Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Applied Catalysis A: General 351 (2008) 148–158

Contents lists available at ScienceDirect

Applied Catalysis A: General


journal homepage: www.elsevier.com/locate/apcata

Modeling residue hydroprocessing in a multi-fixed-bed reactor system


Anton Alvarez a, Jorge Ancheyta a,b,*
a
Instituto Mexicano del Petróleo, Eje Central Lázaro Cárdenas 152, Col. San Bartolo Atepehuacan, México D.F. 07730, Mexico
b
Escuela Superior de Ingenierı´a Quı´mica e Industrias Extractivas (ESIQIE-IPN), México D.F. 07738, Mexico

A R T I C L E I N F O A B S T R A C T

Article history: A three-phase heterogeneous plug-flow reactor model was developed to describe the behavior of residue
Received 27 August 2008 hydroprocessing in a multi-fixed-bed reactor system. The model considers gas–liquid and liquid–solid
Accepted 9 September 2008 mass-transfer phenomena and incorporates reactions such as hydrodesulfurization (HDS), hydrodeni-
Available online 18 September 2008
trogenation (HDN), hydrodemetallization (HDM), hydrodeasphaltenization (HDAs) and hydrocracking
(HCR) as well as hydrogen consumption. HDS reaction was described by Langmuir–Hinshelwood kinetics
Keywords:
while the rest of the reactions were modeled with power law kinetics. To estimate kinetic parameters,
Modeling
experiments were carried out in a multi-reactor pilot plant loaded with a triple catalyst system under the
Residue hydroprocessing
Fixed-bed reactor
following operating conditions: 380–420 8C temperature and 0.25–1.0 h1 liquid hourly space velocity
(LHSV), keeping constant hydrogen-to-oil (H2/oil) ratio at 891 std m3/m3 and pressure at 9.81 MPa.
Model predictions showed good agreement with experimental data in the range of the studied operating
conditions. The model was also applied for simulating an industrial scale residue hydroprocessing unit
with multi-bed adiabatic reactors and hydrogen quenching.
ß 2008 Elsevier B.V. All rights reserved.

1. Introduction alities are the most used process configurations [5]. The catalyst
system is generally composed by a series of CoMo/NiMo alumina
The increasing supply of heavy oils and the availability of supported catalysts designed for specific objectives such as
residues with high sulfur and metal contents to the refineries, as hydrodesulfurization (HDS), hydrodemetallization (HDM), hydro-
well as the growing demand for transportation fuels of high denitrogenation (HDN), hydrodeasphaltenization (HDAs), hydro-
ecological quality, emphasize the importance of processes for cracking (HCR) and Conradson Carbon (CCR) removal [3,6].
upgrading this type of feeds [1]. Catalytic hydroprocessing is a The main disadvantage of fixed-bed reactors is the loss of
relevant process in petroleum refining industry for impurities catalyst activity over time as a result of premature catalyst
removal of hydrocarbon, ranging from straight-run naphtha to deactivation which reduces drastically the length of run [4]. To
vacuum residue or even heavy and extra heavy crude oils. When solve partially this problem, HDM guard beds with wide pore
handling heavy feeds, hydroprocessing has the ability to increase catalysts for deep metals penetration have been used to protect
the yield of distillates and simultaneously reduce the contents of down stream catalysts. Other important factors are the selection of
sulfur, nitrogen, metals (Ni and V) and asphaltenes [2]. adequate reaction severity and feed preparation (e.g. proper feed
Various technologies such as fixed-bed, moving-bed or fractionation) since they control on the one hand product quality
ebullated-bed reactors are available for hydroprocessing of heavy and on the other hand the catalyst deactivating processes such as
oil fractions [3]. The major selection criterion between each type of coking, HDM and sludge formation caused by high levels of
technology is based on catalyst deactivation rate [4], which hydroconversion [7].
depends on the contents of metals and asphaltenes in the feed To avoid some of these problems, the Mexican Institute of
since products formed during their removal are known as catalyst Petroleum (IMP) has developed a heavy petroleum upgrading
deactivating species. Among all the commercially proven tech- process by means of catalytic hydroprocessing operating at
nologies for heavy fraction hydroprocessing, those using fixed-bed moderate reaction severity thereby keeping low the conversion
reactors in series loaded with catalysts with different function- level of the feed [2,8]. The process employs an arrangement of
several reactors in series loaded with selective catalysts for HDM,
HDAs, HDS, HDN and HCR.
* Corresponding author. Tel.: +52 55 9175 8443; fax: +52 55 9175 8429. Mathematical modeling of this type of processes is a difficult
E-mail address: jancheyt@imp.mx (J. Ancheyta). task due to the complex physical and chemical transformations

0926-860X/$ – see front matter ß 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2008.09.010
A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158 149

Nomenclature
HDAs hydrodeasphaltenization
HDN hydrodenitrogenation
aL gas–liquid interfacial area (cm1)
HDS hydrodesulfurization
aS liquid–solid interfacial area (cm1)
in inlet to the following catalytic bed
AS sectional area of the reactor (cm2)
out outlet of the previous catalytic bed
Asph asphaltene
q quench stream
cP mass heat capacity (J g1 K1)
Ci molar concentration of compound i (cm3 mol1)
Superscripts
Ea activation energy (J mol1)
G gas phase
g gas mass flow rate (g s1)
L liquid phase
GL superficial mass velocity (kg m2 s1)
Q quench fluid
Hi Henry’s law constant for compound i
S solid phase
(MPa cm3 mol1)
H2 hydrogen
DHR overall heat of reaction (kJ kg1 sulfur1)
HC hydrocarbon that undergoes the feed along with mass-transfer phenomena and
catalyst deactivation mechanisms in the reaction system. Most of
H 2S hydrogen sulfide
the work in this field deals with trickle-bed reactor (TBR) models
H2/oil hydrogen-to-oil ratio (std m3/m3)
applied to hydroprocessing of middle distillates [9–14] and
kapp apparent reaction rate constant vacuum gas oils [15–19]. Only few models were developed for
kint intrinsic reaction rate constant heavy feeds, principally by major oil companies (Shell, Chevron,
kj reaction rate constant for reaction j Haldor-Tøpsoe) and research institutes (IFP and KISR) for
kLi gas–liquid mass-transfer coefficient for compound monitoring performance of proprietary processes along time-on-
i (cm s1) stream. Shah et al. [20] and Mhaskar et al. [21] studied the effect of
kSi gas–liquid mass-transfer coefficient for compound the quenching position on catalyst cycle life in a residue HDS/HDM
i (cm s1) fixed-bed reactor. Lababidi et al. [22] simulated an industrial
K H2 S Adsorption equilibrium constant for H2S atmospheric residue desulfurization (ARDS) unit with four fixed-
bed reactors in series, considering metals and coke accumulation
(cm3 mol1)
on the catalyst described with the Chiyoda model [23]. The IFP
l liquid mass flow rate of the j stream (g s1)
developed a model called THERMIDOR (Thermal Monitoring for
nj order of reaction j
Isoperformance Desulfurization of Oil Residues) with an elaborate
N nitrogen catalyst deactivation mechanism for evaluating the overall
Ni nickel performance of the HYVAHL process [24]. Kam et al. [25] presented
Ni molar flow of compound i (mol s1) a model for studying catalyst deactivation mechanisms in ARDS
NH3 ammonia units which considers the different catalyst deactivation stages as
pi partial pressure of compound i (MPa) well as HDS, HDM, HDAs and coking reactions. Juraidan et al. [26]
P total pressure (MPa) refined the accuracy of this latter model by adding noncatalytic
q quench fluid mass flow rate (g s1) hydrothermal removal of sulfur, metals and asphaltenes. More
r reaction rate (mol cm3 s1) recently, Verstraete et al. [27] developed a model for the HDS stage
of the HYVAHL process to predict the changes in elemental
S sulfur
composition of the feed.
T temperature (K)
To design and predict the expected behavior of heavy oil
uG gas superficial velocity (cm s1) upgrading processes at commercial scale adequate reactor
uL Liquid superficial velocity (cm s1) modeling tools are required. It is the objective of this work to
V vanadium develop a TBR reactor model based on detailed experimental data,
z axial position along the catalyst bed which accounts for all major residue hydroprocessing reactions
including hydrogen consumption and mass-transfer phenomena
Greek symbols between the phases, capable of describing the behavior of both
h0 effectiveness factor pilot and industrial scale reactors.
hCE external catalyst wetting efficiency
2. Experimental
nHj 2 hydrogen stoichiometric coefficient for reaction j
1
(mol mol )
H2 S Hydroprocessing experiments were carried out in an isother-
nHDS hydrogen sulfide stoichiometric coefficient for mal multi-reactor pilot plant shown in Fig. 1. The reaction section
hydrodesulfurization reaction (mol mol1) comprises two fixed-bed reactors in series with inter-reactor
rG gas density at process conditions (g cm3) hydrogen make up. Each reactor has an internal diameter of
rL liquid density at process conditions (g cm3) 2.94 cm and a total length of 175.88 cm, with a capacity for loading
yG gas molar volume at process conditions (cm3 mol1) up to 500 cm3 of catalyst. The total reactor volume was loaded
with a commercial size triple catalyst system in the following
Subscripts proportions: 30% front-end HDM catalyst, 30% midsection catalyst
with balanced HDM/HDS activity and 40% tail-end highly active
Gas gas production
HDS/HDN/HCR catalyst. Catalysts were in situ activated by
HCR hydrocracking
sulfiding with straight-run gas oil (1.46 wt% sulfur) containing
150 A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158

1.0 wt% dimethyl disulfide (DMDS). A more detailed description of Table 1


Physical and chemical properties of the feedstock.
Section 2 is given elsewhere [28].
An atmospheric residue (AR) derived from a Mexican heavy Property AR
crude oil (138API) was used as feed. The physical and chemical 3
Density at 15.6 8C, g cm 1.0326
properties of the AR feedstock are presented in Table 1. In order to API gravity 5.4
perform the experiments under a relatively steady-state catalyst Kinematic viscosity @ 121 8C, cSt 1637.9
activity, the fresh catalyst system was subjected to an initial C, wt% 82.50
H, wt% 9.80
coking period of 100 h by treating the AR feed under the following
H/C molar ratio 1.410
conditions: temperature of 360 8C, liquid hourly space velocity S, wt% 5.74
(LHSV) of 0.25 h1, hydrogen-to-oil (H2/oil) ratio of 891 std m3/ N, wppm 5960
m3 and pressure of 9.81 MPa. After this period, the tests were
Metals, wppm
carried out by varying reaction temperature and LHSV (based on Ni 102
total catalyst volume) in the ranges of 380–420 8C and 0.25– V 620
1.0 h1, respectively, in order to study the effect of such variables Ni + V 722
and to determine kinetic parameters of hydroprocessing reac- Asphaltenes in C7, wt% 21.77
tions. H2/oil ratio at the entrance of the reaction system was kept Distillation ASTM D-1160
at 891 std m3/m3 and pressure at 9.81 MPa for all experiments. IBP, 8C 296
Hydrogen make-up was supplied at the entrance of the second 5 vol%, 8C 372
10 vol%, 8C 401
reactor at a rate of 151 std m3/m3 for replenishing hydrogen 15 vol%, 8C 438
consumption in the first reactor and diluting the concentration of 20 vol%, 8C 475
H2S and NH3. After 213 h period of experiments, a check-back test 30 vol%, 8C 521
was performed at the conditions used in the first experiment 40 vol%, 8C 541
vol% recovery @ 538 8C 35.25
(LHSV = 0.25 h1 and T = 380 8C) in order to monitor catalyst
activity.
The properties of the feedstock and collected liquid products
samples from the pilot plant were analyzed using the following
standard methods: API Gravity, ASTM D-287; total sulfur, ASTM- To develop the mass and heat balances the following
4294; total nitrogen, ASTM-4629; asphaltenes insoluble in nC7, assumptions were taken into account:
ASTM-3279; metals (Ni and V), ASTM D-5863 (atomic adsorption);
and yield of residue with boiling point above 538 8C, ASTM D-1160 ^ Catalyst deactivation is considered to be negligible since
distillation. samples were taken at steady-state conditions after initial
period of deactivation.
3. Reactor model ^ Fluid velocities are constant across the reactor.
^ Vaporization of the AR feed in the range of studied operating
A three-phase plug-flow reactor model based on a well-known conditions is less than 6 vol% as shown in Fig. 2, therefore, it is
trickle-bed reactor model [16] was developed in this work. The considered to be negligible.
model considers the two film theory which incorporates mass- ^ Liquid velocity is assumed to be constant through the reactor
transfer at the gas–liquid and liquid–solid interfaces and employs since liquid expansion under moderate hydrocracking condi-
correlations taken from the literature to estimate mass-transfer tions used in the experiments is about 3–5 vol%.
coefficients, gas solubilities and hydrocarbon properties at process ^ Gas velocity is variable due to volumetric contraction caused by
conditions. hydrogen consumption.

Fig. 1. Experimental setup for hydroprocessing experiments.


A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158 151

assumes that there is no vaporization of the feed at the process


conditions, therefore, the mass balance equation for these
compounds is the following:

dCiL 1
¼  kSi aS ðCiL  CiS Þ (5)
dz uL

where i = S, N, Ni, V, Asph and HC.


The compounds transported between the liquid and solid
phases are consumed or produced by chemical reaction on the
solid catalyst. The balance for all compounds at the catalyst
particle can be written as follows:

kSi aS ðCiL  CiS Þ ¼ r j (6)

where i = H2, H2S, NH3, CH4, S, N, Ni, V, Asph and HC and j = HDS,
HDN, HDNi, HDV, HDAs and HCR. The ‘‘’’ sign is for the reactants
while the ‘‘+’’ sign is for the products.
Because of the exothermal nature of hydroprocessing reactions,
an energy balance was included in the model to predict temperature
Fig. 2. Calculated AR vaporization using Hysys process simulator. (~) 710 std m3/ rise along the reactors during commercial operation [29]:
m3, (&) 890 std m3/m3, (*) 1070 std m3/m3.
dT 1
¼ ½ðDHR Þr HDS  (7)
dz uG rG cPG þ uL rL cPL
^ There are no radial concentration and temperature gradients.
^ Chemical reactions take place only at the solid catalyst. The energy balance employs an overall heat of reaction
^ Pressure is constant along the reactors. (7820 kJ/kgsulfur) typical for the atmospheric residue HDS pro-
cesses [30]. Overall heats of reaction are mean values derived from
3.1. Model equations heat balances of several similar HDT processes and include the
contribution of all the reactions [31].
The reactor model considers major residue hydroprocessing The experimental setup comprises an inter-reactor hydrogen
reactions such as HDS, HDN, HDM (nickel (HDNi) + vanadium stream, which affects gas flow rate, superficial velocity and partial
(HDV) removals), HDAs and HCR (conversion of the 538 8C+ pressures at the entrance of the second reactor. The latter effect is
fraction), which take place only on the catalyst surface. Hydrogen, considered in the mass balances as follows [32]:
being the main constituent of the gas phase, is transferred to the
G G G
liquid phase and then to the catalyst surface in order to react with NH 2 ;in
¼ NH 2 ;out
þ NH 2 ;q
(8)
the organic compounds (S, N, Ni, V and asphaltenes (Asph)) and G G G
where NH , NH and NH are the molar flow rates of hydrogen
hydrocarbon (HC). Products such as H2S, NH3 and gases generated 2 ;in 2 ;out 2 ;q
entering the following bed, leaving the previous bed and of the
by chemical reaction on the catalyst surface are released to the gas
inter-reactor stream termed ‘‘quench’’, respectively. For the
phase. Since no reactions occur in the gas phase, the change in
commercial unit, it is necessary to estimate the quench rate for
molar flow of gaseous compounds is equal to the mass-transfer
a desired cooled mixture temperature [33]:
rate to the liquid phase, which can be expressed as follows:
! Z T in Z T in Z T in
dNiG P NiG lout cPL dT þ g out cPG dT þ qcPQ dT ¼ 0 (9)
¼ AS kLi aL P G  CiL (1) T out T out TQ
dz Hi Ni
A list of correlations for determining oil properties at the
where i = H2, H2S, NH3 and gases (represented by CH4). Gas process conditions, gas solubilities and mass-transfer coefficients
superficial velocity and composition in terms of partial pressures at the gas–liquid and liquid–solid interfaces that are used in the
can be determined with the change in molar flow rates (NiG ) using model equations is given in Table 2. Fig. 3 shows a flow diagram of
the following expressions: the computational reactor model.
P
yG NiG
uG ¼ (2) 3.2. Reaction kinetics
AS
Developing mechanistic kinetic models for residue hydroproces-
NG
pGi ¼ P P i G (3) sing reactions is a difficult task due to the complexity of oil
Ni
composition and its analysis. Heteroatoms are present as more than
The concentration of a gaseous compound in the liquid phase one class of compound in heavy fractions, for instance, sulfur occurs
changes along the reactor as result of gas–liquid equilibrium and as mercaptanic, sulfidic, thiophenic, benzothiophenic, dibenzothio-
liquid–solid mass-transfer. The mass balance equation for the phenic and their alkyl derivatives forms, while metals can be found
gaseous compounds in the liquid phase can be written as follows: in porphyrinic and non-porphyrinic form, vanadyl or non-vanadyl,
! ! associated or not to asphaltene molecules [6]. Each class is
dCiL 1 L P NiG L S L S
characterized by its own reactivity and complex reaction paths
¼ ki aL P G  Ci  ki aS ðCi  Ci Þ (4) that are specific to each feed. To account for such a complexity of the
dz uL Hi Ni
feed, the rate of reaction is generally lumped into a single power law
where i = H2, H2S, NH3 and CH4. reaction [22]. Sometimes simplified Langmuir–Hinshelwood mod-
The organic compounds and liquid hydrocarbon are only els are used in order to include the effect of inhibiting species (e.g.
transported from the liquid to the solid catalyst since the model H2S and NH3); however, heteroatom species are still lumped into a
152 A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158

Table 2
Correlations used in the model equations [16].

Parameter Correlation

Oil density rL ð p; TÞ ¼ r0 þ DrP  DrT


   2
P P
DrP ¼ ½0:167 þ ð16:181  100:0425r0 Þ  0:01½0:299 þ ð263  100:0603r0 Þ
1000 1000
DrT ¼ ½0:0133 þ 152:4ðr0 þ DrP Þ2:45 ½T  520  ½8:1  106  0:0622  100:764ðr0 þDrP Þ ½T  5202

Dynamic oil viscosity mL ¼ 3:141  1010 ðT  460Þ3:444 ½log10 ðAPIÞa


a ¼ 10:313½log10 ðT  460Þ  36:447 !
Oil heat capacitya cPL ¼ 4:1868 p
0:415
ffiffiffiffiffiffiffiffiffi þ 0:0009½T  288:15
15:6 rL

y
Henry coefficient Hi ¼ l rN
i L  
Solubility of H2 lH2 ¼ 0:559729  0:42947  103 T þ 3:07539  103 r20
T
þ 1:94593  106 T 2 þ 0:835783
20 2
L ðr Þ
L
b 3
Solubility of H2S lH2 S ¼ 2:51182  1:40751  10 T
Solubilityb of CH4 lCH4 ¼ 0:56631 þ 1:30212  103 T
  
y0:267
Molecular diffusivity DLi ¼ 8:93  108 yL0:433 mT
L
i

Liquid molar volume y ¼ 0:285y1:048


C

3 0:7666
Critical specific volume ym
C ¼ 7:5214  10
0:2896
ðTMeABP Þðr15:6
L Þ

kL aL
 0:4  1=2
G mL
Gas–liquid mass-transfer coefficient i
DL
¼ 7 mL rL DLi
L
i

kS
 1=2  1=3
G mL
Liquid–solid mass-transfer coefficient DL aS
i
¼ 1:8 a mL r DL
S L L i
i

6
Specific surface area aS ¼ dP
ð1  eÞ
a
Taken from [34].
b
Determined at the studied process conditions using Hysys process simulator with the Peng–Robinson equation of state.

single group. Therefore, lumped kinetics are feedstock dependant reactors, produce intrinsic kinetics, whereas, the information
reason why their applicability is limited to a single system. collected from pilot trickle-bed reactors loaded with commercial
HDS reaction is described by the following Langmuir–Hinshel- size catalysts lead to hydrodynamically distorted kinetics. Generally,
wood model that accounts for H2S inhibiting effect [16]: pilot plant tests for generating kinetic data are carried out at the
industrial scale LHSV, however, the dimensions of industrial reactors
nHDS 0:5
ðCSS Þ ðCHS 2 Þ are normally 10–20 times greater than those of pilot reactors [13].
r HDS ¼ kHDS (10)
ð1 þ KHS 2 S CHS 2 S Þ
2 Therefore, superficial mass velocity is much lower in pilot reactors;
this causes problems such as incomplete catalyst wetting, channeling
HDN, HDM, HDAs and HCR reactions are modeled by power law and wall effects producing apparent kinetic rate constants. Apparent
models of nth orders, which are characteristic for the type of feed kinetics can be related with the intrinsic kinetics considering internal
employed for the experiments: diffusion and TBR hydrodynamic effects as follows [37]:

r j ¼ k j ðC i Þn j (11) ka p p ¼ h0 hCE kint (15)


H2S generation and H2 consumption can be determined using
where internal diffusion is represented by the catalyst effective-
overall stoichiometric coefficients in molar basis for the different
ness factor (h0) and the hydrodynamics by the external catalyst
hydroprocessing reactions:
wetting efficiency (hCE).
H2 S
r H2 S ¼ nHDS r HDS (12) Apparent kinetics are frequently used in the industry for scale-
up when experiments were carried out with exactly the same
X
r H2 ¼ nHj 2 r j (13) catalyst (type, size and shape) that has been chosen for commercial
j application [38], this allows for predicting the performance of the
catalyst during commercial operation without needing extra
Gas production is assumed to be exclusively by HCR of the modeling for including internal diffusion [39]. Scale-up procedure
538 8C+ fraction as observed by Sánchez et al. [35], which can be for liquid limited reactions consists in matching pilot and
written as follows: industrial scale LHSV, only when it is certain that the reactors at
r Gas ¼ kGas ðC 538  Cþ ÞnHCR (14) both scales yield the same conversion, in other words, if kinetic
studies were carried out under conditions where catalyst
utilization is complete (hCE  1) [37]. However, it is well known
3.3. Internal diffusion and catalyst wetting efficiency that catalyst wetting efficiency in pilot reactors is in the range of
0.12–0.60 [40], therefore, when matching LHSV, the industrial
The reliability of the collected kinetic data to design commercial reactor will yield higher performance.
units depends on the degree of external and internal mass-transfer In the present work, since experiments were carried out using
limitations which are characteristic of the experimental conditions commercial size catalysts, the effectiveness factor is already
[36]. Kinetic studies using fine particle size under conditions included in the kinetic rate constants. For scale-up, catalyst
where external mass-transfer limitations are reduced, e.g. in batch wetting efficiency was used for correlating apparent and intrinsic
A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158 153

kinetics through superficial liquid mass velocity. The partial


wetting model suggested by Satterfield [40], which has been
applied with good results for correlating pilot plant and industrial
HDT data [13], was considered in this work:

1 1 A
 ¼ (16)
ka p p kin GBL

4. Results and discussion

4.1. Experimental results

During hydroprocessing of heavy oils it is widely recognized


that two periods of deactivation are present: (1) deactivation by
coke deposition, and (2) deactivation by metals deposition. Coke
deposition occurs during the first hours of operation, while metals
continue to be deposited during more time-on-stream. Catalyst
deactivation when hydroprocessing heavy oils can be quite
significant depending mainly on the design of the catalytic system
(e.g. textural properties), reaction conditions and impurities level
in the feed. To improve catalyst stability it is highly recommended
to use a front catalyst with high-metal retention capacity, so that
subsequent catalysts, designed for other purposes, e.g. HDS, can
operate with low-metal content feeds.
The catalytic system employed in this work has been designed
to treat heavy oils and is composed by a high-metal retention
capacity HDM catalyst, followed by HDM/HDS and HDS catalysts.
Therefore, it is expected that catalyst activity after the first and
unavoidable period of deactivation by coke deposition, remains
more or less constant. This is clearly demonstrated in Fig. 4, which
shows the profiles of metals and sulfur contents in hydrotreated
products as function of time-on-stream during the whole
experimentation. At the end, a check-back test was conducted
to monitor catalyst activity and differences of less than 1% were
found compared with those results obtained at the beginning of
experiments (first and check-back experiments are highlighted
with circles in Fig. 4). Therefore, correction of kinetics due to
catalyst deactivation is not necessary and the model assumption of
constant catalyst activity is valid.
The effect of LHSV and temperature on sulfur, nitrogen, metals,
asphaltene and 538 8C+ fraction contents is presented in Fig. 5. As
expected, when LHSV is decreased and temperature is increased
Fig. 3. Flow diagram of the computational reactor model.
the impurities removal and conversion are higher. Conversion
levels are fairly proportional to LHSV decrease for all reactions;
however, temperature effect is greater when temperature is

Fig. 4. Profiles of metals and sulfur content during time-on-stream. (&) Ni + V, (*) S, (*) check-back experiment; T1 = 380 8C, T2 = 400 8C, T3 = 420 8C.
154 A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158

Fig. 5. Effect of LHSV and temperature on product quality. (&) 380 8C, (*) 400 8C, (~) 420 8C.

increased from 400 to 420 8C than from 380 to 400 8C. This square errors between experimental and calculated data using the
behavior of temperature effect has been attributed to a change in Levenberg–Marquardt’s algorithm. The obtained kinetic para-
reaction selectivity from being hydrogenation dominated at low meters are presented in Table 3. A comparison between experi-
temperature to hydrocracking dominated at high temperature mental and calculated values is shown in Fig. 6. The calculated
[41]. values are in good agreement with the experimental ones reason
why the overall trends are well predicted, although with small
4.2. Estimation of kinetic parameters deviations for HDN, HCR and gases.
Reaction orders vary from 0.55 to 2.0, which is typical for
Experimental data were used to estimate kinetic parameters lumped kinetics. Several studies report that residue hydroproces-
of the reactions and stoichiometric coefficients. The optimal set sing reactions follow half to second order kinetics [6,42–47].
of kinetic parameters was determined by minimizing the sum of Lumping large number of components with a wide distribution of

Table 3
Kinetic parameters of hydroprocessing reactions.

Reaction n k0 Ea, kJ/mol


9 0.17 3 0.5 1
HDS 1.17 3.568  10 wt% (mol/cm ) h 104.04
HDN 2.00 1.580  103 wppm1 h1 94.25
HDNi 0.55 8.732  106 wppm0.45 h1 85.39
HDV 1.56 1.323  106 wppm0.56 h1 98.52
HDAs 0.75 7.178  108 wt%0.25 h1 116.79
HCR 2.00 2.843  108 wt%1 h1 141.72
Gas – 8.439  1011 wt%1 h1 198.97
A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158 155

Fig. 6. Comparison of experimental and predicted data. (a) Conversions of hydroprocessing reactions: (&) HDS, (*) HDN, (~) HDNi, (^) HDV, ( ) HDAs, (*) HCR. (b) Gases:
(*) H2 consumption, ( ) H2S yield, (~) gas yield.

reactivities that undergo a complex network of reactions can give phase and on the catalyst surface are presented in Fig. 7.
any apparent reaction order [48]. Apparent reaction orders up to 2 Experimental values are also included for comparison. As
or higher may arise when the feed comprises a broad distribution observed, there is a concentration gradient between the liquid
of reactivities which causes a rapid reaction rate decline since most and solid phases which is reduced progressively with bed depth.
reactive species disappear quickly leaving the most refractory Such a gradient is governed by liquid–solid mass-transfer rate,
ones. This is the case of HDN and HCR reactions which present a calculated from the correlations shown in Table 2, which depends
relatively flat slope with respect to LHSV decrease. The latter seems mainly on the physical properties of the liquid (density and
to be true also for fractional reactions orders (HDNi and HDAs) viscosity) and liquid superficial mass velocity. Since the feed
where the same trend is observed. HDS is closer to first order becomes lighter due to hydrocracking reactions and thereby
kinetics which probably means that the AR feed has sulfur physical properties are improved, liquid–solid mass-transfer is
compounds with a narrower range in reactivities. enhanced reducing such a concentration gradient. The evolution of
the calculated liquid–solid mass-transfer coefficient and liquid
4.3. Simulation of the pilot plant reaction system density as function of bed depth and reaction temperature is
presented in Fig. 8. Liquid–solid mass-transfer increases substan-
The developed trickle-bed reactor model was applied to tially with bed depth and reaction temperature, which explains
simulate the behavior of the pilot reaction system. Molar why in Fig. 7 at 400 8C the concentration gradient disappears more
concentration profiles of sulfur, nitrogen and metals in the liquid rapidly.
Fig. 9 shows the molar concentration profiles of hydrogen and
hydrogen sulfide in the liquid and solid phases generated at a
reactor temperature of 380 8C and LHSV of 0.25 h1. As is well
known, their shapes are determined by a balance between
chemical reaction and gas–liquid mass-transfer. From the entrance
to a relative reactor length of 0.05 (5% of the total length) H2
concentration falls down quickly while H2S concentration rises
substantially due to high reaction rates at that section of the
reactor. After that position, H2 begins to concentrate progressively
in the liquid while H2S concentration has the opposite behavior

Fig. 7. Sulfur, nitrogen and metals molar concentration profiles. Simulated: (—) Fig. 8. Evolution of the liquid–solid mass-transfer coefficient and liquid density. (—)
liquid phase, (  ) solid phase; (*) Experimental. Simulated; (&) experimental.
156 A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158

hand, the contribution of H2S and gases to the total gas volumetric
flow rate is negligible compared to that of hydrogen. The hydrogen
stream injected at a relative reactor length of 0.5 also produces a
jump in H2/oil ratio and superficial gas velocity.

4.4. Scale-up and simulation of industrial scale reactors

The trickle-bed reactor model was also employed to simulate


the commercial heavy oil upgrading process shown in Fig. 12. The
industrial configuration presents significant differences such as
adiabatic operation, hydrogen quenching and gas recycling,
reason why temperature and H2/oil ratio profiles differ substan-
tially from that of the pilot plant. Normally, pilot plant
experiments are compared with adiabatic operations by operat-
Fig. 9. Hydrogen and hydrogen sulfide molar concentration profiles. Simulated: (—) ing the pilot plant isothermally at a temperature that is equivalent
liquid phase, (  ) solid phase. to the average temperature of the commercial unit (i.e. weight
average bed temperature, WABT) [17]. Hence, to validate
since it is released to the gas phase, which implies that this section predicted performance of the commercial unit, a suitable reactor
is governed by gas–liquid equilibrium. In this case, there is also a configuration must be established in order to match average
liquid–solid concentration gradient, although of a greater magni- temperature and H2/oil ratio used in the experiments. This is
tude, which is also diminished as the reaction proceeds. accomplished by (1) determining the number and length of
The evolution of gas phase composition at a reaction catalytic beds and feed temperatures and (2) calculating a
temperature of 380 8C and LHSV of 0.25 h1 is depicted in hydrogen balance for distributing the amount of hydrogen for
Fig. 10. Experimental data were also included for verifying the quenching, recycling and make up [49]. The following operating
accuracy of the model. As observed, H2 partial pressure decreases average conditions were considered for the commercial unit:
rapidly as a result of H2 consumption, on the other hand, H2S and LHSV of 0.25 h1, temperature of 380 8C, H2/oil ratio of
gases generation lead to an increase in partial pressure along the 887 std m3/m3 and pressure of 9.81 MPa.
reactor length. At a relative reactor length of 0.5, which Fig. 13 presents one possible configuration of industrial reactors
corresponds to the inter-reactor zone, a notorious jump in H2 along with the simulated temperature and H2/oil ratio (relative to
partial pressure and a fall in H2S and gases pressures can be the same basis as in Fig. 11) profiles, which average values (- - -)
appreciated. This is caused by the inter-reactor hydrogen stream as
described in Section 2. Such a stream enriches the gas phase with
H2 and has the advantage of diluting HDT reaction inhibitors such
as H2S and NH3 towards the end of the reaction system. This type of
gas composition profile is typical in industrial HDT reactors since
they employ multiple hydrogen quench streams.
Fig. 11 shows the profiles of superficial gas velocity and H2/oil
ratio, this latter relative to that of the entrance of the first reactor,
at reactor temperatures of 380 and 400 8C and LHSV of 0.25 h1. For
the H2/oil ratio profiles, experimental values at the exit of the
reaction system are shown. In general, the profiles of both variables
are similar to that of H2 partial pressure shown in Fig. 10. Such
similarity is owed to H2 consumption along the reactors which
decreases H2/oil ratio and produces gas volumetric contraction
reducing superficial gas velocity. As expected, the decrease in H2/
oil ratio and superficial gas velocity is more rapid at higher
temperatures as a result of higher H2 consumption. On the other

Fig. 11. Superficial gas velocity and H2/oil ratio profiles. Simulated: (—) 380 8C, (  )
Fig. 10. Evolution of gas phase composition. (—) Simulated; (*) experimental. 400 8C; (*) experimental.
A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158 157

Fig. 12. Basic configuration of the heavy oil upgrading process [8].

match adequately the conditions used in pilot experiments. To H2/oil ratio profile by solving hydrogen balances; since quench
limit the temperature rise caused by the hydroprocessing hydrogen comes from the hydrogen recycle stream, any extra
reactions, the total catalyst volume of the first reactor (R1) was hydrogen quench stream would reduce substantially the H2/oil
divided in 4 catalytic beds while the second reactor (R2) required 2 ratio at the entrance of R1 leading to lower average values. When
beds. Bed inlet temperatures and delta-Ts for each reactor were hydrogen quenching is restricted, heat exchange is used for cooling
adjusted to be more or less equal in order to match the average the rest of the bed effluents.
temperature in all the catalytic beds. Due the uniformity of bed The figure also depicts the superficial gas velocity profile.
delta-Ts, each bed length increases progressively towards the end Differently from pilot plant, superficial gas velocity is influenced
of a reactor. In R1 a very sharp temperature rise is observed as a besides H2 consumption, by the temperature rise. While in pilot
result of high reaction rates in this section, therefore, quenching is plant gas velocity decreases rapidly, in the industrial reactors gas
required. It can be noticed that only two hydrogen quench streams velocity in the first three beds is more or less constant and in the
injected in specific positions in R1 are employed, causing a sudden rest it decreases very slowly. Temperature rise produces expansion
jump in H2/oil ratio. This arrangement results from matching the of the gas which increases velocity. Such an effect copes with the

Fig. 13. Industrial reactor configuration and simulated profiles of temperature, H2/oil ratio, superficial gas velocity, hydrogen partial pressure and molar concentration. (- - -)
Average values.
158 A. Alvarez, J. Ancheyta / Applied Catalysis A: General 351 (2008) 148–158

Table 4 configuration with multi-bed adiabatic reactors and hydrogen


Comparison of the model predictions for the industrial and pilot reactors.
quenching. The results indicated that industrial reactors had from
Industrial Pilot Difference (Ind-Pilot) 8 to 18% higher performance than pilot plant, which is attributed
to complete catalyst utilization.
HDS, % 88.14 75.86 12.28
HDN, % 59.60 48.62 10.98
HDNi, % 65.43 47.38 18.05 References
HDV, % 84.80 75.96 8.84
HDAs, % 59.08 43.17 15.91 [1] J. Ancheyta, M.S. Rana, E. Furimsky, Catal. Today 109 (2005) 3–15.
HCR, % 31.40 23.47 7.93 [2] J. Ancheyta, G. Betancourt, G. Marroquı́n, G. Centeno, L.C. Castañeda, F. Alonso, J.A.
H2 consumptiona, % 19.23 14.03 5.20 Muñoz, M.T. Gómez, P. Rayo, Appl. Catal. A 233 (2002) 159–170.
a [3] M.S. Rana, V. Sámano, J. Ancheyta, J.A.I. Diaz, Fuel 86 (2007) 1216–1231.
Values are relative to the amount of fed hydrogen in pilot plant experiments.
[4] S.T. Sie, Appl. Catal. A 212 (2001) 129–151.
[5] E. Furimsky, Appl. Catal. A 171 (1998) 177–206.
[6] A. Marafi, S. Fukase, M. Al-Marri, A. Stanislaus, Energy Fuels 17 (2003) 661–668.
loss in gas volume by H2 consumption. Also, there are two gas [7] E. Furimsky, F.E. Massoth, Catal. Today 52 (1999) 381–495.
velocity jumps caused by hydrogen quenching along with three [8] J. Ancheyta, G. Betancourt, G. Marroquı́n, G. Centeno, J.A.D. Muñoz, F. Alonso, U.S.
drops corresponding to the inter-bed heat exchangers which cause Patent 2007/0187294 A1.
[9] G.F. Froment, G.A. Depauw, V. Vanrysselberghe, Ind. Eng. Chem. Res. 33 (1994)
volumetric contraction by thermal effect. 2975–2988.
H2 partial pressure and molar concentration profiles are also [10] B.W. van Hasselt, P.J.M. Lebens, H.P.A. Calis, F. Kapteijn, S.T. Sie, J.A. Moulijn, C.M.
shown in Fig. 13. In R1, the two slight jumps in partial pressure van den Bleek, Chem. Eng. Sci. 54 (1999) 4791–4799.
[11] R. Chowdhury, E. Pedernera, R. Reimert, AIChE J. 48 (2002) 126–135.
are produced by hydrogen quenching. The overall shape of the [12] D.G. Avraam, I.A. Vasalos, Catal. Today 79–80 (2003) 275–283.
hydrogen concentration profile differs slightly from that of the [13] M. Bhaskar, G. Valavarasu, B. Sairam, K.S. Balaraman, K. Balu, Ind. Eng. Chem. Res.
pilot plant. Towards the end R1 and in R2 there are 3 small saddles 43 (2004) 6654–6669.
[14] A. Murali, R.K. Voolapalli, N. Ravinchander, D.T. Gokak, N.V. Choudary, Fuel 86
which correspond to the inter-bed heat exchangers. Since H2
(2007) 1176–1184.
solubility increases with temperature, the inter-bed heat removal [15] S. Mohanty, D.N. Saraf, D. Kunzru, Fuel Process. Technol. 29 (1991) 1–17.
produces a H2 release to the gas phase reducing concentration. The [16] H. Korsten, U. Hoffmann, AIChE J. 42 (1996) 1350–1360.
same effect could be expected for the hydrogen quenching [17] M.A. Rodrı́guez, J. Ancheyta, Energy Fuels 18 (2004) 789–794.
[18] F.S. Mederos, J. Ancheyta, Appl. Catal. A 332 (2007) 8–21.
positions; however, the quench stream increases partial pressure [19] H. Kumar, G. Froment, Ind. Eng. Chem. Res. 46 (2007) 5881–5897.
which counterbalances the temperature effect on solubility. [20] Y.T. Shah, R.D. Mhaskar, J.A. Paraskos, Ind. Eng. Chem., Process Des. Dev. 15 (1976)
Finally, a comparison of the model predictions for the industrial 400–406.
[21] R.D. Mhaskar, Y.T. Shah, J.A. Paraskos, Ind. Eng. Chem. 17 (1978) 27–33.
and pilot reactors is given in Table 4. According to the model, the [22] H.M.S. Lababidi, H.I. Shaban, S. Al-Radwan, E. Alper, Chem. Eng. Technol. 21 (1998)
industrial unit yields from 8 to 18% more in conversion of the 193–200.
different reactions which is consequently accompanied by 5% more [23] S. Kodama, H. Nitta, T. Takatsuka, T. Yokoyama, J. Jpn. Petrol. Inst. 23 (1980) 310–
320.
H2 consumption. This represents a considerable difference [24] H. Toulhoat, D. Hudebine, P. Raybaud, D. Guillaume, S. Kressmann, Catal. Today
between the performances of both scales. Higher performance 109 (2005) 135–153.
of industrial scale trickle-bed reactors is only owed to full catalyst [25] E.K.T. Kam, M. Al-Shamali, M. Juraidan, H. Qabazard, Energy Fuels 19 (2005) 753–
764.
utilization in absence of mass-transfer limitations since operating [26] M. Juraidan, M. Al-Shamali, H. Qabazard, E.K.T. Kam, Energy Fuels 20 (2006)
conditions were matched to that of pilot scale reactors. As 1354–1364.
discussed earlier, the difference in performance when scaling-up is [27] J.J. Verstraete, K. Le Lannic, I. Guibard, Chem. Eng. Sci. 62 (2007) 5402–5408.
[28] A. Alvarez, J. Ancheyta, Ind. Eng. Chem. Res. (2009) in press.
given by the catalyst wetting efficiency factor. Model predictions
[29] O.M. Tarhan, Catalytic Reactor Design, McGraw-Hill, New York, 1983.
must be compared with real industrial data in order to verify the [30] Y.T. Shah, J.A. Paraskos, Chem. Eng. Sci. 30 (1975) 1169–1176.
accuracy of the partial wetting model. [31] W. Döhler, M. Rupp, Chem. Eng. Technol. 10 (1987) 349–352.
[32] A. Alvarez, J. Ancheyta, Chem. Eng. Sci. 63 (2008) 662–673.
[33] G.D. Stefanidis, G.D. Bellos, N.G. Papayannakos, Fuel Process. Technol. 86 (2005)
5. Conclusions 1761–1775.
[34] R.H. Perry, D.W. Green, Perry’s Chemical Engineers’ Handbook, McGraw-Hill,
A trickle-bed reactor model for describing residue hydro- New York, 1999.
[35] S. Sánchez, M.A. Rodrı́guez, J. Ancheyta, Ind. Eng. Chem. Res. 44 (2005) 9409–
processing in a fixed-bed reactor system was developed in the 9413.
present contribution. The model took into account major residue [36] S.K. Bej, Energy Fuels 16 (2002) 774–784.
hydroprocessing reactions (HDS, HDN, HDM, HDAs and HCR), [37] M.H. Al-Dahhan, in: Proceedings - International Symposium on Advances in
Hydroprocessing of Oil Fractions, Morelia, Mexico, June 26–29, 2007.
hydrogen consumption and mass-transfer phenomena between [38] S.T. Sie, Revue de L’Institute Francais du Pétrole 46 (1999) 501–515.
the gas–liquid and liquid–solid phases. A set of experiments was [39] M.H. Al-Dahhan, M.P. Dudukovic, AIChE J. 42 (1996) 2594–2606.
carried out in a multi-reactor pilot plant in order to evaluate the [40] C.N. Satterfield, AIChE J. 21 (1975) 209–228.
[41] J. Ancheyta, G. Centeno, F. Trejo, G. Marroquı́n, Energy Fuels 17 (2003) 1233–
effect of LHSV and temperature on hydroprocessing reactions.
1238.
Experimental data were used to determine kinetic parameters and [42] M.A. Callejas, M.T. Martı́nez, Energy Fuels 13 (1999) 629–636.
stoichiometric coefficients of the involved reactions. The devel- [43] M.A. Callejas, M.T. Martı́nez, Energy Fuels 14 (2000) 1304–1308.
[44] M.A. Callejas, M.T. Martı́nez, Energy Fuels 14 (2000) 1309–1313.
oped model was employed to simulate the behavior of the pilot
[45] A. Marafi, H. Al-Bazzaz, M. Al-Marri, F. Maruyama, M. Absi-Halabi, A. Stanislaus,
plant reaction system. The reactor model allowed for observing Energy Fuels 17 (2003) 1191–1197.
the evolution of the concentration of various compounds in [46] F. Trejo, J. Ancheyta, Catal. Today 109 (2005) 99–103.
the gas, liquid and solid phases along the reactor. In general, the [47] S. Sánchez, J. Ancheyta, Energy Fuels 21 (2007) 653–661.
[48] M.R. Gray, A.R. Ayasse, E.W. Chan, M. Veljkovic, Energy Fuels 9 (1995) 500–506.
predictions showed good agreement with experimental data. The [49] J.A.D. Muñoz, A. Alvarez, J. Ancheyta, M.A. Rodrı́guez, G. Marroquı́n, Catal. Today
model was also applied for scaling up pilot plant data to a 109 (2005) 214–218.

You might also like