Li-ion Battery State of Charge Estimation of Multi-type Working Conditions Using a Segmented Multiple Independent Forgetting Factors Recursive Least Squares Method

Abstract

The accuracy and speed of online state of charge (SOC) estimation of Li-ion batteries is still challenging because of the time-varying parameters and the complexity of the working conditions of electric vehicles (EVs). In this paper, a segmented multiple independent forgetting factors recursive least squares (SMIFFRLS) is proposed for accurate and fast online SOC estimation in EVs under severe and steady conditions. Multiple independent forgetting factors are used to track time-varying parameters in equivalent circuit models (ECMs), which show good performance on finding the initial value of Segmented Ampere-Hour integral method online. Numerous experiments under multi-type working conditions (UDDS and UNpart2) indicate that the average estimation errors of both Li(NiCoMn)O 2 (NCM) and LiFePO 4 (LFP) batteries are less than 2% and 5%, respectively, which are lower than the ampere-hour integral method and Recursive Least Squares (RLS). In addition, the robustness is better than the ampere-hour integral method; and the average computation load is significantly reduced compared to RLS.