基于ResNet-Bi-LSTM-Attention的锂离子电池剩余使用寿命预测

doi:10.19799/j.cnki.2095-4239.2022.0652

摘要/Abstract

摘要：

锂离子电池剩余使用寿命(RUL)预测是锂离子电池研究的一个重要方向，通过对RUL的准确预测，可以降低锂离子电池出现事故的概率。针对锂离子电池RUL的准确预测，该研究提出一种综合残差神经网络(ResNet)和双向长短期记忆网络(Bi-LSTM)的优势，并且加入注意力机制(Attention)的锂离子电池RUL预测模型。首先选取能够表现电池寿命的特征参数作为输入量，利用ResNet提取输入数据的隐含特征信息，然后利用Bi-LSTM对时间序列信息进行预测，并且结合注意力机制对预测结果进行权重分配，得到最终的锂离子电池的RUL预测结果。通过美国马里兰大学(CALCE)提供的开源数据集进行锂离子电池RUL预测试验，并与现有的预测模型进行对比试验，对比模型的预测结果，试验结果表明提出的ResNet-Bi-LSTM-Attention模型能够准确地进行锂离子电池RUL预测，各项误差都比较低，具有较好的精度和准确性。最后使用美国航空航天局(NASA)提供的锂离子电池开源数据集进行泛化性实验，证明了ResNet-Bi-LSTM-Attention模型在不同电池RUL预测中具有良好的准确性，可以被广泛使用。

关键词: 锂离子电池, 残差神经网络, 双向长短期记忆网络, 注意力机制, 剩余使用寿命预测

Abstract:

Accurate prediction of remaining useful life (RUL) of lithium-ion batteries is an important research topic as it can help reduce the risks of lithium-ion battery accidents. Thus, this research proposes a prediction model of RUL of lithium-ion batteries that comprises an attention mechanism and combines the advantages of residual neural network (ResNet) and bidirectional short-term memory network (Bi LSTM). For this, the characteristic parameters that can represent the battery life were selected as the input quantity. ResNet was used to extract the implicit characteristic information of the input data and Bi LSTM was used to predict the time series information. The attention mechanism was used to distribute the weight of the prediction results so as to obtain the final RUL prediction results of lithium-ion batteries. The RUL prediction test of lithium-ion batteries was carried out using the open-source dataset provided by the University of Maryland (CALCE) of the United States, and the obtained results were compared with that of the existing prediction models. The test results show that the proposed model can accurately predict the RUL of lithium-ion batteries, with relatively low errors and good accuracy. Finally, the generalization experiment was carried out using the open-source dataset of lithium-ion batteries provided by NASA, and its results confirmed that the proposed model has good accuracy in predicting the RUL of different batteries and, thus, has wide applications.

Key words: lithium-ion battery, residual neural network, bidirectional long short-term memory network, attention mechanism, remaining service life prediction

中图分类号:

TM 911

王朋凯, 张新燕, 张光昊. 基于ResNet-Bi-LSTM-Attention的锂离子电池剩余使用寿命预测[J]. 储能科学与技术, 2023, 12(4): 1215-1222.

Pengkai WANG, Xinyan ZHANG, Guanghao ZHANG. Remaining useful life prediction of lithium-ion batteries based on ResNet-Bi-LSTM-Attention model[J]. Energy Storage Science and Technology, 2023, 12(4): 1215-1222.

图/表 14

图1

图2

图3

图4

图5

图6

图7

表1

图8

表2

表3

图9

表4

图10

参考文献 20

1	胡振恺, 雷博, 李勇琦, 等. 储能用锂离子电池安全性测试与评估方法比较[J]. 储能科学与技术, 2022, 11(5): 1650-1656.
	HU Z K, LEI B, LI Y Q, et al. Comparative study on safety test and evaluation methods of lithium-ion batteries for energy storage[J]. Energy Storage Science and Technology, 2022, 11(5): 1650-1656.
2	闫啸宇, 周思达, 卢宇, 等. 锂离子电池容量衰退机理与影响因素研究[J/OL]. 北京航空航天大学学报:1-13.[2022-04-20]. https://doi.org/10.13700/j.bh.1001-5965.2021.0458.
	YAN X Y, ZHOU S D, LU Y, et al. Study on the capacity decline mechanism and influencing factors of lithium-ion batteries[J/OL]. Journal of Beijing University of Aeronautics and Astronautics: 1-13.[2022-04-20]. https://doi.org/10.13700/j.bh.1001-5965.2021.0458.
3	蔡艳平, 陈万, 苏延召, 等. 锂离子电池剩余寿命预测方法综述[J]. 电源技术, 2021, 45(5): 678-682.
	CAI Y P, CHEN W, SU Y Z, et al. Review of remaining useful life prediction for lithium ion batteries[J]. Chinese Journal of Power Sources, 2021, 45(5): 678-682.
4	索梦磊, 李俊夫, 赵明, 等. 锂离子电池寿命预测研究进展[J]. 电池工业, 2020, 24(5): 255-263.
	SUO M L, LI J F, ZHAO M, et al. Research progress of lithium ion battery life prediction[J]. Chinese Battery Industry, 2020, 24(5): 255-263.
5	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/OL]. 电源学报:1-15.[2022-04-20]. http://kns.cnki.net/kcms/detail/12.1420.tm.20210824.1932.014.html.
	ZHANG R K, GUO Y F, YU X Y, et al. A review of RUL prediction of lithium-ion batteries based on data-driven[J/OL]. Journal of Power Sources: 1-15. [2022-04-20]. http://kns.cnki.net/kcms/detail/12.1420.tm.20210824.1932.014.html.
6	VIANI L, ROSI P, ROTUNNO E, et al. Enhancing electron computational ghost imaging using artificial neural networks[J]. Microscopy and Microanalysis, 2022, 28(S1): 2242-2244.
7	ZHANG Y Y. MOOC teaching model of basic education based on fuzzy decision tree algorithm[J]. Computational Intelligence and Neuroscience, 2022: doi:10.1155/2022/3175028.
8	GANAIE M A, TANVEER M, SUGANTHAN P N, et al. Oblique and rotation double random forest[J]. Neural Networks, 2022, 153: 496-517.
9	TEPE C, DEMIR M C. Real-time classification of EMG myo armband data using support vector machine[J]. IRBM, 2022, 43(4): 300-308.
10	KARTHIK C R, RAGHUNANDAN, ASHWATH RAO B, et al. Forecasting variance of NiftyIT index with RNN and DNN[J]. Journal of Physics: Conference Series, 2022, 2161(1): 012005.
11	USHARANI B. ILF-LSTM: Enhanced loss function in LSTM to predict the sea surface temperature[J]. Soft Computing, 2022: 1-13.
12	吕明珠. 基于改进LSTM的滚动轴承性能退化趋势预测[J]. 轴承, 2022(4): 70-76.
	LÜ M Z. Performance degradation trend prediction of rolling bearings based on improved LSTM[J]. Bearing, 2022(4): 70-76.
13	BOU-RABEE M A, NAZ M Y, ALBALAA I E, et al. BiLSTM network-based approach for solar irradiance forecasting in continental climate zones[J]. Energies, 2022, 15(6): 2226.
14	SU Y, KONG X W, LIU G B. Advertising popularity feature collaborative recommendation algorithm based on attention-LSTM model[J]. Security and Communication Networks, 2021: 1-11.
15	朱张莉, 饶元, 吴渊, 等. 注意力机制在深度学习中的研究进展[J]. 中文信息学报, 2019, 33(6): 1-11.
	ZHU Z L, RAO Y, WU Y, et al. Research progress of attention mechanism in deep learning[J]. Journal of Chinese Information Processing, 2019, 33(6): 1-11.
16	GE F, ZHANG Y, XU J, et al. Prediction of disease-associated nsSNPs by integrating multi-scale ResNet models with deep feature fusion[J]. Briefings in Bioinformatics, 2022, 23(1): doi:10.1093/bib/bbab530.
17	王英楷, 张红, 王星辉. 基于1DCNN-LSTM的锂离子电池SOH预测[J]. 储能科学与技术, 2022, 11(1): 240-245.
	WANG Y K, ZHANG H, WANG X H. Hybrid 1DCNN-LSTM model for predicting lithium ion battery state of health[J]. Energy Storage Science and Technology, 2022, 11(1): 240-245.
18	ZHU M L, WANG Q Q, LUO J L. Emotion recognition based on dynamic energy features using a Bi-LSTM network[J]. Frontiers in Computational Neuroscience, 2022, 15: doi:10.3389/fncom.2021.741086.
19	史永胜, 施梦琢, 丁恩松, 等. 基于CEEMDAN-LSTM组合的锂离子电池寿命预测方法[J]. 工程科学学报, 2021, 43(7): 985-994.
	SHI Y S, SHI M Z, DING E S, et al. Combined prediction method of lithium-ion battery life based on CEEMDAN-LSTM[J]. Chinese Journal of Engineering, 2021, 43(7): 985-994.
20	刘芊彤, 邢远秀. 基于VMD-PSO-GRU模型的锂离子电池剩余寿命预测[J]. 储能科学与技术, 2023, 12(1): 236-246.
	LIU Q T, XING Y X. Remaining life prediction of lithium-ion battery based on VMD-PSO-GRU model[J]. Energy Storage Science and Technology, 2023, 12(1): 236-246.

特征参数	相关系数
CCCT	0.964 485
CVCT	-0.715 147
循环次数	-0.904 819

电池	LE/%	MAE	RMSE
CS2_35	0	0.016 780 087	0.024 443 291
CS2_36	0.5	0.017 120 992	0.024 411 781
CS2_37	0	0.013 785 957	0.020 911 526
CS2_38	0.33	0.011 880 822	0.018 031 097

电池	评价指标	CNN-Bi-LSTM	RESNET-Bi-LSTM	Bi-LSTM-Attention	RESNET-Bi-LSTM-Attention
CS2_35	MAE	0.017 697 76	0.036 269 485	0.019 153 179	0.016 780 087
CS2_35	RMSE	0.025 043 553	0.045 468 638	0.024 663 573	0.024 443 291
CS2_36	MAE	0.034 204 462	0.026 138 713	0.019 026 492	0.017 120 992
CS2_36	RMSE	0.041 248 78	0.034 371 126	0.024 257 293	0.024 411 781
CS2_37	MAE	0.018 991 438	0.014 542 032	0.031 645 693	0.013 785 957
CS2_37	RMSE	0.022 644 327	0.020 071 071	0.043 750 413	0.020 911 526
CS2_38	MAE	0.015 664 939	0.016 727 846	0.028 442 435	0.011 880 822
CS2_38	RMSE	0.020 777 135	0.025 571 059	0.041 147 682	0.018 031 097

电池	评价指标	CNN-Bi-LSTM	RESNET-Bi-LSTM	Bi-LSTM-Attention	RESNET-Bi-LSTM-Attention
CS2_35	MAE	0.017 061 121	0.015 047 912	0.015 392 527	0.011 880 822
CS2_35	RMSE	0.023 313 838	0.022 000 08	0.019 768 48	0.018 031 097
CS2_36	MAE	0.032 041 899	0.020 812 132	0.015 037 692	0.013 449 2
CS2_36	RMSE	0.044 247 904	0.023 471 57	0.020 134 884	0.019 324 627
CS2_37	MAE	0.016 239 846	0.012 047 632	0.022 348 693	0.012 346 628
CS2_37	RMSE	0.023 458 687	0.016 432 147	0.031 275 775	0.015 378 46
CS2_38	MAE	0.015 629 569	0.013 534 646	0.028 442 435	0.010 612 722
CS2_38	RMSE	0.018 485 614	0.020 826 706	0.031 246 735	0.015 324 861

[1]	宋缙华, 张兴浩, 丰震河, 程广玉, 顾洪汇, 顾海涛, 王可. 基于原位参比的氧化亚硅-石墨复合负极循环衰减机制[J]. 储能科学与技术, 2023, 12(4): 1059-1065.
[2]	胡力月, 姚行艳. 基于正交试验的锂离子电池热失控仿真[J]. 储能科学与技术, 2023, 12(4): 1268-1277.
[3]	刘峰, 陈海忠. 基于CEEMDAN和ISOA-ELM的锂电池荷电状态预测[J]. 储能科学与技术, 2023, 12(4): 1244-1256.
[4]	翟智, 王福金, 邸一, 马珮羽, 赵志斌, 陈雪峰. 基于分层对齐迁移学习的锂离子电池容量估计[J]. 储能科学与技术, 2023, 12(4): 1223-1233.
[5]	赵立禹, 孙桓五, 刘世闯, 闫志远. 重卡辅助动力电池加热系统能耗对比及优化[J]. 储能科学与技术, 2023, 12(4): 1139-1147.
[6]	杨妮, 苏岳锋, 王联, 李宁, 马亮, 朱晨. 聚焦离子束显微镜技术在锂离子电池领域的研究进展[J]. 储能科学与技术, 2023, 12(4): 1283-1294.
[7]	成雪莉, 张维福, 罗城城, 袁小亚. 一步水热法制备三维石墨烯/Fe₃O₄ 复合材料及其储锂性能[J]. 储能科学与技术, 2023, 12(4): 1066-1074.
[8]	李奇松, 陈荣, 李慧芳. 阻抗分析法在锂离子电池析锂阈值检测中的应用[J]. 储能科学与技术, 2023, 12(4): 1278-1282.
[9]	董渊昌, 庞晓琼, 贾建芳, 史元浩, 温杰, 李笑, 张鑫. 基于SVD-SAE-GPR的锂离子电池RUL预测[J]. 储能科学与技术, 2023, 12(4): 1257-1267.
[10]	管敏渊, 沈建良, 徐国华, 汤舜, 张炜鑫, 曹元成. 锂离子电池储能系统靶向消防装备设计与性能[J]. 储能科学与技术, 2023, 12(4): 1131-1138.
[11]	常修亮, 李希超, 贾隆舟, 韦守李, 王敬豪, 戴作强, 郑莉莉. 过充循环老化电池产热特性[J]. 储能科学与技术, 2023, 12(3): 685-697.
[12]	刘树港, 蒙波, 李政隆, 杨亚雄, 陈建. Li _x （Mg， Ni， Zn， Cu， Co） ₁_-_x O高熵氧化物负极材料电化学储锂特性[J]. 储能科学与技术, 2023, 12(3): 743-753.
[13]	姚逸鸣, 栾伟玲, 陈莹, 孙敏. 基于光学显微镜的锂离子电池材料老化衰减原位研究进展[J]. 储能科学与技术, 2023, 12(3): 777-791.
[14]	李奎杰, 楼平, 管敏渊, 莫金龙, 张炜鑫, 曹元成, 程时杰. 锂离子电池热失控多维信号演化及耦合机制研究综述[J]. 储能科学与技术, 2023, 12(3): 899-912.
[15]	封迈, 陈楠, 陈人杰. 锂离子电池低温电解液的研究进展[J]. 储能科学与技术, 2023, 12(3): 792-807.