电力系统自动化2026，Vol.50Issue(2)：103-114,12.DOI:10.7500/AEPS20250313002
基于自注意力扩散强化学习的配电网光伏承载力提升方法
Enhancement Method for Photovoltaic Accommodation Capacity of Distribution Networks Based on Self-attention Diffusion Reinforcement Learning
摘要
Abstract
To address the issue of insufficient accommodation capacity of distribution networks caused by high penetration of distributed photovoltaic(PV)integration,an enhancement method for PV accommodation capacity of distribution networks based on self-attention diffusion reinforcement learning is proposed.Aiming to enhance the PV accommodation capacity of the distribution network,a multi-objective optimization mathematical model is established from the perspective of collaborative optimization of distributed PV inverters on the source side,static var compensators on the grid side,flexible loads,and battery energy storage controllers.The proposed model takes into account both voltage deviation and system network loss indicators,with the goals of maximizing the distributed PV allowable capacity and minimizing the total operation cost.While satisfying the operation constraints,the method uses a diffusion mechanism to rapidly adapt to complex and uncertain environments,and leverages a self-attention mechanism to enable parallel reading of batch data,thereby improving training speed and solution efficiency.Finally,simulation analysis of a case study verifies the effectiveness and economic feasibility of the proposed method.关键词
配电网/分布式光伏/承载力/协同优化/强化学习/扩散/自注意力Key words
distribution network/distributed photovoltaic/accommodation capacity/collaborative optimization/reinforcement learning/diffusion/self-attention引用本文复制引用
杨文伟,彭显刚,蔡伟聪,欧阳昇,王星华,赵卓立..基于自注意力扩散强化学习的配电网光伏承载力提升方法[J].电力系统自动化,2026,50(2):103-114,12.基金项目
国家自然科学基金资助项目(62273104). This work is supported by National Natural Science Foundation of China(No.62273104). （62273104）
