Predicting Energy and Emissions in Residential Building Stocks: National UBEM with Energy Performance Certificates and Artificial Intelligence

Abstract

To effectively decarbonize Europe’s building stock, it is crucial to monitor the progress of energy consumption and the associated emissions. This study addresses the challenge by developing a national-scale urban building energy model (nUBEM) using artificial intelligence to predict non-renewable primary energy consumption and associated GHG emissions for residential buildings. Applied to the case study of Spain, the nUBEM leverages open data from energy performance certificates (EPCs), cadastral records, INSPIRE cadastre data, digital terrain models (DTM), and national statistics, all aligned with European directives, ensuring adaptability across EU member states with similar open data frameworks. Using the XGBoost machine learning algorithm, the model analyzes the physical and geometrical characteristics of residential buildings in Spain. Our findings indicate that the XGBoost algorithm outperforms other techniques estimating building-level energy consumption and emissions. The nUBEM offers granular information on energy performance building-by-building related to their physical and geometrical characteristics. The results achieved surpass those of previous studies, demonstrating the model’s accuracy and potential impact. The nUBEM is a powerful tool for analyzing residential building stock and supporting data-driven decarbonization strategies. By providing reliable progress indicators for renovation policies, the methodology enhances compliance with EU directives and offers a scalable framework for monitoring decarbonization progress across Europe.

Keywords

• urban building energy model (UBEM)
• energy performance certificates (EPCs)
• machine learning
• national building stock
• data driven approaches
• progress indicators