17th European Symposium on Computer Aided Process Engineering ESCAPE17

V. Plesu and P.S. Agachi (Editors)

2007 Elsevier B.V. All rights reserved.

Modeling of trickle bed reactor for hydrotreating of

vacuum gas oils: effect of kinetic type on reactor
modeling
Favio Jimneza, Karina Ojedaa, Eduardo Sncheza, Viatcheslav Kafarova,
Rubens Maciel Filhob
a

Industrial University of Santander, Tel. +57-76344746, Bucaramanga, Colombia, Email: [email protected]

b
Universidade Estadual de Campinas, UNICAMP

Abstract
The hydrodesulfurization (HDS) processes research requires a lot of
experimental work to define promotion and inhibition effects and find out
correlations between the variables that participate on the sulphur removal. By
means of modeling and simulation is possible to determine the scope of the
process, and by means of the sequential design of experiments (SDE) is possible
to reduce the experimental work required through comparison of kinetic
models, at the same time that it predicts experimental conditions that allow to
select a unique model and estimate its parameters. The purpose of this work was
to simulate the HDS process using the mathematical model developed in
previous work and several kinetic models founded in literature. Together
simulation is proposed a SDE-based software with Runge-Kutta routines,
orthogonal collocation methods an Sequential Quadratic Programming to
develop the steps of the Design and to be used on pilot plant applications. The
good agree between theoretical and experimental data led to the development of
a user-friendly program to simulate the complex process, make easier the
interpretation of simultaneous reactions, and become a useful tool for to
improve the operation conditions of hydrotreating industrial reactors.
Keywords
Sequential, experiments, hydrodesulfurization, kinetics, optimization

F. Jimnez et al.

1. Introduction
To accomplish with the actual legislation and at the same time improve the
quality of liquid fuels, it is useful to simulate the petroleum fraction
hydrotreating process at different operation conditions. However, it is necessary
to carry out series of experimental tests, to study and characterize this process,
in order to determine the best operation conditions that allow establishing
mathematic correlations between the studied variables. The kinetic is maybe the
major element of reactor modeling and simulation, and therefore particular
attention must be paid. This work is further development of our research in the
area
of
computer
aided
modeling
for
hydrodesulfurization,
hydrodenitrogenation and hydrodearomatization simultaneous reactions in a
hydrotreating industrial process. In previous work [1] the process model was
created by the combination of two multiphase models reported in literature for
gas and liquid phase and for solid phase respectively. The main purpose of this
work is to compare the behavior of HDT reactions taking into account several
kinetic models founded in literature, and apply techniques of SDE in order to
select the kinetic model that best represents the HDS process according with the
operative conditions suggested by the software itself. A complete review of
kinetic models for different petroleum fractions was carried out, likewise
comparative analysis for type of processing, type and nature of reacting
systems, modes of operation, process conditions, expressions for mass transfer,
physico-chemical properties, and numerical methods to solve the respective
equations systems was made, and used as a base for carry out new
experimentation at pilot plant for hydrotreating of heavy petroleum fractions.
2. Reactor Modeling
For liquid and gas phases the three-phase model reported by Korsten and
Hoffman (1996) was combined with the solid phase model reported by Froment
et al. (1994). The main assumptions and model simplifications were: reactor
model is one-dimensional heterogeneous and operates isothermally, the
reactions occur in the liquid phase in contact with the catalyst surface, external
mass transfer is negligible, catalyst wetting is complete, there is no catalyst
deactivation; and there is no evaporation of liquid
The expressions for estimation of the physico chemical properties (in order to
solve the differential equations) were obtained from authors [2-3]. Two
different data set of basic information for the simulation, corresponding with
reactor geometry (length and diameter), catalyst properties (density, superficial
area, equivalent size), inlet operative conditions (temperature, pressure, liquid
and gas flows), and inlet concentrations of key components, among others, were
selected in order to compare all the kinetic models into the reactor model. The
first data set was obtained from industrial information [4] the second data set
was obtained from experimental tests made at pilot plant [1]. Both data sets are

Modeling of trickle bed reactor for hydrotreating of vacuum gas oils: effect of kinetic
type on reactor modeling
3

presented in Table 1. Finally, a user-friendly computational program was

developed based on programming of Matlab 7.0, which have many tools that
allow to apply optimization process and differential equations solutions
approximation with a wide range of defined matricial functions. The axial
direction in reactor was performed using a fourth order Runge-Kutta routine
with variable step size, and the intra-particle integration was carried out with an
orthogonal collocation method (second order differential equations).
3. Kinetic Model
To represent the HDS process have been incorporated nine kinetic models
(considering the presence of sulphur in form of dibenzotiophene-DBT)
established by several authors [4-11]. Some of these models consider two
kinetics expressions, one for hydrogenolysis and other for hydrogenation of
dibenzothiophene. The selected kinetic models to use in simulation of HDS
process are presented in Table 2 (donde ri = Reaction velocity of i component,
Ci= Concentration, Pi = Pressure, ki = Reaction velocity coefficients, Ki =
Absorption equilibrium constant.
4. Results
Concentrations profiles for DBT, H2S and H2 in liquid and gas phases
corresponding with the considered kinetic models were obtained. In Figures 1-4
are shown behaviors for concentration of DBT in liquid phase along reactor (for
four kinetic models, two Langmuir-Hinshelwood type and two power law type,
like representatives of the other kinetic models), whereas in Figures 5-8 are
shown for H2S. These behaviors are based on data set of industrial information.
Table 1. Basic data sets for reactor simulations
Property
Value
Operation Conditions
590*
Temperature
(K)
603.15**

Property
Value
Feed Characteristics
968.12*
Liquid Den.
(kg/m3)
863**

Propetry
Value
Catalyst Characteristics
1420*
Real Dens.
(kg/m3)
337.8**

Reactor Characteristics

Bula Dens.
(kg/m3)

5e6*
Pressure (Pa)
Gs Flow
(m3/s)
Liquid Flow
(m3/s)

9e6**
3.63*

2.82*
19e3**
7.62*

Particle
Diameter
(m)
2.09e-2*
Especific
Lenght (m)
Area
1.1e-7**
0.73**
(m2/m3r)
*Data from Froment et al. (1994)
**Data from Pilot Plant
2.2e-5**

Diameter
(m)

Table 2. Kinetic models for simulation of HDS process.

710*
327.3**
1.3e-3*
1.6e-3**
8.9e7*
3.6e8**

Property
Value
Inlet Conditions
3.1e6*
Pressure
H2
9e6**
(Pa)
0*
Pressure
H2S
0**
(Pa)
525.4*
DBT Conc.
(mol/m3)
66**
H2 Conc.
(mol/m3)
H2S Conc.
(mol/m3)

256*
0**
0*
0**

F. Jimnez et al.
Model

Hydrogenolysis

1. Girgis and
Gates (1991)

2. Froment et
al (1994)
3. Broderick et

r=

k KDBT KH2 CDBT CH2

(1+ K

) (
2

DBT CDBT + KH2S CH2S 1+ KH2 CH2

rDBT CHB =

r=

al (1981)

Hydrogenation

(1+ K

k 2 K DBT , C DBT C H 2

(1 + K

C DBT )

k CDBT CH2
C

) (
2

DBT DBT

+ KH2S CH2S 1+ KH2 CH2

(1 + K

S
K11E C HS 2 C DBT
HDS
H 2S

S
C HS 2S + K AHDS C DBT

6. Van Hasselt et al (1999)

rA =

Kr C

2
DBT

CH2

1 + K C H 2S

8. Cotta and Maciel-Filho (2000)

1 .2
rS = k S C DBT
PH1.25

k1 K DBT , K H 2 , C DBT C H 2

C
1 + K DBT , C DBT + K H , C H + K H S , H 2 S
2
2
2

C H 2

rDBT =

4. Avraam-Vasalos (2003)

rHDS =

(1 + K ' DBT C DBT )

rDBT BPH =

DBT ,

k ' K ' DBT K ' H 2 C DBT C H 2

r =

k 2 C DBT C H 2
(1 + K DBT C DBT )(1 + K H 2 C H 2 )

5. Chen et al (2001)
1 . 12
r A = K .C DBT
.C H0 .285

7. Tsamatsoulis and Papayannakos (1998)a

rHDS =

(1 + K

K HDS PH 2 C S2.3
H 2S

PH 2 S + K H 2 PH 2

9. Tsamatsoulis and Papayannakos (1998)b

rHDS =

K HDS PH2 .23 C S

(1 + K

H 2S

PH 2 S

There are strong differences on the concentration predicted values along the
reactor when are used Langmuir Hinshelwood (L-H) kinetic models with
respect to power law (P-L) kinetic models. The formers predict DBT
concentrations with a gradual diminution along the reactor whereas the others
are in agree with a pronounced reduction of DBT concentration on the first
section of the reactor. Is remarkable point out that the employed parameters of
this models depends on the specific experimental work conditions (different for
most of the models), therefore these differences could result in tendencies and
predictions out of established original range. On the other hand, the amount of
sulphydric acid (H2S) on the liquid arises very fast on the first sector of the
reactor. Results using L-H kinetic show almost constant production of H2S
along reactor, whereas using P-L kinetic the DBT is consumed very fast and the
H2S diminish to almost zero after the first sector of the reactor. The second data
set of basic information (pilot plant) present higher temperature, almost twice
the pressure, and some catalytic properties are better for HDT process, related
with the first data set. Therefore the concentration profiles (not shown here)
result very similar to the presented in Fig.1-8, however the consume of DBT
and production of H2S are very much faster and the slopes more pronounced.

Modeling of trickle bed reactor for hydrotreating of vacuum gas oils: effect of kinetic
type on reactor modeling
5

Figure 1. Broderick et al.(1981)

Figure 2. Froment et al.(1994)

Figure 3. Van Hasselt et al.(1999)

Figure 4. Chen et al.(2001)

Figures 1-4. Concentrations profiles for DBT (vertical axis) along reactor (horizontal axis)

Figure 5. Broderick et al.(1981)

Figure 6. Froment et al.(1994)

Figure 7. Van Hasselt et al.(1999)

Figure 8. Chen et al.(2001)

Figures 5-8. Concentrations profiles for H2S (vertical axis) along reactor (horizontal axis)

F. Jimnez et al.

On the other hand, it was proposed the use of SDE in order to compare all the
kinetic models and to select that which best represent de HDS process over the
operative conditions suggested by the software itself. Techniques like
Sequential Quadratic Programming (SQP) for each one of the restrained
optimization process necessary for the application of criteria of the SDE,
Orthogonal Collocation Method for the solution of the second order differential
equations, and Runge-Kutta method for the systems of differential equations,
were used in the software. For the validation of discrimination of kinetic models
were used results reported in literature [12], obtaining the same final kinetic
model and estimating parameters in the same order de magnitude with respect to
those presented by the authors. Trough a SQP process, it was founded the
experimental point where the nine selected kinetic models for HDS process had
the maximum difference in their responses. Then, each time some models had to
be eliminated until completes the process.
5. Conclusions and future work
Were compared several kinetic models into the modeling and simulation of a
trickle bed reactor. It was developed specific software based on Matlab 7.0 for
simulation and sequential design of experiments (SDE) and were validated with
several papers reported in literature, obtaining results that are in concordance
with data and behaviors reported. Also, it was shown that power law kinetic
models generate concentration profiles for consume of DBT and production of
H2S very different with respect to Langmuir-Hinshelwood kinetic models,
however the software reproduced very well the behaviors for reactive and
products with two different data set: industrial and pilot plant information.
References
1.
2.
3.
4.

F. Jimenez, M. Nunez, V. Kafarov, 2006, Comp. Aided Chem. Eng., 21A, 651-656
M. Rodriguez, J. Ancheyta, 2004, Energy & Fuels, Vol.18, 789 794
H. Korsten, U. Hoffman, 1996, AIChE Journal, Vol. 42, No. 5.
G. Froment, G. Depauw, V. Vanrysselberghe, 1994, Ind. Eng. Chem. Res., Vol. 33,
2975.
5. D. Avraam, I. Vasalos, I., 2003, Cat. Today, Vol. 79-80, 275-283.
6. J. Chen, Z. Ring and M. Dabros, 2001, Ind. Eng. Chem. Res., Vol. 40, 3294-3300.
7. M. Girgis, B. Gates, 1991, Ind. Eng. Chem. Res., Vol. 30, 2021-2058.
8. B. Van Hasselt, P. Lebens, H. Calis, C. Van D. Bleek, 1999, Chem. Eng. Sci., Vol 54,
4791.
9. D. Tsamatsoulis, N. Papayannakos, 1998, Chem. Eng. Sci., Vol. 53, N19, 3449-3458.
10. R. Cotta, M. Wolf-Maciel, R. Maciel Filho, 2000, Comp. & Chem. Eng., Vol. 24, 1731.
11. D. Broderick, B. Gates, 1981, AIChE Journal,Vol. 27.
12. I. Van Parijs, L. Hosten and G. Froment, 1986, Ind. Eng. Chem. Prod. Res, Vol 25, 437.