A Hybrid Fuzzy–Support Vector Machine Framework for Real‐Time Dust Detection and Thermal Mapping in Photovoltaic Panels.
Saved in:
| Title: | A Hybrid Fuzzy–Support Vector Machine Framework for Real‐Time Dust Detection and Thermal Mapping in Photovoltaic Panels. |
|---|---|
| Authors: | Sarker, Debasish1 (AUTHOR) dsarker.me@iubat.edu, Karim, S. M. Rezaul2,3 (AUTHOR), Hossain, MD. Shouquat2,4 (AUTHOR) shouquat64@gmail.com, Hossain, Md. Moshaddek1 (AUTHOR), Kabir, Md. Monirul4 (AUTHOR) |
| Source: | Energy Science & Engineering. Apr2026, Vol. 14 Issue 4, p2040-2050. 11p. |
| Subject Terms: | *Fuzzy clustering technique, *Support vector machines, *Machine learning, *Real-time computing, *Thermography, *Statistical models, *Dust measurement, *Photovoltaic power generation |
| Abstract: | Dust accumulation significantly degrades the energy output of photovoltaic (PV) panels, particularly in arid and semi‐arid regions. While existing studies have separately explored image‐based dust detection, environmental modeling, and machine learning (ML) for performance prediction, few have integrated these approaches to capture the coupled optical‐thermal effects of soiling. This study proposes a novel integrated framework that combines fuzzy clustering for panel segmentation, a hybrid SVM–fuzzy logic classifier for dust detection using intensity‐texture features, and a semi‐empirical plus ML–based thermal model. The framework uniquely fuses image‐derived dust maps with real‐time meteorological data, including humidity, ambient temperature, solar zenith angle (SZA), and global horizontal irradiance (GHI)–form a 5‐year dataset for model spatially non‐uniform solar absorption and thermal behavior. Experimental validation using reference dust loads of 3, 5.001, and 8 g·m⁻2 across multiple PV panels yielded mean absolute errors of approximately 0.07, 0.12, and 0.18 g·m⁻2, respectively. The thermal coefficient α was estimated through environmentally driven regression, providing a physically consistent, dust‐informed temperature assessment suitable for real‐time monitoring and predictive maintenance. This work advances the state‐of‐the‐art by offering a lightweight, interpretable, and integrated solution that outperforms fragmented approaches in accuracy and practical deployability. [ABSTRACT FROM AUTHOR] |
| Database: | Energy & Power Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
Be the first to leave a comment!
