£'/'/ A"

i SERI/TP—2H"197 3

DE83 011963
SERI/TP-211-1978
UC Category: 60
DE83011963
(lofOF-Sf30^<?S(—/<?

MASTER
Height Extrapolation of
Short- and Long-Term
Averaged Wind Data

Amir S. Mikhail

May 1983

To be presented at the 1983 Annual Meeting of the

American Solar Energy Society
Minneapolis, Minnesota
1-3 June 1983

Prepared under Task No. 1067.10

WPA No. 171

Solar Energy Research Institute

A Division of Midwest Research Institute

1617 Cole Boulevard

Golden, Colorado 80401

Prepared for the

U.S. Department of Energy
Contract No. EG-77-C-01-4042

DISTRIBUTIOK OF THIS DOCUMENT IS UNLIMITED

 DISCLAIMER

This report was prepared as an account of work sponsored by an

agency of the United States Government. Neither the United States
Government nor any agency thereof, nor any of their employees,
makes any warranty, express or implied, or assumes any legal liability
or responsibility for the accuracy, completeness, or usefulness of any
information, apparatus, product, or process disclosed, or represents
that its use would not infringe privately owned rights. Reference
herein to any specific commercial product, process, or service by
trade name, trademark, manufacturer, or otherwise does not
necessarily constitute or imply its endorsement, recommendation, or
favoring by the United States Government or any agency thereof. The
views and opinions of authors expressed herein do not necessarily
state or reflect those of the United States Government or any agency
thereof.

D IS C L A IM E R

Portions of this document may be illegible in electronic image

products. Images are produced from the best available
original document.
 SERI/TP-1978

Height Extrapolation of Short- and Long-Term

Averaged Wind Data

Amir S. Mikhail
Solar Energy Research Institute
1617 Cole Boulevard
Golden, Colorado 80401

ABSTRACT aging period presents a distinctly different

problem. The long-term averaged wind data
Various models that are used for height are functions of the statistics of occurrence
extrapolation of short and long-term averaged of various atmospheric parameters and can
wind speeds are discussed. Hourly averaged only be predicted through empirical models.
data from 3 tall meteorological towers (the The extrapolation of short-term average wind
NOAA Erie Tower in Colorado, the Battelle speeds over flat terrain is well understood
Goodnoe Hills Tower in Washington, and the by the Monin-Obukhov similarity theory (1).
WKY-TV Tower in Oklahoma), together with data The theory applies to the inertial sublayer,
from 17 candidate sites (selected for pos­ where the turbulent shear stresses and heat
sible installation of large WECS) were used fluxes are independent of height. The depth
to analyze the variability of short-term of this layer can vary from 20 to 200 m (2)
average wind shear with atmospheric and sur­ depending on the stability conditions.
face parameters, and Che variability of the
long-term Weibull distribution parameter with 2. EXTRAPOLATTOM OF SHORT-TERM
height. AVERAGE WIND SPEEDS OVER RELA­
TIVELY FLAT TERRAIN
The exponents of a power law model, fit to
the wind speed profiles at the three meteor­ 2.1 The Similarity Model
ological towers, showed the same variability
with anemometer level wind speed, stability, The similarity model Is based on the
and surface roughness as the similarity law Monin-Obukhov similarity theory. It only
model. Of the four models representing applies for relatively flat terrains and
short-term wind data extrapolation with within the depth of the inertial sublayer of
height (1/7 power law, logarithmic law, power the atmospheric boundary layer. To express
law, and modified power law), the modified the relationship in a power law form between
power law gives the minimum rms for all can­ heights Z and Za, the wind speed U is given
didate sites for short-term average wind by
speeds and Che mean cube of the speed. The
U (Z)/U(Za) - (Z/Za)*6 , [1]
modified power law model was also able to
predict Che upper-level scale factor for the
WKY-TV and Goodnoe Hill Tower data with where Che effective exponent a is given by
higher accuracy. All models were not suc­
cessful in extrapolation of the Weibull shape 6 (Zjj/L)
factors. ** “ VZg/Z0) - K Ug/1.) (21

1. DITRODCCTION where

The extrapolation o'f wind speed data measured 0 and <li are universal functions of (Z_/L)
(3) 4
at a certain height (e.g., anemometer height)
to another height (e.g., hub height) is a 1Z'5
Zg is the geometric mean height (Z*Za)J-'“
point of interest to many wind energy appli­
cations. The extrapolation of wind speed Z0 is the surface toughness length
data to different heights varies considerably L is the Monin-Obukhov length.
depending on whether the extrapolation is
conducted over complex or relatively flat The surface roughness length is a physical
terrain. It is also dependent on whether parameter that determines the impedance to
short-term averaged (e.g., 10 minute - 1 the wind flow. Water and ice have lower sur­
hour) or long-term averaged (e.g., monthly, face roughness lengths than wooded and urban
annual) wind data are considered. Each aver- areas. L is a measure of the combined mech-

NGTICE
PORTIONS OF THIS REPORT ARE ILLEGIBLE.
It has been reproduced from the best
available copy to permit the broadest l
possible availability.
DISTRIBUTION OF THIS DOCUMENT IS UNLIMITED
 SERI/TP-1973

anical and thermal turbulence in the (Washington); and the WKY-TV Tower data in
atmosphere. This quantity is in turn Oklahoma City (Oklahoma). The power law
physically related to surface layer turbu­ model is given by
lence caused by wind shear, surface heat
flux, and surface roughness length. By U(Z)/U(Za) - (Z/Zj)2? [4]
examining wind speed temperature profiles and
surface radiation from various meterological where
towers, a universal function relating I. to
the anemometer level wind speed, net ap = a + b ia[U(Za)] ;
radiation, and surface roughness is estab­
lished (4). Figure 1 gives the effective a - 0.37/[i - 0.088 ^(Zj/lO)]
exponent for different 10-m level wind
speeds, surface roughness lengths, and radia­ b - -0.088/[1 - 0.088 ^(Za/LO)] .
tion indices. A negative index indicates
outgoing radiation (nighttime), and a large The coefficients a,b were obtained by equat­
positive index indicates a high incoming ing the probability of occurrence of an upper
radiation. The graph shows clearly that as level wind speed at height Z that corresponds
the anemometor level wind speed increases to a lower level wind speed at height Za for
beyond a certain value, which is a function the four sets of tower data.
of surface roughness and insolation, the
effective exponent converges to the 1/7 power 2.3 Modified Power Law Models
law value.
To combine the accuracy of a theoretical
The effect of stability is also clearly model with the simplicity of an empirical
demonstrated at low radiation index value model, the modified power law model was sug­
(nighttime) compared to high radiation index gested (3). The model is surface roughness,
value (daytime). To eliminate the effect of anemometer level wind speed, and height
stability on the effective power law dependent. However it represents average
exponent, limits of equation 2 are taken as stability conditions (Figure 1). The model
l/L —- '■ 0 or L ■■ — -a for neutrally stable is based on the power law model and the
conditions, and the resulting exponent is the similarity model. The modified power law is
logarithmic power law model exponent. given by

ae - 1/^ (Zg/20) [3] ” am + bin[U(Za)] [5]

Hence, the 1/7 power law is the limit of the where

similarity model for high anemometer level
a^ » 1/VZ3/Z0) + [0.088/
wind speeds, and the logarithmic law is the
limit of the similarity model for neutrally
stable conditions. Figures 2 and 3 show the 1 - 0.088 l (Z /10)] .
n a 1 * * * * * * *
*
exponent variability for the Goodnoe Hills
Data taken at the U.S. candidate sites, for
and the Erie Towers, respectively. Goodnoe
Hills and Erie Towers had a surface roughness possible installation of large DOE turbines,
length of 0.05 and 0.11 m (estimated by were used to compare the performance of the
Hattelle-Pacific Northwest Laboratory), modified power law model with the frequently
respectively; a geometric mean height of 32.7 used 1/7 power law and logarithmic models.
and 38.7 m, respectively; and an anemometer Since stability information was not available
level of 10 m. Figure 4 compares the vari­ for candidate sites, the similarity model was
ability of the theoretical Monin-Obukhov wind
not used. Table 1 shows the observed average
speed exponent for Z0= 0.1 m and Z ” 22.3 m
wind speed at hub height V0 versus the
with the WKT-TV tower exponent for the hourly
averaged data with Z “ 25 m and Z0a 0.07 m average predicted wind speed at the hub
for nighttime and daytime. There is fairly height using the modified power law model
good agreement between the observed and the vTl; the power law model vF; the logarithmic
predicted exponents. The slight variance is model VL; and the 1/7 power law model V7.
due to lumping all the stability categories The table also gives the observed and pre­
into nighttime and daytime for the WKY-TV dicted mean cube of the wind speed at hub
Tower data, while the stable and unstable
height
curves represent stability categories -2 and
2 on a scale of -2 (stable) to 6 (unstable).
VO3, VM3, VP3, VL3, V73.
2.2 Power Law Model
The modified power law model gives the mini­
The power law model (5) is an empirical model mum rms error for all candidate sites for the
based on the height variability of long-term wind speed (0.28 m s_ , 5.9Z) and for the
averaged data in four sets of meteorological mean cube of the speed (115.2 mJ s~ ,
tower data: Kennedy Space Flight Center 16.6Z). Although the logarithmic law seems
(Florida); Wallops Island (Virginia); Hanford to perform as well as the modified power law
 SERI/TP-1978

0.7
Zq = 22.36 \ Equivalent ------------ Night
0.6 Power Law
' Exponent
\ a, for
0.5 Radiation ' Z, = 0.1 m
Index \
0.4 Modified \
T Power Law \
£ 0.4 -
'0.3
Model \
1/7 Power Law* 10
0.2

0.1
Anemometer Level Wind Speed
ol i i i i i i i
0 1 2 3 4 6 8 10
10 m Wind Speed (m/s-1)

Figure 1. Effective Exponent for Various Figure 2. Exponent Variability for the
Parameters Goodnoe Hills Tower

-------- Night

--------- All

c 0-3 -
c 0.4

uj 0.2 Staple

1/7 Power Law

Unstable

Anemometer Level Wind Speed

Anemometer Level Wind Speed

Figure 3. Exponent Variability for the Figure A. Exponent Variability for the
Erie Tower WKY-TV Tower

in predicting the average wind speed, the ?(V) - 1 - exp C-(V/G)k] [6]
error associated with predicting the mean
cube of the speed is higher. This will have a where
direct effect in the estimation of power
output at hub height. C is the scale factor
k is the shape factor
3. SXTrtAPQiATIOH OF LONG-TERM
AVERAGE 'WIND DATA OVER RELATIVELY P(V) is che probability density function.
FIAT TERRAIN
3.1 Model Descriptions
Long-term average wind data include wind
speeds averaged over a period of several The extrapolation of long-term wind data,
hours or more and a long-term frequency dis­ that includes average wind speed and Weibull
tribution parameters. The Weibull distri­ distribution parameters, following the power
bution was found to adequately fit observed law model (equation A) is given by
wind speed distributions. The Weibull
distribution is given by: C2/Ci » (Z2/Zi)“ , [7]

3
 SERI/TP-1978

TABLE 1 OBSERVED VS. PREDICTED 3DB HEIGHT VITO SPEEDS (o/s) AND MEAN CUBED
WIND SPEEDS (a3/s3) ?OR CANDIDATE SITE DATA

Sica VP VO3 77 3
Code HI 112 VO VM VL V7 VL3 Vp3

LOG 18.2 45.7 7.56 7.31 7.36 6.52 6.07 815.6 B07.9 924.3 571.4 511.4
CAO 9.1 45.7 7.50 7.35 7.42 7.75 7.01 698.7 693.3 330.5 815.3 699.6

AGP 18.2 45.7 8.92 8.48 3.o9 8.57 8.15 1179.1 963.6 1112.7 995.0 916.7
CLB 9.1 45.7 6.77 6.47 6.57 8.66 7.38 415.7 361.4 411.8 864.5 710.4
KGS 9.1 45.7 6.84 6.64 6.73 7.99 7.26 611.7 518.6 635.3 903.0 796.9
RSL 9.1 45.7 7.56 7.64 7.72 7.82 7.07 710.5 763.6 918.0 820.1 705.2
PAA 9.1 45.7 6.67 6.35 6.35 7.34 6.57 573.2 441.7 507.1 680.5 562.3
SON 9.1 45.7 3.02 7.95 8.29 8.71 8.14 1387.1 1105.7 1475.6 1454.2 1394.1
HON 9.1 45.7 6.94 6.68 6.64 7.05 6.27 600.3 518.0 590.2 608.3 497.2
BON IS.2 45.7 7.96 7.65 7.81 7.73 7.32 1011.7 959.4 1146.6 990.7 944.6
WTK 18.2 45.7 7.49 7.34 7.47 7.74 7.30 747.7 718.6 829.4 340.8 775.0
BID 9.1 45.7 7.46 7.28 7.22 6.86 6.06 696.1 536.4 710.3 530.9 420.3
HOL 18.2 45.7 7.32 7.33 7.34 6.35 5.87 771.4 684.9 751.2 444.2 385.2
AHA 9.1 45.7 8.04 7.92 6.55 3.73 6.47 795.0 715.0 523.0 958.9 504.0
BON 9.1 39.6 5.16 5.84 5.73 5.99 5.30 308.5 398.9 440.0 423.3 348.6
BON 9.1 70.1 5.48 6.40 6.23 6.82 5.75 357.9 494.1 565.4 596.0 445.3
KAN 9.1 27.4 7.40 7.16 7.28 7.43 7.01 837.7 677.8 776.7 756.4 692.0
KAN 9.1 54.9 7.79 8.20 8.43 8.52 7.74 971.4 963.8 1205.6 1083.1 932.1
RMS 0.28 a/s 0.48 a/s 0.82 a/s 0.83 a/s 115.2 a/s 131.0 i3/.3 194.5 »3/s3 188.0 «3/s3
error 5.92 6.32 12.42 11.12 16.62 22.32 37.32 23.52

where — _ 1 ~ log V^/log Vh

012 ' Io 1 - <x0 log^/Z^/log Vh
a - a + b I^C],
[10]

k2/kl * [1 -0.088 (Z1/10)]/ where

[i - 0.088 ln (Z2/10)] ; [8] (20/Zr)0-2

surface roughness exponent
homogeneous wind speed (a2
0) (m s'1)
and
steady wind speed at elevation Z,
V2/V1 - (Z2/Z1)C£l , [9] (n s'1)
reference elevation (10 a)
where surface roughness length (m)

- a + b In V1 ! and where the Weibull parameters C and ’x are

related by:
where C^ and are the Weibull scale and
shape parameters, respectively, corresponding
C2 - C^Zz/ZO** [U]
to height Z|j and C^ and ’*2 are c^e Weibull
scale and shape parameters corresponding to
height Z2- V"2 and are long-term average where
wind speeds at height Z-^ and Z2, 1 - log C]/log Vh
respectively. £Zc 2° i - a0 log (Z1/Zr)/iog Vj,

The height surface roughness, stability and

dependent, modified power law model is given 1 - a0 log (Zi/Zr)/log Vh
by the same relation as equations 7-9 except <Z ’ kl 1 - ot0 log (Z2/Zr)/log Vh •
coefficient a is replaced by a^ from
equation 5. [12]

Spera and Richards (6) developed a surface 3.2 Model Verification

roughness dependent power law model for long­
term average _speeds and Weibull parameters. The hourly averaged data were used to cal­
The exponent a2 is given by culate the lower-level Weibull distribution

4
 SERI/TP-1978

parameters, and the suggested models were Observed

used to predict the upper-level Weibull para­ Velocity - Surface .
meters for the NOAA Erie Tower in Colorado, Roughness Law
the Battelle Goodnoe Hills Tower in Power Law
Washington, and the WKY-TV Tower in S 3.0 Modified Power Law
Oklahoma. The observed upper-level Weibull
parameters were then compared with the pre­
dicted values from the power law and modified
power law models with the velocity-surface
roughness models (equations 10-12). The sur­
face roughness length was 0.11 m for the Erie
3 5.0 -
Tower, 0.07 m for the Goodnoe Hills Tower,
and 0.05 a for the WKY-TV Tower.

Figures 5 through 7 show clearly that the

modified power law model (roughness—,
height-, and velocity-dependent) generally
Aneometer Levels
predicts upper-level parameters with higher
accuracy than the roughness- and velocity-
dependent NASA power law and the power law
models. However, all models failed to accu­ Figure 6. Upper-Level Weibull Parameters
rately predict the scale factor for the upper Predicted at che WKY-TV Tower ^
levels of the Erie Tower. This might be
attributable to the complex terrain for
certain prevailing wind directions near the Observed
Erie Tower. The maximum observed error in Power Law
the C value for the modified power law model Modified Power Law /•
is <22 for the Goodnoe Hills Tower data, 62 Velocity - Surface /' ;
for the WKY-TV Tower data, and 282 for the flouohness Law /' .Y
Erie Tower data. Except for the WKY-TV Tower
data, all tower data show an increase in the u- 6.0
scale factor prediction error with height,
which is attributable to the breakdown of the
similarity assumption with the increase of
height from the ground.

----------- Observed
----------- Velocity - Surface
Roughness Law Anemometer Levels
(NASA)
------- Power Law
(Justus and Mikhail)
------- Modified Power Law Figure 7. Upper-Level Weibull Parameters
(Mikhail and Justus) Predicted at the Erie Tower

Figures 3 through 10 show the predicted shape

factor by the modified and power law models
(identical k projection) and the NASA
model. All models seem to overestimate the
values for the Erie and Goodnoe Hills Towers,
and underestimate the value for che WKY-TV
Tower. The maximum error is about 252, which
can be significant in the prediction of tur­
bine power output at hub height.
Anemometer Levels
4. CONCLUSIONS

The observed power law exponent for all three

towers showed strong dependence on the
Figure 5. Upper-Level Weibull Parameters anemometer level wind speed and atmospheric
Predicted at the Goodnoe Hills stability (nighttime and daytime). It also
Tower exhibited a high degree of dependence on
extrapolation height with respect to

5
 SERI/TP-1978

------------ Observed anemometer height. These dependences became

-------------Velocity - Surface less severe as the anemometer level wind
Roughness Law speeds were increased due to the dynamic mix­
Power and Modified ing of the atmospheric boundary layer.
Power Law
The data from candidate sites were used to
W ejbu|| shape Fac , or

compare the performance of the 1/7 power law

model, the logarithmic law model, the power
law model, and the modified power law
model. The 1/7 power law and the surface
roughness dependent logarithmic law are com­
monly used for height extrapolation. The
power law is an empirical model that is
dependent on the lower level wind speed. The
modified power law is a semi-empirical model
that is height, surface roughness, and lower
wind speed dependent. The modified power law
model had the minimum rms error for all can­
didate sites for wind speed (0.28 m s” ,
Anemometer Levels 5.7J) and tot mean cube of the speed
(115.2 m3 s*3, 16.6Z).
Figure 8. Predicted Shape Factor of
the WKY-TV Tower
The three models used for Weibull distri­
bution parameters extrapolation ware the
power law, the modified power law, and the
-------- Observed
velocity-surface roughness dependent
-------- Velocity - Surface
models. The models projected the scale para­
Roughness Law
meter C fairly accurately for the Goodnoe
-------- Power and Modified
Hills and WKY-TV Towers and were less accu­
W eibull Shape Factor

Power Law
rate for the Erie Tower. However, all models
have overestimated the C value. The maximum
error for the modified power law model was
<2% for Goodnoe Hills, 6% for WKY-TV, and 28Z
for Erie. The error associated with the pre­
diction of the shape factor (k) was similar
for the three models. It ranged from 20% to
25%.

5. Acknowledgment

Anemometer Levels This work was supported in part by Battelle

Pacific Northwest Laboratories and by the De­
Figure 9. Predicted Shape Factor of partment of Energy.
the Goodnoe Hills Tower

6. REFERENCES
3.0 ------------ Observed
------------ Velocity - Surface 1. Monin, A. S.; M. A. Obukhov.Dimensionless
Roughness Law Characteristics of Turbulence in the Surface
-------------Power and Modified Layer. Tr. No. 24, Akad, Nauk. SSSR Geofig.
W eibull Shape Faclor

2.5 Power Law Inst., (1954), pp. 162-187.

Y.Panofsky, H. S."Vind Structures in Strong
Winds Below 150 m," Wind Engineering, 1,
2.0 No. 2, pp. 91-103 (1977).
3. Mikhail, A. S.; C. G. Justtis."Comparison of
Height Extrapolation Models and Sensitivity
Analysis," Wind Engineering, 5, No. 2
(1981).
4. Mikhail, A. S.; C. G. Justus."Dependence of
Monin-Obukhov Length on Surface Roughness
l.Ol---------- 1---------- 1---------- :^i i Length and Its Relationship to Pasquill -
0 1 2 3 4 5 6 Gifford Stability Classes," Wind Technology
Anemometer Levels Journal, No. 3-4 (1978).
5. Justus, C. G.; A. S. Mikhail ."Height Vari­
Figure 10. Predicted Shape Factor of ation of Wind Speed and Wind Distribution
the Erie Tower Statistics," Geophys. Res. Letters, _3^ 261-
264 (1976).

6
t

SERI/TP-1978

6. Spera, D. S.; T. R. Richards. "Modified

Power Law Equations for Vertical Wind Pro­
files," Proceedings of the Conference and
Workshop on Wind Characteristics and Wind
Energy Siting 1979, American Meteorological
Association, Boston, Massachusetts (1979).
7. Justus, C. G. Wind and Wind System Perfor­
mance. Franklin Institute Press,
Philadelphia, Pennsylvania, (1978).