Skip to content
Research Article Peer ReviewedAt least two reviewers Open Access

Energy Loss and Panel Fault Detection in Solar photovoltaic Systems Using Extreme Gradient Boosting algorithm

1Department of Mechanical Engineering, Gaziantep University
✉ Corresponding author: Fatma Zehra Kardas · zhrakrds2@gmail.com
Proceedings International Symposium on AI-Driven Engineering Systems Volume 22 Pages 107–110 Year 2025
PDF BibTeX RIS

Abstract

Solar photovoltaic (PV) systems are widely adopted for sustainable energy production; however, physical defects such as glass breakage significantly reduce panel efficiency and system output. In this study, a real-world case analysis was conducted on a 1003 kWp rooftop PV installation located at the Faculty of Educational Sciences, Gaziantep University, Türkiye monitored via the SolarEdge platform. Voltage and current (V-I) values from both reference and cracked-glass panels were collected and preprocessed to remove overlapping records and outliers below defined thresholds. A machine learning (ML) model based on the Extreme Gradient Boosting (XGBoost) algorithm was developed using Python to classify panels as either "reference" or "cracked" based on their electrical behavior. The model achieved high classification accuracy, enabling early detection of defective panels without physical inspection. Furthermore, the energy loss caused by cracked panels was quantified by comparing daily production metrics with those from fully operational control panels. This approach highlights the potential for data-driven fault detection and performance optimization in PV systems, offering practical benefits for maintenance planning and energy yield improvement.

Keywords

Cite this article

Kardas, F. Z., & Atmaca, A. (2025). Energy Loss and Panel Fault Detection in Solar photovoltaic Systems Using Extreme Gradient Boosting algorithm. In International Symposium on AI-Driven Engineering Systems (pp. 107–110). SET Technology. https://doi.org/10.36287/setsci.22.59.001

Article History

Published
10 Jul 2025

License

© 2025 SET Technology. This article is distributed under the terms of the CC BY 4.0.

Research Contributions (2019+)Peer reviewed by at least two reviewers
Open AccessCC BY 4.0
Open Discovery Metadata · DOI · OAI-PMH
Global Reachresearchers worldwide