World Academy of Science, Engineering and Technology 60 2009

CFD Analysis on Aerodynamic Design

Optimization of Wind Turbine Rotor Blades
R.S. Amano, R.J. Malloy

turbine can be found from its kinetic energy and the power is
Abstract—Wind energy has been shown to be one of the most expressed as follows [2]
viable sources of renewable energy. With current technology, the low
cost of wind energy is competitive with more conventional sources of 1 1
energy such as coal. Most blades available for commercial grade Pavail = m V02 = ρAV03 (1)
wind turbines incorporate a straight span-wise profile and airfoil
2 2
shaped cross sections. These blades are found to be very efficient at
lower wind speeds in comparison to the potential energy that can be If momentum equation is solved across an idealized control
extracted. However as the oncoming wind speed increases the volume about the turbine rotor it can be shown that the
efficiency of the blades decreases as they approach a stall point. This percentage of the total power available that can be extracted
paper explores the possibility of increasing the efficiency of the by a turbine is 16/27 or .59%. This limit is known as the Betz
blades at higher wind speeds while maintaining efficiency at the limit. Therefore the maximum power that a turbine can
lower wind speeds. The design intends to maintain efficiency at produce is expressed as follows
lower wind speeds by selecting the appropriate orientation and size
of the airfoil cross sections based on a low oncoming wind speed and
⎛ 16 ⎞⎛ 1 ⎞ (2)
given constant rotation rate. The blades will be made more efficient Pmax = ⎜ ⎟⎜ ⎟ ρV03 A
at higher wind speeds by implementing a swept blade profile. ⎝ 27 ⎠⎝ 2 ⎠
Performance was investigated using the computational fluid
dynamics (CFD). Most turbines extract the maximum possible energy as
defined above for lower wind speeds but gradually become
Keywords—CFD, wind turbine blade, renewable energy. less efficient as the on coming wind speed increases and the
flow condition across the blades approach the stall condition.
I. INTRODUCTION To illustrate this effect the power produced by a typical

W IND turbine has been getting its attention as one of the

most viable alternative power media. This paper
explores two methods for optimizing blade for
commercial turbine is compared to the amount of power
available to it at various oncoming wind speeds in Fig 2. The
turbine that is used is the Nordtank 500/41 with LM19.1
operation where wind speeds average 7m/s. The two methods blades [1].
of optimization that are investigated are first, the straight edge
blade optimizes the angle of attack and chord length for a
given airfoil cross section at different positions along the
blade and second implements a swept blade profile. Among
the factors outside the scope of the paper, technological
advances in this field have allowed the price of wind energy to
be competitive with more conventional sources. The blades
are an important component of wind turbines and much
research has been done in this area. Most commercial blades
designs incorporate a straight edge span-wise profile with
airfoil cross sections of various sizes and orientations. These
designs have shown to be efficient in comparison to the
amount of energy that can be extracted. The amount of
energy in a column of wind seen by the swept area of a

Fig. 1 NTK500/41 [1]

R.S. Amano is with the University of Wisconsin at Milwaukee, WI 53201 As can be seen from Fig. 2 the power curve of the Nordtank
USA (corresponding author: 414-229-2345; fax: 414-229-6958; e-mail: NTK4500/41 wind turbine follows the maximum power curve
amano@ uwm.edu).
R.J. Malloy is a graduate research assistant at the University of Wisconsin
up until about 10m/s oncoming wind speed. At this point the
at Milwaukee, WI 53201 USA (corresponding author: 414-229-2345; fax: blades begin to stall and the turbine’s performance levels off
414-229-6958; e-mail: rjmalloy@ uwm.edu). at around 600kW. This paper explores the possibility of

71
 World Academy of Science, Engineering and Technology 60 2009

increasing the efficiency of the blades at higher wind speeds

while maintaining efficiency at the lower wind speeds. The Since the direction and magnitude of the wind generated by
design intends to maintain efficiency at lower wind speeds by rotation changes as a function of span-wise position, the chord
selecting the appropriate orientation and size of the airfoil length and angle of attack of the airfoil cross sections must
cross sections based on a low oncoming wind speed and given change as well.
constant rotation rate. The blades will be made more efficient The angle of twist is easily found at any position by
at higher wind speeds by implementing a swept blade profile. adding the oncoming and rotational wind speed vectors. The
The power generated from a blade using only the first chord length can be found by relating the thrust force to the
optimization technique is compared to that generated from a lift force produced by a cross section. Thrust force is defined
blade using both techniques as well as that generated by as the force experienced by the rotor in the stream-wise
NTK500/41 turbine using LM19.1 blades. direction and is expressed as follows

⎛4⎞ (3)
Performance of NTK500/41 with LM19.1 Blades T = ⎜ ⎟ ρV02 A
⎝9⎠
800
700 The lift force is the force experienced perpendicular to the
600
Maximum Possible
resultant wind vector with respect to airfoil and is expressed
Power(kW)

500
Energy as follows.
400
NTK500/41
300
1 (4)
200 L= ρVa2 AC l
100 2
0
0 10 20 30 These two forces can be related as follows.
wind speed (m/s)

T = L cos(θ ) (5)
Fig. 2: Comparison of Available Wind Power and Power Output of
NTK500/41 Turbine with LM19.1 Blades [2].
From equations 3 through 5 the following expression of chord
A. Straight Edge Blade length can be derived assuming a constant coefficient of lift
value of 1 and neglecting drag. [2]
Wind turbine blades rely on lift produced by their airfoil
cross sectional shapes to produce the torque at the base needed ⎛R⎞
to turn the generator. Wind turbine blade profiles are often 16πR⎜ ⎟ cos(θ )
C= ⎝r⎠ (6)
constructed using the Blade Element Momentum (BEM)
theory. This theory produces the angle of twist and chord
9λ 2 B
length for a given airfoil cross section and rotation speed at a
The BEM approach does not correct for rotational motion.
finite number of positions along the span of the blade. From
It is for this reason that CFD analysis is necessary for new
these two dimensional sections a three dimensional shape can
blade designs. CFD does not used predetermined airfoil data
be extruded. The BEM theory accomplishes this by treating a
to predict the blade performance. Instead it solves the
given cross section as an independent airfoil, which processes
governing fluid flow equations at thousands of positions on
wind with a speed and direction that is a vector sum of the on
and around the blade in an iterative process. This approach
coming wind speed and the wind speed generated by rotation
allows the current airfoil model to take into account any span-
(see Fig. 3 for the vectorial view).
wise wind velocity component which BEM theory cannot.
It was this method that was used to construct a straight edge
blade prototype whose optimal on coming wind and rotation
speeds were 7m/s and 20rpm, respectively. The blade has a
length of 20m and uses the constant airfoil shape NACA
4412.

B. Swept Edge Blade

In addition to this straight blade, a swept edge blade was

tested. This swept edge blade will have the same
characteristics as the straight edge except for the trajectory of
the edge. Each cross section will have the same dimensions
and be at the same vertical distance from the hub as its
corresponding section in the straight edge blade. The reason
Fig. 3: Dynamics of a Wind Turbine Blade Cross Section [3] for testing this new geometry is that the straight edge blade is

72
 World Academy of Science, Engineering and Technology 60 2009

constructed using a formulaic approach which treats the maximum size of 1.08 and 0.5 growth rates were attached to
airflow over the blade as perfectly perpendicular to the the blade surfaces and extended into the rectangular volume
leading edge and neglects any span-wise component. The surrounding the blade. The wedge containing this rectangle
swept edge blade profile aims to accommodate the span-wise and the rest of the blade/hub was meshed with a constant
velocity component and delay the stall point of the rotor. This density mesh of 0.5m. The rest of the domain was given a
geometry has largely been uninvestigated using the CFD growth rate of 1.08 and maximum size of growth rate 10
approach. Recently an investigation into the loading and extending from this wedge.
dynamic behavior of a swept blade was published by Larwood
and Zureck [4]. They used codes developed by the National
Renewable Energy Laboratories (NREL) which applied a
more analytical approach. A CFD approach is more suitable
for this investigation since it is purely an aerodynamic study
and CFD yields very accurate results which are quantitative as
well as qualitative.

II. NUMERICAL METHOD

A diffuser-shaped domain was chosen with a 120 degree
slice taken lengthwise along the axis. Each side of the domain
was given periodic boundary conditions. The front and top
planes were given as velocity inlets. The rear plane was given
as a pressure outlet. The domain extended 5 diameters
upstream of the blade and 10 diameters downstream of the
blade. The domain had a radial height of 5 diameters at the
front and 8 diameters at the back. Several different mesh Fig. 5: Computational Domain
schemes were used in an effort to both resolve the boundary
layer surrounding the blade and hub, and obtain a For the swept blade the rectangle was modified slightly to
computationally feasible domain. A 4m×20m rectangle was accommodate the swept edge geometry. This approach is
constructed around the blade, and an unstructured mesh was based off of the work done by Mandas and Carcangiu [5]. The
used throughout the domain. Meshing proved to be a final mesh contained 1.7 million elements.
challenge due to the large differences in length scales in the Seven different oncoming wind speeds were used ranging
domain. To achieve an adequate fine mesh that was still from 5m/s to 25m/s. This allowed for the construction of a
computationally feasible many size functions were utilized. power curve for each blade so they could be compared. The
sides of the wedge were designated as rotational periodic
boundaries. Finally the fluid being chosen as a moving
reference frame was given a rotational speed of 2.09rad/s. The
turbulence closure model used was the the κ-ω SST model.
This model was used with success in a similar application by
Ferrer and Munduante [6]. All results were obtained with a
convergence criteria of the order of 10 -5. For validation of the
results the same mesh scheme and domain were used to
reproduce experimental data of NTK500 commercial wind
turbine with LM19.1 blades. The results are shown below.
800

CFD
700

600

500
Power (kW)

400

Fig. 4: Boundary Layer Resolution Approach 300

A start size of 0.007m was used at the leading and trailing 200

edges of the blade along with a growth rate and maximum size 100

of 1.14 and 0.05 extending onto the blade surfaces as shown

in Fig. 4.
0
0 5 10 15 20 25

This approach adequately represented the aerodynamic Wind Speed (m/s)

shape of the blades. A starting size of 0.05 growth rate and Fig. 6: CFD Predicted Performance of Commercial Wind Turbine
NTK500 with LM19.1 Blades Compared to Experimental Data

73
 World Academy of Science, Engineering and Technology 60 2009

blades besides a slight increase in frontal surface area as stated

As can be seen there is good agreement up until about 12m/s. in Table 1.
Then the model brakes down as the blades begin to stall. The
same effect was found by Mandas and C. Carcangiu [5]. This Com parison of Sw ept and Straight Blade v.s Wind Speed
effect is not present with the two blades under investigation.
600

III. PRESENTATION OF RESULTS 500

Figures 7 and 8 show the pressure contours along the top

Power Output (kW)

400
upstream side of the straight and swept blades, respectively. It
is clear from these two figures that the pressure level is higher 300
Straight Blade
Sw ept Blade
on the swept blade, which confirms a higher lift force on the
blade. This will enhances the power generated at the same 200

upcoming stream velocity (wind speed). 100

0
0 5 10 15 20 25 30
Wind Speed (m /s)

Fig. 9: CFD Predicted Performance of the Straight Edge and Swept

edge Wind Turbine Blades

TABLE I SURFACE AREA COMPARISON

Edge Surface
Geometry Area(m2)
Straight 89.3
Swept 90.77

From Figure 7 it should be noted that both geometries

Fig. 7 Pressure Contours at the Top Upstream Side of the Straight generate relatively the same amount of power as conventional
Blade At 20m/s oncoming wind speed large scale turbines such as NTK500/41 using LM 19.1
blades. The simulation for straight edge and swept edge
blades also predicts stall effectively where as the simulation
for the NTK500/41 turbine does not. Therefore the results
shown in Fig 6 cannot completely verify the results shown in
Fig 7. No experimental results of a turbine with the straight
edge blade geometry, used in this investigation, are known to
the author. Future work should incorporate experimental
verification of results shown in Fig 7. A possible explanation
for the difference in simulation performance is that the airfoil
shape used in this investigation is slender compared to airfoil
shapes used in typical blade design such as the NACA 63-418
used on the NTK500/41 turbine. Because of this shape
characteristic the stall characteristics of the NACA 4412
airfoil are much more distinct and therefore can be modeled
more easily. It is for this reason that this airfoil shape was
chosen for this investigation.
Fig. 8 Pressure Contours at the Top Upstream Side of the Swept
Blade at 20m/s Oncoming Wind Speed IV. CONCLUSIONS
The following conclusions emerge from this study.
Figure 9 shows the predicted power curve of two turbines (1) The simulation of the NTK500/41 satisfactorily validated
with the different blade geometries. From the figure it is clear the results of this investigation.
that there is little performance variation at the lower oncoming (2) It was observed that the swept edge geometry maintains
wind speeds. This is desired because at these speeds the wind maximum efficiency at lower oncoming wind speeds and
turbines are extracting the maximum amount of power as delays the stall point resulting in an increase in power at
defined in equation 2. It can also be seen from Fig. 9 that the higher oncoming wind speeds.
swept edge blade geometry produces an increase in power at (3) The gain in the power is as high as 20 percent at higher
higher wind speeds with the maximum being 15% occurring speeds than 10m/s.
at 20m/s. It is clear that this increase is due to the swept edge There are other methods that are used for optimizing blade
geometry only since this was the only difference between the performance that were not investigated here but should be

74
 World Academy of Science, Engineering and Technology 60 2009

mentioned. One of the most obvious strategies is to increase

the diameter of the rotor, because the swept area will contain
more energy that can be abstracted. However this can be
problematic as the loads at the hub/blade interface will
increase dramatically. Research on relieving these loads
through blade twist has been done by Larwood and Zuteck
[4].

ACKNOWLEDGMENT
The authors are grateful to We Energies for their support on
this project. The author is grateful to Mr. Carl Siegrist and Ms.
Amy Flom for their great support to this research. Thanks are
due to Mr. Jeremy Hogan for his dedicated assistance to the
optimization study of wind turbine blades.

REFERENCES
[1] NTK500/41
http://130.226.17.201/extra/web_docs/nordtank/WT_description.pdf
[2] Hasen M. ,2000, “Aerodynamics of Wind Turbines”
[3] Piggott H. “Small Wind Turbine Design Notes”
http://users.aber.ac.uk/iri/WIND/TECH/WPcourse/index.html
[4] Larwood, S. and Zuteck, M., 2006, “Swept Wind Turbine Blade
Aeroelastic Modeling for Loads and Dynamic Behavior”.
[5] Mandas, N., Cambuliand, F., and Carcangiu, C., 2006, “Numerical
Prediction of Horizontal Axis Wind Turbine Flow,” University of
Caglairi, EWEC 2006, Athens, Business, Science, and Technology.
[6] Ferrer, E. and Munduante, W., 2007,. “Wind Turbine Blade Tip
Comparisons Using CFD.” Journal of Physics Conference series 75,
012005

Ryo S. Amano (M’80––F’01). This author became a Member (M) of ASME

in 1980 and a Fellow (F) in 2001.

75