In Situ Testing of a Building's Overall Heat Loss Coefficient - Embedding Quasi-stationary and Dynamic Tests in a Building Physical and Statistical Framework

The building sector accounts for about 40% of Europe's total final energy use. Buildings thus harbour enormous potential to save energy, and reduce carbon dioxide emissions. More than half of their energy use is attributable to space heating, which is dependent on the thermal performance of their envelopes, that separate the indoor environment from external influences, and their usage. In order to reduce the energy use of buildings, policy makers are increasingly demanding with regard to their energy performance and the thermal performance of their envelopes, and use mandatory energy labelling to enforce compliance. The building sector is transforming as a result. Several studies, however, indicate a discrepancy, between the designed and the actual performance of buildings. With regard to building envelopes, such discrepancies may rise up to 400% on the scale of building envelope parts, and up to 100% on the scale of a whole building. Today, however, there is little transparency with regard to the actual performance of building envelopes. If we are to effectively advance the building stock's energy efficiency, we need reliable methods to verify this performance. In this work, we investigate the thermal performance characterisation of whole building envelopes, by fitting simplified thermal models to data collected during dedicated heating experiments, performed on vacant houses. We thereby focus on the overall heat loss coefficient, H, in W/K. We develop a building physical framework, that captures the physical phenomena that inform a building's dynamic thermal behaviour, during such heating experiments. From this framework, we derive simplified thermal models, that are apt to be fitted, using linear regression, input-output polynomial modelling (ARX), or grey-box modelling, to data collected from quasi-stationary and dynamic heating experiments. The link between the building physical framework and the simplified thermal models reveals physical phenomena that get lumped into the simplified models' parameters, and uncovers guidelines for organising high quality experiments. At the same time, the building physical framework provides the practitioner with an extended model library. By deriving, from a first-order grey-box model, a black-box ARX model, we show that ARX models comprise parameters that remain physically relevant. We thus derive explicit ARX models, but their application to test cases shows that implicit ARX models are much better apt to deal with data from dynamic heating experiments. We define four scenarios, that provide appropriate strategies to estimate H, based on the building typology, the heating experiment, and the analysis method. We illustrate that the analyses commonly found in literature, can only be applied in a limited number of cases. The developed methodology is applied to three quasi-stationary and three dynamic test cases. They are analysed adopting similar statistical techniques; but we show that the selection of an appropriate scenario underpins the reliability and physical relevance of the obtained estimates. We show that scenarios where H is estimated as the stationary gain with regard to the indoor air temperature yield reliable and robust results. The results also indicate that a dynamic test, analysed with ARX modelling, is not more robust than a co-heating test, analysed with either linear regression or ARX modelling, nor does it yield reliable results for a shorter duration. On the other hand, the combination of a dynamic test and grey-box modelling rewarded measurement durations of 4 to 5 days with reliable results. On the basis of the developed methodology, and its application on six test cases, we suggest an onset for a flowchart, to guide the practitioner in reliably characterising the overall heat loss coefficient of buildings.

Publication year:2015

Accessibility:Open