A combined convolutional neural network model and support vector machine technique for fault detection and classification based on electroluminescence images of photovoltaic modules

Abstract

Nowadays, photovoltaic (PV) systems are gaining increasing momentum due to their ability to generate clean and affordable electric power. However, many factors can impede the production of the PV panels totally or partially. As such, it has become imperative to develop fault detection/classification models to ensure best operating conditions at the maximum energy conversion efficiencies. To address this challenge, a new model for detecting and classifying the faults in electroluminescence images of PV panels has been proposed in this paper. The model combines two machine learning algorithms named Convolutional Neural Network (CNN) and Support Vector Machine (SVM) that are employed for features exaction and classification, respectively. The current model is trained and evaluated using two databases D1 and D2 that contains the electroluminescence images of PV cells. By comparing the proposed model with the previous similar works, this study demonstrates that the CNN combined with SVM provides a higher classification performance with an accuracy of 99.49 % and 99.46 % for databases D1 and D2, respectively.