Rede Neural Artificial Aplicada ao Rastreamento de Pontos de Máxima Potência de Painéis Fotovoltaicos Com Sombreamento Parcial

Abstract

In photovoltaic panels, the maximum power point depends on the levels of solar irradiance and temperature. Determining the maximum power value is a complex task, especially when the irradiance is not uniform, i.e. when there is partial shading on the panel. This work applies an artificial neural network (ANN) to identify the point of operation with maximum power in photovoltaic panels under different conditions of shading (solar radiation), total or partial. The solar irradiance (given in W / m²), temperature (° C) and the open circuit voltage of the panel (V) are used as inputs for training and RNA testing. The information sought at the output of the artificial neural network is the voltage value (V) to be imposed to operate at maximum power. In the study, by simulation, an MLP (Multiple Layer Perceptron) type RNA was developed with two hidden layers of 13 neurons each. For ANN training were used 554 samples of parameters, and 92 for verification, generated from a model developed in Matlab / Simulink® software. The results of the tests returned an error rate around 6%, showing that the artificial neural network with the proposed approach can be an efficient solution to detect the maximum power point in several shadowing situations.