Control of a Wind Turbine
Authors
Dolmer, Anders Katborg ; Kristensen, Oliver Engmann ; Nielsen, Sebastian William Vejlgaard
Term
4. term
Education
Publication year
2026
Pages
100
Abstract
This project addresses retrofitting of aging wind turbines by developing a model predictive controller (MPC) that explicitly handles actuator constraints and coupled, nonlinear dynamics while smoothing transitions between control objectives. A physics-based dynamic turbine model was derived using the Euler–Lagrange formulation combined with blade element momentum (BEM) theory and integrated with a parameter estimation algorithm to identify unknown system parameters; the resulting unified model serves as the MPC prediction model. To improve accuracy, gain scheduling with wind speed as the scheduling variable was applied, and a statistically validated wind prediction model was incorporated to enhance performance over longer horizons. The scope was limited by omitting control regions 1 and 4 and assuming perfect yaw alignment. The model predicts rotor speed and fore–aft tower displacement satisfactorily, while side-to-side tower motion is less accurate due to aerodynamic simplifications. Wind prediction performs well over short horizons but deviates more over longer horizons due to wind stochasticity. In OpenFAST simulations, the MPC was benchmarked against a conventional baseline; it met the weighted objectives and improved constraint handling and transitions, but did not consistently outperform the baseline across all performance metrics, with gains mainly in targeted objectives.
Dette projekt undersøger retrofit af ældre vindmøller ved at udvikle en modelprædiktiv styring (MPC), der eksplicit håndterer aktuatorbegrænsninger og ikke-lineære, koblede dynamikker, samtidig med at skift mellem driftsmål gøres mere glidende. En fysikbaseret, dynamisk model af møllen blev opstillet med Euler–Lagrange-formuleringen kombineret med blade element momentum (BEM) teori og koblet til en parameterestimeringsalgoritme for at identificere ukendte systemparametre; modellen bruges som forudsigelsesmodel i MPC’en. For at forbedre nøjagtigheden anvendes gain scheduling med vindhastighed som skaleringsvariabel, og en statistisk valideret vindforudsigelsesmodel integreres for at forbedre præstationen over længere horisonter. Projektets scope afgrænses ved at ignorere kontrolregioner 1 og 4 og antage perfekt yaw-tilpasning. Modellen viser tilfredsstillende forudsigelse af rotorhastighed og for-/aft-tårnafdrift, mens side-til-side-afdrift er mindre nøjagtig pga. aerodynamiske simplifikationer. Vindforudsigelsen er nyttig for korte horisonter, men afviger mere for længere pga. vindens stokastiske natur. I OpenFAST-simulationer blev MPC sammenlignet med en konventionel baseline; MPC’en opfylder de vægtede mål og håndterer begrænsninger og skift bedre, men den overgår ikke konsekvent baseline på tværs af alle ydelser, med forbedringer primært i de målrettede metrikker.
[This apstract has been generated with the help of AI directly from the project full text]