . . . . . . . . . . . . . . . . . . . . Proposed SH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Working flow of GA . . . . . . . . . . . . . . . . . . . . . . . . . . . Steps involved in BPSO algorithm . . . . . . . . . . . . . . . . . . . Main steps of WDO algorithm . . . . . . . . . . . . . . . . . . . . . Steps involved in BFO algorithm . . . . . . . . . . . . . . . . . . . Steps involved in HGPO algorithm . . . . . . . . . . . . . . . . . . Proposed model for HRES . . . . . . . . . . . . . . . . . . . . . . . Proposed system model for HRES . . . . . . . . . . . . . . . . . . . Flowchart of calculating the hybrid system reliability . . . . . . . . Schematic diagram of the process . . . . . . . . . . . . . . . . . . . Flowchart of JLBO algorithm . . . . . . . . . . . . . . . . . . . . .

75 78 81 82 99 101 108 111 113 116 118 119 120 133 138 142 144

5.1 5.2 5.3 5.4 5.5 5.6

Budget 1.5 $/day . . Budget 2.0 $/day . . Budget 2.5 $/day . . Convergence result at Convergence result at Convergence result at

151 156 159 161 161 162

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . budget 1.5 $/day budget 2.0 $/day budget 2.5 $/day xxi

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 5.15 5.16 5.17 5.18 5.19 5.20 5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.30 5.31 5.32 5.33 5.34 5.35 5.36 5.37 5.38 5.39 5.40 5.41 5.42 5.43 5.44

DA-RTP signal . . . . . . . . . . . . . . . . . . Average monthly cost of clothes dryer . . . . . . Average cost of clothes dryer . . . . . . . . . . . Average delay of clothes dryer . . . . . . . . . . Average monthly cost of dishwasher . . . . . . . Average cost of dishwasher . . . . . . . . . . . . Average delay of dishwasher . . . . . . . . . . . Average monthly cost of refrigerator . . . . . . Average cost of refrigerator . . . . . . . . . . . Average delay of refrigerator . . . . . . . . . . . Average monthly cost of appliances . . . . . . . Average cost of appliances . . . . . . . . . . . . Average delay of appliances . . . . . . . . . . . Energy consumption of appliances . . . . . . . . PAR of appliances . . . . . . . . . . . . . . . . Average monthly cost of appliances (knapsack) . Average cost of appliances (knapsack) . . . . . . Average delay of appliances (knapsack) . . . . . Energy consumption of appliances (knapsack) . PAR of appliances (knapsack) . . . . . . . . . . FR cost-energy of aggregate appliances . . . . . FR cost-delay of aggregate appliances . . . . . . DAP signal . . . . . . . . . . . . . . . . . . . . Forecasted outdoor temperature . . . . . . . . . Solar irradiance . . . . . . . . . . . . . . . . . . Estimated renewable energy . . . . . . . . . . . Charging level of ESS . . . . . . . . . . . . . . . Load distribution . . . . . . . . . . . . . . . . . Energy consumption in case 1 . . . . . . . . . . Electricity cost in case 1 . . . . . . . . . . . . . Total electricity cost of the prosumer in case 1 . Case 1 PAR . . . . . . . . . . . . . . . . . . . . Hourly energy consumption in case 2 . . . . . . Hourly electricity cost in case 2 . . . . . . . . . Case 2 total cost . . . . . . . . . . . . . . . . . Case 2 PAR . . . . . . . . . . . . . . . . . . . . Feasible region of objective function . . . . . . . Hourly insolation profile data during a year . . . xxii

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

163 166 166 166 169 169 170 171 172 172 174 175 175 175 176 178 178 178 179 179 181 182 185 185 185 186 187 187 189 189 190 191 192 195 196 197 199 200

5.45 5.46 5.47 5.48 5.49 5.50 5.51 5.52 5.53 5.54 5.55 5.56 5.57 5.58 5.59 5.60 5.61 5.62 5.63 5.64 5.65 5.66

5.67 5.68

5.69

Hourly wind speed profile data during a year . . . . . . . . . . . . . Hourly conumser’s load profile data during a year . . . . . . . . . . Hourly produced PVs power for PV-WT-FC system during a year . Hourly produced WTs power for PV-WT-FC system during a year . Hourly expected mass of stored energy in HFTs for PV-WT-FC system during a year . . . . . . . . . . . . . . . . . . . . . . . . . . Convergence of Jaya algorithm for PV-WT-FC system . . . . . . . Hourly produced PVs power for PV-FC system during a year . . . . Hourly expected mass of stored energy in HFTs for PV-FC system during a year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Convergence of Jaya algorithm for PV-FC system . . . . . . . . . . Hourly produced PVs power for WT-FC system during a year . . . Hourly expected mass of stored energy in HFTs for WT-FC system during a year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Convergence of Jaya algorithm for WT-FC system . . . . . . . . . . Hourly high load profile during a year . . . . . . . . . . . . . . . . . Convergence process of Jaya algorithm for finding the optimum sizing at high load profile . . . . . . . . . . . . . . . . . . . . . . . . . Convergence process of Jaya, GA, and BSA algorithms for finding the optimum sizing at high load profile . . . . . . . . . . . . . . . . PV-FC system results at different LPSP . . . . . . . . . . . . . . . Hourly insolation profile data . . . . . . . . . . . . . . . . . . . . . Hourly ambient temperature profile data . . . . . . . . . . . . . . . Hourly wind speed profile data . . . . . . . . . . . . . . . . . . . . . Hourly consumer load profile data . . . . . . . . . . . . . . . . . . . Hourly produced power and storage level of PV-WT-Battery system by JLBO algorithm during a year at different LP SP max values . . . Hourly produced power and storage level of PV-WT-Battery system by JLBO algorithm during first eight days of the year at different LP SP max values . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hourly produced power and storage level of PV-Battery system by JLBO algorithm during a year at different LP SP max values . . . . Hourly produced power and storage level of PV-Battery system by JLBO algorithm during first eight days of the year at different LP SP max values . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hourly produced power and storage level of WT-Battery system by JLBO algorithm during a year at different LP SP max values . . . .

xxiii

200 202 203 204 205 205 206 207 207 208 208 209 209 210 212 213 214 215 216 217 222

223 224

225 226

5.70 Hourly produced power and storage level of WT-Battery system by JLBO algorithm during first eight days of the year at different LP SP max values . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 5.71 Convergence process of JLBO algorithm for optimum results at different LP SP max values . . . . . . . . . . . . . . . . . . . . . . . . . 228

xxiv

List of Tables 1.1

A brief comparison of traditional grid and smart grid. . . . . . . . .

2.1 2.2

Formal and other approaches used in energy management . . . . . . 30 Meta-heuristic algorithms used in literature . . . . . . . . . . . . . 40

3.1

Schemes and user comfort . . . . . . . . . . . . . . . . . . . . . . . 48

4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10

Energy consumption and cost . . . . . . . Time based priority . . . . . . . . . . . . . Device based priority . . . . . . . . . . . . Absolute comfort . . . . . . . . . . . . . . Load categorization . . . . . . . . . . . . . GA parameters . . . . . . . . . . . . . . . BPSO parameters with values . . . . . . . WDO parameters . . . . . . . . . . . . . . BFO parameters . . . . . . . . . . . . . . Hybrid system components and parameters

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

Budget 1.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . EACA performance at budget = 1.5 $/day . . . . . . . . . . . . . Total achieved user comfort table due to the energy allocation pattern at budget 1.5 $/day . . . . . . . . . . . . . . . . . . . . . . . 5.4 Base case sample scenario-1 with budget 1.5 $/day . . . . . . . . 5.5 Base case sample scenario-2 with budget 1.5 $/day . . . . . . . . 5.6 Comparison between optimal and randomly generated cases at budget = 1.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7 Total operational time comparison of all appliance at budget = 1.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.8 Budget 2.0 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.9 EACA performance at budget = 2.0 $/day . . . . . . . . . . . . . 5.10 Total achieved user comfort table due to the energy allocation pattern at budget 2.0 $/day . . . . . . . . . . . . . . . . . . . . . . .

5.1 5.2 5.3

xxv

. . . . . . . . . .

5

76 79 81 82 107 110 113 115 115 142

. 149 . 150 . 151 . 152 . 153 . 153 . 154 . 155 . 155 . 156

5.11 Comparison between optimal and randomly generated cases at budget = 2.0 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.12 Total operational time comparison of all appliance at budget = 2.0 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.13 Budget 2.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.14 Total achieved comfort table due to the energy allocation pattern at budget 2.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . 5.15 EACA performance at budget = 2.5 $/day . . . . . . . . . . . . . 5.16 Comparison between optimal and randomly generated cases at budget = 2.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.17 Total operational time comparison of all appliance at budget = 2.5 $/day . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.18 Energy profile of appliances . . . . . . . . . . . . . . . . . . . . . 5.19 Comparison of cost and delay: clothes dryer . . . . . . . . . . . . 5.20 Comparison of cost and delay: dishwasher . . . . . . . . . . . . . 5.21 Comparison of cost and delay: refrigerator . . . . . . . . . . . . . 5.22 Comparison of cost and delay: aggregate appliances . . . . . . . . 5.23 Comparison of cost and delay: aggregate appliances (knapsack) . 5.24 Possible cases - clothes dryer . . . . . . . . . . . . . . . . . . . . . 5.25 Computational time . . . . . . . . . . . . . . . . . . . . . . . . . 5.26 Description of the home appliances . . . . . . . . . . . . . . . . . 5.27 Thresholds of energy consumption . . . . . . . . . . . . . . . . . . 5.28 Comparison of case 1 cost . . . . . . . . . . . . . . . . . . . . . . 5.29 Comparison of case 1 PAR . . . . . . . . . . . . . . . . . . . . . . 5.30 Comparison of case 2 cost . . . . . . . . . . . . . . . . . . . . . . 5.31 Comparison of case 2 PAR . . . . . . . . . . . . . . . . . . . . . . 5.32 Possible cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.33 System component parameters . . . . . . . . . . . . . . . . . . . . 5.34 Jaya results for the proposed hybrid systems . . . . . . . . . . . . 5.35 Jaya results for the proposed hybrid systems at high load . . . . . 5.36 Jaya results for PV-FC system at different LPSP using high consumer’s load profile . . . . . . . . . . . . . . . . . . . . . . . . . . 5.37 Summary of mean, standard deviation, best performance, worst performance, and ranks of the schemes over the proposed hybrid systems for LP SP max = 1% . . . . . . . . . . . . . . . . . . . . . 5.38 Summary of Jaya and JLBO results for the proposed hybrid systems at different LP SP max values . . . . . . . . . . . . . . . . . . . . .

xxvi

. 157 . 157 . 158 . 159 . 159 . 160 . . . . . . . . . . . . . . . . . . .

160 163 165 168 171 174 177 180 184 184 188 190 191 195 197 198 201 203 208

. 213

. 215 . 219

List of Abbreviations A ACs ABSO ACO AMI AEDB AI ADA AMR ANN ABC B BPSO BFO BSA C CPP CD CP COP CR CRF CSA CSOA CS D DG DSM DR DAP -

Air conditioners Artificial bee swarm optimization Ant colony optimization Advanced metering infrastructure Alternative energy development board Artificial intelligence Activity dependent appliances Automatic meter reading Artificial neural network Artificial bee colony Binary particle swarm optimization Bacterial foraging optimization Backtracking search algorithm Critical peak pricing Clothes dryer Convex programming Cut-off point Crossover rate Capital recovery factor Crow search algorithm Cuckoo search optimization algorithm Cuckoo search Distributed generation Eemand side management Demand response Day ahead pricing xxvii

DLC DA-RTP DRP DERs DBR DP DC DB DP DOD DE E EMC ESS EA EDE EDTLA EACA EMS ET HEMDAS EP F FC FCFS FASA FPA FA G GA GC GCM H HEMC HGPO HRESs HEM HS -

Direct load control Day ahead RTP Demand response program Distributed energy resources Declining block rate Dynamic programming Direct current Device based Dynamic programming Depth of discharge Differential evaluation Energy management controller Energy storage system Evolutionary algorithm Enhanced differential evolution Enhanced differential teaching-learning algorithm Evolutionary accretive comfort algorithm Energy management system Electricity tariff Electrical and thermal appliance scheduling Evolutionary programming Fuel cell First come first serve Firefly algorithm-based sizing algorithm Flower pollination algorithm Firefly Algorithm Genetic algorithm Grid connected Graphical construction method Home energy management controller Hybrid GA-PSO Hybrid renewable energy systems Home energy management Harmony search xxviii

HSA HSDE HEMS HOMER HAN HFT HVAC I ICTs IMUs ILP ITLBO IBR IoT IPSO L LOS LoT LP LCC LPSP LTE LSA M MINLP MILP MPPT MKP MFCFS MF MCS MO MC MBA N NIST NAN O

Harmony search algorithm Harmony search differential evolution Home energy management system Hybrid optimization model for multiple energy resources Home area network Hydrogen fuel tanks Heating ventilation and air conditioning Information and communication technologies Information management units Integer linear programming Improved TLBO Inclined block rate Internet of things Improved-PSO Loss of supply Length of operational time Linear programming Life cycle cost Loss of power supply probability Long term evolution Lighting search algorithm Mixed integer nonlinear programming Mixed integer linear programming Maximum power point tracking Multiple knapsack problem Modified FCFS Mutation factor Master controller system Microwave oven Mobile charger Mine blast algorithm National institute of standards and technology Neighbourhood area network

xxix

OHEMS OSR ODA OIA OD OI OTI OLA P PLP PAR PSO PV PEEDF PHEVs PP R RESs RSERs RTP RSM ROI RSC RL S SCs SHs SG SMs SFL SCADA SA SKP SS SOC Std. T TG -

Optimized HEMS Optimal stopping rule Occupancy dependent appliances Occupancy independent appliances Occupancy dependent Occupancy independent Operational time interval Observe, learn and adapt algorithm Peak load pricing Peak-to-average ratio Particle swarm optimization Photovoltaic Priority enables early deadline first Plug-in hybrid electric vehicles Peak pricing Renewable energy sources Renewable and sustainable energy resources Real-time pricing Realistic scheduling mechanism Return on investment Required storage capacity Reinforcement learning Smart cities Smart homes Smart grid Smart meters Shuffled frog leaping Supervisory control and data acquisition system Stand alone Single knapsack problem Smart scheduler State of charge Standard deviation Traditional grid xxx

ToU ToUP TNPC TS TLBO TLGO TB TPR TEA TW TSR TAC TF U UI UMTS W WM WTs WDO WAN WH -

Time-of-Use Time of Use pricing Total net percent cost Tabu search Teaching learning based optimization Teacher learning genetic optimization Time based Total power rating Total energy available Tera watts tip speed ratio Total annual cost Teaching factor User inconvenience Universal mobile telecommunication system Washing machine Wind turbines Wind driven optimization Wide area network Water heater

xxxi

List of Symbols χρτ or t τr βµµdesired µachieved Texp T EA n or i Acd Adw Aref Pcd χb ρep ρad M ζcd Y ζcd H ζcd εcd Pdw εdw D ζdw M ζdw Y ζdw H ζdw Pref -

Cost per unit comfort index Appliances priority Time interval or slot Specific time-slot range between t20 and t24 Consumer budget limit User comfort Total desired comfort Total achieved comfort Total users’ expenditure Total energy available Number of appliances Appliance clothes dryer Appliance dishwasher Appliance refrigerator Power rating of the clothes dryer Boolean integer Electric price DA-RTP Air density Clothes dryer’s monthly cost Clothes dryer’s yearly cost Clothes dryer’s hourly cost Total energy consumption per day of clothes dryer Power rating of dishwasher Total energy consumption per day of dishwasher Total cost per day of dishwasher Dishwasher monthly cost Dishwasher yearly cost Dishwasher hourly cost Power rating of the refrigerator xxxii

εref D ζref M ζref Y ζref H ζref εAn ζAn P AR µp Z∗ ρep p ep ρo Γjmax ui vi xi pc pm vi φ1and2 sig(i, j) EP V Xgbest η P V or η pv Xlbest AP V or Apv Fpg Ir or I FC Ta FG α1 , α2 β1 , β2 ES FF κFG η ESS -

Total energy consumption per day of the refrigerator Refrigerator per day cost Refrigerator monthly cost Refrigerator yearly cost Refrigerator hourly cost Total energy consumption by three appliances Total cost of three appliances Peak to average ratio Time factor (Appliance’s priority) Threshold Maximum electricity price value Minimum electricity price value Knapsack capacity limit Trial vectors Mutant vector Target vector Crossover rate of GA Mutation rate of GA Initial velocity of BPSO particles Acceleration constants of BPSO Sigmoid function Available energy from PV system Global best value Efficiency of PV system Local best value Area occupied by PV panels Pressure gradient force Solar radiation Coriolis force Ambient temperature Gravitational force Shape factors Scale factors Stored energy Friction force Duration of one time-slot Gravitational force Efficiency of ESS xxxiii

Pold EP Ch Pmax EP Dch αEPUCh B ωDch EPLB ∆t ES U B νnew M νold N REa Sig Eb E total P DAP EPa Ci EPb θEP Ωa XmM δν b XnβN wEgrid r1 , r2 ,r min Eunsch c1 τ0 c2 τsch dimM ax -

Pressure at current location Charge rate of ESS at time High pressure point Discharge rate of ESS at time Constant for update position Upper charge limit of ESS Earth rotation Lower discharge limit of ESS Unit step time Upper limit of energy storage Updated velocity Number of controllable appliances Current velocity Number of un-shiftable appliances Universal gas constant Energy consumption of shiftable appliances Sigmoid function Energy consumption of non-shiftable appliances Total energy consumption DAP signal Electricity cost of shiftable appliances energy consumption Step size Electricity cost of non-shiftable appliances energy consumption Position of bacteria Total bill of energy consumption Rotation of earth ON/OFF status of shiftable appliances Finite volume of air ON/OFF status of non-shiftable appliances Inertia factor Available grid energy Random numbers Minimum energy consumed in unscheduled scenario Local pull Lower limit of scheduling horizon Global pull Scheduling time Upper limit of WDO dimensions xxxiv

τmax dimM in ∆Lµvw Vit+1 Vjt xt+1 i xti Ne Pc Nc Pm Np wi Ns wf Nr vmax Ped vmin LP SP max LP SP P OW pv N pv ξ pv P OW wt κvωp θPrwt vr v ci v co N wt ξ wt ξ gen -

Upper limit of scheduling horizon Lower limit of WDO dimensions Pressure gradient Length of chromosomes Velocity vector of wind Particle upcoming velocity Particle current velocity Particle upcoming position Particle current position Number of elimination steps Probability of crossover Number of chemotaxis steps Probability of mutation Number of population steps Initial weight constant Number of swimming steps Final weight constant Number of reproduction steps Upper limit of velocity Probability of elimination-dispersal Lower limit of velocity Maximum loss of power supply probability Loss of power supply probability PV power Number of PV systems Total produced PV power WT power Density of air velocity of wind Coefficient of power Pitch angle WT rated power Rated wind speed Cut-in wind speed Cut-out wind speed Number of WTs Total produced WT power Accumulative electricity generation from PV and WT xxxv

ηi ξ ld pχξ Store ξ Store (t) ξ Store (t − 1) ηe ηf c Nt RSC temp max(temp) min(temp) ζ tot ζ c or ζ cap ζ m or ζ mtn CRF RoI or ir nn or n ζF E ρF C and ρElect ρinst. F C/Elect ζ inv/conv ρInv/Conv ζ wt ζ pv ζt ζF E ζ inv/conv pv ζm wt ζm FC ζm E ζm Store Store ξmax and ξmin Rl,m,t ‘ Rl,m,t Rl,best,t -

Efficiency of the inverter Consumer’s load Power ratings Boolean integer showing an appliance status Amount of hydrogen stored in the HFT Stored amount of energy in the hydrogen tanks at time-slot t Stored amount of energy in the hydrogen tanks at time-slot t − 1 Efficiency of electrolyzer Overall efficiency of DC/DC converter along with FC Total number of HFTs Required storage capacity Temporary storage variable Maximum generation points of temp Minimum generation points of temp Total annual cost Annual capital cost Annual maintenance cost Capital recovery factor Rate of interest Lifespan of the system in years Present worth of the FC and electrolyzer systems Price of FC and electrolyzer Installation fee for both FC and electrolyzer Present worth of the inverter/converter Price of the inverter/converter Cost of WT Cost of PV Cost of hydrogen storage tank Cost of FC/electorlyser Cost of inverter/converter Annual maintenance costs of PV Annual maintenance costs of WT Annual maintenance costs of FC Annual maintenance costs of electrolyzer Maximum and minimum storage capacity of HFTs Candidate solution during the tth iteration Updated value of Rl,m,t Best candidate at tth iteration xxxvi

Rl,worst,t Worst candidate at tth iteration rand1,l,t and rand2,l,t - Random numbers P OP or X Initial population oldP OP Historical population T Trial population I Solar radiation Rated PV power Prpv ref I Solar radiation at reference conditions T cof PV panels temperature coefficient ref T PV cell temperature at reference conditions Tc Temperature of PV cell amb T Ambient air temperature T noct Normal operating cell temperature v Speed of the wind Prwt Nominal power of WT ηi Efficiency of the inverter ξ StoreB Stored amount of energy in the battery bank ιSelf discharging state ηb Battery bank charging efficiency tTime-slot rsc T Difference between max and min points in the temp curve b N Number of batteries needed for battery bank LOP S Loss of power supply tac ζ Total annual cost b ζ Present worth of the battery ρb Battery price b ζ Unit cost of battery DOD Depth of discharge of battery pv,max N Maximum number of PV panels N wt,max Maximum number of WTs b,max N Maximum number of batteries l Xteacher Teacher in TLBO algorithm Tf actor Teaching factor l Xm Learner m in population X l Xn Learner n in population X Xnew New population generation

xxxvii

Chapter 1 Introduction

1

Chapter 1

1.1

1.1. ENERGY MANAGEMENT IN SMART CITIES

Energy management in smart cities

Currently, 54% of the world’s population is living in urban areas, which was only 30% in 1950 and it is expected to rise to 66% by 2050 [1]. The shift towards smart cities (SCs) is emerging due to the usage of information and communication technologies (ICTs), including sensors, context awareness, and cognitive learning which provides the opportunities to make their citizen lives more effective, efficient, sustainable, and comfortable [2, 3]. A SC is defined by six characteristics: smart environment, smart people, smart governance, smart economy, smart mobility, and smart living [4]. Smart environment depicts an attractive and clean natural condition with the least pollution and sustainable management of resources. Smart people are distinguished by creativity, qualification level, open-mindedness, affinity for learning, and public life participation. Smart governance deals with the transparent government that involves its citizens in decision-making and provides easy access to the public, and social services. The elements that contribute to making a city with a smart economy are an entrepreneurial spirit, labor flexibility, productivity, and international embeddedness. The key factors for smart mobility are safe, sustainable, innovative transport systems, and ICT infrastructure availability. Finally, citizens health conditions, living standards, safety, cultural, and educational facilities are the key indicators that contribute to smart living. Thus, to make SCs more sustainable and competitive, an optimal management of SC activities and resources is needed. These resources require energy for its functioning; therefore, energy efficiency and its management are the key requirements of the SCs. The SCs requirements for energy are abundant and complex. For instance, the five energy-related intervention areas identified are infrastructure, generation, transport, facilities, and storage [5]. All these areas contribute to the energy system using distinct ways: infrastructure helps energy distribution and provides user

2

Thesis by: Asif Khan

Chapter 1

1.2. SMART GRID

interfaces; generation produces energy; transport and facilities are consumer’s of energy, and storage provides reliability via energy availability. To overcome the energy problems in SCs, traditional grids (TGs) are being replaced by smart grids (SGs).

1.2

Smart grid

The national institute of standards and technology (NIST) in its document framework and roadmap for SG interoperability standards defined SG as a combination of computing and communication services integrated with power infrastructure. It enables a two-way communication flow of energy and control capabilities. The SG will open an array of new applications and functionalities that will go well beyond smart meters (SMs) for businesses and homes [6]. The European technology platform (European Commission, 2006) defines the SG as, “ SG consists of an electricity network that can intelligently integrate the user’s actions and incorporates the latest technologies to efficiently deliver economic, sustainable, and secure electricity supplies”. SG has different kinds of energy and operational measures like SMs, renewable energy, smart appliances, and storage resources. The vital aspect of the SG is the control of power production, transmission, and distribution through advanced ICTs. The ICTs enable SG to send control commands within the time limits defined by numerous international standards, e.g., IEEE standards 1547 (i.e., the standards defined for the control and management of distributed energy resources (DERs)) [7]. Moreover, SG makes possible the access of the power system operator and end-users at the same time intelligently and efficiently. Thus the vital objectives of the SG are to improve the safety, efficiency, and reliability of the entire system [8]. In recent decades, energy demand around the globe has shown an increasing trend. In past, fossil fuels are the most used sources for power generation. However, to fulfill the growing electricity demand with minimal greenhouse gas emissions, 3


Chapter 1

1.2. SMART GRID

scientists have discovered the new methods of electricity generation: renewable and sustainable energy resources (RSERs). However, this integration of renewable energy sources (RESs) has increased the dynamics and complexity of the power system. It is difficult for the existing power system to maintain its stability if RESs integration and distributed generation (DG) are done on a large scale [9]. In this context, present solutions offer the transformation of the existing power grid into a SG with cutting-edge ICTs [10]. Thus SG enabled with ICTs enhances the reliability and stability of the power system via integration of RESs and DG. The TG suffers from deficiencies like monitoring of transmission lines, bi-directional communication flow, fault detection, real-time support, and self-healing [11]. The advent of SG has emerged to cope with the existing limitations of TG. The SG integrates advanced metering infrastructure (AMI) that supports bi-directional communication between a utility and electricity consumers using advanced communication technologies along with an intelligent control system [12]. The key factors that make SG superior over TGs are two-way communication, AMI, and information management units (IMUs). The SG introduces intelligence, automation, and real-time control to the power system. Further, the two-way communication in SG keeps the end-users well informed about the varying electricity prices, maintenance schedules of the distribution network, and events/failures that come either due to equipment failures or natural disasters. It also enables an operator to monitor the real-time data of energy consumption to make a real-time decision about the operational activities. A comprehensive comparison of TG and SG features is shown in Table 1.1. In SG, the energy management is broadly classified into two different categories: supply side management and demand side management (DSM). The former category deals with the efficient and reliable energy generation, transmission, and distribution to the consumers at minimum cost [13]. The latter category that aims at planning, monitoring, and scheduling activities within a smart home (SH) [14, 15, 16]. 4


Chapter 1

1.2. SMART GRID

Table 1.1: A brief comparison of traditional grid and smart grid. Infrastructure

Traditional grid Centralized generation and uni-directional flow of energy from utility to the consumers.

Smart grid Decentralized generation and bi-directional flow of energy between the prosumers and the utility.

Power losses

High power losses due to centralized structure and inadequate storage facilities.

Significantly reduces the power losses due to DG at distribution level (i.e., the DG eliminates the losses of transmission network).

Information system

Aged metering and monitoring system.

Communication infrastructure

The technology used is wired.

Advanced metering and monitoring system: AMI and supervisory control and data acquisition system (SCADA). Both wireless and wired technologies are used.

Energy storage systems (ESSs)

Main storage facility is pump-hydro power plants.

Facilitate the distributed ESSs integration.

RSERs

Mainly includes dispatchable RESs (Hydro-power plants)

Provides decentralized control for RSERs (solar, wind, tidal, geothermal and biomass energies, etc.).

Self-healing

Find system faults and react to stop further damage by protecting assets.

Automatically senses and reacts to emerging and actual contingencies. The focus is on prevention.

Power system

Optimization of assets

Consumer engagement

Power quality

1.2.1

Negligible incorporation of limited operational data and time based (TB) maintenance. No proper involvement of the consumers in DSM and DR activities (i.e., no mechanism to send the varying electricity prices to the consumers in real-time and forced load shedding is carried out to maintain the balance between supply and demand). The focus only on the reduction of failures and interruptions.

Expanded measurement and sensing of grid technologies and conditions to effectively manage assets.

Provides dynamic pricing, net metering, and other incentive-based schemes.

Ensure the quality of electricity for the smooth operations of sensitive electronic devices.

Demand side management

The SG makes the integration of RESs and DGs practicable. It involves the residential as well as commercial users into the DSM and demand response (DR) activities [17]. The DSM allows consumers to change their energy demand during on-peak hours to minimize electricity usage. This goal of DSM is achieved via different methods such as providing financial incentives to electricity consumers or causing behavioral change through education. Usually, the objective of the DSM is to encourage the electricity consumers to use less amount of energy during the on-peak hours or shift their load to off-peak hours [18]. For the stability of the grid, it is necessary to balance the demand and supply of energy. This was previously achieved via turning on the peak power plants 5


Chapter 1

1.2. SMART GRID

to meet the additional energy requirements. DSM comprises various strategies and techniques that are used to encourage electricity consumers’ to reduce their domestic load during on-peak hours. In this way, consumers can optimize energy consumption and efficiently utilize the limited energy resources. Thus, DSM plays a key role in enhancing the overall power system efficiency. The DSM strategies are peak clipping, load shifting, valley filling, and energy conservation [19]. Peak clipping is defined as a reduction in consumer load demand during high peak hours. This strategy can be achieved by switching-off the unnecessary consumer’s load via direct load control (DLC) which results in a reduced amount of total load. In load shifting technique, incentive-based schemes are used to encourage consumer’s to shift the domestic load from on-peak time-slots to off-peak time-slots. Here, the total load is not reduced as energy consumption at the end of the time interval remains the same. Valley filling strategy is used to build the load during low peak hours, as energy generation is high. Finally, energy conservation is an effort made to decrease the energy consumption via using less of an energy service, i.e., driving less.

1.2.2

Demand response

The term DR represents programs designed to encourage electricity consumers to make short-term demotions in their energy consumption in response to an electricity price signal received from the utility. Thus, the DR refers to various actions that are used by the utility to reduce the energy demand of consumer’s in response to the variable market electricity prices [20]. Therefore, DR adjusts the power demand instead of adjusting the power supply. However, the delay in appliance usage because of load shifting to the off-peak hours may cause discomfort to the electricity consumers [21]. Thus, consumer’s comfort is an essential factor to be considered with DR programs (DRPs).

6


Chapter 1

1.2. SMART GRID

The DRPs are classified into incentive based and price based categories [22, 23]. In the incentive-based category, electricity consumers are being provided with some incentives to take part in DRPs and curtail their energy usage in system stress time or critical peak time-slots. Some of the incentive-based programs are DLC, interruptible programs, emergency programs, and demand bidding. In DLC, on a short notice or prior agreement with the consumer, the utility shuts down the electric power for a short period of time. Appliances with high consumption power, i.e., ACs, heaters, water pumps, etc. are usually controlled by DLC [24]. Interruptible programs provide incentives to the consumers in bills if a timely response is triggered (30 to 60 min), otherwise, penalized based on the terms of the program. This occurs for limited hours, i.e., 200 hrs/year approximately [25]. In emergency programs, consumers are encouraged voluntarily to curtail their load during an emergency situation for incentives [26]. Customers provide a bid price to the utility on which they are willing to curtail their load. The customers are penalized if they do not abide by the rules of the program once the bid is finalized. The benefit of this program is continuous energy supply as the load is not disconnected. In traditional electromechanical meters where the usage of electricity is recorded once a month use flat rate pricing (FRP) which charge consumers at fixed price/kWh. However, the SMs are now replacing these traditional electromechanical meters where electricity usage is recorded in real time. Thus, utility companies are motivated to introduce different pricing schemes to control the gap between supply and demand. These prices based DRPs are in the form of financial incentives or other time-based rates, which provide an ample opportunity for consumers in shifting or reducing appliances’ load to off-peak hours. Thus, DR is considered the most cost-effective, economical, and reliable solution that not only reduces the peak energy demand but also smoothens the demand curve during the system stress time. The motivation is offered by the DR in the form of price signals, including critical peak pricing (CPP), time-of-use (ToU), inclined block rate (IBR), peak load 7


Chapter 1

1.3. ENERGY FLEXIBILITY AND BUILDINGS

pricing (PLP), real-time pricing (RTP), day-ahead pricing (DAP) or day-ahead RTP (DA-RTP). In ToU pricing, the complete day is divided into three different time-slots having low, mid, and high peak hours. Mostly, ToU prices are monthly or season based and announced in advance to the consumers. Various pricing signals have limitations; for instance, in ToU scheduling is performed on a day-ahead basis while usage of RTP requires continuous real-time communication between the consumers and the utility. Thus usage of RTP may cause network congestion problem along with data losses. Therefore, an alternate solution proposed for the RTP is DA-RTP, where real-time prices are predicted and announced to the customers beforehand. The consumers are then charged on the basis of this dayahead price signals [27]. The CPP has a very high rate for a short time interval encouraging consumers to consume less energy during that span. The common goals of DRPs and DSM in SG are to establish two-way interaction between consumers and utility so that users are motivated to reduce their electricity cost or energy usage during peak time-slots. In this way, not only the reliability of the grid is improved, but also utilities have a chance to supply uninterrupted electricity to its consumers. Thus, the core role of DRPs is to provide energy flexibility.

1.3

Energy flexibility and buildings

Flexibility refers to the ability of an electric system which is capable of maintaining a balance between supply and demand [28]. Energy flexibility and stability are very important factors needed for the power system. The core advantages of energy flexibility are the improvement in the grid reliability, reduction in the peak load, and addressing generation-load imbalances [29]. Buildings as electric prosumers have a distinct role due to their flexible energy consumption and distribution. However, flexibility varies among different types of buildings [30].

8


Chapter 1

1.4. SMART HOME

The three types of buildings are residential, commercials, and industrials. Residential buildings are equipped with smart appliances that can provide great energy flexibility. Depending on the types of appliances flexibility also varies. For instance, refrigerators and freezers are less flexible as compared heating and ventilation appliances [31]. Commercial buildings comprising of hotels, hospitals, offices, and stores are more reluctant to participate in DRPs because of the effect on their profits and business [32]. Industrial buildings are composed of various industries that are engaged in various processes, including food, steel, textile, etc., and other technologies [33]. Industrial buildings are also reluctant to shift their energy usage considering their big profits [32]. Since residential buildings are more flexible as compared to commercial and industrial buildings, therefore, in this dissertation, we have focused on energy management of the residential sector. Further, for this work, the terms buildings and homes are used synonymously.

1.4

Smart home

According to [34], 40% of global energy is approximately consumed by buildings. The demand for energy in buildings’ is growing worldwide [35, 36]. The residential sector consisting of SHs has great potential to improve energy efficiency. The SH uses ICT to control and automate its features, including appliances, lighting, windows, etc., [37]. In a SH, various appliances are installed, which need to be efficiently managed by a home energy management controller (HEMC). The HEMC plays a very crucial role in finding an optimum, economical, and reliable energy solution that achieves the highest user comfort (UC) value at a reduced cost and also curtails the peak load. The UC is defined as an indicator or measure of the degree to which a consumer of a SH is pleased with the installed appliances’ utility. The appliance utility refers to the usage of electrical devices within the user desired horizon. The key components of SH are discussed next.

9


Chapter 1

1.4.1

1.4. SMART HOME

Smart appliances

Smart appliances are embedded with the intelligence and communication features that allow them to be connected to a HEMC. These devices shift their load from on-peak to off-peak hours based on the decision signals received from HEMC. These appliances may shift their entire operational cycle to other time-slots or work at reduced power to save consumer’s energy cost. Washing machine (WM), refrigerator, clothes dryer (CD), air conditioners (ACs) are some examples of domestic appliances which are being made smart. For instance, a smart WM can operate only in low peak hours or it can utilize photovoltaic (PV) energy generation. Similarly, a smart refrigerator is able to shift its defrost cycles to low peak hours in the night. Many appliance companies: Panasonic, LG, Samsung, and Whirlpool have invested in smart appliances’ production [38].

1.4.2

Sensors and actuators

In recent days, most of the devices are equipped with sensors and actuators facilities. For instance, in the SH, window blinds can be open and closed, appliances turned ON and OFF, and control of the thermostat enabled appliances is managed by consumer sitting away from the home. Thus, home sensors for energy management may comprise for detection and measuring of motion, temperature, voltage, current, light and occupancy [39]. Intrusive and non-intrusive methods are used to monitor the load [40]. Intrusive load monitoring requires an installation of measuring devices at every load of interest. The non-intrusive method does not need individual or group of appliances to monitor their specific power consumption thus, are more cost-effective and simple.

10


Chapter 1

1.4.3

1.4. SMART HOME

Advance metering infrastructure

Automatic meter reading (AMR) enables the utility to read consumers’ units from the meters in an automated manner. The unidirectional communication flow from meters to utility does not provide maximum benefits to both. In order to overcome this deficiency advance metering infrastructure (AMI) has emerged which has the facility of bidirectional communication. The AMI supports SMs which receives DR price signals from the utility. Thus, AMI provides a link between the SMs and utility. It also uses dynamic tariffs to intelligently control the consumer’s load.

1.4.4

Home communication network

The SG network include home area network (HAN) which supports wired and wireless technologies, including ZigBee, WiFi, WiMAX, etc. [41]. The HAN provides a solid platform that enables and establishes a communication link among various home appliances, SMs, and electricity consumers. The proliferation of the Internet of Things (IoT) along with a widespread low-cost wireless technologies: ZigBee, Wi-Fi, etc., have enabled to accelerate the deployment of HAN. Thus, HAN is characterized by low data rate requirements to control the devices installed in the SH.

1.4.5

Home energy management controller

The HEMC uses advanced tools to monitor, manage, and control the generation along with the household loads. It is essential for HEMC to support various technologies comprising of ICT infrastructure, algorithms, software platform from different vendors. The HEMC connects two subdomains: HAN and AMI. It also provides bidirectional communication. The utility sends DR signals to all SMs installed in the residential area through a neighborhood area network (NAN). The HEMC makes energy management decisions based on the intelligent algorithms

11


Chapter 1

1.4. SMART HOME

and provides facilities to the users in terms of reduced electricity cost along with increased UC. Thus, HEMC receives DR signals via some gateway and makes optimal power scheduling decisions [20]. The key enablers that make a home into a SH are represented in Fig. (1.1).

Smart appliances

HEMC

AMI, SMs

Smart home

Communication Infrastructure

Sensors

Figure 1.1: Key components enabling SH

The SHs are not only an important element of SCs but also play a vital role in the transition towards the SGs. A SH can be considered as nano-grid, which is a reduced model of a public grid because it also comprises of energy generation features, loads and energy distributions. As consumers become prosumers in an electricity market, they are able to sell energy back to the grid. The communication infrastructure and AMI provide bi-directional information flow between the SHs and SG applications. The integration of SHs and SG is depicted in Fig. (1.2). Here, SHs are equipped with smart appliances and their own RESs. When total consumer demand is less than the available generation, additional energy is sold back to the main or other nano grid. Thus, SCs countenance multifarious energy challenges, among which the energy efficiency of SHs is a vital requirement. Further, SGs are emerging grids that 12


Chapter 1

1.5. RESEARCH CHALLENGES

Wind Turbines Industrial Plant

Photovoltaic systems

Smart home 1

Smart Grid

Thermal plant

Smart home 2

Cities and societies Hydro power plant

Nuclear power plant

Smart home N

Figure 1.2: Smart home as integral part of SG

utilizes new technologies, including autonomous controllers, intelligent hardware, robust software and other resources for data management, and reliable energy delivery. This dissertation is focused on the efficient energy utilization to fulfill consumer’s electricity demand at reduced discomfort in a SH under the SG domain. The energy in a SH is managed by using HEMC that controls and monitors the installed appliances in an efficient way to provide maximum UC and convenience to its occupants at a reduced cost.

1.5

Research challenges

The scheduling of devices based on DR results in an operational delay of the appliance, that may cause inconvenience to the electricity consumer [21]. Some studies reveal that the consumers wish to reduce their costs, but do not want to compromise on their UC. Thus, a very important issue in SG is the UC which

13


Chapter 1


is often neglected in the cost minimization problems. User-defined priorities assigned to appliances are also vital as it results in increased UC. Balancing of load that results in reduced peak-to-average ratio (PAR) is another challenge that also needs to be tackled by HEMC. Optimum unit sizing of RESs and energy storage system (ESS) that result in reduced cost and carbon emissions for a sustainable environment are some other challenges focused in this dissertation.

1.5.1

Increased user comfort and priority-induced demand response strategy

The UC is a relative term which varies among consumers and situations. In literature, many authors have associated the UC with appliance delay occurred in shiftable appliances, while others relate it to thermal comfort. In some situations, the UC is also taken as a combination of electricity cost and discomfort associated with shiftable or thermal based appliances. A home energy management system (HEMS) is widely applied using various methodologies to determine an optimal load scheduling pattern with a motive to reduce consumers’ electricity cost [42]. Muralitharan et al. opted a multiobjective evolutionary algorithm (EA) to curtail electricity cost and appliance’s waiting time [43]. The authors presented a trade-off between electricity cost and waiting time. In [44], the authors proposed a novel scheme based on the Dijkstra algorithm for the consumer’s cost reduction. The proposed algorithm has a very low computational complexity. The noncooperative game theory approaches are also used to reduce the user’s electricity bills [45, 46]. An EA with a multi-objective function is presented to reduce the consumer’s electricity cost and dissatisfaction [47]. Bharathi et al. [48] demonstrated a genetic algorithm (GA) to minimize energy consumption in three sectors: commercial, residential, and industrial. Here, a comparison of GA with other EAs is also presented. While in [49], the authors used EAs, including GA, binary particle swarm optimization (BPSO), and cuckoo 14


Chapter 1


search to reduce the electricity cost and also maximize UC in terms of reducing the appliance’s waiting time. However, the existing work has only focused on the consumer’s cost minimization problem and did not consider user changing behaviors. Thus, unpredictable consumer behavior of residential consumers is the main factor to be considered. One way is to incorporate the consumer’s defined appliance’s priorities which may account for an increased UC. Static and dynamic are two kinds of priority choices, which can be assigned by the electricity consumers on their preferences to the usage of home appliances. The static priorities are time-independent, where appliances comparison is made with respect to some defined criteria based on the electricity cost, welfare, or an emergent usage. Here, the appliances are assigned some priority weights in a given range of [0 1], on the user’s preferences over a day. The appliances with high priority weights are chosen to operate initially if a DRP confronts a situation to meet some energy threshold limit. For instance, a consumer may assign a high priority value to microwave oven as compared to the WM. The latter denotes the time-dependent priorities of the appliances, where the consumers can modify the priority weights with respect to the time. For instance, a dishwasher (DW) priority value may be set high in the morning, which is latterly decreased by the consumer in the evening, since, another appliance, i.e., the WM is needed to operate at the same time [50]. In [51], the authors assigned dynamic priorities to the shiftable home appliances and proposed an algorithm based on the prediction of RES to decrease the consumer’s cost within comfort constraints in response to a dynamical electricity market price. The work [50, 51] have incorporated the appliances’ priority, however, they did not consider the user’s budget limitation, which is one of the most dominant constraints. In the recent era, the energy management problem considering peak load is solved using the demand side rather than the supply side management. However, the electricity consumers have different interests and willingness when considering peak

15


Chapter 1


load management. In this regard, efficient energy management solutions are required where the priority of an appliance is considered according to the user’s interest. Rastegar et al. addressed the issue by considering a value of lost load to indicate the appliance operational priority based on consumer’s perspective. The ToU and IBR pricing schemes are used. The results show a lower cost for the IBR [52]. In [50], the authors assigned static and infinite priorities to shiftable and non-shiftable appliances. The results reduced the electricity cost with different threshold values. Therefore, it is vital to consider the priorities of appliances defined by the consumers in HEMS. Shifting of the load based on priorities will be beneficial for the consumers during time-slots where the electricity prices are high and also to a utility to control the peak load demand. However, manual response to DR incentives by the electricity consumers is difficult due to the lack of interest, busy schedules or unwillingness to participate. Therefore, in order to take full benefits of such incentives, the need for the priority-induced energy management controller (EMC) capable of making smart decisions without jeopardizing the UC is the need of the hour.

1.5.2

Mitigation of rebound peaks

The PAR is described as a ratio of peak load to average load, which is consumed by the user over a scheduling horizon. The PAR depicts a relationship between the energy consumption behavior of the consumer and the operation of utility peak plants. Thus, it is beneficial for both the utility and the consumer to reduce PAR, so that a balance between power demand and supply can be maintained. When considering cost minimization, rebound peaks which are again generated in the off-peak hours is a vital challenge that needs to be addressed using the appropriate load capacity limit. Thus, efficient methodologies that consider PAR along with cost are essential.

16


Chapter 1

1.5.3


Environmental issues and towards green energy

Traditional energy generation is widely dependent on the use of fossil-fuel resources such as coal, oil, and natural gas. These resources are exhausted and depleted with consumption [53]. Further, the usage of these sources has caused problems like environmental pollution and global warming. The present-day demands new ways of creating energy sources that are more environment-friendly, clean, sustainable, and inexhaustible by nature. The RESs are widely used to generate electricity from solar, wind, geothermal, hydropower, and other sources that are naturally replenished and also have great potential to produce energy [54]. Among the other RESs, wind turbines (WTs) and PV panels are the most dominant and encouraging technologies that are used to meet the consumer load demand [55]. The RESs can be implemented using two ways, i.e., grid-connected (GC) or standalone (SA) modes. In GC mode, the RESs inject the produced electricity to a power utility network while in the SA mode, it directly powers up the consumer’s electrical demands [56]. The SA system causes reliability concerns due to the nonavailability of electricity backup from a utility network. Further, the intermittent nature of solar energy and wind systems cause a non-linear and unpredictable RESs output power.

1.5.3.1

Integration of renewable energy sources and energy storage system in a grid-connected system

In [57] and [58], the authors have used HRES in GC mode for the electrification of the SHs in order to save consumer’s cost and also avoid peak creation. In [49], the authors used PV and batteries to save consumer’s cost during peak hours. The simulation results revealed that the performance of the cuckoo search algorithm provided better results (43.10% and 6.93%) in comparison to the BPSO and GA. Ahmed et al. used PV and ESS to minimize the user’s cost [59]. The hybrid GA-PSO (HGPO) scheme performance was found better than the other proposed 17


Chapter 1


algorithms: BPSO, GA, wind-driven, and binary foraging. The results showed that HGPO reduced cost by 40.05% and the PAR by 41.07% as compared to the non-scheduled load scenario. In [60], the authors used priority-induced DSM strategy to shift the appliance peak load and also reduce consumer’s cost. Multiple knapsacks are utilized to mitigate the rebound peaks. The results depict a tradeoff between priority value and consumer’s cost. Ma et al. in [21] used a multiobjective approach to have the desired trade-off between appliances operational delay and cost. In GC mode, user inconvenience (UI) and discomfort are caused by scheduling of home appliances based on the DR or limiting its time of use or devices operated at reduced energy [21, 60]. This occurs because of the trade-off between the appliance’s delay and consumer’s cost.

1.5.3.2

Integration of renewable energy sources and energy storage system in a stand-alone system

Since, SA system is not connected to the main grid, which causes reliability concerns when using a single RES. Thus, using a single RES in SA environment results in energy variations. This effect causes an energy mismatch situation where the consumer’s load requirements are not met by the generation capacity. In order to overcome the reliability and aforesaid challenge, HRES along with an ESS, including the FC and batteries are used to meet up the user load demand [61]. The complementary features of wind, solar, and other energies are combined in HRES with the backup of ESS which further makes it more sustainable and reliable as compared to single renewable energy system [62]. The major issue in HRES is the optimum sizing of individual components, including PVs, WTs, and battery combination required for strategic decisions, i.e., feasibility study or an initial capital investment cost calculation. A methodology used to determine the right and accurate sizing of HRES components by maintaining the system reliability at minimum system cost is described as unit sizing [63]. Oversizing of system components may overcome the reliability problem, however, 18


Chapter 1

1.6. PROBLEM STATEMENT

it also results in an increased system cost. On the other side, the undersizing of system components can lead to the loss of supply (LOS) problem, where generation is less than the consumer’s load requirement. Therefore, an optimum unit sizing of HRES is essential for the determination of the exact number of system components that leads to system reliability at reduced cost [64].

1.6

Problem statement

In literature, some limitations have been identified. Based on those limitations, we propose a new HEMS for managing the energy consumption in order to curtail the electricity cost, PAR, and user discomfort parameters using meta-heuristic algorithms [65, 66]. In addition to all aforesaid performance parameters, our objective is to consider the consumer’s lifestyle by assigning priorities through which they can schedule appliances as per their requirements. In the previous work, UC is associated with thermal comfort [67] or related to the appliances’ delay, cost savings, and return on investment (ROI) parameters [21, 68, 69]. Ogunjuyigbe et al. have considered the user’s satisfaction from a different perspective which is based on three satisfaction postulates [70]. However, none of the referenced work has considered the UC which may be derived from time and device based (DB) varying priorities. The consumer’s budget limitation is also one of the prominent constraints to the electricity usage, which is mostly neglected in the literature as their primary focus is on cost minimization. Further, rebound peaks generated during off-peak hours are also dangerous and may harm the grid stability if it is not handled carefully [71]. Due to this fact, the utility encourages users to reduce their load during the on-peak time-slots by incentivizing in terms of DRPs. In order to minimize the rebound peaks, it is vital for EMC to consider threshold and load limits in the design. Moreover, actual load profiles are replaced by maximum or the average load of devices which may not give accurate results when compared with the real load profile, which takes different energy cycles of 19


Chapter 1

1.7. RESEARCH CONTRIBUTIONS

an appliance [38, 72]. Thus, the actual load profile of appliances is taken into account. In addition, the use and combustion of fossil fuel cause toxic air emission that results in environmental problems causing great risk to children [73]. Carbon dioxide accounts for an estimated 77% which are caused by the human-generated greenhouse gas emissions [74]. All these factors contribute to toxic air emission in the environment which also have an adverse impact on climate change. The RESs consist of solar and wind are focused in the literature since these reduce carbon dioxide and are ecological and universal [75, 76]. However, these resources have unpredictable and intermittent nature due to natural conditions. Thus, the reliability of RESs is a major concern which needs to be tackled at minimum consumer cost. In order to overcome the environmental challenges, there is a great need to consider RESs in GC and SA modes within the SH.

1.7

Research contributions

The five major contributions of this dissertation are stated as follows: (i) At first, this dissertation proposes an evolutionary accretive comfort algorithm (EACA) for DSM that is being capable of generating an optimal energy consumption pattern, which yields maximum absolute UC. Absolute UC is being derived from the TB and DB priority values assigned to the appliances by a typical consumer. The proposed EACA satisfies the predetermined user budget constraint and also achieves maximum absolute comfort as compared to the randomly generated base cases. A cost per unit comfort index (χ) is also developed, which relates the total consumer expenditure to the total derived UC. The first contribution is published in [77] and its extended version in [78]. (ii) Shifting the residential domestic load from the on-peak hours to the off-peak hours without any load limit results in the creation of rebound peaks that 20


Chapter 1


are generated in the off-peak time-slots. This situation also increases PAR, where users shift the operational load of their SH appliances to the off-peak hours. Due to the aforementioned limitations, we have used knapsack capacity limit to control the rebound peaks. The DA-RTP DSM strategy adopted uses four optimization techniques GA, BPSO, optimal stopping rule (OSR), and enhanced differential evolution (EDE) which successfully shifts various energy cycles of an appliance during system stress time. The proposed EMC gives freedom to the consumers to set low and high appliance priorities which are then considered in the scheduling. The simulations are conducted using two scenarios where individual and aggregate appliances are taken. The aggregate scenario is further categorized using no load limit and the knapsack capacity limit. The results reveal substantial electricity cost savings when no load limit is taken. When the knapsack capacity limit is applied, it reduces the SG peak load demand and also mitigates the rebound peaks. The second contribution contributes to our published work in [60, 79]. (iii) This thesis also presents an optimized HEMS (OHEMS) that not only facilitates the integration of GC RES and ESS into the SH but also reduces the prosumer’s electricity bill along with the PAR. In addition, the performance of the meta-heuristic algorithms: BPSO, GA, bacterial foraging optimization (BFO), wind driven optimization (WDO), and hybrid GA-PSO (HGPO) are also evaluated in terms of electricity bill, energy consumption pattern, and PAR minimization. An energy management system (EMS) is proposed that uses exogenous grid signals; DAP signal, ambient temperature, and solar irradiance. Then, the meta-heuristic algorithms are implemented to get an optimum solution. The simulations are conducted in two stages, in the first case, the benefits of RES and ESS integration are highlighted, while in the second case the performance of proposed algorithms: WDO, GA, BPSO, BFO, and HGPO based HEMSs is compared in terms of PAR, electricity bill reduction, and uniform distribution of energy consumption patterns. 21


Chapter 1


The third contribution is published as [59]. (iv) Pakistan is one of the South Asian countries which is situated at a latitude of 23.45◦ N − 36.75◦ N and longitude of 61◦ E − 75.5◦ E. Pakistan is geographically located in an area where solar irradiation is immense, i.e., 5 − 5.5kW h/m2 /day in Punjab and 7 − 7.5kW h/m2 /day in Baluchistan, respectively. Further, it has great potential of 346GW of wind power production, approximately [80]. Alternative energy development board (AEDB) is established in Pakistan, with an aim to support, facilitate, and encourage the implementation of RESs in the country. With the support of the World Bank, AEDB is carrying out an assessment and mapping activities in major areas of the country. Considering these RESs potentials, in this thesis, a recent proposed algorithm Jaya is implemented to find an optimum unit sizing of HRES using real wind speed and solar irradiance data for Hawksbay, Pakistan. (v) Motivated from non-algorithmic parameter algorithms, this thesis is further centered on finding the optimum sizing of a SA PV-WT-Battery hybrid system considering reliability. Firstly, Jaya and teaching learning-based optimization (TLBO) methods are applied to resolve the optimization problem for a new site located in Rafsanjan, Iran. Next, a hybrid Jaya learningbased optimization (JLBO) algorithm is presented which is based on Jaya and learning phase of TLBO algorithms. Finally, the algorithm’s performance in terms of total annual cost (TAC) values is also compared with GA, which requires algorithmic-specific parameters: crossover and mutation. Finally, the fourth and fifth contributions are also submitted in the prestigious journals. The detailed problem statement, constituting of five subproblem statements along with contributions are given in section 3.6. Fig. 1.3 provides a pictorial view of the relationship among research challenges, problem statements, contributions, methods, results, chapters, and publications of the whole thesis. 22


Chapter 1

1.8. THESIS ORGANIZATION

Co

J nfe ourn ren al ce pap Pr oce er-1 Re edi su ng lts s-1 an d Se disc c. us Me 5.1 sio tho n d Ch Ch s an p5 d p4 s Sec olu tio Su . 4 n .1 an bp ro d Ch contr blem p3 i Sec butio 1 Re .3.6 ns sea c . h 1 Ch all rch p1 eng Sec e . 1. 5.1

-5 -2 gs e r in ap ed 5 e l p oc hp C na Pr n ur e io Jo nc ss u re sc 2 fe d i 5. n on d c. C tio an Se lu ts so 2 . ul nd . 4 es c a e R -2 s ds S ho p4 em on et h bl ti 2 M C ro ibu .6. bp tr .3 c n h Su co Se rc d p3 ea ge .2 an Ch es len 1.5 R al c. ch Se p1

Ch

di

itte

bm

Su

R e hp ch sea 1 alle rc Se n h an Su c. ge d b 1. C co pro 5. hp n b 3. M t l 1 e 3 r et m i b S ho ec ut -3 ds .3 io C R .6 ns hp a es .3 4 nd ul Se s ts o c. lu an 4. ti d 3 on Se dis c. cu s 5. si 3 on Jo C ur hp na 5 lp ap er -3 C

Re

e ng le 2 al 3. . ch .5 5 h .1 & s rc ec -4 ion ea S es 1 em t -5 R hp bl ibu .4 C ro tr .3.6 bp con ec Su nd 3 S d a hp an s d C 4.5 tho ion Me olut .4 & 5 s 4 . hp Sec nC p4 sio Ch cus .5 5 dis & nd 4 . ts a . 5 ls sul Sec rna ou nJ

Main problem statement Chp-1 Sec. 1.6

Figure 1.3: Relationship diagram of research challenges, problem statements, contributions, methods, results, and publications

1.8

Thesis organization

The thesis is organized into six chapters. In Chapter 2, formal and meta-heuristic based optimization techniques are discussed with their applications in the energy management domain. In Chapter 3, related work is discussed which is categorized into four different threads. The first thread discussed various challenges associated with the DSM together with cost minimization, UC, PAR reduction, mitigation of rebound peaks, balancing of the load. The second thread deals with the priority assigned to various appliances. The third thread is focused on RESs and ESS integration in SA and GC modes. In the fourth thread, optimization methods proposed in the literature are focused. Finally, based on the analysis of the recent state of the art five

23


Chapter 1

1.8. THESIS ORGANIZATION

subproblem statements and detailed contributions are presented in this chapter. Chapter 4 discusses system models, problem formulation, and proposed metaheuristic solutions. The five system models and proposed solutions elaborated in this chapter are relevant to the respective four subproblems as given in section 3.6, respectively. Each problem is mathematically formulated and different metaheuristic solutions are being demonstrated to solve the optimization problem based on the objective function. Chapter 5 deals with results and discussion. The five parts of this chapter are mapped to the five subproblems and models discussed in Chapters 2 and 3, respectively. Here, the first part focuses on UC results considering three different budget scenarios. The second part deals with the priority-induced strategy results to shift the load as per consumer willingness. The RESs and ESS are integrated with GC mode and various meta-heuristic along with our proposed hybrid approach results are given in the third part. The fourth and last parts deal with results of unit sizing when RESs and ESS are considered via SA mode. Finally, conclusion along with future work are stated in Chapter 6.

24


Chapter 2 Meta-heuristic and other optimization schemes along with their application

25

2.1. APPLICATIONS OF FORMAL AND OTHER OPTIMIZATION Chapter 2 TECHNIQUES USED FOR ENERGY MANAGEMENT

Chapter summary The energy management problems involve a search for finding the global optimum solution in a space of the system parameters. In general, both formal and artificial intelligence (AI) based meta-heuristic methods are equally recognized and used for the energy management problems. Distributed solutions together with multiagent systems and game theoretic models are also utilized recently in the energy problems. In section 2.1, formal and recursive techniques used in the literature are discussed which are followed by meta-heuristic algorithms in section 2.2. Further, for this thesis, the terms algorithms, schemes, methods, and techniques are used synonymously.

2.1

Applications of formal and other optimization techniques used for energy management

The appliances load scheduling problem is an optimization problem which is solved using various schemes in the literature. Appliance load shifting is done via task or energy schedule. In former scheduling, devices are turned ON and OFF within the allocated time-slots. In latter case of scheduling the power consumption of devices are reduced and their length of operational time (LoT) extended during system stress time [81]. Here authors have used the recursive approach to find peak demand under compressed, delay and postponement request scenarios and compared it with non-scheduled default scenario using RTP scheme for an infinite number of appliances in a residential HEMS. User participation in the energy management program was also considered along with RES integration. The simulation results show satisfactory accuracy while analytical models calculate peak demand results in very small computational time. The limitation of the work includes power consumption overestimation due to an infinite number of appliances assumption. To overcome the overestimation the authors in [82] proposed four scenarios for a finite 26


2.1. APPLICATIONS OF FORMAL AND OTHER OPTIMIZATION Chapter 2 TECHNIQUES USED FOR ENERGY MANAGEMENT number of appliances. In the compress delay scenario, the operational time of the appliances is expanded by decreasing their power consumption using a recursive approach for peak demand calculation. Researchers have proposed recursive formulas for the calculation of peak demand under different power demand scheduling scenarios. The customer participation in HEM is also considered in order to find social welfare. The analytical results produce low timely results which are essential for near real-time decisions. In [83], the scheduling problem is considered as an optimal stopping problem and solved via OSR. In [84], the authors proposed an electricity load scheduling scheme to satisfy the user demands within budget limits through manipulating energy consumption of appliances by using ToU pricing strategy. The mixed integer nonlinear programming (MINLP) problem was solved by the use of generalized Bender’s decomposition approach. The results show optimal electricity load scheduling and electricity consumption of many appliances in a residence was managed in a unified way. In some studies, the appliance scheduling problem is solved using LP. Shirazi et al. proposed a HEM system, which is integrated with DERs along with electrical and thermal appliance scheduling (HEMDAS) [85]. The home appliances are categorized as controllable electrical, thermal, and uncontrollable appliances. Energy minimization problem is formulated via MINLP. The multi-objective function used considers the cost and UC parameters. The results show that HEMDAS architecture has reduced the cost and also achieved a feasible solution to the energy management problem. To achieve energy management objectives numerous algorithms for an efficient HEMS have been proposed, such as integer linear programming (ILP) [86], mixed ILP (MILP) [87], multi-parametric programming [88], etc. Similarly, these techniques are used for unit sizing of HRES. In [89], the authors proposed an iterative based method, in which the optimum configuration is achieved by searching all possible outcomes iteratively based on the objective function. MILP is used in [90], to find out the optimum combinations of PVs and its components. 27


2.1. APPLICATIONS OF FORMAL AND OTHER OPTIMIZATION Chapter 2 TECHNIQUES USED FOR ENERGY MANAGEMENT Formal techniques and other optimization schemes are used as optimization schemes in the SG domain. Wang et al. [91] proposed the BPSO with a combination of ILP to tackle the household load scheduling problem. In [10, 92, 93, 94], the load scheduling problem was formulated via MILP. In [93], the authors considered multi-objective optimization in 30 homes using three different pricing schemes to reduce the consumers’ cost and CO2 emissions. The framework proposed in [94] has considered the integration of electric vehicle and ESS that had significantly reduced the cost by 65%. In [10], the authors devised a MILP model in combination with the heuristic algorithm considering load scheduling for a household consumer. In [95], the authors formulated the problem using MINLP and used a multi-objective model to maintain a balance between user’s cost and comfort within a SH. Moreover, the above mathematical techniques discussed are highly computational and their complexity exponentially rises with problem size growth. Lee et al. [96], present linear programming (LP) based EMS for reduction of PAR and electricity cost by charging the ESS from the utility in the off-peak time-slots and is discharged in peak hours. The integration of RESs into the residential sector is not considered and the charging of ESS from the utility is not an economical solution. In [72], the authors propose an ILP based HEMS with integrated RES to shift a domestic load of shiftable appliances from peak time-slots to off-peak time-slots. However, the integration of ESS into the system model along with UC parameter is not considered. In [97], the authors discussed home appliances scheduling for a designed objective of electricity cost minimization or optimization of electricity consumption pattern. The MILP is used to schedule the household appliances. An in-house RES not only reduces the electricity bill but also sells the surplus energy back to the utility for generating revenue. The growing interests in SG raises the need for new schemes to achieve the best saving in SHs. Gentile et al. [98] proposed a model-based approach for energy billing profiles via the formalism of fluid stochastic Petri nets. The model applied was capable of capturing both discrete and continuous system aspects. The 28


2.1. APPLICATIONS OF FORMAL AND OTHER OPTIMIZATION Chapter 2 TECHNIQUES USED FOR ENERGY MANAGEMENT compositionality feature allowed a modeler to easily customize configuration and system inner behavior. The results show savings in three different policies in domestic SG. In [99], smart appliances and their integration with a control algorithm were also modeled with Petri nets. The results show cost reduction of about 10% to 24% along with peak reduction of 38% to 53% as compared to conventional appliance usage. Various studies in the literature have mapped the energy and task scheduling problem of home appliances to single and multiple knapsack problems [100, 101]. The single knapsack problem (SKP) is associated with a knapsack having an upper capacity limit which has to be filled up with items having value and weight such that the total weight of inserted items does not exceed the knapsack capacity limit while the objective is to maximize the profit [102]. Multiple knapsack problem (MKP) is considered as a combinatorial optimization problem that consists of multiple knapsacks and items. Here, the goal is to find a subset of items that result in the maximum profit without exceeding the capacity limit [103]. The summary of formal and other approaches used in energy management domain is shown in Table 2.1.

2.1.1

Limitations of formal approaches

These classical and formal approaches [84, 85] are good in finding the optimal solution in a deterministic environment, yet, are highly computational in nature since every solution is computed and checked in the given search space. Further, these schemes cannot tackle a large number of appliances having non-linear, unpredictable energy usage patterns, and their complexity also rises exponentially with the increase in the problem size. Furthermore, these deterministic approaches are formulated with known parameters. Dynamic programming (DP) finds an optimal solution to a stochastic environment, however, has its own limitation and suffers from the curse of dimensionality problem [104]. For a stochastic environment, 29


2.1. APPLICATIONS OF FORMAL AND OTHER OPTIMIZATION Chapter 2 TECHNIQUES USED FOR ENERGY MANAGEMENT Table 2.1: Formal and other approaches used in energy management Techniques used

Objectives

Timebased DR schemes

Demand/user classification

Features

Limitations

OSR, distributed and centralized scheduling algorithms [38]

Cost and average delay minimization

RTP

Considers peak demand cycles of appliances

Real-time scenario implemented, considers appliance peak demand duty cycles

PAR and a large number of appliances are ignored

ILP [72]

Appliances scheduling and integration of RES

ToU

Shiftable, interruptible and weather based load

Recursive formulas [81]

Calculation of peak demand under four different scenarios

RTP

Compress, postponement and no-participation appliances categorization

Recursion [82]

Peak demand calculation

RTP

Compress, postponement and no-participation

FCFS, MFCFS, OSR, and PEEDF [83]

RESs integration to minimize user cost and energy consumption

RTP

Active, passive, and semi-autonomous users classification

MINLP [84]

Minimize energy consumption and maximize utility function

ToU

Elastic and inelastic appliances

Multi-residence and multi-class appliances are considered, robust algorithm for dealing with large number of homes

Considers appliance scheduling in multiple residences and ignores energy cycles of an appliance at micro-level

MINLP [85]

Minimize cost and maximize UC

RTP and natural gas fixed price

Controllable electrical, thermal and uncontrollable appliances

Integrated RESs, ESS, and combined heat and power unit to reduced cost

Increased computational time

MILP [97]

HEMS, and integration, as well as grid interconnection of RES

CPP

Daily fluctuations in load for the week, is considered.

Reduction of cost and PAR via RES utilization

Infeasible for small-scale residential consumers

Electricity cost and peak loads are reduced via RES integration and optimizing energy consumption pattern User participation and RESs integration, fast and accurate calculation of peak user demand via analytical model RESs integration, social welfare calculation via user participation, fast and accurate calculation of peak user demand via analytical model Real-time scenario implemented, integration of RESs and storage system is considered

UC and ESS are not considered Overestimation of power consumption due to the infinite number of appliances assumption

Considers only finite set of appliances

Individual appliance scheduling along with monthly and yearly cost savings are ignored

where consumers’ behavior and intermittent nature of RESs add unknown variables to the environment, then these classical approaches become inefficient. In the SG domain, scheduling of home appliances is dependent on the consumers’ desires, which adds a random factor in the environment [105]. Therefore, the deterministic approaches used are not suitable for the stochastic environment. However, computational intelligence-based techniques like EDE, GA, BPSO, etc. can possess some inherent randomness and provide an optimum solution to the above problems at a low computational time. To overcome the limitation of the formal approaches, some studies have proposed the meta-heuristic schemes. Meta-heuristic techniques based HEMCs are widely used in the literature to reduce the peak load demand along with electricity cost.

30


Chapter 2

2.2

2.2. META-HEURISTIC APPROACHES

Meta-heuristic approaches

In this thesis, various meta-heuristic methods are used to solve energy management problems due to following motivations: high flexibility of these techniques during reusability of the software and high heuristic performance, that leads to efficiently handle big-size instances of given problems. For instance, considering reusability factor, we have used and mapped the GA to achieve maximum UC in section 4.1.3, priority-induced strategy to mitigate rebound peaks in section 4.2.6, GC appliances scheduling in section 4.3.8 and unit sizing of RESs and ESS components in SA mode in section 4.4.7. Similarly, we have tested the algorithms considering big instances of problems, i.e., considering aggregate load and high consumers’ load. These meta-heuristic approaches are broadly categorized into two categories: meta-heuristic approaches with algorithmic-specific parameters and meta-heuristic approaches without algorithmic specific parameters. Both meta-heuristic schemes are focused in this thesis.

2.2.1

Meta-heuristic approaches with algorithmic-specific parameters

The population-based meta-heuristic algorithms are broadly classified into two classes: evolutionary class and swarm intelligence based class. The popular EAs are GA, EDE, BFO, etc. The EAs mimic natural evolution processes based on the concept of best survival or fittest individual. On the other side, some popular swarm intelligence based schemes are ACO, particle swarm optimization (PSO), etc. However, there exist another class of algorithms that work on using various natural phenomena, i.e., HS, WDO, gravitational search algorithm (GSA), etc. All these algorithms require common controlling parameters along with their own algorithmic-specific parameters. Next, the algorithms used in this thesis are briefly

31


Chapter 2


discussed, followed by their applications in the energy management problems under the SG domain.

2.2.1.1

Genetic algorithm

The GA is a population-based EA. The population is composed of individuals, also referred as chromosomes. A chromosome states the characteristics and properties of the individual in the given population. In the GA process, the gene is the term used to refer to the individual characterization. The corresponding gene value is referred to as an allele. During each generation of the GA process, individuals with best survival capabilities compete to reproduce offsprings. These offsprings are produced via crossover or mutation steps. In the crossover, offsprings are produced by combining parts of two parents while in mutation, some of the alleles of the chromosomes are altered. The individual’s survival strength is measured via a fitness function composed of various objectives and constraints of the optimization problem. During each iteration, the elitism step is also applied that refers to the individual’s tendency to survive for the next generation so that weak chromosomes with lowest fitness values die. The steps of GA are the initialization, selection, reproduction, and termination [106]. In the initialization step, a population composed of many individual solutions is randomly generated in a given search space of the optimization problem. The size of the population depends on the problem’s nature. While in the selection step, a portion of individuals based on the objective function is selected to breed and produce a new generation. Some popular selection methods are tournament and roulette wheel selections. Reproduction refers to the generation of the second population based on crossover and mutation. All steps are repeated until a termination criterion is met. The termination criterion is dependent upon the maximum iteration size or a solution that satisfies the minimum criteria, etc.

32


Chapter 2 2.2.1.2

2.2. META-HEURISTIC APPROACHES Enhanced differential evolution algorithm

Differential evolution (DE) technique is proposed by Storn et al. in 1995 [107]. This scheme is enhanced by Arafa et al. in 2014 [108]. The EDE scheme is proposed to enhance the convergence speed and accuracy of the DE algorithm. In order to simplify the tuning process, control parameters in EDE were reduced to two as compared with DE which has three parameters. The algorithmic-specific parameters of DE are population size P OP , mutation factor M F , and crossover rate CR. The P OP parameter affects the ability to search the space. M F controls the convergence speed while CR is relevant to the number of changes in population. The limitation of DE is low accuracy and slow convergence rate, which is improved in EDE. In EDE the control parameters are reduced to two (P OP and M F ). In EDE steps, a random population is generated first. The crossover step in DE is replaced by the generation of five trial vectors in EDE. The initial three vectors are formed by applying three crossover rates values, i.e., 0.3, 0.6, and 0.9. The fourth vector is used to fasten the convergence speed and last trial vector aims to increase the population diversity. A fitness function is used to evaluate all, these trial vectors. The trial vector having the highest fitness value is considered as target vector, that survives for the next generation. Individuals having the least fitness function values are removed from the population to have an efficient performing algorithm [108].

2.2.1.3

Binary foraging optimization algorithm

The BFO is a new addition to the family of nature-inspired optimization algorithms. The BFO algorithm is inspired by the social foraging behavior of Escherichia coli. In BFO algorithm, the bacteria swim in search of nutrients (solutions) and select best nutrients to maximize its energy. The BFO algorithm consists of four steps; Chemotaxis, Swimming, Reproduction, and Eliminationdispersal. The detail elaboration of steps is given in section 4.3.8.4. 33


Chapter 2 2.2.1.4

2.2. META-HEURISTIC APPROACHES Binary particle swarm optimization algorithm

The PSO was first proposed in 1995 [109]. PSO is used to find optimal solutions to problems which are continuous by nature. However, this algorithm cannot be directly applied to discrete problems. The authors extended PSO to BPSO in 1997 to solve discrete nature of problems. The BPSO is based on the concept of swarm intelligence. In swarm intelligence, the emergent behavior occurs when agents interact locally with their environment and results in coherent global patterns. It is a nature-inspired swarm intelligence optimization technique which mimics the social behavior of a flock of birds, fishes, bees or ants. The individuals start the search for food in random directions and reach a food source by sharing information. This phenomenon is also called the social concept of PSO [104]. The individuals are represented by the particles that make a swarm; moves around a search space for finding the optimal solution.

2.2.1.5

Wind driven optimization algorithm

The WDO is another nature-inspired optimization scheme based on the atmospheric motion of air parcels. In this algorithm, infinitely small air parcels move in N-dimensional search space. The main difference between the WDO and other meta-heuristic techniques is the use of four different forces to control the motion of air parcels in the atmosphere. These forces include pressure gradient force, gravitational force, friction force, and Coriolis force. The force that moves the air parcels in the forward direction is a pressure gradient, while the frictional force resists their motion in the forward direction. Furthermore, the gravitational force is a vertical force in three-dimensional search space that attracts the air parcels towards the origin and the deflection of air parcels in the atmosphere is due to the Coriolis force.

34


Chapter 2 2.2.1.6

2.2. META-HEURISTIC APPROACHES Backtracking search algorithm

Civicioglu proposed an evolutionary backtracking search algorithm (BSA) for solving numerical problems [110]. Like GA, BSA uses a new crossover and mutation operators for generation of the trial population. It is a memory-driven algorithm which stores population from the previous generation to be utilized in next generation based on a search-direction matrix. Thus, it gains knowledge from the previous generation experience and utilizes it in generating new trials. Due to its good exploration and exploitation abilities, BSA is widely adopted and used by researchers in power domain problems [111, 112, 113]. The five steps of the BSA are initialization, selection-1, mutation, crossover, and selection-II.

2.2.2

Applications of meta-heuristic approaches with algorithmicspecific parameters

The multiobjective function was formulated for DSM to achieve multiple goals: cost minimization and UC maximization. In-home appliance scheduling, a tradeoff exists between UC and electricity cost. Ogunjuyigbe et al. in [70], proposed load satisfaction algorithm based on GA to maximize UC at minimum cost. The simulation results depict that the proposed GA based scheme has achieved minimum cost and maximum user satisfaction on three different budget scenarios. Sensitivity analysis was also carried out on the user budgets and it was found that user’s satisfaction increases with the increase in the user’s budget. Muralitharan et al. used a multiobjective EA to reduce the consumers’ cost and waiting time in SG [43]. The authors have applied a threshold policy in order to avoid the peak and balance the load. The penalty in the form of additional charges has been incorporated in their proposed model if consumers exceed the price threshold limit. The simulation results, minimize both the waiting time and the electricity cost. In [114], the author’s used WDO algorithm and min-max regret-based knapsack algorithm to reduce the electricity cost, peak load, and appliances’ waiting time by 35


Chapter 2


using ToU pricing signal. The simulation results were compared with PSO. The WDO algorithm was found efficient. Rahim et al. designed HEMC based on three heuristic algorithms: GA, ant colony optimization (ACO), and BPSO to reduce the consumer cost, PAR, and user discomfort [66]. The problem is formulated as multiple knapsacks and a combined model of IBR and ToU tariff is implemented. The ESS, along with RESs are also considered in the model. The simulation results show that GA-based HEMC performed well as compared to ACO and BPSO. In addition, some studies hybridized meta-heuristic schemes for better exploratory and exploitative search results. Manzoor et al. [115] combined GA with TLBO and proposed a teacher learning genetic optimization (TLGO) algorithm, which has comparable results with LP at the reduced complexity and computational efforts. Six appliances were categorized into time-flexible and power-flexible categories. Multi-objective function considering electricity cost and user discomfort is used to solve the load scheduling problem. Javaid et al. [116] combined the EDE with TLBO and proposed enhanced differential teaching-learning algorithm (EDTLA) to minimize the electricity cost while also maintaining the desired UC level. In [117], the authors merged GA with BPSO and proposed genetic BPSO (GBPSO) to minimize the consumer bill considering single and multiple home scenarios using the RTP signals. The simulation results show that GBPSO based HEM controller performed better and reduced 36% cost and 34% PAR. In [59], authors hybridize GA and PSO and proposed HGPO to reduce consumer electricity cost. The integration of RES and ESS were considered in the model. The results show that the HGPO outperforms and significantly reduces the bill by 25.12% and PAR by 24.88%, respectively. The DR helps consumers in making decisions for better and improved consumption patterns. Different optimization techniques have been opted in various studies to achieve the cost minimization objective for a large number of consumers. Derakhshan et al. scheduled appliances of four residential houses in Tehran city of

36


Chapter 2


Iran using different types of DR price schemes [118]. They proposed two algorithms, called TLBO and shuffled frog leaping (SFL), for scheduling appliances in the SG. The objective was to reduce the consumer cost. The results showed that DRPs successfully shift load to minimize the electricity cost. Mhanna et al. proposed large-scale households DR distributed algorithm for aggregating a huge number of appliances based on a non-convex DR decomposition [119]. The method was applied and tested on 2560 households, with average 10 devices resulted in a near-optimal solution. Similarly, Logenthiran et al. used the DSM controller to schedule load for 2604 devices present in residence, 808 devices in commercial, and 109 devices in industrial areas, respectively, using DAP strategy [65]. An EA was used for load shifting to reduce the PAR and cost performance parameters. The results brought the load curve close to the objective load curve. The limitation consists of applying the proposed technique in real-time scenarios to deal with the stochastic environment. In [120], a comparative analysis of meta-heuristic algorithms based on HEMC using the schemes: EDE, harmony search (HS) algorithm (HSA), and HS differential evolution (HSDE) were evaluated in terms of PAR and electricity cost minimization. The RES is also integrated to achieve higher savings. The hybrid HSDE based controller showed a balanced load pattern as compared to the other schemes. Similarly, GA, cuckoo search optimization algorithm (CSOA), and crow search algorithm (CSA) were used to evaluate the load, cost, PAR, and appliances’ waiting time parameters in thirty homes using CPP and RTP schemes [121]. The CSOA performed better and reduced the cost well as compared to CSA and GA. Similarly, a hybrid approach by combining GA with BPSO named (GAPSO) was proposed for residential load management in [122]. The GAPSO performed better and showed comparable performance when it was compared to the DP. In [123], the authors discuss the problem of peak demand during certain hours. The concept of clustering and smart charging has been introduced to maximize the benefits in term of cost reduction and UC. A GA-based EMS is designed which 37


Chapter 2


efficiently utilizes energy within the clusters by scheduling appliances. Results demonstrate that by appropriate scheduling of appliances in RTP environment along with ESS, the customer can get maximum saving over the electricity bill. In [124], the authors present the comparison of PSO and GA in terms of computational cost and computational efforts. Results show that PSO needs less computational effort and computational cost to reach an optimal solution as compared to GA. In [125], the authors presented an efficient HEMS architecture for the residential sector. The simulation results show that the hybridization of RTP and ToUP schemes is effective for reduction of bill and PAR. Moreover, an acceptable trade-off between the UC and cost reduction is also achieved. P. Chavali et al. [126], present a distributed mechanism for EMS and grid optimization using a greedy iterative algorithm. In the proposed technique, the electricity tariff is utilized as an invisible hand to optimize the energy consumption and appliances scheduling. The proposed scheme did not consider the UC in the problem formulation. In [127], the authors proposed an improved-PSO (IPSO) that brings the consumer domestic load curve near to the objective curve. Here the electricity price and objective curve have an inverse relationship. One of the objective functions is power system stability and the proposed scheme compromise on the UC by rejecting the load in peak hours. Meta-heuristic approaches are widely used in the literature for unit sizing [128, 129, 130, 131]. In [128], the authors used the firefly algorithm to determine the optimal and right-sizing of the SA PV system and its components. In [129], HS optimization technique is proposed for an off-grid hybrid solution consisting of PVs and biomass power generators. Agricultural wells located in Bardsir, Iran are targeted with an objective function that reveals minimization of the system total net percent cost (TNPC) while also considering the reliability factor. The comparison results with PSO and GA optimization schemes depict that HS performed better in terms of reducing TNPC. In [130], Maleki et al. investigated optimal unit sizing of HRES, including PV, WT, and batteries. The authors analyzed and compared 38


Chapter 2


EAs consisting of simulated annealing, PSO, tabu search (TS) along with artificial bee swarm optimization (ABSO). The results show that ABSO performed better among other meta-heuristic algorithms with reduced cost for unit sizing of HRES. In [131], the authors used an improved ACO scheme for the unit sizing of HRES consisting of PV, WT, batteries, and fuel cell (FC). In [83], Rasheed et al. presented the OSR scheme to reduce the electricity cost and maximize UC in a real-time pricing environment. The consumers are categorized as active, passive and semi-active users depending upon the available resources of renewable energy and storage system. The home architecture proposed utilizes three schemes: first come first serve (FCFS), modified FCFS (MFCFS), and priority enable early deadline first (PEEDF). The simulation results show that cost is minimized by the FCFS algorithm. However, the appliances’ LoT does not meet in the FCFS as compared to the other two algorithms. Further, the PEEDF algorithm has reduced the cost and also achieved the maximum UC as compared to MFCFS scheme. The meta-heuristic algorithms used in the literature are summarized in Table. 2.2.

2.2.3

Meta-heuristic approaches without algorithmic-specific parameters

The proposed meta-heuristic algorithms in section 2.2.1 require algorithm-specific parameters for its functioning. For instance, the HS scheme uses harmony memory, pitch adjusting, consideration rate along with a number of improvisations; GA requires crossover and mutation probabilities with a selection operator; PSO needs cognitive and social parameters in addition to inertia weight; ABSO uses a number of scouts, employed, and onlooker bees with a limit specifier. The ACO and other algorithms also require performance tuning of these algorithmic-specific parameters, otherwise, may halt in local optimum solutions or yield in an increased computational time. Therefore, meta-heuristic algorithms that do not depend on 39


Chapter 2

2.2. META-HEURISTIC APPROACHES Table 2.2: Meta-heuristic algorithms used in literature

Techniques used

Objectives

Timebased DR schemes

Demand/user classification

Features

Limitations

Multiobjective EA [43]

Minimize cost and delay

ToU

Permanent and schedulable devices

Threshold policy along with penalty has been considered

Dominant or peak energy cycles of an appliance is not taken

Load satisfaction algorithm [70]

Maximize UC and minimize electricity cost

ToU

Categorization based on different sections of the user house

Multi-objective function, three different budget scenarios implemented

Monthly cost savings and PAR are not considered

EA [65]

Minimize cost and peak load

DAP

Controllable devices only

Residential, commercial and industrial areas are considered with large number of appliances

Delay and UC are ignored

ACO, BPSO and GA [66]

Minimize cost, PAR, energy consumption and maximize UC

ToU and IBR

Passive, semi-active and active users, fixed, shiftable and elastic appliances

Integration of RESs and storage, cost reduction, minimizing PAR and user satisfaction is considered

Dominant duty cycles of an appliance is ignored

WDO and min-max regret-based knapsack algorithm [114]

Minimize energy consumption, cost, peak and waiting time

ToU

User dependent, interactive schedulable and unschedulable

WDO performed well when compared with PSO

RES is not exploited. Dominant duty cycles of an appliance is ignored

TLBO and SFL [118]

Reduce cost

ToU, RTP, and CPP

TLBO provides more optimized results than SFL

Delay, UC and PAR are ignored

Distributed algorithm [119]

Robust algorithm

-

Consider large number of households, i.e., 2560, fast convergence and scalibility

Individual appliance cost saving is ignored

GA [123]

REMS

DA-RTP

GA [125]

HEMS

RTP and ToU

-

Greedy iterative algorithm [126]

EMS, and optimization of grid operation

-

Optimization of consumers energy consumption pattern for grid station stability

RTP signal is used an invisible hand to optimized the energy consumption pattern

UC is compromised

IPSO [127]

Grid station stability

ToU

Schedulable and non-schedulable loads

The desired objective is gained by rejecting additional load requests in on-peak time-slots

Only passive appliances are considered, and UC is compromised.

Shiftable, sheddable and non-sheddable loads Must-run, inflexible, flexible and non-interruptible devices N number of appliances considered

New peaks formation is avoided by dividing the appliances into clusters Hybrid of RTP and ToU is used to avoid the peaks formation

UC is compromised, and RES is not considered UC is not considered

algorithm-specific parameters for its functioning has achieved a wide acceptance among the researchers [132]. Next, discussed are some algorithms used in this thesis that do not require any algorithmic-specific parameters.

2.2.3.1

Jaya algorithm

In Jaya, only common control parameters: population size, termination criteria, etc. are required. In Jaya, the objective function f (r) is to be minimized at iterations t, having “p” number of decision variables (l = 1, 2, . . ., p), and “q” number of candidate solutions for a population size, (m = 1, 2, 3, . . ., q). The best candidate achieves the best value of f (r) in the entire candidate solutions and is represented by f (r)best . Similarly, the worst value of f (r) denoted as f (r)worst is assigned to the worst candidate in the entire population. If Rl,m,t represents 40


Chapter 2


the value of lth variable for the mth candidate during the tth iteration, then it is changed as per criteria defined by the following formula [133]: 0

Rl,m,t = Rl,m,t + rand1,l,t (Rl,best,t − |Rl,m,t |)

(2.1)

− rand2,l,t (Rl,worst,t − |Rl,m,t |), where, Rl,best,t and Rl,worst,t are the values of variable l for the best and the worst 0

candidates at tth iteration, respectively. The Rl,m,t depicts the updated value of Rl,m,t while rand1,l,t and rand2,l,t denote the two random numbers for the lth variable during the tth iteration in the range [0, 1]. The expression “rand1,l,t (Rl,best,t − |Rl,m,t |)” shows the tendency of the solution to move towards the best solution and the expression “rand2,l,t (Rl,worst,t − |Rl,m,t |)” indicates the tendency to avoid the 0

worst solution. The Rl,m,t is only accepted if it achieves better function value. At the end of each iteration, all the accepted solutions are updated and maintained which acts as input for the next generation. To avoid negative and decimal values, absolute and floor functions of MATLAB are used, respectively, to have an integer value for the decision variables. The general flowchart of Jaya algorithm is depicted in Fig. 2.1 [133]. The advantage of Jaya lies in its simplicity because it does not need any algorithmspecific control parameters for its functioning. Jaya has attracted a wider research community and it is being applied to various optimization problems. For instance, Jaya is used to solve traffic light scheduling problem in urban areas [41], facial emotion recognition [134], PV model parameter identification [135], and PV maximum power point tracking (MPPT) problem [136]. The TLBO [137] and improved TLBO (ITLBO) [138] are techniques that also do not require any algorithmic-specific parameters.

41


Chapter 2

2.2. META-HEURISTIC APPROACHES Start

1) Initialize population, i.e., POP 2) Set a termination criteria

Select the Best and Worst vectors (solutions) based on the objective function and constraints in population POP

Modify vectors based on the Best and Worst vectors using Jaya Eq. (24)

Is modified vectors have better fitness values to the corresponding POP vectors? Yes

No

Accept modified vectors and replace them with corresponding POP vectors

Keep the previous POP vectors

No

No

Is termination criteria satisfied? Yes Report optimum solution

Stop

Figure 2.1: Flowchart of Jaya algorithm

2.2.3.2

Teaching learning-based optimization

The TLBO is a population-based algorithm inspired by the process of teaching and learning in a class environment [137]. It starts by generating a random population, which is then updated during each iteration of the optimization process. The rows and columns of the population represent the learners and subjects, respectively. Each subject of the learner is related to the decision variable, whereas the total number of subjects of learner corresponds to a solution. The TLBO process is 42


Chapter 2


divided into two different phases: the teacher phase and the learner phase. The former phase shows learning from the teacher and later phase is associated with learning via interaction among the learners. In teacher phase, the mean of the learners is calculated as subject wise. All the learners are evaluated through fitness function and the best learner having l . The algorithm now tries to minimum TAC will be chosen as a teacher Xteacher

shift the learners mean towards the teacher. Thus, a new vector formed by current and best mean vectors is added to the existing population, as shown in Eq. 2.2.

l l l Xnew(t) = Xold(t) + r × Xteacher − (Tf actor × M l ) ,

(2.2)

here, r represents a random number in range of [0, 1], Tf actor shows the teaching factor (TF). This TF is selected as either 1 or 2. It is necessary to mention that Tf actor is not an input parameter rather it is randomly decided with an equal probability, by the algorithm during the optimization process as given in the formula:

Tf actor = round[1 + r × (2 − 1)].

(2.3)

l in Eq. 2.2 is only accepted if it provides a better fitness function value. The Xnew

In the learner phase, each learner randomly interacts to other learners for the sake of sharing and increasing their knowledge. The process starts by randomly l selecting two learners: Xm and Xnl , from the existing population, such that (m 6=

n). Based on the fitness values of the learners, the population is updated by the formula:

l Xnew(t)

  l l l l  Xl old(t) + r.(Xm(t) − Xn(t) ), if Xm(t) ≤ Xn(t) =  l l l  Xold(t) + r.(Xn(t) − Xm(t) ), otherwise.

(2.4)

l Xnew(t) is only accepted in case if it achieves a better fitness function value. The

optimization process continues until some termination criterion is met. 43


Chapter 3 Literature review: subproblem statements and contributions

44

Chapter 3

3.1. DEMAND SIDE MANAGEMENT CHALLENGES

Chapter summary This chapter describes some recent literature pertinent to energy management in SG. The related work is organized into four threads comprising of different challenges associated with the DSM. In the first thread, challenges like the electricity cost, PAR, and users’ discomfort minimization are discussed in section 3.1. While in the second thread as given in section 3.2, the work related to the appliances’ priority in SHs is focused. The third thread in section 3.3 is focused on RESs and ESS integration via GC and SA modes in the SHs. In the fourth thread 3.4, optimization methods proposed in the literature are focused. Finally, the literature analysis is given in section 3.5 followed by five thesis subproblem statements and detailed contributions in section 3.6.

3.1

Demand side management challenges

In [139], the authors focused on the electricity cost minimization of industrial and commercial buildings under a dynamic pricing tariff. The results show 13% decrease in the electricity bill savings when the schedules of office building appliances are shifted to an hour earlier. Similarly, Logenthiran et al. [65] also considered a large number of appliances in residential, commercial and industrial sectors to achieve the objectives pertaining to the peak load and cost reduction. A meta-heuristic-based EA is developed and results depict cost savings of 5 to 10% with the peak load reduction. In [140], the authors proposed a convergent distributed algorithms to deal with the uncertainties in RESs and ESS installed in the user’s premises. The simulation results showed 7% of monetary savings. The authors in [141] proposed a DSM scheduling scheme considering the PAR constraint. A multi-objective function proposed was found cost-effective along with consideration of the PAR constraint and consumers’ preferences. In [142], customer participation is considered using UI indices for peak load reduction.

45


Chapter 3


Hassan et al. [143] considered two customer engagement plans, including constant deviation and proportional deviation plans for residential peak load reduction. The UC varies from consumer to consumer and situation to situation. In [144], the authors considered thermal comfort as a metric of UC and reduced the electricity cost along with peak demand of 30 residences. Ma et al. [145] developed a costefficient load scheduling algorithm. Cost efficiency is considered as a ratio of user’s total consumption benefit to its accumulative electricity payments. The results revealed that the algorithm significantly improved the cost efficiency. In [146], the authors proposed a model where coordination among 30 and 90 homes was performed using DERs. The main objective was customers’ requirement satisfaction, cost, and peak load reduction by efficiently managing DERs operation. Authors in [71] focused on the issue pertaining to the peak load demand. A dynamic energy management framework using multi ToU and CPP was proposed to mitigate rebound peaks. In [60], the authors used multiple knapsack limit to reduce the peak load demand and also minimized the creation of rebound peaks in off-peak hours. There is a trade-off between the electricity cost savings and UC in terms of appliance waiting delay. In [147], the authors proposed a game theoretic approach to take this trade-off into account and reduced the cost while preserving users’ preferences. Fakhrazari et al. in [148] described a model that aimed to optimize the user’s cost considering UC preferences. Naeem et al. categorized customers’ behavior towards DR management based on UC [149]. In [21], the authors proposed a discomfort function for two types of appliances: flexible starting time and flexible power devices. The discomfort of the flexible starting time appliances is caused by delaying their operation. The discomfort of power flexible devices comes from a power deviation from normal, which was calculated by a Taguchi loss function. A multi-objective function is considered for both the discomfort and the electricity cost parameters. The UC was taken as appliances’ utility which carries 60% weight and the remaining 40% was assigned to

46


Chapter 3


the electricity cost savings [68]. Ten smart appliances were categorized as activitydependent appliances (ADA), occupancy dependent appliances (ODA), and occupancy independent appliances (OIA). The optimization problem was solved using a realistic scheduling mechanism (RSM) and BPSO using DA-RTP signals. A uniobjective cost function was considered and the scheduling horizon comprising of a day was subdivided into four sub-slots of 6 hours each. The UC is not considered in the objective function and appliances’ priority also ignored. A UC level framework is proposed in [69], considering three parameters, including electricity cost savings, delay, and ROI. Nine appliances are categorized in occupancy dependent (OD) and occupancy independent (OI) classes. Six different scenarios are taken into account in order to measure the performance metrics comprising of the cost, energy consumption, and UC with respect to a baseline model that has no HEMS installed. In [67], the authors considered bidirectional energy trading between a residential building and the SG. The surplus energy is obtained from the RESs. The ESS containing batteries is also incorporated in the model. The authors proposed a comfort demand function based on a declining block rate (DBR) and power supplies based on an IBR approach, with a goal to maintain a balance between the UC and the electricity cost. The advantage of comfort demand function was merging the comfort gains based on a linear relationship. Here the appliance load is categorized into fixed, non-interruptible, interruptible categories. In [70], a user’s satisfaction is made quantifiable, fuzzy, comparative, and relative based on the three postulates. A GA-based load satisfaction algorithm is proposed to evaluate the user’s satisfaction within a defined budget limit. A single home with twelve different appliances varying in quantity is considered. The electricity tariff (ET) is fixed at the rate of 0.115 $/kWh. The results on three different budget scenarios show that the proposed algorithm has achieved the maximum user’s satisfaction as compared to the random base scenarios, which did not utilize the load satisfaction algorithm. Some recent work relating to UC is summarized in Table 3.1. 47


Chapter 3

3.1. DEMAND SIDE MANAGEMENT CHALLENGES Table 3.1: Schemes and user comfort

Technique used

Pricing scheme

Classification of appliances

Number of homes vs. appliances 1-6

Convex optimization [21]; 2016

DAP

Flexible starting time and flexible power

Comfort demand function based on DBR [67]; 2015

DARTP

1-13

Realistic scheduling mechanism & BPSO [68]; 2016

DA RTP

Fixed, noninterruptible, interruptible and comfort appliances load ADA, ODA & OIA

UC level framework for HEMS [69]; 2017

ToU

ADA, ODA & OIA

1-9

Load satisfaction algorithm [70]; 2017

Fixed price

Appliance load based on different home sections

1-12

1-10

User comfort

Objective tion

func-

Limitations

User discomfort is calculated using delay and Taguchi loss function Power and comfort function calculated using IBR and DBR

Multi-objective function to minimize cost and discomfort Multi-objective function considering buying, selling and comfort factors

Appliances’ priority is not considered

UC calculated using the appliances utility which carries 60% & consumer electricity cost savings adds 40% UC calculated using the appliances utility, electricity cost savings and ROI User satisfaction calculated on three postulation

Uni-objective function to minimize cost

UC not taken as an objective parameter, Appliances’ priority is not considered

Uni-objective function to minimize cost


Uni-objective function to maximize absolute satisfaction


Only thermal comfort of AC is considered

In [150], the authors present a review of current trends in HEMS and DR in the residential sector. The importance of HEMS for relocating and curtailment of the load is also discussed. They give insights into existing optimization techniques: mathematical optimization, model predictive control, and heuristics algorithms. The impact of forecasting uncertainty, devices heterogeneity, computational limitation and timing consideration in the design of optimization algorithms are also discussed. However, the UC and appliances waiting time have not been discussed. The authors in [151], present a compact survey of the current trends in HEMS. The challenges in the implementation of HEMS are discussed and give the insights on current literature regarding DR, DSM, appliance scheduling, and on single or multiple objective optimizations in HEMS. However, the integration of RESs and ESSs into the residential sector, their impacts on electricity bill and PAR have not been addressed. In [5], the authors discussed the importance of energy management and planning in SC. The paper presents a review of the planning and optimization of SC energy system. Four areas of SC energy system: generation, storage, transportation, and end user are addressed in detail. An ESS is integrated to support the residential load in grid events. However, the scheduling of appliances or relocating of load 48


Chapter 3


and ESS in response to the dynamic pricing of electricity market has not been discussed. The authors in [152], give insights regarding the ways and manners that facilitate the integration of RESs and DG in the SG and the concept of SCs have been comprehensively addressed. The obstacles to the integration of DG into the existing distribution network have been highlighted. The effect of DG on voltage stability and control at low and medium voltages are also discussed. Moreover, the impact of DG on power quality, power system stability, and related events of voltage sag and swell due to the failure of distributed sources are addressed in a comprehensive way. Although, the authors discussed the impact of DGs and RESs integration on power quality and power system stability. However, the integration of RESs and ESSs into the residential sector and their role in DSM programs have been ignored. In [153], the authors present a detailed review of evaluating trends in SHs and SG. They investigate the effectiveness of various communication technologies: Zig-Bee, Z-Wave, Wi-Fi, and wired protocols. The authors point out the merits and demerits of existing technologies and products available in the market. Moreover, the barriers, challenges, benefits, and future trends regarding the communication technologies and their role in the SG and SHs are also discussed. The authors in [154], demonstrate a residential energy monitoring system (REMS). Three SHs powered by an in-house RES (i.e., PV system) are considered for demonstration. A data logger system is integrated within a SH to measure and record the electricity production and demand patterns. The internal and external temperature, as well as the humidity, is recorded to accurately schedule the operation of heating ventilation and air conditioning (HVAC) system. However, this paper did not address the integration of ESS to utilize the RES more efficiently. Mehmood et al. [155], present an in-depth review of load forecasting (LF), current LF techniques in existing power system, future trends, and its importance for the implementation of future SG is discussed. They proposed two major types of LF: mathematical modeling and AI-based computational models along with their 49


Chapter 3

3.2. PRIORITY MANAGEMENT OF APPLIANCES

subcategories as well. The authors also present a comparative study of various dynamic pricing schemes: RTP, ToU pricing (ToUP), and CPP. In [156], the authors proposed a hybrid technique of artificial neural network (ANN) and lighting search algorithm (LSA) to generate optimal operation patterns of the appliances. Four appliances are modeled and developed in MATLAB/SIMULINK, with pre-defined preferences of the consumer. To optimize the appliances ON/OFF time, and to complete the assigned tasks with minimum cost, a hybridization of LSA and ANN is used. The proposed technique significantly reduces the electricity cost and outperforms the hybrid of PSO based ANN. However, RES and ESS has not been taken into account to more effectively minimize the electricity bill. Further, the reduction of PAR and its impact on power system has not been discussed. Wen et al. [157], demonstrate reinforcement learning (RL) techniques for appliance scheduling problem and energy management problem. An observe, learn and adapt (OLA) strategy is opted which significantly reduced the cost and PAR. However, the UC is compromised.

3.2

Priority management of appliances

In [158], the authors proposed prioritizing the operation of power units from a consumer perspective. The power units comprise of a PV system, battery power, and main grid supply. The solar energy is given the highest priority and chosen as a primary power unit to reduce the consumer cost and also avoid the peak creation. The controller shifts the load to other power generating units in case of high user demand in response to the ToU. However, the home energy management algorithm proposed for the source optimization has not considered the appliance’s priorities that also play a key role in increasing the UC level. A HEMS is used for improving the energy efficiency by minimizing the electricity cost, PAR, and also maintaining the desired UC level. An appliance-based rolling

50


Chapter 3

3.2. PRIORITY MANAGEMENT OF APPLIANCES

wave planning algorithm was developed that interacts with the appliances in a priority order [99]. The appliance’s priority is characterized on the basis of UC, which also depends on the devices working principles. The results show significant cost and peak reduction. However, the appliances’ priority order is determined only by its operational characteristics. A HEMS supported by the PV and battery system has been considered to manage the temperature of thermal appliances in a home [159]. A threshold level is utilized to control the power consumption of electrical appliances. Three scenarios, considering the electricity cost and UC based on the priority are assessed. The proposed control algorithm using an ESS reduced the costs by 20% per day as compared to the electrical appliances operating at maximum UC without using an optimization scheme. In [50], the home appliances are categorized based on interruptible and shiftable features. Static priorities are allocated to the shiftable home appliances based on the consumer’s preferences while an infinite priority has been assigned to the non-shiftable appliances since these do not undergo any scheduling criteria. The results yield reduction in peak demand and total electricity cost at the maximum user’s satisfaction. In [160], the load demands have been classified as flexible and essential. The flexible demands are further categorized as delay-tolerant and delay-sensitive demands. A queuing based model depicting high priority has been considered for delay-sensitive demands. The delay-tolerant demand is also allowed to be shifted to a high-priority queue, based on the user’s preference. In [161], the authors designed a decision support algorithm for a press-shop factory that may choose to accept or reject a DRP. If the offer is accepted only higher priority equipment installed will receive the major energy allocation while the load is curtailed for other low priority types of equipment. However, the DRP is not widely applied to industrial units because it may cause disruption in the production process, which ultimately results in losses.

51


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION

3.3

Renewable energy sources and energy storage system utilization

One of the goals of energy management is to reduce CO2 emissions for providing a clean environment. Morales et al. [162] used ToU pricing signal to shift the load curves using polynomial functions in Galapagos Islands belonging to Ecuador country. Different strategies are applied to minimize the rebound-effect by considering the peak load reduction and shifting energy in a day. The authors proposed the utilization of SG architecture based on real life along with the integration of ESS as a future direction. In the aggregate residential demand, Muratori and Rizzoni found that simple DR pricing schemes might create pronounced rebound peaks. The authors used the DP to find a global solution and proposed an energy management framework based on multi-ToU and multi-CPP to deal with the rebound peaks by synchronizing the individual residential demands along with plug-in hybrid electric vehicles (PHEVs) consideration in a decentralized manner. In multi-ToU, electricity consumers are placed into different groups, where each group sees different ToU prices. Simulation results serve as a tool for energy policy solutions where different electricity price structures may be applied to developing an effective and robust DRP [71]. In [91], Wang et al. considered system sizing of HRES consisting of PV panels, batteries, WTs, and a diesel generator. SA home consisting of single-family is considered where diesel generator is used as a backup resource for solar and wind systems. The authors have used a receding horizon optimization method considering day-ahead DR and weather forecast to minimize the capital and operating costs. Since in the SA system, consumers are also acting as electricity producers, therefore, the pricing issues have been substituted by taking the amount of electricity. The simulation results confirmed that the proposed strategy has achieved a globally optimal solution. In [163], the authors proposed WTs and PVs as a 52


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION primary power for five homes in the Pacific Northwest area. The FC and ESS are used as a backup for the SA homes. The system proposed is called green power generation due to the fact that all sources being used are environment-friendly. The simulation is carried out for two scenarios considering typical summer and winter days. The objective of unit-sizing achieved is to minimize the gap between energy generation from HRES and household demand over a period of time. Kellogg et al. in [164] contemplated the unit sizing of energy generation using WT, PV, and hybrid WT/PV systems. The authors considered cost analysis of SA home situated in Montana. The authors have adopted an iterative optimization procedure for selecting the optimum unit sizing of WT and PV panels. In [165], Ayop et al. considered the optimum size of PV panels and also gave a review of seven different types of sizing methods for a microgrid in the SA environment. The authors have opted a methodology based on an iterative approach to determine the optimal size of PV panels using the loss of power supply probability (LPSP) and life-cycle cost (LCC) concepts. The proposed methodology has minimized the LCC and also brought the LPSP value equal to 1%. A building load profile and a year’s meteorological data are taken for simulations. The two sizing optimization methods used are graphical construction method (GCM) and the iterative method. The iterative method has a simple procedure yet required long simulation time to obtain the optimum sizing for the PV panels. Emerging energy issues have resulted in the concept of zero net energy houses. UI or discomfort is caused by shifting the appliance’s domestic load from peak to off-peak time-slots. Kang et al. in [166] solved UI problems by considering PV and ESS. The ESS is considered as a domestic load when the battery is charging and considered as an energy resource when it is discharging electricity. A home consisting of 28 appliances is considered with time-slot of 15 min. The ESS control algorithm has proposed, which make the home as constant energy that consumed a constant load under all conditions. The simulations have been performed on four different weather condition scenarios consisting of rainy, cloudy, sunny, and 53


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION overcast days. The simulation results validate maximum utilization of the ESS, along with making the SH as a constant load of energy network. Okoye and Solyal in [90] applied MILP technique to find the accurate sizing of PV systems in Bursari, Nigeria. The objective set is to reduce the annual cost of finding the unit size of PV components. The results illustrated that the use of the PV system is 30% cheaper as compared to the usage of diesel generators. In [128], the authors performed optimal sizing of the SA PV system by using firefly algorithm-based sizing algorithm (FASA) as an optimization strategy. The sizing problem considered four decision variables consisting of PV modules, inverters, charge controllers, and batteries. The objective function has targeted to reduce the LPSP. The FASA results have been compared with conventional, iterative, PSO, GA, and evolutionary programming (EP) considering two different cases, i.e., a system with MPPT and standard charge controller. The results showed that the iterative method, taking every solution into account took the highest computational time of 14 and 16 hours with LPSP values 0.0259 and 0.0235 in two scenarios, respectively. Whereas FASA obtained the same output values with computational time of 11.66 sec and 40.51 sec for two scenarios, respectively. It is also 1.93 times faster as compared to the GA, EP, and PSO. In [167], Maleki et al. considered a hybrid combination of PVs, WTs, and FCs. In unit sizing of this optimization problem, the decision variables have integer values which are based on the cost minimization objective function and LPSP value given by the user for the reliability of the system. A yearly data consisting of insolation, wind speed at height 10 m, and load profiles are taken for simulation. A population-based meta-heuristic scheme based on ABSO algorithm has applied to obtain a global optimum solution. Using different LPSP values, i.e., 0%, 0.3%, 1%, and 2% ABSO is applied to have the cost-effective hybrid system combination in three different combinations: PV-WT-FC, PV-FC, and WT-FC systems. The results showed that at lower LPSP values, the most cost-effective solution has been obtained by a combination of PV-WT-FC. In [168], the authors considered 54


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION the economic aspect of using a diesel generator with other RES, including PVs, WTs, and FCs for the electrification of a remote area. The total annual generation cost has been optimized using a discrete simulated annealing algorithm. Allot of power loss is caused by home appliances due to alternating current (AC) and direct current (DC) conversion. Castillo et al. in [169] considered the requirements of an off-grid building by utilizing a DC PV microgrid. The building under study is assumed to have an equivalent DC module installed. The authors considered real consumption data of the building along with local temperature and solar irradiance measures gathered for the last thirty years, located in the Bilbao, Spain. The simulations have been performed under five different scenarios considering winter, summer, manipulated summer, lowest, and typical solar irradiance days showed a higher mean performance of 75% and reduced transmission losses, i.e., up to three times as compared to the AC network. Various unit sizing analytical tools, including hybrid optimization model for multiple energy resources (HOMER) [170], H2RES, etc. are used for energy optimization, sensitivity analysis, and planning. However, both HOMER and H2RES require input parameters to be inserted by the user in term of components size and sensitivity analysis is also required to select the optimal solution [171]. Some researchers have proposed mathematical approaches like [172], while others have focused on artificial intelligence based meta-heuristic techniques [130]. In [80], Ahmad et al. investigated the potential of HRES for inhabitants of Kallar Kahar, located in the Punjab province of Pakistan. The authors performed a technoeconomic analysis of wind, PV, and biomass system using a software called as a HOMER. The authors concluded that Kallar Kahar has significant potential for energy-producing via HRES because of the suitable availability of solar irradiation, wind, and sufficient animal manure. The cost of HRES calculated is 180.2 million USD for a peak load of 73.6 M W . Similarly, in [173], the authors have applied HOMER to find PV-FC-battery system size optimization for an un-electrified village in India. The seasonal load variations in summer, winter, and rainy seasons 55


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION are considered for the simulation to reduce TNPC. The cost of energy obtained from the optimized solution is estimated at 0.1959 $/kW. Recently, the efficient utilization of generator in HRES has attracted significant attention [174]-[177]. The diesel generator installed in the HRES is not efficiently utilized if it is not running at more than 40% of its rated capacity. Further, the frequent starts and stops may also lead to fuel wastage and mechanical wear-and-tear of the generator. In [174], the authors used a hybrid WT-Diesel-Battery system to fulfill consumer’s load in an off-grid mode. The day-ahead wind energy and load are forecasted in order to optimize the usage of diesel generator via a control algorithm based on GA. The results depicted that the diesel generator operated with fixed output power is believed to operate in a more stable manner. In [175], Ahadi et al. suggested a hybrid PV-WT-Diesel-Battery system for diesel-free communities located remotely in South Korea. The proposed objective function minimized the total cost while also considered the reliability constraint. The results show that the WT played an important part in HRES because when its usage fraction is increased from 8% to 33% has reduced the size of PV panels and batteries, thus ultimately resulted in a reduced total cost. It is further contemplated that increasing WT fraction above 33% resulted in a non-cost-effective solution. Ogunjuyigbe et al. [176] considered a PV-WT-Split-diesel-Battery HRES to fulfill a typical building load demand in a grid-independent scenario. A try-objective GA based model is considered to minimize the LCC, CO2 emissions, and dump energy. The simulation results showed that the PV-WT-Split-diesel-Battery HRES provided the most optimal results with LCC of 11,273$, which equals to 46% of reduced cost as compared to the use of a single big diesel generator. Further, the CO2 emission and dumb energy are reduced to 82% and 94%, respectively. Thus, the use of the n-split-diesel generators resulted in a better LCC performance as compared to the single biggest generator case.

56


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION The authors in [177] considered PV-WT-Diesel-Battery system in the SA environment of the South China Sea, Malaysia (SCSM). The SCSM is a tourist attraction place, completely dependent on the usage of diesel generators. The use of fuels in the diesel generators has resulted in the environmental problem. The HOMER software tool is used for technical and economic HRES analysis to fulfill 13,048 kW of average load per day requirements with an estimated peak of 1185 kW. The PV-WT-Diesel-Battery system provided optimal results of 17.15 million dollars and 2,571,131 kg per year in terms of net present cost (NPC) and CO2 emissions, respectively. In contrast, the electricity produced by a single diesel generator resulted in NPC of 21.09 million dollars and 5,432,244 kg per year CO2 emissions. Similarly, in [170], Mamaghani et al. suggested PV-WT-Diesel generator HRES for electrifying three remotely off-grid villages in Colombia. The techno-economic feasibility analysis of cost and environmental evaluation in terms of CO2 are conducted for the proposed hybrid system via HOMER software. The simulation results showed a reduction of 2.6%, 3.6%, and 5.4% CO2 per year for Puerto Estrella, Unguia, and Jerico villages, respectively as compared to the single diesel-based system. In terms of cost, PV-WT-Diesel HRES provided the optimal solution for Puerto Estrella while the PV-Diesel system is found efficient in the other two villages. In [178], Karmaker et al. conducted an environmental and economic feasibility assessment of using the PV-Biomass-Battery system to supply power to electric vehicles in Bangladesh. The results obtained by the HOMER pro software showed 34.68% deduction in CO2 emissions in contrast to grid-based EV charging along with 12$ to 18$ savings per month. A study showing techno-economic feasibility is conducted to assess the utilization of hybrid PV-Biomass-Battery system in a village located in Egypt via off-grid mode [179]. Here, three different battery types are considered as flooded lead-acid, nickel iron, and lithium ferrous phosphate. The optimization process employed include flower pollination algorithm (FPA), Firefly Algorithm (FA), artificial bee colony (ABC), and HS techniques. The optimization 57


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION function proposed required to minimize the NPC subject to LPSP constraint. The results depicted that the FA has a higher performance and a minimum execution time among the others proposed schemes. The optimal solution comprised of 24, 298, and 4 combinations of PV panels, nickel-iron batteries, and biomass power systems, respectively. Duman and Guler in [180] considered a techno-economical analysis of an off-grid hybrid PV-WT-FC-battery system for regular and seasonal vacation homes in Turkey. The authors used HOMER software to investigate seasonal and regular household scenarios using two different storage options: hydrogen and battery. The results revealed that battery storage is more economical and superior as compared to the hydrogen tanks. Fathy in [181] proposed PV-WT-FC HRES for Helwan city, which is located in a remote area of Egypt. A mine blast algorithm (MBA) is proposed and its results are compared with three meta-heuristics optimization schemes: PSO, cuckoo search (CS), and ABC. The author tried to minimize the annual cost and results indicate savings of 24.8%, 11.5576, and % 8.956% as compared to PSO, ABC, and CS, respectively. Additionally, the computational time of the MBA calculated is far less than the other optimization schemes. The papers [172, 182, 183] considered the integration of HRES using GC mode. Here, the authors considered an economic benefit by selling energy back to the grid at high tariff. Ren et al. in [172] considered a GC scenario wherein residential energy demands are fulfilled through PV-FC-Battery system. The authors proposed a multi-objective function to reduce the annual cost along with CO2 emissions. The authors formulated the optimization problem using mixed integer linear programming (MILP) and also considered selling energy back to the grid. The results demonstrated that the battery contributed to economic benefit while PV module is found effective for environmental performance. Similarly, in [182], the authors consider the hybrid PV-FC-Natural gas system to satisfy consumer’s thermal and electrical loads in GC mode. The authors propose two 58


3.3. RENEWABLE ENERGY SOURCES AND ENERGY STORAGE Chapter 3 SYSTEM UTILIZATION heuristic optimization schemes, i.e., PSO with adaptive inertia weight and PSO with constriction factor. An objective is derived to curtail the system cost subject to relevant constraints. The proposed algorithms provided better results for PV-FC system in comparison to an imperialist competition algorithm, GA, and original PSO. Authors in [183] also considered energy management via PV-Battery hybrid system using GC mode. SimPowerSystem is used to model and simulate the proposed hybrid system. The surplus energy is partially or fully supplied back to the grid. The PV panels generated 15.3 kW of energy via maximum power point under irradiation of 1200 W/m2 wherein 5 kW is powered for the complete day to fulfill the high priority load demand and 4 kW is utilized to satisfy the normal load for 23 hours. Islam in [184] conducted a techno-economic feasibility analysis for the incorporation of PV system via GC mode for a large office building located in France. The study also considered the incorporation of hydrogen and electric cars in the system with an objective to curtail the electricity consumption usage from the grid. The results revealed that 43% electricity consumption is curtailed when PV is utilized. Further, the proposed system also resulted in 90% minimization of carbon emission. Maleki and Poufayaz in [130] proposed a hybrid PV-WT-Battery HRES to fulfill consumer’s demand for electricity with a minimum annual cost for site Rafsanjan, Iran. The hourly consumer’s load, wind speed, insolation, and ambient temperature profiles data during a year are used for the HRES. The results showed that ABSO performed better and also considered robust in contrast to other metaheuristic algorithms. Zahboune et al. in [171] proposed a method called as modified electric system cascade analysis for hybrid electricity generation using PVWT-Batteries at site Oujda city, located in northeast Morocco. A daily energy demand of 18.7 kWh is considered with an aim to reduce the annual cost with minimum LPSP. The optimized results obtained from MATLAB/SIMULINK are also validated through HOMER Pro. A difference of 0.04%, 5.4%, and 0.07% are observed in energy production, electricity excess, and energy cost, respectively. 59


Chapter3.4. 3 OPTIMIZATION METHODS PROPOSED IN THE LITERATURE Various energy sources and ESS for hybrid systems are summarized in Fig. 3.1. Various energy sources

Renewable energy sources

Non-renewable energy sources

Energy storage systems

Photovoltaic

Diesel generator

Fuel cell

Wind

Gas turbine

Batteries

Flooded lead-acid

Hydro

Nickel iron Biomass Lithium ferrous phosphate

Figure 3.1: Energy sources for hybrid system

3.4

Optimization methods proposed in the literature

In the fourth thread, optimization methods proposed in the literature are focused. These optimization schemes are majorly classified into two broad categories: formal and meta-heuristics. The formal, mathematical or deterministic methods find an exact optimal solution for convex and linear nature of problems. The metaheuristic approaches follow a different approach, for instance, the word meta means higher level or beyond and heuristic means to discover something by trial and error or to find something [185]. Thus, meta-heuristics methods are utilized to find an approximate or optimal solution by iteratively searching the space.

60


Chapter3.4. 3 OPTIMIZATION METHODS PROPOSED IN THE LITERATURE The PSO method is widely applied meta-heuristic approach used to find unit sizing solution for HRES [181, 182, 186, 187]. In [186], the authors employed PSO to minimize try-objective; problems of cost, CO2 emission, and load probability loss of PV-Diesel-Battery HRES for an isolated land located in Kilis, Turkey. The results revealed that the PV-Battery system significantly reduced the energy cost by cutting the amount of diesel used under unmet load conditions. Similar to [187], the authors applied PSO to minimize a tri-objective for an isolated village Llama, located in Algeria. The PV-Battery system reduced the cost and CO2 emission. The GA is used in unit sizing problems because of its advantages: dealing with multi-objective problems, good convergence, flexibility to adapt and high efficiency as compared to formal approaches [188, 189]. Further, it is also found effective in the handling of non-linear problems efficiently and its capabilities to globally search the domain for an optimal solution. The GA is applied to minimize energy cost in work [174, 70, 182, 190]. In [190], Merei et al. considered a GA optimization technique to reduce the overall cost of the SA system using PV-WT-Diesel hybrid system for a site located in Quneitra, Syria. The results depicted that PVs in conjunction with batteries are both ecological and economical. In addition, the researchers applied numerous other meta-heuristic techniques: ABSO [130], HS [129, 179], CS, and ABC [181] and so on. It is vital to state that all above proposed meta-heuristic techniques require common control as well as algorithmic-specific parameters and their performance tuning to achieve the optimal results. Therefore, it is vital to opt algorithms that do not depend upon the algorithm-specific parameters for an optimal solution. The next section describes our motivation for such algorithms.

61


Chapter 3

3.5

3.5. ANALYSIS OF LITERATURE

Analysis of literature

Based on the four threads discussed in the related work and optimization schemes in Chapter 2, it can be summarized that the major challenges associated with DSM are the consumer’s cost, PAR, and discomfort minimization. Formal techniques as well as metaheuristics, both are widely applied optimization schemes in the SG domain. The static priorities assigned to the appliances are more frequently applied by the researchers in the SHs as compared to the dynamic priorities. UC is mostly associated with the thermal or the delay caused by the appliances’ utility deviation from its normal usage pattern. In the references [65, 66, 81, 83, 85] the work carried out do not take actual load profiles of appliances. The actual load profile is replaced by the maximum or average load profile of appliances. As with respect to time load profile of appliance varies thus, scheduling without considering the actual load profile may not always give a feasible solution. In [72], actual load profile of appliances considering different energy cycles is considered along with sequential execution constraint without operational interruption. The optimization problem was solved using MILP technique. However, the UC has not been addressed in [72]. Further, the UC is a relative term which also varies among consumers and situations. In literature, many authors have associated the UC with appliance delay occurred in shiftable appliances, while others relate it to thermal comfort. On the DSM, UC is a vital characteristic that may also be obeyed with the cost minimization problem. Moreover, mathematical techniques used in [21, 72, 84, 85], are highly computational and their complexity increases exponentially with the growth of problem size. These deterministic approaches are formulated with known parameters. Further, integration of RESs as used in works [66, 72, 80, 85] may also be considered in HEM schemes to reduce the peaks and CO2 emissions. Based on this analysis of review next discussed is the problem statement along with contributions.

62


Chapter 3

3.6

3.6. SUBPROBLEM STATEMENTS AND CONTRIBUTIONS

Subproblem statements and contributions

The five subproblems along with contributions are as follows:

3.6.1

Subproblem-1 and contributions

The subproblem-1 is summarized in the following points: (i) In [67], the UC is associated with only thermal comfort while others [21, 68, 69] relate it to the appliances’ delay, cost savings, and ROI parameters. Ogunjuyigbe et al. considered the user’s satisfaction which is based on three satisfaction postulates [70]. However, none of the referenced work has considered the UC derived from time-varying priorities. (ii) In [120, 121, 122], the authors primarily focused on the electricity cost minimization problem and the PAR reduction without considering the user’s priorities for the home appliances. The priority management is a major concern in the electricity load management problem as it increases user satisfaction level. (iii) In [158], the authors prioritized the operation of power units from a consumer perspective. For instance, the solar power source unit is assigned a high priority as compared to other installed power units. However, the priorities assigned to the home appliance’s have not been addressed, as it plays a key role in maximizing the UC. (iv) While others specifically have worked on assigning static priorities to the appliances [50, 99, 159, 161]. The static priorities are time-independent and remain constant throughout a day. However, the existing approaches have not assigned the dynamic priorities to the appliances. In dynamic priorities, the weights assigned to the appliances vary with the passage of time according to the consumer’s willingness.

63


Chapter 3


(v) The consumers are bound to various constraints in their attempt to the electricity usage. The budget limitation is one of the prominent constraints which needs to be satisfied. In this dissertation, the consumer budget limit is considered and time-varying priorities are used to derive an absolute UC. In TB priority, time-varying priorities are assigned to an appliance by the consumer in various time-slots in a day. In DB priority approach, the consumer assigns relative priorities to the appliances in each operational time interval (OTI). The relative priorities assigned also vary in the different time-slots within a day. The contributions are listed below: (i) In the view of static and dynamic priorities assigned to the appliances by a typical consumer, this thesis further considers the dimension of deriving absolute comfort. (ii) A GA based EACA is developed that is capable of generating an optimal energy consumption pattern which would give maximum UC. (iii) The proposed EACA satisfies the predetermined user budget and energy constraints and also achieves maximum absolute comfort when it is compared to the randomly generated base cases. (iv) A χ index is also developed, which relates the total users’ expenditure (Texp ) to the total achieved comfort.

3.6.2


The work carried out in references [65, 66, 84, 85, 118, 119], lack prioritizing the operation of shiftable appliances based on consumer’s viewpoint. HEM schemes enabled with appliance priorities defined by consumer willingness is essential. In [50, 52], the authors assigned priorities to the appliances. In [38], Yi et al. have used OSR to shift the operation of appliances from peak hours based on priorities to save electricity cost using RTP scheme. The authors have applied a threshold policy and considered waiting along with electricity cost in their objective function. 64


Chapter 3


However, the authors [38, 52] have not addressed the PAR, which is an essential factor to control the peak load demand in the SG. Further, the RTP signal used in [38] has its own limitations and increases network bandwidth. Similar to [38], we take a single home wherein different energy cycles of a CD, DW, and refrigerator are considered for scheduling using DA-RTP scheme. Our proposed DSM-based strategy uses four optimization techniques: GA, BPSO, EDE, and OSR to shift various energy cycles of an appliance based on consumer-defined priority values. Individual appliance and their aggregate load are considered in two different scenarios. The aggregate load scenario is further categorized using no load limit and knapsack capacity load limit. The simulation results in section 5.2 validate that the proposed DSM-based strategy performs well and significantly reduces both the consumer electricity cost and the appliances’ delay with respect to the low and high priorities set by the consumer. The contributions are given as follows: • Monthly and yearly electricity cost minimization based on consumer-defined priority. • Average delay (user discomfort) minimization with respect to consumer defined priority. • Various energy cycles of an appliance are considered for load shifting. • Appliance priority (µp ) and pricing threshold policy are incorporated. • Problem formulation using multiple knapsacks to mitigate the rebound peaks is demonstrated. Based on the contributions above, the significance of our work is beneficial to both consumers and utility. By maintaining the knapsack capacity limit and shifting of appliances operational time from on-peak to off-peak hours not only reduces the consumer’s electricity cost but also ensure the grid’s stability by maintaining PAR level.

65


Chapter 3

3.6.3



The rapid increase in household electronic appliances significantly increase the electricity demand of the residential sector. Almost 40% of the generated electricity is consumed by the residential sector [191]. Currently, power generation is heavily dependent on fossil fuels and the hour of need is to fulfill the inevitably increasing electricity demand with minimal emissions of greenhouse gases. Therefore, scientists have worked to figure out new means of electricity generation, i.e., RSERs. In this context, the integration of RSERs and ESSs become lucrative for the researchers. Moreover, the conventional power grid is already vulnerable to instability due to heavy loaded conditions and will be unable to maintain its stability if the integration of RESs is done at a large scale. So, the research is going on to implement the SG in a distributed manner along with an integrated HEMS to optimize the energy utilization. Keeping this objective in mind, we seek to develop an OHEMS. Our objective is twofold: i) integration of RES and ESS into the residential sector, ii) energy management through appliances and resource scheduling. In order to achieve the above-mentioned objectives, and to design an efficient HEMS, various algorithms such as LP [192], ILP [193], MILP [194], DP [195], and convex programming (CP) [196] have been proposed. However, these techniques have a very slow convergence rate, and in some cases, they are unable to cope with a huge number of appliances. So, the heuristic algorithms such as GA [197], BPSO [104], WDO [198] and BFO [199] are introduced to overcome these problems. The heuristic optimizations are used where it is very difficult to find the exact optimal/feasible points. For example, in the case of LP, it is understood that the optimal solution must lies within the solutions points in that pool. However, in heuristic optimization, there might be infinite or more solution points in that solution space and optimal solutions should be anyone among those. Moreover, in the aforementioned techniques (LP, ILP, MILP, DP, and CP) based

66


Chapter 3


HEMSs, the integration of RESs, maximization of UC and adaptability with dynamic pricing are ignored. Therefore, in this part of thesis contribution, not only the meta-heuristic algorithms (BPSO, GA, WDO, BFO, and HGPO) are used to design an OHEMS but their performance is also evaluated in terms of energy consumption pattern, PAR, and electricity bill reduction. Some contributions to this subproblem are given as follows: • RESs are integrated into SH via GC mode. • Classification of home appliances into two different categories considering user preferences. • A hybrid solution based on two different meta-heuristic is proposed to have better exploratory and exploitative innovation. • Aggregate appliances load is considered.

3.6.4


The iterative method used as a benchmark in [128] and MILP technique applied in [90] for unit sizing problems are highly computational. In addition, these classical approaches suffer from the curse of dimensionality problem and may not find the optimum solution within a reasonable amount of time [104, 131]. In [80, 173], the authors used HOMER software tool for unit sizing and performance analysis of HRES. The tool used has some limitations, for instance, it does not support multi-objective problems, its formulation, depth of discharge (DOD) of the battery bank, and ranking of HRES based on energy levelized cost. Further, it also does not consider intra-hour and bus voltage variations, which make it difficult to apply in many situations [129, 200]. Optimal sizing of the HRES is a non-convex and non-linear optimization problem which requires an efficient optimization technique for its solution [167]. Considering these limitations, some authors proposed meta-heuristic approaches to find the globally optimized solution with less computational complexity [128, 129, 130, 67


Chapter 3


131, 167]. These meta-heuristic approaches require algorithmic-specific parameters and its performance tuning, which, if not properly tuned may result in high computation time along with local optimal solutions [132]. Therefore, in this part of thesis contribution, a recently proposed algorithm Jaya is applied, which does not require any algorithmic-specific or control parameters for its functioning. The main contributions of this dissertation pertaining to this subproblem are given as: • System components are formulated and elaborated using an informative HRES model. • Motivated from the non-algorithmic-specific meta-heuristic approaches, Jaya algorithm is implemented to find an optimal number of components required in HRES to minimize consumer annual energy cost. • The real data for wind speed and solar irradiance is used to find out the optimal unit sizing of PV-WT-FC, PV-FC, and WT-FC systems, for Hawksbay, Pakistan. The system reliability is considered using different LP SP values provided by the consumer. • To test the efficacy of the proposed methodology, Jaya is implemented in two different consumer’s scenarios, i.e., low and high user load profiles. • Further, a comparison of Jaya results with two algorithmic-specific schemes: GA and BSA is also presented.

3.6.5


Following are some limitations identified in the literature, which have motivated us to utilize the non-algorithmic-specific meta-heuristic methods for unit sizing problem.

3.6.5.1

HOMER software limitations

The HOMER software is widely applied for techno-economic feasibility analysis and optimal solution for unit sizing of HRES [170, 171, 177, 178, 180]. The 68


Chapter 3


HOMER has been effective yet it has following limitations [182, 200]: • HOMER software is not capable to deal with multi-objective optimization and its formulation. It only supports uni-objective function which is focused on the minimization of NPC. • In HOMER software, the solutions obtained are optimized only via TNPC values and it does not rank HRES solutions based on the levelized energy cost. • It also does not support intra-hour basis variability. • Further, it acquires huge computational time for larger design points. • Additionally, HOMER software does not consider the DOD of the battery and different variations in bus voltage.

3.6.5.2

Formal techniques limitation

Some work like [172, 90, 201] solved the optimal sizing problem via MILP mathematical technique. The formal techniques used have some limitations. • MILP technique applied for unit sizing is highly computational because the complete search space of the problem domain is explored to find the exact optimal solution. • These techniques are not suitable for stochastic environments or real-time environments. • In addition, these classical schemes suffer from the curse of dimensionality problem and may not find the optimum solution within a reasonable amount of time [131, 104].

3.6.5.3

Algorithmic-specific meta-heuristic limitations

Considering the limitations of HOMER software and formal methods, some studies suggest the use of meta-heuristic algorithms: ABSO [130], GA [174, 176, 182], HS [129, 179], PSO [181, 182, 186, 187], etc., which are more successive and have a high efficiency rate [202]. The following are some of its limitations: 69


Chapter 3


• The techniques: HS, PSO, GA, ABSO, etc., require algorithmic-specific parameters for their functioning. For instance, the HS algorithm uses pitch adjusting, harmony memory, and the consideration rate with a number of improvisations. The PSO requires cognitive and social parameters with inertia weight values. The GA needs a selection operator along with crossover and mutation probabilities. Similarly, ABSO cannot be executed without initialization and adjustment of the algorithmic-specific parameters: number of employed, scout, and onlooker bees with a limit specifier. • The algorithms like GA, ACO, etc., also require performance tuning of the algorithmic-specific parameters for achieving the optimal results. • The algorithmic-specific parameters, if not tuned properly, may yield to locally optimal solutions or result in an increased computational time. Due to non-algorithmic-specific parameters advantages, the research community has widely applied these Jaya, TLBO, and ITLBO algorithms in various optimization domains. For instance, the TLBO is implemented for appliance scheduling for energy management [203, 204], parameter estimation of PV model [205], and optimal scheduling of wind-thermal power system [206]. Considering these limitations, in this thesis, we are motivated to solve the unit sizing problem of HRES via Jaya, TLBO, and their hybrid JLBO algorithms. We consider the hybrid PV-WT-Battery system, which is more eco-friendly and costeffective as compared to other hybrid systems that utilize generators. Thus, the contributions listed below are an extension of our previous work [207]: • System components of PV-WT-Battery is formulated and elaborated using an informative HRES model. • Motivated by non-algorithmic-specific approaches: Jaya and TLBO algorithms are proposed to find an optimal number of HRES components to reduce the consumer’s annual electricity cost in the SA environment. • A hybrid approach JLBO is proposed by combining Jaya and learning phase of TLBO algorithms for better search space. 70


Chapter 3


• In addition, the results obtained by non-algorithmic-specific schemes are also compared with GA that depends upon algorithm-specific parameters, including crossover and mutation. • The yearly real data for wind speed, ambient temperature, and solar irradiance is utilized to find out an optimal unit sizing of PV-WT-Battery, PV-Battery, and WT-Battery systems for Rafsanjan, Iran. • The system reliability is considered using different LP SP max values provided by the consumer.

71


Chapter 4 System models, problem formulations, and proposed solutions

72

Chapter 4

4.1. PRIORITY CONCEPT AND POSTULATES

Chapter summary This chapter is comprised of system models, problem formulation, and proposed meta-heuristic solutions. The first part of the chapter describes the priority concept and postulates in section 4.1, EACA based proposed solution in section 4.1.2 and mapping steps for the priority induced scheduling problem with GA are given in section 4.1.3. The second part deals with the appliance scheduling problem, which is formulated as an optimization problem using MKP. The problem formulation of three appliances: CD Acd , DW Adw , and refrigerator Aref in terms of load categorization, PAR, threshold, and the objective function with constraints are discussed in section 4.2. A proposed solution consisting of four optimization schemes EDE, GA, BPSO, and OSR is elaborated in section 4.2.6. The third part depicted in section 4.3 shows GC PV, ESS, and appliances problem formulation. Proposed system architecture and meta-heuristic scheduling algorithms are given in section 4.3.7, and section 4.3.8, respectively. The fourth part as described in section 4.4 integrates renewable and storage systems in the SA environment. Next, its system configuration, sizing formulation, and proposed methodology in section 4.4.7 are elaborated. The last part as given in section 4.5 elaborates system model, proposed formulation, and hybrid-based approach for solving unit sizing problem in SA mode for another site location.

4.1

Priority concept and postulates

This section is focused on the subproblem discussed in section 3.6.1. In general, various types of electronic and electrical appliances are installed in the SH. The appliances demand different power ratings, which are essential to perform its operation. Further, these appliances have different priority levels assigned by the consumers, which also varies across different hours of the day. Consumers often have budget limitations to spend on their energy demands. In general, the challenge associated with the HEMC is to determine device and time usage pattern 73


Chapter 4


that may assign power to those appliances that yield maximum comfort value within a predefined user budget. Here, the consumer Texp is related to the derived UC and a χ index is developed, which is expressed in Eq. (4.1).

χ($) =

Texp ($) . µ

(4.1)

Where Texp is the total users’ expenditure, and µ represents the total achieved comfort (µachieved ) value. It is obvious from Eq. (4.1) that the consumer wants to maximize the µ value within defined budget and energy constraints to have a minimum value of χ. In order to achieve the objectives, four priority postulates are derived. (i) Priority, ρ is quantifiable and it can be numerically evaluated. (ii) Priority, ρ is not crisp. It is fuzzy in nature. The values of priority exist in between the states of the lowest priority, ρ = 0 and the highest priority value, ρ = 1. (iii) ρ = 1 value shows the prominent user’s satisfaction as compared to ρ = 0 value assigned to an appliance. (iv) Priority is relative and comparative in nature. The two ρ levels are defined as TB and DB relativities. In TB relativity, the time-varying priorities are assigned to an appliance which change with respect to different time-slots in a day. For instance, if an appliance exists #J, then the priority set by the consumer in time τ1 (ρJ (τ1 )) is relative and can also be compared to the priority, it provides at time τ2 (ρJ (τ2 )). In case of DB relativity, the priority assigned to an appliance #J (ρJ (τ1 )) can be compared to other appliance #K (ρK (τ1 )) for that particular hour. It is similar to the concept of static priorities assigned to the appliances in a day. In this thesis, a new concept of dynamism is introduced in which the time-varying priorities are assigned to the appliances based on the consumers’ preferences. Further details are given in section 4.1.1.2. 74


Chapter 4

4.1.1


User’s priority and comfort enabled system model WAN

Utility

AMI

Figure 4.1: Consumer priority enabled system model for a home

The proposed system model consists of one SH, which is further divided into six sections. Each section is equipped with a number of smart appliances. The details of each section, appliances, their power ratings, quantity, total energy consumption along with energy cost are elicited in Table 4.1. Here scheduling of the appliances is performed on total power rating (TPR). Further, we have considered TPR in our proposed work and each appliance is different from each other. However, quantity varies (Table 4.1), for instance, all lights in the TV lounge have the same priority. Therefore, the scheduling is only performed on the aggregate (total appliances) and we have not considered similar appliances individually. The ET obtained from the utility is assumed to be constant at 0.115 $/kWh. This ET is used to calculate the total energy available (TEA) to the user by the energy consumption of appliances at a defined budget. The appliances’ scheduling is considered for a day horizon, which is equally divided into 24 time-slots. In this model, we do not consider energy dissipation, vampire loads or other standby power, which 75


Chapter 4


may cause electricity losses while they are switched off and not performing their primary function. The overall architecture of the proposed system model consists of a HAN, AMI, HEMC, SM, and user-assigned priorities. The consumer’s priority enabled system model for a home is demonstrated in Fig. 4.1. The transmission and communication infrastructure is briefly elaborated in subsection 4.1.1.1. The priorities and absolute comfort derived are discussed in subsection 4.1.1.2. The schematic of the process is drawn and briefly discussed in subsection 4.1.1.6. Table 4.1: Energy consumption and cost S/N

Section

1 2

TV lounge

3

Appliance with short form

Power rating (kW)

Quantity

LCD TV (TV)

0.1500

1

Total power rating (kW) 0.1500

Cost 0.115 ($/kWh) 0.0172

Lighting (Ligt.)

0.0200

5

0.1000

0.0115

AC

1.2000

1

1.2000

0.1380

Water heater (WH)

2.0000

1

2.0000

0.2300

Lighting (Ligt.) Washing machine (WM) CD

0.0200

2

0.0400

0.0046

0.7000

1

0.7000

0.0805

1.8000

1

1.8000

0.2070

8

Lighting (Ligt.)

0.0300

2

0.0600

0.0069

9

Microwave oven (MO)

1.5000

1

1.5000

0.1725

Juicer

0.4000

1

0.4000

0.0460

DW

1.4000

1

1.4000

0.1610

12

Refrigerator (Ref.)

2.1000

1

2.1000

0.2415

13

Lighting (Ligt.)

0.0300

3

0.0900

0.0104

Lighting (Ligt.)

0.0200

8

0.1600

0.0184

CCTV camera

0.0090

3

0.2700

0.0311

4

Bathroom

5 6 7

10 11

14

Laundry

Kitchen

Surveillance

15 16 17 18

4.1.1.1

Master bedroom

Lighting (Ligt.)

0.0200

5

0.1000

0.0115

Laptop

0.0600

1

0.0600

0.0069

Mobile charger (MC)

0.0060

1

0.0060

0.0001

Transmission and communication infrastructure

Basic transmission and communication infrastructure between a consumer and the utility is assumed to be present. The power and data flow are represented by solid and dashed lines, respectively. The utility sends electricity through wired routes indicated by the solid lines. While data flow occurs bi-directionally via wireless routes using AMI. These wireless routes forming wide area network (WAN) use advanced communication technologies like universal mobile telecommunication 76


Chapter 4


system (UMTS), long-term evolution (LTE), and WiMAX for robust and reliable communication between the consumer and the utility [208, 209]. The HAN installed at home is responsible for communication among SM, HEMC, and appliances. Various low-cost technologies, including Wi-Fi, Zig-Bee, and Bluetooth are mostly adapted to provide effective and reliable communication in a home [208, 210]. HEMC is one of the main components which receives appliance power ratings, ET, user priorities, and budget constraints to provide a schedule for the home appliances. The EACA is an embedded part of HEMC that takes decisions by generating an optimal pattern that would give maximum UC within a defined budget limit. HAN is responsible for transferring HEMC scheduled pattern decisions to the appliances. Based on the signals received from the HEMC, the appliances change their operational status to ON/OFF modes accordingly.

4.1.1.2

Absolute comfort derivation using time and device based priorities

Like UC, priority is also a relative term which varies from consumer to consumer and situation to situation. In literature, authors have assigned the static and dynamic priorities to the home appliances. In this thesis, absolute comfort is derived from the time-varying relative priorities assigned to the appliances by a typical consumer. To achieve this objective, two tables (TB priority Table 4.2 and DB priority Table 4.3) are taken as an input from the consumer. The absolute comfort table is then derived using the TB and DB priority tables which are discussed next.

4.1.1.3

Time based priority table

In the TB priority Table 4.2, the consumer fills up the table with fuzzy values for all time-slots and appliances which are as follows: (i) If appliance #J is used for a day, then the consumer may decide in which hour, τmax , its utility is the highest. 77


Chapter 4

4.1. PRIORITY CONCEPT AND POSTULATES Start

Select an appliance i.e. #J

Select a time slot (Tmax) where appliance utility is highest

Assign priority value 1 to that time slot (Tmax)

Initialize time slot index i=1

Assign priority values Pi to all time slots (Ti) in comparison to highest priority

Increment time slot index i= i+1

No

If i > Total time slots in a day

Yes

Is there any more appliances left?

Yes

No Stop

Figure 4.2: Flow chart of time based appliances priority

(ii) The maximum priority value, ρJ (τmax ) = 1 is assigned to the base hour (τmax ). (iii) The remaining hours of the day (τi ) are compared with (τmax ) and the user decides which level of priority is assigned, if appliance #J is used. Based on the aforesaid points, it is obvious that time-varying priorities ρJ (τi ) are assigned to an appliance at different hours that range between the lowest and the

78


Chapter 4


highest priority values, which are expressed in Eq. (4.2).

0 ≤ ρJ (τi ) ≤ 1,

∀ i = [1, 24].

(4.2)

The TB priority Table 4.2 filled by a typical consumer is selected for this study. As already stated that the priority of an appliance may vary from person to person and situation to situation. Here, the consumer priority values are assigned to the appliances which are filled horizontally for all the time-slots. For instance, in Table 4.2, it can be noticed that the user has assigned a maximum TB priority value to LCD TV in time-slots 21 and 22 because of favorite talk show timings. Similarly, the rest of the hours are compared and the priorities are assigned accordingly based on the user’s preferences. The flow chart of TB priority is represented in Fig. 4.2.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

TV lounge

TV

0.0

0.0

0.0

0.0

0.0

0.0

0.4

0.3

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.2

0.4

0.7

0.8

1.0

1.0

0.7

Ligt.

0.0

0.0

0.0

0.0

0.0

0.9

0.8

0.5

0.3

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.4

0.6

0.7

0.8

1.0

0.9

0.7

AC

0.0

0.0

0.0

0.1

0.2

0.4

0.3

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.2

0.2

0.3

0.5

1.0

0.9

0.5

Bath.

WH

0.0

0.0

0.0

0.2

0.4

1.0

0.6

0.4

0.3

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

Ligt.

0.0

0.0

0.0

0.0

0.5

1.0

0.4

0.2

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.3

0.3

0.4

1.0

0.7

0.4

Laundry

WM

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

1.0

0.3

0.2

0.1

0.0

0.0

0.0

0.0

0.1

0.2

0.5

0.0

0.0

CD

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

1.0

0.3

0.1

0.1

0.0

0.0

0.0

0.1

0.2

0.5

0.0

0.0

Ligt.

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

1.0

1.0

0.4

0.2

0.1

0.0

0.0

0.0

0.1

0.2

0.2

0.0

0.0

MO

0.0

0.0

0.0

0.0

0.3

0.9

1.0

0.5

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.3

0.3

0.4

0.0

0.0

Juicer

0.0

0.0

0.0

0.1

0.2

1.0

0.8

0.3

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.5

0.4

0.2

0.1

0.0

0.0

0.0

0.0

DW

0.0

0.0

0.0

0.1

0.2

0.3

0.5

1.0

0.2

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

1.0

0.2

0.1

0.0

Ref.

0.0

0.0

0.0

0.0

0.5

1.0

0.7

0.4

0.1

0.1

0.2

0.0

0.0

0.0

0.0

0.1

0.1

0.9

1.0

0.5

0.3

0.2

0.1

Ligt.

0.0

0.0

0.0

0.0

0.1

0.2

1.0

0.3

0.1

0.1

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.5

0.8

0.0

0.0

Surveil.

Ligt.

1.0

1.0

1.0

0.9

0.8

0.5

0.3

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.3

0.5

0.8

0.9

1.0

CCTV

1.0

1.0

1.0

0.9

0.8

0.5

0.3

0.2

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.5

0.8

1.0

0.8

Ligt.

0.0

0.0

0.0

0.0

0.3

1.0

0.8

0.4

0.1

0.1

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.5

0.8

0.9

1.0

Laptop

0.0

0.0

0.0

0.0

0.2

0.4

0.4

0.1

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.5

0.8

1.0

0.8

MC

1.0

0.7

0.5

0.4

0.3

0.3

0.2

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.3

0.4

0.5

Kitchen

Sc. App.

M.bedroom

Table 4.2: Time based priority

79


24 0.5 0.4 0.2 0.0 0.3 0.0 0.0 0.2 0.0 0.0 0.0 0.0 0.2 1.0 0.7 0.2 0.7 0.8

Chapter 4 4.1.1.4

4.1. PRIORITY CONCEPT AND POSTULATES Device based priority table

The DB priority table data is filled vertically by the consumer. The following procedure is adopted to fill up the DB priority table: (i) The consumer selects a single time-slot or a particular hour τi . (ii) If there are numerous appliances, i.e., #J, #K, #L, etc., available to the consumer, then the consumer may set a priority value of 1 for a device #J, because of its maximum utility at that respective time-slot ti . (iii) The rest of the appliances are compared to this base device #J and the relative priority values ρK (τi ), ρL (τi ), etc. are assigned by the consumer according to their willingness at τi . These values range from peak priority value of 1 to the lowest priority value of 0. Table 4.3 shows the DB priority values for a typical electricity consumer. For instance, at a time-slot 2, the maximum device-base priority value 1 is assigned to the lighting and CCTV appliances. The rest of the devices are assigned relative priorities with respect to this maximum priority value.

4.1.1.5

Absolute comfort table

The absolute comfort is derived from both TB and DB priority values. The absolute comfort of an appliance #J is calculated in Eq. (4.3) [70]: p (ρtJ (τi ))2 + (ρdJ (τi ))2 √ , µJ (τi ) = 2

∀ i = [1, 24].

(4.3)

Where ρtJ (τi ) and ρdJ (τi ) represent the TB and DB priority values of the appliance #J at time τi , respectively. In order to normalize the absolute comfort to a √ maximum value of 1, denominator of 2 is used in Eq. (4.3). The absolute UC is illustrated in Fig. 4.3. For instance, Table 4.2 reveals the TB priority value which is equal to 0.5 for the mobile charger at τ3 . Whereas Table 4.3 shows the DB priority value 0.7 at the same time τ3 . Thus, the absolute comfort of the mobile charger at τ3 can be calculated as: 80


Chapter 4


2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

0.0

0.0

0.0

0.0

0.0

0.0

0.3

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.2

0.4

0.7

0.8

1.0

1.0

0.7

Ligt.

0.0

0.0

0.0

0.0

0.0

0.3

1.0

1.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.3

0.6

0.4

0.8

1.0

0.9

0.7

AC

0.0

0.0

0.0

0.0

0.1

0.2

0.3

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.2

0.2

0.3

0.5

1.0

0.9

0.5

Bath.

WH

0.0

0.0

0.0

0.2

0.3

1.0

0.3

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

Ligt.

0.0

0.0

0.0

0.1

0.3

0.9

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.5

0.3

0.4

0.1

0.0

0.0

Laundry

WM

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

1.0

0.3

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.2

0.0

0.0

CD

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.5

1.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.2

0.0

0.0

Ligt.

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.9

0.5

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.2

0.0

0.0

MO

0.0

0.0

0.0

0.0

0.3

0.6

1.0

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

1.0

1.0

1.0

0.5

0.3

0.2

0.0

Juicer

0.0

0.0

0.0

0.1

0.2

0.5

1.0

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.2

0.7

1.0

0.4

0.1

0.1

DW

0.0

0.0

0.0

0.1

0.2

0.1

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

1.0

0.1

0.7

0.0

Ref.

0.0

0.0

0.0

0.3

0.3

0.6

0.4

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.8

0.9

1.0

0.5

0.3

0.2

0.1

Ligt.

0.0

0.0

0.0

0.1

0.2

0.1

1.0

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.2

0.0

0.0

Surveil.

Sc. App.

Ligt.

1.0

1.0

1.0

1.0

0.9

0.3

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.4

0.5

0.7

1.0

CCTV

1.0

1.0

1.0

0.9

1.0

0.3

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.3

0.4

0.3

0.4

Ligt.

0.2

0.1

0.1

0.2

0.3

0.5

0.3

0.5

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.1

0.3

0.3

1.0

Laptop

0.1

0.2

0.3

0.3

0.3

0.4

0.2

0.3

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.1

0.2

1.0

0.8

MC

1.0

0.9

0.7

0.3

0.2

0.1

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.1

0.1

0.1

0.2

Kitchen

TV

1

TV

M.bedroom

Table 4.3: Device based priority

(4.4)

co m lu te bs o A

DB priority

fo

rt

p (0.5)2 + (0.7)2 √ ≈ 0.6. µJ (τ3 ) = 2

TB priority

Figure 4.3: Absolute user comfort derived from time and device based priorities

Table 4.4 shows the absolute comfort obtained from both TB and DB priority tables using Eq. (4.3). 81


24 0.3 0.2 0.1 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.5 0.4 1.0 0.9

Chapter 4


1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

TV lounge

TV

0.0

0.0

0.0

0.0

0.0

0.0

0.4

0.3

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.2

0.4

0.7

0.8

1.0

1.0

0.7

Ligt.

0.0

0.0

0.0

0.0

0.0

0.7

0.9

0.8

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.4

0.6

0.6

0.8

1.0

0.9

0.7

AC

0.0

0.0

0.0

0.1

0.2

0.3

0.3

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.2

0.2

0.3

0.5

1.0

0.9

0.5

Bath.

WH

0.0

0.0

0.0

0.2

0.4

1.0

0.5

0.3

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

Ligt.

0.0

0.0

0.0

0.1

0.4

1.0

0.3

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.4

0.3

0.4

0.7

0.5

0.3

Laundry

WM

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

1.0

0.3

0.1

0.1

0.0

0.0

0.0

0.0

0.1

0.2

0.4

0.0

0.0

CD

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.4

1.0

0.2

0.1

0.1

0.0

0.0

0.0

0.1

0.2

0.4

0.0

0.0

Ligt.

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

1.0

0.8

0.3

0.1

0.1

0.0

0.0

0.0

0.1

0.2

0.2

0.0

0.0

MO

0.0

0.0

0.0

0.0

0.3

0.8

1.0

0.4

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.7

0.7

0.7

0.4

0.4

0.1

0.0

Juicer

0.0

0.0

0.0

0.1

0.2

0.8

0.9

0.2

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.4

0.3

0.2

0.5

0.7

0.3

0.1

0.1

DW

0.0

0.0

0.0

0.1

0.2

0.2

0.4

0.7

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

1.0

0.2

0.5

0.0

Ref.

0.0

0.0

0.0

0.2

0.4

0.8

0.6

0.3

0.1

0.1

0.1

0.0

0.0

0.0

0.0

0.1

0.6

0.9

1.0

0.5

0.3

0.2

0.1

Ligt.

0.0

0.0

0.0

0.1

0.2

0.2

1.0

0.3

0.1

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.4

0.6

0.0

0.0

Surveil.

Ligt.

1.0

1.0

1.0

1.0

0.9

0.4

0.2

0.2

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.3

0.5

0.7

0.8

1.0

CCTV

1.0

1.0

1.0

0.9

0.9

0.4

0.2

0.2

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.4

0.6

0.7

0.6

Ligt.

0.1

0.1

0.1

0.1

0.3

0.8

0.6

0.5

0.1

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.2

0.4

0.6

0.7

1.0

Laptop

0.1

0.1

0.2

0.2

0.3

0.4

0.3

0.2

0.1

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.4

0.6

1.0

0.8

MC

1.0

0.8

0.6

0.4

0.3

0.2

0.2

0.1

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.1

0.1

0.2

0.2

0.3

0.4

Kitchen

Sc. App.

M.bedroom

Table 4.4: Absolute comfort

4.1.1.6

Schematic of the process

The schematic of the process (i.e., inputs, process, and outputs) is drawn in Fig. 4.4. The input includes the time and device based priority values, consumer budget limit along with appliances TPR. Based on these inputs the HEMC computes the absolute comfort and finds the appliance’s scheduling pattern that achieves maximum comfort, which is the output. Inputs

Time-based priority

Consumer’s budget limit

Output

Processing Devicebased priority

Total power rating

Absolute comfort

Maximum comfort achieved?

HEMC

Yes

Appliances scheduling pattern

No

Figure 4.4: Schematic of the process

82


24 0.4 0.3 0.2 0.0 0.2 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.1 1.0 0.6 0.3 0.9 0.9

Chapter 4

4.1.2


Proposed solution

The EDE, GA, PSO, WDO, etc. are widely applied metaheuristic algorithms in different areas of SG [80, 120, 211, 212]. The GA is inspired by the evolutionary process of a living organism which is based on an iterative procedure for finding the global optimal solution with certain execution time limits. The GA is effective because it has a good convergence rate, flexible nature and also efficiently handles both: the non-linear and non-smooth optimization problems [70, 213, 214]. Considering these advantages, the proposed EACA is also based on GA. The GA includes genetic properties and offers a potential solution in the form of a generic chromosome. The chromosome consists of a binary string. A set of chromosomes is produced in the initialization phase of the GA. For instance, the initialization phase may generate a Z c number of chromosomes, then, a fitness function is used to compute and evaluate these chromosomes. The chromosomes which represent candidate solutions are ranked based on the fitness function values. After the evaluation of chromosomes, the GA principle will repeat to generate a new generation of chromosomes until a stopping criterion is met. The new chromosomes are also evaluated based on the fitness function. At each generation phase, the GA principle retains the best solutions and also discards the worst ones to achieve a globally optimal solution.

4.1.2.1

Objective of evolutionary accretive comfort algorithm

The objective of EACA is to generate a scheduling pattern that yields maximum absolute comfort as given in Eq. (4.5).

Obj(µ, β) = max(µ),

(4.5)

where, β is subject to the user defined budget. It can also be observed from Eq. (4.1), the χ index, is a function of both the total users’ expenditures Texp and 83


Chapter 4


Algorithm 1 EACA 1: procedure EACA 2: Input : 3: Initialize parameters like population size, crossover rate, mutation rate, termination criteria 4: Read TB and DB priority tables, energy consumptioncost table, user’s budget limitation, ET 5: Initialization: Generate an initial population; 6: Compute absolute comfort values by using Eq. (4.3) 7: Compute total energy availability by using Eq. (4.10) 8: Compute total desired comfort (µdesired ) by using Eq. (5.1) 9: While stopping criteria is not met do 10: Selection Evaluate the population using f itness function by using Eq. (4.5) and sort solutions using convergence 11: Selects only those solutions that satisfy the budget and energy constraints by using Eq. (4.7) and Eq. (4.11) 12: Reproduction : 13: if Crossoverrate > rand(1) then Parent vectors are selected as: P arent1 and P arent2 14: Reproduce the offspring (newchild) by applying a crossover operation 15: of f spring1 = [parent1(1 : COP )P arent2(COP + 1 : end)] 16: if M utationrate > rand(1) then 17: Select an individual from the population as P arent3 and assigned it to variable T emp 18: Randomly invert two bits of selected individual 19: if mutant row == 0 then 20: mutant row = randi(length(P arent3), 1, 1) 21: Status bit = P arent3(mutant row) 22: if Status bit == 1 then 23: T emp(mutant row) = 0 24: else 25: if Status bit == 0 then 26: T emp(mutant row) = 1 27: of f spring2=T emp 28: Evaluation and replacement 29: Evaluate the new population using f itness and sort solutions using convergence 30: Perform elitism to save the best solutions and discard the worst 31: end While 32: Return output 33: Return best solution Gbest 34: Compute total achieved comfort (µachieved ) based on scheduling pattern Gbest 35: Compute percentage comfort by using Eq. (5.2) 84


Chapter 4


the consumer’s comfort µ. Therefore, the objective function of EACA can be re-written as: Obj(χ) = min(χ).

4.1.2.2

(4.6)

Consumer’s budget and energy constraints

The EACA is subject to the budget constraint that the total consumer’s expenditure for electricity is less than or equal to the specified budget, which is expressed in Eq. (4.7). Texp ≤ β,

(4.7)

where Texp represents the consumer’s electricity cost, which is the product of total energy consumed during a day by ET obtained from the utility. Mathematically, it can be expressed by using Eq. (4.8).

Texp = × ET,

(4.8)

where, represents the total energy consumption of all appliances, calculated for each solution of the population by EACA using Eq. (4.9).

=

Z X

(T OTn × T P Rn ),

(4.9)

n=A

where, TOT and TPR represent total operational time and total power rating of the n number of appliances, respectively. Consequently, we can also formulate TEA constraint, since both parameters, including β and ET are known, thus, TEA is calculated by using Eq. (4.10).

T EA =

β($) . ET ($/kW h)

(4.10)

Where, TEA depicts the maximum power usage limit available to the consumers for a day, which cannot be violated. Thus, the energy constraint imposed by the consumer is given as: 85


Chapter 4


≤ T EA.

4.1.3

(4.11)

Mapping the priority induced scheduling problem with genetic algorithm

The following steps are followed by EACA to produce an optimal scheduling pattern. (i) To map the priority values, two tables: TB priority Table 4.2 and DB priority Table 4.3 are given as an input to the EACA. The other inputs include energy consumption/cost Table 4.1, ET, consumer’s budget limit β, and other GA initializing parameters. (ii) From the two input tables, the EACA generates absolute comfort values as given in Table 4.4 using Eq. (4.3). (iii) Initialization: A random population of 2000 solutions consisting of matrix size [18×24] is generated, where 18 represents the total number of appliances and 24 shows the total time-slots/ hours in a day. (iv) Selection: In the selection phase, the EACA evaluates the entire population using two fitness functions: f itness and convergence. In the f itness function all the solutions are evaluated using an objective function as given in Eq. (4.5). After the evaluation, only those solutions that satisfy the budget and energy constraints as given in Eq. (4.7) and Eq. (4.11) are selected for the next generation. Those solutions that do not satisfy these constraints are discarded by assigning a fitness value of zero. While the convergence function sorts the solutions based on the fitness value calculated by the f itness function. A high fitness value depicts a prominent UC (µ). (v) Reproduction: In this phase, the next generation of the chromosomes is produced using a crossover and mutation operators. The new population generated in this phase differs from the previous generation, however, it may 86


Chapter 4


share plenty of similar characteristic of their parents. The crossover operator is considered as one of the most important operators for the generation of new chromosomes. The crossover is performed between the best solutions evaluated by the convergence function. At this stage, a new solution is formed by exchanging genes selected from the two parents. The main goal is to have descendants which are better in term of fitness and can be selected for the next process. Here, a single-point crossover method is chosen, in which two parents called parent1 and parent2 are selected randomly from the population. Now, a cut-off point (COP) is also randomly selected between 1 and the total number of appliances, i.e., 18. The crossover is performed that results in the production of a newchild. The newchild first part values ranging from [1 COP] are taken from the parent1 while the second part values [COP+1 up-to 18] are obtained from the parent2. The mutation operator is used to avoid the premature convergence. In the mutation stage, a parent is selected and the status of its two genes in the chromosome is randomly modified so that a new solution is formed. Thus, a new population set is generated by applying mutation and crossover steps. (vi) Evaluation and replacement: In this stage, the new population obtained from the crossover and mutation steps will be evaluated using the fitness functions. Based on the fitness values the new population will replace the old population. This procedure is also known as elitism. (vii) Termination: Once a termination criterion is met, the best solution based on the objective function is considered and its values are returned accordingly. Based on these steps, the EACA is demonstrated as algorithm 1.

87


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION

4.2

Appliance scheduling problem formulation and proposed solution

In this section, appliances scheduling problem is formulated using MKP. A system model is proposed which is followed by four optimization strategies. This section is mainly focused on the subproblem discussed in section 3.6.2.

4.2.1

Load categorization

Let An represents a set of three appliances consisting of CD Acd , DW Adw , and refrigerator Aref . Thus, An = {Acd , Adw , Aref }. Appliances are scheduled in 24 hours time horizon τ T ∀ T = {τ1 , τ2 , τ3 , · · ·, τ24 }. In the following subsections, we discuss the energy consumption and cost calculation for appliances.

4.2.1.1

Clothes dryer

The CD’s operation time coinciding peak hours is shown in Fig. 4.6 when considering the DA-RTP signal as given in 5.7. Let Pcd represents the power rating of the CD. The total energy consumption per day of CD (εcd ) can be computed by the following equation:

εcd =

T X

Pcd × χb (τ ).

(4.12)

τ =1

In the above equation, χb (τ ) = [0, 1] is a boolean integer variable which shows appliance status. χb (τ ) = 1 if the appliance is ON in time-slot τ and otherwise. D The CD per day total cost ζcd in the total time interval T is calculated through a

formula:

D ζcd =

T X

Pcd × χb (τ ) × ρep (τ ).

(4.13)

τ =1

88


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION Electric price is represented by ρep (τ ) which in this case is DA-RTP signal. SimiY M costs are given as: and yearly ζcd larly, CD’s monthly ζcd

M ζcd

=

M X

D ζcd (d),

∀ M = 30,

(4.14)

M (y), ζcd

∀ Y = 12.

(4.15)

d=1

Y ζcd

=

Y X y=1

Here, we minimize the hourly cost of the CD, which results in the overall cost H is given by the formula: reduction. The hourly cost of CD ζcd

H ζcd = Pcd × χb (τ ) × ρep (τ ),

4.2.1.2

∀ τ = {1, 2, ...., T }.

(4.16)

Dishwasher

The DW’s usage spikes after breakfast and dinner, coinciding peak hours Fig. 4.6. Let Pdw denotes the power rating of the DW. The total energy consumption per day of the DW (εdw ) can be calculated by the following formula:

εdw =

T X

Pdw × χb (τ ).

(4.17)

τ =1 D The DW per day total cost ζdw in T is calculated through a formula:

D ζdw

=

T X

Pdw × χb (τ ) × ρep (τ ).

(4.18)

τ =1

Similarly, monthly and yearly cost of DW’s is given in Eq. (4.19) and Eq. (4.20).

M ζdw =

M X

D ζdw (d),

∀ M = 30,

(4.19)

d=1

89


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION

Y ζdw

=

Y X

M (y), ζdw

∀ Y = 12.

(4.20)

y=1

The hourly cost of the DW is minimized, which results in the overall cost reduction. H is given in Eq. (4.21). The hourly cost of DW ζdw

H = Pdw × χb (τ ) × ρep (τ ), ζdw

4.2.1.3

∀ τ = {1, 2, ...., T }.

(4.21)

Refrigerator

The refrigerator has a different energy pattern when compared to DW and CD. Let Pref shows the power rating of the refrigerator. The total energy consumption per day of the refrigerator (εref ) is represented by the following equation:

εref =

T X

Pref × χb (τ ).

(4.22)

τ =1 D The refrigerator per day total cost ζref in T is calculated through a formula:

D ζref

=

T X

Pref × χb (τ ) × ρep (τ ).

(4.23)

τ =1 M Y Similarly, the refrigerator monthly ζref and yearly ζref costs are given as:

M ζref

=

M X

D ζref (d),

∀ M = 30,

(4.24)

M ζref (y),

∀ Y = 12.

(4.25)

d=1

Y ζref

=

Y X y=1

90


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION We minimize the hourly cost of the refrigerator, which results in an overall cost H is given by a formula: reduction. The hourly cost of refrigerator ζref

H ζref = Pref × χb (τ ) × ρep (τ ),

∀ τ = {1, 2, ...., T }.

(4.26)

The total energy consumed and total cost per day are given by equations 4.27 and 4.28:

4.2.2

εAn = εcd + εdw + εref ,

(4.27)

D D D ζAn = ζcd + ζdw + ζref .

(4.28)

Peak to average ratio

It is essential to maintain a balance between demand and supply by reducing PAR. The PAR is a ratio of peak domestic load to average domestic load, consumed by the user in 24 hour time-slots. The mathematical equation for PAR is given in Eq. (4.29). P AR =

4.2.3

1 T

max(ετn ) PT PAn τ =1

n

(ετn )

,

∀ T = 24.

(4.29)

Threshold

The threshold of an appliance is dependent on the time factor µp set by the consumer and energy consumption ε and is calculated by the equation given below [38]. r Z∗ =

ep p 2(ρep p − ρo )µ τ + ρep o . ε

(4.30)

ep Here, ρep p and ρo represents the maximum and minimum electricity price values at

time-slot τ when the appliance is using energy consumption ε. µp represents time factor of an appliance set by the consumer. Here, the term time factor represents the appliance’s priority value. We have implemented the above threshold equation 91


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION in proposed HEMC. We compared the threshold value with electricity price signals ρep in each time-slot and select those operational slots which have lower values as compared to the defined threshold.

4.2.4

Knapsack problem formulation

In this section, we formulate home appliance scheduling problem as an optimization problem and map it to the MKP. The SKP is a combinatorial problem which consists of one knapsack with a specific capacity limit. Many objects with different weights and values need to be assigned to knapsack such that the profit is maximized considering the capacity limit. On the other hand, MKP is a resource allocation problem which consists of m resources or knapsacks and a set of n objects. Like SKP, each object in MKP is associated with a certain weight and value. Each resource or knapsack j has a capacity limit of Γjmax , which depicts the maximum weight that can be supported. The objective of MKP is to find a combination of objects that can be packed within the knapsacks so that total profit or net value of objects in all knapsacks is maximized [101]. The assumptions of mapping the scheduling problem using the MKP are given below: • The objects in the MKP are considered as a number of appliances. • The weight of each object is considered its power rating of the appliance consumed in each time-slot. • The object value in a specific time-slot is specified by the cost of power consumption of appliance in that time-slot. • The capacity of a j number of knapsacks represents the threshold of power limit in each time-slot. Here, we assume this limit to be fixed and equal to the maximum non-scheduled load limit. • The binary variable χb ∈ [1, 0] shows ON or OFF status of the appliance. These assumptions provide consumer to participate in energy management schemes to reduce their electricity cost along with PAR. The knapsack capacity limits Γjmax 92


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION in the proposed scenario is equal to the maximum non-scheduled peak load which is used to mitigate the rebound peaks, decreases PAR, and also increases grid sustainability.

4.2.5

Objective function

The objective function considering cost minimization is given as:

min

T X An X

χbn (τ ) × ζnD (τ ).

(4.31a)

τ =1 n=1

Here χbn (τ ) is a boolean integer [0, 1], which shows the status of appliance n either ON or OFF at hour τ . ζnD represents the daily electricity cost of the appliance n in time-slot τ . This cost can be computed by an equation given below:

ζnD

=

T X An X

Pn × χbn (τ ) × ρep n (τ ),

(4.31b)

τ =1 n=1

subject to: T X An X

Pn × χbn (τ ) ≤ γcap ,

(4.31c)

τ =1 n=1 T X An X

χbnN sch. (τ )

=

τ =1 n=1

T X An X

χbnS ch. (τ ),

(4.31d)

τ =1 n=1

sch. εN = εSch. An An ,

(4.31e)

N sch. Sch. ζAn > ζAn ,

(4.31f)

M −N sch. M −Sch. ζAn > ζAn ,

(4.31g)

Y −N sch. Y −Sch. > ζAn , ζAn

(4.31h)

P AR ≤ Γjmax .

(4.31i)

93


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION Eq. (4.31a) shows the objective function to minimize the electricity cost. The daily electricity cost is represented in Eq. (4.31b). The Eqs. (4.31c-4.31i) show constraints of objective function. Eq. (4.31c) ensures the total load demand of appliances should be less than or equal to the grid capacity (γcap ). Constraint Eq. (4.31d) represents that LoT of all appliances before and after scheduling must be equal. The constraint in Eq. (4.31d) also guarantees that total energy consumed by appliances before and after scheduling must also be equal is given in Eq. (4.31e). Constraints Eq. (4.31f), Eq. (4.31g), and Eq. (4.31h) shows daily, monthly, and yearly costs of nonscheduled must be greater than scheduled costs respectively. Eq. (4.31i) states that the PAR value must be less than or equal to the knapsack capacity limit which is Γjmax . In the proposed scenario, Γjmax capacity is equal to a maximum non-scheduled peak load value.

4.2.6

Proposed solution

We apply four optimization schemes EDE, GA, BPSO, and OSR to solve the MKP. The meta-heuristic schemes EDE, GA, and BPSO are similar in nature because of their population generation based search methods. Deterministic and probabilistic rules are applied to improve the new generation of populations during each iteration. In the following subsections, we shortly discuss the proposed schemes.

4.2.6.1

Enhanced differential evolution

The basic information relevant to EDE is given in section 2.2.1.2. The modification in EDE is done at the stage of generating trial vectors CR. Five groups of trial vectors are generated in each iteration. The first three trial vectors are obtained by taking three distinct CR values, i.e., 0.3, 0.6, and 0.9. The fourth trial vector aims to speed up the convergence rate while the last trial vector increases the population diversity. The equations for generating five groups of trial vectors are

94


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION given in [108]:    vi ui =   x    vi ui =   x    vi ui =  x 

if (rand(d) ≤ 0.3, i

(4.32)

if (rand(d) > 0.3 if (rand(d) ≤ 0.6,

i

(4.33)

if (rand(d) > 0.6 if (rand(d) ≤ 0.9,

i

(4.34)

if (rand(d) > 0.9

ui = rand(d). xi ,

(4.35)

ui = rand(d).vi + (1 − rand(j)).xi .

(4.36)

Here, d = 1, ui represents the five trial vectors, vi is mutant vector and xi is the target vector. A fitness function is used to evaluate generated trial vectors. The trial vector with minimum fitness function value is considered.

4.2.6.2

Genetic algorithm

GA is a meta-heuristic based optimization algorithm which is inspired by the biogenetic process of living organisms. Here new genes are formed which carry properties and characteristics of their parents. In the GA process, a random population of chromosomes is initially generated. Each chromosome represents a candidate solution to the problem. Here, a binary GA is implemented in which each bit of a chromosome is associated with on/off status of the appliance. The number of bits in the chromosome is equal to different energy cycles of the appliance. A fitness function based on the objective function evaluates each chromosome of the initial population. The best value based on the fitness function is chosen as a current best value in that iteration. Based on these current best solutions, a new population is generated by using the crossover and mutation steps.

95


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION In the GA process, crossover and mutation are two important algorithmic-specific control parameters. In the crossover step, a COP is randomly selected and two parents bits are crossed with each other to form a new child. The probability of crossover rate value specifies the convergence rate. A high value of the crossover will result in faster convergence at the cost of accuracy. The best crossover rate specified in the literature is given as:

pc = 0.9.

(4.37)

To create randomness in population, so that it may not halt in the local minima, mutation function is applied. In mutation, one or more genes in a chromosome are changed. The probability of mutation is very low and is calculated by:

pm = 1 − pc .

(4.38)

Finally, new population based on crossover and mutation is generated. This population is again evaluated by fitness function and compared with the previous population to find the global best optimal solution.

4.2.6.3

Binary particle swarm optimization

In BPSO, the initial population is randomly generated in the form of position matrix. Each bit position in the matrix represents the state of the appliance. Each row in the position matrix depicts a candidate solution to the optimization problem. The initial velocity of individuals is generated by a given formula:

vi = vmax × 2 × (rand(swarm, n) − 0.5).

96

(4.39)


4.2. APPLIANCE SCHEDULING PROBLEM FORMULATION AND Chapter 4 PROPOSED SOLUTION The particles move freely in the search space and are evaluated through fitness function and updated via velocity equation given as [215].

vi (t) = w.vi (t − 1) + φ1 .rand(1). pbest − xi (t − 1) +φ2 .rand(1). gbest − xi (t − 1) .

(4.40)

In the above equation, φ1 and φ2 are acceleration constants which controls the movement of particle towards pbest and gbest positions. In our case, both acceleration constants have equal value of 2. w is a weighted factor and is calculated by using a formula. w = wi +

wf − (wi × k) . nitra

(4.41)

Where wi and wf have values 2 and 0.4, respectively. nitra represents the total number of generations while k is the loop index value. Since the velocity values of the particles are real numbers which are mapped between 0 and 1 by using a function called sigmoid. The sigmoid function is calculated by the formula:

sig(i, j) =

1 1 + exp − vi (t)

.

(4.42)

The position matrix is updated by comparing sigmoid function values with random function values using the formula:    1 sig(i,j) v co ,   P OW wt (t) = x.(v(t)3 − y.Prwt ) v ci < v(t) < v r ,       P OW wt (t) = P wt v r < v(t) < v co , r

(4.112)

where v shows the speed of the wind, Prwt denotes the nominal power of WT, v r , v co , and v ci represent rated, cut-out, and cut-in wind speed, respectively. The parameters x and y can be obtained by Eq. 4.113:    x = P wt /((v r )3 − (v ci )3 ), r   y = (v ci )3 /((v r )3 − (v ci )3 ).

134

(4.113)


Chapter 4

4.5. STAND-ALONE HRES CONFIGURATION AND SIZING FORMULATION FOR PV-WT-BATTERY SYSTEM

If there are N wt number of WTs installed in an area, then the overall produced wind power is obtained by the following formula:

ξ wt (t) = N wt × P OW wt (t).

4.5.2.3

(4.114)

Accumulative power generation by RESs and consumer’s load formulation

The PV and WT accumulative electricity generation can be expressed as:

ξ gen (t) = ξ pv (t) × ηi + ξ wt (t) × ηi2 ,

(4.115)

where ηi shows the efficiency of the inverter. In homes, the consumer’s load ξ ld at time-slot t depends upon the appliances usage. Thus, the ξ ld can be computed by Eq. (4.116):

ld

ξ (t) =

z X

pi (t) × χ(t),

(4.116)

i=a

where i and p denote the number of appliances and their power ratings, respectively. χ(t) represents a boolean integer showing an appliance status. When χ(τ ) = 1, the appliance status is ON in hour t and otherwise, it is considered OFF.

4.5.2.4

Sizing formulation of battery bank

The energy storage capacity of the battery bank is subject to constant changes due to the intermittent nature of PV-WT-Battery based system. When ξ gen (t) is greater than ξ ld (t), the battery bank is in state of charge (SOC). Thus, the amount

135


Chapter 4


of charging quantity of battery bank at time t is obtained by Eq. 4.117 [130]. ξ StoreB (t) = ξ StoreB (t − 1) × (1 − ι) " # ld ξ (t) + ξ gen (t) − × η b , ∀ ξ gen (t) > ξ ld (t), ηi

(4.117)

where ξ StoreB (t) and ξ StoreB (t−1) show the stored amount of energy in the battery bank at time-slots t and t − 1, respectively. ι represents self discharging state, ξ ld represents the load demand, ηi depicts the efficiency of the inverter, and η b denotes battery bank charging efficiency. In case, when ξ gen (t) is less than ξ ld (t) at time-slot t, then stored energy in the battery bank is utilized to meet the load demand. Now, the battery bank state is changed to discharging. The battery bank discharging efficiency is assumed to be 1, and we also ignore the temperature effects in this part of the thesis. Thus, the battery bank charging quantity at time-slot t is given by the following formula: ξ StoreB (t) = ξ StoreB (t − 1) × (1 − ι) # " ξ ld (t) − ξ gen (t) /ηi , ∀ ξ gen (t) < ξ ld (t). − ηi

4.5.3

(4.118)

Calculation of batteries for battery bank

An important decision variable in the PV-WT-Battery HRES is the calculation of the total number of batteries (N b ) required for the battery bank. The N b depends upon the consumer’s load requirement and the RESs generation capacity. In order to find N b , we suppose a temporary storage variable temp, which is initially initialized as 0. In case, when RESs power generation is higher than the consumer’s load demand at an instant time t, the variable temp stores energy as per Eq. (4.117). In another case, when power generation produced by RESs is smaller than the consumer’s load demand at time-slot t, the temp variable is updated using Eq. (4.118). Thus, finding the total number of batteries for a system is dependent 136


Chapter 4


upon the curve of variable temp. The positive temp values indicate the generation availability by RESs while the negative values show generation deficiency in the respective time-slots. The total required storage capacity (T rsc ) is the difference between the maximum point and the minimum point in the temp curve, which can be obtained by the following formula:

T rsc = max(temp) − min(temp).

(4.119)

Here, max(temp) and min(temp) represent the maximum and minimum generation points on the temp curve, respectively. Thus, the calculation for the N b required for a given system can be derived by using the formula [164]: T rsc N = , 1.35 b

(4.120)

where N b shows the number of batteries needed for the battery bank and 1.35 shows the nominal capacity of a battery.

4.5.4

System reliability

Reliability is an essential factor that needs to be considered in the SA system. Therefore, in this paper, the concept of LP SP is regarded and implemented to have a reliable HRES. LP SP is elucidated by a number in the range of 0 and 1. When LP SP = 0, it shows that the system is very reliable and the consumer’s load will always be fulfilled. LP SP = 1 states that the consumer’s load is never fulfilled. The LP SP for one year (T = 8760 h) can be expressed as:

LP SP

P8760 t=1 LOP S(t) , = P 8760 ld t=1 ξ (t)

(4.121)

where LOPS depicts loss of power supply. The LOPS occurs when the total energy generated ξ gen by HRES is less as compared to the total consumer’s load demand

137


Chapter 4


energy ξ ld at any time-slot. LOPS is defined in Eq. 4.122.

LOP S(t) = ξ ld (t) − ξ gen (t),

∀ t∈T

(4.122)

The flowchart of calculating the system reliability is presented in Fig. 4.15. The flowchart is presented for a population size (X = 50). Start

X = 1:50

Take input data for Xi Npv, Nwt, Nb, Load profile (Eld), LPSPmax, T = 8760

t=1

Calculate Energy generation Egen(t), Energy stored in batteries, Excess energy, Deficit energy No t=t+1

If t = 8760 Yes Calculate LPSP

No If LPSP