Power System Critical Clearing Time Prediction Based on Graph Attention Network

In recent years, data-driven related methods have been widely applied in fields such as power system transient stability assessment. However, most traditional data-driven methods are used to analyze Euclidean data, and their depiction of the topological connection relationship of power network is limited, resulting in insufficient generalization ability of traditional methods in new topologies. This paper, combining with the idea of graph deep learning, introduces the graph attention network (GAT) into power system critical clearing time (CCT) prediction, designing a multi-head attention scheme. The paper conducts simulations on the IEEE 39-bus system. By comparing the performance with critical time prediction models established under other deep learning models, the superiority of the proposed model is verified.