基于图神经网络与强化学习的配电网电压与无功功率优化方法  

Voltage and Reactive Power Optimization Method for Distribution Networks Based on Graph Neural Network and Reinforcement Learning

作　　者：朱涛 海迪 李文云 黄伟 周胜超 吴明贺 王逸飞 ZHU Tao;HAI Di;LI Wenyun;HUANG Wei;ZHOU Shengchao;WU Minghe;WANG Yifei(Honghe Power Supply Bureau,Yunnan Power Grid Co.,Ltd.,Honghe,Yunnan 661100,China;Power Dispatching and Control Center,Kunming Power Supply Bureau,Kunming 650011,China;Yunnan Power Grid Co.,Ltd.,Kunming 650217,China;School of Electrical Engineering,Southeast University,Nanjing 210096,China)

机构地区：[1]云南电网有限责任公司红河供电局,云南红河661100 [2]昆明供电局电力调度控制中心,昆明650011 [3]云南电网有限责任公司,昆明650217 [4]东南大学电气工程学院,南京210096

出　　处：《南方电网技术》2024年第11期67-78,共12页Southern Power System Technology

基　　金：国家自然科学基金资助项目(52007032)~~。

摘　　要：高比例分布式光伏的接入改变了配电网的运行方式,并导致配电网出现有功功率损耗过大、调压设备寿命下降、节点电压越限等一系列问题。基于此背景,首先将电压无功优化问题建模为一个马尔科夫决策过程,然后使用无模型的深度强化学习方法进行求解,该方法可以从历史运行数据中捕获光伏发电的间歇性和负荷的波动性。提出了一种图卷积网络的近端策略优化算法(graph convolutional network-proximal policy optimization,GCN-PPO),该算法通过嵌入图卷积网络来提高强化学习智能体对配电网图数据的感知能力。最后以改进的IEEE 33节点测试系统开展算例分析,验证了所提方法的有效性和相比其他方法的优势。结果表明,基于图卷积网络训练的强化学习智能体在配电网拓扑发生变化和测量数据丢失时仍表现出较好的性能。High propotional distributed photovoltaic integration changes the operation mode of the distribution networks,and leads to a series of problems such as excessive active power losses,reduced service life of regulating equipment,and exceeding node voltage limits in the distribution networks.Based on this background,firstly the voltage and reactive power optimization problem is modelled as a Markov decision process,which is solved by using a model-free deep reinforcement learning method that captures the intermit⁃tence of PV and load fluctuation from historical operating data.A graph convolutional network-proximal policy optimization(GCNPPO)algorithm is proposed which improves the perception of reinforcement learning agent on graph data of distribution networks by embedding the graph convolutional network.Finally,an arithmetic analysis is carried out with a modified IEEE 33-node test system to verify the effectiveness of the proposed method and its advantages over other methods.The results show that the trained reinforce⁃ment learning agent based on graph convolutional networks exhibits better performance when the topology of the distribution network changes and the measurement data are lost.

关 键 词：分布式光伏 配电网 电压无功优化 深度强化学习 图卷积网络 

分 类 号：TM711[电气工程—电力系统及自动化] TM727