Basin insulator is an important insulation component in gas insulated switchgear (GIS), and the accurate identification of its different defective partial discharge (PD) signals is of great significance to ensure the long-term safe and stable operation of GIS. In this paper, a classification and identification method for PD of basin insulators based on local mean decomposition (LMD) and long short-term memory (LSTM) neural network is proposed. Firstly, the PD signal is decomposed using the LMD method assisted by paired Gaussian white noise, and then the components obtained from the decomposition are segmented to extract the energy share, Renyi entropy and Hurst index of each segment to form feature matrices, and finally the feature matrices are sent to LSTM for training and classification. A basin insulator PD experimental platform is established in the laboratory to simulate the actual working conditions, and the PD signals of four different defects are collected for analysis and processing. The results show that the proposed method can effectively identify PD signals of different defects of basin insulators, and the feature parameters extracted by LMD decomposition can effectively characterize the characteristics of the PD signals in different frequency bands, and the recognition accuracy is significantly higher than that without LMD decomposition.