Use of Hydrogen in Off-Grid Locations, a Techno-Economic Assessment

Diesel generators are currently used as an off-grid solution for backup power, but this causes CO2 and GHG emissions, noise emissions, and the negative effects of the volatile diesel market influencing operating costs. Green hydrogen production, by means of water electrolysis, has been proposed as a feasible solution to fill the gaps between demand and production, the main handicaps of using exclusively renewable energy in isolated applications. This manuscript presents a business case of an off-grid hydrogen production by electrolysis applied to the electrification of isolated sites. This study is part of the European Ely4off project (n◦ 700359). Under certain techno-economic hypothesis, four different system configurations supplied exclusively by photovoltaic are compared to find the optimal Levelized Cost of Electricity (LCoE): photovoltaic-batteries, photovoltaic-hydrogen-batteries, photovoltaic-diesel generator, and diesel generator; the influence of the location and the impact of different consumptions profiles is explored. Several simulations developed through specific modeling software are carried out and discussed. The main finding is that diesel-based systems still allow lower costs than any other solution, although hydrogen-based solutions can compete with other technologies under certain conditions.


Introduction
Population density and urbanization rates are key parameters to planning electricity grid extension. If certain thresholds are not reached, it is not cost-effective to provide access to the electricity grid for a large part of the population. Most of the incremental electrification over the period of 1990-2010 was in urban areas; even with this significant expansion, electrification only just kept pace with rapid urbanization in the same period [1]. On the other hand, both in developed and developing countries, renewable energies are growing. However, in developing countries, it is estimated that 1.16 billion people do not have access to electricity [2].
Another factor to take into account is that the addition of renewable power is limited in locations where there is no grid, weak grids, or grids that are already saturated with renewables due to the unpredictable or unsteady character of RES generation.
The world needs more RES generation. The Paris Agreement aims to strengthen the global response to the threat of climate change by keeping global temperature rise in this century well below 2 • C above pre-industrial levels and to pursue efforts to limit the temperature increase even further to 1.5 • C. To achieve this objective, the share of renewable energies must be 65% in 2050 [3], which will also have an important impact on the quality life of thousands of people across the globe [4].

•
Only electrical application • The electrical load profile comes from two different uses: isolated telecom antenna (1 and 4 kW steady load), which is becoming an interesting niche market, as is mentioned in other studies [11,12]; and off-grid home (variable load). • Two locations with different amount of solar resource: Tenerife (Spain) with high potential and Edinburgh (Scotland) with low potential.
Recent studies mention the scarcity and limitation of data sources [13] to compare off-grid renewable energy systems based on hydrogen to the other more common technologies mentioned before. This fact stresses that green hydrogen should still be subject to further techno-economic analysis in order to obtain valuable data. This paper uses literature from different sources, including first-hand Energies 2018, 11, 3141 3 of 16 information from the partners of the project, aiming to create original and valuable information regarding the use of hydrogen in off-grid environments and contributing to the scarce current literature, being a relevant input for further studies. The results show different scenarios in which a wide range of energy systems can provide electrical loads in an off-grid environment from a techno-economic point of view, obtained through several simulations with a specialized modeling software. This software permits greater detailing in modeling several equipment, as well as offering strategies and big data, obtaining very accurate and realistic solutions.
The interpretation of the results not only identifies scenarios that can be profitable with the use of an Ely4off-based system, but also offers an economic comparison with other competitors. The specialized software, the approach used, and the modeling of a novel system based on green hydrogen produced by PEM technology provide valuable information and will boost the integration of an off-grid and renewable hydrogen cycle technology in the European market.

Materials and Methods
By using a specialized software recently created by one of the project partners, a simulated microgrid was modeled, which represented an isolated energy installation. Both economic and technical data are well provided in the literature or by project partners, who have a broad experience in the manufacturing and working of the model. Figure 1 shows the methodology followed to perform the assessment.
Energies 2018, 11, x FOR PEER REVIEW 3 of 16 current literature, being a relevant input for further studies. The results show different scenarios in which a wide range of energy systems can provide electrical loads in an off-grid environment from a techno-economic point of view, obtained through several simulations with a specialized modeling software. This software permits greater detailing in modeling several equipment, as well as offering strategies and big data, obtaining very accurate and realistic solutions. The interpretation of the results not only identifies scenarios that can be profitable with the use of an Ely4off-based system, but also offers an economic comparison with other competitors. The specialized software, the approach used, and the modeling of a novel system based on green hydrogen produced by PEM technology provide valuable information and will boost the integration of an off-grid and renewable hydrogen cycle technology in the European market.

Materials and Methods
By using a specialized software recently created by one of the project partners, a simulated microgrid was modeled, which represented an isolated energy installation. Both economic and technical data are well provided in the literature or by project partners, who have a broad experience in the manufacturing and working of the model. Figure 1 shows the methodology followed to perform the assessment.

Simulation Tool
The simulation tool used in this work is ODYSSEY (Optimization and Design of hYbrid Storage Systems for rEnewable energY), a CEA-LITEN proprietary tool. It is an optimization platform developed to perform comprehensive techno-economic assessments of energy systems comprising renewable energy sources and energy storage units. The use of this tool makes the approach completely new, as the tool has only been used by CEA in other studies completely different from the present one.

Assessment and conclusion drawing
Sensitivty analysis to cover wider range of boundary conditions Comparison of optimized cases between the proposed system and competitors Obtaining optimized configurations under certain criteria Modeling the system proposed (based on H 2 ) and its competitors with the simulation tool Defining the case study: boundaries, competitors, configuration, strategies and input data

Simulation Tool
The simulation tool used in this work is ODYSSEY (Optimization and Design of hYbrid Storage Systems for rEnewable energY), a CEA-LITEN proprietary tool. It is an optimization platform developed to perform comprehensive techno-economic assessments of energy systems comprising renewable energy sources and energy storage units. The use of this tool makes the approach completely new, as the tool has only been used by CEA in other studies completely different from the present one.
The software has been entirely developed in C++ with an object-oriented approach and allows the following actions:

•
The possibility to integrate different degrees of technical modeling of the different system components (production units, storage units, power converters, auxiliary components).

•
The possibility to represent ageing of a given component through different approaches. • A high level of modularity to represent different architectures of electrical and fluidic systems (choice of AC or DC buses, location of power converters).

•
The ability to run simulations on small time steps in order to fully consider the impacts of intermittency and limited predictability of renewable energy sources on the economic results. Typical time steps range between 1 s and 1 h.

•
The possibility to systematically optimize both the system sizing and the power control strategy.
Once the architecture is defined, the user configures each element (production units, storage units, converters, etc.). To do so, a techno-economic model and model parameters must be chosen for each component. The user then defines the power control strategies in charge of operating the whole system. This methodology is well explained in previous papers developed by CEA [14,15].

Business Case Boundaries
The analysis is focused on the following case study: "Electrical powering of an isolated site load". This scenario refers to locations where a connection to the grid is not possible, either because it is too expensive or because the location is too remote. In this situation, hotels or mountain huts in remote locations are very common examples, as well as small isolated villages in locations where the grid quality is very weak. Telecommunication systems, solar water pumps, refrigeration, street lighting or traffic signs are other common examples [4].
Two specific locations with different and representative solar profiles were chosen to be considered: Tenerife (Spain) with a great amount of solar radiation through the year, and Edinburgh (Scotland) with a lower and very seasonal solar radiation.
In this study, four different customer electrical loads are created and introduced into ODYSSEY software using a 10-min time step series: Heating is not considered due to the time restrictions and lack of representative data, but it would be interesting to use the residual heat that comes from the fuel cell for this purpose.
The construction of "home loads" is based on data provided by ITER (Technology Institute on Renewable Energies) in Tenerife and Edinburgh. From these data, a 1-min time step load power profile was built using an academic excel tool [16,17].

Ely4off Configuration
Ely4off consists on an off-grid autonomous system to be connected to photovoltaic (PV) panels. The energy produced by the renewable source is used to produce hydrogen with the electrolyser through specially designed DC-DC converts (Direct Current to Direct Current). This eases the operation of the electrolyzer under the variability of renewable sources and improves efficiency.
For correct operation of the system, it is mandatory to supply the so-called essential load continuously. This load includes communication and control devices (safety sensors) and also a heater to prevent damage due to the low temperatures. When no radiation is available, a Hybrid Storage System composed of Lead-Acid batteries and a fuel cell is in charge of the essential load supply. This redundant Hybrid Storage System allows a self-sufficient operation of the Ely4off for several days without solar radiation.
Once the hydrogen is produced at a high purity, it is stored in a low-pressure buffer at 20 bars directly from the electrolyzer. The hydrogen can be used for any purpose (in the project, the main purpose is mobility), but also to supply the fuel cell and replenish the essential load. The overall configuration is shown in Figure 2.

Ely4off Configuration
Ely4off consists on an off-grid autonomous system to be connected to photovoltaic (PV) panels. The energy produced by the renewable source is used to produce hydrogen with the electrolyser through specially designed DC-DC converts (Direct Current to Direct Current). This eases the operation of the electrolyzer under the variability of renewable sources and improves efficiency.
For correct operation of the system, it is mandatory to supply the so-called essential load continuously. This load includes communication and control devices (safety sensors) and also a heater to prevent damage due to the low temperatures. When no radiation is available, a Hybrid Storage System composed of Lead-Acid batteries and a fuel cell is in charge of the essential load supply. This redundant Hybrid Storage System allows a self-sufficient operation of the Ely4off for several days without solar radiation.
Once the hydrogen is produced at a high purity, it is stored in a low-pressure buffer at 20 bars directly from the electrolyzer. The hydrogen can be used for any purpose (in the project, the main purpose is mobility), but also to supply the fuel cell and replenish the essential load. The overall configuration is shown in Figure 2. One of the characteristics of solar photovoltaics is the intermittency of the availability of power (due to cloud, fog or any other climate event). The PEMWE developed within Ely4off is capable of changing the H2 production from 0% to 100% in less than 30 s if the system is warm (typical condition if a cloud passes). If the climate event lasts longer and the system is cold, the nominal production of H2 is reached in less than 300 s, which is quite a good value, considering that the system performs safety checks to assure no leaks during turning on.

Power Systems in Competition
Three different systems are considered as potential competitors for an Ely4off system based on hydrogen production by PEM electrolysis: These configurations (and similar variants) are nowadays the most commonly used in off-grid locations, being therefore the most mature solutions [18].
In order to simulate the Ely4off system and its competitors, Figure 3 shows the conceptual configurations to be modeled by the software. One of the characteristics of solar photovoltaics is the intermittency of the availability of power (due to cloud, fog or any other climate event). The PEMWE developed within Ely4off is capable of changing the H 2 production from 0% to 100% in less than 30 s if the system is warm (typical condition if a cloud passes). If the climate event lasts longer and the system is cold, the nominal production of H 2 is reached in less than 300 s, which is quite a good value, considering that the system performs safety checks to assure no leaks during turning on.

Power Systems in Competition
Three different systems are considered as potential competitors for an Ely4off system based on hydrogen production by PEM electrolysis: These configurations (and similar variants) are nowadays the most commonly used in off-grid locations, being therefore the most mature solutions [18]. In order to simulate the Ely4off system and its competitors, Figure 3 shows the conceptual configurations to be modeled by the software.
Energies 2018, 11, x FOR PEER REVIEW 6 of 16 Figure 3. Systems in competition with Ely4off (diesel generator is not represented due to its simplicity).

Techno-Economic Hypothesis
Technical and economic hypothesis used for the development of simulations with ODYSSEY are presented in this section. The general set of hypothesis is represented in Table 1 and describes the simulation timeframe as well as time steps of the radiation profile and the electrical loads.

Techno-Economic Hypothesis
Technical and economic hypothesis used for the development of simulations with ODYSSEY are presented in this section. The general set of hypothesis is represented in Table 1 and describes the simulation timeframe as well as time steps of the radiation profile and the electrical loads. The reason for choosing 1 June is based on a timeframe scheduled in the project for the demo period, which will take eight months of operation. The simulation time step is 10 min, which is considered enough for the kind of system simulated.
Additional technical and economic hypothesis for each component within the systems are also necessary for system simulations. Economic hypothesis, such as investment, operation and maintenance (O&M), and replacement costs are presented in Table 2. As mentioned before, the data used is from previous studies using ODYSSEY [14,15,19] and from project partners' knowledge. The technical hypothesis for each of the element to be simulated is presented in Table 3. The data used in this table has been collected from recent studies developed by Ely4off partners, which includes hydrogen injection to the gas grid (to be published in 2019) and benchmarking of energy storage technologies [20]. Solar profiles are obtained from public databases developed by NREL [21].

Optimization Criteria and Methodology
ODYSSEY software allows for optimization of system components size in order to minimize user-defined indicators. Two criteria are used for the optimization of the system: • Economic criterion: levelized cost of energy (LCOE). It can be defined as the total cost required to install, operate, and maintain a power-generating asset over its lifetime, divided by the total energy output of the asset over that lifetime. Equation is defined in [22].
Provided electricity (y) (1+d) y−1 where n-life of the system, d-discount rate, y-year Both criteria must be minimized in order to achieve the best results. To obtain them, the parameters mentioned in Table 3 as "to be optimized by ODYSSEY" are shown in Table 4. In order to achieve the lowest criteria of LCOE and unsatisfied load, simulations are performed for each interval value among the range. The results presented in Section 3 show the optimized component value for which the system obtains the lowest LCOE and 0% of the unsatisfied load.
A multi-criterion, multi-parameter optimization algorithm is integrated into ODYSSEY. This genetic algorithm is used in this study to rapidly approach the optimal solution.

Sensitivity Analysis
In addition to the optimization criteria, a sensitivity analysis is carried out through additional simulations to obtain a wider outlook of the results. The parameters to which this analysis is applied are shown in Table 5.

Results
The results illustrate conditions wherein off-grid hydrogen production may present a techno-economic interest for isolated site electrification.
Results are presented for six different scenarios, considering the two different locations and the defined electrical loads.
Steady load 1 kW in Edinburgh 5.
Steady load 4 kW in Edinburgh 6.
Variable load (home) in Edinburgh Firstly, the results show that the available photovoltaic surface is a very important parameter to consider when comparing scenarios. The required photovoltaic peak power to be installed in Edinburgh is between two-three times greater than in Tenerife, depending on the final application as shown in Figure 4. Firstly, the results show that the available photovoltaic surface is a very important parameter to consider when comparing scenarios. The required photovoltaic peak power to be installed in Edinburgh is between two-three times greater than in Tenerife, depending on the final application as shown in Figure 4. The photovoltaic power installed is clearly different for every location, being very large in Edinburgh in order to supply 100% of the load. As a consequence, technical issues such as available space to install the PV panels should be considered in real scenarios. Figure 5 shows the optimal values observed for each simulated case, categorized by the location, type of load, and system in competition.
These results show that LCOE is three to four times higher in Edinburgh compared to Tenerife, when power solution is based on PV and storage means (no diesel generator). For the same load to satisfy, PV surface required in Edinburgh is about twice larger than in Tenerife and the required storage capacity to compensate for intermittency and variability is bigger, which explains the increased LCOE.
In Tenerife, for whichever type of load considered, PV-Bat-H2 is always the highest cost solution. This tends to show that hydrogen is not appropriate compared to a PV-battery solution in locations with high solar radiation.  The photovoltaic power installed is clearly different for every location, being very large in Edinburgh in order to supply 100% of the load. As a consequence, technical issues such as available space to install the PV panels should be considered in real scenarios. Figure 5 shows the optimal values observed for each simulated case, categorized by the location, type of load, and system in competition.
These results show that LCOE is three to four times higher in Edinburgh compared to Tenerife, when power solution is based on PV and storage means (no diesel generator). For the same load to satisfy, PV surface required in Edinburgh is about twice larger than in Tenerife and the required storage capacity to compensate for intermittency and variability is bigger, which explains the increased LCOE.
In Tenerife, for whichever type of load considered, PV-Bat-H 2 is always the highest cost solution. This tends to show that hydrogen is not appropriate compared to a PV-battery solution in locations with high solar radiation. In Edinburgh, for 1 kW and 4 kW steady loads, the hydrogen-based solution seems preferable to a battery solution, when coupled with PV, but LCOE remains much higher than the diesel-based solution. Scenarios 4 and 5 are the most promising cases for Ely4off-based solutions. Thereby, in order to discuss them in more detail, further results of Scenario 5 are presented in Figure 6. It shows in the same axis the unused PV (percentage of energy produced and not used) and the power consumed by the electrolyzer. The LCOE is shown in the other axis. The figure shows the ODYSSEY internal calculations done during the optimization process of Scenario 5. As the electrolyzer power decreases, the total LCoE also does, even if the PV power installed is increased. The figure shows a point in which the LCoE increases because the PV power also does, but the electrolyzer power cannot be lower in order to supply the electrical loads defined in the boundaries case. This allows stating a point in which the parameters are optimized to find the lowest LCoE (2.873 €/kWh) and 0% of unsatisfied load, with 41.8% of unused PV (oversized). This point is the one represented in the previous Figure 5 for Scenario 5.   In Edinburgh, for 1 kW and 4 kW steady loads, the hydrogen-based solution seems preferable to a battery solution, when coupled with PV, but LCOE remains much higher than the diesel-based solution. Scenarios 4 and 5 are the most promising cases for Ely4off-based solutions. Thereby, in order to discuss them in more detail, further results of Scenario 5 are presented in Figure 6. It shows in the same axis the unused PV (percentage of energy produced and not used) and the power consumed by the electrolyzer. The LCOE is shown in the other axis. The figure shows the ODYSSEY internal calculations done during the optimization process of Scenario 5. In Edinburgh, for 1 kW and 4 kW steady loads, the hydrogen-based solution seems preferable to a battery solution, when coupled with PV, but LCOE remains much higher than the diesel-based solution. Scenarios 4 and 5 are the most promising cases for Ely4off-based solutions. Thereby, in order to discuss them in more detail, further results of Scenario 5 are presented in Figure 6. It shows in the same axis the unused PV (percentage of energy produced and not used) and the power consumed by the electrolyzer. The LCOE is shown in the other axis. The figure shows the ODYSSEY internal calculations done during the optimization process of Scenario 5. As the electrolyzer power decreases, the total LCoE also does, even if the PV power installed is increased. The figure shows a point in which the LCoE increases because the PV power also does, but the electrolyzer power cannot be lower in order to supply the electrical loads defined in the boundaries case. This allows stating a point in which the parameters are optimized to find the lowest LCoE (2.873 €/kWh) and 0% of unsatisfied load, with 41.8% of unused PV (oversized). This point is the one represented in the previous Figure 5 for Scenario 5.   As the electrolyzer power decreases, the total LCoE also does, even if the PV power installed is increased. The figure shows a point in which the LCoE increases because the PV power also does, but the electrolyzer power cannot be lower in order to supply the electrical loads defined in the boundaries case. This allows stating a point in which the parameters are optimized to find the lowest LCoE (2.873 €/kWh) and 0% of unsatisfied load, with 41.8% of unused PV (oversized). This point is the one represented in the previous Figure 5 for Scenario 5. In order to understand the cost structure of the solution based on Ely4off (lesser known than the competitors), Figure 7 shows the cost breakdown for 20 years of operation of both the overall system and specifically the elements directly related to hydrogen technology (electrolyzer, fuel cell, compressor and storage). A major part of the cost is for the photovoltaic source and the conversion device for coupling to the electrolyzer; the hydrogen chain accounts for the second greatest cost, followed by the batteries. Within the hydrogen chain, the greatest cost is for hydrogen storage, followed by the electrolyzer, and the fuel cell. The compressor is the lowest cost because of its low power size.
Energies 2018, 11, x FOR PEER REVIEW 12 of 16 In order to understand the cost structure of the solution based on Ely4off (lesser known than the competitors), Figure 7 shows the cost breakdown for 20 years of operation of both the overall system and specifically the elements directly related to hydrogen technology (electrolyzer, fuel cell, compressor and storage). A major part of the cost is for the photovoltaic source and the conversion device for coupling to the electrolyzer; the hydrogen chain accounts for the second greatest cost, followed by the batteries. Within the hydrogen chain, the greatest cost is for hydrogen storage, followed by the electrolyzer, and the fuel cell. The compressor is the lowest cost because of its low power size. Li-ion battery results are 10% to 20% higher than lead-acid ones, probably caused by the replacement cost hypothesis. The same cost of replacement after 10 years of operation has been taken for both, although a fall in the cost of Li-ion can be expected in the next years.
In Figure 6, the concept of unused PV was presented for Scenario 5, showing values of around 40%. Similar results are obtained for the remaining technologies (for example, values as high as 30% to 40% of unused PV are obtained for Tenerife or Edinburgh. This outcome shows show that the low cost of PV installation favors oversizing, when compared to other elements of the system. However, it is very difficult to find a situation where 50% of the PV is unused (due to surface availability, for example), although it appears to be more economically interesting.
Within the frame of the Ely4off project, the nominal power envisioned for the electrolysis system is 50 kW. None of the optimized scenarios from this work led to a suggested size of electrolysis of 50 kW. Only the 4 kW steady load scenario in Edinburgh showed results close to it. It is worth pointing out the characteristics of this scenario in Table 6, which shows the values of the most competitive scenario based on Ely4off, compared against its competitors. Li-ion battery results are 10% to 20% higher than lead-acid ones, probably caused by the replacement cost hypothesis. The same cost of replacement after 10 years of operation has been taken for both, although a fall in the cost of Li-ion can be expected in the next years.
In Figure 6, the concept of unused PV was presented for Scenario 5, showing values of around 40%. Similar results are obtained for the remaining technologies (for example, values as high as 30% to 40% of unused PV are obtained for Tenerife or Edinburgh. This outcome shows show that the low cost of PV installation favors oversizing, when compared to other elements of the system. However, it is very difficult to find a situation where 50% of the PV is unused (due to surface availability, for example), although it appears to be more economically interesting.
Within the frame of the Ely4off project, the nominal power envisioned for the electrolysis system is 50 kW. None of the optimized scenarios from this work led to a suggested size of electrolysis of 50 kW. Only the 4 kW steady load scenario in Edinburgh showed results close to it. It is worth pointing out the characteristics of this scenario in Table 6, which shows the values of the most competitive scenario based on Ely4off, compared against its competitors.
It should be highlighted that these values are obtained based on a certain set of boundaries and other parameters, and should be merely considered as indications, on conditions allowing for a 50-kW electrolysis used in isolated site electrification, and as economically competitive compared to PV-Bat solutions. On the other hand, the hydrogen stored in Tenerife and Edinburgh for Scenario 5 clearly shows a seasonal tendency (see Figure 8). Energy is stored as H 2 during summer and consumed by the fuel cell during winter in both locations. It should be highlighted that these values are obtained based on a certain set of boundaries and other parameters, and should be merely considered as indications, on conditions allowing for a 50-kW electrolysis used in isolated site electrification, and as economically competitive compared to PV-Bat solutions.
On the other hand, the hydrogen stored in Tenerife and Edinburgh for Scenario 5 clearly shows a seasonal tendency (see Figure 8). Energy is stored as H2 during summer and consumed by the fuel cell during winter in both locations.
The left-hand section shows the H2 storage profile in Tenerife, indicating that H2 is produced and stored during summer and used in winter, with slights variations during the weeks. The right one shows the same situation in Edinburgh, where the seasonality of the H2 storage is more evident.
The effect of diesel price has been also assessed (see Table 5). In the results presented until now, a value of 2 €/L has been used, and the effect up to 3 €/L is assessed. It is observed (Figure 9) that LCOE values offer very slight variations (roughly 0.2 €/kWh per €/L) when price of diesel varies from 1 €/L to 3 €/L in Scenario 5. These variations do not change the comparison of technologies discussed throughout the manuscript.  The left-hand section shows the H 2 storage profile in Tenerife, indicating that H 2 is produced and stored during summer and used in winter, with slights variations during the weeks. The right one shows the same situation in Edinburgh, where the seasonality of the H 2 storage is more evident.
The effect of diesel price has been also assessed (see Table 5). In the results presented until now, a value of 2 €/L has been used, and the effect up to 3 €/L is assessed. It is observed (Figure 9) that LCOE values offer very slight variations (roughly 0.2 €/kWh per €/L) when price of diesel varies from 1 €/L to 3 €/L in Scenario 5. These variations do not change the comparison of technologies discussed throughout the manuscript.  It should be highlighted that these values are obtained based on a certain set of boundaries and other parameters, and should be merely considered as indications, on conditions allowing for a 50-kW electrolysis used in isolated site electrification, and as economically competitive compared to PV-Bat solutions.
On the other hand, the hydrogen stored in Tenerife and Edinburgh for Scenario 5 clearly shows a seasonal tendency (see Figure 8). Energy is stored as H2 during summer and consumed by the fuel cell during winter in both locations.
The left-hand section shows the H2 storage profile in Tenerife, indicating that H2 is produced and stored during summer and used in winter, with slights variations during the weeks. The right one shows the same situation in Edinburgh, where the seasonality of the H2 storage is more evident.
The effect of diesel price has been also assessed (see Table 5). In the results presented until now, a value of 2 €/L has been used, and the effect up to 3 €/L is assessed. It is observed (Figure 9) that LCOE values offer very slight variations (roughly 0.2 €/kWh per €/L) when price of diesel varies from 1 €/L to 3 €/L in Scenario 5. These variations do not change the comparison of technologies discussed throughout the manuscript.

Discussion
The main lessons that can be learned based on the work performed is: • LCOE for PV-Bat and PV-Bat-H 2 observed in Tenerife are three times lower than those observed in Edinburgh. This is not surprising, but the results offer a quantitative tool worthy for assessing the implementation of these technologies in locations with different solar resources.

•
The elements that appear to be giving hydrogen an advantage are: • A high seasonality for PV production requiring a large quantity of energy stored in summer to be reused in winter, as is shown for Edinburgh in Figure 8. In all of the scenarios assessed for Tenerife, the most expensive technology is PV-Bat-H2, while in Edinburgh, PV-Bat-H2 is the third-best technology in two scenarios. • It powers a demanding profile requiring daily energy shift from day to night. • A constrained PV production due to limited available surface.
It is interesting to notice that minimizing PV is not an optimal configuration from an economic point of view, as it requires a larger storage capacity, which has a strong impact on LCoE. It is always more economically interesting to oversize a PV plant, if possible, even though a potentially large part of the PV production is unused.
Additional results of the sensitivity analysis defined before lead to the following conclusions: 1.
For PV-Bat-H 2, in scenarios where hydrogen seems to have a potential interest compared with the rest of the competitors (steady loads in Edinburgh), a conclusion is that when a PV surface available is limited, increasing the electrolysis size may present a techno-economic interest.

2.
Diesel-based systems still allow lower costs than any other solution, even if we consider 3 €/L fuel cost, after the sensitivity analysis performed for different values of fuel cost as defined in the economic hypothesis.

Comparison with Other Studies
Other studies on the use of hydrogen as a power do not consider the same approach as the present study. Relevant differences on the methodology and the data used are described below:

•
To begin with, it is common to use the software HOMER to obtain LCOE values in these off-grid scenarios; however, this software has an important limitation of editable parameters when hydrogen is part of the system [10,23]. The use of ODYSSEY allows the modeling and programming of an infinite number of flows of mass, energy, etc. Elements such as the electrolyzer are modeled to include complex parameters, such as stoichiometry, pressure, operation temperature, and polynomial efficiency. This is reflected in the results and strategies to follow. • Some economic and technical data from Tables 2 and 3 are part of the current knowledge of the different project partners. Some values come from a specialized company in the manufacturing of integrated hydrogen energy systems (ITM Power) and from an innovative converter manufacturing company (Epic Power). Their business experience and market knowledge has a great impact over the main assumptions.

•
The Ely4off object of this assessment is a novel system, which allows total independence of the grid, even in very cold locations where anti-freezing action is needed. It also guarantees that the essential load is covered and allows direct coupling of the photovoltaic source to the PEM electrolyzer, avoiding losses. Other assessments based on renewable hydrogen for re-electrification do not consider a 100% of renewable penetration [24]. • This paper presents the use of hydrogen-based power to power systems and is compared against their main competitors that are common current solutions. This allows a wide outlook of the current status in off-grid scenarios, not only the ones based on hydrogen, but also on batteries and diesel generators.

•
Other approaches consider alkaline technologies and larger systems (>1 MW) [25]. The values obtained here are for different purposes and the use of PEM technology is more suitable to the variability of the renewable sources than alkaline, thanks to its dynamic operation, as shown in previous studies [26].
Based on these results, further research and simulations can be performed with different renewable sources, such as on-shore wind, off-shore wind or hydroelectric, in order to look for lower LCOE of the Ely4off system in off-grid locations for electrification case studies. It will also be included in different case studies: green mobility, power to hydrogen for gas grid injection and industrial applications.

Conclusions
The use of an energy system based on the Ely4off system is feasible but only profitable in some of the different scenarios discussed. The re-electrification of hydrogen is still very expensive, when compared to its competitors, which makes the installation of batteries and diesel generators more suitable. However, when seasonality is important, room for installing PV is limited, and electrical load is high during nights, the use of hydrogen can be profitable.
Thereby, it is encouraging and worth to point out that the system analyzed based on PEM technologies and able to operate in a self-sufficient way with 0% grid penetration has given economic values that can compete with other technologies under certain conditions. This is achieved thanks to the high dynamism of the PEM electrolyzer, the capability of hydrogen to be stored for long periods, its great energy density, and the high efficiencies achieved in the Ely4off system.