Hotspot defects in solar photovoltaic arrays significantly impact energy efficiency and pose safety risks, necessitating automated detection solutions to replace time-intensive manual inspections. This study presents a comparative analysis of YOLOv8 and YOLOv11 instance segmentation models for hotspot detection in thermal imagery. A dataset of 766 manually annotated thermal images from solar farms in Kelantan, Malaysia, captured via drone-based inspections, was used for model training and evaluation. Multiple variants of both YOLO architectures were systematically compared across precision, recall, F1-score, mean Average Precision (mAP), and computational efficiency metrics. Results demonstrate that YOLOv11x achieves superior performance with the highest mask mAP@50 of 0.987, balanced precision of 0.959, and recall of 0.970, significantly outperforming YOLOv8 models. While YOLOv8 models generally process faster, YOLOv11x provides optimal accuracy-speed balance at 30.4 FPS, making it suitable for real-time monitoring applications. The study concludes that YOLOv11x is the most effective model for automated hotspot detection in solar panel monitoring systems, offering superior detection accuracy essential for preventing system failures and ensuring operational safety.
| Item Type: | Article |
|---|---|
| Creators: | Creators Email / ID Num. Khiruddin, Nur Raihah UNSPECIFIED Daud, Kamarulazhar UNSPECIFIED Osman, Muhammad Khusairi UNSPECIFIED Ab Hamid, Mohd Zulhamdy UNSPECIFIED Ismail, Ahmad Puad UNSPECIFIED Abd. Samat, Ahmad Asri UNSPECIFIED Ahmad, Fadzil UNSPECIFIED Ismail, Lyly Nyl UNSPECIFIED Jadin, Mohd Shawal UNSPECIFIED Kanata, Sabhan UNSPECIFIED |
| Subjects: | T Technology > TA Engineering. Civil engineering > Applied optics. Photonics > Optical data processing T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Production of electric energy or power > Production from solar energy |
| Divisions: | Universiti Teknologi MARA, Shah Alam > Faculty of Electrical Engineering |
| Journal or Publication Title: | Journal of Electrical and Electronic Systems Research (JEESR) |
| UiTM Journal Collections: | UiTM Journals > Journal of Electrical and Electronic Systems Research (JEESR) |
| ISSN: | e-ISSN : 3030-640X |
| Volume: | 29 |
| Number: | 1 |
| Page Range: | pp. 122-131 |
| Keywords: | Computer Vision, Hotspot detection, Instance segmentation, Solar photovoltaic, Thermal imaging, YOLOv8, YOLOv11 |
| Date: | 31 October 2026 |
| URI: | https://ir.uitm.edu.my/id/eprint/149163 |
149163.pdf
