Adaptive fault detection and tracking for a wind turbine generator using Kalman filter

This paper describes a wind turbine (WT) condition monitoring technique that uses the measurement of stator current and rotational speed to derive a fault detection signal. The detection algorithm uses a Kalman filter (KF) to extract and track the strength of particular frequency components, characteristic of faults in the stator current signal. This has been done by an extensive simulation studies to develop an on-line detection and monitoring of mechanical faults in permanent magnet synchronous generators (PMSGs), recentlly used in modern variable-speed WTs. The model is developed and validated with operational data of five 2.5MW turbines were recorded by the supervisory control and data acquisition (SCADA) system over the period of 1 year. The simulation results show that the KF algorithm can provide a reliable indication of the presence of a fault with low computational times, from director indirectdrive fixed- or variable-speed WTs. The proposed algorithm can indicate the severity of the fault, where in contrast with traditional methods, they failed to extract the fault features from non-stationary current measurements, due to variable-speed operating conditions of WTs.