A physics-guided data-driven framework for joint SOC and multi-point temperature estimation of lithium-ion batteries

With the widespread use of lithium-ion batteries in electric vehicles and energy storage systems, accurate and robust joint estimation of battery temperature and state of charge (SOC) is essential to prevent temperature-induced degradation and thermal runaway. However, achieving this remains challenging due to the limitations of isolated physics-based or data-driven methods. In this study, a physics-guided data-driven framework is proposed for joint SOC and multi-point temperature estimation. First, a diffusion-enhanced electrical equivalent circuit model is coupled with a lumped thermal network model comprising three heat sources and six states, and an extended Kalman filter is employed for physics-based SOC and temperature estimation. Simultaneously, a data-driven gated recurrent unit network is constructed for end-to-end SOC and temperature estimation. Building on this, a physics-guided framework is developed by embedding physics-based prior estimates into the input layer of the data-driven model, while a result-level fusion framework is additionally introduced as a comparative benchmark. The four methods are systematically evaluated over a wide temperature range and under diverse dynamic driving cycles, together with four robustness test scenarios. The results demonstrate that the proposed physics-guided framework significantly outperforms all baseline methods in both estimation accuracy and robustness. For instance, at an ambient temperature of 25 °C, the mean absolute error of the five estimated temperature states is as low as 0.12 °C and 0.10 °C under the UDDS and WLTC cycles, respectively. This study offers a promising pathway toward safe and reliable battery management systems.