Use of ARIMA and SVM for forecasting time series of the Brazilian electrical system

• Autor(a) principal: Nunes, Lucas Renan Maués
• Data de Publicação: 2023
• Outros Autores: Veras, Juam Sousa, Silva, João Pedro Ribeiro, Conte, Thiago Nicolau Magalhães de Souza, Santos, Wilker José Caminha dos, Oliveira, Roberto Célio Limão e
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/40438
• topic: ARIMA; SVM; Aprendizaje automático; Serie de Tiempo; Sector Eléctrico Brasileño.; ARIMA; SVM; Aprendizado de Máquina; Séries Temporais; Setor Elétrico Brasileiro.; ARIMA; SVM; Machine Learning; Time Series; Brazilian Electric Sector.

The present work proposes to forecast time series of the Brazilian electricity sector. For this purpose, an attempt was made to make predictions for the Settlement Price of Differences (PLD) and the wind speed for moving wind turbines, which transforms the kinetic energy of air currents into electrical energy, based on the ARIMA methodology, based on statistics computational, and the SVM model, from the area of ​​artificial intelligence, and the period analyzed corresponds from 2001 to 2009 for the PLD and from 2004 to 2017 for the wind. The results provide an analysis tool for the free energy market, as they demonstrate price trends and electricity production, serving as an aid to decision-making, with ARIMA being the predictive model that performed best in short-term forecasts. Despite this, it is concluded that the SVM has the potential to produce more assertive results for long-term forecasts, since the model has many characteristics that can be exploited and thus enhance forecasts with large volumes of data in more complex situations.