全球能源危機帶動再生能源的需求日益增加，引進新的科技，解決問題。光伏技術是可持續的潔淨能源，是全球能源挑戰的最具有潛力的方案之一。因此，光伏模組被大量採用，模組與系統定期維護是維持持高效能運轉必要的工作。大型光伏電廠位於人員攀爬不易的位置，維護作業危險性高且效益不佳，利用無人機(UAV)提供維護鑑驗是最適切的技術。
本研究利用無人機從光伏電廠收集數據，從紅外線熱成像(Infrared Thermal Image, IRTI)與(Red Green Blue, RGB)影像分光及至數據庫建立與辨識分析，建立一套完整的技術架構。本研究將以人工智慧(Artificial Intelligent, AI)來分析檢測光伏電廠影像數據，找到缺陷和故障位置、分析嚴重程度，製成維修計畫，以利完整的維護工作。建構一套整合無人機熱影像巡檢技術與資料庫之智慧化太陽能板維運系統。透過無人機搭載紅外線與可見光影像模組，進行太陽能板之熱點巡檢，並運用 YOLOv5 與 YOLOv8 等深度學習演算法進行異常熱點之自動辨識。本研究以五組不同規模之訓練資料集進行模型訓練與比較，並以 Precision、Recall、mAP 及 F1-score 等指標進行評估。經實測，本研究最終採用之 YOLOv5 模型（V5-4）於測試集之 mAP@0.5 達 91.3%、F1-score 為 0.89，展現良好辨識能力。為達到模組層級之空間定位，本研究亦整合熱影像 EXIF 座標資訊與可見光正射影像，透過地理資訊系統（Geographic Information System, GIS）平台進行疊圖分析，建立模組熱點分布圖。最終結合維修報表資訊，實現從影像拍攝、AI 辨識、空間定位至異常追蹤與報修的完整流程，提升太陽能場域之智慧維運效率。檢測辨識的結果，對有缺陷或瑕疵的光伏板，將透過資料庫之GPS位置確認，檢出精確位置，並在Google地圖上進行標記。本研究與現有的替代方案相比，這種方法可能顯著提高光伏模組檢查和健康管理的準確性和效率。本研究從實務的檢測經驗，獲得完整的太陽能光伏電廠數據進行分析、標定及維護工程，所建議系統檢測成果可以有效且準確地發揮無人機技術與影像辨識技術的成效。

With the rapid growth of global photovoltaic (PV) capacity, maintaining efficient and reliable operations has become a critical challenge for renewable energy development. Traditional inspection methods, which rely heavily on manual labor and handheld thermal cameras, are time-consuming, unsafe, and inconsistent in accuracy. This study presents a comprehensive UAV-based solar panel inspection system that integrates multi-sensor data acquisition, artificial intelligence (AI)-based anomaly detection, and geographic information system (GIS)-enabled spatial analysis. By combining unmanned aerial vehicles (UAVs), infrared thermal imaging, visible imagery, and deep learning models, the proposed system provides a practical and scalable solution for intelligent PV field management.
Thermal and RGB imagery captured by UAVs are processed using a YOLOv5 deep learning model, achieving a high detection performance with mAP@0.5 of 91.3%, precision of 88.6%, and recall of 89.5%. A custom Python tool is developed to extract GPS metadata from the images, enabling accurate geospatial localization of detected anomalies. These results are visualized through GIS-based mapping tools, generating anomaly distribution maps and actionable maintenance reports.

[1] Intergovernmental Panel on Climate Change (IPCC), Sixth Assessment Report, 2023. [Online] Available: https://www.ipcc.ch/report/ar6/
[2] International Renewable Energy Agency (IRENA), Renewable Capacity Statistics 2023, Abu Dhabi, 2023. [Online] Available: https://www.irena.org/Statistics
[3] C. F. J. Kuo, “Performance optimization of solar photovoltaic systems through intelligent fault detection,” Energy Conversion and Management, Vol. 276, 15 January 2023, Art. No. 116437, doi: 10.1016/j.enconman.2022.116437.
[4] 行政院能源及減碳辦公室。《能源轉型白皮書》，2020。 [Online] Available: https://energywhitepaper.tw//
[5] 經濟部能源局。《再生能源發電統計月報》，2023。 [Online] Available: https://www.moeaboe.gov.tw/
[6] 行政院國家發展委員會。《台灣2050淨零排放路徑藍圖》，2022。 [Online] Available: https://www.ndc.gov.tw/
[7] 中華民國政府，《再生能源發展條例》，2019。[Online] Available: https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=J0130032。
[8] The Climate Group and CDP, RE100 Campaign Overview, 2014. [Online]. Available: https://www.there100.org/
[9] 工業技術研究院（2022）。台灣綠能產業發展現況分析報告，[Online] Available:https://itrisdgs.itri.org.tw/chi/%E5%B7%A5%E7%A0%94%E9%99%A22022%E6%B0%B8%E7%BA%8C%E5%A0%B1%E5%91%8A%E6%9B%B8.pdf.
[10] P. B. Quater, F. Grimaccia, S. Leva, M. Mussetta, and M. Aghaei, “Light unmanned aerial vehicles (UAVs) for cooperative inspection of PV plants,” IEEE Journal of Photovoltaics, Vol. 4, No. 4, pp. 1107–1113, July 2014, doi: 10.1109/JPHOTOV.2014.2323714.
[11] M. Aghaei, A. Gandelli, F. Grimaccia, S. Leva, and R. E. Zich, “IR real-time analyses for PV system monitoring by digital image processing techniques, ” Proceedings of IEEE International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP), Krakow, Poland, June 2015, pp. 1–6, doi: 10.1109/EBCCSP.2015.7300708.
[12] C. Henry, S. Poudel, S.-W. Lee, and H. Jeong, ”Automatic Detection System of Deteriorated PV Modules Using Drone with Thermal Camera, ” Applied Sciences, Vol. 10, No. 11, Art. 3802, May 2020, doi: 10.3390/app10113802.
[13] A. Gurras, L. Gergidis, C. Mytafides, L. Tzounis, and A. S. Paipetis, ”Automated detection-classification of defects on photo-voltaic modules assisted by thermal drone inspection, ” MATEC Web of Conferences, Vol. 349, p. 03015, 2021, doi: 10.1051/matecconf/202134903015.
[14] J. Liu and N. Ji, “A bright spot detection and analysis method for infrared photovoltaic panels based on image processing, ” Frontiers in Energy Research, Vol. 10, January 2023, doi: 10.3389/fenrg.2022.978247.
[15] S. J. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Hoboken, NJ, USA: Pearson, 2020.
[16] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning, ”Nature, vol. 521, no. 7553, pp. 436–444, May 2015, doi: 10.1038/nature14539.
[17] Y. Gao and S. Li, “A Deep Learning-Based Method Detects Dust from Solar PV Panels through Unmanned Aerial Vehicles, ” Journal of Physics, Conference Series, Vol. 2584, No. 1, Article 012019, 2023, doi: 10.1088/1742-6596/2584/1/012019.
[18] Pruthviraj, U.; Kashyap, Y.; Baxevanaki, E.; Kosmopoulos, P."Solar Photovoltaic Hotspot Inspection Using Unmanned Aerial Vehicle Thermal Images at a Solar Field in South India"Remote Sensing, 2023, 15(7), 1914. doi: 10.3390/rs15071914
[19] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-Based Learning Applied to Document Recognition,” Proceedings of the IEEE, Vol. 86, No. 11, pp. 2278–2324, 1998, doi: 10.1109/5.726791.
[20] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” Proceedings of the 25th International Conference on Neural Information Processing Systems (NeurIPS), Lake Tahoe, NV, USA, 2012, pp. 1097–1105.
[21] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, “Going deeper with convolutions,” Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, 2015, pp. 1–9, doi: 10.1109/CVPR.2015.7298594.
[22] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556, 2015. [Online] Available: https://arxiv.org/abs/1409.1556.
[23] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 779–788, doi: 10.1109/CVPR.2016.91.
[24] J. Redmon and A. Farhadi, “YOLOv3: An Incremental Improvement,” arXiv preprint arXiv:1804.02767, April 2018. [Online] Available: https://arxiv.org/abs/1804.02767
[25] L. Li, Z. Wang, and T. Zhang, "Photovoltaic panel defect detection based on Ghost convolution with BottleneckCSP and tiny target prediction head incorporating YOLOv5," arXiv preprint arXiv:2303.00886, March 2023.[Online] Available: https://arxiv.org/abs/2303.00886
[26] J. Huang, K. Zeng, Z. Zhang, and W. Zhong, "Solar panel defect detection design based on YOLO v5 algorithm," Heliyon, Vol. 9, No. 8, p. e18826, August 2023, doi: 10.1016/j.heliyon.2023.e18826.
[27] L. Liu, Q. Li, X. Liao, and W. Wu, "Bird droppings defects detection in photovoltaic modules based on CA-YOLOv5," Processes, vol. 12, no. 6, p. 1248, 2024, doi: 10.3390/pr12061248.