last issues | tethys - microscale variability of wind …...microscale variability of wind power...
TRANSCRIPT
Microscale variability of wind power density assessment in complex terrain and wind
regime trends in future climate projections
Darko Koracin1,2, Radian Belu3, Craig Smith1, John Mejia1, Kristian Horvath4, and Ramesh Vellore5, and Greg
McCurdy1
1Desert Research Institute, Reno, Nevada, USA2University of Split, Croatia
3University of Alaska Anchorage, USA4Meteorological and Hydrological Service, Zagreb, Croatia
5Indian Institute of Tropical Meteorology, Pune, India5th International Conference on Meteorology and Climatology of the Mediterranean,Istanbul, Turkey, 2-4 March 2015
Objectives
• To understand winds, turbulence, and wind power density in complex terrain of western Nevada using wind towers and sodar data
• Evaluate mesoscale-microscale numerical models in single and ensemble modes using towers’ and sodar data
• Provide initial estimates of climate projections of the winds and wind power density for western Nevada
Outline
• Field program setup• Analysis of winds, turbulence intensity, and
wind power density • Skills of the WRF forecasts for increasing
horizontal resolution and approaching microscales
• Climate model results for winds and wind power density in the 21st century
• Conclusions
Field program
• The main components of the project were a field campaign during a period of more than one year and subsequent forecasting studies
• Two 60-m towers and a collocated acoustic sounder were deployed over the mountain ridges in the area of developed topography
• The towers were located 2700 m apart with a vertical distance of 140 m elevation between their bases
• Both towers were equipped with standard anemometers at three levels (20, 40, 60 m) and sonic anemometers at two levels (20, 60 m).
Wind towers in the Reno and Washoe Valley ridge areas. “2 NC WT1 CEC” and“3 NC WT2 CEC” are two towers that were operational were supported project.
Topographical setup of the wind towers
Tower 1 (60 m)
Tower 2 (60 m)
Sodar near Tower 1
Scatter plots of cup (x-axis) vs. sodar (y-axis) wind speed at 60 m (10/6/12 – 11/1/13)at 60 m on tower WT1 for all sodar observations (QC=0; slope=0.82; R2 =0.63; numberof observations=42540) (left panel) and with QC > 95 (slope=0.93; R2 =0.91; numberof observations=18032) (right panel).
Tower vs. sodar (60 m)
Large scatter reduced when the quality control threshold for the sodar is increased
Cup anemometer Tower 1 vs. Tower 2 – 60 m
Period 11/1/12 – 6/1/13
Differences in mean hourly wind speed are less than 1 m/s
60 mSonic vs. cup
Sodar vs. cup
Sodar vs. sonicR2 = 0.96
R2 = 0.91
R2 = 0.90
Correlation:- Overall high CC- Best – sonic vs. cup- Sodar overestimates both cup and sonic
Turbulence intensity (TI) vs. wind speed20 m 40 m 60 m
TI decreases with height, especially for mid-range speeds.The decrease is stronger at higher levels.
Weibull distribution coefficients and available power density
Tower 1
Tower 2
(10/5/12 – 2/24/14)
Large differences in wind power density in spite of the relatively smallbase height and horizontal distance
Weibull distribution coefficients and available power density - sodar
Significant increase in the power density beyond 60 m, higher speeds, and greater k and c coefficients compared to the towers. Much higher power density. (10/30/12 – 11/11/13)
TW 1
TW 2
Autocorrelation coefficient
TW 1 keeps higher autocorrelation –higher wind speeds
Both towers reach first zeroautocorrelationcoefficient at about 60 hrs
Mesoscale -> Microscale
WRF model with increasingresolution: 18 km – top6 km – middle2 km – bottom
Red – sonicBlue – cupGreen - WRF
60 m height
Only the highest WRF resolution approaches some wind maxima andwind ramps; however, it alsooverestimates some peaks
Correlation coefficient vs. grid spacing: Tower 1 (top) and Tower 2 (bottom)
WRF ensemble #1 red, #2 blue, #3 cyan, #4 magenta for the period 10/6/12 – 8/12/13
WRF ensemble Forecasting
Similar correlation- uniformly increasing withIncreasing gridspacing
RMSE and bias vs. grid spacing: Tower 1 (top) and Tower 2 (bottom)
WRF ensemble #1 red, #2 blue, #3 cyan, #4 magenta for the period 10/6/12 – 8/12/13
WRF ensemble Forecasting
Non - uniform behavior
RMSE and bias vs. grid spacing – diurnal variation: WRF at Tower 1 (60 m)
WRF ensemble: 18 km (red), 6 km (blue), 2 km (green) for the period 10/6/12 – 8/12/13
WRF ensemble Forecasting
Largest differencesin nighttime;mainly for bias
Power spectral density for various grid spacing: WRF, Tower 1 – 60 m
Tower 1 (black), WRF: 18 km (red), 6 km (blue), 2 km (green) for the period 10/6/12 – 8/12/13
WRF ensemble forecasting
Uniform behavior inspectral domain:increasing powerspectrum for reducing grid spacing
Future wind projections for Nevada
CCSM3 model grid covering western U.S.
Future wind projections for Nevada
Histograms of the measured (solid) and climate model simulated (dashed) wind speed (left panel) and wind powerdensity (right panel) at the Tonopah tower site for the periodAugust 2003-December 2007. Number of sample pairs = 5895Weibull distribution fits are overlaid (left panel).
Future wind projections for Nevada
Annual mean wind speed (left panel) and annual mean windpower density (right panel) simulated by CCSM3 under fourdifferent emission scenarios from 2000 to 2100 for all ofNevada at 50 m.
Most of emission scenarios predict slight decrease in averagewind speed, but increase in variability
Conclusions (I)• In spite of relatively small differences in base heights
(140 m) and the horizontal difference (2700 m) between the towers, multi-year tower and sodar data over the ridges in western Nevada show significant differences in the estimation of Weibull parameters and wind power density
• There is a high correlation among the towers’ (cup and sonic) and sodar data; however, sodar overestimates both cup and sonic anemometers
• Hourly mean wind speeds are quite similar diurnally and differences are generally less than 1 m/s
• Turbulence intensity decreases at a faster rate at higher levels
Conclusions (II)• The sodar data indicates that the wind power density
increases almost exponentially beyond 60 m• Although WRF shows a uniform increase in correlation
with the increase of resolution (from 18 to 2 km), RMSE and bias do not show uniform increase with increasing resolution
• Only the highest WRF resolution (2 km) approaches some wind maxima and wind ramps; however, it also overestimates some peaks
• Power spectral density is significantly increasing for reducing grid spacing
• Future projection by a climate model indicate slight decrease of wind speeds, but higher wind variability in the 21st century
Acknowledgment: DOE-NREL support (DOE Award #: NAX-9-66014-02 (DE-AC36-08G028308)