Empirical OCV Model Benchmarking for ECM Identification in LFP Battery Management Systems

Accurate battery models are essential for the safe and efficient management of lithium-ion cells in electric ve- hicles. Equivalent circuit models (ECMs) are widely used in embedded battery management systems, and the accuracy of these models depends heavily on the open-circuit voltage (OCV) model used as the voltage baseline. This paper presents an empirical benchmarking of OCV model families for ECM iden- tification, using a fast-charge dataset acquired experimentally at the iCube laboratory on a Lithium Werks APR18650M1B LFP cell. Six OCV models from three families are evaluated in combination with 1RC and 2RC ECM structures: 4th and 8th order polynomial, 2- and 3-term logistic sigmoid, and Nernst- Log and Plett Enhanced thermodynamic. A charge/discharge voltage averaging method is employed to extract the OCV from continuous cycling data without the need for long rest phases. This method is applicable to EV operating profiles where hybrid pulse power characterisation (HPPC) is unavailable. The results show that, for the cell and identification protocol considered here, an empirical OCV fit RMSE threshold of around 28 mV separates two 2RC identification regimes. Below this value, the slow time constant remains physically interpretable, whereas above it, the second RC branch reaches an identification bound and primarily absorbs OCV bias. Of the models evaluated, the 3-Sigmoid/2RC configuration achieves the lowest validation voltage RMSE of 134.9 mV, while the Nernst-Log model offers the optimal accuracy-complexity balance with just five OCV parameters. These results provide practical guidance for selecting an OCV model in an ECM-based system for electric vehicles. Index Terms—Lithium-ion battery, LFP, electric vehicle, bat- tery management system, equivalent circuit model, open-circuit voltage, parameter identification, 1RC, 2RC.