(139c) Fixed Bed Residue Hydroprocessing: Modeling Hydrotreating Performances and Catalyst Deactivation

Crude oils contain large fractions of heavy products for which only few outlets exist. Refining processes converting heavy oils into more valuable products are therefore vital to a refinery. However, the heaviest cuts, termed residues, also contain the largest amounts of impurities, such as sulfur, nitrogen and metals. Fixed bed residue hydroprocessing is industrially carried out at high temperature and under a high hydrogen partial pressure. Under these conditions, the catalysts employed promote a complex network of parallel and consecutive reactions, among which are hydrodemetallization (HDM), hydrodesulfurization (HDS), hydrodenitrogenation (HDN), hydrocracking and hydrogenolysis. To enable all these chemical transformations, the residue feed successively passes over several fixed adiabatic beds containing different catalysts.

The reactivity of a residue feed depends both on the concentration and the nature of the various species. As it is not possible to fully separate and identify the various compounds in these mixtures, they are generally subdivided into several classes: saturates, aromatics, resins and asphaltenes (SARA). The latter are colloid structures whose radius of gyration typically lies between 10 Å and 500 Å, depending on the conditions. Asphaltenes can therefore only penetrate into the outer layers of porous catalysts. Their penetration is not only governed by molecular and Knudsen diffusion, but also by so-called configurational diffusion or surface diffusion. Hence, the performances of residue hydrotreating catalysts also strongly depend on the porous texture of the catalysts.

In this work, the performance of fixed bed residue hydrotreating units is predicted based on a model that accounts for the various physical and chemical phenomena. The porous structure of the catalyst is represented by Random Spheres, Random Needles and Random Coins models depending on the type of catalyst (Toulhoat et al., 2005). The hydrodynamic description accounts for transport of the species by convection and by Fickian diffusion inside the catalyst pellets. The molecular diffusion coefficients are calculated from the Stokes-Einstein equation and combined with Knudsen's law. Finally, as the molecular species can have a radius of gyration of the same order of magnitude as the pore size, configurational diffusion is also accounted for (Spry and Sawyer, 1975).

To describe the reactions, the hydrocarbons were first separated according to their boiling point into a low boiling (<520°C) and a high boiling (>520°C) fraction. Each fraction was subsequently separated by a SARA fractionation, which separates oil fractions into Saturates, Aromatics, Resins and Asphaltenes. To account for the purification reactions, the atomic composition (C, H, S, N, O, Ni and V) of each of these 7 lumps was also analysed and tracked in the model, leading to a total of 84 reactions.

The kinetic parameters were determined from experimental data obtained on a representative pilot unit operating on actual vacuum residues. From the simulations (Figure 1), it can be concluded that the model is able to correctly predict the unit operation. Indeed, a good agreement with the experimental data has been obtained, with regard not only to the concentration profiles of the various lumped families but also to the elemental composition of each family. As shown in Figure 1a, significant intraparticle profiles exist, even at the reactor outlet, illustrating the need to account for intraparticle diffusion.

In residue hydrotreating, catalyst aging is extremely rapid and run lengths are typically around 10 to 12 months. The build-up of deposited coke and metal sulfides over time strongly affects the catalyst accessibility and pore structure, and hence its activity. The use of Random Spheres, Random Needles and Random Coins models for the catalyst structure allows to advantageously represent catalyst aging. Indeed, the nucleation and growth of solid deposits inside the porosity can easily be accounted for by superimposing three random media (fresh catalyst, metal sulfides, coke), and calculate the residual porosity, surface area, volumes of coke deposit and of metal sulfides deposit, etc.

With this model, the simulations allow to predict not only the concentration profiles of the various chemical lumps, but also the evolution of elemental composition of each lump throughout the reactor. It is also shown that significant intraparticle profiles exist for asphaltenes and heavy resins, even at the reactor outlet, illustrating the need to account for intraparticle diffusion. The predicted intraparticle profiles also correspond well to the measured intraparticle profiles of the metal deposits (V and Ni).

Finally, the full-scale problem was also tackled with this model by simulating an industrial multi-bed adiabatic reactor with inter-reactor quenches. The various simulation outputs were then examined to assess the effects of the various operating variables.

References 1. Toulhoat, H., Hudebine D., Raybaud, P., Guillaume D. and Kressmann S.; A new model for combined simulation of operations and optimization of catalysts in residues hydroprocessing units, Catalysis Today, 109, p 135-153, 2005 2. Spry J. and Sawyer W., Paper presented at 68th Annual AIChE Meeting, Los Angeles, November 1975

1-2009-->