A Foundation Model for State of Health Prediction of Lithium-ion Battery in Electric Vehicles

Batteries are pivotal in electric vehicles (EVs), serving as the primary source of power. To ensure the safe and efficient operation of EVs, it is essential to accurately predict the state of health (SOH) of the battery, typically achieved through a battery management system (BMS). However, the complex coupling reactions and nonlinear degradation processes inherent in lithium-ion batteries (LIBs) present significant challenges in SOH prediction. Current data-driven models often require extensive datasets and prolonged training periods, while also exhibiting limited generalization capabilities. To address these challenges, this paper proposes a novel method based on a foundation model which named Lag-Llama for predicting SOH and other critical battery characteristics synchronously, such as temperature, internal resistance. Remarkably, under zero-sample conditions, our approach achieves a Continuous Ranked Probability Score (CRPS) of 0.0079, demonstrating robust zero-shot generalization capabilities. Furthermore, the model’s predictive performance is significantly enhanced following fine-tuning.