为了更精准地预测锂离子电池剩余寿命,针对电池的容量回升现象导致的预测不准确问题,提出一种融合自适应噪声完全集合经验模态分解（CEEMDAN）分解算法、堆栈式稀疏自编码器（SSAE）、改进的时间卷积网络（LATCN）与改进的Transformer（SWFormer）的混合预测模型。模型首先通过CEEMDAN算法对数据进行分解,利用稀疏堆叠自编码器提取电池分解后的模态分量稀疏特征,随后构建改进的双分支网络模型LATCN-SWFormer进行寿命预测,并应用提出的特征融合注意力机制进行特征融合,最终采用单一的柯尔莫哥洛夫-阿诺德网络（KAN）层作为增强网络进行电池剩余寿命的最终预测。采用CALCE数据集以及NASA泛化实验数据集进行验证,实验结果表明,相较于XGBoost、BiTCN和iTransformer方法,所提模型的均方根误差（RMSE）、平均绝对误差（MAE）和平均绝对百分比误差（MAPE）最高降幅分别达73.6%、68.6%和66.2%。在CS2_36和B0006电池的容量回升阶段,模型可实现零误差预测,充分验证了其在应对非线性退化特性与容量回升干扰方面的显著优势。同时,在容量快速衰减阶段,模型展现出更强的拟合能力,在多个子数据集上均可实现显著的误差下降,体现出良好的泛化能力和鲁棒性。
Abstract
To more accurately predict the remaining useful life （RUL） of lithium-ion batteries and address the prediction inaccuracy caused by the capacity recovery phenomenon, this paper proposes a hybrid prediction model that integrates the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise （CEEMDAN） algorithm, a Stacked Sparse Autoencoder （SSAE）, an improved Temporal Convolutional Network （LATCN）, and an improved Transformer （SWFormer）. The model first decomposes the data using the CEEMDAN algorithm, and then employs the stacked sparse autoencoder to extract sparse features from the decomposed modal components. Subsequently, an improved dual-branch network model, LATCN-SWFormer, is constructed for life prediction, and a novel feature fusion attention mechanism is applied to fuse the features. Finally, a single Kolmogorov-Arnold Network （KAN） layer serves as an enhancement network for the final prediction of battery remaining useful life. The CALCE dataset and the NASA generalization experimental dataset are used for validation. Experimental results show that compared with XGBoost, BiTCN, and iTransformer methods, the proposed model reduces the root mean square error （RMSE）, mean absolute error （MAE）, and mean absolute percentage error （MAPE） by 73.6%, 68.6%, and 66.2%, respectively. In the capacity recovery stage of CS2_36 and B0006 batteries, the model achieves zero-error prediction, fully demonstrating its significant advantages in handling nonlinear degradation characteristics and capacity recovery interference. Meanwhile, during the rapid capacity decay stage, the model exhibits stronger fitting ability and achieves substantial error reduction on multiple sub-datasets, reflecting good generalization ability and robustness.
关键词
锂离子电池 /
剩余使用寿命 /
自适应噪声完全集合经验模态分解 /
特征融合注意力 /
局部注意力时间卷积网络
Key words
lithium ion batteries /
remaining useful life /
CEEMDAN /
feature fusion attention /
temporal convolutional network /
SWFormer
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 李炳金, 韩晓霞, 张文杰, 等 . 锂离子电池剩余使用寿命预测方法综述[J]. 储能科学与技术, 2024, 13(4): 1266-1276.
Li B J, Han X X, Zhang W J, et al.Review of the remaining useful life prediction methods for lithium-ion batteries[J]. Energy Storage Science and Technology, 2024, 13(4): 1266-1276.
[2] 何冰琛, 杨薛明, 王劲松, 等 . 基于 PCA-GPR的锂离子电池剩余使用寿命预测[J]. 太阳能学报, 2022, 43(5): 484-491.
He B C, Yang X M, Wang J S, et al.Prediction of remaining useful life of lithium-ion batteries based on PCA-GPR[J]. Acta Energiae Solaris Sinica, 2022, 43(5): 484-491.
[3] 晋殿卫, 顾则宇, 张志宏 . 锂电池健康度和剩余寿命预测算法研究[J]. 电力系统保护与控制, 2023, 51(1): 122-130.
Jin D W, Gu Z Y, Zhang Z H.Lithium battery health degree and residual life prediction algorithm[J]. Power System Protection and Control, 2023, 51(1): 122-130.
[4] Zhao L Y, Wang T S, Zuo F K, et al.A fast-charging/ discharging and long-term stable artificial electrode enabled by space charge storage mechanism[J]. Nature Communications, 2024, 15: 3778.
[5] Shu X, Liu Y, Shen J, et al.Capacity prediction for lithium-ion batteries based on improved least squares support vector machine and box-cox transformation[J]. Journal of Mechanical Engineering, 2021, 57(14): 118-128.
[6] 张若可, 郭永芳, 余湘媛, 等 . 基于数据驱动的锂离子电池RUL预测综述[J]. 电源学报 , 2023, 21(5): 182-190.
Zhang R K, Guo Y F, Yu X Y, et al.Review of data driven RUL prediction for lithium-ion batteries[J]. Journal of Power Supply, 2023, 21(5): 182-190.
[7] Alipour M, Yin L T, Tavallaey S S, et al.A surrogate-assisted uncertainty quantification and sensitivity analysis on a coupled electrochemical - thermal battery aging model[J]. Journal of Power Sources, 2023, 579: 233273.
[8] 宋胜, 李云伍, 赵颖, 等 . 锂离子电池片段数据的荷电状态估计研究[J]. 电源技术, 2022, 46(7): 734-738.
Song S, Li Y W, Zhao Y, et al.Research on SOC estimation based on fragment data of lithium-ion battery[J]. Chinese Journal of Power Sources, 2022, 46(7): 734-738.
[9] 刘博, 尹杰, 李然 . 基于改进粒子滤波的锂离子电池剩余寿命预测[J]. 电力系统保护与控制, 2024, 52(9): 123-131.
Liu B, Yin J, Li R.Improved particle filter algorithm for remaining useful life prediction of lithium-ion batteries[J]. Power System Protection and Control, 2024, 52(9): 123-131.
[10] 徐洲常, 王林军, 刘洋, 等 . 采用改进回归型支持向量机的滚动轴承剩余寿命预测方法[J]. 西安交通大学学报, 2022, 56(3): 197-205.
Xu Z C, Wang L J, Liu Y, et al.A prediction method for remaining life of rolling bearing using improved regression support vector machine[J]. Journal of Xi'an Jiaotong University, 2022, 56(3): 197-205.
[11] 周航, 程泽, 弓清瑞, 等 . 基于TCN编码的锂离子电池SOH估计方法[J]. 湖南大学学报(自然科学版), 2023, 50(4):185-192.
Zhou H, Cheng Z, Gong Q R, et al.SOH estimation method for lithium-ion batteries based on TCN encoding[J]. Journal of Hunan University(Natural Sciences), 2023, 50(4): 185-192.
[12] Chen D Q, Hong W C, Zhou X Z.Transformer network for remaining useful life prediction of lithium-ion batteries[J]. IEEE Access, 2022, 10: 19621-19628.
[13] 刘斌, 吉春霖, 曹丽君, 等 . 基于自适应噪声完全集合经验模态分解与BiLSTM-Transformer的锂离子电池剩余使用寿命预测[J]. 电力系统保护与控制, 2024,52(15):167-177.
Liu B, Ji C L, Cao L J, et al.Remaining Useful Life Prediction of Lithium-Ion Batteries Based on CEEMDAN and BiLSTM-Transformer[J]. Power System Protection and Control, 2024, 52(15): 167-177.
[14] 刘金凤, 陈浩玮, Herbert h C I. 基于VMD和DAIPSO-GPR解决容量再生现象的锂离子电池寿命预测研究[J]. 电子与信息学报, 2023, 45(3):1111-1120.
Liu J, Chen H, Herbert Ho-Ching Iu. Li-ion batteries life prediction based on variational mode decomposition and DAIPSO-GPR to solve the capacity regeneration phenomenon[J]. Journal of Electronics & Information Technology, 2023, 45(3): 1111-1120.
[15] Wang Z Q, Guo Y M, Xu C.An HI extraction framework for lithium-ion battery prognostics based on SAE-VMD[J]. Journal of Northwestern Polytechnical University, 2020, 38(4): 814-821.
[16] Zhou J, Zhang S, Wang P.Fault diagnosis for power batteries based on a stacked sparse autoencoder and a convolutional block attention capsule network[J]. Processes, 2024, 12(4): 816.
[17] 郭喜峰, 王凯泽, 单丹, 等 . 多角度基于 CEEMDANCNN-BiLSTM模型的锂离子电池RUL预测[J]. 太阳能学报, 2024, 45(7): 181-189.
Guo X F, Wang K Z, Shan D, et al.Rul prediction for lithium ion batteries based on CEEMDAN-CNN-BiLSTM model from multiple perspectives[J]. Acta Energiae Solaris Sinica, 2024, 45(7): 181-189.
[18] 王昆, 郭迎清, 赵万里, 等. 基于SSAE和相似性匹配的航空发动机剩余寿命预测[J]. 北京航空航天大学学报, 2023, 49(10): 2817-2825.
Wang K, Guo Y Q, Zhao W L, et al.Remaining useful life prediction of aeroengine based on SSAE and similarity matching[J]. Journal of Beijing University of Aeronautics and Astronautics, 2023, 49(10): 2817-2825.
[19] 吴诗淼, 王文波, 朱婷, 等 . 基于跳跃连接多尺度CNN的锂离子电池剩余寿命预测[J]. 太阳能学报, 2024, 45(7): 199-208.
Wu S M, Wang W B, Zhu T, et al.Prediction of residual life of lithium ion battery based on multi-scale CNN with jump connection cnn[J]. Acta Energiae Solaris Sinica, 2024, 45(7): 199-208.
基金
国家自然科学基金（62241307）; 甘肃省科技计划（22YF7FA166; 23CXGA0060）; 兰州市科技计划（2022-RC-60; 2023-RC-26）
PDF(4856 KB)
