Research And Realization Of IGBT Overheating Failure Monitoring System

Insulated Gate Bipolar Transistor(IGBT),as the core component of the power electronic system,has been widely used in aerospace,industrial automation,transportation and other fields.However,the explosive growth of the new energy field in recent years,and provide a new opportunity for the in-depth popularization of IGBT.With the performance improvement and capacity expansion of various IGBTs,the device overheating fault has become the main factor restricting their rapid development,and the root cause of IGBT overheating failure is often caused by comprehensive factors such as device overcurrent or various short-circuit failures.This failure often occurs in complex and highly integrated power electronic systems.Moreover,the IGBT’s own withstand characteristics are only in the subtle level,which brings great difficulties for the monitoring and protection of devices.At this stage,researchers mostly focus on the analysis of a certain parameter of IGBT,but the comprehensive system for IGBT overheating fault monitoring and analysis has not been realized.Collecting and analyzing multiple thermoelectric parameters of IGBT to monitor the working status of IGBT or predict the health degree of the device has become a key research topic in this field.Aiming at such difficult problems,this article focuses on the IGBT chip by designing an overheating fault monitoring experimental platform,studying and analyzing the core thermoelectric parameters,combined with the optimized neural network,so as to realize the IGBT status monitoring and reduce the chip overheating failure rate.The following content is the main research work of this article:(1)At the beginning of the research of this subject,the basic principles and device characteristics of the IGBT were analyzed,and the internal loss status and the essential causes of component failure were deeply understood.The theoretical analysis was combined with simulation software to demonstrate;The analysis and introduction of IGBT overheating faults or closely related to each class fault types,and provide corresponding IGBT protection methods according to the characteristics of different fault types;(2)Aiming at the characteristics of IGBT condition monitoring and protection methods,this article proposes an overheating prediction based on the LSTM algorithm,using it as an auxiliary method of the IGBT overheating fault monitoring system,combining software and hardware,thereby designing an IGBT fault prediction algorithm based on the optimized LSTM,And demonstrate the superiority of the design algorithm through comparative experiments;(3)In order to realize the condition monitoring and fault prediction of the IGBT,this experiment designed and implementation an IGBT overheating fault monitoring system at the hardware level,including collector-emitter voltage and collector current monitoring module,IGBT overheating fault prediction algorithm,and fault protection module based on the grid Miller platform.Basically realize the comprehensive experimental platform of IGBT overheating fault monitoring,prediction and protection and conducted multiple sets of experiments for different working conditions of the device,Moreover,collecting a data set containing multiple thermoelectric signals to prepare for the prediction of IGBT overheating faults in the next stage;(4)Combined with the hardware monitoring platform,the work content of the upper-level software in this thesis is mainly based on the optimization of the LSTM-based overheating fault prediction algorithm.Use the experimental data provided by the underlying hardware to train and optimize the network model,and finally analyze and evaluate the performance of the system in many aspects,and the prediction accuracy of the algorithm can reach 95.1%;later,a comparative analysis experiment based on BP neural network is introduced to further confirm This algorithm has superiority in prediction accuracy and stability in overall performance.