欢迎访问过程工程学报, 今天是

›› 2007, Vol. 7 ›› Issue (2): 366-369.

• 系统与集成 • 上一篇    下一篇

采用神经网络和主成分分析方法建立催化重整装置收率预测模型

郭彦,李初福,何小荣,龚真直,陈勃,张秋怡   

  1. 清华大学化学工程系
  • 出版日期:2007-04-20 发布日期:2007-04-20

Prediction on Product Yields of Catalytic Reforming Unit by BPNN-PCA

GUO Yan,LI Chu-fu,HE Xiao-rong,GONG Zhen-zhi,CHEN Bo,ZHANG Qiu-yi   

  1. Department of Chemical Engineering, Tsinghua University
  • Online:2007-04-20 Published:2007-04-20

摘要: 催化重整装置收率预测的准确性对生产计划的制定具有重要意义. 采用神经网络和主成分分析方法,并运用改进的遗传算法进行网络训练,建立了催化重整装置收率预测模型. 将这一模型程序作为催化重整装置收率预测模块加入软件GIOPIMS(Graphic I/O Petro-chemical Industry Modeling System)中,得到更为精确的产品收率预测结果. 在实际生产中,将收率预测模块计算的催化重整装置各侧线的收率代入生产计划模型中,可为生产计划的制定提供依据.

关键词: 催化重整, 收率, 预测, BP神经网络, 主成分分析

Abstract: An accurate prediction model of product yields of the catalytic reforming unit is important for keeping the production plan in optimal operation while the external factors are changing in a refinery. The prediction model based on back propagation-principal component analysis (BPNN-PCA) method was appended to the GIOPIMS (Graphic I/O Petro-chemical Industry Modeling System). In the actual production planning, the prediction results were accurate. The prediction model on product yields of the catalytic reforming unit can effectively assist the production planning.

Key words: catalytic reforming, yield, prediction, back propagation network, principal component analysis