Equivalent Circuit Model-Based State of Charge Estimation of Lithium-Ion Batteries Using Kalman Filter Algorithms

Abstract: Estimation of accurate SoC of Li-ion batteries is the basic requirement for a safe and efficient operation of the Battery Management System (BMS) of EVs. A systematic method for SoC estimation using a second order ECM (2RC ECM) calibrated from the data of HPPC tests is proposed in this paper. The OCV-SoC relation is extracted from C/20 charge-discharge test and fitted by a seventh order polynomial function. Parameters of ECM passive components (R₀, R₁, C₁, R₂, C₂) are identified from ten discrete SoC values with Levenberg-Marquardt (LM) nonlinear least-square algorithm, then they are represented by seventh order polynomials with regard to the SoC. EKF and UKF, both recursive Bayesian estimators, are implemented and evaluated for Turnigy Graphene 4.6928 Ah lithium cell at 0C with the standard test profiles (C/20 charge-discharge and HPPC). Error estimation (RMSE, MAE and MAX) is used for SoC and terminal voltage estimation respectively. Experiments show that both filters can always converge at the conditions and the UKF is slightly better than EKF at the SoC estimation under most of drive profiles due to the avoid of Jacobian linearisation. The sensitivity of SoC estimation accuracy with regard to parameters' identification accuracy is discussed and future works are summarized as adaptive noise covariance setting and order elevation to 3RC model.

Keywords: State of Charge (SoC), Equivalent Circuit Model (ECM), Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), HPPC, Levenberg–Marquardt, Battery Management System (BMS), Electric Vehicle (EV).