基于MIV-GA-BP神经网络的铅酸蓄电池SOC预测

SOC estimation of lead-acid battery based on MIV-GA-BP neural network

建立了铅酸蓄电池充电过程中SOC的神经网络预测模型,采用平均影响值(MIV)算法对预测模型的输入变量进行了分析和筛选。在MIV算法的基础上,比较了基于遗传算法优化的BP神经网络(MIV-GA-BP)与传统MIV-BP神经网络对蓄电池充电过程中SOC的预测误差。测试样本的验证结果表明,MIV-GA-BP神经网络模型对蓄电池充电过程的SOC预测精度更优。

A neural network model was established to predict the SOC of lead-acid battery in charging process,and the input variables of the prediction model were analyzed and selected by using mean impact value(MIV)algorithm.On the basis of MIV algorithm,the prediction errors of SOC in the battery charging process were compared between the BP neural network optimized by genetic algorithm(MIV-GA-BP)and the traditional MIV-BP neural network.The verification of the testing samples shows that the prediction accuracy of MIV-GA-BP neural network is better,and the SOC prediction in lead-acid battery charging process can be realized.