You are on page 1of 8

See

discussions, stats, and author profiles for this publication at:


https://www.researchgate.net/publication/245169637

Weibull parameters for wind speed


distribution in Saudi Arabia

Article in Solar Energy · December 1994


DOI: 10.1016/0038-092X(94)90126-M

CITATIONS READS

139 621

3 authors, including:

Shafiqur Rehman Tahir Husain


King Fahd University of Petroleum and… Memorial University of Newfoundland
173 PUBLICATIONS 4,878 CITATIONS 132 PUBLICATIONS 1,905 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Indexing of Hybrid Wind/Solar Energy Resources over Saudi Arabia View project

FAILURE OF POWER TRANSFORMERS DUE TO CORROSIVE SULFUR View project

All content following this page was uploaded by Shafiqur Rehman on 23 October 2017.

The user has requested enhancement of the downloaded file.


Solar Energy, Vol. 53, No. 6, pp. 473-479, 1994
Copyright0 1994Ekevier Science Ltd
Pergamon Printed in the USA. All rights reserved
0038-092X/94$6.00 + .OO

003%092X( 94)00081-6

WEIBULL PARAMETERS FOR WIND SPEED


DISTRIBUTION IN SAUDI ARABIA

SHAFIQUR REHMAN,* T. 0. HALAWANI, and TAHIR HUSAIN


Meteorology, Standards & Materials Division, Research Institute, King Fahd University of
Petroleum & Minerals, Dhahran-3 126 I, Saudi Arabia

Abstract-The shape and scale parameters of a Weibull density distribution function are calculated for 10
locations in Saudi Arabia. The daily mean wind speed data from 1970 to mid-1990 are used for this purpose.
It is found that the numerical values of the shape parameter vary between I .7 and 2.7, whereas the value
of the scale parameter is found to vary between 3 and 6. It is also concluded from this study that wind data
are very well represented by the Weibull distribution function.

1. INTRODUCTION much attention has been given to the Weibull function


because it is found to give a good fit to the observed
Wind speed frequency distribution is an important pa-
wind speed data both at the surface (Wentink, 1976;
rameter for predicting the energy output of a wind en-
Justus, Hargraves, Yalcin, 1976) and in the upper air
ergy conversion system. During the last two decades,
(e.g., Baynes and Davenport, 1975). Wentink com-
ample attention was paid toward the development of
pared the Weibull functions with other distributions
an adequate statistical model for describing wind speed
like Plank’s frequency distribution, Rayleigh distri-
frequency distribution. As seen from the literature,
bution, and Gamma distribution in his study of Alas-
kan wind power potential. Justus et al. concluded that
* ISES Member the Weibull distribution gave the best fit to wind speed

Fig. I. Location map of the sites used in this study.


474 S. REHMAN,T. 0. HALAWANI,and T. HUSAIN

Table I. Summarv of dailv averaee wind sneed for all locations analvzed

Years Altitude* Instrument No. daily avg. Mean wind


Location of Data (m) height* (m) wind speed data speed (m/s)

Dhahran 19 22 10 6894 4.42


Riyadh 18 624 10 6483 3.28
Jeddah 19 17 10 6867 3.56
Tabou k 17 771 9 6274 3.05
Gizan 17 5 8 6330 3.51
Al-Jouf 15 562 7 5231 3.76
AI-Wajh 17 22 10 6316 4.28
Gassim 14 648 7 5229 3.02
Medina 17 646 10 6297 3.46
Nejran II 1275 8 429 1 2.71

* Taken from Saudi Arabian Wind Energy Atlas (1986).

data from more than 100 stations of the United States where u is the expected mean wind speed and r is the
National Climatic Center. In a study conducted in gamma function.
Denmark, Petersen et al. ( 198 1) found that Weibull As summarized by Stevens and Smulders ( 1979),
distribution gave an excellent fit to the wind speed dis- the values of k and c can be obtained from the method
tribution. of moments, the method of energy pattern factor, the
The Weibull function is a two-parameter function, maximum likelihood method, the Weibull probability
namely, shape parameter k and scale parameter c. It paper method, or by the use of percentile estimators.
is used in this paper for describing the wind speed fre- Stevens and Smulders calculated the values of both the
quency distribution. There are several methods avail- parameters using all five methods mentioned above.
able in the literature for the calculation of these two They obtain almost the same values with all five meth-
parameters as stated by Stevens and Smulders ( 1979). ods. However, in this paper, the authors have adopted
In this study, the shape parameter k is estimated by the method described by Hennessey ( 1978) because
the nomogram of Kotel’nikov (Johnson and Kotz, of its simplicity and availability of sample mean and
1970). The scale parameter c is calculated using the standard deviation values. This method requires the
wind speed relationship given in Hennessey (1978). values of the sample mean wind speed and standard
In this paper, daily average wind data which cover the deviation. Knowing these two values for a sample of
period from 1970 to 1990 are used. The details of the data set, the shape parameter k is conveniently obtained
wind data are described, in detail, in Section 3. Section from the nomogram of Kotel’nikov. The scale param-
2 deals with the mathematical formulation of the Wei- eter c is calculated from eqn (3).
bull density distribution function.

3. DESCRIFTION OF OBSERVED WIND DATA


2. WEIBULL DlSTRlBUTlON FUNCTION
The daily observed mean wind speed data is used
The Weibull distribution function, which is a two- for determining the values of Weibull parameters for
parameter function, is expressed mathematically as IO sites in Saudi Arabia. The locations of these sites
follows: are shown in Fig, 1. A summary of all the data sets
analyzed in this study, is given in Table I. The daily
average wind speed data is obtained from the Data
f(u) = (~)(~)ew[
- (F)k] (1)

Table 2. Numerical values of Weibull


where k is the scale parameter, c is the shape parameter, parameters for annual wind speed
and u is the wind speed. The cumulative distribution distribution
function is obtained by integrating eqn ( 1) and takes
Location k c
the form:
Dhahran 2.35 4.98
F(u) = 1 - exp{ -(v/c)“} (2) Riyadh 1.95 3.70
Jeddah 2.35 4.02
Tabouk 2.00 3.44
The mean of the Weibull distribution, i.e., the mean Gizan 2.50 3.96
wind speed, is expressed in terms of the gamma func- Al-Jouf 2.60 4.24
Al-Wajh 2.50 4.83
tion and is given by:
Gassim 1.95 3.41
Medina 2.40 3.91
Nejran 2.05 3.07
u=cr(l+ l/k) (3)
Weibull parameters 415

15

35 40 46 50 6s
LONGITUDE

Fig. 2. Contour map showing the variation of shape parameter k.

3s -

30

25
w
2
;2
4
20

15

Fig. 3. Contour map showing the variation of scale parameter c.


476 S. REHMAN,T. 0. HALAWANI,and T. HUSAIN

0 2 4 6 8 10 12 0 2 C 6 8 10 12

WIND SPEED [m/s) WIND SPEED [m/s)

._

35 -1 TABOUK k = 2.00
c q 3.44

0 2 4 6 8 10 12 0 2 4 6 8 10 12

WIND SPEED (m/s) WIND SPEED (m/s)

Fig. 4. Wind frequency histograms for Dhahran, Riyadh, Jeddah, and Tabouk. The superimposed solid
curves are the Weibull fit obtained from calculated values of k and c

Processing Center of the King Fahd University of Pe- installed at 7, 8, and 9 m above ground surface. The
troleum and Minerals, Dhahran. This data is well monitoring instruments meet the WMO Class-I-spec-
maintained by the Saudi Arabian Meteorological and ifications. As noted in the Saudi Arabian Wind Energy
Environmental Protection Administration (MEPA) Atlas, the instruments were inspected and cleaned on
and covers the period from 1970 to mid 1990. No data a daily basis during strong wind periods. The instru-
was collected during 1976. As summarized in Table 1, ments were maintained using detailed maintenance
at some stations, the data was not collected in 1970, procedures on a regular basis once every 3 months. In
197 1, and 1972. In 1984, the data was available only order to maintain the WMO standards, the wind sen-
for the first 5 months of the year for all the stations. sors were calibrated in the wind tunnel at least once
In 1990, the data was available only for the first 6 every 6 months. The surface roughness coefficient, Z0
months for the Dhahran, Jeddah, and Riyadh sites. In (dimensionless), varies between 0.10 and 0.4 for
calculating the long term averages of the wind speed, smooth surfaces such as sea, sand, and snow to very
the missing values were not considered. rough surfaces such as urban areas and tall buildings,
The Saudi Arabian Wind Energy Atlas ( 1986) re- respectively.
ports that the wind sensors are located at a height of As noticed from the mean wind speed data, Dhah-
10 m above ground surface at most of the meteoro- ran observed a maximum mean wind speed of’4.42
logical data acquisition stations. This height of the sen- m/s while a minimum average wind speed of 2.7
sors is recommended by the World Meteorological Or- m/s is observed at Nejran. At other locations the av-
ganization (WMO). At some stations, the sensors are erage wind speed lies between 3 m/s and 3.75 m/s.
Weibull parameters 477

45 I
k = 2.50 AL-JOUF k = 2.60
c = 3.96 c = 4.34

30

t;

h
= 25
2
E 20
LL

is 15

r, 6 8 10 12 0 2 I 6 8 10 12
WIND SPEED (m/s1 WIND SPEED (m/s)

1 GASSIM k q 1.95
c = 3.&l

30

t
$ 25

I
2
2 20
IL

is 15

IO

0
0 2 4 6 8 lo 11 2 6 B 10 12

WIND SPEED (m/s) WIN0 SPEED (m/s1

Fig. 5. Wind frequency histograms for &an, Al-Jouf, Al-Wajh, and Gassim. Superimposed solid curves
are the Weibull fit obtained from calculated values of k and c.

i.e.. greater than 4 at Dhahran, Jeddah (coastal sites),


4. RESULTS AND DISCUSSION
and Al-Jouf or AI-Wajh (northern areas). In Riyadh
The Weibull parameters, namely, shape parameter and Gassim this value lies around 3.5. Variation of
k and scale parameter c, are calculated for 10 locations the scale parameter c is shown in Fig. 3. A maximum
in Saudi Arabia based on annual and monthly daily value of c equals 4.98 is observed at Dhahran and a
average wind speed data. Table 2 summarizes the nu- minimum of 3.07 at Nejran. Regions of high, medium,
merical values of k and c for all the locations studied and low values of the scale parameter are clearly ex-
in this paper. It is noticed that the value of the shape plained in this figure.
parameter lies around 2.5 in the northern and southern Figures 4, 5, and 6 show the wind speed histograms
regions ofthe country, whereas at coastal sites this value of the percent frequency distribution for all 10 loca-
is found to be around 2.35. A lower value of k, i.e., tions. The Weibull curves obtained using the numerical
1.95, is obtained in and around the central part of the values of k and c for different locations are superim-
country. The contour map plotted for the shape pa- posed on the wind speed histograms. The wind speed
rameter is shown in Fig. 2. This map depicts lower data for Dhahran, Riyadh, Jeddah, Tabouk, Al-Jouf,
values of the shape parameter in the central region, Gassim, Medina, and Nejran are very well represented
namely, Riyadh and Gassim, and higher values in the by the Weibull distribution density function.
northern and southern regions. Moderate values are The Weibull parameters obtained for the monthly
seen in the eastern and western parts of the country. wind speed data are summarized in Tables 3 and 4 for
The scale parameter c is found to have higher values, Dhahran, Riyadh, Jeddah, Tabouk, Gizan, Al-Jouf,
S. REHMAN, T. 0. HALAWANI, and T. HUSAIN

c=
418

NAJF \N k = 2.05

1
35 3.07

30 30
2-
z” 25 t
z 25
3 !i
g 20 2 20
B E
>0 15

10 10

0 $ 15
0 2 4 6 8 IO 12 0 2 4 6 a 10 12

WIND SPEED [m/s) WIND SPEED (m/s]


Fig. 6. Wind frequency histograms for Medina and Nejran. Superimposed solid curves are the Weibull fit
obtained from calculated values of k and c.

Al-Wajh, Gassim, Medina, and Nejran. It is observed February. In Tabouk and Al-Jouf, the shape parameter
that in the eastern and western regions (Dhahran, Jed- k shows an increasing trend from January to July and
dah, and Medina), k had its maximum value in June then a decreasing trend towards the end of the year.
and its minimum in December. In Dhahran, k varies In Al-Wajh, k varies between 2.6 and 2.7, and no def-
between 2 and 2.6, in Jeddah it varies between 2.45 inite trend is observed. In southern cities, i.e., Gizan
and 2.55, and in Medina between 2.15 and 2.8. The and Nejran, maximum values of k are noticed in Oc-
monthly variation of k shows an increasing trend from tober and August, respectively, and minimum values
January to June and a decreasing trend toward De- are noticed in December. The shape parameter in
cember for all the locations in the eastern and western Nejran show an increase from January to August and
regions of Saudi Arabia. This region also has a maxi- then a decrease toward December.
mum value of k parameter. No definite trends in the variation of the scale pa-
In Riyadh and Gassim, the maximum values of k rameter care observed in any city. Certain similarities
are noticed in July and June and the minimum values in the occurrence of maximum and minimum values
are noticed in November and December. Over the year, are noticed. In Dhahran, Jeddah, and Medina, the
the shape parameter increases at the start of the year maximum values are noticed in June and March while
and, after reaching its maximum in June and July, it the minimum values are found in the month of Oc-
decreases toward the end of the year. tober. In Riyadh and Gassim, the maximum is found
The maximum values of the shape parameter are in July and May while the minimum is found in Oc-
found in the month of July for Tabouk, Al-Jouf, and tober and September, respectively. The northern cities
Al-Wajh cities (in the northern region of the country) have the maximum value of scale parameter in March
and the minimum values are found in December and and April and the minimum in November and De-

Table 3. Numerical values of Weibull parameters for monthly wind speed distribution
for Dhahran. Riyadh, Jeddah, Abouk, and Gizan

Dhahran Riyadh Jeddah Tabouk Gizan

Month k c k c k C k C k C

JAN 2.40 4.11 1.90 3.55 2.10 4.17 1.75 2.92 2.45 3.81
FEB 2.45 4.85 1.85 3.89 2.05 4.27 1.70 3.42 2.45 3.82
MAR 2.55 5.15 2.00 4.31 2.25 4.46 1.90 4.0 I 2.50 3.96
APR 2.40 5.06 1.95 3.93 2.25 4.35 2.10 4.26 2.45 3.82
MAY 2.40 5.52 2.00 3.86 2.37 4.20 2.00 3.97 2.50 3.94
JUN 2.60 6.51 2.00 4.42 2.45 4.34 2.40 3.85 2.50 4.04
JUL 2.50 5.54 2.05 4.56 2.40 4.02 2.50 3.78 2.50 4.69
AUG 2.30 4.9 I 1.85 3.87 2.55 3.97 2.40 3.59 2.50 4.57
SEP 2.20 4.18 1.90 3.12 2.40 3.82 2.40 3.33 2.60 3.79
OCT 2.05 4.09 1.75 2.56 2.05 3.16 2.15 3.02 2.60 3.59
NOV 2.20 4.38 1.72 2.85 2.10 3.43 1.85 2.56 2.50 3.7 1
DEC 2.00 4.68 1.83 3.24 2.00 3.90 1.70 2.49 2.45 3.13
Weibull parameters 479

Table 4. Numerical values of Weibull parameters for monthly wind speed distribution
for Al-Jouf, Al-Wajh, Gasim, Medina, and Nejran

Al-Jouf Al-Wajh Gassim Medina Nejran

Month k C k C k c k C k c

JAN 1.85 3.93 2.45 4.83 1.81 3.15 2.20 3.80 1.95 2.48
FEB 1.90 4.64 2.40 4.98 1.75 3.57 2.25 3.95 1.95 2.98
MAR 2.00 5.03 2.55 5.27 2.00 4.09 2.45 4.43 2.10 3.14
APR 1.95 5.10 2.55 5.18 2.07 4.00 2.50 4.23 2.35 3.34
MAY 2.00 4.51 2.45 5.02 2.15 4.26 2.65 4.05 2.35 3.39
JLJN 2.19 4.6 I 2.60 5.09 2.1 I 3.61 2.60 3.96 2.35 3.35
JUL 2.35 4.48 2.60 4.58 2.05 3.50 2.80 4.30 2.40 4.00
AUG 2.15 4.00 2.60 4.51 1.91 3.08 2.65 4.08 2.50 3.85
SEP 2.00 3.71 2.70 4.96 I .90 2.66 2.45 3.48 2.40 3.29
OCT 2.00 3.71 2.50 4.43 1.75 2.85 2.30 3.31 2.10 2.65
NOV I.91 3.68 2.60 4.42 1.80 3.15 2.30 3.60 2.00 2.31
DEC I .82 3.55 2.58 4.53 1.75 2.91 2.15 3.54 1.90 2.32

cember. The maximum values of scale parameter for culated at hub height by using the power law, which
the Gizan and Nejran cities are observed in July. Their generally holds good up to a height of 100 meters
minimum values are noticed in October and No- (Gupta, 1986).
vember. The contour maps plotted for the shape and scale
parameters may be a good tool for the rough estimate
of k and c at locations where the wind data is not avail-
5. CONCLUSIONS able. These contour maps may further be improved if
It is found that the values vary between 1.95 and more locations are added.
2.6 for annual wind speed distribution. It is concluded A~knoM,/~~~~ment-The authors wish to acknowledge the sup-
that the central region (the Riyadh and Gassim area) port of the Research Institute of King Fahd University of
observes the minimum values of the k parameter while Petroleum and Minerals, Dhahran, Saudi Arabia.
the northern region (Al-Wajh, Al-Jouf) observes the
maximum values. The same values of k for Dhahran REFERENCES
and Jeddah show a similarity between the two sites of
T. Wentink. Study of Alaskan wind power potential and its
being located at the coast. The scale parameter varies
possible applications. Final Rep.. Rep. No. NSF/RANN/
between 3.07 and 4.98. As for the k parameter, the SE/AER 74-00239/FR-76/ 1, Geophysical Institute, Un.
values of the c parameter have their maximum at Al- of Alaska (1976).
Jouf and Al-Wajh, but their minimum value is found C. G. Justus. W. R. Hargraves. and A. Yalcin, Nationwide
assessment of potential output from wind powered gen-
at Nejran. The central region has moderate value of c
erators, J. .4ppl. M&wr. 15, 673-678 ( 1976).
around 3.5. C. J. Baynes, and A. G. Davenport, Some statistical models
For monthly wind speed distribution, it isconcluded for wind climate prediction. Preprints Fourth Conf. Prob-
that the shape parameter has its maximum values in ability and Statistics in the Atmospheric Sciences. Talla-
the months of June and July for all the cities except hassee. Fla.. Amer. Meteor. Sot., l-7 ( 1975).
E. L. Petersen, I. Troen, S. Frandsen, and K. Hedegaard, Wind
Gizan and Nejran. The minimum values for all the
atlas for Denmark., RISO, Denmark ( 1981).
cities are found in December. In general, an increasing M. J. M. Stevens and P. T. Smulders. The estimation of the
trend of the shape parameter is noticed from January parameters of the weibull wind speed distribution for wind
to June/July and then a decreasing trend towards the energy utilization purposes. Win.gEng. 3,132-145 ( 1979).
N. L. Johnson and S. Katz. Continuous univariate distribu-
end of the year. It is also concluded that the scale pa-
tions, Vol. I, Houghton Miffin, New York ( 1970).
rameter has its maximum values in March, April, May, J. P. Hennessey, Jr.. A comparison ofthe Weibull and Rayleigh
and June and its minimum in October, November, distribution for estimating wind power potential. Bind
and December. Eng. 2, 156-164 (1978).
The values of the shape parameter can be used at Saudi Arabian Wind Energy Atlas, KACST, KFUPM. Dhah-
ran, Saudi Arabia, 67 pp. ( 1986).
hub height without any modification (Gupta, 1986), B. K. Gupta, Weibull parameters for annual and monthly
but the values of the scale parameter c need modifi- wind speed distribution for five locations in India. .So/ur
cation. The values of the scale parameter could be cal- Energy. 37 (6), 469-471 (1986).

View publication stats

You might also like