Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Data-driven statistical modelling of real energy use in modern Flemish single-family houses: a feasibility study

Matthias Yvan C. Van Hove 1, Mieke Deurinck 2, Wim Lameire 2, Jelle Laverge 1, Arnold Janssens 1, M. Delghust 1
1 Ghent University, Department of Architecture and Urban Planning, Building Physics Research Group, Belgium
2 Flemish Energy and Climate Agency (VEKA), Brussels, Belgium

Abstract: How accurately do regulatory energy calculations predict the real energy use in modern Flemish single-family houses and how does their performance relate to comprehensive data-driven statistical models built from common dwelling parameters? This question was investigated by using data from the EPB-registry and real energy uses from the energy utilities for 135166 singlefamily houses. The paper comprises outlined methodology choices, enlightening database analyses and linear regression model results. The results clearly demonstrate that the applied statistical modelling method outperforms the official EPB-calculation method, however the model performance is still limited and a large part of the variance is left unexplained.
Keywords: statistical modelling, linear regression, EPB data, Flanders
Pages: 1295 - 1302