Arabian Journal for Science and Engineering

https://doi.org/10.1007/s13369-020-04376-y

RESEARCH ARTICLE-CHEMICAL ENGINEERING

A Detailed Reaction Kinetic Model of Heavy Naphtha Reforming

Zaidoon M. Shakor1 · Adnan A. AbdulRazak1 · Khalid A. Sukkar1

Received: 22 June 2019 / Accepted: 20 January 2020

© King Fahd University of Petroleum & Minerals 2020

Abstract
A detailed reaction kinetic model was developed to describe heavy naphtha reforming reactions. The kinetic model involved
32 lumps and 132 reactions; the lumps were one to 11 carbon atoms n-paraffins, four to 11 carbon atoms iso-paraffins, methyl-
cyclopentene, and six to 11 carbon atoms for naphthenes and aromatics. All computations in the present study were predicted
using the particle swarm optimization (PSO) method coded by MATLAB 2015a software. This optimization method was
used to estimate the optimum set of kinetic parameters of heavy naphtha reforming reactions. All 150 kinetic and deactiva-
tion parameters that were predicted in this work were fine-tuned using PSO. The proposed kinetic model was validated by
benchmarking the model results with the data collected over 5 years for a commercial naphtha reforming unit. The mean
absolute error for all component compositions within the process was found to be 0.0079. The catalyst deactivation rate was
also predicted. It was found that catalyst activity decayed to 58.8% after 1225 operating days.

Keywords Kinetic model · Heavy naphtha reforming · Deactivation · Particle swarm optimization
List of Symbols Pt Reactor pressure (bar)
A, B, C, D Constants of specific heat polynomial R Gas constant (J/mol K)
Ac Reactor cross-sectional area (m2) r Reaction rate (kmol/kg cat h)
a Catalyst activity coefficient S Reaction stoichiometry
CP Specific heat (kJ/kmol K) T Reaction temperature (K)
Dp Catalyst pellet diameter (m) To Reference temperature (K)
EA Activation energy (J/mol) w Catalyst weight (kg)
Fi Species i molar flow rate (kmol/hr) y Mole fraction
G Mass flux (kg/m2 s) Z Length of reactor (m)
Hj Enthalpy (kJ/kmol) ∆HRj Heat of jth reaction (kJ/kmol)
Hfo Standard enthalpy of formation (kJ/kmol)
Greek Letters
gc Acceleration due to gravity (m/s2) 𝜀b Void fraction of reactor bed (m3/m3)
I Component i α Power of pressure effect
j Reaction j 𝜇 Viscosity (kg/m s)
kio Pre-exponential factor ρ Density (kg/m3)
ki Reaction rate constant (kmol h−1) ρcat Catalyst density (kg/m3)
kd Catalyst deactivation rate constant (day−1)
MCP Methylcyclopentene
Pi Component i partial pressure (bar) 1 Introduction

Naphtha reformate is the main feedstock for a gasoline

Electronic supplementary material The online version of this
article (https://doi.org/10.1007/s13369-020-04376-y) contains blending unit. The reformate octane number relates directly
supplementary material, which is available to authorized users. to the isomer and aromatic compositions. The naphtha
reforming process is one of the most effective methods to
* Adnan A. AbdulRazak increase the isomers and aromatics in reformates. In spite of
adnansss2002@yahoo.com
the fact that a heavy naphtha reforming unit has economi-
1
Department of Chemical Engineering, University cal benefits, it also protects the environment by minimizing
of Technology, Baghdad, Iraq

13
Vol.:(0123456789)
 Arabian Journal for Science and Engineering

the amount of harmful gases emitted into the atmosphere, equation in a kinetic model of 22 lumps. Ramage et al. [9]
such as carbon dioxide and nitrous oxide, while also sup- presented Mobil’s reaction kinetic model, which was bench-
plying hydrogen to other refining units, such as hydrotreat- marked with data collected from an industrial pilot plant. In
ing reactors. In this unit, the hydrotreated heavy naphtha 1987, Froment [10] extended the Hougen–Watson rate equa-
was converted to more valuable gasoline, having a higher tion to describe simultaneous reactions, choosing naphtha
octane number [1, 2]. Heavy naphtha is a petroleum fraction reforming as an example. Taskar and Riggs [11] presented
consisting of more than 300 hydrocarbons having carbon a detailed reaction model of 35 lumps in the range C5–C10
numbers ranging from C1 to C12, and each one of these and 35 reactions using the Langmuir–Hinshelwood reaction
components undergoes several reactions within the reform- rate equation. The kinetics of coke formation was included
ing process [3]. in their model to represent the catalyst deactivation term. In
Generally, the naphtha reforming unit is comprised of 2000, Ancheyta et al. [12] extended the Krane model [9],
three or four adiabatic reactors operated at 450 to 520 °C and taking into consideration the effect of the reactor pressure
10–35 bar pressure, in a 3–8 hydrogen/hydrocarbon molar on the reaction rate constants. Their kinetic model involved
ratio [4, 5]. The typical reforming catalyst is Pt supported 21 lumps and 51 reactions. Hu et al. [13, 14] used molecu-
on alumina [5]. However, other metals may be used with Pt lar series matrices to represent the naphtha composition
in the preparation of the reforming catalysts, such as Pd, Sn, within a reforming process. The reaction model included
and Gr. These metals work to promote the desired reactions 21 molecular classes and 51 reactions. Weifeng et al. [15]
with low deactivation rate. Several chemical reactions occur developed an 18-lump kinetic model. They tuned kinetic
in the reforming reactors, such as hydrocracking, dehydroge- parameters by comparing simulation results estimated by
nation, hydroisomerization, isomerization, and dehydrocy- Aspen Plus software, using data from the industrial unit.
clization. The quality and quantity of the reformate products’ Wei et al. [16] used Kinetic Model Editor (KME) software,
distribution mainly depends on the catalyst type, feed qual- which includes a standard model library, to develop and
ity, and performance of the catalyst. Therefore, the reform- customize the reforming reaction network. Mirko et al. [17]
ing catalyst performance is highly affected by the reaction predicted a semiempirical kinetic model by benchmarking it
temperature, operating pressure, deactivation rate, and reac- with industrial data. “Activation energy lumps” were intro-
tion kinetics in the reactors [2, 4]. duced to represent the activation energies within specific
Modeling, optimization, and control of petroleum reac- reaction classes.
tors require an accurate kinetic model. Sixty years ago, vari- One of the most successful kinetic models was proposed
ous studies were undertaken to describe the reaction kinetics in 2011 by Rodríguez and Ancheyta [18], who developed
of heavy naphtha reforming. Some of these kinetic models a detailed kinetic model of 71 chemical reactions and 33
are very simple, whereas others are very sophisticated. Fur- lumps, plus 41 paraffin isomers. Their model takes into
thermore, some of these kinetic models distinguish between account hydrocarbons up to 11 carbon atoms, benzene for-
iso and normal paraffins, while others do not. These kinetic mation reactions, 41 iso-paraffins, and pressure effects on
models differ in their number of reactive lumps, number of kinetic parameters. Iso- and n-paraffin distributions were
reactions, and type of reaction equations. estimated by thermodynamic equilibrium relations. How-
The lumping method is widely applied in the simulation ever, the catalyst deactivation rate was not considered in
process to reduce the complexity of the reforming reaction [18].
networks. Therefore, it is important to note that the reaction Shakor [19] used data from a commercial unit to identify
network and kinetic parameters are highly dependent on the the parameters of a kinetic model involving 29 pseudo-com-
number of lumps in the simulation system [3]. Splitting par- ponents and 83 reactions. Six isomerization reactions were
affin lumps into n-paraffins and iso-paraffins is critical when introduced to describe the equilibrium relations between iso-
evaluating the reformate composition and octane number. and n-paraffins within the range of the carbon molecules
Thus, it is necessary to develop a detailed kinetic model C5–C10.
to predict the reformate composition during the reforming Taillar [20] estimated the kinetic and catalyst deactiva-
process. tion parameters using commercial and micro-pilot plant
In 1959, Smith [6] proposed the first successful kinetic data. Their study considered the effect of temperature, two-
model, which consists of three lumps (i.e., paraffins, naph- dimensional pressure drops, and two-dimensional catalyst
thenes, and aromatics). Krane’s [7] model consisted of 20 activity variation. Arani et al. [3] applied Hougen–Wat-
individual hydrocarbon lumps and 53 reactions. The prod- son–Langmuir–Hinshelwood in a kinetic model containing
ucts were yielded and their resulting compositions were 17 lumps, with a hydrocarbon range from C6 to C8+ and 15
assessed and compared with the experimental data. In 1973, reactions. Zagoruiko et al. [21] proposed a thermodynami-
Kmak and Stuckey [8] used the Langmuir–Hinshelwood rate cally consistent kinetic model of 62 reactant groups and 146

13
Arabian Journal for Science and Engineering

reactions. The catalyst deactivation by coke deposition was counted all of these targets and is based on real data from a
also presented in their model. Iranshahi et al. [22] developed commercial naphtha reforming unit.
a two-dimensional model to simulate a naphtha reforming
unit. The reaction network consisted of 32 lumps and 84
reactions, including the catalyst deactivation in the model. 2 Commercial Naphtha Reforming Unit
Polovina et al. [23] used the Levenberg–Marquardt optimi-
zation method to estimate the parameters of a kinetic model Figure 1 represents a schematic diagram of an Al-Doura
of 51 chemical species (16 pure components and 35 pseudo- Refinery semiregenerative heavy naphtha reforming unit that
components). Various assumptions were used to reduce the contains four reactors with interstage heaters in series and
kinetic parameters from 126 to 44, and the absolute relative occupied by a Pt/Al2O3 catalyst. The samples taken from
error between the model predictions and the data from the the feed and from four reactors’ effluents were analyzed by
commercial continuous catalytic reformer unit was lower gas chromatography–mass spectroscopy (GC–MS). The unit
than 2%. Zhou et al. [24] developed a detailed kinetic model was operated at 467–481 °C feed temperature, 27.5 bar feed
involving 447 naphtha molecules, 1469 reactions, isomers pressure, and a 7 mol/mol hydrogen to hydrocarbon feed
of paraffin and naphthene up to C9, aromatic isomers up ratio. The reactors’ dimensions and catalyst weight for each
to C10, and a catalyst deactivation term. Babaqi et al. [25] of the four reactors are displayed in Table 1.
investigated the kinetic model of 36 lumps and 55 reactions.
They found that the average absolute deviation (AAD%) was
1.03%, 2.6%, 1.3%, 0.43%, 0.93%, and 2.5% for temperature, 3 Mathematical Modeling
pressure, octane number, hydrogen yield, light gases, and
reformate yield, respectively. 3.1 Model Assumptions
The instantaneous iso-paraffin concentration cannot be
predicted by most new models. Additionally, most of the The following assumptions were considered in the math-
previous studies did not consider the catalyst deactivation ematical modeling:
rate, while others calculated the catalyst deactivation rate
using pilot plant data. Obtaining the catalyst activity coef- Table 1 Heavy naphtha reforming reactor specifications of the Al-
ficient is essential in calculating the catalyst life and decid- Doura Refinery in Baghdad
ing the catalyst replacement duration. There was a need to Reactor number 1 2 3 4
develop a new kinetic model for naphtha reforming reactions
Catalyst weight, kg 2700 4500 4750 5875
that would include the prediction of instantaneous change in
Reactor length, m 6 6 6 6
iso-paraffin concentrations in addition to counting the cata-
Reactor diameter, m 2.4 2.4 2.4 2.4
lyst deactivation rate. The model proposed in this research

Recycle gas Compressor Off gas

Reactor 1 Reactor 2 Reactor Reactor 4

3

Separator

Heater 1 Heater 2 Heater 3 Heater 4

Reformate

Feed
Cooler
Heat exchanger

Fig. 1 Al-Doura semiregenerative naphtha reforming process

13
 Arabian Journal for Science and Engineering

1. Plug flow in reactor [10]. 3.3 Kinetic Model

2. Negligible radial dispersion [26].
3. Homogeneous phase reactions [12]. The developed reaction kinetic model is an extension of
4. Pseudo-first-order reactions with respect to hydrocarbon Rodríguez and Ancheyta’s [18] model, which included
[12, 27, 28]. three groups: paraffins (1 to 10 carbon atoms), naphthenes
(6 to 10 carbon atoms), and aromatics (6 to 10 carbon
3.2 Model Equations atoms). These authors grouped the individual n- and iso-
paraffins as single lumps, and they used equilibrium rela-
By applying the basic principles of material and energy bal- tions to determine the isomers’ compositions within each
ance, the following differential equations were developed paraffin lump. Additionally, they grouped the C10 and
[17, 29]:
+
higher carbon atoms using single lumps as P+10 , N10 , and A+10
for paraffins, naphthenes, and aromatics, respectively.
m
dFi ∑ In the proposed kinetic model, it was assumed that the
=− Si,j rj (1) iso-paraffins with 4–11 carbon atoms undergo the same
dw j
chemical reactions as the n-paraffins. Moreover, the par-
∑m affin isomerization reactions of four to 11 carbon atoms
dT − j=1 rj ΔHRj (n-Pi ↔ iso-Pi) were studied. The model employs a math-
= ∑n (2)
dw F Cpi ematical representation of 32 lumps that undergo 132
i=1 i
chemical reactions within the reaction network, as shown
The reaction rate and reaction rate constant were calcu- in Fig. 2. The activation energies were classified into nine
lated using Eqs. (3) and (4): groups, according to reaction type. Each one of these 132
reaction rate constants (ki) was fine-tuned using particle
rj = kj Poi (3) swarm optimization, by selecting 132 pre-exponential
factors, nine activation energies, and nine pressure effect
EAj
[ ( )]( )𝛼
1 1 pt parameters. Table 2 includes a comparison between the
kj = kjo exp − (4) values predicted by the present model and the actual data
R To T pto
reported for the activation energy and pressure effect fac-
where i represents the number of the component and j rep- tors [18]. The great difference between these values can be
resents the number of reactions. attributed to the difference between the reaction mecha-
The heat capacity was evaluated using the following nisms, feedstock compositions, catalyst types, and oper-
third-order polynomial [30]: ating conditions. Furthermore, the values of the reaction
rate constant are a nonlinear combination of three terms:
Cpi = Ai + Bi T + Ci T 2 + Di T 3 (5) (1) the pre-exponential factor, (2) the negative exponential
The heat of the reaction was calculated by the following: of the activation energy, and (3) the power law of pressure
∑ ∑ effect. In contrast, for the same reaction, different activa-
ΔHro = o
ΔHr,prodcts − o
ΔHr,reactants (6) tion energies were predicted by various scholars using a
number of reaction schemes. For example, the activation
T
energy for the paraffin dehydrocyclization was calculated
to be about 711.52–720.65 kJ/mol by Babaqi et al. [25],
∫
[∑ ∑ ]
ΔHt = ΔHro + ΔCpprodcts − ΔCpreactants dT (7) 275.276 kJ/mol by Rahimpour et al. [32], and 188.28 kJ/
To mol by Rodríguez et al. [18], while in the present study, it
was calculated to be 52.712 kJ/mol.
An Ergun equation was used to compute the pressure drop
[31]:
3.4 Kinetic Parameters Estimation
( )( ( ) )
dPt G 1 − 𝜀b 1 − 𝜀b 𝜇
=− 150 + 1.75G (8) A differential evaluation–optimization algorithm was
dz 𝜌gc Dp 𝜀3 b
Dp
used to estimate a global optimum set of kinetic param-
eters by minimizing the objective function (Eq. 10),
dw ( ) which represents the deviations of the compositions and
= AC 𝜌C 1 − 𝜀b (9)
dz the temperature of the reactors’ effluents. The model
involved 35 ordinary differential equations (33 mass

13
Arabian Journal for Science and Engineering

Fig. 2 Proposed reaction network for naphtha reforming

Table 2 Compression between Predicted Reported [18]

the predicted and reported
activation energy and pressure EA (kJ/mol) 𝛼 EA (kJ/mol) 𝛼
effect parameters
Dehydrocyclization of paraffins ( Pn → Nn) 52,712 − 0.66 188.28 − 0.7
Hydrocracking of paraffins ( Pn → Pn−1 + Pi) 72,254 0.20 230.12 0.433
Isomerization of paraffins (iPn ↔ nPn) 135,455 0.00 – 0.37
Dehydrogenation of naphthenes ( Nn → An) 40,528 0.11 125.52 0
Hydrodealkylation of naphthenes ( Nn → Nn−1 + Pi) 186,470 0.49 230.12 0.5
Ring opening of naphthenes ( Nn → Pn) 23,345 0.96 188.28 –
Hydrodealkylation of aromatics ( An → An−1 + Pi) 138,888 0.29 167.36 0.5
Ring opening of aromatics ( An → Pn) 138,635 1.17 188.28 –
Hydrogenation of aromatics ( An → Nn) 149,622 0.70 125.52 –

balance equations, one heat balance equation, and one nexp nreac ⎛
� � 1 n� � T exp − T pred �⎞
1
Com
� exp pred � � i,j i,j �
Ergun equation). The hydrogen mass balance equation f = ⎜ �yi,j,k − yi,j,k � + �� �⎟
nexp nreac i=1 j=1 ⎜ nCom k=1 � � exp �⎟
was also included. A fourth-order Runge–Kutta integra- � Ti,j �
⎝ � �⎠
tion method was used to solve these differential equa- (10)
tions, which are represented in Eqs. (1), (2), and (8),
for each individual reactor within the process to obtain
concentrations, temperature, and pressure drop profiles.
4 Results and Discussion
All computations within this study were predicted using
programs coded using MATLAB 2015a software. Par-
4.1 Validity of the Predictive Kinetic Model
ticle swarm optimization was used to estimate the opti-
mum set of kinetic parameters. MATLAB functions pso
To prove the validity of the developed kinetic model, the
and ode45 were used in optimization and integration. The
model predictions were compared with data collected from
values of the pre-exponential factors (A1 to A132) were
a commercial reforming process, as shown in Figs. 3, 4,
estimated and are presented in Table 3.
5, 6, 7, 8, 9 and 10.

13
 Arabian Journal for Science and Engineering

Table 3 Kinetic constants of the Reaction C4 C5 C6 C7 C8 C9 C10 C11

proposed kinetic model
nPn → iPn 0.0034 0.0073 0.0763 0.0670 0.1052 0.4088 0.0592 0.0019
iPn → nPn 0.0031 0.0162 0.0261 0.0351 0.0344 0.0149 0.0155 0.0296
Nn + H2 → nPn 0.0035 0.0016 0.0373 0.0020 0.0317 0.1008
Nn → An + 3H2 0.6178 0.7168 0.5233 0.8036 7.5196 0.0009
An + 4H2 → nPn 0.0126 0.0029 0.0165 0.0033 0.0179
An + 4H2 → iPn 0.0152 0.0098 0.0508 0.7510 0.0038
Nn + H2 → iPn 0.4533 0.0086 0.1945 0.0182 0.1057 1.4912
nPn → Nn + H2 0.0016 0.0412 0.0116 0.0070 0.0074 0.0024
iPn → Nn + H2 0.0005 0.0155 0.0861 0.3145 0.0039 0.0248
An + 3H2 → Nn 0.0137
nPn + H2 → nPn−1 + nP1 0.0002 0.0053 0.0014 0.0003 0.0017 0.4301 0.0010
nPn + H2 → nPn−2 + P2 0.0111 0.0001 0.0006 0.0011 0.0023 0.0331 0.0007
nPn + H2 → nPn−3 + P3 0.0000 0.0006 0.0006 0.0035 0.0010
nPn + H2 → nPn−4 + nP4 0.0001 0.0004 0.0005
nPn + H2 → nPn−5 + nP5 0.0012
iPn + H2 → iPn−1 + P1 0.0003 0.0021 0.0099 0.0037 0.0047 0.0012 0.0072
iPn + H2 → Pn−2 + P2 0.0126 0.0001 0.0045 0.0044 0.0004 0.0021 0.3902
iPn + H2 → iPn−3 + P3 0.0001 0.0003 0.0008 0.0011 0.0021
iPn + H2 → iPn−4 + iP4 0.0002 0.0000 0.0003
iPn + H2 → iPn−5 + iP5 0.0007
Nn + H2 → Nn−1 + P1 0.0013 0.0026 0.0058 0.0018
Nn + H2 → Nn−2 + P2 0.0011 0.0083 0.0025
Nn + H2 → Nn−3 + P3 0.0041 0.0011
An + H2 → An−1 + P1 0.0010 0.0033 0.9012 0.0038
An + H2 → An−2 + P2 0.0028 0.0038 0.0435
An + H2 → An + P3 0.0008
nPn + H2 → 2nPn∕2 0.0013 0.0001 0.0052
iPn + H2 → 2Pn∕2 0.0009 0.0001 0.0018
iPn + H2 → nPn−4 + iP4 0.0001
nPn → MCP + H2 0.0037
iPn → MCP + H2 0.0037
MCP + H2 → nPn 0.0230
MCP + H2 → iPn 0.0994
MCP → Nn 0.0097
Nn → MCP 0.0100

Practical data include analysis of feedstock to the first consideration in the simulation program, and the cata-
reactor, analysis of the output of each of the four reactors, lyst was not changed or regenerated during the sampling
recording the feedstock temperatures of the first reactor period.
and the temperatures of the outputs from the four reactors Figure 3 shows the comparison between the actual and
as well as the feedstock pressure before the first reactor predicted compositions of four reactors’ effluents. The pre-
and after the last reactor. These data were obtained over dicted compositions agree very well with the actual process
1225 days from a heavy naphtha reforming unit of the Al- data, with a mean absolute deviation equal to 0.0079. This
Doura Refinery, located in southern Baghdad. The first error level results from taking nine data sets at various time
data set was taken after exchanging the consumed catalyst periods, starting with a fresh catalyst. Adding the pressure
of the four reactors with a fresh catalyst to ensure that the effect factor (α) reduced the deviation level by approxi-
catalyst was highly efficient initially. Ten data sets were mately 5%.
taken over the entire time period after the catalyst was Figure 4 presents a comparison between the actual and
replaced. Changes in the process conditions (i.e., tem- predicted composition of the iso-paraffins, n-paraffins, naph-
perature, pressure, and feed composition) were taken into thenes, and aromatics with respect to the catalyst weight

13
Arabian Journal for Science and Engineering

Fig. 3 Comparison between the 0.25 0.25

Predicted weight fractions

actual and predicted results

Predicted weight fractions

0.2 0.2

0.15 0.15

0.1 n-paraffins 0.1

n-paraffins
i-paraffins
i-paraffins
0.05 Naphthenes 0.05 Naphthenes
Aromatics
Aromatics
0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
Experimnental weight fractions Experimnental weight fractions
a Reactor 1 b Reactor 2
0.25 0.25
Predicted weight fractions

Predicted weight fractions

0.2 0.2

0.15 0.15

0.1 0.1
n-paraffins n-paraffins
i-paraffins i-paraffins
0.05 Naphthenes 0.05 Naphthenes
Aromatics Aromatics
0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
Experimnental weight fractions Experimnental weight fractions
c Reactor 3 d Reactor 4

0.6 0.1 n-p4

n-p5
n-p6
n-paraffin (exp.) 0.08 n-p7
0.5
Weight Fraction

n-paraffin(pred.) n-p8
i-paraffin (exp.) n-p9
i-paraffin (pred.)
Weight Fraction

n-p10
0.4 Naphthene (exp.) 0.06 n-p11
Naphthene (pred.) n-p4 (exp.)
Aromatics (exp.) n-p5 (exp.)
Aromatics (pred.) 0.04 n-p6 (exp.)
0.3
n-p7 (exp.)
n-p8 (exp.)
0.02 n-p9 (exp.)
0.2 n-p10 (exp.)
n-p11 (exp.)

0.1 0
0 5000 10000 15000 20000
Catalyst weight (kg)
0
0 5000 10000 15000 20000
Catalyst weight (kg) Fig. 5 Comparison between the actual and predicted n-paraffin com-
position (lines = predicted; symbols = actual)
Fig. 4 Comparison between the actual and predicted n-paraffin, i-par-
affin, naphthene, and aromatic total composition (lines = predicted;
symbols = actual) about 5% of the reformate, and the rate of the naphthene
disappearance in the last three reactors was very low. The
increase in the aromatic concentrations in the second and
within four reactors. The naphthene dehydrogenation and third reactors was due to the disappearance of the n-par-
aromatic production occurred more quickly than the other affins. Hydrocracking of naphthenes and paraffins occurs
reforming reactions. Therefore, these reactions took place slowly in exothermic reactions. The iso-paraffins and aro-
in the first reactor, where the variation in the naphthene matics were obtained by converting the n-paraffins and
and aromatic concentrations was highly significant. After naphthenes to iso-paraffins and aromatics. Cyclization and
the second reactor, the naphthene compositions approached aromatization of paraffins are desired reactions because they

13
 Arabian Journal for Science and Engineering

i-p4
750
0.12
i-p5
i-p6
740
Weight Fraction

0.1 i-p7

Teamperature (K)
i-p8
i-p9
0.08 i-p10
730
i-p11
i-p4 (exp.) 720
0.06 i-p5 (exp.)
i-p6 (exp.) Temperature (exp.)
i-p7 (exp.) 710 Temperature (pred.)
0.04 i-p8 (exp.)
i-p9 (exp.)
0.02 i-p10 (exp.) 700
i-p11 (exp.)

0 690
0 5000 10000 15000 20000 0 5000 10000 15000 20000
Catalyst weight (kg) Catalyst weight (kg)

Fig. 6 Comparison between the actual and predicted i-paraffin com- Fig. 9 Comparison between the actual and predicted temperature dis-
position (lines = predicted; symbols = actual) tribution

0.08 MCP 28
N6
N7 27
Weight Fraction

0.06 N8

Pressure (Bar)
N9 26 Real Pressure
N10 Predicted Pressure
0.04 N11 25
MCP(exp.)
N6 (exp.) 24
0.02 N7 (exp.)
N8 (exp.) 23
N9 (exp.)
0 22
N10 (exp.)
0 5000 10000 15000 20000 0 5000 10000 15000 20000
N11 (exp.)
Catalyst weight (kg) Catalyst weight (kg)

Fig. 7 Comparison between the actual and predicted naphthene com- Fig. 10 Comparison between the actual and predicted pressure drop
position (lines = predicted; symbols = actual)

whereas isomerization is moderately fast, and dehydrocycli-

0.25 zation and hydrocracking are the slowest reactions. Paraffin
A6
A7 isomerization is a moderately fast reaction on the reforming
0.2 A8
catalyst, not as rapid as the dehydrogenation of naphthenes,
Weight Fraction

A9
A10 but more rapid than any other reactions. For the isomeriza-
0.15 A11
A6 (exp.)
tion reactions, temperature and pressure have a slight influ-
0.1 A7 (exp.) ence on the equilibrium.
A8 (exp.)
A9 (exp.)
Figures 5, 6, 7, and 8 compare the actual and predicted
0.05 A10 (exp.) individual component compositions of the n-paraffins,
A11 (exp.)
i-paraffins, naphthenes, and aromatics, respectively. Figure 5
0 illustrates that amount of all n-paraffins decreased within
0 5000 10000 15000 20000
Catalyst weight (kg) the four reactors at different rates. The major decrease was
observed in C8 to C11, especially in the first two reactors,
Fig. 8 Comparison between the actual and predicted aromatics com- while the composition of C4 to C7 showed only a small
position (lines = predicted; symbols = actual) change along the reactor chain. From Fig. 6 it can be seen
that the amount of some iso-paraffins decreased, while oth-
ers increase within the four reactors, but the total compo-
increase the number of branches and cyclohydrocarbons, sition of iso-paraffins changed only slightly, as shown in
hence increasing the octane number, while hydrocracking Fig. 3. As shown in Fig. 7, a higher naphthene composi-
and hydrodealkylation are mostly undesirable. Dehydrocy- tion was associated with a rapid decrease in the first reac-
clization and dehydrogenation reactions produce hydrogen tor, a moderate decrease in the second reactor, and a slow
as a by-product. Dehydrogenation is the fastest reaction, decrease in the third and fourth reactors. The effect of the

13
Arabian Journal for Science and Engineering

aromatic composition is displayed in Fig. 8, where there 100

was a rapid increase in the first reactor, a moderate increase

% Catalsyt activity
in the second reactor, and a slow increase in the third and 90
fourth reactors. Within the reforming process, A6 and A7 80
increased slightly, A8 and A9 experienced a larger increase,
while A10 and A11 remain unchanged. For n-paraffins and 70

naphthenes, a sharp change was observed in the molecules 60

with higher numbers and a moderate change was seen in
50
molecules with smaller numbers due to the hydrocracking 0 200 400 600 800 1000 1200 1400
of larger molecules into smaller molecules. High aromatic Time (day)
yields are most likely to occur at high temperatures and low
hydrogen partial pressures. However, at higher temperatures, Fig. 11 Catalyst activity profile of heavy naphtha reforming unit of
the rates of hydrocracking reactions increase sharply, which the Al-Doura refinery
is undesirable. Also, pressures below a certain level are not
feasible in commercial operations because of rapid catalyst
deactivation. These reactions are mildly endothermic; hence, The integration form of Eq. (12) is expressed by the fol-
temperature does not greatly affect their equilibrium. lowing formula [37]:
Figure 9 shows a comparison between the actual and pre- ( )
dicted temperature distribution within the four reactors. In a = exp −kd t (12)
the first reactor, the major reactions, such as the dehydroge-
The catalyst activity profile 1226 days after starting
nation of paraffins and naphthenes, were endothermic and
with a fresh catalyst is plotted in Fig. 11. Catalyst activity
very rapid, as can be seen in Fig. 4, while in the second reac-
is assumed to be equal to 1 on the first operating day. An
tor, a lower-temperature drop was observed due to paraffin
exponential decrease was seen over time. Catalyst activity
isomerization and the dehydroisomerization of the remain-
dropped to 58.8% after 1225 operating days. The Arrhenius
ing naphthenes. The temperature drop through the third and
form, such as Eq. (13), was used to describe the effect of
fourth reactors was low compared to the first two reactors
temperature on the deactivation rate constant.
due to paraffins hydrocracking and dehydrocyclization reac-
tions. For this reason, commercial catalytic reformers are
[ ( )]
217985 1 1
designed with multiple reactors and with heaters between kd = 0.00057194exp − (13)
8.314 To T
the reactors to maintain the required reaction temperature.
As can be seen, the temperature differences in the four reac-
tors were consistent with those experienced in a real-life
situation. The mean relative error between the actual and
predicted temperatures was 0.058%.
5 Conclusions
Figure 10 shows both the actual and predicted pressure
The following conclusions were drawn:
drop with respect to the accumulative catalyst weight within
the reforming process. The actual final pressure was lower
1. The proposed kinetic model agrees very well with the
than the predictive pressure drop due to the effect of catalyst
experimental data from the Al-Doura naphtha reforming
plugging. The deviation in the pressure drop within the four
unit.
reactors was about 1.27%.
2. Particle swarm optimization technology can be used
effectively to estimate parameters and optimize pro-
4.2 Catalyst Deactivation cesses.
3. Adding pressure effect factors to the Arrhenius rate con-
Catalyst activity decay occurred due to the coking, poison- stant reduced the total predicted error by 5%.
ing, sintering, and mechanical deactivation through attrition 4. Using an industrial naphtha reforming unit, the catalyst
or erosion, and corrosion/leaching by the reaction mixture activity decay to about 58.8% of its initial activity after
[33–35]. In many processes, the concentration-independent 1225 operating days.
decay functions are widely used to predict the catalyst deac-
tivation rate, as shown in Eq. (11) [36]:
Acknowledgements Zaidoon M. Shakor, Adnan A. AbdulRazak, and
da Khalid A. Sukkar wish to express their gratitude to the Department of
= −kd a (11)
dt Chemical Engineering, University of Technology, Iraq, for their sup-
port. Also, the authors acknowledge the collaboration of the Al-Doura

13
 Arabian Journal for Science and Engineering

Refinery in Baghdad/Iraq for their efforts in providing the industrial 20. Taillar, R.G.: Cross-flow naphtha reforming in stacked-bed radial
data used in this study. reactors with continuous solid circulation: catalyst deactivation
and solid circulation between reactors. Energy Fuels 26, 6938–
6959 (2012)
21. Zagoruiko, A.N.; Belyi, A.S.; Smolikov, M.D.; Noskova, A.S.:
Unsteady-state kinetic simulation of naphtha reforming and coke
References combustion processes in the fixed and moving catalyst beds. Catal.
Today 220–222, 168–177 (2014)
1. Ciapetta, F.; Wallace, D.: Catalytic naphtha reforming. Catal. Rev. 22. Iranshahi, D.; Karimi, M.; Amiri, S.; Jafari, M.; Rafiei, R.; Rahim-
5, 158–167 (1972) pour, M.R.: Modeling of naphtha reforming unit applying detailed
2. George, J.A.; Abdullah, M.A.: Catalytic Naphtha Reforming. Mar- description of kinetic in continuous catalytic regeneration process.
cel Dekker, New York (2004) Chem. Eng. Res. Des. 92, 1704–1727 (2014)
3. Arani, H.M.; Shirvani, M.; Safdarian, K.; Dorostkar, E.: Lump- 23. Polovina, S.; Vojtech, M.; Dejanović, I.; Grujić, A.; Stijepović,
ing procedure for a kinetic model of catalytic naphtha reforming. M.: Modeling a reaction section of a commercial continuous cata-
Braz. J. Chem. Eng. 26(04), 723–732 (2009) lytic reformer. Energy Fuels 32, 6378–6396 (2018)
4. Chang, A.; Pashikanti, K.; Liu, Y.A.: Refinery Engineering. Wiley, 24. Zhou, X.; Hou, Z.; Wang, J.; Fang, W.; Ma, A.; Guo, J.; Klein,
Hoboken (2012) M.T.: A molecular level kinetic model for C12 continuous cata-
5. Fahim, M.A.; Al-Sahhaf, T.A.; Elkilani, A.: Fundamentals of lytic reforming. Energy Fuels 32, 7078–7085 (2018)
Petroleum Refining. Elsevier, Amsterdam (2009) 25. Babaqi, B.S.; Takriff, M.S.; Kamarudin, S.K.; Othman, N.A.:
6. Smith, R.B.: Kinetic analysis of naphtha reforming with platinum Mathematical modeling, simulation, and analysis for predicting
catalyst. Chem. Eng. Prog. 55, 76–80 (1959) improvement opportunities in the continuous catalytic regen-
7. Krane, H.G.: Reactions in catalytic reforming naphtha. In: Pro- eration reforming process. Chem. Eng. Res. Des. 132, 235–251
ceeding of the 5th World Petroleum Congress, pp. 39–51 (1959) (2018)
8. Kmak, W.S.; Stuckey, A.N.: powerforming process studies with 26. Liang, K.M.; Guo, H.Y.; Pan, S.W.: A study on naphtha catalytic
a kinetic simulation model. AIChE National Meeting, Paper. No. reforming reactor simulation and analysis. J. Zhejiang Univ. Sci.
56a, New Orleans, March (1973) 6B, 590–596 (2005)
9. Ramage, M.P.; Graziani, K.R.; Krubeck, F.J.: Development of 27. Ancheyta, J.; Villafuerte, E.; Diaz, L.; Gonzalez, E.: Modeling and
mobil’s kinetic reforming model. Chem. Eng. Sci. 35, 41–48 simulation of four catalytic reactors in series for naphtha reform-
(1980) ing. Energy Fuels 15, 887–893 (2001)
10. Froment, G.: The kinetic of complex catalytic reactions. Chem. 28. Ancheyta, J.; Villafuerte, E.; Schachat, P.; Aguilar, R.; Gonzalez,
Eng. Sci. 42, 1073 (1987) E.: Simulation of a commercial semiregenerative reforming plant
11. Taskar, U.; Riggs, J.B.: Modeling and optimization of a semi- using feedstocks with and without benzene precursors. Chem.
regenerative catalytic naphtha reformer. AIChE J. 43, 740–753 Eng. Technol. 25, 541–546 (2002)
(1997) 29. Behin, J.; Kavianpour, H.R.: A comparative study for the simu-
12. Ancheyta, J.; Villafuerte, E.: Kinetic modeling of naphtha cata- lation of industrial naphtha reforming reactors with considering
lytic reforming reactions. Energy Fuels 14, 1032–1037 (2000) pressure drop on catalyst. Pet. Coal 51, 208–215 (2009)
13. Hu, S.; Towler, G.; Zhu, X.X.: Combine, molecular modeling with 30. Reid, R.C.; Prausnitz, J.M.; Poling, B.E.: The Properties of Gases
optimization to stretch refinery operation. Ind. End. Chem. Res. and Liquids, 4th edn. McGraw-Hill Book Company, New York
41, 825 (2002) City (1987)
14. Hu, S.Y.; Zhu, X.X.: Molecular modeling and optimization for 31. Ergun, S.: Fluid flow through packed columns. Chem. Eng. Prog.
catalytic reforming. Chem. Eng. Commun. 500, 191 (2004) 48, 89–94 (1952)
15. Weifeng, H.; Hongye, S.; Yongyou, H.; Jian, C.: Modeling, simu- 32. Rahimpour, M.R.: Operability of an industrial catalytic naphtha
lation and optimization of a whole industrial catalytic naphtha reformer in the presence of catalyst deactivation. Chem. Eng.
reforming process on aspen plus platform. Chin. J. Chem. Eng. Technol. 29, 616–624 (2006)
14, 584–591 (2006) 33. Moulijn, J.A.; Van, D.A.; Kapteijn, F.: Catalyst deactivation: is
16. Wei, W.; Bennett, C.A.; Tanaka, R.; Hou, G.; Klein, M.T.: it predictable? What to do? Appl. Catal. A: General. 212, 3–16
Detailed kinetic models for catalytic reforming. Fuel Process. (2001)
Technol. 89, 344–349 (2008) 34. Bartholomew, C.H.: Mechanisms of catalyst deactivation. Appl.
17. Mirko, Z.S.; Aleksandra, V.O.; Ivan, M.; Patrick, L.: Develop- Catal. A: General 212, 17–60 (2001)
ment of a kinetic model for catalytic reforming of naphtha and 35. Forzatti, P.; Lietti, L.: Catalyst deactivation. Catal. Today 52,
parameter estimation using industrial plant data. Energy Fuels 23, 165–181 (1999)
979–983 (2009) 36. Afshar, E.A.; Mousavi, H.; Bayesteh, H.; Towfighi, J.: Nine-
18. Rodríguez, M.A.; Ancheyta, J.: Detailed description of kinetic lumped kinetic model for VGO catalytic cracking; using catalyst
and reactor modeling for naphtha catalytic reforming. Fuel 90, deactivation. Fuel 231, 118–125 (2018)
3492–3508 (2011) 37. Naik, D.V.; Karthik, V.; Kumar, V.; Prasad, B.; Grag, M.: Kinetic
19. Shakor, Z.M.: Optimization of Al-Doura catalytic naphtha reform- modeling for catalytic cracking of pyrolysis oils with VGO in a
ing process using genetic algorithm. Eng. Technol. J. 31, 1276– FCC unit. Chem. Eng. Sci. 170, 790–798 (2017)
1296 (2013)

13