汽车工程 ›› 2021, Vol. 43 ›› Issue (1): 10-18.doi: 10.19562/j.chinasae.qcgc.2021.01.002

• • 上一篇    下一篇

用新陈代谢极限学习机实现电池健康状态估算

陈琳1,2,王惠民1,李熠婧1,张沫1,黄江1,潘海鸿1,2()   

  1. 1.广西大学机械工程学院,南宁 530004
    2.广西电化学能源材料重点实验室(广西大学可再生能源材料协同创新中心),南宁 530004
  • 收稿日期:2020-04-29 修回日期:2020-07-10 出版日期:2021-01-25 发布日期:2021-02-03
  • 通讯作者: 潘海鸿 E-mail:hustphh@163.com
  • 基金资助:
    国家自然科学基金(51667006)

Battery State⁃of⁃Health Estimation by Using Metabolic Extreme Learning Machine

Lin Chen1,2,Huimin Wang1,Yijing Li1,Mo Zhang1,Jiang Huang1,Haihong Pan1,2()   

  1. 1.School of Mechanical Engineering,Guangxi University,Nanning 530004
    2.Guangxi Key Laboratory of Electrochemical Energy Materials (Collaborative Innovation Center of Sustainable Energy Materials),Nanning 530004
  • Received:2020-04-29 Revised:2020-07-10 Online:2021-01-25 Published:2021-02-03
  • Contact: Haihong Pan E-mail:hustphh@163.com

摘要:

针对因电池内部电化学反应的复杂性、算法泛化性差或可用已知数据量少导致的锂离子电池SOH估算精度下降的问题,提出使用极限学习机(ELM)构建强泛化性电池退化状态模型来描述不同电池的共性退化规律;引入新陈代谢机制来更新退化状态模型的输入数据进而实现对SOH的新陈代谢估算,在保证估算精度的同时降低对输入数据量的需求。利用两种不同材料电池在不同工况下所测数据对所提出方法进行验证,结果表明该方法能在仅用4个数据样本的情况下准确估算电池SOH,估算误差不超过2.18%。

关键词: 锂离子电池, 健康状态, 极限学习机, 退化状态模型, 新陈代谢机制

Abstract:

In view of the low state of health (SOH) estimation accuracy of lithium?ion battery due to the complexity of internal electro?chemical reaction of battery, the poor generalization of the algorithm or the small amount of available data samples, a strongly generalized degradation state model of batteries is constructed based on the extreme learning machine (ELM) to describe the common degradation regularities of different batteries. Moreover, the metabolic mechanism is introduced to update the input data of the degradation state model and then to realize the metabolic estimation of SOH, which reduces the demand for the amount of input data, while ensuring the accuracy of estimation. The method proposed is verified by using the measured data of two batteries with different materials under different working conditions. The results show that the method proposed can accurately estimate the SOH of battery with only four data samples, and the estimation errors are no more than 2.18%.

Key words: lithium?ion battery, state of health, extreme learning machine, degradation state model, metabolic mechanism