Surrogate Model Development for Naturally Ventilated Office Buildings

Building energy simulation tools are very helpful to achieve thermal performance for buildings. However, modeling can require much detail, specially related to input data. The use of machine learning to develop surrogate models can support architects and builders to get useful information on buildings thermal performance, in a fast and simple way. The aim of this study is to present a machine learning methodology to develop a surrogate model for naturally ventilated office buildings, using artificial neural networks. The output of the surrogate model is the Exceedance Hour Fraction (EHF), a thermal comfort indicator. The final surrogate model has 12 input parameters that can estimate thermal comfort for offices with a wide range of characteristics. The mean absolute error measured for the surrogate model was 0,04.