Brad Isom and Bob Palmer

1
Signal Processing Research for Wind Turbine
Clutter (WTC) Mitigation

• Brad Isom and Bob Palmer
• School of Meteorology
• University of Oklahoma

http//arrc.ou.edu/turbine/
ROC Briefing, August 29, 2006
2
Gray County Wind Farm

• Located 25 miles SW of Dodge City, Kansas
• 170 towers
• Height of Tower 217 feet (66m)
• Length of Blades 77 feet (23.5m)
• RPM 28.5
• Tip Velocity 70.14 ms-1

Photohttp//www.cimarronkansas.net/WINDMILLFARMPI
CS/WindfarmE200TU-Main.jpg
3
Gray County Wind Farm
Photo from.maps.google.com
4
The Doppler SpectrumUnraveling radial velocities

• The Doppler spectrum is a power-weighted
  distribution of radial velocities within the
  resolution volume
• Examples of unique velocity distributions
• Tornados
• Ground Clutter
• Sea Clutter
• Birds
• Wind Turbine Clutter (WTC)

5
Measured Doppler Spectrum From UK

• QinetiQ Radar Study (UK)
• S-band
• PRT0.5 ms
• spotlight mode
• single turbine in resolution volume
• 3.45 km range from turbine
• blade length 30.8 m
• rotation speed 23 rpm

50 m/s
Wind Farm Impact on Radar Aviation Interests,
QinetiQ, September 2003.
6
KDDC Level I Experiment September 20, 2005
1400-1600 UTC
Scanning Mode

• VCP 21 and 32
• PRT 1.2 ms and 3.1 ms
• ra 175 km and 466 km
• va 22.17 ms-1 and 8.33 ms-1
• Range Resolution is 250 m

Spotlight Mode

• One azimuth angle 248o
• 6 files each containing 2 minutes of data

7
Spotlight DataTemporal Evolution of Doppler
Spectra

• Periodogram Algorithm
• Non-overlapping 32-pt windows
• Velocity Resolution 1.95 ms-1
• Zero-pad to 128 points
• va 31.05 ms-1

8
Spotlight DataSpatial Evolution of Doppler
Spectra
9
KDDC Level I Experiment II March 30-31, 2006

• VCP 21
• Approximately 16 hours of scanning
• To be used in final testing of mitigation
  algorithm
• RHI
• 10 elevation angles 0.5-2.9o
• Spotlight
• Oversampled to achieve 25 meter range gates
• Used the highest PRF 1282 Hz (PRT 0.78 ms)
• Selected azimuth angles to isolate turbines
• Some data taken simultaneously with mobile
  SMART-R
• Smart-R
• Spotlight mode on isolated turbine from three
  different angles
• PRF 2550 Hz (PRT 0.39 ms)
• Wavelength 5.4 cm
• Beamwidth of 1.5o

10
KDDC Level I Experiment IISurface Wind Data

• Given the wind direction, it is expected that the
  tip velocity will produce a radial velocity of
  approximately 50 ms-1
• Aliasing will occur

Wind 300o
40o
250o Az
Blade Face
Data taken from the Dodge City Forcast Office
Surface Station
11
KDDC Level I Experiment IISurface Synoptic
Weather Map
Image obtained from http//weather.unisys.com/
12
KDDC Level I Experiment IIWSR-88D Beam Pattern
2.0o 2.5o 3.0o
Image from WSR-88D Antenna Polarization Change,
Dale Sirmans 1993
13
KDDC Level I Experiment IIRHI Power/Range vs
Elevation Angle
14
KDDC Level I Experiment IIRHI Doppler Spectra
vs Elevation Angle
15
KDDC Level I Experiment IISpotlight Data Single
Turbine
16
Non-linear Filtering Methods

• Unwanted WTC signals are transient (0.1 sec).
  Weather echoes are approximately stationary over
  seconds.
• Solution for Spotlight Data Use a simple
  variation of a median filter to remove the
  artifacts
• Sort the input to the filter by power level and
  average the 15-70 range
• Filter order used is 20 spectral samples

17
Filter Design for vr 30 ms-1
Primary Flashes
Aliased Secondary Flashes
18
Filter Design for vr 5 ms-1
Primary Flashes
19
Filter DesignWindow Optimization
20
Filtering ResultsMedian Filter on Single Turbine
21
Addition of Simulated Weather
Simulated time-series weather added Zrnic,
1969 ? 3 ms-1, vr varies with time, SNR
20 dB
22
Filtering ResultsMedian Filter with Simulated
Weather
23
OUGMAPStationary Clutter Target
OUGMAP designed based on Gaussian Model
Adaptive Processing (GMAP) for Improved Ground
Clutter Cancelation and Moment Calculation,
Siggia and Passarelli.
24
Results w/ OUGMAPSingle Wind Turbine
25
Results w/ OUGMAP Median Filter
26
Smart-R Experiment

• Spotlight mode on an isolated turbine
• Data were taken from three different positions
• Wavelength 5.4 cm
• PRF 2550 Hz
• va 34.425 ms-1

Data collected with the help of Mike
Biggerstaff and Tim Wiegman
27
Smart-R Locations
28
Smart-R Data
Position 1
Position 2
29
Simulated Wind Turbine

• Create an EM scattering model for a wind turbine
  to be placed in the field of view of our Next
  Generation Radar Simulator
• time-series, Doppler spectra
• control over number of turbines, orientation,
  rotation rate, etc.

30
Simulation
31
Next Step in Experiments

• Optimize processing scheme for spotlight data
• Qualitatively compare Smart-R data to KDDC
  results
• Refine the simulation for cancellation study
• Explore methods of non-stationary clutter
  mitigation for VCP scans adaptive filtering,
  wavelet processing, Fractional-Fourier Transform
• Future experiment may be needed for more
  difficult case of multi-path scatter, multi-trip
  echoes, etc. KTFX Great Falls, MT?

32
Questions
33
backup slides
34
Simulation - Radial VelocityThree Blades, Three
Positions, 10o Off
35
Simulations Parameters

• ? orientation of blade-face with radar radial
• pitch angle of the blades
• u horizontal wind speed

36
Aliasing