储能科学与技术 ›› 2021, Vol. 10 ›› Issue (3): 1137-1144.doi: 10.19799/j.cnki.2095-4239.2020.0400

• 储能测试与评价 • 上一篇    下一篇

基于RLS-DLUKF算法的锂电池SOC预测方法研究

朱磊(), 刘子博, 李路路, 潘庭龙(), 杨玮林   

  1. 江南大学物联网应用技术教育部过程研究中心,江苏 无锡 214122
  • 收稿日期:2020-12-11 修回日期:2020-12-17 出版日期:2021-05-05 发布日期:2021-04-30
  • 通讯作者: 潘庭龙 E-mail:1270560524@qq.com;tlpan@jiangnan.edu.cn
  • 作者简介:朱磊(1994—),男,硕士研究生,研究方向为新能源控制及节能技术,E-mail:1270560524@qq.com
  • 基金资助:
    国家自然科学基金(61672266);北京市自然基金项目(21JC0026)

Research on a battery SOC prediction method based on the RLS-DLUKF algorithm

Lei ZHU(), Zibo LIU, Lulu LI, Tinglong PAN(), Weilin YANG   

  1. Engineering Research Center of Internet of Things Technology Applications Ministry of Education, Jiangnan University, Wuxi 214122, Jiangsu, China
  • Received:2020-12-11 Revised:2020-12-17 Online:2021-05-05 Published:2021-04-30
  • Contact: Tinglong PAN E-mail:1270560524@qq.com;tlpan@jiangnan.edu.cn

摘要:

本工作以钴酸电池为研究对象,针对锂电池在复杂工况下电流的剧烈变化导致荷电状态(SOC)无法有效预测的问题,建立了以精确参数为基础的双层无迹卡尔曼滤波算法(DLUKF)架构来更精准的估算SOC。首先,以脉冲功率特性实验数据获取二阶电路模型中电池开路电压与荷电状态的函数关系及特性曲线;其次,为增强模型辨识过程中的自适应学习能力并解决模型参数估计不准确的问题,应用递推最小二乘(RLS)算法在线准确地识别出模型中的未知变量;最后根据输入的变量信息,利用UKF算法相互嵌套形成的DLUKF算法实现对SOC的快速预测来解决单一的UKF算法在高阶非线性系统里估算不准确、精度低的问题。在UDDS工况和FUDS工况下对DLUKF算法和单一的UKF算法进行比较,通过对比分析两种算法估计出的SOC曲线、SOC误差曲线、端电压曲线及端电压误差曲线,表明DLUKF算法预测SOC的平均误差比UKF的低且预测精度更高。

关键词: SOC, RLS算法, DLUKF, 参数辨识, 二阶RC电路

Abstract:

With the lithium cobalt oxide battery as the research object, to solve the problem of the ineffective prediction of the state of charge (SOC) due to drastic changes in the battery current under complex working conditions, we established a double-layer unscented Kalman filter (DLUKF) architecture based on accurate parameters to estimate the value of SOC accurately. First, the functional relationship and characteristic curve between the open circuit voltage and SOC of the battery in the second-order circuit model were obtained from the experimental data of hybrid pulse power characteristics. Second, the recursive least squares method was applied to accurately identify the unknown variables in the model online to enhance the adaptive learning capability in the process of model identification and solve the problem of inaccurate model parameter estimation. Finally, the unknown variables in the model were identified accurately based on the output value of the model. Then, the DLUKF algorithm was used to realize the prediction of SOC fast to solve the problem of inaccurate estimation and large error of a single UKF algorithm in a strong nonlinear system. This paper compared the DLUKF algorithm with the single UKF algorithm under UDDS and FUDS conditions. The SOC curve, SOC error curve, terminal voltage curve, and terminal voltage error curve estimated by the two algorithms were compared. The results showed that the average error of DLUKF algorithm was lower than that of UKF algorithm, and when comparing predictions, the result of DLUKF algorithm was more accurate.

Key words: SOC, RLS algorithm, DLUKF, parameter identification, second-order RC circuit

中图分类号: