Data-driven energy performance certification for Danish single-family homes: a proof of concept

Source abstract

Abstract Energy Performance Certificates (EPC) often misrepresent real-world energy use of buildings due to their reliance on standardized assumptions, leading to a well-documented performance gap. This study introduces a data-driven alternative, the dEPC, leveraging Bayesian Energy Signature modeling on smart meter data from 2877 single-family homes in Denmark. The dEPC framework directly estimates the energy performance characteristics of the in-use building, offering a probabilistic assessment with quantified uncertainty. Results indicate that traditional EPC labels fail to systematically correlate with measured heat loss coefficients (HLC). Comparison of EPC and dEPC ratings reveals consistent discrepancies: highly rated buildings tend to have overestimated efficiencies, while lower-rated buildings are often underestimated. The dEPC methodology provides actionable insights at both the building stock and individual building levels, allowing for a more accurate validation of EPC assessments. The results demonstrate the potential for integrating data-driven methods into existing EPC frameworks, improving transparency and decision-making for building owners, engineers, and policy makers. By bridging the gap between theoretical assessments and real-world performance, dEPC improves the credibility and utility of energy performance certification, supporting a data-driven transition to sustainable building management.