Scholarlink

Photovoltaic (PV) panels can experience various defects due to operational conditions, environmental factors, or human errors, leading to performance degradation and general risks such as system failures, inefficiencies, and potential fire hazards. This study presents a defect classification method using the k-nearest neighbors (kNN) algorithm, optimized with current-voltage curves. This method identifies three specific faults: Partial shading, shunted modules, and ground faults. To enhance classification, preprocessing techniques, notably normalization, are applied. The performance of the kNN algorithm is compared using datasets of different sizes (500 and 2000 samples) and varying test sizes (20%, 30%, 40%), with both normalized and non-normalized data. Results show that normalization significantly improves the mean score and reduces RMSE, with a statistically significant difference, demonstrating the effectiveness of normalization in fault classification. This study is essential for advancing the detection and classification of defects in photovoltaic (PV) panels. By enhancing these capabilities, the research aims to significantly optimize system performance and safety, thereby mitigating potential risks and improving the overall reliability and efficiency of PV installations.

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

Improved Fault Classification in Photovoltaic Panels Using kNN Machine Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Abdelilah Khlifi,
  • Yamina Khlifi

摘要

Photovoltaic (PV) panels can experience various defects due to operational conditions, environmental factors, or human errors, leading to performance degradation and general risks such as system failures, inefficiencies, and potential fire hazards. This study presents a defect classification method using the k-nearest neighbors (kNN) algorithm, optimized with current-voltage curves. This method identifies three specific faults: Partial shading, shunted modules, and ground faults. To enhance classification, preprocessing techniques, notably normalization, are applied. The performance of the kNN algorithm is compared using datasets of different sizes (500 and 2000 samples) and varying test sizes (20%, 30%, 40%), with both normalized and non-normalized data. Results show that normalization significantly improves the mean score and reduces RMSE, with a statistically significant difference, demonstrating the effectiveness of normalization in fault classification. This study is essential for advancing the detection and classification of defects in photovoltaic (PV) panels. By enhancing these capabilities, the research aims to significantly optimize system performance and safety, thereby mitigating potential risks and improving the overall reliability and efficiency of PV installations.