基于外特征图像与深度卷积神经网络的锂离子电池一致性分类方法及应用

夏向阳; 赵晓悦; 岳家辉; 吴晋波; 彭铖; 柏文琦

doi:10.19912/j.0254-0096.tynxb.2025-0667

PDF(3322 KB)

太阳能学报 ›› 2026, Vol. 47 ›› Issue (9) : 28-39. DOI: 10.19912/j.0254-0096.tynxb.2025-0667

基于外特征图像与深度卷积神经网络的锂离子电池一致性分类方法及应用

夏向阳¹, 赵晓悦¹, 岳家辉¹, 吴晋波², 彭铖³, 柏文琦⁴

作者信息 +

CONSISTENCY CLASSIFICATION METHOD AND APPLICATION OF LITHIUM-ION BATTERIES BASED ON EXTERNAL FEATURE IMAGES AND DEEP CONVOLUTIONAL NEURAL NETWORKS

Xia Xiangyang¹, Zhao Xiaoyue¹, Yue Jiahui¹, Wu Jinbo², Peng Cheng³, Bo Wenqi⁴

Author information +

文章历史 +

摘要

针对储能电站电池一致性改善手段较为单一、主要依赖外置离线设备调节电池箱间电压差的问题,提出一种将外特征图像与TL-GoogLeNet卷积神经网络模型相结合的储能电站锂离子电池一致性分类方法。该方法以电池电压、温度及IC曲线构建外特征图像,利用灰度层突显曲线峰值与拐点特征,以明暗区域强化特征表现,增强电池状态的有效表征并捕捉其老化过程中的关键变化。采用迁移学习策略将预训练GoogLeNet模型参数迁移至特定电池数据集,构建TL-GoogLeNet模型以提升模型泛化性能。通过锂电池公开数据集和实际储能站运行数据进行综合验证,所提方法准确率均保持在95%以上,结果表明该方法具有较高的分类准确性和鲁棒性,可为储能电站运行后期的不一致性问题和运行安全维护提供有效指导。

Abstract

Inconsistency problems in energy storage stations are often resolved by adjusting the voltage difference of battery packs with external offline devices, while relevant solutions remain limited. This paper proposes an energy storage lithium-ion battery consistency classification method combining external feature images with the Transfer Learning （TL）-GoogLeNet convolutional neural network （CNN） model. The method constructs external feature images from battery voltage, temperature, and IC curves. It utilizes grayscale layers to highlight peaks and inflection points of curves via different brightness regions, which improves the effective representation of battery status and captures key changes during the aging process. The transfer learning strategy transfers parameters of the pre-trained GoogLeNet to target battery datasets to build the TL-GoogLeNet model for better generalization performance. Comprehensive validation is performed using both publicly available lithium battery datasets and practical operating data from energy storage stations. The proposed method achieves an accuracy rate of over 95%, demonstrating its high classification accuracy and robustness. The results suggest that this method can provide effective guidance for inconsistency elimination and safe maintenance of energy storage stations during their later stages of operation.

导出引用

夏向阳, 赵晓悦, 岳家辉, 吴晋波, 彭铖, 柏文琦. 基于外特征图像与深度卷积神经网络的锂离子电池一致性分类方法及应用[J]. 太阳能学报. 2026, 47(9): 28-39 https://doi.org/10.19912/j.0254-0096.tynxb.2025-0667

Xia Xiangyang, Zhao Xiaoyue, Yue Jiahui, Wu Jinbo, Peng Cheng, Bo Wenqi. CONSISTENCY CLASSIFICATION METHOD AND APPLICATION OF LITHIUM-ION BATTERIES BASED ON EXTERNAL FEATURE IMAGES AND DEEP CONVOLUTIONAL NEURAL NETWORKS[J]. Acta Energiae Solaris Sinica. 2026, 47(9): 28-39 https://doi.org/10.19912/j.0254-0096.tynxb.2025-0667

中图分类号： TM912

参考文献

[1] Yi Z X, Chen Z L, Yin K, et al.Sensing as the key to the safety and sustainability of new energy storage devices[J]. Protection and Control of Modern Power Systems, 2023, 8: 27.
[2] 黎博, 陈民铀, 钟海旺, 等. 高比例可再生能源新型电力系统长期规划综述[J]. 中国电机工程学报, 2023, 43(2): 555-581.
Li B, Chen M Y, Zhong H W, et al.A review of long-term planning of new power systems with large share of renewable energy[J]. Proceedings of the CSEE, 2023, 43(2): 555-581.
[3] 郭喜峰, 王凯泽, 单丹, 等. 多角度基于CEEMDAN-CNN-BiLSTM模型的锂离子电池RUL预测[J]. 太阳能学报, 2024, 45(7): 181-189.
Guo X F, Wang K Z, Shan D, et al.Rul prediction for lithium ion batteries based on CEEMDAN-CNN-BiLSTM model from multiple perspectives[J]. Acta Energiae Solaris Sinica, 2024, 45(7): 181-189.
[4] 夏向阳, 吕崇耿, 吴小忠, 等. 基于混合模型的储能用锂离子电池剩余寿命预测方法研究[J]. 太阳能学报, 2024, 45(10): 726-735.
Xia X Y, Lyu C G, Wu X Z, et al.Research on remaining life prediction method of lithium-ion battery for energy storage based on hybrid model[J]. Acta Energiae Solaris Sinica, 2024, 45(10): 726-735.
[5] 王楠, 周喜超, 彭勇, 等. 基于证据KNN分类器的蓄电池一致性诊断[J]. 太阳能学报, 2022, 43(4): 13-19.
Wang N, Zhou X C, Peng Y, et al.Battery consistency diagnosis based on evidential KNN classifier[J]. Acta Energiae Solaris Sinica, 2022, 43(4): 13-19.
[6] Schneider E L, Oliveira C T, Brito R M, et al.Classification of discarded NiMH and Li-ion batteries and reuse of the cells still in operational conditions in prototypes[J]. Journal of Power Sources, 2014, 262: 1-9.
[7] Liao Q Q, Mu M M, Zhao S Q, et al.Performance assessment and classification of retired lithium ion battery from electric vehicles for energy storage[J]. International Journal of Hydrogen Energy, 2017, 42(30): 18817-18823.
[8] 骆凡, 黄海宏, 王海欣. 基于短时脉冲放电与电化学阻抗谱的退役动力电池快速分选与重组方法[J]. 仪器仪表学报, 2022, 43(1): 229-238.
Luo F, Huang H H, Wang H X.A fast screening and recombinant method based on short-time pulse discharge and electrochemical impedance spectroscopy for decommissioned power batteries[J]. Chinese Journal of Scientific Instrument, 2022, 43(1): 229-238.
[9] Garg A, Yun L, Gao L, et al.Development of recycling strategy for large stacked systems: Experimental and machine learning approach to form reuse battery packs for secondary applications[J]. Journal of Cleaner Production, 2020, 275: 124152.
[10] Lyu C, Song Y K, Wang L X, et al.A new method for lithium-ion battery uniformity sorting based on internal criteria[J]. Journal of Energy Storage, 2019, 25: 100885.
[11] Ran A H, Zhou Z H, Chen S X, et al.Data-driven fast clustering of second-life lithium-ion battery: mechanism and algorithm[J]. Advanced Theory and Simulations, 2020, 3(8): 2000109.
[12] 段双明, 王恩吉, 朱微, 等. 基于锂电池电压片段的退役电池快速筛选重组[J]. 电网技术, 2024, 48(10): 4264-4274.
Duan S M, Wang E J, Zhu W, et al.Rapid screening and recombing of retired batteries based on voltage segments of lithiumion baterries[J]. Power System Technology, 2024, 48(10): 4264-4274.
[13] Li C B, Wang N B, Li W, et al.Regrouping and echelon utilization of retired lithium-ion batteries based on a novel support vector clustering approach[J]. IEEE Transactions on Transportation Electrification, 2022, 8(3): 3648-3658.
[14] Lai X, Qiao D D, Zheng Y J, et al.A rapid screening and regrouping approach based on neural networks for large-scale retired lithium-ion cells in second-use applications[J]. Journal of Cleaner Production, 2019, 213: 776-791.
[15] Zhou Z K, Duan B, Kang Y Z, et al.An efficient screening method for retired lithium-ion batteries based on support vector machine[J]. Journal of Cleaner Production, 2020, 267: 121882.
[16] Qiang H, Liu Y L, Zhang W J.A screening method for retired lithium-ion batteries based on support vector machine with a multi-class kernel function[J]. Journal of Electrochemical Energy Conversion and Storage, 2024, 21(2): 021005.
[17] 夏珺羿. 基于机器学习方法的液态金属电池分选研究[D]. 武汉: 华中科技大学, 2023.
Xia J Y.The research on battery sorting of liquid metal batteries based on machine learning methods[D]. Wuhan: Huazhong University of Science and Technology, 2023.
[18] Yu J H, Liu K P, Liu W L, et al.Recurrence plot image and GoogLeNet based historical abuse backtrace for li-ion batteries[J]. Journal of Energy Storage, 2023, 74: 109378.
[19] Xie J L, Li Y K, Hou Z S, et al.Inference of historical abusive operations on Li-ion batteries using series voltage synchronicity[J]. IEEE Transactions on Reliability, 2025, 74(3): 3977-3989.
[20] Mao B H, Yuan J Y, Li H, et al.Gramian angular field-based state-of-health estimation of lithium-ion batteries using two-dimensional convolutional neural network and bidirectional long short-term memory[J]. Journal of Power Sources, 2025, 626: 235713.
[21] 赵波, 刘相万, 章雷其, 等. 基于GoogLeNet与迁移学习的质子交换膜燃料电池集成系统故障诊断[J]. 中国电机工程学报, 2024, 44(13): 5147-5157.
Zhao B, Liu X W, Zhang L Q, et al.Fault diagnosis of proton exchange membrane fuel cell integrated system based on GoogleNet and transfer learning[J]. Proceedings of the CSEE, 2024, 44(13): 5147-5157.
[22] 柳坤鹏, 刘连起, 李元锴, 等. 基于递归图和GoogLeNet模型的锂离子电池的历史滥用回溯[J]. 电力科学与工程, 2023, 39(9): 20-27.
Liu K P, Liu L Q, Li Y K, et al.Historical abuse backtrace for Li-ion batteries based on recurrence plot image and GoogLeNet model[J]. Electric Power Science and Engineering, 2023, 39(9): 20-27.
[23] Zhu J G, Wang Y X, Huang Y, et al.Data-driven capacity estimation of commercial lithium-ion batteries from voltage relaxation[J]. Nature Communications, 2022, 13: 2261.