Decarbonization potentials of energy citizen clusters
Coupling machine learning clustering algorithms with energy simulation models to explore the decarbonization potential of strategic energy citizen clusters
Introduction
A successful energy transition constitutes a deeply societal matter that has created the imperative to address its societal implications. In this regard, the European Union outlines policies that envision a central role for citizens in the energy transition and a more decentralized and democratic energy system that could encourage renewable energy production and empower citizens to engage and take responsibility for energy production and consumption. Aiming to realize this vision, the sociopolitical concept of “energy citizenship”, consistently resurfaces in recent scientific and political discourse, due to its potential to bridge the gap between energy transition policies and social participation by emphasizing the role that citizens can play through their engagement and involvement in the energy system, as well as the impact they can have on the European Union’s future energy landscape.
However, current climate change mitigation strategies have predominantly emphasized the technical and infrastructural aspects of the energy transition, often overlooking the critical sociopolitical dimensions that underpin the success of these efforts. This is also reflected in contemporary energy modeling practices and scenario-based research approaches, used at a large extent by policymakers to assess and design policies and regulations to address climate-related challenges, focus primarily on cost optimization and technoeconomic implications of the energy transition, often disregarding social aspects, matters of inclusivity and equity, as well as citizens’ preferences. Omitting such considerations may result in monolithic modeling exercises and oversimplified models that while they produce cost-optimized and computationally neat decarbonization pathways, they lack the necessary depth to inform policymakers about the broader societal dimensions crucial for a sustainable and inclusive energy transition. In this regard, a comprehensive and integrative approach that encompasses not only technological innovations and policy reforms but also profound behavioral changes across society is imperative.
In this study, we integrate energy citizenship trends and patterns into energy modeling to create more comprehensive and socially informed modeling exercises. This allows us to explore the relationship between the various expressions of energy citizenship and their potential impact on decarbonization efforts under a variety of scenarios and external conditions. To this end, we use an ensemble of modeling tools to produce insights regarding the decarbonization potential of energy citizenship at different contexts across the European Union.
With the goal of performing socially informed modeling exercises and producing relevant, actionable insights, we pair the strengths of energy simulation models with results derived from machine learning clustering algorithms. More specifically, through the use of machine learning algorithms, clusters of energy citizens are identified by utilizing data on energy consumption and carbon emissions for specific Member States. This results in the identification of commonalities within the utilized dataset and a subsequent categorization of citizens in distinct groupings. Citizen clusters are then analyzed based on sociodemographic characteristics and cognition-related attributes to produce a set of citizens’ profiles that can be used to draw narratives for energy scenario development.
Leveraging this work, we integrate clustering and profiling results into energy system modeling and simulation tools, by translating the clusters' characteristics and energy behaviors into inputs for the modeling ensemble. This integration quantifies the decarbonization potential of different citizen groups, and demonstrates which clusters are more crucial and responsive towards accelerating decarbonization efforts. Targeting these clusters with tailored policies and interventions can significantly enhance the overall effectiveness and efficiency of decarbonization initiatives.
This study presents outcomes of model applications corresponding to two (2) different geographical contexts and socioeconomic environments, with our aim being, eventually, a sound and robust understanding of the relationship between patterns related to the identified citizen clusters’ energy behaviors and decarbonization pathways.
Utilizing the clustering results on citizens’ energy consumption, future patterns regarding energy consumption and electricity demand as well as relevant historical data, we designed scenarios matching the characteristics of the respective citizen clusters, for Greece and the Netherlands respectively. To simulate scenarios, two (2) models of the ENCLUDE modeling ensemble were employed, namely:
- ATOM (Agent-based Technology adOption Model), developed to simulate the expected effectiveness of policy schemes on technology adoption (e.g., small-scale solar photovoltaics (PV), battery energy storage systems (BESS), heat pumps, electric vehicles, etc.) in the residential sector, for the geographical and socio-economic context under study.
- OSeMOSYS-GR (Open-Source energy MOdeling SYStem for GReece), a country-specific implementation of the OSeMOSYS model generator, which models the characteristics of the Greek power system for the period 2021-2050. Specifically, OSeMOSYS-GR simulates the capacity requirements for, and the electricity generation from fossil-fired sources, like lignite, natural gas, and oil power plants, as well as renewable energy sources (RES), such as hydro, onshore and offshore wind, utility, commercial, and rooftop solar PV, biomass, and geothermal. It also incorporates various flexibility options including pumped hydro, BESS, and hydrogen systems powered by electrolyzers and fuel cells.
More details about the presented model applications can be found in Deliverable 5.5: Report on the decarbonization potential of strategic energy citizen clusters of the ENCLUDE project.
Model application I: Adoption of rooftop solar photovoltaic (PV) systems in the housing sector for different energy citizen clusters in Greece and the Netherlands
Citizens are increasingly becoming individual owners of solar PV assets, thus consuming their own electricity and playing a supportive role in driving the energy transition. Considering that, we use ATOM to evaluate the adoption potential of small-scale solar PV systems in the residential sector in Greece and the Netherlands toward 2030. Specifically, we explore how different policy schemes can empower prosumerism and further citizen adoption of small-scale PV systems for different citizen clusters as well as their decarbonization potential.
Using the clustering data provided by machine learning algorithms as well as historical data, we created scenarios of energy consumption patterns and carbon footprint for citizens of Greece and the Netherlands. To compare and validate findings, we also established baseline scenarios based on data retrieved from the Greek NECP and the Eurostat Energy Balances.
The modeled housing electricity demand scenarios for Greece and the Netherlands by 2030 are depicted in the following figures.
Using ATOM, we explore how FiT, net metering, and net metering with BESS support schemes can empower prosumerism and further citizen adoption of small-scale PV systems as well as estimate the related decarbonization potential, answering two (2) research questions (RQs).
RQ1: How could different policy schemes empower prosumerism and further citizen adoption of small-scale PV systems in Greece and the Netherlands by 2030 for the different citizen clusters that have been defined?
Explore more case specific details regarding the effectiveness of policy schemes on solar PV adoption in Greece and the Netherlands in the following subsections.
Greece
Currently, there is a residential solar PV installed capacity of approximately 435 MW, while the target of the Greek National Energy Climate Plan by 2030, is 1 GW of installed residential solar PV capacity. Our results show that the the application of a FiT scheme with a fixed price equal to 87 €/MWh has the greatest impact on the adoption of small-scale solar PV systems in the residential sector. However, it would not be enough to achieve the national target by 2030.
This finding highlights the insufficiency of supportive policy schemes when applied individually. Therefore, it is possible that a combination of different policy schemes needs to be applied in order to further enhance the adoption of small-scale solar PV systems in the Greek residential sector.
The uncertainty that governs the decision-making process of Greek citizens when adopting solar PV systems appears to be very low. This outcome shows the perception of Greek citizens toward PV investments, and more specifically, it highlights that on average, Greek citizens have a somewhat clear perception about the profitability potential of the investment. This becomes more obvious under the net metering and net metering with BESS policy schemes where the uncertainty bounds are generally lower when compared to the uncertainty bounds under the FiT scheme. This highlights the need for higher and long-term fixed prices to make citizens feel more certain about the profitability of their investments.
Explore in detail the results for solar PV capacity additions by 2030 based on the three (3) identified clusters and policy schemes in Greece in the figures below.
The Netherlands
So far in the Netherlands the installed capacity of residential solar PV systems reaches approximately 10,106 MW. For this application, we assumed that the national target for the installed capacity of residential solar PV systems by 2030 equals the technical potential for residential solar PV systems, i.e., approximately 14,000 MW.
Similarly to Greece, applying a FiT scheme, in this case with a fixed price equal to 109.7 €/MWh appears to have the greatest impact on the potential adoption of small-scale solar PV systems in the Netherlands. This policy scheme is capable of not only driving capacity additions to meeting the national target of 14 GW by 2030 but could also lead to overshooting the target by a large margin. According to modeling results, the projected PV capacity additions are approximately 2.5 GW over the 2030 target. However, the 2030 target could possibly be achieved even if the Netherlands continues to apply only the existing net metering scheme for small-scale solar PV systems. This means that in contrast to the Greek case study, there is no need to combine policy schemes to achieve the 2030 target.
The uncertainty gaps for the PV capacity additions of the Dutch citizen clusters are calculated to be higher when compared to the Greek case study, with “Cluster 3” having the largest uncertainty among all the clusters. This highlights that despite the higher PV capacity additions achieved by “Cluster 3”, citizens who belong to this cluster have not a “clear” perspective about the profitability of the investment. This can be also attributed to the lower annual electricity consumption of “Cluster 3” citizens which results to higher ambiguity about the potential economic benefits of prosumerism. Another factor that affects the final PV capacity additions of the clusters is their population. “Cluster 3” has the larger population since around 44% of the total population belongs to this cluster.
Explore in detail the results for solar PV capacity additions by 2030 based on the three (3) identified clusters and policy schemes in the Netherlands in the figures below.
RQ2: What are the carbon emissions that could be avoided through the further empowering of prosumerism in Greece and the Netherlands by 2030 for the different citizen clusters that have been defined?
Greece
Our analysis also provides estimations of the decarbonization potential of adopting small-scale solar PV systems by 2030 by considering the carbon intensity of the power sector. As expected, the FiT policy scheme shows the greatest decarbonization potential since the carbon emissions that could be avoided through this scheme is more than double when compared with the emissions that could be avoided under the net metering and net metering with BESS policy schemes.
While the FiT policy scheme has the greatest potential for achieving decarbonization in the Greek residential sector by 2030, the potential combination of more supportive policy schemes that incentivize prosumers to invest in small-scale solar PV systems could increase the decarbonization potential of the residential sector.
From a cluster-based perspective, we see that “Cluster 1” results to higher capacity additions and avoided carbon emissions compared to the other two (2) Greek citizen clusters and thus contributes the most to small-scale solar PV adoption and decarbonization of the residential sector. This is mainly attributed to its higher population and lower electricity demand when compared to the other two (2) clusters.
Take a closer look at the projected decarbonization potential of the Greek residential sector though the installation of small-scale PV systems for the three (3) policy schemes in the figures below.
The Netherlands
Our analysis results in similar to the Greek case study findings with regards to the estimations of the avoided CO2 emissions from the projected adoption of small-scale solar PV systems under the three (3) policy schemes for the Netherlands. More specifically, the most effective policy scheme seems to be the FiT, followed by Net Metering and Net Metering with BESS. Specifically, prosumerism in the Dutch residential sector is capable to reduce the total carbon emissions of the country by 1% on average.
When it comes to the Dutch citizen clusters, citizens that belong to “Cluster 3” appear to have the most prominent role in the decarbonization of the residential sector since they contribute the most to the adoption of PV systems and thus to the reduction of carbon emissions. “Cluster 3” citizens are characterized by the lowest electricity demand among all clusters under analysis while their population is the largest.
Explore in more detail the projected decarbonization potential of the Greek residential sector though the installation of small-scale PV systems for the three (3) policy schemes in the figures below.
Model application II: Power sector capacity buildout based on electricity consumption patterns of different energy citizen clusters in the housing and mobility sectors in Greece
Citizens’ adoption rate of electrification technologies and their energy use patterns will significantly impact the future energy system’s design in terms of capacity and flexibility requirements. Drawing from the need for better understanding of the potential impacts wielded by citizen groups with different profiles in terms of their sociodemographic characteristics, energy consumption levels, etc., we integrated different citizen groups’ energy consumption patterns into OSeMOSYS-GR. This allowed us to assess their electricity consumption patterns alongside other aspects like technical feasibility and cost to explore decarbonization pathways in the Greek power system by 2050.
Using the clustering data from machine learning algorithms, we generated electricity demand projections for each one of the three clusters identified within the Greek residential and mobility sectors. These scenarios allowed for a more granular understanding of sector-specific electricity demands, facilitating a more informed approach to future energy planning and sectoral decarbonization efforts. The scenario design for this case study also includes a business-as-usual (“BAU”) scenario, serving as a reference point against which the emissions and electricity demand trajectories of the scenarios based on the different energy citizen clusters are compared. The “BAU” scenario for this case study was constructed using projections by 2050 from the revised draft Greek NECP.
The modeled housing and mobility electricity demand scenarios for Greece by 2050 are illustrated in the following figures.
Using OSeMOSYS-GR, we explore capacity additions to decarbonize the power sector and the resulting electricity mix, the capital investments per technology, and the CO2 footprint in the Greek power sector by 2050, answering two (2) RQs:
RQ3: How do different citizen groups’ energy consumption patterns affect the capacity and flexibility requirements as well as the resulting electricity mix of decarbonization pathways in the power sector by 2050?
In the “BAU” scenario, Greece’s power sector capacity is expected to increase by about 250%, reaching 106.8 GW by 2050 compared to 2025. The “BAU” scenario shows that RES expansion is mainly driven by variable renewable energy (VRE) sources, including utility-scale, commercial, and rooftop solar PV systems as well as onshore and offshore wind installations. It results in a total RES capacity of 63.8 GW by 2050. Supporting this growth, flexible technologies like BESS, pumped hydro, and electrolyzers also play a key role in stabilizing the system. Solar commercial PV and offshore wind see the greatest growth, each adding about 14 GW by 2050, while electrolyzers and fuel cells are projected to surge with 15.1 GW and 5.6 GW additions, respectively. Utility-scale and rooftop BESS as well as pumped hydro, are also expected to rise substantially, reaching a combined total of 17.1 GW by 2050. Specifically, capacity additions of 12.7 GW of utility-scale BESS, 1.1 GW of rooftop BESS, and 2.3 GW of pumped hydro are expected to take place by 2050 compared to 2025. Fossil fuels like lignite and oil are set to be completely phased out by 2028 and 2040, although natural gas will still contribute approximately 5.2 GW, supported by Carbon Capture and Storage (CCS) technology.
In the “Cluster 1” scenario, which represents citizens with the lowest energy expenditures in both the housing and mobility sectors and higher climate awareness, total capacity additions by 2050 are projected to be the smallest, reaching only 94.3 GW. Here, RES capacity is the lowest among all scenarios, totaling 56.6 GW by 2050. The scenario sees stabilization in capacity growth by 2045 due to early adoption of short-term storage solutions like utility BESS, which compensate for a relative underinvestment in longer-duration storage options such as pumped hydro.
In contrast, “Cluster 2” scenario, which is characterized by high energy use in the mobility sector and more neutral climate perceptions, is the only scenario where projected capacity additions surpass the “BAU” scenario’s projections. Total capacity in this scenario is estimated to reach 108.4 GW by 2050, with a particularly large RES expansion (65 GW) to meet the increased electricity demand. This scenario results in the most significant capacity growth across solar commercial PV, offshore wind, electrolyzers, and utility-scale BESS.
“Cluster 3” scenario, which includes citizens with higher energy use in housing, moderate mobility consumption, and positive climate attitudes, falls between the other two in terms of capacity additions. Capacity is expected to expand by 74.2 GW from 2025 to 2050, slightly below the “BAU” scenario. RES installations in this scenario are projected to reach a total capacity of approximately 62.7 GW by 2050, mirroring the growth trends seen in the “BAU” scenario.
The resulting capacity mix of the Greek power sector by 2050 for the baseline scenario and the scenarios informed by the clustering analysis is depicted in the figure below.
Power generation by 2050
By 2050, Greece’s total annual power generation will rely heavily on RES, particularly VRE sources, and hydrogen across all scenarios. In the “BAU” scenario, RES and hydrogen make up around 89.6% of the total generation, led by wind energy. Wind alone accounts for 42.7% of electricity generation, with offshore wind farms producing approximately 178.2 PJ and onshore wind farms contributing around 83.2 PJ. Solar energy also plays a key role, delivering a combined 168.9 PJ from rooftop, utility-scale, and commercial PV, underscoring the critical role of wind and solar in Greece’s energy future.
“Cluster 1” scenario, which has the lowest electricity demand, results in 560.3 PJ of total power generation in 2050. RES and hydrogen constitute 88.7% of the total power generation, with wind energy being the largest contributor, followed closely by solar. This scenario falls short of the 2050 RES penetration target (98.3%) by the largest margin of all scenarios. “Cluster 2” scenario results in the highest total power generation in 2050, driven by the elevated electricity demand, especially in the mobility sector. RES and hydrogen dominate the electricity mix, making up 89.8% of it—the highest share among all scenarios. “Cluster 3” scenario, which reflects moderate energy use in both mobility and housing, results in a total generation close to the “BAU” scenario, reaching 605.1 PJ. In this scenario, 89.5% of the power generation in 2050 comes from RES and hydrogen.
While lignite and oil are phased out by 2028 and 2040, respectively, natural gas remains part of the energy mix in 2050 due to CCS deployment. Natural gas contributes between 8.4% and 9.4% of total generation, with the lowest share in “Cluster 2” scenario and the highest in “Cluster 3” scenario.
RQ4: How do different citizen groups’ energy consumption patterns affect the capital investments and the carbon footprint of decarbonization pathways in the power sector by 2050?
Capital investments by 2050
The energy transition in Greece relies heavily on wind and solar energy, with offshore wind turbines requiring the highest capital investments across all scenarios. In the “BAU” scenario, average annual investments in offshore wind are projected to reach approximately €1.14 billion.
In “Cluster 1” scenario, where citizens exhibit the least energy-intensive behaviors, offshore wind investments are the lowest, namely around €1.02 billion annually. Conversely, “Cluster 2” scenario, with its higher electrification in mobility and housing, requires the greatest average annual investment at €1.16 billion. For “Cluster 3” scenario, closely aligned with NECP consumption trends, necessary offshore wind investments are expected to be €1.12 billion per year.
The “BAU” scenario requires a combined annual investment of €1.65 billion across both onshore and offshore wind, with “Cluster 3” scenario requiring €1.63 billion. “Cluster 1” and “Cluster 2” scenarios require €1.46 billion and €1.69 billion, respectively, with wind investments making up around 39% of total spending across all scenarios.
While “Cluster 3” scenario shows similar electricity demand patterns to the “BAU” scenario, slight differences emerge in total average annual capital investment needs. The “BAU” scenario requires total average annual capital investments of €4.24 billion, compared to €4.17 billion for the “Cluster 3” scenario. In “Cluster 2” scenario, total average annual investments total €4.26 billion, aligning closely with the “BAU” scenario despite the differences in electricity demand patterns. This is largely due to pumped hydro storage investments, which are higher in the “BAU” scenario than in “Cluster 2” scenario.
“Cluster 1” scenario requires the least investment overall, with total average annual investments of €3.71 billion, saving 12.5% compared to the “BAU” scenario. This investment saving reflects the cluster’s larger population, more energy-conscious behaviors, and stronger climate awareness, which together reduce the need for extensive infrastructure development.
The figure below depicts the average annual capital investments that will be needed considering the electricity demand patterns of each scenario towards 2050.
Regarding the carbon footprint of the energy transition in the Greek power sector, “Cluster 1” scenario exhibits the lowest carbon emissions between 2025 and 2050, driven by the environmentally conscious energy consumption patterns of the citizens within this cluster.
Interestingly, despite the higher energy consumption observed in the “Cluster 2” scenario, its total cumulative carbon emissions between 2025 and 2050 are projected to be lower than those in the “BAU” scenario. Citizens in this cluster tend to consume more electricity in the mobility sector and less electricity in the housing sector. However, in the short term (i.e., 2025-2030) when the power generation from gas is at its peak, the amount of electricity required from the mobility sector is much less than the amount required from the housing sector due to the exponential patterns of mobility demand by 2050. Thus, this scenario leads to lower short-term consumption of gas in the short-term and thus a lower total carbon footprint in the power sector.
Regarding “Cluster 3” scenario, its total cumulative carbon emissions between 2025 and 2050 are projected to be the closest to those of the“BAU” scenario among all examined scenarios. This is due to the fact that citizens in this cluster tend to consume electricity with very similar patterns with those of the Greek NECP both in the mobility and the housing sector.
The total annual CO2 footprint in the Greek power sector by 2050 for the baseline scenario and the scenarios informed by the clustering analysis is depicted in the following figure.
More studies by ENCLUDE
You can explore results from more modelling studies by ENCLUDE project in the following links:
- Scenario design for energy citizenship
- Citizen-led transitions towards justice and inclusivity
- Impacts of Collective Energy Initiatives in Europe by 2030
- People-centered national energy systems
- "Power to the People" transitions in EU cities