Estimation of state-of-charge and capacity of used lithium-ion cells

Abstract

We describe an approach to estimate state-of-charge and faded capacity of cobalt-based lithium-ion cell based on timedomain analysis of a short-term transient. This approach requires a relatively short-duration test and is suitable for repurposing cells for less demanding applications. The successful estimation requires previous characterization of the cells for the given family because lithium ion chemistries differ significantly. Two algorithms were considered for estimation of unknown state-of-charge and capacity: Bayesian inference and boosted regression trees. The achieved accuracy was 95 % of capacity estimations; estimations were within ±2 % of the nominal cell capacity from the true value.