Abstract
The change-point detection of Power Quality (PQ) events has been a critical issue in smart grids because of the wide adoption of delicate power electronic devices. In this work, we design algorithms to achieve two goals. The first one is to detect various PQ events in the quickest sense under the false alarm constraints. The second goal is to mitigate the severe misdetection caused by the malfunctioned or affected sensors. To achieve these goals, we adopted the matrix cumulative sum (CUSUM) algorithm to perform the hypothesis test of different PQ events. Then, we propose a simplified matrix CUSUM algorithm to reduce the computational complexity by utilizing the statistical characteristics of PQ events. To overcome the degradation caused by malfunctioned or affected sensors, a fault tolerant decision fusion mechanism was proposed with slight cost of detection delay. Finally, simulation results are shown to validate the proposed algorithm and mechanism.