Model predictive control of wind turbines
Author
Katsifas, Vasilis
Term
4. semester
Education
Publication year
2016
Submitted on
2016-06-08
Pages
60
Abstract
Denne afhandling undersøger, hvordan modelprædiktiv regulering (MPC) kan anvendes til at styre en vindmølle og sammenlignes med klassiske PI-regulatorer. Der opbygges en fysikbaseret, ikke‑lineær model af møllens centrale delsystemer (aerodynamik, drivtog, generator, tårn og aktuatorer), som valideres under forskellige ydre forstyrrelser og efterfølgende lineariseres omkring drift i det nominelle (rated) område. Med udgangspunkt i receding‑horizon‑princippet formuleres et optimeringsproblem, og der designes tre MPC‑regulatorer med løsningskode genereret i CVXGEN. Regulatorerne tunes for forskellige horisonter og afprøves i simuleringer for at vurdere deres evne til at håndtere afvigelser og begrænsninger. Ydelsen sammenlignes med PI‑strategier ved hjælp af blandt andet Power Spectral Density‑analyse, hvorefter den bedst egnede MPC implementeres på den ikke‑lineære model og holdes op mod en baseline‑regulator. Resultaterne indikerer, at de foreslåede MPC‑regulatorer kan håndtere systemets afvigelser og belyser fordele og ulemper ved at anvende avancerede metoder på relativt simple modeller; detaljerede kvantitative forbedringer fremgår ikke af dette uddrag.
This thesis examines the use of model predictive control (MPC) for wind turbine control and compares it with classical PI strategies. A physics‑based, nonlinear model of key turbine subsystems (aerodynamics, drivetrain, generator, tower, and actuators) is developed, validated under various external disturbances, and then linearized around operation in the rated region. Based on the receding‑horizon concept, an optimization problem is formulated and three MPC controllers are designed using code generated with CVXGEN. The controllers are tuned for different horizon lengths and tested in simulations to assess their ability to handle deviations and constraints. Performance is compared with PI controllers using, among other metrics, Power Spectral Density analysis, after which the best‑performing MPC is implemented on the nonlinear model and evaluated against a baseline controller. The results indicate that the proposed MPCs can handle system deviations and summarize the advantages and disadvantages of applying advanced control to relatively simple models; detailed quantitative improvements are not available in this excerpt.
[This summary has been generated with the help of AI directly from the project (PDF)]
Keywords
model ; predictiveq ; control
Documents