Lithium Battery State Estimation and Life Prediction Based on Multi-Model Fusion

An approach to estimating the state of lithium battery and predicting its lifetime based on multi-model fusion is presented. The electrochemical behavior model is combined with data-driven model, and adaptive weighted PSO is introduced in order to improve prediction precision. Through multi-model weighted fusion, it can fully capture the multiple levels of battery state, which improves accurate prediction of battery aging process. This paper is experimentally verified on the MATLAB/Simulink simulation platform, using a lithium battery data set from actual tests. Experimental results show that the multi-model fusion method is better than traditional single-model method both in capacity forecast and residual life prediction. The MAE of capacity prediction is O.008Ah and the RMSE is O.012Ah, which is about 15% lower than the traditional method; the RMSE of remaining life prediction is 22.5 hours, which is about 25% less than the traditional method. This method not only maintains a high prediction accuracy under various usage conditions, but also shows strong robustness and adaptability.