基于拉盖尔模拟神经网络的过热汽温直接自适应控制系统.pdfVIP

26 卷 23 期 中 国 电 机 工 程 学 报 Vol.26 No.23 Dec. 2006 2006 年12 月 Proceedings of the CSEE 2006 Chin.Soc.for Elec.Eng. 文章编号：0258-8013 (2006) 23-0067-05 中图分类号：TP273 ；TK323 文献标识码：A 学科分类号：47040 基于拉盖尔模拟神经网络的过热汽温 直接自适应控制系统 军 (绍兴文理学院机电系，浙江省 绍兴市 312000) Direct Adaptive Control System of Super-heated Steam Temperature Based on Laguerre Analog Neural Network YE Jun (Department of Mechatronic Engineering, Shaoxing College of Arts and Sciences, Shaoxing 312000, Zheji ang Province, China) ABSTRACT: A kind of Laguerre analog compound 作为直接自适应控制器，这种单隐层正交神经网络是基于 orthogonal neural network was presented and was applied to Laguerre 复合正交多项式函数，并具有在线连续学习的简单 the direct adaptive control of super-heated steam temperature in 算法，且学习算法与被控对象模型无关。由于采用3 层网络 a power plant. Herein, the analog neural network was 结构，输入层与隐层之间不用权值调整，在学习算法中只要 developed to employ the direct adaptive controller. This 在输出层与隐层之 寻找最佳权值，因此网络学习速度较快。 single-hidden layer orthogonal neural network is based on 网络隐层节点(处理元)是Laguerre 多项式展开项，展开项的 Laguerre compound orthogonal polynomial function and has a 多少决定着网络的学习速度和精度。通过对具有严重参数不 simple learning algorithm that is derived from an on-line 确定性、扰动以及大迟延的电厂过热汽温被控对象进行仿真 continuously learning, and the learning algorithm is not relative 研究，结果表明控制系统性能优于常规的PI 控制系统。 to the controlled plant model. Owing to using the three-layer 关键词：模拟复合正交神经网络；拉盖尔复合正交多项式； network structure and no weight adjustment