Constant Voltage Charging Curve Based Features and State of Health Estimation of Li-ion Batteries: A Comprehensive Study

The State of Health (SOH) of Li-ion batteries is usually estimated from the charging or discharging curves. While the discharging curves of EV batteries are highly dynamic in nature, the Constant Voltage Constant Current (CCCV) protocol is the most common charging method. This makes the charging curve based SOH estimation of EV batteries a preferred choice over the discharging curves. CCCV protocol can have single or multiple CC sections and can start from different State of Charge level of batteries. On the other hand, CV charging section is reported to be stable and is complete enough for estimating SOH, especially, in case of Level-I charging. However, the CV charging based SOH estimation is underexplored. More importantly, there are multiple health indicating features reported in the literature and many more can be derived, but details discussion on their impact on the SOH estimation are underexplored. In this paper, we focus on providing the correlations between the various health indicating features and their potential in estimating SOH in EV batteries. We have used a total fifteen features and the most popular datasets of NASA and MIT for the study. The study, based on Pearson coefficient of correlation, shows that the time to reach a particular current (in this case 0.5Amp) is the most dominant health indicating feature in both the dataset. Further assessment using machine learning model shows that SOH estimation using this feature, ‘time to reach 0.5 Amp’ provides highest accuracy or minimum Root Mean Square Error compared to using other features. This paper sets a clear guideline for future researchers in selecting the appropriate CV features for SOH estimation of Li-ion batteries in EVs.