Modeling of Energy Management Systems using Artificial Intelligence

Using Machine Learning techniques is possible to analyze larges amounts of data collected by Energy Management Systems, which can be used to, for example, predict the energy bill. This paper presents the employment of Machine Learning techniques to numerically model the electrical consumption of one building of Ministries' Esplanade (headquarter of the Brazilian Executive Government), alongside with the applications of such models to Energy Management Systems. The Machine Learning model consists of an ensemble of 6 techniques, namely: Ridge Regression, Random Forest Regression, Extremely randomized trees, Gradient Tree Boosting, Support Vector Machine, and Artificial Neural Network. The 6 techniques were trained in parallel on historical data from 2018 and 2019 obtained from the PGEN project, which describe the building's instant characteristics through 12 electrical and climatic variables, and the dataset is entirely available. The ensemble model predicts the building's power consumption (averaged in a period of 10 minutes) with an error of MAE=9.32 KW (equivalent to 5.10% of the mean value of the power) and were listed four practical applications to the model.