Fault detection and identification for wind turbine systems: a closed-loop analysis

This paper assesses the usability of time-domain model-based Fault Detection and Identification (FDI) methods for application to a horizontal axis wind turbine (HAWT) that uses pitch-to-vane control. Two scenarios are considered: the estimation of an unknown actuator gain and an unknown actuator delay. The wind acts as a disturbance to the system, for which no measurement is available. A linear HAWT model has been designed and used as a simulation environment. Using System Identification, no reliable model could be estimated because of the closed-loop measurements. Two observer-based approaches, the discrete-time Kalman filter and the Interacting Multiple-Model (IMM) estimator, have also been investigated. Both methods allowed simultaneous estimation of the wind speed and the faulty pitch actuator gain. When compared to the discrete-time Kalman filter, the IMM estimator is more versatile: it has been applied successfully to estimate an unknown actuator delay.