Modelo híbrido baseado em séries temporais e redes neurais para previsão da geração de energia eólica

Abstract

Electric power generation through wind turbines is one of the practically inexhaustible alternatives for clean energy. However, advances in science and technology are still needed to ensure greater uniformity in generation and increase the participation of wind energy in the energy matrix. This thesis proposes the development of prediction models for wind speed and energy generation over different time horizons — ultra-short, short, medium, and long-term — using computational intelligence techniques, such as SARIMA models and Artificial Neural Networks (ANN), as well as hybrid models combining these approaches. The methodology was applied using meteorological data from the Petrolina/PE station, from the SONDA database, between 2004 and 2017. The results demonstrated that the proposed hybrid model showed better performance, especially for the hourly forecast horizon, significantly contributing to the reliability and security of wind energy integration into the electrical system.