基于短期充电数据和增强鲸鱼优化算法的锂离子电池容量预测  

Lithium battery capacity prediction based on short-term charging data and an enhanced whale optimization algorithm

作　　者：陈峥[1] 彭月 胡竞元 申江卫[1] 肖仁鑫[1] 夏雪磊 CHEN Zheng;PENG Yue;HU Jingyuan;SHEN Jiangwei;XIAO Renxin;XIA Xuelei(Faculty of Transportation Engineering,Kunming University of Science and Technology,Kunming 650500,Yunnan,China)

机构地区：[1]昆明理工大学交通工程学院,云南昆明650500

出　　处：《储能科学与技术》2025年第1期319-330,共12页Energy Storage Science and Technology

基　　金：国家自然科学基金项目(52267022)。

摘　　要：为解决采用数据驱动的方法对锂离子电池容量进行预测时,难以获取完整充电数据、数据采样精度低和特征因子提取质量不佳等问题,本工作提出了一种基于短期充电数据和增强鲸鱼优化算法的锂离子电池容量预测方法。首先,为提升数据精度,利用三次样条插值对充电数据进行补充。其次,通过挖掘充电电压曲线与容量衰退之间的规律,确定特征因子为某充电时间区间的电压增量,并利用增强鲸鱼算法,从短期充电数据中实现了老化特征的有效提取。随后,构建了高斯过程回归容量预测模型,在确定训练数据量后,对比了不同算法的预测结果,验证了所构建模型的有效性。最后,将该方法在不同电池上进行测试,验证了预测精度和泛化能力。结果表明:对于实验室数据集,将前15%老化特征作为训练集时,可将该类电池最大误差控制在2.49%以内,且97%的预测误差控制在1.5%内;对于公开数据集,仅12组训练数据就能将该类电池最大误差控制在1%以内,实现了利用低精度和短期充电数据对电池容量的准确预测。The prediction of lithium-ion battery capacity using data-driven methods involves many challenges,such as the difficulty in obtaining complete charging data,low data sampling precision,and poor quality of feature extraction.To overcome these challenges,this study proposes a lithium-ion battery capacity prediction method based on short-term charging data and an enhanced whale optimization algorithm.First,to improve data precision,the charging data were supplemented by cubic spline interpolation.Next,by exploring the relationship between the charging voltage curve and capacity degradation,the voltage increment over a specific charging time interval was identified as the feature factor.An enhanced whale optimization algorithm was then utilized to extract aging features effectively from the short-term charging data.Subsequently,a Gaussian process regression model was constructed for capacity prediction.After determining the amount of training data,the predictive results of different algorithms were compared to verify the effectiveness of the proposed model.Finally,the method was tested on different batteries to validate its prediction accuracy and generalization capability.The results demonstrate that for the laboratory dataset,by using the first 15%of aging features as the training set,the maximum error of this type of battery can be controlled within 2.49%,with 97%of the prediction errors being within 1.5%.For a publicly available dataset,by using only 12 groups of training data,the maximum error of this type of battery can be controlled within 1%,achieving accurate capacity prediction using low-precision and short-term charging data.

关 键 词：锂离子电池 短期充电数据 容量预测 增强鲸鱼优化算法 高斯过程回归 

分 类 号：TM912[电气工程—电力电子与电力传动]