Improving wind turbine aerodynamic performance using iterative learning control applied to smart rotors

Currently there is significant research into the inclusion of localised active flow control on wind turbine rotor blades, with the aim, in conjunction with collective and individual pitch control, of improving the aerodynamic performance of the rotor. These blades are termed smart rotors. The unique contribution of this research is the application of Iterative Learning Control to wind turbine smart rotors to reduce blade loading from lift disturbances. The smart devices act locally, at different spanwise positions, and include actuation to manipulate local lift (e.g. trailing edge flaps, blowing/suction, circulation control); sensing to determine the current turbine loading (e.g. pressure sensors, strain gauges, LIDAR); and a suitable control scheme to achieve predefined objectives. The principal objective is to reduce fatigue loads, although mitigating the effects of extreme loads is also of interest. The reduction of these loads leads to lighter, larger and more reliable turbines. Traditionally blade loads have been managed using stall regulation, pitch control, torque control or a combination of all three. Smart rotors are an evolutionary step in the control of turbines and have the advantage of deploying variable control along the blade with quicker response times to variations in flow conditions, leading to a potential increase in energy production, an increase in turbine reliability and a reduced energy requirements. The aerodynamic loads on a wind turbine blade have periodic and non-periodic components, and the nature of these strongly suggests the application of iterative learning control. The research within this PhD thesis employs a 2D computational fluid dynamics model (vortex panel method), with nonlinear wake effects, to represent flow past an aerofoil. The CFD model uses a potential flow approximation which is valid for inviscid and attached flow only. This is acceptable because smart devices typically operate under such conditions. Circulation control (actuation) and pressure sensors (load sensing) are modelled to represent a 2D section of a smart rotor. The model is used in conjunction with a first-order lag actuator model to undertake a detailed investigation into the level of control possible by, as in other areas, combining iterative learning control with classical control action with emphasis on how performance can be eeffectively measured. Typical turbine flow regimes are simulated by generating multiple upstream vortices, drifting turbine time periods, stochastic in flow conditions and a combination of all three regimes. Results indicate that cyclical and stochastic loadings on turbine blades can be effectively managed using Iterative Learning Control, with significant reductions in both fatigue and extreme loads for a range of flow conditions