Solar panels are integral components of renewable energy systems, but their efficiency can be compromised by the accumulation of dust and debris. In this study, we propose an automated approach for the binary classification of solar panels as either clean or dusty using deep learning techniques. Specifically, we apply the ResNet18 pre-trained convolutional neural network (CNN) model with transfer learning to classify images of solar panels.
A dataset comprising 194 images of clean solar panels and 191 images of dusty solar panels was collected and utilized for model training and evaluation. The images were preprocessed to enhance features relevant to dust detection while maintaining image integrity. The ResNet18 model, pre-trained on large-scale image datasets, was fine-tuned using the collected dataset to adapt its parameters to the task of binary classification.
During experimentation, the trained model exhibited remarkable performance, achieving an accuracy rate of 95.39% in distinguishing between clean and dusty solar panels. This high accuracy demonstrates the efficacy of the proposed approach in automated detection. Moreover, the model's ability to accurately classify solar panels in real-world conditions suggests its potential for practical deployment in monitoring and maintenance systems for solar energy installations.
The success of our study highlights the importance of leveraging deep learning and transfer learning techniques for addressing environmental challenges in renewable energy systems. By automating the detection of dusty solar panels, our approach offers a cost-effective and efficient solution to optimize solar panel performance and maximize energy generation. Future research may explore the scalability of the proposed method to larger datasets and its integration into real-time monitoring systems for sustainable energy management. ORCID NO: 0000-0001-9105-508X
Anahtar Kelimeler: CNN, Transfer Learning, Deep Learning, Solar panels
|
