预测动车组牵引系统故障率的TSOBP-ARIMA-Prophet组合模型  

TSOBP-ARIMA-Prophet Combined Model for Predicting the Failure Rate of EMU Traction System

作　　者：张雨晨 戴贤春 刘敬辉 李秋芬 代成烨 ZHANG Yuchen;DAI Xianchun;LIU Jinghui;LI Qiufen;DAI Chengye(Graduate Department,China Academy of Railway Science,Beijing 100081,China;Railway Science&Technology Research&Development Center,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;Railway Safety Research Center,China State Railway Group Co.,Ltd.,Beijing 100081,China)

机构地区：[1]中国铁道科学研究院研究生部,北京100081 [2]中国铁道科学研究院集团有限公司铁道科学技术研究发展中心,北京100081 [3]中国国家铁路集团有限公司铁路安全研究中心,北京100081

出　　处：《中国铁道科学》2024年第4期147-157,共11页China Railway Science

基　　金：中国国家铁路集团有限公司科技研究开发计划课题(J2023B003,P2023T002)。

摘　　要：针对单一模型预测故障率时的适用性差异问题,在考虑动车组牵引系统故障率数据特点的基础上,提出TSOBP-ARIMA-Prophet组合模型。首先,针对动车组牵引系统故障率的复杂非线性,引入金枪鱼群算法(TSO)优化BP模型,训练出TSOBP预测模型;其次,针对故障率的非平稳波动性,选取ARIMA预测模型;然后,针对故障率的季节周期性,选取Prophet预测模型;最后,运用方差倒数法对3个模型的预测结果赋权,得到TSOBP-ARIMA-Prophet组合模型的预测结果。以某动车组牵引系统为例,采用该组合模型预测故障率,并与3个单一模型及TSOBP-ARIMA组合模型对比验证其预测能力。结果表明:该组合模型预测时均方误差为0.0752,较TSOBP,ARIMA和Prophet模型单独预测时分别降低了45.83%,61.65%和53.42%,预测精度显著提高,且较TSOBP-ARIMA组合模型对数据趋势的感知力更优,可有效提升对动车组牵引系统故障率的预测能力。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 0.0752,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.

关 键 词：动车组牵引系统 故障率预测 组合模型 BP模型 金枪鱼群算法 ARIMA模型 Prophet模型 

分 类 号：U266[机械工程—车辆工程]