Applied Energy 247 (2019) 438–453

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

A hydrogen supply chain with spatial resolution: Comparative analysis of T

infrastructure technologies in Germany
Markus Reußa, , Thomas Grubea, Martin Robiniusa, Detlef Stoltena,b
⁎

a
Forschungszentrum Jülich, IEK-3: Institute of Electrochemical Process Engineering, D-52425 Jülich, Germany
b
RWTH Aachen University, Chair for Fuel Cells, Faculty of Mechanical Engineering, Kackertstraße 9, D-52072 Aachen, Germany

HIGHLIGHTS

• Techno-economic modeling of future hydrogen supply chains with spatial resolution.

• Spatially-resolved infrastructure design for hydrogen transport.
• Large-scale seasonal storage for renewable hydrogen production.
• Implementation of different technologies for hydrogen storage and transportation.
• Evaluation of optimization potential for hydrogen distribution.

ARTICLE INFO ABSTRACT

Keywords: Hydrogen could play a key role in future energy systems by enabling the storage of excess electricity from
Renewable energy renewable power sources, like solar and wind, and fueling emission-free fuel cell electric vehicles. Nevertheless,
Hydrogen mobility the temporal and spatial gap between the fluctuating production in electrolysis plants and the demand at fueling
Hydrogen infrastructure stations necessitates the construction of infrastructures. Different technologies are available for storing and
Hydrogen supply chain
transporting hydrogen in its gaseous or liquid states, or even via liquid organic hydrogen carriers. To select and
Spatial infrastructure design
Liquid organic hydrogen carriers
compare these different infrastructure options on a nationwide scale in Germany for an energy system 2050, we
carried out an infrastructure assessment with spatial resolution to analyze the resulting costs and CO2 emissions,
as well as the primary energy demand. To do so, methods for designing a spatially-resolved infrastructure are
presented. In particular, the setup of a transmission pipeline with gaseous trailer distribution has not been well
represented and investigated in the literature so far. The results show that salt caverns, as well as transmission
pipelines, are key technologies for future hydrogen infrastructure systems. The distribution should be handled
for low penetration of fuel cell vehicles rates with gaseous compressed trailers and replaced by distribution
pipelines in areas with high fueling station densities. This ensures the cost-effective supply during the transition
to higher fuel cell vehicle fleets.

1. Introduction Nevertheless, the production and utilization of hydrogen are char-

acterized by a temporal and spatial gap that must be closed by infra-
The progressive expansion of renewable energy technologies uti- structure for storage and transport. Different options are available for
lizing wind and solar resources is seen as a major step towards tackling each purpose like transport by truck or pipeline or storage in liquid or
climate change, even if the fluctuating nature of electricity thus pro- gaseous state.
duced remains a challenge for the power sector. Meanwhile, the Bolat and Thiel [1,2] give an overview of modeling approaches to
transport sector lags behind in reaching CO2 abatement targets. assess hydrogen infrastructure, separating these analytically and
Hydrogen offers a solution that would be beneficial for both sectors: mathematically. Analyzing in depth the application of the developed
Producing hydrogen during times of oversupply would lead to addi- methodologies, two different application cases can be recognized.
tional revenues in the power sector, while fuel cell electric vehicles On the one side, several studies have examined a spatially-resolved
(FCEV) would offer emission-free mobility in the transport sector. infrastructure design for national supply strategies: Seydel [3] used a

⁎
Corresponding author.
E-mail address: m.reuss@fz-juelich.de (M. Reuß).

https://doi.org/10.1016/j.apenergy.2019.04.064
Received 4 December 2018; Received in revised form 8 April 2019; Accepted 14 April 2019
0306-2619/ © 2019 Elsevier Ltd. All rights reserved.
M. Reuß, et al. Applied Energy 247 (2019) 438–453

GIS-based optimization model to elaborate the cost-optimal hydrogen hydrogen requires renewable energy sources. Yang and Ogden [22]
supply chain with regard to hydrogen production, transport, and re- compared hydrogen transport options for transmission and distribution,
fueling for a future FCEV-dominated vehicle market. Regarding hy- but not as an entire supply chain, from production to refueling. The
drogen transport, he implemented pipeline and truck transport. Pipeline Hydrogen Delivery Scenario Analysis Model (HDSAM) [23], as well as
routing was conducted across the minimum cost paths with respect to the Hydrogen Refueling Station Analysis Model (HRSAM) [24], belong
territorial properties, whereas the truck routing was conducted using the to the H2A model family, which offers a detailed calculation of supply
German street network. Krieg [4] and Baufumé et al. [5] designed a chain costs without spatial resolution. Yang and Ogden [22], as well as
nationwide pipeline system with an iterative network development al- the HDSAM [23], separated the transport of hydrogen into transmission
gorithm based on a shortest path algorithm weighted by costs. The focus for long-distance transport and distribution for the “last mile” to the
here was on technical feasibility. They investigated two production op- fueling station carried out by different technologies. Furthermore,
tions: Lignite gasification and wind electrolysis. They implemented sea- Teichmann et al. [25–27] and Moroz et al. [28] compare the costs of
sonal storage, as well as a reserve capacity of 90 days. Robinius et al. technologies, but both represent specific case studies to highlight a
[6,7] used the pipeline algorithm from Baufumé et al. [5] to examine a technology without taking the entire supply chain into account.
hydrogen supply chain based solely on the utilization of excess electricity Teichmann et al. [25–27] highlighted the application of LOHCs, while
from the power sector. The authors showed that electrolysis from wind Moroz et al. [28] propose the utilization of metal hydrides for hydrogen
energy in the North of Germany would be sufficient to supply 75% of the storage and transport. Wulf et al. [29] conducted a life cycle assessment
German LDV market with hydrogen, installing 28 GW of electrolysis (LCA) of a whole hydrogen supply chain, but do not have a closer look
capacity operating at 5300 full load hours. With regard to hydrogen at the final cost of the system.
storage, they considered a 60-day storage capacity that took seasonal All in all, the national supply systems aim to design one cost-opti-
fluctuations into account, as well as a strategic reserve. Furthermore, mized system. Most of them, however, lack information about alter-
Moreno-Benito et al. [8], Almansoori and Betancourt-Torcat [9], as well native solutions and competing technologies, as well as technological
as Samsatli et al. [10] calculated a hydrogen infrastructure design for the details. In addition, the production technology mostly sets the para-
UK by means of mixed-integer linear optimization. Samsatli et al. [10] meters of the analysis. Technology-focused studies elaborate on the
aimed to design a renewable-based hydrogen infrastructure. Therefore, competitiveness of different technologies for various application areas
they need a temporal resolution and the implementation of storage and parts of the supply chain with regard to a variety of evaluation
technologies for the fluctuating power production of renewable energy criteria like CO2 emissions and energy demand. So far, the literature is
sources. A similar approach was used in Welder et al. [11] in a case study missing a bridge between technology-focused and spatially-resolved
for Germany. The implementation of storage was neither considered by studies, evaluating different technologies on a nationwide scale.
Almansoori and Betancourt-Torcat [9], nor by Moreno-Benito et al. [8]. This study aims to close this gap by evaluating all parts of the supply
Instead, they implemented various hydrogen production options. Ochoa chain, from hydrogen production to refilling, on a nation-wide scale in
Biqué und Zondervan [12] extended the study of Almansoori and Be- the case of Germany for the target year 2050, considering a spatial
tancourt-Torcat [9] by a pure renewable-based supply chain. However, resolution with regard to costs, primary energy demand and CO2
the calculated costs were being too high due to a calculation error in the emissions. To examine the competitiveness of different technologies
electrolysis calculation by 2.4 USD/kg. Yáñez et al. [13] promoted a against one another, a simulation approach, which allows a higher
hydrogen supply chain network supported by waste hydrogen to lower degree of detail in each step of the supply chain compared to an opti-
fuel costs during the transition to a renewable system in northern Spain. mization, is chosen. To do so, the model of Reuß et al. [18] is extended
De-Léon Almaraz et al. [14,15] investigated the design of a hydrogen by spatial resolution to adequately represent the transport technologies
supply chain network in France based on liquid hydrogen through a of pipeline and truck. The adequate representation of the separation of
multi-objective optimization incorporating costs, CO2 emissions and risks transport into transmission and distribution has not been performed in
into the objective function. André et al. [16,17] investigated the devel- the literature so far and will be examined in detail in this publication.
opment of a pipeline network for France to minimize pipeline costs. They Section 2 will explain the used methods and data, separated into the
did not consider production, storage, and distribution at all in their work. technological background, the explanation of the supply chain model
All of these studies have in common the fact that the final results are fully developed by Reuß et al. [18], the spatial extension, and the scenario
determined by the production technology. Hydrogen transport is con- assumptions.
sidered to set the balance between subregions and not to get deeper in-
sights into the behavior of the transport technology or to design a rea- 2. Method and data
listic network. Detour factors for the truck transport to represent real
routing in streets are also considered by Seydel [3] and De Léon Almaraz This section will introduce the developed methodology, as well as
et al. [14,15], while pipeline detouring to avoid natural reserves and the data used for acquiring the results. Thus, this work will use a two-
densely populated areas are considered by Baufumé et al. [5], as well as step approach: The basic model for the comparison of different supply
Seydel [3] and Robinius et al. [7]. Almansoori and Betancourt [9], as chains is adapted from Reuß et al. [18], who developed a modular
well as Ochoa Biqué and Zondervan [12] even consider rail transport hydrogen supply chain architecture model for investigating the appli-
without modeling the rail infrastructure. The distribution of hydrogen cation areas of technology combinations and conducting a well-to-tank,
within a subregion is neither considered in Almansoori und Betancourt as well as a well-to-wheel, analysis. As part of the preprocessing, we
[9] nor Ochoa Biqué und Zondervan [12] nor Moreno-Benito et al. [8]. extend the model by means of spatial resolution to calculate the
Samsatli et al. [10] estimate the distribution and fueling station costs, transport costs of the supply chain based on realistic distances. There-
without considering possible fueling station locations. fore, additional algorithms and models for allocating sources and sinks,
In contrast to studies focusing on a national supply chain setup to as well as designing a pipeline network and the route determination of
design a spatially-resolved network, there exists a second group of trucks, are necessary.
technology-focused analyses that are mostly independent of regional
boundaries. Reuß et al. [18] investigated hydrogen supply chains based 2.1. Hydrogen technologies
on a varying framework to elaborate the application areas of the
technology pathways. The well-to-wheel analyses of the JEC con- The main challenge for hydrogen storage and transport is its ex-
sortium consisting of the Joint Research Center, EUCAR and CONCAWE tremely low density (0.09 kg/m3), in accordance with its being the
[19–21] compared different supply chain options for the supply of lightest element [30,31]. Therefore, the density of hydrogen must be
FCEVs by a broad well-to-tank analysis and shows that low-emission improved if it is to be economical. Compressed gaseous (GH2) and

439
M. Reuß, et al. Applied Energy 247 (2019) 438–453

liquid hydrogen (LH2) are the current state of the art in hydrogen sto- costs. Reuß et al. [42] estimate fueling station investment costs on the
rage options. Nevertheless, many alternative storage methods like hy- basis of previous estimations from Robinius et al. [7] and results from
drogen sorption by metal organic frameworks as well as chemical or the HRSAM [24], as well as own considerations. Nevertheless, Melaina
metal hydrides are addressed by current research activities [32–35]. and Perez [43] claim that, as with other technologies, fueling stations
Amongst these, Reuß et al. [18] state that the liquid organic hydrogen are subjected to learning and scaling effects. Today, there are 64 op-
carriers (LOHCs) offer the most promising alternative technology for erating hydrogen fueling stations in Germany according to H2MOBIL-
logistical needs such as storage and transport. ITY [44]. Most of them are type S stations [45] with one dispenser and
For supplying hydrogen from production to the vehicle, five dif- 212 kg/day fueling capacity. For supplying a meaningful fleet of FCEVs
ferent supply chain steps are necessary, in equivalence to Reuß et al. in the future, the numbers of stations as well as the station capacity
[18]. have to rise significantly. The learning and scaling effects will have an
Production: Electrolysis is considered as hydrogen production immense impact on fueling station costs. In addition, the state of the
technology, as it allows for utilizing surplus electricity from the power supplied hydrogen influences the investment as well as operating costs.
supply system. According to Götz et al. [7], PEM electrolysis is con- Therefore, we conducted a bottom up analysis of hydrogen fueling
sidered the most suitable option for operation together with fluctuating station investment costs based on the approach of the HRSAM [24] for
renewables, even if additional cost reductions, as well as efficiency GH2, LH2, and LOHC fueling station types in order to determine the
improvements, are necessary. Saba et al. [36] investigated the invest- scaling effects, as well as base investment costs for all supply types.
ment costs of the past three decades and estimated that future invest- According to Bonhoff [46], it is expected that the uninstalled fueling
ment costs until 2030 would be between 327 and 787 EUR/kWel. This station costs – equipment only – for GH2 trailer-supplied fueling sta-
study considers 500 EUR/kWel to be linked to an average efficiency of tions with onsite storage and 212 kg/day fueling capacity will drop to
70%, in accordance to Reuß et al. [18] and Robinius et al. [7]. The 600,000 EUR through 2023 if 400 fueling stations are set up in ac-
hydrogen is produced as GH2 at a 30 bar outlet pressure. Further cordance with H2MOBILITY [44]. This value plus 30% installation costs
techno-economic parameters for electrolysis are shown in Table 2. is set as a starting value for learning effects according to the learning
Seasonal storage: Seasonal storage is determined by the contain- equation of Melaina and Perez [43]:
ment system. GH2 can be stored on the small- to mid-scale by storage
vessels of different sizes. Large-scale storage options are offered by C1 C1 n
geological formations like salt caverns or aquifers. According to Reuß I1 = 1, 3 600.000EUR kg
(1 )log2 kg
212 Tag 212 Tag 400
et al. [18], the surface-near storage of GH2 for seasonal storage is not
competitive with the storage in salt caverns, LH2 or LOHC, as the (1)
pressure vessel containment is too expensive. Therefore, only geological
In this study, we assume a capacity of 1000 kg/day for all fueling
formations and especially salt caverns for GH2 storage are considered in
stations. All values are given by Table 7. The fueling stations can be
this study, since they are the most suitable geological storage for hy-
supplied by pipeline, as with GH2, or by trailer, as with GH2, LH2 or
drogen according to the HyUnder Study [37]. The excavation incurs
LOHCs. The scaling factor α varies from 0.6 to 0.75 depending on the
costs with each cavern, while the above-ground infrastructure is a one-
supply mode. The number of fueling stations n is dependent on the
time investment. Therefore, we assume fixed investment costs of 23
demand scenario. The hydrogen is then decanted to an onsite storage
million EUR plus 38.8 million for each cavern of 500,000 m3 volume.
system at the fueling station. Elgowainy et al. [40] propose using the
LH2 requires super-insulated storage vessels, which are assumed to cost
GH2 trailer as onsite storage. This would simplify the fueling station
25 EUR/kgH2, while LOHC tanks are comparable to today’s diesel sto-
design and reduce the compressor’s power demand. Nevertheless, the
rage options, with respective costs of 50 EUR/kgH2 (incl. cost of LOHC),
trailer’s investment must be considered in addition. The learning rate of
in equivalence to Reuß et al. [18]. The storage-related techno-economic
β is set in accordance to Melaina and Perez [43] to 6%. With doubled
parameters are given in Table 3.
installed capacity, the investment for a new station will decrease by 6%.
Transport: This study distinguishes between two transport options:
According to McDonald [47], a learning rate of 6% is a conservative
Pipeline and truck. Pipeline transport requires high capital investment
assumption. Due to the effect of missing information about worldwide
costs at limited flow rates, but with low operational costs. In addition,
fueling station installations, the increase in fueling stations in Germany
the pipeline network is separated into transmission and distribution
is considered to happen in accordance with the worldwide develop-
segments. The pipeline costs are largely determined by the diameter of
ment. All related techno-economic parameters for different fueling
the pipeline, which is again dependent on the hydrogen mass flow and
station types are displayed in Table 7. The fueling station investment
the respective pressure drop. Robinius et al. [38] present a modeling
costs dependent on the total installed fueling stations for each con-
approach for considering pressure drops, as well as the mass flow into a
sidered station are shown in Fig. 1 to clarify the effect of learning ef-
pipeline – a sizing optimization that will be used in this study. The
fects. Obviously, the LOHC supplied fueling station has the highest
pipeline costs are given in Table 4.
investment costs which is caused by the high investment into a dehy-
Truck transport allows for the utilization of GH2, LH2 or LOHC, in
drogenation unit as well as a high-performance compressor to recom-
equivalence to the storage options. The main indicator for comparing
press the hydrogen from ∼2 bar to 950 bar.
different trailer delivery methods is the hydrogen storage capacity of
Conversion: The utilization of different storage methods necessi-
the trailer, which is limited due to its maximum allowable volume and
tates conversion between different states. Regarding the three con-
weight. Current developments for GH2 trailers have achieved net-load
sidered storage methods, GH2, LH2 and LOHCs, five technological
capacities of up to 1100 kgH2, with composite containers of up to
capabilities are necessary, namely: Compression, liquefaction, eva-
500 bar [39,40]. According to Lahnaoui et al. [41], pressure levels of
poration, hydrogenation and dehydrogenation. Detailed discussions
250–350 bar will be the most used option in mid-term to 2030, but the
about these can be found in Reuß et al. [18], Teichmann et al. [26], the
500 bar trailer will be mainly used for increasing demands in 2050. LH2
Nexant Report [48] and Aasadnia and Mehrpooya [49]. Table 8 shows
trailers can carry up to 4300 kgH2 [18]. The capacity for transported
all techno-economic assumptions employed in this study.
hydrogen in an LOHC amounts to roughly 1680 kg, considering a
maximal dehydrogenation rate of 90% [18]. Truck transport-related
techno-economic assumptions are given by Tables 5 and 6. 2.2. Supply chain model
Fueling: The fueling station is the final link in the supply chain,
supplying GH2 at 700 bar to an FCEV tank. The assumptions of the Reuß et al. [18] developed a hydrogen supply chain model that is
characteristics of the fueling stations have a major impact on the final capable of comparing pathways with different technological

440
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 1. 1000 kg/day (L-type) fueling station investment dependent on the total number of fueling stations installed with a learning rate β = 0.06, starting with 400
fueling stations of 2012 kg/day (S-type) in 2023.

combinations in a modular setup. There are five technology modules

based on the introduced application types, namely: production, storage,
transport, fueling and conversion (connector). Each module calculates
its specific hydrogen costs (total expenditures – TOTEX), as well as the
total investment, energy demand and the CO2 emissions caused. A si-
milar approach was used by Linnemann and Steinberger-Wilckens [50]
or Andrea et al. [51]. Necessary input parameters of each module are
the average daily throughput, the TOTEX of the previous module, as
well as the technology database drawn in the Appendix (Tables 2–8).
The transport distances, as well as the storage capacities, are exogenous
inputs for the model to calculate the transport and storage costs. Ad-
ditional information about the technology modules can be drawn from
Reuß et al. [18].
These technology modules are ultimately interconnected to a supply
chain, which is represented in Fig. 2. This study will only consider
hydrogen production by electrolysis, fed with fluctuating excess elec-
tricity in accordance to the results from Robinius et al. [7]. Therefore,
storage as well as a bypass, is necessary to take the fluctuation into
account. In this study, the transport task is separated by transmission
and distribution. Even if both are transport tasks, the technology could
differ. However, the transport technologies are considered to operate
continuously.

2.3. Spatial methodology

The application of the supply chain model developed by Reuß et al. Fig. 2. Supply chain setup in this study.
[18] to a nationwide supply chain network necessitates spatial resolu-
tion to realistically describe transport distances. Therefore, additional vehicle fleet is used as exogenous input. Based on the vehicle fleet
algorithms and models are necessary. First, the sources and sinks must distribution on the county-level (German: “Landkreise”) given by the
be allocated. Second, the transport system must be drawn. Therefore, German Kraftfahrt-Bundesamt [53], the FCEV fleet is assumed to be to
pipeline and truck transport are considered. Since pipeline networks are proportional to the vehicle fleet in each county. With an average
separated in this study into transmission and distribution, the allocation mileage and the specific hydrogen consumption, the hydrogen demand
of the hub, which separates the transmission from distribution, is an per county is calculated. Based on the countywide distribution, the
additional aspect to be investigated. specific locations of fueling stations must be selected. The existing lo-
Distribution of sources and sinks: The hydrogen production cations of gasoline stations from OpenStreetMap [54] are used as can-
sources are based on the results from Robinius et al. [7,52]. They al- didate locations. Considering a target capacity of each fueling station
located the electrolysis locations based on spatially resolved negative and a predefined connection sequence, fueling stations in each county
residual loads that occur due to a highly-renewable dominated power are selected until demand can be satisfied. The connection sequence is
sector scenario supplying the electrical load of 2015. based on the following territorialities: metropolitan area (> 1million
The hydrogen sinks are represented by fueling stations. In ac- inhabitants), urban area (< 1 million inhabitants) and rural area based
cordance to Robinius et al. [7] the share of FCEV on the passenger

441
M. Reuß, et al. Applied Energy 247 (2019) 438–453

on the results from the Global Urbal-Rural Mapping Project (GRUMP) FCEVs and the production potential. The general scope of this analysis
project [55], which offers a global distinction for rural and urban areas. is a hydrogen infrastructure for the year 2050. According to Robinius
Urban fueling stations are selected first, rural stations last. If there are et al. [7], Germany could install 28 GW of electrolysis capacity, to
not enough fueling stations available in a county, an additional fueling produce 3.07 million tons of hydrogen per year out of renewable hy-
station at an already used location will be added to the fueling station. drogen, through to 2050. As was already mentioned, the locations of
The fueling station capacity is set to 1000 kg/day, which corresponds to these electrolysis capacities are viewed as potential production sites.
an L-sized fueling station from H2MOBILITY [45]. Since it is not rea- The electricity price for excess electricity is set to 0.06 EUR/kWhel,
listic that fueling stations are continually operated at full capacity, a which represents the average levelized cost of electricity of onshore
utilization of 70% is assumed. wind energy, that is used in the underlaying scenario and can be con-
Truck routing: To examine the traveled distance of trucks, a truck sidered as rather conservative assumption [7]. In addition, further
routing model is developed. The street grid from DLM1000 (Digitales electricity is necessary for operating the liquefaction unit, for instance.
Landschaftsmodell 1000) [56] is used as a basic graph with nodes and Therefore, we assumed the electricity prices to be analogous to Reuß
edges and representing the highway and other roads. The average truck et al. [18] based on historic values from Eurostat [61] without taxes
speed is set to 60 km/h on highways and 30 km/h on all other roads. and levies for different annual consumption, but set the excess elec-
Each fueling station is connected by the Euclidian distance to the closest tricity price of renewable energy equal to the value of the highest
street node. For each edge, the travelled distance and time is calculated. consumption band. The resulting electricity prices are shown in
Based on this information, a linear flow optimization is conducted, Table 9.
minimizing the used time. Next to electricity, some infrastructure technologies consume nat-
Pipeline system design: In this work, the methodology to establish ural gas or diesel. According to the Long Term Scenarios of the German
the pipeline system design is separated into four steps: Candidate grid Government [62], the natural gas price will rise to 0.046 EUR2010/
development, topology selection, mass flow determination and dia- kWhNG through to 2050, while the diesel price will rise to 1.73 EUR/
meter selection, as well as cost calculations. lDiesel (without taxes). The associated greenhouse gas emissions, as well
The methodology used to develop a candidate grid is derived from as the primary energy demand of all energy carriers used, are shown in
the work of Baufumé et al. [5], wherein the existing lines of the high Table 10. The weighted average cost of capital (WACC) is set to 8%.
pressure natural gas grid are used as potential routes for the hydrogen Regarding the FCEV population, three different penetration rates
pipeline grid. In addition, distance matrices based on Euclidian dis- are investigated in this study: 25%, 50% and 75%. The average mileage
tances between the source, sinks and natural gas grid are implemented of each FCEV is set to 14,000 km, which represents the average in-
to allow pipeline routing in sparsely-populated areas. Thus, for these dividual vehicle mileage in Germany in 2017 according to the
Euclidian distances a detour factor of 1.4 in accordance to Baufumé Kraftfahrt-Bundesamt [63]. The specific hydrogen consumption of
et al. [5] is added to address additional costs by exploiting the area. FCEVs is set to 0.63 kg/100 km, which represents a future driving
The topology selection is then performed by applying a minimum profile of a C-segment car with an MVEG driving cycle according to
spanning tree [57] on the candidate grid. The minimum spanning tree Grube and Stolten [64]. As a comparison: the JEC well to tank report
forces the topology to use existing lines as far as possible due to the high [20] expected in 2014 a value of 0.58 kg/100 km for the year 2020 and
resolution of the existing lines. On the resulting tree, a linear flow following. A recent study from Cambridge Econometrics [65] assumed
optimization similar to the truck routing is conducted to get the re- a drop from today’s roughly 0.85 kg/100 km for a medium sized FCEV
sulting flow on each pipeline section. Afterwards, the arc-sizing opti- to roughly 0.6 kg/100 km based on a 10% drop per decade and with
mization developed by Robinius et al. [58] determines the diameters, regards to real world driving consumption.
costs and pressure drop of each pipeline section. The analyzed pathways are shown in Fig. 3. All in all, nine different
Determination of hubs: The pipeline network is often separated into pathways are analyzed. They are defined by the storage, transmission
a transmission and distribution pipeline [5,23,59], whereas the transmis- and distribution technology.
sion has the task of transport across long distances, while the distribution The storage capacity of the storage module is set to 60 days, in ac-
is used for local distribution to single consumers/fueling stations. On a cordance to Reuß et al. [18] and the worst case of Robinius et al. [7]. It
nationwide scale, the placement of the hub is important, because it defines is assumed that it is possible to set up the seasonal storage facility close
the ratio between the pipeline lengths of transmission to the distribution. to the production location, so that after storage, a continuous flow can
Baufumé et al. [5] place the hubs into the centroids of counties in- be established.
dependent of the total hydrogen demand. Seydel uses NUTS11-Regions to
transmission and distribution. By doing so, they are using political borders 3. Results and discussion
to separate the transport tasks. In this work, we determine the hub position
by clustering the fueling station locations using a k-means clustering. As This section shows the results of the analysis. Three scenarios are
such, we are independent of political boundaries and more freely to ex- analyzed: 25%, 50% and 75% penetration of FCEVs. First, we will show
amine the influence of the hub allocation. the resulting fueling station distribution to provide an overview,
wherein the demand of hydrogen occurs. Second, the allocation of hubs
to separate transmission from distribution is further analyzed before, in
2.4. Scenario assumptions a third step, the spatial infrastructure designs for the pipeline system,
the truck routing and the combination of a pipeline transmission and
Besides the techno-economic details given in the Appendix, some truck distribution are drawn. Finally, a techno-economic analysis of the
scenario assumptions are necessary to serve as a framework for the different supply chain pathways will be presented.
whole analysis, which will determine energy prices, the amount of
3.1. Fueling station distribution
1
NUTS: Nomenclature des unites territoriales statistiques (Nomenclature of
The distribution and selection of fueling stations are important in-
Territorial Units for Statistics) is a geocode standard dividing countries into
subregions inside the European Union [60] Europäische Union (EU). Ver- puts, as they represent the hydrogen sinks for the transport system.
odnung (EG) Nr 1059/2003 des europäischen Parlaments und des Rates vom Fig. 4 shows fueling station locations, with the territorialities for the
26. Mai 2003 über die Schaffung einer gemeinsamen Klassifikation der Ge- three investigated FCEV share scenarios. Obviously, the number of
bietseinheiten für die Statistik (NUTS). In: Parlament E, editor. 1059/ fueling stations rises with an increasing share of FCEVs. In the 75%
20032003. scenario, 12,063 fueling stations are installed, fueling 3.03 million tons

442
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 3. Investigated pathways for hydrogen supply chain architectures.

Fig. 4. Spatial distribution of fueling station locations in three investigated scenarios for 25%, 50% and 75% fleet penetration of FCEVs.

of hydrogen per year. This is in line with the used scenario from Ro- number of fueling stations, which are clustered by a k-means cluster
binius et al. [7] who presented a potential hydrogen production of 3.07 algorithm from scikit-learn [66], which is a python module for machine
million tons of hydrogen per year. The fueling station density is highly learning, to one cluster center. For comparative reasons, the allocation
distributed. Especially in the metropolitan areas of Rhein-Ruhr and of the hub by political borders like counties, as was done by Baufumé
Rhein-Main, as well as in big cities like Munich, Berlin and Hamburg, et al. [5], is also performed here.
the density is very high with a peak of 0.6 fueling stations per km2 in Fig. 5 shows the resulting transport costs for a transmission and
Munich in the 75% scenario. For the same scenario, the fueling station distribution by pipeline for varying cluster sizes. Depending on the
density in the Uckermark county, which is located in the north-east of scenario, the cumulative cost of transmission and distribution does not
Germany, is at just 0.0058 fueling station per km2. The average utili- show major variation in each scenario. This is caused by a redistribu-
zation of each fueling station rises from 665.5 kg/d in the 25% scenario tion of costs from transmission to distribution: with an increasing
to 688 kg/d in the 75% scenario. number of fueling stations per cluster, the specific costs of transmission
decrease, while the distribution costs are rising. Nevertheless, in all
scenarios the zoning by political borders lead to higher costs compared
3.2. Clustering to the clustered hub placement. However, it must be kept in mind that
the transmission pipeline is routed through the existing natural gas
The allocation of hubs is carried out in this study by a predefined

443
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 5. Specific transport cost for a transmission and distribution pipeline dependent on varying cluster sizes for 25%, 50% and 75% fleet shares of FCEVs. The
location of each hub or cluster center is either calculated by k-means clustering or based on political borders of the counties.

grid, while the distribution pipeline uses Euclidian distances with de- storage)
tour factors. The costs of the transmission pipeline are consequently • Pipeline transmission and GH2 trailer distribution (with GH2 cavern
more reliable than the distribution costs. or LOHC storage)
In contrast, Fig. 6 depicts the results for a transmission by pipeline • Trailer transmission to the fueling station as GH2, LH2 or LOHC
and distribution by GH2 trailer. Clearly, the cluster size now has a major (with GH2 cavern, LH2 or LOHC storage)
influence on the cumulative transport costs and a cost minimum occurs
for cluster sizes of about 100 fueling stations per cluster in each sce- Fig. 7 presents the results of pipeline transmission and distribution
nario. According to an average demand of each fueling station of for the three scenarios of 25%, 50% and 75%. Due to the location of
700 kg/day, the beneficial throughput of the hub is 70 t/day. Smaller electrolysis plants- according to Robinius et al. [7] – in the North of
cluster sizes lead to higher costs due to the high costs of the transmis- Germany, the transmission pipeline has one large pipeline section that
sion pipeline, while keeping a fixed cost value for the trailer filling supplies the hydrogen to the demand centers in the south of the
station, as well as the trailer fixed costs. High cluster sizes result in low country, and such cities as Munich and Stuttgart. This backbone pipe-
costs for the transmission pipeline, but at the same time the travelled line features in all three scenarios. With an increasing share of FCEVs,
distances for the trailer increase substantially. Comparing the results of the transmission pipeline length increases from 4936 km in the 25%
the political borders with those of the clustering approach, it is clear scenario to 8825 km in the 75% scenario, this accounts for an increase
that, especially for a lower penetration rate of 25%, the clustering ap- of 79%. Meanwhile, the distribution pipeline distance rises by 50%,
proach decreases transport costs from 1.17 EUR/kg to 0.95 EUR/kg from 39,439 km to 59,367 km. Keeping in mind that the number of
which constitutes to savings of 19%. fueling stations increases from 4158 to 12,063, the average specific
On the basis of the results of the variation of cluster sizes, the fur- distribution pipeline length per fueling station decreases from 9.5 km to
ther analyses employ a cluster size of 100 fueling stations since that 4.92 km. This will have a major effect on the specific costs of the pi-
represents the most suitable amount for a combination of pipeline and peline distribution. The investment costs for the transmission pipeline
trailer. For pipeline transmission and distribution, Fig. 4 shows just rise from 2.5 billion EUR to 5.12 billion EUR. In turn, the distribution
small differences and as explained above, a higher number of fueling costs rise from 13.7 billion EUR to 20.7 billion EUR.
stations per cluster would result in less reliability since the distribution The infrastructure design of a transmission pipeline, together with a
grid is based on Euclidian distances with detour factors. Therefore, the GH2 trailer distribution, is drawn in Fig. 8. The transmission pipeline is
same cluster size is chosen as for the pipeline trailer combination. analogous to that in Fig. 7 and has up to 8825 km for 5.12 billion EUR
in the 75% scenario. For the distribution of hydrogen to the fueling
3.3. Spatial design of hydrogen supply networks stations, 2532 trucks are necessary in the 25% scenario and 6480 trucks
in the 75% scenario. The number of trucks is not linear to the demand
In this subsection, the spatial design of the hydrogen transport in- due to the decreasing average transport distances of the truck dis-
frastructure is presented and analyzed. In total, three different transport tribution. While the average travelled distance from the hub to the
designs based on the 9 pathways outlined in Fig. 3 arise: fueling station amounts to 42.2 km in the 25% scenario, this distance
decreases to 27.5 km in the 75% scenario due to the higher number of
• Pipeline transmission and distribution (with GH 2 cavern or LOHC clusters. Compared to today’s average transport distances of mineral oil

Fig. 6. Specific transport cost for transmission by pipeline and distribution by GH2 trailer dependent on of varying cluster sizes for 25%, 50% and 75% fleet shares of
FCEVs. The location of each hub or cluster center is either calculated by k-means clustering of fueling stations or based on political borders of the counties.

444
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 7. Infrastructure design of pipeline transmission and distribution in the three investigated scenarios for 25%, 50% and 75% fleet penetration of FCEVs.

products of 95 km [67], this is rather low. GH2 trailer distribution from Fig. 8 has a peak load of 217 trucks per
Finally, Fig. 9 shows the routing of truck transport from the elec- day.
trolysis/storage location to the fueling station. Like the transmission With these results, it is already apparent without checking the
pipeline, there are several main routes from the North to the south. The specific costs that a truck-based supply of hydrogen in the form of GH2
average travelled distance (one-way) increases from 433 km to 440 km, or LOHC for the considered distances is hardly feasible and would have
which means that the trailer is almost unaffected by additional demand. a high impact on the road network.
Nevertheless, comparing this value with today’s average mineral oil
product transport distance of 95 km [67], the travelled distance is en- 3.4. A techno-economic comparison between infrastructure options
ormous. While a pipeline would be explicitly for hydrogen transport,
the trucks would drive on existing road infrastructure and influence The techno-economic results of the costs are shown in Fig. 10. The
this. Depending on the trailer capacity of LH2, GH2 and LOHC, the cheapest system in the 25% scenario appears to be a combination of salt
annual transport capacity leads to a different number of trucks that cavern storage, pipeline transmission and GH2 trailer distribution with
must drive on specific road sections. To address the effect this would specific costs of 6.40 EUR/kg. With increasing demand, these costs
have, we calculated the number of trucks that must drive on one road decrease to 6.19 EUR/kg in the 50% scenario and 6.10 EUR/kg in the
section per day. The largest effect is caused by a GH2 trailer in the 75% 75% scenario. The second best option in the 25% scenario is a pipeline
scenario due to the lowest payload of 1100 kg. The peak load amounts distribution instead of the GH2 trailers with 6.96 EUR/kg. With rising
to 2017 trucks per day. Wietschel et al. [68] calculate 3200 heavy duty hydrogen demand, this represents the cheapest solution, with 6.05
trucks (> 12 t) per day as a maximum truck load occurring through EUR/kg in the 75% scenario. The third option, which remains compe-
2030 on German highways and caused by all transported goods. In titive with costs of 6.98 EUR/kg in the 25% scenario, is salt cavern
comparison to that, an additional load of 2017 trucks per day caused storage with subsequent liquefaction and LH2 trailer transport. In all
solely by supplying hydrogen for FCEVs is no longer reasonable. With scenarios, this induces approximately 10% additional costs compared to
an increasing capacity, the load is decreasing, but even LOHC trailers the best option. All other solutions, like GH2 or LOHC trailers, are not
still imply 1300 trucks per day, which is high compared to the 3200 competitive with additional costs of more than 20% compared to the
trucks for all transported goods. LH2 trailers alone would entail 516 cheapest option.
trucks per day – a reasonable value that is feasible, but still high. With a The cheapest solution without geological storage is LH2 storage with
decreasing fleet size of FCEVs, the truck load will decrease as well. an LH2 trailer supply. This would lead to additional costs of 16–18%
Nevertheless, the peak truck load of the 25% scenario for GH2 trailers is compared to salt cavern storage pathways, which confirms that geological
still 913 trucks per day. For comparison, the additional truck load of storage options will be a key element in future hydrogen supply chains.

Fig. 8. Infrastructure design of pipeline transmission and GH2 trailer distribution in the three investigated scenarios for 25%, 50% and 75% fleet penetration of
FCEVs.

445
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 9. Infrastructure design of trailer transmission to the fueling station in three investigated scenarios for 25%, 50% and 75% fleet penetration of FCEVs.

To better address the development of costs with rising hydrogen the additional compression after dehydrogenation of 133–159 MJ/km.
demand, Fig. 11 shows the diagram of nozzle costs for the three To further classify the results, Fig. 13 compares the findings of this
cheapest pathways with salt cavern storage, pipeline transmission and study with the results of the JEC well-to-wheel analysis [19] for FCEVs
distribution, pipeline transmission with GH2 trailer distribution and supplied with hydrogen from wind-powered electrolysis (WDEL1-CH2)
LH2 trailer transport. The distribution pipeline profits the most from and conventional ICE vehicles fueled with gasoline or diesel in a time-
rising FCEV fleet. The main reason for this lies in the reduction of frame of 2020+. The primary energy demand for all investigated path-
distribution pipelines per fueling station, which was noted in Fig. 7. ways in this study is higher compared to the WDEL1-CH2 pathways in
Nevertheless, the combination of pipeline transmission and GH2 trailers the JEC. This is due to different hydrogen consumption values: the JEC
appears to be the cheapest solution for an FCEV fleet of 65–70%. The assumes 0.48 kg/100 km while this study assumes 0.63 kg/100 km.
LH2 solution has similar scaling effects as the pipeline/trailer combi- Meanwhile, the CO2 emissions are lower for some gaseous as well as
nation, but imposes additional costs of around 0.60–0.70 EUR/kg. liquid pathways compared to the JEC value. Only the LOHC pathway
Based on these results, the two key elements of a future supply chain causes clearly higher emissions. Nevertheless, comparing the CO2 emis-
are geological formations for hydrogen storage like salt caverns, and sions of all pathways with the CO2 emissions from conventional com-
transmission pipeline systems for long-distance transport to a hub. bustion engines, all FCEV pathways are significantly lower than ICE.
These two elements are the main technologies, even at lower fleet
shares of FCEVs of 25%. The distribution mode from this hub finally 3.5. Sensitivity analysis
changes with rising demand. While the distribution by GH2 trailer is
cheaper than the pipeline for 25% and 50% fleet penetration, this To give an impression of the relevant input assumptions, a sensi-
changes in the 75% scenario. tivity analysis across all input parameters was conducted on the 50%
As well as the costs, the hydrogen supply chain model calculates the scenario as well as its most cost-efficient pathway, namely salt caverns,
well-to-wheel CO2 emissions as well as the primary energy demand. The pipeline transmission and GH2 trailer distribution. To better understand
different FCEV fleets exhibit no influence on the final results. Therefore, the results of the sensitivity analysis, the detailed results of that
Fig. 12 only shows the results of the 75% scenario. The lowest emissions pathway are given by Table 1.
are produced by LH2 storage option with LH2 trailer supply at 1.6 gCO2/ The sensitivity analysis varied each input parameter from Tables 2
km, followed by the salt cavern storage with pipeline transmission and to 9 by +20%/−20%. The results are shown in Fig. 14. By far the most
either distribution by pipeline (1.7 gCO2/km) or GH2 trailer (2.5 gCO2/ important assumption is the electricity costs for the electrolysis op-
km). As was already discussed by Reuß et al. [18], a flexible liquefac- eration. The electrolysis itself causes 3.70 EUR/kg, whereas 2.896 EUR/
tion powered by excess renewable power could lower emissions. kg is assigned to the electricity costs. This is by far the largest individual
Nevertheless, the flexible operation of such a plant would also need to cost parameter. Due to the high assumed operating hours of the elec-
be proven in the future, as the part-load behavior must be further ex- trolysis, the investment costs of the electrolysis plant are less important
amined. Today’s plants are operating continuously to reduce capital than the electricity costs. The WACC has the second highest impact,
expenditures. With continuous operation, the electricity demand would since the WACC is used for all investment costs that arise. While the
be supplied by the electricity grid and an alternative storage option electricity costs affect one individual parameter substantially, the
would be necessary. This is presented by salt cavern storage with LH2 WACC increases the costs across all supply chain steps. As shown in
trailer transport. Utilizing grid electricity for the liquefaction process Table 1, the total CAPEX accounts to 2.24 EUR/kg, which is affected by
would increase CO2 emissions to 7.0 gCO2/km. In accordance to the the WACC. The second highest individual cost contribution is the
findings of Reuß et al. [18], LOHC transport causes by far the highest fueling station. The scaling factor, as well as the base investment costs,
CO2 emissions due to the burning of natural gas at the fueling station to of the fueling station influence this individual parameter. As a last
supply heat for the dehydrogenation. Reuß et al. [18] investigated al- important parameter, the trailer investment costs were identified due to
ready alternative options for heat supply, but finally, the increase in the fact that the trailer imposes costs at the distribution level, as well as
energy demand from the dehydrogenation in combination with com- at the fueling stations.
pression will remain an issue for fully decarbonize the LOHC pathways.
Regarding primary energy demand, we can confirm the conclusions
of Reuß et al. [18]: GH2 transport options require the least energy, with 3.6. Optimization potential
117–125 MJ/km, followed by LH2 with 133–139 MJ/km. LOHCs have
the highest energy demand due to the dehydrogenation heat, as well as As was already stated, the transmission pipeline and salt cavern
storage represent key elements of future hydrogen supply systems for

446
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 10. Specific hydrogen costs at the fueling station dependent on of the investigated scenarios for 25%, 50% and 75% fleet penetration of FCEVs.

Fig. 11. Specific hydrogen costs at the fueling

stations dependent on of the investigated sce-
narios for 25%, 50% and 75% fleet penetration
of FCEVs for three pathways: Cavern storage,
pipeline transmission and GH2 trailer distribu-
tion; cavern storage, pipeline transmission and
distribution; cavern storage and LH2 trailer
transport.

447
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 12. Specific CO2 emissions and primary energy demand for 75% fleet penetration of FCEVs.

transport. In contrast, the distribution mode from a hub to the fueling trailer. In particular, the big demand centers of Munich, Berlin and
station varies with increasing fueling station density and FCEVs, re- Rhein Ruhr area are now supplied by pipeline. The 75% scenario now
spectively. However, this fueling station density not only varies from already has 57 of 121 hubs supplied by pipeline. Now, only sparsely
scenario to scenario, but also from hub to hub. Therefore, an analysis of populated areas like Mecklenburg in western Pomerania or lower
the distribution costs for each hub is conducted to estimate an opti- Franconia are still supplied by GH2 trailers.
mization potential between distribution by pipeline or trailer. The re- Regarding the overall cost optimization potential, Fig. 16 shows the
sults of this analysis are shown in Fig. 15. With respect to the 25% cost of different distribution types in comparison to an optimized dis-
scenario, distribution by trailer is cheaper for all 42 hubs. With an in- tribution. While the 25% scenario does not have additional optimiza-
creasing number of FCEVs, the 50% scenario already shows that 23 tion potential, since all fueling stations are supplied by trailer, the 50%
hubs of 81 are cheaper when supplied by a distribution pipeline than by scenario already shows a cost decrease in distribution and fueling

Fig. 13. Specific CO2 emissions and primary energy demand for all pathways in comparison to results from JEC well-to-wheel analysis [19].

448
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Table 1
Detailed cost analysis of the most cost-efficient pathway for the 50% scenario, namely salt caverns, pipeline transmission and GH2 trailer distribution.
Technology TOTEX CAPEX fixOPEX varOPEX Investment
[EUR/kg] [EUR/kg] [EUR/kg] [EUR/kg] [billion EUR]

Production Electrolyzer 3.700 0.670 0.135 2.896 9.069

Connector1 None 0.000 0.000 0.000 0.000 0.000
Connector2 Compressor 0.034 0.009 0.003 0.022 0.162
Storage GH2-Cavern 0.193 0.157 0.035 0.000 3.574
Connector3 None 0.000 0.000 0.000 0.000 0.000
Connector4 Compressor 0.041 0.004 0.001 0.036 0.070
Transmission Pipeline 0.168 0.152 0.016 0.000 3.650
Connector5 Compressor 0.130 0.027 0.009 0.093 0.473
Distribution GH2-Truck 0.502 0.263 0.203 0.037 3.765
Station GH2 (Trailer) 1.419 0.957 0.256 0.207 13.546

Total costs 6.187 2.238 0.659 3.290 34.308

Fig. 14. Sensitivity analysis of hydrogen infrastructure based on salt cavern storage, pipeline transmission and GH2 trailer distribution for 50% FCEV fleet share.

Fig. 15. Optimal selection of distribution types from hub to fueling station per cluster for 25%, 50% and 75% fleet penetration of FCEVs.

station costs of 4%, which amounts to 0.09 EUR/kg. The 75% scenario model to design and analyze hydrogen supply chains based on different
offers the highest cost decrease in the optimized case of 0.12 EUR/kg, technologies for storage and transport was developed.
or 6%. Since the costs of pipeline distribution are now already cheaper The most beneficial hydrogen supply chain in terms of the three
than truck distribution, the cost potential for even higher penetration investigated scenarios was salt cavern storage, in combination with
rates will decrease due to the fact that a full pipeline distribution will no pipeline transmission and GH2 trailer distribution. This setup offers the
longer offer any optimization potential. best compromise for hydrogen supply with the lowest costs at low CO2
emissions and primary energy demand. In addition, the fueling stations
have almost the same setup as for today’s fueling stations, which sim-
4. Conclusion
plifies the transition to a future hydrogen infrastructure by installing
additional transmission pipelines on demand. Distribution by pipeline is
In this investigation, the aim was to assess hydrogen infrastructure
also a suitable option, but only relevant for high hydrogen demand
alternatives with respect to the resulting cost, CO2 emissions and pri-
densities. This is supported by varying the distribution mode per
mary energy requirements of hydrogen provision. For this purpose a

449
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Fig. 16. Resulting distribution and fueling station costs for an optimal distribution type from hub to fueling station per cluster in comparison to only truck and only
pipeline distribution for 25%, 50% and 75% fleet penetration of FCEVs.

clustered area. The pipeline distribution in metropolitan areas benefits In addition, the average distance of 440 km is linked to the production
of scaling effects with an increasing number of fueling stations to supply scenario in this study that is dominated by electrolysis plants located in
while the costs of GH2 trucks offer low-cost supply for sparsely popu- the North of Germany. Demand-oriented electrolysis or the expansion
lated areas. of the analysis to different countries would lower the distances between
These findings support the step-by-step transition of today’s gaseous production and demand. This influences the technology choice as well
truck supply to a future hydrogen transmission grid with location and and will be part of future analyses.
demand dependent distribution by truck or pipeline. With regard to the total cost of the hydrogen supply, the electricity
The spatial infrastructure design for trucks shows that GH2 and costs of the electrolysis as well as the fueling station investment costs
LOHC trailers have a high impact on the existing road network if con- are the main issue to solve in future supply chains. A significant cost
sidered for long-distance transport. Comparing the resulting average decrease necessitates cheap renewable energy technologies as well as
truck distances of 440 km with conventional mineral oil transports of innovative fueling station design. Storage and transport have mean-
95 km, the discrepancy is obvious. The utilization of a transmission while costs of 1.00–1.20 EUR/kg, which represents a share of less than
pipeline is a suitable solution to cut down the distance to 27 km for 75% 16% to the total supply chain costs.
fleet share of FCEVs. GH2 trailers for the distribution from a distribution
center/hub to a fueling station have consequently much less of an effect
on the road network compared to a trailer transport without pipeline Acknowledgement
transmission. However, other options for long-distance transport like
ship or rail transport could have a significant impact for future supply This work was supported by the Helmholtz Association under the
chains as well by utilizing existing infrastructure, even if they are Joint Initiative “EnergySystem 2050 – A Contribution of the Research
heavily restricted to the rail network and the accessibility to waterways. Field Energy”.

Appendix

See Tables 2–10.

Table 2
Production assumptions based on Reuß et al. [18] and Robinius et al. [7].
System Outlet pressure Investment costs Depreciation period O&M Electricity consumption Water consumption
bar EUR/kWel a kWhel/kgH2 m3/kgH2

Electrolyzer 30 500 10 3% 47.6 0.01

Table 3
Storage assumptions based on a variety of references [18,25,26,69].
System Pressures Investment costs Depreciation period Losses O&M
bar EUR years 1/d 1/year

GH2-Cavern 58–175 23Mio. EUR + 38.8Mio . EUR

x 30 0% 2%
cavern
LH2-Tank 1 25EUR
x 20 0.03%
kg
LOHC-Tank 1 50EUR
x 20 0%
kg

Table 4
Pipeline assumptions based on Mischner et al. [70].
System Investment costs Maximal pressure Minimal pressure Depreciation period O&M
EUR/m bar bar a EUR/(m * a)

Pipeline 0.0016 d 100 70 40 5

292.152 e mm

450
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Table 5
Tractor assumptions based on Teichmann et al. [26]. Fuel consumption based on developments from Shell [71].
System Investment Depreciation period Utilization O&M Diesel demand
EUR a h/a 1/a lDiesel/100 km

Truck 160,000 8 2000 12% 27.6

Table 6
Trailer assumptions based on Teichmann et al. [26] and Stolzenburg et al. [69].
System Pressure range Investment costs Depreciation period Utilization O&M Payload (total capacity) Loading time
bar EUR a h kg h

GH2-Trailer 15–500 660,000 EUR 12 2000 2% 1100 (1200) 1,5

LH2-Trailer 1 860,000 EUR 12 2000 2% 4300 (4500) 3
LOHC-Trailer 1 150,000 EUR 12 2000 2% 1620 (1800) 1,5

Table 7
Fueling station assumptions. The chosen capacity was set to 1000 kg/day. Own calculations based on the HRSAM [24], Melaina and Penev [43], Bonhoff [46] and
Teichmann et al. [26].
System α γ Depreciation time O&M Energy demand Losses β

Electricity Natural gas

a 1/a kWhel/kgH2 kWhNG/kgH2

GH2 (Pipeline) 0.7 0.8 10 5% 2 0 0.5% 0.06

GH2 (Storage) 0.75 1 10 5% 2 0 0.5%
GH2 (Trailer) 0.7 0.6 10 5% 1.6 0 0.5%
LH2 0.6 0.9 10 5% 0.6 0 3%
LOHC 0.66 1.4 10 5% 4.4 11.7 0.5%

Table 8
Conversion assumptions in accordance to Reuß et al. [18]. Liquefaction O&M adjusted from Stolzenburg [72].
System Investment costs Deprec. period O&M Energy demand/supply Loss

Electricity Natural gas Heat

EUR a 1/a kWh el kWhNG kWhth
kgH2 kgH2 kgH2

Liquefaction 0,66 20 4% 6.78 0 0 1.65%

x
105 million EUR t
50 H 2
d
Hydrogenation 0,66 20 3% 0.37 0 –8.8 1%
x
40 million EUR t
300
d
Dehydrogenation 0,66 20 3% 0.37 11.7 0 1%
x
30 million EUR tH 2
300
d
Compressor 15, 000
EUR
x 0,6089 finst 15 4% Calculated 0 0 0.5%
kW
Evaporator 3000 EUR
x 10 3% 0.6 0 0 0
tH 2
d
LH2 pump 30, 000 EUR
x 10 3% 0.1 0 0 0
tH 2
d
LOHC pump 500 EUR
x 10 3% 0.1 0 0 0
tH 2
d

Table 9
Comparison of historical electricity prices from Eurostat 2017 [61] and assumed electricity costs for grid
electricity.
Annual demand Eurostat 2017 [61] Assumption 2050

MWhel EUR/kWhel EUR/kWhel

< 20 0.12065 0.1362
20–500 0.0987 0.11425
500–2000 0.07735 0.0929
2000–20,000 0.06365 0.0792
20,000–70,000 0.0476 0.06315
> 70,000 0.04445 0.06

451
M. Reuß, et al. Applied Energy 247 (2019) 438–453

Table 10
CO2 emissions and primary energy demand of energy carriers assumed. Excess electricity: renewable-
based, grid electricity: based on Schlesinger et al. [73] and prognos [74], diesel and natural gas according
to JEC [20].
CO2 emissions Primary energy demand
gCO2/kWh MJ/kWh

Excess electricity 0 3.6

Grid electricity 122 5.04
Diesel 318 4.32
Natural gas 238 3.96

References [22] Ogden J, Nicholas M. Analysis of a “cluster” strategy for introducing hydrogen
vehicles in Southern California. Energy Pol 2011;39:1923–38.
[23] Argonne (Argonne National Laboratory). Hydrogen Delivery Scenario Analysis
[1] Bolat P, Thiel C. Hydrogen supply chain architecture for bottom-up energy systems Model (HDSAM) V3.0. Department of Energy USA; 2016.
models. Part 1: Developing pathways. Int J Hydrogen Energy 2014;39:8881–97. [24] Argonne (Argonne National Laboratory). Hydrogen Refueling Station Analysis
[2] Bolat P, Thiel C. Hydrogen supply chain architecture for bottom-up energy systems Model (HRSAM) V1.1. Argonne National Laboratory; 2016.
models. Part 2: Techno-economic inputs for hydrogen production pathways. Int J [25] Teichmann D. Konzeption und Bewertung einer nachhaltigen Energieversorgung
Hydrogen Energy 2014;39:8898–925. auf Basis flüssiger Wasserstoffträger (LOHC) [Doctoral dissertation]. Aachen:
[3] Seydel P. Entwicklung und Bewertung einer langfristigen regionalen Strategie zum Friedrich Alexander Universität Erlangen-Nürnberg; 2015.
Aufbau einer Wasserstoffinfrastruktur – aus Basis der Modellverknüpfung eines [26] Teichmann D, Arlt W, Wasserscheid P. Liquid Organic Hydrogen Carriers as an
Geografischen Informationssystems und eines Energiesystemmodells [PhD Thesis]. efficient vector for the transport and storage of renewable energy. Int J Hydrogen
Zürich: ETH Zürich; 2008. Energy 2012;37:18118–32.
[4] Krieg D. Konzept und Kosten eines Pipelinesystems zur Versorgung des deutschen [27] Teichmann D, Arlt W, Wasserscheid P, Freymann R. A future energy supply based
Straßenverkehrs mit Wasserstoff [Doctoral dissertation]. Jülich: Forschungszentrum on Liquid Organic Hydrogen Carriers (LOHC). Energy Environ Sci 2011;4:2767.
Jülich, Zentralbibliothek; 2012. [28] Moroz S, Tan XF, Pierce J, Greaves M, Duguid A, Dumur K, et al. Systems based on
[5] Baufumé S, Grüger F, Grube T, Krieg D, Linssen J, Weber M, et al. GIS-based sce- hypo-eutectic Mg–Mg2Ni alloys for medium to large scale hydrogen storage and
nario calculations for a nationwide German hydrogen pipeline infrastructure. Int J delivery. J Alloy Compd 2013;580:S329–32.
Hydrogen Energy 2013;38:3813–29. [29] Wulf C, Reuß M, Grube T, Zapp P, Robinius M, Hake J-F, et al. Life Cycle
[6] Robinius M, Otto A, Heuser P, Welder L, Syranidis K, Ryberg D, et al. Linking the Assessment of hydrogen transport and distribution options. J Cleaner Prod
power and transport sectors—Part 1: The principle of sector coupling. Energies 2018;199:431–43.
2017;10:956. [30] Klell M. Storage of hydrogen in the pure form. Handbook of hydrogen storage.
[7] Robinius M, Otto A, Syranidis K, Ryberg DS, Heuser P, Welder L, et al. Linking the Wiley-VCH Verlag GmbH & Co. KGaA; 2010. p. 1–37.
power and transport sectors—Part 2: Modelling a sector coupling scenario for [31] Di Profio P, Arca S, Rossi F, Filipponi M. Comparison of hydrogen hydrates with
Germany. Energies 2017;10:957. existing hydrogen storage technologies: energetic and economic evaluations. Int J
[8] Moreno-Benito M, Agnolucci P, Papageorgiou LG. Towards a sustainable hydrogen Hydrogen Energy 2009;34:9173–80.
economy: optimisation-based framework for hydrogen infrastructure development. [32] Zhang Y-H, Jia Z-C, Yuan Z-M, Yang T, Qi Y, Zhao D-l. Development and application
Comput Chem Eng 2017;102:110–27. of hydrogen storage. J Iron Steel Res Int 2015;22:757–70.
[9] Almansoori A, Betancourt-Torcat A. Design of optimization model for a hydrogen [33] Durbin DJ, Malardier-Jugroot C. Review of hydrogen storage techniques for on
supply chain under emission constraints – a case study of Germany. Energy board vehicle applications. Int J Hydrogen Energy 2013;38:14595–617.
2016;111:414–29. [34] Niaz S, Manzoor T, Pandith AH. Hydrogen storage: materials, methods and per-
[10] Samsatli S, Staffell I, Samsatli NJ. Optimal design and operation of integrated wind- spectives. Renew Sustain Energy Rev 2015;50:457–69.
hydrogen-electricity networks for decarbonising the domestic transport sector in [35] Fakiolu E. A review of hydrogen storage systems based on boron and its compounds.
Great Britain. Int J Hydrogen Energy 2016;41:447–75. Int J Hydrogen Energy 2004;29:1371–6.
[11] Welder L, Ryberg DS, Kotzur L, Grube T, Robinius M, Stolten D. Spatio-temporal [36] Saba SM, Müller M, Robinius M, Stolten D. The investment costs of electrolysis – a
optimization of a future energy system for power-to-hydrogen applications in comparison of cost studies from the past 30 years. Int J Hydrogen Energy
Germany. Energy 2018. 2018;43:1209–23.
[12] Ochoa Bique A, Zondervan E. An outlook towards hydrogen supply chain networks [37] LBST, cenex, eon GasStorage, CEA, DEEP, ECN, et al. Assessment of the potential,
in 2050—design of novel fuel infrastructures in Germany. Chem Eng Res Des the actors and relevant business cases for large scale and long term storage of re-
2018;134:90–103. newable electricity by hydrogen underground storage in Europe. FCH JU; 2014.
[13] Yáñez M, Ortiz A, Brunaud B, Grossmann IE, Ortiz I. Contribution of upcycling Report available at < http://hyunder.eu/wp-content/uploads/2016/01/D8.1_
surplus hydrogen to design a sustainable supply chain: the case study of Northern HyUnder-Executive-Summary.pdf > .
Spain. Appl Energy 2018;231:777–87. [38] Schmidt M, Stolten D, Thürauf J, Welder L. Robust optimal discrete arc sizing for
[14] De-León Almaraz S, Azzaro-Pantel C, Montastruc L, Domenech S. Hydrogen supply tree-shaped potential networks. Comput Optim Appl 2019:1–29.
chain optimization for deployment scenarios in the Midi-Pyrénées region, France. [39] Baldwin D. Development of High pressure hydrogen storage tank for storage and
Int J Hydrogen Energy 2014;39:11831–45. gaseous truck delivery. DOE hydrogen and fuel cells programm – annual progress
[15] De-León Almaraz S, Azzaro-Pantel C, Montastruc L, Pibouleau L, Senties OB. report 2015. U.S. Department of Energy; 2015. Report available at < https://www.
Assessment of mono and multi-objective optimization to design a hydrogen supply hydrogen.energy.gov/pdfs/progress15/iii_5_baldwin_2015.pdf > .
chain. Int J Hydrogen Energy 2013;38:14121–45. [40] Elgowainy A, Reddi K, Sutherland E, Joseck F. Tube-trailer consolidation strategy
[16] André J, Auray S, Brac J, De Wolf D, Maisonnier G, Ould-Sidi M-M, et al. Design and for reducing hydrogen refueling station costs. Int J Hydrogen Energy
dimensioning of hydrogen transmission pipeline networks. Eur J Oper Res 2014;39:20197–206.
2013;229:239–51. [41] Lahnaoui A, Wulf C, Heinrichs H, Dalmazzone D. Optimizing hydrogen transpor-
[17] André J, Auray S, De Wolf D, Memmah M-M, Simonnet A. Time development of new tation system for mobility by minimizing the cost of transportation via compressed
hydrogen transmission pipeline networks for France. Int J Hydrogen Energy gas truck in North Rhine-Westphalia. Appl Energy 2018;223:317–28.
2014;39:10323–37. [42] Robinius M, Linßen J, Grube T, Reuß M, Stenzel P, Syranidis K, et al. Comparative
[18] Reuß M, Grube T, Robinius M, Preuster P, Wasserscheid P, Stolten D. Seasonal analysis of infrastructures: hydrogen fueling and electric charging of vehicles.
storage and alternative carriers: a flexible hydrogen supply chain model. Appl Jülich: Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag; 2018.
Energy 2017;200:290–302. [43] Melaina M, Penev M. Hydrogen station cost estimates – comparing HSCC results
[19] JEC. Well-to-wheels report version 4.a: well-to wheels analysis of future automotive with other recent estimates. NREL; 2013. Report available at < https://www.nrel.
fuels and powertrains in the European context. JEC well-to-wheels analysis. JRC, gov/docs/fy13osti/56412.pdf > .
EUCAR, CONCAWE (JEC); 2014. Reports available at < http://iet.jrc.ec.europa.eu/ [44] H2Mobility. H2 tanken – Wasserstoffmobilität beginnt jetzt. H2Mobility; 2019.
about-jec/downloads > . [45] H2Moblity. 70 MPa hydrogen refuelling station standardization – function de-
[20] JEC. Well-to-tank report version 4.a: well-to wheels analysis of future automotive scription of station modules; NOW; 2010. Report available at < https://www.now-
fuels and powertrains in the European context. JEC well-to-wheels analysis. JRC, gmbh.de/content/5-service/4-publikationen/4-nip-wasserstoff-und-
EUCAR, CONCAWE (JEC); 2014. Reports available at < http://iet.jrc.ec.europa.eu/ brennstoffzellentechnologie/h2mobility_hrs_functional_description.pdf > .
about-jec/downloads > . [46] Bonhoff K. Supporting market ramp-up of hydrogen and fuel cell technologies.
[21] JEC. Tank-to-wheels report – version 4.0: well-to wheels analysis of future auto- World hydrogen energy conference 2016. Zaragoza; 2016.
motive fuels and powertrains in the European context. JEC well-to-wheels analysis. [47] McDonald A. Learning rates for energy technologies. Energy Pol 2001;29:255–61.
JRC, EUCAR, CONCAWE (JEC); 2013. Reports available at < http://iet.jrc.ec. [48] Nexant Inc., Air Liquide, Argonne National Laboratory, Chevron Technology
europa.eu/about-jec/downloads > . Venture, Gas Technology Institute, National Renewable Energy Laboratory, et al.

452
M. Reuß, et al. Applied Energy 247 (2019) 438–453

H2A hydrogen delivery infrastructure analysis models and conventional pathway Langfristszenarien für die Transformation des Energiesystems in Deutschland –
options analysis results. Nexant; 2008. Report available at < https://www.energy. Modul 3: Referenzszenario und Basisszenario. BMWi; 2017. Reports available
gov/sites/prod/files/2014/03/f9/nexant_h2a.pdf > . at < https://www.bmwi.de/Redaktion/DE/Artikel/Energie/langfrist-und-
[49] Aasadnia M, Mehrpooya M. Large-scale liquid hydrogen production methods and klimaszenarien.html > .
approaches: a review. Appl Energy 2018;212:57–83. [63] Kraftfahr-Bundesamt. Gesamtfahrleistung und durchschnittliche Fahrleistung nach
[50] Linnemann J, Steinberger-Wilckens R. Realistic costs of wind-hydrogen vehicle fuel Fahrzeugarten im Jahr 2017. Kraftfahr-Bundesamt; 2018.
production. Int J Hydrogen Energy 2007;32:1492–9. [64] Grube T, Stolten D. The impact of drive cycles and auxiliary power on passenger car
[51] Andrea MF, Sara RH, Luca DZ, Giovanni SS, Enrico B. Techno-economic analysis of fuel economy. Energies 2018;11:1010.
in-situ production by electrolysis, biomass gasification and delivery systems for [65] Cambridge Econometrics. Low-carbon cars in Germany. Cambridge Econometrics;
Hydrogen Refuelling Stations: Rome case study. Energy Proc 2018;148:82–9. 2017.
[52] Robinius M, Stein FT, Schwane A, Stolten D. A top-down spatially resolved elec- [66] Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-
trical load model. Energies 2017;10:361. learn: machine learning in {P}ython. J Mach Learn Res 2011;12:2825–30.
[53] Kraftfahrt-Bundesamt. Bestand am 1. Januar 2017 nach Zulassungsbezirken und [67] Radke S. Verkehr in Zahlen 2016/2017. DIW; 2017. Report available at < https://
Gemeinden. Kraftfahrt-Bundesamt; 2017. www.bmvi.de/SharedDocs/DE/Anlage/VerkehrUndMobilitaet/verkehr-in-zahlen-
[54] OpenStreetMap contributors. Planet dump retrieved from < https://planet.osm. pdf-2016-2017.pdf?__blob=publicationFile > .
org > . Dataset available at < https://download.geofabrik.de/europe/germany. [68] Wietschel M, Gnann T, Kühn A, Plötz P, Moll C, Speth D, et al. Machbarkeitsstudie
html > Open Street Map; 2017. zur Ermittlung der Potentiale des Hybrid-Oberleitungs-Lkw. Karlsruhe 2017.
[55] Center for International Earth Science Information Network (CIESIN). [69] Stolzenburg K. Integration von Wind-Wasserstoff-Systemen in das Energiesystem –
Documentation for the global rural-urban mapping project, version 1 (GRUMPv1): Abschlussbericht. Oldenburg; 2014.
urban extent polygons, revision 01. NY: NASA Socioeconomic Data and [70] Mischner J, Fasold H-G, Kadner K. gas2energy.net. Systemplanerische
Applications Center; 2017. Grundlagender Gasversorgung: Deutscher Industrieverlag 2011.
[56] GeoBasis-DE. Digitales Landschaftsmodell 1:1000000. BKG; 2018. [71] Adolf J, Balzer C, Haase F, Lenz B, Lischke A, Knitschky G. Shell Nutzfahrzeug-
[57] Kruskal Joseph B. On the shortest spanning subtree of a graph and the traveling studie – diesel oder alternative Antriebe – Womit fahren LKW und Bus morgen? In:
salesman problem. Proc Am Math Soc 1956;7:48–50. Shell Deutschland Oil GmbH, editor. Hamburg; 2016.
[58] Robinius M, Schewe L, Schmidt M, Stolten D, Thürauf J, Welder L. Robust optimal [72] Stolzenburg K, Mubbala R. Hydrogen liquefaction report – whole chain assessment.
discrete arc sizing for tree-shaped potential networks. Preprint; 2018. Integrated Design for Demonstration of Efficient Liquefaction of Hydrogen
[59] Yang C, Ogden J. Determining the lowest-cost hydrogen delivery mode. Int J (IDEALHY): FCH JU; 2013.
Hydrogen Energy 2007;32:268–86. [73] Schlesinger M, Hofer P, Kemmler A, Kirchner A, Koziel S, Ley A, et al. Entwicklung
[60] Europäische Union (EU). Verodnung (EG) Nr 1059/2003 des europäischen der Energiemärkte – Energiereferenzprognose. Bundesministerium für Wirtschaft
Parlaments und des Rates vom 26. Mai 2003 über die Schaffung einer gemeinsamen und Technologie; 2014. Report available at < https://www.bmwi.de/Redaktion/
Klassifikation der Gebietseinheiten für die Statistik (NUTS). In: Parlament E, editor. DE/Publikationen/Studien/entwicklung-der-energiemaerkte-
1059/20032003. energiereferenzprognose-endbericht.pdf?__blob=publicationFile&v=7 > .
[61] Eurostat. Strompreise nach Art des Benutzers. 08.05.17 ed. < http://ec.europa.eu/ [74] Thamling N, Pehnt M, Kirchner J. Hintergrundpapier Energieeffizienzstrategie
eurostat/2017 > . Gebäude. Prognos, ifeu, IWU; 2015. Report available at < https://www.prognos.
[62] Bernath C, Bossmann T, Deac G, Elsland R, Fleiter T, Kühn A, et al. com/uploads/tx_atwpubdb/20151220_Prognos_Hintergrundpapier_EES.pdf > .

453