LAUSR

Utility-scale wind turbines are equipped with a supervisory control and data acquisition (SCADA) system for remote supervision and control. The SCADA system accumulates a large amount of data that contains the health conditions of the wind turbines. Thus, it is interesting to mine the health status-related information from SCADA data for wind turbine condition monitoring. In this article, an ensemble approach is proposed to detect anomalies and diagnose faults in wind turbines. Historical SCADA data collected from healthy wind turbines are used to model their normal behaviors and build a Mahalanobis space as a reference space. By comparing the predicted behavior of the wind turbine by a trained model with the reference space, anomalies can be detected. Finally, wind turbine faults are diagnosed through the analysis of the distributions and correlations of their SCADA data. The proposed approach is validated by using the SCADA data collected from two field wind turbines. Results show that it can detect anomalies and diagnose the corresponding failure components before the wind turbines have to be shut down for maintenance.