linear parameter varying model

In this paper, a novel linear parameter varying (LPV) model of a wind turbine is developed based on a benchmark model presented by Aalborg University and KK-electronic a/c. The observability and validity of the model are investigated using real aerodynamic data. In addition, a robust fault detection and reconstruction method for linear parameter varying systems using second-order sliding mode observer is developed and implemented on the linear parameter varying model. The fault signal is reconstructed using a nonlinear term named equivalent output error injection during sliding motion and a proper transformation. The effect of uncertainties and incorrect measurements are minimized by employing an oriented method that requires solving a nonlinear matrix inequality. During numerical simulations, an actuator fault in the pitch system is considered and the performance of the method in fault reconstruction is investigated. Wind speed range is considered from 14 m/s to 16  and it is regarded as a stochastic input exerting aerodynamic torque. Fast and accurate fault reconstruction happens in 0.6 seconds with less than one percent error. The observer performance is not affected by the fault and fault is estimated in 2.5 seconds with an error smaller than 2.48 percent. Results illustrate fast and accurate fault reconstruction and accurate state estimations in the presence of actuator fault.