Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Hydrocracking is a petroleum refining process where cracking occurs simultaneously with hydrogenation to
convert heavy petroleum feedstocks into desired lighter products. In this study, multi-objective optimization
of an industrial hydrocracking unit has been performed for the first time, to find scope for further improvements
and to provide a range of optimal solutions. The reactor model was simulated based on a discrete lumped
model approach to kinetic modeling. The kinetic and product distribution parameters were fine-tuned using
available industrial data. The real-coded elitist nondominated sorting genetic algorithm was used to carry out
the multi-objective optimization study. The Pareto-optimal solutions for the hydrocracker unit are presented,
and their significant features are discussed.
Introduction
Hydrocracking is a catalytic cracking process for the conversion of complex feedstock like heavy vacuum gas oils into more
valuable lower boiling products, such as gasoline, kerosene, and
diesel, in hydrogen-rich atmosphere at elevated temperature and
pressure.1 The reactions that occur in hydrocracking are cracking
of the high molecular weight compounds giving rise to low
molecular weight compounds and hydrogenation of olefins,
aromatic rings, and sulfur, nitrogen, and oxygen compounds to
some extent. The cracking reaction provides the olefins for
hydrogenation, while the hydrogenation liberates the heat
required for cracking. The cracking reactions are slightly
endothermic while the hydrogenation reactions are highly
exothermic, making the overall hydrocracking process highly
exothermic in nature. Hydrocracking is the best source of lowsulfur and low-aromatics diesel fuel as well as high-smokepoint jet fuel. The higher hydrogen content of the products
provides better combustion characteristics, making hydrocracking technology necessary to meet the current and future fuel
specifications. Moreover, hydrocracking has the capability of
processing a wide range of feedstocks of different characteristics
to produce a broad range of products.
Even though considerable amount of research has been
reported on reaction mechanism, catalysts, kinetic modeling,
hydrodynamics, and reactor simulation of hydrocrackers, hardly
any work on the optimization, either single or multiple objectives, of hydrocracking units has been reported in the open
literature. Due to increasing worldwide demand for jet fuel and
diesel and the easy availability of H2 from the catalytic
reforming unit within the plant premises, the hydroprocessing
industry has considerable growth potential. Although hydrocracking technology is quite advanced, new developments in
catalyst and changes in reactor design influence the operation
of the entire unit. This requires attention to optimize catalyst
design, minimize utility consumption, and to have the best
resource utilization. Multiple objectives are relevant in such
cases. In multi-objective optimization,2-4 a solution that is the
best (global optimum) with respect to all objectives may not
exist. Instead, an entire set of optimal solutions that are equally
* Corresponding author. Tel.: (65) 6516 2187. Fax: (65) 6779 1936.
E-mail: chegpr@nus.edu.sg.
1356
ki ) Kik where k )
( )
FHVGO
Ae-E/RT
Fcatalyst
(2)
(3)
(4)
(5)
yij )
(Tbi - 2.5)
[(Tbj - 20) - 2.5]
(7)
dT
i MiCpi dW ) j (-H)RjkjCj
i ) 1 to N + 1, j ) 14 to N (8)
1358
TBP (C)
FPC (kg/m3)
Watson K (-)
xHVGO (-)
xHD (-)
xHN (-)
xLN (-)
yHN (-)
yLN (-)
1
8
9
10
11
12
13
14
15
16
17
18
36
37
38
39
40
41
49
50
51
52
53
54
55
56
57
58
2.50
64.98
74.60
84.33
94.62
103.98
114.79
124.91
134.44
144.66
154.28
163.98
344.93
354.78
365.11
375.33
385.25
395.25
475.03
485.02
494.99
504.97
514.95
524.76
535.06
545.75
553.59
565.16
583.00
689.83
701.12
712.33
721.87
730.67
740.95
749.40
756.88
764.21
769.62
775.32
858.67
861.94
866.48
872.26
877.61
882.49
915.61
919.02
922.20
925.15
927.84
930.18
932.25
935.74
938.47
942.79
13.360
12.285
12.201
12.120
12.073
12.028
11.974
11.941
11.916
11.900
11.906
11.907
12.067
12.085
12.087
12.071
12.058
12.052
12.061
12.069
12.080
12.094
12.110
12.129
12.154
12.162
12.165
12.166
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00593
0.00689
0.00842
0.01056
0.01331
0.01698
0.06311
0.06525
0.06582
0.06513
0.06390
0.05874
0.05326
0.06563
0.03522
0.03452
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.1175
0.1028
0.0957
0.0931
0.0874
0.0772
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.0052
0.0354
0.1519
0.1893
0.1862
0.1586
0.1314
0.0915
0.0312
0.0194
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.1580
0.1287
0.1292
0.0626
0.0421
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.0000
0.0862
0.3899
0.8498
0.9129
1.00
0.9690
0.8647
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.0000
0.9138
0.6101
0.1502
0.0871
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Table 2. Experimental Data for Feed and Products Obtained from the Refinerya
feed
IBP
5%
10%
20%
30%
40%
50%
60%
70%
80%
90%
95%
FBP
F (15 C)
N2 (ppm)
S (wt %)
LN
HN
KS
LD
HD
UCO
310.00
363.50
383.50
407.50
424.50
HTEffluent
34.70
42.50
46.80
52.00
56.50
101.10
107.30
109.00
111.10
113.60
163.90
179.70
183.50
191.00
198.10
199.20
206.60
219.00
305.20
321.80
326.00
227.80
333.00
456.50
65.00
119.10
216.00
244.60
342.40
487.00
504.00
526.50
543.00
578.00
0.933
1413
2.656
73.20
126.60
237.80
258.00
356.00
85.10
91.90
101.40
0.68
138.00
142.70
160.50
0.75
263.70
272.50
277.20
0.81
272.20
278.00
280.00
0.82
373.40
378.60
380.60
0.83
315.50
361.00
380.50
403.00
417.50
429.00
440.50
453.00
467.00
483.50
507.00
526.50
566.00
0.84
7.20
0.03
0.883
64
0.092
0.99
0.03
Note that the percent in column 1 refers to wt % for feed and UCO and to vol % for LN, HN, KS, LD, and HD. Boiling points in the table are in degrees
Celcius.
1360
HC
bed-1
m3
Vcat,
Mcat, kg
22.70
19 997
bed-2
25.00
23 586
bed-3A
5.6
4469
bed-3B
20.3
19 975
bed-4A
7.2
5738
bed-4B
18.2
17 800
a Two types of catalysts are present in bed 3 and bed 4, hence given as
3A and 3B and 4A and 4B.
operating data
HT
FHVGO, kL/h
Pinlet, atm
Tinlet, K
Tout, K
QHT, QHC, tons/h
LHSV, 1/h
RG1, RG2, kg/h
Tinlet1, K
Tout1, K
Tinlet2, K
Tout2, K
Tinlet3, K
Tout3, K
Tinlet4, K
Tout4, K
a
176.60
655.99
689.72
a
a
HC
170.54
672.25
680.04
a
1.43
a
672.25
680.89
668.94
680.27
670.87
679.04
672.25
680.04
parameter
reported value13
tuned value
-HHcon, kcal/kmol
A1, vol feed/(vol cat h)
A2, vol feed/(vol cat h)
A3, vol feed/(vol cat h)
A4, vol feed/(vol cat h)
D1
D2
D3
D4
C
10 000
1.0 107
1.0 107
1.0 107
1.0 107
0.494
5.2 10-3
-2.185 10-5
3.12 10-8
0.35
0.01
0.70
6310
9.797 106
9.952 106
9.843 106
9.844 106
0.5335
5.616 10-3
-2.425 10-5
3.37 10-8
0.70
0.02
0.64
recycle oil temperature, recycle gas flow rate, and quench flow
rates. The computation time for the HC unit simulation on a
2.4 GHz Pentium IV computer with 512 MB of SDRAM was
about 4 s. The predicted values of bed exit temperatures and
product flow rates were compared with the industrial values.
The HC model parameters were fine-tuned using the consistent industrial data. The twelve parameters selected for finetuning the model were the amount of heat released per unit
consumption of hydrogen, -HHcon in kcal/kmol, the frequency
factor in the reaction rate constant corresponding to the four
catalyst beds in the HC reactor (A1-A4 in eq 2), parameters for
relative reaction rate kinetics (D1-D4 in eq 3), and the product
distribution parameters (C, , and B) in eqs 4 and 5. The finetuning of the 12 parameters was based on minimization of the
root of sum of the squared error between the model predictions
and the measured values in the plant. The overall objective
consisted of the minimization of error in product flow rates,
the total of all product flow rates, and the exit temperature from
the four catalyst beds.
(x ( ) ( ) ( ) )
Objective ) Minimize
8
w1
k)1
1-
fk,S
fk,I
+ w2 1 -
FT,S
FT,I
i)1
+ w3
1-
Touti,S
Touti,I
(9)
Figure 3. Comparison of model predicted and industrial values of pseudocomponent flow rates.
day 1
day 2a
day 3
day 4
day 5
LE
LN
HN
KS
LD
HD
UCOT
UCOP
H2,makeup
13.67
5.35
12.88
-1.58
1.24
-30.61
12.03
-2.64
19.98
-13.16 (5.43)
-4.24 (-34.46)
6.08 (-45.24)
-1.38 (-3.45)
2.90 (-1.69)
-18.93 (32.39)
27.42 (36.69)
11.14 (23.67)
11.0 (10.9)
5.75
3.71
10.86
-1.59
1.61
-28.23
15.16
2.27
16.45
4.04
-0.18
11.94
-0.77
2.28
-28.12
13.06
8.87
14.73
17.22
6.08
14.62
-0.34
0.79
-31.22
5.52
-0.23
21.69
-5%
5%
FRO
-5%
MRG
5%
-5%
2.56
1.68
1.19
-0.03
-0.21
-1.14
-2.81
-2.81
-2.81
0.67
TRG
5%
-2%
TRO
2%
-2%
-2.36 -30.76
36.39 -35.16
-1.55 -21.08
23.11 -24.53
-1.11 -15.81
15.63 -18.76
0.03 -1.19 -1.56 -2.53
0.20
1.17 -3.86
0.12
1.09
14.85 -14.84
15.77
2.73
46.32 -31.16
52.87
2.73
46.32 -31.16
52.87
2.73
46.32 -31.16
52.87
-0.61 -8.69
9.01 -10.66
0.00
0.03
0.00
0.04 -0.04 -0.60
-0.02 -0.69
0.70 -10.07
0.54 -0.74
8.16 -11.33
-11.29
15.79 -0.03
8.05 -10.69 -1.42
1.42 -2.20 -1.80
0.57 -0.90 -0.30
0.02 -0.68
0.66
11.20
1.31
0.78 -0.74 -11.51
1.69
0.15 -0.14 -2.39
0.29 -0.57
0.56 -2.15
-7.54
12.77
9.26 -13.27
1.62 -2.80
2.22 -2.64
Q1
Q2
Q3
2%
-5%
5%
-5%
5%
-5%
5%
26.84
17.24
11.84
-0.83
-2.57
-11.06
-24.32
-24.32
-24.32
6.80
8.01
5.23
3.68
-0.14
-0.69
-3.51
-8.41
-8.41
-8.40
2.07
-7.58
-5.03
-3.62
0.00
0.55
3.47
9.09
9.09
9.09
-2.01
3.82
2.51
1.78
-0.05
-0.31
-1.69
-4.14
-4.14
-4.14
1.00
-3.67
-2.43
-1.74
0.02
0.28
1.66
4.24
4.24
4.24
-0.97
1.44
0.95
0.67
-0.02
-0.12
-0.64
-1.60
-1.60
-1.60
0.37
-1.42
-0.94
-0.67
0.02
0.12
0.65
1.63
1.63
1.63
-0.37
0.41
6.25
0.07 -0.08
2.08 -2.15
0.04 -0.04
1.01 -1.02
0.01 -0.01
0.39 -0.39
-5.89 -2.03
2.20 -1.00
1.03 -0.39
0.39
7.13
2.36 -2.44
1.15 -1.16
0.44 -0.44
1.27
0.43 -0.47
0.21 -0.22
0.08 -0.08
1.68 -1.39
1.40 -1.30
1.30 -1.26
1.26
1362
objective functiona
decision variables
constraints
fixed variables
case 1
KS/
Tinlet e 680 K
Tout,1 e 686 K
Tout,2 e 685 K
Tout,3 e 684 K
Tout,4 e 684 K
0.5 e LHSV (1/h) e 2.0
85 e Xoverall (%) e 98
65 e Xpass (%) e 80
case 2
case 3
a The
reason.
max
min H/2,makeup
max HD/
min H/2,makeup
max HE/
min LE/
symbol indicates that the value is normalized with respect to simulated values for the industrial operation. b Value not given due to proprietary
estimate for 5 days of industrial operation is in the range 1122%. The hydrogen losses in the downstream separation (i.e.,
with Off-gas1 and Off-gas2) for any days simulation are fixed
at their industrial value. For example, for day 2, it is fixed at
4.2% of the H2,makeup. Some amount of hydrogen also gets lost
with other gases to flare from distillation column 2 (not shown
in Figure 1) and in the wash water separator. Hydrogen has a
high solubility in the hydrocarbon mixture at high pressure in
the reactor, and it transfers to the gas phase during fractionation
at low pressure in the distillation column. Hydrogen is also
soluble in the liquid used for wash water separation. Such losses
are not measured in industry and, hence, are not considered for
simulation. These contribute to the error in the estimate of
H2,makeup.
Sensitivity Analysis
To formulate the multi-objective optimization problem appropriately, sensitivity analysis was performed by varying one
variable at a time around a reference set of values and noting
its effect on several performance indicators of the HC unit. The
calibrated model was used for sensitivity analysis and was later
interfaced with real-coded NSGA-II40 for plantwide multiobjective optimization. The sensitivity analysis results are
summarized in Table 7. The increase in the feed flow rate
(FHVGO) increases the production of heavier products (KS, LD,
and HD) and unconverted oil (UCO) and decreases the production of lighter products (LE, LN, and HN), as a result of a
decrease in the operating temperature and residence time in the
reactor. The liquid hourly space velocity (LHSV) increases,
thereby decreasing the conversion per pass (Xpass) by 10%
and overall conversion (Xoverall) by 2%. The increase in FHVGO
has a counter effect on the inlet temperature (Tinlet) of the HC.
Due to increased LHSV and reduced Tinlet, and consequently
reduced conversion, H2 consumption in the HC reduces and the
need for H2,makeup decreases by 1%. For the hydrocracking of
any heavier pseudocomponent, the formation of lighter
pseudocomponents consumes more hydrogen than the formation
of heavier components as the C/H ratio for lighter hydrocarbons
is much less than that for heavier hydrocarbons.
The increase in the fraction of unconverted oil recycled (FRO)
to the reactor is equivalent to processing heavier feed in the
reactor. The heat of reaction for hydrocracking of heavier feed
is more in comparison to that for hydrocracking of lighter feed.
Though the inlet temperature (Tinlet) to the reactor is comparable
with the plant value (Table 7), a higher operating temperature
in the catalyst beds is expected because of the higher exothermic
heat of reaction of the heavier feed being processed in the
reactor. Hence, both Xpass and Xoverall increase due to the increase
in the recycle oil flow rate and the increase in bed temperatures.
The overall effect of the increase in LHSV, Xpass, and Xoverall
increases the requirement for hydrogen supply to the HC for
reaction as well as for quenching.
value/specification
number of generations
population size
encoding technique
crossover probability
distribution index for real-coded crossover
mutation probability
distribution index for real-coded mutation
seed for random number generator
crossover parameter in the SBX operator
variable bounds
selection strategy
200
50
real coding
0.85
10
0.05
20
0.857
10.0
rigid
tournament selection
The increase in the mass flow rate of the recycle gas (MRG),
RG2 in Figure 1, increases the amount of gas fed to the reactor/
m3 of liquid feed. This ultimately decreases Tinlet and the bed
temperatures, decreasing Xpass and Xoverall. The need for H2,makeup
goes high due to the increase in the gas flow rate in the process
loop. The yield of heavier products can be improved slightly
by increasing the recycle gas flow to the HC at the expense of
increased cost for recycle gas compression and an increased
hydrogen requirement in the process loop. The increase in Tinlet,
as a result of the increase in recycle gas temperature (TRG) or
the recycle oil temperature (TRO), increases the rates of all
hydrocracking reactions leading to increase in the production
of lighter products such as LE, LPG, LN, and HN but decreases
the production of heavier products such as KS, LD, and HD.
This ultimately increases the amount of hydrogen required for
chemical consumption and for temperature control in the reactor
beds by increasing the quench flow rates. The hydrocracking
reaction is relatively more sensitive to TRG than to TRO as given
by the increase in conversion to 9% and 7% for a 2% increase
in each of these temperatures.
Quenching requires a good amount of H2, and an increase of
the quench flow rate (Q1-Q3) has a countereffect on temperature. The increase in H2,makeup is expected due to the increased
quench flow rate, which also increases the overall gas flow rate
to the HC. The increased quench ultimately leads to reduced
reaction rates, lower yields of light products, and lower hydrogen
consumption for hydrocracking, which is directly related to
fractional yield of lighter products.
Multi-objective Optimization
The real life optimization problems faced in industries usually
deal with more than one competing objective.3 Traditionally,
solving such problems involves taking a weighted average of
all objectives and treating it as a single objective optimization
problem. However, the solution then depends on the chosen
weights, which in turn are subject to individual perception and
knowledge on the process. This is arbitrary, and a deficiency is
inherent in the method.4 The best way to solve and represent
the solution of a multi-objective optimization problem is through
Figure 4. Results for two-objective optimization: the maximization of kerosene and minimization of the makeup H2 flow rate. (a) Plot of Pareto optimal
solutions set. Plots of decision variables corresponding to Pareto optimal solution set in part a are shown in Figure 4b-i: (b) recycle gas flow to hydrocracker;
(c) fresh feed flow rate; (d) recycle gas temperature; (e) unconverted recycle oil fraction; (f) unconverted recycle oil temperature; (g) quench flow between
bed 1 and 2 in the HC; (h) quench flow between bed 2 and 3 in the HC; (i) quench flow between bed 3 and 4 in the HC. The industrial operating point is
shown by the filled square symbol.
1364
Figure 5. Plot of constraints corresponding to the optimal solution in Figure 4a: (a) inlet temperature to the HC; (b) outlet bed temperatures; (c) LHSV;
(d) conversion per pass; (e) overall conversion. The industrial operating point is shown by the filled square symbol.
Figure 6. Results for two-objective optimization: the maximization of heavy diesel and minimization of the makeup H2 flow rate. (a) Plot of Pareto optimal
solutions set. Plots of decision variables corresponding to Pareto optimal solution set in part a are shown in Figure 6b-i: (b) recycle gas flow to hydrocracker;
(c) fresh feed flow rate; (d) recycle gas temperature; (e) unconverted recycle oil fraction; (f) unconverted recycle oil temperature; (g) quench flow between
bed 1 and 2 in the HC; (h) quench flow between bed 2 and 3 in the HC; (i) quench flow between bed 3 and 4 in the HC. The industrial operating point is
shown by the filled square symbol.
1366
Figure 7. Plot of constraints corresponding to optimal solution in Figure 6a: (a) inlet temperature to the HC; (b) outlet bed temperatures; (c) LHSV; (d)
conversion per pass; (e) overall conversion. The industrial operating point is shown by the filled square symbol.
Figure 9. Plots of decision variables corresponding to Pareto optimal solution in Figure 8: (a) recycle gas flow to hydrocracker; (b) fresh feed flow rate;
(c) recycle gas temperature; (d) unconverted recycle oil fraction; (e) unconverted recycle oil temperature; (f) quench flow between bed 1 and 2 in the HC;
(g) quench flow between bed 2 and 3 in the HC; (h) quench flow between bed 3 and 4 in the HC. The industrial operating point is shown by the filled square
symbol.
the increase in TRG (Figure 4d) and TRO (Figure 4f) and of the
corresponding increase in the enthalpy of the feed with an
increase in the feed flow rate and MRG. More recycle gas at the
HC inlet (Figure 4b) is required to saturate the liquid feed as
the production of heavier product is favored at a higher degree
of saturation and lower temperature. As a result of the increase
in the operating inlet temperature and bed temperatures with
the increase in the flow rate of the feed, the cooling requirement
also increases. The intermediate quench rates (Figure 4g-i)
depend on the cooling requirement. The fraction of unconverted
oil recycled, FRO (Figure 4e), is found to have a direct relation
to Xoverall (Figure 5e). The value of Xpass (Figure 5d) is lower in
comparison to the plant operating value of 0.675 and remains
constant at 0.65 until FHVGO is close to its upper bound. At this
point, one of the ways to increase the production rate of KS/ is
to increase Xpass by cooling less between the beds, as indicated
1368
Figure 10. Plot of constraints corresponding to optimal solution in Figure 8: (a) inlet temperature to the HC; (b) outlet bed temperatures; (c) LHSV; (d)
conversion per pass; (e) overall conversion. The industrial operating point is shown by the filled square symbol.
Figure 11. Results for two-objective optimization: the maximization of heavy diesel and minimization of the makeup H2 flow rate. (a) Plot of Pareto
optimal solutions set. Plots of decision variables corresponding to Pareto optimal solution set in part a are shown in Figure 11b-i: (b) recycle gas flow to
hydrocracker; (c) recycle gas temperature; (d) unconverted recycle oil fraction; (e) unconverted recycle oil temperature; (f) quench flow between bed 1 and
2 in the reactor; (g) quench flow between bed 2 and 3 in the reactor; (h) quench flow between bed 3 and 4 in the reactor. The industrial operating point is
shown by the filled square symbol.
1370
Figure 12. Plot of constraints corresponding to optimal solution in Figure 11a: (a) inlet temperature to the HC; (b) outlet bed temperatures; (c) LHSV; (d)
overall conversion. The industrial operating point is shown by the filled square symbol.
1372