Abstract: Faults of urban rail transit traction substation switchgear directly affect the security and steady operation of power supply system, also the safe and stable operation of urban rail transit system in case of fault, and even lead to major safety accidents in serious cases. Therefore, accurately predicting the change trend of switchgear operation state and realizing fault inference can scientifically guide the maintenance of switchgear and effectively reduce operation risks. A method based on LSTM (long short term memory) to predict the change trend of each switchgear state quantity is proposed, and WOA (whale optimization algorithm) is adopted to optimize LSTM model parameters. Grey Relation Analysis method is used to analyze the correlation degree between the predicted value of the traction substation switchgear state quantity and the fault. Based on the historical operation data of the switchgear in a southern city urban rail transit traction substation, the model is used for predictive analysis, and the information is provided to the operation maintenance personnel to guide the maintenance decision of the switchgear. The results show that the proposed method based on WOA-LSTM can accurately predict the switchgear faults and provide reliable guarantee for the operation of urban rail transit traction power supply equipment.