Robust Abnormal Detection in Large-scale Power Systems with Unknown Noise Statistics

Recent years have witnessed the broad deployment of phase measurement units (PMUs), enabling intelligent control of power systems, i.e., accurate abnormal detection. However, it is challenging to implement a model-driven indicator of the unbearable complexity in solving the high-dimensional power flow equations. In this paper, based on high dimensional statistics, we first propose a novel data-driven abnormal detection method for large-scale power systems with unknown information of noise, which relieves the pain of thick assumptions needed in classical data-driven methods. Furthermore, we propose to employ the Eigen-inference theory to estimate the unknown parameters in the noise model. Lastly, comprehensive experiments are carried out to verify the superiority of the proposed method over different scale benchmarks.