基于AG-BP算法的锂电池模型参数在线估计

On-line estimation of lithium battery model parameters based on AG-BP algorithm

对现有锂电池DP(dual-polarized)模型及其参数辨识方法进行研究,提出基于AG-BP算法的双极化锂电池模型参数在线辨识方法。结合混合动力脉冲测试和MATLAB CF曲线拟合工具箱得到神经网络的训练及测试数据库。结合正弦脉冲电流注入法,将注入电池的电流幅值、频率、电池电压响应幅值、温度、SOC(state of charge)作为BP神经网络的输入参数,实现了基于BP (back propagation)神经网络的锂电池DP(double polarization)模型的参数估计。将BP神经网络的权值矩阵、阈值向量作为基因编码,运用遗传算法实现了对神经网络的初始权值和阈值进行优化,提高了参数估计精度,并通过仿真实验对系统的参数估计精度进行了验证。

This paper studies the existing dual-polarized lithium battery model and its parameter identification method,and proposes an online parameter identification method based on AG-BP algorithm.The training and testing database of neural network is obtained by combining hybrid impulse test and MATLAB CF curve fitting toolbox.Combined with the sinusoidal pulse current injection method,the current amplitude,frequency,voltage response amplitude,temperature and SOC(state of charge) injected into the battery were taken as input parameters of the BP neural network to realize the parameter estimation of the double polarization model of lithium battery based on the BP(back propagation) neural network.The weight matrix and threshold vector of BP neural network were taken as gene coding,and the initial weight and threshold of BP neural network were optimized by genetic algorithm to improve the accuracy of parameter estimation.