Prediction of cooling load via machine learning on building envelope design parameters

Abstract

Energy demand in buildings has increased significantly in recent years. With this increase, ensuring energy efficiency in buildings and accurately estimating energy performance is critical for sustainable construction and energy management. Existing machine learning (ML)-based methods in the literature are generally developed with limited data sets, which limits the accuracy of the models. This study applies machine learning techniques using an extensive data set to estimate the annual cooling loads of residential buildings. In this context, a large data set containing 12960 scenarios was used, and the scenarios were created by changing the wall layers, plan type, orientation, and window type through simulation programs using simulation-based calculation. This study aims to develop an accurate and reliable model using machine learning algorithms to estimate the energy consumption of residential buildings at the early design stage. The size and diversity of the data set used to estimate energy performance provide more reliable and generalizable results, unlike previous studies in the literature. The study proposes the artif icial neural network (ANN) model that provides the best prediction performance by comparing various machine learning algorithms such as support vector machine (SVM), random forest (RF), decision tree (DT), linear regression (LR) and sub-models. That model showed the best performance with a 98 % accuracy rate. These results increase the effectiveness of machine learning techniques in predicting energy consumption of buildings and significantly contribute to predicting energy performance, especially in early design stages.