کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Hybrid Attention-Based Defect Classification in Photovoltaic Modules for Edge Deployment

نویسندگان :

Sepehr Shakibi¹ Melika Vafaee Taj Khatooni² Farnia Ayati³ Mohsen Hamzeh⁴

1- دانشگاه تهران، دانشکده مهندسی برق و کامپیوتر 2- دانشگاه تهران، دانشکده مهندسی برق و کامپیوتر 3- دانشگاه تهران، دانشکده مهندسی برق و کامپیوتر 4- دانشگاه تهران، دانشکده مهندسی برق و کامپیوتر

کلمات کلیدی :

Photovoltaic systems،Hybrid attention networks،CBAM،Coordinate Attention،Deep learning،Edge deployment

چکیده :

Accurate defect detection in photovoltaic (PV) modules is critical for system efficiency, safety, and long-term energy yield. This paper introduces a novel hybrid attention mechanism, CCA (Combined Convolutional Block Attention Module and Coordinate Attention), which enhances feature representation for PV defect classification under complex imaging conditions. By integrating channel, spatial, and positional attention, CCA enables precise defect localization and improved semantic understanding. The method is evaluated on real-world PV images. Experimental results across multiple backbone networks show that CCA consistently boosts classification accuracy, especially in lightweight and deeper architectures, while maintaining computational efficiency. Deployment on a Raspberry Pi demonstrates real-time performance with minimal inference overhead. Overall, the proposed CCA mechanism offers a robust, generalizable, and edge-deployable solution for reliable PV module defect detection.