Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Data-driven transfer function modelling of building dynamics using a frequency domain approach

Sophie Bernard 1,2, Valery Ann Jacobs 1, Bert Belmans 1,3, Arjen Mentens 1, Filip Descamps 1,4, John Lataire 2
1 Vrije Universiteit Brussel, Belgium
2 Université libre de Bruxelles, Belgium
3 Universiteit Antwerpen, Belgium
4 Daidalos Peutz, Belgium

Abstract: 1 In this research, a modelling tool is presented to derive surrogate models of thermal energy transfers in buildings, to support the development and testing of smart control algorithms. A data-driven approach is used to identify a model that is able to predict the indoor temperature in a case-study building when an electric heater is turned on. The approach is demonstrated on data obtained from EnergyPlus simulations, which resolve the heat balance equations to simulate the thermal response of a building. The model structure that was selected is a second-order ARMAX transfer function whose parameters were identified with a Least Squares optimisation criterion. The model inputs were limited to the heater's power, the global horizontal solar radiation and the outdoor dry-bulb temperature.
Keywords: Building thermal modelling, EnergyPlus, system identification, black-box model, surrogate modelling
Pages: 1810 - 1817