云南师范大学学报（自然科学版）2025，Vol.45Issue(6)：26-31,6.DOI:10.7699/j.ynnu.ns-2025-066
基于多时间尺度强化学习的光伏—储能协同电压调节方法
Photovoltaic Energy Storage Collaborative Voltage Regulation Method Based on Multi-time Scales Reinforcement Learning
摘要
Abstract
The integration of photovoltaic power generation into the distribution network could cause the local voltage fluctuation,therefore a photovoltaic energy storage collaborative voltage regulation method based on multi-time scales reinforcement learning was proposed.The voltage regulation mainly focuses on energy storage unit regulation,with the optimization goal of minimizing the daily operating cost of the distribution network.By optimizing the charging and discharging strategies of the energy storage system and the distribution of photovoltaic output,the complete consumption of photovoltaic power generation was achieved,ensuring the economic stability of the distribution network.The intra-day voltage regulation was further subdivided into upper level long-term economic optimization and lower level short-term voltage stability control.Based on the current voltage regulation and the intra-day voltage regulation,a photovoltaic energy storage collaborative voltage regulation model was built using multi-time scale reinforcement learning.The model used deep Q-network algorithm and actor-critic algorithm to train upper and lower level reinforcement learning model agents,respectively.Through a double-layer deep reinforcement learning strategy,the photovoltaic energy storage collabo-rative voltage regulation amount was obtained,thereby achieving photovoltaic energy storage voltage collaborative regulation.The experimental results showed that after adjusted by the design method,the voltage of each node was basically stable around the rated value of 1.0 p.u.,and the fluctuation range was relatively small and within the safe operating range,indicating that this method can effectively solve the problem of local power imbalance in the distribution network.关键词
多时间尺度/强化学习/光伏—储能协同/电压调节/深度Q网络/智能体训练Key words
Multi-time scales/Reinforcement learning/Photovoltaic energy storage synergy/Voltage regulation/Deep Q-network/Intelligent agent training分类
信息技术与安全科学引用本文复制引用
张济凡,夏泰宝,周磊,周晨,李晓亮,王德顺..基于多时间尺度强化学习的光伏—储能协同电压调节方法[J].云南师范大学学报（自然科学版）,2025,45(6):26-31,6.基金项目
国网江苏省电力有限公司科技资助项目(J2024069). （J2024069）
