Ensemble Learning Models Applied in Energy Time Series of a University Building

During 2020, the construction and operation of buildings globally accounted for more than a third (36%) of final energy used and 37% of carbon dioxide emissions. Hence, in the last decade, there has been a great interest in analyzing energy efficiency in buildings from different approaches. In this paper, machine learning-based black-box methods are proposed to predict the energy consumption of a university building. In related works, little has been explored on machine learning techniques such as decision trees and ensemble learning for predicting time series of energy consumption in homes and buildings. Hence, this work proposes an analysis of forecast accuracy and computational times of decision tree and ensemble learning techniques such as Random Forest and Extreme Gradient Boosting applied particularly to time series of the total active power of a university building. The results show that Random Forest presents the best forecast error metrics RMSE, MAE, and MAPE. However, this model ranks second in computational time, below the decision tree technique. These models can be beneficial in predictive control systems considered in buildings to forecast the behavior of buildings' energy consumption in the short term with outstanding precision.