Cross-capacity internal temperature estimation in lithium-ion batteries using multiple impedance features from the negative electrode

The internal temperature (IT) of lithium-ion batteries (LIBs) is a critical parameter influencing their operational safety and reliability. Electrochemical impedance spectroscopy (EIS), as a non-destructive diagnostic method, shows a strong correlation with IT. However, conventional EIS measurements at a single frequency point are often affected by variations in the state of charge (SOC) and battery capacity, which limits their accuracy in IT estimation. To address this issue, a negative electrode (Neg) EIS acquisition method based on a reference electrode is proposed. The performance of raw and analyzed EIS features in IT estimation is systematically compared. A gradient boosting decision tree model is constructed, with EIS data from 1.7 Ah small-capacity battery used for training and data from 3.4 Ah large-capacity battery used for test, enabling generalized modeling across different capacities. The results demonstrate that IT estimation based on Neg EIS significantly outperforms that based on battery EIS. Notably, the prediction error of the solid electrolyte interphase (SEI) impedance features reduces by nearly 3 °C, and the minimum estimation error of the Neg impedance phase angle reaches 1.33 °C. This improvement is attributed to the selected feature frequencies falling within the SEI-responsive region of the Neg, which ensures high temperature sensitivity while effectively mitigating SOC interference. Furthermore, the optimized impedance phase angle feature is obtainable from a single frequency point (15.8 Hz), providing a feasible, efficient, and low-cost solution for real-time online IT monitoring in LIBs.