储能科学与技术 ›› 2022, Vol. 11 ›› Issue (1): 253-257.doi: 10.19799/j.cnki.2095-4239.2021.0297

• 储能测试与评价 • 上一篇    下一篇

基于改进的高斯过程回归的SOC估计算法

孙爱芬1(), 赤娜2   

  1. 1.郑州电力职业技术学院,河南 郑州 451450
    2.郑州工业应用技术学院,河南 郑州 451100
  • 收稿日期:2021-06-29 修回日期:2021-07-21 出版日期:2022-01-05 发布日期:2022-01-10
  • 通讯作者: 孙爱芬 E-mail:sunaifen@163.com
  • 作者简介:孙爱芬(1984—),女,讲师,研究方向为电气自动化,E-mail:sunaifen@163.com
  • 基金资助:
    郑州市社会科学调研课题《加快推动郑州创新型城市建设》(ZSLX2019336);2021年度河南省高等学校重点科研项目《发电厂直流系统故障主动监测与隔离技术研究及装置开发》(21B470012)

SOC estimation algorithm based on improved Gaussian process regression

Aifen SUN1(), Na CHI2   

  1. 1.Zhengzhou Electric Power Vocational and technical college, Zhengzhou 451450, Henan, China
    2.Zhengzhou Institute of industrial application technology, Zhengzhou 451100, Henan, China
  • Received:2021-06-29 Revised:2021-07-21 Online:2022-01-05 Published:2022-01-10
  • Contact: Aifen SUN E-mail:sunaifen@163.com

摘要:

为了提高锂离子电池荷电状态(SOC)的估计精度,文中采用基于高斯过程回归(GPR)机器学习的锂离子电池数据驱动方法,首先选取数据集,将电池测量参数电流和电压作为模型的输入向量,SOC作为模型的输出向量来训练模型,为了提高模型精度,文中改进了高斯过程回归模型。将上一时刻估计的SOC值加入到移动窗口中,并与电流和电压一起作为输入向量。通过窗口的大小不断更新训练集,从而训练出高精度SOC估计模型。通过实验采集的数据,并和GPR、最小二乘支持向量机(LSSVM)、支持向量机(SVM)和神经网络(NN)相比,所提模型估计的SOC均方根误差(RMSE)控制在1.5%以内,验证了提出方法的有效性。

关键词: SOC, 高斯过程回归, 电压, 电流

Abstract:

A data-driven method based on Gaussian process regression (GPR) machine learning is adopted herein to improve the estimation accuracy of the state of charge (SOC) of lithium-ion batteries. The current and the voltage measured by the battery are taken as the input vectors of the model, while the SOC is taken as the output vector of the model for model training. The GPR model is improved to improve the model accuracy. The SOC-estimated values are then added to the moving window and used as the input vectors together with the current and the voltage. A high-precision SOC estimation model is trained by updating the training set with the window size. Compared with the GPR, least square support vector machine, support vector machine, and neural network, the root mean square error of the SOC estimated by the proposed model is controlled within 1.5%, verifying the effectiveness of the proposed method.

Key words: SOC, gauss process regression, voltage, current

中图分类号: