![Author ORCID: We display the ORCID iD icon alongside authors names on our website to acknowledge that the ORCiD has been authenticated when entered by the user. To view the users ORCiD record click the icon. [opens in a new tab]](https://chemrxiv.org/engage/assets/public/chemrxiv/images/logos/orcid.png)
Abstract
Previous studies aimed at predicting the lifetime of lithium-ion batteries often rely on cycling data characterized by charge and discharge rates that do not reflect the typical usage conditions of electric vehicles (EVs), or they focus on batteries that degrade slowly, limiting their applicability to scenarios where batteries degrade more rapidly. This study introduces a new cycling experiment dataset that mirrors realistic EV charge and discharge profiles and includes conditions that simulate rapid pressure increases due to various side reactions or mechanical impacts.We establish a criterion for identifying cells that experience abrupt capacity fades. Utilizing a two-dimensional convolutional neural network, we accurately classify cells prone to rapid degradation using only the voltage, current and temperature data from the first 3 cycles, achieving over 99% accuracy. Furthermore, we predict the state of health (SOH) of each classified cell at 270 cycles with a root mean square error (RMSE) of less than 2%.
