Digital energy performance certification of single-family houses in Denmark

Regulatory Energy Performance Certificates (EPCs) often diverge substantially from actual building energy use, leading to the well-known performance gap. With widespread deployment of smart heat meters and an increasing policy emphasis on operationally grounded EPC assessments under the revised EPBD, new methods are needed to reflect in-use building behaviour at scale. This work proposes a complementary, data-driven Energy Performance Certificate (dEPC) framework based on a Bayesian ARMAX Energy Signature (ARMAX-ES) model. Using daily heat meter readings and weather data, the model yields probabilistic estimates of envelope heat loss (UA0), wind-driven infiltration losses (UAw), base heat load (𝜙0), solar gains (gA), and the standardised annual heating demand under a reference climate. The framework was applied to 2788 single-family houses in Denmark, resulting in the first national-scale empirical characterisation of residential building energy performance derived solely from operational data. Results reveal limited stratification of thermal parameters across the regulatory EPC classes and frequent, systematic deviations between regulatory EPC labels and dEPC-assigned in-use energy labelling. High-rated dwellings often consume more heat than expected, while lower-rated houses sometimes consume less heat than predicted. At the building level, the dEPC diagnostics identify performance anomalies and support targeted retrofit planning. At stock level, the inferred parameter distributions offer actionable insight for policymakers, retrofit prioritisation, and grid-scale heating demand planning. Overall, the dEPC framework provides a scalable, evidence-based complement to regulatory EPCs, enhancing transparency, credibility, and policy relevance of building energy certification.