Switch Cabinet Discharge Pattern Recognition Based on Ensemble Learning

Power switch cabinet is composed of circuit breakers, measuring appliances and protective appliances, and it plays an important role in the acceptance and distribution of electricity. In the process of design and manufacture of power switch cabinets, it is inevitable that there will be some defects which are closely related to the generation of partial discharge. Therefore, it is essential to identify and warn partial discharge fault. In this paper, a fault identification algorithm based on UHF signal of partial discharge in switch cabinet is proposed. Firstly, partial discharge (PD) of switch cabinet was simulated in the laboratory, and typical insulation defects were set up. Pd signals were collected by ultrahigh frequency sensors, a three-dimensional PRPS pattern of different defect types were drawn, and characteristic parameters of PRPS images were extracted. Ensemble learning algorithms XGBoost and LightGBM were used to recognize pd PRPS images of three typical defects of switch cabinet, and the recognition accuracy of each algorithm was compared by five-fold cross-validation. The results show that the characteristic parameters extracted in this paper can well reflect the pd characteristics of different defects and realize the task of fault diagnosis and defect recognition.