Multi-Fault Detection and Classification of Wind Turbines Using Stacking Classifier.

Area of Science:

• Renewable Energy Engineering
• Machine Learning Applications
• Condition Monitoring Systems

Background:

• Wind turbines are crucial for clean energy but suffer from failures and downtime, increasing operational costs.
• Traditional fault detection methods struggle with the complexity of wind turbine operations.
• Supervisory Control and Data Acquisition (SCADA) systems provide valuable data for monitoring wind turbine health.

Purpose of the Study:

• To develop an accurate and reliable method for classifying wind turbine faults.
• To improve the dependability and performance of wind turbines through timely maintenance.
• To leverage SCADA data for effective wind turbine condition monitoring.

Main Methods:

• Proposed a novel AdaBoost, K-nearest neighbors, and logistic regression-based stacking ensemble (AKL-SE) classifier.
• Preprocessed SCADA data by cleaning and removing abnormal data for improved validity.
• Utilized the Pearson correlation coefficient for input variable selection and trained the stacking ensemble classifier.

Main Results:

• The proposed AKL-SE classifier successfully identified faults in wind turbines.
• The stacking ensemble approach demonstrated enhanced accuracy in fault classification.
• The method was validated on local 3 MW wind turbines, showing effective fault detection capabilities.

Conclusions:

• The developed AKL-SE classifier offers a promising solution for wind turbine fault diagnosis.
• Accurate fault detection can significantly reduce wind turbine downtime and maintenance costs.
• This approach supports the reliable use of wind energy, promoting clean energy adoption.