A comparative study of battery state of charge estimation with equivalent circuit model and empirical model

Abstract

The equivalent circuit model and empirical model are commonly used for battery prognostics for electric vehicles in industry. This paper gives a comprehensive comparison study from system modeling to parameter estimation and state of charge (SOC) estimation between the two models. For the equivalent circuit model, an accurate parameter identification method is applied by deducing least square form with the pulse discharging data, and the SOC is estimated by designing a reduced-order observer. For the empirical model, the parameters are determined directly from dynamic testing data, and then a novel modified Luenberger observer is designed for SOC estimation for the nonlinear system, where the convergence is validated by designing a Lyapunov function. Finally, a comprehensive comparison analysis is demonstrated in model structure, modeling accuracy, SOC estimation accuracy, feasibility, initial bias situations and hyperparameters, which provide practical guide for industrial applications.