Impact of Topology Noise on Power Systems State Estimation Using a Temporal GCN Framework

Graph Convolutional Networks (GCNs) have demonstrated great potential in analyzing energy data for learning the complex interactions and dynamics for supporting various functions within power systems including state estimation. However, these models are susceptible to noise in their underlying graph structure. In this paper, topology noise refers to the presence of a few additional or missing links in the power system graph model, caused by inaccurate information about the structure of the system and state of the lines or adversarial attacks on the graph’s structure. The focus of this work is on evaluating the effects of topology noises or attacks and their location on the performance of a Temporal Graph Convolutional Network (TGCN) framework for power system state estimation.The results of this study demonstrate the TGCN framework’s sensitivity in the presence of topology noises and attacks for state estimation in power systems. This study provides new insight regarding areas of vulnerability that could be exploited by such disturbances.