Energy Economics in European Union Countries—Typological Analysis Using Kohonen Networks

Abstract

Energy is a key resource determining economic and social development. The aim of the study was to identify and analyze patterns in the energy economy of European Union countries in 2019 and 2023 using the self-organizing maps (SOMs) method, which is an artificial intelligence tool. This approach enables unsupervised clustering of countries based on complex data, revealing hidden relationships between energy indicators. Analysis of Eurostat data showed that over the analyzed period, five countries improved their cluster position, one country experienced a decline, and the remaining 21 countries retained their previous assignment. The grouping of the countries in 2023 was more diverse, with a clear concentration of countries with favorable development parameters. The results of the study provide valuable guidance for EU energy policy, supporting the planning of a sustainable transition in the context of challenges such as the COVID-19 pandemic and the war in Ukraine.

1. Introduction

EU countries currently face challenges related to energy transition, rising energy prices, climate change, and the need to achieve climate neutrality. The degree of advancement of a country’s energy economy demonstrates its ability to efficiently manage energy resources, guarantee stable energy supplies, reduce energy costs, and combine economic priorities with environmental protection and sustainable development. Understanding the mechanisms shaping energy markets and their impact on the economies of Member States is crucial for shaping energy and climate policy across the EU.

This study addresses the following research objectives:

• Identification and analysis of patterns in EU countries’ energy economies using an artificial intelligence-based method (SOM).
• Comparison of the energy-economy classifications of EU countries in 2019 and 2023 and assessment of changes in the structure of these classifications.

Achieving the assumed study objectives required answering the following research questions:

• Are there clear patterns in energy economics across EU countries that allow for their classification according to structural similarities and levels of energy management?
• Has the structure of the EU country classifications changed in 2023 compared to 2019, which could indicate varying rates of energy transition across EU countries?
• Do changes in the structure of the EU country classifications reflect a trend toward the levelling out of development differences, with countries with lower levels of energy development moving closer to those with higher levels of this phenomenon?

The study is based on statistical data for 2019 and 2023, obtained from the Eurostat database.

The main motivation for this research is the timeliness of the issue and the increasingly significant link between energy economics and the economies of EU countries. By analyzing various aspects of energy economics, it is possible to assess the effects of energy transformation, investments in renewable energy sources, and changes in energy prices on economic growth, competitiveness, and energy security of countries. In the context of the EU, where all member states strive for an efficient, sustainable, and stable energy sector, EU-wide research is particularly important.

Despite growing interest in the subject of energy economics, there is still a lack of comprehensive comparative studies covering all European Union countries and analyzing changes over time using advanced data analysis methods. Previous studies have focused mainly on the national level and have been based on classical statistical methods, which do not allow for the identification of complex, non-linear relationships between energy and economic indicators. The identified research gap concerns the lack of comparative studies of the energy economies of EU countries that use artificial intelligence methods in a dynamic and multidimensional way. This study addresses this gap by applying the Self-Organizing Map (SOM) method to identify hidden patterns and compare the results of country clustering in 2019 and 2023. The use of SOM, which belongs to the trend of empirical methodologies based on artificial intelligence, enables unsupervised clustering of countries based on multidimensional data, which allows for capturing non-obvious relationships and dynamics of changes in the energy economy. This approach enriches the body of research in the field of energy economics by combining empirical analysis with modern machine learning tools, and the results provide practical guidance for shaping energy policy and supporting sustainable transformation in the European Union. The paper is structured as follows: the introduction presents the authors’ motivation for conducting the study, including the study objectives, research questions, data sources, and methodological approach. Section 2 provides a literature review, focusing on energy economics. Section 3 provides detailed information on the data and methods used. Section 4 provides a detailed discussion of the research findings. The paper concludes with a discussion and concluding remarks.

2. Literature Review

Energy is a fundamental resource necessary for everyday human functioning. It is crucial not only for the needs of individuals and households, but also for the development of society and the economy [1]. Its availability, price, and source of origin are key factors determining this development. The challenges associated with energy policy in Europe are becoming increasingly important, especially in the context of rising energy costs, limited availability of certain fuels, ever-growing needs, and climate change. An important aspect of these challenges is ensuring energy efficiency and security [2] and creating conditions for the achievement of sustainable development goals [3]. Contemporary challenges related to energy production, distribution, and consumption require not only technological solutions, but also in-depth economic analysis. In response to these challenges, energy economics allows for the assessment of the costs, effectiveness, and effects of energy policies on the economy and society.

Energy economics is an interdisciplinary field of science that studies various aspects of energy resources and raw materials, including factors that drive companies, organizations, and consumers to supply, process, transport, and use energy. It also deals with the disposal and processing of used energy waste, the creation of market and regulatory structures, and the analysis of the impact of energy consumption on distribution and the environment. The overarching goal of energy economics is to promote the most economical and efficient use of energy [4,5,6,7]. Like any field of economics, energy economics deals with the fundamental economic issue of allocating limited resources in the economy. Therefore, microeconomic issues of energy supply and demand and macroeconomic issues of investment, financing, and economic linkages with the rest of the economy are an important part of this field. However, the problems facing the energy sector are changing, bringing new issues to the fore [8]. This requires further analysis and research on specific aspects of energy economics or a comprehensive, multi-faceted approach to the issue.

Energy economics covers many broad subfields, which leads researchers to focus their analyses on specific elements of the topic rather than on the entire field.

Table 1. Review of research on selected aspects of energy economics.

Aspect of Energy Economics	Key Observations and Conclusions
Supply, demand, and energy prices and their impact on the economy	Russia’s war in Ukraine and disruptions in trade in raw materials (oil, gas, coal) have contributed significantly to the increase in energy prices in Poland, both through production costs and through increased energy costs for households and businesses. This may weaken Poland’s position relative to countries with more energy-efficient industries or a better energy mix. The publication fills a gap in the comprehensive approach to the relationship between the energy commodities market (including price volatility, mix structure, and energy costs) and the competitiveness of the Polish economy—both in sectoral and international terms, taking into account the forecast perspective [9]. Oil and coal prices have a significant impact on inflation in Poland. Natural gas prices had a limited impact due to lower consumption and mitigation measures. The development of the energy market should be closely monitored in terms of its inflationary potential. The publication fills a research gap consisting in the lack of empirical analyses of the impact of global energy commodity prices (oil, coal, gas) on aggregate inflation in Poland—a country heavily dependent on energy imports [10] and concerning the sectoral transmission of the impact of energy commodity prices on various components of inflation in Poland, and not only on the overall price level [11]. Spikes in gas, coal, fuel, and electricity prices lead to a significant increase in inflation and a slowdown in economic activity. The energy and food sectors are key components of inflationary pressure. The publication fills a gap in knowledge concerning the macroeconomic effects of sharp increases in energy prices in Poland, both in real terms (GDP, consumption) and in price terms (inflation), using advanced modeling tools [12]. Rising energy prices are a significant driver of inflation in EU countries. The transmission and differentiation of the impact of energy price increases on the structure of inflation (and not just the level of inflation) in EU countries has been identified, taking into account the crisis year of 2022 [13]. Periods of crisis (e.g., the energy crisis, the war in Ukraine) increase synchronization between countries in terms of energy price dynamics. The authors supplement scientific achievements with information on how energy price changes between Poland and other EU countries synchronize, which had not been analyzed previously in such a time-frequency perspective [14]. Energy consumption, including renewable energy, affects economic growth. The increase in consumption and production of energy from renewable sources, both on a national scale (Poland) and in individual regions of the country, has contributed to the growth of measures characterizing economic development. The authors supplement scientific achievements with an answer to the question of how the development and use of renewable energy affect economic growth in Poland and its individual regions [15].
Energy poverty	The publication fills a gap in the lack of comparative analyses of energy poverty in the EU that take into account both objective and subjective indicators, showing that the energy situation of households may vary depending on the measurement perspective adopted. The use of objective and subjective indicators shows that the perception of energy poverty often exceeds what is perceived based on hard data. National policies should consider both perspectives to more accurately identify needs and design interventions [16]. There is a tendency to reduce the differences in energy poverty levels between EU countries, but the process is slow and uneven. The authors used multidimensional indicators for 27 EU countries and examined whether countries with higher energy poverty are closing the gap with better-off countries and what factors influence the pace of this convergence [17]. Transfer policies (benefits) and improving home energy efficiency—both types of interventions significantly reduce energy poverty. The effect varies depending on the country’s circumstances and the extent of poverty. Combinations of both approaches are most effective. The publication fills a research gap in empirical comparative assessments of the effectiveness of policies to combat energy poverty [18]. The article fills a gap in comparisons of energy poverty reduction measures in the context of energy transition between countries with different energy systems, using Poland and Sweden as examples. In Poland—finding a balance between alleviating energy poverty and switching to cleaner energy sources, given the country’s high reliance on affordable coal; in Sweden—maintaining energy affordability while achieving a high level of renewable energy integration [19]. Rising energy prices (raw materials) are deepening energy poverty. The time it takes to return to pre-crisis levels (COVID-19 + inflation) depends on support policies and economic conditions. Without intervention, many households could remain in energy poverty for several years. The publication fills a gap in the lack of forecasts and models assessing the impact of rising energy commodity prices on future levels of energy poverty in EU countries [20].
Energy efficiency	Analyzing the energy efficiency levels of sectors such as industry, transport, and households, progress was identified from 2011 to 2021. Further reductions in energy intensity will require greater investments and structural changes. The authors fill a gap in the analysis by measuring and comparing their energy efficiency over time and forecasting its development [21]. The publication fills a gap in the field of energy development prospects—in particular, increasing the share of RES and improving energy efficiency—in the face of the European Green Deal strategy and related challenges (the COVID-19 pandemic, the war in Ukraine) for countries with diverse socio-technical and economic conditions. The need for faster action was identified, especially in countries starting from a disadvantaged position. Adapting infrastructure, regulations, and financial support is necessary [22]. The report fills a gap in global guidance on how to effectively implement energy efficiency policies on a large scale, combining economic (employment, investment) and social (energy access, reduction in energy poverty) aspects [23]. Improving household energy efficiency has gained importance due to the increase in energy prices in recent years. The publication fills a gap in research on the actual actions and attitudes of Polish households towards improving energy efficiency, combining statistical data with an analysis of consumer behavior [24].
Energy security	Energy security is increased by reducing consumption, not imports. Energy security taxes are politically attractive, but more difficult to justify economically. The publication fills a gap in integrated, dynamic models that combine economics, technology, and systemic risks in measuring energy security [25]. Energy security tends to worsen income distribution in the early stages of economic development, while improving it after reaching a certain level of development. The publication fills a gap in the empirical analysis of the impact of multidimensional energy security on income inequality, taking into account different levels of economic development and the potential non-linearity of this impact [26]. The energy security of European countries strongly depends on the structure of the energy mix and the share of domestic/local resources. Economic development facilitates the implementation of strategies that improve energy security but does not automatically guarantee resilience—policies and investments are necessary. The authors have introduced an assessment of the energy security of European countries in the context of the availability of their own energy resources and economic conditions, using a consistent, comparable index [27]. The financial stability of energy companies is fundamental to energy security—its absence can lead to supply disruptions and investment problems. Legal aspects also have a significant impact—predictability of the law promotes investment and the resilience of the sector. The authors link the financial and economic stability of energy sector companies with national energy security, particularly in the context of energy transition and geopolitical shocks [28].
Energy policy and legal regulations	The degree of integration of the Polish energy sector with the European energy market is increasing with the introduction of new EU energy market regulations. The author proposes quantitative, advanced analyses assessing the impact of the integration of the Polish energy market with the EU market on prices, competition, and energy security [29]. A strong rule of law and the quality of legal institutions correlate with better implementation of energy policies and better energy outcomes (e.g., in terms of efficiency, the share of renewable energy sources, and the stability of supply). The authors fill a gap in the knowledge on the impact of EU countries’ compliance with the rule of law on the implementation of energy policy by examining the correlations between rule of law indicators and the objectives of the “green” and “brown” energy economy [30]. A just transition is a key component of policy—it requires safeguards for regions/industries dependent on fossil fuels. Cooperation between institutions, a clear legal framework, and predictability are necessary to make the transition socially acceptable. The economic aspects—the costs and benefits of the transition—must be distributed over time and space, and compensatory instruments should be legally established. The publication fills a gap by combining the legal, economic, and social aspects of a just energy transition in Poland by 2040, with particular emphasis on different regions (e.g., mining regions), the role of EU funds, and regulatory mechanisms [31]. Energy efficiency requires appropriate legal regulations and appropriate policies implemented by public authorities. EU and national legal instruments are crucial for modernization and investments in energy efficiency. Implementation of these regulations varies across countries, depending on administrative capacity, financial resources, and public acceptance. The publication analyzed the legal and economic aspects of energy efficiency in the EU, combining EU regulations with national practices in applying the law [32]. The authors examine how energy market regulations affect the financial stability of energy companies in the context of the global energy crisis. The pursuit of an integrated electricity market in the EU has inadvertently increased the interdependence between gas and electricity prices. Electricity market regulations should be flexible and well-thought-out to avoid deepening financial crises in the sector and discouraging investments in new energy sources [33].
Energy transition, sustainable development	Energy and energy policies in EU countries are increasingly linked to the idea of a circular economy. It is necessary to modernize infrastructure, reduce energy losses, and utilize waste and biofuels. The publication fills a gap by comparing how EU and former Eastern Bloc countries implemented circular economy principles in the energy sector between 2013 and 2020, and what economic and structural factors influenced the observed differences [34]. The diversity of EU countries in terms of environmental conditions has been studied—not all countries are able to achieve the ambitious EU energy policy goals in the field of environmental protection without having to make sacrifices in the socio-economic area. The author supplements the literature with a publication in which he quantitatively compares electricity production in the EU and Poland in the context of sustainable development, taking into account both environmental and socio-economic factors, and identifying the extent to which Poland can meet EU climate targets without serious social costs [35]. The authors fill the gap by combining technological innovations, economic aspects, and decarbonization goals in the long term (until 2040), showing how these three areas can work together in the energy transition. The current energy transition offers a unique opportunity to reconcile three objectives simultaneously provided that regulations, energy systems (in particular transmission and distribution networks), and financing are adapted to the pace and scale of change [36]. It is necessary to transition to systems based on renewable energy sources that are local, decentralized, and more resistant to external disruptions. The biggest obstacles are not technological or economic issues, but mentality, power structures, the influence of the conventional energy industry, politicians dependent on these structures, the subsidy system, and privileged investments. The authors fill this gap by quantitatively and empirically examining how realistic it is to achieve full energy independence based on renewable sources, and what the costs and challenges would be in different countries [37]. The authors fill a gap by examining how green transformation and sustainable management affect the financial performance of Polish energy companies. The positive consequences of implementing green energy include a reduction in CO2 emissions and access to new investment funds, while the negative aspects result from the rising costs of transformation and regulatory pressure. Comprehensive sustainable development management strategies need to be implemented to balance the risks and benefits of the green transition in a rapidly changing market environment [38].

presents scientific achievements in the most frequently discussed aspects of energy economics in the literature, focusing on research gaps filled by researchers. First, research on the market aspects of the phenomenon is presented, followed by its consequences, including energy poverty, energy security, and the need for legal regulations.

The literature [39,40,41,42,43,44] emphasizes that the assessment of energy economics requires consideration of many aspects of this phenomenon and the use of a set of multiple indicators simultaneously, as only their combination provides a complete picture—from technical efficiency and economic profitability to social and environmental impact. This does not in any way detract from research covering specific aspects of energy economics or conducted in selected countries. However, the authors decided to take a comprehensive approach to assessing the spatial diversity of EU countries in terms of energy economics, using a method based on artificial intelligence (SOM), taking into account a multifaceted, specific set of indicators and comparing changes in the phenomenon over time. The study fills a conceptual and methodological gap observed in previous studies.

3. Materials and Methods

3.1. Characteristics of the Research Material

This study assumes that information for 2023 plays a key role in examining the energy economy in the 27 EU countries. The year 2023 reflects the situation after the COVID-19 pandemic and the effects of global energy crises, including rising commodity prices and changes in energy policy caused by the war in Ukraine. The year 2019 was taken as a reference point because it was the last year before the pandemic, characterized by stable economic and energy trends. Data from that year allows for an assessment of standard energy consumption, raw material prices, and investments in renewable energy sources before the onset of global disruptions.

The selection of energy efficiency indicators was based on an analysis of the literature on the subject and a review of relevant documents [2,3]. The study primarily considered indicators available in the Eurostat database. The justification for using data from this database is its high reliability and accuracy. Eurostat is the official statistical body of the European Union, which uses uniform and proven methods of data collection and processing. This approach allows for comparison between countries in a European context. In this study, all indicators were given equal weight, as they come from a single, reliable, and standardized data source—the Eurostat database—and represent different but equally important aspects of the energy economy. This assumption avoids subjective differentiation of the importance of individual variables and is consistent with the approach used in exploratory studies aimed at identifying general patterns. At the same time, it should be noted that equal weighting may not reflect the varying political significance or variability of certain indicators, which is a potential limitation of the approach adopted.

The selected indicators provide information on various aspects of the phenomenon under study, such as cost, efficiency, and safety.

Table 2. Indicators to assess energy economics.

No.	Indicators	Unit
1	Final energy consumption	Tonnes of oil equivalent (TOE) per capita
2	Energy intensity of GDP in purchasing power standards	Kilograms of oil equivalent (KGOE) per thousand euro in purchasing power standards (PPS)
3	Energy productivity	Purchasing power standard (PPS) per kilogram of oil equivalent
4	Share of renewable energy in gross final energy consumption	%
5	Domestic net greenhouse gas emissions	Tonnes per capita
6	Per capita electricity demand	kWh
7	Per capita electricity generation	kWh
8	Share of primary energy consumption that comes from fossil fuels	%
9	Energy import dependency	%
10	Population unable to keep home adequately warm	%
11	Arrears on utility bills	%
12	Total population living in a dwelling with a leaking roof, damp walls, floors or foundation, or rot in window frames or floor	%
13	Expenditure on electricity, gas, and other fuels as a proportion of the total household expenditure	%
14	Average annual electricity prices for household consumers (with consumption from 2500 kWh to 4 999)	kWh
15	At-risk-of-poverty rate	%
16	Total energy supply	Thousand tonnes of oil equivalent

presents the indicators selected for the study [45].

In the next step of the study, an analysis was performed in terms of the information provided by the selected indicators. Due to the complex nature of the analyzed phenomenon, it is characterized by many variables that may duplicate certain information about the studied phenomenon. For this reason, the selection of diagnostic variables using a substantive and statistical approach is an extremely important step. According to the substantive criterion, variables should play a significant role in characterizing the phenomenon under study. The statistical criterion, on the other hand, emphasizes variability and correlations between variables, which should be characterized by sufficiently high variability and be weakly correlated with each other. To ensure the completeness of the data, the values of the diagnostic variables should be available.

The coefficient of variation was used to characterize variability, and the correlation of indicators was analyzed using the inverse correlation coefficient matrix. The procedure for eliminating strongly correlated indicators according to this method is as follows [46]:

• On the basis of the matrix of correlation coefficients R, the inverse matrix is determined $R^{-1}=\left[r^{\left(kl\right)}\right]$, where $r^{\left(kl\right)}$$\left(k, l=1, \dots , K\right)$ are elements of an inverse matrix $R^{-1}$, and each diagonal element $r^{kk}\in \left(1, +\infty \right)$, K—number of indicators. Indicators that are overcorrelated with the rest have diagonal elements $r^{kk}$ matrices $R^{-1}$ much greater than unity, indicating a bad conditioning of the matrix.
• Overcorrelated indicators to which diagonal elements correspond $r^{kk}$ values greater than 10 shall be eliminated from the set of potential indicators. If such elements are not present—the procedure ends.
• The inverse matrix is determined again $R^{-1}$ for a reduced set of indicators and its diagonal elements shall be analysed. The procedure is repeated until the stability of the matrix is achieved $R^{-1}$, i.e., occurrences of diagonal elements whose values do not exceed 10.

The indicators included in Table 2 meet the first condition—they are highly volatile, as measured by the coefficient of variation, which is well above 10% for each indicator, and they are complete for each EU country. In addition, the relationships between the indicators were examined using correlation measures. Two indicators—energy intensity of GDP in purchasing power standards and share of primary energy consumption that comes from fossil fuels—showed a strong correlation with others and were therefore excluded from further analysis. The other indicators are poorly correlated with each other, so they meet the condition of not duplicating information about the phenomenon under study.

3.2. Application of the Self-Organizing Map Method in the Classification of EU Countries

The SOM method, developed by Kohonen, belongs to the group of unsupervised neural networks used for the classification and visualization of multidimensional data [47]. Its core principle lies in mapping the variable space onto a two-dimensional grid of neurons while preserving the topological relationships among the objects. Each neuron is described by a weight vector that adjusts to the input data during the learning process, so that units with similar characteristics are positioned in proximity on the map [48]. The learning process is iterative and consists of gradually adapting the neuron weights to the input data. In each iteration, the Best Matching Unit (BMU) is selected, and its weights—together with those of neighboring neurons—are updated while the neighborhood radius and learning rate gradually decrease [49]. As a result, the network self-organizes, forming a map that reflects the structure of the analyzed data.

In this study, the SOM method was applied to classify EU countries based on energy–economic indicators. The diagram presented in Figure 1 illustrates the main stages of the network training process, which lead to the creation of an ordered neural map representing the relationships among the objects.

As an unsupervised learning method, SOM enables the identification of similarity patterns in multidimensional datasets without requiring assumptions about variable distributions or linear relationships. This makes it well suited for analyzing the diverse energy, cost, and social indicators used in this study. It is important to emphasize that SOM provides a typological representation of the data structure and does not permit causal inference—the resulting clusters reflect the configuration of values in multidimensional space rather than the influence of individual factors.

Figure 1 presents the sequence of learning steps in the SOM method—from calculating the distance between the input data and neurons, through selecting the Best Matching Unit (BMU) and updating the weights of neighboring neurons, to the successive reduction in the neighborhood radius and learning rate, leading to the convergence of the map.

SOM has been increasingly applied in energy–economic analyses, enabling the identification of structural similarities and developmental disparities among countries. In European studies, SOM has been used to classify EU member states in terms of energy security levels [50] and to examine the relationships between the use of renewable energy sources and environmental degradation [51]. It has also facilitated the analysis of links between energy efficiency and socio-economic factors [52], as well as the assessment of sustainable energy development considering the economic and demographic potential of countries [53]. In subsequent studies, SOM was applied to evaluate the progress of the energy transition [54], to map the level of sustainability of energy systems [55], and to identify research gaps in the field of sustainable transport and energy innovation [56]. The results of these studies confirm the high flexibility of SOM in analyzing multidimensional energy processes and its usefulness in shaping energy and climate policy.

4. Results

4.1. Statistical Characteristics of the Analyzed Indicators

All the indicators examined had a high capacity to discriminate between EU countries. The indicator with the highest variability was total energy supply (the coefficient of variation exceeded 130% in both years), the smallest—expenditure on electricity, gas, and other fuels as a proportion of the total household expenditure (the coefficient of variation was approximately 20% in both years). In terms of the analyzed indicators, many positive changes were noted for 2023 compared to 2019.

The average energy productivity has increased, which proves the increase in energy efficiency and efficiency of the economies of the EU countries. The share of renewable energy in gross final energy consumption also increased from 22.43% to 25.75%, indicating progress in the energy transition and the achievement of the EU’s climate goals. A positive change is also the reduction in the EU average per capita electricity generation, which is reflected in increasing energy efficiency and structural changes in the economies of EU countries. Another positive change is the indicator showing a reduction in dependence on energy imports in European Union countries, which translates into growing self-sufficiency and energy security in the region. It is also worth noting the changes concerning arrears on utility bills, where a decrease in their average level has been observed in EU countries. This is a desirable direction for change, indicating an improvement in the financial situation of households and, in particular, an improvement in the energy efficiency of buildings.

Despite the reduction in the EU average level of arrears on utility bills, the percentage of the population unable to keep their homes adequately warm remains significant in some countries, although it is decreasing. Such countries include Bulgaria (20.8%) or Greece (19.2%).

The significant differences in the levels of the analyzed indicators are also evidenced by the strong asymmetry of the distributions of the indicator values, in most cases to the right. Regarding the indicators adopted in the study, which have a disincentive effect on energy efficiency, this situation is desirable and shows that few EU countries perform very poorly in terms of the adopted indicators.

4.2. Classification of EU Countries Using the SOM Method Based on Energy-Economy Indicators

The classification of EU countries was conducted using SOM in the Statistica 13.3 environment. Given the number of analyzed units (27 countries), a two-dimensional grid with a 3 × 1 structure was applied, allowing for the identification of three distinct types of energy-economy profiles. The training process consisted of 1000 epochs, with an initial learning rate of 0.1 gradually reduced to 0.02, and a neighborhood radius decreasing from 3 to 0. These parameters ensured stable and convergent self-organization within the model.

The normalization of variables and the initialization of neuron weights were performed automatically by the SOM module before each run of the algorithm, ensuring consistent preparation of the input data. After the training process was completed, three groups of countries differing in their energy-economy profiles were obtained. These groups were then used to compare the typological structure of EU countries in 2019 and 2023 within the scope of the study.

4.3. Structure and Characteristics of SOM Classes in 2019 and 2023

The comparison of the SOM-based classification of EU countries for 2019 and 2023 made it possible to assess changes in the typological structure identified by the method. The cluster numbers were assigned based on a comparison of the mean indicator values within each group with the average values for the EU, which enabled the characterization of their profiles in terms of energy-economy performance. The numbering of classes is not hierarchical; instead, it reflects differences in the multidimensional data structure revealed by the SOM.

The analysis focused on assessing the stability of the identified classes over time and on identifying potential shifts in countries between them, interpreted as typological changes in their energy-economy profiles. This approach made it possible to determine the direction of transformation indicated by the analyzed indicators without drawing causal conclusions. The classification results for both years are presented in Figure 2.

In 2019, the SOM analysis made it possible to distinguish three groups of countries with distinct energy-economy profiles. Cluster 1 included countries characterized by a large scale of energy activity, high levels of electricity consumption and generation, and relatively favorable social conditions. This group also exhibited a higher share of renewable energy, indicating a more diversified structure of energy carriers used.

Cluster 2 included countries with lower energy efficiency and less favorable social and cost conditions, including, among other factors, greater household difficulties related to energy use and poorer housing conditions. The technical indicators of energy systems in this group reached lower values than in the remaining classes, reflecting a less advanced energy profile of these countries.

Cluster 3 included countries with a high scale of energy consumption, generation, and supply. The efficiency indicators in this group had moderate values, and the share of renewable energy was lower than in the other classes. The social conditions were generally close to the EU average. The profile of this cluster reflects the functioning of economies with high energy intensity.

To ensure the clarity of the comparison between the two periods,

Table 3. Mean values of the analyzed diagnostic variables for individual classes in 2019 and 2023.

No.	Diagnostic Variable	Year	EU Average	Cluster 1	Cluster 2	Cluster 3
1	Final energy consumption	2019	2.39	3.44	1.71	2.20
		2023	2.48	2.71	1.66	1.82
3	Energy productivity	2019	9.13	8.74	8.73	9.92
		2023	12.51	12.64	10.99	10.09
4	Share of renewable energy in gross final energy consumption	2019	22.43	27.94	22.56	17.38
		2023	25.71	27.60	25.73	21.38
5	Domestic net greenhouse gas emissions	2019	7.87	9.29	5.79	8.93
		2023	8.13	8.28	5.94	7.46
6	Per capita electricity demand	2019	6.70	9.59	4.66	6.39
		2023	7.23	8.17	4.77	5.22
7	Per capita electricity generation	2019	5926.74	8267.38	3809.50	6198.67
		2023	6641.37	7801.08	4240.40	4422.00
9	Energy import dependency	2019	60.02	54.44	65.18	59.24
		2023	58.78	56.09	62.47	49.22
10	Population unable to keep home adequately warm	2019	8.18	2.35	15.13	5.64
		2023	8.12	5.98	15.05	6.72
11	Arrears on utility bills	2019	8.23	4.13	13.59	5.93
		2023	6.64	4.97	11.63	5.44
12	Total population living in a dwelling with a leaking roof, damp walls, floors or foundation. or rot in window frames or floor	2019	13.64	11.68	16.33	12.40
		2023	14.67	14.07	15.37	10.34
13	Expenditure on electricity, gas, and other fuels as a proportion of the total household expenditure	2019	21.60	24.60	17.66	23.31
		2023	23.69	25.10	18.28	23.74
14	Average annual electricity prices for household consumers (with consumption from 2500 kWh to 4999 kWh)	2019	0.14	0.12	0.15	0.14
		2023	0.13	0.13	0.15	0.14
15	At-risk-of-poverty rate	2019	21.04	16.14	25.17	20.82
		2023	14.81	13.62	19.07	16.56
16	Total energy supply	2019	52,475.81	32,214.81	14,659.51	112,503.69
		2023	55,908.76	63,188.46	22,312.60	56,043.02

Note: The units of the indicators are consistent with the values presented in Table 2.

presents the mean values of the analyzed variables for the individual SOM classes in 2019 and 2023, which serve as the basis for the subsequent interpretation of the results.

Based on the data presented in Table 3, the classification of EU countries for 2023 was also analyzed. These results formed the basis for assessing changes in the structure of the clusters identified by the SOM.

Table 4. Characteristics of EU country classes identified using SOM in 2023.

Class	n	Favorable Features	Unfavorable Features
1	12	Higher energy productivity and a greater share of renewable energy than the EU average. Lower values of final energy consumption, emissions, difficulties in heating the home, and arrears in paying utility bills.	Slightly higher final energy consumption and dependence on energy imports, with moderately weaker social indicators. Higher electricity demand, higher electricity generation, and a higher level of energy imports than the EU average.
2	10	Final energy consumption, emissions, and both electricity demand and generation are lower than the EU average.	Energy productivity and the share of renewable energy are lower than the EU average. The cluster is characterized by a higher level of energy imports, higher indicators of difficulties in heating the home and arrears in paying bills, and higher electricity prices.
3	5	Final energy consumption, electricity demand, and the share of energy expenditures are lower than the EU average.	Energy productivity and the share of renewable energy are lower than the EU average. Emissions, electricity prices, and the level of energy imports are higher, as are the indicators of difficulties in heating the home and arrears in paying bills.

n—number of countries in the given cluster.

provides a concise summary of the characteristic features of the individual clusters in 2023, developed from their mean indicator profiles. The resulting cluster structure reflects the persistent differentiation of profiles in terms of energy-economy performance among the member states, stemming from the varying values of the analyzed variables. This division includes both countries with relatively more favorable energy and social parameters and those with less advantageous profiles, highlighting the multidimensional nature of the data revealed by the SOM algorithm.

The differentiation of the clusters indicates that, despite improvements in some indicators in 2023, the profiles of EU countries still differed in terms of their energy-economy structure. The classification results show that among the Member States there were both groups with more favorable energy and social parameters and groups with less advantageous profiles. The distinct nature of these configurations reflects the multidimensional character of the analyzed data and the way it is represented by the SOM algorithm. It should be emphasized that the obtained results do not allow for drawing conclusions about the causes of the observed differences; rather, they describe their typological nature.

4.4. Changes in SOM Classification Between 2019 and 2023

A comparison of the classification of EU countries in 2019 and 2023 shows small but noticeable shifts in the structure of the clusters identified by the SOM algorithm. For most countries, the cluster assignment remained stable, which confirms the persistence of the typological structure during the study period. The changes mainly occurred in countries with similar economic potential, where improvements in selected energy and social indicators did not follow the trends observed in the other Member States. The direction and size of these shifts are shown in Figure 3, which presents the classification differences (Delta change) between 2019 and 2023.

The shifts shown in Figure 3 are reflected in the detailed analysis presented in

Table 5. EU countries that changed their position in the SOM classification (2019—2023).

Country	Δ	Direction	Reasons for Change
Belgium	3 → 1	↑	Belgium moved from cluster 3 in 2019 to cluster 1 in 2023, indicating a clear shift in the country’s profile toward the group characterized by higher energy efficiency and more favorable technical and economic conditions. The key driver of this change was the improvement in efficiency-related indicators—specifically, an increase in energy productivity (X3) and decreases in final energy consumption (X1) and greenhouse gas emissions (X5). At the same time, the country’s dependence on energy imports (X9) declined, which made Belgium more similar to cluster 1, where the level of this variable is the lowest. The values of other indicators—including the share of renewable energy (X4)—also shifted the country’s profile in the direction of the structure typical of the more efficient cluster. Although some social indicators slightly deteriorated (e.g., difficulties in maintaining adequate indoor temperatures—X10), their levels remained much closer to the averages of cluster 1 than to the characteristics of cluster 3, where the scale of energy poverty is considerably higher. Overall, the dominant processes—improvements in efficiency, reductions in emissions, and lower import dependence—outweighed the moderate deteriorations in the social dimension and resulted in Belgium’s shift to cluster 1.
France	3 → 1	↑	France moved from cluster 3 in 2019 to cluster 1 in 2023, indicating a clear shift in the structure of its energy economy toward the most efficient cluster. The most important changes occurred in the in decarbonization. An increase in the share of renewable energy (X4), a reduction in greenhouse gas emissions (X5), and a lower dependence on energy imports (X9) shifted France’s profile toward cluster 1, which is characterized by the highest level of efficiency and relatively low emissions. At the same time, stable electricity prices (X14) and an improvement in energy productivity (X3) also contributed to this shift. Some social indicators (X10—X12) deteriorated; however, their 2023 levels remain closer to the cluster averages of cluster 1 than to the clearly less favorable values typical of cluster 3. This means that despite declines in housing conditions and energy poverty indicators, France’s overall indicator structure aligns more strongly with the profile of the more energy-efficient clusters. The shift to cluster 1 reflects the dominance of positive changes in the environmental and efficiency dimensions, which outweighed the moderate deterioration of social indicators.
Germany	3 → 1	↑	Germany moved from cluster 3 in 2019 to cluster 1 in 2023, indicating a clear shift in the structure of its energy economy toward the group with the highest energy efficiency and the most favorable technical and economic profile. The change was driven primarily by improvements in energy efficiency, including an increase in energy productivity (X3), a reduction in greenhouse gas emissions (X5), and a decrease in electricity consumption per capita (X6). These directions are consistent with the characteristics of cluster 1, which is defined by lower emission levels and higher efficiency in energy use. During the same period, several social indicators (X10—X13) deteriorated, and total energy supply (X16) decreased. However, their 2023 levels remain closer to the values observed in cluster 1 than to those typical of cluster 3, where the scale of social challenges related to energy poverty and cost burdens is considerably higher. Overall, Germany’s shift to cluster 1 reflects the dominance of improvements in efficiency, emissions, and energy supply stability, which had a stronger influence on the country’s profile than the unfavorable changes in the social dimension.
Ireland	3 → 1	↑	Ireland moved from cluster 3 in 2019 to cluster 1 in 2023, indicating a clear shift in the structure of its energy profile toward the group characterized by the highest efficiency and a favorable technical and economic balance. The main processes driving this change were improvements in energy efficiency, including a substantial increase in energy productivity (X3) and a reduction in greenhouse gas emissions (X5). These developments are consistent with the characteristics of cluster 1, where efficiency and decarbonization indicators reach the highest levels. At the same time, some demand-related factors evolved favorably—electricity consumption per capita (X6) remained stable. Several social and cost-related indicators (X10, X13—X14) deteriorated, and Ireland’s dependence on energy imports (X9) increased. Nevertheless, their 2023 levels remain closer to the averages of cluster 1 than to the clearly less favorable values observed in cluster 3, where the intensity of social and cost-related problems is significantly higher. As a result, the dominant influence came from positive changes in efficiency, emissions, and the structure of energy production, which shifted Ireland toward cluster 1 despite simultaneous deteriorations in the social and import dimensions.
Spain	3 → 2	↑	In the SOM analysis, Spain moved from cluster 3 in 2019 to cluster 2 in 2023, indicating a moderate convergence of its diagnostic profile toward the group with more favorable efficiency characteristics, while still maintaining social conditions that differ from those in cluster 1. The most important shifts occurred in the energy dimension. A decline in final energy consumption (X1), a reduction in import dependence (X9), an increase in energy productivity (X3), and a higher share of renewable energy (X4) made Spain’s profile more similar to the cluster 2 profile, which features higher efficiency and a moderate level of emissions. These changes simultaneously moved the country away from the typical configuration of cluster 3, where efficiency indicators are notably weaker. At the same time, the social situation deteriorated significantly—especially the share of people having difficulty maintaining adequate indoor temperatures (X10), which increased by 13 percentage points, as well as household cost-burden indicators (X11—X14). Nevertheless, the 2023 values remained closer to the averages of cluster 2 than to the distinctly unfavorable conditions observed in cluster 3. Overall, Spain’s reclassification is moderate in nature and results primarily from improvements in energy-related parameters and a partial shift toward the efficiency profile of cluster 2, while the deterioration in social indicators did not alter the general direction of this movement.
Hungary	2 → 3	↓	Hungary moved from cluster 2 in 2019 to cluster 3 in 2023, indicating stronger similarity of its diagnostic profile to units characterized by higher social vulnerability and less favorable structural conditions. A key factor behind this shift was the marked deterioration of social and cost-related indicators. The share of people having difficulty maintaining adequate indoor temperatures (X10) increased substantially, as did the share of household expenditures on energy (X13). These developments are typical of the cluster 3 profile, where the intensity of energy poverty and cost burdens is higher than in cluster 2. In addition, total energy supply (X16) declined, further weakening the structural stability of the energy system. At the same time, some technical and economic variables improved. Energy productivity (X3) increased, and dependence on energy imports (X9) decreased. However, these improvements were not strong enough to maintain similarity to the average profile of cluster 2, which is characterized by better social conditions and greater cost stability. As a result, the dominant deteriorations in the social and cost dimensions shifted Hungary’s diagnostic structure toward cluster 3, reflecting the intensification of problems typical of units with less favorable living conditions and higher energy vulnerability.

Δ—cluster change, →—shift, ↑—improvement in position (advance to a group with a higher level of development), ↓—deterioration in position (decline to a group with a lower level of development).

, which indicates the direction of the classification changes and describes their nature in the individual countries.

A comparison of the classification of EU countries in 2019 and 2023 shows that several Western and Northern European countries moved to clusters with more favorable energy-economy profiles, reflecting changes in the values of the analyzed indicators over this period. In the case of Hungary, an opposite shift was observed, indicating a different dynamic of change compared with most EU Member States. Overall, the 2023 classification displayed greater internal differentiation than in 2019 and a clearer concentration of some countries in clusters characterized by more favorable indicator values.

5. Discussion

The subject matter and results of this study indicate that energy economics is a complex phenomenon and a key element of sustainable development. Its complexity stems from the fact that it combines many different fields and factors—technical, economic, social, political, and environmental. The importance of energy economics is demonstrated by publication [57], which presents a bibliometric analysis of the entire output of the journal Energy Economics from 1979 to April 2024.

The authors analyzed 6563 documents in which topics related to such aspects of energy economics as: oil and oil price, carbon dioxide emissions, climate change, renewable energy, energy transition, electricity markets, energy efficiency and energy poverty. Researchers studying energy economics analyze its various determinants, usually focusing on a specific problem area. In recent years, there has been a significant increase in interest in renewable energy sources, as evidenced by numerous studies on this topic [51,58,59,60,61]. Also, in relation to energy poverty, there is an increase in interest, which translates into scientific studies [62,63,64,65].

A review of the literature on the subject shows that researchers very often analyze one of the aspects of energy economics. In our study, we propose a more comprehensive approach to this issue, including indicators from various areas of energy economics in the set of variables: energy security, the market, energy poverty, and energy efficiency. According to the literature review and our knowledge, this approach has not yet been considered in the context presented in this paper, which allows us to fill a research gap concerning the assessment of these phenomena.

To achieve the objective of the study, indicators from the Eurostat database were used, which are a reliable source for comparative analyses between EU countries. The literature on the subject contains a great many studies conducted for specific countries [65,66,67,68,69,70,71,72], which are of great importance mainly for practical reasons and are a source of information for decision-making at the national level. They also provide a starting point for comparative analyses between.

AI-supported methods are increasingly used in research on energy and its various aspects. There is a growing number of studies using advanced artificial intelligence and machine learning techniques, e.g., for energy price forecasting [73], market analysis [74], and renewable energy production modeling [75]. This study uses the SOM method, which enabled unsupervised data analysis, allowing for the identification of hidden patterns and structures without the need to make prior assumptions about the relationships between variables. The use of SOM in the context of energy economics is still not widespread, which makes the presented study original and a potential reference point for future work using unsupervised data mining methods in energy analysis.

6. Conclusions

In this study, the authors considered all 27 EU countries, which made it possible to use advanced quantitative methods, such as neural networks, to identify new patterns among EU countries in terms of energy economics. The approach used is innovative in that it covers all 27 EU countries and uses neural networks to identify hidden patterns in energy economics. The result of our study is a classification of EU countries and the identification of new patterns that are not visible when using traditional methods. It should also be emphasized that the study was conducted in two years, 2019 and 2023, i.e., before and after the COVID-19 pandemic. The information for 2023 is the latest data available in the Eurostat database, which increases the practical value of the study’s conclusions. The study, conducted over two years, identified changes in the positions of countries based on selected indicators in the context of shocking events such as the COVID-19 pandemic and the war in Ukraine.

The use of Kohonen’s self-organizing network allowed for the identification of the following patterns:

• In 2019, the group of EU countries characterized by high energy productivity, a significant share of renewable energy sources, low greenhouse gas emissions, and a favorable social and cost profile mainly included countries in Northern and Western Europe. In 2023, the group of countries achieving the best results in terms of energy economy expanded to include Western European countries such as France, Germany, Ireland, and Belgium.
• The weakest results in terms of energy efficiency indicators were recorded in 2023 for Central and Eastern European countries. Italy was also included in this group. It is also worth noting that countries such as Poland, Slovakia, Estonia, and Italy were classified as having the weakest energy management in both years of the study.
• Hungary was the only country where social and cost indicators deteriorated and total energy supply declined. Despite improvements in energy productivity and reduced import dependency, Hungary moved to the group with a lower energy-economy level.
• The classification of EU countries in 2023 was characterized by greater internal diversity than in 2019, with a clearer concentration of countries with higher energy economy in classes with more favorable development parameters.
• Despite progress in some countries, differences in economic indicators between EU countries persist.

The results of the study allowed the authors to give affirmative answers to all the research questions formulated in the introduction.

This work contributes several significant elements to the existing scientific literature on the assessment and identification of issues related to energy economics. The innovation of the study lies in the use of neural networks, based on artificial intelligence methods, to classify European Union countries in terms of level of energy economy. Such an approach enables the identification of new patterns among countries, providing a valuable complement to studies that employ traditional quantitative methods.

In addition, the assessment of the classifications from two years—2019 and 2023—allows for comparisons and an evaluation of changes in energy-economy levels in EU countries. The study fills a gap in the literature on comprehensive comparative energy economics assessment using current Eurostat data and advanced data analysis tools.

The results of the study provide important information and guidance for decision-making in the field of European Union energy policy by identifying differences in energy economy and determining areas requiring support in the energy transition process.

This study fills a methodological gap by demonstrating the potential of artificial intelligence in assessing selected aspects of sustainable development in EU countries.

The authors of this study note both the limitations of the research and the potential for further development of research on energy economics. Future studies plan to extend the analysis to a longer time horizon and to additional factors affecting energy management efficiency. The issue of selecting indicators for the study is directly related to the problem of assigning appropriate weights to them. Adopting uniform weights may limit the precision of interpretation, especially when individual indicators differ in their significance for energy policy or are characterized by different variability. Therefore, in future studies, it would be reasonable to consider alternative weighting methods, such as principal component analysis (PCA) or expert approaches.

Another direction of analysis could be the use of more advanced machine learning techniques, such as k-means, DBSCAN, HDBSCAN, or supervised models, which include random forest and gradient boosting. These methods could be used to deepen the structural or predictive analysis of numerical values or the classification of various objects.

This approach will allow for deeper conclusions about energy economics and increase the practical value of the results for policymakers and other researchers.