基于GNN的复杂电网薄弱节点辨识  

Identification of weak nodes in complex power grids based on GNN

作　　者：李垚逸 吕飞鹏[1] 刘利文 魏韶韶 朱玉勇 LI Yaoyi;LU Feipeng;LIU Liwen;WEI Shaoshao;ZHU Yuyong(College of Electrical Engineering,Sichuan University,Chengdu 610065,China;China Three Gorges Construction Engineering Corporation,Chengdu 610000,China)

机构地区：[1]四川大学电气工程学院,四川成都610065 [2]中国三峡建工(集团)有限公司,四川成都610000

出　　处：《武汉大学学报(工学版)》2025年第3期407-415,共9页Engineering Journal of Wuhan University

基　　金：国家自然科学基金项目(编号:51907097)。

摘　　要：为提高电网薄弱节点辨识模型的准确度,将节点特征与网络参数相结合,提出基于图神经网络(graph neural network,GNN)和复合节点特征的薄弱节点辨识方法。首先,分析电网结构,计算节点各项电气指标并将其作为染色体编码数据,以准确描述节点电气特征,同时增强节点间差异,提高辨识结果准确度和区分度;其次,利用GNN模型,建立节点间通过多条互异通路产生的关联关系,并对传统辨识方法中的节点间最短潮流路径理论进行优化,使得辨识模型更贴合电网实际,计算结果更加合理。算例分析表明,所述方法的辨识结果准确度高,且节点薄弱指标区分明显。In order to improve the accuracy of weak node identification model of power grid,a weak node identification method based on graph neural network(GNN)and composite node features is proposed by combining node features with network parameters.Firstly,the power grid structure is analyzed and each electrical index of nodes is calculated as chromosomal coding data to accurately describe the electrical characteristics of nodes.At the same time,the differences between nodes are enhanced to improve the accuracy and differentiation of the identification results.In addition,GNN model is used to establish the correlation between nodes through a number of different paths,and the theory of the shortest power flow path between nodes in traditional identification methods is optimized,so that the identification model is more suitable for the actual power grid and the calculation results are more reasonable.The example analysis shows that the identification results of the proposed method are more accurate and the node weakness index is distinguished clearly.

关 键 词：图神经网络 复杂网络 薄弱指标 染色体编码 稳定性研究 

分 类 号：TM711[电气工程—电力系统及自动化]