Scholarlink

The research in the field of photovoltaic fault diagnosis has mainly focused on single faults. To achieve accurate diagnosis of compound faults in photovoltaic modules, a method based on feature extraction and machine learning was proposed in this paper. A PV cell simulation model was established through equivalent circuit analysis, and a mathematical model for the PV array was built based on its series-parallel connection structure. A PV fault simulation model was constructed within the PV power generation simulation system to analyze the volt-ampere output characteristic curves of PV modules with single and compound faults. Parameters were normalized based on changes in surface characteristics, and new characteristic parameters were merged to accurately characterize faults. A mapping relationship between the characteristic parameters and faults was established. A fault diagnosis model for PV modules was constructed using the CatBoost algorithm of Bayesian optimization. The algorithm model proposed was verified through simulation, and the diagnostic accuracy for both single faults and compound faults exceeded 97%.

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

Complex Fault Diagnosis of Photovoltaic Modules Based on Feature Extraction and Machine Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Qiming Wang,
  • Kun Xia,
  • Pingyang Jiao,
  • Bingqi Liu,
  • Peijie Zhang

摘要

The research in the field of photovoltaic fault diagnosis has mainly focused on single faults. To achieve accurate diagnosis of compound faults in photovoltaic modules, a method based on feature extraction and machine learning was proposed in this paper. A PV cell simulation model was established through equivalent circuit analysis, and a mathematical model for the PV array was built based on its series-parallel connection structure. A PV fault simulation model was constructed within the PV power generation simulation system to analyze the volt-ampere output characteristic curves of PV modules with single and compound faults. Parameters were normalized based on changes in surface characteristics, and new characteristic parameters were merged to accurately characterize faults. A mapping relationship between the characteristic parameters and faults was established. A fault diagnosis model for PV modules was constructed using the CatBoost algorithm of Bayesian optimization. The algorithm model proposed was verified through simulation, and the diagnostic accuracy for both single faults and compound faults exceeded 97%.