Data‐Driven Entropy Coefficient Estimation for Large Format Lithium‐Ion Batteries

Large format lithium‐ion cell operations are nonisothermal and require an additional entropy term to the open‐circuit potential expression in physics‐based electrochemical models to accurately predict and optimize cell designs for a given application. However, the experimental estimation of entropy coefficient via potentiometric and colorimetric methods is time‐consuming and expensive. A fast method based on parameter optimization of empirical expressions and data‐driven approaches is presented to accelerate the estimation of the entropy coefficient term. This involves the use of (dis) charge and hybrid pulse power characterization (HPPC) data from a 100 Ah prismatic cell comprising lithium iron phosphate and graphite electrodes to determine the entropy coefficient via least‐squares method. The estimated entropy coefficient is implemented in the Doyle–Fuller–Newman model to predict the voltage profiles at different temperatures and validated against experiments. Complex empirical functions and data‐driven estimated entropy from the HPPC data give the best model predictions, especially at low temperatures.