Online state-of-charge estimation for lithium-ion batteries via a high-degree-of-freedom robust observer with model parameter identification

Lithium-ion battery state-of-charge (SOC) estimation faces persistent challenges, such as model errors, inaccurate parameter identification, and measurement noise. To address these limitations, this study proposes an innovative SOC estimation strategy. First, a recursive least squares (RLS) algorithm online estimates model parameters for a combined second-order resistor–capacitor (RC) model. Second, the battery dynamics are partitioned into three linearized submodels based on SOC ranges (0%–20%, 20%–30%, 30%–100%) to reduce nonlinear effects. Subsequently, a novel-structure high-degree-of-freedom robust observer (HDFRO) is developed for SOC estimation. Compared to traditional robust observers, the HDFRO provides greater design flexibility, with its convergence rigorously proven via Lyapunov stability theory. Moreover, linear matrix inequalities (LMIs) are introduced and transformed into a constrained optimization problem to obtain observer parameter matrices and gain matrices for the proposed HDFRO. Finally, the strategy’s efficacy is validated under Dynamic Stress Test (DST) and Urban Dynamometer Driving Schedule (UDDS) conditions using NCR18650 cells. Experimental results demonstrate a mean absolute error (MAE) below 1.4% and a root mean square error (RMSE) under 1.62%.