Online parametric adaptive state of charge estimation for lithium batteries

State-of-charge (SOC) estimation methods based on battery model rely heavily on the accuracy of model parameters. And these parameters could vary with environment and the types of batteries. Online battery modeling methods can improve the robustness of SOC estimation algorithms through updating model constantly with real-time data. These methods have far more profound significance on algorithm adaptability rather than the accuracy improvement of single parameter on some particular conditions. After all, we cannot acquire accurate referential values of these parameters, since they vary with many factors and some of them are not measurable at all. In this paper, damped recursive least squares with extended Kalman filter (DRLS-EKF) was adopted to identify model parameters and estimate SOC online. Robustness of the proposed algorithm was tested in harsh conditions that could occur in real applications. Besides, the performances of offline model identification and the online method we adopt were compared and analyzed. Last but not least, experiments on NCM and LiFePO4 cells and cells with different aging conditions were designed, which aim at verifying algorithm's adaptability on different battery types and slowly time-varying model parameters, respectively.