基于图注意力网络的配电网故障区段判断  

Research on Fault Section Diagnosis of Distribution Network Based on Graph Attention Network

作　　者：李志强 LI Zhiqiang(State Grid Zhuzhou Power Supply Company,Zhuzhou 412000,China)

机构地区：[1]国网株洲供电公司,湖南株洲412000

出　　处：《电工技术》2025年第4期112-114,共3页Electric Engineering

摘　　要：配电网的结构通常包含多个分支和大量的设备,且配电网故障类型多种多样,增加了配电网故障区段判断的难度,为此研究了一种基于图注意力网络的配电网故障区段判断方法。收集配电网的拓扑结构数据、节点特征数据,构建配电网图结构。通过多个独立的注意力头计算每个节点与邻居节点的注意力系数,实现基于图注意力网络的配电网关键区段节点类别划分。结合风险理论法计算配电网中输电的风险值,根据不同线路的风险值与风险造成的负荷损失进行配电网不同区段输电线路权重的计算。根据配电网的拓扑结构将配电网划分为多个区段,从而实现故障区段划分。实验测试结果表明,所设计方法可以实现对配电网故障区段的精准判断,实际应用效果好。The structure of distribution networks usually includes multiple branches and a large number of devices,and the types of faults in distribution networks are diverse,which increases the difficulty of identifying fault sections in distribution networks.Therefore,a fault section identification method for distribution networks based on graph attention networks is studied.Collect topology data and node feature data of the distribution network,and construct a distribution network diagram structure.By using multiple independent attention heads to calculate their attention coefficients with neighboring nodes,the classification of key section nodes in distribution networks based on graph attention networks is achieved.Calculate the risk value of transmission in the distribution network using the risk theory method,and calculate the weight of transmission lines in different sections of the distribution network based on the risk value of different lines and the load loss caused by the risk.Divide the distribution network into multiple sections based on its topology,in order to achieve fault section division.The experimental test results show that the designed method can achieve accurate judgment of fault sections in the distribution network,and the practical application effect is good.

关 键 词：图注意力网络 节点划分 关键区段 判断 故障区段 配电网 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]