1 School of Mechanical and Electrical Engineering, Wuhan University of Technology, Wuhan 430070, China
2 SAIC-GM-Wuling Automobile Co., Ltd., Liuzhou 545000, China
Show less
文章历史+
Received
Published
Revised
2025-10-22
2026-08-28
2025-12-16
Online Date
2026-09-04
PDF (6845K)
摘要
为应对锂离子电池衰减的强非线性和显著时序依赖性,并解决传统预测模型缺乏不确定性量化能力的挑战,提出了一种高精度的长短期记忆网络(long short-term memory,LSTM)-高斯过程回归(Gaussian process regression,GPR)多维混合预测模型架构。该方法先利用LSTM处理电压、电流、温度、时间及循环索引等多维时序监测数据。LSTM凭借其门控机制,能够自动深入挖掘数据中蕴含的深层衰减特征,有效捕捉复杂的时序依赖关系。然后模型引入GPR对LSTM提取的30维深层特征向量进行高精度回归预测。GPR基于贝叶斯框架,其创新点在于能够实现多维时序监测以及同步输出预测结果的95%置信区间,实现了预测不确定性的可靠量化。基于美国国家航空航天局(National Aeronautics and Space Administration,NASA)电池公共数据集的验证试验结果显示,该混合模型性能优异:决定系数(R2)最低值达到0.997 5。均方根误差(root mean square error,RMSE)和平均绝对误差(mean absolute error,MAE)的最大值分别仅为0.004 99 Ah和0.004 63 Ah。性能指标显著优于对比基准模型卷积神经网络(convolutional neural network,CNN)和支持向量回归(support vector regression,SVR)。研究证实,所提出的LSTM-GPR混合模型有效结合了高精度点预测与可靠不确定性量化,为电池管理系统(battery management systems,BMS)提供了可行的在线电池健康状态(state of health,SOH)预测工具。
Abstract
A highly accurate long short-term memory (LSTM)-Gaussian process regression (GPR) multivariate hybrid prediction model architecture was proposed to address the challenges of strong nonlinearity, significant time-series dependency in battery degradation, and the lack of uncertainty quantification in conventional models. The method utilized a LSTM network to process multi-dimensional time-series data (voltage, current, temperature, etc.). The LSTM effectively captured complex time-series dependencies and automatically extracted deep degradation features. Subsequently, GPR was introduced, operating on the extracted 30-dimensional feature vectors. GPR’s innovation lay in its ability to simultaneously output the 95% confidence interval, achieving reliable uncertainty quantification. Validation on the National Aeronautics and Space Administration (NASA) battery dataset demonstrates superior performance: the minimum coefficient of determination (R2) reaches 0.997 5, with the maximum root mean square error (RMSE) and mean absolute error (MAE) being only 0.004 99 Ah and 0.004 63 Ah, respectively. The performance metrics significantly surpass those of the comparative benchmark models, namely the convolutional neural network (CNN) and support vector regression (SVR). Research confirms that the proposed LSTM-GPR model, by combining high-precision point prediction with reliable uncertainty quantification, provides a viable solution for online state-of-health (SOH) prediction within battery management systems (BMS).
锂离子电池作为一种重要的能量存储设备,具有能量密度高、寿命长、自放电率低等优点。然而,电池在长期的使用过程中,由于复杂的内部反应和外部干扰,电池的容量持续下降[1]。电池健康状态(state of health,SOH)直接衡量了电池的性能衰退情况和剩余使用寿命[2-5]。SOH通常定义为电池当前最大可放电容量与其出厂标称容量的比值。依据 IEEE Recommended Practice for Maintenance, Testing, and Replacement of Valve-Regulated Lead-Acid (VRLA) Batteries for Stationary Applications(IEEE Std 1188—2025)标准规定,当SOH下降至70%~80%的范围时,电池的实际使用寿命便宣告终结[2]。SOH估计的准确性具有多重价值,有助于优化电池的充放电策略,从而提高能量利用率,也可以预防电池过度放电和保障整个系统安全稳定运行[6]。然而,电池的衰减是一个复杂的过程,涉及电化学、热力学和机械应力等多重因素的相互耦合,呈现出明显的非线性特性。而且电池的状态具有时序依赖性,即其当前的健康状况与历史的使用过程紧密相关[7]。当前,中国新能源汽车和储能产业快速发展,对电池寿命、安全与可靠健康管理的需求不断提升。随着车载数据量的显著增加,数据驱动的SOH预测已成为行业主流方向。但现有模型仍普遍存在难以处理多维时序数据、不具备不确定性量化等问题。因此,发展兼具深度特征提取能力与预测可信度的混合模型,对提升中国电池管理系统的可靠性具有重要意义。
使用美国国家航空航天局(National Aeronautics and Space Administration,NASA)艾姆斯研究中心发布的电池老化数据集,包括B0005、B0006和B0007三组电池的循环数据如图3所示。该数据集记录了每个循环的电压、电流、温度、时间等时序数据,以及实际容量值,适合构建基于时序特征的LSTM模型。
WangZhifu, YangZhongyi, LuoWei, et al. Review of remaining useful life prediction for lithium-ion battery based on data-driven[J]. Science Technology and Engineering, 2023, 23(15): 6279-6289.
ZhouRen, ZhangXiangwen. State of health estimation for electric vehicle power batteries based on real-vehicle data and BP-AdaBoost algorithm[J]. Science Technology and Engineering, 2022, 22(21): 9398-9406.
[5]
KumarB V, FarhanM A A. Optimalsimul-taneous allocation of electric vehicle charging stations and capacitors in radial distribution network considering reliability[J]. Journal of Modern Power Systems and Clean Energy, 2024, 12(5): 1584-1595.
[6]
HuangW, HuJ, LuoS. The technological innovation pathway for green, low-carbon, and durable pavement construction and maintenance[J]. Science China Technological Sciences, 2024, 67(12): 3959-3961.
[7]
DingG R, WangW B, ZhuT. Remaining useful life prediction for lithium-ion batteries based on CS-VMD and GRU[J]. IEEE Access, 2022, 10: 89402-89413.
FuXin, WangJingyue, ZhuNan, et al. Battery remaining useful life prediction based on SSA-GPR and WPD[J]. Science Technology and Engineering, 2025, 25(23): 10023-10030.
LiYingshun, KanHongda, GuoZhannan, et al. Lithium-ion battery remaining useful life prediction based on data preprocessing and VMD-LSTM-GPR[J]. Transactions of China Electrotechnical Society, 2024, 39(10): 3244-3258.
WuShengli, OuHua, XingWenting. SOC estimation of Li-ion ba-ttery based on multi innovation extended Kalman filtering[J]. Science Technology and Engineering, 2024, 24(16): 6742-6748.
[14]
WuJ, ZhangC B, ChenZ H. An online method for lithium-ion battery remaining useful life estimation using importance sampling and neural networks[J]. Applied Energy, 2016, 173: 134-140.
[15]
DengZ W, HuX S, LinX K, et al. General discharge voltage information enabled health evaluation for lithiumion batteries[J]. IEEE/ASME Transactions on Mechatronics, 2020, 26(3): 1295-1306.
WeiZixuan, HanXiaojuan, LiXuan. State of health assessment for second-life batteries based on deep neural networks[J]. Acta Energiae Solaris Sinica, 2022, 43(5): 518-524.
ZangWeihong, LiZhonghua, WangFacheng, et al. Research on data-driven methods for lithium battery capacity degradation modeling[J]. Journal of Chongqing University of Technology (Natural Science), 2023, 37(2): 12-18.
[20]
ZhuJ, TanT X, WuL F, et al. RUL prediction of lithium-ion battery based on improved DGWO-ELM method in a random discharge rates environment[J]. IEEE Access, 2019, 7: 125176-125187.
XuJianing, NiYulong, ZhuChunbo. Remaining useful life prediction of lithium batteries based on improved support vector regression[J]. Transactions of China Electrotechnical Society, 2021, 36(17): 3693-3704.
WuZhongqiang, HuXiaoyu, MaBoyan, et al. Remaining useful life prediction of lithium batteries based on PF-LSTM[J]. Acta Metrologica Sinica, 2023, 44(6): 939-947.
[27]
LiuD T, ZhouJ B, PanD W, et al. Lithium-ion battery remaining useful life estimation with an optimized relevance vector machine algorithm with incremental learning[J]. Measurement, 2015, 63: 143-151.
YangYanru, WenJie, ShiYuanhao, et al. Remaining useful life prediction of lithium-ion batteries based on CEEMDAN and SVR[J]. Journal of Electronic Measurement and Instrumentation, 2020, 34(12): 197-205.
[30]
PhattaraK, YodoN. A data-driven predictive prognostic model for lithium-ion batteries based on a deep learning algorithm[J]. Energies, 2019, 12(4): 660.
[31]
LiX, WangZ, YanJ. Prognostic health condition for lithium battery using the partial incremental capacity and Gaussian process regression[J]. Journal of Power Sources, 2019, 421: 56-67.