论文部分内容阅读
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。
Aimed at the combustion process of a circulating fluidized bed boiler with multidimensional nonlinearity and pure hysteresis characteristics, a radial basis function (RBF) neural network based on PLS learning algorithm and OLS learning algorithm was used to study the modeling. Firstly, the network training data and generalized data for modeling experiments are generated by the CFB boiler simulation platform. Then the OLS and PLS algorithms are used respectively to conduct network training and generalization studies. Finally, the algorithm parameters affecting the modeling results are discussed And its selection method, and focuses on the influence and selection of 4 network parameters of PLS algorithm. Compared with the modeling experiment based on wavelet network, RBF network modeling is used to simulate the combustion process of CFB boiler with complicated characteristics, while the accuracy of modeling is guaranteed, the selection of algorithm parameters is also more convenient and easy.