Optimisation of the Mexican Refining System for low sulphur

diesel production.

Julian Castellanos, IMP, Mexico; Alberto Alcaraz, PEMEX, Mexico; Jorge
Cuellar, IMP, Mexico; Gustavo Sanchez, PEMEX, Mexico.

Abstract.

The Mexican refining industry is facing the challenge of producing ultra-clean fuels, while
processing increasingly proportions of heavy and sour crude oil. Of particular relevance in the
strategies devised for optimising the overall domestic petroleum industry, are the actions for
assuring the supply of low sulphur diesel within a schedule that considers different stages for
achieving reductions from 500 to 300 ppm in the near future, to 50 ppm for the year 2006. Other
specifications considered are cetane, aromatics, density and distillation points.
The Mexican Refining System includes 6 refineries with a total processing of 8,140.6 m
3
/h.
Diesel feedstock is mainly constituted by atmospheric gasoil; all streams are hydrotreated in 14
units operating in average at 52.6 x 10
5
N/m
2
g. that produce diesel fuel with less than 500 ppm
sulphur. Future streams to be incorporated, resulting from ongoing projects, include cracked
products from coking and FCC plants. Additionally the refinery reconfiguration in progress will
permit the processing of a higher proportion of heavy crude oil, representing a movement from
2.2% to 2.6% sulphur in the crude blend.
The reported strategies in this paper are based on a formal mathematical model and an
optimisation exercise, considering operating condition adjustments, catalyst replacement,
revamp and new installations. Technology projects for process and catalyst development were
considered as scenarios for the decision taking process.

Introduction (Diesel specification scenarios).

The environmental legislation that affects fuel specifications vary from country to country and
even in the same country may be different for particular regions or jurisdictions. In any case, the
overall trend is towards a severe reduction of sulphur content, with limits that will surely evolve
in response to the combination of the economies and environmental problems of each region
and the progress in process and catalyst technologies. In particular for diesel, an ultra-low
sulphur product is foreseen, looking for a drastic reduction of SO
2
emissions and the possibility
of applying advanced catalytic converter technologies for NOx reduction in the exhaust gases.
There are some other diesel specifications which future is more difficult to predict.

Region USA European Union Sweden Mexico
EPA California (1)
Year 1994 2006 1993 2006 2000 2005 1991 2000 2006
Sp. gr. min. 0.83 0.83 0.8
Sp. gr. max. 0.876 0.876 0.86 0.86 0.845 0.825 -
0.845
0.82 0.85
Sulphur, ppm max 500 15 500 15 350 50 10 500 50
Cetane Index min. 40 40 50
Cetane Number min. 48 48 51 54 - 58 48 48
Aromat, vol% max. 35 35 10 10 5 30 20
PNA, wt. max. 1.4 1.4 11 1-6 0.02
vol.
11 5
T90, K max. 611.15 611.15 594.15 594.15 613.15-
633.15
558.15 618.15 613.15
T95, K max. 639.15 639.15 633.15 573.15
EP, K max. 621.15 621.15
(1) Reference emissions fuel
Table 1 Diesel specifications trend.

In the long run, the motor engines based on fuel cells will also demand ultra-low sulphur fuels.



Diesel hydrotreating chemistry.

Up to now catalytic hydroprocessing is the technology extensively used for reducing the sulphur
content of fuels. The combination of more active catalysts, more efficient processes and
reactors with better design concepts has allowed the elaboration of diesel products that are
presently specified in the range of 10 to 500 ppm S, as shown in Table 1. The process
simultaneously incorporates hydrogen into the non-saturated molecules, promoting a reduction
of density, volumetric yields over 100%, and improvement on the cetane number.

The chemistry associated with the process seems to be very simple: cleavage of C-S bonds
with H
2
S release, hydrocracking of hydrocarbons and hydrogenation of aromatic rings, but
indeed it is rather complex when considering the diversity of chemical species participating in
the reactions, each with very different reactivity behaviour (Table 2). At the same time, there are
catalyst options that selectively promote hydrogenation or hydrogenolysis reactions.

Relative Reactivity
@ 320 C (1)
DBT 1.000
1-MeDBT 1.778
4-MeDBT 0.199
46-DiMeDBT 0.077
24-DiMeDBT 0.325
13-DiMeDBT 2.117
23-DiMeDBT 2.402
4-PrDBT 0.288
4Et6MeDBT 0.110
146-TriMeDBT 0.171
346-TriMeDBT 0.445
(1) Derived from G. F. Froment et al.
1

Table 2 Hydrodesulphurisation reactivity of some dibenzothiophene molecules.

For an adequate prediction of the performance of a Diesel Hydrodesulphurisation Process (D-
HDS), it is necessary to consider a minimum of the participating species for representing the
whole of the reacting system. For the same level of total sulphur content, the response of a
process-catalyst combination may be quite different depending on the feedstock characteristics.
The nature of the crude oil is important, so it is the source of the feedstock. Streams resulting
from upstream processing, like coking, FCC or vacuum gasoil hydroprocessing will have a very
different response from that of a virgin atmospheric gasoil.

wt.% Istmo Maya
BT 0.0000 0.2298
C1-BT 0.7030 3.9977
C2-BT 3.2040 7.1109
C3-BT 5.4346 8.3630
C4-BT 2.1495 10.9650
C5-BT 5.4346 5.4279
DBT 2.3523 0.6548
C6-BT 2.0549 0.0000
C1-DBT 16.2228 13.9920
C2-DBT 11.7210 3.7335
C3-DBT 9.3822 3.4118
C4-DBT 3.6366 4.1700
C5-DBT 5.9213 2.2286
C6-DBT 9.4768 0.2183
BNT 22.3063 35.4968
TOTAL 100 100

Table 3 Benzothiophenes distribution in Mexican crude oils.
Reaction model.

The fitting of kinetic models for a set of chemical species representative of the overall reactivity
of the sulphur compounds becomes absolutely necessary for a formal optimisation analysis of
D-HDS units. The model should represent the behaviour of the catalyst and feedstock of the
existing or projected plant. The kinetic model associated to a reactor model, and adapted as a
user subroutine in a commercial process simulator, is a valuable analysis tool. The simulator
may then be used for predicting the performance of different feedstock in an existing unit, for
planning operating conditions, or for designing or revamping plants with optimisation criteria.

A model based on adaptations from Broderik and Gates
2
and M. Castellanos
3
for two species,
DBT and 46-DiMeDBT, has shown to be a good approach for representing a set of
16 experimental runs performed at the Mexican Petroleum Institute pilot plants, as described in
Table 4:
















Figure 1 - Dibenzothiophenes hydrogenolysis and hydrogenation reactions.




























Figure 2 - Hydrocracking reactions
3
.










Figure 3 - Aromatic saturation reactions














Figure 4 - Reactor material balance for compound i



































) 1 ( ) 1 (
785 . 0

1
2 2
1
2
4
2
3 2

1
2
m
DBT
m
H BP S H BP BP
H DBT
BP
cat BP
P k C k C k
P C k
F dL
dC
+ + +

=

DBT CHB
H DBT CHB cat CHB
C k
P C k
F dL
dC
m
2
1
2
1


785 . 0

1
2
+

=

dL
dC
dL
dC
dL
dC CHB BP DBT
=

) ' 1 ( ) ' ' 1 (
'

785 . 0

2
2 2
2
2
4
2
3 46 2
46
1
2
46
m
m
H BP S H BP DiMeDBT BP
H DiMeDBT
BP
cat DiMeBP
P k C k C k
P C k
F dL
dC
+ + +

=

DiMeDBT CHB
H DiMeDBT CHB cat DiMeCHB
C k
P C k
F dL
dC
m
46 2
46 1
2
46
' 1
'

785 . 0

2
2
+

=

dL
dC
dL
dC
dL
dC DiMeCHB DiMeBP DiMeDBT 46 46 46
=
































Sulphur, ppm Pressure,
N/m
2
x10
-5
H
2
/Hc ratio,
m
3
std/m
3
LHSV, h
-1
Temperature,
K
Experimental Model
43.1 374.1 1.8 623.15 566 628
43.1 374.1 1.8 643.15 220 206
43.1 445.3 2.1 623.15 778 720
43.1 445.3 2.1 633.15 404 398
43.1 445.3 2.1 643.15 214 240
54.9 445.3 1.0 623.15 313 294
54.9 445.3 2.1 623.15 683 681
54.9 445.3 2.1 633.15 338 334
54.9 445.3 2.1 643.15 163 168
59.8 374.1 1.0 623.15 306 318
59.8 374.1 1.0 643.15 89 91
59.8 445.3 1.0 623.15 260 254
59.8 445.3 1.0 633.15 131 109
59.8 445.3 2.1 623.15 572 597
59.8 445.3 2.1 633.15 297 281
59.8 445.3 2.1 643.15 125 129

Table 4.- Pilot plant information and reaction model fitness.


Mexican Refining System (MRS) description.

The MRS consists of 6 refineries, presently processing 8,140.6 m
3
/h of a blend of 36.7/63.3
vol.% of Maya/Istmo crude oils.


=
(

=


= + =
1
, 1
PM - PM
PM - PM

785 . 0

1 n
1 i k
H i
H k
2
2
2
i n i
n i K C C K
F dL
dC
i
i j
ij Di Dk ki
cat Di
PNA PNA
H PNA PNA cat
PNA
C k
C C k
F dL
dC
2
1
2
1


785 . 0

2
+

=


1

1


785 . 0

2
1
2
1
2
2 2
|
.
|

\
|
+

+

=
A A
H A A
PNA PNA
H PNA PNA cat
A
C k
C C k
C k
C C k
F dL
dC
( )
dL
dC
n
dL
dC
n n
dL
dC
dL
dC
dL
dC
dL
dC
dL
dC A PNA DiMeCHB CHB DiMeBP BP H
2 2 1
46 46
5 2
2
+ + +
|
.
|

\
|
+
|
.
|

\
|
+ =
|
.
|

\
|
+ =
dL
dC
dL
dC
dL
dC DiMeDBT DBT S H 46
2














Figure 5 Current product slate of the Mexican Refining System

Property Istmo Maya
Sp. gr. 0.851 0.921
Viscosity @ 25C, cst 7.3 155.3
Sulphur, w.% 1.5 3.4
Carbon Ramsbottom, w.% 3.5 10.6
Asphaltenes, w.% 2.4 14.8
Vanadium, w. Ppm 41.4 308.0
Yield, vol. %
Light naphtha 341.15 K 5.1 2.9
Heavy naphtha 448.15 K 20.9 12.8
Kerosene 533.15 K 16 11.5
Atmospheric gasoil 633.15 K 15 14.6
Vacuum gasoil 813.15 K 23.5 19.7
Vacuum residue 19.5 38.5

Table 5 - Istmo and Maya crude oils main characteristics

14 D-HDS units operating with atmospheric gasoil (AGO) produce 1,755.3 m
3
/h of diesel fuel
with an average of 340 ppm of S. The units, with variations described in Table 6, have the
following average operating conditions: Pressure 52.6 x 10
5
N/m
2
g., temperature 630.17 K,
space velocity 2.2 h
-1
and H
2
/ hydrocarbons ratio 311.4 m
3
std/ m
3
. Nine units employ IMP
catalyst and the other five a commercial product.





















Table 6- Current D-HDS operating conditions and performance

Optimisation of the Mexican Refining System.

The development of catalysts and the use of the reaction model have supported the
optimisation of the Mexican Refining System.

Several refinery schemes were considered and analysed with the aim of accomplishing
volumetric and specification demands of different fuels. However the volumetric satisfaction was
limited by the achievable increase in atmospheric distillation capacity in accordance with the
feasible de-bottlenecking of downstream units. The total refining capacity will reach 10,134.7
m
3
/h (24.5% increase).

A large reconfiguration project of the MRS is now being implemented with the additional goal to
allow the processing of a larger proportion of heavy Maya crude oil (36.7 55.7%). In the
overall, 89% more m
3
/h of this crude will be processed in the system, with one of the refineries
receiving up to 83.9% of Maya. The new whole crude blend represents a challenge for
achieving qualitative and quantitative goals. Sulphur will increase from 2.2 to 2.7%. On the other
hand, as volumetric targets are accomplished by new conversion units, diesel feedstock will
include cracked material (398.5 m
3
/h from three new coking units, plus 151.4 m
3
/h of LCO).
Diesel production (2,734 m
3
/h) increases by 55.7%. The diesel hydrodesulphurisation capacity
is expanded with seven new plants and one revamp (Table 6).
















Figure 6 Slate products of the reconfigured Mexican Refining System

Refinery Unit (1) Project
scope
Flowrate,
m
3
/h
Sulphur, wt
ppm
Feedstock
Cadereyta U-700-I No change 148.6 500 AGO
HDD-3 New 242.9 500 AGO/LCO (73%/27%)
U-3800 New 123.5 500 CokGO
Total 515.0 500
Cd. Madero U-501 Revamp 181.4 200 AGO
U-502 New 137.2 200 AGO/LCO (87%/13%)
Total 318.6 200
Minatitln HDD No change 94.2 1,000 AGO
U-100 No change 117.6 500 AGO
U-24000 New 160.0 200 AGO/LCO (80%/20%)
U-11000 New 120.4 200 CokGO
Total 492.1 420
Salamanca U-7 No change 81.1 500 AGO
U-8 No change 81.1 500 AGO/HOIL-LGO (76%/24%)
U-14 No change 152.8 500 AGO
Total 315.0 500
Tula U-800-I No change 135.9 500 AGO
U-700-II No change 135.9 500 AGO
U-800-II No change 134.4 500 AGO
HDD-5 No change 173.0 500 AGO/LCO (80%/20%)
U-24000 New 18.8 200 HDSHVGO
Total 598.0 490
Salina Cruz U-700-I No change 112.2 500 AGO
U-800-I No change 114.2 500 AGO
U-800-II No change 114.1 500 AGO
U-HDG New 154.6 200 CokGO
Total 495.1 410
TOTAL 2,734 433

(1) Some units change service from jet fuel to diesel hydrodesulphurisation and vice versa

Table 7 Diesel streams after reconfiguration

The unit revamp and the design of the new installations performed by IMP were defined by
optimisation exercises that involved the incorporation of operating and investment cost functions
associated to the process. An objective function was defined based on the net present value
and connected with the following decision variables: reaction temperature and pressure, space
velocity, H
2
/Hc ratio, hydrogen purity (adjusted with binary decisions related with the inclusion or
not of PSA and DEA sweetening units for purifying H
2
feed or removing H
2
S in the H
2
recycling
stream). The analysis was performed with a commercial simulator and a user subroutine based
on the described reactor model.

The new process schemes include three reaction beds inside the same reactor vessel, the use
of intercooling in the reactor for controlling the temperature when using cracked feedstock and a
better thermal integration.

The diesel hydrotreating units of the reconfigured system, using the latest more active IMP
catalyst versions, will allow the production of a 300-ppm product.
















Figure 7 Current and reconfigured process schemes

Ultra-low sulphur diesel production.

In prevision for the future specifications, several studies are in progress focused to the
economic production of ultra-low sulphur diesel. Sulphur reduction to very low levels implies its
removal out of complex molecules that exhibit a refractory behaviour for hydrogenolysis
reactions with presently available catalysts. 46-DiMeDBT is one example of such molecules,
although there are many others, particularly multiple-substituted dibenzothiophenes of high
molecular weight present in high boiling point fractions. Therefore the importance of a short cut
distillation when preparing the feedstock for this type of diesel. A sharp 603.15 K cut could be
recommended for avoiding heavy sulphur molecules.

The design of future diesel hydrodesulphurisation processes will require to consider many other
factors than just the selection of operating conditions. It will be of paramount importance a
precise characterisation of the sulphur species present in the feedstock, and their performance
evaluation in the new developed catalysts. While real efficient catalysts are not available, a first
approach together with a lower cut point gasoil, will be the restriction to virgin feedstock.
Cracked streams like FCC light cycle oil or coking gasoil, rich in difficult sulphur compounds,
may require other direction in the refining schemes.

There is already some progress in the comprehension of the performance of different sulphur
compounds
5,6
. It is well known that the hydrodesulphurisation of more reactive molecules, like
DBT, takes place preponderantly through the hydrogenolysis route. From the experimental work
referred in this paper, this fact became evident when the reaction velocity constants of both
routes, hydrogenolysis vs hydrogenation, were compared. Other interesting fact resulting from
the experiments is the low influence of the hydrogen partial pressure on the hydrogenolysis rate
of reaction: the P
H2
exponent in the kinetic model is very low. An inference in consistence with
the process design criteria is the unnecessary use of high operating pressure for the production
of a conventional diesel product where only easy sulphur compounds need to be eliminated.

For ultra-low sulphur diesel production the much more refractory molecules need to be
removed. These molecules, like 46DiMeDBT, exhibit a preferred route through hydrogenation,
but still with a very low overall reaction velocity. According to data in Table 2, the reactivity at
593.15 K of this molecule is 13 times smaller than that of DBT, and under more typical
operating conditions (613.15-633.15 K) the ratio is in the range of 20:1. These difficult
molecules limit the specification of low sulphur diesel and become the target to attain an ultra-
low sulphur product.

Although the reactivity of the hydrogenolysis hydrogenation routes inverts for difficult
molecules, the absolute rate of hydrogenation remains very low when using conventional CoMo
catalysts. To promote the reaction, a hydrogenation active catalyst based on NiMo is a better
option. On the other hand, hydrogenation is a much more H
2
demanding route. This implies a
higher H
2
consumption and a greater influence of hydrogen partial pressure in the performance.
Ultra-low diesel production requires higher pressure and H
2
/Hc ratio.

Some ideas are presently being developed for the revamp of old low pressure D-HDS units
based on the incorporation of a second high pressure reaction loop. The replacement of CoMo
catalyst by NiMo version is only recommended for high pressure operating conditions.

Additionally it will be necessary to consider the negative effect of light hydrocarbon compounds
normally accompanying the H
2
feed, as they reduce its partial pressure. Therefore a purification
of hydrogen input may benefit the process performance. In the same line of thought is the
negative effect of H
2
S in the H
2
recycled stream, as inferred from the kinetic model. A H
2
S
removal from these streams becomes another recommended action for the process design.

A fundamental explanation of the refractory behaviour of 46DiMeDBT molecules and alike is
found in the molecular geometry. These molecules are basically flat, promoting a low access to
active centres of the catalyst, due to the steric hindrance of the alkyl groups in the position 4,6.
An explanation of the preferred hydrogenation route may be the modification of the molecule
geometry by the steps yielding 46DiMeTHDBT and 46DiMeHHDBT (Figure 8). In this latter
molecule the steric factor is mitigated as it finds more incidence approach angles to the active
sites. This is reflected in a better reactivity for the hydrogenolysis of this molecule, the last stage
of the route.

Another factor to consider is the equilibrium controlled reaction 46DiMeTHDBT + H
2

46DiMeHHDBT. Being an exothermic reaction, the conversion is negatively influenced by high
temperature, as required by kinetic considerations. This effect is experimentally demonstrated
by the asymptotic behaviour on the sulphur reduction with temperature increase.

Two non-exclusive solutions may be anticipated for this equilibrium limitation: One is the
development of much more active catalyst to allow the reaction at lower temperatures; this will
favour the reaction in the desired direction. The other is to employ a catalytic distillation
approach in the process design. As above mentioned, one action already being considered is
the use of NiMo based catalyst. An extension would be the use of combined NiMo CoMo
beds. The design of active species in the catalyst is now a new optimisation problem.























Figure 8 Reactions routes to remove 46-DiMeDBT

Conclusions.

Several conclusions may be derived from the arguments presented in this paper:
The use of fundamental kinetic and reactor models may become a critical tool for the
analysis and design of D-HDS units.
It is demonstrated that the proposed model reasonably reproduce experimental information.
The model may be improved by the inclusion of several sulphur species kinetics.
It is highly recommended to fit this or any other model to specific catalyst feedstock.
The model has been successfully applied for redefining the Mexican Refining System and it
is considered for planning the production of ultra-low sulphur diesel.
Ultra-low sulphur diesel production for Pemex is particularly challenging given the nature of
the available crude oils and dependency on conversion units to accomplish volumetric
targets.
Ultra-low sulphur diesel plant analysis and design will require:
The development of catalyst with higher hydrogenation activity.
The optimisation of the catalyst bed, considering different catalytic functions.
The selection of proper feedstock.
The analytical information of sulphur chemical species in feedstock.
The use of higher pressure and lower temperature.
The supply of additional hydrogen.
The use of hydrogen purification units.
The re-conception of D-HDS process schemes, specially for revamping old low
pressure units.

Notation.

Catalytic bed diameter

cat
Catalyst density
Reactor stream molar density
F Reactor stream molar flowrate
Fi Reactor stream molar flowrate of compound i
L Reactor length
C
i
Molar concentration of component i
P
H2
Hydrogen partial pressure
r
i
Reaction rate of formation per unit catalyst mass of compound i


k
iBF
DBTBF reaction kinetic and adsorption parameters
k
iCHB
DBTCHB reaction kinetic and adsorption parameters
k
iBF
46DiMeDBT46DiMeBF reaction kinetic and adsorption parameters
k
iCHB
46DiMe DBT46DiMeCHB reaction kinetic and adsorption parameters
PMi Molecular weight of hydrocarbon i
K
ij
Hydrocracking kinetic constant of hydrocarbon i yielding hydrocarbon j
k
iA
PNAA reaction kinetic and adsorption parameters
k
iN
AN reaction kinetic and adsorption parameters

Compound name abbreviations.

DBT Dibenzothiophene
BP Biphenyl
CHB Cyclohexylbenzene
46-DiMeDBT 46-Dimetyl dibenzothiophene
46-DiMeBP 46-Dimetyl biphenyl
46-DiMeCHB 46-Dimetyl cyclohexylbenzene
46-DiMeTHDBT 46-Dimetyl 1,2,3,4,4a,9b tetrahydrodibenzothiophene
46-DiMeHHDBT 46-Dimetyl 1,2,3,4,4a,9b hexahydrodibenzothiophene
PNA Poly-nuclear aromatics
A Aromatics
N Naphtenes



Acknowledgement.

The authors acknowledge the contributions of: Arturo Deloiza (IMP) for elaborating the
computer programmes used to fit the kinetic model parameters, Susana Rodriguez (IMP) for
gathering relevant information and preparing the electronic version of the manuscript, and
Monica Perez (IMP) for performing test runs of the process simulation.

References.

1. G. F. Froment, G.A. Depauw, V. Vanrysselberghe, Kinetics of the catalytic removal of the
sulphur components from the light cycle oil of a catalytic cracking unit, Hydrotreatment and
Hydrocracking of oil fractions, G.F. Froment, B. Delmon and P. Grange, editors, 1997
Elsevier Science B.V.
2. D. H. Broderik and B.C. Gates Hydrogenolysis and hydrogenation of dibenzothiophene
catalyzed by sulfided CoO-MoO3/-Al2O3: The reaction kinetics.
3. Mariajose Castellanos, Estudio de la cintica de las reacciones de hidrodesulfuracin de
diesel, anlisis de modelos termodinmicos simulacin y optimizacin del proceso, BSc
Thesis on Chem. Eng. Faculty of Chemistry UNAM, Mxico, 1999.
4. J.L. Cano, F. Murrieta et al. Diesel sulphur reduction, report of the TCM-2 research project,
IMP Research Programme on Maya Crude Treatment.
5. S. Mayo, E. Brevoord, L.Gerritsen, and F.Plantenga, Process ultra-low sulfur diesel,
Hydrocarbon Processing, February 2001, 84A-84F.
6. W. Shiflett, The drive to lower and lower sulfur: Criterions new catalysts help refineries
tackle sulfur, NPRA 2001 Annual Meeting, March 18-20, 2001 New Orleans, LA