Assessing the Carbon Intensity of e-fuels Production in European Countries: A Temporal Analysis

Abstract

The transport sector heavily relies on the use of fossil fuels, which are causing major environmental concerns. Solutions relying on the direct or indirect use of electricity through e-fuel production are emerging to power the transport sector. To ensure environmental benefits are achieved over this transition, an accurate estimation of the impact of the use of electricity is needed. This requires a high temporal resolution to capture the high variability of electricity. This paper presents a previously unseen temporal analysis of the carbon intensity of e-fuels using grid electricity in countries that are members of the European Network of Transmission System Operators (ENTSO-E). It also provides an estimation of the potential load factor for producing low-carbon e-fuels according to the European Union legislative framework. This was achieved by building on top of the existing EcoDynElec tool to develop EcoDynElec_xr, a python tool enabling—with an hourly time resolution—the calculation, visualisation, and analysis of the historical time-series of electricity mixing from the ENTSO-E. The results highlight that, in 2023, very few European countries were reaching low carbon intensity for electricity that enables the use of grid electricity for the production of green electrolytic hydrogen. The methodological assumptions consider the consumption of the electricity mix instead of the production mix, and the considered time step is of paramount importance and drastically impacts the potential load factor of green hydrogen production. The developed tools are released under an open-source license to ensure transparency, result reproducibility, and reuse regarding newer data for other territories or for other purposes.

1. Introduction

In 2022, more than 95% of the final energy used in the transport sector was sourced from fossil fuels globally [1], illustrating the heavy reliance of the transport sector on fossil fuels. While this share slightly decreased in the last decades as a result of policies promoting the use of renewable energy [2], fuel consumption in transport followed continuous growth, with the exception of a brief period during the COVID crisis [3]. As a consequence, the direct greenhouse gas (GHG) emissions of the transport sector increased from 5.0 GtCO₂eq in 1990 to 8.7 GtCO₂eq in 2019 [4]. This accounts for 23% of global energy-related greenhouse gas emissions and more than a third when including the impacts of transport infrastructure [4]. In the European Union (EU), the transportation sector is currently the only sector with a growing trend in GHG emissions [5]. Increasing the use of renewable and low-carbon energy sources in transport is thus becoming essential to achieve EU carbon neutrality goals.

In addition to GHG emissions, fossil fuels are also responsible for environmental impacts, from their extraction to their final use in internal combustion engines with local pollutant emissions [6]. Solutions relying on the direct or indirect use of electricity are being investigated and are emerging as alternatives to power the transport sector. Electricity is expected to play an increasing role in the future of the transport sector [3,4] due to the following:

• The direct use of electricity in battery electric vehicles, electric trains, or other trolley vehicles;
• The indirect use of electricity to produce fuels that require a high amount of electricity as input, such as hydrogen or hydrogen-derived fuel (the so-called e-fuels or electro-fuels).

Transport solutions that directly use electricity are already being largely deployed and expanding, while solutions relying on an indirect use of electricity through the use of e-fuels are at a lower commercial readiness level [7]. Several e-fuels or electro-fuel pathways, also referred to as power-to-liquid (PtL), are candidates for the production of synthetic fuels using hydrogen produced through water electrolysis [8,9,10]. Despite some technological and nontechnological barriers to deployment, such electro-fuel pathways are promising options for fuel synthesis, enabling the achievement of high GHG emission savings when using low-carbon electricity [7,11]; these appear to be essential in terms of meeting the EU climate targets of the transport sector [12,13].

In order to facilitate the emergence of a hydrogen economy and a low-carbon hydrogen industrial ecosystem, the EU has developed a comprehensive legislative framework through the Renewable Energy Directive 2018/2001 (RED) [14] that promotes the use of renewable energy sources across the EU economy. The Renewable Energy Directive (EU) 2023/2413 (the so-called RED III) [15] has increased the level of ambition in terms of the use of renewable energy in the EU and the use of renewables in transport by 2030. Member States have the possibility to choose between either a binding target of a 14.5% reduction in GHG intensity in transport from the use of renewables or a binding share of at least 29% of renewables in the final consumption of energy in the transport sector. The Directive also sets a binding combined subtarget of 5.5% for the use of advanced biofuels and renewable fuels of non-biological origin (RFNBO), i.e., renewable hydrogen or derived synthetic fuels to be used in the transport sector. Within this target, there is a minimum requirement of a share of 1% of RFNBOs in the energy supplied to the transport sector in 2030.

Directive 2018/2001 (RED II) [14] defined RFNBO as fuels produced from renewable energy and requires greenhouse gas emissions savings from their use of at least 70% compared to the fossil fuel comparator of 94 gCO₂eq/MJ. Two delegated regulations on RFNBOs set out the detailed rules for their production and the methodology for calculating emissions savings: EU 2023/1184 [16] and EU 2023/1185 [17]. The corresponding framework, represented in Figure 1, introduces the three essential pillars aiming at ensuring that hydrogen is effectively produced from low-carbon electricity:

• Additionality: The aim is to stimulate the deployment of new renewable electricity generation capacities and avoid competition with other uses of renewable electricity;
• Temporal correlation: The aim is that RFNBO production takes place at times when electricity is produced
• Geographical correlation: The aim is that RFNBO production takes place in grid areas where renewable electricity is produced.

The criteria for RFNBO production make sense from a GHG emissions perspective by using additional electricity when and where it is available. The regulation EU 2023/1184 proposed a flexible application of the criterion to temporal correlation in the initial phase. From 1 January 2030, the temporal correlation condition requires the production of the RFNBO during the same 1-h period as the renewable electricity production. Until 1 January 2030, the temporal correlation condition shall be considered complied with if RFNBO is produced during the same calendar month as the renewable electricity. Regulation EU 2023/1185 sets the methodology to assess and calculate the life cycle GHG emissions savings.

Figure 1. Legislative European framework for low-carbon hydrogen production [18].

Figure 1. Legislative European framework for low-carbon hydrogen production [18].

These two regulations are part of a wider context where hydrogen and derived fuels are expected to become one of the main pillars of the EU decarbonisation strategy as a clean solution for mobility, power generation, and industrial applications, according to the European Green Deal (EGD) [19]. In order to promote renewable hydrogen production, the European Commission adopted “A hydrogen strategy for a climate-neutral Europe” in 2020 [20] by setting targets for the installation of at least 6 GW of electrolysers powered by renewables in the EU by 2024 and 40 GW by 2030. The EU also recently proposed the REPowerEU plan [21] for rapidly reducing the dependence of the EU on fossil fuels by accelerating the green energy transition. RePower EU considers renewable hydrogen as key to replacing natural gas, coal, and oil in hard-to-decarbonise industries and transport. Thus, hydrogen can be used to replace fossil-based hydrogen for transport and industrial processes and new industrial products, such as green fertilisers and steel. REPowerEU proposed a target of producing 10 Mt of hydrogen using electrolysis within the EU, which we expect to be equivalent to 60 to 70 GW of electrolysis capacity that has to be installed in the EU by 2030; another 10 Mt of renewable hydrogen will have to be imported by the EU in 2030, according to REPower EU. Hydrogen can be used to produce liquid synthetic kerosene or other synthetic fuels to decarbonise the aviation and maritime sectors, in addition to other uses.

While technological solutions are emerging, the European legislative framework requires the saving of at least 70% of GHG emissions to avoid a simple shift of transport-related pollution from fossil-based internal combustion engines to power plants generating the electricity used for e-fuel production. Even though electric, hydrogen or e-fuel vehicles are sometimes promoted as zero-emission or climate-neutral by considering only tailpipe emissions, the production of both the vehicles and the energy used to power them is not burden-free [22,23]. This is why performing a life cycle analysis (LCA) and a systemic analysis [24,25,26] is necessary to ensure that the environmental impact is effectively reduced and not simply shifted to another stage of the system or another environmental impact [22].

Many LCAs of electric, hydrogen, or e-fueled vehicles already exist. Reviews and meta-analyses based on those numerous publications highlight how determinant the impact of the electricity used is on the environmental footprint [27,28,29]. Alongside this, other papers have focused on assessing the impact of the electricity consumed and not produced. Few papers provide details on the carbon intensity of used electricity, such as Scarlat et al. [30], who considered annual data for electricity production and trade and provided the carbon intensities of electricity produced and consumed as well. The study from Tranberg et al. [31] makes use of hourly data for carbon accounting and confirms that the carbon intensity of electricity consumed in a country can significantly differ from the carbon intensity of the produced electricity. While studies regularly neglect greenhouse gas emissions associated with e-fuels when produced from renewable electricity [32], other studies [30] consider combustion emissions, upstream and downstream emissions, and also the emissions from the manufacture of equipment, e.g., including GHG emissions from renewable electricity. Thus, there is an essential need for an integrated method that is consistent with the existing European legislative framework; this would provide an accurate estimation of the environmental impact of the electricity used to power the transport sector (at the right time resolution) to correctly inform decision-making [33,34].

The purpose of this paper is to propose an integrated method using the associated Python tool to assess the carbon intensity of electricity used, directly or indirectly, in the transportation sector. This method relies on the data from the European Network of Transmission System Operators (ENTSO-E) to provide a comprehensive and retrospective analysis of the historical carbon intensity of produced and consumed electricity with an hourly time resolution for each member country of ENTSO-E. The consumed electricity is calculated using the exchange data between countries and the tracking algorithm presented by Tranberg et al. [31] and implemented in the Ecodynelec tool [35]. By building on top of this, we developed EcoDynElec_xr, which contains some improvements over its previous version to make its use easier, less prone to mistakes, and more efficient. Section 2 details the step-by-step approach followed, relying on the improved tool to calculate a time-series of the carbon footprint of electricity. The obtained time-series of carbon intensities was then used to propose a retrospective statistical analysis of the potential load factor of hydrogen (or derived fuels) under the European Union’s legislative framework. Those results are presented and discussed for e-hydrogen, as it is a necessary intermediary step for other e-fuels that can be derived from it; however, the discussion can be generalised to the e-fuels mentioned in Section 3. This results section discusses the importance of key methodological assumptions, such as considering the production or consumption of electricity mix and the chosen time scale for the analysis. Finally, Section 4 outlines the conditions for producing low-carbon e-fuels and their impact on economic performance. It emphasizes the importance of considering the outcomes of this work in future studies related to the environmental footprint of e-fuels or LCAs of other electro-intensive products that require accurate and dynamic assessments of electricity impacts.

2. Material & Methods

2.1. Methodological Framework Overview

Figure 2 represents the approach to assess the greenhouse gas emissions from e-fuel production. The e-fuel can be e-H₂, the simplest form of e-fuel, or any synthetic fuel derived from it. As the last steps are optional, dashed arrows were used in Figure 2. When an e-fuel is derived from H₂, the GHG emissions are calculated as the sum of emissions from H₂ production and from all other inputs, such as the energy required for the final transformation or emissions from carbon capture. However, the scope of this paper covers H₂ production only, while the GHG emissions from all other potential e-fuels can be calculated by expanding the boundaries to cover all subsequent conversion steps. Although further transformation steps induce additional emissions and losses, the effects of H₂ production can be generalised to other e-fuels since the carbon intensity of electricity has the highest impact on their production.

The GHG emissions of hydrogen are calculated from the GHG emissions associated with electricity used for hydrogen production. The amount of electricity consumed is directly calculated considering the efficiency of water electrolysis. The thermodynamic limit for dissociating water at ambient temperature through electrolysis is around 40 kWh/kgH₂. For low-temperature electrolysis, around 50 to 55 kWh of electricity is needed to produce 1 kg of hydrogen. The typical electrolysis efficiency ranges from about 60% for proton exchange membrane (PEM) electrolysers to 65 % for alkaline electrolysers [36,37,38]. Considering the potential future improvements in terms of efficiency, this value of 65% is considered by default in the later calculations in this paper. One can easily adjust this parameter in the code and generate all the graphs considering other electrolyser efficiency values.

Previous studies highlighted the fact that the carbon intensity of H₂ is dominated by the impact of electricity used [39] and the impact of the equipment and infrastructure can be neglected; only emissions related to the use of electricity for H₂ production are considered in the scope of this work. In this paper, the carbon intensity of electricity is calculated as the electricity mix multiplied by the emission factors of each electricity source. While emission factors exclude emissions related to equipment and infrastructure, they include both combustion emissions and upstream emissions related to fuel supply (extraction, transport, and refining for fossil fuels and enrichment for nuclear fuel).

The calculations were made for the production mix of each country member of ENTSO-E, listed in Appendix A, at 1-h resolution using the ENTSO-E production data, and then for the tracked consumption mix calculated based on the production and exchange of ENTSO-E data. The most recent ENTSO-E data available (for the year 2023) were used. By using the tracking algorithm [31] implemented in the EcoDynElec tool [35], it is possible to obtain the tracked electricity consumption mix. The algorithm was re-implemented in the developed EcoDynElec_xr tool. With the assumption that H₂ electrolysis uses electricity at a high-voltage level, as expected for a Gigawatt scale project, the calculations considered only the losses after transport, based on the grid loss data sourced from the EUROSTAT dataset NRG_CB_E. Those transport losses referred to the difference between the fed-in (generation) electricity at a high-voltage grid and the electricity delivered to distributors. The tool also enables the inclusion of the grid distribution losses on top of the transport losses in the case where an electrolyser is connected to a lower voltage grid. The use of the EUROSTAT data [40] enables the calculation of average annual grid losses, which is a simplification in the calculation of the hourly carbon intensity of electricity. The emission factors per unit of electricity produced were calculated by considering power plant efficiency for each country and each type of fuel input based on the EUROSTAT dataset NRG_BAL_C; this determines the amount of primary energy needed to produce electricity. The most recent available EUROSTAT data, corresponding to the year 2022, were used. This enables the calculation of the average annual plant efficiency, which is assumed to be constant over the year, which is a simplification in the calculations. Indeed, the efficiency varies over the year, especially for plants generating both heat and power if the heat is not or less used during the warm season [41]. Following the approach presented in Figure 2, it is possible to calculate the carbon intensity with a time resolution of 1 h. For some graphics, the data were later resampled at a daily time resolution, only for display reasons. Moreover, in order to test the sensitivity of the assessed carbon intensity, the calculations were also carried out by down-sampling the hourly data for the calculation of carbon intensity at a lower time resolution.

2.2. Methodological Implementation and Execution

This subsection presents the steps involved in the calculation of the time-series of the carbon intensity of electricity produced and consumed. Some graphics, corresponding only to the input data, are presented in this section to help the reader visualise and understand the steps of the method. The graphics corresponding to the output data of this work are presented and discussed in the Results section. The structure of this section is strongly linked to that of the Jupyter Notebook, which the reader can refer to to reproduce the work or use for other purposes. The whole code is available on github. For this work, a new python library, named EcoDynElec_xr was developed and implemented in the tracking algorithm behind [31]. It relies on the previous library EcoDynElec [35] to download the ENTSO-E data. Even though this EcoDynElec tool was developed in the context of electricity consumption in buildings, the tool is not sector-specific, and its application to the transport sector is relevant. The EcoDynElec_xr tool enables the calculation and analysis of the historical time-series of electricity mix published by the European Network of Transmission System Operators (ENTSO-E). By making use of this tool, we downloaded, calculated, and visualised the time-series of carbon footprint for both the electricity produced and consumed in each country member of the ENTSO-E with an hourly time resolution. Then, a new data structure was used to make the code more efficient and readable by using the xarray datasets structure and their explicit, labelled dimensions. Such a structure is fully adapted to manipulate the multidimensional data with explicit names for all dimensions. The tracking mix algorithm involved in the calculation was also completely re-implemented using the pandas dataframe structure to allow for data manipulation using explicit index and column names instead of numbered indexes that are more prone to mistakes and are less readable. Some grouping and visualsation features were added to allow for data quality checks and results analysis. Finally, a part of the previous code specific to Switzerland was removed to keep ENSTO-E as the only source of electricity data and avoid mixing data of heterogeneous quality.

2.2.1. Downloading the ENTSO-E Data

The first step consists of downloading the ENTSO-E data. The European Union Member States must provide information related to electricity generation, load, transmission, and balancing for publication through the ENTSO-E Transparency Platform, according to Regulation (EU) 543/2013 on submission and the publication of data in electricity markets [42]. The data on the production, trade, and consumption of electricity were downloaded for the year 2023. This task can be carried out manually or by using the EcoDynElec library, as was carried out in the Jupyter Notebook, where more information can be found; the user needs to create an account to access the ENTSO-E platform and get a token to download the data. The username and password have to be indicated in the settings.py file. As this download can take a significant amount of time, if data already exists, the tool asks for confirmation before downloading them again. The user simply needs to reply “yes” or “y” when confirmation is asked for.

2.2.2. Formatting the Downloaded Data

When downloaded, the data are contained in large csv files. These files, corresponding to a 1-month period each, are weighted at around 160 Mb for the generation data and 80 Mb for the exchange data. One of the reasons why these files are so big is because they also contain data at the bidding zone level that are not used. The EcoDynElec_xr tool has some functions to extract, format, and save the data in a nicely structured and labelled dataset. Once formatted, the data can be loaded and accessed much faster than from the csv. The manipulation of the data is also facilitated and the user can filter data by country, energy type, and datetime, which are the three dimensions for the generation of the data. A similar function exists for the exchange data, and data can be filtered based on exporting countries, importing countries, and datetime.

2.2.3. Checking Data Quality & Missing Values

Figure 3 represents heat maps of the ratio of missing data in the previously downloaded and formatted data. The table of country codes and corresponding country names can be found in Appendix A. With the exception of production data for the United Kingdom, which is no longer a member of the European Union, the overall quality of the data is really good, with only a few missing data. In many situations, some data seem to be missing but they simply do not exist. An example is the exchange data between the United Kingdom (GB) and Denmark (DK). Indeed, when looking at the time-series, we see some data for the end of December but no data before that date. In this case, the reason is that GB and DK inaugurated their first electric connection, the so-called Viking Link. The same situation happens for France and offshore wind production. In France, the first offshore wind power plant entered into operation in June 2023, so there is no data before this date. More graphics can be found in the Jupyter Notebook. By default, if there are missing values for less than 2 h, the values are filled with linearly interpolated values.

2.2.4. Visualising Data

For exemplification, Figure 4 shows the generation and exchange time-series for Italy. To represent these data in a visible way, the data have been resampled to the daily mean values before being plotted. The type of electricity was grouped into a reduced number of categories: wind, solar, hydro, other renewable, nuclear, fossil, and other, as the initial dataset has up to 20 energy categories. Those grouping functions were added to the EcoDynElec_xr tool in order to make graphs that are more readable by limiting the number of categories. However, the user is able to generate the same plot with the ungrouped data or can change categories by modifying the grouping dictionary in the energy_grouping.py file. Figure 4 also shows the exchange data for Italy with the same resampling at the daily time step. Looking at the scale of the y-axis, the data highlight that Italy imports a significant amount of electricity mainly from its northern neighbours and is marginally exporting electricity.

At this stage, we can combine the generation and exchange data together to obtain the consumption data. However, we cannot yet assess the nature of the electricity imported or exported, so we cannot know the nature of the electricity mix consumed. To do so, we need to run the tracking algorithm as explained below, but it is first necessary to identify countries that are directly exchanging electricity between themselves.

Figure 4. Generation and exchange data for Italy.

Figure 4. Generation and exchange data for Italy.

2.2.5. Neighbours Identification

Before running the tracking algorithm, we need information about the neighbouring countries that exchange electricity. The previous EcoDynElec tool contained a file that provided (for each country member of the ENTSO-E) the list of its electrical neighbours. The new EcoDynElec_xr tool automatically identifies the electrical neighbours by using the ENTSO-E exchange data. We were thus able to notice, for instance, that the first electrical connection between GB and DK was recently inaugurated. Such features enable the tool to deal with the creation of new electrical connections without the need to update the file with the list of its electrical neighbours. In 2023, ENTSO-E covered 35 countries that were exchanging electricity between themselves, but some countries also exchanged electricity with countries outside the ENTSO-E area. Figure 5 presents the share of electricity exchanged with external neighbours as relating to the total electricity production of ENTSO-E members. Exports and imports from neighbours of ENTSO-E countries represent around half a per cent of ENTSO-E production. Moreover, we can see that the exchanges are rather balanced, as the net import values are even lower, with an order of magnitude of one per thousand (0.1%) only. In the calculation, the external countries are considered as one country producing its electricity from gas, as a previous study [30] showed that the exporting countries to ENTSO-E have an average carbon intensity that is not very different from the carbon intensity of gas electricity. This assumption can be modified, but the impact is globally negligible. It can, at most, have a local impact on the carbon intensity of small neighbouring countries heavily relying on import; a detailed analysis shows that only Georgia, Lithuania, Montenegro, Kosovo, and Latvia have more than 5% of their consumption sourced from outside ENTSO-E.

2.2.6. Tracking the Electricity Mix

The mathematical essence of the program lies in this step. From the existing EcoDynElec tool, the mathematics are strictly unchanged, but the code was re-implemented to use explicit names to access and manipulate data instead of index numbers that are less readable and more prone to mistakes. Solving this mathematical problem of countries, directly and indirectly, importing electricity consists of inverting the Leontieff matrix of an input-output problem [31,35]. This is carried out using the existing function from the scipy library instead of using the previously implemented algorithm. More information can be found in the existing EcoDynElec documentation [43]. This function, called track_mix, has four parameters that allow the possibility of considering net production (or not), net exchange between countries (or not), grid loss, and time step. By setting net production to True, which is the default value, the tool subtracts the electricity consumed by the production plant from its gross production. Setting net_exchange to False means that export and import are both considered at a given moment in time. If the user prefers net_trade, he/she should set this parameter to True. A last parameter enables the user to ignore the grid losses consider the losses occurring on the transport grid only, or use the distribution losses as well. The values for losses are sourced from the EUROSTAT dataset NRB_CB_E, which are automatically downloaded by running the notebook Eurostat.ipynb. It is important to note that once calculated, we can estimate the nature of the electricity that is consumed in all countries (if it has been produced with fossil, nuclear, or renewable energy, depending on grouping); we can also know the countries of production for each of the considered time steps. It is also important to note that those values result from a model that may differ from reality, which is obviously more complex. This model aims to give the best possible and reasonable estimates of the reality, but the reality may differ, considering the impedance of electric connections between countries. The underlying assumption of the model is that when given the case of country A exchanging electricity with a country B, there is no preference over the nature of the electricity exchanged; country A cannot claim specifically exporting electricity from coal power plants exclusively, even if the power plant is closer to the border. Exchanges correspond to the mix of countries A and B. Their mix also takes into account the fact that country B may have, itself, exchanged electricity with another country (C), which may itself exchange with another country (D), and so on. The second aspect is that there is no differentiation in the mix exported between neighbours; if country A exports electricity to countries B and C, the electricity mix exported is identical. Country A cannot claim to export hydroelectricity to country B and coal electricity to country C, even if the power plants may be closer or have more direct electrical connections. Those assumptions (even though reasonable) may differ from the reality, where it could be, in theory, possible that a power plant is turned on to export production to a determined country or that the limited transport capacity creates a situation where the mix exported to different countries would depend on a plant’s location, etc. Thus, the obtained results can be considered as the best possible estimate of the reality, but the user must be aware that those values are the outcome of a model with the specific assumptions mentioned above.

2.2.7. Calculating the Impact of Electricity Mix

At this stage of the calculation, we can track which type of electricity is consumed in a country and in which country the electricity has been produced. To get an estimation of the carbon intensity, we now simply need to multiply the electricity mix by the emission factor, which depends on the type of electricity used and the country in which it is produced. Such multiplication is easy to conduct when using the xarray datastructure; the only condition is to ensure that the name of countries and the origin of electricity types are identical, as well as the dimension names. Having conversion efficiency depending on the type of electricity is obviously necessary, but having it also depend on the country of production is very relevant when significant differences between countries occur. In the context of the European legislative framework, emission factors have to be calculated following a specific method that accounts for the combustion emissions and for the upstream emissions. For the calculations, we have the emission factors per type of fuel input, but these values correspond to the emissions per unit of primary energy and not per final electricity generated. To convert the tracked electricity data into the primary energy, we need to calculate the efficiency per type of electricity and per country. In a standard LCA approach, emission factors could be obtained using LCA databases. Even if the present work focuses on carbon intensity, it is possible to do the calculation for other impact categories relating, for instance, to the emissions of particulate matter. Those efficiency values are calculated using the Eurostat API (application programming interface) [44]. The corresponding code can be found in the notebook Eurostat.ipynb. There are specific assumptions for allocation; the allocation of the primary energy for electricity production and for heat in cogeneration was made by considering alternative heat production with average efficiencies of 85%. More details can be found in [30]. The data for the year 2022 from Eurostat statistics were used as the most recent available data, whereas the ENTSO-E data were already available for the year 2023. However, we expect this to have a very low impact on the results since the changes in electricity conversion efficiency are low at the scale of a country from one year to another. The efficiency values were calculated from the IEA data for the countries having no available data for each type of power plant and fuel input. The preference is given to Eurostat data over IEA data, as they offer a higher level of detail and are more rapidly available. The calculated efficiencies are represented in Figure 6, which shows high variations due to diverse conditions in different countries for different energy sources.

Once those efficiencies (by fuel input and by country) are calculated, we can easily calculate the amount of primary energy consumed. These primary energy data then have to be multiplied by the emission factors and extracted from the Commission Delegated Regulation (EU) 2023/1185 [17], as represented in Figure 7, to obtain the emissions resulting from electricity production for each source. Those emission factors include the emissions of the following greenhouse gases: carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O). These emissions are converted into carbon dioxide equivalent emissions using the Global Warming Potential (GWP) relative to CO₂ over the 100-year time horizon, as recommended in the IPCC Fourth Assessment Report [4].

2.2.8. Statistical Analysis Under a Regulatory Limit

Once multiplied by the emission factor, we obtained a time-series of the carbon intensity of electricity produced or consumed in every country member of the ENTSO-E. In order to assess the number of hours an electrolyser could operate in each country to produce low-carbon hydrogen, as is specific in case 2b in Figure 1, we simply test when the time-series values are below this limit of 18.3 gCO₂eq/MJ (66.0 g CO₂eq/kWh) of electricity that the electrolyser uses. This value corresponds to the threshold of reducing the GHG intensity of e-fuel by 70% in comparison to the 94 g CO₂eq/MJ fuel from the fossil fuel comparator when assuming an efficiency for the electrolyser of 65% (Threshold value calculated based on that assumed efficiency is indicated with an asterix (*) on Figures 8 and 9). The results obtained are presented and discussed in the next section.

3. Results & Discussion

This section discusses the results obtained by following the method previously described. The temporal variability in carbon intensity for electricity was calculated using a 1-h time-series and then the carbon intensity of hydrogen produced through water electrolysis using grid electricity. For further analysis, the carbon intensity of hydrogen has been calculated at different time steps, in line with the timing operation of electrolyser and electricity production or consumption, to assess the impact on the carbon intensity of e-hydrogen.

3.1. Time-Series of the Electricity Mix and Its Carbon Footprint

The time-series of the electricity mix and its carbon intensity has been produced for all countries that are members of the ENTSO-E. This covers 35 countries, including all Member States of the European Union, except Malta and Cyprus, which do not belong to or provide data for the ENTSO-E. Figure 8 presents the electricity dashboard for Italy, Germany, and Austria; these were chosen to illustrate interesting and dissimilar cases. For all the other countries, the figures are available at the following github repository. For display purposes only, the data represented in the figures were resampled using the daily mean for better readability of the graph, but all data were calculated with an hourly resolution. The various subplots present the following:

• The production mix of electricity in the country: The energy categories were grouped to limit the number of energy categories to represent. Except for the grouping, those data directly come from the ENTSO-E data and do not involve any modelling or calculation;
• The consumption mix of electricity: Those data are the outcome of the tracking algorithm. They correspond to the output of a model assuming no preference in direction and energy type in exchanges between countries (i.e., countries are always exchanging electricity corresponding to their domestic mix);
• The power balance is also represented;
• The carbon intensity, per unit of electricity, of the production electricity mix. To ease the readability, the same colours are used as for the above graph;
• The carbon intensity of the consumption electricity mix, per unit of electricity;
• A comparison of the carbon intensity of the production and consumption electricity mix, as well as the limit for the used grid electricity to produce low-carbon hydrogen. For the last three graphs, the left axis gives the values in gCO₂eq/kWh while the right axis gives the values in gCO₂eq/MJ of electricity.

Figure 8 shows that, for instance, Italy consumes nuclear electricity, whereas Italy does not produce nuclear electricity. This is because Italy imports a significant amount of electricity from France and Switzerland, where nuclear is significant in the national electricity mix. The graph for Italy shows that all over the year, Italy is importing electricity to satisfy a demand that is higher than the national production. In August, the electricity demand reduced and imports were at the lowest level. When looking at the previous graph, it also corresponds to the moment when the consumption of nuclear electricity was also at the lowest level.

Figure 8 also shows that Germany’s consumption mix includes a share of nuclear and hydroelectricity, which contribute to the reduction in its carbon intensity. It is also interesting to note that there is seasonality in electricity exchange. Germany exports electricity to neighbours globally in winter when electricity production relies more on fossil fuels and has a higher carbon intensity. It also globally imports electricity during the summer, which reduces the use of fossil fuel plants. These aspects explain why the consumption mix in Germany has a slightly lower carbon intensity than its production mix. Even if the share of renewables is high during some periods, the remaining share of electricity produced from fossil fuels prevents it from reaching the 18 gCO₂eq/MJ limit.

Austria was also chosen for illustration purposes since, in comparison to Italy or Germany, its electricity mix sometimes reaches the threshold. Figure 8 shows that, in the absence of production from fossil fuels, the carbon intensity of the electricity production mix in Austria is very low. However, in winter, when the electricity consumption increases, Austria becomes a net importer of electricity with high-carbon intensity. During those periods, this imported consumption of fossil electricity leads to a significant increase in carbon intensity. As a result, most of the time, the carbon intensity of electricity production is below the threshold, but it is above the threshold most of the time when considering electricity consumption.

Annual averaged carbon intensities are presented for all country members of ENTSO-E in Section 3.2. The statistics on the number of hours below the threshold are presented and discussed for all countries in Section 3.3. The carbon intensity of the consumption mix is calculated and displayed for various time steps. The effect of the calculation time step is discussed in Section 3.4.

Figure 8. Dashboard of electricity mix and carbon intensities for selected cases.

Figure 8. Dashboard of electricity mix and carbon intensities for selected cases.

3.2. Annual Averaged Carbon Intensity of the Electricity Mix and e-H₂

Figure 9A presents the annual averaged carbon intensity of electricity by countries. This annual average is calculated after calculating the corresponding carbon intensity for each hour of the year using the method previously described. The black dashed line corresponds to the threshold for the carbon intensity of electricity used, enabling 70% of the GHG emissions of e-hydrogen to be saved, assuming a 65% electrolysis efficiency, compared to the fossil fuel comparator value represented by a red dashed line in Figure 9B. Countries are sorted by the carbon intensity of their consumption mix. Only a few countries have an average annual carbon intensity satisfying this criterion. We can also see that some countries, such as Denmark, have a consumption mix cleaner than the production mix due to the import of cleaner electricity. On the other hand, some countries have a consumption mix with a higher carbon intensity than the production mix.

Figure 9B presents the carbon intensity of e-hydrogen production using grid electricity (at high voltage) in each ENTSO-E country. The blue dashed line represents the value calculated for the production of e-H₂ in the EU27, with the exemption of Cyprus and Malta, which do not provide data to ENTSO-E. The red and blue dashed lines are extremely close, meaning that producing e-H₂ with the electricity mix of the EU27 does not save any GHG emissions in 2023 compared to the fossil fuel comparator. Out of the 34 countries that are members of ENTSO-E, only five countries would have been able to produce, when averaged over the whole year, e-H₂ complying with the 70% GHG emission threshold, 13 would still reach a lower carbon intensity than the fossil fuel comparator without reaching the 70% savings, and 16 would produce e-H₂ with a higher carbon intensity than fossil fuels. Such a result is expected as long as a significant share of electricity is produced by burning fossil fuels.

3.3. Statistical Analysis of Potential Load Factor of Low-Carbon H₂

As presented in the introduction, the European legislative framework sets a threshold for the use of grid electricity for hydrogen production through water electrolysis, achieving a GHG emission reduction of at least 70%. Assuming an electrolysis efficiency of 65%, the corresponding limit for the carbon intensity of electricity used is 18.3 gCO₂eq/MJ of electricity. Relying on the previously presented carbon intensity time-series, we retrospectively calculated the potential load factor that could be achieved while respecting this limit for the year 2023 (e.g., how many hours can the electrolysis operate in a year to produce low-carbon hydrogen), which is shown in Figure 10. To reveal the impact of the produced versus consumed electricity, this figure shows the impacts of the type of electricity considered in the calculations on the load factor. We first see that many countries are never able to reach the limit. For the countries able to reach this limit, many are reaching it for a limited number of hours per year. There are, finally, a low number of countries that have a carbon intensity of the production mix below this limit for an important number of hours a year. However, when looking at the consumption mix and not the production mix, the potential load factor is reduced for most of them; Luxembourg is an extreme case because the carbon intensity of the production mix, mainly relying on hydropower, is often below the limit, whereas the consumption mix, involving the significant import of fossil-based electricity, is rarely lower. Denmark and Finland are the only countries performing better with their consumption mix due to imports of low-carbon electricity from Sweden and Norway.

3.4. Influence of the Considered Time Resolution

For the exemplification of the influence of the considered time resolution, Figure 11 presents a time-series for the carbon intensity of e-H₂ in Austria. This case study was selected as a clear and relevant case for illustration purposes, as its carbon intensity of e-H₂ is sometimes below the threshold and sometimes above it. The time-series are calculated and represented for a daily, weekly, and monthly time resolution. It highlights the importance of time resolution when comparing the average value to the threshold, as it can be observed that when the monthly value is just above the threshold, such as in August, some weekly or daily values can be lower than the threshold, and the contrary can be observed in June.

Figure 12 presents the potential load factor of low-carbon hydrogen production considering both the production and consumption mix for the different time steps, from annual values to hourly values. When considering an annual time step, the load factor is 0 or 1, depending on the annual average and if it is above the limit or not. The load factor is slightly affected by the considered time step. Indeed, as exemplified for Austria in Figure 11, even if the monthly carbon intensity is above the limit for 1 month, it is possible that within this month, there are some weeks, days, or hours for which the carbon intensity is lower than the limit. On the other hand, there might be months with a monthly intensity below the limit, and within this month, there might be some weeks, days, or hours over the limit. The larger the time step is, the higher the risk of including periods of time where the electricity is sourced from high-carbon intensity sources. Having a time step as low as an hourly time step strongly limits this risk. A larger time step may facilitate the running of an e-fuel power plant, but Figure 12 also shows that, for most countries, when looking at the consumption mix, the shorter the time step is, the higher the potential load factor is. This is due to the fact that every single hour below the limit can be used for plant operation.

4. Conclusions & Perspective

The major contribution of this paper is the quantification of the impact of electricity production (at high time resolution) and the effect of electricity exchange between different countries (at the time it occurs) on the carbon intensity of electricity and e-fuels. This is the most relevant aspect in determining whether any e-fuels produced are indeed “green” and to what extent they really contribute to the decarbonisation goals for transport. The results highlight that e-fuels, either hydrogen or fuels derived from it, can be produced using low-carbon intensity. The condition is to use electricity only when its carbon intensity is low enough, which significantly affects the potential operational load factor of an electrolysis unit. A reduced load factor means a reduced number of hours when low-carbon fuels can be produced, reduced operation power, or a combination of both. Thus, with the per unit of power capacity of electrolysis installed, this limits both the amount of hydrogen that can be produced and makes reaching economic viability more challenging. From that perspective, additional sources of income other than the fuel production itself may help. For instance, e-fuels could play a role in balancing the grid as storage, which could provide potential income for the provision of ancillary services to the grid. Indeed, even though combining environmental and economic performance is challenging, this is a necessary step to ensure the sustainable production of e-fuels.

Making use of ENTSO-E data, covering the largest network of national grid operators in the world, we showed that exchanges of electricity between countries have the potential to dramatically change the carbon intensity of the national electricity mix. Some countries may have a carbon intensity of their production mix low enough to produce low-carbon e-fuels under the European legislative framework, but they may depend at some moments on electricity imports with a higher carbon intensity, which would result in an increase in the carbon intensity of their consumption mix above the limit. Another key parameter for the carbon intensity of e-fuels, whether hydrogen or derived fuels, is the considered temporal resolution. When considering an annual or monthly average of the carbon intensity, the country might satisfy (or not) the criteria of having an electricity carbon intensity low enough to produce low-carbon e-fuel over that period. However, with a finer time resolution, within that period, the carbon intensity of electricity may be low enough to be used, even if, on average, over the whole period, it is not, or the opposite. The risk of having a time step that is not fine enough is using electricity when the production is not sourced from low-carbon sources.

Regarding the limits of this study, it is important to mention that massive electrolytic hydrogen production would require a massive amount of electricity and would involve major changes to the electricity mix to produce the expected low-carbon e-fuels. It would then become necessary to apply the proposed methods to prospective scenarios. This represents an interesting perspective, but that would require prospective production and exchanging time series between the considered territories, data the authors never found published in the scientific literature. In addition, it is important to remember that emissions related to the manufacture of infrastructures are not considered in the scope of this work; we only cover combustion emissions and upstream emissions related to fuel production. The inclusion of those emissions will necessarily increase the e-hydrogen footprint and lower the load factor below the threshold.

Considering the importance e-hydrogen is expected to play inside the EU or outside for its international trading partners [45], the same methodology might be needed for the calculation of its carbon intensity. The work can be replicated in any other area of the world depending on the availability of data at the same level of detail, in particular, a time series by energy type and electricity exchange data at the right resolution.

Finally, the developed python tool, EcoDynElec_xr, is released under an open-source license to ensure transparency, result reproducibility, reuse on data that are more recent than 2023 (when they will be available), reuse for other geographical areas, or reuse in other applications where a fine and dynamic assessment of electricity impact is needed.