Login
By 2050, the European council proposed to achieve total decarbonization in buildings. In this way, building energy models are key factors to predict the energy consumption in the design, use and retrofit stages. However, these models may present a relevant gap between predicted and measured energy performance, which should be minimised by cutting uncertainties with real data. Air leakage is one of the main uncertainties and causes of increasing building loads by renovating the indoor air in an uncontrolled way. Nevertheless, many energy modellers do not have a solution for this parameter.
Therefore, the two main goals of this study are to find the most accurate dynamic infiltration model and to verify if it can be extrapolated to different periods and wind data. For this reason, an experiment of tracer gas with CO2 was carried out in the south room of a flat in Pamplona, Spain. The experiment was conducted for 40 days, 18 in summer (9 for training and 9 for checking), 11 in winter and 11 in spring for checking. The Design Flow Rate EnergyPlus object was chosen to calculate infiltration, which, in turn, fed the multi-points decay equation to generate the simulated CO2 curve. Then, to find the best coefficients of this object, the performance of multivariable regressions was done based on the objective function of minimising the mean absolute error between predicted and measured CO2 concentrations. As wind plays an important role in the calculation of air leakage, this process was made using different wind data: one from in-situ sensor and three from a nearby meteorological station (a global wind with all directions, a westbound wind and an eastbound wind), in order to analyse which one was the best to predict the air leakage. The most precise training model was applied in the checking periods to test its robustness to time and wind data. To evaluate these models, the ASTM D5157 Standard Guide for Statistical Evaluation of Indoor Air Quality Models and Taylor Diagrams were used.
As a result, the models created from the in-situ data and from the west wind of the weather station best represent the measured CO2. They present 14% better performance than the model generated with the global wind from the weather station, the latter usually applied in building energy simulations. The in-situ wind data developed coefficients specific to the test space that can be extrapolated to other seasons and weather conditions without losing their quality. Even models that did not meet ASTM D5157 criteria in the training period passed the standard with in-situ coefficients. This study is a step forward in reducing the infiltration uncertainty and corresponds to a cost-effective solution, since with only 9 days of training, it is possible to obtain coefficients that generate accurate air leakage values at other seasons and with wind from the weather station, which is easier to collect than in field measurements.