Multi-kernel SVM Analysis Method for Remaining Capacity in Battery Health Management

杨军; 王仲夏; 许云辰; <table></table>

English Version Journal Login

您当前的位置：

首页 >

文章列表页 >

Multi-kernel SVM Analysis Method for Remaining Capacity in Battery Health Management

更新时间：2026-05-28

- Multi-kernel SVM Analysis Method for Remaining Capacity in Battery Health Management
- Vol. 19, Pages: 1-7(2027)
- 作者机构：
  
  1. 国电南京自动化股份有限公司,南京,210003
  2. 皖江工学院电气与信息工程学院,安徽,马鞍山,243031
- 作者简介：
- 基金信息：
- DOI：
  CLC：
- Received：08 April 2026，
  
  Accepted：06 May 2026，
  
  Online First：28 May 2026，
  
  Published：2027
- 稿件说明：
移动端阅览
杨军, 王仲夏, 许云辰, et al. Multi-kernel SVM Analysis Method for Remaining Capacity in Battery Health Management[J/OL]. 2027, 191-7.
DOI：

杨军, 王仲夏, 许云辰, et al. Multi-kernel SVM Analysis Method for Remaining Capacity in Battery Health Management[J/OL]. 2027, 191-7. DOI：

摘要

为解决锂离子电池剩余容量预测中非线性特征提取不足、参数寻优易陷入局部最优、预测精度偏低的问题，提升电池健康管理的可靠性，本文提出混沌步长粒子群（CSPSO）优化多核支持向量机的电池容量预测算法，构建拉普拉斯核函数与径向基函数（RBF）核函数融合的多核SVM 模型，利用CSPSO算法实现模型关键参数全局寻优，采用均方误差倒数法自适应分配核函数子模型权重。结果表明：相较于传统PSO多核SVM预测方法，混沌步长粒子群算法凭借混沌动态调整学习因子与步长的特性，有效避免陷入局部最优，提升参数寻优精度。相较于单一核SVM预测方法，拉普拉斯核凭借适配高维稀疏数据的特性，可精准捕捉电池退化全局特征，RBF核凭借对局部非线性波动敏感的特性，可高效拟合容量衰减局部规律，二者融合大幅提升非线性特征提取能力。均方误差倒数法具备自适应权重分配的特点，避免等权重融合带来的性能平均化缺陷，提升模型整体预测精度与鲁棒性。CSPSO优化多核支持向量机的容量预测算法可有效提取电池退化多维非线性特征，参数寻优精度高、预测结果稳定，为锂离子电池剩余容量精准预测与健康状态管理提供了一种可靠的新途径。

Abstract

To address the problems of insufficient nonlinear feature extraction

susseptibility to local optimum trapping in parameter optimization and low prediction accuracy in lithium-ion battery remaining capacity prediction

and improve the reliability of battery health management

this paper proposes a battery capacity prediction algorithm based on multi-kernel support vector machine (SVM) optimized by chaotic step particle swarm optimization (CSPSO). A multi-kernel SVM model integrating Laplacian kernel and radial basis function (RBF) kernel is constructed. The CSPSO algorithm is adopted for global optimization of key model parameters

and the reciprocal mean square error method is used to adaptively assign weights to sub-models of kernel functions. The results show that

compared with the traditional PSO-based multi-kernel SVM prediction method

CSPSO dynamically adjusts learning factors and step sizes via chaos theory

which effectively avoids local optima and improves parameter optimization accuracy. In contrast to single-kernel SVM

the Laplacian kernel adapts well to high-dimensional sparse data and accurately captures global battery degradation characteristics

while the RBF kernel is sensitive to local nonlinear fluctuations and efficiently fits local capacity fading rules. The fusion of the two kernels greatly enhances nonlinear feature extraction capability. The reciprocal mean square error method realizes adaptive weight allocation

eliminates the performance averaging defect caused by equal-weight fusion

and boosts the overall prediction accuracy and robustness of the model. The proposed CSPSO-optimized multi-kernel SVM capacity prediction algorithm can effectively extract multi-dimensional and nonlinear battery degradation features with high parameter optimization precision and stable prediction performance. It provides a novel and reliable approach for accurate remaining capacity prediction and health state management of lithium-ion batteries.

关键词

Keywords

references

赵璐,吴鹏江,齐绍通,等.基于多物理场耦合的磷酸铁锂电池组热失控及其蔓延特性仿真分析[J].能源研究与管理,2026,18(1):7-15.

康振华,凌鑫晨,万里平,等.基于活化能的功率型动力电池寿命影响因素分析[J].能源研究与管理,2024,16(3):79-84.

高国章,刘长昊,袁裕鹏,等.基于PKO-XGBoost-TCN 的锂电池SOH 估计[J].电源技术,2025,49(12):2553-2562.

杜江龙,王旭峰,陈润凯.磷酸铁锂电池正极压实密度的多维度作用及对循环老化的影响[J].能源研究与管理,2026,18(1):89-97.

夏宇.基于健康特征变换的锂离子电池健康状态评估及预测方法[D].重庆:重庆大学,2023.

邓昊,杨林,邓忠伟,等.基于电化学机理模型的锂离子电池参数辨识及SOC 估计[J].上海理工大学学报,2018,40(6):557-565.

何法,韩志,李彦超,等.基于遗传算法优化BP 神经网络的锂电池容量预测研究[J].汽车技术,2026(1):45-50.

王瑞,陈泽坤,逯静.基于VMD 和IBA-LSSVM 的短期风电功率预测[J].河海大学学报(自然科学版),2021,49(6):575-582.

李雨佳,欧阳权,刘灏仪,等.基于支持向量机与改进高斯过程混合模型的车用电池容量预测方法[J].电气工程学报,2024,19(1):87-96.

蔡仕云,吴肖龙,李豫,等.基于融合聚类和RBF-SVM 的固体氧化物燃料电池系统故障识别[J].武汉大学学报(理学版),2025,71(6):841-850.

鞠红梅,周菲,孟凡.基于模糊隶属度与粒子群算法协同优化的双子支持向量机鲁棒性分类方法[J/OL].系统科学与数学,(2025-09-17).https://link.cnki.net/urlid/11.2019.o1.20250916.1042.002.

Pan Jing,Tong Shu,Li Yang,et al.State of health estimation for lithium-ion batteries based on improved Gaussian process regression[C].2021 IEEE 12th Energy Conversion Congress & Exposition,2021:1-6.

Putra M R A,Nizam M,Mujiato A,et al.Robust SVM optimization using PSO and ACO for accurate lithium-ion battery health monitoring [J].Mechanical Engineering for Society & Industry,2025,5(1):1-12.

刘勇,江沙里,廖士中.基于近似高斯核显式描述的大规模SVM 求解[J].计算机研究与发展,2014,51(10):2171-2177.

Baker J S,Park G,Yu K,et al.Parallel hybrid quantum-classical machine learning for kernelized time-series classification [J].Quantum Machine Intelligence,2024,6(1):1-18.

Gao Dong,Huang Miaohua.Prediction of remaining useful life of lithium-ion battery based on multi-kernel support vector machine with particle swarm optimization [J].Journal of Dalian University of Technology,2017,17 (5):1288-1297.

李丽,车金香,赵云,等.基于Subbagging 的支持向量回归集成及应用[J].数学的实践与认识,2019,49(15):167-172.

刘鑫,李彬周,王军生,等.一种改进型拉普拉斯核的GRNN模型在重轨性能预测中的应用[J].铁道科学与工程学报,2025,22(8):3471-3480.

胡林静,李在伟.基于灰狼优化算法和组合核函数GPR 模型的锂电池剩余使用寿命预测[J].综合智慧能源,2025,47(12):25-33.

潘冯超,刘勤明,叶春明,等.基于改进萤火虫算法的分布式能源供应链配置研究[J].能源研究与信息,2022,38(1):46-52.

Williard N,He Wei,Osterman M,et al.Comparative analysis of features for determining state of health in lithium-ion batteries[J].International Journal of Prognostics and Health Management,2013,4(1):14-20.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Research Progress of Additives for Li-ion Battery Electrolytes

氢气存储与加注设备技术现状及发展趋势

超临界水气化驱动氢冶金（竖炉-电弧炉）：低碳化路径与技术挑战

能源电力期刊集群

E方知库期刊群

Related Author

陈高明

胡立新

王超

Related Institution

杭州中美华东制药有限公司

湖北工业大学化学与环境工程学院

AI问答

⁰