Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA

     

Machine learning techniques for the daylight and electric lighting performance predictions

Chantal Basurto, Oliver Paul, Jérôme H. Kämpf
Idiap Research Institute, 1920 Martigny, Switzerland

DOI: https://doi.org/10.26868/25222708.2021.30387
Abstract: The use of external blinds is a common strategy for the design of energy efficient buildings, its performance evaluation in this study involves an integrated assessment of daylight, electric lighting and blinds controls. Nowadays, such evaluations are mostly performed with the use of computer simulations, which, due to the complexity of the issue, are still highly demanding in terms of computing time and performance capabilities. In order to improve the response time of daylight and electric lighting performance-predictions, machine learning techniques are employed in this study as surrogate models. The workflow for producing daylight surrogate models from RADIANCE simulations was validated for an individual office room, and the obtained accuracy for predicting daylight performance resulted in 98.91% for work-plane illuminance (WPI) and 99.92% for daylight glare probability (DGP).
Keywords: Daylighting and electric lighting, Machine Learning, blinds and lighting controls, illuminance and glare index, RADIANCE matrix multiplication methods.
Pages: 2483 - 2490
Paper:
bs2021_30387