芳烃抽提过程收率软测量技术应用  

Application of soft sensing technology for extraction of aromatics

作　　者：刘思成[1] 李悦[1] 杨秀[1] 王娟[1] 张蒙蒙[1] 

机构地区：[1]辽宁石油化工大学信息与控制工程学院,抚顺113001

出　　处：《自动化与仪器仪表》2015年第5期53-56,共4页Automation & Instrumentation

摘　　要：为了预测芳烃抽提过程收率对生产计划的制定准确性的目的。采用遗传算法和粒子群算法优化BP神经网络方法,并运用改进的遗传算法和粒子群算法进行网络训练,分别建立了芳烃抽提过程的芳烃收率建立软测量模型,并结合实际生产工艺,选取16个过程变量。并对收率分类建立相应软测量模型进行仿真研究。获得该两种模型的性能优于单纯的BP神经网络模型,且都能够较准确预测芳烃收率和较高精度的结果。为实际生产计划的制定提供参考的依据并具有良好的应用前景。In order to predict yield of aromatics extraction process for the purpose of the accuracy of the production plan.Basing on genetic algorithms（GA） and particle swarm optimization（PSO） by optimizing BP artificial neural network（BP）,two soft-sensing models for aromatics yield in aromatics extraction distillation process were presented respectively.Then two models by combining the actual production process which choose 16 process variable and classify the yield were applied in simulation research.Simulation results show that the model based on GA-BP and PSO-BP have higher precision.To provide reference basis for the establishment of the actual production plan and has good application prospect.

关 键 词：芳烃抽提 收率 预测 BP神经网络 软测量 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]