XGBoost Classifier for Fault Identification in Low Voltage Neutral Point Ungrounded System

In the neutral point ungrounded power supply system (IT system), detecting and judging the electric current of human body effectively from the total leakage current is difficult. It is necessary to identify whether electric shock fault occurs and the type of electric shock fault. This paper has proposed a new method based on signal processing and machine learning algorithm to solve them, only using phase voltage and current data. First, we view no-fault, ground fault, zero-line fault as three results of fault identification. Then, preprocess the data and extract features from them using wavelet analysis. At last, input the features and corresponding fault types into XGBoost algorithm to get a model. The accuracy can reach 90% when applying to test set. It shows effectiveness of the proposed method in neutral point ungrounded system fault classification.