基于RUN-GRU-attention模型的实车动力电池健康状态估计方法

doi:10.19799/j.cnki.2095-4239.2024.0576

摘要/Abstract

摘要：

实车动力电池的健康状态(state of health，SOH)评估存在数据质量差、工况不统一、数据利用率低等问题，本文面向阶梯倍率充电工况构建多源特征提取及SOH估计模型。首先，通过数据清洗、切割、填充，获取独立的充电片段；其次，基于不同电流阶段计算容量，实现原始数据利用率达96.9%，并与单独限定SOC范围计算容量的方法相比，误差降低48.1%以上；然后，从当前工况、历史累积两个维度提取多个健康因子，对于当前工况特征值，通过灰色关联度及干扰性随机森林重要度分析双重筛选。对于历史累积特征值，利用Spearson相关性分析和核主成分分析方法(kernel principal component analysis，KPCA)降低信息冗余；最后，对门控循环单元网络模型(gated recurrent unit，GRU)引入注意力机制和龙格库塔优化算法(Runge Kutta optimizer，RUN)，建立RUN-GRU-attention模型，基于实车运行数据集与现有5种模型进行对比，实验结果表明，无论是包含单阶段还是多阶段电流的测试样本，优化模型的估计精度更佳，误差不高于0.0086，并且随着充电循环次数增加表现出良好的误差收敛性，可有效预测SOH波动趋势。

关键词: 实车动力电池, 阶梯倍率充电, 健康状态估计, 多源特征提取, 龙格库塔优化算法, 机器学习

Abstract:

The evaluation of the state of health (SOH) for real-vehicle batteries is challenging owing to poor data quality, inconsistent operating conditions, and limited data utilization. This paper presents a multisource feature extraction and SOH estimation model specifically designed for step-rate charging conditions. First, charging segments are obtained through data cleaning, segmenting, and filling processes. Next, capacity is calculated using data from various current stages, achieving a raw data utilization rate of 96.9%. Compared to methods that calculate capacity within a restricted state of charge (SOC) range, this approach reduces error by over 48.1%. Finally, health factors are extracted based on current operating conditions and historical data accumulation. For current operating condition feature values, dual screening is performed using grey correlation analysis and random forest importance analysis to manage interference. For historical cumulative feature values, Spearman correlation analysis and Kernel Principal Component Analysis (KPCA) are employed to reduce information redundancy. Finally, an attention mechanism and Runge-Kutta optimizer (RUN) are integrated into the Gated Recurrent Unit (GRU) network model. The performance of this optimized model is then compared with five existing models using an actual vehicle operation dataset. The experimental results demonstrate that the optimized model achieves superior estimation accuracy, with an error margin of no more than 0.0086, regardless of whether the test samples include single-stage or multi-stage currents. Additionally, the model shows excellent error convergence as the number of charging cycles increases and effectively predicts the trend of SOH fluctuations.

Key words: real-vehicle battery, step rate charging, SOH estimation, multisource features extraction, Runge Kutta optimization algorithm, machine learning

中图分类号:

TM 911

刘定宏, 董文楷, 李召阳, 张红烛, 齐昕. 基于RUN-GRU-attention模型的实车动力电池健康状态估计方法[J]. 储能科学与技术, 2024, 13(9): 3042-3058.

Dinghong LIU, Wenkai DONG, Zhaoyang LI, Hongzhu ZHANG, Xin QI. Estimation of real-vehicle battery state of health using the RUN-GRU-attention model[J]. Energy Storage Science and Technology, 2024, 13(9): 3042-3058.

图/表 24

图1

图2

图3

图4

图5

表1

表2

图6

图7

图8

图9

图10

表3

图11

图12

图13

图14

表4

图15

表5

图16

表6

表7

图17

参考文献 30

1	熊庆, 邸振国, 汲胜昌. 锂离子电池健康状态估计及寿命预测研究进展综述[J]. 高电压技术, 2024, 50(3): 1182-1195. DOI: 10.13336/j.1003-6520.hve.20221843.
	XIONG Q, DI Z G, JI S C. Review on health state estimation and life prediction of lithium-ion batteries[J]. High Voltage Engineering, 2024, 50(3): 1182-1195. DOI: 10.13336/j.1003-6520.hve.20221843.
2	卢婷, 杨文强. 锂离子电池全生命周期内评估参数及评估方法综述[J]. 储能科学与技术, 2020, 9(3): 657-669. DOI: 10.19799/j.cnki. 2095-4239.2019.0263.
	LU T, YANG W Q. Review of evaluation parameters and methods of lithium batteries throughout its life cycle[J]. Energy Storage Science and Technology, 2020, 9(3): 657-669. DOI: 10.19799/j.cnki.2095-4239.2019.0263.
3	LIU W, XU Y. A comprehensive review of health indicators of Li-ion battery for online state of health estimation[C]// 2019 IEEE 3rd Conference on Energy Internet and Energy System Integration (EI2). IEEE, 2019: 1203-1208. DOI: 10.1109/EI247390.2019.9062037.
4	戴彦文, 于艾清. 基于健康特征参数的CNN-LSTM&GRU组合锂电池SOH估计[J]. 储能科学与技术, 2022, 11(5): 1641-1649. DOI: 10.19799/j.cnki.2095-4239.2021.0623.
	DAI Y W, YU A Q. Combined CNN-LSTM and GRU based health feature parameters for lithium-ion batteries SOH estimation[J]. Energy Storage Science and Technology, 2022, 11(5): 1641-1649. DOI: 10.19799/j.cnki.2095-4239.2021.0623.
5	刘轶鑫, 张頔, 李雪, 等. Research on SOH estimation method based on SOC-OCV curve characteristics[C]// 中国汽车工程学会(China Society of Automotive Engineers). 2019中国汽车工程学会年会论文集(2), 2019: 1.
	LIU Y X, ZHANG D, LI X, et al. Research on SOH estimation method based on SOC-OCV curve characteristics[C]// China Society of Automotive Engineers. Proceedings of the 2019 Annual Meeting of the Society of Automotive Engineers of China(2), 2019: 1.
6	梁新成, 张勉, 黄国钧. 基于BMS的锂离子电池建模方法综述[J]. 储能科学与技术, 2020, 9(6): 1933-1939. DOI: 10.19799/j.cnki.2095-4239.2020.0166.
	LIANG X C, ZHANG M, HUANG G J. Review on lithium-ion battery modeling methods based on BMS[J]. Energy Storage Science and Technology, 2020, 9(6): 1933-1939. DOI: 10.19799/j.cnki.2095-4239.2020.0166.
7	俞志骏, 安斯光, 汪伟. 锂电池SOC和SOH的自适应联合在线估算方法[J]. 计算机应用与软件, 2023, 40(10): 142-149. DOI: 10.3969/j.issn.1000-386x.2023.10.022.
	YU Z J, AN S G, WANG W. An adaptive joint online estimation method of lithium battery SOC and SOH[J]. Computer Applications and Software, 2023, 40(10): 142-149. DOI: 10.3969/j.issn.1000-386x.2023.10.022.
8	SUN F C, XIONG R. A novel dual-scale cell state-of-charge estimation approach for series-connected battery pack used in electric vehicles[J]. Journal of Power Sources, 2015, 274: 582-594. DOI: 10.1016/j.jpowsour.2014.10.119.
9	HE H W, XIONG R, FAN J X. Evaluation of lithium-ion battery equivalent circuit models for state of charge estimation by an experimental approach[J]. Energies, 2011, 4(4): 582-598. DOI: 10.3390/en4040582.
10	颜湘武, 邓浩然, 郭琪, 等. 基于自适应无迹卡尔曼滤波的动力电池健康状态检测及梯次利用研究[J]. 电工技术学报, 2019, 34(18): 3937-3948. DOI: 10.19595/j.cnki.1000-6753.tces.171452.
	YAN X W, DENG H R, GUO Q, et al. Study on the state of health detection of power batteries based on adaptive unscented Kalman filters and the battery echelon utilization[J]. Transactions of China Electrotechnical Society, 2019, 34(18): 3937-3948. DOI: 10.19595/j.cnki.1000-6753.tces.171452.
11	徐东辉. 锂电池一阶RC等效电路模型的动力学特性分析[J]. 电源技术, 2021, 45(11): 1448-1452. DOI: 10.3969/j.issn.1002-087X. 2021.11.018.
	XU D H. Dynamic analysis of first-order RC equivalent circuit model for automotive lithium ion power batteries[J]. Chinese Journal of Power Sources, 2021, 45(11): 1448-1452. DOI: 10. 3969/j.issn.1002-087X.2021.11.018.
12	高仁璟, 吕治强, 赵帅, 等. 基于电化学模型的锂离子电池健康状态估算[J]. 北京理工大学学报, 2022, 42(8): 791-797. DOI: 10.15918/j.tbit1001-0645.2021.310.
	GAO R J, LYU Z Q, ZHAO S, et al. Health state estimation of Li-ion batteries based on electrochemical model[J]. Transactions of Beijing Institute of Technology, 2022, 42(8): 791-797. DOI: 10.15918/j.tbit1001-0645.2021.310.
13	GAO R J, LYU Z Q, ZHAO S, et al. Estimation of state of health of lithium-ion battery based on electrochemical model [J]. Transactions of Beijing institute of Technology, 2022, 42 (08): 791-797. DOI: 10.15918/j.tbit1001-0645.2021.310.
14	PRADA E, BERNARD J, SAUVANT-MOYNOT V. Ni-MH battery ageing: From comprehensive study to electrochemical modelling for state-of charge and state-of-health estimation[J]. IFAC Proceedings Volumes, 2009, 42(26): 123-131. DOI: 10.3182/20091130-3-FR-4008.00017.
15	陆楠, 孙越, 彭鹏, 等. 数据驱动的钠离子电池健康状态评估方法研究[J]. 电源学报, 2024, 22(1): 1-10. DOI: 10.13234/j.issn.2095-2805.2024.1.1.
	LU N, SUN Y, PENG P, et al. Data-driven state of health estimation for sodium-ion batteries[J]. Journal of Power Supply, 2024, 22(1): 1-10. DOI: 10.13234/j.issn.2095-2805.2024.1.1.
16	GUO P Y, CHENG Z, YANG L. A data-driven remaining capacity estimation approach for lithium-ion batteries based on charging health feature extraction[J]. Journal of Power Sources, 2019, 412: 442-450. DOI: 10.1016/j.jpowsour.2018.11.072.
17	陈吉清, 李子涵, 兰凤崇, 等. 基于非线性降维IC特征的实车电池SOH估计[J]. 汽车工程, 2023, 45(2): 199-208. DOI: 10.19562/j.chinasae.qcgc.2023.02.005.
	CHEN J Q, LI Z H, LAN F C, et al. Real-vehicle battery health state estimation based on nonlinear reduced-dimensional IC features[J]. Automotive Engineering, 2023, 45(2): 199-208. DOI: 10.19562/j.chinasae.qcgc.2023.02.005.
18	SCHIFFER Z J, CANNARELLA J, ARNOLD C B. Strain derivatives for practical charge rate characterization of lithium ion electrodes[J]. Journal of the Electrochemical Society, 2015, 163(3): A427-A433. DOI: 10.1149/2.0091603jes.
19	WENG C H, CUI Y J, SUN J, et al. On-board state of health monitoring of lithium-ion batteries using incremental capacity analysis with support vector regression[J]. Journal of Power Sources, 2013, 235: 36-44. DOI: 10.1016/j.jpowsour.2013.02.012.
20	RICHARDSON R R, OSBORNE M A, HOWEY D A. Battery health prediction under generalized conditions using a Gaussian process transition model[J]. Journal of Energy Storage, 2019, 23: 320-328. DOI: 10.1016/j.est.2019.03.022.
21	陈媛, 段文献, 何怡刚, 等. 带降噪自编码器的锂离子电池健康状态估计算法 [J/OL]. 电工技术学报, 1-17[2024-07-20]. https://doi.org/10.19595/j.cnki.1000-6753.tces.231644.
	CHEN Y, DUAN W X, HE Y G, et al. Lithium ion battery health state estimation algorithm with denoising autoencoder [J/OL]. Journal of Electrical Technology, 1-17 [2022-07-20]. https://doi.org/10.19595/j.cnki.1000-6753.tces.231644.
22	赵旭. 基于数据挖掘的电动汽车电池健康状态估计研究[D]. 成都: 西南交通大学, 2022. DOI: 10.27414/d.cnki.gxnju.2022.001673.
	ZHAO X. Research on battery health state estimation of electric vehicle based on data mining[D]. Chengdu: Southwest Jiaotong University, 2022. DOI: 10.27414/d.cnki.gxnju.2022.001673.
23	MA Y, SHAN C, GAO J W, et al. A novel method for state of health estimation of lithium-ion batteries based on improved LSTM and health indicators extraction[J]. Energy, 2022, 251: 123973. DOI: 10.1016/j.energy.2022.123973.
24	LIU Q, ZHENG M, LI P K. A review of data-driven SOH and RUL estimation for lithium-ion batteries[C]// 2023 42nd Chinese Control Conference (CCC). IEEE, 2023: 8769-8774.
25	RAMANI P, VAGHELA R, POPAT Y, et al. Lithium-ion battery health prediction using a GRU model with protective characteristic[C]//2023 International Conference on Modeling, Simulation & Intelligent Computing (MoSICom). IEEE, 2023: 521-525. DOI: 10.1109/MoSICom59118.2023.10458820.
26	LI Z C, CHEN D W, LU J J, et al. Li-ion battery SOH estimation based on BO-SVR model[C]// 2023 IEEE 7th Conference on Energy Internet and Energy System Integration (EI2). IEEE, 2023: 3737-3742. DOI: 10.1109/EI259745.2023.10513320.
27	OJI T, ZHOU Y L, CI S, et al. Data-driven methods for battery SOH estimation: Survey and a critical analysis[J]. IEEE Access, 2021, 9: 126903-126916. DOI: 10.1109/ACCESS.2021.3111927.
28	朱振宇, 高德欣. 基于混合网络的锂离子电池健康状态与剩余使用寿命联合估计方法[J]. 信息与控制, 2024, 53(1): 120-128. DOI: 10.13976/j.cnki.xk.2023.2378.
	ZHU Z Y, GAO D X. Joint estimation method of state of health and remaining useful life for lithium-ion batteries based on hybrid networks[J]. Information and Control, 2024, 53(1): 120-128. DOI: 10.13976/j.cnki.xk.2023.2378.
29	SUN C B, QIN W H, YUN Z H. A state-of-health estimation method for lithium batteries based on fennec fox optimization algorithm-mixed extreme learning machine[J]. Batteries, 2024, 10(3): 87. DOI: 10.3390/batteries10030087.
30	JAVAID M U, SEO J, SUH Y K, et al. Battery state of health estimation from discharge voltage segments using an artificial neural network[J]. International Journal of Precision Engineering and Manufacturing-Green Technology, 2024, 11(3): 863-876. DOI: 10.1007/s40684-024-00602-2.

方案	样本数目	极差	四方位差	与累积里程的回归模型	R²
ΔSOC≥15	291	0.334	0.0666	y=-1.62×10^-7x+1.4738	0.4185
ΔSOC≥30	233	0.2665	0.0621	y=-1.59×10^-7x+1.4653	0.3817
ΔSOC≥45	173	0.2665	0.0578	y=-1.77×10^-7x+1.4806	0.4395
ΔSOC≥60	119	0.1858	0.0557	y=-1.74×10^-7x+1.4690	0.5100
Ⅰ阶段	148	0.1653	0.0495	y=-1.73×10^-7x+1.2994	0.6985
Ⅱ阶段	170	0.1747	0.0572	y=-2.46×10^-7x+1.3155	0.6367
Ⅲ阶段	83	0.4379	0.1782	—	—
Ⅳ阶段	52	1.7872	0.1819	—	—

样本	HF1	HF2	HF3	HF4	HF5	HF6	HF7	HF8	HF9
关联系数-Ⅰ样本	0.7868	0.7893	0.8003	0.7706	0.8865	0.8839	0.8905	0.8354	0.9335
关联系数-Ⅱ样本	0.7662	0.8087	0.7962	0.6677	0.9588	0.9545	0.9640	0.7850	0.8993

方案序号	初始学习率	正则化系数	隐藏层节点数
方案1	0.003	1×10^-6	20
方案2	0.005	1×10^-6	20
方案3	0.008	1×10^-6	20
方案4	0.01	1×10^-6	20
方案5	0.005	5×10^-7	20
方案6	0.005	5×10^-5	20
方案7	0.005	5×10^-3	20
方案8	0.005	5×10^-1	20
方案9	0.005	5×10^-5	10
方案10	0.005	5×10^-5	30
方案11	0.005	5×10^-5	50
方案12	0.005	5×10^-5	70

模型类型	容量样本计算	总样本数目/个	MAE	MAPE	RSME	R²
方案1	限制SOC在60%～75%(ΔSOC=15%)	181	0.0159	0.0181	0.0192	0.6225
方案2	限制SOC在35%～65%(ΔSOC=30%)	149	0.0137	0.0148	0.0165	0.6816
方案3	限制SOC在30%～90%(vSOC=60%)	98	0.0163	0.0198	0.0217	0.5799
方案4	本文提出容量计算方法	318	0.0071	0.0075	0.0084	0.9210

数据项目		1#数据集	2#数据集
累计行驶里程		58853.7 km	150513.3 km
阶梯电流大小		110 A-97 A-46 A-14 A	115 A-102 A-44 A-13 A
有效充电样本数目		318(包含Ⅰ阶段148个，Ⅱ阶段170个)	1587(包含Ⅰ阶段853,Ⅱ阶段734个)
attention层		自注意力模式，注意力头数为2，键-查询-值通道数为2
隐藏节点数，初始学习率，正则化系数最优组合	RUN	(52，1.0577×10^-6，0.0771)	(58，1.3682×10^-6，0.0594)
	SMA	(56，0.9493×10^-6，0.0596)	(49，1.1034v，0.0326)
	AVOA	(31，8.0211×10^-7，0.0042)	(42，0.9213×10^-6，0.0467)
LSTM/GRU网络		最大训练次数300，梯度阈值为0.81，自适应学习率为分段常数衰减模式学习率下降周期为120
运行环境		AMD Ryzen 77840H with Radeon 780M Graphics 3.80 GHz 处理器和32.00GB RAM 的计算机，MATLAB R2023a