Fault estimation and fault‐tolerant control of the FAST NREL 5‐MW reference wind turbine using a proportional multi‐integral observer. (1st August 2017)

• Record Type: Journal Article
• Authors: Kühne, Patrick; Pöschke, Florian; Schulte, Horst
• Other Names: Puig Vicenç guestEditor.; Odgaard Peter Fogh guestEditor.; Schulte Horst guestEditor.
• Is Part Of: International journal of adaptive control and signal processing. Volume 32:Number 4(2018)
• Page Start: 568
• Page End: 585
• Publication Date: 2017-08-01
• Subjects: fault estimation -- fault‐tolerant control -- Takagi‐Sugeno model approach -- unknown inputs -- wind energy systems; Adaptive control systems -- Periodicals; Adaptive signal processing -- Periodicals; 629.836
• DOI: 10.1002/acs.2800
• Languages: English
• ISSNs: 0890-6327

Abstract

Summary: A proportional multi‐integral (PMI) observer scheme is used to reconstruct actuator and sensor faults in large‐scale wind turbines. For the nonlinear full‐load region, a PMI observer in Takagi‐Sugeno form is used. Two different methods are proposed for the observer design: a Lyapunov‐based and a modified linear quadratic regulator approach, where the stability is ensured by a negative derivative of the Lyapunov function candidate. The actuator and sensor faults are directly reconstructed from an augmented state‐space vector. The control inputs and measurement signal are then modified with the reconstructed fault signals to achieve fault‐tolerant control in the presence of faults with a similar behavior to the fault‐free case. It is shown by extensive simulation studies that the PMI observer is able the reconstruct offset and ramp‐shaped faults with high accuracy. The performance of the fault reconstruction and fault‐tolerant control scheme based on a 4‐degree‐of‐freedom observer model is tested in simulations by a high‐order 24‐degree‐of‐freedom model with the aeroelastic code FAST by the National Renewable Energy Laboratory. The new results are also compared with those of the previously published fault reconstruction scheme using Takagi‐Sugeno sliding‐mode observers.