wtn09_bullmore

7/29/2019 wtn09_bullmore

1/20

Wind Farm Noise Predictions and Comparison with Measurements Page 1 of 20

Third International Meetingon

Wind Turbine NoiseAalborg Denmark 17 19 June 2009

Wind Farm Noise Predictions and Comparison with Measurements

Andrew Bullmore, Justin Adcock, Mark Jiggins, Matthew Cand

Hoare Lea Consulting Engineers140 Aztec West Business Park

Bristol, UKBS32 4TX

[email protected]

[email protected]@hoarelea.com

[email protected]

Abstract

Conservative approaches to the prediction of noise immissions from wind farmsreduce the risk of compliance failure. However, overly conservative approachesintroduce the risk of not capturing the true energy generating capacity of a given windfarm site. Unlike other forms of development, conservative planning of wind farms

cannot be offset by increased mitigation without incurring such lost energygenerating potential. The large scale of modern wind farms means that seeminglysmall conservatisms in the prediction of noise immission levels can translate tosubstantial lost development opportunities.

A worst case assessment methodology assumes that a receiver is located downwindof every turbine, all turbines experience the same wind conditions as the first upwindturbine, the ground acts as a hard reflective surface, and all turbines are emittingsound power greater than test levels. Minimal reductions, if any, are factored for theexcess attenuation provided by the atmosphere and barrier effects. In practice, allthese factors are unlikely to transpire simultaneously. To gauge the pessimism of this

approach, several campaigns of long term measurements were carried out nearoperational wind farms and compared to noise levels predicted using severaltechniques. Initial results were presented previously; this paper presents furtherresults and analysis and discusses the opportunities for more realistic predictiontechniques. The paper then continues to discuss the potential impact that the use ofmore realistic prediction techniques may have on increasing the potential generatingcapacity of wind farm sites.
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]


2/20


Introduction

It is common practice in many countries to control the noise impact of proposed windfarm developments by setting limits for the maximum level of noise that may occur atsurrounding noise sensitive receptor locations. An important distinction in this

practice is that the test of compliance in some countries may be based on predictednoise levels alone, whilst in others the test is based on the actual noise levels thatoccur in practice, as demonstrated by measurements. The latter method offersbenefits for regulators in that there is a definitive limit for the noise that may occur inpractice. However, measurement based compliance places the onus on developersand their advisors to plan and design wind farms in a way that adequately addressesthe risk of a failed compliance test and the subsequent power generation losses thatcould be incurred to address the failure. The prediction of wind farm noise immissionlevels (i.e. the noise occurring at the receiver location, in contrast to the emissionlevel that defines the sound power output of the sources) is therefore an integralelement of the planning process for measurement based noise compliance regimes.

However, prediction of environmental noise immissions from wind farms is influencedby a range of variables. This means that choices have to be made in the calculationparameters adopted for these variables in any assessment, and these choices canhave a significant bearing on the outcome results.

The propagation of environmental noise immissions, and therefore its prediction, isinfluenced by a number of variables. This is evident in the measured noise levelsobserved around wind farms. Such noise levels tend to show a relatively wide rangeof temporal variation, even under relatively stable downwind propagation conditions.

The key focus of a prediction exercise is usually the upper noise level that will occurunder such downwind conditions, for which the following factors must be addressed:

the turbine sound power and any associated uncertainties;

the source height;

the receiver height;

the wind speed experienced by the turbine rotors, how this may vary acrossthe rotor diameter, how it may vary between individual turbines across a multi-turbine site, and how these variations relate to reference wind speeds derivedat other heights or locations;

the wind direction, and the range of angles to the direct line between thesource and receiver for which downwind conditions are considered to occur,

along with the portion of an expansive site that can simultaneously lie withinthese angles;

temperature and humidity;

the terrain profile with respect to intervening ground height and noisescreening features;

the ground characteristics of the surrounding area and any regional orseasonally varying changes to its composition;

the selection of noise index adopted to quantify the calculated noise immissionlevels (LAeq, LA90, LA50, etc.).


3/20


These variables, and the manner in which they are accounted for, will impact on thelikelihood of the predicted noise immission levels being higher or lower than theactual immissions that occur in practice. Understanding the nature of these factors toenable informed selection of prediction input parameters is therefore vitally important

for designers and developers alike if they are to make truly informed selections oftheir noise prediction methodologies and likewise the relevant input parameters fortheir selected methodology. Ultimately, there will be a trade-off of considerations,which will need to strike the appropriate balance between the potential generatingcapacity losses of conservative approaches and the compliance risks of moreoptimistic approaches.

In most instances, environmental noise predictions are made on the basis ofestablished engineering methods such as ISO 9613[1]. The use of these methods hasbeen supported by various studies such as the EU J oule report[2] which found thatsuch methods offered a robust means of estimating downwind noise levels that

would not generally be exceeded in practice, and generally offered a margin ofconservatism depending on the choices made regarding input parameters. However,despite the relative simplicity of such prediction methods (when compared toadvanced numerical or analytic methods), informed choices still need to be made ona site by site basis. Experience suggests, however, that these choices can have asignificant effect on the outcome findings of individual noise assessments and thuscan often become the focus of considerable dispute between developers/designersand other interested parties.

Previous papers[3] produced by Hoare Lea Acoustics highlighted the effect thatseemingly minor assessment choices can have: differences of less than 3 dB(A) can

translate in effect into large differences in the potential generating capacity ofindividual wind farm sites, which has implications for national-scale wind energypotential. To provide an improved basis for making noise prediction choices, HoareLea Acoustics have carried out noise monitoring exercises around a number of UKwind farm sites for the purpose of comparing predicted and actual noise immissionlevels. A key element of this investigation was to compare predicted noise levels thatare derived from techniques that are normally used during the design phase of awind farm with the actual immission levels that occur in practice. This investigationhas considered predicted and measured turbine noise levels that occur for the windspeed experienced at the turbine rotors, thus focussing the analysis on sound powerand propagation effects by limiting the influence of uncertainties related to reference

wind speeds and wind shear.This paper presents the findings of the measurements and analysis completed todate and sets out the requirements of any further studies.

Site Descriptions

For commercial reasons, it is not possible to disclose the full details of the wind farmschosen for this study, and thus the following general descriptions are provided. Allthree sites (A, B and C) were located in rural areas and comprise wind farms withmore than 20 turbines. The turbines in all three cases were two speed active stallregulated machines rated at over 2 MW generating capacity per machine, with hub

heights of 60 to 70 meters.


4/20


Site A

The wind farm is located on a relatively high plateau characterised by moderatelyundulating terrain and minimal vegetation. Ground conditions were a mix of partlygrassland and mainly peat bog, but given the undulation, the land was not prone to

complete saturation. In addition, ground conditions were effectively frozen for themost part of the measurements because of low ambient temperatures.

Site B

The wind farm is located in reasonably flat terrain with minimal vegetation. Theground surrounding the wind farm was almost entirely composed of peat bog. Theseground characteristics, coupled with the very high rainfall in the area, meant that theground is believed to have been totally water logged for the entire duration of thesurvey, thus providing effectively hard ground propagation conditions.

Site C

The terrain was lightly undulating but effectively flat in acoustic terms, and groundconditions were a mix of grassland and flooded areas. Minimal vegetation waspresent in the immediate surroundings of the turbines, but large areas of forestrywere located further away.

Survey Description

At each site, automated Type 1 sound logging meters (SLMs) were positioned atvarying distances from the nearest turbine, with an installed microphone height ofapproximately 1.5 m above ground.

For all sites, the SLMs were set to log continuous periods of 10 minute noise levels,recording statistical and equivalent noise level parameters. The internal clocks on theSLMs were synchronised with the wind farm control system. All systems werecalibrated on deployment, during interim data collection and following collection fromsite, no significant drifts in sensitivity were found (typically below 0.5 dB(A)).

Supplementary non-acoustic data was obtained from the Supervisory Control & DataAcquisition (SCADA) System of each wind farm for the operation of the turbines andmet mast during the period of noise monitoring. The SCADA data provided thefollowing information:

date/time at the end of each 10 minute period;

primary wind speed from the turbine nacelle (mean);

turbine power output (kW) (mean);

turbine rotor speed (min/mean/max);

turbine nacelle orientation (mean);

met mast wind speeds at hub height (mean);

met mast wind direction at hub height (mean);

rainfall indication;


5/20


temperature and humidity (not used in the present assessment but effects tobe studied).

Site A

3 sound level meters were positioned to the northwest of the wind farm at distancesranging from 415 m to 920 m. The equipment comprised Svantek SVAN949 loggingSLMs housed in environmental enclosures with battery power. The enclosures havean integral pole to provide a mounting for the microphone and windshield system. Atwo layer windshield system was used to reduce wind induced noise on themicrophone. The primary windshield and rain protection were provided by a 01dBBAP21 outdoor microphone adaptor which enclosed the standard microphone andpre-amplifier. The secondary windshield was custom made from open cell foamapproximately 25 mm thickness formed as a domed cylinder 170 mm diameter and300 mm high. A lower disc of 40 mm thick open cell foam formed total enclosure ofprimary windshield. The outer windshields were custom designed following the

guidance given in the report Noise Measurements in Windy Conditions

[4]

. The reportindicates that the insertion loss of this type of windshield assembly is likely to be lessthan 1 dB between 50 Hz and 5 kHz. The positioning of the meters was largelydriven by practical access constraints. A total measurement period of approximately47 days was obtained at this site.

Site B

5 sound level meters were positioned along a single line directed just to the west ofnorth. The alignment of this array of meters was chosen for the availability of stableground conditions and to avoid local streams to the north east of the site which wouldhave been sufficient to contaminate the measurements with water flow noise. Themeasurement distances were 101 m, 270 m, 466 m and 754 m. The equipment wasthe same as that used at Site A. A total measurement period of approximately 34days was obtained at the 100 m location and more than 57 days at the otherlocations.

Whilst the northwest positioning of the meters for both sites A and B was out of thedirect down-wind line according to the prevailing UK south westerly wind direction, itoffered a broader mix of wind directions to be acquired enabling both downwind andcrosswind noise propagation conditions to be investigated.

Site C

At site C, five sound level meters were positioned along two lines directed to theNorth and North-East, at distances of 100 m to 820 m from the closest turbine, inrecognition of prevailing wind directions and site constraints such as streams andforestry. Individual positions surrounding the site, at distances of 700 m to 1000 mwere also installed. In addition to the Svantek SLMs, systems based on RION NL-31SLMs in similar enclosures were used, and the microphones equipped with largediameter windshields.

To reduce uncertainties due to ground absorption effects and help characterise theeffective noise emissions of the sources, the closest position consisted of amicrophone installation on a circular hard board with a double hemispherical wind-


6/20


shield arrangement, following the guidance of IEC 61400-11[5] for turbine soundpower certification. In addition, several other individual positions were located atvarying distances around the wind farm. A total of approximately 2 monthsmeasurement data was obtained.

Analysis

The first element of the data analysis was to correlate the measured noise levelinformation with the prevailing wind conditions. At the design stage of a wind farm,predictions would normally be based on a single reference free condition windspeed value which is taken to be experienced simultaneously by all wind turbines(with the exception of very large sites where more than one reference may be used).

Thus, for this study, it was initially chosen to relate the measured noise levels to asingle wind speed and direction representation for the site. Generally, thisinformation was acquired from the site meteorological masts. However, at Site B, it

was known that under certain directions, the reference meteorological masts wouldbe downwind of the wind farm and the wind speed measurement would therefore beinfluenced by the wind farms presence. For these directions, the wind speed wastaken from turbine locations that were upwind of the remainder of the site. The windspeed data at the turbines were deduced from the nacelle anemometer readings,subsequently corrected to free-flow conditions (using site-specific nacelle correctionssupplied by the site operators). In all instances, the wind speed reference for thecorrelation related to hub height wind speeds. Due to differing client requirements forsites, the wind speed data was either corrected to 10m wind speed heights assumingreference roughness conditions (z=0.05) (Sites A and C), or raw hub height windspeed data was referenced (Site B).

The correlated noise and wind speed data were then filtered to eliminate any periodsin which rainfall was indicated to have occurred, or during the times when servicepersonnel were known to have been near the sound level meters.

At site A, additional data filtering comprised reduction of the data set to winddirections from 90 to 200 degrees to provide a 110 degree wide arc of downwindpropagation conditions (required to encompass the distributed measurementlocations).

At Site B and C, data filtering resulted in the production of two datasets for downwindangle ranges +/-15 degrees from directly downwind and +/-45 degrees from directlydownwind. The former angle range is specified in the relevant turbine sound poweroutput standard (BS EN 61400 Part 11:2003[5]) whilst the latter represents anextended range often regarded as still representing downwind conditions (althoughdownwind propagation can ultimately occur for wider angles due to the range of windspeeds considered for wind farm sound propagation). An additional dataset was alsoformed for wind directions within +/-45 degrees of directly upwind conditions for thenearest turbine and the measurement line. In addition, the study focussed on theperiods in which all turbines were generating in high speed mode.


7/20


For each of these correlated 10 minute records, predicted noise levels weregenerated using the ISO 9613 prediction methodology according to the followingparameters:

source height equal to hub height;

receiver height equal to 4 m and free-field conditions;

10 degrees Celsius and 70% relative humidity;

flat and level ground cover for two separate conditions: G = 0 and G = 0.5 (forsource, middle, and receiver ground) to consider hard and mixed groundcover conditions, according to site-specific considerations;

turbine sound power data provided by the manufacturers and measuredaccording to BS EN IEC 61400 Part 11:2003[5], excluding any margin for testuncertainty or manufacturers warranties (i.e. raw measurement turbine noiselevels). The turbine sound power data was converted to hub height wind

speeds (assuming reference ground roughness of 0.05 m) and plotted toobtain a 3rd order polynomial best fit curve. This curve was then used toobtain the sound power value for non-integer wind speeds when required;

the subtraction of 2 dB(A) to correct for the use of the LA90 rather than LAeqindex, according to ETSU-R-97[6]. This assumption was found to be consistentwith the analysis of the measured data.

The predictions are first made without inclusion of any margin for test uncertainty ormanufactures warranties. It is common practice for manufacturers to addapproximately 2 dB, although actual values may be considerably different to thisaccording to commercial factors. The IEC 61400-11 standard requires an estimate of

the test uncertainty to be presented along with the determined turbine sound power,and reported values tend to be less than 2 dB.

Results & Discussion

Site A

The analysis of Site A was subject to a greater degree of complexity due to factorssuch as ground terrain profile and the varied orientations of the measurementlocations relative to the wind farm. Thus, whilst the analysis of Site A was supportive

and consistent with that of the other sites, for brevity only partial results arepresented within this paper.

Figure 1 presents the results of measurements at the most distant measurementlocations: position 3, approximately 900 m from the nearest turbine. Predictions weremade using hard ground cover (ISO 9613 G=0 for the source, middle and receiverground) because of site observations of a frozen ground surface for which G=0would be expected to be the appropriate ground characterisation. The predictions foreach 10 minute period only included for the turbines which are known to have beenoperating, as the number of operational turbines varied during the survey; this resultsin lower predicted levels, particularly apparent in lighter wind conditions.


8/20


The predicted immission levels, excluding any margin for uncertainty, generallyexceed measured levels; the margin above typical, average immission levels isapproximately 3 dB(A). The addition of a further 2 dB(A) sound power uncertaintymargin would then correspond to a significant over-estimate of typical immissionlevels. Furthermore, background ambient levels have likely influenced the

measurement to some extent; this aspect is discussed in more detail for site B below.

Figure 1 - Comparison of Site A measured (grey) and predicted (red) noise levels at measurementlocation 3 only.

Site B

It was more straightforward for this site to directly compare the 4 differentmeasurement locations, situated at increasing distance from the wind farm.

The group of charts presented in Figures 2a to 2d relate purely to the Site Bmeasurements, with associated noise immission predictions presented for a single

prediction methodology, which is based on hard ground cover (ISO 9613 G=0 for thesource, middle and receiver ground). This is because site observations indicated thesoil to be almost totally submerged by ground and surface water for which G=0would be expected to be the appropriate ground characterisation. A single referencefree-field wind speed for each ten minute period is assumed to occur at all of theturbines

The indicated upwind measurements for all four sites are the +/-45 degree upwindmeasurements taken from the furthest position (where upwind noise levels are morelikely to relate to background noise levels). The indicated downwind measurementsare for the +/-15 degrees angle only. Comparison of the results for the four separatemeasurement locations indicates the following:

Location 3

(Wind Directions 90 to 200 Only)

20

25

30

35

40

45

50

55

60

3 4 5 6 7 8 9 10 11 12 13

Wind Speed Calculated at 10m Assuming 0.05 Roughness (m/s)

L90d

B(A)


9/20


The upwind and downwind measurements show a very clear noise leveldifference which supports the view that downwind measurements have beenstrongly influenced, if not controlled by, the wind farms emissions. At thefurthest location, the difference between the downwind measurement trendline and the general trend of the upwind values is around 6 to 7 dB(A).

Previous studies such as the EU J oule project[2]

have indicated differencesbetween upwind and downwind turbine noise levels of the order of 10 dB(A) to15 dB(A) at distant locations. The observation of a lower difference at thefurthest position may indicate that either the 10 dB(A) to 15 dB(A) reductionhas not been realised at this site, or more likely, that the background noise isdominating the upwind measurements, thus limiting the observed difference.

At all locations, the predicted immissions trend line generally exceeds (by upto approximately 1 dB(A) at the nearest measurement location) or just equalsthe downwind measured data trend line. The exact background noise levelinfluence at each measurement location for each 10 minute period cannot be

known. It is however likely that the background noise level may havecontributed 1 dB(A) or more to the total measured noise levels at the furthestlocation. The margin between actual turbine contribution and the predictedimmission levels will therefore be greater than indicated by the totalmeasurement comparison represented in the charts.

At hub height wind speeds up to approximately 12 m/s, the margin betweenthe prediction trend line and measurement trend line is relatively constant forincreasing wind speed at each site. At higher wind speeds, the prediction andmeasurement trend lines diverge at each site, with the predictions showing anincreasing margin above the measured noise levels. Subsequent resultsdiscussed later tend to suggest this is due to the increased significance ofwind speed variations across the wind farm at higher wind speeds.

The margin between the prediction trend line and measurement trend linetends to progressively decrease with increasing distance from the turbines.

The most obvious potential cause of this effect is the increasing influence ofbackground noise at increasing distance. However, another importantconsideration is the changing angle between the turbines and themeasurement locations relative to the wind direction with increasing distancefrom the turbines. As the receiver location approaches the turbine locations itbecomes increasing unlikely that the receiver location could lie downwind ofevery turbine simultaneously. This means that some turbines at peripheral

positions may contribute less than the directly downwind propagationassumed in the prediction. At increasing distance, this effect is reduced, andthe turbines located at the periphery of the site will then increasingly contributeto the total wind farm noise level (i.e. a greater portion of the turbines at thewind farm site will be propagating sound under conditions closer to directdownwind propagation).


10/20


Site B - Measurement Location 1 (100 m)

30

35

40

45

50

55

60

8 9 10 11 12 13 14

Hub Height Wind Speeds (m/s)

TotalLA90

NoiseLevels(dB)

Measured

Predicted

Upwind (45)

Poly. (Predicted)

Poly. (Measured)


30

35

40

45

50

55

60

8 9 10 11 12 13 14


TotalLA90

NoiseLevels(dB)

Measured

Predicted

Upwind (45)

Poly. (Predicted)

Poly. (Measured)

Figure 2(a & b): Sample analysis group set to compare Site B measured and predicted noise levelsat the 4 measurement distances. All downwind angles restricted to +/-15 degrees.Predicted noise levels based on a single site wind speed reference and G =0.


11/20



30

35

40

45

50

55

60

8 9 10 11 12 13 14


TotalLA90

NoiseLevels(dB)

Measured

Predicted

Upwind (45)

Poly. (Predicted)

Poly. (Measured)


30

35

40

45

50

55

60

8 9 10 11 12 13 14


TotalLA90

NoiseLevels(dB)

Measured

Predicted

Upwind (45)

Poly. (Predicted)

Poly. (Measured)

Figure 2 (c & d): Sample analysis group set to compare Site B measured and predicted noise levelsat the 4 measurement distances. All downwind angles restricted to +/-15 degrees.Predicted noise levels based on a single site wind speed reference and G =0.


12/20


The group of charts presented in Figures 3a and 3b relate only to measurementlocation 4 (754 m) of Site B. The 2 charts for this location present +/-45 degreesdownwind conditions, and differ in terms of the wind speed reference used for thepredictions: initially a single wind speed reference as presented in figure 2, but thenmodifying the predictions to account for the actual wind speed seen by each

individual turbine in each 10 minute period. Comparison of the results indicates that:o Expansion of the range of downwind angles from +/-15 degrees to +/-45

degrees indicates the predictions exceed the total measured noise levels by aslightly greater margin for the widened downwind angle. This may be due tothe increased number of data samples offering a better representation of thetrue relationship between measurements and predictions. Alternatively, thismay indicate that the contribution of the dominant/nearest turbines to themeasured levels is progressively reduced as the wind direction moves awayfrom directly downwind conditions and this effect is not represented in thepredictions.

o The predictions made on the basis of the individual wind speed experiencedby each turbine rather than a single site wind speed reference indicateimmission levels which no longer diverge from the measurement trend line athigher wind speeds. This tends to suggest that the margin of conservatismdemonstrated at higher wind speeds is strongly related to the reduced level ofwind seen by the nearest turbines to the measurement location which may bedue to sheltering and/or wake effects of upwind turbines. To investigate thisfurther, a statistical analysis of the difference between the single wind speedreference and the wind speed of each of the turbines indicated the followingkey figures:

oMean difference = -0.5 m/s

oStandard deviation of differences = 1.2 m/s

oMaximum decrease = -5.1 m/s

oMaximum increase = 4.3 m/s

[Values shown are derived from the individual turbine specific wind speeds minus the single sitereference wind speed. A negative number indicates the reference wind speed is overestimatingthe wind speed at each individual turbines.]

A wind speed changes of the order of 0.5 m/s would correspond to a 0.4dBdifference in the sound emissions of the turbines. The values above areaverages, but immission levels are dominated by the closest turbines whichare also the most shielded in down-wind conditions. The scatter of thepredicted data also appears closer to that of the measurement data whenusing the variable wind speed reference, which suggests this effect could be a

significant source of the variability observed in practice.

o Although the predictions appear to match with measurements at the mostdistant location, the latter were likely influenced by background noise to acertain degree. If the typical upwind measured noise levels are taken as anestimate of the background levels, this influence would then equate to anincrease of 1 dB to 2 dB. The true margin between predictions andmeasurements would then be similar to the margin observed at closerlocations. The further addition of a 2 dB(A) uncertainty margin would thencorrespond to a significant over-estimate of typical immission levels.


13/20



30

35

40

45

50

55

60

8 9 10 11 12 13 14

Hub Height Wind Speed (m/s)

TotalLA90

NoiseLevel(dB)

Measured

Predicted

Upwind (45)

Poly. (Predicted)

Poly. (Measured)

Figure 3a: Comparison of Site B measured and predicted noise levels at measurement location 4only. All downwind angles restricted to +/-45 degrees. Predicted noise levels based on asingle site wind speed reference and G =0.


30

35

40

45

50

55

60

8 9 10 11 12 13 14

Hub Height Wind Speed (m/s)

TotalLA90

NoiseLevel(dB)

Measured

Predicted

Upwind (45)

Poly. (Predicted)

Poly. (Measured)

Figure 3b: Comparison of Site B measured and predicted noise levels at measurement location 4only. All downwind angles restricted to +/-45 degrees. Predicted noise levels basedon turbine specific wind speeds and G =0.


14/20


Site C

The prediction methodology for this site was based on a mixed ground cover(ISO 9613 G=0.5 for the source, middle and receiver ground) based on siteobservations described above.

The results obtained for site A and B suggest that adding a 2 dB uncertaintymargin, typical of the commercial warranty margin used for some turbine models, tothe raw tested sound power values, is likely to lead to an over-estimate of actualturbine noise immission levels. The stated test uncertainty, as required by therelevant IEC 61400-11 standard, has therefore been referenced instead. For theturbines installed at site C, the predictions were made by using the measured soundpower data and adding the corresponding stated uncertainty value of 1 dB.

Figure 4 shows the noise levels measured at position 1, situated 100 m from to theclosest turbine on a reflective board. At this position, the influence of backgroundnoise levels is minimal. However, comparisons with predictions at this location were

complicated by the proximity of the measurement location to the turbines whichmakes it unlikely it could lie downwind of every turbine simultaneously. The effect ofthe wind speed variations across the site (described below) was also identified assignificant.

15

20

25

30

35

40

45

50

55

60

65

3 4 5 6 7 8 9 10 11 12

Wind Speed at 10m (m/s)

NoiseLevelL90dB(A)

Figure 4: Levels measured at site C, position 1. Green points correspond to high-speed operation ofthe site, and purple points to some turbines operating at low-speed or not at all.

The 2-speed pattern of operation of the turbine installed at site C is clearly apparentin Figure 4, with the turbines operating in a lower speed in lighter wind, and a higherfixed speed (with corresponding louder noise emissions) in stronger winds.Intermediate levels correspond to times where only part of the site was operating in ahigh-speed mode. In the remainder of the analysis, this region of high-speedoperation was the focus of the study, in order to obtain the highest signal-to-noise


15/20


ratio. In particular, it is expected that the region where turbines have switched tohigh-speed operation while wind speeds remain moderate will correspond to theleast background-affected measurements.

Figures 5a and 5b display the levels measured at the two locations situatedrespectively 450 and 750 m from the closest turbine, along the North-East line.

Similar general observations can be made as were made for site B, above.Compared to the latter site, the more distant measurement locations of site C weresituated close to forestry and vegetation which created higher wind-relatedbackground noise levels, affecting the measurements; it was not possible to operatesite shutdowns to characterise more precisely the background noise levels, but figure5b indicates that the margin between noise levels measured in upwind anddownwind conditions is low. Comparing figures 5a and 5b, it is apparent that themargin between raw measured values and predictions is constant over the wind-speed range in the first case, but increases in the second: this suggests that thisincreased margin is not related to changes in the noise source but in background-

related effects. Therefore, in both cases the turbine immission noise levels arethought to be close or lower than the predictions.

The exact background noise level influence at each measurement location for each10 minute period cannot be known. It is however likely that the background noiselevel may have contributed approximately 1 dB(A) to the total noise levels measuredat location 3 (750 m distance). In addition, historical background data measured innearby locations suggests that the background levels measured in downwindconditions were marginally higher, because of site-specific effects, suggesting thatthe background levels described above are underestimated to some extent.

Figures 6a and 6b display a similar comparison for the measurements made at

positions 4 and 5, which were situated respectively 700 and 820 m from the closestturbine. Similar observations can be made as for locations 2 and 3 above.


16/20


15

20

25

30

35

40

45

50

55

60

65

3 4 5 6 7 8 9 10 11 12


NoiseLevelL90dB(A)

Measured (turbine +background)

Predicted turbine (G=0.5)

Poly. (Measured (turbine +background))

Figure 5a: Levels Comparison of Site C measured and predicted noise levels at measurementlocation 2 (450 m distance). All downwind angles restricted to +/-45 degrees.Predicted noise levels based on a single site wind speed reference and G =0.5.

15

20

25

30

35

40

45

50

55

60

65

3 4 5 6 7 8 9 10 11 12


NoiseLevelL90dB(A)



Upwind


Figure 5b: Levels Comparison of Site C measured and predicted noise levels at measurementlocation 3 (750 m distance). Downwind and upwind angles restricted to+/-45 degrees. Predicted noise levels based on a single site wind speed referenceand G =0.5. The low margin above background levels suggests that actual turbinenoise levels will be closer to the predictions.


17/20


15

20

25

30

35

40

45

50

55

60

65

3 4 5 6 7 8 9 10 11 12


NoiseLevelL90dB(A)




Figure 6a: Levels Comparison of Site C measured and predicted noise levels at measurementlocation 4 (700 m distance). All downwind angles restricted to +/-45 degrees.Predicted noise levels based on a single site wind speed reference and G =0.5.

15

20

25

30

35

40

45

50

55

60

65

3 4 5 6 7 8 9 10 11 12


NoiseLevelL90dB(A)


Predicted


Figure 6b: Levels Comparison of Site C measured and predicted noise levels at measurementlocation 5 (820 m distance). All downwind angles restricted to +/-45 degrees.Predicted noise levels based on a single site wind speed reference and G =0.5.


18/20


Noise levels were also measured at several positions surrounding the site, atdistances of 700 to 1000 m: for these locations, complexities due to the influence ofbackground noise levels hindered the interpretation of the results, but observationswere generally consistent with the comments made above.

Finally, observations could be made for site C in terms of the difference between the

single wind speed reference and the wind speed of each of the turbines, with thefollowing results:

oMean difference = -0.4 m/s

oStandard deviation of differences = 1.4 m/s

oMaximum decrease = -8.9 m/s

oMaximum increase = 7.8 m/s

Figure 7 demonstrates the trend that the reduction in effective wind speeds betweenthe upwind and downwind extremities of the wind farm site tends to be higher withincreasing wind speeds. These observations are similar to those made for site B.

-8

-6

-4

-2

0

2

4

6

0 1 2 3 4 5 6 7 8 9 10 11 12 13

Upwind turbine wind speed (standardised to 10m)

Differencebetweenupwindanddownindturbine

Figure 7: Difference between the wind speed measured at hub height (standardised to 10 m)between the most upwind and the most downwind turbine at site C, over a period ofapproximately 3 months. A negative number indicates that the downwind wind speedis lower.


19/20


Conclusions

The results of the study of noise emissions from large operating wind-farm sites havesupported the view that engineering methods such as ISO 9613 offer a robust meansof determining the upper turbine immission levels that may occur in practice under

favourable, downwind propagation conditions.Detailed analyses made to date of completed large-scale measurement studiesillustrate the extent of conservatism that may be inherent to certain predictionmethods and choices. Our studies also illustrate the difficulties encountered in noiseimmission measurements, and in particular evaluating the measurement contributiondirectly attributable to turbine immissions alone and defining a relevant wind speedreference. Measurements made closer to the source can help in evaluating thedifferent contributions within the measurements.

Predictions using relatively conservative methods tend to equal or exceed totalmeasured noise levels in practice. In particular, the addition of relatively high

uncertainty margins corresponding to the use of commercial warranted sound power,as well as the choice of pessimistic propagation parameters, both have thepropensity to result in significant design conservatism. Whilst these conservatismsmay seem numerically small, and will be of limited significance subjectively, theconsequences in power generating losses can be substantial.

The results are in line with recommendations for best practice in wind farm noisepredictions as recently set out by several practitioners in the field in the UK[7]. Whilstcertain choices of parameters have been found to be effective in practice, furtherdetailed studies would be required to determine exact propagation effects occurring.

The findings have also shown that the assumption of a single wind speed reference

for all turbines that form a large wind farm site may overestimate the actual windspeed seen by each individual turbine. This is particularly the case for the turbinesnearest to a location of interest which may be partly shielded by the furthest upwindturbines which experience uninterrupted (by the wind farm) and higher wind speedconditions. This means that a single wind speed reference will likely overestimate thesound emissions of the turbines nearest to a location of interest. This effect appearsto be most significant at higher wind speeds for the sites studied.

In summary, better knowledge of the relationship between predicted and actual noiseimmission levels has the potential to reduce un-necessary conservatism and enablesubstantially enhanced generating capacity during the design phase of a wind farm.

This requires that careful account is taken of the specifics of each site underconsideration and that compatible design choices are made to avoid cumulativepessimism which may be unlikely to simultaneously occur in practice.

Further study works would be beneficial to identify in more detail and in isolation theinfluence of different ground conditions, of more complex terrains profiles, and othertypes of turbines such as variable speed machines. Further study of directionalpropagation effects would be beneficial given their relevance to large wind farm siteand cumulative impacts.


20/20


References

[1] ISO 9613-2 Acoustics Attenuation of sound during propagation outdoors Part 2: General method of calculation, International Standards Organisation, ISO9613-2, 1996.

[2] J OR3-CT95-0091 Development of a Wind farm Noise Propagation PredictionModel, Bass J H, Bullmore A J , Sloth E, Final Report for EU Contract

J OR3-CT95-0051, 1998.

[3] Wind Farm Noise Predictions - The Risks of Conservatism, J . Adcock, A.Bullmore, M. J iggins and M. Cand, Second International Meeting on Wind

Turbine Noise, Lyon, France, September 2007

[4] ETSU W/13/00386/REP Noise Measurements in Windy Conditions, Davis R A,Lower MC, 1996

[5] BS EN IEC 61400 Part 11:2003. Wind turbine generator systems - Part 11:Acoustic noise measurement techniques. British Standards Institute.

[6] The Assessment and Rating of Noise from Wind Farms - ETSU-R-97. TheWorking Group on Noise from Wind Turbines. September 1996.

[7] Prediction and Assessment of Wind Turbine Noise - agreement about relevantfactors for noise assessment from wind energy projects, Proceedings of theInstitute of Acoustics (Acoustics Bulletin), Vol34, No 2, March/April 2009, pp35-37; Bowdler, Bullmore, Davis, Hayes, J iggins, Leventhall, McKenzie.

wtn09_bullmore

Documents