Identifying urban energy-vulnerable areas: a machine learning approach

Abstract

Access to energy services is essential for preserving health and well-being. However, energy poverty is a challenge affecting millions of citizens worldwide, which could even worsen due to the predicted severity of climate change. Energy poverty vulnerability and social problems are often linked to energy-inefficient buildings. Thus, identifying energy-inefficient dwellings in energy-vulnerable urban areas is crucial for formulating and implementing effective public policies. Consequently, this study proposes a multidimensional methodological approach to determine these urban areas and support decision-making to develop public policies that can help lift dwellings out of or prevent them from falling into energy poverty. The suggested methodology utilizes public data from existing databases and applies unsupervised machine–learning classification algorithms. Applying such methodology to the case study of Seville identified different clusters of urban areas with similar characteristics, providing key information for creating specific public policies tailored to the needs of each area and community. The study’s findings support Building Renovation Wave strategies to improve energy efficiency in dwellings, define specific policies for access to financial resources for low-income families, and provide personalized support for vulnerable populations.