基于LSTM&GRU-Attention多联合模型的锂离子电池SOH估计

doi:10.19799/j.cnki.2095-4239.2023.0514

摘要/Abstract

摘要：

锂离子电池的健康状态(state of health，SOH)准确估计对于储能电站的稳定高效运行至关重要。为了进一步提高数据驱动方法对SOH估计的精度，本团队提出了一种利用交叉验证训练的线性回归加权融合模型的方法。首先，从放电电压曲线、充电和放电温度曲线中提取了健康特征，并使用Pearson相关系数对所选特征进行了相关性分析，确定了网络模型输入的健康因子参数。随后，通过在LSTM与GRU中加入注意力机制，建立了LSTM-Attention与GRU-Attention模型，分别以NASA电池老化数据集B0005、B0006、B0007和B0018电池的前50%作为模型训练集，用剩余数据对模型进行验证，分别得到了模型对应的 $y^L - A$ 与 $y^G - A$ 估计值，然后使用所提融合模型方法对两个估计值进行线性回归加权，结果显示该方法的最大均方根误差和平均绝对误差分别为0.00291和0.00200。最后，为验证所提模型的抗干扰能力，在输入模型的健康因子中加入不同比例的高斯白噪声，实验结果显示融合模型的抗干扰能力较强，最大均方根误差和平均绝对误差仅为0.03562和0.02889。

关键词: 锂离子电池, 健康状态, 健康因子, LSTM-Attention, GRU-Attention, 线性回归加权法

Abstract:

The accurate estimation of the state-of-health (SOH) of lithium-ion batteries (LiBs) plays a critical role in ensuring the stable and efficient operation of energy storage systems. This study proposes a fusion model based on cross-validation-trained linear regression weighting to enhance the precision of data-driven methods for SOH estimation. First, health features are extracted from the discharge voltage curve as well the as charging and discharging temperature curves. Second, Pearson correlation coefficients are used to analyze the selected features, determining the health-indicator parameters for the network model inputs. Finally, attention mechanisms were incorporated into the long short-term memory (LSTM) and gated recurrent unit (GRU) to establish the LSTM-Attention and GRU-Attention models, respectively. These models are trained using the first 50% of data from NASA's battery aging datasets, B0005, B0006, B0007, and B0018, with the remaining 50% used for validation. The LSTM- and GRU-Attention models produce SOH estimates of $y^L - A$ and $y^G - A$ , respectively. Then, the fusion model proposed in this study performs linear regression weighting on these two estimates, yielding a maximum root mean square error (RMSE) and mean absolute error (MAE) of 0.00291 and 0.00200, respectively. Furthermore, the robustness of the proposed model is demonstrated by subjecting the health factors input to various proportions of Gaussian white noise. The results indicate that the fusion model exhibits strong resistance to interference, with a maximum RMSE and MAE of only 0.03562 and 0.02889, respectively.

Key words: lithium-ion battery, state of health, health indicator, LSTM-Attention, GRU-Attention, weighted linear regression

中图分类号:

TM 911

毛百海, 覃吴, 肖显斌, 郑宗明. 基于LSTM&GRU-Attention多联合模型的锂离子电池SOH估计[J]. 储能科学与技术, 2023, 12(11): 3519-3527.

Baihai MAO, Wu QIN, Xianbin XIAO, Zongming ZHENG. SOH estimation of lithium-ion batteries based on LSTM&GRU-Attention multijoint model[J]. Energy Storage Science and Technology, 2023, 12(11): 3519-3527.

图/表 13

表1

图1

图2

图3

表2

Pearson相关系数评价表"

项目	$F 1$	$F 2$	$F 3$	$F 4$	$F 5$	$F 6$
B0005	-0.8345	0.9962	-0.4871	0.9999	-0.4755	0.9998
B0006	-0.9573	0.9949	-0.4897	0.9999	0.3501	0.9999
B0007	-0.8362	0.9978	-0.3147	0.9997	0.0407	0.9992
B0018	-0.8997	0.9978	0.1176	0.9998	-0.1413	0.9994

表2

图4

图5

图6

图7

图8

图9

表3

图10

参考文献 24

1	中国电力企业联合会. 2022年度电化学储能电站行业统计数据[EB/OL]. [2023-03-15]. https://cec.org.cn/index.html.
2	GANDOMAN F H, JAGUEMONT J, GOUTAM S, et al. Concept of reliability and safety assessment of lithium-ion batteries in electric vehicles: Basics, progress, and challenges[J]. Applied Energy, 2019, 251: 113343.
3	SUI X, HE S, VILSEN S B, et al. A review of non-probabilistic machine learning-based state of health estimation techniques for Lithium-ion battery[J]. Applied Energy, 2021, 300: 117346.
4	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/OL]. 高电压技术: 1-14[2023-08-16]. https://doi.org/10.13336/j.1003-6520.hve.20221843.
	XIONG Q, DI Z G, JI S C. A review of research progress on health state estimation and life prediction of lithium-ion batteries[J/OL]. High Voltage Technology:1-14[2023-08-16]. https://doi.org/10.13336/j.1003-6520.hve.20221843.
5	LUO K, CHEN X, ZHENG H R, et al. A review of deep learning approach to predicting the state of health and state of charge of lithium-ion batteries[J]. Journal of Energy Chemistry, 2022, 74: 159-173.
6	魏梓轩, 韩晓娟, 李炫. 基于深度神经网络的梯次利用电池健康状态评估[J]. 太阳能学报, 2022, 43(5): 518-524.
	WEI Z X, HAN X J, LI X. State of health assessment for echelon utilization batteries based on deep neural network[J]. Acta Energiae Solaris Sinica, 2022, 43(5): 518-524.
7	VAN C N, QUANG D T. Estimation of SoH and internal resistances of Lithium ion battery based on LSTM network[J]. International Journal of Electrochemical Science, 2023, 18(6): 100166.
8	GONG Y D, ZHANG X Y, GAO D Z, et al. State-of-health estimation of lithium-ion batteries based on improved long short-term memory algorithm[J]. Journal of Energy Storage, 2022, 53: 105046.
9	GAO M Y, BAO Z Y, ZHU C X, et al. HFCM-LSTM: A novel hybrid framework for state-of-health estimation of lithium-ion battery[J]. Energy Reports, 2023, 9: 2577-2590.
10	LIN M Q, WU J, MENG J H, et al. State of health estimation with attentional long short-term memory network for lithium-ion batteries[J]. Energy, 2023, 268: 126706.
11	ZHENG Y X, HU J X, CHEN J J, et al. State of health estimation for lithium battery random charging process based on CNN-GRU method[J]. Energy Reports, 2023, 9: 1-10.
12	孙丙香, 苏晓佳, 马仕昌, 等. 基于低频阻抗谱和健康特征融合的锂离子电池健康状态主动探测方法研究[J]. 电力系统保护与控制, 2022, 50(7): 23-30.
	SUN B X, SU X J, MA S C, et al. An active detection method of li-ion battery health state based on low-frequency EIS and health feature fusion[J]. Power System Protection and Control, 2022, 50(7): 23-30.
13	ZOU C Y, CHEN X, ZHANG Y D. State of health prediction of lithium-ion batteries based on temporal degeneration feature extraction with deep cycle attention network[J]. Journal of Energy Storage, 2023, 65: 107367.
14	王凡, 史永胜, 刘博亲, 等. 基于注意力改进BiGRU的锂离子电池健康状态估计[J]. 储能科学与技术, 2021, 10(6): 2326-2333.
	WANG F, SHI Y S, LIU B Q, et al. Health state estimation of lithium-ion battery based on attention improvement BiGRU[J]. Energy Storage Science and Technology, 2021, 10(6): 2326-2333.
15	JIANG Y Y, CHEN Y, YANG F F, et al. State of health estimation of lithium-ion battery with automatic feature extraction and self-attention learning mechanism[J]. Journal of Power Sources, 2023, 556: 232466.
16	张帅涛. 基于机器学习的锂电池荷电状态及健康状态预测研究[D]. 桂林: 广西师范大学, 2022.
	ZHANG S T. Research on prediction of state of charge and health of lithium battery based on machine learning[D].Guilin: Guangxi Normal University, 2022.
17	National Aeronautics and Space Administration Prognostics Center of Excellence. PCoE Datasets [EB/OL].[2023-06-01]. https://ti.arc.nasa.gov/tech/dash/groups/pcoe/prognostic-data-repository/#battery.
18	HU X S, CHE Y H, LIN X K, et al. Battery health prediction using fusion-based feature selection and machine learning[J]. IEEE Transactions on Transportation Electrification, 2021, 7(2): 382-398.
19	DUAN W X, SONG S X, XIAO F, et al. Battery SOH estimation and RUL prediction framework based on variable forgetting factor online sequential extreme learning machine and particle filter[J]. Journal of Energy Storage, 2023, 65: 107322.
20	XU H W, WU L F, XIONG S Z, et al. An improved CNN-LSTM model-based state-of-health estimation approach for lithium-ion batteries[J]. Energy, 2023, 276: 127585.
21	PARK K, CHOI Y, CHOI W J, et al. LSTM-based battery remaining useful life prediction with multi-channel charging profiles[J]. IEEE Access, 2020, 8: 20786-20798.
22	XU P H, WANG C C, YE J L, et al. State-of-charge estimation and health prognosis for lithium-ion batteries based on temperature-compensated Bi-LSTM network and integrated attention mechanism[J]. IEEE Transactions on Industrial Electronics, 2865, PP(99): 1-11.
23	董波, 陈艾睿, 张明. 机器学习在解决过拟合现象中的作用[J]. 心理科学, 2021, 44(2): 274-281.
	DONG B, CHEN A R, ZHANG M. The role of machine learning in solving overfitting[J]. Journal of Psychological Science, 2021, 44(2): 274-281.
24	张孝远, 张金浩, 蒋亚俊. 基于改进TCN模型的动力电池健康状态评估[J]. 储能科学与技术, 2022, 11(5): 1617-1626.
	ZHANG X Y, ZHANG J H, JIANG Y J. Power battery health evaluation based on improved TCN model[J]. Energy Storage Science and Technology, 2022, 11(5): 1617-1626.

项目	充电			放电		其他
项目	电流/A	电压/V	截止电流/mA	恒流/A	截止电压/V	初始容量/Ah
B0005	1.5	4.2	20	2	2.7	1.86
B0006	1.5	4.2	20	2	2.5	2.04
B0007	1.5	4.2	20	2	2.2	1.89
B0018	1.5	4.2	20	2	2.5	1.86

评价指标	模型	电池
评价指标	模型	B0005	B0006	B0007	B0018
MAE	LSTM	0.02526	0.01629	0.01468	0.01550
	GRU	0.02143	0.01305	0.00947	0.01462
	LSTM-Attention	0.01019	0.00807	0.00723	0.01142
	GRU-Attention	0.00757	0.00672	0.00609	0.00840
	线性回归加权法	0.00079	0.00200	0.00138	0.00184
RMSE	LSTM	0.02911	0.02012	0.01545	0.01669
	GRU	0.02368	0.01349	0.01047	0.01523
	LSTM-Attention	0.01026	0.00903	0.00892	0.01360
	GRU-Attention	0.00880	0.00772	0.00777	0.00998
	线性回归加权法	0.00146	0.00291	0.00232	0.00218

[1]	李悦, 王博, 吴楠. 石墨烯/Si/SiO _x 纳米复合材料的制备及储锂性能研究[J]. 储能科学与技术, 2023, 12(9): 2752-2759.
[2]	曾少鸿, 吴伟雄, 刘吉臻, 汪双凤, 叶石丰, 冯振宇. 锂离子电池浸没式冷却技术研究综述[J]. 储能科学与技术, 2023, 12(9): 2888-2903.
[3]	李纪伟, 刘睿涵, 吕桃林, 潘隆, 马常军, 李清波, 赵芝芸, 杨文, 解晶莹. 基于局部离群点检测和标准差方法的锂离子电池组早期故障诊断[J]. 储能科学与技术, 2023, 12(9): 2917-2926.
[4]	黄晓伟, 李少鹏, 张校刚. 负极补锂锂化裕度对电芯性能的影响及机理研究[J]. 储能科学与技术, 2023, 12(9): 2727-2734.
[5]	陈欣, 李云伍, 梁新成, 李法霖, 张志冬. 基于模态分解的Transformer-GRU联合电池健康状态估计[J]. 储能科学与技术, 2023, 12(9): 2927-2936.
[6]	江婉薇, 梁呈景, 钱历, 刘梅城, 朱孟想, 马骏. 锡基三维石墨烯泡沫调控及其锂电池负极性能[J]. 储能科学与技术, 2023, 12(9): 2746-2751.
[7]	高欣, 王若谷, 高文菁, 邓泽军, 梁睿祺, 杨騉. 基于运行数据的储能电站电池组一致性评估方法[J]. 储能科学与技术, 2023, 12(9): 2937-2945.
[8]	周向阳, 胡颖杰, 梁家浩, 周其杰, 文康, 陈松, 杨娟, 唐晶晶. 天然鳞片石墨球化尾料的高性能负极材料制备及储锂特性研究[J]. 储能科学与技术, 2023, 12(9): 2767-2777.
[9]	官亦标, 沈进冉, 刘家亮, 渠展展, 高飞, 刘施阳, 郭翠静, 周淑琴, 付珊珊. 以安全高质量应用为导向的储能锂离子电池综合性能评价标准[J]. 储能科学与技术, 2023, 12(9): 2946-2953.
[10]	郭煜, 王亦伟, 钟隽, 杜进桥, 田杰, 李艳, 蒋方明. 基于增量容量曲线的锂离子电池微内短路故障诊断方法[J]. 储能科学与技术, 2023, 12(8): 2536-2546.
[11]	唐程波, 锁要红, 何昭坤. 基于正弦函数的液冷板上流体流向对锂离子电池散热性能的影响[J]. 储能科学与技术, 2023, 12(8): 2547-2555.
[12]	左安昊, 方儒卿, 李哲. 锂离子电池单颗粒动力学表征方法综述[J]. 储能科学与技术, 2023, 12(8): 2457-2481.
[13]	陈满, 程志翔, 赵春朋, 彭鹏, 雷旗开, 金凯强, 王青松. 锂离子电池储能集装箱爆炸危害数值模拟[J]. 储能科学与技术, 2023, 12(8): 2594-2605.
[14]	黄永浩, 臧国景, 朱霨亚, 廖友好, 李伟善. LiF添加剂改善含锂陶瓷隔膜与4.35 V LiNi_0.8Co_0.1Mn_0.1O₂ 正极的界面稳定性[J]. 储能科学与技术, 2023, 12(8): 2361-2369.
[15]	杨佳兴, 张恒运, 徐屹东. 基于电化学-热耦合模型的锂离子电池组件产热分析[J]. 储能科学与技术, 2023, 12(8): 2615-2625.