Topology-Adaptive Small-Signal Stability Assessment Based on An Edge-Graph Learning Scheme

Long-term power system operation security requires data-driven small-signal assessment (SSA). Existing SSA schemes neglect edge features like branch powers and become vulnerable when applied to inter-area low-frequency oscillation (LFO) mode prediction under unseen topologies. We propose an edge-graph learning-based SSA scheme (EGL-SSA) based on steady-state measurements. The graph construction focuses on the edge features, before edge-node convolution (ENC) layers finish the feature extraction on both edges and nodes. Then the frequencies and damping ratios of various modes can be predicted in parallel via a multi-task structure with different downstream networks. Test results on the IEEE 39 Bus system indicate the superiority of our scheme under unseen topologies, especially aimed at the inter-area LFO modes.