Evaluation of energy forecasting mechanisms with renewable sources for maximizing the brazilian energy matrix using machine learning

Abstract

The present study focuses on the Brazilian energy scenario and highlights the progressive increase in the use of renewable sources in the country's electricity matrix. The main objective of this study is to contribute to the search for solutions and to stimulate debates and reflections on the future actions necessary for energy planning. To achieve this, the research employs computational tools based on machine learning and data mining, using government and energy market data sources. The research methodology encompasses the use of computational tools to project the forecast of the electricity market in the country. The methodology employed includes the execution of forecasting models, highlighting the behavior of the energy market over time, using methods such as Multilayer Perceptron Neural Networks (MLP), Gaussian Process Regression (GPR), and Linear Regression to project electricity generation by source in Brazil. The results indicate a considerable growth of renewable sources in the national energy market by the year 2030, approaching the goal of the Ten-Year Energy Expansion Plan to achieve 90% renewability, covering sources such as hydroelectric, biomass, wind, and solar. The Linear Regression method achieved 86% renewability, while the GPR method reached 90%, and the MLP method reached 88%. The projection of the electricity market forecast allowed for the identification of market behavior patterns, enabling the anticipation of trends and changes in the market. These forecasts aim to provide information to support the development of actions in the energy planning process, contributing to the transition to more sustainable and renewable energy sources in Brazil.