基于CEEMDAN–LSTM組合的鋰離子電池壽命預測方法

史永勝; 施夢琢; 丁恩松; 洪元濤; 歐陽

doi:10.13374/j.issn2095-9389.2020.06.30.007

基于CEEMDAN–LSTM組合的鋰離子電池壽命預測方法

doi: 10.13374/j.issn2095-9389.2020.06.30.007

1.
陜西科技大學電氣與控制工程學院，西安 710021
2.
江蘇潤寅石墨烯科技有限公司，揚州 225600

基金項目: 國家自然科學基金資助項目（61871259）

詳細信息

通訊作者:
E-mail：84770540@qq.com

中圖分類號: TM912
計量
- 文章訪問數: 3454
- HTML全文瀏覽量: 1122
- PDF下載量: 196
- 被引次數: 0
出版歷程
- 收稿日期: 2020-06-30
- 網絡出版日期: 2020-09-24
- 刊出日期: 2021-07-01

Combined prediction method of lithium-ion battery life based on CEEMDAN–LSTM

1.
College of Electrical and Control Engineering, Shaanxi University of Science and Technology, Xi’an 710021, China
2.
Jiangsu Runyin Graphene Technology Co., Yangzhou 225600, China

More Information

Corresponding author: E-mail: 84770540@qq.com

摘要

摘要: 針對目前鋰離子電池壽命預測結果不準確的問題，提出了一種多模態分解的鋰離子電池組合預測模型，從而學習鋰離子電池退化過程的微小變化。該方法在單一長短期記憶（LSTM）預測模型的基礎上，采用了自適應噪聲完全集成的經驗模態分解(CEEMDAN)算法將鋰電池容量分為主退化趨勢和若干局部退化趨勢，然后使用長短期記憶神經網絡（LSTMNN）算法分別對所分解的若干退化數據進行壽命預測，最后將若干預測結果進行有效集成。結果表明，所提出的CEEMDAN?LSTM鋰離子電池組合預測模型最大平均絕對百分比誤差不超過1.5%，平均相對誤差在3%以內，且優于其他預測模型。
- 電池健康管理 /
- 鋰離子電池 /
- 剩余使用壽命 /
- 長短期記憶神經網絡 /
- 自適應噪聲完全集成經驗模態分解
Abstract: As a new generation of new energy battery, lithium-ion battery is widely used in various fields, including electronic products, electric vehicles, and power supply, due to its advantages of high energy density, light weight, long cycle life, small self-discharge, no memory effect, and no pollution. With the wide application of lithium-ion battery, numerous research on its performance has been done, including its health assessment as one of the hot spots. Repeated charging and discharging of a lithium-ion battery that was run under full charge state results to internal irreversible chemical changes leading to a fall in the maximum available capacity. Specifically, a decline to 70%–80% of the rated capacity results in lithium-ion battery failure. Battery failure may lead to electrical equipment damage, resulting in safety accidents. Therefore, it is of great significance to predict the remaining usable life of lithium-ion battery for improving system reliability. In this paper, a combination prediction model for lithium-ion batteries with multimode decomposition was presented based on the long and short-term memory (LSTM) prediction model to learn about small changes in its degradation process. A complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) algorithm was used to divide the capacity into main degradation trend and some local degradation trend. Long Short-Term Memory Neural Network (LSTMNN) algorithm was then introduced to perform the capacity prediction of decomposed degradation data. Finally, some prediction results were integrated effectively. The maximum mean absolute percentage error (MAPE) of the proposed CEEMDAN–LSTM lithium-ion battery combination prediction model does not exceed 1.5%. The average relative error is less than 3%, which is better than the other prediction model.

HTML全文

圖 1 LSTM總體結構圖

Figure 1. LSTM structure diagram

下載: 全尺寸圖片幻燈片

圖 2 組合模型預測框圖

Figure 2. Block diagram of combination prediction model

下載: 全尺寸圖片幻燈片

圖 3 鋰離子電池容量衰減數據。（a）CS33、CS34；（b）CS37、CS38、CX36、CX37

Figure 3. Capacity degradation data of lithium-ion batteries: (a) CS33, CS34; (b) CS37, CS38, CX36, CX37

下載: 全尺寸圖片幻燈片

圖 4 基于CEEMDAN的CS33容量序列分解

Figure 4. CS33 capacity sequence decomposition based on CEEMDAN

下載: 全尺寸圖片幻燈片

圖 5 CS33兩種算法重構誤差對比

Figure 5. Comparison of reconstruction errors between two algorithms

下載: 全尺寸圖片幻燈片

圖 6 50%訓練集電池預測結果。（a）CS33；（b）CS34；（c）CS37；（d）CS38；（e）CX36；（f）CX37

Figure 6. Battery prediction results under 50% training set: (a) CS33; (b) CS34; (c) CS37; (d) CS38; (e) CX36; (f) CX37

下載: 全尺寸圖片幻燈片

圖 7 相對誤差曲線（a）CS33；（b）CS34；（c）CS37；（d）CS38；（e）CX36；（f）CX37

Figure 7. Relative error curve: (a) CS33; (b) CS34; (c) CS37; (d) CS38; (e) CX36; (f) CX37

下載: 全尺寸圖片幻燈片

圖 8 50%訓練集電池預測誤差。（a）CS33；（b）CS34；（c）CS37；（d）CS38；（e）CX36；（f）CX37

Figure 8. Battery prediction error under 50% training set: (a) CS33; (b) CS34; (c) CS37; (d) CS38; (e) CX36; (f) CX37

下載: 全尺寸圖片幻燈片

圖 9 30%訓練集鋰離子電池預測結果。（a）CS33；（b）CS34；（c）CS37；（d）CS38；（e）CX36；（f）CX37

Figure 9. Battery prediction error under under 30% training set: (a) CS33; (b) CS34; (c) CS37; (d) CS38; (e) CX36; (f) CX37

下載: 全尺寸圖片幻燈片

圖 10 30%訓練集電池預測誤差。（a）CS33；（b）CS34；（c）CS37；（d）CS38；（e）CX36；（f）CX37

Figure 10. Battery prediction error under 30% training set: (a) CS33; (b) CS34; (c) CS37; (d) CS38; (e) CX36; (f) CX37

下載: 全尺寸圖片幻燈片

表 1 LiCoO₂電池詳細參數

Table 1. LiCoO₂ battery details

Battery	Anode and cathode materials	Size /mm	Weight /g
CS	LiCoO₂ cathode/graphite anode	5.4×33.6×50.6	21.1
CX	LiCoO₂ cathode/graphite anode	6.6×33.8×50	28

下載: 導出CSV

表 2 LSTM預測模型參數設置

Table 2. LSTM prediction model parameter setting

Number of iterations	Number of hidden layers	Number of hidden cells	Initial learning rate
500	1	200	0.002

下載: 導出CSV

表 3 50%訓練集鋰電池壽命預測誤差

Table 3. Lithium battery life prediction error under 50% training set

Model	Battery	RUL_tr	RUL_pr	RUL_er	P_er
LSTM	CS33	198	201	3	0.0152
	CS34	176	269	93	0.5284
	CS37	167	169	2	0.0120
	CS38	201	223	22	0.1095
	CX36	191	192	1	0.0076
	CX37	224	227	3	0.0134
EMD–LSTM	CS33	198	198	0	0
	CS34	176	183	7	0.0398
	CS37	167	169	2	0.0114
	CS38	201	222	21	0.1045
	CX36	191	192	1	0.0062
	CX37	224	225	1	0.0045
CEEMDAN– LSTM	CS33	198	197	1	0.0051
	CS34	176	176	0	0
	CS37	167	171	4	0.0239
	CS38	201	223	22	0.1095
	CX36	191	192	1	0.0062
	CX37	224	224	0	0

下載: 導出CSV

表 4 30%訓練集鋰電池壽命預測誤差

Table 4. Lithium battery life prediction error under 30% training set

Model	Battery	RUL_tr	RUL_pr	RUL_er	P_er
LSTM	CS33	323	346	23	0.0712
	CS34	301	325	24	0.0797
	CS37	347	527	180	0.5187
	CS38	381	408	27	0.0709
	CX36	381	385	4	0.0105
	CX37	414	419	5	0.0121
EMD–LSTM	CS33	323	324	1	0.0031
	CS34	301	304	3	0.0100
	CS37	347	354	7	0.2018
	CS38	381	408	27	0.0708
	CX36	381	430	49	0.1286
	CX37	414	414	0	0
CEEMDAN– LSTM	CS33	323	323	0	0
	CS34	301	309	8	0.0266
	CS37	347	353	6	0.0173
	CS38	381	406	25	0.0656
	CX36	381	376	5	0.0131
	CX37	414	414	0	0

下載: 導出CSV

表 5 不同算法預測精度

Table 5. Prediction accuracy of different algorithms

Battery	Training proportion/%	algorithm	RMSE/ (A·h)	MAPE	MAE/ (A·h)
CS33	50	BP	0.1471	0.1649	0.1043
		ELM	0.0650	0.0626	0.0367
		SVR	0.0297	0.0304	0.0244
		CEEMDAN–LSTM	0.0120	0.0123	0.0077
CS33	30	BP	0.1727	0.1794	0.1172
		ELM	0.1216	0.1244	0.0805
		SVR	0.0708	0.0726	0.0692
		CEEMDAN–LSTM	0.0327	0.0312	0.0200

下載: 導出CSV

中文字幕在线观看

參考文獻(26)

[1]	Li L B, Ji L, Zhu Y Z, et al. Investigation of RUL prediction of lithium-ion battery equivalent cycle battery pack. Chin J Eng, 2020, 42(6): 796 李練兵, 季亮, 祝亞尊, 等. 等效循環電池組剩余使用壽命預測. 工程科學學報, 2020, 42(6):796
[2]	Liu D T, Zhou J B, Liao H T, et al. A health indicator extraction and optimization framework for lithium-ion battery degradation modeling and prognostics. IEEE Trans Syst Man Cybern:Syst, 2015, 45(6): 915 doi: 10.1109/TSMC.2015.2389757
[3]	Pan H P, Lv Z Q, Fu B, et al. Online estimation of lithium-ion battery’s state of health using extreme learning machine. Automot Eng, 2017, 39(12): 1375 潘海鴻, 呂治強, 付兵, 等. 采用極限學習機實現鋰離子電池健康狀態在線估算. 汽車工程, 2017, 39(12):1375
[4]	Xiong R. Battery Management Algorithm for Electric Vehicles. Beijing: China Machine Press, 2020
[5]	Zhu J, Tan T X, Wu L F, et al. RUL prediction of lithium-ion battery based on improved DGWO–ELM method in a random discharge rates environment. IEEE Access, 2019, 7: 125176 doi: 10.1109/ACCESS.2019.2936822
[6]	Chen L, Chen J, Wang H M, et al. Prediction of battery remaining useful life based on wavelet packet energy entropy. Trans China Electrotech Soc, 2020, 35(8): 1827 陳琳, 陳靜, 王惠民, 等. 基于小波包能量熵的電池剩余壽命預測. 電工技術學報, 2020, 35(8):1827
[7]	Zhang Y Z, Xiong R, He H W, et al. Lithium-ion battery remaining useful life prediction with Box-Cox transformation and Monte Carlo simulation. IEEE Trans Ind Electron, 2019, 66(2): 1585 doi: 10.1109/TIE.2018.2808918
[8]	Yun Z H, Qin W H. Remaining useful life estimation of lithium-ion batteries based on optimal time series health indicator. IEEE Access, 2020, 8: 55447 doi: 10.1109/ACCESS.2020.2981947
[9]	Park K, Choi Y, Choi W J, et al. LSTM-based battery remaining useful life prediction with multi-channel charging profiles. IEEE Access, 2020, 8: 20786 doi: 10.1109/ACCESS.2020.2968939
[10]	Zhang Y Z, Xiong R, He H W, et al. Long short-term memory recurrent neural network for remaining useful life prediction of lithium-ion batteries. IEEE Trans Veh Technol, 2018, 67(7): 5695 doi: 10.1109/TVT.2018.2805189
[11]	Wei H Y, An J J, Chen J, et al. RUL prediction of lithium-ion battery based on improved particle filtering algorithm. Automot Eng, 2019, 41(12): 1377 韋海燕, 安晶晶, 陳靜, 等. 基于改進粒子濾波算法實現鋰離子電池RUL預測. 汽車工程, 2019, 41(12):1377
[12]	Yu J B. State of health prediction of lithium-ion batteries: Multiscale logic regression and Gaussian process regression ensemble. Reliab Eng Syst Saf, 2018, 174: 82 doi: 10.1016/j.ress.2018.02.022
[13]	Zhou Y P, Huang M H. Lithium-ion batteries remaining useful life prediction on a mixture of empirical mode decomposition and ARIMA model. Microelectron Reliab, 2016, 65: 265 doi: 10.1016/j.microrel.2016.07.151
[14]	Li X Y, Zhang L, Wang Z P, et al. Remaining useful life prediction for lithium-ion batteries based on a hybrid model combining the long short-term memory and Elman neural networks. J Energy Storage, 2019, 21: 510 doi: 10.1016/j.est.2018.12.011
[15]	Sayah M, Guebli D, Al Masry Z, et al. Robustness testing framework for RUL prediction Deep LSTM networks. ISA Trans, https://doi.org/10.1016/j.isatra.2020.07.003
[16]	Wang C S, Lu N Y, Wang S L, et al. Dynamic long short-term memory neural-network-based indirect remaining-useful-life prognosis for satellite lithium-ion battery. Appl Sci, 2018, 8(11): 2078 doi: 10.3390/app8112078
[17]	Bruneo D, De Vita F. On the use of LSTM networks for predictive maintenance in smart industries // 2019 IEEE International Conference on Smart Computing (SMARTCOMP). Washington, 2019: 241
[18]	Yan H R, Qin Y, Xiang S, et al. Long-term gear life prediction based on ordered neurons LSTM neural networks. Measurement, 2020, 165: 108205 doi: 10.1016/j.measurement.2020.108205
[19]	Hu T Z, Yu J B. Life prediction of lithium-ion batteries based on multiscale decomposition and deep learning. J Zhejiang Univ Eng Sci, 2019, 53(10): 1852 胡天中, 余建波. 基于多尺度分解和深度學習的鋰電池壽命預測. 浙江大學學報: 工學版, 2019, 53(10):1852
[20]	Zhang C L, He Y G, Yuan L F. Prediction approach for remaining useful life of lithium-ion battery based on EEMD and MKRVM. Proc CSU–EPSA, 2018, 30(7): 38 doi: 10.3969/j.issn.1003-8930.2018.07.006 張朝龍, 何怡剛, 袁莉芬. 基于EEMD和MKRVM的鋰電池剩余壽命預測方法. 電力系統及其自動化學報, 2018, 30(7):38 doi: 10.3969/j.issn.1003-8930.2018.07.006
[21]	Qi H M. Research on Prediction Method of Remaining Life of Lithium Battery Based on Deep Learning [Dissertation]. Harbin: Harbin Institute of Technology, 2019 齊昊明. 基于深度學習的鋰電池剩余壽命預測方法研究[學位論文]. 哈爾濱: 哈爾濱工業大學, 2019
[22]	Li J L, Li X Y, He D. A directed acyclic graph network combined with CNN and LSTM for remaining useful life prediction. IEEE Access, 2019, 7: 75464 doi: 10.1109/ACCESS.2019.2919566
[23]	Zhou Y T, Huang Y N, Pang J B, et al. Remaining useful life prediction for supercapacitor based on long short-term memory neural network. J Power Sources, 2019, 440: 227149 doi: 10.1016/j.jpowsour.2019.227149
[24]	Yu Y, Hu C H, Si X S, et al. Averaged Bi–LSTM networks for RUL prognostics with non-life-cycle labeled dataset. Neurocomputing, 2020, 402: 134 doi: 10.1016/j.neucom.2020.03.041
[25]	Yang F F, Zhang S H, Li W H, et al. State of charge estimation of lithium-ion batteries using LSTM and UKF. Energy, 2020, 201: 117664 doi: 10.1016/j.energy.2020.117664
[26]	Liu J, Chen Z Q, Huang D Y, et al. Remaining useful life of lithium-ion batteries based on time interval of equal charging voltage difference. J Shanghai Jiaotong Univ, 2019, 53(9): 1058 劉健, 陳自強, 黃德揚, 等. 基于等壓差充電時間的鋰離子電池壽命預測. 上海交通大學學報, 2019, 53(9):1058