1 Branch of Inner Mongolia Electric Power Research Institute, Inner Mongolia Power (Group) Co., Ltd., Hohhot 010040, China
2 Hohhot Power Supply Company, Inner Mongolia Power (Group) Co., Ltd., Hohhot 010011, China
3 Inner Mongolia Power (Group) Co., Ltd., Hohhot 010010, China
Show less
文章历史+
Received
Published
2025-03-09
2026-01-08
Online Date
2026-05-27
PDF (8504K)
摘要
为全面评估站用2 V/500 AH蓄电池健康状态(state of health,SOH)以延长电池寿命和提高监测准确性,通过正交实验分析温度、充放电电流和放电深度的影响,结合超声无损检测技术捕捉电极微观变化,采用希尔伯特黄变换(Hilbert-Huang transform,HHT)降噪并构建灰狼优化算法-人工神经网络(grey wolf optimizer-artificial neural network, GWO-ANN)模型实现SOH高精度预测。实验表明,温度每升高10 ℃缩短寿命40%~50%,放电深度50%时寿命为100%时的3倍;HHT-GWO-ANN模型预测精度(R2=0.941)显著优于ANN模型(R2=0.749),复杂工况下准确率超90%。提出的超声无损检测与HHT-GWO-ANN模型相结合的方法,为蓄电池SOH评估提供了高精度、非破坏性解决方案,“浅充浅放”策略可延长电池寿命20%以上,对电池寿命延长和智能监测具有重要价值。
Abstract
In order to comprehensively assess the state of health (SOH) of station-use 2 V/500 AH batteries to prolong the battery life and improve the monitoring accuracy. The effects of temperature, charge/discharge current and discharge depth were analyzed through orthogonal experiments, combined with ultrasonic nondestructive testing technology to capture the micro changes of electrodes, and state of health(HHT) noise reduction and construction of grey wolf optimizer-artificial neural network(GWO-ANN) model to realize the SOH prediction with high accuracy. The experiments show that the life span is shortened by 40%~50% for every 10 ℃ increase in temperature, and the life span at 50% discharge depth is three times of that at 100%. The prediction accuracy of the HHT-GWO-ANN model (R2=0.941) is significantly better than that of the ANN model (R2=0.749), and the accuracy rate is more than 90% under complex working conditions. The proposed method combining ultrasonic nondestructive testing and the HHT-GWO-ANN model provides a high-precision and nondestructive solution for the SOH evaluation of batteries, and the “shallow charging and shallow discharging” strategy can extend the battery life by more than 20%, which is of great value for the extension of the battery life and intelligent monitoring.
WEI Bing-ling, TANG Li, ZHOU Xin, NIU Zi-xuan, AN Sheng-dong.
Evaluation of Battery Health by Fusion of Ultrasonic Data and Hybrid Machine Learning[J].
Science Technology and Engineering, 2026, 26(1): 205-215 DOI:10.12404/j.issn.1671-1815.2501568
随着智能电网建设对电力供应可靠性要求的提升,铅酸蓄电池作为变电站不间断电源(uninterrupted power system, UPS)系统的核心储能单元[1-2],其健康状态直接影响着电网后备电源的应急响应能力。作为变电站中的紧急备用电源装置,蓄电池运行稳定性及使用寿命对于确保电力系统的安全稳定运行以及提升应急响应能力具有至关重要的影响。在电力行业中,确保蓄电池的长寿命和高效性能是保障电网稳定运行的关键。然而,蓄电池的寿命受多种因素影响,如充放电循环次数、工作环境等,这些因素导致电池性能逐渐下降,直至失效[3-5]。因此,准确、快速地检测蓄电池的寿命,对于预防电池故障、延长电池使用寿命、降低运维成本具有重要意义。
现基于混合人工神经网络,开展变电站用蓄电池健康状态(state of health,SOH)的超声检测表征研究。利用正交实验法进行加速条件下蓄电池健康状态评估,并将此数据记录后作为模型输出。其次,采用超声检测对完成充电后的样品进行检测,使用希尔伯特黄变换(Hilbert-Huang transform,HHT)[27]对超声原始数据降噪处理以提升其建模效率,并以降噪后的结果为模型输入。通过构建混合人工神经网络模型,实现对超声检测信号的智能分析和处理并提取出反映蓄电池性能的关键特征参数,从而建立超声检测表征与蓄电池SOH之间的映射关系模型。从多角度对所构建的训练模型和预测模型的效果进行评价以获得更优异的蓄电池健康状态评估模型。以期为变电站用蓄电池的寿命检测提供一种新的、有效的技术手段和理论依据,从而为整套输电系统的稳定运行提供保障。
然而,ANN存在训练时间长、模型复杂度高、对数据质量与特征选择敏感以及模型解释性差等不足。这些局限性促使研究者们不断探索和优化ANN的训练过程[31],其中,灰狼优化算法(grey wolf optimizer,GWO)[32-33]作为一种高效的启发式优化算法,被广泛应用于ANN的参数优化中。GWO优化ANN的流程如图3所示。
ZhongGuobin, LiuXintian, HeYao, et al. Estimation of SOH of lead-acid batteries in substation[J]. Chinese Journal of Power Sources, 2016, 40(12): 2407-2410.
HuYan, YangYang, XieJiayu, et al. Aviation alkaline storage battery state-of-health estimation based on the EP-weighted static SOR algorithm[J]. Science Technology and Engineering, 2015, 15(27): 203-206.
[9]
王博. 变电站阀控铅酸蓄电池充电均衡控制的研究[D]. 广州: 华南理工大学, 2020.
[10]
WangBo. Research on charging balance control of VRLA battery for substation[D]. Guangzhou: South China University of Technology, 2020.
MaoJunlong, ChenWengang. Life evaluation of valve-regulated lead-acid battery based on improved LS-SVM based on particle swarm optimization[J]. Batteries, 2022, 59(2): 70-74, 86.
[13]
ButlerP, DunleaveyJ, Farber-DeandaM, et al. Performance of valve-regulated lead-acid batteries in real-world stationary applications: utility installations[J]. Journal of Power Sources, 2001, 96(1): 94-101.
[14]
GruensternR G, TaylorM E. High temperature application accelerated cycle life test for 12 Volt lead-acid SLI automotive storage batteries[R]. Wisconsin: SAE Technical Paper, 2007.
[15]
HeB, WangH, HeX. Vibration test methods and their experimental research on the performance of the lead-acid battery[J]. Journal of Power Sources, 2014, 268: 326-330.
[16]
钟国彬. 铅酸蓄电池寿命评估及延寿技术[M]. 北京: 中国电力出版社, 2018.
[17]
ZhongGuobin. Life assessment and life extension technology for lead-acid batteries[M]. Beijing: China Electric Power Press, 2018.
[18]
KerleyR, HyunJ H, HaD S. Automotive lead-acid battery state-of-health monitoring system[C]// IECON 2015-41st Annual Conference of the IEEE Industrial Electronics Society. New York: IEEE, 2015: 3934-3938.
ShuChengcai. Research on the co-estimation of SOC and SOH and on charging-discharging strategy of the vehicular lead-acid battery[D]. Hefei: Hefei University of Technology, 2018.
[21]
ShahriariM, FarrokhiM. State of health estimation of VRLA batteries using fuzzy logic[C]// 2010 18th Iranian Conference on Electrical Engineering. New York: IEEE, 2010: 629-634.
[22]
FestigoE, JuanicoD, NonatP, et al. Acoustic non-invasive estimation of lead-acid battery state of health: applications for cell-level charge balancing[J]. Energy Reports, 2022, 8: 372-377.
DingYi, LiuShengzhong, WangXudong, et al. The SOH estimate method research on lead-acid battery based on mixture kernels function[J]. Electric Drive, 2021, 51(22): 56-62.
YuZeyu, YuanHongqiang, WeiXiaolong, et al. Defect identification method of pipeline weld ultrasonic testing based on deep learning[J]. Science Technology and Engineering, 2022, 22(30): 13288-13292.
WuLifeng. Portable rapid battery capacity detection device based on ultrasonic detection technology[D]. Beijing: North China Electric Power University (Beijing), 2022.
YangZhengyu. Overall architecture analysis system of abnormal electricity behavior based on big data and machine learning[J]. Electrical Measurement & Instrumentation, 2023, 60(6): 167-173.
KongXiangqing, ZhangMingliang, KangRan, et al. Advances in machine learning for predicting concrete properties[J]. Science Technology and Engineering, 2015, 25(5): 1764-1773.
YuHonghui, LinShenghong, ZhuJianquan, et al. On-line optimization of micro grid based on deep reinforcement learning[J]. Electrical Measurement & Instrumentation, 2024, 61(4): 9-14.
[39]
ÇelikÖ. A research on machine learning methods and its applications[J]. Journal of Educational Technology and Online Learning, 2018, 1(3): 25-40.
[40]
HegdeJ, RoksethB. Applications of machine learning methods for engineering risk assessment: a review[J]. Safety Science, 2020, 122: 104492.
HeQing, LiNing, LuoWenjuan, et al. A survey of machine learning algorithms for big data[J]. Pattern Recognition and Artificial Intelligence, 2014, 27(4): 327-336.
WuXiaomeng, SunAnlei, LiChenchen, et al. Hybrid energy storage double-layer optimal configuration considering wind power output volatility[J]. Science Technology and Engineering, 2024, 24(24): 10313-10320.
[45]
MaoY, QueP. Application of Hilbert-Huang signal processing to ultrasonic non-destructive testing of oil pipelines[J]. Journal of Zhejiang University-Science A, 2006, 7(2): 130-134.
National Lead-acid Storage Battery of Standardization Administration of China. Lead-acidbatteriesforstationaryvalve-regulated—Part1:Technicalrequirements:GB/T19638.1—2014[S]. Beijing: Standards Press of China, 2014.
DingJiangqiao, WenYi, LüQiansu, et al. Abnormal state detection method of power equipment based on time series and neural network[J]. Electrical Measurement & Instrumentation, 2024, 61(2): 185-190.
ZhangChi, GuoYuan, LiMing. Review of development and application of artificial neural network models[J]. Computer Engineering and Applications, 2021, 57(11): 57-69.
[52]
AbdolrasolM G M, HussainS M S, UstunT S, et al. Artificial neural networks based optimization techniques: a review[J]. Electronics, 2021, 10(21): 2689.
[53]
ShiG, ZhaoJ, WangH, et al. Characteristics and GWO-ANN model based on prior knowledge for contact resistance in a pantograph-catenary[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-11.
[54]
HattaN M, ZainA M, SallehuddinR, et al. Recent studies on optimisation method of grey wolf optimiser (GWO): a review (2014-2017)[J]. Artificial Intelligence Review, 2019, 52: 2651-2683.
[55]
HodsonT O. Root mean square error (RMSE) or mean absolute error (MAE): When to use them or not[J]. Geoscientific Model Development Discussions, 2022, 15(14): 5481-5487.
[56]
MyttenaereA, GoldenB, GrandB, et al. Mean absolute percentage error for regression models[J]. Neurocomputing, 2016, 192: 38-48.
[57]
SchuurmannG, EbertR U, ChenJ, et al. External validation and prediction employing the predictive squared correlation coefficient test set activity mean vs training set activity mean[J]. Journal of Chemical Information and Modeling, 2008, 48(11): 2140-2145.
LiuShanshan, TanYue, QiuChidong, et al. Research for electrolyte density measurement of battery based on ultrasonic[J]. China Instrumentation, 2007 (1): 30-32.