In order to solve the problem of the difference in applicability of a single model in predicting the failure rate,TSOBP-ARIMA-Prophet combined model was proposed on the basis of considering the characteristics of the EMU traction system failure rate data.Firstly,in view of the complex nonlinearity of the EMU traction system failure rate,the tuna swarm algorithm(TSO)was introduced to optimize the BP model and train the TSOBP prediction model.Secondly,aiming at the non-stationary fluctuation of the failure rate,the ARIMA prediction model was selected.Then,according to the seasonal periodicity of the failure rate,the Prophet prediction model was selected.Finally,the reciprocal variance method was used to weight the prediction results of the three models,and the prediction results of the TSOBP-ARIMA-Prophet combined model were obtained.Taking an EMU traction system as an example,the combined model is used to predict the failure rate,and its prediction ability is verified by comparing with three single models and the TSOBP-ARIMA combined model.The results show that the mean square error of the combined model is 0.075 2,which is 45.83％,61.65％and 53.42％lower than that of the TSOBP,ARIMA and Prophet models respectively,and the prediction accuracy is significantly improved,and the perception of data trend is better than that of the TSOBP-ARIMA combined model,which can effectively improve the prediction ability of the failure rate of the EMU traction system.
维普
万方数据