The use of energy for space cooling is growing faster than any other end use in buildings, justifying the search for improvements in the energy efficiency of these systems. A simplified model to predict cooling energy consumption in Brazilian office buildings was developed. Artificial neural networks (ANNs) were trained from consumption data obtained by building simulation. As it is intended to be applicable to different climates, a new climate indicator also appropriate for hot and humid climates was proposed and validated. The Sobol sensibility analysis was performed to reduce the number of input factors and thus the number of cases to be simulated. The data was built with the simulation of 250,000 cases in Energyplus. Studies were conducted to define the sample size to be used for the ANN training, as well as to define the best ANN architecture. The developed metamodel was used to predict the consumption of Heating, Ventilating and Air Conditioning (HVAC) system of 66,300 new unseen cases. The results showed that the new proposed climate indicator was more accurate than the usual climate correlations, such as cooling degree hours. The developed metamodel presented good performance when predicting annual HVAC consumption of the cases used to obtain the model (R2 = 0.9858 and NRMSE = 0.068) and also of the unseen cases (R2 = 0.9789 and NRMSE = 0.064).
End use energy consumption data base: transportation sector