Classificação Multiestágio Aplicada Ao Monitoramento Não-Intrusivo De String Fotovoltaica

Abstract

The great current demand for photovoltaic electric energy generation has required the search for solutions to problems of operation of such systems. Among these, the so-called atypical photovoltaic string operating conditions stand out, leading the system to a loss of efficiency in its generation capacity or even to severe failures. Regardless of the cause of the atypical string condition, its effect is somehow registered in the electrical generation itself. Each causative factor results in an electrical signature, which allows it to be identified. Thus, the objective of this article is to identify the operating condition of a photovoltaic string among twenty possible states, one is the normal condition, and the others are atypical ones. It is used a non-intrusive monitoring method based on the use of electrical voltage and current samples generated by the PV string itself. A multistage classification methodology is adopted to divide a complex problem into smaller and less complicated sub-problems. Therefore, they are considered two classification stages and, in each one, the K-nearest Neighbors, Support Vector Machine, and Multi- Layer Perceptron classification technics are used. The results achieved led to an average accuracy of 93.9% when using the classifier with the best performance in each subproblem treated.