The effect on noise emission from wind turbines due to ice accretion on rotor blades

(1)

The eﬀect on noise emission from wind turbines

due to ice accretion on rotor blades

TRITA-AVE-2012:62

Peter Arbinge

Stockholm, July 6, 2012

Master’s Project in Technical Acoustics, 30 credits.

The Marcus Wallenberg Laboratory for Sound and Vibration Research, Department of Aeronautical and Vehicle Engineering,

(2)

Abstract

Swedish EPA (Naturvårdsverket) noise level guide-lines suggest that equivalent (A-weighted) noise levels must not exceed 40 dB(A) at residents. Thus, in the planning of new wind farms and their location it is crucial to estimate the disturbance it may cause to nearby residents. Wind turbine noise emission levels are guaranteed by the wind turbine manufacturer only under ice-free conditions. Thus, ice accretion on wind turbine may lead to increased wind turbine noise resulting in noise levels at nearby residents to exceed 40 dB(A).

The purpose of the project is to evaluate the eﬀect on wind turbine noise emission due to ice accretion. This, by trying to quantify the ice accretion on rotor blades and correlate it to any change in noise emission. A literature study shows that the rotor blades are to be considered the primary noise source. Hence, ice accretion on rotor blades are assumed to be the main influence on noise character.

A field study is performed in two parts; as a long term measurement based on the method out-lined by IEC 61400-11 and as a short term measurement in strict accordance with IEC 61400-11. These aim to obtain noise emission levels for the case of icing conditions and ice-free conditions (reference conditions) as well as background noise levels.

An analysis is performed, which sets out to correlate ice measurements with wind turbine performance and noise emission. Data reduction procedures are performed according to IEC 61400-11.The apparent sound power levels are eval-uated. This is performed for the case of icing conditions as well as for the case of ice-free conditions. A statistical evaluation of icing event is carried out.

The results show that ice accretion on wind turbine (rotor blades) may leads to drastically higher noise emission levels. The sound power levels show an increase of 10.6 dB at 8 m/s. However, this can occur at all wind speeds from 6 m/s to 10 m/s. Higher levels of noise, (55 to 65 dB) may be caused by very small amounts of ice accretion. Occurrences of higher levels of noise, in the range of 50 to 65 dB, are not common. Noise levels exceeding 50 dB are to expected 10,3% of the time during the winter or 3% of the time during one year. Correlation between measured ice accumulation and noise level is weak apart from large amounts of ice. This due to statistical noise.

Taking into account the noise level guide-lines of 40 dBA at residents, as is recommended by Swedish EPA (Naturvårdsverket), the increased levels of wind turbine noise under icing conditions may force the power production to a halt.

(3)

Acknowledgements

I would like to thank Daniel Gustafsson at Vattenfall and Paul Appelqvist at ÅF Sound & Vibration for the opportunity of doing this master’s project with their respective companies. Further, I would like to thank professor Mats Åbom at KTH for providing theoretical insight to the acoustical diﬃculties of this project. I would also like to thank Jan-Åke Dahlberg, Peter Krohn and Ingemar Forsgren at Vattenfall for their time and eﬀort of guiding me through the measurements and to interpret data. Last, but no least, I would like to thank Manne Friman, Jens Fredriksson and Niklas Törnqvist at ÅF Sound & Vibration for the time taken to explain and discuss various matters of the project, theoretical as well as practical.

Peter Arbinge

(4)

Introduction

1.1 Background

In Sweden electric power generated by wind is growing strongly. In 2010 wind turbines produced 3.5% of Sweden’s total electric power production. In 2011 the electric power produced by wind power was 6.1% of the total power production. During 2012 it is estimated that more than 8% of Sweden’s power production is from wind power.

Although the benefits of wind power being many, and with the environ-mental aspect of “clean electricity” perhaps being the largest, there are also consequences. Wind turbine noise, and its eﬀect on physical and mental health is frequently debated. Through Swedish EPA (Naturvårdsverket) noise level guide-lines have been established, suggesting that the equivalent A-weighted noise level (caused by wind turbines) may not exceed 40 dB at residents. To keep wind turbine noise levels from exceeding 40 dB, diﬀerent types of measure-ments and calculations may be performed. These can be:

• immission measurements (at resident) • emission measurements (at wind turbine)

• noise emission/immission calculation (noise propagation)

The first and second types require in situ measurements, thus there must be one or more wind turbine(s) acting as measurement object(s). These may therefore, only verify actual noise levels. The method of measuring noise emission, is out-lined in international standard IEC 61400-11 [1].

The third type aims to evaluate noise levels at a resident by calculating how the noise propagates. This by considering the atmospheric conditions (wind di-rection and speed, atmospheric pressure, air humidity etc) as well as the topog-raphy surrounding an actual or an imaginary wind turbine. This may require an in situ measurement to verify the actual noise emission, but may also be calculated by using the guaranteed noise levels (sound power levels) provided

(8)

by the wind turbine manufacturer. The latter proves to be important when planning new wind farms and the location of the individual wind turbines.

The wind turbine noise levels (sound power levels) provided by the wind turbine manufacturer, is guaranteed only under ice-free conditions. This is assumably because of the risk of higher noise levels, which would exceed the guaranteed levels under icing conditions. Since the guaranteed noise levels may have been used when planning the location of a wind turbine/farm, ice accre-tion on wind turbine may lead to increased wind turbine noise levels at nearby residents.

1.2 Purpose

The purpose of the project is to evaluate the eﬀect on wind turbine noise emis-sion due to ice accretion. This, by trying to quantify the ice accretion on rotor blades and correlate it to any change in noise emission. The rotor blades are considered to be the primary noise source. Thus, ice accretion on rotor blades are assumed to be the main influence on noise character.

1.3 Limitations

Regarding the analysis of wind turbine noise, and the methods presented in IEC 61400-11, the project has the following limitations:

• 1/3 octave band analyses are not part of this project. Neither are any other frequency analyses.

• Analyses of tonality are not part of this project.

(9)

Chapter 2

Method

2.1 Problem formulation

In accordance with IEC 61400-11 the noise emission of the measurement object is evaluated. By performing concurrent (time synchronized) measurements of ice accumulation near the measurement object, a correlation between noise emission and ice accretion may be established.

2.2 Literature study

The literature study includes a summary of the sound/noise sources associated with wind turbine. This is important as to understand where/if noise is gener-ated and how ice accretion may alter the emitted noise. Further, the literature study aims to describe the basics of the ice accretion mechanisms associated with wind turbines.

2.3 Field study

The field study is performed in two parts. These are:

• a long term measurement based on the method out-lined by IEC 61400-11 • a short term measurement in strict accordance with IEC 61400-11 The goal is to obtain noise emission levels for the case of icing conditions and ice-free conditions (reference conditions) as well as background noise levels.

2.4 Analysis

By gathering data during field study, the analysis sets out to correlate ice mea-surements with wind turbine performance and noise emission. Measured sound

(10)

pressure levels (of wind turbine noise) are corrected for the influence of back-ground noise. The apparent sound power levels are evaluated. This is performed for the case of icing conditions as well as for the case of ice-free conditions. Data reduction procedures are performed according to IEC 61400-11 [1]. A shorter statistical evaluation of icing event is carried out.

(11)

Chapter 3

Literature study

3.1 Wind turbine noise sources

The noise sources associated with wind turbines can be divided into aeroacousti-cal and mechaniaeroacousti-cal types. Among the aeroacoustiaeroacousti-cal noise sources, according to S. Oerlemans et al. three types can be distinguished [2]. These are tower-blade interaction noise, inflow turbulence noise and airfoil self-noise. The first two noise types have distinct noise characters, while the third type, airfoil self-noise, may vary in nature and is assumed to be dependent of the specifications of the wind-turbine. According to T. F. Brooks et al. airfoil self-noise consists of five sound generating mechanisms [3]. These depend individually on flow speed and rotor blade geometry.

3.1.1 Mechanical noise sources

Mechanical parts, usually in the form of rotating machinery, is in most cases the most obvious source of noise. The character may be broadband as well as tonal. Due to wear of gears, bearings etc, the noise character is altered over time to generate more high frequency noise as well as increased tonality.

According to S. Wagner et al., regarding wind turbines, mechanical noise levels are negligible compared to total noise levels. This, if mechanical noise is adequately treated, for instance by replacement of worn mechanical parts [4]. Also H. H. Hubbard and K. P. Shepherd conclude that for larger wind turbines mechanical noise is considered to be of little importance to the emitted noise [5].

3.1.2 Aeroacoustical noise sources

3.1.2.1 Tower-blade interaction noise

Due to tower-blade interaction, low frequency noise may be generated depending on blade passing frequency. For wind turbine configurations having rotor in

(12)

the upwind direction from tower, this noise type is considered to be of little importance compared to total noise emission. However, it might become an issue if considering total noise emitted from several wind turbines [4]. In the case of rotor placement in the downwind direction of tower, the tower wake (caused by the mean flow) interacts with rotor blades and may cause low frequency noise of higher amplitudes.1 _{Alterations in tower design to reduce tower-blade}

interaction noise have been performed with various results [6, 7]. 3.1.2.2 Inflow turbulence noise

The interaction of atmospheric turbulence and leading edge (LE) of rotor blade at turbine inflow generates broadband noise, although this is not yet fully quan-tified [4].

3.1.2.3 Airfoil self-noise

Being somewhat more complex than the previous two noise types, airfoil (rotor blade) self-noise is generated by several mechanisms and can be broadband as well as tonal in character. Self noise, in itself, assumes a constant (undisturbed) mean flow, in contrast to inflow turbulence noise (see Section 3.1.2.2). The com-plexity of airfoil self-noise may suggest that this noise type is more dependent of the turbine rotor blade specifications (angle-of attack, airfoil geometry etc) and atmospheric conditions (flow speed, radiation direction etc).

One source of self-noise is caused by the interaction of the turbulent bound-ary layer and the trailing edge (TE) of the airfoil (see Figure 3.1.1). This is usually referred to as turbulent-boundary-layer-trailing-edge noise, or TBL-TE noise. For a turbulent boundary layer to exist, higher mean flow speeds are needed. Hence, the TBL-TE noise is more prominent towards (but not exactly at) the tip of the turbine rotor blade [2]. The noise is broadband in character, and is the main source of higher frequency noise (700-2000 Hz). TBL-TE is considered to be the largest contributor to the noise emitted by wind turbines.

Figure 3.1.1: Generation of turbulent-boundary-layer-trailing-edge (TBL-TE) noise [8].

For lower flow speeds, usually towards the rotor hub, a laminar boundary layer (LBL) replaces the turbulent. As the laminar flow passes over the TE of

1_{Within the field of acoustics, this phenomenon is usually referred to as amplitude}

(13)

the rotor blade, periodic vortex shedding may couple to the laminar flow waves upstream (see Figure 3.1.2). This feedback mechanism can result in LBL-TE noise which has a quasi-tonal character.

Figure 3.1.2: Generation of laminar-boundary-layer-trailing-edge (LBL-TE) noise [8].

If non-zero angle of attack (stall) is taken into consideration, so called separation-stall (S-S) noise may occur. For small angles, this is caused by flow separation on the suction side, close to the TE of the airfoil (see Figure 3.1.3). Broadband noise is generated by shedding of turbulent vorticity at rotor blade TE.

Figure 3.1.3: Generation of separation-stall (S-S) noise at a low angle of attack [8].

For higher angles of attack (deep stall), the flow separation close to TE may generate vorticity on a larger scale (see Figure 3.1.4). This causes the airfoil, as a whole, to generate low-frequency noise.

Figure 3.1.4: Generation of separation-stall (S-S) noise at a high angle of attack [8].

(14)

For rotational motion, the highest speed is obtained at rotor periphery. Thus, at the very tip of the airfoil, this causes local flow separation, which generates turbulent vorticity. As the vorticity interacts with the TE of the tip region, tip-vortex-formation (TVF) noise is generated (see Figure 3.1.5).

Figure 3.1.5: Generation of tip-vortex-formation (TVF) noise [8]. If considering the bluntness of the TE of the airfoil, yet another noise type may arise. As the turbulent boundary layer passes over the blunt TE, vortices are shed (see Figure 3.1.6) and noise is generated. This noise type is referred to as turbulent-boundary-layer-trailing-edge-bluntness (TBL-TEB) noise. The character of this noise appears to be neither tonal nor broadband [5].

Figure 3.1.6: Generation of turbulent-boundary-layer-trailing-edge-bluntness (TBL-TEB) noise [8].

3.2 Ice accretion on wind turbines

The location of a wind turbine is crucial for its production of electric power, and hence, its profitability. A wind exposed location often means exposure to other atmospheric instabilities and/or weather conditions. Ice accretion on wind turbines, and on the rotor blades in particular, is one of many issues, which accompanies these unsteady atmospheric conditions. The eﬀects of ice accretion, or icing, can be complete loss of, or drastically reduced, power output [8].

(15)

3.2.1 Ice accretion mechanisms

The international standard ISO 12494 defines two types of atmospheric ice ac-cretion; in-cloud icing and precipitation icing [9]. Based on these definitions, M. C. Homola et al. concludes the icing mechanisms of importance to wind turbines are [10]: • In-cloud icing – Hard rime – Soft rime – Glaze • Precipitation icing – Wet snow – Freezing rain

The in-cloud icing mechanism occurs as sub-cooled water droplets (existent in clouds) freeze on contact with a surface, enabling/initiating the crystallization process. The droplet size determines the types of rime and/or glaze that form on the surface. For small droplet sizes, freezing is instantaneous, creating soft rime. For larger sized droplets, freezing is somewhat slower, resulting in hard rime. Glaze forms when liquid water is present on the surface during freezing.

Several simplified mathematical methods exist for modeling of atmospheric ice accretion [9, 10]. T. G. Myers suggests a more in-depth method [11].

Precipitation icing is a result of rain and/or wet snow freezing on contact with a surface. Precipitation icing caused by rain requires surface temperatures of below 0 °C, while wet snow sticks to surfaces at temperatures of 0 to 3 °C. Compared to in-cloud icing, the rate of mass accumulation is higher. The risk of damage to machinery is therefore higher.

According to M. C. Homola the ice accretion on wind turbine rotor blades is larger towards the tip. This because of three factors:

• higher (flow) velocities lead to higher rate of ice accumulation.

• rotor blades collect atmospheric water from larger area/volume of rota-tional plane.

• blade tips may reach low clouds, leading to in-cloud icing.

Analytical modeling of the effect of water droplet size on ice accretion of wind turbines, as well as the possible effect on noise emission caused by ice accretion, have recently been conducted by L. Fuchs at KTH[12]. The research shows that ice accretion is located in the hub area, and at around 2/3 of the blade span, mainly on the leading edge. For wind turbines acting together (as a wind farm), the rate of ice accretion on downstream wind turbines is lower at the hub area, but larger at 2/3 of blade span. Also, the research recognizes the different noise sources and their location on the wind turbine as

(16)

• monopoles, at trailing edge of blade • dipoles, at blade tip and leading edge

• quadrupoles, as turbulent vorticity arises from blade trailing edge.

3.3 Acoustical measurements of wind turbines

What follows is a summation of the relevant parts of the wind turbine noise emission measurement standard IEC 61400-11 [1]. An alternative instrumenta-tion setup and measurement procedure is presented, which is not part of the standard (see Section 4.1.4). The standard suggests measurements and mea-surement procedures to acquire

• A-weighted sound pressure level (SPL), LA (dB), with ref. 20 µPa

Measurements of SPL and of wind speeds are performed simultaneously. The wind speeds are converted to corresponding wind speeds at a reference height of 10 m and a reference ground roughness length of 0.05 m (Sections 3.3.3.1 and 3.3.3.2). This allows for a standardized analysis2_of

• apparent sound power level, LWA (dB), with ref. 1 pW

This, by assuming a point source located at rotor centre, having the same noise emission in the downwind direction as the wind turbine being measured. Both SPL and sound power level are determined at each wind integer speed from 6 to 10 m/s at 10 meters height for an individual wind turbine at atmospheric reference conditions (see Table 3.3.1).

Atmospheric reference conditions Temperature, 288 K (15 °C) Atmospheric pressure, 1013 mbar

Table 3.3.1: Atmospheric reference conditions according to IEC 61400-11. The apparent sound power level is analyzed to determine changes in source strength due to ice accretion on wind turbine. The parts of IEC 61400-11 which are utilized to a larger extent are the measurement procedures and the following data reduction procedures.

3.3.1 Instrumentation

3.3.1.1 Acoustic instruments

For the determination of the equivalent A-weighted sound pressure level, a sound level meter is utilized. This shall be of type 1 to fulfill the requirements of

2_{IEC 61400-11 allows for analysis of 1/3 octave band levels, tonality and directivity,}

(17)

the standard3_{. The sound level meter is connected to a microphone, which is}

mounted on a circular measurement board with microphone diaphragm in a plane normal to the board, with the axis of the microphone pointing towards the wind turbine (see Figure 3.3.1). This to reduce wind noise generated at the microphone and to compensate for diﬀerent ground types. The measurement board should be made from a material that is acoustically hard.

Figure 3.3.1: Mounting of the microphone on the measurement board. Top view. [1]

A primary, and if necessary, a secondary windscreen is used. The primary (cell foam) wind screen shall be of hemispherical shape and is to be centered around the microphone diaphragm (see Figure 3.3.2). The secondary (open cell foam) wind screen is placed symmetrically over the primary wind screen.

(18)

Figure 3.3.2: Mounting of the microphone on the measurement board, along with the primary and the secondary wind screens. Side view. [1]

When using the secondary wind screen, its influence on the frequency re-sponse must be documented and corrected for.

The sound measurement system, in its entirety, is calibrated immediately before and after each measurement session, using an acoustical calibrator. IEC 61400-11 requires the calibrator to be of class 1.4

3.3.1.2 Non-acoustic instruments

To perform measurements of the wind speed, according to IEC 61400-11, the nacelle anemometer may be used. It is calibrated in situ, meaning that fur-ther calibration is not necessary. The measured wind speed data from the nacelle anemometer is provided by the wind turbine control system. From tur-bine control system is also acquired electric power production data as well as wind direction data. Electric power transducers shall be of class 1.5 _Wind

di-rection transducer shall be accurate within ± 6 °. Distance measurements are performed using a laser-optic range meter.

3.3.2 Measurements and measurement procedures

3.3.2.1 Measurement position

To minimize the influence of terrain eﬀects, atmospheric conditions and wind induced noise, measurements are made close to the wind turbine. According to IEC 61400-11 at least one microphone position is required. This is referred to as the reference position and is located downwind of wind turbine (see Figure

4_{Classification according to IEC 60942} 5_{Classification according to IEC 60688}

(19)

3.3.3). This position shall be within ±15° relative to the wind direction at the time of the measurement.6

Figure 3.3.3: Standard layout of microphone measurement positions. Top view. [1]

For horizontal axis wind turbines the distance, R0, from wind turbine tower

to the reference position is calculated as

6_{Because of the noise directivity of the wind turbine rotor, as well as the wind speed}

gra-dient, downwind positions are generally considered to be exposed to higher levels of noise. Hence, a “worst case scenario” of the noise emission can be estimated. Furthermore, the replicability of the measurements is higher when performed downwind of wind turbine. If considering an upwind measurement position, this may have inconsistencies in the noise dis-tribution caused by the combination of the wind direction, the (positive) wind speed gradient and the direction of sound propagation.

(20)

R0= H +

D

2 (3.3.1)

where H is the vertical distance from ground to rotor centre and D is the diameter of the rotor itself. Hence, R0is equal to the total height of the wind

turbine, see Figure 3.3.4. A tolerance of 20% is allowed and the distance, R0,

shall be measured with an accuracy of ±2%.

Figure 3.3.4: Definition of measurement layout parameters for horizontal axis wind turbine. Side view. [1]

3.3.2.2 Acoustic measurements at reference position

At reference position the continuous equivalent A-weighted SPL, LAeq, is

mea-sured

• by a series of not less than 30 data points

• where each data point is integrated over at least 1 minute

• with at least three data points within ±0.5m/s at each integer wind speed • and with concurrent wind speed measurements

The same applies to background noise measurements. 3.3.2.3 Wind speed measurements

According to Method 2 in IEC 61400-11, the wind speed may be determined by an anemometer, at a height of between 10 m and hub height. This may be done

(21)

for measurements of both wind turbine noise and background noise. The data are to be averaged over the same time as SPL data.

To account for variations in atmospheric pressure and temperature, the wind speed is standardized (corrected) to atmospheric reference conditions (see Sec-tions 3.3.1). The wind speed correction equation is

VH= VD ✓ prefTk pTref ◆1 3 (3.3.2) where

• VH is the corrected wind speed at hub height (m/s)

• VD is the wind speed at hub height derived from the power curve (m/s)

• pref is the atmospheric pressure under reference conditions (kPa)

• Tref is the air temperature under reference conditions (K)

• p is the measured atmospheric pressure (kPa) • Tk is the measured air temperature(K)

3.3.2.4 Wind direction measurements

Wind direction is measured with a wind direction transducer to ensure that measurements are kept within ±15° of nacelle azimuth position compared to upwind. The data are to be averaged over the same time as SPL data. 3.3.2.5 Atmospheric conditions

Air temperature and atmospheric pressure is measured and recorded hourly.

3.3.3 Data reduction procedures

3.3.3.1 Wind speed

To standardize the wind speed measurement data to reference conditions re-garding height and ground type, a logarithmic wind profile is assumed. The following equation is used:

Vs= Vz 2 4ln ⇣ zref z0ref ⌘ ln⇣H_z₀⌘ ln⇣ H z0ref ⌘ ln⇣z z0 ⌘ 3 5 (3.3.3) where

• z0ref is the reference roughness length of 0.05 m.

(22)

• H is the rotor centre height (m).

• zref is the reference roughness height of 10 m.

• z is the anemometer height (m).

If the nacelle anemometer is used (see Section 3.3.1.2), from which follows that H = z, Equation 3.3.3 is reduced to Vs= Vz 2 4ln ⇣ zref z0ref ⌘ ln⇣_z_0refH ⌘ 3 5 = VH 2 4ln ⇣ zref z0ref ⌘ ln⇣_z_0refH ⌘ 3 5 (3.3.4) 3.3.3.2 Roughness length

According to IEC 61400-11 the site terrain/ground roughness length may be estimated using Table 3.3.2.

Type of terrain Roughness length z0(m)

Water, snow or sand surfaces 0.001 Open, flat land, mown grass, bare soil 0.01

Farmland with some vegetation 0.05 Suburbs, towns, forests, many trees and bushes 0.3

Table 3.3.2: Table of diﬀerent terrain types with corresponding roughness lengths.

3.3.3.3 Correction for background noise

The measured sound pressure levels shall be corrected for the influence of back-ground noise. If backback-ground noise levels are 6 dB or more below the combined level of wind turbine and background noise, a logarithmic subtraction is made using the following equation:

Ls= 10log

h

10(0.1Ls+n) ₁₀(0.1Ln)i (3.3.5)

where

• Ls is the continuous equivalent SPL (dB), of the wind turbine when

op-erating alone

• Ls+n is the continuous equivalent SPL (dB), of the wind turbine and

background noise

(23)

3.3.3.4 Bin analysis

To determine the sound pressure level at each integer wind speed, a bin analysis is utilized. Within each bin between 6 and 10 m/s

• a linear regression is performed • where each bin is 1 m/s wide

• and each bin must have data points on both sides of the integer wind speed

The individual integer wind speeds are to be combined by using a second order regression. This yields

• the A-weighted SPL at integer wind speed (k), LAeq,k(dB) under reference

conditions

from which the apparent sound power levels may be determined (Section 3.3.3.5) The same procedure is performed for the background noise at integer wind speeds between 6 and 10 m/s. To correct for background noise logarithmic subtraction is done by using Equation 3.3.5

3.3.3.5 Apparent sound power levels

The apparent sound power levels corresponding to the integer wind speeds be-tween 6 and 10 m/s is calculated using the following equation:

LWA,k= LAeq,c,k 6 + 10log

_4⇡R2 1

S0 (3.3.6)

where

• LAeq,c,kis the background corrected A-weighted SPL at integer wind speed

(k) under reference conditions.

• R1is the slant distance (m) from rotor centre to microphones (see Figure

3.3.4).

(24)

Chapter 4

Field study

The purpose of the field study is to acquire noise emission data of a wind turbine during icing conditions. This whilst trying to have an idea of the amount of ice accreted on the wind turbine. If weather conditions allow, ice accretion on wind turbine, and the rotor blades in particular, can cause severe losses in power production. Power production may drop from maximum to zero in a matter of hours. A long term noise measurement setup, which can withstand these kinds of weather conditions, enables a study of the entire icing process, from full power production to a standstill.

Complementary manual noise emission measurements are carried out ac-cording to IEC 61400-11. This as a reference of ice free conditions as well as verification and validation of the long term measurement data. Apparent sound power levels are calculated using these measurement data. Both manual (short term) noise measurements and long term noise measurement use the same mea-surement position, and diﬀer only in instrumentation setup (see Sections 4.1.4 and 4.1.5).

The use of an ice detector system which makes concurrent (time synchro-nized) measurements parallel to the noise measurements, allows for an objective quantification of the ice accretion. The ice detector can only estimate the ic-ing on the wind turbine caused by weather conditions, and does not actually measure the amount of ice on the rotor blades themselves. Still, an assumption is made that there exists a strong correlation between the ice detector indica-tions and the actual amount of ice on the rotor blades. This will be further investigated in Section 5.5.1.

4.1 Wind turbine noise measurements

4.1.1 Measurement object

The manufacturer of the measurement object is Vestas and the model is V90. This wind turbine is a horizontal axis wind turbine, with a electric power pro-duction capacity of 1.8 MW. The maximum propro-duction capacity of a Vestas

(25)

V90 wind turbine is 2 MW, but because of the weather exposed location, the measurement object has been set to a production maximum of 1.8 MW.1

Be-tween the two types, it is only a matter of diﬀerences in software configuration. Some basic information about the Vestas V90 wind turbine is given in Table 4.1.1.

Vestas V90-1.8 MW

Rated (maximum) output power 1.8 MW Cut-in wind speed 4 m/s Rated wind speed 12 m/s Cut-out wind speed 25 m/s

Wind class IEC IIA

Operating temperature 20°C Power control type Active Axis orientation Horizontal

Rotor diameter 90 m Swept area 6362 m2 Nominal revolutions 14.5 rpm Operational interval 9.3-16.6 rpm Tower height 95 m Total height 140 m

Table 4.1.1: A concise specification of Vestas V90 wind turbine.

4.1.2 Measurement site

The topography of the measurement site in the nearest 1 km is hilly, however closer to the measurement object it is more flat. Under normal conditions the site features a mixture of low pine forest and/or bushes, lower vegetation (rough grass) closer to the measurement object. During measurements, however, the site is covered in approximately 1.5 m of snow with very smooth surfaces, covering smaller bushes and trees (see Figure 4.1.2). No sand or water surfaces. The measurement site roughness length is shown in Table 4.1.2. There are no larger reflecting structure(s) nearby, however there are other wind turbines perhaps influencing the background noise levels. Since snow has covered the trees, background levels due to vegetation noise is expected to be low.

Measurement site ground type Roughness length z0(m)

Snow 0.001

Table 4.1.2: The ground type of the measurement site and the corresponding roughness length.

1_{According to the manufacturer, the V90-1.8 MW model is ideal for IEC II sites, while}

V90-2.0 MW model is suitable for IEC III sites. Wind turbine classification according to IEC 61400.

(26)

4.1.3 Measurement position

From a statistical point of view, the dominating wind direction for the wind turbine of interest is from the north-west. Converted into degrees(°), with north being the reference (0°), this wind direction would correspond to 315°. Hence, the measurement position is chosen to be south-east of the wind turbine. Using a range meter the layout of the measurement is determined, see Table 4.1.3.

Notation Distance Horizontal distance, R0 138 m

Slant distance, R1 163 m

Table 4.1.3: Layout parameters of measurement site. Notations according to IEC 61400-11.

4.1.4 Long term wind turbine noise measurement

As previously mentioned, the acoustical measurements consist of two separate measurement setups. To gather long term noise data a Sigicom-Infra Master system logs the emitted sound pressure level. For a time period starting on De-cember 1 2011 and ending on March 23 2012, noise data are logged continuously as

• equivalent A-weighted sound pressure level (SPL), LAeq or LAeq,Sigicom

(dB)

• in 1-minute (time) averages • as a total of 163,140 data points

The long term measurement setup is displayed in Figure 4.1.1. A microphone is mounted on a stand and is connected to an Sigicom Infra system which meters and logs the measured SPL data. A primary (cell foam) and a secondary (open cell foam) wind screen is used. Both are of spherical shapes. The influence of the wind screens (primary and secondary) on the measured SPL is recorded to be 0.8 dB.

(27)

Figure 4.1.1: Long term noise measurement setup and the immediate surround-ings. The Sigicom Infra system is covered by snow, and hence not visible.

4.1.5 Short term wind turbine noise measurement

The short term noise measurement is carried out according to international standard IEC 61400-11. Of interest is the A-weighted sound pressure level (SPL) at reference position (see Section 3.3.2 and Figure 3.3.3). The short term measurement is carried out during a time period starting on February 20 2012 and ending on February 23 2012. The requirements to fulfill the standard are presented in 3.3. The short term measurements logs data continuously as

• equivalent A-weighted sound pressure level (SPL), LAeq or LAeq,IEC(dB)

(28)

• a total of 4,166 data points

The short term measurement setup, and its immediate surroundings, is displayed in Figure 4.1.5. As for the long term measurement, the influence of the wind screens (primary and secondary) on the measured SPL is recorded to be 0.8dB.

Figure 4.1.2: The short term measurement setup. A photo of the microphone, on the measurement board and the immediate surroundings. The measurement object is visible in the background.

4.1.6 Background noise measurement

Background noise measurements are performed by the long term measurements. The measurement object has for various reasons been at standstill, which enables for extraction of background noise data during these time periods.

(29)

4.2 Complementary data acquisition

4.2.1 Wind turbine data

Along with the noise measurement data, concurrent (time synchronized) wind turbine data are provided by wind turbine control system as a continuous log of wind speed and direction at rotor hub, electric power production and rotor rpm. These, along with a so called power curve (see Figure 5.2.1), provided by the wind turbine manufacturer, establishes a relation between noise emission and power output. The wind turbine data are

• measured (active) electric power, Pm (kW)

• wind speed measured by the nacelle anemometer, VH (m/s)

• wind direction2_{, ✓}

nacelle (°)

• rotor rpm3_{, n}

rotor (min 1)

These data are available in both 1-minute and 1-second averages, to comply with the two diﬀerent measurements covered in Sections 4.1.4 and 4.1.5.

4.2.2 Ice data

Ice and temperature data are provided by a weather mast located approximately 70 meters from the measurement object. The measurements are performed continuously at a the height of 58.3 meters. Data are logged in 10 minute averages. The data contain:

• ice accretion data, mice(g)

• temperature data, Tc, Tk(°C, °K)

Ice data is registered by a Combitech IceMonitor. It consists of a vertical metal rod, which at its base is mounted on a weight/pressure transducer. The rod is subject to icing when weather conditions allow, and may rotate freely around its vertical axis. Thus, as ice accretes on one side of the rod, it is possible for the wind to make it rotate. The bearing is electrically heated for a constant temperature of 1°C. Table 4.2.1 presents the specifications of the IceMonitor.

Sensor surface area 0.05 m2

Measuring range 0to 50 kg or 0 to 100 kg

Accuracy ±50 g

Operating temperature range 40to 50°C Table 4.2.1: Combitech IceMonitor specifications.

2_{Although being a non-standard parameter (when utilizing IEC 61400-11), the wind}

direc-tion data may be used to verify correct wind condidirec-tions. The parameter proves to be necessary when treating measurement data obtained without manual observation.

3_{A non-standard parameter when utilizing IEC 61400-11, however may prove to be}

(30)

4.2.3 Ice prognoses data

Available are prognoses of the possibility of ice accretion at the measurement site. These are updated every hour and the resolution is 9, 3 and 1 km. These prognoses forecast weather conditions at 100 meters height relative to the mea-surement site. The parameters presented in the forecasts are not part of the data analysis, but are utilized for verification etc. The forecasts are available via an online-server and contain

• ice load, (kg/m) • production loss, (%) • temperature, (°C) • cloud water flux, (g/m2_s)

4.2.4 Weather data

Since the weather conditions, such as atmospheric pressure and air tempera-ture, eﬀect the performance of wind turbines, these factors need to be taken into consideration. IEC 61400-11, for instance, suggests a standardization of the measured wind speed with respect to both these parameters to reference conditions, see further Section 5.5.

An SMHI weather station, which is located approximately 6.7 kilometers from the measurement site, provides long term data of the atmospheric pressure as well as the relative air humidity. These weather data contain

• atmospheric pressure, p (kPa) • relative air humidity4_{, (%)}

4_{No specific notation since not part of data analysis. The relative humidity, however, is to}

(31)

Chapter 5

Analysis

To enable processing of large amounts of measurement data, the analysis is performed solely in MATLAB. The separate data types are imported from Mi-crosoft Excel sheets and compiled in MATLAB before analysis. Table 5.0.1 displays the imported raw data types, along with their respective notations (according to IEC 61400-11) and units.

Measurement data type Notation Unit

Time t min, s

Equivalent A-weighted sound pressure level (SPL) LAeq dB

Active electric power Pm W

Wind speed at hub height VH m/s

Nacelle direction ✓nacelle °

Rotor RPM nrotor min 1

Ice accretion mice g

Air temperature Tc,Tk °C,K

Atmospheric pressure p hPa

Table 5.0.1: The raw data types imported to and analyzed in MATLAB.

5.1 Time synchronization

5.1.1 Conversion of averaging times

To (time) synchronize the different kinds of data, adjustment to the same aver-aging times is necessary. Table 5.1.1 shows the averaver-aging times of the different kinds of measurement data before treatment. The target averaging time is 1 minute. MATLAB offers conversion of regular date and time to a built-in serial time, which simplifies calculations of averaging times, durations etc. The data points are assigned a certain date and time, which indicates the end of an aver-aging time period. Thus, the data points corresponding to a date and time of

(32)

2012-01-01 00:01:00, are time averages obtained over the past minute, starting at 2012-01-01 00:00:00.

Notation Averaging time t 1 min, 1 s LAeq 1 min Pm 1 min VH 1 min ✓nacelle 1 min nrotor 1 min mice 10 min Tc,Tk 10 min p 10 h

Table 5.1.1: Averaging times of the diﬀerent kinds of measurement data. The measurement data which have longer averaging times than the target time of 1 minute, is treated with linear interpolation to fill out empty time points. To determine the interpolants in-between two subsequent raw data points, the formula used is

xi= x1+ i

x2 x1

t2 t1

, t2 t1= Tavg, i = 1, 2, 3...Tavg (5.1.1)

where x1and x2are two known subsequent raw data points, t1and t2are the

corresponding time points and Tavg is the averaging time before interpolation.

The formula is iterated to yield interpolants for each two subsequent points of the raw data, for all data points. The averaging time before interpolation, Tavg,

is either 10 min or 60 min, and consequently i = 1, 2, 3...10 or i = 1, 2, 3...60 respectively.

Since the SPL data, LAeq, have shorter averaging times than the target time,

these need to be adjusted by logarithmic averaging.

5.1.2 Synchronization and trimming of data

The data type series are checked to determine the earliest and the latest common time points for which there is a complete set of data. The redundant, incomplete, data are removed and the time synchronized data are compiled into matrix form.

5.2 Measured electric power

According to IEC 61400-11, there exists a strong correlation between the mea-sured electric power and the meamea-sured sound pressure level (SPL), and hence the sound power of the wind turbine, in the power production range of 5 95% of rated power. For the measurement object, having a rated power of 1.8 MW this corresponds to

(33)

• a measured electric power, Pm, of 90 1710 MW

Thus, a corresponding filtering criterion is used to evaluate the measurement data with respect to its power production (Table 5.2.1). Time periods of the wind turbine not running, may be used to evaluate background noise levels.

Wind turbine production state Filtering criterion Wind turbine running 90 MW_{ P}m1710 MW

Wind turbine not running Pm0 MW

Table 5.2.1: Electric power production states and the corresponding filtering criteria.

5.2.1 Power curve

As previously mentioned, the wind turbine power curve holds the correlation between the wind speed at hub height and the electric power production. The standard power curves of a Vestas V90-2.0 MW, along with the power curve of the measurement object, a Vestas V90-1.8 MW, is shown in Figure 5.2.1. The latter is used in the subsequent analysis.

Figure 5.2.1: Power curves of Vestas V90-2.0 MW and Vestas V90-1.8 MW. The power curves are provided by the wind turbine manufacturer (Vestas) as discrete values at integer wind speeds in the range of 0 to 30 m/s. To find points on the power curve in-between the discrete values, a linear interpolation

(34)

i performed. This proves to be far more accurate, and time saving, compared to the option of performing a polynomial fit of the power curve and solving for each wanted value. The linear interpolation is performed according to the following equation

VD= Vi+ (Vi+1 Vi)

Pm Pi

Pi+1 Pi

, i = 0, 1, 2...29 (5.2.1) where VDis the wind speed derived from the power curve, Pmis the measured

electric power, Pi is a discrete electric power value on the power curve and Vi

is the corresponding integer windspeed. Since Vi+1 Vi= 1for all values of i,

Equation 5.2.1 is simplified to VD= Vi+ Pm Pi Pi+1 Pi , i = 0, 1, 2...29 (5.2.2)

5.3 Wind

5.3.1 Wind direction

Measurements of wind turbine noise emission, is to be performed downwind of the wind turbine. Because of its active wind direction control system, the wind turbine is assumed to be angled towards the incoming wind. This, at all time points when actively producing electric power. Consequently, the nacelle direction ✓nacelle, may be utilized to determine the wind direction. The reference

direction of 0°, corresponds to north.

As discussed in Sections 3.3.2 and 4.1.3, the wind direction, according to IEC 61400-11 and the measurement position chosen in the field study, is optimally 311_{± 15°. However, analyses of wind turbine noise directivity suggest that the} tolerance span of ±15° may be increased to ±30° with only negligible diﬀerences in measured sound pressure levels [13]. Hence, in the subsequent analysis ±30° is considered the default wind direction tolerance. Choosing a wider tolerance span is not in accordance with IEC 61400-11, but is done to increase the number of data points in the sector of interest. Thus, the first filtering criterion in Table 5.3.1 is applied to the measurement data in general. In some cases, a narrower tolerance may be applied, when the number of data points allows (see Section 5.6.1).

Wind direction scenario Filtering criterion IEC 61400-11, tolerance span increased 281°  ✓nacelle 341°

Table 5.3.1: Wind direction scenarios and the corresponding filtering criterion.

5.3.2 Wind speed

By calculating the range of integer wind speeds corresponding to a measured electric power, Pm, of 90 1710 MW, a basic wind speed filtering criterion

(35)

according to Table 5.3.2 is utilized. This to remove redundant data points, which simplifies data analysis in MATLAB.

Basic wind speed span Filtering criterion Wind speed corresponding to a Pm of 90 1710 MW 4 m/s  VH 12 m/s

Table 5.3.2: Basic wind speed span and the corresponding filtering criterion. Wind speed at 10 meters height.

5.3.3 Wind speed binning

In the cases of the turbine actively producing electric power, to further remove redundant data and sort out wanted data of diﬀerent kinds, wind speed bin filtering criteria is utilized. This results in wind speed bins at integers from 6 to 10 m/s. Each bin is set to be 1 m/s wide. This to comply with IEC 61400-11. Table 5.3.3 displays the filter criteria.

Wind speed bin Filtering criterion k = 6 5.5 m/s VH< 6.5 m/s

k = 7 6.5 m/s_{ V}H< 7.5 m/s

k = 8 7.5 m/s VH< 8.5 m/s

k = 9 8.5 m/s_{ V}H< 9.5 m/s

k = 10 9.5 m/s_{ V}H< 10.5 m/s

Table 5.3.3: Wind speed bins and the corresponding filtering criteria. Wind speeds at 10 meters height.

5.4 IceMonitor data

The ice accretion data provided by the IceMonitor indicates large negative values at various times during the long term measurement. These are interpreted to have an uncertain meaning, and are removed from the analysis. The remaining data points are treated as either an indication of ice or an indication of no ice on the wind turbine, depending on IceMonitor indication (see Table 5.4.1). An ice indication filtering criterion is utilized, according to Table 5.4.2, to remove uninterpretable data.

IceMonitor indication (g) Interpretation Ice accretion. 0 Ice accumulation uncertain Ice accretion ⇡ 0 No ice accumulation on wind turbine Ice accretion& 0 Ice accumulation on wind turbine

Table 5.4.1: Interpretation of IceMonitor data relative to ice accretion on wind turbine.

(36)

Notation Filtering criterion mice mice 10 g

Table 5.4.2: Basic filtering criterion applied to IceMonitor indication data.

5.5 Wind sensitivity ratio

As previously mentioned, ice accretion on wind turbine rotor blades may de-crease its performance, resulting in power production losses of various mag-nitudes. This problem is commonly referred to as a degradation of the wind turbine power curve. A degraded power curve means that the measured elec-tric power production is less than expected taking into account wind speed and other atmospheric conditions at the time of measurement. Hence, the quotient between measured electric power, Pm, and power production under reference

conditions, PD, may be a good indicator of degraded power production, as

Pm

PD (5.5.1)

Implicitly, the quantities in Equation 5.5.1 correspond to certain atmospheric conditions, such as wind speed, air temperature and atmospheric pressure at the time of the measurement and under reference conditions (See Table 3.3.1).

By using the standard power curve for a specific wind turbine together with the measured electric power production, one may obtain the corresponding wind speed under reference conditions. If taking into consideration the air tempera-ture and atmospheric pressure at the time of the measurement, and by utilizing the wind speed correction introduced in Section 3.3.2.3, one may compare the performance of the wind turbine from an atmospheric conditions point of view.

To remind ourselves, the wind speed correction equation is VH= VD ✓_p refTk pTref ◆1 3 (5.5.2) where

• VH is the corrected wind speed at hub height (m/s)

• VD is the wind speed at hub height derived from the power curve (m/s)

• pref is the atmospheric pressure under reference conditions (kPa)

• Tref is the air temperature under reference conditions (K)

• p is the measured atmospheric pressure (kPa) • Tk is the measured air temperature (K)

(37)

Now, put simply, the wind sensitivity ratio of the turbine may be studied. For the purposes of this project report, it is defined as

VH

Vn (5.5.3)

where Vn is the measured wind speed at hub height, VH is the wind speed

at hub height derived from the standard power curve (corresponding to Pm)

standardized to the atmospheric conditions at the time of the measurement. The workflow may be illustrated as

Vn() Pm=) Standard power curve =) VD=) Equation 5.5.2 =) VH

Table 5.5.1 shows the possible interpretations of the wind sensitivity ratio.

VH

Vn < 1

• Wind turbine producing below normal rate • Reason uncertain

VH

Vn = 1

• Wind turbine producing at normal rate • Reason uncertain

VH

Vn > 1

• Wind turbine producing above normal rate • Reason uncertain

Table 5.5.1: Simplified interpretations of the wind sensitivity ratio.

5.5.1 Correlation between wind sensitivity ratio and

Ice-Monitor indication

By determining the correlation, or lack of correlation, between the wind sen-sitivity ratio and the IceMonitor indication, one may be able to predict the possibility of power production losses due to ice accretion on the wind turbine rotor blades. For the purpose of this project, this correlation may also lead to a correlation between IceMonitor indication and measured sound pressure level (SPL). The filter criteria in Table 5.5.2 is used to acquire data for ice-free conditions. Notation Criterion t -LAeq 0 dB < LAeq< 65 dB Pm 90 kW < Pm< 1710 kW VH 5.5 m/s < VH 10.5 m/s ✓nacelle -nrotor -mice 5 g < mice< 5 g

Table 5.5.2: Filter criterion to display distribution of wind sensitivity ratio relative to IceMonitor indication under ice-free conditions.

(38)

Under normal (ice-free) conditions the wind sensitivity ratio is theoretically equal to 1. If measurement data are averaged of a long period of time this may prove to be true. Figure 5.5.1 shows an example of this as function of time, during a longer time period. Figure 5.5.2 shows the wind sensitivity ratio as function of IceMonitor indication (ice accretion). The wind sensitivity ratio varies in a stochastic manner, and a deviation of ±0.2 around mean value is not unusual even for ice-free conditions.

Figure 5.5.1: Example of wind sensitivity ratio as function of time under ice-free conditions.

(39)

Figure 5.5.2: Example of wind sensitivity ratio as function of IceMonitor indi-cation under ice-free conditions.

The statistics presented in Table 5.5.3 corresponds to the variations of the wind sensitivity ratio. The histogram (Figure 5.5.3) shows that it is normally distributed around the mean value 1.091, rather than around exact unity. This can be for several reasons, the most likely perhaps being that the wind turbine normally produces slightly above its rated electric power production. Since it is configured to a rated power of 1.8 MW, but has a potential of producing 2.0 MW, this may very well be the reason.

Statistics Wind sensitivity ratio Mean value 1.0910 Min. value 0.7830 Max. value 1.5859 Standard deviation 0.0790

Variance 0.0062

Table 5.5.3: Statistics of the wind sensitivity ratio under ice-free conditions. This shows that it is normally distributed around the mean value of 1.09, rather than around unity.

(40)

Figure 5.5.3: A histogram of the wind sensitivity ratio under ice-free condition. This shows that it is normally distributed around the mean value of 1.09, rather than around unity.

Consequently, a strong correlation between wind sensitivity ratio and Ice-Monitor data in the range of 0.8 to 1.2 cannot be made. An estimation of ice accretion on wind turbine rotor blades for wind sensitivity ratios may prove to be diﬃcult. The simplified interpretations of the wind sensitivity ratio in Table 5.5.1 have to be revised. Since ice accretion on rotor blades are consid-ered to have only negative eﬀects on the power production, a wind sensitivity upper limit of 0.8 is adequate for removing most of the stochastic variations of measurement data. Table 5.5.4 shows the revised interpretations of the wind sensitivity ratio.

VH

Vn  0.8

• Wind turbine producing under normal rate • Ice accretion on wind turbine likely

VH

Vn > 0.8

• Wind turbine producing at normal rate • Ice accretion on wind turbine not likely Table 5.5.4: Revised interpretations of the wind sensitivity ratio.

5.6 Sound pressure level data

The analysis of the equivalent A-weighted SPL data may be separated into three categories, depending on the type of investigation:

(41)

• Turbine operating under ice-free (normal) conditions • Turbine operating under icing conditions

• Turbine not operating - background noise

A basic filtering criterion applied to SPL data are only to remove technical errors and possible external disturbances. Table 5.6.1 shows the criterion.

Notation Criterion LAeq 0 dB < LAeq< 65 dB

Table 5.6.1: Basic filtering criteria applied to SPL data.

5.6.1 SPL under ice-free conditions (long term

measure-ment)

To obtain noise level data under ice-free conditions (as a reference), the long term measurement data, acquired using the Sigicom-Infra system is filtered ac-cording to Table 5.6.2.

The filtering criteria applied to the SPL data, the measured electric power data, wind speed and nacelle direction are the basic filter functions previously introduced. The ice accretion data criterion is set to around zero, while wind sensitivity ratio criterion is disabled. Filtering criteria are set “generously” to not loose too much data, since amount of data in the specified nacelle direction is not large. Notation Criterion t -LAeq 0 dB < LAeq< 65 dB Pm 90 kW < Pm< 1710 kW VH 5.5 m/s < VH 10.5 m/s ✓nacelle 281°  ✓nacelle 341° nrotor -mice 10 g mice 10 g VH Vn

-Table 5.6.2: Filtering criteria for determination of noise emission under ice-free conditions, using long term measurement data.

Figure 5.6.1 and Table 5.6.3 show the SPL levels under ice-free conditions. A complete set of graphics is available in Appendix A.

(42)

Figure 5.6.1: Bin analysis of SPL for ice-free conditions using the long term measurement data.

Statistical values ₆ ₇ Bin (m/s)₈ ₉ _10* Second order regression (dB) 47.4 47.3 47.3 47.4 47.48 Linear regression (dB) 47.4 47.2 47.1 47.4 -Mean value (dB) 47.5 47.4 47.1 47.4 -Min. (dB) 47.1 46.5 46.5 46.8 -Max. (dB) 48.2 48.5 47.9 48 -Standard deviation (dB) 0.37 0.49 0.38 0.60 -Variance (dB) 0.14 0.24 0.14 0.36 -Number of data points 8 16 17 3 0

Table 5.6.3: Bin analysis of ice free conditions using the long term measurement data. *No data points available after filtering.

5.6.2 SPL under icing conditions

To obtain noise level data under icing conditions, the long term measurement data, acquired using the Sigicom-Infra system is filtered according to Table 5.6.4.

The filtering criteria applied to the SPL data, the measured electric power data, wind speed and nacelle direction are the basic filter functions previously

(43)

introduced. The ice accretion data criterion is set to above zero, while wind sensitivity ratio criterion is set to a maximum of 0.8. The amount of data in the specified nacelle direction still is not large.

Notation Criterion t -LAeq 0 dB < LAeq< 65 dB Pm 90 kW < Pm< 1710 kW VH 5.5 m/s < VH 10.5 m/s ✓nacelle 281°  ✓nacelle 341° nrotor -mice 10 g mice VH Vn VH Vn  0.8

Table 5.6.4: Filtering criteria for determination of noise emission under icing conditions, using long term measurement data.

Figures 5.6.2 and Table 5.6.5 show the resulting SPL levels under icing con-ditions (a complete set of graphics is available in Appendix B). As is discussed in Section 5.5.1, the correlation between wind sensitivity and IceMonitor indica-tion is weak in the range of 0.8 and 1.2 due to stochastic variaindica-tions (statistical noise) around unity. Since the filtering criterion in Table 5.6.4 regarding the wind sensitivity ratio is set to less and/or equal to 0.8, most of the noise is removed. However, the criterion is set “generously” with the intention not to influence the results more than necessary, and to have enough data points. In the 40 to 50 dB range, Figure 5.6.2 displays some of the noise which manages to leak through the filter. These data points influence the statistics displayed in Table 5.6.5, generally lowering the SPL’s calculated through regressions and mean values. Thus, making it interesting to take the maximum values into consideration.

(44)

Figure 5.6.2: Bin analysis of SPL for icing conditions using the long term mea-surement data

Statistical values ₆ ₇ Bin (m/s)₈ ₉ ₁₀ Second order regression (dB) 56.8 57.8 57.9 57.1 55.5 Linear regression (dB) 56.8 57.8 57.8 57.1 33.8 Mean value (dB) 46.8 57.8 57.8 57.2 52.0 Min. (dB) 46.5 55.9 54.0 52.3 48.8 Max. (dB) 58.4 59.0 58.9 59.5 58.3 Standard deviation (dB) 2.53 0.70 0.88 2.06 4.23 Variance (dB) 6.38 0.50 0.77 4.23 30.1 Number of data points 82 129 136 35 3

Table 5.6.5: Bin analysis of icing conditions using the long term measurement data.

5.6.3 Background noise

To obtain background noise level data, the long term measurement data, ac-quired using the Sigicom-Infra system is filtered according to Table 5.6.6.

The filtering criteria applied to the SPL data, the measured electric power data, wind speed and nacelle direction are the basic filter functions previously introduced. The ice accretion data criterion is disabled, as well as the wind

(45)

sensitivity ratio criterion. The amount of data in the specified nacelle direction is not large. Notation Criterion t -LAeq 0 dB < LAeq< 65 dB Pm Pm 0 kW VH 5.5 m/s < VH 10.5 m/s ✓nacelle* 10°  ✓nacelle 40° nrotor nrotor 0 mice -VH Vn

-Table 5.6.6: Filtering criteria for determination of background noise, using long term measurement data. *Nacelle direction and span closest to 311°, with avail-able data after filtering.

Figure 5.6.3 and Table 5.6.7 show the resulting SPL levels under ice-free conditions. A complete set of graphics is available in Appendix C.

Figure 5.6.3: Bin analysis of background noise using long term measurement data.

(46)

Statistical values ₆ ₇ Bin (m/s)₈ ₉ ₁₀ Second order regression (dB) 30.1 30.6 31.0 31.4 31.7 Linear regression (dB) 30.1 30.5 31.1 31.4 31.8 Mean value (dB) 30.1 30.5 31.1 31.4 31.8 Min. (dB) 28.1 28.2 29.1 30.5 30.9 Max. (dB) 33.7 33.5 33.0 32.2 33.7 Standard deviation (dB) 1.10 1.00 0.67 0.46 0.77 Variance (dB) 1.20 1.00 0.45 0.22 0.59 Number of data points 154 101 90 75 12

Table 5.6.7: Bin analysis of background noise using the long term measurement data.

5.6.4 SPL under ice-free conditions (short term

measure-ment)

The short term measurements are analyzed for verification of long term mea-surements, and to determine the sound power data for the measurement object for ice-free conditions using the method outlined by standard IEC 61400-11. The result would be seen as a reference state of the measurement object. The short term measurement data are filtered according to Table 5.6.8.

The filtering criteria applied to the SPL data, the measured electric power data, wind speed and nacelle direction are the basic filter functions previously introduced. The ice accretion data criterion is disabled, as well as the rpm criterion and the wind sensitivity ratio criterion. The amount of data in the nacelle direction of 311° ± 30°, is more than expected, hence the tolerance span is narrowed to ±20°, which is almost in accordance with IEC 61400-11.

For clarification, the analysis does not comply with the IEC standard re-garding:

• The tolerance span of the nacelle direction, ✓nacelle, is set to ±20° (see

Section 5.3.1)

• Number of points in 10 m/s bin is too low. A minimum of two data points is required.

Figure 5.6.4 and Table 5.6.9 show the background noise levels. A complete set of graphics is available in Appendix D.

(47)

Notation Criterion t -LAeq 0 dB < LAeq< 65 dB Pm 90 kW Pm 1710 kW VH 5.5 m/s < VH 10.5 m/s ✓nacelle 291°  ✓nacelle 331° nrotor -mice* -VH Vn

-Table 5.6.8: Filtering criteria for determination of measured equivalent A-weighted SPL, using short term measurement data. *Filtering criterion disabled because of manual observation and verification of ice-free conditions.

Figure 5.6.4: Bin analysis of measured equivalent A-weighted SPL, using short term measurement data.

(48)

Statistical values ₆ ₇ Bin (m/s)₈ ₉ _10* Second order regression (dB) 50.1 50.3 50.4 50.4 50.3 Linear regression (dB) 50.1 50.3 50.3 50.4 49.5 Mean value (dB) 50.1 50.3 50.3 50.5 49.1 Min. (dB) 48.2 47.8 48.2 49.2 49.1 Max. (dB) 51.7 52.3 52.1 51.1 49.1 Standard deviation (dB) 0.85 0.90 1.00 0.60 0 Variance (dB) 0.72 0.82 1.00 0.36 0 Number of data points 54 79 40 11 1

Table 5.6.9: Bin analysis of measured equivalent A-weighted SPL, using short term measurement data. *Only one point in bin, not in accordance with IEC 61400-11.

5.6.5 Microphone SPL diﬀerence

For a shorter period of time, parallel measurements are performed using the long term measurement setup. By mounting the two microphones next to each other on the stand, the diﬀerence in SPL between the microphones may be determined (see Figure 5.6.5). The results are used for subsequent corrections. The diﬀerence is calculated as time average (mean value) of

LAeq,mic= LAeq,Sigicom LAeq,IEC

for parallel time points. The result is an SPL diﬀerence of 0.36 dB. This is interpreted as an error, or deviation, which most likely arises from the calibration and/or of the measurement equipment (microphones, sound level meter, cords etc).

(49)

Figure 5.6.5: Time period for analysis of diﬀerence in measured equivalent A-weighted SPL for microphones.

5.6.6 Measurement setup SPL diﬀerence

According to IEC 61400-4, wind turbine noise emission measurements shall be performed using a measurement board. The board is estimated to influence measured SPL by +6 dB. This error is corrected for when the apparent sound power level is calculated (second term on right hand side in Equation 3.3.6). Since the long term measurement setup consists of a stand mounted microphone, it is of interest to evaluate and/or verify the diﬀerence in measured SPL between the two measurement setups. With the analysis performed in Section 5.6.4 as a reference, the same time segment for the long term measurement is extracted and analyzed. The data are filtered in the same manner (Table 5.6.8. Figure 5.6.6 and Table 5.6.10 show the resulting SPL during the time segment.

(50)

Figure 5.6.6: Bin analysis of long term measurement SPL data for evaluation of SPL diﬀerences between measurement setups.

Statistical values ₆ ₇ Bin (m/s)₈ ₉ _10* Second order regression (dB) 46.4 46.7 46.9 47.2 47.6 Linear regression (dB) 46.4 46.6 46.9 47.2 47.4 Mean value (dB) 46.4 46.5 46.9 47.1 47 Min. (dB) 44.6 44.2 45.1 46.3 47 Max. (dB) 47.7 47.9 48.1 48.7 47 Standard deviation (dB) 0.73 0.85 0.78 0.62 0 Variance (dB) 0.53 0.71 0.61 0.39 0 Number of data points 54 79 40 11 1

Table 5.6.10: Bin analysis of measured equivalent A-weighted SPL using the long term measurement data during the same time period as in Section 5.6.4.

The diﬀerence in measured SPL due to diﬀerence between measurement setups, is calculated as

LAeq,k,setup= LAeq,k,Sigicom LAeq,k,IEC

(51)

Bin (m/s)

6 7 8 9 10

LAeq,k,IEC(dB) 50.1 50.3 50.4 50.4 50.3

LAeq,k,Sigicom(dB) 46.4 46.7 46.9 47.2 47.6

LAeq,k,setup (dB) 3.7 3.6 3.5 3.2 2.7

Table 5.6.11: Diﬀerence in measured SPL due to diﬀerence in measurement setups.

5.6.7 Apparent sound power levels

To obtain the apparent sound power levels at each wind speed bin, the back-ground noise level must be determined. This, as it would be measured by the short term measurement setup (in accordance with IEC 61400-11). Hence, the following is needed

• Background noise levels for long term measurement (from Section 5.6.3) • Correction factor for microphone calibration/sensitivity diﬀerence (from

Section 5.6.5)

• Correction factor for diﬀerence between measurement setups (from Section 5.6.6)

Simple addition and subtraction is performed according to Table 5.6.12 with notation according to Equation 3.3.5.

Bin (m/s)

6 7 8 9 10

Background noise levels (dB) 30.1 30.6 31.0 31.4 31.7 Microphone correction (dB) +0.36

Measurement setup correction (dB) +3.7 +3.6 +3.5 +3.2 +2.7 Background noise level, Ln,k (dB) 34.2 34.6 34.9 35.0 34.8

Table 5.6.12: Analysis of background noise levels. The results are calculated from short term measurement data.

The measured equivalent A-weighted SPL is corrected for the influence of background noise according to Equation 3.3.5, since diﬀerence is greater than 6dB. Further, the SPL is corrected for the influence of the wind screens (primary and secondary), see Section 4.1.5. By using Equation 3.3.6 the sound power levels for each (wind speed) bin may be calculated. Table 5.6.13 displays the results.

(52)

Bin (m/s)

6 7 8 9 10

Combined level, Ls+n,k(dB) 50.1 50.3 50.4 50.4 50.3

Wind screens correction +0.8

Background noise level, Ln,k (dB)* -34.2 -34.6 -34.9 -35.0 -34.8

Background corrected SPL, LA,c,k(dB) 50.8 51.0 51.1 51.1 51.0

A-weighted sound power level, LWA,k(dB) 100.0 100.2 100.3 100.3 100.2

Table 5.6.13: The apparent sound power level of the measurement object. *Log-arithmic subtraction.

The effect on noise emission from wind turbines due to ice accretion on rotor blades

The eﬀect on noise emission from wind turbines

due to ice accretion on rotor blades

TRITA-AVE-2012:62

Peter Arbinge

Stockholm, July 6, 2012

Contents

Chapter 1

Introduction

1.1 Background

1.2 Purpose

1.3 Limitations

Chapter 2

Method

2.1 Problem formulation

2.2 Literature study

2.3 Field study

2.4 Analysis

Chapter 3

Literature study

3.1 Wind turbine noise sources

3.1.1 Mechanical noise sources

3.1.2 Aeroacoustical noise sources

3.2 Ice accretion on wind turbines

3.2.1 Ice accretion mechanisms

3.3 Acoustical measurements of wind turbines

3.3.1 Instrumentation

3.3.2 Measurements and measurement procedures

3.3.3 Data reduction procedures

Chapter 4

Field study

4.1 Wind turbine noise measurements

4.1.1 Measurement object

4.1.2 Measurement site

4.1.3 Measurement position

4.1.4 Long term wind turbine noise measurement

4.1.5 Short term wind turbine noise measurement

4.1.6 Background noise measurement

4.2 Complementary data acquisition

4.2.1 Wind turbine data

4.2.2 Ice data

4.2.3 Ice prognoses data

4.2.4 Weather data

Chapter 5

Analysis

5.1 Time synchronization

5.1.1 Conversion of averaging times

5.1.2 Synchronization and trimming of data

5.2 Measured electric power

5.2.1 Power curve

5.3 Wind

5.3.1 Wind direction

5.3.2 Wind speed

5.3.3 Wind speed binning

5.4 IceMonitor data

5.5 Wind sensitivity ratio

5.5.1 Correlation between wind sensitivity ratio and

Ice-Monitor indication

5.6 Sound pressure level data

5.6.1 SPL under ice-free conditions (long term

measure-ment)

5.6.2 SPL under icing conditions

5.6.3 Background noise

5.6.4 SPL under ice-free conditions (short term

measure-ment)

5.6.5 Microphone SPL diﬀerence

5.6.6 Measurement setup SPL diﬀerence

5.6.7 Apparent sound power levels