an efficient noise modeling tool for wind tur- · pdf filebines including sound propagation in...
TRANSCRIPT
1
AN EFFICIENT NOISE MODELING TOOL FOR WIND TUR-BINES INCLUDING SOUND PROPAGATION IN ARBITRARY WEATHER CONDITIONS
Sterling McBride and Ricardo Burdisso
Virginia Tech – Department of Mechanical Engineering, Durham Hall, Blacksburg, Virginia 24060
email: [email protected]
José David Parra
Universidad Austral de Chile – Instituto de Acústica, General Lagos 2086, Valdivia, Chile
Wind energy is the world´s fastest-growing renewable energy source. Thus, the number of people
exposed to wind farm noise is increasing. Due to its broadband amplitude modulated characteris-
tic, wind turbine noise (WTN) is more annoying than noise produced by other common commu-
nity/industrial sources. Aerodynamic noise on the blades is the dominant noise source of modern
large wind turbines. Physically accurate methods for the prediction of acoustic noise produced
by wind turbines and farms are crucial for their environmental impact assessment, including their
amplitude modulation behavior. The state-of-the-art approach to model the aerodynamic noise
from wind turbines is to divide the blades into a number of radial segments. A noise source is then
associated to each element. From the wind profile, blade geometry and aerodynamic parameters
(e.g. angle of attack, etc.), the strength and directivity of each of the sources are estimated. Finally,
the noise from each source is coupled to a propagation code to account for weather conditions.
This process has to be performed at each angular position of the blades as it completes a full
rotation. This approach results in hundreds of noise sources needed to model a single turbine.
The result of this process is extremely computationally intensive calculations, unfeasible to model
realistic wind farms. We propose a novel method for modeling wind turbine's noise and its atmos-
pheric propagation. The approach consists of computing an equivalent noise source placed at the
turbine hub with a strength and directivity that is a function of the rotor angular position. This
single equivalent source is then coupled to a curved ray tracing propagation code. The approach
is demonstrated for a 5MW modern wind turbine over a flat acoustically soft terrain. The results
show the proposed modeling approach to be computationally efficient and accurate.
1. Introduction
Power production by renewable wind energy has risen sharply and worldwide in the last 20 years.
This energy source has the highest growth rate of all renewable sources at >20% increase of installa-
tions annually. Most of the onshore wind turbines are subjected to noise constraints and ever more
stringent regulations. Additionally, wind turbines are continuously increasing in size to achieve
higher outputs per unit for profitability and efficiency. Thus, these larger turbines are louder and
potentially more annoying to the people living nearby.
The blade broadband aerodynamic noise is the dominant noise source in modern large wind tur-
bines. Moreover, it is also responsible for the amplitude modulation (AM) observed in mainly large
turbines, i.e. pulsating broadband sound [1]. This AM, commonly referred as “swishing”, “whoosh-
ing” or “pulsating noise”, is currently considered as the main cause of annoyance for residents near
wind farms [2,3].
The 23rd International Congress on Sound and Vibration
2 ICSV23, Athens (Greece), 10-14 July 2016
Prediction of the noise emission from wind turbines and farms is very important, in particular for
new farms. Accurate noise predictions can be used for environmental impact analyses, developing
noise control strategies, and optimizing wind farm layout. Accurate predictions require the modelling
of the flow on the blades and the resulting noise source strengths and directivity. For long propagation
distances (a few kilometres), atmospheric conditions including thermal stratification, humidity, and
uneven terrain need to be accounted for.
There are several commercial codes such as SoundPLAN, CadnaA, Predictor-LIMA, WindPRO
and so forth for predicting wind turbine noise with models of increasing degree of sophistication.
However, all of them assume a monopole at the hub, which is a drawback because it does not consider
the actual radiation characteristics of the wind turbine. They simply rely on estimates of the turbine’s
sound power levels from manufacturers’ databases or on user input. The most advanced propagation
model implemented corresponds to NORD2000. It consists of a simplified ray tracing method that
can support moderate atmospheric refraction by assuming that sound speed varies linearly with
height, and that all rays follow a circular path. It also takes into account atmospheric absorption for
refracting and non-refracting media [4].
The state-of-the-art aerodynamic noise from wind turbine is based on solving the 2D flow field
over the blades and using these results to predict the aerodynamic noise using semi-empirical models
[5] or scaling of wind tunnel data [6]. However, these models haven’t yet been coupled to propagation
codes. They use simple straight ray propagation over flat terrains for the prediction of noise near the
turbine [7,8].
In this work, a state-of-the-art model of the turbine noise was coupled to a curved ray tracing code
for the prediction of noise over long distances (several kilometres). However, the code is computa-
tionally intensive with the propagation part taking the overwhelming majority of the computational
effort. Thus, a new modelling turbine noise method is proposed to properly capture the radiation
characteristics of the turbine and the atmospheric variation (wind and temperature profiles) in the
propagation at a reasonable computational effort. The approach consists of defining an equivalent
sound source placed at the turbine hub with a source strength and radiation pattern that is a function
of the turbine rotation or azimuth position of rotor.
2. Turbine noise model
As shown in Fig. 1, the WTN model consists of five modules, which are briefly described here.
The input to the code consists of the turbine and blade geometry, operating conditions, atmospheric
data, ground impedance, and execution control parameters. The turbine blades are assumed rigid, the
terrain flat, and the atmospheric conditions uniform over the domain but arbitrary with height. The
blades are then split in span-wise direction elements and the blade rotation approximated as a discrete
set of azimuth positions. Thus, this approach defines a finite number of positions on the rotor plane
to perform aerodynamic and noise calculations as shown in Fig. 2a. The sound sources characterizing
the turbine noise radiation will be defined at these points. The second module is the Aerodynamic
Module, which uses a blade element momentum method (BEM) to compute the aerodynamic param-
eters needed for noise calculations, i.e. angle-of-attack (AoA) and relative flow Mach for all blade
elements while the turbine is operating [9]. To this end, the airfoil section polars are either computed
using XFoil [10] or taken from data collected in a wind tunnel [6] to account for the induction effects.
Turbine yaw, tilt, and conning angles are accounted for in the calculation of the AoAs. Fig. 2b illus-
trates the resulting AoAs for all the positions shown in Fig. 2a for a particular wind profile.
The flow conditions around the blades of a wind turbine govern wind turbine aerodynamic noise
generation mechanisms. In the Noise Source Module, the aerodynamic noise sources (leading and
trailing edge noise) are computed for the selected blade elements and the set of azimuth blade posi-
tion. This module uses the code NAFNoise [11] or wind tunnel data [6,12] to predict the aerodynamic
noise in 1/3rd octave bands at a single point in the direction normal to the airfoil chord line at a distance
of 1 meter. Five different noise mechanisms for airfoils can be included. They are turbulent boundary
The 23rd International Congress on Sound and Vibration
ICSV23, Athens (Greece), 10-14 July 2016 3
layer-trailing edge noise (TBLTE), separation stall flow noise (SSF), laminar boundary layer noise
(LBLVS), trailing edge bluntness noise (TEB), and turbulent inflow noise (TI). The TBLTE, SSF,
LBLVS, and TEB are known as airfoil self-noise. Semi empirical noise solutions to these mechanisms
were developed by Brooks et al. [5] and implemented in NAFNoise. On the other hand, TI corre-
sponds to the noise produced by the interaction of the external inflow with the airfoil [13]. Amiet [14]
and Moriarty [11] developed methods to solve for TI. The radiation directivity of the sources pro-
posed by Brooks et al. [5] are applied to define sound spheres to couple with the propagation module.
An example of the resulting noise spectrum for position 4 in Fig. 2a as computed by NAFNoise is
shown in Fig. 2c. Upon implementing the radiation directivity, the resulting sound sphere centred at
the trailing (or leading) edge of the airfoil element contained is shown in Fig. 2d.
The next step is the Propagation Module that implements a curved ray tracing propagation of the
individual sound spheres. This module uses the wind and temperature profiles and other atmospheric
conditions (humidity, pressure, etc.) to predict the noise at an array of microphones in the domain,
typically over a plane parallel to the ground. To this end, a large number of rays are emitted from the
sound sources. The acoustic losses, due to atmospheric attenuation and ground reflections, and Dop-
pler effect of the moving sound sources are accounted for in this module. The ray tracing code im-
plemented here is based on a NASA code developed for prediction of noise from fixed wind aircrafts
and helicopters [15]. A new Hamiltonian ray tracing formulation will be implemented in the near
future [16]. Fig. 2e illustrates the propagation of rays from a sound source.
The final module, Turbine Noise, concatenates the noise at the microphones produced by all the
blade element sound sources on the 3 blades and rotor azimuth positions for all 1/3rd octave frequency
bands. Fig. 2f shows a typical resulting noise map corresponding to a particular azimuth position.
Figure 1: Turbine noise modelling approach.
The coupling of the blade noise sources to the ray tracing propagation method is one of the novel
aspects of the work presented here. To the best of the author’s knowledge, this is the first time such
coupling of the turbine noise source to a propagation code has been reported in the open literature.
Input/Control
Module
Aerodynamic
Module AoA
Relative Mach
Noise Source
Module
Propagation
Module
Noise Spectrum NAFNoise Wind tunnel data
Input blade geometry, wind and temperature pro-files, ground impedance, and code control param-
eters:
Turbine Noise
Module
Polars XFoil Wind tunnel data
Ray Tracing Conventional Hamiltonian
Turbine Noise Spectra Maps Animation
Noise Spectrum
Sound sphere
Rays
Blade divided
in elements
The 23rd International Congress on Sound and Vibration
4 ICSV23, Athens (Greece), 10-14 July 2016
(a) Points for aero and noise calculations
(b) Angle of attack on rotor plane
(c) Noise Spectrum source 4
(c) Sound sphere 4 at 500 Hz 3rd octave band
(e) Ray tracing propagation
(f) Noise maps
Figure 2: (a) Points on rotor plane for aerodynamic and noise calculation, (b) AoA as the blade makes a rota-
tion in a non-uniform flow, (c) noise source 4 spectrum computed by NAFNoise, (d) sound sphere for
source 4, (e) ray trajectories, and (f) resulting OASPL noise map due to turbine at 48 ̊azimuth position.
3. Proposed simplified turbine noise model
The most computationally intensive process in the noise modeling approach described above is
the propagation module. The main reason is that each sound sphere on the blades at each azimuth
position must be propagated through the domain accounting for the weather conditions. This process
overwhelmingly requires the most computational time (> 95%) which makes this direct approach
very difficult to implement, in particular at the farm level with many installed wind turbines.
In order to mitigate this problem, a simplified approach is proposed consisting of modeling the
turbine noise with a single equivalent sound source placed at the turbine hub that accounts for the
acoustic characteristic of the blades as they rotate. Fig. 3 illustrates the approach. Fig. 3a shows the
rotor at a particular azimuth position and the sound spheres accounting for the aerodynamic noise
AoA (deg)
SPL(dBA)
OASPL(dBA) Azimuth position = 48 ̊
Wind Turbine
The 23rd International Congress on Sound and Vibration
ICSV23, Athens (Greece), 10-14 July 2016 5
produced by the blade sections. The sound spheres have varying strength and directivity from the
hub to the tip as observed in Fig. 3a. This is a consequence of the changing inflow, blade twist and
airfoil geometry along the blade. Furthermore, the sound sphere will also change as a function of the
rotor azimuth position as the blade sees a non-uniform inflow (not shown in the figure). In order to
create an equivalent sound source located at the hub, all the sources observed in Fig. 3a are added in-
coherently. In adding the sound spheres to generate the equivalent one, the key issue is to select a
distance from the hub to perform the addition. In other words, the equivalent sound source at the hub
in Fig. 3b will match the noise due to all the sound sources in Fig. 3a over a sphere centered on the
hub in free-field. Thus, the method being described is accurate only in the geometric far field of the
turbine as a whole. For a case where the distances from the sources to observers are small relative to
the rotor diameter, near field effects take place, where distance and direction information become
important. The equivalent sound sphere is not expected to be an accurate representation of the turbine
noise in the near field. However, the noise prediction away from the turbine is of most interest. It is
suggested that the distance to use for the calculation of the equivalent sound source be a few (2-4)
rotor diameters. Fig. 3b illustrates the result for a specific rotor azimuth position. As expected, the
generated equivalent sound source is characterized by a strength and directivity pattern that changes
with rotor azimuth positions (not shown).
Computational time is significantly reduced with a single equivalent source located at the hub. By
having a single source, the propagation computation has to be performed only once. This should be
contrasted to the full formulation where propagation has to be performed from all the points on the
rotor plane (see Fig. 2a).
Figure 3: (a) Array of noise source representing the turbine noise at a particular azimuth position, (b) the
equivalent noise source representing the in-coherent addition of these sources.
4. Numerical Example
4.1 Description of Turbine and Weather Conditions
A simple example problem is presented here to illustrate the proposed approach. The selected
generator is the NREL 5MW reference turbine [17]. The reason for using this turbine is that the blade
geometry and other parameters are available in the open literature. The rated rotor speed is 12.1 rpm.
The length of the blades is 61.5 meters and its maximum chord is 4.65 meters. The blade airfoil
sections are composed of a series of circular, DU and NACA airfoils. For the simulations, it is as-
sumed the hub height is 100 m and the turbine operates at 10 rpm with an inflow of 10 m/s at the hub.
The turbine yaw and tilt angles are set to zero and the rotor is not conned either. The weather condition
consists of the non-uniform wind and temperature profiles shown in Fig. 4. In this figure, the black
line in the wind profile plot sketches the position of the rotor plane. They were generated by modify-
ing experimentally measured data [18]. There is no vertical wind component in the simulations. The
terrain was assumed flat and covered with short grass with a uniform flow resistivity of 225 rays. The
flow resistivity was used to compute the ground impedance and absorption.
Sound spheres
at blades
Equivalent Sound
sphere at the hub
The 23rd International Congress on Sound and Vibration
6 ICSV23, Athens (Greece), 10-14 July 2016
The blades were divided in 5 span-wise elements and the rotation accounted for by taking 15 azi-
muth positions for a total of 75 sound sources distributed on the rotor plane. A total of 702 rays were
emitted from each sound sphere. NAFNoise was used to predict the trailing edge noise for the 75
sound sources, e.g. leading edge noise was not modelled. As explained before, the equivalent sound
source was computed such as to match the noise levels at a distance of 400 meters from the hub. Once
again 702 rays were emitted from the equivalent sound source and propagated through the medium.
The formulations were implemented using Matlab and the simulations were performed on a 3.42-
GHz quad-core personal computer with 16 GB of RAM. It is important to mention that NAFNoise
and the ray tracing code are coded in Fortran.
Figure 4: Wind and temperature profile used in the simulations.
4.2 Results
The code computes the 1/3rd octave band spectrum for an array of microphones at each azimuth
position of the rotor. In these simulations, a square grid of 1600 microphones was placed on the
ground over an area of 2 km by 2 km. The turbine is at the centre in the domain. Background noise
was not added to the turbine noise results. The resulting noise maps at the 250 and 500 Hz 3rd octave
bands for the rotor in the zero azimuth position for the full or direct (left column) and equivalent
sound source (right column) formulations are shown in Fig. 5. The turbine position and wind direction
are shown in this figure. It is clear that the equivalent source formulation is capable of modelling the
turbine noise reasonably well, in particular in the downwind direction. However, the equivalent sound
source tends to smooth out the results and thus it cannot capture areas where the turbine noise is
amplified, most likely due to the contribution of different sections of the blades that radiates towards
these local regions due to refraction effects. The single equivalent sound source at the hub clearly
cannot capture this behaviour.
Fig. 6 shows a comparison of the average spectrum for microphones at 150 and 400 m distance
downwind the turbine, respectively. It can be observed that the equivalent sound source approach
predicts very well the turbine noise in particular as the observer moves away from the turbine.
Finally, the computational time was significantly reduced from 5.8 hours for the full formulation
to just 40 minutes for the equivalent source approach.
Rotor plane
The 23rd International Congress on Sound and Vibration
ICSV23, Athens (Greece), 10-14 July 2016 7
(a) 250 Hz
(b) 250 Hz
(c) 500 Hz
(d) 500 Hz
Figure 5: Noise map for rotor on zero azimuth position predicted by (a) and (b) full formulation (array of
sound sources on blades) and (c) and (d) a single equivalent sound source at the hub.
Figure 6: Average spectrum in 1/3rd Octave bands for microphones at 150 and 400 meters distance down-
wind from turbine for full and proposed equivalent formulations.
5. Conclusions
A state-of-the-art wind turbine noise prediction tool was presented. The code models the flow
conditions over the blades over a full revolution. The aerodynamic variables are then used to compute
the aerodynamic noise sources along the trailing and leading edge of the blades. These noise sources
are then coupled to a curved ray tracing propagation tool to account for atmospheric effects over long
Mic. at 150 m
Mic. at 400 m
Formulation
Full
Equivalent source
Wind
The 23rd International Congress on Sound and Vibration
8 ICSV23, Athens (Greece), 10-14 July 2016
propagation distances. The main limitation of this tool is the excessive computational tool required
for the propagation of the noise in the atmosphere. To this end, a new modelling approach is proposed
which consists of defining a single equivalent source with azimuth varying strength and directivity
placed at the hub. This single source is then coupled to the curved ray tracing propagation code. The
approach was tested using the NREL 5MW reference turbine. As expected, the results show accurate
predictions in the far field with a significant computation time improvement. However, the approach
did not capture the built up of sound pressure levels at localized areas due to the contribution of
different sections of the blades that radiates towards these local regions because of refraction effects.
Future work includes the implementation of a more computational efficient Hamiltonian ray tracing
tool and the noise predictions of wind farms.
REFERENCES
1. Stigwood M., Large S., Stigwood D. Audible amplitude modulation – results of field measurements and investigations
compared to psychoacustical assessment and theoretical research, 5th International Conference on Wind Turbine Noise,
Denver, Colorado, USA, 28-30 August, (2013).
2. Van den Berg, G. The Beat is Getting Stronger: The Effect of Atmospheric Stability on Low Frequency Modulated
Sound of Wind Turbines, Journal of Low Frequency Noise, Vibration and Active Control. 24 (1), 1−24, (2005).
3. Bolin K., Bluhm G., Eriksson G. Nilsson M. E. Infrasound and low frequency noise from wind turbines: exposure
and health effects. Environmental Research Letters, 6 (3), 035103, (2011).
4. Plovsing, B. DELTA acoustics report AV 1851/100, Nord2000. Comprehensive Outdoor Sound Propagation Model.
Part 2: Propagation in an Atmosphere with Refraction, (2001).
5. Brooks, t., Pope, S., & Marcolini, M. National Aeronautics and Space Administration, NASA Reference Publication
1218. Airfoil Self-Noise and Prediction, (1989).
6. Devenport W., Burdisso R.A., Camargo H., Crede E., Remillieux M., Rasnick M., and Van Seeters P. National
Renewable Energy Laboratory, Technical Report NREL/SR-500-4347, Aeroacoustic Testing of Wind Turbine Air-
foils, (2010).
7. Moriarty, P. and Migliore, P. National Renewable Energy Laboratory, Technical Report NREL/TP-500-34478,
Semi-Empirical Aeroacoustic Noise Prediction Code for Wind Turbines, (2003).
8. Fuglsang P and Madsen HA. Risø National Laboratory, Risø-R-867(EN), Implementation and verification of an
aeroacoustic noise prediction model for wind turbines, (1996).
9. Sanderse B. Energy Research Centre of the Netherlands, Report ECN-e--09-016, Aerodynamics of wind turbine
wakes, (2009).
10. Drela, M. and Youngren, H. XFOIL 6.94 User Guide, Massachusetts Institute of Technology, Cambridge, Massa-
chusetts, (2001).
11. Moriarty, P. NAFNoise User’s Guide. National Wind Technology Centre, National Renewable Energy Laboratory.
Golden, Colorado, (2005).
12. Migliore P.J. & Oerlemans S. Wind Tunnel Aeroacoustic Tests of Six Airfoils for Use on Small Wind Turbines.
AIAA Wind Energy Symposium, Reno, Nevada, (2004).
13. Oerlemans, S. Detection of aeroacoustic sound sources on aircraft and wind turbines. Thesis Dissertation, Univer-
sity of Twente, (2009).
14. Amiet, R. Acoustic Radiation from an Airfoil in a Turbulent Stream, Journal of Sound and Vibration. 41 (4) 407-
420, (1975).
15. Burley, C., Pope, S. APET User Guide. NASA-LaRC, AS&M, Inc, (2014).
16. Brown, D., Garcés, M. Ray Tracing in an Inhomogeneous Atmosphere with Winds. Handbook of Signal Processing
in Acoustics. Springer, New York, NY, (2008).
17. Jonkman, J., Butterfield, S., Musial, W., Scott, G. National Renewable Energy Laboratory, Technical Report
NREL/TP-500-38060, Definition of a 5-MW Reference Wind Turbine for Offshore System Development, (2009).
18. Slawsky L., Zhou L., Roy S., Xia G., Vuille M., and Harris R., Observed Thermal Impacts of Wind Farms Over
Northern Illinois, Sensors, 15 (7), 14981-15005, (2015).