Comparative study of methods for integrated model identification and state of charge estimation of lithium-ion battery

Abstract

Model-based observers appeal to both research and industry utilization due to the high accuracy and robustness. To further improve the robustness to dynamic work conditions and battery ageing, the online model identification is integrated to the state estimation, giving rise to the co-estimation methods. This paper systematically compares three types of co-estimation methods for the online state of charge of lithium-ion battery. This first method is dual extended Kalman filter which uses two parallel filters for co-estimation. The second method is a typical data-model fusion method which uses recursive least squares for model identification and extended Kalman filter for state estimation. Meanwhile, a noise compensating method based on recursive total least squares and Rayleigh quotient minimization is exploited for online model identification, which is further designed in conjunction with the extended Kalman filter to estimate the state of charge. Simulation and experimental studies are carried out to compare the performances of three methods in terms of the accuracy, convergence property, and noise immunity. The computing cost and tuning effort are further discussed to give insights to the application prospective of different methods.