BCRLS-EKF-Based Parameter Identification and State-of-Charge Estimation Approach of Lithium-Ion Polymer Battery in Electric Vehicles

摘要

It is very importance for the BMS (battery management system) in electric vehicles to estimate the SOC (state of charge) of LiB (lithium-ion battery) accurately. In this paper a 3.7V/2.4Ah LiB is chosen as the research object. This paper firstly established the Thevenin battery model, and the parameters of which are determined by off-line identification method; To address the interference caused by colored noise on recursive least squares (RLS) method in on-line parameter identification, the bias compensation recursive least squares with forgetting factor (BCRLS) method is used to on-line identify the parameters of the battery, which can effectively reduce the interference of the noise on the estimation results; Finally, the extended Kalman filter (EKF) method is used to estimate the SOC. In order to improve the accuracy and efficiency of parameter identification and state estimation, BCRLS and EKF are combined to realize joint estimation of model parameters and states. In the identification process, BCRLS provides new parameters for EKF estimation, and EKF provides a relatively accurate OCV value for BCRLS. In addition, EKF estimates that the state variable U_p can be used as a reference to the BCRLS output value, and the output U_rc of the BCRLS can also be used to correct the measured value of the next cycle of EKF. That is to say, the output values of the two algorithms can be compared with the measured values to improve the identification accuracy of the next cycle. Because StroPower BMS HIL test system can well simulate the measurement error and non-ferrous noise in the practical operation of BMS, we use the BMS test system of StroPower to conduct a hardware in-loop experiment to verify the accuracy of the BCRLS-EKF algorithm in actual use, And compare the result with the result of RLS-EKF algorithm to verify the validity of the algorithm. Experimental results show: In the DST condition, the RLS-EKF algorithm estimates the absolute error of SOC is 3%, while the BCRLS-EKF algorithm estimates the absolute error of SOC is not more than 1%. Compared with RLS-EKF algorithm, the accuracy of SOC estimation is improved obviously by using BCRLS-EKF algorithm. Therefore, the SOC estimated by BCRLS-EKF algorithm based on the Thevenin model can meet the requirements of the BMS of electric vehicle.