Scholarlink

This study introduces an innovative IoT-based framework for predictive maintenance, specifically developed for small wind turbines in decentralized renewable energy systems. The paper focuses on overcoming the challenges posed by legacy monitoring techniques through the use of economical sensor arrays, which include vibration, temperature, and rotational speed sensors, streamlined through an Arduino Uno microcontroller for seamless and uninterrupted data gathering. The hybrid machine learning method that mixes Random Forest and Long Short-Term Memory (LSTM) algorithms is applied for the processing of the sensor data, providing for the effective prediction of turbine issues like bearing deterioration and blade imbalance. The experimental evaluation testifies a prediction of the fault at 92%, with spike drops in unplanned downtime by 35–40%, and an increase in annual energy production by 15–20%. The maintenance system will enable users to make better decisions by providing them with a tailored web dashboard that shows the condition of the turbine and its history over time, which reduces maintenance costs by 20% to the average. This study contributes to the overall development of cost-effective predictive maintenance systems considerably, thus enabling the proper and more reliable functioning of the small resource wind energy systems.

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

Advance IoT and Machine Learning Framework for Predictive Maintenance of Small-Scale Wind Turbines

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rajesh Dhake,
  • Soniya Warade,
  • Siddhika Zanje,
  • Rizwan Shaikh,
  • Aditya Bhandare,
  • Rohit Yeole

摘要

This study introduces an innovative IoT-based framework for predictive maintenance, specifically developed for small wind turbines in decentralized renewable energy systems. The paper focuses on overcoming the challenges posed by legacy monitoring techniques through the use of economical sensor arrays, which include vibration, temperature, and rotational speed sensors, streamlined through an Arduino Uno microcontroller for seamless and uninterrupted data gathering. The hybrid machine learning method that mixes Random Forest and Long Short-Term Memory (LSTM) algorithms is applied for the processing of the sensor data, providing for the effective prediction of turbine issues like bearing deterioration and blade imbalance. The experimental evaluation testifies a prediction of the fault at 92%, with spike drops in unplanned downtime by 35–40%, and an increase in annual energy production by 15–20%. The maintenance system will enable users to make better decisions by providing them with a tailored web dashboard that shows the condition of the turbine and its history over time, which reduces maintenance costs by 20% to the average. This study contributes to the overall development of cost-effective predictive maintenance systems considerably, thus enabling the proper and more reliable functioning of the small resource wind energy systems.