摘要：
锂离子电池剩余使用寿命(RUL)是电池健康管理的一个重要指标。本工作采用电池容量作为健康状况的指标，使用模态分解和机器学习算法，提出了一种CEEMDAN-RF-SED-LSTM方法去预测锂电池RUL。首先采用CEEMDAN分解电池容量数据，为了避免波动分量里的噪音对模型预测能力的影响，且又不完全抛弃波动分量里的特征信息，本工作提出使用随机森林(RF)算法得到每个波动分量的重要性排序和数值，以此作为每个分量对原始数据解释能力的权重。然后将权重值和不同波动分量构建的神经网络模型得到的预测结果进行加权重构，进而得到锂离子电池的RUL预测。文章对比了单一模型和组合模型预测精度，加入了RF的组合模型预测精度让五种神经网络的表现都有进一步的提升。最后，对表现较好的两种网络——LSTM和GRU引入了简单编码解码(SED)的机制，让其更好地学习到序列数据全局时间上的特征和远程的依赖关系。以NASA数据集作为研究对象进行该方法的性能测试。实验结果表明，CEEMDAN-RF-SED-LSTM模型对电池RUL预测表现效果好，预测结果相比单一模型具有更低的误差。
Abstract:
Lithium-ion battery's remaining useful life (RUL) is an important indicator of battery health management. Therefore, in this paper, using battery capacity as an indicator of health status, including modal decomposition and machine learning algorithms, a CEEMDAN-RF-SED-LSTM method was proposed to predict lithium battery RUL. First, adaptive white-noise full-ensemble empirical-mode decomposition (CEEMDAN) was used to decompose the battery capacity data. Then, to avoid the influence of the noise in the fluctuation component on the prediction ability of the model and not completely discard the characteristic information in the fluctuation component, this paper used the Random Forest algorithm to obtain important values for each fluctuation component, after which sexual ranking and numerical values were used as weights for each component's ability to explain the original data. Subsequently, the weight value and prediction result obtained by the neural network model constructed by different fluctuation components were weighted and reconstructed, resulting in the RUL prediction of the lithium-ion battery. Next, this research compared the prediction accuracy of the single model and the combined model, followed by the addition of the combined model prediction accuracy of RF, to improve the performance of the five neural networks further, after which the Simple Encoder-Decoder (SED) mechanism was introduced for the two networks with better performance, LSTM and GRU, to better learn the global temporal features and long-range dependencies of sequence data. We finally tested the method's performance using the NASA dataset as the research object. The experimental results showed that although the CEEMDAN-RF-SED-LSTM model performed well in battery RUL prediction, the prediction results had lower errors than the single model.
Key words:
Li-ion battery,
life prediction,
adaptive white noise full ensemble empirical mode decomposition,
random forest,
neural network
B0005预测评价指标结果"
|
电池编号
|
神经网络模型
|
模型方法
|
运行时间/s
|
MAE
|
RMSE
|
MAPE
|
RE
|
|
B0005
|
LSTM
|
LSTM
|
18.77
|
0.058592
|
0.070959
|
0.043809
|
0.040650
|
|
CEEMDAN-LSTM
|
70.65
|
0.043490
|
0.057525
|
0.026992
|
0.026260
|
|
CEEMDAN-RF-LSTM
|
74.93
|
0.040182
|
0.053559
|
0.024962
|
0.024390
|
|
RNN
|
RNN
|
38.07
|
0.280588
|
0.360674
|
0.180698
|
0.335772
|
|
CEEMDAN-RNN
|
87.56
|
0.245580
|
0.332067
|
0.169977
|
0.390244
|
|
CEEMDAN-RF-RNN
|
91.95
|
0.071269
|
0.087691
|
0.047033
|
0.195122
|
|
GRU
|
GRU
|
17.84
|
0.060329
|
0.072215
|
0.044820
|
0.073171
|
|
CEEMDAN-GRU
|
61.86
|
0.045586
|
0.060776
|
0.028049
|
0.028130
|
|
CEEMDAN-RF-GRU
|
66.39
|
0.041444
|
0.055729
|
0.025512
|
0.024390
|
|
CNN
|
CNN
|
13.09
|
0.068268
|
0.081806
|
0.039025
|
0.440650
|
|
CEEMDAN-CNN
|
47.96
|
0.062904
|
0.073718
|
0.040348
|
0.304553
|
|
CEEMDAN-RF-CNN
|
52.11
|
0.062832
|
0.073593
|
0.040437
|
0.280813
|
|
MLP
|
MLP
|
13.02
|
0.070968
|
0.081720
|
0.047639
|
0.325203
|
|
CEEMDAN-MLP
|
45.63
|
0.068850
|
0.080262
|
0.045182
|
0.329496
|
|
CEEMDAN-RF-MLP
|
49.97
|
0.068734
|
0.080241
|
0.044737
|
0.305152
|
B0006预测评价指标结果"
|
电池编号
|
神经网络模型
|
模型方法
|
运行时间/s
|
MAE
|
RMSE
|
MAPE
|
RE
|
|
B0006
|
LSTM
|
LSTM
|
20.72
|
0.054751
|
0.069080
|
0.037688
|
0.068037
|
|
CEEMDAN-LSTM
|
64.83
|
0.044735
|
0.058687
|
0.028165
|
0.056075
|
|
CEEMDAN-RF-LSTM
|
69.37
|
0.039161
|
0.049791
|
0.024893
|
0.054112
|
|
RNN
|
RNN
|
38.66
|
0.207933
|
0.325623
|
0.177686
|
0.439252
|
|
CEEMDAN-RNN
|
87.25
|
0.249714
|
0.345865
|
0.180759
|
0.355140
|
|
CEEMDAN-RF-RNN
|
91.36
|
0.086660
|
0.110239
|
0.061186
|
0.196262
|
|
GRU
|
GRU
|
17.84
|
0.064288
|
0.082282
|
0.044646
|
0.046729
|
|
CEEMDAN-GRU
|
58.64
|
0.042328
|
0.052737
|
0.027216
|
0.065421
|
|
CEEMDAN-RF-GRU
|
62.91
|
0.041394
|
0.051566
|
0.026716
|
0.065421
|
|
CNN
|
CNN
|
13.19
|
0.083262
|
0.095125
|
0.054467
|
0.233645
|
|
CEEMDAN-CNN
|
47.86
|
0.046205
|
0.062880
|
0.030333
|
0.029597
|
|
CEEMDAN-RF-CNN
|
51.37
|
0.041894
|
0.054645
|
0.027848
|
0.028037
|
|
MLP
|
MLP
|
13.06
|
0.085937
|
0.096994
|
0.064232
|
0.088037
|
|
CEEMDAN-MLP
|
39.78
|
0.051744
|
0.066709
|
0.034459
|
0.084112
|
|
CEEMDAN-RF-MLP
|
44.08
|
0.048342
|
0.063234
|
0.032276
|
0.037383
|
四组电池预测评价指标结果"
|
电池编号
|
模型方法
|
MAE
|
RMSE
|
MAPE
|
RE
|
|
B0005
|
CEEMDAN-RF-SED-LSTM
|
0.025569
|
0.031899
|
0.016515
|
0.018130
|
|
CEEMDAN-RF-SED-GRU
|
0.037491
|
0.044218
|
0.024056
|
0.089431
|
|
B0006
|
CEEMDAN-RF-SED-LSTM
|
0.032727
|
0.039157
|
0.021635
|
0.056075
|
|
CEEMDAN-RF-SED-GRU
|
0.041408
|
0.049697
|
0.028029
|
0.093458
|
|
B0007
|
CEEMDAN-RF-SED-LSTM
|
0.021852
|
0.026331
|
0.013319
|
0.059524
|
|
CEEMDAN-RF-SED-GRU
|
0.028275
|
0.036218
|
0.017499
|
0.061905
|
|
B0018
|
CEEMDAN-RF-SED-LSTM
|
0.025835
|
0.031901
|
0.016380
|
0.094737
|
|
CEEMDAN-RF-SED-GRU
|
0.030818
|
0.040062
|
0.019301
|
0.081053
|
ZHANG R K, GUO Y F, YU X Y, et al. A review of data-driven Lithium-ion battery RUL predictions[J/OL]. Journal of Power Sources,2021[2022-03-02].https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJLAST&filename=DYXB202108 24003&uniplatform=NZKPT&v=wYAa586U8l2FcQbuNntpFqahCvLt EGyCU5ZM5Yf0wRMkJw67wWLkq3r54EysBEvW.
XING Y J, MA E W M, TSUI K L, et al. An ensemble model for predicting the remaining useful performance of lithium-ion batteries[J]. Microelectronics Reliability, 2013, 53(6): 811-820.
WANG J W, DENG Z W, YU T, et al. State of health estimation based on modified Gaussian process regression for lithium-ion batteries[J]. Journal of Energy Storage, 2022, 51: doi: 10.1016/j.est.2022.104512.
周亚鹏, 郭彪, 王一纯. 基于PF-ARIMA的锂离子电池剩余寿命预测[J]. 电池工业, 2022, 26(1): 19-22.
栗志展, 秦金磊, 梁嘉宁, 李峥嵘, 王瑞, 王得丽.
高镍三元层状锂离子电池正极材料：研究进展、挑战及改善策略
[J]. 储能科学与技术, 2022, 11(9): 2900-2920.
牛少军, 吴凯, 朱国斌, 王艳, 曲群婷, 郑洪河.
锂离子电池硅基负极循环过程中的膨胀应力
[J]. 储能科学与技术, 2022, 11(9): 2989-2994.
陈晓宇, 耿萌萌, 王乾坤, 沈佳妮, 贺益君, 马紫峰.
基于电化学阻抗特征选择和高斯过程回归的锂离子电池健康状态估计方法
[J]. 储能科学与技术, 2022, 11(9): 2995-3002.
贡淑雅, 王跃, 李萌, 邱景义, 王洪, 文越华, 徐斌.
锂离子电池负极材料
TiNb
2
O
7
的研究进展
[J]. 储能科学与技术, 2022, 11(9): 2921-2932.
张群斌, 董陶, 李晶晶, 刘艳侠, 张海涛.
废旧电池电解液回收及高值化利用研发进展
[J]. 储能科学与技术, 2022, 11(9): 2798-2810.
张越, 孔得朋, 平平.
液冷板抑制锂离子电池组热失控蔓延性能及优化设计
[J]. 储能科学与技术, 2022, 11(8): 2432-2441.
卓萍, 朱艳丽, 齐创, 王聪杰, 高飞.
锂离子电池组过充燃烧爆炸特性
[J]. 储能科学与技术, 2022, 11(8): 2471-2479.
曹志成, 周开运, 朱家立, 刘高明, 严慜, 汤舜, 曹元成, 程时杰, 张炜鑫.
锂离子电池储能系统消防技术的中国专利分析
[J]. 储能科学与技术, 2022, 11(8): 2664-2670.
马勇, 李晓涵, 孙磊, 郭东亮, 杨景刚, 刘建军, 肖鹏, 钱广俊.
基于三维电化学热耦合析锂模型的锂离子电池参数设计
[J]. 储能科学与技术, 2022, 11(8): 2600-2611.
唐亮, 尹小波, 吴候福, 刘鹏杰, 王青松.
电化学储能产业发展对安全标准的需求
[J]. 储能科学与技术, 2022, 11(8): 2645-2652.
霍丽萍, 栾伟玲, 庄子贤.
锂离子电池储能安全技术的发展态势
[J]. 储能科学与技术, 2022, 11(8): 2671-2680.
徐成善, 鲁博瑞, 张梦启, 王淮斌, 金昌勇, 欧阳明高, 冯旭宁.
储能锂离子电池预制舱热失控烟气流动研究
[J]. 储能科学与技术, 2022, 11(8): 2418-2431.
韦荣阳, 毛阗, 高晗, 彭建仁, 杨健.
基于
TWP-SVR
的锂离子电池健康状态估计
[J]. 储能科学与技术, 2022, 11(8): 2585-2599.
喻航, 张英, 徐超航, 余思瀚.
锂电储能系统热失控防控技术研究进展
[J]. 储能科学与技术, 2022, 11(8): 2653-2663.
凡注明本刊独家的版权为《储能科学与技术》所有，欢迎转载但请务必注明来源。如果本文侵犯您的权益，请联系本站删除！
凡注明“来源：XXX（非《储能科学与技术》）”，均转载自其它媒体，转载目的在于传递更多信息，并不代表本网赞同其观点和对其真实性负责。其他媒体如需转载，请与稿件来源方联系，如产生任何问题与本网无关。
地址：北京东城区青年湖南街13号化工出版社3层 电话：86-10-64519601/64519602/64519643 E-mail：
[email protected];
[email protected]
京公网安备 11010102001997号
京ICP备12046843号-1
