Research On Prediction Methods Of Remaining Useful Life Of Lithium-ion Battery Based On Filtering

Lithium-ion battery is widely used,and there is no doubt that evaluate its health state becomes a research hotspot.But lithium-ion battery has complex electrochemical properties and the capacity will degrade gradually with the charge and discharge process.When the capacity is reduced to the failure threshold(usually 70%~80% of the rated capacity),it is considered that the life of lithium-ion battery reaches the end of life.Therefore,predicting the remaining useful life of battery is beneficial to improve system reliability,to prevent accident,and has vital application value.Predicting the remaining useful life of lithium-ion battery has become an important research content in the field of battery system health assessment.In recent years,there are two main types of lithium-ion battery remaining useful life prediction methods: empirical prediction and performance prediction.Filtering technology is a performance prediction method.In predicting battery remaining useful life,extended kalman filter and particle filter are commonly used methods.They can start from the data point of view,to obtain the change rules between test data and time,or to obtain recursive relations of system state.This kind of mathematical model is relatively easy to build,it has a wider scope of application.At the same time,kalman filter has good convergence and no standard requirement when the initial value setting.The initial value is usually selected according to the experience,and the estimated results are ideal.Particle filtering use the monte carlo to solve bayesian estimation.Combined with experience degradation model in predicting remaining useful life of lithium-ion battery and describe probability by particle set.Therefore,the adaptive ability is good in nonlinear non-gaussian degradation process and it is applicable to any form of state space model.In view of the above situation,this paper carried out the research on prediction methods of remaining useful life of lithium-ion battery based on filtering.First of all,this paper analyzes the lithium-ion battery life degradation process based on the degradation test data.We select the performance model prediction method and use extended Kalman filter to predict remaining battery useful life,then analyze the prediction results of experiment.Secondly,in view of the extended kalman filter is suitable for weak nonlinear model and poor adaptability for lithium-ion battery with complex electrochemical properties,the standard particle filter based on bayesian estimation is used to predict remaining useful life.And compare the results with the extended kalman filter.Finally,for the problem of particle degradation in standard particle filter resampling,we make the life prediction experiments of polynomial resampling and stratified resampling respectively.Based on this two mathod s,we use the improved resampling technique to predict remaining useful life,then to analyze the prediction results of the three resampling algorithms,and to compare them with the standard particle filter.This article using public battery test data that from the N ASA PCo E research center to complete the related experiments,comparing the predicted results between extended kalman filter and standard particle filter,and then compare the predicted results of the later with improved resampling particle filter.Through the error analysis for many times,so as to verify the improved resampling particle filter can effectively predict remaining cycle useful life of lithium-ion battery.To verify the prediction performance is more suitable and prediction precision is more accurate.