Flexible production of green hydrogen and ammonia from variable solar and wind energy. Case study of Chile and Argentina

Abstract

We report a techno-economic modelling for the flexible production of hydrogen and ammonia from water, air, and optimally combined solar and wind energy. We use hourly data in four locations with world-class solar in the Atacama desert and wind in Patagonia steppes. We find that hybridization of wind and solar can reduce hydrogen production costs by a few percents, when the effect of increasing the load factor on the electrolyser overweighs the electricity cost increase. For ammonia production, the gains by hybridization can be substantially larger, because it reduces the power variability, which is costly, due to the need for intermediate storage of hydrogen between the flexible electrolyser and the less flexible ammonia synthesis unit. Our modelling reveals the crucial role of flexibility in the synthesis, because it cuts the cost of variability, especially for the more variable wind power. Our estimated near-term production costs for green hydrogen, around 2 USD/kg, and green ammonia, below 500 USD/t, are encouragingly close to competitiveness against fossil-fuel alternatives.

Full text

Flexible production of green hydrogen and ammonia from variable solar
and wind energy. Case study of Chile and Argentina
Julien Armijo1,* and Cédric Philibert2,†
1: Centro de Investigación en Recursos Naturales y Sustentabilidad, Universidad Bernardo O’Higgins,
Santiago, Chile, 2: International Energy Agency, OECD, Paris, France
* julienarmijo@gmail.com, † cedric.philibert@iea.org
Abstract: We report a techno-economic modelling for the flexible production of hydrogen and
ammonia from water, air, and optimally combined solar and wind energy. We use hourly data
in four locations with world-class solar in the Atacama desert and wind in Patagonia steppes.
We find that hybridization of wind and solar can reduce hydrogen production costs by a few
percents, when the effect of increasing the load factor on the electrolyser overweighs the
electricity cost increase. For ammonia production, the gains by hybridization can be
substantially larger, because it reduces the power variability, which is costly, due to the need
for intermediate storage of hydrogen between the flexible electrolyser and the less flexible
ammonia synthesis unit. Our modelling reveals the crucial role of flexibility in the synthesis,
because it cuts the cost of variability, especially for the more variable wind power. Our
estimated near-term production costs for green hydrogen, around 2 USD/kg, and green
ammonia, below 500 USD/t, are encouragingly close to competitiveness against fossil-fuel
alternatives.
Keywords):)energy)transition,)renewable)energy,)hydrogen,)ammonia,)flexibility)
1) Introduction
Since recent studies have shown that climate change is becoming an "existential threat" to
humanity, with possible warming above 2oC in the next three decades and beyond 4oC to 6oC
within the next several decades [1], it has become clear that aggressive reductions of
greenhouse gas emissions to near zero level must be implemented as soon as possible to avoid
global warming turning from dangerous to catastrophic [2–4]
To address this challenge, renewable electricity and biomass are not sufficient. Green
hydrogen production via water electrolysis is the paradigm of industrial conversion of
renewable electricity into chemical energy, storable and transportable. As such, water
electrolysis and the conversion of green electricity into heavier green molecules and fuels
beyond hydrogen, is attracting rapidly increasing attention, because of the promise of
unleashing the potential of low-emission renewable energies (RE) way beyond electricity, and
pave the way for realistic planning of a global economy with net zero emissions [5–9]
However, a key remaining question concerns the temporal variability of wind and solar, the
main resources expected to play the largest role in the increase of RE in the global energy
system (see, e.g., [10]). How does their variability affect the production costs of H2, NH3 or
carbon-based synthetic fuels? Some studies, like [11] for methane, or [12] for methanol, have
found that if the steps after the electrolysis are not flexible, the required intermediate buffer
storage of H2 could be a prominent cost driver. Yet, most previous studies modelling green
hydrogen or green fuels production, have avoided to treat the variability issue, assuming either

!
dispatchable RE power from the grid (e.g., [13] for ammonia), perfectly flexible synthesis (e.g.,
[14] for ammonia), or underground buffer storage of hydrogen, which is so cheap that it
almost erases the cost of variability (e.g., [15] for Fischer-Tropsch fuels). For ammonia, flexible
production was modelled in [16] wind power from grid with 80% down flexibility, and in [17]
from an islanded wind farm and flexibility varied from 0% to 80%. A complete understanding
of the role of flexibility is thus still in progress, especially when considering also the open
question of the possible interest to combine wind and solar power, as suggested, e.g., in [15]
to reduce green fuels production costs.
In this paper, we present a techno-economic model that plainly addresses these questions,
using four examples locations in Chile and Argentina, two countries having world-class variable
renewable energy (VRE) resources, and thus tremendous potentials for becoming leading RE
producers, and exporters of RE stored in H-rich chemicals [8,18]. Chile has the world’s
strongest solar resource and good wind in the north, and excellent wind in the far south.
Argentina has among the world’s best onshore wind potentials in its vast desert Patagonian
steppes, and very good solar in the North.
For each location, our modelling estimates the short-term costs of mass production of H2 and
NH3 based on full meteorological yearly data with hourly resolution, determining the optimal
sizes of the solar, wind and NH3-synthesis units, relative to the electrolyser capacity. We focus
on NH3 because, apart from H2 itself (and the more complex hydrazine), it is the only potential
H-based fuel that does not contain carbon. It is easier to liquefy, store and transport than H2,
and has a greater energy density. Furthermore, it is a large global market already,
internationally traded with pipelines and oceangoing tankers [19,20]. However, its production
from natural gas currently entails production of 2 t CO2/t NH3.
Our goal is essentially to treat the situation of abundant and cheap renewable resources, in
locations far from consumption centers [5], where grids are generally weak, therefore, our
modelling is based on dedicated wind and solar resources, so that we face the questions of
variability completely. Our modelling allows to understand the benefits of hybridization of
wind and solar power, which is much stronger when the less flexible process of ammonia
synthesis follows the production of hydrogen. We also analyse the interplay between the
variability of the RE sources, the flexibility of the ammonia plant, and the optimal sizing of the
wind, solar, and ammonia synthesis units.
2) The potential for renewables in Chile and Argentina)
2.1) Chile
In Chile, the potential for renewable electricity production via solar PV, CSP, wind, and hydro,
has been estimated in a comprehensive overview study [21], carried by the Ministry of Energy
and GIZ. The study covered the country from Arica to Chiloé, that is, from the northern end of
the country, to the southern end of the National Electric System (SEN). Southern Patagonian
parts of Chile have two islanded electric systems.
)
Figure 1 shows the capacity factors estimated in [21] for solar PV with 1-axis tracking, and
wind, in the northern part of the country. In each case, the map on the right side takes into
account land restrictions (national parks, vicinity <1km to towns, excessive elevation >2000m,
etc.). The results shown are based on meteorological Weather Research and Forecast (WRF)
simulations, and technical performance of solar panels and wind turbines are computed with
realistic models that include temperature dependence of solar panel efficiency, air density

!
dependence of wind turbine efficiency, etc. In [21], the simulation results were compared in
detail to field meteorological and plant production data in several representative cases, giving
good reliability to the model.
In Fig. 1, estimated solar potentials reach capacity factors CF up to 40%, taking into account an
assumption of 15% of annual technical losses. In [21], the solar potential in the North of Chile,
was estimated to 1 260 GW for PV and 550 GW for CSP, i.e., a total above 1 800 GW available
in North of Chile, considering only capacity factors CF > 30%, and with a conservative land use
assumption of 20 MW/km2 (projects often reach 50 MW/km2).
Fig 1: Potential capacity factors for solar and wind energy in the North of Chile. Maps on the right side
show the resource combined with land restrictions. Source: [GIZ 2014]
In [21], wind potentials were estimated from the model, to which a correcting factor of
0.75=0.85 x 0.88 was applied to take into account technical losses (assumed at 15%) and
where the factor 0.88 accounts for the noted overestimation of the modelled wind speeds,
with regards to measurements (using a station in the Taltal area). Following this procedure,
the map in Fig. 1 shows potential capacity factors above 40% in several areas of the North,
especially near Taltal, where wind average speeds of about 10 m/s at 80 m height were
recorded in 2010 ([21] p. 21), and in several valleys such as the Calama valley. Note, however,
that the specific power (i.e. power over swept area) of wind turbines considered was not
specified in this work.)
The available wind potential in the 5 northernmost regions shown in Fig. 1, considering only
capacity factors >30%, and a land use of 10 MW/km2, was estimated to about 14.5 GW,
including 11.5 GW in the interior of Taltal. Very windy places on the Altiplano plateau near the
border with Argentina were discarded due to excessive elevation and remoteness. Wind
potentials with CF>40% are also available in several locations in the central North and central
South, in particular, about 23 GW were identified in the southern regions of Bio-Bio,
Araucanía, Los Rios, and Los Lagos (including Chiloé island), essentially along the Pacific coast
or the coastal mountain ranges (see Fig. 17 in [21]) However the best wind resource with
excellent potentials and capacity factors CF>60%, are located in the Patagonian far South
(Magallanes region).)
From Fig. 1 it is clear that the North of Chile offers highly interesting possibilities for solar,
wind, or combinations of them. In this study, we pick two locations (see Fig. 1) with excellent
solar resource and also very good wind ("Taltal"), or good wind ("Calama"). It also is very
convenient that the Ministry of Energy disclosed two online explorer tools [22,23] that allow to
simulate the production of solar and wind farms and download hourly data, with various
“Tal Tal”
25.07ºS, 69.84ºW
“Calama Valley”
22.05ºS, 69.08ºW
Solar PV (1 axis tracking) Wind

!
technological assumptions. Note that the wind explorer covers the whole country, including
the far southern Patagonian regions and even parts of Argentina.
2.2) Argentina
For Argentina, [24] provided overviews of the solar and wind potentials, reproduced in Fig. 2.
Schematically, most of the southern half of the country has enormous wind power with
average wind speeds above 10 m/s in extensive areas, and potential capacity factors often
exceeding 60%, even for class 1 turbines. The north-western part of the country, next to the
Andes mountains, is arid and boasts high solar irradiation above 5.5 kWh/m2 day, very close to
the levels reached in Chile. The potentials for wind and solar in Argentina can also be visualized
in [25] and [26].
Fig. 2: Wind and solar resources in Argentina, with land-use restrictions. Wind speed is estimated at
50 m. Source: [24]. Red dots show the "Patagonia Argentina" and "Patagonia Chile" study cases.
Taltal
Calama
Pat. Chile
Pat. Arg.
Latitude (oS)
25.07)
22.05)
52.48)
44.34)
Longitude (oW)
69.84)
69.08)
70.94)
70.77)
Altitude (m)
2)200)
2)021)
331)
967)
Air density (kg/m3)
0.95)
0.97)
1.21)
1.13)
Table 1: The four locations studied. Data source: [23]
To explore the possibility to use both the extreme wind and good solar potentials, we select a
"Patagonia Argentina" location marked as red spot on Fig. 2, in the western Chubut province.
The red spot in the Chilean Magallanes region denotes the "Patagonia Chile" study case.
3) Energy situation and strategies for renewables and hydrogen
44.34ºS, 70.77ºW
“Patagonia Chile”
52.48ºS, 70.94ºW
“Patagonia Argentina”

!
3.1) Current energy situations in Chile and Argentina
In Argentina, the primary energy matrix is strongly dominated by a 90% share of fossil fuels
and 52% of gas in 2016 [27]. In Chile, fossils are also well dominant, providing 70% of the
primary energy in 2016, but biomass provided 18.5% in Chile, versus 4% in Argentina, and
hydro provided 5% in 2016 in Chile, and 3.5% in Argentina. Geothermal and non-conventional
renewables are rising rapidly [28], although in 2016 they provided only 1.1% of the primary
energy in Chile and 0.06% in Argentina.
As reported in [29], the electrical installed capacity in Chile is 23.3 GW, with already 2.3 GW of
solar, 1.5 GW of wind, 6.8 GW of hydro, 500 MW of biomass. The total annual production is
about 80 TWh. Due to rapid development of variable renewable energies (VRE) in recent years
(and transmission issues), electricity prices have fallen strongly in some nodes in the northern
parts of the former Central Interconnected System (SIC). For example, as reported in [30], in
December 2017 the Diego de Almagro spot price averaged approximately 7 USD between
11am and 6pm.
In Argentina, the installed electric capacity was 34 GW in 2017, and the total generation is
about 150 TWh [28]. In Argentina, renewables have just started to boom in 2018. The
Argentinian wind market was in 2018 the fastest growing globally, with an installed capacity
growing 181%, from 228 MW to 640 MW during 2018 [31]. 2018 was also the year of the first
massive additions of PV capacity, in the 500 MW range. Analysis and forecasts for renewables
in both countries are also available in [32].
3.2) Electricity prices in the latest auctions
Following global trends, renewable electricity prices have been going down dramatically in
recent years in both countries [33]. Table 2 shows the latest auctions results, in November
2017, for electricity to be supplied by 2023.
Chile
Argentina
Avg. price solar
32.5)
43.5)
Avg. price wind
42.7)
41.2)
Min. price solar
21.48)
40.4)
Min. price wind
n.a.)
37.3)
Table 2: Latest electricity prices (USD/MWh) in November 2017 auctions. Sources: [34][35]
3.3) The value and market of NH3
Ammonia is the second most produced synthetic chemical today, after sulphuric acid. In 2017,
its production was of 166 Mt [36] of which about 80% were used to produce nitrogen
fertilizers. Ammonia represents the second largest demand for pure H2 after oil refining (about
32 Mt, i.e. 44% in 2018) [6]. As for today´s uses, ammonia production (essentially from gas and
coal), causes 420 Mt CO2 emissions, that is, more than 1% of global emissions. Therefore,
greening today´s ammonia already represents a large opportunity for near-term
decarbonisation.

!
But ammonia is also already the main carrier of hydrogen energy, which avoids the main
difficulties and costs of storing and transporting gaseous H2. Ammonia could thus be a major
enabler for renewable energy storage, transport and use [9,37,38]. It may be the most
interesting pathway for long-duration storage of chemical hydrogen energy. In particular,
ammonia can be combusted with air to provide power, which may represent one solution of
choice in many local or regional scenarios towards 100% renewable electricity, to deliver
power at times when the wind and solar are not producing enough [39]. Currently, active
research is being conducted to find optimal techniques to combust ammonia in gas turbines,
or co-combust it with coal or gas in existing steam plants, the two main challenges being to
overcome the low flammability, and to ensure the NOx emissions remain below acceptable
standards [40][41]. Also, green ammonia is increasingly being perceived as probably the most
cost-efficient option for decarbonizing maritime transport [6,42] and engine manufacturers are
preparing for a shift to green ammonia [43].
Since ammonia is a directly usable chemical with an existing market, in agriculture and other
end uses, including the explosives industry for mining in Chile, technology for transporting and
storing it are already demonstrated, widely deployed and functional worldwide, including, e.g.,
about 5 000 km of pipelines in the U.S. Midwest, and a 2 500 km long pipeline from Russia to
Ukraine. In general, the costs for transporting or storing ammonia are well lower than those
for H2, and the efficiencies higher [20].
Issues concerning security are also well-known for decades, and the safety track record is
excellent. Despite a toxicity that can be lethal to humans at high concentrations, deadly
accidents have been very rare for decades of massive use (< 3 dead / 1 billion pers.). Therefore
it is fair to say that "NH3 is a chemical that should be respected, but not feared" [44].
3.4) National strategies for the energy transition and expansion plans for H2 and H-rich
molecules
In Chile, the Long-Term Energy Planning [45] foresees a doubling of the energy consumption
between 2016 and 2050 (p71). A carbon tax at 5 USD/t CO2 is already applied in the electricity
generation sector, and, in some prospective scenarios, it is considered to rise to 30 USD/t CO2
in 2030 (p. 63). Scenarios for the energy matrix in [45] do not consider a decrease of the
installed capacities of fossil fuel based electricity generation, however, progressive phasing-out
carbon based generation is underway.
In Chile, hydrogen production is being promoted by the Chilean Economic Development
Agency (CORFO). Two international consortia have been created at the initiative of CORFO,
both for H2 use for mobility in mining. The production of green NH3 is also being pursued in the
mining context by the company Enaex, which today imports 360 000t/yr of NH3 to produce
explosives, and aims at cutting its dependency on price volatility and reduce costs by
producing solar-based green ammonia.
In Argentina, only one private initiative for H2 is today visible: the company Hychico, which has
developed wind farm-powered electrolysers, and is currently developing hydrogen cavern
storage where methanation occurs due to bacteria [46].
4) Renewable resources in Taltal and Patagonia Argentina locations
4.1) Data sources and technological assumptions

!
The estimation of power production from renewables depends on the atmospheric physical
data, and technological assumptions.
Wind. Concerning wind turbines, as discussed for example in [47–49], the current trend
globally in the industry is to increase the swept areas (i.e. decrease the specific power in
W/m2) for similar wind resources. This has allowed to continuously crank up the capacity
factors of newly installed plants, and increase the value of the produced energy [47] because it
allows to reduce fluctuations in the production, spreading it over more hours.
For example, the average specific power among U.S. turbines, has declined from 395 W/m2 for
projects installed in 1998-1999 to 230 W/m2 for projects installed in 2018 [49] and the majority
of installed turbines currently are of class 3. This trend has been accompanied by a continuous
increase in the average capacity factor for newly installed turbines, from about 25-30% to 42%
today, even though the quality of the wind resources used has declined by about 15% in the
same period (see Fig. 39 in [49])
Vestas 90-2 (class 2)
Nordex N 100-3.3 (class 1)
Rotor diameter (m)
90)
100)
Turbine power (MW)
2)
3.3)
Hub height (m)
93)
93)
Specific power (W/m2)
314)
420)
CAPEX (USD/kW)
1)300)
1)200)
Table 3: Wind turbines assumptions for Northern Chile (class 2) and southern Patagonia (class 1)
In our modelling, for all sites we use hourly data from [23] assuming hub height 93 m, and we
consider two types of turbine. For northern Chile, we use Vestas V 90-2, which are class 2
turbines of specific power 314 W/m2, similar to the Vestas 112-3 (with 3 075 kW generator) of
specific power 312 W/m2, which are the model installed in the 99 MW Taltal Enel Green Power
plant [50]. However, as mentioned above, the data in [23] have been noticed to overestimate
the resource in the North. Thus, we expect that, in the near future, the best options to
optimally exploit the wind potential in northern Chile, should probably rather be class 3
turbines. For our two modelled southern Patagonia locations, due to the extremely strong
wind, we consider Nordex 100-3.3 turbines, which are class 1 and have specific power 420
W/m2, as summarized in Table 3.
Concerning technical losses, [51] considering wake effects due to multiple turbine siting,
availability (maintenance) and aging, assumes a factor of reduction of 18.5% (p. 246). In [52],
the losses (including wake effects, electric and transmission losses) are estimated to 15%, and
the availability of the plant to 98%. In our modelling, we assume technical losses of 15% for the
southern locations, and 20% for the North, to account also for the overestimation of wind
speeds in [23].
Taltal
Calama
Pat. Chile
Pat. Arg.
Deviation from 1980-2013 (%)
-2.5)
+1)
+19)
-2.5)
Table 4: Deviations of wind capacity factors in 2013 from the 1980-2013 average. Source : [23].
For all locations, both for wind and solar data, we use data for 2013, which is the latest year
available in [23]. Because inter-annual variability is important for wind, in Table 4 we show the

!
deviation of the capacity factors in 2013, used in this work, compared to the climatic average
over 1980-2013. The deviation is significant only in the Patagonia Chile case, but the model
displays there a sustained increase of the capacity factor over the period, and the 2013 data
are good enough to represent present conditions.
Solar. For solar power production, we use data from [53], assuming 1-axis azimuthal tracking
with and zero tilt. In [21], a loss factor of 15% was used to take into account all technical
losses, although in [54], larger losses of 20-40% are discussed. For all our locations, we assume
an overall technical loss factor of 15%.)
Note that both for solar and wind, even though technical losses in real plants may be higher in
average than our standard assumption of 15%, their levelized impact over the plant lifetime is
lower, because their performance is systematically better in their first half-life of operation
[49,54], and that the production in earlier years of life has higher economic value than in later
years.)
4.2) Analysis of renewable energy resources
Fig. 3: Wind and solar resources in (A, B) : Taltal, Chile and (C, D) : Patagonia Argentina. In each panel : a)
Average daily cycle. b) Yearly cycle with day (blue) and weekly average (red). c) Load statistic, bins
width: 5%. d) Year-day cycle. Data sources: [23,53]
Taltal. Figures 3.A and B show the estimated wind and solar resources for 2013 in our Taltal
location. Interestingly, Fig 3.A.a shows that the average day cycle has good complementarity
with the solar cycle, as wind blows strongly from 10 pm to 8 am. In Fig. 3.A.b and d, one sees
that the wind is also more intense in winter (June-September), opposite to the solar radiation.
0 10 20
Hour
0
0.5
1
Capacity factor
Wind Taltal 2013. a) Day cycle
100 200 300
Day
0
10
20
Hour
d) Year-day cycle
0
0.2
0.4
0.6
0 100 200 300
Day
0
0.2
0.4
0.6
0.8
Capacity factor
b) Day / week avg; CF=43.8%
0 0.2 0.4 0.6 0.8
Capacity factor
0
1
2
3
nb of hours (x1000)
c) Statistics of load
0 10 20
Hour
0
0.5
1
Capacity factor
Wind Pat. Arg. 2013. a) Day cycle
100 200 300
Day
0
10
20
Hour
d) Year-day cycle
0
0.2
0.4
0.6
0.8
0 100 200 300
Day
0
0.2
0.4
0.6
0.8
Capacity factor
b) Day / week avg; CF=52.7%
0 0.2 0.4 0.6 0.8
Capacity factor
0
1
2
3
nb of hours (x1000)
c) Statistics of load
0 10 20
Hour
0
0.5
1
Capacity factor
Solar Taltal 2013. a) Day cycle
100 200 300
Day
0
10
20
Hour
d) Year-day cycle
0
0.2
0.4
0.6
0.8
0 100 200 300
Day
0
0.2
0.4
Capacity factor
b) Day / week avg; CF=32.5%
0 0.2 0.4 0.6 0.8
Capacity factor
0
1
2
nb of hours (x1000)
c) Statistics of load
0 10 20
Hour
0
0.5
1
Capacity factor
Solar Pat. Arg. 2013. a) Day cycle
100 200 300
Day
0
10
20
Hour
d) Year-day cycle
0
0.2
0.4
0.6
0.8
0 100 200 300
Day
0
0.2
0.4
Capacity factor
b) Day / week avg; CF=20.7%
0 0.2 0.4 0.6 0.8
Capacity factor
0
1
2
nb of hours (x1000)
c) Statistics of load
A) B)
C) D)

!
Note that the occurrence of night winds is also interesting in other areas in northern Chile,
such as the area of Calama, where the wind variability is also not only seasonal and synoptic
(days to weeks), but has a marked thermal (daily) cycle. In Fig. 3.A.c, one sees that the load
statistics has accumulations near 0 and the maximum at 80% (due to our assumption of 20%
losses), and is continuously distributed in between. The large number of hours (>3000) with
80% load, is partly an artefact, due to the above-mentioned overestimation of wind speeds by
the model [21].
Fig. 3.B shows the Taltal solar resource, with data from [53], assuming 15% technical losses,
which explains the maximal values at 0.85 on Fig. 3.B.c. Solar electricity production in Taltal is
very stable year round, as seen in Fig. 3.B.b and d: only 8 days in 2013 had capacity factors
under 20%. The average day cycle (Fig. 3.B.a) has very steep slopes at the beginning and end of
the day, typical of 1-axis tracking PV plants, which results, in Fig. 3.B.c, in a concentration of
load statistics at 0% (night) and near the maximal value of 85%.
Patagonia Argentina. Figure 3.C and D show the wind and solar resource in our Patagonia
Argentina (Chubut) location. The wind is extremely strong, with an average capacity factor of
52.7% (including the 15% losses). The average cycle is rather flat both daily and yearly (Fig.
3.C.a, b, d). However, fluctuations can be very strong on the daily and weekly scale, as seen on
Fig. 3.C. Note that, in the 6.3MW Diadema Wind Park in eastern Chubut Province (one of the
oldest in Argentina), an average net CF of 47.7% over 2012-2017 was reported [46].
Considering that those turbines are Enercon 44-900 at only 45m height [55], with a high
specific power of 592 W/m2, brings credibility to the performance of 52.7% in our model.
The solar resource, shown on Fig. 3.D, is weaker, and has a marked yearly cycle with a
minimum in winter months (Fig. 3D.b and d).
4) Optimized production of H2 from hybrid wind and solar energy
4.1) Description of the model
To estimate the production costs of H2 in Chile and Argentina, we use the economic
assumptions in Table 5, including Capital expenditures (CAPEX), operational expenditures
(OPEX), and weighted average capital cost (WACC). All our assumptions are conceived for the
short term (2020).
Electrolyser. The electrolyser is assumed to be alkaline, with perfectly flexible behaviour, i.e., it
can follow in real time the variable electricity supply and maintain a constant efficiency
h
=
70%. Such assumptions are consistent with latest technologies reported, e.g. in [56] where the
load range is 10-110%, and for which the maximal ramping rates (up or down) of +/- 20% / s is
largely sufficient to follow wind or solar variations from utility-scale farms. According to
[57,58], 70% efficiency seems optimistic but it is in fact realistic given recent achievements and
current progress. Furthermore, efficiencies at partial load are usually higher than at full load
(e.g. 67%), which increases the effective average efficiency for renewable power with a
statistically important number of hours with partial load [16].
The CAPEX assumed is higher than the 450 USD/kW given in [59] for 2023 to account for
installation, but lower than values given in [57,58]. We can assume that cost reductions by
scaling should go fast in the next years, because large scale electrolysers are still in
development. In the near future, industrials and the review [60] claim that the large scale

!
(>10MW) PEM electrolysers can have the same cost reductions as alkaline electrolysers did, so
similar prices can be expected for both technologies.
Value
References
CAPEX electrolyser (USD/kW)
600)
[56–62]))
OPEX electrolyser (% CAPEX/yr)
2)
ibid.)
lifetime N (yr)
30)
ibid.)
stack lifetime Ns (h)
80,000)
ibid.)
stack repl. cost (% CAPEX)
40)
Ibid.)
Electrolyser efficiency
h
(LHV)
70%)
ibid.)
)
)
CAPEX solar (USD/kW)
740)
[33,45,63,64])
CAPEX wind (USD/kW)
1200)-)1300)
ibid.,)[49])
OPEX solar (% CAPEX/yr)
1.7)
[33])
OPEX wind (% CAPEX/yr)
2)
[33][49])
tech. losses solar
15%)
[21,54,65])
tech. losses wind (North /
South)
20%)/)15%)
[21,66])
lifetime vRE plants (yr)
25)
-)
)
)
WACC (Chile / Arg.)
7%))/))10%)
-)
Table 5: Techno-economic parameters for the electrolysers and solar and wind power plants
VRE plants. For the electricity, the CAPEX values that we take are roughly in line with [45,63],
however we adjusted them so that the obtained LCOE are similar to most recent auction prices
(see Table 2). In particular, the CAPEX for wind plants in [45] is 1800 USD/kW, but such high
values are probably out of date and could not reproduce the lowest price bids for wind energy,
both in Chile and Argentina. Our assumptions for CAPEX of wind and solar, are aligned with the
ones for China in [64] for 2020, to which we apply a heuristic 20% cost increment. Note that
our estimated CAPEX in this fashion, are slightly higher than those of 700 USD/kW for solar,
and 1100 USD/kW for wind, assumed for 2020 in Argentina in [67].
In 2016, in the USA, the average OPEX for wind power were of 52 USD/kW for CAPEX of
1 590 USD/kW [49]. According to [33], OPEX for wind can be as low as 20 USD/kW for full-
service initial contracts, and rather on the order of 30 USD/kW for full service renewal
contracts. We assume 25 USD/kW/yr. According to [IRENA 17], OPEX for solar are between 10-
18 USD/kW. We assume 13 USD/kW/yr.)
Financial cost. A higher WACC of 10% is assumed for Argentina, to match the higher prices,
probably related to the country risk factor, which is known to be higher than in Chile.
Cost calculations. Our model considers a fixed electrolyser nominal size PH2, powered by
dedicated solar and/or wind farms of capacities Psolar and Pwind. For a range values of the ratios
as = Psolar/PH2 and aw = Pwind/PH2, we compute the levelized cost of hydrogen LCOH. For each
couple (as, aw), the full hourly data sets of solar and wind electricity production (presented in
Section 3) are combined for the year 2013. The total production of H2 is computed, as well as
the total curtailment of electricity, and the hybrid capacity factor, which is the load factor on
the electrolyser. Finaly the LCOH is computed using the costs of electrolyser, electricity, water
(including desalination), and oxygen sales as:

!
LCOH(as , aw)= cH2 + celec + (cH2O - cO2), (1)
where for each plant element k=H2, wind or solar (H2 standing for electrolyser), we use the
standard annualised formula)
ck = CAPEX k x (CRF + OPEXk) x EH2 / (
h
x nhy x CFk), (2)
where, CRF=WACC x (1+WACC)N / ((1+ WACC)N -1) is the capital recovery factor, nhy=8760 is the
number of hours per year, CFk is the capacity factor of the considered plant element k, and
EH2=33.3 kWh/kg is the low heating value (LHV) energy density of H2.
For the electrolyser, the stack lifetime is assumed to be Ns = 80 000 h, in line with published
performances [57]. Stack replacement is done at time Nrep=Ns/(nhy x CFh), where CFh is the
hybrid load capacity factor that effectively is applied to the electrolyser in each location. We
thus add to the electrolyser CAPEXH2 the net present value of stack replacement (CAPEXH2 x
0.4) x (1-WACC)Nrep, using for simplicity the WACC as discount factor.
Concerning water desalination, probably required in northern Chile for H2 production, with the
assumptions of [68] for reverse osmosis, (p. 171) of a CAPEX of 7400 USD/m3/day, fixed OPEX
of 3% of CAPEX, electric consumption of 3 kWh/m3, lifetime of 30 years, and assuming full load
electricity at 60 USD/MWh, the cost of water production is LCOW=5.9 USD/m3, that is, a cost
cH2O= 0.053 USD /kg H2, which is small. For simplicity, the cost of water, as well as the value of
the produced oxygen, which in Chile is cO2= 0.03 USD/kg O2 [69], i.e. 0.24 USD /kg H2, are both
neglected in our calculations for all locations. In [70], the cost of desalination of water is also
seen to be negligible.)
Optimization procedure. A minimum search is then applied to the matrix LCOH(as, aw),
determining the parameters for the optimal hybrid plant in each location (see Fig. 4).
4.2) Example of LCOH optimization in Taltal and Patagonia Argentina
Fig. 4: Optimization of a hybrid H2 plant in Taltal. The red dots show the identified optimum.
0.8 1 1.2 1.4
as
0
0.2
0.4
0.6
0.8
aw
H2 plant Taltal : LCOH*=2.12 USD/kg
2.2
2.3
2.4
2.5
USD/kg
0.8 1 1.2 1.4
as
2.1
2.2
2.3
2.4
USD/kg
0 0.2 0.4 0.6 0.8
aw
2.1
2.15
2.2
USD/kg
CAPEX
H2=600USD/kW;
N=30yrs; WACC=7%,
OPEX=2%/yr; LHV=70%
LCOEs=26.7USD/MWh;
LCOEw=35.8USD/MWh
Opt.: as
*=1.18; aw
*=0.33;
CF*=51.8%; curt *=1.8%
LCOH reduction =1.6%
compared to only solar

!
Taltal. Figure 4 shows the LCOH in Taltal, for various combinations of wind and solar plant
sizing relative to the electrolyser. The computed levelized costs of electricity LCOEs=26.7
USD/MWh and LCOEw= 35.8 USD/MWh are in good agreement with the latest auction prices in
Chile (see Table 2). The lowest LCOH*=2.12 USD/kg, is obtained with the optimal combination
as*=1.18 solar and aw*=0.33 wind.
In this case study where solar electricity is the cheapest, the modest hybridization of solar with
some more expensive wind electricity allows to increase the capacity factor of the electrolyser
from 41.3% (the best case with only solar with as=1.29), to 51.8%. A fraction curt*=1.8% of the
renewable electricity mix is curtailed, and the LCOH is reduced thanks to this hybridization, by
1.6% only.)
Patagonia Argentina. In the Patagonia Argentina case, our model yields electricity prices
LCOEs=51.8USD/MWh and LCOEw=33.8USD/MWh, also in good agreement with latest auction
prices (see Table 2), projecting upcoming learning and deployment of the technologies. In this
location where wind electricity is notably cheaper than the solar, the lowest LCOH is obtained
without hybridization, that is, with as*=0, and with a slight oversizing of the wind farm capacity,
so that the optimal load factor of the electrolyser is 62.2%, with 0.2% of the electricity
curtailed (see Table 6).
4.3) Summary of hydrogen production costs
Fig. 5: Optimal LCOH for hybrid H2 plants in our four locations. Each element includes CAPEX and OPEX
for the considered unit.
Figure 5 shows a comparison of the obtained optimal LCOH in our four selected locations, and
the optimal parameters found are summarized in Table 6. In sum, we note that hybridization
can be slightly profitable for H2 production, only when the electricity costs are close to each
other. But for the cases we run, the gain from hybridization (cost reduction) is only 1.6%, in
Taltal, compared to the best case with only solar. This is because we have treated cases where
one resource is strongly dominant on the other.
Hybridization of course is more favoured when daily and yearly cycles combine favourably,
because they can increase the hybrid capacity factor CF* of the electrolyser. In general, this is
not a strong effect, and the temporal correlation between wind and solar productions is small
[15]. However, it can be an interesting factor to consider, where the wind has a marked daily
cycle, which is often the case in northern Chile, where cycles can be favourable, as in Taltal
(see Fig. 3). Running different test cases, we also noted that the interest for hybridization is
enhanced when CAPEXH2 is larger. )
Taltal Calama Pat. Chile Pat. Arg
0
0.5
1
1.5
2
2.5
USD / kg
LCOH
Electrol.
Solar elec.
Wind elec.

!
Taltal
Calama
Pat. Chile
Pat. Arg.
Capacity factor solar (%)
32.5
32.4
17.4
20.7
Capacity factor wind (%)
43.8
35.6
51.8
52.7
LCOE solar (USD/MWh)
26.7
26.8
49.9
51.8
LCOE wind (USD/MWh)
35.8
44.1
28
33.8
Capacity solar as*
1.18
1.29
0
0
Capacity wind aw*
0.33
0
1.18
1.18
Hybrid load CF* (%)
51.8
41.1
61.2
62.2
curt* (%)
1.8
1.7
0.1
0.2
hybridisation cost reduction (%)
1.6
0
0
0
LCOH* (USD/kg)
2.12
2.16
1.94
2.33
Table 6: Summary of optimal parameters for hybrid H2 plants
Table 6 allows to visualize the driving ingredients of the LCOH in our four studied locations.
First, the lowest LCOH are obtained where electricity is cheapest. In Taltal, with solar at 26.7
USD/MWh, the LCOH is 2.12 USD/kg, however in Patagonia Chile, with slightly more expensive
wind at 28 USD/MWh, the LCOH is lower at 1.94 USD/kg. This is due to the effective capacity
factor of CF=61.2% in Patagonia Chile, quite larger than the CF=51.8% in Taltal. Another remark
is that the larger assumed WACC in Argentina, prevents from reaching the same low LCOH as
in Chile, although the quality of the wind in the far South is as good as in Chile.
5) Optimal flexible production of NH3 from hybrid wind and solar energy
5.1) Description of the model
Modelling a green NH3 plant, as shown in Fig. 6, is substantially more complex than modelling
a green H2 plant. After H2 is produced, it has to go through the Haber-Bosch (HB) process,
where it is combined with N2, previously obtained from an air separation unit (ASU), to give
NH3. This combination, usually run on less variable hydropower sources, has actually been the
workhorse of the fertiliser industry from 1926 to the 1980s [19].
The Haber-Bosch synthesis loop, in general, is much less flexible than an electrolyser:
f
HB, the
intake flux of H2 in the reactor, cannot be vary as rapidly as does
f
H2, the flux of H2 production
by the electrolyser. One thus has to include an intermediate storage of H2, large enough to
buffer the fluctuations of H2 production. For this purpose, we consider pressurised gaseous H2
storage in steel tanks, which is quite costly, but well mature and available off the shelf. If
available, one could also consider underground storage of H2 in caverns, which is much
cheaper. Table 7 summarizes our techno-economic assumptions.

!
Fig. 6: Schematic of a green all-electric Haber-Bosch plant. Light blue lines denoted by
p
show power
flows, other lines represent chemical energy fluxes when denoted by
f
, or mass flows.
Values
References
CAPEXHB (USD/kW) (LHV H2 in)
580)
[39])
CAPEXASU (USD/kW) (LHV H2 in)
224)
ibid.)
OPEXHB+ASU (% CAPEX/y)
2)
ibid.)
lifetime (y)
30)
ibid.)
Elec. HB-ASU (MWh/t NH3)
0.64)
ibid.)
Elec. pre-compression (MWh/t NH3)
0.26)
ibid.))
)
)
flexibility HB total down (% max load)
40)/)80)
[71,72])
flexibility HB up (% max. load)
5)
[71])
max ramp (+/- % load/ h)
20)
[interviews])
cold stop load (%)
0)
-)
cold stop min. duration (h)
48)
-)
)
)
CAPEXstor cylinders (USD/kWh) (LHV)
12)
[39])
CAPEXstor cavern (USD/kWh) (LHV)
0.6)
[73])
OPEXstor storage (% CAPEX/y)
1)
[39])
Cost elec. firm-up (USD/MWh)
100)
-)
Table 7: Techno-economic parameters for flexible Haber-Bosch plant
HB synthesis unit. In our model, the power needed for the HB and ASU – at nominal HB load,
10.8% of the chemical energy flow in the HB loop, i.e., 7.3% of the power
p
H2 flowing in the
electrolyser – is taken as much as possible from the renewable power mix as
p
HB. The required
complement of firm-up electricity
p
firm-up is assumed to cost 100 USD/MWh, whether
purchased from the grid or produced locally with one or another form of generation or
storage. Note that the power needed for the pre-compression of the H2 before storage
(assumed at 60 bars), is here included in the total power pH2 needed for the electrolysis, that
is, a specific consumption of 1.5 kWh/kg H2 (0.26 MWh/t NH3) is added to the (EH2/
h
) = 47.6
kWh/kg H2 that are required in the electrolysis modelled above for H2 production only, so that
the electrolyser efficiency including precompression is
h
2 = 68%. Our assumptions for the HB
CAPEX and OPEX are taken from [39].
PV
Buffer H#
storage
Air
unit
Haber-
Bosch
N2
φNH3
Air
Firm-up
elec.
O2
H2O
Wind
Electrolyser
πsolar
πwind
πhyb
πHB
πH2
πcurt
πfirm-up
φH2 φHB

!
Buffer H2 storage. For the buffer H2 storage, reported values cover a wide range from about
200 to 1500 USD/kg depending on the scale and the pressure, with current costs around 500
USD/kg, and ambitious long term targets down to 80 USD/kg [39]. Having in mind rather low
pressure storage (at 60 bar) at industrial scale, we take an moderately optimistic assumption
of 400 USD/kg (12 USD/kWh). It can be noted that storage of H2 it is much cheaper than the
storage of electricity in batteries, at about 200 USD/kWh, but still much more expensive than
the storage of ammonia, at about 0.14 USD/kWh [39]. Note also that H2 storage could be
realised by variable pressure in pipelines, in case transport is needed between the electrolyser
and the NH3 synthesis plant - as will likely be the case for the Engie-Enaex project in Calama,
Chile, the electrolysers being planned to lie near Calama in altitude, while the NH3 synthesis
would take place in Mejillones, near the coast [74].
Flexibility: load modulation, enhanced flexibility, cold stops. Standard HB machines today
operate at high pressure and high temperature. The syngas (mixture of N2, H2 and inerts such
as argon) must be compressed to 100 - 250 bars and heated to about 400 - 500oC before
entering the synthesis loop [20,75]. The required compressor represents the major part of the
CAPEXHB. It should be noted, however, that several companies are developing catalysts able to
work at lower pressure, so as to make renewables-based HB more agile and possibly less
costly, as reported, e.g., in [76,77] at 10 bar, or in [78] at 50 bar. In [76], ramping the load up
by about 25% in about 1 minute was demonstrated. R&D is also being pursued on
electrochemical techniques to produce NH3 directly from air and water [79], however, all such
ideas are in exploratory phase, and their costs are not well defined. Thus, in our modelling we
consider only the well demonstrated and established HB scheme. This route is already partially
being pursued in the recently announced project by Engie and Yara at Pilbara, Western
Australia, where green H2 from solar power will replace a fraction (about 1%) of the grey H2
from SMR in the existing HB plant [80], and in the project of H2U and ThyssenKrupp at Port
Lincoln, South Australia [81].)
We assume that all of the H2 is transformed into NH3, which has an LHV energy density
ENH3=5.17 kWh/kg, although this is not fully exact when working away from nominal conditions
[Cheema 18]. Thus the output chemical energy flux is
f
NH3 = (3/17) x (ENH3/EH2)
f
HB = 0.88
f
HB,
due to the fact that the HB reaction is exothermic.
We define a "standard flexibility" case, where the maximal reduction of
f
HB from its maximum
is 40%, and the HB-ASU is modelled to work at 3 discrete levels: 100%, 80% and 60% of the
maximal flux
f
HBmax. This assumption for flexibility is realistic with present technologies,
according to plant manufacturers (see, e.g., [82] (p 694), [72]). It is important to note that
including no flexibility at all would lead to anomalously high costs (see below Figs 7 and 8). In
agreement with our interviews with manufacturers, we set the maximal rate of change at
20%/hour (relative to the nominal flux). Since, as illustrated below, the largest needs for buffer
H2 storage are related to renewable power variations at the seasonal or synoptic timescales
(see Figs. 7, 8), for example, a week without wind (see Fig. 8.b), we find that the maximal
ramping rates are not a key limiting factor.
In addition to our "standard flexibility" case, we consider the possibility of an "advanced
flexibility" mode, with 80% possible reduction for
f
HB from its maximum. According to several
interviewed manufacturers and a recent patent [72], a 80% flux reduction is feasible, with
ramping times on the order of a few hours. Also, the recent work [71] carried a rigorous
analysis of the parameter envelopes for the HB loop with standard parameters, focusing on
the conditions to maintain the autothermal behaviour. Six adjustable parameters were studied
one by one, and the flexibility that they allowed was quantified. The H2 : N2 ratio was found to

!
give the largest flexibility in terms of H2 intake flux, allowing to reduce it by 67%, when varying
it from the standard 3 : 1, to 1.18 : 2.82. Allowing to vary several parameters at the same time,
the authors expect down flexibility of up to 80% to be reasonably achievable. It can be noted,
as well, that upwards flexibility up to about +10% without losing too much of efficiency, was
also possible, varying the intake of inerts (argon) in the gas blend, or increasing the pressure.
In our modelling, we also allow for 5% upwards flexibility.)
It should be stressed that the efficiency of the HB loop is expected to be reduced when its load
is varied away from nominal conditions. To take this into account, we assume that the
electricity consumption of the HB-ASU has a fixed component of 20% of the nominal (using
0.64 MWh/t NH3), and a variable component proportional to the intake flux
f
HB of hydrogen in
the reactor. This modelling is rough, however, as is clear below (see Fig. 9), the electrical
consumption of the HB machines is a minor component of the total LCOA, much smaller than
the cost of the buffer H2 storage, and the CAPEX of the oversized HB machines. Hence, our
approximation is justified in the present exercise.
In our "advanced flexibility" mode, we also include the possibility to realize "cold stops", that
is, to shut down the autothermal HB reaction completely to 0%. According to manufacturers,
this process, contrary to load modulation, cannot be realized rapidly: shutting down and
starting again involves a waiting time of several days. In our model, we implement this
possibility in relation with the weather forecast. We assume that good knowledge of the
weather, and that renewable power production estimates are available for several days ahead.
If the predicted integrated future production during the next 48 hours is lower than a
threshold, the plant is shut down, and remains shut down for at least 48 hours.
Cost computation. The levelized cost of ammonia LCOA depends, as the LCOH, on the sizes as
and aw of the solar and wind power plants relative to the electrolyser capacity PH2, but also, on
the capacity of the HB machine PHB, and its relative size aHB = PHB / PH2. We define the more
meaningful factor
a
HB = aHB / <
f
H2 > = 1/CFsyn, (3)
that expresses the oversizing of the HB capacity with respect to the mean flux of H2 production
by the electrolyser, and which is by definition, the inverse of the capcity factor of the HB
synthesis reactor CFsyn. It can be noted that
a
HB=1 corresponds to the situation where the H2
buffer storage is sufficiently that it absorbs all the fluctuations in H2 production, so that the HB
operates at perfectly constant flux. Defining CFH2 = <
p
H2> the load factor on the electrolyser, in
this limiting case one would have aHB =
h
2 CFH2.
For each combination of the 3 optimization parameters (as, aw,
a
HB) we compute the LCOA as:)
LCOA(as, aw,
a
HB) = cH2 + celec + cHB-ASU + cfirm-up + cstor + (cH2O - cO2), (4)
where cHB-ASU is the cost of the HB-ASU unit, cfirm-up the cost of the firm-up electricity for the HB-
ASU which cannot be obtained from the renewable power mix, and cstor is the cost of the
buffer H2 storage. We evaluate the latter from the full yearly curve of the tank level (see Fig.
7.c), computing the minimal required tank size tank0 = max(tank) - min(tank), where tank(i), is
the level of H2 in the buffer tank for each hour i=1,...,8760 of the year, and where at each hour
step, the variation of the tank level is computed as :
tank(i+1)- tank(i) =
f
H2 -
f
HB. (5)

!
Note that tank(1)=0, i.e., the tank level is set to 0 for the first hour of the year, but the
absolute value of the tank variable is arbitrary, i.e., only its maximal and minimal values are
relevant and determine the required storage capacity to be installed.
From interviews with manufacturers, we find that buffer storage for N2 is already part of
existing ASU, and is not a significant addition to CAPEX if about 1 day of production is required
as buffer capacity, therefore, we do not consider any variation in the size of the ASU relative to
the NH3 production capacity.
Optimization procedure. The determination of optimal relative sizes of plant sub-units is
analogous to the one for H2 plants: we firstly compute the LCOA for a 3-dimensional matrix (as,
aw,
a
HB), and then find the optimal combination yielding the minimum LCOA.
The summary of our hybrid NH3 plant optimization results is presented below in Fig. 9 and
Tables 8 and 9. Before commenting the whole set of results, we analyse two modelled
examples of flexible NH3 synthesis, with standard and advanced flexibility.
5.2) Example 1 : NH3 plant in Taltal with standard flexibility (40%)
Fig. 7: Optimal plant for flexible NH3 production in Taltal, with 40% flexibility. a) Hourly H2 production
flux from the electrolyser
f
H2* (blue), daily average of
f
H2* (red), and hourly HB flux
f
HB* (green), all
normalized to the maximal electrolyser output
h
2 PH2. b) Zoom. c) H2 storage level, without flexibility
(blue), and with 40% flexibility (green). Black horizontal lines show the trigger level (+/- 0.5 hours full
load) to change the HB load level. d) Zoom.
Figure 7 shows the behaviour of the found optimal hybrid NH3 plant in Taltal, Chile, in the
standard flexibility case. Fig. 7.a shows the hourly production flux of H2 normalized to maximal
electrolyser output
f
H2* =
f
H2 / (
h
2 PH2) (blue curve), as well as its daily average (red curve) and
the intake flux of H2 in the HB reactor with the same normalization
f
HB* =
f
HB / (
h
2
PH2)
.
In the
50 100 150 200 250 300 350
0
0.2
0.4
0.6
0.8
1
H2
* , HB
*
NH3 plant Taltal; a s=1.2; a w=0.43; HB =1.12; flex=40%; a) H 2 flux
0 50 100 150 200 250 300 350
Day
-6
-4
-2
0
days full load
c) H2 tank level - required storage: 0.5 days
126 128 130 132
0
0.2
0.4
0.6
0.8
1
b) H2 flux zoom
126 128 130 132
Day
-0.5
0
0.5
d) H2 tank level zoom

!
zoom of Fig. 7.b, one sees how
f
HB* (green curve) follows, with a delay of a few hours, the
variations
f
H2* (blue curve), going up and down from 100% to 60% during most days.
On Fig 7.c, we see that, without flexibility (blue curve), the required storage tank size, i.e., the
difference between the maximal and minimum tank levels during the year, would be about 6
days of H2 production at full load, which would be prohibitively costly (see also Fig. 11). And we
see that this storage need is dominated by the seasonal cycle. On the contrary with 40%
flexibility (green curve), the seasonal cycle is completely absorbed, the maximal and minimal
tank levels are due to fluctuations at shorter time scales, typically, one or few days with less
sun, and the storage need is only 0.51 days equivalent full electrolyser load. The daily cycle of
the solar, well visible on Fig. 7.b and d, induces a storage need that is about +/- 3 hours of full
load.
5.3) Example 2: NH3 plant in Patagonia Argentina with advanced flexibility (80% + stops)
Figure 8 shows the behaviour of the hybrid optimal HB plant in Patagonia Argentina, with 80%
flexibility. In Fig. 8.b, one sees how the HB flux
f
HB* (red curve) follows closely the hourly H2
production flux
f
H2* (blue curve). In this simulation, 10 "cold stops" to 0% flux (of minimal
duration 48 hours) are recorded, one of them being visible in Fig. 8.b. These stops are
triggered when the tank level (green curve in Fig. 8.c and d) goes below the red line level, on
the condition explained above, that the tank will not fill up too fast during the minimum stop
duration of 48 hours.
Fig. 8: Optimal flexible NH3 production in Patagonia Argentina, with 80% flexibility + stops. a-d) Same as
in Fig. 7. In c and d, the red horizontal line shows the tank level below which a stop can be triggered.
In Fig.8.a and b, one clearly sees how more flexibility implies more capacity oversizing. Indeed,
the average of the working flux
f
HB* is substantially smaller than its maximal value (which is 5%
higher than its nominal value), so that the optimal HB oversizing factor is now
a
HB*=1.48, while
it is only
a
HB*=1.23 in the standard flexibility case for Patagonia Argentina (see Table 8).
5.4) Summary of optimal NH3 production costs
50 100 150 200 250 300 350
0
0.2
0.4
0.6
0.8
1
H2
* , HB
*
NH3 plant Pat. Arg.; as=0; a w=1.4; HB =1.48; flex=80%; a) H2 flux
0 50 100 150 200 250 300 350
Day
-1
-0.5
0
0.5
days full load
c) H2 tank level - required storage: 0.58 days; 10 Cstops
190 192 194 196 198
0
0.2
0.4
0.6
0.8
1
b) H2 flux zoom
190 192 194 196 198
Day
-1
-0.5
0
0.5
d) H2 tank level zoom

!
Figure 9 summarizes the optimal LCOA in our four modelled locations, for buffer storage of H2
in steel tanks, in the standard (left bars) and advanced (right bars) flexibility cases. The found
optimal parameters are summarized in Tables 8 and 9.
Fig. 9: Comparison of optimal LCOA for hybrid NH3 plants. For each location the left bar is standard
flexibility (40%) and the right bar advanced flexibility (80% + stops). Each element includes CAPEX and
OPEX for the considered unit.
Several general observations can be made from Fig. 9. First, concerning hybridization, we note
that it is substantially more favoured for NH3 plants than for H2 plants (see Fig. 5), especially at
lower flexibility, where hybrid renewables mixes occur in all locations but Patagonia Chile,
providing cost reductions (gains) ranging from 5% to 13%, which is substantially more than the
maximum of 1.6% found for H2 production in Taltal (see Table 6).
A second remark concerns the very large cost of the H2 buffer storage for the wind-dominated
Patagonia locations, when flexibility is limited to “standard”. Indeed, in both southern cases,
where the wind variability is very strong, large buffer storage capacities of 4.4 and 3.5 days
equivalent full load are needed with standard flexibility (see Table 8). As this is very costly, the
optimal configuration requires substantial oversizing of the renewable power supply, leading
to hybrid load factors of 67.6% and 72% (see Table 8), and large fractions of 11% and 13% of
curtailed electricity. Only with such oversizing can the variability of the wind power and thus,
the storage costs be reduced.
A third remark is that advanced flexibility (80% + stops) strongly reduces the costs of H2
storage, especially in the wind-dominated Patagonia locations. There, buffer storage needs are
divided almost by 10 by enhanced flexibility, falling from about 4 days (Table 8) to 0.5 days
(Table 9) of full load H2 production. Also, as seen on Fig. 9, because the enhanced flexibility
dramatically cuts the costs of the wind fluctuations, it allows including more of this cheaper
wind power in the mix. The need to oversize the wind plant capacity in order to mitigate its
variability, also vanishes. Indeed, the electrolyser load factors CFH2 are reduced on average for
Patagonia locations from 70% to 63.5% with enhanced flexibility, and curtailment falls on
average from 12% to 3.4%. On the other hand, for both Atacama locations, where fluctuations
of the solar resource are much more moderate, advanced flexibility reduces the storage need
only slightly, from 0.58 days on average to 0.33 days.
A fourth remark is that more flexibility results in larger HB oversizing factors aHB everywhere,
but this effect is quite stronger in the windy Patagonia locations, where aHB increases from 1.3
on average to 1.48, while in the solar-dominated Atacama locations, it only increases on
Taltal Calama Pat. Chile Pat. Arg
0
100
200
300
400
500
600
700
USD / t NH3
LCOA with HB flex = 40% / 80% + stops (# stops = 0 0 11 10)
Electrolyzer
Solar elec. H 2
Wind elec. H 2
Electricity HB-ASU
HB-ASU
H2 buffer storage
HB firm-up elec.

!
average from 1.13 to 1.18, due again, to the much more regular solar power, compared with
the strongly varying wind productions.
Flexibility 40%
Taltal
Calama
Pat. Chile
Pat. Arg.
Capacity solar as*
1.2
1.3
0
0.86
Capacity wind aw*
0,43
0.34
1.57
1.36
HB oversizing
a
HB*
1.13
1.13
1.36
1.27
Hybrid load CF* (%)
52.8
49.2
67.6
72
curt* (%)
2.2
3.7
11
13
stor* (days full load)
0.5
0.66
4.4
3.5
firm-up* (% of HB elec.)
6
15
5.3
3.8
hybridization cost reduction (%)
8.8
5.4
0
13
LCOA* (USD/t)
487
521
580
710
Table 8: Optimal parameters for hybrid NH3 plants with standard HB flexibility of 40%
Table 9: Optimal parameters for hybrid NH3 plants with advanced HB flexibility of 80% + stops
Cost performances. With advanced flexibility, Patagonia Chile yields the lowest achievable
LCOA at 462 USD/t, lower than the 483 USD/t in Taltal, even though the wind at 28 USD/MWh
(see Table 6) is more costly than the solar at 26.7 USD/MWh in Taltal. As seen in Fig. 9, the
driving factor is the capacity factor on the electrolyser of 63.3% in Patagonia Chile, larger than
the 52.8% in Taltal, which reduces the CAPEX in electrolyser, even though the CAPEX for the
HB is larger in Patagonia Chile, where the oversizing of the HB (with respect to the NH3
production) is 1.48, while it is only 1.18 in Taltal.
Finally, as for the LCOH, we see that in Patagonia Argentina, the wind resource is as good or
better as in neighbouring Patagonia Chile, and here the solar resource is worth being included
in hybridization, when flexibility is low. Still, the unfavourable financial conditions (WACC of
10% versus 7% in Chile) make the LCOA substantially higher in Argentina.)
5.4) Analysis of the gain (LCOA reduction) from hybridization
Since hybridization is found to possibly bring about interesting cost reductions, especially at
lower HB flexibility, let us now analyse the mechanisms involved, focusing on our two favourite
example locations. Figure 10 shows the LCOA cost structure for Taltal (a) and Patagonia
Argentina (b), for standard flexibility and for hybrid RE mixes (right bars) versus single
technology cases (left bars). Note that in all cases, optimal configurations are shown, where for
Flexibility 80% + stops
Taltal
Calama
Pat. Chile
Pat. Arg.
Capacity solar as*
1.2
1.3
0
0
Capacity wind aw*
0.43
0.17
1.36
1.36
HB oversizing
a
HB*
1.18
1.18
1.48
1.48
Hybrid load CF* (%)
52.8
44.5
63.3
63.9
curt* (%)
2.2
2
2.8
3.9
stor* (days full load)
0.31
0.35
0.41
0.58
firm-up* (% of HB elec.)
6
18
2.3
5.1
hybridization cost reduction (%)
5.4
1.4
0
0
LCOA* (USD/t)
483
506
462
571

!
single technology cases, the optimization concerns only two parameters: the relative size of
the RE pant, and of the HB unit.
Fig. 10: Analysis of hybridization gain for NH3 production in a) Taltal and b) Patagonia Argentina, with
standard flexibility (40%). For each location, left bars show the cost splitting for optimal single
technology case, and right bars for optimal hybrid case
For Taltal (Fig.10.a), adding wind power increases the load factor end thus reduces the
electrolyser cost, however, this is mostly offset by the increase in electricity cost due to more
expensive wind. The 9% reduction of the LCOA is here essentially obtained by the reduction in
the need for firm-up electricity, and for buffer H2 storage. In fact the key mechanism, is that
the added wind electricity, although more expensive, allows to stabilize the power supply, so
that less firming electricity has to be acquired, and less H2 needs being stored.
For Patagonia Argentina (Fig. 10.b), the key driver of the 13% cost reduction is the buffer H2
storage, which is much larger with only wind. In the hybridized case, even though some more
expensive solar is blended with the cheaper wind, the reduced oversizing of the wind plant and
thus reduced curtailment makes the spending in electricity equivalent to the case with only
wind. Here again, the use of some solar allows to stabilize the wind resource, and reduces the
cost induced by its variability.)
5.5) The impact of HB flexibility
To better understand the impact of HB flexibility on green NH3 plants, Figs. 11 shows the
optimal parameters for hybrid NH3 plants in Taltal (A) and Patagonia Argentina (B), without
cold stops, and flexibility from 0 to 100%. The most important observation is that, in both
cases, for increasing HB flexibility, both the LCOA and the H2 buffer storage capacity decrease
in a very similar way, which confirms that the H2 storage is the key cost driver related to the
(lack of) flexibility. In the solar dominated Taltal case (Fig. 11.A), most of the decrease in H2
storage size and LCOA is obtained within the first 20% of flexibility, which corresponds to the
seasonal variability for this RE mix, and which is essentially already available in present day HB
units. In the wind dominated Patagonia Argentina case (Fig 11.B), both the LCOA and the
required H2 storage decrease rapidly for HB flexibility going from 0 to 40%, which also
corresponds to absorbing the seasonal variability, but they also keep decreasing substantially
when flexibility goes all the way up to 100%.
0
100
200
300
400
500
600
700
800
900
USD / t NH3
Pat. Arg.; flex=40%; LCOA=710 USD/t HB
*=1.27
No hybri : LCOA =815 USD/t; HB
*=1.27. Gain =13%
Electrolyzer
Solar elec. H2
Wind elec. H2
Elec. HB-ASU
HB-ASU
H2 buffer stor.
HB firm-up
0
100
200
300
400
500
600
USD / t NH3
Taltal; flex=40%; LCOA=487 USD/t HB
*=1.12
No hybri : LCOA =535 USD/t; HB
*=1.12. Gain =8.8%
Electrolyzer
Solar elec. H2
Wind elec. H2
Elec. HB-ASU
HB-ASU
H2 buffer stor.
HB firm-up
a) b)

!
Fig. 11: Impact of HB flexibility on optimal NH3 plant in A) Taltal and B) Patagonia Argentina, without
cold stops. For each location: a) LCOA. b) Capacities as and aw of VRE plants, HB oversizing factor
a
HB and
fraction %w of wind in energy mix c) Required H2 storage capcity. d) Fractions of curtailed and firm-up
electricity for HB loop, and LCOA gain by hybridization.
In Taltal, the wind energy fraction %w (Fig 11.A.b), and thus also the amount of firm-up
electricity (close to 6%) are not strongly affected by increasing flexibility. With increasing
flexibility, the HB oversizing factor
a
HB continuously increases from 0.99 to 1.18, while the
0 0.5 1
HB flexibility
500
600
700
800
900
1000
USD/t
NH3 plant Pat. Arg. ; CFs=20.7% ; CFw=52.7%
LCOA
0 0.5 1
HB flexibility
0
0.5
1
1.5
aw as HB %w
0 0.5 1
HB flexibility
0
5
10
days full load
H2 storage
0 0.5 1
HB flexibility
0
5
10
15
20
25
%
firm-up elec.
curt. elec.
hibridiz. gain
LCOEs=51.8USD/MWh
LCOEw=33.8USD/MWh
0 0.5 1
HB flexibility
450
500
550
600
650
700
USD/t
NH3 plant Taltal ; CFs=32.5% ; CFw=43.8%
LCOA
0 0.5 1
HB flexibility
0
0.5
1
1.5
aw as HB %w
0 0.5 1
HB flexibility
0
2
4
6
days full load
H2 storage
0 0.5 1
HB flexibility
0
5
10
15
20
%
firm-up elec.
curt. elec.
hibridiz. gain
LCOEs=26.7USD/MWh
LCOEw=35.8USD/MWh
a)
a)
b)
b)
c)
c) d)
d)
A)
B)

!
solar capacity as and the curtailed fraction continuously decrease. The gain by hybridization is
larger at flexibility less than 40%, i.e., as expected, when the RE variability is costly.
In Patagonia Argentina, for HB flexibility under 40%, the share %w of wind in the mix is a bit
enhanced, due to the important seasonal variability of the solar (see Fig. 4). However, for
flexibility above 40%, one sees a continuous increase in the share %w of wind, which is locally
the cheapest energy. The HB oversizing also continuously increases from 0.96 to 1.48, as
expected for more flexible operation, while the fraction of curtailed electricity continuously
decreases from about 20% at low flexibility to almost 0% at 100% flexibility. The fraction of
firm-up electricity decreases for increasing flexibility, but presents a slight increase for the last
point at 100% flexibility, when solar electricity disappears from the mix, due to a sudden
increase of the number of day hours with no renewable power supply.
The maximal gain by flexibility (from 0 to 100%), is a reduction of the LCOA by 27% in Taltal,
and 40% in Patagonia Argentina, the difference being due to the stronger variability of the
wind. Said differently, without any flexibility, the variability of the RE power causes a cost
increase of 67% in Patagonia Argentina, and 37% in Taltal, compared to the case with 100%
flexibility. These estimates are quite lager than the about 8% cost difference found in [17] for
flexibility varying from 80% to 0%, for mostly wind-based ammonia production in the UK.
5.6) Strategic view for NH3 production
Based on these results, NH3 production in northern Chile seems a particularly low hanging
fruit, due to the existence in the area of a mature solar market, with an excellent resource,
interesting wind options, available water desalination technology, and large local consumption
of ammonia due to mining, in particular for the direct use in explosives (and oxygen as a by-
product). Fig. 23 shows the market case, comparing recent prices of imported grey ammonia
(from natural gas), and our estimate for 2020 conditions in Taltal with advanced flexibility.
Fig. 12: Market costs of ammonia in Chile, compared with our Taltal estimate with advanced flexibility at
483 USD/t. Source: [83]
In Patagonia, both Chilean and Argentinian, very promising opportunities could also emerge
for producing green H2 or NH3 from wind, and perhaps some solar in the northernmost
locations, possibly firmed-up by some stranded hydro resources that could also help to buffer
the variability. Due to the quality of the resource and the vast desert spaces, such cases could
aim at large scale export of synthetic fuels via pipelines and ships, for example towards Japan
[18].
2004 2006 2008 2010 2012 2014 2016 2018
0
100
200
300
400
500
600
700
USD / t NH3
Ammonia cost in Chile
SMR imported
LCOA green Tal Tal 2020

!
Finally, for illustration, Table 10 shows the details for a 35 000 t/yr green ammonia plant in
Taltal, with advanced flexibility. The largest share of the CAPEX, and also of the footprint,
would be the 72.3 MW solar farm, complemented by a 25.8 MW wind farm. The required
buffer storage tank size is of 309 MWh/EH2 = 9.3 t H2, that is, about or 1 700 m3 at 60 bars and
the total CAPEX would be 148 MUSD.
)
Table 10: Summary of optimal capacities and CAPEX, example case in Taltal, for an ammonia plant
producing 35 000 t/yr, with advanced flexibility. (*: MWh)
6) Conclusion
In this work we have developed a methodology that can be applied to any location in the
world, provided that reliable meteorological and realistic economic data are available, to
compute costs of flexible production of H2 and NH3)from variable wind and solar energy.)
Our study brings novel understanding about hybridization for RE-based H2 and alternative fuels
production, beyond some general insights provided, e.g. in [15], where hybrid wind-solar
plants were studied only in fixed ratios of capacities. For H2, we find that combining solar and
wind resources can reduce production costs only in small amounts, e.g., by 1.6% in our most
favourable case of Taltal. On the other hand, for NH3 production, we find that gains by
hybridization can be substantially larger, in the 5-20% range, because it allows to stabilize the
RE power supply, and thus reduce the need for expensive buffer storage of H2.
This work also clarifies the key role of flexibility in the production flux of green ammonia. Our
modelling shows how enhanced flexibility continuously reduces the storage cost of gaseous H2,
and the total LCOA. The cost increase due to RE variability is substantially larger for the
strongly variable wind resource: up to 67% in our Patagonia Argentina case with 0% flexibility
compared to 100% flexibility. For solar power, which is more regular, the cost increase reaches
up to 37% in our Taltal case. Our work also reveals how flexibility can modify the whole
structure of the optimal plants considered, and points towards the key value of flexibility in
green ammonia projects, and the importance of devoting sufficient R&D to address this major
cost driver. Note, of course, that if underground storage of H2 is locally available, the storage
and variability cost could be drastically reduced.
In these locations all boasting world-class wind or solar, we have found that green H2 and NH3
could be soon produced at costs almost competitive with the traditional fossil fuel alternatives.
Our estimated LCOH of about 2 USD/kg, comparable to the estimated 2.16 USD/kg for wind in
Argentina in [18] or the 1.8 - 3 USD /kg for solar in Northern Chile for 2023 [68], are starting to
challenge the 1 - 2 USD/kg for steam methane reforming (SMR) [5], and our LCOA below
500 USD/t are already commensurate to the 300-600 USD/t market value of imported NH3
from SMR in Chile. Thus, the expected cost reductions for electrolysers and solar and wind
energy, make the production of green H2 and NH3, well identifiable business cases in the short
term. Factoring in the economic value of price stability (versus fossil fuels price volatility), as
well as climate benefits or possibly increased carbon taxes would make the case even more
compelling.
HB + ASU
Electrolyser
Solar
Wind
H2 buffer
Total
Capacity (MW)
27.6
60.2
72.3
25.8
309*
-
CAPEX (MUSD)
21.3
36.1
53.5
33.6
3.7
148

!
Future further analysis could refine the modelling, e.g., including the variable efficiency of the
electrolyser at partial load and beyond nominal load. Also, the flexibility assumptions will need
to be tested and their costs assessed, in particular concerning the behaviour of the catalysts, in
upcoming real world projects.
Acknowledgements: This work was funded by the IEA, thanks to grants from the French
government through ADEME, and the German government. The views expressed are those of
the authors and do not necessarily reflect those of the IEA or its member countries.
Declaration of interests: none.
7) References:
[1] Ramanathan V, Molina L, Zaelke D, Borgford-Parnell N, Xu Y, Alex K, et al. Well under 2
Degrees Celsius: Fast Action Policies to Protect People and the Planet from Extreme
Climate Change 2017.
[2] World Bank. Turn down the heat : why a 4°C warmer world must be avoided. 2012.
[3] IPCC. Global Warming of 1.5°C. An IPCC Special Report on the impacts of global warming of
1.5°C above pre-industrial levels and related global greenhouse gas emission pathways, in
the context of strengthening the global response to the threat of climate change,
sustainable development, and efforts to eradicate poverty. 2018.
[4] UNEP. Emissions Gap Report 2018. 2018.
[5] Philibert C. Renewable energy for industry. Paris Int Energy Agency 2017.
[6] IEA. The Future of Hydrogen. OECD, Paris; 2019.
[7] IRENA. Hydrogen from renewable power: Technology outlook for the energy transition.
2018.
[8] IRENA. Hydrogen: A renewable energy perspective. 2019.
[9] Andersson J, Grönkvist S. Large-scale storage of hydrogen. Int J Hydrog Energy 2019.
[10] IEA. WEO 2018. OECD, Paris; 2018.
[11] Götz M, Lefebvre J, Mörs F, Koch AM, Graf F, Bajohr S, et al. Renewable Power-to-Gas: A
technological and economic review. Renew Energy 2016;85:1371–90.
[12] Hank C, Gelpke S, Schnabl A, White RJ, Full J, Wiebe N, et al. Economics & carbon dioxide
avoidance cost of methanol production based on renewable hydrogen and recycled
carbon dioxide–power-to-methanol. Sustain Energy Fuels 2018;2:1244–61.
[13] Tunå P, Hulteberg C, Ahlgren S. Techno-economic assessment of nonfossil ammonia
production. Environ Prog Sustain Energy 2014;33:1290–7.
[14] Morgan E, Manwell J, McGowan J. Wind-powered ammonia fuel production for remote
islands: A case study. Renew Energy 2014;72:51–61.
[15] Fasihi M, Bogdanov D, Breyer C. Techno-economic assessment of power-to-liquids (PtL)
fuels production and global trading based on hybrid PV-wind power plants. Energy
Procedia 2016;99:243–68.
[16] Schulte Beerbühl S, Fröhling M, Schultmann F. Combined scheduling and capacity planning
of electricity-based ammonia production to integrate renewable energies. Eur J Oper Res
2015;241:851–62.
[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;57:14607–16.

!
[18] Heuser P-M, Ryberg DS, Grube T, Robinius M, Stolten D. Techno-economic analysis of a
potential energy trading link between Patagonia and Japan based on CO2 free hydrogen.
Int J Hydrog Energy 2019;44:12733–47.
[19] Philibert C. Producing ammonia and fertilizers: new opportunities from renewables. IEA,
OECD, Paris; 2017.
[20] Bartels JR. A feasibility study of implementing an ammonia economy. Iowa State
University, 2008.
[21] Santana C, Falvey M, Ibarra M, García M. Energias Renovables en Chile. El potencial eolico,
solar e hidroelectrico de Arica a Chiloe. Santiago Chile MINENERGIA GIZ 2014.
[22] Ministerio de Energia de Chile. Explorador Solar 2013.
http://www.minenergia.cl/exploradorsolar/.
[23] Ministerio de Energia de Chile. Explorador Eólico 2013.
http://walker.dgf.uchile.cl/Explorador/Eolico2/.
[24] Sigal A, Leiva EPM, Rodríguez CR. Assessment of the potential for hydrogen production
from renewable resources in Argentina. Int J Hydrog Energy 2014;39:8204–14.
[25] World Bank. Global Wind Atlas n.d. https://globalwindatlas.info.
[26] World Bank. Global Solar Atlas n.d. https://globalsolaratlas.info/.
[27] IEA. Countries n.d. https://www.iea.org/countries/.
[28] IEA. Energy Policies Beyond IEA Countries: Chile 2018 Review. OECD, Paris; 2018.
[29] Ministerio de Energia de Chile. Capacidad instalada – Energía Abierta n.d.
http://energiaabierta.cl/visualizaciones/capacidad-instalada/.
[30] BNEF. The Clean Technology Fund and Concessional Finance. 2019.
[31] Wind Power Monthly. The 7 best performing wind markets in 2018 2019.
[32] IEA. Renewables 2018. Analysis and Forecast to 2023. 2018.
[33] IRENA. Renewable Power Generation Costs in 2017. 2019.
[34] ACERA. Análisis ACERA-Resultados del proceso de Licitación 2017/01. 2017.
[35] Ministerio de Energia de Argentina. Resumen de Ofertas Adjudicadas RenovAr 2 FASE 1
2017.
[36] FAO. World fertilizer trends and outlook to 2020. 2017.
[37] Bañares-Alcántara R, Dericks G, Fiaschetti M, Grünewald P, Lopez JM, Tsang E, et al.
Analysis of islanded ammonia-based energy storage systems. 2015.
[38] Giddey S, Badwal SPS, Munnings C, Dolan M. Ammonia as a renewable energy
transportation media. ACS Sustain Chem Eng 2017;5:10231–9.
[39] Ikäheimo J, Kiviluoma J, Weiss R, Holttinen H. Power-to-ammonia in future North
European 100% renewable power and heat system. Int J Hydrog Energy 2018;43:17295–
308.
[40] Valera-Medina A, Xiao H, Owen-Jones M, David WIF, Bowen PJ. Ammonia for power. Prog
Energy Combust Sci 2018;69:63–102.
[41] Kobayashi H, Hayakawa A, Somarathne KKA, Okafor EC. Science and technology of
ammonia combustion. Proc Combust Inst 2019;37:109–33.
[42] Ash N, Scarbrough T. Sailing on solar: Could green ammonia decarbonise international
shipping? Environmental Defense Fund; 2019.
[43] Laursen RS. Ship Operation Using LPG and Ammonia As Fuel on MAN B&W Dual Fuel ME-
LGIP Engines 2018.
[44] Anderson MK. Ammonia safety, a global perspective. 2017.
[45] Ministerio de Energia de Chile. Proceso de Planificación Energética de Largo Plazo. 2018.
[46] Perez A. Hychico. Clean Hydrogen Production and Renewable Energy Storage 2017.
[47] Hirth L, Müller S. System-friendly wind power: How advanced wind turbine design can
increase the economic value of electricity generated through wind power. Energy Econ
2016;56:51–63.
[48] Dalla Riva A, Hethey J, Vītiņa A. Impacts of wind turbine technology on the system value of
wind in europe. Tech. rep., International Energy Agency; 2017.

!
[49] Wiser R, Bolinger M, Barbose G, Darghouth N, Hoen B, Mills A, et al. 2018 Wind
Technologies Market Report 2018:103.
[50] The Wind Power. Taltal (Chile) - Wind farms - Online access n.d.
https://www.thewindpower.net/windfarm_en_20644_taltal.php.
[51] Brower M. Wind resource assessment: a practical guide to developing a wind project. John
Wiley & Sons; 2012.
[52] Stehly TJ, Beiter PC, Heimiller DM, Scott GN. 2017 Cost of Wind Energy Review. National
Renewable Energy Lab.(NREL); 2018.
[53] Renewables.ninja n.d. https://www.renewables.ninja/.
[54] Marion B, Adelstein J, Boyle K ea, Hayden H, Hammond B, Fletcher T, et al. Performance
parameters for grid-connected PV systems. Conf. Rec. Thirty-First IEEE Photovolt. Spec.
Conf. 2005, IEEE; 2005, p. 1601–6.
[55] The Wind Power. Diadema (Argentina) - Wind farms - Online access n.d.
https://www.thewindpower.net/windfarm_en_15923_diadema.php.
[56] ThyssenKrupp. Power to gas > Water Electrolysis > Products > Home. Uhde Chlorine Eng
n.d. https://www.thyssenkrupp-uhde-chlorine-engineers.com/en/products/water-
electrolysis-hydrogen-production/power-to-gas.
[57] Schmidt O, Gambhir A, Staffell I, Hawkes A, Nelson J, Few S. Future cost and performance
of water electrolysis: An expert elicitation study. Int J Hydrog Energy 2017;42:30470–92.
[58] 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;82:2440–54.
[59] Lokke JA. Nel Group 2017.
[60] Proost J. State-of-the art CAPEX data for water electrolysers, and their impact on
renewable hydrogen price settings. Int J Hydrog Energy 2019;44:4406–13.
[61] Saba SM, Müller M, Robinius M, Stolten D. The investment costs of electrolysis–a
comparison of cost studies from the past 30 years. Int J Hydrog Energy 2018;43:1209–23.
[62] Parra D, Valverde L, Pino FJ, Patel MK. A review on the role, cost and value of hydrogen
energy systems for deep decarbonisation. Renew Sustain Energy Rev 2019;101:279–94.
[63] Comision Nacional de Energía. Informe costos de generación 2019.pdf. 2019.
[64] CNREC. China Renewable Energy Outlook 2018. 2018.
[65] Maghami MR, Hizam H, Gomes C, Radzi MA, Rezadad MI, Hajighorbani S. Power loss due
to soiling on solar panel: A review. Renew Sustain Energy Rev 2016;59:1307–16.
[66] Staffell I, Green R. How does wind farm performance decline with age? Renew Energy
2014;66:775–86.
[67] Kind S. Argentina - Renewables vs. natural gas (economics) 2019.
https://www.linkedin.com/feed/update/urn%3Ali%3Aactivity%3A6530634661180297216/
.
[68] Tractebel. Oportunidades para el desarrollo de una industria de hidrógeno solar en las
regiones de Antofagasta y Atacama: Innovación para un sistema energético 100%
renovable. 2018.
[69] GIZ. Tecnologías del Hidrógeno y perspectivas para Chile. 2018.
[70] Agora Energiewende and Frontier Economics. The future cost of electricity-based synthetic
fuels. 2018.
[71] Cheema II, Krewer U. Operating envelope of Haber–Bosch process design for power-to-
ammonia. RSC Adv 2018;8:34926–36.
[72] Ostuni R, Zardi F. Method for load regulation of an ammonia plant, 2016.
[73] ITP. Comparison of Dispatchable Renewable Electricity Options - Technologies for an
orderly transition. ITPEnergised 2018. http://www.itpenergised.com/wp-
content/uploads/2018/11/Comparison-Of-Dispatchable-Renewable-Electricity-Options-
ITP-et-al-for-ARENA-2018.pdf (accessed September 25, 2019).
[74] Enaex. Private communication 2019.

!
[75] Morgan ER. Techno-economic feasibility study of ammonia plants powered by offshore
wind 2013.
[76] Beach J. Starfire Energy : Rapid Ramp NH3 reactor update 2018.
[77] Welch A. Starfire Energy : Advanced Heterogeneous Catalysts for Renewable Ammonia
Synthesis 2018.
[78] JGC corp. Demonstration of CO2-free Ammonia Synthesis using Renewable Energy-
Generated Hydrogen 2018.
[79] NH3 fuel association. 2018 NH3 Fuel Conference, Session “Ammonia Energy Technology
Roadmap.” NH3 Fuel Assoc 2018. https://nh3fuelassociation.org/conference2018/.
[80] Engie. ENGIE and YARA take green hydrogen into the factory 2019.
https://www.engie.com/en/news/yara-green-hydrogen-factory/ (accessed September 26,
2019).
[81] Brown T. ThyssenKrupp’s “green hydrogen and renewable ammonia value chain.” Ammon
Ind 2018. https://ammoniaindustry.com/thyssenkrupps-green-hydrogen-and-renewable-
ammonia-value-chain/.
[82] Stolten D, Scherer V. Transition to renewable energy systems. John Wiley & Sons; 2013.
[83] Enaex. 2017 Annual Report. 2018.