Classification of defect photovoltaic panel images using deep learning in computer vision

As one of renewable energy sources, the uses of solar panels is seen to be more widespread nowadays along with the development of technology. Large-scale maintenance has long been seen as a great challenge and needs attention. Currently, electrical performance measurement or image processing is used to carry out condition monitoring of photovoltaic panels. This method has the limited ability to detect defects, is time-consuming and has the inability to determine the exact location of defects quickly. To overcome this challenge, deep learning techniques are used for detection in this classification tasks. This research is focused on classifying defect PV panels using Matrox Imaging Library software which are to be installed in computer vision applications. This application provides a deep learning algorithm that is able to classify images according to its respective group. The image data set has been carefully compiled and divided into training and development datasets during the training to ensure the highest accuracy for the prediction of the presence or absence of defects on PV panels. A statistical measure that is the average accuracy rate for the training model and the total prediction was implemented to evaluate the classification performance of the defect PV panel models. The results show a remarkable total model accuracy of 99.9% for each class and the prediction results confirms show that nearly 90% of PV panel defects are detected from the test dataset. Furthermore, a comparative analysis was conducted to benchmark the findings against other algorithms. The findings of this research demonstrate the effectiveness of deep learning algorithms and its compatibility in computer vision applications in use. By leveraging this technique, solar panel users can improve maintenance management, control quality and reduce financial losses by promptly identifying and addressing panel defects.