Ensemble Deep Learning Applied to Predict Building Energy Consumption

Buildings use energy and produce carbon dioxide considerably. Despite progress in economics, this trend of rising emissions continues. Research on building energy consumption allows for determining building energy efficiency and developing energy-saving strategies. Additionally, it helps to forecast trends in future building energy consumption. The research on building energy consumption has emerged as one of the critical issues for achieving carbon neutrality. Therefore, we propose an ensemble deep learning applied to predict building energy consumption. The proposed algorithm employs ensemble architecture to improve deep learning's effectiveness in predicting error of the reduction in building energy consumption. Furthermore, with negative correlation learning (NCL), learning across all samples is improved. The dataset from the American Society of Heating and Air-Conditioning Engineers (ASHERE) is used to compare several approaches, including the proposed algorithm, deep learning, decision trees, and linear regression. The results demonstrate that the proposed algorithm enhances and lessens the prediction of the root mean squared error (RMSE). Among the approaches compared, the proposed algorithm has the lowest RMSE. The proposed ensemble deep learning algorithm outperforms the other approaches compared.