Browsing by Author "Sindareh-Esfahani, Peyman"

Now showing 1 - 1 of 1
  • Thumbnail Image
    ItemOpen Access
    Machine Learning Modeling and Robust Model Predictive Control of a Wind Turbine
    (2019-10-17) Sindareh-Esfahani, Peyman; Pieper, Jeff; Chen, Tongwen; De La Hoz Siegler, H.; Ramírez Serrano, Alejandro
    In the era of fast-emerging technologies for large-scale wind turbines, demand on control systems is significantly increased. Control systems improve the capability of the turbine to efficiently capture wind energy and increase its lifespan. The aim of this research is to develop a comprehensive model-based controller for a wind turbine. To do so, the problem of Robust Model Predictive Control (RMPC) of discrete-time hybrid systems is addressed. A hybrid model of the wind turbine is identified with machine learning techniques and the proposed controller is implemented to demonstrate the effectiveness of the proposed control system to the wind turbine. For controller system design, this study contributes a new H-infiniti RMPC strategy for PieceWise Affine (PWA) systems and also RMPC scheme for a class of discrete-time switched linear systems, as the first and second topic of this thesis, respectively. Both PWA and switched systems are special classes of hybrid systems. The problem of minimization of the cost function for model predictive control (MPC) design is converted to minimization of the worst case of the cost function. Then, this objective is reduced to minimization of a supremum of the cost function subject to a terminal inequality by considering the induced l2-norm. Simulation results are provided to show good convergence properties along with the capability of the proposed controller to reject disturbances. PWA models of the wind turbine during Maximum Power Point Tracking (MPPT) and Power Regulation (PR) regions are identified, as the third and fourth topic, respectively. PWA models provide opportunity to identify the affine model through multi-dimensional operating point linearization using clustering-based identification technique. Then, as the fifth topic, the proposed RMPC technique for PWA models is applied to the PWA models of the wind turbine for different operations, satisfying the objectives of different working conditions. Transient between MPPT and PR regions is a challenging problem since the different dynamics and control objectives are associated with these regions. Therefore, as the sixth topic, the concept of switched linear systems is proposed for transition mode. Afterward, the proposed RMPC for switching systems is applied to the switching model of the wind turbine so that the whole operation of the turbine covered. The proposed control techniques, as well as the developed models, are validated and examined through a wind turbine simulator named FAST, and a commercial software package. The characteristics of the 5MW NREL offshore variable speed variable pitch wind turbine are considered for implementation and study purposes.