Gas Insulated Switchgear (GIS) plays a core role in power systems, and its insulation performance is crucial for the stable operation of the system. Partial Discharge (PD) is one of the main indicators of the deterioration of GIS insulation performance. This paper proposes a GIS insulation defect partial discharge pattern recognition method based on Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) and ResNet50. By establishing four typical partial discharge defect models and collecting PRPD spectra to build a dataset, the WGAN-GP network is used to perform data augmentation. The augmented dataset is then input into the ResNet50 classification network to achieve accurate identification and classification of partial discharge defect types. Experimental results show that this method demonstrates excellent accuracy and stability in partial discharge defect recognition, significantly improving the defect identification accuracy.