Transient stability assessment of power systems with graph neural networks considering global features

Currently, the transient stability assessment of power systems using graph neural networks often overlooks the multidimensional characteristics of transmission lines and exhibits limited utilization of overarching features. To address this issue, this paper introduces a novel framework for graph neural networks, termed Global Features-Exploiting Edge Features for Graph Convolutional Networks (G-EGCN), specifically designed for transient stability assessment in power systems while considering global features. The proposed framework effectively harnesses the complete graph information of the power system by aggregating node features, edge features, and global features. Ultimately, a comprehensive validation of the proposed model's performance is conducted through simulation and comparative analysis on a 10-machine 39-node system.