Partial discharge (PD) is an important reason for the deterioration of GIS insulation performance. Accurate PD pattern recognition is of great significance to GIS operation and maintenance. Due to the small number of samples, low recognition accuracy and long running time of traditional PD pattern recognition methods, this paper proposed a GIS PD pattern recognition method based on improved WGAN and MobileNet-V2 network. Firstly, the test platform for PD was designed and built to obtain UHF signals under typical defects, and the PRPD spectrum of UHF signals was generated. Then, the improved WGAN was used to expand the PRPD spectrum. Finally, the pattern recognition of PD was realized based on MobileNet-V2 network. The results show that the proposed method which has less parameters can effectively solve the problem of insufficient data volume, and it has a high accuracy. So the model can be applied to the GIS operation and maintenance process, which has practical engineering value.

GIS,MobileNet,WGAN,Pattern Recognition,Partial Discharge