Electrical Fault Diagnosis of Solar PV Array Using Machine Learning Techniques

Abstract

Because of the increased use of renewable energy in photovoltaic (PV) arrays as a secondary source of energy, they became susceptible to various types of faults, so that fault detection and diagnosis (FDD) become a necessary process for extending the life of these arrays, prevent power losses in the system and avoid the safety hazards resulting from it. This paper proposes to utilize four machine learning classifier techniques which are (Decision Tree (DT), K-Nearest Neighbors (KNN), Naive Bayes (NB), and Support Vector Machines (SVM)) to diagnose six common types of PV array faults (open circuit, intra-string line-to-line, inter-string line-to-line, line-to-ground, short-circuited bypass diode, and short-circuited blocking diode) occurring in a 4x4 PV array with a total power of 5kW. The data set collected from the proposed Matlab/Simulink model consists of 1210 samples and 29 features. The dataset is divided into 70% for training the models and 30% for testing them. The best model based on the SVM approach has achieved 99.7% classification accuracy with a training time of 94.42 seconds.