基于改进型RBF神经网络的PEM FC电特性建模

Electric Characteristic Modeling of PEMFC Based on Improved RBFNN

常规能源短缺的今天,开发利用新型清洁、绿色能源已成为各国科学家共同追求的目标。质子膜燃料电池(PEMFC)以其高功率密度,启动迅速,无污染等优点成为21世纪首选清洁能源系统。但其原理涉及热力学、电化学、流体力学、传质学等理论,形成一个非线性复杂系统,难以建立数学模型。因此,采用一种动态自适应网络即最近邻聚类径向基函数神经网络,它能够动态调节网络的规模和参数,具有较强的逼近能力以及自学习能力。并利用测试数据作为训练样本,在氢气流速给定的条件下,以空气(或氧气)压力和冷却水流速作为模型的输入量,电池的电压为输出量,建立了在工作温度为60℃和80℃时的PEMFC电特性模型。表明该方法具有简单、可行、精度高等优点。并为PEMFC控制系统的设计和电池性能的优化提供了基本依据。

Due to today＇s conventional energy shortage, developing and utilizing new - type clean, green energy become the goal of scientists the world over. Proton membrane fuel cell（PEMFC） with its advantages such as high power density, rapid start and pollution - free becomes the first - selectcd clean energy system in 21st century. But its operation principle involves such theories as thermodynamics, electrochemistry, hydrodynamics, mass transfer etc. , thus forming a nonlinear complicated system, so it is difficult to establish mathematical model. A dynamic self- adaptive network, the nearest neighbor - clustering RBF network can regulatethe scale and parameter of network dynamically, has stronger approaching and self - learning ability. Utilizing the test data as training samples,on terms the velocity of flow of hydrogen is given, the pressure of air （or oxygen ） and velocity of flow of cooling water as the input of model, the voltage of cell as the output of model, the electric characteristic model of PEMFC is set up when operating temperature is 60℃ and 80℃. It indicates that the method has the advantages of simplicity , feasibility , high precision and provides a basis for the design of PEMFC control system and optimization of cell performance .