Data-Driven-Based Internal Temperature Estimation for Lithium-Ion Battery Under Variant State-of-Charge via Electrochemical Impedance Spectroscopy

Internal temperature estimation is critical to the safe operation of lithium-ion batteries (LIBs), and electrochemical impedance spectroscopy (EIS)-based methods have been demonstrated to be promising. However, accurate internal temperature estimation under variant state-of-charge (SoC) is still challenging due to the combined impact of temperature and SoC on the EIS. Accordingly, this work proposes a novel EIS-based internal temperature estimation approach, for which SoC-insensitive EIS features are quantitatively selected and utilized for temperature estimation using support vector regression (SVR) with unknown SoC. First, the EIS feature selection is performed to select SoC-insensitive features from the imaginary of impedance spectrum. Subsequently, an SVR-based framework and a well-trained SVR model are created to estimate the internal temperature of LIBs. The performance of the proposed model is validated by its lowest estimation error (0.57 degrees C) under known and unknown SoCs compared to that of the existing methodologies. The results confirmed that the proposed method holds the advantage in estimating the internal battery temperature with different SOCs.