In the lack of experimental data, one of the most elaborate building simulation codes, the TRNSYS software, was used to calculate the cooling loads for the selected residential
buildings. The use of such an advanced software (i. e. transient, multi-zone and interactive) was necessary as the reliability of simulation results strongly affects the performance of the, thereinafter, used ANN.

The weather inputs to TRNSYS include external dry bulb temperature, relative humidity, total and diffuse solar radiation on the horizontal and wind speed. The simulations were carried out for a period of one year, using the Typical Meteorological Year of Athens.

Five residential building are selected for this study. Table 1 presents some general details concerning these buildings. The building materials used in the models were based on the construction materials of each building.

In residential buildings there are some crucial factors, which are quite difficult to be measured or estimated, mainly due to inhabitant’s influence. In office or commercial buildings, the use and operation of buildings are in many cases well defined, making feasible to determine workings hours and estimate parameters like ventilation, lighting or other internal gains. Unlike this, residential buildings are highly dependent on the, often without a specific pattern, presence of occupants and more over on their personal habitudes, making the estimation of above factors a complex task.

This being the case, the validation of simulation results was thought necessary. To this effect, measurements of the indoor and ambient air temperature were performed during a period of two months, for April to May when heating and cooling loads are minimal and the building operate under free floating conditions. Figures 1 and 2 show the comparison of TRNSYS simulated and measured values, which are in very good agreement.

Fig. 1. Comparison of TRNSYS and actual values of indoor air temperature for buildings: House 1 and House 4