Petroleum Science and Technology

ISSN: 1091-6466 (Print) 1532-2459 (Online) Journal homepage: http://www.tandfonline.com/loi/lpet20

A new lumping kinetic model for fluid catalytic

cracking

Shiyuan Sun, Fandong Meng & Hongfei Yan

To cite this article: Shiyuan Sun, Fandong Meng & Hongfei Yan (2018): A new lumping
kinetic model for fluid catalytic cracking, Petroleum Science and Technology, DOI:
10.1080/10916466.2018.1519576

To link to this article: https://doi.org/10.1080/10916466.2018.1519576

Published online: 08 Oct 2018.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=lpet20
PETROLEUM SCIENCE AND TECHNOLOGY
https://doi.org/10.1080/10916466.2018.1519576

A new lumping kinetic model for fluid catalytic cracking

Shiyuan Sun, Fandong Meng, and Hongfei Yan
Department of reaction engineering, Sinopec Engineering (Group) Company LuoYang R&D Center of
Technology, LuoYang, China

ABSTRACT KEYWORDS
In the modeling of fluid catalytic cracking reaction, it is necessary to take catalytic cracking; kinetic
slurry as an independent lump due to the great difference between feed- model; MIP; product
stock properties and slurry properties. Based on the commercial data and prediction; slurry lump
the reaction mechanism of catalytic cracking, the 8 lumped kinetic model
including slurry lump and 7 lumped kinetic model excluding slurry lump
were developed combined with the characteristics of the MIP (maximizing
isoparaffins) reaction. By means of Fourth-order Runge-Kutta integral
method and two-swarm cooperative particle swarms optimization method,
the kinetic model parameters were estimated, respectively. The average
relative error between commercial data and calculated values, as well as
the results of product predictions, indicate that the eight-lumped kinetic
model has higher prediction accuracy, and the predicted results of slurry
yield are more in line with the actual industrial conditions.

1. Introduction
Catalytic cracking is one of the important heavy oil lightening processes and plays an important
role in the production of light oil products such as gasoline and diesel. The design and operation
of the catalytic cracking units have a major impact on the overall economic efficiency of the
refinery(Pinheiro et al. 2012). In the design and operation stage of the catalytic cracking unit, the
catalytic cracking reaction system process simulation model which is used to optimize the design
and operation of the reaction system, or improve the design and operation of the entire plant,
can significantly increase the economic efficiency of the refinery (Lopes et al. 2011).
For the simulation and optimization of the catalytic cracking reaction system, an accurate and
practical process calculation model is needed. The process calculation model has been widely
used in the process design and operation of catalytic cracking reaction system, such as the
ASPENTECH’s catalytic device reaction regeneration simulation optimization software Aspen
FCC (Khandeparker 2012). The catalytic cracking reaction process calculation models all adopt a
lumped reaction model, in which the feedstock and products are artificially divided into several
lumped components, and the changes in the flow rate and composition of these lumped compo-
nents are regarded as the changes in the flow rate and composition of feedstock and product. In
the current process simulation and optimization software and catalytic cracking kinetic model
(Ancheytaju
arez, L
opezisunza, and Aguilarrodr
ıguez 1997; Mehran, Habib, and Bahram 2010;
Gao et al. 2014; Xiong et al. 2015; Ouyang, Wang, and Qiao 2016), there is no slurry lump in the
lumped product division, meanwhile the feedstock is treated as slurry lump. For kinetic models
that do not subdivide the feedstock according to the distillation range and the structure group

CONTACT Shiyuan Sun sunsy.lpec@sinopec.com Sinopec Engineering (Group) Company LuoYang R&D Center of
Technology, LuoYang 471003, China.
ß 2018 Taylor & Francis Group, LLC
2 S. SUN ET AL.

Figure 1. MIP riser reactor is here.

composition, there is great difference between the properties of the feedstock and the slurry.
While the kinetic model is used to predict the product, in which the feedstock is regarded as
slurry, the prediction result of slurry yield will be poor and the error between the prediction and
the actual situation is large.
In addition, for some catalytic cracking processes, such as MIP process, due to the existence of
two reaction zones, the reaction time is long. If only cracking reaction is considered, and conden-
sation reaction is not taken into consideration, as in the previous lumped kinetic model, it can’t
really reflect the reaction law of the catalytic cracking reaction process. In this case, it is necessary
to consider the reaction of gasoline and diesel oil condensation to form slurry. In this paper, the
lumped kinetic model of catalytic cracking is established on the basis of the measured data of
maximizing isoparaffin process (MIP) (Jiang and Huang 2016) industrial units, in which the
slurry component is treated as a separate lumped model in the product lumped.
The MIP technology adopts a new reaction system and corresponding process conditions
(Zhu, Li, and Zheng 2017). The sketch of MIP reactors is shown in Figure 1. The reactor is div-
ided into two reaction zones connected by direct cascade with different diameters. The first reac-
tion zone is similar as the traditional FCCU, where the cracking reaction takes place with higher
reaction intensity. The second reaction zone has a larger diameter, in order to increase the reac-
tion time. Meanwhile, it used lower temperature to increase the hydrogen transfer and isomeriza-
tion reactions, which can improve the production of iso-paraffins.

2. Kinetic modeling
In this study, the lumped kinetic model including slurry component is named model 1, and the
model excluding slurry component is named model 2. Model 1 divides the feed oil into two
lumps of less than 500
C and more than 500
C according to the distillation range. The products
are divided into six lumped parts: dry gas, liquefied gas, gasoline, diesel oil, slurry and coke. In
this way, the raw materials and products are divided into eight lumped parts. Model 2 is 7
lumped kinetic model without slurry component.
 PETROLEUM SCIENCE AND TECHNOLOGY 3

Figure 2. The reaction network for 8-lump kinetic model of catalytic cracking is here.

Unlike conventional riser, MIP riser reactor has a longer reaction time. Therefore, the cracking
reaction and condensation reaction should be taken into account in the establishment of kinetic
model of MIP riser reactor. According to the catalytic cracking reaction mechanism of MIP pro-
cess, the model has been hypothesized as follows: (1) All reactions in the reaction network are
irreversible first-order. (2) The feedstock experience cracking reaction and condensation to form
all the products (3) As an intermediate product of catalytic cracking, diesel oil can be further
cracked to form dry gas, liquefied gas, gasoline, or condensed to form slurry and coke. (4)
Gasoline can be cracked to form dry gas, liquefied gas, or condensed to form coke. (5) Dry gas,
liquefied gas and coke are the final product of the reaction. (6) Two reaction zones have the
same reaction network, but have the different reaction temperatures and material concentrations
that cause the differences of reaction rate. Accordingly, the reaction networks of model 1 and
model 2 are established, as shown in Figures 2 and 3, respectively. Model 1 contains 21 reactions
and model 2 contains 17 reactions.
There are two reaction zones in MIP riser reactor, in which the reaction temperature in the
second zone is obviously lower than that in the first zone. Generally, the temperature difference
between the outlet of the two zones is above 15
C. Therefore, the two reaction zones are
regarded as isothermal, gas-phase and piston flow reactors at different temperatures respectively.
The diffusion within the particle is neglected, so the basic equation of the model can be derived
from the continuity equation and the reaction rate equation:

da P 1 1
¼ f ðAÞ  uðtc Þ  X   Ka (1)
dX RT ai SWH

where i is lumped component, j is the j – th reaction. f ðAÞ ¼ 1þKA1ðacatÞ indicates the effect of cata-
lyst activity. uðtc Þ ¼ ebtc indicates the effect of coke formation on the activity of the catalyst.
And it is considered that the coke formation rate of catalyst is only a function of the residence
time of catalyst.
4 S. SUN ET AL.

Figure 3. The reaction network for 7-lump kinetic model of catalytic cracking is here.

3. Results and discussion

3.1. Solution of the model
Estimation of the kinetic parameters was performed by minimizing the error objective function,
which is the sum of square errors between the calculated and actual values:

X
nc exp X
nexp
U¼ ðaij  ^a ij Þ2 (2)
j¼1 i¼1

where nexp is number of trials, nc exp is number of components.

A program written in FORTRAN has been used to calculate the kinetic parameters. The pro-
gram used an algorithm based on the fourth-order Runge-Kutta method to solve the ordinary dif-
ferential equation (Sun and Li 2014a). This program is combined with a two-swarm cooperative
particle swarms optimization (Sun and Li 2014b, 2015), also written in FORTRAN, for calculating
the kinetic parameters of best fit.
The data of kinetic parameter estimation in this study come from the data of a refinery from
July 2017 to December 2017. The representative 16 groups of raw oil and catalyst properties data
are shown in Table 1, and operating conditions data are shown in Table 2, and product distribu-
tion data are shown in Table 3. The parameters of model 1 and model 2 are estimated, and the
average relative error between the actual and calculated values of product distribution is shown in
Table 4. It can be seen from Table 4 that the predicted product distribution is closer to the actual
value and the average relative error of slurry component is reduced from 23.89% in Model 2 to
7.95% in Model 1 when the slurry components are lumped separately.

3.2. Prediction of product distribution

By investigating the ability of model 1 and model 2 to predict the distribution of products, it is
possible to judge the advantages and disadvantages of the models.
 PETROLEUM SCIENCE AND TECHNOLOGY 5

Table 1. Properties of feedstock and catalyst of MIP commercial units

Feedstock oil
Catalyst
No. Density/(kg/m ) 3
Distillation range <500
C/% Distillation range >500
C/% Activity/%
1 927.5 71.2 28.8 60.4
2 923.4 70.5 29.5 55.9
3 932.3 69.1 30.8 57.1
4 921.9 69.5 30.5 56.5
5 925.3 72.5 27.5 54.1
6 927.2 61.0 39.0 54.9
7 922.8 66.0 34.0 56.1
8 926.2 60.5 39.5 56.4
9 926.6 68.2 31.2 57.1
10 928.0 66.5 33.5 56.9
11 925.1 71.5 28.5 57.4
12 928.7 66.5 33.5 56.5
13 924.8 69.3 30.7 56.2
14 925.1 63.9 36.1 56.8
15 926.9 62.8 37.2 58.0
16 925.8 66.1 33.9 56.8

Table 2. Operating conditions of MIP process

Temperature/
C Quantity/(t/h)
Outlet for Outlet for Fresh Recycle Atomizing Pre-lift
No. first reactor second reactor Preheat Pressure/MPa feedstock oil steam steam
1 530.61 513.47 242.4 0.27 316.47 14.25 17.07 0.91
2 529.21 512.8 246.9 0.27 303.33 16.04 17.33 1
3 528.13 513.17 243.2 0.26 304.83 13.05 17.5 1.01
4 533.89 514.75 248.1 0.27 288.81 28.66 16.66 0.75
5 534.44 519.78 251.9 0.27 293.72 32.89 16.89 0.79
6 533.15 517.13 239.2 0.27 297.7 23.57 17.01 0.13
7 532.28 518.77 239.3 0.26 296.81 18.42 16.99 1.79
8 535.7 513.21 232.6 0.26 291.74 4.69 17.06 0.66
9 531.32 517.18 233.2 0.26 300.5 1.25 17.06 0.31
10 534.96 512.86 237.5 0.26 287.53 6.52 17.05 0.46
11 530.34 513.49 245.0 0.27 302.01 12.55 17.32 0.97
12 530.76 513.33 248.4 0.26 299.34 25.24 17.1 0.9
13 533.58 516.01 247.0 0.27 292.56 25.13 16.84 0.77
14 534.52 517.76 241.1 0.26 289.87 22.83 16.88 0.49
15 532.98 516.13 236.9 0.26 278.71 11.73 16.92 0.82
16 533.81 513.84 236.5 0.26 287.34 10.78 17.04 0.59

Table 3. Product distribution of MIP commercial units

No. Dry gas LPG Gasoline Diesel oil Slurry Coke
1 4.37 16.53 44.20 23.81 4.30 6.79
2 3.49 15.84 47.11 20.77 5.77 7.02
3 2.63 15.81 47.27 20.47 6.08 7.73
4 2.83 14.59 44.55 25.63 4.52 7.88
5 2.71 15.80 46.17 21.95 5.60 7.77
6 3.06 15.73 45.21 22.95 5.56 7.49
7 3.76 15.88 43.26 24.55 4.56 7.99
8 2.44 14.08 44.76 24.6 6.02 8.10
9 3.39 17.20 44.74 22.55 4.09 8.03
10 3.12 16.98 43.87 23.41 4.32 8.30
11 3.52 16.42 46.55 21.12 5.58 6.80
12 2.73 16.09 45.22 25.07 4.34 6.55
13 2.94 15.70 45.48 24.08 5.03 6.78
14 3.49 16.86 46.44 21.71 5.74 5.76
15 3.65 15.52 45.01 23.49 5.00 7.33
16 4.16 16.19 44.50 24.38 4.11 6.66
6 S. SUN ET AL.

Table 4. Average relative error between the actual and predicted values
Product distribution/w% Model 1 Model 2
Dry gas 17.74% 26.07%
LPG 2.26% 3.07%
Gasoline 1.42% 1.99%
Diesel 1.36% 3.18%
Slurry 7.95% 23.89%
Coke 14.20% 15.93%

Table 5. Predicted product distribution of model 1

Operating condition
Outlet temperature for first reactor/
C 530 530 530 530 530 520 525 530 535 540
Outlet temperature for second reactor /
C 500 505 510 515 520 510 510 510 510 510
Pressure/MPa 0.265 0.265 0.265 0.265 0.265 0.265 0.265 0.265 0.265 0.265
Recycle ratio 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04
Product distribution/w%
Dry gas 1.77 1.89 2.05 2.25 2.50 1.85 1.96 2.05 2.08 1.81
LPG 16.13 16.32 16.48 16.65 16.82 16.39 16.48 16.48 16.40 16.79
Gasoline 45.32 45.27 45.05 44.68 44.23 44.66 44.72 45.05 45.54 43.81
Diesel 24.02 24.25 24.37 24.42 24.44 25.26 24.97 24.37 23.46 24.03
Slurry 5.82 5.12 4.74 4.54 4.43 4.62 4.58 4.74 5.21 5.57
Coke 7.06 7.21 7.33 7.45 7.58 7.25 7.31 7.33 7.35 8.11

Table 6. Predicted product distribution of model 2

Operating condition
Outlet of the first reactor/
C 530 530 530 530 530 520 525 530 535 540
Outlet for second reactor /
C 500 505 510 515 520 510 510 510 510 510
Pressure/MPa 0.265 0.265 0.265 0.265 0.265 0.265 0.265 0.265 0.265 0.265
Recycle ratio 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04
Product distribution/w%
Dry gas 1.43 1.57 1.72 1.87 2.05 1.27 1.47 1.72 2.01 2.37
LPG 15.27 15.63 15.96 16.27 16.55 15.34 15.64 15.96 16.32 16.72
Gasoline 44.35 45.32 46.16 46.86 47.41 46.28 46.28 46.16 45.89 45.45
Diesel 22.65 22.90 23.06 23.14 23.13 22.79 22.95 23.06 23.11 23.11
Slurry 9.90 7.97 6.28 4.83 3.62 8.06 7.13 6.28 5.50 4.78
Coke 6.76 6.92 7.07 7.23 7.38 6.58 6.81 7.07 7.39 7.75

According to actual industrial operating parameters, in the recovery ratio of 0.04, reaction
pressure (gauge) of 0.265 MPa, range of reaction temperature from 500
C to 540
C, predict the
influence of different first and secondary outlet temperatures on product distribution, respect-
ively. The results are shown in Table 5 and Table 6, respectively. It can be seen that the slurry
yield predicted by Model 2 is relatively high under certain operating conditions, even reaching
9.9 W%, which would not be achieved in industrial. At the same time, the product distribution
predicted by model 1 is within a reasonable range, which shows that the product prediction abil-
ity of the model can be significantly improved by adding slurry lumped into the lumped model.

4. Conclusions
1. Since the properties of feedstock are quite different from those of slurry, it is necessary to
consider slurry as an independent lump. At the same time, due to the longer reaction time
of MIP, the condensation reaction should be fully considered.
2. According to the mechanism of catalytic cracking reaction, 8 lumped kinetic models includ-
ing slurry and 7 lumped kinetic models without slurry were established respectively.
 PETROLEUM SCIENCE AND TECHNOLOGY 7

3. Based on the industrial data, the fourth order Runge-Kutta method and particle swarm opti-
mization algorithm are used to search and determine the parameters of the dynamic model.
The calculated results show that the predicted product distribution of 8 lumped kinetic
model is closer to the actual value than that of 7 lumped kinetic model. According to the
prediction results of product distribution, 8 lumped kinetic model is more consistent with
the actual industrial situation.

References
Ancheytaju
arez, Jorge, Felipe L
opezisunza, and Enrique Aguilarrodr
ıguez. 1997. A strategy for kinetic parameter
estimation in the fluid catalytic cracking process. Industrial & Engineering Chemistry Research 36 (12):5170–4.
Gao, Haohua, Wang Gang, Chunming Xu, and Jinsen Gao. 2014. Eight-lump kinetic modeling of vacuum residue
catalytic cracking in an independent fluid bed reactor. Energy & Fuels 28 (10):6554–62.
Jiang, Hongbo, and Shuai Huang. 2016. Eight-lump reaction kinetic model for the maximizing isoparaffin process
for cleaning gasoline and enhancing propylene yield. Energy & Fuels 30 (12):10770–6.
Khandeparker, Ashish. 2012. Study of different operating parameters of fcc unit with aspen-hysys. National Institute
of Technology, Rourkela.
Lopes, G. C., L. M. Rosa, M. Mori, J. R. Nunhez, and W. P. Martignoni. 2011. Three-dimensional modeling of
fluid catalytic cracking industrial riser flow and reactions. Computers & Chemical Engineering 35 (11):2159–68.
Mehran, Heydari, Ale Ebrahim Habib, and Dabir Bahram. 2010. Study of seven-lump kinetic model in the fluid
catalytic cracking unit. American Journal of Applied Sciences 7 (1):71–6.
Ouyang, Fusheng, Yongqian Wang, and Ling Qiao. 2016. A lumped kinetic model for heavy oil catalytic cracking
FDFCC process. Liquid Fuels Technology 34 (4):335–42.
Pinheiro, Carla I. C., Joana L. Fernandes, Lu
ıs Domingues, Alexandre J. S. Chambel, In^es Graça, Nuno M. C.
Oliveira, Henrique S. Cerqueira, and Fernando Ram^ oa Ribeiro. 2012. Fluid catalytic cracking (FCC) process
modeling, simulation, and control. Industrial & Engineering Chemistry Research 51 (1):1–29.
Sun, Shiyuan, and Jianwei Li. 2014a. Parameter estimation of methanol transformation into olefins through
improved particle swarm optimization with attenuation function. Chemical Engineering Research & Design 92
(11):2083–94.
Sun, Shiyuan, and Jianwei Li. 2014b. A two-swarm cooperative particle swarms optimization. Swarm and
Evolutionary Computation 15:1–18.
Sun, Shiyuan, and Jianwei Li. 2015. Kinetics of the entire methanol to olefins process over SAPO-34 catalyst.
Progress in Reaction Kinetics and Mechanism 40 (1):22–34.
Xiong, Kai, Chunxi Lu, Zhifeng Wang, and Xionghou Gao. 2015. Kinetic study of catalytic cracking of heavy oil
over an in-situ crystallized FCC catalyst. Fuel 142:65–72.
Zhu, Wenjun, Shaoyuan Li, and Yi Zheng. 2017. Modeling and parameter estimation for MIP-CGP riser reactor of
FCCU. Paper presented at the Chinese Automation Congress (CAC), Jinan, China.