摘要
针对前馈网络的不足,提出利用改进的Elman神经网络模型对兰州石化某厂600kt乙烯精馏塔出口成分含量进行建模和预测,并采用主元分析法对高维输入变量进行预处理,以简化网络建模的结构。实例分析表明:改进的Elman神经网络收敛速度快、预测精度高,能够更好的实现对乙烯精馏塔出口成分含量的建模和预测。
Aiming at the shortage of feed-forward neural network, a modified Elman neural network is proposed, which is applied to modeling and prediction the production of Ethylene Rectifying Column of an industrial. And the principal component analysis is applied to preprocessing high dimensional input variable so that simplified Neural Networks structure. Through application in real case, the modified Elman neural network has high precision, fast convergence rate and it performances better in modeling and forecasting the production of Ethylene Rectifying Column.
出处
《微计算机信息》
北大核心
2008年第22期248-249,284,共3页
Control & Automation
关键词
精馏塔
改进ELMAN神经网络
主元分析法
建模
Rectifying Column
Modified Eiman neural network
principal component analysis
modeling
