Developing an accurate correlation for estimating heat generation rate by an 18650 lithium iron phosphate cell through a machine learning model trained on experimental measurements

Accurate prediction of heat generation rate (HGR) in lithium-ion batteries is critical for ensuring thermal safety and performance in electric vehicles (EVs). This study presents a machine learning (ML)-based approach for estimating HGR of an LFP-cell using experimentally measured data from an isothermal battery calorimeter. Data spans a wide range of operating conditions—discharge rates from 0.5C to 6C and ambient temperatures from 10 °C to 60 °C. Key input features include C-rate, current, DOD, and temperature. Three ML models—Feedforward Neural Network (FNN), Long Short-Term Memory (LSTM), and Support Vector Machine (SVM)—are developed and compared. Among them, SVM demonstrates the highest accuracy, achieving a mean absolute error (MAE) of 0.0063, root mean square error (RMSE) of 0.0112, and R2 of 0.9994. Furthermore, a compact and accurate correlation for HGR prediction is developed using the best performing SVM model under diverse operating conditions. To enhance interpretability and reduce complexity, symbolic regression is applied using the AI-based PySR framework. The final correlation is expressed as a function of C-rate, current, DOD, and temperature. It reliably estimates HGR across 10 °C-60 °C and 0.5C–6C. The proposed correlation enables fast and reliable HGR estimation under different operating conditions, supporting real-time thermal management of EVs.