循环流化床锅炉燃烧过程建模研究  

Modeling of the combustion process of circulating fluldized bed boiler

作　　者：樊诚[1] 江青茵[1] 曹志凯[1] 周荣[1] 

机构地区：[1]厦门大学化学化工学院化学工程与生物工程系,福建厦门361005

出　　处：《计算机与应用化学》2007年第3期371-374,共4页Computers and Applied Chemistry

基　　金：福建省科技计划重点项目(2005H044)

摘　　要：针对具有多维非线性和纯滞后特性的循环流化床锅炉燃烧过程,采用基于PLS学习算法和OLS学习算法的径向基函数(RBF)神经网络进行建模研究。首先通过循环流化床锅炉仿真平台产生用于建模实验的网络训练数据和泛化数据,然后分别采用OLS算法和PLS算法进行网络训练和泛化研究,最后讨论了影响建模结果的算法参数及其选取方法,重点讨论了PLS算法的4个网络参数的影响和选取。与基于小波网络的建模实验比较,对具有复杂特性的循环流化床锅炉燃烧过程,采用RBF网络建模在保证建模精度的同时,算法参数的选取也较为方便易行。To get a online dynamic model of the combustion process of Circulating Fluidized Bed （CFB） Boiler which was difficult to be modeling because of its complex characteristics such as non-linear, time-delay, time-variation and multidimensional, the RBF network based on the PLS arithmetic and the OLS arithmetic was adopted. The data used as the network training samples and the data used in the generalization test of the network was produced by a CFB Boiler simulation system. The network training and generalization study was done by means of PLS arithmetic and OLS arithmetic individually. The arithmetic parameters, especially the parameters of PLS arithmetic were discussed to study their influence and selected methods. Contrasting to the online dynamic model of this process which was credited by the wavelet network, the model put forward in this paper keep the same precision and is more easily to select the parameters in the modeling.

关 键 词：RBF神经网络 过程辨识 循环流化床锅炉 

分 类 号：TQ015[化学工程]