Abstract:
In an effort to reduce cost of energy from wind, wind turbines have grown to immense sizes. This has led to large, flexible, lightly damped towers and rotors that can be excited by the wind. Reducing the resulting fatigue loading and maintaining power capture are primary objectives for advanced controllers. In this thesis, a synthesis procedure for creating a multivariable linear parameter varying (LPV) controller suitable for the wind turbine control problem is created. The LPV controller uses the current wind speed estimate from an Extended Kalman Filter (EKF) for gain scheduling in order to accommodate system nonlinearities. The synthesis procedure allows the use of a parameter dependance Lyapunov function without having to choose the form of the parameter dependence. Additionally, the synthesis procedure is designed for discrete time systems, allowing digital implementation of the controller. While the LPV controller is suitable for the wind turbine problem, its performance is limited by constrained actuators, as well as persistent disturbances to the system. Therefore a model predictive controller (MPC controller) that builds on the advantages of the LPV controller is created. The MPC controller utilises future wind speed information to increase controller performance and can maintain stability in the presence of constrained actuators. The ability of both controllers to reduce fatigue loading in the drivetrain, tower and blades whilst maintaining power capture relative to a baseline is tested in simulation. The testing includes six hours of simulations using a high order, nonlinear aeroelastic model of a three-bladed, 600kW wind turbine in full-field turbulent winds. The simulation conditions include above rated, below rated, and transitional winds. The LPV controller shows overall reductions in tower, drivetrain and blade loads relative to the baseline. The MPC controller shows poor performance in below rated winds due to high errors in the prediction model. In above rated winds, the MPC controller shows the ability to reduce loads in the blades, drivetrain and tower relative to the LPV controller. Furthermore, the MPC controller shows less pitch actuator usage and maintains performance in situations that cause the LPV controller to saturate the pitch actuators and lose performance.