See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327500032

Small 500 kW onshore wind farm project in Kribi, Cameroon: Sizing and
checkers layout optimization model

Article in Energy Reports · September 2018

DOI: 10.1016/j.egyr.2018.08.003

CITATIONS READS

38 1,484

1 author:

Nkongho Ayuketang Arreyndip

Università Ca' Foscari Venezia
18 PUBLICATIONS 130 CITATIONS

SEE PROFILE

All content following this page was uploaded by Nkongho Ayuketang Arreyndip on 07 September 2018.

The user has requested enhancement of the downloaded file.

 Energy Reports 4 (2018) 528–535

Contents lists available at ScienceDirect

Energy Reports
journal homepage: www.elsevier.com/locate/egyr

Research paper

Small 500 kW onshore wind farm project in Kribi, Cameroon: Sizing

and checkers layout optimization model
∗
Nkongho Ayuketang Arreyndip a,b,c , , Ebobenow Joseph a
a
Department of Physics and Applied Physics, University of Buea, 63, Buea, Cameroon
b
African Institute for Mathematical Science (AIMS), Limbe, Cameroon
c
Polytechnic, Saint Jerome Catholic University of Douala, 5949, Akwa, Douala, Cameroon

article info a b s t r a c t

Article history: For the purpose of providing cheap, affordable and reliable electrical energy to communities in need
Received 30 January 2018 and power-up new industries and enterprises, a small onshore wind farm with an estimated capacity
Received in revised form 18 May 2018 of 500 kW is designed and studied. With the help of a wind rose, wind resource map, and results from
Accepted 20 August 2018
Weibull statistics, a potential site is selected and the wind farm is positioned along Cameroon’s coastline
Available online xxxx
for maximum energy capture. Using the PARK model for wind turbine layout optimization, two different
Keywords: layout patterns (Checkers pattern and column pattern) are studied for the purpose of minimizing the
Wind farm wake effect and thereby, maximizing the output power from the farm. The Checkers model was found
Layout suitable as compared to the column model to be used on Grand Batanga, a small locality South of the city
Optimization of Kribi, Cameroon.
© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction compared to solar in Cameroon because it is cheaper and yields

more power per square area of land than solar.
The quest to become an emerging state by the year 2035 has To select the best site for the installation of a small wind farm,
pushed Cameroon’s government to heavily invest in the energy the first step is to carry out a wind energy assessment which is
sector, as energy is the driving force behind the rapid develop- usually done using wind resource maps. A wind resource map
ment of any country. Significant progress has already been made for Cameroon is shown in Fig. 1 (Anon, 2014). Here we see that
with a newly constructed dam in the south region of the country. favourable wind speeds to support a small wind farm are mostly
But giving that dams are not ecologically friendly and also given available in the Northern parts of the country and the coastal
problems posed by global warming with the earth warming by 1◦ regions. Also given that the population density in the coastal re-
every 10 years, has caused most sub-saharan African countries like gions is increasing in an alarming rate and more industries and
enterprises are being created in these regions, we choose to look
Cameroon to be affected by droughts. Long periods of droughts or
for a potential wind energy site for the installation of a small wind
dry season as is often the case in the tropics give rise to seasonal
farm along the coast.
rivers. Hence there is usually very low energy output from these
Next we have applied the Weibull statistics to estimate and
dams during the dry seasons which causes some large communi-
compare the wind power densities, average wind speeds, most
ties to go days and sometimes weeks without electricity.
probable wind speeds and wind speed carrying maximum energy
Our objective in this work is to contribute in the energy sec-
of three different locations in three different coastal regions to
tor by identifying potential renewable energy sources, sites and
select the best site with sufficient wind speed (Arreyndip et al.,
propose efficient ways of harnessing them that can add significant 2016). When the best site is selected, next step is to select an
megawatts to the national grid. Which will go a long way to make appropriate wind turbine with wind speed characteristics capable
electrical energy available in rural areas that are often forgotten or of fully functioning on the chosen site and finally given the energy
deprived of basic energy and water. needs of the consumer, a wind farm is constructed to satisfy these
For a larger and more industrious community like Kribi, wind energy needs.
energy has been found suitable as an alternative energy source To construct a wind farm, wind turbines layout within the farm
is a very crucial step towards harvesting higher energy from the
∗ Corresponding author at: Department of Physics and Applied Physics, Univer- farm as wrong positioning leads to maximizing the wake effect
sity of Buea, 63, Buea, Cameroon. and minimizing power capture. The problem of wake (which is
E-mail address: ayuketang@aims-cameroon.org (N.A. Arreyndip). a low wind field created by a wind turbine that is sometimes

https://doi.org/10.1016/j.egyr.2018.08.003
2352-4847/© 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-
nd/4.0/).
 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535 529

2. Methods

We will start this section with a brief review of the widely used
Weibull statistics for wind energy assessment in which we have
already presented some results which are stated here-in in our
previous work. We will also revisit the actuator disc concept and
the PARK model which will help readers without a background
in wind energy to know how and where some constants and
parameters like the axial induction factor, power coefficient are
derived and assigned values.

2.1. Weibull statistics

The Weibull distribution with probability density function, cu-

Fig. 1. Wind resource map of Cameroon showing distribution of wind speed over
mulative probability distribution and quantile distribution given
the territory (Anon, 2014). by (Arreyndip et al., 2016; Fagbenle et al., 2011; Oyedepo et al.,
2012; Kollu et al., 2012; Ozerdem and Turkeli, 2003; Carta et al.,
2009; Montgomery and Runger, 2003; Akpinar and Akpinar, 2005;
Arreyndip and Joseph, 2016):
experienced by another wind turbine downstream) elimination in
v k−1 v k
( )( ) [ ( )]
a large wind farm with clusters of large wind turbines still remains k
f (v ) = exp − , (1)
a challenge (Wang et al., 2016b; Kusiak and Song, 2010; Song c c c
et al., 2013). But this challenge can equally be eliminated by the
constructing stand-alone wind turbines farm that will definitely v k
( ( ))
take an incredible land space. Hence given that we are constrained F (v ) = 1 − exp − , (2)
c
in a particular land surface area, researchers have developed and
simulated complex wake models to solve the wake problem (Wang 1
et al., 2016b; Kusiak and Song, 2010; Song et al., 2013). They Q (v ) = [− log(1 − v )] k . (3)
have shown that an efficient wind farm design can be carried The mean value of the wind speed vm and standard deviation
out through optimization of wind farm layout and also carrying ν is defined in terms of the Weibull parameter k and c are given
out some control strategies (Wang et al., 2016b; Kusiak and Song, as (Arreyndip et al., 2016; Fagbenle et al., 2011):
2010; Song et al., 2013). Which can go a long way to minimize the ( )
1
wake effect by cutting down power loses due to turbulence and vm = c Γ 1+ , (4)
maximize the wind farm output power. k
Here, we will limit ourselves to simple analytic wake models where Γ () is the gamma function. and
without going deeply into Computational Fluid Dynamics (CFD)  [
 ( ) [ ( )]2 ]
simulation since we do not have access to the software. More  2 1
complex wake models have been developed like the Ainslie-based ν = c Γ 1+
√ 2 − Γ 1+ . (5)
k k
models, the parabolic Navier–Stokes (N–S) equation and the com-
plete 3D (N–S) models (Wang et al., 2016b). Researchers who have
2.1.1. Mean power and energy densities
already developed and studied different analytic models for wind When assessing the wind energy potential of a site, there are
farm layout optimization are; Longyan et al. applied and compared two wind speeds that are of interest and must therefore be taken
the effectiveness of three different analytic wake models (PARK, seriously into consideration. These are the most probable wind
Larsen and B–P model) for wind farm with variable and constant speed vmp and the wind speed carrying maximum energy vEmax .
hub heights. One of their result is that, when using PARK model, the They are given by the expression (Oyedepo et al., 2012; Kollu et al.,
surface roughness value must be carefully tuned to achieve good 2012):
performance in predicting the wind farm power production (Wang ( )(1/k)
et al., 2016b). Kusiac and Song also applied the PARK model for k−1
vmp = c , (6)
the purpose of maximizing output power from a circular shape k
wind farm. Their complex nonlinear model was solved by the and
evolutionary strategy algorithm with the quality of the generated ( )(1/k)
k+2
solutions acceptable for industrial applications (Kusiak and Song, vEmax = c . (7)
2010). k
In this work, we will limit ourselves to the most popular and Theoretically, the mean power density is proportional to the cube
widely used wake model for wind farm layout optimization which of the mean velocity given by,
is the PARK model and focus more on wind turbine sizing and the
P(vm ) 1
layout patterns with minimal wake effects. The remaining sections PD = ρv 3 .
= (8)
are organized and explained as follows. In Section 2, we will revisit A 2 m
the widely used Weibull model for wind energy assessment, the It can also be calculate from the Weibull probability density func-
actuator disc concept and the PARK model. We will also make use tion given by,
P(v )
( )
of the wind rose which will help us in positioning wind turbines 1 3
in the farm. Sizing and layout patterns over the selected site for PD = = ρc Γ3
1+ , (9)
A 2 k
maximum energy capture are also discussed. Wind power output
will also ba calculated and compared using the different layout and the energy density is given by,
( )
patterns. Section 3 is dedicated to discussing the results of our 1 3
findings and we will end with a conclusion in Section 4.
ED = ρc3Γ 1+ T. (10)
2 k
530 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535

Table 1 Table 3
Data descriptive statistics. Summary statistics for Weibull fit using the method of maximum likelihood.
Data Kribi Douala Limbe Data Kribi Douala Limbe
min (m/s) 1 0.8 1.45 k 2.9 3.9 3.0
max (m/s) 6.45 3.45 4.62 s.e 0.1 0.14 0.128
median (m/s) 2.3 1.6 2.04 c (m/s) 2.8 1.8 2.58
mean (m/s) 2.5 1.7 2.31 s.e (m/s) 0.05 0.026 0.05
s.d 0.87 0.43 0.7 likelihood −475.6 −232.27 −333.17
skew 1.45 0.9 1.35 P-value of K–S 0.14 0.118 0.18
kurt 5.17 4.43 3.88
p-value 2.011e−10 2.89e−05 4.19e−09

2.2. Actuator disc theory

Table 2
Wind characteristics of coastal regions.
The actuator disc concept is based on the fact that a rotating
Data Kribi Douala Limbe
rotor produces a disclike structure that spins wind in circular
Longitude 9.98 9.7 9.21
direction behind the turbine to produce a stream tube shown in
Latitude 2.88 4.05 4.02
Elevation (m) 15 41 36
Fig. 4 (Castellani et al., 2013; Archer et al., 2013). This aerodynamic
vmp (m/s) 2.426 1.67 2.27 simulation gives us an idea about the variation of wind speeds and
vmax (m/s) 3.35 2.0 3.30 pressure before and after the rotating rotor. The resulting spinning
PD (W/m2 ) 33.7 8.0 25.42 velocity behind the rotor is what is called the wake which greatly
Pmax (W/m2 ) 160 30 60 affects wind turbines positioned downstream as their power cap-
ture drops because of turbulence.
Let us consider the actuator disc shown in Fig. 4. The velocity
Where PD is the power density, ED the energy density, P(v ) wind and pressure in front of the disc are denoted as vu and P0+ re-
power, A cross sectional area of rotor, ρ the air density assumed spectively. While those behind the disc are denoted as vw and P0−
respectively (Munteanu et al., 2014). Momentum change across the
here to take the value 1.225, c scale parameter, k shape parameter,
wind turbine is given by,
T period, and Γ () the gamma function (Fagbenle et al., 2011;
Oyedepo et al., 2012; Kollu et al., 2012). The shape and scale pa- △ T = m(vu − vw ), (13)
rameters here, have been estimated from the maximum likelihood
which is the momentum transmitted to the disc by air of mass
method given by the expressions:
m passing through the disc with cross sectional area A. The force
on the blades due to this momentum change is then given by
[ ∑n k ∑n ]−1
i=1 vi ln(vi ) i=1 ln(vi )
(Munteanu et al., 2014):
shape(k) = − , (11)
△ m(vu − vw )
∑n k
i=1 vi
n △T
F = = = ρ Av0 (vu − vw ). (14)
where vi is the wind speed in time step i and n is the number of △t △t
data points. The force can also be written in the form
The scale parameter is given by (Fagbenle et al., 2011): F = A(P0+ − P0− ), (15)
[ n
]1/k
1∑ and also making use of the Bernoulli’s equation, the pressure
scale(c) = vik . (12) difference is given by (Munteanu et al., 2014):
n
i=1
1
The data set used in the previous work (Arreyndip et al., 2016) is P0+ − P0− = ρ (vu2 − vw2 ). (16)
2
NASA satellite wind data over Limbe, with latitude 4.02◦ N, longi- Immediately before and after the disc, v0 = vu and v0 = vw
tude 9.21◦ E and elevation of about 69 m above sea level, Douala respectively, the force is then given by,
with longitude 9.76◦ E, latitude 4.05◦ N and with an elevation of
1
19m above sea level and Kribi with longitude 9.98◦ E, latitude F = ρ A(vu2 − vw2 ). (17)
2.88◦ N with an elevation of 10 m above sea level. These data sets 2
contains 31 (1983–2013) years of mean daily data with missing Using Eqs. (14) and (17), we find that v0 = 1
2
v +vw ) → vu −vw =
( u
values in the years 1992 and 1994 for the Limbe data which is taken 2(vu − v0 ).
at Debouncha, on the western coast of Limbe. Hence these years The power
were considered as outliers. 1
The results of the parameter estimation, mean power densities, P = ρ Av0 (vu2 − vw2 ), (18)
2
maximum power densities, most probable wind speeds and wind
speed carrying maximum energy over three of Cameroon’s coastal now become
cities have already been presented in our previous work (Arreyndip 1
P = ρ Av 3 4a(1 − a)2 . (19)
et al., 2016) but displayed in Tables 1–3 and Fig. 2 for clarity. Fig. 3a 2
and b are the probability density function (PDF) and cumulative v
Where a = 1 − v0 is called the axial induction factor (Munteanu
u
distribution function (CDF) of NASA mean monthly data of Kribi et al., 2014).
fitted to Weibull distribution (Arreyndip et al., 2016). The power coefficient Cp given by Cp = P /Pt where Pt = 12 ρ Av 3
From Tables 1–3 and comparative plots for Weibull power is the ideal theoretical power now becomes
density in Fig. 2, we see that Kribi is the best site given that it has a
higher Weibull power density and mean wind speed. But we also Cp = 4a(1 − a)2 . (20)
see that wind speed in Kribi that is estimated about 3 m/s is very A Cp value of 0.59 is the maximum possible value and is called
low to support a modern large wind turbine energy farm. Hence the Betz limit wish corresponds to the maximum energy extracted
small wind turbines with cut-in wind speed a little lower than the from the wind. Cp value of 0.59 corresponds to an induction factor
mean speed over Kribi are employed to construct the wind farm. a = 1/3 (Munteanu et al., 2014).
 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535 531

Fig. 4. Actuator disc concept.

Fig. 2. A comparative plot of Weibull power density over Douala, Limbe, and Kribi.

2.3. PARK model

The PARK model assumes a linear expansion of the wake behind

a wind turbine as shown in Fig. 5 with the wake velocity at a
distance x down stream given by (Wang et al., 2016b; Kusiak and
Song, 2010; Song et al., 2013; Samorani, 2014; Jensen, 1983; Katic
et al., 1986):
[ ( )2 √ ]
rr Aw
vx = v0 1 − 2a , (21)
rr + α x Ar

where a is the axial induction factor, rr is the downstream rotor

Fig. 5. An illustration of PARK model showing full and partial wake cases.
radius, Aw is the overlap area due to the wake effect and Ar is the
turbine area of cross section. α is the wake spreading coefficient
which depends on the surface roughness length z0 and the wind
turbine hub height h given by the expression (Wang et al., 2016b; where R is the rotor radius. The wake radius down stream is given
Samorani, 2014): by

0.5 r x = r r + α x. (24)
α= ( ). (22)
ln h For full wake layout, the area of overlap Aw is same as rotor area
z0 Ar . Hence the wake velocity for full wake becomes
Downstream rotor radius is given by (Wang et al., 2016b): [ ( )2 ]
√ rr
1−a
, vx = v0 1 − 2a , (25)
rr = R (23) rr + α x
1 − 2a

Fig. 3. Weibull probability wind speed distribution of NASA satellite data of Kribi. (a) is the probability density function and (b) is the cumulative distribution function.
532 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535

with the term

( )2
rr
vdef = 2a (26)
rr + α x
call the velocity deficit for full wake (Wang et al., 2016b). from
Eq. (26), we see that the vdef → 0 as x → ∞. Hence velocity deficit
disappears at larger distance of separations between turbines and
vx ≈ v0 . For partial wake, the area of overlap Aw is given by (Wang
et al., 2016b):
⎡
rx2 + d2 + △ h2 − R2
( )
⎢
Aw = rx2 ⎢
⎣ cos
−1
√
2rx d2 + △ h2

rx2 + d2 + △ h2 − R2
( ( )) ⎤
sin 2 cos −1 √
2rx d 2 + △ h2 ⎥
− ⎥
2 ⎦

⎡
R2 + d2 + △ h2 − rx2
( )
⎢ −1
+ R2 ⎢
⎣ cos √
2R d2 + △ h2

R2 + d2 + △ h2 − rx2
( ( )) ⎤
sin 2 cos −1 √
2R d2 + △ h2 ⎥ Fig. 6. Wind rose of Kribi showing dominant wind directions.
− ⎥ (27)
2 ⎦

with velocity deficit vdef given by,

N N
∑ ∑
Pix ,
[ )2 √ ]
Ptot = Pi0 + (33)
(
rr Aw
vx = v0 1 − 2a . (28)
rr + α x Ar i=1 i=1

where Pi0 is the power from the first row of wind turbines with no
This is for the case of a wind turbine under a single wake. For the wake effect and Pix is the power from other wind turbines assuming
case of a wind turbine under multiple wakes, the velocity deficit is all experience partial wake. So for no wake effect, the total power
given by (Wang et al., 2016b): from the farm is calculated to be Ptot = 500 kW. For partial wake,

 N ( ( )2 the total power from the farm is calculated to be Ptot = 330.7 kW
and for full wake, the total power is Ptot = 288.4 kW.
)2 √
∑ rr Aw
vdefi =√ 2a . (29)
rr + α x Ar
r =1 2.4. Wind turbine positioning
The wake velocity due to these wind turbines is given by
A wind rose is a graphic tool used by scientist to give a succinct
vi = v0 (1 − vdefi ) (30) view of how wind speed and direction are typically distributed at
a particular location. They are very useful in that, they give the
A single wind turbine power according to the onshore wind turbine direction of wind flow at a particular period of time which greatly
power curve is given by, helps for wind turbine positioning for maximum energy capture.
1 The wind rose for Kribi in Fig. 6 shows that wind appears to be
P = ρ Aν 3 Cp , (31) blowing from both the Southwest and West direction with West-
2
South-West being the dominant wind direction. Hence East-North-
where ν is the velocity that depends on the wind farm layout East square wind farm positioning with wind turbine upstream
model. For the case of zero wake affect, ν is the free stream velocity facing West-South-West is the best for maximum energy capture.
and for the case of a partial or full wake, ν is the wake velocity in
full or partial wake. The total power from the farm is given by, 2.5. Sizing and layout models
N
∑
Ptot = Pi (32) For sizing, we assume the availability of a 1 km × 1 km flat
i=1
piece of land for the installation of 100 5 kW wind turbines with
rotor radius of 10 m and same hub height of 40 m. The turbines are
where N is the number of wind turbines. As an example, let us placed 100 m apart which gives us the ideal positioning of at least
consider a case where the area of overlap for partial wake Aw = 5 blade diameter distance apart.
0.7Ar . The wake velocity becomes vx = 0.314v0 . Full wake is for Now, the way wind turbines are positioned for minimal wake
the case where Aw = Ar with wake velocity vx = 0.297v0 . Taking effect and maximizing the total wind energy production from the
ρ = 1.225 kg/m3 , rr = 10 m, a = 0.33, z0 = 0.01, h = 40 m, farm is the main area of focus for this work. Many researchers have
x = 100 m and Cp = 0.5. The power of one wind turbine under the used different wake models and advance computer simulation
influence of partial wake become Pi = 2.99 kW and Pi = 2.52 kW software like Computational Fluid Dynamics (CFD) to study differ-
for full wake. The estimated power from the wind farm is given by, ent wake models and layout models that will drastically reduce the
 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535 533

Fig. 7. An illustrated case of wind turbine layout patten (a) is the column model for full wake and (b) is the checkers model for partial wake. ‘X’ represents the wind turbine.

wake effect while economizing total space to be used (Samorani, Table 4

2014; Jensen, 1983; Katic et al., 1986; Feng and Shen, 2015; Wang Small wind farm estimated parameters and installation cost.

et al., 2016a, 2015; Frandsen, 1992; Shahriar et al., 2015). Here, we Parameters Values

have used the most popular and widely used model for wind farm Wind Farm estimated power 500 kW
1 Wind turbine power 5 kW
optimization layout which is the PARK model.
Number of wind turbines 100
Users of the PARK model through computer simulation using Rotor diameter 20 m
CFD have found that this model experience the maximum possible hub height 40 m
wake effect called full wake for wind turbines placed in column Cut-in-speed 2.0 m/s
patten. Other wind farm developers have found that irregular Rated speed 8 m/s
Cut-out-speed 12 m/s
arrangement of wind turbines in a farm have more wake minimiza- Total cost (shipping plus installation) $15,000
tion than wind turbines arranged in columns. So our model of the for 1 turbine
wind farm stems from the game of checkers. A checkers pattern 100 turbines cost $1,500,000
is shown in Fig. 7b. Our idea is that while using same dimension of Average cost of 1 m2 of land $20
Total cost of land $20,000,000
land, the second row of wind turbines are placed at mid distance of
Total project cost $21,500,000
the two neighbouring turbines in the first row while those in the
3rd row are also placed at mid point between tow neighbouring
turbines in the second row and are on same column as those in
the first row. Also working on the assumption that the wake effect for column layout that accounts for full wake and Fig. 7b for
is drastically reduced at wider distances between two turbines on Checkers layout that accounts for partial wake which is based on
same column, only partial wake is experienced in this checkers the assumption that, the further downstream a wind turbine is
layout model. from a wake, the lesser it experiences the wake effect and hence the
wake velocity approaches the free stream velocity. In the last part
3. Results and discussion of Section 2.3, we have calculated and compared the total power
output of the farm in different three layout patterns. Here, we see
We have previously compared wind energy potentials of three that the ideal layout of wind turbines is for all wind turbines to be
cities in three different coastal regions of Cameroon and found placed in a row which gives an estimated power of 500 kW. But this
that Kribi in the South region is the best site (Arreyndip et al., layout model will take an incredible amount of space. The next best
layout model is the Checkers model with an estimated power of
2016). We also proposed the use of Savonius wind turbines for
330.7 kW and finally the column model with an estimated power
stand alone low energy needs. But this site can also support small
of 288.4 kW. The estimated power can then be met by using an
wind turbines up to 100 kW. Table 4 shows a suggestion of the
appropriate step-up transformer.
wind turbine to be used, its parameter specifications and total
Finally, Fig. 10 is a google map showing the position where the
cost including shipping and installation. In Figs. 8 and 9, using
wind farm is to be constructed at Grand Batanga locality in Kribi.
the PARK model, we study the effects of distance(x) from the
We estimated the availability of a 1 km × 1 km square piece of
wind turbine upstream of the wake on wake velocity at different land that will be taken away from the indigents and compensated
surface roughness as shown in Fig. 8a. Here, we see that the surface by funds allocated in Table 4. Fig. 11 is a satellite map showing
roughness has a very small effect on the wake velocity but greatly positioning of the farm and Checkers model layout patten over the
affects the velocity deficit for full wake as shown in Fig. 8b. site following wind direction from wind rose of Kribi.
Fig. 9a studies the effect of surface roughness on velocity deficit
for partial wake which is similar to that of full wake but with a 4. Conclusion
slight decrease. In Fig. 9b, power deficit for partial and full wake
are compared at a surface roughness of z = 0.001. We see that the Due to mass migration of working population from landlocked
power deficit is slightly less in partial wake layout as compared to regions to the coastal regions in search of jobs and a better liveli-
full wake layout. The various layout pattens are shown in Fig. 7a hood, the population of these coastal regions is increasing yearly
534 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535

Fig. 8. (a) is a plot of full wake velocity against distance from wind turbine at different values of the surface roughness. (b) is a plot of the velocity deficit at a distance from
the turbine at different values of the surface roughness.

Fig. 9. (a) is a plot of partial wake velocity deficit against distance from wind turbine at different values of the surface roughness. (b) is a plot of wake power deficit comparing
full wake with partial wake at a distance from the turbine.

Fig. 11. Wind farm positioning south of Grand Batanga using the checkers model.
Fig. 10. Google image of Kribi showing chosen site for wind farm installation.

in an alarming rate. Also to become an emerging state by the 100 5 kW (with rotor diameter of 20 m) wind turbines to meet this
estimated power.
year 2035, the government of Cameroon has heavily invested in
To harvest maximum power from the farm, we have applied
industrialization. There is therefore the demand for more and more
the PARK model for wind turbine layout optimization. We have
energy. To solve these ever increasing energy crises in the densely
studied two different layout pattens which are the column layout
populated coastal cities, a thorough wind energy potential has pattern where the wind turbines are laid out in columns in the farm
been carried out to identify a potential site for the installation and the Checkers layout pattern (being inspired from the game
of a small onshore wind farm. Kribi has been found suitable and of Checkers) where wind turbines of the next row are displaced
the wind speed characteristics over Kribi, shows that it can only half the distance of separation between turbines of the previous
support a wind farm made up of small wind turbines. We have row. Through calculations, we have seen that the Checkers layout
estimated the power output from this wind farm of dimension is better than the column pattern in minimizing the wake effect
1 km × 1 km to be 500 kW. We have proposed the installation of and thereby maximizing the total output power from the farm.
 N.A. Arreyndip, E. Joseph / Energy Reports 4 (2018) 528–535 535

Acknowledgements Feng, J., Shen, W.Z., 2015. Solving the wind farm layout optimization problem using
random search algorithm. Renew. Energy 78, 182–192.
Frandsen, S., 1992. On the wind speed reduction in the centre of large clusters of
This work was supported by the government of Canada’s In-
wind turbines. J. Wind Eng. Ind. Aerodyn. 39 (1), 251–265.
ternational Development Research Centre (IDRC), and within the Jensen, N.O., 1983. A note on wind generator interaction.
framework of the AIMS Research for Africa Project Katic, I., Højstrup, J., Jensen, N., 1986. A simple model for cluster efficiency.
(CMMCM2014025S). Kollu, et al., 2012. Mixture probability distribution functions to model wind speed
The authors will like to thank Prof. Alain Moise Dikande, head distributions. Int. J. Energy Environ. Eng. 3, 27, http://www.journal-ijeee.com/
content/3/1/27.
of the LaRaManS research group of the Department of Physics, Kusiak, A., Song, Z., 2010. Design of wind farm layout for maximum wind energy
University of Buea for fruitful discussions that let to solid ideas capture. Renew. Energy 35 (3), 685–694.
used to produce this manuscript. Montgomery, D.C., Runger, G.C., 2003. Applied Statistics and Probability for Engi-
neers, third ed. Wiley, New Jersey.
Munteanu, I., Bratcu, A.I., Cutululis, N.-A., Ceanga, E., 2014. Optimal Control of Wind
References
Energy Systems. Towards a Global Approach, Springer, pp. 15–16 (Chapter 2).
http://www.springer.com/gp/book/9781848000797.
Akpinar, E.K., Akpinar, S., 2005. An assessment on seasonal analysis of wind energy Oyedepo, et al., 2012. Analysis of wind speed data and wind energy potential in
characteristics and wind turbine characteristics. Energy Convers Manage 46, three selected locations in south-east Nigeria. Int. J. Energy Environ. Eng. 3, 27,
1848e67. http://www.journal-ijeee.com/content/3/1/7.
Anon, 2014. Vestas Wind Systems A/S, derived from International Energy Agency Ozerdem, B., Turkeli, M., 2003. An investigation of wind characteristics on the
(IEA), the World Bank and Vestas data https://www.africa-eu-renewables.org/ campus of Izmir Institute of Technology, Turkey. Renew. Energy 28, 1013e27.
market-information/cameroon/renewable-energy-potential/. Samorani, Michele, 2014. Michele Samorani The Wind Farm Layout Optimiza-
Archer, C.L., Mirzaeisefat, S., Lee, S., 2013. Quantifying the sensitivity of wind farm tion Problem. Springer, p. 16, https://link.springer.com/chapter/10.1007/978-
performance to array layout options using large-eddy simulation. Geophys. Res. 3-642-41080-2_2.
Lett. 40 (18), 4963–4970. Shahriar, M.R., et al., 2015. Fault detection of wind turbine drivetrain utilizing
Arreyndip, Nkongho Ayuketang, Joseph, Ebobenow, 2016. Generalized extreme power- speed characteristics. In: 9th WCEAM Research Papers. Springer, pp.
value distribution models for the assessment of seasonal wind energy potential 143–155.
of Debuncha, Cameroon. J. Renew. Energy 9357812, 9, http://dx.doi.org/10. Song, M.X., et al., 2013. Bionic optimization for micro-siting of wind farm on
1155/2016/9357812. complex terrain. Renew Energy 50, 551–557.
Arreyndip, Nkongho Ayuketang, Joseph, Ebobenow, David, Afungchui, 2016. Wind Wang, Longyan, Tan, Abdy C.C., Cholette, Michael, Gu, Yuantong, 2016b. Comparison
energy potential assessment of Cameroon’s coastal regions for the installa- of the effectiveness of analytical wake models for wind farm with constant and
tion of an onshore wind farm. Heliyon 2, e00187. http://dx.doi.org/10.1016/j. variable hub heights. Energy Convers. Manage. 124, 189–202.
heliyon.2016.e00187. Wang, L., Tan, A., Gu, Y., 2016a. A novel control strategy approach to optimally
Carta, J.A., Ramıez, P., Velázquez, S., 2009. A review of wind speed probability design a wind farm layout. Renew. Energy 95, 10–21.
distributions used in wind energy analysis: case studies in the Canary Islands. Wang, L., et al., 2015. A new constraint handling method for wind farm layout
Renew. Sustain. Energy Rev. 13, 933e55. optimization with lands owned by different owners. Renew. Energy 83, 151–
Castellani, F., et al., 2013. A practical approach in the cfd simulation of off-shore 161.
wind farms through the actuator disc technique. Energy Proc. 35, 274–284.
Fagbenle, R.O., et al., 2011. Assessment of wind energy potential of two sites in
North-East, Nigeria. Renew. Energy 36, 1277–1283.

View publication stats