INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH

I. Ifanda et al., Vol.13, No.3, September, 2023

Optimizing Turbine Siting and Wind Farm Layout in

Indonesia

I. Ifanda*‡ , Didik Rostyono* , Rudi Purwo Wijayanto* , Nurry Widya Hesty* , Amiral Aziz*‡
, Khotimatul Fauziah* , Toha Zaky* , Ario Witjakso* , Ahmad Fudholi*,**‡

*Research Center for Conversion and Conservation Energy, National Research and Innovation Agency Republic of Indonesia
(BRIN), South Tangerang 15314 Indonesia
**Solar Energy Research Institute, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia
([email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected],
[email protected], [email protected])

‡
Corresponding Author; Ifanda, Amiral Aziz, Ahmad Fudholi, Research Center for Conversion and Conservation Energy,
National Research and Innovation Agency Republic of Indonesia (BRIN), South Tangerang 15314,
[email protected], [email protected], [email protected]

Received: 01.04.2023 Accepted:20.05.2023

Abstract- Wind resource assessments are required to identify a specific area capable of producing valuable energy from wind
speeds. This paper aims to optimize wind assessment through wind farm siting and layout in Indonesia’s semi-arid region. Wind
data collected on Sumba Island over a one-year period was analyzed to assess the area's wind energy potential. Wind Atlas
Analysis and Application Programme (WAsP) and Windographer were used to generate a generalized wind climate and resource
maps for the area. Wind farm layout and preliminary turbine micro-sitting were completed with various scenarios in mind to
achieve the best possible result. Four different scenarios are considered to maximize power output. There are 34 identical wind
turbines with a unit capacity of 90 kW in Scenario 1. Scenario 2 includes 20 identical wind turbines with a total capacity of 3000
kW. In Scenario 3, 14 identical wind turbines with 225 kW of unit capacity are used. There are 12 identical wind turbines with
a unit capacity of 250 kW in Scenario 4. The results showed that Scenario 1 produced the highest total net Annual Energy
Production (AEP) of 11,287 MWh/year with a 3.73% wake loss. The minimum wake loss seemed to be 2.62% in Scenario 4,
with a total net AEP of 10,221 MWh/year.
Keywords Wind resource assessment, semi-arid region, WAsP, wake loss.

1. Introduction energy mix by 2025, as stated in the National Electricity

Master Plan (RUKN), as well as a reduction in greenhouse gas
One of the main factors influencing every country's emissions of 29–41% by 2030 and net-zero emissions by
progress is its access to energy. A country's ability to maintain 2060. To meet the target, several studies assessing renewable
its energy utilization is a key factor in its success [1]. energy potential have been conducted. Hesty et al. [4] estimate
According to Tang et al. [2], the utilization of renewable the national wind energy potential, while Pranoto et al. [5]
energy has grown globally over time because of the need to estimate the hydro energy potential. A web-based application
adhere to international climate agreements that forbid the use has been created to calculate the energy potential of a rooftop
of fossil fuels as an energy source. Wind energy has recently solar PV system installed on a home by Nurliyanti et al. [6].
emerged as the leading sustainable energy alternative for According to the Indonesian Ministry of Mineral and Energy
reducing the effects of energy on anthropogenic pollutants in Resources (MMER), the total solar energy potential in
the atmosphere, with a total installed capacity of 651 GW Indonesia is 3,294.36 gigatons peak (GWp), which is
worldwide in 2019 [3]. As a rapidly developing economy, distributed evenly across the country. East Nusa Tenggara
Indonesia has set a target of 23% renewable energy in its total (369.5 GWp), Riau (290.41 GWp), and South Sumatra (285.1
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

GWp) are the three provinces with the greatest solar energy testing in New Zealand's established infrastructure. WAsP
potential. MMER estimates that Indonesia has 154.88 GW of was used to assess the wind power potential in Turkey's
wind energy potential, with onshore potential of 60.65 GW Mardin province, as well as the potential to meet specific
and offshore potential of 94.2 GW. Onshore locations with energy demands [18]. Kamdar et al. [19] investigated wind
high wind energy potential, including wind speeds of 6–8 m/s, energy potential using WAsP Tool for a small–scale wind
power densities of 400–500 W/m2, and AEP of 4-5 GWh/year, farm in south–eastern Thailand. Ratjiranukool [20] used the
can be found on the south coast of Java, South Sulawesi, Weather Research and Forecasting (WRF) meso-scale climate
Maluku, and East Nusa Tenggara. model to assess the wind energy potential of Thailand's
northern region. Meanwhile, Wang et al. [21] proposed wind
East Nusa Tenggara province, particularly Sumba Island,
energy potential assessment method based on wind speed,
has more than 200 MW of wind energy potential, with wind
wind direction and its power using finite mixture statistical
speeds ranging from 5 to 9 m/s. Sumba also had solar energy
distributions. Kwon [22] presented uncertainty analysis of
potential due to its daily solar insolation of 5 kWh/m2 [7].
wind energy potential assessment using probability models
Sumba was designated as the Iconic Island of Renewable
and Monte-Carlo based simulation procedure. Moreover,
Energy (RE) by a ministerial decree in 2015. The island's
several studies on predicting wind energy production using
sparsely populated settlements and huge open savannah plains
machine learning methods have also been developed to
provide plenty of open space for the construction of wind and
improve forecasting accuracy [23,24].
solar farms. Sumba, like the rest of Indonesia, has only two
seasons: dry and rainy. Indonesia's climate and wind surface Studies on the investigation of wind energy based on the
direction are mostly influenced by the Asian-Australian use of simulation software have gained attention in recent
Monsoon (AAM) system [8]. According to Alifdini [9], the years. The programs utilized for wind energy can be used to
Asian (Australian) monsoon is distinguished by a assess the preliminary economic and technical viability of the
northwesterly (southerly) wind that blows from Asia projected investments. One of the most significant design
(Australia) to Australia (Asia), bringing humid (dry) air and challenges is the high initial investment cost of wind farms,
creating a rainy (dry) season in the majority of Indonesian which could result in significant financial losses if the
regions. Due to its wind patterns, the AAM system is also feasibility analysis is flawed. WAsP is a well-known wind
known as the northeast monsoon (rainy season), which peaks energy application program and a powerful tool for assessing
from December to February, and the southwest monsoon (dry wind resources. However, there has been little research on the
season), which peaks from June to August. Because of its application of the WAsP model to wind resource assessment
lengthy dry season, which lasts from April through November in semi-arid areas. Using a methodology that complies with
(7 months) every year, Sumba is considered a dry island with international standards, this research will offer a scientific
an unusually semi-arid climate [10]. Sumba, as a semi-arid method for evaluating wind resources in semi-arid regions like
region, receives the least rainfall in the country. The monsoon Sumba Island. According to the authors' knowledge, there is a
brings steady, strong winds from June to August, while from gap in the literature regarding optimizing wind assessment
December to April it brings calm winds. through wind farm siting and layout using the WAsP tool in
the semi-arid region. The WAsP and Windographer tools were
Despite Sumba's high renewable energy potential, the
used in this study to optimize wind farm siting and layout after
region's energy security index remains low [11]. Sumba's
a thorough analysis of the wind resources. The WAsP
electrical system currently consists of three electrical systems:
simulation calculates the mean wind speed and power density
East Sumba, West Sumba, and Southwest Sumba. All these
for selected eastern Sumba locations. The power and capacity
systems rely heavily on diesel power plants spread across
factors of the sites with different wind turbine capacities were
Sumba Island. According to PLN's (Indonesia's state utility)
calculated and analyzed. This will enable decision-makers to
long-awaited 2021–2030 Electricity Supply Business Plan
set goals about how to meet rising power demands through
(RUPTL), a total capacity of approximately 42 MW of various
renewable wind power. Furthermore, this study will contribute
thermal and renewable power plants is planned for the island
to the National Electricity Master Plan, which aims to replace
of Sumba, with a total wind power generation capacity of 3
fossil fuels with up to 23% renewable energy in its total energy
MW. RUPTL also indicates that a 70 kV transmission line and
mix by 2025. The paper discusses some results of the study
70 kV substations will be built to improve power transmission
that was conducted to optimize the turbine siting and wind
efficiency between the east and west Sumba systems. The
farm layout in the semi-arid region of Indonesia.
potential for wind power generation at any scale, whether for
grid-connected or stand-alone systems, must be thoroughly
2. Material and Methods
assessed [12,13]. Furthermore, the optimal use of the available
wind resources and the viability of a wind power project
2.1. Description of Object Study
depend on the siting of the wind turbines as described by
Mathew [14] and Manwell et al. [15]. A precise and accurate
analysis of the data for the relevant area is required to This research was carried out in Hambapraing, East
construct the wind farm to its full potential [16]. Numerous in- Sumba Regency. Sumba Island has a semi-arid climate type
depth scientific studies on wind energy have recently been with a short rainy season that lasts only for four months
undertaken on a global scale. In order to enhance wind speed (December to March) and total annual rainfall of 547 mm up
forecasting techniques for the wind pattern over complex to 1100 mm during the period of 2018-2022 (statistics of
terrain, Flay et al. [17] employed CFD, Wind Atlas Analysis Sumba Timur Regency 2018, 2019, 2020, 2021, 2022).
and Application Program (WAsP) models, and wind-tunnel According to statistics for Sumba Timur Regency (2022), the

1352
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

average temperature varies between 36 C in October and 18 C The optimum wind farm layout should maximize energy
in July. The average rainfall in East Sumba Regency in 2021 production from an energy standpoint. This is mostly
was 102.17 mm, with the highest level in March at 286 mm. accomplished in wind farms by efficiently spreading the
Despite the fact that the most rain fell in March, the most rainy distance between wind turbines within the available region,
days occurred in January, with 25 days in contrast; from April maximizing the available wind resource while minimizing
to November, there was very little rain in East Sumba wake loss. However, from an economic standpoint, an
Regency. The short, wet season with little rainfall results in optimized wind farm layout is one that minimizes the unit cost
limited water availability. This area is characterized by a vast of produced energy. This is mostly accomplished by
expanse of savanna and short thorny plants that reflect the examining the trade-off between the energy yield benefits
nature of the region's dry climate, which deteriorates from increased wind turbine spacing and the greater
continuously due to successive drought years. with soils expenditures associated with the land costs, the electrical
containing sand, limestone, and coral rock. The brittle rock, cabling costs, and losses. Because the central part of Sumba
limestone, and soil in the area, which are unable to withstand where the wind farm is proposed is barren and uninhabited,
the force of the winds that carry the bits of disintegrating land prices are not an important consideration in the
material, are another problem. Due to stiff winds during the optimization process. In this study, we solely apply single-
dry months, there is a phenomenon where sand accumulates objective optimization to reduce the total wake effect of the
in some areas. wind turbine farm, without taking into account the cost of
connecting between wind turbines.
The measuring mast is at 9° 31' 5.82" S latitude and 120°
10' 3.01" E longitude, with a height of 100 meters above sea Figure 2 shows the flow diagram of this study. The main
level as shown in Fig. 1. Meteorological instruments such as goal is to optimize the geographical location of every single
anemometers, wind vane, barometric pressure, thermometers, WT (wind turbine) in the selected region, including its
rain gauges, and hygrometers are housed in the measuring capacity, capacity factor, and AEP. The objective function is
mast. From September 2014 to September 2015, the to minimize the wake effect of the WT’s array on the wind
instrument measured and recorded a 10-min average of wind farm. Referring to Azlan [26], some previous studies have
speed and direction, solar radiation, temperature, and air provided several guidelines prior to WT positionings, such as
pressure, as given in Table 1, which are primarily considered land availability, forbidden zones, maximum investment, WT
over an international standard period for wind measurement quantity and capacity, WT spacing, and electrical
[25]. infrastructure. In this study, we consider capacity and spacing.
There is no issue regarding land availability or forbidden
zones on Sumba Island since it is a semi-arid region with no
conservation area. This research was carried out in
Hambapraing, East Sumba Regency. Sumba Island has a
semi-arid climate type with a short rainy season that lasts only
for four months (December to March), with a total annual
rainfall of 547 mm up to 1100 mm during the period of 2018–
2022.
We performed two types of preprocessing to simplify the
optimization algorithm. Firstly, the geospatial location and its
information regarding the wind potential; and secondly, we
did image manipulation by dividing it into the number of rows
and columns. The geospatial location and information refer to
Fig. 1. The wind speed map for Sumba Island. the Indonesia Geospatial Information Agency (BIG) via
DEMNAS with the numbers 2106-13 and 2106-41, and a map
of Sumba Timur is used as the data map reference. The
Table 1. Data set description of the location number of rows and columns in the matrix of a wind farm
represents the location of a single wind turbine. The distance
Parameter Unit Value between the columns is the spacing between two neighboring
Average wind speed (m/s) m/s 6.16 wind turbines or crosswind spacing. Meanwhile, the distance
Max wind speed (m/s) m/s 19.3 between two rows indicates the spacing between the turbines
Min wind speed (m/s) m/s 0.4 in the downstream direction or downwind spacing. Its spacing
Average Wind direction ° 128 depends on the WT diameter, which means that influenced by
Average temperature °C 26.8 capacity, a higher WT capacity usually has a wider diameter.
Average Solar Radiation W/m2 250 The crosswind spacing is between 4D and 6D, and the
Average Barometric Pressure mB 998 downwind spacing is almost twice that, between 8D and 12D.
The priority and non-priority of a single location of WT are
decided by its potential energy based on the annual wind
velocity.
2.2. Methodology

1353
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

unit capacity above 300 kW. A single WT capacity in this

study is 90 kW, 150 kW, 225 kW, and 250 kW, which means
we have four scenarios of a WT farm layout, and the total
capacity for each scenario is about 3 MW. A 90 kW WT has a
rotor diameter of 20 m, whereas a 250 kW WT's rotor diameter
is 50 m. We set the crosswind spacing between 2D and 3D and
the downwind spacing between 4D and 6D.

2.3. The Weibull Distribution

To design and select wind farms, wind speed is the main

element that must be measured, while the performance of wind
turbines is significantly influenced by the Weibull function
probability distribution (PDF) [27]. In explaining the wind
speed histogram, two parameters are often used in the
calculation, namely the Weibull probability distribution. It is
also used in the WAsP to study wind characteristics in all
directions characterized by sectors [28]. The probability
distribution function (PDF) of the Weibull distribution is
defined by Eq. (1) [14,25,29]:
k −1 k
U 
k U  − 
 A
f (U ) =    e (1)
A  A

where k  0; U  0; A  0 .
The Weibull probability density function of observed
wind speed U (m/s) is represented by f (U), the Weibull scale
parameter in m/s is defined by A, and the dimensionless
Weibull shape parameter is represented by k. The Weibull
shape parameter k has values between 1 and 3 and describes
the behaviour of wind in accordance with its speed; where
small values of k show variations in wind variables, a rather
constant wind speed can be shown with a large k value [25,28].
Fig. 2. Flow diagram of the study. Then, the Weibull distribution is expressed according to
the cumulative probability function in Eq. (2) [30,31]
Figure 3 shows the two preprocessing datasets before feeding
them to the optimization algorithm as a matrix. This research
was carried out in Hambapraing, East Sumba Regency. Sumba U 
k
− 
Island has a semi-arid climate type with a short rainy season f (U ) = 1 − e  A
(2)
that lasts only for four months (December to March) and total
annual rainfall of 547 mm up to 1100 mm during the period of where f(U) is the cumulative distribution function of the
2018-2022 (statistics of Sumba Timur Regency 2018, 2019, observed wind speed U. The cumulative distribution is the
2020, 2021, 2022). According to statistics for Sumba Timur integral of the density or PDF with respect to speed [25].
Regency (2022), the average temperature varies between 36 C
in October and 18 C in July. The average rainfall in East 2.4. Calculation of Wake Effect
Sumba Regency in 2021 was 102.17 mm, with the highest
level in March at 286 mm. The Jensen wake model is mostly used in previous
We proposed a total capacity of 3 MW of WTs based on studies, and other wake models are Gaussian or Frandsen [26].
Indonesian electricity planning [RUPTL] for the region. Due The Jansen wake model is shown in Fig. 4 [32], there are no
to the limited facilities and infrastructure in the study region, external forces acting upon the control volume and only the
including transportation, facilities, port capacity, bridges, angular momentum is conserved inside the wake.
roads, etc., it is not possible to choose a wind turbine with a

1354
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

(a) (b)
Fig. 3. (a) Pre-processing using WAsP and (b) Pre-processing image with the value of the rectangular area to identify
nonpriority and priority area as constraints.

1−
rd = r0 (6)
1 − 2a
where r0 is rotor radius, and a is axial induction factor.
The wind slows down as it approaches the turbine. The
ratio of this reduced wind speed and free stream velocity, u0,
is axial induction factor, a, calculates use the Thrust
Coefficient ct, as given in Eq. (7)

ct = 4a (1 − a ) (7)

laying between latitudes 12°N and 120°S with an optimum

temperature of between 24 – 38°C, average radiation of 5-7
Fig. 4. Schematic diagram of the Jensen wake model.
Here, power output is calculated using Eq. (3)  
 
1 u  2a 
pi =   Aui
3
= 1−
(3)  2 
(8)
 1 + ax  
2 u0
where pi, η, ρ, A, and u are the power output of wind turbine   r  
  d  
in watt, Betz Constant, air density in kg/m3, wind turbine rotor
swept area in m2, and wind speed in m/s, respectively. The where uω is wake velocity and u0 is free velocity.
Betz constant cannot exceed 0.59. In the Jensen model, wake
Eq. (8) is used for a single wind turbine in the Jensen
radius, rω is calculated using Eq. (4)
wake model. Builtjes et al. [33] researched the calculation of
r = rd +  x (4) the wake effect on the wind turbine farm by defining the ratio
of the average energy output of a wind turbine positioned in a
where rω, rd, α, x is wake radius, wake radius immediately farm and the energy output of an isolated wind turbine as the
behind the turbine, entrainment constant, and downstream formula:
distance, respectively. The calculation of the entrainment
constant (α) reflects the speed of wake expansion given by Eq. P
P ( n) = (9)
(5) P0
0.5 where P(n) is the ratio of the average energy output to the
= (5)
 z isolated single wind turbine of the n-th wind turbine due to the
log   wake effect, P is the average power of the wind turbine in a
 z0  wind farm and P0 is the power of an isolated single wind
turbine. In a simple way, we define the wake value based on
where z, and z0 is hub height and surface roughness considered the differentiation of down spacing from Eq. (9) become the
constant in the current work, respectively. To calculate wake following equation.
radius rd behind the turbine, Eq. (6) is used as follows:

1355
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

 ( n ) = 1 − P ( n) (10)

The Builtjes’ experiments done on the WT farm by the

down spacing of 2D, 5D, 7D, 10D, and 20D with the variation
of degrees in 0°, 5°, 15°, and 30° as shown in Fig. 5.
We redefine Fig. 5 and Eq. (10) to Table 2.
Table 2. The ⍵ value in the various down spacing and the
upwind direction.
Down The direction
P(n) ω = 1 - P(n)
spacing to the upwind
2D 0° 0.15 0.85
2D 5° 0.22 0.78
2D 15° 0.4 0.6
2D 30° 0.95 0.05
5D 0° 0.3 0.7
5D 10° 0.7 0.3
7D 0° 0.35 0.65
7D 10° 0.71 0.29
7D 15° 0.88 0.12
10D 0° 0.57 0.43
10D 5° 0.86 0.14
20D 0° 0.84 0.16 Fig. 5. The wake effect of wind turbine [33].

2.5. The Genetic Algorithm Optimization 3. Results and Discussions

The objective function is to minimize the total wake effect 2.6. Variations of Wind Speed
of the wind turbine farm as defined by Eq. (11)
Knowledge of wind variations at a site is required to
6 10
total =  
m = 2 n −1
( m , n ) x( m , n ) x( m −1, n ) (11)
demonstrate the seasonal behavior of wind speed at a specific
location. Such differences are critical for the design and
selection of suitable wind turbines, energy storage, and
where ωtotal is wake effect of all installed wind turbines, ω(m,n) scenario planning. Figure 6 depicts these variations. Figure 6
is a wake constant for a wind turbine in the m rows and n (a) depicts how the wind fluctuates significantly throughout
columns, m is wind turbine in rows, n is wind turbine in the day and night based on 12 months of data collected at 33
columns, and x(m-1,n) is the existence of wind turbine in the m. There are 10-minute intervals with wind speeds of less than
previous row m and column n. 3 m/s, as seen. Similarly, wind speeds can exceed 10 m/s at
times.
The wake constant refers to Table 2, with only using 2D
and 5D downwind spacing on this study. The wake calculation A box plot representing the monthly maximum wind
starts from the second row of the Wind Turbine Generator speed, average daily high, average daily low, and monthly
(WTG) farm with the assumption that is wake effect on the mean wind speed of the data gathered at altitudes of 34 m is
first row is zero since there is no obstacle from the upwind displayed in Fig. 6 (b). Maximum wind power may be
direction. captured due to the faster speed. The figure clearly shows that
from June to August, the wind speed is higher. The box plot
We used two constraints on this optimization; the first is can also be used to identify outliers for a typical thunderstorm
the limitation of the total capacity of the WTG and the second wind.
is the priority and nonpriority location value based on Fig. 3
(b) for the single wind turbine on the location x(m,n). The matrix Figure 6 (c) shows the average monthly wind speed for
size is (6,10) means 6 rows and 10 columns with the binary Hambapraing. From the graph, it is clear that there are
value 0 and 1 for the location x(m,n). The value of 0 means the significant monthly variations in average wind speed. While
non-existence of WTG and 1 means the existence of a single June through August have the strongest seasonal prevailing
WT for the location x(m,n). winds, October through December have the lowest wind
speeds. Due to the fact that it offers a reliable estimate of the
The wind farm has singular capacities, which means using amount of electricity a wind farm can supply to the power
the same capacities for a single wind turbine. In this study, we system in a given month, this information is essential for wind
proposed 90 kW, 150 kW, 225 kW, and 250 kW for a single farm installation. This research was carried out in
wind turbine. On results, we have four scenarios of a WTG’s Hambapraing, East Sumba Regency. Sumba Island has a
configuration based on its single WTG capacity. semi-arid climate type with a short rainy season that lasts only
for four months (December to March) and a total annual
rainfall of 547 mm.

1356
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

(a) (b) (c)

Fig. 6. Variation of wind speed: (a) time series, (b) monthly mean wind speed, and (c) monthly boxplot statistic.

2.7. Wind Direction wind energy possible [27]. Figure 7 depicts the annual and
monthly sector-wise distribution of wind roses, divided into
The dominant wind direction is crucial when assessing a 16 distinct portions spaced at 22,50 consecutive intervals. It
wind energy resource and looking for ideal locations to put was found that the dominant strong winds for the entire year
wind turbines [34]. If most of the wind energy moves in one come from the southeast sector. The dominant wind direction
direction, there shouldn't be any obstructions or interference is northwest from December to March and southeast for the
in the path of the wind. The wind generator's orientation needs remaining eight months.
to be perpendicular to the wind's direction to capture the most

Fig. 7. Annual and monthly wind rose profile

1357
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

at Hambapraing yielded the Weibull's shape parameter k =

2.8. Wind Diurnal
2.67, the scale parameter c = 7.04 m/s, and the WPD = 225.4
W/m2. Figure 9 depicts the Weibull distribution frequencies
A certain location's wind speed varies throughout the day. for the entire year and monthly at an altitude of 34 m. The
The behavior of wind speed over the course of a year is power produced by the on-site wind turbine can be predicted
examined using the diurnal profile. The wind speed does not using these frequencies. Throughout the year, the most
significantly vary throughout the year; however, it is a little frequent speeds were between 5 and 10 m/s. Strong winds
bit higher during the day than at night (see Fig. 8). The average occur from June to August, with the most frequent speeds
wind speed is lowest at 6 a.m., then increases and peaks appearing to be between 8 and 9 m/s.
around 12 p.m. from January to April and September to
December. In May–August, the average wind speed exceeds Table 3. Weibull's parameters and WPD.
7 m/s throughout the day.
Weibull Parameter WPD
Down spacing
k c (m/s) (W/m2)
2.9. Weibull Frequency Distribution and Wind Power Density
(WPD) Annual 2.67 7.04 225.4
January 1.92 7.05 299.0
Table 3 shows the annual and monthly Weibull's shape
parameters k and scale c, as well as the WPD. The scale February 1.59 5.05 142.4
parameter can be used to indicate wind strength, whereas the March 1.96 5.33 125.9
shape parameter can be used to indicate wind stability and April 2.85 6.01 135.9
distribution at a specific location. WPD is one of the best
May 3.43 8.31 335.4
indicators of wind resources because it considers the
frequency distribution of wind velocity, air density, and the June 5.15 9.61 484.8
cube of wind speed. According to the table, the lowest shape July 5.09 8.69 358.2
parameter during the study was 1.59 in February and the Augustus 4.80 8.55 342.8
highest was 5.15 in June. The lowest value of the scale
parameter, on the other hand, was 4.31 in December, and the September 3.22 6.98 203.1
highest was 9.61 in June. The highest WPD was recorded in October 3.03 5.78 118.0
June, at 484.8 W/m2, while the lowest was recorded in November 2.53 4.92 79.8
December, at 72.0 W/m2. Therefore, the wind potentiality is December 1.84 4.31 72.0
lowest in December, increases until it reaches its peak in June,
and then begins to fall again. The overall 1-year measurement

Fig. 8. Annual and monthly mean diurnal profiles in Hambapraing.

1358
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

Fig. 9. Annual and monthly Weibull frequency distribution.

2.10. Optimization Results in WTGs’ Array 4 has the lowest wake loss of 2.62% and produces a net AEP
of 10,221 MWh/yr.
Figure 10 shows the optimal result of micro-siting WTG
with identical 90 kW, 150 kW, 225 kW, and 250 kW as
Scenario 1, 2, 3, and 4, respectively. Using the software Arc
GIS 10.8 and WAsP version 12.7, Fig. 11, Fig. 12. Fig. 13,
and Fig. 14 show a location map and the layouts of Scenario
1, Scenario 2, Scenario 3, and Scenario 4, respectively, for the
wind turbine. It is based on optimization calculations in the
East Sumba Regency of East Nusa Tenggara Province using
the Universal Transverse Mercator (UTM) system coordinates
for Zone 51. Scenario 1 is a configuration of 34 wind turbines
with a capacity of 90 kW each; Scenario 2 is a configuration
of 20 wind turbines with a capacity of 150 kW each; Scenario
3 is a configuration of 14 wind turbines with a capacity of 225
kW each; and Scenario 4 is a configuration of 12 wind turbines
with a capacity of 250 kW each.
By using WAsP version 12.7, the results of the
calculations for the four scenarios can be seen in Table 4-8.
From Table 4-8, it can be seen that Scenario 1, with a
configuration of 34 WTGs, produces the largest total net AEP
of 11,287 MWh/yr with a wake loss of 3.73%. The wake loss
and total net AEP for Scenarios 2 and 3 are 3.6%, 10,742
MWh/yr, and 2.98%, 11,429 MWh/yr, respectively. Scenario Fig. 10. Configuration of identical WTG in various
scenarios.

1359
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

Fig. 11. Scenario 1. Fig. 14. Scenario 4.

Table 4. Turbine coordinates, wake loss, Gross and Net AEP

for Scenario 1.
Gross Net
Number Wake
X- Y- AEP AEP
of loss
UTM51S UTM51S (MWh/ (MWh/
turbine (% )
yr) yr)
1 190940 8951613 1.3 323 318
2 190832 8951444 1.27 322 318
3 190867 8951778 1.59 341 336
4 190813 8951694 2.32 351 343
5 190760 8951610 2.4 344 335
6 190706 8951525 2.51 345 337
7 190652 8951441 1.78 331 325
8 190741 8951859 3.09 318 308
9 190687 8951774 3.85 350 336
10 190633 8951690 4.1 347 333
11 190579 8951606 3.93 360 346
12 190526 8951522 3.17 356 345
Fig. 12. Scenario 2. 13 190472 8951438 1.88 345 339
14 190561 8951855 5.02 318 302
15 190507 8951771 4.57 361 345
16 190453 8951687 4.57 366 350
17 190399 8951603 4.21 370 355
18 190345 8951518 3 382 371
19 190292 8951434 1.8 368 361
20 190238 8951350 1.28 341 336
21 190184 8951266 1.01 311 307
22 190327 8951768 6.21 323 303
23 190273 8951684 5.22 353 335
24 190219 8951599 4.44 366 350
25 190165 8951515 3.85 362 348
26 190111 8951431 3.33 365 353
27 190057 8951347 2.54 352 343
28 190004 8951262 1.07 324 321
29 190146 8951764 6.71 303 283
30 190093 8951680 5.41 345 327
31 190039 8951596 4.31 379 363
32 189985 8951512 3.7 384 370
Fig. 13. Scenario 3. 33 189931 8951427 3.93 336 323
34 189877 8951343 2.74 316 308

1360
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

Table 5. Turbine coordinates, wake loss, Gross and Net AEP Table 7. Turbine coordinates, wake loss, Gross and Net AEP
for Scenario 2. for Scenario 4.
Gross Net Gross Net
Number Wake Number Wake
X- Y- AEP AEP X- Y- AEP AEP
of loss of loss
UTM51S UTM51S (MWh/ (MWh/ UTM51S UTM51S (MWh/ (MWh/
turbine (% ) turbine (% )
yr) yr) yr) yr)
1 190943 8951613 1.52 519 511 1 190813 8951694 0.58 854 849
2 190817 8951694 2.95 562 546 2 190706 8951525 0.97 843 835
3 190763 8951609 2.41 553 540 3 190633 8951690 3.19 856 828
4 190709 8951525 1.34 557 550 4 190526 8951522 1.85 877 861
5 190636 8951690 6.09 559 525 5 190453 8951687 4.16 903 865
6 190529 8951522 2.03 575 564 6 190345 8951518 1.84 946 928
7 190421 8951353 1.15 520 514 7 190238 8951350 0.64 838 832
8 190564 8951855 4.46 512 489 8 190273 8951683 5.12 869 825
9 190510 8951771 5.38 579 548 9 190165 8951515 3.57 899 867
10 190456 8951687 4.54 589 562 10 190093 8951680 5.79 844 795
11 190349 8951518 3.19 614 595 11 189985 8951512 2.39 940 918
12 190241 8951350 1.81 552 542 12 189931 8951427 1.25 828 818
13 190061 8951347 1.65 568 558
14 190007 8951262 0.34 525 523
15 190168 8951515 4.47 585 559 Table 8. The proportional wake loss and Annual Energy
16 190276 8951683 4.91 570 542 Production for various scenarios.
17 190330 8951768 6.88 524 488 Sce Numb Turbi Total Proport Total Total
18 189988 8951512 3.41 618 597 nari er of ne Rating ional gross net
19 190096 8951680 5.4 559 529 o Wind Ratin Capacit wake AEP AEP
20 190149 8951764 8.29 492 451 Turbi g y loss (MW (MWh/
ne (kW) (kW) (% ) h/yr) yr)
Table 6. Turbine coordinates, wake loss, Gross and Net AEP 1 34 90 3060 3.73 11,681 11,287
for Scenario 3. 2 20 150 3000 3.60 11,133 10,732
3 14 225 3150 2.98 11,780 11,429
Gross Net 4 12 250 3000 2.62 10,496 10,221
Number Wake
X- Y- AEP AEP
of loss
UTM51S UTM51S (MWh/ (MWh/
turbine (% ) generators (WTG) with a unit capacity of 90 kW in Scenario
yr) yr)
1. Scenario 2 includes 20 identical wind turbine generators
1 190813 8951694 0.71 836 830
with a total capacity of 3000 kW. In Scenario 3, 14 identical
2 190706 8951525 1.15 825 816
WTG with 225 kW of unit capacity are used. In Scenario 4,
3 190598 8951357 1.19 759 750
there are 12 identical WTGs with a unit capacity of 250 kW.
4 190633 8951690 2.89 832 808
The results showed that Scenario 1 produced the highest total
5 190526 8951522 3.15 853 826
net AEP of 11,287 MWh/year with a 3.73% wake loss. The
6 190418 8951354 2.11 779 763
minimum wake loss seemed to be 2.62% in Scenario 4, with a
7 190453 8951687 4.06 874 838
total net AEP of 10,221 MWh/year.
8 190345 8951518 3.63 912 878
9 190238 8951350 2.06 819 802 The findings of this study are one of the solutions to the
10 190273 8951683 5.15 847 803 proposed wind farm project in the same location by Win Rock
11 190165 8951515 4.22 868 832 International and NREL in 2015, which was delayed owing to
12 190058 8951347 1.62 842 828 the state of infrastructure and land transportation on the island.
13 190093 8951680 5.59 827 781 Because of these issues, we limited the wind turbine capacity
14 189985 8951512 3.74 909 875 to less than 500 kW.
This single-objective optimization has drawbacks in that
it does not simultaneously investigate minimizing the total
4. Conclusion
wake effect of the wind turbine farm and minimizing energy
production costs. The investment cost includes not only the
In this work, an optimization method for minimizing the infrastructure of the WT towers and the cost of the land
total wake effect of a wind turbine farm on the island of occupied but also the cost of connecting and maintaining grid
Sumba, Indonesia, using a genetic algorithm is developed and stability between the WT generators and the load distributions.
simulated with WaSP software to estimate the AEP. The wind
For future studies, we will investigate multi-objective
farm layout and preliminary turbine micrositing were optimization with a focus on the maximization of annual
completed with various scenarios in mind to achieve the best energy production and the simultaneous minimization of
possible result. Four different scenarios are considered to investment costs.
maximize power output. There are 34 identical wind turbine

1361
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

References Renewable Resources: A Case Study of East Sumba,

East Nusa Tenggara, Indonesia,” Bisnis Birokrasi J., vol.
[1] M. A. Saeed, Z. Ahmed, S. Hussain, and W. Zhang, 26, no. 1, May 2019.
“Wind resource assessment and economic analysis for
[12] M.C. Brower, B.H. Bailey, P. Beaucage, D.W.
wind energy development in Pakistan,” Sustain. Energy
Bernadett, J. Doane, M.J. Eberhard, K. V. Elsholz, M. V.
Technol. Assessments, vol. 44, p. 101068, April 2021.
Filippelli, E. Hale, M.J. Markus, D. Ryan, M.A. Taylor,
[2] X. Y. Tang, S. Zhao, B. Fan, J. Peinke, and B. and J.C. Tensen, “Wind Resource Assessment: A
Stoevesandt, “Micro-scale wind resource assessment in Practical Guide to Developing a Wind Project,” John
complex terrain based on CFD coupled measurement Wiley & Sons, Ltd, May 2012.
from multiple masts,” Appl. Energy, vol. 238, pp. 806–
[13] E. P. P. Soares-Ramos, L. De Oliveira-Assis, R. Sarrias-
815, March 2019.
Mena, P. Garcia-Trivino, C. A. Garcia-Vazquez, and L.
[3] C. Jung and D. Schindler, “Introducing a new approach M. Fernandez-Ramirez, “Control of a Grid-connected
for wind energy potential assessment under climate Wind Turbine with Quasi-Z-Source Inverter,” in 9th
change at the wind turbine scale,” Energy Convers. International Conference on Renewable Energy
Manag., vol. 225, p. 113425, December 2020. Research and Applications, pp. 309–314, September
2020.
[4] N. W. Hesty, D. G. Cendrawati, R. Nepal, and M. I. Al
Irsyad, “Wind Energy Potential Assessment Based-on [14] S. Mathew, Wind energy: Fundamentals, resource
WRF Four-Dimensional Data Assimilation System and analysis and economics. New York: Springer-Verlag,
Cross-Calibrated Multi-Platform Dataset,” IOP Conf. March 2006.
Ser. Earth Environ. Sci., vol. 897, no. 1, pp. 0–7,
[15] J. F. Manwell, J. G. McGowan, and A. L. Rogers, Wind
November 2021.
Energy Explained: Theory, Design and Application,
[5] B. Pranoto, H. Soekarno, D.G. Cendrawati, I.F. Akrom, Second Edition, John Wiley & Sons, Ltd, December.
M.I.A. Irsyad, N.W. Hesty, Aminuddin, I. Adilla, L. 2009.
Putriyana, A.F. Ladiba, Widhiatmaka, R. Darmawan,
[16] M. Djamai and N. Kasbadji Merzouk, “Wind farm
S.R. Fithri, R. Isdiyanto, V.J. Wargadalam, M.
feasibility study and site selection in Adrar, Algeria,”
Magdalena, and M. Aman, “Indonesian hydro energy
Energy Procedia, vol. 6, pp. 136–142, April 2011.
potential map with run-off river system,” IOP Conf. Ser.
Earth Environ. Sci., vol. 926, no. 1, November 2021. [17] R.G.J. Flay, A.B. King, M. Revell, P. Carpenter, R.
Turner, P. Cenek, and A.A. Safaei Pirooz, “Wind speed
[6] V. Nurliyanti, K. Ahadi, R. Muttaqin, B. Pranoto, G. P.
measurements and predictions over Belmont Hill,
Srikandi, and M. I. Al Irsyad, “Fostering Rooftop Solar
Wellington, New Zealand,” J. Wind Eng. Ind. Aerodyn.,
PV Investments Toward Smart Cities through e-SMART
vol. 195, p. 104018, December 2019.
PV,” 5th Int. Conf. Smart Grid Smart Cities, pp. 146–
150, June 2021. [18] U. Yilmaz, F. Balo, and L. S. Sua, “Simulation
Framework for Wind Energy Attributes with WAsP,”
[7] Winrock International, “Fuel Indenpendent Renewable
Procedia Comput. Sci., vol. 158, pp. 458–465, June
Energy ‘Iconic Island’ Preliminary Resource
2019.
Assessment Sumba & Buru Islands - Indonesia,” 2010.
[Online]. Available: [19] I. Kamdar, S. Ali, J. Taweekun, and H. M. Ali, “Wind
https://www.bibalex.org/Search4Dev/document/437343 farm site selection using WAsP tool for application in the
(accessed Mar. 9, 2023) tropical region,” Sustain., vol. 13, no. 24, December
2021.
[8] I. Pramuwardani, Hartono, Sunarto, and A.
Sopaheluwakan, “Indonesian rainfall variability during [20] P. Ratjiranukool and S. Ratjiranukool, “Evaluating Wind
Western North Pacific and Australian monsoon phase Speed by WRF Model over Northern Thailand,” Energy
related to convectively coupled equatorial waves,” Arab. Procedia, vol. 138, pp. 1171–1176, October 2017.
J. Geosci., vol. 11, no. 21, November 2018.
[21] Z. Wang and W. Liu, “Wind energy potential assessment
[9] I. Alifdini, T. Shimada, and A. Wirasatriya, “Seasonal based on wind speed, its direction and power data,” Sci.
distribution and variability of surface winds in the Rep., vol. 11, no. 1, pp. 1–15, August 2021.
Indonesian seas using scatterometer and reanalysis
[22] S. D. Kwon, “Uncertainty analysis of wind energy
data,” Int. J. Climatol., vol. 41, no. 10, pp. 4825–4843,
potential assessment,” Appl. Energy, vol. 87, no. 3, pp.
April 2021.
856–865, March 2010.
[10] R. Fisher, W. E. Bobanuba, A. Rawambaku, G. J. E. Hill,
[23] A. Geçmez and Ç. Gençer, “Wind Energy Production
and J. Russell-Smith, “Remote sensing of fire regimes in
Estimation with ANN and ANFIS,” 9th International
semi-arid Nusa Tenggara Timur, eastern Indonesia:
Conference on Smart Grid, October 2021.
Current patterns, future prospects,” Int. J. Wildl. Fire,
vol. 15, no. 3, pp. 307–317, January 2006. [24] O. Ozturk, B. Hangun, and M. Shoaeinaeini, “Utilizing
Machine Learning to Predict Offshore Wind Farm Power
[11] B. Yulianto, S. Maarif, C. Wijaya, and H.
Output for European Countries,” 11th International
Hardjomidjojo, “Energy Security Scenario based on

1362
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
I. Ifanda et al., Vol.13, No.3, September, 2023

Conference on Renewable Energy Research and [30] P. Tiam Kapen, M. Jeutho Gouajio, and D. Yemélé,
Application, October 2022. “Analysis and efficient comparison of ten numerical
methods in estimating Weibull parameters for wind
[25] F. H. Mahmood, A. K. Resen, and A. B. Khamees,
energy potential: Application to the city of Bafoussam,
“Wind characteristic analysis based on Weibull
Cameroon,” Renew. Energy, vol. 159, pp. 1188–1198,
distribution of Al-Salman site, Iraq,” Energy Reports,
October 2020.
vol. 6, pp. 79–87, February 2020.
[31] E. Rodrigo and A. Larico, “Wind Energy Potential by the
[26] F. Azlan, J. C. Kurnia, B. T. Tan, and M. Z. Ismadi,
Weibull Distribution at High-Altitude Peruvian
“Review on optimisation methods of wind farm array
Highlands,” Int. J. Smart grid, vol. 5, no. 3, September
under three classical wind condition problems,” Renew.
2021.
Sustain. Energy Rev., vol. 135, p. 110047, January 2021.
[32] M. Bin Ali, Z. Ahmad, S. Alshahrani, M. R. Younis, I.
[27] H. S. Ramadan, “Wind energy farm sizing and resource
Talib, and M. Imran, “A Case Study: Layout
assessment for optimal energy yield in Sinai Peninsula,
Optimization of Three Gorges Wind Farm Pakistan,
Egypt,” J. Clean. Prod., vol. 161, pp. 1283–1293,
Using Genetic Algorithm,” Sustain., vol. 14, no. 24,
September 2017.
December 2022.
[28] D. Romanic, D. Parvu, M. Refan, and H. Hangan, “Wind
[33] P. J. H. Builtjes and J. Smit., “Calculation of Wake
and tornado climatologies and wind resource modelling
Effects in Wind Turbine Parks,” in Wind Engineering,
for a modern development situated in ‘Tornado Alley,’”
vol. 2, no. 3, Sage Publications, Ltd., 1978, pp. 135–145.
Renew. Energy, vol. 115, pp. 97–112, January 2018.
[34] V. Sharma, M. Calaf, M. Lehning, and M. B. Parlange,
[29] A. Hossieni, V. Rasouli, and S. Rasouli, “Wind energy
“Time-adaptive wind turbine model for an LES
potential assessment in order to produce electrical energy
framework,” Wind Energy, vol. 19, no. 5, pp. 939–952,
for case study in Divandareh, Iran,” in 3rd International
May 2016.
Conference on Renewable Energy Research and
Applications, pp. 133–137, October 2014.

1363