Deep-Discharging Li-Ion Battery State of Charge Estimation Using a Partial Adaptive Forgetting Factors Least Square Method

Abstract

State of charge (SOC) estimation of deep-discharging Li-ion batteries under complicated working conditions at different temperatures is still challenging. Nowadays, the depth of discharge (DOD) of batteries in electric vehicles (EVs) is generally low, resulting in the insufficient use of battery energy. This paper proposes a SOC estimation method using a novel partial adaptive forgetting factors recursive least square (PAFFRLS), which adjusts the forgetting factors based on the own physical properties of each parameter in equivalent circuit models (ECMs) to accommodate to greatly changing under deep-discharging range and high dynamic working conditions. The gain matrix in the proposed method is split to update independently according to each parameter, which solves the issue of mutual influence between parameters vary with different rates. In addition, four typical test profiles, including DST, UDDS, US06, and EUDC, are employed to simulate different working conditions of EVs. Eventually, numerous simulations and experiments results at different temperatures are employed to verify the validity of the proposed method. All average errors of the SOC estimation under four different kinds of working conditions are less than 1.3% as well as all peak errors are less than 5%. All peak errors are less than 3% while DOD is larger than 90%, which illustrates the effectiveness of the proposed method in the case of deep-discharging and provides better guidance to the design of battery management system (BMS) in EVs.