Detection and isolation of sensor faults of wind turbines using sliding mode observers

This paper proposes a sliding mode observer (SMO)-based fault detection and isolation (FDI) scheme for wind turbines. The actuator faults in pitch systems of the wind turbine are transformed as sensor faults. A reduced order model of the drive train system is constructed to eliminate the effects of unknown aerodynamic rotor torque. Based on the new system representation, a bank of SMOs are designed such that the output signal can be accurately estimated in the presence of faults. The proposed method can accurately determine the location of the faults by comparing the estimated outputs with measurements. The effectiveness of the proposed FDI scheme is illustrated via simulations.