Adaptive electro-thermal digital twin for state-of-charge estimation of lithium-ion batteries

Abstract Accurate state-of-charge (SOC) estimation is essential for ensuring the safety, reliability, and energy efficiency of lithium-ion battery packs in electric vehicles. Conventional estimation methods often couple state and parameter estimation, resulting in error propagation and reduced robustness under dynamic operating conditions. To overcome these limitations, this paper proposes an adaptive electro-thermal digital twin (ADT) framework that continuously synchronizes a battery-specific electro-thermal model with the measured operating data. The synchronized digital representation performs online adaptation of internal resistance through recursive least squares (RLS) while providing physically consistent parameters to a dedicated nonlinear SOC observer. This decoupled architecture minimizes parameter-state interaction, enhances observability, and improves estimation stability during rapid load transients and temperature variations. The proposed framework is validated under UDDS, US06, and HWFET driving cycles over a wide temperature range from − 20 to 40 °C. Experimental results demonstrate consistent superiority over conventional Extended Kalman Filter (CEKF) and adaptive Extended Kalman Filter (AEKF) methods. Under the UDDS cycle at 25 °C, the proposed ADT achieves an SOC root mean square error (RMSE) of 0.425%, compared with 6.775% and 8.182% for the AEKF and CEKF, respectively. Even under the highly dynamic US06 cycle, the SOC RMSE remains below 1.81% across all investigated temperatures while maintaining significantly lower voltage tracking error and maximum SOC estimation error than the benchmark methods. These results demonstrate that continuous synchronization between the physical battery and its adaptive electro-thermal digital counterpart enables accurate, robust, and computationally efficient battery state estimation for next-generation battery management systems.