UC Davis

Lithium-ion batteries are widely used in the Electric Vehicle (EV) and the Energy Storage Systems (ESS) applications. In such systems, the batteries are connected in combinations of series and parallel connections in such a way that the overall current and voltage of the pack meets the performance requirements of the application. However, due to the inherent degradation mechanisms and inconsistencies in operating conditions for cells within a series connected battery pack, there is a differential in the State-of-Health (SoH) of cells which subsequently affects the utilizable energy of the overall pack. Thus, it is imperative to have models which rely on short duration tests to rapidly estimate the SoH of the battery packs. In this project, two of the most widely used Li ion battery types – Nickel-Manganese-Cobalt (NMC) and Lithium-Iron-Phosphate (LFP) chemistry cells are evaluated.An empirical model is proposed which relies on two quick 10 second pulses to estimate the SoH of a battery pack consisting of 4 series connected cells. The empirical models are trained with a variety of series connected battery packs constructed using combinations of differently aged cells. The model yields excellent results for the NMC chemistry battery packs with Root-Mean-Squared Error (RMSE) of 2.05 Total SoH % points and Mean Absolute Error (MAE) of 1.60 Total SoH % points. However, the LFP chemistry of cells only achieve an RMSE of 6.09 Total SoH % points and MAE of 4.81 Total SoH % points. The hypothesis is that this performance differential exists due to the much steeper Open Circuit Voltage – State of Charge (OCV-SoC) curve for NMC cells over LFP cells. Further tests expanding the range of SoH of packs is recommended along with additional models to also estimate information about the distribution of SoH of the constituent cells are suggested as part of future work.