电力系统自动化2026，Vol.50Issue(1)：97-107,11.DOI:10.7500/AEPS20250611008
基于知识嵌入型深度强化学习的电力系统频率紧急控制方法
Emergency Frequency Control Method for Power Systems Based on Knowledge-embedded Deep Reinforcement Learning
摘要
Abstract
The rapid development of new power systems has exacerbated frequency security challenges,making emergency control crucial for restoring stability during faults.This paper proposes a power system emergency frequency control method based on knowledge-embedded deep reinforcement learning.First,the emergency frequency control problem is formulated as a Markov decision process,with a simulation system serving as the reinforcement learning environment,and a deep reinforcement learning(DRL)agent is constructed based on the deep deterministic policy gradient(DDPG)algorithm.Furthermore,theoretical knowledge guides the action space optimization,integrating both over-frequency generator tripping and under-frequency load shedding scenarios.Finally,the proposed method is validated on the IEEE 39-bus system,demonstrating that the DRL agent can generate effective emergency frequency control strategies to ensure system security,and the knowledge-embedded technique enhances training stability and significantly improves policy learning efficiency and decision quality.关键词
人工智能/新型电力系统/频率安全/频率紧急控制/深度强化学习/深度确定性策略梯度/高频切机/低频减载Key words
artificial intelligence/new power system/frequency security/emergency frequency control/deep reinforcement learning(DRL)/deep deterministic policy gradient(DDPG)/over-frequency generator tripping/under-frequency load shedding引用本文复制引用
LI Jiaxu,WU Junyong,SHI Fashun,ZHANG Zhenyuan,LI Lusu..基于知识嵌入型深度强化学习的电力系统频率紧急控制方法[J].电力系统自动化,2026,50(1):97-107,11.基金项目
国家重点研发计划资助项目(2018YFB0904500) （2018YFB0904500）
国家电网有限公司科技项目(SGLNDK00KJJS1800236). This work is supported by National Key R&D Program of China(No.2018YFB0904500)and State Grid Corporation of China(No.SGLNDK00KJJS1800236). （SGLNDK00KJJS1800236）
