The Sensitivity of In-Use Heat Transfer Coefficient Results to Practical Constraints in Measurement of Thermal Performance in Occupied Homes

In-use heat transfer coefficient (HTC) measurements are useful for retrofit evaluation, heating system sizing, and thermal performance assessment in occupied homes. Quantifying the variance in the in-use HTC when, necessarily, utilising a range of assumptions and simplifications is therefore crucial, and also critical for further method development. Two empirical sensitivity analyses were used to explore how changes in commonly assumed in-use factors affect HTC estimates in occupied homes. The factors investigated were measurement uncertainty, solar gains, metabolic gains, boiler efficiency, water use, party wall heat transfer, and ventilation rate - parameters that are impractical or impossible to routinely measure, and as such default values are generally adopted. The sensitivity analyses used data from seven occupied homes and a single, common, HTC estimation method. Input distributions for each factor were derived from available data and current assumptions. A local sensitivity analysis examined how changes in each input affect the HTC and a global analysis quantified the contribution of the inputs’ variance to the HTC’s variance. Conducting parallel analyses enabled a more complete picture to be obtained, and the alignment of the two approaches provided confidence in their results. The factors with the greatest overall effects on the HTC were ventilation and party wall heat transfer; however, this was not the case for every home. In particular, HTCs from homes with higher occupancy exhibited stronger HTC sensitivity to metabolic gains and water use. The use of real data from occupied homes enables the results to be applicable to typical imperfect datasets. The results will inform future applications of in-use HTC measurements and methods for determining their uncertainty. Further work expanding this analysis to a larger dataset with more building typologies, and gathering data to define the sensitivity analysis more accurately would strengthen these conclusions.