Analyses of Photovoltaic Systems Performance with Fuzzy Inference Systems and Bayesian Networks

Renewable energy sources have been increasingly targeted as an alternative for reducing fossil fuel emissions. This scenario requires a better understanding of energy systems' performance and limitations, notably using photovoltaic setups under diverse meteorological conditions and determining device measurement uncertainties. Accordingly, this study conducts a performance evaluation of a photovoltaic system of copper, indium, and gallium (di)selenide technology. It compares the results provided by a reference normalized equation with two interpretable artificial intelligence techniques, Fuzzy Inference System and Bayesian Networks, when temperature is also considered. In the experiments, the study found significant percentage differences between the normalization procedure and the fuzzy system on days with lower irradiation values. On the other hand, the results provided by Bayesian networks presented lower difference percentages in relation to the normalization equation values.