基于LSTM网络的牵引变流器IGBT故障预测方法研究  

A Fault Prediction Method of IGBT in Traction Converter Based on LSTM

作　　者：高伟[1] 张琼洁[1] 李长留 李建龙[1] GAO Wei;ZHANG Qiongjie;LI Changliu;LI Jianlong(Vehicle Engineering Department,Zhengzhou Railway Vocational&Technical College,Zhengzhou He’nan 450052,China)

机构地区：[1]郑州铁路职业技术学院机车车辆学院,河南郑州451460

出　　处：《电子器件》2020年第4期804-808,共5页Chinese Journal of Electron Devices

基　　金：河南省高等学校重点科研项目《基于PLC-S7200的HXD系列电力机车控制与故障监测系统的研究开发》。

摘　　要：绝缘栅双极型晶体管(Insulated Gate Bipolar Transistor,IGBT)作为牵引变流器失效率较高且最易损器件,对其进行状态监测和故障预测,可以有效减少日常维护成本,提高列车安全性和可靠性。首先,对IGBT失效原因和性能退化参数进行研究,选择集电极-发射极关断峰值电压作为性能退化参数,研究了一种基于三层长短记忆(Long Short-Term Memory,LSTM)网络的IGBT故障预测方法,并给出IGBT维护决策理论依据。最终在NASA PCoE研究中心提供的IGBT加速老化数据集上开展验证实验,实验结果表明多层LSTM网络相比于传统浅层模型具有更高的预测精度。IGBT has the highest failure rate and it is the most fragile component of traction converter.The condition monitoring and fault prediction of IGBT not only can reduce the daily maintenance costs,but improve the safety and reliability of metro train.Firstly,the failure mechanism of IGBT is deeply studied,the transient peak voltage of collector-emitter turn-off is selected as the characteristic parameter of fault prediction.A fault prediction method for IGBT based on LSTM network is proposed,and the theoretical basis of decision support for maintenance is provided.Finally,the experiment is carried out on the accelerated aging data of IGBT provided by NASA PCoE research center,and the results show that the three layers LSTM network can achieve higher prediction accuracy than the traditional shallow machine learning model.

关 键 词：牵引变流器 绝缘栅双极型晶体管 故障预测 长短记忆网络 

分 类 号：TM464.55[电气工程—电器]