Prediction Of Remaining Useful Life Of Lithium-ion Battery Based On Model And Data Fusion Method

As the main power supply energy of most equipment,the reliability and safety of lithium-ion battery affect the accurate realization of the system function goal,On the one hand,the prediction of its life in advance can improve the utilization rate of lithium-ion battery,on the other hand,it can avoid the sudden stagnation of equipment in operation or other phenomena,which also attracts many scholars’ attention on the prediction of the remaining useful life of lithium-ion battery.According to the current research results can be divided into three series of prediction methods,model-based method,data-based method and fusion method.With the continuous development of model-based and data-based methods,the fusion method combining the advantages of the two will become the mainstream prediction scheme.At the same time,the fusion method still has a large optimization space in the prediction application of RUL.Aiming at the prediction of RUL for lithium-ion batteries,this thesis firstly analyzes the application effect of nonlinear filtering algorithm,and shows the advantages and disadvantages of five nonlinear filtering algorithms through verification experiments.Then,the unscented Kalman filter algorithms is combined with BP neural network to predict the RUL of lithium-ion battery by error compensation,which effectively solves the problem of the unscented Kalman filter lack of measurement value in the update stage.In order to make full use of various model information of lithium-ion battery degradation process,an interacting multiple model(IMM)algorithm based on UKF method is proposed.By setting filter for each sub model,and updating model weight coefficient and model update probability in real time.After weighted calculation,combined with the observation value predicted by wavelet neural network,the optimal estimation of state parameters is realized,so as to further optimize the predictability of RUL.At the same time,the experimental results show that the IMMUKF-WAVENN algorithm can estimate the state of the prediction system more accurately than the UKF-WAVENN algorithm of its sub model,which improves the prediction accuracy effectively.