基于预演算法的灾后配电网应急抢修决策优化  

Rollout Algorithm-Based Post-Disaster Distribution System Emergency Repair Optimization

作　　者：李明轩 魏韡[1] 许寅 张琪祁 LI Mingxuan;WEI Wei;XU Yin;ZHANG Qiqi(Department of Electrical Engineering,Tsinghua University,Beijing 100084,China;School of Electrical Engineering,Beijing Jiaotong University,Beijing 100044,China;State Grid Shanghai Municipal Electric Power Company,Shanghai 200122,China)

机构地区：[1]清华大学电机工程与应用电子技术系,北京市100084 [2]北京交通大学电气工程学院,北京市100044 [3]国网上海市电力公司,上海市200122

出　　处：《电力建设》2023年第12期85-94,共10页Electric Power Construction

基　　金：国家重点研发计划青年科学家项目(2022YFB2405500);国家电网有限公司科技项目(52094023001H)。

摘　　要：为应对极端灾害导致的停电损失,需合理构建灾后恢复模型,探索有效的灾后应急决策方法。构建了考虑关键负荷失效事件的灾后应急抢修马尔科夫决策过程模型,在计算损失时考虑了停电时长对不同类型关键负荷功能性的影响;对于灾后应急决策,提出了基于预演算法的在线决策方法,通过多场景仿真的方式快速评估动作-状态价值函数(即Q函数),进而生成在线优化策略,算例证明了基于预演仿真生成Q函数的决策方案优于基础策略提供的方案。所提方法不依赖精确的数学模型和参数,可以根据系统状态实时更新策略,比传统离线优化模型更具实用性。In order to cope with the loss from power outages caused by extreme disasters,it is necessary to construct a reasonable post-disaster recovery model and explore effective post-disaster emergency decision-making methods.In this study,a Markovian decision process model for post-disaster emergency recovery considering critical load failure events is constructed.It also considers the impact of outage duration on the functionality of different types of critical loads when calculating the losses.For post-disaster emergency decision-making,an online decision-making method based on a preview algorithm is proposed to quickly evaluate the action-state value function(i.e.,Q function)by means of multi-scenario simulation and generate an online optimization strategy.It is demonstrated that the decision scheme based on the Q function generated by the rollout algorithm is better than that provided by the underlying strategy in the simulation.The proposed method does not rely on precise mathematical models and parameters and can update the policy in real time according to the system state,which is a more practical approach than that of the traditional offline optimization model.

关 键 词：韧性 配电网 灾后故障恢复 在线决策 预演算法 

分 类 号：TM712[电气工程—电力系统及自动化]