Artificial Intelligence Empowered Smart Energy Systems

Qiang Yang, PhD, is a Full Professor at the College of Electrical Engineering at Zhejiang University, China. He is a Senior Member of the IEEE and a Distinguished Member of the China Computer Federation. His primary research interests include smart energy systems, learning-based optimization and control. Gang Huang, PhD, is an Assistant Professor at the College of Electrical Engineering at Zhejiang University, China. He is a Senior Member of the IEEE and a Senior Member of the China Computer Federation. His primary research interests include artificial intelligence for power and energy systems.

• List of Contributors xvAbout the Editors xxiForeword xxiiiPreface xxvAcknowledgments xxvii1 Machine Learning-Based Applications for Cyberattack and Defense in Smart Energy Systems 1Sha Peng, Mengxiang Liu, Zhenyong Zhang, and Ruilong Deng1.1 Introduction to Machine Learning 11.1.1 Overview of Machine Learning Approaches 11.1.1.1 Unsupervised Learning 11.1.1.2 Supervised Learning 31.1.1.3 Semi-Supervised Learning 41.1.1.4 Reinforcement Learning 41.1.2 Advantages of Machine Learning 51.2 Machine Learning in Attack Design 61.2.1 System Information Inference 71.2.2 Attack Resource Allocation 91.3 Machine Learning in Attack Protection 101.3.1 Data Security 101.3.2 Vulnerability Analysis 111.4 Machine Learning in Attack Detection 121.4.1 Anomaly Detection 131.4.2 Line Outage Detection 181.4.3 Electricity Theft Detection 181.5 Machine Learning in Impact Mitigation 191.5.1 Compromised Measurement Clearing 191.6 Future Directions 20References 222 Enhancing Cybersecurity in Power Communication Networks: An Approach to Resilient CPPS Through Channel Expansion and Defense Resource Allocation 27Yingjun Wu, Yingtao Ru, Jinfan Chen, Hao Xu, Zhiwei Lin, Chengjun Liu, and Xinyi Liang2.1 Introduction 272.2 Mechanisms for the Classification and Propagation of Cyberattacks 282.2.1 Popularity in Academic Research and Frequency of Actual Cyberattacks 302.2.2 Propagation Mechanism of Hotspot Cyberattacks 302.2.2.1 Denial of Service (DoS) Attack 302.2.2.2 False Data Injection Attacks (FDIAs) 322.2.2.3 The Black Hole Attack (BHA) 342.2.2.4 Address Resolution Protocol (ARP) Spoofing Attack 352.2.2.5 Man-In-The-Middle (MITM) Attack 362.2.2.6 Eavesdropping Attack (EA) 372.2.2.7 Replay Attack 382.2.2.8 Load Altering Attack (LAA) 392.2.3 Propagation Mechanism of Non-Hotspot Cyberattacks 412.2.3.1 Active Non-Hotspot Cyberattack 412.2.3.2 Passive Non-Hotspot Cyberattack 422.2.4 Targeting Equipment of Cyberattacks in TAL 432.3 Power Communication Network Planning Based on Information Transmission Reachability Against Cyberattacks 472.3.1 Introduction 472.3.2 Planning Framework for Grid Communication Network Planning Object 482.3.2.1 Planning Goals 492.3.2.2 Planning Framework 502.3.3 Topology Planning Model of Power Communication Network 512.3.3.1 Planning Goal of the TP to Ensure Information Transmission of Regular Operation 512.3.3.2 Enhancing Cyberattack Defense Capabilities in ACAP Planning Goals 532.3.4 A Game Theory-Based Planning Method 592.3.4.1 Criterion Derived from the Nash Equilibrium 592.3.4.2 Model Solution Using Enhanced Particle Swarm-Based Optimization Algorithm 602.3.5 Simulations 612.3.5.1 System Overview Under Study 612.3.5.2 Cyberattacks Factored in During the Planning Phase 632.3.5.3 Planning Results in Different Cases 632.3.5.4 Relative to Planning Methods that Disregard Cyberattacks 652.3.5.5 Practical Case Application 672.3.5.6 PSO-Based Analysis of Planning Outcomes 692.4 Survivability-Oriented Defensive Resource Allocation for Communication and Information Systems Under Cyberattack 712.4.1 Introduction 712.4.2 Systematic Evaluation of Survivability for CPPS Communication and Information Systems 732.4.2.1 Breaking Down Power Businesses into Atomic Services 732.4.2.2 Indexes for Evaluating the Survivability of Atomic Services 742.4.2.3 Survivability Evaluation Framework 742.4.2.4 Calculation Method for Survivability Evaluation Indices 742.4.2.5 Calculation Method for Survivability Evaluation Indexes 792.4.3 Defensive Resource Allocation Model for Enhancing CPPS Survivability Against Cyber Threats 792.4.3.1 Objective Function 802.4.3.2 Constraints 802.4.4 Modified Genetic Algorithm for the Proposed Model 802.4.5 Simulations 822.4.5.1 Introduction of the Studied System 822.4.5.2 Simulation Results and Analysis 85References 913 Multi-Objective Real-Time Control of Operating Conditions Using Deep Reinforcement Learning 101Ruisheng Diao, Tu Lan, Zhiwei Wang, Haifeng Li, Chunlei Xu, Fangyuan Sun, Bei Zhang, Yishen Wang, Siqi Wang, Jiajun Duan, andDiShi3.1 Introduction 1013.2 Principles of Deep Reinforcement Learning 1023.2.1 Deep Q Network (DQN) 1033.2.2 Proximal Policy Optimization (PPO) 1043.2.3 Soft Actor-Critic (SAC) 1063.3 Real-Time Line Flow Control Using PPO 1073.3.1 Problem Formulation 1073.3.1.1 State Space 1073.3.1.2 Action Space 1073.3.1.3 Reward Function 1073.3.1.4 Training Process of PPO-Based Agents 1083.3.2 Case Studies 1093.4 Dueling DQN-Based Topology Control for Maximizing Available Transfer Capabilities 1103.4.1 Problem Formulation 1123.4.1.1 Architecture Design 1133.4.1.2 Dueling DQN Agent 1133.4.1.3 Imitation Learning 1143.4.1.4 Guided Exploration Training 1153.4.1.5 Early Warning 1163.4.2 Case Studies 1163.4.2.1 Environment and Framework 1163.4.2.2 Effectiveness of Imitation Learning 1173.4.2.3 Improved Performance with Guided Exploration 1173.4.2.4 Performance Comparison of Different Agents 1183.5 Real-Time Multi-Objective Power Flow Control Using Soft Actor-Critic 1193.5.1 Problem Formulation 1193.5.1.1 Architecture Design 1203.5.1.2 Episode and Terminating Conditions 1223.5.1.3 State Space 1223.5.1.4 Control Space 1223.5.1.5 Reward Definition 1223.5.2 Case Studies 123References 1244 Smart Generation Control Based on Multi-Agents 127Lei Xi, Yixiao Wang, Lu Dong, and Jianyu Ren4.1 Overview 1274.2 Research on Intelligent Power Generation Control Based on Multi-Agents 1284.2.1 Function and Architecture of Multi-Agent System 1284.2.2 Virtual Power Generation Tribe Control Based on Multi-Agent Theory 1294.2.2.1 First-Order Multi-Agent Consistency Algorithm 1304.2.2.2 Consistent AGC Power Allocation Algorithm 1324.2.2.3 Robust Consistency Algorithm in Nonideal Communication Networks 1364.3 Intelligent Power Generation Control for Islands and Microgrids 1394.3.1 AGC Cooperative Control Based on Equal Incremental Rate Consistency Algorithm 1414.3.1.1 Intelligent Distribution Network Decentralized Autonomy Framework 1414.3.1.2 AGC Power Allocation Model of Smart Distribution Network 1424.3.1.3 Consistency Algorithm of Equal Incremental Rate 143References 1475 Power System Fault Diagnosis Method Under Disaster Weather Based on Random Self-Regulating Algorithm 149Tao Wang, Liyuan Liu, Ruixuan Ying, Chunyu Zhou, Hanyan Wu, and Quanlin Leng5.1 Introduction 1495.2 Analytic Model for Fault Diagnosis 1515.2.1 Three Types of Self-Regulating Trust Factors 1535.2.1.1 Self-Regulating Expectation Trust Factor and Self-Regulating Warning Trust Factor 1535.2.1.2 Self-Regulating Weather Trust Factor 1555.2.2 Expected States of Protection Devices 1565.2.2.1 Expected States of Main Protective Relays 1575.2.2.2 Expected States of Primary Backup Protective Relays 1575.2.2.3 Expected States of Secondary Backup Protective Relays 1575.2.2.4 Expected States of Breaker Failure Protective Relays 1575.2.2.5 Expected States of Circuit Breakers 1575.3 Random Self-Regulating Algorithm 1575.3.1 Random Self-Regulating Algorithm 1585.3.2 Bionic Self-Regulating Function 1605.3.2.1 Increment Operator of Guiding Probability 1605.3.2.2 Iterative Mutation Operator 1625.3.3 Fault Diagnosis Process 1645.4 Experiment and Analysis 1655.4.1 Case Study 1665.4.2 Accuracy Test 169References 1716 Statistical Machine Learning Model for Production Simulation of Power Systems with a High Proportion of Photovoltaics 173Xueqian Fu, Feifei Yang, Qiaoyu Ma, Na Lu, and Chunyu Zhang6.1 Introduction 1736.2 Methodology 1746.2.1 Long Time Scale 1746.2.1.1 Bidirectional LSTM Style-Based Generative Adversarial Networks 1746.2.1.2 Clustering with Adaptive Neighbors 1766.2.2 Short Time Scale 1806.2.2.1 Decomposition Strategy and Attention-Based Long Short-Term Memory Network 1806.2.2.2 Forecasting with Dynamic Mask 1836.3 Case Studies 1856.3.1 Long Time Scale 1856.3.1.1 Year-Round Photovoltaic Scenario Generation 1856.3.1.2 Typical Scenarios Extraction 1886.3.2 Short Time Scale 1916.3.2.1 Power Load Forecast 1916.3.2.2 Photovoltaic Power Generation Forecast 193References 1957 Dynamic Reconfiguration of PV-TEG Hybrid Systems via Improved Whale Optimization Algorithm 199Bo Yang, Jiarong Wang, and Yulin li7.1 Introduction 1997.2 PV-TEG Hybrid System Model 2027.2.1 PV System Model 2027.2.2 TEG System Model 2047.2.3 PV-TEG Hybrid System Model 2067.2.4 Objective Function 2077.2.5 Performance Evaluation 2087.3 Improved Whale Optimization Algorithm 2087.3.1 Whale Optimization Algorithm 2087.3.1.1 Encircling Prey 2087.3.1.2 Bubble-Net Attack 2097.3.1.3 Random Search 2097.3.2 Design of Improved Whale Optimization Algorithm 2107.3.2.1 The Roulette Mechanism 2107.3.2.2 Nonlinear Convergence Factor 2127.4 Case Study 2127.4.1 6 × 6SquareArray 2137.4.2 6 × 10 Non-Square Array 2197.5 Conclusion 226References 2288 Coordinating Transactive Energy and Carbon Emission Trading Among Multi-Energy Virtual Power Plants for Distributed Learning 233Peiling Chen and Yujian Ye8.1 Introduction 2338.1.1 Background and Motivation 2338.1.2 Review of Previous Work 2348.2 Overall Transactive Trading Market in Heterogeneous Networked MEVPPs 2368.3 Mathematical Formulation of MEVPP Coordination Problem 2388.3.1 Objective Function 2388.3.2 Constraints 2398.3.2.1 Energy Balance Constraint 2398.3.2.2 CER Balance Constraint 2398.3.2.3 Operating Constraints of Conversion Devices 2418.3.2.4 Constraints of Energy Storage Devices 2428.3.2.5 Trading Constraints 2428.3.2.6 Network Constraints 2428.4 Adaptive Consensus ADMM 2438.4.1 Consensus ADMM 2438.4.2 Adaptive Tuning of Penalty Parameters 2448.4.3 AC-ADMM-Based Energy and CER Trading for Networked MEVPPs 2448.5 Case Studies 2498.5.1 Experimental Setup 2498.5.2 Impact of Transactive Trading 2518.5.2.1 Impact of Transactive CER Trading 2528.5.2.2 Impact of Transactive Heat Trading 2548.5.3 Impact of Network Constraints 2558.5.4 Convergence Analysis 2568.6 Conclusions 258References 2589 Cluster-Based Heuristic Algorithm for Collection System Topology Generation of a Large-Scale Offshore Wind Farm 263Jincheng Li, Zhengxun Guo, and Xiaoshun Zhang9.1 Introduction 2639.1.1 Background and Importance 2639.1.2 Research Status 2649.1.2.1 Radial Topologys Optimization Methods 2649.1.2.2 Graph Theory-Based Methods 2659.1.2.3 Meta-Heuristic Optimization Methods 2659.2 Mathematical Model for CS Optimization in LSOWFs 2669.2.1 Composition of the LSOWF 2669.2.2 Objective Function 2679.2.3 Constraints 2689.2.3.1 Constraint on the Number of WTs per Feeder 2689.2.3.2 Constraint on the Load Capacity of Submarine Cable 2699.2.3.3 Non-Crossing Constraint of Submarine Cables 2699.2.3.4 Voltage Constraint 2709.3 Cluster-Based Topology Generation Method 2709.3.1 Polar Coordinate Clustering 2709.3.2 Radial Clustering 2719.3.2.1 Fuzzy C-Means Algorithm 2729.3.2.2 Radial Fuzzy C-Means Algorithm 2729.3.3 Dynamic Minimum Spanning Tree 2749.3.4 Firefly Algorithm-Based Optimization 2769.4 Case Study 2779.4.1 Test Case # 1 2779.4.2 Test Case # 2 2789.4.3 Cost Comparison 2809.5 Conclusion 282References 28310 Transmission Line Multi-Fitting Detection Method Based on Implicit Space Knowledge Fusion 287Qianming Wang, Congbin Guo, Xuan Liu, and Yongjie Zhai10.1 Introduction 28710.2 Overall Overview of Methods 29210.3 Implicit Spatial Knowledge Fusion Structure 29410.3.1 Spatial Box Setting Module 29510.3.2 Spatial Context Extraction Module 29710.3.3 Spatial Context Memory Module 29810.4 Improved Post-Processing Structure 30110.5 Experimental Results and Analysis 30310.5.1 Experimental Dataset and Environment 30310.5.2 Comprehensive Comparative Experiment 30410.5.3 Ablation Experiment 30710.5.4 Visual Comparison Experiment 30910.6 Summary 311References 312Index 315