2011 International Conference on Alternative Energy in Developing Countries and Emerging Economies
- 344 -
high variability both in space and time. Furthermore,
wind energy is intermittent exhibits large variability in
the production of energy due to various factors, such as
wind speed, air density, and turbine characteristics. The
theoretical energy of wind is proportional to the cube of
wind speed and slight changes in the wind speed might
cause significant changes in the total electricity produced
by wind. In practical manner, the power output of wind
turbine and the wind speed are matched using a power
curve which usually perform S-shape characteristics.
Consequently, more accurate forecasting of the wind
speed is of great importance in energy production. It has
already been demonstrated that the forecasting of wind
power is beneficial for the optimum operation of a power
system with a significant penetration level from wind. An
accurate wind forecast is helpful in maximization of
power system reliability [10]. Although in literature many
researchers have used different techniques to forecast the
wind speed, however, it is obvious that for short-term
forecasting, statistical technique offer better results while
for long-term forecasting it relies on meteorological
methods. This paper presents the short-term forecast of
wind speed along the coast of Chana district in Songkhla
province, southern Thailand based on two different
statistical methods for comparison between the Box-
Jenkins and decomposition models. Generally, model
performance is evaluated in a variety of ways e.g. mean
error (ME), mean absolute error (MAE), mean squared
error (MSE), root mean squared error (RMSE),
improvement over persistence, and correlation with real
data. In this work, the MSE was used as a main criterion
for judgment of the appropriate method.
II. M
ATERIAL AND
M
ETHOD
Wind speed and wind direction at 20 m, 30 m, and 40
m were measured using 3-cup anemometers and wind
vanes model HOBO at Chana district in Songkhla
province, southern Thailand. The geographical coordinate
and elevation above sea level (a.s.l.) of wind met station
is given in Table I. The geographical position of wind
energy research station is shown in Fig. 1. Resolution of
wind speed measurement is 0.19 m/s and wind direction
measurement is 1.4
o
. Accuracy is
r
3% at wind speed of
17-30 m/s. Wind speed and direction sensors were
connected to a HOBOWare data logger. A PV battery
charging system is used for power supply backup.
Ambient air temperature was also measured. The
lightening arrester is used for protecting the measuring
equipments from severe thunder. The main component
and picture of wind energy research station is shown in
Fig. 2. Sampling interval is 1 min and logging interval is
10 min. Wind speed and direction data were recorded
starting in August 2007.
The source of data for wind speed along the coast of
Chana, Songkhla, Thailand, to consider in this paper was
collected by the Solar and Wind Energy Research Unit,
Thaksin University. As the observed wind speed at 30 m
height has the highest data recovery rate that the number
of missing data is zero, therefore, we use this data set in
analysis. This paper used the wind speed at 30 m height
for statistical modeling. A first wind speed data set of 736
observations manipulated by the three-hour (3-h) in May
1-July 31, 2010 is used to create the models in order to
forecast a second wind speed data set of 248 observations
in August 1-31, 2010. Two methodologies to create the
forecasting models in this paper are the Box-Jenkins and
the decomposition methods. We then assess the quality of
forecast values from those two methods by MSE.
TABLE
I
G
EOGRAPHICAL COORDINATE OF WIND MET STATION
N
O
.
S
ITE NAME
L
AT
(
O
N)
L
ONG
(
O
E)
E
LEVATION
(A.S.L.)
1.
C
HANA
06
o
58’
100
o
37’
7
Fig. 1. Geographical position of wind energy research station at Chana
district, Songkhla province in southern Thailand.
Fig. 2. Wind speed measurement at 30 m height a.g.l..
