Model Predictive Control of a Wind Turbine using Short Term Wind Field Predictions

Abstract

As wind turbines become larger and hence more flexible the design of advanced controllers to mitigate fatigue damage and optimise power capture is becoming more important. The majority of the existing literature focuses on feedback controllers which use measurements from the turbine itself and possibly an estimate of the current local wind speed. This thesis investigates the design of controllers which can use short-term predictions about the approaching wind field in order to improve performance by compensating for measurement and actuation delays. Simulations are carried out using the FAST aeroelastic design code with the NREL 5MW reference turbine, and controllers are designed in both above and below rated wind conditions using Model Predictive Control and then tested in various wind conditions and with different future wind field information available. It is shown that in above rated wind conditions, significant fatigue load reductions are possible compared to a controller which knows only the current wind profile, although this depends very much on the speed of the pitch actuator response and the wind conditions. In below rated wind conditions the goals of power capture and fatigue load control were considered separately. It was found that power capture could only be improved using wind preview if the wind speed changed rapidly during the simulation, and that fatigue loads were not consistently reduced when wind preview was available, indicating that wind preview is of limited benefit in below rated wind conditions.