Abstract: In order to solve the problem of misoperation and misdiagnosis of train control caused by abnormal data acquisition, this paper studied the intelligent control and diagnosis method of EMU based on BP neural network, established the prediction model based on BP neural network, obtained the optimal network model by using the train actual operation data for multiple training and parameter adjustment, combined the predicted value and actual value of the model to the final trusted value, and integrated into the existing train control logic for control and diagnosis. Through the experimental verification, the prediction results of the training model have higher accuracy compared with the actual acquisition value, and can achieve the prediction effect.The experimental results show that it is feasible to use BP neural network model to predict the state of the train, and combine with the relevant processing strategy to control the train operation and diagnose the fault.