The efficiency of solar photovoltaic (PV) systems is greatly affected by the condition of panel surfaces, as dust buildup and other impurities can noticeably reduce power output. Traditional maintenance methods rely largely on manual inspection, which can be subjective, inconsistent, and impractical for large-scale installations. To address these challenges, this work investigates a deep learning-based computer vision approach for the automatic assessment of solar panel cleanliness. In this work, a lightweight YOLOv8n classification model is utilized to categorize solar panels into four levels of surface condition: Clean, Almost Clean, Slightly Dirty, and Completely Dirty. The objective is to develop a fast and efficient model that can operate in real time and be deployed in both drone-based and stationary monitoring setups. The proposed approach involves constructing a labeled image dataset corresponding to the defined categories, followed by systematic training, validation, and performance analysis of the model. The model’s performance is assessed using common evaluation measures, including accuracy, precision, recall, and F1-score, along with an analysis of confidence levels for each predicted class. This study contributes to the advancement of intelligent maintenance strategies in renewable energy systems by demonstrating the adaptability of a modern object detection architecture for multi-class classification tasks. The results highlight the model’s capability to distinguish subtle variations in soiling levels while also emphasizing the importance of dataset quality and parameter tuning for real-world applications. Overall, the proposed solution supports proactive and data-driven cleaning decisions, which can enhance energy efficiency and lower operational costs in solar PV systems.

[1] A. A. Kazem, M. T. Chaichan, and H. A. Kazem, “Dust effect on photovoltaic utilization in Iraq: Review article,” Renewable and Sustainable Energy Reviews, vol. 37, pp. 734–749, 2014. [2] M. Mani and R. Pillai, “Impact of dust on solar photovoltaic (PV) performance: Research status, challenges and recommendations,” Renewable and Sustainable Energy Reviews, vol. 14, no. 9, pp. 3124–3131, 2010. [3] S. Ilse, C. Figgis, A. Naumann, et al., “Techno-economic assessment of soiling losses and mitigation strategies for solar power generation,” Joule, vol. 3, no. 9, pp. 2303–2321, 2019. [4] A. A. Merrouni, F. Elalaoui, A. Mezrhab, et al., “Investigation of soiling effect on photovoltaic systems performance,” Energy Procedia, vol. 42, pp. 172–181, 2013. [5] J. Bosch, L. Kleissl, and M. Mehos, “Soiling losses for solar photovoltaic systems in the southwestern United States,” Solar Energy, vol. 85, no. 7, pp. 1605–1616, 2011. [6] Y. Li, J. Chen, and X. Zhang, “Deep learning-based image recognition for solar panel defect detection,” IEEE Access, vol. 8, pp. 184468–184480, 2020. [7] S. Deitsch, C. Christlein, S. Berger, et al., “Automatic classification of defective photovoltaic module cells in electroluminescence images,” Solar Energy, vol. 185, pp. 455–468, 2019. [8] H. Zhang, J. Liu, and Y. Chen, “Vision-based dust detection for photovoltaic panels using deep learning,” Energies, vol. 13, no. 8, pp. 1–15, 2020. [9] M. H. Alajmi, M. A. Alotaibi, and S. Rehman, “Image-based solar panel condition monitoring using convolutional neural networks,” IEEE Access, vol. 9, pp. 123456–123468, 2021. [10] J. Redmon and A. Farhadi, “YOLOv3: An incremental improvement,” arXiv preprint arXiv:1804.02767, 2018. [11] A. Bochkovskiy, C.-Y. Wang, and H.- Y. M. Liao, “YOLOv4: Optimal speed and accuracy of object detection,” arXiv preprint arXiv:2004.10934, 2020. [12] G. Jocher, A. Chaurasia, and J. Qiu, “YOLOv8: Ultralytics next-generation real-time object detection and classification,” GitHub Repository, 2023. [13] H. Tan and L. Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” Proceedings of ICML, 2019. [14] T. Fawcett, “An introduction to ROC analysis,” Pattern Recognition Letters, vol. 27, no. 8, pp. 861–874, 2006. [15] M. Sokolova and G. Lapalme, “A systematic analysis of performance measures for classification tasks,” Information Processing & Management, vol. 45, no. 4, pp. 427–437, 2009. [16] A. Mellit, S. Kalogirou, L. Hontoria, and S. Shaari, “Artificial intelligence techniques for sizing photovoltaic systems: A review,” Renewable and Sustainable Energy Reviews, vol. 13, no. 2, pp. 406–419, 2009. [17] M. Benghanem, “Monitoring of solar photovoltaic systems,” Energy Conversion and Management, vol. 62, pp. 50–57, 2012. [18] A. Kimber, L. Mitchell, S. Nogradi, and H. Wenger, “The effect of soiling on large grid-connected photovoltaic systems in California and the Southwest region of the United States,” IEEE 4th World Conference on Photovoltaic Energy Conversion, pp. 2391–2395, 2006. [19] J. K. Kaldellis and D. Zafirakis, “Experimental investigation of the effect of dust deposition on photovoltaic panels’ performance,” Renewable Energy, vol. 37, no. 1, pp. 1–6, 2012. [20] A. Sayyah, M. N. Horenstein, and M. Mazumder, “Energy yield loss caused by dust deposition on photovoltaic panels,” Solar Energy, vol. 107, pp. 576–604, 2014. [21] C. Buerhop, P. Pickel, M. Reindl, et al., “Quality control of PV-modules in the field using infrared thermography,” Energy Procedia, vol. 27, pp. 532–539, 2012. [22] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016. [23] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” Advances in Neural Information Processing Systems (NeurIPS), 2012. [24] Y. Lin, T.-Y. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” IEEE International Conference on Computer Vision (ICCV), pp. 2980–2988, 2017. [25] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 785–794, 2016. [26] A. Dosovitskiy, L. Beyer, A. Kolesnikov, et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” International Conference on Learning Representations (ICLR), 2021. [27] S. Woo, J. Park, J.-Y. Lee, and I. S. weon, “CBAM: Convolutional block attention module,” European Conference on Computer Vision (ECCV), pp. 3–19, 2018. arXiv:1807.06521 [28] J. Long, E. Shelhamer, and T. Darrell, “Fully convolutional networks for semantic segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 4, pp. 640–651, 2017. arXiv:1411.4038 [29] S. Han, H. Mao, and W. J. Dally, “Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding,” International Conference on Learning Representations (ICLR), 2016. arXiv:1510.00149 [30] A. Howard, M. Zhu, B. Chen, et al., “MobileNets: Efficient convolutional neural networks for mobile vision applications,” arXiv preprint arXiv:1704.04861, 2017.