Anda di halaman 1dari 8

J. Wind Eng. Ind. Aerodyn.

145 (2015) 115–122

Contents lists available at ScienceDirect

Journal of Wind Engineering


and Industrial Aerodynamics
journal homepage: www.elsevier.com/locate/jweia

Comparison of BEM and CFD results for MEXICO rotor aerodynamics


Borja Plaza a, Rafael Bardera b, Sergio Visiedo c,n
a
Computational Fluid Dynamics Branch, INTA, ISDEFE consultant, Torrejón de Ardoz, 28850, Spain
b
Experimental Aerodynamics Branch, INTA, Torrejón de Ardoz, 28850, Spain
c
Computational Fluid Dynamics Branch, INTA, ISDEFE subcontracted, Torrejón de Ardoz, 28850, Spain

art ic l e i nf o a b s t r a c t

Article history: This work aims to conduct an aerodynamic analysis of the MEXICO wind turbine rotor, establishing
Received 3 October 2014 comparisons between the results of two radically different computational techniques and measure-
Received in revised form ments, obtained in wind tunnel tests under controlled conditions in the framework of the MEXICO
12 May 2015
project (Boorsma and Schepers, 2009; Schepers et al., 2012). Forces, pressures, and torque generated in
Accepted 17 May 2015
the rotor are computed usingthe blade element momentum (BEM) and computational fluid dynamics
Available online 7 July 2015
(CFD) codes to investigate their main strengths and weaknesses at different wind speeds, highlighting
Keywords: the quality of the results obtained at different blade stations. In general terms, both methods tend to
MEXICO overestimate the spanwise distribution of forces and pressures. Reynolds-averaged Navier–Stokes
Mexnext
(RANS)-CFD simulations maintain a uniform level of accuracy across the studied velocity range, whereas
Wind Turbines
BEM calculations outperform CFD estimates at low wind velocities but fail at higher velocities due to
Aerodynamics
CFD separated flow conditions. Blade tip loss and three-dimensional (3D) effects are partly responsible for
BEM inaccuracies in calculation, especially for the BEM code. These 3D effects are discussed briefly in relation
3D effects to force estimates.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction massive volume of data generated in these projects, with the aim of
improving their algorithms (Schepers et al., 2012; Boorsma et al.,
In the wind turbine industry,efficiency in costs and calculation 2014) or even developing basic theories (Schepers et al., 2012). Most
time at the design stages of their machines is imperative. The of the advancements in the codes have been directly related to the
simpler codes are based on the blade element momentum (BEM) accuracy of three-dimensional (3D)effect prediction as the predic-
theory (Burton et al., 2011), and their accuracy and low computa- tion becomes poorer at higher wind speeds due to rotational
tion time make them highly preferred by many companies that augmentation, especially at the inboard sections. Unfortunately,
design wind turbines. On the other hand, computational fluid these phenomena remain highly debated (Guntur and Sørensen,
dynamics (CFD) codes, governed by the Navier–Stokes equations 2014; Yelmule and EswaraRaoAnjuri, 2013), andonly few semi-
(Anderson, 1995), are usually intended for research purposes due empirical corrections are implemented in the two-dimensional
to their high-performance computing (HPC) requirements. How- (2D) polar curves of airfoils for BEM codes, as performed by Du-
ever, in recent years, the rapid development of computational Selig (Du and Selig, 1998).
hardware has resulted in wind turbine designers using CFD codes This work compares the accuracy of the two abovementioned
increasingly more, at the same level of more efficient and profit- methods in performing aerodynamic calculations on the MEXICO
able codes as BEMs. wind turbine rotor. This is carried out by contrasting numerical
From the first international project in this field, the UAE Phase VI results, given by the BEM and CFD codes, with the data measured
(Leishman, 2002; Simms et al., 2001), to the more recent MEXICO in the wind tunnel to establish some clarifications trying to
project, several fruitful studies have been carried out, causing contribute to the choice of the suitable technique from a technical
significant advances in the understanding of wind turbine aero- point of view.
dynamics. The most important developers of aerodynamic codes, The paper discusses the results obtained from the two
based on the lifting line theory, BEM, or CFD, continue to use the methods and their adequacy at different incoming wind speeds,
thus attempting to explain inaccuracies in calculation. For the
n
CFD technique, a previous study obtained good results using
Corresponding author at: Department of Aerodynamics, National Institute for
Aerospace Technology, Carretera de Ajalvir km 4. 28850, Torrejón de Ardoz. Madrid,
multi-block structured meshes and unsteady simulations
Spain. Tel.: +34 91 520 1165. (Schepers et al., 2012), but it also demonstrated that an
E-mail address: visiedoms.pers_externo@inta.es (S. Visiedo). unstructured mesh combined with the moving reference frame

http://dx.doi.org/10.1016/j.jweia.2015.05.005
0167-6105/& 2015 Elsevier Ltd. All rights reserved.
116 B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122

method (steady simulation) gives results close to experimental


values, thus also reducing the preprocessing and calculation
times significantly.

2. Methods

The MEXICO experimental wind turbine is an instrumented


three bladed machine, with a horizontal axis and a 4.5-m rotor
diameter. It was tested at the German-Dutch DNW wind tunnel,
where large volumes of experimental data were recorded for a
great variety of operating conditions, but this work only investi-
gates three cases at different wind speeds. These cases represent
the turbulent wake state (U 1 ¼ 10 m=sÞ, design conditions
(U 1 ¼ 15 m=sÞ, and separated flow conditions (U 1 ¼ 24 m=sÞ, and
they demonstrate the effect of rotor loading on the calculations.
The remaining operating conditions for the simulations are set as Fig. 2. Twist and chord of computed 26 blade elements.

follows: axial flow (noyaw), a pitch angle of  2.31, and a rotation


speed of 424.5 rpm. All wind tunnel conditions (wind speed, 2.2. CFD CODE
turbulence intensities, etc.), different measurements (pressures,
forces, etc.), and instrumentation (PIV, pressure sensors, etc.) used CFD simulations were performed usingthe commercial software
in these tests are described in the section “Description of Experi- Fluent 14.5 (ANSYS), with an HPC cluster to carry out the calcula-
mental Setup” (Boorsma and Schepers 2009). tions with the steady Reynolds-averaged Navier–Stokes (RANS)
The calculations were carried out at the National Institute for approach. The grid was modeled using the meshing ANSYS software,
Aerospace Technology of Spain (INTA), with the following variables ICEM. It was defined as an unstructured mesh because the number
studied: normal force, tangential force, axial force, and torque. The of elements is not high enough to compute time savings.
pressure coefficient (Cp) distribution is analyzed using few airfoils of The 3D-modeled hub, tower, and nacelle are simplifications of
the blade to support the explanations for the forceresults; the 3D the real tested model. A faithful reproduction of the blade
effects are exposed with several images of streamlines on the geometry was achieved using a fine mesh. The fluid domain was
suction side of the MEXICO blade, extracted from CFD simulations. divided into two regions. The first represents the fluid inside the
The first step before starting the simulations is the transforma- wind tunnel, and it extends 5 R from the rotor plane upwind and
tion of the physical problem into a mathematical one, configuring 25 R downwind with a width and height of 20 R, as shown in Fig. 3
a calculation domain with appropriate initial and boundary con- (a), where R is the radius of the rotor.
ditions: the preprocessing stage. The second region is a cylinder of 2.75 R diameter and 0.5 R
height, which contains all the moving parts, blades, and hub(Fig. 3
2.1. BEM CODE (b)). A moving reference frame is applied to this domain,which
rotates at a design wind turbine speed of 424.5 rpm.
For BEM calculations, the FAST code (developed by NREL) has A prism layer is created covering all wind turbine surfaces to
been chosen, in particular, the aerodynamic module AeroDyn, as it improve the calculation accuracy in the boundary layer. The turbu-
has been validated for many different conditions and it is widely lence model k-ω shear stress transport (SST) is used, necessitating a
used for wind turbine certifications.This code requires the blades factor of y þ o1 after the computation. To ensure this, the thickness
to be discretized into smaller elements with the same aerody- of the first layer should be 5  10  6 m. The successive layers grow
namic and geometric properties. Fig. 1 shows the seven different progressively, as shown in Fig. 4, up to a total of 20 layers. The final
parts that compose the swept area of the multi-airfoil MEXICO result is an unstructured tetra-prism layer mesh of about 12 million
blade. The blade is split into 26 elements, the twist and chord elements to run the steady-RANS simulations.
values of which are presented in Fig. 2. A higher density of The boundaryconditions are set as follows: “velocity inlet”
elements is used near the blade tip for precision as the BEM (blue surface in Fig. 3) for the airflow entrance, “Pressure outlet”
calculation in this region is less accurate due to blade tip losses. (red surface), “Symmetry“ for the rest of the tunnel faces (gray
The aerodynamic coefficients C l , C d ,and C m of the airfoils DU91- surfaces), and “wall” for the surface of the wind turbine model.
W2-250, RISØ-A1-21, and NACA 64-418, respectively, used in BEM The cylindrical surface between the rotational and static domain is
simulations for this work were taken from the Mexnext project. set as “interface” for a correct computation.
These coefficients are results of wind tunnel tests for each airfoil at The computed CFD simulations employ a steady pressure-
the appropriate Reynolds number, as explained in the experimen- based model, in addition to the Semi-Implicit Method for
tal description (Boorsma and Schepers, 2009). For transition air- Pressure-Linked Equation algorithm (SIMPLE) to solve the RANS
foils, the aerodynamic coefficients were extracted from the equations (Jasak, 1996).
average of polar curves of the adjacent airfoils.

3. Results and Discussion

The results obtained by BEM and RANS-CFD computation with


the simulations described in the previous section are presented in
this section. Forces, torque, and pressure are compared with the
MEXICO wind tunnel data, and they are evaluated using a
Fig. 1. MEXICO blade airfoil configuration. comparative analysis. Finally, the 3D effects are discussed briefly.
B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122 117

Fig. 3. MEXICO wind turbine grid.

Fig. 4. Prism layer and tetra-mesh.

the underprediction at the inner part of the blade and the over-
prediction near the tip are evident at an incoming velocity of
24 m/s. This limitationarises from the rotational augmentation
produced by 3D effects, which cannot be quantified directly with
any BEM code owing to their two-dimensional nature. 2D polar
curves (aerodynamics data input for BEM calculations) are com-
monly modified to improve the accuracy of the calculations. To
highlight the lack of accuracy of BEM calculations at high wind
speeds, no changes were made to the experimental polar curves in
the present study. However, in other calculations with different
BEM codes used by Mexnext project partners, no substantial
improvement was observed when 3D effects were taken into
consideration.
At a lower wind speed of 10 m/s, BEM calculations were closer
to the experimental results than the CFD estimates. The high load
on the rotor under this flow condition results in higher calcula-
Fig. 5. Normal and tangential forces. tions of pressures at blades than real measurements, which
increasesuncertainty (Daniel Micallef, 2009) and leads to over-
estimation of the velocity in the boundary layer at the suction side
3.1. FORCES of the airfoil.
In spite of the general tendency of the CFD code to overpredict
The spanwise distributions of normal force (Fn) and tangential normal loads, its results show good agreement with experimental
force (Ft) per unit length determine the blade’s aerodynamic and data at high wind speeds, as can be seen in Fig. 6(c). At this
structural performance. Both are represented in Fig. 5. incoming flow velocity, the calculations exhibit result oscillations
The normal force distribution is shown in Fig. 6 at three at the outer sections of the blades, likely due to the use of an
different wind speeds. At wind speeds of 10 and 15 m/s, there is unstructured mesh under stalled flow conditions.
a good agreement between calculations and measurements, espe- It is important to note the difficulties encountered for both codes
cially for BEM code results. in the accurate prediction of the normal force at the blade tip region,
Fig. 6 shows, in general,good agreement between the experi- where a certain level of overprediction was observed for the three
mental data and the calculations from both codes. However, a cases. The high values in the BEM results indicate that the Prandtl
closer look comparing Figs. 5 and 6 reveals that accuracy of the tip loss correction used in FAST is too optimistic, and the tip loss
BEM calculation decreases with increase in wind speed, because factor should be lower at the inner sections near the tip region. The
118 B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122

Fig. 6. Normal force distribution.

disagreement could also be due to an erroneous calculation of the 3.2. GLOBAL VARIABLES
induction factors altering the local angle of attack, thus resulting in a
higher force than the experimental value. The studied global variables, torque and axial force (Fig. 8), are
Fig. 7 represents the tangential force distribution. In general, it mainly generated by the tangential and normal force, respectively.
presents a similar trend to the distribution of normal force, but Their importance lies in their contribution to define the general
some emerging differences must be pointed out. wind turbine performance, as power coefficient or tower struc-
The results from the BEM code approximate experimental tural requirements. Fig. 9 represents the comparison between
measurements to some extent at wind speeds of 10 m/s and calculation results and experimental data of axial force and torque
15 m/s, but the calculation curves are completely incongruent as a function of wind speed.
with measured data for massive flow detachment at 24 m/s. Again, As discussed previously, the general tendency to overpredict
this shows that BEM simulations are inaccurate at high wind normal loads with increase in wind speed is reflected in the axial
speeds as can be seen at the inner sections of the blade, where the force estimates of Fig. 9(a). It is interesting to note that some CFD
estimates decay drastically. Moreover, CFD simulations tend to calculations of axial induction velocity through the rotor plane
overestimate the tangential loads at the blade tip region for every (not seen in this article) produced higher values than experimental
wind speed becausecalculations are highly sensitive in the tan- data, leading to a pressure jump lower than real. This is incon-
gential direction, being fundamentally governed by pressure gruent with the actuator disc theory because axial force computed
gradients in the leading and trailing edges, precisely where the by the CFD code should be lower than the experimental value;
major differences in measurements exist. The tangential force however,this is not the case.
computed with different codes by Mexnext partnersis highly BEM code calculations also overestimate the axial force, but it is
dispersed, as can be seen in Ref. Schepers et al., (2012). This interesting to note that the calculation and the experimental value
highlights the difficulty of computing the tangential loading, are very close at 24 m/s. From Fig. 6(c), it can be deduced that this
considering that only pressure data were used to obtain the force coincidence is merely due to the compensation of the under- and
and disregarding the friction terms (Schepers et al., 2012) that overestimation in the calculation of the normal force at the inner
would have produced an even higher level of dispersion. and outer parts of the blade, respectively.
B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122 119

Fig. 7. Tangential force distribution.

appear above the experimental values, so that the tangential force


is highly overestimated at the blade tip, which indicates the
importance of the accuracy of the calculation in this region.
Simulations with FAST provide good estimations of the torque
at 10 and 15 m/s, in agreement with the tangential force calcula-
tions; however, the BEM code fails again at the highest wind
speed, and the torque results remain at about 10% below the
experimental data.

3.3. Pressure Distribution

The pressure distribution along the blade can explain the


results in forces. For the three wind speed cases, the pressure
coefficient (C P ) is studied over two airfoils, at 35% and 92% of span,
representing the inner and outer sections on the blade. This
coefficient is defined as follows:
Fig. 8. Axial force and torque. p  pref
CP ¼
q1
Fig. 9(b) shows the results of torque calculations using the BEM wherep is the absolute pressure over the airfoil surface, pref the
and CFD codes compared with the experimental data as a function corresponding static pressure, and q1 is the dynamic pressure of
of wind speed. As already mentioned, torque estimates must be the steady flow in the wind tunnel given as q1 ¼ ð1=2ÞρU 1 2 .
consistent with the spanwise tangential force distribution,which Fig. 10 presents the distribution of the pressure coefficient at
gives rise to the rolling moment. The curves of the CFD results the two sections along the non-dimensionalized chord “x/c,”
120 B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122

Fig. 9. Axial force and Torque at different wind speeds.

where x is the local chord-wise coordinate. At 35% of span, the may occur at higher angles of attack than the equivalent 2D curve
pressure over the airfoil is, by far, lower than that on the outer from the static wind tunnel test at the same Reynolds number. It is
section as can be observed, and it is also lower than the calculation known that adverse gradients of pressure suffered by the airflow
error. The calculation is observed to be less accurate on the suction over the suction side of the blade slow down the flow near the
side than on the pressure side, and the differences in experimental trailing edge, as in the boundary layer. Under the combination of
measurements increase with the wind velocity in the blade tip both effects, the flow velocity may decay to zero or even reverse
direction. The overprediction in pressure calculation could be a flow can occur. If this inversion of flow direction occurs, the
specific limitation of the RANS simulation, as URANS offers a separation takes place and the airfoil is in stall conditions. The
slightly better solution as demonstrated in Ref. Pascal (2009). This adverse pressure gradient is reduced for unknown reasons by the
limitation could also arise from the refinement of the unstructured blade rotation; however, the following explanation is commonly
mesh used, and a structured mesh could give better results. accepted: the rotation produces an increase of dynamic pressure in
For the highest wind speed case, with stall conditions at the the spanwise direction,which in turn adds to the centrifugal force
blades, the prediction of the separation point location is unlike the experienced by air particles, resulting in a radial induction to the
measurements shown. As can be observed in Fig. 10(c), the sudden flow. This radial component of velocity generates the appearance
pressure drop computed on the suction side, which indicates the of the Coriolis force, which redirects the flow towards the trailing
detachment of the flow, occurs behind measurements. This high- edge inducing a positive pressure gradient that causes reattach-
lights the weakness of the RANS model for massive flow separa- ment of the flow and delays the separation (Lee and Chua, 2012;
tions,because the turbulent model used (k-ω SST) is too dissipative Tony Burton et al., 2011).
to capture many length scales and the complex turbulent struc- The streamline pattern on the suction side of the blade at
tures generated in detachments. Some investigations with DES1 different wind speeds explains the behavior of the near-wall flow.
simulations (Yuwei Li et al., 2012; Meng and van Rooij, 2007) give When fluid particles start to experience the radial component of
better results as many of these turbulent structures are resolved velocity, the rotational augmentation begins. Fig. 11 shows the
and therefore the separation point is better predicted, although streamlines (oil flow) at 10, 15 and 24 m/s, where vortex regions
the required computational resources increase significantly. and reverse-flow zones, which determine the flow separation, as
The general overprediction in the pressure calculation at the deduced by Guntur et al. Guntur and Sørensen (2014) and Mukesh
suction side generates a similar trend in force to that seen (Yelmule and EswaraRaoAnjuri, 2013), are seen.
previously. The difference between the measured and computed At 10 m/s, the flow is attached to the whole blade, as shown by
pressures in the outer section is more evident,corresponding with Fig. 11 (a); however, at the inner region near the hub, a radial
the overestimation of forces at the blade tip region. For the 92% component can be seen probably due to vortex shedding in this
spanwise station, the location of the peaks of pressure at the zone. When U1 ¼15 m/s,a strong radial component appears near
suction side near the leading edge is well predicted; however, its the trailing edge of the blade, which covers the length from the
magnitude is overpredicted, leading to the high values in tangen- hub to the 45% station, where the flow starts to reverse and hence
tial forces seen in Fig. 7. detach. The line of flow separation can be observed, which is the
confluence of reverse streamlines. It is interesting to observe the
mid-span vortex near the 60% section (rE 1.35 m) in Fig. 11 (b).
3.4. 3D EFFECTS The cause of this vortex is unknown, although it may be inducted
by the abrupt transition between airfoils RISØ-A1-21 and NACA
The origin of the physical phenomenon to explain the stall 64-418, very close to this section. This vortex is responsible for the
delay remains a matter of debate. When an airfoil rotates, the stall distortion in the force curves in this region, as seen in Fig. 6 and
Fig. 7 at a wind speed of 15 m/s. At a wind speedof 24 m/s (Fig. 11
1
DES: Detached Eddy Simulations, is a combined method which solves the
(c)), massive reverse flow appears from the trailing to leading
near wall region by mean of RANS models, and uses LES (Large Eddy Simulations) edges on the suction side of the whole blade, causing the
calculations when the turbulent length scale exceeds the grid dimension. boundary layer to separate by about 30% chord-wise.
B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122 121

Fig. 10. Pressure distribution at sections 0.35 R and 0.92 R.

Fig. 11. Suction side streamlines on MEXICO blade at different wind speed. (a) U∞ = 10m/s (b) U∞ = 15m/s (c) U∞ = 24m/s.

The 3D effects are very important, particularly for inner the discrepancies between both codes, as the CFD technique can
sections, and its calculation should be taken into consideration predict these effects,providing good results at this region of
to improve the accuracy of BEM codes. This is the main reason for the blade.
122 B. Plaza et al. / J. Wind Eng. Ind. Aerodyn. 145 (2015) 115–122

4. Conclusions and Future Work K. Boorsma, J.G. Schepers, Sugoi Gomez-Iradi, Helge Aagaard Madsen, NielsSøren-
sen,Wen Zhong Shen, and Christoph Schulz, Mexnext-II: The Latest Results on
Experimental Wind Turbine Aerodynamics, EWEA 2014 Annual Event, Barce-
Different results from the BEM and CFD calculations with the lona, Spain, March 2014.
MEXICO experimental wind turbine were explored and discussed Burton, Tony, Jenkins, Nick, Sharpe, David, Bossanyi, Ervin, 2011. Wind Energy
in comparison with experimental measurements. Handbook. John Wiley & Sons, Ltd., West Sussex, PO19 8SQ, United Kingdom.
This work confirms that BEM calculations achieve some great Z. Du, M. Selig, “A 3-D Stall-Delay Model for Horizontal Axis Wind Turbine
Performance Prediction.” 1998 ASME, Wind Energy Symposium. AIAA-1998-
results, surprisingly better than RANS-CFD at low and medium 21, January 1998.
wind speeds. However,the BEM technique is not suitable when the Guntur, Srinivas, Sørensen, Niels N., 2014. A study on rotational augmentation
incoming wind velocity is high enough to produce detachment at using CFD analysis of flow in the inboard region of the MEXICO rotor blades.
Wiley Online Library.
the blades. Rotational 3D effects are to be taken into consideration
Jasak, H., 1996. Error analysis and estimation for the finite volume method with
in these simulations, especially at high wind speed, because the applications to fluid flows PhD thesis. Imperial College of Science, Technology
accuracy of the calculations decays considerably with velocity. On and Medicine.
the other hand, RANS-CFD simulations produce good results over Hsiao Mun Lee and LeokPoh Chua, Investigation of the stall delay of a 5 kW horizontal
axis wind turbine using numerical method, International Conference on Renew-
the whole range of wind velocity, although theBEM calculations able Energies and Power Quality (ICREPQ 12), Santiago de Compostela, Spain,
are slightly better for turbulent wake and design conditions. March 2012.
Overestimates near the blade tip region highlights the difficulties Leishman, J. Gordon, 2002. Challenges in Modeling the Unsteady Aerodynamics of
Wind Turbines PAPER 2002-0037. AMERICAN INSTITUTE OF AERONAUTICS
in calculation for both codes as a result of the strong loss
AND ASTRONAUTICS, Reston, VA 20191 4344, USA.
produced. F. Meng and R. P.J.O.M. van Rooij, CFD investigations with respect to model sensitivity
For this work, the preprocessing and calculation timesof the for the non-rotating flow around the NREL Phase VI Blade, Proceeding of the
two techniques are vastly different. By far, BEM simulations European Wind Energy Conference EWEC, Milan, Italy, 2007.
L. Pascal, Analysis of MEXICO measurements, Energy research Centre of the Nether-
tookless time than RANS-CFD. This time efficiency makes the lands (ECN), 2009.
BEM method the tool of choice for large-scale use in the wind Daniel Micallef, 2009. MEXICO Data Analysis, Stage III - Cp Distributions Using the
turbine industry; however, to improve the designs and the under- Stagnation Point Method and Sensitivity Analysis. Energy research Centre of
standing of rotor flow physics, CFD techniques are required the Netherlands (ECN).
Schepers, J.G., Boorsma, K., Cho, T., Gomez-Iradi, S., Schaffarczyk, P., Jeromin, A.,
because of the vast amount of information they provide for all Shen, W.Z., 2012. Review of computational fluid dynamics for wind turbine
the flow variables. wake aerodynamics. Energy research Centre of the Netherlands (ECN).
Future work for BEM codes should be focused on improving the Schepers, J.G., Boorsma, K., Chon, T., 2012. NielsSørensen,Final report of IEA Task 29,
Mexnext (Phase1): Analysis of Mexico wind tunnel measurements, ECN-E-12-004.
accuracy of results at high wind speeds, studying rotational 3D
Energy research Centre of the Netherlands (ECN).
effects and developing new models to compute them, as the Simms, D., Schreck, S., Hand, M., Fingersh, L.J., 2001. NREL Unsteady Aerodynamics
calculation of separated flow is unaffordable by this technique. Experiment in the NASA-Ames Wind Tunnel: A Comparison of Predictions to
Tip loss correction models should be reviewed to estimate forces Measurements. National Renewable Energy Laboratory, Golden, Colorado
80401-3393, USA.
at the blade tip region with more accuracy, because the Prandtl Tony Burton, Nick Jenkins, Sharpe, David, Bossanyi, Ervin, 2011. Wind Energy
correction is generally optimistic. With the improvement of Handbook. John Wiley & Sons, Ltd., West Sussex, PO19 8SQ, United Kingdom.
computational performance, the research community must seize Yelmule, Mukesh M., EswaraRaoAnjuri, VSJ, 2013. CFD predictions of NREL Phase VI
Rotor Experiments in NASA/AMESWind tunnel. INTERNATIONAL JOURNAL of
the opportunity to perform CFD simulations with more complex
RENEWABLE ENERGY RESEARCH.
models such as DES or LES to gain a deep insight into some Yuwei Li, Kwang-Jun Paik, Xing, Tao, Carrica, Pablo M., 2012. Dynamic overset CFD
unknown physical mechanisms,thus allowing the development of simulations of wind turbine aerodynamics. Renewable Energy 37, 285–298.
new, faster models for wind turbine calculation.

References

Anderson, John, 1995. Computational Fluid Dynamics: The Basics With Applica-
tions. McGraw-Hill Education.
Boorsma, K., Schepers, J.G., 2009. Description of experimental setup MEXICO
measurements, ECN-X-11-120. Energy research Centre of the Netherlands (ECN).

Anda mungkin juga menyukai