Received: 20 November 2020 Revised: 23 April 2021 Accepted: 8 May 2021

DOI: 10.1002/er.6886

RESEARCH ARTICLE

A highly carbon-efficient and techno-economically

optimized process for the renewable-assisted synthesis of
gas to liquid fuels, ammonia, and urea products

Mohammad Ziaei1 | Mehdi Panahi1 | Mohammad Ali Fanaei1 |

1,2 3
Ahmad Rafiee | Kaveh Khalilpour

1
Chemical Engineering Department,
Faculty of Engineering, Ferdowsi Summary
University of Mashhad, Mashhad, Iran Carbon dioxide conversion into beneficial products has received very much
2
Department of Theoretical Foundations attention in recent years to decrease industrial CO2 emissions. In this context,
of Electrical Engineering, Faculty of
integration of gas to liquids (GTL) process with an iron-based Fischer-Tropsch
Energy, South Ural State University,
Chelyabinsk, Russian Federation (FT) reactor with ammonia and urea synthesis plants was investigated. The
3
Faculty of Engineering and IT, main motivation of the proposed integration is to reuse a released CO2 stream
University of Technology Sydney, Ultimo, from the GTL process and to enhance the commercial process economy. The
New South Wales, Australia
required hydrogen for ammonia comes from polymer electrolyte membrane
Correspondence (PEM) electrolyzers running by solar power. Latin hypercube design (LHD)
Mehdi Panahi, Chemical Engineering
approach was applied to model the profitability and carbon efficiency of the
Department, Faculty of Engineering,
Ferdowsi University of Mashhad, process. Optimization was conducted to maximize the carbon efficiency and
Mashhad, Iran. profit index of the overall process using the model-based calibration (MBC)
Email: mehdi.panahi@um.ac.ir
toolbox of MATLAB. The results demonstrated that at the optimum case, the
proposed integration is capable of producing 48 t/h of urea and also utilizing
about 35 t/h of CO2 produced in the GTL process. The results were compared
with another configuration in which a cobalt-based FT reactor was integrated
with ammonia and urea processes. The results suggest that profitability, car-
bon efficiency, and urea production of the process configuration with a Co-
based FT reactor is higher than the iron-based configuration while the wax
production rate of the iron-based configuration is higher than that of the Co-
based process. Techno-economic feasibility study of the zero CO2 emission pro-
cess represents that the carbon efficiency of around 100% could be obtained.

KEYWORDS
carbon efficiency, CO2 utilization, computer experiment, GTL, urea, zero CO2 emission

1 | INTRODUCTION direct impact of CO2 emissions on global warming

which is resulting in drastic problems such as rising sea
One of the most worrying environmental subjects for levels and forest fires.2 In 2020, according to the reports,
the human of the 21st century is the greenhouse gas the average concentration level of the CO2 was raised to
(GHG) emissions and the phenomenon of the global 412.4 ppm with an average increase of circa 2.5 ppm per
warming.1 There has been growing consensus on the year since 2015.3

16362 © 2021 John Wiley & Sons Ltd. wileyonlinelibrary.com/journal/er Int J Energy Res. 2021;45:16362–16382.
ZIAEI ET AL. 16363

There are three approaches to overcome the irrepara- analysis. It is deduced from the sensitivity analysis that
ble troubles of the CO2 emission: substituting CO2 emit- the electricity and the capital investment costs of the elec-
ting fossil fuels by renewable sources, carbon capturing trolyzer are the most significant terms in the methanol
and storing underground (CCS), and carbon production economy. From the CO2 utilization point of
capturing and utilization (CCU).4 The vented CO2 is cap- view, their results show that circa 5800 t/y of CO2 could
tured and stored underground in the CCS processes, be reused in the proposed integrated process and saved
while in the CCU approaches, the emitted CO2 is cap- about 5.2 Nm3/y of natural gas. In a research study con-
tured and utilized to produce value-added chemicals. The ducted by Fern
andez-Dacosta et al,21 a refinery and a
drawback of CCS is the extra cost that is imposed on the recovered CO2 stream were integrated with DME and
plant. The idea of using CCU as a technique to reuse the polyols production units. Two different process configu-
emitted CO2 for the manufacturing of valuable products rations were discussed including cascade and parallel. In
was applied in the 1970s for the first time.5 The main comparison with the base-case without CO2 recovering
advantage of CO2 utilization is that, CO2 is known as a unit, results show that the cost of process and fuel con-
noteworthy inexpensive source of carbon and even nega- sumption decreased by 9% and 10% respectively.
tive in price when the carbon tax is considered.6 For fur- Abdelaziz and co-workers22 have constructed three inno-
ther reading about likenesses and discrepancies between vative designs which utilize the captured CO2 for the pro-
CCU and CCS see.7 duction of methanol. The environmental and economic
By increasing global warming concerns, more heed considerations and heat integration were evaluated for
has been enticed to uncommon sources of carbon. The three proposed configurations. In another study, a
exploitation of the recovered CO2 in the production of techno-economic study of the integration of a CCS and a -
chemicals is an important research topic in the academia power-to-methanol process with a coal power plant was
and industry communities. This could enhance the profit- accomplished by Bellotti et al.23 The study was performed
ability of the carbon capture process.8-10 CO2 is utilized in three phases as follows:
for the manufacturing of inorganic and organic com-
pounds such as sodium bicarbonate11 and salicylic acid. 1. The influence of the CCS installation and power to
CO2 is also used in the production of synthetic fuels12,13 methanol plant integration on the coal power plant
(like methane, methanol14,15 and dimethyl ether performances was surveyed on four different
(DME)4,16). It can also be used directly in applications scenarios.
comprising welding medium, firefighting equipment, sol- 2. The yearly electricity cost for various scenarios was
vent, refrigerant, dry ice, and process fluid.17,18 analyzed.
In our recently published work, the integration of a 3. Economic analysis for the methanol synthesis process
cobalt-based FT-GTL plant with ammonia and urea pro- was accomplished.23
cess was carried out to utilize the CO2 originated from
the GTL process for urea production.19 Hydrogen and The production of methane and ammonia was presented
nitrogen (the raw materials of the ammonia process) by Castellani and co-workers, where the flue gas was
were produced via polymer electrolyte membrane (PEM) chosen as the nitrogen source and CO2 and also hydrogen
electrolyzer and air separation unit (ASU) of the GTL was provided by electrolysis of water from renewable
process, respectively. The energy demand of the PEM sources of energy.24
electrolyzer was provided from the excess generated Bose et al25 investigated the co-production of power
power of the GTL plant. The multi-objective optimization and urea by integration of a coal-fired power plant with a
was carried out to maximize the wax production rate and urea synthesis process to utilize CO2. The feed rate to the
carbon efficiency. The results suggested that the pres- plant is 46 t/h. The economic benefit of 5% and 10% were
ented integration could produce about 434 000 t/year achieved by the shift reaction conversion of 95% and 90%
urea in addition to GTL products and could also utilize respectively. Zhang and colleagues26 considered two pro-
37 t/h of CO2 for urea production.19 The economic feasi- cess structures where CO2 was added to the conventional
bility of a photovoltaic (PV) power plant integrated with methanol process. They combined steam methane
the 5 MW electrolyzer-base methanol production unit reforming (SMR) with dry methane reforming (DMR) to
was survived by Belloti et al20 using response surface produce synthesis gas (syngas). Both processes were sim-
methodology. The major three design variables were ulated in Aspen Plus. The major difference between the
presented including the size of the PV plant, electrical two configurations is the feeding point of CO2. In the first
power cost, and the reduction percentage in capital cost configuration, CO2 is combined with natural gas, and in
of the electrolyzer. For evaluation of these variables, the the second configuration, CO2 is appended to the metha-
RSM approach was applied to accomplish the sensitivity nol reactor. The results showed mitigation in CO2
16364 ZIAEI ET AL.

emission and an enhancement in the energy efficiency of 12 000 bbl/day of GTL products and 576 t/day of ammo-
the proposed processes in comparison to the conven- nia. As derived from the results, the profitability of the
tional one.26 In 2019, Zhang et al27 executed the concep- integrated process configuration was about 1.5 times
tual design of a power-to-liquids (PTL) process and more than the conventional GTL process. Nguyen et al34
hybrid configuration of PTL with a power-to-gas (PTG) proposed three routes to alleviate CO2 emission compris-
process using Aspen Plus software. According to the ing hydrogenation, bi-reforming, and tri-reforming to
techno-economic studies, it was deduced from the results produce 300, 1500, and 3500 t methanol per day. They
that both processes can be considered as candidate tech- perused annualized capital investment and operational
niques in utilizing CO2 for the production of valuable costs and carried out the comparison among different
products. Furthermore, it is deduced from their results presented routes. They concluded that the methanol pro-
that the hybrid configuration is more efficient and profit- duction from hydrogenation with renewable hydrogen
able. The integration of the sequestered CO2 from a sources is a near-zero CO2 emission process. On the other
cement plant with the methanol production unit was pro- hand, two other presented routes have lower total annu-
posed by Meunier and colleagues.28 As concluded from alized cost in comparison with the hydrogenation route.34
their optimization results, circa 103 t/h of CO2 could be The production of dimethyl carbonate (DMC) was inves-
utilized for the production of 64.5 t/h of methanol. tigated in 2019 by Sanchez and colleagues.35 Further-
Pérez-Fortes et al proposed two CO2 utilization con- more, urea production was considered for comparison of
figurations including urea and methanol production. The the processes from a techno-economic point of view.
requirements of utility, process efficiency, CO2 utiliza- Based on the results of their optimization, the production
tion, and investment cost evaluation were appraised in of DMC is desirable with consideration of urea produc-
their study.29 In 2017, the integration of a coal-fired tion. For a production capacity of 342 t/d, the DMC
power plant and a CO2 recovering unit with a GTL plant investment cost is about 91 MM€ and the production cost
were investigated by Rafiee and co-workers.30 In their of €520/tonne is proposed. From the result of the sensitiv-
procedure, a post-combustion CO2 capture (PCC) process ity analysis, it is obvious that the price of methanol plays
is augmented to a 300 MW power plant to capture and a significant role in the DMC production cost.35 Table 1
purify CO2. The captured CO2 stream is fed to the provides a brief review of the CO2 utilization processes.
reformer of the FT-GTL process. The possibility of CO2 According to the science of catalysts, the catalysts
utilization is perused for three various case studies. function in the FT reactor depends on hydrogen adsorp-
Panahi and co-workers continued the same work and tion capacity and desirable CO hydrogenation activity.36
appointed multi-objective optimization (MOO) to the FT- The FT catalysts with sufficient CO hydrogenation activ-
GTL process for CO2 utilization with staged FT reactor. ity includes Ni-, Co-, Fe- and Ru-based catalysts.37
Three objective functions were proffered, where the Among them, Co- and Fe- based catalysts are broadly
results showed that the production of wax and carbon used in commercial FT-GTL processes. Fe compounds
efficiency could be increased by the staging of the FT are highly abundant in nature, attractive from an eco-
reactor.31 In 2018, Fazeli and co-workers have investi- nomical point of view, and have low selectivity to paraf-
gated the potential of CO2 utilization in an iron-based fins. On the other hand, Co has a low selectivity to
FT-GTL process with two distinct syngas production olefins and high selectivity to paraffins. Co-based cata-
technologies, that is, auto thermal reformer (ATR) and lysts show more resistance to deactivation in comparison
SMR.32 The enhancement of carbon efficiency as well as to Fe-based catalysts.38 Water gas shift (WGS) is active on
wax production rate in two detached conditions were an iron-based catalyst so a fraction of the CO present in
specified. Based on the results of their research, the opti- the syngas feed to the FT reactor is converted to CO2
mal carbon efficiency of about 70% was reported for both through this reaction.
cases and the CO2 intakes of 0 and 166.4 t/h and wax pro- In this paper, the integration of the ammonia and
duction rates of 68.17 and 101.4 t/h were illustrated for urea plants with an iron-based FT-GTL process is sur-
ATR and SMR routes, respectively. veyed. The raw materials of the ammonia plant, nitrogen,
In 2018, Ostadi et al33 integrated a GTL process and and hydrogen are grabbed from the air separation unit
ammonia synthesis unit, in which about 99% pure hydro- (ASU) of the GTL process and PEM electrolyzer respec-
gen is provided by a heat exchange reformer (HGR) and tively. The carbon efficiency and the overall profit index
the nitrogen was taken from an ASU. The proposed inte- of the integrated process were maximized separately. The
grated configuration produces sufficient power and water two mentioned objective functions were specified using
and also there was no necessity for a separate CO2 recov- space-filling design by suggesting five independent vari-
ering unit and water gas shift (WGS) reactor. The pro- ables. Space-filling design is a set of mathematical and
posed configuration was capable to produce circa statistical approach for computer experiment model
ZIAEI ET AL. 16365

TABLE 1 Review of CO2 utilization pathways

CO2 utilization
Reference source Product(s) Highlight(s)
19 Natural gas GTL products, urea Co-based FT-GTL plant integrated with urea-ammonia processes, response
surface methodology applied to model carbon efficiency and profitability
20,23 Coal Methanol Integration of photo voltaic (PV) power plant with PEM electrolyzer
21 Natural gas DME, polyols Recovering CO2 from refinery and integrating with DME, polyols
production plants
22 Natural gas Methanol Economic and environmental study of three novel methanol production
processes from captured CO2
24 Captured CO2 Methane, ammonia The process is a viable CO2 recycling when renewable excess energy is
from flue gas used
25 Coal Power, urea Coal power plant integrated with urea production process
26 Natural gas methanol Adding CO2 to conventional methanol production plant using various
reformer types
27 Captured CO2 Synthetic natural gas Profitability analysis of hybrid configuration of PTL and PTG with CO2
(SNG), liquid fuels utilization
28 Natural gas Cement, methanol Co-production of methanol and cement using captured CO2 from acement
plant
29 Natural gas Methanol, urea Integration of methanol and urea processes
30 Coal GTL products GTL process integrated with coal-fired power plant
33 Natural gas GTL products, Co-production of ammonia and GTL process using heat exchanger
ammonia reformer
34 Captured CO2 Methanol Production of methanol using three different ways containing
from flue gas hydrogenation, bi-reforming and tri-reforming
Present Natural gas GTL products, urea Modeling and optimization of carbon efficiency and profitability of
work integrated GTL and urea-ammonia processes using Latin hypercube
design, zero-emission feasibility study

construction. The main difference between the presented FT reactors in parallel. The FT reactors' volume is equal
study and the previous work19 is the reactions mecha- to 2000 m339 (In this study the upgrading unit has not
nism taking place in the FT reactor leading to different been studied). In the base case of the propounded inte-
product distribution, process carbon efficiency, natural gration, the urea and ammonia production rates are in
gas intake to the fired heater, and CO2 emission to the turn 1503 kmol/h (25.6 t/h) and 765 kmol/h (45.9 t/h).
air. These owe to the fact that water gas shift is negligible The required hydrogen for ammonia comes from the
when cobalt-based catalysts are used in the FT reactor. In PEM electrolyzer. The PEM power demand is supplied by
this research, techno-economic analysis of a zero CO2 surplus produced power from the GTL process. The pro-
emission integrated iron-based FT GTL process with duced oxygen in the PEM electrolyzer is utilized in one
ammonia, and urea is presented. Solar photovoltaic cells auto thermal reformer (ATR) to reduce the size of one of
are used for power production, and the profitability com- the air separation units (ASU) and consequently the total
parison of the optimized overall flowsheet with the Co- cost of investment. In the proposed configuration, the
based integrated GTL process is discussed. feed rate of natural gas which is evenly distributed
between the two trains of the GTL plant is 16 390 kmol/h
(276 t/h). The natural gas composition was taken from
2 | P R O C E S S DE S C R I P T I O N the study of Panahi et al40 (CH4: 95.5%, C2H6: 3%, C3H8:
0.5%, n-C4H10: 0.4%, and N2: 0.6%, all in molar basis).
The integrated GTL process with ammonia and urea pro- The natural gas feed is preheated up to 455
C in a fired
cess layout is illustrated in Figure 1. In the presented heater and is conveyed to the pre-reformer for converting
integration, the GTL plant with the same capacity as the of heavier hydrocarbons than methane to syngas.
Sasol Oryx GTL process was simulated. As shown in In a GTL process, the fired heater is utilized for sup-
Figure 1, there are two similar trains with two iron-based plying the following energy requisitions19,40:
16366 ZIAEI ET AL.

F I G U R E 1 Simple flow diagram of

the integrated iron-based FT-GTL
process with ammonia and urea
synthesis [Colour figure can be viewed
at wileyonlinelibrary.com]

• Heating the oxygen inlet stream of the ATR to 200
C, Due to the high temperature of ATR effluent, the
• Heating the inlet streams of the pre-reformer to 455
C, ATR reactions were supposed to reach equilibrium.19,44
• Heating the outlet of the pre-reformer to 675
C before The effluent of the pre-reformer should be heated up to
entering the ATR, 675
C before entering to the ATR. The ATR outlet is
• Preparing the superheated steam for the ASUs and fixed at 1030
C by adjusting the flowrate of oxygen to the
recycled tail gas compressors, ATR.45 In order to separate the water content of the syn-
gas stream, the ATR exit stream was cooled to 255
C by
An excess 10% of the total considered power of the crossing through a heat exchanger. The heat of this hot
fired heater was propounded to guarantee the energy stream is used to procreate medium pressure (MP) steam
supplement for other power-demanding equipment, and then it is cooled down to 38
C.
which are not included in the flowsheet. Before entering the fresh syngas into the FT section, a
Syngas mostly subtends hydrogen and carbon monox- CO2 removal unit was located to capture CO2 from the
ide with a scant amount of other gases for example car- syngas stream. The main reactions that occur in the iron-
bon dioxide, methane etc. The syngas generation unit based FT reactor are illustrated in reactions (4) to (6).32
transforms natural gas to mainly H2 and CO. There are
numerous paths for synthesis gas production comprising CO þ 3H2 ! CH4 þ H2 O, ð4Þ
SMR, DMR, partial oxidation (POX), ATR, bi-reforming
(BR), tri-reforming (TR), and combined reforming nCO þ 2nH2 ! ðCH2 Þn þ nH2 O, ð5Þ
(CR).32,41-43 In a GTL plant, the most capital-intensive
unit is the syngas production unit in comparison to other CO þ H2 O ! H2 þ CO2 : ð6Þ
units. The presence of oxygen is necessary for ATR and
POX processes, in which a costly ASU is required. In this
study, the ATR is chosen as syngas production technol- Equation (5) denotes olefins and paraffins production.
ogy for the following reasons: The weight fraction of the FT products heavier than
methane follows the reputable Anderson-Schulz-Flory
1. Having H2/CO ratio of about 2, which is a near- (ASF) distribution40,46-48:
optimal condition for the GTL process;
2. The produced nitrogen in ASU is used as the feed- wn ¼ nð1  αÞ2 αn1 , ð7Þ
stock of the ammonia plant.
where n is the carbon atoms number and is larger than
The substantial reactions taking place in the ATR reactor 1, and α is chain growth probability. This concept is
are represented in reactions (1) to (3).19 depicted in Figure 2.32
The kinetic reaction rate and α correlation equation
CH4 þ 1:5 O2 ⇌ CO þ 2H2 O ΔH
298 ðkJ=molÞ ¼ 520, ð1Þ were taken from the work of Fazeli and co-workers,32
Equation (8) and (9).
CH4 þ H2 O ⇌ CO þ 3H2 ΔH
298 ðkJ=molÞ ¼ 206, ð2Þ
9:76  103 P0:5

H2 P CO
r FT ¼ kmol=m3
reactor :s
, ð8Þ
CO þ H2 O ⇌ CO2 þ H2 ΔH
298 ðkJ=molÞ ¼ 41: ð3Þ ð1 þ 0:563 PCO þ 4:5 PH2 O Þ 2
ZIAEI ET AL. 16367

The details of ammonia and urea simulation are pres-

ented in our recently published work.19
Methane, ethane, LPG including propane and butane
components, gasoline/naphtha (C5-C11), diesel (C12-C20),
wax (C21+), and water are the products of the GTL pro-
cess, while ammonia and urea are the products of the
ammonia and urea plants.

3 | BASE-C ASE C ONFIGURATION

The base-case process configuration of the integrated pro-

cess is depicted in Figure 3. The iron-based GTL process
layout is taken from the work of Fazeli and co-workers.32
The capacity of the simulated urea and ammonia plants
are 25.6 and 45.9 t/h, respectively. In the GTL process
flow diagram, the FT reactor is functioning at 210
C and
2259 kPa with an iron-based catalyst. The ratio of the H2/
CO inside the FT reactor is about 1.065, where this ratio
F I G U R E 2 Chain growth to different paraffins probability30
is 1.107 at the outlet of the ATR reformer. The conver-
[Colour figure can be viewed at wileyonlinelibrary.com]
sions of H2 and CO in the FT reactor are 47.32% and
44.08%, respectively. The total mass flowrate of FT-
P H2 O 5 PH2 PCO2
5:84  103 PCO  6:8057  10 P2CO

 derived products is demonstrated in Table 2.
r WGS ¼
2 kmol=m3
reactor :s
: The flowrate of the tail gas is 44 650.1 kmol/h with
P P
1 þ 1:4249 H2 O
PCO þ 6:14 H2 O
PCO the molar composition of H2 (31.32%), CO (31.19%), CO2
ð9Þ (28.58%), CH4 (3.31%), N2 (2.23%), and other light hydro-
carbons (3.37%). About 4.9% of the tail gas is purged and
For more comprehensive reviews of GTL process simu- the unpurged tail gas is divided between the FT reactor
lation, see References 30-32,40,49. Based on the integrated and the pre-reformer with the proportion of 77.45% and
process, the nitrogen requirement for ammonia produc- 22.55%, respectively. The total extent of oxygen demand
tion was taken from the ASU by-product stream. Hydro- for the ATR is circa 6074 kmol/h, of which for one train,
gen, the second feedstock of the ammonia process, was about 20% of the oxygen demand is supplied by the PEM
prepared by the PEM electrolyzer modules. For every electrolyzer and the rest is supplied by the ASU. This will
1 MW power to the PEM electrolyzer, the amount of pro- eventuate in size reduction of the ASU and less make-up
duced oxygen and hydrogen are about 19 kg/h and natural gas consumption for one train of the GTL pro-
151 kg/h, respectively.23,50 As can be seen in Equation (10), cess. The excess produced power in the two trains of the
for every three moles of hydrogen, two moles of ammonia GTL process is about 253.2 MW, which can provide the
are procreated33: energy demand of the PEM electrolyzer for the produc-
tion of about 2405 kmol/h of hydrogen. The produced
3H2 þ N2 ! 2NH3 ΔH
298 ðkJ=molÞ ¼ 91:8: ð10Þ hydrogen is utilized as the feed of the ammonia synthesis
plant. By producing this amount of hydrogen, about
The produced ammonia was mixed with the captured 1503 kmol/h of ammonia is obtained from the ammonia
CO2 from the GTL process to produce urea pursuant to plant, which can be used as the feedstock of the urea syn-
the following reactions35: thesis plant. About 801 kmol/h of the recovered CO2
from the CO2 recovery unit of the GTL process is
2NH3 þ CO2 ⇌ H2 N  COONH4 ΔH
298 ðkJ=molÞ exploited in the urea plant, which for each ton of cap-
¼ 117 ðCarbamate formationÞ, ð11Þ tured CO2, circa 0.19 MWh of electrical power is used.51
About 1940 kmol/h of make-up natural gas is used as an
H2 N  COONH4 ⇌ ðNH2 Þ2 CO þ H2 O ΔH
298 ðkJ=molÞ energy supplier for compression processes and fuel in the
¼ 16 ðUrea conversionÞ: fired heater (for two trains). The carbon efficiency (the
ð12Þ moles of carbon in the products stream divided by the
total number of carbons entering the process including
16368 ZIAEI ET AL.

FIGURE 3 Base-case of the integrated process [Colour figure can be viewed at wileyonlinelibrary.com]

TABLE 2 FT-derived products from two trains of GTL process deterministic model. Computer simulations can be very
FT-derived products tonnes/h
complex including many variables with obscurant inter-
relationships. The deterministic character of the com-
LPG 3.2
puter experiments proposed here brings a paradigm shift
Diesel 64.6 in the design of experiment and analysis routes in com-
Gasoline 71.4 parison to those in the physical experiments. The main
Wax 40.1 differences between physical experiment and computer
simulation experiment are summarized below52:

the moles in the make-up natural gas) of the standalone 1. there is no need for replication, randomization, and
GTL process is 67.02% and for the integrated process, this blocking,
amount is 71.65%. In this integration, about 14.8% of the 2. due to lack of random error, a large experimental
generated CO2 in the FT-GTL process is consumed in the region can be explored in computer experiment,
urea plant. 3. simplicity in changing levels of degrees of freedom.

The purpose of the design of the experiment on these sys-

4 | R ES U L T S A N D D I S C U S S I O N tems is to define a simple empirical model that suffi-
ciently predicts the behavior of the system over limited
4.1 | Model regression ranges of the degrees of freedom. In physical experi-
ments, a few levels were chosen for different degrees of
Computer experiment is one of the most efficacious pro- freedom, and then using factorial structure, the experi-
cedures that can be applied to deal with the industrial mental designs were created for a specified number of
challenges to enhance productivity in various scopes. runs. Given that there is no cost for variation in degrees
Space-filling designs are convenient for modeling deter- of freedom in a computer experiment, the design of the
ministic systems. The major example of a deterministic experiment can completely differ from the ones in
system is a computer simulation. Although numerical the physical experiment. In a computer experiment, the
variances are available in the solvers used in a computer design points could be scattered in the experimental
simulation, for all applied goals, those errors can be region with ingenious techniques without worrying about
neglected and the computer model can be known as a the number of the generated levels. This propels to the
ZIAEI ET AL. 16369

concept of space-filling designs. In experiments on sys-

tems where there is considerable random noise, the
objective is to minimize the variance of prediction. In
computer experiment systems, there is a bias instead of
variance. Bias is defined as the difference between the
estimation model and the true mathematical function.
The purpose of space-filling designs is to bound the
bias.52 There are two approaches to bound the bias.
The first route is to distribute the design points as far
from each other as possible with staying inside the empir-
ical boundaries. The other approach is to spread the
points out equally over the experimental region. Space-
filling designs involve Latin Hypercube Designs (LHD)
and their revisions. An LHD with n runs and d factors is
shown as L(n, d), where Lij signifies the level factor of j in
the ith experimental run. In computer simulations, the
input spaces are usually bounded hyper-rectangles and
can be converted to unit hypercubes. In an LHD, the
levels are scattered equally from the lower bound to
the upper bound of each factor. The Latin hypercube
method selects points for maximizing the minimum F I G U R E 4 Example of Latin hyper cube design46 [Colour
interval between design points with considering a con- figure can be viewed at wileyonlinelibrary.com]
straint. The constraint retains the equal spacing between
factor levels.53 For representing the Latin hypercube, the
experimental region of unit square [0, 1]2 is supposed. To variables of X1, …, Xp, the π j are independent uniform
ðiÞ
obtain a design comprising n points, each axis is divided random permutations of the integers 1, …, n and ξj are
into n evenly spaced distances. This divides the unit independent U[0, 1] random numbers independent of
square into n2 cells of identical size. These cells occupied the π j.
with integers 1, 2,…, n and constructs a Latin square. In a LHD approach was applied to conduct the computer
Latin square, each integer becomes visible just once in experiment. The design of experiment was accomplished
each column and in each row of the grid. The purpose of by using JMP software with space-filling design (LHD
the sampling stage is to procreate a matrix Xn = (x[i]j)i=1… approach).
n,j=1…p, where n and p denote the number of experiments Five independent variables were chosen as follows:
and is the number of variables, respectively.
LHD warrants that all input variables cover the whole A. Unpurged ratio (known as 1- purged ratio). Because
defined bounds. The bounds of the variables were nomi- of convergence issues of the flowsheet, this value is
nated as Xj, j = 1,…, p, and they were divided into kept below 99%.
n evenly probable intervals. The n-sample of the variable B. The split ratio of the recycled tail gas stream to the FT
Xj is noted as Xnj. An LHD of size n is gained from a ran- reactor;
dom selection of n values for each Xj. The ith line of the C. CO2 recovery percentage from fresh syngas;
n  p matrix includes p inputs that correspond to D. Steam/carbon ratio to the pre-reformer in syngas sec-
the execution of ith code. Figure 4 demonstrates 10 sam- tion, which should be kept above 0.3. This is to
ples of two typical random variables obtained with ensure soot-free operation40;
LHD.54 E. CO2 intake to the ATR.
From Mathematical point of view, the LHD of the ith
sample for the jth variable can be formulated as54: For more information about the selection of the degrees
of freedom, see the Data S1.
ðiÞ ðiÞ
!
π j  ξj According to our previous work19 and flowsheet con-
ðiÞ
xj ¼ F 1
j , ð13Þ vergence considerations, the following boundary condi-
n
tions were applied to the degrees of freedom for design of
the experiment (Table 3).
where Fj ( j = 1, …, p) are the invertible continuous distri- A total number of 50 runs were provided by JMP soft-
bution functions with mutually independent random ware to perform the design of experiment. The level of
16370 ZIAEI ET AL.

each independent variable and the resultant carbon effi- tion of various correlations between the responses and the
ciency is shown in Table S1 of Data S1. input variables can be initiated. The statistical MBC model
After defining the codes and attaining the test program, fitting toolbox of MATLAB software was used to investigate
the process simulations were done and responses were log- the regression analysis of data from process simulation.
ged to develop the correlation between the variables and the Based on the analysis results gained from the MATLAB
objective functions. The adopted procedure for the present MBC toolbox, it was deduced that models 1 and 2 are the
study is to maximize carbon efficiency and an economic most adequate models to predict the carbon efficiency and
objective function. Thereby, the carbon efficiency and profit profit index (PI) of the plant, respectively. The models' defini-
index (which will be explained in the next sections) were tion terms are presented in Table 4.
appointed as the responses of the experiments. A combina- The statistical parameters of the regressed models for
carbon efficiency and PI are illustrated in Table 5.
The RMSE is the root mean square of the errors. The
TABLE 3 Boundary conditions of degrees of freedom
RMSE demonstrates how nearby the data points taken
from the simulations are to the predicted values from the
Independent variable Lower bound Upper bound model. Whereas RMSE and R2 are the absolute and rela-
A (Unpurged ratio) 0.95 0.98 tive measure of fit, respectively. A good regression has
B (Recycle to FT) 0.75 0.85 lower RMSE values and R2 values close to unity.
C (CO2 removal) 0.93 0.99 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn
D (Steam/carbon) 0.4 0.6 i¼1 ð^ yi  yi Þ2
RMSE ¼ : ð14Þ
E (CO2 intake) 160 000 205 000 n

TABLE 4 Fitted models definition

1- Carbon efficiency presented model 2- Profit index (PI) presented model

Model for VAR Model for VAR

Input Scaling Input Scaling

A = VAR_1:[0.95,0.98] ! [1,1] A = VAR_1:[0.95,0.98] ! [1,1]
B = VAR_2:[0.75,0.85] ! [1,1] B = VAR_2:[0.75,0.85] ! [1,1]
C = VAR_3:[0.93,0.99] ! [1,1] C = VAR_3:[0.93,0.99] ! [1,1]
D = VAR_4:[0.4,0.6] ! [1,1] D = VAR_4:[0.4,0.6] ! [1,1]
E = VAR_5:[1.6e+5,2.05e5] ! [1,1] E = VAR_5:[1.6e5,2.05e5] ! [1,1]
Quadratic-RBF-Medium Gaussian Process Model (ARD Squared Exponential, Constant)
0.71807 + 0.0070478*VAR_1.0.00085466* VAR _2 + 0.0033271* VAR _3- Kernel: ARD-Squared Exponential
0.010634* VAR _4 + 0.0040955* VAR _5-0.0038988* VAR _1 + 2
Basis: Constant
0.0013882* VAR _1* VAR _2 + 0.00020833* VAR _1* VAR _3+ LenghthScal1: 2.25269
0.00168483* VAR _1* VAR _4+
0.00028125* VAR _1* VAR _5–0.0037742* VAR _22 -0.00079258* VAR LenghthScal2: 4.53677
_2* VAR _3-
0.0015635* VAR _2* VAR _4 + 0.00098543* VAR _2* VAR _5–0.003062 LenghthScal3: 29.1477
* VAR _32-0.0011963* VAR _3* VAR _4+
0.0017253* VAR _3* VAR _5-0.0013025 VAR _42-0.0020318* VAR _4* LenghthScal4: 2.37421
VAR _5-0.001121* VAR _52
LenghthScal5: 1.83654
Plus σ f : 0.06837
A radial basis function network using a multiquadric kernel with 7 σ n : 0.00516612 where σ f and σ n are signal SD and length scale
centers and a global width of 0.22376. of each predictor, respectively.
The regularization parameter, lambda, is 0.0014737.
ZIAEI ET AL. 16371

Predicted residual error sum of squares (PRESS) RMSE Figures 5 and 6 indicate the normal probability of
provide a summary measure of the fit of a model to a the studentized residuals of the objective functions. As
sample of observations that are not themselves used to illustrated in these figures, small deviations from the
estimate the model. From the small values of PRESS straight line of normal probability validates that the
RMSE, it can be concluded that the model can predict presented models are adequate. In Figures 7 and 8,
other points with good accuracy.55 near-zero values of residuals imply the prediction is
As evident from Table 5, the low values of RMSE and exactly correct and the structureless pattern of residuals
predicted RMSE show that the regression models are against predicted value also confirms the model's
adequate. adequacy.

TABLE 5 Statistical parameters of carbon efficiency and PI model regression

PRESS RMSE RMSE R2 Adjusted R2 Validation RMSE

3 4
Carbon efficiency 1.519  10 6.376  10 0.997 0.994 4.745  104
PI 7.192  103 3.898  103 0.991 - 6.385  103

FIGURE 5 Studentized residuals probability plot of carbon efficiency [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 6 Studentized residuals probability plot of PI [Colour figure can be viewed at wileyonlinelibrary.com]
16372 ZIAEI ET AL.

F I G U R E 7 Studentized
residuals probability plot of
carbon efficiency [Colour figure
can be viewed at
wileyonlinelibrary.com]

F I G U R E 8 Residuals vs the
predicted value of PI [Colour figure can
be viewed at wileyonlinelibrary.com]

The comparison between predicted and simulation The common correlation between the obtained cost and
results of the response values of objective functions is capacity of an equipment is given by
depicted in Figures 9 and 10. These figures illustrate a
 
good agreement between the predicted and simulation TCI 2 Capacity 2 n
¼ , ð16Þ
results. The green triangles, which overlap on the y = x TCI 1 Capacity 1
line in Figures 9 and 10, demonstrate that the presented
models predict the carbon efficiency and PI appropriately. where, n is the cost exponent (scaling factor), which var-
ies on different processes or equipment.
The principal units of a GTL plant include the auto-
4.2 | Profitability analysis thermal reforming section, the air separation unit, and
the FT-reactor. Owing to the attendance of the PEM elec-
4.2.1 | Analysis of total capital trolyzer in the integrated plant, ASU size decreased by
Investment (TCI) circa 20%. For one train of the base-case flowsheet, the
TCI of the ASU in the standalone GTL configuration is
The TCIs of the integrated process and individual pro- about 123.23 million USD. This amount is reduced to
cesses are compared at the same capacity in this subsec- 105.6 million USD in the integrated process layout (the
tion. All the TCIs applied in this study were brought to scaling factor of 0.75 was applied for cost estimation of
2018 using Chemical Engineering Plant Cost Index different capacities).56 The flowrate of oxygen from the
(CEPCI), Equation (15): ASU to the ATR of the standalone process and the inte-
grated process are 194.4 and 155.9 t/h, respectively. In
CPECI2018 the integrated process, the TCI reduction of the ASU
TCI2018 ¼ TCIbase year : ð15Þ
CPECIbase year section of one train of the GTL process compared to the
ZIAEI ET AL. 16373

FIGURE 9 Simulation results and predicted values of carbon efficiency [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 10 Simulation results and predicted values of PI [Colour figure can be viewed at wileyonlinelibrary.com]

base-case is 14.31%. The extractable power from medium the above explanation, the TCI for the ammonia-urea
pressure (MP) steam of the GTL plant in the base-case plant in the integrated plant is about 416.39 million USD.
configuration is 253.2 MW. The TCI of a power plant The simple process information of the integrated pro-
with this capacity is circa 246.1 million USD.57 For the cess and standalone processes are showed in Table 6. As
PEM electrolyzer, the TCI of 165.6 million USD and scal- evident from Table 6, the total fixed capital investment of
ing factor (ie, 0.85) were taken from References 23,50. the integrated plant is 58 million USD less than the
According to Reference 58 it is concluded that the TCI of standalone processes. In addition, the integrated process
the PEM electrolyzer could be decreased to 50 million uses less natural gas and releases less CO2 to the atmo-
USD at the time prospect of 30 years. sphere. The values of total capital cost of the integrated
The pivotal units for syngas preparation in a conven- and standalone processes are mentioned in Table 7.
tional urea-ammonia process comprise natural gas pre-
treatment and purification unit, combined reformers, high
and low-temperature shift reactors, and CO2 recovering 4.3 | Equivalent annual operating
unit. Based on References 59,60, the TCI for the men- cost (EAOC)
tioned units is circa 205.97 million USD, while the TCI of
the whole process with the urea production capacity of The aggregation of the cost of manufacturing (COM) and
45.94 t/h (764.96 kmol/h) is 622.35 million USD. Based on the yearly capital cost is defined as the EAOC.61 The
16374 ZIAEI ET AL.

TABLE 6 TCI of standalone (rows 1 and 2) and integrated processes (row 3)

Natural gas to process side and fired heater Process Products (tonnes/h)
(1) Natural gas feed (16 390 kmol/h) Two trains of Oryx GTL plant + Power plant LPG 3.2
Makeup natural gas (1942 kmol/h) (246 M$) Gasoline 71.4
Diesel 64.6
Wax 40.1
(2) Natural gas (1079 kmol/h) Ammonia and Urea (622.3 M$) Urea 46
(3) Natural gas feed (16 390 kmol/h) Two trains of Oryx GTL plant + Power plant Urea 46
Makeup natural gas (1942 kmol/h) (246 M$) + Ammonia and Urea (416.4 M$) LPG 3.2
+ PEM electrolyzer (165.6 M$) Gasoline 71.4
Diesel 64.6
Wax 40.1

TABLE 7 TCI breakdown of standalone and integrated processes

PEM
Power plant electrolyzer FT reactor ASU Ammonia- ATR Total
(M$) (M$) (M$) (M$) urea (M$) (M$) (M$)
GTL + power plant 246.1 - 163 123.2 622.3 413.52 1568.2
+ ammonia-urea
Integrated process 246.1 165.6 163 105.6 416.4 413.52 1510.2

yearly capital cost is the amortized fixed capital cost over As depicted in Figure 11, at the same capacities for each
the operating years of the plant to determine an annual run, the profit index of the integrated process is higher than
cost. For example, for a discount rate of 8% and plant the profit index of the standalone processes. At the pro-
operating life of 20 years, the yearly capital cost is 10%. posed design of experiment, the fourth run has the maxi-
The yearly COM can be computed by Equation (17): mum profit index (PI = 1.183) in comparison to the other
points. The decision variables of the maximum PI are [A, B,
COM ¼ 0:3037  FCI þ 2:73  C OL þ 1:23  ðCUT þ CRM Þ, C, D, E] = [0.978, 0.823, 0.964, 0.420, 184 795.918]. The car-
ð17Þ bon efficiency of this point is 0.7513 and also is the highest
one among 50 other points of the DOE.
where FCI is a term for fixed capital investment, COL
defines the cost of operating labor, CUT represent the
price of utilities, with CRM being the price of raw mate- 4.4 | Investigation of different scenarios
rial. The average charges of feedstocks and products were of the process integration
taken from our previous work19 and represented in
Table 8. 4.4.1 | Optimization of integrated process
The yearly profit is equal to the deduction of revenue based on maximizing of the carbon efficiency
from EAOC and sold products. The profit index (PI) is and PI
specified by Equation (18):
Optimization of the defined objective functions (1- maxi-
ðannual profitÞi mization of the carbon efficiency, 2- maximization of
PI i ¼ , ð18Þ
annual profit of the base_case profit index) were performed separately with two differ-
ent methods using model- based calibration toolbox of
where i is the number of runs represented in Table S1 of MATLAB.
Data S1. The definition of the optimization problem is as
The profit index of the base-case (row 49 in Table S1 below:
of Data S1) was assumed to be 1 (ie, 100%) and the profit Find:
indices of the standalone processes and the overall inte- Unpurged (A), recycle to FT (B), CO2 removal from
grated process were calculated and compared with this syngas (C), steam/carbon (D), and CO2 intake to
value. ATR (E).
ZIAEI ET AL. 16375

TABLE 8 Average prices of raw material and products higher profit index as well. The results of the optimiza-
tions for carbon efficiency are indicated in Table 9. Here-
Utility Cost
inafter, the optimization results of the fmincon in Table 9
LPG $1/kg
is referred to as “optimum case.” Table 10 demonstrates
Diesel $0.73/kg the optimization results of the case that the profit index
Gasoline $0.88/kg is maximized. As it is evident in Tables 9 and 10, the
Electricity $0.06/kWh profit index of the integrated process with carbon effi-
CO2 tax $50/tonne ciency maximization is more than the one with PI maxi-
mization. For each degree of freedom, the optimal value
Ammonia $261.8/tonne
of that decision variable is depicted in Figures 12 and 13,
Urea $254.28/tonne
for carbon efficiency and PI maximization, respectively.
As can be seen from these figures, some of the constraints
are activated at their upper or lower bounds.

4.4.2 | Comparison of the integrated iron-

based GTL-ammonia-urea synthesis with
cobalt-based GTL-ammonia-urea

In this section, the results of this study are compared

with our previous work19 where a GTL process with
cobalt-based FT reactor was integrated with ammonia
and urea processes. For profitability analysis of the two
proposed integrated configurations, the investment cost
of the GTL process with iron- and cobalt-based FT reac-
F I G U R E 1 1 Profit index (PI) of the integrated and standalone tors must be taken into account. According to data from
processes for each run [Colour figure can be viewed at the open literature, the price of cobalt catalysts is about
wileyonlinelibrary.com] 70 times more expensive than iron.62 The following sup-
positions were made to calculate the investment cost of
the catalysts:
By evaluation and maximization of:
• The cobalt mass fraction in FT catalyst is 20% and cata-
1. Carbon efficiency lyst volume is 10% of the total reactor volume.40
2. Profit index (PI) • The density of the catalyst is 2000 kg/m3.
• The composition of the iron catalyst is 74.3% Fe, 3.7%
Subject to: Cu, 3.1% K and 18.6% SiO2.63

0:95 < A < 0:98 The total optimum carbon efficiency in the integrated
process with an iron-based FT reactor is about 75.94%,
0:75 < B < 0:85 where the optimum value of carbon efficiency in our pre-
vious work19 is circa 81.57%. Despite the higher price of
0:93 < C < 0:99 the cobalt catalyst, the profitability index of the inte-
grated process with the Co-based FT-GTL plant is
0:4 < D < 0:6 approximately 5% higher than that of the integrated pro-
cess with an iron-based FT reactor (current study).
160 000 < E < 205 000 Table 11 demonstrates the key differences between the
iron- and cobalt-based integrated processes.
As can be seen from Table 11, the integrated process
For each of the objective functions, the optimizations with Co-based FT reactor consumes less makeup natural
were accomplished using the local fmincon method and gas and the total CO2 emission is about 44.2 t/h less than
genetic algorithm (ga) global optimization method. From the iron-based integrated process. Additionally, the car-
our previous study,19 it was concluded that the case with bon efficiency of the iron-based GTL process is lower
higher carbon efficiency in the integrated process has the than the cobalt-based configuration while the wax
16376 ZIAEI ET AL.

TABLE 9 Maximization of carbon efficiency

Optimized
value Observed
(D) Steam/ (E) obtained value
Optimization (A) Unpurged (B) Recycle (C) CO2 natural CO2 from from the Obtained
technique ratio to FT removal gas intake model flowsheet PI
Fmincon 0.98 0.85 0.9648 0.4 182 500 0.758 0.7594 1.1996
(optimum case)
ga 0.98 0.8488 0.9754 0.4023 202 491 0.762 0.7562 1.1957

TABLE 10 Optimization results of the profit index

(A) (B) Optimized value Observed value

Optimization Unpurged Recycle (C) CO2 (D) Steam/ (E) CO2 obtained from the from the
technique ratio to FT removal natural gas intake model flowsheet
ga 0.9745 0.8285 0.9487 0.4261 183 452.439 1.171 1.178
fmincon 0.9784 0.8225 0.9602 0.4401 182 500 1.18 1.179

FIGURE 12 Independent variables of the carbon efficiency maximization [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 13 Independent variables of the PI maximization [Colour figure can be viewed at wileyonlinelibrary.com]

production rate of the iron-based FT reactor is higher block diagrams of the integrated processes with cobalt-
than that of the cobalt-based process which owes to the and iron-based FT reactor are presented in Figures 14
higher chain growth probability of the FT reactor. Simple and 15, respectively.
ZIAEI ET AL. 16377

TABLE 11 Comparison between iron-based and cobalt-based integrated processes

Products mass flow of the GTL

FT Makeup Urea mass Total CO2 Carbon α inside
process (kg/h)
reactor natural gas flow emission efficiency the FT
catalyst (kmol/h) LPG Gasoline Diesel Wax (tonne/h) (tonne/h) (%) reactor
Iron-based 2548 1306.3 42 289.4 59 436.5 93 434.6 48.1 208.4 75.94 0.9118
19
Co-based 2299 3884.7 56 227.6 67 415 81 679.9 55.1 164.2 81.57 0.8988

F I G U R E 1 4 Simple schematic of
the cobalt-based GTL process
integrated with urea and ammonia
units19 [Colour figure can be viewed
at wileyonlinelibrary.com]

F I G U R E 1 5 Simple schematic of
the cobalt-based GTL process
integrated with urea and ammonia
units [Colour figure can be viewed at
wileyonlinelibrary.com]
16378 ZIAEI ET AL.

4.4.3 | The feasibility investigation of using emission tax was assumed that is, $50, $100, and $200/
solar power in process integration tonne of the emitted CO2.64

In this subsection, the possibility of using solar power in

the integrated process (Figure 3) to produce hydrogen as 4.4.4 | Substituting the smallest ASU by
ammonia feedstock has been investigated. For this pur- solar photovoltaic (PV) panels followed by a
pose, two different scenarios have been considered. The PEM electrolyzer
base-case chosen for this investigation is the optimum
case given in Table 9 where the carbon efficiency is maxi- In this case, the smallest ASU of the GTL plant was elim-
mized. The optimal decision variables are [A, B, C, D, inated and replaced by solar photovoltaic (PV) cells to
E] = [0.98, 0.85, 0.9648, 0.4, 182 500]. Three levels of CO2 produce electrical power followed by a PEM electrolyzer

F I G U R E 1 6 Integration of the
proposed process with solar power
plant (the 58.36 MWASU is
eliminated) [Colour figure can be
viewed at wileyonlinelibrary.com]

F I G U R E 1 7 Integration of the
proposed process with solar power plant
(the 74.5 MW ASU is eliminated)
[Colour figure can be viewed at
wileyonlinelibrary.com]
ZIAEI ET AL. 16379

TABLE 12 PI values of different scenarios

Scenario Capacity of ASU to be CO2 tax

No. replaced ($/tonne) PI
1 58.36 50 0.89
100 1.05
200 1.65
2 60.15 50 0.887
100 1.04
200 1.648
2 61.95 50 0.875
100 1.029
200 1.639

F I G U R E 1 8 The impact of substituting the ASUs on the PI at 3 63.74 50 0.861

different levels of CO2 tax [Colour figure can be viewed at 100 1.015
wileyonlinelibrary.com] 200 1.625
4 65.53 50 0.858
100 1.014
to produce oxygen and hydrogen as the feedstocks for the
200 1.632
ATR and ammonia plant, respectively. The simple pro-
cess layout is depicted in Figure 16. In the mentioned 5 67.32 50 0.843
optimum case, the capacity of the smallest and largest 100 1
ASUs is in turn 58.36 and 74.50 MW. In the presence of a 200 1.618
solar power generation farm, it is possible to utilize about 6 69.12 50 0.833
3810 kmol/h of CO2 (as a feed of urea production unit)
100 0.99
which is about 4.75 times larger than the optimum case
200 1.611
presented in Table 9. The total capital investment for the
solar farm with a capacity of 960 MW is circa 1017 mil- 7 70.91 50 0.822
lion USD.65 Based on Reference 66, the required area for 100 0.981
the construction of a solar farm with the mentioned 200 1.605
capacity is about 28.5 km2. Figure 18 demonstrates the 8 72.71 50 0.811
PIs of the different scenarios. The PI of the optimum case 100 0.97
(Table 9) with CO2 emission tax of $50, $100, and $200/
200 1.6
tonne is respectively 1.2, 1.26, and 1.51. The reason for
9 74.5 50 0.82
the rise in the profit index as an increase in CO2 tax is
that in comparison to the base case, these points utilize 100 0.98
more CO2 and therefore pay less tax. 200 1.615
As evident in Figure 18, when the CO2 tax is $200/t,
the profit index of the case using solar power for water
electrolysis is about 10% larger than the optimum case produce 186.25 t/h of oxygen To produce 186.25 t/h of
(Table 9) of the integrated process. oxygen, a PEM electrolyzer with a capacity of 1225 MW
is required.
The total capital investment of the solar power gener-
4.4.5 | Substituting the largest ASU by solar ation farm to power the PEM electrolyzer is 1298 million
photovoltaic (PV) panels followed by a PEM USD.65 In this case, not only the whole produced CO2
electrolyzer from the GTL process is utilized, but also about 2.8 t/h of
CO2 is needed to convert all of the produced ammonia to
Unlike the previous case, the largest ASU (the one with a urea. In other words, the carbon efficiency of 100% is
capacity of 74.5 MW and oxygen production capacity of reached.
186.25 t/h) was eliminated and replaced by solar panels PI values for different scenarios for two cases
followed by a PEM electrolyzer. The simple process (1-eliminating the smallest ASU, and 2-eliminating the
scheme for this integration is illustrated in Figure 17. To largest ASU) are represented in Figure 18. As can be
16380 ZIAEI ET AL.

C, D, E] = [0.9784, 08225, 0.9602, 0.4401, 182 500],

respectively.
Results show that at the optimum point, approximately
35.2 t/h of the recovered CO2 in the GTL plant can be
reused as the feedstock of the urea synthesis plant. The
CO2 emission of the integrated process is about 48.2 t/h
less than the standalone processes. The total fixed capital
cost of the integrated process is about 58 million USD less
than the standalone processes (Table 6). In comparison
with the standalone processes, the profit index of the inte-
grated process was increased by circa 30%.
According to the economic calculations, the profit-
F I G U R E 1 9 PI values for the case with $200/t of tax [Colour ability of the CO-based integrated process19 is about 5%
figure can be viewed at wileyonlinelibrary.com] more than the iron-based integrated process (present
work) when the same amount of natural gas is fed to the
integrated process. Additionally, the carbon efficiency of
concluded from Figure 18, the most profit index is the integrated process with iron-based FT reactor is lower
obtained by omitting the smallest ASU and considering than the cobalt-based configuration19 while the wax pro-
the CO2 tax of $200/tonne. duction rate of the current study is higher than that of
The possibility of replacing the ASUs with proper the previous study.
PEM electrolyzers with different capacities between A solar PV power plant can be integrated to the pro-
58.36 MW and 74.5 MW is examined. In this investiga- cess in order to improve the carbon efficiency of the
tion, the ASUs with different capacities of (58.36, 60.15, whole process. By eliminating the smallest and the larg-
61.95, 63.74, 65.53, 67.32, 69.12, 70.91, 72.71, and 74.5) est ASUs the carbon efficiencies become 96.7% and 100%,
MW were replaced by the PEM electrolyzers to compen- respectively. Despite the enhancement of the carbon effi-
sate the required amount of oxygen to GTL plants. The ciency, the total capital investment for these cases
PIs of the different cases in three scenarios comprising increases drastically. In the first case study (eliminating
the tax of CO2 emission $50, $100, and $200/tonnes of the smallest ASU), the results of the sensitivity analysis
CO2, is represented in Table 12. indicate that at CO2 tax of $100 and $200/tonne of emit-
According to Table 12, the PI values (for the case with ted CO2, the profitability of the integrated process
a tax of $200/tonne of emitted CO2) were represented in enhances by 5% and 65% respectively in comparison with
Figure 19. As can be seen in Figure 19, the most profit- the base-case. For the second case study (eliminating the
able point is obtained by replacing the smallest ASU with largest ASU), the results indicate that a green process can
a PEM electrolyzer. The increase in PI value for points be achieved in which not only all the produced CO2 in
larger than 72.71 MW is due to all the produced CO2 in the GTL process is utilized, but also there is around
the process is utilized for urea production and the process 2.8 t/h CO2 sink in the proposed integration.
also capable to utilize more CO2 from external sources.
However, due to the land use considerations for solar ORCID
power plants, the replacement of ASU with larger capaci- Mehdi Panahi https://orcid.org/0000-0003-4907-226X
ties of 74.5 MW was not taken into consideration.
RE FER EN CES
1. Olah GA, Mathew T, Goeppert A, Surya Prakash GKGK. Dif-
5 | C ON C L U S I ON ference and significance of regenerative versus renewable car-
bon fuels and products. J Top Catal. 2018;61:522-529.
In this work, the carbon efficiency and profit index of an 2. Rafiee A, Khalilpour KR, Milani D. Chapter 8 - CO2 conversion
integrated GTL process with ammonia and urea plants and utilization pathways. In: Khalilpour KR, ed. Polygeneration
were maximized to utilize the maximum possible amount with Polystorage for Chemical and Energy Hubs. Cambridge,
Massachusetts, United States: Academic Press; 2019:213-245.
of CO2 release from the GTL process. To attain these pur-
3. Global Monthly Mean CO2, in, NOAA/ESRL, 2021.
poses, the design of experiment was implemented using
4. Matzen M, Demirel Y. Methanol and dimethyl ether from
LHD. The results demonstrate that the optimum carbon renewable hydrogen and carbon dioxide: alternative fuels pro-
efficiency and profit index occur at [A, B, C, D, duction and life-cycle assessment. J Clean Prod. 2016;139:
E] = [0.98, 0.85, 0.9648, 0.4, 182 500] and [A, B, 1068-1077.
ZIAEI ET AL. 16381

5. Aresta M, Dibenedetto A. Utilisation of CO2 as a chemical feed- 22. Abdelaziz OY, Hosny WM, Gadalla MA, Ashour FH,
stock: opportunities and challenges. Dalton Trans. 2007;28: Ashour IA, Hulteberg CP. Novel process technologies for con-
2975-2992. version of carbon dioxide from industrial flue gas streams into
6. Peters M, Köhler B, Kuckshinrichs W, Leitner W, Markewitz P, methanol. J CO2 Utiliz. 2017;21:52-63.
Müller TE. Chemical technologies for exploiting and recycling 23. Bellotti D, Sorce A, Rivarolo M, Magistri L. Techno-economic
carbon dioxide into the value chain. ChemSusChem. 2011;4: analysis for the integration of a power to fuel system with a
1216-1240. CCS coal power plant. J CO2 Utiliz. 2019;33:262-272.
7. Bruhn T, Naims H, Olfe-Kräutlein B. Separating the debate on 24. Castellani B, Rinaldi S, Morini E, Nastasi B, Rossi F. Flue gas
CO2 utilisation from carbon capture and storage. Environ Sci treatment by power-to-gas integration for methane and ammo-
Pol. 2016;60:38-43. nia synthesis – energy and environmental analysis. Energy Con-
8. Aresta M, Dibenedetto A, Angelini A. Catalysis for the valori- vers Manag. 2018;171:626-634.
zation of exhaust carbon: from CO2 to chemicals, materials, 25. Bose A, Jana K, Mitra D, De S. Co-production of power and
and fuels. Technological use of CO2. Chem Rev. 2014;114:1709- urea from coal with CO2 capture: performance assessment.
1742. Clean Techn Environ Policy. 2015;17:1271-1280.
9. Huang C-H, Tan C-S. A review: CO2 utilization. Aerosol Air 26. Zhang C, Jun K-W, Kwak G, Lee Y-J, Park H-G. Efficient utili-
Qual Res. 2014;14:480-499. zation of carbon dioxide in a gas-to-methanol process com-
10. Olah GA, Goeppert A, Prakash GKS. Chemical recycling of car- posed of CO2/steam–mixed reforming and methanol synthesis.
bon dioxide to methanol and dimethyl ether: from greenhouse J CO2 Utiliz. 2016;16:1-7.
gas to renewable, environmentally carbon neutral fuels and 27. Zhang C, Gao R, Jun K-W, et al. Direct conversion of carbon
synthetic hydrocarbons. J Org Chem. 2009;74:487-498. dioxide to liquid fuels and synthetic natural gas using renew-
11. Lee JH, Lee DW, Kwak C, Kang K, Lee JH. Technoeconomic able power: techno-economic analysis. J CO2 Utiliz. 2019;34:
and environmental evaluation of sodium bicarbonate produc- 293-302.
tion using CO2 from flue gas of a coal-fired power plant. Ind 28. Meunier N, Chauvy R, Mouhoubi S, Thomas D, Weireld GD.
Eng Chem Res. 2019;58:15533–15541. Alternative production of methanol from industrial CO2.
12. Karjunen H, Tynjälä T, Hyppänen T. A method for assessing Renew Energy. 2020;146:1192-1203.
infrastructure for CO2 utilization: a case study of Finland. Appl 29. Pérez-Fortes M, Bocin-Dumitriu A, Tzimas E. CO2 utilization
Energy. 2017;205:33-43. pathways: techno-economic assessment and market opportuni-
13. Putra ZA, Kurnia JC, Sasmito AP, Muraza O. Process design ties. Energy Procedia. 2014;63:7968-7975.
and techno-economic analysis of ethyl levulinate production 30. Rafiee A, Panahi M, Khalilpour KR. CO2 utilization through
from carbon dioxide and 1,4-butanediol as an alternative bio- integration of post-combustion carbon capture process with
fuel and fuel additive. Int J Energy Res. 2019;433:5932-5945. Fischer-Tropsch gas-to-liquid (GTL) processes. J CO2 Utiliz.
14. Parigi D, Giglio E, Soto A, Santarelli M. Power-to-fuels through 2017;18:98-106.
carbon dioxide re-utilization and hightemperature electrolysis: 31. Panahi M, Yasari E, Rafiee A. Multi-objective optimization of a
a technical and economical comparison between synthetic gas-to liquids (GTL) process with staged Fischer-Tropsch reac-
methanol and methane. J Clean Prod. 2019;226:679-691. tor. Energy Convers Manag. 2018;163:239-249.
15. Lee HW, Kim K, An J, et al. Toward the practical application 32. Fazeli H, Panahi M, Rafiee A. Investigating the potential of car-
of direct CO2 hydrogenation technology for methanol produc- bon dioxide utilization in a gas-to-liquids process with iron-
tion international. J Energy Res. 2020;44:8781-8798. based Fischer-Tropsch catalyst. J Nat Gas Sci Eng. 2018;52:
16. Li C, Gao Y, Wu C. Modeling and simulation of hydrogen pro- 549-558.
duction from dimethyl ether steam reforming using exhaust 33. Ostadi M, Hillestad M. Conceptual design of a once-through
gas. Int J Energy Res. 2015;39:1272-1279. gas-to-liquid process combined with ammonia synthesis. Chem
17. Jiang Z, Xiao T, Kuznetsov VL, Edwards P. Turning carbon Eng Technol. 2018;41:1668-1674.
dioxide into fuel. Philos Trans R Soc A Math Phys Eng Sci. 34. Nguyen TBH, Zondervan E. Methanol production from cap-
2010;368:3343-3364. tured CO2 using hydrogenation and reforming technologies_
18. Quadrelli EA, Centi G, Duplan JL, Perathoner S. Carbon diox- environmental and economic evaluation. J CO2 Utiliz. 2019;34:
ide recycling: emerging large-scale technologies with industrial 1-11.
potential. ChemSusChem. 2011;4:1194-1215. 35. S
anchez A, Gil LM, Martín M. Sustainable DMC production
19. Ziaei M, Panahi M, Fanaei MA, Rafiee A, Khalilpour KR. Maxi- from CO2 and renewable ammonia and methanol. J CO2 Utiliz.
mizing the profitability of integrated Fischer-Tropsch GTL pro- 2019;33:521-531.
cess with ammonia and urea synthesis using response surface 36. Perego C, Bortolo R, Zennaro R. Gas to liquids technologies for
methodology. J CO2 Utiliz. 2020;35:14-27. natural gas reserves valorization: the Eni experience. Catal
20. Bellotti D, Cassettari L, Mosca M, Magistri L. RSM approach Today. 2009;142:9-16.
for stochastic sensitivity analysis of the economic sustainability 37. Davis BH. FischerTropsch synthesis: comparison of perfor-
of a methanol production plant using renewable energy mances of iron and cobalt catalysts. Ind Eng Chem Res. 2007;
sources. J Clean Prod. 2019;240:1-16. 46:8938-8945.
21. Fern
andez-Dacosta C, Stojcheva V, Ramirez A. Closing carbon 38. Khodakov AY, Chu W, Fongarland P. Advances in the develop-
cycles: evaluating the performance of multi-product CO2 ment of novel cobalt FischerTropsch catalysts for synthesis of
utilisation and storage configurations in a refinery. J CO2 long-chain hydrocarbons and clean fuels. Chem Rev. 2007;107:
Utiliz. 2018;20:128-142. 1692-1744.
16382 ZIAEI ET AL.

39. Rafiee A, Panahi M. Optimal design of a gas-to-liquids process 56. Rafiee A, Hillestad M. Techno-economic analysis of a gas-to-
with a staged Fischer-Tropsch reactor. Chem Eng Technol. liquid process with different placements of a CO2 removal unit.
2016;39:1778-1784. Chem Eng Technol. 2012;35:420-430.
40. Panahi M, Rafiee A, Skogestad S, Hillestad M. A natural gas to 57. Breeze P. The Cost of Power Generation, Business Insights Ltd.
liquids process model for optimal operation. Ind Eng Chem Res. 2010.
2012;51:425-433. 58. Bucy JD. The potential of power-to-gas, in, ENEA Consulting.
41. Liu K, Song C, Subramani V. Hydrogen and Syngas Production 2016.
and Purification Technologies. USA: Wiley; 2010. 59. Santos SO, Collodi G, Azzaro G, Ferrari N. Techno-Economic
42. Medrano-García JD, Ruiz-Femenia R, Caballero JA. Multi- Evaluation of Hyco Plant Integrated to Ammonia/Urea or
objective optimization of combined synthesis gas reforming Methanol Production with CCS, in, IEAGHG. 2017.
technologies. J CO2 Utiliz. 2017;22:355-373. 60. Santos SO, Collodi G, Azzaro G, Ferrari N. Techno-Economic
43. Rafiee A, Hillestad M. Synthesis gas production configurations Evaluation of SMR Based Standalone (Merchant) Plant with
for gas-to-liquid applications. Chem Eng Technol. 2012;35: CCS, in, IEAGHG. 2017.
870-876. 61. Turton R, Bailie RC, Whiting WB, Shaeiwitz JA, Debangsu B.
44. Aasberg-Petersen K, Hansen JHB, Christensen TS, et al. Tech- Analysis, synthesis and design of chemical processes. 4th ed.
nologies for large-scale gas conversion. Appl Catal A Gen. 2001; London, United Kingdom: Prentice Hall; 2012.
221:379-387. 62. Beer MD, Kunene A, Nabaho D, Claeys M, Steen EV. Technical
45. Aasberg-Petersen K, Christensen TS, Nielsen CS, Dybkjær I. and economic aspects of promotion of cobalt-based Fischer-
Recent developments in autothermal reforming and pre- Tropsch catalysts by noble metals–a review. J Southern African
reforming for synthesis gas production in GTL applications. Inst Mining Metall. 2014;114:157-165.
Fuel Process Technol. 2003;83:253-261. 63. Laan GPVD. Kinetics, selectivity and scale up of the Fischer-
46. Anderson RB. Forty years with the Fischer-Tropsch synthesis Tropsch synthesis, in, University of Groningen, Netherlands,
1944-1984. Stud Surf Sci Catal. 1984;19:457-461. 1999.
47. Rafiee A, Hillestad M. Staging of the Fischer–Tropsch reac- 64. Ha-Duong M. What Is the Price of Carbon? Five Definitions,
tor with an iron based catalyst. Comput Chem Eng. 2012;39: Surveys and Perspectives Integrating Environment & Society, 2
75-83. (2009).
48. Rafiee A, Hillestad M. Staging of the Fischer-Tropsch reactor 65. Fu R, Feldman D, Margolis R. U.S. Solar Photovoltaic System
with a cobalt-based catalyst. Chem Eng Technol. 2013;36:1729- Cost Benchmark: Q1 2018, in, National Renewable Energy
1738. Laboratory. 2018.
49. Rafiee A, Khalilpour KR, Milani D, Panahi M. Trends in CO2 66. Ong S, Campbell C, Denholm P, Margolis R, Heath G. Land-
conversion and utilization: a review from process systems per- use requirements for solar power plants in the United States,
spective. J Environ Chem Eng. 2018;6:5771-5794. in, National Renewable Energy Laboratory (NREL), 2013.
50. Bellotti D, Rivarolo M, Magistri L, Massardo AF. Feasibility
study of methanol production plant from hydrogen and cap-
tured carbon dioxide. J CO2 Utiliz. 2017;21:132-138. SU PP O R TI N G I N F O RMA TI O N
51. Panahi M, Skogestad S. Economically efficient operation of Additional supporting information may be found online
CO2 capturing process part I: self-optimizing procedure for in the Supporting Information section at the end of this
selecting the best controlled variables. Chem Eng Process Pro- article.
cess Intensif. 2011;50:247-253.
52. Fang K-T, Li R, Sudjianto A. Design and Modeling for Computer
Experiments. 1st ed. Boca Raton, Florida: Chapman and How to cite this article: Ziaei M, Panahi M,
Hall/CRC; 2005. Fanaei MA, Rafiee A, Khalilpour K. A highly
53. Tang B. Orthogonal array-based latin hypercubes. J Am Stat carbon-efficient and techno-economically
Assoc. 1993;88:1392-1397.
optimized process for the renewable-assisted
54. Petelet M, Iooss B, Asserin O, Loredo A. Latin hypercube sam-
pling with inequality constraints. AStA Adv Stat Anal. 2009;94:
synthesis of gas to liquid fuels, ammonia, and urea
325-339. products. Int J Energy Res. 2021;45(11):
55. Tarpey T. A note on the prediction sum of squares statistic for 16362–16382. https://doi.org/10.1002/er.6886
restricted least squares. Am Stat. 2000;54:116-118.