A LSTM Fault Diagnosis Method Based on Zero-sequence Voltage for Small Current Grounding System

A fault line selection method based on clustering and the Long Short-Term Memory(LSTM) is proposed for the single-phase grounding fault line selection problem in distribution networks containing renewable energy. Firstly, the fault recording data of the zero-sequence voltage is collected based on the fault recording device and the D5000 system, and the ratio of harmonics to fundamental wave in the zero-sequence voltage is found as the characteristic quantity. Cluster analysis is used to classify the data, add labels, and determine the fault moment. A LSTM algorithm is used to build a fault line selection model. Finally, the effectiveness of the model is validated with examples, and the results show the effectiveness of the proposed method.