摘要
针对分布式风/光发电大规模接入智能配电网产生的电压越限和系统潮流的不确定性问题,提出了基于强化学习的测控一体化分布式配电网电压协同优化控制算法。以有载调压器将配电网母线平均节点电压控制在安全阈值内为控制目标,采用深度Q网络算法实现在长时间尺度内的全局电压最小误差控制;建立配电网运行节点的实时电压协调控制模型,利用多智能体强化学习算法实现了多时间尺度和跨设备的调压优化目标。采用MATLAB软件建立IEEE 33节点配电网系统仿真模型,验证了所提方法的有效性。结果表明,所提出的控制方法具有可靠性和实用性。
A voltage collaborative optimization control algorithm for distributed distribution network with integrated measurement and control based on reinforcement learning was proposed to address the issues of voltage exceeding limits and system flow uncertainty caused by large-scale connection of distributed wind/solar power generation into intelligent distribution networks.Specifically,the control objective was to use an on load voltage regulator to control the average node voltage of the distribution network bus within a safe threshold,and a deep Q-network algorithm was applied to achieve global voltage minimum error control over a long time scale;a realtime voltage coordination control model for the operation nodes of the distribution network was established,and the multi-agent reinforcement learning algorithm was used to achieve the optimization goals of voltage regulation across muliple time scales and devices.A simulation model of IEEE 33 node distribution network system was built using MATLAB software to verify the effectiveness of the proposed method.The results indicate that the proposed control method is reliable and practical.
作者
王定美
梁利
江兴亮
马吉龙
岳金鑫
Wang Dingmei;Liang Li;Jiang Xingliang;Ma Jilong;Yue Jinxin(Lanzhou Longneng Electric Power Science&Technology Co.,Ltd.,Lanzhou Gansu 730070,China)
出处
《电气自动化》
2025年第3期97-101,共5页
Electrical Automation
基金
企业科技项目“分布式配电网测控一体化控制策略研究”(kjyf-2024-j0105)。
关键词
智能配电网
强化学习
电压控制
数据驱动
马尔可夫决策过程
intelligent distribution network
reinforcement learning
voltage control
data-driven
Markov decision process
