Back to the programme printer.gif Print




Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'Advanced rotor technologies' taking place on Tuesday, 11 March 2014 at 11:15-12:45. The meet-the-authors will take place in the poster area.

Lars Henriksen DTU, Denmark
Co-authors:

(1) DTU, Roskilde, Denmark

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Lars Christian Henriksen received his M.Sc. degree in 2007 electrical engineering from the Technical University of Denmark. In 2011 he received his Ph.D. degree from Risø National Laboratory for Sustainable Energy, Technical University of Denmark. He is currently employed as a Research Scientist at the Technical University of Denmark where he works with his main areas of interest: Control theory, in particular model-based control and fault diagnosis of wind turbines. He is also working on novel concepts including Lidars, trailing edge flaps and floating wind turbines.

Abstract

Investigation of the dependency of wind turbine loads on the simulation time

Introduction

One of the problems of holistic wind turbines numerical optimization is the selection of a cost function. The cost function should be representative of the model physics and should capture the effects of design variables changes on the system. Furthermore when the cost function is based on parameters computed from aeroelastic simulations, it should represent the life time loads and not just the selected conditions described in the set of simulations. In this work we investigate the dependency of several wind turbine parameters with respect to the length of simulations used for their evaluation to identify a suitable cost function.

Approach

The dependency of several wind turbine loads with respect to the number of DLC 1.2 simulations is first investigated. The analysis is performed computing the parameters with a different number of simulations and repeating the computation with different turbulence realizations. The repetition of the computation is performed to identify the scatter of the parameters for a given number of simulations due to the different turbulence realization. To estimate the loads reliability to describe the model physics also their dependency with respect to wind turbine parameters changes is investigated. The analysis is based on aero-servo-elastic simulations performed with HAWC2.

Main body of abstract

Results from the investigations show that even with a high number of turbulence seeds the fatigue loads depend on the turbulence seeds considered for theirs evaluation. It also appears that when increasing the simulation time the loads do not always converge monotonically to a value. Hence even when many simulations are considered to evaluate a load, the addition of another simulation can change significantly the results. This behavior, for example, appears to happen with blade root flapwise bending moment damage equivalent load that is often used as a parameter to evaluate the blade loading in optimization applications. When changing wind turbines parameters the variations of the loads can be of comparable amount with the one due to a different wind turbulence realization. This can be critical because when a design variable is changed the wind turbine response is different and so the wind seen by it, e.g. due to a different rotor speed. Hence if the load dependency on the design variables is comparable with the one on the wind, the optimization algorithm could attribute a change in the loads to a design variable when instead it is due to the wind turbulence. This could lead to erroneous conclusions when using gradient based methods with gradients computed numerically.

Conclusion

From the investigations it appears that even with a high number of turbulent wind simulations wind turbines loads depend on the turbulence realization that is considered. This dependency can be problematic for numerical optimization. Indeed changes in the loads due to the wind seen by the wind turbine can be attributed to changes in the design variables. This attribution would lead to a wrong estimation of the dependency of the cost function from the design parameters. Hence the optimization would not converge or it would but not to a significant minimum point for the design.


Learning objectives
Caution must be taken when performing wind turbine optimization when the cost function is computed from aeroelastic simulations. Wind turbine loads are highly affected by the stochastic nature of turbulent wind even when a large amount of wind conditions is taken into account. Hence a cost function can be not enough reliable and robust with respect to turbulence to guarantee the convergence of a traditional optimization procedure.