www.hoarelea.comoctober 2013 wind turbine amplitude modulation: research to improve understanding as...
TRANSCRIPT
w w w . h o a r e l e a . c o mOctober 2013
Wind Turbine Amplitude Modulation: Research to Improve Understanding as to its Cause & Effect
Project overview and conclusions
w w w . h o a r e l e a . c o mOctober 2013
Project elements
G Blade surface measurements RISO
WP Description LEAD
A1 Source generation effects modeling NLR
A2 Fundamental Research into Possible Causes of Amplitude Modulation ISVR
B1 Development of an Objective AM Measurement Methodology ISVR
B2 Development of an AM Dose-Response Relationship ARC
C Collation and Analysis of Existing Acoustic Recordings HLA
D Measurement and Analysis of New Acoustic Recordings HLA
E Wider Dissemination of Results ALL
F Collation of Work Package Reports and Final Reporting HLA
Sabine von Hunnerbein – ARC University of SalfordHelge Madsen – DTU (formerly RISOE) Denmark
w w w . h o a r e l e a . c o mOctober 2013
Source modelling - Blade Swish
Stefan Oerlemans NLR
Turbulent boundary layer
Trailing edge Wake
Airfoil
w w w . h o a r e l e a . c o mOctober 2013
Source modelling - Blade Swish
w w w . h o a r e l e a . c o mOctober 2013
Source modelling - Blade Swish
w w w . h o a r e l e a . c o mOctober 2013
Source modelling - Blade Swish
Stefan Oerlemans NLR
5 d
B(A
)
crosswindupwind/downwind
Amplitude Modulated (AM) aerodynamic noise
20 seconds
w w w . h o a r e l e a . c o mOctober 2013
7
Blade Swish CharacteristicsCaused by the directivity and convective amplification of the dominant trailing edge noise mechanism:
– mid-high frequency noise (400-1000 Hz)
– peaks when the blade is moving towards the observer
– close to turbine dominant underdownwards sweep of the blade
– at large distances (>3RD) only occurs incross-wind directions
– maximum predicted variation~5 dB peak-to-trough
w w w . h o a r e l e a . c o mOctober 2013
Literature review (historical)UK
• ETSU-R-97
• ETSU WTN source study
• DTI/HMP LFN study
• Salford AM study
• Bowdler review
non-UK
• van den Berg (NL)
• Di Napoli (Finland)
• Legarth
• Lundmark (Sweden)
• Toora Wind Farm (Australia)
• Waubra Wind Farm (Australia)
• West Wind (NZ)
• Lee et al (South Korea)
• internet sources (USA, Canada, etc.)
• AM was not just a UK issue
• reports indicated something other than ‘normal’ blade swish
w w w . h o a r e l e a . c o mOctober 2013
A further definition of AM …..
• commonly termed ‘blade swish’
• part of normal WTN
• ~5dB modulation at source
• dominant crosswind effect
• decreases away from source
• dominated by mid frequencies (400Hz to 1000Hz) ‘swish’
• source mechanism understood
‘Normal’ AM ‘Other’ AM
• atypical, intermittent
• >5dB (>10dB) amplitude at times?
• audible/noticeable at large distances downwind to >1km?
• more impulsive ‘thump’
• additional lower frequency content (200 Hz to 500 Hz)? ‘whooomp’
• source mechanism?
Confusion
w w w . h o a r e l e a . c o mOctober 2013
What is OAM?
• for the purpose of the present project OAM is objectively defined as any AM whose characteristics can not be described by the known source generation mechanisms of NAM (blade swish)
• potential issue that this definition precludes propagation (as opposed to source) effects
w w w . h o a r e l e a . c o mOctober 2013
Possible origin of OAM (Oerlemans)
Turbulent boundary layer
Trailing edge Wake
Airfoil
Airfoil
Boundary layerseparation
Airfoil
Large-scale separation(deep stall)
w w w . h o a r e l e a . c o mOctober 2013
Possible origin of OAM (Oerlemans)
fixed pitch and rpm, different inflow wind speeds
w w w . h o a r e l e a . c o mOctober 2013
• variable inflow conditions across the rotor could lead tolocalised stall on some portions of the blade,
• medium/high wind shear conditions can lead to such inflow conditions, as can other factors …
• occurrence will also be dependent on blade design and control logic
yaw wake topography inflow turbulence
a=0.3
Possible origin of OAM (Oerlemans)
a=0.6a=0.3
w w w . h o a r e l e a . c o mOctober 2013
New measurements
SITE A: measurements at residential properties at which OAM had been reported
SITE B: detailed measurements around turbines operating ‘normally’ at a site where OAM had been reported
SITE C: detailed measurements around a turbine at a site with access granted to modify the control system settings (including blade pitch to induce stall)
w w w . h o a r e l e a . c o mOctober 2013
Sample new measurements
• directivity of OAM in near-field and at 10 Rotor Diameters (~900m)
• practical observations match refined Oerleman’s model
Cross-wind Down-wind
near field
far field
w w w . h o a r e l e a . c o mOctober 2013
• SPL measured in far field of wind turbine at ~ 1000 m
• OAM only identifiable for one (differently pitched) blade
10
dB
(A)
20 seconds
Additional measurements
w w w . h o a r e l e a . c o mOctober 2013
Detecting and measuring AM
Any successful metric must:
• be objective
• be repeatable
• be robust (avoid false positives and false negatives)
• work on real and not just simulated AM noise
• ideally allow automated application
• be relatable to subjective response
w w w . h o a r e l e a . c o mOctober 2013
Examples of existing AM methodsA range of methods have been proposed for measuring amplitude modulated wind farm noise:
1. direct time domain analysis of instantaneous Leq levels
2. frequency domain analysis of the Leq levels
3. Fourier transform of the spectrogram
Methods 2 and 3 are variations of existing techniques used in sonar and condition monitoring
Methods 2 and 3 rely on the periodicity of the noise data at the BPF as a fundamental part of the detection process
w w w . h o a r e l e a . c o mOctober 2013
Objective quantification of AM
A metric has been identified to quantify the level of AM (NAM or OAM) at BPF present in a sample of acoustic data
Sample analysis over a 3 hour period
w w w . h o a r e l e a . c o mOctober 2013
Riso studies – DAN-AERO MW
w w w . h o a r e l e a . c o mOctober 2013
Proof that OAM is a source effect?
w w w . h o a r e l e a . c o mOctober 2013
Proof that OAM is a source effect?
w w w . h o a r e l e a . c o mOctober 2013
Proof that OAM is a source effect?
w w w . h o a r e l e a . c o mOctober 2013
Proof that OAM is a source effect?
w w w . h o a r e l e a . c o mOctober 2013
Surface pressure on suction side
w w w . h o a r e l e a . c o mOctober 2013
Surface pressure on suction side
w w w . h o a r e l e a . c o mOctober 2013
Listening tests on AM
• Reports in the published literature …..• WTN more annoying than other environmental noise
• Speculation in literature• Sound characteristics to blame?
• Response by industry and government• Need for dose-response relation
w w w . h o a r e l e a . c o mOctober 2013
Stimuli overview
Parameter
Modulation depth, dB(A) 0, 2, 3, 4, 5, 6, 9, 12
Sound level of total stimulus LAeq, dB(A)
25, 30, 35, 40 and 45
w w w . h o a r e l e a . c o mOctober 2013
Setup: user interface
w w w . h o a r e l e a . c o mOctober 2013
‘Overall’ absolute annoyance ratings
25 30 35 40 45 500
2
4
6
8
10
AM LAeq, dB(A)
Ann
oyan
ce R
atin
g
0 dB(A)
2 dB(A)3 dB(A)
4 dB(A)
5 dB(A)
6 dB(A)9 dB(A)
12 dB(A)
w w w . h o a r e l e a . c o mOctober 2013
‘AM’ absolute annoyance ratings
0 5 10
0
2
4
6
8
Modulation Depth, dB
An
no
yan
ce R
atin
g
4540353025
w w w . h o a r e l e a . c o mOctober 2013
Adaptive rating – not normalised
0 5 1025
30
35
40
45
50
Modulation Depth, dB
Re
fere
nce
leve
l, d
B(A
)
4540353025
w w w . h o a r e l e a . c o mOctober 2013
Normalising adaptive rating levels
0 5 1025
30
35
40
45
50
Modulation Depth, dB
Re
fere
nce
leve
l, d
B(A
)
4540353025
41.8 – 40 = 1.8reference – test = normalised
w w w . h o a r e l e a . c o mOctober 2013
Adaptive rating – normalised (1)
0 5 10-4
-2
0
2
4
6
8
Modulation Depth, dB
Re
fere
nce
leve
l - T
est
LA
eq, d
B(A
)
4540353025
w w w . h o a r e l e a . c o mOctober 2013
Adaptive rating – normalised (2)
w w w . h o a r e l e a . c o mOctober 2013
Subjective response findings• Consistent ratings between participants
• Average annoyance from AM signals higher than from non-modulated noise, but ….
• Strongest effect of overall level on annoyance
• No clear effects with increasing• modulation depth above a certain threshold• type of modulation (see report)• addition of garden noise (see report)• use of different metrics (see report)
w w w . h o a r e l e a . c o mOctober 2013
Causes of AM
• The principal causal source mechanism of OAM identified as partial, transient blade stall
• No evidence to suggest that OAM should occur in the far field as a result of propagation effects only in the absence of a source mechanism
• Potential speculative causal factors which have been suggested to date shown to have little or no systematic association to the occurrence of OAM: for example interaction between closely spaced turbines in linear arrays, large rotor to tower height ratios, etc..
• But some of these may represent potential contributory factors.
w w w . h o a r e l e a . c o mOctober 2013
Can OAM be mitigated?• Partial stall as a source of OAM could be efficiently mitigated by
avoiding blade stall.
• This could be achieved through a number of potential solutions to be developed and tested, including:
a. Software fixes, e.g. modifying the logic of the turbine control system
b. Physical changes including innovative blade designs or cyclical pitch control
• Mitigation is often likely to only be required in down-wind conditions, but ….
• Need to be aware of possible overlap between OAM and NAM
w w w . h o a r e l e a . c o mOctober 2013
Conclusions• AM can take at least two forms which appear to have fundamentally
different source generation mechanisms
• the different source mechanisms result in different radiation characteristics between NAM (‘normal’ blade swish) and OAM
• there is no evidence to suggest that OAM should occur in the far field in the absence of its presence at source, but ….
• measurements undertaken close to the source (e.g. at IEC61400-11 compliant microphone locations) may not properly characterise AM
• NAM is an inherent feature of wind turbine noise
• OAM only occurs dependent on a number of interacting factors
• robust objective detection and quantification methods for AM are possible