Robust Parameter Estimation of an Electric Vehicle Lithium-Ion Battery Using Adaptive Forgetting Factor Recursive Least Squares

Abstract

In this paper, a battery model suitable for electric vehicle application is analyzed. Open circuit voltage is described by an adaptation of Nernst equation. Thevenin circuit is used to depict the instantaneous and transient regime. Hysteresis effect is outlined by a zero-state correction term. We propose a new algorithm AFFRLS (adaptive forgetting factor recursive least squares) to extract the parameter of the battery model, then to predict the output voltage, and compare it to the original FFRLS (forgetting factor recursive least squares). To evaluate these algorithms, we used experimental data conducted by CALCE Battery Research Group on the Samsung INR 18650-20R battery cell. We fed the data to the algorithms and compared the estimated output voltage for two dynamic tests on MATLAB. Results show that AFFRLS has low distribution in high error range up to 4% less than FFRLS, this means that AFFRLS has a better parameter identification than FFRLS.