Estudio cinético sobre la desulfuración y la desnitrogenación de destilados de petróleo medios

Abstract

El hidrotratamiento de dos fracciones medias de petróleo (Diesel y Coker Gasoil) se analizó mediante simulación utilizando un modelo de reactor flujo pistón y diferentes enfoques cinéticos. El estudio se centró en las reacciones de hidrodesulfuración (HDS) e hidrodesnitrogenación (HDN). Las materias primas se procesaron en un rango de temperaturas de 335 a 365ºC, la presión de operación varió entre 3.5 y 5.1 MPa y el LHSV se ajusto entre 0.7 y 3.5 h-1. Se evaluó la precisión de los distintos enfoques, así como el significado cualitativo de los parámetros cinéticos. Se encontró que uno de los modelos estudiados predijo el 66% de los experimentos en el rango de incertidumbre con un coeficiente de correlación de Pearson de r = 0.941, mientras que el segundo modelo solo alcanzó el 48% de los experimentos en el rango deseado, con un coeficiente de correlación de Pearson de r = 0.789.

Hydrotreating processes play a vital role in petroleum refineries to meet the increasing demand of fuels and oil fractions and to guarantee the long-term viability of the refining business. Different kinetic approaches have been proposed in the past to model this process, each one with specific advantages in simplicity or accuracy. The hydrotreating of medium oil cuts was analysed via simulation using a heterogeneous plug-flow reactor model and different kinetic approaches. The study was focused on the hydrodesulfurization (HDS) and hydrodenitrogenation (HDN) reactions. The dataset comprised two medium oil fractions: diesel fraction (DF) and coker gasoil (CGO). The feeds were processed in a range of temperatures from 335 to 365ºC, the operation pressure varied from 3.5 MPa to 5.1 MPa and the LHSV was set from 0.7 to 3.5 h-1. The accuracy of the different approaches is assessed as well as the qualitative meaning of the kinetic parameters. It was found that with one of the studied models 66% of the experiments were predicted in the given range of experimental uncertainty, with a Pearson correlation coefficient of the linear regression of the parity plot of r = 0.941, while in the second model only the 47% of the experiments were predicted in the desired range, with a Pearson correlation coefficient of the parity plot of r = 0.789. Keywords: