A Smart Step-by-Step Method for Fault Detection and Severity Assessment in Photovoltaic Arrays

This paper proposes a step-by-step fault identification model which includes five steps for detection, classification and severity assessment of electrical faults at the DC side of PV arrays. In this model, three different machine learning (ML)-based classifiers namely Support Vector Machine (SVM), Naïve Bayes (NB), and Logistic Regression (LR) are employed. The proposed method consists of two main substages in each step, one is to implement the feature selection algorithm, which is sequential forward selection (SFS) in this study, in order to select the best features for each classifier, and the other is to select the best classifier based on performance criteria. To train the classifiers, PV array current-voltage (I-V) characteristic curve is carefully explored and consequently sixteen features are extracted under various faulty and normal conditions. The results of the simulation study show high accuracies in the processes of fault detection, classification and severity assessment.