Data-driven Fault Detection and Cause Identification Method for Distribution Systems

Fast detection and correct cause identification of grounding faults are important measures to ensure the safe operation of distribution systems. However, the current methods mainly rely on "manual patrol", which reduces the efficiency and reliability of the cause identification. Based on the fault recording data, a data-driven fault detection and cause identification method for distribution network is proposed. Firstly, the field waveforms are analyzed to obtain the fault characteristics of different causes. And the change rate of zero-sequence current is calculated to detect the fault starting time. Secondly, the waveform is decomposed according to different time scales based on ensemble empirical mode decomposition method to extract the local features. And principal component analysis method is used to extract the main feature quantities. In addition, a fault cause classification model based on temporal convolutional network is proposed. The experimental results using field data show that the proposed method has high accuracy.