Power System Small-signal Stability Assessment Model Based on Residual Graph Convolutional Networks

Abstract

Abstract Small-signal stability (SSA) is important to power system security. A data-driven approach is established for rapid prediction of the power system oscillation characteristics. The key of the approach is the Graph Convolution Networks (GCN) with residual mechanism, which works to aggregate features from high-dimension steady-state operation information and is denoted as ResGCN (RESidual GCN) in the paper. The residual mechanism helps to overcome the network degradation phenomenon. Both the oscillation frequency and damping ratio of multiple modes can be predicted by the proposed model. The performance of the proposed scheme as well as its adaptability to the power system topological changes is verified on the IEEE 39 Bus system.