基于多尺度特征融合的锂离子电池早期剩余使用寿命预测

doi:10.19799/j.cnki.2095-4239.2025.0488

摘要/Abstract

摘要：

锂离子电池作为关键储能器件，在电动汽车等新能源领域具有广泛应用，其性能衰减与寿命预测对电池健康管理至关重要。早期预测不仅能够优化电池使用策略，更能为故障预警提供重要依据。针对锂离子电池早期数据质量差、退化特征微弱等挑战，本工作提出了一种基于多尺度特征融合的锂离子电池早期剩余使用寿命预测的新方法。首先，通过改进蝗虫优化算法(improved grasshopper optimization algorithm，IGOA)自适应调整时变滤波经验模态分解(time-varying filtered empirical mode decomposition，TVF-EMD)的关键参数，优化早期信号的分解过程，有效提取早期退化特征；其次，通过k均值聚类算法将本征模态函数(intrinsic mode function，IMF)划分为表征容量退化的高频、中低频、趋势项三类分量，经加权融合后显著增强早期特征表达能力；进而，构建基于鲸鱼迁徙优化算法(whale migrating algorithm，WMA)优化的预测模型，独立预测各频段IMF最终通过多尺度预测结果的融合重构，实现锂电池全寿命周期的退化轨迹预测与剩余使用寿命的预测。基于CALCE和MIT数据集的实验表明，本方法较传统预测模型具有显著优势，均方根误差(root mean square error，RMSE)始终低于1.4%，为锂离子电池早期寿命预测提供了可靠解决方案。

关键词: 锂离子电池, 早期剩余使用寿命, 时变滤波经验模态分解, 多尺度预测

Abstract:

Lithium-ion batteries, as key energy storage devices, are widely used in new energy fields such as electric vehicles, and their performance degradation and life prediction are crucial for battery health management. Early prediction can optimize battery usage strategies and provide a crucial basis for fault warning. To address challenges of poor early data quality and weak degradation characteristics, this study proposes a method for predicting the early remaining service life of lithium-ion batteries based on multiscale feature fusion. First, the decomposition of early signals is optimized by adaptively adjusting the key parameters of time-varying filtered empirical mode decomposition using an improved grasshopper optimization algorithm to effectively extract early degradation features. Second, intrinsic mode functions (IMFs) are divided into three components—high-frequency, medium-low-frequency, and trend terms representing capacity degradation—using the k-means clustering algorithm, which considerably enhances early feature expression ability after weighted fusion. Furthermore, a prediction model based on whale migration algorithm optimization is constructed to independently predict IMFs in each frequency band. Finally, through the fusion and reconstruction of multiscale prediction results, both the degradation trajectory and remaining service life prediction of the battery are achieved. Experiments based on the CALCE and MIT datasets demonstrate that this method outperforms traditional prediction models, with root mean square error consistently below 1.4%, providing a reliable solution for early-life prediction of lithium-ion batteries.

Key words: lithium-ion batteries, early remaining useful life, time-varying filtered empirical mode decomposition, multiscale prediction

中图分类号:

TM 912

方枷云, 贾建芳. 基于多尺度特征融合的锂离子电池早期剩余使用寿命预测[J]. 储能科学与技术, 2025, 14(11): 4346-4359.

Jiayun FANG, Jianfang JIA. Early remaining useful life prediction of lithium-ion batteries based on multiscale feature fusion[J]. Energy Storage Science and Technology, 2025, 14(11): 4346-4359.

图/表 22

图1

图2

表1

图3

图4

图5

图6

图7

图8

图9

表2

图10

图11

图12

表3

表4

图13

图14

图15

表5

图16

表6

参考文献 26

[1]	KHALEGHI S, HOSEN M S, KARIMI D, et al. Developing an online data-driven approach for prognostics and health management of lithium-ion batteries[J]. Applied Energy, 2022, 308: 118348. DOI: 10.1016/j.apenergy.2021.118348.
[2]	KIM S, CHOI Y Y, KIM K J, et al. Forecasting state-of-health of lithium-ion batteries using variational long short-term memory with transfer learning[J]. Journal of Energy Storage, 2021, 41: 102893. DOI: 10.1016/j.est.2021.102893.
[3]	XIA G S, JIA C Y, SHI Y H, et al. Remaining useful life prediction of lithium-ion batteries by considering trend filtering segmentation under fuzzy information granulation[J]. Energy, 2025, 318: 134810. DOI: 10.1016/j.energy.2025.134810.
[4]	KUMAR R, GOEL V. A study on thermal management system of lithium-ion batteries for electrical vehicles: A critical review[J]. Journal of Energy Storage, 2023, 71: 108025. DOI: 10.1016/j.est.2023.108025.
[5]	LAI X, QIAN L L, TANG X P, et al. Early-stage remaining useful life prediction for lithium-ion batteries based on geometric output construction[J]. Journal of Energy Storage, 2025, 114: 115792. DOI: 10.1016/j.est.2025.115792.
[6]	SUI X, HE S, VILSEN S B, et al. A review of non-probabilistic machine learning-based state of health estimation techniques for lithium-ion battery[J]. Applied Energy, 2021, 300: 117346. DOI: 10.1016/j.apenergy.2021.117346.
[7]	YANG M X, SUN X F, LIU R, et al. Predict the lifetime of lithium-ion batteries using early cycles: A review[J]. Applied Energy, 2024, 376: 124171. DOI: 10.1016/j.apenergy.2024.124171.
[8]	ELMAHALLAWY M, ELFOULY T, ALOUANI A, et al. A comprehensive review of lithium-ion batteries modeling, and state of health and remaining useful lifetime prediction[J]. IEEE Access, 2022, 10: 119040-119070.
[9]	MA G J, WANG Z D, LIU W B, et al. A two-stage integrated method for early prediction of remaining useful life of lithium-ion batteries[J]. Knowledge-Based Systems, 2023, 259: 110012. DOI: 10.1016/j.knosys.2022.110012.
[10]	ZHAO W J, DING W, ZHANG S J, et al. Enhancing lithium-ion battery lifespan early prediction using a multi-branch vision transformer model[J]. Energy, 2024, 302: 131816. DOI: 10.1016/j.energy.2024.131816.
[11]	翟健帆, 李波, 李永利, 等. 基于弛豫电压模型的锂离子电池RUL预测[J]. 电池, 2024,54(4):542-547. DOI: 10.19535/j.1001-1579.2024.04.021.
	ZHAI J F, LI B, LI Y L, et al. RUL prediction of Li-ion battery based on relaxation voltage model[J]. Dianchi(Battery Bimonthly), 2024, 54(4): 542-547. DOI: 10.19535/j.1001-1579.2024.04.021.
[12]	SEVERSON K A, ATTIA P M, JIN N, et al. Data-driven prediction of battery cycle life before capacity degradation[J]. Nature Energy, 2019, 4(5): 383-391. DOI: 10.1038/s41560-019-0356-8.
[13]	KE Y Q, JIANG Y Y, ZHU R, et al. Early prediction of knee point and knee capacity for fast-charging lithium-ion battery with uncertainty quantification and calibration[J]. IEEE Transactions on Transportation Electrification, 2024, 10(2): 2873-2885. DOI: 10.1109/TTE.2023.3304670.
[14]	CELIK B, SANDT R, DOS SANTOS L C P, et al. Prediction of battery cycle life using early-cycle data, machine learning and data management[J]. Batteries, 2022, 8(12): 266. DOI: 10.3390/batteries8120266.
[15]	AFSHARI S S, CUI S H, XU X Y, et al. Remaining useful life early prediction of batteries based on the differential voltage and differential capacity curves[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 71: 6500709. DOI: 10.1109/TIM.2021. 3117631.
[16]	李嘉波, 王志璇, 田迪, 等. 变模态分解下SSA-LSTM组合的锂离子电池剩余使用寿命预测方法[J]. 储能科学与技术, 2025, 14(2): 659-670. DOI: 10.19799/j.cnki.2095-4239.2024.0732.
	LI J B, WANG Z X, TIAN D, et al. Prediction method for remaining service life of lithium batteries using SSA-LSTM combination under variable mode decomposition[J]. Energy Storage Science and Technology, 2025, 14(2): 659-670. DOI: 10.19799/j.cnki.2095-4239.2024.0732.
[17]	JIA J F, WANG K K, SHI Y H, et al. A multi-scale state of health prediction framework of lithium-ion batteries considering the temperature variation during battery discharge[J]. Journal of Energy Storage, 2021, 42: 103076. DOI: 10.1016/j.est.2021.103076.
[18]	WEI M, YE M, ZHANG C W, et al. A multi-scale learning approach for remaining useful life prediction of lithium-ion batteries based on variational mode decomposition and Monte Carlo sampling[J]. Energy, 2023, 283: 129086. DOI: 10.1016/j.energy.2023.129086.
[19]	GAO K P, SUN J J, HUANG Z Y, et al. Capacity prediction of lithium-ion batteries based on ensemble empirical mode decomposition and hybrid machine learning[J]. Ionics, 2024, 30(11): 6915-6932. DOI: 10.1007/s11581-024-05768-y.
[20]	WILLIARD N, HE W, OSTERMAN M, et al. Comparative analysis of features for determining state of health in lithium-ion batteries[J]. International Journal of Prognostics and Health Management, 2013, 4(1): DOI: 10.36001/ijphm.2013.v4i1.1437
[21]	LI H, LI Z, MO W. A time varying filter approach for empirical mode decomposition[J]. Signal Processing, 2017, 138: 146-158. DOI: 10.1016/j.sigpro.2017.03.019.
[22]	LIU C, WU Y Y, XU Z L, et al. Short-term power load prediction method for high voltage power cables based on IGOA-VMD-LSTM-MHSAM[C]// 2024 The 9th International Conference on Power and Renewable Energy (ICPRE). IEEE, 2024: 1685-1690. DOI: 10.1109/ICPRE62586.2024.10768365.
[23]	IKOTUN A M, EZUGWU A E, ABUALIGAH L, et al. K-means clustering algorithms: A comprehensive review, variants analysis, and advances in the era of big data[J]. Information Sciences, 2023, 622: 178-210. DOI: 10.1016/j.ins.2022.11.139.
[24]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. DOI: 10.1162/neco.1997.9.8.1735.
[25]	JIAO M, WANG D Q, QIU J L. A GRU-RNN based momentum optimized algorithm for SOC estimation[J]. Journal of Power Sources, 2020, 459: 228051. DOI: 10.1016/j.jpowsour.2020.228051.
[26]	GHASEMI M, DERICHE M, TROJOVSKÝ P, et al. An efficient bio-inspired algorithm based on humpback whale migration for constrained engineering optimization[J]. Results in Engineering, 2025, 25: 104215. DOI: 10.1016/j.rineng.2025.104215.

参数	WMA-LSTM参数值(IMF-H)	WMA-LSTM参数值(IMF-M&L)	WMA-GRU参数值(IMF-T)
Training set	9.5%	9.5%	9.5%
Test set	90.5%	90.5%	90.5%
Optimizer	Adam	Adam	Adam
MaxEpochs	1100	800	800
InitialLearnRate	0.001	0.001	0.001
NumHiddenUnits	78	80	77
LearnRateDropPeriod	0.9*MaxEpochs	0.9*MaxEpochs	0.8*MaxEpochs
LearnRateDropFactor	0.2	0.2	0.01
L2Regularization	0.0001	0.01	0.0001

电池名称	CS2-35	CS2-36	CS2-37	CS2-38
早期剩余使用寿命实际值	439	397	462	489
早期剩余使用寿命平均预测值	442	399	465	491
平均误差值	3	2	3	2

方法	RMSE	MAE	MAPE	计算时间/s
method1(本工作)	0.012795	0.008655	1.48%	345.14
method2	0.017819	0.011417	2.03%	198.67
method3	0.021085	0.013243	2.20%	108.29
method4	0.040121	0.029786	5.25%	92.2

电池名称	误差指标	Proposed method	LSTM	BiLSTM	GRU	BiGRU	CNN
CS2-35	RMSE	0.012795	0.017819	0.04346	0.042771	0.051151	0.04906
	MAE	0.0086551	0.011417	0.028255	0.028971	0.035395	0.031632
	MAPE	1.4821%	2.0284%	5.4588%	5.4487%	6.5359%	6.1526%
CS2-36	RMSE	0.010392	0.017327	0.056927	0.043013	0.041721	0.2533
	MAE	0.0085821	0.010326	0.038947	0.031266	0.027494	0.17418
	MAPE	1.6543%	3.06%	9.9642%	7.5431%	7.3154%	44.401%
CS2-37	RMSE	0.012757	0.013615	0.039321	0.02109	0.032949	0.052894
	MAE	0.0084692	0.0062687	0.024614	0.015302	0.021141	0.034002
	MAPE	1.677%	1.5531%	5.0805%	2.8412%	4.2859%	6.7799%
CS2-38	RMSE	0.013085	0.019694	0.036283	0.023644	0.048999	0.078381
	MAE	0.010647	0.0095519	0.022347	0.016352	0.032896	0.057991
	MAPE	1.7069%	2.1601%	4.45%	3.0372%	6.2254%	10.3913%

误差指标	CH8	CH9	CH11	CH12
RMSE	0.0093619	0.0074526	0.0095817	0.0083431
MAE	0.0072235	0.0053415	0.0084304	0.0069983
MAPE	0.75174%	0.53872%	0.86675%	0.70607%