Inclusion of Dynamic Inflow in Model Predictive Control of Wind Turbines

Steffensen, Henrik ; Overgaard, Anders ; Jørgensen, Martin Østergaard

4. term

Control and Automation, Master

2014

2014-06-03

152 pages

Denne afhandling analyserer behovet for inkluder- ing af dynamisk inflow i modelprediktiv regulering af vindmøller. De undersøgte hovedområder er effekt- produktion og tårntræthed. Projektet er foreslået og delvist vejledt af Vestas Wind Systems A/S i Aarhus, Danmark. I denne afhandling anvendes en 5 MW vindmølle fra NREL, som er en 3-vinget landbaseret opvinds, horisontal-akset vindmølle, der simuleres i FAST. Der er lavet en analyse af systemet og problemet, der klarlægger hvilke ting, der skal med i designet af en modelprediktiv regulator, der inkluderer dynamisk inflow. Derudover er der lavet en evaluering af de mulige forbedringer, som kan vindes ved inkludering af dynamisk inflow. I analysen er metoden rainflow counting undersøgt og anvendt til at udregne tårn- træthed, hvor en skadesækvivalent belastning bliver udregnet. Der opstilles en model, som inkluderer aerodynamik, drivtog, tårndynamik, vingeaktuator og generator. Herfra bliver modellen delt op i to modeller. En der anvender en quasistatisk inflow model og en som an- vender en simpel dynamisk inflow model. Modellerne er anvendt i to modelprediktive regulatorer, som er sammenlignet. Denne sammenligning er lavet ved hjælp af Pareto kurver, da der eksisterer en afvejn- ing imellem effektproduktion og tårntræthed. Sammenligningen viser en markant forbedring lige over rated vindhastighed, hvor dynamisk inflow er mest markant. Den modelprediktive regulator med information om dynamisk inflow er i stand til at forbedre træthedsækvivalentet med mere end 20%, samtidig med en forbedring på 5% i den gennem- snitlige absolutte effektfejl.

This thesis investigates the need for including dy- namic inflow in Model Predictive Control (MPC) of wind turbines. The main design drivers are gener- ated power and tower fatigue. The project has been proposed and partly supervised by Vestas Wind Sys- tems A/S in Aarhus, Denmark. The wind turbine used in this thesis to generate re- sults is the NREL 5 MW wind turbine, which is a 3 bladed onshore upwind HAWT simulated in FAST. An analysis of the system and the problem is made that clarifies, which things that should be addressed in the design of an MPC including dynamic inflow. Furthermore, an initial evaluation is made of the pos- sible improvements from including dynamic inflow. In this analysis the rainflow counting method is also shown as a measurement of fatigue, where a Damage Equivalent Load is found. A model including aerodynamics, drivetrain, tower dynamics, pitch actuator and generator is estab- lished. From here the model is split into two models. One where the aerodynamics is a quasi-steady model and one where the aerodynamics contain a simple dynamic inflow model. These models of the wind turbine are used for two MPCs, which are then com- pared. The comparison are done as Pareto fronts, as there exist a trade-off between power generation performance and tower fatigue. The results show significant improvements just above rated wind speed, where dynamic inflow is most out- spoken. The MPC with dynamic inflow information is able to improve the fatigue by more than 20%, while also improving the mean absolute error from the reference of the power at about 5%.

Wind Turbine ; MPC ; Dynamic Inflow ; Fatigue ; Model Predictive Control

Download
View record in AAU Student Projects