FCCU分馏塔产品质量预测系统的设计  

Design of quality prediction system for distillation tower in FCCU

作　　者：杨淑莹[1] 郭翠梨[2] 

机构地区：[1]天津理工大学计算机科学与工程系,300191 [2]天津大学化工学院,天津300072

出　　处：《哈尔滨工业大学学报》2005年第4期501-503,共3页Journal of Harbin Institute of Technology

基　　金：天津市科技发展计划资助项目(993109511)

摘　　要：为了解决单纯采用正系统模型无法确保预估计值的准确性问题,以神经网络的内模控制预测系统为基础,对对象-正模型-逆系统进行逆变换,建立了输入可测而输出无法测量的预测系统,实现了FCCU分馏塔轻柴油凝固点、粗汽油干点实时在线测量.以实时采集的数据作为检测标准,结果表明,该模型可对生产提供在线的可靠指导,对稳定油品质量、避免质量事故的发生和提高经济效益具有重要的价值.In order to solve the problem that the positive BP model can't guarantee the accuracy of the predicated value, the internal model control prediction system of neural network was adopted. A prediction system which input but not output can be measured is set up by inverse transformation of the object-positive model-reversal system. It can estimate the real-time qualities for distillation tower in FCCU. The result has been proved by experiments. The model can give online credible guiding and decrease quality accident and increase economic profit.

关 键 词：催化裂化 分馏塔 预测 神经网络 效用函数 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]