Fault Diagnosis of Photovoltaic Modules Using Infrared Images and Hybrid DCNNs: A Comparative Study

With the expansion of solar energy in various fields, effective and reliable fault detection and diagnosis techniques have been developed to ensure the safety and efficiency of photovoltaic (PV) installations. Recently, fault diagnosis of PV plants based on infrared thermography (IRT) images has attracted the attention of researchers working on this topic. In this manuscript, we consider a deep convolutional neural network (DCNN)-based model combined with five hybrid classification models, in order to classify the severity of photovoltaic modules based on the captured IRT images. The hybrid classifiers examined are DCNN-SVM, DCNN-XGBoost, DCNN-RF, DCNN-KNN and DCNN-DT. The results show that DCNN, DCNN-SVM and DCNN-XGBoost provide relatively the same results in terms of classification accuracy (87%), while the other hybrid classifiers offer lower accuracy.