Analysis of physical uncertainties in computer simulations of residential buildings

Abstract

The aim of this paper is to analyse physical uncertainties in a computer simulation of a residential building and the implications for its thermal performance. The simulation experiment was performed using the Monte Carlo method, involving the thermal and physical properties of materials. The sensitivity of the variables was also analysed using regression and correlation coefficients. The degree-hours of discomfort criterion for heating and cooling for the climate of Florianópolis-SC were considered, according to adaptive thermal comfort limits. The results showed that uncertainties in degree-hours are high, i.e., 32% for cooling discomfort and 53% for heating discomfort. The uncertainty is largely due to the average monthly temperature of the ground. Other variables, such as the solar absorptance of the roof and the specific heat of the mortar of the walls had a relevant impact on heat discomfort. The density of the mortar and the specific heat of the brickwork walls affected the cooling discomfort. These variables need proper accuracy, either through more reliable databases and/or through field or laboratory measurements.

Keywords:
Building simulation; Thermal performance; Uncertainty analysis; Sensitivity analysis