基于强化学习的电力系统应急物资仓储控制算法

Reinforcement Learning Based Storage Control Algorithm for Emergency Materials in Power System

在人工智能飞速发展的启发下,智能库存控制和调度系统被认为是电力系统稳定性的一个有效的解决方案。提出了一种“端到端”强化学习方法,用于联合优化库存控制和应急电源交付策略,以平衡维护成本和电力系统稳定性。所提出算法为“端到端”算法,算法不预测需求,直接做出库存控制和调度决策;所提出的算法为“在线”算法,即库存控制和调度决策仅依赖于对过去事件的观察;所提出的算法也是“无模型”算法,即算法不依赖于任何假定的不确定事件随机模型。通过利用真实数据进行数值模拟,表明所提出的“端到端”强化算法平均比有代表性的基准算法在性能上高出138.5%。

Inspired by the rapid development of artificial intelligence,the intelligent inventory control and scheduling system is considered as an effective solution for the stability of the power system.This paper proposes an"end-to-end"reinforcement learning approach to jointly optimize inventory control and emergency power delivery strategies to balance maintenance costs and power system stability.The proposed algorithm is an"end-toend"algorithm,which does not predict demand and directly makes inventory control and scheduling decisions.The proposed algorithm is the"online"algorithm,where inventory control and scheduling decisions rely only on observations of past events.The proposed algorithm is also a"model-free"algorithm,where the algorithm does not rely on any putative stochastic model of uncertain events.By using numerical simulations using real data,it is showed that the proposed"end-to-end"reinforcement algorithm achieves on average 138.5%higher performance than the representative benchmark algorithm.representative benchmark algorithm.