电力信息与通信技术2025，Vol.23Issue(11)：1-7,7.DOI:10.16543/j.2095-641x.electric.power.ict.2025.11.01
基于联邦强化学习的配电网分布式电压控制策略
Decentralized Voltage Control Strategy for Distribution Network Based on Federated Reinforcement Learning
摘要
Abstract
In the growing power grid environment of distributed power generation,traditional voltage control strategies are limited due to privacy protection and scalability issues.In response to this challenge,based on the multi-agent reinforcement learning framework represented by"centralized training and decentralized execution",this paper proposes a decomposition and coordination reinforcement learning algorithm based on federated learning framework.Firstly,by combining the distributed computing and information exchange of local intelligent agents,a federated learning framework for the distribution network is constructed.Secondly,the model parameters are updated using stochastic gradient descent to ensure the convergence of the learning process.Finally,simulation analysis was conducted on the improved IEEE-141 node distribution system test case,and the experimental results showed that the proposed method significantly enhanced the scalability and privacy protection capabilities of the algorithm,and had a learning convergence speed that was not inferior to traditional centralized algorithms.关键词
电压控制/联邦学习/多智能体强化学习/分散控制Key words
voltage control/federated learning/multi-agent reinforcement learning/decentralized control分类
电子信息工程引用本文复制引用
赵猛,沈竹筠,李圆琪,马斌,朱志成,马千里..基于联邦强化学习的配电网分布式电压控制策略[J].电力信息与通信技术,2025,23(11):1-7,7.基金项目
国家自然科学基金项目(62303243). （62303243）
