Pattern Recognition of UHF Partial Discharge Signals in GIS Based on Deep Learning

The diagnosis of type of the ultra high frequency (UHF) partial discharge (PD) signals in gas insulated switchgear (GIS) can effectively prevent the occurrence of equipment failure. Firstly, a GIS basin-type insulator test platform is established to simulate the actual PD defect in GIS. Secondly, according to the characteristics of the UHF PD signals, the spectrogram is established, which characterizes its energy distribution on the time-frequency domain. Then, the dimensionality reduction and feature extraction are carried out by modified MFCCs (MMFCCs). Finally, the depth neural network model based on the gated recurrent unit (GRU) is established for PD type recognition. The results show that the model can effectively identify all kinds of PD defects of GIS in the laboratory conditions, and have a significant advantage over other machine learning algorithms.