文献类型：期刊文献

中文题名：基于动态神经网络的非线性过程在线预测

英文题名：On-line Prediction of Nonlinear Process Based on Dynamic Neural Network

作者：张兵[1];钱锋[1];颜学峰[1];罗娜[1]

机构：[1]华东理工大学信息学院自动化研究所,上海200237

年份：2006

卷号：33

期号：6

起止页码：15

中文期刊名：化工自动化及仪表

外文期刊名：Control and Instruments in Chemical Industry

收录：EI(收录号：20070410389430);Scopus

基金：国家"973"计划(2002CB3122000);上海市自然科学基金(05ZR14038);上海科委科技攻关项目(04DZ11010;05DZ11C02);上海市科委重大基础研究(05DJ14002)

语种：中文

中文关键词：非线性;在线预测;泛回归神经网络;动态神经网络

外文关键词：nonlinearity; on-line prediction; general regression neural network; dynamic neural network

摘要：神经网络需满足以下两个条件方能用于非线性过程的在线预测:①神经网络必需以某种递推的方式出现;②神经网络的学习算法应尽可能简洁快速。为此改造泛回归神经网络(GRNN),运用递推更新的样本数据集训练GRNN,构成动态泛回归神经网络。该动态神经网络训练方便快捷,能够满足在线预测的实时性的要求。仿真实验表明预测值较观测值有一定滞后,但均能尾随观测值而变化,达到了预期的目标。
Neural network can be applied to on-line prediction with the following two requirements ：①neural network appears in recursive form;② learning algorithm of neural network is simple and fast. Therefore general regression neural network （GRNN） is modified to meet the requirements as above. It is trained by using the recursively updated sample date sets and thus GRNN became dynamic neural network. Training the dynamic neural network is convenient and fast,and so it can be applied to on-line prediction. The results of simulation experiments show that there is a little lag between the prediction profile and real profile,but the former vary with the latter. The expected aim is reached.