This thesis aims to develop methods to obtain representative building stock models to benchmark the energy performance of Brazilian schools considering building-level and stock-level perspectives. A literature review regarding the main research gaps in operational building performance was carried out, contrasting both building and stock-level investigation perspectives. Then, an overview of the school building stock in Brazil outlined the main characteristics regarding the energy performance, and a representative building stock database was composed. The database was used to build a top-down building stock model, resulting in a model able to incorporate thermal satisfaction of occupants in the benchmarking classification through machine learning. Furthermore, a framework was proposed to model representative archetypes through entropy and cluster analysis. Artificial Neural Networks were used to model the energy benchmarking model. Results showed a manageable bottom-up model of the building stock able to perform a reliable representation of the actual stock performance. The bottom-up building stock model was used to predict the performance of the building stock performance under unseen conditions, i.e. future climatic conditions and different scenarios of air-conditioning. Five research articles were written to report the research. Conclusions of the thesis outlined the determinant factors that impact the energy performance of the school building stock in Brazil and the suitability of both top-down and bottom-up models to represent the building stock, according to specific purposes. The two stock modelling methods proposed employ different metrics to solve different problems. The top-down method provided a single performance scale to include occupants’ aspects in buildings operational performance evaluation. The bottom-up method is adequate to rate the building performance under standard conditions. Thus, through the models proposed it is possible to evaluate further conditions, such as future climates, in the buildings-level perspective that potentially impact the energy consumption at a stock-level scale.
Veja a defesa da tese clicando aqui.