基于多智能体深度强化学习的随机事件驱动故障恢复策略  

Uncertain event-driven fault recovery strategy based on multi-agent deep reinforcement learning

作　　者：王冲[1] 石大夯 万灿 陈霞[3] 吴峰[1] 鞠平[1] WANG Chong;SHI Dahang;WAN Can;CHEN Xia;WU Feng;JU Ping(School of Electrical and Power Engineering,Hohai University,Nanjing 211100,China;College of Electrical Engineering,Zhejiang University,Hangzhou 310027,China;School of Electrical and Electronic Engineering,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]河海大学电气与动力工程学院,江苏南京211100 [2]浙江大学电气工程学院,浙江杭州310027 [3]华中科技大学电气与电子工程学院,湖北武汉430074

出　　处：《电力自动化设备》2025年第3期186-193,共8页Electric Power Automation Equipment

基　　金：国家自然科学基金资助项目(52277088)。

摘　　要：为了减少配电网故障引起的失负荷,提升配电网弹性,提出一种基于多智能体深度强化学习的随机事件驱动故障恢复策略:提出了在电力交通耦合网故障恢复中的随机事件驱动问题,将该问题描述为半马尔可夫随机决策过程问题;综合考虑系统故障恢复优化目标,构建基于半马尔可夫的随机事件驱动故障恢复模型;利用多智能体深度强化学习算法对所构建的随机事件驱动模型进行求解。在IEEE 33节点配电网与Sioux Falls市交通网形成的电力交通耦合系统中进行算例验证,结果表明所提模型和方法在电力交通耦合网故障恢复中有着较好的应用效果,可实时调控由随机事件(故障维修和交通行驶)导致的故障恢复变化。In order to reduce the load loss caused by distribution network faults and improve the resilience of distribution network,an uncertain event-driven fault recovery strategy based on multi-agent deep reinforce-ment learning is proposed.The uncertain event-driven problem in the fault recovery power-traffic coupling network is presented,which is described as a semi-Markov random decision process problem.An uncer-tain event-driven fault recovery model based on semi-Markov is constructed by considering the optimization objective of system fault recovery comprehensively.Then,the multi-agent deep reinforcement learning algo-rithm is used to solve the uncertain event-driven model.A case study is carried out in the power-traffic coupling system formed by IEEE 33-bus distribution network and Sioux Falls traffic network.The results show that the proposed model and method have good application effects in the fault recovery of power-traf-fic coupling network,and can adjust the fault recovery changes caused by uncertain events(fault mainte-nance and traffic travel)in real time.

关 键 词：随机事件驱动 故障恢复 深度强化学习 电力交通耦合网 多智能体 

分 类 号：TM732[电气工程—电力系统及自动化]