Classificação de faltas em plantas fotovoltaicas com aplicação de inteligência artificial

Abstract

Photovoltaic generation has proven to be a significant alternative in the realm of renewable energy sources, offering flexibility for application in residential settings as well as large-scale photovoltaic plants, serving as a strategically important complementary source for extensive electrical systems. However, this form of generation poses a challenge for fault detection using conventional protection and classification systems. An alternative approach to address this challenge involves the application of Artificial Intelligence techniques for fault classification, yielding benefits in terms of system longevity and maintenance optimization. Common faults in photovoltaic systems were addressed, along with modeling for the implementation of computational models and the application of multilayer perceptron neural networks for fault classification. During the studies, tests were conducted to define the neural network hyperparameters using bayesian optimization and random search techniques, and comparisons of results were made based on accuracy. The obtained results contribute to the understanding and improvement of fault detection in photovoltaic systems through the effective application of neural networks.