Fusing coupled degradation mechanisms with machine learning: A multi-fidelity framework for lithium-ion battery lifespan prediction

可解释性 电池(电) 降级(电信) 计算机科学 可靠性工程 非线性系统 降维 点(几何) 容量损失 还原(数学) 可靠性(半导体) 人工智能 线性化 线性回归 机器学习 偏最小二乘回归 操作点 状态监测 工程类
作者
Lu Dong,Fei Ren,Changlong Li,Haoyu Ming,Naxin Cui
出处
期刊: [Elsevier BV]
卷期号:5 (4): 100367-100367 被引量:15
标识
DOI:10.1016/j.geits.2025.100367
摘要

Accurate prediction of battery remaining useful life (RUL) is critical for safety enhancement and service extension. However, conventional model-based methods suffer from computational complexity, while data-driven methods struggle with inadequate mechanistic interpretability and prohibitive data requirements. Moreover, RUL prediction in real-world applications often faces the challenge of data sparsity, where only limited early-cycle measurements are available due to time and cost constraints. To address these limitations, we propose a multi-fidelity framework that synergistically integrates electrochemical degradation mechanisms with sparse early-life data, achieving precise RUL predictions through physics-informed interpretability and reduced experimental data. First, we develop an enhanced pseudo two-dimensional (P2D) model that integrates solid electrolyte interphase (SEI) growth mechanisms with partially reversible lithium plating dynamics. This advancement captures nonlinear capacity degradation patterns frequently overlooked in existing models. Second, through causative-process analysis of battery aging trajectories, three governing parameters—SEI solvent diffusivity, lithium plating kinetic rate constant, and dead Li decay constant are identified. Quantitative correlations between molecular-scale processes and macroscopic degradation patterns are established. Crucially, a multi-fidelity regression architecture is designed to simultaneously predict knee point and RUL by adaptively fusing sparse capacity measurements with physics-based degradation trajectories. Validated across 169 commercial cells under different fast-charging protocols, the proposed framework achieves satisfactory accuracy. It maintains Mean Absolute Percentage Errors (MAPE) of just 5.7% for the prediction of knee point and RUL. These results can serve as valuable metrics for performance evaluation and safety improvement of battery systems. • A modified battery model considering coupled degradation mechanisms is developed. • The effect of mechanistic parameters on capacity decay trajectory is investigated. • A physic-informed multi-fidelity method for battery knee point and RUL prediction. • The proposed method realizes accurate RUL prediction with sparse aging data.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
袋袋发布了新的文献求助10
刚刚
1秒前
NY发布了新的文献求助10
1秒前
lyj发布了新的文献求助10
1秒前
所所的应助被qd采纳,获得10
2秒前
3秒前
科研通AI6.3的应助被lmy采纳,获得10
3秒前
zhudaxia完成签到,获得积分10
5秒前
5秒前
5秒前
花痴的手套完成签到 ,获得积分10
5秒前
麦粑块发布了新的文献求助10
6秒前
小牛牛发布了新的文献求助10
6秒前
6秒前
8秒前
INT发布了新的文献求助10
9秒前
hyy发布了新的文献求助30
9秒前
研友_VZG7GZ的应助被李瑞程采纳,获得10
9秒前
小蘑菇的应助被醉熏的寄松采纳,获得10
10秒前
我是谁的应助被醉熏的寄松采纳,获得50
10秒前
zhudaxia发布了新的文献求助30
10秒前
领导范儿的应助被醉熏的寄松采纳,获得10
10秒前
慕青的应助被醉熏的寄松采纳,获得10
10秒前
无花果的应助被醉熏的寄松采纳,获得10
10秒前
Lucas的应助被醉熏的寄松采纳,获得10
10秒前
国服躺赢完成签到,获得积分10
10秒前
搜集达人的应助被醉熏的寄松采纳,获得10
11秒前
11秒前
桐桐的应助被醉熏的寄松采纳,获得30
11秒前
11秒前
11秒前
陈富贵完成签到,获得积分10
11秒前
12秒前
12秒前
12秒前
growl发布了新的文献求助10
12秒前
任亮亮发布了新的文献求助10
13秒前
逆光完成签到,获得积分10
13秒前
gg完成签到,获得积分10
13秒前
领导范儿的应助被韩明佐采纳,获得10
15秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
八旗制度の研究 600
Improving the wear resistance of vehicle gear oil by oil soluble LaF3 nanoparticles and typical additives of gear oil 500
Boundaries and Bodhisattvas 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8183263
求助须知:如何正确求助?哪些是违规求助? 9572532
关于积分的说明 21826339
捐赠科研通 7780187
什么是DOI,文献DOI怎么找? 3433993
关于科研通互助平台的介绍 2558131
邀请新用户注册赠送积分活动 2480239