引用本文:商云龙,张承慧,崔纳新,张奇.基于模糊神经网络优化扩展卡尔曼滤波的锂离子电池荷电状态估计[J].控制理论与应用,2016,33(2):212~220.[点击复制]
SHANG Yun-long,ZHANG Cheng-hui,CUI Na-xin,ZHANG Qi.State of charge estimation for lithium-ion batteries based on extended Kalman filter optimized by fuzzy neural network[J].Control Theory and Technology,2016,33(2):212~220.[点击复制]
基于模糊神经网络优化扩展卡尔曼滤波的锂离子电池荷电状态估计
State of charge estimation for lithium-ion batteries based on extended Kalman filter optimized by fuzzy neural network
摘要点击 4496  全文点击 2516  投稿时间：2014-12-15  修订日期：2015-09-01
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2016.41167
  2016,33(2):212-220
中文关键词  动力电池  SOC估计  模型误差  模糊神经网络  扩展卡尔曼滤波
英文关键词  power battery  SOC estimation  model error  fuzzy neural networks  extended Kalman filters
基金项目  国家自然科学基金项目(61527809, 61273097, 51277116, 61403162), 国家自然科学基金重大国际(地区)合作研究项目(61320106011)资助
作者单位E-mail
商云龙 山东大学 shangyunlong@mail.sdu.edu.cn 
张承慧* 山东大学 zchui@sdu.edu.cn 
崔纳新 山东大学  
张奇 山东大学  
中文摘要
      电池荷电状态(state of charge, SOC)的精确估计是判断电池是否过充或过放的重要依据, 是电动汽车安 全、可靠运行的重要保障. 传统基于扩展卡尔曼滤波(extended Kalman filter, EKF)的SOC估计方法过度依赖于精确 的电池模型, 并且要求系统噪声必须服从高斯白噪声分布. 为解决上述问题, 基于模糊神经网络(fuzzy neural network, FNN)建立模型误差预测模型, 并藉此修正扩展卡尔曼滤波测量噪声协方差, 以实现当模型误差较小时对 状态估计进行测量更新, 而当模型误差较大时只进行过程更新. 仿真和实验结果表明, 该算法能有效消除由于模型 误差和测量噪声统计特性不确定而引入的SOC估计误差, 误差在1:2%以内, 并且具有较好的收敛性和鲁棒性, 适用 于电动汽车的各种复杂工况, 应用价值较高.
英文摘要
      The accurate estimation for state of charge (SOC) is the important basis to prevent overcharge or overdischarge of batteries, and is the important guarantee for the electric vehicle safety and reliability. In the traditional SOC estimation methods based on extended Kalman filter (EKF), the SOC estimation precision was highly dependent on an accurate battery model. To solve the above problems, an error prediction model was built based on fuzzy neural network (FNN), by which the measurement noise covariance of EKF was real-time revised. When the predicted model error was small, the measurement model was updated, otherwise, the process model was updated only. The simulation and experimental results show that the proposed algorithm can effectively eliminate the SOC estimation error caused by the model error and the uncertain noise statistical properties, with the maximum error of less than 1:2%. The proposed algorithm has good convergence and robustness, and is applicable to various complicated driving cycles for electric vehicles, with high application value.