Real-time random reference tracking nonlinear model predictive control: a case study on wind turbines

Recently, a research effort to extend nonlinear model predictive control methods from setpoint stabilization to reference tracking has been felt increasingly. On the other hand, uncertainty in the reference signal and the requirement for its dynamic forecasting in applications such as wind turbine control motivate the need for robust tracking nonlinear model predictive control approaches more and more. Therefore, this study proposes a random reference tracking nonlinear model predictive control with dynamic forecasting of stochastic references. Convergence to a robust invariant set is guaranteed by an additional constraint limiting the previous step's tracking stage cost function. The proposed predictive approach is implemented using a parallel Newton-type method to make it more efficient and applicable. The proposed approach for wind turbine control is designed considering the random wind speed reference. Simulations are performed for extreme and fatigue load scenarios. The results show that the proposed controller performs more robustly than the nominal nonlinear model predictive control approach, performing better in optimal power extraction and reducing aerodynamic loads.