汽车工程 ›› 2021, Vol. 43 ›› Issue (10): 1472-1478.doi: 10.19562/j.chinasae.qcgc.2021.10.008

• • 上一篇    下一篇

采用粒子群优化和高斯回归实现电池SOH估计

陈琳,刘博豪,丁云辉,吴淑孝,冯喆,潘海鸿()   

  1. 广西大学机械工程学院,南宁 530004
  • 收稿日期:2021-04-07 修回日期:2021-05-20 出版日期:2021-10-25 发布日期:2021-10-25
  • 通讯作者: 潘海鸿 E-mail:hustphh@163.com
  • 基金资助:
    国家自然科学基金(52067003)

Estimation of Battery State⁃of⁃Health Using Particle Swarm Optimization with Gauss Process Regression

Lin Chen,Bohao Liu,Yunhui Ding,Shuxiao Wu,Zhe Feng,Haihong Pan()   

  1. School of Mechanical Engineering,Guangxi University,Nanning 530004
  • Received:2021-04-07 Revised:2021-05-20 Online:2021-10-25 Published:2021-10-25
  • Contact: Haihong Pan E-mail:hustphh@163.com

摘要:

为准确估算锂离子电池非线性退化过程中的健康状态(SOH),提出融合自适应变异粒子群优化器和高斯过程回归的AMPSO?GPR算法。首先提取欧姆内阻增量和电压样本熵作为电池退化表征量,然后引入自适应变异粒子群(AMPSO)优化高斯过程回归(GPR)核函数的超参数,构建基于AMPSO?GPR的SOH估算框架,用提取的退化表征量实现SOH估算;最后,通过对比AMPSO?GPR采用不同核函数时SOH估算结果,得到最优核函数。实验结果表明,AMPSO?GPR算法可以有效地估算电池SOH,且最大估算误差不超过2.08%。

关键词: 锂离子电池, 健康状态, 高斯过程回归, 自适应变异粒子群优化器, 核函数

Abstract:

In order to accurately estimate the state of health (SOH) of lithium?ion battery in the process of nonlinear degradation, an AMPSO?GPR algorithm is proposed by the fusion of adaptive mutation particle swarm optimizer (AMPSO) with gaussian process regression (GPR). Firstly, the increment of ohmic internal resistance and the sample entropy of voltage are extracted as degradation characterization indicators. Then AMPSO is introduced to optimize the hyperparameters of GPR kernel function, an SOH estimation framework based on AMPSO?GPR is constructed, and the degradation characterization indicators are extracted to perform SOH estimation. Finally, by comparing the results of SOH estimation using AMPSO?GPR with different kernel functions, the optimal kernel function is obtained. The results of experiment indicate that the AMPSO?GPR algorithm can effectively estimate the SOH of battery with a maximum absolute estimation error not more than 2.08%.

Key words: lithium?ion battery, state?of?health, gaussian process regression, adaptive mutation particle swarm optimizer, kernel functions