Variability in the Heat Transfer Coefficient of dwellings

There is a need to accurately measure the thermal performance of a dwelling’s envelope, to identify those dwellings most in need of improvements through retrofit and to quantify the delivered improvement. The Heat Transfer Coefficient (HTC) is a widely used metric for quantifying whole dwelling thermal performance. It describes the average rate of heat loss from transmission through the building fabric and by ventilation and is expressed in Watts per Kelvin (W/K). Specialised tests, such as the co-heating test, developed for directly measuring the HTC of a dwelling have been useful in quantifying the energy performance gap. Such tests have shown that the HTC of a dwelling can vary between measurements, depending on the conditions present. The HTC of a dwelling can be measured continuously whilst a dwelling is occupied in an unobtrusive manner over the long-term. The longitudinal measurement of the in-use HTC provides new insights into the variability of the HTC.

This thesis investigates the variability in the HTC, and how, by explaining the variability, it can be used to reduce the uncertainty in the measurement of a dwelling’s thermal performance. Three main phases of analysis were conducted. Firstly, different methods of calculating the in-use HTC using data from occupied dwellings were evaluated, and the most accurate HTC calculation method was identified through comparison with the co-heating test HTC of each dwelling in the sample. Multiple linear regression using 20 days of data was identified as the most accurate. The analyses also revealed that the in-use HTC of a dwelling can vary, even over a short monitoring period.

The second phase of analysis sought to quantify and characterise the variability in the in-use HTC. The in-use HTC was calculated over rolling 20-day periods for 19 occupied dwellings for a heating season. The in-use HTC of the dwellings had a coefficient of variation (standard deviation divided by mean) of between 1.0% and 11.8% (mean 7.1%). Linear models were developed to explain the variability in the in-use HTC as a function of changes in environmental boundary conditions observed during the measurement. Unique linear models were developed for 17 of the 19 dwellings, identifying that the in-use HTC for each dwelling had a unique relationship with its boundary conditions. Despite being limited to between only 11 and 52 in-use HTC values, the linear models explained at least 80% of the variability in 13 of 17 dwellings.

The third phase of analysis used data collected from a pair of adjoining semi-detached test houses across two heating seasons. The indoor conditions of both test houses were controlled such that the impact of inter-dwelling heat transfer via the party wall on the HTC could be evaluated. Whilst a linear model made from data collected across the whole experiment could not directly account for inter-dwelling heat transfer, measured heat flux enabled the predicted HTC to be adjusted to account for the additional heat loss. Where there was no net inter-dwelling heat transfer, the linear model could accurately predict the HTC of the test house for a range of observed boundary conditions.

This research has demonstrated that variability in the HTC should not be attributed to uncertainty in HTC measurement. In fact, by using linear models to explain the variability in the HTC of dwellings, uncertainty can be reduced, and the HTC can be predicted for pre-defined conditions. The results will be of value for the accurate sizing of heat pump systems, using the HTC predicted for specific conditions, and to academics and innovators aiming to further improve technologies and methods for measuring the thermal performance of occupied dwellings. Equally, the results of this thesis will be of value to policymakers, regulators and building assessors, seeking to quantify the thermal performance of the housing stock, demonstrate the improvements delivered by retrofit investment and close the energy performance gap.