Convolutional versus dense neural networks: comparing the two neural networks’ performance in predicting building operational energy use based on the building shape

A building’s self-shading shape impacts substantially on the amount of direct sunlight received by building and contributes significantly to operational energy use, in addition other major contributing variables, such as materials window-to-wall ratios. Deep Learning has potential assist designers engineers efficiently predicting performance. This paper assesses applicability two different neural networks’ structures, Dense Neural Network (DNN) Convolutional (CNN), for use with respect shape. The comparison between networks shows that DNN model surpasses CNN performance, simplicity, computation time. However, image-based benefit utilizing architectural graphics facilitates design communication.