Date of Award

Spring 1-1-2012

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical, Computer & Energy Engineering

First Advisor

Lucy Y. Pao

Second Advisor

Alan Wright

Third Advisor

Mark Balas

Fourth Advisor

John Hauser

Fifth Advisor

Jason Marden

Abstract

This research investigates the use of model predictive control (MPC) in application to wind turbine operation from start-up to cut-out. The studies conducted are focused on the design of an MPC controller for a 650 KW, three-bladed horizontal axis turbine that is in operation at the National Renewable Energy Laboratory's National Wind Technology Center outside of Golden, Colorado. This turbine is at the small end of utility scale turbines, but it provides advanced instrumentation and control capabilities, and there is a good probability that the approach developed in simulation for this thesis, will be field tested on the actual turbine. MPC is an active area for turbine control research, because wind turbine operation is complicated by multiple factors that are intrinsic to harvesting power from the wind resource:

  • Since the goal of the turbine is to produce as much energy as possible from the available power in the air ow passing through the turbine's rotor plane, either the turbine's blade pitch (used to regulate aerodynamic torque), or the generator load torque (used to regulate rotor speed at the optimal tip-speed-ratio) are routinely set at the limits of their operating range.
  • There is a significant variation in the gain from perturbations in blade pitch to perturbations in bending moments and torque. This variation is dependent on the relative speed between the blade and wind, and the nominal blade pitch. As a result, gain scheduling techniques are found to be necessary in order to obtain adequate speed regulation, and optimal load mitigation.
  • The three individual pitch (IP) commands and the generator load command, along with structural loads that can be in conflict with speed regulation objectives, make the turbine control problem inherently multi-input-multi-output (MIMO) in nature.
  • Advanced measurement technologies like LIDAR (light detection and ranging) make the use of preview control plausible in the near future.

Standard formulations of MPC accommodate each of these issues. Also, a common MPC technique provides integral-like control to achieve offset-free operation [9]. At the same time in wind turbine applications, multiple studies [38, 5, 73] have developed \feed-forward" controls based on applying a gain to an estimate of the wind speed changes obtained from an observer incorporating a disturbance model. These approaches are based on a technique that can be referred to as disturbance accommodating control (DAC) [32]. In this thesis, it is shown that offset-free tracking MPC [52] is equivalent to a DAC approach when the disturbance gain is computed to satisfy a regulator equation. Although the MPC literature has recognized that this approach provides \structurally stable" [20] disturbance rejection and tracking, this step is not typically divorced from the MPC computations repeated each sample hit. The DAC formulation is conceptually simpler, and essentially uncouples regulation considerations from MPC related issues. This thesis provides a self contained proof that the DAC formulation (an observer-controller and appropriate disturbance gain) provides structurally stable regulation.

Share

COinS