This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

Scribd Selects Books

Hand-picked favorites from

our editors

our editors

Scribd Selects Audiobooks

Hand-picked favorites from

our editors

our editors

Scribd Selects Comics

Hand-picked favorites from

our editors

our editors

Scribd Selects Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

P. 1

10.1.1.137|Views: 20|Likes: 0

Published by Thanon Daengmanee

See more

See less

https://www.scribd.com/doc/62563220/10-1-1-137

08/18/2011

text

original

**www.elsevier.com/locate/renene
**

Stochastic generation of hourly mean wind

speed data

Hafzullah Aksoy

Ã

, Z. Fuat Toprak, Ali Aytek,

N. Erdem U

¨

nal

Department of Civil Engineering, Civil Engineering Faculty, Istanbul Technical University,

Hydraulics Division, Maslak, 34469 Istanbul, Turkey

Received 19 September 2003; accepted 23 March 2004

Abstract

Use of wind speed data is of great importance in civil engineering, especially in structural

and coastal engineering applications. Synthetic data generation techniques are used in prac-

tice for cases where long wind speed data are required. In this study, a new wind speed data

generation scheme based upon wavelet transformation is introduced and compared to the

existing wind speed generation methods namely normal and Weibull distributed independent

random numbers, the ﬁrst- and second-order autoregressive models, and the ﬁrst-order Mar-

kov chain. Results propose the wavelet-based approach as a wind speed data generation

scheme to alternate the existing methods.

# 2004 Elsevier Ltd. All rights reserved.

Keywords: Normal distribution; Weibull distribution; Autoregressive models; Markov chain; Wavelet;

Hourly mean wind speed

1. Introduction and existing literature

Climatology is deﬁned as a set of probabilistic statements on long-term weather

conditions [1], and wind climatology as that branch of climatology that specialises

in the study of winds, from which information on extreme winds is provided to

structural designers. Such information is also needed for wind energy producers

and engineers who design coastal civil structures, for example breakwaters. From a

structural engineering point of view, forecasting the maximum wind speed that is

Ã

Corresponding author. Tel.: +90-212-2856577; fax: +90-212-2856587.

E-mail address: haksoy@itu.edu.tr (H. Aksoy).

0960-1481/$ - see front matter #2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.renene.2004.03.011

expected to aﬀect a structure during its lifetime is important to the designer. On the

other hand, in coastal engineering practices, not only the magnitude but also the

directionality of wind becomes important. The duration of wind, in addition to its

magnitude and direction, is also required in wind energy production systems, and

the amount of energy that can be produced depends upon it.

The information required by either structural and coastal engineers or wind

energy producers is related to wind speed data, and is a matter of quality and quan-

tity. The quality of the wind speed data refers to whether the data set is reliable and

micrometeorologically homogeneous. A data set is reliable if (i) the measurement

instrument performs adequately, (ii) the instrument is not inﬂuenced by obstructions

and (iii) the atmospheric stratiﬁcation is neutral. A set of wind speed data is con-

sidered micrometeorologically homogeneous if the data set is obtained under ident-

ical micrometeorological conditions [1]. The size of the data set (quantity) is related

to the time period during which the wind speed data are recorded. The time period

over which wind speed data are recorded is usually shorter than the lifetime of civil

engineering structures. Therefore, the worst case of wind load that the structural

designer expects that the structure will face during its lifetime is determined by mod-

elling the wind speed data record in hand. For this, climatological and physical

modelling techniques are available. Additionally, probabilistic and stochastic models

have been developed, for which the existing literature is reviewed in brief below. The

main aim in those techniques is to determine minimum design loads due to wind [2].

Short records of daily, weekly, and monthly highest wind speeds taken at 36

weather stations in the US were empirically analyzed [3] in order to determine

design wind speeds. Short records of hourly mean wind speed data from normal

regions in the US were used by Cheng and Chiu [4] for determination of the tran-

sition probabilities of the Markov chain upon which the methodology in that study

was based. This methodology was extended later to tropical cyclone-prone regions

[5]. Also, a knowledge-based expert system, principally similar to the mentioned

methodologies, was made available [6,7]. Alternative approaches used in the gener-

ation of simulated wind speed time series were compared by Kaminsky et al. [8].

Sfetsos [9] examined adaptive neuro-fuzzy inference systems and neural logic net-

works and compared them to the traditional autoregressive moving average

(ARMA) models. Dukes and Palutikof [10] employed the Markov chain in order to

estimate hourly mean wind speed with very long return periods. Another Markov

chain based study was conducted by Sahin and Sen [11]. Castino et al. [12] coupled

autoregressive processes to the Markov chain and simulated both wind speed and

direction. A recent study [13] presents a wavelet-based method to generate artiﬁcial

wind data. The Weibull distribution has commonly been ﬁtted to hourly mean

wind speed data [14,15]. The peaks-over-threshold approach has also been com-

monly used in the estimation of extreme quantiles of wind speed data [16–19].

2. Methods

In this study, a number of probabilistic and stochastic methods are used in order

to compare their ability to reproduce long series of hourly mean wind speed data

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2112

with the same statistical behaviour as that of the observations in hand. The normal

and Weibull probability distribution functions are chosen in order to generate

independent and identically distributed random numbers. Autoregressive processes

are useful tools in generating data sets in cases where persistency exists. Persistency

means that large values tend to be followed by large values, and small values by

small values, so that runs of values of similar magnitude tend to persist throughout

the sequence. First- and second-order autoregressive processes are chosen in this

study. Another concept commonly employed in wind speed data generation studies

is the ﬁrst-order Markov chain. The results of these methods are compared to

those obtained from a newly developed wavelet-based approach.

The methods are described below. Only the wavelet-based approach will be

detailed, whereas the remaining ﬁve methods will be outlined brieﬂy as they have

been well documented in literature.

2.1. Normal distribution

Hourly mean wind speed time series are generated by using a sequence of inde-

pendent random numbers from the normal distribution. The normal probability

distribution function is given by

f ðwÞ ¼

1

r

ﬃﬃﬃﬃﬃﬃ

2p

p exp½Àðw ÀlÞ

2

=2r

2

ð1Þ

where w is the variable (hourly mean wind speed, in this study), l mean value of

wind speed, and r standard deviation of wind speed. A number of computational

methods are available for the generation of random numbers with normal prob-

ability distribution of mean l and standard deviation r.

2.2. Weibull distribution

The Weibull distribution is another probability distribution function commonly

used for the frequency analysis of wind speed data [14,15]. It is given by

f ðwÞ ¼

a

b

a

w

aÀ1

exp À

1

b

a

w

a

w ! 0; a; b > 0 ð2Þ

where a and b are shape and scale parameters, respectively, that can be determined

by using either a graphical method or the method of moments. They can also be

determined using the method of probability weighted moments (PWMs) for which

explicit equations are available. It is the method used in this study for the determi-

nation of parameters.

Equations to be used for this purpose are given by

a ¼

lnð2Þ

L

2;ðln wÞ

ð3aÞ

b ¼ exp L

1;ðln wÞ

þ

0:5772

a

¸

ð3bÞ

2113 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

In Eqs. (3a,b), L

1,(ln w)

and L

2,(ln w)

are L

1

and L

2

moments of the logarithm of

the hourly mean wind speed time series. The L

1

and L

2

moments of a series are

given by

L

1

¼ b

0

ð4aÞ

L

2

¼ 2b

1

Àb

0

ð4bÞ

in which b

0

and b

1

are given by

b

0

¼ xx ð5aÞ

b

1

¼

¸

NÀ1

j¼1

ðN ÀjÞ

NðN À1Þ

x

j

ð5bÞ

x

j

in Eq. (5b) comes from the time series sorted in descending order as

x

N

Á Á Á x

i

Á Á Á x

1

. Detailed information on L moments and the method of

PWM is given in [20].

Once the parameters are determined, the generation of Weibull distributed ran-

dom numbers is a matter of a simple computer code, as the cumulative distribution

function of the Weibull distribution can be obtained in closed form.

2.3. AR(1) model

The hourly mean wind speed time series is of high dependence. This property

particularly requires a wind speed data generation model incorporating the depen-

dence structure of the observations. As mentioned, both normal and Weibull dis-

tributed random numbers do not take this property into account as they are

independent, but autoregressive models are of correlated type and hence capable of

simulating this property of the data series.

The use of autoregressive type models is reported in literature very commonly.

The ﬁrst-order autoregressive [AR(1)] model accommodates only the eﬀect of the

previous value in the series in which the observed sequence of wind speed data {w

1

,

w

2

,. . ., w

t

,. . .} is used to ﬁt a model of form

w

i

¼

¸

m

j¼1

a

j

w

iÀj

þe

i

ð6Þ

where w is the hourly mean wind speed, a the autoregressive coeﬃcient, that is,

model parameter, and e a normally distributed independent random variable. It is

noted that Eq. (6) is written for the mth order. The simplest case of Eq. (6) is

obtained for m ¼ 1, which is also called the Markov model. Eq. (6) then becomes

y

i

¼ r

1

y

iÀ1

þe

i

ð7Þ

where y is the standardised (zero mean and unit variance) version of the variable

and r

1

the lag-one serial correlation coeﬃcient of the sequence.

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2114

The random component (e) in AR(1) is of normal distribution with zero mean

and a variance of 1 À r

2

1

. The simulation procedure for the processes is very sim-

ple. It requires only a random number of a normal distribution to be generated.

2.4. AR(2) model

With increase in order of the autoregressive model, the dependence structure in

the observations is better preserved. Therefore, the second-order autoregressive

[AR(2)] model is preferred to AR(1). This becomes more important in cases where

dependence in the data set is very obvious, as in the hourly mean wind speed data.

AR(2) is formulized as

y

i

¼ /

1

y

iÀ1

þ/

2

y

iÀ2

þe

i

ð8Þ

where autoregressive coeﬃcients /

1

and /

2

are given by

/

1

¼ r

1

ð1 Àr

2

Þ=ð1 Àr

2

1

Þ ð9aÞ

/

2

¼ ðr

2

Àr

2

1

Þ=ð1 Àr

2

1

Þ ð9bÞ

in which r

1

is the lag-one autocorrelation coeﬃcient and r

2

the lag-two autocorrela-

tion coeﬃcient of the wind speed time series. The random component in AR(2) is

again of normal distribution, with zero mean and variance equal to 1ÀR

2

, where

R

2

¼

r

2

1

þr

2

2

À2r

1

r

2

1 Àr

2

1

ð10Þ

2.5. Markov chain

In this approach, the observed time series is divided into a number of states. A

wind speed state contains wind speeds between certain values. For example, State 1

might include wind speeds below 2 m/s, State 2 wind speeds between 2 and 4 m/s,

etc. until the ﬁnal wind speed state includes all speeds above the highest observed

value or a predeﬁned upper limit. The upper and lower limits of the states are

highly subjective values. For instance, the hourly mean wind speed data set in this

study was divided into 10 states. In another wind speed study [11], states were

deﬁned depending upon the standard deviation of the data set. Each state in that

study [11] was taken as wide as one standard deviation of the observed hourly

mean wind speed time series. Dukes and Palutikof [10], on the other hand, used a

ﬁxed width for the states, which was equal to 2 m/s.

In the Markov chain approach, the state of wind speed in the current hour can

be deﬁned depending only upon the previous state. This is called the ﬁrst-order or

one-step Markov chain. Two previous states are used in the second-order or two-

step Markov chain in determining the current state of the wind speed. Although

they are not common as the ﬁrst- and second-order Markov chains, higher-order

Markov chains can also be used. However, dramatic increase in the number of

their parameters limits their use.

2115 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

The parameter set of a Markov chain consists of probabilities of transition from

one state to another that are given in transition probability matrices. The tran-

sition probability matrix of a ﬁrst-order Markov chain with m states can be written

symbolically as

P ¼

P

11

P

12

. . . P

1m

P

21

P

22

. . . P

2m

. . . . . . P

ij

. . .

P

m1

P

m2

. . . P

mm

¸

¸

¸

¸

ð11Þ

where P

ij

is the probability of transition from state i to state j. The number of

parameters is mðmÀ1Þ, as the sum of the probabilities is equal to 1 (100%) for

each row of the matrix. If n

ij

is the total number of hours of observation in state j

with the previous state i, the probabilities of transition from state i to state j can be

calculated as

P

ij

¼

n

ij

¸

j

n

ij

i; j ¼ 1; 2; . . . ; m ð12Þ

The procedure for generating the simulated hourly mean wind speed time series

is explained below.

First, the cumulative transition probability matrix is calculated. In the cumulative

transition probability matrix, cumulative summation of probabilities within each row

is carried out; hence, each row in that matrix ends with 1. Then, an initial state is

adopted. No wind (State 1) can, for example, be assumed as the initial state. Using a

uniform random number, the next state of wind speed can be determined. If State 1 is

obtained as the new state of wind speed, then it is ﬁrst checked if the wind speed is

zero. If the wind speed is not zero, then a uniform random number is generated from

the interval of State 1. If the highest state is found to be the new state of wind speed,

then a shifted one-parameter gamma distributed random number is used in order to

ﬁnd the magnitude of the wind speed. The reason for choosing the gamma distribution

will be discussed in the section where results obtained from application of the methods

are presented. For intermediate states, a uniform random number from the interval of

the corresponding state is generated and set as the wind speed at the current hour.

2.6. Wavelet-based approach

A real or complex-value continuous function with zero mean and ﬁnite variance

is called a wavelet [21]. There are many functions that can qualify as wavelets.

Some examples of wavelets are Morlet, Mexican hat, Shannon and Meyer. A sim-

ple wavelet is the Haar wavelet (Fig. 1), deﬁned as

wðtÞ ¼

1 0 t 1=2

À1 1=2 t 1

0 otherwise

ð13Þ

Decomposing a signal and then reconstructing it is the base for the wavelet

transform. In this study, the Haar wavelet was used due to its simplicity. There-

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2116

fore, decomposition of a signal (multiresolution analysis) with the Haar wavelet is

considered and explained in detail below.

For a certain value of k, let us deﬁne f

k

(t) as the average of f(s) over an interval

of size 2

k

:

f

k

ðtÞ ¼

1

2

k

2

k

ðlþ1Þ

2

k

l

f ðsÞds 2

k

l < t < 2

k

ðl þ1Þ ð14Þ

where k and l are integers, k a scale variable (k > 0 means stretching and k < 0

means contracting of the wavelet) and l a translation variable [21]. For

k ¼ À1,. . .,À1, 0, 1,. . ., 1, f

k

(t) is as follows:

f

À1

ðtÞ ¼ f ðtÞ

.

.

.

f

À1

ðtÞ ¼ 2

ðlþ1Þ=2

l=2

f ðsÞds

l

2

< t <

l þ1

2

f

0

ðtÞ ¼

ðlþ1Þ

l

f ðsÞds l < t < ðl þ1Þ

f

1

ðtÞ ¼

1

2

2ðlþ1Þ

2l

f ðsÞds 2l < t < 2ðl þ1Þ

.

.

.

f

1

ðtÞ ¼ 0

ð15Þ

Fig. 1. Haar wavelet.

2117 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

The resolution decreases as k increases. The diﬀerence between the successive

averages f

kÀ1

(t) and f

k

(t) is deﬁned as a detail function:

g

k

ðtÞ ¼ f

kÀ1

ðtÞ Àf

k

ðtÞ ð16Þ

It can be easily seen that

f ðtÞ ¼

¸

1

k¼À1

g

k

ðtÞ ð17Þ

According to Eq. (17), the original signal is obtained when all detail functions

are summed up. Change in data resolution with change in k, the resolution level,

can be seen in the upper part of Fig. 2, in which the average of the time series

taken at diﬀerent resolution levels according to Eq. (15) is shown. Note that the

data sample used in Fig. 2 has 16 elements. Increase in the ordinates of f

k

(t) with

decrease in k shows the change (increase) in the resolution. The middle part of

Fig. 2 shows the detail functions calculated using Eq. (16) for diﬀerent resolution

levels. Note from Eq. (15) that f

4

ðtÞ ¼ 0 for all t. At the bottom of Fig. 2, f(t), the

sum of the four detail functions according to Eq. (17), is seen, and it represents the

original data, f

0

(t). Eq. (17) is the basis for the generation algorithm explained

below.

Let us consider a data sample of size M ¼ 2

K

, where K is a positive integer

(K ¼ 4 for the sequence in Fig. 2) taken from a stochastic process f(t) with zero

mean: f(1), f(2),. . ., f(M). Deﬁne the sample f

k

(i) (k ¼ 0, 1,. . ., K; i ¼ 1,. . ., M) con-

sisting of averages of 2

k

successive elements of the sample. f

0

(i) is the original sam-

ple and f

K

(i) is a sample of all zeros, since the average of M elements is zero. The

detail function g

k

(t) has a sample consisting of M elements given by Eq. (16) for

k ¼ 1, 2,. . ., K.

Thus, for each element f

i

of the original sample, we have K detail function

values, g

k

(i), corresponding to diﬀerent resolutions. Choosing from M elements for

each g

k

(t) randomly, and then summing them up using Eq. (17), one obtains a

simulated value for f(t) as

f ðjÞ ¼

¸

K

k¼1

g

k

ðjÞ ð18Þ

where j is the index for generated elements.

The generation algorithm is given step by step as follows [22] and is illustrated in

Fig. 3 for K ¼ 4.

1. In order to obtain the ﬁrst element of the series ( j ¼ 1), g

k

values (k ¼ 0, 1,. . .,

K) are chosen from M values randomly and summed up to obtain f

1

(Fig. 3).

2. The second element ( j ¼ 2) is generated by choosing, for each k, the g

k

coming

just after the g

k

values chosen in the ﬁrst step. f

2

is obtained by the summation

of these (Fig. 3).

3. Data generation is continued in this way for a desired number of times using,

for the generation of each element f

j

, the detail function values right next to

those of the previous step jÀ1 at each resolution level.

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2118

Fig. 2. Decomposition and reconstruction of a data sequence.

2119 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

This generation algorithm is a newly developed approach for data simulation

purposes. It was ﬁrst used in non-skewed annual and monthly streamﬂow data

simulation studies [22,23]. The approach was later used for the simulation of the

storage capacity of river reservoirs [24]. Modelling suspended sediment discharge

series [25] and annual and monthly rainfall data series [26] was also performed by

this approach successfully. The algorithm generated the mean, standard deviation

and correlation structure of the observed streamﬂow data sets. When one is inter-

ested in the generation of skewed data, it is ﬁrst required to transform the data to

a non-skewed structure, generate them and then transform them back to their

skewed structure.

3. Application

The methods were applied to an hourly mean wind speed data set that will be

introduced in the following subsections. Results obtained from the application of

the methods are presented and discussed below. The performance of the methods

Fig. 3. Construction of a simulated data sequence.

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2120

was measured according to their ability to capture the statistical behaviour of the

observed data set. A comparison of the methods is ﬁnally presented.

3.1. Data

Table 1 shows the main statistical characteristics of the data set of hourly mean

wind speed taken from the State Meteorological Works’ meteorology station in

Diyarbakir, a southeastern Anatolian city. The data set is of four years’ length,

from 1994 to 1997 (35064 hours in total). The region is normal, as is seen in

Table 1. The data set is highly correlated, as expected, and skewed. For the wave-

let-based approach, 32768 hours of data, extending from the ﬁrst hour of April 6,

1994, to the eighth hour of December 31, 1997, were used. This is a choice with no

speciﬁc reason. Characteristics corresponding to that part of the observed series are

also given in Table 1.

3.2. Parameters

The hourly mean wind speed data set used in the study is of skewed structure.

This prevents ﬁtting of the normal distribution to the data. Therefore, power trans-

formation [ y ¼ x

h

; where x the is raw (untransformed) variable, y the transformed

variable, and h the transformation coeﬃcient] was adopted in order to obtain non-

skewed data, to which the normal distribution can be ﬁtted. The transformation

coeﬃcient was obtained as h ¼ 0:38585 for the data set in the study. As the normal

distribution is ﬁtted to the transformed hourly mean wind speed time series (but

not to the raw data series), the parameters of the normal distribution are the mean

and standard deviation of the transformed hourly mean wind speed time series.

Those parameters are presented in Table 2. The normal probability distribution

function based upon the determined parameters was ﬁtted to the transformed wind

speed data series (Fig. 4). It is seen that the distribution performs very well in ﬁt-

ting to the observations as well as to the generated data, to be explained later in

following sections.

The Weibull distribution has two parameters (a, the shape parameter, and b, the

scale parameter). The parameters were determined using the method of L-moments

on which detailed information was given previously. The reason for choosing this

method is that explicit equations are available for determination of the parameters

of the distribution. The method also has the superiority of being less sensitive to

outliers, which means that outliers do not aﬀect the performance of the method in

determining the parameters correctly. The only problem with this method is the

presence of zero wind speeds, which makes the method inapplicable due to the log-

arithm included. In order to overcome this problem, zero wind speeds were ignored

from the observed time series as their number of occurrences was very small, less

than 0.5%. The parameters of the Weibull distribution determined by the method

of L-moments are listed in Table 2. Fig. 5 shows the agreement between the

observed data and the ﬁtted Weibull probability distribution function. It can be

considered a very good ﬁt, although the Weibull probability distribution function

2121 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

T

a

b

l

e

1

S

t

a

t

i

s

t

i

c

a

l

c

h

a

r

a

c

t

e

r

i

s

t

i

c

s

o

f

o

b

s

e

r

v

e

d

h

o

u

r

l

y

m

e

a

n

w

i

n

d

s

p

e

e

d

t

i

m

e

s

e

r

i

e

s

D

a

t

e

N

u

m

b

e

r

o

f

d

a

t

a

M

e

a

n

(

m

/

s

)

S

t

a

n

d

a

r

d

d

e

v

i

a

t

i

o

n

(

m

/

s

)

C

o

e

ﬃ

c

i

e

n

t

o

f

v

a

r

i

a

t

i

o

n

C

o

e

ﬃ

c

i

e

n

t

o

f

s

k

e

w

n

e

s

s

M

a

x

i

m

u

m

w

i

n

d

s

p

e

e

d

(

m

/

s

)

C

o

r

r

e

l

a

t

i

o

n

c

o

e

ﬃ

c

i

e

n

t

r

1

r

2

r

3

r

4

r

5

1

J

a

n

u

a

r

y

1

9

9

4

–

3

1

D

e

c

e

m

b

e

r

1

9

9

7

3

5

0

6

4

2

.

5

3

8

1

.

7

8

6

0

.

7

0

3

1

.

2

8

5

1

4

.

4

0

.

8

6

0

0

.

7

3

2

0

.

6

3

3

0

.

5

4

9

0

.

4

7

6

6

A

p

r

i

l

1

9

9

4

–

3

1

D

e

c

e

m

b

e

r

1

9

9

7

3

2

7

6

8

2

.

5

5

5

1

.

7

9

4

0

.

7

0

2

1

.

2

8

3

1

4

.

4

0

.

8

6

1

0

.

7

3

3

0

.

6

3

5

0

.

5

5

1

0

.

4

3

8

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2122

gives the mode an occurrence probability slightly lower than that in the obser-

vation.

AR(1) is a parametric model with two parameters (a, the autoregression coef-

ﬁcient, and r

2

e

, the variance of the independent normal variable). The model

requires only the lag-one serial correlation coeﬃcient (r

1

), as both parameters are

dependent only upon r

1

.

AR(2) has three parameters (/

1

and /

2

, the autoregression coeﬃcients, and r

2

e

,

the variance of the independent normal variable), all functions of r

1

and r

2

, the lag-

one and lag-two serial correlation coeﬃcients listed in Table 2.

Of the six methods, the Markov chain is the one that requires the highest num-

ber of parameters. The number of parameters required changes with the number of

states used for the wind speed. In this study, 10 states were chosen for the wind

speed, each 1.5 m/s wide. This resulted in 90 transition probabilities to be determ-

ined from the observed wind speed data set, when it is considered that summation

Table 2

Parameter sets of methods

Method Parameter set

Normal

l ¼ 1:347 m=s r ¼ 0:392 m=s

Weibull

a ¼ 1:583 b ¼ 1:973

AR(1), AR(2)

r

1

¼ 0:820 r

2

¼ 0:688

Fig. 4. Normal probability distribution function ﬁtted to the observed and simulated random wind

speed sequences.

2123 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

over any row in the transition probability matrix results in 100% probability. The

transition probability matrix of the data set is given in Table 3. Not only transition

probabilities, but also the wind speed distribution in each state should be known

by this method. In this study, wind speed was assumed to be distributed uniformly

over the states except for the last one (state of highest wind speeds with no upper

limit), where the one-parameter gamma distribution was used. In State 1 with the

lower limit of zero, the probability of occurrence of zero wind speed was also taken

Fig. 5. Weibull probability distribution function ﬁtted to the observed and simulated random wind

speed sequences.

Table 3

Transition probability matrix of the observed hourly mean wind speed data set

P

ij

j ¼ 1 2 3 4 5 6 7 8 9 10

i ¼ 1

0.7053 0.2779 0.0144 0.0015 0.0008 0.0001 0.0000 0.0000 0.0000 0.0000

2

0.2405 0.6089 0.1306 0.0153 0.0041 0.0005 0.0000 0.0001 0.0000 0.0000

3

0.0256 0.2839 0.5317 0.1352 0.0178 0.0048 0.0005 0.0003 0.0002 0.0000

4

0.0042 0.0491 0.3116 0.4954 0.1191 0.0179 0.0023 0.0003 0.0000 0.0000

5

0.0008 0.0176 0.0865 0.3486 0.4311 0.0978 0.0168 0.0008 0.0000 0.0000

6

0.0000 0.0089 0.0266 0.1197 0.3437 0.4013 0.0865 0.0111 0.0022 0.0000

7

0.0000 0.0152 0.0076 0.0455 0.1212 0.3561 0.3485 0.1061 0.0000 0.0000

8

0.0000 0.0000 0.0000 0.0526 0.0000 0.2105 0.3684 0.2632 0.0789 0.0263

9

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1818 0.3636 0.3636 0.0909

10

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2124

into consideration in order to reproduce the zero wind speeds, although their

occurrence was very low.

There is no parameter to be listed for the wavelet approach, as it is a nonpara-

metric method. The length of the series to be used in this method is equal to 2

K

,

where K is a positive integer and equal to 15 in this study. This corresponds to a

series 32768 hours in length. Another requirement for the wavelet approach is that

the data set should be of a non-skewed structure. Therefore, the part of the

observed series used for the wavelet approach was transformed by using the power

transformation with h ¼ 0:3853.

3.3. Simulation and results

A thousand-year (8760000-hour) -long series was generated for each method.

The correlogram, frequency distribution of maximum wind speeds and wind dur-

ation curve obtained from the simulations will be compared to those of the

observed series.

It is obvious that the hourly mean wind speed time series has a highly dependent

structure. The normal and Weibull distributions, however, are of independent

structures (Fig. 6), yet they are very common methods used in generating wind

speed data. These methods may be useful in oﬀering, to the structural designer, the

highest wind speed that the structure will possibly face during its lifetime.

It is seen from Fig. 4 that wind speed data generated by the normal distribution

ﬁt the observed series very well. It is seen in Fig. 5 that the Weibull ﬁt is perfect as

well.

Other than those two methods, the AR(1), AR(2), and Markov chain methods

looked to produce the dependence structure of the series. However, with increasing

lags in time, the success of those methods in reproducing the correlation structure

Fig. 6. Correlogram of the observed and simulated wind speeds.

2125 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

of the series decreases (Fig. 6). The wavelet method of the six studied, was found

to be the best in preserving the correlation structure of the series.

The annual maximum values of the simulation series were compared in Fig. 7. It

is seen that the normal distribution, AR(1) and AR(2) produced similar maxima,

whereas the wavelet approach produced higher, the Markov chain slightly lower

and the Weibull distribution considerably lower maxima. From the structural

engineering point of view, therefore, it is safer to use the wind load due to the

maximum wind speed generated by the wavelet-based approach.

Maxima obtained from the Markov chain method should be discussed speciﬁ-

cally. There are three vertical jumps (one of them is very obvious) in the cumulat-

ive frequency diagram of the maxima of this method, as is seen in Fig. 7. The

reason for those jumps can be explained very simply. It is seen from Table 3 that

the probabilities of transition of the wind speed to the highest states are too low,

making transition of wind speed to those states almost impossible in the simulation

series. It is only possible to make a transition to State 10 if the previous state in the

simulation series is either State 8 or State 9. Otherwise State 10 is not simulated.

This causes the maximum value of the series to be bounded by the upper limit of

State 9, which was taken as 13.5 m/s in this study. A very small jump exists in the

frequency curve in Fig. 7 due to this circumstance. Similarly, State 9 can be simu-

lated if and only if the previous state of the wind speed is one of the following

states: 3, 6, 8, 9 and 10 (Table 3). The big jump in the cumulative frequency curve

in Fig. 7 is due to this situation. It is seen that maximums of the simulated series

are bounded by the upper limit of State 8, which was taken as 12 m/s in this study.

Fig. 7. Cumulative frequency diagram of maximum wind speeds.

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2126

The third jump at the very beginning of the curve (close to the y-axis of the graph)

is due to a similar situation. This is the result of not having simulated wind speeds

from State 8, which limits the maximum wind speed to 10.5 m/s at the upper limit

of State 7. This drawback of the method can be overcome by forcing the simula-

tion series to have at least one value from the highest state that results in a

maximum wind speed data series all generated from the highest state with no upper

limit. Such a forcing can be considered quite reasonable and it does not aﬀect the

transition probability matrix as the number of data is usually very large (of the

order of tens of thousands).

Uniformly distributed wind speeds were accepted for the intermediate states,

whereas the one-parameter gamma distribution was adopted for the highest state

(State 10 in this study). The reason for choosing this distribution is explained

below together with a discussion on other distributions.

A distribution with no upper limit should be used for the highest state so that

maximum wind speeds higher than those in the observed series can possibly be

generated. Therefore, in this study, it was ﬁrst thought to simulate wind speeds of

the highest state by using the exponential distribution shifted to that state as Sahin

and Sen [11] did. This is quite a reasonable choice for simulating the wind speeds

in that state. However, it was seen that the exponential distribution generated

lower maximum wind speeds compared to those generated by other methods.

Therefore, the Gumbel distribution was tested. It was seen that the maximum wind

speeds generated by this distribution were too low compared to those obtained by

the other methods. The Frechet distribution, which is accepted as the distribution

of the maximum wind speeds [1], was also found to be unsuccessful in generating

maximum wind speeds compared to other methods. The distribution generated low

maximum wind speeds. In the end, the two-parameter gamma distribution was ﬁt-

ted, which resulted again in low maximum wind speeds. Finally, the one-parameter

gamma distribution was ﬁtted and results comparable to those of the other meth-

ods (in Fig. 7) were obtained.

The conclusion that can be drawn from those trials is that a one-parameter dis-

tribution can ﬁt to the highest state better than distributions with two or more

parameters. If the standard deviation of the highest state, which is bounded by the

lower and upper limits of the state, is included in the generation scheme, then

lower maximum wind speeds are generated. Therefore, mean-dependent probability

distribution functions are better in the simulation of maximum wind speeds.

The transition probability matrix of the Markov chain based simulation wind

speed series is given in Table 4. It is almost the same as its observed counterpart

given in Table 3, which means that the Markov chain based simulation technique

worked very well in the simulation of the state of the wind speed series.

The wind duration curve is a graph with time percentage as abscissa and wind

speed as ordinate (Fig. 8). It is an important tool used in determining the percent-

age of time that the wind speed exceeds a speciﬁed level. Wind energy production

systems use this graph in order to determine the wind energy potential of the

region under consideration. A very good ﬁt was obtained in Fig. 8, where the wind

duration curves of the six methods were plotted together with the one extracted

2127 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

from the observed series. Although the wind duration curve of the Markov chain

method ﬂuctuates around the others, it has a ﬁt that is good enough as well.

The ﬁrst three central moments of the observed and simulated series are given in

Table 5. The maximum values and the ﬁrst ﬁve lags of the correlation are also lis-

ted. It is seen that the mean values of the simulated series are almost the same as

those of the observations. The wavelet-based method approaches its counterparts

with a relative error of 0.3%. Standard deviation and variation coeﬃcient were

best captured by the normal probability distribution, and AR(1) and AR(2) pro-

cesses. Skewness coeﬃcient in the wind speed time series was best reproduced by

Table 4

Transition probability matrix of hourly mean wind speed data simulated by Markov chain method

P

ij

j ¼ 1 2 3 4 5 6 7 8 9 10

i ¼ 1

0.7049 0.2782 0.0143 0.0016 0.0009 0.0001 0.0000 0.0000 0.0000 0.0000

2

0.2406 0.6086 0.1309 0.0153 0.0041 0.0005 0.0000 0.0001 0.0000 0.0000

3

0.0255 0.2845 0.5318 0.1346 0.0179 0.0048 0.0005 0.0003 0.0002 0.0000

4

0.0042 0.0491 0.3115 0.4958 0.1191 0.0177 0.0023 0.0003 0.0000 0.0000

5

0.0008 0.0181 0.0860 0.3474 0.4322 0.0982 0.0165 0.0008 0.0000 0.0000

6

0.0001 0.0089 0.0259 0.1206 0.3421 0.4028 0.0861 0.0113 0.0022 0.0000

7

0.0001 0.0156 0.0077 0.0478 0.1227 0.3526 0.3491 0.1045 0.0000 0.0000

8

0.0000 0.0000 0.0000 0.0501 0.0000 0.2197 0.3667 0.2606 0.0777 0.0252

9

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1888 0.3619 0.3720 0.0773

10

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000 0.0000

Fig. 8. Wind duration curve of observed and simulated wind speeds.

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2128

T

a

b

l

e

5

S

t

a

t

i

s

t

i

c

a

l

c

h

a

r

a

c

t

e

r

i

s

t

i

c

s

o

f

s

i

m

u

l

a

t

e

d

s

e

r

i

e

s

S

e

r

i

e

s

M

e

a

n

(

m

/

s

)

S

t

a

n

d

a

r

d

d

e

v

i

a

t

i

o

n

(

m

/

s

)

C

o

e

ﬃ

c

i

e

n

t

o

f

v

a

r

i

a

t

i

o

n

C

o

e

ﬃ

c

i

e

n

t

o

f

s

k

e

w

n

e

s

s

M

a

x

i

m

u

m

w

i

n

d

s

p

e

e

d

(

m

/

s

)

C

o

r

r

e

l

a

t

i

o

n

c

o

e

ﬃ

c

i

e

n

t

r

1

r

2

r

3

r

4

r

5

N

o

r

m

a

l

2

.

5

3

8

1

.

7

7

7

0

.

7

0

0

1

.

3

1

5

2

6

.

9

4

0

.

0

0

0

5

À

0

.

0

0

0

2

0

.

0

0

0

5

0

.

0

0

0

2

À

0

.

0

0

0

2

W

e

i

b

u

l

l

2

.

5

2

9

1

.

6

3

4

0

.

6

4

6

0

.

9

8

0

1

6

.

2

2

À

0

.

0

0

0

3

0

.

0

0

0

5

0

.

0

0

0

3

À

0

.

0

0

0

1

À

0

.

0

0

0

7

A

R

(

1

)

2

.

5

3

7

1

.

7

7

6

0

.

7

0

0

1

.

3

0

9

2

7

.

3

0

0

.

8

1

5

0

.

6

6

1

0

.

5

3

6

0

.

4

3

6

0

.

3

5

5

A

R

(

2

)

2

.

5

3

7

1

.

7

7

6

0

.

7

0

0

1

.

3

1

3

2

5

.

2

3

0

.

8

1

5

0

.

6

7

7

0

.

5

6

1

0

.

4

6

6

0

.

3

8

7

M

a

r

k

o

v

2

.

5

8

5

2

.

0

5

8

0

.

7

9

6

0

.

9

8

3

2

1

.

2

9

0

.

7

1

5

0

.

5

8

7

0

.

4

8

3

0

.

3

9

9

0

.

3

3

0

W

a

v

e

l

e

t

2

.

5

6

6

1

.

8

3

3

0

.

7

1

4

1

.

4

3

8

3

1

.

1

3

0

.

7

1

5

0

.

5

8

0

0

.

5

2

3

0

.

4

4

3

0

.

4

2

1

2129 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

AR(1). Higher maximums were obtained by the methods of AR(1), wavelet and

normal distribution and lower maximums by Weibull distribution. Correlation

structure, as discussed earlier, was best simulated by the wavelet-based method.

4. Summary and conclusion

In this study, hourly mean wind speed data sets were generated by traditional

simulation methods—the normal and Weibull probability distribution functions,

the ﬁrst- and second-order autoregressive processes, and the Markov chain.

Additionally, the newly developed wavelet-based approach was used. The normal

and Weibull probability distribution functions consist of independent identically

distributed random numbers. The autoregressive models include the correlation

structure of the observation and hence generate dependent series. The Markov

chain is a two-step method that ﬁrst determines the state of the wind speed and

then generates its magnitude by using a preselected distribution. All the mentioned

methods are parametric and they therefore require the time series to have a speciﬁc

probability distribution. This is a drawback of parametric models more than a

limitation. A nonparametric model, of which the wavelet approach in this study is

one of the best examples, can be applied to data sets with any distribution. How-

ever, it should be kept in mind that the wavelet approach works only with sequen-

ces of zero skewness.

The correlation structure of the observations, distribution of the maximum wind

speeds, wind duration curve and statistical features of the series were used in order

to compare the success of the methods.

The generation of maximum wind speeds requires special attention in Markov

chain based simulation methods. Based upon the application in this study, it is

concluded that the uniform probability distribution function is suitable for use in

the ﬁrst and intermediate states. A probability distribution function with no upper

limit should be used for the highest state. It is concluded that the one-parameter

gamma distribution is good enough in ﬁtting to the wind speed data in the highest

state of the series for normal regions, such as the one used in this study.

Some methods performed better in preserving some particular characteristics

than other methods did. For example, the wavelet method is obviously the best in

preserving the correlation structure of the sequence. This method is as good at pre-

serving other statistical features of the series as other methods. Therefore, in con-

clusion, the wavelet method is proposed as a tool to substitute for the classical

generation schemes for the simulation of hourly mean wind speed data.

Acknowledgements

The wavelet approach presented in this study is a result of an earlier cooperation

between the ﬁrst author (H. Aksoy) and Professor M. Bayazit of Istanbul Techni-

cal University, Turkey, whom the authors sincerely thank.

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131 2130

References

[1] Simiu E, Scanlan RH. Wind eﬀects on structures. New York: John Wiley & Sons; 1986.

[2] American Society of Civil Engineers. Minimum design loads for buildings and other structures.

ANSI/ASCE 7-93 (Revision of ANSI/ASCE 7-88), New York, 1994.

[3] Simiu E, Filliben JJ, Shaver JR. Short-term records and extreme wind speeds. ASCE, Journal of

the Structural Division 1982;108(ST11):2571–7.

[4] Cheng EDH, Chiu ANL. Extreme winds simulated from short-period records. ASCE, Journal of

Structural Engineering 1985;111(1):77–94.

[5] Cheng EDH, Chiu ANL. Extreme winds generated from short records in a tropical cyclone-prone

region. Journal of Wind Engineering and Industrial Aerodynamics 1988;28:69–78.

[6] Cheng EDH. Wind data generator: a knowledge-based expert system. Journal of Wind Engineering

and Industrial Aerodynamics 1991;38:101–8.

[7] Cheng EDH, Chiu ANL. An expert system for extreme wind simulation. Journal of Wind Engin-

eering and Industrial Aerodynamics 1990;36:1235–43.

[8] Kaminsky FC, Kirchhoﬀ RH, Syu CY, Manwell JF. A comparison of alternative approaches for

the synthetic generation of a wind speed time series. Transactions of the ASME 1991;113:280–9.

[9] Sfetsos A. A comparison of various forecasting techniques applied to mean hourly wind speed time

series. Renewable Energy 2000;21:23–35.

[10] Dukes MDG, Palutikof JP. Estimation of extreme wind speeds with very long return periods. Jour-

nal of Applied Meteorology 1995;34:1950–61.

[11] Sahin AD, Sen Z. First-order Markov chain approach to wind speed modelling. Journal of Wind

Engineering and Industrial Aerodynamics 2001;89:263–9.

[12] Castino F, Festa R, Ratto CF. Stochastic modelling of wind velocities time series. Journal of Wind

Engineering and Industrial Aerodynamics 1998;74–76:141–51.

[13] Kitagawa T, Nomura T. A wavelet-based method to generate artiﬁcial wind ﬂuctuation data. Jour-

nal of Wind Engineering and Industrial Aerodynamics 2003;91:943–64.

[14] Garcia A, Torres JL, Prieto E, De Francisco A. Fitting wind speed distributions: a case study.

Solar Energy 1998;62(2):139–44.

[15] Grigoriu M. Estimates of design wind from short records. ASCE Journal of the Structural Division

1982;108(ST5):1034–48.

[16] Heckert NA, Simiu E, Whalen T. Estimates of hurricane wind speeds by ‘peaks over threshold’

method. ASCE Journal of Structural Engineering 1998;124(4):445–9.

[17] Lechner A, Simiu E, Heckert NA. Assessment of ‘peaks over threshold’ methods for estimating

extreme value distribution tails. Structural Safety 1993;12:305–14.

[18] Pandey MD, Van Gelder PHAJM, Vrijling JK. The estimation of extreme quantiles of wind velo-

city using L-moments in the peaks-over-threshold approach. Structural Safety 2001;23:179–92.

[19] Simiu E, Heckert NA. Extreme wind distribution tails: a ‘peaks over threshold’ approach. ASCE,

Journal of Structural Engineering 1996;122(5):539–47.

[20] Stedinger JR, Vogel RM, Foufoula-Georgiou E. Frequency analysis of extreme events. In: Maidment

D, editor. Handbook of hydrology. New York: McGraw Hill Book Co; 1993 [Chapter 18].

[21] Rao RM, Bopardikar AJ. Wavelet transforms, introduction to theory and applications. Reading,

MA: Addison-Wesley; 1998.

[22] Bayazit M, Aksoy H. Using wavelets for data generation. Journal of Applied Statistics 2001;28(2):

157–66.

[23] Bayazit M, Onoz B, Aksoy H. Nonparametric streamﬂow simulation by wavelet or Fourier analy-

sis. Hydrological Sciences Journal 2001;46(4):623–34.

[24] Aksoy H. Storage capacity for river reservoirs by wavelet-based generation of sequent peak algor-

ithm. Water Resources Management 2001;15(6):423–37.

[25] Aksoy H, Akar T, Unal NE. Wavelet analysis for modeling suspended sediment discharge. Nordic

Hydrology 2004;35:165–74.

[26] Unal NE, Aksoy H, Akar T. Annual and monthly rainfall data generation schemes. Stochastic

Environmental Research and Risk Assessment 2044;18(6):in press.

2131 H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

2112

H. Aksoy et al. / Renewable Energy 29 (2004) 2111–2131

expected to aﬀect a structure during its lifetime is important to the designer. On the other hand, in coastal engineering practices, not only the magnitude but also the directionality of wind becomes important. The duration of wind, in addition to its magnitude and direction, is also required in wind energy production systems, and the amount of energy that can be produced depends upon it. The information required by either structural and coastal engineers or wind energy producers is related to wind speed data, and is a matter of quality and quantity. The quality of the wind speed data refers to whether the data set is reliable and micrometeorologically homogeneous. A data set is reliable if (i) the measurement instrument performs adequately, (ii) the instrument is not inﬂuenced by obstructions and (iii) the atmospheric stratiﬁcation is neutral. A set of wind speed data is considered micrometeorologically homogeneous if the data set is obtained under identical micrometeorological conditions [1]. The size of the data set (quantity) is related to the time period during which the wind speed data are recorded. The time period over which wind speed data are recorded is usually shorter than the lifetime of civil engineering structures. Therefore, the worst case of wind load that the structural designer expects that the structure will face during its lifetime is determined by modelling the wind speed data record in hand. For this, climatological and physical modelling techniques are available. Additionally, probabilistic and stochastic models have been developed, for which the existing literature is reviewed in brief below. The main aim in those techniques is to determine minimum design loads due to wind [2]. Short records of daily, weekly, and monthly highest wind speeds taken at 36 weather stations in the US were empirically analyzed [3] in order to determine design wind speeds. Short records of hourly mean wind speed data from normal regions in the US were used by Cheng and Chiu [4] for determination of the transition probabilities of the Markov chain upon which the methodology in that study was based. This methodology was extended later to tropical cyclone-prone regions [5]. Also, a knowledge-based expert system, principally similar to the mentioned methodologies, was made available [6,7]. Alternative approaches used in the generation of simulated wind speed time series were compared by Kaminsky et al. [8]. Sfetsos [9] examined adaptive neuro-fuzzy inference systems and neural logic networks and compared them to the traditional autoregressive moving average (ARMA) models. Dukes and Palutikof [10] employed the Markov chain in order to estimate hourly mean wind speed with very long return periods. Another Markov chain based study was conducted by Sahin and Sen [11]. Castino et al. [12] coupled autoregressive processes to the Markov chain and simulated both wind speed and direction. A recent study [13] presents a wavelet-based method to generate artiﬁcial wind data. The Weibull distribution has commonly been ﬁtted to hourly mean wind speed data [14,15]. The peaks-over-threshold approach has also been commonly used in the estimation of extreme quantiles of wind speed data [16–19]. 2. Methods In this study, a number of probabilistic and stochastic methods are used in order to compare their ability to reproduce long series of hourly mean wind speed data

Persistency means that large values tend to be followed by large values.2. so that runs of values of similar magnitude tend to persist throughout the sequence.and second-order autoregressive processes are chosen in this study. in this study). A number of computational methods are available for the generation of random numbers with normal probability distribution of mean l and standard deviation r. that can be determined by using either a graphical method or the method of moments. Autoregressive processes are useful tools in generating data sets in cases where persistency exists. They can also be determined using the method of probability weighted moments (PWMs) for which explicit equations are available. First.1. and r standard deviation of wind speed. The normal probability distribution function is given by 1 f ðwÞ ¼ pﬃﬃﬃﬃﬃﬃ exp½Àðw À lÞ2 =2r2 r 2p ð1Þ where w is the variable (hourly mean wind speed. Normal distribution Hourly mean wind speed time series are generated by using a sequence of independent random numbers from the normal distribution. The results of these methods are compared to those obtained from a newly developed wavelet-based approach. a. 2. The methods are described below.ðln wÞ 0:5772 b ¼ exp L1. and small values by small values. The normal and Weibull probability distribution functions are chosen in order to generate independent and identically distributed random numbers. It is the method used in this study for the determination of parameters. It is given by a aÀ1 1 a f ðwÞ ¼ a w exp À a w w ! 0. Weibull distribution The Weibull distribution is another probability distribution function commonly used for the frequency analysis of wind speed data [14. respectively. Equations to be used for this purpose are given by lnð2Þ L2.ðln wÞ þ a a¼ ð3aÞ ð3bÞ . / Renewable Energy 29 (2004) 2111–2131 2113 with the same statistical behaviour as that of the observations in hand. Aksoy et al. 2. whereas the remaining ﬁve methods will be outlined brieﬂy as they have been well documented in literature.H. b > 0 ð2Þ b b where a and b are shape and scale parameters. Only the wavelet-based approach will be detailed. Another concept commonly employed in wind speed data generation studies is the ﬁrst-order Markov chain. l mean value of wind speed.15].

. The ﬁrst-order autoregressive [AR(1)] model accommodates only the eﬀect of the previous value in the series in which the observed sequence of wind speed data {w1. Eq.(ln w) and L2. model parameter.. . 2. wt. but autoregressive models are of correlated type and hence capable of simulating this property of the data series. / Renewable Energy 29 (2004) 2111–2131 In Eqs. a the autoregressive coeﬃcient.(ln w) are L1 and L2 moments of the logarithm of the hourly mean wind speed time series. and e a normally distributed independent random variable. (6) is written for the mth order. L1. w2. Once the parameters are determined. which is also called the Markov model. The simplest case of Eq. both normal and Weibull distributed random numbers do not take this property into account as they are independent. The L1 and L2 moments of a series are given by L 1 ¼ b0 L2 ¼ 2b1 À b0 in which b0 and b1 are given by b0 ¼ x b1 ¼ N À1 X j¼1 ð4aÞ ð4bÞ ð5aÞ ðN À jÞ xj NðN À 1Þ ð5bÞ xj in Eq.2114 H. . Aksoy et al. As mentioned.} is used to ﬁt a model of form wi ¼ m X aj wiÀj þ ei j¼1 ð6Þ where w is the hourly mean wind speed. that is..3. as the cumulative distribution function of the Weibull distribution can be obtained in closed form. It is noted that Eq. Detailed information on L moments and the method of PWM is given in [20].. (6) then becomes yi ¼ r1 yiÀ1 þ ei ð7Þ where y is the standardised (zero mean and unit variance) version of the variable and r1 the lag-one serial correlation coeﬃcient of the sequence. . AR(1) model The hourly mean wind speed time series is of high dependence.b). (5b) comes from the time series sorted in descending order as xN Á Á Á xi Á Á Á x1 . the generation of Weibull distributed random numbers is a matter of a simple computer code. This property particularly requires a wind speed data generation model incorporating the dependence structure of the observations. (6) is obtained for m ¼ 1. . The use of autoregressive type models is reported in literature very commonly. (3a.

Although they are not common as the ﬁrst. For example. higher-order Markov chains can also be used. dramatic increase in the number of their parameters limits their use. the observed time series is divided into a number of states. AR(2) is formulized as yi ¼ /1 yiÀ1 þ /2 yiÀ2 þ ei where autoregressive coeﬃcients /1 and /2 are given by /1 ¼ r1 ð1 À r2 Þ=ð1 À r2 Þ 1 /2 ¼ ðr2 À r2 Þ=ð1 À r2 Þ 1 1 ð9aÞ ð9bÞ ð8Þ in which r1 is the lag-one autocorrelation coeﬃcient and r2 the lag-two autocorrelation coeﬃcient of the wind speed time series. where R2 ¼ r2 þ r2 À 2r1 r2 1 2 1 À r2 1 ð10Þ 2. used a ﬁxed width for the states. the hourly mean wind speed data set in this study was divided into 10 states. Two previous states are used in the second-order or twostep Markov chain in determining the current state of the wind speed.H. etc. This becomes more important in cases where dependence in the data set is very obvious. Markov chain In this approach. In another wind speed study [11]. Aksoy et al. as in the hourly mean wind speed data. However. State 1 might include wind speeds below 2 m/s. Therefore. the second-order autoregressive [AR(2)] model is preferred to AR(1). the state of wind speed in the current hour can be deﬁned depending only upon the previous state. Dukes and Palutikof [10]. The simulation procedure for the processes is very sim1 ple. State 2 wind speeds between 2 and 4 m/s. This is called the ﬁrst-order or one-step Markov chain. states were deﬁned depending upon the standard deviation of the data set. on the other hand. For instance. The upper and lower limits of the states are highly subjective values. It requires only a random number of a normal distribution to be generated. which was equal to 2 m/s. the dependence structure in the observations is better preserved.4. .and second-order Markov chains. until the ﬁnal wind speed state includes all speeds above the highest observed value or a predeﬁned upper limit.5. Each state in that study [11] was taken as wide as one standard deviation of the observed hourly mean wind speed time series. In the Markov chain approach. with zero mean and variance equal to 1ÀR2. AR(2) model With increase in order of the autoregressive model. A wind speed state contains wind speeds between certain values. The random component in AR(2) is again of normal distribution. / Renewable Energy 29 (2004) 2111–2131 2115 The random component (e) in AR(1) is of normal distribution with zero mean and a variance of 1 À r2 . 2.

Then. the cumulative transition probability matrix is calculated.2116 H. .6. P2m 7 7 P ¼ 6 21 ð11Þ 4 . . Some examples of wavelets are Morlet. In the cumulative transition probability matrix. . The number of parameters is mðm À 1Þ. . then a shifted one-parameter gamma distributed random number is used in order to ﬁnd the magnitude of the wind speed. 1). . the Haar wavelet was used due to its simplicity. 2. . then a uniform random number is generated from the interval of State 1. If State 1 is obtained as the new state of wind speed. a uniform random number from the interval of the corresponding state is generated and set as the wind speed at the current hour. deﬁned as 8 < 1 0 t 1=2 wðtÞ ¼ À1 1=2 t 1 ð13Þ : 0 otherwise Decomposing a signal and then reconstructing it is the base for the wavelet transform. the next state of wind speed can be determined. . . be assumed as the initial state. as the sum of the probabilities is equal to 1 (100%) for each row of the matrix. j ¼ 1. . . each row in that matrix ends with 1. There are many functions that can qualify as wavelets. an initial state is adopted. / Renewable Energy 29 (2004) 2111–2131 The parameter set of a Markov chain consists of probabilities of transition from one state to another that are given in transition probability matrices. . m ð12Þ The procedure for generating the simulated hourly mean wind speed time series is explained below. Pmm where Pij is the probability of transition from state i to state j. If nij is the total number of hours of observation in state j with the previous state i. Aksoy et al. cumulative summation of probabilities within each row is carried out. . the probabilities of transition from state i to state j can be calculated as nij Pij ¼ P nij j i. P1m 6P P22 . then it is ﬁrst checked if the wind speed is zero. Pij . For intermediate states. Mexican hat. In this study. No wind (State 1) can. First. Using a uniform random number. 2. for example. The reason for choosing the gamma distribution will be discussed in the section where results obtained from application of the methods are presented. . Wavelet-based approach A real or complex-value continuous function with zero mean and ﬁnite variance is called a wavelet [21]. hence. There- . The transition probability matrix of a ﬁrst-order Markov chain with m states can be written symbolically as 2 3 P11 P12 . . . A simple wavelet is the Haar wavelet (Fig. If the wind speed is not zero. 5 Pm1 Pm2 . Shannon and Meyer. . . If the highest state is found to be the new state of wind speed.

H..À1. ð ðlþ1Þ=2 l lþ1 fÀ1 ðtÞ ¼ 2 <t< f ðsÞds 2 2 l=2 ð ðlþ1Þ f0 ðtÞ ¼ f ðsÞds l < t < ðl þ 1Þ l ð 2ðlþ1Þ 1 f ðsÞds 2l < t < 2ðl þ 1Þ f1 ðtÞ ¼ 2 2l . Haar wavelet. 1. f1 ðtÞ ¼ 0 ð15Þ . . k a scale variable (k > 0 means stretching and k < 0 means contracting of the wavelet) and l a translation variable [21]. decomposition of a signal (multiresolution analysis) with the Haar wavelet is considered and explained in detail below. 1. . For k ¼ À1. Aksoy et al.. . 1.. fore. 0. .. . fk(t) is as follows: fÀ1 ðtÞ ¼ f ðtÞ . . . let us deﬁne fk(t) as the average of f(s) over an interval of size 2k: ð k 1 2 ðlþ1Þ fk ðtÞ ¼ k f ðsÞds 2k l < t < 2k ðl þ 1Þ ð14Þ 2 2k l where k and l are integers. / Renewable Energy 29 (2004) 2111–2131 2117 Fig. For a certain value of k. .

gk(i). in which the average of the time series taken at diﬀerent resolution levels according to Eq. . one obtains a simulated value for f(t) as f ðjÞ ¼ K X gk ðjÞ k¼1 ð18Þ where j is the index for generated elements. . (16) for diﬀerent resolution levels.. (17).. where K is a positive integer (K ¼ 4 for the sequence in Fig. K. . . The generation algorithm is given step by step as follows [22] and is illustrated in Fig. . K. 2.. . f0(i) is the original sample and fK(i) is a sample of all zeros. (15) that f 4 ðtÞ ¼ 0 for all t. 2 shows the detail functions calculated using Eq. 2) taken from a stochastic process f(t) with zero mean: f(1). the original signal is obtained when all detail functions are summed up. the detail function values right next to those of the previous step jÀ1 at each resolution level. . the gk coming just after the gk values chosen in the ﬁrst step.. for each k. (15) is shown. 2. K) are chosen from M values randomly and summed up to obtain f1 (Fig. 2... Let us consider a data sample of size M ¼ 2K . the resolution level. Aksoy et al. f(t).2118 H. Note from Eq. . f(M). Choosing from M elements for each gk(t) randomly. . f2 is obtained by the summation of these (Fig. (17) is the basis for the generation algorithm explained below. (17). M) consisting of averages of 2k successive elements of the sample.. 3. gk values (k ¼ 0. since the average of M elements is zero. f0(t). Data generation is continued in this way for a desired number of times using. is seen. 3).. Deﬁne the sample fk(i) (k ¼ 0. (16) for k ¼ 1. 3). and then summing them up using Eq. (17). Eq. for each element fi of the original sample. The detail function gk(t) has a sample consisting of M elements given by Eq. corresponding to diﬀerent resolutions. 1.. Thus. for the generation of each element fj. Increase in the ordinates of fk(t) with decrease in k shows the change (increase) in the resolution. we have K detail function values. 3 for K ¼ 4. The second element ( j ¼ 2) is generated by choosing. 2. The diﬀerence between the successive averages fkÀ1(t) and fk(t) is deﬁned as a detail function: gk ðtÞ ¼ fkÀ1 ðtÞ À fk ðtÞ It can be easily seen that 1 X gk ðtÞ f ðtÞ ¼ k¼À1 ð16Þ ð17Þ According to Eq. Note that the data sample used in Fig. 2 has 16 elements. . / Renewable Energy 29 (2004) 2111–2131 The resolution decreases as k increases. . can be seen in the upper part of Fig. In order to obtain the ﬁrst element of the series ( j ¼ 1). Change in data resolution with change in k.. and it represents the original data. At the bottom of Fig. 1. the sum of the four detail functions according to Eq. f(2). i ¼ 1. The middle part of Fig. 1.

Decomposition and reconstruction of a data sequence. / Renewable Energy 29 (2004) 2111–2131 2119 Fig. Aksoy et al. . 2.H.

generate them and then transform them back to their skewed structure. The performance of the methods . This generation algorithm is a newly developed approach for data simulation purposes. Aksoy et al. Construction of a simulated data sequence. It was ﬁrst used in non-skewed annual and monthly streamﬂow data simulation studies [22. The approach was later used for the simulation of the storage capacity of river reservoirs [24]. Results obtained from the application of the methods are presented and discussed below. standard deviation and correlation structure of the observed streamﬂow data sets. 3. Application The methods were applied to an hourly mean wind speed data set that will be introduced in the following subsections. it is ﬁrst required to transform the data to a non-skewed structure. Modelling suspended sediment discharge series [25] and annual and monthly rainfall data series [26] was also performed by this approach successfully. / Renewable Energy 29 (2004) 2111–2131 Fig. When one is interested in the generation of skewed data.23].2120 H. The algorithm generated the mean. 3.

H. As the normal distribution is ﬁtted to the transformed hourly mean wind speed time series (but not to the raw data series). although the Weibull probability distribution function . 3. This prevents ﬁtting of the normal distribution to the data. Therefore. to be explained later in following sections. For the wavelet-based approach. Characteristics corresponding to that part of the observed series are also given in Table 1. The data set is of four years’ length. Aksoy et al. The Weibull distribution has two parameters (a. power transformation [ y ¼ xh . The parameters of the Weibull distribution determined by the method of L-moments are listed in Table 2. Those parameters are presented in Table 2. Parameters The hourly mean wind speed data set used in the study is of skewed structure. zero wind speeds were ignored from the observed time series as their number of occurrences was very small. The only problem with this method is the presence of zero wind speeds. 3. A comparison of the methods is ﬁnally presented. less than 0. Fig. It is seen that the distribution performs very well in ﬁtting to the observations as well as to the generated data. 1994. The data set is highly correlated. the shape parameter. The region is normal.2. The transformation coeﬃcient was obtained as h ¼ 0:38585 for the data set in the study. The parameters were determined using the method of L-moments on which detailed information was given previously. It can be considered a very good ﬁt. In order to overcome this problem. extending from the ﬁrst hour of April 6. as expected. and h the transformation coeﬃcient] was adopted in order to obtain nonskewed data.1. were used. to which the normal distribution can be ﬁtted. The reason for choosing this method is that explicit equations are available for determination of the parameters of the distribution. the scale parameter). Data Table 1 shows the main statistical characteristics of the data set of hourly mean wind speed taken from the State Meteorological Works’ meteorology station in Diyarbakir. 5 shows the agreement between the observed data and the ﬁtted Weibull probability distribution function. 4). 1997. 32768 hours of data.5%. which makes the method inapplicable due to the logarithm included. where x the is raw (untransformed) variable. and skewed. and b. / Renewable Energy 29 (2004) 2111–2131 2121 was measured according to their ability to capture the statistical behaviour of the observed data set. a southeastern Anatolian city. from 1994 to 1997 (35064 hours in total). to the eighth hour of December 31. The method also has the superiority of being less sensitive to outliers. which means that outliers do not aﬀect the performance of the method in determining the parameters correctly. This is a choice with no speciﬁc reason. The normal probability distribution function based upon the determined parameters was ﬁtted to the transformed wind speed data series (Fig. y the transformed variable. the parameters of the normal distribution are the mean and standard deviation of the transformed hourly mean wind speed time series. as is seen in Table 1.

476 0.4 Coeﬃcient of skewness Correlation coeﬃcient r3 0.549 0.861 r2 0.703 0.4 1.283 14. Aksoy et al.635 r4 0. / Renewable Energy 29 (2004) 2111–2131 6 April 1994–31 32768 December 1997 2.2122 Table 1 Statistical characteristics of observed hourly mean wind speed time series Coeﬃcient of variation r1 0.860 0.702 1.438 Standard deviation (m/s) 1.786 1.732 0.285 14.794 Maximum wind speed (m/s) Date Number of data Mean (m/s) 1 January 1994–31 35064 December 1997 2.555 .733 0.538 H.633 0.551 r5 0.

the variance of the independent normal variable). each 1. The model e requires only the lag-one serial correlation coeﬃcient (r1). AR(2) has three parameters (/1 and /2. 4. AR(2) Parameter set l ¼ 1:347 m=s a ¼ 1:583 r1 ¼ 0:820 r ¼ 0:392 m=s b ¼ 1:973 r2 ¼ 0:688 2123 gives the mode an occurrence probability slightly lower than that in the observation. when it is considered that summation Fig.H. and r2 . Aksoy et al. Of the six methods.5 m/s wide. In this study. The number of parameters required changes with the number of states used for the wind speed. all functions of r1 and r2. This resulted in 90 transition probabilities to be determined from the observed wind speed data set. the lagone and lag-two serial correlation coeﬃcients listed in Table 2. the autoregression coefﬁcient. 10 states were chosen for the wind speed. AR(1) is a parametric model with two parameters (a. . and r2 . the Markov chain is the one that requires the highest number of parameters. the autoregression coeﬃcients. e the variance of the independent normal variable). as both parameters are dependent only upon r1. / Renewable Energy 29 (2004) 2111–2131 Table 2 Parameter sets of methods Method Normal Weibull AR(1). Normal probability distribution function ﬁtted to the observed and simulated random wind speed sequences.

0000 10 0.7053 0.0000 0.0015 0.3636 0.0176 0.0000 0.3636 1.2632 0.0000 0.0008 0.0000 0.0491 0.0000 . In State 1 with the lower limit of zero.0263 0.0111 0.0041 0.3116 0.0000 0.2124 H.0000 0.0256 0. Weibull probability distribution function ﬁtted to the observed and simulated random wind speed sequences.0178 0. / Renewable Energy 29 (2004) 2111–2131 Fig.0152 0.0000 0. Not only transition probabilities.0000 0.0089 0.4013 0.0909 0. wind speed was assumed to be distributed uniformly over the states except for the last one (state of highest wind speeds with no upper limit).0008 0.3684 0.0008 0.0865 0.0000 5 0.0000 7 0.0000 0.2405 0.0000 0. The transition probability matrix of the data set is given in Table 3.1818 0.0000 0.0526 0.0000 6 0.0005 0.0000 0.0005 0.0153 0.3486 0.0000 0.0000 0.0022 0.0000 2 0.0000 4 0.0000 0.1212 0.0168 0.4311 0.0144 0.1352 0.0865 0.4954 0.0000 0.0076 0.2779 0.0001 0.6089 0.0000 0.0266 0.0000 0.2839 0.0042 0.0048 0.0001 0.2105 0.1061 0.0789 0. In this study.0000 8 0.0179 0. where the one-parameter gamma distribution was used.3561 0.0000 0.0455 0.1191 0.0000 0. the probability of occurrence of zero wind speed was also taken Table 3 Transition probability matrix of the observed hourly mean wind speed data set Pij i¼1 2 3 4 5 6 7 8 9 10 j¼1 0.5317 0.3485 0.0003 0. but also the wind speed distribution in each state should be known by this method.0000 9 0.0002 0.0023 0.0000 0.0003 0.0978 0.3437 0.0000 0. Aksoy et al.0000 0. over any row in the transition probability matrix results in 100% probability.0000 3 0.0000 0.1306 0.0000 0. 5.0000 0.0000 0.1197 0.

Simulation and results A thousand-year (8760000-hour) -long series was generated for each method. 5 that the Weibull ﬁt is perfect as well. 4 that wind speed data generated by the normal distribution ﬁt the observed series very well. and Markov chain methods looked to produce the dependence structure of the series. It is seen from Fig. / Renewable Energy 29 (2004) 2111–2131 2125 into consideration in order to reproduce the zero wind speeds. The normal and Weibull distributions. Another requirement for the wavelet approach is that the data set should be of a non-skewed structure. Aksoy et al. 6. however. where K is a positive integer and equal to 15 in this study. 6). Other than those two methods. Correlogram of the observed and simulated wind speeds. the highest wind speed that the structure will possibly face during its lifetime. although their occurrence was very low. are of independent structures (Fig. However. the part of the observed series used for the wavelet approach was transformed by using the power transformation with h ¼ 0:3853. the success of those methods in reproducing the correlation structure Fig. with increasing lags in time. . as it is a nonparametric method. It is obvious that the hourly mean wind speed time series has a highly dependent structure. The correlogram. frequency distribution of maximum wind speeds and wind duration curve obtained from the simulations will be compared to those of the observed series. to the structural designer. The length of the series to be used in this method is equal to 2K. Therefore. These methods may be useful in oﬀering. It is seen in Fig. AR(2). the AR(1).3.H. yet they are very common methods used in generating wind speed data. This corresponds to a series 32768 hours in length. There is no parameter to be listed for the wavelet approach. 3.

which was taken as 12 m/s in this study. it is safer to use the wind load due to the maximum wind speed generated by the wavelet-based approach. State 9 can be simulated if and only if the previous state of the wind speed is one of the following states: 3. of the series decreases (Fig. Aksoy et al. The annual maximum values of the simulation series were compared in Fig. 7 due to this circumstance. was found to be the best in preserving the correlation structure of the series. The reason for those jumps can be explained very simply. Similarly. It is seen that maximums of the simulated series are bounded by the upper limit of State 8. which was taken as 13. A very small jump exists in the frequency curve in Fig. Cumulative frequency diagram of maximum wind speeds. / Renewable Energy 29 (2004) 2111–2131 Fig. The big jump in the cumulative frequency curve in Fig. as is seen in Fig. Maxima obtained from the Markov chain method should be discussed speciﬁcally. .5 m/s in this study. whereas the wavelet approach produced higher. The wavelet method of the six studied. There are three vertical jumps (one of them is very obvious) in the cumulative frequency diagram of the maxima of this method. It is seen that the normal distribution. It is only possible to make a transition to State 10 if the previous state in the simulation series is either State 8 or State 9. 7. making transition of wind speed to those states almost impossible in the simulation series. 6). 8. the Markov chain slightly lower and the Weibull distribution considerably lower maxima. From the structural engineering point of view. It is seen from Table 3 that the probabilities of transition of the wind speed to the highest states are too low. 9 and 10 (Table 3). 6. 7. This causes the maximum value of the series to be bounded by the upper limit of State 9. 7 is due to this situation. therefore. Otherwise State 10 is not simulated. AR(1) and AR(2) produced similar maxima.2126 H. 7.

Such a forcing can be considered quite reasonable and it does not aﬀect the transition probability matrix as the number of data is usually very large (of the order of tens of thousands). The distribution generated low maximum wind speeds. It is an important tool used in determining the percentage of time that the wind speed exceeds a speciﬁed level. then lower maximum wind speeds are generated. it was ﬁrst thought to simulate wind speeds of the highest state by using the exponential distribution shifted to that state as Sahin and Sen [11] did. the Gumbel distribution was tested. where the wind duration curves of the six methods were plotted together with the one extracted . Finally. A very good ﬁt was obtained in Fig. 7) were obtained. In the end. 8). in this study. The Frechet distribution. whereas the one-parameter gamma distribution was adopted for the highest state (State 10 in this study). which means that the Markov chain based simulation technique worked very well in the simulation of the state of the wind speed series. It was seen that the maximum wind speeds generated by this distribution were too low compared to those obtained by the other methods. which limits the maximum wind speed to 10. which resulted again in low maximum wind speeds. it was seen that the exponential distribution generated lower maximum wind speeds compared to those generated by other methods. mean-dependent probability distribution functions are better in the simulation of maximum wind speeds. Uniformly distributed wind speeds were accepted for the intermediate states. 8. the two-parameter gamma distribution was ﬁtted. was also found to be unsuccessful in generating maximum wind speeds compared to other methods. The reason for choosing this distribution is explained below together with a discussion on other distributions.H. This is quite a reasonable choice for simulating the wind speeds in that state. The wind duration curve is a graph with time percentage as abscissa and wind speed as ordinate (Fig. / Renewable Energy 29 (2004) 2111–2131 2127 The third jump at the very beginning of the curve (close to the y-axis of the graph) is due to a similar situation. Therefore. which is accepted as the distribution of the maximum wind speeds [1]. Wind energy production systems use this graph in order to determine the wind energy potential of the region under consideration.5 m/s at the upper limit of State 7. Aksoy et al. This drawback of the method can be overcome by forcing the simulation series to have at least one value from the highest state that results in a maximum wind speed data series all generated from the highest state with no upper limit. A distribution with no upper limit should be used for the highest state so that maximum wind speeds higher than those in the observed series can possibly be generated. However. is included in the generation scheme. Therefore. Therefore. which is bounded by the lower and upper limits of the state. The conclusion that can be drawn from those trials is that a one-parameter distribution can ﬁt to the highest state better than distributions with two or more parameters. The transition probability matrix of the Markov chain based simulation wind speed series is given in Table 4. the one-parameter gamma distribution was ﬁtted and results comparable to those of the other methods (in Fig. This is the result of not having simulated wind speeds from State 8. If the standard deviation of the highest state. It is almost the same as its observed counterpart given in Table 3.

2845 0.2406 0.3474 0.0000 0.0000 0.0003 0. Wind duration curve of observed and simulated wind speeds.0000 0.0259 0.2782 0.0000 0.0143 0.0001 0.0000 from the observed series.0005 0.0000 0. Standard deviation and variation coeﬃcient were best captured by the normal probability distribution.3526 0. Although the wind duration curve of the Markov chain method ﬂuctuates around the others.0000 0.0000 0.0000 10 0.0042 0. The wavelet-based method approaches its counterparts with a relative error of 0.3491 0.0005 0. . Aksoy et al.0001 0. Skewness coeﬃcient in the wind speed time series was best reproduced by Fig.0000 0.0000 0.0000 3 0. 8. The maximum values and the ﬁrst ﬁve lags of the correlation are also listed.4028 0.0000 0.0000 2 0.0777 0. and AR(1) and AR(2) processes. it has a ﬁt that is good enough as well.0023 0.0000 0.0000 0.0255 0.0000 0.0501 0.6086 0.0773 0.3115 0.0001 0.0048 0.0478 0.0113 0.0003 0.3421 0.0089 0.0177 0.1888 0.1206 0.4322 0.0181 0.0000 8 0.1191 0.1045 0.0022 0. It is seen that the mean values of the simulated series are almost the same as those of the observations.3720 1.0860 0.0000 9 0.0000 0.0000 0.1346 0.0016 0.4958 0.0000 0.0179 0.2128 H.3667 0. The ﬁrst three central moments of the observed and simulated series are given in Table 5.0000 4 0.0008 0.0002 0.0041 0. / Renewable Energy 29 (2004) 2111–2131 Table 4 Transition probability matrix of hourly mean wind speed data simulated by Markov chain method Pij i¼1 2 3 4 5 6 7 8 9 10 j¼1 0.0861 0.0000 0.0000 5 0.0000 0.0000 0.0153 0.0165 0.2606 0.0000 7 0.3%.0252 0.2197 0.0000 0.1309 0.3619 0.1227 0.0008 0.0000 0.0077 0.0156 0.0000 0.0000 6 0.0009 0.0982 0.5318 0.7049 0.0491 0.0000 0.0000 0.0001 0.0000 0.

715 r2 À0.436 0.537 2.566 1.387 0.537 2.815 0.443 r5 À0.536 0.700 0. / Renewable Energy 29 (2004) 2111–2131 Normal Weibull AR(1) AR(2) Markov Wavelet 2.776 2.058 1.523 r4 0.983 1.714 1.777 1.0003 0.585 2.776 1.Table 5 Statistical characteristics of simulated series Coeﬃcient of variation r1 0.634 1.0002 0.796 0.30 25.13 Coeﬃcient of skewness Correlation coeﬃcient r3 0.315 0.661 0.309 1.23 21.483 0.538 2.0003 0.0005 À0.313 0.421 Maximum wind speed (m/s) Series Mean (m/s) Standard deviation (m/s) H.0007 0.29 31.561 0.438 26.0002 À0.0002 À0. Aksoy et al.815 0.833 2129 .0005 0.700 0.529 2.677 0.980 1.580 0.94 16.700 0.466 0.646 0.399 0.22 27.355 0.330 0.587 0.715 0.0001 0.0005 0.

the newly developed wavelet-based approach was used. This method is as good at preserving other statistical features of the series as other methods. Therefore. Some methods performed better in preserving some particular characteristics than other methods did. whom the authors sincerely thank. Additionally. 4. The generation of maximum wind speeds requires special attention in Markov chain based simulation methods.2130 H. For example. / Renewable Energy 29 (2004) 2111–2131 AR(1). Higher maximums were obtained by the methods of AR(1). such as the one used in this study. All the mentioned methods are parametric and they therefore require the time series to have a speciﬁc probability distribution. It is concluded that the one-parameter gamma distribution is good enough in ﬁtting to the wind speed data in the highest state of the series for normal regions. wind duration curve and statistical features of the series were used in order to compare the success of the methods. Correlation structure. in conclusion. Summary and conclusion In this study. However. A nonparametric model. This is a drawback of parametric models more than a limitation. The correlation structure of the observations. The normal and Weibull probability distribution functions consist of independent identically distributed random numbers. wavelet and normal distribution and lower maximums by Weibull distribution. of which the wavelet approach in this study is one of the best examples. Aksoy) and Professor M.and second-order autoregressive processes. hourly mean wind speed data sets were generated by traditional simulation methods—the normal and Weibull probability distribution functions. The Markov chain is a two-step method that ﬁrst determines the state of the wind speed and then generates its magnitude by using a preselected distribution. Bayazit of Istanbul Technical University. the wavelet method is obviously the best in preserving the correlation structure of the sequence. and the Markov chain. Turkey. The autoregressive models include the correlation structure of the observation and hence generate dependent series. the ﬁrst. can be applied to data sets with any distribution. Acknowledgements The wavelet approach presented in this study is a result of an earlier cooperation between the ﬁrst author (H. as discussed earlier. Based upon the application in this study. the wavelet method is proposed as a tool to substitute for the classical generation schemes for the simulation of hourly mean wind speed data. it is concluded that the uniform probability distribution function is suitable for use in the ﬁrst and intermediate states. Aksoy et al. distribution of the maximum wind speeds. it should be kept in mind that the wavelet approach works only with sequences of zero skewness. . A probability distribution function with no upper limit should be used for the highest state. was best simulated by the wavelet-based method.

Wavelet analysis for modeling suspended sediment discharge.111(1):77–94. [19] Simiu E. ASCE Journal of Structural Engineering 1998. Unal NE. [9] Sfetsos A. Journal of Wind Engineering and Industrial Aerodynamics 1988. Extreme winds simulated from short-period records. Nomura T. An expert system for extreme wind simulation. [5] Cheng EDH. [18] Pandey MD. Solar Energy 1998.23:179–92. [10] Dukes MDG. [25] Aksoy H. Extreme wind distribution tails: a ‘peaks over threshold’ approach. Scanlan RH. [21] Rao RM. MA: Addison-Wesley. [20] Stedinger JR. Handbook of hydrology. Journal of the Structural Division 1982. A comparison of various forecasting techniques applied to mean hourly wind speed time series. New York: McGraw Hill Book Co. Extreme winds generated from short records in a tropical cyclone-prone region. [4] Cheng EDH. [22] Bayazit M. [16] Heckert NA. Using wavelets for data generation. Simiu E.38:101–8. 1994. Wavelet transforms. Festa R.H.124(4):445–9. introduction to theory and applications. Storage capacity for river reservoirs by wavelet-based generation of sequent peak algorithm. Akar T. Journal of Wind Engineering and Industrial Aerodynamics 1998. A comparison of alternative approaches for the synthetic generation of a wind speed time series.36:1235–43. Bopardikar AJ. Shaver JR.108(ST5):1034–48. Aksoy H. Frequency analysis of extreme events.35:165–74. Wind data generator: a knowledge-based expert system. The estimation of extreme quantiles of wind velocity using L-moments in the peaks-over-threshold approach. Hydrological Sciences Journal 2001. [13] Kitagawa T. Short-term records and extreme wind speeds. Torres JL. editor. Water Resources Management 2001. [23] Bayazit M. Estimation of extreme wind speeds with very long return periods. Heckert NA. Stochastic modelling of wind velocities time series. Kirchhoﬀ RH. Chiu ANL.18(6):in press. Palutikof JP. [3] Simiu E. Assessment of ‘peaks over threshold’ methods for estimating extreme value distribution tails.21:23–35. Ratto CF.89:263–9. ASCE Journal of the Structural Division 1982. New York. [14] Garcia A. [12] Castino F. [11] Sahin AD. Renewable Energy 2000. Onoz B. Nordic Hydrology 2004. Prieto E. Transactions of the ASME 1991. Vogel RM.108(ST11):2571–7. Journal of Wind Engineering and Industrial Aerodynamics 2001.91:943–64. . ASCE. Minimum design loads for buildings and other structures. First-order Markov chain approach to wind speed modelling. ANSI/ASCE 7-93 (Revision of ANSI/ASCE 7-88). Aksoy H. Estimates of design wind from short records. 1998. Fitting wind speed distributions: a case study. Nonparametric streamﬂow simulation by wavelet or Fourier analysis. Journal of Wind Engineering and Industrial Aerodynamics 1990. Wind eﬀects on structures.15(6):423–37.122(5):539–47. [6] Cheng EDH. / Renewable Energy 29 (2004) 2111–2131 2131 References [1] Simiu E. Reading.34:1950–61. Estimates of hurricane wind speeds by ‘peaks over threshold’ method. [26] Unal NE. 1993 [Chapter 18]. ASCE.62(2):139–44. Journal of Structural Engineering 1985. [15] Grigoriu M. ASCE.28(2): 157–66. A wavelet-based method to generate artiﬁcial wind ﬂuctuation data. De Francisco A. [8] Kaminsky FC. Journal of Wind Engineering and Industrial Aerodynamics 1991. Heckert NA. In: Maidment D. Stochastic Environmental Research and Risk Assessment 2044.74–76:141–51. [24] Aksoy H. 1986. Chiu ANL.46(4):623–34. [2] American Society of Civil Engineers. [17] Lechner A. Filliben JJ. Chiu ANL. Syu CY. Manwell JF. Simiu E.113:280–9. Akar T. Sen Z. [7] Cheng EDH. Journal of Structural Engineering 1996. Structural Safety 2001. Structural Safety 1993. Annual and monthly rainfall data generation schemes.28:69–78. Aksoy H. Foufoula-Georgiou E. Journal of Applied Statistics 2001. New York: John Wiley & Sons. Aksoy et al. Van Gelder PHAJM. Vrijling JK. Journal of Wind Engineering and Industrial Aerodynamics 2003. Journal of Applied Meteorology 1995.12:305–14. Whalen T.

- Read and print without ads
- Download to keep your version
- Edit, email or read offline

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

CANCEL

OK

You've been reading!

NO, THANKS

OK

scribd