Variable recursive least square algorithm for online battery equivalent circuit model parameters identification for electric vehicles

In state-of-charge (SOC) estimation approaches which rely on electric circuit models, the accuracy of the model's parameters is influenced by factors such as battery aging and temperature, leading to SOC estimation errors. To tackle this issue effectively, a constant update of battery parameters is proposed. Our novel approach introduces the variable recursive least squares (VRLSs) algorithm, specifically designed to upgrade the parameters of a 2-capacitor-resistor network and accurately evaluate the terminal voltage of the battery. To assess the effectiveness of the VRLS algorithm, we conducted a comparison with two other methods: recursive least squares (RLS) and adaptive forgetting factor RLS (AFFRLS) algorithms. We employed experimental data from the CALCE Battery Research Group, with the Samsung and the A123 lithium-ion cells for testing. The results of the tests revealed that VRLS outperformed both AFFRLS and RLS methods. Notably, VRLS demonstrated significantly lower distribution in high error range and exhibited superior predictive performance indicators, including root-mean-square error, mean absolute error, and mean absolute percentage error across all tests.