Scholarlink

<p>Fault detection in photovoltaic (PV) plants is essential to ensure reliability, safety, and maximum operating efficiency while reducing maintenance costs. Conventional protection devices often fail to detect subtle PV faults, leading to safety risks and performance losses. This study proposes a machine learning–based approach for fault detection using key electrical measurements such as voltage, current, power, and irradiance. Labelled datasets were generated from real-time data collected through a LabVIEW-integrated Data Acquisition (DAQ) system in a 1.3 kWp grid-connected PV plant. Multiple supervised classification algorithms, including Naïve Bayes, Logistic Regression, Decision Trees, Random Forest, and Gradient Boosting, were trained and validated to identify line-to-line faults and distinguish them from partial shading conditions. Experimental results demonstrate that ensemble methods, particularly Random Forest and Gradient Boosting, achieved superior accuracy, with classification models consistently exceeding 99% detection performance. The findings confirm the effectiveness of the proposed framework in accurately detecting PV faults and differentiating them from non-critical disturbances, thereby improving diagnostic accuracy and system reliability.</p>

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

Machine Learning Approaches for Solar PV Fault Identification

  • 期刊论文

  • 发表日期：

    2025年9月19日

查看全文
  • M. V. Prashanth,
  • K. N. Bharath,
  • Nasreen Fathima,
  • B. M. Latha,
  • M. S. Sathisha,
  • B. R. Ramji,
  • G. R. Yathiraj

摘要

Fault detection in photovoltaic (PV) plants is essential to ensure reliability, safety, and maximum operating efficiency while reducing maintenance costs. Conventional protection devices often fail to detect subtle PV faults, leading to safety risks and performance losses. This study proposes a machine learning–based approach for fault detection using key electrical measurements such as voltage, current, power, and irradiance. Labelled datasets were generated from real-time data collected through a LabVIEW-integrated Data Acquisition (DAQ) system in a 1.3 kWp grid-connected PV plant. Multiple supervised classification algorithms, including Naïve Bayes, Logistic Regression, Decision Trees, Random Forest, and Gradient Boosting, were trained and validated to identify line-to-line faults and distinguish them from partial shading conditions. Experimental results demonstrate that ensemble methods, particularly Random Forest and Gradient Boosting, achieved superior accuracy, with classification models consistently exceeding 99% detection performance. The findings confirm the effectiveness of the proposed framework in accurately detecting PV faults and differentiating them from non-critical disturbances, thereby improving diagnostic accuracy and system reliability.