基于近端策略优化的变电站运维备件库存动态优化研究  

Research on Dynamic Optimization of Substation Maintenance Spare Parts Inventory Based on Proximal Strategy Optimization

作　　者：陈延秋 杨勇 王豹 李俊杰 张国栋 CHEN Yanqiu;YANG Yong;WANG Bao;LI Junjie;ZHANG Guodong(Guangzhou Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Guangzhou 510620,China)

机构地区：[1]广东电网有限责任公司广州供电局,广东广州510620

出　　处：《电工技术》2025年第3期33-38,共6页Electric Engineering

摘　　要：在变电站的整个生命周期内会随机出现多种重要程度各不相同的维修任务。不同种类的维修任务通常需要不同类型的维修备件,而同一类维修备件往往可以用于多种维修任务中。结合知识迁移和折扣方法,提出了一种基于近端策略优化的高效强化学习算法,并将该算法在一系列大型变电站维修备件库存系统上进行了测试,结果表明在大多数条件下,由该算法生成的库存控制策略始终优于其他文献中的先进启发式算法。Throughout the life cycle of a substation,a variety of maintenance tasks of varying importance randomly occur.Different types of maintenance tasks usually require different types of maintenance spares,and the same type of maintenance spares can often be used in multiple maintenance tasks.Combining knowledge transfer and discounting methods,an efficient reinforcement learning(RL)algorithm based on proximal policy optimization(PPO)is proposed.The algorithm was tested on a series of large-scale substation maintenance spare parts inventory systems.Under most conditions,the inventory control policies generated by the algorithm consistently outperform state-of-the-art heuristics in the literature.

关 键 词：变电站维修 备件管理 库存控制 近端策略优化 强化学习 

分 类 号：TM63[电气工程—电力系统及自动化]