Parameter identification and state of charge estimation for lithium-ion batteries based on trust region least squares

To address issues of inaccurate parameter identification and large state of charge (SOC) estimation error for lithium-ion batteries in electric vehicle (EV) applications, we proposed a parameter identification and corresponding SOC estimation method based on a trust region least squares (TRLS) algorithm. First, a second-order resistor-capacitor (RC) equivalent circuit model for the lithium-ion battery was established. To effectively describe the dynamic characteristics of the lithium-ion battery, discharge experiments were designed and conducted to collect the data of terminal voltage, current, and time. The initial parameters of this model were then obtained through a segmented identification based on the Levenberg-Marquardt (LM) algorithm. Subsequently, an improved parameter identification algorithm based on the TRLS was constructed. This algorithm conducted the iterative optimization of model parameters by searching for the best solution within the trust region, thereby achieving precise identification of model parameters. Finally, the SOC estimation was achieved using the TRLS-based optimization solution of model parameters in conjunction with the dual extended Kalman filter (DEKF) algorithm. With the proposed TRLS-based method, the Mean Absolute Error (MAE) values of the estimated SOC were 0.0031, 0.0066, 0.0052, and 0.0044 under the static open-circuit voltage (OCV), the dynamic Urban Dynamometer Driving Schedule (UDDS), Hybrid Pulse Power Characterization (HPPC), and Dynamic Stress Test (DST) conditions, respectively. The comparative results successfully verified the reliability and effectiveness of the proposed TRLS-based method, which demonstrated significant potential and advantages for battery management applications.