电动车蓄电池荷电状态估计的神经网络方法

针对电动车蓄电池电能容量判别问题,将神经网络方法应用于电动车蓄电池荷电状态估计.对多种神经网络方法的估计性能进行了分析,包括多层感知器网络、径向基函数网络、线性支持向量机、使用MLP核函数的支持向量机、使用RBF核函数的支持向量机.实验结果表明:神经网络经过训练后,可以通过蓄电池的工作电压、工作电流和表面温度参数估计蓄电池的SOC实时值,其中多层感知器和支持向量机估计性能最好,同时,支持向量机较多层感知器有更高的噪声容忍能力.

Battery state of charge Estimation Basedon Neural-Network for Electric Vehicles

In this paper, the neural-network (NN) is proposed based approach to estimate the battery's state of charge (SOC) of the electric vehicles (EVs). In this approach, NNs are trained to extract important features from the battery's voltage, current, and temperature to estimate the battery's SOC. Such automated, noninvasive estimation will be critical in future EVs' energy monitoring and enhancement systems. Several NN-based estimation models including multilayer perceptron (MLP) network, radial basis function (RBF) network, linear support vector machines (SVM) network, and support vector machines network with MLP, and RBF kernels are developed for SOC estimation. The performance of these estimators is compared in terms of their accuracy and noise tolerance limits. The results showed that MLP and SVM are both able to estimate the SOC with high accuracy. SVM is found to be the best estimation method because of its high noise tolerating ability.