Simultaneous fault detection and sensor selection for condition monitoring of wind turbines

Abstract

Data collected from the supervisory control and data acquisition (SCADA) system are used widely in wind farms to obtain operation and performance information about wind turbines. The paper presents a three-way model by means of parallel factor analysis (PARAFAC) for wind turbine fault detection and sensor selection, and evaluates the method with SCADA data obtained from an operational farm. The main characteristic of this new approach is that it can be used to simultaneously explore measurement sample profiles and sensors profiles to avoid discarding potentially relevant information for feature extraction. With K-means clustering method, the measurement data indicating normal, fault and alarm conditions of the wind turbines can be identified, and the sensor array can be optimised for effective condition monitoring.