Optimized parameter selection for assessing building energy efficiency

This study analyzes the influence of various physical parameters on estimating the heating and cooling load requirements of a building. Building energy models represent an important starting point for research in advanced building systems. In order to determine the most relevant combination of parameters for assessing energy use in buildings, we have created a framework which combines a stochastic classification method, namely a Gaussian Mixture Model, with a combinatorial optimization procedure. The framework was evaluated using a benchmark database consisting of 768 simulated buildings for which annual data on energy demands related to heating and cooling were available. The results show that feature selection is an important phase in developing building models by decreasing the computation time while still allowing for a high accuracy in the prediction of energy demand.