Estratégia para classificação de efeitos em arranjos fotovoltaicos através de redes neurais artificiais

Abstract

The characteristic curve of a photovoltaic module is na important parameter to verify the operational condition of any photovoltaic system. In addition, to each voltage and current characteristic, there is a mold of the I-V curve. This curve modifies with the parameters change, such temperature and irradiance. Besides, there are intrinsic parameters of the photovoltaic modules, such, the series resistance and the shunt resistance, which change the shape of the I-V curve. For which defect that may happens in a photovoltaic array, the I-V curve has a pattern, changing the I-V curve parameters. These defects in a system reduce the array generate power, cause the decrease on useful life and cause irreversible damages on the photovoltaic cells. Therefore, there is a need to identify the photovoltaic array defects, the defects are chosen according with the standard NBR 16274. The defects identification is based on the I-V curve, the array irradiance and the array datasheet, therefore, an artificial neural network, such, multilayer perceptron with one hidden layer is used for classification of the chosen conditions. The proposed system results shown that the neural network was able to classify the defects with an accuracy of 99,3% and moreover the system has a compact structure.