Академический Документы
Профессиональный Документы
Культура Документы
REACTOR ENGINEERING
Volume 9 2011 Article A4
Comparison of Lumping Approaches to
Predict the Product Yield in a Dual Bed
VGO Hydrocracker
Sepehr Sadighi
Arshad Ahmad
Mansoor Shirvani
v
=
= (15)
) ( ). (
0
i i C m
j j
v = (16)
=
=
=
G
F j
j
G
F j
j
j
m
m
i
.
) (
1
0
0
(17)
0
) ( ). (
f
l l j
j
m
N N C
Y
v
= (18)
In above equations,
j
is density of lumps (Table 3); i is the number of cells
which ranges from 1 to 200; C is the mass concentration of lumps; q is the
effectiveness factor; ' c is the catalyst volume fraction; ) (i V
cat
is the volume of
hydrocracking catalyst in each cell;
b
V is the volume of the bed; N
l
is the number
of cells (200);
0
f
m is the mass flow rate of fresh feed, and
j
Y is the yield of each
lump in the product stream leaving the reactor. The effectiveness factor for
cylindrical catalyst in trickle bed regime and the bed void fraction are 0.7 (Mills
& Dudukovic) and 0.35, respectively.
10 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
For parameter estimation, sum of squared error, SQE , as given below, is
minimized:
2
1
) (
pred
jn
t
N
n
meas
jn
G
F j
Y Y SQE =
=
=
(19)
In Eq.19,
t
N ,
meas
jn
Y and
pred
jn
Y are the number of test runs, the measured and the
predicted yields, respectively.
For the combined bed model, equations 1 to 7 and 13 to 18 should be
solved simultaneously by applying the following boundary conditions:
f
f
m
v
0
) 0 ( = ;
f f
C = ) 0 ( ; 0 ) 0 ( =
D
C ; 0 ) 0 ( =
N
C ; 0 ) 0 ( =
g
C (20)
But, for the dual bed model, equations 1 to 7 and equations 12 to 18
should be solved simultaneously by applying the following boundary conditions:
) 0 (
) 0 (
0
v
=
=
N
f j
jh
m
(21)
=
+ +
=
N
f j
jh N
Nh
N
f j
jh D
Dh
N
f j
jh f
fh
m
m
m
m
m
m
0
0
0
0
0
0
1
) 0 (
(22)
) 0 (
) 0 (
0
v
fh
f
m
C = ;
) 0 (
) 0 (
0
v
Dh
D
m
C = ;
) 0 (
) 0 (
0
v
Nh
N
m
C = ; 0 ) 0 ( =
g
C (23)
To develop the model, Aspen Custom Modeler (ACM) programming
environment (AspenTech, 2004) is used. Then Eq.19 is minimized by sequencing
NL2Sol and Nelder-Mead algorithm which are both existed in Aspen Custom
Modeler software. NL2Sol algorithm is a variation on Newton's method in which
a part of the Hessian matrix is computed exactly and a part of that is
approximated by a secant (quasi-Newton) updating method. To promote
convergence from a poor initial point, a trust-region is used along with a choice of
model Hessian. Hence, the approximate region is found with NL2Sol; then to fine
tune the parameters; Nelder-Mead method is used.
To evaluate the estimated kinetic parameters, absolute average deviation
of predictions ( % AAD ) is calculated by using the following expression.
11 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
100
) (
%
1
2
2
=
=
=
t
t
N
n meas
jn
pred
jn
meas
jn
G
F j
N
Y
Y Y
AAD (24)
Moreover, the goodness of fitting of developed models is checked with
analysis of variance (ANOVA) using Fischer test with 99% probability.
4. Results and discussions
At first for the combined bed model, twelve kinetic parameters, frequency factors
and activation energies were estimated by using the experimental data. After
estimating parameters and predicting yields, the AAD% of the model was 8.28%
in comparison to the measured data.
In Table 6, kinetic constants and the rate order of reactions at the mean
operating temperature (375
0
C) to the highest one (k
DN
or distillate to naphtha) are
presented. It is found that for the combined bed model, the rate orders of
converting VGO to naphtha (k
FN
) and distillate to gas (k
DG
) are significantly
lower than the highest value (k
DN
). It means that these reactions can be ignored.
After eliminating the low-reaction-rate pathways and re-estimating the parameters
(Table 7), the AAD% of the reduced model was found to be 8.23%. It can be
concluded that the model reduction can improve the accuracy of the yield
prediction which is the similar scenario with respect to the previous works
(Sadighi et al., 2010a; Sadighi et al., 2010b; Sadighi et al., 2010c).
From Table 7, it is obvious that apparent activation energies of VGO to
middle distillate and gas are 14.97 kcal/mol and 7.32 kcal/mol, respectively. The
reported ones by Aboul-Ghiet (1989) were about 13-17.5 kcal/mol, and 18-19
kcal/mol, respectively. It is thought that the lower estimated activation energy for
hydrocracking of feed to gas in this work is due to the higher hydrocracking
activity of zeolite type catalyst. Additionally, the estimated activation energy
(Botchwey et al., 2004) for formation of naphtha from middle distillate in the low
severity temperature regime (340 to 370
0
C) was about to 8.8 kcal/mol. Moreover
the reported value for hydrocracking of kerosene to heavy naphtha in an industrial
hydrocracking process charged with amorphous catalyst was 7.43 kcal/mol
(Sadighi et al., 2010 a). It can be concluded that the estimated activation energy of
distillate to naphtha in this work (7.11 kcal/mol) is not far from the others.
The simplified reaction-path network for the four-lump combined bed
model is shown in Figure 4.
12 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
Table 6. Kinetic parameters for complete network of combined bed model
Frequency Factor
(m
3
.hr
-1
.m
3
total cat
-1
)
Activation Energy
(kcal/mol)
Rate order
k
0FD
3.49E+05 E
FD
15.80 k
FD
0.61
k
0FN
2.44E+07 E
FN
69.30 k
FN
3.79E-17
k
0FG
3.34E-01 E
FG
2.10 k
FG
0.024
k
0DN
900.84 E
DN
7.49 k
DN
1
k
0DG
2.48E+07 E
DG
34.07 k
DG
2.98E-05
k
0NG
1.10E+00 E
NG
0 k
NG
0.41
Table 7. Kinetic parameters for reduced network of combined bed model
Frequency Factor
(m
3
.hr
-1
.m
3
total cat
-1
)
Activation Energy
(kcal/mol)
Rate order
k
0FD
1.82E+05 E
FD
14.97 k
FD
0.62
k
0FN
- E
FN
- k
FN
-
k
0FG
2.29E+01 E
FG
7.32 k
FG
0.03
k
0DN
660.20 E
DN
7.11 k
DN
1
k
0DG
- E
DG
- k
DG
-
k
0NG
8.33E-01 E
NG
- k
NG
0.32
Figure 4. The reduced 4-lump kinetic network for the combined bed model
After following again the described strategy for the dual bed model, it was
found that the AAD% of the complete and reduced approaches were 6.77% and
5.87%, respectively. The apparent kinetic constants of those are presented in
Tables 8 and 9, respectively. Also, the reduced kinetic network of this strategy is
depicted in Figure 5.
13 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
Table 8. Kinetic parameters for the complete network of the dual bed model
Frequency Factor
(m
3
.hr
-1
.m
3
HCR cat
-1
)
Activation Energy
(kcal/mol)
Rate order
k
0FD
6.72E+07 E
FD
23.37 k
FD
0.57
k
0FN
1.04E+08 E
FN
24.63 k
FN
0.33
k
0FG
0 E
FG
13.08 k
FG
0
k
0DN
0 E
DN
2.14 k
DN
0
k
0DG
6.90E+03 E
DG
28.15 k
DG
1.43E-06
k
0NG
1.55E+00 E
NG
0 k
NG
1
Table 9. Kinetic parameters for the reduced network of the dual bed model
Frequency Factor
(m
3
.hr
-1
.m
3
HCR cat
-1
)
Activation Energy
(kcal/mol)
Rate order
k
0FD
1.32E+07 E
FD
21.25
k
FD
0.59
k
0FN
1.42E+10 E
FN
31.02
k
FN
0.32
k
0FG
- E
FG
- k
FG
-
k
0DN
- E
DN
- k
DN
-
k
0DG
- E
DG
- k
DG
-
k
0NG
1.514 E
NG
- k
NG
1
Figure 5. The reduced 4-lump kinetic network for the dual bed model
It should be noted that the kinetic constants and the kinetic network of the
combined bed model (Table 7 and Figure 4) show the performance of
hydrotreating and hydrocracking sections together. But the related ones for the
14 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
dual bed model (Table 9 and Figure 5) only demonstrate the performance of the
hydrocracking catalyst.
Therefore, it can be understood that in comparison to the combined bed
approach, the dual bed one can predict the yield of hydrocracking reactor with
lower AAD% about 2.36%. The brilliant point in this approach which can be
understood from Table 9 is its requirement to only five kinetic parameters
(frequency factors and activations energies) to predict twelve sets of test runs
(three levels of LHSV and four levels of temperature). But, this approach needs
complete information from hydrogen consumption as well as sulfur, nitrogen and
aromatic contents of the feed and product. In contrast, combined bed model needs
seven parameters to predict the yield of products with AAD % of 8.23. But it is a
simplex in which no hydrotrating or hydrogen consumption data is needed.
In the previous works (Sadighi et al., 2010a), it was reported that the
activation energies for both light and heavy naphtha to gas were about to 9
kcal/mol for a dual-functional amorphous hydrocracking-hydrotreating catalyst.
But, from Tables 7 and 9, it can be concluded that in this range of operating
temperature, the activation energy for hydrocarcking of naptha to gas is
independent to temperature (E
NG
=0). The reason for this phenomenon is supposed
to be higher ability of zeolite type catalysts for hyrocracking (Shimada et al.,
1997). So, gas formation from naphtha may be influenced by the nature of
catalyst, and it is independent to temperature within the operating range.
The AAD percentages for all lumps are presented in Table 10. It can be
concluded that the predictions of the combined and dual bed models are close
together and they are acceptable for all products except to the case of gas for the
combined bed model. Additionally, in the Table 11, the ANOVA of both
strategies has been presented. The positive point in this table is acceptable
difference between the F-critical and the F-test of the proposed models.
Table 10. The AAD% for the different dual bed lumping models
Lump
Combined bed
(Completed)
Combined bed
(Reduced)
Dual bed
(Completed)
Dual bed
(Reduced)
Gas 21.04 20.7 11.00 9.26
Naphtha 4.92 5.34 8.29 7.04
Distillate 5.33 5.11 4.87 4.34
Un.VGO 1.83 1.76 2.90 2.85
Ave. 8.28 8.23 6.77 5.87
15 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
Table 11. ANOVA results of the developed models
Combined bed
(complete)
Combined bed
(reduced)
Dual bed
(complete)
Dual bed
(reduced)
DF* of regression 11 6 11 4
DF* of residual 36 41 36 43
R
2
-adjusted (%) 99.944 99.974 99.916 99.976
F-value 1789.69 3865.29 1189.87 4193.26
F-critical (1%) 2.79 3.28 2.79 3.79
*DF is degree of freedom
Figures 6 to 9 show comparisons between the measured yields and the
predicted ones. To evaluate the accuracy of the prediction, the corresponding
deviation plot is also presented. The deviation error reported in these figures was
calculated as follows:
100
Pr
%
=
yield Measured
yield edicted yield Measured
Error (25)
As it was resulted from Table 10, close mappings between the measured
and predicted yields by using both approaches can be understood. Moreover, it
can be found that deviations are acceptably distributed evenly around the zero; but
in the Figure 6, more deviation for the gas lump can be found for the combined
bed model.
Additionally, the deviation plot for the predicted gas by the combined bed
model in the Figure 6 demonstrates that in most of temperatures and LHSVs, the
dual bed model predicts lower yield for the gas lump in comparison to the
measured data. It is supposed that this deviation is because of disregarding the
reactions which can produce gas in the hydrotreating layer. Conversely, for most
of temperatures and LHSVs, the combined bed model predicts higher values for
the yield of gas. It is supposed that the main reason for this deviation is the
consideration of gas producing reactions in all points along the catalytic bed.
Because of lower ability of hydrotreating catalyst to produce gas than that of
hydrocracking one, the model predicts more gas in comparison to the measured
values. Therefore the yield of gas can be laid between the values predicted by the
combined and dual bed models.
16 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
Figure 6. Predicted yields (, &), measured yields (, & ) and
deviation plots for the prediction of gas lump
17 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
Figure 7. Predicted yields (, &), measured yields (, & ) and
deviation plots for the prediction of naphtha lump
18 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
Figure 8. Predicted yields (, &), measured yields (, & ) and
deviation plots for the prediction of distillate lump
19 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
Figure 9. Predicted yields (, &), measured yields (, & ) and
deviation plots for the prediction of VGO lump
5. Conclusions
It was demonstrated that the product yields of a pilot scale VGO-hydrocracker
could be predicted with the AAD% of 5.87% by using a four-lump rigorous model
approach, called the dual bed model. This model could predict the yield of gas,
naphtha, diesel and residue with the AAD% of 9.26%, 7.04%, 4.34% and 2.85%,
respectively. These deviations for 12 test runs (48 observations) in three levels of
LHSV and four levels of temperature can be satisfying. But to develop such a
20 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
model, the hydrogen consumption of the process was needed. Moreover, sulfur,
nitrogen and aromatic content of the feed and product should be analyzed. In
contrast, there was a simpler four-lump approach, called the combined bed model
which was capable of predicting the yield of products with the AAD% of 8.23%.
This model could predict the yield of gas, naphtha, diesel and residue with the
AAD% of 20.7%, 5.34%, 5.11% and 1.76%, respectively. The enormous deviation
for the gas lump can be the disadvantage of this model. But, the advantage of this
model over the dual bed model was its simplicity because it only required product
yields to tune the model parameters.
6. Nomenclature
6.a Notations
AAD Absolute Average Deviation, %
C
Mass concentration, kg/m
3
D Distillate
E Apparent activation energy, kcal/mol
G
Gas
k
Reaction rate constant, m
3
.hr
-1
.m
3
cat
-1
0
k
Frequency factor, m
3
.hr
-1
.m
3
cat
-1
LHSV
Liquid Hourly Space Velocity, hr
-1
0
m Mass flow rate, kg/hr
Mw
Molecular weight
S
Mw
Molecular weight of sulfur (32)
N
Mw
Molecular weight of nitrogen (14)
N
Naphtha
l
N
Number of cells (200)
t
N
Number of experiments
R
Ideal gas constant, 1.987 kcal.kmol
-1
.K
-1
j
R
Reaction rate of lump j, kg.hr
-1
.m
3
cat
-1
21 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
T Temperature, K or R
b
V
Total Volume of bed, m
3
cat
V
Volume of catalyst per cell, m
3
VGO
Vacuum Gas Oil
X Mass fraction of lumps
Y
Yield of products
6.b Greek letters
o Consumed mass of hydrogen per mass of VGO
12 22 11 2 1 0
, , , , , | | | | | | Coefficient values for the hydrogen consumption
' c Catalyst void fraction
q
Effectiveness factor
v Volume flow rate, m
3
/hr
Density, kg/m
3
6.c Subscripts
Af
Aromatic in feed
Ap
Aromatic in product
' Dj
Distillate to lighter lumps
Dh Diesel in the output stream of hydrotrating bed
fh
VGO feed or residue in the output stream of hydrotrating bed
Fj
Feed to lighter lumps
i Cell number
j
Distillate, naphtha and gas lumps
' j
naphtha and gas lumps
2
H Hydrogen
S H
2
Hydrogen sulfide
n Number of experiments
Nc Nitrogen lumped component
22 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
Nf
Nitrogen in feed
Nh Naphtha in the output stream of hydrotrating bed
3
NH Ammonia
Sc Sulfur lumped component
Sf
Sulfur in feed
References
Aboul-Gheit K., "Hydrocracking of vacuum gas oil (VGO) for fuels production",
Erdol Erdgas Kohle, 1989, 105, 1278.
Almeida R. M. and Guirardello R., "Hydroconversion kinetics of Marlim vacuum
residue", Catalysis Today, 2005, 109, 104.
Alvarez A. and Ancheyta J., "Modeling residue hydroprocessing in a multi-fixed-
bed reactor system", Applied catalysis A: General, 2005, 351, 148.
Ancheyta J., Sanchez S. and Rodriguez M. A., "Kinetic modeling of
hydrocracking of heavy oil fractions: A review", Catalysis Today, 2005,
109, 76-92.
Ancheyta-Juarez J., Lopez-Isunza F. and Aguilar-Rodriguez E., "5-Lump kinetic
model for gas oil catalytic cracking", Applied Catalysis A: General, 1999,
177, 227.
Aoyagi K., McCaffrey W. C. and Gray M. R., "Kinetics of hydrocracking and
hydrotreating of coker and oilsands gas oils", Petroleum Science
Technolology, 2003, 21, 997.
Basak K., Sau M., Manna U. and Verma R.P., "Industrial hydrocracker model
based on novel continuum lumping approach for optimization in
petroleum refinery", Catalysis Today, 2004, 98, 253.
Balasubramanian P. and Pushpavanam S., "Model discrimination in
hydrocracking of vacuum gas oil using discrete lumped kinetics", Fuel,
2008, 87, 1660.
23 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011
Botchwey C., Dalai A.K. and Adjaye J., "Kinetics of Bitumen Derived Gas Oil
Upgrading Using a Commercial NiMo/Al
2
O
3
Catalyst", The Canadian
Journal of Chemical Engineering, 2004, 82, 478.
Callejas M. A. and Martinez M. T., "Hydrocracking of a Maya residue, kinetics
and product yield distributions", Industrial Engineering Chemical
Research, 1999, 38, 98.
Chen C., Yang B., Yuan J., Wang Z. and Wang L., "Establishment and solution of
eight-lump kinetic model for FCC gasoline secondary reaction using
particle swarm optimization", Fuel, 2007, 86, 2325.
Elizalde I., Rodrguez M.A. and Ancheyta J., "Application of continuous kinetic
lumping modeling to moderate hydrocracking of heavy oil", Applied
Catalysis A: General, 2009, 365, 237.
Krambeck F.J., "Kinetics and thermodynamics lumping of multicomponent
mixtures", Elsevier, 1991, Amsterdam, 111.
.
Meng X., Xu C., Gao J., Li L., "Catalytic pyrolysis of heavy oils: 8-lump kinetic
model", Applied Catalysis A: General, 2006, 301, 32.
Mills P.L., Dudukovic M.P., "A Dual-Series Solution for the Effectiveness Factor
of Partially Wetted Catalysts in Trickle-Bed Reactors", Ind. Eng. Chem.
Fund., 1979, 18, 2.
Mohanty S., Saraf D. N. and Kunzro D., "Modeling of a hydrocracking reactor",
Fuel Processing Technology, 1991, 29, 1-17.
Sadighi S., Arshad A. and Mohaddecy S. R., "6-Lump Kinetic Model for a
Commercial Vacuum Gas Oil Hydrocracker", International Journal of
Chemical Reactor Engineering, 2010a, 8, A1.
Sadighi S., Arshad A. and Irandoukht A., "Modeling a Pilot Fixed-bed
Hydrocracking Reactor via a Kinetic Base and Neuro-Fuzzy Method",
Journal of Chemical Engineering Japan, 2010b, 43 (2), 174.
Sadighi S., Arshad A. and Irandoukht A., "Kinetic Study on a Commercial
Amorphous Hydrocracking Catalyst by Weighted Lumping Strategy",
International Journal of Chemical Reactor Engineering, 2010c, 8, A60.
24 International Journal of Chemical Reactor Engineering Vol. 9 [2011], Article A4
http://www.bepress.com/ijcre/vol9/A4
Sadighi S., Arshad A. and Rashidzadeh M., "4-Lump kinetic model for vacuum
gas oil hydrocracker involving hydrogen consumption", Korean Journal of
Chemical Engineering, 2010d, 27 (4), 1099.
Shimada H., Yoshitomi S., Sato T., Matsubayashi N., Imamura M., Yoshimura Y.
and Nishijima A., "Dual-functional Ni-Mo sulfide catalysts on zeolite-
alumina supports for hydrotreating and hydrocracking of heavy oils",
Studies in Surface Science and Catalysis Hydrotreatment and
Hydrocracking of Oil Fractions, Proceedings of the 1st International
Symposium/6th European Workshop, 1997, 106, 115-128.
Singh J., Kumar M., Saxena A. K. and Kumar S., "Reaction pathways and product
yields in mild thermal cracking of vacuum residues: A multi-lump kinetic
model", 2005, Chemical Engineering Journal, 108, 239.
Valavarasu G., Bhaskar M. and Sairam B., "A Four Lump Kinetic Model for the
Simulation of the Hydrocracking Process", Petroleum Science
Technology, 2005, 23, 1323.
Yui S. M. and Sanford E. C., "Mild hydrocracking of bitumen-derived coker and
hydrocracker heavy gas oils: kinetics, product yields and product
properties, Industrial Chemical Research, 1989, 28, 319.
25 Sadighi et al.: Lumping Model for a Dual Bed VGO Hydrocracker
Published by The Berkeley Electronic Press, 2011