作者：周震尘, 金涛

作者单位：福州大学电气工程与自动化学院，福建 福州 350016

关键词：电力系统弹性;脆弱性分析;拓扑攻击;深度强化学习

摘要：

由于全球变暖等原因，导致越来越多的极端天气事件发生，此类高影响低概率事件的序列攻击对配电网造成重大破坏。为了分析配电网在极端事件下的脆弱性，该文提出一种基于深度强化学习（deep reinforcement learning，DRL）的极端事件攻击序列确定方法。利用马尔科夫决策过程（Markov decision process，MDP）捕捉灾害的时空特性，同时结合元件故障的随机性，应用连锁故障模型模拟系统的行为，即系统受极端事件影响而带来的更大的故障现象，引入基于价值的DRL方法确认对系统影响最大且故障率较高的关键线路序列。该文在IEEE测试系统中进行仿真，验证所提方法对极端事件中的配电网进行脆弱性分析的有效性和准确性。

A novel deep reinforcement learning approach for vulnerability analysis of distribution systems under extreme event
ZHOU Zhenchen, JIN Tao
College of Electrical Engineering and Automation, Fuzhou University, Fuzhou 350016, China
Abstract: Due to global warming, more and more extreme weather events are occurring, and sequential attacks of such high-impact, low-probability events cause significant damage to distribution systems. In order to analyze the vulnerability of distribution networks under extreme events, this paper proposes a deep reinforcement learning-based approach to determine the sequence of attacks of extreme events. The Markov decision process (MDP) is used to capture the spatiotemporal characteristics of extreme events, incorporating the uncertainties of component failures. Then, the cascading failure model is utilized to simulate the behavior of the system, i.e., the larger failure scenarios that may result from extreme events impacting the system. Finally, a value-based DRL method is introduced to find critical line sequences that have the greatest impacts on the system and have high failure rates. In this paper, simulations are carried out in the IEEE test system to validate that the proposed approach is effective and accurate for vulnerability analysis of distribution networks in extreme events.
Keywords: power system resilience;vulnerability analysis;topology attack;deep reinforcement learning
2022, 48(2):98-104  收稿日期: 2020-10-26;收到修改稿日期: 2020-12-19
基金项目: 国家自然科学基金面上项目资助(51977039)
作者简介: 周震尘(1996-),男,福建宁德市人,硕士研究生,专业方向为智能电网理论及技术
参考文献
[1] 鞠平, 王冲, 辛焕海, 等. 电力系统的柔性、弹性与韧性研究[J]. 电力自动化设备, 2019, 39(11): 1-7
[2] JIA H, PAN D. Tornado disaster impacts and management: Learning from the 2016 tornado catastrophe in Jiangsu Province, China[J]. Natural hazards, 2017, 89(1): 457-471
[3] MAHZARNIA M, MOGHADDAM M P, BABOLI P T, et al. A Review of the Measures to Enhance Power Systems Resilience[J]. IEEE Systems Journal, 2020, 14(3): 4059-4070
[4] 别朝红, 林超凡, 李更丰, 等. 能源转型下弹性电力系统的发展与展望[J]. 中国电机工程学报, 2020, 40(9): 2735-2745
[5] PANTELI M, TRAKAS D N, MANCARELLA P, et al. Power systems resilience assessment: Hardening and smart operational enhancement strategies[J]. Proceedings of the IEEE, 2017, 105(7): 1202-1213
[6] BAGHERI A, ZHAO C, QIU F, et al. Resilient transmission hardening planning in a high renewable penetration era[J]. IEEE Transactions on Power Systems, 2019, 34(2): 873-882
[7] WANG X, LI Z, SHAHIDEHPOUR M, et al. Robust line hardening strategies for improving the resilience of distribution systems with variable renewable resources[J]. IEEE Transactions on Sustainable Energy, 2019, 10(1): 386-395
[8] ARROYO J M. Bilevel programming applied to power system vulnerability analysis under multiple contingencies[J]. IET generation, transmission & distribution, 2010, 4(2): 178-190
[9] YUAN W, WANG J, QIU F, et al. Robust optimization-based resilient distribution network planning against natural disasters[J]. IEEE Transactions on Smart Grid, 2016, 7(6): 2817-2826
[10] MA S, CHEN B, WANG Z. Resilience enhancement strategy for distribution systems under extreme weather events[J]. IEEE Transactions on Smart Grid, 2018, 9(2): 1442-1451
[11] PAUL S, DING F. Identification of worst impact zones for power grids during extreme weather events using Q-learning[C]//2020 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT), 2020: 1-5.
[12] 吴熙, 唐子逸, 徐青山, 等. 基于Q学习算法的综合能源系统韧性提升方法[J]. 电力自动化设备, 2020, 40(4): 146-152
[13] WU J, FANG B, FANG J, et al. Sequential topology recovery of complex power systems based on reinforcement learning[J]. Physica A:Statistical Mechanics and its Applications, 2019, 535: 122487
[14] 叶興, 薛家祥. 改进型LSTM网络光伏发电功率预测研究[J]. 中国测试, 2019, 45(11): 14-20
[15] MNIH V, KAVUKCUOGLU K, SILVER D, et al. Human-level control through deep reinforcement learning[J]. Nature, 2015, 518(7540): 529-533
[16] 刘全, 翟建伟, 章宗长, 等. 深度强化学习综述[J]. 计算机学报, 2018, 41(1): 1-27
[17] HUANG Q, HUANG R, HAO W, et al. Adaptive Power System Emergency Control Using Deep Reinforcement Learning[J]. IEEE Transactions on Smart Grid, 2020, 11(2): 1171-1182
[18] SALMAN A M, LI Y, STEWART M G. Evaluating system reliability and targeted hardening strategies of power distribution systems subjected to hurricanes[J]. Reliability Engineering & System Safety, 2015, 144: 319-333