Multiclass Deep Learning for Automated Solar Panel Surface Defect Detection Using the Solar Augmented Dataset
Mohamed HADADI, Walid HOUMAIDI, Anas RIAD
Abstract. In this study, we evaluated the performance of deep learning methodologies on the classification of surface conditions of solar panels on the open-solar dataset known as the Solar Augmented Dataset. The task entailed classifying six unique operational conditions in the dataset, namely: Clean, Dusty, Bird Drop, Physical Damage, Electrical Damage, and Snow-Covered. To test and compare our results against other methodologies on the dataset, we fine-tuned two pre-trained architectures of Convolutional Neural Networks on the dataset; namely, the VGG16 and Xception networks. Results on the dataset were that the best model was the VGG16 for our task since it gave over 91% on the validation set and had excellent recall values for the classes of interest-protecting safety-protective conditions, that is, Electrical Damage and Snow-covered Conditions. However, error analysis of the approach on the dataset through the creation of confusion matrices indicates that the model had difficulty distinguishing between the Clean and Dusty conditions on the solar panels, which could be so because the latter condition only represented the accumulation of light dust on the panels.
Keywords
Solar Panel Defect Detection, Deep Learning, Convolutional Neural Networks, Xception, VGG16, Image Classification
Published online 4/25/2026, 8 pages
Copyright © 2026 by the author(s)
Published under license by Materials Research Forum LLC., Millersville PA, USA
Citation: Mohamed HADADI, Walid HOUMAIDI, Anas RIAD, Multiclass Deep Learning for Automated Solar Panel Surface Defect Detection Using the Solar Augmented Dataset, Materials Research Proceedings, Vol. 64, pp 519-526, 2026
DOI: https://doi.org/10.21741/9781644904091-65
The article was published as article 65 of the book Energy Futures
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 license. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
References
[1] A. Rahman, Solar Panel Surface Defect and Dust Detection: Deep Learning Approach, J. Imaging 11 (2025) 287. https://doi.org/10.3390/jimaging11090287
[2] H. Ling, M. Liu, Y. Fang, Deep Edge-Based Fault Detection for Solar Panels, Sensors 24 (2024) 5348. https://doi.org/10.3390/s24165348
[3] I. Polymeropoulos, S. Bezyrgiannidis, E. Vrochidou, G.A. Papakostas, Enhancing Solar Plant Efficiency: A Review of Vision-Based Monitoring and Fault Detection Techniques, Technologies 12 (2024) 175. https://doi.org/10.3390/technologies12100175
[4] R.A.M. Rudro, K. Nur, Md.F.A.A. Sohan, M.F. Mridha, S. Alfarhood, M. Safran, K. Kanagarathinam, SPF-Net: Solar panel fault detection using U-Net based deep learning image classification, Energy Rep. 12 (2024) 1580–1594. https://doi.org/10.1016/j.egyr.2024.07.044
[5] S.R. Joshua, S. Park, K. Kwon, Solar Panel Fault Detection: Applying Convolutional Neural Network for Advanced Fault Detection in Solar-Hydrogen System at University, in: 2024 IEEE 24th International Conference on Software Quality, Reliability, and Security Companion (QRS-C), IEEE, 2024: pp. 289–298. https://doi.org/10.1109/QRS-C63300.2024.00045
[6] K. Masita, A. Hasan, T. Shongwe, H.A. Hilal, Deep learning in defects detection of PV modules: A review, Solar Energy Adv. 5 (2025) 100090. https://doi.org/10.1016/j.seja.2025.100090
[7] A. Mellit, S. Kalogirou, Artificial intelligence and internet of things to improve efficacy of diagnosis and remote sensing of solar photovoltaic systems: Challenges, recommendations and future directions, Renew. Sustain. Energy Rev. 143 (2021) 110889. https://doi.org/10.1016/j.rser.2021.110889
[8] S.H. Han, T. Rahim, S.Y. Shin, Detection of Faults in Solar Panels Using Deep Learning, in: 2021 International Conference on Electronics, Information, and Communication (ICEIC), IEEE, 2021: pp. 1–4. https://doi.org/10.1109/ICEIC51217.2021.9369744
[9] R. G, D.S. Manish, R. Narasimhan A, A.U. Dhavale, R.R. John, SparkNet–A Solar Panel Fault Detection Deep Learning Model, IEEE Access 13 (2025) 75599–75617. https://doi.org/10.1109/ACCESS.2025.3564714
[10] Y. Ledmaoui, A. El Maghraoui, M. El Aroussi, R. Saadane, Enhanced Fault Detection in Photovoltaic Panels Using CNN-Based Classification with PyQt5 Implementation, Sensors 24 (2024) 7407. https://doi.org/10.3390/s24227407
[11] G. Navnath, Solar Augmented Dataset, Kaggle, n.d. https://www.kaggle.com/datasets/gitenavnath/solar-augmented-dataset (accessed November 29, 2025).
