Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Using a surrogate model to analyze the impact of geometry on energy efficiency of buildings.

Bhumika Bhatta 1, Ralph Evins 2, Paul Westermann 3
1 University of Victoria, Canada
2 University of Victoria, Canada
3 University of Victoria, Canada

Abstract: Parametric exploration and optimization of building geometry is a powerful tool for designing energy efficient buildings. However, in practice this process is computationally expensive and time-consuming. In this research, we explore the use of surrogate models, i.e. efficient statistical approximations of expensive physics-based building simulation models, to lower the computational burden of large-scale building geometry analysis. For this purpose, we developed a novel dataset of 38,000 residential building models derived from real world floor plans from (Wu et al. (2019)) and train a surrogate model to emulate their simulated annual energy performance. We extract up to 20 parameters as surrogate model inputs to represent the building geometry and show that the trained surrogate model reaches a high accuracy (R2 score = 0.999, MSE = 0.007 and RMSE = 0.022) on test data. The current setup forms the basis for further research where the complexity of the building models will be increased.
Keywords: Energy efficiency, Building geometry, Heating and Cooling demand, Surrogate Model, Machine learning.
Pages: 1833 - 1840