Multivariable Optimization of Dynamic Translucent Solar Screen on West-Facing Offices

Abstract

Discomfort glare frequently occurs in the west-facing rooms of office buildings during the afternoon. Using conventional opaque and fixed solar screens can obstruct the direct solar radiation, and reduces the glare. However, they may considerably reduce indoor daylighting levels and block the view outside at the same time. Thus, dynamic translucent panels are proposed in this study as external solar shading screens on the west-facing facades. Most importantly, a systematic method is developed to optimize the positions of the panels, addressing both daylighting, glare and view. The optimization algorithm - Artificial Neural Networks, combined with daylighting performance simulations is used to determine the optimal design parameters of the proposed solar shading screens. An office room in Guangzhou is used as a case study to test and optimize the proposed method. The whole study process consists of three steps: simulation, regression, and optimization. The results show that the neural network regression results are very close to the simulation ones. It means that deep neural networks can achieve a relatable regression performance and reduce the optimization time significantly. The optimal positions of the solar screens effectively reduce the glare level and maintain a relatively high indoor daylighting level and a large view field. Furthermore, this study also discusses the different effects of the weights for evaluation indicators on the final optimization results.