Machine Learning for Forecasting Building System Energy Consumption

Abstract

Today, the use and demand response of electric power has become increasingly important in industry, transportation, commercial and residential buildings. to better optimize the consumption of electrical energy and avoid as much as possible the loss of this source and bring a benefit for energy operators as well as for the end user, the prediction of electric energy with the use of the AI artificial intelligence science and the different prediction methods that have emerged under the field of machine learning namely ANN, SVM… becomes promoter and quite paramount. The purpose of this article is to introduce the principle and the relationship of ML and smart grid and to make a study of the state of the art of current electrical energy prediction methods and to compare all methods based on the latest proven research in the field of energy prediction at building level. In this paper the comparison and study is about measuring the accuracy of prediction with a low error percentage. There are also hybrid prediction methods that are improved over the single methods, also a new approach is proposed in this paper called SMEP (Smart Model of Electric Power Prediction).