Predição de séries temporais de radiação solar utilizando modelos estatísticos e aprendizagem profunda na região de Quixeramobim - CE

Abstract

Weather forecasting is essential in the renewable energy sector, and it is indispensable to use climate information such as air humidity, atmospheric pressure, temperature, wind speed and solar irradiation, which can be considered variables for forecasts in a certain region. In particular, there is a notable potential for the use of photovoltaic solar energy in the Brazilian Northeast due to the high solar irradiation levels in the region and, from the climatic time series, it is possible to train deep learning models that seek to predict it at short and long term. This work aims to statistically analyze and predict a time series of total daily solar irradiation in the municipality of Quixeramobim, Ceará, through machine learning methods. The results obtained indicate that the models predict solar irradiance with low prediction error when compared to existing results in the literature.