基于神经网络的软测量技术在精馏塔上的应用

针对扬子石化公司丁二烯精馏塔原控制系统存在的问题, 利用从集散控制系统(DCS)采集的大量现场数据和用机理模型得到的模拟数据, 运用前向反馈(BP)神经网络软测量技术，构造了产品丁二烯和总炔含量的自适应软测量仪表, 设计了一套控制系统. 实际监测数据表明, 这套控制系统可实现产品质量的闭环控制.

Application of the Soft-sensing Technique Based on Neural Network to a Distillation Column

In view of the existing problem of the former control system of the Yangzhi Petrochemicals butadiene distillation column, voluminous plant operation data collected by DCS and simulated results from a theoretical model are pooled together and used to build the adaptive soft-sensor instrument for butadiene and alkynes contents in the distillation column top based on the BP neural network technique. Then, an inferential control system was designed according to the targets of product quality, in which the on-line estimating values of soft sensor instrument were used. As a result of increasing logic calculation in the inferential control arithmetic, the robustness of the control system is strengthened. Application of the control system to the column showed that the control system can run smoothly over a long period in worksites, and has realized the close-loop control of product quality.