Previsão da Irradiação Solar utilizando Aprendizagem de Máquina com Seleção de Características por Algoritmo Genético

Abstract

This study explores solar irradiation forecasting by integrating feature selection techniques with genetic algorithms. Meteorological data from Fortaleza is employed for training and evaluation. Various machine learning algorithms, including support vector regression, k-nearest neighbors, random forest, adaptive boosting, gradient boosting trees, and extreme gradient boosting trees, are utilized to predict solar irradiation. Performance evaluation metrics such as MAE, RMSE, MAPE, and R² are employed to assess model performance. The results demonstrate the effectiveness of the genetic algorithm in selecting model inputs, highlighting Extreme Gradient Boosting as the most effective across various forecast horizons.