Research On The Prediction Of The Remaining Life Of Power Electronic Device IGBT

Insulated gate bipolar transistor(IGBT)has been widely used in new energy generation system and electric vehicle power conversion device due to its good performance.However,IGBT has high reliability and safety requirements.If it fails,it will bring great loss to the whole system.Therefore,it is very important to evaluate the health status of IGBT.If the accurate prediction of the remaining life of IGBT can be realized,the corresponding maintenance plan can be made according to the prediction results,and an early warning can be given before the failure of IGBT,so as to replace new devices,so as to avoid the occurrence of equipment failure.Therefore,on the basis of analyzing the failure mechanism of IGBT,the percentage of U_CE change of IGBT is used as the monitoring characteristic of IGBT failure,and then an optimal scale Gaussian process model based on the improved bird swarm optimization algorithm is proposed,which is used to predict the total cycle life of IGBT and calculate the remaining service life of IGBT.This paper mainly includes the following research contents:First of all,the basic mechanism and working characteristics of IGBT are introduced,and the aging failure mechanism and failure mode of IGBT are described.After comparing and analyzing several main failure characteristics and failure standards of IGBT,U_CE is selected to increase to 5%of the initial value as the failure criterion of IGBT,which provides a theoretical basis for predicting the remaining life of IGBT.Secondly,the accelerated aging test is introduced to obtain the time series data of three different types of U_CE change percentage.On this basis,wavelet denoising method is used to preprocess the data and evaluate the results of wavelet denoising.Thirdly,the method of phase space reconstruction of non-stationary time series is introduced and applied to U_CE change percentage data.On the basis of Gauss process model,the method of different scale representation in wavelet theory is introduced,and the optimal scale Gauss process(OSGP)model is established.Based on the improved bird swarm optimization algorithm,the optimal scale and optimal scale function required by OSGP model are found.Compared with the original BSA algorithm,the improved IBSA algorithm has improved the global search ability,stability and convergence speed.Finally,the IBSA-OSGP model is used to predict the remaining life of three different models,and the predicted results are compared with those of SVM,ELM,GP and MGP models.The root mean square error of the prediction results of the IBSA-OSGP model is the smallest of the five models,the difference between the predicted value and the actual value of the remaining life is relatively small,the prediction effect is basically not affected by the number of training samples,and the confidence interval of the prediction results is also the smallest The prediction accuracy,stability,adaptability and reliability of the IBSA-OSGP model is obviously better than the other four comparative models,which can effectively predict the remaining life of IGBT.