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 'How does the wind blow behind wind turbines and in wind farms?' taking place on Tuesday, 11 March 2014 at 16:30-18:00. The meet-the-authors will take place in the poster area.

Luca Lavaroni Loughbourough University, United Kingdom
Co-authors:
Luca Lavaroni (1) F P Simon Watson (1) Malcolm Cook (1) Mark Dubal (2)
(1) Loughbourough University, Loughborough, United Kingdom (2) E.ON New Build & Technology Ltd, Nottingham, United Kingdom

Printer friendly version: printer.gif Print

Presenter's biography

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

Mr Lavaroni is currently in the last year of his Engineering Doctorate studies at Loughborough University and E.On New Built & Technology Ltd in the UK, using computational fluid dynamics to model air flows through large off-shore wind farms. Previously he has worked in project management and has a Masters degree in renewable energy and energy efficiency from the University of Nottingham.

Abstract

COMPARISON OF UNSTEADY TURBULENCE MODELS FOR COMPUTATIONAL FLUID DYNAMICS (CFD) SIMULATIONS OF AN OFFSHORE WIND FARM.

Introduction

The installation of wind turbines in large off-shore clusters leads to known issues such as a reduced power production due to the induced velocity reduction and higher fatigue loads generated by increased turbulence [1]. The present paper investigates advanced computational fluid dynamics turbulence models to better understand unsteady wake propagation downstream of the wind turbines in off-shore conditions.

Approach

The methodology used in this paper employs computational fluid dynamics (CFD) methods and modelling techniques that are implemented to improve the knowledge of the wake propagation. In particular, enhanced physics and rotor modelling techniques are embedded into the commercial CFD software Ansys CFX. This is achieved through a set of pre and post processing tools, called WindModeller [2], that builds up the physics through the use of the incompressible flow equations, the modelled turbulence, the atmospheric stability conditions and the boundary layer profile. The modelling of each wind turbine is obtained through the use of an actuator disc model which represents a constant load distribution on the rotor through the exerted force on the surface of the disc [2]. Atmospheric stability conditions are implemented in the model, imposing neutral stability conditions in the surface layer and stable conditions in the free stream. The potential temperature θ is embedded in the vertical momentum and turbulent transport equations to compute the buoyancy force [3]. The momentum equations include the Coriolis force, making use of the assumption that the free stream flow satisfies geostrophic balance. As for the velocity components, the Zilitinkevich et al. [4] similarity theory is used to define the inflow profiles. Expressions for the turbulent kinetic energy, k, and the dissipation rate of turbulent kinetic energy, ε, are then used to achieve preservation of the profiles defined at the inlet over the development of the boundary layer [3]. Appropriate time step selection is chosen on the basis of the maximum Courant Friedrichs Levy (CFL) number smaller than 1 to achieve solution stability of the simulations.

Main body of abstract

In this paper a computational fluid dynamics (CFD) simulation of the wind flow at Nysted off-shore wind farm will be presented. The aim of the study is to compare several unsteady modelling techniques and achieve accurate results in the representation of the wind turbine wake flow. In particular, Unsteady Reynolds-Averaged-Navier-Stokes (URANS), Scale Adaptive Simulation (SAS) and Zonal Large Eddy Simulation (ZLES) are used to investigate the flow pattern downstream of a turbine within neutral atmospheric stability conditions in the surface layer and stable in the free surface. SAS is a hybrid model that makes use of flow instabilities to produce turbulent structures in regions with large separation. The von Karman length-scale is the main parameter introduced into SAS with regard to the turbulence scale equation. SAS models are able to adapt to resolved structures in a dynamic way in a URANS simulation through the von Karman length-scale: consequently unsteady areas of the flow field present LES-like capabilities, while in the regions where stable flow occurs SAS generates RANS results [5]. Whilst URANS model results can describe the large scale unsteadiness of the flow, the SAS Shear Stress Transport (SST) turbulence model dynamically adapts to the resolved scales, developing a spectrum of turbulence in the areas characterised by detached flow [5]. However, past studies have shown that in some cases the flow does not present the necessary instability or the unsteadiness is not sufficient [6] for SAS to function correctly. In these situations, RANS and LES models can be set-up in domain areas that are established in a pre-processing stage. At the interface between RANS and LES domains the modelled turbulent kinetic energy is explicitly transferred to the resolved turbulent kinetic energy through a forcing term [7].
In this paper simulations are performed using the standard actuator disc methodology and embedding neutral stability conditions in the surface layer and stable conditions in the free stream. A mesh of approximately 50 million cells is generated in order to resolve the wake structure, setting-up a domain containing two Siemens wind turbines with a nameplate capacity of 2.3 MW, a rotor diameter of 82.4 m and a hub height of 68.8 m. The distance between the two turbines is 6 rotor diameters, the radius of the cylindrical computational domain is 3 km and the height 1 km. The most refined area in the domain stretches from 5 rotor diameters upstream of the first turbine to 8 diameters downstream of the second turbine, while the RANS-LES interface is located 20 cells on the inside of the refined area. The results are compared to the field data recorded at Nysted wind farm over a period from September 2006 to July 2008. The field data for an upstream wind speed of 10±1m/s at hub height are filtered using the Gradient Richardson Number, to select neutral atmospheric stability conditions with the wind coming from a southerly direction. The results indicate that the SAS model remains in RANS mode, as the flow is not unsteady enough to produce LES-like results, while the Zonal LES satisfactorily describes the unsteady flow field and the turbulent spectrum.


Conclusion

This paper shows the results of unsteady turbulence model simulations of the flow around two wind turbines at Nysted off-shore wind farm with the aim of analysing the wake propagation with better accuracy. The simulations make use of the actuator disc to model the turbines and take into account neutral atmospheric stability conditions in the surface layer and stable conditions in the free stream and Coriolis force. The findings are then compared to field data filtered by direction, wind speed and neutral atmospheric stability conditions. It is found that whilst the URANS simulation generates an acceptable flow field with large scale unsteadiness, the SAS model is not able to trigger the LES-like behaviour because of the likely insufficient flow unsteadiness. For this reason the SAS model presents RANS results in the whole domain and therefore does not produce realistic models of the wake structure. Conversely, the zonal LES successfully propagates the transition from bulk modelled turbulent kinetic energy into resolved turbulence from the RANS to the LES region through an interface and with the use of a forcing term known as a harmonic flow generator. It is also concluded that when unsteady simulations are considered, the Courant Friedrichs Levy (CFL) number needs to be less than 1. This ensures temporal accuracy, as the CFL number describes the stability of the solution. With regard to the LES mesh, it is assumed that an acceptable resolution is achieved when the resolved turbulent kinetic energy is not less than 80% of the total turbulent kinetic energy [8].


Learning objectives
The present work gives an insight into the application of unsteady turbulence models and, in particular, on the set-up of a zonal LES model. Recommendations are given with regard to mesh set-up and refinement.


References
[1] Barthelmie RJ, Frandsen ST, Rathmann O, Hansen K, Politis ES, Prospathopoulos J, Cábezon D, Rados K, van der Pijl SP, Schepers JG, Schlez W, Phillips J, Neubert A. Flow and wakes in large wind farms in complex terrain and offshore. Proceedings of AWEAC, 2008, Houston.

[2] Montavon C, Jones I, Staples C, Strachan C, Gutierrez I. Practical issues in the use of CFD for modelling wind farms. EWEA proceeedings, 2009, Marseille.

[3] WindModeller Manuals, 2012, ANSYS UK Ltd.

[4] Zilitinkevich S, Johansson PE, Mironov DV, Baklanov A. A similarity-theory model for wind profile and resistance law in stably stratified planetary boundary layers. J. Wind Engineer. Industr. Aerodyn. 1998, 74-76 : 209-218.

[5] Menter FR and Egorov Y. A Scale-Adaptive Simulation Model using Two-Equation Models. AIAA paper 05-1095, 2005, Reno/NV.

[6] Davidson L, Billson M. Hybrid LES-RANS using synthesized turbulent fluctuations for forcing in the interface region. International Journal of Heat and Fluid Flow, 2006, 27 (6), 1028-1042.

[7] Menter FR, Garbaruk A, Smirnov P. Scale Adaptive Simulation with Artificial Forcing. Proc. 3rd Symposium on Hybrid RANS-LES Methods, 2009, Gdansk.

[8] Pope SB. Ten questions concerning the large-eddy simulation of turbulent flows. New Journal of Physics 6, 2004.