基于一种改进的RBF神经网络的直接甲醇燃料电池建模

针对直接甲醇燃料电池(DMFC)系统过于复杂，难以数学建模。应用一种改进的RBF神经网络对DMFC系统进行辨识建模。模型以甲醇的浓度和流速为神经网络辨识模型的输入量，电池电压／电流密度为输出量，利用1000组实验数据作为训练和测试样本，建立了不同甲醇浓度和流速下电池电压／电流密度动态响应模型。应用仿真对建模的有效性和精度进行了检验，并与BP神经网络辨识的效果进行了对比。仿真结果证明RBF神经网络比BP神经网络收敛得快，建模精度高，从而为设计DMFC实时控制系统奠定了基础。

Modeling Direct Methanol Fuel Cell (DMFC) Based on Improved RBF Neural Networks Identification

For the inner complexity of Direct Methanol Fuel cell (DMFC), mathematical modeling of DMFC is very difficult. We try to establish the voltage and current dynamic model of DMFC by using an improved RBF neural networks identification technique. The concentration and flow rate of Methanol are taken as the input and the voltage and current response as the output of the neural networks model. 1000 groups of experimental data are used, and the structure and an improved RBF algorithm of neural networks identification system are described. The validity and accuracy of modeling are tested by simulations, and the simulation results of the comparison between RBF neural networks and BP neural networks identification are given. The simulation tests demonstrate that the modeling method is feasible, and that the training of RBF networks is much faster than BP networks. The RBF neural networks identification model established above makes it possible to design on-line controller of DMFC.