Loading...
Snow Covered Solar Photovoltaics Identification: Exploring GANs as Feature Learners for Classification
Anwar, Naveed Khan, Mian Farhan Ullah
Anwar, Naveed
Khan, Mian Farhan Ullah
Type
Supervisor
Date
2026-02-09
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
Solar photovoltaic (PV) systems are installed worldwide to promote green energy and mitigate climate change. In cold and snowy regions, however, snow accumulation on PV surfaces blocks sunlight and halts power generation until the snow melts or is manually removed. This phenomenon not only decreases energy yield but also complicates grid stability during winter seasons. Therefore, timely and accurate detection of snow-covered PV panels is crucial to ensure continuous and reliable energy production. This study presents a simple yet novel approach using an isolated Generative Adversarial Network (GAN) discriminator to classify PV panels as fully exposed, partially covered, or completely covered by snow. GAN-based discriminators are advantageous because they can learn high-level spatial and textural features directly from images without extensive preprocessing or handcrafted filters. The proposed model was validated on a publicly available University of Waterloo dataset and trained using the Adam optimizer with a learning rate of 0.001 and momentum of 0.9. It achieved 100% accuracy in both training and testing, with a 0% error rate. These results demonstrate a robust framework for automated PV health monitoring, real-time snow detection, and large-scale renewable energy optimization.
Department
Publisher
Sponsor
No