Quantitative validation of calendar aging models for lithium-ion batteries

摘要

Calendar aging prediction is a key technique to develop durable and robust electric vehicles. Automotive grade pouch cells based on LiNi1/3Mn1/3Co1/3O2and graphite are tested in an extensive accelerated calendar aging matrix and analyzed for capacity loss and evolved gas volume. This study derives an extended semi-empirical calendar aging model considering an initial solid electrolyte interface layer grown during the formation process. The extent of the thus lost active lithium is derived by open circuit voltage curve fitting as well as by inductive coupled plasma experiments. For this analysis the LiNi1/3Mn1/3Co1/3O2first cycle inefficiency is considered. Additionally, a validation technique based on the split of aging data into training and validation data is introduced with which it is possible to quantify the predictive capability of aging models. Using this technique the developed calendar aging model of this study is compared with competing aging models in literature. The derived global aging model is quantitatively shown to exceed other models in terms of their predictive ability, especially when little data is provided to the model.