Aerodynamic Noise-Based Fault Detection for Wind Turbine Blades: An Online Unsupervised Approach

Wind turbine blades are vulnerable to environmental damage, which can compromise the energy conversion efficiency of wind turbines and precipitate safety hazards. This paper proposes an online unsupervised method for detecting faults in wind turbine blades based on aerodynamic noise. The proposed method uses the Empirical Mode Decomposition (EMD) algorithm to process the aerodynamic noise signals and extract feature vectors. A domain of normal operation is delineated using these feature vectors derived from standard aerodynamic noise. During the real-time operation, the latest aerodynamic noise sample is integrated into this normative domain and analyzed using the DBSCAN clustering algorithm to pinpoint anomalies. Case studies conducted on the MIMII dataset indicate that the proposed method is more effective in detecting anomaly status of wind turbine blades from the aerodynamic noise. The proposed unsupervised approach is designed to address the challenge of limited anomaly samples in real-world datasets, demonstrating significant robustness. It is anticipated to facilitate rapid deployment within wind farm SCADA systems.