高维小波网络及其在丙烯腈收率预测中的应用  

Large-dimensional wavelet networks and its application in (predicting) yield of acrylonitrile

作　　者：赵恒平[1] 俞金寿[1] 

机构地区：[1]华东理工大学自动化研究所,上海200237

出　　处：《控制与决策》2005年第3期353-355,360,共4页Control and Decision

摘　　要：针对基于单尺度小波框架的高维小波网络,提出一种系统化的设计方法.首先在小波框架内提出一种小波基初始化方法;然后根据样本的分布特点,提出一种改进的小波基粗选方法;最后将自适应投影算法与AIC准则相结合,对小波基进行精选,同时完成网络参数的辨识.将该方法应用于丙烯腈收率的预测,研究结果表明了该方法的有效性.To large-dimensional wavelet networks based on single scaling wavelet frames theory, a systematic design method is proposed. First, method for initialization of wavelet basis in wavelet frames is proposed) second, according to the data distribution, a modified method for rough selection of wavelet basis is given; and the last, an adaptive projection algorithm combined with AIC criterion is used to purify the wavelet basis, meanwhile finishing the parameters identification. The proposed method is applied in predicting the yield of acrylonitrile, and the results show the effectiveness of this method.

关 键 词：小波框架 小波网络 自适应投影算法 AIC准则 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]