Repetitive Control of Individual Pitch to Reduce Wake Effect on Wind Turbines
Translated title
Authors
Term
10. term
Publication year
2012
Submitted on
2012-05-30
Pages
166
Abstract
Efterhånden som windmøller bliver større, med rotordiametre over 100 m, vil vingerne dække et større vindfelt som indeholder en række forskellige vindfænomener, såsom wakes, vindforskydning, tårnskygge, hvilket vil påføre store strukturelle laster på vindmøllen. I denne these er en løftet repetitive regulator udviklet, som reducere de strukturelle laster ved brug af individuel pitching. Til dette formål er en dynamisk model af en vindmølle udledt. Modellen indeholder en aerodynamiskmodel, mekaniskmodel, en strukturelmodel og en model af pitchingsystemet. Modellen er blevet lineariseret og valideret i forhold til simulerings koden FAST. Fra modellen er en løftet repetitive regulator blevet udviklet, ved at lave en løftet system beskrivelse, og herfra blev en reduceret output-feedback formulering fundet, hvilken lagde grunden til LQR designet af regulatoren. I en accepttest blev den løftede repetitive regulator samlignet med regulatoren fra FAST som var implementeret i Matlab. Resultatet herfra var at accepttesten ikke blev godkendt, selvom bøjningen af tårn og vinger blev reduceret. Det formodes at dette skyldes en forskel mellem implementeringen i Matlab og modellen.
As wind turbines get larger, with rotor diameters above 100 m, the blades will sweep a large wind field, containing different wind phenomenas such as, wakes, wind shear and tower shadow; applying a big structural load upon the wind turbine. In this thesis, a lifted repetitive controller is developed which reduces the structural loads by the use of individual pitching. For this purpose a dynamic model of a wind turbine has been developed. The model contains an aerodynamic model, mechanical model, a structural model and a model of the pitch system. The model has been linearized and validated in accordance with simulation code FAST. From the model a lifted repetitive controller was design, by making a lifted system description where made, and from this a reduced output-feedback formulation was found, making it possible to use LQR design to calculate the controller gain. In an acceptance test, the lifted repetitive controller was compared to the controller from FAST which was implemented in Matlab. The results from this were, that the controller designed did not pass, even though the deflection of tower and blades were reduced. It is assumed that it is caused by a mismatch between the model and the implementation in Matlab.
Keywords
Documents