Proceedings of the Institute of Acoustics

 

 

Field validation of octave band sound modeling for wind turbines

 

Dana Lodico¹, RSG, Denver, CO, USA

Emma Butterfield², RSG, Denver, CO, USA

Hugo Rost³, RSG, Denver, CO, USA

Kenneth Kaliski⁴, RSG, White River Junction, VT, USA

Shawn Fitzgerald⁵, RSG, White River Junction, VT, USA

 

ABSTRACT

 

The Sugar Creek Wind Project (“the Project”) is a 57-turbine wind farm in Logan County, Illinois, with a capacity of up to 202 megawatts. Preconstruction sound modeling of the final Project layout was conducted in July 2019. Following construction of the Project, postconstruction attended sound measurements were made at 38 sites in the vicinity of Project turbines in September, October, and November 2021. Measured postconstruction sound levels were compared to the octave band sound level limits applicable under the Project’s Conditional Use Permit. A comparison of the modeled and background-adjusted measured octave band sound levels for the Project indicates that with the proper modeling parameters, octave band sound levels can be conservatively predicted.

 

1. INTRODUCTION

 

Over the past 20+ years, there have been several high-quality studies to validate the modeling of wind turbine sound pressure levels at receivers. Some examples include [1], [2], [3], and [4]. These studies focused on comparing the modeled overall A-weighted broadband sound level to that measured in the field. The focus on the A-weighted level is appropriate for the following reasons:

• Most noise standards use only the A-weighting.

• The ISO 9613-2 methodology predicts the broadband sound power and does not have a sophisticated spectral ground absorption algorithm.

• Wind turbine manufacturers generally do not warranty octave band sound power.

• Low-frequency octave bands are more difficult to measure under windy conditions but are less important to the overall A-weighted sound level.

 

However, there are standards that are based on octave band sound levels. In these situations, it is important to understand whether the sound modeling parameters used to conservatively calculate the broadband A-weighted sound level can also be used to conservatively calculate octave band sound levels at the receiver. Specifically, do the ISO 9613-2 modeling parameters of G=0.5 + 2 dB at a 4- meter height, or approximately equivalently G=0 + 0 dB [5], [6], provide a reasonable estimate of the higher range of equivalent continuous sound levels downrange of wind turbines?

 

This paper describes a measurement campaign to assess the octave band sound levels at 38 sites around a 57-turbine wind project in Logan County, Illinois. It then compares the measured sound levels to those predicted at each site using the ISO 9613-2 model.

 

2. METHODOLOGY

 

2.1. Sound Propagation Modeling

 

Sound modeling was conducted in accordance with the standard ISO 9613-2, “Acoustics – Attenuation of sound during propagation outdoors, Part 2: General Method of Calculation.” The acoustical modeling software used was CadnaA^® V4.4, from DataKustik GmbH. ISO 9613-2 assumes down wind sound propagation between every source and every receiver. Consequently, all wind directions, including the prevailing wind directions, are considered. The following inputs and assumptions were used in the model:

• Ground type: half porous/half hard ground (G=0.5).

• Meteorological Conditions: Temperature of 10 °C (50 °F) and 70% Humidity.

• Ground elevations: input from the USGS Digital Terrain Model.

• Receptor height: 4 meters (13 feet).

• Vestas V150 4.2 MW turbines were modeled at the maximum sound power at each octave band for any wind speed based on manufacturer specifications.

• Vestas V110 2.0 MW turbines were modeled using octave bands from an IEC 61400-11 test. The octave bands were measured at a 9.5 m/s wind speed, representing the maximum A weighted sound power during the test.

• 2 dB was added to all results to account for sound power and modeling uncertainty.

• Foliage was not included in the model.

 

The model also considered surface reflection and absorption, atmospheric absorption, geometric di vergence, and default meteorological conditions of moderate downwind winds or equivalently a moderate nighttime inversion.

 

2.2. Sound Monitoring

 

2.2.1 Site Selection and Location

 

Sound monitoring was conducted outside all sound-sensitive structures located within 5 dB of the modeled nighttime sound limits at any frequency. These were locations whose landowners had given written permission to access their property. A total of 38 sites in the vicinity of the Project met these criteria.

 

At each site, measurements were conducted on the side of the residential portion of each property with the highest exposure to wind turbine sounds. Sound-level meters were mounted on tripods at a height of approximately 1.5 meters (5 feet) and located no closer than 25 feet from reflective surfaces to minimize acoustic reflections.

 

2.2.2 Instrumentation

 

Sound monitoring was performed with ANSI/IEC Type 1 Cirrus CR:171B sound-level meters with a minimum frequency range of 6.3 Hz to 20 kHz. Sound-level meters logged one-third octave band sound levels once each second for the entire measurement period and also recorded audio internally in .wav format to aid in sound source identification. Sound-level meters were covered with 180 mm (7 inch) windscreens to minimize the impact of wind distortion on measurements. The sound-level meters were field calibrated before and after each measurement period.

 

2.2.3 Meteorological Data

 

Meteorological data was measured using an Onset HOBO anemometer, located at microphone height at one of the monitor locations during each monitoring period. The temperature, average wind speed, and maximum wind gust speed were logged once per minute. At every monitor location, a handheld anemometer was used to record wind gusts, temperature, and relative humidity. Additional relative humidity and temperature data were obtained from the nearby Automated Surface Observation Station (“ASOS”) located at the Logan County Airport (AAA) in Lincoln, Illinois.

 

Wind speeds of 8 m/s (18 mph) or greater at the turbine hub height were needed to ensure that wind turbines were operating at or near maximum sound power levels. However, wind speeds in excess of 5.4 m/s (12 mph) at the microphone result in extraneous wind sounds in the measurement data.

 

2.2.4 Monitoring Period

 

Monitoring was conducted during nighttime periods when background sound levels are lower to reduce extraneous sound sources, such as vehicular traffic and other human activity. Wind shear is also higher at night leading to more favorable propagation conditions and lower ground wind speeds. Sound levels were measured for a minimum total monitoring period of one hour at each location, including 45 minutes or longer during turbine operations (i.e., turbine-on period) followed immediately by 15 minutes or longer with wind turbines shut down (i.e., the background period).

 

2.2.5 Turbine Operations

 

Sound monitoring was conducted when Project turbines were operating at a level sufficient to produce maximum sound emissions (±1 dB). Based on a review of manufacturer data for the two turbine types, the Vestas V150 4.2 MW and the Vestas V110 2.0 MW, it was determined that ±1 dB of maximum sound power level is achieved at 75% of full power or greater. This generally equates to a wind speed at the turbine hub height of 8 to 10 m/s (18 to 22 mph) or greater. To assess when these conditions may occur for valid sound monitoring, production forecasts were reviewed daily. After measurements were completed, hub height wind speed and turbine active power was obtained from Sugar Creek Wind’s Supervisory Control and Data Acquisition system in 10-minute intervals for all monitoring periods.

 

During the monitoring of background sound, all turbines that were anticipated to substantively contribute to the sound levels at a specific site (0.5 dB or more to the turbine operational level at the receptor) were shut down.

 

2.3. Data Processing

 

2.3.1 Data Exclusions

 

After data collection, all logged one-second one-third octave band sound-level data was downloaded, processed, and aggregated into 10-second periods. Field notes, meteorological data, audio recordings, and analysis of sound-level spectrograms were used to identify exclusion periods. At each monitoring location, periods under the following conditions were excluded:

• Wind gust speeds at the monitoring location exceeding 5.4 m/s (12 mph).

• Precipitation and thunder (did not occur).

• Humidity greater than 90% (did not occur).

• Temperatures below -18° C (0° F) (did not occur).

• Anomalous sounds such as nearby cars, planes, and animals.

• Equipment interactions by staff, other people, or animals.

 

Biogenic sounds including insects and birds were present in all data. Since it was not possible to collect data without these sounds and the sounds persist both during the period of turbine operations and during the shutdown period, these sounds were not directly removed from the data, but were rather canceled using background subtraction (see Section 2.3.3).

 

2.3.2 Identification of Valid Data

 

Once data exclusions were removed, the one-third octave band data was summed into octave bands and then averaged for the turbine-on and shutdown periods. The sound monitoring protocol required a minimum of 15 minutes of valid turbine-on data and 5 minutes of valid background (turbine-off) data.

 

Forecasting nights where wind turbines are operating at or near their maximum sound output is difficult. As a result, in situations where turbines were operating near, but not at, maximum sound power levels, sound-level adjustments were made, as follows:

• For periods where active power of all turbines that were anticipated to substantively contribute (0.5 dB or more) to the operational sound levels at a specific site were 75% or greater, data was considered valid, and no adjustments were used.

• Where active power was below 75%, adjustments were calculated based on wind turbine manufacturer data, correlated with sound modeling results from the preconstruction sound study, as follows:

  • Where adjustments of +1 dB or less were required at all frequency bands, sound levels were adjusted upwards based on the difference between the actual and maximum sound power contribution for that turbine.

  • Where adjustment greater than 1 dB at any frequency band would be needed, data was considered invalid and monitoring repeated.

 

If data did not meet these duration requirements, measurements at that site were repeated.

 

2.3.3 Calculation of Turbine-Only Data

 

Data acquired during turbine operations includes turbine sounds, as well as other ambient background sounds such as insects, distant traffic, and wind blowing through foliage. To calculate the turbine only sound levels, corrections were applied to the turbine-on data based on the difference between the turbine-on and turbine-off sound levels using corrections specified in Section 910.106(a) of the Illinois Administrative Code (Title 35, Subtitle H, Chapter I). The resulting turbine-only sound levels were then compared to the modeled sound levels at these same measurement locations.

 

2.4. Results

 

Sound monitoring was conducted at 38 sites in the vicinity of the Project to satisfy county and state requirements. Figure 1 shows the monitoring locations and the modeled contour of the 1,000 Hz nighttime sound-level limit minus 5 dB.

 

 

Figure 1: Sound monitoring locations.

 

A comparison of the measured to the modeled sound levels for each site is given in Figure 2. Table 1 shows a summary of the results of the comparison between measured and modeled sound levels. If the difference between the turbine-on and background sound was less than 3 dB at any frequency, the turbine-only sound level at that frequency was not determinable and, therefore, is not reported. Figure 3 shows the percentage of valid sites at each frequency band.

 

As shown in Figure 2, almost all sites resulted in measured levels below modeled levels at all valid frequencies. Measured sound levels were, on average, 1 to 6 dB below modeled levels in the octave bands from 31.5 Hz to 2 kHz. Measured sound levels were as much as 10 to 12 dB below modeled sound levels at select sites in some frequencies. It may be that the most favorable weather conditions to propagation were not captured in these cases, since only a short time period is measured under county requirements. Three sites had measured sound levels that exceeded modeled sound levels in an individual frequency band; in all three cases where individual frequencies resulted in levels above modeled levels, turbine sound levels were noted in the field observations to be similar in level to background sounds. Thus, the higher levels could be due to background contamination.

 

Table 1: Difference between measured and modeled results.

 

 

 

Figure 2: Comparison of measured and modeled results.

 

 

Figure 3: Percent of valid measurement sites.

 

High-frequency turbine-generated sound levels above 2,000 Hz were determinable at only two out of 38 (5%) of the sites due to turbine levels being below background biogenic sounds, including insects and birds. Likewise, sound levels in the 2,000 Hz octave band were determinable at only 11 of the 38 sites (29%). The two sites where high-frequency turbine-only data were considered valid were measured simultaneously during a nighttime period (approximately 12 AM to 1 AM). See Figure 4 for an example of one of the two sites that resulted in valid high-frequency data. Note that the grayed-out area is the turbine shutdown period, which includes the transitional period that was not included in the background-level calculation. Review of the 2 to 8 kHz data, and to some degree the 1 kHz data, indicates higher sound levels in the early turbine-on data that are not indicated in the later turbine-on data with the same turbine active power. Given that turbine sound levels should be relatively consistent given the same operating conditions, this is attributed to differences in the background sounds. These sounds were difficult to discern in the recordings due to their high-frequency content and low recording sample rate (low Nyquist frequency) and were therefore not excluded from the data set.

 

 

Figure 4: Measured sound levels at an example site with varying high-frequency background sounds.⁶

 

As described previously, measurements were made at all sites within the predicted noise contour where landowner permission was given to access their property. Sites were not eliminated due to high background levels. As a result, turbine sound levels at a few of the sites were indiscernible from background at most or all frequencies. Figure 5 shows an example of a site where background sound sources dominated the sound environment. Figure 6 shows an example site that achieved a satisfactory signal-to-noise ratio at frequencies below 4,000 Hz.

 

 

Figure 5: Measured sound levels at an example site dominated by background sounds.⁶

 

 

Figure 6: Measured sound levels at an example site with satisfactory signal-to-noise ratio.⁶

 

The site shown in Figure 5 was measured on two separate days during conditions favorable to sound monitoring. Turbine sound was indistinguishable from background sound during both measurement periods. Figure 6 indicates a visible difference in sound levels between the turbine-on and turbine-off periods in the frequencies below 4,000 Hz.

 

3. DISCUSSION

 

3.1. Octave Band Noise Standards

 

The analysis presented above show that sound modeling of octave bands is feasible and conservative using the typical modeling parameters cited in such noise standards as ANSI/ACP 111-1-2022 [4]. However, this observation includes the following caveats:

• The sound monitoring conducted for this validation study was limited to one hour of attended measurement at each of 38 locations. While each measurement was made during nighttime meteorology and within 1 dB of the maximum sound power level of the closest wind turbines, a wider range of meteorological conditions could be obtained from unattended measurements. Additional monitoring would provide additional data points and reduce uncertainty.

• Octave band sound measurements are particularly difficult for very low frequency and very high frequency sound. Low-frequency measurements are easily contaminated by both wind induced sound (such as wind blowing through vegetation) and pseudonoise from turbulence over the microphone. High-frequency sound, such as that above about 2,000 Hz should be disregarded, as atmospheric attenuation of sound at these frequencies is usually sufficient at receptor distances to reduce sound below background. Many contaminating high-frequency sources create anomalies in the data. In addition, high-frequency sound is normally well attenuated through building structures and is thus irrelevant for indoor audibility.

• Octave band sound modeling is not supported in ISO 9613-2. While the model can predict octave bands from 63 Hz to 8,000 Hz, the standard clearly states, “The method predicts the equivalent continuous A-weighted sound pressure level (as described in parts 1 to 3 of ISO 1996) under meteorological conditions favorable to propagation from sources of known sound emission.” That is, the methodology is intended to predict broadband sound. In particular, the spectral calculation of ground effects is somewhat limited especially for high sources such as wind turbines.

• Octave band sound power levels are often not guaranteed/warranted by the wind turbine manufacturer. That is, if there is an exceedance of an octave band noise standard caused by the wind turbine, where the wind turbine’s A-weighted sound power is within the warranted level, the developer may have no recourse from the manufacturer.

• Mitigation of specific octave bands is difficult. If noise limits are exceeded in only a single octave band, the most commonly implemented solution is noise-reduced operations (NRO). However, NRO affects all octave bands, sometimes some more than others, and results in the loss of energy production from that turbine. The authors know of no mitigation measures specific to individual frequencies of sound emissions.

 

For these reasons, it is not recommended that jurisdictions adopt octave band noise standards for wind turbines. In this case, the noise standard applied to the Project is a statewide standard that applies to all noise sources. As such, it was the default noise standard applied to the Project. However, given a choice, jurisdictions should shy away from octave bands and focus on A-weighted broadband equivalent continuous sound levels.

 

3.2 Attended vs. Unattended Sound Monitoring

 

This study used attended sound monitoring to collect data. The advantage of attended monitoring is that there is a person observing the measurement and identifying anomalous sounds. The disadvantages are that it can be difficult to capture the most appropriate weather conditions and that only a short period of time is measured.

 

The alternative is unattended sound monitoring. In this case, sound monitors are set up for days to weeks at a time, logging sound data and recording audio to help identify anomalies. The advantages of unattended monitoring are the ability to collect data over a wide range of meteorological conditions. The disadvantages are that it is more difficult to identify anomalies in the audio recordings, fewer sites can be measured, and the methodology results in a large amount of data that must then be analyzed.

 

The experience in this case highlighted several issues with attended monitoring. First, the monitoring was done at night to capture low background sound levels. However, there are substantive safety concerns operating around people’s homes at night, especially with respect to potential misunderstandings. In addition, it took over two months and several visits to the site to collect the one hour of data at the 38 locations. This was due to issues related to weather, turbine operations, and unforeseen turbine maintenance. Nevertheless, high-quality data was successfully collected in the end, albeit at a cost.

 

The general recommendation is to collect more data at fewer representative sites using unattended sound monitoring. In this way, a greater variety of conditions can be monitored.

 

3.3. Site Selection

 

Sites for this study were selected based on landowner permissions in accordance with county requirements. As a result, many of the sites had turbine sound levels that were not discernable from background sound levels. Although valid data was achieved at these sites, turbine-only sound levels were unable to be calculated at most or all frequencies. For a more cost-effective monitoring program, a more discriminating site selection process should prioritize sites where a higher signal-to-noise ratio can be achieved.

 

4. CONCLUSIONS

 

Postconstruction attended sound measurements were made at 38 sites in the vicinity of Project turbines. Measured postconstruction sound levels were compared to the octave band sound level limits applicable under the Project’s Conditional Use Permit and to preconstruction modeled sound levels. A comparison between measured and modeled octave band sound levels showed that octave band sound levels can be conservatively predicted using proper modeling parameters. However, due to modeling, measurement, and mitigation constraints on octave band levels, it is recommended that jurisdictions should shy away from octave bands and focus on A-weighted broadband equivalent continuous sound levels when setting noise standards for wind turbines. Additionally, for cost effectiveness and to ensure that a variety of meteorological conditions are collected, long-term unattended monitoring at a few representative sites is recommended over attended monitoring at a larger number of sites. It is also recommended that site selection prioritize acoustical considerations such as low background sound levels and good signal-to-noise ratios.

 

5. ACKNOWLEDGEMENTS

 

The work reported here was funded by Sugar Creek Wind One LLC, a wholly owned subsidiary of Liberty Algonquin Business Services(APCo). Considerable support and coordination assistance were provided by Andrea Berenkey, Dean Waldinger, and Anthony Jones of APCo and Stan Komperda of Highlander Renewables.

 

6. REFERENCES

 

1. Kaliski, K. and Duncan, E. “Propagation modeling Parameters for Wind Power Projects,” Sound & Vibration Magazine, Vol. 24 no. 12, December 2008.

2. RSG et al. “Massachusetts Study on Wind Turbine Acoustics,” Massachusetts Clean Energy Center and Massachusetts Department of Environmental Protection, 2016.

3. Delta, “Validation of the Nord2000 propagation model for use on wind turbine noise,” October 2009.

4. Evans, T. & Cooper, J., “Comparison of predicted and measured wind farm noise levels and implications for assessments of new wind farms.” Acoustics Australia. 40, 2012.

5. ANSI/ACP 111-1-2022. “Wind Turbine Sound Modeling,” American Clean Power Association, 2022.

6. Institute of Acoustics. “A good practice guide to the application of ETSU-R-97 for the assessment and rating of wind turbine noise,” 2013.

 

---

¹ Dana.Lodico@rsginc.com

² Emma.Butterfieldf@rsginc.com

³ Hugo.Rost@rsginc.com

⁴ Ken.Kaliski@rsginc.com

⁵ Shawn.Fitzgerald@rsginc.com

⁶ Gaps indicate data excluded from averaging due to anomalous sounds or high wind gusts at the monitoring location. Grayed areas are turbine-off periods.