Effect of wind speed and wind direction on amplitude modulation of wind turbine noise. Thileepan PAULRAJ1; Petri VÄLISUO2;

Transcription

1 Effect of wind speed and wind direction on amplitude modulation of wind turbine noise Thileepan PAULRAJ1; Petri VÄLISUO2; 1,2 University of Vaasa, Finland ABSTRACT Amplitude modulation of wind turbine noise occurs intermittently at the location of the receiver who is at a considerable distance from the turbine. This phenomenon could be because the weather conditions affect the noise source and the propagation of noise between the source and the receiver. Hence it becomes necessary to understand how amplitude modulation of wind turbine noise depends on the weather conditions. In order to understand this relationship, this paper utilizes the method proposed by the Amplitude Modulation Working Group (AMWG) of the UK Institute of Acoustics to detect amplitude modulation of wind turbine noise and measure it s depth. Further analysis will find out how the speed and direction of the wind affects the amplitude modulated sound pressure levels (SPL). Long term measurement of wind speed (WS), wind direction (WD) and sound pressure time series recording has been done since 2016 at Kikokkalio wind farm in Finland. The data used in this paper is collected approximately 1 kilometer from the nearest wind turbine. Periodic amplitude modulation within a certain frequency range is assumed to come only from wind turbines. The accuracy of amplitude modulation detection methods are uncertain, as demonstrated by Larsson and Öhlund (2014). Hence, AMWG s algorithm will also be tested on noise data which were collected from places far away from the influence of wind turbine noise. Keywords: Wind turbine noise, Amplitude modulation, Environmental noise, Social acceptance I-INCE Classification of Subjects Number(s): 11,12,24,61,71 1. INTRODUCTION Periodic fluctuations in the level of audible noise from one or more wind turbines is defined as amplitude modulation[1]. These fluctuations are perceived as regular swishing or thumping noise depending on how far away the perceiver is from the turbine or wind farm [3]. These fluctuations or modulations of audible noise level occur at the blade passing frequency of the turbine[4]. The swishing noise from wind turbines is due to the trailing edge noise of the turbine blades and is directed towards the direction of rotation of the blade [4]. Renewable UK terms swishing noise as Normal Amplitude Modulation (NAM) and the thumping phenomenon as Other Amplitude Modulation (OAM). Although the mechanics behind the thumping phenomenon is not completely understood yet [2][4] Renewable UK mentions that OAM is observed occasionally at large distance from the turbine and is dependent on atmospheric factors including wind speed and direction [4]. A receiver, at a considerable distance (approximately 1 KM) from the wind turbine perceives intermittent AM which could also be due to the effect of weather on both the noise source and the propagation path. Larsson and Öhlund also confirm that AM is more common under certain meteorological conditions [2]. 1 thileepan.paulraj@uva.fi petri.valisuo@uva.fi 2

2 Residents living in the vicinity of a wind farm often express annoyance from wind turbine noise. In Finland, there is an increased concern about the effect of this noise on the soundscape and the landscape of locations in the vicinity of a wind farm [6]. Results of the survey conducted among residents living between 2.5 and 10 km from 5 wind farms in Finland by authors in [6] indicate that % of the respondents were somewhat worried about wind turbine noise and % think low frequency wind turbine noise is hazardous to human health. Van Den Berg in [3] has recorded the annoyance reported by a Dutch resident living at 1.5 km from a German wind farm in GermanyNetherlands border. The author argues that, on quite nights when the turbines rotate at high speed, the noise can be heard several kilometers away from the wind farm and on the same days thumping noise can be heard between 500 and 1000 meters. Authors in [5] have conducted a listening test among 30 participants using modified recorded wind turbine noise. Their results indicate that amplitude modulation of wind turbine noise has a statistically significant effect on annoyance and the annoyance increases with the increase in amplitude modulation. The annoyance response to wind turbine AM is dependent on modulation depth, modulation rise time, modulation frequency, average sound level etc., [8]. Thus, it becomes clear that AM of wind turbine noise is dependent on weather and annoys people. Therefore, studying the effect of particular weather parameters on AM becomes important. Hence, this paper aims to study the effect of wind speed and wind direction on the depth and frequency of occurrence of AM in wind turbine noise in the vicinity of a wind farm in Finland. Figure 1: An aerial view of the Kirkkokallio wind farm showing the location of the 9 turbines and the microphone (courtesy- Google Earth)

3 Continuous noise recordings are carried out at 3 locations around the Kirkkokallio wind farm in Honkajoki, Finland since The farm has 9 Nordex N117/2400 turbines with a rated output of 2.4 MW each. The hub height of the turbines are 120 m and the rotor diameter is 117 m. The noise data used in this paper is measured through a microphone located at a road called Ristilantie close to the wind farm. Figure 1 shows the locations of the 9 turbines in the wind farm, the location of the microphone and the SODAR (SOnic Detection And Ranging) weather data measurement equipment. The distance to the closest turbine (turbine1) from this microphone is about 1.1 Km. Noise recording through this microphone started on and is still ongoing. The noise data analyzed in this paper is from to One G.R.A.S 46AE microphone is used to record noise at a sampling frequency of Hz combined with a 24 bit analog to digital converter. In this paper amplitude modulation (AM) is identified and depth of AM quantified in the analyzed noise data. Wind speed and direction data corresponding to the time instances when AM was detected is used to study the relationship between wind speed, wind direction and AM. The sound pressure levels (SPL) measured when AM was detected is also matched with the wind speed and direction data to see how SPL is affected by the weather data when AM was detected METHODS Amplitude modulation (AM) detection method This paper uses the amplitude modulation detection and depth measurement method proposed by the Amplitude Modulation Working Group (AMWG) of the UK Institute of Acoustics (IOA). This method can measure and assess AM in large upwind turbines with 3 bladed rotors with a rotational speed of 20 RPM. However according to IOA [1] it can also capture all the first 3 harmonics in the noise measured from a turbine rotating at 32 RPM. This method not only identifies amplitude modulated wind turbine noise within the analyzed sound data but also measures the depth of the modulation. The recorded noise data is stored in time series format in HDF5 files with each file containing data for 5 minutes. The noise data from these files are read and converted into L Aeq values of 100 millisecond chunks to be used as the input data for the AMWG s method. AM is measured in three frequency bands Hz, Hz and Hz. Whichever frequency band exhibits higher amplitude modulation levels are used for further analysis. The AM detection method splits the input time-series data into 10 second (minor time interval for analysis) blocks. The method transforms this data blocks into the frequency domain using Fast Fourier Transform (FFT) to obtain the modulation spectrum. In the resultant spectrum if a peak is present at the fundamental frequency which is equal to the blade passing frequency of the turbine then a window around this frequency and the first and second harmonics is selected for further analysis. If no peak is detected in the power spectrum then that 10 second block is rejected. An inverse Fourier transform is applied to the filtered window and the modulation depth is calculated by subtracting the L 5 from L 95 of the reconstructed time series data. 60 successive modulation depth values of 10 seconds blocks are aggregated. Out of these 60, if at-least 30(50%) of the 10 second blocks have an amplitude modulation depth value, then the 90 th percentile of the distribution of these 30 values is reported as the amplitude modulation depth of the aggregated 10 minute period. If less than 50% of these 10 second blocks are amplitude modulated then the 10

4 minute period is reported as not amplitude modulated.[1] In the remaining part of this paper all the 10 minute periods with an AM depth value will be referred to as AM incidents. Complete exclusion of background noise from wind turbine noise cannot be achieved through this method. So, in order to identify if the peak in the power spectrum is created by wind turbine noise amplitude modulation and not by other background noise, this method uses a power spectrum peak prominence detection method. This prominence detection method calculates the amplitude of the power spectrum peak(l pk ) at the fundamental frequency and divides it by the average amplitude(l m) of the power spectrum at the 2 nd and 3 rd frequencies on either side of the peak. While calculating L m, the amplitude of the power spectrum at frequencies immediately adjacent to the power spectrum peak is neglected. If the ratio L pk/ L m is greater than or equal to 4 then the peak is considered a prominent one. [1] 2.2 Matching amplitude modulation (AM) detection method s output and weather data All the recorded noise data between and were analyzed using AMWG s AM detection method. The analysis gave us both 10 seconds and 10 minutes amplitude modulation depths for the entire period in all the 3 frequency bands (50 to 200 Hz, 100 to 400 Hz and 200 to 800 Hz). The 10 minute amplitude modulation depth values in the frequency band 50 to 200 Hz were chosen for further analysis since the depth of AM was high in this band for the entire analysis period. Figure 2 shows the density plot of AM depth for the entire analysis period. The peak AM depth for the bands 50 to 200(Hz), (Hz) and (Hz) are 3.19(dB ), 2.69(dB) and 2.24(dB) respectively. The weather data used in this paper are measured using SODAR (SOnic Detection And Ranging) measurement equipment about 3.4 km North-East from the microphone location and about 1.1 km North-East from the nearest wind turbine (turbine8) in the wind farm. The location of the SODAR could also be seen in figure 1. The SODAR measures wind speed and wind direction at 31 heights starting from 50 m from the ground level to 200 m with 5 m increment in steps. Wind speed and direction are recorded as 10 minute averages. SODAR measurements started on and is still on-going. The wind speeds measured at 120 meters from ground level is used in this paper because this height matches the hub height of all the turbines at Kirkkokallio wind farm. Sound pressure level (SPL) values are calculated from the same data collected by the microphone The time stamps of the AM depth values in the 50 to 200 Hz band is matched with the time stamps of the wind speed and direction values and only those values are used for the statistics done in the remaining part of this paper.

5 Figure 2 - Amplitude Modulation dept from to in the frequency bands Hz, Hz, Hz 3. RESULTS 3.1 Accuracy of the amplitude modulation detection method In order to determine that the accuracy of the AM detection method of AMWG the same method was tested on the data collected approximately 9.5 KM north east from the Santavuori wind farm in Ilmajoki, Finland. This data was collected between to for approximately 36 hours. Results indicate that the AMWG s AM detection method detects AM for 1.24% of the time in all the three pass bands ( (hz), (Hz) and (Hz)). Hence there is some uncertainty in this method. 3.2 Relationship between AM, wind speed and wind direction After detecting amplitude modulation and measuring it s depth in the recorded noise data further analysis was done to study if wind speed and wind direction has any influence on AM. Analysis indicate that amplitude modulation depth was high when the wind blew from certain directions. Figure-3 shows the relationship between wind speed, wind direction and AM depth for the entire analysis period. In figure 3 the outer circle indicates wind direction with 0 degree representing North, 90 degrees representing East, 180 degrees representing South and 270 degrees representing West. The inner circles represent wind speed in (m/s). The color of the dots represent AM depth of the

6 wind turbine noise. This figure shows with little clarity that more number of high amplitude modulation depths (between 4.5 db to 8.5 db) were detected when the wind blew from south west and west directions. Figure 3 - Relationship between wind speed (at 120m from ground level), wind direction and AM depth. Higher AM depths are concentrated on the south-west and west directions ( º). The relationship between wind direction and AM depth is seen clearly in the box-plot shown in figure 4. The figure shows a clear increase in the maximum and median amplitude modulation depths when the wind blew from south-west and west directions. To confirm this relationship, a ttest was conducted between the amplitude modulation depth data when the wind was blowing from the west and south west direction and the same data when the wind was blowing from all other directions. The resultant p-value of the test was Hence we can say with more than 95% confidence that the noise from the wind turbines were behaving differently when the wind was blowing from the west and south west directions.

7 Figure 4 - Relationship between wind direction and AM Figure 5: Relationship between wind speed (m/s) and amplitude modulation depth (db)

8 Figure 5 is a box plot of the amplitude modulation depths with respect to each observed wind speeds for the entire analysis period. No visible pattern was observed to prove the dependency of amplitude modulation depth on wind speed. Figure 6 shows the frequency (left y-axis) and percentage (right y-axis) of amplitude modulation incidents occurred during the period. The number of times AM was detected at a particular wind speed is compared with the total number of times that particular wind speed was recorded during the entire measurement period. This comparison shows that more number of amplitude modulation incidents were detected when wind speed increased and it declined sharply after a certain wind speed. Figure 6: Percentage of AM incidents detected during the measurement period. Primary Y-axis on the left shows the frequency of occurence of wind speeds and the secondary y-axis on the right shows the percentage of AM incidents detected indicated by the red line Table 1 shows the percentage of amplitude modulation incidents detected by AMWG s method for all the 12 months of data that was analyzed. Except April, May and December 2016, AM was detected for about 30% of the time in the remaining months. The reason for lower amplitude modulation incidents during April and May, 2016 and higher AM incidents during December, 2016 is not clearly understood yet. According to our analysis the average amplitude modulation depth for the entire analysis period is 3.95 db.

9 Table 1 - Percentage of amplitude modulation incidents detected each month during the measurement period. Month Percentage of amplitude modulated 10 minute periods in the frequency band Hz April, % May, % June, % July, % August, % September, % October, % November, % December, % January, % February, % March, % 3.3 Relationship between wind speed, wind direction and Sound Pressure Level(SPL) when AM was detected. Figure 7 shows the relationship between wind speed, wind direction and A weighted sound pressure level (SPL). It includes all the wind speed, wind direction and SPL data during the times when AM was detected. Due to some error in the recording instruments, the recorded noise was corrupted between 8:40pm on to 7:54 pm on and between 06:03 am on and 08:35 am on Hence these data were not included in the analysis. High Sound Pressure Levels (SPL) were recorded at higher wind speeds. These higher SPLs could either be caused by the wind turbine noise or by the wind itself on the microphone or by the wind noise on vegetation. Due to the uncertainty of the AM detection method it is possible that non wind turbine related noise data might also be mixed in our data. Pearson and Spearman correlation coefficients between wind speed and SPL are 0.28 and 0.32 respectively. These correlation results indicate a weak correlation between the wind speed and SPL. It is evident from figure 7 that there is no concrete correlation between wind direction and SPL, since higher SPLs are mesured in all directions. Pearson and Spearman correlation coefficients indicate a very weak negative correlation between the two parameter. The coefficients are and for Spearman and Pearson correlation respectively. Hence we were not able to find any strong correlation between wind speed, wind direction and sound pressure levels (SPL) of the wind turbine noise, during AM incidents in the analyzed data. In figure 7 the outer circle indicates wind direction with 0 degree representing North, 90 degrees representing East, 180 degrees representing South and 270 degrees representing West. The

10 inner circles represent wind speed in (m/s). The color of the dots represent A-weighted SPL from 30 (dba) represented in blue to 50 (dba) represented in red. Figure 7 - Relationship between wind speed(120m from ground level), wind direction and sound pressure level (dba) when amplitude modulation incidents occurred. 4. CONCLUSIONS Analysis results of the noise and weather data collected 1.1 km away from the Kirkkokallio wind farm indicates that higher amplitude modulation depth was observed in the frequency band 50 to 200 Hz during the 12 months period from to The amplitude modulation detection method of the AMWG of the UK Institute of Acoustics (IOA) have some uncertainty. The method indicates that 1.24% of the noise data collected from 9.5 km away from a wind farm is amplitude modulated in all the three pass bands ( (hz), (Hz) and (Hz)). It is also possible that amplitude modulated wind turbine noise from Santavuori wind farm could have propagated to this measurement location. Accuracy of the AM detection method could be determined with more certainty if the method is tested using more data collected at even farther distances from wind turbines. From our analysis we can conclude that AM depth is dependent on wind direction but not on wind speed. Higher levels of AM depth were measured when the wind blew from south-west and west directions, which means the microphone was located at the downwind and crosswind directions during these times

11 with respect to the turbines. Frequency of AM incidents increase with the increase in wind speed and drops sharply after a particular speed. The reasons for this drop is neither studied not understood in this paper. Further analysis is needed to understand the reason for this drop. The average percentage of amplitude modulation incidents for the entire 12 months is 24.95%. The reasons for low percentage of AM incidents detected during April and May 2016 and high percentage of AM incidents detected in December, 2016 are not known yet. Higher sound pressure levels (SPL) were measured at higher wind speeds, but this high SPL levels could also be due to the noise of the wind on the microphone and/or the noise from vegetation. There is also no strong evidence to prove correlation between wind speed, wind direction and sound pressure levels (SPL) of wind turbine noise during the instances when AM was detected. REFERENCES 1. Institute of Acoustics Amplitude Modulation Working Group. (2016). AMWG Final report. Available from: [Accessed 23 May 2017] 2. Larsson, C., & Öhlund, O. (2014). Amplitude modulation of sound from wind turbines under various meteorological conditions. The Journal of the Acoustical Society of America, 135(1), Van den Berg, G. P. (2004). Effects of the wind profile at night on wind turbine sound. Journal of sound and vibration, 277(4), Renewable UK (2013), Wind Turbine Amplitude Modulation:Research to Improve Understanding as to its Cause and Effect. 5. Lee, S., Kim, K., Choi, W., & Lee, S. (2011). Annoyance caused by amplitude modulation of wind turbine noise. Noise Control Engineering Journal, 59(1), Turunen, A., Tiittanen, P., Taimisto, P., & Lanki, T. (2016, August). Noise annoyance and sleep disturbance in the vicinity of five wind farms in Finland. In INTER-NOISE and NOISE-CON Congress and Conference Proceedings (Vol. 253, No. 6, pp ). Institute of Noise Control Engineering. 7. Large, S. (2016, August). A quantitative and qualitative review of amplitude modulation noise from wind energy development. In INTER-NOISE and NOISE-CON Congress and Conference Proceedings (Vol. 253, No. 3, pp ). Institute of Noise Control Engineering. 8. UK Institute of Acoustics Amplitude Modulation Work ing Group. Methods for Rating Amplitude Modulation in Wind Turbine Noise. Available from: %20Discussion%20Document.pdf [accessed on 31st May, 2017]