To identify the type of partial discharge signal quickly and accurately for the safe operation of equipment, a method of partial discharge feature extraction and discharge recognition based on S transform and genetic algorithm optimization probabilistic neural network (GA-PNN) is proposed in this paper. Firstly, the time-frequency diagram and time-frequency matrix A of partial discharge signal are obtained by S-transform, to reduce the dimension of the matrix and remove the Gauss noise at the same time, the time-frequency matrix B of the main characteristics of partial discharge is extracted from A by S-transform time-frequency diagram and spectral kurtosis algorithm. Then singular value decomposition of time-frequency matrix B is carried out to extract appropriate number of singular values as eigenvectors to be the input of probabilistic neural networks (PNN), and genetic algorithms (GA) is used to optimize the parameters of PNN to realize the identification of partial discharge signals. Results show that the proposed method can identify the type of cable partial discharge very well, and has a better recognition effect than GA-BP.