Modelling and Optimal Operation of a Crude Oil Hydrotreating Process with Atmospheric Distillation Unit Utilising Stacked Neural Networks

Abstract

Abstract Data-driven modelling and optimisation of a crude distillation process with crude oil hydrotreating (HDT) using stacked neural network are presented in this paper. HDT of crude oil is a new process that has not been considered widely in the literature. Process optimisation should be conducted to improve the operation efficiency of HDT process. In this paper, stacked neural networks are used to model crude oil HDT process and to improve model prediction accuracy and reliability. A crude oil HDT process with crude distillation unit (CDU) simulated using Aspen HYSYS software is used as a case study. The application results show that accurate models for crude oil HDT process with CDU can be developed from process operational data using stacked neural networks. One of the most important findings from this paper is that stacked networks can generate more accurate and reliable predictions than single networks, which in turn leads to reliable optimisation results. Goal-attainment method for multi-objective optimisation is used. The obtained optimisation results are validated on Aspen HYSYS simulation and the effectiveness of the proposed scheme is demonstrated.