4

Modeling and Simulation of Water

Gas Shift Reactor: An Industrial Case
Douglas Falleiros Barbosa Lima1, Fernando Ademar Zanella1,
Marcelo Kaminski Lenzi2 and Papa Matar Ndiaye2
1Refinaria Presidente Getúlio Vargas – REPAR / PETROBRAS
2Universidade Federal do Paraná – UFPR

Brazil

1. Introduction
Recently, refineries finished products hydro treatment became critical due to changes in fuel
regulations. These changes are related to the specification of more clean fuels with special
focus in sulfur content reduction. In order to achieve these goals more hydro treatment is
needed. Hydrogen is the main raw material to hydro treatment units.
In refining plants, a hydrogen generation unit usually is necessary to supply the hydrogen
demand to all demanding processes.
The process known as steam reform unit is the most widely adopted technology. In large
scale, it has the highest energetic efficiency and the best cost-benefit ratio (Borges, 2009).
In this process the hydrogen conversion is carried out in two reactors in series. The first one,
the steam reform reactor converts steam and a hydrocarbon (naphtha or natural gas) into
syngas. In the sequence, a reactor known as water gas shift reactor (WGSR) converts the
carbon monoxide present in syngas into carbon dioxide and more hydrogen is generated.
Consequently, the WGSR, an intermediate step of hydrogen generation process, plays a key
role in a petrochemical plant due to hydrogen increasing demand.

1.1 Hydrogen generation unit

The hydrogen generation unit, based on the steam reform technology, is responsible for
approximately 95% of generated hydrogen (Borges, 2009).
A simplified unit block diagram is showed in figure 1.1.1.

Fig. 1.1.1. Hydrogen generation unit – process block diagram.

www.intechopen.com
54 Petrochemicals

First, sulfur is removed from the hydrocarbon stream (usually natural gas), in order to
prevent catalyst poisoning and deactivation with the use of a guard bed. Steam is mixed
in the main stream in a fixed steam to carbon molar basis. The steam reform reactor (SRR)
is a multitubular catalyst filled furnace reactor where the hydrocarbon plus steam are
converted into syngas at high temperatures (700ºC – 850ºC) according to the following
reaction:

o
CH 4 + H 2O ⇔ CO + 3 H 2 ΔH 298 = 205,9 kJ / mol (1)

o
CO + H 2O ⇔ CO2 + H 2 ΔH 298 = −41,1kJ / mol (2)

The reaction (1) is endothermic and reaction (2) is moderately exothermic. Both are
reversible reactions. In the SRR, (1) is the main reaction and generates most of the hydrogen.
The reaction (2) due to its endothermic nature occurs in a lower extension in the SRR. The
syngas stream composition is CO and H2, in this process CO2 and H2O are also present in
gas state. The main purpose of the water gas shift reactor (WGSR) is to carry out the reaction
(2) reducing the CO fraction and increasing the hydrogen yield. Finally, the WGSR stream is
conducted to a purification section, where hydrogen purity is increased according to the
process needs.

2. The water gas shift reaction

2.1 kinetic rate expression
The water gas shift reaction (reaction 2) is a heterogeneous reaction (gas/solid).
According to (Smith et al., 2010) in this kind of application, there are two options in the
WGSR step. Using a high temperature shift (HTS) catalyst based reactor or a series of HTS
followed by a low temperature shift catalyst based reactor (LTS) with intercooling stage to
increase the overall conversion and high purity hydrogen is needed (Newsome, 1980).
The chapter focuses on a HTS ferrochrome catalyst industrial reactor modeling.
The HTS usually is an iron oxide – chromium oxide based catalyst. Also reaction promoters
such as Cu may be present in catalyst composition. Operational temperatures vary from
310ºC to 450ºC. Inlet temperatures are usually kept at 350ºC to prevent the catalyst bed
temperature from damage. Exit CO concentrations are in the order of 2% to 4%. Industrial
reactors can operate from atmospheric pressure to 8375 kPa. Sulfur is a poison for Fe-Cr
catalysts.
LTS reactors are copper based catalyst. Typical compositions include Cu, Zn, Cr and Al
oxides.
Recent catalysts can be operated at medium temperatures around 300ºC. Copper is more
sensitive to catalyst thermal sintering and should not be operated at higher temperatures.
Sulfur is also a poison to LTS reactors. Typical exit concentration is of 0,1% of CO.
The reaction is operated adiabatically in industrial scale, where the temperature increases
along the length of the reactor.

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 55

According to Arrhenius law of kinetics, increasing temperature increases the reaction rate.
By the other side, the thermodynamic of equilibrium or Le Châtelier principle states that
increasing the temperature of an exothermic reaction shifts the reaction to reactants side
decreasing its equilibrium conversion. Therefore the water gas shift reaction is a balance
between these effects and the reactor optimal operational point takes into account the
tradeoff between kinetics and equilibrium driving forces.
In (Chen et al., 2008) experimental data indicates that increasing temperature in HTS will
promote the performance of WGSR. For the LTS, the reaction is not excited if the reaction is
bellow 200ºC. Once the temperature reaches 200 ºC the reaction occurs, but the CO
conversion decreases with increasing temperature. This fact reveals that that the water gas
shift reactions with the HTS and the LTS are governed by chemical kinetics and
thermodynamic equilibrium, respectively in industrial conditions.
(Smith et al., 2010) classifies the reaction kinetic models in microkinetic approach and the
empirical method.
Basically, the micro kinetic approach explores the detailed chemistry of the reaction. On the
other hand, the empirical models are based on the experimental results and are typically
expressed in the Arrhenius model and provide an easy and computationally lighter way to
predict the rate of reaction. The main disadvantage is the fact that the adjusted model
cannot be extrapolated to different composition and types of catalysts.
Many empirical expressions have been reported in literature for HTS. (Newsome, 1980) and
(Smith et al., 2010).
An empirical rate expression succesfully used to describe the WGSR in ferrochrome
catalysts is a power law type: (Newsome, 1980)

 1 PCO 2 ⋅ PH 2 
2 ⋅ PH 2 ⋅  1 − 
− Ea

 
l q
⋅ PHm2O ⋅ PCO
n

 eq PCO ⋅ PH 2 O 
r = k0 ⋅ e RT ⋅ PCO ⋅ (3)
K

Where:
r – reaction rate.
Ea – activation energy.
Ko – pre exponential factor.
Keq – reaction equilibrium constant.
l, m, n, q – estimated parameters by experimental data.
Py – partial pressure of component y.
R – universal gases constant.
T – absolute temperature.
The reaction equilibrium constant derived from thermodynamics as function of temperature
is given by (Smith et al., 2010):

5693, 5 49170
( )
ln K eq =
T
+ 1,077 ⋅ ln (T ) + 5, 44 ⋅ 10 −4 ⋅ T − 1,125 ⋅ 10 −7 ⋅ T 2 −
T2
− 13,148 (4)

T – temperature in Kelvin.

www.intechopen.com
56 Petrochemicals

Several authors have published estimated parameters for specific catalysts types or
catalysts classes. (Newsome, 1980) and (Smith et al., 2010)
Table 2.1.1 summarizes some previous publicated values and authors for HTS catalysts:

Author Catalyst Ko Ea l m n q
information (kJ/mol)
Bohlbro et al. Fe2O3/ Cr2O3 - 105,9 0,93 0,24 -0,31 0,00
(Newsome, 1980) Commercial
reduced particle
size
Bohlbro et al. Fe2O3/ Cr2O3 - 59,8 0,87 0,26 -0,18 0,00
(Newsome, 1980) Commercial
Large particle
size
S.S. Hla et al. Fe2O3/ Cr2O3/CuO 102,845 111 1 0 -0,36 -0,09
(Hla et al., 2009) type 1
S.S. Hla et al. Fe2O3/ Cr2O3/CuO 100,659 88 0,9 0,31 -0,156 -0,05
(Hla et al., 2009) type 2
Adams and Barton. Fe2O3/ Cr2O3/CuO 725 110 1 0 -0,32 -0,083
(Adams & Barton, 2009) Commercial type 1
Same as Hla et al.
Table 2.1.1. Estimated parameters for power law HTS catalysts.

In industrial reactors, catalysts are loaded in pellets. Therefore, intrinsic rate expressions
cannot be used in pseudo-homogeneous models without some type of compensation to
account for diffusion effects.
According to (Newsome, 1980) two different methods can be applied to model industrial
reactors as pseudo-homogeneous reactions. The first was proposed by Bohlbro and
Jorgensen consists of estimating the empirical rate expression in laboratory using
commercial size catalysts. The diffusion effects remain implicit in the rate expression. First
and second rows of Table 2.1.1 illustrate this method. The catalyst in the first row was
grounded to avoid diffusion effects. In the second row the same catalyst was used in
commercial pellet size. It can be noted slight differences between l, m, n e q parameters. The
disadvantage of this method is that the rate expression can model successfully only a reactor
with this specific type and size of catalyst.
(Hla et al., 2009) estimates the intrinsic rate parameters for two commercial catalysts as can
be seen in rows three and four.
(Adams & Barton, 2009) also model a WGSR reactor using a heterogeneous modeling
approach. The intrinsic rate expression is from Hla et al. previous article. The parameters are
result for the best fit estimation.
The second approach of pseudo-homogeneous rate modeling consists of using correction
factors to compensate for the pore diffusion phenomena in the catalyst, catalyst age,

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 57

operating pressure and hydrogen sulfide concentration with intrinsic rate expressions.
(Singh and Saraf, 1977) is a good example of this method.

2.2 Mathematical modeling

In this section, mathematical expressions for the fixed bed adiabatic catalytic WGSR
fundamental principles (conservation equations) are developed.
A basic ideal flow steady state one-dimensional model is presented.
The differential molar balance simplified to a fixed bed reactor can be expressed as equation
5: (Froment and Bischoff, 1990)

dX a r
= a (5)
dW Fa0

Xa – component “a” conversion.

W – catalyst weight.
ra – rate of reaction of component “a”.
Fa0 – molar feed rate of reactant “a”.
Fa – molar flow of component “a” leaving the reactor.
The ideal model assumes that concentration and temperature gradients only occur in the
axial direction. The only transport mechanism operating in this direction is the overall flow
itself, and is considered to be of the plug flow type.
In case of a reactor bed with a fixed cross sectional area (S), the differential molar balance
can be rewritten as a function of the reactor differential length as can be seen in equation 6:

dX a ra ⋅ ρB ⋅ S
= (6)
dz Fa0

ρB – catalyst bulk density.

S – reactor bed constant section area.
z – length of reactor (z axis – axial direction).

Fig. 2.2.1. Reactor differential volume representation.

www.intechopen.com
58 Petrochemicals

The differential equation of energy conservation may be written as equation 7:

 Fi ⋅ Cpi ⋅ dz = ( −ΔHR ) ⋅ ra ⋅ ρB ⋅ S − 4 d
dT U (7)
⋅ (T − TR )
t

ΔHR – enthalpy of reaction.

Cpi – specific heat of component “i”.
Fi – component “i” molar flow.
T, TR – reactor temperature and temperature at radius R of internal tube.
dt – internal tube diameter.
U – overall heat transfer coefficient.
For an adiabatic reactor U equals zero and equation 3 can be simplified to:

 Fi ⋅ Cpi ⋅ dz = ( −ΔHR ) ⋅ ra ⋅ ρB ⋅ S
dT (8)

The momentum differential equation can be defined as in equation 9:

dP ρg ⋅ u s2
− = −f ⋅ (9)
dz dp

P – reactor pressure in position z.

f – friction factor.
ρg – gas density.
us – superficial velocity.
dp – particle diameter.
The aim of the next topics is to develop this set of equations to model an industrial high
temperature WGSR.

2.3 Some design aspects

When designing WGSR (Chen et al, 2008), it is known that a proper selection on certain
parameters is of the most importance because the reaction result depends strongly on the
combination of these parameters. Typically, the important parameters include the catalyst
type, residence time of reactants in a catalyst bed, reaction temperature and feeding
reactants ratio or CO/steam ratio.

2.3.1 Types of catalysts

According to the reaction temperature (Chen et al, 2008), the WGSR falls into two categories:
high-temperature shift catalyst (HTC) and low-temperature shift catalyst (LTC). The catalyst
commonly used in the former is an iron–chromium-based catalyst, whereas a copper–zinc-
based catalyst is frequently adopted in the latter.

2.3.2 Residence time

It is known that the catalyst amount in a reactor is highly related to the overall project cost.
In other words, if a reaction can be developed with least catalyst, both the operation (or

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 59

catalyst) cost and the facility (or space) cost can be reduced effectively. The least catalyst
amount can be evaluated by determining the residence time of reactants in a catalyst bed.
For both categories of reaction, a residence time of 0.09 s is long enough to establish an
appropriate WGSR (Chen et al, 2008), a design factor can be used to compensate
deactivation during the catalyst life.

2.3.3 Reaction temperature

The reaction temperature on the design of the WGSR, varies according with the type of
catalyst used (Chen et al, 2008), For HTC, increasing reaction temperature, the concentration
of CO declines with respect to the temperature, thus the conversion of WGSR increases, but
between 400ºC and 500 ºC the change in this propriety is small, so temperatures between
350ºC and 400ºC the major reaction occur.
In the case of LTC reaction, for temperatures below 150ºC, no reaction take place, increasing
to 200ºC the conversion rise to above 90%. If temperature is increased further the conversion
begins to fall. So, an optimal reaction, for the WGSR with the LTC is obtained in
temperatures around 200ºC.
The WGSR is an exothermic reaction in nature, from thermodynamics it is known that an
increase in reaction temperature will impede the forward reaction for H2 production in the
WGSR. Because the behavior, it is realized that the WGSR with the LTC catalyst is governed
by chemical equilibrium.
In contrast, for the HTC catalyst, the CO conversion is highly sensitive to the reaction
temperature and an increase in temperature is conducive to the hydrogen generation. It
follows that the reactions with the HTC are controlled by chemical kinetics (i.e. Arrhenius
law).

2.3.4 CO/steam ratio

For both kind of catalysts when the CO/steam ratio is large than 1/4, the performance of the
WGSR is sensitive the variation of the ratio. Alternatively, if the ratio is smaller than 1/4,
varying the ratio merely has a slight influence on the performance(Chen et al, 2008).

3. Experimental data
Experimental data of an industrial shift reactor of REPAR/PETROBRAS were used for the
modeling studies. The reactor performs CO oxidation to CO2 using industrial steam. The
reaction occurs in a fixed bed filled with a Fe-Cr-based catalyst, which is shown in Figure
3.1.
The historical data set comprises 4 years of operation. More specifically, samples withdrawn
from feed and exist streams were analyzed by chromatography, following the technical
standard norm (NBR-14903, 2002). Temperature measurements are performed using K type
thermocouples connected to Honneywell STT3000 transmitters positioned according to
Figure 3.2. Finally, it is important to mention that feed flow rate measurements were
performed using orifice plates connected to a Honeywell STD900 differential pressure

www.intechopen.com
60 Petrochemicals

transmitter and the reactor pressure drop was measured using manometric pressure
coupled to Honeywell STD900 transmitters.

Fig. 3.1. Catalyst.

Fig. 3.2. Temperature measurement positioning.

Experimental data of CO conversion are presented in Figure 3.3. These data were
normalized by dividing the experimental value by the reactor design conversion value. It is
important to state that all experimental data were normalized to the design values in order
to avoid numerical convergence issues along the modeling and parameter estimation tasks.

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 61

20.0
18.0
Normalized CO conversion

16.0
14.0
12.0
10.0
8.0
6.0
4.0
2.0
0.0
00 01 02 03 04
Years

Fig. 3.3. Normalized experimental CO conversion.

The modeling work will focus on CO conversion, however this variable is not directly
measured, but actually obtained from chromatographic analysis using Eq. (10), which
considers the measurement of CO, CO2 and CH4 in the feed and exit streams.

CO s ( CH 4e + CO e + CO 2e )
X co = 100 − 100 * * (10)
CO e CH 4s + CO s + CO 2s

However, it is important to estimate the experimental uncertainty (i) of CO conversion the

measurement, i.e., an estimate of the CO conversion variance. This calculation can be done
by applying an error propagation approach to Eq. (10), more specifically, an estimate can be
obtained by Eq. (11), which considers the variance of CO, CO2 and CH4 concentration
measurements, while covariance was considered zero. The variance of each measurement
was calculated according to the technical standard norm NBR-14903. This analysis yielded
an average standard deviation of 1% of the CO measurement.

 ∂X  2 ∂X   ∂X 
  * i CH 4 _ e   * i CO _ e +   * i CO 2 _ e + ...
2 2 2
2 2

 ∂CH 4_e   ∂CO _e   ∂CO 2_e 

 ∂X  2 ∂X   ∂X 
i= (11)
... +   * i CH 4 _ s   * i CO _ s +   * i CO 2 _ s
2 2 2

 ∂CH
2 2

 4_s   ∂CO _s   ∂CO 2 _ s 

www.intechopen.com
62 Petrochemicals

4. Model parameter estimation and validation

Based on the kinetic expressions, the development of a fundamental model to describe the
shift reactor behavior represents an important tool for in-depth studies and performance
optimization.
For modeling purposes, the experimental data set was firstly modeled considering a model
with a reasonable amount of simplifying hypothesis and literature reported parameters.
Afterwards, some of the hypotheses need to be disregarded, allowing an increase in model
complexity, leading to further parameter estimation. Different hypotheses were disregarded
until the model could successfully describe the experimental data set behavior. Only steady-
state analysis was performed.
Parameter estimation was performed using a simplex-based method (Himmelblau et al.,
2002) focusing on minimizing the classical least square objective function based on the
difference between experimental and predicted CO conversion values. The experimental
industrial data set was divided into two groups, the first for parameter estimation and the
second for model validation.
Despite the relaxation of some simplifying hypotheses, in all models, both axial and radial
diffusion were not considered. This choice occurred due to some features of the studied
reactor, for example, it presents a high length to diameter ratio, in order to assure a
turbulent flow for different operating conditions. Consequently, effects of a possible external
diffusive resistance tend to be reduced. A second common simplifying hypothesis present in
all models concerns the absence of pressure drop, therefore an isobaric reactor was
considered. This hypothesis was assumed because experimental data revealed a pressure
drop less than 3% along the catalyst bed, bellow literature recommendation threshold of
10% to disregard possible pressure drop effects (Iordanidis, 2002).

4.1 Model 1 – Pseudo-homogeneous isothermal reactor

The first modeling attempt represents a very simplified reactor model as given by Eq. (12)

dX a ra ⋅ ρB ⋅ S
= (12)
dz Fa0

The reaction rate is based on power law kinetics and is given by Eq. (13)

 1 yCO 2 ⋅ y H 2 
2 O ⋅ yCO 2 ⋅ y H 2 ⋅  1 − 
− Ea

 K eq yCO ⋅ y H 2 O 
l m n q


ra = FPRESS ⋅ k0 ⋅ e RT ⋅ yCO ⋅ yH ⋅ (13)

For modeling purposes, reaction rate considered molar fractions instead of the components
partial pressure, according to (Adams & Barton, 2009). According to the authors, a pressure
correction factor, Fpress, given by Eq. (14) also needs to be considered as a pressure increase
tends to increase the effect of molecular diffusion inside the catalyst pore.

 P 
 0.5 − 
FPRESS = P  250 
(14)

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 63

where: P is the absolute pressure in bar.

The equilibrium constant, Keq, is given by Eq. (15), obtained by considering temperature
dependent heat capacities.

5693, 5 49170
( )
ln K eq =
T
+ 1,077 ⋅ ln ( T ) + 5, 44 ⋅ 10 −4 ⋅ T − 1,125 ⋅ 10 −7 ⋅ T 2 −
T2
− 13,148 (15)

Finally, it is worth mentioning that the reaction rate constants were obtained from (Hla et al.
,2009) as presented in Table 4.1.1 and also that the catalyst activity was considered constant.

l m n q
k0 Ea
CO H2O CO2 H2
700 111 1.0 0 – 0.36 – 0.09
Table 4.1.1 – Reaction rate exponents.

After solving the reactor model, normalized experimental CO conversion plotted against
normalized CO conversion predicted values are shown in Figure 4.1.1. It can be observed
that a systematic deviation occurs, probably because of the number of simplifying
assumptions. The first hypothesis to be revisited is the reactor isothermal behavior, mainly
because of the direct temperature effect on CO conversion and also because of the reaction
exothermal behavior.

1.2

1.15
Normalized experimental conversion

1.1

1.05

0.95

0.9
0.900 0.950 1.000 1.050 1.100 1.150 1.200
Normalized predicted conversion

Fig. 4.1.1. Model 1 results.

www.intechopen.com
64 Petrochemicals

4.2 Model 2 – Pseudo-homogeneous non-isothermal reactor

This model represents and extension of Model 1. Due to the non-isothermal behavior, the
reaction rate is not constant along the reactor length. The exothermal characteristic of the
reaction also contributes to performing a heat balance on the reactor, yielding Eq. (16):

 Fj ⋅ cpj ⋅ dz = ( −ΔHR ) ⋅ ra ⋅ ρB ⋅ S
dT
(16)

Consequently, the reactor model is now comprised by a system of ordinary differential

equations. Figure 4.2.1 is obtained after plotting normalized experimental CO conversion
against normalized CO conversion predicted values. It can be observed, after comparison
with Figure 4.2.1, that non-isothermal feature actually resulted in worse model predictions.
By analyzing Figure 7, it can be observed that for higher experimental conversion values,
systematic higher predictions are obtained. This probably happens because the non-
isothermal behavior leads to higher temperature values, consequently not only leading to
higher reaction rates overestimating CO conversion, but also affecting diffusion phenomena
inside the catalyst particles. Therefore, an alternative modeling approach would be
considering an effectiveness factor in order to correct the reaction rate.

1.2

1.15
Normalized experimental conversion

1.1

1.05

0.95

0.9
0.900 0.950 1.000 1.050 1.100 1.150 1.200
Normalized predicted conversion

Fig. 4.2.1. Model 2 results.

4.3 Model 3 – Pseudo-homogeneous non-isothermal reactor with reaction rate

correction
In this model, the reaction rate (Eq. 13) is corrected by an effectiveness factor as presented
by Eq. (17).

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 65

ra − effective = η ⋅ ra (17)

It is important to realize that in Model 1 and Model 2 only simulation studies were carried
out, no model parameter estimated. Now, parameter η, which remains in the interval (0,1],
will be estimated in order to improve data fitting the model predictions. As mentioned
before, only part of the experimental data set will be used for parameter estimation using
the minimum least square method, while the other part will be used for validation purposes.

After parameter estimation, an optimal value of η = 0.56 was obtained. Considering this
value, the mathematical model was then used to predict CO conversion as presented in
Figure 4.3.1. It can be seen a considerable improvement in the model prediction, which was
achieved by estimating only one parameter. However, the experimental data set refers to a
considerable production horizon; consequently, catalyst activity may have changed, leading
to another opportunity for model improvement.

1.2

1.15
Normalized experimental conversion

1.1

1.05

0.95

0.9
0.900 0.950 1.000 1.050 1.100 1.150 1.200
Normalized predicted conversion

Fig. 4.3.1. Model 3 validation.

4.4 Model 4 – Pseudo-homogeneous non-isothermal reactor with reaction rate

correction and catalyst deactivation
This model considers the reaction rate given by Eq. (18)

ra − effective = η ⋅ a ( t ) ⋅ ra (18)

In this effective reaction rate, the catalyst activity, a(t), can be described by a hyperbolic rate
expression (Eq. (19)) as previously reported (Keiski et al., 1992):

www.intechopen.com
66 Petrochemicals

1
a(t) = 1 (19)
(1 + α ⋅ t) 3

It must be stressed that α is the only parameter to be estimated in Model 4 as η was kept
equal to 0.56. Parameter α was estimated as 10–5. Figure 4.4.1 presents experimental
normalized CO conversion plotted against predicted normalized values considering only
the validation data set. It can be observed that the use of catalyst deactivation improved
model performance, indicating an important role played by catalyst deactivation. However,
the intrinsic reaction rate was not yet used for parameter estimation as exponents i, m, n, p
values were considered the ones reported in the literature (see Table 4.1.1), consequently,
this creates an alternative for model improvement.

1.2

1.15
Normalized experimental conversion

1.1

1.05

0.95

0.9
0.900 0.950 1.000 1.050 1.100 1.150 1.200
Normalized predicted conversion

Fig. 4.4.1. Model 4 validation.

4.5 Model 5 – Revisiting model 4 intrinsic reaction rate

It can be observed that the order of H2O in the reaction rate is equal to 0 as reported in Table
4.1.1. This indicates that H2O is present in such a large excess that concentration changes
does not considerably affect the reaction rate (Hla et al., 2009). However, in the industrial
reactor, where the feed composition may change due to the nature of a petrochemical plant,
H2O may be in excess, but not in enough amount to be considered constant throughout the
reactor length. Therefore, exponent m needs to be re-estimated, while the others will be kept
at the same values. However, changes in m may affect the effective rate, consequently, η
value also needs to be revisited and estimated. To sum up, Model 5 basically keeps the same
structure of Model 4 only different parameters need to be estimated.

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 67

After performing the estimation task, a m value of 0.2 was obtained, simultaneously to a η
value equal to 0.575. Figure 4.5.1 presents experimental and predicted values of CO
conversion, showing that model predictions improved considering the new m value.
In order to compare the model predictions, the sum of the square difference of experimental
and predicted values of CO conversion was performed and Model 2 led to the highest sum
being the worst model. The sum of the squares of Model 2 was used as normalization factor
and the normalized sum of all models are presented in Figure 4.5.2. It can be concluded that
two hypotheses played a key role in the reactor modeling: firstly the isothermal behavior;
secondly, the effectiveness factor. Further changes were import to improve, but not
considerably, the data fit. It must also be noted that a few number of parameters were
estimated, indicating an important likelihood feature of the reactor model. Finally, it is
worth stressing that the correlation coefficient between experimental data and model
predictions of Model 5 was equal to 0.8.

1.2

1.15
Normalized experimental conversion

1.1

1.05

0.95

0.9
0.900 0.950 1.000 1.050 1.100 1.150 1.200
Normalized predicted conversion

Fig. 4.5.1. Model 5 validation.

www.intechopen.com
68 Petrochemicals

1.15

1.05
1.00
Normalized sum of squared errors

0.95

0.85

0.75

0.65
0.57

0.55

0.45 0.42
0.41
0.39

0.35
1 2 3 4 5
Proposed model number

Fig. 4.5.2. Model comparison.

5. A novel approach to optimal process control strategy

Due to the shift reaction features, for a given reactor feed stream composition and flow rate,
the temperature of the feed stream can be manipulated in order to maximize CO conversion.
Consequently, if the industrial reactor has in-line/on-line composition analyzers as well as
flow rate measurement instrumentation, on-line optimization can be successfully
implemented. The problem is that this kind of in-line/on-line instrumentation is not only
expensive but also may need continuous and excessive calibration and maintenance.
Therefore, some question arises: can other real-time measurements be used for CO
conversion? How does temperature influence CO conversion correlated? This is an
important issue as temperature measurements are usually reliable, accurate, real-time and
low cost. Moreover, fixed bed reactors can have temperature instruments installed along the
reactor length, providing a temperature profile. Towards this, a novel and alternative
approach will be presented in order to overcome this issue, focusing on CO conversion
control.
Figure 5.1 presents historical normalized CO conversion data plotted against normalized
feed flow-rate values, obtained from PETROBRAS shift reactor unit. As mentioned before,
the normalized variables equal to 1 represent the reactor design values. These data refer to
the same reactor temperature; however, due to the nature of petrochemical process, raw
material composition can fluctuate, explaining the different CO conversion values obtained
for the same experimental conditions. As expected, the higher the feed flow rate, the lower
the reactor conversion.

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 69

In Figure 5.1, one observes, for example, conversions of 1.02 or 1.04, indicating that in these
scenarios the experimental values were over the design values. The same observation can be
made for the feed flow rate, indicating some operating conditions either bellow or above the
design values. Consequently, it is important to emphasize that the reactor feed composition
represents an important disturb variable affecting reactor performance, enhancing the
importance of well tuned and designed regulatory control loops.

1.06

1.04

1.02
Normalized CO Conversion

0.98

0.96

0.94
R = -0,72
0.92
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
Normalized Feed flow rate

Fig. 5.1. Historical data: CO Conversion versus feed flow rate.

The studied shift reactor has thermocouples installed at the reactor feed and exit streams. It
also has four thermocouples equally distributed along the reactor bed. After performing
extensive sensitivity analysis, the temperature difference between the thermocouple placed
on the reactor exit stream and last thermocouple of the bed provided the highest sensitivity,
here denominated Delta_T. After analyzing the historical data set, Figure 5.2 presents
experimental data of CO conversion plotted against normalized Delta_T. It can be observed
a good correlation between both variables, which corroborate the idea of monitoring CO
conversion by using lower cost measurements. Finally, Figure 5.3 presents the relationship
between normalized Delta_T and normalized feed flow rate values, as expected indicating
strong correlation and also good sensitivity.

www.intechopen.com
70 Petrochemicals

1.06

1.04

1.02
Normalized CO Conversion

0.98

0.96
R = -0,76
0.94

0.92
-1 1 3 5 7 9 11 13 15

Normalized Delta_T

Fig. 5.2. Historical data: CO Conversion versus Delta_T.

The developed mathematical model was used for sensitivity analysis in order to study the
CO conversion behavior for different operating conditions. More specifically, this analysis is
aimed at providing not only the control feasibility itself, but also the feasibility of leading to
operating conditions which may allow optimum conversion values. Figure 5.4 shows the
CO conversion behavior for different feed flow rate values, considering in all simulations
the feed composition design value. More specifically, the design value of the feed flow rate
(FLOW1), the design value increased by 35% (FLOW2) and the design value increased by
55% (FLOW3) were used. Firstly, one can observe that due to the shape of the curve, the
feed flow rate temperature can be manipulated in order to reach a maximum CO
conversion. Secondly, the simulation results show that keeping the reactor feed temperature
constant, for example at 1.03, an increase in the flow rate may lead CO conversion reduction,
similar to the behavior of the historical data exhibited by Figure 5.1. It is also important to
observe the effect of the temperature of the reactor feed stream on the CO conversion. After
careful analysis, it can be also noted that to keep track of CO conversion, for example, at
0.955, the feed flow rate needs to be increased for a feed stream temperature increase.
Consequently, feed flow rate and feed temperature can be regarded as important
manipulated variables for CO conversion either in servo or regulatory control.

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 71

15

13

11
Normalized Delta_T

3 R = 0,94

-1
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
Normalized Feed Flow Rate

Fig. 5.3. Historical data: Delta_T versus feed flow rate.

0.98
Normalized CO conversion

0.93

FLOW1
0.88
FLOW2

0.83 FLOW3

0.78
0.97 0.99 1.01 1.03 1.05 1.07 1.09 1.11

Normalized feed temperature

Fig. 5.4. Sensitivity Analysis.

www.intechopen.com
72 Petrochemicals

However, it is important not only to provide control feasibility, but also to assure that the
control can lead to optimal conversion values. Figure 5.5 presents a sensitivity analysis
study focusing on the effect of feed composition. The reactor feed flow rate was kept
constant in all simulations, while two different feed compositions were chosen, more
specifically, the design feed composition (COMP1) and a feed composition involving an
increase in CO amount (COMP2). For each feed composition, the temperature of the feed
stream was changed in order to evaluate the correspondent resulting Delta_T values and the
maximum conversion was determined and plotted. For COMP1, the maximum conversion
is reached at a Delta_T value equal to roughly 3.7. Consequently, for this feed composition
and feed flow rate, the feed temperature set-point needs to be roughly 1.01. The same
analysis can be performed for composition 2.

18 COMP1 COMP2
16 Delta_T of Maximum CO Conversion

14
Normalized Delta_T

12

10

4 3.7
3.3
2

0
0.95 0.97 0.99 1.01 1.03 1.05 1.07 1.09
Normalized Feed Temperature
Fig. 5.5. Sensitivity Analysis: Composition.

On the other hand, Figure 5.6 presents a sensitivity analysis study focusing on the effect of
feed flow rate. The reactor feed composition was set equal to COMP2 and different values of
the flow rate, FLOW1, FLOW2, FLOW3 were chosen. For each feed flow rate value, the
temperature of the feed stream was changed in order to evaluate the correspondent
resulting Delta_T values. For each feed flow rate, the maximum conversion was also
determined and plotted. Considering the feed composition COMP2, for the feed flow rate
close to the design value, Delta_T should be equal to 3.3 to lead to the maximum CO
conversion, consequently, the feed temperature set-point should be roughly 1.02. Keeping
the composition unchanged and increasing the flow rate by 35%, Delta_T should be equal to
3.6 and, therefore, the feed temperature should be increased to approximately 1.05.
Increasing the design value of the feed flow rate by 55%, Delta_T should be changed to 4.4,
so the feed temperature should be increased to approximately 1.06.

www.intechopen.com
Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case 73

18
FLOW1 FLOW2 FLOW3
16
Delta_T of Maximum CO Conversion
14
Normalized Delta _T

12

10

4 4.4
3.3 3.6
2

0
0.97 0.99 1.01 1.03 1.05 1.07 1.09 1.11
Normalized Feed Temperature

Fig. 5.6. Sensitivity Analysis: Feed Flow.

Based on the historical data and on the sensitivity analysis results, an alternative CO
conversion control loop having the following features arises:

• controlling Delta_T represents the same as controlling CO conversion;

• controlling Delta_T can lead to optimal conversion values;
• cascade structure can be used with Delta_T control mastering the loop by providing
appropriate set-points to the reactor feed temperature control loop;
• feedforward features may also be considered as if feed flow rate changes the Delta_T
value leading to the optimal conversion needs to be updates as observed in Figure 7;
• feedforward features may also be considered if feed analyzers can be installed, as any
change in feed composition would lead to a different Delta_T set-point.

6. References
Adams II, Thomas A. and Barton, Paul I. (2009). A dynamic two-dimensional heterogeneous
model for water gas shift reactors. International Journal of Hydrogen Energy 34, 8877–
8891.
Borges, Joana Lopes. (2009). Diagrama de fontes de hidrogênio. Dissertação de Mestrado em
Tecnologia de Processos Químicos e Bioquímicos. Universidade Federal do Rio de
Janeiro. Rio de Janeiro. Brasil.
Chen, Wei-Hsin et al. (2008). An experimental study on carbon monoxide conversion and hydrogen
generation from water gas shift reaction. Energy Conversion and Management 49,
2801–2808.

www.intechopen.com
74 Petrochemicals

Edgar, Himmelblau and Lasdon. (2001). Optimization of Chemical Processes. 2nd Edition. New
York – McGraw-Hill, pp. 10-11.
Froment, Gilbert F. and Bischoff, Kenneth B. (1990). Chemical reactor analysis and design. John
Wiley and Sons, ISBN 0-471-51044-0, second edition, pp. 403-404.
Hla, San Shwe et al. Kinetics of high-temperature water-gas shift reaction over two iron-based
commercial catalysts using simulated coal-derived syngases. Chemical Engineering
Journal 146 (2009) 148–154.
Iordanidis, A. A. (2002). Mathematical Modeling of Catalytic Fixed Bed Reactors. Ph.D. thesis,
University of Twente. ISBN 9036517524, Enschede, The Netherlands.
Keiski, R. L. et al. (1992). Deactivation of the high-temperature catalyst in nonisothermal
conditions. Applied Catalysis A: General, Vol. 87, pp 185-203.
NBR 14903 (2002) Gás Natural – Determinação da composição por cromatografia gasosa.
Associação Brasileira de Normas Técnicas.
Newsome, David S. (1980). The water-gas shift reaction. Catal. Rev. Sci. Eng., 21(2), 275-318.
Singh, C. C. P. and Saraf, D. N. (1977). Simulation of high-temperature water-gas shift Reactors.
Ind. Eng. Chem. Process Des. Dev., Vol. 16, No. 3, 313-319.
Smith, Byron R. J. et al. (2010). A review of the water gas shift reaction kinetics. International
Journal of Chemical Reactor Engineering Vol. 8, Review R4, ISSN 1542-6580.

www.intechopen.com
 Petrochemicals
Edited by Dr Vivek Patel

ISBN 978-953-51-0411-7
Hard cover, 318 pages
Publisher InTech
Published online 28, March, 2012
Published in print edition March, 2012

The petrochemical industry is an important constituent in our pursuit of economic growth, employment
generation and basic needs. It is a huge field that encompasses many commercial chemicals and polymers.
This book is designed to help the reader, particularly students and researchers of petroleum science and
engineering, understand the mechanics and techniques. The selection of topics addressed and the examples,
tables and graphs used to illustrate them are governed, to a large extent, by the fact that this book is aimed
primarily at the petroleum science and engineering technologist. This book is must-read material for students,
engineers, and researchers working in the petrochemical and petroleum area. It gives a valuable and cost-
effective insight into the relevant mechanisms and chemical reactions. The book aims to be concise, self-
explanatory and informative.

How to reference
In order to correctly reference this scholarly work, feel free to copy and paste the following:

Douglas Falleiros Barbosa Lima, Fernando Ademar Zanella, Marcelo Kaminski Lenzi and Papa Matar Ndiaye
(2012). Modeling and Simulation of Water Gas Shift Reactor: An Industrial Case, Petrochemicals, Dr Vivek
Patel (Ed.), ISBN: 978-953-51-0411-7, InTech, Available from:
http://www.intechopen.com/books/petrochemicals/modeling-and-simulation-of-water-gas-shift-reactors-an-
industrial-case

InTech Europe InTech China

University Campus STeP Ri Unit 405, Office Block, Hotel Equatorial Shanghai
Slavka Krautzeka 83/A No.65, Yan An Road (West), Shanghai, 200040, China
51000 Rijeka, Croatia
Phone: +385 (51) 770 447 Phone: +86-21-62489820
Fax: +385 (51) 686 166 Fax: +86-21-62489821
www.intechopen.com