Optimizing Energy Consumption in Smart Buildings by Using Machine Learning Algorithms

Master’s Thesis Summary

Abstract

In this article, an approach for optimizing energy consumption in smart buildings using machine learning algorithms is presented. Utilizing TDSP, data on climatic conditions, occupancy, and energy consumption obtained from EnergyPlus software are integrated. Feature selection and feature importance techniques, as well as statistical analyses, are implemented to select variables that are used to train machine learning models such as MLP neural networks, support vector machines, random forest, and XGBRegressor for predicting energy consumption, with accuracy evaluated using RMSE. It was demonstrated that models based on neural networks offer better accuracy, thereby enabling measures to achieve energy optimization.