Misalignment Fault Diagnosis for Wind Turbines Based on Information Fusion

Area of Science:

• Engineering
• Renewable Energy Systems
• Signal Processing

Background:

• Conventional wind turbine fault diagnosis often relies on single signal types, limiting performance.
• Misalignment is a critical fault in wind turbines, impacting operational efficiency and lifespan.

Purpose of the Study:

• To develop an advanced wind turbine misalignment diagnosis model using multiple signal types.
• To improve the accuracy and robustness of fault detection in wind turbine transmission systems.

Main Methods:

• Utilized vibration, temperature, and stator current signals for misalignment diagnosis.
• Integrated Dempster-Shafer (D-S) evidence theory with Least Square Support Vector Machine (LSSVM).
• Applied t-distributed stochastic neighbor embedding (t-SNE) for dimensionality reduction and an artificial bee colony algorithm for LSSVM parameter optimization.

Main Results:

• The proposed model demonstrated high accuracy in diagnosing normal, parallel, angular, and integrated misalignments.
• Simulation and experimental results confirmed the model's superiority over existing methods.
• Evidence fusion using D-S theory effectively integrated multi-signal information.

Conclusions:

• The multi-signal integrated diagnosis model offers a more accurate and reliable approach for wind turbine misalignment detection.
• This method enhances wind turbine health monitoring and predictive maintenance strategies.
• The combined signal processing and evidence theory approach provides a robust framework for complex fault diagnosis.