Transient Stability Assessment of Improved Graph Attention Networks Considering Dense Electric Transmission Channel

In engineering, the performance of data-driven transient stability assessment of power systems will continue to deteriorate due to factors such as communication delays and dense channel failures, as well as the problem of missing dense channel data. This paper proposes a transient stability assessment model based on Long Short Term Memory Graph Attention Network (LSGAT) to address this issue. After obtaining the mask matrix representing the observability of the system based on the original network topology and PMU configuration scheme, observable data is obtained through the mask matrix and the model is trained under any missing PMU condition; During the training process, the LSGAT network extracts spatial features of incomplete measurement data through a multi head attention mechanism, captures spatiotemporal correlations through a long short-term memory network, and enhances the model's learning ability for unstable samples using a focus loss function. The simulation results show that the proposed method can maximize the utilization of observable data, has high accuracy and strong robustness, and is not limited by network topology, making it easy to transfer.