Fast-charging lithium-ion batteries are crucial for accelerating the adoption of electric vehicles (EVs) by reducing charging time and improving operational efficiency. However, fast charging presents a significant challenge due to the knee point in the battery degradation trajectory, beyond which capacity decreases rapidly. Optimizing fast-charging protocols that consider knee point capacity and cycle life is essential but requires extensive and costly cycling aging tests to obtain the necessary degradation labels. To address this challenge, a multiobjective Bayesian optimization framework is proposed for fast-charging protocol optimization, jointly considering knee point capacity and cycle life. To reduce experimental costs, two deep learning-based early prediction models are developed to predict knee point capacity and cycle life using data from the first 60 cycles. The framework employs a noisy expected hypervolume improvement acquisition function to handle prediction uncertainties during multiobjective optimization. Validation on publicly available battery datasets demonstrates that the proposed framework achieves effective optimization of fast-charging protocols while reducing experimental costs by approximately 90%.
Fast-Charging Protocols Design of Lithium-Ion Battery: A Multiple-Objective Bayesian Optimization Perspective
IEEE Transactions on Transportation Electrification ; 11 , 3 ; 8327-8338
01.06.2025
2741803 byte
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Multi-Objective Bayesian Optimization of Lithium-Ion Battery Cells
British Library Conference Proceedings | 2022
|Multi-Objective Bayesian Optimization of Lithium-Ion Battery Cells
SAE Technical Papers | 2022
|Multi-Objective Bayesian Optimization of Lithium-Ion Battery Cells
British Library Conference Proceedings | 2022
|