基于深度强化学习的电网自适应保护策略优化研究

打开文本图片集

中图分类号：TM77 文献标志码：A 文章编号：1003-5168（2025）21-0013-04

DOI：10.19968/j.cnki.hnkj.1003-5168.2025.21.003

Abstract： [Purposes] To address the high real-time requirements and diverse fault types in power grid protection systems，this study proposes an adaptive protection strategy optimization method based on deep reinforcement learning.[Methods] A hybrid architecture combining a Deep Q-Network （DQN） with a policy gradient algorithm is adopted to construct a state-aware model and action decision mechanism for intelligent parameter adjustment.The proposed method was validated through testing on the IEEE-39 node system. [Findings] Compared with traditional protection strategies，the proposed method reduces the false trip rate to 1.1% ，decreases the failure-to-trip rate to 2.1% ，and shortens the protection action time to 38.9 ms across various fault scenarios.[Conclusions] Experimental results demonstrate that deep reinforcement learning effectively enhances the adaptive capability and reliability of power grid protection systems，providing new theoretical support for the development of smart grid protection technologies.

Keywords： deep reinforcement learning; power grid protection; adaptive strategy; intelligent optimization; fault diagnosis

0 引言

随着可再生能源渗透率不断提升与电网互联规模持续扩大，现代电网呈现出高度复杂性与强耦合性特征。（剩余4129字）