基于DWD-SVR模型的锂离子电池剩余使用寿命预测  

RUL PREDICTION FOR LITHIUM-ION BATTERIES BASED ON DWD-SVR MODEL

作　　者：王小明[1,2] 何叶 王路路 吴红斌 徐斌[2] 赵文广 Wang Xiaoming;He Ye;Wang Luu;Wu Hongbin;Xu Bin;Zhao Wenguang(Anhui Province Key Laboratory of Renewable Energy Utilization and Energy Saving(Hefei University of Technology),Hefei 230009,China;Electric Power Research Institute of State Grid Anhui Electric Power Co.,Ltd.,Hefei 230601,China)

机构地区：[1]新能源利用与节能安徽省重点实验室(合肥工业大学),合肥230009 [2]国网安徽省电力有限公司电力科学研究院,合肥230601

出　　处：《太阳能学报》2025年第2期52-59,共8页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金区域创新发展联合基金(U19A20106);安徽省高校协同创新项目(GXXT-2022-023)。

摘　　要：针对锂离子电池容量退化特性的非线性和多尺度特性,提出一种基于离散小波分解(DWD)和支持向量回归(SVR)模型的锂离子电池RUL预测方法。首先,利用DWD对容量时间序列进行多尺度解耦,以降低局部再生和波动现象对预测结果的影响;其次,利用K-均值聚类方法将各尺度信号中样本熵与排列熵相近的子序列进行聚类,根据聚类结果将复杂度与随机性相近的子序列进行重构,以减少建模次数,提高预测效率;最后,通过SVR预测模型精确捕捉不同尺度下容量信号的变化情况,实现电池RUL准确预测。实验结果表明,提出的基于DWD-SVR模型的锂离子电池RUL预测方法能在保证全局退化趋势预测准确性的同时对波动进行及时地响应,可提高预测性能。The remaining useful life(RUL)prediction of Lithium-ion batteries is crucial to ensure the safe and stable operation of Lithium-ion battery energy storage devices.Aiming at the problem of low RUL prediction accuracy due to the non-linear and multi-scale characteristics of the capacity degradation characteristics of Lithium-ion batteries,a RUL prediction method for Lithium-ion batteries based on the discrete wavelet decomposition(DWD)and support vector regression(SVR)model is proposed in the paper.First,the global degradation trend,local regeneration and fluctuation components of capacity are extracted using discrete wavelet decomposition,which can reduce the influence of local regeneration and fluctuation phenomena on the machine learning algorithm to predict the capacity degradation process.Then,each decomposition subsequence is reconstructed based on sample entropy,alignment entropy and K-means clustering method to reduce the number of local regeneration and fluctuating subsequences in the decomposition signals and improve the prediction efficiency.Experimental results based on the NASA lithium-ion battery dataset show that the proposed prediction method is able to ensure the accuracy of global degradation trend prediction while responding to fluctuations in a timely manner to improve the RUL prediction accuracy.

关 键 词：锂离子电池 支持向量回归 K-均值聚类 剩余使用寿命 离散小波分解 

分 类 号：TM911.3[电气工程—电力电子与电力传动]