Sharing Economy in Energy Markets: Modeling, Analysis and Mechanism Design 9811676445, 9789811676444

Citation preview

Jianxiao Wang · Haiwang Zhong · Qing Xia · Gengyin Li · Ming Zhou

Sharing Economy in Energy Markets Modeling, Analysis and Mechanism Design

Sharing Economy in Energy Markets

Jianxiao Wang · Haiwang Zhong · Qing Xia · Gengyin Li · Ming Zhou

Sharing Economy in Energy Markets Modeling, Analysis and Mechanism Design

Jianxiao Wang School of Electrical and Electronic Engineering North China Electric Power University Beijing, China Qing Xia Department of Electrical Engineering Tsinghua University Beijing, China

Haiwang Zhong Department of Electrical Engineering Tsinghua University Beijing, China Gengyin Li School of Electrical and Electronic Engineering North China Electric Power University Beijing, China

Ming Zhou School of Electrical and Electronic Engineering North China Electric Power University Beijing, China

ISBN 978-981-16-7644-4 ISBN 978-981-16-7645-1 (eBook) https://doi.org/10.1007/978-981-16-7645-1 Jointly published with Science Press The print edition is not for sale in China (Mainland). Customers from China (Mainland) please order the print book from: Science Press. © Science Press 2022 This work is subject to copyright. All rights are reserved by the Publishers, whether the whole or part of the material is concerned, specifically the rights of reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publishers, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publishers nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publishers remain neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Foreword by Saifur Rahman

With the core idea of “access over ownership”, the concept of sharing economy has gained substantial popularity in the housing and transportation sectors in recent years. Sharing economy—the theme of this book—refers to a market model that enables individuals or entities to share their idle resources with others upon payment for the purpose of efficient resource allocation and social welfare maximization, which will bring new challenges and opportunities for deregulated energy markets. In recent decades, the importance of achieving a high share or even 100% renewable penetration has become a global aspiration. While the ever-increasing proliferation of renewables contributes to a more sustainable energy sector, considerable challenges remain for the secure and economic operation of electric power systems. For a long time, the popularized locational marginal pricing-based market settlement rule has been considered to give strong incentives for profit-seeking participants to make strategic bids for price manipulation, leading to market efficiency loss. On the other hand, an effective market model deserves further development for sharing the availability of ubiquitous idle demand-side energy resources. There remain enormous tasks to take a further step toward a deeper renewable penetration on the premise of the present electricity market framework and mechanism, which prompts us to ponder how to improve the utilization of resources. This book aims at conducting a systematic examination of the current research and practice of energy sharing and identifying the potential merits of such an emerging business model in the energy sector. In light of sharing economy, energy sharing can contribute to a more accurate match between energy supply and demand, thereby making efficient use of idle resources. Based on a fair and reasonable profit-sharing mechanism, Pareto improvement of an energy system or market can be achieved, which guarantees sufficient incentives for participants’ involvement. In this book, the authors analyze the modeling and application in various forms of energy markets, e.g., electricity spot markets, multi-area electricity markets, retail markets and integrated energy markets. In addition, the enabling technologies for the implementation of the energy sharing are discussed, which provides the readers with an explicit sense about the cyber-physical nexus.

v

vi

Foreword by Saifur Rahman

Hopefully, this book will provide a fundamental reference for the development of sharing economy-related technologies and business models in the energy sector. Saifur Rahman, Ph.D. Joseph Loring Professor and Director Virginia Polytechnic Institute and State University Arlington, VA, USA IEEE President-elect 2022 President, IEEE Power & Energy Society 2018 and 2019 IEEE Life Fellow

Foreword by Xiaoxin Zhou

The recent decades have witnessed China’s great efforts to a sustainable ecological environment and society. In September, 2020, the carbon neutrality target was declared with China committing to peak carbon dioxide emissions before 2030 and to achieve carbon neutrality before 2060. In March, 2021, China further declared to construct a novel paradigm of renewable-dominated power systems toward a lowcarbon and efficient energy transition. In this instance, China has set an ambitious goal for an over 1200 GW wind and photovoltaic portfolio by 2030. In addition to large-scale renewable clusters, distributed energy resource (DER) technology has been advocated as another promising solution to facilitate the accommodation of local clean energy in smart cities and rural communities, e.g., offshore wind power and waste to biomass. In a foreseeable future, the ever-increasing proliferation of renewables will pose great challenges to the secure and efficient operation of the power grids as well as the electric power industry reform in China. Traditional locational marginal pricingbased market framework has already raised concerns that the merit order effect of zero marginal cost renewables will bring down the electricity market prices. In addition, the design of distribution-level retail markets is arousing a public interest regarding how to manage large-scale intermittent DERs into wholesale markets. A series of energy policies and studies have been proposed to enhance the reliability of renewable-dominated power systems in a market-oriented fashion. The ambition of the authors of this book has been to produce a fundamental reference that can take advantage of sharing economy to improve Pareto efficiency of energy markets. Based on the core idea of “access over ownership”, energy sharing can be interpreted as the sharing economy in the energy sector, namely designing incentive-compatible market mechanisms for Pareto improvement by facilitating the utilization of idle energy resources via advanced information and communication technologies. For example, the capacity of a large-scale centralized storage can be shared among a set of customers for individual use, and a number of distributed storages can be aggregated as a single entity as well. Energy sharing is not only a novel business model, but a transformation of our thinking way. Such a concept enables the maximization of production by making full use of limited resources. I vii

viii

Foreword by Xiaoxin Zhou

think that the authors have made an admirable success in their objective and task. The chapters in this book present an up-to-date analysis and modeling for energy sharing in a comprehensive framework of energy markets, with a considerable innovation in terms of theories and practices. I hope that this book will provide well-founded guidance and direction for the research and refinement of sharing economy in the energy sectors. Xiaoxin Zhou China Electric Power Research Institute Beijing, China Academician of Chinese Academy of Sciences International Member of the United States National Academy of Engineering (NAE) IEEE Fellow

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Bibliometric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Concept of Energy Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 History and Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.2 Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.3 Taxonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Sharing Economy in Wholsesale Markets . . . . . . . . . . . . . . . . . . . . . . 1.4.1 Electricity Spot Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.2 Multi-area Electricity Markets . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.3 Integrated Energy Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5 Sharing Economy in Retail Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.1 Agent-Based Energy Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.2 Peer-To-Peer Energy Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.3 Integration of Distributed Energy Resources into Wholesale Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Enabling Technology and Business Models . . . . . . . . . . . . . . . . . . . . 1.6.1 Energy-Related Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.2 Information-Related Technology . . . . . . . . . . . . . . . . . . . . . . . 1.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1 2 5 5 6 6 8 8 9 10 11 11 12 13 14 14 17 19 21

2 Mechanism Design for Sharing Economy . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Wholesale Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Retail Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Profit Sharing Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Social Welfare Maximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Individual Rationality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

27 27 28 28 29 30 31 31

ix

x

Contents

2.3.3 Incentive Compatibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.4 Budget Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Profit Sharing Mechanism in Wholesale Markets . . . . . . . . . . . . . . . . 2.4.1 Wholesale Market Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.2 Sufficient N − 1 Power Generation Capacity . . . . . . . . . . . . . 2.4.3 Proof of the Mechanism Property . . . . . . . . . . . . . . . . . . . . . . . 2.4.4 Insufficient N − 1 Power Generation Capacity . . . . . . . . . . . 2.4.5 Major Challenges of VCG in Electricity Markets . . . . . . . . . 2.4.6 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Profit Sharing Mechanism in Retail Markets . . . . . . . . . . . . . . . . . . . . 2.5.1 Retail Market Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.2 Mechanism Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.3 Closed Form of Profit Sharing Mechanism . . . . . . . . . . . . . . . 2.5.4 Proof of the Mechanism Property . . . . . . . . . . . . . . . . . . . . . . . 2.5.5 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

32 32 33 33 34 35 37 39 40 41 42 43 45 45 48 51 51

3 Sharing Economy in Electricity Spot Markets . . . . . . . . . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Electricity Spot Market Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Mathematical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 Settlement Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Revenue Inadequacy Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Theoretical Analysis on Budget Imbalance . . . . . . . . . . . . . . . 3.3.2 Revenue Inadequacy Allocation Strategy . . . . . . . . . . . . . . . . 3.4 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Non-congested Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.2 Congested Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.1 IEEE 30-Bus System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.2 IEEE 118-Bus System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.3 Polish 2383-Bus System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

53 53 54 54 55 58 59 62 64 64 65 67 67 72 75 76 77

4 Sharing Economy in Multi-area Electricity Markets . . . . . . . . . . . . . . . 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 VSC-HVDC Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Multi-area Economic Dispatch Model . . . . . . . . . . . . . . . . . . . 4.3 Incentive Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Marginal-Pricing Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Incentive-Compatible Mechanism . . . . . . . . . . . . . . . . . . . . . .

79 79 81 81 83 85 86 86

Contents

xi

4.3.3 Revenue Inadequacy Allocation . . . . . . . . . . . . . . . . . . . . . . . . 4.3.4 Toy Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Decomposed Regional Sub-problem . . . . . . . . . . . . . . . . . . . . 4.4.3 Improved Lagrangian Multiplier . . . . . . . . . . . . . . . . . . . . . . . 4.5 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.1 Impacts of Strategic Bids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.2 Performance of the Proposed Mechanism . . . . . . . . . . . . . . . . 4.5.3 Impacts of Inter-area Transmission Capacity . . . . . . . . . . . . . 4.5.4 Impacts of Thermal Generation Flexibility . . . . . . . . . . . . . . . 4.5.5 3-Area 354-Bus Power System . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

87 89 91 92 92 94 96 97 97 99 100 101 102 104

5 Sharing Economy for Renewable Energy Aggregation . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Aggregation of Wind Farms and Concentrating Solar Power . . . . . . 5.2.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Offering Strategy Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.3 Profit Sharing Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Aggregation of Distributed Energy Resources in Energy Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Energy Sharing Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Profit Sharing Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.4 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Aggregation of Distributed Energy Resources in Energy and Capacity Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Energy Sharing Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.3 Profit Sharing Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1 Aggregation of Wind Farms and Concentrating Solar Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2 Aggregation of DERs in Energy Markets . . . . . . . . . . . . . . . . 5.5.3 Aggregation of DERs in Energy and Capacity Markets . . . . 5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

107 107 108 108 110 113

6 Sharing Economy in Energy Systems Integration . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Integrated Energy Sharing Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1 Status and Challenge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.2 Market Operation Practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

143 143 144 144 146

117 117 119 122 125 126 126 129 131 133 133 135 137 140 142

xii

Contents

6.2.3 Equilibrium-Based Integrated Energy Market . . . . . . . . . . . . 6.2.4 Arbitrage Models for Integrated Energy Markets . . . . . . . . . 6.3 Sharing Economy in Joint Electricity-Heat Markets . . . . . . . . . . . . . 6.3.1 Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Solar-Powered Heat Pump Planning Model . . . . . . . . . . . . . . 6.3.3 Pricing Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.4 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Sharing Economy in Joint Electricity-Gas Markets . . . . . . . . . . . . . . 6.4.1 Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.2 Modeling for Power-To-Gas . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.3 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Sharing Economy in Transportation-Energy Systems . . . . . . . . . . . . 6.5.1 Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.2 Model of Fuel-Cell Hybrid Electric Vehicle . . . . . . . . . . . . . . 6.5.3 Optimal Scheduling Model for Trans-Energy Systems . . . . . 6.5.4 Shortest Path Search Algorithm . . . . . . . . . . . . . . . . . . . . . . . . 6.6 Integrated Demand Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6.1 Basic Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6.2 Value Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6.3 Techo-Economic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6.4 Key Issues and Potential Research of IDR . . . . . . . . . . . . . . . 6.7 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.7.1 Sharing Economy in Joint Electricity-Heat Markets . . . . . . . 6.7.2 Sharing Economy in Joint Electricity-Gas Markets . . . . . . . . 6.7.3 Sharing Economy in Transportation-Energy Systems . . . . . . 6.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

147 149 150 151 152 154 156 158 159 160 163 167 168 169 173 177 177 178 181 182 183 187 187 189 190 192 192

7 Sharing Demand Side Resources for Regional Market Bidding . . . . . 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Sharing Demand Side Resources in Wholesale Markets . . . . . . . . . . 7.2.1 Co-optimization of Energy and Ancillary Service Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.2 Uncertainty Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.3 Optimal Bidding Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Sharing Demand Side Resources Toward Available Transfer Capability Enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3.1 Available Transfer Capability Evaluation Framework . . . . . . 7.3.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4 Sharing Demand Side Resources for Carbon Trading . . . . . . . . . . . . 7.4.1 Internet of Things Platform for Sharing Economy . . . . . . . . . 7.4.2 Model of Electric Vehicle Fleets . . . . . . . . . . . . . . . . . . . . . . . . 7.4.3 Optimal Bidding Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

195 195 196 196 198 199 203 203 204 208 208 210 211

Contents

7.5 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5.1 Wholesale Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5.2 Available Transfer Capability Enhancement . . . . . . . . . . . . . . 7.5.3 Energy and Carbon Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Sharing Non-wire Alternatives for Transmission Expansion Deferral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Overall Nodal Price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.1 Basic Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.2 Existing Transmission Cost Allocation Methods . . . . . . . . . . 8.3 Mechanism Design for Non-wire Alternative Planning . . . . . . . . . . . 8.3.1 Strucutral Transmission Cost Identification . . . . . . . . . . . . . . 8.3.2 Usage-Based Transmission Cost Allocation . . . . . . . . . . . . . . 8.3.3 Optimal Planning Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3.4 Solution Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.4 Sharing Non-wire Alternatives for Expansion Deferral . . . . . . . . . . . 8.4.1 Tri-level Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.4.2 Solution Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.5 Theorem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.6 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.6.1 Non-wire Alternative Planning . . . . . . . . . . . . . . . . . . . . . . . . . 8.6.2 Joint Planning for Non-wire Alternatives and Transmission Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Information and Communication Technology for Sharing Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 Cloud-Edge Computing Technology for Energy Sharing . . . . . . . . . 9.2.1 Tri-layer System Architecture of Energy Sharing . . . . . . . . . 9.2.2 Coordinated Demand Response Program . . . . . . . . . . . . . . . . 9.2.3 Sensitivity Analysis of Convex Optimization . . . . . . . . . . . . . 9.2.4 Lagrangian Multiplier Optimal Selection Approach . . . . . . . 9.3 Influencing Factor: Communication Connectivity . . . . . . . . . . . . . . . 9.3.1 Communication Topology Connectivity . . . . . . . . . . . . . . . . . 9.3.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4 Influencing Factor: Communication Reliability . . . . . . . . . . . . . . . . . 9.4.1 Communication Reliability Model . . . . . . . . . . . . . . . . . . . . . . 9.4.2 System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.3 Linearization Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xiii

214 214 217 221 224 224 227 227 229 229 230 231 231 233 236 239 242 242 247 250 251 251 260 267 268 271 271 273 273 275 279 282 285 285 286 288 289 292 293

xiv

Contents

9.5 Resilience Amidst Rare Weather Events . . . . . . . . . . . . . . . . . . . . . . . 9.5.1 Simplified Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5.2 Reliability Value and Optimal Investment Problem . . . . . . . . 9.5.3 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.6 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.6.1 Performance of Cloud-Edge Computing Algorithm . . . . . . . 9.6.2 Impact of Communication Connectivity on Energy Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.6.3 Impact of Communication Reliability on Energy Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

296 296 298 300 304 304 310 312 317 318

Chapter 1

Introduction

1.1 Background and Motivation The growing awareness of serious environmental challenges and energy shortage issues entails a renewable and sustainable energy transition. In recent decades, the importance of achieving a high share or even 100% renewable penetration has become a global consensus. Regarding the “30·60” Carbon–neutral Target, China declared that great efforts should be made to construct a novel paradigm of renewabledominated power systems [1]. China has set an ambitious goal for an over 1200 GW wind and photovoltaic (PV) portfolio by 2030, accounting for approximately 34% of the national total installed generation capacity [2]. In addition to large-scale renewable clusters, distributed energy resource (DER) technology has been advocated by many countries around the world as another promising solution to facilitate the integration of near-zero-emission (NZE) generation by matching regional supply and demand [3]. In California, the total installed capacity of DERs has exceeded 12 GW, with 33% of the electric load served by renewable energy since 2020 [4]. While the ever-increasing proliferation of renewables contributes to a more sustainable energy sector, considerable challenges have been posed to the secure and economic operation of electric power systems. Therefore, recent years have witnessed a wide variety of studies and practices striving to enhance the reliability of renewable-dominated power systems in a market-oriented fashion. For example, China has enacted a series of mechanisms to promote the consumption of renewable energy in the Northwest and Southwest in recent years [5]. Another example is the pilot project initiated in US, allowing DER end-users to participate in market bidding for peer-to-peer (P2P) transactive energy [6]. There remain enormous tasks to take a further step toward high-share renewable penetration on the premise of the present electricity market framework and mechanism. On the one hand, in most of the wholesale markets around the world, locational marginal pricing (LMP)-based market clearing and settlement are popularized, depending on the marginal cost of balancing the last-MWh load demand.

© Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_1

1

2

1 Introduction

However, such a paradigm has already raised concerns that the merit order effect of zero marginal cost renewables will bring down the electricity market prices. Meanwhile, profit-seeking thermal generators have incentives to make strategic bids for price manipulation, resulting in market efficiency loss. Recent empirical evidence shows that the strategic bidding of the thermal generators in China’s load centers may distort global-optimal dispatch, thereby leading to additional curtailment of the wind and PV power from the northwest [7]. On the other hand, the design of distribution-level electricity markets is arousing a public interest regarding how to manage large-scale intermittent DERs into wholesale markets. Despite the success of the world’s first blockchain-based solar power trading, it remains challenging to efficiently organize ubiquitous DERs owing to considerable transaction costs. In contrast to the generators in wholesale markets, end-users may be reluctant to serve as participants in retail markets and be involved in frequent bidding processes. What end-users actually need is a well-escrowed service for sharing idle DERs and a bidding-free reward mechanism to receive payment. To this end, with the core idea of “access over ownership”, there is growing concern about the concept of sharing economy in the energy sector in recent years. Sharing economy refers to a market model that enables individuals or entities to share their idle resources with others upon payment for the purpose of efficient resource allocation and social welfare maximization [8]. To date, sharing economy-based business models have achieved substantial success in the housing (i.e., Airbnb) and transportation (i.e., Uber) fields by matching individuals to enjoy underutilized products [9]. Essentially, the physical interconnected networks of energy systems provide a natural platform for sharing economy application. In bulk power systems, for example, PJM has initiated coordinated transaction scheduling (CTS) via interregional tie-lines with NYISO and MISO since 2014 and 2017, respectively, to improve generation utilization and enhance price predictability [10]. In addition, transactive energyrelated pilot projects have been launched by Pacific Northwest National Laboratory (PNNL), enabling DER owners to share demand side resources with their neighbors through distribution grids while smoothing the fluctuations of electric load [11]. The concept and business model of sharing economy will bring new challenges and opportunities for deregulated energy markets. Therefore, in this chapter, we call upon an overview of the potential market design for energy sharing, and conduct a systematic review of energy sharing-related research and practice, which will provide a useful reference and insight for the development of the sharing economy in the energy sector.

1.2 Bibliometric Analysis To provide an overview of the existing research on the sharing economy in energy markets, a bibliometric analysis was conducted on January 1, 2021 using Web of Science (WoS) database. Keywords used for WoS were as follows: TS = ((sharing

1.2 Bibliometric Analysis

3

economy OR collaborative consumption) AND (energy market OR electricity market OR mechanism design)). The number of publications retrieved by WoS from 2003 to 2021 are shown in Fig. 1.1. In summary, 964 publications were identified. The number of publications before 2008 was relatively small, while it has been rapidly increasing since 2013. This proliferation of sharing economy was driven by the Economic Crisis and Great Recession in 2007–2008. It takes several years to bring sharing economy-related research and practice to publication. The number of relevant publications in popular journals since 2010 are listed in Fig. 1.2. 350 P2P

Sharing economy

Incentive compatibility

Collaborative consumption

Number of pubilcations

300 250 200 150 100 50 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 Year

Fig. 1.1 Number of publications retrieved by WoS Number of pubilcations 0

10

20

30

IEEE Trans. Power Syst. Energy Policy IEEE Trans. Smart Grid J. Clean. Prod. IEEE Trans. Industr. Inform. IEEE Trans. Sustain. Energy Renew. Sust. Energ. Rev. Energy Nat. Commun. Appl. Energy Energy Econ. Nature Nat. Energy Last 5 years

2011-2015

Before 2010

Fig. 1.2 Number of relevant publications in popular journals

40

50

60

4

1 Introduction

In the energy sector, the concept of sharing economy is generally proposed for the purpose of incentive-compatible and individual-rational market participation using mechanism design theory. As explained in the Introduction, the traditional marginal pricing (MP)-based mechanism may not be able to elicit truthful bidding in wholesale markets. In practice, a thermal generator may not fully share its availability by withholding its capacity or making strategic bids. Thus, the sharing economy should contribute to a fair and reasonable pricing mechanism for truthful bidding. On the other hand, the sharing economy should achieve an efficient aggregation of DERs in retail markets and identify the unique value of a participant for rational profit sharing. We briefly review the existing studies related to the sharing economy in energy markets from the perspective of market structure. For wholesale markets, ref. [12] proposes an incentive mechanism that elicits truthful information on strategic wind power producers supplying stochastic resources for wholesale markets. Ref. [13] applies the Vickrey-Clarke-Groves (VCG) mechanism to electricity spot markets and conducts a comparative study with a marginal pricing mechanism. In [14], the VCG mechanism is adopted in a two-stage electricity market to prevent strategic behaviors of market participants with a high penetration of variable renewables. For retail markets, ref. [15] proposes a nucleolusbased cost allocation method for incentivizing multi-microgrids within a grand coalition. Ref. [16] applies a bargaining game in an agent-based hierarchical framework on the retail side and implements a distributed optimization program for privacy protection. Ref. [17] designs a profit-sharing mechanism based on cooperative game theory, and the cooperative surplus is allocated according to each participant’s externality. Ref. [18] proposes a cooperative energy sharing market using generalized Nash bargaining (NB), and develops a linearization solution algorithm. In recent years, some review articles about sharing economy in terms of P2P transactive energy and demand side management have been published. Ref. [19] reviews energy sharing on the demand side and analyzes its potential for balancing services provision. Ref. [20] conducts a review of the principle of the sharing economy in electricity markets and assesses the development of sharing economy based on economic, social and environmental perspectives. Ref. [21] contributes an overview of the emerging P2P markets that consists of motivation, challenge and mechanism design and proceeds to potential application. Most existing review articles provide an interpretation of energy sharing as collaborative consumption in retail markets and focus on transactive energy among P2P sharing. However, from the perspectives of game and mechanism design theory, there exists no systematic overview or taxonomy for the sharing economy in energy markets, including wholesale, retail, integrated energy, and a high share of renewables.

1.3 Concept of Energy Sharing

5

1.3 Concept of Energy Sharing 1.3.1 History and Development Sharing economy is a hybrid market that refers to the sharing of the right to use goods and services. The umbrella concept of sharing economy can be explained in contained different labels, such as collaborative economy, P2P economy and other interpretations. The concept of sharing economy is not new and can be dated back to “collaborative consumption”, which was first proposed in 1978 [22]. The original idea was that by sharing idle resources, people would be able to improve the utilization of goods and services, thus achieving Pareto improvements based on existing resources. Over the past decades, the leapfrog development of the Internet and information and communication technologies (ICTs) has led to a dramatic improvement in computing power and the diversity of display modes, as well as an increasing number of business models related to sharing economy. In 2002, Yochai Benkler from Harvard University proposed the concept of “commons-based peer production”, and then extended the idea to “shareable goods” in 2004 [23]. With the advent of the Great Recession during 2007–2009, there was a growing sense of urgency about global population growth and resource depletion, leading to people’s awareness of the importance and necessity of sharing economy. Conventionally, customers may possess too many belongings that are not frequently used, thus yielding a huge waste of resources. In US, for example, more than $1.2 trillion was spent on nonessential goods each year [24]. There have also been attempts to mitigate the “Tragedy of the Commons”, the idea that when we need to make more efficient use of idle resources when maintain our quality of life [25]. In 2010, the first book of sharing economy was published, which systematically presented the definition, business model and significance of sharing economy. In recent years, the maturation of various enabling technologies has provided a possible commercialization of sharing economy. Many businesses have been influenced by this phenomenon, including hospitality, transportation as well as insurance industry. The leading companies that are driven by sharing economy are no longer insurgents and newcomers. Uber, Airbnb and a handful of others have gained the capability and scale to compete with, or even surpass, some of the world’s largest players in transportation, hospitality and other industries [26]. Sharing economy has expanded the choice of transaction subjects and the space for welfare improvement so that people can stay at home and employ all kinds of resources for individual use. These business models have endowed sharing economy with new significance.

6

1 Introduction

1.3.2 Characteristics Based on the core idea of “access over ownership”, Energy Sharing can be interpreted as the sharing economy in the energy sector, namely designing incentivecompatible market mechanisms for Pareto improvement by facilitating the utilization of idle energy resources via advanced information and communication technologies. According to such an interpretation, the characteristics of energy sharing are summarized as follows: • Utilization: Energy sharing can contribute to a more accurate match between energy supply and demand, thereby making more efficient use of idle resources. Such an accurate energy balance benefits from Internet technologies. For example, Airbnb developed an Internet-based platform for guests with short-term activities, which is able to improve the utilization of idle housing. Similarly, an Internetbased platform is required to support the energy sharing among customers on distribution power networks. In addition to individual use, DER owners can share surplus availability with neighbors, for example, rooftop solar energy transactions. • Efficiency: Energy sharing helps achieve Pareto improvement of an energy system or market, which has to be supported by optimization-based strategies. Generally, an increase in the utilization of idle resources represents a higher efficiency of energy system operation. For example, a virtual power plant (VPP) enables the aggregation of shared DERs to achieve peak load shaving and off-peak wind accommodation. However, a poorly-designed energy sharing strategy may even depress the overall social welfare, e.g., storage sharing. Sharing in-home battery storage with neighbors can accelerate the degradation of the battery bank, and thus extra expenses have to be incorporated into decision-making process. • Mechanism: Energy sharing requires a fair and reasonable settlement rule that defines the payments for the shared resources, which guarantees sufficient incentives for customer involvement. In the energy sector, for example, a well-designed pricing mechanism is the key to eliciting marginal generators’ truthful bids. On the other hand, a profit sharing or cost allocation mechanism is needed for efficient and stable aggregation of DERs in retail markets to reward good behavior and penalize bad one.

1.3.3 Taxonomy We propose a taxonomy for energy sharing-related research in terms of market structure, supply chain and energy attributes, as shown in Fig. 1.3. From the perspective of market structure, the sharing economy in energy markets can be divided into wholesale and retail markets. Wholesale markets involve the sale of energy among utilities and energy traders before it is eventually sold to consumers, while the retail markets involve the sale of energy to end-use consumers. Generally,

1.3 Concept of Energy Sharing

7

Fig. 1.3 Taxonomy of energy sharing

we focus on the concept of the sharing economy for mechanism designs in wholesale and retail markets. From the perspective of supply chain, energy sharing is involved in energy production, transmission, storage and consumption. Regarding energy production, energy sharing enables the coordination of various forms of energy. For example, with the gradual maturation of energy conversion technologies such as heat pumps and wasteto-biomass, sharing the cogeneration capability can greatly improve the efficiency of combined power and heating systems. For the transmission sector, energy sharing can realize the coupled transportation of different energy carriers in a single transmission facility, thus reducing the redundancy in materials and corridor coverage. One typical example is superconductor cable, which enables the transmission of electric power and liquid hydrogen. Additionally, the technology and business model of storage sharing have been widely investigated around the world, with a focus on the bidirectional sharing of energy storage. The capacity of a large-scale centralized storage can be shared among a set of customers for individual use, and a number of distributed storages can be aggregated as a single entity as well. In the consumption sector, energy sharing can satisfy the heterogeneous preferences of individual users to the greatest extent. For example, a blockchain-based platform can help end-use customers to bid on rooftop solar power and obtain maximal rewards. From the perspective of energy attributes, different forms of energy carriers can be shared and traded in interconnected energy markets. For example, power-to-gas (P2G) and fuel cell technologies have been advocated as an appealing solution to provide additional flexibility and facilitate energy sharing in joint gas/hydrogenelectricity markets.

8

1 Introduction

1.4 Sharing Economy in Wholsesale Markets 1.4.1 Electricity Spot Markets The core of an electricity spot market is the pricing mechanism, which is generally based on marginal pricing. The marginal pricing mechanism has been widely used in many electricity markets around the world. In PJM, cost bidding and locational marginal pricing have been applied in real-time markets since 1998 and subsequently in day-ahead and regulation markets since 2000 [27]. Nord Pool is the first transnational electricity market around the world. It receives bids and offers from producers and consumers, and calculates market clearing prices to balance supply and demand curves based on marginal price settlement. The marginal pricing mechanism meets the requirement of maximizing social welfare in perfectly competitive markets, where market participants cannot manipulate prices and the market prices are determined only by supply and demand. However, there exists potential for market participants to exercise market power under this pricing mechanism, as many actual cases show that market-oriented generators could manipulate market prices by making strategic bids. For example, one of the reasons that electricity prices of California soared between 2011 and 2017 is that the San Onofre Nuclear Generating Station (SONGS) closed. In this case, some power plants made strategic bids to earn more profits. There is the possibility of market participants exercising market power under the marginal price mechanism. Therefore, efficient market mechanisms should be carefully designed to mitigate the market power of generators. Mechanism design theory, also known as reverse game theory, studies the approaches of economic incentives or cost allocation toward designed objectives, where market participants act rationally through strategic behavior. A lot of research has focused on how to make market participants submit truthful information, and the Vickey-Clarke-Groves (VCG) theory is widely adopted. The payment for a unit based on VCG mechanism is equal to the substitution benefit of the unit for other units, i.e., the change in total cost of the market before and after the unit participates in the market. The VCG mechanism could accurately identify the value created by market participants and motivate market participants to submit truthful information to the market operator. Therefore, under the VCG mechanism, it’s the best choice for a generator to make truthful bids, no matter whether other generators make truthful bids or not. The social welfare is shown to be maximized at the dominant strategy equilibrium where every market participant submits truthful information. Ref. [28] applies the VCG mechanism to supply and demand bidding and the VCG mechanism is proved to elicit bidders to bid truthfully, then the feasibility of applying VCG mechanism to power and gas pipeline capacity auctions is evaluated. Ref. [29] applies the VCG mechanism to wholesale electricity markets, but the network constraints and renewable generation are not considered. Ref. [30] improves the standard VCG mechanism and applies it to wholesale electricity markets and the result shows that an efficient Nash

1.4 Sharing Economy in Wholsesale Markets

9

equilibrium exists when every market participant submits truthful information. Ref. [31] proposes a VCG-based profit distribution mechanism for wind power aggregators to elicit private information truthfully. VCG mechanism design theory has been widely applied to the design of incentive compatibility mechanisms for general commodities, renewable energy, energy storage and demand response. Theoretically, this mechanism perfectly satisfies incentive compatibility. However, the VCG mechanism has not been implemented in practical applications due to some defects, such as complicated computation and sacrificing budget balance.

1.4.2 Multi-area Electricity Markets The essence of multi-area electricity markets is to determine the optimal sharing strategy among multiple connected power grids for the concerns of price predictability, renewable accommodation, etc. In recent years, the PJM market has conducted coordinated transaction scheduling (CTS) with electricity markets in other regions of the United States, to improve the optimal allocation of resources and reduce the fluctuation of market prices. Additionally, several European electricity markets have focused on the coordination of multi-area market in recent years in the context of the gradual increase in the penetration rate of renewable energy [32]. Compared with isolated operation, multi-area electricity market realizes the coordination of multiarea power systems, and clean electricity can be shared by each regional power grid, which improves the overall operation economy. Many scholars regard the lack of algorithmic support as the main barrier to coordinating multi-area electricity markets. Some studies focus on the pricing mechanism or the clearing algorithm. In [33], a joint energy and reserve pricing mechanism is proposed to enable the balance of supply and demand in a multi-area market. In [34], the reliability criteria in a multi-area market are established and evaluated using probabilistic metrics. The proposed criteria are incorporated into the multi-area market clearing formulation. In [35], a market-based cross-border trading mechanism for multi-regional energy markets is designed. Some decentralized coordination strategies have also been proposed to protect information privacy in multi-area energy markets, e.g., optimality condition decomposition [36], alternating direction multiplier method [37], and augmented Lagrangian relaxation [38]. However, in these studies, the marginal pricing mechanism, which may not theoretically meet the requirement of incentive compatibility, is commonly used. The problem of exercising market power is even more serious in multi-area markets because there is information asymmetry between the multi-area markets. As a matter of fact, generators in the power-receiving area have to provide reserves for the inter-area power to ensure the safe and stable operation of the power system. More inter-area power usually requires more reserves in the power-receiving area. Therefore, when the generators submit higher reserve costs, market operators aiming to minimize system costs will decline the amount of inter-area power to reduce the costs of electricity market. In this case, the generators in the power-receiving area

10

1 Introduction

could get more revenue by forcing inter-area power out of the electricity market, but the efficiency of multi-area energy system will decrease and resources cannot be optimally allocated. A real-world case shows that in China, the generators in the power-receiving area prevent renewable energy from the northwest by submitting high reserve costs, leading to the curtailment of wind and solar. Therefore, it is necessary to design a market mechanism which is incentivecompatible to elicit market participants to make truthful bids in multi-area market. Ref. [7] first applies VCG auctions to joint market clearing in multi-area power systems. The thermal units are elicited to make truthful bids and provide reserves to help accommodate renewable energy. However, how to improve the efficiency of multi-area market coordination and VCG auctions in real-world cases remains a public interest.

1.4.3 Integrated Energy Markets To date, the market scheduling of electricity, thermal energy and natural gas generally takes other energy systems as static boundary conditions, which leads to inadequate sharing among different energy carriers. However, with the increasing coupling of different energy resources, the interaction between multi-energy markets can no longer be ignored. Therefore, existing studies and pilot projects have proposed constructing an integrated energy market, taking charge of coordinated sharing among different energy market entities. Recent decades have witnessed a rapid development of integrated energy markets in different regions around the world. For example, the U.S. government has initiated integrated energy trading projects with a total investment of 650 million dollars since 2007, which further guarantees national energy supply adequacy and security [39]. Due to a high proportion of gasfired generation (~34%), Great Britain has been focusing on the construction of joint electricity and natural gas markets, especially technical solutions regarding the challenges brought by the increasing penetration of wind power [40]. Japan has issued a series of policies to establish integrated energy markets for energy sharing. In April 2010, the Japan Smart Community Alliance (JSCA) was founded to balance demand side energy supply and demand [41]. A 100% hydrogen-powered city, Harumi Flag, will be built by 2024 [42]. A wide variety of the existing literature has quantified the cost and benefit of sharing energy resources among integrated energy markets. In [43], a coordinated operation and long-term planning strategy of electricity and natural gas systems is developed based on real-world cases in Spain. In [44], a hybrid gas-electricity model is proposed, in which the potential coupling effects between gas and electric power systems are evaluated. In [45], a day-ahead market clearing framework is designed to investigate the optimal operation strategy of gas-fired power plants in electricity-gas combined markets. In [46], a joint market framework of integrating power grids and heating systems is proposed to evaluate the cobenefits of sharing solar-powered heat pumps.

1.4 Sharing Economy in Wholsesale Markets

11

Additionally, some studies have investigated the strategic behavior of multi-energy market participants. In [47], a computational game theoretic investment model is proposed considering the strategic market behaviors of natural gas participants and its influence on the electricity market and carbon emission market. In [48], the strategic behavior of market participants in a multi-energy market is analyzed, in which a multienergy participant is allowed to aggregate the local energy system and maximize the expected profits in the whole electricity market. In [49], a heat and electricity coupled system is introduced, and the concept of integrated demand response is proposed in a heat and electricity combined market to investigate the demand flexibility of smart buildings. In most of the existing literature, the marginal pricing mechanism is adopted for integrated energy market settlement. While this may not be the case in realistic heating or natural gas markets, ref. [50] has explored the marginal pricing design of gas and thermal energy. However, information asymmetry between different energy systems can lead to a much more severe impact on market efficiency than that in a single-energy carrier market. For example, to prevent the strategic bidding of gasfired power plants, many regions have enacted market regulation and supervision policies, e.g., the Electric Reliability Council of Texas (ERCOT) sets a price cap of $ 2000/MWh when contingencies occur [51]. Therefore, the mechanism design theory of energy sharing may be another promising choice to elicit truthfulness of the participants in integrated energy markets, which deserves in-depth study in the future.

1.5 Sharing Economy in Retail Markets 1.5.1 Agent-Based Energy Sharing Agent-based energy sharing refers to the case where various DERs are coordinated and organized by an external operator. For agent-based sharing, developing suitable coordinated strategies and mechanisms for different types of DERs is an effective way to incentivize DER owners to share their idle resources. The existing literature can be divided into two categories: (i) intrusive strategies that allow the operator to access individual DERs, e.g., direct load control, and (ii) non-intrusive strategies in which the self-dispatch of DER owners is influenced by the incentive signals sent by the operator, e.g., price-based demand response. For intrusive strategies, the operator collects the information and sends control signals to DERs. Many realistic cases of intrusive strategies have been developed. For example, the U.S. military has implemented demand-side management projects in many regions and adopted intrusive strategies to dispatch DERs [52]. Demandside management has been implemented to realize the direct control of DERs in Ningxia, China [53]. Many scholars have conducted research on intrusive strategies. The direct load control algorithm of electric water heaters applied to wind power

12

1 Introduction

accommodation is studied in [54]. Based on the state sequence control algorithm, ref. [55] studies the direct load control algorithm of electric heat pumps based on low-pass filtering to suppress the power fluctuation of the connection between a microgrid and distribution grids. However, intrusive strategies bring some problems, such as privacy concern, heavy computational burdens and large data exchange. Considering the basic nature of a sharing economy structure, non-intrusive strategies, such as edge computing techniques, have been developed. Distributed algorithms such as Lagrangian relaxation and ADMM algorithms are promising candidates. The distributed edge computing framework is designed for energy management that can be applied to renewable energy to improve the control response speed of DERs. Ref. [56] applies decomposition techniques for large-scale distributed prosumers in demand-side equipped with IoT devices. Agent-based energy sharing is dominated by an external operator or an energy sharing platform. The key issue is to investigate optimal pricing methods and profit sharing between the operator and DER owners. Such mechanisms are uniformly designed and organized by the operator while trying to fulfill Pareto optimality, budget balance, incentive compatibility and other axioms. However, there remains an open question regarding how to design a fair and reasonable profit-sharing mechanism, and some studies have focused on the measurement of “fairness” [57].

1.5.2 Peer-To-Peer Energy Sharing In contrast to the agent-based model, which is a kind of business-to-customer (B2C) service, P2P energy sharing requires customers to make self-decisions, and is thus defined as a customer-to- customer (C2C) service. P2P energy sharing refers to energy transactions in a P2P trading platform or transactive market among diversified DER owners, including residential and enterprise prosumers. The earliest commercialized energy trading platform is Vandebron, which was launched in the Netherlands in 2014 [58]. Users of the platform first choose different trading contract periods according to personal preference; then, the platform recommends appropriate power suppliers. By this means, the source of power supply can be tracked through P2P trading. The platform has already provided clean electricity for more than 100,000 households, but it is still limited to electricity trading and has not yet covered the costs of ancillary services [59]. Another example is the P2P trading platform in Brooklyn, which enables DER owners to provide clean energy to households on low incomes [60]. P2P energy sharing reduces the threshold of energy transactions and allows smallscale individual DER owners to directly participate while retaining control over their DERs. As a form of energy sharing with a high degree of freedom, P2P energy transactions are not only related to the interests of all parties in energy sharing but also related to wider social interests. As a result, more complicated rules need to be followed than in the existing sharing economy. Unlike the market rules in traditional energy markets, the rules for emerging P2P sharing lack uniform standards and show a diversification trend. The application of game theory and auction theory in P2P

1.5 Sharing Economy in Retail Markets

13

energy trading is examined to improve energy efficiency in [61]. Heterogeneous risk aversion of different parties in community-based energy sharing is considered, and a novel definition of fairness is introduced in profit allocation in [62]. As a core component of the market mechanism, the limitations of LMP-based pricing mechanisms in the wholesale market also exist in retail markets. Several studies focus on the mechanism design for facilitating DER aggregation, including ex-post profit sharing based on cooperative game theory and bilateral contracts. The cooperative surplus can be allocated among the DERs based on the Shapley value and nucleolus methods [63]. In [64], a new stability concept is introduced, leading to a trail-stable outcome whenever the preferences of agents are able to satisfy full substitutability. In practice, there may be a large number of market participants in P2P energy sharing, which may cause the problem of computational complexity. Some studies design a Nash bargaining-based profit-sharing mechanism to overcome this problem. In [65], an energy sharing scheme is established among DER owners, and the benefits brought by sharing DERs are allocated based on the contribution rate of each participant. Some technical challenges remain to be addressed in terms of P2P energy sharing. For example, more efficient consensus-based algorithms need to be investigated to coordinate P2P energy sharing for fast convergence with acceptable negotiations. Additionally, some recent studies have focused on developing cryptocurrency, digital currency and other derivatives for P2P sharing settlement.

1.5.3 Integration of Distributed Energy Resources into Wholesale Markets With the increasing penetration of DERs, how to efficiently manage large-scale DERs in wholesale markets has become a public interest. In essence, the aggregation of large-scale DERs characterizes the feasible region formed by the operating constraints of different DERs. A distribution system operator can aggregate largescale DERs to behave as a controllable flexible power plant. Generally, the feasible region of a traditional thermal power plant is described by “static parameters”, e.g., installed capacity, minimum power output and ramping rate. However, the feasible region of a number of DERs is not only restricted by the static parameters of distribution networks but also shaped by the DERs and shiftable/curtailable loads with dynamic spatiotemporal dependency. For example, the “maximum available capacity” of a distribution grid can be changed under the impact of distributed photovoltaic power, power flow, and node voltage along different time slots. The “ramping rate” depends on the operating status and dynamic performance of resources, such as demand response. In [66], a geometric approach is proposed to explore the flexibility potential of demand response, which facilitates the integration of demand-side resources into system-level operation. In [67], the Fourier-Motzkin elimination method is adopted

14

1 Introduction

to represent the dispatchable region of power systems. Ref. [68] proposes a method for approximately calculating the equivalent active-reactive power feasibility region of the energy networks. This method approximates the equivalent feasibility region of distribution networks by heuristically selecting active-reactive four-quadrant operating points. In addition, some papers explore the application of demand-side resources or renewable energy equivalent feasible regions in the optimized operation of transmission grids. Ref. [69] applies equivalence theory to the field of unit commitment and proposes a safety-constrained unit commitment model, which takes the uncertainty of variable renewables into account. The aggregation of DERs makes it possible for distributed individuals to provide upstream grid services for wholesale markets. The aggregation of various DERs can also serve as non-wire alternatives (NWA) to defer investment of energy networks as well as capacity expansion. In [70], the reliability value is embedded in the planning framework to determine the capacity of rooftop photovoltaic and storage amidst rare weather events when distribution network contingencies occur. Ref. [71] notes that the existing DER pilot projects may help to defer generation and transmission expansion, thereby reducing the systemwide costs by 20–50%. Ref. [72] allows DERs to act as NWAs in a joint planning framework considering DER investment and power system expansion. Ref. [73] evaluates the role of DERs as NWAs against wire investment in traditional distribution network planning.

1.6 Enabling Technology and Business Models 1.6.1 Energy-Related Technology 1.6.1.1

Energy Conversion Technology

In human history, from the replacement of firewood by coal in the sixteenth century to the replacement of coal by oil in the twentieth century, every revolution in energy technology has promoted the course of human civilization. In the future, with the continuous development of renewable energy, energy conversion technology will play a structural role in energy sharing. Efficient conversion of energy can reduce the mining of fossil energy while reducing environmental pollution, yielding great significance for energy security and the development of human civilization. Recent decades have witnessed a wide variety of emerging energy conversion technologies, including concentrating solar power (CSP), fuel cells and biomass gasification. One of the most important is power-to-gas or -hydrogen (P2G/P2H). Hydrogen energy is a resource-rich, low-carbon and widely-used secondary energy source, and is becoming a critical energy carrier for future clean energy transition. In this instance, P2G enables the transformation from surplus renewable energy to green hydrogen, which is extremely beneficial for the development of renewabledominated power systems and the decarbonization of the industrial, transportation

1.6 Enabling Technology and Business Models

15

and energy sectors. In 2017, the installed capacity of P2G in the European demonstration project was approximately 30 MW. It is estimated that by 2050, 10–65% of the energy consumption of the EU’s industrial field will come from P2G, and 30–65% of the energy in the heating industry and transportation will come from P2G. There are many technical routes for P2G, which are mainly divided into alkaline electrolysis, proton exchange membranes (PEMs) and solid oxide electrolyzer cells (SOECs). PEM water electrolysis hydrogen production technology has the advantage of flexibility and is able to match the volatility of renewable energy power generation. At present, PEM water electrolysis hydrogen production has entered the 10 MW-level demonstration stage. Additionally, 100 MW PEM electrolyzers are under development, and NEL-Proton, SIEMENS, and ITM Power are in a leading position in the relevant technology and equipment manufacturing. The 718 Research Institute of China Shipbuilding Industry Corporation has also carried out many studies on PEM water electrolysis technology [74]. The 10 MW P2G project under construction in Guyuan, Hebei, is the largest P2G conversion demonstration project in China, and the hydrogen can be used in industries and refueling stations [75].

1.6.1.2

Energy Transmission Technology

The hybrid energy transfer line (HETL) enables long-distance transportation of electricity and cryogenic fuel, e.g., liquid hydrogen and liquefied methane, in a single transmission device. The basic structure of an HETL is similar to that of an ordinary superconducting cable, and the major difference lies in that the cooling medium of an ordinary superconducting cable is supercooled liquid nitrogen, while an HETL uses cryogenic fuels [76]. This means of sharing the same transmission device with different energy carriers can greatly improve energy efficiency and is especially economical for long-distance transmission. Meanwhile, this technology is also a critical support for future energy sharing applications in the transmission sector. As early as the beginning of this century, scholars proposed the idea of a similar hybrid energy transfer line [77]. With the maturity of the material technologies, pilot projects are continuously emerging. In July 2019, the Chinese Academy of Sciences successfully developed the principle prototype of a “superconducting direct current power/gas transmission integrated energy transfer line”. However, this technology is still in the laboratory stage and has not yet been widely popularized in practice.

1.6.1.3

Energy Storage Technology

Energy storage technology is an essential means for the transformation of human energy structure from fossil energy to renewable energy, which can smooth the volatility of renewables. In recent years, transportation electrification has created potential for portable energy storage sharing. The Swedish government has declared that all vehicles must use non-fossil fuels by 2030 [78]. In September 2019, China

16

1 Introduction

began to set up strong transportation networks with wide coverage and high speed, which requires the optimization of the transportation energy structure and the deep utilization of roadside renewables. At present, one of the lowest hanging fruit is vehicle-to-grid (V2G), which allows bidirectional power exchange between electric vehicles (EVs) and power grids. The essence of V2G is to share idle battery resources for various grid supports. According to forecasts, by the end of 2030, the number of EVs in China will reach over 100 million, and the aggregated capacity will exceed 1000 GW, which is equivalent to China’s installed thermal power capacity [79]. Through effective aggregation technology, the aggregated electric vehicle can be treated as a single controllable storage device and respond to dispatch signals from the upstream power grid. Some of the existing literature has evaluated the benefits of V2G applications and designed related business models. In [80], the integration value of EVs in Midcontinent Independent System Operator (MISO) grid is evaluated based on a multiday optimization model. The results show that with the support of V2G technology, orderly bidirectional charging of EVs can provide flexibility for peak shaving and ramping. In [81], the feasible region of EVs aggregation is formulated and applied in microgrid bidding toward connected bulk power systems. In recent years, an emerging concept and technology termed fuel-cell hybrid electric vehicles (FCHEVs) has prompted the integration of transportation and energy systems. With the complementarity of hydrogen and power systems, it is possible to electrolyze water to produce hydrogen during peak hours and store it for further use. In addition, energy storage has become a generic supporting technology in various industries, e.g., communication, data centers, architecture, robotics, manufacturing and national defense security, among which the concept of sharing economy will bring about novel business models. Uber announced a new sharing model, “UberAir”, in Dubai and the Dallas-Fort Worth area that will launch in 2023 [82]. The maturity of all-electric helicopter technology can greatly reduce people’s commuting and travel time while relieving traffic pressure and can serve as large-scale portable storage as well. On the other hand, Huawei has provided Pakistani operators with communication facilities that share lithium batteries as backup power to solve the problem of communication interruption caused by unstable power supply. Furthermore, some researchers have proposed business models of cloud energy storage (CES) [83]. In [84], a CES operator is supposed to invest in centralized storage and share virtual storage for individual customers when needed. By contrast, in [85], individual customers decide to invest in distributed solar and storage, which can be shared and utilized by a CES operator to hedge against wholesale market risks and achieve peak shaving.

1.6.1.4

Energy Consumption Technology

In recent years, the wide use of intelligent instruments and sensors and the application of the Internet of Things (IoTs) have created conditions for deeper energy sharing on the demand side and provided a platform and an effective way to decentralize energy

1.6 Enabling Technology and Business Models

17

transactions, e.g., P2P energy trading. One of the emerging energy consumption entities comes from digitalization, which can be interpreted as converting data into realistic value. About 700 million smart meters have been installed worldwide, including 400 million in China alone. By the end of 2025, it is estimated that 75 billion devices will be used to connect cyber and physical systems across the world, which will provide abundant information for demand-side energy sharing [86]. For example, in Belgium, DER owners are allowed to provide flexibility services to a distribution system operator via a centrally-shared data platform [87]. In Germany, the DERs of more than 30,000 households can be effectively aggregated by sonnenCommunity, which has formed one of the largest virtual storage systems in the world [88]. Additionally, as the core of information exchange, processing and communication, data centers are gaining substantial popularity among Internet companies such as Google and Apple. As stated in the 2018 Data Center World Global Conference, by 2021, the average number of data centers per organization will reach 10.2, which is estimated to account for 8% of global power consumption. In this instance, various existing studies have focused on the flexibility of shared data centers. Lawrence Berkeley National Laboratory has published a report that a data center can reduce energy consumption by 10–25% for emergency demand response within their abilities [89]. Its reasonability has been proven by what occurred on July 22, 2011; many data centers were managed to prevent an emergency blackout by shutting off their servers in the U.S. [90].

1.6.2 Information-Related Technology 1.6.2.1

Power and Communication Tower-Sharing

With the deployment of ultradense networks in the 5G era, network functions are largely dispersed to base stations (BSs). It is estimated that the global power consumption of communication BSs accounts for approximately 70–80% of that of the entire communication system [91]. In 2025, the deployment density of BS will grow to more than ten times the current level, thus consuming more electric power and occupying larger floor space [92]. Power and communication tower-sharing refers to mounting communication base stations on power line towers, which can not only solve the power supply problem of BSs but also significantly reduce corridor occupation and enhance signal coverage. Such technology enables full use of tower resources while saving the construction costs of 5G bases, including material, labor and land costs. As early as 2017, China Southern Power Grid Company (CSPGC) started to cooperate with Yunnan Tower Company to build China’s first 4G communication BS with an antenna attached to a 110 kV power tower [93]. Since 2018, China Tower has successively signed strategic cooperation agreements with the CSPGC and State Grid Company of China (SGCC), making it clear that both sides will promote the installation of BSs on power towers

18

1 Introduction

[94]. In 2020, the SGCC has completed the sharing of 37 towers in Changzhou, Jiangsu [95]. In addition, as a non-negligible load, the energy consumption of BSs may have a considerable impact on the power flow of distribution networks. Some existing literature has investigated such an impact and proposed an energy sharing strategy for potential demand response. To share renewable generators and energy storage, ref. [96] proposes a two-stage radio resource allocation scheme to maximize the total rate of cellular users while ensuring the minimum throughput of each sensor and control unit. In order to assign a reliable path to the key information flow, ref. [97] proposes an information flow importance evaluation method based on the physical sensitivity of information. Ref. [98] proposes an energy sharing strategy to jointly dispatch DERs and BSs while incorporating linearized communication reliability constraints.

1.6.2.2

Network Slicing Technology

With the advent of the 5G era, IoTs allow massive and diversified devices to be connected on communication networks, and such networks will be oriented to different application scenarios. Massive control signal exchange and processing need to be completed in a relatively short time, which enforces a strict requirement for low communication delay. To improve the utilization of communication networks while reducing time delay, a physical network can be dynamically divided into multiple virtual logical networks without the need to construct specific channels for each type of application scenario. This technology is termed network slicing [99]. Network function virtualization (NFV) is a necessary condition for network slicing. Essentially, NFV aims to transfer the hardware and software functions of special devices in the network to a virtual machine (VM)., e.g., mobility management entities (MMEs) and policy and charging rules functions (PCRFs) in the core network and digital units (DUs) in the wireless network. These VMs are based on industry standard commercial servers, thus replacing dedicated network element devices with servers, storage and networking devices based on industry standards. After NFV, the VMs in the core network (i.e., core cloud) and those in the wireless access network (i.e., edge cloud) are interconnected through a software defined network (SDN). By this means, based on NFV and SDN, network slicing is easy to implement, and the network can be horizontally cut into multiple virtual subnetwork slices for different energy sharing application scenarios [100]. Network slicing technology enables virtual networks to provide personalized services of energy sharing and is considered one of the most important developments in the 5G era. Some existing literature focuses on the design of network slicing, but few studies have investigated its application in the energy sector. In [101], efficient network slicing is designed via a hybrid machine learning and deep learning algorithm. In [102], an independently-tailored network-slicing architecture is proposed to improve the interaction between different application scenarios.

1.6 Enabling Technology and Business Models

1.6.2.3

19

Blockchain Technology

Blockchain is a peer-to-peer network with no centralized hardware or management organization. Each node in the network has equal status and can simultaneously act as a client and a server. Each node in the blockchain system stores all the data information in the entire network. Therefore, the data of each node are jointly owned, managed and supervised by all entities. To ensure that all transactions and data in the P2P energy sharing process are credible, the blockchain uses a decentralized mechanism. Due to its transparency and decentralization, blockchain technology ensures that different entities can trust each other, thereby greatly reducing the cost of reshaping or maintaining truthfulness. Therefore, blockchain technology can be further applied in other areas. Ref. [103] designs a reputation recording system based on blockchain that can be used in multiple networks in response to security vulnerabilities in current reputation systems. In response to problems such as the leakage of personal privacy caused by the collection of large amounts of user information by third parties, a blockchain-based decentralized personal data management system is proposed in [104]. To avoid the sampling robot from being affected by improper human intervention, ref. [105] proposes a blockchain-based method to ensure the authenticity of sensor data. In addition to theoretical research, several blockchain-based energy sharing pilot applications have been initiated around the world. Blockchain is adopted in the Brooklyn Microgrid project to facilitate the transaction of excess rooftop photovoltaic power on local P2P trading platforms [106]. Two demonstration projects have been constructed in Europe, where blockchain is used to process large-scale decentralized energy trading [107]. In China, a series of blockchain-based energy trading platforms have been under development, accounting for approximately 80% of the sales in distributed energy markets and effectively reducing time, labor and communication costs [108].

1.7 Conclusions In this chapter, we conduct a systematic review of the sharing economy in energy markets, and analyze the characteristics of energy sharing toward a more sustainable energy system. Based on the proposed taxonomy for energy sharing in terms of market structure, supply chain and energy attributes, we focus on how to elicit truthfulness while satisfying incentive compatibility in wholesale markets and how to design profit-sharing mechanisms while reducing transaction costs in retail markets. A discussion about the energy- and information-related supporting technologies is conducted along with some real-world business models.

20

1 Introduction

Four challenges deserve in-depth investigation in future work. (1)

(2)

(3)

(4)

A comprehensive sphere of idle resources should be precisely modeled and incorporated in the energy sharing paradigm. While the existing literature contributes research on the sharing economy in terms of storage, electric vehicles and even multiple energy sources, a wide variety of novel forms of idle energy resources should be further evaluated. For example, some recent studies have proposed the concept of gravity storage, which enables bidirectional transformation between gravitational potential energy and electricity [109]. By this means, vehicles parked in lots can be lifted for charging and dropped for discharging, serving as a new form of energy carriers. Advanced optimization strategies should be formulated with detailed full lifecycle impacts of energy sharing embedded. Most of the existing literature has focused on the short-term value brought by energy sharing without quantifying the impacts of a sharing strategy in a full life-cycle horizon. However, the absence of such a consideration may lead to severe incompatibility between sharing benefits and depreciation costs. For example, no literature has investigated the economic and environmental impacts of ride sharing on the global economy and climate change, especially for electric vehicles. Efficient market mechanisms should be developed considering the bounded rationality of market participants. On the one hand, with the explosive growth of the categories of shared goods and services, market participants are generally confronted with complicated combinatorial auctions, which is well known as an NP-hard problem. Additionally, a VCG-based pricing mechanism can hardly be implemented in practice due to its computational complexity. On the other hand, the existing literature in the field of energy markets seldom considers bounded rationality. However, real-world customers are always gifted with the property of bounded rationality, which may result in a theoretical deviation. For example, customer stickiness refers to a customer’s emotional, psychological or physical engagement caused by repeated interactions [110]. Prospective research and profitable business models should be designed to explore the potential of sharing economy. For example, scientists have begun to launch satellites for space energy collection, which can be shared back to the earth via microwaves. Another example is the Agriport Data Center in Amsterdam developed by Microsoft, which enables sharing waste heat from data centers to greenhouse planting and biomass [111].

Hopefully, this book will provide a fundamental reference for the development of sharing economy-related technologies and business models in the energy sector.

References

21

References 1. China promises to achieve the time from carbon peak to carbon neutral, much shorter than the time used in developed countries. [Online]. Available: http://www.gov.cn/xinwen/202104/22/content_5601515.htm 2. Total installed capacity forecast for green energy generation in China in 2020 and a forecast up to 2050. [Online]. Available: https://www.statista.com/statistics/977295/china-total-ren ewable-power-generation-installed-capacity-forecast/ 3. Wang, J., Zhong, H., Yang, Z., et al.: Exploring the trade-offs between electric heating policy and carbon mitigation in China. Nature Commun. 11, 6054 (2020) 4. California adopts 2030 Preferred System Portfolio with 12 GW new wind, solar, storage, geothermal. [Online]. Available: https://www.utilitydive.com/news/california-adopts-2030preferred-system-portfolio-with-12-gw-new-wind-sola/553584 5. 13th Five Year Plan for Energy Development. [Online]. Available: http://www.nea.gov.cn/ 135989417_14846217874961n.pdf 6. DOE Global Energy Storage Database. [Online]. Available: http://www.energystorageexcha nge.org/projects 7. Wang, J., Zhong, H., Yang, Z., et al.: Incentive mechanism for clearing energy and reserve markets in multi-area power systems. IEEE Trans. Sustain. Energy 11(4), 2470–2482 (2019) 8. Schlagwein, D., Schoder, D., Spindeldreher, K.: Consolidated, systemic conceptualization, and definition of the “sharing economy.” J. Assoc. Inform. Sci. Technol. 71(7), 817–838 (2020) 9. Zervas, G., Proserpio, D., Byers, J.W.: The rise of the sharing economy: estimating the impact of Airbnb on the hotel industry. J. Market. Res. 54(5), 687–705 (2017) 10. PJM, MISO coordinated transaction scheduling. In: Proceedings Interchange Optimisation Workshop, Apr. 18, 2014. [Online]. Available: http://www.pjm.com/∼/media/committeesgroups/stakeholder-meetings/pjm-misojoint-common/20140418/20140418-miso-pjm-jcminterchangeoptimizationpresentation.ashx 11. Transactive Energy: Negotiating New Terrain. [Online]. Available: https://www.pnnl.gov/ news-media/transactive-energy-negotiating-new-terrain 12. Tang, W., Jain, R.: Market mechanisms for buying random wind. IEEE Trans. Sustain. Energy 6(4), 1615–1623 (2015) 13. Xu, Y., Low, S.H.: An efficient and incentive compatible mechanism for wholesale electricity markets. IEEE Trans. Smart Grid 8(1), 128–138 (2016) 14. Exizidis, L., Kazempour, J., Papakonstantinou, A., et al.: Incentive-compatibility in a twostage stochastic electricity market with high wind power penetration. IEEE Trans. Power Syst. 2846–2858 (2019) 15. Du, Y., Wang, Z., Liu, G., et al.: A cooperative game approach for coordinating multi-microgrid operation within distribution systems. Appl. Energy 222, 383–395 (2018) 16. Dehghanpour, K., Nehrir, H.: An agent-based hierarchical bargaining framework for power management of multiple cooperative microgrids. IEEE Trans. Smart Grid 10(1), 514–522 (2017) 17. Wang, J., Zhong, H., Qin, J., et al.: Incentive mechanism for sharing distributed energy resources. J. Modern Power Syst. Clean Energy 7(4), 837–850 (2019) 18. Zhong, W., Xie, S., Xie, K., et al.: Cooperative P2P energy trading in active distribution networks: an MILP-based Nash Bargaining solution. IEEE Trans. Smart Grid 12(2), 1264– 1276 (2021) 19. Chen, Y., Liu, F., Wei, W., et al.: Energy sharing at demand side: concept, mechanism and prospect, automation of electric power systems 45(2), 1–11 (2021) 20. Filipovi, S., Radovanovi, M., Lior, N.: What does the sharing economy mean for electric market transitions: a review with sustainability perspectives. Energy Res. Social Sci. 58, 101258 (2019) 21. Sousa, T., Soares, T., Pinson, P., et al.: Peer-to-peer and community-based markets: a comprehensive review. Renew. Sust. Energ. Rev. 104, 367–378 (2019)

22

1 Introduction

22. Felson, M., Spaeth, J.L.: Community structure and collaborative consumption: a routine activity approach. Am. Behav. Sci. 21(4), 614–624 (1978) 23. Benkler, Y., Nissenbaum, H.: Commons-based peer production and virtue. J. Polit. Phil. 14(4) (2006) 24. Number of the Week: Americans Buy More Stuff They Don’t Need. [Online]. Available: https://www.wsj.com/articles/BL-REB-13793 25. Lloyd, W.F.: Two lectures on the checks to population. JH Parker (1833) 26. Prieto, M., Baltas, G., Stan, V.: Car sharing adoption intention in urban areas: what are the key sociodemographic drivers? Transp. Res. Part A Policy Pract. 101, 218–227 (2017) 27. Overview of PJM: Looking Back to Look Forward. [Online]. Available: https://www.meti. go.jp/press/2018/05/20180529003/20180529003-6.pdf 28. Hobbs, B.F., Rothkopf, M.H., Hyde, L.C., et al.: Evaluation of a truthful revelation auction in the context of energy markets with nonconcave benefits. J. Regulat. Econ. 18, 5–32 (2000) 29. Schummer, J., Vohra, R.V.: Auctions for procuring options. Oper. Res. 51(1), 41–51 (2003) 30. Tang, W., Jain, R.: A Nash equilibrium need not exist in the locational marginal pricing mechanism (2013). arXiv:1310.4282 31. Tang, W., Jain, R.: Aggregating correlated wind power with full surplus extraction. IEEE Trans. Smart Grid 9(6), 6030–6038 (2017) 32. 2020 Report on the State of the Energy Union. [Online]. Available: https://ec.europa.eu/ene rgy/sites/ener/files/report_on_the_state_of_the_energy_union_com2020950.pdf 33. Vlachos, A.G., Biskas, P.N.: Simultaneous clearing of energy and reserves in multi-area markets under mixed pricing rules. IEEE Trans. Power Syst. 26(4), 2460–2471 (2011) 34. Ahmadi-Khatir, A., Bozorg, M., Cherkaoui, R.: Probabilistic spinning reserve provision model in multi-control zone power system. IEEE Trans. Power Syst. 28(3), 2819–2829 (2013) 35. Karimi, A., Seifi, H., Sheikh-El-Eslami, M.K.: Market-based mechanism for multi-area power exchange management in a multiple electricity market. IET Gen. Trans. Distrib. 9(13), 1662– 1671 (2015) 36. Bakirtzis, A.G., Biskas, P.N.: A decentralized solution to the DC-OPF of interconnected power systems. IEEE Trans. Power Syst. 18(3), 1007–1013 (2003) 37. Erseghe, T.: Distributed optimal power flow using ADMM. IEEE Trans. Power Syst. 29(5), 2370–2380 (2014) 38. Ahmadi-Khatir, A., Conejo, A.J., Cherkaoui, R.: Multi-area unit scheduling and reserve allocation under wind power uncertainty. IEEE Trans. Power Syst. 29(4), 1701–1710 (2014) 39. Wang, J., Xia, Q., Li, G., et al.: Mechanism design for integrated energy markets based on multi-market equilibrium. In: Proceedings of the CSEE, Early access (in Chinese) (2021) 40. The future of gas-fired generation. [Online]. Available: https://www.nationalgrid.com/uk/ gas-transmission/insight-and-innovation/gas-future-operability-planning-gfop/future-gasfired-generation 41. Japan Smart Community Alliance. [Online]. Available: https://www.smart-japan.org/english/ about/index.html 42. HARUMI FLAG, the City That Brings the Future to Today’s Tokyo. [Online]. Available: https://news.panasonic.com/global/stories/2019/71600.html 43. Gil, J., Caballero, A., Conejo, A.J.: Power cycling: CCGTs: the critical link between the electricity and natural gas markets. IEEE Power Energy Mag. 12(6), 40–48 (2014) 44. Dueñas, P., Leung, T., Gil, M., et al.: Gas–electricity coordination in competitive markets under renewable energy uncertainty. IEEE Trans. Power Syst. 30(1), 123–131 (2015) 45. Chen, R., Wang, J., Sun, H.: Clearing and pricing for coordinated gas and electricity day-ahead markets considering wind power uncertainty. IEEE Trans. Power Syst. 33(3), 2496–2508 (2018) 46. Wang, J., Zhong, H., Tan, C.W., et al.: Economic benefits of integrating solar-powered heat pumps in a CHP system. IEEE Trans. Sustain. Energy 9(4), 1702–1712 (2018) 47. Spiecker, S.: Modeling market power by natural gas producers and its impact on the power system. IEEE Trans. Power Syst. 28(4), 3737–3746 (2013)

References

23

48. Yazdani-Damavandi, M., Neyestani, N., Shafie-khah, M., et al.: Strategic behavior of multienergy players in electricity markets as aggregators of demand side resources using a bi-level approach. IEEE Trans. Power Syst. 33(1), 397–411 (2017) 49. Shao, C., Ding, Y., Siano, P., et al.: A framework for incorporating demand response of smart buildings into the integrated heat and electricity energy system. IEEE Trans. Industr. Electron. 66(2), 1465–1475 (2017) 50. Dueñas, P., Leung, T., Gil, M., et al.: Gas–electricity coordination in competitive markets under renewable energy uncertainty. IEEE Trans. Power Syst. 30(1), 123–131 (2014) 51. Texas cuts ERCOT cap from $9,000/MWh to $2,000; summer exemption mulled. [Online]. Available: https://www.spglobal.com/platts/en/market-insights/latest-news/electric-power/ 030321-texas-cuts-ercot-cap-from-9000mwh-to-2000-summer-exemption-mulled 52. The US military has developed a “microgrid” to ensure a steady supply of electricity. PLA Daily. [Online]. Available: https://mil.huanqiu.com/article/9CaKrnJA24P 53. Tian, X., Yu, L., Dong, X., et al.: Analysis and prospect of electricity demand in Ningxia under China’s new normal. Electr. Power 50, 27–32 (2017) 54. Kondoh, J.: Direct load control for wind power integration. IEEE Power Energy Society General Meeting. IEEE, San Diego, CA, pp. 1–8 (2011) 55. Wang, C., Liu, M., Lu, N.: A tie-line power smoothing method for microgrid using residential thermostatically-controlled loads. Proc. CSEE 32(25), 36–43 (2012) 56. Xin, L., Huang, Q., Wu, D.: Distributed large-scale co-simulation for iot-aided smart grid control. IEEE Access 5(99), 19951–19960 (2017) 57. Lin, W., Bitar, E.: Forward electricity markets with uncertain supply: cost sharing and efficiency loss. In: IEEE Conference on Decision and Control, pp. 1707–1713 (2014) 58. Park, C., Yong, T.: Comparative review and discussion on P2P electricity trading. Energy Procedia 128, 3–9 (2017) 59. Essential renewable energy from Dutch soil. [Online]. Available: https://vandebron.nl/bestebod?pc=duurzaam-voordeel&gclid=EAIaIQobChMIlZH5tZ6W8QIV2qqWCh2xqQkyEAA YASAAEgLhfPD_BwE&gclsrc=aw.ds 60. Morstyn, T., Farrell, N., Darby, S.J., et al.: Using peer-to-peer energy-trading platforms to incentivize prosumers to form federated power plants. Nat. Energy 3(2), 94–101 (2018) 61. Tushar, W., Yuen, C., Mohsenian-Rad, H., et al.: Transforming energy networks via peer-topeer energy trading: the potential of game-theoretic approaches. IEEE Signal Process. Mag. 35(4), 90–111 (2018) 62. Moret, F., Pinson, P., Papakonstantinou, A.: Heterogeneous risk preferences in communitybased electricity markets. European J. Oper. Res. 287(1) (2020) 63. DabbaghS, R., Sheikh-El-Eslami, M.K.: Risk-based profit allocation to DERs integrated with a virtual power plant using cooperative Game theory. Electric Power Syst. Res. 121, 368–378 (2015) 64. Tamás, F., Zsuzsanna, J., Tamura, A., et al.: Trading networks with bilateral contracts. EAI Endorsed Trans. Serious Games 3(11) (2015) 65. Wang, H., Huang, J.: Incentivizing energy trading for interconnected microgrids. IEEE Trans. Smart Grid 9(4), 2647–2657 (2018) 66. Zhao, L., Zhang, W., Hao, H., et al.: A geometric approach to aggregate flexibility modelling of thermostatically controlled loads. IEEE Trans. Power Syst. 32(6), 4721–4731 (2017) 67. Jahromi, A.A., Bouffard, F.: On the loadability sets of power systems—part I: characterization. IEEE Trans. Power Syst. 32(1), 137–145 (2017) 68. Silva, J., Sumaili, J., Bessa, R.J., et al.: Estimating the active and reactive power flexibility area at the TSO-DSO interface. IEEE Trans. Power Syst. 33(5), 4741–4750 (2018) 69. Kalantari, A., Restrepo, J.F., Galiana, F.D.: Security-constrained unit commitment with uncertain wind generation: the loadability set approach. IEEE Trans. Power Syst. 28(2), 1787–1796 (2012) 70. Wang, J., Qin, J., Zhong, H., et al.: Reliability value of distributed solar+storage systems amidst rare weather events. IEEE Trans. Smart Grid 10(4), 4476–4486 (2019)

24

1 Introduction

71. Sidhu, A.S., Pollitt, M.G., Anaya, K.L.: A social cost benefit analysis of grid-scale electrical energy storage projects: a case study. Appl. Energy 212, 881–894 (2018) 72. Contreras-Ocaña, J.E., Siddiqi, U., Zhang, B.: Non-wire alternatives to capacity expansion. In: 2018 IEEE Power & Energy Society General Meeting (PESGM), pp. 1–5. IEEE (2018) 73. Andrianesis, P., Caramanis, M., Masiello, R.D., et al.: Locational marginal value of distributed energy resources as non-wires alternatives. IEEE Trans. Smart Grid 11(1), 270–280 (2019) 74. The application of hydrogen energy industry. [Online]. Available: http://www.peric.ac.cn/en/ cpzs.asp?id=162 75. Wind power conversion to hydrogen project of Guyuan, Hebei. [Online]. Available: https:// chuneng.bjx.com.cn/news/20201208/1120775.shtml 76. Nagamatsu, J., Nakagawa, N., Muranaka, T., et al.: Superconductivity at 39 K in magnesium diboride. Nature 410(6824), 63–64 (2001) 77. Zhang, J., Qiu, M., Chen, P., et al.: Research status and key technologies of hybrid energy transfer line. CAryo&Supercond 49(2), 1–7+31 (2020) 78. Sweden’s long-term strategy for reducing greenhouse gas emissions. [Online]. Available: https://unfccc.int/sites/default/files/resource/LTS1_Sweden.pdf 79. China’s electric vehicle ownership may exceed 100 million by 2030. [Online]. Available: https://www.sohu.com/a/288819079_115737 80. Zhang, C., Greenblatt, J.B., MacDougall, P., et al.: Quantifying the benefits of electric vehicles on the future electricity grid in the midwestern United States. Appl. Energy 270, 115174 (2020) 81. Zhou, M., Wu, Z., Wang, J., et al.: Forming dispatchable region of electric vehicle aggregation in microgrid bidding. IEEE Trans. Ind. Inform. 17(7), 4755–4765 (2021) 82. The future of urban mobility. [Online]. Available: https://www.uber.com/us/en/elevate/ 83. Liu, J., Zhang, N., Kang, C., et al.: Cloud energy storage for residential and small commercial consumers: a business case study. Appl. Energy 188, 226–236 (2017) 84. Chakraborty, P., Baeyens, E., Poolla, K., et al.: Sharing storage in a smart grid: a coalitional game approach. IEEE Trans. Smart Grid 10(4), 4379–4390 (2018) 85. Wang, J., Zhong, H., Wu, C., et al.: Incentivizing distributed energy resource aggregation in energy and capacity markets: an energy sharing scheme and mechanism design. Appl. Energy 252, 113–471 (2019) 86. Internet of Things (IoT) connected devices installed base worldwide from 2015 to 2025 (in billions). [Online]. Available: www.statista.com/statistics/471264/iot-number-of-connecteddevices-worldwide/ 87. Deployment of a datahub shared by all system operators to support electrical flexibility. [Online]. Available: http://www.elia.be/~/media/files/Elia/PressReleases/2018/201 80308_SYN_Persbericht_Datahub_EN.pdf 88. Sonnen’s ‘Community’: Aggregating domestic battery storage. [Online]. Available: https:// greycellsenergy.com/examples/sonnens-community-aggregating-domestic-battery-storage/ 89. Air pollution: Dirty diesel generators test epa, demand response industry. [Online]. Available: http://www.eenews.net/stories/1059967047 90. Liu, Z., Wierman, A., Chen, Y., et al.: Data center demand response: avoiding the coincident peak via workload shifting and local generation. Perform. Eval. 70(10), 770–791 (2013) 91. Ghazzai, H., Yaacoub, E., Alouini, M.S., et al.: Optimized smart grid energy procurement for LTE networks using evolutionary algorithms. IEEE Trans. Veh. Technol. 63(9), 4508–4519 (2014) 92. Chih-Lin, Rowell, C., Han, S., et al.: Toward green and soft: a 5G perspective. IEEE Commun. Mag. 52(2), 66–73 (2014) 93. China Southern Power Grid joins hands with China Tower. [Online]. Available: http://finance. sina.com.cn/roll/2018-03-27/doc-ifysrwvt1908466.shtml 94. State Grid, China Southern Power Grid and China Tower signed strategic cooperation agreement. [Online]. Available: https://baijiahao.baidu.com/s?id=1598695037176899780&wfr= spider&for=pc 95. 5G base station + power towers! Check 18 communication and power sharing tower landing projects. [Online]. Available: https://www.sohu.com/a/427207468_203761

References

25

96. Kong, P., Song, Y.: Joint consideration of communication network and power grid topology for communications in community smart grid. IEEE Trans. Ind. Inform. 16(5), 2895–2905 (2020) 97. Xu, L., Guo, Q., Yang, T., Sun, H.: Robust routing optimization for smart grids considering cyber-physical interdependence. IEEE Trans. Smart Grid 10(5), 5620–5629 (2019) 98. Chen, L., Wang, J., Wu, Z., et al.: Communication reliability-restricted energy sharing strategy in active distribution networks. Appl. Energy 282 (2021) 99. Rost, P., Mannweiler, C., Michalopoulos, D.S., et al.: Network slicing to enable scalability and flexibility in 5G mobile networks. IEEE Commun. Mag. 55(5), 72–79 (2017) 100. Alotaibi, D.: Survey on network slice isolation in 5G networks: fundamental challenges. Procedia Comput. Sci. 182, 38–45 (2021) 101. Abidi, M.H., Alkhalefah, H., Moiduddin, K., et al.: Optimal 5G network slicing using machine learning and deep learning concepts. Comput. Stand. Interf. 76, 103518 (2021) 102. Al-Makhadmeh, Z., Tolba, A.: Independent and tailored network-slicing architecture for leveraging industrial internet of things job processing. Comput. Netw. 187, 107827 (2021) 103. Dennis, R., Owen, G.: Rep on the block: A next generation reputation system based on the blockchain. In: 10th International Conference for Internet Technology and Secured Transactions (ICITST), pp. 131–138. IEEE (2015) 104. Zyskind, G., Nathan, O.: Decentralizing privacy: using blockchain to protect personal data. In: 2015 IEEE security and privacy workshops, pp. 180–184. IEEE (2015) 105. He, Z., Xiaofeng, L., Likui, Z., et al.: Data integrity protection method for microorganism sampling robots based on blockchain technology. J. Huazhong Univ. Sci. Technol. (Natural Science Edition) 43(Suppl 1), 216–219 (2015) 106. Mengelkamp, E., Gärttner, J., Rock, K., et al.: Designing microgrid energy markets: a case study: the Brooklyn Microgrid. Appl. Energy 210, 870–880 (2018) 107. Blockchain in energy trading closer to realization. [Online]. Available: https://group.vat tenfall.com/press-and-media/news--press-releases/newsroom/2018/blockchain-in-energytra ding-closer-to-realisation 108. State Grid has released ten scenarios in the field of blockchain technology and application. http://www.xinhuanet.com/energy/2019-12/20/c_1125370233.html 109. Chaturvedi, D.K., Yadav, S., Srivastava, T., et al.: Electricity storage system: a gravity battery. In: 2020 Fourth World Conference on Smart Trends in Systems, Security and Sustainability (WorldS4), pp. 412–416. IEEE (2020) 110. What is Customer Stickiness and Why Does It Matter? [Online]. Available: https://www.par lor.io/blog/customer-stickiness/ 111. Microsoft’s e2bn Netherlands data center revealed. [Online]. Available: https://www.datace nterdynamics.com/en/news/microsofts-2bn-netherlands-data-center-revealed/

Chapter 2

Mechanism Design for Sharing Economy

2.1 Introduction A reasonable market mechanism is a prerequisite for ensuring the efficient operation of energy markets. One of the most important issues is to motivate market participants to make rational bids, which is an essential condition for Pareto efficiency [1]. A well-designed market mechanism will operate in accordance with the goals and rules expected by its mechanism designer [2]. Generally speaking, the key issues in energy market mechanism design are: (1) How to realize optimal economic dispatch via mechanism design which elicits truthful behaviors? (2) How to guarantee a stable grand coalition that all the market participants would like to join in? Mechanism design is also called reverse game theory [3], which studies how to design an economic incentive or profit sharing rule so that rational market participants can achieve the goals and properties expected by the mechanism designer. The essence of mechanism design is the rigorous theoretical derivation and mathematical structure in game theory [4]. Such a theory and concept have been widely used in many fields [5], including auction theory [6], taxation policy [7], principal-agent [8], etc. At present, the marginal pricing mechanism is generally adopted by most electricity markets [9]. However, both theory and practice have proved that since system marginal price is directly related to the bid of the marginal generator, generators may fail to report their true generation costs, and thus raise market clearing prices [10]. In 2000, the electricity markets in the U.S., e.g., PJM and CAISO, encountered a situation where the price spikes exceeded the normal electricity prices by over 10 times when power supply was in shortage [11]. To ensure efficient competition in electricity markets and suppress the price manipulation, many regions have introduced complicated market supervision mechanisms and suffered high market supervision costs. On the other hand, in future smart distribution grids, as massive distributed energy resources (DERs) are connected to the grid and participate in market transactions [12], retail markets gradually tend to be perfectly competitive. At this time, DER © Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_2

27

28

2 Mechanism Design for Sharing Economy

owners will declare their true costs, otherwise they cannot win the bid. However, with the expansion of retail market scale, the traditional centralized-bidding market framework may not work well: The first reason is that it is extremely costly to collect the bids from thousands of DERs and optimize the market clearing in a centralized fashion; Secondly, distinguished from wholesale markets, the market participants in retail markets lack the motivation and willingness to bid on a daily basis, and they actually pursue a simple profit sharing mechanism [13]. In other words, the only thing that DER owners need to do is to share DERs and get paid. For the problem of incentivizing truthful bids in wholesale markets, a profit sharing mechanism based on the Vickrey–Clarke–Groves (VCG) theory is proposed. In contrast to the marginal pricing mechanism, the VCG-based mechanism defines the payment to a generator as other generators’ incremental costs after removing it in the market clearing process. Such a substitution benefit determines the payment allocated to a generator, which is proved to elicit truthfulness. For the problem of reducing transaction costs in retail markets, a decentralized market framework is developed, and a bidding-free mechanism based on Nash bargaining is proposed. The concept of sharing contribution rate is designed to evaluate each customer’s or DER’s contribution to energy sharing. A closed-form expression of the profit sharing payment is obtained by analytically solving the Nash bargaining model, which is proved to fulfill individual rationality and budget balance.

2.2 Problem Description 2.2.1 Wholesale Market In a wholesale market, the traditional marginal pricing mechanism can hardly suppress the willingness of generating units to make strategic bids. Marginal units may raise their bids to increase individual income when the market supply is in shortage. One critical necessary condition for optimal resource allocation is that all market participants declare true marginal costs. Due to the existence of market power, market participants may submit strategic bids deviating from true marginal costs for price manipulation, causing market efficiency loss. Therefore, this section focuses on studying how to incentivize truthful bids in wholesale markets via mechanism design under the condition of information asymmetry [14]. Here we consider N generators participating in a wholesale market. The cost function ci : R+ → R+ of generator i is private information. Each generator declares its supply function ci : R+ → R+ to a market operator. Note that the declared cost ci can be different from the true cost ci . After collecting all generator bids, the market operator will schedule the generators’ output through economic dispatch, as follows: 



N    min f X, cˆ = cˆ i (Pi ), s.t. X ∈ χ X

i=1

(2.1)

2.2 Problem Description

29 

where X denotes the vector of decision variables, f (X, c) is the objective function of the market clearing model, c is the vector of all generators’ bids, Pi is the variable representing generator i’s output, and χ is the feasible region depicted by load balance, line flow, unit output and other constraints. The market operator aims at minimizing the declared costs over the generator set, satisfying the operation constraints of the power system, and optimally determines the winning bid quantity P∗ of each generator in the market. Finally, the operator make settlements for the bid-winning generators according to the proposed profit sharing mechanism. The settlement mechanism defines a function that maps generator bid c to the generators’ payment π, where πi represents the payment of the market operator to generator i, thus forming a game of N generators. Therefore, the goal of the mechanism designer in a wholesale market is to make “declaring the true marginal cost” to be the optimal bidding strategy through mechanism design. 



2.2.2 Retail Market With the integration of a large number of DERs, the market will tend to be perfectly competitive. Perfectly competitive market refers to a market structure with sufficient competition without any hindrance and interference. At this time, any market participant is a price taker and cannot change the market price by strategically bidding. Thus, market participants will faithfully report their marginal costs. In this instance, we make a mild assumption that the market participants are price takers and will not abuse market power in retail markets. In a distribution network, we consider an aggregator acting as an agent for N users. Each user has a local power load and DERs, such as rooftop photovoltaics, battery energy storage, heat pumps, etc. The aggregator organizes retail sales and is responsible for market scheduling and settlement among the users. At the same time, the aggregator is responsible for purchasing electricity from the upstream transmission grid to meet users’ load demands, and selling the surplus power from the shared DERs to the transmission grid. The revenue function r A of the aggregator indicates the revenue that the aggregator gets from the users, minus the cost paid by the aggregator to the transmission grid. The superscript A represents aggregator. Initially, when the users do not participate in the retail market without sharing DERs, they only purchase electricity from the aggregator at a fixed retail rate, and sell the surplus power of the DERs back to the aggregator at a fixed net metering rate (NMR). Here all the users are assumed to be rational, with the goal of minimizing the net costs of purchasing and selling electricity, while optimizing DERs and managing local load, shown as follows: min ciU , s.t. Xi ∈ χi Xi

(2.2)

30

2 Mechanism Design for Sharing Economy

where the superscript U represents user.Xi is the vector of the decision variables of user i, ciU is the cost function of user i, and χi is the feasible region of user i. Under the condition that users do not participate in the retail market, the net cost of user i is denoted by ciU ,0 , and the net income of the aggregator is recorded as r A,0 . Then, we discuss about the case where the aggregator organizes a retail market enabling users to share DERs with neighbors. In this case, each user is not individually optimal, and should follow the instructions from the aggregator to share the available DERs. The objective of the aggregator is to minimize the total costs of the entire entity formed by the aggregator and all users. The optimization model is formulated as follows: min −r A + X

N 

ciU , s.t. X ∈ χ ∩ χ S

(2.3)

i=1

where χ is the local constraint set of all users, and χ S indicates the balance constraint in the retail market. At this time, the optimal value of the net income of the aggregator is denoted by r A,1 , and the net cost of user i is ciU ,1 . In contrast to the case where users do not participate, the retail market actually maximizes the social welfare of the aggregator and all users. By this means, the aggregator can respond to the wholesale prices by dispatching the DERs shared from the users, thus improving the utilization of DERs while saving the power purchase costs charged by the transmission grid. Furthermore, the shared DERs can help to achieve the effect of peak shaving and other kinds of ancillary services, thereby enhancing the Pareto efficiency of the distribution grid. It should be noted that some of the users share DERs, which may deviate from individual optimum, i.e., ciU ,1 > ciU ,0 . Therefore, the aggregator need a profit sharing mechanism to identify the value created by each user and make reasonable settlements. Here we denote πi as the payment of the aggregator to user i, so that the  net cost of user i becomes ciU − πi , and the net income of the aggregator is r A − i πi . Basically, the market mechanism should distinguish the users’ unique contributions to energy sharing, and guarantee that the net income of each user after participating in the market is positive.

2.3 Profit Sharing Mechanism In the study of economics and mechanism design theory, profit sharing mechanism refers to the process by which multiple market participants decide how to share the benefits of a public good or service. Existing research proposes a variety of profit sharing mechanisms to solve diverse settlement problems for different commodities, different services and different cost functions. The core idea of profit sharing is to carry out fair benefit allocation according to the value created by each market participant. For wholesale markets, due to the insufficient incentive paid to the marginal

2.3 Profit Sharing Mechanism

31

generator, market participants have an essential motivation for strategically bidding. Thus, the profit sharing mechanism in wholesale market should distinguish the value created by the marginal generator. For retail markets, however, the profit sharing mechanism should be bidding-free, and enable the identification of each DER or energy user for his/her shared availability [15]. Before designing detailed market mechanisms, we focus on some standard properties to quantify the performance of a mechanism, i.e., Social Welfare Maximation (SWM), Individual Rationality (IR), Incentive Compatibility (IC), and Budget Balance (BB) [16].

2.3.1 Social Welfare Maximation In economics, social welfare is the sum of individual welfare. SWM refers to maximizing the difference between the utilities and the costs of all individuals. Specifically, in electricity markets, SWM means that the power generation and load dispatch obtained via market clearing can achieve the operating condition of minimizing the true costs of the entire system, shown as follows: X∗∗ ∈ arg min f (X, c) s.t. X ∈ χ X

(2.4)

where X∗∗ represents the optimization result when all generators declare the true marginal costs. When all the power generation resources participate in the market and the marginal cost is faithfully declared, the cost minimization-oriented market clearing model will automatically fulfill the property of SWM. It is worth noting that if the strategic bid of the generator set deviates from the true marginal cost, or a DER owner exits from the market, SWM may not be satisfied.

2.3.2 Individual Rationality Individual rationality refers to the voluntary participation of market participants in market competition. For a mechanism that satisfies IR, the net profit of each market participant is greater than or equal to zero, otherwise the market participant will choose to quit from the market, shown as follows:   πi − ci Pi∗ ≥ 0, ∀i

(2.5)

Only by satisfying IR constraints can market participants have incentives and willingness to participate in markets.

32

2 Mechanism Design for Sharing Economy

2.3.3 Incentive Compatibility In 1972, American economist Leonid Hurwicz put forward the concept of incentive compatibility. In a deregulated market, every rational market participant aims to maximize his/her individual benefits. If a market mechanism achieves the consistency between individual optimum and the collective goal of maximizing social welfare, this mechanism satisfies IC. Specifically, in wholesale markets, IC indicates that it is the best choice for each generator to declare the true marginal cost, rather than other strategic bids, i.e.,         πi ci , cˆ −i − ci Pi∗∗ ≥ πi cˆ − ci Pi∗ , ∀i

(2.6)

IC can be further divided into Domiant-Strategy Incentive Compatibility (DSIC) and Bayesian-Nash Incentive Compatibility (BNIC). DSIC means that truthfully bidding is the dominant strategy, regardless of whether the declaration of other units is true or not. BNIC indicates that only if other units bid truthfully, a generator’s faithful declaration is the best choice. It can be observed that DSIC is a stricter property. In this book, we mainly focus on DSIC.

2.3.4 Budget Balance In real-world electricity markets, independent system operator (ISO) is a non-profit organization (NPO). For example, the ISO in U.S. is supervised by the Federal Energy Regulatory Commission (FERC), which is mainly responsible for coordinating, controlling and monitoring the operation of power systems. Therefore, BB means that the payment to generation supply must be less than the charges from load demand. Otherwise, there will exist revenue inadequacy. It should be noted that, according to Hurwicz impossibility theorem, there does not exist a market mechanism that simultaneously satisfies the aforementioned four properties. Therefore, in practice, mechanism designers should develop specific market mechanisms based on the realistic requirements to find an appropriate solution to different challenges. In wholesale markets, this chapter focuses on how to design a market mechanism to meet incentive compatibility, i.e., the optimal bidding strategy is to declare true marginal costs. In retail markets, this chapter focuses on how to design market mechanisms to satisfy individual rationality, i.e., the net income of all users should be non-negative.

2.4 Profit Sharing Mechanism in Wholesale Markets

33

2.4 Profit Sharing Mechanism in Wholesale Markets In mature electricity markets, the marginal pricing mechanism is widely used. When the market tends to be perfectly competitive, all the generators will truthfully bid. Therefore, the marginal pricing mechanism is efficient and can maximize social welfare. However, as system marginal price is directly related to the bid of the marginal generator, the generators may speculate on the bidding price of the unsuccessful unit, and strategically report the generation cost. As a result, the market clearing price can be forced to increase, thereby leading to an imbalance between producer and consumer surplus. The principle of our proposed profit sharing mechanism is to make market participants get rewards based on the value they create. This section focuses on the theory and idea of the Vickrey-Clarke-Groves (VCG) mechanism for eliciting truthfulness, while applying such a mechanism in wholesale markets with respect to the operation constraints of power systems. VCG mechanism naturally satisfies dominant-strategy incentive compatibility, and is named after the initials of the three economists, William Vickrey, Edward H. Clarke and Theodore Groves, to commemorate their contribution in mechanism design theory. William Vickrey and another economist also won the 1996 Nobel Prize in Economics.

2.4.1 Wholesale Market Stage According to the degree of generation adequacy, the operation of an electricity market is divided into three stages: the market tends to be perfectly competitive; the generation tends to be inadequate but the N-1 generation capacity is still sufficient; and the N − 1 generation capacity is insufficient, as shown in the Fig. 2.1. Stage 1: When the market tends to be perfectly competitive, power generation resources are much more abundant than loads. At this time, all generators will truthfully bid the marginal costs, and the marginal pricing mechanism can achieve incentive compatibility.

Capacity Sufficient

N-1 capacity insufficient The degree of generation adequacy

N-1 capacity sufficient Fig. 2.1 Three stages of electricity market operation according to the degree of generation adequacy

34

2 Mechanism Design for Sharing Economy

Stage 2: With the growth of load demands, the generation supply will approach inadequacy. Here we consider that load demand can still be balanced after removing any generator from the market, namely N − 1 generation capacity is sufficient. At this time, under the marginal pricing mechanism, the marginal generator has the motivation to strategically report the generation cost for price manipulation. Stage 3: When the load demands futher increase so that N − 1 generation capacity is insufficient, i.e., removing some generator from the market will lead to load imbalance, market failure will occur due to monopoly. The marginal generator will keep submitting high prices until the price cap is reached. In this chapter, we focus on designing VCG-based mechanism for Stage 2, and a supplementary mechanism based on market supervision for Stage 3.

2.4.2 Sufficient N − 1 Power Generation Capacity Based on VCG auction, the value of a generator is defined as the substitution benefit of the generator to others, which can be calculated as the change in the total costs of the other generators before and after a generator participates in the market. The payment received by generator i is:          πi cˆ = f−i X∗−i , cˆ −i − f X∗ , cˆ − cˆ i X∗i

(2.7)

∗ where f−i (·) and X−i represent the system cost function and optimal dispatch result when generator i does not participate in the market, respectively. Therefore, the first ∗ , c−i ) represents the total cost of N− 1 generators when generator i term f−i (X−i does not participate, and the second term f X∗ , c − ci (Xi∗ ) represents the total cost of N − 1 generators when all the generators participate, i.e., the total system cost minus that of generator i. Next, we compare the proposed VCG-based mechanism with marginal pricing, and explain how VCG achieves incentive compatibility (Fig. 2.2). There are N generators participating in the electricity market, and their true marginal costs are c1 ≤ c2 ≤ · · · ≤ cN . Suppose the cost-minimization market clearing mechanism makes generators from 1 to N − 1 win the bids. Under marginal pricing, the system marginal price is directly related to the bid of the marginal generator. Thus the marginal generator has incentive to speculate on the Nth unit’s bid and 



N units

c1

c2

Fig. 2.2 Market clearing process



M arket clearing

cN-1

cN

2.4 Profit Sharing Mechanism in Wholesale Markets

35

try to raise the bid. Such a strategic behavior may lead to the increase in the settlement for all the generators, thereby resulting in a severe deviation between producer and consumer surplus. However, under the VCG-based profit sharing mechanism, the value of the marginal unit is determined by the incremental cost of other generators after removing it from the market. Reporting strategic bids may decline the marginal unit’s generation share and even lead to an unsuccessful bid. Therefore, the marginal unit will intuitively bid its true cost. Compared with marginal pricing, the VCG-based profit sharing mechanism can encourage generators to elicit truthfulness, while satisfying the principle of incentive compatibility. Theoretical proofs are provided as follows.

2.4.3 Proof of the Mechanism Property 2.4.3.1

Incentive Compatibility

To prove that the profit sharing mechanism satisfies DSIC, we can equivalently prove that the net income of any generator when truthfully bidding is not less than that when the generator strategically bids. When generator i declares its true cost ci , and other N − 1 generators declare arbitrary cost c−i (whether true or not), the payment to generator i is: 

         πi ci , cˆ −i = f−i X∗−i , cˆ −i − f X∗∗ , ci , cˆ −i − ci X∗∗ i

(2.8)

where X∗∗ represents the optimal generation dispatch when generator i declares ci and others report c−i , expressed as follows: 

  X∗∗ ∈ arg min f X, ci , cˆ −i s.t. X ∈ χ X

(2.9)

The net income of generator i is the payment minus its realized generation cost:     πi ci , cˆ −i − ci X∗∗ i          = f−i X∗−i , cˆ −i − f X∗∗ , ci , cˆ −i − ci X∗∗ − ci X∗∗ i i ⎡ ⎤ N    ∗    ⎦ + ci X∗∗ = f−i X−i , cˆ −i − ⎣ cˆ j X∗∗ j i

(2.10)

j=1,j =i 

When generator i strategically declares ci , and other N − 1 generators still declare c−i , generator i receives the following payment: 

         πi cˆ = f−i X∗−i , cˆ −i − f X∗ , cˆ − cˆ i X∗i

(2.11)

36

2 Mechanism Design for Sharing Economy

Then the net income of generator i is:     πi cˆ − ci X∗i          = f−i X∗−i , cˆ −i − f X∗ , cˆ − cˆ i X∗i − ci X∗i ⎡ ⎤ N    ∗    = f−i X−i , cˆ −i − ⎣ cˆ j X∗j + ci X∗i ⎦

(2.12)

j=1,j =i

By comparing Eqs. (2.10) and (2.12), one can observe that the first term ∗ f−i (X−i , c−i ) at the right hand is only related to other N − 1 generators’ bids, which has nothing to do with the bid of generator i. As shown in (2.9), the optimal dispatch result X∗∗ satisfies: ⎡ ⎤ N    X∗∗ ∈ arg min ⎣ cˆ j Xj + ci (Xi )⎦ s.t. X ∈ χ (2.13) 

X

j=1,j =i

Therefore, the following inequality can be obtained: N 

N     ∗∗    + c X ≤ cˆ j X∗∗ cˆ j X∗j + ci X∗i i j i

j=1,j =i

(2.14)

j=1,j =i

        πi ci , cˆ −i − ci X∗∗ ≥ πi cˆ − ci X∗i i

(2.15)

In other words, the net income of any generator when truthfully bidding is not less than that when the generator strategically bids, regardless of the bids of other N − 1 generators. Therefore, we prove that the profit sharing mechanism satisfies DSIC.

2.4.3.2

Individual Rationality

According to the definition of the payment function, the net income of generator i can be interpreted as the change in system costs before and after generator i participates in the market, shown as follows:      ∗∗  = f−i X∗∗ πi (c) − ci X∗∗ i −i , c−i − f X , c

(2.16)

The left-hand term represents the net income of generator i. The first term on the right hand indicates the system cost without generator i’s participation, and the second term represents the system cost. It should be noted that we can model the non-participation of generator i by adding constraint Xi = 0, forcing the outputs of generator i to be 0. Therefore, the optimization problem

2.4 Profit Sharing Mechanism in Wholesale Markets

37

min f−i (X−i , c−i ) s.t. X−i ∈ χ−i

(2.17)

min f (X, c) s.t. X ∈ χ , Xi = 0

(2.18)

X−i

is equivalent to X

Therefore, when generator i does not participate in the market, the feasible region of the market clearing optimization is cut by Xi = 0, whose optimal objective value is not less than that when all the generators participate, i.e.,    ∗∗  f−i X∗∗ −i , c−i − f X , c ≥ 0

(2.19)

Therefore, the net income of any generator is greater than or equal to 0. The profit sharing mechanism satisfies IR.

2.4.3.3

Social Welfare Maximation

The previous sections have proved that the profit sharing mechanism satisfies dominant strategy incentive compatibility and individual rationality. Therefore, any generator declares its true marginal cost and has incentive to participate in the market. The profit sharing mechanism automatically satisfies SWM. It should be noted that the VCG-based profit sharing mechanism has fulfilled three of the aforementioned four properties. According to Hurwicz impossibility theorem, the budget imbalance issue will be discussed in Chap. 3 in a detailed spot market model.

2.4.4 Insufficient N − 1 Power Generation Capacity 2.4.4.1

Motivation

The VCG-based profit sharing mechanism requires to remove each generator from the market, so as to calculate its substitution value in place of other units. However, when the N − 1 generation capacity is insufficient to balance electric loads, VCG mechanism cannot work. Intuitively, the generation adequacy is extremely insufficient and market failure tends to occur due to monopoly. Therefore, this subsection proposes a supplementary market supervision mechanism. In practice, market supervision is an important measure to maintain the orderly operation of markets, which can be classified into two major types: (i)

Service cost supervision, e.g., rate of return regulation, which limits the rate of return of power generation companies.

38

(ii)

2 Mechanism Design for Sharing Economy

Incentive-based supervision, e.g., price cap regulation, limiting generators’ highest bids or price fluctuations during a period.

Due to the severe market failure in California ISO in 2000, the electricity markets in many regions across the world have realized the importance of market supervision. CAISO requires that the average price increase within three months must not exceed 10%. PJM proposes three pivotal supplier (TPS) test for market power detection. Some other market operators have designed other price-cap rules, e.g., setting a 200 or 300% cap of the average price over past one month. The existing market supervision methods rely heavily on the experience from market operation, e.g., to what extent a market operator sets a price cap, which lacks a theoretical basis and may suppress the enthusiasm of market participants.

2.4.4.2

Mechanism Design

Therefore, we theoretically design a price-cap mechanism by taking advantage of other N − 1 generators’ average bids, which is a similar idea with VCG. During a relatively long time horion, the fuel consumption rate of a generator is fixed, and its bid mainly reflects the power generation cost, approximately equaling the fuel consumption rate timed by fuel ruling price. Thus, the bids of different generators reveal their differences in fuel consumption rates that are technical parameters. We can take advantage of this observation for mechanism design. The steps are elaborated as follows: (1)

(2) (3)

(4)

Collect the bidding data of all the generators when N − 1 generation capacity is sufficient. Note that based on VCG theory, these bids reveal true marginal costs. Let the bid of generator i be ci , and the average bid of other N − 1 generators is denoted by c−i . Calculate the relative fuel consumption rate of generator i as ηi = ci /c−i . When the situation with insufficient N − 1 generation capacity occurs, namely removing generator i leads to load imbalance, we can use the current bids of other N-1 generators to set the price cap for generator i. The average bid of

others at this time is denoted by c−i , and the price cap for generator i is defined



as ci = ηi c−i . The settlement price for each generator is the minimum of its price cap and its



bid, i.e., min{ci , ci }.

Then we explain how this supplementary mechanism works. Since the upper limit (price cap) of the marginal generator’s bid is determined according to other N − 1 generators’ bids, the marginal generator has no incentive to raise its bid. On the other hands, to prevent the marginal generator from strategically reporting low prices thus damaging the producer surplus of other units, we propose that the settlement price for each generator is the minimum of its price cap and its bid. Reporting prices lower than the true costs will lead to the loss of the generators’ profits.

2.4 Profit Sharing Mechanism in Wholesale Markets

39

To conclude, the VCG-based profit sharing mechanism has been designed in different stages of a market. In contrast to the traditional heuristic supervision mechanisms, our proposed mechanism takes into a reasonable consideration the fluctuation of fuel ruling prices. Therefore, the entire mechanisms can work well for different market stages even when N − 1 generation capacity is insufficient.

2.4.5 Major Challenges of VCG in Electricity Markets Electricity is different from other general commodities. The operation of electricity markets must satisfy certain physical and economic constraints. In this book, we make a discussion about the following major challenges of VCG for its real-world implementation. (1)

What is the solution to budget imbalance problem caused by the VCG-based profit sharing mechanism?

From Hurwicz’s impossibility theorem, it can be seen that the profit sharing mechanism that achieves DSIC does not satisfy budget balance, i.e., the charges from electric loads may not cover the payments to generators. As a result, the market operator has to suffer revenue inadequacy, violating its role as a non-profit organization. Such a revenue inadequacy is the extra payment to the marginal generator, which can be interpreted as the profit earned by the marginal generator by taking advantage of information asymmetry, or the payment from the market operator to avoid the marginal generator’s strategic bidding. In economics, this fraction is termed as “information rent”, which is designed to be allocated among customers in this book. (2)

What method can be used to improve the optimization efficiency of the VCGbased profit sharing mechanism?

According to VCG, the settlement for N generators requires N + 1 times of optimization. However, the traditional marginal pricing mechanism only needs one time of optimization for market clearing and pricing. Therefore, how to improve the optimization efficiency of the VCG-based profit sharing mechanism has become one of the key problems in practical application. In fact, removing one generator from the market is equivalent to that other N − 1 generators take charge of the removed generator’s market share. Theoretically, the sensitivity analysis of a general convex optimization program can be used to avoid some steps of re-optimization. To this end, we will further investigate the algorithm for improving the efficiency of the profit sharing mechanism, which will be elaborated in next chapter based on a detailed market operation model.

40

2 Mechanism Design for Sharing Economy

2.4.6 Case Studies In this section, we compare the performance of traditional marginal pricing and the VCG-based profit sharing mechanism. In the Fig. 2.3, the true costs of unit 1, 2 and 3 are c, 2c and 3c, respectively. The load at bus 1 is denoted by D. The output ranges of the three generators are all [0, 2D]. The line capacity between bue 1 and 2 is 0.5D.

2.4.6.1

Marginal Pricing Mechanism

The marginal price reflects the bid level of the marginal unit that meets the last MWh of power systems. Under the marginal pricing mechanism, generators are willing to make strategic bids and take advantage of market power to increase their own revenues, which may eventually lead to suboptimality of system operation. When each unit declares the true marginal cost, the cost-minimization dispatch result is: unit 1 generates D, and unit 2 and 3 generate 0. At this time, the marginal price of bus 1 is c, and the net incomes of the three units are all 0. However, suppose when unit 1 and 2 declare 3c − ε at the same time, where ε is a very small positive number, and unit 3 still declares 3c, the optimal dispatch result will be changed: unit 1 and 2 generate 0.5D, respectively. At this time, the marginal prices of bus 1 and 2 are both 3c − ε, and the net incomes of the three units are (2c − ε) × 0.5D, (c − ε) × 0.5D, and 0, respectively. It can be seen that the incomes of unit 1 and 2 when strategically bidding are greater than those when truthfully bidding. Also, note that the strategic behaviors can lead to a deviation from the optimality. Therefore, it is necessary to elicit truthfulness while fulfilling the maximization of social welfare.

2.4.6.2

Profit Sharing Mechanism

Here we focus on verifying that the profit sharing mechanism realizes the principle of incentive compatibility. (1)

For unit 1, if unit 1 and 2 declare 3c − ε at the same time, they both generate 0.5D. The generation costs of unit 2 and 3 are (3c − ε)×0.5D+0 = (3c − ε)×

Fig. 2.3 Structure diagram of 2 nodes and 3 units

Unit 1

Unit 2 Line

Unit 3

Load

Bus 1

Bus 2

2.4 Profit Sharing Mechanism in Wholesale Markets

41

0.5D. Then we remove unit 1 from the market, and the dispatch result is that unit 2 and 3 both generate 0.5D. The generation costs of unit 2 and 3 are (3c − ε) × 0.5D + 3c × 0.5D. Therefore, the VCG-based payment to unit 1 is:   π1 cˆ = (3c − ε) × 0.5D + 3c × 0.5D − (3c − ε) × 0.5D = 1.5cD

(2.20)

In this case, the net income of unit 1 is cD. (2)

If unit 1 declares the true marginal cost c, but unit 2 still declares 3c − ε, the dispatch result will be that unit 1 generates D, and unit 2 and 3 generate 0. At this time, the generation costs of unit 2 and 3 are 0. Then we remove unit 1 from the market, the generation costs of unit 2 and 3 are (3c − ε)×0.5D+3c×0.5D. Therefore, the VCG-based payment to unit 1 is:   π1 c1 , cˆ −1 = (3c − ε) × 0.5D + 3c × 0.5D = (3c − 0.5ε)D

(2.21)

In this case, the net income of unit 1 is (2c − 0.5ε)D. It can be seen that the net income of unit 1 when truthfully bidding is greater than that when strategically bidding. (3)

If both unit 1 and 2 declare their true marginal costs, the payment to unit 1 can similarly calculated below: π1 (c) = 2c × 0.5D + 3c × 0.5D = 2.5cD

(2.22)

Note that the strategic bidding of unit 2 will increase the net income of unit 1, and thus unit 2 has no incentive to do so. In summary, the profit sharing mechanism satisfies DSIC. In addition, when all units declare the true marginal costs, individual rationality and social welfare maximization can be easily verified.

2.5 Profit Sharing Mechanism in Retail Markets With a large number of DERs, controllable loads, microgrids and other resources participating in retail markets, the traditional centralized market model can hardly work. One of the reasons is that a market operator has to collect massive DERs’ bids, which causes huge transaction costs. The second reason is that different from profitmaximizing generators in wholesale markets, energy users are pursuing a simple market mechanism, and are reluctant to frequently bid on a daily basis. Therefore, in distribution-level retail markets, if the value created by each user and his/her DERs

42

2 Mechanism Design for Sharing Economy

can be precisely identified and allocated without frequently bidding, huge transaction costs can be avoided, which can be readily implemented for Pareto efficiency. To this end, we propose a sharing economy-based market model and profit sharing mechanism to explore the possibility of identifying the value created by each user without frequently bidding. Such an incentive mechanism in retail markets can be modeled as a typical profit sharing problem, i.e., the allocation of cooperative surplus jointly created by a number of market participants. Here we adopt Nash bargaining (NB) theory.

2.5.1 Retail Market Model In this section, the model of a retail market is formulated. In a distribution network, an aggregator is considered to act as an agent for N users to participate in market transactions. At the beginning, each user achieves local optimum in an independent fashion, with the goal of minimizing the cost of purchasing electricity, while optimizing the operation of DERs and controllable loads, i.e., min ciU , s.t. Xi ∈ χi Xi

(2.23)

At this time, the net cost of user i is ciU ,0 , and the net income of the aggregator is recorded as r A,0 , which is called Disagreement Point. After establishing a retail market, the aggregator aims at reaching social welfare maximization, i.e., minimizing the total costs of all users minus his/her income respecting all the DERs and loads: min −r A + X

N 

ciU , s.t. X ∈ χ ∩ χ S

(2.24)

i=1

The optimal value of the net income of the aggregator is denoted by r A,1 , and the cost of user i is ciU ,1 . Note that the aforementioned retail market model requires the aggregator to collect all users’ private information, such as DER parameters, controllable load adjustment range, etc. To avoid transaction costs caused by users’ frequent bids, two decentralized market models are briefly introduced here. One is an intrusive market model based on direct load control. Users who are willing to participate sign a contract to entrust their controllable resources to the aggregator, allowing the aggregator to directly control the users’ DERs, controllable load, etc. The second is a non-intrusive market model based on decentralized decisionmaking, say Lagrangian relaxation. The optimization model can be decomposed into an aggregator program and each user program. The aggregator program only

2.5 Profit Sharing Mechanism in Retail Markets

43

requires the users’ net load information, and a user program only schedules local resources. The aggregator and the users coordinate in an iterative manner, and finally achieve optimal market clearing. The aggregator send coordinated price signals to the users. The users respond to the prices, and optimize local programs to obtain the net load for the next iteration. The iteration continues until converged. Both of the two retail market models can avoid frequent bids from users, thereby effectively simplifying the market transaction process. However, after users participate in the retail market, they may share their own DERs, leading to an increase in individual costs, i.e., ciU ,1 > ciU ,0 . Therefore, the key problem of the retail market is how to identify the value created by each user, and carry out the profit sharing mechanism, so as to ensure each user’s individual rationality.

2.5.2 Mechanism Design According to the core concept of profit sharing mechanism, the key problem of a retail market is how to identify the value created by each user. Nash bargaining theory is adopted to instead of VCG or marginal pricing mechanism. This is because in contrast to wholesale markets, there are quite a lot of users and DERs in retail markets, and thus we do not need to elicit users’ truthfulness. Meanwhile, VCG requires too much computation, which is hardly possible to be implemented in retaile markets considering large-scale DERs. On the other hand, the marginal pricing is based on frequently bidding from market participants, which is just what we are trying to avoid. Here the payment to user i from the aggregator is denoted by πi , which is obtained by optimizing the following Nash bargaining model:  max f X,π

NB

= −r

A,0

+r − A

N 

τ A πi

i=1

N  αi  ciU ,0 − ciU + πi

(2.25)

i=1

s.t. X ∈ χ ∩ χS −r A,0 + r A −

N 

(2.26)

πi ≥ 0

(2.27)

ciU ,0 − ciU + πi ≥ 0, ∀i

(2.28)

i=1

The first term in the objective function represents the profit increase after the aggregator organizes a retail market. The second term represents the cost reduction after user i participates in the market. The objective function f NB represents

44

2 Mechanism Design for Sharing Economy

the product of maximizing the surplus of the aggregator and all users, which is called Cobb–Douglas Utility Function. The economic significance lies in the total system surplus can be allocated to the aggregator and all users according to a certain weight, so as to ensure that each market participant has sufficient incentives to participate. τ A ∈ (0, 1) is the rate of return of the aggregator, which can be determined by the government via market supervision, or the competition in a multi-aggregator market. Equation (2.26) expresses the physical constraints of DERs and retail market constraints. Equation (2.27) shows that the aggregator’s income increases after organizing the retail market, i.e., the individual rationality of the aggregator. Equation (2.28) shows that the cost of user i decreases after participating, i.e., the individual rationality of user i. Parameter αi represents the bargaining power of user i, i.e., the weight coefficient of user i to obtain benefits from the total system surplus. The weight coefficient satisfies: N 

αi = 1 − τ A

(2.29)

i=1

αi ≥ 0, ∀i

(2.30)

It is an open question as to how to design bargaining power. Here we propose a weight setting method reflecting the economic value of DERs. After the market is cleared, the shared power of user i is denoted by Pi∗ . Pi∗ > 0 means that user i is a producer and shares its DERs. Pi∗ < 0 means that user i is a consumer and purchases electricity from the market. Then the bargaining power weight coefficient αi is:  ∗ P  αi = N i  ∗  , ∀i   i=1 Pi

(2.31)

It can be seen that a user’s bargaining power is related to the amount of electricity that has been cleared. A greater amount of electricity indicates a larger value created by this user, who deserves a higher level of bargaining power. Therefore, users have incentive to earn more profits by increasing the bargaining power. Note that we use the absolute value of the cleared electricity, indicating that both producers and consumers should share social welfare. From the above Nash bargaining model, the objective function of the model is to maximize the product of the utility increment of all market participants. In realistic retail markets, the number of users is huge and the model is quite nonlinear, which is extremely complicated to solve. Therefore, we propose the closed form of a desired solution to the profit sharing mechanism without optimizing the Nash bargaining model.

2.5 Profit Sharing Mechanism in Retail Markets

45

2.5.3 Closed Form of Profit Sharing Mechanism The basic idea is to transform the product-based Nash bargaining model into a sum form by logarithm, and derive the first-order optimality condition (KKT). The theoretical derivation and proof will be discussed in the next section. We firstly provide the closed form as follows: πi∗ = αi c + ciU ,1 − ciU ,0 c = r A,1 − r A,0 +

N   ciU ,0 − ciU ,1 ≥ 0

(2.32)

(2.33)

i=1

In the formula, c is called Cooperative Surplus, indicating the total benefits created by the aggregator and users. It can be seen from that in the profit sharing mechanism, the optimal payment πi∗ is composed of two parts: The first part αi c is the income that the user obtains from the cooperative surplus according to weight αi . The second part ciU ,1 − ciU ,0 equals user i’s incremental cost after participating in the market. Such a profit sharing mechanism has very intuitive economic explanation: Firstly, the cooperative surplus is allocatd to each user according to his/her bargaining power. Secondly, each user gets paid according to how much he/she sacrifices individual optimality. These terms can be directly obtained after the market is cleared without users’ frequently bidding. According to (2.32), the net incomes obtained by the aggregator and the users and aggregators are:  −r

A,0

+ r

A,1

−

N 

 πi∗

= τ A c ≥ 0

(2.34)

i=1

 ciU ,0 − ciU ,1 − πi∗ = αi c ≥ 0, ∀i

(2.35)

The net incomes are all non-negative, thus ensuring the willingness of market participants to participate, i.e., individual rationality.

2.5.4 Proof of the Mechanism Property 2.5.4.1

Social Welfare Maximization

The NB-based profit sharing mechanism is obtained by using the logarithm form of Nash bargaining. Therefore, we firstly prove that the logarithm Nash bargaining is convex, and then the closed form is derived.

46

2 Mechanism Design for Sharing Economy

The Nash bargaining model is reformulated as follows: max ln f X,π

NB

N     A,0 A r = τ ln −r + r + π + αi ln ciU ,0 − ciU + πi A

(2.36)

i=1

s.t. X ∈ χ ∩ χS πr +

N 

πi = 0

(2.37)

(2.38)

i=1

Note that the individual rationality constraints of aggregators and users have been automatically considered in the domain of the logarithmic function. To prove that the logarithm Nash bargaining is convex, we calculate the Hessian matrix of the logarithmic objective function. The second-order partial derivative of the objective function is:  U 2 ∂ci ∂ 2 ciU αi αi ∂ 2 ln f NB = − −  2 ∂X ∂X2i ciU ,0 − ciU + πi ∂X2i i ciU ,0 − ciU + πi

(2.39)

∂ 2 ln f NB =0 ∂Xi ∂Xj

(2.40)

∂ 2 ln f NB αi = − 2 2 ∂πi ciU ,0 − ciU + πi

(2.41)

∂ 2 ln f NB =0 ∂πi ∂πj

(2.42)

The user’s objective function ciU is a convex function, so we have: ∂ 2 ciU ≥0 ∂X2i

(2.43)

Therefore, Eq. (2.39) is less than or equal to 0, and Eq. (2.41) is also less than or equal to 0. The Hessian matrix of the logarithmic objective function is a diagonal matrix, and the diagonal elements are all less than or equal to 0. So the Hessian matrix is non-positive definite, and the logarithmic objective function is concave. The logarithmic Nash bargaining model is a convex programming problem.

2.5 Profit Sharing Mechanism in Retail Markets

47

Since the logarithmic Nash bargaining model is convex, we can use the Karush– Kuhn–Tucker (KKT) condition to acquire a unique solution to the model: αj ∂ ln f NB −τ A = + U ,0 = 0, ∀j N A,0 A ∂πj cj − cjU + πj −r + r − i=1 πi

(2.44)

Euation (2.44) includes the equations of N users, which can be solved simultaneously: N 

  N  N      U ,0 A A,0 A U ci − ci ciU − ciU ,0 −r − r + πi = 1 − τ +

i=1

i=1

(2.45)

i=1

Furthermore:  πj = αj −r

A,0

+r + A

= αj c(X) + cjU −

N   i=1 cjU ,0 ,

ciU ,0

−

ciU



+ cjU − cjU ,0

(2.46)

∀j

One can observe that the payment to a user depends on the cooperative surplus and the changes in their own costs after participating in the market. By substituting (2.46) into the objective function (2.36), we can get: N    αi ln[αi c(X)] max ln f NB = τ A ln τ A c(X) + X

i=1

(2.47)

∝ ln[c(X)] Therefore, the Nash bargaining model is actually equivalent to the retail market model, both of which minimize the total costs, thereby satisfying SWM. X∗ ∈ arg max f NB ⇔ X∗ ∈ arg min −r A + X

2.5.4.2

X

N 

ciU

(2.48)

i=1

Individual Rationality

If the Nash bargaining model has a feasible solution, the constraints (2.27) and (2.28) in the model make the profit sharing mechanism naturally satisfy IR. The following proof highlights that there must be a feasible solution for the Nash bargaining model. According to the constraints (2.26)–(2.28), the existence of a feasible solution is equivalent to the fulfillment of Nash bargaining. Note that the constraint (2.26) is feasible, otherwise there is no feasible solution for the retail market. It is easy to prove

48

2 Mechanism Design for Sharing Economy

that the necessary and sufficient condition for the satisfaction of individual rationality is that the cooperative surplus is non-negative, i.e., the existence of {πi ≥ 0, ∀i} enables the users and aggregator to satisfy IR, which is equivalent to non-negative cooperative surplus: 

 −r A,0 + r A − Ni=1 πi ≥ 0 ⇔ c ≥ 0 ciU ,0 − ciU + πi ≥ 0, ∀i

(2.49)

Therefore, we only need to prove that the cooperative surplus is non-negative. Since the optimization variables of different users are independent, each user’s individual model is equivalent to the following model: ⎧ min c1U s.t. X1 ∈ χ1 ⎪ ⎪ ⎪ X1 ⎪ ⎪ ⎨ min cU s.t. X2 ∈ χ2 2 X2

⎪ ··· ⎪ ⎪ ⎪ ⎪ ⎩ min cNU s.t. XN ∈ χN

⇔ min

N 

X

ciU s.t. X ∈ χ

(2.50)

i=1

XN

The total costs of the aggregators and users of the model (2.50) must be greater than those of the cost-minimization retail market model. Therefore, c ≥ 0 holds. The Nash bargaining model must have a feasible solution, and the profit sharing mechanism satisfies IR.

2.5.4.3

Budget Balance

Budget balance means that the aggregator and users share the total cooperative surplus. Based on equations (2.34) and (2.35), the sum of the revenues of all market participants can be obtained as: τ c + A

N  i=1

 αi c = τ + A

N 

 αi c = c

(2.51)

i=1

Therefore, the profit sharing mechanism satisfies BB.

2.5.5 Case Studies Here we consider an aggregator organizing two users, as shown in Fig. 2.4. For simplicity, the cost and benefit of the aggregator is not taken into account. We adopt a tiered tariff for charging users: If load is higher than D, the price will be c; if load is lower than or equal to D, the price will be 0.95c. Only two time slots, i.e., peak and off-peak hours, are considered (Table 2.1).

2.5 Profit Sharing Mechanism in Retail Markets

49

Transmission grid

Fig. 2.4 Illustration of an aggregator organizing two users

Aggregator

User1

S torage1

Table 2.1 Load of two users

User 2

Load 1

Load 2

User\Hour

Peak

Off-peak

User 1

0.2D

0.1D

User 2

1.5D

0.1D

Total load

1.7D

0.2D

In addition, user 1 is equipped with an energy storage unit with an capacity of D, whose efficiency is 90%. The first case is that the two users are independent. The peak price is c, and the off-peak price is 0.95c. Due to a small difference between the peak and off-peak prices, user 1 has no incentive to use energy storage for arbitrage. Therefore, the cost of user 1 is c × 0.2D + 0.95c × 0.1D = 0.295cD, and the cost of user 2 is c × 1.5D + 0.95c × 0.1D = 1.595cD. The following cases will compare the performance of marginal pricing and the Nash bargaining mechanisms.

2.5.5.1

Marginal Pricing Mechanism

If the aggregator organizes a retail market to allow user 1 to share the energy storage, user 1’s storage will be used for peak shaving, thereby reducing the peak-to-valley difference and peak price. The optimal operation strategy of the energy storage is: discharging 0.7D during peak hour and charging 0.7D/90% = 0.778D during offpeak hour. This strategy reduces the total peak load to D, and the total off-peak load increases to 0.978D. Therefore, the marginal prices during both peak and off-peak hours are 0.95c. Under the marginal pricing mechanism, the costs of the two users are shown in the Table 2.2. It can be seen from the results that user 1 shares the energy storage after participating in the market, thereby reducing the total costs of the two users. However,

50

2 Mechanism Design for Sharing Economy

Table 2.2 Costs of two users under marginal pricing

User\Scenario

Independent case

Retail market

User 1

0.295cD

0.359cD

User 2

1.595cD

1.52cD

Total costs

1.89cD

1.879cD

after user 1 contributes to peak shaving in the market, he/she deviates from individual optimum with a significant cost increase from 0.295cD to 0.359cD. Due to the absence of market bidding, user 1 cannot cover individual revenue inadequacy via marginal pricing. Thus, it is difficult to guarantee the market participants’ individual rationality, i.e., the willingness to participate in the market.

2.5.5.2

Profit Sharing Mechanism

To ensure that users are willing to participate in the retail market, the profit sharing mechanism is adopted to make settlements. First, we establish the Nash bargaining model for the two users:  max π

c1U ,0 − c1U ,1 + π1

0.5  0.5 c2U ,0 − c2U ,1 + π2

(2.52)

s.t. π1 + π2 = 0

(2.53)

ciU ,0 − ciU ,1 − πi ≥ 0, i = 1, 2

(2.54)

In the model, the payments to user 1 and 2 are denoted by π1 and π2 , respectively. The objective function is to maximize the cost reduction before and after the two users participate in the market. The weight coefficients of the two users are set to 0.5. Constraint (2.53) means budget balance, and constraint (2.54) means individual rationality of the users. The solution is: π1 = 0.07cD, π2 = −0.07cD

(2.55)

The settlement result indicates that user 2 needs to pay 0.07cD to the aggregator, and the aggregator pays user 1 0.07cD, which is budget balanced. Under the profit sharing mechanism, the costs of users 1 and 2 are shown in the Table 2.3. It can be seen that after the two users participate in the market, the total costs can be reduced, while both user 1 and 2 benefit from the retail market. Therefore, the proposed profit sharing mechanism can identify the contribution of different market

2.5 Profit Sharing Mechanism in Retail Markets Table 2.3 Costs of two users under profit sharing

51

User\Scenario

Independent case

Retail market

User 1

0.295cD

0.289cD

User 2

1.595cD

1.59cD

Total costs

1.89cD

1.879cD

participants while fulfilling individual rationality, thereby ensuring users’ willingness to participate.

2.6 Conclusion In this chapter, the current situation and mechanism requirements of wholesale and retail markets are analyzed. The main idea of profit sharing is to precisely identify the actual contributions of each market participant. According to the generation adequacy, wholesale markets can be classified into three types. When the generation capacity tends to be insufficient, we define the payment to a generator as its substitution value by using VCG auctioin theory instead of traditional marginal pricing. Such a mechanism is proved to fulfill incentive compatibility. When N − 1 capacity is even insufficient, we take advantage of the idea of VCG and design a market supervision mechanism based on historical bidding data. The effectiveness is well explained. In retail markets, however, the bottleneck becomes how to make settlements in a bidding-free market due to large-scale DERs and demand side resources. It should be noted that transaction costs could be considerable in practice if a centralized market is organized with ubiquitous resources frequently bidding. To this end, we design a decentralized market framework and propose a Nash bargaining mechanism to allocate the profits that all the market participants jointly create. The mechanism is proved to fulfill budget balance and individual rationality, which guarantees the willingness of users to participate.

References 1. Mcafee, R.P., Mcmillan, J.: Auctions and bidding. J. Econ. Literature 25(2), 699–738 (1987) 2. David, A.K., Wen, F.: Market power in electricity supply. IEEE Trans. Energy Convers. 21(12), 67–68 (2001) 3. Guan, X., Ho, Y.C., Pepyne, D.L.: Gaming and price spikes in electric power markets. IEEE Power Eng. Rev. 21(8), 58–58 (2007) 4. Helman, U., Hobbs, B.F.: Large-scale market power modeling: analysis of the U.S. eastern interconnection and regulatory applications. IEEE Trans. Power Syst. 25(3), 1434–1448 (2010) 5. Kahn, A.E., Cramton, P.C., Porter, R.H., et al.: Uniform pricing or pay-as-bid pricing: a dilemma for California and beyond. Electricity J. 14(6), 70–79 (2001)

52

2 Mechanism Design for Sharing Economy

6. Elmaghraby, W., Oren, S.S.: The efficiency of multi-unit electricity auctions. Energy J. 20(4), 89–116 (1999) 7. Cole, R., Gkatzelis, V., Goel, G.: Mechanism design for fair division: allocating divisible items without payments. In: Proceedings of the Fourteenth ACM Conference on Electronic Commerce, pp. 251–268. ACM, Pennsylvania, USA (2013) 8. Tang, W., Jain, R.: Aggregating correlated wind power with full surplus extraction. IEEE Trans. Smart Grid (2018, accepted) 9. Wang, T., Xu, Y., Ahipasaoglu, S.D., et al.: Ex-post max-min fairness of generalized AGV mechanisms. IEEE Trans. Autom. Control 62(10), 5275–5281 (2017) 10. Contreras-Ocana, J.E., Ortega-Vazquez, M.A., Zhang, B.: Participation of an energy storage aggregator in electricity markets. IEEE Trans. Smart Grid (2018, accepted) 11. Hobbs, B.F., Rothkopf, M.H., Hyde, L.C., et al.: Evaluation of a truthful revelation auction in the context of energy markets with non-concave benefits. J. Regul. Econ. 18(1), 5–32 (2000) 12. Samadi, P., Mohsenian-Rad, H., Schober, R., et al.: Advanced demand side management for the future smart grid using mechanism design. IEEE Trans. Smart Grid 3(3), 1170–1180 (2012) 13. Nekouei, E., Alpcan, T., Chattopadhyay, D.: Game-theoretic framework for demand response in electricity markets. IEEE Trans. Smart Grid 6(2), 748–758 (2015) 14. Wang, Q., Hodge, B.M.: Enhancing power system operational flexibility with flexible ramping products: a review. IEEE Trans. Ind. Inform. (2018, accepted) 15. Zou, X.: Double-sided auction mechanism design in electricity based on maximizing social welfare. Energy Policy 37(11), 4231–4239 (2009) 16. Silva, C., Wollenberg, B.F., Zheng, C.Z.: Application of mechanism design to electric power markets (republished). IEEE Trans. Power Syst. 16(4), 862–869 (2001)

Chapter 3

Sharing Economy in Electricity Spot Markets

3.1 Introduction In 1998, locational marginal price (LMP) mechanism was introduced into the PJM interconnection, in which bidding-based day-ahead energy and frequency regulation markets were established in 2000. California, Texas, New England, New York, Midcontinent and other regional electricity markets in the United States followed PJM to adopt LMP as a major mechanism for market clearing and settlement. UK electricity market initiated an electric power industry reform in 2001, which changed its Pool-based model into Neta. Nordic electricity market, Nord Pool for energy and Nasdaq for financial trading, has organized a centralized-bidding model based on LMP settlement. Currently, most of the mature electricity markets have adopted LMP as a basic trading mechanism. However, both theory and practice have proved that marginal pricing cannot well address the information asymmetry problem, especially when generation adequacy tends to be insufficient or transmission networks are heavily congested. In this case, thermal generators have incentive to strategically report high prices, so as to raise the marginal clearing price and significantly expand their own revenues. PJM, CAISO and some other markets had encountered severe price spikes due to generators’ withholding capacity and strategically bidding in 2000s [1]. Ref. [2] points out that the existence of market power will suppress effective market competition, and market manipulation and transmission congestion are the two main causes of market power. Ref. [3] simulates the phenomenon of price spikes and the loss of market efficiency in CASIO by using a strategic bidding model. Ref. [4] studies the market power of the U.S. eastern market, and evaluates the strategic participation of 2725 generators. Policy suggestions are provided for market supervision in the U.S. Efficient allocation of resources relies on the truthful bidding of all market participants. Due to the existence of market power, truthfulness-deviated bidding may distort market operation and lead to loss of market efficiency. To this end, some existing references have proposed to apply VCG theory in electric power sectors. © Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_3

53

54

3 Sharing Economy in Electricity Spot Markets

Incentive compatibility refers to the mechanism for market participants to rationally maximize individual interests, which is consistent with the strategy expected by mechanism designers. In electricity markets, such consistency generally refers to the truthfulness of market participants. Ref. [5] studies a profit sharing mechanism for general goods among several market participants, which realizes the truthful bids of market participants on the basis of proportional fairness. In [6] and [7], a profit sharing mechanism based on VCG theory is proposed, which helps wind power aggregators to encourage wind producers to truthfully report probability distributions. Ref. [8] conducts a theoretical comparison between LMP and VCG-based pricing mechanisms. In [9], it is proved that VCG mechanism satisfies the properties of incentive compatibility, and a subgradient algorithm is proposed for decentralized solution. In [10], VCG theory is applied to the settlement for battery energy storage systems, so as to achieve true declaration of battery operators. The existing research has applied the VCG mechanism to the design of incentive compatibility mechanism for general commodities, wind power and energy storage agents. However, few studies have explored the application of VCG in a realistic electricity spot market. In contrast to general commodities, electricity markets must obey specific physical and economic constraints. For example, the VCG-based mechanism does not satisfy budget balance, leading to revenue inadequacy of market operators. Additionally, VCG naturally requires multiple rounds of optimization, which is time-consuming and deserves improvement in real world. In this chapter, we are going to propose a revenue inadequacy allocation strategy for budget imbalance issue. To improve the computation efficiency of profit sharing mechanism, an efficient solution algorithm based on sensitivity analysis is formulated by initializing base variables of the simplex method.

3.2 Electricity Spot Market Model This section will establish the clearing model of a spot market. Then the marginal pricing and profit sharing mechanisms will be revisited.

3.2.1 Mathematical Model In the existing mature electricity markets, generators participate in day-ahead energy and ancillary service markets by submitting bidding curves to an independent system operator (ISO). The ISO runs a market clearing model to schedule generators and generate LMPs for settlement [11]. Without loss of generality, the following assumptions are made: 1.

A day-ahead energy market is considered as a typical example to verify the effectiveness of the proposed profit sharing mechanism.

3.2 Electricity Spot Market Model

2. 3. 4.

55

The cost function of each generator is a positive convex function. Direct current optimal power flow (DCOPF) is adopted without considering network loss and reactive power. Only generation side bidding is considered and settled by the profit sharing mechanism, and demand side bidding is ignored with LMP settlement.

Considering a power system with N nodes and K lines, the market clearing model is formulated as follows: min f (X, cˆ ) = X

N    cˆ i PiG

(3.1)

i=1

s.t. N   G  Pi − PiD = 0 : λE

(3.2)

i=1

−

N 

  L Fk−i PiG − PiD ≥ −Pk,max : μ+ k , ∀k

(3.3)

i=1 N 

  L Fk−i PiG − PiD ≥ −Pk,max : μ− k , ∀k

(3.4)

i=1 G G ≤ PiG ≤ Pi,max , ∀i Pi,min

(3.5)

In this model, the decision variable is X, representing the active power of all the generators PiG . The objective function f (X, c) is to minimize the total procurement from all the generators, and the bidding vector is denoted by c. Constraint (3.2) represents load balance. PiD is the load at node i. The Lagrangian multiplier of this constraint is denoted by λE . Constraints (3.3) and (3.4) represent the upper and lower limits of line power flow, and Fk−i is the generation shift distribution factor (GSDF) L is the transmission capacity of line k. The Lagrangian of node i to line k. Pk,max − multipliers of line power flow constraints are denoted by μ+ k ≥ 0 and μk ≥ 0. Constraint (3.5) is the upper and lower limits of generator i’s output. The optimal strategy of the generators can be obtained by optimizing the model above. 



3.2.2 Settlement Mechanism Based on the day-ahead market model, this section will revisit marginal pricing and profit sharing mechanism respectively.

56

3 Sharing Economy in Electricity Spot Markets

3.2.2.1

Marginal Pricing Mechanism

LMP represents the incremental system operation cost caused by a marginal load increase at a node, as shown below: = λLMP i

∂L , ∀i ∂PiD

(3.6)

where L is the Lagrangian function of the day-ahead market clearing model: N N 

   G  G D E L= −λ Pi − Pi cˆ i Pi i=1 i=1

K N  G    D L − − P + P μ+ F − P k−i i i k,max k k=1 Ni=1

K    −  L − μk Fk−i PiG − PiD + Pk,max k=1

(3.7)

i=1

Thus, the marginal price of node i is: λLMP = λE − i

K 

  − Fk−i μ+ k − μk , ∀i

(3.8)

k=1

The payment to a generator at node i is denoted by πiLMP : PiG , ∀i πiLMP = λLMP i

(3.9)

In a perfectly competitive electricity market, each generator is a price taker, i.e., the generators cannot expect the impact of their bidding upon market clearing and LMP. Therefore, the generators will bid their true marginal costs, and social welfare can be maximized. However, it is hardly possible to reach a perfectly competitive market in practice. The generators, especially the marginal one, has incentive to speculate on the bid of unsuccessful units and strategically submit profitable bids. As a result, the system marginal price could be significantly raised in some cases, leading to a severe deviation of social welfare allocation.

3.2.2.2

Profit Sharing Mechanism

Based on VCG theory, the payment to a generator is defined as the substitution costs of other generators, i.e., the change in the total costs of other generators before and after generator i participates:        πiVCG = f−i X∗−i , cˆ −i − f X∗ , cˆ − cˆ i X∗i , ∀i

(3.10)

3.2 Electricity Spot Market Model Fig. 3.1 Structure diagram of 2 nodes and 3 units

57

Unit 1

Unit 2 Line

Unit 3

Load

Bus 1

Bus 2

where f−i (·) represents the system cost function when generator i does not participate ∗ is the optimal dispatch strategy of other N − 1 generators in the market, and X−i ∗ , c−i ) represents the when generator i is removed. Therefore, the first term f−i (X−i total cost of other N − 1 generators after removing generator i, and the second   term f X∗ , c − ci (Xi∗ ) represents the total cost of other N-1 generators with the participation of generator i. According to the aforementioned proof, the profit sharing mechanism naturally satisfies dominant-strategy incentive compatibility (DSIC) [12], namely: 





        ≥ πiVCG cˆ − ci X∗i , ∀i πiVCG cˆ −i , ci − ci X∗∗ i

(3.11)

The net income of any generator when truthfully bidding is not less than that when strategically bidding, regardless of whether other N-1 generators are truthful or not. However, according to Hurwicz Impossibility Principle [13], no mechanism can simultaneously fulfill DSIC, individual rationality (IR), social welfare maximization (SWM) and budget balance (BB). In this instance, the proposed VCG-based profit sharing mechanism cannot satisfy BB, indicating that the charges that ISO levies from load may not fully cover the payments to generators. This will lead to system revenue inadequacy. Take the following toy case as an example. The true costs of unit 1, 2 and 3 are c, 2c and 3c, respectively. The load at bus 1 is denoted by D. The output ranges of the three generators are all [0, 2D]. The line capacity between bus 1 and 2 is 0.5D (Fig. 3.1). When the three generators bid their true costs, the marginal price at bus 1 will be c. The charges from load is cD, but the payments to the three generators are 2.5cD, 0 and 0, respectively. A 1.5cD revenue inadequacy will be generated.

3.2.2.3

Comparison of Market Mechanisms

This section compares the implementation process of the marginal pricing and profit sharing mechanisms. Under the MP mechanism, the market clearing price is directly affected by the bid of the marginal generator, while the profit of each generator is determined by the product of LMP and the winning quantity. Therefore, the marginal generator always has incentive to speculate on the bid of next unsuccessful unit and make strategic bids for individual profit maximizing.

58

3 Sharing Economy in Electricity Spot Markets Marginal pricing

Profit sharing

Units submit bids to the market operator.

Units submit bids to the market operator.

The operator clears the market and optimally dispatches the units.

LMP and reserve price are calculated. Each unit gets MP-based payment.

The operator clears the market and optimally dispatches the units. The VCG-based payment to each unit is calculated by removing it from the market. The operator allocates the revenue inadequacy.

Fig. 3.2 Comparison of market organization and implementation process under the MP and profit sharing mechanisms

However, under the profit sharing mechanism, the profit of a generator is determined by the bids of other generators. If a generator bids a higher price than its margin, the winning quantity of market shares will decline, and the VCG-based payment to this generator will be reduced. Therefore, in contrast to MP, the VCGbased profit sharing mechanism develops an irrelative or even negative map between the generators’ bidding and their profits. On the other hand, a generator’s overbidding cannot increase individual profits but push up the payments to other N − 1 generators. Once again, the VCG-based profit sharing mechanism can suppress the strategic behaviors of market participants. Figure 3.2 compares the implementation process of the two market mechanisms. The profit sharing mechanism does not change the traditional electricity market framework based on a series of processes from market bidding, clearing to settlement. The only difference is the determination of settlement prices, in which VCG is an extension of LMP rather than an overturn. In view of the budget imbalance issue, we then propose a revenue inadequacy allocation strategy in next section, enabling the VCG-based profit sharing mechanism to work in realistic electricity markets.

3.3 Revenue Inadequacy Allocation This section firstly analyzes the reasons of the budget imbalance, theoretically deriving that the VCG-based price is always higher than LMP. Then a revenue inadequacy allocation strategy is developed to reach a post budget balance.

3.3 Revenue Inadequacy Allocation

59

3.3.1 Theoretical Analysis on Budget Imbalance Here we show that under the arbitrary bidding of generators, the VCG-based price is always higher than LMP.

3.3.1.1

Cost Function of N-1 Generators

When the output of generator i is PiG , we define the cost function of other N − 1 generators as follows: N

      hi PiG , cˆ −i = min cˆ j PjG , s.t. X−i , PiG ∈ χ X−i

(3.12)

j=1,j=i

This function indicates the minimum costs of other N-1 generators on the condition that power system constraints χ are satisfied and the output of generator i is fixed at PiG . Here we discover that hi (PiG , c−i ) is a monotonically decreasing convex function with respect to PiG . The domain of the optimization variable PiG G G is {0} ∪ {x ∈ R+ |Pi,min ≤ x ≤ Pi,max }. With the increase of PiG in its domain, the surplus load balanced by other N-1 generators decreases, thus leading to a reduction in the value of hi (PiG , c−i ). 





Additionally, the function hi (PiG , c−i ) aims at minimizing the total costs of other N − 1 generators. With the increase of PiG , the expensive generators have priority to be substituted. Therefore, the reduction rate of hi (PiG , c−i ) gets low with an increasing PiG , i.e., the second-order derivative of hi (PiG , c−i ) is greater than 0. The function hi (PiG , c−i ) is convex. A schematic diagram of the function hi (PiG , c−i ) is as follows (Fig. 3.3). 







Fig. 3.3 Schematic diagram of the cost function of N − 1 generators

O

Capacity of unit 1

……

Capacity of unit M

60

3 Sharing Economy in Electricity Spot Markets

A theoretical proof is derived as follows: G G ≤ x1 < x2 ≤ Pi,max , the optimal solutions are x1∗ and x2∗ . The For any Pi,min constraint set χ includes a set of linear constraints, so (x1∗ + x2∗ )/2 is also a feasible solution. Since the model (3.12) is a convex programming, the following inequality can be obtained:  ∗     x1 + x2∗ 1  , cˆ −i ≤ hi x1∗ , cˆ −i + hi x2∗ , cˆ −i (3.13) hi 2 2 

Therefore, the function hi (PiG , c−i ) is a convex function. 3.3.1.2

Comparison Between MP and VCG

Based on the cost function of N − 1 generators, the day-ahead market clearing model (3.1)–(3.5) can be transformed into     G G ≤ PiG ≤ Pi,max min hi PiG , cˆ −i + cˆ i PiG , s.t. Pi,min PiG

(3.14)

According to the first-order optimality condition (Karush–Kuhn–Tucker, KKT), 

  dhi d cˆ i UB LB  + + w − w i i  G G∗ = 0 G G dPi dPi P =P i

(3.15)

i

where wiUB and wiLB are the Lagrangian multipliers of the upper and lower limits of of generator’s output, respectively. Therefore, the locational marginal price λLMP i node i equals the generation marginal cost and the associated Lagrangian multipliers of generation output, i.e., = λLMP i

  d cˆ i  dhi  UB LB + w − w = − i i dPiG PG =PG∗ dPiG PG =PG∗ i

i

i

(3.16)

i

Similarly, the payment to a generator via the profit sharing mechanism can be expressed as:     πiVCG = hi 0, cˆ −i − hi PiG∗ , cˆ −i

(3.17)

can be expressed as: Therefore, the VCG-based price λVCG i λVCG i

    hi 0, cˆ −i − hi PiG∗ , cˆ −i πiVCG = G∗ = Pi PiG∗

(3.18)

3.3 Revenue Inadequacy Allocation

61

Fig. 3.4 Schematic diagram of the comparison between LMP and VCG

O

  Since hi PiG , c−i is a monotonically decreasing convex function, the following inequality can be obtained: 

    hi 0, cˆ −i − hi PiG∗ , cˆ −i PiG∗

≥−

 dhi  dPiG PG =PG∗ i

(3.19)

i



“=” holds if and only if hi (PiG , c−i ) is linear. A schematic diagram of the comparison between LMP and the VCG-based represents the absolute price is shown in the Fig. 3.4. One can observe that λLMP i represents value of the tangent slope of hi (PiG , c−i ) at PiG∗ , and λVCG i  the absolute value of the slope of the connection between the two points (0, h i 0, c−i ) and   LMP ≥ λ . (PiG∗ , hi PiG∗ , c−i ). Therefore, it is clear that λVCG i i 





3.3.1.3

Budget Imbalance

Under the MP mechanism, LMP can always fulfill budget balance, i.e., N 

 D  Pi − PiG ≥ 0 λLMP i

(3.20)

i=1

In the formula, the first term λLMP PiD represents the charges from load, and the i LMP G second term λi Pi represents the payments to generators. Therefore, the charges from load are not less than the payments to generators, and “=” holds if and only if power networks are not congested. However, if we apply the VCG-based profit sharing mechanism to pay for generators, budget balance may not be guaranteed since the VCG-based price is always higher than LMP. In other words, we sacrifice BB to pursue DSIC so as to elicit the truthfulness of generators. The revenue inadequacy is expressed as follows:

62

3 Sharing Economy in Electricity Spot Markets

Price

Price up

Strategic bid of the marginal unit

Settlement for Price Settlement the marginal unit for N-1 units Supply curve

Supply curve

Demand curve

Demand curve

Quantity

Quantity

Incremental surplus a

Information rent

Producer surplus under LMP

b

Producer surplus under VCG

Fig. 3.5 A schematic diagram of the budget imbalance caused by the profit sharing mechanism

π = π

VCG

−π

LMP

N   VCG  πi = − πiLMP

(3.21)

i=1

which can be interpreted as the additional payment to generators to suppress the strategic behaviors in contrast to LMP. In economics, such revenue inadequacy is called “information rent”. The information rent is essentially paid to generators for truthfulness in order to avoid a great loss of market efficiency caused by strategically bidding and market manipulation. In our case studies, we will show that information rent actually accounts for a small fraction of the total payments. The budge imbalance caused by the profit sharing mechanism is illustrated as follows (Fig. 3.5). Under the MP mechanism, the market clearing price is directly related to the bid of the marginal unit. The strategic bidding of the marginal unit can lead to a significant increase in the producer surplus, thus deviating from the optimal allocation of social welfare. Under the profit sharing mechanism, the marginal unit has no incentive to strategically bid since its settlement price is the bid of the next unsuccessfully unit. In addition, the other N − 1 units are settled by the bid of the marginal unit. In contrast to a perfectly competitive market, information rent will be generated to compensate the marginal generator.

3.3.2 Revenue Inadequacy Allocation Strategy Information rent is an amount of extra payments to the marginal generator in order to prevent its strategic bids. In practice, the strategic behaviors of generators can have a

3.3 Revenue Inadequacy Allocation

63

great impact on consumers. As aforementioned, if the marginal unit makes strategic bids, producer surplus will be significantly increase, indicating that consumers have to suffer losses. Therefore, consumers essentially have incentive to pay information rents to elict the truthfulness of generators, which will avoid a greater losses caused by strategic bids. On top of LMP, the price caused by information rent λIR can then be allocated to load, expressed as follows: π D λIR = N D i=1 Pi

(3.22)

 where Ni=1 PiD indicates the total load at a time slot. Therefore, the adjusted LMP can be formulated below: = λLMP + λIR , ∀i λALMP i i

(3.23)

The basic idea is to allocate information rent to load. In practice, we can determine the responsibility according to voltage class, industry category, load elasticity and time-of-use. For example, we consider two revenue inadequacy allocation strategies according to time-of-use. (1)

Time-independent allocation strategy

In this case, we allocate information rent to hourly load regardless of time-of-use. The information rent price at time slot t λIR t is: πt λIR , ∀t t = N D i=1 Pi,t

(3.24)

D where πt and Pi,t are the information rent and load at time slot t.

(2)

Critical-peak allocation strategy

To further mitigate peak-hour load, information rent can be aggregated to be allocated to critical peak hours for demand response implementation. Here we allocate information rent to peak hours only.  λIR t

=

0, t ∈ OP P D , t ∈  P P

 π i,t∈

(3.25)

i,t

where π is a daily aggregation of information rent, and OP and P are the sets of off-peak and peak hours. A schematic diagram is illustrated as follows (Fig. 3.6). Importantly, the revenue inadequacy allocation strategy will not violate DSIC since the payment function to generators remains unchanged.

64

3 Sharing Economy in Electricity Spot Markets

Fig. 3.6 A schematic diagram of allocating information rent for critical peak price implementation

Information rent allocation Price

Critical peak price

Peak price

Hour 0

8

18

23

3.4 Solution Algorithm In practice, another bottleneck for implementing the VCG-based profit sharing mechanism lies in its computation. To measure the value of a generator, it is necessary to reoptimize the market clearing model N + 1 times by removing one unit for each time, where N denotes the number of generators. In contrast to LMP that is cleared for one time, the proposed profit sharing mechanism is extremely time consuming with a large number of candidate generators. Removing one unit from the market is equivalent to setting its output as 0, and the extra load should be optimally allocated to other N − 1 generators. Therefore, we propose a solution algorithm that adapts to the cases regarding whether transmission networks are congested or not.

3.4.1 Non-congested Case When power systems are not congested, the market clearing model is only subject to load balance constraints. The removal of a generator requires that other N − 1 units balance the extra load at an equal increment coal consumption rate. The output increase vector of the N − 1 generators is denoted by [P1 , . . . , Pi−1 , Pi+1 , . . . , PN ]T . The principle of efficiently achieving the optimum is explained as follows: (1) (2)

The generator with lower marginal cost is given priority to increase its output until reaching the maximum. If the generators with residual capacity (not reaching its installed capacity) have equal increment coal consumption rates, the optimum will be achieved.

Therefore, the steps of the solution algorithm in the non-congested case are elaborated below: Step 1: Step 2:

Remove generator i and its optimal output Pi∗ from the market. Select the generators with residual capacity from the remaining ones. UP −i denotes this generator set.

3.4 Solution Algorithm

Step 3: Step 4:

If there is only one unit in UP −i , this unit will balance the remaining load, and the algorithm terminates. Otherwise, go to Step 4. According to the principle of equal increment coal consumption rate, solve the following equation set: ⎧ ⎨ ⎩

Step 5:

65



dcj (Pj )  dPj P =P ∗ +P j

j

j

= 



dcj (Pj )  , dPj P =P ∗ +P j

j∈UP −i

j

Pj = Pi∗

j

∀j, j ∈ UP −i

(3.26)

Remove the generators that reach the installed capacity from UP −i , and set these generators directly to their maximum output. Go to Step 2.

Therefore, when the power system is not congested, we can apply the aforementioned solution algorithm, which does not need to reoptimize the market clearing model. The computation efficiency for the VCG-based profit sharing can be greatly improved.

3.4.2 Congested Case When power systems are congested, we cannot rely on the principle of equal increment coal consumption rate due to network congestion. It should be noted that removing one unit from the market is equivalent to setting its output as 0. Therefore, we can use sensitivity analysis method to add a new constraint to the original market clearing model. The market clearing model can be written its compact form as follows: min cT X X

s.t. AX = b

(3.27)

X≥0 where vector c is generators’ bids, X is generators’ output, and AX = b includes load balance, line flow, unit output, etc. Note that the inequality constraints can be rewritten as equality by introducing positive slack variables. Removing generator i from the market is equivalent to adding constraint Xi = 0 to the above model. The general expression of adding a new constraint is: pm+1 X ≤ bm+1

(3.28)

where pm+1 = [0, . . . , 1, . . . , 0, 1], indicating that the position of generator i and the newly added auxiliary variable is 1, and the others are 0. If the optimal solution of

66

3 Sharing Economy in Electricity Spot Markets

the original problem satisfies the newly added constraints, the solution will also be optimal for the modified model. In other words, if a generator’s optimal output is 0, indicating that it is unsuccessful in the market bidding, there will be no influence after removing it from the market, and the optimal clearing results will not be changed. If the optimal solution of the original problem does not satisfy the newly added constraints, we can add these constraints to the optimal table obtained by using simplex method. The inequality constraint is firstly transformed by introducing slack variables XN +1 : pm+1 X + XN +1 = bm+1 , XN +1 ≥ 0

(3.29)

Let the optimal basis of the original problem be B, and the optimal solution is X∗ :

XB X = XN ∗





B−1 b = 0

(3.30)

where XB and XN are the basic and the non-basic variables, respectively. Note that the value of the non-basic variable is 0 when a linear programming achieves optimum. After adding a new constraint, the new basis B , the inverse of the new basis (B )−1 and the right-hand vector b are expressed as follows: B =







−1 0 B 0  −1 B b , B , b = = −pBm+1 B−1 1 pBm+1 1 bm+1

(3.31)

Note that pBm+1 is the coefficient vector of the new constraint corresponding to the basic variables of the original problem. Since the new basic variable is XN +1 , the basic solution of the modified model after adding new constraints is:

XB XN +1



−1



−1  −1 0 b B b B = B = (3.32) b = −pBm+1 B−1 1 bm+1 bm+1 − pBm+1 B−1 b

Therefore, if bm+1 − pBm+1 B−1 b ≥ 0, the current dual feasible basic solution will be the optimal solution of the new problem. Otherwise, the model needs to be solved by dual simplex method until the optimal table is acquired. It is worth noting that, the aforementioned solution algorithm still needs optimization process after removing one generator. However, this algorithm is a warm-start method since we derive the basis with newly added constraints for simplex, and there is no need to initiate a reoptimization from the very beginning. By this means, the simplex method can continue the optimization on the premise of our derived transitional basis, thereby greatly saving computation time.

3.5 Case Studies

67

3.5 Case Studies 3.5.1 IEEE 30-Bus System 3.5.1.1

Data Description

The structure of IEEE 30-bus system is shown below. In IEEE 30-bus system, there are 6 generators, with minimum output 0. The cost functions of the generators are quadratic and convex. The detailed parameters of the generators are shown in the Table 3.1. Here we study a day-ahead market clearing. The daily average load of PJM electricity markets from January 1 to December 31, 2017 is selected as the 24-h load profile at each bus, as shown in the Figs. 3.7 and 3.8. Table 3.1 Parameters of the generators in IEEE 30-bus system No

Quadratic cost coefficient ($/MW2 )

Linear cost coefficient ($/MW)

Capacity (MW)

1

0.0200

2.00

80

2

0.0175

1.75

80

3

0.0625

1.00

50

4

0.0083

3.25

55

5

0.0250

3.00

30

6

0.0250

3.00

40

Fig. 3.7 IEEE 30-bus system

68

3 Sharing Economy in Electricity Spot Markets 5

1.5

x 10

1.4 1.3

Load (MWh)

1.2 1.1 1 0.9 0.8 0.7 0.6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour (h) Fig. 3.8 PJM electricity market daily load in 2017

3.5.1.2

Verification of Mechanism Properties

Here we verify that the profit sharing mechanism fulfills incentive compatibility, individual rationality and social welfare maximization. To verify DSIC, it is assumed that the generators declare the cost coefficients within a certain range. Figure 3.9 shows the net profit of each generator when bidding different coefficients. For the 6 generators, only when the bid over the true cost equals 1.0, i.e., when the generators truthfully bid, maximum net profits can be obtained. Therefore, the proposed profit sharing mechanism satisfies incentive compatibility, while eliciting truthfulness of generators. 0.8 G1 G2 G3 G4 G5 G6 100%

0.4 0.0 -0.4 -0.8 -1.2 0.0

0.4

0.8

1.2

Bid/True cost (a) Quadratic

1.6

Net profit ratio

Net profit ratio

0.8

0.4 0.0 -0.4 -0.8 -1.2 -1.6

2.0

-2.0 0.0

0.4

0.8

G1 G2 G3 G4 G5 G6 100% 1.2 1.6

2.0

Bid/True cost (b) Linear

Fig. 3.9 Net profit of the generators when strategically bidding in the IEEE 30-bus system

3.5 Case Studies

69

Table 3.2 Generation and profit results of 6 generators

No

Daily power generation (MWh)

VCG-based payment ($)

Net profit ($)

1

996.88

3809.19

983.85 1527.28

2

1310.72

5077.66

3

511.00

1908.70

716.60

4

592.03

2260.96

206.81

5

317.50

1183.25

122.99

6

317.50

1183.25

122.99

To verify the individual rationality of the proposed mechanism, the Table 3.2 shows the net profits of the 6 generators. When truthfully bidding, the net profit of each generator is greater than 0, ranging from $122.99 to $1527.28. Therefore, the profit sharing mechanism ensures that the generators are willing to participate in the market. The output profiles and the relationship between output and payment are shown in the Figs. 3.10 and 3.11. It can be seen that there is a significant positive correlation between the generators’ output and the associated payment. A larger amount of market share wined by a generator indicates a higher substitution cost from other generators. 200 180

Generation (MWh)

160 140

G6 G5 G4 G3 G2 G1

120 100 80 60 40 20 0

5

10

15

20

Hour

Fig. 3.10 Output profiles of the generators when truthfully bidding

70

3 Sharing Economy in Electricity Spot Markets

5000

Payment ($)

4000

3000

2000

1000

400

600

800

1000

1200

1400

Generation (MWh)

Fig. 3.11 Relationship between the generators’ output and the associated payment

3.5.1.3

Revenue Inadequacy Allocation

Figure 3.12 compares the MP and profit sharing mechanisms to pay for the generators. On can observe that the VCG-based payment is always higher than the MP-based one when the generators truthfully bid. In this case study, the ratio of incremental payment ranges from 1.30 to 5.63%. Therefore, the market operator has to pay slightly more for generation by the profit sharing mechanism than MP, accounting for 3.81% of the total payments. It should be noted that such a small fraction payment is acceptable in practice. 1.10

18 Marginal pricing Profit sharing

15 Income / cost ( 103$)

Ratio of payment

1.05

1.00

12 G6 G5 G4 G3 G2 G1

9 6

0.95

3 0.90

1

2

3

4

Unit No.

5

6

0 Generator

Load

Fig. 3.12 Comparisons between the marginal pricing and profit sharing mechanisms

3.5 Case Studies

71

Time-independent allocaƟon Marginal price

4.0 3.9

4.0 3.9 3.8

LocaƟonal price ($/MWh)

LocaƟonal price ($/MWh)

3.8

CriƟcal-peak allocaƟon Marginal price

3.7 3.6 3.5 3.4 3.3

3.7 3.6 3.5 3.4 3.3

3.2

3.2

3.1

3.1

3.0

5

10

15

3.0

20

5

10

15

20

Hour

Hour

Fig. 3.13 Pricing results under two allocation strategies in the IEEE 30-bus system

To solve the problem of budget imbalance caused by the profit sharing mechanism, two strategies are proposed here: time-independent and critical-peak allocation. For time-independent allocation, the daily total inadequate revenue is $565.65, and the daily total load is 4045.63 MWh. Thus, the information rent price is $0.14/MWh. For critical-peak allocation, the time period from 9:00 to 22:00 is selected as the peak hour, and the total peak-hour load is 2541.23 MWh. Thus, the information rent price is $ 0.22/MWh. The pricing results are shown in Fig. 3.13. Under the time-independent allocation strategy, the information rent prices only account for 3.69–4.02% of locational prices. Even if the information rents are allocated to peak hours, the information rent prices will account for 5.81–5.99% of locational prices, which is a reasonable proportion for implementing demand response. By this means, the budge imbalance issue can be well addressed.

3.5.1.4

Computational Efficiency Analysis

Here we verify the computation efficiency of the proposed solution algorithm in non-congested and congested cases (Table 3.3). As one can observe, in the non-congested case, the fast solution method can improve the computation efficiency by 97.97% in contrast to re-optimization. In the congested case, the proposed algorithm enables a well-selected warm start for simplex method, thus improving the efficiency by 31.50% in contrast to re-optimization. Table 3.3 Computation time in IEEE 30-bus system with or without congestion Case

Re-optimization (s)

Fast solution algorithm (s)

Improvement (%)

Non-congested

1.48

0.03

97.97

Congested

2.00

1.37

31.50

72

3 Sharing Economy in Electricity Spot Markets 45

25

40 35

Number of units

Number of nodes

20

15

10

30 25 20 15 10

5

5 0

4000

3000

2000

1000

0

0 100

6000

5000

500

400

300

200

Load (MWh)

700

600

800

Installed capacity (MW)

Fig. 3.14 Load level and installed capacity in IEEE 118-bus system

3.5.2 IEEE 118-Bus System 3.5.2.1

Data Description

In IEEE 118-bus system, there are 54 generators and 186 lines (Fig. 3.14). Load profiles of each bus is also selected as the average daily load of the PJM electricity market in 2017. The minimum power output of the generators is set to 0, and the cost functions are all quadratic convex.

3.5.2.2

Verification of Mechanism Properties

1.0

1.0

0.8

0.8

Net profit ratio

Net profit ratio

To verify the incentive compatibility, it is assumed that the generators declare the cost coefficient within a certain range. Figure 3.15 shows the net profit of each online generator when strategically bidding.

0.6

0.4

0.4

0.2

0.2

0.0

0.6

0.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Bid/True cost

1.4

1.6

1.8

2.0

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

Bid/True cost

Fig. 3.15 Net profit of the generators when strategically bidding in the IEEE 118-bus system

2.0

3.5 Case Studies

73 500 450

Payment Net profit

400

Profit (

103$)

350 300 250 200 150 100 50 0

0

2

4

6

8

10

12

14

16

18

20

Unit No.

Fig. 3.16 System payment and net profit when truthfully bidding

As one can observe, the generators can obtain maximum net profits only when the generators truthfully bid. In this instance, the generators have no incentive to strategically bid. Truthfully bidding is a generator’s dominant strategy, which fulfills the dominant-strategy incentive compatibility. The payment to the generators and their net profits are shown in the Fig. 3.16. Note that only 19 generators are online in this case study, and the others are offline. The payments to the online generators vary from $3104.70 to $48,520.27, and their net profits range from $723.82 to $12,390.47. The net profits are greater than 0, which fulfills the individual rationality.

3.5.2.3

Revenue Inadequacy Allocation

Fig. 3.17 compares the ratio of the payments under the MP and profit sharing mechanisms (Table 3.4). Under the profit sharing mechanism, we discover that the VCG-based payment to every generator is always higher than the MP-based payment. The ratio of incremental payment ranges from 0.02 to 3%. Therefore, the total payments to the generators could increase by 2.01%. The revenue inadequacy allocation results are shown as follows (Fig. 3.18). According to the time-independent allocation strategy, the total daily load is 90705.95 MWh, and thus the information rent price is $0.75/MWh, ranging from 1.91 to 2.20% of the locational prices. For critical-peak allocation, the total peak load is 56976.11 MWh, and the information rent price is $1.20/MWh, ranging from 3.21 to 3.05% of the locational prices.

74

3 Sharing Economy in Electricity Spot Markets 1.04

Marginal price Profit sharing

Ratio of payment

1.03

1.02

1.01

1.00

0.99

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19

Unit No.

Fig. 3.17 Marginal price and profit sharing mechanisms pay for generating units

Table 3.4 Total payments to the generators under the two mechanisms Mechanism Total payment

$)

Marginal pricing mechanism

Profit sharing mechanism

3396.90

3465.15

Time-independent allocation Marginal price

40

Locational price ($/MWh)

Locational price ($/MWh)

40

(×103

38

36

34

32

5

10

15

20

Critical-peak allocation Marginal price

38

36

34

32

Hour

5

10

15

20

Hour

Fig. 3.18 Pricing results under two allocation strategies in the IEEE 118-bus system

3.5.2.4

Computational Efficiency Analysis

Table 3.5 compares the computation time in the congested and non-congested cases. When power networks are not congested, the fast solution algorithm based on equal increment coal consumption rate can significantly improve the computation

3.5 Case Studies

75

Table 3.5 Computation time in IEEE 118-bus system with or without congestion Case

Re-optimization (s)

Fast solution algorithm (s)

Improvement (%)

Non-congested

162.54

0.10

99.94

Congested

212.83

153.60

27.83

efficiency by 99.94%, reducing computation time from 162.54 to 0.10 s. However, since the sensitivity analysis-based algorithm for the congested case actually requires a warm-start re-optimization. Thus the computation time can be only reduced by 27.83%.

3.5.3 Polish 2383-Bus System This section further demonstrates the efficiency of the proposed mechanism and method in a even larger case. In the Polish’s 2383-bus system, there are 327 generators and 2896 lines. The load profiles of each bus are also the average daily load of the PJM electricity market in 2017. Each day is divided into 6 periods. Figure 3.19 shows the data of load and installed capacity. When truthfully bidding, the market shares of the generators are shown below, in which 312 generators are online (Fig. 3.20). 50% of the generators win the market shares for over 36 MWh. Table 3.6 compares the computation time in the congested and non-congested cases. When power networks are not congested, the fast solution algorithm can significantly improve the computation efficiency by 99.98%, reducing computation time from 940.92 to 0.20 s. However, since the sensitivity analysis-based algorithm for the congested case actually requires a warm-start re-optimization. Thus the computation time can be only reduced by 10.34%. Here we discover that a large fraction of the computation time is consumed during the preprocess of CPLEX solver, which should be further addressed in our future research. 800

250

200

600

Number of units

Number of nodes

700

500 400 300 200

150

100

50

100 0 -2000

0

2000

4000

Load (MWh)

6000

8000

0

0

500

1000

1500

2000

Installed capacity (MW)

Fig. 3.19 Load level and installed capacity in Poland 2383-bus system

2500

3000

76

3 Sharing Economy in Electricity Spot Markets 200

Number of units

150

100

50

0

0

2000

4000

6000

8000

10000

Market shares of generators (MWh)

Fig. 3.20 Market shares of the generators in Polish 2383-bus system

Table 3.6 Computation time in Polish 2383-bus system with or without congestion Case Non-congested Congested

Re-optimization (s)

Fast solution algorithm (s)

Improvement (%)

940.92

0.20

99.98

1034.10

927.16

10.34

3.6 Conclusion This chapter proposes the sharing economy-based mechanism design in electricity spot markets. Since the profit sharing mechanism defines the value of a generator as the substitution cost in place of other generators, it will be a dominant strategy to truthfully bid. Thus, the proposed profit sharing mechanism can achieve dominantstrategy incentive compatibility. By comparing marginal pricing with the profit sharing mechanism, we provide a theoretical proof that the VCG-based price is always higher than MP, which induces a higher payment to generators and leads to budget imbalance. Therefore, we discover the essence of revenue inadequacy, which can be interpreted as information rent. In this chapter, a revenue inadequacy allocation strategy is developed to distribute the information rent among electric load. Two methods, i.e., time-independent and critical-peak allocation, are investigated for demand response implementation. In addition, the computation efficiency of the profit sharing mechanism needs to be improved with the increase of the number of generators. Therefore, an efficient solution algorithm is proposed in this chapter in terms of congested and non-congested cases. When power networks are not congested, we apply an equal increment coal assumption rate-based method, in which the other N − 1 generators balance the quota of the removed one. In the congested case, a sensitivity analysis-based method

3.6 Conclusion

77

is first developed, deriving a warm-start basis without re-optimization from the very beginning. Case studies based on IEEE 30-bus, IEEE 118-bus and Polish 2383-bus systems demonstrate that the properties of the profit sharing mechanism can be verified. Additionally, the information rent price only accounts for a small fraction of locational prices, thus yielding limited impacts on electric load. Besides, the proposed fast solution algorithm can greatly reduce the computation time for market clearing and settlement. Hopefully, the proposed mechanism and algorithm can be helpful for spot market construction.

References 1. Amjady, N., Keynia, F.: Electricity market price spike analysis by a hybrid data model and feature selection technique. Electric Power Syst. Res. 80(3), 318–327 (2010) 2. David, A.K., Wen, F.: Market power in electricity supply. IEEE Trans. Energy Convers. EC 21(12), 67–68 (2001) 3. Guan, X., Ho, Y. C., Pepyne, D.: Gaming and price spikes in electric electricity markets. In: Power Industry Computer Applications, 2001. PICA 2001. Innovative Computing for Power— Electric Energy Meets the Market. 22nd IEEE Power Engineering Society International Conference on IEEE (2001) 4. Helman, U., Hobbs, B. F.: Large-scale market power modeling: analysis of the U.S. eastern interconnection and regulatory applications. IEEE Trans. Power Syst. 25(3), 1434–1448 (2010) 5. Cole, R V., Gkatzelis, Goel, G. Mechanism design for fair division: allocating divisible items without payments In: Fourteenth ACM Conference on Electronic Commerce ACM (2013) 6. Tang, W., Jain, R.: Market mechanisms for buying random wind. IEEE Trans. Sustain. Energy 6(4), 1615–1623 (2015) 7. Tang, W., Jain, R.: Aggregating correlated wind power with full surplus extraction. IEEE Trans. Smart Grid 99, 1–1 (2017) 8. Xu, Y., Low, S.H. An efficient and incentive compatible mechanism for wholesale electricity markets. IEEE Trans. Smart Grid (2017) 9. Tao, W., et al.: Max-min fairness of generalized AGV mechanisms. Decis. Control IEEE (2016) 10. Contreras-Ocana, J.E., Ortega-Vazquez, M.A., Zhang, B.: Participation of an energy storage aggregator in electricity markets. IEEE Trans. Smart Grid 99, 1–1 (2017) 11. Wang, J., et al.: Tri-level expansion planning for transmission networks and distributed energy resources considering transmission cost allocation. IEEE Trans. Sustain. Energy 1–1 (2018) 12. Wu, F., et al.: Folk theorems on transmission access: proofs and counterexamples. J. Regulatory Econ. 10(1) (1996) 13. Jain, R., Varaiya, P.: Efficient market mechanisms for network resource allocation. In: Decision and Control, 2005 and 2005 European Control Conference. CDC-ECC ‘05. 44th IEEE Conference on IEEE (2006)

Chapter 4

Sharing Economy in Multi-area Electricity Markets

4.1 Introduction The global renewable and sustainable power industry has witnessed a rapid development over the past decades. By the end of 2017, the worldwide wind and photovoltaic installed capacities have reached 540 and 401 GW, respectively, undergoing a prosperous development in the future [1]. In China and many other regions across the world, the renewable energy clusters are located far from the load centers [2]. Thus, interchange trading in multi-area power systems (MAPS) has been advocated as a promising solution to facilitating the accommodation of renewable energy. Interchange trading in multi-area power systems is a sharing economy in multi-area electricity markets. Interchange trading in multi-area power systems transports idle renewable energy to load centers, and realize regional interconnection and surplus renewable energy transaction. MAPS can promote renewable energy consumption, reducing carbon emissions and coal consumption of thermal power, reducing overall cost and improving social welfare. On the other hand, when a large-scale power system failure occurs in a certain area, other areas can transport surplus energy to the failure area, which enhances the safety to a certain extent. In 2016, China launched a project to build the world first ±1100 kV high-voltage direct current (HVDC) transmission link, which is capable of transmitting 12 GW of electricity from Xinjiang in the northwest to Anhui in eastern China. Since 2014 and 2017, the PJM market has initiated coordinated transaction scheduling (CTS) with NYISO and MISO [3], respectively, to enhance the interchange predictability and price convergence. Additionally, several market operators in Europe have focused on the multi-area market coupling to integrate the increasing renewable energy. In some existing multi-area electricity markets, marginal prices have been used to make settlements. However, the marginal pricing (MP) mechanism may not guarantee incentive compatibility and is known to lose market efficiency when profitmaximizing generators make strategic bids for price manipulation. Empirical studies have been conducted in real-world electricity markets [4]. In recent years, China has issued a series of reforms to construct deregulated electricity markets, and tried to © Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_4

79

80

4 Sharing Economy in Multi-area …

adopt the MP mechanism for settlement. However, without mature market regulation at the early stage, power plants have incentive to manipulate market prices to increase individual profits [5]. In multi-area markets, the problem of price manipulation becomes even more severe because of the inter-area information asymmetry. In practice, the generators in the power-receiving area have to provide spinning reserve for the inter-area power to smooth out the fluctuations. A larger amount of inter-area power generally requires more reserve in the power-receiving area. Hence, when the generators bid on higher reserve prices, the cost-minimized economic dispatch will schedule less inter-area power to reduce the spinning reserve requirement. As a result, the generators in the power-receiving area can earn more profits by occupying higher generation share, but the optimality of MAPS operation may be violated. Realistic evidence shows that in China, the generators in load centers prevent the wind and solar from the northwest by reporting high ancillary service prices, leading to renewable energy curtailment. Thus, it is imperative to design an incentive-compatible mechanism to elicit truthful bids in multi-area markets and achieve the optimal operation of MAPS. A wide variety of studies have been proposed over the past decades to coordinate MAPS. In [6], a decentralized coordination strategy is proposed using optimality condition decomposition (OCD). The inter-area Lagrangian multipliers (LMs), interpreted as marginal prices, are used to iteratively coordinate MAPS. In [7], a centralized frame-work for multi-area markets is developed in which genera-tors bid on electricity prices. The market clearing price is determined by the MP mechanism. Reference [8] extends multi-area economic dispatch (MAED) for joint energy and reserve clearing, and marginal prices are adopted as the trading price. Among these references, the MP mechanism is commonly used in multi-area markets, which may not guarantee incentive compatibility. To elicit truthful information, a number of studies have focused on incentivecompatible mechanism design, among which the Vickey-Clarke-Groves (VCG) theory is widely adopted. In contrast to the MP mechanism, the VCG mechanism is designed to satisfy incentive compatibility by eliminating a market participant’s bids from individual profits. In [9], the VCG mechanism is applied in electricity spot markets. A comparative result is obtained that the VCG mechanism always leads to higher electricity prices than MP, yielding revenue inadequacy. In [10], the VCG mechanism is adopted to incentivize consumers to report true preferences to a utility company, thereby achieving social welfare optimality in distribution grids. In [11], an auction paradigm is designed in which wind farms bid probability distributions of generation. The VCG mechanism is applied to elicit true probability of wind farms. In [12], the VCG mechanism is implemented to handle the strategic participation of emergency demand response. In [13], a VCG-based incentive mechanism is designed for users to share energy storage. In [14], an incentive market mechanism is designed to al-locate electricity costs among a number of strategic consumers. In the existing literature, the VCG mechanism has been successfully applied for truth-telling in electricity spot markets and demand side management. However, not much is understood

4.1 Introduction

81

about the multi-area market efficiency loss induced by MP. No studies have investigated incentive-compatible mechanisms for energy and ancillary ser-vice markets in MAPS. To fill the aforementioned gap, we propose an incentive-compatible mechanism based on VCG auction for energy and reserve market clearing in MAPS. Distinguished from the MP mechanism, the VCG-based mechanism defines the payment to a unit as other generators’ incremental costs after removing it in the market clearing process. Thus, the thermal units are incentivized to truthfully bid and provide spinning reserve for integrating the high penetration of renewable energy in MAPS. Note that different market operators across the world have enforced various regulations to limit the strategic behaviors of market participants, e.g., price caps and three pivotal supplier test. This chapter aims at eliciting truthfulness in MAPS from the perspective of mechanism design.

4.2 System Model To jointly clear the energy and reserve markets, a two-stage MAED model is established in this section. Considering the stochastic nature of wind and solar power, scenario-based stochastic programming is adopted. Hybrid AC/DC links are incorporated in MAED for inter-area transmission due to an increasing application of HVDC technology in real-world power grids. The assumptions are listed as follows: (1) Load is assumed to be inelastic. (2) Renewable units are not considered to provide reserves. (3) Thermal units incorporate the opportunity costs in reserve bidding1. (4) DC links are only used for inter-area connections considering high capital costs. (5) DC optimal power flow (OPF) is adopted for day-ahead market clearing and settlement without considering power loss and reactive power, which has been widely adopted in existing literature and practice, e.g. the Electric Reliability Council of Texas (ERCOT) [15].

4.2.1 VSC-HVDC Model Firstly, a linearized DC link model is developed. Voltage source converter (VSC) technology has gained global popularity and rapid development in recent years, which enables independent control of active and reactive power. A number of VSC-HVDC projects are under construction around the world, such as a ±160 kV VSC-HVDC project in Nanao Island supported by China Southern Power Grid [16]. The schematic of a two-area power system connected by hybrid AC-DC transmission networks is shown in Fig. 4.1. The terminal voltages on the AC and DC sides of the converter are coupled by the amplitude index of pulse width modulation (PWM) Mc , i.e.,

82

4 Sharing Economy in Multi-area …

AC transmission AC grid

Transformer

bus a

AC grid

VSC-HVDC

bus k

bus i

Phase reactor

AC filter

bus m Converter

Fig. 4.1 Schematic of a two-area system connected by AC/DC lines

Mc Um = √ UiDC , Mc ∈ [0, 1], 2 2

(4.1)

where Um and UiDC are the AC and DC terminal voltages, respectively. The current on the DC transmission line IijDC is:   IijDC = Gij UiDC − UjDC .

(4.2)

Then the power injection at bus i PiDC is:    2 PiDC = −UiDC IijDC = Gij − UiDC + UiDC UjDC 2  2 2   UiDC − UjDC − UiDC + UjDC = Gij − Gij , 2 2

(4.3)

 2 where the second term Gij UiDC − UjDC /2 is related to the power loss on the DC transmission line, which is a relatively small number. Thus, power loss is ignored for simplification, and PiDC is approximately the first term. Then a linearized VSC-HVDC model is shown as follows. Note that the power on HVDC is induced by the difference between DC terminal voltages.  DC 2 ≥ 8(Um )2 , Ui

(4.4)

 2  2  , PiDC = 0.5GijDC − UiDC + UjDC

(4.5)

4.2 System Model

83 DC DC −Pij,max ≤ PiDC ≤ Pij,max ,

(4.6)

where UiDC is the voltage of DC bus i, Um = 1.0 represents the AC terminal voltage, PiDC is the power injection, GijDC is the conductance of DC transmission line and DC Pij,max is the transmission capacity. In this model, the decision variables are the 2  DC square of DC voltages Ui/j and the power injection PiDC . Bus i and j are the DC buses connected by a VSC-HVDC. Constraint (4.4) shows the relationship of terminal voltages between connected AC and DC sides. Equation (4.5) shows the relationship of power injection related to DC voltages. (4.6) shows the transmission capacity.

4.2.2 Multi-area Economic Dispatch Model A two-stage MAED model is adopted to clear the day-ahead (DA) market anticipating potential real-time (RT) adjustments. The objective is to minimize the DA generation and reserve costs and expected RT adjustment costs, shown as follows:  



min f X, cˆ = X

 E cˆ iE Pi,t

+

R qˆ iR Pi,t

A t∈T i∈A,G

+



E−RT γs cˆ iE pi,s,t

.

(4.7)

s∈S

In the objective (4.7), the vector of decision variables is denoted by X, classified into scenario-independent (here-and-now) DA variables XDA and scenario-related (wait-and-see) RT variables XRT . The DA variables include the generation of a E R RG , the reserve of a thermal unit Pi,t , the renewable power Pi,t , the thermal unit Pi,t DC AC power on an AC transmission line Pmn,t , the power on a DC transmission line Pi,t , DC the voltage of a DC bus Ui,t and the phase angle of an AC bus θm,t . The RT variables E−RT include the RT power adjustment pi,s,t and other variables with the superscript “RT”. cˆ is the vector of all thermal units’ bids, including the bids for energy cˆ iE and reserve qˆ iR . A is the area index. S , T and A,G are the sets of scenarios, time slots and thermal units in area A, respectively. γs represents the probability of scenario s. The LMs of nodal power balance and regional reserve requirement are denoted by A,R λLMP m,t and λt , respectively, representing the DA locational marginal price (LMP) at bus m and the reserve price in area A. The constraints for the multi-area economic dispatch model are listed as follows: (1)

Power balance at each bus.

i∈A,G m

E Pi,t +

i∈A,RG m

RG Pi,t −

n∈Am ∪A,N m

D A = Pm,t : λLMP m,t , ∀t, m ∈  .

AC Pmn,t +



DC Pi,t

i∈A,DC m

(4.8)

84

4 Sharing Economy in Multi-area …

The nodal load is satisfied by the generation from thermal and renewable units, A,RG and Am are the sets of the and power on the connected AC/DC lines. A,G m , m thermal units, renewable units and regional buses connected to bus m, respectively. (2)

Power limit for intra- and inter-area AC transmission lines.   AC = Bmn θm,t − θn,t , ∀t, m ∈ A , n ∈ Am ∪ A,N Pmn,t m , AC AC AC −Pmn,max ≤ Pmn,t ≤ Pmn,max , ∀t, m ∈ A , n ∈ Am ∪ A,N m .

(4.9) (4.10)

Constraint (4.9) shows the power flow on AC transmission lines related to the AC is the phase angles, which is restricted by its capacity shown in (4.10). Pmn,max capacity of AC line mn. (3)

Power limit for each renewable generator. RG FRG ≤ Pi,t , ∀t, i. 0 ≤ Pi,t

(4.11)

In (4.11), the generation of a renewable generator is restricted by the forecasted FRG . renewable power, denoted by Pi,t (4) (5)

Power and voltage limits for VSC-HVDC, which can be obtained by extending (4.4)–(4.6) to multiple time slots. Power limit for each thermal unit. E R G E R + Pi,t ≤ Pi,max , Pi,t , Pi,t ≥ 0, ∀t, i. Pi,t

(4.12)

In (4.12), the capacity for generation and reserve is restricted by the installed G . capacity of a thermal unit, denoted by Pi,max (6)

Ramp-up and -down limits for each thermal unit.  E  R E U Pi,t + Pi,t − Pi,t−1 ≤ Pi,max , ∀t, i,

(4.13)

 E  R E D Pi,t−1 + Pi,t−1 − Pi,t ≤ Pi,max , ∀t, i.

(4.14)

The constraints (4.13) and (4.14) show the ramp-up and –down limits for thermal U /D units, respectively, denoted by Pi,max . (7)

Reserve requirement in each area.

These constraints are shown in (4.20), in which the maximal inter-area power is expressed as follows: 

AC PtA,N ≥ max Pnm,t , ∀t, |m ∈ A , n ∈ A,N m

(4.15)

4.2 System Model

85



DC PtA,N ≥ max Pi,t |i ∈ A,DC , m ∈ A , ∀t. m

(4.16)

These constraints show the maximal power on AC/DC lines. (8)

Other real-time constraints.

RT constraints can be obtained by expanding (4.4)–(4.6), (4.8)–(4.11) to multiple scenarios. The RT available adjustment and ramp capacity of each thermal unit are expressed below: E−RT E R ≤ pi,s,t ≤ Pi,t , −Pi,t

(4.17)

 E   E  E−RT E−RT U Pi,t + pi,s,t , ∀s, t, i, − Pi,t−1 + pi,s,t−1 ≤ Pi,max

(4.18)

 E   E  E−RT E−RT D Pi,t−1 + pi,s,t−1 − Pi,t + pi,s,t ≤ Pi,max , ∀s, t, i.

(4.19)

In this chapter, an area’s reserve requirement is formulated as a linear sum of the reserve capacity needed by different events, i.e., inter-area power, regional load and renewable power [17]:

R Pi,t ≥ η1A PtA,N + η2A PtA,D + η3A

i∈A,G



RG Pi,t , ∀t, A,

(4.20)

i∈A,RG

where PtA,N and PtA,D represent area A’s maximal power import from neighbor areas and regional load. ηiA , i = 1, 2, 3 are the fractions of capacity reserved for inter-area power, load and renewable power, respectively. Collecting the costs reported by all thermal units, a market operator can solve the two-stage MAED model, and schedule the optimal generation and reserve. Given the bids cˆ , the optimal solution is denoted by X* (ˆc). Then the units get payments from the market operator. A market mechanism sets up a rule that maps the bids cˆ into the payments to units.

4.3 Incentive Mechanism In this section, we firstly revisit the MP mechanism and explain why thermal units have incentive to strategically bid in MAPS. Then we apply a VCG auction to elicit truthful bids in multi-area markets. The difference between MP and VCG mechanisms lies in the market settlement, i.e., the MP mechanism defines MP-based payments to units, and the VCG-based payment to a unit is calculated according its externality to the market. Additionally, a revenue inadequacy allocation sector is added in the implementation of VCG.

86

4 Sharing Economy in Multi-area …

4.3.1 Marginal-Pricing Mechanism Under the MP mechanism, given the marginal prices and optimal solution to MAED, the MP-based payment to a thermal unit πiMP is calculated as follows:     LMP E∗ R∗ λm,t Pi,t + λA,R πiMP cˆ = t Pi,t ,

(4.21)

t∈T

A,R R∗ E∗ where the term λLMP m,t Pi,t is the payment for energy, and the term λt Pi,t is the payment for reserve. Note that the market clearing prices are directly related to the bids of the marginal unit. In a competitive market, as price-taking generators do not anticipate the influence of their bids on marginal prices, all thermal units truthfully bid and the MP mechanism achieves social welfare optimality. However, in an imperfect competition environment, the marginal unit can manipulate market clearing prices by making strategic bids. Thus the marginal unit can raise the market clearing price to increase individual profits. Meanwhile, the producer surplus can be greatly improved, which deviates from Pareto efficiency of the market. Specially, the thermal units in MAPS have incentive to strategically bid on reserve prices to reduce inter-area power exchange. As (4.20) shows, an area’s reserve requirement is expressed as a function of inter-area power. Thus, a larger amount of inter-area power trading requires a higher level of reserve provided by thermal units in the power-receiving area. When the thermal units bid on higher reserve prices, the MAED will dispatch less inter-area power to achieve the minimization of the reported costs. As a result, the thermal units can improve the market share of generation and individual profits. However, the strategic behaviors of thermal units can influence the effectiveness and optimality of MAED. Therefore, an incentive-compatible mechanism is needed to elicit truthful bids and encourage thermal units to provide reserve for the accommodation of renewable energy.

4.3.2 Incentive-Compatible Mechanism The VCG mechanism is designed to eliminate market power and prevent market participants from profitably altering prices [18]. The VCG-based payment function πiVCG is designed as:     ∗      , cˆ −i − f X∗ , cˆ − cˆ i Xi∗ , πiVCG cˆ = f−i X−i

(4.22)

 ∗  where f−i X−i , cˆ −i is the total cost of MAED without the participation of unit i, ∗ ∗ and cˆ −i are the optimal dispatch X−i  ∗  and thermal units’ bids without unit i. Xi is the the and optimal dispatch of unit i, and cˆ i Xi is unit i’s cost, including   DAgeneration  reserve costs as well as the RT adjustment cost. Thus, f X∗ , cˆ − cˆ i Xi∗ represents

4.3 Incentive Mechanism

87

the costs of all units except unit i when all units participate. In this chapter, we assume that there exists a feasible dispatch after removing any unit. The VCG-based payment can be explained as the incremental costs of other generation after removing a unit, which reflects the externality the unit causes to the MAPS. Thus, the proposed mechanism can incentivize thermal units to bid truthful costs since the payment to a unit is valued by others’ bids. If a thermal unit bids on a higher price, the optimal output of this unit may decrease, and the incremental costs of others can be reduced after removing this unit. As a result, the payment to this unit will decrease. Then we have the following theorem. Theorem 1 The proposed mechanism (4.22) fulfills dominant-strategy incentive compatibility (DSIC), individual rationality (IR) and social welfare optimality (SWO), explained as follows: (1) (2) (3)

(DSIC) For each unit, the optimal strategy is to bid truthful information regardless of its belief on the costs from others. (IR) Each unit’s net profit is nonnegative, which guarantees that units have incentive to participate. (SWO) The optimal dispatch is efficient, which achieves social welfare maximization. In another word, the optimal solution to MAED minimizes the total costs of the MAPS.

See proof in [9]. Hurwicz impossibility theorem proves that there exists no mechanism that is efficient, individual rational, dominant-strategy incentive compatible and budget-balanced [19]. According to the existing literature, the VCG mechanism always results in higher per-unit prices than marginal prices given any reported supply curves. This indicates that the VCG mechanism yields higher payments to thermal units than MP. Thus, the proposed mechanism does not satisfy budget balance, leading to system revenue inadequacy. In this chapter, budget balance requires that the total payments to generators are not more than the total charges from electrical load [20]. Otherwise, the market operator has to suffer from the financial deficit. Therefore, we propose a revenue inadequacy allocation scheme to address the budget imbalance issue.

4.3.3 Revenue Inadequacy Allocation An illustrative schematic of revenue inadequacy caused by VCG is shown in Fig. 4.2. Under the MP mechanism, the marginal unit can manipulate market clearing prices by strategically bidding, which may significantly expand producer surplus. In contrast to MP, the VCG-based payment is defined as the incremental cost after removing a unit. In Fig. 4.2, suppose the marginal unit has sufficient capacity. The marginal unit will fill the imbalance after removing another unit, and the next unit out of the market will fill the imbalance after removing the marginal unit. Thus, the other units are settled by the bid of the marginal unit, while the marginal unit is settled by the bid

88

4 Sharing Economy in Multi-area … Marginal unit’s bid

Clearing price

Producer surplus

Price

Price

Supply

Supply

Demand

Demand

Quantity

Quantity

VCG mechanism

MP mechanism

Fig. 4.2 Schematic of revenue inadequacy caused by VCG

of the next unit out of the market, which prevents the marginal unit from submitting high bids. Compared with MP, the VCG mechanism requires an additional payment to the marginal unit, which can be interpreted as the financial incentive to the marginal unit for truth-telling. A plausible solution to address the revenue inadequacy is that electrical load takes responsibility for the additional payment to the marginal unit. While such additional payment induced by VCG leads to a higher charge for electrical load, the consumers can avoid greater loss caused by the strategic bidding of thermal units when settled by MP. As illustrated in Fig. 4.2 thermal units’ strategic bidding can result in even larger increase of producer surplus than the additional payment to the marginal unit under the VCG mechanism. We denote the total revenue inadequacy in MAPS as π , which is expressed as the difference between the VCG-based and MP-based payments when truthfully bidding: π =



 πiVCG (c) − πiMP (c) .

(4.23)

A i∈A,G

We propose to averagely allocate the revenue inadequacy to electrical load in MAPS, which is a straightforward and simple idea to fulfill fairness like ancillary service cost recovery or transmission cost allocation [21]. Thus, an additional price λRI is used to charge electrical load on top of marginal prices, expressed as follows: π λRI =   

D Pm,t

,

(4.24)

A t∈T m∈A

D is where the denominator represents the total electrical load in MAPS, and Pm,t the load at bus m. Note that the allocation scheme does not violate the properties of the proposed VCG-based mechanism in Theorem 1 because the VCG-based payments to thermal units are not changed. There are alternative allocation methods

4.3 Incentive Mechanism

89

for revenue inadequacy, and we make no claim as to the optimality of such an allocation scheme. The revenue inadequacy allocation is an open and challenging question. Other methods deserve an in-depth study in the future. For example, the revenue inadequacy can be allocated to the load demands during peak hours to further mitigate peak load and price spikes.

4.3.4 Toy Example We provide an illustrative example to demonstrate the potential market efficiency loss of the MP mechanism, and validate incentive compatibility of the proposed mechanism. A two-area system is shown in Fig. 4.3. The load in area 1 is 15 MWh, and the capacity of the tie-line is 10 MW. In area 1, G1 and G2 provide reserve for load and inter-area power. 1 MWh load or inter-area power requires 0.1 MW reserve capacity. An MAED model is optimized to schedule the three generators, as follows: min

3

P1E ,P1R ,P2E ,P2R ,P3E ≥0

cˆ iE PiE +

i=1

2

qˆ iR PiR ,

(4.25)

i=1

s.t. P1E + P2E + P3E = 15,

(4.26)

P1R + P2R ≥ 0.1 × P3E ,

(4.27)

PiE + PiR ≤ 20, i = 1, 2,

(4.28)

P3E ≤ 10,

(4.29)

where, cˆ iE and qˆ iR represent thermal unit i’s bid in energy markets and thermal unit i’s bid in reserve markets, respectively.

Fig. 4.3 Schematic of a two-area system

Area 1

Area 2

G1 Tie-line

G2 Load

G3

90

4 Sharing Economy in Multi-area …

Table 4.1 Parameters of three generators when truthfully and strategically bidding Unit

Capacity (MW)

Energy bid ($/MWh)

Reserve bid ($/MW)

T

S

T

S

G1

20

100

100

100

110

G2

20

120

120

120

120

G3

10

89

89

/

/

Table 4.2 Optimal solution when G1 truthfully and strategically bids Unit

Generation (MWh)

Reserve (MW)

Net profit ($)

T

S

T

S

T

S

G1

5

15

2.5

1.5

0

15

G2

0

0

0

0

0

0

G3

10

0

0

0

110

0

The parameters of the three generators are shown in Table 4.1. “T” indicates the three generators truthfully bid, and “S” indicates the reserve bid of G1 is 110 $/MW, which is slightly higher than its true cost.

4.3.4.1

MP Mechanism

Under the MP mechanism, we discover that G1 has incentive to strategically bid on reserve prices, thus influencing the optimality of MAPS. Table 4.2 compares the optimal dispatch results when G1 truthfully and strategically bids. When G1 truthfully bids, G3 is dispatched to produce 10 MWh electricity for area 1. In area 1, the load and inter-area power reach 25 MWh, and thus the reserve requirement is 2.5 MW. The LMPs at the two buses equal the energy bid of G1, i.e., 100 $/MWh. The net profit of G3 is calculated as (100–89) $/MWh × 10 MWh = $ 110. In this case, the total costs of the two-area system can be minimized, which equals $ 1640. However, when G1 raises the reserve price, the market operator will not dispatch G3. G1 can improve its net profit from 0 to $ 15, but the total operation costs will increase to $ 1650, which deviates from the cost-minimization level.

4.3.4.2

Incentive-Compatible Mechanism

By comparing a generator’s net profit under different bids, we can validate whether truth-telling is a dominant strategy. The proposed mechanism defines the payment to a generator as others’ incremental costs after removing it. For example, after removing G1, the optimal dispatch is that the generation and reserve capacity of G2 are 5 MWh and 2.50 MW, respectively, and G3 produces 10 MWh. Thus the term

4.3 Incentive Mechanism

91

Table 4.3 VCG-based payments when G1 truthfully and strategically bids Unit

Payment ($)

Net profit ($)

T

S

T

S

G1

900

1790

150

140

G2

0

0

0

0

G3

900

0

10

0

 ∗  f−1 X−1 , cˆ −1 in (4.22) equals $ 1790. Table 4.3 compares the VCG-based payments to the generators whether G1 truthfully and strategically bids. When bidding true cost, G1 can earn $ 150 net profit. However, strategically bidding can decrease the net profit of G1 to $ 140. Therefore, G1 has no incentive to strategically bid, and thus the proposed mechanism fulfills incentive compatibility.

4.3.4.3

Revenue Inadequacy Allocation

When bidding true cost, the total MP-based payment is $ 1750, and the total VCGbased payment is $ 1800. Thus, the revenue inadequacy induced by the incentive mechanism is $ 50, accounting for 2.86% of the MP-based payment. The VCG mechanism yields a slight influence on electrical load. However, if G1 makes strategic bids for generation and reserve at 110 $/MWh and 110 $/MW, respectively, the total MP-based payment will be $ 1925, which is greater than the VCG-based payment. In another word, compared with the incentive payment induced by VCG, the strategic bidding of thermal units may yield even greater charges for electrical load. Therefore, electrical load has motivation to pay incentives to the marginal unit for truth-telling.

4.4 Solution Algorithm In practice, a centralized multi-area economic dispatch model requires all regional market operators to submit detailed information to a central coordinator. However, this is sometimes impractical because such a coordinator may not exist, e.g., the Eastern Interconnection of the U.S., or a coordinator may not have the authority for MAED, e.g., State Grid Corporation of China. Therefore, we propose a decentralized algorithm based on improved OCD to preserve regional autonomy and efficiently coordinate MAPS. Note that the proposed algorithm is used for market clearing based on the day-ahead bids, which does not require market bidding during each iteration.

92

4 Sharing Economy in Multi-area … Operator 2

Information layer

Operator 3 Operator 1

Area 2

Information flow Power flow

Physical layer

G G G Area 1

Area 3

G

G

Fig. 4.4 Framework of the decentralized multi-area economic dispatch

4.4.1 Framework The framework of the decentralized MAED is shown in Fig. 4.4. In the physical layer, different areas are connected via transmission lines, and exchange power with the neighbors. In the information layer, a regional market operator is responsible for collecting bids from local market participants and scheduling local generation resources. Additionally, different market operators coordinate with adjacent areas by exchanging boundary information. To efficiently achieve such coordination, some existing literature has proposed optimal condition decomposition, which is essentially a modified version of Lagrangian relaxation. Based on the decomposition of coupling constraints, the OCD method can split MAED into regional sub-problems, and iteratively update the boundary information until convergence is achieved. In the following subsections, the decomposed regional sub-problem will be formulated, and the information exchange mechanism will be designed.

4.4.2 Decomposed Regional Sub-problem In an MAPS, Lagrangian relaxation is performed to decouple the constraints for inter-area transmission lines, shown in Fig. 4.5. The terms topped with “~” are the boundary information provided by neighbors. In the sub-problem of area A, the decomposed constraint for an AC line is formulated below:   AC = Bmn θm,t − θ˜n,t : λmn,t , ∀t, m ∈ A , n ∈ A,N (4.30) Pmn,t m , AC is the power on AC line mn, and Bmn is the susceptance. θ˜n,t is the phase where Pmn,t angle provided by the neighbor area N. For DC lines, the decomposed constraint is

4.4 Solution Algorithm

93

Area A

Area N

Fig. 4.5 Schematic of decoupling inter-area AC/DC transmission constraints

shown as follows:

DC Pi,t = GijDC

 DC 2  DC 2 − Ui,t + U˜ j,t 2

A,DC : λDC . i,t , ∀t, i ∈ m

(4.31)

DC DC where Pi,t is the power on DC line ij. U˜ j,t is the voltage provided by the neighbor DC from area N, the area N. Therefore, given the boundary information θ˜n,t and U˜ j,t DC operator in area A can schedule the inter-area power by adjusting θm,t and Ui,t . In DC addition, the neighbor area N provides the LMs λ˜ nm,t and λ˜ j,t for area A, representing the marginal costs for inter-area AC and DC power transmission, respectively. Then the sub-problem of area A is formulated as follows:

 

min f XA , cˆ A = A

XA

t∈T



  E R cˆ iE Pi,t + qˆ iR Pi,t

i∈A,G



−

m∈A ,n∈A,N m

AC λ˜ nm,t Pmn,t +

   + E C RT XA−RT , s



⎤ DC ⎦ λ˜ DC j,t Pi,t

m∈A ,i∈A,DC m

(4.32)

subject to the regional constraints in Sect. II.B, as well as Eqs. (4.30) and (4.31). DC AC and λ˜ DC In the objective (4.32), the terms λ˜ nm,t Pmn,t j,t Pi,t represent the profit/cost of A exchanging power with the neighbor area N. X is the vector of decision variables, and A,DC are and cˆ A is the vector of the thermal units’ bids in area A. A , A,N m m the sets of the buses in area A, the buses in neighbor areas connected to bus m, and the DC buses connected to bus m, respectively. The function C RT (·) represents the RT adjustment cost, including the adjustment costs of thermal units and inter-area power. XA−RT represents the RT variables in area A. The traditional OCD algorithm is elaborated in the following algorithm:

94

4 Sharing Economy in Multi-area …

Traditional optimality condition decomposition 1

Initialization. Set inter-area AC/DC power to 0, and each area optimizes the decomposed regional sub-problem. Set the iteration index k to 0

2

k=k+1

3

LM calculation. Given the current dispatch results, each area directly calculates the value of LMs associated with the inter-area transmission constraints (4.30) and (4.31)

4

Information exchange. For each area, collect the phase angles, bus voltages and LMs of inter-area links from the neighbor areas, and send local boundary information to the neighbors

5

Sub-problem updating. Based on updated boundary information, each area optimizes the decomposed regional sub-problem

6

Convergence check. Given the tolerance σ , check the convergence criteria:    AC  P [k] + P AC [k] ≤ σ, P DC [k] + P DC [k] ≤ σ, ∀m, n, i, j, t. mn,t nm,t i,t j,t

7

If the convergence criteria are satisfied, terminate the algorithm. Otherwise, go to Step 2

In practice, the determination of LMs is a key factor leading to the performance of algorithmic convergence. During each iteration of the traditional OCD method, the LMs from the neighbors are regarded as constants in the objective (4.32) of a regional sub-problem. However, such static point-wise fashion cannot reveal the elasticity of an area’s aggregated supply functions. In other words, interpreted as the marginal costs for inter-area power transmission, the LMs can greatly vary with the change in inter-area power. To identify the elasticity of the LMs and generate a more accurate step size, an improved OCD method is proposed in this chapter.

4.4.3 Improved Lagrangian Multiplier Instead of using the traditional point-wise constants, the LMs are approximated as linear functions of inter-area power by using parameter sensitivity. For simplification, a regional sub-problem is written in a compact form:   min f A XA , αA , XA   s.t. hA XA , αA = 0 : λA ,   g A X A , α A ≤ 0 : μA ,

(4.33)

where the decision variable is XA ∈ Rv×1, the parameter vector is αA ∈ Rp×1 , the equality constraint vector is hA XA , αA and the inequality constraint vector is   gA XA , αA , and λA ∈ Rh×1 and μA ∈ Rg×1 are the LMs of equality and inequality constraints, respectively. Then we can differentiate the Karush-Kuhn-Tucher (KKT)

4.4 Solution Algorithm

95

conditions at the optimal point of (4.33). The following theorem gives the solution to the sensitivity of XA , λA and μA with respect to αA . Theorem 2 The first-order derivative of XA , λA and μA with respect to αA , denoted by SA , is expressed as follows: SA =



∂XA ∂λA ∂μA ∂αA ∂αA ∂αA

T

= −J1−1 J2 ,

(4.34)

where J1 and J2 represent the Jacobian matrices of the KKT conditions with respect T  to XA , λA , μA and αA , respectively. Proof Let XA∗ , λA∗ and μA∗ denote the optimal solution to the model (4.33). Then we can differentiate the KKT conditions at the optimal point, expressed as follows: ⎡ ⎤ ⎡ ⎤ ⎤ dX dX FXX HX GX FXα ⎢ ⎥ ⎢ ⎥ ⎣ HT 0 0 HT ⎦⎢ d λ ⎥ = [J1 , J2 ]⎢ d λ ⎥ = 0, α X ⎣ dμ ⎦ ⎣ dμ ⎦ GXT 0 0 GαT dα dα ⎡

(4.35)

where FXX = ∇XX f + A

h

A λA∗ j ∇XX hj

+

j=1

FXα = ∇Xα f A +

h

g∗

A μA∗ j ∇XX gj ,

(4.36)

A μA∗ j ∇Xα gj ,

(4.37)

j=1

A λA∗ j ∇Xα hj +

j=1

g∗

j=1

 T  T  T  T HX = ∇X hA , Hα = ∇α hA , GX = ∇X gA , Gα = ∇α gA .

(4.38)

Note that g ∗ in (4.36) and (4.37) represents the number of binding inequality constraints. J1 and J2 are the Jacobian matrices of the KKT conditions with respect T  to XA , λA , μA and αA , respectively. Thus Eq. (4.34) can be obtained according to (4.35). For a regional sub-problem, we set the inter-area AC/DC power as parameters αA . During each iteration, when a regional sub-problem is optimized, the sensitivity of the LMs of (4.30) and (4.31) with respect to the inter-area power can be calculated according to Theorem 2. Then we can obtain the linear approximations of LMs:    AC  AC AC + λ˜ nm , = SAPAC Pnm − P˜ nm λnm Pnm nm

(4.39)

96

4 Sharing Economy in Multi-area …

    PjDC = SAPDC PjDC − P˜ jDC + λ˜ DC λDC j j ,

(4.40)

j

where λnm (·) and λDC j (·) are the linear approximations of LMs provided by neighA A bors. SPAC and SPDC represent the submatrices extracted from SA associated with the nm

j

constraints for AC/DC links. Instead of exchanging point-wise multipliers λ˜ nm and λ˜ DC j , the proposed LM functions are exchanged among areas. In another word, the LMs in Step 4 of the traditional OCD are replaced with Eqs. (4.39) and (4.40) during each iteration of the proposed algorithm. By identifying the elasticity of the marginal costs for inter-area power, a more accurate step size can be calculated during each iteration. Thus, compared with the traditional OCD, the proposed algorithm can greatly improve the convergence performance and efficiently coordinate MAPS.

4.5 Case Studies In Fig. 4.6, the 2-area power system consists of two IEEE 30-bus systems connected by an AC line and a VSC-HVDC. The parameters of the inter-area transmission lines are shown in Table 4.4. There are 6 thermal units located in area 1. The parameters of the thermal units are listed in Table 4.5. Area 2 has another 6 thermal units with identical parameters. In this chapter, ηiA = 0.1, i = 1, 2, 3. There are 4 renewable units in area 2, the capacities Area 1

Area 2

Fig. 4.6 Schematic of a 2-area 60-bus power system

Table 4.4 Parameters of the inter-area AC line and the VSC-HVDC Type

From

To

B/G (p.u.)

Capacity (MW)

AC

28

34

5

50

DC

30

53

200

100

4.5 Case Studies

97

Table 4.5 Parameters of the thermal units No

Bus

Capacity (MW)

Generation ($/MWh)

Reserve ($/MW)

1

1

80

2.00

3.00

2

2

80

1.75

2.63

3

13

40

3.00

4.50

4

22

50

1.00

1.50

5

23

30

3.00

4.50

6

27

55

3.25

4.88

of which are set to 60 MW. Wind farm 1 is at bus 37, and wind farm 2 is at bus 51. Solar station 1 is at bus 42, and solar station 2 is at bus 60. Wind and solar power scenarios are collected from the yearly data in PJM markets and National Renewable Energy Laboratory (NREL) [24], respectively.

4.5.1 Impacts of Strategic Bids As aforementioned, thermal units have incentive to make strategic bids when settled by the MP mechanism, which may greatly influence the optimality of MAED. Figure 4.7 shows the renewable energy accommodation and inter-area power under different reserve bids from the thermal units in area 1. The inter-area power is the total power on inter-area AC/DC links. With the increase in the reserve bids of thermal units in area 1, the inter-area power from area 2 to 1 keeps decreasing because the MAED will dispatch less inter-area power. When the bids over true costs vary from 1.0 to 5.0, the peak value of the inter-area power decreases by 34.65%, and the total amount is reduced by 43.61%. As a result, the renewable energy in area 2 cannot be effectively transmitted to area 1, leading to a higher curtailment rate. The accommodation rate drops from 80.36 to 76.47%, curtailing another 65.91 MWh renewable energy. Therefore, it is of vital importance to elicit truthful bids from generators to ensure the optimality of MAED.

4.5.2 Performance of the Proposed Mechanism To validate incentive compatibility, the net profits of the thermal units in area 1 under different levels of bids are shown in Fig. 4.8. The x-axis represents a unit’s bid over its true cost, and the y-axis shows a unit’s net profit scaled to [0,1]. As one can observe, when a unit’s bid over its true cost equals 1.0, the net profits can be maximized, indicating that truthfully bidding is a unit’s dominant strategy to get the maximal net profits. Therefore, incentive compatibility is satisfied.

4 Sharing Economy in Multi-area … 1400

Accommodation rate

Renewable energy accommodation (MWh)

98

Energy Rate

1360

1320

Inter-area power (MWh)

1280

1.0

2.5

2.0

1.5

3.5 3.0 Bid/True cost

4.0

5.0

4.5

20 15

Bid/true cost

1.0

10 5

5.0

0

5

10

15 Hour (h)

20

25

Fig. 4.7 Renewable energy accommodation and inter-area power under different reserve bids of the thermal units in area 1 of the 60-bus system Generation cost

Ratio of net profit

1

0.5

0 0

0.2

0.4

0.6

1.2 1 0.8 Bid/True cost Reserve cost

1.4

1.6

1.8

2

0

0.2

0.4

0.6

1.2 1 0.8 Bid/True cost

1.4

1.6

1.8

2

Ratio of net profit

1

0.5

0

Fig. 4.8 Net profits of the thermal units in area 1 under different levels of bids

4.5 Case Studies

99

Table 4.6 Payments and net profits of all thermal units No

1

2

3

4

5

6

MP payment

91.44

3025.08

18.62

2673.65

60.61

0

VCG payment

113.32

3404.80

19.07

4709.86

63.22

0

VCG net profit

21.89

482.64

0.45

3402.63

2.61

0

No

7

8

9

10

11

12

MP payment

0

28.68

0

690.78

0

0

VCG payment

0

30.76

0

1169.06

0

0

VCG net profit

0

2.08

0

486.89

0

0

Unit $

To validate individual rationality, the payments and net profits of thermal units under truthful bids are show in Table 4.6. From the results, all thermal units have nonnegative net profits settled by the proposed mechanism when truthfully bidding, which guarantees the rationality of thermal units’ participation. In addition, the proposed mechanism always results in higher payments than the MP mechanism. Compared with MP, the incremental payments caused by the VCG mechanism vary from $ 0.45 to $ 2036.22. The total revenue inadequacy induced by the VCG mechanism is $ 2921.23, averagely allocated to load in MAPS. The total daily load is 4166.05 MWh, and thus λRI = 0.70 $/MWh, which is relatively small and acceptable.

4.5.3 Impacts of Inter-area Transmission Capacity In real-world power grids, multiple regions are mainly connected by AC transmission lines due to the high capital cost of DC links. Thus, we evaluate the impacts of interarea DC capacity on the performance of the proposed mechanism. Figure 4.9 shows the renewable energy accommodation and inter-area power when thermal units in area 1 truthfully and strategically bid. Note that “S” indicates that the reserve bids over true costs reach 5.0, and “T” represents that all thermal units bid true costs. With the increase in DC capacity, the amounts of renewable energy accommodation can be improved in both “S” and “T” cases because more generation can be transmitted from area 2 to 1. Without the DC link (DC capacity equals 0), 1195.18 MWh and 1231.58 MWh of renewable energy can be accommodated in “S” and “T” cases, respectively. However, with a 25-MW DC link, the renewable energy accommodation can increase to 1295.68 MWh and 1361.59 MWh in the two cases. In addition, the marginal accommodation of renewable energy gets low with the increase in DC capacity. When DC capacity varies from 15 to 25 MW, the renewable energy accommodation and inter-area power nearly remain unchanged.

Renewable energy (MWh)

100

4 Sharing Economy in Multi-area … 1400

S T

1300

1200

1100

0

5

10

15

20

25

20

25

DC capacity (MW)

Inter-area power (MWh)

300

S T

200

100

0

0

5

10

15

DC capacity (MW)

Fig. 4.9 Renewable energy and inter-area power with the increase in DC transmission capacity

The potential renewable energy curtailment caused by strategic bids is positively related to the DC capacity. Compared with the case without DC, the potential renewable energy curtailment can be greatly increased by 81.07% when DC capacity reaches 25 MW. It’s worth mentioning that the potential curtailment reaches the lowest, equaling 7.93 MWh, when DC capacity is 5 MW. This is because the DC link is almost congested in both “S” and “T” cases. As shown in Fig. 4.9, the interarea power in both “S” and “T” cases can be greatly improved when DC capacity switches from 0 to 5 MW.

4.5.4 Impacts of Thermal Generation Flexibility In practice, various types of thermal units may have different levels of minimum power limit. Such diversity of thermal generation flexibility can influence the potential benefits of the proposed mechanism. Therefore, we conduct a sensitivity analysis about the minimum power limit of thermal units. Figure 4.10 shows the renewable energy accommodation and inter-area power when thermal units in area 1 truthfully and strategically bid. Note that “S” indicates that the reserve bids over true costs reach 5.0, and “T” represents that all thermal units bid true costs.

Renewable energy (MWh)

4.5 Case Studies

101

1500

T S 1000

500

Inter-area power (MWh)

0

0.1

0.2 0.3 Minimum power

0.4

0.5

0.1

0.2 0.3 Minimum power

0.4

0.5

0

-500 T S -1000

0

Fig. 4.10 Renewable energy and inter-area power with the increase in thermal units’ minimum power

The increase of thermal units’ minimum power indicates the improvement of thermal generation share and the reduction of thermal generation flexibility. Therefore, the renewable energy accommodation significantly decreases with the minimum power varying from 0 to 40% of the installed capacity. Note that when the minimum power is switched from 30 to 40%, there is a steep drop of renewable energy accommodation from 1137.77 to 479.94 MWh and from 1226.12 to 468.92 MWh in “S” and “T” cases, respectively. This is because the surplus thermal generation in area 1 forces the transmission power to flow from area 1 to 2. When the minimum power increases from 40 to 50% of the installed capacity, however, the renewable energy accommodation is slightly improved to 557.20 and 555.19 MWh in “S” and “T” cases, respectively. Compared with the minimum power equaling 40%, more power is transmitted from area 1 to 2 in the 50% case, which takes the place of thermal generation in area 2. Such substitution further leaves space for the spinning reserve provided by thermal units in area 2, thereby facilitating the penetration of renewable energy.

4.5.5 3-Area 354-Bus Power System To validate the efficiency and effectiveness of the proposed mechanism and method, a 3-area 354-bus power system is tested, which consists of three connected IEEE 118-bus systems. The parameters of the tie-lines are shown in Table 4.7.

102

4 Sharing Economy in Multi-area …

Table 4.7 Parameters of the tie-lines in the 3-area 354-bus system Area

From

To

B/G (p.u.)

Capacity (MW)

1–2

19

165

100

100

2–3

137

283

100

100

3–1

255

47

100

100

1–2

31

214

200

150

2–3

149

332

200

150

3–1

267

96

200

150

In this case, area 1 has the highest load demands, and area 2 and 3 have five 250-MW wind farms and three 150-MW solar stations, respectively. Wind and solar power scenarios are collected from the yearly data in PJM markets and National Renewable Energy Laboratory (NREL), respectively. Therefore, the cost-minimized MAED schedules the renewable energy from area 2 and 3 to flow into area 1. However, when settled by MP, thermal units in area 1 have incentive to strategically bid, thereby replacing renewable generation and improving market revenues. Thus, it is imperative to guarantee the truthful bids from market participants. To evaluate the potential efficiency loss induced by MP, the renewable energy accommodation and inter-area power under different reserve bids are shown in Fig. 4.11. The inter-area power is the total power flowing from area 2 and 3 to area 1. With the increase in the reserve bids of thermal units in area 1, the inter-area power flowing into area 1 keeps decreasing. When the bids over true costs vary from 1.0 to 5.0, the total amount of inter-area power decreases from 9.04 to 0.46 GWh. As a result, the renewable energy accommodation rate drops from 77.31 to 65.69%, leading to additional 3.15 GWh renewable energy curtailment. In this chapter, the market clearing for multi-area power grids is realized by using the traditional and improved OCD algorithms. The performances of the two algorithms are compared in Table 4.8, where “Tra.” And “Imp.” represent the traditional and improved OCD algorithms, respectively. Given different levels of tolerance σ , both traditional and improved algorithms can achieve high accuracy in solving MAED. However, compared with the traditional one, the proposed algorithm can significantly reduce iterations by 56.82–58.90%, which demonstrates the robustness and efficiency.

4.6 Conclusion In this chapter, an incentive-compatible mechanism based on VCG auction is proposed to elicit truthful bids of generators and coordinate the economic operation of MAPS. By valuing a unit according to others’ bids, the proposed mechanism defines the payment to a unit as other’s incremental costs after removing it, which is proven to fulfill DSIC and other properties. Case studies based on 2-area 60-bus

Inter-area power (MWh)

103

4

2.4

x 10

0.80

Energy Rate

2.3 2.2

0.75 0.70

2.1

0.65

2

0.60

1.9

1.0

1.5

2.0

2.5 3.0 3.5 Bid/True cost

4.0

4.5

5.0

AccommodaƟon rate

Renewable accommodation (MWh)

4.6 Conclusion

0.55

600

Bid/true cost 1.0

400 200 0

5.0 5

10

15 Hour (h)

20

25

Fig. 4.11 Renewable energy accommodation and inter-area power under different reserve bids of the thermal units in area 1 of the 354-bus system

Table 4.8 Performance of traditional and improved OCD algorithms σ (MWh)

Iterations Tra

Imp

Tra

Imp

10

220

95

8.69

9.71

1

309

128

0.96

0.97

0.1

399

164

0.09

0.09

Inter-area power error (MWh)

and 3-area 354-bus systems demonstrate: (1) Under the MP mechanism, thermal units have incentive to bid on high reserve price to prevent inter-area power. As a result, the optimality of MAED cannot be guaranteed, leading to inter-area renewable energy curtailment. (2) Under the proposed mechanism, the dominant strategy of a thermal unit is to truthfully bid. In contrast to MP, an additional payment should be given to thermal units for truth-telling, which is proposed to be allocated to load. (3) The proposed improved OCD algorithm can dramatically improve the efficiency of MAED. Two challenges need an in-depth study in future work: (1) A two-sided market with demand side bidding and the associated incentive-compatible mechanism should

104

4 Sharing Economy in Multi-area …

be investigated. (2) An efficient revenue inadequacy allocation scheme should be designed considering demand response.

References 1. Wang, J., Zhong, H., Lai, X., et al.: Exploring key weather factors from analytical modeling toward improved solar power forecasting. IEEE Trans. Smart Grid 10(2), 1417–1427 (2019) 2. Li, Z., Wu, W., Shahidehpour, M., et al.: Adaptive robust tie-line scheduling considering wind power uncertainty for interconnected power systems. IEEE Trans. Power Syst. 31(4), 2701– 2713 (2016) 3. PJM: MISO. In: Proceedings of Interchange Optimization Workshop, April 18, 2014 [Online] (2014). Available: http://www.pjm.com/~/media/committees-groups/stakeholder-meetings/ pjm-miso-joint-common/20140418/20140418-miso-pjm-jcm-interchange-optimization-pre sentation.ashx 4. Joskow, P., Kahn, E.: A quantitative analysis of pricing behavior in California’s wholesale electricity market during summer 2000. In: Proceedings of IEEE Power Engineering Society Meeting, pp. 392–394. Vancouver, BC, Canada (2001) 5. Wang, P., Sun, H., Hong, X.: A unifying method to supervise market power in bilateral-auction electricity market. In: IEEE Conference on Energy Internet and Energy System Integration, pp. 1–5 (2018) 6. Bakirtzis, A.G., Biskas, P.N.: A decentralized solution to the DC-OPF of interconnected power systems. IEEE Trans. Power Syst. 18(3), 1007–1013 (2003) 7. Karimi, A., Seifi, H., Sheikh-El-Eslami, M.K.: Market-based mechanism for multi-area power exchange management in a multiple electricity market. IET Gener. Transm. Distrib. 9(13), 1662–1571 (2015) 8. Xu, Q., Zhang, N., Kang, C., et al.: A game theoretical pricing mechanism for multi-area spinning reserve trading considering wind power uncertainty. IEEE Trans. Power Syst. 31(2), 1084–1095 (2016) 9. Xu, Y., Low, S.H.: An efficient and incentive compatible mechanism for wholesale electricity markets. IEEE Trans. Smart Grid 8(1), 128–138 (2017) 10. Samadi, P., Mohsenian-Rad, H., Schober, R., et al.: Advanced demand side management for the future smart grid using mechanism design. IEEE Trans. Smart Grid 3(3), 1170–1180 (2012) 11. Tang, W., Jain, R.: Market mechanisms for buying random wind. IEEE Trans. Sustain. Energy 6(4), 1615–1623 (2015) 12. Tran, N.H., Pham, C., Nguyen, M.N.H., et al.: Incentivizing energy reduction for emergency demand response in multi-tenant mixed use buildings. IEEE Trans. Smart Grid 9(4), 3701–3715 (2018) 13. Mediwaththe, C.P., Shaw, M., Halgamuge, S., et al.: An incentive-compatible energy trading framework for neighborhood area networks with shared energy storage. IEEE Trans. Sustain. Energy, early access (2019) 14. Mhanna, S., Verbic, G., Chapman, A.C.: A faithful distributed mechanism for demand response aggregation. IEEE Trans. Smart Grid 7(3), 1743–1753 (2016) 15. Eldridge, B., O’Neill, R.P., Castillo, A.: Marginal loss calculations for the DCOPF. In: Federal Energy Regulatory Commission, Tech. Rep. (2017) 16. Li, X., Yuan, Z., Fu, J., et al.: Nanao multi-terminal VSC-HVDC project for integrating largescale wind generation. In: Proceedings of IEEE Power Energy Society General Meeting, pp. 1– 5. National Harbor, MD, USA (2014) 17. Holttinen, H., Milligan, M., Ela, E., et al.: Methodologies to determine operating reserves due to increased wind power. IEEE Trans. Sustain. Energy 3(4), 713–723 (2012) 18. Vickrey, W.: Counterspeculation, auctions, and competitive sealed tenders. J. Financ. 16(1), 8–37 (1961)

References

105

19. Jain, R., Varaiya, P.: Efficient market mechanisms for network resource allocation. In: IEEE Conference on Decision and Control, pp. 1056–1061 (2008) 20. Wang, J., Zhong, H., Wu, C., et al.: Incentivizing distributed energy resource aggregation in energy and capacity markets: An energy sharing scheme and mechanism design. Appl Energy 252, 113471 (2019) 21. Wang, J., Zhong, H., Tang, W., et al.: Tri-level expansion planning for transmission networks and distributed energy resources considering transmission cost allocation. IEEE Trans. Sustain. Energy 9(4), 1844–1856 (2018) 22. The IBM ILOG CPLEX website. [Online]. Available: http://www-01.ibm.com/software/web sphere/products/optimization/academic-initiative/index.html/ 23. PJM System Operations: Wind Generation and Load. [Online]. Available: http://www.pjm. com/markets-and-operations/ops-analysis.aspx 24. The National Renewable Energy Laboratory website: PVWatts Calculator. [Online]. Available: http://pvwatts.nrel.gov

Chapter 5

Sharing Economy for Renewable Energy Aggregation

5.1 Introduction To achieve the Carbon Neutrality Target issued by China in 2020 [1], it has become an inevitable requirement for energy transition process to vigorously develop renewable energy technologies, e.g., wind, photovoltaic, biomass, etc., toward the construction of a renewable-domiated power system. In recent years, the renewable industries in China have been rapidly developing. For example, China has set an ambitious goal for an over 1,200 GW wind and photovoltaic (PV) portfolio by 2030, accounting for approximately 34% of the national total installed generation capacity. It is estimated that by 2050, the proportion of renewables in primary energy supply will increase to 67% [2]. Renewable energy is of great significance to promote the low-carbon, reliable and efficient development of national energy systems. Renewable energy such as wind and photovoltaics has come to occupy a prominent place on the agenda of governments in most industrialized countries [3]. However, the natural uncertainty and fluctuation of renewable generation results in substantial reliability defects and brings about great challenges for power system operation, which significantly weakens the competitiveness of renewable energy in electricity markets. Considering the demand for energy transition and the importance of power system stability, the solutions to such problems are urgently needed. Generally, the power producer can choose to purchase spinning reserve for renewable energy like wind farms to help reduce the risks of incurring penalties, which could be costly in high penetration of renewable power systems. Concentrating solar power (CSP), which can generate dispatchable renewable power using solar themal energy, is an appealing alternative to help mitigate the uncertainty of volatile wind power. The joint offering of wind and CSP is a viable solution to enhance the competitiveness of variable renewables and achieve a high share of renewable power system. CSP has the ability to store wind energy via electric heating during off-peak hours, and then discharge the energy to alleviate peak load while gaining greater profits. Additionally, the complementarity between wind and solar power can improve the utilization of the thermal storage in CSP, which could © Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_5

107

108

5 Sharing Economy for Renewable Energy Aggregation

reduce its levelized cost of energy while enhancing the market competitiveness of renewables. On the other hand, the increasing penetration of distributed energy resources (DERs) has imposed great challenges to the reliable and economic operation of power grids [4]. To improve the controllability of DERs and enhance system flexibility, the concept of energy sharing has been adopted to balance local demands and make full use of idle DERs in a market fashion. In practice, the idle DERs cannot only be shared among neighbors for load balance, but also be aggregated to provide ancillary service for the connected power grid. For example, the aggregator can take advantage of the surplus capacity of users’ battery storage systems for peak shaving, thereby saving the reliability charges in capacity markets. However, it remains an open question as to how to identify the specific contributions of the participants and achieve a fair profit sharing among them. For example, it would be desirable to develop an effective profit sharing mechanism to compensate CSP plants and incentivize their participation in the joint offering with wind power. Since a CSP plant may have to deviate from its optimal operation to provide flexible regulation for wind power in real-time markets, the renewable aggregation may sacrifice some of their market revenues. On the other hand, the DER owners may behave differently according to their heterogeneous preferences and resources. It is of essential importance to design a profit sharing mechanism to allocate their values that are jointly created in retail markets. To this end, we investigate the profit sharing mechanisms in both large-scale renewables and DERs aggregation in this chapter.

5.2 Aggregation of Wind Farms and Concentrating Solar Power 5.2.1 Problem Description 5.2.1.1

Market Framework

A short-term electricity market is considered in this section, which includes two successive trading floors: the day-ahead market and the balancing market. In the day-ahead market, the power producers submit their hourly energy offers, and a market operator performs a cost-oriented economic dispatch algorithm for market clearing. In the balancing market, the difference between the real-time generation and cleared quantity in the day-ahead market is settled at the imbalance price. When the real-time generation is less than the cleared quantity, the imbalance is negative (under-production), and the power producer has to pay for the deviation at a negative imbalance price. Conversely, the imbalance is positive (extra-production) when realtime generation is more than the cleared quantity, the power producer is paid the deviation at a positive imbalance price.

5.2 Aggregation of Wind Farms and Concentrating Solar Power

109

In this section, a two-price mechanism is applied for the balancing market, which has been adopted in the Spanish electricity market, as well as in some other European electricity markets. Compared with the one-price mechanism (energy imbalance settled at one price, neglecting the imbalance sign), there is no settlement for the energy imbalance that helps to restore the system balance and no arbitrage opportunity for a stochastic producer. The pricing mechanism of the two-price balancing market can be explained as follows:  λ+ t

=

,

(5.1)

λRT ,t , if λRT ,t ≥ λDA,t , λDA,t , if λRT ,t < λDA,t

(5.2)

λRT ,t , if λRT ,t < λDA,t 

λ− t =

λDA,t , if λRT ,t ≥ λDA,t

+/−

where λ(·) are positive and negative imbalance prices in time slot t of the balancing market. λRT ,t is the balancing market prices in time slot t. λDA,t is The day-ahead market prices in time slot t. Under a two-price mechanism, the energy imbalance is priced differently depending on the imbalance sign. Deviation in the opposite direction to the overall system imbalance, which mitigates the system imbalance and helps the system restore the generation-load balance, is priced at the day-ahead market price. In contrast, deviation in the same direction as the overall system imbalance is settled at the clearing price of the balancing market. The positive imbalance price is not more than the dayahead market price, and the negative imbalance price is not lower than the day-ahead market price. More details of the imbalance pricing mechanism can be found in [5].

5.2.1.2

Joint Offering Framework

Based on the above market framework, wind power producers are forced to submit their offers in the day-ahead market in the presence of uncertain generation and market prices. This will cause profit loss in the imbalance settlement. Thus, the optimal offering strategy for wind power producers is to make a tradeoff between the profit in the day-ahead market and the imbalance costs in the balancing market. As the imbalance prices tend to exhibit volatility and can be difficult to forecast, the key approach for wind power producers in maximizing market profit is to find a way to reduce the uncertainty of real-time output. Therefore, wind power and CSP both have incentives. Figure 5.1 illustrates the joint offering model of wind power and CSP in spot markets.

110

5 Sharing Economy for Renewable Energy Aggregation

Fig. 5.1 Schematic of the joint offering of wind and CSP

5.2.2 Offering Strategy Model The following assumptions are made in the proposed model. (1) (2)

The uncertainties of real-time wind power, solar power and electricity price are characterized by a series of scenarios. Day-ahead electricity prices are assumed to be known by forecasting, and wind power producers and CSP plants are considered as price takers.

Multiple wind power producers and CSP plants are considered in this section. The individual strategy without collaboration and the joint offering strategy with collaboration are presented.

5.2.2.1 (1)

Individual Strategy

Offering Strategy of Wind Power Producers

The offering strategy of wind farm i can be formulated as the following two-stage stochastic optimization model: = max RWPP i

WPP PDA,i,t

 t∈

WPP [λDA,t PDA,i,t +



WPP+ − WPP− γs (λ+ s,t Ps,i,t −λs,t Ps,i,t )],

(5.3)

s∈ WPP WPP s.t. 0 ≤ PDA,i,t ≤ Pi,max ,∀t,

(5.4)

5.2 Aggregation of Wind Farms and Concentrating Solar Power

111

WPP+ WPP− WPP WPP Ps,i,t − PDA,i,t = Ps,i,t − Ps,i,t , ∀s, t,

(5.5)

WPP− WPP 0 ≤ Ps,i,t ≤ Pi,max , ∀s, t,

(5.6)

WPP+ WPP 0 ≤ Ps,i,t ≤ Ps,i,t , ∀s, t,

(5.7)

WPP where RWPP is the expected market revenue of wind power producer i. PDA,i,t is the i day-ahead offering power of wind power producer i. γs is weight of scenario s. is WPP + WPP and Ps,i,t are positive and negative imbalance of WPP i the set of scenarios. Ps,i,t WPP in real-time in scenarios s respectively. Pi,max is the maximum output power of WPP WPP i. Ps,i,t is the output power of WPP i in scenarios s. The objective function aims at maximizing the market revenue of wind power WPP . producer i. Constraint (5.4) limits the day-ahead offering power to its capacity Pi,max Constraint (5.5) calculates the positive and negative output deviations from day-ahead offers. Constraints (5.6) and (5.7) present the output deviation limits.

(2)

Offering Strategy of CSP Plants

Similarly, the offering strategy of CSP j is also formulated as a two-stage stochastic optimization model as follows: max RCSP = j

CSP PDA,j,t

 t∈

CSP [λDA,t PDA,j,t +



CSP+ − CSP− γs (λ+ s,t Ps,j,t − λs,t Ps,j,t )],

(5.8)

s∈

CSP+ CSP− CSP CSP s.t. Ps,j,t − PDA,j,t = Ps,j,t − Ps,j,t , ∀s, t,

(5.9)

CSP + CSP 0 ≤ Ps,j,t ≤ Pj,max , ∀s, t,

(5.10)

CSP CSP 0 ≤ Ps,j,t ≤ Pj,max , ∀s, t,

(5.11)

pc

solar thd EH thc PC Ps,j,t +ηd Ps,j,t +Ps,j,t − Ps,j,t /ηc − us,j,t PjSU =Ps,j,t /ηPC , ∀s, t,

(5.12)

th Ej,th,min ≤ Es,j,t ≤ Ej,th,max , ∀s, t,

(5.13)

thc thc 0 ≤ Ps,j,t ≤ Pj,max · φj,c , ∀s, t,

(5.14)

thd thd 0 ≤ Ps,j,t ≤ Pj,max · φj,d , ∀s, t,

(5.15)

φj,c + φj,d ≤ 1,

(5.16)

112

5 Sharing Economy for Renewable Energy Aggregation

  th th thc thd , ∀s, t, Es,j.t = (1 − γ )Es,j,t−1 + Ps,j,t−1 − Ps,j,t−1

(5.17)

PC PC PC Ps,j,t,min ≤ Ps,j,t ≤ Ps,j,t,max , ∀s, t,

(5.18)

CSP EH PC Ps,j,t + Ps,j,t /ηEH =Ps,j,t , ∀s, t,

(5.19)

CSP is expected market revenue of CSP j.  is the set of time. PDA,j,t is the where RCSP j CSP + day-ahead offering power of wind power producer j. γs is weight of scenario s. Ps,j,t CSP and Ps,j,t are positive and negative imbalance of CSP j in real-time in scenarios s CSP CSP respectively. Pj,max is the maximum output power of CSP j. Ps,j,t is the output power solar of CSP j in scenarios s. Ps,j,t is the solar power j in scenarios s. ηd is the discharging thd is the discharging thermal power of efficiency of thermal energy storage (TES). Ps,j,t EH thc thermal energy storage. Ps,j,t is the electric heater (EH) power j in scenarios s. Ps,j,t is the charging thermal power of thermal energy storage. ηc is the charging efficiency pc of TES. us,j,t is the binary variable of the start-up status of Power cycle (PC) of CSP. pc PjSU is the energy consumption of the start-up of CSP. Ps,j,t is the output of PC in CSP. th is the state-of-charge of TES. Ej,th,min ηPC is the conversion efficiency of PC. Es,j,t and Ej,th,max are the minimum and maximum state-of-charge of TES respectively. thc thd is the maximum charging thermal power of thermal energy storage. Pj,max is Pj,max the maximum discharging thermal power of thermal energy storage. φj,c is the binary variable of the charging status of CSP j. φj,d is the binary variable of the discharging PC PC and Ps,j,t,max are the minimum and maximum output of PC status of CSP j. Ps,j,t,min in CSP. ηEH is the conversion efficiency of EH. Constraint (5.12) represents the instantaneous balance of thermal power in CSP plants. Constraint (5.13) depicts the feasible operation interval. Constraints (5.14) and (5.15) set the charging and discharging rates within the operation bound. Constraint (5.16) guarantees that the charging and discharging states will not occur simultaneously. Constraints (5.17) and (5.18) limit the state of charge (SOC) of TES and the output power of PC, respectively. Constraint (5.19) formulates the net output of CSP with TES and EH.

5.2.2.2

Joint Offering Strategy

The joint offering strategy of the aggregation of wind power producers and CSP plants is modeled as the following two-stage stochastic optimization model: ⎛ max

WPP CSP PDA,i,t ,PDA,j,t

RA = ⎝

 i∈φ WPP

RWPPA + i

 j∈φ CSP

⎞ ⎠ RCSPA j

5.2 Aggregation of Wind Farms and Concentrating Solar Power

=



A λDA,t PDA,i,t

t∈

s∈



A s.t. PDA,i,t =

+



WPP PDA,i,t +

i∈φ WPP

 i∈φ WPP

WPP WPP (Ps,i,t − PDA,i,t )+



113

 A+ γs (λ+ s,t Ps,i,t



−

A− λ− s,t Ps,i,t )

,

CSP PDA,j,t , ∀i, j, t,

(5.20) (5.21)

j∈φ CSP

A+ A− CSP CSP (Ps,j,t −PDA,j,t ) = Ps,i,t − Ps,i,t , ∀s, i, j, t,

(5.22)

j∈φ CSP A− 0 ≤ Ps,i,t ≤



WPP Pi,max +

i∈φ WPP A+ ≤ 0 ≤ Ps,i,t



CSP Pj,max , ∀s, i, j, t,

(5.23)

CSP Pj,max , ∀s, i, j, t,

(5.24)

j∈φ CSP WPP Ps,i,t +

i∈φ WPP A 0 ≤ PDA,i,t ≤





 j∈φ CSP

WPP Pi,max +

i∈φ WPP



CSP Pj,max , ∀i, j, t,

(5.25)

j∈φ CSP

Constraints(5.12) − (5.19), where superscript A denotes the variable in joint offering. RA is the expected market is the revenue of the aggregation of wind power producers and CSP plants.RWPPA i is the expected expected market revenue of the WPP i of the aggregation. RCSPA j market revenue CSP j of the aggregation.WPP is the set of WPPs. CSP is the set of A is the day-ahead offering power of joint offering of aggregation. CSP plants. PDA,i,t A+ A− Ps,i,t and Ps,i,t are positive and negative imbalance of joint offering of aggregation in real-time in scenarios s respectively. In the above joint offering strategy model, the imbalance is computed over the joint offer of aggregation. Obviously, due to the benefits of the joint offering of wind power producers and CSP plants, the expected revenue of the joint offering of aggregation RA is supposed to exceed the sum of the expected revenue of individual offerings for each wind power producer or CSP plant. In the following section, we propose profit sharing schemes to distribute the extra revenue within the coalition based on Nash bargaining.

5.2.3 Profit Sharing Mechanism A fair profit sharing mechanism should identify the different contributions of participants in aggregation. In this section, a contribution ratio in joint offering is presented to quantify the relevant contributions of wind power producers and CSP plants in both the day-ahead market and the balancing market. Based on the indices, an asymmetric Nash bargaining (ANB) model is formulated for profit sharing.

114

5 Sharing Economy for Renewable Energy Aggregation

The incentive mechanism is required to satisfy the following properties: (i)

(ii) (iii) (iv) (v)

(Pareto Optimality) There is no other solution, in which every participant’s cost is no greater than that in the proposed joint sharing scheme, and some participants’ cost is strictly less than that in the proposed scheme. (Individual Rationality) All participants should reduce their costs by joint offering compared with their costs from individual offering. (No Exploitation) The one not in aggregation should not be given any benefits. (Monotonicity) If a participant makes more contributions to joint offering, this participant should gain more benefits. (Budget Balance) The total benefits are allocated among all participants.

5.2.3.1

Contribution Ratio in Joint Offering

The contributions of wind power producers and CSP plants in aggregation consist of two components. One is the revenue in the day-ahead market according to the energy offered by wind power producers and CSP plants. The other is the expected performance in the balancing market. For wind power producers, different prediction accuracies may lead to different levels of imbalance in real time. For CSP plants, some of them may deviate from their individual schedule to maximize the profit of aggregation in joint offering. Hence, the contribution ratio in the joint offering proposed considers these two aspects. (1)

Contribution in the Day-ahead Market

The contribution in the day-ahead market can be measured using the corresponding revenue, which can be expressed as the product of the day-ahead market price and offered energy. Therefore, the contribution of wind power producers and CSP plants can be defined as follows:  WPP WPP∗ = (λDA,t PDA,i,t ), ∀i, j, t, (5.26) CDA,i,t t∈ CSP CDA,j,t =



CSP∗ (λDA,t PDA,j,t ), ∀i, j, t,

(5.27)

t∈

WPP where the term with superscript * denotes the optimal joint offering strategy. CDA,i,t CSP is the contribution of WPP plant i in the day-ahead market. CDA,j,t is the contribution of CSP plant j in the day-ahead market.

(2)

Contribution in the Balancing Market

Let ζs,i and ζs,j denote the imbalance sign of wind power producers and CSP plants in scenario s, respectively. If the imbalance is in the opposite direction to the overall aggregation imbalance, which helps the aggregation reduce the total imbalance in real time, then ζ = 1. Conversely, ζ = −1 when the imbalance directions of participants

5.2 Aggregation of Wind Farms and Concentrating Solar Power

115

and aggregation are the same. The contribution in the balancing market can be defined as follows:  WPP∗+ WPP∗− WPP ζs,i γs (λ+ +λ− ), ∀s, i, t, (5.28) CRT ,i,t = s,t Ps,i,t s,t Ps,i,t t∈ s∈ CSP CRT ,j,t =



CSP∗+ CSP∗− ζs,j γs (λ+ +λ− ), ∀s, j, t, s,t Ps,j,t s,t Ps,j,t

(5.29)

t∈ s∈ WPP∗+ WPP WPP∗ Ps,i,t = (Ps,i,t − Ps,i,t )+ , ∀s, i, t,

(5.30)

WPP∗− WPP∗ WPP Ps,i,t = (Ps,i,t − Ps,i,t )+ , ∀s, i, t,

(5.31)

CSP∗+ CSP CSP∗ Ps,j,t = (Ps,j,t − Ps,j,t )+ , ∀s, j, t,

(5.32)

CSP∗− CSP∗ CSP Ps,j,t = (Ps,j,t − Ps,j,t )+ , ∀s, j, t,

(5.33)

WPP where (·)+ = max(0, ·). CRT ,i,t is the contribution of WPP plant i in the balancing CSP market. CRT ,j,t is the contribution of CSP plant j in the balancing market. Therefore, based on the contributions in both the day-ahead market and balancing market, the contribution ratio in the joint offering of wind power producers and CSP plants can be formulated as follows: WPP WPP t∈ (CDA,i,t +CRT ,i,t ) WPP , ∀i, j, t, (5.34) τi = WPP WPP CSP CSP t∈ (CDA,i,t + CRT ,i,t +CDA,j,t +CRT ,j,t ) CSP CSP t∈ (CDA,j,t +CRT ,j,t ) CSP τj = , ∀i, j, t, (5.35) WPP WPP CSP CSP t∈ (CDA,i,t + CRT ,i,t +CDA,j,t +CRT ,j,t )

where τiWPP and τjCSP denote the contribution ratio of wind power producers i and CSP plants j, respectively.

5.2.3.2

Mechanism Design

In this subsection, a profit sharing scheme based on Nash bargaining theory is formulated. Nash bargaining theory is widely applied to allocate the profit in a cooperative game by maximizing the product of market participants’ excess utilities. The existing profit sharing schemes are mostly based on the symmetric Nash bargaining (SNB) method, in which the extra profits are allocated with identical shares. However, each participant in aggregation makes variable contributions in joint offering, and a well-designed joint offering incentive mechanism should be

116

5 Sharing Economy for Renewable Energy Aggregation

able to identify the different contributions of joint offering participants and reward good behavior. Therefore, we propose a profit sharing scheme based on the asymmetric Nash bargaining method, in which different contribution ratios in joint offering are considered. The profit sharing schemes proposed in this section can be realized by optimizing the following Nash bargaining model:

max

πiWPP ,πjCSP

×

(RWPPA∗ − RWPP* + πiWPP )τi i i

WPP

i∈φ WPP

(RCSPA∗ − RCSP∗ + πjCSP )τj , ∀i, j, j j CSP

(5.36)

j∈φ CSP

− RWPP∗ + πiWPP ≥ 0, ∀i, s.t. RWPPA∗ i i

(5.37)

RCSPA∗ − RCSP∗ + πjCSP ≥ 0, ∀j, j j

(5.38)

 i∈φ WPP

πiWPP +



πjCSP = 0, ∀i, j,

(5.39)

j∈φ CSP

where πiWPP is the payment in Nash bargaining for WPP i. πjCSP is the payment in Nash bargaining for CSP j. Objective function (5.36) indicates the increase in expected revenue when wind power producers and CSP plants jointly offer in the day-ahead market. The objective function is the Cobb–Douglas utility function. The economic significance is that through the associated payment among CSP plants and wind power producers, the additional revenue obtained by the joint offering is distributed based on a certain weight, so they all have sufficient incentives to participate in the joint offering. Here, the associated payments of wind power producers i and CSP j are denoted as πiWPP and πjCSP , respectively. Constraints (5.37) and (5.38) denote that the total expected revenue of each participant in aggregation will increase when jointly offering. Constraint (5.39) is market clearing constraint. Then, the analytical expression of the Nash bargaining model proposed in this section is as follows (the proof is the similar as that in Sect. 2.5.4): − RWPPA∗ + τiWPP R, ∀i, πiWPP* = RWPP* i i

(5.40)

πjCSP* = RCSP∗ − RCSPA∗ +τjCSP R, ∀j, j j

(5.41)

where R is the aggregation surplus, i.e., the increment of market returns after aggregation. From constraints (5.40) and (5.41), the optimal payment of wind power producers i, denoted by πiWPP* , or CSP j, denoted by πjCSP* , consists of two parts. Here,

5.2 Aggregation of Wind Farms and Concentrating Solar Power

117

we take the payment of wind power producers as an example: the first part, RWPPA* – i , is the change in the expected market revenue of wind power producers when RWPP* i jointly offering in the day-ahead market, and the second part, τi R, is the revenue from aggregation surplus based on contribution ratio in the joint offering. The extra profits are allocated to each participant in aggregation according to their contributions in both the day-ahead and balancing market. Thus, the proposed profit sharing mechanism can achieve a fair and reasonable distribution of profits among participants of the aggregation.

5.3 Aggregation of Distributed Energy Resources in Energy Markets 5.3.1 Energy Sharing Scheme In this section, we consider that one aggregator organizes N energy users in a distribution grid in a day-ahead market. Each user has a PV system, an ESS and local load. The energy trading without and with energy sharing are compared.

5.3.1.1

Energy Trading Without Energy Sharing

In this case, energy users are assumed to be price-takers, who purchase electricity from the aggregator at a retail rate, and sell back the surplus power of DERs at a net metering rate (NMR). The following trading events happen in order. (i) (ii) (iii)

Each user schedules his/her local DERs and determines the net load to minimize individual costs under fixed rates. The aggregator collects user’ net load information, and trades with the connected power grid for energy balance. The aggregator charges each user for the net load at a retail rate, and pays each user for the net power at an NMR.

In practice, as the retail rate is generally higher than the NMR, the users cannot get sufficient and reasonable payments to justify the investments of DERs. For example, the NMR is about 3 cents/kWh in Pacific Gas and Electric Company (PG&E) in California in 2017. However, the retail rate during peak hours can reach 0.263 $/kWh [6].

5.3.1.2

Energy Trading with Energy Sharing

As electricity is an undifferentiated good, a pool-based energy sharing platform is considered, in which each energy user schedules the amount of shared energy and the

118

5 Sharing Economy for Renewable Energy Aggregation

aggregator determines the associated payment for each user. The following events happen in order. (i) (ii) (iii) (iv)

All users enroll sharing contracts with the aggregator, which sets up a rule that determines the sharing incentives to users for an amount of shared energy. Each energy user schedules his/her DERs and deviates from individual optimum to share DERs. The aggregator collects the users’ net load and shared energy information, and trades with the connected power grid for energy balance. The aggregator charges each user for the net load at a retail rate, and pays each user for the net power at an NMR, and for the shared energy with the sharing incentives.

Instead of only trading with the aggregator, the users can share DERs with each other in the platform. The proposed energy sharing scheme enables the aggregator to organize users to cooperate as a single interest entity, and minimize the total costs. In contrast to the case without energy sharing, the proposed scheme achieves Pareto optimality of the aggregator and all users.

5.3.1.3

Decentralized Implementation

The proposed energy sharing scheme aims at maximizing the total benefits of the aggregator and users, which requires detailed information about users’ preferences and DERs. However, it is challenging to collect users’ private information and schedule energy sharing in a centralized manner. Thus, a decentralized framework is developed to preserve users’ privacy. The schematic is shown in Fig. 5.2. Each user is equipped with an energy management controller (EMC), which controls the hourly load consumption and communicates this load information to the

Aggregator

Surplus energy Purchased energy Shared energy Information flow

Aggregator

Price

Net power

EMC

EMC

EMC

User

User

Without sharing Fig. 5.2 Decentralized framework for energy sharing

EMC

User

User

Energy sharing

5.3 Aggregation of Distributed Energy Resources in Energy Markets

119

aggregator. The EMC also receives the price signals from the aggregator. Therefore, the bidirectional communication makes the interactions easy between the aggregator and users. Based on the price signal, the EMC of each user optimally schedules local load and DERs. Then the EMCs communicate the net power to the aggregator. After collecting all users’ net power, the aggregator updates the price signal and sends it back to the EMCs of users. The proposed energy sharing scheme can be applied in day-ahead and intra-day markets.

5.3.2 System Model Pursuing clarity and simplicity, the modeling assumptions are as follows: (i) We consider one aggregator organizes N energy users in the distribution grid. (ii) The locational marginal prices (LMPs) in the connected power system are predefined and constant. (iii) The distribution grid network and energy losses are not considered in this section.

5.3.2.1

The Aggregator’s Net Cost

To depict the uncertainties in load and solar power, scenario-based stochastic programming is adopted. Thus, the net cost of the aggregator OA is calculated below: O = A



NL γs λin t,s Pt,s

−

NS λout t,s Pt,s

t,s

+

λSt



NS Pi,t,s

−

λRt



i

 NL Pi,t,s

,

(5.42)

i

where    NL NS Pi,t,s − Pi,t,s = , ∀t, s,

NL Pt,s

i

NS Pt,s

+



  NS NL Pi,t,s − Pi,t,s = , ∀t, s. i

(5.43)

(5.44)

+

out In (5.42), γs is the probability of scenario s. λin t,s and λt,s are the prices for purchasing and selling electricity with the power grid, which are different in practice due to transmission service charges, tax, etc. λSt and λRt are the retail rate and net NL NS and Pt,s are users’ aggregated net load and surplus power, and metering rate. Pt,s NL NS Pi,t,s and Pi,t,s are the net load and surplus power of user i. In (5.42), the first term represents the aggregator’s cost for purchasing electricity from the power grid, and the second term represents the revenue by selling electricity back. The third and fourth terms are the costs and revenues of the aggregator trading

120

5 Sharing Economy for Renewable Energy Aggregation in/out

with users. Two ratios ωin/out are used to reflect the difference between λt,s locational marginal prices (LMPs) λLMP t,s : in LMP out out LMP λin t,s =ω λt,s , λt,s =ω λt,s , ∀t, s.

and

(5.45)

In (5.43) and (5.44), (·)+ = max{0, ·}. Equations (5.43) and (5.44) calculates user’s net load and power surplus. In practice, the net loads of distribution grids and the LMPs are influenced by each other. Analyzing the mutual influence needs solving the market equilibrium between the distribution grid and the connected power system, deserving an in-depth study in the future. To focus on energy sharing in a distribution grid, we do not consider the impacts of the net loads on LMPs, and model the LMPs as predefined constants.

5.3.2.2

A User Model Without Energy Sharing

In this section, each user is modeled as an agent solving the following stochastic program: min

   L   ESS  NL NS ESS + ciESS Pi,t,s,α , γs λRt Pi,t,s − λSt Pi,t,s − Ui Pi,t,s + Pi,t,s,β

(5.46)

t,s

subject to NL NS L PV ESS ESS Pi,t,s − Pi,t,s = Pi,t,s − Pi,t,s + Pi,t,s,α − Pi,t,s,β , ∀t, s,

 NL  NS L PV ESS ESS ESS Xi = Pi,t,s , Pi,t,s , Pi,t,s , Pi,t,s , Pi,t,s,α , Pi,t,s,β , Ei,t,s , ∀t, s ∈ χi ,

(5.47) (5.48)

where Ui (·) is user i’s utility function. Without loss of generality, we use a quadratic concave utility function. ciESS is the operation cost of user i’s ESS, which is caused by the degradation of the ESS’s charging and discharging [7] In practice, an ESS’s cost mainly comes from the capital cost, while the operation cost is measured as the replacement cost of the storage bank over the lifetime throughput, which is relatively a small fraction. The decision variables are denoted by Xi , including user i’s net load NL NS L PV , surplus power Pi,t,s , hourly load Pi,t,s , solar power Pi,t,s , the charging and Pi,t,s ESS ESS ESS discharging power Pi,t,s,α and Pi,t,s,β , and the stored energy in the ESS Ei,t,s . In (5.48), the feasible region χi includes user i’s constraints, shown as follows: NL NS C , Pi,t,s ≤ Pi,max , ∀t, s, 0 ≤ Pi,t,s

(5.49)

PV APV 0 ≤ Pi,t,s ≤ Pi,t,s , ∀t, s,

(5.50)

5.3 Aggregation of Distributed Energy Resources in Energy Markets L L L Pi,t,s,min ≤ Pi,t,s ≤ Pi,t,s,max , ∀t, s,



L L Pi,t,s ≥ Qi,s , ∀s,

121

(5.51) (5.52)

t ESS ESS 0 ≤ Pi,t,s,α ≤ Pi,α,max , ∀t, s,

(5.53)

ESS ESS 0 ≤ Pi,t,s,β ≤ Pi,β,max , ∀t, s,

(5.54)

ESS ESS ESS ESS Ei,t,s = Ei,t−1,s + ηiESS Pi,t,s,α − Pi,t,s,β /ηiESS , ∀t, s,

(5.55)

ESS ESS ESS Ei,min ≤ Ei,t,s ≤ Ei,max , ∀t, s,

(5.56)

ESS ESS Ei,N T ,s = Ei,0,s , ∀s,

(5.57)

APV L is user i’s available solar power, Pi,t,s,min/max represent the where Pi,t,s L minimum/maximum of user i’s hourly load, Qi,s is user i’s daily load requirement, ESS Pi,α/β,max are the maximal power of charging/discharging, ηiESS is the efficiency of ESS user i’s ESS, Ei,min/max are the minimum/maximum of the stored energy in the ESS, T and N is the number of daily time slots. Constraint (5.49) shows the limit for user i’s net power. Constraint (5.50) shows user i’s solar power is bounded by the forecast value. In (5.51), user i’s hourly load is limited by the lower and upper bounds. Constraint (5.52) shows user i’s daily minimal load requirement. Constraints (5.53) and (5.54) show the bounds of charging and discharging power of user i’s energy storage. Constraint (5.55) shows the dynamics of the stored energy in user i’s energy storage, restricted by the lower and upper bounds in (5.56). In (5.57), the stored energy in the final time slot is equal to the initial value. In (5.46), let OiU denote user i’s objective function, which is to minimize the difference between the total costs and his/her utility. In (5.47), the net power equals the difference between the load and the power of the PV and ESS. Note that the NL NS and Pi,t,s minimization of the objective function guarantees that at least one of Pi,t,s NL NS is 0. When the user’s net load is positive, Pi,t,s > 0 and Pi,t,s = 0. When the user’s NL NS net load is negative, Pi,t,s = 0 and Pi,t,s > 0. In (5.48), the feasible region χi includes the constraints for user i’s PV, ESS and load. In model (5.46)–(5.48), user i can only trade with the aggregator, without sharing DERs with others. Let OiU ,0 and OA,0 be user i’s objective value and the aggregator’s net cost without energy sharing, known as the disagreement point [8, 9].

122

5.3.2.3

5 Sharing Economy for Renewable Energy Aggregation

The Energy Sharing Model

In the proposed energy sharing scheme, the aggregator organizes all users to cooperate as a single interest entity, and the shared power from users can be optimized. Energy sharing requires users to deviate from individual optimal schedule to accommodate the surplus or demand from their neighbors. Thus, the aggregator should incentivize the users to share DERs by allocating the sharing benefits. The payment from the aggregator to user is πiES . User i’s net cost is OiU − πiES , and the aggre i ES A gator’s net cost is O + πi . The energy sharing model is to minimize the total i

costs of the aggregator and all users: min OA +



πiES +

i

   OiU , OiU − πiES = OA + i

(5.58)

i

subject to (5.43)–(5.45), and NL NS ES L PV ESS ESS − Pi,t,s − Pi,t,s = Pi,t,s − Pi,t,s + Pi,t,s,α − Pi,t,s,β , ∀i, t, s, Pi,t,s



ES Pi,t,s = 0 : λES t,s , ∀t, s,

(5.59) (5.60)

i C ES C ≤ Pi,t,s ≤ Pi,max , ∀i, t, s, X i ∈ χi , −Pi,max

(5.61)

ES where Pi,t,s is the variable for shared power, restricted by the capacity of a building’s C fuse Pi,max . In contrast to (5.47), the decision variables for shared power are added to (5.59). Equation (5.60) is the constraint for balancing the shared power. Note that the loss of energy sharing and the distribution network are not considered in this section [8]. The Lagrangian multiplier of constraint (5.60) is denoted by λES t,s , interpreted as the clearing price for shared power. Let OiU ,1 and OA,1 be user i’s objective value and the aggregator’s net cost by solving the energy sharing model. Although the energy sharing model defines the amount of users’ shared power, it cannot reveal the payments to users that will incentivize the deviation from individual optimum. Thus, an incentive mechanism is proposed for benefit allocation according to users’ contributions.

5.3.3 Profit Sharing Mechanism As energy sharing requires users to deviate from individual optimal schedule, thus increasing individual costs, an incentive mechanism is needed for benefit allocation so as to incentivize users to participate in energy sharing. In this section, an incentive mechanism is required to satisfy the properties like those illustrated in Sect. 5.2.3.

5.3 Aggregation of Distributed Energy Resources in Energy Markets

123

We firstly propose a novel index, termed as SCR, to evaluate users’ contributions to energy sharing. Then an ANB model considering SCRs is developed for benefit allocation.

5.3.3.1

Sharing Contribution Rate

In this section, a user’s contribution to energy sharing is defined as the economic U value of the shared DERs, Ci,t,s :  ES∗  U   = λES Ci,t,s t,s Pi,t,s , ∀i, t, s,

(5.62)

 ES∗   is the amount of a user’s shared power, and λES is the clearing price where Pi,t,s t,s U for shared power. Ci,t,s can be interpreted as the economic value for sharing DERs, which is a straightforward choice to measure users’ contributions to energy sharing. Alternative definitions of users’ contributions deserve in-depth study in the future. Therefore, user i’s SCR, denoted by SCRi , is defined as his/her contributions over the total contributions of all energy users: U  t,s γs Ci,t,s  A , ∀i, SCRi = 1 − τ U j,t,s γs Cj,t,s

(5.63)

where τ A ∈ (0, 1) is the aggregator’s rate of return, which is predefined and constant in this section. In regulated environment, the permitted rate of return of an aggregator is regulated by the government. In deregulated markets, the selection of τ A involves an aggregator’s pricing strategy: A higher τ A can improve the aggregator’s rate of return, but lead to the loss of users. A lower τ A may attract more users, but the rate of return decreases. Note that for the users without participating in energy sharing, their shared power is always 0. According to (5.63), their SCRs equal 0.

5.3.3.2

Mechanism Design

The Nash bargaining problem studies how market participants share a surplus that they jointly generate by maximizing the product of market participants’ excess utilities [10]. In a few existing studies, symmetric Nash bargaining (SNB) models are adopted for cost allocation, in which market participants are assigned with identical contribution rates regardless of their distinct behaviors [8, 9]. However, users can make different contributions to the energy sharing. For example, a user shares 1 kWh electricity while another user shares 10 kWh electricity. These two users should be allocated with distinct benefits. An asymmetric Nash bargaining model based on users’ SCRs is proposed for benefit allocation:

124

5 Sharing Economy for Renewable Energy Aggregation

 max

O

A,0

−O − A



τ A

SCRi

 U ,0 Oi − OiU (Xi ) + πiES ,

πiES

i

(5.64)

i∈ES

subject to (5.43)–(5.45), (5.59)–(5.61) and OA,0 − OA −



πiES ≥ 0,

(5.65)

i

OiU ,0 − OiU + πiES ≥ 0, ∀i ∈ ES ,

(5.66)

  ES where the decision variables are πiES , Pi,t,s , Xi , ∀i, t, s , and ES represents the set of the users whose SCRs are positive. Constraints (5.65) and (5.66) are the individual rationality for the aggregator and users. By solving the ANB model, the aggregator’s payments to users can be obtained. The proposed incentive mechanism satisfies (i) Pareto optimality, (ii) individual rationality, (iii) no exploitation, (iv) monotonicity, and (v) budget balance. The solution to the ANB model defines the payments to users: πiES∗ = SCRi · + OiU ,1 − OiU ,0 , ∀i,

(5.67)

where πiES∗ is the optimal payment to user i, and the total benefit induced by energy sharing is: =OA,0 − OA,1 +



 OiU ,0 − OiU ,1 ≥ 0.

(5.68)

i

The proof of (5.67) is the similar as that in Sect. 2.5.4. According to (5.67), the optimal payment to a user can be interpreted as two parts. The first is OiU ,1 − OiU ,0 , representing the incremental cost after sharing DERs. The second is SCRi , indicating a user’s share of the total benefits that all market participants jointly generate. The net benefits of the aggregator and users are derived below:   (5.69) OiU ,0 − OiU ,1 − πiES∗ =SCRi · , ∀i,  O

A,0

− O

A,1

+



 πiES∗

=τ A · .

(5.70)

i

Each market participant’s benefit is related to his/her SCR, reflecting the contributions he/she creates. Therefore, all participants can benefit from the energy sharing.

5.3 Aggregation of Distributed Energy Resources in Energy Markets

125

5.3.4 Solution Algorithm With the rapid growth of DERs, it becomes challenging to schedule the energy sharing among a large number of users and allocate the sharing benefits in a centralized manner mainly due to the privacy concern. Distinguished from the generators in wholesale markets, electricity users may be reluctant to bid frequently, and thus we decentralize the proposed energy sharing and settlement models to preserve users’ privacy and avoid users’ bidding. In this section, the alternating direction method of multipliers is adopted for decentralized implementation. To decompose the energy sharing model, the following constraints are embedded: x x x = Pi,s,t : wi,s,t , ∀i, s, t, x ∈ {NL, NS, ES}, Pˆ i,s,t

(5.71)

x where Pˆ i,s,t is an auxiliary variable, interpreted as user i’s net load, net power and x is the Lagrangian multishared power recommended by the aggregator, and wi,s,t plier, i.e., the coordination price sent by the aggregator. By relaxing the constraints (5.71), the energy sharing model can be decomposed into a local program for the aggregator and N individual programs for users. Then, the energy sharing model can be iteratively solved by coordinating the aggregator’s local program and the users’ local programs, which is shown in Fig. 5.3. Similarly, we can decompose the asymmetric NB model by introducing and relaxing the following constraints:

πˆ iES = πiES : wiπ , ∀i,

(5.72)

Aggregator Optimize the local program for the aggregator min Lagrangian function of aggregator s.t. i) total net load and surplus power ii) sharing power balance

s net cost

Net load, surplus power and shared power Coordination price and recommended value

Coordination price and recommended value

User i User 1

Optimize individual program

User N

Optimize individual program

min Lagrangian function of user i's net cost s.t. i) individual power balance ii) DER operation

Optimize individual program

Fig. 5.3 Iterative coordination between the aggregator and all users

126

5 Sharing Economy for Renewable Energy Aggregation

where πˆ iES is an auxiliary variable, interpreted as the recommended payment to user i, and wiπ is the Lagrangian multiplier. As the energy sharing model and the asymmetric NB model are convex, the convergence and optimality of the solution algorithm can be guaranteed [11].

5.4 Aggregation of Distributed Energy Resources in Energy and Capacity Markets 5.4.1 Energy Sharing Scheme 5.4.1.1

Market Framework

Time Event Participant Three years

Aggregator User

Capacity market aucƟon (Base residual aucƟon)

10

20

1

Reliability charge for daily peak load

2

Time

User At least three incremental aucƟons

Delivery year

Daily load

User

One year

Daily load

In this section, we consider a distribution grid with an aggregator that serves N energy users. An energy user is assumed to be a price-taker in joint energy and capacity markets. Capacity markets have been developed in many regions to address longterm resource adequacy problems, e.g., Midcontinent independent system operator (MISO), New York ISO (NYISO) and Pennsylvania-New Jersey-Maryland (PJM) [12]. Without loss of generality, PJM’s capacity market that coexists with the energy market is adopted, in which an auction is held every year. The framework for one complete auction process is shown in Fig. 5.4. PJM’s capacity market has a multi-auction structure designed to procure resources that balance the region’s long-term load. A base residual auction (BRA) is held for the delivery year, which is three years in the future. After BRA, at least three incremental auctions are conducted for additional resource commitments to satisfy potential load changes prior to the start of the delivery year. Participation by load serving entities (LSEs) in PJM’s capacity market is mandatory. During each delivery year, the aggregator is responsible for paying two charges. On the one hand, the aggregator participates in the capacity market and pays a reliability charge [13]. Based on the aggregator’s daily peak load and capacity market prices, the reliability charge CdR is calculated daily as follows:

10

1

Time

Fig. 5.4 PJM’s capacity market that coexists with the energy market

20

2

Electricity charge for hourly load

5.4 Aggregation of Distributed Energy Resources …

  CdR = max PdNL,t · λCd , t

127

(5.73)

  where the first term max PdNL,t represents the aggregator’s peak load during day d, t

and the second term λCd is the capacity market price. On the other hand, during each day of the delivery year, the aggregator participates in the day-ahead energy market to satisfy N energy users’ hourly loads. The aggregator’s daily electricity cost CdE is calculated as follows:    (5.74) λEd,t PdNL,t − PdNS,t , CdE = t

  where λEd,t is the electricity price on day d and time t, and the term PdNL,t − PdNS,t is the aggregator’s net load on day d and time t. From Eqs. (5.73) and (5.74), sharing DERs can help to decrease the aggregator’s total costs. DERs can be shared among users for local power balance, substituting the electricity from power grids. More importantly, the aggregator can take advantage of the users’ shared DERs for peak shaving, which flattens load profiles and enhances system flexibility. In this section, we only consider the aggregator’s participation in BRA, without the incremental auctions and the realization in the delivery year. In other words, we focus on the decision making of the aggregator and users before the delivery year considering the uncertainty in renewable generation and prices. Before the delivery year, each user strategically invests in DERs and minimizes annualized individual costs. Then the aggregator implements an energy sharing scheme and an incentive mechanism to encourage users to share DERs. The energy sharing scheme is detailed below.

5.4.1.2

Energy Trading Scheme Comparison

In this subsection, the energy trading schemes with and without energy sharing are compared, shown in Fig. 5.5. As one can observe, in the case without energy sharing, users are assumed to be price-takers who only trade with the aggregator. The users purchase electricity from the aggregator at a retail rate, and sell back the surplus energy from DERs at a net metering rate (NMR). Without energy sharing, each user has no information about the demand or surplus of neighbors’ DERs. Thus, only individual costs are minimized by scheduling DERs and local load. Then, the aggregator’s net cost is calculated as the difference between the price charged by the power grid and the revenues from users. However, users may not receive sufficient and reasonable benefits to justify the investment costs of DERs. For example, the NMR is about 3 cents/kWh for the Pacific Gas and Electric Company (PG&E) in California in 2017, which is much lower than retail rates. Thus, DER planning may be limited without energy sharing.

128

5 Sharing Economy for Renewable Energy Aggregation

Without energy sharing

Energy sharing

Each user minimizes individual costs by planning DERs.

Users enroll sharing contracts with the aggregator.

Each user invests in and schedules DERs considering energy sharing. The aggregator collects users’ net load and trades with power grids. The aggregator uses the shared DERs for power balance and peak shaving.

The aggregator charges users for load, and pays users for surplus DERs.

The aggregator charges users for load, and pays users for surplus DERs and sharing benefits.

Fig. 5.5 Events in cases with and without energy sharing

In the case with energy sharing, a pool-based sharing platform is established, in which users purchase or sell shared energy and the aggregator determines the associated payments to each user. Before the delivery year, all the users enroll sharing contracts with the aggregator, which set up the rule of sharing benefit allocation. Given the sharing contract, each user determines the sizing of DERs anticipating the optimal scheduling for DERs. In the energy sharing platform, the aggregator collects information regarding all users’ demands and surpluses and helps match DER sharing among users. At the same time, the aggregator takes advantage of shared DERs for peak shaving to reduce the reliability charge. To encourage users to invest in DERs and participate in energy sharing, an incentive mechanism is designed by identifying different users’ contributions. The following events occur in order before the delivery year: (i) (ii) (iii)

(iv)

Each energy user enrolls in a sharing contract with the aggregator, which sets up the rules for sharing payments to users. Each user optimizes the DER capacity and schedules the DERs for energy sharing by deviating from individual optimum. To reduce the electricity and reliability charges, the aggregator uses the shared DERs to satisfy local loads and reduce peak loads. Then, the aggregator collects the information on users’ hourly net load and power. The aggregator charges each user for the net load at retail rates, and pays each user for the net power at NMRs and for the shared DERs with sharing payments.

In this case, the aggregator and users cooperate as a single interest entity to minimize the total costs. The models are formulated in next section.

5.4 Aggregation of Distributed Energy Resources …

129

5.4.2 System Model In this section, we firstly establish the models for the aggregator and users without energy sharing. Then the energy sharing model is formulated to evaluate the sharing benefits.

5.4.2.1

The Model Without Energy Sharing

As shown in Eqs. (5.73) and (5.74), the aggregator is responsible for paying the electricity and reliability charges. As the charges are recovered by retail sales to energy users, the aggregator’s net cost f A is expressed as follows: f = A



 γs

CsR

+

CsE

−

s



NL λRt Pi,s,t

i,t

+



 NS λSt Pi,s,t

,

(5.75)

i,t

where  

    NL NS NS NS NL Pi,s,t Pi,s,t = − Pi,s,t , Ps,t = − Pi,s,t , ∀s, t.

NL Ps,t

i

+

i

(5.76)

+

In (5.75), the aggregator’s net cost consists of the reliability charge is CsR , the E are electricity R NL cost with the power grid is Cs , the profits from retail sales to Susers NS λt Pi,s,t and the costs of purchasing surplus energy from users are λt Pi,s,t . In i,t

i,t

(5.76), (·)+ = max{0, ·}. Note that the aggregator is assumed to be a price-taker in the energy and capacity markets, and thus the market prices λCd and λEs,t are predefined constants. In the case without energy sharing, each user aims to minimize individual costs. We model each user as an agent solving the following stochastic program for optimal sizing and scheduling for DERs: min fi U = ciPV QiPV + ciBS QiBS + Xi

   L  NL NS , γs λRt Pi,s,t − λSt Pi,s,t − Ui Pi,s,t

(5.77)

s,t

subject to NL NS L PV cha dis Pi,s,t − Pi,s,t = Pi,s,t − Pi,s,t + Pi,s,t − Pi,s,t , ∀s, t,



(5.78)

 NL NS L PV cha dis BS QiPV , QiBS , Pi,s,t , Pi,s,t , Pi,s,t , Pi,s,t , Pi,s,t , Pi,s,t , Ei,s,t , ∀s, t = Xi ∈ χi , (5.79)

130

5 Sharing Economy for Renewable Energy Aggregation

where the decision variables are denoted by Xi , including the capacity of a solar NL NS , surplus power Pi,s,t , the electrical panel QiPV , battery storage QiBS , the net load Pi,s,t L PV cha dis and Pi,s,t , load Pi,s,t , the solar power Pi,s,t , the charging and discharging power Pi,s,t BS and the stored energy in the battery storage Ei,s,t . In (5.77), the objective of each energy user is to minimize the difference between his or her total costs and utility.  L is a quadratic concave function [11], related to his User i’s utility function Ui Pi,s,t or her electric load. Equation (5.78) shows that a user’s net power is equal to the difference between his or her load and the solar and battery power. In (5.79), a user’s constraint set is denoted by χi , including constraints as described in Sect. 5.3.2.2. In the model (5.77)–(5.79), each user minimizes individual costs and trades only with the aggregator without energy sharing. User i’s optimal cost without energy sharing is denoted by fi U ,0 . The aggregator can then collect all the users’ net load, and trade with the power grid. The aggregator’s cost is denoted by f A,0 .

5.4.2.2

The Energy Sharing Model

In the proposed energy sharing scheme, energy users share DERs with the aggregator. The aggregator can then employ the shared power for peak shaving. To share DERs with others, an energy user has to deviate from the optimal individual schedule. Thus, the aggregator should provide incentives to the users who share DERs. The payment from the aggregator to user i is πiES , and user i’s net cost is fi U − πiES . The A aggregator’s net cost is f + πiES . The energy sharing model minimizes the total i

costs of the aggregator and all users: fA+

min

ES Xi ,Pi,s,t



πiES +

i

   fi U − πiES = f A + fi U , i

(5.80)

i

subject to (5.76), NL NS ES L PV cha dis − Pi,s,t − Pi,s,t = Pi,s,t − Pi,s,t + Pi,s,t − Pi,s,t , ∀i, s, t, Pi,s,t



ES Pi,s,t = 0 : λES s,t , ∀s, t,

(5.81) (5.82)

i C ES C ≤ Pi,s,t ≤ Pi,max , ∀i, s, t, Xi ∈ χi , −Pi,max

(5.83)

  ES where the decision variables are Xi , Pi,s,t , ∀i, s, t . In contrast to (5.78), the decision ES variables for sharing DERs Pi,s,t are added in the users’ power balance constraints (5.81). Equation (5.82) is the constraint for balancing shared power. The Lagrangian multiplier of Eq. (5.82) is denoted by λES s,t , which is the clearing price for shared U ,1 A,1 power. Let f and fi be the aggregator’s and user i’s optimal net cost with energy

5.4 Aggregation of Distributed Energy Resources …

131

sharing, respectively, by solving the model (5.80)–(5.83). The total benefits brought on by energy sharing are denoted by f , known as the cooperative surplus: f = f A,0 − f A,1 +

   U ,0 fi − fi U ,1 ≥ 0.

(5.84)

i

As shown in (5.80), the proposed model cannot reveal the payments to users that incentivize users to deviate from individual optimum for energy sharing. Thus, an incentive mechanism is proposed for benefit allocation.

5.4.3 Profit Sharing Mechanism A well-designed incentive mechanism should identify different users’ contributions and allocate the associated benefits to users. In this section, we propose a sharing contribution rate to quantify different users’ contributions to both energy sharing and peak shaving. Based on the indices, an asymmetric NB model is formulated for benefit allocation.

5.4.3.1

Sharing Contribution Rate

An energy user’s contributions to energy sharing are twofold. On the one hand, the user deviates from the optimal individual schedule to provide or consume shared power, which helps decrease the electricity costs of the aggregator. On the other hand, the user curtails electrical load or increases DERs during peak hours, which helps reduce the aggregator’s reliability charges. Therefore, our proposed SCR considers these two aspects. In (5.82), λES s,t is the clearing price for shared power, which indicates the economic value of per-kWh shared energy. Therefore, we define user i’s contribution of shared DERs in time t and scenario s as follows:    ES,1  ES (5.85) = γs λES Ci,s,t s,t Pi,s,t , ES,1 where Pi,s,t is user i’s shared power in time t and scenario s from the solution to    ES,1  ES is measured in the model (5.80)–(5.83); Pi,s,t  is the amount of shared power; Ci,s,t dollars, interpreted as the economic value of user i’s shared DERs. represent the peak hours of the aggregator’s load in scenario s. User i’s Let PH s contribution to peak shaving in scenario s is defined as follows:

  NL,0 NS,0 NL,1 NS,1  PS = γs λCs Pi,s,t − Pi,s,t − Pi,s,t + Pi,s,t Ci,s 

t∈PH s

,

(5.86)

132

5 Sharing Economy for Renewable Energy Aggregation

NL,0 NS,0 NL,1 NS,1 where Pi,s,t −Pi,s,t represents user i’s individual optimal net load, and Pi,s,t −Pi,s,t is user i’s optimal net load after energy sharing. The difference between these two PS is terms represents the aggregator’s peak load reduction contributed by user i. Ci,s also measured in dollars, interpreted as the economic value of user i’s peak shaving. Therefore, we define user i’s contribution rate as his or her contributions over the total contributions of all energy users:

  PS ES    s Ci,s + t Ci,s,t + A  , SCRi = 1 − τ   PS ES C + C j,s j s t j,s,t

(5.87)

+

where τ A ∈ (0, 1) is the aggregator’s rate of return, which is a predefined constant. In practice, τ A can be regulated by the government or optimally determined in a marketbased environment. In (5.87), the numerator term represents user i’s total contributions to energy sharing and peak shaving, and the denominator is all users’ contributions. Note that, each user’s total contribution takes a positive value to guarantee that an SCR is nonnegative. In this section, the users’ economic values of energy sharing and peak shaving are naturally chosen to measure their relative contributions. There are alternative characterizations for proportional cost sharing mechanisms, which need an in-depth study in the future.

5.4.3.2

Mechanism Design

Nash bargaining theory studies how market participants share a surplus that they jointly generate by maximizing the product of market participants’ excess utilities, which fulfills the properties like those illustrated in Sect. 5.2.3. In contrast to the symmetric NB model, the objective of the proposed asymmetric NB model is to maximize the product of market participants’ excess utilities with different sharing contribution rates, as shown in (5.88):  max πiES

f

A,0

−f

A,1

−

 i

τ A πiES

SCRi

 U ,0 fi − fi U ,1 + πiES ,

(5.88)

i∈ES

subject to f A,0 − f A,1 −



πiES ≥ 0,

(5.89)

fi U ,0 − fi U ,1 + πiES ≥ 0, ∀i,

(5.90)

i

5.4 Aggregation of Distributed Energy Resources …

133

  where the decision variables are πiES , ∀i . Constraints (5.89) and (5.90) indicate individual rationality of the aggregator and users, respectively, which means that the aggregator and users can reduce costs after energy sharing. Note that in the objective function (5.88), the exponent of each user is his/her SCR, reflecting different users’ contributions to energy sharing. However, the exponents in the symmetric NB model are identical for different users. Then we propose the following theorem: The optimal payment to user i is: πiES∗ = fi U ,1 − fi U ,0 + SCRi · f , ∀i,

(5.91)

where πiES∗ is the optimal payment to user i, i.e., the solution to the proposed asymmetric NB model. The proposed incentive mechanism satisfies the aforementioned four properties, i.e., (i) Pareto optimality, (ii) budget balance, (iii) individual rationality and (iv) monotonicity. The proof is the similar as that in Sect. 2.5.4. As one can observe, the incentive mechanism defines the optimal payment from the aggregator to each user. In (5.91), the optimal payment πiES∗ consists of two parts: The first term fi U ,1 − fi U ,0 indicates the incremental cost of user i after deviating from individual optimum for energy sharing, and the second term SCRi · f represents the allocation of the total benefits from energy sharing, which is related to user i’s contributions to energy sharing.

5.5 Case Studies 5.5.1 Aggregation of Wind Farms and Concentrating Solar Power We study the optimal joint offering strategy of aggregating three wind power producers and two CSP plants, denoted as WPP1-WPP3 and CSP1-CSP2, respectively. The installed capacity of the wind farms is 600 MW, 300 MW and 300 MW, with forecasting errors equal to 20%, 10% and 20%, respectively. The technical parameters of the CSP plants are listed in Table 5.1. Table 5.1 Parameters of CSP plants

Parameter

Value

Parameter

Value

P Ru , P Rd

40%P max

ηEH /ηe

80%/85.66%

2(2) hours

ηPB

45%

8h

ηc ,

98.5%

Off

On (T Tmin min )

TES capacity

Initial capacity of TES 30%E max Minimum capacity of TES

10%E max

ηd

γ

0.031%

P max

50/100 MW

134

5 Sharing Economy for Renewable Energy Aggregation

Fig. 5.6 Relationship between contribution ratio and revenue in DA and balancing markets

Compared to the traditional symmetric Nash bargaining method, in which all participants have identical bargaining power without distinguishing their individual contribution to aggregation, the profit sharing scheme proposed in this chapter can distinguish the bargaining power of participants in aggregation with different contribution ratios, considering both the day-ahead and balancing market. Figure 5.6 shows the contribution in the day-ahead and balancing market of wind power producers and CSP plants in aggregation, and the size of each circle indicates the contribution ratio in the joint offering of each participant. As one can observe, in the proposed profit sharing scheme, the contribution ratios are positively related to the offered energy in the day-ahead market and the expected dispatchable ability in imbalance settlement. In another word, a participant can obtain a higher contribution ratio by offering more energy in the day-ahead market or accommodating greater uncertainty in real-time. All participants’ contribution ratios equal 0.2 in the traditional Nash bargaining method, regardless of the distinct market performances. Therefore, the proposed profit sharing scheme can identify the contribution of different participants in aggregation and then allocate the extra profits fairly. For example, WPP2 and WPP3 have the same capacity, so their expected profits in the day-ahead market are similar. However, as a result of higher prediction error, the expected revenue in the real-time imbalance settlement of WPP3 is lower, which leads to a smaller contribution ratio compared with that of WPP2. Based on the above contribution ratio, the extra profit earnings of participants in aggregation, resulting from the SBN and the proposed profit sharing scheme, are shown in Fig. 5.7. Note that the extra profit in SNB of all participants is the same, while the extra profits in the proposed scheme (ANB) are related to the contribution ratio of participants, which is defined as the proportion of expected economic values of the offering strategy.

5.5 Case Studies

135

Fig. 5.7 Extra profits of the participants in renewable aggregation

5.5.2 Aggregation of DERs in Energy Markets The quadratic and linear coefficients of users’ utility functions are randomly generated from uniform distributions, i.e.,ai ∈ U [−0.5, −0.1], bi ∈ U [20, 50]. In the 10-user case, the load and solar data are described in Fig. 5.8. Users’ minimal and maximal loads are set to 0.8 and 1.2 times of the actual loads. Users’ daily minimum loads are set to his/her actual daily load demand. The parameters of users’ ESSs are listed in Table 5.2. The initial stored energy is randomly

Daily load Daily solar energy Peak load

Load and solar power (MWh)

90 80

8 7

70

6

60

5

50

4

40

3

30

2

20

1

10 0

Peak load (MW)

100

1

2

3

4

5

6

7

8

9

10

0

User index

Fig. 5.8 Users’ load and solar power data

Table 5.2 Parameters of users’ energy storage systems ESS Pi,t,s,α/β

ηiESS

ESS Ei,min

ESS Ei,max

ciESS

5 MW

95%

5 MWh

30 MWh

3.7 $/MWh

5 Sharing Economy for Renewable Energy Aggregation 5 4 3 2 1 0 -1 -2 -3 -4 -5

ESS PV Load Net load

Power (MW)

Power (MW)

136

2 4 6 8 10 12 14 16 18 20 22 24

Hour (h)

(a) W/O sharing

5 4 3 2 1 0 -1 -2 -3 -4 -5

ESS PV Load Net load

2

4

6

8 10 12 14 16 18 20 22 24

Hour (h)

(b) Energy sharing

Fig. 5.9 A user’s power profiles without and with energy sharing

generated from U [5, 5] MWh. The ESS’s operation cost is estimated as the replacement cost of storage bank over the lifetime throughput. The replacement cost per year is about 13,400 $/MW, and the yearly charging/discharging hours are 1800 h. Therefore, the operation cost ciESS is expressed as follows: ciESS =

1800



13400  = 3.7 $/MWh. + 1/ηiESS

ηiESS

(5.92)

An energy user’s average power curves without and with energy sharing are compared in Fig. 5.9. In the case without energy sharing, the user has no incentives to arbitrage with the ESS against retail rates. The ESS is only used to store the surplus of solar energy, and discharge to satisfy the night-time load. Due to the low NMR, the user will minimize his/her net load instead of selling back the surplus power. However, in the case with energy sharing, the utilization of the ESS is greatly improved for shifting the day-time load to the night hours. As the net load shows, the user consumes electricity at night but supplies DERs at day time. In addition, by comparing the electrical load curves, one can observe that the load is also shifted from day time to night time after energy sharing. Figure 5.10 shows the average curves of the LMPs and aggregated loads. In the energy sharing scheme, the aggregator organizes users to respond to the LMPs. Com-pared with the case without sharing, energy sharing can provide additional 132.35 MWh power for the power grid during peak hours from 6:00 to 20:00, thereby contributing to power balance in the power grid. In SNB, the symmetric Nash bargaining model is used to allocate the sharing benefits. All users have identical weights, without distinguishing the users’ contributions to sharing. However, the proposed ANB identifies users’ contributions for benefit allocation. The cost savings of users by ANB and SNB are shown in Fig. 5.11. As one can observe, the cost savings of all users are 0.49 thousand dollars in SNB. However, the cost savings range from 0.44 to 0.53 thousand dollars in ANB.

5.5 Case Studies

137 W/O sharing Energy sharing LMP

60 50

Power increment 50 40

30 30

20 10

20

Price ($/MWh)

Load (MWh)

40

0 10

-10 -20 4

8

12

Hour (h)

16

20

24

0

Cost saving (thousand $)

Fig. 5.10 Average profiles of the LMPs and aggregated loads 0.55

SNB ANB 0.5

0.45

0.4 1

2

3

4

5

6

7

8

9

10

User index

Fig. 5.11 Cost savings of users by ANB and SNB

As aforementioned, a user’s cost savings are related to his/her SCR, defined as the user’s proportion of the economic values of shared DERs. Figure 5.12 shows the relationship between users’ SCRs and shared DERs. Note that a user’s shared DERs refers to the total amount of the absolute value. In ANB, users’ SCRs are positively related to the shared DERs, indicating that the more DERs a user shares, the higher level of contributions this user makes. However, all users’ SCRs equal 0.08 regardless of the distinct behaviors in SNB. Therefore, the pro-posed ANB can reveal the contributions of different users and then allocate the benefits.

5.5.3 Aggregation of DERs in Energy and Capacity Markets The quadratic and linear coefficients of users’ utility functions are randomly generated from uniform distributions, i.e., ai ∈ U [−0.5, −0.1], bi ∈ U [20, 50], ∀i. Users’

138

5 Sharing Economy for Renewable Energy Aggregation

Fig. 5.12 Relationship between users’ SCRs and shared DERs

SNB ANB

SCR

0.085

0.080

0.075

0.070

74

76

78

80

82

84

86

88

90

Shared DERs (MWh)

1

1

0.8

0.8

0.6

0.6

CDF

Solar power (MWh)

minimal and maximal loads are set to 0.9 and 1.1 times the actual loads for each time slot. Users’ daily minimum load requirements are set to his or her actual daily load demand. The daily solar production of a 1-MW solar panel is shown in Fig. 5.13. The parameters of a 1-MW battery storage system are listed in Table 5.3. The annualized investment costs for a 1-MW solar panel and 1-MW battery storage system are $240,000 and $80,000, respectively. PG&E summer retail prices are 0.212 $/kWh from 1:00 to 8:00 and from 22:00 to 24:00, 0.239 $/kWh from 8:00 to 12:00 and from 18:00 to 22:00, and 0.263 $/kWh from 12:00 to 18:00. The net metering rate is 0.03 $/kWh. The day-ahead energy market prices and the capacity market prices are provided by PJM [13]. The average

0.4

0.2

0.2 0 0

0.4

5

10

15

Hour (h)

20

25

0 0

2

4

6

Solar production (MWh)

Fig. 5.13 Daily solar production of a 1-MW solar panel

Table 5.3 Parameters of a 1-MW batter storage system

Parameter

BS Pi,max

BS Ei,max

ηiBS

Value

1 MW

2.7 MWh

90%

8

5.5 Case Studies

139

zonal capacity market price for the 2020–2021 BRA is 120.98 $/MW-day, and the 2017 yearly LMPs are shown in Fig. 5.14. The optimal capacities for solar and battery storage in the cases without and with energy sharing are shown in Fig. 5.15. Compared with the case without energy sharing, the optimal capacities for solar and storage significantly increase after energy sharing. The total solar capacity increases from 23.99 to 91.16 MW, and the total storage capacity increases from 0.89 to 4.17 MW. These results conclude that the proposed energy sharing scheme effectively incentivizes users to invest in DERs. Note that 10 users have different 350

LMP ($/MWh)

300

250

200

150

100

50 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Hour (h)

Solar capacity (MW)

Fig. 5.14 Yearly LMPs at a bus in PJM

10

5

0

1

2

3

4

5

6

7

8

9

10

Storage capacity (MW)

User index

W/O sharing Energy sharing

0.8 0.6 0.4 0.2 0

1

2

3

4

5

6

7

8

9

10

User index

Fig. 5.15 Optimal capacities for solar and battery storage without and with energy sharing

140

5 Sharing Economy for Renewable Energy Aggregation

Fig. 5.16 Aggregator’s average net loads without and with energy sharing

20

Net load (MW)

10

0

-10

-20

-30

-40

W/O Sharing Energy Sharing 2

4

6

8

10

12

14

Hour (h)

16

18

20

22

24

utility levels and electrical load profiles, leading to different investment decisions. For example, the optimal storage capacity of User 6 decreases after energy sharing, indicating that User 6 prefers to consume shared energy by others compared with individual investments. The aggregator’s average net power profiles without and with energy sharing are shown in Fig. 5.16. With more solar and storage integrated, the aggregator’s peak load is reduced by 17.65%, from 20.11 to 16.56 MW. Thus, the annual reliability charges for the aggregator decrease from 0.89 to 0.73 million dollars. In addition, with more solar resources installed, the aggregator can provide the power grid with net power during peak hours from 8:00 to 17:00. The aggregator annually sells 87.75 GWh of electricity back to the power grid. In SNB, all users  are weighted identically, and thus each user’s contribution rate in SNB is equal to 1 − τ A /N = 0.09. However, the proposed ANB model can identify different users’ contributions to energy sharing and peak shaving. The users’ SCRs and cost savings in ANB and SNB are compared in Fig. 5.17. Compared with NS, all users can reduce the costs after participating in energy sharing. In SNB, as all users are weighted identically, each user’s annual cost savings equal 0.13 million dollars. However, the proposed ANB method can identify the contributions of different users, so the users’ cost savings and SCRs are distinguishable. In ANB, users can reduce their costs by 0.09–0.18 million dollars.

5.6 Conclusion This chapter introduces the profit sharing mechanisms for renewable aggregation, which allocate the jointly created benefits according to the different contribution rates of the aggregation participants.

5.6 Conclusion

141

SCR

0.1 0.05

Annual cost saving (million $)

0

1

2

3

4

5

6

7

8

9

10

User index

SNB ANB

0.2 0.15 0.1 0.05 0

1

2

3

4

5

6

7

8

9

10

User index

Fig. 5.17 Users’ SCRs and cost savings in ANB and SNB

An optimal joint offering strategy in wholesale markets is proposed for the aggregation of wind power producers and CSP plants. An incentive mechanism based on ANB is designed for profit sharing. To avoid violating the privacy of individual operations, a decentralized framework is developed for practical implementation. Case studies validate the effectiveness of the joint offering strategy and profit sharing mechanism. In contrast to the traditional SNB model, the proposed profit sharing scheme can identify different contributions of wind producers and CSP plants in the joint offering. For the aggregation of DERs, a novel energy sharing scheme is proposed to enable the peer-to-peer trading among customers. In the energy sharing scheme, an aggregator organizes users to cooperate as a single interest entity, and an incentive mechanism based on ANB is designed for profit sharing according to users’ contributions. Case studies demonstrate that in contrast to the case without energy sharing, the utilization of in-home energy storage can be dramatically improved, and the joint benefits of the aggregator and all users increase significantly. In addition, users can be incentivized by the extra profits earned by energy sharing, thus investing in more solar and storage. Therefore, an energy sharing scheme is designed for DER aggregation in joint energy and capacity markets. A concept of sharing contribution rate is designed to measure the merit of a DER for energy trading and peak shaving. Case studies show that the propose energy sharing scheme can effectively incentivize the investments in DERs while making full use toward grid interaction. Meanwhile, the ANB-based mechanism can distinguish the users’ specific contributions, thus achieving fair benefit allocation.

142

5 Sharing Economy for Renewable Energy Aggregation

References 1. China strives to achieve carbon neutrality by 2060. http://www.tanpaifang.com/tanzhonghe/ 2020/0923/74144.html. Accessed 13 Oct 2020 2. The International Renewable Energy Agency publishes a new version of the Global Energy Transition: Roadmap to 2050. http://www.casisd.cn/zkcg/ydkb/kjzcyzxkb/kjzczxkb2019/kjz czxkb201906/201908/t20190830_5373985.html. Accessed 30 Aug 2019 3. Kroposki, B., Johnson, B., Zhang, Y., et al.: Achieving a 100% renewable grid: operating electric power systems with extremely high levels of variable renewable energy. IEEE Power Energy Mag. 15(2), 61–73 (2017) 4. Hu, J., Yang, G., Kok, K., et al.: Transactive control: a framework for operating power systems characterized by high penetration of distribution energy resources. Mod. Power Syst. Clean Energy 5(3), 451–464 (2017) 5. Morales, J.M., Conejo, A.J., Pérez-Ruiz, J.: Short-term trading for a wind power producer. IEEE Trans. Power Syst. 25(1), 554–564 (2010) 6. The Pacific Gas and Electric Company website: https://www.pge.com/en_US/residential/solarand-vehicles/. Accessed 13 Oct 2020 7. Yan, G., Liu, D., Li, J., et al.: A cost accounting method of the Li-ion battery energy storage system for frequency regulation considering the effect of life degradation. Prot. Control Mod. Power Syst. 3(4), 1–9 (2018) 8. Wang, H., Huang, J.: Incentivizing energy trading for interconnected microgrids. IEEE Trans. Smart Grid 9(4), 2647–2657 (2018) 9. Nguyen, H.K., Khodaei, A., Han, Z.: Incentive mechanism design for integrated microgrids in peak ramp minimization problem. IEEE Trans. Smart Grid (early access) (2018) 10. Nash, J.F.: The bargaining problem. Econometrica 18(2), 155–162 (1950) 11. Boyd, S., Parikh, N., Chu, E., et al.: Distributed optimization and statistical learning via the alternating direction method of multipliers. Found. Trends Mach. Learn. 3(1), 1–122 (2011) 12. Zhao, F., Zheng, T., Litvinov, E.: Constructing demand curves in forward capacity markets. IEEE Trans. Power Syst. 33(1), 525–535 (2018) 13. PJM Capacity Market Operations: http://www.pjm.com/markets-and-operations/rpm.aspx. Accessed 13 Oct 2020

Chapter 6

Sharing Economy in Energy Systems Integration

6.1 Introduction It has become an inevitable choice to develop integrated energy technologies involving the coupling and complementarity of coal, oil, gas, conventional and renewable energy toward a low-carbon, reliable and efficient energy system transition. The Chinese government has declared that it is essential to construct “Internet+” smart energy and promote the interconnection and transformation of various forms of energy [1]. In May 2020, the government further stated to establish an electric power system-centered energy internet [2]. At present, China has established integrated energy service companies in 26 provinces and cities, and is actively promoting integrated energy technology innovation and industry development. The essence of energy systems integration is to achieve an interaction of energy networks and “Internet+” toward the construction of smart energy systems; to realize the coupling and complementarity of heterogeneous forms of energy for economy and reliability enhancement; to establish an efficient and economic multi-energy markets to enable the trading among different energy providers. The economic significance of such “integration” lies in the prospect that it forms a flexible elasticity between supply and demand due to the substitution of different energy resources, while realizing a Pareto efficiency improvement on a larger domain than independent market operation. In recent years, it has become a major strategic competition focus of different countries to develop integrated energy technology and industry. A wide variety of hightechs have emerged in this research area, and relevant pilot projects have been initiated. Admittedly, the development of integrated energy technologies worldwide has achieved remarkable results, and the industrial benefits are considerable. However, the development of the industry depends not only on the technology, but also upon the effectiveness of market environment and mechanism. Technological innovation is the outcome of a well-designed market, while market mechanisms are the driving force of technology. Therefore, it is of vital importance to propose and operate an

© Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_6

143

144

6 Sharing Economy in Energy Systems Integration

effective profit sharing mechanism for a high-quality development of energy systems integration. The existing research generally regards an integrated energy market as a centralized optimization problem, in which a universal integrated market operator is conceived to organize all the energy markets including electricity, natural gas and heat markets. However, it is hardly possible to establish such a centralized universal market in practice. One reason is that the government has to carry out a huge project for integrated market reform, which is extremely costly to develop an integrated market operator on top of the existing ones, e.g., independent system operators in electricity markets. Secondly, what the integrated market operator confronts is a large-scale optimization program for integrated market clearing and settlement, which may be time-consuming and difficult for market regulation and supervision. Another reason lies in information asymmetry among the existing energy markets. Constructing a higher level of integrated markets indicates that the trading and information barriers among different markets have to be broken through. In this instance, a universal integrated market may not be realistic in the existing market environment. In order to solve the above-mentioned problems, we propose an equilibrium-based multi-energy market framework without violating the rules of each individual market. Motivated by the price difference among multi-energy markets, four types of arbitrage including inter-commodity, inter-exchange, calendar spread and future-spot are proposed. Based on the concept of multi-energy markets, we formulate the sharing economy models in electricity-heat, electricity-gas/hydrogen and transportationenergy systems, while quantifying the associated sharing benefits. In addition, a novel vision termed as integrated demand response (IDR) is studied in this chapter.

6.2 Integrated Energy Sharing Market 6.2.1 Status and Challenge 6.2.1.1

Global Overview

In recent years, energy systems integration technologies have been developing rapidly across the world. A wide variety of the countries are actively promoting the construction of integrated energy sharing markets, attracting extensive attention from industry and academia. The United States is committed to establishing a smart grid-centered integrated energy market. The research and development (R&D) of distributed energy resource and combined cooling, heating and power technologies has been a major task, aiming to achieve the dependency of various forms of energy. Europe initiated the R&D of integrated energy technology in an earlier year. The Great Britain has focused on the coupled market of electricity and natural gas, and Germany has been working on the construction of a cyber-physical energy system toward efficient trading. Due to a limited and constrained natural resources, Japan has long

6.2 Integrated Energy Sharing Market

145

been focusing on the improvement of energy conservation and emission reduction system driven by integrated energy technology. Some 100% renewable-powered communities have been built with distributed generation and multi-energy microgrid supported.

6.2.1.2

Challenges

The existing studies and pilot projects across the world have focused on the construction of a centralized energy systems integration, in which a universal market operator takes charge of all different markets. As shown in Fig. 6.1, a centralized integrated energy market requires market participants, e.g., coal-fired generators, renewable producers and integrated energy service providers, to submit the detailed individual constraints and bids. By this means, an optimization program for market clearing and settlement can be solved. However, some challenges remain to be addressed: (1)

The reform cost of a centralized integrated energy market In practice, different energy markets have long been independent, and been governed by individual market operators. There have been trade barriers between energy industries, and thus it would be extremely costly to break such barriers to establish a universal market on top of the existing ones. For example, the electricity markets in the U.S. are scheduled by Independent System Operator (ISO)/Regional Transmission Organization (RTO), while the natural gas markets are operated by locational pipeline and retail companies. It is hardly possible to abolish the existing market frameworks and establish a higher-level integrated energy market.

Universal market operator

Installed capacity Minimum output ramping

GenCo

Forecast power Frequency regulation

Quantity

Consumer

Renewable

Quantity

Electricity-gas coupling: power-to-hydrogen Electricity-heat coupling: heat pump Gas-heat coupling: combined heating and power

Integrated energy service provider Price

Price

Price

Price

Load shedding Peak shaving

Quantity

Quantity

Fig. 6.1 Illustration of a centralized optimization-based integrated energy market

146

6 Sharing Economy in Energy Systems Integration

(2)

The information asymmetry of different energy markets The existing energy markets have formed each mature information disclosure mechanism. However, a universal integrated energy market would require a wide information exchange. As shown in Fig. 6.1, coal-fired generators have to submit the technical parameters to an integrated energy market operator. Integrated energy service providers should report individual detailed energy conversion units. Therefore, it is difficult to break through the information asymmetry among different markets. Additionally, some other types of transaction costs can be induced. For example, it is time-consuming to solve a large-scale centralized market clearing optimization. The market regulation and supervision can be hard to be implemented. These transaction costs can lead to a severe Pareto efficiency deviation.

6.2.2 Market Operation Practice At present, the market operation of electricity, heat and gas generally takes other energy systems as boundary conditions. Each energy market is managed by individual operator. It should be noted that the operation and mechanism of different energy markets can be quite diversified.

6.2.2.1

Electricity Market

In order to guarantee power supply security, electricity markets are organized and implemented in a sequential transaction fashion, including medium/long-term market, day-ahead market, real-time market, etc. For example, the bid of a day-ahead market is an hourly market share during the delivery day. An ISO cleares the market on the basis of participants’ bids, and make settlements via locational marginal prices. In addition, different regions may have specific market mechanisms, e.g., Nord Pool and PJM.

6.2.2.2

Natural Gas Market

The marketization and spatial scale of natural gas markets lie in the middle of electricity and heat markets. Most of the natural gas markets are based on medium and long-term contracts. Natural gas was originally developed as a substitute for coal, fuel oil and other traditional fossil fuels. The international standard pricing of natural gas keeps a high correlation with the prices of oil and coal. Specifically, there are four types of representative pricing mechanisms. North America and the Great Britain adopt a market pricing mechanism, regarding the natural gas future price in Henry

6.2 Integrated Energy Sharing Market

147

Hub and National Balancing Point (NBP) as a benchmark, respectively. The Continental Europe uses the oil-linked natural gas price, and Japan relies on the connection of imported oil price. Currently, the natural gas price in China is determined by the National Development and Reform Commission with a low degree of marketization. There exists a general imbalance in the pricing of natural gas industry chains, and the pricing is based on the principle of cost recovery, which is difficult to reveal the supply–demand relationship. In recent years, Shanghai, Chongqing and some other regions have set up natural gas trading centers, aiming to introduce marketization while weakening centrally-planned economy.

6.2.2.3

Heat Market

In contrast to electricity that can be transmitted over a large range, a heat market is restricted by limited heating radius and distribution losses. Thus, a heat market is generally localized. The existing heat markets across the world are not based on bilateral bidding, but operated and planned by a regional energy regulation department. In other words, the degree of marketization is relatively low compared with electricity and natural gas markets. The current major pricing mechanisms rely on two types, i.e., heating quantity and heating area. Heating charge is still based on the principle of cost recovery rather than profit maximization. The centralized bidding framework has not yet been established in practice. However, some recent studies have noticed this issue and explored the improvement strategy of district heating pricing mechanism.

6.2.3 Equilibrium-Based Integrated Energy Market In recent years, the development of energy conversion technologies, e.g., powerto-gas, combined cooling, heating and power, etc., has strengthened the coupling between multi-energy markets. However, there are natural information asymmetry and trade barriers between different energy markets, which may lead to a certain degree of market failure. In order to avoid market risks and improve market operation efficiency, we propose an equilibrium-based integrated energy market that enables the energy sharing among various forms of energy. As shown in Fig. 6.2, we do not expect to change the market operation and rules of each energy market. Every market operator maintains the original energy trading, while the coupling of different markets can be achieved by the arbitrage of the integrated energy service providers with energy conversion abilities. Instead of a centralized integrated energy market, the energy trading is driven by the price difference among different markets. Integrated energy service providers can fully explore individual arbitrage capability for profit maximization, thus leading to the convergence of different market prices.

148

6 Sharing Economy in Energy Systems Integration

Here we claim that information disclosure will play a critical role to guide the decision making of market participants, e.g., the clearing prices and quantities of different markets. The market information can be divided into four types: (i) Public information, e.g., the clearing prices of electricity, natural gas and heat, which refers to the data accessed by public. (ii) Published information, e.g., load forecasting, bidding categories, market shares, etc., which indicates the data that can be published to market participants. (iii) Private information, e.g., long-term contracts, market bids, etc., which can be accessed by specific participants and should not be published without admission. (iv) Classified information, e.g., network and topology of energy systems, critical load, which can only be accessed according to legal requirements. In essence, the proposed integrated energy market does not change the existing independent mechanisms of each energy market, but approaches an equilibrium in a step-by-step way. Driven by the price difference, rational service suppliers can maximize individual profits by arbitrage while mitigating the imbalance of energy resource allocation, which reveals the incentive compatibility of mechanism design. On the hand, energy users are given opportunities to flexibly change the source of energy supply, which significantly increases load elasticity and thus forms demand response over various forms of energy. Electricity market

Power-to-gas Longterm

Secondary

Dayahead

Price

Intraday

Real-time

Ancillary service Contract

Quantity

Heat pump

Right trading

Heat market

…

Storage

Cost recovery Marginal price

Information disclosure

CCHP Local distribution network

Provider Natural gas market

Big consumer

CCHP Production

Pipeline

Information Fund Energy

End-use

Fuel cell

CCHP

Fig. 6.2 Illustration of an integrated energy market system

Chemical industry

Agent

6.2 Integrated Energy Sharing Market

149

6.2.4 Arbitrage Models for Integrated Energy Markets The free choice of consumers is a major requirement of market mechanisms. In the future integrated energy market, consumers are given the right to freely choose energy categories, and the potential arbitrage behavior will effectively restrain the price deviations between different trading types and markets in different time scales. Arbitrage, originated from international financial markets, refers to a trading method where market participants take advantage of the difference of short-term spot interest rates in different countries or regions to obtain interest spread income. In integrated energy markets, the difference of energy prices can encourage energy service providers to participate in arbitrage. We propose four types of arbitrage including inter-commodity, inter-exchange, calendar spread and future-spot, as shown in Fig. 6.3. (1)

Inter-commodity arbitrage

Inter-commodity arbitrage refers to the arbitrage by taking advantage of the price gap between two commodity markets when there exists mutual substitution between the commodities. In integrated energy markets, for example, when natural gas prices are low, providers can purchase natural gas and convert the gas into electricity or heat via combined heating and power, which can be sold in the associated markets. Therefore, the price divergence can be suppressed between different energy markets, thus promoting efficient competition among different energy resources.

Year

Month

Day ahead

Real time

Energy finance market

Integrated energy service provider Inter-commodity arbitrage

Inter-exchange Calendar spread Future - spot arbitrage arbitrage arbitrage

Gas Electricity Heat Energy spot market

Fig. 6.3 Diagram of arbitrage and trading in integrated energy markets

150

(2)

6 Sharing Economy in Energy Systems Integration

Inter-exchange arbitrage

Inter-exchange arbitrage refers to the arbitrage in which when the prices of the same commodity in two regions or markets are different, participants can make use of the price gap to earn profits. In integrated energy markets, for example, when the price of natural gas in other regions is lower than that of local gas, providers can purchase natural gas from remote regions while satisfying local demand so as to reduce the spatial price deviation. (3)

Calendar spread arbitrage

Calendar spread arbitrage refers to that when commodity prices differ in different time scales, market participants can forecast the price diviation for arbitrage by buying commodities at a low price and selling them at a high price. For instance, when the long-term natural gas price is lower than the spot gas price, providers can purchase an amount of long-term contracts and sell them in the spot markets. Some other examples include the arbitrage between day-ahead and real-time markets. (4)

Future-spot arbitrage

Future-spot arbitrage refers to the cases where there is a price gap between financial and physical markets. In integrated energy markets, providers can take advantage of the price difference between electricity future and spot markets for arbitrage, while suppressing the divergence of financial derivatives and energy resources.

6.3 Sharing Economy in Joint Electricity-Heat Markets In response to severe environmental challenges and energy shortages caused by the massive consumption of fossil fuels, integrating renewable and distributed energy resources has become a global consensus. Electricity and heat are the major energy usages in both the industrial and residential sectors. For example, approximately 1410 billion kWh of electricity was consumed by the residential sector in the United States in 2016, accounting for 24% of total residential energy consumption. Over 35% of the consumed electricity was used for heating [3]. The increase in electric and heating loads poses significant challenges to maintaining the reliable and economic operation of a power grid. Therefore, it is necessary to deploy renewable and distributed energy technologies to meet the increasing loads. Combined heat and power (CHP) technology is the use of power plants to simultaneously generate electricity and useful heat, which recovers wasted thermal energy for heating. As an efficient production technology, CHP has been widely adopted in various regions worldwide. In Finland, highly efficient CHP units have gained substantial popularity in producing electricity and heat. In 2007, 74% of district heating energy was generated by CHP units [4]. In Germany, the government aims to increase the country’s CHP production to 25% by 2020 [5]. In China, particularly in the northern provinces, CHP units are commonly used for industrial and residential

6.3 Sharing Economy in Joint Electricity-Heat Markets

151

space heating. The national CHP installed capacity reached 250 GW by the end of 2015 [6]. The integration of SPHPs will contribute to a more sustainable operation of a CHP system. As a promising distributed energy resource (DER), solar-powered heat pumps (SPHPs) can efficiently provide heat energy for consumers using solar power and electricity from the power grid. An SPHP consists of three devices: (1) a solar panel, (2) a heat pump (HP) and (3) a thermal energy storage (TES) system, e.g., a water tank. From the consumer perspective, SPHPs can produce heat energy with a high coefficient of performance (COP). In addition, the surplus of solar generation can be locally consumed to reduce electricity charges. From the perspective of a CHP system, SPHPs can help reduce net electrical and heating loads, decreasing generation costs and improving the flexibility of centralized thermal units. Therefore, it is imperative to fully investigate the grid-friendly benefits of SPHPs. In the previous studies, the economic and environmental benefits of SPHPs have been widely evaluated. However, few studies focus on the sharing market problem of SPHP investments and how electricity and heating prices influence the SPHP planning strategy. With an increased use of SPHPs on the demand side, both electric and heating loads that rely on energy supply from CHP systems will be substantially decreased. Thus, electricity and heating prices will change with the planning strategy for SPHPs, which will further influence user decisions on SPHP investments. Therefore, we attempt to adopt consumers as price-makers in the SPHP planning problem and thoroughly evaluate the benefits of integrating SPHPs into a CHP system in this section.

6.3.1 Framework The schematic of the CHP system with SPHPs is shown in Fig. 6.4. Rooftop solar panels produce electricity for consumers during the daytime and have gained substantial popularity worldwide. By making full use of the electricity from solar panels, HPs can efficiently generate heat for dwellings. As a result, consumers can further justify the investment costs by electricity and heating cost reductions.

PV CHP system

Electric load TES Electricity Heat

HP

Fig. 6.4 The schematic of the CHP system with SPHPs

Heating load Dwelling

152

6 Sharing Economy in Energy Systems Integration

6.3.2 Solar-Powered Heat Pump Planning Model On the upper level of the proposed framework, the consumers located at different nodes strategically invest in SPHPs to maximize individual revenues. In this section, the planning model for SPHPs is established. To depict the uncertainties in the solar power, load demands and ambient temperatures, scenario-based stochastic programming is adopted.

6.3.2.1

Thermal Behavior of a Dwelling

To model the heating load demands of the consumers, the thermal behavior of a dwelling is described. The indoor temperature of consumer i’s dwelling at time slot t in scenario s is determined as follows: CiD

D dTi,t,s

dt

 D  H A = μCi Pi,t,s , ∀t, s ∈ HS − μIi Ti,t,s − Ti,t,s

(6.1)

Equation (6.1) shows that the indoor temperature is related to the heating power supply and the heat dissipation. HS is set of heating scenarios; CiD is heat capacity D is temperature of consumer i’s dwelling at time slot of consumer i’s dwelling; Ti,t,s H t in scenario s; Pi,t,s is heating power injected in consumer i’s dwelling at time slot t A in scenario s; Ti,t,s is ambient temperature of consumer i’s dwelling at time slot t in C scenario s; μi is the energy conversion efficiency; μIi represents the heat insulation level. The smaller μIi is, the better insulation the dwelling has. The discrete form of Eq. (6.1) is derived below:  D  D D H A − CiD Ti,t−1,s − μCi Pi,t,s T = μIi Ti,t,s , ∀t, s ∈ HS Ci +μIi Ti,t,s

(6.2)

The indoor temperature should be maintained within the comfortable level of the consumers, i.e., D D D ≤ Ti,t,s ≤ Ti,max , ∀t, s ∈ HS Ti,min

(6.3)

D D where Ti,min and Ti,max are minimal and maximal temperature of consumer i’s dwelling, respectively. Note that the heating power can be obtained from both the central heat supply of the CHP system and the invested SPHPs.

6.3.2.2

SPHP Planning Model

We assume that a rational consumer makes strategic investments for SPHPs by optimizing the following program:

6.3 Sharing Economy in Joint Electricity-Heat Markets

153

min CiI + CiE + CiH

(6.4)

subject to Eqs. (6.2) and (6.3): N  PV

CiI =

N  HP

ujPV ICjPV +

j=1

ujHP ICjHP +

TES N 

j=1

CiE =



γs



N 



ELD λEB(i),t,s Pi,t,s T

γs



HLD λH D(i),s Pi,t,s T

N  PV

HP,in Pj,t,s

+

LD Pi,t,s

j=1

HLD Pi,t,s

(6.7)

t∈T

HP

=

(6.6)

t∈T

s∈HS

ELD Pi,t,s

(6.5)

j=1

s∈S

CiH =

ujTES ICjTES ≤ ICi,max

=

−

PV Pj,t,s , ∀t, s ∈ S

(6.8)

j=1

H Pi,t,s

−

TES N 

TES Pj,t,s,β , ∀t, s ∈ HS

(6.9)

j=1

ujPV , ujHP , ujTES ∈ {0, 1}, ∀j 

HP,out/in

ELD HLD PV Pi,t,s , Pi,t,s , Pj,t,s , Pj,t,s

 TES TES , Pj,t,s,α/β , Ej,t,s , ∀j, t, s ∈ χi

(6.10) (6.11)

The objective (6.4) is to minimize consumer i’s total costs, including the investments in SPHPs and the electricity and heating costs. CiI ,CiE and CiH are consumer i’s investment cost, electricity cost and heating cost, respectively. The thermal behavior of consumer i’s dwelling is shown in (6.2) and (6.3). Equation (6.5) shows that the investments costs include those for PVs, HPs and TESs. uj(·) is the Binary decision variable of PV/HP/TES j, the value 1 indicates investing in PV/HP/TES j, Otherwise 0; ICj(·) is investment cost of PV/HP/TES j; ICi,max is investment budget of consumer i. In (6.6) and (6.7), the electricity and heating costs equal consumer i’s net load multiplied by the prices.γs is weight of scenario s; λE(·) is locational marginal price; ELD λH (·) is district heating price; Pi,t,s is net electric load of consumer i at time slot t in HLD scenario s; Pi,t,s is net heating load of consumer i at time slot t in scenario s; T is time interval. In (6.8), the net electrical load equals the electrical load and the HP,in LD is input power of heat pump; P(·) is electric load for HPs minus PV output. P(·) PV H load demands; P(·) is solar power; Pi,t,s is heating power injected in consumer i’s TES is discharging power of TES j at time t in scenario s. In (6.9), the dwelling; Pj,t,s,β net heating load equals the difference between the heating load and the TES output. (6.10) and (6.11) are the constraints for decision variables, and χi represents the set of consumer i’s constraints. The LMPs in (6.6) and the DHPs in (6.7) are influenced by the planning strategies of all consumers.

154

6 Sharing Economy in Energy Systems Integration

Therefore, the objective (6.4) and the constraints (6.2), (6.3) and (6.5)–(6.11) form consumer i’s planning model for SPHPs.

6.3.3 Pricing Model 6.3.3.1

LMP Model

DCOPF has been widely used for computing LMPs in existing research and electricity markets. In DCOPF, the power losses are not considered for simplification. ∀s ∈ HS , the DCOPF model in scenario s at time slot t is as follows:  G ciG × Pi,t,s (6.12) min i∈G

subject to 

G Pi,t,s =

i∈G N 



ELD Pi,t,s : λt,s

(6.13)

i∈LD

 G  ELD LN ≤ Pk,max GSDFk−i Pi,t,s − Pi,t,s : μ+ k,t,s , ∀k

(6.14)

 G  ELD LN ≥ −Pk,max GSDFk−i Pi,t,s − Pi,t,s : μ− k,t,s , ∀k

(6.15)

i=1 N  i=1

ηCHP

 i∈CHP x

G θiG Pi,t,s =



HLD Pi,t,s , ∀x

(6.16)

i∈LD x

G G G ≤ Pi,t,s ≤ Pi,max , ∀i Pi,min

(6.17)

The DCOPF model seeks to minimize the operational costs of conventional generators in the power system, subject to electric power balance constraint (6.13), line power flow constraints (6.14) and (6.15), thermal power balance constraint (6.16) G is and the limits for generators output (6.17). ciG is unit cost of thermal unit i; P(·) power of thermal generators; GSDFk−i is generation shifting distribution factor of LN is transmission capacity of line k; ηCHP is heating efficiency line k to node i. Pk,max G G of the CHP system; Pi,min and Pi,max are minimal and maximal power of thermal unit i, respectively. Note that thermal power balance constraints are removed from the DCOPF model for ∀s ∈ S \HS . The Lagrangian multipliers of constraints (6.13)– − (6.17) are denoted by λt,s , μ+ k,t,s and μk,t,s , respectively. Then the LMP at node i in Eq. (6.6) is as follows:

6.3 Sharing Economy in Joint Electricity-Heat Markets

λEi,t,s = λt,s +



  − GSDFk−i μ+ k,t,s + μk,t,s

155

(6.18)

k∈LN

In the DCOPF model, the heating demands are supplied by the CHP units in the power system. Because the thermal power cannot be transmitted over a long distance, the CHP units in district x are required to satisfy the heating demands in the same district, as shown in (6.18). Note that to satisfy the heating loads, the flexibility of a CHP system is reduced, and the operational costs increase. Therefore, it is important to develop SPHPs to locally satisfy heating demand.

6.3.3.2

DHP Model

In recent decades, the district heating pricing problem has been widely investigated. There are two main types of district heating market: regulated and deregulated. Thus, two representative methods are used for DHPs: (i) the cost-plus pricing method and (ii) the marginal-cost pricing method. In this subsection, we adopt the cost-plus pricing method to calculate district heating prices, which is commonly used in regulated markets, e.g., China. In certain regions in the world, e.g., the Netherlands and London, UK, the costs for district heating depend on three main factors: (1) the connection costs for customers, (2) the costs of a distribution network and (3) the production costs of thermal energy. The total heating costs in district x in scenario s can be calculated as: H H = VCx,s + ADxH + PPxH TCx,s

(6.19)

H is the variable heating costs, related to the production of thermal In (6.19), VCx,s energy in the district. The expression is: H = VCx,s

  t∈T

G ciG Pi,t,s

(6.20)

i∈CHP x

ADxH is the annual depreciation and PPxH is the permitted profits. These two costs can be regarded as fixed costs. Thus, the heating price per unit in scenario s in district x is as follows: H TCx,s  = λH x,s HLD Pi,t,s i∈LD x

H where TCx,s is total heating costs in district x in scenario s.

(6.21)

156

6 Sharing Economy in Energy Systems Integration

6.3.4 Solution Algorithm The bi-level model is formulated as a market equilibrium problem because the LMPs and DHPs are influenced by the planning strategies of all the consumers. A number of studies have focused on the solution algorithm of market equilibrium problems. One approach is the diagonalization method, which iteratively solves the model until a stationary point is obtained. The other is the nonlinear complementarity problem (NCP), which combines the Karush–Kuhn–Tucker (KKT) conditions of each equilibrium sub-problem. Then, the NCP is solved as a single-level model. In this subsection, the diagonalization method is adopted to iteratively solve the proposed bi-level equilibrium model in 3.3.1. In 3.3.2, convergence analysis is conducted, and a criterion is designed to achieve desired solutions due to the nonlinearity and concavity of the equilibrium model.

6.3.4.1

Diagonalization Method

The procedure of the solution algorithm is shown in Fig. 6.5. (1) (2) (3) (4) (5) (6)

Initialize the LMPs, the DHPs and iteration index. Given the current LMPs and DHPs and other users’ planning strategies, each user optimizes his/her individual SPHP planning model. After obtaining all users’ planning strategies for SPHPs, update the nodal net electric and heating load. Given the updated nodal electric load and zonal heating load, optimize the proposed DCOPF model to calculate LMPs. Update the generation output of CHP units, and calculate DHPs by using the cost-plus method. If a stationary planning strategy is achieved, i.e., the current planning strategies for SPHPs are identical to the previous ones, output the results, and the solution procedure terminates. Otherwise, go to step 2).

6.3.4.2

Convergence Analysis

The bi-level model can be reformulated into a single-level optimization program by integrating the first-order optimality conditions of each equilibrium sub-problem. This single-level model is a mixed integer nonlinear programming (MINLP) problem. Because of the nonlinearity and concavity of this model, the optimality and convergence of the proposed solution algorithm cannot always be guaranteed. The planning strategies for SPHPs may oscillate, usually among two or more different decisions, because of the possible occurrence of the non-existence of a pure Nash equilibrium solution. The previously mentioned oscillations are illustrated in a dilemma between two situations: (i) The users invest in the marginal PV/HP/TES. As the LMPs and DHPs

6.3 Sharing Economy in Joint Electricity-Heat Markets

157

Start

Initialize LMPs , DHPs and iteration index

User index j=1

Upper level : SPHP planning

j>Number of users ? No Yes

Optimize SPHP planning model for user

j

j=j+1

Update nodal net electric and heating load

Optimize DCOPF model and calculate LMPs

Lower level : Pricing model

Update generation output of CHP units

Calculate DHPs by the cost- plus method

A stationary planning strategy?

No

Update nodal LMPs and DHPs

Yes Output the optimal strategy

End

Fig. 6.5 Procedure of the diagonalization method

vary with the planning strategy for SPHPs, the users may find it not beneficial to invest in the marginal PV/HP/TES. ii) The users choose not to invest in the marginal PV/HP/TES. Then, the nodal net electric and heating loads increase, leading to an increase in users’ costs. Then, the users find it beneficial to use the marginal PV/HP/TES instead of electricity and heat from CHP units. The dilemma is illustrated in Fig. 6.6.

158

6 Sharing Economy in Energy Systems Integration

Fig. 6.6 Illustration of the dilemma

Table 6.1 Criterion for desired solutions Criterion for desired solutions 1

Find all the situations in which oscillations occur during the iterations;

2

In each oscillation situation, calculate the LMPs and DHPs based on the planning strategies for SPHPs;

3

Apply the LMPs and DHPs for settlement: Calculate all users’ electricity and heating costs;

4

Calculate the system-wide total costs for each oscillation situation, including users’ investments for SPHPs, users’ electricity and heating costs and the generation costs of all units in the CHP system

5

Find the situation with minimal system-wide total costs; the planning strategies in the selected situation are the desired solutions

To resolve this dilemma, a convergence criterion is proposed. The idea is to search for the situation with minimal system-wide total costs. The LMPs and DHPs are used for settlement (Table 6.1).

6.4 Sharing Economy in Joint Electricity-Gas Markets Increasing penetration of uncertain and variable renewable generation in evolving power system leads to fast unexpected changes in net load. California has declared that there will be an increasing demand on ramp-down capacity during sunrise and ramp-up capacity during sunset up to 2030, resulting in “duck curve” on the system net load [7]. Two consequences caused by high penetrations of renewable resources are: (1) increasing ramping requirements posed to online generators, and (2) possibility of ramp capability shortages. To deal with the inadequacy of the system’s ramping capacities, California Independent System Operator (CAISO) and Midcontinent Independent System Operator (MISO) have proposed flexible ramping product (FRP) to improve the dispatch flexibility [8, 9]. FRPs consist of separate products in the upward and downward directions as energy imbalances may be positive or negative. Any real-time schedulable

6.4 Sharing Economy in Joint Electricity-Gas Markets

159

resource has the ability to provide FRPs, while only available from thermal power units currently. Some literature has analyzed the ability of resources such as wind turbines, energy storage and electric vehicles to provide FRPs. In recent years, power-to-gas (P2G) technologies have gained substantial popularity in the role of virtual storage to transform surplus renewable energy to hydrogen that can be utilized in the future. Traditionally, the connection between natural gas and power systems is mainly through gas-fired turbines (GTs). With the rapid development of P2G, such emerging technology has been advocated as a promising solution to enhance the flexibility of renewable energy-dominated power grid. In practice, P2G can convert the hard-to-distribute wind power into hydrogen during valley hours, and thus store the surplus renewable energy as gas to support GTs during peak hours. Therefore, the coordination between P2G and GT/fuel cell will play a critical role in contributing to system operational flexibility in future smart grids. Some studies have focused on the modeling and scheduling for P2G. Among the existing research, to our best knowledge, two issues deserve an in-depth investigation: (i) The existing technoeconomic analyses mainly model P2G with constant efficiency, ignoring the operational characteristics of electrolyzed water process. (ii) Few studies have evaluated the impact of P2G on improving power system flexibility. Therefore, we aim to fill the aforementioned gap by investigating the contribution of P2G in ramping-constrained unit commitment.

6.4.1 Framework The framework is designed in Fig. 6.7. We quantify the merits of P2G toward system operational flexibility in a stochastic unit commitment (UC) model, which minimizes

ISO UC Dispatch

Net load

GT

P2G

Upward uncertainty

FRU

Volatility Downward uncertainty

FRD

Thermal

Wind t

Natural gas market

Fig. 6.7 Framework of unit commitment considering P2G as FRP providers

t+1

Time

160

6 Sharing Economy in Energy Systems Integration

the total operation costs with FRP requirements embedded. According to the characteristics of proton exchange membranes, we simulate the nonlinear relationship between hydrogen production and electricity consumption. A linearization method is then proposed to transform the efficiency function into a piecewise linear form.

6.4.2 Modeling for Power-To-Gas 6.4.2.1

Power-To-Gas

The electrolyzed water is dissociated into hydrogen and oxygen by an electrochemical process under the action of direct current. The electrode reaction is described by (6.22). According to different diaphragms, power-to-gas can be divided into alkaline liquid electrolysis tank, proton exchange membrane electrolyzer and high temperature solid oxide electrolysis tank. Due to high power density, fast start-up and environmental friendliness, proton exchange membrane electrolyzers have broad application prospects. Therefore, this subsection chooses the proton exchange membrane electrolyzer technology as the energy conversion module. ⎧ + ⎨ Anode : H2 O − 2e− → 1/2O2 + 2H + − Cathode : 2H + 2e → H2 ⎩ Reaction : H2 O → H2 + 1/2O2

(6.22)

The water electrolysis step is formed by a plurality of electrolyzer stacks connected in parallel. Each stack is formed by a series of current controlled cells. Hydrogen production can be adjusted by controlling the start and stop of the electrolyzer stacks. The output voltage of one cell can be written by the following equation. Vcell = ENerst +ηact +ηohm

(6.23)

where Vcell is the operating voltage. ENerst is the open circuit voltage. ηact is the activation overvoltage. ηohm is the ohmic overvoltage. Open circuit voltage is calculated using Nernst equation, which takes into account the effect of temperature and species concentration on the cell electromotive force: 0 + ENerst = Erev

RT 1 [ln(PH2 ) + ln(PO2 )] 2F 2

(6.24)

where T is battery temperature (K). R and F are the ideal gas constant and Faraday constant, respectively. PH2 and PO2 are the partial pressures of H2 and O2 respectively. 0 is the dependent value of reversible cell voltage, and is given as follows: Erev 0 = 1.229 − 0.9 × 10−3 (T − 298) Erev

(6.25)

6.4 Sharing Economy in Joint Electricity-Gas Markets

161

Activation overpotential is affected by current density id . This overpotential can be described by Butler–Volmer (B–V) equation for both anode and cathode as follows: ηact

  RT id id RT + arcsinh arcsinh = αc,an F 2id,an αc,cat F 2id,cat

(6.26)

where αc,an and αc,cat are the charge transfer coefficients of the anode and cathode. id,an and id,cat are the exchange current densities of the anode and cathode. The ohmic overvoltage is caused by the resistance (Rm ) in the cell, which can be written as: ηohm =Rm id

(6.27)

According to the above analysis, the relationship between the operating voltage and the current density is obtained. The total voltage drop across a stack is obtained as follows: Vstack = ncell · Vcell

(6.28)

where ncell is the number of series cells placed in a stack. According to the structure of the electrolysis water and the magnitude of the current and voltage, the power consumption (P cons ) and, hydrogen production flow (FH2 ) are expressed as: P cons = 0.001 ·

N stack 

stack ut,i Vstack Ic

(6.29)

i=1

FH2 = 3600 ·

Nstack Ic RT0  stack · ut,i ηf Nc P0 i=1 2F

(6.30)

stack where ut,i is the 0–1 variable characterizing whether the ith electrolyzer stack is operating. Ic = Acell · id , where Acell is the area of electrode. Nstack is the number of electrolyzer stacks. ηf is Faraday efficiency. T0 and P0 are standard temperature and pressure, respectively.

6.4.2.2

Efficiency Linearization

Through the above analysis, the dynamic efficiency expression of water electrolysis can be obtained. The relationship between efficiency and power consumption is shown in Fig. 6.8.

162

6 Sharing Economy in Energy Systems Integration 0.9

Efficiency

0.85 0.8 0.75 0.7 0.65 0.6 0

0.5

1.0

1.5 P(MW)

2.0

2.5

3.0

Fig. 6.8 Nonlinear efficiency function and piecewise linearization

ηH2 ,t =

HHVH2 · FH2 Pte

(6.31)

where HHVH2 is the high heat value of hydrogen under standard conditions. In fact, as the power consumption increases, the operating efficiency of the P2G is continuously reduced. In order to improve calculation speed and decrease the difficulty of solving the mixed integer nonlinear model, this subsection proposes a piecewise linearization of operational efficiency.  For the univariate function η(P) on the interval p, p , we introduce a number of segmentation points αi (i ∈ {0, 1, . . . , m}), making p = α0 ≤ α1 , · · · ≤ αm = p. The corresponding variables pi and zi (i ∈ {1, . . . , m}) are expressed as follows: p=

m 

pi

(6.32)

i=1

ai−1 zi ≤ pi ≤ ai zi , ∀i ∈ {1, . . . , m} m 

zi = 1, zi ∈ {0, 1}

(6.33)

(6.34)

i=1

The value of the variable p is only located in one of the segmentation intervals, and the corresponding function value is the P2G operation efficiency at the power consumption level. An illustration is shown in Fig. 6.8.

6.4 Sharing Economy in Joint Electricity-Gas Markets

163

6.4.3 System Model 6.4.3.1

FRP Requirement

FRP aims to reserve enough ramping space for the system to simultaneously address the variability and uncertainty of net load. The FRP requirements are composed of two parts: the forecasted net load changes between the current period and the next period, and the uncertainty of forecasted net load at next period. FRP includes both upward and downward types. It is called upward ramping products (FRU) and downward ramping products (FRD). The formulas for calculating FRP requirements are: sys

  FRU = max RFRU V,t + RU,t , 0

(6.35)

sys

  FRD = max RFRD V,t + RU,t , 0

(6.36)

URRt DRRt sys

sys

where URRt and DRRt is the total FRU and FRD requirements for time t, respecFRD tively. RFRU V,t and RV,t is the FRU and FRD requirements caused by the system net FRD load variability, respectively. RFRU U,t and RU,t are FRU and FRD requirements for the uncertainty of the net load forecast, respectively. The FRP requirement for multiple time periods is shown in Fig. 6.9. The black line is the net load forecasting curve, and the difference between the two periods is called the net load volatility. The blue line is the uncertain demand caused by the deviation of the system’s net load forecast within a certain confidence interval. The red line (green line) is the FRU (FRD) requirement of the current time period.

Net load

Upward uncertainty

FRU Demand

Volatility Downward uncertainty

FRD Demand

t

t+1

Fig. 6.9 FRP requirement for multiple time periods

t+2

t+3

Time

164

6 Sharing Economy in Energy Systems Integration

6.4.3.2

Ramping-Constrained Unit Commitment (UC) with P2G

The proposed day-ahead unit commitment model is formulated as a mixed-integer programming problem with the objective function given in (6.15). The objective of the UC is to minimize operation and start-up/shut-down costs while anticipating the real-time generation and FRP costs. min

N N T  U 

(CiSU yi,t + CiSD zi,t )

t=1 i=1

+

NS  s=1

rs

N N T  U 

G CiG (Pi,s,t )

t=1 i=1

NT  RU UP RD DN + (Cs,t Ss,t + Cs,t Ss,t )

 (6.37)

t=1

where NT is number of time periods. i is index of units, including the thermal power unit (m is index of thermal power unit) and the gas turbine (g is index of gas turbines). NU is number of units. NTU is number of thermal power units, NGT is number of gas turbines, NU = NTU + NGT . s is the index of scenarios. NS is number of scenarios. CiSU and CiSD are the up and down costs respectively. yi,t and zi,t are startup and shutdown state variable of a unit. rs is the probability of the scenario s. CiG is the G is the power generation of unit i at time t under power generation cost of unit i. Pi,s,t RU RD is flexible ramp down surplus scenario s. Cs,t is flexible ramp up surplus price. Cs,t price. The FRP surplus price is obtained by estimating the probability of power imbalance under a certain amount of FRP by historical data, and multiplying it by UP DN and Ss,t are flexible ramp up and down surplus at the value of the load loss. Ss,t time t. The constraints of the proposed UC model are listed below: (1)

FRP requirements

FRP aims to reserve enough ramping space for the system to simultaneously address the variability and uncertainty of net load. The FRP requirements are composed of two parts: the forecasted net load changes between the current period and the next period, and the uncertainty of forecasted net load at next period. It is called upward ramping products and downward ramping products. The formula for calculating FRP requirements is: NTU 

RUP m,s,t +

m=1 NTU 

NGT 

sys

(6.38)

sys

(6.39)

UP RUP g,s,t + Ss,t ≥ URRs,t

g=1

RDN m,s,t

m=1

+

N P2G 

DN RDN p,s,t + Ss,t ≥ DRRs,t

p=1 sys

sys

DN UP DRRs,t ≤ Ss,t ≤ 0 ≤ Ss,t ≤ URRs,t

(6.40)

6.4 Sharing Economy in Joint Electricity-Gas Markets

165

DN where RUP m,s,t and Rm,s,t are ramping up and down capacity available from thermal power unit m at time t. p is index for P2G systems. RUP g,s,t is ramp up capacity available from gas turbine g at time t. NP2G is number of P2G systems. RDN p,s,t is ramp down capacity available from P2G system p at time t.

(2)

Available capacity constraints G G Pi,s,t+1 + RUP i,s,t ≤ ui,t Pi,max , ∀t, i, s

(6.41)

G G Pi,s,t+1 − RDN i,s,t ≥ ui,t Pi,min , ∀t, i, s

(6.42)

G where ui,t ∈ {0, 1} indicates the start and stop state variables of unit i at time t. Pi,max G and Pi,min are the maximum and minimum power generation of unit i respectively.

(3)

Ramping capability constraints G G Pi,s,t+1 − Pi,s,t + RUP i,s,t ≤ Lt · URi,t ,∀t, i, s

(6.43)

G G Pi,s,t+1 − Pi,s,t − RDN i,s,t ≥ −Lt · DRi,t ,∀t, i, s

(6.44)

where URi,t and DRi,t represent the up/down ramping rate limit for unit i at interval t. Lt represents the length of interval t. (4)

System balance constraints NU 

G Pi,s,t +

i=1

NK  k=1

RG Pk,s,t =

NB 

D Pb,s,t +

b=1

N P2G 

P2G Pp,s,t

(6.45)

p=1

where k is index for wind turbines. NK is number of wind turbines. b is index for buses. NB is number of buses. (5)

(6)

Unit commitment constraints ui,t+1 − ui,t = yi,t − zi,t ,∀t, i

(6.46)

yi,t + zi,t ≤ 1

(6.47)

yi,t + zi,t ≤ 1

(6.48)

Power flow constraints L −Pl,max ≤

NB  b=1

G RG D P2G L Fl−b (Pi,s,t + Pk,s,t − Pb,s,t − Ph,s,t ) ≤ Pl,max ,∀t, ∀s

(6.49)

166

6 Sharing Economy in Energy Systems Integration

L where Pl,max is the maximum transmission capacity of line l. Fl−b is the distribution factor of the branch l to the bus b.

(7)

Wind power output constraints RG FRG ≤ Pk,s,t ,∀t, k, s 0 ≤ Pk,s,t

(6.50)

FRG is the predicted output of wind turbine k at time t in scenario s. where Pk,s,t

(8)

P2G operating constraints

The P2G operating constraints include a nonlinear relationship between hydrogen production flow rate and the power consumption, and the current density adjustment range constraint: id,min ≤ id ≤ id,max

(6.51)

where id,max and id,min are the upper and lower limits of current density, respectively. (9)

Gas turbine operating constraints

For a power system, gas-fired turbines are energy supply resources. The power generation is proportional to the natural gas consumption: GT GT Pg,s,t = ηG2P · Vg,s,t , ∀t, g, s

(6.52)

GT where g is index of GTs. ηG2P is the conversion efficiency of GT. Vg,s,t is the volume of natural gas consumed by GT.

(10)

Natural gas market contract

The gas consumption of GTs is provided by natural gas market contract with P2G hydrogen production. In this subsection, we consider a day-ahead contract, and the total gas consumption should be restricted as follows: NT  NGT 

GT Vg,s,t ≤ V H →NG + V NG

(6.53)

NT N P2G  HHVNG  = · V P2G HHVH2 t=1 p=1 p,s,t

(6.54)

t=1 g=1

V

H →NG

where V NG is the contracted gas volume purchased by the gas unit from the natural gas market. V H →G is the volume of hydrogen equivalent to the natural gas required by the gas turbine. HHVNG is the high heat value of natural gas under standard conditions.

6.5 Sharing Economy in Transportation-Energy Systems

167

6.5 Sharing Economy in Transportation-Energy Systems Power and transportation industries are the main sources of global greenhouse gas emissions, of which electric power and transportation sectors generally account for 40% and 24%, respectively [10]. In addition, traffic emissions are important causes of air pollution. Due to the emission of particulate matter and NOx, more than one million people might die prematurely in Europe each year. The direct costs of diseases, production losses, crop yield reductions, and building damage due to air pollution are about 24 billion euros per year, while external costs are estimated to be 330–940 billion euros per year [11]. To cope with the severe challenges caused by resource shortage, climate change and environmental pollution, developing the coordinated revolution of energy and transportation system is a fundamental way. Sweden requires that all vehicles must use non-fossil fuels by 2030 and achieve net zero-carbon emissions by 2050. Britain and France have announced a nationwide ban on fossil-fuel vehicles by 2040 [12]. In September 2019, China issued the “Outline of Constructing a Powerful Transportation Country”, proposing to establish a multilevel and integrated transportation hub to optimize the transportation energy structure and the utilization of renewable energy. International Energy Agency (IEA) has pointed out that battery electric vehicles (BEVs) and fuel cell vehicles (FCVs) are the technologies that can provide a sustainable road transportation system with near-zero emissions. Due to lower fuel costs and ready-to-use infrastructure, EV has become a key development target for countries all over the world prior to FCV. In 2019, 2.3 million EVs were sold worldwide, which was 9% more than those in 2018. Compared with fossil-fuel vehicles, EVs cannot only help maintain a clean environment, but also reduce the operating costs of transportation. However, the integration of large-scale EVs will significantly enlarge the peak-valley difference of power grids, thus yielding a great burden of power system operation. By 2030, State Grid Corporation of China (SGCC) has expected a 153-TW peak load caused by EV fleet charging. In order to alleviate the burden of large-scale EVs integrated to power systems, many studies have discussed how to manage EV fleet charging in aspects of dispatch, control and planning. The existing references mainly focus on electric power system scheduling considering EVs integration, without exploring the synergy and complementary of different forms of energy toward a low-carbon transportation-energy (trans-energy) system. The concept of energy systems integration has provided a novel pathway to manage the penetration of large-scale EVs. As a promising secondary energy, hydrogen has the advantages of zero emissions and high specific energy relative to most batteries, and is becoming an important link between trans-energy systems in recent years. Green hydrogen can be generated from renewable energy via powerto-gas (P2G), and be transformed into other chemical products or stored for future use. The relatively lower cost of green hydrogen enables FCVs to be a potential development direction.

168

6 Sharing Economy in Energy Systems Integration

The current research on FCHEVs focuses on the optimal energy management within a vehicle rather than the optimal scheduling strategy for FCHEVs in transenergy systems. In this section, an optimal scheduling framework is proposed for trans-energy systems considering the participation of FCHEVs. During peak load periods, FCHEVs can choose the hydrogen refueling mode instead of the electricity charging mode to reduce the operating burden of the grid. When wind power is rich during night hours, FCHEVs can be charged by electricity to improve the accommodation of wind power, then surplus wind power can be converted into hydrogen for future use. By this means, FCHEVs connect trans-energy systems, thereby promoting the consumption of renewable energy and improving the flexibility of power grids.

6.5.1 Framework In this subsection, a trans-energy system framework considering FCHEVs’ participation is designed to aggregate hydrogen production and transportation, conventional energy and renewable energy power generation plants. The schematic is shown in Fig. 6.10. On the one hand, Fig. 6.10 describes hydrogen supply chains (HSCs) that include hydrogen production through P2G system, compression, transportation and end-use by FCHEVs. During periods of low load and high incidence of wind power at night, surplus wind power can be converted into storable hydrogen through P2G, thereby facilitating the accommodation of wind power and providing downward flexibility for the power system. The generated hydrogen is then transported to energy consumption terminals for FCHEVs’ hydrogen refueling. In addition, large-scale hydrogen storage

Generation

Hydrogen Central Production

Hydrogen Distribution

Transportation

H2 P2G system

Hydrogen

Compression

Tube trailers

Fuel cell

Renewable energy Hydrogen storage

FCHEV

Conventional energy

Electricity network

Electricity Distribution

Fig. 6.10 Schematic of the trans-energy system connected by FCHEVs

Battery

6.5 Sharing Economy in Transportation-Energy Systems

169

is one of the few low-carbon solutions that can balance the intermittency of wind and solar power generation. On the other hand, the electricity produced by conventional energy and renewable energy is transmitted to the charging station through the electricity network to charge the batteries of FCHEVs. FCHEVs can also release electricity to the grid, providing upward flexibility during peak load hours. We have made the following selections for the equipment. Hydrogen production is considered as proton exchange membrane (PEM) electrolyzers, which has quick response and quick start with a wider dynamic range (5%–150%). Thus, it is more suitable for intermittent power supplies such as wind power. The produced hydrogen needs to be compressed or liquefied to reach sufficient energy density. Currently, highly compressed gas is the most cost-efficient choice for on-board vehicle storage. Excessive compressed hydrogen can be stored in a hydrogen storage tank. In terms of transportation, due to lower infrastructure costs and risks, we select gas tube trailers to transport compressed hydrogen. FCHEV is a key component that connects transportation system with power grids. By taking advantage of FCHEV’s ability to be charged by both hydrogen and electricity, we can fully explore the complementary effects of hydrogen and electricity, thus achieving an in-depth integration of trans-energy systems.

6.5.2 Model of Fuel-Cell Hybrid Electric Vehicle 6.5.2.1

Techno-Economic Analysis

BEVs, FCVs and FCHEVs are the mainstreams as being capable of delivering a sustainable road transportation system with near-zero emissions. They have different performances in terms of cost, mileage and energy supplement as follows: (1)

Capital cost: At present, the investment and operating cost of an FCV are higher than those of a BEV. Figure 6.11 shows the capital cost comparison between

Vehicle capital cost(US$)

70000 60000

BEV(2017) FCV(2017 ) FCHEV(2017 )

50000 40000

BEV(2040 ) FCV(2040) FCHEV(2040 )

30000 20000 0

100 200 300 400 500 600 700 800 900 1000

Mileage(km)

Fig. 6.11 Capital cost comparison between BEV, FCV and FCHEV

170

6 Sharing Economy in Energy Systems Integration

BEV, FCV and FCHEV in 2017 and in 2040. In 2040, the capital costs of the three vehicles tend to be the same. (2)

(3)

Mileage and charging time: “mileage anxiety” is the main obstacle for BEVs, causing 30% of American consumers to be reluctant to pay for it. But this is not a problem for FCVs, as hydrogen has high specific energy relative to batteries, it has longer driving mileage. In terms of charging time, the FCVs can be refueled quickly within 5 min while BEVs take several hours. Energy supplement: The fuel of FCVs, hydrogen, is very expensive. There are few hydrogen refueling stations currently. BEVs use electric energy as energy source, and the cost is low.

Technically, FCHEVs combine the advantages of BEVs and FCVs, and will have a better performance with longer mileage and shorter charging time. In terms of cost, due to the peak load shifting characteristics of wind power, a large amount of wind power is not fully utilized at night. In practice, wind power curtailment can be used to produce hydrogen via P2G, which can greatly reduce the cost of hydrogen. Thus, the operating cost of FCHEVs is lower. Therefore, if hydrogen and electricity are used simultaneously as the energy supply for vehicles, it will solve the problems of high fuel cost of FCV and insufficient mileage of BEV. FCHEV can satisfy the above-mentioned needs.

6.5.2.2

FCHEV Mileage Model

FCHEVs use both hydrogen and electricity as energy sources so that they can choose to be charged by electricity or refueled by hydrogen according to dispatching signals. The configuration mainly includes fuel cell and battery as its primary sources, shown in Fig. 6.12. The fuel cell uses a PEM fuel cell, and the battery uses a lithium-ion battery. To uniformly measure the hydrogen / electricity charging status of FCHEV, mileage is an important indicator which is modeled by the stored electricity and hydrogen mass. The mileage of an FCHEV (Rv ) can be expressed as follows: Rv =

EB ϕB (MV +

+

EB km,B SEBC

+ MH 2 + MFC PFC + MHT ) MH 2

ϕH (MV +

EB km,B SEBC

+ MH 2 + MFC PFC + MHT )

(6.55)

where EB is the battery pack energy, MH 2 is the mass of stored hydrogen. ϕB and ϕH are energy consumption efficiency and hydrogen consumption efficiency of FCHEVs, respectively. MV is the vehicle mass excluding the battery pack. km,B is the battery pack mass.SEBC is the specific energy of the battery cell. MFC and MHT are the mass

6.5 Sharing Economy in Transportation-Energy Systems

171

Power Line Vehicle controller

e

Motor controller

Transmission

Control Signal

e

Electronic interfaces

Battery

Electric machine

Charge

Load

Electrolytes (lithium salts) Cathode (lithium metal oxides)

Discharge

e

Anode (carbons)

e

Hydrogen tank H2

Fuel Regulation

PEMFC Stack

Fuel Cell System

Humidified Air

Air

Fig. 6.12 Configuration of fuel-cell hybrid electric vehicle

of fuel cell system and hydrogen tank, respectively. PFC is the power of fuel cell system. The formula (6.55) can be simplified into the following form: Rv =

EB MH 2 + a1 + b1 · MH 2 + c1 · EB a2 + b2 · MH 2 + c2 · EB

(6.56)

where ai , bi and ci are constants. The mileage of FCHEV is sensitive to the battery pack energy and the mass of stored hydrogen. In the denominator, (b1 · MH 2 + c1 · EB ) and (b2 · MH 2 + c2 · EB ) account for a small proportion, about 4.85–9.23%. Thus, the denominator can be set as a constant to simplify the computational complexity. We simplify the formula to Eq. (6.2). The mileage of FCHEV is determined by the battery’s state of charge (SOC) and the mass of stored hydrogen. EV EV H · Ev,max · kiE + mEV Rv =SOCv,s,t v,s,t · ki ,

(6.57)

where kiE is the conversion coefficient of battery capacity to mileage and kiE = 1/[ϕB (Mv + C1 )]; kiH is the conversion coefficient of hydrogen mass to mileage and EV is the SOC of the vth kih = 1/[ϕH (Mv + C2 )]; C1 and C2 are constant; SOCv,s,t EV FCHEV at time t;Ev,max is the capacity of the battery pack; mEV v,s,t is the mass of stored hydrogen in the vth FCHEV at time t.

172

6 Sharing Economy in Energy Systems Integration

6.5.2.3

Charging/Refueling Station Model

In this chapter, the charging/refueling station will dispatch FCHEVs’ charging/discharging and refueling according to the peak or valley situations of the net load, thereby improving the system’s operational flexibility. FCHEVs dep will leave the station when the target mileage is met, and the target mileage (Rv ) is determined by the battery’s state of charge (SOC EV dep ) and hydrogen mass (mEV

v,s,Tv

dep v,s,Tv

). The station model is formulated as follows: CSC Pj,s,t =





EVC CSD Pv,s,t , Pj,s,t =

v∈EV

EVD Pv,s,t ,

  CSC CSD CS , , Pj,s,t ∈ 0, Pmax Pj,s,t    EV mEV dep − mv,s,T arr , v v,s,T

mRS q,s =

v∈EV

v

EV EV E EV H SOCv,s,T ≥ Rdep dep · Ev,max · ki + m dep · ki v , v,s,T v

v

EV EV SOCv,s,t = SOCv,s,t−1 +

(6.58)

v∈EV

 EV  1  EV EVC EVD ηv Pv,s,t−1 − Pv,s,t−1 ηv , EV Ev,max

(6.59) (6.60) (6.61) (6.62)

EVC EVC EVC 0 ≤ Pv,s,t ≤ αv,s,t Pmax ,

(6.63)

EVD EVD EVD 0 ≤ Pv,s,t ≤ αv,s,t Pmax ,

(6.64)

EV arr EV arr SOCv,s,T arr = SOCv,s , mv,s,T arr = mv,s , v v

(6.65)

EV 0 ≤ SOCv,s,t ≤ SOC,

(6.66)

EV marr v,s ≤ mv,s,t ≤ m,

(6.67)

EVC EVD where Pv,s,t and Pv,s,t are the charging and discharging power of the vth FCHEV,   EV respectively;  = 1, 2, . . . , N EV represents the set of FCHEVs; N EV is the CSC CSD and Pj,s,t are the total charging and discharging power number of FCHEVs; Pj,s,t CS of the station;Pmax is the capacity of the station; mRS q,s is the total hydrogen mass EVC EVD provided by the station q; Pmax and Pmax are the maximum power of FCHEV dep charging and discharging, respectively; Tvarr and Tv are arrival time and deparEV ture time of FCHEVs, respectively.ηv is the coulomb efficiency of battery; SOC is the upper limits of FCHEV’s SOC; SOCvarr and marr v are initial state of charge and

6.5 Sharing Economy in Transportation-Energy Systems

173

initial hydrogen mass, respectively; m is the maximum mass of hydrogen that can be EVC EVD and αv,s,t are 0–1 variables that characterize whether the stored by an FCHEV.αv,s,t vth FCHEV is connected to the grid and expressed as (6.58). At any time, FCHEV can only be in one of the charging, discharging or hanging-up state. 

EVC EVD αv,s,t = αv,s,t = 0, t < Tvarr or t > Tvdep , EVC EVD αv,s,t + αv,s,t ≤ 1, Tvarr ≤ t ≤ Tvdep .

(6.68)

In Eq. (6.58), the power of station is the aggregation of FCHEV fleets. Equation (6.59) shows the capacity of the station. The hydrogen consumption of the refueling station throughout the day is shown in Eq. (6.60). In constraint (6.61), an FCHEV will only leave after meeting its mileage requirement. In Eq. (6.62), the SOC of an FCHEV is related to the charging and discharging power during a time slot. In constraints (6.63) and (6.64), the charging and discharging of a single FCHEV is limited by its maximum power. In Eq. (6.65), the initial SOC and the initial hydrogen mass is set to the value of an FCHEV in arrival time. Constraints (6.66) and (6.67) show that the SOC and hydrogen refueling mass of an FCHEV are limited by its battery pack and hydrogen tank, respectively.

6.5.3 Optimal Scheduling Model for Trans-Energy Systems In this section, an optimal scheduling model for trans-energy systems is proposed considering the participation of FCHEVs. Through the scheduling of FCHEV’s charging, discharging and hydrogen refueling at different times, the strategy for economically optimal operation of FCHEVs in the trans-energy systems is obtained.

6.5.3.1 (1)

Optimal Scheduling Model

Objective function

The objective of the scheduling strategy is to minimize the trans-energy systems costs including the operation cost of power system and hydrogen transportation cost. In Eq. (6.69), the first term is the cost of unit start-up and shutdown, and the second term is the generation cost of thermal power units (the generation cost of renewable energy units is taken as zero). Due to the long distance between P2G and the load center, the transportation cost of hydrogen cannot be ignored, as shown in the third term.

174

6 Sharing Economy in Energy Systems Integration

min

⎧ NT NI ⎫ NS  NT  NI   ⎪ ⎪ ⎪ SU SD G G ⎪ ⎪ (Ci yi,t + Ci zi,t ) + rs Ci (Pi,s,t )⎪ ⎪ ⎪ ⎪ ⎪ ⎨ t=1 i=1 ⎬ s=1 t=1 i=1 Q ⎪  ⎪ ⎪ 2 ⎪ [C HT · mH + ⎪ l→q ]dl→q ⎩

⎪ ⎪ ⎪ ⎪ ⎪ ⎭

l→q=1

(6.69)

whereNT , NI , NS are the sets of time slots, thermal power units and scenarios, respectively; Q is the set of paths in the hydrogen transportation network;yi,t and zi,t are start-up and shut-down state variable of unit i; CiSU and CiSD are the start/shut cost of the uniti; rs is the probability of the scenario s. CiG is the power generation cost G is the power generation of thermal power unit i at time t, respectively; of unit i; Pi,s,t 2 HT C is fuel unit cost of hydrogen transportation; mH l→q is hydrogen transportation flow of path l → q; dl→q is the distance between path l and q. (2)

FCHEV charging/refueling station constraints

Constraints (6.58)–(6.68) (3)

Electricity-hydrogen energy coupling constraints

P2G converts the surplus wind power into hydrogen, so its electricity input can be regarded as the load of the system, and the hydrogen output can be regarded as the source of the hydrogen supply chain. The electricity-to-hydrogen process is to dissociate water molecules into hydrogen and oxygen through an electrochemical reaction. The relationship between power consumption and hydrogen production is shown in Eq. (6.70), and the power consumption in the period t is limited by the capacity of the P2G system as shown below. P2G 2 mH · p,s,t = ϕp

P2G · t Pp,s,t

LHVH2

P2G P2G 0Pp,s,t Pp,max

(6.70) (6.71)

P2G where ϕpP2G is the efficiency of converting electricity into hydrogen; Pp,s,t is the H2 power consumed by P2G; mp,s,t is the mass of hydrogen produced by P2G; LHVH2 P2G is the low heat value of hydrogen under standard conditions; Pp,max is the capacity limits of P2G system.

(4)

Hydrogen supply chain constraints

The hydrogen produced by the P2G systems is compressed and then injected into the hydrogen storage tank or directly injected into tube trailers transporting to refueling stations. 0≤

NP 24   t=1 p=1

HS 2 mH p,t ≤ mp,max

(6.72)

6.5 Sharing Economy in Transportation-Energy Systems NP 24  

2 mH p,t

·ϕ

comp

t=1 p=1

Q 

=

175

2 mH l→q

(6.73)

l→q=1 RS 2 mH l→q = mq,s

(6.74)

where NP is the set of P2G systems; ϕ comp is the efficiency of compressing hydrogen; mHS p,max is the capacity limits of hydrogen storage tank. Constraint (6.72) shows the total amount of hydrogen produced in one day needs to be less than the capacity of the hydrogen storage tank. Due to the loss of compressed hydrogen, the actual amount of hydrogen in transportation is shown in (6.73). In Eq. (6.74), the amount of hydrogen in the refueling station q is equal to the sum of the hydrogen transported from each P2G to the refueling station q. (5)

Power system constraints

NI  i=1

G Pi,s,t +

NK 

ui,t+1 − ui,t = yi,t − zi,t , yi,t + zi,t ≤ 1

(6.75)

G G G ui,t Pi,min ≤ Pi,s,t+1 ≤ ui,t Pi,max

(6.76)

G G Pi,s,t+1 − Pi,s,t Lt · URi,t

(6.77)

G G Pi,s,t+1 − Pi,s,t ≥ −Lt · DRi,t

(6.78)

RG Pk,s,t =

k=1

NB  b=1

L − Pl,max ≤

NB 

D Pb,s,t +

NP 

P2G Pp,s,t +

p=1

NEV 

EVC EVD (Pv,s,t − Pv,s,t ) (6.79)

v=1

G RG D P2G Fl−b [Pi,s,t + Pk,s,t − Pb,s,t − Pp,s,t

b=1

−

NEV 

EVC EVD L (Pv,s,t − Pv,s,t )] ≤ Pl,max

(6.80)

v=1

where NB ,NEV , NK are the sets of load nodes, FCHEVs and renewable energy units, G G and Pi,max respectively; ui,t ∈ {0, 1} indicates the start and stop state variables; Pi,min are the upper and lower limits of the output of thermal power unit i, respectively; URi,t and DRi,t are the up/down ramping rate limit, respectively; Lt is the length RG D is the power generation of renewable energy unit k; Pb,s,t is the of interval t; Pk,s,t L power consumed by the load;Pl,max is the maximum transmission capacity; Fl−b is the distribution factor of the branch l to the node b.

176

6 Sharing Economy in Energy Systems Integration

Constraints (6.75) to (6.80) represent unit commitment constraints, available capacity constraints, up/down ramping capability constraints, system balance constraints and power flow constraints, respectively.

6.5.3.2

Benefits of FCHEV

Hourly power (MWh)

Hourly power (MWh)

The potential benefits of FCHEV are analyzed in Fig. 6.13. Figure 6.13a illustrates the operation situation of BEVs connected to the power system. Compared with BEVs that do not interact with the power system, BEVs’ participating in power system dispatching can achieve peak shaving and valley filling. Unfortunately the wind power may not be fully accommodated at night. In Fig. 6.13b, an illustration for the scenario of using FCHEVs is provided. When the wind power output is large during valley load periods, the surplus wind power can be converted into storable hydrogen energy via P2G, which facilitates the accommodation of renewable energy. The produced hydrogen can be then refueled by FCHEVs during peak electric load, thereby reducing FCHEVs’ demand for charging electricity. Compared with scenario (a), FCHEVs in scenario (b) have a better performance in peak-shaving and valley-filling. With the connection of FCHEV, the transportation and energy systems are integrated, and hydrogen energy and electric energy are multi-energy complementary, thus optimizing the allocation of resources on a larger scale.

(a)

Hour(h)

(b)

Hour(h)

Load

Wind power

Power shift

Discharging of EV/FCHEV

Charging of EV/FCHEV

FCHEV hydrogen charging

Wind curtailment

Hydrogen converted by wind

Fig. 6.13 Benefit analysis brought by FCHEVs

Power charging curtailment

6.5 Sharing Economy in Transportation-Energy Systems

177

6.5.4 Shortest Path Search Algorithm To measure the cost of hydrogen transportation, a shortest path search algorithm is used to optimize the transportation system. Dijkstra algorithm is a popular method to find the shortest path in traffic network. The main feature of Dijkstra algorithm is to expand from the source node to the outer layer until it reaches the destination node, and find the shortest path for each passing point. Assume that each node has a pair of labels (wj , pj ). wj is the distance of the shortest path from the source node s to any node j(the shortest path length from the source node to itself is 0); pj is the node before node j in the shortest path from source node s to node j. The direct connection distance from node i to node j is denoted dij . Dijkstra algorithm 1

Initialization. For the source node s, set ws to 0, set ps to null; for other nodes, set wi = ∞,pi = s. Mark the source node s, set the marked node k = s, and set other nodes as unmarked.

2

Check the shortest path. Check the distance from all marked node k to its directly   connected unmarked node j, and set wj = min wj , wk + dkj .

3

Select the next point i.wi = min wj (all unmarked nodes j), Node i is selected as the node in the shortest path and set as a marked node.

4

Find the previous node. Find the point j’ directly connected to node i from the marked nodes as the previous node, and set j’ .

5

Mark nodei. If all nodes are marked, the algorithm finds the shortest path. Otherwise, set k = i and go to Step 2.

6.6 Integrated Demand Response Faced with the serious environmental challenges and energy shortage brought on by the massive consumption of fossil fuels, it has become a global consensus to develop renewable and sustainable energy. To solve this worldwide problem, smart grid is actively constructed to facilitate the high penetration of renewable generation. However, due to the limited accommodation capability of smart grids, renewable energy curtailment is still a severe issue in existing energy systems. Therefore, the concept of energy internet is emerging at the historic moment. In the book The Third Industrial Revolution, Jeremy Rifkin presents that by integrating various forms of energy, the smart grid-centered energy internet will change the conventional energy consumption patterns and stimulate the development of renewable and sustainable energy [13]. In 2015, the Chinese government issued the initiative of Internet+ smart energy, indicating that the integration of electricity, thermal energy and natural gas is an imperative basis to construct the energy internet [14]. With the rapid development

178

6 Sharing Economy in Energy Systems Integration

of multi-energy systems and energy internet, the coupling of various forms of energy has been growing tighter along the energy production, delivery and consumption sectors. Demand response is a critical and effective measure to stimulate the demand side resources to interact with renewable generation in the power system. According to the reports from Department of Energy (DOE) in the U.S., the electricity users can participate in price-based and incentive-based DR programs to shift or reduce load demands. When the secure operation of the power system is jeopardized or there exist large amounts of renewable energy to be consumed, DR can effectively help keep the balance between electricity supply and demand, and accommodate more renewable energy. In most existing research, the DR techniques are applied for a single energy carrier system, e.g. electricity, which are barely feasible when electricity users have some shiftable or curtailable loads. However, inelastic electricity users with only must-run loads cannot participate in any DR program at all. In addition, due to high discomfort costs, most of the electricity users would operate as must-run loads and are reluctant to interrupt or delay their electricity consumption. Therefore, the single energy carrier system cannot fully utilize the demand side resources to implement DR programs. The core concept of multi-energy systems brings new insights for demand response. The integration of electricity, thermal energy, natural gas and other forms of energy enables all the energy users to be active in DR programs. With the complementarity of MESs, the energy users, including must-run loads, can actively participate in DR programs by converting various forms of energy to electricity in peak periods, instead of purchasing electricity from the power system. From the power system perspective, the energy users reduce electricity demands in peak periods. From the users’ point of view, their energy consumption is not changed at all, maintaining consumers’ comfort. This idea can be extended to the multi-energy systems, where DR becomes a critical measure to improve the economy and reliability of MESs. This novel vision of DR programs is termed as “Integrated Demand Response”. By implementing the IDR programs, the MESs, e.g., district heat/cooling, natural gas, biomass and electric power systems, will backup each other to constitute a more economical and reliable entity. In addition, the response capability of users can be fully exploited without any loss of energy users’ comfort. Up to present, the research on IDR in the MES has been drawing wide attention from the world. In this chapter, the review and prospect of IDR in the MES are conducted to provide a reference for the future investigation.

6.6.1 Basic Concept Integrated demand response is a novel vision of DR programs. Taking advantage of the complementarities of different inertia of multi-energy, IDR is aimed at fully exploiting the DR capabilities of all the users and improving the economic and reliable operations of multi-energy systems. The conventional DR programs are merely

6.6 Integrated Demand Response

179

focused on the electric power sector. Energy users are encouraged to participate in incentive-based and price-based DR programs. When the power system is jeopardized, incentive payments or discounted rates will be broadcast to energy users for pre-contracted load reductions. Or the energy users with some shiftable loads will actively respond to the change of real-time prices (RTPs) and shift load demands away from peak hours to valley hours. These options usually involve a temporary loss of comfort. The changes of the electricity consumption behavior cause dissatisfaction and discomfort of energy users. In addition, only a single energy carrier, e.g., electricity, is utilized to implement the DR programs, which may not be applicable for all users. For instance, in power systems, the users with only must-run loads cannot participate in DR programs. Even if some loads are shiftable or curtailable, the DR capability of energy users cannot be fully exploited merely through electricity. With the development of multi-energy systems, the integration of electricity, thermal energy, natural gas and other forms of energy enables all the energy users to be active in DR programs. The smart energy hub, as the center of converting various kinds of energy, has played an important role in the MES. By co-optimizing and deploying the synergy of MESs, the flexibility and economy of the energy systems will be significantly improved via the strategic complementarity among different forms of energy. Coupling with different energy carriers, SEHs can economically optimize the input energy flows to satisfy arbitrary energy outputs by converting internal energy resources. In an IDR program, energy users can not only shift their energy consumption, but also change the source of the consumed energy. Hence, all the energy users, including the must-run loads, can be active to provide DR capabilities for the MESs. A diagram of IDR is shown in Fig. 6.14. In the diagram, the electric energy and natural gas from the power and gas systems are integrated by various devices in the SEH, including energy storage systems (ESSs), micro-turbines (MTs), gas furnaces

Energy Consumption

ESS

Smart Energy Hub

Wind

Electric load Solar

AC MT

HP LiBr

Gas

Cooling load

GF Heating load Natural gas Electric energy

TES

Cooling energy Heating energy

Gas load

Fig. 6.14 The diagram of integrated demand response

180

6 Sharing Economy in Energy Systems Integration

Smart energy hub

+

Power grid Transformer

Electricity load

Time Time

Price

EMS Micro turbine Control signal

Natural gas

Thermal load

+

Gas furnace

+

Electricity

Thermal energy

Natural gas

Information

Time Time

Fig. 6.15 The schematic of a combined gas and electric power system

(GFs), heat pumps (HPs), air conditioning (AC) ssystems, refrigeration system (e.g., LiBr chiller), thermal energy systems (TESs), etc. In addition, different types of load demands can be satisfied with the optimal operations of the SEH. One of the major contributions of IDR is to improve the reliability of the MESs. For instance, taking advantage of the inertia of the thermal loads and natural gas systems, the reliable operation of the electric power system can be enhanced with the help of IDR. Take the combined gas and electric power system as an example. The schematic is shown in Fig. 6.15. The energy inputs of the system are natural gas and electricity from the power system. To satisfy the electricity and heat loads, SEHs can optimize and convert the input energy flows. When the electricity prices are high at peak hours, the natural gas will be mostly supplied for the electricity loads through micro-turbines. Equivalently, the electricity from the power system can be effectively reduced. In addition, due to the thermal inertia of the buildings, the comfortable temperatures will be maintained for a period. When wind output is high at night leading to low electricity prices, the SEHs will control the users’ appliances to directly consume electricity from the power system. From the power system’s perspective, the energy users reduce electricity demands in peak periods and increase electricity demands in valley hours. From the users’ point of view, the electricity consumption is not changed at all and the temperatures can be maintained within the comfortable levels. Therefore, by means of IDR, the DR capability of users can be fully utilized without any loss of energy users’ comfort.

6.6 Integrated Demand Response

181

6.6.2 Value Analysis With the expansion of multi-energy systems towards the demand side, IDR will break down the barriers between electricity and other forms of energy, thereby achieving the deep integration of multiple energy and information streams in the demand side. The concept of IDR fully considers the strategic complementarity of multiple energy consumption, yielding the synergy effects of MESs. The values of IDR in terms of the system operation and users’ benefits are analyzed as follows: (1)

Improve the economy of energy systems.

By integrating electricity, thermal energy, natural gas and other forms of energy, IDR makes it possible for the system operator to maximize the social welfare in a greater optimization space. IDR breaks the barriers among different forms of energy, which enables energy users to flexibly switch the energy sources according to different energy prices. By converting electricity to gas and thermal energy, large amounts of renewable energy can be further accommodated and reused in the future. As a result, the total operational costs can be dramatically reduced. In addition, the strategic complementarity of various energy can improve the utilization of the electric distribution networks, natural gas and heat supply pipelines. This system-friendly manner will defer the expansion planning of energy distribution systems and save a large amount of investment costs. (2)

Enhance the reliability of energy systems.

Taking advantage of the complementarities of different energy, IDR can help enhance the reliability of energy systems. Different energy systems backup each other to satisfy the load demands of the energy users. When the natural gas is in shortage, the electricity from the bulk power system will be used to meet the requirements of the consumers. Moreover, distinguished from the electric power system where power balance must be kept at any given instant, the energy unbalance is acceptable for the thermal and gas system because of the storage ability. Considering the inertia of natural gas systems, power-to-gas (P2G) technique can be applied to incorporate the surplus of electric power by converting electricity to hydrogen or methane. In addition, the storage ability of the heat/cooling systems can be used to smooth out the fluctuations of electric power and reduce the peak loads. (3)

Exploit the capability of demand side resources.

IDR enables energy users to consume energy in a more flexible manner and fully utilize the capability of DR resources. With the integration of various energy in MESs, SEHs make energy users flexibly change the energy inputs responding to the requirements from the power system or the price signals of different energy. Without any loss of users’ comfort, energy users can switch the use of electricity, thermal energy, natural gas or other forms of energy. In addition, the capability of demand side resources can be fully exploited with the natural storage capability of thermal and gas systems. The surplus of renewable energy can be economically stored in thermal and gas systems and reused later. Hence, the MESs will benefit from the

182

6 Sharing Economy in Energy Systems Integration

IDR capability in terms of improving the global economy, enhancing the system reliability, etc.

6.6.3 Techo-Economic Analysis Coupling with different energy carriers in the SEHs, the strategic complementarity of various forms of energy makes it possible for all the consumers to actively participate in IDR programs without violating his/her comfort level. In practice, given the prices of electricity, heating, natural gas and other forms of energy, the consumers would like to manually switch the energy consumption to minimize the energy costs. To automatically implement the IDR programs, an electronic multi-energy system, supported by the advanced ICT (Information & Communication Technology) infrastructure needs developing. Therefore, it is worth conducting a techno-economic analysis of implementing the IDR programs. One major cost of IDR is to construct the multi-energy facilities and upgrade the sectors of various energy to make the consumers flexibly switch the energy sources. Electricity, cooling, heating and gas networks need tight interactions through various distributed energy resources (DERs) such as combined heat and power (CHP), HPs, GFs, ACs, refrigeration, etc. Similarly, the electricity, the fuel chain as well as the transportation sector will be highly coupled with the increase of hybrid vehicles. In the future, it becomes convenient for the consumers to switch natural gas and electric power as the fuel of the vehicles. With the development of smart grids, advanced metering, communications, control devices and other information technologies will support the implementation of IDR and become another major cost. The application of ICT for the control of the MESs will lead to the development of an integrated energy and communication system architecture. There have been a number of initiatives across the world making efforts to the use of real-time information with IDR programs to maximize the system operation efficiency while providing the consumers with flexible energy choices. For example, the GridWise and IntelliGrid in the USA and SmartGrids in the EU. The implementation of IDR requires a large amount of investment and operation costs to support the MESs and ICT infrastructures. However, there will be significant benefits brought by the technology upgrading. The benefits of IDR in both economic and environmental perspectives have been widely investigated in the existing research. IDR is able to break the barriers among different forms of energy, which highly improves the DR ability of the demand side resources. With the consumers’ active participation in the IDR programs, the operation efficiency of the energy systems can be further increased and the expansion planning can be effectively deferred. According to the simulation results in a real UK multi-energy district, the deployment costs of the district can be further minimized by switching different energy consumptions without compromising end-users’ comfort. The case studies demonstrate that

6.6 Integrated Demand Response

183

micro-CHP is an energy efficient technology that provides heat and electricity to households. With the optimal scheduling of IDR with micro-CHP, the operation costs for households can be reduced by up to 14%. The additional value created by IDR is the improvement of the environmental benefits of the energy systems. A better scheduling of different energy sectors can be achieved, leading to pollutant emission reduction. Moreover, with the complementarities of the heat and electric power, the consumers in the IDR programs can provide more ancillary services (e.g., spinning reserve service) for the electric power system instead of the conventional centralized power plants. Hence, the fuel costs and the pollutant emissions can be further decreased. Therefore, with the development of the MES and ICT, it is beneficial to promote the demand side resources towards integrated demand response.

6.6.4 Key Issues and Potential Research of IDR According to the literature review, the research and engineering projects about multienergy systems and integrated demand response have been widely investigated in the past decades. Most of the existing studies are focused on the optimal operation of MES considering DR. However, there remain more key issues and potential research to be addressed. (1)

The precise modeling of multi-energy consumption

The existing literature mainly focuses on the generation, distribution and conversion of different energy systems. The modeling of new-type energy carriers and the applications have been research highlights, including CCHP, fuel cells, energy storage systems, electric vehicles (EVs), natural gas systems, ground source heat pumps (GSHPs), ice storage, etc. However, it remains as open questions that (i) how to precisely model the multi-energy consumption sector to reveal the inertia characteristics of IDR; (ii) how to evaluate the contributions of the detailed models of multi-energy consumption. An illustrative figure is shown in Fig. 6.16. In contrast to electric power system where the power consumption must be balanced with the power supply at any instance, the heat/cooling and natural gas consumptions at the demand side are not instant. The processes of the thermal and gas consumptions must obey thermodynamics and hydromechanics principles, respectively. In addition, different facilities have distinct physical characteristics as well as working conditions. For example, the heat transfer rate and efficiency are different when the thermal energy is supplied for the indoor heating system and the water heater. In addition, the consumption flows are different when the natural gas is used for cooking and heating. A precise modeling of multi-energy consumption leads to accurate implementation of IDR. Therefore, it deserves an in-depth investigation in the future research. (2)

The scheduling strategy for MESs with IDR considering the influences of different energy prices

184

6 Sharing Economy in Energy Systems Integration

Voltage driven

Power supply and consumption must be balanced at any instance.

Temperature driven

Different heat transfer rate and efficiency with different facilities.

Precise modeling of energy consumption

Pressure driven

Different natural gas flows with different facilities.

Fig. 6.16 Illustration of the precise modeling of multi-energy consumption

Energy prices are critical incentive signals to lead the energy users to rationally consuming different forms of energy. Hence, it is imperative to incorporate welldesigned pricing mechanism in the scheduling strategies for MESs with IDR. According to the incentive compatible principle, energy prices should reflect the actual usage of the energy resources among each sector in different energy systems. However, in most existing literature, the energy prices are taken as exogenous conditions, which are predicted as constants or modeled as random variables. How to design effective prices according to the response characteristics of IDR resources still needs an in-depth study. For example, it remains as an open question how the consumers will respond to the price changes of different energy in a market-based environment. Another example is how to design effective price signals to alleviate the congestions in the distribution networks and pipelines by taking advantage of the complementarities of different energy. A scheduling framework of the MESs with IDR is illustrated in Fig. 6.17. On the upper level of the framework, the consumers optimize the consumptions of different energy given the energy prices. On the lower level of the framework, the energy prices are generated from the market clearing based on the scheduling of different consumers. The problem can be formulated as a non-cooperative game model. The model, the solution methodology and the environmental and economic impacts of the IDR scheduling have not yet been thoroughly investigated. (3)

The data-driven consumption strategy of IDR

In most existing research, the objectives of the energy users are to maximize the individual revenues in the scheduling. However, this may not be the case in the real world. To simulate the energy users’ irrational consumption behaviors, Experimental Economics and Consumer Psychology need to be studied by applying the data-driven statistical methods.

6.6 Integrated Demand Response

185

Fund flow Energy flow

Fig. 6.17 Illustration of scheduling for MESs with IDR considering the influence of different energy prices

With the development of advanced metering technology, the real-world energy consumption data can be collected. Then the data-driven technique is adopted to investigate the consumption strategies of IDR. In contrast to the conventional analytic models, the data-driven model and method can reflect the real consumption behaviors of the energy users. The data-driven consumption strategy of IDR is illustrated in Fig. 6.18. The applications of data-driven IDR are promising and there remain a number of research highlights to be addressed: (i) how to interpret the meanings of the data-driven IDR and make the data-driven IDR more practicable to the system operator; (ii) how data-driven IDR can be incorporated in the MES model and influence the operation of the MESs; (iii) how data-driven IDR will influence other market participators’ decision-making. (4)

The market operation and mechanism design considering IDR.

To the best knowledge of the authors, the market mechanism of IDR has not been thoroughly investigated. There remain a number of open questions about the market mechanism of IDR: (i) The optimal bidding strategy of IDR resources in the energy and ancillary service markets, which evaluates the capability of IDR as price-takers. (ii) The market equilibrium of multiple IDR resources; (iii) How to coordinate the bidding strategies of IDR resources in different energy markets along the time horizon, e.g., how to determine the available capacities in the day-ahead electricity markets considering the bi-lateral contracts in the mid-term natural gas market. (iv) How to design a market mechanism to effectively eliminate the asymmetric information in the demand side. These research topics deserve further investigation. A framework of the market research on IDR is shown in Fig. 6.19.

186

6 Sharing Economy in Energy Systems Integration

Technique Data cleaning Clustering Feature selection Data compression ...

Data input

Technique

Data-driven IDR model

Artificial Neural Network Support Vector Machine Decision Tree Machine Learning ...

Data output

Fig. 6.18 Illustration of the data-driven consumption strategy of IDR

Energy carrier

Electricity Bidding

Contract

Market equilibrium

Real-time

Day-ahead

Mid- & long-term

Model

Heat Gas

Time Research

Element Energy

Reserve

Regulation

Ramping

Service

Region Cross-region

Price-taker

Price-maker

Spatial scale

Fig. 6.19 The framework of the market research on IDR

Participator

6.7 Case Studies

187

6.7 Case Studies 6.7.1 Sharing Economy in Joint Electricity-Heat Markets The modified IEEE 118-bus system consists of three heating districts in Fig. 6.20. The users at buses 11, 17, 47, 51, 75 and 106 are planning SPHPs, including 10 PVs, 5 HPs and 5 TESs. The budget for each user is $ 450,000. After 4 iterations, the planning strategies reach equilibrium, and the computation time is 223.99 s. The planning strategies of the users at six buses are shown in Table 6.2. The annual total costs of the CHP system with/without investing in SPHPs are listed in Table 6.3. Heating district 2

Heating district 1

Potential SPHPs investment

Heating district 3

Fig. 6.20 Modified IEEE 118-bus system

Table 6.2 Planning strategies of the users at six buses Bus No

11

17

47

51

75

106

PV

6

6

3

4

3

5

HP

3

3

3

3

3

3

TES

4

5

5

5

5

5

188

6 Sharing Economy in Energy Systems Integration

Table 6.3 Annual total costs of the CHP system (Unit: million dollars) CHP system

Demand Electricity

Heat

Investment

Total

W/O SPHPs

726.19

683.52

123.66

0

807.18

W/ SPHPs

723.12

682.51

119.53

2.188

804.23

As one can observe, with the investment in the SPHPs, both the generation costs and the demand charges decrease. In the CHP system, the operational costs of the centralized thermal units can be reduced by 3.07 million dollars after investing in SPHPs. On the demand side, the total costs for the consumers can be reduced by 2.95 million dollars. Therefore, it is beneficial for the users to strategically invest in SPHPs. With the increase in the capacity of a PV, the annual generation costs and the invested PVs are shown in Fig. 6.21. As one can observe, the system generation costs decrease with the increase in the capacity of a PV. With the solar capacity increasing from 1 to 10 MW, the generation costs are reduced by 1.18%. In addition, when the solar capacities increase from 1 to 3 MW, the users will invest in additional PVs to maximize the individual benefits. However, because of the congestion of the power system, the revenues from PVs are limited with the increase in solar capacity. Thus, the number of PVs that are invested in slightly decreases. 40

725 724

35 30

722 25

Generation cost Invested PV

721 720

20

719

15

718

PV number

Cost (million dollars)

723

10

717 5

716

0

715 1

2

3

4

5

6

7

8

9

10

Capacity of a PV (MW)

Fig. 6.21 Annual generation costs and invested PVs with the increase in the capacity of a PV

6.7 Case Studies

189

6.7.2 Sharing Economy in Joint Electricity-Gas Markets The case is based on a 13-machine system. The system includes 10 thermal power units, 1 wind turbine, a gas turbine and a P2G system. The flexible ramp up/down service price is 247 $/ MW. The forecast error of wind power output and daily load is 10%, and the scheduling period is 24 h with 1 h as the interval. Considering P2G/ GT and FRP, the total operating cost of the system within 24 h is 5.87 × 105 dollars, and the wind power utilization rate is 96.3%. Power generation plan is shown in Fig. 6.22. The FRP demand curve considering the net load volatility and uncertainty is shown in Fig. 6.23. It can be seen that during the peak period of 10-12 h and 20 h, the thermal power unit needs to generate to respond users, so that it is difficult to reserve enough upward ramping capacity. Gas turbines are committed to assist in providing FRU. Since the user consumes less power during the wind power output increases, the net load drops rapidly at 13–17 h and 20–24 h. When there is only thermal power units in the power system, it is difficult to meet the system FRU demand due to the ramping rate limit, resulting in wind power curtailment. P2G starts or stops quickly, which can meet the needs of FRU. As can be seen, P2G meets most of the FRU requirements. Sufficient FRP capacity will fully address the system’s uncertainty and volatility, thereby increasing system flexibility. 1800 1600 1400

P(MW)

1200 1000 800 600 400 200 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Gen1

Gen2

Gen3

Gen4

Gen5

Gen6

Gen8

Gen9

Gen10

wind

GT

P2G

Fig. 6.22 Unit commitment Considering P2G/ GT and FRP

Gen7

Time(h)

190

6 Sharing Economy in Energy Systems Integration 200

Gen(u) GT FRU

150

Gen(d) P2G FRD

P(MW)

100 50 0 -50 -100 -150 -200

1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Time(h)

Fig. 6.23 FRU/FRD demand curve

6.7.3 Sharing Economy in Transportation-Energy Systems In this subsection, a modified IEEE 30-node electricity network and 16-node transportation network are used to verify the feasibility and effectiveness of the proposed transportation-energy model. The modified IEEE 30-node network has 41 branches and 6 thermal power units, with total installed capacity of 335 MW and wind power installed capacity of 80 MW. The four scenarios of wind power and loads are obtained by clustering the actual data of one year, r1−4 = [0.29, 0.40, 0.16, 0.14]. There are 18 roads in the transportation network. The transportation system and the power system are coupled at 5 charging / refueling stations. There are 1500 FCHEVs in the trans-energy system. To demonstrate the effectiveness and benefits of the proposed framework, three modes in the same system are compared in the case studies as follows (Table 6.4). (i) M1 is the proposed optimal schedule strategy of FCHEV in joint trans-energy systems. As a hub for coupling the energy system and the transportation system, FCHEV can participate in the economic dispatch of the joint system and utilize the hydrogen generated by P2G to absorb excess renewable energy. (ii) In M2, the EVs are interactive with the power system, and can be charged or discharged to alleviate the peak pressure of the power system. (iii) M3 is the traditional case that EVs are Table 6.4 Comparison case

Mode

Vehicle type

Interaction

M1

FCHEV

Yes

M2

BEV

Yes

M3

BEV

No

6.7 Case Studies

191

connected to the power system as inelastic load, and the randomness of the EVs will increase the operation cost of the power system. Compared with M2 and M3, M1 can optimize the allocation of resources in a larger range. At the same time, the integrated energy system of electricity and hydrogen also provides flexibility for economic dispatch. The combination of the power system and the transportation system can expand the spatial and temporal distribution of resources, thereby facilitating the optimal allocation of resources. Under the conditions of the same number of vehicles and arrival/departure time, the results of the UC problems in M1, M2, and M3 modes are calculated as follows: Figure 6.24a shows the wind power consumption of the three modes. The wind power utilization rate is the lowest in the M3 mode, and the wind curtailment reaches 414 MWh. The interaction between the EV and the power system in M2 mode has improved wind power consumption to a certain extent, but due to the limited charging time of the EV, the wind power at night cannot be absorbed. Compared with M2, the wind curtailment in M1 is reduced by 75.8%. In M1 mode, night wind power can be consumed through P2G, so the amount of wind curtailment is the least. Figure 6.24b shows the total load in different modes. The total load includes basic load, P2G load and EV charging load. Due to the centralized charging of EVs, the peak-valley difference of the total load in M3 mode reaches 172.4 MW, which is significantly larger than that of M2 and M1. If the number of vehicles in the system increases, it will cause power transmission congestion. Compared with M2, the peakvalley difference is reduced by 27.7%, which proves that FCHEVs can play a better role in peak load shifting compared with EVs. where Cop is the operation cost; Css is the startup-shutdown cost; Ct is the transportation cost;Uw is the wind utilization. Table 6.5 shows the system cost and wind power consumption in one day under different modes. Although the hydrogen transportation cost is taken into account in the M1 mode, the total operating cost is the lowest. On the one hand, because FCHEV can use the hydrogen generated by wind power at night, the utilization 100

Max

M1

M2

300

M3

M1

M2

M3

Power(MW

Power(MW)

80 60 40 20 0

1 3 5 7 9 11 13 15 17 19 21 23

Hour(h)

(a) Wind power consumption Fig. 6.24 Comparison of M1-M3

250 200 150 100

1 3 5 7 9 11 13 15 17 19 21 23

Hour(h)

(b) Total load

192

6 Sharing Economy in Energy Systems Integration

Table 6.5 Cost comparison in different modes Mode

Cop ($)

Css ($)

Ct ($)

Total cost($)

Uw (%)

M1

10,887

500

740

12,127

88.23

M2

12,636

900

/

13,536

61.95

M3

12,924

1400

/

14,324

53.93

rate of wind power is greatly improved, and the operating cost is reduced; On the other hand, as seen in Fig. 6.24b, the peak-valley difference in M1 mode is small, since the thermal power unit does not need to start up and shut down frequently to meet the load volatility. Thus the start-up/shut-down costs are reduced. Compared with M2, the transportation cost in M1 increases, but due to the increase in wind power utilization, the operating and start-up/shut-down costs are reduced. In this case, the hydrogen energy system reduces the power system cost by about 10.4%.

6.8 Conclusion This chapter presents the sharing economy in energy systems integration. The structure and concept of integrated energy sharing market are introduced and the value of sharing economy in joint electricity-heat, electricity-gas and transportation-energy systems are analyzed. In addition, we study the value and key issues of integrated demand response. In this chapter, the case studies are carried out in terms of the aforementioned three aspects. In electricity-heat markets, sharing economy enables the coupled trading of electricity and heat. Both electric and heating sectors benefit from the increased use of solar-powered heat pumps with lower operational costs and higher generation flexibility. In electricity-gas markets, the flexibility of power-togas devices can be aggregated to provide power systems with flexible ramping products, thus enhancing the reliability of electric power grids. In transportation-energy systems, we propose an energy sharing framework considering hydrogen supply chain from water electrolysis production to fuel cell hybrid electric vehicle consumption. Cases demonstrate that the popularization of FCHEVs can help accommodate surplus renewable energy while alleviating users’ mileage anxiety.

References 1. National Development and Reform Commission, National Energy Administration. Revolutionary strategy of energy production and consumption. (2016–2030)[EB/OL], 17 May 2017. https://www.ndrc.gov.cn/fggz/fzzlgh/gjjzxgh/201705/t20170517_1196767.html 2. National Energy Administration. Announcement of the Comprehensive Department of the National Energy Administration on Soliciting Opinions on the Guiding Opinions on Establishing and Improving the Long-term Mechanism of Clean Energy Absorption (Draft for

References

193

Comments)[R/OL], 19 May 2020. http://www.nea.gov.cn/2020-05/19/c_139069819.html 3. U.S. Energy Information Administration: Energy consumption by sector. Available: https:// www.eia.gov/totalenergy/data/monthly/pdf/sec2.pdf 4. International Energy Agency, “CHP/DHC country scorecard: Finland,” Available: https://www. iea.org/media/topics/cleanenergytechnologies/chp/profiles/Finland.pdf 5. International Energy Agency: CHP/DHC country scorecard: Finland. Available: https://www. iea.org/media/topics/cleanenergytechnologies/chp/profiles/germany.pdf 6. Wu, D.W., Wang, R.Z.: Combined cooling, hearing and power: a review. Prog. Energy Combust. Sci. 32, 459–495 (2006) 7. Guo, H., Chen, Q., Xia, Q., Zhou, P.: Flexible ramping product in electricity markets: basic concept, equilibrium model and research prospect. In: Proceedings of the CSEE, vol. 37, no.11, pp. 3057–3066+3361 (2017) (in Chinese). 8. Xu, L., Tretheway, D.: Flexible ramping products—draft final proposal. California ISO (2014) 9. Midcontinent ISO (MISO). [Online]. Available: https://www.misoenergy.org/Pages/Home. aspx 10. Saber, A.Y., Venayagamoorthy, G.K.: Plug-in vehicles and renewable energy sources for cost and emission reductions. IEEE Trans. Industr. Electron. 58, 1229–1238 (2011) 11. European Commissio,: A Clean Air Programme for Europe (2013) 12. Staffell, I., Scamman, D., Velazquez Abad, A., Balcombe, P., Dodds, P.E., Ekins, P., Shah, N., Ward, K.R.: The role of hydrogen and fuel cells in the global energy system. Energy Environ. Sci. (2019) 13. Rifkin, J.: The third industrial revolution: how lateral power is transforming energy, the economy and the world. Palgrave MacMillan, New York (2011) 14. The State Council: Guidelines of the state council on actively pushing “Internet+” action. Available via DIALOG. http://www.gov.cn/zhengce/content/2015-07/04/content_10002.htm. Accessed 9 June 2016

Chapter 7

Sharing Demand Side Resources for Regional Market Bidding

7.1 Introduction Sharing economy refers to a market model that enables individuals or entities to share their idle resources with others upon payment for the purpose of efficient resource allocation and social welfare maximization. An important benefits of sharing economy is to realize the aggregation and utilization of idle resources. Up till present, sharing economy-based business models have won huge success in housing and transportation fields by matching individuals to offer/enjoy underutilized products like ridesharing (Uber, Lyft) and real estate (Airbnb, Expedia). In the field of power system, it is of great significance to aggregate distributed energy resources (DERs) and demand side resources (DSRs) to respond to the demand of electricity market. Essentially, the interconnected transmission and distribution networks of energy systems provide a natural platform for energy sharing. In bulk power systems, for example, PJM has initiated coordinated transaction scheduling (CTS) via inter-regional tielines with NYISO and MISO since 2014 and 2017, respectively, to improve generation utilization and enhance price predictability. Besides, transactive energy-related pilot projects have been launched by Pacific Northwest National Laboratory (PNNL), enabling DER owners to share demand side resources through distribution grids with their neighbors while smoothing out the fluctuations of electric loads. The development of renewable energy has drawn attention across the world in the past decade. California, for example, announced its ambitious goal of achieving a 50% renewable portfolio standard by 2030 [1]. While the integration of renewable energy contributes to a more sustainable future, the variability and uncertainty of the renewable sources pose great challenges to the economic and reliable operations of the power system [2]. With the increasing penetration of renewable energy, rapid ramping of generation resources may be insufficient to smooth out the huge fluctuations in renewable energy production. Thus, it is critical to facilitate the accommodation of renewable generation while economically and reliably operating the power system. © Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_7

195

196

7 Sharing Demand Side Resources for Regional Market Bidding

The concept of the aggregate bidding based on sharing economy assumes a cluster of loads and DERs operating as a single controllable system in the regional market. Taking advantage of the synergy among various DERs, the renewable generators can cooperate with controllable energy resources to provide both energy and ancillary services (ASs) for the bulk power system [3]. For example, in a stand-alone mode, a wind farm must deviate from its maximum power output status and leave a margin to provide ramping services for the system. However, in an aggregate bidding based on sharing economy, the wind farm is able to leave the ramping margin by charging a Na/S battery without deviating from its maximum power. Hence, an aggregate bidding based on sharing economy can stably provide both energy and ASs by integrating various DERs [4]. From the system point of view, the aggregate bidding based on sharing economy show the advantages of low investment costs, low pollutant emission and high operational flexibility. The flexibility of a regional power systems (RPS) provided by the DERs can be aggregated for power system operations, thereby replacing high-cost centralized units and deferring the generation expansion. In addition, the RPS have abundant DSRs, efficiently offering capacities to meet the local requirements [5]. Compared with centralized thermal units, sharing economybased RPS can achieve localized energy balance without the loss accompanied with long-distance power transmission and difficulties caused by transmission congestions. Therefore, the concept of the aggregate bidding based on sharing economy provides new insights for exploring the grid-friendly manner of DERs and DSRs. Aggregate bidding based on sharing economy has many benefits, e.g., participating in the market of different trading varieties to improve economy, to enhance capacity adequacy while ensure power energy security and to promote renewable energy accommodation toward carbon abatement.

7.2 Sharing Demand Side Resources in Wholesale Markets 7.2.1 Co-optimization of Energy and Ancillary Service Markets Without loss of generality, the co-optimization of energy and AS markets is implemented in this chapter [6]. In the pool-based day-ahead markets, it is assumed that a sharing economy-based regional power system can simultaneously bid in joint energy and AS markets. Considering its relatively small capacity, a sharing economy-based RPS is reasonably assumed to be a price-taker. As a controllable entity, due to the mechanism of sharing economy, the RPS will strategically allocate available capacities in day-ahead markets to maximize the revenues. The RPS bids energy and ASs, including spinning reserve service and flexible ramping products (FRPs) [7]. The FRP provided by a sharing economy-based regional power system is referred to the potential power output change from time slot t to t + 1, which reflects the available capacity of the RPS reserved for the power system to satisfy load following.

7.2 Sharing Demand Side Resources in Wholesale Markets

Energy market

Wind turbine Sharing Solar station economyEnergy storage Integration based RPS

197

Market

Ancillary service market

bidding

Micro turbine

Spinning reserve

Market clearing

Upward FRP

Results released

Downward FRP

DERs & DSRs

t0

t1

t2

Fig. 7.1 Framework of a sharing economy-based RPS’s participation in day-ahead energy and AS markets

FRPs are specifically designed to relieve the system-wide ramping constraints, which are first introduced in California and MISO markets in the United States. The models and applications of FRPs have been investigated recently. In [8], the mathematical model of FRPs is formulated according to the California market. In [9], the security-constrained economic dispatch (SCED) model is presented to incorporate the ramping constraints. Numerical results demonstrate the effectiveness of the ramping constraints for reducing the instances of short-term scarcity conditions. In [10], a risk-constrained SCED scheme is proposed to optimize the dispatch and provision of FRPs. With the increasing demand for the ramping resources in the power system, the FRP market is expected to be fully operational in the near future. The framework of a sharing economy-based RPS’s participation in day-ahead energy and AS markets are shown in Fig. 7.1. In this framework, integrated with various DERs, a sharing economy-based RPS bids in the joint energy, reserve and FRP markets at time t 0 , when the independent system operator (ISO) opens the dayahead market. Then the day-ahead market bid period closes at time t 1 . The ISO begins to run the market clearing software to determine the hourly dispatch schedules and the locational marginal prices for the day-ahead market. Finally, the market clearing results will be released at time t 2 . The co-optimization of the energy and AS markets is implemented in most electricity markets operated by the ISO [11]. The bids of the sharing economy-based RPS must be determined before the closure of the day-ahead markets for the next day [12]. By introducing FRP markets, on one hand, the sharing economy-based RPS is able to increase the revenues from market bidding; on the other hand, the power system will benefit from the improvement of the dispatch flexibility and the adequacy of the ramping resources.

198

7 Sharing Demand Side Resources for Regional Market Bidding

7.2.2 Uncertainty Modeling In this chapter, a hybrid stochastic/robust optimization approach is adopted to address the uncertainties in renewable generation and day-ahead market prices [6, 12]. The prices in the energy and AS markets are modeled via scenario-based stochastic programming. The uncertainties in wind and photovoltaic power are addressed using RO. For market prices, a sharing economy-based RPS is concerned with the profiles of market prices to optimally allocate its available capacities in each market. The prices in different markets have strong correlations, which cannot be modeled with independent confidence intervals. For example, both prices of energy and spinning reserve are relatively high during peak hours because of high load demands. In addition, to allocate the capacities in different markets, the relative differences of prices are the major concern instead of the absolute values of market prices. Therefore, stochastic programming with multiple price scenarios is more appropriate than RO to model the uncertainty of market prices [12]. For renewable generations, the absolute capacities of wind and photovoltaic power have large impacts on the bidding strategy of an RPS as well as the operation of the DERs in the sharing economy-based RPS. Moreover, because the intervals of wind and photovoltaic power can be obtained according to historical data, RO is an effective tool to address the uncertainties in renewable generation. Therefore, the robust mixed-integer linear programming (RMILP) in [13] is applied in this chapter. The available power of the WT and PV in the sharing economy-based RPS at time slot t, denoted by PtAWT and PtAPV , are modeled as independent and bounded random variables. Under a confidence level σ, PtAWT takes values from the minimum power AWT APV to the maximum power P t , while PtAPV takes values from P APV to P t . To P AWT t t obtain the confidence intervals of PtAWT and PtAPV , the forecast errors are analyzed based on historicalWT datasets. For example, P t and PtWT are the point forecast and actual wind power at time slot t. Based on the historical data from the Wind Integration Datasets of the NREL  WT WT of wind power forecast errors εtWT [14], the probability distribution f WT P t /Pmax can be acquired, i.e., 



εtWT =

  Pˆ tWT −PtWT WT WT ˆ WT WT , ε ∼ f /P P t t max WT Pmax

(7.1)

Then the upper and lower bounds of wind power forecast errors under the WT WT confidence level σ, εt,min and εt,max , can be calculated as follows: WT εt,min

⎧ ⎨

 w ⎫   1−σ ⎬ = Inf w ∈ [0, 1] f WT (x)dx ≥ ⎩ 2 ⎭  0

(7.2)

7.2 Sharing Demand Side Resources in Wholesale Markets 50

Positive error Negative error

40

Forecast error (%)

Fig. 7.2 The intervals of forecast errors with 95% confidence level under different levels of forecasted wind power

199

30 20 10 0 -10 -20 -30 0

20

40

60

80

100

Forecasted wind power (% installed capacity)

WT εt,max

⎧ ⎨

 w ⎫   1+σ ⎬ = Inf w ∈ [0, 1] f WT (x)dx ≥ ⎩ 2 ⎭ 

(7.3)

0

Figure 7.2 shows the intervals of forecast errors with 95% confidence level under different levels of forecasted wind power. According to the forecasted wind power, the minimum power and the maximum power can be calculated as follows: WT WT WT =Pˆ tAWT − Pmax · εt,max , εt,max >0 P AWT t

(7.4)

WT WT WT =Pˆ tAWT − Pmax · εt,min , εt,min 0. indicates that the carbon allowance is sold, and E C − Es < 0 means the RPS purchases the carbon allowance. Note that as the carbon price is constant in DA and RT, there is no need to apply the two-settlement mechanism. In the traditional case without IoT, PEVs are charged according to users’ preferences, which behave as fixed load rather than a controllable storage. During day hours, the solar power may threaten the security threshold of the RPS (e.g., voltage and power flow limits), thus leading to clean energy curtailment. During night time, distributed generation (e.g., micro-turbines and gas furnaces) may come online to balance local load and submit market bids, thereby producing carbon emissions. However, IoT allows the PEVs to be aggregated as a storage system, shifting the surplus solar power for the night-time load. As a result, the accommodation of renewable energy can be facilitated, the market revenues of the RPS can be improved.

7.4.2 Model of Electric Vehicle Fleets In this section, a charging station can schedule the charging/discharging processes of PEV fleets to operate as a battery storage. The operational flexibility of a charging station is determined by the available capacity of PEVs. The behavior of each PEV is characterized by a set of parameters, including arrival time (Tiarr ), departure time dep dep (Ti ), initial state of charge (SOCiarr ), target SOC (SOCi ) and the feasible region of output power. Therefore, the charging station model is formulated as follows: PtCSC =



EVC Pi,t , PtCSD =

i∈EV



EVD Pi,t , ∀t,

(7.55)

i∈EV

  CS , ∀t, PtCSC , PtCSD ∈ 0, Pmax

(7.56)

EVC EV EVC 0 ≤ Pi,t ≤ αi,t Pi,t,max , ∀i, t,

(7.57)

EVD EV EVD 0 ≤ Pi,t ≤ αi,t Pi,t,max , ∀i, t,

(7.58)

EV EV SOCi,t = SOCi,t−1 +

1 EV Ei,max

 EV EVC  EV  EVD ηi Pi,t−1 − Pi,t−1 ηi , ∀i, t, dep

EV arr EV SOCi,T arr = SOC i , SOCi,T dep ≥ SOCi , ∀i, i i

(7.59) (7.60)

7.4 Sharing Demand Side Resources for Carbon Trading

211

EV SOCi,t ∈ [0, 1], ∀i, t,

(7.61)

  where EV = 1, 2, . . . , N EV represents the set of PEVs, where N EV is the PEV number; PtCSC and PtCSD are the total charging and discharging power of the charging EVC EVD CS and Pi,t are the charging and discharging power of the ith EV; Pmax is station; Pi,t EV the capacity of the charging station; αi,t is a parameter indicating the plug-in status of the ith EV, expressed as follows:  EV αi,t

=

dep

0, t < Tiarr or t > Ti , dep 1, Tiarr ≤ t ≤ Ti .

(7.62)

In Eq. (7.55), the power of the charging station is the aggregation of PEV fleets. Equation (7.56) shows the capacity of the charging station. In constraints (7.57) and (7.58), the charging and discharging power of an EV are restricted by its maximum power. In Eq. (7.59), the SOC of an EV is related to the charging and discharging energy during a time slot. In constraint (7.60), the initial SOC is set to the SOC of an EV in arrival time, and the SOC in departure time must be greater than the target. Note that we assume the SOC of each EV can vary from 0 to 1 shown in (7.61).

7.4.3 Optimal Bidding Model In this section, an optimal market bidding model of a regional power system is proposed considering the participation in electricity and carbon markets.

7.4.3.1

Bidding Model

The problem of the optimal bidding strategy is formulated as a two-stage stochastic linear programming model as follows: max

XDA ,XRT



γs REs,t +

s∈S t∈T



γs RCs −

s∈S

 

MT ,RT γs ci Pi,s,t ,

(7.63)

s∈S t∈T i∈MT

s.t. PtDA =

 i∈MT

MT ,DA Pi,t +



PV ,DA Pi,t + PtCSD,DA

i∈PV

− PtCSC,DA − PtL,DA , ∀t,

(7.64)

212

7 Sharing Demand Side Resources for Regional Market Bidding RT Ps,t =



MT ,RT Pi,s,t +

i∈MT CSC,RT − Ps,t

−



PV ,RT CSD,RT Pi,s,t + Ps,t

i∈PV L,RT Ps,t , ∀s, t,

 RPS RPS  RT PtDA , Ps,t ∈ −Pmax , Pmax , ∀s, t, Es =

 

MT ,RT ei Pi,s,t , ∀s,

(7.65) (7.66) (7.67)

t∈T i∈MT MT ,DA MT ,RT MT MT ≤ Pi,t − Pi,s,t ≤ δiADJ Pi,max , ∀i, s, t, −δiADJ Pi,max

(7.68)

  MT ,DA MT ,RT MT Pi,t , Pi,s,t ∈ 0, Pi,max , ∀i, s, t,

(7.69)

 

PV ,DA PVF,DA PV ,RT PVF,RT , Pi,s,t , ∀i, s, t, Pi,t ∈ 0, Pi,t ∈ 0, Pi,s,t

(7.70)



L PtL,DA ≥ Emin ,

(7.71)

L,RT L Ps,t ≥ Es,min , ∀s,

(7.72)

t∈T

 t∈T

 

L,DA L,DA L,RT L,RT L,RT , Ps,t , ∀s, t, , Pt,max ∈ Ps,t,min , Ps,t,max PtL,DA ∈ Pt,min

(7.73)

constraints (7.55)–(7.62) for DA and RT, where the decision variables are denoted by XDA and XRT , representing the DA- and RT-related variables, respectively, including RPS’s DA energy bids PtDA , the power MT ,DA PV ,DA of the ith MT Pi,t , the power of the ith PV Pi,t , the charging/discharging CSC,DA CSD,DA L,DA power of the station Pt / Pt , the load Pt , and the associated variables with the superscript “RT”; T , MT and PV are the sets of time slots, microRPS is the capacity of the RPS, turbines and photovoltaic systems, respectively; Pmax indicating the capacity of a substation or a distribution grid; ci and ei are the cost MT is the capacity of the ith MT; δiADJ and carbon emission rate of the ith MT; Pi,max PVF,DA PVF,RT is the limit for the RT adjustment power of the ith MT; Pi,t and Pi,s,t are L the DA forecasted power and RT scenario-related forecast, respectively; Emin and L are the DA forecasted daily load requirement and RT forecast, respectively; Es,min L,DA L,DA L,RT L,RT Pt,min , Pt,max , Ps,t,min and Ps,t,max are the DA and RT minimum and maximum load, respectively. The objective of the bidding strategy is to maximize the DA profits anticipating the RT adjustments. In (7.63), the first term is the profits from DA and RT markets, the second term is the profits from carbon trading and the third term is the generation costs of MTs. In Eqs. (7.64) and (7.65), the energy bids of the RPS equal the power from

7.4 Sharing Demand Side Resources for Carbon Trading

213

MTs, PVs and PEVs minus electric load. In (7.66), the bidding of the RPS is restricted by its capacity. Equation (7.67) shows the carbon emissions are released from MTs. In (7.68), the RT adjustments-of MTs are bounded. Constraints (7.69), (7.70) and (7.73) show the lower and upper limits of MTs, PVs and load. Constraints (7.71) and (7.72) show the minimum requirement for daily load. Note that the constraints for the charging station (7.55)–(7.62) are extended to DA and RT forms.

7.4.3.2

Benefits of Sharing Economy

The potential benefits of sharing economy are analyzed in Fig. 7.6. Figure 7.6a illustrates the traditional case without IoT, in which PEVs behave as fixed load rather than a controllable energy storage. During peak hours, gas-fired MTs have to start up for load balance, thus yielding carbon emissions. However, when solar power is extremely high at noon, the RPS may not have sufficient hosting capacity to accommodate all the solar power, and some clean energy has to be curtailed. In Fig. 7.6b, an illustration for the IoT-enabled operation is provided, in which PEVs are aggregated and managed as a battery storage. In practice, a number of PEVs can work in a relatively flexible range while satisfying the required SOC at departure time. Hence, PEVs can absorb the surplus solar energy at noon, improving the energy efficiency of the RPS. On the other hand, PEVs can discharge during peak hours in place of MTs, which significantly reduces the carbon emissions. With the help of IoT, the accommodation of renewable energy can be facilitated toward beneficial market bidding, thus achieving Pareto Improvement.

Fig. 7.6 Benefit analysis of sharing economy brought on by the IoT technology

214

7 Sharing Demand Side Resources for Regional Market Bidding

7.5 Case Studies 7.5.1 Wholesale Markets 7.5.1.1

Basic Data

Historical data of the Electric Reliability Council of Texas (ERCOT) day-ahead market prices [18] from July 1, 2016, to September 30, 2016, are used to generate 20 typical scenarios to address the uncertainties in day-ahead market prices. These price scenarios are generated by K-means clustering. The average hourly prices in the energy and AS markets are shown in Fig. 7.7. The wind and solar power are the real-world data from a wind farm and a photovoltaic station in a province in China. The parameters of the other DERs in the RPS are shown in Table 7.1. To evaluate the ramping capabilities and the benefits of the RPS in joint energy and AS markets, three strategies are considered:

(ii) (iii)

S1, where the RPS bids in the joint energy, reserve and FRP markets with tWT = tPV = 1; S2, where the RPS bids in the joint energy and reserve markets with tWT = tPV = 1; S3, where the RPS only bids in the energy market with tWT = tPV = 1.

Energy and reserve price ($/MWh)

160

20

Energy Reserve Upward FRP Downward FRP

140 120

18 16 14

100

12

80

10 8

60

6

40

4

20 0

2 2

4

6

8

10

12

14

16

18

20

22

Time (h)

Fig. 7.7 Average day-ahead hourly prices in the energy and AS markets

24

0

FRP price ($/MWh)

(i)

7.5 Case Studies

215

Table 7.1 Parameters of the DERs in the RPS DER

ciMT ($/MWh)

MT ,RAMPU Pi,max (MW/h)

MT (MW) Pi,max

MT ,RAMPD Pi,max (MW/h)

MT-1

13

3

2

2

MT-2

10

4

2

2

MT-3

18

2

2

2

ESS-1

ESS ηi,α

ESS ηi,β

ESS SOCi,min

ESS SOCi,max

0.95

0.95

0.1

0.9

CiESS (MWh)

ESS (MWh) Ei,0

ESS Pi,α,max (MW)

ESS Pi,β,max (MW)

12

6

2

2

ESS ηi,α

ESS ηi,β

ESS SOCi,min

ESS SOCi,max

0.95

0.95

0.1

0.9

CiESS (MWh)

ESS (MWh) Ei,0

ESS Pi,α,max (MW)

ESS Pi,β,max (MW)

20

10

2.5

2.5

ESS-2

7.5.1.2

Optimal Bidding Strategy

In S1, the RPS bids in joint energy, reserve and FRP markets. The optimal bidding strategies of the RPS are shown in Fig. 7.8. The RPS strategically allocates the available capacity in each hour to maximize the revenues from the joint energy and AS markets. The bidding strategy of the RPS depends on the physical constraints of the DERs and the opportunity costs in each market. In the energy market, the RPS will generate electricity for the bulk power system, which leads to the operational costs of the RPS. In the AS markets, the RPS will leave a margin for the AS capacities, which may not cause the operational costs. As shown in Fig. 7.7, when the prices of spinning reserve are high and the opportunity Fig. 7.8 Optimal bidding strategies of the RPS in the base case

Energy Reserve

15

Upward FRP Downward FRP

Capacity (MW)

10 5 0 -5 -10 -15 2

4

6

8

10

12

14

Time (h)

16

18

20

22

24

7 Sharing Demand Side Resources for Regional Market Bidding PV WT

Capacity (MW)

16

ESS-1 ESS-2

MT-3 Load

MT-1 MT-2

ESS-1 energy ESS-2 energy 16

14

12

12

8

10

4

8

0 -4

6

-8

4

-12

2

-16

0

Stored energy (MWh)

216

-20

-2

-24

-4

-28 2

4

6

8

10

12

14

16

18

20

22

24

Time (h)

Fig. 7.9 Optimal bidding strategies of the DERs in the energy market

costs are relatively low, the capacity is provided for reserve instead of bidding in the energy market. Similar conclusions can be drawn from the bidding for the upward FRPs. In addition, by curtailing renewable generation, decreasing the MTs’ output and making the ESSs charge, the RPS is able to provide downward FRPs. The optimal bidding strategies of the DERs in the energy market are shown in Fig. 7.9. Because the operational costs of wind and photovoltaic power are zero, the capacity of WT and PV is fully used in the energy market to maximize the energy revenues. The operational cost of M-2 is relatively low; thus, all the available capacity is provided for energy. However, the costs of the other two MTs are higher, thereby driving MT-1 and MT-3 to bid the available capacity for ancillary services during some periods. In the process of arbitrage, the ESSs will charge during the valley hours and discharge during the peak hours. In addition, because the ESSs can flexibly adjust the consumption or production, the ESSs will strategically bid for energy and ancillary services. The energy capacity of the RPS is equal to the difference between the capacity offered by the DERs and the load demands. The expected revenues of the DERs in different markets are shown in Table 7.2. By strategically allocating the capacity of the DERs in different markets, the RPS can obtain the optimal expected revenues with 20.35% energy, 67.77% reserve and 11.88% FRP. As one can observe, it is beneficial for the RPS to participate in the joint energy and AS markets, in which FRPs are also important fractions.

7.5.1.3

Comparative Study

Table 7.3 shows the expected revenues from each market in Case 1.

7.5 Case Studies

217

Table 7.2 Expected revenues of the DERs in different markets Revenue WT

Energy ($) 275.99

Reserve ($)

FRP ($)

0

Total revenue ($)

15.01

291.00

PV

1189.46

0

16.71

1206.17

MT-1

1470.21

419.02

78.30

1967.53

MT-2

2805.89

0

75.68

2881.57

MT-3

797.51

123.88

1179.38

ESS-1

−534.68

257.99

1598.88

235.74

1299.94

ESS-2

−687.50

2014.55

301.70

1628.75

RPS

1450.18

4829.96

847.02

7127.16

Table 7.3 Expected revenues from each market in Case 1 Revenue

Energy ($)

Reserve ($)

FRP ($)

Total revenue ($)

S1

1450.18

4829.96

847.02

7127.16

S2

1498.84

5099.03

0

6597.87

S3

5713.36

0

0

5713.36

Comparing the results in S1 with those in S2, one can observe that the RPS can increase its revenues by 8.02% if providing FRPs. Comparing the results in S1 and those in S3, one can observe that the RPS can increase its revenues by 24.75% if participating in joint energy and AS markets. Therefore, by participating in joint energy, reserve and FRP markets, the RPS can further increase its revenues from the day-ahead markets. Meanwhile, the RPS is able to provide ramping capacities for the bulk power system, which fully utilizes the grid-friendly potentials of the RPS.

7.5.2 Available Transfer Capability Enhancement 7.5.2.1

Basic Data

To depict the uncertainty in renewable power and load, we apply k-means clustering to generate 10 daily scenarios from yearly data. The IEEE 14-bus system with 24-h time slots is studied, and the schematic is shown in Fig. 7.10. There are five thermal generators. The startup and shutdown costs are set to $1000 × PG max for each thermal generator [19]. The minimal power of each thermal generator is half of the installed capacity. A wind farm and a solar station are located at bus 4 and 7, respectively. The wind power and load profiles are collected from the PJM market in the U.S. [20], and the solar power data are from the National Renewable Energy Laboratory [21]. The installed capacities of wind and solar are

218

7 Sharing Demand Side Resources for Regional Market Bidding

G

G

1 S

2

W G

3

G

5

6

12

11

13

4

ATC

G

7

10 9

8

14

Fig. 7.10 IEEE 14-bus system

Table 7.4 Parameters of energy storage

Parameter

Maximal power (MW)

Capacity (MWh)

Efficiency

Value

12

96

90%

360 MW and 70 MW, respectively. There is an interchange line starting from bus 7, and the selling price is $15/MWh. The minimum accommodation rate of renewable energy is 75%. A commercial ES is located at bus 10. The parameters of the ES are shown in Table 7.4 Note that the sustaining time (ST) of the ES is 96/12 = 8 h, indicating the amount of hours that the ES can sustain when discharging at maximal power.

7.5.2.2

Impacts of Energy Storage

The daily ATC and system load in the cases without and with ES are compared in Fig. 7.11. In contrast to the case without ES, the daily total ATC can be improved by 15.98% during peak hours and 17.95% during off-peak hours with ES. The daily maximal ATC increases from 107.48 to 109.24 MW, and the daily minimal ATC increases from 13.48 to 64.56 MW. In addition, 29.94 MWh load demands can be effectively reduced during peak hours by using ES, thus significantly enhancing the reliability of interchange power trading. The charging/discharging power and stored energy of the ES are shown in Fig. 7.12. As one can observe, the ES charges electricity during valley hours and discharges during peak hours, which achieves load shifting and enhances real-time ATC. Therefore, with the responsive capability of ES, renewable energy can be further accommodated. The accommodation rates of renewable energy are shown in Fig. 7.13.

7.5 Case Studies

219

W/ ES W/O ES

Peak Hours

W/O ES W/ ES W/O ES W/ ES Off-peak hours Peak hours (10:00-21:00)

Fig. 7.11 Daily ATC and system load without and with ES 100

Discharging Power Charging Power Stored Energy

10

80

Power (MW)

5 60 0 40 -5 20

-10

-15 0

5

10

15

20

0 25

Time (h)

Fig. 7.12 Charging/discharging power and stored energy of ES Fig. 7.13 Accommodation rates of renewable energy without and with ES

Windpower Solarpower Maximum Accommodation Rate

3.4%

Stored Energy (MWh)

15

220

7 Sharing Demand Side Resources for Regional Market Bidding

From the comparative results, energy storage in the demand side can increase the renewable energy accommodation by 126.68 MWh, and the accommodation rate can be improved from 75.0 to 77.5%. Moreover, the maximum accommodation rate increases from 77.0 to 80.4%.

7.5.2.3

Cost and Benefit Analysis

Table 7.5 lists the generation costs and the profits from interchange power trading. The negative value of interchange power trading cost indicates the power trading can earn profits. As one can observe, the total costs can be effectively reduced by 47.53% after integrating ES. With more renewable energy consumed after using ES, the generation costs of thermal generators decrease from 146.98 to 97.47 thousand dollars. More importantly, ES can flatten system load profiles and thus avoid thermal generators frequently switching on/off. The startup and shutdown costs are dramatically reduced. Figure 7.14 shows the hourly online capacity of thermal generators without and with ES. In the case with ES, the thermal generators do not need to start up or shut Table 7.5 Generation costs and profits from interchange power Cost structure Generation costs

W/ES (103 $)

W/O ES

97.47

146.98

Startup and shutdown costs (103 $)

440.00

830.90

Interchange power trading costs (103 $)

−48.71

−46.37

Total costs (103 $)

488.76

931.51

Unit 4 on

Unit 4 off

Unit 5 on

Fig. 7.14 Online capacity of thermal generators

Unit 5 off

7.5 Case Studies

221

down frequently. However, without ES, Unit 4 and 5 have to shut down at 2:00 and 21:00, respectively, to accommodate renewable energy, and start up at 20:00 and 5:00, respectively, to satisfy the peak loads. By integrating ES, the generation adequacy of the power system can be highly improved, thereby enhancing the ATC of interchange transmission networks.

7.5.3 Energy and Carbon Markets 7.5.3.1

Basic Data

A regional power system with 50 electric vehicles, 10 PV systems and 5 MTs is selected as the base case. To depict the uncertainty in PEVs, solar power, load and electricity market prices, 10 scenarios are generated from the yearly data and probability distributions. The battery capacity of each EV is shown in Table 7.6. Here, the electricity market prices are collected from the PJM market [22]. The solar power data are collected from [23]. The price of carbon emission credits is set to 12.3 $/ton [16], and the daily carbon emission cap is set to 125.0 kg. The charging/discharging efficiency of PEVs is 95%, and the maximal power of an EV is 5 kW. Electric load data are collected from the PJM market [24], and the hourly adjustment range is set to 0.9 and 1.1 times of the original data. The maximum bidding capacity of the RPS is 250 kW. The parameters of MTs are listed in Table 7.7. The RT adjustment limit of MTs is 10%. To demonstrate the effectiveness and benefits of the proposed framework, three methods are compared in the case studies: (i) M1, which is the proposed framework with V2G-enabled controllable PEV fleets; (ii) M2, in which V2G is not available but PEV fleets can be controlled to change the charging rates; and (iii) M3, which is the traditional case without IoT, indicating that PEV fleets are charged at the maximum rate as soon as they arrive. Table 7.6 Battery capacity of 50 electric vehicles Capacity (kWh)

15

20

25

30

35

40

45

Number

8

6

10

6

10

4

6

Table 7.7 Parameters of 5 gas-fired micro-turbines

No

Capacity (kW)

Cost ($/kWh)

Emission rate (kg/kWh)

1

50

0.055

0.76

2

50

0.065

0.90

3

50

0.075

1.04

4

50

0.085

1.18

5

50

0.095

1.32

222

7.5.3.2

7 Sharing Demand Side Resources for Regional Market Bidding

Optimal Bidding Strategy

The DA optimal bidding strategies of the RPS and PEVs are shown in Figs. 7.15 and 7.16. In Fig. 7.15, the RPS can bid more surplus power in M1 than in M2 and M3, especially during peak hours from 8:00 to 11:00. This is because the PEVs can flexibly charge and discharge in M1. As shown in Fig. 7.16, the y-axis represents the difference of the discharge and charge power. In M1, the PEVs charge from 3:00 to 6:00 when market prices are low, or from 11:00 to 16:00 when solar power is high. During peak hours from 9:00 to 10:00 and from 17:00 to 21:00, the PEVs discharge to satisfy electric load in place of MTs. In M2, the PEVs can only adjust the charging rates, and thus charge at noon to accommodate the surplus solar power. In M3, however, the PEVs are charged according users’ preferences, i.e., at the maximum charging rate after arrival. Hence, without the energy management brought on by IoT, a large number of the PEVs are charged during peak hours. The solar power and carbon emissions are shown in Fig. 7.17. Fig. 7.15 DA optimal bidding strategies of the RPS

300 M1 M2 M3

DA bidding (kWh)

200 100 0 -100 -200 -300

0

5

10

15

20

25

Hour (h)

Fig. 7.16 DA optimal bidding strategies of the PEVs

200

EV output (kWh)

0 -200 0 200

5

10

15

20

25 M1 M2 M3

0 -200 0 200

5

10

15

20

25

5

10

15

20

25

0 -200 0

Hour (h)

Solar power (kWh)

7.5 Case Studies

223

800 M1 M2 M3

600 400 200 0 0

5

10

15

20

25

Hour (h)

Emission (kg)

200 150

Carbon emission Emission cap

100 50 0

M1

M2

M3

Fig. 7.17 Comparisons of solar power accommodation and carbon emissions

From the results, with the help of V2G, the RPS can accommodate more solar power in M1 than in M2 and M3. The total amounts of daily solar energy in the three methods are 5.74, 5.52 and 5.09 MWh, respectively. Hence, the solar power accommodation in M1 can be improved by 3.99% and 12.77% compared with those in M2 and M3, respectively. Since more clean energy can be integrated, the daily carbon emissions in M1 are much less than those in M2 and M3. The daily carbon emissions in the three methods are 46.76, 123.72 and 171.98 kg, respectively. Thus, with the emission cap equaling 125 kg, the RPS in M1 and M2 can earn profits by selling CECs, while the RPS in M3 has to purchase an extra 46.98 kg CECs from the carbon market.

7.5.3.3

Market Revenue

The market revenues of the three methods are compared in Table 7.8. As compared in Table 7.8, the RPS in M1 can significantly increase its market revenues compared with in M2 and M3. In the DA markets, since more solar energy Table 7.8 Market revenues of the three methods

Revenue\method

M1

M2

M3

DA

$55.86

$44.49

$33.67

RT

$7.96

$−13.92

$−7.03

Carbon

$0.96

$0.02

$−0.58

Total

$64.78

$30.59

$26.06

224

7 Sharing Demand Side Resources for Regional Market Bidding

can be accommodated, M1 can improve the DA revenues by 25.56% and 65.90% compared with M2 and M3, respectively. In the RT markets, taking advantage of the V2G-enable PEVs, the RPS can earn profits in M1, but pay for the adjustment costs in M2 and M3. Additionally, the expected generations from MTs in M1, M2 and M3 are 61.52, 160.86 and 224.35 kWh, producing 46.76, 123.72 and 171.98 kg carbon, respectively. Hence, M1 can achieve the highest carbon revenues among the three method.

7.6 Conclusion This chapter introduces the significant value of the aggregate bidding based on sharing economy. Sharing economy can realize the aggregation and utilization of idle demand side resources. Based on the concept of sharing economy, we propose an optimal bidding framework for regional power systems considering flexible ramping product provision. A two-stage ATC evaluation framework is proposed integrating demand side resources. Additionally, we conduct research about the economic and environmental performance of sharing demand side resources in a deregulated market. Case studies based on an RPS with various demand side resources demonstrate the effectiveness of the proposed model and method. According to the simulation results, (i) the energy sharing model can aggregate the idle resources to behave as a controllable entity to earn more profits in energy and ancillary service markets; (ii) the aggregation of demand side resources to enhance real-time ATC, thereby improving the security and reliability of bulk power grids; and (iii) surplus renewable energy can be further accommodated, yielding both economic and environmental benefits.

References 1. Edmund, B.: Draft integrated energy policy report. California Energy Commission, Tech. Rep. (2015) 2. Yan, J., Chou, S.K., Chen, B., et al.: Clean, affordable and reliable energy systems for low carbon city transition. Appl. Energy 194, 305–309 (2017) 3. Wu, J., Yan, J., Desideri, U., et al.: Synergies between energy supply networks. Appl. Energy 192, 263–267 (2017) 4. Saraiva, J., Gomes, M.: Provision of some ancillary services by microgrid agents. In: 7th International Conference on the European Energy Market, pp. 1–8 (2010) 5. Kuznetsova, E., Li, Y., Ruiz, C., et al.: An integrated framework of agent-based modelling and robust optimization for microgrid energy management. Appl. Energy 129, 70–88 (2014) 6. He, G., Chen, Q., Kang, C., et al.: Optimal offering strategy for concentrating solar power plants in joint energy, reserve and regulation markets. IEEE Trans. Sustain. Energy 7(3), 1245–1254 (2016) 7. Xu, L., Tretheway, D.: Flexible ramping products—draft final proposal. California ISO 2014 (2014)

References

225

8. Abdul-Rahman, K., Alarian, H., Rothleder, M., et al.: Enhanced system reliability using flexible ramp constraint in CAISO market. In Proceedings of IEEE PES General Meeting, pp. 1–6 (2012) 9. Navid, N., Rosenwald, G.: Market solutions for managing ramp flexibility with high penetration of renewable resource. IEEE Trans. Sustain. Energy 3(4), 784–790 (2012) 10. Wu, C., Hug, G., Kar, S.: Risk-limiting economic dispatch for electricity markets with flexible ramping products. IEEE Trans. Power Syst. 31(3), 1990–2003 (2016) 11. Zou, P., Chen, Q., Xia, Q., et al.: Evaluating the contribution of energy storages to support large-scale renewable generation in joint energy and ancillary service markets. IEEE Trans. Sustain. Energy 7(2), 808–818 (2016) 12. Liu, G., Xu, Y., Tomsovic, K.: Bidding strategy for microgrid in day-ahead market based on hybrid stochastic/robust optimization. IEEE Trans. Smart Grid 7(1), 227–237 (2016) 13. Bertsimas, D., Sim, M.: Robust discrete optimization and network flows. Math. Program. 98, 49–71 (2003) 14. Milligan, M., Ela, E., Lew, D., et al.: Assessment of simulated wind data requirements for wind integration studies. IEEE Trans. Sustain. Energy 3(4), 620–626 (2012) 15. Wang, J., Zhong, H., Tang, W., et al.: Optimal bidding strategy for microgrids in joint and ancillary service markets considering flexible ramping products. Appl. Energy 205, 294–303 (2017) 16. Chen, Q., Kang, C., Xia, Q., et al.: Optimal flexible operation of a CO2 capture power plant in a combined energy and carbon emission market. IEEE Trans. Power Syst. 27(3), 1602–1609 (2012) 17. Li, X., Yu, C., Luo, F., et al.: A multimarket decision-making framework for GENCO considering emission trading scheme. IEEE Trans. Power Syst. 28(4), 4099–4108 (2013) 18. Electric Reliability Council of Texas, Inc. Available: http://www.ercot.com/mktinfo/prices 19. Wang, J., Zhong, H., Tan, C.W., et al.: Economic benefits of integrating solar-powered heat pumps in a CHP system. IEEE Trans. Sustain. Energy 9(4), 1702–1712 (2018) 20. PJM Data Directory. PJM wind generation. [Online]. Available: https://dataminer2.pjm.com 21. The National Renewable Energy Laboratory website. NREL’s PVWatts Calculator. [Online]. Available: http://pvwatts.nrel.gov 22. PJM Day-Ahead Hourly LMPs website. [Online]. Available: https://dataminer2.pjm.com/feed/ da_hrl_lmps 23. Wang, J., Zhong, H., Lai, X., et al.: Exploring key weather factors from analytical modeling toward improved solar power forecasting. IEEE Trans. Smart Grid 10(2), 1417–1427 (2019) 24. PJM Historical Load Forecasts website. [Online]. Available: https://dataminer2.pjm.com/feed/ load_frcstd_hist

Chapter 8

Sharing Non-wire Alternatives for Transmission Expansion Deferral

8.1 Introduction Power grids have been traditionally designed to be reliable during normal conditions and in response to foreseeable contingencies on adverse conditions. However, due to a dramatically increasing penetration of renewables and distributed energy resources (DERs), high-impact and low-probability events, e.g., rare weather cases, can pose great challenges to the secure operation of power grids and the reliability of power supply. A list of reliability outages around the U.S. caused by rare weather events is shown in Table 8.1. In the United States, for example, the annual economic losses caused by weather-related damage to transmission and distribution networks range from $20 to 55 billion. Despite low probability, rare weather events usually cause severe consequences and damage to power grids, thereby leading to inadequate power supplies and even blackouts. The problem of reliability issues and power outages should be enhanced in the design of power system planning. Traditionally, centralized generation capacity and transmission networks should be expanded, thus inducing costly investments [1, 2]. On the other hand, with an increase in alternating current lines, the electrical distance among substations becomes closer, which can result in significant shortcircuit current and some unexpected stability concerns [3]. Instead of “wire” solutions, more attention has been paid to non-wire alternatives (NWA) in recent years, e.g., investing in local distributed energy resources. The I-5 corridor project in the Pacific Northwest of the United States explores the DER alternative to improve available transfer capability [4]. The Brooklyn-Queens Demand Management Program in New York continues to pay attention to the improvement of the hosting capacity of distribution networks [5]. These methods of solving longterm planning problems based on DERs are typical NWAs, which can suppress peak load and enhance system reliability, thus postponing infrastructure upgrades. As an emerging business model, the concept of energy sharing has been adopted to balance local demands and make full use of idle DERs, which meets the requirements of NWA. In practice, the idle DERs cannot only be shared among neighbors for load © Science Press 2022 J. Wang et al., Sharing Economy in Energy Markets, https://doi.org/10.1007/978-981-16-7645-1_8

227

228

8 Sharing Non-wire Alternatives for Transmission Expansion Deferral

Table 8.1 Reliability outages caused by rare weather events Time

Location

Events

Impacts

2015.11

Spokane, Washington

A windstorm

Power lines were damaged, leading to over 161,000 people living without electricity.

2016.9

Tallahassee, Florida

Hurricane Hermine

Over 350,000 people lived without power for one week.

2017.2

Scranton, Pennsylvania

A storm

The storm caused 285,000 people to go without power for more than 3 days.

2017.3

Northeastern USA

A thunderstorm

10 million people from New York, New Jersey, Maryland and Pennsylvania lived without electricity.

2017.3

Michigan

A windstorm

Approximately 1 million people were without power for two days.

2017.9

Southeastern USA

Hurricane Irma

6 million in Florida, 1.3 million in Georgia and 0.2 million in South Carolina lost power.

2017.9

Puerto Rico, Dominica, etc.

Hurricane Maria

Over 547 people were killed, and the total economic losses reached $103.45 billion. Many people in Puerto Rico lost power for several months.

balance, but also be aggregated to provide ancillary services for connected power grids. For example, an aggregator can take advantage of the surplus capacity of users’ battery storage systems for peak shaving, thereby saving the reliability charges in capacity markets. Therefore, the purpose of delaying investment and improving social welfare has been achieved. In addition to the incentive compatibility and individual rationality discussed in the previous chapters, a well-designed profit sharing mechanism should also reveal each participant’s externality. However, locational marginal pricing (LMP) is known to only reflect price difference when power networks are congested. In this chapter, we propose a novel concept, termed as overall nodal pricing (ONP), which incorporates transmission cost allocation (TCA) into traditional LMP. On the one hand, the ONP can induce a usage-based price difference ahead of congestion by identifying the actual occupation on transmission networks. On the other hand, this pricing mechanism can generate effective price signals for congestion management in a preventable manner, thus enhancing power system reliability.

8.2 Overall Nodal Price

229

8.2 Overall Nodal Price 8.2.1 Basic Concept Locational marginal price refers to the marginal cost of increasing unit load demand at a certain node on condition that the constraints of transmission networks (TNs) and generation resources are satisfied. The research on LMP can data back to the late 1970s, and arouse a wide attention in 1980s. In the past decades, scholars have made great progress in the research and practice of LMP, e.g., electricity market models and optimization algorithms. In the traditional pricing paradigm, the Lagrangian multiplier-based LMP is a byproduct of optimization, which can be acquired after a market model is cleared. Such an inexplicit calculation process has long been criticized for many deficits. For example, due to the essence of Lagrangian multiplierbased congestion component, LMP can only reflect price difference when power networks are congested. However, the traditional LMP cannot incorporate the impact of fixed transmission assets. Transmission costs, employed to recover the fixed costs of transmission assets, are critical components of users’ electricity bills. Traditionally, transmission costs for most energy users are determined according to the share of annual demand or summer peak demand. Therefore, the transmission prices for energy users are predefined and constant, ignoring the impacts of time-varying load on the hourly actual usage of transmission assets. A fair transmission cost allocation scheme should reveal users’ hourly actual usage of transmission assets. On the one hand, fair transmission prices should distinguish the contributions of user’ hourly loads: A user should be allocated with a lower transmission price if devoted to reducing transmission capacity usage. Therefore, we propose a novel concept in this chapter, termed as overall nodal pricing, which incorporates a usage-based structural TCA into LMP, as shown in Fig. 8.1. For LMP, when power transmission increases from 0 to 100%, nodal price difference only occurs when the network is congested. Note that loss component is ignored here. However, the proposed ONP takes advantage of the concept of scarcity pricing, and thus nodal price gap can be enlarged with the increase in power transmission. By this means, the ONP can induce a usage-based price difference ahead of congestion so as to avoid congestion in a preventable manner. In this chapter, the ONP can be expressed as follows: Fig. 8.1 Comparison between locational marginal price and overall nodal price

Nodal price gap

Nodal price gap Step induced by congestion

0

Increase caused by scarcity 0

100%

(a) LMP

Power

100%

(b) ONP

Power

230

8 Sharing Non-wire Alternatives for Transmission Expansion Deferral

λONP = λSMP + λCong + λTCA

(8.1)

where λONP , λSMP , λCong and λTCA represent overall nodal price, system marginal price, congestion price and transmission cost allocation price, respectively. It should be noted that we aim at designing the concept of overall nodal price in this chapter, and thus only the scarcity for transmission capacity is incorporated. Some other components, e.g., the scarcity for primary frequency regulation and flexible ramping, can be embedded in future research.

8.2.2 Existing Transmission Cost Allocation Methods In Table 8.2, the transmission costs for most energy users are determined according to the share of annual demand or summer peak demand, known as the postage stamp method [6]. Therefore, the transmission prices for energy users are predefined and constant, ignoring the impacts of time-varying load on the hourly actual usage of transmission assets. The power flow tracking method is suitable for users or units participating in the unified clearing market to share the fixed costs of the power grid. This method forms a tracking matrix and finally obtains the cost that should be shared by each node. However, there are the following shortcomings: Firstly, the method only considers the allocation in space, but not in time series. Secondly, the calculation process of Table 8.2 Transmission cost allocation in different regional transmission organizations (RTOs) RTO

Responsibility

Method

CAISO

All loads

Share of annual demand [7]

ERCOT

All loads

Share of summer peak demand [8]

ISO-NE

All loads

Share of monthly coincident peak demand [9]

MISO

≥345 kV: 20% regionally and 80% sub-regionally 100–345 kV: sub-regionally

Regional loads: share of annual demand Sub-regional loads: line outage distribution factor analysis [10]

NYISO

All loads

Share of annual demand and contribution to reliability violation [11]

PJM

≥500 kV: all loads 2a i

bi 2ai

(9.117)

where ai is the opposite of the quadratic coefficient, and bi is the monomial coefficient. The parameters of the 10 consumers are listed in Table 9.1. Because the utility functions are only known to the consumers themselves, the LSE has to estimate the second derivatives of consumers’ utility functions. The estimated Table 9.1 Parameters of 10 consumers Consumer

ai ($/MW2 )

bi ($/MW)

pti (MW)

pti (MW)

TEi (MWh)

1

1.44

20.00

6.93

0.5

117.84

2

0.91

19.85

10.90

0.5

185.35

3

0.72

19.70

13.65

0.5

232.10

4

0.62

19.55

15.73

0.5

267.39

5

0.56

19.40

17.37

0.5

295.37

6

0.51

19.24

18.72

0.5

318.27

7

0.48

19.10

19.85

0.5

353.44

8

0.46

18.94

20.81

0.5

353.72

9

0.43

18.79

21.63

0.5

367.72

10

0.42

18.64

22.34

0.5

379.85

306

9 Information and Communication Technology for Sharing Economy

Table 9.2 Absolute real and estimated values of second derivatives of consumers’ utility functions Consumer

1

2

3

4

5

6

7

8

9

10

Real value

2.88

1.82

1.44

1.24

1.12

1.02

0.96

0.92

0.86

0.84

Estimated value

2.93

1.85

1.46

1.26

1.13

1.04

0.97

0.92

0.88

0.85

values can be obtained by observing the energy demand increments in response to congestion price perturbation, shown in Eq. (9.40). According to Table 9.1, the real second derivatives are −2ai . The estimated and real values are shown in Table 9.2. From these results, the second derivatives can be estimated accurately, which are the basis of the LMOS approach. In the following part of the case study, the LSE uses estimated second derivatives to optimize the Lagrangian multipliers while the consumers optimize their energy demands with real utility functions. The supply capacity of the distribution grid is 125.96 MW, and the tolerance is 10−3 MW. The optimal congestion prices and energy demands are shown in Fig. 9.9. From the optimal results, one can observe: (1) Energy demands are related to RTP. When RTPs are high from 12:00 to 20:00, the consumers decrease their energy demands to reduce electricity costs; (2) Congestion prices reflect the power supply shortage. From 12:00 to 20:00, the total energy demands of 10 consumers are less than the supply capacity of the distribution grid. Thus, the congestion prices during this period are zero, reflecting that the power supply is sufficient. To test the best performance of M1, C1 and C2 are screened. The results of iteration times by M1 are demonstrated in Fig. 9.10. In Fig. 9.10, the values of C1 and C2 greatly influence the iterations. When C1 = 0.20 and C2 = 0.97, M1 achieves the lowest iterations, equaling 108. RTP Congestion Price

120

35 30 25

100

20 15

80

10 5

60

0 -5

40

-10 -15 -20

20

-25 0

0

2

4

6

8

10

12

14

16

18

Time(h)

Fig. 9.9 Optimal results in the distribution grid with 10 consumers

20

22

24

-30

Price($/MWh)

Energy demand(MWh)

10 9 8 7 6 5 4 3 2 1

9.6 Case Studies

307

Fig. 9.10 Results of iterations by M1 screening C1 and C2

Although the same optimum is achieved by all three methods, the fluctuations in the Lagrangian multipliers take as many as 108 and 87 iterations before stability was observed for M1 and M2. In contrast, the iteration number of M3 is only 28, which is reduced by 74.07% and 67.82% compared with those of M1 and M2, respectively. Figure 9.11 illustrates a comparison of the congestion prices at 1:00 and the energy demand of consumer 1 at 1:00 by three methods. 7

40

6

32 5

28 24

M1-Energy demand M2-Energy demand M3-Energy demand

20 16 12

4 3 2

8

M1-Congestion price M2-Congestion price M3-Congestion price

4 0 0

10

20

30

40

50

60

Iteration

Fig. 9.11 Iteration of results by M1, M2 and M3

70

80

90

100

1 0 110

Energy demand(MWh)

Congestion price($/MWh)

36

308

9 Information and Communication Technology for Sharing Economy

From the iteration results, M3 converges much faster than M1 and M2. The congestion price at 1:00 converges to 12.43 $/MWh and the energy demand of consumer 1 at 1:00 converges to 5.19 MWh by three approaches. However, in M1, the search direction is the sub-gradient projection, and the step size is obtained in a fixed manner, shown in Eq. (9.117); in M2, although the binary search method is employed for coordination, the congestion prices still rely on sub-gradient projection, whereas in M3, the Lagrangian multipliers are optimized in a predictive way according to the iterative energy demands. Therefore, M3 converges much faster.

9.6.1.2

Real-World Distribution Grid with 14 Commercial Consumers

To further verify the efficiency of the proposed approach, a real-world distribution grid with 14 commercial consumers is tested. The time horizon is 24 and the current hour is 1. The energy usage, revenue and cost data are from Illinois, USA. These commercial consumers are engaged in different business, including merchandising, manufacturing, communication, etc. The users can gain revenues by consuming electricity. Thus, the revenues are regarded as the utility of commercial consumers. The historical data and the regressed utility of consumer 1 is shown in Fig. 9.12. Hence, the utility function of consumer 1 is,  2 U1 = −2.863 p1t + 50.16p1t − 14.84

(9.118)

By regressing the quadratic relationship between electricity consumption and the revenues, the utility function of 14 consumers can be obtained. The typical daily load demands of consumer 1 are shown in Fig. 9.13. The upper and lower limits are set to be 1.2 and 0.8 times of the load demands. The daily load requirements are the demands of each consumer’s typical daily load. The supply capacity of the distribution grid is 60 MW. The tolerance is 204

Fig. 9.12 Historical data and the regressed utility of consumer 1

202

Utility(thousand $)

200

Real value Regressed curve

198 196 194 192 190 188 186 184

6.0

6.2

6.4

6.6

6.8

7.0

7.2

7.4

Electricity consumption(MWh)

7.6

7.8

9.6 Case Studies

309 8

Typical load Upper limit Lower limit

Energy demand(MWh)

7 6 5 4 3 2 1 2

4

6

8

10

12

14

16

18

20

22

24

Time(h)

Fig. 9.13 The typical daily load demands of consumer 1

10−3 MW. The energy demands before and after optimal scheduling are shown in Fig. 9.14a, and the optimal congestion prices are shown in Fig. 9.14b. Restricted by the supply capacity of the distribution grid, the consumers will shift the peak loads to the valley hours to maintain the daily load requirements. When the total loads reach the supply capacity during peak hours from 9:00 to 23:00, the congestion prices are generated to leverage the balance between the energy supply and demand. To test the best performance of M1, C1 and C2 are screened. When C1 = 0.05 and C2 = 0.50, M1 achieves the lowest iterations. Table 9.3 compares the iteration Total load before optimal scheduling

Energy demand(MWh)

60 50 40 30 20 10 0

2

4

6

8

10

12

14

Time(h)

16

18

20

22

35

Congestion price RTP

30 25

Price($/MWh)

14 13 12 11 10 9 8 7 6 5 4 3 2 1

70

20 15 10 5 0

24

2

4

6

8

10

12

14

16

18

Time(h)

b

a

Fig. 9.14 Energy demands before and after optimal scheduling

Table 9.3 Iterations of M1, M2 and M3

Method

M1

M2

M3

Iteration

181

106

41

20

22

24

310

9 Information and Communication Technology for Sharing Economy

results of M1, M2 and M3. From the iteration results, M3 converges faster than M1 and M2, demonstrating the effectiveness and efficiency of the proposed method.

9.6.2 Impact of Communication Connectivity on Energy Sharing A distribution grid with 10 energy users is tested in different communication topology connectivity. Here, the users refer to large consumers in a distribution grid, including industrial parks and commercial buildings. The coefficients of quadratic utility function of different energy users are randomly generated from uniform distribution, which is shown in Table 9.4 The minimal/maximal loads of energy users are range from 80 to 120% of the actual load, respectively, and the minimal daily loads are set to the actual daily load demand of energy users. The parameter of ESS are shown in Table 9.5, and the initial stored energy is randomly generated from standard normal distribution. The retail rate and net metering rate of retailer are provided in Table 9.6. In this subsection, we investigate the impact of communication connectivity on energy sharing. Each energy user can exchange information and share energy with the interconnected energy users. The power curves of a certain energy user without communication network (communication topology connectivity = 0) and with an Table 9.4 Parameter of the stochastic variables

Table 9.5 Parameter of energy storage

Table 9.6 Retail rate and net metering rate

Variable

Min

Max

Quadratic coefficients

−0.5

−0.1

Liner coefficients

20

50

Parameter ESS (MW) Pistc ESS (MW) Pistd ηiESS

Values

Parameter

Values

5

ciESS ESS Eimin ESS Eimax

−0.5

5 95%

5 30

Type\hour

Time period

Price ($/kWh)

Retail rate

1:00–8:00

0.212

8:00:12:00

0.239

12:00:18:00

0.263

18:00:22:00

0.239

22:00–24:00

0.212

All period

0.03

Net metering rate

9.6 Case Studies

(a) Connectivity=0

311

(b) Connectivity=100%

Fig. 9.15 Power curves of an energy user

IoT-based communication network (communication topology connectivity = 100%) are shown in Fig. 9.15a, b, respectively. As can be observed, in the case without communication network, the ESS of energy user is only used to accommodate surplus solar energy to satisfy the load at nighttime. This is because the net metering rate is much lower than retail rate, and the energy user has no incentive to arbitrage with the ESS against retail rates, which prevents the energy sharing. The energy users minimize their individual net load rather than selling back the surplus power to the retailer. However, if there exists a fully connected communication network among energy users, in which energy users can exchange information and energy with each other. The utilization of the ESS can be greatly improved for shifting the day-time solar power to the night hours. The net load curve can be greatly changed, representing that the ESS stores energy at night and share the stored energy during the day-time. Additionally, one can observe that the electrical load curve can be smoothed through energy sharing. To further explore the value of communication network on energy sharing, the cost savings in different communication topologies are shown in Fig. 9.16. The communication topology connectivity varies from 0 to 100%. For each communication topology, we randomly generate the connectivity matrix for 10 rounds, and the average cost savings are adopted. As one can observe, with the increase of the connectivity, the cost savings brought by energy sharing can be improved while the marginal benefit declines. Such a decreasing margin can help decision-makers to determine the optimal sizing of IoT devices. In other words, it may not be an economic optimum to achieve a fully connected communication systems, and the tradeoff between the investment costs and the associated benefits should be balanced.

312

9 Information and Communication Technology for Sharing Economy

Fig. 9.16 Impact of communication connectivity on energy sharing

9.6.3 Impact of Communication Reliability on Energy Sharing In this section, the IEEE 33-bus feeder is adopted to validate the effectiveness and efficiency of the proposed framework and method. There are 32 branches, and the reference voltage is 12.66 kV. The three-phase power standard value is 10 MVA, and the total active power and reactive power demands are 5084.26 kWh and 2547.32 kVar, respectively. The case study covers 33 nodes, 6 of which are connection nodes (located at 1, 3, 6, 8, 12, and 30), and the others are load nodes. We generate 27 MGs uniformly distributed within a range of 4 km. The distribution of the MGs in the structure of the IEEE 33-bus is shown in Fig. 9.17. The bubbles represent the MGs with base stations. The colors indicate different base station transmit power levels. The quadratic and linear coefficients of the MG utility functions are randomly generated from uniform distributions, i.e., ai ∈ U [−0.5, −0.1], bi ∈ U [20, 50]. The MG minimal and maximal loads are set to 0.8 and 1.2 times the actual loads, respectively. The MG daily minimum loads are set to their actual daily load demand. The parameters of the MG ESSs and retail rates are listed in Tables 9.5 and 9.6. The initial stored energy is set to 40% of the maximal capacity. The price of BS load consumption is 0.22 $/kWh at all times. The values of the BS rated power are within the range of [40, 150] W. Here, we set it to 110 W. The equivalent rated transmit power of all base stations in the microgrid is 33 kW. The BS minimal and maximal transmit powers are set to 0.7 and 1.2 times the rated powers, respectively. The load consumption model depends on the type of base station. We choose a macroscale base station to be modelled. Therefore,   the path-loss model used here is ITU UMa NLOS: PL = 11.6 + 39.1log10 d n−i . The

9.6 Case Studies

313 BS' rated transmit power

Connection nodes 2000

80

Load nodes

30

31

1500

95

25 29 23

Vertical axis (m)

1000

32 33

2

0

26

5

1

8

4

17

21

20

125

18 18

7

3

19

110

28

6

24

500

-500

27

15 16

9

-1000 22

14

-1500

13

10 11

-2000 -2500 -2500 -2000 -1500 -1000

-500

12

0

500

1000

1500

2000

Horizontal axis (m)

Fig. 9.17 Microgrid distribution map in the structure of the IEEE 33-bus feeder

related coefficients are e = 70.22 and f = 894.54. The other related parameters are listed in Table 9.7. To verify the impact of the CR, 3 methods are compared, as summarized in Table 9.8. All the BSs under the NS method are operating at the rated transmit power (33 kW) without considering the CR. ATP has an adjustable BS transmit power without considering the CR. ACR is the most comprehensive energy sharing scheme, which adjusts the BS transmit power with the CR constraints (α = 0.93). Table 9.7 Parameters of the BSs

Table 9.8 Three methods compared in the case studies

Parameter

Values

Parameter

Values

B (MHz)

10

l (bit)

125

BN (MHz)

1.449

R (Mbps)

1000

Method

Adjustable transmit power

Considering CR

NS

× √

×

ATP ACR

√

× √

314

9 Information and Communication Technology for Sharing Economy

9.6.3.1

Base Case

With the connection of communication networks, microgrids exchange information and energy with each other. Due to PV’s zero marginal cost, the PV generation is used to balance daytime loads. To reduce electric bills, ESS is required to further discharge at daytime and charge at night. Similar conclusions can be observed from the net load profile, in which the MG stores energy at night and shares energy during daytime. By this means, the load profile can be smoothed through energy sharing. The load consumption of the BS does not reflect significant changes but fluctuates around at a fixed value instead. All the BSs must meet the requirements of the CR at each time slot to provide a stable information exchange for the physical layer. In addition, we notice that most of the MGs in the center of the distribution grid reach a relatively low CR due to the interference among the adjacent BSs. Thus, the CR of an MG is highly related to its location. Furthermore, we explain the location-related CR in details. Through the proposed linearization method, 27 linearized functions regarding the transmit power of the BSs can be obtained. The linear coefficients of these linearized functions reflect the selfand mutual-influence of the transmit power of one BS on the SNR to the other. Additionally, one can observe that such influence is highly related to the distance between two BSs. To examine the relationship between the distance and CR, the values in the 27 × 27 coefficient matrix are normalized and expressed in the form of a heat map, as shown in Fig. 9.18a. The darker color in the grid indicates a greater impact (a larger linear coefficient) on the CR. As one can observe from the darkest color of the diagonal elements, the dominant influencing factor of an MG’s CR is the transmit power of the inside BS. On the other hand, we apply K-means clustering method to partition the locations of the base stations into 2, 3, 4 and 5 clusters, and the clustering results are shown in Fig. 9.18b–e, respectively. According to the clustering results, the coefficient matrix is reorganized and sorted according to the distance between the BSs and the clustering center. The reorganized coefficient matrix forms several block matrices composed of dark-color elements, and the number of the block matrices just equals the number

SNR of the BS

(b) Restructured matrix (N=2) 1.056

-0.6800 BS BS2 BS4 BS5 B 7 BSS9 BS 10 B S1 1 BS 1 3 BS 1 4 BS15 B S1 6 BS17 BS18 BS19 BS20 BS21 BS 2 2 B S23 BS24 BS25 BS26 BS27 BS 2 8 BS29 BS31 BS32 33

2 4 5 7 9 10111314151617181920212223242526272829313233

SNR of the BS

(a) Original matrix

Fig. 9.18 Heat map of the coefficient matrix

2419 2 23332520282629 7 5 273218311714 9 101115 4 16132122

SNR of the BS

(c) Restructured matrix (N=3)

BS' transmit power

BS11 11 BS21 21 BS4 4 BS22 22 BS14 14 BS10 10 BS9 9 BS13 13 BS27 27 BS15 15 BS18 18 BS28 28 BS16 16 BS7 7 BS26 26 BS17 17 BS29 29 BS23 23 BS31 31 BS5 5 BS33 33 32 BS32 25 BS25 20 BS20 2 BS2 19 BS19 24 BS24

2419 2 20253233 5 3123291726 7 162818152713 9 101422 4 2111

SNR of the BS

(d) Restructured matrix (N=4)

BS20 20 BS4 4 BS9 9 BS22 22 BS21 21 BS16 16 BS11 11 BS13 13 BS15 15 BS10 10 BS14 14 BS27 27 BS28 28 BS17 17 BS18 18 BS7 7 BS26 26 BS23 23 BS29 29 BS5 5 BS31 31 BS33 33 BS32 32 BS25 25 BS19 19 BS2 2 BS24 24

BS2 B 4 BS1S2 BS29 BS35 BS32 BS33 B 1 BS2S5 BS29 BS23 B 6 BS1S7 BS18 BS27 BS28 BS17 BS14 BS10 BS15 BS13 BS11 BS26 BS21 BS92 B BS2S4 0

BS' transmit power

BS' transmit power 2419233321 2 322025312215 7 1418161727 4 2611 5 9 13282910

2.792

BS22 22 BS21 21 BS13 13 BS16 16 BS4 4 BS15 15 BS11 11 BS10 10 BS9 9 BS14 14 BS17 17 BS31 31 BS18 18 BS32 32 BS27 27 BS5 5 BS7 7 BS29 29 BS26 26 BS28 28 BS20 20 BS25 25 BS33 33 BS23 23 BS2 2 BS19 19 BS24 24

BS2 BS14 B 9 BS2S2 BS20 BS35 BS32 B 3 BS3S5 BS21 BS23 BS19 BS27 B 6 BS1S7 BS26 BS18 BS18 BS25 BS17 B 3 BS1S9 BS10 BS24 B 2 BS2S4 BS11 1

4.528

BS10 10 BS29 29 BS28 28 BS13 13 BS9 9 BS5 5 BS11 11 BS26 26 BS4 4 BS27 27 BS17 17 BS16 16 BS18 18 BS14 14 BS7 7 BS15 15 BS22 22 BS31 31 BS25 25 BS20 20 BS32 32 BS2 2 BS21 21 BS33 33 BS23 23 BS19 19 BS24 24

BS2 BS14 B 9 BS2S2 BS33 BS23 BS25 BS20 BS28 BS26 BS79 B BS2S5 BS37 BS12 BS38 BS11 BS17 BS94 BS1 BS10 BS11 B 5 BS1S4 BS16 BS23 BS21 2

6.264

BS' transmit power

8.000

BS2 BS14 BS29 BS33 BS23 B 1 BS3S2 BS22 BS20 BS35 BS21 BS12 B 5 BS1S7 BS14 BS18 BS16 BS27 B 7 BS2S4 BS16 BS51 B BS1S9 BS23 BS28 BS19 0

BS' transmit power

Value 33 BS33 32 BS32 31 BS31 29 BS29 28 BS28 27 BS27 26 BS26 25 BS25 24 BS24 23 BS23 22 BS22 21 BS21 20 BS20 19 BS19 18 BS18 17 BS17 16 BS16 15 BS15 14 BS14 13 BS13 11 BS11 10 BS10 9 BS9 7 BS7 5 BS5 4 BS4 2 BS2

24 2 1925323331 5 292326 7 181728271410151311162122 9 4 20

SNR of the BS

(e) Restructured matrix (N=5)

9.6 Case Studies

315

of the clusters, indicating that the influence of adjacent BSs is far greater than that of remote ones. Table 9.9 provides the daily energy charged/discharged by the ESSs, the accommodated solar energy, and the total costs of all MGs via the three methods. Since the CR is not considered in ATP, all base stations operate at the minimum transmit power based on the principle of cost minimization. In contrast, a BS’s transmit power varies between 77 and 132 W under ACR. The utilization rate of the ESS is mainly affected by two aspects: one is its capacity limitation, and the other is the power flow constraint. The ESS cannot reach a greater charging power when the power flow constraints are bounded at night under NS. Thus, the utilization of the ESSs under NS is the lowest. Due to the zero-marginal costs of PV, solar energy could be fully accommodated in the three methods. In addition, since the base stations are fixed at a high load consumption level, the cost under NS is the highest. In contrast, to reduce the mutual signal influence, most base stations operate below the rated power under ACR. Thus, the cost under ACR is lower than that under NS. On top of adjustable BSs, ACR incorporates the CR constraints, which may induce additional operational costs than ATP. The detailed performances regarding cost reduction, storage utilization and CR are compared among the three methods in Fig. 9.19. From the results, it is a straightforward conclusion that the CR under ACR can be ensured. Also, it demonstrates that in practice, the CR constraints must be considered in the energy sharing since NS and ATP may lead to the violation of CR. With the Table 9.9 MG daily storage energy, solar energy and total costs

Method

Storage energy (MWh)

Solar energy (MWh)

Cost ($)

NS

1351.78

2839.09

747,761.38

ATP

1393.92

2839.09

642,782.29

ACR

1383.58

2839.09

663,093.85

Fig. 9.19 Performance of the ESS utilization, CR, and cost reduction under the 3 methods

NS ATP ACR

Profit

ESS

CR

316

9 Information and Communication Technology for Sharing Economy

support of adjustable BSs, the utilization of ESS can be further improved in contrast to ACR.

9.6.3.2

IEEE 141-Bus Feeder

To validate the efficiency and effectiveness of the proposed mechanism and method, we conduct case studies based on the IEEE 141-bus distribution network with an aggregator and 84 load nodes, which are uniformly distributed across an area 8 km in diameter. The parameter settings are the same as those in the IEEE 33-bus case study. Figure 9.20 shows the difference in the CR value between the cases with and without the CR constraints. Compared to the IEEE 33-bus case study, the interference between the base stations in the IEEE 141-bus feeder becomes much more notable, and we therefore relax the CR constraint to α = 0.85. As one can observe, the CR values of all the MGs are higher than 0.85. In contrast, approximately 30% of the MGs achieve the CR lower than 0.85 without the CR constraint, 2 of which reach even as low as 0.6. On the premise of different CR requirements under ACR, the storage utilization and the total cost of the distribution grid are shown in Fig. 9.21. The results reveal that the storage utilization and the total cost can be greatly improved with the increase in CR requirements. Most base stations have to increase the transmit power to enhance the communication reliability, thereby leading to higher total costs of the distribution grid. 30 28 26

Without CR requirement CR requirement

Microgrid number

24 22 20 18 16 14 12 10 8 6 4 2 0

0.60 0.64 0.68 0.72 0.76 0.80 0.84 0.88 0.92 0.96 1.00

Communication reliability Fig. 9.20 Communication reliability with and without the CR requirements

317

6500

Cost

6400

ESS

1.56

6300

1.54

6200

1.52

6100

1.50

6000

1.48

5900

1.46

Cost (106 $)

Storage energy (MWh)

9.7 Conclusion

5800 1.44

5700

1.42

5600 5500

1.40 0.80

0.82

0.84

0.86

0.88

The requirement of CR Fig. 9.21 Storage utilization and the total cost under different CR requirement (α)

9.7 Conclusion This chapter first introduces the cloud-edge computing technology that supports energy sharing, and then discusses the impact of information and communication technology on energy sharing from two aspects, i.e., communication connectivity and communication reliability. An energy sharing scheme based on ICT is proposed to illustrate the interaction of ICT and energy sharing, in which N microgrids in a citylevel distribution network exchange information and cooperate as a single-interest entity. For cloud-edge computing technology, we propose a Lagrangian multiplier optimal selection method to overcome the oscillation and miscovergence of traditional Lagrangian relaxation. Case studies based on a small case and a realistic case demonstrate the robustness and efficiency of the proposed cloud-edge computing technology. A communication connectivity matrix is designed to reflect the possibility of peer-to-peer energy sharing. Here we discover that the marginal benefit brought by IoT declines with the increase in P2P connectivity. Then we analyze the impact of communication reliability on energy sharing. The communication reliability can be modeled as a nonlinear function with respect to BSs’ transmit power, which is linearized via a least-square regression method. Case studies based on the IEEE-33 and -141 feeder systems compare the performance of three methods. Finally, we discuss a worst case where IoT does not work and energy sharing cannot be achieved amidst extreme weather events. The resilience value of distributed solar + storage systems are quantified via a simplified formulation without the need of optimization. Hopefully, the development of information and communication technology can improve the performance and merits of energy sharing in the near future.

318

9 Information and Communication Technology for Sharing Economy

References 1. DOE Global Energy Storage Database. [Online]. Available: http://www.energystorageexcha nge.org/projects 2. Wang, J., Zhong, H., Wu, C., Du, E., Xia, Q., Kang, C.: Incentivizing distributed energy resource aggregation in energy and capacity markets: an energy sharing scheme and mechanism design. Appl. Energy 252, 113–471 (2019) 3. Zhang, C., Wu, J., Zhou, Y., Cheng, M., Long, C.: Peer-to-Peer energy trading in a Microgrid. Appl. Energy 220, 1–12 (2018) 4. Long, C., Wu, J., Zhou, Y., Jenkins, N.: Peer-to-peer energy sharing through a two-stage aggregated battery control in a community microgrid. Appl. Energy 226, 261–276 (2018) 5. Wang, Z., Yu, X., Mu, Y., Jia, H.: A distributed peer-to-peer energy transaction method for diversified prosumers in Urban Community Microgrid System. Appl. Energy 260, 114–327 (2020) 6. Wang, J., Zhong, H., Qin, J., Tang, W., Rajagopal, R., Xia, Q., Kang, C.: Incentive mechanism for sharing distributed energy resources. J. Modern Power Syst. Clean Energy 7(4), 837–850 (2019) 7. Wang, J., Wu, Z., Du, E., Zhou, M., Li, G., Zhang, Y.: Constructing V2G-enabled regional energy internet toward cost-efficient carbon trading. J. Power Energy Syst. 6(1), 31–40 (2020) 8. Hui, H., Ding, Y., Shi, Q., Li, F., Song, Y., Yan, J.: 5G network-based Internet of Things for demand response in smart grid: a survey on application potential. Appl. Energy 257, 113–972 (2020) 9. Shafi, M., Molisch, A.F., Smith, P.J., Haustein, T., Zhu, P., De Silva, P., et al.: 5G: a tutorial overview of standards, trials, challenges, deployment, and practice. IEEE J. Sel. Areas Commun. 35(6), 1201–1221 (2017) 10. Chang, K., Chu, K., Wang, H., Lin, Y., Pan, J.: Energy saving technology of 5G base station based on internet of things collaborative control. IEEE Access 8, 32935–32946 (2020) 11. Kong, P., Song, Y.: Joint consideration of communication network and power grid topology for communications in community smart grid. IEEE Trans. Industr. Inf. 16(5), 2895–2905 (2020) 12. Xu, L., Guo, Q., Yang, T., Sun, H.: Robust routing optimization for smart grids considering cyber-physical interdependence. IEEE Trans. Smart Grid 10(5), 5620–5629 (2019) 13. Kwonsik, S., Kyle, A., SangHyun, L.: An energy-cyber-physical system for personalized normative messaging interventions: identification and classification of behavioral reference groups. Appl. Energy 260, 114–237 (2020) 14. Chen, J., Liang, G., Cai, Z., Hu, C., Xu, Y., Luo, F.: Impact analysis of false data injection attacks on power system static security assessment. J. Modern Power Syst. Clean Energy 4(3), 496–505 (2016) 15. Kexing, L., Mahesh, I., Karthikeyan, S.: A tri-level optimization model to mitigate coordinated attacks on electric power systems in a cyber-physical environment. Appl. Energy 235, 204–218 (2019) 16. Deng, R., Xiao, G., Lu, R., et al.: Fast distributed demand response with spatially- and temporally-coupled constraints in smart grid. IEEE Trans Ind. Inform. 11(6), 1597–1606 (2015)