Santiago L. Torres, Yuzo Sakamoto
Bibliographic info:
Building Simulation, 2007, Beijing, China

The aim of the present study was to determine the applicability of a genetic algorithm for the optimization of daylighting systems, as well as the requirements for the lighting simulations to be used. Furthermore, by testing the daylighting performance of a building's facade when several parameters are allowed to change simultaneously, the results were used as a complement of previous parametric studies. The goal of the optimization was to maximize energy savings by reducing visual discomfort while maintaining good daylight penetration. The results were obtained by dynamic simulations using Radiance. Discomfort glare produced by daylight was calculated for several viewpoints inside the building, adapting the blinds' position accordingly for each time step tested. Fitness was defined as the proportion of the annual lighting requirements that can be replaced by daylight. The results show a fast convergence in the beginning, followed by a minimal improvement in subsequent generations. Several trials over 200 generations showed similar evolution and consistent results, suggesting that genetic algorithms can be used effectively for facade design optimization considering daylight performance. Finally, some possible improvements and modifications are further  discussed.