Técnicas de redes neurais artificiais aplicadas na previsão de geração de uma usina fotovoltaica

Abstract

The insertion of photovoltaic (PV) systems in electricity distribution networks, through distributed generation, is promising, as it guarantees greater availability of electricity to the distribution system. However, as it is an energy source with these intermittent characteristics, limited by climatic variations, the dissemination of large-scale systems can generate instabilities in the network. In order to predict the output power of these PV systems, research related to energy generation prediction methods was developed. Obtaining an accurate forecast can help to stabilize and protect the operation of the grid, in addition to enabling the insertion of photovoltaic energy on a large scale. This work aims to present techniques of Artificial Neural Networks (ANN) as tools for predicting photovoltaic energy generation. As a differential, a Multilayer Perceptron Network (PMC) was used combined with a preliminary classification of the input data, which validated the hypothesis that this previous classification has an influence on the ANN results.