Detection and classification of faults in photovoltaic arrays using a 3D convolutional neural network

Abstract

Operations in an electric distribution system become challenging because the penetration of renewables (such as large photovoltaic (PV) arrays) becomes high nowadays. The detection and classification of faults in PV arrays are crucial for energy management and the mitigation of financial losses. The convolutional neural network (CNN) is one of the most popular deep learning-based methods for solving detection and classification problems. The use of CNNs in various fields has yielded promising results owing to their ability to extract features of signals. This study presents a 3D CNN for PV fault detection and classification. Both direct current (DC) and alternating current (AC) signals in the PV system are converted to 3D images by using the Gramian Angular Field (GAF) transform for signal pre-processing. The proposed method yields promising results in terms of overall accuracy (OA) of testing data. Simulation results indicate that the proposed 3D CNN outperforms other machine learning (ML) methods, such as k-nearest neighbor, Random Forest, Decision Tree, and Support Vector Machine, when applied to the problem of interest.