Application of Multilayer Extreme Learning Machine for Efficient Building Energy Prediction

Building energy efficiency is vital, due to the substantial amount of consumed in buildings and associated adverse effects. A high-accuracy prediction model considered as one most effective ways understand building efficiency. In several studies, various machine learning models have been proposed for However, existing are based on classical approaches small datasets. Using a dataset inefficient may lead poor generalization. addition, it not common see studies examining suitability methods forecasting consumption during early design phase so that more energy-efficient can be constructed. Hence, these purposes, we propose multilayer extreme (MLELM) annual consumption. Our MLELM fuses stacks autoencoders (AEs) with an (ELM). We designed autoencoder ELM concept, used feature extraction. Moreover, were trained layer-wise manner, employed extract efficient features from input data, was using least squares technique fast speed. decision making. this research, large residential capture sizes. compared other commonly predicting From results, validated outperformed comparison prediction. experiments study, identified predictive use before construction, which make informed decisions about, manage, optimize construction.