The increasing adoption of photovoltaic systems has created a growing need for reliable and efficient defect detection methods to maintain energy yield and reduce maintenance costs. However, solar panels are highly susceptible to surface anomalies such as dust accumulation, bird droppings, snow cover, physical damage and electrical faults, all of which can significantly degrade performance. In this study, a cost-effective deep learning-based monitoring framework is proposed for the automatic classification of solar panel surface defects. The proposed approach employs a pre-trained Swin Transformer, which benefits from hierarchical feature extraction and shifted window-based self-attention to capture both local and global visual patterns under diverse environmental conditions. A six-class image dataset consisting of clean, dusty, bird-dropping, snow-covered, physical-damage and electrical-damage samples was used for training and validation. Experimental results demonstrate that the proposed model achieves a validation accuracy of 98.82%, an F1-score of 98.83%, a recall of 98.66% and a precision of 98.66%. Comparative analysis further shows that the proposed method outperforms several existing deep learning models, including VGG16, MobileNetV3 and SparkNet. These findings confirm that the Swin Transformer provides a robust and practical solution for intelligent solar panel defect classification in real-world monitoring applications.
