储能蓄电池电能动态模型辨识与SOC预测  

Energy Storage Battery Dynamic Model Identification and SOC Prediction

作　　者：郑坤[1] 赵志刚[1] 王思远[1] 王龙军 

机构地区：[1]兰州交通大学机电工程学院,甘肃兰州730070

出　　处：《计算机仿真》2017年第9期120-124,共5页Computer Simulation

摘　　要：在蓄电池能量管理系统中,根据蓄电池外部电化学特性建立蓄电池模型和估计其动态容量值是一个挑战性问题,也是一个急需解决的问题,为了获得蓄电池荷电状态(SOC)准确估计,为蓄电池管理提供理论和实践基础。针对蓄电池极化、滞后等非线性动态特性,RC电路方程引入到蓄电池模型;在一阶RC模型基础上分别采用Laplace变换和前向欧拉转换算法得到适用于最小二乘辨识算法的递推模型,针对辨识过程中前后采样点相互影响和"数据饱和"现象,在递推最小二乘算法基础上加入遗忘因子;模拟蓄电池实际工况制定实验方案,根据实验数据辨识得到模型参数;基于得到蓄电池模型推导得到适用于无迹卡尔曼滤波算法的状态空间方程形式,采用无迹卡尔曼滤波算法估计得到蓄电池动态荷电状态。仿真结果表明,一阶RC电路模型可以很好的仿真蓄电池实际工况,基于上述模型得到蓄电池动态容量估计值。In order to attain estimation of the battery charged state accurately and provide theoretical and practical basis for the battery management system, in the first place, accurate battery model is set up. Aiming at the dynamic characteristics such as battery polarization and lag, RC circuit equation is introduced into the battery model. In the meantime, battery impedance transfer function was obtained by Laplace transform based on first-order RC model. Forward Euler transformation algorithm was used to attain recursive model which is suitable for the Least Squares i- dentification algorithm. The forgetting factor was added to Recursive Least Squares algorithm to avoid mutual influence of sampling points in the process of identification and data saturation phenomenon. Then the experiment solution of battery was made based on its actual working condition, and the method of nonlinear online identification function was used to extract model parameters in real time by using the forgetting factor least square algorithm. And then, the form of the state space equation, which is suitable for Unscented Kalman filtering algorithm, was derived based on the bat- tery model. The model estimation function of Unscented Kalman filtering algorithm was adopted to attain state of charge estimation. With state of charge estimation of battery compared with its actual state of charge estimation, the identifiction precision was verified. The results show that the First-order RC circuit model is not only closer to the ac- tual working condition of battery, but also can estimate the SOC values.

关 键 词：电池模型 最小二乘 卡尔曼滤波 辨识 

分 类 号：TM912[电气工程—电力电子与电力传动]