Scholarlink

The frequent occurrence of insulated gate bipolar transistor (IGBT) failures in wind turbines, together with the lack of exhaustive analysis in this field, demands the development of reliable methods for detecting and diagnosing these failures. Recurrent neural networks (RNN), particularly gated recurrent units (GRU), have demonstrated their potential in predicting new time series data. In this manuscript, a method that combines statistical analysis to find patterns and relationships to explain the behavior of IGBTs is presented; an RNN-GRU model is implemented to predict data coming from the SCADA system. Additionally, the method incorporates the Isolation Forest (IF) algorithm and the analysis of operation and maintenance (O&M) records to establish a threshold for identifying possible failures related to IGBTs. The results demonstrate the proposed method’s effectiveness and ability to detect faults, significantly improving the reliability and maintenance of wind turbine systems, and providing a valuable tool for the wind energy industry.

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

Fault Prediction in Wind Turbines: A Recurrent Neural Network Approach for IGBT Failure Detection

  • 章节

  • 发表日期：

    2025年

查看全文
  • Génesis Vásquez,
  • David Rosales,
  • Jorge Maldonado-Correa

摘要

The frequent occurrence of insulated gate bipolar transistor (IGBT) failures in wind turbines, together with the lack of exhaustive analysis in this field, demands the development of reliable methods for detecting and diagnosing these failures. Recurrent neural networks (RNN), particularly gated recurrent units (GRU), have demonstrated their potential in predicting new time series data. In this manuscript, a method that combines statistical analysis to find patterns and relationships to explain the behavior of IGBTs is presented; an RNN-GRU model is implemented to predict data coming from the SCADA system. Additionally, the method incorporates the Isolation Forest (IF) algorithm and the analysis of operation and maintenance (O&M) records to establish a threshold for identifying possible failures related to IGBTs. The results demonstrate the proposed method’s effectiveness and ability to detect faults, significantly improving the reliability and maintenance of wind turbine systems, and providing a valuable tool for the wind energy industry.