on optimizing energy consumption with combined

0 downloads 0 Views 4MB Size Report
Khalid Usmani, Dr. Munam Ali Shah, Dr. Yasir Hafeez, Tariq Ali; all of whom I .... fuels and high demand of electricity lead to a dependency on RES (Mhanna,.
ON OPTIMIZING ENERGY CONSUMPTION WITH COMBINED OPERATIONS OF MICROGRIDS FOR DEMAND SIDE MANAGEMENT IN SMART HOMES

ZAFAR IQBAL 12-arid-2415

UNIVERSITY INSTITUTE OF INFORMATION TECHNOLOGY PIR MEHR ALI SHAH ARID AGRICULTURE UNIVERSITY RAWALPINDI PAKISTAN 2018

ON OPTIMIZING ENERGY CONSUMPTION WITH COMBINED OPERATIONS OF MICROGRIDS FOR DEMAND SIDE MANAGEMENT IN SMART HOMES

by

ZAFAR IQBAL 12-arid-2415

A thesis submitted in partial fulfillment of the requirement for degree of

Doctor of Philosophy in Computer Science

UNIVERSITY INSTITUTE OF INFORMATION TECHNOLOGY PIR MEHR ALI SHAH ARID AGRICULTURE UNIVERSITY RAWALPINDI PAKISTAN 2018 i

CERTIFICATION I hereby undertake that this research is an original one and no part of this thesis falls under plagiarism, If found otherwise at any stage, I will be responsible for the consequences. Name: ZAFAR IQBAL

Signature: ______________

Registration Number: 12-arid-2415

Date:

______________

Certified that the contents and form of thesis entitled on optimizing energy consumption with combined operations of microgrids for demand side management in smart homes submitted by ZAFAR IQBAL has been found satisfactory for requirements of the degree.

Supervisor:___________________ Dr. Saleem Iqbal

Co-Supervisor: __________________ Dr. Nadeem Javaid

Member: _____________________ Dr. Saud Altaf

Member: _______________________ Dr. M. Azeem Abbas [DD-MM-YEAR]

Date of Viva Voce: _______________

External Examiner: _______________

(Name)

Director: _________________________

Director Advanced Studies:

___________________________ ii

DEDICATION This humble work is dedicated to my Family, Friends, the ComSens (Research Lab) students and above all, To the Almighty Allah (S.W.T) and Prophet Muhammad (P.B.U.H).

iii

Publications Journal 1. Iqbal, Z., Javaid, N., Iqbal, S., Aslam, S., Khan, Z. A., Abdul, W., and Alamri, A. (2018). A Domestic Microgrid with Optimized Home Energy Management System. Energies, 11(4), 1002. (Impact Factor: 2.676) 2.

Iqbal, Z, Nadeem Javaid, Syed Muhammad Mohsin, Syed Muhammad Abrar Akber, Muhammad Khalil Afzal and Farruh Ishmanov. (2018). Performance Analysis of Hybridization of Heuristic Techniques for Residential Load Scheduling. Energies. (Impact Factor: 2.676)

3. Aslam, S., Iqbal, Z., Javaid, N., Khan, Z. A., Aurangzeb, K., and Haider, S. I. (2017). Towards efficient energy management of smart buildings exploiting heuristic optimization with real time and critical peak pricing schemes. Energies, 10(12), 2065. (Impact Factor: 2.676) 4. Naz, M., Iqbal, Z., Javaid, N., Khan, Z. A., Abdul, W., Almogren, A., and Alamri, A. (2018). Efficient Power Scheduling in Smart Homes Using Hybrid Grey Wolf Differential Evolution Optimization Technique with Real Time and Critical Peak Pricing Schemes. Energies, 11(2), 384. (Impact Factor: 2.676) 5. Rasheed, M. B., Javaid, N., Awais, M., Khan, Z. A., Qasim, U., Alrajeh, N., Zafar Iqbal and Javaid, Q. (2016). Real time information based energy management using customer preferences and dynamic pricing in smart homes. Energies, 9(7), 542. (Impact Factor: 2.676) 6. Javaid, N., Ullah, I., Akbar, M., Iqbal, Z., Khan, F. A., Alrajeh, N., and Alabed, M. S. (2017). An intelligent load management system with renewable energy integration for smart homes. IEEE Access, 5, 13587-13600. (Impact Factor: 3.775) 7. Javaid, N., Naseem, M., Rasheed, M. B., Mahmood, D., Khan, S. A., Alrajeh, N., and Iqbal, Z. (2017). A new heuristically optimized Home Energy Management controller for smart grid. Sustainable Cities and Society, 34, 211227. (Impact Factor: 3.073)

iv

Conference Proceedings 1. Iqbal, Z., Javaid, N., Khan, M. R., Khan, F. A., Khan, Z. A., and Qasim, U. (2016, March). A Smart Home Energy Management Strategy Based on Demand Side Management. In Advanced Information Networking and Applications (AINA), 2016 IEEE 30th International Conference on (pp. 858- 862). IEEE. 2. Iqbal, Z., Javaid, N., Khan, M. R., Ahmed, I., Khan, Z. A., and Qasim, U. (2016, March). Cost and load reduction using heuristic algorithms in smart grid. In Advanced Information Networking and Applications Workshops (WAINA), 2016 30th International Conference on (pp. 24-30). IEEE. 3. Nazir, S., Shaffiq, S., Iqbal, Z., Zeeshan, M., Tariq, S., and Javaid, N. (2018, September). Cuckoo Optimization Algorithm Based Job Scheduling Using Cloud and Fog Computing in Smart Grid. In International Conference on Intelligent Networking and Collaborative Systems (pp. 34-46). Springer, Cham. 4. Ullah, R., Javaid, N., Iqbal, Z., Ahmad, I., Jan, A., and Jadoon, Y. K. (2018, July). CRRP Analysis of Cloud Computing in Smart Grid. In Conference on Complex, Intelligent, and Software Intensive Systems (pp. 64-74). Springer, Cham. 5. Hafeez, G., Javaid, N., Ullah, S., Iqbal, Z., Khan, M., and Rehman, A. U. (2018, July). Short Term Load Forecasting based on Deep Learning for Smart Grid Applications. In International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing (pp. 276-288). Springer, Cham. 6. Ghulam Hafeez, 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. International Conference on P2P, Parallel, Grid, Cloud and Internet Computing November 2018, DOI: 10.1007/978-3-319-69835-9-13

v

7. Hussain, S. M., Rahim, M. H., Nadeem, Z., Fatima, I., Iqbal, Z., Asif, S., and Javaid, N. (2017, November). The trends of integrating renewable energy sources: a survey. In International Conference on Broadband and Wireless Computing, Communication and Applications (pp. 627-636). Springer, Cham. 8. Jamil, A., Javaid, N., Iqbal, Z., Abdullah, M., Riaz, M. Z., and Akbar, M. (2018, July). Hierarchical Based Coordination Strategy to Efficiently Exchange the Power Among Micro-grids. In International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing (pp. 242-251). Springer, Cham. 9. Aslam, S., Javaid, N., Iqbal, Z., Asif, M., Iqbal, U., and Sarwar, M. A. (2018). A mixed integer linear programming based optimal home energy management scheme considering grid-connected microgrids. In 2018 14th International Wireless Communications & Mobile Computing Conference (IWCMC)(pp. 993998). IEEE. 10. Saboor, A., Javaid, N., Iqbal, Z., Abbas, Z., Khan, A. J., Rashid, S., and Awais, M. (2018, May). Home Energy Management in Smart Grid Using Evolutionary Algorithms. In 2018 IEEE 32nd International Conference on Advanced Information Networking and Applications (AINA) (pp. 1070-1080). IEEE. 11. Bukhsh, R., Javaid, N., Iqbal, Z., Ahmed, U., Ahmad, Z., and Iqbal, M. N. (2018, May). Appliances Scheduling Using Hybrid Scheme of Genetic Algorithm and Elephant Herd Optimization for Residential Demand Response. In 2018 32nd International

Conference

on

Advanced

Information

Networking

and

Applications Workshops (WAINA). IEEE. 12. Bukhsh, R., Javaid, N., Iqbal, Z., Ahmed, U., Ahmad, Z., and Iqbal, M. N. (2018, May). Appliances Scheduling Using Hybrid Scheme of Genetic Algorithm and Elephant Herd Optimization for Residential Demand Response. In 2018 32nd International

Conference

on

Advanced

Information

Networking

and

Applications Workshops (WAINA). IEEE. 13. Khan, A. J., Javaid, N., Iqbal, Z., Anwar, N., Saboor, A., and Qasim, U. (2018, May). A Hybrid Bacterial Foraging Tabu Search Heuristic Optimization for Demand Side Management in Smart Grid. In 2018 32nd International

vi

Conference on Advanced Information Networking and Applications Workshops (WAINA). IEEE. 14. Bukhsh, R., Iqbal, Z., Javaid, N., Ahmed, U., Khan, A., and Khan, Z. A. (2018, March). Appliances Scheduling Using State-of-the-Art Algorithms for Residential Demand Response. In International Conference on Emerging Internetworking, Data and Web Technologies (pp. 292-302). Springer, Cham. 15. Rahim, S., Iqbal, Z., Shaheen, N., Khan, Z. A., Qasim, U., Khan, S. A., and Javaid, N. (2016, March). Ant colony optimization based energy management controller for smart grid. In Advanced Information Networking and Applications (AINA), 2016 IEEE 30th International Conference on (pp. 1154-1159). IEEE. 16. Rahim, S., Khan, S. A., Javaid, N., Shaheen, N., Iqbal, Z., and Rehman, G. (2015, September). Towards multiple knapsack problem approach for home energy management in smart grid. In Network-Based Information Systems (NBiS), 2015 18th International Conference on (pp. 48-52). IEEE. 17. Shaheen, N., Javaid, N., Iqbal, Z., Muhammad, K., Azad, K., and Chaudhry, F. A. (2015, September). A hybrid algorithm for energy management in smart grid. In Network-Based Information Systems (NBiS), 2015 18th International Conference on (pp. 58-63). IEEE. 18. Rasheed, M. B., Awais, M., Javaid, N., Iqbal, Z., Khurshid, A., Chaudhry, F. A., and Ilahi, F. (2015, September). An energy efficient residential load management system for multi-class appliances in smart homes. In Network- Based Information Systems (NBiS), 2015 18th International Conference on (pp. 53-57). IEEE. 19. Awais, M., Javaid, N., Shaheen, N., Iqbal, Z., Rehman, G., Muhammad, K., and Ahmad, I. (2015, September). An efficient genetic algorithm based demand side management scheme for smart grid. In 2015 18th International Conference on Network-Based Information Systems (pp. 351-356). IEEE. 20. Fahim, H., Javaid, N., Khan, Z. A., Qasim, U., Javed, S., Hayat, A., Zafar Iqbal and Rehman, G. (2015, July). Bio-inspired Routing in Wireless Sensor Networks. In Innovative Mobile and Internet Services in Ubiquitous Computing (IMIS), 2015 9th International Conference on (pp. 71-77). IEEE. vii

21. Fahim, H., Javaid, N., Khan, Z. A., Qasim, U., Javed, S., Mahmood, D., and Iqbal, Z. (2015, July). Multilevel Routing Protocol for Energy Optimization in Wireless Sensor Networks. In 2015 9th International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing (IMIS) (pp. 86-93). IEEE. 22. Fahim, H., Javaid, N., Qasim, U., Khan, Z. A., Javed, S., Hayat, A., Zafar Iqbal and Rehman, G. (2015, July). Interference and bandwidth aware depth based routing protocols in underwater WSNs. In Innovative Mobile and Internet Services in Ubiquitous Computing (IMIS), 2015 9th International Conference on (pp. 78-85). IEEE. 23. Ilyas, N., Javaid, N., Iqbal, Z., Imran, M., Khan, Z. A., Qasim, U., and Shoaib, M. (2015, March). AAEERP: Advanced AUV-aided energy efficient routing protocol for underwater WSNs. In Advanced Information Networking and Applications (AINA), 2015 IEEE 29th International Conference on (pp. 77-83). IEEE. 24. Ilyas, N., Javaid, N., Iqbal, Z., Imran, M., Khan, Z. A., Qasim, U., and Shoaib, M. (2015, March). Extended Lifetime Based Elliptical Sink-Mobility in Depth Based Routing Protocol for UWSNs. In Advanced Information Networking and Applications Workshops (WAINA), 2015 IEEE 29th International Conference on (pp. 297-303). IEEE. 25. Hasnat, M. A., Akbar, M., Iqbal, Z., Khan, Z. A., Qasim, U., and Javaid, N. (2015, February). Bio inspired distributed energy efficient clustering for Wireless Sensor Networks. In Information Technology: Towards New Smart World (NSITNSW), 2015 5th National Symposium on (pp. 1-7). IEEE. 26. Umar, A., Akbar, M., Iqbal, Z., Khan, Z. A., Qasim, U., and Javaid, N. (2015, February). Cooperative partner nodes selection criteria for cooperative routing in underwater WSNs. In Information Technology: Towards New Smart World (NSITNSW), 2015 5th National Symposium on (pp. 1-7). IEEE.

viii

Conference proceedings published as book chapters 1. Bukhsh, R., Iqbal, Z., Javaid, N., Ahmed, U., Khan, A., and Khan, Z. A. (2018, March). Appliances Scheduling Using State-of-the-Art Algorithms for Residential Demand Response. In International Conference on Emerging Internetworking, Data and Web Technologies (pp. 292-302). Springer, Cham. 2. Jamil, A., Javaid, N., Iqbal, Z., Abdullah, M., Riaz, M. Z., and Akbar, M. (2018, July). Hierarchical Based Coordination Strategy to Efficiently Exchange the Power Among Micro-grids. In International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing (pp. 242251). Springer, Cham. 3. Hafeez, G., Javaid, N., Ullah, S., Iqbal, Z., Khan, M., and Rehman, A. U. (2018, July). Short Term Load Forecasting based on Deep Learning for Smart Grid Applications. In International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing (pp. 276-288). Springer, Cham. 4. Ullah, R., Javaid, N., Iqbal, Z., Ahmad, I., Jan, A., and Jadoon, Y. K. (2018, July). CRRP Analysis of Cloud Computing in Smart Grid. In Conference on Complex, Intelligent, and Software Intensive Systems (pp. 64-74). Springer, Cham. 5. Nazir, S., Shaffiq, S., Iqbal, Z., Zeeshan, M., Tariq, S., and Javaid, N. (2018, September). Cuckoo Optimization Algorithm Based Job Scheduling Using Cloud and Fog Computing in Smart Grid. In International Conference on Intelligent Networking and Collaborative Systems (pp. 34-46). Springer, Cham.

ix

CONTENTS Page List of Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x List of Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxvii Chapter 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 INTRODUCTION . . . .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . 1 1.1.1 Why Smart Grid? . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.2 SG and its Role in Traditional Grid . . . . . . . . . . . . . . . . . .. . . . . . . . . . 3 1.1.3 Role of RESs and MGs. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .. . . . .5 1.1.4 Optimization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.1.5 Energy Consumption and its Scheduling Effects on Consumers . . . . . . 7 1.2 PROBLEM STATEMENT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 9 1.2.1 Sub-problem 1 . . . . . . . . . . . . . . . .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . 9 1.2.2 Sub-problem 2 . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.2.3 Sub-problem 3 . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . .. . . . . . . 11 1.2.4 Sub-problem 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.3 RESEARCH QUESTIONS OF THE THESIS . . . . . . . . . . . . . . . . . . . . . 12 1.4 CONTRIBUTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.4.1 Contribution 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.4.2 Contribution 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.4.3 Contribution 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.4.4 Contribution 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . 15 Chapter 2 RELATED WORK . . . . . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . . 16 2.1 INTRODUCTION . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 16 2.2 OPTIMIZATION TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2.2.1 Heuristic Techniques . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . 17 2.2.2 Hybrid Heuristic Techniques . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . 18 2.2.3 Conventional Techniques . . . . . . . . . . . .. . . . . . . . .. . . . . . . . . . . . . . . . 18 x

2.3 DR BASED TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.4 PRICING SCHEMES BASED TECHNIQUES . . . . . . . .. . . . . . . . . . . . 22 2.4.1 Combined Pricing Model Based Techniques . . . . . . . . . . . . . .. . . . . . 22 2.5 RES INTEGRATION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.6 MG BASED TECHNIQUES . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . 25 2.7 LOAD BALANCING AND PEAK LOAD SHAVING BASED TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . .28 2.8 APPLIANCES SCHEDULING BASED TECHNIQUES.. . . .. . . . . . . . . .29 2.9 USER PREFERENCES BASED TECHNIQUES . . . . . . . .. . . . . . . . . . . .30 2.10 ELECTRICITY PRICE REDUCTION TECHNIQUES . . . . . . . . . . . . .31 2.11 PAR AND UC ORIENTED TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . 32 2.12 HEMS BASED TECHNIQUES . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . .33 2.13 ESS BASED TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.14 MKP FOR PROBLEM FORMULATION . . . . . . . . . .. . . . . . . . . . . . . .37 2.15 ENERGY SHARING AND TRADING . . . . . . . . . . . . . . . . . . . . . . . . .38 Chapter 3 PROPOSED SYSTEM MODELS AND SOLUTIONS . . . . . . . 54 3.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 3.2 ENERGY MANAGEMENT OF SMART BUILDINGS EXPLOITING HEURISTIC OPTIMIZATION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .54 3.2.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .54 3.2.2 Classification of Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 3.2.2.1 Deferrable appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.2.2.2 Non-deferrable appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 3.2.2.3 Base load appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 3.2.3 Electricity Cost . . . . . . . . . .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . .61 3.2.4 Electricity Storage System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .62 3.2.5 Proposed Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63 3.2.6 Genetic Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . 64 3.2.7 Cuckoo Search Optimization Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.2.8 Crow Search Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . .66 3.3 FEASIBLE REGION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .67 xi

3.3.1 Feasible Region for Electricity Cost and Electricity Consumption using RTP Signals . . . . . . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . .68 3.3.2 Feasible Region for Electricity Cost and Electricity Consumption using CPP Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .70 3.3.3 Feasible Region for Electricity Cost and User WT using RTP Signals .72 3.3.4 Feasible Region for Electricity Cost and User WT using CPP Signals . . 74 3.4 POWER SCHEDULING IN SMART HOMES USING HGWDE OPTIMIZATION TECHNIQUE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 3.4.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .76 3.4.2 Classification of Appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 3.4.2.1 Shiftable appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 3.4.2.2 Controllable appliances . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . 78 3.4.2.3 Non-shiftable appliances . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 79 3.4.3 Proposed Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .79 3.4.4 EDE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . .80 3.4.5 GWO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .82 3.4.5.1 Encircling prey . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 3.4.5.2 Hunting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 3.4.6 HGWDE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 3.5 A DOMESTIC MG WITH OPTIMIZED HOME ENERGY MG SYSTEM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 90 3.5.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .90 3.5.2 Formulation of the Problem Statement . . . . . . . . . . . . . . . . . . .. . . . . . . .90 3.5.3 PV Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . .90 3.5.4 Wind Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . .91 3.5.5 Battery Bank System (BBS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .92 3.5.6 Energy Consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93 3.5.7 Energy Pricing and Electricity Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . .94 3.5.8 PAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . 95 3.5.9 AWT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 3.5.10 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .96 xii

3.5.11 System Model . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 3.5.12 Optimization Techniques . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 3.5.13 GWO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 3.5.14 GA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 3.5.15 BPSO. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 3.5.16 WDO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 3.5.17 WDGA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 3.5.18 WDGWO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 3.5.19 WBPSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 3.6 ANALYSIS OF HYBRIDIZATION OF HEURISTIC TECHNIQUES FOR RESIDENTIAL LOAD SCHEDULING.. . . . . . . . . . . . . . . . . . . . . 114 3.6.1 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 3.6.2 Mapping of Load Scheduling to MKP. . . . . . . . . . . . . . . . . . . . . . . . . . 114 3.6.3 Mathematical Modeling of Objective Function and Constraints. . . . . . 115 3.6.3.1 Energy consumption model . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . 115 3.6.3.2 Energy cost model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 3.6.3.3 PAR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 3.6.3.4 Waiting time. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . .. . . . . . . . . . . . 117 3.6.4 Optimization Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 3.6.5 Proposed System Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 3.6.6 Optimization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 3.6.7 Existing Optimization Techniques . . . . . . . .. . . . . . . . . . . . . . . .. . . . . . 123 3.6.7.1 GA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . 123 3.6.7.2 TLBO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 3.6.7.3 BAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 3.6.7.4 FPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 126 3.6.8 Proposed Optimization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . .. . . 126 3.6.8.1 GTLBO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 3.6.8.2 FBAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 3.6.8.3 FTLBO . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 3.6.8.4 FGA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 xiii

Chapter 4 SIMULATIONS AND RESULTS . . . . . . . . . . . . . . . . . . . . . . . 135 4.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 4.2 ENERGY MANAGEMENT OF SMART BUILDINGS EXPLOITING HEURISTIC OPTIMIZATION. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . 135 4.2.1 Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 4.3 POWER SCHEDULING IN SMART HOMES USING HGWDE OPTIMIZATION TECHNIQUE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 4.3.1 Simulations and Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 4.3.2 Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 4.3.2.1 Cost using RTP. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . 152 4.3.2.2 Cost using CPP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 4.3.3 Energy Consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 4.3.3.1 Load using RTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 4.3.3.2 Load using CPP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 4.3.4 PAR . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . 162 4.3.4.1 PAR using RTP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 4.3.4.2 PAR using CPP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 4.3.5 WT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 4.3.5.1 WT using RTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 4.3.5.2 WT using CPP. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . 166 4.3.6 Convergence of the Fitness Function . . . . . . . . .. . . . . . . . . . . . .

. . . . 166

4.3.7 Feasible Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 4.3.7.1 Feasible region using RTP

. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . 167

4.3.7.2 Feasible region using CPP

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

4.3.8 Performance Trade-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 4.4 A DOMESTIC MG WITH OPTIMIZED HOME ENERGY MG SYSTEM. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 4.4.1 Simulation Results and Discussion. . . . . . . . . . . . . . . . . . .. . . . . . . . . . 175 4.4.2 Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 4.4.3 RTP Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 4.4.4 Energy Consumption Profile .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 xiv

4.4.4.1 Energy consumption with and without RESs . . . . . . . . . . . . . . . . . . 178 4.4.5 Electricity Cost .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 4.4.5.1 Electricity cost profile with and without RESs. . . . . . . . . . . . . . . . . . 186 4.4.6 PAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 4.4.6.1 PAR with and without RESs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 4.4.7 AWT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 4.4.8 Energy Generation Profile of MG . . . . . . . . . . . . . . . . . . . . . .. . . . . . . 190 4.4.8.1 Energy generation with wind turbine and solar panel

. . . . . . . . . . . 190

4.5 ANALYSIS OF HYBRIDIZATION OF HEURISTIC TECHNIQUES FOR RESIDENTIAL LOAD SCHEDULING. . . . . . . . .. .. . . . . . . . . . . . 193 4.5.1 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . 193 4.5.2 Power Consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 4.5.3 Electricity Consumption Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 4.5.4 User Discomfort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 4.5.5 PAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 4.5.6 Feasible Region for Electricity Cost and User Discomfort. . . . . . . . . . 206 4.5.7 The Performance Parameters Trade-off . . . . . . . . . . . . . . . . . . . . . . . . 207 Chapter 5 CONCLUSION AND FUTURE WORK . . . . . . . . . . . . . . . . . 210

xv

List of Tables 2.1 Summarized literature review . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 40 3.1 Parameters of appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 3.2 GA parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.3 CSOA parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 3.4 CSA parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 3.5 Parameters of appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 3.6 EDE mapping on HEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 3.7 Hybrid gray wolf differential evolution (HGWDE) mapping on HEM . . 88 3.8 Power rating of system model components . . . . . . . . . . . . . . . . . . . . . . . .99 3.9 Appliances classification . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . 99 3.10 Parameters of GWO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 3.11 Parameters of GA. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 3.12 Parameters of BPSO. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 3.13 Parameters of WDO . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 3.14 Parameters of WDGA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .105 3.15 Parameters of WDGWO. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . 106 3.16 WBPSO parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.17 GA parameters . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 4.1 Summary of the results for a single smart home . . . . . . . . . . . . . . . . . . . . 145 4.2 Summary of the results for thirty smart homes . . . . . . . . . . . . . . . . . . . . . 150 4.3 Cost comparison of different techniques for 24 h . . . . . . . . . . . . . . . . . . 157 4.4 PAR comparison of different techniques for 24 h . . . . . . . . .. . . . . . . . . . 164 4.5 WT comparison of different techniques for 24 h. . . . . . . . . . . . . . . . . . . 166 4.6 Possible cases: OTI 15-min using RTP. . . . . . . . . . . . .. . . . . . . . . . . . . . . 169 4.7 Possible cases: OTI 30-min using RTP. . . . . . . . . . . . . . . . . . . . . . . . . . . 170 4.8 Possible cases: OTI 60-min using RTP. . . . . . . . . .. . . . . . . . . . . . . . . . . . 171 4.9 Possible cases: OTI 15-min using CPP. . . . . . . . . . . . . . . . . . . . . . . . . . . 171 4.10 Possible cases: OTI 30-min using CPP . . . . . . . . . . . . . . . . . . . . . . . . . . 174 xvi

4.11 Possible cases: OTI 60-min using CPP. . . . . . . . . . . . . . . . . . . . . . . . . . 174 4.12 Energy consumption with and without RESs. . . . . . . . . . . . . . . . . . . . . . 176 4.13 Hourly cost with and without RESs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 4.14 Total cost of one day with and without RESs. . . . . . . . . . . . . . . . . . . . . 177 4.15 PAR with and without RESs.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 4.16 Effects of wind speed and temperature on wind and PV generation . .

192

4.17 Comparison of heuristic techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 4.18 Performance trade-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

xvii

List of Figures 3.1 Status of appliances' execution pattern . . . . . . . . . . . . . . . . . . . . . . . . . . .55 3.2 Proposed HEMS architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.3 Pricing signals. CPP, critical peak pricing . . . . . . . . . . . . . . . . . . . . . . . . 62 3.4 Feasible region for cost and electricity consumption using RTP. . . . . . . . . 69 3.5 Feasible region for cost and electricity consumption using CPP. . . . . . . . . 71 3.6 Feasible region for cost and WT RTP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 3.7 Feasible region for cost and WT using CPP. . . . . . . . . . . . . . . . . .. . . . . . . 75 3.8 Proposed system model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 3.9 Pricing schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.10 Wind turbine power output and wind speed . . . . . . . . . .. . . . . . . . . . . . . 91 3.11 Block diagram of system model . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 99 3.12 Functional model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 3.13 Abstract view diagram of power flow

. . . . . . . . . . . . . . . . . . . . . . . . . 120

3.14 System architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 4.1 Hourly electricity consumption for single and multiple homes . . . . . . . . . 137 4.2 Hourly electricity cost using RTP signals. . . . . . . . . . . . . . . . . . . . . . . . . 139 4.3 Total electricity cost using RTP signals. . . . . . . . . . . . . . . . . . . . . . . . . . 142 4.4 Hourly electricity cost using CPP signals. . . . . . . . . . . . . . . . . . . . . . . . . 144 4.5 Total electricity cost using CPP signals. . . . . . . . .. . . . . . . . . . . . . . . . . . 147 4.6 PAR for single and multiple homes with RTP signals . . . . . . . . . . . . . . 148 4.7 PAR for single and multiple homes with CPP signals. . . . . . .. . . . . . . . . 149 4.8 Appliances' WT. . . . . . . . .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . 151 4.9 Total cost after scheduling using RTP and CPP. . . . . . . . . . . . . . . . . . . . 153 4.10 Cost per time slot for RTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .156 4.11 Load per time slot for RTP.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 4.12 Load per time slot for CPP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 4.13 PAR for different OTI’s using RTP and CPP. . . . . . . . .. . . . . . . . . . . . . 163 4.14 AWT of appliances using RTP and CPP. . . . . . .. . . . . . . . . . . . . . . . . . .165 xviii

4.15 Evolution of the cost function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 167 4.16 Feasible region for multiple OTIs using RTP.. . . . . . . . . . . . . . . . . . . . .. 169 4.17 Feasible region for multiple OTIs using CPP... . . . . . . . . . . . . . . . . . . . .173 4.18 RTP pricing scheme. . . . . . . . . . . . . . . . . . . . . . . . .. . .. . . . . . . . . . . . . . 175 4.19 Energy consumption profile by WDGA . . . . . . . . . . . . . . . . . . . . . . . . . 179 4.20 Energy consumption profile by WDGWO . . . . . . . . . . . . . . . . . . . . . . . 180 4.21 Energy consumption profile by WBPSO. . . . . . . .. . . . . . . . . . . . . . . . . 181 4.22 Electricity cost of WDGA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 4.23 Electricity cost of WDGWO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 4.24 Electricity cost of WBPSO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 4.25 Total electricity cost. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 4.26 PAR with and without RESs.. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . 188 4.27 Appliances waiting time by WDGA, WDGWO and WBPSO . . . . . . . . 189 4.28 Hourly electricity generation.. . . . . . . . . . . . . . . . . . . . . . . .. . . . . 191 4.29 Relationship between wind generation and wind speed . . . . . . . . . . . . . 192 4.30 Power consumption of appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 4.31 Hourly power consumption of appliances . . . . . . . . . . . . . . . . . . . . . . . . 195 4.32 Hourly power consumption of appliances . . . . . . . . . . . . . . . . . . . . . . . 197 4.33 Hourly electricity cost of appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 4.34 Hourly electricity cost of appliances . . . . . . . . . . . . . . .. . . . . . . . . . . . . 199 4.35 Daily discomfort of appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 4.36 Hourly discomfort of appliances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 4.37 PAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 4.38 Feasible region.. . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 4.39 Cost per time slot for RTP... . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 4.40 DAP signal . . . . .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . 209

xix

ABBREVIATIONS u

Electricity users/homes

m

Total number of electricity users/homes

t

Time interval

T

Total time slots

h

Hour

An

All appliances in smart homes

Ad

Deferrable appliances

And

Non-deferrable appliances

Ab

Base load appliances

ad

Each appliance in deferrable class

and

Each appliance in non-deferrable class

ab

Each appliance in base load class

𝜆d

Power rating of deferrable appliances

𝜆nd

Power rating of non-deferrable appliances

𝜆b

Power rating of base load appliances

Ɛd

Total load of deferrable appliances for single home

ϑd

Total load against deferrable appliances for multiple homes

Ɛnd

Total load of non-deferrable appliances for single home

ϑnd

Total load against non-deferrable appliances for multiple homes

Ɛb

Total load of base load appliances for single home

ϑb

Total load against base load appliances for multiple homes

αd

On/off status of deferrable appliances

αnd

On/off status of non-deferrable appliances

αb

On/off status of base load appliances

Total 𝛿αd

Per day cost against deferrable appliances for single home

Total 𝜑αd

Per day cost against deferrable appliances for multiple homes

Total 𝛿αnd

Per day cost against non-deferrable appliances for single home

Total 𝜑αnd

Per day cost against non-deferrable appliances for multiple homes

Total 𝛿bd

Per day cost against base load appliances for single home xx

Total 𝜑αb

Per day cost against base load appliances for multiple homes

t 𝜎αd

Per hour cost against deferrable appliances for single home

t 𝜎αnd

Per hour cost against non-deferrable appliances for single home

t 𝜎αb

Per hour cost against base load appliances for single home

δTotal

Total cost per day against all appliances for single home

ξTotal

Total cost per day against all appliances for multiple homes

𝜎t

Total cost per hour against all appliances for single home

ζt

Total cost per hour against all appliances for multiple homes

η

Starting execution of any appliance

τ

Waiting time against each appliance

EStch

Charging of electricity storage system

ES (max)

Maximum charging of electricity storage system

EStdis

Discharging of electricity storage system

ES(max)

Maximum charging of electricity storage system

ESS(upl)

Upper limit of charging of electricity storage system

ESStch

ESS charging level

kW

Kilowatt

Pc

Crossover

Pm

Mutation

D

Total appliances

Ds

Set of shiftable appliances

Dc

Set of controllable appliances

Dns

Set of non-shiftable appliances

ρ

Power rating

χ

Status of appliance

Th

Time slots

µ

Trial vector

u

Mutant vector

Uj

Upper bound

lj

Lower bound

CR

Crossover rate xxi

POP

Population

Max.iter

Maximum iteration



Position of the α search agent



Position of the α search agent



Position of the α search agent

F

Scaling factor

A, C

Fitness coefficients

α

Best search agent

δ

Third best optimal solution

PPV-out

PV panel output power

Gref

Solar radiation at reference conditions

KT

Temperature coefficient of the PV panel

𝜎

Air density

Pcoff

Wind turbine power coefficient

S

Set of appliances α

Vcut-in

Cut-in wind speed

CWh

BBS storage capacity

ηV

BBS voltage

CB(t)

Available power from BBS at time slot t

𝜆

BBS self-discharge rate

CBmin

Minimum allowable energy level remain in the BBS

α

Appliance

PSrtp(t)

Real time PS in time slot t

YilSIα(t)

ON/OFF state of IL appliances

ndl Eps (t)

EP of NDL appliances in time slot t

mrl Eps (t)

EP of MRL appliances in time slot t

SNα

Represents NDL appliances

SMα

Represents MRL appliances

ilSIα

SI represents the number of appliances of IL

ΓPAR

PAR of the demanded load

γtαwt

Waiting time of appliance

xxii

𝑠𝑡 Tαw

Appliance α start time

Tmw

Appliance α maximum waiting time

Eil(t)

EC of IL appliances in time slot t

EPV(t)

Available energy from PV in time slot t

BS(t)

Available energy from battery in time slot t

Eug(t)

Available energy from utility grid in time slot t

t0

Lower limit of scheduling horizon

tsch

Scheduling time of appliance

Xid(t-1)

Position of particle i in the d dimension at time slots t

Vid(t-1)

Velocity of particle i in the d dimension at time slots t-1

Gbestid(t-1) Best positions obtained by particle i and swarm in d dimension in time slot t-1 c

Coriolis force

r

Variable value for the rank of air parcels

FCr

Coriolis force

v

Wind velocity

g

Acceleration of gravity

Δ

Pressure gradient

𝜑

Friction coefficient

p

𝑣i+1

Current and new velocity of the air parcels

xgbest

Global best position

PN-PV

Rated or nominal power of PV cell at reference conditions

EαNDL

Energy consumption by NDL appliances EαNDL

EαMRL

Energy consumption by MRL appliances

LA

Average load

X (t)



In X (t), t is current iteration Prey position vector Best search agent



Third best search agent

G

Solar radiation

Tref

Cell temperature at reference conditions

Ars

Rotor swept area

X

𝑝

xxiii

V3

Average wind velocity

A

Set of appliances

𝐸𝑇𝑝𝑒

Total energy consumption of power elastic appliances

𝐴𝑡𝑒

Set of time elastic appliances

𝐸𝑐𝛼 (t)

Energy consumption of interruptible appliances

𝐸𝑒𝑝

Set of power elastic appliances

𝑃𝑟𝑎

Power rating of interruptible appliances

𝑆𝑡𝑖

Current position of an appliance i

𝐸𝑇𝑎

Total energy consumption of interruptible appliances

𝑖 𝑆𝑡+1

Position of appliance at the next time slot

𝐸𝑐b (𝑡)

Energy consumption of non-interruptible appliances per time slot

𝑇𝑠0

ON time

𝑇𝑟b

Power rating of non-interruptible appliances

β

Operation end time

𝑇𝑇b

Total energy consumption by β appliances

pe

𝐶𝑇

Total cost of power elastic appliances

𝑟𝑡n

Number of remaining time slots

ρ(t)

Power rating

𝑤𝑡n

Number of waiting time slots

𝐶𝑇te

Total cost of time elastic appliances

Xt

ON/OFF status

wi

Waiting time of appliance i

i 𝑝𝑚𝑎𝑥

Maximum power of appliance i

ti

Time of appliance i

i 𝑝𝑚𝑖𝑛

Minimum power of appliance i

𝑝(Lt)

Combined electricity price at time slot t

βi

End time slot of Appliance i

Rt

Real-time price

CT

Total cost of all appliances

bt

Electricity price when IBR threshold exceeded

ET

Total energy consumption of all appliances xxiv

Lt

Total load

MDi

Mean difference

Lth

Threshold of load

Meannew

Outcome of the best learner

j

𝐸𝑐 (t)

Energy consumption of power elastic appliances per time slot

ri

Random number between 0 and 1 j

𝑝𝑟

Power rating of power elastic appliances

Tfactor

Teaching factor

𝐸𝑇te

Total energy consumption of time elastic appliances

DSM

Demand side management

GWO

Gray wolf optimization

EDE

Enhanced differential evaluation

HGWDE Hybrid gray wolf differential evaluation PSO

Particle swarm optimization

BPSO

Binary particle swarm optimization

TLBO

Teaching and learning-based optimization

IDSS

Intelligent decision support system

MOO

Multi-objective optimization

EMC

Energy management controller

LOT

Length of operational time

CPP

Critical peak pricing

OTI

Operational time interval

DR

Demand response

WDO

Wind-driven optimization

TOU

Time of use

FRP

Flat rate pricing

IBR

Inclined block rate

DHP

Day-ahead pricing

PS

Pareto sets

SG

Smart grid

SM

Smart meter

xxv

GA

Genetic algorithm

SFL

Shuffled frog leaping

MPP

Multi-parametric programming

MRL

Must-run load

MRLA

Must run load appliances

NCP

Non-critical peak

NDL

Non-deferrable load

NDLA

Non-deferrable load appliances

OCM

Optimal control method

PEV

Plug-in electric vehicle

PC

Personal computer

PCPM

Predictor corrector proximal multiplier

PHEV

Plug-in hybrid electric vehicle

PMU

Phaser measurement unit

PMU

Power management unit

PP

Peak pricing

PS

Price signal

PSO

Particle swarm optimization

AD

Autonomy days

HP

Hourly pricing

PS

Price signal

RE

Renewable energy

RESs

Renewable energy sources

RTMP

Real-time market pricing

RTP

Real time pricing

SCADA

Supervisory control and data acquisition

SI

Set of IL appliances

SM

Smart meter

SM

Set of must-run appliances

SMSU

Smart scheduler unit

SN

Set of NDL appliances

T

Temperature xxvi

TOU

Time of use

UC

User comfort

USA

United States of America

WBPSO

Wind driven BPSO

GA

Genetic algorithm

WDGA

Wind driven genetic algorithm

WDGWO Wind driven GWO algorithm WDO

Wind driven optimization

ACO

Ant colony optimization

HEMA

Home energy management architecture

AMI

Advanced metering infrastructure

HEMS

Home energy management system

AWT

Appliances waiting time

HEM

Home energy management

BAB

Branch and bound

HEMC

Home energy management controller

BESS

Battery energy storage system

HGPO

Hybrid of GA and PSO

SHEMS

Smart home energy management system

BFOA

Bacterial foraging optimization algorithm

ICT

Information and communication technology

CPCT

Conventional programming communication thermostat

iDES

Incentive-driven distributed energy sharing system

IHD

In-home display

CPR

Critical peak rebates

IPCT

Intelligent programming communication thermostat

CPSO

Cooperative particle swarm optimization

kW

Kilowatt

CSA

Cuckoo search algorithm

K-WDO

knapsack based WDO

DAP

Day-ahead pricing

LP

Linear programming xxvii

DEMS

Distributed energy management system

MFCFS

Modified first come first serve

MIPO

Mixed integer programming optimization

MKP

Multiple knapsack problem

ECSU

Energy consumption scheduling unit

MTPSO

Multi-team particle swarm optimization

NCG

Non-cooperative game

EDTLA

Enhanced differential teaching learning algorithm

NE

Nash equilibrium

OEMS

Optimized energy management system

EMU

Energy management unit

OSR

Optimal stopping rule

ESS

Energy storage system

PAR

Peak-to-average ratio

EV

Electric vehicle

PDH

Power distribution hub

EA

Evolutionary algorithms

PEEDF

Priority enabled early deadline first

FBAT

Flower pollination BAT

FCFS

First come first serve

PV

Photovoltaic

FGA

Flower pollination GA

FPA

Flower pollination algorithm

FREC

Federal regulatory energy commission

FTLBO

Flower pollination TLBO

RSM

Realistic scheduling mechanism

SEH

Smart energy hub

GBPSO

Genetic BPSO

SG

Smart grid

G-DSM

Generic demand side management

GTLBO

Genetic teaching learning based optimization

ToU

Time of use xxviii

UC

User comfort

HA

Heuristic algorithms

UP

User preferences

HEGS

Hybrid energy generation system

xxix

Acknowledgements My cordial gratefulness goes to Allah Almighty who has provided all means that was needed to complete this research work. There was some very hard times. Throughout this entire study, my motivation goes up and down, but he took care of everything that would have stopped me in my tracks and supported me even through my most tough times. I am thankful to the head of ComSens and my Co-supervisor Dr. Nadeem Javaid, who provided me the opportunity to do my research work in his Lab and allowing me to COMSATS University, Islamabad, where I have been trained to take up my fortune. My Co-supervisor whose support and constructive criticism has pressed me to disburse the kind of efforts I have exerted to make this work as original as it can be. Thanks to him I have experienced true research and my knowledge on the subject matter has been widened. His continuous support, motivation and extended guidance enable me to complete this research work. I will never forget you sir. Special thanks goes to my supervisor Dr. Saleem Iqbal, whose continuous guidance and support motivated me to complete this journey. He always pushed me and ask me about my work progress. Thank you sir very much. I also appreciate the Director of UIIT, Dr. Mubashir Ali Khan, who has shown exemplary cooperation and guidance as a mentor. I cannot forget deputy director advanced studies and research, Dr. Farooq, who is a real mentor and guided me in synopsis and thesis write up. He provided me motivation and continuous support during my research work. I also appreciate the support of Dr. Sohail Asghar and my supervisory committee members: Dr. Saud, and Dr. Mohammad Azeem for their continuous support and cooperation. I sincerely appreciate Dr. Faraz Ahsan, Dr. Khalid Usmani, Dr. Munam Ali Shah, Dr. Yasir Hafeez, Tariq Ali; all of whom I have had direct contact with and who have impacted me during this program. I say a big thank you. My appreciation also goes to the program coordinator Assistant Professor Saqib Majeed, deputy director Ghulam Mustafa and assistant director Mazoor of UIIT department for their support and help.

xxx

A very special thanks to Imran and Ahmad Bhatti in advanced studies and research, Liaqat, examination Director UIIT staff: Khurram, Fawad, shahid and IT support team: Zeeshan for their continuous support, cooperation and coordination. My appreciation and thanks also goes to my Lab colleagues: Sakeena Javaid (PhD Scholar), Muhammad Awais, Mr. Ihsan Ullah, Dr. Muhammad Babar Meher, Sheraz Ahmad, Muqadas Naz, Mudassar, Anwar Khan, Dr. Saleem Khan, Dr. Danish Mahmood, Samia shah, Sahar, Madam Nusrat, Mr. Mohsin (PhD scholar), Asif Khan, Adia Khalid, Rasool Bakish, Manzoor for their continuous help and technical discussion on the subject matter. My sincere thanks also goes to the whole academic staff both past and present of the Department of Computer Science, UIIT and COMSATS University. My highest regard also goes to my Parents, who thoroughly laid the foundation for my education giving it all it takes. I am also thankful to my Siblings, who gave me courage and help me out through this journey, especially my brother Mohammad Imran who support me in every aspect.

xxxi

ABSTRACT The smart grid plays a vital role in decreasing electricity cost via Demand Side Management (DSM). Smart homes, being a part of the smart grid, contribute greatly for minimizing electricity consumption cost via scheduling home appliances. However, user waiting time increases due to the scheduling of home appliances. This scheduling problem is the motivation to find an optimal solution that could minimize the Peak to Average Ratio (PAR) and electricity cost with minimum user waiting time. There are many studies on Home Energy Management (HEM) for cost minimization and peak load reduction. However, none of the systems gave sufficient attention to tackle multiple parameters (i.e., electricity cost and peak load reduction) at the same time where user waiting time is considered to be minimum for residential consumers with multiple homes. Hence, in contribution 1, we propose an efficient HEM scheme using the well-known meta-heuristic Genetic Algorithm (GA), the recently developed Cuckoo Search Optimization Algorithm (CSOA) and the Crow Search Algorithm which can be used for electricity cost and peak load alleviation with minimum user waiting time. The integration of a smart electricity storage system is also taken into account for more efficient operation of the HEM System. Furthermore, we took the real-time electricity consumption pattern for every residence, i.e., every home has its own living pattern. The proposed scheme is instigated in a smart building which is comprised of thirty smart homes (apartments). Critical Peak Pricing (CPP) and Real-Time Pricing (RTP) signals are examined in terms of electricity cost assessment for both a single smart home and a smart building. In addition, feasible regions are presented for multiple and single smart homes, which show the relationship among the electricity cost, electricity consumption and user waiting time. Experimental results prove the effectiveness of our proposed scheme for multiple and single smart homes concerning electricity cost and PAR minimization. Moreover, there subsists a tradeoff between electricity cost and user waiting. With the emergence of automated environments, energy demand by consumers is increasing rapidly. More than 80% of total electricity is being consumed in the residential sector. This brings a challenging task of maintaining the balance between xxxii

demand and generation of electric power. In order to meet such challenges, a traditional grid is renovated by integrating two-way communication between the consumer and generation unit. To reduce electricity cost and peak load demand, DSM is modeled as an optimization problem and the solution is obtained by applying metaheuristic techniques with different pricing schemes. In contribution 2, an optimization technique, the Hybrid Gray Wolf Differential Evolution (HGWDE) is proposed by merging the Enhanced Differential Evolution (EDE) and Gray Wolf Optimization (GWO) schemes using the same RTP and CPP tariffs. Load shifting is performed from on-peak hours to off-peak hours depending on the electricity cost defined by the utility. However, there is a trade-off between User Comfort (UC) and cost. To validate the performance of the proposed algorithm, simulations have been carried out in MATLAB. Results illustrate that using RTP, the PAR is reduced up to 53.02%, 29.02% and 26.55%, while the electricity bill is reduced up to 12.81%, 12.012% and 12.95%, respectively, for 15-min, 30-min and 60-min operational time intervals (OTI). On the other hand, the PAR and electricity bill are reduced up to 47.27%, 22.91%, 22% and 13.04%, 12%, 11.11% using the CPP tariff. Microgrid is a community-based power generation and distribution system that interconnects smart homes with renewable energy sources. Microgrid generates power for electricity consumers and operates in both islanded and grid-connected modes more efficiently and economically. In contribution 3, we propose optimization schemes for reducing electricity cost and minimizing PAR with maximum UC in a smart home. We consider a grid-connected microgrid for electricity generation which consists of wind turbine and photovoltaic (PV) panel. First, the problem was mathematically formulated through Multiple Knapsack (MKP) then it is solved by existing heuristic techniques: GWO, binary particle swarm optimization (BPSO), GA and Wind Driven Optimization (WDO). Furthermore, we also propose three hybrid schemes for electricity cost and PAR reduction: (1) hybrid of GA and WDO named as WDGA; (2) hybrid of WDO and GWO named as WDGWO; and (3) WBPSO, which is the hybrid of BPSO and WDO. In addition, a battery bank system has also integrated to make our proposed schemes xxxiii

more cost-efficient and reliable to ensure stable grid operations. Finally, simulations have been performed to verify our proposed schemes. Results show that our proposed schemes efficiently minimize the electricity cost and PAR. Moreover, our proposed techniques: WDGA, WDGWO and WBPSO outperform the existing heuristic techniques. The advancements in smart grid, both consumers and electricity providing companies can benefit from real-time interaction and pricing methods. In contribution 4, a smart power system is considered, where consumers share a common energy source. Each consumer is equipped with a Home Energy Management Controller (HEMC) as scheduler and a smart meter. The HEMC keeps updating the electricity proving utility with the load profile of the home. The smart meter is connected to power grid having an advanced metering infrastructure which is responsible for two way communication. Genetic teaching-learning based optimization, flower pollination teaching learning based optimization, flower pollination BAT and flower pollination genetic algorithm based energy consumption scheduling algorithms are proposed. These algorithms schedule the loads in order to shave the peak formation without compromising UC. The proposed algorithms achieve optimal energy consumption profile for the home appliances equipped with sensors to maximize the consumer benefits in a fair and efficient manner by exchanging control messages. Control messages contain energy consumption of consumer and RTP information. Simulation results show that proposed algorithms reduce the PAR by 34.56% and help the users to reduce their energy expenses by 42.41% without compromising the comfort. The daily discomfort is reduced by 28.18%.

xxxiv

CHAPTER 1 INTRODUCTION 1.1 INTRODUCTION In this chapter, we will discuss the significance of the Smart Grid (SG), the role of SG in traditional grid, role of Microgrids (MGs) and other Renewable Energy Sources (RESs), optimization techniques and energy consumption after proper scheduling of the appliances in the residential sector. The focus of this research is towards the appliances’ scheduling in the residential area using heuristics optimization techniques. Later, problem statements and research questions regarding existing work are also discussed. 1.1.1 Why Smart Grid? World’s increasing population, global warming, the rise in carbon emissions and increasing electricity demand create an alarming situation for electricity producing and distributing companies as well as for governments to take any strong action against these alarming situations. The electricity producing companies are detained from integrating RESs to overcome global warming and carbon emissions (Benzi, Anglani, Bassi, & Frosini, 2011). The present fossil fuel based electric grid is working on the centralized approach; only a few large electricity producing plants are operating at 50 Hz or more. High-power electricity plants are operating at very high voltage (i.e., 400 kV or more). Then, the produced electricity from large plants is distributed to the electricity consumers. A large number of supply lines supply the high voltage load to heavy industries and low voltage load to residential consumers and small-scale industries. The power flow in the present power system is unidirectional due to the centralized approach. Electricity consumers are considered just passive users; they cannot play any role in the stability and reliability of the electric grid. According to (Evangelisti, Lettieri, Clift, & Borello, 2015), more than 65% of total electricity is wasted during generation, transmission and distribution of electricity. The basic reasons behind electricity wastage are that the present electricity system has 1

unidirectional communication and a lack of monitoring technologies. The novel approach is distributed based on bidirectional communication. It provides the widely and highly distributed intelligence in electric power generation, distribution and flow of information. Furthermore, the novel approach provides the multiple opportunities for electricity consumers to manage their electricity consumption for bill reduction and reliable grid operations. Enhancement of the smart grid is increasing every day in order to make a robust and reliable system. The world is now moving towards automation, hence increasing the electricity consumption due to excessive use of automatic equipment. Reduction in fossils fuels and high demand of electricity lead to a dependency on RES (Mhanna, Chapman, & Verbic, 2016). However, usage of electricity can be controlled through scheduling and coordination of appliances. According to the authors in (Energy Reports, 2007), 38% augmentation in consumption by the power sector is expected by the year 2020. For the residential and commercial sector, there is an expectation of a 16% upturn. Recently, increasing energy consumption has been observed around the globe. Presently, most of the power is produced from fossil fuels which increases carbon emissions. To minimize carbon emissions and fulfill the inevitably increasing electricity demand, scientists have explored the alternative sources of energy generation, i.e., RESs. Moreover, complexity of power system is significantly increased due to the penetration of RESs. However, the large-scale installation of RESs to the existing conventional power system will increase the vulnerability of already heavily loaded power system (Guo, Pan, & Fang, 2012). For this purpose, the transformation of the current electric power system to the smart grid, i.e., the unification of advanced Information and Communication Technologies (ICTs) with conventional power grid is one of the best solutions (Agnetis, Pascale, Detti, & Vicino, 2013). These technologies not only exploit the stability and reliability of the power system; however, they also enable the SG to efficiently incorporate the RESs and DG.

2

1.1.2 SG and it’s Role in Traditional Grid Considering the growth in electricity consumption, there is a need to reorganize the pattern of energy consumption and production on the demand and supply side (Logenthiran, Srinivasan, & Shun, 2012). For this purpose, the entire focus is to replace the traditional grid via SG with the integration of the latest telecommunication technologies. Advanced Metering Infrastructure (AMI) provides two-way communication between the utility and consumer (Shirazi, & Jadid, 2015). In a SG, the utility manages the electricity consumption and demand of the user by a specific set of rules. These rules are called Demand Response (DR) (Bradac, Kaczmarczyk, & Fiedler, 2014). Multiple incentives are declared by the electric utility to encourage the consumer for using electricity resources efficiently as explained in (Teets, 1992). Widely used pricing signals are: Real Time Pricing (RTP), Time of Use (TOU), Critical Peak Pricing (CPP), Flat Rate Pricing (FRP), Day-Ahead Pricing (DAP) and Inclined Block Rate (IBR). The Energy Management Controller (EMC) within a smart home receives the pricing signals from the utility. The schedule is generated by EMC using two inputs: pricing signal and load demand of the user. Appliances are synchronized with these schedules through Wi-Fi, infrared or ZigBee (Mahmood, Javaid, Khan, & Razzaq, 2016). In SGs, two-way communication provides an opportunity to optimize consumption costs along with peak formations. Due to the advent of a smart grid, a lot of studies have focused in regard to cost and PAR reduction via Demand Side Management (DSM) (Khalid et al., 2018; Albadi, & El-Saadany, 2008; Avci, Erkoc, Rahmani, & Asfour, 2013; Yang, Zhang, & Ma, 2014). However, none of this work has included the capability to generate and store electricity for future use. The authors in (Ahmad et al., 2017) have proposed a DSM scheme by considering different types of electricity consumers. Optimum electricity consumption with maximum user comfort (UC) is determined in (Aslam et al., 2017) within a smart home containing different types of smart appliances. They also investigate their proposed scheme on smart buildings comprised of multiple smart homes with different living patterns (power rating and load demand). A new cost efficient home energy management scheme has been proposed in (Sander, AlSkaif, & Wilfried, 3

2018). The authors of (Sander, AlSkaif, & Wilfried, 2018) also integrate the RESs to minimize the electricity cost and carbon emissions. Furthermore, consumers are able to store excess electricity in batteries for future use; when electricity rates are high, stored electricity is consumed. In (Liu, & Hsu, 2018), authors have proposed a new SG based architecture for electricity consumers. They also integrate the RESs for electricity generation. According to their work, consumers are able to consume, generate, store and sell excess electricity. The excess electricity is sold back to the electric grid for earning profit maximization. Most of the recently proposed schemes have discussed the problem from the grid or electricity consumer’s perspective. To make an advance and automated energy management and distribution system, SG incorporates new, smart and intelligent technologies. Smart controllers and relays along with intelligent software tools are used for data management. The best feature of the SG is the bidirectional communication between power companies and consumers. The exchange of information enables the utility companies and consumers to control their load, reduce bill and Peak-to-Average Ratio (PAR). The gain of UC, implementation of user preferences and integration of Renewable Energy (RE) is another advantage. The addition of these new and intelligent technologies in the next generation power grid is going to be incorporated across the entire power system. SG incorporates new technologies from generation, transmission and distribution of power consumption at the consumer’s side. These technologies are used for the purpose of enhancing the safety, reliability and efficiency of the power system. The novel approach is the integration of cost-efficient RESs which holds promise to tackle the above-discussed problems in the traditional electric grid. In the SG, electricity is generated via cheaper and efficient resources and then distributed to electricity consumers through smart transmission lines. Electricity prosumers are the consumers because they can utilize as well as produce the electricity from their own local MG which consists of multiple RESs, i.e., solar panel, wind turbine, hydro power plant, etc. They utilize electricity from their own generation and are also interconnected with the commercial grid. In case of less electricity generation as compared to load demand, they purchase electricity from utilities or neighbors. If the 4

electricity production from their own MG is more than the load demand, then the excess electricity is sold back to the commercial grid or stored in batteries for future use when electricity generation is low. The batteries may discharge only when electricity production from the MG is low or per unit electricity price is high. Generally, the main focus lies in load shifting from on-peak hours to off-peak hours by applying different strategies such as load clipping and valley filling. Load shifting helps to achieve cost minimization by shifting the operation of appliances to low-price hours (Veras, Pinheiro, Silva, & Rabelo, 2017). Although load shifting reduces the electricity bill; however, it leads to increase in user discomfort. In order to achieve these objectives, global optimization algorithms are mostly used due to their low complexity and fast convergence towards optimal points. 1.1.3 Role of RESs and MGs At present, RESs generate few kilowatts (kW) or megawatts (MW) of electricity in residential areas and integration of RESs on a large-scale is widely diffused around the globe. Furthermore, electricity generation via RESs and integration of storage systems are enabling smart homes and small-scale industries to gain profit by selling excess electricity to the grid or neighbors. Moreover, end users may purchase energy when electricity tariffs are low and sell back electricity when prices are high. In addition, approximately 10–30% electricity consumption can be saved through DSM (Tascikaraoglu, Boynuegri, & Uzunoglu, 2014). There are many dynamic pricing schemes to calculate electricity consumption cost for consumers’ motivation to alleviate their electricity consumption in on-peak hours. These pricing schemes include RTP, ToU, CPP and critical peak rebate (CPR) (Economics, & First, 2012). Electricity consumers have the option to select the best electricity tariff according to their satisfaction. RESs have gained prominence over traditional and fossil fuel-based energy sources and they also contribute in environmental degradation. Therefore, policy makers and researchers are being compelled to think about changing the form of energy generation. The Distributed Generation (DG) emerges with the emergence of

5

RESs (Hossain, Kabalci, Bayindir, & Perez, 2014). A MG is considered as a lower layer of the SG and is an independent small scale power generation system that supplies power to the electricity consumers (Farhangi, 2010). MG operates in three different modes: grid-connected mode, where it is needed to sell power back or purchase to/from main grid; off-grid mode, where power is not available from the utility grid; and isolated mode, where utility grid is in far and remote areas. Numerous articles have been published about isolated MG. The authors have discussed stand-alone MG consisting of Photovoltaic (PV) source, wind turbine and storage that are mathematically formulated to design voltage regulation policy and control-based load tracking system. They have proposed a control and energy management policy. According to this strategy, the storage can be charged by constant current and voltage which increases its lifespan. It is also considered in this work that the power demand is less than the generated power (Dizqah, Maheri, Busawon, & Fritzson, 2015). The burgeoning population continuously increases the use of electric appliances which results in increasing power demand. To fulfill this increasing demand of electricity, RESs become lucrative for scientists because of the conventional sources of electricity generation are costly and cause high carbon emissions. Hence, it is necessary to generate more power locally from RESs. In addition, we have to optimize the existing power sources to create alternative methods of power generation. To achieve this end, researchers are working on the utilization of renewable energy (RE) generation in power sector to make it more efficient. According to the concept of a MG, the power could be used in a reliable and optimized way, and more energy will be locally generated. The power of a MG will fulfill the energy requirement along with the considerable reduction in cost and PAR. 1.1.4 Optimization Techniques In SG, the common goals of different DR and DSM strategies are the reduction in electricity bill and PAR. Load shifting schemes are used to achieve balance energy consumption. To design an effective home energy management system (HEMS),

6

different algorithms are used by research community, e.g., Mixed Integer Linear Programming (MILP) (Tsui, & Chan, 2012), Dynamic Programming (DP), Multi Parametric Programming (MPP) (Oberdieck, & Pistikopoulos, 2016), Integer Linear Programming (ILP) (Hubert, & Grijalva, 2011), etc. However, these algorithms have unpredictable energy consumption patterns and cannot handle a large range of different home appliances. Furthermore, authors in (Corchero, Cruz-Zambrano, & Heredia, 2014; Graditi, Ippolito, Telaretti, & Zizzo, 2016; Ippolito, Telaretti, Zizzo, & Graditi, 2013; Molderink, Bakker, Bosman, Hurink, & Smit, 2010; Siano, Graditi, Atrigna, & Piccolo, 2013; Pipattanasomporn, Kuzlu, & Rahman, 2012) have focused on electricity cost reduction and PAR minimization through stochastic, mathematical and heuristics techniques. 1.1.5 Energy Consumption and its Scheduling Effects on Consumers Residential and commercial buildings consume 50% of the global power (Vardakas, Zorba, & Verikoukis, 2015). The active collaboration and energy sharing by buildings and homes are not included in the present electricity system. All the energy management efforts carried out are less effective and their efficiency is affected by current demand of the users. SG enables the next generation energy efficiency and sustainability. In this kind of architecture as discussed above, the aggregation of houses acts as an intelligent collaboration of networked system. Whereas, in the conventional system, they act as passive and isolated units (Kok et al., 2011). With the help and cooperation of SG technologies, investment on a traditional grid can be changed or reduced by implementing DSM rules. The participants of the DSM or DR programs try to reduce their energy utilization at certain instances by showing a little flexibility. By offering flexibility, consumers achieve benefits (Kohlmann, Van, Knigge, Kobus, & Slootweg, 2011). Since 1982, peak electricity demand and electricity usage have increased due to the growth in electric appliances and increasing demand of power by industry. The increase is almost 25% by each year according to the U.S department of energy (Leon, Salcedo, Ran, & Martinez, 2015). Moreover, considering the residential sector in the U.S, the electricity sales are expected to increase 24% from 2011 to 7

2040 (Conti, 2013). Peak energy demand is expected to be far more than the available transmission, generation and distribution capability of the existing grid. This dilemma can be solved by enhancing the existing transmission capability, decreasing peak load, increasing distributed generation and exploring new methods of energy generation such as RESs. Researchers are trying to expand traditional grid infrastructure to meet new challenges; however, it is very expensive job (Leon, Salcedo, Ran, & Martinez, 2015). There are also monitory benefits added to the physical system consideration. The peak power stations can be eliminated by reducing the load during peak hours. This ensures decrease in cost of electricity for consumers. As an example, during the California energy crisis of 2000-2001, a 5% peak demand reduction decreases the highest wholesale prices by 50% as stated in (Osborne, & Warrier, 2007). The authors attempt to decrease peak load demand by intelligent and smart coordination amongst customer’s appliances scheduling. The peak demand is avoided by scheduling the appliances in low peak hours and thus, benefiting both the consumers and utility companies (Hansen, Roche, Suryanarayanan, Maciejewski, & Siegel, 2015). There is an emerging trend towards scheduling the load of residential homes for reducing electricity cost and balancing the energy consumption across 24 hours of a day. The use of MG comprising of RESs are in spotlight. Normally, PV source, wind turbines and micro-combined heat and power generators are used. The current power generation system is in a transition state to become a large scale distributed power generation system. This transition state to SG will be completed with the addition of distributed RESs (Koutsopoulos, Tassiulas, & Hellas, 2011), (Ramachandran, Srivastava, Edrington, & Cartes, 2011). SG also supports changing electricity prices and this change is according to the dynamic status of electric power demand and generation system. It is now a promising method for the RE generation, resources management, use in the context of increasing energy demand and increasing prices (Smart Grid Australia, 2010). The recent smart appliances and SG technologies enable residential and commercial sector to use power efficiently using its advanced features. Such 8

electrical appliances have the capability to make their operation according to the changing electricity prices. Peak load management can reduce the cost of electricity consumption. The reliability of electric grid can be improved by smart appliances and their load management characteristics knowing every minute details of each appliance (Zhao, Ding, Cooper, & Perez, 2014). The researchers are trying to add intelligence for energy management in order to improve the efficiency, comfort, convenience in services and home-based healthcare support (Cook, 2012), (Davidoff, Lee, Yiu, Zimmerman, & Dey, 2006). There are many articles, in which especially smart homes energy management systems (SHEMS) considering energy efficiency and load management are discussed (Kofler, Reinisch, & Kastner, 2012; Giorgio, & Pimpinella, 2012; Iqbal et al., 2018; Khan et al., 2018). Most of the literature considers reduction of cost via load management by following variations in electricity prices. Our work proposes a novel approach for appliances scheduling in residential buildings and SHEMS as a detailed solution. The proposed approach minimizes the overall daily electricity cost of home appliances. 1.2 PROBLEM STATEMENT In SG, DSM techniques are used to overcome the irregular use of the electricity by the consumers. There are multiple cases for the improper usage of the energy from the consumers’ side especially in the residential sector. In first case, there can be no local power generation station (i.e., MG or other RESs) for fulfilling the consumers’ requirements. In second case, no proper scheduling of the appliances is done which are consuming the heavy load. They create burden on the utility without EMC for their scheduling. Third, user preferences are not taken into consideration. Fourth, feasible cost optimization is not performed. Relating to these problems, some existing studies are explored in the subsequent subsections and highlighted the detailed descriptions for these problems which are discussed below. 1.2.1 Sub-problem 1 Growing electrical energy cost and environmental pollution reduction are two increasing international problems (Fuselli et al., 2013). At present, power supply 9

systems are dependent on few large electricity-producing plants by using usual fossil fuels. Afterward, the electricity that is generated by the conventional power plants is distributed to the electricity customers with the help of distribution and transmission networks. According to (Evangelisti, Lettieri, Clift & Borello, 2015), during electricity generation, transmission and distribution, more than 65% of the power produced is wasted. The evolution of the smart grid provides opportunities to organize energy management systems and functions to maximize energy savings. Multiple research studies have focused in PAR and cost reduction via DSM techniques (Khalid et al., 2018; Albadi, & El-Saadany, 2008; Avci, Erkoc, Rahmani, & Asfour, 2013; Yang, Zhang, & Ma, 2014) with the advent of the SG; however, these studies lack the MG integration for consumers’ own generation and storage requirements. The authors in (Ahmad et al., 2017) have planned a DSM scheme by taking into account different types of electricity customers. This scheme lacks the consumer comfort. Optimum electricity consumption with maximum UC is determined in (Aslam et al., 2017) in a smart home comprising of various elegant appliances. They test their proposed scheme on smart buildings with multiple smart homes and different existing patterns (power rating and load demand). A cost efficient HEM scheme has been proposed in (Sander, AlSkaif, & van, 2018). The authors of (Sander, AlSkaif, & van, 2018) also incorporate the RESs to minimize the power cost and carbon emissions. In addition, the customers are able to store surplus electrical energy in batteries for future use; when electrical energy rates are far above the ground, stored electrical energy is consumed. The authors of (Liu, & Hsu, 2018) have proposed a novel SG architecture for electricity customers. They also incorporate the RESs for electrical energy generation. Motivated to these schemes, there is a need to develop and EMC which integrates the local storage (i.e., Energy Storage System (ESS)) for consumers’ cost and PAR reduction. 1.2.2 Sub-problem 2 Main focus of the DSM techniques lie in load shifting or load curtailing from onpeak hours to off-peak hours. Load shifting effectively helps in cost reduction by scheduling the operation of appliances to low-price hours (Veras, Pinheiro, Silva, & 10

Rabelo, 2017). Even though, load shifting reduces the electricity cost, consumer discomfort increases due to the trade-off. PAR in the off-peak hours can also increase. In order to attain these objectives, global optimization techniques are mostly used due to their low difficulty and fast convergence towards most favorable points. Based on the aforesaid issues, a new hybrid scheme is required to obtain cost minimization and PAR reduction by maintaining the user preferences. 1.2.3 Sub-problem 3 With the rapid growth in population, the electricity demand in residential area is also increasing. The residential sector consumes almost 40% of the electricity (Lior, 2010). To meet the energy demand, various methods of power generation have been explored. The existing and outdated power systems cannot meet the current power demand required by consumers. In addition, the existing old power system is often subjected to power interruptions due to cumbersome maintenance procedures. The authors design a HEMS model considering ToU pricing scheme and RES integration in (Javaid et al., 2017). Their model uses evolutionary algorithms, i.e., Cuckoo Search Optimization Algorithm (CSOA), Binary Particle Swarm Optimization (BPSO) and Genetic Algorithm (GA) to optimally consume RESs and grid energy. The proposed HEMS model significantly reduces high peaks and electricity cost. However, they have not considered minimization of Average Waiting Time (AWT) for enhancing UC. In (Yu, Kim, & Son, 2013), the authors study the sizing of the storage units and the scheduling of RESs in MG. They have considered the uncertain nature of the MG and associated load. However, they have not considered peak reduction and UC maximization. The authors in (Moon, & Lee, 2016) have provided study of domestic load scheduling problem. To satisfy the budget of the consumers, the authors have proposed a load scheduling algorithm. This problem is difficult to solve and have high computational complexity. To reduce the computational complexity and solve the problem easily, they introduce the generalized bender decomposition approach. They solve the optimization problem providing optimal load scheduling of appliances having different operation characteristics and energy consumption pattern. However, by scheduling the 11

appliances and satisfying the budget limit, the UC has been compromised. These problems can be resolved by the efficient integration of the EMC using the optimization techniques. 1.2.4 Sub-problem 4 Since 1982, electricity usage has increased due to the growth in electric appliances which increases the peak electricity demand. According to the United States department of energy (Leon, Salcedo, Ran, & Martinez, 2015), this demand is increased to almost 25% by each year. In addition, the electricity sales are expected to increase 24% from 2011 to 2040 (Conti, 2013) by considering the residential sector in the U.S. In this scenario, peak energy demand is likely to be far more than the available transmission, generation and distribution capability of the contemporary grid. Peak demand can be resolved by enhancing the current transmission capability, decreasing peak load, increasing distributed generation and exploring new methods of energy generation such as RESs. Existing studies are trying to incorporate the new technologies in contemporary grid infrastructure to meet new challenges; however, it is very expensive job (Leon, Salcedo, Ran, & Martinez, 2015). Some monitory benefits are also added to the physical system consideration. Peak power can be prevented by scheduling the load during the peak hours which ensures the decrement in cost for the consumers. 1.3 RESEARCH QUESTIONS OF THE THESIS This thesis aims to investigate how home load management can help to overcome the energy crisis. For this purpose, following research questions are addressed: Q 1: How ESS and local generation (MGs) helps in cost minimization and PAR reduction for both customers and utility company? How customers can maintain their comfort standards by utilizing the MGs? Q 2: When load shifting is carry out from on-peak hours to off-peak hours depending on the electricity cost defined by the utility then how optimization techniques help in load shifting?

12

Q 3: How hybrid metaheuristic algorithms optimize the cost, PAR and UC? Q 4: How EMC effects the system cost, PAR and UC after scheduling the appliances in comparison to the unscheduled scenario? How EMC is helpful in achieving the monitory benefits for the consumers? 1.4 CONTRIBUTIONS In this section, we will discuss the list of contributions against each problem statement and research question. 1.4.1 Contribution 1 In contribution 1, we propose a novel DSM approach for residential consumers to tackle the home appliances’ scheduling problem, which is based on meta-heuristic techniques, which are implemented in an energy management controller; an embedded system. The main focus of our research work is to alleviate the electrical energy cost and PAR reduction with minimal consumer waiting time simultaneously. We also integrate the ESS to enhance the performance of the proposed scheme. We consider an intelligent building that consists of multiple smart homes, i.e., apartments, for implementation of simulations. Three heuristic techniques, GA, the Crow Search Algorithm (CSA) and the CSOA have been put into practice along with RTP and CPP signals. Simulation results show the effectiveness in terms of electricity bill and PAR reduction with adjustable consumer waiting time. However, a tradeoff occurs between user waiting time and electrical energy cost. If the waiting time is minimum, then the electricity cost is high, otherwise, electrical energy cost is minimum when consumer waiting time is maximum. 1.4.2 Contribution 2 In order to decrease the electricity bill and PAR, load shifting is performed during the on-peak hours. It effects the consumers preferences and the PAR which can be disastrous for both consumers and utility. In order to resolve these challenges (as highlighted in Sub-problem 2), global optimization techniques are mostly used due to their low difficulty and fast meeting towards optimal points. Based on the 13

aforesaid issues, we propose Hybrid Gray Wolf Differential Evolution (HGWDE) algorithm to obtain cost reduction and load shifting. As the demand of electricity in the residential sector is increasing rapidly, there is a need to adopt optimization strategies for efficient utilization of electricity resources. This fact attracts researchers from around the globe to the issues of the sustainability of the electric grid. DR is very beneficial to encourage consumers by introducing incentive-based programs. In this paper, we have also considered a model using three different classes of appliances: shiftable, controllable and non shiftable. Out of 17 appliances, each appliance has been assigned a class on account of their behavior and power ratings. Length of Operational Time (LOT) is different for each appliance. Four parameters are taken into consideration such as power consumption, cost and PAR reduction and waiting time. RTP and CPP schemes are used as the pricing signals. Simulations have been carried out in MATLAB to estimate the behavior of our proposed scheme. 1.4.3 Contribution 3 The work presented in (Corchero, Cruz, & Heredia, 2014; Graditi, Ippolito, Telaretti, & Zizzo, 2016; Ippolito, Telaretti, Zizzo, & Graditi, 2013; Molderink, Bakker, Bosman, Hurink, & Smit, 2010; Siano, Graditi, Atrigna, & Piccolo, 2013; Pipattanasomporn, Kuzlu, & Rahman, 2012) have either focused on specific parameter (i.e., cost reduction, PAR minimization, etc.) or failed to gain full benefits from the current SG technologies to present efficient HEMS. The motivation for this work was to reduce the deficiencies of existing HEMSs. This work introduces an optimized HEMS. The proposed HEMS minimizes the consumer’s electricity cost and PAR with maximum UC while integrating Battery Bank System (BBS) and RESs simultaneously in the residential sector. Moreover, the RTP tariff has used for electrical energy cost estimate along with the implementation of four heuristic techniques, grey wolf optimization (GWO), BPSO, GA, and wind-driven optimization (WDO) to achieve aforesaid objectives. We have also proposed three hybrid optimization algorithms: Wind-Driven GA (WDGA), Wind-Driven GWO

14

(WDGWO) and Wind- Driven BPSO (WBPSO). This work is the extension of (Iqbal et al., 2016). The major contributions of this work are as follows: • We have proposed three hybrid schemes: WDGA, WDGWO and WBPSO. • This work has considered the grid-connected MG system with multiple appliances. • Our proposed work has minimized the electricity cost and PAR. • By implementing our proposed schemes, user can enjoy maximum comfort. • Imported electricity is also reduced by integrating MG. 1.4.4 Contribution 4 In contribution 4, residential load is scheduled using DSM techniques for smart homes. Considering smart power system for consumers, where a common energy source is used. Each consumer uses smart meter and energy consumption scheduling unit (ECSU). The electric grid and smart meter is connected via AMI. The AMI communicates between the electric grid and smart meter. Four algorithms are proposed, i.e., Genetic Teaching-Learning based Optimization (GTLBO), Flower Pollination Teaching Learning based Optimization (FTLBO), flower pollination BAT (FBAT) and flower pollination GA (FGA). These proposed techniques are used to schedule the load for reducing electricity cost, user discomfort and PAR. Simulation outcome show that proposed techniques perform superior as compared to the existing techniques. On the other hand, there is trade-off between cost and consumer discomfort. The discomfort decreases with the increasing cost and increases with decreasing cost.

15

CHAPTER 2 RELATED WORK 2.1 INTRODUCTION In this chapter, related work from the existing studies is discussed based on the optimization techniques, heuristic techniques, hybrid heuristic techniques, conventional techniques, DR based techniques, pricing schemes based techniques, combined pricing model based techniques, MG based techniques, load balancing and peak load shaving, appliances scheduling, user preferences based techniques, electricity price reduction and appliances Waiting Time (WT), PAR and UC oriented techniques, HEMS based techniques, ESS based techniques, MKP for problem formulation based techniques and energy sharing and trading based techniques. 2.2 OPTIMIZATION TECHNIQUES In literature, significant amount of work has been done in SG, MG, macrogrid and hybrid energy generation to optimize energy consumption, energy consumption cost and PAR. Researchers are further working to introduce alternative ways of local energy generation that are less expensive, easy to generate and environment friendly. Some research work indicates that the combination of RESs into residential sector provides the most cost effective solutions. This hybridization of RESs and use of distributed energy resources (DERs) make energy more flexible, reliable, sustainable and remove redundancies. Some related work has been cited below and summary of cited work is presented in Table 2.1. In the last few years, numerous optimization techniques in the SG have been proposed by researchers so as to achieve a general objective such as cost reduction. Hard work have been made for the inexpensive use of electricity. In this section, existing research work on these optimization techniques is offered.

16

2.2.1 Heuristic Techniques The GA and BFOA were proposed with their amalgam scheme in (Ahmad et al., 2017). Three pricing schemes, TOU, RTP and CPP, were used to calculate the performance of these techniques. The key objective of this paper was to attain PAR minimization and electrical load shifting from on-peak hours to off-peak hours. In (Javaid et al., 2017), three optimization techniques, BPSO, EDE and GA, were implemented to attain cost minimization and electrical load shifting, while carbon emission was reduced by integrating RES. DHP is used as a pricing tariff. The primary concern was not only scheduling of appliances, but also prioritizing their operation according to the consumer’s demand. The authors ensured grid sustainability by matching electrical load at the production and consumptions units. However, they ignored UC. The correlation between electricity price and load was developed to alter the energy consumption pattern in slots (Shayeghi, Ghasemi, Moradzadeh, & Nooshyar, 2015), where the electricity price was high. The multiple input and multiple output model was adopted to formulate the problem with the help of Wavelet Packet Transform (WPT) and Generalized Mutual Information (GMI). Forecasting of electricity prices was performed in order to analyze the pattern of variation. The Ant Colony Optimization (ACO) technique was applied for optimization purposes. In (Yang, Yang, Tsai, Chen, & Chen, 2015), the improved version of PSO (IPSO) is used for optimization. The goal of IPSO is to minimize cost. Results illustrate that the user load curve and the objective curve nearly become the same by the proposed IPSO, on the other hand, electrical energy price and the objective curve have an opposite relationship. Power system constancy is one of the objective functions, and the proposed scheme refusing the load in peak hours and thus UC is compromises. The authors in (Peyvandi, Zafarani, & Nasr, 2011) discussed the comparison of GA and PSO for computational complexity. Their results illustrate that PSO have lesser computational complexity to attain a desired result as compared to GA. 17

Genetic algorithm are used for power loss minimization (Jaramillo, Munoz, Ortiz, Lopez, & Albarracin, 2018). For minimizing two conditions, a fitness function is used. Mismatching between power charge ability and desired charge ability and nominal load voltage and changing the top position. The proposed method is for any number of Transformers Connected in Parallel (TCP). A stand-alone MG and decentralized control strategy is proposed in this work. It consists of different modules and each modules consists of wind and solar power, load, three port converter and a battery storage. The economic dispatch of the MG is solved using Mixed-integer Programming Problem (MIP) (Zhang et al., 2018). A day-ahead operation optimization model is proposed, which consists of battery operation cost, fuel cost, and power transmission cost. An Improved Simplified Swarm Optimization (iSSO) technique is therefore proposed to find a feasible solution. Survival of the fittest policy and new update mechanism is presented by iSSO scheme. 2.2.2 Hybrid Heuristic Techniques In (Setlhaolo, & Xia, 2014), the authors considered multi-objective optimization problem by making a test on Pareto Sets (PS) using DE. The core purpose of this model was to deal with multifaceted PS shapes. This model was known as the Multiobjective Evolutionary Algorithm (MOEA). The issue of varying electricity prices was catered by (Osorio, Matias, & Catalao, 2014) and a forecasting model was proposed based on the hybrid evolutionary approach. These forecasts were calculated for 24 and 168 h ahead of time. The hybrid evolutionary approach includes PSO with a neuro-fuzzy logic network in order to eliminate the uncertainty in the market’s electricity rates. 2.2.3 Conventional Techniques The Ant Colony Optimization (ACO) technique for Optimal Power Flow (OPF) in the grid was presented by (Tuaimah, Abd, & Hameed, 2013). The key objective of OPF was to calculate the load of each time period so that power demand could be

18

satisfied at the consumer end. Methods used to solve such problems are LP, NonLinear Programming (NLP), MILP, Mixed Integer Non-Linear Programming (MINLP), the Newton method and the interior point method. In (Bradac, Kaczmarczyk, & Fiedler, 2014), the authors proposed a scheduling technique to reduce the total electricity expenditures and balance the load using MILP for domestic consumers. The proposed technique efficiently reduces the electricity expenditures and PAR. However, UC is not taken into consideration in this work. Another optimization strategy is proposed by in (Agnetis, Pascale, Detti, & Vicino, 2013), using a heuristic method for optimal domestic energy consumption, climate relaxation level and timeliness. Their proposed work efficiently achieved their claimed objectives. In (Setlhaolo, & Xia, 2014), the authors suggested a model for a domestic area for cost reduction by using MINLP under the ToU pricing scheme. They claim that the residents can save 25% or more electricity expenses through adopting their proposed model. Conversely, PAR is not talked in their work. The authors proposed the ILP based scheme to balance the electricity demand and supply in a residential area (Zhu, Tang, Lambotharan, Chin, & Fan, 2012). This scheme is confident about shifting peak power and optimal operation time for powershiftable and time-shiftable appliances accordingly. When more than one home are scheduling their appliances, a more balanced load is attained. Simulation results represent that the proposed scheme efficiently achieved the desired objectives that are claimed. However, the UC is negated here. 2.3 DEMAND RESPONSE BASED TECHNIQUES The authors in (Rahimi, & Ipakchi, 2010) considered DR under the corporate sector. In order to manage energy consumption, certain measures were adopted to maintain a balance between production and consumption. Two different pricing schemes were used, such as DAP and RTP.

19

An optimization model such as GA for balancing electric load was proposed by the authors in (Yang, Yang, Tsai, Chen, & Chen, 2015; Yang, Yang, Tsai, Chen, & Chen, 2015). They took time horizon T equal to 24 h. The load profile maker tool was used to compute energy consumption and the arrangement of load in each time horizon. However, UC was ignored, and the execution time was compromised. In (Jayabarathi, Raghunathan, Adarsh, & Suganthan, 2016), the authors presented DSM practices for the domestic sector. Multiple homes with smart meter were considered for bidirectional communication between the customer and utility company. EMC was set up in each home to schedule the appliances. GA, BPSO, WDO and BFOA were applied to assess the performance regarding cost, UC and PAR. However, trade-off parameters were not discussed. In the above-mentioned literature, the authors did not get a complete benefit from the SG. They minimized PAR or cost, while some authors worked on UC. However, the above-mentioned parameters were not catered to by any literature work simultaneously. Furthermore, multiple Operational Time Intervals (OTIs) were not taken into account. In this research work, the cost and PAR minimization with affordable WT are considered simultaneously to achieve the objective of efficient power usage in the residential sector. The work (Pipattanasomporn, Kuzlu, & Rahman, 2012) discuss an intelligent HEM algorithm. The algorithm manages the power consumption of domestic appliances with DR analysis. The household load is managed using priority and constrained the total domestic load below a certain threshold level. The work provide an insight for performing DR activities for residential consumers. In the next section, the motivation and problem is explained in detail. In (Zakariazadeh, & Jadid, 2013), authors present an operational planning model of MG considering multiple DR programs. They defined two objective functions, cost and carbon dioxide emission, which have been optimized using epsilon constraint multi-objective optimization. The authors used MILP for solving the optimization problem; however, they did not consider PAR and optimization of UC.

20

The authors discussed the contribution in two layers (Pan, Liu, Wu, & Hao, 2018). To optimize the load curve and the maximum load satisfaction, the first layer is optimized. The second layer is discussed to optimize the power utilization ratio and make the MG system economically feasible and environmentally safe. To improve the economy of the system and optimize the operation state of battery, a control strategy for isolated MG is proposed. To explain the scheduling problem of isolated MG Non-dominated Sorting Genetic Algorithm-II (NSGA-II) is used. The scheduling strategy proposed, can improve the environmental conservation of the system as well as operational economy. Furthermore, MG power supply can be improved and energy wastage can be reduce. The authors proposed a hierarchal and energy management framework for MG. This framework optimize operation for the uncertainties in energy load and RESs (Fan, Ai, & Piao, 2018). There are three stages of scheduling: hour-ahead scheduling (HAS), real-time scheduling (RTS) and day-ahead scheduling (DAS). To reduce the operation cost of MG and ensure its safe operation, an optimization model is developed. A decomposition-based algorithm is adopted to settle the model. In this paper the strategic behavior of retailer is implemented. The strategic behavior includes, distributed energy sources, forward contracts, and DR programs (Imani, Zalzar, Mosavi, & Shamshirband, 2018). The aims of these programs is to increase profit and reduce risks. The retail price is kept as low as possible. Risk management problem of the retailer companies is modeled through-level programming approach. A comprehensive policy for retail electricity pricing is provided. The retailer maximizes its expected profit for a given risk level of profit variability. Additionally, the customers minimize their consumption cost. The proposed optimization problem is solved using MIP. Dynamic pricing approach is used for increasing profit and reducing the retailer risk. This work presents a coordinated and centralized energy management system for optimal grid operation. The optimization is done for Supply Side Management System (SSMS), HEMS and Transmission Line Management System (TLMS). Objective function is implemented to reduce the SSMS and HEMS electrical energy 21

cost (Monyei, Viriri, Adewumi, Davidson, & Akinyele, 2018). Externally Constrained GA (ExC-GA) is used to vary DR load intelligently. Day-ahead and time of use pricing scheme are used for comparison. 2.4 PRICING SCHEMES BASED TECHNIQUES Energy management in the SG was presented in (Derakhshan, Shayanfar, & Kazemi, 2016) using TLBO and Shuffled Frog Leaping (SFL). Four scenarios were considered with different pricing schemes such as TOU, RTP, CCP and the last one without n electricity tariff. Scheduling was performed by varying electricity prices in different intervals. In this research paper the authors proposed a novel pricing scheme for MG. To understand consumers load profile and then assign real-time prices based on their load profile pattern, the MG deploy clustering techniques (Liu, Mahmoudi, & Chen, 2018). To cluster load curve of consumers k-mean algorithm are used. The load profile of each customers is determined in each clusters, in an optimal number of clusters. When the load profile of each cluster is find, real-time prices are assigned to each cluster customers having the best price in than cluster. 2.4.1 Combined Pricing Model Based Techniques An appliance scheduling model is suggested in (Shirazi, & Jadid, 2015), for cost reduction and PAR minimization using GA for domestic consumers. RTP and IBR signals are used for electrical energy cost assessment in this model. Simulation results demonstrate that the proposed model attained the objectives for a single consumer, as well as multiple consumers. In (Rahim et al., 2016), the authors proposed an HEMC by using heuristic algorithms such as BPSO, GA and ACO. A generic architecture is proposed for DSM model. They formulated the problem via multiple knapsack problem (MKP). They used a combined model of ToU and IBR. The GA centered EMC is more efficient than BPSO and ACO in term of cost saving, PAR minimization and UC

22

maximization. The cost and PAR are reduced in this work; however, Average Waiting Time (AWT) is not considered. In (Deconinck, & Decroix, 2009), the authors presented an efficient HEMS for DSM in residential area. They used the combination of two pricing schemes for cost calculation: ToU and RTP. To minimize peak creation, GA is used in this work. Simulation results demonstrate that the combination of TOU and RTP is favourable for electricity cost and PAR reduction. However, there subsists a trade-off between the UC and energy cost. 2.5 RENEWABLE ENERGY SOURCES INTEGRATION Energy cost reduction in the SG was targeted in (Mary, & Rajarajeswari, 2014) using GA via integration of RES and storage units, i.e., batteries. The stored electrical energy is used in specific time intervals when the electrical energy price or demand is high. Charging and discharging bounds of the batteries are assumed by the authors, and also, a controller is installed to avoid any damage of batteries. On the other hand, the authors pay no attention to the installation cost and repairs expenses of RES and storage devices. The authors in (Bharathi, Rekha, & Vijayakumar, 2017) proposed a method to balance the electrical load in industrial, commercial and domestic areas. They compare the electricity consumption of various datasets through DSM without GA and DSM with GA (GA-DSM). The ample simulation results show that the proposed approach GA-DSM achieved the objective: decrease in electricity consumption, which is 21.91% during peak hours having high price. However, the authors did not talk over PAR and UC. Another electricity cost minimization scheme is offered by (Samadi, Wong, & Schober, 2016). For shifting the household appliances in different time intervals dynamic programming is used, and for the interaction of the user with surplus electricity generation, game theory is implemented. According to this research work, inhabitants generate electricity from RES for their household usage, and surplus

23

produced electricity is sold to the neighbors or utility company. The authors disregarded the installation and repairs cost of RES. In (Ahmad et al., 2017), an optimized energy management system (OEMS) is proposed including RESs integration and ESS. They also discussed the residential sector. They used the MKP for problem formulation. To resolve the problem of electricity cost and PAR reduction they used BPSO, GA, WDO, BFOA and hybrid of GA and PSO named GA-PSO (HGPO) algorithms. By the integration of RES and ESS, they achieved 19.94% and 21.55% cost and PAR reduction respectively. Moreover, they achieved 25.12% and 24.88%, bill and PAR reduction respectively by implementing HGPO algorithm. This research did not discuss the AWT for maximizing UC. The authors proposed a HEM model in (Javaid et al., 2017). They used ToU pricing scheme with RESs and without RESs. They used evolutionary algorithms (EV) based BPSO, GA and CSA for DSM model for scheduling the appliances. They used ToU pricing scheme and consider old-style homes, smart homes and smart homes with RESs. The cost saving by CSA is 6.93% and 43.10% with and without RESs in comparison to GA and BPSO respectively. The authors reduced cost using HEMS; however, they did not consider UC. In (Ippolito, Telaretti, Zizzo, & Graditi, 2013), a controller strategy is proposed which acts as a controller between grid and PV or wind generators with battery storage system. The connection device provides the ancillary services. In this research paper, the authors proposed a three step control strategy. The methodology is used to manage the collaboration between RESs and DERs, keeping in view the use of domestic energy. The main objectives of this study are: UC, peak shaving and forming a virtual power plants (Molderink, Bakker, Bosman, Hurink, & Smit, 2010). The authors developed a MG consisting of storage, renewable generation and residential buildings. The building require heating/cooling. An optimization model for MG is developed based on MINLP. The optimization objectives are thermal

24

comfort and economy (Liu et al., 2018). Results show that the proposed approach effectively lower the operating cost and also ensure the thermal comfort level of the customers. 2.6 MICRO GRID BASED TECHNIQUES Authors present and design a distributed EMS (DEMS) for the best operation of MG. They considered the problem as an optimum power flow problem (Shi, Xie, Chu, & Gadh, 2014). In this model, the MG central controller and the local controller compute an optimal schedule. They applied the proposed distributed EMS to a real MG consisting of solar, wind turbines, diesel generators and a BBS. They tested the distributed EMS together in islanded and grid connected mode and shown that their proposed techniques converges fast. The authors did not consider UC optimization. In (Liu, Chen, & Yuan, 2015), authors propose a hybrid energy MG model and discuss energy scheduling problem. Their model consists of solar, wind power, combined heat energy storage system and Electric Vehicle (EV). The objective function is cost optimization; which includes operational, gas, electric power, storage and EV charging discharging cost reduction. They proposed a Multi-team PSO (MTPSO) and units, groups and swarm information are used to update velocity. MTPSO has stable conversion as compared to PSO. However, the authors did not consider UC. The authors proposed a residential MG consisting of RESs in (Corchero, Cruz, & Heredia, 2014). To obtain an efficient and realistic management, the domestic load is divided into three different types. They introduced the anxiety range concepts for consumers behavior. The designed model makes a schedule for all components of the MG when operating day-ahead and results show daily cost saving of 10%. The authors discuss the TOU based EMS along with ESS integration in (Graditi, Ippolito, Telaretti, & Zizzo, 2016). The economic and technical evaluation is carried out using different battery technologies. Their experimental results show that the integration of ESS with TOU significantly reduce the cost.

25

A MILP based HEM scheme was proposed in (Aslam et al., 2018) for electric cost and imported load reduction from external grid. They integrated MG which comprises of wind turbine and solar panel with EV (mobile storage). Simulation results illustrate that their proposed scheme decreases the total cost and imported load. In (Boopathy, & Sivakumar, 2014), authors proposed an autonomous hybrid power system (HPS), they used supervisory control and data acquisition (SCADA) for this purpose. In autonomous HPS they integrated diesel generation with wind and solar power which increases the availability of power. Solar generates DC electricity which is converted to AC via inverter. However, they did not consider minimization of AWT for maximum UC. The authors considered locational pricing in the market for power losses (Menshari, Salehi, & Ghiamy, 2018). This work considered domestic market and upstream market modeling to play a dual role. However, to maximize social welfare, market will simultaneously implement active and reactive power respectively. For modeling the uncertainties in power generation, Weibull probability density function (PDF) and Beta methods are used. To determine the system prices, locational marginal pricing methods (LMP) are used. The problem is optimized using intelligent GA-PSO hybrid algorithm. The

authors

discussed

a

statistical

model

for

electric

vehicle

charging/discharging behavior. The behaviors is described considering the randomness in the initial state of charge (SOC) in electric vehicle batteries (Cai, 2013). The authors used serial quadratic programming (SQP) for the optimal charging/discharging schedules of EV. The day-ahead scheduling is used for power generation and load demand. In grid-connected operation mode, the network loss and in islanded operation mode the required ESS capacity is decreased respectively. The authors proposed MG Energy Management Distributed Optimization Algorithm (MEM-DOA) for efficient energy management (Longe, Ouahada, Rimer, Ferreira, & Han, 2017). According to consumer type this algorithm is used in a distributed fashion in the network. Each customers optimizes the trading and energy

26

consumption for profit and discomfort. The proposed model gave better user satisfaction, higher cost savings and reduced PAR demand on the utility. Reduced investment on peak power plants, grid reliability and environmental benefits are some other advantages. This work proposes an optimal EMS and a control algorithm for grid-connected MG (El-Hendawi, Gabbar, El-Saady, & Ibrahim, 2018). The proposed algorithm minimize the operating cost of the MG. The MG is made up of PV, wind turbine and a BSS. To determine the optimal hourly scheduling for the MG, Interior Search Algorithm (ISA) is used. The control system of the MG includes three stages: local control, supervisory control and EMS. The proposed model was tested on a real case study with various load conditions. In MG operation economic dispatch is a central problem which efficiently schedule various energy resources. The resources are scheduled along these lines to satisfy the energy demand, while minimize the operating cost of MG (Zhang, Hajiesmaili, Cai, Chen, & Zhu, 2018). The authors proposed a peak-aware online dispatch algorithm to solve three challenges. It is proved that the proposed deterministic and randomize algorithm perform better as compare to existing algorithms. A small-scale power dispatch MG is proposed in Colombian area for power supply. An energy management system is used to generate power dispatch, based on MILP (Logenthiran, Srinivasan, & Vanessa, 2014; Henao, Saavedra, & Ramos, 2018). The MG minimizes the cost of operating the grid in the given constrains. The optimization problem is solved using energy management system and MILP. The MG consist of PV, wind, diesel generators, battery bank and domestic load. The power dispatch obtained with heuristic algorithm and with proposed solution is compared. The authors proposed an isolated MG consisting of RESs, typical load, and a Hybrid Energy Storage System (HESS) composed of batteries. To optimize the HESS capacity, a Quantum-behaved PSO (QPSO) algorithm is proposed. Compensation power is corrected considering the rated power of each energy storage (Wang et al., 2018). In addition, a mathematical model is derived for minimizing the daily cost of HESS. This work take an isolated MG in China to validate the 27

effectiveness of the proposed scheme. A comparison is made between QPSO and traditional PSO, which shows that QPSO can find the optimal solution very faster as traditional PSO. The HESS daily cost is also low in case of QPSO. 2.7

LOAD

BALANCING

AND

PEAK

LOAD

SHAVING

BASED

TECHNIQUES Load shifting and balancing were the main objectives of (Logenthiran, Srinivasan, & Vanessa, 2014). For this purpose, a multi-agent system was designed in which each consumer acted as a single representative. The electrical load for each agent was distributed into time frames. Power was transmitted for each agent according to a specific time frame. Domestic, commercial and industrial users were considered in this research paper. In (Islam, Das, Ghosh, Roy, & Suganthan, 2012), the rudimentary concepts of the EDE technique were discussed. An enhanced version of differential evolution was used in which five test vectors were used in place of one trial vector. A new populace was generated using three arbitrary vectors. In EDE, the modified vector was produced, and then, trial vectors were created by comparing fitness with target and mutant to bring up-to-date the population. The Differential Evolution (DE) algorithm was presented in (Li, & Zhang, 2009) with different crossover and mutation approaches in order to improve the performance of the algorithm for scheduling resolve. In contrary to (Islam, Das, Ghosh, Roy, & Suganthan, 2012), a solitary trial vector was used to bring up to date the population. In (Javaid et al., 2017), the authors proposed a mathematical optimization model to control residential energy load and customer preferences. Considering appliance classes, customer preferences and weather conditions, they modeled the customer comfort. For UC and electricity cost they used WDO while for the electricity cost and PAR reduction they used knapsack based problem (K-WDO). Their simulation results show that they achieved optimized results of electricity cost, PAR and UC. The authors control residential load and minimize cost and PAR using WDO and KWDO but ignore UC.

28

The scheduling of appliances operation is described (Qayyum et al., 2015). They proposed a mixed integer programming optimization (MIPO) algorithm. They minimize the peak load and cost reduction. They used the branch and bound (BAB) algorithm for problem formulation and solution. They used PV as MG and export electricity to grid back. This work did not consider PAR reduction. 2.8 APPLIANCES SCHEDULING BASED TECHNIQUES Scheduling home appliances optimally for cost reduction and balancing load between electricity supply and demand is an exciting research issue targeted by the research communal. In the past few years, many methods have been proposed for price minimization, PAR reduction and UC extension. (Zhang, Evangelisti, Lettieri, & Papageorgiou, 2016) proposed a MILP model in for reduction in energy consumption, PAR and energy cost with the incorporation of RESs. Contrary wise, the energy optimization and energy cost reduction strategy is adopted for residential consumers using the GA by (Mohamed, & Koivo, 2012). A heuristic-based model is presented in (Javaid et al., 2017) to tackle the energy optimization problem in a residential area. The simulation results indicate the effectiveness of their proposed model in terms of electricity expense and PAR reduction by scheduling the home appliances in a given time interval. In (Liu, Chen, & Yuan, 2015), authors propose a hybrid energy generation system (HEGS) and discuss appliances scheduling problem. Their model consists of PV, wind turbine, combined with heat and power energy storage and electric vehicle (EV). They reduce cost optimization which includes minimum total operation cost, cost of gas consumption, power purchased from the electric grid, storage system cost, and charging-discharging costs of EV. They have proposed an efficient algorithm i.e., MTPSO which uses different information to update velocity. MTPSO is more stable as compared to PSO. This work reduces cost efficiently but ignores UC, as there is trade-off cost is reduced more UC will be compromised. In (Zhu, Lauri, Koukam, & Hilaire, 2015), the authors used Cooperative PSO (CPSO) to optimize scheduling and operation of appliances. They considered two 29

types of appliances i.e., time shiftable and power shiftable. They achieved reduced electricity cost, maximize UC and balanced the total load on the main grid. This work reduces cost efficiently using CPSO but did not consider PAR and AWT. The authors in (Yoo et al., 2012) describe the scheduling of home appliances. Their objectives is to optimize electricity consumption pattern. For appliances and the RESs scheduling MILP is investigated. They generates electricity locally from RESs and it reduces the electricity cost and the excess energy generated is traded back to the commercial grid to further minimize the electricity cost. Although, RESs combined with HEMS is useful for both the consumers and the utility, however, the installation cost of RESs is expensive for a single home or consumer. 2.9 USER PREFERENCES BASED TECHNIQUES Authors proposed different DSM programs (Javaid et al., 2017). They used the TLBO, GA, the enhanced differential evolution (EDE) algorithm and the proposed enhanced differential teaching-learning algorithm (EDTLA) to manage energy consumption and UC, while taking into consideration the human preferences and energy consumption pattern. They have considered the power consumption pattern for shiftable appliances to get monitory benefits. They have considered cost reduction, UC, reduce carbon emission and RESs integration. They also considered PAR reduction and the trade-off between cost and UC. Without RESs the electricity cost and PAR are reduced up to 36% and 43%, respectively. With RESs electricity cost, PAR and carbon emissions reduction is up to 67%, 29% and 55%, respectively. The author’s considered reduction of cost and carbon emissions but ignored AWT. The authors discusses real-time information based energy management algorithm to reduce cost, PAR while keeping the UC (Rasheed et al., 2016). They classified the appliances into different categories. They considered customer preferences, cost saving and UC. Air conditioner and refrigerator are modeled using Intelligent Programming Communication Thermostat (IPCT). This work proposed a new Decision Support and Management System (DSEMS) keeping in view the residential load consumption (Siano, Graditi, Atrigna, & Piccolo, 30

2013). The designed system acts as Finite State Machine (FSM). The FSM consists of different scenarios based on consumers preferences. 2.10 ELECTRICITY PRICE REDUCTION TECHNIQUES In (Ullah et al., 2015), a new scheduling model is proposed for bill minimization and PAR reduction in a domestic area using heuristic techniques. BPSO and GA are used for optimization, and simulation results explain the effectiveness of the proposed model. Energy consumption in the residential sector plays an important role in reducing electricity prices and its consumption. Considering this fact, residential appliances were scheduled using an Artificial Neural Network and GA (ANN-GA) scheme in (Ghasemi, Shayeghi, Moradzadeh, & Nooshyar, 2016). The scheduling was performed on a single home with four bedrooms on a weekly basis. The objective was to find the best possible solutions out of the total search space. Results were obtained with a 25%, 40% and 10% reduction in electricity usage from the grid. However, multiple OTIs were not taken into consideration Aslam et al. (2017) proposed a cost-efficient scheme using cuckoo search and GA. Their proposed scheme efficiently minimized the electricity cost but UC is not taken into account. The authors proposed a Generic DSM (G-DSM) model for residential users (Khan, Javaid, Mahmood, Khan, & Alrajeh, 2015). They consider reducing PAR, electricity cost and appliances WT. They use GA for appliances scheduling and consider 20 users. They obtained 39.39% and 45.85% cost reduction for single and 20 users respectively. The PAR reduction for a single user and 20 users are 17.17% and 45.24%, respectively. The bill reduction in cents on day-to-day basis is 25.62%. They discussed the correlation between price and waiting time of appliances. The authors reduced cost and PAR but UC is ignored. An opportunistic scheduling algorithm is discussed in (Rasheed et al., 2016). They used RTP scheme and Optimal Stopping Rule (OSR). They have assigned

31

priority to consumers based on energy consumption pattern. They used First Come First Serve (FCFS) algorithm to reduce cost and WT for appliances. They have used Priority Enabled Early Deadline First (PEEDF) to maximize the UC. FCFS save 65.95% cost while MFCFS save 42.58% cost which is 23.34% less than FCFS. Moreover, PEEDF save cost up to 48.28% which is 5.7% more than FCFS. They also used RE during peak hours and sell it back to the grid, when the energy is surplus. The authors used different algorithms to reduce cost and AWT but ignored the UC. To optimize electricity cost and carbon dioxide emission an optimization model is proposed. The authors consider consumers preferences in buildings, equipped with distributed energy resources (DERs) (Pooranian, Abawajy, & Conti, 2018). The DER operation and controllable and uncontrollable appliances are scheduled. The authors used real-time pricing scheme and reduce peak demand on the grid. The optimization problem is formulated using MILP technique in multiple smart homes. The load is scheduled keeping in view the environmental and economic perspective. The excessive experiments shows that the proposed approach reduced carbon dioxide emission and energy cost. 2.11 PEAK-TO-AVERAGE RATIO AND USER COMFORT ORIENTED TECHNIQUES In (Javaid et al., 2017), authors have discussed HEMC for the residential load. They used four algorithms for the bill and PAR reduction i.e., GA, BPSO, WDO, BFOA and Genetic BPSO (GBPSO). They achieved 34% PAR and 36% cost reduction. They calculated the cost for single, ten and fifty homes. On the other hand, the authors did not consider the UC and consumer preferences. The authors suggested a Smart Energy Hub (SEH) and modern energy management technique considering electricity and natural gas consumption. They formulated the SEH as a Non-cooperative Game (NCG) (Sheikhi, Rayati, Bahrami, & Ranjbar, 2015). They also proved the Nash Equilibrium (NE). They obtained cost

32

and PAR minimization by the proposed SEH. The authors use NCG to reduce cost and PAR but ignore the UC. In (Mahmood et al., 2016), authors proposed a realistic scheduling mechanism (RSM) for reducing electricity cost and enhancing appliance utility. They divided a 24 hours’ time horizon into 4 sub-logical time slots of 6 hours each. They used BPSO for appliances utility and cost reduction. They also consider UC. To create a balance between appliance utility and cost-effectiveness, RSM with power bank is proposed that gives UC gain of 0.185 with respect to unscheduled load and 0.149 with respect to BPSO on a scale of 0 to 1. The authors reduce cost and optimize UC but ignore PAR. GA is used to increase intelligence of Conventional Programming Communication Thermostat (CPCT). They considered electrical energy cost, reduction in PAR and maximization of UC. Their proposed algorithms effectively manage the energy utilization by scheduling home appliances. Their proposed algorithm reduced the energy cost and PAR up to 22.63% and 22.77%, respectively. The authors reduced cost and PAR using GA but did not consider AWT. The authors proposed GA based Evolutionary Creative Comfort Algorithm (EACA) to increase UC in three budget scenarios (Khan, Javaid, & Khan, 2018). 2.12 HOME ENERGY MANAGEMENT SYSTEM BASED TECHNIQUES A lot of research has been done on residential load scheduling and DSM for smart homes and HEM. Some work has been cited below. (Huang, Wang, Guo, Kang, & Wu, 2016) offered a DR scheme to optimize the operation of the home appliances in the HEMS. To reduce the computational load burden, Two Point Estimation Method (2PEM)-embedded PSO-based approach was proposed. They compare the proposed scheme, gradient-based PSO-2PEM, results with the GPSO-Latin Hypercube Sampling (LHS) method’s results. However, the offered scheme was found to be more effective compared to GPSO-LHS in terms of reduced computational load burden. However, the authors did not address the PAR and electricity cost, which is paid by the residents.

33

In (Han, Choi, Park, Lee, & Kim, 2014), the authors proposed a smart HEM architecture for the reduction of electrical energy consumption cost and PAR. In the proposed architecture, they also consider a PV panel for smart electrical energy generation. The ZigBee model is used for the assessment of home appliances; this model gathers the information from home appliances and transfers it to the home server. A Power Line Communication (PLC) model is used for PV panel assessment and maintenance; which gathers the information about the PV panel (e.g., forecasting about electricity generation, temperature or maintenance problem) and transmits it to the home server. Simulation results show the effectiveness of the proposed architecture according to the defined objectives. The authors in (Hassan et al., 2017) propose a HEMS for cost reduction and load balancing. They evaluate the performance of HEMS by using GWO and BFOA. They reduce cost and PAR by dividing the appliances into two classes based on energy consumption pattern. They consider peak and off-peak hours to manage energy consumption. They used critical peak pricing scheme and achieved 10% more cost reduction by GWO as compared to BFA. In this work the bill and PAR is reduced but UC is overlooked. Beaudin & Zareipour (2015) provided a comprehensive review of HEMS. HEMS is an efficient tool for shifting and reducing the energy production and consumption of a residential area. HEMS plays an important role for DR. By considering multiple objectives i.e., UC, energy costs, load profile and environmental concerns, HEMS creates an optimal energy consumption schedule for home appliances. A Stochastic Programming Model (SPM) has been presented in (Somma, Graditi, Heydarian-Forushani, Shafie-Khah & Siano, 2018). The authors considered the environmental and economic aspects using RESs. Using Monte Carlo method and roulette wheel mechanism the uncertain parameters are modeled for 24 hours duration considering demand and supply. The authors formulated the optimization problem as a stochastic multi-objective linear programming problem.

34

The authors in (Erdinc, Paterakis, Pappi, Bakirtzis, & Catalao, 2015) proposed a framework for the HEMS modeling and techno-economical sizing using MILP. The sizing of additional DG and BBS are discussed for smart home appliances. They investigated the DR activities for daily energy consumption demand profile as compared to normal daily energy consumption profile of household appliances. They have focused on decreasing cost, varying load and distributed generation profile for different seasons for DG and BBS. They have also considered different sensitivity analysis keeping in view the impact of variations of economic input for the provided model for a long-term analysis. However, the authors did not considered minimization of AWT for UC maximization. This paper presents an efficient strategy for charging control of vehicle batteries for optimizing the charging/discharging process (Ito et al., 2018). This technique used the information from home power load and vehicle state for use of domestic purpose. Dynamic programming and semi-Markov model is used to develop future prediction algorithm. The prediction algorithm find the future vehicle state. This research work presents a new approach for the representation of reliability and flexibility of distributed energy resources. To represent the devices that acts as a surrogate model, ANN are used. The advantage of this approach is the arbitrary energy flexibility (Forderer, Ahrens, Bao, Mauser, & Schmeck, 2018). 2.13 ENERGY STORAGE SYSTEM BASED TECHNIQUES Power storage in the case of sensitive load demand from the consumer was discussed in (Soares, Ghazvini, Vale, & Moura, 2016). Virtual power play helped to schedule load and energy consumption ahead of time. For distributed generation of electricity, load demand increased simultaneously. To cater to the intensive load demand, the Vehicle to Grid (V2G) strategy was used. BPSO along with MILP was proposed for balancing the load among multiple distribution units. It involved a complex mathematical formulation to design a model for day-ahead electricity price forecasting. One hundred and eighty distributed units along with 1000 electric vehicle stations were used to perform this evaluation on an experimental basis.

35

The authors of (Wu, Tazvinga, & Xia, 2015) proposed an optimal energy management model for a grid-connected solar power and battery hybrid system is discussed. Their model optimizes the electricity cost while keeping in view the constraints like power balance, solar output and battery capacity limits. They use the open and closed loop method to dispatch the power flow in real-time based on uncertain distributions. These two methods lead to great cost saving and robust control performance. Furthermore, the authors did not consider UC. A power scheduling problem with RESs and energy storage is investigated in (Isikman, Yildirim, Altun, Uludag, & Tavli, 2013). They have prioritized the appliances into five classes and proposed a novel formulation and solution for this model using MIP. The MIP makes the problem more complex and cannot handle a large number of appliances. However, they also ignored the UC. The author’s presents a sizing and control strategy of battery-operated energy storage systems for PV generation farm (Brenna, Foiadelli, Longo, & Zaninelli, 2016). The BESS is dispatched for ahead and day-ahead markets. A predictive model is developed for forecasting solar irradiation and load power consumption. The model is established on feed-forward neural network. The neural network is educated with the Gutenberg–Margaret Back-Propagation Learning Algorithm (GMBPLA). The authors proposed a Real-time Decentralized DSM (RDCDSM) for adjusting the real-time domestic load (Tushar, Zeineddine, & Assi, 2018). Based on predicted customer aggregate load, day-ahead energy generation is preplanned. At the time of energy consumption, a deviation from the predicted demand results in additional cost of the deviated customers. To formulate our problem, game theory with mixed strategy is implemented for cost reduction of customers. To avoid demand deviation, customers used RESs and ESSs. RDCDSM helps to reduce uncertainties in renewable integration. In this way, the power quality is increased. In this work, the authors proposed an optimal operation policy for a hybrid energy storage scheme, which is hierarchical in nature (Yang, Yang, Tsai, Chen, & Chen, 2015). This strategy can be used in distribution network for voltage regulations, price arbitrage and PV power smoothing with high PV penetration. To 36

smooth power fluctuation, fuzzy logic is used. A coordinated control comprised of centralized and local control for lithium-ion battery is proposed. The coordinated control perform price abridge and voltage regulation. The authors discussed two area: design of a BESS and power management of the MG. MG Central Controller (MGCC) performs power management (Salas, Marzal, Gonzalez, Figueres, & Garcera, 2018). The key goals of the power management unit is to reduce the cost of the electricity imported from the grid and to maximize the battery life following proper procedure for battery charging/discharging. In addition, the authors suggested a Power Management Algorithm (PMA) for the MG. 2.14

MULTIPLE

KNAPSACK

PROBLEM

FOR

PROBLEM

FORMULATION A general cost model was proposed by (Ogwumike, Short, & Abugchem, 2015). Some problems were solved through an Intelligent Decision Support System (IDSS). IDSS linked to AMI provides excellent communication between the customer and utility. WDO with Knapsack (K-WDO) was implemented to attain the objective function of cost reduction and maximizing UC (Rasheed et al., 2015). Min-max constraints were defined for K-WDO. Appliances were classified according to their category and power ratings. TOU was used as the electrical energy pricing signal for on-peak and off-peak hours in agreement with consumer preference; however, PAR and system complexity were compromised. The authors in (Mahmood, Javaid, Khan, & Razzaq, 2016) had further incorporated an optimization technique using GA and PSO for electrical energy cost and PAR minimization. The MKP is used for the problem formulation, and three different pricing signals ToU, RTP and CPP are examined for electricity cost estimation. Simulation results validate the effectiveness of their proposed scheme, and GA performs superior to the other counterparts.

37

In (Rasheed et al., 2016), authors propose an energy optimization technique. They scheduled household appliances to find the electricity prices, weather conditions and dynamic behaviors of users. They considered cost and UC optimization. They solved their objective function via MKP. They obtained energy savings by 11.77% and 5.91%, with and without people occupancy respectively. The authors minimized cost but AWT is not considered. In (Mahmood et al., 2016), authors have proposed a HEM architecture (HEMA). They integrate multiple appliance scheduling and load classification. Their model contains of six layers and validated by simulation using Matlab. They used knapsack optimization technique for appliances scheduling. They considered four cases of appliances cost reduction. They have also considered fault identification and electricity theft control. They have also calculated carbon footprints for user awareness. Their simulation results show that peak load decrease of 22.9% for an unscheduled electrical load with person presence controller (PPC), 23.15% for the scheduled load with PPC and 25.56% for the scheduled load with UC index. In the same way, total cost reduction of 23.11%, 24% and 25.7% has been observed, respectively. The authors reduced cost and carbon emissions but did not consider AWT. 2.15 ENERGY SHARING AND TRADING Authors have proposed a DEMS called an Incentive-Driven Distributed Energy Sharing System (iDES). They considered the reduction of communication overhead of appliances and make sure effective load sharing among different homes appliances in (Zhong et al., 2014). They have used a new pricing scheme for different incentives. The load sharing price means the cost of renewable and storage system, changes of power supply-demand and the remaining energy level of the battery storage system. The authors efficiently shared energy and reduce carbon footprints but ignored the UC. In (Mahmood et al., 2017), the authors propose a multi-agent power distribution hub. The proposed hub optimize energy consumption and management of ESS. They share the power among neighbors with no profit and loss. They considered on-peak,

38

off-peak and mid-peak prices. They considered three scenarios for their model i.e., without BESS, with BESS and with sharing the power of BESS. 21% and 6% power savings are achieved for baseline and without sharing ESS consumption from the utility. 36% of electricity cost is reduced if compared with baseline cost and 9% for sharing of BESS. The authors reduced cost using ESS and multi-agent concept but ignored the UC. In (Inam, Strawser, Afridi, Ram, & Perreault, 2015), the authors introduce the concept of peer-to-peer energy sharing by those who can afford renewable and nonrenewable electricity generation sources such as solar panel, generator and a windmill to those who cannot afford such sources of renewable energy generation or lack access to main grid. This will create a marketplace for electricity and selfsufficiency in power market by people itself. Up to this end, an ad hoc MG is introduced using a power management unit (PMU). However, the authors did not reduce PAR. The authors used game theoretic approach for energy management (Tushar et al., 2018). P2P energy trading is an efficient and feasible solution for energy management. The authors discussed some features and testbed of P2P. Energy trading among peers, i.e., P2P has arisen as a next-generation energy supervision strategy for the SG (Tushar et al., 2018). Hence, enable each prosumers to participate with other prosumers of energy and grid in power trading. The decision process of the participants is very difficult especially consumers with different interests and motivating the consumers to participate in energy trading process. To ensure an efficient operation of electric power grid, such a decision making method need to be built on solid statistical and mathematical principles. When electric network and thermal network integrates into Community Energy Internet (CEI) with Combined Heat and Power (CHP) to improve the energy system economically (Liu, Guo, Liu, & Wang, 2018). An energy sharing network is proposed for prosumers, equipped with Heat Pump (HP) and PV-Thermal (PVT). In addition, a social welfare optimization model is introduced, which consists of CHP

39

system, utility grid and PVT-HP prosumers. The social welfare optimization problem is solved using language multiplier method. A distributed algorithm is also used for problem solution. Six residential buildings are used in the study. Table 2.1: Summarized literature review Technique(s)

Objective (s)

Finding (s)

Limitation (s)

BPSO, GA, WDO,

Cost and PAR

Achieved 25.12%

AWT is not

BFOA, HGPO,

reduction

and 24.88%, bill and

considered

MKP (Ahmad et al.,

PAR reduction

2017)

by HGPO

WDO, K-WDO

Control

Achieved optimized

UC is not

(Javaid et al., 2017)

residential load

results of cost and

considered

PAR

BPSO and neuro

Electricity price

Electricity market

Trade-off

fuzzy logic

forecasting

price forecasting

between

(Shayeghi,

by a hybrid

using mean absolute

MAPE and

Ghasemi,

evolutionary-

percentage error

computational

Moradzadeh,

adaptive

(MAPE)

time

& Nooshyar, 2015)

methodology

IPSO (Yang, Yang,

Peak load

They achieve

UC is

Tsai, Chen, &

reduction

desired objectives

compromised

Chen, 2015)

and only passive appliances is considered.

PSO and GA

To minimize

Cost is reduced

To reduce cost

(Peyvandi, Zafarani,

electricity cost

using PSO and GA

UC is

& Nasr, 2011)

compromised 40

MILP (Bradac,

Optimal

Kaczmarczyk, &

scheduling of

Fiedler, 2014)

domestic

Minimize cost

Ignored UC

appliances

Greedy algorithm

Heuristic

Minimized user

PAR is

(Ogwumike, Short,

optimization of

frustration

ignored,

& Abugchem, 2015)

Customer

and cost

System

electrical energy

complexity

costs using a

increased

generic cost model

Multi-agent Model

Demand-side

Load shifting and

User comfort

(Logenthiran,

management

cost minimization

is

Srinivasan, &

of the SG: load

Vanessa, 2014)

shifting and

compromised

incentives

BPSO and neuro

Electricity price

Electricity market

Trade-off

Fuzzy logic

forecasting

price forecasting

between

(Shayeghi, Ghasemi,

by a hybrid

using mean absolute

MAPE and

Moradzadeh, &

evolutionary-

percentage error

computational

Nooshyar, 2015)

adaptive

(MAPE)

time

methodology

Multiple input

Instantaneous

Load and price

Real-time

multiple output

day-ahead

signal forecasting

forecasting is

model (MIMO)

forecasting of

(Derakhshan,

electrical energy

not considered

Shayanfar, &

41

Kazemi, 2016)

price and load in SGs

TLBO and shuffled

The optimization

frog leaping (SFL)

of DR programs

(Soares, Ghazvini,

in SGs

Cost optimization

RES not integrated

Vale, & Moura, 2016)

PSO and MILP

A multiobjective

Virtual power play

Requirements

(Ghasemi, Shayeghi,

model for the

Scheduling

of the

Moradzadeh, &

next-day energy

customers for

Nooshyar, 2016)

reserve

reliable power

scheduling of a

grids is not

SG with great

considered

penetration of sensitive loads

MIMO (Jayabarathi,

New hybrid

Electricity price and

Computational

Raghunathan,

technique for

load forecasting

time is not

Adarsh, &

electrical energy

Suganthan,

price and load

2016)

forecasting in

practical

SGs with demand-side management

GWO (Huang,

Economic

Solving non-linear

The user has

Wang, Guo,

dispatch using

economic load

to come up

Kang, & Wu,

the hybrid gray

dispatch problems

with ways of

2016)

wolf optimizer

42

handling the constraints

ANN-GA

ANN-GA smart

Reduction in grid

Cannot be

(Setlhaolo, & Xia,

appliance

energy usage

applied to

2014)

scheduling for

different

optimized

building

energy

types

management in

involving a

the domestic

higher number

sector

of appliances

Game theory

A DR modeling

Scheduling the

RES not

algorithm (GTFT)

for residential

appliances using the

integrated

(Logenthiran,

consumers in a

game theory strategy

Srinivasan, &

SG environment

and PAR reduction

Vanessa, 2014)

using a game theory-based energy scheduling algorithm

MILP and heuristic

Load scheduling

Load balancing

Cost

algorithms (Agnetis,

for household

minimization

Pascale, Detti, &

energy

is not

Vicino, 2013)

consumption

considered

optimization

MINLP (Setlhaolo,

Optimal

Cost reduction is

PAR is

& Xia, 2014)

scheduling of

achieved

ignored

43

Household appliances for DR

GWO and ILP

Grey wolf

Load dispatching in

Solved

(Pradhan, Roy,

optimization

off-peak hours

economic

& Pal, 2016)

applied to

load dispatch

economic load

(ELD)

dispatch

problems in

problems

the current study

MILP and heuristic

Load balancing

Their proposed

Electricity

algorithm (Agnetis,

scheme balances the

cost neglected

Pascale, Detti, &

load

Vicino, 2013)

MINLP (Setlhaolo,

Cost reduction

& Xia, 2014)

Residents enjoy a

They ignore

lesser cost

the PAR

through this scheme

GA, PSO and

To reduce cost

Beneficial model for

Neglected the

MKP (Mahmood,

and PAR

electricity cost and

UC

Javaid, Khan, &

PAR reduction

Razzaq, 2016)

GA (Mary, &

To decrease

Simulations

Ignored the

Rajarajeswari,

electricity

represent that

installation

2014)

cost and PAR

they minimized the

cost of RES

cost and PAR

44

GA-DSM

To reduce cost

(Jayabarathi,

User enjoys a lesser

Neglected the

cost with this model

PAR and UC

Raghunathan, Adarsh, & Suganthan, 2016)

PSO (Huang,

Computational

Simulation

Ignored the

Wang, Guo,

load burden

represents that

electricity

Kang, & Wu,

reduced

they minimized the

cost and PAR

2016)

computational load burden

ILP (Zhu, Tang,

This approach

They did not

Lambotharan,

efficiently balanced

address the

Chin, & Fan, 2012)

the load

UC

Potentially reduced

Ignored the

the cost

installation

Dynamic

Load balancing

Cost reduction

programming (Samadi, Wong, &

and

Schober, 2016)

maintenance cost of RES

GA, BPSO (Ullah

To decrease

The proposed model

User comfort

et al., 2015)

PAR and

efficiently decrease

refuted

electricity cost

the electricity cost and PAR

GA (Muralitharan,

To reduce cost

The proposed

User comfort

Sakthivel, & Shi,

and PAR

scheme minimized

ignored

2016)

the user bill and PAR

45

GA, BPSO, WDO,

Bill and PAR

Achieved 34% PAR

UP is not

GBPSO, BFA

reduction

and 36% cost

considered

(Javaid et al., 2017)

reduction

MIPO, BAB

PAR and cost

PAR and cost

PAR is

(Qayyum et al.,

reduction

reduction

ignored

GA (Khan, Javaid,

PAR and

The PAR and cost

UC is not

Mahmood, Khan,

electricity cost

are reduced by

considered

& Alrajeh, 2015)

reduction

17.17% and 25.62%.

MKP (Rasheed

Cost and UC

Obtained energy

AWT is not

et al., 2016)

Optimization

savings by 11.77%

considered

2015)

and 5.91%.

GWO and BFOA

Cost and PAR

Achieved 10% more

(Hassan et al., 2017)

reduction

cost reduction by

UC is ignored

GWO as compared to BFA

MKP (Mahmood

Cost and carbon

Peak load reduction

AWT not

et al., 2016)

footprints

of 22.9%, 23.15%,

considered

reduction

25.56% and total cost reduction of 23.11%, 24% and 25.7% are achieve respectively.

46

BPSO (Mahmood

User comfort

User comfort gain of

PAR is not

et al., 2016)

and cost

0.185 and 0.149 on a considered

reduction

scale of 0 to 1

TLBO,GA, EDE,

Cost and carbon

Cost, PAR and

AWT is not

EDTLA (Javaid

emission

carbon emissions

considered

et al., 2017)

reduction

reduction is up to 67%, 29% and 55%, respectively.

MTPSO (Liu

Reduction of

Cost reduction

UC is

et al., 2014)

cost

achieved

compromised.

BPSO, GA (Zhong

Energy sharing

Cost minimization

UC is not

et al., 2014)

and carbon

and energy sharing

considered

footprint

efficiency achieved.

reduction

BPSO, GA and

Reduction of

Cost saving is

UC is not

CSA (Javaid

cost

6.93% and 43.10%

considered

et al., 2017)

with and without RES

BPSO, GA, ACO,

Cost and PAR

GA based EMC is

AWT is not

MKP (Rahim

reduction

efficient in-term of

considered

et al., 2016)

cost and PAR reduction

Multi-Agent

Cost reduction

36% of electricity

UC is not

(Mahmood et al.,

using ESS

cost is reduced

considered

2017)

47

FCFS, PEEDF,

Reduce cost and

Cost reduction by

UC is not

MFCFS (Rasheed

AWT

FCFS, MFCFS and

considered

et al., 2016)

PEEDF are 65.95%, 42.58% and 48.28% respectively.

GA (Rasheed et al.,

Reduce cost and

Cost and PAR

AWT is not

2016)

PAR

reduction is 22.63%

considered

and 22.77%, respectively.

CPSO (Zhu, Lauri,

Electricity cost

Cost reduction

PAR and

Koukam, & Hilaire,

reduction

achieved

AWT is not

2015)

considered

NCG (Sheikhi,

Cost and PAR

Cost and PAR is

UC is not

Rayati, Bahrami,

reduction

reduced

considered

GA (Deconinck,

Electricity cost

Cost and PAR is

UC is not

& Decroix, 2009)

and PAR

reduced using RTP

considered

reduction

and TOU

MILP (Yoo et al.,

PAR and cost

Cost is reduced

2012)

reduction with

small scale

RESs integration

residential

& Ranjbar, 2015)

Expensive for

Users

OCM and MPC

Electricity bill

Optimal energy

AWT for UC

(Wu, Tazvinga, &

reduction

management

is not taken

solution and cost

into account

Xia, 2015)

48

saving

MIP (Isikman,

Optimal

Maximizes energy

UC is not

Yildirim, Altun,

scheduling of

utilization

considered

Uludag, & Tavli,

energy resources

2013)

among users

SCADA (Boopathy,

A hybrid power

Design of hybrid

UC is not

& Sivakumar,

generation

power system

considered

2014)

model

MILP (Erdinc,

HEMS modeling

Used single step

UC and cost

Paterakis, Pappi,

and techno-

methodology

reduction are

Bakirtzis, &

economic sizing

to size additional

ignored

Catalao, 2015)

PV and ESS

PSO and GA

To minimize

Cost is reduced

To reduce cost

(Peyvandi, Zafarani,

electricity cost

using PSO and GA

UC is

& Nasr, 2011)

compromised

MILP

Reduction of

CO2 emission and

UC and PAR

(Zakariazadeh,

CO2 emission

cost minimization

are not

& Jadid, 2013)

and cost

PCPM (Shi, Xie,

Design of a

EMS is designed

PAR and UC

Chu, & Gadh,

distributed EMS

using optimal

are not

operation of MGs

addressed

Cost reduction

UC is not

achieved

discussed

considered

2014)

MTPSO (Liu,

Cost reduction

Chen, & Yuan, 2015)

49

IPSO (Yang,

Peak load

They achieve

UC is

Yang, Tsai,

reduction

desired objectives

compromised

Chen, & Chen,

and only

2015)

passive appliances is considered.

PMU (Inam,

To reduce

Cost is reduced by

UC is

Strawser, Afridi,

electricity cost

peer-to-peer

decreased

Ram, & Perreault,

electricity sharing

2015)

GA-PSO (Menshari,

To maximize

Social welfare

uncertainty

Salehi, & Ghiamy,

social welfare

maximized

cost is ignored

Game theory

P2P energy

Energy trading

PAR and cost

(Tushar et al., 2018)

trading to

achieved efficiently

is ignored

2018)

increase UC

GA (Khan,

To increase UC

High UC achieved at Consider fixed

Javaid, & Khan,

a higher budget

2018)

electricity price only

NSGA-11 (Pan,

Load power

High UC achieved at Consider fixed

Liu, Wu, &

consumption

a higher budget

Hao, 2018)

Forecasting

GMBPLA (Brenna,

To reduce cost

electricity price only

Foiadelli, Longo, & Zaninelli, 2016)

50

Load forecasting

BESSs cost is

achieved

not considered

Game theory

Energy trading

The cost is reduced

PAR and

(Tushar, Zeineddine,

BESSs cost is

& Assi, 2018)

not considered

Game theory

Optimal load

(Tushar et al., 2018) demand and

Energy trading

PAR is not

efficiently achieved

considered

generation

Serial quadratic

To increase

Optimal load

Network loss

programming

profit and

demand and

and ESS

(SQP) (Cai, Du,

discomfort

generation achieved

capacity is

Yu, Gao, Littler,

decreased

& Wang, 2013)

MEM-DOA (Longe,

To minimize the

Profit and comfort

Cost increased

Ouahada, Rimer,

operating cost

achieved

with increased

Ferreira, & Han,

comfort

2017)

ISA (El-Hendawi,

Price arbitrage

Operating cost

PAR

Gabbar, El-Saady,

and power

minimized

increased with

& Ibrahim, 2018)

smoothing

decreased comfort

Fuzzy logic (Salas,

Power loss

Power smoothing

Battery cost is

Marzal, Gonzalez,

minimization

achieved

ignored

GA (Jaramillo,

Operating cost

Power loss

PAR

Munoz, Ortiz,

and energy

minimized

increasing

Figueres, & Garcera, 2018)

51

Lopez, &

demand

during power

Albarracin, 2018)

loss minimization

BED and RED

Energy

Operating cost and

Cost increased

(Zhang, Hajiesmaili,

flexibility

energy demand

with energy

satisfied

demand

Cai, Chen, & Zhu, 2018)

satisfaction

ANN (Forderer,

To minimize the

Energy flexibility

PAR and cost

Ahrens, Bao,

operation cost

achieved

is not

Mauser, &

considered

Schmeck, 2018)

MINLP (Fan, Ai, &

To optimize the

Operation cost

PAR not

Piao, 2018)

social welfare

minimized

considered

Graph Theory

To minimize

Social welfare

Cost is

(Liu, Guo, Liu, &

electricity cost

optimized

increased with

Wang, 2018)

social welfare optimized

PMA (Salas,

To minimize

Electricity cost

PAR is

Marzal, Gonzalez,

operating cost

optimized

ignored

MILP (Henao-

To maximize

Operating cost

PAR is

Munoz, Saavedra-

profit and reduce

optimized

ignored

Montes, & Ramos-

risk

Figueres, & Garcera, 2018)

Paja, 2018)

52

MIP (Imani, Zalzar,

To find best

Mosavi, &

price

Profit maximized

PAR and UC is ignored

Shamshirband, 2018)

K-Mean algorithm

To reduce

(Liu, Mahmoudi,

energy cost and

& Chen, 2018)

carbon dioxide

Best price assigned

Only RTP is used

emission

MILP (Pooranian,

To reduce

Energy cost and

UC is

Abawajy, & Conti,

energy cost

carbon dioxide

decreased

2018)

emission reduced

ExC-GA (Monyei,

To reduce

Energy cost reduced

UC is

Viriri, Adewumi,

operating cost

Davidson,&

and optimize

Akinyele, 2018)

thermal comfort

MINLP (Liu et

To reduce MG

Operating cost

PAR is

al., 2018)

operating cost

reduced

compromised

MIP and iSSO

To reduce daily

Operating cost

PAR and

(Zhang, Yeh, Jiang,

cost

reduced

discomfort

compromised

Huang, Xiao, & Li,

Increased

2018)

QPSO (Wang

To reduce

et al., 2018)

electricity cost

Daily cost reduced

Daily PAR increased

53

CHAPTER 3 PROPOSED SYSTEM MODELS AND SOLUTIONS 3.1 INTRODUCTION With reference to the problems identified in section 1.2 in chapter 1. This chapter proposed various solutions regarding the energy optimization in residential area. We model various types of the appliances in a home by categorizing them based on the consumers’ preferences. Various heuristics optimization techniques, i.e., CSA, SA, GWO, EDE, BPSO, WDO, GA and their hybrid techniques are applied for designing the EMCs. These techniques are also applied to optimize the energy consumption patterns of the residential users based on their demands because these techniques gives the best optimal solution within the feasible computational time. Effects of integrating the local MGs and RESs are also tested and verified in various cases for fulfilling the consumers’ requirements and optimization of the cost and comfort of the consumers. We have presented four solutions for the HEM after considering multiple household appliances. 3.2 ENERGY MANAGEMENT OF SMART BUILDINGS EXPLOITING HEURISTIC OPTIMIZATION An overview of the proposed system model for related problems is discussed in section 1.2.1. This research work examines a design of future SGs that targets the reduction of electrical energy costs for customers in a domestic area. The proposed system makes electrically powered grid stable and minimizes the total peak load during electric grid operation. 3.2.1 System Model Efficient and consistent operations in the SG are authorized by DSM. DSM main functions are energy management and demand side control accomplishments for electricity customers. We undertake a system model having an electrical power system with a single utility and a smart building which is comprised of 30 smart residential homes (apartments). The demand of the electricity consumers is met by 54

the electric grid (utility) directly or ESS. Let  denotes the set of all electricity consumers in a building and each consumer u



 has his/her own living pattern,

i.e., LOT and power rating of appliances. In order to calculate the hourly electrical power consumption of each electricity consumer u, smart meters are connected at every residence to communicate the extracted data of consumers’ demand to the utility company. It also links price information from the utility to the electricity consumer. We consider one day for our implementation and the whole day is allocated into 24 operational time slots where each time slot contains of 1 h as described below using Equation (3.1). T = {t1, t2, t3, t4 …t24}

(3.1)

We have categorized the home appliances based on the requirements of the electricity consumers considering deferrable (ad), non-deferrable (and) and base load (ab) appliances where ad, and, ab



An (i.e., An represents the set of all appliances).

We assume that each appliance is capable of connecting with the EMC, which is linked to the Internet. The EMC is an embedded computing platform that schedules the appliances subject to the obtainability of the time slot. Each appliance must complete its runtime period between the original starting time and least finishing time. The execution pattern for each appliance is described in Figure 3.1, where α shows the original starting time, β shows the least finishing time and η nominates the time when the appliance starts its execution. The difference between α and η shows the waiting time for each appliance. The original starting time and the minimum finishing time for each appliance are defined by the electricity customer, explained in Table 3.1. The proposed system model is illustrated in Figure 3.2.

Figure 3.1: Status of appliances’ execution pattern. The categorization of home appliances is done by the electricity consumers according to their requirements, and it can vary from time to time, i.e., in the summer 55

season, the refrigerator is a non-deferrable appliance. However, the operation of the refrigerator is different in the winter season. Furthermore, we discuss in detail the categorization of home appliances and the mathematical formulation of the proposed system in Section 3.2.2. Table 3.1: Parameters of appliances Appliance

Appliance

Power

Earliest

Finishing

LOT

Rating

Starting

Time (h)

(h)

(kWh)

Time (h)

Cooker hub

3–4

6

10

1–3

Cooker oven

4–5

15

20

1–3

Microwave

1.7–2.5

6

10

1–3

Laptop

0.1–0.2

18

24

2–4

Desktop

0.3–0.5

18

24

3–5

Vacuum cleaner

1.2–2

9

17

1–3

Electrical car Dish washer

3.5–5 1.5–2

18 9

8 17

3–5 2–4

Washing machine 1.5–2

7

12

2–3

2.5–3.5 0.84–1

13 16

18 24

1–3 6–8

0.3–0.5

1

24

22–24

Class

Base Load Appliances

Deferrable Appliances

Spin dryer Non-deferrable Interior lighting Appliances

Refrigerator

3.2.2 Classification of Load To assess the performance of our proposed system, we have considered two scenarios in our work. First, we implement our proposed system in a solitary smart home, and then, we examine the performance of our proposed system on multiple smart homes. Each home consists n number of smart appliances with different categories. These smart appliances are scheduled in a 24-h time horizon t



T

according to electricity consumer preferences. For the domestic appliance scheduling problem, we study twelve different rudimentary appliances for all homes in a smart building; typically, every home 56

contains the following appliances as mentioned in (Zhang, Evangelisti, Lettieri, & Papageorgiou, 2016).

Figure 3.2: Proposed HEMS architecture 3.2.2.1 Deferrable appliances In this research work, the smart appliances that can be interrupted or moved in any of the given time slots in a day subject to their need are called deferrable appliances. The washing machine, dish washer and spin dryer are contained within this class. Let Ad denotes the combination of deferrable appliances and ad



Ad

denotes each appliance in the deferrable class. In Equation (3.2), 𝜆d indicates the power rating of each appliance in this class. The overall electricity utilization Ɛd of deferrable appliances in a day time is shown by the following precise formula: T

d   t 1

( 

d

  d (t ))

ad òAd

57

(3.2)

The total cost per hour that is paid by the customer against all deferrable appliances can be considered as:

 (

 At 

d

d

.

  (t )   d (t ))

(3.3)

ad òAd

The overall electricity cost of one day that is paid by the customer to the utility

contrary to all deferrable appliances is given by the following equation:



Here,

T

Total Ad

 t 1

(  (

d

  (t )   d (t )))

(3.4)

ad òAd

 d (t ) denotes the ON/OFF status of deferrable appliances in the form of

one or zero. 1 If ad is ON 0 If ad is OFF

 d (t )  

(3.5)

Similarly, the total electricity consumption and electricity cost for several smart homes contrary to deferrable appliances in a day is considered by Equations (3.6) and (3.7), respectively. 

d   ( d )

(3.6)

u 1



 Total   ( ATotal ) A d

In Equation (3.6),

u 1

(3.7)

d

 d represents the daily electricity consumption and  aTotal d

represents the daily cost for a single consumer in Equation (3.7). 3.2.2.2 Non-deferrable appliances An appliance is consider non-deferrable which cannot be shifted or interrupted during execution. These appliances requirement are the right time slot for the

58

finishing point of their execution. Refrigerator and interior lighting are assumed as non-deferrable appliances. Let

and  And denotes each appliance in the class of non-

deferrable appliances. The electrical power rating of each appliance is energy consumption

nd

and total

 nd per day is presented by the following mathematical formula:

T

 nd   ( t 1



(nd   nd (t )))

(3.8)

and òAn d

Due to the un-interruptible and non-shiftable behavior of these appliances, consumers pay the maximum cost because for the demanded slot of these appliances, the utility has greater prices. The purpose behind high prices is an upturn in PAR. In order to sustain the sense of balance between generation and consumption, the utility charges greater prices. The electricity cost of each day for all appliances in the nondeferrable class can be calculated using Equation (3.9). T

 ATotal   ( nd

t 1



(nd   (t )   nd (t )))

(3.9)

and òAn d

Similarly, the cost for non-deferrable appliances over an individual time slot can be computed according to Equation (3.10).

 At  nd

Here,



(nd   (t )   nd (t ))

(3.10)

and òAn d

 nd (t ) denotes the ON/OFF status of non-deferrable appliances. 1 If and is ON 0 If and is OFF

 nd (t )  

(3.11)

The total electricity consumption and electricity cost against non-deferrable appliances for  number of users in an individual day is calculated by Equations (3.12) and (3.13), respectively.

59



nd   ( nd )

(3.12)

u 1



 Total   ( ATotal ) A nd

u 1

(3.13)

nd

3.2.2.3 Base load appliances The third category of smart appliances is known as base load appliances and these appliances comprise base load for every smart home. The operation time interval cannot be modified or interrupted for these types of appliances. Conversely, the scheduler has to schedule base load smart appliances between the earliest beginning and least finishing time which is defined by the consumers. Let

ab  Ab

represents each appliance in this category. The power rating and electrical energy consumption of these appliances can be represented by

b and  b , correspondingly.

Per day electrical energy consumption is given in Equation (3.14). T

 b   (  (b   b (t ))) t 1

(3.14)

abòAb

Per hour electricity cost and overall electrical energy cost against 24 h for a solitary smart home paid by customers are calculated by Equations (3.15) and (3.16), respectively.

 At  b



 (

  (t )   b (t ))

(3.15)

  (  (b   (t )   b (t )))

(3.16)

b

abòAb

T

Total Ab

t 1

abòAb

The total electrical energy consumption and electricity cost that is paid contrary to base load appliances for multiple households in 24 h are described in Equations (3.17) and (3.18), respectively.

60



b   ( b )

(3.17)

u 1



 Total   ( ATotal ) A b

Here,

b (t )

u 1

(3.18)

b

denotes the ON/OFF status of base load appliances in the form of

one or zero.

1 If base load appliance is ON 0 If base load appliance is OFF

 b (t )  

(3.19)

3.2.3 Electricity Cost The cost is paid against the consumed electricity which is calculated with respect to the per unit price of electricity. There are many schemes for defining per unit cost of electricity provided by the utility to motivate consumers to manage their electricity load. Some of the pricing schemes are: ToU, CPP, IBR and RTP. In our work, RTP and CPP signals are used for electricity cost estimation which are represented in Figure 3.3. The pricing scheme is represented by  (t ) for each time interval t ò T . The total electricity cost per hour for all classes is equal to the product of total load that is consumed in the particular time slot ‘t’ and the price signal  in the same hour. The total electricity cost against the individual time slot for single and multiple smart homes is denoted by Equations (3.20) and (3.21).

t 



(ad ,and ,ab   (t )   (t ))

(3.20)

ad , and , ab



 t   ( t )

(3.21)

u 1

Similarly, the total cost per day contrary to all kinds of appliances for single and multiple smart homes is calculated by Equations (3.22) and (3.23), respectively.

61

24

 Total   t 1

( 

(ad ,and ,ab   (t )   (t )))

(3.22)

ad , and , ab

T

 Total   ( Total )

(3.23)

t 1

3.2.4 Electricity Storage System We have assumed a small capacity ESS for the storage of electricity. Basically, ESS is integrated to exploit the efficiency of HEMS. ESS is a special type of deferrable electricity load, which can be shifted in any time slot in an adoptable way, i.e., charging and discharging of electricity. We have considered the ESS minimum storage level and maximum storage level to improve ESS efficiency, and the ESS charging must be less than or equivalent to the maximum charging level presented in Equation (3.24).

Figure 3.3: Pricing signals. CPP, Critical Peak Pricing. We have considered the least possible and maximum electricity storage levels of 10% and 90%, correspondingly, of the total capacity of ESS. ESS charges electricity

62

only when the electrical energy price is low and the charging level of ESS is below the upper charging level limit explained in Equation (3.25). ESS discharges only when the electricity rate is high and ESS storage is more than the ESS minimum level explained in Equation (3.26).

EStch  ES (max)

(3.24)

ESStch  ESS (upl )

(3.25)

EStdis  ES (min)

(3.26)

The stored electricity in ESS at time ‘t’ is explained by Equation (3.27), the electricity charging and discharging taken into account. There are some electricity losses due to charging and discharging ESS; therefore, turn-around ESS efficiency is taken into account.

SE (t )  SE (t  1)  k. ESS .ES (ch) (t )  k.ES dis (t ) /  ESS

(3.27)

where SE is the stored electrical energy in kWh at time ‘t’, k is the time slot (h),

 ESS is the ESS efficiency,

ES ( ch ) is charging to ESS at time ‘t’ and ES dis is the

discharging from ESS at time ‘t’. 3.2.5 Proposed Schemes In the SG, each smart home is linked with a smart meter and the smart meter is further linked with an EMC. The smart meter and EMC permit the common bidirectional communication between electrical energy customers and the utility. It motivates the consumer to perform electrical load shifting from peak to off-peak hours for bill minimization. Conversely, load shifting in off-peak hours may cause the generation of peaks in off-peak hours. This problem is measured as an optimization problem and in the past few years, several techniques have been proposed to tackle this problem. However, no technique gave sufficient attention to minimize the electrical energy cost and PAR with minimum consumer waiting time 63

simultaneously. Therefore, the key targets of our work are electrical energy cost and PAR reduction with minimum consumer WT. On behalf of this optimization problem, three meta-heuristic techniques: GA, CSA and CSOA are used. In this unit, a momentary introduction of GA, CSA and CSOA is offered and then, the procedure of these heuristic techniques in our work is elaborated. 3.2.6 Genetic Algorithm The GA is a distinguished algorithm from the meta-heuristic procedures which is also called the global exploration algorithm. The GA shows effectiveness for solving high complexity and big problems because the convergence rate of GA is very high. Due to the high merging rate, GA finds a precise optimized explanation for any computing problem. GA was described in (Whitley, 1994) and works on the base of a hereditary population like chromosomes. GA is essentially used to solve the optimization problem in calculating and the solution is achieved in binary form (1, 0); conversely, other encodings are also used for the answer. GA has three main steps for searching for the global solution. The first step is to prepare the population randomly; the assessment of fitness is next step; and the most recent step of GA is the reproduction of the first-hand population. Three key tasks, assortment, crossover and modification, are completed for the reproduction step. In our scenario, GA starts searching from randomly-initialized (binary valued 1, 0) chromosomes. The bit 1, 0 displays the ON or OFF state of the home appliances, correspondingly. The whole number of home appliances is denoted by the length of chromosomes. A random population (or solution set) generated for a specific time interval shows the status of every appliance in the home. Every possible solution is evaluated according to the fitness function; which is PAR and cost minimization with minimum user waiting time. The novel population (or solution set) is replicated iteratively using crossover and mutation operators. In crossover, two parentages are nominated on the basis of fitness values, and first-hand offspring are produced. The crossover possibility of GA is represented by pc in Equation (3.28).

pc  0.9

(3.28) 64

To bring modifications in the new population, the mutation operator is used by shifting one or more bits. Mutation possibility is very low in the natural genomic process, so paramount mutation probability is described in Equation (3.29). The complete process has restarted for searching the global best solution and the factors of GA are described in Table 3.2.

Pm  1  Pc

(3.29)

Table 3.2: GA parameters. Parameters

Values

Population size

200

Number of iterations

500

Crossover

0.90%

Parameters Mutation n

Values 0.10% 12

3.2.7 Cuckoo Search Optimization Algorithm CSOA is a heuristic nature-inspired technique which is used to find the optimal answer of any computing problem. It is built on the reproduction behavior of some distinctive types of cuckoo species and topographies of Lèvy flights of some birds as explained in (Yang, & Deb, 2009). Some cuckoos rubbish dump their eggs in other birds’ cases called host nests. Host birds determine the eggs that are left by other cuckoos for reproduction. The amount of host nests is fixed, and the determining probability of the host is P = [0, 1]. CSOA offers the best solution through accepting the following three rudimentary rules. • Every single cuckoo put only one egg at a moment in a nest which is arbitrarily selected. • Only the nests comprising a higher quality of eggs are nominated for the next production. • The number of host nests is persistent and the host birds discover the eggs that are placed by cuckoos. In our work, CSOA starts finding the local solution from randomly-laid eggs (in the form of 1, 0) in the host nests. The pattern of eggs (1, 0) represents the ON or 65

OFF state of the home appliances correspondingly in a specific time interval. Every egg (possible solution) is assessed according to our fitness function which is cost minimization with minimum user WT and PAR. After searching for the local solution, the reproduction step is performed by cuckoos. The best feature eggs (local preeminent solutions) are discovered by host cuckoos and determining possibility rate is 0.250. For the reproduction phase, only the nests having the best features eggs (best native solutions) are nominated. To find the overall best solution, Lèvy flights are implemented for producing new solutions X (t 1) for a Cuckoo i (Yang, & Deb, 2009).

X i(t 1)  X i(t )    Levy( ),

(3.30)

By nature, animals or birds search for food randomly. A Lèvy flight is a haphazard walk in which step intervals have a high probability distribution. It is performed for searching for the global best answer and generating randomness for the next generation. These steps continue till the global best solution is not found and the factors of CSOA are clarified in Table 3.3. Table 3.3: CSOA parameters. Parameters

Values

Parameters

Values

Number of host nests

50

Discovery rate

0.250

Number of iterations

2000

n

12

3.2.8 Crow Search Algorithm The CSA is meta-heuristic technique based on the very intelligent performance of crows, proposed in (Askarzadeh, 2016). The crows have the largest brain according to their body size. Crows save/hide their food in different places for future use and can remember the faces of other birds and food hiding points for some months. Moreover, crows can communicate with each other and remember the faces of each other. Crows can try to obtain the information about places where the other

66

birds/crows save their food, then steal the food when the owners of the food leave the food hiding place. In reality, crows can determine the risk-free places to secure their collected food from being stolen using their personal knowledge of having been a thief to foresee the behavior of a thief. The parameters of CSA are shown in Table 3.4. The CSA follows the four basic principles mentioned below: • Crows live in the form of a group. • Crows remember the hiding places. • Crows follow each other to steal food. • Crows secure their collected food from being stolen by a probability. In our work, CSA starts searching for the positions (local solutions) from random hiding places on the basis of the large memory of crows, where the other birds’ food is hidden. In this study, we have considered a 0.10 awareness probability of crows. Every solution shows the appliance in the form of zero or one (OFF or ON). At any position, when a crow finds the food, best solution according to our fitness function is considered for the best position and the appliance is turned ON, otherwise, it is turned OFF. After collecting all food hiding positions (local solutions), it finds out the global best solution by comparing all local solutions. Table 3.4: CSA parameters. Parameters

Values

Parameters

Flock size

100

Awareness probability 0.10

Number of iterations

100

Fight length

2

n

Values

12

3.3 FEASIBLE REGION An area that is covered by a set of some points covering all potential solutions according to our fitness function is called the feasible region. In this optimization problem, our target is to reduce the electrical energy cost and decrease PAR. Hence, electricity cost is based on these two main factors: electricity consumption and electricity price which are offered by the utility in a specific time horizon and we 67

have no authority over electricity price. However, we have authority on load only via shifting load from higher price to lower price hours to achieve our target. For electricity cost calculation, we have the following four situations using pricing signals: • Minimum load, minimum price. • Minimum load, maximum price. • Maximum load, maximum price. • Maximum load, minimum price. The WT of the user also affects the electricity cost. Therefore, we calculate the feasible region of electricity cost with user WT. When user WT is maximal, the electricity cost is low, and when WT is minimal, electricity cost is high. The feasible region of electricity cost and user WT lies in the following conditions: • Minimum waiting time, minimum cost. • Minimum waiting time, maximum cost. • Maximum waiting time, minimum cost. • Maximum waiting time, maximum cost. 3.3.1 Feasible Region for Electricity Cost and Electricity Consumption Using Real Time Pricing Signals The feasible region of the electricity cost for multiple smart homes using RTP signals is represented in Figure 3.4b. An area that is covered by these points, p1(12, 97.2), p2(12, 328.2), p3(262, 7165.7), p4(262, 2122.2), p5(213, 5823) and p6(262, 5823) demonstrates the feasible region of the electricity consumption cost for multiple homes; where p1(12, 97.2) shows the minimum electricity cost against the minimum electricity consumption and minimum electricity rate for any time interval ‘t’; p2(12, 328.2) represents the maximum cost against the minimum electricity consumption and maximum electricity price for any time interval ‘t’; moreover, p1(12, 97.2) and p2(12, 328.2) represent the electricity consumption cost when the maximum appliances are in the off status in the minimum and maximum electricity rates, respectively.

68

However, p3 (262, 7165.7) and p4 (262, 2122.2) represent the maximum and minimum cost against the same electricity consumption with maximum and minimum electricity rates accordingly. This means the electricity cost through load scheduling must not exceed the unscheduled cost for any time interval ‘t’ which is 5823 cents.

(a)

(b) Figure 3.4: Feasible region for cost and electricity consumption using RTP. (a)Feasible region for a single home; (b) Feasible region for multiple homes. 69

After implementing the threshold of 5823 cents with respect to electrical energy consumption cost, the feasible region of electrical energy consumption cost is shown by the pentagon of p1(12, 97.2), p2(12, 328.2), p4(262, 2122.2), p5(213, 5823) and p6(262, 5823); where the p4(262, 2122.2) and p6(262, 5823) show the minimum and maximum electricity cost with the same electricity consumption and for unscheduled electricity consumption, the maximum load limit is 262 kW. Therefore, scheduled scenario electricity consumption in any time interval ‘t’ must not exceed the load limit which is defined in p4 (262, 2122.2). The feasible region of electricity cost and consumption for a single smart home is represented in Figure 3.4a. In Figure 3.4a, an area is shown that is the combination of these points: p1 (0.3, 2.43), p2 (0.3, 8.20), p3 (8, 218.8), p4 (8, 64.8), p5 (5.9, 104.27) and p6 (8, 104.27) is the feasible region of electricity cost and electricity consumption for a single smart home. Here, p1 (0.3, 2.43) and p2 (0.3, 8.20) represent the minimum and maximum electricity cost with minimum load 0.3 kW, and p3 (8, 218.8) and p4 (8, 64.8) represent the minimum and maximum electricity cost with maximum electricity consumption. In the scheduled case, electricity cost must not increase the limit of the maximum electricity cost for any time period. The maximum electricity cost limit is 104.27 which is represented in point p6 (8, 104.27) with maximum electricity consumption of 8 kW. 3.3.2 Feasible Region for Electricity Cost and Electricity Consumption Using Critical Peak Pricing Signals Figure 3.5(a, b) demonstrates the feasible area for electrical energy cost and electrical energy consumption against solitary and multiple smart households using CPP signals. In Figure 3.5a, an area that is covered by these points: p1(0.3, 2.43), p2(0.3, 16.50), p3(8, 440), p4(8, 64.8), p5(3.75, 206.64) and p6(8, 206.64) shows the feasible region of the electricity cost and electricity consumption for a single smart home, where the points p1(0.3, 2.43) and p4(8, 440) represent the minimum and maximum electricity cost with minimum and maximum electricity consumption, respectively. When we implement the threshold of 206.64 cents on electricity cost, the feasible region of electricity cost is shown by the pentagon of the following 70

points: p1 (0.3, 2.43), p2 (0.3, 16.50), p4 (8, 64.8), p5 (3.75, 206.64) and p6 (8, 206.64). Points p5 (3.75, 206.64) and p6 (8, 206.64) clearly show that electricity cost is the same with different electricity consumptions. Therefore, in the scheduled case, electricity cost for any time interval t must not exceed the threshold which is 206.64 cents.

(a)

(b) Figure 3.5: Feasible region for cost and electricity consumption using CPP. (a) Feasible region for a single home; (b) Feasible region for multiple homes. 71

Figure 3.5b represents the feasible region for cost and electricity consumption against multiple smart homes using CPP signals. The combination of these points p1 (12, 97.2), p2 (12, 660), p3 (262, 14,410), p4 (262, 2122.2), p5 (215, 11,835) and p6 (262, 11,835) shows the feasible region in Figure 3.5b. The points p1 (12, 97.2) and p2 (12, 660) represent the minimum and maximum electricity cost with the same load which is 12 kW. The points p3 (262, 14,410) and p4 (262, 2122.2) show the maximum electricity load 262 kW with minimum and maximum electricity cost. However, after implementing the cost limit, which is 11,835 cents with maximum load 262 kW, the electricity cost for any time slot ‘t’ must not exceed the defined limit. 3.3.3 Feasible Region for Electricity Cost and User Waiting Time Using Real Time Pricing Signals The feasible region of electricity cost and user WT for multiple homes is represented in Figure 3.6b. A region that is covered by these points: p1 (0, 104), p2 (0, 6654), p3 (1.92, 111), p4 (1.92, 2590), p5 (0.63, 5823) and p6 (0, 5823) shows the feasible region of electricity cost with user WT for multiple smart homes, where the points p1 (0, 104) and p2 (0, 6654) demonstrate the minimum and maximum electricity cost with no WT. However, p3 (1.92, 111) and p4 (1.92, 2590) represent the minimum and maximum cost with the same WT which is 1.92 (h). The point p2 (0, 6654) represents maximum user discomfort because the user pays the maximum cost of 6654 cents against zero WT. Therefore, p2 (0, 6654) has omitted from feasible area by implementing the all-out electricity cost boundary which is 5823 cents in any time slot ‘t’. However, p5 (0.58, 5823) and p6 (0, 5823) show the same electricity cost at different WTs. This means deferrable appliances can minimize sound electricity cost by reducing their electricity demand throughout peak hours without increasing user WT. Figure 3.6a shows the feasible region of electricity cost with user WT for a single smart home. A feasible region of electricity cost and user WT is generated with concatenation of these six points: p1 (0, 2.43), p2 (0, 218.8), p3 (1.66, 2.43), p4 (1.66, 63.24), p5 (1.22, 104.27) and p6 (0, 104.27). Minimum WT with minimum and

72

maximum electricity cost is represented by the points p1 (0, 2.43) and p2 (0, 80), respectively.

(a)

(b) Figure 3.6: Feasible region for cost and WT RTP. (a) Feasible region for a single Home; (b) Feasible region for multiple homes. Points p3 (1.66, 2.43) and p4 (1.66, 63.24) demonstrate the maximum user WT with minimum and maximum electricity cost, respectively. The electricity consumer pays 73

maximum electricity cost of 218.8.27 cents without WT on point p2 (0, 218.8). After implementing the maximum electricity cost limit which is 104.27 cents, p2 (0, 218.8) is negated from the feasible region and it is shown in Figure 3.6a. Points p5 (1.2, 104.27) and p6 (0, 104.27) represent the same electricity cost with different user WT. Therefore, the electricity consumer can minimize the electricity cost via maximum consumption in off-peak hours. 3.3.4 Feasible Region for Electricity Cost and User Waiting Time Using Critical Peak Pricing Signals The feasible region of electricity cost with user WT for solitary and multiple smart homes by means of CPP signals is demonstrated in Figure 3.7(a,b) correspondingly. In Figure 3.7a, an area covered by these points: p1 (0, 2.43), p2 (0, 440), p3 (1.66, 2.43), p4 (1.66, 65.40), p5 (1.04, 206.64) and p6 (0, 206.64), shows the feasible region for a single smart home using CPP signals. In this region, points p1 (0, 2.43) and p2 (0, 440) represent the minimum cost of 2.43 cents and maximum cost of 440 cents with no WT. However, points p3 (1.66, 2.43) and p4 (1.66, 65.40) show the minimum and maximum cost with maximum WT of 1.66 h. After implementing the maximum cost limit of 206.64 cents, electricity cost must be less than the defined limit. When we exclude the point p2 (0, 440) due to the higher cost compared to points p5 (1.04, 206.64) and p6 (0, 206.64), the electricity cost limit is 206.64 cents with the WT of 0 and 1.04 h. This clearly demonstrates that we minimize the electricity cost with bearable WT. In Figure 3.7b, a region overspread by the following points, p1 (0, 104), p2 (0, 14,440), p3 (1.92, 111), p4 (1.92, 5208), p5 (0.54, 11,835) and p6 (0, 11,835) shows the feasible region against electricity cost with user WT for multiple smart homes using CPP signals. However, we negate the point p2 (0, 14,440) due to the high cost with zero WT by implementing a threshold on the electricity cost. The threshold is 11,835 cents which is shown on points p5 (0.54, 11,835) and p6 (0, 11,835) with 0 and 0.54 h WT. After excluding the point p2 (0, 14,440), the feasible region of electricity cost and user WT is shown by the following five points: p1 (0, 104), p3

74

(1.92, 111), p4 (1.92, 5208), p5 (0.54, 11,835) and p6 (0, 11,835). This clearly shows that the users can reduce their electricity cost with affordable WT.

(a)

(b) Figure 3.7: Feasible region for cost and WT using CPP. (a) Feasible region for a single home; (b) Feasible region for multiple homes. 75

3.4 POWER SCHEDULING IN SMART HOMES USING HGWDE OPTIMIZATION TECHNIQUE In this section, proposed solution 2 is described with reference to the subproblem 2 (i.e., section 1.2.2), which is described in the chapter 1. The solution is discussed in detail as given below. 3.4.1 System Model The basic structural design of our proposed scheme is explained in Figure 3.8. A single smart home with seventeen appliances is taken into consideration (Javaid et al., 2017).

Figure 3.8: Proposed system model. Each appliance has different power ratings. Appliances are scheduled to achieve the objective function. The HEM system is incorporated for mutual communication between customer and utility. 76

(a) RTP Signal

(a) RTP Signal

(b) CPP Signal Figure 3.9: Pricing schemes. 77

Different electrical energy pricing signals are defined by the utility such as CPP, RTP, DAP, TOU, adjustable time pricing and IBR. Time slots in which load demand from the consumer ranges the maximum value are called peak hours. Electricity prices are usually high in on-peak hours. A bidirectional flow is presented in the model. First, electricity price is sent to the consumer. Considering the electrical energy rates, EMC decides which appliances should be turned on. Heavy appliances cannot be switched on in peak hours. EMC is incorporated to induce automation in order to decrease electrical energy bills. After the final judgement, the load demand is dispatched to the utility through SM, and ultimately, the requested power is sent to the consumer via the utility. RTP and CPP schemes for 24 h are given in Figure 3.9 (a, b). The data for these two pricing schemes are collected from (Vardakas, Zorba, & Verikoukis, 2016; Ogunjuyigbe, Ayodele, & Akinola, 2017). 3.4.2 Classification of Appliances Appliances are classified on the basis of their working behavior and electrical energy consumption pattern. Particulars of each classification are specified below. 3.4.2.1 Shiftable appliances Such appliances are also called deferrable appliances. They can be moved to any time slot, however cannot be disturbed during their operational time. As soon as their operation is started, they cannot be stopped till the time slot ranges zero. Washing machines and dishwashers are uninterpretable appliances. The whole number of appliances is specified by D. Shiftable appliances are a subcategory of total appliances. From Equation (3.32), P  Ds denotes the set of shiftable appliances. The power rating of each appliance is denoted by  and  P (h) , where tòT displays the power rating and status of

Ds appliance at a specific time slot. Equation (3.33)

represents the cases as zero when the appliance is OFF and one when the appliance is ON. 3.4.2.2 Controllable appliances Controllable appliances are also called interruptible appliances. The functioning time of these appliances cannot be altered; for instance, air conditioner, lightning and 78

heating system.

Dc is the group of controllable appliances. The cost for controllable

appliances is given in Equation (3.32). 3.4.2.3 Non-shiftable appliances Non-shiftable appliances are also called base appliances. Electrical energy consumption and functioning time cannot be altered for such appliances. Televisions, refrigerators and telephones are classified as non-shiftable appliances. From Equation (3.32), P  Dns is the group of non-shiftable appliances.

P  Ds , Dns , Dc

(3.31)

H

P    P  P (h)

(3.32)

h 1 P D

1, if appliance is ON  P (h)   0, otherwise

(3.33)

Classification and the power rating of each appliance are given in Table 3.5. 3.4.3 Proposed Scheme In this section, different meta-heuristic optimization algorithms are debated for DSM. These techniques are implemented on a single home with seventeen appliances. Each appliance has different power ratings relating to their energy consumption patterns. There are four elementary steps of electrical energy such as generation, transmission, distribution and consumption. Electrical energy is consumed by the domestic, corporate and industrial sector. On the other hand, our key goal is to attain efficient electrical power scheduling in the domestic sector. On behalf of DSM in the domestic area, diverse optimization techniques have been proposed by numerous researchers. In this regard, we have suggested an optimization method, HGWDE, which is a mixture of two different meta-heuristics: EDE and GWO to optimize the electricity consumption. All three techniques are implemented to assess the performance of our suggested scheme with respect to existing schemes. 79

Table 3.5: Parameters of appliances. Appliance

Appliance

Power

Earliest

Finishing

Rating

Starting

Time (h)

(kWh)

Time (h)

Washing Machine

1.4

6

10

1–3

Dish Washer

1.32

15

20

1–3

Hair Straightener

0.055

18

8

1–2

Hair Dryer

1.8

18

8

1–2

Microwave

1.2

18

8

3–5

Telephone

0.005

9

17

1–24

Computer

0.15

18

24

1–3

Oven

2.4

6

10

2-4

Cooker

0.225

18

24

3-5

Iron

2.4

18

24

1-2

Toaster

0.8

9

17

1-2

Kettle

2

18

8

1-2

Non-Shiftable

Printer Television

0.011 0.083

18 9

24 17

1-2 6–1

Appliances

Refrigerator

1.666

24

1

1-24

Controllable

Air Conditioner

1.14

16

24

6–8

Appliances

Lightning

0.1

1

24

12–20

Class

Shiftable Appliances

LOT(h)

3.4.4 Enhanced Differential Evolution EDE is an advanced form of DE, which was offered by Storn in 1995 for the very first time. The aforementioned is a population-based algorithm where the initial population is randomly generated. The four key steps of EDE include early generation of the population, mutation, crossover and assortment. The population is created arbitrarily using Equation (3.34).

X i, j  l j  (rand  (U j  l j ))

(3.34)

80

Generate a random function to produce three vectors xr1 , xr 2 ,

xr 3 in order to form

a modified vector. The first vector is professed as the target vector. The mutant vector is formed according to Equation (3.35) given below:

Vi ,G1  xr1,G  F ( xr 2,G  xr 3,G )

(3.35)

where F is a scaling factor. After forming the mutant vector, three experimental vectors are formed. Then, the finest trial vector is nominated to compare it with the target vector, so that the off-spring can be populated with the finest vector. The first three experimental vectors are formed using Equations (3.36)–(3.38). V j ,i ,G 1 if randb( j )  0.3 U j ,i ,G 1   Otherwise  x j ,i ,G

(3.36)

V j ,i ,G 1 if randb( j )  0.6 U j ,i ,G 1   Otherwise  x j ,i ,G

(3.37)

V j ,i ,G 1 if randb( j )  0.9 U j ,i ,G 1   Otherwise  x j ,i ,G

(3.38)

The fourth and fifth trial vectors are created using Equations (3.39) and (3.40).

U j ,i,G1  randb( j)  x j ,i ,G

U j ,i ,G1  randb( j )  v j ,i ,G  (1  randb( j ))  x j ,i ,G

(3.39) (3.40)

The pseudocode for EDE is presented in Algorithm 1. Max.iter represents the maximum iterations; POP is the total population, which is the number of possible solutions in the case of HEMS; and h is the total time slots. CR is the crossover ratio, which is taken as 0.3, 0.6 and 0.9.  ,  and x are the mutant, trial and target vectors. EDE in terms of HEM is shown in Table 3.6.

81

Table 3.6: EDE mapping on HEM. EDE Parameters

HEM Parameters

Values

Population

Possible solution

50

Number of dimensions

Number of appliances

17

Gradient of problem

Scheduling

vary

3.4.5 Grey Wolf Optimization GWO is a heuristic optimization technique, motivated by the hunting nature and ordered leadership of wolves. There are four stages of headship, i.e., alpha  , beta  , delta  and gamma  . Alpha is measured as the smartest leaders of the

collection; they are responsible for controlling other wolves on hunting policies. After alpha, there comes beta and delta in the ranked level, and gamma is the weakest associates of the group. Therefore, gamma cannot be considered for leadership abilities. In HEM, alpha



is in use as the fittest member to attain the objective

function of cost reduction. The primary population is created arbitrarily using Equation (3.41): X (i, j )  rand ( POP, D)

(3.41)

Where POP denotes the entire population of gray wolves and D is the overall number of appliances. Coefficients A and C are considered to assess the objective function (distance from prey) of respective search agent. 3.4.5.1 Encircling prey Gray wolves encircle a prey before hunting. For the mathematical formulation of encircling behavior of gray wolves, the following equations are considered from (The Wind Power Program and the UK NOABL Wind Speed Database, 2018).

X (t  1)  X p (t )  A  D

(3.42)

D | C  X p(t )  X (t ) |

(3.43) 82

X p denotes the location of prey, while X is the location of the gray wolf at the

tth iteration, which is given by Equation (3.42). The vectors A and C are calculated according to Equations (3.44) and (3.45):

A  2a  r1  a

(3.44)

C  2  r2

(3.45)

where r1 and r2 are random vectors within the range of [0, 1]. After multiple iterations, the value of

a

is reduced from two to zero. The value of C is taken

randomly between zero and two, which is used to define the weight for the attractiveness of prey. Diverse spaces around the finest agent can be reached with respect to the present position by regulating the value of A and B vectors. Algorithm 1: EDE 1: Initialize all parameters Max: iter, CR, POP, h 2: Generate initial population using Equation (3.34) 3: for h = 1: H do 4: Calculate mutant vector using Equation (3.35) 5: for iter= 1:Max.iter do 6:

Calculate first trial vector with CR 0.3

7:

if rand ()  0.3 then

j j

8: 9: 10:

else

 j  xj

11:

end if

12:

Calculate second trial vector with CR 0.6

13:

if rand ()  0.6 then

14:

j j

83

15:

else

 j  xj

16: 17:

end if

18:

Calculate third trial vector with CR 0.9

19:

if rand ()  0.9 then

j j

20: 21:

else

 j  xj

22: 23:

end if

24:

Generate fourth and fifth trial vector using Equations (3.39) and (3.40)

25:

Find the best trial vector

26:

X new  best of  j

27:

Compare trial vector with target vector

28:

if f ( X new )  f ( X j ) then

X j  X new

29: 30:

end if

31: end for 32: end for 3.4.5.2 Hunting Hunting is mainly guided by  , while  and  are the secondary participants. They follow the lead of  , which has the best knowledge about the position of prey. The first three best solutions are obtained so that other members of the group such as gamma  update its position according to the best solution. The position of wolves is updated according to Equation (3.46).

X t 1 

x1  x2  x3 3

(3.46)

84

x1 , x2 and x3 are determined by using Equations (3.47)–(3.49). x1  x  A1  (d )

(3.47)

x2  x  A2  (d )

(3.48)

x3  x  A3  (d )

(3.49)

x , x and x are the best solutions obtained at the tth iteration; A1 , A2 , A3 are determined using Equation (3.44), while D , D , D

are determined using

Equations (3.50)–(3.52):

where

D  C1  x  x

(3.50)

D  C2  x  x

(3.51)

D | C3  x  x

(3.52)

C1 , C2 and C3 are calculated using Equation (3.45). The last step is the

gradation of variable a ; it controls the trade-off between exploration and exploitation by taking the value from two to zero in each iteration as shown in Equation (3.53). a  2t

2 Max.iter

(3.53)

Equation (3.54) demonstrations the objective function, which is considered as the power rating of each appliance multiplied by the rank of the appliance.

Fitness  D  D (h)

(3.54)

From Algorithm 2, Max: iter represents the maximum iterations, POP is the total population, D is the number of appliances and fitness is the objective function.

 is

considered as the best solution or participant in the hunting behavior, while  and 85

 are the second and third optimal solutions, respectively. The fitness function is compared with the fitness of  ,  and  to evaluate the best hunting leader. Their final positions are updated according to Equations (3.50)–(3.52). 3.4.6 Hybrid Grey Wolf Differential Evolution In this section, our proposed scheme is discussed in detail. In EDE, the new population is generated using four steps: initialization, mutation, crossover and selection. The population is updated by selecting the best trial vector out of five vectors and then comparing it with the target vector. The selection procedure of EDE is effective as it considers the best trial vector from all the available vectors. In GWO, there are three basic steps: encircling prey, hunting and updating the position of wolves within the pack.



is considered as the leader. All the search agents of the

pack update their position according to

 . In GWO, their is no comparison of 

with  and  . It might be possible that  and  are much closer to the prey as compared to  . In order to make a clear comparison among all the search agents, crossover from EDE is performed. After selection of the best search agent, updating the position of search agents is performed according to GWO. HGWDE is adopted as it has the best features of EDE and GWO. The detailed steps of HGWDE are presented in Algorithm 3. The steps of HGWDE are initialization, encircling prey, selection of best search agent and updating the position. A random population of wolves Xi(i  1, 2,..., n) is generated using Equation (3.32). Selection is performed according to the steps given in Algorithm 3. A mutant vector



is created by using three random vectors according

to Equation (3.35).  ,  and  are initialized as three vectors. The fitness of



and

 ,  ,  is calculated by using Equation (3.54). The crossover is performed by using

the equations given below:

 if fitness of  j    new   j Otherwise 

86

(3.55)

 if fitness of  j    new   j Otherwise   if fitness of  j    new   j Otherwise 

(3.56)

(3.57)

After selection, the position of search agents is updated relatively according to the steps of GWO, which are given in Equation (3.46). Mapping of HGWDE parameters with HEM is given in Table 3.7. Algorithm 2 GWO 1: Initialize all parameters Maxiter , POP, D,  ,  ,  2: Arbitrarily generate primary population of gray wolves Xi(i  1, 2,..., n) 3: X (i, j )  rand ( POP, D) 4: while iter < Maxiter do 5: for i=1: POP do 6:

Compute fitness as objective function using Equation (3.54)

7: end for

  score

8: if fitness 9:

 score = fitness

10:

 Pos =

X (i,:)

11: end if 12: if fitness

  score and fitness   score then

13:

 score = fitness

14:

 Pos =

X (i,:)

15: end if 16: if fitness

  score and fitness   score and fitness   score then

17:

 score = fitness

18:

 Pos =

X (i,:)

87

19: end if 20: for i = 1: POP do 21:

for j = 1: D do

22:

Create r1 and r2 arbitrarily between 0 and 1

23:

Compute fitness coefficients A and C via Equations (3.44), (3.45) Inform values of  ,  ,  using Equation (3.50)–(3.52)

24: 25:

end for

26: end for 27: end while Each step of Algorithm 3 is discussed below in detail. In Step 1, the proposed algorithm starts by initializing the required parameters. The population is generated randomly in Step 2. After population generation, the counter is set to maximum iterations. Crossover is performed by comparing the fitness of the mutant vector with  ,  ,  using EDE. Update positions of search agents using GWO. The last step is

to repeat until the termination criteria are satisfied Table 3.7: Hybrid gray wolf differential evolution (HGWDE) mapping on HEM. HGWDE parameters

HEM parameters

Population

Possible solution

50

Wolfs in each pack

Number of appliances

17

Minimum distance from prey

Min (cost)

vary

Status of leader

status of appliance

Values

1 or 0

Algorithm 3 HGWDE 1: Prepare all parameters Maxiter , POP, D,  ,  ,  2: Arbitrarily generate primary population of gray wolves Xi(i  1, 2,..., n) 3: X (i, j )  rand ( POP, D)

4: while iter < Maxiter do 5: for i = 1: POP do 88

6:

Generate a mutant vector using Equation (3.35) from EDE

7:

Calculate fitness of mutant vector as cost cost  j

8:

Randomly generate  ,  ,  Calculate fitness of  ,  ,  as the objective function using Equation

9:

(3.54) 10:

if fitness of  ( j )   score then

 position   j

11:

12:

end if

13:

if  j   score and upsilon j  score then

 position   ( j)

14:

15:

end if

16:

if x j   score and fitness x j  score and fitness x j   score then

 position   j

17:

18:

end if

19:

for i = 1: POP do

20:

for j = 1: D do

21:

Produce r1 and r2 arbitrarily between 0 and 1

22:

Compute fitness coefficients using Equations (3.44) and (3.45) Inform values of  ,  ,  using Equations (3.50)–

23:

(3.52) 24: 25:

end for end for

26: end for 27: end while

89

3.5 A DOMESTIC MG WITH OPTIMIZED HOME ENERGY MG SYSTEM In this section, we have discussed our proposed solution 3 according the identified sub-problem 1.2.3 explained in the problem statement of the chapter 1. The details of the system components and their problem formulation is described below. 3.5.1 System Model This section contains the problem formulation and the components of the contribution 3. 3.5.2 Formulation of the Problem Statement In this section, we have mathematically formulated our problem by defining an objective function along with few constraints. The detailed description of the formulation is presented as follows. 3.5.3 PV Generation We proposed a smart home which is equipped with a rooftop PV generation system as solar energy is less costly than other RESs (biomass, wind, biogas, tidal and geothermal) and available everywhere. The earth receives a huge amount of solar radiations and most of the areas with population have insulation levels of 150–300 watts/m2 (Solar Energy, 2017). The output power from a PV panel is given in Equation (3.58), (Lotfi, Tarazouei, & Ghiamy, 2013; Ismail, Moghavvemi, & Mahlia, 2012; Daud, & Ismail, 2012).

PPV out  PN  PV  (

G )  [1  KT (Tc  Tref )] Gref

where total electricity generated by PV is presented by

(3.58)

PPV out , G shows solar

irradiation (W/m 2 ), Gref is solar radiation at reference conditions (Gref 1000\W/m 2 ), the cell temperature at reference conditions is (Tref  25



C), 

=

KT is

the temperature coefficient (in this work, a value of KT  3.7 10  3(1/ C) is 90

adopted) and

Tc depicts the cell temperature which is calculated by the Equation

(3.59), (Daud, & Ismail, 2012).

Tc  Tamb  (0.0256  G) where,

(3.59)

Tamb represents the ambient temperature.

Figure 3.10: Wind turbine power output and wind speed 3.5.4 Wind Generation Wind is a very promising source of RE. USA, China, Germany, Spain, Denmark and India are the leading countries in power generation from wind using wind turbine. Power from the wind can be given by the following Equation (3.60), (Fathima, & Palanisamy, 2015; Sinha, & Chandel, 2015).

P  0.5 Ars V 3 Pcoff

(3.60)

91

where

Ars shows the rotor swept (blade) area of wind turbine in m2, the air

density is represented by



in kg/m2, the average wind velocity is shown by V in

m/s and Pcoff is a power coefficient which shows the efficiency of a wind turbine (maximum value of 0.59). The output power available from wind turbine depends on wind speed and can be given by Equation (3.61) taken from (Liu, Kong, Liu, Peng, & Wang, 2015). 0, v  vcut in , v  vcut out   (v  vcut in ) Pwt (v)   prated , vcut in  v  vrated  vrated  vcut in  vrated  v  vcut out  Prated ,

where

(3.61)

vrated , vcut out and vcut in are the rated, cut-out and cut-in wind speeds,

respectively. The rated output power of wind turbine is shown by Prated . Wind turbine power output and wind speed is given in Figure 3.10 (Wang, Palazoglu, & El-Farra, 2015). (Wang, Palazoglu, & El-Farra, 2015; The Wind Power Program and the UK NOABL Wind Speed Database, 2018). 3.5.5 Battery Bank System (BBS) The capacity of the BBS

(CWh ) is calculated by Equation (3.62) in time slot t

(Khatib, Mohamed, Sopian, & Mahmoud, 2011).

CWh (t )  ( EL  AD) / (V B  DOD) where DOD shows the allowable depth of discharge,

(3.62)

EL is daily energy

consumption, AD is a number of autonomy days, V and  B are the voltage and BBS efficiency, respectively. Energy charging and discharging by the BBS during the time period from t-1 to t can be given by Equation (3.63) (Suryoatmojo, Elbaset, Pamuji, Riawan, & Abdillah, 2014).

CB (t )  CB (t  1).(1   )  PBAT (t ) 92

(3.63)

Where

CB (t ) and CB (t  1) show the available power (which may be consumed

by consumer) in BBS at time slot t and (t-1), respectively. The symbol  denotes the self-discharge rate of the BBS and it is assumed as 0.002 in our study. the power from battery bank in time slot t. The value of

PBAT (t ) is

CB (t ) remains between

(CBmin ) and (CBmax ) during charging operation of the BBS as given by Equation (3.64). CBmin  CB(t )  CBmax Where

(3.64)

CBmin and CBmax are minimum and maximum allowable energy levels in

the BBS. Furthermore, the BBS is charged from own MG and commercial grid when electricity prices are low, and charged electricity is used in high price hours. 3.5.6 Energy Consumption We assume that t represents a single time slot and T represents total time horizon, which is 24 h. Set of appliances in home is denoted by S and each appliance is denoted by



and consumes an amount of energy

E (t ) in time slot t such that t  T.

S  1 ,  2 ,  3 , ...,  n .

(3.65)

The daily energy consumption by non-deferrable load (NDL), interruptible load (IL) and must-run load (MRL)



and the total energy consumed in whole day by all

appliances are calculated in Equations (3.66)–(3.69), respectively (Ahmad et al., 2017). 24

SN

t 1

sn 1

NDL NDL NDL ENDL   ( EtNDL , snSN )  {Et1, snSN  Et 2, snSN  Et 24, snSN }

24

SI

t 1

si 1

E   ( EtIL,siSI )  {EtIL1,siSI  EtIL2,siSI  EtIL24,siSI } IL 

93

(3.66)

(3.67)

24

SM

t 1

sm 1

MRL MRL MRL EMRL   (  EtMRL , smSM )  {Et1, smSM  Et 2, smSM  Et 24, smSM }

24

SN

SI

SM

t 1

sn 1

si 1

sm 1

(3.68)

IL MRL Etotal  ENDL  EIL  EMRL   ( EtNDL , snSN   Et , siSI   Et , smSM ) (3.69)

NDL NDL NDL IL IL IL Where Et1,snSN  Et 2,snSN  Et 24,snSN , Et1, siSI  Et 2, siSI  Et 24, siSI and MRL MRL EtMRL 1, smSM  Et 2, smSM  Et 24, smSM are the energy consumption of NDl, IL and MRL

appliances, respectively, denoted by  . E

total

and is the total energy consumption

of NDL, IL and MRL appliances. 3.5.7 Energy Pricing and Electricity Cost Many pricing schemes in existence are defined per unit energy cost. Some of the pricing schemes are RTP, peak pricing (PP), CPP, TOU pricing, real-time market pricing (RTMP), non-critical peak (NCP), locational marginal pricing (LMP), hourly pricing (HP), and critical peak pricing with rebate (CPP-R) (Javaid, Khan, Ullah, Mahmood, & Farooq, 2013; Wolak, 2018). However, most of the work on appliances scheduling consists of the DAP or TOU pricing scheme. In TOU scheme, the time is divided into multiple time slots. In this work, we use RTP scheme which remains constant for one time slot and varies from one slot to another slot. The electricity cost against each class (NDL, IL and MRL) of appliances and total electricity cost of all appliances are calculated by Equations (3.70)–(3.73), respectively. 24

SN

t 1

sn 1

E (t )   (  (E( ndlsn ,t )  Yndlsn (t ))  PS rtp (t )) ndl ps

24

SI

t 1

si 1

E (t )   (  (E(ilsi ,t )  Yilsi (t ))  PS rtp (t )) il ps

24

SM

t 1

sm 1

E (t )   (  (E( mrlsm ,t )  Ymrlsm (t ))  PS rtp (t )) mrl ps

94

(3.70)

(3.71)

(3.72)

ndl il mrl E total ps (t )  E ps (t )  E ps (t )  E ps (t )

(3.73)

1 if sn is on Yndlsn (t )   0 if sn is off

(3.74)

1 if si is on Yilsi (t )   0 if si is off

(3.75)

1 if sm is on Ymrlsm (t )   0 if sm is off

(3.76)

where

PS rtp (t ) is RTP signal in time slot t. Yndlsn (t ) , Yilsi (t ) and Ymrlsm (t ) are

the ON/OFF state of NDL, IL and MRL appliances as shown in Equation (3.74)– ndl il mrl (3.76), respectively. E ps (t ) , E ps (t ) and E ps (t ) are the electricity prices of NDL, IL

and MRL appliances in time slot t, respectively. ps show the price signal, while , si ,

sm , ndlsn , ilsn , and mrlsn

sn

represent NDL, IL and MRL appliances and

SN, SI and SM represent the set of appliances of NDL, IL and MRL, respectively. 3.5.8 Peak-to-Average Ratio PAR balancing is necessary to bring equilibrium of demand and supply between consumers and utility. The PAR is very important for cost savings, achieving stable system and increasing spinning reserve system capacity. The PAR also helps in reducing peak load demand, peak power plants cost, transmission line losses, increasing of electrical equipment life, etc. Let the peak and average load of the smart home be denoted by

LP and LA , respectively. Then, the PAR of the demanded load

 PAR can be given in Equation (3.77) (Liu et al., 2014).  PAR 

LP TmaxtT Lt  LA  Lt

(3.77)

tT

where

Lt is a total load of all consumers. 95

3.5.9 Appliances Waiting Time Let,

t wt denote the AWT of all smart appliances. T w is an AWT term which

introduces the start time, stop time, maximum waiting time, LOT and minimum waiting time of appliances



such that

0  T w  Tmw . Now, the AWT  can be

given by Equation (3.78) (Rasheed et al., 2015).

t wt 

where,

Tow  Tstw Tmw  Tl

(3.78)

t wt is appliance  waiting time, To Tow is the appliance  ON time,

Tstw is the appliance  start time, Tl is the appliance  LOT and Tmw is the appliance



maximum waiting time. The AWT is 0 if start time and on time are

equal, i.e., (T w  T w ) . On the other hand, if the earliest starting time of any o

st

appliance and ON time when any appliance starts execution is different, i.e.,

(Tow  Tstw ) , then consumer has to wait to perform the operation of the appliance. 3.5.10 Objective Function The reduction in electricity cost, PAR and AWT for maximum UC were the basic objectives of this study which are achieved by proper management of smart appliances. PAR reduction is important for both utility and consumers to minimize the operation time of peak power plants and backup generators. We supposed that there is single smart home (electricity consumer) in a residential area with HEMS and is consuming electricity from commercial grid and owned MG. Furthermore, our proposed schemes did not affect any liberalized electricity market. We formulated the optimization problem using MKP. From a set of appliances, we selected an appliance for a particular hour to be ON or OFF. Each appliance has a single unique weight and value, which shows the ON/OFF state and power rating, respectively. To allow any smart appliance to perform its operation in particular time slot, EMC decides based on defined objective function (see Equation (3.79)). The constraints

96

must be satisfied by the total weight i.e., the total energy consumed by appliances is explained in Equation (3.79) and the constraints in Equations (3.80)–(3.82). Objective function:

min((E ndl (t )  E il (t )  E mrl (t )  (E PV (t )  EWD (t ))  BS (t ))  P rtp (t )) (3.79) Subject to:

E nim (t )  Eug (t )  E PV (t )  EWD (t )  BS (t ),  1  t  24 NDL IL MDL E nim (t )  Ereq  Ereq  Ereq

(3.81)

t0  tsch  tmax where E

ndl

(3.80)

(3.82)

(t ) , E il (t ) and E mrl (t ) are the energy consumption of NDL, IL and

MRL appliances in time slot t, respectively.

E PV (t ) ,

WD(t ) and BS (t ) are the

available energy from PV, wind and battery in time slot t, respectively.

E nim (t ) is

the total energy consumption caused by all smart appliances in particular time slot t.

Eug (t ) is the available energy from utility grid that a consumer can import in time min

slot t. Eunsch is the minimum amount of energy consumed in unscheduled case. and

t0

tmax are the lower and upper limit of scheduling horizon, respectively. tsch shows

the scheduling time of appliances. In Equation (3.81), constraint is defined to bring balance between the energy of demand and supply. 3.5.11 System Model In proposed system model, each electricity consumer has a HEMS. The smart user uses RESs, BBS, and energy from the electric grid for meeting their load requirements. Here, the RESs consists of PV and wind turbine and Table 3.8 displays the assumed rating of the system model components. The appliances are scheduled to reduce electricity expenses, PAR and maximum UC. Additionally, the appliances 97

are categorized into three classes by consumers, i.e., NDLAs, which cannot be moved to another time slot; NDLAs have start and end points to describe its timespan and we assumed that consumer cannot compromise in these type of appliances; and ILA, the load of particular consumers that, according to the agreement, can be cut off by the supply undertaking for a limited period. Its operation can be suspended in the middle. The interruptible appliance has a task having several orders of operation which can be interrupted. MRLA must be run immediately at any time. These appliances are not shiftable, non-deferrable and non-interruptible. They must be run at any cost, as presented in Table 3.9. Furthermore, all appliances considered in this work are connected to an AC system. The suggested system model is shown in Figure 3.11 and consists of smart meter, EMC, smart scheduler unit (SMSU), AMI, PV and wind power generation system, solar charge controller, DC/AC inverters, appliances and BBS. Mutual communication between consumers and utility is only possible by integrating AMI and works as a backbone for the SG. The responsibility of AMI includes the hourly load demand and the electricity rates between utility and smart meter. Utility and smart home interconnect with each other via smart meter which acts as a communication gateway between them. The processing, reading, sending of energy consumption data and receiving of pricing signals are the main functions of a SM. RESs are the real alternative sources of local power generation to fossil fuel. RESs mainly consist of PV system, fuel cells, hydro and wind turbines, while, in our case, they consist of PV and wind turbine. PV panel generates electricity which is DC and then converted to AC via converter. A BBS works as a both sink and source of energy, and is considered as a suitable solution for RES integration in the residential sector. Therefore, in the proposed system model, BBS power is used to exploit the PV and wind system energy efficiently and to alleviate the electricity cost and PAR. An SMSU is programmed using heuristic algorithms and works in between the smart meter and EMC. The optimal energy consumption pattern for all appliances is generated by SMSU and then it sends the scheduling pattern to the EMC for further processing.

98

EMC controls the BBS and operation of all appliances according to the generated scheduled by SMSU. EMC is the core of the proposed system model. Table 3.8: Power rating of system model components. Component Rating Battery

1.2 kWh

Wind turbine

10 kW

Solar panel

230 W

Table 3.9: Appliances classification. Non-deferrable loads Interruptible loads Must-run loads Home lightings

Water heater

Fan

PEV

Optional lightings

Exhaust fan

Iron

Heated towel rails

Desktop PC

Pool Pump

Personal computer

ESS

Refrigerator

Television

Washing machine

Out-door lightings

Electric clock

Figure 3.11: Block diagram of system model. 99

Fans

3.5.12 Optimization Techniques The objective function for optimization of appliances scheduling discussed in section 3.5.10 was solved using nature-inspired algorithms such as GWO, WDGWO, GA, BPSO, WDO, WDGA and WBPSO. Therefore, we adopted heuristic techniques to solve our optimization problem. Furthermore, heuristic algorithms provide alternative ways for solving complex problems and have better performance as compared to other techniques. 3.5.13 Grey Wolf Optimization GWO algorithm represents the hunting mechanism and leadership hierarchy of grey wolves which is proposed in (Mirjalili, Mirjalili, & Lewis, 2014). To understand the leadership hierarchy, there are four types of wolves, i.e., alpha, beta, delta and omega. To perform optimization, four main steps are implemented in GWO, i.e., hunting, searching, encircling and attacking prey. Grey wolves always live in a pack with average size of 12–15. Encircling prey can be mathematically given in Equation (3.83) and (3.84) below.

D | C  X p (t )  X (t ) |

(3.83)

X (t  1)  X p (t )  A  D

(3.84)

where A and C are coefficient vectors, t represents the current iteration, the position vector of the prey is represented by X p , and the position vector of a grey wolf is shown by X . To update the position of the search agents, Equations 3.85 and 3.91 are used.

D | C1  X   X |

(3.85)

D | C2  X   X |

(3.86)

D | C3  X   X |

(3.87)

100

X1  X   A1  ( D )

(3.88)

X 2  X   A2  ( D )

(3.89)

X 3  X   A3  ( D )

(3.90)

X (t  1) 

X1  X 2  X 3 3

(3.91)

With the help of these equations, a search agent updates its position in ndimensional search space. The position is updated according to

,

 and  . The

parameters used for GWO are given in Table 3.10. 3.5.14 Genetic Algorithm GA belongs to heuristic optimization family and is inspired from the genes of living organisms. GA works on the basis of iteration and having different possible iterations with different possible solutions (Logenthiran, Srinivasan, & Shun, 2012). The structure of GA consists of binary coded chromosomes which are randomly initialized. The ON/OFF state of the appliances is represented by the binary coded chromosomes pattern of GA and length of chromosomes show the total number of smart appliances represented in Equation (3.92). Chromosomes length = Number of household appliances

(3.92)

With the creation of initial population, the fitness function of GA is evaluated as the objective function of this study. The new population is generated by implementing mutation and crossover operators. The parameters used in this research work are presented in Table 3.11. Table 3.10: Parameters of GWO. Parameters

Value

Parameters

Value

Total iterations

50

Random vectors r1, r2

0,1

Population size

200

n

18



2 to 0 101

Table 3.11: Parameters of GA. Parameters

Value

Parameters

Value

Number of iterations

50

Probability of crossover

0.9

Population size

200

n

18

Probability of mutation

0.1

A whole new generation will be produced if crossover probability is 100%, while the new generation produced will be an exact copy of the parents if the probability of crossover is 0%. However, pre-mature convergence to the suboptimal solution is avoided by larger crossover rate for optimization problems, which is why 90% is the best crossover rate, as given in Equation (3.93) below. Probability of crossover = 0.9

(3.93)

The mutation process is used for randomness creation in the results. One or more genes are mutated in a chromosome from its original state. The probability of mutation is given below: Probability of mutation = 1-Probability of crossover

(3.94)

After crossover and mutation process, the generated population and its fitness are compared with the previous individuals. 3.5.15 Binary Particle Swarm Optimization BPSO is a discrete variant of PSO and consists of four main steps, i.e., particle’s initial position and initial velocity, and local and global best positions among the particles. The PSO randomly generates and disperses population in the search space. BPSO updates velocity and position by following Equations (3.95) and (3.96), respectively (Tuaimah, Abd, & Hameed, 2013; Zhu, Tang, Lambotharan, Chin, & Fan, 2012).

Vid (t )  Vid (t  1)  c1r1 ( Pbestid (t  1))  X id (t 1))  c2r2 ( gbestid (t 1)  X id (t 1)) (3.95)

102

where

X id (t ) , X id (t  1) , Vid (t ) , and Vid (t  1) are the position and velocity of

particle i in the d dimension at time slots t and t-1, respectively.

Pbestid (t  1) and

gbestid (t  1) are the best positions obtained by particle i and swarm in d dimension c1 and c2 are the two acceleration coefficients.

in time slot t and t-1, respectively.

r1 and r2 are arbitrary numbers between 0 and 1. The sigmoid function for position is given below.

sig (Vi t 1 ( j )) 

where Vi

t 1

1 1  exp(Vi t 1 ( j ))

(3.96)

shows the velocity of the particle. The parameters of BPSO are

explained in Table 3.12. 3.5.16 Wind Driven Optimization WDO is a heuristic optimization algorithm. Instead of particles in BPSO, it works on the basis of atmospheric motion of air parcels. WDO is differentiated from the other heuristic techniques due to the existence of forces. Friction force resists the motion of air parcels in forward direction. Gravitational force is a straight up force. coriolis force deflects the air parcels in the atmosphere. Pressure gradient moves the parcels in the forward direction. These forces can be mathematically represented in Equations (3.97) and (3.100) (Bayraktar, Komurcu, & Werner, 2010). Table 3.12: Parameters of BPSO. Parameters

Value

Number of iterations Swarm size Maximum velocity Minimum velocity

Parameters

50 200 4 -4

Initial weight constant

Final weight constant Local pull Global pull n

2

FCr  2

(3.97) 103

Value 0.4 2 2 18

where velocity,

FGv    g

(3.98)

Fprg  

(3.99)

FFr  

(3.100)

FCr and  show Coriolis force and earth rotation, respectively.  is wind

FGv is gravitational force,  is air density,  is air finite volume, g is

acceleration of gravity, pressure gradient force is presented by Fprg ,  shows pressure gradient,

FFr is friction force and  is friction coefficient. Equations

(3.101) and (3.102) (Bayraktar, Komurcu, & Werner, 2010) represent the air parcel’s position and velocity.



p i 1

c ip 1 p  ((1   ) i  gx  [ RT |  1| ( xgbest  xt )]  r r p

p i

(3.101)

and

xip1  xip  ip1 where

(3.102)

 ip and  ip1 represent the current and new velocity of the air parcels, p

p

respectively. xt and xi 1 show the current and new positions of the air parcels, respectively. xgbest is the global best position. R, T,  , g and c are universal gas constant, temperature, the coefficient of friction, gravity and Coriolis force, respectively. r is a variable value for the rank of air parcels. Random solutions are created by WDO and then a new population is generated by updating the velocities and evaluating the fitness function. To attain an optimal appliances pattern, the fitness function of updated and previous generation of air

104

parcels are compared. The term pressure in WDO is fitness function, i.e., in GA, PSO and BPSO. The parameters of WDO are described in Table 3.13. Table 3.13: Parameters of WDO. Parameters

Value

Parameters

Value

Total iterations

50

vmax

0.3

Population size

200

Universal gas constant

3

dimMin

-5

n

18

dimMax

5

Coefficient of friction

0.4

Gravity

0.2

vmin

-0.3

3.5.17 Wind Driven Genetic Algorithm WDGA is the hybrid of GA and WDO. In WDGA, first, steps of the WDO are performed, i.e., initialization of population and selection. Then, instead of using velocity updating step of WDO, crossover and mutation operators from GA are performed for the generation of new population to ensure diversity in the solution. The reason for replacing velocity step of WDO with crossover and mutation operators of GA is the increase in time complexity which degrades the performance of WDO when the input value is large. Thus, in our work, WDGA generated random solutions in the form of 0 and 1 (0 and 1 show the appliance status OFF and ON, respectively). After initialization of population, these solutions were evaluated according to our defined objective function (minimum electricity cost and PAR) in Equation (3.79). The proposed WDGA algorithm is given in Algorithm 4 and its parameters are given in Table 3.14. Table 3.14: Parameters of WDGA. Parameters Number of iterations

Value

Parameters

50

Gravity Coefficient of friction

Value 0.2

Parcels size

200

Dimensions

[-1, +1]

Crossover rate

0.9

Maximum velocity

0.4

Mutation rate

0.1

Universal gas constant

3.0

105

0.4

3.5.18 Wind Driven Grey Wolf Optimization WDGWO is an optimization technique that is developed by GWO and WDO algorithms. The WDGWO works initially the same as GWO; however, the position of the search agents of the GWO is replaced by iterative velocity updating parameter of the WDO. In WDO, velocity updating step is better for new generation as compared to GWO updating method. This hybrid version of GWO and WDO gives better results than GWO and WDO, separately. In our work, the initial population (in the form of 0 and 1) was generated on the basis of wolves, i.e., alpha, beta, delta and omega. The selection was also performed according to our objective function by GWO and velocity updating is performed to regenerate best population. In every iteration, our proposed WDGWO found local best solutions and finally it found the global best on the basis of local solutions. The pseudocode of the WDGWO is shown in Algorithm 5. Table 3.15 shows the parameters that are used for simulations in WDGWO. Table 3.15: Parameters of WDGWO. Parameters Number of iterations Population size Dimensions

Value

Parameters

50 200 [-1, +1]

Value

Gravity

0.2

Coefficient of friction

0.4



2 to 0

Maximum velocity

0.4

Random vectors r1, r2

0, 1

Universal gas constant

3.0

n

18

3.5.19 Wind Binary Particle Swarm Optimization In this section, we discuss WBPSO algorithm which merges WDO and BPSO. The WBPSO algorithm is more efficient in solving optimization problems as compared to WDO and BPSO because WBPSO consists of the best properties of both aforementioned algorithms. The pseudocode of this hybrid algorithm is shown 106

in Algorithm 6. The WBPSO works similarly to BPSO, i.e., random population generation, finding the local and global best positions of particles by changing velocity, and updating position of the particles in each iteration. When the stopping criteria are fulfilled, the algorithm stops working and generates random solutions that are different from the previous results. In our work, population generation step was performed by BPSO in binary form (0 and 1 show OFF and ON status of each appliance, respectively). After population generation, best solutions were selected on the basis of our defined objective function in Equation (3.79). WDO wind pressure was used for new solution generation. Finally, the global best solutions for each hour were selected having minimum electrical energy cost and PAR. The parameters of WBPSO are denoted in Table 3.16. Table 3.16: WBPSO parameters Parameters

Value

Parameters

Value

Number of iterations

50

Final weight constant

Population size

200

Local pull

2

Global pull

2

Dimensions

[-1, +1]

Maximum velocity

0.4

Gravity

Minimum velocity

-4

Coefficient of friction

Initial weight constant

2

n

0.4

0.2 0.4 18

Algorithm 4: WDGA 1: Input: Set of appliances



or population

2: Initialization of GA parameters: peak hour, off peak hour, t = 0, H, vmax, vmin, no of Iteration 3: Initialization of WDO parameters: dimMin, dimMax, dim, param.rt, param.g, param.alp, param.c 4: for t  1  24 do 5: for h  1  H do 6:

for p  1  P do

7:

Generate population randomly

8:

Fitness calculation 107

9:

Select best population, pop save in pop1

10:

Status check of appliance using peak hour and off peak hour

11:

if t == peak hour then

12:

Shift on RESs and BBS OR wait for off peak hour

13:

if Consumption == high then

14:

Check LOT of all apps until it is 0

15: 16:

end if end if

17:

end for

18:

Generate new population

19:

Replace the genetic operators by particles pressure

20:

Evaluate and find air parcels (population) pressure

21:

for K  1  swarm do

22:

for h  1  n do x (K, h) = (dimMax - dimMin) * ((x (K, h) +1). /2) + dimMin

24: 25:

Pres (K, h) = sum

( x ( K , h) 2 )

end for

26:

end for

27:

Save air parcels value in pop2

28:

Check and find air parcels velocity

29:

Vel = min (vel, maxV); and vel = max (vel, -maxV)

30:

Find and update air parcel positions

31:

x = x + vel; and x = min(x, 1.0) and x = max (x, -1.0)

32:

Finding best particle in population

33:

Globalpres, indx = min (pres); and globalx = x (indx, :)

34:

Find min location for this iteration

35:

Minpres, indx = min (pres); and minpos = x (indx, :)

36:

Rank the air parcels: sortedpres rankind = sort (pres)

37:

Sort the air parcels position, velocity and pressure

38:

Pres = sortedpres

108

39:

Updating the global best

40:

Better = minpres < globalpres

41:

if Solution = better then

42:

Globalpres = minpres

43:

Globalpos = minpos

44:

end if

45:

Save the velocity and position value in pop3

46:

Select from pop2 and pop3

47:

New velocity and position of air parcels

48:

if Solution == infeasible then

49:

Update result

50:

Update with sol in pop2 and pop3

51:

end If

52:

Update pop preeminent solution

53: end for 54: end for

Algorithm 5: WDGWO 1: Input: Set of appliances



or population

2: Initialization GWO parameters: Max iter, Np, D, alpha, beta, delta, search agents 3: Initialization WDO parameters: dimMin, dimMax, dim, param.rt, param.g, param.alp, param.c 4: Arbitrarily initialize the position of exploration agents, i.e., positions = rand (Np, D) 5: Assess the position of search agents 6: while

iter  itermax do

7: for i  1: size ( positions,1) do 8:

Compute objective function for each search agent

9:

Fitness = sum (electricity cost * positions)

10:

Update alpha, beta and delta

11:

if Fitness  alpha  score then 109

12:

Alpha  score  fitness

13:

Alpha  pos  positions(i :1)

14:

end if

15:

if Fitness  alpha  score and fitness  beta  score then

16:

Beta  score  fitness

17:

Beta  pos  positions(i :1)

18:

end if if Fitness  alpha  score andfitness  beta  score and

19:

fitness  delta  score then

20:

Delta  score  fitness

21:

Delta  pos  positions(i :1)

22:

end if

23: end for 24: a  2  l *((2) / Max  iter ) ; a value linearly from 2 to 0 25: for i  1: size ( positions,1) do 26:

for j  1: size ( positions, 2) do

27:

r1, r2 arbitrarily prepare the value between 0 to 1

28:

Vel = maxV * 2 * (rand (Np, D)-0.5)

29:

for i  1  Np do

30: 31:

for j  1  D do Velot (i, j) = vel (i, j) Vel (i, j) = (1 - alp) * vel (i, j) - (g * positions (i, j)) + abs (1 - 1/i)*(((positions (i, j) positions (i, j))).* RT) + (c * velot (i, j) / i) if ((vel (i, j) < vmax) and (vel (i, j) > vmin)) then

35:

Velot (i, j) = vel (i, j)

36:

else if (vel (i, j) < vmin) then

37: 38:

Vel (i, j) = vmin else if (vel (i, j) > vmax) then 110

39:

Vel (i, j) = vmax

40:

end if

41:

Position updating

42:

Sig (i, j) = 1/ (1+exp (-vel (i, j)))

43:

if rand (1) < sig (i, j) then

44:

Positions (i, j) = 1

45:

else

46:

Positions (i, j) = 0 Update the positions and find the optimal solution

48:

end if

49:

end for

50: 51:

end for end for

52: end for 53: end while Algorithm 6: WBPSO 1: Input: Number of particles, swarm size, t max , electricity price, LOT, 2: Appliance power consumption rating, vmax, vmin, no of iter, c1, c2, param.RT, param.g, param.alp, param.c, dimMin, dimMax, dim, randomly generate the particles’ 4: Locations and velocities: Pgbest   5: for t  1 to swarmsize do 6:

Initialize (swarmsize, tbits)

7:

Pvel  randomvelocity ()

8:

Ppos  randomposition (swarmsize)

9:

Plbest  Ppos

10: end for 11: 12:

for h = 1 to 24 do Authenticate constraints 111

13:

for i = 1 to M do if f ( i )  f ( plbest , i ) then

14:

plbest , i   i

15: 16:

end if

17:

if f ( Plbest , i )  f ( Pgbest , i ) then

Pgbest , i  Plbest , i

18: 19:

else

Pgbest , i  Pgbest , i

20: 21:

end if

22:

Reduce one from the TOT of the occupied appliance

23:

if

24:

Ecost > Emaxcost then if ETOTRESs  Eloadh then

25: 26:

Change the load to RESs and BBS else

27: 28:

Consume the electric grid energy end if

29:

end if

30:

Return Pgbest , i

31:

Vel = maxV * 2 * (rand (swarm, n)-0.5)

32:

for i = 1: swarm do

33: 34:

for j = 1: n do Velot (i, j) = vel (i, j) Vel (i, j) = (1 - param.alp) * vel (i, j) - (param.g * pres (i, j)) + abs (1- 1/i) *(((pres (i, j) - pres (i, j))).*param.RT) + (param.c * velot (i, j) /i)

37:

if ((vel (i, j) < = vmax) and (vel (i, j) >= vmin)) then

38:

vel (i, j) = vel (i, j)

39:

elseif (vel (i, j) < vmin)

40:

vel (i, j) = vmin 112

41:

if (vel (i, j) > vmax) then

42:

vel (i, j) = vmax

43:

end if

44:

end if

45:

Sig (i, j) = 1 / (1 + exp (-vel (i, j)))

46:

if rand (1) < sig (i, j) then

47:

x (i, j) = 1

48:

else

49:

x (i, j) = 0

50:

end if

51:

end for

52:

end for

53:

Check velocity: vel = min (vel, maxV); vel = max (vel, -maxV)

54:

Update air parcel positions: x = x + vel; x = min (x, 1.0); x = max (x, -1.0) 55:

Evaluate population: (pressure)

56:

Finding best particle in population

57:

Globalpres, indx = min (pres); globalx = x (indx, :)

58:

Min location for this iteration

59:

Minpres, indx = min (pres); minpos = x (indx, :)

60:

Rank the air parcels

61:

Sorted-pres rank-ind = sort (pres)

62:

Sort the air parcels position, velocity and pressure

63:

Pres = sorted-pres

64:

Updating the global best

65:

Better = minpres < globalpres

66:

if Better then

67:

Globalpres = minpres

68:

globalpos = minpos

69:

end if

70: end for 71: end for 113

3.6 ANALYSIS OF HYBRIDIZATION OF HEURISTIC TECHNIQUES FOR RESIDENTIAL LOAD SCHEDULING We have discussed the solution 4 according to the sub-problem 4 (i.e., section 1.2.4) defined in the chapter 1. System model and the relevant descriptions are elaborated in the subsequent subsections. 3.6.1 System Model Proposed model specifications include the following: 1. Power elastic loads have some flexibility in operation within predefined time slots. These appliances operate between the minimum and maximum power within the scheduling time horizon. For example, air conditioner and refrigerator can regulate their power consumption from the minimum to the maximum power. 2. Time elastic load have flexibility in their operating time. They can either be interruptible or non-interruptible. The interruptible appliances can be delayed and interrupted if required such as washing machine and clothes dryer. Alternatively, non-interruptible appliances can only be delayed before it starts the operation such as electric kettle and toaster. 3.6.2 Mapping of Load Scheduling To Multiple Knapsack Problem In this section, scheduling problem is mapped to MKP. MKP in engineering and computer science is a combinatorial problem, i.e., finding an optimal item from the set of items. It is a standard mathematical technique in which many optimization problems are mapped. MKP is a generalized form of a single knapsack with multiple instances of the items. It has m knapsacks and a group of n items. Each item in this set has two attributes: the value of the object and weight. Every knapsack has a capacity constraint that represents the maximum mass that it can support. The objective of the multiple knapsacks is to find the subset of the items that can be filled within the knapsacks such that the value of the objects inside the knapsack is maximized. The mapping of scheduling problem to a multiple knapsacks is as follow: 114

• The time intervals t correspond to m knapsack. • The number of appliances correspond to n objects that must be packed within the knapsack. • The electrical energy consumption of appliances correspond to the weight of each object. • The cost of energy consumed corresponds to the value of the object in specific time-slots. • The utmost energy that can be drained from the grid at any time correspond to the capacity of the knapsacks. For the consumer, it ensures that the electrical energy cost can be restricted, and for the utility company, this boundary ensures that the grid is not over-burdened. In proposed system model, this limit is considered for a single household. 3.6.3 Mathematical Modeling of Objective Function and Constraints The description of mathematical modeling of electrical energy consumption, cost, UC and PAR are as follow: 3.6.3.1 Energy consumption model The electrical energy consumption is the energy consumed by the home appliances during scheduling time horizon. As discussed earlier, there are two types p

t

of appliances: power elastic appliances Ae and time elastic appliances Ae . The energy consumed at each time slot and during the day is given by the following Equations (3.103-3.109):

Ecj (t )  Pr j  X t , 24

(3.103)

N

ETpe   Ecj (t ),

(3.104)

Eca (t )  Pra  X t ,

(3.105)

t 1 j 1

24

N

E   Eca (t ), a T

(3.106)

t 1 a 1

115

Ecb (t )  Prb  X t , 24

(3.107)

N

ETb   Ecb (t ),

(3.108)

t 1 b 1

ETte  ETa  ETb .

(3.109)

The aggregated daily energy consumption of all appliances is calculated using the Equation 3.110:

ET  ETpe  ETte .

(3.110)

Figure 3.12: Functional Model 3.6.3.2 Energy cost model For electricity cost calculation, utility offered various pricing schemes such as ToU, RTP, CPR and CPP to benefit both utility and consumers (Boopathy, & Sivakumar, 2014). However, 60% of benefit for 2009 come from altered pricing schemes, as predicted by federal regulatory energy commission (FREC) (Primer, 2014). The energy consumed by the consumers is charged by the utility with respect to these pricing schemes. In this work, combined RTP and IBR are used for electricity cost calculation since in case of merely RTP, there is a chance that peaks will arise in demand during off-peak hours. The peaks will arise during off-peaks hours because all the load burden will be in off-peaks hours having price variations 116

in each hours. The cost paid by the consumers to the utility for the energy consumption can be calculated using the Equations 3.111 and 3.112: 24  N  CTpe     Ecj (t )  X t   (t ) , t 1  j 1 

(3.111)

24 N  N  CTte    ( Eca (t )   Ecb (t ))  X t   (t )) . t 1  a 1 b 1 

(3.112)

The total cost, i.e.,

CT of the consumer energy consumption can be calculated

using the Equation 3.113:

CT  CTpe  CTte .

(3.113)

3.6.3.3 Peak-to-average ratio The utility gives motivations to the consumers to move some load from on-peak to off-peak hours in order to reduce the peaks in demand that results in PAR reduction. PAR is the proportion of peak demand to average demand of the consumers during the scheduling time horizon. It is beneficial for both utility and consumers because it balances the demand curve and tries to manage the space between demand and supply. The PAR is considered by the following formula as in Equation 3.114:



 max Ecj (t ), Eca (t ), Ecb (t ) PAR  24    ET 

  .  

(3.114)

3.6.3.4 Waiting time The waiting time is the time for which the appliances must wait before starting operation. The waiting time defines the comfort or discomfort of the consumers. For example, if the appliance starts operation at a later time, it has larger waiting time. There is a trade-off between the waiting time and electricity cost. When the consumers wait more during peak load, they pay less cost and those consumers, who

117

do not tolerate wait, in fact, pay more. The formula of waiting time emerges as in Equation 3.115:

wi 

ti   i . i  Tso,i   i

(3.115)

For the consumers, the waiting time is expected to be as smaller as possible. In this paper, the waiting time along with the electricity cost are minimized. 3.6.4 Optimization Problem Formulation From the optimization viewpoint, it is wanted for the consumer to make use of the offered capability provided by the utility. The consumer utilizes the available energy in such a manner that the sum of the utility functions is enhanced and the electricity cost is minimized. The general objective function of residential consumer load scheduling is to reduce the electricity cost and PAR without compromising the comfort of the consumer. Objective function is prepared as an optimization problem using the Equations 3.116, 3.117 (a-c):

min  CT , wi , PAR  ,

(3.116)

subjected to;

ET  Capacity,

(3.117a)

ETsch  ETunsch ,

(3.117b)

Tso,sch  Tso,unsch .

(3.117c)

The constraint in Equation 3.117a defines the capacity, which ensures that electricity grid is not overstressed as well as provides control of the total energy usage of a home. The energy consumption of unscheduled load is equal to the scheduled load as indicated in Equation 3.117b. Equation 3.117b ensures the proper scheduling of the household load. Constraint defined in Equation 3.117c is the completion constraint. The constraint 3.117c ensures that the operation of the 118

appliances is completed within the specified time-slots. Constraint in Equation 3.117c also provides support for fair comparison of the scheduled load with the unscheduled load. 3.6.5 Proposed System Model The proposed system considers one electrical power system which is comprised of both supply and demand side with several consumers as shown in Figure 3.13. Each consumer is equipped with energy consumption scheduling unit (ECSU), smart meter, remote control, in-home display (IHD) and control. The ECSU is the key factor in HEM to control the consumer energy consumption and to coordinate each consumer with the utility company. The ECSU is connected to the utility company through the network interface such as local area network as shown in Figure 3.14. The ECSU receives control parameters setting from users through the user interface and pricing information from the utility to schedule household appliance’s consumption behaviour. Each user inputs control parameters such as starting time, finishing time and length of operation time, etc., through the user interface to the scheduling unit. The scheduling unit receives price information from utility, forwards control parameters and price information to managing unit. Using historical data and data received from scheduling unit; the managing unit schedules appliances consumption behavior using optimization techniques. Finally, ECSU exchanges the schedule with the utility to optimally control the consumer’s consumption behavior as shown in Figure 3.12. The scheduling time scope is divided into T time slots, where T = {1, 2, 3,…, T}. The utility generates day-ahead RTP signal. This division and day-ahead RTP is based on the behavior of consumers and their demand patterns, such as on-peak time slots, off-peak time slots, and mid-peak time slots. The load demand is classified into two types: elastic load and inelastic load. Moreover, the elastic loads is further classified into two classes: time elastic appliances and power elastic appliances. Each time elastic appliance can either be interruptible or non-interruptible. The operation of interruptible appliances can be 119

delayed, interrupted and adjusted or moved to the time slots other than on-peak time slots, while for non-interruptible appliances, it is only feasible to holdup its operation, when needed. Conversely, there is the inelastic load, having price inflexible nature. The detailed description of appliances is as follow.

Figure 3.13: Abstract view diagram of power flow Let the group of appliances are indicated by: A  {Ae , Ae } , such that, t

p

Aet  {Aten , Atein } . Aet is the set of time elastic appliances and Aep are the set of power elastic appliances. For each appliance i , the current position, St  (rt , wt ) and the i

n

n

i

position at the next time slot, St 1 is defined. In the time elastic appliances, the operation of interruptible appliances can be adjusted and interrupted, if necessary. The initial position and the position at next time slot for interruptible appliances can be modeled as shown below in Equations 3.118 and 3.119:

Sti  {Tso ,     Tso  1},

120

(3.118)

r n , wn  1 if X t  0, wtn  1, Sti1   t n t n n rt  1, wt if X t  1, rt  1.

(3.119)

In time elastic appliances, the non-interruptible appliances can only be delayed on the user requirement, as discussed above. The initial position and position at next time slot for non-interruptible appliances are modeled as in Equations 3.120 and 3.121:

Sti  {Tso ,     Tso  1},

(3.120)

r n , wn  1 if X t  0, wtn  1, Sti1   t n t n n rt  1, wt if X t  1, rt  1.

(3.121)

The power elastic appliances have elasticity in their power rating, and tolerate flexibility in their operation time. The position of power elastic appliances at the next time slot is modeled as in Equation 3.122:  p i if X t  1, 8 AM  t  9 PM, Sti1   max  pmini if X t  1, 10 PM  t  7AM.

(3.122)

The inelastic appliances start operation immediately and need to be power-on at all times during the day. The initial position and the position at next time slots are given as in Equations 3.123 and 3.124:

Sti  (Tso, 0),

S

i t 1

(3.123)

rt n , 0 if X t  1, rt n  1,  0, 0 otherwise.

(3.124)

Among the various pricing schemes discussed below, RTP scheme is chosen because it has more flexibility for appliances scheduling as compared to the ToU tariff, CPP and CPR. To avoid the making of peaks during off-peak hours, collective RTP and IBR pricing scheme are considered. We assume that the utility has no control over the consumer’s consumption and it may only influence the load by providing price flexibility. Load synchronization and making of peaks during off121

peak hours can be avoided by adopting combined RTP with IBRs, where the marginal price has a direct relationship with the load. This combined pricing scheme encourages the consumer to shift the load from on-peak to off-peak hours in order to reduce electricity cost and PAR.

Figure 3.14: System architecture Using this combined pricing scheme, electricity price depends on time and also on total load. Let,  ( L ) indicates the electrical energy price at time slot t, as a function t

of consumer’s consumption at that time slot as shown using Equation 3.125:  Rt if 0  Lt  Lth ,

 ( Lt )  

t bt if L  Lth .

Where

(3.125)

Rt is the electricity price, when the overall consumption is fewer than the

threshold of IBRs at time slot t, and

bt is the electricity price at time slot t, when the

total consumption exceeds the threshold of IBRs. 122

3.6.6 Optimization Techniques Many mathematical and heuristics techniques have been used for appliances scheduling. The main objectives of using these techniques for optimization are the reduction of electricity bill, PAR and energy balancing for demand and supply. In addition, optimizing grid stability, UC and to bring power quality are some other objectives. One of the key features of heuristics techniques is the low execution and computation time. These techniques provide a realistic answer to the problem. The details of these techniques used in this work are provided below. 3.6.7 Existing Optimization Techniques This section provides a brief discussion of GA, TLBO, BAT and FPA. The description of each algorithm is elaborated below. 3.6.7.1 Genetic algorithm GA is an adaptive heuristic search technique which is based on the evolutionary thoughts of inheritance and natural choice. GA is used to explain optimization problems and represents a smart exploitation of an arbitrary search. It tends to transfer random search towards a better performance region within the search space. GA performs a search in a multi-model state-space, large state-space or ndimensional surface and offers significant benefits over many other typical optimization techniques. To solve a problem over consecutive generations, GA simulates the survival of the fittest among individuals (Ramachandran, Srivastava, Edrington, & Cartes, 2011). Table 3.17: GA parameters. Parameters

Value

Parameters

Value

Number of iterations

500

Probability of crossover

0.9

Population size

200

Number of appliances

11

Probability of mutation

0.1

123

Each individual has a set of characteristics called chromosome or genotype bearing genes, which can be altered or mutated to produce individuals with best characteristics than the parents. Generally, GA candidate solutions are represented as a binary string of 0s and 1s; however, further encodings are also feasible (Smart Grid Australia, 2010). The genes pattern of chromosome represents the ON/OFF state of appliances and the length of chromosome represents the number of appliances. Once the population is created, i.e., the ON/OFF conditions of appliances are initialized in a special time slot. The fitness of each candidate solution is computed, related to the objective function defined for the optimization problem. After evaluating the fitness of each candidate solution, crossover and mutation process is applied to produce off-springs known as a new population which is better than the parents. The parameters for GA are described in Table 3.17. 3.6.7.2 Teaching learning based optimization This algorithm was originally proposed by (Zhao, Ding, Cooper, & Perez, 2014). TLBO is computationally very efficient because it’s parameters do not require any tuning. TLBO consists of a teacher and a student. Teacher is considered the most knowledgeable person. The teacher shares its knowledge with the students to improve their output or performance. The quality of the student knowledge or performance can be found by evaluating their grades. Moreover, the students also improve their knowledge by discussion, which ultimately improves their performance (Cook, 2012). TLBO is a populationbased technique motivated by the teaching and learning environment of the classroom. TLBO includes of two stages: teacher phase and student phase. In the first phase, a teacher is chosen as the best solution for the population and the remaining population is considered as students. To raise the level of the students, the teacher gave his information to the students for changing the average value of the students knowledge. The mean difference between the teacher and the student’s knowledge in a specific subject can be given by the following Equations (3.126 - 3.128):

MDi  ri (Meannew  Tfactor  Meani ), 124

(3.126)

T factor  round [1  rand (0,1){2  1}],

X new,i  X old ,i  MDi . Where students,

(3.127) (3.128)

MDi presents the mean difference of knowledge between the teacher and

Meannew presents the outcome of the best learner in particular subject. ri

indicates the random number between 0 and 1. Tfactor presents a teaching factor and its value can be 1 or 2 and is a heuristic step determined arbitrarily with equal possibility. The value of the teaching factor is decided by Equation (1b). The updated value of X new,i is chosen in the population if it is better than the X old ,i . In the student phase, the students intermingle with each other and the student with less knowledge gain knowledge from the student with additional knowledge. Two students, i.e., s1 and s2 are arbitrarily selected from the initial populace such that X total s1,i  X total s1,i . However, the optimization problem can be given as below using Equations 3.129 and 3.130.

X fnew,s1,i  X new,s1,i  ri ( X new,s1,i  X new,s 2,i ), ifX new,total s1,i  X new,total s1,i ,

(3.129)

X fnew,s1,i  X new,s1,i  ri ( X new,s 2,i  X new,s1,i ), ifX new,total s 2,i  X new,total s 2,i . (3.130) The solution X fnew,s1,i is selected, if its results are better than the existing one, otherwise discarded. These selected values are given as input to the teacher phase in the next iteration. This procedure repeats until the stopping conditions are reached. 3.6.7.3 BAT The BAT is a meta-heuristic technique used for global optimization. It was developed by Xin-She Yang in 2010 and motivated by the echolocation behavior of microbats. The changing pulse rates of discharge and loudness of the bat technique helps in finding the particular location via sound waves. The idea of the echolocation 125

of microbats can be explained as follows: Each cybernetic bat flies arbitrarily with a velocity

vi at location (solution) xi . The occurrence and loudness of the bat vary

continuously. It changes frequency, pulse emission rate and loudness; as it searches and finds its prey via local arbitrary walk. Choice of the best prey lasts until certain ending conditions are met. To control the active behavior of a swarm of bats, essentially a frequency-tuning method is used. In BAT algorithm, tuning algorithmdependent parameters are very essential for controlling the balance between exploration and exploitation (Davidoff, Lee, Yiu, Zimmerman, & Dey, 2006). 3.6.7.4 Flower pollination algorithm Flower pollination is an interesting process in the natural world. FPA is motivated by the fertilization process of flowers in flowering plants. The aim of the flower fertilization is the existence of the fittest and the best reproduction of plants. This is actually an optimization process of the flowering plants. There are two types of pollination: biotic and abiotic. Biotic pollination takes place in 90% of the flowering plants. Insects and animals act as pollinators which transfer pollen grain in flowering plants. Pollinators are also called pollen vectors, which are very diverse in nature. Estimated 200,000 of pollinators exists such as animals, birds, bats and insects. Cross and biotic pollination is a global pollination process. Due to random flights, the pollinators carry pollens to various places. Local pollination is considered as abiotic and self-pollination process. A switch probability p



[0, 1] controls the

global and local pollination process. Due to the wind and physical proximity, local pollination has p, as a fraction of the overall pollination process. 3.6.8 Proposed Optimization Techniques A brief overview of the proposed techniques: GTLBO, FBAT, FTLBO and FGA is provided in this section. In this work, basically four heuristic algorithms are used, i.e., GA, TLBO, FPA and BAT. Based on these algorithms, four hybrid algorithms, i.e., GTLBO, FTLBO, FBAT and FGA are proposed.

126

3.6.8.1 Genetic teaching learning based optimization GTLBO is a hybrid technique that is made by merging the parameters of GA and TLBO. TLBO performs better in searching of an optimal solution as well as in utilization mode, i.e., discovery the best solution in indigenous search space; conversely, TLBO performs very poor in search mode. It is designed for fast and local search, while in global search, the searching time is too long. So, there must be a balance between exploration and exploitation to find the best solution. GA performs best in exploration mode (global search) and as well as good convergence rate. GA have the efficiency to search in large spaces without trapping in local optima. In scheming the algorithm, search and utilization modes are two significant facets that are taken into concern. To overcome the imbalance between exploration and exploitation mode of TLBO, hybrid technique is proposed by relating the crossover and mutation operator of GA to TLBO. Hybrid techniques, initially, works as TLBO. After updating the population in learner mode, crossover and mutation are applied. Then new vectors are added to the population to calculate fitness. The procedure continues until end is reached. The hybrid method takes the benefit of both exploration and exploitation mode of TLBO and GA. The proposed technique is able to maintain an equilibrium between indigenous and global search by applying crossover and mutation operator of GA to TLBO as well as show better convergence rate. Algorithm 7: GTLBO 1: Initialize number of students, number of subjects, i.e., variables, d, termination criteria 2: Calculate the mean of each variable or subjects 3: Find the best solution and best variable value 4: while ending condition not met do 5: for each learner X, of the class do 6:

Teaching phase = round (1+rand (0, I))

7:

for j = 1: d do

8:

NewX ij = X ij + rand (0, 1)  (teacher (j) - teaching phase  mean (j) 127

9:

end for

10:

Accept

newX i if f (newX i ) is improved than f ( X i )

11: end for 12: for each learner

X i of the class do

13:

Arbitrarily select one learner X k , such that i  k

14:

if

f ( X i ) better f ( X k ) then

15:

for j = 1: d do

newX ij  X ij  rand (0,1)  ( X ij  X kj )

16: 17: 18:

end for else

19:

for j = 1: d do

newX ij = X ij +rand (0, 1)  ( X kj  X ij )

20: 21:

end for

22:

end if

23:

Perform GA crossover operation on newX ij

24:

Perform mutation operation on the new population

25:

Use objective function to take new population as input after crossover and mutation

26:

if off-spring is better than previous population then

27: 28:

Accept new population else

29:

Transfer to the next repetition and repeat the process

30:

Until termination criteria reached or first learner phase end

31:

end if

32:

From learning phase one accept results and run second learning phase

33:

Perform crossover and mutation operation and run objective function

34:

Repeat the process for each learning phase

35:

Accept

newX i if f (newX i ) is better than f ( X i )

36: end for 128

37: Update the teacher and the mean 38: end while 3.6.8.2 FBAT FBAT is a hybrid algorithm made of FPA and BAT. Originally, the population is arbitrarily generated for finding initial finest solution using objective function. After evaluating the population; the BAT velocity, frequency and position steps are replaced by the main steps of FPA, i.e., local pollination is started and random flowers are found in the neighborhood. Then new solutions are evaluated by checking their fitnesses, if fitness improves, i.e., better solutions are found, then update the existing solutions and update the current global best solution. In the same manner, update the search space accordingly. The hybrid technique, i.e., FBAT performance is enhanced than FPA and BAT in terms of cost, PAR and discomfort reduction. The daily cost reduced by FPA, BAT and FBAT are 3.87%, 12.32% and 25.23%, respectively. Hence, the daily cost reduction by FBAT is more than FPA and BAT algorithms. In addition, the daily discomfort and PAR reduction by FPA and BAT are 30.91%, 5.55% and 48.36%, 2.46%, respectively, while 22.18%, and 27.16% by FBAT. PAR and daily cost are reduced by FBAT as compared to FPA and BAT; however, the daily discomfort is reduced less by FBAT as compared to FPA and BAT due to trade-off between cost and daily discomfort. Algorithm 8: FBAT 1: Population size, number of appliances d, frequency q, velocity v, no of iteration, 2: probability switch, maximum and minimum frequency 3: Initialize the population/solutions 4: Generate population 5: Find the initial best solution 6: Initialize the best solution 7: Declare coefficients of objective function 8: Evaluate the population 129

9: FPA parameters replaces the BAT frequency, velocity and position calculation method 10: for j = 1: Niteration do 11: for i = 1: n do 12:

Local pollination will start

13:

if rand > p1, then

14:

L = Levy (d);

15:

dS = L.*(Sol (i, :)-gbest)

16:

Snew (i, :) = Sol (i, :) + dS

17:

else

18:

Epsilon = rand

19:

Find random flowers in the neighbourhood

20:

JK = randperm (n)

21:

Snew (i, :) = Snew (i, :) + epsilon * (Snew (JK (1), :) - Snew (JK (2), :))

22:

Check if the simple limits/bounds are acceptable

23:

Sol1 (i, :) = simplebounds (Sol1 (i, :), Lb, Ub)

24:

Improvise a new harmony vector

25:

end if

26:

Evaluate new solutions by checking their fitnesses

27:

Fnew = Fun (Snew (i, :))

28:

If fitness improves (better solutions found), update the population

29:

if Fnew  F (i) , then

30:

Sol (i, :) = Snew (i, :)

31:

F1 (i) = Fnew

32:

end if

33:

Update the current global best

34:

if Fnew p1 then

14:

L = Levy (d)

15:

dS = L.* (Sol (i, :) - gbest)

16:

Snew (i, :) = Sol (i, :) + dS

17:

else

18:

Epsilon = rand

19:

Find random flowers in the neighbourhood

20:

JK = randperm (n) Snew(i,:) = Snew(i,:) + epsilon *(Snew(JK(1) ,:) - Snew(JK(2),:)) Check if the simple limits/bounds are acceptable

23:

Sol1 (i, :) = simplebounds (Sol1 (i, :), Lb, Ub)

24:

Improvise a new harmony vector

25:

end if

26:

Evaluate new solutions by checking their fitnesses

27:

Fnew = Fun (Snew (i, :)) If fitness improves (better solutions found) and update the new population

29:

if Fnew  F (i)) then

30:

Sol (i, :) = Snew (i, :)

31:

F1 (i) = Fnew;

32:

end if

33:

Update the current global best

34:

if Fnew  fmin then,

35:

gbest = Snew (i, :) 132

36:

fmin = Fnew

37:

end if

38:

update the best value

39:

fmin

40: 41:

end for end for

42: end for 43: Update the teacher and the mean value 44: end while 3.6.8.4 FGA FGA is made by FPA and GA. In FGA, all the steps of GA are followed except two main steps, i.e., crossover and mutation, which generate new solutions in each iteration. The steps are replaced by the FPA pollination process as already discussed in section GTLBO and FTLBO. Algorithm 10: FGA 1: Initialize population, population size, number of iteration, crossover and 2: mutation rate, number of appliances, probability switch 3: Generate an primary random population 4: while Ending condition not met do 5: Iteration

 Maximum iteration

6: Iteration = iteration + 1 7: Define objective function; 8: Calculate the fitness of each individual via objective function 9: Select the individuals which are fittest among the population generated 10: FPA parameters replaces the TLBO teaching and learning phases 11: for j = 1: Niteration do 12:

for i = 1: n do

13:

Local pollination will start

14:

if rand > p1 then

15:

L = Levy (d) 133

16:

dS = L.* (Sol (i, :) - gbest)

17:

Snew (i, :) = Sol (i, :) + dS

18:

else

19:

Epsilon = rand

20:

Find random flowers in the neighbour hood

21:

JK = randperm (n) Snew (i,:) = Snew (i,:) + epsilon * (Snew (JK(1) ,:) Snew (JK(2),:))

23:

Check if the simple limits/bounds are OK

24:

Sol (i, :) = simplebounds (Sol (i, :), Lb, Ub)

25:

Improvise a new harmony vector

26:

end if

27:

Evaluate new solutions by checking their fitnesses

28:

Fnew = Fun (Snew (i, :))

29:

If fitness improves (better solutions found), update the new population if Fnew  F (i)) then

30: 31:

Sol (i, :) = Snew (i, :)

32:

F1 (i) = Fnew

33:

end if

34:

Update the current global best

35:

if Fnew