GIS Partial Discharge Fault Diagnosis Method Based on Synchrosqueezing Wavelet Transform and Vision Transformer

Gas Insulated Switchgear (GIS) is widely used in transmission and distribution systems because of its good insulation and high power supply reliability. The diagnosis of partial discharge (PD) defect types is of great significance for the stable and reliable operation of GIS. In this paper, a partial discharge fault diagnosis method based on synchrosqueezing wavelet transform and Vision Transformer is proposed. Firstly, the partial discharge pulse signal is processed by synchronous squeeze wavelet transform, and the gray level of pulse frequency spectrum is obtained, which constitutes the data sample of partial discharge diagnosis. Secondly, the data samples are input into the deep learning model for training. In order to improve the feature perception and extraction ability of SWT spectrum, the vision transformer model combined with the self-attention mechanism is adopted to enable the model to better analyze the position relationship of the input spectrum and understand the spatial structure of different regions in the spectrum. Finally, the sample to be tested is sent into the ViT model as input to test the diagnostic accuracy of partial discharge. The results show that the PD diagnosis method proposed in this paper can effectively identify different types of PD pulses. Compared with the two commonly used machine learning models, the accuracy rate increased by 6.4% and 1.8%.