基于滑动窗口和LSTM神经网络的锂离子电池建模方法

doi:10.19799/j.cnki.2095-4239.2021.0373

摘要/Abstract

摘要：

为提高锂离子电池在复杂工况下的预测能力和建模精度，提出一种基于滑动窗口和长短时记忆(long short term memory，LSTM)神经网络的锂离子电池建模方法。首先建立了基于神经网络的锂离子电池模型，确定了神经网络的基本结构，通过LSTM层、向量拼接层和全连接层分别实现了时序特征提取、特征融合和回归预测。然后提出了滑动窗口的输入向量处理方法，滑动窗口每次向前推进一个时间点，通过限制时间窗口内所能处理的最大信元数对数据量进行限制，为多个LSTM层的并行计算和深隐层的拼接层和全连接层预留了计算量的裕度，实现了对模型中循环网络层深度的优化选择。为解决模型在多工况下运行的泛化问题，提出使用离线数据集的预训练和在线数据的参数修正的训练方法，通过大量离线数据集的反复训练，使模型学习电池的共性部分；再使用部分在线数据，对网络参数进行调整，将其应用于预测中。最后使用恒流/恒压、随机电流脉冲、大功率脉冲等多个工况的数据分别进行测试。结果表明，基于长短时记忆神经网络的建模方法能够准确预测电池输出电压和荷电状态。

关键词: 锂离子电池, 模型, 神经网络, 长短时间记忆, 多时序特征提取, 滑动窗口

Abstract:

This study proposes a lithium-ion battery model based on the sliding window and long short-term memory (LSTM) neural network to improve the model accuracy under complex working conditions. First, a lithium-ion battery model based on the LSTM neural network is established. Next, the basic structure of the neural network is determined. The time series feature extraction, feature fusion, and regression prediction are realized by combining the LSTM, vector splicer, and full connection layers. A sliding window input vector processing method is then proposed. The sliding window is advanced one time point at a time, and the data volume is limited by restricting the maximum number of letter elements within the time window. A computational margin is reserved for the parallel computation of the multiple LSTM layers and the deep hidden layers of the splicing and fully connected layers. Subsequently, the optimal selection of the depth of the recurrent network layer in the model is achieved. A training method using offline data set pre-training and online data parameter modification is proposed to solve the generalization problem under various complex working conditions. The model learns the common parts of the battery through the repetitive training of a large number of offline data sets. The network parameters are adjusted and used in the prediction by using a part of the online data. Finally, the datasets of the constant current/constant voltage, random current pulse, high-power pulse, and other working condition test profiles are applied for validation. The results show that the proposed modeling method can accurately predict the battery's output voltage and state of charge.

Key words: lithium-ion battery, battery model, neural network, long short term memory, feature extraction, sliding window

中图分类号:

TM 912

张少凤, 张清勇, 杨叶森, 苏义鑫, 熊斌宇. 基于滑动窗口和LSTM神经网络的锂离子电池建模方法[J]. 储能科学与技术, 2022, 11(1): 228-239.

Shaofeng ZHANG, Qingyong ZHANG, Yesen YANG, Yixin SU, Binyu XIONG. Lithium-ion battery model based on sliding window and long short term memory neural network[J]. Energy Storage Science and Technology, 2022, 11(1): 228-239.

图/表 15

图1

图2

表1

表2

图3

表3

图4

图5

图6

图7

表4

图8

图9

图10

表5

参考文献 26

1	ZHAO E W, LIU T, JÓNSSON E, et al. In situ NMR metrology reveals reaction mechanisms in redox flow batteries[J]. Nature, 2020, 579(7798): 224-228.
2	范兴明, 王超, 张鑫, 等. 基于增量学习相关向量机的锂离子电池SOC预测方法[J]. 电工技术学报, 2019, 34(13): 2700-2708.
	FAN X M, WANG C, ZHANG X, et al. A prediction method of Li-ion batteries SOC based on incremental learning relevance vector machine[J]. Transactions of China Electrotechnical Society, 2019, 34(13): 2700-2708.
3	陈翌, 白云飞, 何瑛. 数据驱动的锂电池健康状态估算方法比较[J]. 储能科学与技术, 2019, 8(6): 1204-1210.
	CHEN Y, BAI Y F, HE Y. Comparison of data-driven lithium battery state of health estimation methods[J]. Energy Storage Science and Technology, 2019, 8(6): 1204-1210.
4	黄凯, 郭永芳, 李志刚. 基于信息反馈粒子群的高精度锂离子电池模型参数辨识[J]. 电工技术学报, 2019, 34(S1): 378-387.
	HUANG K, GUO Y F, LI Z G. High precision parameter identification of lithium-ion battery model based on feedback particle swarm optimization algorithm[J]. Transactions of China Electrotechnical Society, 2019, 34(S1): 378-387.
5	张卓识. 锂离子电池建模与故障预测方法研究[D]. 大连: 大连海事大学, 2016.
	ZHANG Z S. Study of lithium-ion battery modeling and prognostics method[D]. Dalian: Dalian Maritime University, 2016.
6	凡旭国. 锂离子电池组等效电路建模及SOC估算的研究[D]. 绵阳: 西南科技大学, 2017.
	FAN X G. Research on equivalent circuit modeling and state of charge estimation of lithium-ion battery[D]. Mianyang: Southwest University of Science and Technology, 2017.
7	杨帆, 乔艳龙, 甘德刚, 等. 不同充电模式对锂离子电池极化特性影响[J]. 电工技术学报, 2017, 32(12): 171-178.
	YANG F, QIAO Y L, GAN D G, et al. Lithium-ion battery polarization characteristics at different charging modes[J]. Transactions of China Electrotechnical Society, 2017, 32(12): 171-178.
8	郑志坤, 赵光金, 金阳, 等. 基于库仑效率的退役锂离子动力电池储能梯次利用筛选[J]. 电工技术学报, 2019, 34(S1): 388-395.
	ZHENG Z K, ZHAO G J, JIN Y, et al. The reutilization screening of retired electric vehicle lithium-ion battery based on coulombic efficiency[J]. Transactions of China Electrotechnical Society, 2019, 34(S1): 388-395.
9	郑继利. 基于机理模型的锂离子电池电化学行为及优化研究[D]. 哈尔滨: 哈尔滨工业大学, 2018.
	ZHENG J L. Research on electrochemical performance and optimization of lithium ion batterries based on the mechanism model[D]. Harbin: Harbin Institute of Technology, 2018.
10	LUZI M, PASCHERO M, RIZZI A, et al. A novel neural networks ensemble approach for modeling electrochemical cells[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30(2): 343-354.
11	HUSSEIN A A. Adaptive artificial neural network-based models for instantaneous power estimation enhancement in electric vehicles' Li-ion batteries[J]. IEEE Transactions on Industry Applications, 2019, 55(1): 840-849.
12	CHEMALI E, KOLLMEYER P J, PREINDL M, et al. State-of-charge estimation of Li-ion batteries using deep neural networks: A machine learning approach[J]. Journal of Power Sources, 2018, 400: 242-255.
13	DAI H D, ZHAO G C, LIN M Q, et al. A novel estimation method for the state of health of lithium-ion battery using prior knowledge-based neural network and Markov chain[J]. IEEE Transactions on Industrial Electronics, 2019, 66(10): 7706-7716.
14	LI W H, JIAO Z P, DU L, et al. An indirect RUL prognosis for lithium-ion battery under vibration stress using Elman neural network[J]. International Journal of Hydrogen Energy, 2019, 44(23): 12270-12276.
15	SHEN S, SADOUGHI M, CHEN X Y, et al. A deep learning method for online capacity estimation of lithium-ion batteries[J]. Journal of Energy Storage, 2019, 25: doi: 10.1016/j.‍est.‍2019. 100817.
16	ZHAO R X, KOLLMEYER P J, LORENZ R D, et al. A compact methodology via a recurrent neural network for accurate equivalent circuit type modeling of lithium-ion batteries[J]. IEEE Transactions on Industry Applications, 2019, 55(2): 1922-1931.
17	李超然, 肖飞, 樊亚翔. 基于循环神经网络的锂电池SOC估算方法[J]. 海军工程大学学报, 2019, 31(6): 107-112.
	LI C R, XIAO F, FAN Y X. Approach to lithium battery SOC estimation based on recurrent neural network[J]. Journal of Naval University of Engineering, 2019, 31(6): 107-112.
18	WU Z Q, SHANG M Y, SHEN D D, et al. SOC estimation for batteries using MS-AUKF and neural network[J]. Journal of Renewable and Sustainable Energy, 2019, 11(2): doi: 10.1063/1.5064479.
19	TONG S J, LACAP J H, PARK J W. Battery state of charge estimation using a load-classifying neural network[J]. Journal of Energy Storage, 2016, 7: 236-243.
20	BIAN C, HE H L, YANG S K, et al. State-of-charge sequence estimation of lithium-ion battery based on bidirectional long short-term memory encoder-decoder architecture[J]. Journal of Power Sources, 2020, 449: doi: 10.1016/j.jpowsour.2019.227558.
21	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[J]. IEEE Transactions on Vehicular Technology, 2018, 67(7): 5695-5705.
22	LIU V T, SUN Y K, LU H Y, et al. State of charge estimation for lithium-ion battery using recurrent neural network[C]//2018 IEEE International Conference on Advanced Manufacturing (ICAM), 2018: 376-379.
23	赵明愿. 新型时间序列复杂网络建网及在风序列分析中的应用研究[D]. 天津: 天津大学, 2017.
	ZHAO M Y. Time series analysis based on complex network and application research in wind records[D]. Tianjin: Tianjin University, 2017.
24	刘琴, 王恺乐, 饶卫雄. 不等长时间序列滑窗STS距离聚类算法[J]. 计算机科学与探索, 2015, 9(11): 1301-1313.
	LIU Q, WANG K L, RAO W X. Non-equal time series clustering algorithm with sliding window STS distance[J]. Journal of Frontiers of Computer Science and Technology, 2015, 9(11): 1301-1313.
25	黄喆. 基于RBF神经网络的上证指数预测研究[D]. 合肥: 中国科学技术大学, 2009.
	HUANG Z. Prediction of Shanghai stock exchange composite index based on RBF neural network[D]. Hefei: University of Science and Technology of China, 2009.
26	艾虎, 李菲. 基于改进的长短期记忆神经网络方言辨识模型[J]. 科学技术与工程, 2019, 19(2): 163-169.
	AI H, LI F. Dialect identification model based on improved long short-term memory neural network[J]. Science Technology and Engineering, 2019, 19(2): 163-169.

预测目标	层定义	神经元数	激活函数
SOC	隐层1	2L_tw	tanh
	隐层2	L_tw	tanh
	输出层	L_tw	linear
电压	隐层	2L_tw	tanh
电压	输出层	L_tw	linear

编号	工况	来源
1	电池1恒流恒压充放电	实验平台
2	电池2恒流恒压充放电	实验平台
3	低温随机电流脉冲	NASA RW10
4	高温随机电流脉冲	NASA RW23
5	平抑波动的随机脉冲	Simulink模型
6	大倍率脉冲放电	Simulink模型
7	温度多变环境	Simulink模型

性能	数值
性能	镍钴锰酸锂 (HLY18650-2500)	钴酸锂 (ICR18650/15P)
额定容量/(mA·h)	2500	1500
最小容量/(mA·h)	2500	1500
工作电压/V	2.75～4.20	2.50～4.20
内阻/Ω	≤60	≤18
放电温度/℃	-20～60	-20～60

训练方法	Step1		Step2		Step3		Step4		Step5
训练方法	SOC	电压	SOC	电压	SOC	电压	SOC	电压	SOC	电压
离线	32.1×10^-4	14×10^-4	25.1×10^-4	14.2×10^-4	31.4×10^-4	14.4×10^-4	28.6×10^-4	14.5×10^-4	30×10^-4	14.7×10^-4
在线	8.95×10^-4	22×10^-4	9.61×10^-4	22.1×10^-4	8.54×10^-4	22.1×10^-4	9.22×10^-4	22.2×10^-4	9.31×10^-4	22.5×10^-4
离线+在线	1.71×10^-4	12.3×10^-4	1.32×10^-4	12.3×10^-4	1.31×10^-4	12.4×10^-4	1.50×10^-4	12.4×10^-4	1.21×10^-4	12.7×10^-4

模型	平抑波动		大倍率放电		温度多变
模型	SOC	电压	SOC	电压	SOC	电压
MLP1	4.9×10^-5	1.4×10^-5	1.37×10^-4	1.16×10^-3	2.01×10^-3	1.53×10^-4
MLP2	2.11×10^-4	1.85×10^-4	2.94×10^-3	1.48×10^-3	8.4×10^-3	3.62×10^-4
提出的模型	9.52×10^-4	8.57×10^-5	6.66×10^-3	1.97×10^-2	1.30×10^-2	1.81×10^-3