THE INFLUENCE OF MICROPHONE HEIGHT IN THE MEA- SUREMENT OF LOW FREQUENCY AND INFRASONIC WIND TURBINE NOISE AT TYPICAL RECEIVERS

Transcription

1 23 rd International Congress on Sound & Vibration Athens, Greece July 2016 ICSV23 THE INFLUENCE OF MICROPHONE HEIGHT IN THE MEA- SUREMENT OF LOW FREQUENCY AND INFRASONIC WIND TURBINE NOISE AT TYPICAL RECEIVERS Kristy L. Hansen Flinders University, Adelaide 5042, Australia Branko Zajamšek University of New South Wales, Sydney 2052, Australia Colin H. Hansen University of Adelaide, Adelaide 5005, Australia For environmental noise measurements, the most common height chosen for the outdoor measurement microphone is 1.5 m, as this corresponds approximately to the ear height of an average receiver. The main disadvantage of measuring at this height is that wind-induced noise can affect the results, particularly at low frequencies. Therefore, it is advantageous to measure at ground level or below ground, where the wind speed approaches zero and the wind-induced noise is minimised. On the other hand, results obtained from any height that does not represent the human receiver should be interpreted with caution, particularly at low frequencies when large propagation distances are involved. The reason for this is that the difference in noise level at various heights above the ground is a function of the frequency, ground surface properties, distance from the source and the height of the source. Therefore, it may not be valid to assume that the noise level at ground level and at a height of 1.5 metres is equivalent for low frequencies with corresponding large wavelengths. Also, the noise level difference between these heights at mid- and high-frequencies may not be exactly 3 db, as it will depend on the ground impedance. Therefore, further research in this area is necessary before measurement at ground level or underground becomes standardised. This paper presents results of low-frequency measurements taken at two distances from an operating wind farm, where microphones were mounted at a height of 1.5 m and at ground level and protected using secondary wind screens of the same diameter and material. The low-frequency results are compared to a theoretical model that incorporates the parameters described above and similarities and differences are explained. 1. Introduction Wind turbine noise contains a significant amount of low frequency noise (<200 Hz) (1) and during propagation, noise at mid- to high-frequencies is selectively attenuated due to air and ground absorption. Therefore, at a typical residence located more than km from the nearest turbine in a wind farm, the noise due to wind farm operation is dominated by low frequencies. Low frequency noise has the propensity to be annoying, particularly in cases where the noise spectrum is unbalanced (2). Therefore, it is important to accurately characterise the wind farm noise at a typical residence, and this requires that wind-induced noise is kept to a minimum. Wind-induced noise is caused by pressure fluctuations at the microphone that originate primarily from atmospheric turbulence but also from vortex shedding behind the microphone and associated 1

2 accessories. One way to minimise wind-induced noise is to use a relatively large, secondary wind screen, providing a relatively large surface area over which incoherent wind noise is averaged (3). The volume of air between the primary and secondary wind screens in such a design also provides a region for viscous dissipation to occur before the turbulence reaches the microphone (4). The effectiveness of a secondary wind screen increases with size (5); however, for measurements of wind turbine noise, issues of practicality must be considered, since instrumentation needs to be transported to and from various measurement sites. Therefore, an additional method of reducing wind-induced noise is to locate the microphone in close proximity to the ground where the wind speed and turbulent fluctuations are minimised. An additional advantage of ground-mounting is that vortex shedding about the wind screen is reduced, resulting in a further reduction of wind-induced noise. There are several examples of secondary wind screens that have been designed to be mounted in close proximity to the ground and that have been developed for use in measurements of wind farm noise. These include the hemispherical secondary wind screen with a one metre diameter ground board, described in (6), the dodecahedral secondary wind screen described by (7) and the underground box developed by (8). According to these three designs, the microphone is located on the ground, 20 cm above the ground and under the ground, respectively. When the noise source is significantly above the surface of the ground, such as in the case of a wind turbine, noise can reach the receiver via a number of propagation paths. At the receiver, the incident sound waves at each frequency arrive with different phase angles depending on the number of ground reflections that have occurred. In a downward refracting atmosphere, it is possible that more than one ground-reflected ray will arrive at the receiving microphone. Relative differences in phase of the incoming sound waves are caused by different path-length differences between the direct and ground-reflected waves and phase changes that occur when sound waves are reflected at the ground surface. The direct and reflected waves from a given source can either reinforce or interfere, respectively increasing or decreasing the sound level, depending on their relative phase. This paper examines the results of measurements taken at two distances from an operating wind farm, where microphones were mounted at a height of 1.5 m and at ground level, and protected using secondary wind screens of the same diameter and material. The results are compared to a theoretical model that takes into account the path-length differences between the direct wave and a ground reflected wave. The contribution of waves that have undergone more than one reflection and the effects of phase change due to ground impedance are not taken into account in the theoretical analysis. 2. Methodology 2.1 Field measurements Outdoor measurements were carried out at two dwellings that are located near the Waterloo wind farm, in South Australia. The wind farm is made up of 37 wind turbines each of rated power 3 MW, 90 m hub height and 90 m rotor plane diameter. The wind farm is located along the centreline of a ridge and therefore, the average height of the wind turbine base is approximately 200 m above the measurement locations. Apart from the ridge, which is approximately 500 m wide, the topography between the wind farm and the two receivers is flat. The dwellings are referred to as measurement locations H1 and H2 respectively, and are indicated in Fig. 1. Locations H1 and H2 are situated 2.7 km and 3.3 km, respectively, from the nearest wind turbine. This turbine is located at 161 m and 154 m above H1 and H2, respectively. For location H1, the measurement period was from 8/05/2013 to 13/05/2013 (autumn) and for location H2 from 24/07/2013 to 1/08/2013 (winter). 2 ICSV23, Athens (Greece), July 2016

3 35 m 1.5 m The 23 rd International Congress of Sound and Vibration Wind turbines H2 ( , ) Wind turbines 3.3 km ( , ) ( , ) 2.7 km N H1 ( , ) Figure 1: Measurement locations with respect to the wind farm. 2.2 Instrumentation At all locations, acoustic measurements were performed outdoors using two different wind screens, as shown in Fig. 2. At measurement location H1, the microphones at ground level and 1.5 m height were separated by approximately 3 m and at location H2, wind screens were approximately 20 m apart. At location H1, the data were recorded using a National Instrument 9234, 24-bit data acquisition card with a sampling frequency of Hz and a G.R.A.S. type 40 AZ 1/2 inch microphone with a 20 CG G.R.A.S. pre-amplifier. This microphone has a flat frequency response from 0.5 Hz to 20 khz and noise floor of 17 dba. At location H2, Brüel and Kjær (B&K) LAN-XI data acquisition hardware was used with B&K Pulse software to aquire the time series data. The microphones were B&K type 4955 with a noise floor of 6.5 dba and minimum frequency (ensuring measurements were within 10% of flat response) of 6 Hz. Microphones were placed more than 20 m away from the residence and more than 10 m away from surrounding vegetation in order to minimise sound reflections from the façade and vegetation noise, respectively, as shown in Fig. 2. The measurement set-up at location H2 was similar to the one shown in Fig. 2 for location H1 and is thus not described here. side view microphone hemispherical windshield side view steel post tree aluminium plate 1.5 m weather station 15 m = 1000mm wind farm N = 450mm microphone sphericalwindshield 10 m 25 m house Figure 2: Outdoor measurement set-up at location H1, where φ is the wind screen diameter. To protect the microphones from wind-induced noise, primary 90 mm wind screens and two types of secondary wind screens, namely, a sphere and a hemisphere, were used. The hemispherical wind screen consisted of a 16 mm layer of acoustic foam covered by a layer of SoundMaster acoustic fur, supported by a 450 mm diameter steel frame of hemispherical shape, made out of thin-wire steel. The wind screen was attached concentrically to a 1 m diameter aluminium plate of 3 mm thickness and a microphone with a primary wind screen was placed in the centre of the aluminium plate. The ICSV23, Athens (Greece), July

4 hemispherical design follows the specifications in IEC standard (6). The spherical wind screen of a similar design was mounted at a height of 1.5 m on a star-dropper, as shown in Fig. 2. A microphone with a primary wind screen was placed in the middle of the sphere of 450 mm outer diameter. The wind speed at heights of 1.5 m and 10 m was measured near the microphones using Davis Vantage Vue weather stations, and was averaged over 10 minute time intervals. The wind speed accuracy of this weather station is of the order of ± 0.4 m/s. Hub height wind speed data were available from the manufacturer for the measurements at H1 but not at H2. Therefore for H2, the wind speed at hub height was measured using a SODAR that was positioned on the same ridge-top as the wind turbines. This device has a resolution of ± 0.01 m/s, according to the manufacturer. Waterloo wind farm output power information was obtained from the Australian Energy Market Operator website (9) in 5 minute averages. 2.3 Data selection criteria All acoustic and wind data were averaged over 10 minute long time blocks for which the wind farm power output was >40% of its maximum, the wind speed at 1.5 m height was less than or equal to 0.5 m/s and the residence was downwind (±45 ) from the nearest wind turbine. These two criteria were imposed in order to guarantee that wind farm noise was present and to minimise wind-induced noise on the microphones and wind noise caused by movement of the wind through the surrounding vegetation. The number of measurements fulfilling these two criteria was 15 at location H1 and 10 at location H2. The average wind farm power output at location H1 was 45% and at H2 it was 57%. The absence of extraneous noise was verified through listening to the audio files, viewing the data in the time domain and looking into the spectral characteristics of the measured signals. 3. Model description To model the difference in noise level as a function of frequency for the microphones mounted on the ground and at a height of 1.5 m, it is assumed that the direct and reflected waves arrive in phase whereas for a microphone mounted at a height of 1.5 m, the direct and reflected waves do not arrive in phase due to differences in the path length travelled by the sound waves. Therefore, the noise level measured at a height of 1.5 m can be different to the one measured on the ground and the difference is frequency dependent. In the following analysis, it is assumed that the ground is a solid reflector and therefore the phase shift that occurs during each reflection is negligible. This assumption is only valid for low frequencies and therefore the analysis is restricted to frequencies below 100 Hz. The incorporation of ground effects into the analysis is the subject of future work. It is important to determine the number of reflected waves, n gr, that arrive at the microphone mounted at a height of 1.5 m and this can be calculated approximately using (10) N gr 8h max,1 h S + h R (1) where h S and h R are the source and receiver heights and the maximum height, h max,n, above the source reached by the ray is h max,n = D n Bm 2πc 0 (2) where D is the horizontal distance between the source and the point at which a sound wave reaches the equivalent height of the source after reaching its maximum height, c 0 is the speed of sound in air = 343 m/s and B m is a coefficient that is related to the wind velocity profile. The value of D was found using the procedure outlined in (10), which involves calculating the radius of curvature of the 4 ICSV23, Athens (Greece), July 2016

5 sound waves, the centre of the associated circular arc and the distance between the horizontal plane at source height and the centre of the circular arc, using geometrical relationships. The coefficient B m was determined using the hub height wind speed data and Equation 3. The wind speed measured at heights of 1.5 m and 10 m as well as at hub height at both residences is shown in Figure 3 and it can be seen that the best curve fit to the average data is approximately linear. B m = ( log h0 e z ) (3) where U 0 is the wind velocity at reference height, h 0 and z 0 is the roughness length. The latter was assumed to be 0.04 and 0.1 for houses H1 and H2, respectively, based on the surrounding farm crops whose height, and hence roughness length, is a function of the time of year. The resulting values of B m were 1.03 and 1.36 for houses H1 and H2, respectively. U 0 Height relative to residence (m) All wind speeds Avg. wind speed Linear fit Wind Speed (m/s) (a) Location H1 Height relative to residence (m) All wind speeds Avg. wind speed Linear fit Wind Speed (m/s) (b) Location H2 Figure 3: Wind speed at heights of 1.5 m, 10 m and hub height and linear fit to data It was found that a single reflection occurred at H1, whereas three reflections were predicted for H2. The location of the reflection point closest to the receiver, d G, was determined by solving the following cubic equation (11) 2d 3 G 3dd 2 G + (b 2 R + b 2 S + d 2 )d G b 2 Sd = 0 (4) where d is the horizontal source/receiver separation distance, and b 2 S = h S ξ n (2 + ξ n h S ) (5) b 2 R = h R ξ n (2 + ξ n h R ) (6) ξ n = c(h R) c(h S ) c 0 (h R h S ) where the sound speeds, c(h R ) and c(h S ), are determined by using the wind speed measured at 1.5 m height and hub height, respectively and adding this value to the speed of sound, c 0. This is a reasonable approximation since the residences were downwind from the nearest wind turbine for the data considered in this analysis. Once the reflection point had been determined, the difference in path length for the direct and reflected sound waves,, could be found, allowing the relative phase angle, φ, to be calculated using (7) φ = 2π /λ (8) ICSV23, Athens (Greece), July

6 where λ is the wavelength of the frequency of interest. The sound field due to interference between the direct and reflected waves, p tot, was then calculated as follows p tot = p 1 + p 1 cosφ (9) where p 1 is the rms pressure associated with the direct sound wave. Figure 2 illustrates the path of the direct wave and one reflected wave for the microphone with the spherical wind screen that was mounted at a height of 1.5 m. The direct and reflected sound waves are assumed to arrive at the ground-mounted microphone simultaneously. source r 1 windshield hs r 2 h m Z 1 ground ψ ψ Z 2 Figure 4: Schematic showing the path of the direct and reflected wave for low frequency noise. 4. Results Figure 5 shows a comparison between the average sound pressure levels (SPL), in one-thirdoctave bands, measured using microphones protected with spherical and hemispherical wind screens, at locations H1 and H2. The average was determined by finding the logarithmic mean of the data that fulfilled the criteria described in Section 2.3, in each 1/3 octave band. As can be seen in Fig.5, there is negligible difference (<1 db) between the results obtained using the different mounting configurations and wind screens at both H1 and H2 for frequencies up to 63 Hz. The reason for this is that the path length difference between the direct and reflected waves is relatively small and therefore, at low frequencies, where the wavelengths are relatively large, the corresponding phase difference between the sound waves is negligible. Above 63 Hz, the difference in SPL increases gradually and reaches a maximum at 250 Hz. It is interesting that this maximum is observed at both H1 and H2 and this is likely the result of the phase difference caused by reflection from the finite impedance ground. Above 250 Hz, the difference between the results obtained using the different wind screens decreases. Figure 6 shows the simulated frequency response obtained using the phase shift resulting from the path length differences for the microphone heights associated with the spherical and hemispherical wind screens. The predictions agree well with the measurements for frequencies less than 63 Hz, where ground effects are negligible. Between 63 Hz and 100 Hz, the model predicts that the SPL at the spherical wind screen microphone is slightly lower and this agrees with the trends indicated by the measurement results. Ground effects appear to cause a further shift in the phase angle at these frequencies, resulting in a greater measured than predicted difference between the two microphone mounting configurations. Slightly larger differences in the predicted and measured results at H2 are attributed to the larger horizontal separation distance between the microphones. Figure 6 (a) shows that the influence of the path length difference is maximum for rays originating from the nearest wind turbine (WTG 1). Incorporating the path length differences for the fifth turbine (WTG5) and tenth turbine (WTG10) relative to the southernmost turbine, would have a negligible effect on the results 6 ICSV23, Athens (Greece), July 2016

7 60 60 SPL (db re 20µPa) L Aeq (avg) = 32.0 db 20 hemisphere, h 0 m sphere, h 1.5 m Frequency 1/3 (Hz) SPL (db re 20µPa) L Aeq (avg) = 34.3 db 20 hemisphere, h 0 m sphere, h 1.5 m Frequency 1/3 (Hz) (a) Location H1 (b) Location H2 Figure 5: Sound pressure level, one-third-octave band spectra. and therefore it is reasonable to assume that modelling the nearest wind turbine only is sufficient for this analysis. SPL (db re 20µPa) 5 0 Measured (avg) Predicted WTG 1 Predicted WTG 5 Predicted WTG 10 SPL (db re 20µPa) 5 0 Measured Predicted Frequency 1/3 (Hz) Frequency 1/3 (Hz) (a) Location H1 (b) Location H2 Figure 6: Predicted and measured correction, where the measured value is determined by subtracting the noise levels measured at a height of 1.5 m (spherical wind shield) from those measured at ground level (hemispherical windshield). 5. Conclusions Results from both predictions and measurements have indicated that for 1/3 octave bands below 63 Hz, there is negligible difference between the SPL obtained by mounting a microphone on the ground or at a height of 1.5 m with a hemispherical or spherical secondary wind screen, respectively. At frequencies greater than or equal to 63 Hz, the model predicts that the SPL measured by the microphone with the spherical wind screen will be lower than that measured with the hemispherical wind screen microphone. The reason for this is that the relative phase angle between the direct and reflected sound waves becomes more significant as the wavelength decreases and the frequency increases. In this case, this causes destructive interference of the direct and reflected sound waves at the 1.5 m height. The predicted trends agree with results from the measurements, although the difference between the two mounting configurations is larger for the measurements and this is attributed to ground effects. These effects cause a further variation in the relative phase angle between the sound waves ICSV23, Athens (Greece), July

8 incident on the microphones and they have not been considered in the simplified analysis presented here. At frequencies higher than 100 Hz, ground effects become increasingly more significant and therefore the corresponding predicted results are not shown here. The effect of ground impedance on the frequency response of a microphone mounted at a height of 1.5 m will be explored in future work. It is suggested that for measuring the infrasonic and low frequency component of wind turbine noise outdoors, microphones are mounted at ground level, as this reduces the wind-induced noise on the microphone as well as minimising the phase difference between the direct and reflected sound waves incident at the microphone. Converting the SPL measured on the ground to an equivalent SPL at a height of 1.5 m must take into account frequency, ground surface properties, distance from the source and the height of the source. Therefore it is recommended that the correction factor be determined on a case by case basis. The most comprehensive results can be obtained by measuring at both ground level and at a height of 1.5 m and this is the recommended approach. 6. Acknowledgements Financial support from the Australian Research Council, Project DP , is gratefully acknowledged. The authors extend their thanks to the mechanical and electrical workshop staff at the University of Adelaide and to the rural residents in South Australia who participated in this study. REFERENCES 1. Møller, H. and Pedersen, C. S. Low-frequency noise from large wind turbines, Journal of the Acoustical Society of America, 129 (6), , (2011). 2. Blazier Jr, W. Rc mark ii: A refined procedure for rating the noise of heating, ventilating, and airconditioning (HVAC) systems in buildings, Noise Control Engineering Journal, 45 (6), , (1997). 3. Strasberg, M. Nonacoustic noise interference in measurements of infrasonic ambient noise, Journal of the Acoustical Society of America, 66, , (1979). 4. Morgan, S. and Raspet, R. Low frequency wind noise contributions in measurement microphones, Journal of the Acoustical Society of America, 92 (2), , (1991). 5. Raspet, R., Yu, J. and Webster, J. Low frequency wind noise contributions in measurement microphones, Journal of the Acoustical Society of America, 123 (3), , (2008). 6. IEC Ed.3.0, (2012), Wind turbines - Part 11: Acoustic noise measurement techniques. 7. Tachibana, H. Outcome of systematic research on wind turbine noise in Japan, Internoise 2014, Melbourne, Australia, (2014). 8. Betke, K., Schultz-von Glahn, M., Goos, O. and Remmers, H. Messung der infraschallabstrahlung von windkraftanlagen (In German), Proceedings of DEWEK â96, 3rd German Wind Power Conference, Wilhelmshaven, Germany, (1996). 9. Australian Energy Market Operator. Market-Operations/Dispatch/AWEFS, accessed: 17/12/ Salomons, E., Computational atmospheric acoustics, Springer Science & Business Media (2001). 11. Plovsing, B. Delta, Comprehensive outdoor sound propagation model. part 2: Propagation in an atmosphere with refraction, (2006). 8 ICSV23, Athens (Greece), July 2016