内蒙古电力技术2025，Vol.43Issue(4)：56-67,12.DOI:10.19929/j.cnki.nmgdljs.2025.0049
基于深度强化学习的新型终端配电网源荷储协同控制
Source-Load-Storage Collaborative Control for New Terminal Distribution Network Based on Deep Reinforcement Learning
摘要
Abstract
With the high proportion of distributed new energy aceess to the distribution network,the conventional decoupling voltage control,maximum consumption control and other methods can not meet the distribution network control needs under the condition of high proportion of new energy integration.Considering the operational economy,new energy consumption and voltage qualification rate of the distribution network,a new reward function for evaluating the overall economic benefits of active distribution network intelligent agents is proposed by combining time-of-use electricity prices,system losses and related economic losses.A distribution network control mathematical model containing high penetration distributed photovoltaics is constructed to improve the action effect of active distribution network agents under the source-load-storage collaborative control technology based on deep reinforcement learning.Through case analysis of the IEEE 33-node model,the universal applicability of distribution network source-load-storage collaborative optimization control technology based on deep reinforcement learning is verified in improving photovoltaic consumption rate,reducing line loss and suppressing voltage fluctuations.The per unit value of maximum voltage fluctuation decreases by 0.029 7.The maximum line loss decreases by 26.09％,and the photovoltaic consumption rate increases by 4.73％.关键词
配电网/分布式能源/深度强化学习/源荷储协同控制/新能源消纳Key words
distribution network/distributed energy/deep reinforcement learning/source-load-storage collaborative control/new energy consumption分类
信息技术与安全科学引用本文复制引用
齐军,马鹏,周生存,何予莹,杜鑫,赵爱国,杨勇,罗毅,刘海文..基于深度强化学习的新型终端配电网源荷储协同控制[J].内蒙古电力技术,2025,43(4):56-67,12.基金项目
内蒙古电力(集团)有限责任公司阿拉善供电分公司科技项目"园区级配电网源网荷储协同控制技术与示范应用项目(二次)"(ALSYS-2023-5-016) （集团）
