State Of Charge Estimation Of Lithium-Ion Batteries Based On Fusion Of Cubature Kalman Filters

Against the background of oil shortage and environmental pollution,green cars quickly became popular,and lithium-ion batteries have therefore been widely used.A battery management system(BMS)is essential to ensure the batteries’ safe,efficient,and long-life operation.As a key state of BMS,State of Charge(SOC)is critical for its steady operation,efficient energy management of vehicles,and accurate driving range prediction.Therefore,it is of great significance to study SOC estimation methods to improve its estimation accuracy.The accurate simulation of batteries’ dynamic characteristics is important for improving SOC estimation performance.However,an equivalent circuit model of batteries tends to have changing simulation accuracy of the battery’s dynamic characteristics during SOC estimation.Although the adaptive high-degree cubature Kalman filter(AHCKF)has a more accurate estimation of the dynamic characteristics simulated by the battery model than that of the adaptive cubature Kalman filter(ACKF),AHCKF may have lower estimation accuracy of the real dynamic characteristics of batteries.Therefore,AHCKF does not always outperform ACKF at each step during SOC estimation.In view of the above problem,this paper explored the changing characteristics of the accuracy difference between ACKF and AHCKF during SOC estimation.To improve the accuracy of SOC estimation,a method based on the absolute minimum error of SOC estimation and a probability-based method were designed to combine the estimation results of ACKF and AHCKF,respectively.After comparison,it is concluded that the SOC estimation method based on probabilistic fusion is more robust and can effectively improve the SOC estimation accuracy.Further,this paper explored the influence of the SOC estimation error sequences’ length and the filters’ initial weights on the probabilistic fusion-based SOC estimation method.The results show that it is more accurate to update the filter weights based on the error sequences in the adaptive noise window.Besides,this paper considered the filters’ initial weight optimization as a one-dimensional multi-objective optimization problem.Specifically,the evaluation function was constructed by the ideal point method after comparative analysis.Then,the golden section method and quadratic function interpolation method were combined to find the optimal solution.The combined approach can reduce the time required for optimization while ensuring convergence.The results show that the use of optimized initial weights for probabilistic fusion-based estimation can further decrease the SOC estimation error.