s chool of g eo s ciences offshore wind mapping using synthetic aperture radar and meteorological...
TRANSCRIPT
SCHOOL of GEOSCIENCES
Offshore wind mapping using synthetic aperture radar and meteorological model data
Iain CameronDavid MillerNick WalkerIain Woodhouse
SCHOOL of GEOSCIENCES
Offshore Wind
• UK has largest offshore wind resource in EU
• 10 km off shore ~ 25% more energy than on land
• BUT more expensive than land based technology
• Accurate understanding of wind resource vital
Blade Diameter90m
TowerHeight 80m
2 MW Vestas
SCHOOL of GEOSCIENCES
Overview
Retrieved Wind
Synthetic Aperture Radar
Data
UKMO Unified Mesoscale Model (UMM)
SAR Wind Inversion
Analysis cells
Hilbre Island
SCHOOL of GEOSCIENCES
Retrieved Wind
SAR Scenes
UMM Data
Inversion
Envisat ASAR
• ASAR (advanced SAR• C-band (5.6 cm λ, 5.3 GHz)• Multiple modes of operation• Image mode
– High res (12.5 m2)
– Low repeat time (~25 days)
• Wide swath– Medium resolution (75 m2)
– High repeat time (~3-5 days)
SCHOOL of GEOSCIENCES
Apriori Data
Retrieved Wind
SAR Scenes
UMM Data
Inversion
• UKMO Unified Mesoscale Model (UMM)
• 6 hourly analysis levels– Interpolated to SAR time– Interpolated to ~2.5 km Grid
SCHOOL of GEOSCIENCES
Forward Model
Retrieved Wind
SAR Scenes
UMM Data
Inversion
NRCS (σ0)Wind direction +Wind speed
GMF
=
CMOD5
SCHOOL of GEOSCIENCES
Model Inversion
Retrieved Wind
SAR Scenes
UMM Data
Inversion
a) Image Directions• Roll vortices/streaks• Fourier, wavelet, Sobel
filters, cross spectra analysis
• Not visible in all scenes (~60% of cases)
b) NWP winds• Always available• Poor resolution
– Spatial (0.125 deg)– Temporal (every 6 hrs)
SCHOOL of GEOSCIENCES
Model Inversion
Retrieved Wind
SAR Scenes
UMM Data
Inversion
Retrieves wind speed assuming NWP wind direction is true
Problems
•Assumes SAR variation only due to wind speed changes
•Doesn’t account for known retrieval errors
1) “Directional Wind Speed Algorithm” (DWSA)
SCHOOL of GEOSCIENCES
Model Inversion
Retrieved Wind
SAR Scenes
UMM Data
Inversion
•Estimates optimal wind vector given the σ0 and apriori wind vector
1 1
ˆ=minTT
a a a e
0 0x x x S x x y f x S y f x
Observation Term
AprioriTerm
2) Maximum Aposteriori Probability (MAP)
•Apply Gauss-Newton minimisation•Stabilises within 3-5 iterations
SCHOOL of GEOSCIENCES
Sensitivity Analysis
• Generate σ0 using wind speeds 5-25 ms-1 and directions 0-180o
• Add 5% Gaussian noise to σ0
• Retrieve speed
SCHOOL of GEOSCIENCES
Validation Results
R2= 0.715RMSE = 1.57 m/s
R2= 0.576RMSE = 1.7 m/s
UKMO UMM
R2= 0.609RMSE = 2.28 m/s
SCHOOL of GEOSCIENCES
Mean Wind Speeds
UKMO UMM MAP CMOD5DWSA CMOD5
0 10Speed m/s
SCHOOL of GEOSCIENCES
Conclusions & Future Directions
• The MAP methodology shows promise for SAR wind field retrieval
• BUT there are limitations in the resolution of the weather model data
• Future work will:– Introduce SAR wind direction analysis
– Consider the applicability of these data products for wind farm planning
SCHOOL of GEOSCIENCES
Offshore wind mapping using synthetic aperture radar and meteorological model data
Iain CameronDavid MillerIain Woodhouse
SCHOOL of GEOSCIENCES
Sensitivity Analysis
Gaussian Noise on apriori
Why the high speed, Downwind bias?
SCHOOL of GEOSCIENCES
Methodology
Hierarchical Inversion Method For Improving Retrieval Resolution
SCHOOL of GEOSCIENCES
Sensitivity Analysis
CMOD5 shows increased
saturation effects at
high speeds
Wind direction relative to antenna
σ0