To improve the accuracy of estimating the state of health (SOH) of lithium-ion batteries and overcome the limitations of existing methods to fully characterize the details of battery decay, this paper proposes a method integrated by distance intersection over union loss (DIoUloss), simple, parameter-free attention module (SimAM) and convolutional neural network-bidirectional long short term memory network (CNN-BiLSTM) for battery SOH estimation. IEA-T features, composed of the incremental energy area (IEA) and charging time (T) of lithium-ion batteries are used for the battery SOH estimation. The DIoUloss function and SimAM mechanism are integrated into the CNN-BiLSTM model to establish the CNN-BiLSTM-SimAM battery SOH estimation model. The cyclic aging experiments of lithium-ion batteries are detected. Compared with other methods such as GRU, SVR, CNN-LSTM, and CNN-BiLSTM, the proposed method can more effectively characterize the details of battery health decline. The coefficient of determination is higher than 0.96, and the maximum root-mean-square error is less than 0.020, showing favorable accuracy and efficiency.
数据驱动方法中的一项关键技术是模型构建。近年来,一些深度学习类算法,如神经网络[13]、高斯过程回归[14]、支持向量机[15]和极限学习机[16]展现出极大的优势。门控循环单元(gated recurrent units,GRU)[17]、长短期记忆网络(long short-term memory,LSTM)[18]等循环神经网络,擅长识别数据中的复杂模式和映射规律,在电池SOH中得到广泛应用。LSTM能够处理长序列数据时遇到的梯度爆炸问题,表现出更好的泛化能力。双向长短期记忆网络(bidirectional long short-term memory,BiLSTM)[19]通过组合两个LSTM,在正向和反向上处理输入数据序列,能够更全面地提取数据特征。卷积神经网络-双向长短期记忆网络(convolutional neural network-bidirectional long short-term memory network,CNN-BiLSTM)算法则结合了CNN的局部特征提取能力和BiLSTM的时序特征分析能力,表现出优异的SOH估计性能。但在实际应用中,由于测量误差和噪声干扰,SOH估计的准确性仍有提高空间。通过为时间步或特征赋予不同的权重,加入注意力机制,帮助估计模型聚焦关键的特征信息,同时加入数据处理和损失函数,能有效提高模型的SOH估计能力。距离交并比损失函数(distance intersection over union loss function,DIoUloss)能够更准确地衡量评估结果与实际标签之间的重叠程度。无参注意力机制(simple, parameter-free attention module,SimAM)允许模型自动聚焦于数据中的关键特征,无需手动设置参数,有助于模型在长序列数据中捕捉到更重要的信息。DIoUloss、SimAM可融合于CNN-BiLSTM模型,使得CNN-BiLSTM模型在SOH估计时更加精确和高效。
本文在IC分析和DV分析的基础上,提出了一种增量能量(incremental energy,IE)分析方法,通过从锂电池充电电压、电流中提取数据,开展增量能量分析,结合与电池SOH相关度较高的充电时长和电池增量能量曲线面积,构建增量能量面积-时长(incremental energy area-time,IEA-T)特征,再将DIoUloss损失函数和SimAM机制融合于CNN-BiLSTM模型,建立了一个CNN-BiLSTM-SimAM锂离子电池SOH估计模型,并开展锂离子电池SOH估计实验,验证方法的有效性和先进性。
在增量能量面积(incremental energy area,IEA)特征的基础上,对测量数据进行深入分析,进一步提取电池健康特征。恒压充电时长、充电电流的时间积分和电池温度的时间积分能够反映电池在充电过程中的动态变化,且与电池SOH相关。将上述特征归一化处理,使用皮尔逊相关系数[21]量化这些特征与SOH之间的相关程度:“1”表示完全正相关;“-1”表示完全负相关;“0”表示没有线性相关性。上述特征的皮尔逊相关系数计算如下:
DIoUloss是一种用于目标检测中边界框回归的损失函数。在交并比损失函数(intersection over union loss,IoUloss)和广义交并比损失函数(generalized intersection over union loss,GIoUloss)函数的基础上进行了扩展。IoUloss函数主要用于评估两个边界框的重叠程度,目标是离散的边界框,无法直接适用于SOH连续值回归问题。GIoUloss函数是对IoUloss函数的优化,能解决IoUloss在边界框不重叠的情况下梯度信息不足的问题,但由于离散化的特性,在SOH估计中的适用性仍然有限。相比之下,DIoUloss函数通过引入距离惩罚项衡量预测框和真实框的中心点距离,同时结合重叠面积和中心点距离量化预测框和真实框的距离交并比误差,可以更有效地指导边界框回归。
PENGC. Electrochemical energy storage and conversion technologies: Opportunities and challenges[J]. Journal of Wuhan University (Natural Science Edition), 2023, 69(4): 417-418. DOI: 10.14188/j.1671-8836.2023. 0112(Ch ).
ZHOUY J, WANGZ P, HONGJ C, et al. Review of overcharge-to-thermal runaway and the control strategy for lithium-ion traction batteries in electric vehicles[J]. Journal of Mechanical Engineering, 2022, 58(10): 112-135. DOI: 10.3901/JME.2022.10.112(Ch ).
[5]
WUJ, MENGJ H, LINM Q, et al. Lithium-ion battery state of health estimation using a hybrid model with electrochemical impedance spectroscopy[J]. Reliability Engineering & System Safety, 2024, 252: 110450. DOI: 10.1016/j.ress.2024.110450 .
[6]
LIC F, YANGL, LIQ, et al. SOH estimation method for lithium-ion batteries based on an improved equivalent circuit model via electrochemical impedance spectroscopy[J]. Journal of Energy Storage, 2024, 86: 111167. DOI: 10.1016/j.est.2024.111167 .
[7]
XUX D, TANGS J, RENH H, et al. Joint state estimation of lithium-ion batteries combining improved equivalent circuit model with electrochemical mechanism and diffusion process[J]. Journal of Energy Storage, 2022, 56: 106135. DOI: 10.1016/j.est.2022.106135 .
[8]
ZHENGY J, CUIY F, HANX B, et al. A capacity prediction framework for lithium-ion batteries using fusion prediction of empirical model and data-driven method[J]. Energy, 2021, 237: 121556. DOI: 10.1016/j.energy.2021.121556 .
[9]
WEIZ B, LIUK L, LIUX H, et al. Multilevel data-driven battery management: From internal sensing to big data utilization[J]. IEEE Transactions on Transportation Electrification, 2023, 9(4): 4805-4823. DOI:10.1109/TTE.2023.3301990 .
[10]
KONGJ Z, YANGF F, ZHANGX, et al. Voltage-temperature health feature extraction to improve prognostics and health management of lithium-ion batteries[J]. Energy, 2021, 223: 120114. DOI: 10.1016/j.energy.2021.120114 .
[11]
TIANJ P, XIONGR, SHENW X. State-of-health estimation based on differential temperature for lithium ion batteries[J]. IEEE Transactions on Power Electronics, 2020, 35(10): 10363-10373. DOI: 10.1109/TPEL.2020.2978493 .
[12]
YANGJ F, CAIY F, MIC C. State-of-health estimation for lithium-ion batteries based on decoupled dynamic characteristic of constant-voltage charging current[J]. IEEE Transactions on Transportation Electrification, 2022, 8(2): 2070-2079. DOI: 10.1109/TTE.2021.3125932 .
[13]
LINK R, HUANGC C, SOUK C. Lithium-ion battery state of health estimation using simple regression model based on incremental capacity analysis features[J]. IEEE Transactions on Transportation Electrification, 2021, 8(2): 2070-2079.
[14]
KRUPPA, FERGE, SCHULDTF, et al. Incremental capacity analysis as a state of health estimation method for lithium-ion battery modules with series-connected cells[J]. Batteries, 2021, 7(1):2. DOI: 10.3390/batteries7010002 .
[15]
QUJ T, LIUF, MAY X, et al. A neural-network-based method for RUL prediction and SOH monitoring of lithium-ion battery[J]. IEEE Access, 2019, 7: 87178-87191. DOI: 10.1109/ACCESS.2019.2925468 .
[16]
FENGH L, SHIG L. SOH and RUL prediction of Li-ion batteries based on improved Gaussian process regression[J]. Journal of Power Electronics, 2021, 21(12): 1845-1854. DOI: 10.1007/s43236-021-00318-5 .
ZHANGC L, ZHAOS S, HEY G. Remaining useful life prediction method for lithium-ion batteries based on ensemble empirical mode decomposition and ensemble machine learning[J]. Power System Protection and Control, 2023, 51(13):177-186. DOI: 10.19783/j.cnki.pspc.221746(Ch ).
[19]
LIY, WANGS L, CHENL, et al. Multiple layer kernel extreme learning machine modeling and eugenics genetic sparrow search algorithm for the state of health estimation of lithium-ion batteries[J]. Energy, 2023, 282: 128776. DOI: 10.1016/j.energy.2023.128776 .
[20]
ZHANGC L, LUOL J, YANGZ, et al. Battery SOH estimation method based on gradual decreasing current, double correlation analysis and GRU[J]. Green Energy and Intelligent Transportation, 2023, 2(5): 100108. DOI: 10.1016/j.geits.2023.100108 .
ZHANGC L, ZHAOS S, HEY G. State-of-health estimate for lithium-ion battery using information entropy and PSO-LSTM[J]. Journal of Mechanical Engineering, 2022, 58(10):180-190. DOI: 10.3901/JME.2022.10.180(Ch ).
[23]
ZHUZ Y, YANGQ, LIUX, et al. Attention-based CNN-BiLSTM for SOH and RUL estimation of lithium-ion batteries[J]. Journal of Algorithms & Computational Technology, 2022, 16: 17483026221130598. DOI: 10.1177/17483026221130598 .
ZHANGX. Interface evolution and engineering for lithium-ion batteries[D]. Beijing: University of Chinese Academy of Sciences (Institute of Physics, Chinese Academy of Sciences), 2023. DOI: 10.27604/d.cnki.gwlys.2023.000072(Ch ).
[26]
CHANGC, WANGQ Y, JIANGJ C, et al. Lithium-ion battery state of health estimation using the incremental capacity and wavelet neural networks with genetic algorithm[J]. Journal of Energy Storage, 2021 38: 102570. DOI: 10.1016/j.est.2021.102570 .
[27]
REHMERA, KROLLA. On the vanishing and exploding gradient problem in gated recurrent units[J]. IFAC-PapersOnLine, 2020, 53(2):1243-1248. DOI:10.1016/j.ifacol.2020.12.1342 .