Sistemas fotovoltaicos de geração de energia: modelo de previsão de curto prazo a partir de dados meteorológicos

Abstract

This thesis, carried out by students of the Control and Automation Engineering course, has the premise to verify the possibility of carrying out a system for predicting the efficiency of a photovoltaic panel using data from a local solarimetric station. For this, data from the years 2021 and 2022 from a station in Curitiba, provided by the Solar Energy Laboratory of the Federal Technological University of Paraná (LABENS) were used. Linear models were carried out using Excel, as well as the prediction of future efficiency of solar panels. A linear regressive model (ARX) was also performed as a proof of concept, using data from one month. It was found that the models have a high degree of mathematical complexity due to the mathematical chaos, as well as the large computational weight of the models, making accurate models computationally costly. It was observed that the models generated with data referring to a shorter period of time, with a forecast of efficiency of the panels in the short term, have a higher degree of reliability than the models generated with data from large periods of time. It was also observed that the choice of input variables is extremely important, since models generated with linear inputs have a better result than with linear and non-linear inputs. This makes the modeling of systems based on meteorological data extremely difficult, requiring high-performance computers to generate accurate models.