一种改进的NARX神经网络的锂离子电池荷电状态估算

doi:10.19799/j.cnki.2095-4239.2021.0135

摘要/Abstract

摘要：

准确估算锂离子电池的荷电状态（state of charge，SOC）是电动汽车电池管理系统的关键技术之一。文中利用动态非线性自回归模型（nonlinear autoregressive with exogeneous， NARX）来完成SOC的估算。由于NARX模型在训练过程中可能会出现梯度爆炸和梯度消失，导致参数易陷入局部最优，从而影响SOC的估算精度。因此，为了提高模型的估算能力，文中提出了一种新颖的NARX模型。长短期记忆算法（long short-term memories ，LSTM）用来保存NARX模型输出的重要信息，并与当前时刻测量的电池参数共同作为模型的输入向量，完成SOC估算。实验结果表明，所提模型估算的SOC误差控制在1%以内，验证了所提算法的有效性。

关键词: 锂离子电池, 动态非线性自回归模型（nonlinear autoregressive with exogeneous， NARX）, 长短期记忆模型（long short-term memories ，LSTM）, 荷电状态（state of charge，SOC）

Abstract:

Accurate estimation of state of charge (SOC) of lithium-ion battery is one of the key parameters in battery management system of electric vehicles. The dynamic nonlinear autoregressive with abnormal (NARX) model is used to estimate SOC in this paper. Because of the NARX model may appear gradient explosion and gradient disappearance in the training process, the parameters are easy to fall into local optimum, which affects the estimation accuracy of SOC. Therefore, in order to improve the estimation ability of the model, a novel NARX model is proposed. The long short-term memories (LSTM) algorithm is used to save the important information of NARX model output, and together with the battery parameters measured at the current time are used as the input vectors of the model to complete SOC estimation. Experimental results show that the SOC error of the proposed model is controlled within 1%, which verifies the effectiveness of the proposed algorithm.

Key words: lithium-ion battery, nonlinear autoregressive with abnormal (NARX), long short-term memories (LSTM), state of charge (SOC)

王语园, 安盼龙, 惠亮亮. 一种改进的NARX神经网络的锂离子电池荷电状态估算[J]. 储能科学与技术, doi: 10.19799/j.cnki.2095-4239.2021.0135.

Yuyuan WANG, Panlong AN, Liangliang HUI. The improved NARX neural network for battery state of charge estimation[J]. Energy Storage Science and Technology, doi: 10.19799/j.cnki.2095-4239.2021.0135.

图/表 10

图1

图2

图3

表1

基于所提模型的算法流程表"


算法流程：
Step1. 收集电池在不同工况下测量的电流，电压和SOC数据；
Step2. 选取数据集D= (X,y)，其中X是表示采样k时刻的电压V(k)和电流，y(k)表示SOC(k)； Step3. 训练NRAX 模型 $y (t) = f (y (t - 1), y (t - 2) . . ., y (t - n y), u (t - 1), u (t - 2) . . ., u (t - n u))$ Step4. 利用NRAX完成第一次SOC估算，并将预测的SOC加入到数据集中 Step5. 将重新构成的数据集划分训练集和验证集，首先完成LSTM模型的训练: 输入门 $i t = σ (w x i x t + w h i x t - 1 + b i)$ ；遗忘门 $f t = σ (w x f x t + w h f x t - 1 + b f)$ 和 $c t = f t c t - 1 + t a n h (w x c x t + w h c h t - 1 + b c)$ ；输出门 $o t = σ (w x o x t + w h o x t - 1 + b o)$ 和 $h t = o t t a n h c t$
Step6. 重复Step.2- Step.5，完成最终的SOC估算。

表1

图4

图5

表2

图6

图7

表3

参考文献 16

1	WEI Meng,YE Min,LI Jiabo,et al.State of charge estimation of lithium-ion batteries using LSTM and NARX neural networks[J].IEEE Access,2020,8:189236-189245.
2	马向平,靳皓晴,朱奇先,王晓兰.锂离子电池荷电状态的在线融合估计方法[J].兰州理工大学学报,2020,46(05):78-84.
	Ma Xiangping, Jin Haoqing, Zhu Qixian, Wang Xiaolan. Online fusion estimation method for state of charge of lithium ion batteries [J]. Journal of Lanzhou University of technology, 2020,46 (05): 78-84
3	章军辉,李庆,陈大鹏,赵野.基于自适应UKF的锂离子动力电池状态联合估计[J].东北大学学报(自然科学版), 2020,41 (11): 1557-1563.
	Zhang Junhui, Li Qing, Chen Dapeng, Zhao Ye. State joint estimation of lithium ion power battery based on adaptive UKF [J]. Journal of Northeast University (NATURAL SCIENCE EDITION), 2020,41 (11): 1557-1563
4	焦自权,范兴明,张鑫,罗奕,刘阳升.基于改进粒子滤波算法的锂离子电池状态跟踪与剩余使用寿命预测方法[J].电工技术学报,2020,35(18):3979-3993.
	Jiao Ziquan, fan Xingming, Zhang Xin, Luo Yi, Liu Yangsheng. State tracking and residual service life prediction method of lithium ion battery based on improved particle filter algorithm [J]. Acta electrotechnics Sinica, 2020,35 (18): 3979-3993
5	陈剑,肖振锋,李达伟,罗磊鑫,夏向阳.基于RLS和UKF算法的锂离子电池荷电状态估计[J].电源技术,2020,44(11): 1600-1657.
	Chen Jian, Xiao Zhenfeng, Li Dawei, Luo Leixin, Xia Xiangyang. State of charge estimation of lithium ion battery based on RLS and UKF algorithm [J]. Power technology, 2020,44 (11): 1600-1657
6	李静,石求军,洪良,刘鹏.基于车辆状态估计的商用车ESC神经网络滑模控制[J].吉林大学学报(工学版),2020,50(05):1545-1555.
	Li Jing, Shi Qiujun, Hong Liang, Liu Peng. Neural network sliding mode control of commercial vehicle ESC based on vehicle state estimation [J]. Journal of Jilin University (Engineering Edition), 2020,50 (05): 1545-1555
7	Shrivastava P, Soon TK, Idris MYIB, Mekhilef S. Overview of model-based online state-of-charge estimation using Kalman filter family for lithium-ion batteries. Renewable and Sustainable Energy Reviews. 2019;113:109233,DOI: 10.1016/j.rser.2019.06.040.
8	Zhong F, Li H, Zhong S, Zhong Q, Yin C. An SOC estimation approach based on adaptive sliding mode observer and fractional order equivalent circuit model for lithium-ion batteries. Communications in Nonlinear Science & Numerical Simulation. 2015;24:127-44
9	Chen C, Xiong R, Shen W. A lithium-ion battery-in-the-loop approach to test and validate multi-scale dual H infinity filters for state of charge and capacity estimation. IEEE Transactions on Power Electronics. 2018;33:332-42.
10	陈剑,肖振锋,刘顺成,罗磊鑫,夏向阳.基于EKF-SVSF的锂离子电池SOC和SOH准确估计[J].电源技术,2020,44(10):1483-1487.
	Chen Jian, Xiao Zhenfeng, Liu Shuncheng, Luo Leixin, Xia Xiangyang. Accurate estimation of SOC and SOH of Li ion battery based on EKF-SVSF [J]. Power technology, 2020,44 (10): 1483-1487.
11	陈涛,郭俊文,张芮.基于双UKF滤波器的锂电池SOC-SOH联合估计方法[J].船电技术,2020,40(S1):95-100.
	Chen Tao, Guo Junwen, Zhang Rui. Joint estimation method of lithium battery soc-soh based on dual UKF filters [J]. Marine electric technology, 2020,40 (S1): 95-100
12	谈发明,张士荣.基于自适应UKF算法的锂电池荷电状态估算策略[J].工业控制计算机,2021,34(01):136-139.
	Tan Faming, Zhang Shirong. State of charge estimation strategy of lithium battery based on adaptive UKF algorithm [J]. Industrial control computer, 2021,34 (01): 136-139
13	张立佳,徐国宁,赵向阳,杜晓伟,周翔.基于神经网络的老化锂电池SOC估计方法的研究[J].电源学报,2020,18(01):54-60.
	Zhang Lijia, Xu Guoning, Zhao Xiangyang, Du Xiaowei, Zhou Xiang. Research on SOC estimation method of aging lithium battery based on neural network [J]. Acta Sinica Sinica, 2020,18 (01): 54-60.
14	李嘉波,魏孟,李忠玉,叶敏,焦生杰,徐信芯.一种改进的支持向量机回归的电池状态估计[J].储能科学与技术,2020,9(04): 1200- 1205. Li Jiabo, Wei Meng, Li Zhongyu, ye min, Jiao Shengjie, Xu Xinxin. Battery state estimation based on improved support vector machine regression [J]. Energy storage science and technology, 2020,9 (04): 1200-1205
15	LH, a KW, CW, JG, XP. Hybrid method based on particle filter and NARX for real-time flow rate estimation in multi-product pipelines. Journal of Process Control. 2020;88:19-31.
16	Bian C, He H, Yang S. Stacked bidirectional long short-term memory networks for state-of-charge estimation of lithium-ion batteries. Energy. 2020;191:116538,DOI: 10.1016/j.energy.2019.


方法	MAE	RMSE
LSTM-NARX	0.82%	0.95%
LSTM	1.78%	1.07%
NARX	2.14%	1.12%
LSSVM	2.68%	1.59%


方法	MAE	RMSE
LSTM-NARX	0.87%	0.98%
LSTM	1.89%	1.11%
NARX	2.01%	1.18%
LSSVM	2.79%	1.67%

[1]	李海涛, 孔令丽, 张欣, 余传军, 王纪威, 徐琳. N/P设计对高镍NCM/Gr电芯性能的影响[J]. 储能科学与技术, 2022, 11(7): 2040-2045.
[2]	刘显茜, 孙安梁, 田川. 基于仿生翅脉流道冷板的锂离子电池组液冷散热[J]. 储能科学与技术, 2022, 11(7): 2266-2273.
[3]	陈龙, 夏权, 任羿, 曹高萍, 邱景义, 张浩. 多物理场耦合下锂离子电池组可靠性研究现状与展望[J]. 储能科学与技术, 2022, 11(7): 2316-2323.
[4]	易顺民, 谢林柏, 彭力. 基于VF-DW-DFN的锂离子电池剩余寿命预测[J]. 储能科学与技术, 2022, 11(7): 2305-2315.
[5]	祝庆伟, 俞小莉, 吴启超, 徐一丹, 陈芬放, 黄瑞. 高能量密度锂离子电池老化半经验模型[J]. 储能科学与技术, 2022, 11(7): 2324-2331.
[6]	王宇作, 王瑨, 卢颖莉, 阮殿波. 孔结构对软碳负极储锂性能的影响[J]. 储能科学与技术, 2022, 11(7): 2023-2029.
[7]	孔为, 金劲涛, 陆西坡, 孙洋. 对称蛇形流道锂离子电池冷却性能[J]. 储能科学与技术, 2022, 11(7): 2258-2265.
[8]	霍思达, 薛文东, 李新丽, 李勇. 基于CiteSpace知识图谱的锂电池复合电解质可视化分析[J]. 储能科学与技术, 2022, 11(7): 2103-2113.
[9]	邓健想, 赵金良, 黄成德. 高能量锂离子电池硅基负极黏结剂研究进展[J]. 储能科学与技术, 2022, 11(7): 2092-2102.
[10]	欧宇, 侯文会, 刘凯. 锂离子电池中的智能安全电解液研究进展[J]. 储能科学与技术, 2022, 11(6): 1772-1787.
[11]	韩俊伟, 肖菁, 陶莹, 孔德斌, 吕伟, 杨全红. 致密储能：基于石墨烯的方法学和应用实例[J]. 储能科学与技术, 2022, 11(6): 1865-1873.
[12]	辛耀达, 李娜, 杨乐, 宋维力, 孙磊, 陈浩森, 方岱宁. 锂离子电池植入传感技术[J]. 储能科学与技术, 2022, 11(6): 1834-1846.
[13]	燕乔一, 吴锋, 陈人杰, 李丽. 锂离子电池负极石墨回收处理及资源循环[J]. 储能科学与技术, 2022, 11(6): 1760-1771.
[14]	沈秀, 曾月劲, 李睿洋, 李佳霖, 李伟, 张鹏, 赵金保. γ射线辐照交联原位固态化阻燃锂离子电池[J]. 储能科学与技术, 2022, 11(6): 1816-1821.
[15]	丁奕, 杨艳, 陈锴, 曾涛, 黄云辉. 锂离子电池智能消防及其研究方法[J]. 储能科学与技术, 2022, 11(6): 1822-1833.