Predicting Cooling Energy Demands of Adaptive Facades Using Artificial Neural Network

Abstract

Adaptive Façades (AFs) have proven to be effective as a building envelope that can enhance energy efficiency and thermal comfort. However, evaluating the performance of these AFs using the current building performance simulation (BPS) tools is complex, time-consuming, and computationally intensive. These limitations can be overcome by using a machine learning (ML) model as a method to assess the AF system efficiently during the early design stage. This study presents an alternative approach using an Artificial Neural Network (ANN) model that can predict the hourly cooling loads of AF in significantly less time compared to BPS. To construct the model, a generative parametric simulation of office tower spaces with an AF shading system were simulated in terms of energy consumption using Honeybee add-on in Grass-hopper which are linked to EnergyPlus for training the ANN model. The prediction results showed a highly accurate model that can estimate cooling loads within seconds.