Large-scale statistical analysis and modelling of real and regulatory total energy use in existing single-family houses in Flanders. Issue 2 (17th February 2023)

• Record Type: Journal Article
• Authors: Van Hove, M. Y. C.; Deurinck, M.; Lameire, W.; Laverge, J.; Janssens, A.; Delghust, M.
• Is Part Of: Building research and information. Volume 51:Issue 2(2023)
• Page Start: 203
• Page End: 222
• Publication Date: 2023-02-17
• Subjects: Existing single-family houses -- energy performance registry -- real annual energy use -- true energy savings -- data-driven statistical modelling -- multicollinearity; Building -- Periodicals; Building -- Research -- Periodicals; 690.015
• DOI: 10.1080/09613218.2022.2113023
• Languages: English
• ISSNs: 0961-3218

Abstract

ABSTRACT: Large-scale statistical studies on the gap between the real and regulatory energy use in residential buildings in Europe have shown that the regulatory calculation overestimated the real energy use, inflated true energy savings and undermined national energy policy making. Using data from 122, 680 Flemish existing single-family houses, this research builds further on existing studies by contributing results for Flanders. The study also examines to what extent available aggregated variables explain the real annual total building energy use using statistical linear models and addresses the problem of multicollinearity and the importance of bootstrapped confidence intervals for model quality control. The overestimation of the real total energy use (and potential energy savings) by the Flemish regulatory method is exceedingly large compared to studies from other EU countries. The Flemish labels prove very poor indicators of the real energy use. Statistical linear models explain up to 46.6% of all variability and indicate that a significant extent of multicollinearity had to be corrected. Half of the variability has been left unexplained and has to be attributed to variables that were not available and the fact that the data were insufficiently accurate. Future analysis will explore whether more complex models identify more evidence.