Power Network Topology Identification based on Graph Attention Network

Abstract

Identifying and tracking power grid topology changes is a prerequisite requirement for grasping the operating behavior of the power system. The data collected by the power information system can be used to obtain the characteristic quantities describing the state of lines and nodes in the power grid. The traditional classification method ignores the correlation between samples, which affects the identification accuracy. In this paper, we propose a topology identification method based on graph attention network. Firstly, we describe the line features according to the grid information data. Secondly, based on the graph attention network method, we identify line status by line classification. Finally, the power network topology is generated by combining the adjacency matrix and line classification decision to achieve topology identification. IEEE59 node system and an actual provincial power system are tested. The result shows that the proposed method is effective and more accurate in topology identification than traditional methods.