基于神经网络的常压塔柴油凝点软测量的研究与应用

Research and Application of Soft Sensor of the Diesel Oil Solidifying Point Based on the Neural Network on a Crude Distillation Unit

研究某炼油厂常压塔三线柴油凝点的软测量建模问题,分析过程变量对柴油凝点的影响。基于在线分析仪6min采样数据,利用前向网络和时延前向网络(TDNN)分别建立了三线柴油凝点的静态软测量模型和动态软测量模型,并结合在线分析仪对模型实现了在线修正。通过两种模型的仿真和在线实施效果,表明基于神经网络的软测量模型取得了较好的应用效果,而且动态模型的实施效果优于静态模型。

A soft sensor modeling of the diesel oil solidifying point (DOSP) on a crude distillation unit is studied. The impact on the solidifying point is analyzed by the process variables. A static model and a dynamic model of DOSP soft sensor are built based on forward neural network and time delay neural network (TDNN) respectively. The model outputs are modified online according to the values given by the online analysis device. The results of simulation and implementation state that the proposed models take effect in the soft sensor of the crude distillation unit, and the dynamic model is better than the static model.