A method of Wasserstein generative adversarial network with gradient penalty (WGAN-GP) data enhancement method for pattern recognition of partial discharge (PD) is proposed. Firstly, denoising and pulse extraction are carried out to obtain local emission pulse signal. Secondly, Wasserstein distance and gradient penalty function are introduced into generative adversace network (GAN) to improve the training stability of the generated model and the diversity of generated samples. Partial discharge pulse signals are used as samples to train the network to realize the enhancement of partial discharge data based on WGAN-GP. The proposed method is used to enhance the partial discharge sample data, and the enhanced partial discharge samples are used to train the common pattern recognition algorithms. The experimental results show that compared with traditional data enhancement method, the proposed method can more effectively enhance the partial discharge pulse sample data, the recognition accuracy of the data discharge pulse pattern can be increased by 12.9%.