Ricardo Guilherme Duraiski

Comparison of linear and nonlinear model predictive control of wind turbines using LIDAR

Recent developments in remote sensing are offering a promising opportunity to rethink conventional control strategies of wind turbines. With technologies such as LIDAR, the information about the incoming wind field - the main disturbance to the system - can be made available ahead of time. Feedforward control can be easily combined with traditional collective pitch feedback controllers and has been successfully tested on real systems. Nonlinear model predictive controllers adjusting both collective pitch and generator torque can further reduce structural loads in simulations but have higher computational times compared to feedforward or linear model predictive controller. This paper compares a linear and a commercial nonlinear model predictive controller to a baseline controller. On the one hand simulations show that both controller have significant improvements if used along with the preview of the rotor effective wind speed. On the other hand the nonlinear model predictive controller can achieve better results compared to the linear model close to the rated wind speed.

NMPC with state-space models obtained through linearization on equilibrium manifold

Abstract This paper presents a Nonlinear Model Predictive Control approach for nonlinear state-space models obtained with the modelling and identification technique recently proposed in literature as Linearization on the Equilibrium Manifold (LEM). The predictive controller that will be applied to the LEM uses the Local Linearization on the Trajectory algorithm (LLT) which simulates the nonlinear plant and calculates optimal control actions based on local linearizations around the simulated trajectory by online minimization of an objective function. The proposed combination of the LEM and LLT techniques is tested with a nonlinear SISO system.

RPN tuning strategy for model predictive control

Abstract A novel tuning strategy based on RPN for MIMO MPC is presented. The RPN indicates how potentially difficult it is for a given system to achieve the desired performance robustly. It reflects both the attainable performance of a system and its degree of directionality. These systems properties are the basis of the proposed RPN-MPC tuning strategy, which is applied in the controller design of an air separation plant. Although it was only used a linear nominal model, the results can also be applied at least at some extent for nonlinear systems with uncertainties.