Techniques to establish accurate background noise levels in areas affected by operational wind turbine noise Jason Baldwin 1 TNEI Unit S12, Synergy Centre, TU Dublin Tallaght Campus, Tallaght, Ireland, D24 A386 Alex Dell 2 TNEI 7th Floor, West One, Forth Banks, Newcastle Upon Tyne, England, NE1 3PA James Mackay 3 TNEI 7th Floor, West One, Forth Banks, Newcastle Upon Tyne, England, NE1 3PA ABSTRACT During the planning process for an onshore wind farm, the establishment of accurate background noise levels at nearby noise sensitive receptors is vital in determining the allowable noise limits at these locations. This is typically achieved through background noise monitoring at a selection of representative noise sensitive receptors. As onshore windfarms have become more prevalent, areas suitable for wind farm development are increasingly within the vicinity of operational wind farms. The result of this is background noise monitoring at receptors is now often influenced by nearby operational wind farms and is not a true representation of the unaffected noise levels at these locations. Current good UK practice (as set out in the Institute of Acoustics document ‘A good practice guide to the application of ETSU-R-97 for the assessment and rating of wind turbine noise’) proposes a number of solutions to account for this additional wind farm noise. One solution is to consider directional filtering or the subtraction of the predicted noise levels from the measured baseline data. Whilst both approaches are valid, it can often be unclear which approach is most appropriate for a particular scenario. Within this paper, the limitations of current good practice are discussed, and an alternative methodology is proposed which combines both approaches to establish the unaffected background noise levels. An example of where this alternative approach has been used are presented within the paper as validation .

1 jason.baldwin@tneigroup.com

2 alex.dell@tneigroup.com

3 james.mackay@tneigroup.com

ae 2022

1. INTRODUCTION

1.1. Wind farm noise guidance In the UK and Ireland, wind farm noise limits are set relative to background noise levels. Noise limits are set 5 dB above the average prevailing background noise level (measured using the L A90 10 min descriptor) subject to fixed minimum limits which apply when background noise levels are low. In the UK the relevant guidance is set out in ETSU-R-97 ‘The assessment and rating of noise from wind farms’ (ETSU-R-97) [1] which is supplemented by ‘A good practice guide to the application of ETSU-R-97 for the assessment of wind turbine noise’ published by the Institute of Acoustics (IOA GPG) [2]. In Ireland the relevant guidance is set out in the ‘Wind Energy Development Guidelines’ published in 2006 by the Department for Housing (WEDG) [3]. All three guidance documents make it clear that background noise levels should be established in absence of turbine noise. ETSU-R-97 details that “It is clearly unreasonable to suggest that, because a wind farm has been constructed in the vicinity in the past which resulted in increased noise levels at some properties, the residents of those properties are now able to tolerate higher noise levels still. The existing wind farm should not be considered as part of the prevailing background noise” , whilst the WEDG states that “Any existing turbines should not be considered as part of the prevailing background noise.”

1.2. Consideration of operational turbines Since ETSU-R-97 and the WEDG were published (in 1996 and 2006 respectively) the number of operational wind turbines has increased significantly. Data produced by RenewableUK [4] and Wind Energy Ireland [5] suggest there are now 300 operational wind farms in Ireland and 1500 in the UK. The proliferation of operational wind farms, combined with a tendency for wind farms to cluster together in areas that are technically and environmental suitable for such developments, means that it is increasingly common for background noise measurements to be taken at locations which are potentially influenced by operational wind turbine noise levels. In 2013, the IOA GPG provided some additional guidance setting out four methods which can be used to derive appropriate background noise levels in the presence of operational wind turbines: 1) “switching off the existing wind farm during the background noise level survey (with associated significant cost implications)” 2) “accounting for the contribution of the existing wind farm in the measurement data e.g. directional filtering (only including background data when it is not influenced by the existing turbines e.g. upwind of the receptor, but mindful of other extraneous noise sources e.g. motorways) or subtracting a prediction of noise from the existing wind farm from the measured noise levels” 3) “utilising an agreed proxy location removed from the area acoustically affected by the existing wind farm/s” or

ae 2022

4) “utilising background noise level data as presented within the Environmental Statement/s for the original wind farm/s (the suitability of the background noise level data should be established).” Whilst all four methods can and are used, this paper focuses on the third option, specifically the inclusion of the word ‘or’. The text as written implies that assessments should consider the use of a directional filter or the use of subtraction; this paper considers the strengths and weakness of the two approaches and explores whether both options can be combined. 1.3. Data This paper presents analysis of measurements undertaken by TNEI proximate to an operational wind farm. All data has been collected and analysed in accordance with the guidance set out in ETSU-R- 97 and the IOA GPG. Average 10 minute wind speed data was measured at wind turbine hub height and standardised to 10 m height using a roughness length of 0.05 m as recommended in the IOA GPG. Noise levels were measured as 10 minute L A90 values using equipment which complies with all the requirements set out in ETSU-R-97 and the IOA GPG. Concurrent noise and wind condition data were collected over several weeks resulting in a dataset which covered a range of wind speeds and wind directions. Data potentially affected by rainfall was removed for the dataset. 1.3. Terminology The following terms are used throughout this paper (all expressed as dB L A90 10 minutes ): Background Noise (BN) – The noise level measured in the absence of turbine noise Predicted Wind Turbine Noise (WTN) – Predictions of the specific noise level from the turbines Total Noise (TN) – The noise level measured (which is comprised of background noise and turbine noise) Calculated Background Noise (CBN) – This is calculated by logarithmically subtracting WTN from TN 2. METHODOLOGY

The measured data have been analysed using three different methods: Method A – CBN levels have been calculated for each individual 10 minute period during the survey. Wind speed and direction data was used to predict WTN levels at the measurement location using the ISO 9613-2 propagation methodology [6] and the guidance in the IOA GPG. The predicted WTN for each 10 minute period was then logarithmically subtracted from the measured TN level to determine

ae 2022

the CBN. Where WTN exceeds TN then it is not possible to calculate the CBN. In such circumstances the BN is assumed to be 10 dB below the TN; this is considered to be a cautious approach. Method B - The data have been filtered to exclude all periods when the measurement location was downwind or crosswind of the operational wind turbines. Only data collected when the measurement location was upwind of the operational turbines (+/- 45 degrees). Method C – This method combines methods A and B with data collected when the measurement location was downwind and crosswind of the operational turbines removed, and subtractions used to ensure that the influence of the WTN on the remaining data was also considered.

3. RESULTS

The unedited TN data collected at the measurement location is presented below as Figure 1. Wind speed is correlated with wind direction (top graph) and then with noise level (bottom graph). Each blue data point represents a 10 minute period, the average noise level is shown as a line of best fit (derived using a third order polynomial). Predicted worst case (downwind) WTN is also shown (as the solid green line). Comparison of the predicted WTN levels with the measured TN levels suggests that the operational turbines have the potential to influence the measured level, particularly under worst case propagation conditions (when the wind is blowing from the turbines towards the property, ~15 degrees in this case). Use of the data without further analysis may overestimate the background noise levels. Figure 2 presents the results of the analysis undertaken using Method A. The black data points represent CBN. On inspection of Figures 1 and 2, it can be seen that CBN levels are lower than the TN levels (as would be expected). It can however be seen on Figure 2 that a split has appeared in the data for wind speeds in the range of 4 to 9 m/s. This has occurred where a 10 dB subtraction has been applied to the TN data (because WTN was greater than TN). This cautious approach is likely to underestimate BN levels for some datapoints. Figure 3 presents data which has been filtered using Method B. Only data collected when the measurement location was upwind of the turbine has been retained. To add some context, predicted WTN for the upwind wind direction (~195 degrees) are shown as a dashed green line. The directivity of wind turbine noise is discussed in the IOA GPG which suggests (in Section 4.4.2) that upwind levels (presented in Figure 3) can be 10 dB lower than downwind levels (presented originally in Figure 1). Whilst filtering out data when the measurement location was downwind and crosswind of the operational wind turbines leaves the data which should be influenced the least by the operational turbines, it is still possible that some of the remaining TN still comprises noise from the wind turbines. Figure 4 sets out the results obtained using Method C, the TN data presented in Figure 3 has then be corrected using predicted WTN levels to derive CBN data. A comparison of polynomials derived in Figures 1 – 4 is presented in Figure 5.

ae 2022

ae 2022

Figure 1 : Measured TN level data

oe

Figure 2: Calculated CBN level data (Method A)

ae 2022

Figure 3: Filtered TN level data (Method B)

Figure 4: CBN level data (Method C)

ae 2022

Figure 5: Comparison of polynomials 4. DISCUSSION

4.1. Consideration of results The analysis suggests, in relation to the data presented in this paper, that:  Use of the unedited data would not be appropriate given the potential for turbine noise to be influencing the measured level.  Use of Method A may be unduly cautious given the use of a 10 dB correction when WTN is greater than BN.  Use of Method B may not fully account for the influence of the operational turbines which may still have a small impact on the measured levels even during upwind conditions. This may overestimate the true background noise level.  Use of Method C filters out data which is likely to be heavily influenced by the turbines and provides appropriate corrections for the remaining data.

4.1. Wider considerations There are a number of important wider considerations which are not covered in the IOA GPG and have not been explored in this paper either:  There are a range of angles which can be used to filter the data. The optimum angle is likely to depend on how much data has been collected, the location of the operational wind turbines relative to the measurement location (i.e. are all the turbines in a single direction or do they surround the location) plus the predicted WTN level and how that compares to the measured TN levels. It may be that several options will need to be explored / compared by the assessor.

 How low wind speeds should be considered, sound power level data is not always available at lower wind speeds ‒ should assumptions be made to enable the calculation of CBN for wind speeds where WTN can not be predicted?  The variation in wind speed between the operational wind farm site being considered and the proposed wind development being assessed (noting that the wind speed measurements would usually be undertaken on the later but that it is the wind speed on the former that will determine the WTN).  What should be done when WTN is greater than TN, is a 10 dB correction the most appropriate approach? 4. CONCLUSIONS

The need to undertake background noise monitoring in the proximity of operational wind turbines is becoming a more frequent occurrence. The analysis presented in this paper demonstrates that whilst the wording of good practice suggests that assessments can consider the influence of nearby operational turbines by undertaking directional filtering or by subtracting predictions of wind turbine noise, it is possible to combine both approaches. Whilst a combined approach was found to be useful for the dataset presented in this paper, there are a number of factors which will determine the optimum approach for a given noise monitoring location and assessments will need to be undertaken on a case by case basis. 6. REFERENCES

1. ETSU for the DTI (Department of Trade and Industry. The Working Group on Noise from Wind Turbines ETSU-R-97 The Assessment and Rating of Noise from Wind Farms’ . 1996. 2. Institute of Acoustics. Good Practice Guidance on the application of ETSU-R-97 for wind turbine noise assessment . 2013. 3. Department of the Environment, Heritage and Local Government. Planning Guidelines . 2006. 4. RenewableUK, https://www.renewableuk.com/page/WindEnergy (last accessed 29-04-2022) 5. Wind Energy Ireland, https://windenergyireland.com/about-wind/facts-stats (last accessed 29-04-

2022) 6. International Standards Organisation. ISO9613:1996 ‘Acoustics – Attenuation of sound during propagation outdoors’ – Part 2: General method of calculation . 1996.

ae 2022