文献类型：期刊文献

英文题名：Transformer-based model predictive control with anode potential awareness for online fast charging optimization of lithium-ion batteries

作者：Zhuang, Zixian[1,3];Chen, Hongxu[1];Chen, Ying[1];Luan, Weiling[1];Chen, Haofeng[1,2];Ji, Xiaoyan[3]

机构：[1]East China Univ Sci & Technol, Sch Mech & Power Engn, Key Lab Adv Battery Syst & Safety CPCIF, Shanghai 200237, Peoples R China;[2]Univ Strathclyde, Dept Mech & Aerosp Engn, Glasgow G1 1XJ, Scotland;[3]Lulea Univ Technol, Div Energy Sci, Energy Engn, SE-97187 Lulea, Sweden

年份：2026

卷号：404

外文期刊名：APPLIED ENERGY

收录：;EI(收录号：20254819620729);WOS:【SCI-EXPANDED(收录号:WOS:001633110100001)】；

基金：The authors gratefully acknowledge the support from the National Natural Science Foundation of China (52375144 and 52205153) , Shanghai Leading Project under the Eastern Talent Program during the course of this work.

语种：英文

外文关键词：Lithium-ion battery; Model predictive control; Anode potential; Transformer; Charging control

摘要：Achieving efficient and safe charging while effectively mitigating degradation induced by lithium plating is crucial for fully unleashing the performance and extending the lifespan of lithium-ion batteries. This study proposes a fast charging optimization control method that integrates anode potential awareness with a Transformer-based model predictive control (MPC) framework to address the complex multi-physics coupling constraints during fast charging. To enable multi-step prediction of the anode potential, 30 candidate features are initially constructed based on measurable parameters and their temporal derivatives. A robust feature set is then established by selecting 5 most discriminative input variables through correlation analysis and the Null Importance method. Subsequently, a Transformer-based state predictor is developed to perform accurate joint prediction of voltage, temperature, and anode potential under typical conditions, including dynamic loads and constant current charging. The root mean square errors (RMSE) for voltage, temperature, and anode potential predictions are 11.62 mV, 0.251 degrees C, and 3.67 mV, respectively. Building on the prediction model, an MPC framework is further developed using the particle swarm optimization (PSO) algorithm. This framework enables real-time optimization of the charging current trajectory under multi-dimensional safety constraints, including voltage upper limit, temperature upper limit, and anode potential lower limit. Results demonstrate that the proposed method can achieve a closed-loop integration of accurate state prediction and optimal control during charging, effectively suppressing constraint violations of key variables while balancing charging efficiency, cycle life, and safety. The method exhibits strong engineering adaptability and promising scalability for practical applications.