A Hypothesis Testing Approach for Topology Error Detection in Power Grids

When the grid topology is changed due to incidents and the state estimator is not updated with the topological change, it is considered a topology error. In this paper, we develop a new method for detecting topology errors in power grids. The proposed method considers the measurement data as a nonstationary Gaussian process, explores the dependence structure of the underlying process. It detects errors by testing the hypothesis of whether the mean vector of a nonstationary Gaussian process is zero and does not rely on the convergence of the standard weighted least-squares (WLS) state estimation algorithm. It is very effective in detecting topology errors, in which multiple conforming errors may occur and the traditional state estimation algorithms may fail to converge. Simulation results show that it can accurately identify the abnormal measurements caused by the topology error.