Scholarlink

Electroluminescence (EL) imaging serves as a valuable modality for scrutinizing photovoltaic (PV) modules, offering high spatial resolution capable of detecting even the most subtle defects on module surfaces. However, manual analysis of EL images is inherently expensive, time-consuming, and demands expertise across various defect types. Ensuring efficient defect detection in solar cell manufacturing is imperative for stable green energy technology production. Implementing anomaly detection techniques for PV cells can yield substantial cost savings in operation and maintenance (O&M). The study highlights the importance of detecting defects, in cell manufacturing and maintaining PV power stations effectively due to the difficulties in EL image analysis and the rapid changes in the industry landscape using advanced technologies like convolutional neural networks (CNNs). Introducing the Bat Algorithm enhanced CNN model (BA-CNN), the research focuses on identifying and categorizing defects, in PV cells. The BA-CNN demonstrates an accuracy of 98. It surpasses Visual Geometry Group 16 (VGG16) and outperforms MobileNet variants with lower averages of around 90 and 83 respectively. This highlights the models suitability for use in quality control systems and its practical application in projects. According to the research, researchers emphasize the BA-CNNs tradeoff between performance and computational efficiency that is crucial, for solar cell manufacturing. They also provide results to support their findings.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Photovoltaic Cell Quality Evaluation and Anomaly Detection Through Deep Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Eman Ashraf,
  • Warda M. Shaban,
  • Shady Y. EL Mashad,
  • A. E. Kabeel,
  • Essam Bahgat,
  • M. M. Eissa

摘要

Electroluminescence (EL) imaging serves as a valuable modality for scrutinizing photovoltaic (PV) modules, offering high spatial resolution capable of detecting even the most subtle defects on module surfaces. However, manual analysis of EL images is inherently expensive, time-consuming, and demands expertise across various defect types. Ensuring efficient defect detection in solar cell manufacturing is imperative for stable green energy technology production. Implementing anomaly detection techniques for PV cells can yield substantial cost savings in operation and maintenance (O&M). The study highlights the importance of detecting defects, in cell manufacturing and maintaining PV power stations effectively due to the difficulties in EL image analysis and the rapid changes in the industry landscape using advanced technologies like convolutional neural networks (CNNs). Introducing the Bat Algorithm enhanced CNN model (BA-CNN), the research focuses on identifying and categorizing defects, in PV cells. The BA-CNN demonstrates an accuracy of 98. It surpasses Visual Geometry Group 16 (VGG16) and outperforms MobileNet variants with lower averages of around 90 and 83 respectively. This highlights the models suitability for use in quality control systems and its practical application in projects. According to the research, researchers emphasize the BA-CNNs tradeoff between performance and computational efficiency that is crucial, for solar cell manufacturing. They also provide results to support their findings.