Sunteți pe pagina 1din 9

EACWE 5 Florence, Italy 19th 23rd July 2009

Flying Sphere image Museo Ideale L. Da Vinci

CFD Modeling of the wake interactions of two wind turbines on a Gaussian hill
Alexandros Makridis, John Chick
Institute for Energy Systems, University of Edinburgh a.makridis@ed.ac.uk Kings Buildings, Edinburgh Institute for Energy Systems, University of Edinburgh john.chick@ed.ac.uk Kings Buildings, Edinburgh

Keywords: wind turbines, wakes, complex terrain, CFD modeling.

ABSTRACT One part of the solution to the problem of climate change is the use of wind energy for electricity generation. Onshore wind farms usually consist of a number of horizontal axis wind turbines closely spaced in clusters and they are often sited on complex terrain. Various models have been developed in the past to estimate the wake effects and optimize the siting of a wind farm, but they were mainly developed and validated for flat terrain and not for complex terrain sites. Computational Fluid Dynamics (CFD) can be a useful tool towards the understanding of wake behavior on complex terrain. The present work involves the use of FLUENT 6.3 commercial CFD code to study the wake interaction of two wind turbines closely spaced next to each other at the top of an ideal Gaussian hill.

Contact person:

Alexandros Makridis, Institute for Energy Systems University of Edinburgh, Mayfield road, Kings Buildings, EH9 3JL, Edinburgh, telephone: +447833521488, FAX: +44 (0)131 6506554. E-mail: a.makridis@ed.ac.uk

1. BACKGROUND The operation of a wind turbine (WT) in a wind farm will inevitably affect the others in its vicinity due to velocity deficit and increased level of turbulence in its wake reducing the power output and increasing the dynamic loading on the machines. Wind farms are often built as close as possible to each other in clusters due to land use restrictions causing wind turbine wakes to interact and further contribute to power losses and dynamic loads. It is estimated, that losses in gross energy production due to wakes may vary from 5% to more than 15% depending on the wind farm layout, Smith et al. (2006). It can be easily concluded that understanding of the factors affecting wake characteristics is important for the optimal design of a wind farm. Wind turbine wakes have been a topic of research for the last 30 years. Crespo et al. (1999) presented an overview of the rich literature on different wake modelling methods and Vermeer et al. (2003) conducted a more up-to-date assessment of the latest research on wind turbine wakes. Most of the simple (kinematic) models which are commonly used have been validated for flat terrain cases and some of them assume uniform unperturbed wind velocity, which is not the case for many applications such as wind farms in complex terrain. A more recent review of the simple models used for wake modeling has recently been published by Duckworth and Barthelmie (2008). Ainslie (1988) developed a parabolic eddy viscosity model (EVMOD) assuming axisymmetric wake flow and ignoring ground effects and flow variations with height. The approach provided reasonable results when compared to wind tunnel experiments. Improvements to the model were applied by the University of Oldenburg developing the FLaP model aimed at offshore wind farm applications, Lange et al. (2003). Crespo et al. (1985) developed the UPMWAKE model taking into account the ground roughness and atmospheric stability and using the k model for turbulence closure. The model predictions were validated using the commercial CFD PHOENICS code as well as wind-tunnel and full-scale experiments providing acceptable agreement with the exception of the near wake region, where the predicted velocity deficit was smaller than the measured, Hernandez and Crespo (1990). ECN developed the WAKEFARM model based on the UPMWAKE code, Schepers (2003). The UPM-ANIWAKE model was also developed as an extension of the UPMWAKE code to consider anisotropic turbulence in the wake effects, Gomez-Elvira et al. (2003). The anisotropy of turbulence in wind turbine wakes was further studied using an algebraic stress model by Gmez-Elvira et al. (2005). A common approach in CFD wind turbine wake modeling is to use the actuator disk concept rather than model the actual rotor geometry. According to this approach, the rotor is represented by forces acting as momentum sinks in the equations. Ammara et al. (2002) used the actuator-disk concept by applying time-averaged forces on the Reynolds Averaged Navier-Stokes (RANS) equations. Srensen et al. (1999) and Mikkelsen et al. (2001) also used the actuator-disk concept and later developed the actuator-line technique where the forces act on rotating lines which represent the blades, Srensen and Shen (2002). This approach has recently been used to model a row of wind turbines, Mikkelsen et al. (2007) and with Large Eddy Simulation (LES), Troldborg et al. (2007). Dobrev et al. (2007) used the actuator surface approach, representing the blades with their mean surfaces and a pressure jump boundary condition in FLUENT 6.2 commercial CFD code. Most of the above models were applied and validated for flat terrain cases. The literature on wind turbine wakes is rather limited. UPMWAKE model was applied in a moderately irregular terrain wind farm by superposing wake and terrain effects, Crespo et al. (1993). The study showed that if the velocity irregularities of the terrain and the wake velocity deficit are of similar order of magnitude, the linear superposition approach seems to provide acceptable results, although for wake interaction this assumption is generally not valid. The effect of complex terrain on wake behavior was studied in a wind tunnel by Taylor and Smith (1991), showing that complex terrain can have an important effect on the wake characteristics. Stefanatos et al. (1996) and Helmis et al. (1995) also used wind tunnel as well as large scale tests and they obtained some guidelines for modelling the wake-terrain interaction. Chaviaropoulos and Douvikas (1999) developed an in-house Navier-Stokes solver for wind prediction over complex terrain taking into account distributed arrays of wind turbines acting as

momentum absorbers. The k model was used for turbulence closure and the method was used to investigate single wake and wake interaction on flat terrain. More recently, Prospathopoulos et al. (2008) also used a Navier Stokes solver along with k closure to investigate a single wake on two Gaussian hill cases considering varying wind directions and different ambient turbulence intensities corresponding to different roughness lengths. The work was part of a larger project (UpWind project) investigating wakes of large wind farms in complex terrain and offshore, Barthelmie et al. (2008). The current work involves the investigation of the interaction of two wind turbines in complex terrain, focusing on the complex terrain effect on wakes. Two different turbulence models, k- Shear Stress Transport (SST) and the anisotropic Reynolds Stress Model (RSM) were used along with the 3D RANS equations in the commercial CFD code FLUENT 6.3. The rotor disk is simulated using FLUENTs Virtual Blade Model (VBM), Ruith et al. (2004). Simulations are performed for the case of neutral atmospheric flow over a 3D Gaussian hill terrain with the two wind turbines on the hilltop facing the wind.

2. METHODOLOGY

2.1

Virtual Blade Model

VBM is based on the Blade Element Theory (BET) according to which the WT blade is divided in a number of spanwise sections and the rotor performance is then calculated by summing the contribution of each section taking into account varying characteristics such as chord length, airfoil type, and blade twist and pitch angle. The effect of each airfoil type is considered indirectly by using aerodynamic tables of lift and drag coefficient vs. angle of attack without the need of creating and meshing the actual rotor geometry, significantly reducing the required number of cells and computational time. The rotor diameter, collective pitch and rotational speed, as well as the WT hub height are also inputs in the model. VBM considers all the above properties and applies the calculated aerodynamic forces of each section in the RANS equations allowing the pressure drop across the disk to vary with radius and azimuth. It is, thus, a more detailed approach than simply applying the actuator disk concept by means of a uniform pressure drop without the rotational effect in the wake. The modeled WT properties were chosen to match the general characteristics of the NREL 5MW machine, Jonkman (2007). The chosen rotor diameter is 126 m and the hub height is 90 m. The 3 bladed rotor is upwind orientated and is controlled by variable speed and collective pitch. Rated wind speed and rotor speed is 11.4m/s and 12.1rpm respectively. The rotor blade was divided into 19 sections of varying chord length, twist and airfoil type. Airfoil aerodynamic tables for each of the different airfoil type were also added. 2.2 Turbulence modeling

Two different turbulence models (k- SST and RSM) are used along with the Reynolds Averaged Navier-Stokes equations to model the atmospheric wind flow. The k- SST (Shear Stress Transport) model belongs to the family of two-equation turbulence models. It is a hybrid model combining the good near-wall behavior of the k- model with the robustness of the k- model away from the wall, Versteeg and Malalasekera (2007). It is expected to perform better in flows with adverse pressure gradients and wake regions, but it is still based on the Boussinesq hypothesis of isotropic turbulence. Some of the model constants were modified according to the assumptions made by Prospathopoulos et al. (2008). The Reynolds Stress Model (RSM) accounts for the anisotropy in turbulent flow. It is a complex

model which includes extra equations for each of the Reynolds Stresses and thus becoming computationally expensive. Nevertheless it was chosen due to the fact that it can better represent the anisotropic atmospheric turbulence as well as the anisotropy of turbulence in the wakes. 2.3 Domain Mesh

The properties of the Gaussian hill were chosen to match the test case in the Upwind project, Prospathopoulos et al. 2008. The idealized complex terrain investigated is a 3D (quasi-2D) Gaussian hill. The hill is chosen to be 700m high with a mean slope value of 0.4 (fig. 1). The rotors are placed at the hilltop with a spacing of 2D (fig. 1). The complex terrain domain has a size of -80D<x<120D; -75D<y<75D; 0<z<80D (fig.1, left), where y=0 is the symmetry plane and x=0 is the plane along the hilltop. The two rotors are positioned at a distance of 1D on each side of the y=0 plane (fig.1, right). The simulation of the two rotors on flat terrain was also considered. The size of the domain is -10D<x<50D; -10D<y<10D; 0<z<8D on that case (fig. 2).

Figure 1: The quasi-2D Gaussian hill domain

Figure 2: The flat terrain domain

A hybrid approach for meshing was successfully applied. The wall adjacent cell height was chosen so that the cell centroids are above the height equal to the roughness length. The mesh around the rotors consists of unstructured tetrahedral elements (fig.3) while the rest of the mesh is structured comprising of hexahedral elements of gradually increasing size from the ground to the top of the domain as well as from the site of the turbines towards the right and left symmetry planes (fig.1-3). The need for denser mesh is due to the high pressure and velocity gradients on some parts of the

domain (mainly close to the rotors, in the wakes as well as close to the ground).

Figure 3: Rotor meshing at x=0 plane (left) and y=0 plane (right)

The final size of the grid is 1,503,679 cells for the flat and 1,550,459 for the complex terrain case and the discretization scheme is 2nd order Upwind and PRESTO! for pressure. 2.4 Boundary Conditions

The inlet values for velocity and turbulence intensity were chosen as 10 m/s and 13% respectively at the hub height of 90 m. Appropriate roughness height was chosen to match the levels of turbulence and a logarithmic inlet velocity profile was applied by means of user defined functions in order to correspond to fully-developed flow in a neutral Atmospheric Boundary Layer, as shown in (1)
U ( z) = z ln z o U ( zref ) where u * ABL = is the friction velocity and zref = 90m : reference height z ref ln z zo = 0.0445m : roughness height o u*

(1)

= 0.4 : Von Karman const.


Turbulence profiles for the SST k- model are shown in eq. (2).
k ( z) = u *2

( z) =

u*

(2)

z
*

where * is a model constant taken as * = 0.033. Appropriate inflow turbulence was also considered for the RSM, shown in eq. (3). Apart from k and , profiles were also taken for each of the Reynolds stresses assuming anisotropic turbulence. The normal Reynolds stresses were taken from Panofsky and Dutton (1984) and the profile for the Reynolds shear stress u w is from ESDU (1985).
u *2 k ( z) = , C

( z) =

u *3 z

(3)
2 z = 1.56, u w = u * 1 , u v = vw = 0 z h

u2
u*
2

= 5.71,

v2
u*
2

= 3.68,

w2
u*
2

(4)

where C is a model constant taken as C = 0.033. FLUENTs standard wall functions were chosen for the wall treatment and a constant horizontal velocity was applied at the top boundary.
3. RESULTS

In the results, horizontal wind velocity is normalized with the velocity at the top boundary, as shown in (5), whereas x and y coordinates are normalized with the rotor diameter, eq. (6). As shown in fig. 4 and 5, the rotor hubs are positioned at (xnorm , ynorm) = (0 , 1) and (xnorm , ynorm) = (0 , -1) respectively. Ux = Ux/ U ynorm = y/D, xnorm = x/D (5) (6)

Figures 4 and 5 show the contour plots of the normalized horizontal velocity at the hub height using RSM (fig. 4) and k- SST (fig.5). The flat terrain simulation results are shown on the left and the complex terrain results on the right. As in all cases, the initial conditions are the same for wind speed and turbulence; the wind speed up on the hill is clear (although the plot is mostly focused in the wake region).

y/D

y/D

Ux x/D Ux

Figure 4: normalized x-velocity contours at hub height for the flat (left) and the hill (right) case using RSM.

x/D

y/D

y/D

Ux x/D

Figure 5: normalized x-velocity contours at hub height for the flat (left) and the hill (right) case using k SST

Ux

x/D

One notable difference between the flat and the hill case is that in the flat case the wind deficit is extended almost 2D upstream the WTs. The k- SST model predicts this region to be bigger. In the hill case, the upwind deficit seems to be a relatively small. Another difference between the two models is shown clearly in the flat terrain case. According to RSM predictions, the two wakes seem to merge after around 8 diameters downstream. The wake also seems to widen in the y-direction. There

is no indication of wake merging or any significant lateral wake expansion in the k- SST results. An important remark after comparing the flat and hill case results is that the wake seems to be symmetrical in the flat terrain case, but not in the complex terrain, especially in the far wake region, more than 10D downwind, which is also the region of rapid wind slowdown and increase in turbulence intensity on the lee side of the hill, most noticeable in the k- SST results. This indicates a non-linear behavior as it is clear that the result would have been different if we superposed the flat terrain wakes and the wind slowdown due to the hill. The above remarks are also shown in the plots of velocity deficit in the lateral (y) direction (fig. 6, 7). The velocity deficit is calculated taking into account the simulation without the WT, as shown in (7). U x ,no WT is the calculated velocity at each specific point without the presence of the WT. This is done mainly due to the fact that in the hill case there is a significant slowdown of the flow downwind, therefore the velocity deficit would need to be related to the local wind speed. U x ,with _ WT U (7) Velocity deficit = x ,noWT U x ,noWT The results were taken for the lateral direction, where -4D < ynorm < 4D, at the hub height considering various distances downwind (mainly focusing on the far wake): 3D, 5D, 8D, 10D, 12D, 14D and 18D. The k- SST model predicts higher deficit in the flat terrain, but lower in the complex terrain. The merging of the two wakes in the RSM simulation is clearly shown in fig.6 (left), whereas in fig. 7 (left) it is clear that in the k- SST simulation each wake almost remains separate and the velocity deficit peaks are found behind each rotor on the ynorm=1 and ynorm=-1 planes.

Figure 6: velocity deficit for flat (left) and hilly terrain (right) using the RSM

Figure 7: velocity deficit for flat (left) and hilly terrain (right) using k- SST model

After a comparison between flat and complex terrain velocity deficits, they are found in general to be higher in the flat terrain cases, but the maximum velocity deficit is increasing in the complex terrain, after around 10D downwind. This can be attributed to the fact that the wind speed is slowing down significantly in that region. The non-linearities in the complex terrain wake interactions are

clearly shown in fig. 6 (right) and especially using the k- SST model in fig. 7 (right), where the wake seems to drift away from its path. Validation of the modeling approach with measurements or wind-tunnel experiments would be needed to further investigate whether the complex terrain wake interaction was modeled satisfactory with the current approach.

ACKNOWLEDGEMENT

A.M. would like to acknowledge the Greek State Scholarship Foundation (I.K.Y.) for the financial support.

REFERENCES

Smith, G., Schlez, W., Liddell, A., Neubert, A., and Pea, A. (2006). "Advanced wake model for very closely spaced turbines." EWEC 2006, Athens. Crespo, A., Hernandez, J., and Frandsen, S. (1999). "Survey of modelling methods for wind turbine wakes and wind farms." Wind Energy, 2(1), 1-24. Vermeer, L. J., Srensen, J. N., and Crespo, A. (2003). "Wind turbine wake aerodynamics." Progress in aerospace sciences, 39(6-7), 467-510. Duckworth, A., and Barthelmie, R. J. (2008). "Investigation and Validation of Wind Turbine Wake Models." Wind Engineering, 32(5), 459-475. Ainslie, J. F. (1988). "Calculating the flowfield in the wake of wind turbines." Journal of Wind Engineering and Industrial Aerodynamics, 27(213-224). Lange, B., Waldl, H. P., Guerrero, A. G., Heinemann, D., and Barthelmie, R. J. (2003). "Modelling of offshore wind turbine wakes with the wind farm program FLaP." Wind Energy, 6(1), 87-104. Crespo, A., Manuel, F., Moreno, D., Fraga, E., and Hernandez, J. (1985). "Numerical Analysis of wind turbine wakes." Proceedings of Delphi Workshop on Wind turbine applications. Hernandez, J., and Crespo, A. (1990). "Wind turbine wakes in the atmospheric surface layer'." PHOENICS J. Comput. Fluid Dyn, 3, 330-361. Schepers, J. G. (2003). "ENDOW: Validation and improvement of ECNs wake model." ECN-C-03-034, ECN. Gomez-Elvira, R., Crespo, A., Migoya, E., and Manuel, F. (2003). "An explicit algebraic turbulent model to reproduce the anisotropy of the momentum turbulent flows in a wind turbine wake." EWEC 2003. Gmez-Elvira, R., Crespo, A., Migoya, E., Manuel, F., and Hernndez, J. (2005). "Anisotropy of turbulence in wind turbine wakes." Journal of Wind Engineering & Industrial Aerodynamics, 93(10), 797-814. Ammara, I., Leclerc, C., and Masson, C. (2002). "A Viscous Three-Dimensional Differential/Actuator-Disk Method for the Aerodynamic Analysis of Wind Farms." Journal of Solar Energy Engineering, 124, 345. Srensen, J. N., Shen, W. Z., and Munduate, X. (1999). "Analysis of wake states by a full-field actuator disc model." Wind Energy, 1(2), 73-88. Mikkelsen, R., Sorensen, J. N., and Shen, W. Z. (2001). "Modelling and analysis of the flow field around a coned rotor." Wind Energy, 4(3). Srensen, J. N., and Shen, W. Z. (2002). "Numerical modeling of wind turbine wakes." Journal of fluids engineering, 124(2), 393-399. Mikkelsen, R., Srensen, J. N., ye, S., and Troldborg, N. (2007). "Analysis of Power Enhancement for a Row of Wind Turbines Using the Actuator Line Technique." The Science of Making Torque from Wind, 012044. Troldborg, N., Srensen, J. N., and Mikkelsen, R. (2007). "Actuator line simulation of wake of wind turbine operating in turbulent inflow." The Science of Making Torque from Wind, 012063.

Dobrev, I., Massouh, F., and Rapin, M. (2007). "Actuator surface hybrid model." The Science of Making Torque from Wind, 012019. Crespo, A., Manuel, F., Grau, J. C., and Hernandez, J. (1993). "Modelization of wind farms in complex terrain. Application to the Monteahumada wind farm." 1993 European Community Wind Energy Conference, Lbeck-Travemnde, Germany, 440-443. Taylor, G. J., and Smith, D. (1991). "Wake measurements over complex terrain." Proceedings of the 13th BWEA Wind Energy Conference, Swansea, UK, 33542. Stefanatos, N. C., Morfiadakis, E. E., and Glinou, G. L. (1996). "Wake measurements in complex terrain." EWEC 1996, Gteborg, Sweden, 7737. Helmis, C. G., Papadopoulos, K. H., Asimakopoulos, D. N., Papageorgas, P. G., and Soilemes, A. T. (1995). "An experimental study of the near-wake structure of a wind turbine operating over complex terrain." Solar Energy, 54(6), 413-428. Chaviaropoulos, P. K., and Douvikas, D. I. (1999). "Mean wind field prediction over complex terrain in the presence of wind turbines." Proceedings of the 1999 European Union Wind Energy Conference, Nice, France, 120811. Prospathopoulos, J. M., Politis, E. S., and Chaviaropoulos, P. K. (2008). "Modelling Wind Turbine Wakes in Complex Terrain." EWEC 2008, Brussels. Barthelmie, R. J., Frandsen, S. T., Rathmann, O., Hansen, K., Politis, E., Prospathopoulos, J., Cabezn, D., Rados, K., van der Pijl, S., and Schepers, G. (2008). "Flow and wakes in large wind farms in complex terrain and offshore." EWEC 2008, Brussels, Belgium, 10. Ruith, M. R., Inc, F., Shashidhar, S., Vishak, T., Kumar, R., and India, F. (2004). "A Powerful Wind of Change." Fluent News, S4-S5. Jonkman, J. M. (2007). "Dynamics Modeling and Loads Analysis of an Offshore Floating Wind Turbine." National Renewable Energy Laboratory. Versteeg, H. K., and Malalasekera, W. (2007). An introduction to computational fluid dynamics: the finite volume method, Prentice Hall. Panofsky, H. A., and Dutton, J. A. (1984). Atmospheric turbulence, New York. ESDU. (1985). "Characteristics of atmospheric turbulence near the ground." Part II, Single point data for strong winds (neutral atmosphere), ESDU Item, No. 85020.

S-ar putea să vă placă și