Machine learning-based lifelong estimation of lithium plating potential: A path to health-aware fastest battery charging

Abstract

To enable a shift from fossil fuels to renewable and sustainable transport, batteries must allow fast charging and exhibit extended lifetimes—objectives that traditionally conflict. Current charging technologies often compromise one attribute for the other, leading to either inconvenience or diminished resource efficiency in battery-powered vehicles. For lithium-ion batteries, the way to meet both objectives is for the lithium plating potential at the anode surface to remain positive. In this study, we address this challenge by introducing a novel method that involves real-time monitoring and control of the plating potential in lithium-ion battery cells throughout their lifespan. Our experimental results on three-electrode cells reveal that our approach can enable batteries to charge at least 30% faster while almost doubling their lifetime. To facilitate the adoption of these findings in commercial applications, we propose a machine learning-based framework for lifelong plating potential estimation, utilizing readily available battery data from electric vehicles. The resulting model demonstrates high fidelity and robustness under diverse operating conditions, achieving a mean absolute error of merely 3.37 mV. This research outlines a practical methodology to prevent lithium plating and enable the fastest health-conscious battery charging.