Machine Learning-Based Predictive Models for Energy Consumption Estimation in Energy-Efficient Building Envelope Design

Recently, the construction of energy-efficient buildings has gained increasing importance. The estimation of a building's energy consumption, which takes into account envelope parameters such as wall type, glass type, window-to-wall ratio, orientation, and others, is crucial at the project's early stage for managers. Currently, building energy estimation methods rely on mathematical formulas or simulations using specialized energy BIM software. However, these initial estimates are often inaccurate due to the lack of detailed BIM models, resulting in an inefficient and challenging process of energy analysis during the early design stage of the project. This research employs various machine learning techniques, including Support_Vector Machine, Artificial_Neural_Network, Generalized Linear Regression, Deep_Learning Neural Network (DLNN), Random_Forest, and Gradient_Boosting to predict a building's preliminary energy consumption. These machine-learning models were trained and tested on data gathered from simulations using the BIM-Design Builder software. Comparative results show that Gradient Boosting, an ensemble learning technique, outperforms all other machine learning algorithms in terms of accuracy and performance. Based on these findings, energy estimation experts can more efficiently select the best model for predicting a building's preliminary energy consumption during the early design stage of the project.