Anda di halaman 1dari 10

APPLICATION OF A RAY THEORY MODEL TO THE PREDICTION OF NOISE EMISSION FROM ISOLATED WIND TURBINES AND WIND PARKS

John Prospathopoulos, Spyros G. Voutsinas National Technical University of Athens, School of Mechanical Engineering, Department of Fluids, 9 Heroon Polytechniou Str., 15773 Zografou, Athens, Greece Tel.: (+)30210-7721098, Fax.: (+)30210-7721056

Summary Various propagation models have been developed to estimate the level of noise near residential areas. Predictions and measurements have proven that the proper modeling of the propagation medium is of particular importance. In the present work, calculations are performed using a ray theory methodology. The ray trajectory and transport equations are derived from the linear acoustics equations for a moving medium in three dimensions. Ground and atmospheric absorption, wave refraction and diffraction, and atmospheric turbulence are taken into account by introducing appropriate coefficients to the equations. In the case of a wind turbine (W/T), it is assumed that noise is produced by a point source located at the rotor centre. Given the sound power spectrum, the noise spectrum at the receiver is obtained by solving the axisymmetric propagation problem. The procedure consists of a) finding the eigenrays, b) calculating the energy losses along the eigenrays, and c) synthesizing the sound pressure level (SPL) by superposing the contributions of the eigenrays. In the case of a wind park, the total SPL is calculated by superposing the contributions of all W/Ts. Application is made to five cases of isolated W/Ts in terrains of varying complexity. In flat or even smooth terrain, the predictions agree well with the measurements. In complex terrain, the predictions can be considered satisfactory, taking into account the assumption of constant wind velocity profile. Application to a wind park shows clearly the significance of the terrain on the wind velocity and consequently on the SPL.

1 Introduction During the last decade, the rapidly growing demand in energy capacity has created the need for exploiting new resources. In this respect, the environmental concerns favour the usage of the renewable energy sources among which wind energy is the most developed and cost effective. Considering the fact that most of the off-shore locations have already been utilized, focus is now concentrated on the mainland, and especially in complex terrains, where unexploited sites with high potential still exist. The siting of W/Ts in such locations raises the issue of noise immission over complex terrain. Most of the

previous research in W/T noise propagation is related to off-shore sites or flat terrain in the mainland, leading to the development of relatively simple models which can produce sensible results in these cases. In 1991, the Danish Ministry of Environment described a simple method [1] for the determination of the SPL at a height of 1.5m above ground. It takes into account the spherical spreading and the air absorption and assumes a constant 3dB decrease due to the interaction with hard ground. A similar method was formulated later by the National Environment Protection Board of Sweden [2], based on the experimental results of a literature survey [3]. This method makes a distinction between ground and water and also between short and long range propagation. In 1993, the Renewable Energy Systems Ltd. made an experimental campaign in England in order to assess the noise emissions at the Carland Cross and Coal Clough wind farms [4]. Four of the existing propagation models were used for the prediction of noise levels and comparison with measurements: The IEA model [5] which is based on hemispherical propagation and includes air absorption; the NORDFORSK model [6] which accounts for geometrical spreading, air absorption, reflection and screening by obstacles, vegetation and ground effect; the CONCAWE model [7] which takes into account geometrical spreading, air absorption, ground effect, meteorological effects and barrier shielding; and finally the ENM model [8] which allows for directivity correction and includes geometrical spreading, terrain discontinuities, air absorption, wind and temperature effects and ground attenuation. Both ENM and CONCAWE models can model wind speed and direction effects unlike the IEA or NORDFORSK model. In addition, ENM can take as input the digitized terrain data of the site. The comparison with the measurements showed that the simple hemispherical model IEA performs equally well with the others for the flat terrain of the Carland Cross wind farm. However, in the complex terrain of the Coal Clough site, the most advanced model ENM performs best whereas the IEA model gave the poorest correlation with measurements. This led to the conclusion that more sophisticated models are required to model noise propagation over complex terrain. In the context of the Noise immission from wind turbines EU project, DELTA developed the WiTuProp sound propagation model [9], which is based on the so-called heuristic model concept [10]. It takes into account the refraction of the rays by the

wind velocity profile, the atmospheric absorption and the ground characteristics. Screening is considered according to Maekawa, modified for curved rays as described in [6]. The model gave good agreement with the measurements, however it is valid only for flat terrain. Nord2000 [11,12] is a more general model, applicable to any kind of ground surfaces. The ground effect is calculated using the theory of Fresnel zones in the case of a non-homogeneous flat surface. For nonflat surfaces, the ground effect is estimated using the valley method: the terrain is approximated by straight segments which are categorized according to their perpendicular distance from the source and the receiver. A different formula of calculation is applied for each category. The effect of refraction in the presence of wind and temperature gradients is considered by modifying the source and receiver heights and applying the straight line propagation method. Van den Berg [13] applied the simple national propagation models of the Netherlands [14] and Germany [15] for the 17 turbine Rhede wind park in Germany, and found a very close agreement between the predicted and measured overall values of the sound immission level at 400m from the closest machine. These models account only for geometrical spreading, air absorption and ground effect. The result reinforced the conclusion that simple propagation models can perform well in flat terrain. In the context of the NOISEPARK EU project [16], NTUA developed the NOISEPRO axisymmetric model [17] for estimating the propagated W/T noise in complex terrain sites. The model is a combination of two methodologies: a normal-modes method, taken from underwater acoustics, for the low frequency part, and a ray tracing approach for the high frequency part. The vertical profile of the mean wind velocity, the range dependency of the medium characteristics and the terrain variation are taken into account. The comparison with the measurements showed good agreement in the high frequency part, however, with the exception of few cases, significant differences were observed in the low frequency part. This led to the conclusion that the ground effect was not predicted satisfactorily by the normal modes method. In this connection, an advanced propagation model, based on ray tracing, has been developed to accurately simulate the ground and air effect. It takes into account the spatial variation of the terrain, wind velocity and temperature in detail [18]. In a previous study [19], this model was validated for flat terrain and used to assess the effect of the propagation medium in complex terrain. In the present paper, implementation is made on the prediction of noise immission from isolated W/Ts and wind parks in terrains of varying complexity.

at the rotor center and propagating in all directions. In order to calculate the noise spectrum at the receiver, the eigenrays are traced by resolving the axisymmetric ray trajectory equations [19,20]. Sound propagation is considered to occur only on the plane which contains the source and the receiver, meaning that the effect of the lateral wind component has been ignored. The contribution of the j eigenray to the SPL at the receiver is given by the relationship: L p , j = L p ,0 Ageo Agr Aair Arefr , (1)

where L p ,0 is the SPL at 1m from the source and Ageo , Agr , Aair , Arefr are the attenuation factors due to geometrical spreading, ground absorption, air absorption and refraction of sound respectively. Considering that only spherical spreading occurs close to the source, the SPL at 1m from the source is related to the sound power level of the source through the relationship [18]:

Lp ,0 = Lw + 10log

cWref , 2 4 pref

(2)

where , c stand for the air density and the sound speed respectively. The reference pressure pref is taken equal to 20Pa for air and the reference power Wref is 10-12 W. The acoustic pressure at the receiver is obtained from the superposition of the contributions of the eigenrays. The contribution of each eigenray to the complex pressure field is p j with modulus p j = 10
L p , j / 20

pref ,

(3)

where L p , j is given by (1) in dB. The phase of p j is calculated through the elapsed time until the ray arrives at the receiver:

j = 2 f t j ,

(4)

with f standing for the frequency of the acoustic source. The total acoustic pressure at the receiver is the sum of the contributions from the N eigenrays:
p = p j cos ( j ) + i sin ( j ) . j =1
N

(5)

In case there is diffraction of sound by an obstacle, the acoustic pressure due to diffraction is added:
p = p j cos ( j ) + i sin ( j ) + pdiff (6) j =1
N

2 Methodology The sound propagation model used for noise predictions is based on ray theory for moving media [19]. In the case of a W/T, it is assumed that noise is produced by a point source of specified power, located

The SPL at the receiver is determined in dB as

L p = 20log

p pref

Aturb ,

(7)

where the attenuation factor Aturb due to the turbulence effect has been included. This factor can only be added last because it is related to the interaction between the eigenrays. In order to calculate the SPL, the attenuation factors in (1), (7) and the pressure pdiff in (6) should be determined. The attenuation factor due to geometrical spreading denotes a logarithmic attenuation with respect to the distance from the source: Ageo = 20log r (dB) Agr (8) is calculated

forest, modified relations should be used as formulated by Attenborough et al.[27]. The air absorption attenuation factor Aair is calculated by integrating the absorption coefficient a along the eigenray:

Aair = 8.69

r = rmax r =0

a dr

(dB).

(12)

The ground attenuation factor

through the spherical reflection coefficient Q as:


Agr = 20log 1 Q

Molecular absorption in the atmosphere has been systematically measured and theoretically predicted. Here, the improved analytical relationship of Bass et al. [28] is used, which estimates a as a function of temperature, relative humidity and atmospheric pressure. The interaction of the sound rays with an inhomogeneous wind velocity field results in their refraction. The attenuation factor due to the refraction is: Arefr = 10log10
2 Pnon refr 2 Prefr

(dB)

(9)

(dB),

(13)

For local reaction conditions, when the penetrating sound is refracted normally to the surface, the spherical reflection coefficient is connected to the plane wave reflection coefficient Rp through the equation [21]: Q = R p + (1 R p ) F , (10)

where Prefr is the amplitude of the sound pressure for a refracted ray and Pnon refr is the amplitude of the sound pressure in the case of a homogeneous medium (rectilinear propagation). The calculation of the pressure ratio is made geometrically, by applying the Blokhintzev invariant for the refracted and the nonrefracted ray tubes. The procedure is described in detail by Lamancusa and Daroux [20]. In the presence of obstacles, the contribution of the diffracted sound, pdiff , to the acoustic pressure field, is estimated using the analytical expressions formulated by Pierce [29] for an edge or a three-sided barrier. In this sense, an obstacle of arbitrary shape is substituted by an equivalent three-sided barrier. The sound pressure due to the diffraction is obtained as a function of the geometrical characteristics of that equivalent barrier. The interaction of sound waves with the atmospheric turbulence may cause fluctuations in their phase and amplitude which eventually affect significantly the mean SPL. According to the concept described by Daigle et al [30], the mean time-averaged value of the squared turbulent sound pressure is:
2 pturb

with the boundary loss coefficient F being a function of the distance, the ground impedance and the angle of incidence [22,23]. When the angle of incidence is greater than 5o, then F 0 and Q R p , meaning that the orbicularity of the wavefront need not be taken into account. The plane wave reflection coefficient is determined from the relationship [24]:

Rp =

sin 1/ , sin + 1/

(11)

where stands for the angle of incidence and Z for the complex ground impedance. Different approximations have been formulated for the ground impedance which is a function of the microscopic ground properties. Most commonly used is the singleparameter approximation formulated by Chessell [24] and Nicolas et al. [25], which expresses the impedance in terms of the specific flow resistivity of the ground. In several cases, the simulation of the ground as a single semi-infinite layer is not appropriate; a thin layer above semi-infinite hard material or multiple layers of absorbent materials constitutes a better approximation. In such cases, an equivalent impedance is inserted in (11), calculated from the impedances of the separate materials (see [26] for details). For extended reaction conditions, occurring for low values of the specific flow resistivity (<300000 kg/sm3), like a snow layer or a moisture layer in the

p
N i 1

2 0

= 1 + A2 Ci2 +
i =1

Q 2 CiC j cos 2 f ( j i ) + Arg j ij i = 2 j =1 Qi

, (14)

where p0 is the pressure 1m away from the source and Ci = ai Qi / ri ; Qi , Qj are the spherical reflection coefficients, i , j are the absorption coefficients and ri , rj are the total distances covered

by the rays i , j respectively; A2

is the variance of

the amplitude fluctuations and ij is the coherence coefficient between rays i and j . The coherence coefficient ij takes into account the phase fluctuations and the interaction between the rays i , j . Both phase and amplitude fluctuations are calculated in terms of the turbulent index of refraction which is a function of the measured temperature and velocity variances. The attenuation factor Aturb due to turbulence is obtained as Aturb = 10log10
2 pnon turb 2 pturb

noise) is compared to the measurements at 100 and 250m horizontal distance from the loudspeaker. The agreement can be considered satisfactory.

(15)

where pnon turb is the acoustic pressure field in the absence of turbulence given by (14) for a 2 = 0 and Figure 1: Terrain contours of the Carland Cross site
60

each W/T is calculated by considering the axisymmetric problem of sound propagation from the rotor centre (point source) to the receiver. For each of the N S axisymmetric problems considered, the wind velocity is projected in the direction of propagation and the normal component of the velocity is ignored (it does not affect propagation). The total SPL is calculated by superposing all contributions:
L p ,tot = 10log

Sound pressure level (dB)

ij = 1 . In the case of a wind park, the contribution L p , m of

50 40 30 20 10 0 100
2 4 6 8

Predictions Measurements Background noise 1000


2 4 6 8

10
m =1

NS

Frequency (Hz)

L p ,m /10

(16)
Sound pressure level (dB)

(a)
60 50 40 30 20 10 0 100
2 4 6 8

3 Results

The method is applied to five cases of noise propagation from isolated W/Ts in terrains of varying complexity. Predictions are compared with the available experimental data in each case. The effect of the complexity of the terrain on noise propagation is also investigated in the case of a wind park. Case 1. Carland Cross in England is a fairly smooth terrain with simple topographic features as shown in Fig. 1. During the experimental campaign [4], the Vestas Windane 34 W/T noise was simulated by a loudspeaker raised to 29m above ground (it is noted by TB). Microphones were located at various positions around the source, mounted on 1.2m tripods. In order to apply the propagation model, ground is simulated as a semi-infinite layer with 105 kg/sm3 specific flow resistivity, indicative of a typical grassland. Atmospheric conditions are assumed neutral with 5oC temperature and 90% relative humidity. Predictions are compared to the measurements taken along the line TB-A on 20 January 1992. The wind direction was 110o, almost perpendicular to the direction of propagation. In Fig. 2, the predicted SPL spectrum (including background

Predictions Measurements Background noise 1000


2 4 6 8

Frequency (Hz)

(b) Figure 2: Carland Cross - SPL at distance (a) 100m and (b) 250m from the source (line TB-A) In Fig. 3, the predicted overall noise level is compared with the measurements and the predictions of the IEA, ENM, CONCAWE (CWE) and NORDFORSK (ND) models. The predictions of the present model are in close agreement with the experimental data, whereas the simple models IEA and NORDFORSK give also satisfactory results, probably better than the more advanced ENM and CONCAWE models. This could be a matter of the assumptions made for the atmospheric and the terrain input data.

70 60 50 40 30 20 10 0 50 100 Predictions ENM CWE ND Measurements 150 200 250 300 350

Overall noise level (dB)

variation with height; thus, wind velocity is modelled with a neutral logarithmic or exponential profile fitted to the experimental data. The mean value of roughness length was estimated 0.10m and the flow resistivity of the ground, used for the calculation of impedance, is taken equal to 5105 kg/sm3.
10 5

Excess attenuation (dB)

0 -5 -10 -15 -20 -25 100


2 4 6 8

Horizontal distance (m)

Figure 3: Carland Cross - Variation of the overall SPL with distance (line TB-A)
Case 2. Tammhausen site in Lower Saxony, Germany, is characterized by flat terrain, covered by wet grass. The relative experiment was performed in the context of the Noise Immission from Wind Turbines project [9] and concerns the noise propagation from a MONOPTEROS 50 W/T with 60m hub height and 56m diameter [31]. In cases of noise propagation from W/Ts, the SPL is often measured at two points - one near the turbine (about 1.5 rotor diameters from the center of the rotor) called the emission point, and another at the receiver called the immission point. In this case, the emission point lied on the ground, 80m far from the W/T base, whereas the immission point was at 1.5m height, 530m far from the W/T base. The excess attenuation (E.A) is estimated through the relationship [19]:

Predictions, = 3 10-5 Predictions, = 2 10 Predictions, = 10 Measurements 1000


2 4 6

Frequency (Hz)

(a)
10 5

= 2 10

-5

Excess attenuation (dB)

0 -5 -10 -15 -20 -25 100


2 4 6 8

Predictions, refraction loss not included Predictions, refraction loss included Measurements

Frequency (Hz)

1000

r E. A. = Lp , e Lp , i + 20log e , ri

(17)

(b) Figure 4: Tammhausen - Excess attenuation at 530m distance for downwind propagation. (a) Effect of refraction and (b) Effect of turbulence

where L p , e and Lp ,i are the measured SPLs at the emission and the immission points respectively; re , ri denote the corresponding distances from the center of the rotor. In the present simulation, the ground is considered as a semi-infinite layer with 3105 kg/sm3 flow resistivity. The vertical profiles of the velocity and temperature are calculated according to similarity theory [32] using the available measurements and the Nieuwstadt method [33]. In Fig.4, the predictions are compared with the measurements [31] for downwind propagation conditions. Agreement is satisfactory and becomes even better by properly adjusting the turbulence level through the turbulent index of refraction (Fig.4a). Furthermore, the predictions improve by taking into account the refraction loss (Fig.4b). Case 3. Toplou site in Crete (Greece) exhibits a particularly complex terrain with many hills and slopes varying from 3 to 20o (Fig.5). The experiment was performed in the context of the Noise Immission from Wind Turbines project [9] and concerns the noise propagation from a Tacke 500kW W/T with 35m hub height and 36m diameter [34]. Measurements of temperature showed little



 

   

 

Figure 5: The terrain of the Toplou site. The emission measurement microphones (1, 3, 7) were placed at ground level in a horizontal

distance between 48 and 64m from the W/T. The immission measurements microphones (2, 4, 8) were positioned at 1.5m height above ground at a horizontal distance between 230 and 270m from the W/T. In Fig.6, the excess attenuation between the 12 and 3-4 points is compared with the measurements [34]. The comparison is satisfactory for frequencies higher than 200Hz, however, in the low frequency part, the experiment [34] shows a remarkable amplification of noise which is not reproduced by the predictions. The effect of the turbulent index of refraction on the maxima of the predicted spectrum is also depicted in the same figures.
10 5 0 -5 -10 -15 -20
4 6 8

100

80

60

40

20

0 0 20 40 60 80 100 120 140 160 180 200 220 240 260

(a)
10 5 0 -5 -10 -15 -20
4 6 8

Excess attenuation (dB)

Excess attenuation (dB)

Predictions, = 10 -5 Predictions, = 3 10 Predictions, = 5 10 Measurements 100


2 4 6 8

Predictions, refraction loss included Predictions, refraction loss not included Measurements 100
2 4 6 8

Frequency (Hz)

1000

Frequency (Hz)

1000

(b) Figure 7: Toplou (a) Eigenray diagram and (b) excess attenuation between 7 and 8 points. Effect of refraction.
15 10

(a)
10 5 0 -5

Excess attenuation (dB)

Excess attenuation (dB)

-10 -15 -20


4 6 8

5 0 -5 -10 -15 -20


4 6 8

Predictions, = 5 10 Predictions, = 10 Predictions, = 3 10-5 Measurements 100


2 4 6 8

-6

Frequency (Hz)

1000

(b) Figure 6: Toplou Excess attenuation between points (a) 1 and 2 and (b) 3 and 4. Effect of turbulence. In downwind propagation conditions, the wind velocity profile combined with the complexity of the terrain, may increase the number of the eigenrays reaching the receiver (Fig.7a). In this case, the inclusion of the refraction loss in the calculations can significantly improve the predicted spectrum of excess attenuation as observed in Fig.7b. Case 4. Lyse site is located to the west coast of Sweden and exhibits a rocky terrain. The experiment was conducted on a small peninsula, 500m far from the shore, where a NWP 400 W/T with 40m hub height and 35m diameter is installed. All emission and immission points are located on that peninsula. The experimental configuration and the meteorological data are described in detail in [35], written in the context of the Noise Immission from Wind Turbines project [9].

Predictions, = 5 10 Predictions, = 10 -6 Predictions, = 10 Measurements 100


2 4 6 8

Frequency (Hz)

1000

(a)
10 5 0 -5 -10 -15 -20
4 6 8

Excess attenuation (dB)

Predictions, = 5 10 Predictions, = 10 -6 Predictions, = 10 Measurements 100


2 4 6 8

Frequency (Hz)

1000

(b) Figure 8: Lyse Excess attenuation at (a) point 1 and (b) point 2. Effect of turbulence.

The emission measurement microphones were placed on a board at ground level, at a distance of 85m from the W/T. The receivers at the immission points were free microphones, elevated at 1.35m above ground level. For the present simulation, wind velocity profiles and molecular absorption are estimated through the meteorological data reported in [35]. On account of the fact that the measurement points are very close to the sea, it is not clear whether the reflection of the sound waves occurs on the sea or on the ground. Consequently, there is an uncertainty about the value of the flow resistivity to be used for the simulation of the ground impedance. In Fig. 8, the predicted excess attenuation spectrum is presented for the two immission points 1 and 2. In the first case, the flow resistivity has been set to 106 kg/sm3. For frequencies higher than 100Hz, the comparison with the measurements is satisfactory, especially when the turbulent index of refraction is adjusted to 10-5. For the low frequencies, under 100Hz, the experiment presents particularly amplified noise levels, probably due to some local weather conditions which favour the propagation of lowfrequency noise. In the second case, the flow resistivity has been set to 5105 kg/sm3. The comparison is satisfactory for the whole frequency range. Again, the maximum excess attenuation is best predicted when = 105 .
40

Case 5. The experiment in Medicine Bow, Wyoming, was conducted by Willshire, and concerns the propagation of low frequency noise from a WTS-4 downwind W/T with 80m hub height and 80m rotor diameter. The results of the experiment are analyzed in Hubbard et al [36]. On account of the fact that the low frequency noise comes from the passage of the blade tip through the turbulent velocity deficit of the tower, the source height is considered equal to 40m [37]. Microphones were located on the ground at zero height. The terrain is flat with 0.1m roughness length and the friction velocity was estimated at 0.64m/s. The available acoustic data refer to a 10Hz frequency and large distances from the W/T (up to 10km). Ground is simulated as a semi-infinite layer with 106 kg/sm3 flow resistivity. A logarithmic wind velocity profile (neutral atmospheric conditions) is used, calculated via the estimated values of the roughness length and the friction velocity. Measurements are reported in the form of transmission loss (TL) which can be written as:
TL = 20log 10
j =1 N A j / 20

(18)

where Aj denotes the total attenuation in dB along the j eigenray. In Fig.9a, the predictions of the ray-tracing model are compared with the measurements. The spherical spreading law dominates up to 2km (two eigenrays). At that distance, two new eigenrays appear, significantly increasing the noise level, which is verified by the measured data. The number of eigenrays remains constant up to about 4.8km where two new eigenrays are added resulting in a similar increase in noise. This is also in agreement with the experiment. However, for distances larger than 6km, the numerical error of the method is evident leading to inaccurate eigenray tracing. Such errors relate to the accuracy of the integration technique and the interpolations used for the velocity calculations. They increase with integration time and become significant for large distances irrespectively of the choice of the numerical parameters. The computational time required also increases rapidly for large distances. In view of the above, a modification of the numerical model was implemented based on the definition of cells which surround the calculation points. The SPL at each point is given by the superposition of all rays that pass through the surrounding cell. The reduction of computational time is substantial because only one launch of the ray beam is required. In Fig.9b, the results of the modified model are presented. The spherical spreading and the abrupt increase at 2km are well represented. For larger distances, the predictions present fluctuations, due to the changing number of eigenrays that pass through the numerical cells. It should be noted, though, that the mean value of the predictions follows the measurements.

Transmission loss (dB)

50

Predictions Measurements

60

70

80

90 0 2000

Horizontal distance (m)

4000

6000

8000

10000

(a)
40

Transmission loss (dB)

50

Predictions Measurements

60

70

80

90 0 2000

Horizontal distance (m)

4000

6000

8000

10000

(b) Figure 9: Medicine Bow Prediction of TL with distance using (a) eigenray tracing and (b) numerical cell averaging.

Wind park case. The Kalivari site is located to the north part of the Andros island, in Greece. It exhibits a fairly complex terrain with altitudes varying from 0 up to 300m (Fig.10). Because of the sparse vegetation, roughness length is estimated 3cm. The wind park consists of 7 VESTAS 27 W/Ts with rotor diameter 32m marked with dots in Fig.10. In order to investigate the effect of the terrain, the SPL is calculated at three positions, P1, P2 and P3 marked with squares on the contour map.

1200

1000

800

y (m)

7 6
600

5 4 3 2 1

400

200

The velocity field is calculated by resolving the full 3D Navier-Stokes equations [18]. A first resolution is made for the broader domain of the north part of the island using a coarse grid and then a successive focusing procedure is followed to finely estimate the velocity field over the area of the wind park installation. To calculate the noise spectrum at P1, P2, P3, the axisymmetric propagation problem is solved for each W/T; then all W/T contributions are superposed at each point according to (16). The flow acceleration caused from the irregular relief of the site, affects significantly the eigenray tracing and therefore the SPL predicted. In Fig.11, the eigenrays for three of the separate axisymmetric problems are depicted. The steep velocity gradients cause a strong refraction of the rays, resulting in the existence of eigenrays even in cases that the receiver point is not visible from the source (Fig.11a, 11b). They also cause the existence of multiple eigenrays, creating a sound channel close to the ground (Fig.11a,b,c). The SPL spectra predicted are represented in Fig.12. The overall noise levels are 53.79, 53.76 and 47.70 dB respectively. The reduced level at P3 point is due to its longer distance from the nearest W/Ts and to the existence of a hill in front of the 4,5 and 6 W/Ts (see Fig.10).
50

200

400

600

800

1000

1200

x (m)

Figure 10: Terrain contours of the Kalivari site.


Sound pressure level (dB)
60 50 40

40

30

z (m)

30 20 10 0 0 100 200 300 400 500

20

W/T spectrum (53 m)

10

1 point 2 point 3 point


100 1000 10000

x (m)

(a) W/T 6 P1 point


60 50 40

Frequency (Hz)

Figure 12: Kalivari SPL at P1, P2 and P3 points.

z (m)

30 20 10 0 0 100 200 300 400 500

Conclusions

x (m)

(b) W/T 1 P3 point


50 40

z (m)

30 20 10 0 0 100 200

x (m)

300

400

500

(c) W/T 7 P2 point Figure 11: Kalivari Eigenray diagrams

An advanced ray tracing propagation model has been used for the prediction of noise emission from wind parks. Application was made to five cases of isolated W/Ts in terrains of varying complexity and one wind park in complex terrain. In cases of flat or smooth terrain, like the Carland Cross site in England or the Tammhausen site in Germany, the predictions were in good agreement with the measurements, especially when the turbulence level was properly adjusted. Simple models, like IEA and NORDFORSK, gave also satisfactory results for the overall SPL in the Carland Cross case. In cases of complex terrain, like the Toplou site in Greece, simple propagation models are no longer valid. The same applies for the case of Lyse site in

Sweden, although the W/T is located to a flat peninsula surrounded by sea, because the broader area exhibits a rocky complex terrain which shapes the boundary layer of wind velocity. In both cases, the predictions of the present model can be considered satisfactory, given the fact that the spatial variation of the wind velocity profile was ignored due to lack of sufficient data for the resolution of the flow field. Some of the differences between predictions and measurements can be attributed to the assumption of constant wind velocity profile. In the low frequency part, at both sites, free microphone measurements showed a significant amplification of noise which was not observed in the predictions. Two possible explanations are: a) the presence of topographic or weather local effects and b) the inability of the ray tracing model to perform equally well in the low frequencies. Possibility (a) is reinforced by the fact that in the Lyse site, a relevant amplification phenomenon did not appear in the second set of measurements taken by the vertical board technique for the same conditions [35]. The SPL spectrum prediction was improved by properly adjusting the turbulence level and including the refraction loss calculation. Application to the flat terrain case of Medicine Bow in Wyoming proves that the ray tracing model performs well even in the low frequency of 10Hz, thus reducing the (b) possibility above. However, for long distances (>6km), the decreasing accuracy of the ray tracing method and the substantial increase of its computational cost indicate the need of using a cellaveraging procedure which provides more satisfactory mean results. Finally, the application of the propagation model to the wind park of Kalivari in Greece, verifies the significant effect of the complexity of the terrain on noise propagation. The flow acceleration caused from the relief of the site, changes the gradient of the boundary layer, bearing an increase in the number of eigenrays and therefore in SPL.

References

1. Danish Environmental Protection Agency. Noise from W/Ts; Statutory order No.304, 14th May 1991 (In Danish). 2. Naturvrdsverket. Ljud frn vindkraftverk (Sound from wind power station); ISBN 91-620-6249-7, 2001 (In Swedish). 3. Ljunggren S. Ljudutbredning kring havsbaserede vindkraftverk. Resultat frn en litteraturstudie (Sound propagation around a sea-based wind power station. Result from a literature study); Department of civil and architectural engineering KTH, Stockholm, AR 1996:6. (in Swedish) 4. Bass J. Noise assessment at Coal Clough wind farm site; ETSU W/13/00354/037 Report; Renewable Energy Systems Ltd., 1994. 5. Expert Group Study on Recommended practices for W/Ts and evaluation. 4. Acoustic measurement of noise emission from W/Ts; Submitted to the Executive Committee of the International Energy Agency Programme for Research and Development on Wind Energy Conversion Systems; 3rd ed., 1994. 6. Kragh J, Andersen B, Jakobsen J. Environmental noise from industrial plants. General prediction method; Lydteknisk Laboratorium Report No. 32, Denmark, 1982. 7. Marsh KJ. The CONCAWE model for calculating the propagation of noise from open air industrial plants. Applied Acoustics 1982; 15:411-428. 8. Tonin R. Estimating noise levels from petrochemical plants, mines and industrial complexes. Acoustics Australia 1985; 13(2): 59-67. 9. Kragh J. et al. Noise immission from wind turbines; Final Report of the EU funded project JOR3_CT960065; DELTA Acoustics & Vibration, AV 1509/98, 1998. 10. LEsperance A, Herzog P, Daigle GA, and Nicolas JR. Heuristic model for outdoor sound propagation based on an extension of the geometrical ray theory in the case of a linear sound speed profile. Appl. Acoust. 1992; 37: 111-139. 11. Plovsing B, Kragh J. Nord 2000. Comprehensive outdoor sound propagation model. Part 1: Propagation in an atmosphere without significant refraction; DELTA Acoustics & Vibration Report AV 1849/00, 2001. 12. Plovsing B, Kragh J. Nord 2000. Comprehensive outdoor sound propagation model. Part 2: Propagation in an atmosphere with refraction; DELTA Acoustics & Vibration Report AV 1851/00, 2001.

Acknowledgments

This work was partially financed by the European Union in the context of the JOR3_CT96-0098 project entitled as Investigation of noise emissions from wind parks and their impact to the design of parks (NOISEPARK). The authors are grateful to the Research Institutes of DEWI, CRES and KTH for providing the measurements carried out in the context of the JOR3_CT96-0065 EU funded project (Noise Immission from Wind Turbines).

13. Van den Berg GP. Effects of the wind profile at night on W/T sound. Journal of Sound and Vibration 2004, 277: 955-970. 14. Handleiding meten en rekenen industrielwaai (Manual for measuring and calculating industrial noise), Den Haag, 1999 (in Dutch). 15. TA-Lrm: Technische Anleitung zum Schutz gegen Lrm (Technical guideline for noise protection), 1998 (in German). 16. Voutsinas S. Investigation of noise emissions from wind parks and their impact to the design of parks; Final Report of the EU funded project JOR3_CT960098; National Technical University of Athens, Greece, 1998. 17. Pothou K, Voutsinas S, Huberson S, Kuhlmann M, Rawlinson-Smith R. Investigation of noise emissions from wind parks and their impact to the design of parks by means of the NOISEPARK software. Proceedings of the European Wind Energy Conference 1999; E.L. Petersen et al, eds., James & James Ltd., London, UK, pp.97-100. 18. Prospathopoulos JM. Propagation of acoustic fluctuations in the atmospheric environment and application to the prediction of noise from W/Ts; Ph.D. thesis, National Technical University of Athens, Greece, 2002. 19. Prospathopoulos JM, Voutsinas SG. Noise propagation issues in wind energy applications. Journal of Solar Energy Engineering 2005; 127:234241. 20. Lamancusa JS, Daroux PA. Ray tracing in a moving medium with two-dimensional sound-speed variation and application to sound propagation over terrain discontinuities. Journal of the Acoustical Society of America 1993; 93(4): 1716-1726. 21. Attenborough K. Review of ground effects on outdoor sound propagation from continuous broadband sources. Applied Acoustics 1988; 24: 289319. 22. Chien CF, Shoroka WW. Sound propagation along an impedance plane. Journal of Sound and Vibration 1975; 43; 9-20. 23. Ingard U. On the reflection of a spherical wave from an infinite plane. Journal of the Acoustical Society of America 1951; 23: 329-335. 24. Chessell CI. Propagation of noise along a finite impedance boundary. Journal of the Acoustical Society of America 1977; 62(4): pp.825-834. 25. Nicolas J, Berry JL, Daigle GA. Propagation of sound above a finite layer of snow. Journal of the Acoustical Society of America 1985;.77(1): pp.67-73 .

26. Attenborough K. Ground parameter information for propagation modeling. Journal of the Acoustical Society of America 1992; 92(1): 418-427. 27. Attenborough K, Hayek SI, Lawther M. Propagation of sound above a porous half-space, Journal of the Acoustical Society of America 1980; 68(5): pp.1493-1501. 28. Bass HE, Sutherland LC, Zuckerwar AJ, Blackstock DT, Hester DM. Atmospheric absorption of sound: Further developments. Journal of the Acoustical Society of America 1995; 97(1): pp.680683. 29. Pierce AD. Diffraction of sound around corners and over wide barriers. Journal of the Acoustical Society of America 1974; 55(5): 941-955. 30. Daigle GA, Piercy JE, Embleton TFW. Effects of atmospheric turbulence on the interface of sound waves near a hard boundary. Journal of the Acoustical Society of America 1978; 64(2): pp.622-630. 31. Osten T, Klug H. Noise propagation measurements in the vicinity of W/Ts; Report AS 971214, Deutcshes Windenergie, Wilhelmshaven, Germany, 1997. 32. Businger JA, Wyngaard JC, Izumi Y, Bradley EF. Flux-profile relationships in the atmospheric surface layer, Journal of the Atmospheric Sciences 1971; 28:181-189. 33. Nieuwstadt F. The computation of the friction velocity u* and the temperature scale T* from temperature and wind velocity profiles by least-square methods, Boundary Layer Meteorology1978; 14: 235246. 34. Theofiloyiannakos D. Noise immission from wind turbines; Report CRES.WE.NIWT.06, Centre for Renewable Energy Sources, Pikermi, Greece, 1998. 35. Fegeant O. Noise propagation around wind turbines: Second set of measurements at the Lyse wind power station; Report ISSN 0349-6562, KTH, Sweden, 1997. 36. Hubbard HH, Shepherd KE, Willshire WL. Results of simultaneous acoustic measurements around a large horizontal axis wind turbine. Journal of the Acoustical Society of America 1986; 80: S28(A). 37. Hawkins JA. Application of ray theory to propagation of low frequency noise from wind turbines; Report 178367, NASA, Langley Research Center, Hampton, Virginia, 1987.

Anda mungkin juga menyukai