Energy Conversion and Management 99 (2015) 210–220

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Aerodynamic noise prediction of a Horizontal Axis Wind Turbine using

Improved Delayed Detached Eddy Simulation and acoustic analogy
Masoud Ghasemian, Amir Nejat ⇑
School of Mechanical Engineering, University of Tehran, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents the results of the aerodynamic and aero-acoustic prediction of the flow field around
Received 4 January 2015 the National Renewable Energy Laboratory Phase VI wind turbine. The Improved Delayed Detached Eddy
Accepted 6 April 2015 Simulation turbulence model is applied to obtain the instantaneous turbulent flow field. The noise
prediction is carried out using the Ffowcs Williams and Hawkings acoustic analogy. Simulations are
performed for three different inflow conditions, U = 7, 10, 15 m/s. The capability of the Improved
Keywords: Delayed Detached Eddy Simulation turbulence model in massive separation is verified with available
Horizontal Axis Wind Turbine
experimental data for pressure coefficient. The broadband noises of the turbulent boundary layers and
Wind turbine noise
Aerodynamic noise
the tonal noises due to the blade passing frequency are predicted via flow field noise simulation. The
Turbulence contribution of the thickness, loading and quadrupole noises are investigated, separately. The results
indicated that there is a direct relation between the strength of the radiated noise and the wind speed.
Furthermore, the effect of the receiver location on the Overall Sound Pressure Level is investigated.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction mechanism. Therefore, it is important to identify and predict the

An increasing need for energy coupled with global warming has
caused to exploring new alternatives to meet energy requirements
[1]. One of the most promising renewable sources is wind energy
[2]. In particular, there is an interest to develop small wind tur-
bines for urban and suburban applications [3]. However, wind
energy also has several disadvantages that stand in the way of
wind turbine technology becoming popular. One of its major prob-
lems is societal rejection of wind turbines in developed areas due
to acoustic pollution. Aeroacoustic noise from wind turbine may
be a cause of annoyance from people living in the neighborhood
of the turbines, particularly those neighborhoods with the low
ambient noise level [4]. Noise emitted from an operating wind tur-
bine can be divided into mechanical noise and aerodynamic noise.
Mechanical noise originates from different machinery components,
such as the generator and the gearbox. Aerodynamic noise is radi-
ated from the blades and is mainly associated with the interaction
of turbulence with the blade surface [5]. Machinery noise can be
reduced efficiently by well-known engineering methods Such as
vibration suppression, vibration isolation and fault detection tech-
niques [4], whereas reduction of aerodynamic noise still represents
a problem, and aerodynamic noise is the dominating noise
⇑ Corresponding author.
E-mail address: nejat@ut.ac.ir (A. Nejat).
most important noise sources.
Traditionally, noise prediction was performed using empirical
or semi-empirical considerations. In recent years, because of the
computational technology advances, the interest in the computa-
tional aeroacoustics has grown noticeably. The use of
Computational Fluid Dynamics (CFD) for wind turbines has made
it possible to achieve an accurate design tool, but this use has been
limited to date because of the difficulties encountered by tradi-
tional Reynolds-Averaged Navier Stokes (RANS) turbulence models
in accurately predicting the highly unsteady features that are key
in the prediction of the noise of the wind turbine [6]. Large Eddy
Simulation (LES) method has superiority compared with RANS.
Unlike the RANS method, it resolves directly all the large scales,
which contain the most energy. The small scales, or eddies behave
in a universal way, and are simply modeled. However, LES is still
computationally expensive for high Reynolds numbers and indus-
try applications. This issue is tackled by using hybrid RANS-LES
methods. The hybrid methods are a combination between the sta-
tistical RANS and LES methods. The basic principle in these meth-
ods is to model the boundary layer using a RANS turbulence model,
whereas LES is used to resolve the detached eddies and separation
regions in the farfield.
Tadamasa and Zangeneh [7] predicted the noise radiated from
the rotating Horizontal Axis Wind Turbine (HAWT) blades. They
used a RANS approach with Shear Stress Transport (SST) k  x

http://dx.doi.org/10.1016/j.enconman.2015.04.011
0196-8904/Ó 2015 Elsevier Ltd. All rights reserved.
 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220 211

Nomenclature

a0 speed of sound (m/s) sij tensor of stress

Cp pressure coefficient s retarded time (s)
d distance to the nearest wall x rotational speed (rpm)
f frequency (Hz)
H Heaviside function Subscripts
k turbulent kinetic energy ðm2 =s2 Þ L loading noise
lIDDES IDDES length scale n component in surface normal direction
lRANS RANS length scale T thickness noise
M mach number 0 undisturbed condition

p acoustic pressure (Pa)
Pij compressive stress tensor
Abbreviations
r distance to the receiver (m) CFD Computational Fluid Dynamic
Sij strain rate FW-H Ffowcs Williams and Hawkings
t time (s) FFT Fast Fourier Transform
T ij Lighthill stress tensor
IDDES Improved Delayed Detached Eddy Simulation
Uj velocity (m/s) HAWT Horizontal Axis Wind Turbine
Un fluid velocity in the normal direction (m/s) LES Large Eddy Simulation
vn normal velocity of the integration surface (m/s)
NREL National Renewable Energy Laboratory
yþ wall unit OASPL Overall Sound Pressure Level
PISO Pressure Implicit with Splitting of Operator
Greeksymbols RANS Reynolds-Average Navier Stokes
dð f Þ Dirac delta function SPL Sound Pressure Level
l molecular viscosity (kg/ms) URANS Unsteady RANS
lt turbulent viscosity (kg/ms) TKE Turbulent Kinetic Energy
q density (kg/m3) URANS Unsteady RANS

based turbulence model for aerodynamic calculation and Ffowcs
Williams-Hawking (FW-H) equation for aeroacoustics prediction.
They investigated the effects of wind speed and rotational speed
on the total noise radiation. Their results showed that the thickness
noise increases continuously with increment in the rotational
speed. Cho et al. [8] measured acoustic noise for a scaled model
of NREL Phase VI wind turbine. They used a microphone array to
identify the noise source position of the blade. Their results
showed that the main acoustic noise source position moves toward
the blade tip as the frequency increases and the noise level at low
frequency below 2 kHz has much higher when the blade is in a stall
condition. Chourpouliadis et al. [9] carried out a comparative study
of the noise emissions from two interconnected wind farms. Their
results showed that the predicted noise signals prove to be within
the limits of recent regulations concerning the installation of wind
farms. The aerodynamic and acoustic optimization process with
decreasing the noise emission levels while increasing the aerody-
namic performance can be found in study conducted by Göçmen
and Özordem [10]. Their results show that redesigned airfoils have
lower levels of noise emission and higher lift to drag ratios. Mo and
Lee [11] numerically predicted the characteristics of aerodynamics
noise generated from rotating wind turbine blades using incom-
pressible large eddy simulation. The far-field aerodynamic noise
for frequency below 500 Hz was modeled using FW-H analogy.
They studied the aerodynamic noise due to the tip vortex-trailing
edge interaction by local cross flows along the trailing edges. Lee
and Lee [12] predicted aerodynamic noise from a 10 kW wind tur-
bine using semi-empirical models. They found that trailing edge
bluntness noise can be a dominant noise source for small wind tur-
bines unless the wind turbine blades have very sharp trailing
edges. Recently, Mohamed [13] carried out several noise evalua-
tions of H-rotor Darrieus wind turbines. He studied the blade
Several previous studies of the HAWTs noise used URANS
approaches which these models tend to be overly dissipative and
have been found to be poorly suited for prediction separated flow
typically encountered at high wind speed. Furthermore, due to
their inherent time-averaged nature, direct acoustic predictions
derived from RANS are questionable. This study addresses some
aspects of wind turbine noise generation and propagation not cov-
ered or not fully understood in the literature such as the role of
thickness, loading and quadrupole noises at different frequencies
and the effects of wind speed on flow separation and noise
generation.
This paper has used Improved Delayed Detached Eddy
Simulation (IDDES) to predict aerodynamic noise radiated from
the NREL Phase VI wind Turbine. The current study is an accurate
three-dimensional CFD unsteady simulation for aerodynamic noise
prediction of the flow around the NREL Phase VI wind turbine. The
IDDES turbulence model is conducted to obtain the instantaneous
turbulent flow field. The noise predictions are performed by the
FW-H acoustic analogy formulation. This paper focuses on the
broadband noises of the turbulent boundary layers and tonal
noises related to the passage of the blade. The surface
pressure coefficients for three different inflow conditions
U 1 ¼ 7; 10; 15 m=s were compared with the experimental data
by [14]. The effect of distance and wind speed on the Sound
Pressure Level spectrum and the Overall Sound Pressure Level
(OASPL) is studied. Furthermore, the contribution of the thickness,
loading and quadrupole noises were investigated, separately.
The purpose of this study is to evaluate the capability of the
IDDES turbulence model in the aerodynamic prediction of wind
Table 1
The contribution of different sources in the total noise.

shape, the tip-speed ratio and the solidity effects on radiated noise. U 1 (m/s) Thickness and loading noises Quadrupole noise Total noise
Results indicated that the higher solidity and higher tip-speed ratio (dB) (dB) (dB)
rotors produce much more noise than the normal turbines. His 7 46.7 49.3 51.2
study was based on a two-dimensional URANS (Unsteady RANS) 10 53.5 58.6 59.8
simulation which neglected three-dimensional effects. 15 60.7 64.6 66.1
212 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220

turbines and also, investigate the noise generation mechanisms in T ij ¼ qui uj þ Pij  a20 ðq  q0 Þdij ð4Þ
operating HAWTs.
where Pij is the compressive stress tensor that includes the surface
pressure and viscous stress.
2. Governing equations
The second term is related to the dipole or loading noise and the
last term correspond to the monopole or thickness noise. Thickness
2.1. IDDES formulation
and loading terms are surface distribution sources as indicated by
the presence of the delta function dðf Þ while the quadrupole term
The three-dimensional unsteady incompressible Navier–Stokes
is a volume distribution of sources indicated by Heaviside function
equations have been solved by the Improved Delayed Detached
Hðf Þ. The wave Eq. (3) can be integrated analytically under the
Eddy Simulation technique. The IDDES turbulence model is a
assumptions of the free-space flow and the absence of obstacles
hybrid RANS-LES model based on modification in the length scale
between the sound sources and the receivers. The complete solution
of the dissipation rate term in the Turbulent Kinetic Energy (TKE)
consists of surface integrals and volume integrals. The surface inte-
transport equation of the Menter’s (SST) k  x model [15].
grals represent the contributions from monopole and dipole acoustic
    
@ ðqkÞ @ quj k @ 1 @k qk3=2 sources and partially from quadrupole sources, whereas the volume
þ ¼ l þ lt þ sij Sij  ð1Þ integrals represent quadrupole (volume) sources in the region out-
@t @xj @xj rk @xj LIDDES
side the source surface. The contribution of the volume integrals
where q; k; t; uj ; l; lt ; sij and Sij are the density, turbulent kinetic becomes small when the flow is low subsonic, and the source surface
energy, time, velocity, molecular viscosity, turbulent viscosity, ten- encloses the source region and can be neglected. Thus, acoustic pres-
sor of stress and mean strain rate, respectively. sure p0 which is mentioned in Eq. (3) is composed as:
The IDDES length scale LIDDES is defined as [16]:
p0 ð~
x; tÞ ¼ p0T ð~
x; tÞ þ p0L ð~
x; tÞ ð5Þ

LIDDES ¼ ~f d ð1 þ f e ÞLRANS þ 1  ~f d LLES ð2Þ where ~ x is the observer position, t is the observer time, the sub-
scripts T and L correspond to thickness (monopole) and loading
where the length scales of RANS and LES are defined as:
(dipole) components, respectively, and as follows [21]:
k
1=2
Z " #
LRANS ¼ LLES ¼ C DES D q0 ðU_ n þ U n_ Þ
b x 4pp0T ð~
x; tÞ ¼ dS
f ¼0 rð1  M r Þ2 ret
For the IDDES, the grid scale is redefined as Z " #
D ¼ minfmaxfC w Dmax ; C w d; Dmin g; Dmax g, where C w is a fundamental
q0 U n ðrM_ r þ a0 ðMr þ M2 ÞÞ
þ dS ð6Þ
empirical constant, d the distance to the nearest wall, Dmin is f ¼0 r 2 ð1  Mr Þ3 ret



min Dx ; Dy ; Dz and Dmax is max Dx ; Dy ; Dz . Function ~f d is defined and
as maxfð1  f dt Þ; f B g, which is determined by both the geometry " # " #
Z Z
part f B and the flow part ð1  f dt Þ. The detailed information can be 1 L_ r Lr  LM
4p p0L ð~
x; tÞ ¼ dS þ dS
found in [17]. a0 rð1  M r Þ2 ret
f ¼0 f ¼0 r 2 ð1  M r Þ
3
ret
2  3
Z _ r þ a0 ðM r þ M 2 Þ
rM
2.2. Aeroacoustic formulation 1 4Lr 5 dS
þ ð7Þ
a0 f ¼0 r2 ð1  M r Þ3
The Ffowcs-William and Hawkings (FW-H) method [18], the ret

most general form of Lighthill acoustic analogy [19], is used to pre- where
dict the farfield noise. The FW-H method is appropriate for the pre-
q
diction of sound generated by rigid bodies in arbitrary motion and Ui ¼ v i þ ðu  v i Þ
is an inhomogeneous wave equation that can be derived by manip-
q0 i
ulating the continuity equation and the Navier–Stokes equations.
^ j þ qui ðun  v n Þ
Li ¼ Pij n
The FW-H equation can be written as [20]:
The various subscripted quantities appearing in (6) and (7) are the
1 @ 2 p0 @2

2 2
 r2 p0 ¼ T ij Hðf Þ inner products of a vector and a unit vector implied by the sub-
a0 @t @xi @xj ! !

@

 script. For instance, Lr ¼ L ~
 Pij nj þ qui ðun  v n Þ dðf Þ
@xi
@
þ f½q0 v n þ qðun  v n Þdðf Þg ð3Þ
@t
where un is the fluid velocity in the direction normal to the integra-
tion surface, v n is the normal velocity of the integration surface, dðf Þ
is Dirac delta function and Hðf Þ is Heaviside function. Subscript, 0,
defines the value in undisturbed medium and the primed value rep-
resents the difference between the value in real state and in the
undisturbed medium (e.g. p0 ¼ p  p0 Þ.
The shape and the motion of the control surface are defined by
f ð~
x; tÞ ¼ 0, with f < 0 for its interior and f > 0 for its exterior.
The terms in the right-hand side of Eq. (3) refer to different
mechanisms of sound production. The first source term involves
Lighthill stress tensor and shows that the time-dependent stresses
generate sound. Such sources are called quadrupoles. Lighthill
stress tensor is defined as:
r and U n ¼ U ~
n where ~r and ~
n denote
the unit vectors in the radiation and wall-normal directions, respec-
tively. The dot over a variable denotes the source-time differentia-
tion of that variable. The Mach number vector M i is the local surface
velocity vector divided by the freestream sound speed. The sub-
script ret denotes that the integrand is evaluated at the retarded
time, s defined as:
r
s¼t
a0
where t; r and a0 are receiver time, the distance to the receiver and
the speed of sound, respectively.
3. Problem description
The NREL Phase VI wind turbine is a two-bladed rotor. The
geometry of the blade is based on the S809 airfoil. The radius of
the blades is 5.029 m and the rated power of the wind turbine is
 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220 213

Fig. 1. Schematic of computational grid near the blade (a) computational domain, (b) grid near blade, and (c) boundary layer grid.

450
19.8 kW. The details about the blade geometry can be found in the
NREL/NASA Ames Phase VI unsteady experiment [14].
400
A 3-D incompressible unsteady computational fluid dynamics
solver, Ansys Fluent 15, based on the finite volume method is
employed to solve the Navier–Stokes equations using IDDES tech- 350
nique. The fluid is assumed to be incompressible as the free-stream
Torque [N.m]

Mach number is low ðM ¼ 0:0206Þ. Due to the incompressibility of 300

the flow, the pressure-based solver is chosen, which is traditionally
implemented to solve low-speed incompressible flows. The PISO
250
(Pressure Implicit with Splitting of Operator) algorithm is chosen
for coupling the velocity–pressure equations. Spatial discretization
has been preceded using a least squares cell based algorithm for 200
gradients, the bounded central differencing scheme for momen-
tum, second order upwind scheme for both turbulent kinetic
150
energy and specific dissipation rate. A bounded second order impli-
cit scheme was used for transient algorithm.
The computational domain consists of two regions; a rotating 100
0 2E+06 4E+06 6E+06 8E+06
region and a stationary region. To simulate rotating blade with Number of elements in grid
respect to the stationary outer domain the sliding mesh method
is used with a half-cylindrical interface between both domains. Fig. 2. Grid refinement study for NREL Phase VI blade at for U ¼ 7 m=s.
214 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220

While building mesh, special attention is given to minimize mesh
non-orthogonality and skewness. A good mesh quality is necessary
to obtain accurate results and to reduce the computational time.
Fig. 1 shows the computational domain and boundary layer
mesh near the blade. In order to resolve the boundary layer, a
boundary layer grid was attached to the blade surface. The height
of the first row of cells that bounding the blade is set to be 105 m
which ensures yþ < 1 for cells immediately adjacent to the blade.
To decrease both cell number of the grid and computation time,
only one blade is considered, and periodic boundary condition is
used. The numerical time step size is set to be 2:5  105 s in order
to ensure an accurate temporal discretization and small Courant
number.
Calculations were run for four revolutions, and data were col-
lected over the three last revolutions. Then the time-dependent
surface pressure fluctuations were used as the acoustic source field
data, and they were converted to the frequency domain using Fast
Fourier Transform (FFT).

U=7m/s @ r/R=0.30 U=7m/s @ r/R=0.467

4 4

IDDES IDDES
k-w SST [22] k-w SST [22]
3 Experiment [14] 3 Experiment [14]

2 2

-Cp
-Cp

1 1

0 0

-1 -1

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

U=7m/s @ r/R=0.633 U=7m/s @ r/R=0.80

4 4

IDDES IDDES
k-w SST [22] k-w SST [22]
3 Experiment [14] 3 Experiment [14]

2 2
-Cp
-Cp

1 1

0 0

-1 -1

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c
U=7m/s @ r/R=0.95
4

IDDES
k-w SST [22]
3 Experiment [14]

2
-Cp

-1

-2
0 0.2 0.4 0.6 0.8 1
x/c

Fig. 3. Pressure coefficient distribution of NREL Phase VI blade at wind speed of 7 m/s.
 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220 215

4. Grid independency study
It is necessary to perform a grid refinement study in every
Computational Fluid Dynamics simulation. In order to confirm
the independence of the results on the grid resolution, several grid
sizes were tested. Fig. 2 shows the torque of the blade for five dif-
ferent numbers of elements corresponded to U ¼ 7 m=s. As Fig. 2
shows, as the number of the grid exceeds about 5 million elements,
there is a negligible variation in the blade torque. So the final mesh
consists of 5,211,537 cells in two volumes.

U=10m/s @ r/R=0.30 U=10m/s @ r/R=0.467

6 6

IDDES IDDES
5 k-w SST [22] 5 k-w SST [22]
Experiment [14] Experiment [14]
4 4

3 3
-Cp

-Cp
2 2

1 1

0 0

-1 -1

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

U=10m/s @ r/R=0.633 U=10m/s @ r/R=0.80

7 6

IDDES IDDES
6 5 k-w SST [22]
k-w SST [22]
Experiment [14] Experiment [14]
5
4

4
3
3
-Cp
-Cp

2
2
1
1

0
0

-1 -1

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

U=10m/s @ r/R=0.95
4

IDDES
k-w SST [22]
3 Experiment [14]

2
-Cp

-1

-2
0 0.2 0.4 0.6 0.8 1
x/c

Fig. 4. Pressure coefficient distribution of NREL Phase VI blade at wind speed of 10 m/s.
216 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220

5. Results p  p1
Cp ¼ 
1
2
q u21 þ ðrxÞ2
5.1. Aerodynamics results
where p  p1 is the gage pressure, r is the radius of the blade sec-
In order to validate the simulation, the surface pressure coeffi- tion, and x is the rotational speed. Figs. 3–5 depict the surface pres-
cient on the blade was compared with the available experimental sure coefficient distribution on the five spanwise sections at 30%,
data [14]. The surface pressure coefficient is defined as: 46.7%, 63.3%, 80%, 95% of the blade span for three different inflow

U=15m/s @ r/R=0.30 U=15m/s @ r/R=0.467

5 3

IDDES IDDES
4 k-w SST [22] k-w SST [22]
Experiment [14] Experiment [14]
2

1
2

-Cp
-Cp

1
0

-1
-1

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

U=15m/s @ r/R=0.633 U=15m/s @ r/R=0.80

4 4

IDDES IDDES
k-w SST [22] k-w SST [22]
3 Experiment [14] 3 Experiment [14]

2 2
-Cp

-Cp

1 1

0 0

-1 -1

-2 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/c x/c

U=15m/s @ r/R=0.95
6

IDDES
5 k-w SST [22]
Experiment [14]
4

3
-Cp

-1

-2
0 0.2 0.4 0.6 0.8 1
x/c

Fig. 5. Pressure coefficient distribution of NREL Phase VI blade at wind speed of 15 m/s.
 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220 217

conditions, U ¼ 7; 10; 15 m=s. At 7 m/s inflow velocity, the pressure
coefficient at all spanwise sections show good agreement with
experimental data. At 7 m/s wind speed which is categorized as
low wind speed, the flow is mostly attached. At 10 m/s wind speed
which the stall commences to occur, a discrepancy is seen near the
leading edge of the suction side at 46.7% of the blade span. This dis-
crepancy is due to flow separation near the mid-span of the blade.
At 15 m/s, wind speed which is corresponds to stall flow, some
deviations were found near the leading edge of suction side at
30% of the blade span. The value of these discrepancies in the
IDDES turbulence model is less than the RANS turbulence models.
At high wind speeds, where the flow is massively separated over
the entire blade span, RANS models fail to predict the correct sepa-
ration and the lift and torque is over predicted [22]. These results
show the superiority of the IDDES turbulence model in the separa-
tion prediction as compared to the conventional RANS models.
Fig. 6 shows the instantaneous pressure contours on the blade
pressure side for three inflow conditions. For all conditions, the
contours show considerable variations in the both spanwise and
chordwise directions.
Turbulent vortical structures of the flow field for two flow con-
ditions U 1 ¼ 7; 15 m=s are visualized in Fig. 7. To identify these
vortex structures, the Q-criterion isosurface colored with the val-
ues of the velocity magnitude is shown.
The Q-criterion which is the second invariant of the velocity
gradient tensor is defined as [23]:
1 
Q¼ Xij Xij  Sij Sij
2

Fig. 6. Instantaneous pressure contour on the NREL Phase VI blade pressure side.

Fig. 7. Isosurface of turbulent vertical structures for Q = 200 s2 contoured by the velocity magnitude for U 1 ¼ 7 m=s (left) and U 1 ¼ 15 m=s (right).
218 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220

80 80

70 IDDES U=7 [m/s]

Tadamasa & Zangeneh [7] f = 2.4 [Hz] U=10 [m/s]
60 U=15 [m/s]
60
Sound Pressure Level [dB]

Sound Pressure Level [dB]

50 40

40

20
30

20
0

10

-20
0
200 400 600 800 1000
0 1 2
Frequency [Hz] 10 10 10
Frequency [Hz]
Fig. 8. Comparison of Sound Pressure Level the NREL Phase VI wind turbine at
U 1 ¼ 15 m=s with Tadamasa and Zangeneh [7]. Fig. 10. Wind speed effects on the noise generation of NREL Phase VI wind turbine.

80

Thickness & Loading Noise 80 D=14.74 [m]

Total Noise D=20.87 [m]
2.4 [Hz]
D=36.18 [m]
60 D=69.29 [m]
D=137.02 [m]
60
Sound Pressure Level [dB]
Sound Pressure Level [dB]

40
40

20 20

0
0

-20

-20 0 1 2
10 10 10 10
0
10
1
10
2

Frequency [Hz]
Frequency [Hz]
Fig. 9. Sound Pressure Level of different mechanism of noise generation from the
NREL Phase VI wind turbine at U 1 ¼ 15 m=s. Fig. 11. The effect of distance on Sound Pressure Level at U 1 ¼ 7 m=s.

where And p^ref has the value of 2:105 Pa.

    In the present computation, only one blade of the whole wind
1 @ui @uj 1 @ui @uj
Sij ¼ þ Xij ¼  turbine of Tadamasa and Zangeneh [7] is simulated. It is assumed
2 @xj @xi 2 @xj @xi that every blade generates the same acoustic pressure; therefore
It can be observed that in U 1 ¼ 7 flow condition, the flow is the Sound Pressure Level of the whole wind turbine is computed
attached and the turbulent vortical structures are only shed from as follows:
the blade tip. Whereas, at U 1 ¼ 15, the flow is massively separated
over the entire blade span. ! !!
^2one blade
p ^2one blade
p
Ltwo blade ¼ 10log 10 þ
5.2. Aeroacoustics results ^2ref
p ^2ref
p
!!
To confirm the accuracy of the aero-acoustic simulation, the ^2one blade
p
¼ 10log 10 2 
Sound Pressure Level is compared with the available numerical ^2ref
p
results of Tadamasa and Zangeneh [7]. Sound Pressure Level is a !!
^2one blade
p
logarithmic measure of the effective sound pressure of a sound rel- ¼ 10 log 10 ð2Þ þ log 10 ¼ 3:01 þ Lone blade
ative to a reference value and is defined as follows [5]: ^2ref
p
!
^2
p Fig. 8 presents a comparison of the Sound Pressure Level
LP ¼ 10  Log 10 between the current simulation and the numerical results of
^2ref
p
Tadamasa and Zangeneh [7] at U 1 ¼ 15 m=s inflow condition and
 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220 219

rotational speed of 72 rpm. The Sound Pressure Level is reported
for a receiver located at the reference distance. This reference dis-
tance is according to IEC 61400-11 international standard [11].
It can be observed that there is a good agreement between
IDDES turbulence model and published results [7].
In order to show the low frequency noises in acoustic spectra,
the Sound Pressure Level at U 1 ¼ 15 m=s inflow condition and
rotational speed of 72 rpm is shown in Fig. 9 in logarithm scale.
It can be seen that there is a tonal peak in the sound spectrum at
f ¼ 2:4 Hz that matches with the blade rotation frequency.
1 min
x ¼ 72 rpm ! x ¼ 72 rpm  ffi 1:2
60 s
Number of blades ¼ 2
Rotation frequency ¼ Number of blades  number of
¼ 2  1:2 ¼ 2:4
It is generally known that the Sound Pressure Level at the recei-
ver position is highly depending on the distance between the recei-
ver and the noise source. The further away from the wind turbine,
the lower SPL at the receiver position is observed. The receiver
position effect is shown in Figs. 11–13 where the Sound Pressure
Level is computed at different distances from the wind turbine. It
can be seen that the Sound Pressure Level spectrums show signif-
icant peaks in the sound intensity at the rotating frequency at each
distance between the receiver and the wind turbine.
Figs. 11–13 are related to the noise spectrum for the combina-
tion of thickness and loading noise at U 1 ¼ 7; 10; 15 m=s, respec-
revolution
tively. It can be concluded that the noise amplitude is reduced
second with increasing the distance between the receiver and the wind
turbine. However, there is no change in the behavior of the spec-
trum and the location of the tonal peaks.
The effect of distance on the Overall Sound Pressure Level
revolution
(OASPL) for three mentioned flow conditions is shown in Fig. 14.
second

As previously mentioned, sources of aerodynamic noise generated 100

by the wind turbine consist of the thickness, loading, and quadru- D=14.74 [m]
pole noise. The combination of thickness and loading noise is D=20.87 [m]
obtained by integrating on the blade surface and the total noise is 80 D=36.18 [m]
D=69.29 [m]
calculated by integrating over the interface surface. The difference D=137.02 [m]
between the total noise and the combination of thickness and load- Sound Pressure Level [dB]
60
ing noise indicates the quadrupole noise. Fig. 9 shows the total and
the combination of thickness and loading noise spectrum. These
spectrums indicate that the quadrupole noise has negligible influ-
40
ence on the tonal noise, but the combination of thickness and load-
ing noise is the dominant noise sources at these frequencies. Fig. 9
shows several peaks in the high frequency region for the total noise
20
that cannot be seen in the combination of thickness and loading
noise spectrum. It would be concluded that the quadrupole noise
is related to the turbulent structures behind the wind turbine.
0
Fig. 10 shows the effects of increasing wind speed on the com-
bination of thickness and loading noise. The Sound Pressure Level
(SPL) is reported for a receiver located at the reference distance
-20
to the wind turbine. There is a direct relation between the strength 10
0
10
1
10
2

of radiated noise and the wind speed. It can be concluded that the Frequency [Hz]
noise amplitude is increased with increasing the wind velocity.
However, there is no significant change in the behavior of the spec- Fig. 13. The effect of distance on Sound Pressure Level at U 1 ¼ 15 m=s.
trum and the location of the tonal peaks.

70

65 U=7 [m/s]
80 D=14.74 [m]
D=20.87 [m] U=10 [m/s]
D=36.18 [m] U=15 [m/s]
D=69.29 [m] 60
D=137.02 [m]
60
55
Sound Pressure Level (dB)

OASPL [dB]

50
40
45

20 40

35

0
30

25
50 100 150 200
-20 0 1 2
10 10 10 Distance [m]
Frequency (Hz)
Fig. 14. The overall Sound Pressure Level due to combination of thickness and
Fig. 12. The effect of distance on Sound Pressure Level at U 1 ¼ 10 m=s. loading noise at different distance from the wind turbine for U 1 ¼ 7; 10; 15 m=s.
220 M. Ghasemian, A. Nejat / Energy Conversion and Management 99 (2015) 210–220
The OASPL was evaluated by integration over the frequency
spectrum.
It can be seen that the OASPL does not vary linearly with the
receiver distance. This means that doubling the receiver distance
does not lead to the halving of Sound Pressure Level. The OASPL
varies logarithmically with the receiver distance. Furthermore,
Fig. 14 indicates that there is a direct relation between the
OASPL and the wind speed. It can be observed that the OASPL is
6 dB greater at U 1 ¼ 15 m=s in comparison with U 1 ¼ 10 m=s
and 8 dB greater at U 1 ¼ 10 m=s in comparison with U 1 ¼ 7 m=s
in each distance from the wind turbine.
To indicate the importance of the quadrupole noises in the total
noise, the contribution of every source in the OASPL for three dif-
ferent inflow conditions, U 1 ¼ 7; 10; 15 m=s, is reported in
Table 1. The receiver is located at the reference distance. These
data indicate that the quadrupole noises form a major part of the
total noise.
6. Conclusion
In the present paper, the aerodynamic and aero-acoustic of the
flow field around the NREL Phase VI wind turbine was studied. The
IDDES turbulence model was conducted to obtain the instanta-
neous turbulent flow field. The Ffowcs William and Hawkings
(FW-H) acoustic analogy was applied to predict the Farfield noise.
Simulations were performed for three different inflow conditions,
U ¼ 7; 10; 15 m=s. The surface pressure coefficient at five blade sec-
tions were verified by experimental data. The good agreement in
the surface pressure shows the advantages of the IDDES turbulence
model for the separation prediction as compared to the conven-
tional RANS models. To illustrate the tip vortex in low velocity con-
dition and massive separation in high velocity circumstance,
turbulent vortical structures around blades were visualized. The
acoustic pressure spectra of the thickness, loading and quadrupole
noise sources were presented and the contributions of each source
were investigated. The results indicate that the quadrupole noise
has negligible influence on the tonal noise, but the combination
of thickness and loading noise are the dominant noise sources at
those frequencies. The effect of the wind speed and the distance
between the wind turbine and the receivers were studied. The
results showed a direct relation between the strength of the radi-
ated noise and the wind speed. The Sound Pressure Level spectrum
indicated relatively significant peaks in the sound intensity at fre-
quency of 2.4 Hz for rotational speed of 72 rpm. This tonal peak in
the sound spectra matches with the corresponding blade passing
frequency for this rotational speed. The Overall Sound Pressure
Level (OASPL) at different distances ranging from 20 m to 140 m
were calculated and the results showed it decreases logarithmi-
cally with the receiver distance. The OASPL is 6 dB greater at
U 1 ¼ 15 m=s in comparison with U 1 ¼ 10 m=s and 8 dB greater
at U 1 ¼ 10 m=s in comparison with U 1 ¼ 7 m=s in each distance
from the wind turbine.
Therefore, the results of this paper can improve a better under-
standing of the noise generation mechanisms as well as the design
of quieter small Horizontal Axis Wind Turbine. Furthermore, these
results can be exploited to determine the efficient location for
wind turbine installation that the noise radiated has the minimum
annoying to the vicinity residents.
References
[1] Mostafaeipour A. Economic evaluation of small wind turbine utilization in
Kerman, Iran. Energy Convers Manage 2013;73:214–25.
[2] Mohamed M, Janiga G, Pap E, Thévenin D. Optimal blade shape of a modified
Savonius turbine using an obstacle shielding the returning blade. Energy
Convers Manage 2011;52:236–42.
[3] Al-Hadhrami LM. Performance evaluation of small wind turbines for off grid
applications in Saudi Arabia. Energy Convers Manage 2014;81:19–29.
[4] Jianu O, Rosen MA, Naterer G. Noise pollution prevention in wind turbines:
status and recent advances. Sustainability 2012;4:1104–17.
[5] Wagner S, Bareiss R, Guidati G. Wagner–Bareiß–Guidati. In: Wind turbine
noise; 1996.
[6] Stone C, Lynch CE, Smith MJ. Hybrid RANS/LES simulations of a Horizontal Axis
Wind Turbine. In: 48th AIAA aerospace sciences meeting, AIAA-2010-459,
Orlando, FL2010.
[7] Tadamasa A, Zangeneh M. Numerical prediction of wind turbine noise. Renew
Energy 2011;36:1902–12.
[8] Cho T, Kim C, Lee D. Acoustic measurement for 12% scaled model of NREL Phase
VI wind turbine by using beamforming. Current Appl Phys 2010;10:S320–5.
[9] Chourpouliadis C, Ioannou E, Koras A, Kalfas AI. Comparative study of the
power production and noise emissions impact from two wind farms. Energy
Convers Manage 2012;60:233–42.
[10] Göçmen T, Özerdem B. Airfoil optimization for noise emission problem and
aerodynamic performance criterion on small scale wind turbines. Energy
2012;46:62–71.
[11] Mo J-O, Lee Y-H. Numerical simulation for prediction of aerodynamic noise
characteristics on a HAWT of NREL Phase VI. J Mech Sci Technol
2011;25:1341–9.
[12] Lee S, Lee S. Numerical and experimental study of aerodynamic noise by a
small wind turbine. Renew Energy 2014;65:108–12.
[13] Mohamed M. Aero-acoustics noise evaluation of H-rotor Darrieus wind
turbines. Energy 2014;65:596–604.
[14] Simms DA, Schreck S, Hand M, Fingersh L. NREL unsteady aerodynamics
experiment in the NASA-Ames wind tunnel: a comparison of predictions to
measurements. In: National Renewable Energy Laboratory Colorado, USA;
2001.
[15] Menter FR. Two-equation eddy-viscosity turbulence models for engineering
applications. AIAA J 1994;32:1598–605.
[16] Xiao L, Xiao Z, Duan Z, Fu S. Improved-Delayed-Detached-Eddy Simulation of
cavity-induced transition in hypersonic boundary layer. Int J Heat Fluid Flow
2014;51:138–50.
[17] Shur ML, Spalart PR, Strelets MK, Travin AK. A hybrid RANS-LES approach with
delayed-DES and wall-modelled LES capabilities. Int J Heat Fluid Flow
2008;29:1638–49.
[18] Williams JF, Hawkings DL. Sound generation by turbulence and surfaces in
arbitrary motion. Philos Trans R Soc London Ser A, Math Phys Sci
1969;264:321–42.
[19] Lighthill MJ. On sound generated aerodynamically. I. General theory. Proc R
Soc London Ser A Math Phys Sci 1952;211:564–87.
[20] Di Francescantonio P. A new boundary integral formulation for the prediction
of sound radiation. J Sound Vib 1997;202:491–509.
[21] Farassat F, Succi GP. The prediction of helicopter rotor discrete frequency
noise. In: American Helicopter Society, Annual Forum, 38th, Anaheim, CA, May
4–7, 1982, Proceedings (A82-40505 20-01) Washington, DC, American
Helicopter Society; 1982. p. 497–507.
[22] Yelmule MM, VSJ EA. CFD predictions of NREL Phase VI Rotor Experiments in
NASA/AMES Wind tunnel. Int J Renew Energy Res (IJRER) 2013;3:261–9.
[23] Chakraborty P, Balachandar S, Adrian RJ. On the relationships between local
vortex identification schemes. J Fluid Mech 2005;535:189–214.