Modelling and simulation of diesel catalytic dewaxing reactors

Neto, Eugenio Turco (2017) Modelling and simulation of diesel catalytic dewaxing reactors. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (3MB)

Abstract

Designing catalytic dewaxing reactors is a major challenge in petroleum refineries due to the lack of kinetic studies related to this operation. Also, the measurements of the cloud point of diesel fuels produced from those units are still being carried out using inaccurate visual procedures, which bring difficulties to the process control design. In this thesis, a single event kinetic model is developed for catalytic dewaxing on Pt/ZSM-5, which application has not been explored in the scientific literature. A total of 14 kinetic parameters have been estimated from experimental data, which are independent on the feedstock type. Then, the obtained parameters were used to propose a soft-sensor consisting of three different modules that handles the feedstock distillation data and integrates a mechanistic reactor model to a solid-liquid flash algorithm to predict the cloud point of the diesel product. Finally, a surrogate model is developed using a sequential design of experiments to simplify the sensor framework and reduce the computing time so that it can be industrially implemented to perform on-line cloud point estimations. The results have showed that the proposed kinetic model was in agreement with observed data and suitable to simulate the industrial operation. Pressure, temperature, and liquid hourly space velocity (LHSV) were found to be the main process variables controlling the conversion in the hydrodewaxing mechanism. The proposed sensor showed to be suitable to study the reactor performance for different set of operating conditions. Also, the surrogate model drastically reduced the computing time to obtain the cloud point estimations and showed to be suitable for on-line prediction purposes. Finally, the employed sequential design strategy revealed that nonlinearities can strongly affect the sensor accuracy if not properly handled.

Actions (login required)

View Item