Renewable and Sustainable Energy Reviews 173 (2023) 113057

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Original research article

Techno-economic assessment of green ammonia production with different

wind and solar potentials
Nicolas Campion a ,∗, Hossein Nami b,c , Philip R. Swisher a , Peter Vang Hendriksen b ,
Marie Münster a
a
DTU Management, Technical University of Denmark, Anker Engelunds Vej 101, Kongens Lyngby, 2800 Denmark
b DTU Energy, Technical University of Denmark, Anker Engelunds Vej 101, Kongens Lyngby, 2800 Denmark
c SDU Life Cycle Engineering, Department of Green Technology, University of Southern Denmark, Campusvej 55, Odense M, 5230, Denmark

ARTICLE INFO ABSTRACT

Dataset link: Input data (Original data) This paper focuses on developing a fast-solving open-source model for dynamic power-to-X plant techno-
economic analysis and analysing the method bias that occurs when using other state-of-the-art power-to-X
Keywords:
cost calculation methods. The model is a least-cost optimisation of investments and operation-costs, taking
Green
Hydrogen
as input techno-economic data, varying power profiles and hourly grid prices. The fuel analysed is ammonia
Techno-economic synthesised from electrolytic hydrogen produced with electricity from photovoltaics, wind turbines or the grid.
Renewable Various weather profiles and electrolyser technologies are compared. The calculated costs are compared with
Ammonia those derived using methods and assumptions prevailing in most literature. Optimisation results show that a
Semi-islanded semi-islanded set-up is the cheapest option and can reduce the costs up to 23% compared to off-grid systems
but leads to e-fuels GHG emissions similar to fossil fuels with today’s electricity blend. For off-grid systems,
estimating costs using solar or wind levelized cost of electricity and capacity factors to derive operating hours
leads to costs overestimation up to 30%. The cheapest off-grid configuration reaches production costs of 842
e/t𝑁𝐻3 . For comparison, the "grey" ammonia price was 250 e/t𝑁𝐻3 in January 2021 and 1500 e/t𝑁𝐻3 in April
2022 (Western Europe). The optimal power mix is found to always include photovoltaic with 1-axis tracking
and sometimes different types of onshore wind turbines at the same site. For systems fully grid connected,
approximating a highly fluctuating electricity price by a yearly average and assuming a constant operation
leads to a small cost (<4%) overestimation compared to the more advanced optimisation method applied here
.

1. Introduction conducted to estimate electro-fuel production costs and to identify cost

The latest IPCC reports on climate consequences of increased emis-
sions shows that urgent action is needed to reduce the anthropogenic
greenhouse gas emissions to mitigate severe adverse climate effects [1].
Focusing on the transport sector, direct electrification powered by re-
newable energy is identified as the most efficient and cheapest alterna-
tive to replace fossil fuels [2,3]. However, electrification of long-range
transport like aviation or shipping is severely challenged by the weight
of batteries and limited autonomy [4]. On the other hand, the large-
scale development of biofuels raises concerns regarding the available
sustainable biomass resources [5]. Thus, electro-fuels synthesised from
renewable hydrogen produced through water electrolysis requiring less
or no biomass (like e-ammonia) are also envisaged as substitutes. One
barrier to electro-fuel development is the economic viability of its
production. Therefore, in the last decade, many studies have been
reduction potentials following various methodologies.
Some studies analyse the potential development of electro-fuels,
considering multiple fuel plants, using aggregated data, and modelling
the whole energy system for a region or a country [6–8]. Usually, the
fuel costs presented make sense from a system perspective but can
hardly be used for a single plant’s cost analysis. Focusing instead on
the plant level allows a better perspective on the investment decisions,
operation or plant design that a private stakeholder must consider.
One common approach in the literature, referenced as ‘‘process
modelling and economic analysis’’ (PM) in this study, consists in mod-
elling the plant process at a high level of detail and using it as a basis for
economic analysis (plant investment, operation expenses and fuel cost).
Proposals for reducing production costs are usually based on process
improvements such as heat integration or component scaling. Usually,

∗ Corresponding author.
E-mail address: njbca@dtu.dk (N. Campion).

https://doi.org/10.1016/j.rser.2022.113057
Received 7 June 2022; Received in revised form 22 October 2022; Accepted 13 November 2022
Available online 30 November 2022
1364-0321/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Nomenclature Parameters

This list presents the relevant abbreviations and symbols that 𝜌𝑖𝑛
𝑢 Storage charging efficiency of unit u [%]
are used within the body of this work. 𝜌𝑜𝑢𝑡
𝑢 Storage discharging efficiency of unit u [%]
Abbreviations 𝐴𝑢 Annuity factor of unit u [–]
𝐶𝑜𝑛𝑠𝐸𝐿
𝑢 Electrical consumption of unit u if sign
𝐴𝐸𝐶 Alkaline electrolyser is negative, electrical production if sign is
𝐴𝑆𝑈 Air separation unit positive [kWh/Output quantity]
H
𝐶𝑎𝑝𝑒𝑥 Capital expenditure 𝐶𝑜𝑛𝑠𝑢 2 Hydrogen consumption of unit u if sign
𝐶𝐹 Capacity factor is negative, hydrogen production if sign is
𝐷𝑃 𝑆_𝑆𝑦𝑠𝑡_𝑂𝑝𝑡 Dynamic Power Supply and System Optimi- positive [kgH2 /Output quantity]
sation 𝐶𝑜𝑛𝑠𝐻𝑒𝑎𝑡
𝑢 Heat consumption of unit u if sign is neg-
𝐸𝑆𝐼 Energy system integration ative, heat production if sign is positive
𝐹 𝑖𝑥𝐸𝑙_𝑃 Fixed electricity price [kWh/Output quantity]
𝐹 𝐿𝐻 Full load hours 𝑑 Discount rate (8%)
𝐻𝐵 Haber–Bosch 𝐷𝑢 Annual demand of unit u [Metric tonnes]
𝐻𝐻 Hub height 𝐸𝑛𝑒𝑟𝑔𝑦𝑃 𝑟𝑜𝑑𝑦 Total of the energy produced by the power-
𝐼𝑃 𝐶𝐶 Intergovernmental Panel on Climate generating technology in one year [MWh]
Change 𝐹𝑢 Fixed operation and maintenance cost of
𝐿𝐶𝑂𝐸 Levelised Cost of Electricity unit u [e/Capacity/year]
𝐿𝐻𝑉 Lower heating value 𝐼𝑢 Investment expenditure of unit u
[e/Capacity]
𝑂𝑝𝑒𝑥 Operational expenses
𝐿𝑚𝑖𝑛
𝑢 Minimal load of unit u [% of installed
𝑃 𝐸𝑀𝐸𝐶 Proton exchange membrane electrolyser
capacity]
𝑃𝑀 Process modelling
𝑛 Technical lifetime of unit u [years]
𝑃 𝑡𝑋 Power-to-X
𝑃𝑢,𝑡 Price of the output from unit u at time t
𝑆𝑖𝑚_𝐿𝐶𝑂𝐸 Simulation and use of LCOE
[e/Output quantity]
𝑆𝑂𝐸𝐶 Solid oxide electrolyser
𝑃 𝑃𝑢,𝑡 Normalised power profile of unit u at time
𝑆𝑃 Specific power t [kW]
𝑆𝑦𝑠𝑡_𝑂𝑝𝑡 System Optimisation 𝑃 𝑅𝑢 Quantity of product synthesised in unit u
Sets and subsets per kilo of reactant [kg output/kg input]
𝑅𝑑𝑜𝑤𝑛
𝑢 Ramping rate down of unit u [% of installed
𝑡 Time in hours capacity/h]
𝑢 Unit of the fuel plant system 𝑅𝑢𝑝 Ramping rate up of unit u [% of installed
𝑢
𝑦 Time in years capacity/h]
𝜓 Subset of renewable intermittent power 𝑉𝑢 Variable operation and maintenance cost of
units unit u [e/Output quantity]
𝐼𝑛𝑗 Subset of units used to fill a storage facility
𝑀𝑖𝑛𝐷 Subset of fuel plant units with an external Variables
output demand 𝐶𝑢 Invested capacity of unit u [kg𝑜𝑢𝑡𝑝𝑢𝑡 /h or
𝑂𝑢𝑡𝑗 Subset of units used to empty a storage kW]
facility 𝑋𝑢,𝑡 Output flow of mass or energy from unit u
𝑃𝑖 Subset of units synthesising a product using at time t [kg𝑜𝑢𝑡𝑝𝑢𝑡 or kWh]
a reactant from unit 𝑅𝑖
𝑅𝑖 Subset of units providing a reactant to unit
𝑃𝑖
𝑇𝑗 Subset of units acting as a storage facility programming optimisation have also been used [19–23]. However,
whereas the increased complexity may improve the model realism, this
usually comes with increased computational expenses, reducing the
practicality and accessibility of the modelling tool.
these studies do not model the electricity supply extensively and often In recent literature reviews focusing on e-fuels [24,25], the studies
assume a fixed electricity price and a steady plant operation [9–12]. are explicitly compared in terms of their techno-economic assumptions.
This is a simplification of the conditions of real future green e-fuel This is useful but the comparison does not include an analysis of the
plant, which will have to face either fluctuating electricity prices (‘‘grid implications of the type of modelling method used. In this study, we
have classified the different studies via the method applied and later
connected’’) or fluctuating electricity availability (‘‘off-grid’’ operation,
elucidate the effects of the method for the cost prediction. Specifically,
also called ‘‘islanded’’).
the focus is on comparing two generic types of methods identified
Some studies here called ‘‘individual plant models with local energy above: ‘‘individual plant models with local energy system integra-
system integration’’ (ESI) have a greater focus on the electricity supply tion’’ and ‘‘process modelling and economic analysis’’. The question
that is assumed variable and the dynamics of the system. Identified cost addressed in this paper is then: which system design and cost estimate
reduction potentials are generally related to increased plant flexibility, can be expected with an advanced individual plant model with local
reduced electricity tariffs, and combining various renewable sources energy system integration, and how does it compare with the results
for the electricity [13–18]. Other fuel-plant modelling approaches us- obtained from typical ‘‘process modelling and economic analysis’’ using
ing mixed integer linear programming or mixed integer non-linear fixed electricity prices, in calculating the fuel costs?

2
N. Campion et al.
Taking e-ammonia as an example, the production process is mod-
elled following the typical process modelling, including heat integra-
tion and comparing the process when using alkaline (AEC) or solid
oxide (SOEC) electrolysers for hydrogen production. This process model
is used directly to calculate the cost of the e-ammonia for a given
electricity price and a given number of operating hours. The results
will later be compared with the costs obtained with an ‘‘ESI’’ type of
model that uses the same input data.
The ‘‘ESI’’ model developed in this analysis, combines the lat-
est state-of-the-art modelling features of individual electrofuel plant
models, including:
• Extensive modelling of the renewable power supply, taking into
account different wind turbines and photovoltaic (PV) technolo-
gies, hourly profiles, and weather profile types.
• Dynamic operation of the fuel plant and minimal production
load constraints combined with the possibility to use intermediate
storage (hydrogen and batteries).
• Optimisation of both investments and the plant’s operation on an
hourly basis.
• Sale of by-products.
In addition, the following features currently not, or only partially
addressed in the literature, are here added to the ‘‘ESI’’ fuel plant
model:
• Multiple wind turbines and PV technologies are assumed avail-
able for every site, and each technology is associated with a
specific cost and power profile (weather-dependent). The model
invests in the turbines or mix of technologies leading to the lowest
fuel production cost for each site.
• The possibility to combine a renewable power supply (solar or
wind) with grid power and optimise the use of each given the
hourly electricity spot price. In the literature, most studies fo-
cus on off-grid renewable solutions [15,17,18,26–28] or on-grid
solutions [11,15,29–31]. Fewer use semi-islanded solutions, but
with a marginal grid usage (only the ammonia plant is grid pow-
ered) [16], no dynamic operation (constant inflexible ammonia
production) [32] or using more complex models with a relatively
high solving time [23].
These features are combined to determine the least-cost power-to-X
(PtX) system design and assess the consequent fuel production cost
for different power-supply configurations (effectively different plant
locations) and with different assumptions on plant flexibility. Until
now, all studies performing a PtX cost assessment use or develop only
one single method to calculate the fuel production cost (or the levelised
cost of fuel). In this study, two methods of cost calculations are tested
with the same techno-economic assumptions and power profile. By
doing so, it is possible to identify and quantify usual ‘‘method bias’’,
which, to the best knowledge of the authors, has not been done before.
To summarise, the novelty of this work is to gather in one study
an extensive modelling of the power supply (including various wind
turbines types, solar PV technologies with off-grid, on-grid and semi-
islanded set-ups in different locations), different electrolysis technolo-
gies (SOEC and AEC) integrated with the fuel plant, cost optimisation
of investments, and dynamic operation. In addition, this work identifies
and quantifies the methodological bias of some specific studies when
calculating e-fuel costs.
By doing so, this study identifies cost-effective solutions to develop
alternative fuels which are crucial to phase out fossil fuels and reduce
global GHG emissions. Given the various types of technologies and
weather type configurations used in this study, the conclusions can be
used as a reference for other areas in the world with similar character-
istics. Finally, in the literature, while comparing the result of different
studies, the main differences are usually explained based on differences
in the assumed input data and the method bias is disregarded. This
3
Renewable and Sustainable Energy Reviews 173 (2023) 113057
study shows how the cost calculation method used can influence the
results and where modellers should focus to avoid important method
bias.
Section 2 discusses the relevant literature focusing on single plant
power-to-X models. Section 3 describes the methods developed in
this work, and Section 4 presents the input data used in the model.
Results are presented in Section 5, which is divided in two parts, one
presenting the results obtained with the model developed in this work
(Section 5.1), the other one comparing it with the results obtained
using other methods and assumptions (Sections 5.4, 5.2, 5.3). Section 6
summarises the main conclusions.
2. Literature review
This section presents a literature review on relevant studies focusing
on single plant power-to-X models.
2.1. Review of state-of-the-art electro-fuel technologies
Given their potential to decarbonise the hard-to-abate sectors such
as shipping or aviation, electrofuels are gaining more and more atten-
tion. A recent literature review [25] on electrofuels shows that a large
variety of fuels have been studied with a large focus on e-methanol,
e-methane and Fischer–Tropsch liquid fuels. In the latest years e-
ammonia has been studied more extensively, notably because ammonia
synthesis does not require carbon. This is a significant advantage given
the limited availability of biogenic carbon sources [5] and the technical
challenges of extracting carbon from the air at a reasonable cost [33].
For this reason, the authors of this study decided to focus on ammonia
as a study case. The exact same method could however be applied
to any of the other e-fuels and the methodological conclusions of this
analysis remain valid for the other type of e-fuels.
The main component of a PtX plant is the electrolyser. In gen-
eral three types of electrolysis are studied: solid oxide electrolysis
(SOEC), alkaline electrolysis (AEC) and proton exchange membrane
electrolysis (PEM). PEM-based electrolysis units have relatively good
performance but are subject to high material supply risks, due to the use
of very scarce elements like iridium and platinum, which may hinder
its large-scale deployment [34]. For this reason, PEM electrolysis is
not considered in this study. On the other hand, alkaline electrolysers
(AEC) are characterised by long operational lives and use materials
that are abundant, but are hampered by modest efficiencies. Solid
oxide electrolysis units are the most efficient, but some materials
currently used (yttrium) may pose challenges if deployed on a large
scale [34]. However, this technology is still at a lower technology
readiness level, so further developments and material changes can be
expected. Therefore, both AEC and SOEC electrolysis types are included
in this study. The dynamics of the PtX plants and flexibility assumptions
of the different plant components are very different from one study
to another. Some do not address the question of a flexible system and
directly assume that the PtX operates steadily all year long [31]. This
represents a ‘‘best-case’’ extreme and thus gives a lower bound cost
estimate. Others consider the possibility to operate the electrolyser and
the fuel plant flexibly with a given operating load range [16–18]. When
the flexibility of the system is addressed, studies usually also include the
modelling of an additional intermediate storage of hydrogen (typically
in the form of hydrogen tanks or caverns [16]) or electricity (with
batteries) [18]. In this study, an underground pressurised hydrogen
pipe storage is considered as we believe this to be a more cost effective
solution than hydrogen tanks for medium scale applications [35].
A summary of the system characteristics of recent PtX cost assess-
ment studies can be found in Table 1.
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Table 1
Summary of the representative literature review of power-to-X plant models.
Selected studies Method System
Study System Elec. cost Power Hydrogen Plant flexibility Storage Fuel plant
type opt. method supply production (load range) FLH (h/y)
(Gomez et al. 2020) [31] PM – FixEl_P On-grid PEMEC Steady operation Hydrogen buffer 8400
SOEC
SMRa
(Funez guerra et al. 2020) PMb I FixEl_P On-gridc PEMEC d
– 7500
[30] (1000–8000)
(Bicer & Dincer, 2019) [36] PMe If FixEl_P Off-grid PECg H2 : 0%–100% – 2500
(Solar) NH3 : 0%–100%
(Zhang et al. 2020) [11] PM O+If FixEl_P On-grid SOEC Steady operation – 7200
(Collet et al. 2017) [29] PMh – FixEl_P On-grid AEC H2 : 0–100%/100% – 2000j –8760
CH4 : 0–100%i /100%
(Nami et al. 2022) [37] PM – FixEl_P Off-grid AEC H2 : 10%–100% – 4380
SOEC Fuelk : 10%–100%
(Morgan et al. 2017) [32] ESI – Sim_LCOEl Semi-islanded AEC Steady operation – 8760
(Offshore wind, grid)
(Hank et al. 2018(a)) [15] ESI O Sim_LCOE On-grid PEMEC H2 : 0%–100% H2 tanks 8400/4620m
MeOH: 20%–100%
/100%
(Hank et al. 2018(b)) [15] ESI O Sim_LCOE Off-grid PEMEC H2 : 0%–100% H2 tanks 3585/3896n
(Onshore wind) MeOH: 20%–100%
/100%
(Nayak-Luke et al. 2018) ESI O+Io Sim_LCOE Off-grid Not clear H2 : 0–100% H2 buffers Not specified
[17] (Solar PV, wind) NH3 : 20%–100%
(Armijo & Philibert, 2020) ESI O+I Syst_Opt Off-gridp AEC H2 : 0–100% H2 tank or cavern Not specified
[16] (Solar PV, wind) NH3 : 20/60%–100%
(Osman et al. 2020) [18] PM, ESI O+I Syst_Opt Off-grid AEC H2 : 20–100% Batteries, H2 tank, Not specifieds
(Solar PV, CSP) PEMEC N2 : 60–100% Thermal storage
NH3 : 100%r (molten salts)
This study PM, ESI O+I Syst_Opt On-Grid AEC H2 : 0/20–100% Batteries, See Table 8
Off-grid (Solar PV, windq ) SOEC NH3 : 10/40%–100% H2 storage
Semi-Islanded (buried pipes)

PM: Process Modelling; ESI: Single plant with local Energy System Integration; O: optimisation of the operation; I: optimisation of the investments; FLH: Full Load Hour, which is
a virtual number of hours where the plant would operate at full load over a year.
a
Ammonia is produced via electrochemical reactor or Haber–Bosch (HB) loop and hydrogen is produced via electrolysis or steam methane reforming (SMR).
b
Data from an industrial partner instead of extensive modelling.
c From solar PV, assumed to be bought at a fixed price with a power purchase agreement.

d Technical constraints for HB flexibility are not addressed.

e Based on experimental set-up.

f
At process level.
g
Photo-electrochemical water splitting.
h
Data from literature.
i
The flexibility of the methanation unit is not mentioned.
j Low FLH when using the PtX facility to store overproduced electricity.

k Focus of this study is mostly on hydrogen but the possibility to produce hydrogen-based fuels is discussed.

l Simulation-based model but electricity prices are calculated using Capex and Opex of the wind farm plus monthly grid electricity price.

m
FLH of the electrolyser with constant grid supply/electricity price-driven.
n
FLH of the electrolyser with wind-driven production and flexible/non-flexible fuel plant.
o
Without using formal optimisation methods.
p
But HB plant is powered with grid electricity.
q Different type of turbines are compared and the optimal ones are chosen for each site.

r To avoid catalyst damaging.

s But the FLH of solar PV supply is 2705 h/y.

2.2. Review of methods for modelling electro-fuel cost for single plants
The two primary types of approach to estimating the production
costs of electrofuels for individual plants could be classified as ‘‘process
modelling and economic analysis’’ (PM) [9–12,30,31,36–38] and ‘‘in-
dividual plant models with local energy system integration’’ (ESI) [15–
18,32,39]. The PM approach consists in modelling the plant process
at a high level of detail and using it as a basis for economic analysis.
Usually, these studies do not model the electricity supply extensively
and often assume a fixed electricity price and a steady plant opera-
tion [9–12]. The electricity price used is usually the grid electricity
price average for a specific year [11], a hypothetical power purchase
agreement [30] or the levelised cost of electricity (LCOE) of the gen-
erating technology [37]. This would result again in a lower bound on
the fuel cost as it does not factor in the effects of not having power
full time, having to oversize the plant for a given production. It also
neglects that in an energy system with significant amounts of PtX and
a large share of renewable production, the cost of electricity will vary
over time.
The ESI approach models the electricity supply with further details,
considering a weather-dependent electricity profile and sometimes a
variable cost for grid electricity. In this case, the fuel plant can usually
operate flexibly by use of some intermediate storage (hydrogen buffer,
batteries) or backup power [13–17]. From one study to another, the

4
N. Campion et al.
plant power supply is described at different levels of detail, going from
a single source available to an optimised mix of different power supplies
with variable availability and prices. In the latest studies of ESI type,
the size of the plant, the power supply system and hourly operation are
usually optimised together [18].
For both ESI and PM methods, the system sizing and operation
can be either pre-determined or optimised, taking into account the
economic and technical aspects. The optimisation can include the plant
operation, the plant capacities (investments), or both. The type of
optimisation used in the different studies is summarised in Table 1.
The electricity price is an important e-fuel cost driver [25] and the
way it is calculated differs between studies. Three different calculation
methods have been identified:
• Fixed electricity price methods (FixEl_P)
• Use of LCOE and fuel plant simulation with variable power supply
(Sim_LCOE)
• Using the Capex and Opex of the power supply system after
optimising this (Syst_Opt)
The principle of FixEl_P methods is, first, to estimate the electricity
consumption of every plant component (usually obtained from process
modelling) and later use this to calculate the annual electricity con-
sumption of the fuel plant for a given number of operating hours at full
load. Then, the annual electricity cost is calculated by multiplying the
yearly electricity consumption by the average annual electricity cost
(or sometimes LCOE) in e/MWh. The electricity cost is a part of the
Opex of the fuel plant. The fuel cost is obtained by adding investment
expenditure and other operation and maintenance costs. This method
is mostly used in the process modelling type of analysis such as [10–
12,30,31,36,37]. For systems that are fully grid-connected, this method
has the problem that it disregards any effects of potential grid price
variation on the plant’s operation that could benefit from operating
more flexibly and using intermediate storages. To ensure that the fuel
produced is renewable, this method also requires the ambitious assump-
tion that ‘‘green’’ electricity can be bought through a power purchase
agreement ensuring a fixed electricity price at all times like in [30].
For off-grid systems, using the LCOE of a single renewable technology
as price input associated with a given capacity factor [37] also has
some limitations. Indeed, the method cannot identify an optimal mix of
different renewable power sources or how to benefit from intermediate
storage systems. This method is also completely inadequate for the
study of ‘‘semi-islanded’’ systems.
The Sim_LCOE methods used by [15,17] put more focus on the
variability of the power supply and the flexibility of the fuel plant.
At first, the power supply and plant capacities are fixed. Then, given
renewable power capacity, production profile, and fuel plant techni-
cal specifications (electrical consumption, minimal load), a simulation
tool estimates the fuel plant’s annual electricity consumption and the
amount of fuel produced. The cost of electricity is obtained by mul-
tiplying its consumption by the levelised cost of electricity (LCOE) of
the renewable power supply. The fuel cost is obtained by adding fuel-
plant investment expenditures and other operating and maintenance
costs and dividing the sum by the total amount of fuel produced. [17]
iterates the process with different sizing configurations to optimise
the levelised cost of fuel production as a ‘‘brute-force’’ optimisation
approach. The method is further improved in [26] using a genetic
algorithm optimisation. However, finding an optimal design for semi-
islanded solutions that includes both local renewable power and some
purchase from the grid spot market remains challenging with Sim_LCOE
methods.
The Syst_Opt methods for example used by [16,18,21] do not use
the LCOE or any averaged electricity price directly to calculate the elec-
tricity cost. Instead, the cost of electricity is obtained using the Capex
and Opex of the renewable plant and the installed renewable capacities.
The sizing of the renewable plant is usually done by minimising the fuel
production cost for given fuel demand. This is a very sound approach
5
Renewable and Sustainable Energy Reviews 173 (2023) 113057
for ‘‘island’’ operation where any back up by grid purchase is neglected.
However, the possibility of using both local renewable power and some
electricity from the grid with dynamic pricing is usually not addressed
in the fast-solving models of this type. Studies that use more advanced
modelling techniques, sometime consider this set-up as an option [23].
However, the high solving time of these techniques limits the number
of possible system configurations and sensitivities that can be analysed
in practice.
The model developed in this study aims, on one hand, to com-
pensate for some of the shortcomings in the existing methods and,
on the other hand, to be used as a reference to quantify the method
bias that occurs when using fixed electricity prices for e-fuel costs
calculation. To do so, the model is designed to be able to use the
same input data set as for the fixed electricity price methods. The
analysis method is also designed to be flexible and fast-solving in
order to be usable with many different technologies, scenarios and
sensitivity-analyses. This also facilitates the use of the tool by others
to validate the results (see data and code availability). The model
has similar features compared to models using fast-solving Syst_Opt
methods (such as Osman et al. [18] or Armijo et al. [16]), but some
notable improvements have been included such as the possibility to
study optimal semi-islanded configurations with dynamic grid pricing
while maintaining a reasonable solving time. The possibility to use an
optimal mix of different wind turbine types in the same site is also
included.
Using formal optimisation methods to design the model allows also
to take into account dynamic grid pricing or variable electricity input
associated with potential flexible plant operation, which cannot be
done with methods that use a fixed electricity price. In addition, the
method presented in this study facilitates the assessment of an optimal
renewable power mix with multiple renewable sources and gives a
precise assessment of the hours of operation that can be reached using
100% renewable sources and an optimised use of the storage systems.
This method can be extended to the study of semi-islanded systems that
include both local renewable power and a connection the grid with a
dynamic electricity pricing. This gives an advantage over the Sim_LCOE
methods currently challenged with designing optimal semi-islanded
systems [26].
Finally, by restraining the ‘‘freedom’’ of the optimisation model in
a specific way, it is possible to replicate the ‘‘process modelling with
fixed electricity price’’ approach. By doing so, it is possible to compare
the results obtained with a fixed price method and the results from
the optimisation model using exactly the same data set. This gives
the possibility to identify the systematic differences (‘‘method bias’’)
between the two.
2.3. Summary
Table 1 summarises the state-of-the-art review of the systems used
in PtX analysis and classify the selected studies depending on the
method used. The methods selected are characterised by their relatively
low computational demand and their focus on individual PtX plant.
3. Methods
This section explores the model framework of the method developed
in this study described as ‘‘dynamic power supply and system optimi-
sation’’ DPS_Syst_Opt. This method will be used as a reference for the
comparison with the other state-of-the-art methods.
N. Campion et al.
Fig. 1. Overview of the cases under study.
3.1. Overview of the ‘‘dynamic power supply and system optimisation’’
model
The model is a linear programming model which minimises the costs
of investments and operation of storage, power-supply and fuel pro-
duction units. The system under study is a large-scale ammonia plant
and its surrounding energy system, including power supply and by-
product markets. Three specific types of weather profiles are analysed.
The fuel producer owns the fuel production system, including a desali-
nation plant, storage systems (hydrogen storage and battery park), an
electrolyser and an ammonia plant (Air Separation Unit and Haber–
Bosch loop). The method presented could be used to analyse any fuel
plant if the techno-economic data are available. Process side-products
such as heat and oxygen can be sold at market price. The fuel producer
pays the investment and operating costs of each owned facility. The
system is constrained to produce 430 000 tonnes of ammonia per year,
comparable to existing conventional large-scale ammonia plants [40],
which is still much larger than all current renewable ammonia projects.
Two different ways to power the plant are considered: ‘‘off-grid’’
(also known as behind-the-meter) or ‘‘semi-islanded’’: a mix of local
renewable production and grid supply. In the off-grid set-up, the fuel
producer owns renewable power facilities and cannot purchase or sell
electricity to the public grid. Excess electricity can be curtailed without
extra costs. In the islanded set-up, the fuel producer owns renewable
power facilities that are not connected to the grid and another connec-
tion allowing the purchase of electricity from the grid at the market
spot price plus grid fees. The fuel plant is assumed to be a marginal
player and does not influence the market prices. An overview of the
studied cases is presented in Fig. 1, illustrating (i) weather profiles
(plant location), (ii) how the plant is powered, and finally (iii) the
process chosen for producing the hydrogen.
The method used to determine the optimal ammonia plant design
and integration within the local energy system consists of three steps
presented in Fig. 2.
6
Renewable and Sustainable Energy Reviews 173 (2023) 113057
3.2. Power supply model
The renewable power supply side is modelled with a high level
of detail using a separate model to generate power time series and
estimate technology costs based on data from the literature. Three types
of representative weather profiles relevant to electro-fuel production
are selected:
• Profile 1, (solar+): very high solar capacity but low wind (north-
ern Chile)
• Profile 2, (wind+): very high wind capacity but low solar (Den-
mark)
• Profile 3, (wind/solar): high wind and solar capacity (South
Australia)
One objective of this paper is to understand the dynamics of typical
weather profiles, rather than identify optimal placements. Thus other
locations with higher potentials, such as western Sahara (see [41]), may
give better financial results but are not studied. Instead, the focus is
on three areas with ‘‘representative’’ renewable potential and where
data on electricity spot prices are publicly available. The precise plant
location, assumed to be on the coastline, facilitate water accessibility
and fuel bunkering. Some sites corresponding to these criteria and pre-
sented in Fig. 3 are Esbjerg in Denmark, Arica in Chile and Ceduna in
Australia. As the study aims to compare different types of profile rather
than performing an actual case study, possible practical limitations
such as available space, regulations or existing infrastructure are not
considered.
The renewable power devices considered are eight different onshore
wind turbines, four different offshore wind turbines, fixed-axis PV, and
single-axis tracking PV. Each power generating technology is charac-
terised by its cost and the associated hourly power-generating profile
specific to each selected location. DTU Wind’s CorRES model [44] was
used to determine the annual energy production of each wind turbine
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Fig. 2. Methodology.

Fig. 3. Selected sites for the study [42,43].

from wind-farm power curves (found through computational fluid dy-
namics modelling using PyWake tool) and hourly MERRA-2 weather
satellite data with reanalysis. The annual energy production of fixed
and 1-axis tracking solar generating technologies is found using the
open-source reanalysis model based on satellite data developed by [45]
(renewable.ninja). However, as the authors of the model mentioned,
the re-analysis data used (MERRA-2) tends to overestimate the solar
potential. Especially outside Europe, there is also a lack of validation
with specific site measured data. Therefore, the results obtained in this
study are likely to underestimate the capacities of solar PV actually
needed and under-size the plant infrastructures compared to a real
life PtX project. The costs obtained in the result section must then be
interpreted as a low cost boundary. Ideally, the optimisation model
would use on-site (or near-site) measured PV production profile but this
type of data is usually hard to obtain.
3.3. Fuel plant model
Ammonia synthesis plants are mainly based on the Haber–Bosch
(HB) process, which utilises hydrogen and nitrogen produced by cryo-
genic air separation (ASU). Hydrogen is produced via an electrolysis
reaction in a ‘‘green’’ ammonia synthesis plant. This study considers the
alkaline (AEC) and solid oxide (SOEC) types of electrolyser. AEC is a
mature technology with a long history of green hydrogen production
with a lower Capex compared with SOEC. However, due to a high
operating temperature, the SOEC has higher efficiency, and the reduced
electricity consumption may be worth the investment. In addition, the
high-temperature SOEC and Haber–Bosch process can be heat inte-
grated. The process modelling of the two set-ups and the related costs
are presented in Section 4.2. The cost and technical assumptions are
valid for a large-scale ammonia plant (430 kt of ammonia produced per

7
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Fig. 4. Main components of the e-fuel plant.

Table 2 The different energy system units are constrained by their technical
Input and output flows per unit. limitations for the minimal and maximal load (constraint (3)). To avoid
Unit u of the fuel plant Output Units for 𝑋𝑢,𝑡 and 𝐶𝑢 a mixed integer linear or non-linear problem formulation and to keep
Ammonia plant including ASU Ammonia kgNH3 /h reasonable calculation time, start and stop costs are not implemented.
Desalination plant Purified water1 kgH2 O /h In this regard, it is assumed that all load-constrained units operate
Electrolyser (SOEC or AEC) Hydrogen kgH2 /h
above their minimal load all year round.
Hydrogen storage Stored hydrogen kgH2
Wind turbines (on and offshore) Electricity MWh/h (=MW)
𝐶𝑢 ∗ 𝐿𝑢 𝑚𝑖𝑛 ≤ 𝑋𝑢,𝑡 ≤ 𝐶𝑢 ∀𝑢,𝑡 (3)
Solar PV (fixed and 1 axis tracking) Electricity MWh/h (=MW)
Grid Electricity MWh/h (=MW) The fuel production units can also have ramping constraints (constraint
Battery park Stored electricity MWh
(4)), with a maximal ramping rate up 𝑅𝑢𝑝 𝑢 and down 𝑅𝑢
𝑑𝑜𝑤𝑛 .

𝑋𝑢,𝑡 − 𝑋𝑢,𝑡−1 ≤ 𝑅𝑢𝑝

𝑢 ∗ 𝐶𝑢 ∀𝑢,𝑡>1
(4)
year). Techno-economic assumptions from the other components of the 𝑋𝑢,𝑡−1 − 𝑋𝑢,𝑡 ≤ 𝑅𝑑𝑜𝑤𝑛
𝑢 ∗ 𝐶𝑢 ∀𝑢,𝑡>1
fuel plant, such as hydrogen storage, batteries and a desalination plant, The electricity produced by unit u equals the installed capacity 𝐶𝑢
are taken from the literature. The different components considered in multiplied to the normalised power profile 𝑃 𝑃𝑢,𝑡 of the units 𝑢 ∈ 𝜓
the fuel plant model are presented in Fig. 4. at time t, 𝜓 being the subset of the renewable intermittent power units
(constraint (5)):
3.4. Mathematical description of the optimisation model
𝑋𝑢,𝑡 − 𝑃 𝑃𝑢,𝑡 ∗ 𝐶𝑢 = 0 ∀𝑢∈𝜓,𝑡 (5)
The optimisation model is formulated as a linear problem to allow
Then, power/heat balance must be respected (constraint (6)): power/
fast solving and the possibility to run multiple scenarios (around 500
heat generation equal power/heat consumption 𝐶𝑜𝑛𝑠𝑢 𝐸𝐿 /𝐶𝑜𝑛𝑠𝑢 𝐻𝑒𝑎𝑡
for this study). Solving one scenario takes between 1 and 5 min using
being the electric/heat consumption (if negative) or production (if
Gurobi solver with Julia/JuMP language on a desktop computer (Intel
positive) of unit u in kWh per output produced. To avoid non-linear or
Core(TM) i5-8265U CPU 1.80 GHz and 8 GB RAM).
piece-wise problem formulation that usually imply the use of integer
The decisions variables are the hourly output flows of mass or
variables, the electrolyser efficiency is assumed non-load dependent
energy from the unit ‘‘u’’ 𝑋𝑢,𝑡 and the capacities of unit u invested 𝐶𝑢
thus omitting the better efficiencies at lower load and underestimating
(both positive variables) (see Table 2).
The objective function minimises the annualised fuel production the performance of the system. This is however balanced by other
cost. approximations that overestimate the performance of the system such
∑ ∑ ∑ as no start and stop costs or no components lifetime reduction due to
min 𝑃𝑢,𝑡 ∗ 𝑋𝑢,𝑡 + (𝐼𝑢 ∗ 𝐴𝑢 + 𝐹𝑢 ) ∗ 𝐶𝑢 + 𝑉𝑢 ∗ 𝑋𝑢,𝑡 (1) dynamic operation.
𝑋𝑢,𝑡 , 𝐶𝑢
𝑢,𝑡 𝑢 𝑢,𝑡
∑
Where 𝑃𝑢,𝑡 is the hourly price of the output from unit u at time t. 𝐹𝑢 and 𝐶𝑜𝑛𝑠𝑢 𝐸𝐿 ∗ 𝑋𝑢,𝑡 = 0 ∀𝑡
𝑢
𝑉𝑢 are the fixed and variable operation and maintenance costs of unit ∑ (6)
u. 𝐼𝑢 ∗ 𝐴𝑢 is the total investment expenditure of unit u multiplied by 𝐶𝑜𝑛𝑠𝑢 𝐻𝑒𝑎𝑡 ∗ 𝑋𝑢,𝑡 = 0 ∀𝑡
𝑑(1+𝑑)𝑛 𝑢
the annuity factor 𝐴𝑢 = (1+𝑑) 𝑛 −1 . d is the discount rate (always 8%) and
n the technical lifetime of unit u. Some of the production units need to Similarly, the hydrogen production and consumption balance must
fulfil an annual demand 𝐷𝑢 . The units constrained to produce a given be in balance (equilibrium constraint (7)). Hydrogen consumption or
H
yearly amount of output (here only the ammonia plant) belong to the production (depending on the sign) 𝐶𝑜𝑛𝑠𝑢 2 is expressed in kgH2 per
subset 𝑀𝑖𝑛𝐷 (constraint (2)): output produced.
∑ ∑
𝑋𝑢,𝑡 = 𝐷𝑢 ∀𝑢 ∈ 𝑀𝑖𝑛𝐷 (2) 𝐶𝑜𝑛𝑠𝑢 H2 ∗ 𝑋𝑢,𝑡 = 0 ∀𝑡 (7)
𝑡 𝑢

8
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Table 3
PV Capex, Opex, capacity factor (CF) and LCOE [45,46].
PV type Capexa Fixed O&M Chile (solar+) Denmark (wind+) Australia (wind/solar)
(e/kW) (e/kW/y) CF LCOE CF LCOE CF LCOE
(–) (e/MWh) (–) (e/MWh) (–) (e/MWh)
Fixed-axis 550 9.1 23.3% 27.7 12.2% 53.1 20.0% 32.3
1-axis tracking 650 11.2 30.3% 25.1 15.3% 49.7 24.6% 30.9
a
This is PV farm 2020 Capex, meaning this includes the panel cost, cost of installation, and financial costs. Currency year is 2019.

Other chemicals used to produce hydrogen or synthesised using hydro- Table 4

Grid tariffs.
gen respect the mass balance presented in constraint (8), where 𝑃𝑖 is
Grid fees Grid connection Life-time grid connectionb
the subset of units synthesising a product utilising a reactant from unit
𝑅𝑖 . The production rate 𝑃 𝑅𝑢 is the quantity of product synthesised at Price 16.65 e/MWh𝑒𝑙 a 66 e/kW 35 years

unit u per kilogram of reactant. a Market spot price is added on top of the grid tariff to calculate the electricity price.
b
Needed to calculate the grid connection annuity factor.
𝑋𝑃𝑖 ,𝑡 − 𝑋𝑅𝑖 ,𝑡 ∗ 𝑃 𝑅𝑃𝑖 = 0 ∀𝑖,𝑡 (8)

Finally, the storage systems balance constraint (constraint (9)) is ex-

pressed as:
𝑋𝑢∈𝑇𝑗 ,𝑡 −𝑋𝑢∈𝑇𝑗 ,𝑡−1 −𝜌𝑖𝑛 𝑜𝑢𝑡
𝑣∈𝐼𝑛 ∗ 𝑋𝑣∈𝐼𝑛𝑗 ,𝑡 +𝜌𝑤∈𝑂𝑢𝑡 ∗ 𝑋𝑤∈𝑂𝑢𝑡𝑗 ,𝑡 = 0
𝑗 𝑗
where 𝑇𝑗 is the subset of units behaving like a tank (e.g., H2 tank or
batteries), 𝐼𝑛𝑗 is the subset of units used to charge a tank (e.g., com-
pressor or charger) and 𝑂𝑢𝑡𝑗 is the subset of units used to discharge a
tank (e.g., valve). 𝜌𝑖𝑛
𝑢 and
𝜌𝑜𝑢𝑡
𝑢 are respectively the charging and discharging efficiencies. At t
= 0, the tanks are empty.
4. Data
4.1. Power supply
4.1.1. Wind power
In this study, multiple turbines designed for different wind speeds
are available for each site. The eight onshore wind turbines comprise
four rotors (with rotor diameters of 117 m, 126 m, 136 m, and 142 m)
at two hub heights (100 m and 150 m). All onshore turbines have power
capacities between 3.15 MW and 3.45 MW. Larger rotors and higher
hub heights lead to higher capacity factors, albeit at a cost. The four
offshore turbines have the same rotor, with a diameter of 164 m. What
differentiates them is their power ratings (8 MW and 9.5 MW) and their
hub heights (100 m and 150 m). The offshore turbine with the lowest
power rating of 8 MW has a higher capacity factor, optimised for lower
wind speeds. Windfarm Capex values were found using the national
renewable energy laboratory wind-plant integrated system design &
engineering model built on NASA’s OpenMDAO software [47,48]. The
results were then scaled to match the costs for onshore and offshore
wind in the danish energy agency energy catalog [46]. Opex values
for onshore and offshore turbines were taken from [46]. The levelised
cost of electricity of each turbine is also indicated and calculated
according to Eq. (10) using a discount rate d of 8%, and a technical
lifetime of n. Techno-economic values used to model the different wind
turbines, capacity factors and LCOE are summarised in Table A.11 in
the supplementary material.
∑𝑛
𝑠𝑢𝑚 𝑜𝑓 𝑐𝑜𝑠𝑡𝑠 𝑜𝑣𝑒𝑟 𝑙𝑖𝑓 𝑒𝑡𝑖𝑚𝑒 𝑦=1
𝐿𝐶𝑂𝐸 = = ∑
𝑠𝑢𝑚 𝑜𝑓 𝑒𝑛𝑒𝑟𝑔𝑦 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑 𝑜𝑣𝑒𝑟 𝑙𝑖𝑓 𝑒𝑡𝑖𝑚𝑒 𝑛 𝐸𝑛𝑒𝑟𝑔𝑦𝑃 𝑟𝑜𝑑𝑦
𝑦=1
𝐸𝑛𝑒𝑟𝑔𝑦𝑃 𝑟𝑜𝑑𝑦 is the total of the energy produced by the power-
generating technology in one year in MWh
4.1.2. Solar PV
Both fixed-axis and single-axis PV were included as generation
∀𝑗,𝑡>1 options. Renewables.ninja [45] was used to find the annual energy
production of both fixed-axis and single-axis solar generating technolo-
(9)
gies. The optimal tilt angles for each location are taken from global
solar atlas [43], and 90% inverter efficiency was assumed. Capex and
Opex were taken from [46]. Costs, capacity factors, and LCOE for the
included PV technologies can be seen in Table 3. LCOE is calculated
using Eq. (10).
4.1.3. Grid electricity
When grid electricity is used, the electricity prices are taken from
the same place and year (2019) as the weather profile to retain any
potential correlation between electricity price and weather profile.
Prices originate from coordinador eléctrico national (Arica, northern
Chile, Solar site) [49], Nordpool (DK1, Denmark, Wind) [50], and
the Australian energy market operator (Ceduna, South Australia, So-
lar/Wind) [51]. A description of the spot market electricity feature for
each location can be seen in Fig. 5.
In South Australia, there are a few hours per year where the
electricity price is either extremely high (9070 e/MWh) or low (−555
e/MWh). Most of the time, prices are also higher in South Australia
than in northern Chile or Denmark, which themselves have relatively
similar electricity price profiles.
Grid tariffs are added to the electricity spot price. The same grid
tariffs are applied everywhere to limit the comparison to the electric-
ity market price and weather profile. The tariffs are 16.65 e/MWh
consisting of 10.74 for transmission, 5.37 for distribution, and 0.54
for taxes based on an analysis from the Danish transmission system
operator Energinet adapted to electrolysis systems [52]. Like [15], the
grid connection cost is set at 66 e/kW (see Table 4). Other types of grid
pricing, such as dynamic tariffs and tariffs adapted to the local network
features that may include the costs of potential grid reinforcement
could also be investigated in further research.
4.2. Green ammonia synthesis plant description and layouts
𝐶𝑎𝑝𝑒𝑥+𝑂𝑝𝑒𝑥𝑦 Figs. 6 and 7 present the ammonia production process based on AEC
(1+𝑑)𝑦
and SOEC, respectively. Many other layouts leading to slightly higher
(1+𝑑)𝑦 or lower efficiencies but reduced investment could also be investigated.
One of them increases the efficiency by 3% with a relatively limited
(10)
Capex increase and is presented in Appendix C. The results obtained
with this improved layout are still very similar to those obtained with
the original layout of Fig. 7 and could be explored in further studies.

9
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Fig. 5. Spot price spreads in 2019 for DK1, South Australia and Chile.

Fig. 6. Ammonia synthesis plant with AEC.

Fig. 7. Ammonia synthesis plant heat integrated with SOEC.

10
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Table 5
Technical assumptions used in the optimisation model.
Input/Output Production rate Usable excess heat Minimal load Ramping ratea Electrical consumption
in/out kg𝑜𝑢𝑡 /kg𝑖𝑛 kWh𝑡ℎ /kg𝑜𝑢𝑡 % of capacityb % of capacityb /h kWh𝑒𝑙 /kg𝑜𝑢𝑡
HB + ASU (AEC)c H2 /NH3 5.29 0d 40f 20g 0.38
HB + ASU (SOEC)c H2 /NH3 5.29 0e 40f 20g 0.83
Electrolyser AECh H2 O/H2 0.07 7.07 0i 100j 51.50
Electrolyser SOECh H2 O/H2 0.07 0e 0i 100j 37.90
Desalination plantk –/H2 O – 0 0 100 0.004
Hydrogen storagel H2𝑖𝑛 /H2𝑜𝑢𝑡 – 0 9m 100 0.94n
Battery park kWh𝑖𝑛 /kWh𝑜𝑢𝑡 – 0 20 100 0.09o
a
Up and down.
b Check Table 6 for capacity units.
c Plant is under maintenance 10% of the year (876 h).

d Not recovered.

e
Due to heat integration with the ammonia synthesis plant.
f
Based on [53].
g
Based on [16].
h
See Appendix B for the detailed assumptions.
i It is assumed (optimistically) that the plant operator is able to put the electrolyser park on stand-by regularly.

j Both electrolysers are assumed to operate very dynamically and follow the power variations in real-time [54].

k Based on [55] using reverse osmosis. More recent desalination techniques such as membrane distillation [56] or MED-TVC [57] could also be investigated in further research.

l
In the form of hydrogen compressed at 100 bars in buried steel pipes [35].
m
9% of cushion gas is needed in the underground pipes storage [35].
n
0.94 kWh/kg of compressed hydrogen from 20 to 100 bars.
o
0.04 kWh of losses per kWh charged and 0.05 kWh losses per kWh discharged with a total round-trip efficiency of 91%, based on [58].

Table 6
Economic assumptions used in the optimisation model. The chemical engineering plant cost index was adopted to account for the inflation
effect except for batteries where inflation is from [59].
Unit𝑜𝑢𝑡 Capacity Capex Fixed cost Variable cost Lifetimea
e2019 /Capacity e2019 /Capacity/y e2019 /Unit𝑜𝑢𝑡 Years
HB + ASU (AEC) kgNH3 kgNH3 /h 4192 436 0 25
HB + ASU (SOEC) kgNH3 kgNH3 /h 4192 828 0 25
Electrolyser AECb kgH2 kgH2 /h (or kW) 56 467 (1096)c 1129 (22) 0 25
Electrolyser SOECb kgH2 kgH2 /h (or kW) 131 493 (3469)d 3945 (104) 0 25
Desalination plante kgH2 O kgH2 O /h 26 0f 0.0003 20
Hydrogen storageg kgH2 𝑠𝑡𝑜𝑟𝑒𝑑 kgH2 𝑠𝑡𝑜𝑟𝑒𝑑 461 1 0 50
Battery park kWh kWh 362h 5i 0 20j
a Lifetime is used to calculate the annuity factor (capital recovery factor) using a discount rate of 8%.
b See Appendix B for the detailed assumptions.
c Including stack replacement every 10 years (using stack net present value).

d
Including stack replacement every 5 years (using stack net present value).
e
Based on [55].
f
Included in the variable cost.
g
In the form of hydrogen compressed at 100 bars in buried steel pipes [35].
h Based on [60].

i Based on [6].

j
Based on [58].

4.3. Fuel plant techno-economic metrics 5.1. Least-cost ammonia production system using the DPS_Syst_Opt method

Techno-economical features of the ammonia synthesis plant and
electrolysers obtained from the literature and the process simulation
are summarised in Tables 5 and 6. Economic assumptions are relevant
for a large-scale ammonia plant (430 kt/y). The supplementary mate-
rial provides a detailed description of the process models, technology
choices and economic assumptions in Appendix B. All assumptions are
valid for present and near term future (2020–2025).
5. Results and discussion
This section is divided into two parts, one presenting and discussing
the results obtained with our method ‘‘DPS_Syst_Opt method’’ for dy-
namic power supply and system optimisation (described in Section 5.1),
the other comparing these results with those obtained using methods
and assumptions commonly used in the literature in Sections 5.4, 5.2,
and 5.3.
The results presented are based on wind, solar, and electricity price
time series from 2019 and may differ using time series from other
years. The plant operation and sizing are also made assuming perfect
foresight meaning that the plant operates ‘‘knowing’’ the future power
production and grid prices. 2 stage stochastic programming, myopic
foresight or robust optimisation methods could be used to decide on
investment decisions that consider weather and grid price uncertainties.
Taking more robust investment decisions adapted to a large variety of
possible events would likely oversize the infrastructures thus increasing
the total costs compared to the method used in this study.
The sizing of the system is driven by an annual demand for ammonia
of 430 kt.
5.1.1. Examples of optimal sizing and operation
As an example, Fig. 8 shows the yearly masses and energy balances
and the optimal system sizing obtained for off-grid (also known as
behind-the-meter) ammonia production in South Australia (wind/solar)
using an alkaline electrolyser.

11
N. Campion et al.
Fig. 8. Installed capacities and flows with Profile 3 (South Australia, wind/solar) in the off-grid set-up using an AEC electrolyser. The conversion losses (not represented in the
figure) at each transformation unit explain the mass difference between the input and output flows.
To produce 430 ktNH3 per year (representing the yearly energy
consumption of 2 of the world’s largest containerships −18 000 TEU-
cruising at 23 kts [61]), the least-cost off-grid solution in South Aus-
tralia (wind/solar), that has both good wind and solar resources, sug-
gests that about 500 MW of onshore wind and 1.66 GW of PV power
with a tracking system should be installed. This would require around
30 km2 (2700 ha) of land (using space requirements from [46]) while
the largest solar PV plant in the world (Bhadla solar park in India
of 2.5 GW) occupies 5665 ha. In addition, a bit more than 1 GW of
electrolysers would be required. Only 600 MW are installed globally in
2021 but several 1 GW alkaline electrolysers are planned in the coming
years [62]. Then, 100 MWh of batteries are used. In comparison,
the world’s largest battery park in 2017 has a 129 MWh capacity
(Tesla/Neoen hornsdale power reserve [63]). Finally, 140 tonnes of
pressurised hydrogen underground pipes need to be installed requiring
to excavate 0.1 km2 so 10 ha of land [35]. These comparisons give
an idea of the (challenging) sizes required to reach the same yearly
production as standard fossil ammonia plants.
Two examples of optimal plant operation on an hourly basis are
presented in Fig. 9. With a semi-islanded power supply in northern
Chile presented on the left of Fig. 9, the plant’s operation depends on
the electricity price assumptions and the renewable power availability.
The storage systems needed are reduced to a small hydrogen buffer (8
tonnes) without batteries. During the day, the fuel plant and electrol-
yser operate at full load using solar power to maximise the amount
of ammonia delivered at a low cost and fill up the hydrogen storage.
Some electricity has to be curtailed around midday when the power
production is at its maximum. During the night, the fuel plant and
electrolyser operate at a lower load to limit the grid-related expenses
and use some of the hydrogen stored. If the electricity price goes low
enough during the night, the plant operates again at a higher load
taking advantage of low electricity prices. In the off-grid power supply
set up in South Australia (wind/solar), presented on the right of Fig. 9,
renewable power production is much higher during the day than in the
semi-islanded case. This is done (a) to produce enough ammonia during
the day to compensate for the lower production at night and respect the
yearly ammonia demand; and (b) to fill up the hydrogen storage system
during the day to use it at night. The complementarity between wind
and solar power can be observed, the solar tracking system providing
most of the electricity during the day and the onshore turbines during
12
Renewable and Sustainable Energy Reviews 173 (2023) 113057
the night. Similar optimal off-grid operational behaviour has been
observed by [16], which also stresses the beneficial effects of combining
wind and solar power (with a tracking system) to increase the total
capacity factor.
5.1.2. Optimal sizing and operation in other configurations
Fig. 10 and Table 7 present the optimal sizing of the plant in the
other sites for different system configurations.
As seen in Fig. 10, offshore wind turbines are not invested in any
of the cases because their ratio cost/capacity factor is less favourable
than the solar and onshore wind power supply. Due to limited land
availability and local resistance, offshore turbines may however still
be deployed in some locations. Solar PV with 1-axis tracking is always
preferred to fixed support solar PV. Even if the tracking system is more
expensive, the total system cost decreases thanks to the higher capacity
factor, limiting the need for intermediate storage systems and infras-
tructure oversizing. Other studies like [26] consider fixed-axis solar PV
technology, which would tend to either reduce the share of solar PV in
optimal solutions or overestimate the fuel-production costs compared
to a solution using 1-axis tracking systems. Due to the relatively high
storage cost, running the fuel plant at full load is not economically
viable in the off-grid configuration. Operating the plant flexibly and
oversizing the infrastructure is preferable to deliver the yearly ammonia
demand. This effect has also been identified by [16,18,26], which
confirms the validity of the results obtained.
Fig. 10 shows that, in the off-grid case, at sites where both wind
and solar resources are available (wind/solar in South Australia), in-
vestments in both solar and wind power capacities are the least-cost
option. Thanks to the complementarity between wind and solar pro-
files, smaller plant capacities and less intermediate storage are needed.
Batteries are still used to power the ammonia plant at a minimal
load when no renewable electricity is produced. The site with limited
solar potential (wind+ in Denmark) still uses solar PV with a tracking
system due to its low cost and complementarity with the wind power
profile. Similar investment behaviour is also observed in [16] except
that firm-up electricity from the grid is used instead of batteries. When
the wind potential is high enough (in Denmark (wind+) and South
Australia (wind/solar)), the onshore turbine SP198-HH100 is always
implemented, as it provides the best balance between cost and capacity
factor among the proposed turbines. Studies focusing on wind-based
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Fig. 9. Example of optimised hourly system operation. Normalised H2 storage level (or electricity price) trend is obtained dividing the hourly quantity of hydrogen in the storage
(or hourly electricity price) by the highest capacity stored over the year (or highest electricity price) and then multiplying by a factor large enough to see the trend appearing on
the figure.

Fig. 10. Optimal installed capacities for a 430 kt/y ammonia PtX plant. The electrolyser capacity is expressed as the nominal power consumption.

Table 7
Optimal installed capacities.
Unit Off-grid Semi-islanded
Chile (solar+) Denmark (wind+) Australia (wind/solar) Chile (solar+) Denmark (wind+) Australia (wind/solar)
AEC SOEC AEC SOEC AEC SOEC AEC SOEC AEC SOEC AEC SOEC
NH3 plant tNH3 /h 62 62 72 68 68 66 57 55 55 55 60 56

H2 storage tH2𝑠𝑡𝑜𝑟𝑒𝑑 162 152 268 186 153 133 16 0 0 0 66 28

(buried pipes)

Batteries MWh 339 742 103 372 102 217 0 0 0 0 1 8

e-fuel production usually consider one specific type of turbine per
site like [16]. Here, in Denmark (wind+), investing in two turbines
with different hub-heights adapted for different wind speeds (Onshore
SP198-HH100 and SP198-HH150) leads to the lowest fuel production
cost, which shows that using a mix of different types of wind turbine
instead of one for a specific site should be considered in further studies.
Results also show that the power supply technology providing the
lowest fuel production cost is not always the one with the lowest
LCOE. In some cases (i.e., in Denmark (wind+), off-grid), a turbine
with higher LCOE but a high capacity factor (reported in Table A.11 in
the supplementary materials) lowers the total system cost by limiting
intermediate storage size and electrolyser oversizing. This example
stresses that choosing a power technology with the lowest LCOE to
estimate fuel costs has its limitations and can omit the potential gains
of using technologies with a higher LCOE and a higher capacity factor.
When a semi-islanded set-up is possible, the size of the storage
system can be drastically reduced as seen in Table 7. Batteries and
hydrogen storage are however still used in South Australia (wind/solar)
to cope with peak electricity prices. Furthermore, a small hydrogen
storage is used in northern Chile to avoid running the electrolyser at
night when the electricity prices are too high. When the solar resource
is sufficient as in Chile and Australia, the least-cost solution uses both

13
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Table 8
Full load hours and average load of the Haber–Bosch and electrolysis unit in the optimal set-up.
Off-grid Semi-islanded
Chile (solar+) Denmark (wind+) Australia (wind/solar) Chile (solar+) Denmark (wind+) Australia (wind/solar)
AEC SOEC AEC SOEC AEC SOEC AEC SOEC AEC SOEC AEC SOEC
Electrolyser FLHa 3217 3435 4696 5641 3714 5506 7190 7873 7809 7883 5510 7516
Electrolyser average load (%) 41 44 60 72 47 70 91 100 99 100 70 95
Haber–Bosch plant FLHa 6968 6960 5957 6309 6344 6562 7575 7873 7809 7883 7171 7641
Haber–Bosch plant average load (%) 88 88 76 80 80 83 96 100 99 100 91 97
a
Full load hours: virtual number of hours where the plant would produce at full load over the year, 876 h are used for maintenance time, so maximum FLH possible is 7884 h.

Fig. 11. Production and consumption of the different PtX units and overall efficiency calculated as the ratio between the energy content of the fuel produced (2.2 TWhNH3 LHV) and
the electricity produced. The electrical consumption due to desalination and battery losses are negligible compared to curtailment, ammonia plant, hydrogen storage compression
and electrolyser consumptions.

grid supply and local solar power production with 1 axis tracking as
seen in Fig. 10. In Denmark, where electricity prices are relatively low,
investing in wind turbines seems more expensive than using the grid
alone. However, these results do not consider the possible extra costs
due to grid reinforcement, and it is assumed that the infrastructure and
power production capacity is already there (and paid), which may not
be the case in reality. These results show that the most economical
solution for a grid-connected power-to-X plant required to produce a
given amount of fuel per year is overall to maximise its number of
full load hours and reduce the capital invested. Therefore, assumptions
made by some studies of PtX plant using mainly grid-surplus electricity
as a power source such as [39] should be taken carefully.
Table 8 shows that the ammonia plant has higher full load hours
(FLH) compared to the electrolyser due to the limited flexibility as-
sumed, which leads to the requirement for hydrogen storage. With a
stable power supply, the least-cost solution tends to reduce the plant’s
capacities and operate at a high or full load most of the time as seen
in Table 8. The plant is operated more flexibly in South Australia
(wind/solar), where electricity prices vary with amplitude.
5.1.3. System efficiency
Fig. 11 presents the shares of electricity production and consump-
tion per type of units in the different configurations. The overall system
efficiency is also indicated.
It can be seen on Fig. 11, that, in the off-grid set-up, some of
the renewable electricity produced is curtailed, especially for SOEC-
based systems. The very high Capex of SOEC (2020 numbers) push
the optimisation model to use batteries and high-capacity factor power
supply such as wind power to a higher extent in order to increase the
plant full load hours instead of oversizing the electrolyser. Typically, in
the off-grid case, in Denmark (wind+) and South Australia (wind/solar)
sites, more wind turbines are invested to increase the electrolyser load
at night but the model also invest into large PV plants to cover the
electricity demand during the day. In repeated occasions, when wind
and solar produce simultaneously, some of the excess electricity is used
to fill the batteries, but a significant share remains curtailed when the
battery capacities is exceeded. Investing into larger battery capacities
would be less profitable compared to curtailment. For this reason, even
with the higher SOEC efficiency, the overall PtX efficiency is only
improved by a few percent (up to 7%). Therefore, considering only the
electrolyser and processes efficiencies without analysing economic ra-
tional system investments decisions and dynamic operation behaviours
(similarly to [64]) may be misleading when comparing different elec-
trolysis technologies, especially in an off-grid set-up. Similar dynamics
explain the curtailment for AEC-based systems. If a grid connection is
possible, storage and electrolyser oversizing are no longer an issue,
and the system can fully benefit from the higher SOEC efficiency.
Therefore, in the semi-islanded case, the overall system efficiency is
higher, and the cost difference with AEC systems is reduced. For off-
grid solutions, the overall efficiencies obtained and presented in Fig. 11
are in general slightly lower than those encountered in the literature,
which lay between 50 and 55 for alkaline-based systems [17,27,33] and
between 70%–75% for SOEC-based systems [11,33]. This is mostly be-
cause the method used here considers the least-cost plant design, which

14
N. Campion et al.
Fig. 12. Cost breakdown and production cost. The costs include the annualised investment and the operation and maintenance costs (fixed and variable), and the purchase of
electricity. The fuel production cost is obtained by dividing the annualised system cost by the amount of fuel produced per year (around 2.2 TWh). The ammonia energy density
used is 18.6 MJ/kg LHV.
sometimes prefers a reduced overall efficiency (due to curtailment)
to limit excessive oversizing and storage costs. When the electricity
supply is constant, the overall efficiency or specific consumption is very
close to that obtained in other studies (for alkaline system, 50% LHV
efficiency in [32] or 10.43 kWh/kgNH3 in [18], among other examples).
As the electrolyser is the primary electricity consumer, the effect of
its efficiency on the overall ammonia production cost was studied in a
sensitivity analysis available in the supplementary material (Appendix
D.2). It was found that an increase of 1 kWh/kgH2 in electricity con-
sumption increases the production cost between 1.1 and 2.6 e/MWh,
so around 1–1.5%.
5.1.4. Carbon emissions
As shown in Fig. 11, when the grid connection is used, more than
50% of the electricity consumed originates from the grid in the least-
cost solution. Using the 2019 world average of the electricity carbon
intensity of 472 gCO2e/kWh [65] and the grid electricity consumption
levels indicated in Fig. 11, the emissions related to the grid electricity
usage would be around 440 gCO2e/kWhNH3 in the case with the
lowest grid electricity share (northern Chile using SOEC). With cleaner
electricity at 239 gCO2e/kWh (European Union levels in 2019 [65]),
emissions related to the grid would be around 220 gCO2e/kWhNH3 . As
a comparison, the very low sulphur oil used in the maritime industry
and the ammonia produced from natural gas emit around 325 and
510 gCO2e/kWh (Well-to-Wake) respectively [66]. This puts a question
over the sustainability of fuels relying partially on grid electricity with
today’s mix. Indeed, according to Ref. [3], the electricity must be at
least 90% from renewable origin to have a better climate footprint
than the fossil counterpart (for e-petrol, e-diesel and e-kerosene). The
quantification of GHG emissions could be further refined including the
emissions from the grid in real time but also the life cycle emissions
of the plant infrastructure. By excluding these aspects, the calculated
e-fuels carbon footprint is likely to be underestimated.
5.1.5. Production costs
Fig. 12 presents the annualised fuel production cost depending on
the location and electrolyser type and the system cost breakdown. In
all cases, producing the fuel off-grid is more expensive than using a
grid connection (semi-islanded) because the cost of the electrolyser
15
Renewable and Sustainable Energy Reviews 173 (2023) 113057
and power supply is doubled due to oversizing. The additional storage
systems slightly increase the costs. The cost of the desalination plant is
negligible compared to the rest as observed by [16,67].
The site with the highest solar potential has the lowest fuel pro-
duction cost. In the off-grid case, the production cost reaches 163
e/MWhNH3 (842 e/tNH3 ) using an AEC electrolyser. Adding a connec-
tion to the grid reduces the production cost down to 147 e/MWhNH3
(759 e/tNH3 ). This result differs from that in [26], which found the
cheapest locations were usually those with a high wind capacity factor.
However, [26] did not consider the 1-axis tracking PV technology,
which could explain the difference.
These values are relatively high compared to some of the values
reported in the literature: 463 e/tNH3 in [16], 416–664 e/tNH3 in [18]
or 431–528 e/tNH3 in [6], mostly due to different costs assumptions for
the infrastructure and different power profiles. However, the values of
this study are more in line with those reported by Nayak et al. [17]
and Pawar et al. [8], who reported around 670 and 723 e/tNH3 ,
respectively.
At 2020 price levels, even if SOEC systems have a higher efficiency,
the fuel plant relying on SOEC leads to ammonia production costs that
are significantly higher, especially when the power supply fluctuates.
A sensitivity analysis of SOEC Capex and Opex values can be found in
the supplementary material in Figure D.20. This sensitivity is used to
assess which cost reduction is needed to reach the cost level of AEC
systems. In the off-grid configuration, the SOEC-based system can be
cost-par with the current AEC system with a Capex reduction (including
stack replacement cost), of around 50% (down to 1300–1600 e/kW)
compared to today’s level. For systems relying fully on grid electricity,
only a 42% Capex reduction is needed (down to 2000 e/kW). This is
considering an AEC consumption of 51.5 kWh/kgH2 and a Capex of
1100 e/kW, including stack replacement. The efficiency considered for
the SOEC heat integrated with the ammonia plant is 37.9 kWh/kgH2 . A
50% SOEC cost reduction is in line with 2030 expectations.
5.2. Effect of plant flexibility assumptions
Among the Sim_LCOE and Syst_Opt types of method, the assump-
tions about plant and electrolyser flexibility can differ significantly
from one study to another. As mentioned in [16], these assumptions can
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

Table 9
Technical assumptions used in the optimisation model.
Scenario label Minimal load (%) Ramping constraint Comments
AEC HB
AEC[20–100%]_HB[100%] 20 100 – Worst case scenario
AEC[20–100%] 20a 40 20%/h Effect of a less flexible AEC
HB[100%] 0 100b – Effect of a non-flexible HB plant alone
HB_NoRamping 0 40 No Effect of the ramping constraint
HB[10–100%] 0c 10d 20%/h Best case scenario
AEC[0–100%]_HB[40–100%] 0 40 20%/h Base case used for comparison
a
As Buttler et al. mentioned in his review, the most common minimal load used for AEC is 20% to avoid hydrogen cross-over [54].
b
Motivated by [18], who stated that the Haber–Bosch loop should operate at a constant load to avoid catalyst damage and efficiency
loss.
c
Assuming the possibility to put the electrolyser on stand-by regularly with negligible extra costs.
d
Some industries claim that reducing the minimal load to 10% is feasible [68].

Fig. 13. Effect of flexibility assumptions on fuel production cost compared to base case AEC[0-100%]_HB[40-100%].

significantly impact the results. To study that influence, one reference
case is compared to various scenarios addressing the assumptions that
are usually encountered in the literature regarding minimal loads and
ramping rates. The different cases studied are presented in Table 9. To
simplify the comparison, the focus was on the AEC-based system. The
differences in terms of fuel production costs are presented in Fig. 13.
In Fig. 13, it can be observed that, when the plant has access to the
grid, the flexibility of the electrolyser and ammonia plant have a limited
impact on the production cost. It does however have a noticeable, but
small, influence when the electricity market price has a high variation
amplitude with price peaks as in South Australia (wind/solar).
In the off-grid set-up, when production relies primarily on solar
power, an electrolyser with a minimal load of 20% (AEC[20–100%])
increases the production cost up to 36.6% compared to the reference.
This is mainly because very large and expensive battery capacities are
needed to operate at minimal load, and the power supply needs to be
oversized accordingly to fill the batteries. A non-flexible electrolyser
in locations where the power supply is more regular (South Australia,
wind/solar) limits the additional costs to ‘‘only’’ an 19.5% increase.
A non-flexible Haber–Bosch plant (HB[100%]) with a fully flexible
electrolyser increases the cost by around 20% compared to the base
case and by 25% with the wind/solar profile (in South Australia). The
costs increase because running the ammonia plant at full load requires
larger hydrogen storage and hence larger electrolyser and power supply
capacities. On the other hand, additional HB plant flexibility (down
to 10% instead of 40%) (HB[10–100%]) does not reduce the cost
significantly.
Combining an electrolyser that is not fully flexible with a non-
flexible ammonia plant as a ‘‘worst-case scenario’’ (AEC[20–100%]
_HB[100%]) increases the production price by nearly 50% in sites
where the power supply depends mostly on one source and 40% in the
Australia (wind/solar) case.
Removing the ramping constraint of the HB plant (HB_NoRamping)
gives results very similar to the base case. When considering time steps
of one hour, the constraint is usually non-binding and could be removed
to improve the computational time. The possibility that the constraint
may have a stronger effect with a smaller time resolution could be
investigated in further research.
To better understand the minimal load effect on the costs, multiple
optimisation runs are made with different HB minimal load values and
the costs obtained are displayed in Fig. 14. It can be observed that at
40% minimal load, the costs even out for all cases and for solar even
earlier at 60%.
Fig. 14 shows that in the off-grid case, the ammonia plant’s flexi-
bility has a key influence on production costs, especially for ammonia
plant that have a minimal load above 60%. Armijo et al. [16] observed
cost-flexibility curves with similar trends and stated that ‘‘H2 storage
is the key cost driver related to the lack of flexibility’’. To validate
this observation, the cost breakdown of the optimal energy system was
calculated for every ammonia plant’s minimal load and can be found in
Figure E.22 in the supplementary materials. Results show that hydrogen
storage indeed plays a major role in increasing costs. However, what
mostly drives the costs upwards is actually the need to invest in larger
power supply capacities (and larger electrolysers) to provide enough

16
N. Campion et al.
Fig. 14. Effect of HB plant minimal load on off-grid ammonia production cost using a fully flexible AEC. The ramping constraint is still applied.
hydrogen for both the fuel plant and the storage facility. Another
interpretation of these results, not mentioned by [16], is that the HB
plant’s flexibility has a strong influence on the costs, but only down to
a specific minimum load that depends on the type of power profile.
For example, in northern Chile (solar+), which relies on relatively
regular solar power, operating the plant below a 60% load does not
significantly reduce the costs. With the production sites relying more on
wind power (Denmark (wind) and South Australia (wind/solar)), which
has a less regular profile pattern, operating at a lower load is more
beneficial. However, below a 40% minimal load the cost reductions
are also less significant. Indeed, when the system is sized to produce
a given amount of fuel per year, the optimal operation is to run the
ammonia plant with a high number of full load hours to make the
best use of the available capacities and thus avoid excessive oversizing.
Therefore, with an optimal operation and a fully flexible ammonia
plant, the actual number of hours at which the ammonia plant would
operate at low loads is very limited (167, 394, and 225 h below a 10%
HB load in northern Chile, Denmark and South Australia (wind/solar)
respectively), so a very flexible ammonia plant does not significantly
improve the plant’s economic performance. In addition, this study did
not consider the extra plant costs of a chemical process designed to be
more flexible. With the limited benefits of additional flexibility, these
extra costs may actually make systems using a very flexible plant more
expensive than ones with ‘‘standard’’ flexibility. This result is a rather
positive outcome for the industry. Indeed, if the industrial research
should still focus on designing flexible fuel plants for off-grid systems,
the challenge is limited to the design of ‘‘flexible enough’’ plants that
probably require a lower amount of technical innovations than fully
flexible plants.
Summing up, the assumed flexibilities of plants are highly important
for the fuel production costs of off-grid solutions and it is therefore key
to consider these features when modelling e-fuel production off-grid or,
in a lower extent, in systems with highly fluctuating electricity prices.
However, a very flexible fuel plant able to operate below 40% load has
a limited impact on economic performances thus further research on
very flexible fuel plant processes may not so heavily be needed.
5.3. Effect of assumptions regarding by-product sales
Some studies, like [69], consider the sale of by-products such as
heat and/or oxygen in their analysis. Using the model developed in this
study, we study the influence of by-product selling price on ammonia
17
Renewable and Sustainable Energy Reviews 173 (2023) 113057
cost estimates. Oxygen produced through electrolysis has a very high
purity, making it suitable for any application, including medical use.
It is assumed that 7.8 kgO2 is recovered and sold per kgH2 produced.
The oxygen prices reported in the literature vary between 27 and
150 e/tO2 [69,70]. Quantities of recoverable heat from AEC have
been estimated at 7.07 MWh𝑡ℎ /tH2 [71]. The heat has a relatively low
temperature, still suitable for a local district-heating network. Low-
temperature district heating prices are reported between 23 and 52
e2019 /MWh [69,72,73].
Cost estimate reductions that can be expected from the sale of heat
and oxygen are presented in Fig. 15.
It is observed in Fig. 15 that selling oxygen and/or heat strongly
influences the cost, which can be reduced down to 30%–40% if heat
and oxygen sales are cumulated and selling prices are in the highest
range. Oxygen sale represents the highest costs reduction potential
which is between 5 and 29% if all the oxygen can be sold at the prices
stated in the literature. Costs reduction related to heat sale are between
4 and 10%.
The possibility of selling excess renewable electricity was also stud-
ied. Here it is assumed that the renewable plant can be connected
to the grid without extra cost and the surplus electricity sold at the
market price instead of curtailing it. Results showed that it does not
make a significant difference, as excess electricity is usually generated
when the price of grid electricity is low. Hence, in the studied sites, the
benefits from the electricity sale reduce the fuel costs by less than 1%.
In conclusion, selling by-products entails are large cost-decrease
potential, which should be taken into account when modelling the
e-fuel production plants and when deciding on locations.
5.4. Estimate of ammonia production costs using FixEl_P methods
The objective of this section is to identify and quantify the method
bias that occurs when calculating the costs using the system opti-
misation method presented in this work (DPS_Syst_Opt) or a fixed
electricity price method (FixedEl_P). Unfortunately, the simulation
and LCOE methods (Sim_LCOE) cannot be directly compared with
the DPS_Syst_Opt method (presented in Section 3), as the Sim_LCOE
methods results are driven by the installed renewable capacity and not
by the fuel demand.
The FixEl_P methods estimate the plant’s annual electricity con-
sumption for a given number of operating hours and calculate the
electricity cost by multiplying it by a fixed electricity price. The fuel
N. Campion et al.
Fig. 15. Cost-reduction potentials for heat and oxygen sale. Reference price to calculate the reduction percentages is 147 e/MWhNH3 obtained in northern Chile with a semi-islanded
power supply.
cost is obtained by adding fuel plant investment expenditure and other
operational and maintenance costs. Two cases are differentiated: one
where the PtX plant is powered exclusively by the grid (like [11,31]
or [29]), the other where it is powered exclusively by variable re-
newable sources (like [30,36] or [37]). To make a methodological
comparison possible, the FixedEl_P method is reproduced using exactly
the same data-set as is used with the DPS_Syst_Opt presented in the
previous sections. The electricity cost is calculated in the three locations
presented in Section 4.1 with two different methods depending on
whether the PtX plant is powered by the grid or by wind and solar
PV.
If the PtX plant is powered exclusively by the grid, the electricity
costs are calculated in terms of the yearly average of the hourly
electricity prices presented in Section 4 (similarly to [11] or [31]). No
intermediate storage is used, and the plant does not operate flexibly.
This method is labelled ‘‘FixedAvGrid’’ in Fig. 16.
In the case where the PtX plant is powered by wind or solar power
alone, the electricity price is determined using the lowest LCOE among
the power supply technologies for each site using the data presented in
Section 4.1. The size of the electrolyser and ammonia plant required to
produce 430 ktNH3 /y is determined by the number of operating hours
at full load derived from the wind or solar capacity factor (as done
by [36]). No storage system is considered.
The calculated LCOEs and full load hours are presented in Table 10.
The investment and operating costs are derived from the plant sizing
and added on top of the electricity cost to obtain the ammonia cost
labelled ‘‘MinLCOE_LowFLH’’ in Fig. 16.
Other studies arbitrarily set the plant operating hours (for example,
to 85% of yearly operation as a base case in [30] or 50% in [37]). In the
method labelled ‘‘MinLCOE_HighFLH’’, the ammonia cost is calculated
using a high FLH (7884 FLH) and the lowest LCOE of renewables for
the electricity price.
Fig. 16, shows the comparison between the results obtained with
the DPS_Syst_Opt method and the other methods.
The ‘‘off-grid’’ case in Fig. 16 illustrates the method bias between
the fixed electricity price and the DPS_Syst_Opt methods in an off-
grid set-up. Results show that using the minimal LCOE available while
assuming a high number of full load hours underestimates the fuel
production cost by more than 50% compared to the DPS_Syst_Opt
method. This method is not used as such in the literature, but studies
assuming that renewable electricity can be available with high full load
18
Renewable and Sustainable Energy Reviews 173 (2023) 113057
hours (for example, through the grid) at a cost close to PV or wind
power LCOE like in [30] are likely to generate a similar methodological
bias. This can be corrected by assuming a lower fuel plant full load hour
calculated from the renewable power capacity factor (like in [36]).
However, with this method, the estimated costs can be up to 30%
over-estimated compared with the DPS_Syst_Opt method. Indeed, the
MinLCOE_LowFLH method do not consider the possibility of optimally
combining different power sources that complement each other, nor
the use of intermediate storage systems which finally underestimate the
plant full load hours that would remain economically feasible. The type
of electrolyser used seems to have a limited influence on the percentage
costs difference between the two methods.
The ‘‘only-grid connected’’ case in Fig. 16 illustrates the limited
methodological bias between the averaged grid electricity price method
(‘‘FixedAvGrid’’ for example used by [11]) and optimisation methods
that consider optimal flexible plant operation (DPS_Syst_Opt method).
Results show that, when the amplitude of electricity price variation is
low, as in Denmark (wind+) or northern Chile (solar+), the calculated
costs are very similar between the two methods with around 1%
difference. Indeed, with the fixed electricity price method, the PtX plant
runs at a constant load and is sized at the minimum to respect the
annual demand. Given the relatively stable electricity prices, stable
ammonia production is also favoured by the optimisation method, and
the plant is run flexibly for only a few hours per year (which explains
the remaining cost difference between the two methods). With a higher
electricity price spread like in South Australia (wind/solar) (see Fig. 5),
the average electricity price method calculates production costs up to
5% higher compared to the DPS_Syst_Opt method. This is because using
a fixed average price and full load production omits the savings that
could be made when operating the plant at a lower load or using
an intermediate storage system when the electricity price is high. An
electricity market price with high variation is specific to Australia,
but it is also likely to occur in future energy systems with a high
variable renewable energy penetration [41]. Thus, studies focusing
on late future and using a grid supply should consider the grid-price
variation for more accuracy.
Summing up, the method used has a high impact on the calculated
fuel production costs, in particular when analysing off-grid solutions.
It also has some impact, albeit less, when looking at grid connected
solutions with high fluctuations in electricity prices.
N. Campion et al.
Table 10
Full load hours and LCOE used for the methods MinLCOE_LowFLH and MinLCOE_HighFLH.
Location (profile type) Northern Chile (solar+)
Technology used Solar PV tracking
Capacity factor (%) 30.3
LCOE (e/MWh) 25.1
Low FLHb 2400
High FLHc 7884d
a Using a turbine of 237 W/m2 specific power and 100 m hub height.
b
Ammonia plant full load hours for the MinLCOE_LowFLH method, including maintenance time.
c
Ammonia plant full load hours for the MinLCOE_HighFLH method, including maintenance time.
d
Operation 90% of the year at full load.
Fig. 16. Fuel production costs using methods of FixEl_P and comparison with ‘‘dynamic power supply and system optimisation’’ method: ‘‘only grid connected’’ and ‘‘off-grid’’
scenarios.
6. Conclusion
This paper has focused on the modelling and quantification of
the costs related to e-ammonia production using and comparing dif-
ferent methods of estimating e-fuel costs. E-ammonia is synthesised
from electrolytic hydrogen produced using renewable power and/or
electricity from the grid. Different electrolyser types, power connec-
tion set-ups and production sites are compared. Building upon typical
models features from the previous literature, an optimisation model
was developed and used to determine least-cost e-fuel plant designs
and estimate the consequent fuel costs. Using the same input data,
the e-ammonia cost was calculated using other methods often used in
the literature (referred to as ‘‘fixed electricity price methods’’ in this
analysis) and compared with the results obtained using the developed
optimisation method. The key questions addressed in this study were:
(1) which system design and cost estimate can be expected with an
advanced individual plant model with local energy system integration,
and (2) how does it compare with the results obtained from typi-
cal ‘‘process modelling and economic analysis’’ using fixed electricity
prices, in calculating the fuel costs? Thus, this paper provides power-
to-X plant design recommendations based on a fast solving optimisation
model, and points to the bias that may occur while using other simpler
methods to estimate e-fuel costs.
Results concerning the power-to-X plant design and costs show that,
for off-grid e-fuel production systems, the optimal power mix always
includes solar PV with one axis tracking even if the solar potential
is limited (15% capacity factor). In some places, using different types
of onshore turbines in the same site provides the cheapest option.
19
Renewable and Sustainable Energy Reviews 173 (2023) 113057
Denmark (wind+) South Australia (wind/solar)
Wind Onshorea Solar PV tracking
36.0 24.6
51.3 30.9
3200 2150
7884d 7884d
Notably, the technologies with the lowest LCOE were not always those
minimising the overall system costs, as high capacity factors may be
more important. The electrolyser park should be as flexible as possible
to avoid prohibitive battery costs. A flexible ammonia plant gives
significant fuel cost reduction down to a minimal load of ca. 40%.
Below this, there is no further cost gain. Thus, designing fully flexible
processes may not be a research objective so relevant for the industry.
The overall power to ammonia efficiency lies between 47%–48% for
AEC-based systems and 52%–54% for SOEC-based systems depending
on the production site. The lowest off-grid ammonia production cost
obtained was 163 e/MWhNH3 (842 e/tNH3 ) in northern Chile using an
alkaline electrolyser powered with 1-axis tracking solar PV.
A mix of local renewable power production and a connection to the
grid (semi-islanded set up) is the cheapest option to produce e-fuels
when potential grid-reinforcement costs are not considered. Overall
system efficiency is also increased compared to an off-grid system
thanks to reduced curtailment and ranges between 50%–51% for AEC-
based systems and 63%–65% for SOEC-based systems. The production
costs are reduced to 147 e/MWhNH3 (760 e/tNH3 ) in northern Chile
using an alkaline electrolyser and an optimal mix of grid electricity and
on-site solar power with one axis tracking. In comparison, in Western
Europe, ‘‘grey’’ ammonia had a selling price of around 40 e/MWh
(200 e/tNH3 ) in 2019 and 290 e/MWh (1500 e/tNH3 ) in 2022 [74].
Thus, if the fossil fuel prices remain globally at high levels, the e-fuels
may become competitive even with limited governments support. The
fuel cost would still remain four times higher compared to previous
price levels which may be economically unsustainable in the long run.
However, the electrolysis technology which today represents between
N. Campion et al.
30 (Chile, AEC) and 66% (Chile, SOEC) of the total costs is still an
‘‘immature’’ technology and suggests a large cost reduction potential
which could be achieved via mass production gains and technology
improvements. Typically, the cost of e-ammonia could drop to 110
e/MWh (570 e/tNH3 ) if SOEC becomes available at a cost of 460
e/kW as predicted in some studies [37]. In addition, in the semi-
islanded case, the size of the infrastructures and intermediate storage is
significantly reduced, making that solution potentially more appealing
from an industrial and technical point of view. However, in the least-
cost solution, the grid provides more than 50% of the electricity.
With the 2019 global grid carbon intensity levels (472 kgCO2/kWh),
the e-ammonia carbon footprint would be similar to ‘‘grey’’ ammonia
produced from natural gas.
Results on method bias identification and quantification show that,
for systems that are fully grid-connected, with electricity prices fluc-
tuating hourly, approximating the electricity cost by an average grid
electricity price with a full load plant operation (FixEl_P methods) gives
results very close (max 1% cost difference) to the optimisation method
if the hourly grid price variation amplitude is limited. With larger
variations in price amplitude, using an average electricity price disre-
gards the potential savings due to flexible plant operation and slightly
overestimates the production costs (but only up to 4% in the cases
studied). Using power purchase agreements ensuring, for example, that
the electricity is ‘‘green’’ should be addressed carefully. Assuming that
‘‘green electricity’’ is available at all times at a price similar to solar or
wind LCOE will underestimate the cost of e-fuels significantly. For off-
grid e-fuel production systems, estimating the production costs using
the renewable technology LCOE and deriving the fuel-plant operating
hours from the power capacity factors overestimates the costs by up
to 30% compared to the optimisation method because potential costs
savings due to the use of complementary power sources or intermediate
storages are disregarded.
In the off-grid set up, assumptions made on the ammonia plant flexi-
bility have a significant impact on the results. An ammonia plant with a
minimal load of 40%–60% (depending on the weather profile) provides
significant cost reductions. Nevertheless, ammonia plants that are more
flexible, with minimal loads below these levels, reduce the system costs
only marginally. If a grid connection is available, the flexibilities of
the ammonia plant and electrolyser have a limited effect on the system
costs (with 2019 electricity prices). The consideration of ammonia plant
ramping constraints did not have a significant influence on the results.
Assumptions on by-products sale, such as heat and oxygen, entails large
cost-decrease potential (up to 40%), thus the possibility to sell these
products should be looked at carefully by industrials before choosing
a site for PtX production. Considering the sale of surplus electricity at
the market price does not significantly reduce the cost estimates as the
price will be low when there is a surplus.
Since this study has mostly focused on comparing fast-solving meth-
ods, further development could include the comparison with more
complex methods that use integer or non-linear programming. The
methods that handle weather profile uncertainty such as stochastic or
robust optimisation could also be compared.
Finally, the results show that using both stable power supply like
grid and local renewable power is the cheapest option, but runs the
risk of having counterproductive effects on greenhouse gas emissions
reduction as long as the electricity blend on the grid includes large
fractions of fossil power. This leads to the conclusion that the grid usage
for e-fuel production should be carefully controlled. In further studies,
the pertinence of control mechanisms such as green certificates, power
purchase agreements, forced off-grid solutions or carbon taxes could
then be assessed in terms of costs, technical feasibility and life cycle
assessment performances.
20
Renewable and Sustainable Energy Reviews 173 (2023) 113057
CRediT authorship contribution statement
Nicolas Campion: Conceptualization, Methodology, Software, Val-
idation, Investigation, Data curation, Writing – original draft, Visual-
ization. Hossein Nami: Methodology, Software, Investigation, Visu-
alization, Writing – original draft. Philip R. Swisher: Methodology,
Investigation, Data curation, Writing – original draft. Peter Vang Hen-
driksen: Conceptualization, Project administration, Funding acquisi-
tion, Writing – review & editing. Marie Münster: Conceptualization,
Project administration, Supervision, Funding acquisition, Writing –
review & editing.
Declaration of competing interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Data availability
The input data files and the code used for this study are freely
available at
Input data (Original data) (GitHub)
Acknowledgement
This paper is published as part of the MarE-fuel project funded
by the Danish Maritime Fund and the Lauritzen Fund. We thank the
partners of the project for their cooperation and inspiration.
Appendix A. Supplementary data
Supplementary material includes wind-turbines data, extended fuel
plant description and additional sensitivity analyses.
Supplementary material related to this article can be found online
at https://doi.org/10.1016/j.rser.2022.113057.
References
[1] Intergovernmental Panel on Climate Change. Sixth assessment reports. 2022, URL
https://www.ipcc.ch/.
[2] Connolly D, Mathiesen BV, Ridjan I. A comparison between renewable transport
fuels that can supplement or replace biofuels in a 100% renewable energy system.
Energy 2014;73:110–25. http://dx.doi.org/10.1016/j.energy.2014.05.104.
[3] Ueckerdt F, Bauer C, Dirnaichner A, Everall J, Sacchi R, Luderer G. Potential
and risks of hydrogen-based e-fuels in climate change mitigation. Nature Clim
Change 2021;11(5):384–93. http://dx.doi.org/10.1038/s41558-021-01032-7.
[4] USEnergy Information Administration. Few transportation fuels surpass the
energy densities of gasoline and diesel. 2013, URL https://www.eia.gov/
todayinenergy/detail.php?id=9991.
[5] Harnesk D. Adding fuel to the fire : North-South dynamics in the geographies of
transport energy: the case of EU biofuels (Ph.D. thesis), Lund University; 2018, p.
212, URL http://portal.research.lu.se/portal/en/publications/adding-fuel-to-the-
fire(63b9e062-11c3-4223-b697-653262f0ef22).html.
[6] Ikäheimo J, Kiviluoma J, Weiss R, Holttinen H. Power-to-ammonia in future
north European 100 % renewable power and heat system. Int J Hydrogen Energy
2018;43(36):17295–308. http://dx.doi.org/10.1016/j.ijhydene.2018.06.121.
[7] Blanco H, Nijs W, Ruf J, Faaij A. Potential for hydrogen and Power-to-
Liquid in a low-carbon EU energy system using cost optimization. Appl Energy
2018;232(October):617–39. http://dx.doi.org/10.1016/j.apenergy.2018.09.216.
[8] Pawar ND, Heinrichs HU, Winkler C, Heuser PM, Ryberg SD, Robinius M,
Stolten D. Potential of green ammonia production in India. Int J Hydro-
gen Energy 2021;46(54):27247–67. http://dx.doi.org/10.1016/j.ijhydene.2021.
05.203.
[9] Clausen LR, Houbak N, Elmegaard B. Technoeconomic analysis of a methanol
plant based on gasification of biomass and electrolysis of water. Energy
2010;35(5):2338–47. http://dx.doi.org/10.1016/j.energy.2010.02.034.
[10] Pérez-Fortes M, Schöneberger JC, Boulamanti A, Tzimas E. Methanol synthesis
using captured CO2 as raw material: Techno-economic and environmental as-
sessment. Appl Energy 2016;161:718–32. http://dx.doi.org/10.1016/j.apenergy.
2015.07.067.
N. Campion et al.
[11] Zhang H, Wang L, Van herle J, Maréchal F, Desideri U. Techno-
economic comparison of green ammonia production processes. Appl En-
ergy 2020;259(November):114135. http://dx.doi.org/10.1016/j.apenergy.2019.
114135.
[12] Bos MJ, Kersten SR, Brilman DW. Wind power to methanol: Renewable methanol
production using electricity, electrolysis of water and CO2 air capture. Appl
Energy 2020;264(February):114672. http://dx.doi.org/10.1016/j.apenergy.2020.
114672.
[13] Mignard D, Sahibzada M, Duthie JM, Whittington HW. Methanol synthesis from
flue-gas CO2 and renewable electricity: A feasibility study. Int J Hydrogen Energy
2003;28(4):455–64. http://dx.doi.org/10.1016/S0360-3199(02)00082-4.
[14] Matzen M, Alhajji M, Demirel Y. Chemical storage of wind energy by renewable
methanol production: Feasibility analysis using a multi-criteria decision matrix.
Energy 2015;93:343–53. http://dx.doi.org/10.1016/j.energy.2015.09.043.
[15] Hank C, Gelpke S, Schnabl A, White RJ, Full J, Wiebe N, Smolinka T, Schaadt A,
Henning HM, Hebling C. Economics & carbon dioxide avoidance cost of methanol
production based on renewable hydrogen and recycled carbon dioxide-power-to-
methanol. Sustain Energy Fuels 2018;2(6):1244–61. http://dx.doi.org/10.1039/
c8se00032h.
[16] Armijo J, Philibert C. Flexible production of green hydrogen and ammonia
from variable solar and wind energy: Case study of Chile and Argentina. Int
J Hydrogen Energy 2020;45(3):1541–58. http://dx.doi.org/10.1016/j.ijhydene.
2019.11.028.
[17] Nayak-Luke R, Bañares-Alcántara R, Wilkinson I. "Green" ammonia: Impact of
renewable energy intermittency on plant sizing and levelized cost of ammonia.
Ind Eng Chem Res 2018;57(43):14607–16. http://dx.doi.org/10.1021/acs.iecr.
8b02447.
[18] Osman O, Sgouridis S, Sleptchenko A. Scaling the production of renewable am-
monia: A techno-economic optimization applied in regions with high insolation. J
Clean Prod 2020;271:121627. http://dx.doi.org/10.1016/j.jclepro.2020.121627.
[19] Schulte Beerbühl S, Fröhling M, Schultmann F. Combined scheduling and ca-
pacity planning of electricity-based ammonia production to integrate renewable
energies. European J Oper Res 2015;241(3):851–62. http://dx.doi.org/10.1016/
j.ejor.2014.08.039.
[20] Lythcke-Jørgensen C, Ensinas AV, Münster M, Haglind F. A methodology for
designing flexible multi-generation systems. Energy 2016;110:34–54. http://dx.
doi.org/10.1016/j.energy.2016.01.084.
[21] Mallapragada DS, Gençer E, Insinger P, Keith DW, O’Sullivan FM. Can industrial-
scale solar hydrogen supplied from commodity technologies be cost competitive
by 2030? Cell Rep Phys Sci 2020;1(9). http://dx.doi.org/10.1016/j.xcrp.2020.
100174.
[22] Sánchez A, Martín M. Optimal renewable production of ammonia from water
and air. J Clean Prod 2018;178:325–42. http://dx.doi.org/10.1016/J.JCLEPRO.
2017.12.279.
[23] Allman A, Palys MJ, Daoutidis P. Scheduling-informed optimal design of systems
with time-varying operation : A wind-powered ammonia case study. Am Inst
Chem Eng 2018;65(7). http://dx.doi.org/10.1002/aic.16434.
[24] Grahn M, Malmgren E, Korberg AD, Taljegard M, Anderson JE, Brynolf S,
Hansson J, Skov IR, Wallington TJ. Review of electrofuel feasibility - Cost and
environmental impact. Prog Energy 2022;4(3). http://dx.doi.org/10.1088/2516-
1083/ac7937.
[25] Brynolf S, Taljegard M, Grahn M, Hansson J. Electrofuels for the transport
sector: A review of production costs. Renew Sustain Energy Rev 2018;81(May
2017):1887–905. http://dx.doi.org/10.1016/j.rser.2017.05.288.
[26] Nayak-Luke RM, Bañares-Alcántara R. Techno-economic viability of islanded
green ammonia as a carbon-free energy vector and as a substitute for conven-
tional production. Energy Environ Sci 2020;13(9):2957–66. http://dx.doi.org/10.
1039/d0ee01707h.
[27] Bañares-Alcántara R, Dericks III G, Fiaschetti M, Philipp Grünewald, Lopez JM,
Tsang E, Yang A, Ye L, Zhao S. Analysis of islanded ammonia-based energy
storage systems. Tech. rep., University of Oxford; 2015, p. 1–150.
[28] Sánchez M, Amores E, Abad D, Rodríguez L, Clemente-Jul C. Aspen plus model of
an alkaline electrolysis system for hydrogen production. Int J Hydrogen Energy
2020;45(7):3916–29. http://dx.doi.org/10.1016/j.ijhydene.2019.12.027.
[29] Collet P, Flottes E, Favre A, Raynal L, Pierre H, Capela S, Peregrina C. Techno-
economic and Life Cycle Assessment of methane production via biogas upgrading
and power to gas technology. Appl Energy 2017;192:282–95. http://dx.doi.org/
10.1016/j.apenergy.2016.08.181.
[30] Fúnez Guerra C, Reyes-Bozo L, Vyhmeister E, Jaén Caparrós M, Salazar JL,
Clemente-Jul C. Technical-economic analysis for a green ammonia produc-
tion plant in Chile and its subsequent transport to Japan. Renew Energy
2020;157:404–14. http://dx.doi.org/10.1016/j.renene.2020.05.041.
[31] Gomez JR, Baca J, Garzon F. Techno-economic analysis and life cycle assessment
for electrochemical ammonia production using proton conducting membrane.
Int J Hydrogen Energy 2020;45(1):721–37. http://dx.doi.org/10.1016/j.ijhydene.
2019.10.174.
[32] Morgan ER, Manwell JF, McGowan JG. Sustainable ammonia production from
U.S. offshore wind farms: A techno-economic review. ACS Sustain Chem Eng
2017;5(11):9554–67. http://dx.doi.org/10.1021/acssuschemeng.7b02070.
21
Renewable and Sustainable Energy Reviews 173 (2023) 113057
[33] Nami H, Butera G, Campion N, Frandsen HL, Hendriksen PV. MarE-fuel: Energy
efficiencies in synthesising green fuels and their expected cost. Tech. rep., Den-
mark Tekniske Universitet; 2021, URL https://orbit.dtu.dk/en/projects/electro-
fuels-for-long-range-maritime-transport.
[34] Kiemel S, Smolinka T, Lehner F, Full J, Sauer A, Miehe R. Critical materials for
water electrolysers at the example of the energy transition in Germany. Int J
Energy Res 2021;45(7):9914–35. http://dx.doi.org/10.1002/er.6487.
[35] Papadias DD, Ahluwalia RK. Bulk storage of hydrogen. Int J Hydrogen Energy
2021;46(70):34527–41. http://dx.doi.org/10.1016/j.ijhydene.2021.08.028.
[36] Bicer Y, Dincer I. Exergoeconomic analysis and optimization of a concentrated
sunlight-driven integrated photoelectrochemical hydrogen and ammonia produc-
tion system. Int J Hydrogen Energy 2019;44(34):18875–90. http://dx.doi.org/
10.1016/j.ijhydene.2018.10.074.
[37] Nami H, Rizvandi OB, Chatzichristodoulou C, Hendriksen PV, Frandsen HL.
Techno-economic analysis of current and emerging electrolysis technologies for
green hydrogen production. Energy Convers Manage 2022;269:116162. http:
//dx.doi.org/10.1016/J.ENCONMAN.2022.116162.
[38] Ozturk M, Dincer I. An integrated system for ammonia production from re-
newable hydrogen: A case study. Int J Hydrogen Energy 2021;46(8):5918–25.
http://dx.doi.org/10.1016/j.ijhydene.2019.12.127.
[39] Lee B, Lim D, Lee H, Lim H. Which water electrolysis technology is appropriate?:
Critical insights of potential water electrolysis for green ammonia production.
Renew Sustain Energy Rev 2021;143(March):110963. http://dx.doi.org/10.1016/
j.rser.2021.110963.
[40] Statista. Ammonia plant production capacity in the United States in 2019,
by facility. 2021, URL https://www.statista.com/statistics/1266392/ammonia-
plant-capacities-united-states/.
[41] Campion N, Backer M, Swisher P, Münster M. MarE-fuel: LCOE and optimal
electricity supply strategies for P2X plants. Tech. rep., Denmark Tekniske Uni-
versitet; 2021, URL https://orbit.dtu.dk/en/projects/electro-fuels-for-long-range-
maritime-transport.
[42] DTU wind and World Bank, Global Wind Atlas. URL https://globalwindatlas.
info/.
[43] World Bank, Global Solar Atlas. URL https://globalsolaratlas.info/map.
[44] Koivisto M, Das K, Guo F, Sørensen P, Nuño E, Cutululis N, Maule P. Using time
series simulation tools for assessing the effects of variable renewable energy
generation on power and energy systems. Wiley Interdiscip Rev: Energy Environ
2019;8(3):1–15. http://dx.doi.org/10.1002/wene.329.
[45] Pfenninger S, Staffell I. Long-term patterns of European PV output using 30
years of validated hourly reanalysis and satellite data. Energy 2016;114:1251–65.
http://dx.doi.org/10.1016/j.energy.2016.08.060.
[46] Danish Energy Agency and Energinet. Technology data for generation of
electricity and district heating. Tech. rep., Danish Energy Agency; 2020.
[47] NREL. WISDEM. 2021, URL https://www.nrel.gov/wind/systems-engineering-
models-tools.html.
[48] NREL. WISDEM2. 2021, URL https://wisdem.readthedocs.io/en/master/index.
html.
[49] CEN, Electricity market Chile. URL https://www.coordinador.cl/.
[50] Nordpool, Electricity market North Europe. URL https://www.nordpoolgroup.
com/.
[51] AEMO, Electricity market Australia. URL https://www.aemo.com.au/.
[52] Energinet. PTX in Denmark before 2030. Tech. Rep. April, Energinet; 2019, p.
34.
[53] IEA. The Future of Hydrogen for G20. Seizing today’s opportunities. Tech. Rep.
June, IEA; 2019.
[54] Buttler A, Spliethoff H. Current status of water electrolysis for energy storage,
grid balancing and sector coupling via power-to-gas and power-to-liquids: A
review. Renew Sustain Energy Rev 2018;82(September 2017):2440–54. http:
//dx.doi.org/10.1016/j.rser.2017.09.003.
[55] Kettani M, Bandelier P. Techno-economic assessment of solar energy cou-
pling with large-scale desalination plant: The case of Morocco. Desalination
2020;494(June):114627. http://dx.doi.org/10.1016/j.desal.2020.114627.
[56] Al-Turki YA, Mebarek-Oudina F, Ahmadian A, Baleanu D. Flat sheet direct
contact membrane distillation desalination system using temperature-dependent
correlations: thermal efficiency via a multi-parameter sensitivity analysis based
on Monte Carlo method. J Therm Anal Calorim 2021;144(6):2641–52. http:
//dx.doi.org/10.1007/s10973-020-10503-6.
[57] Abusorrah AM, Mebarek-Oudina F, Ahmadian A, Baleanu D. Modeling of a MED-
TVC desalination system by considering the effects of nanoparticles: energetic
and exergetic analysis. J Therm Anal Calorim 2021;144(6):2675–87. http://dx.
doi.org/10.1007/s10973-020-10524-1.
[58] Danish Energy Agency and Energinet. Technology data energy storage. Tech.
rep., Danish Energy Agency; 2020.
[59] Inflation tool, Inflation timeline in the euro-zone (2015–2021). URL https://
www.inflationtool.com/euro?amount=1&year1=2015&year2=2019?.
[60] IRENA. Electricity storage and renewables: Costs and markets to 2030.
Tech. Rep., IRENA; 2017, p. 132, URL http://irena.org/publications/
2017/Oct/Electricity-storage-and-renewables-costs-and-markets%0Ahttps:
//www.irena.org/-/media/Files/IRENA/Agency/Publication/2017/Oct/IRENA_
Electricity_Storage_Costs_2017.pdf.
N. Campion et al. Renewable and Sustainable Energy Reviews 173 (2023) 113057

[61] Teodorović D, Janić M. Transportation, environment, and society. Elsevier Inc.; [69] Hannula I. Co-production of synthetic fuels and district heat from biomass
2017, p. 719–858. http://dx.doi.org/10.1016/b978-0-12-803818-5.00011-1. residues, carbon dioxide and electricity: Performance and cost analysis. Biomass
[62] Odenweller A, Ueckerdt F, Nemet G, Jensterle M, Luderer G. Probabilistic Bioenergy 2015;74:26–46. http://dx.doi.org/10.1016/j.biombioe.2015.01.006.
feasibility space of scaling up green hydrogen supply. Nat Energy 2022;1–27. [70] Bellotti D, Rivarolo M, Magistri L, Massardo AF. Feasibility study of methanol
http://dx.doi.org/10.1038/s41560-022-01097-4.
production plant from hydrogen and captured carbon dioxide. J CO2 Util
[63] NEOEN. Hornsdale power reserve. 2017, URL https://hornsdalepowerreserve.
2017;21(May):132–8. http://dx.doi.org/10.1016/j.jcou.2017.07.001.
com.au/.
[64] Jang D, Kim J, Kim D, Han WB, Kang S. Techno-economic analysis and Monte [71] Danish Energy Agency and Energinet. Technology data for renewable fuels. Tech.
Carlo simulation of green hydrogen production technology through various rep., Danish Energy Agency; 2020.
water electrolysis technologies. Energy Convers Manage 2022;258:115499. http: [72] Dansk Fjernvarme. Power-To-X and district heating. Tech. rep., Dansk
//dx.doi.org/10.1016/J.ENCONMAN.2022.115499. Fjernvarme; 2021.
[65] IEA. Tracking power 2021. 2021, URL https://www.iea.org/reports/tracking- [73] DIN Forsyning Varme A/S. Din forsyning. 2020, URL https://dinforsyning.dk/da-
power-2021. dk/varme/priser-for-varme-2020.
[66] Franz S, Campion N, Shapiro-Bengtsen S, Bramstoft R, Keles D, Münster M. [74] USEnergy Information Administration. Today in energy. 2022, URL https://www.
Requirements for a maritime transition in line with the paris agreement. iScience eia.gov/todayinenergy/detail.php?id=52358#.
2022. http://dx.doi.org/10.1016/J.ISCI.2022.105630.
[67] Cheema II, Krewer U. Operating envelope of Haber-Bosch process design for
power-to-ammonia. RSC Adv 2018;8(61):34926–36. http://dx.doi.org/10.1039/
c8ra06821f.
[68] Alfa Laval, Hafnia, Haldor Topsoe, Vestas, Siemens Gamesa. Ammonfuel-an
industrial view of ammonia as a marine fuel. Tech. Rep. August, Ammonfuel;
2020, p. 1–59, URL https://hafniabw.com/news/ammonfuel-an-industrial-view-
of-ammonia-as-a-marine-fuel/.

22