Renewable Energy 87 (2016) 403e414

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Wind resource assessment for urban renewable energy application in

Singapore
*
B.R. Karthikeya a, Prabal S. Negi b, N. Srikanth a
a
Energy Research Institute @ Nanyang Technological University, #06-04, 1 Cleantech Loop, Cleantech One, Singapore 637141, Singapore
b
School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: In highly urbanized and energy intensive countries like Singapore all possible avenues for power gen-
Received 10 February 2015 eration need attention. In this context, rooftop installations of both solar and wind energy are of
Received in revised form particular interest for Singapore, especially because of Singapore's condition of land limitation. Decen-
5 August 2015
tralized and distributed energy sources such as rooftop wind and solar installations have numerous
Accepted 7 October 2015
Available online 10 November 2015
advantages. However, the potential for wind energy is not fully understood in built-up areas and thus not
fully exploited. Hence it is important to study wind flow patterns in built-up areas and also develop
technologies tuned for these conditions. The demand for technologies that deliver energy for low flow
Keywords:
Urban wind potential
wind conditions is of paramount importance to Southeast Asia region and especially to Singapore. In this
Wind energy paper, two measurement systems, namely stationary rooftop wind mast and mobile Light Detection and
Wind speed Ranging (LiDAR) profiler, have been discussed. Measured wind data from various sites across Singapore
Urban wind turbines using have also been presented. Wind roses, Weibull distribution, roughness lengths and other statistical
Weibull distribution analyses were carried out to understand the prevailing wind characteristic, which is used for evolving the
Resource assessment basic criteria for economic viability of roof top wind turbines in the tropical conditions of Singapore.
© 2015 Published by Elsevier Ltd.

1. Introduction
Singapore's installed electricity generation capacity is 12.5 GW,
which is almost entirely derived from fossil fuels [1]. Household
electricity consumption is 15% of the total electricity consumption
and consumption by commercial and services sector is 37%.
Singapore is committed to reducing its emissions by 7%e11% below
2020 business-as-usual levels [2]. Hence, decentralized and
distributed energy sources such as rooftop wind and solar in-
stallations require due attention in terms of resource estimation
and techno-economic evaluation. Based on the space availability in
Singapore, maximum cumulative capacity of rooftop photovoltaic
installations is estimated to be 5 GWp by 2030, with 80% of the
installed capacity on rooftops and facades [3]. However, the po-
tential for wind energy is not fully understood in built-up areas and
hence not fully estimated. In order for Singapore to achieve this
goal and to diversify the energy mix, several government agencies
are working with Institutes of Higher Learning and local SMEs.
National Environment Agency (NEA) owns and maintains several
* Corresponding author.
E-mail address: NSrikanth@ntu.edu.sg (N. Srikanth).
met masts that measure the surface wind across the island nation.
But this data is inadequate to estimate the wind potential for the
installation of rooftop wind turbines in densely urbanized
Singapore. Housing Development Board (HDB) of Singapore, a
government body responsible for public housing in Singapore, has
been working with the Energy Research Institute at Nanyang
Technological University (ERI@N) to study the feasibility of rooftop
wind turbines in Singapore. Hence, ERI@N has developed two wind
measurement systems for Singapore, namely remote sensing and
mobile LiDAR measurement system and rooftop standard
anemometry measurement system with wireless data transfer ca-
pabilities. ERI@N is the first and so far the only institute in
Singapore that conducts wind measurement campaigns in the re-
gion using SODAR (SOnic Detection and Ranging) and LiDAR wind
profilers.
Singapore has a wet equatorial climate with fairly uniform mean
monthly temperatures between 26
C and 28
C throughout the
year [4]. Singapore's weather can be classified into four seasons.
These four seasons are the Northeast Monsoon (December to early
March), the Southwest Monsoon (June to September) and two
relatively short inter-monsoon seasons. Singapore experiences
light and variable winds during the transition between these

http://dx.doi.org/10.1016/j.renene.2015.10.010
0960-1481/© 2015 Published by Elsevier Ltd.
404 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414
seasons [5]. Wind power is affected by the monsoon weather
pattern [6]. Singapore's weather is characterized by high humidity
and frequent rainfall [7]. Singapore receives a considerable amount
of rainfall throughout the year. The mean monthly rainfall can be
between 150 mm and 275 mm depending on the season [4].
Fig. 1 indicates the locations where ERI@N either has permanent
masts for long term wind measurements or has undertaken LiDAR
campaigns for short-term measurements. Green ‘rhombus’
markers indicate the four HDB rooftops where wind measurement
masts have been installed. In this paper, wind measurements of
over two years (2012e2014) from the three sites viz. Woodlands
Crescent, Pandan Gardens and Marine Drive have been presented
and discussed. Results of short-term LiDAR campaigns undertaken
at Sentosa and SSC Tanah Merah have been presented, especially to
understand the roughness lengths in various direction sectors and
turbulence intensity variation with increasing wind speeds. A
stripped down techno-economic analysis has been presented for
the case of small wind turbine installation.
2. Urban wind energy potential
Urban areas include considerable turbulence and local aberra-
tions. Urban wind energy problem has two main aspects: One,
understanding of flow pattern within the lowest urban canopy
layer where individual building affects the flow; Two, vertical
extrapolation of wind in roughness and inertial sublayers. Wind
characteristics such as surface drag, vertical shear profile and tur-
bulence intensity are affected by the roughness lengths of the ur-
ban canopy. Estimation of spatially averaged wind profile for urban
Fig. 1. ERI@N’s wind measurement sites in Singapore.
surface sublayer and roughness sublayer has been discussed in
Ref. [8]. In Ref. [9] Ishugah et al. discuss three methods to estimate
surface roughness in urban areas viz. Davenport classification,
morphometric and meteorological methods. Giovanni and Sauro
[10] have discussed in detail the effect of roughness length and
wind shear coefficients on the Annual Energy Yield (AEP) for three
coastal sites in Southern Italy. Hee-Chang and Tae-Yoon [11] discuss
the wind resource assessment of a wind energy site in Jeju Island
and spectra analysis of wind data.
There can be significant increase of wind speed at specific lo-
cations due to concentration effect of buildings [12]. Hence, esti-
mation of wind parameters is more complicated in built up areas.
Sara Louise Walker [13] has discussed in detail various methods of
estimating urban wind resource. Wind potential in urban areas can
be evaluated using standard anemometry at the site, computational
fluid dynamics (CFD) simulation and wind tunnel experiments on
the physical model of the building and the surrounding area [14]. A
combination of two of these methods is employed to verify the
results. A CFD simulation of wind flow around simple building and
subsequent wind tunnel measurements to verify CFD simulation
have been discussed in Ref. [15]. A method describing a combina-
tion of CFD and measurements from a single anemometer for wind
energy assessment has been discussed in Ref. [16]. Subsequent to a
thorough understanding of the flow patterns, various wind energy
systems suitable for exploiting wind potential in that specific urban
environment can be considered [17].
D. Elliott and D. Infield [18] have studied the effect of averaging
time wind measurements on turbine energy capture prediction and
turbulence intensity calculation. The AEP for low wind speed
 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414 405

Table 1
Wind measurement locations.

Location Instruments Elevation (m)

Woodlands Crescent WindSonic 2-axis anemometer, cup anemometer, wind vane 64.2
Pandan Gardens WindMaster 3-axis anemometer, cup anemometer, wind vane 78
Marine Drive WindSonic 2-axis anemometer 83
Sentosa Remote sensing LiDAR NA
SSC Tanah Merah Remote sensing LiDAR NA

regimes does not vary more than 1e2% for different averaging vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u P 3
periods. However the calculated turbulence intensity changes u Vi
u
significantly with the averaging time period [18]. We've chosen 10- c¼u N
3
u
 (7)
min average wind data has been used in this paper. Referring to t
NG 3k þ 1
various sources [19e21], a brief description of meteorological and
technical nomenclature applicable for this paper is as follows.
Weibull distribution is the probability distribution function used Turbulence intensity is defined as
to describe the distribution of wind speeds over time duration.
TI ¼ s=Vav (8)
Weibull k (shape factor) parameter controls the width of the dis-
tribution and Weibull c (scale factor in m/s) parameter controls the
where s ¼ 10-min standard deviation, Vav ¼ 10-min average wind
average wind speed. The two-parameter Weibull probability
speed.
function is given by
Wind speed variation with height above ground is given by
logarithmic profile and power law profile. As per logarithmic pro-


k
k V k1 V file, wind speed at height z1 is given by
PðV; c; kÞ ¼  exp (1)
c c c
ln zz10
vðz1 Þ ¼ vðz2 Þ (9)
Cumulative distribution function is given by ln zz20

    where z0 ¼ roughness length, z2 ¼ reference height, v(z2) ¼ wind

vo k
pðv  vo Þ ¼ 1  exp  (2) speed at the reference height.
c
Power extracted by a wind turbine from the incident wind can
There are various methods to estimate the Weibull parameters. be expressed as
In this paper, we have considered maximum likelihood, least
squares and WAsP methods. Maximum likelihood method is the P ¼ 0:5CP rV3av A (10)
most widely used method. Shape factor k according to maximum
likelihood method is given by Ref. [22]. where A ¼ rotor area, CP ¼ power coefficient of the rotor,
Vav ¼ wind speed and r ¼ air density [25].
2 31 AEP is given in Watt hour and over one year is calculated using
P
6 n vk ln v 7 the Weibull distribution and power curve of the wind turbine.
6 i¼1 i i 1Xn
7
k¼6 P  ln vi 7 (3)
4 n
vk n 5 Z∞
i¼1 i i¼1
AEP ¼ 8; 760 PðV; c; kÞPðVÞdv (11)
0
Scale factor c is given by
3. Wind measurement system
" #1=
1X n k
Table 1 shows the locations where wind parameters were
c¼ vk (4)
n i¼1 i measured and the set of instruments used for taking these mea-
surements. There are three stationary wind measurement in-
In least squares method, the slope and intercept of the line of stallations for long term measurements and two locations where
best fit given by the equation below are calculated [23]. the mobile LiDAR unit was deployed for short term measurements.
Locations chosen for stationary measurement installation are on
 
1 the rooftops of the tallest building in their respective vicinity.
ln ln ¼ k ln V  k ln c (5)
½1  pðvÞ
3.1. Stationary installation: rooftop anemometry
WAsP algorithm was originally described by Troen and Petersen.
According to [24], WAsP algorithm calculates c and k to fit the
IEC61400 e Part 2 provides some general guidelines regarding
power density in the time series instead of the mean wind speed. k
the locations suitable for the installation of small wind turbines
is given by the equation
under inclined flow and turbulent wind conditions on rooftops
[20]. Same guidelines, depicted in Fig. 2, have been adopted for the

k
1 placement of anemometers on rooftops.
lnX ¼ G þ1 (6)
k Fig. 3 shows the measurement mast set up on one of the roof-
tops. Many commercial wind measurement systems are available.
where X represents the proportion of the wind speeds that exceed But these systems are less flexible in terms of the data transmission
mean wind speed. And c is given by and types of sensors that can be installed.
406 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414

Fig. 2. General guidelines for the placement of anemometers on the rooftop to overcome turbulence conditions and inclined flow.
Source: [20].

Fig. 4 shows the process flow diagram of the cost-effective and horizontal extrapolation and significantly reduce uncertainties
measurement setup that was designed. Measured parameters are in wind speed and energy yield prediction. For a site in the
wind speed and direction. Different types of wind measurement Netherlands, uncertainty in terms of wind speed was reduced by up
instruments deployed are - wind vane, cup anemometers, 3-axis to 4% and uncertainty of energy yield reduced by up to 7% [30].
sonic anemometers, 2-axis sonic anemometers. Cup anemome- Continuous wave, dual mode LiDAR profiler ZephIR 300 has been
ters convert the rate of rotation into wind speed whereas sonic used for wind measurement campaigns at multiple locations to get
anemometers use ultrasonic pulses to measure the wind speed and
direction. 2-axis anemometers measure ‘U’ and ‘V’ horizontal
components of the wind and 3-axis anemometers measure an
additional vertical wind vector ‘W’. Hence, 3-axis sonic anemom-
eters have the ability to provide detailed information on the tur-
bulence of the wind. Sonic anemometers are designated as primary
sensors and cup anemometers, wherever used, are designated as
secondary sensors. Missing or invalid data points from sonic ane-
mometers are substituted by measurement data from cup ane-
mometers [26]. From the measured 4 Hz data, 10 min averaged
samples were derived. As documented in the calibration certifica-
tion, the NRG cup anemometers were factory calibrated with
maximum deviation of 0.035 m/s at 14e15 m/s. This complies
with the MEASNET requirement for anemometers of absolute un-
certainty less than 0.1 m/s at a mean wind velocity of 10 m/s [27]. 3-
axis sonic anemometers and 2-axis sonic anemometers have wind
speed accuracy of ±2% and direction accuracy of ±20 at 12 m/s. No
onsite calibration was carried out for any of the sensors.
On building rooftops, the height of wind mast is limited to 4 m
due to civil aviation and other regulatory requirements. Datalogger
has integrated 3G modem that transmits 10-min average data to
the FTP server. Using Python and Matlab scripts, FTP data is
retrieved, stored locally and processed further. From Equation (3) it
can be seen that air density has direct effect on the power gener-
ated by a wind turbine. However, for this study measurement masts
have no mounted pressure and temperature measurement sensors.

3.2. Roving installation: LiDAR unit

LiDAR profiles work based on detection and processing of the

Doppler shift of backscattered laser beam due to aerosols in the
wind [28]. A LiDAR profiler is not a replacement for wind masts due Fig. 3. Measurement setup at Woodlands Crescent showing cup anemometer, 2-axis
to the costs involved [29]. But short term LiDAR campaigns signif- sonic anemometer and wind vane. DataTaker with integrated 3G modem is power
icantly improve wind statistics, long term representation, vertical by solar panels.
 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414 407

Fig. 4. Process flow diagram from measurement to reporting for rooftop wind measurement installations.

holistic wind measurement data including turbulence intensity and
wind shear profile. The LiDAR profiler can measure wind charac-
teristics at ten user defined heights between 10 and 200 m at the
sampling rate of 50 Hz [31]. Fig. 5 shows the mobile remote wind
sensing system implemented with ZephIR 300. Gill 3-axis sonic
anemometer was deployed on a mobile mast of height between 1
and 10 m. A Campbell Scientific CR800 datalogger is connected to
ZephIR 300 and 3-axis anemometer, which transmits data over
modem. This mobile sensing unit was deployed for short term
measurements at several locations of interest to gain the complete
understanding of prevailing wind characteristics. Wherever
possible IEA recommended practices [32] for deploying and

Fig. 5. Remote sensing system using LiDAR profiler ZephIR 300 and Campbell Scientific CR800 datalogger.
408 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414

Fig. 6. A plot of daily mean wind speed at Marine Drive showing two instances of outage. Due to lack of a secondary sensor, missing data from the primary sensor could not be filled
for this specific site.

operating the LiDAR profiler and data analysis have been followed.
The LiDAR profiler was calibrated by the manufacturer and the
calibration accuracy was validated against the measurements from
a met mast. Calibration uncertainty of less than 0.1% and overall
variation in calibration of less than ±0.5% is promised [33].
4. Results
Our experience shows that 3-axis sonic anemometers are more
susceptible to lightning strikes despite proper lightning protection
schemes. There was some data loss due to outage of power supply

Fig. 7. Diurnal profile variation of mean wind speed and wind direction by month at Marine Drive. A diurnal variation of mean wind speed is weak in the month of August.
 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414 409

Fig. 8. Wind direction frequency distribution and Weibull distribution for wind speeds at Marine Drive.

and anemometer faults. Data recovery is between 75 and 90% for for two locations without any significant disruptions. However, at
various sensors at different locations (Fig. 6). At Woodlands Cres- Marine Drive no secondary sensor was installed during the period
cent and Pandan Gardens, cup anemometers were installed as of measuring considered for this paper.
secondary sensors. Due to secondary sensors, wind data is available Southern shore of Singapore experiences better wind speeds

Fig. 9. Wind direction frequency distribution and Weibull distribution for wind speeds at Pandan Gardens.

Fig. 10. Wind direction frequency distribution and Weibull distribution for wind speeds at Woodlands Crescent.
410 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414

Fig. 11. Monthly average wind speeds measured from all the sensors at Marine Drive, Woodlands Crescent and Pandan Gardens.

due to open sea to the South. Hence LiDAR campaigns are mostly
restricted to the southern shore of Singapore as can be seen in Fig. 1.
Fig. 7 shows the monthly diurnal variation of mean wind speed and
wind direction. Figs. 8e10 shows the Weibull distribution for the
three measurement sites. Fig. 11 indicates the monthly variable of
average wind speeds for all the sensors, separately, at all the three
locations.
From the wind measurements from stationary installations it
was observed that southern coast of Singapore is better in terms of
wind resources. So, LiDAR campaigns were undertaken at three
locations - Tuas View Extension, SSC Tanah Merah and Sentosa, all
located on the southern coast of Singapore. A thorough under-
standing of the Turbulence intensity of wind is important to un-
derstand the loadings on the turbine and hence also important for
the optimum turbine design. The Normal Turbulence Model (NTM),
defined in IEC61400-2, does not accurately estimate the turbulence
in urban environments [34]. Figs. 13 and 14 are results from the
LiDAR campaigns undertaken at these locations. Fig. 13 shows the
variation of mean turbulence intensity, for different heights, ac-
cording to varying wind speeds. Fig. 14 shows the 16-sector
roughness roses for the three coastal locations.
wind speeds are observed in the month of January at all the
locations. However, during Southwest Monsoon from June to
September, Site B experiences high wind speeds in June and
other two locations in the month of August.
 Turbulence intensities as the wind speed approaches 15 m/s are
0.15 at Site C and 0.25 at Site A. For Site B, there are not enough
data points for wind speeds higher than 11 m/s. As can be seen
in the Fig. 12, the turbulence intensity decreases with increasing
wind speeds.
 The Turbulence Intensity versus the wind speeds at different
heights is depicted in Fig. 13. It can be seen that the turbulence
intensity grows in magnitude after 8 m/s. The turbulence in-
tensity is fairly uniform for wind speeds between 2 m/s and 8 m/
s.
 Fig. 14 shows the 16-sector roughness rose obtain from the
LiDAR measurements. Vertical shear profile of the wind in-
dicates substantial surface roughness lengths at Sentosa and
surrounding areas near the central business district. To the
north of Sentosa island there is Singapore's business district

5. Discussion

For the three sites where long term wind measurements were
taken, Table 2 summarises the Weibull parameters estimated based
on three algorithms viz. maximum likelihood method, least squares
method and WAsP method. From the assessment of measured two-
year wind data for the three sites, originally chosen as prospective
sites for small wind turbine installation, following outcomes can be
listed out:

 Maximum wind power density at Site A, Site B and Site C are 45,
35 and 15 W/m2 respectively.
 The highest average diurnal wind speeds are observed between
3 pm and 4 pm.
 Based on monthly distribution of mean wind speed, two high
wind seasons are observed. Even though Northeast Monsoon Fig. 12. Turbulence intensity at Marine Drive, Woodlands Crescent and Pandan
extends from December to March, as observed in Fig. 11, highest Gardens.
 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414 411

Fig. 13. Turbulence Intensity and wind speed distribution at three heights (50 m, 80 m, and 100 m) from measurements at SSC Tanah Merah using LiDAR device. For low wind
speeds, from 1 m/s to 8 m/s, the turbulence intensity gradually decreases.

with tall skyscrapers and to the south there are small islands.
This explains significantly higher surface roughness lengths at
Sentosa. The highest roughness length at Sentosa is 11.6 m in the
direction sector 146.25
e168.75
.
 Changi airport lies to the north of SSC Tanah Merah, which is the
reason why roughness lengths at SSC Tanah Merah are signifi-
cantly lower than that of Sentosa. The highest roughness length
at SSC Tanah Merah is 0.26 m in the direction sector
78.75
e101.25
.
 The roughness lengths at Tuas View Extension are 0.23 m in the
direction sector 168.75
e191.25
and 0.127 m in the direction
sector 348.75
e11.25
. However, roughness lengths in other
direction sectors are found to be close to zero. Hence, Tuas View
Extension is found to be the best location out of the three
locations.
 There is significant diurnal variability and seasonal variability
that has to be taken into account to estimate wind energy
potential.
5.1. Techno-economic analysis
The installed costs per kW for small wind turbines are highly
variable depending on the technology, country of origin and the
market place conditions. The installed cost per kW for Skystream
3.7 is $7917 and annual operation and maintenance cost is esti-
mated at $0.01/kWh [35]. In the USA, the cost per installed kW is
between $2300 and $10,000. However, in China the cost per
installed kW is as low as 12,000 Yuan (US $1900) [36]. The indus-
try's cost target is to bring the cost per installed kW to
412 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414

Fig. 14. 16-sector roughness rose derived from measured vertical shear profiles at Tuas View Extension, Sentosa and SSC Tanah Merah using the LiDAR device. Axis is in logarithmic
scale and six circles represent roughness length ZO values from 0.0001 m to 100 m.

$3000e$4500 [37]. Table 3 shows the basic tabulation to calculate
some key techno-economic parameters [38]. It is difficult to un-
derstand the direct correlation between the cost of electricity,
payback period and the turbine price, with too many variables at
play. Hence, this calculation assumes no grant, incentive or loan.
Only Site A and Site C are considered for economic analysis. The
table indicates Annual Energy Production (AEP) and payback pe-
riods for a 2.5 kW turbine installed at Site A and Site C respectively.
For various commercially available turbines, calculated Net Ca-
pacity Factor (NCF) is between 2 and 12% at Site C and between 3
and 15% at Site A. Due to better wind conditions, capacity factor at
Site A is would be 25% higher than that of Site C. NCF is assumed to
be 8.75% and 7% for Site A and Site C respectively. The cost of
electricity in Singapore is directly dependent on the cost of im-
ported natural gas. In the past 5 years, electricity tariff in Singapore
has fluctuated between 0.25 and 0.30 S$/kWh [39]. Assuming
0.35 S$/kWh for exported electricity, expected payback period for
Site A is 14 years and expected payback period for Site C is 19 years,
longer than the payback period for Site C, due to relatively poor
capacity factor. At both locations, payback period has to be less than

Table 2
Summary table.

Algorithm Weibull k (shape factor) Weibull c (scale factor in m/s) Mean (m/s) Proportion above [Mean] m/s

Pandan Gardens (Site A) Maximum likelihood 2.037 3.835 3.398 0.457

Least squares 2.005 3.846 3.408 0.458
WAsP 2.086 3.868 3.426 0.466
Actual data 3.4 0.466

Woodlands Crescent (Site B) Maximum likelihood 1.634 2.267 2.029 0.435

Least squares 1.689 2.252 2.01 0.433
WAsP 1.615 2.258 2.023 0.432
Actual data 2.026 0.432

Marine Drive (Site C) Maximum likelihood 2.312 3.793 3.361 0.468

Least squares 2.174 3.83 3.391 0.47
WAsP 2.489 3.854 3.419 0.489
Actual data 3.367 0.489
 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414 413

Table 3
Payback period and AEP at Site A and Site C.

# Parameters Formula Site A Site C Unit

a) Estimating the value of generated electricity

1 Turbine capacity (TC) TC 2.5 2.5 kW
2 Net capacity factor (CF) CF 8.75 7 %
3 Annual energy production (Y) Y ¼ TC$CF$8760 h/year 1916 1533 kWh/year
4 Export fraction (F) F 100 100 %
5 Electricity exported (Eexp) Eexp ¼ Y $ F 1916 1533 kWh/year
6 Electricity used directly (Edir) Edir ¼ Y $ (1eF) 0 0 kWh/year
7 Unit price of exported electricity (Pexp) Pexp 0.35 0.35 S$/kWh
8 Unit cost of imported grid electricity (Pimp) Pimp 0.25 0.25 S$/kWh
9 Value of exported electricity (Vexp) Vexp ¼ Eexp $ Pexp 670.7 536.6 S$
10 Value of electricity used directly (Vdir) Vdir ¼ Edir $ Pimp 0 0 S$
11 Total value of electricity (Vtot) Vtot ¼ Vexp þ Vdir 670.7 536.6 S$/year
b) Estimating the simple payback
1 Total capital cost of turbine installation (CC) CC 7500 7500 S$
2 Annual operating cost (Co) Co 150 150 S$/year
3 Simple payback period CC/(Vtot  CO) 14.4 19.3 years

20 years which is normally the life time of a wind turbine. This gives
the precondition on the maximum cost per installed kW, which is
S$ 3000.
6. Conclusions
In this paper the challenges and prior research on urban wind
measurements and wind resource assessment have been exam-
ined. Two measurement systems and their implementation have
been discussed-rooftop wind measurement systems and data
integration for urban canopy and remote sensing, mobile wind
profiler based on a LiDAR device. Key results of several wind
measurement campaigns in Singapore are presented. Wind mea-
surements from the three rooftop wind measurement installations
have been discussed. Also, Turbulence Intensity and roughness rose
derived from LiDAR campaigns have been presented. It is found that
the southern shore of Singapore is ideal for wind turbines. Wind
power densities at Pandan Gardens and Marine Drive are 45 and
35 W/m2 respectively. A CFD simulation approach would be ideal to
understand the actual flow patterns in these localities to decide the
best locations for wind turbine installations. For the cost per
installed kW of less than S$ 3,000, it is possible to achieve simple
payback period of less than 20 years. By employing materials that
are cost-effective and locally sourced and unique design concepts, it
is possible to achieve such low capital cost without substantially
degrading the quality and performance. Since this the first study of
its kind for Singapore, results presented are from the groundwork
towards a comprehensive urban wind flow simulation for
Singapore.
Acknowledgements
Authors would like to thank the Housing Development Board of
Singapore for partially funding this research project. Authors would
like to express their heartfelt gratitude to the Singapore Economic
Development Board for their generous funding to facilitate further
research under the Energy Innovation Research Programme
(NRF2013EWT-EIRP003-032). Thanks to Do Dinh Tho David, Shu
Guixin, P Mohan Kumar and Dr Giuseppe Cavallaro for their assis-
tance in various matters related to the work.
References
[1] Energy Market Authority, Singapore Energy Statisctics 2014, Research and
Statistics Unit, Energy Market Authority, Republic of Singapore, Republic of
Singapore, 2014.
[2] Sectoral Measures to Reduce Emissions (Up to 2020), NCCS, 13.11.2014
(Online). Available: http://app.nccs.gov.sg/nccs-2012/sectoral-measures-to-
reduce-emissions-up-to-2020.html?AspxAutoDetectCookieSupport¼1
(accessed 13.11.2014).
[3] J. Luther, T. Reindl, Solar Photovoltaic (PV) Roadmap for Singapore (A Sum-
mary), Solar Energy Research Institute of Singapore, Singapore, 2014.
[4] W.T.L. Chow, M. Roth, Temporal dynamics of the urban heat island of
Singapore, Int. J. Climatol. 26 (15) (2006) 2243e2260.
[5] National Environment Agency, WEATHERwise Singapore, 2009 (Online).
Available: http://www.nea.gov.sg/docs/default-source/training-knowledge-
hub/guidebook-on-climate-of-singapore.pdf?sfvrsn¼0 (accessed 20.01.2015).
[6] G.K. Nathan, T.N. Goh, Wind measurements in an equitorial region
(Singapore), Sol. Energy 26 (1981) 275e278.
[7] Singapore's Climate Information & Data, 13.11.2014 (Online). Available:
http://app2.nea.gov.sg/weather-climate/climate-information/local-
climatology (accessed 14.11.2014).
[8] J.T. Millward-Hopkins, A. Tomlin, D. Ingham, M. Pourkashanian, The predict-
ability of above roof wind resource in the urban roughness sublayer, Wind
Energy 15 (2012) 225e243.
[9] T.F. Ishugah, Y. Li, R.Z. Wang, J.K. Kiplagat, Advances in wind energy resource
exploitation in urban environment: a review, Renew. Sustain. Energy Rev. 37
(2014) 613e626.
[10] G. Gualtieri, S. Secci, Wind shear coefficients, roughness length and energy
yield over coastal locations in Southern Italy, Renew. Energy 36 (3) (2011)
1081e1094.
[11] H.-C. Lim, T.-Y. Jeong, Wind energy estimation of the Wol-Ryong coastal re-
gion, Energy 35 (12) (2010) 4700e4709.
[12] L. Lu, K. Sun, Wind power evaluation and utilization over a reference high-rise
building in urban area, Energy Build. 68 (2014) 339e350.
[13] S.L. Walker, Building mounted wind turbines and their suitability for the ur-
ban scaleda review of methods of estimating urban wind resource, Energy
Build. 43 (8) (2011) 1852e1862.
[14] A. Gagliano, F. Nocera, F. Patania, A. Capizzi, Assessment of micro-wind tur-
bines performance in the urban environments: an aided methodology
through geographical information systems, Int. J. Energy Environ. Eng. 4 (1)
(2013) 43.
[15] M.A. Heath, J.D. Walshe, S.J. Watson, Estimating the potential yield of small
building-mounted wind turbines, Wind Energy 10 (3) (2007) 271e287.
[16] M.X. Song, K. Chen, Z. He, X. Zhang, Wind resource assessment on complex
terrain based on observations of a single anemometer, J. Wind Eng. Ind.
Aerodyn. 125 (2014) 22e29.
[17] F. Toja-Silva, A. Colmenar-Santos, M. Castro-Gil, Urban wind energy exploi-
tation systems: behaviour under multidirectional flow con-
ditionsdopportunities and challenges, Renew. Sustain. Energy Rev. 24 (2013)
364e378.
[18] D. Elliott, D. Infield, An assessment of the impact of reduced averaging time on
small wind turbine power curves, energy capture predictions and turbulence
intensity measurements, Wind Energy 17 (2) (Feb. 2014) 337e342.
[19] E.L. Petersen, N.G. Mortensen, L. Landberg, J. Højstrup, H.P. Frank, Wind power
meteorology. Part I: climate and turbulence, Wind Energy 1 (S1) (1998)
25e45.
[20] International Standard, Wind Turbines e Part 2 Small Wind Turbines, Inter-
national Electrotechnical Commission, 2013.
[21] P.M. Pardalos, S. Rebennack, M.V.F. Pereira, N. Illiadis, V. Pappu, Handbook of
Wind Power Systems, Springer Berlin Heidelberg, 2013.
[22] J.A. Carta, P. Ramirez, S. Velazquez, A review of wind speed probability dis-
tributions used in wind energy analysis case studies in the Canary Islands,
Renew. Sustain. Energy Rev. 13 (5) (2009) 933e955.
[23] M.A. Al-Fawzan, Methods for Estimating the Parameters of the Weibull Dis-
tribution, 2000 (Online). Available: http://interstat.statjournals.net/YEAR/
2000/articles/0010001.pdf (accessed 01.01.2015).
414 B.R. Karthikeya et al. / Renewable Energy 87 (2016) 403e414
[24] H.S. Bagiorgas, M. Giouli, S. Rehman, L. Al-Hadhrami, Weibull parameters
estimation using four different methods and most energy-carrying wind
speed analysis, Int. J. Green Energy 8 (2011) 529e554.
[25] A.P. Schaffarczyk, Introduction to Wind Turbine Aerodynamics, Springer
Berlin Heidelberg, Berlin, 2014.
[26] M.C. Brower, Wind Resource Assessment: a Practical Guide to Developing a
Wind Project, John Wiley & Sons, Inc., May 2012.
[27] MEASNET, Anemometer Calibration Procedure, October 2009 (Online).
Available: http://www.measnet.com/documents/ (accessed June 2015).
[28] J. Gottschall, M.S. Courtney, R. Wagner, H. Jørgensen, I. Antoniou, Lidar pro-
filers in the context of wind energyea verification procedure for traceable
measurements, Wind Energy 15 (1) (2012) 147e159.
[29] P.J.M. Clive, Remote sensing best practice, in: EWEA Offshore, Amsterdam,
2011.
[30] Erik Holtslag, Improved Bankability e the Ecofys Postion on LiDAR Use,
ECOFYS, 2013. Available online: http://www.ecofys.com/files/files/ecofys-
2013-position-paper-on-lidar-use_02.pdf.
[31] “zephirlidar.com,” ZephIR, (Online). Available: http://www.zephirlidar.com/
support/tech-specs/. (accessed 14.11.2014).
[32] International Energy Agency, Ground-based Vertically-Profiling Remote
Sensing for Wind Resource Assessment, International Energy Agency, 2013.
[33] T. Rutherford, M. Harris, W. Barker, E.B.D. Roziers, M. Pitter, R. Scullion,
C. Slinge, Lidar Calibration and Performance Validation Process, Natural Po-
wer, Hollybush Ledbury, UK, 2014.
[34] N. Carpman, Turbulence Intensity in Complex Environments and its Influence
on Small Wind Turbines, Department of Earth Sciences Geotryckeriet, Uppsala
University, Uppsala, 2011.
[35] ICF International, “The Cost and Performance of Distributed Wind Turbines,
2010e35, June 2010. (Online). Available: http://www.eia.gov/analysis/
studies/distribgen/system/pdf/appendix-b.pdf. (accessed May 2015).
[36] New Energy Husum, 2015 Small Wind World Report Summary, 2015 (Online).
Available: http://small-wind.org/wp-content/uploads/2014/12/Summary_
SWWR2015_online.pdf (accessed May 2015).
[37] DWEA, DWEA Distributed Wind Vision e 2015e2030, March 2015 (Online).
Available: http://distributedwind.org/wp-content/uploads/2012/08/DWEA-
Distributed-Wind-Vision.pdf (accessed May 2015).
[38] Small-scale Wind Energy: Policy Insights and Practical Guidance, Carbon
Trust, 2008.
[39] Energy Market Authority. https://www.ema.gov.sg, 30 December 2014 (On-
line). Available: https://www.ema.gov.sg/Residential_Electricity_Tariffs.aspx
(accessed 20.01.2015).