Accurately estimating internal temperature of lithium-ion batteries based on the distribution of relaxation time and data-driven

Accurately estimating the internal temperature of lithium-ion batteries (LIBs) represents a crucial means of minimizing battery thermal accidents. An effective way to estimate the battery's internal temperature is electrochemical impedance spectroscopy (EIS). Distribution of relaxation times (DRT) methodologies can achieve a rapid decomposition of EIS, but DRT is primarily used for battery analysis. In this work, a method for estimating the internal temperature of LIBs based on the combination of DRT and machine learning was proposed, and more than 1200 EIS spectra of commercial LIBs were collected at different states of charge and internal temperatures, which is very scarce in the field of battery internal temperature. DRT calculations were performed on the collected data and the features strongly correlated to internal temperature were confirmed. The mapping of input features to the internal temperature is established with six machine-learning algorithms. The results indicate that although the battery status is unknown, the models can accurately estimate the internal temperature, and the mean absolute error (MAE) is below 0.319 °C. A comparison to the same machine learning algorithms based on EIS shows that using DRT as input can reduce model outliers and improve model stability.