Enhancing fast-charging protocols with section-based Bayesian optimization for lithium-ion batteries to prevent Li-plating

Abstract

This study presents a model-based optimization framework for fast-charging protocols in lithium-ion batteries (LIBs), combining a physics-based electrochemical model with Bayesian optimization (BO). Two BO-based multi-step constant current (MCC) protocols, namely a single-section and a bi-section strategy, were developed and experimentally validated using a commercial 55.6 Ah pouch-type LIB cell under various conditions. By incorporating physics-informed safety constraints such as Li-plating potential, voltage, and temperature, the proposed BO-MCC protocols reduced charging time by up to 20 percent compared to the conventional constant current constant voltage (CCCV) method, while maintaining plating-free operation and thermal stability. In particular, the bi-section strategy further reduced charging time by up to 11 percent relative to the single-section approach, while effectively suppressing Li-plating and SEI growth. Furthermore, under a high-temperature condition with pre-heated cells at 60 °C, the BO-MCC protocol enabled charging from 0 % to 80 % state of charge within 629 s, thereby satisfying the USABC target for extreme fast charging. Finally, experimental cycling and post-mortem analyses confirmed that the BO-MCC protocols mitigate capacity degradation more effectively than the CCCV method. This work provides a practical and experimentally validated framework for designing efficient and safe fast-charging strategies for electric vehicle(EV) batteries operating under diverse thermal conditions.