Research On IGBT Life Prediction Method Based On Improved Particle Filter Algorithm

In power electronic systems,the failure of IGBTs can lead to equipment damage or failure,which may lead to equipment downtime,resulting in production and economic losses.Evaluate and predict the failure mechanism and life of IGBT to formlate more scientific operation and maintenance strategies,which can reduce equipment maintenance costs and failure rates,and improve equipment reliability and operating efficiency.Therefore,aiming at the life prediction problem of IGBT,this paper improves the particle filter algorithm and proposes an IGBT life prediction method based on hunter-prey optimized particle filter(HPO-PF).Firstly,based on the analysis of the failure mechanism of IGBT,this paper selects the saturation voltage drop as the input parameter of the prediction model from the consideration of characterization factors,aging trend and online measurement feasibility.Taking the aging data of IGBT published by NASA as a dataset,the saturation pressure drop degradation model was obtained,the model parameters were determined,and the remaining life of IGBT was predicted by the elementary particle filter algorithm,and the relative error accuracy was about 80%after simulation analysis.Secondly,the remaining lifetime of IGBT is predicted by using the unscented Kalman improved particle filter algorithm,and the simulation shows that the relative accuracy of the unscented particle filter is about 93%.Aiming at the problems of low accuracy of particle filter algorithm and long iteration time of unscented particle filter algorithm,this paper uses HPO to improve the particle filter algorithm,drive particles closer to the real state and optimize particle weight,and after simulation analysis,the relative error accuracy can reach about 95%,and the iteration speed is increased by about 1/3 compared with the unscented particle filter algorithm.Finally,using the accelerated aging experimental platform,Vce(on)in the degradation process is measured under constant IC,and a data set is established to determine the prediction model parameters.A verification system was designed to verify the HPO-PF-based IGBT life prediction method.The results show that the module fails in the actual 347 cycle group,and the avearge prediction error group of the prediction statistics is 5 groups,and the minimum error is only 2 groups.Experimental results show that the proposed method has high precision,does not require too much sample data,and can solve the life prediction problem of IGBT,which provides a solution for the life prediction of multiple IGBT modules of different models in the system.