Explainable Electrochemical Impedance Spectroscopy Features for Multi-Temperature State of Charge Estimation of Lithium Iron Phosphate Batteries

Accurate estimation of state of charge (SOC) for lithium iron phosphate (LFP) batteries remains a challenge due to the poor observation into the internal states through measurable electric signals. To overcome this challenge, the electrochemical impedance spectroscopy (EIS), which is a potentially accessible signal onboard in the near future, is used to indicate the SOC of LFP batteries. EIS tests on batteries are conducted under different temperatures (covering 10 °C–40 °C) and dynamic conditions, with Shapley Additive exPlanations analysis used to evaluate the contribution of real and imaginary EIS features across full frequency range to battery SOC. The analysis identifies three dominant imaginary impedance features that are highly related to battery SOC. A Gaussian Process Regression model is then developed to describe the relationships between these features and SOC. Validation experiments are conducted on 24 cells under various temperatures and dynamic conditions. Results demonstrate that the proposed method achieves an average root-mean-square-error (RMSE) of SOC estimation being 2.6% at 25 °C over 15 repeated runs. At different temperatures within 10 °C–40 °C, the maximum RMSE and absolute error are 2.4% and 5.4%, respectively. Under dynamic conditions, the maximum RMSE is 2.3%, and except for a few SOC points, all estimation errors are below 3%.