Desenvolvimento preliminar de um pacote para previsão automática de potência solar baseado em dados históricos do INMET

As a clean energy alternative, the installation of power plants that utilize renewable sources of electricity has gained prominence in Brazil’s energy sector, already accounting for 48% of installed capacity. Among the various existing options, photovoltaic solar energy has been growing exponentially, especially in the context of electrical power, through the implementation of photovoltaic panels. However, incorrect sizing, meaning when designed without proper preliminary analysis, of these grid-connected panels can cause disturbances and technical losses in the existing distribution network. Therefore, there is a great need for reliable preliminary tools that facilitate analysis for proper sizing. Consequently, this study presents the development of a preliminary tool for analyzing data provided by INMET (National Institute of Meteorology) that directly influence the generation of photovoltaic energy, followed by appropriate data treatment, particularly the detection of outliers, which can be used in the future for historical series forecasting models. The entire process is designed to occur automatically using the Julia programming language. The methodology was developed in three parts: data acquisition, data treatment, and outlier detection. The employed method allowed for graphical visualization of the identified outlier points, as well as filtering the data set to obtain more reliable data without the presence of anomalies.