上海交通大学学报 ›› 2024, Vol. 58 ›› Issue (9): 1454-1464.doi: 10.16183/j.cnki.jsjtu.2023.051

• 新型电力系统与综合能源 • 上一篇    下一篇

基于RBF-BLS面向电动汽车低碳安全出行的SOH估计方法

李春喜1, 乔涵哲2, 姚刚1, 姜淏予3(), 崔向科4, 葛泉波5   

  1. 1.上海海事大学 物流工程学院,上海 200135
    2.杭州电子科技大学 自动化学院,杭州 310018
    3.广东海洋大学 电子与信息工程学院, 广东 湛江 524008
    4.北京交通大学 经济与管理学院,北京 102603
    5.同济大学 电子与信息工程学院,上海 201804
  • 收稿日期:2023-02-15 修回日期:2023-02-15 接受日期:2023-02-24 出版日期:2024-09-28 发布日期:2024-10-11
  • 通讯作者: 姜淏予,副研究员,电话(Tel.):0759-2383761;E-mail:john_h_y_walter@163.com.
  • 作者简介:李春喜(1982—),博士生,从事充电安全研究.
  • 基金资助:
    国家自然科学基金(61803136)

SOH Estimation Method Based on RBF-BLS for Low-Carbon and Safe Travel of Electric Vehicle

LI Chunxi1, QIAO Hanzhe2, YAO Gang1, JIANG Haoyu3(), CUI Xiangke4, GE Quanbo5   

  1. 1. Logistics Engineering College, Shanghai Maritime University, Shanghai 200135, China
    2. School of Automation, Hangzhou Dianzi University, Hangzhou 310018, China
    3. School of Electronic and Information Engineering, Guangdong Ocean University, Zhanjiang 524008, Guangdong, China
    4. School of Economics and Management, Beijing Jiaotong University, Beijing 102603, China
    5. College of Electronic and Information Engineering, Tongji University, Shanghai 201804, China
  • Received:2023-02-15 Revised:2023-02-15 Accepted:2023-02-24 Online:2024-09-28 Published:2024-10-11

摘要:

电动汽车充电过程的安全性与动力电池组的健康状态(SOH)紧密相关,因此SOH的高性能实时估计是充电过程中安全检测的重要基础.由于动力电池组的SOH受复杂结构、电芯类型、驾驶习惯、环境温度和充电行为等因素的深度影响,现有基于单个或少量特定电池电芯实验数据的方法研究在面对整车动力电池组实时SOH估计时遭遇模型复杂、数据缺失、实时性差、精度不足等难题.针对建模困难、实时性和精度不足等问题,应用多方法集成融合思想,在电池经验退化模型上引入径向基函数(RBF)优化的宽度学习(BLS)神经网络,提出一种高性能的动力电池组SOH估计方法.首先,该方法采用经验退化模型和离线历史充电数据得到初步的SOH值;其次,应用RBF神经网络给出一种BLS系统中初始权重矩阵的确定方法,建立经验退化与径向基函数优化的宽度学习神经网络(RBF-BLS);再次,采用RBF-BLS神经网络和实时充电数据训练得到估计误差,并对经验退化模型得到的SOH进行补偿,从而得到更高精度的SOH估计值;最后,采用基于充电运营企业实际充电数据的计算机仿真实例来验证新方法的有效性和优越性.

关键词: 充电安全, 健康状态, 经验退化模型, 宽度学习

Abstract:

The charging safety of electric vehicle (EV) is closely related to the state of health (SOH) in power battery pack. Therefore, the high-performance and real-time estimation of SOH is an important basis for safety detection in the charging process. Power battery is deeply effected by factors such as complex structure, types of battery cell, driving habits, temperature, and charging behavior. Compared to SOH estimation methods based on experimental data from one or few battery cells, research on real-time SOH estimation of the power battery meets with insufficient problems in battery model, data getting, real-time, accuracy, and so on. Aimed at these drawbacks, a high performance SOH estimation method in power battery pack is proposed by introducing the broad learning system(BLS) optimized by radial basis function (RBF) into the empirical battery degradation model based on the idea of multi-method integration and fusion. First, the empirical degradation model and offline historical charging data are used to obtain the initial SOH value. Then, a radial basis function neural network is applied to get the initial weight matrix of the BLS to optimize the BLS method, and establish the RBF-BLS neural network. The estimation error can be trained by the RBF-BLS neural network and real-time charging data, and compensate for the initial SOH to gain a higher precise SOH value. Finally, a computer simulation example based on actual charging data from a charging operation enterprises is used to verify the effectiveness and superiority of the proposed method.

Key words: charging safety, state of health (SOH), empirical degradation model, broad learning system (BLS)

中图分类号: