On the detection of patterns in electricity prices across European countries: An unsupervised machine learning approach

Abstract

The year 2022 is characterized by a generalized energy crisis, which leads to steadily increasing electricity prices around the world, while the corresponding salaries remain stable. Therefore, examining trends in electricity prices relative to existing income levels can provide valuable insights into the overpricing/underpricing of energy consumption. In this article, we examine the tendencies of 35 European countries according to their national kWh prices and the average household incomes. We use a series of established clustering methods that leverage available information to reveal price and income patterns across Europe. We obtain important information on the balance between family earnings and electricity prices in each European country and are able to identify countries and regions that offer the most and least favorable economic conditions based on these two characteristics studied. Our analysis reveals the existence of four price and income patterns that reflect geographical differences across Europe. Countries such as Iceland, Norway, and Luxembourg exhibit the most favorable balance between prices and earnings. Conversely, electricity prices appear to be overpriced in many southern and eastern countries, with Portugal being the most prominent example of this phenomenon. In general, average household incomes become more satisfactory for European citizens as we move from east to west and south to north. In contrast, the respective national electricity prices do not follow this geographical pattern, leading to notable imbalances. After identifying significant cases of inflated prices, we investigate the respective causes of the observed situation with the aim of explaining this extreme behavior with exogenous factors. Finally, it becomes clear that the recent increase in energy prices should not be considered as a completely unexpected event, but rather as a phenomenon that has occurred and developed gradually over the years.

Keywords

• clustering algorithms
• electricity prices
• partition around medoids
• hierarchical clustering
• household incomes
• unsupervised machine learning