Lithium-ion battery degradation mode quantification for 2nd life using half-cell data and machine learning techniques

The reactions inside lithium-ion batteries make it complex to quantify the aging behavior, especially those to be used in second-life applications. The hidden mechanisms can suddenly overturn the observable aging mechanism, generating a sudden decrease in the battery performance rate. In this context, this study presents a machine learning model that observes the hidden aging mechanisms of lithium-ion batteries. To do so, anode- and cathode-level data have been obtained to artificially generate several aging paths defined by the loss of lithium inventory, loss of active material, and incremental capacity curves. Once the artificial data were obtained, an artificial neural network was built to correlate the aging modes and incremental capacity curve. As a result, a model to quantify the loss of lithium inventory and active material was obtained. This model has been applied to the sorting activity of Nickel Manganese Cobalt (NMC) and Lithium Iron Phosphate (LFP) batteries for second-life applications. The results show that the developed aging mechanism quantification model reaches errors below 1.5 % for NMC batteries.