Research On Prediction Of Remaining Service Life Of Power Batteries Based On Particle Filterin

As the main component of battery electric vehicles,lithium-ion power battery has attracted much attention from the public for its reliability and safety.Since the life decay of the vehicle power battery is unavoidable,in order to ensure the safe and stable operation of the vehicle,the power battery needs to be replaced when it decays to a certain life threshold,so as to avoid accidents or economic losses,or even casualties caused by the failure of the power battery.The life degradation trend of lithium-ion power batteries has strong nonlinear and non-Gaussian characteristics,and its remaining useful life(RUL)cannot be directly obtained,so accurate RUL prediction is required.This paper analyzes the structural principle and life decline mechanism of the Lithiumion power battery,and evaluates the remaining service life of the power battery with the remaining capacity.Based on the built battery test experimental platform,the life acceleration experiment was carried out on the vehicle power battery model INR21700M50 T,and the battery capacity degradation data was obtained.According to the trend of the data,two empirical models were selected to fit it.Based on the comparative analysis of the fitting results,the double-exponential model was selected as the battery capacity degradation model,and the model parameters were obtained through data fitting.The particle filter(PF)algorithm with strong nonlinear non-Gaussian system processing ability is studied,and the problems of poor selection of proposed density distribution,particle degradation and loss of particle diversity in PF algorithm are improved.Unscented Kalman Filtering(UKF)is used to optimize the selection of the proposed density distribution,and Improved Residual Resampling(IRR)and Gaussian Particle Swarm Optimization(GPSO)algorithms are introduced to alleviate the problem of particle degradation and loss of particle diversity.Improved Residual Resampling Unscented Particle Filter(IRR-UPF)algorithm and Gaussian Particle Swarm Optimization based Unscented Particle Filter(GPSO-UPF)algorithm.Based on the PF algorithm and its improved algorithm,the battery RUL prediction simulation is carried out,and the prediction results are compared with two sets of battery capacity data.And use the absolute error(AE)and relative error(RE)to evaluate the prediction accuracy.It is concluded that the improved particle filter algorithm has higher prediction stability while improving the prediction accuracy.And it has relatively stable and accurate RUL prediction results,as well as filter tracking with low error for different batteries,indicating that it has a certain general applicability.By improving the standard PF algorithm,this paper improves the prediction accuracy and prediction stability of the PF algorithm for the RUL of the vehicle power battery,and uses the experimental data to verify it.