Improved adaptive state-of-charge estimation for batteries using a multi-model approach

Adaptive estimation of the state-of-charge (SoC) for batteries is increasingly appealing, thanks to its ability to accommodate uncertain or time-varying model parameters. We propose to improve the adaptive SoC estimation using multiple models in this study, developing a unique algorithm called MM-AdaSoC. Specifically, two submodels in state-space form are generated from a modified Nernst battery model. Both are shown to be locally observable with admissible inputs. The iterated extended Kalman filter (IEKF) is then applied to each submodel in parallel, estimating simultaneously the SoC variable and unknown parameters. The SoC estimates obtained from the two separately implemented IEKFs are fused to yield the final overall SoC estimates, which tend to have higher accuracy than those obtained from a single-model. Its effectiveness is demonstrated using simulation and experiments. The notion of multi-model estimation can be extended promisingly to the development of many other advanced battery management and control strategies.