Modelling Decentralized Hydrogen Systems: Lessons Learned and Challenges from German Regions

Abstract

Green hydrogen produced by power-to-gas will play a major role in the defossilization of the energy system as it offers both carbon-neutral chemical energy and the chance to provide flexibility. This paper provides an extensive analysis of hydrogen production in decentralized energy systems, as well as possible operation modes (H₂ generation or system flexibility). Modelling was realized for municipalities—the lowest administrative unit in Germany, thus providing high spatial resolution—in the linear optimization framework OEMOF. The results allowed for a detailed regional analysis of the specific operating modes and were analyzed using full-load hours, share of used negative residual load, installed capacity and levelized cost of hydrogen to derive the operation mode of power-to-gas to produce hydrogen. The results show that power-to-gas is mainly characterized by constant hydrogen production and rarely provides flexibility to the system. Main drivers of this dominant operation mode include future demand for hydrogen and the fact that high full-load hours reduce hydrogen-production costs. However, changes in the regulatory, market and technical framework could promote more flexibility and support possible use cases for the central technology to succeed in the energy transition.

1. Introduction

With the Paris Climate Agreement and the goal of limiting the global annual temperature increase to 1.5 °C, the EU and Germany have now set themselves the goal of being climate-neutral by 2050 [1,2,3]. In order to achieve this goal, both the EU and the German government have a hydrogen (H₂) strategy for the integration of CO₂-neutral—in long-term only renewable—H₂ [4,5]. The transfer of climate-neutral, renewable energy (RE) produced by sources such as solar or wind power to the consumption sectors is made possible by sector coupling [6,7,8]. Power-to-gas (PtG) technology is of central importance for a systemic solution, as it fulfils the functions of a flexibility option and a long-term storage option for electricity and can provide CO₂-neutral substitutes for gas and fuels in the form of H₂ or methane for usage in other consumption sectors [6,9]. Thus, green H₂ enables the defossilization of power production, as well as the reduction in emissions in other sectors, such as heat, industry and transportation [10,11,12,13,14]. H₂ will play an important role in reaching climate goals in some sectors and applications [15]. Therefore, PtG technologies—as one sector-coupling technology—will play a crucial role in the defossilization of the energy system [16]. However, the use of H₂ from PtG has also drawbacks. For example, direct electrification is more efficient [17], and the optimal system design remains unclear. Given these aspects, this paper helps to assess possible operation modes of electrolysis and whether changes in the regulatory, market or technical framework can support the realization of the necessary contribution of PtG to the energy transition.

1.1. Definition of Power-to-Gas

The term PtG is used to refer to several different concepts and processes. PtG technology is based on the electrolysis process. Electrolysis for converting electricity into H₂ is the core technology of PtG. Some concepts include further processes in which H₂ is processed into methane (synthetic natural gas (SNG/CH₄)) or into fuels (synthetic fuels (SF)) [9,16,18]. An overview of PtG technologies and its concepts is shown in Figure 1, according to Zapf (2017) and Adolf et al. (2017) [19,20]. In the following, the term PtG refers only to the conversion of renewable electricity into H₂. To align with the goals of sector coupling, it is assumed that electricity consumed from the grid is renewable from a balanced perspective in the base year. For the year 2050, we assume a 100% renewable electricity supply.

In theory, PtG can take on different functions. The technology can be used as a flexibility option to balance fluctuating RE generation or to produce large quantities of H₂ to replace fossil fuels. In practice, the function of PtG lies between the two extremes [21].

As a flexibility option, PtG offers an advantage for the energy system regarding the higher fluctuation of energy production in a renewable power sector. The necessary flexibility is provided by the demand side through PtG. In this context, PtG as a flexibility option also offers a storage option for electricity. PtG can offer the opportunity to use chemical storage to shift energy into periods with low supply of RE sources both in the short and long term. Thus, PtG is a flexibility option for fluctuating RE sources in cases of high power production but low demand and/or in cases of local power-grid congestion [22,23]. The flexible operation mode allows for the storage of H₂. In of higher demand than can be met by renewable power production (e.g., dark doldrums), a reconversion to electricity is possible.

Produced H₂ can be exported [9], and PtG is used for the defossilization of sectors beyond the power sector. It is possible to feed PtG as admixture into the gas network or to export pure H₂. In addition, produced H₂ can be used to meet an existing or future demand for H₂. The latter results from potential use in the sectors such as heating, industry or traffic. The demand can be material or energetic [9,24,25].

Given these facts, a need for H₂ exists. The necessity can be explained by the necessary defossilization, as well as the increasing need for flexibility in the energy system. However, the respective regional deployment and operation mode must be investigated.

The regional expansion of electricity generation based on renewable energies causes a change in the regional energy balance. Regions with high installed renewable-energy (RE) generation capacity have energy quantities that exceed their local energy demand. In Germany, for example, this is the case in regions with high wind-energy potential in the north. The German federal state of Mecklenburg-Vorpommern (MV) in the northeast had the highest share of RE in gross electricity consumption of 147% among the German states in 2016 [26]. Due to the volatile feed-in, there are deviations between RE supply and energy demand during the year, with the extent of these differences varying from region to region. As one consequence, the resulting operating mode for electrolyzers may differ between an urban region and a rural region. In addition to the difference in demand quantity and structure, the installed RE power capacity and generation in each region must also be considered [27].

1.2. Power-to-Gas in the Energy System

There is extensive literature on PtG in energy systems. Most studies address the potential of PtG at the global or national level of energy-system analysis. In contrast, the analysis of local, decentralized energy systems is less prominent and is mostly based on one concrete, local use case (see

Table 1. General information of investigated literature.

Author (Year)	Energy-System Analysis	Spatial Resolution	Model (Methodology)
Fraunhofer ISI et al. (2017) [28]	Yes	Input regionalized, calculation with six regions for Germany	Enertile (optimization of dispatch and investment), combined with detailed grid and demand models
Fraunhofer ISI et al. (2021) [29]	Yes	Input regionalized, calculation with six regions for Germany	Enertile (optimization of dispatch and investment), combined with detailed grid and demand models
Robinius et al. (2020) [30]	Yes	Input regionalized, calculation with Germany as one node	FINE-NESTOR (optimization of dispatch and investment), combined with detailed grid and demand models
Dena (2018) [31]	Yes	Input regionalized, calculation with Germany as one node	DIMENSION + (optimization of dispatch and investment)
Prognos et al. (2020) [32]	Yes	Input regionalized based on reference regions, calculation with Germany as one node	System model based on model combination (optimization and simulations)
Dena (2016) [33]	No	Analysis of four cluster regions	No simulation or optimization
Zdrallek et al. (2018) [34]	No focus	Modelling based on nine reference municipalities	Dispatch simulation
Schröer et al. (2015) [35]	No focus	Using one reference region with 17 sub regions	P2IONEER (simulation)
Hey (2012) [36]	No	Reference use case	Matlab/Simulink (simulation)
Sterchele et al. (2020) [37]	Yes	Germany as one node	REMod (simulation and optimization)
Kelch et al. (2019) [38]	Yes	Modelling counties and municipalities	Ren-Model (Simulation)
Andresen and Schmitz (2016) [39]	No	Using Hamburg as a reference municipality	TransiEnt.EE (simulation)
RLI (2013) [40]	Yes	Germany represented by 14 nodes	Optimization model for dispatch and investment

and

Table 2. Representation of power-to-gas, electrolysis operation and influencing factors in investigated literature.

Authors (year)	Power-to-Gas Types	Electricity/Gas/Heating Sector	Electrolysis Operation	Influencing Factors
Fraunhofer ISI et al. (2017) [28]	CH4, E-Fuels, H2	+/+/+	No	Legal and political framework, technological development, grid expansion
Fraunhofer ISI et al. (2021) [29]	CH4, E-Fuels, H2	+/+/+	No	Legal and political framework, technological development, grid expansion
Robinius et al. (2020) [30]	CH4, E-Fuels, H2	+/-/+	Yes	Legal and political framework, technological development, renewable expansion
Dena (2018) [31]	CH4, E-Fuels, H2	+/+/+	Yes	Legal and political framework, technological development, social acceptance
Prognos et al. (2020) [32]	CH4, E-Fuels, H2	+/+/+	Yes	Technological development, carbon price
Dena (2016) [33]	H2, CH4	+/+/+	Yes	-
Zdrallek et al. (2018) [34]	H2, CH4	+/+/-	Yes	Legal and political framework, technological development
Schröer et al. (2015) [35]	H2, CH4	+/+/+	Yes	Legal and political framework, technological development
Hey (2012) [36]	H2, CH4	+/-/-	Yes	Legal and political framework
Sterchele et al. (2020) [37]	CH4, E-Fuels, H2	+/-/+	Yes	Social acceptance
Kelch et al. (2019) [38]	-	+/-/+	No	Legal and political framework, technological development
Andresen and Schmitz (2016) [39]	H2	-/+/-	Yes	Variation of possible admixture in gas grids
RLI (2013) [40]	-	+/-/-	No	Technological development

).

Sterchele et al. (2020) examined the development of the German energy system to achieve the 2050 emission-reduction targets. Technical feasibility, costs and change in societal behavior were considered, showing that climate targets can be achieved. Green H₂ is a key energy carrier in an RE system [37].

Further energy-system studies for Germany come to comparable conclusions. The energy-system analysis by Kelch et al. (2019) focused on potential energy autarky by 2030 in two exemplary communities. The focus lay on the electricity and heat sectors. The authors showed that an autarky level of 80% can be achieved with an energy system based on RE. Sector coupling is an important part. However, PtG was not part of the analysis [38].

Prognos et al. (2020) [32], Fraunhofer ISI et al. (2021) [29], Robinius et al. (2020) [30], Dena (2018) [31], Fraunhofer ISI et al. (2017) [28] and RLI (2013) [40] also have in common that they analyzed long-term greenhouse-gas reductions. Electrolyzers play an essential role in ambitious reduction scenarios. The reconversion of H₂ into electricity to provide flexibility for the system is also seen as central element in the future. H₂ based on electrolysis as a chemical-energy carrier is the starting point for the defossilization of further sectors in the case of greenhouse-gas-neutral scenarios. Thus, two roles for PtG are addressed in system studies: on the one hand, the permanent provision of H₂ as an energy carrier; on the other hand, PtG enables provision of the required flexibility for the system. The exact role in decentralized energy systems remains without detailed analysis in all studies. Changes in the legal and policy framework are rarely (or only in very general terms) considered, without specifically addressing PtG.

In its potential atlas for PtG, Dena (2016) examined existing utilization options. In addition, the authors discussed the short- and medium-term market development of the technology. Market opportunities for PtG were identified in all consumption sectors. Necessary recommendations for action were made, especially in the regulatory area. Concrete approaches for regional implementation or the value creation in the decentralized energy system were missing [33].

The potential study by Zdrallek et al. (2018) goes into detail about the potentials in distribution networks. The study aimed to estimate the potential of current and future installation and operation of PtG plants in German electricity- and gas-distribution networks. Nine municipal classes are analyzed with regard to their typical supply tasks of electricity and gas networks. Increasing gas consumption results in greater technical potential and savings in network-expansion costs. Economic revenue potential was found to exist in only one of the municipal classes, and possible regulatory changes were addressed. No statements were made on the specific role or function of the PtG system at the decentralized level [34].

Schröer et al. (2015) investigated PtG technology in a model-based manner at the decentralized level. Different configurations and autarky lines of supply were considered. The use of H₂ storage, fuel cells or direct H₂ injection into the gas grid, as well as power exchange across the system boundary of the model region, were not considered in the model. PtG was identified as an important system component in distributed energy systems [35].

Hey (2012) also dealt with another concrete PtG use case. He investigated storage options for surplus electricity and components of demand-side management. He considered to what extent a PtG plant is capable of providing control energy as a system service. In addition, he also analyzed the contribution to balancing forecast deviations in wind -energy feed-in. The integration of PtG plants into an energy system was not included in the model [36].

Andresen and Schmitz (2016) investigated the suitability of PtG as a storage option using an application in Hamburg. The focus lay on the design and the technical simulation of a PtG plant. The investigation showed that the dimension of a PtG plant strongly depends on given boundary conditions and restrictions. H₂ storage was identified as a necessary component. Repercussions and interactions with the (local) energy system were not part of the analysis [39].

In summary, the focus of all investigations was either energy-system analysis or PtG technology. In energy-system analyses, the focus was rarely on distributed generation. PtG was not the exclusive focus in any case and was not even considered in some analyses. The modelling of PtG only addressed specific use cases. They are short-term-oriented and only refer to the long-term time horizon up to 2050 in some studies. Only the study by Zdrallek (2018) [34] and the work of Schröer et al. (2015) [35] focused on both distributed energy systems and PtG. Furthermore, it is unclear how the legal, policy, and technical framework will affect the deployment of PtG in regional energy systems.

In this context, this paper investigates which operation mode of PtG (H₂ generation or system flexibility) is primarily realized and whether there could be a profitable use case. In addition, the influence of the regulatory market and technical framework on the use of PtG must be considered. The regional consideration of municipalities enables a more detailed analysis of specific operating modes and the impact of the regulatory market and technical framework on PtG. This will be analyzed in the context of different regional supply tasks, which are defined by the exploitation of regional RE potentials, regional electricity demand and potential H₂ demand. This question will be answered by using the operating modes of PtG in the local energy system.

For this purpose, this paper is structured in the following way: Section 2 provides an overview of the methods used. In Section 3, the database used for analysis is discussed. The results of the investigated scenarios are presented in Section 4. The operation mode of PtG in regional energy systems is presented in Section 4.1, and the effect of regulatory measures is addressed in Section 4.2. The results are then interpreted and discussed in Section 5 and Section 6, where the question of profitable use cases will be addressed, with our conclusions provided in Section 7.

2. Method

The research questions are answered based on the modeling of decentralized energy systems. For this purpose, we focused on municipalities (the different regions) in Nordwestmecklenburg (NWM) county in the federal state of MV. The selection of representative municipalities within NWM was realized by applying a cluster algorithm.

The research question was then answered in the next three steps (Figure 2) by modelling the different decentralized energy systems at different points in time (status quo and 2050 scenario) and by a criteria-based (key performance indicators) comparison of the regions. The open-source framework OEMOF was used for the modeling [41]. We referred to levelized cost of H₂ (LCOH₂) to assess the potential profitability of the regional PTG system. Together with the operations hours and the use of local negative residual load, this was used as an indicator for the operation mode. The impact of the regulatory, market and technical framework was determined by varying the corresponding input parameters. The results provide information on the influence of specific measures discussed in the literature.

2.1. Cluster Analysis

The first step of analyzing and deriving basic statements about decentralized energy systems is the identification of representative municipalities. For this purpose, a cluster analysis of municipalities in NWM was carried out.

The literature presents examples of cluster analyses in which regions in Germany are clustered according to different criteria [42,43,44,45,46]. To analyze decentralized energy systems, the studies use technoeconomic energy indicators at the municipal level [47,48]. With regard to decentralized energy systems, a wide range of possible parameters can be used. Typical data of decentralized energy systems are, for example, electricity, heat or fuel demand by consumption sectors; installed electricity or heat generation capacities; as well as social and geographic data, such as area, population or number of households [42,43,47]. We also included information about the potential role of hydrogen for the municipalities in NWM in our analysis (see Section 3.1).

First, a correlation analysis was carried out to identify independent indicators. This was to avoid bias due to overestimation of certain parameters. A strong correlation was assumed for a correlation coefficient above 0.9 [42,49,50]. For the cluster analysis, only the independent indicators were used.

The aim of the clustering analysis was to identify clusters that that are heterogeneous among themselves and homogeneous within the cluster. In

Table 3. Overview of the aspects of the cluster analysis.

Characteristics	Attribute
Method:	Hierarchical-agglomerative, partitioning
Cluster algorithm:	Ward, K-Means
Distance measures:	Squared Euclidean distance
Variables/indicators	25 indicators (see Section 3.1)
Objects:	86 municipalities in Germany
Criteria:	Elbow criteria, test of Mojena

, the relevant information about the cluster analysis is listed.

The Ward method was chosen as the cluster algorithm. It is a hierarchical-agglomerative method. The two clusters that produce the minimum increase in variance in the new cluster are merged. The aim is to minimize the loss of homogeneity that occurs by merging two clusters. The algorithm produces very homogeneous groups and is therefore considered the most powerful among the agglomerative procedures [49,51]. To verify the results, the clusters are checked with the K-Means algorithm in a final step [51]. In both cases, the quadratic Euclidean distance is used to determine the similarity of objects based on the distance between two objects. Thus, the used statistic measure of similarity is one of the most frequently used approaches [51].

Based on the cluster result, it was finally necessary to determine the number of clusters for further analysis of the decentralized energy systems. The elbow criteria and the test of Mojena were used to determine the number of clusters [51]. For this purpose, the standardized fusion coefficient per fusion level was used. In conclusion, the representative examples for each cluster were selected.

2.2. Energy-System Modeling

An H₂-model was developed using the OEMOF framework, employing the oemof.solph module. OEMOF is a generic graph-based representation of the energy system. The model framework is modular, based on a linear optimization approach and designed for flexible modeling tasks [41]. The framework is based on the Pyomo package. Pyomo is an open-source Python software package for optimization models [52,53]. Further information on the mathematical model formulations and the modular structure can be found in the documentation of OEMOF [54]. Thus, OEMOF allows for a simple representation of energy systems (see Figure 3).

We use the investment option to model the individual regional energy systems. The energy systems were optimized for the time horizon of the investigated year (status quo or 2050) in hourly resolution. The linear problem was solved by using the Gurobi solver [55].

The applied energy-system model minimizes the total system costs over one year and takes seasonal fluctuations on the demand and supply side into account. According to Löffler et al. (2017) [56] and Zerrahn and Schill (2017) [57], we considered annuities for the investment costs, annual fixed costs and variable costs (e.g., fuel costs) for each component to determine the system costs. The mapped components include RE capacities, biomass combined heat and power (CHP) capacities, electricity import, PtG capacities, H₂ storage, reconversion and transport capacities for H₂ export. For each time increment, a price-inelastic demand for electricity and H₂ must be met. A description of the sets, parameters, scalars and variables used is given in the Appendix.

2.3. Key Performance Indicators

Further energy-related key performance indicators can be determined in post-processing of the model outputs. The analysis of the results included the following variables: full-load hours, installed capacity and the share of used negative residual load by the electrolyzers to indicate the operation mode. High full-load hours indicate a constant production of H₂, and low full-load hours suggest a flexible operation mode. The installed capacity was used to further specify the result, as it is possible to indicate whether power peak-load situation of RE should be reduced by PtG. The share of used negative residual load also supports, in combination with the other two parameters, the analysis of the operation mode. A low share, together with a high installed capacity and few full load hours, bolsters the conclusion of a flexible operation mode. However, the same conditions, in combination with a high use of negative residual load, suggest a constant operation of H₂.

We also focused on LCOH₂ to assess costs and, in a second step, the possible economic profitability. LCOH₂ enables the comparison of different production options or locations. Comparatively, low LCOH₂ means that H₂ can be produced at low cost. Such a location or production option has an advantage over others [58,59]. This allowed us to compare the regions with one another and to assess changes in the regulatory, market and technical framework. The calculated LCOH₂ values were also compared with values from other studies for further assessment.

LCOH₂ (Equation (1)) was calculated from the sum of the annuity costs of the investment and the annual fixed costs ($C_{PtG}^{inv}$) of the PtG plant and the sum of the variable costs ($C_{PtG,t}^{car}$) divided by the annual H₂ production ($E_{PtP}$). We assumed that the variable cost and the production amount were constant over the years following the modelled year.

$$LCO{H}_2=\frac{C_{PtG}^{inv}+{{\displaystyle \sum }}_t{C}_{PtG,t}^{var} }{E_{PtG}}$$

(1)

The annuity of the investment results from the investment module in OEMOF. In this case, the variable operating costs of the electrolyzer can be calculated from the local electricity costs. The electricity price (Equation (2)) (determined on a weighted basis according to the locally existing sources ($E_{E{L}_{import,t}}$, $E_{BiomassCHP,t}$, $E_{RES,t}$)). The electricity can be obtained from the upstream grid (electricity exchange price—$C_{E{l}_{import},t}^{flow}$), the RE plants ($C_{RES,t}^{flow}$) or a biomass CHP unit ($C_{BiomassCHP,t}^{flow}$).

$$P_{el,local,t}=\frac{C_{E{l}_{import, t}}^{flow}+C_{BiomassCHP,t}^{flow}+C_{RES,t}^{flow}}{E_{E{L}_{import,t}}+E_{BiomassCHP,t}+E_{RES,t}}$$

(2)

This price time series represents the hourly local electricity costs. The local variable costs for the electrolyzer (Equation (3)) were derived by a multiplication of the hourly electricity demand of the electrolyzer and the hourly local electricity costs. For this purpose, the efficiency (${\eta }_{Pt{h}_2,t}$) of the electrolyzer was included.

$$C_{PtG,t}^{car}=P_{el,local,t}\ast \frac{E_{PtG}}{{\eta }_{PtG,t}}$$

(3)

2.4. Parameters to Assess Changes in the Regulatory, Market and Technical Framework

To answer the research question, the status quo and 2050 scenarios were used for each region. The input data are described below. In addition to the two observation points, further sensitivities were considered. The influence of parameters on the results of the status quo and 2050 scenario can be determined in this way (see

Table 4. Manipulated input parameters in the status quo and in 2050.

Parameter	Framework Parameter	Change/Measure
H2 demand	Market	Change in demand
H2 price	Market	Subsidization
Electricity price components	Regulatory	Considering levies
Electricity price variation	Market/Regulatory	Wholesale price
Electricity grid	Technical	Grid capacity

for the varied parameters.) Thus, we can show whether the different possible operation strategies are robust solutions. Furthermore, this allows for the assessment of the regulatory framework.

Short-term H₂ demand is a crucial parameter for investment in electrolysis capacities. An increase or decrease in the size of the sales market can significantly affect the profitability of PtG. From a political point of view, given quotas could boost the demand for H₂. The variation of H₂ export in the model allows for an assessment of the role of the current incentives for H₂ export. Effects of subsidies were analyzed using a variation of the sales price.

Another influencing factor is electricity price components. We focused primarily on levies, such as grid-utilization fees or the EEG-Umlage (levy to promote renewable energy sources based on the German Renewables Act (EEG)) as crucial parameters. The variation has an influence on the exchange with the upstream electricity network and the electricity source of electrolysis. In addition, we looked at the influence of the electricity price in general by a manipulation of the variation of the applied wholesale electricity prices.

From a technical point of view, the operating strategy can be influenced by transportation networks. H₂ demand is probably limited by demand and not by export facilities. Therefore, our focus was on measures in the electricity grid. Grid-expansion measures can have an impact on local capacity and operation.

3. Input Data

The input data are made up of three components. Section 3.1 presents the regional energy-related data on the level of municipalities. These were initially included in the clustering and form the basis for the modelling of the energy systems in the representative municipalities. The data of the representative municipalities will be discussed in more detail after the presentation of the cluster results in Section 4.2. General energy-market data are presented in Section 3.2 and include information about electricity wholesale prices and potential hydrogen prices. The last component of the input data is technoeconomic data, which are presented in Section 3.3 and include investment and operational cost information about the technologies used in the decentralized energy systems.

3.1. Regional Data

The subjects of this study were the municipalities in the region of NWM in the federal state of MV, Germany. For the modelling of decentralized municipal energy systems, data with high spatial resolution are necessary. The data basis is shown in

Table 5. Data basis for the cluster analysis and energy-system modelling.

Socio-Demographic	Energy Demand	RE
Area	Electricity demand—households (MWh)	Capacity wind (MW)
Population	Electricity demand—industry (MWh)	Capacity pv (MW)
Number of households	Electricity demand—commercial (MWh)	Capacity—bioenergy (MW)
Number of flats	Electricity demand—mobility (MWh)	Capacity—energy (other) (MW)
	Heat demand—households (MWh)	Wind potential (MW)
	Heat demand—industry (MWh)	PV roof potential (MW)
	Heat demand—commercial (MWh)	PV open-space potential (MW)
	Hydrogen demand—household (MWh)	Biomass potential (MWh)
	Hydrogen demand—industry (MWh)	Biogas potential (MWh)
	Hydrogen demand—commercial (MWh)	Emission (tCO2eq)
	Hydrogen demand—mobility (MWh)

. The data consist of sociodemographic data and data concerning local demand and supply of energy for the year 2016 [60]. All sectors (households, industry, commercial and mobility) were included. In addition, installed capacity and potential for decentralized (renewable) power plants were included in the dataset. The basis of the dataset was a bottom-up modelling for all municipalities in NWM [61].

For the cluster analysis, all indicators shown in Table 5 were used. For the OEMOF model of the different decentralized energy systems, the indicators for electricity and hydrogen systems were used. Energy-system modeling focuses on the electricity sector, as the electricity supply is crucial for the operation of an electrolyzer (as described above). Local energy supply was represented by the installed capacities of wind, solar energy (PV) and biomass. For the status quo, the installed capacities in 2016 were used. For the year 2050, we assumed a complete utilization of local RE potentials.

The feed-in profiles for wind and PV were based on weather data from Germany provided by the Open Power System Data (OPSD) platform [62]. The same profiles were used for the status quo and 2050.

H₂ demand was aggregated from the heat, industry and transport sectors. There are currently no standard load profiles for H₂ supply. Thus, the distribution over time was based on sector-specific assumptions. The demand time revenues were also passed on to the model in hourly resolution.

To derive the H₂ demand in the heat sector, the heat demand of the household and the commercial sectors was used [62]. Here, it is possible to derive the demand based on existing load profiles for gas. As these sectors are supplied via the gas grid, we considered the current limit between 2% and 20% of the gas-volume flow for the H₂ feed-in [63,64]. The feasibility of higher feed-in quotas was assumed for the future scenario.

The future annual H₂ demand in transport was developed based on the studies by e-mobil BW et al. (2013) and Prognos and thinkstep (2019) [65,66]. The energy demand can be determined based on the current number of inhabitants, the regional vehicle fleet, the expected share of H₂-based vehicles and the predicted annual mileage [67,68]. The temporal distribution of demand was based on Grüger et al. (2018) [69]. In the first step, an H₂ demand profile for vehicle types was derived, which was transferred to the demand of H₂ filling stations in the second step. Based on this, it was assumed there is 20% less freight transport on Saturdays and none on Sundays due to the ban on non-essential truck driving on Sundays and public holidays in Germany (according to § 30 para. 3,4 StVO). For buses, 20% less traffic was assumed during weekends. The H₂ consumption of vehicles is also temperature dependent, considered a seasonal factor [70,71]. The sum of the demand groups results in the H₂ demand at petrol stations. It was assumed that supply is delivered by trailer.

Industry H₂ demand is made up of feedstock demand and energy demand. Feedstock H₂ demand mainly consists of refineries, ammonia and methanol production. Livestock fields are not located in the investigated regions. Thus, feedstock use was not considered in the aggregated H₂ demand The energy demand is to more than 85% driven by process heat [72]. Thus, we concentrated on process heat in industry. The demand was estimated based on Naegler et al. (2016), Blesl and Kessler (2017) and Sterchele et al. (2020). The share of industrial process heat substituted with H₂- or H₂-based gases in the future (2050) was estimated at 40%. For the hourly resolution, a constant demand in the industrial sector was assumed, and a band-load profile of the local network operator, E.DIS Netz (2020), was used for the hourly resolution [73].

3.2. General Energy-Related Data for All Decentralized Energy Systems

The costs for imported electricity and revenues for exported electricity were based on the day-ahead price for the German bidding zone. The price was stored as a time series (see Figure 4). A trading surcharge of 10 €/MWh was applied to electricity exports in order to exclude arbitrage opportunities [74]. The electricity price time series for Germany in the status quo was taken from the POMMES model [75]. The mean price for the status quo was 28.98 €/MWh_el. For the year 2050, this time series was scaled with an assumed mean value of 40 €/MWh and a standard deviation of 1.9. Current demand was also taken from “ENavi” data [76]. The temporal resolution was modelled using the standard load profiles (SLP) of the grid operator in the region, E.DIS Netz [73].

In addition to the electricity price, regional electricity-grid capacities are also relevant. The local electricity demand can be covered by local RES or by imports from the transmission grid. A limitation was made for the estimated regional grid capacity. This was based on the maximal value of either the local load or the local RES generation supplied via the grid without congestion. In addition, a security surcharge of 60% was applied.

Imports and exports of H₂ in and out of a region were also considered. The results can be used to determine whether a region can supply itself with H₂ or act as an exporter. Filling-station prices for H₂ are currently 9.50 €/kg_H2, which corresponds to a price of 275 €/MWh_H2. Sterchele et al. (2020b) were in agreement with this number and stated that taxes and levies were included. Thus, we used this price for the import of H₂. For 2050, a reduction of around 54% was assumed [77,78,79]. The export price was more complex, as only profit was considered. Thus, current levies, such as grid-usage fees or the EEG-Umlage and taxes must be subtracted [80]. This results in income of 52 €/MW_H2 for the status quo. In the future, a reduction of around 54% was also assumed for export revenues. These values are also in line with other studies on future costs of H₂ products [77,81].

3.3. Technoeconomic Data

The cost of electricity from local wind and PV capacities was assumed to be zero. The biomass CHP plant was also represented by the existing capacity, based on the energy atlas [60]. Electricity and heat generation were set at a ratio of 50:50 of the installed capacity, with electrical and thermal efficiency of 45% each. The variable costs of the CHP result from the fuel input. The required biomass was set at 30 €/MWh [37,77]. Other variable operating costs were neglected.

The electrolysis used in the model was based on PEMEL electrolysis. The byproducts of electrolysis, such as oxygen and waste heat, were neglected. The efficiency was assumed to be 67%. Based on Sterchele et al. (2020b), an efficiency increase to 73% was taken into account by 2050 [77,82]. For dimensioning of the electrolyzer, investment costs of 738 €/kW_el and 495 €/kW_el were assumed for the status quo and the future scenario, respectively [37]. The annual fixed operating costs (OPEX) were set at a share of 3.5% and 3.9% [77] of the total investment. In the literature lifetime ranges between 20 and 30 years [36,77], depending on the year under consideration. A life span of 20 years was assumed for the status quo and 30 years for the future scenario. assumptions of the weighted average cost of capital (WACC) range between 5 and 7% [36,37,83], which is why a uniform WACC of 6% was chosen for all investment components.

The second investment component in our model is H₂ storage. The dimensioning of the storage tank has an influence on the dimensioning and operation of the electrolyzer and can therefore also be determined by an endogenous investment. In addition to the storage capacity, maximum injection and withdrawal rate, annual loss rate and initial maximum and minimum storage levels were considered. Injection and withdrawal were assumed to be ideal due to small losses in reality [84,85]. The ratio between the injection and withdrawal capacity and the storage capacity is another relevant parameter. An injection and withdrawal ratio of 0.5 was assumed for the modelling [86]. The investment parameters of the H₂ storage system were taken from Sterchele et al. (2020b) [77].

The third investment component is the reconversion option, the dimension of which also has an influence on the dimensioning of the electrolyzer. A gas turbine was applied, with an efficiency of 40%. The investment parameters were also based on the data from Sterchele et al. (2020b) [77]. Information about the basic data for the investment components is provided in

Table 6. Input parameters of the investment components.

Input Parameter	Unit	Value (2050)	Source
Electrolyzer investment	€/kWel	738 (495)	Sterchele et al. (2020b)
Electrolyzer operational cost	€/kWa	3.5% of investment	Sterchele et al. (2020b)
Electrolyzer lifetime	a	20	Sterchele et al. (2020b)
H2 storage investment	€/kW	163	Sterchele et al. (2020b)
H2 storage operational cost	€/kWa	2.5% of investment	Sterchele et al. (2020b)
H2 storage lifetime	a	30	Sterchele et al. (2020b)
H2 storage yearly loss rate	%	1	Sterchele et al. (2020b)
Hydrogen turbine investment	€/kW	500 (385)	Sterchele et al. (2020b)
Hydrogen turbine operational cost	€/kWa	2.5% of investment	Sterchele et al. (2020b)
Hydrogen turbine lifetime	a	40	Sterchele et al. (2020b)

.

4. Clustering

First, the results of the cluster analysis are presented, followed by a detailed description of the selected regions.

4.1. Results of the Cluster Analysis

In the correlation analysis, eight independent indicators were determined from the 25 indicators shown in Table 5. The indicators used in the cluster analysis are shown in

Table 7. Indicators for the cluster analysis.

Input Parameter	Description	Unit
population_s	Population	Number per km2
dem_elec_Ind_s	Electricity consumption— industrial sector	MWh/km2
cap_WEA_s	Installed capacity—wind	MW/km2
cap_PV_s	Installed capacity—PV	MW/km2
cap_Bio_s	Installed capacity—bioenergy	MW/km2
cap_other_s	Installed capacity—other	MW/km2
pot_cap_WEA_s	Wind potential	MW/km2
pot_cap_PVarea_s	PV potential	MW/km2

.

Figure 5 shows the solution space for the choice of the number of clusters. Using the elbow criteria, the number of clusters is between five and eleven (red lines in Figure 5). According to the test of Mojena, the optimal number of clusters is 15 (black line in Figure 5).

However, it is also important that the differences between the final clusters can be described verbally. After further analyses of the data and the different numbers of clusters, the five-cluster solution was selected for the following analysis. It should be noted that the selected municipality-level spatial resolution is of a higher level of differentiation than in comparable cluster analyses [47]. Figure 6 shows the municipalities in NWM, with different colors representing the clusters of the five-cluster solution.

Cluster 1 is, by far, the cluster with the most municipalities (58), including Zierow, a rural region with no wind power and low PV capacity, as representative municipality. Cluster 2 and 3 contain 12 and 14 regions, respectively. Cluster 2 is represented by Schildetal, a rural municipality with high wind-power capacity. Cluster 3 is represented by Grevesmühlen, a comparable urban municipality with some industrial demand, mean wind-power capacity and high PV (potential). There are two clusters with only one municipality each. Both regions are clearly differentiated from the other clusters. Cluster 5, the municipality Wismar, is clearly differentiated, as it is the only larger city with high demand in NWM. Cluster 4, Selmsdorf, is characterized by high installed wind-power capacities (like other rural regions) but with some electricity demand from the industrial sector.

The two single-region clusters could be regarded as outliers. To test whether the cluster result were stable, we repeated the cluster analysis without these two municipalities. In this case, the other clusters remained comparable to and homogeneous with the first result. This outcome of the cluster analysis corresponds with the results reported by Weinand et al. (2019) [47]. In their study, the NWM region was also considered as a rather homogeneous region with only small differences between municipalities, supporting our choice of five representative regions.

4.2. Description of the Representative Municipalities

For further analysis, the municipalities shown in

Table 8. Characteristics of the analyzed municipalities.

Municipality	Cluster	Population (Number)	Electricity Demand Industry (GWh/a)	Installed Capacity (MW)		Additional Potential (MW)
				Wind	PV	Wind	PV
Grevesmühlen	3	10,440	20.6	9.2	4.8	0	21.7
Schildetal	2	753	0.7	32.2	0.5	40.6	2.7
Selmsdorf	4	2906	18.4	27.5	2	15.5	8.7
Wismar	5	42,992	172.6	0	15.1	0	59.1
Zierow	1	791	1.1	0	0.4	0	3.5

were considered. These are the representative municipalities of the five clusters (for more details see Supplementary Materials).

Grevesmühlen and Wismar are two rather urban regions or regions with urban structures. In contrast, Schildetal, Selmsdorf and Zierow are rural regions with fewer inhabitants. Wismar has a comparatively high demand for electrical energy in the industrial sector. Grevesmühlen and Selmsdorf have comparatively average demand for electrical energy in the industrial sector, and Schildetal and Zierow have no noteworthy demand from industry. Compared with the rural region of Schildetal, there are neither existing nor potential capacities for wind power in Zierow.

In the following section, essential parameters for modelling the decentralized energy systems are presented. Existing capacity and potential for wind is higher in rural municipalities, such as Schildetal or Selmsdorf. In contrast, the installed capacity and potential for PV is higher in urban areas, such Wismar and Grevesmühlen. The rural municipality of Zierow is an exception, as wind power is non-existent, and the existing and potential capacities of solar power are small. The shares of installed and potential capacities of renewable energies in the municipalities are shown in Figure 7.

The limitation for imports from the transmission grid and the estimated regional grid capacity are shown in

Table 9. Assumed capacities between transmission grid and the municipality.

Municipality	Maximal Load (MW)	Grid Capacity (MW)
Gevesmühlen	7.37	11.79
Schildetal	24.59	39.34
Selmsdorf	21.00	33.60
Wismar	22.77	36.44
Zierow	0.38	0.6

. As described in Section 3.2, the maximal load in a municipality was used to assume the local transmission-grid capacity. In three of the five municipalities, installed RE drives grid capacities. In the city of Wismar, the load of electricity consumption is the critical factor that determines grid capacity. This is also the case in the rural municipality of Zierow; however, due to low electricity consumption, the required grid capacity is, by far, the lowest.

The hydrogen demand in the representative municipalities was calculated based on the information provided in Section 3.1. In urban areas, such as Wismar and Grevesmühlen, with more demand from the industrial sector or the building sector, H₂ demand is higher than in rural municipalities (see Figure 8).

5. Results

In the first step, the results (for details see Supplementary Materials) of the operation mode were presented for the status quo and the year 2050. In the second step, the results of variation of the considered parameters were analyzed.

5.1. Operation Modes of Power to Gas

The operation mode differs between regions and the investigated years. Changes are observed in all the used key performance indicators.

5.1.1. Operation Modes in the Status Quo

In case of the status quo, the realized capacity of electrolysis does not significantly exceed the minimum capacity of 1 MW (see

Table 10. Regional results for electrolysis in the status quo.

Community	Electrolysis Capacity [MW]	Full-Load Hours [h]	H2 Production [MWh]	Share of Used Residual Load [%]
Grevesmühlen	1	5823	5823	72.1
Schildetal	2.5	7109	17,712	47.3
Selmsdorf	1.7	6875	11,354	41.0
Wismar	1	7770	6770	-
Zierow	1 [0.48] *	3576	1715	100.0

* Due to limited electricity supply, it is not possible to use the whole capacity.

). Only the regions of Selmsdorf and Schildetal, with the highest wind-power capacities, have a slightly higher capacity. The full-load hours are close to 6000 h or significantly higher. The municipality of Zierow shows a clear difference, the whole available capacity cannot be used. This is caused by the low grid capacity, in combination with low installed solar-power potential. The complete electrolysis capacity cannot be supplied by sufficient electricity.

Another aspect is that the entire negative residual load is not consumed by electrolysis in all municipalities. This effect can be explained by the option to export electricity for wholesale market prices. Prices for RE electricity are often higher than the price for H₂.

Concerning the operation mode, the following can be concluded for the status quo: High full-load hours suggest that continuous production of H₂ seems most attractive under current conditions. The main driver is the possible export of H₂, as local demand is very low. However, an electrolysis operation does not appear to be sufficiently attractive, as only two regions have slightly more than 1 MW of capacity installed, and one region does not even use the whole 1 MW capacity. This region is Zierow, where 100% of the negative RL is used. Thus, is seems that under current conditions, a flexible operation mode can only be realized if grid capacity is low, and the electrolyzer depends on hours with cheap local RE surplus.

5.1.2. Operation Modes in 2050

The situation changes for the year 2050. In all regions, except Zierow, installed capacity is considerably greater than in the status quo (see

Table 11. Regional results for electrolysis in the year 2050.

Community	Electrolysis Capacity [MW]	Full-Load Hours [h]	H2 Production [MWh]	Share of Used Residual Load [%]
Grevesmühlen	6.4	5629	36,181	67.4
Schildetal	12.7	840	17,712	12.1
Selmsdorf	3.2	5.697	11.354	26.9
Wismar	26.6	7167	190,898	99.8
Zierow	1.29	1494	1924	61.3

). Differences are shown in full-load hours and the produced amount of H₂. The increase in electrolysis capacity is caused by the increase in regional demand for H₂. Larger peaks of RE production also contribute to higher installed capacity.

In 2050, local H₂ demand has a greater impact on the operation mode. In regions with high H₂ demand in 2050, production is drastically higher than in the status quo (see Wismar and Grevesmühlen). More rural regions show an equal or even lower H₂ production in 2050. In addition, the share of the used negative residual load is significantly lower in these regions.

The operation mode is still represented by constant production in three out of five regions. Only the rural region of Schildetal shows another operation mode and is the region with the second largest installed capacity (12.7 MW) but by far the lowest full-load hours (840). The operation is flexible in that most of the electricity (94.3%) for electrolysis is provided by the negative residual load of local RE sources. In 2050, the municipality of Zierow is still limited by low RE potential and low grid capacity. Thus, the local electricity supply limits the operation of electrolysis and causes a rather flexible operation mode, as hydrogen is produced by PV at times with local supply peaks.

5.1.3. LCOH₂ in the Status Quo and the Year 2050

A more detailed look at LCOH₂ (see

Table 12. LCOH₂ in the status quo and the year 2050.

Community	LCOH2 Status Quo [€/MWhH2]	LCOH2 2050 [€/MWhH2]
Grevesmühlen	38.57	52.33
Schildetal	23.52	90.23
Selmsdorf	23.99	24.79
Wismar	55.63	66.78
Zierow	108.58	74.99

) shows differences depending on the electricity and H₂ demand. In regions with high electricity and H₂ demand, PtG is used for local H₂ production, and parts of the negative RL are absorbed by the electrolyzer. Operation is based on H₂ demand and is less influenced by electricity prices. Local production is not always the cost-optimal solution, and H₂ imports are more relevant. On the other hand, if local electricity demand can be completely covered by RE, cheaper H₂ production is possible. This is only the case in Selmsdorf and Schildetal, where H₂ demand and export possibilities can be completely covered by local RE. This is demonstrated by the lower LCOH₂ compared to the average electricity price (Section 3.1). If, in contrast, the local demand for H₂ is very low, the possibility of short-term export is limited, and high load peaks occur. PtG is only used to minimize costs in the event of negative or low electricity-exchange prices. In extreme cases, this results in the entire electrolysis capacity not being used, as is the case in Zierow, which is why LCOH₂ values are particularly high there.

In the Selmsdorf region, demand is low, and increased electricity prices in 2050 lead to a preferential export of RE surpluses. In this region, LCOH₂ values are lower. Simultaneously, it is the only region with LCOH₂ values that are lower than the average electricity price. Likewise, there is no use for H₂ in Schildetal and Zierow due to a lack of local H₂ demand and limited export potential. Therefore, electrolysis is used exclusively for total system-cost minimization when electricity prices are negative or low. In the case of Schildetal, this operation mode results in only a few full-load hours, which is why the local LCOH₂ values increase to 90.23 €/MWh_H2. The decrease in the case of Zierow can be explained by the fact that in 2050, the whole capacity of electrolysis is used, and no overcapacity exists.

5.2. Variations of the Parameters

The results of the base scenario show that a constant production is the best operation mode in most cases. In addition, under current conditions, there seems to be no incentive to build large H₂ capacities.

5.2.1. Local H₂ Demand

Variation of the (local) H₂ demand indicates a change in the operation mode. Electrolysis is used with a higher number of full-load hours if the LCOH₂ is below the import price of H₂. More full-load hours also cause a reduction in LCOH₂. This is driven by the increased use of local RE generation instead of exporting the electricity. It should be noted that this would reduce the yields of local RE generation, but electrolysis can be operated more profitably. Consequently, export opportunities, which act as demand drivers, have a particular impact.

5.2.2. H₂ Price Subsidization

This leads to another relevant driver of the operating mode: the achievable price for H₂. An increase in the assumed sales price of 52 €/MW_H2 in the status quo and 24 €/MW_H2 in 2050 increases the incentive for the model to produce H₂. Its influence depends on the original LCOH₂ level compared to the original H₂ and electricity wholesale price. As long as the profit margin for exchange sales is greater than the difference between LCOH₂ and H₂ price, the model favors selling the electricity. In regions with already low LCOH₂, these costs will only start to decrease when it becomes more attractive to purchase local renewable electricity for PtG. In regions with high LCOH₂, the H₂ price must at least compensate the electricity price. This is reasoned by the fact that before this threshold is reached, the sale of H₂ results in losses.

The exact H₂ price at which the penny-switching effect occurs varies from region to region due to varying local LCOH₂. An increase in H₂ price of 50% increases the operating time of the electrolyzer and, because of longer operating hours, LCOH₂ decreases, on average, by 42.9%.

5.2.3. Electricity Price Components

It should be noted that electricity consumption in the status quo and in 2050 was initially modelled without regard to the variation in electricity price components. Thus, quasi-ideal conditions were assumed, and the variation focused on the impact of assuming higher electricity prices. For example, grid charges for importing electricity increase the incentives to use local RE electricity. This is the case in all regions, as the complete local negative residual load is not consumed by PtG. However, the effect only occurs if local RE electricity is not charged with electricity price components. In this case, the LCOH₂ declines. However, if local RE cannot fully replace previously imported electricity, LCOH₂ increases, as it cannot offset the higher electricity-import costs.

5.2.4. Electricity Price Variation

Fluctuations in exchange electricity prices are also a relevant variable influencing the operation and the function of PtG. This is particularly evident in low-demand but high-supply communities, where PtG is increasingly used as an arbitrage device in response to price signals. Consequently, underlying electricity prices and prominent fluctuations are central to the operation and local LCOH₂.

5.2.5. Electricity Grid

The capacity of the electricity grid was the last influencing factor that we investigated. The capacities for exchange with the transmission grid have an influence on the local use of the RE sources and thus on the utilization of electrolysis and LCOH₂. It was found that depending on regional renewable capacities and energy demand within the energy system, either electricity import or electricity export is decisive for the change in LCOH₂.

In regions such as Zierow, limited grid capacity causes high LCOH₂, as local RE capacities are rather scarce. Thus, electrolysis cannot be supplied with sufficiently cheap electricity from the wholesale market, and the investment costs are not justified by high-capacity utilization. In the case of a 50% higher grid capacity, LCOH₂ drops to 78.81 €/MWh in the status quo. The decline to 72.99 €/MWh in 2050 is rather small, since local RE capacity in 2050 is almost sufficient to supply electrolysis with cheap electricity.

In contrast, Schildetal, with a large surplus of renewable electricity, is affected by export restriction of the grid. The restriction of export possibility leads to an increased use of renewable electricity in local electrolysis. Consumption of cheap electricity and increased full-load hours lead to a drop in LCOH₂ and can improve the economic efficiency of the plant. However, this effect is counteracted by increased capacity of electrolysis and higher investment costs in the case of increased availability of cheap energy. Thus, in Schildetal, with a reduction in export capacity by 50% of regional electricity consumption, the installed PtG capacity increases by a factor of 6.4 in the status quo and 3.5 in the year 2050. Thus, LCOH₂ doubles in the status quo to 56.99 €/MWh due to the higher investment cost and falls to almost one-third of the original cost (33.54 €/MWh) in 2050, when the renewable electricity surplus is significantly higher.

In regions with low renewable generation and high demand, such as Grevesmühlen, restriction on the import side causes this change. Regarding regions with high electricity demand, limitation of electricity purchase is primarily used to meet local electricity demand and is not available to optimize the operation of electrolysis. Consequently, increased H₂ imports are necessary to meet local H₂ demand, and the local use of electrolysis becomes unattractive. In the status quo, the capacity of the electrolysis is not changed, and maintains a minimal capacity of 1 MW. In the year 2050, a reduction in import capacity by 50% leads to a capacity of 63.7% of the original capacity. Nevertheless, LCOH₂ is not much affected and increases in the status quo by 7% to 41.3 €/MWh and decline significantly in the year 2050, by 17.5% to 34.91 €/MWh.

6. Discussion

The discussion interprets the results of the decentralized energy systems for the five exemplary regions. First, we look at the operation modes. Secondly, we focus on the influences of the key performance indicators. Thirdly, we discuss limitations regarding the chosen method.

6.1. Operation Modes of Power to Gas

The prevailing mode is constant H₂ production. In 2050, all but one region exhibit constant H₂ production. This effect is mainly driven by H₂ demand, with regional H₂ demand increasing in the future. In regions such as Wismar and Grevesmühlen, with the highest demand for H₂, the dimensioning of the electrolyzer is larger, although operation requires additional and more expensive electricity imports. The running costs increase, and the investment costs of PtG are higher. This consequence cannot be fully compensated for by increased utilization, which is why LCOH₂ is above average wholesale electricity price in both scenarios for Wismar and Grevesmühlen. In Zierow, limited RE and grid capacities drive the dimensioning of the electrolyzer and storage. Additionally, the local sales potential and export opportunities for H₂ are low, and the profits are higher when most of the renewable electricity is sold.

The constant operation mode can be also be evaluated based on the key performance indicator of LCOH₂. In the status quo, LCOH₂ is comparably low in all regions. The higher cost in Wismar (55.63 €/MWh_H2) can be explained by low RE capacities and considerable import of electricity. The LCOH₂ in 2050 also supports the assumption of a constant H₂ production in most of the regions.

Only in Schildetal we can observe a flexible operation of PtG due to high LCOH₂, in combination with low full-load hours and a small share of used negative RL. Additionally, the key performance indicator of full-load hours only indicates a flexible operation mode in the region Schildetal in 2050. Thus, the impact of local RE generation and electricity prices, which should cause a rather flexible operation mode, is low. This can be explained by the fact that local negative residual load is not completely consumed by local electrolysis capacity, even if it is available at zero cost. This conclusion is also supported by the key performance indicator, share of negative RL. On the one hand, this is due to the revenue opportunities associated with exporting surplus RE electricity, especially when exchange electricity prices are high. On the other hand, in the case of very high load peaks (e.g., in Schildetal), investments for the expansion of electrolysis capacity and other components cannot be covered by the possible income.

6.2. Variations of the Parameters

Nevertheless, the future energy system of Germany needs flexibility options. PtG is supposed to be one of them; however, the results suggest that the existing incentives do not promote it realization in regional energy systems. In both investigated years, just one region shows a flexible operation mode. Variation of the regulatory, market and technical framework in Section 4.2 indicated that specific changes in parameters promote a flexible constant operation mode.

Firstly, increasing H₂ demand and subsidized H₂ price promote constant production, as they foster the importance of H₂. Nevertheless, both options promote the profitability of PtG, as they support an increase in full-load hours, thereby reducing LCOH₂.

Secondly, a change in electricity price components is suitable to support a flexible operation mode. This is shown by a variation of constant price components, as well as variation of electricity prices. Higher price components, such as grid-usage fees, support the use of local RE and, if set right (e.g., in a flexible way), they can lead to the use of local RE peaks for electrolysis instead of exporting the power or ending RE power production. Thus, an adaptation of a grid-fee design may help to realize the necessary flexibility for energy systems at a local level. Higher price peaks on the electricity market work in a similar way. A wider spread of power prices offers an arbitrage option in response to price signals. Thus, price peaks (positive and negative) can be seen as a possibility to encourage a flexible operation mode.

Thirdly, the capacity of the electricity grid has a greater impact on LCOH₂ than on the operation mode. Grid-capacity restrictions tend to use more local RE and foster constant production. The capacity of the electrolyzer is increased to absorb more power. It is worth mentioning that grid capacity is not likely to be reduced. However, the analysis can be seen in the context of constant grid capacity with increasing RE generation. A prerequisite for the use of local RE in PtG is that the power can be used at low cost and without grid-usage fees. The latter assumption was made based on the analysis of increasing electricity price components. In the case of low local RE cost for PtG and high RE generation, LCOH₂ decreases. This contributes to better economic viability of PtG.

Finally, the question of profitable use cases is addressed. Green H₂ from PtG competes with H₂ from fossil sources. The main source of H₂ is currently steam reformation from methane at production prices between 2 and 3 €/kg [87], with prices at fuel stations at 9.5 €/kg, including taxes and levies (see Section 3.1). Given these facts, H₂ for PtG is still too expensive, as our LCOH₂ does not include further costs, such as transportation or storage. In the future, PtG could create a profitable application when the price of carbon leads to more expensive fossil H₂. Additionally, the analysis shows that competition with local RE generation also has an impact. PtG becomes more profitable if cheap and locally existing RE are used. However, if sales via the stock exchange are more attractive, this option will not be realized. Analysis of the electricity price components shows that changes in the regulatory framework could improve the economic situation in the future by decreasing LCOH₂.

6.3. Limitations and Further Research

One limitation of the presented model concerns LCOH₂. Real, local LCOH₂ differs from that used in the model due to the chosen system boundaries and assumptions. However, the calculated LCOH₂ is within the range or below that repored in other publications [88,89], and differences can be explained by our modeling approach, as we did not include all the cost components, and local electricity can be used at a variable cost of zero [59,90]. For 2050, the results are better than the best-case scenarios, which is due to increasing import-electricity prices, as electricity price is the main driver of LCOH₂ in the case of high full-load hours [24]. Local RE generation is not consumed by PtG, as electricity export is more attractive economically. Thus, for the comparison of municipalities between time points, the key performance indicator of LCOH₂ is sufficient to provide information about which municipalities have more favorable conditions for H₂ production, as well as the resulting operating mode.

Variable RE costs represent a further limitation, as they were modelled at marginal costs of zero. In the model, local RE production can be sold (exported) at any time at the market price. In reality, operators of electrolyzers compete with the wholesale market and do not get local RE for free. This could be covered by implementing local electricity prices for every region.

Another approach could be to consider the levelized cost of electricity (LCOE) [91] for local RE. This approach would change the picture provided in Table 12 (see

Table 13. LCOH₂ in the status quo and the year 2050, including LCOE for RE.

Community	LCOH2 Status Quo [€/MWhH2]	LCOH2 2050 [€/MWhH2]
Grevesmühlen	83.79	75.32
Schildetal	82.85	141.57
Selmsdorf	85.20	71.22
Wismar	65.88	76.36
Zierow	118.32	108.84

). In this case, LCOH₂ increases in all regions in both cases: the status quo and the year 2050. Especially in regions with high RE consumption by the electrolyzer, the effect is crucial. Higher LCOH₂ would affect the competitiveness of PtG in a negative way.

However, the current electricity market is not designed to secure the earning of RE with respect to investment cost and to fund the LCOE. In a future energy system, this must be ensured to guarantee the necessary expansion of RE [92,93]. Future defossilized energy systems must cover this issue either on the electricity side, e.g., with further feed-in tariffs, or on the hydrogen side by an adequate consideration for the design of future hydrogen markets.

Furthermore, the model used fixed RE capacities. The future scenario did not consider any investment decision for the expansion of RE, as our focus was not on optimal expansion from the perspective of plant operators but on the operating strategy of electrolysis under given conditions. The mutual influence could be considered in a common investment model.

Lastly, the data availability for H₂ demand and transport capacities had an influence on the model results. There is a lack of appropriately resolved data at the local level, so assumptions were made. The influence was considered by analyzing the results with varying parameters in Section 4.2. Nevertheless, specific data with good spatial and temporal resolution would lead to more accurate results. The addressed points can be improved in further studies and must be considered in the interpretation of the presented results.

7. Conclusions

The energy transition and the shift towards a climate-neutral energy system is a major challenge. Sector-coupling technologies, such as PtG, can play a curial role regarding the transfer of climate-neutral energy produced by sources such as solar or wind power to consumption sectors. In decentralized energy systems, PtG can play a crucial role by offering flexibility to the system and acting as a greenhouse-gas-neutral energy carrier. In this paper, the two possible operating modes of flexibility or constant H₂ production, as well as factors influencing the operation mode, were investigated

Investigation of the operation modes in decentralized energy systems was realized by a linear optimization model in OEMOF for the status quo and the year 2050 in five municipalities. The results were analyzed by using key performance indicators: levelized cost of hydrogen, full-load hours, used share of negative RL and installed capacity. Furthermore, changes in the regulatory, market and technical framework were investigated to observe possible changes in the operation mode of H₂ production. To assess the impact of regulatory, market and technical framework, we focused on H₂ prices, electricity prices and price components, as well as H₂ demand and the demand affecting electricity exchange capacities.

In the status quo, the dominant operating mode is a constant H₂ production, and in one region, a flexible operation mode is observed. The operation mode is driven by H₂ demand, and low electrolysis capacities are sufficient to fulfill the demand. Higher demand leads to lower LCOH₂, as the demand incentives higher full-load hours. This only applies if local RE provide sufficient power at low cost. In this context, electrolysis can also enable the local use of parts of the negative RL. The flexible operation mode is only realized if both electricity grid and local RE capacities are too small but local negative RL is available and can be used to reduce system costs.

In the year 2050, demand is higher, and this results in higher installed electrolysis capacities. The prevailing operation mode is constant H₂ production, but under certain conditions, a flexible operation mode is observed. This is particularly the case in regions with relatively high RE capacities and low demand for electricity and H₂.

The results show that determinants of the operation modes and LCOH₂ in the different energy systems are local H₂ demand, regional RE capacities, short-term export opportunities and arbitrage opportunities that exist through electricity wholesale prices. Depending on their interdependencies and the opposing income from electricity exports, local RE surpluses are used for local H₂ production.

However, the LCOH₂ is not competitive under current market conditions, even in regions that have better conditions for economic H₂ production. Possible business models would probably not be viable under current market conditions, including all the neglected price components or cheaper H₂ production pathways. H₂ production costs are, in most cases, above the expected market prices for H₂, especially for the year 2050. To enable profitable business models, locally generated renewable electricity must be available cheaply, or the price of H₂ needs to be sufficiently high. The investigated variation of the regulatory, market and technical framework conditions shows that subsidized H₂ price and changes in electricity price components, such as more flexible grid-usage fees, are good options. In this context, flexible grid-usage fees would also contribute to more flexibility, and we assume that adjustments towards flexible grid-usage fees are more likely to be realized than fundamental changes in the design of the electricity market. In contrast, the new market for green hydrogen should offer more opportunities to design it accordingly in order to realize defossilization, secure RE funding and support business cases for H₂. Thus, the market will play a crucial role in the success of PtG. In order to realize profitable use cases in the future and to enable the potential of PtG to contribute to the energy transition, it is important to stimulate the identified parameters accordingly to fully utilize the local H₂ potential of the regions. This requires a suitable incentive structure and the implementation of a new market design for green hydrogen. However, further research is needed to determine a suitable market design.