Detection of Hot Spots in PV Modules with the CBAM-Enhanced Yolo Model

In photovoltaic (PV) systems, early detection of hot spot faults is important for energy efficiency and reliability. In this study, a YOLOv8s based model is proposed for hot spot detection on thermal images of PV panels obtained with infrared cameras. The Convolutional Block Attention Module (CBAM), which is an attention mechanism, is integrated in order to increase the fault detection capability of the model. The developed model achieved results of 0.938 in mAP@ 0.5 metric and 0.491 in [email protected]:0.95 metric. In addition, the recall metric of the developed model is 0.913 and the F1-score metric is 0.891, providing higher accuracy and stability in fault detection. The obtained findings demonstrate that integrating attention mechanisms into lightweight object detection models is an effective approach for thermal image-based fault diagnosis in PV systems. In future studies, it is aimed to examine the fault detection performance of the YOLO algorithm by applying it on electroluminescence (EL) images.