Identifying the causes of discrepancy between measured and modelled indoor temperatures in two synthetically occupied test houses

Dynamic thermal models are used to predict indoor temperatures for overheating building regulation compliance (Part O) assessments. The models must therefore be reliable in their predictions, but previous work has shown differences between measured and modelled temperatures. The present study uses bespoke datasets measured in test houses to examine the influence of various model inputs on the accuracy of the overheating predictions. Standard metrics were used for determining model accuracy. The results showed that accuracy was improved by reducing window ventilation rate by 20%, reducing the U-value of curtains, carefully modelling curtains with respect their interaction with incoming solar radiation, artificially reducing the diffuse solar radiation in the weather file, and increasing the thermal mass in the model to beyond the levels that exist in the real test house.