Aiming at the problems of high dimensionality and too much invalid information in traditional signal feature extraction methods, this paper proposes a switchgear partial discharge pattern recognition method based on linear discriminant algorithm and radial basis neural network. This method combines the two algorithms to achieve the purpose of both the recognition rate and the recognition accuracy. First, establish three ultrasonic partial discharge (PD) models of the switchgear. Then, time-frequency analysis and wavelet decomposition are used to extract the time-frequency features and wavelet coefficient features of the signal, and the extracted feature vectors are reduced by linear discriminant algorithm (LDA). Finally, the radial basis basis (RBF) neural network is used to analyze the local The types of discharge defects are classified, and the recognition accuracy is above 90%, and the training time is reduced by more than 50%, which proves that the recognition method is practical.