Energy consumption prediction has been recognized as one of the key strategies to develop energy efficiency policies. The objective of this research was to develop predictions of energy consumption and compare which one of them presented the best performance, especially focusing on the residential sector. For this, four Machine Learning (ML) models, named Linear Regression, Random Forest, XGBoost, and K-Nearest Neighbors, were established to predict the energy consumption of Brazil, considering a dataset of registers from 2004 to 2024. Then, they were compared considering four performance metrics named R-Squared (R2), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). The results showed accurate predictions from the four models, highlighting Random Forest and XGBoost as the ones with the best performance (R2 > 0.95). This research contributes to enhancing the usage of Industry 4.0 technologies, such as Machine Learning, to accomplish predictions. It helps in the decision-making process of energy systems managers, especially in the Brazilian context or in regional levels, and permitting to create more accurate energy consumption policies towards increasing the sustainability levels for the populations.