汽车工程 ›› 2023, Vol. 45 ›› Issue (2): 199-208.doi: 10.19562/j.chinasae.qcgc.2023.02.005

所属专题: 新能源汽车技术-动力电池&燃料电池2023年

• • 上一篇    下一篇

基于非线性降维IC特征的实车电池SOH估计

陈吉清,李子涵,兰凤崇(),蒋心平,潘威,陈继开   

  1. 1.华南理工大学机械与汽车工程学院,广州  510640
    2.华南理工大学,广东省汽车工程重点实验室,广州  510640
  • 收稿日期:2022-07-25 修回日期:2022-08-20 出版日期:2023-02-25 发布日期:2023-02-21
  • 通讯作者: 兰凤崇 E-mail:fclan@scut.edu.cn

Real-Vehicle Battery Health State Estimation Based on Nonlinear Reduced-Dimensional IC Features

Jiqing Chen,Zihan Li,Fengchong Lan(),Xinping Jiang,Wei Pan,Jikai Chen   

  1. 1.School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou  510640
    2.South China University of Technology,Guangdong Provincial Automobile Engineering Key Laboratory,Guangzhou  510640
  • Received:2022-07-25 Revised:2022-08-20 Online:2023-02-25 Published:2023-02-21
  • Contact: Fengchong Lan E-mail:fclan@scut.edu.cn

摘要:

基于实车电池复杂的运行数据,本文使用增量容量分析方法提取IC峰特征作为电池充电片段的有效特征,使用t-SNE非线性降维方法处理IC峰特征,消除多维特征冗余性,以解决实车数据难以提取可靠特征的问题。另外构建支持向量回归模型,实现对电池健康状态的估计。结果表明,本文使用的增量容量曲线峰特征能有效表征电池健康状态衰退变化;对实车数据的平滑、降噪方法可以较好地提升训练数据质量;基于t-SNE降维特征的SVR模型提升了对电池健康状态的估计精度,保证了有限样本数据集上实现准确估计。

关键词: 动力电池, SOH估计, 非线性降维, 机器学习

Abstract:

Based on the complex operation data of real-vehicle batteries, the IC peak features are extracted as effective features of battery charging segments using incremental capacity analysis method in this paper, and the IC peak features are processed using t-SNE nonlinear dimensionality reduction method to eliminate the redundancy of multidimensional features and solve the problem that it is difficult to extract reliable features from real-vehicle data. In addition, a support vector regression model is constructed to estimate the battery health status. The results show that the use of incremental capacity curve peak features can effectively characterize the recession changes of the battery health state. The smoothing and noise reduction methods for real vehicle data can improve the quality of training data better. The SVR model based on t-SNE dimensionality reduction features improves the estimation accuracy of battery health state and ensures accurate estimation on a limited sample data set.

Key words: power battery, SOH estimation, nonlinear dimensionality reduction, machine learning