Renewable Energy 115 (2018) 533e546

Contents lists available at ScienceDirect

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

A geometrical optimization method applied to a heaving point

absorber wave energy converter
Milad Shadman, Segen F. Estefen*, Claudio A. Rodriguez, Izabel C.M. Nogueira
Ocean Engineering Department, COPPE e Federal University of Rio de Janeiro, P.O. Box 68508, CEP 21945-970, Rio de Janeiro, RJ, Brazil

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

Article history: A methodology for the geometrical optimization of wave energy converters (WEC) based on statistical
Received 4 May 2016 analysis methods and the hydrodynamics of the system in the frequency domain is presented. The
Received in revised form optimization process has been applied on a one-body heaving point absorber for a nearshore region of
25 July 2017
the Rio de Janeiro coast. The sea characteristics have been described using a ve-year wave hindcast and
Accepted 21 August 2017
are based on a third generation wind wave model WAVEWATCH III. The optimization procedure is
Available online 23 August 2017
performed based on the resultant wave spectrum and joint probability distribution. The optimization
process aims at maximizing both WEC absorbed power and absorption bandwidth when providing a
Keywords:
Geometry optimization
natural period close to the predominant wave periods of the sea site. The optimized geometry of the WEC
Wave energy converter is determined by running a few simulations in the frequency domain and using the design of experiment
Statistical analyses (DOE) method. The software ANSYS-AQWA is used for the hydrodynamic diffraction analysis, and the
Point absorber DOE method is applied through the Minitab software to determine the optimized geometry. The two
DOE method primary advantages of this optimization method are the reduced computational time and the possibility
of performing parametric analyses for the WEC geometry.
2017 Elsevier Ltd. All rights reserved.

1. Introduction objective of the geometry optimization of the offshore platforms,

Wave energy conversion technology is in the pre-commercial
stage. The cost of a wave energy converter (WEC) increases with
its size. Therefore, the geometry optimization of the system has a
signicant role in the design process to have an economically
feasible system. So far, one of the most promising WEC concepts
under technical and economic evaluation is the point absorber (PA).
It consists of a oater body, which has small dimensions relative to
the incident wavelength and, a support system, which could be
mooring cables connected to the seabed or bottom-mounted
structures. A heaving PA is a wave energy device in which the
heave motion due to the wave-body interaction is absorbed to
produce power. The system that receives the mechanical power
resulting from heave motion and delivers electricity, for instance, is
called power take-off (PTO). Some examples of the PA WEC can be
found in Refs. [1e4]. Geometry optimization techniques applied on
marine structures have been used for seakeeping of ships and
minimizing the dynamic response of moored vessels [5e8]. The
* Corresponding author.
E-mail address: segen@lts.coppe.ufrj.br (S.F. Estefen).
oating breakwaters etc. is to minimize the motion of the structure
to improve the seakeeping ability. In contrast, in the case of point
absorbers, the geometry of WEC must be optimized to make the
system oscillates in the range of sea predominant wave frequencies
allowing maximum possible motion amplitudes to produce more
power. Until now, different works have been presented for geom-
etry optimization of a PA. Vantorre et al. [9] presented an optimi-
zation study of a point absorber with controllable inertia through
linear frequency domain analyses and experimental tests for the
Belgian coastal area of the North Sea. The geometry of the heaving
buoy, PTO damping and supplementary inertia are considered as
variable parameters to maximize the power absorption. Soulard
et al [10] presented a numerical modeling of a two-body heaving
PA. they determined the optimum dimensions of the WEC applying
a statistical analyses method called design of experiments (DOE).
Sjokvist et al. [11] and Goggins and Finnegan [12] applied a fre-
quency domain approach to optimize a heaving point absorber,
with a pure damper as PTO system, for the Lysekil test site in
Sweden and Atlantic marine energy test site (AMETS) respectively.
The objective of the optimization was to determine the buoy di-
mensions that maximize its heaving velocity for that sea state and a
certain PTO damping. To attain this, several runs with different

http://dx.doi.org/10.1016/j.renene.2017.08.055
0960-1481/ 2017 Elsevier Ltd. All rights reserved.
534 M. Shadman et al. / Renewable Energy 115 (2018) 533e546

diameters and drafts were performed for different PTO damping

values. Several studies have also investigated the geometrical
optimization process for other WEC concepts. For instance, Kramer
and Frigaard [13] used the boundary element method to optimize
wave reectors by exploring their orientation and angle to increase
the wave energy absorption for the wave energy converter Wave
Dragon. Genetic algorithms have been used by Babarit and Clement
[14] and McCabe [15] to the shape optimization of the SEAREV
device, which is based on a pendulum that is placed in a closed
buoy actuated by the swell through excitation forces [16], and a
surging wave energy collector, respectively. The SEAREV optimi-
zation process goals were to maximize the absorbed power and
minimize the costs. The objective of the shape optimization of the
surging wave energy collector was to maximize the mean power,
the ratio of the mean power to the characteristic length of the
device shape, and the ratio of the mean power to the displaced
volume of the collector. Kurniawan and Moan [17] used a multi- Fig. 1. Generic point absorber with a PTO system in reference to the seabed.
objective optimization algorithm to determine the optimal geom-
etry of a wave absorber oscillating about a xed submerged hori-
zontal axis. They used two objective functions to be minimized,
which were the ratios of the submerged surface to the maximum
absorbed power and the maximum reaction force to the maximum
absorbed power. Recently, Mahnamfar and Altunkaynak [18] and
Son et al. [19] deployed experimental tests beside the numerical
analyses for the shape optimization of a oscillating water column
and a specic PA WEC respectively.
This paper presents a geometry optimization methodology, as a
preliminary approach, applied on a simple one-body point
absorber, considering the nearshore region of Rio de Janeiro, Brazil
as the sea site. The WEC consists of a oating cylinder, which reacts
against the seabed with an ideal pure damper as the PTO system
(PTO stiffness is zero). The optimization process has been devel-
oped through linear hydrodynamic frequency domain analyses and
the design of experiments (DOE) approach. Minitab [20] is used to
apply the DOE and perform statistical analysis of the optimization Fig. 2. Geometry optimization process of the WEC.
process. The frequency domain analysis of the system hydrody-
namics is performed using ANSYS-AQWA [21]. The diameter and
draft of the oating cylinder (buoy) are considered as the the sea characteristics of the site where the WEC will be installed,
geometrical parameters to be optimized. The objective of the e.g., signicant wave height, mean period and water depth (or ba-
optimization process is to determine the buoy that absorbs the thymetry, for shallow water cases). The second step (immature
maximum wave energy over the largest range of frequencies for the determination) is the denition of the upper and lower bounds for
site's predominant waves. The presented methodology can be the buoy diameter and draft. Within these bounds, a set of
applied to geometrical optimization of different types of wave en- geometrical parameters are obtained for the buoy. The third step is
ergy converters. to compute the primary WEC requirements, what we call re-
sponses in this study, (heave natural frequency, resonance band-
2. Methodology approach width and maximum mechanical power) in the frequency domain
and apply the DOE method. In the next step (mature determina-
The methodology to be used aims at obtaining a set of tion), the results of the previous steps are analyzed, and the opti-
geometrical parameters (factors) to optimize the point absorber in mized buoy geometries are selected based on the statistical
relation to the following three primary requirements (responses): analyses results, which will be explained in the results and dis-
cussion section. Due to the application of the DOE method as a
- Buoy heave natural frequency, which should be within the statistical analysis, a few runs (13 runs) must be performed to
prevailing sea wave frequencies; optimize the responses. It should be noted that for the optimization
- Resonance bandwidth that is dened as the frequency interval with more than two geometrical parameters, the number of
in which the absorbed power is more than half of its maximum required runs could be slightly increased. Furthermore, the
value; graphical results of the statistical analyses such as contour and
- Maximum mechanical power. surface plots offer a broad view of the optimization problem
domain that give us the ability of discussing and comparing the
A simple wave energy converter is used as the WEC system for effect of all the possible sets of geometrical parameters on the
the optimization study. The wave energy converter is a generic responses.
point absorber of a cylinder (buoy) with a diameter D and a draft L,
as illustrated in Fig. 1. The point absorber is a single body device 3. Sea characteristics
with reference to the seabed.
Fig. 2 illustrates the optimization process. As indicated, the rst The information on the wave climate is obtained by either in situ
step in the optimization process is to provide information regarding measurements or numerical modeling. Third generation spectral
 M. Shadman et al. / Renewable Energy 115 (2018) 533e546 535

Table 1 used in this study was obtained from the ETOPO1 database of the
Computational grids used in the simulations: Dx,y is the grid spacing; Nx and Ny are National Oceanic and Atmospheric Administration/National
the number of points in x and y, respectively; Latin and Lonin are the initial latitude
and longitude, respectively, and Lat and Lon are the nal latitude and longitude,
Geophysical Data Centre (NOAA/NGDC). The spectral space was
respectively. resolved in 24 directions (15
angular bandwidth) and 25 loga-
rithmically spaced frequencies between 0.042 Hz (23.8 s) and
Domain Dx,y Nx Ny Latin Lonin Lat Lon
0.5 Hz (2 s).





Grid 1 1 86 81 80 80 0 5
The wave model conguration was validated against data ob-
Grid 2 150 181 121 55
65
15
20

tained from meteo-oceanographic buoys of Brazil's PNBOIA (Na-
Grid 3 5 km 97 73 27 47 21 39
Grid 4 2 km 120 70 24 43.9 22.6 41.5 tional Buoys Program). The nearest PNBOIA buoy to the Rio de
Janeiro coast is the Santos buoy, which is located at 25
170 0.0200 S
and 44
560 0.0200 W and approximately 200 m deep. Data from this
wave models have emerged as a reliable tool for forecasting and buoy has been recorded since 2011; however, the period for which
hindcasting ocean conditions. The use of these models in hindcast there are continuous measurements without gaps was between
mode allows for an assessment of the global wave climate [22] and 2013 and 2014.
the energy resources [23e25]. The Santos wave data recorded from 2013/05 to 2014/01 was
The wave power assessment along the Brazilian coast has been applied to validate the model wave grids. A qualitative view of the
discussed in several studies [26,27]. To describe the nearshore wave validation results in terms of the signicant wave height (Hs), peak
climate on the Rio de Janeiro coast, a wave hindcast for ve years period (Tp), and peak direction (Dp) is illustrated in Fig. 3. The
was developed with the third generation wind wave model comparison between the model simulations and the Santos buoy
WAVEWATCH III (WW3) version 3.14 [28] using the NCEP's Climate data shows a good agreement. Despite an underestimation of the
Forecast System Reanalysis (CFSR) wind database [29]. The CFSR extreme Hs values, results from the WW3 are relatively close to the
covers the period from 1979 to 2010 and has considerably higher observed peak values. Furthermore, the WW3 was capable of
temporal and spatial resolutions than other reanalysis. The ad- modeling rapidly changing sea conditions, and it achieved accurate
vantages of this latest reanalysis can be given as follows: (1) timing of the observed peak storm waves.
landeatmosphereeocean coupling during the generation of the 6- To quantify the accuracy of the model, standard error metrics of
h guess eld, (2) interactive sea ice model and (3) assimilation of Bias (Bias), Root Mean Square Error (RMSE), Scatter Index (SI) and
satellite radiances for the entire period. In this report, to perform Pearson's correlation coefcient (r) are used. If xi represents the
the wave climate hindcast, the last ve years (2006e2010) of the measured values, yi represents the simulated values and n is the
CFSR were selected. The WW3 wave model was run for 4 different number of observations, the statistical expressions can be dened
grids. Grid details can be found in Table 1. The bathymetric data as follows:

Fig. 3. Comparison between the time series of the signicant wave height Hs, peak period Tp, and peak direction Dp for the results from the Santos Basin buoy and the WW3 model.
536 M. Shadman et al. / Renewable Energy 115 (2018) 533e546

Table 2 results. The SI, which is dimensionless, denotes the error percent-
Bias, Root Mean Square Error (RMSE), Scatter Index (SI) and Pearson Correlation's age. The correlation is a measure of the strength and direction of a
Coefcient (r) for the correlations between the wave parameters of the buoy and the
WW3.
linear relationship between the measured and modeled values. A
correlation of 1 indicates a perfect one-to-one linear relationship
Buoy Wave Parameters BIAS RMSE SI r whereas 1 indicates a negative relationship.
Santos Hm0 (m) 0.03 0.33 0.15 0.91 The values of these statistical parameters are listed in Table 2.
Tp (s) 0.32 1.7 0.17 0.75 For the Santos buoy, the results for Hs are considerably better than
Dp (
) 18 39 0.23 0.75
for all of the other considered parameters. The correlation coef-
cient was 0.91 for Hs, 0.75 for Tp and 0.75 for Dp. The BIAS, RMSE
and SI values are small in both parameters, thus demonstrating that
Pn the wave model ts well with the observations [22,30,31]. There-
i1 xi  yi
Bias (1) fore, this calibrated model conguration forced with the CFSR
n winds was used to perform a ve-year wave hindcast between
s 2006 and 2010 of the nearshore wave climate in Rio de Janeiro.
Pn 2 The wave statistics on the Rio de Janeiro coast are presented in
i1 xi  yi
RMSE (2) terms of their joint probability distribution (JPD) in Fig. 4, which
n
indicates the probability of both a signicant wave height and a
wave peak period occurring for a given location. There is a domi-
RMSE
SI (3) nance of peak periods between 7 and 9 s for an Hs between 0.5 and
y
1 m with occurrence of 17.3%. Furthermore, the majority of waves
Pn have a height of less than 2 m, and only 2.2% of the waves are more
i1 xi  xy  y than 3 m in height. The predominant waves have frequencies be-
r q
Pn
(4)
2 Pn 2 tween 7 and 13 s with an occurrence of 86%. The average signicant
i1 i x  x i1 i y  y
height was Hs 1:33 m, and the wave peak period was
Tp 9:7 s.
where the overbar in x and y denotes the average values. The Bias is
The average wave energy spectrum based on the wave hindcast
used to indicate the quantity of deviation in the model results. By
is illustrated in Fig. 5. The primary spectral characteristic deter-
taking the average of the bias over the entire period, the underes-
mined from the average spectrum is the presence of large amounts
timation (negative value) or overestimation (positive value) of the
of energy at a frequency between 0.14 and 0.076 Hz (7.14 and
model can be identied. The RMSE, which is a positive value,
13.16 s). The peak of this average spectrum was 0.0908 Hz (11 s).
demonstrates the accuracy and magnitude of the error in the model

Fig. 4. Joint Probability Distribution (%) for the nearshore region of Rio de Janeiro.
 M. Shadman et al. / Renewable Energy 115 (2018) 533e546
The total stored energy, E, in a wave per unit area of sea surface
in terms of the signicant wave height (Hs) can be computed
directly from the hindcast data as follows [32]:
rgHs2
E (5)
16
where r is the density of the uid; and g is the acceleration due to
gravity. The wave power level, P, per width unit in a wave in terms
of the signicant wave height (Hs) and the energy period (Te) can
be given as follows:
rg2 Hs2 Te
P (6)
64p
The measured sea states are often specied in terms of the
signicant wave height Hs and either the peak period Tp or the
mean period Tz. The energy period Te is rarely specied and must
be estimated from other variables. When the peak period Tp is
known, one potential approach can be assumed as follows:
Te aTp (7)
where the coefcient a depends on the frequency spectrum model,
which was assumed to be equal to 1 [33].
Equation (6) is a deep water approximation and has been used
in this study to describe the wave power for the nearshore region of
Rio de Janeiro. By considering an average signicant height,
Hs 1:33 m, and a wave energy period, Te Tp 9:7s, for the
sea site and applying equation (6), the resultant wave power level
per unit width is approximately 8.5 kW/m.
Fig. 6 illustrates a combined scatter and energy diagram to
visualize the composition of the wave energy resource in terms of
wave heights and periods. The numerical values represent the
probability of occurrence of a combination of signicant wave
heights and peak period in percentage. The peak period and sig-
nicant wave height are divided into intervals of 2 s and 0.5 m,
respectively. Color scale represents annual wave power level (in
MWh=m year). As it shown, the most energetic waves have sig-
nicant wave heights between 1.5 m and 2.0 m with peak periods
between 9 s and 11 s, which represents only 7.1% of the total
number of waves per year.
4. Hydrodynamics of the system
4.1. Equations of motion
In the preliminary hydrodynamic modeling of the WECs, it is
typically assumed that the hydrodynamic forces of the oating
body in waves are those obtained from the linear diffraction theory,
i.e., viscous effects are neglected and only potential forces are
considered. Thus, the response of a single oating body in a wave is
generally described using a mass-spring system. By assuming a
linear system with 6
of freedom, the equations of motion for this
analysis can be represented as follows:
6 
X 
Mkj Akj xj Bkj x_j Ckj xj Fk k 1; 26 (8)
j1
where k and j are the subscripts for denoting the hydrodynamic
properties in the k-mode as a result of motion in the j-mode; Mkj is
the mass of the structure; Akj , Bkj and Ckj are the added mass matrix,
damping matrix and hydrostatic restoring matrix, respectively; and
Fk represents the other external forces in the kth mode. A detailed
discussion on the linear diffraction theory can be found in Ref. [34].
537
Because the primary mechanism for the energy extraction of the
proposed point absorber is the heave motion, at this preliminary
stage, only the heave motion of the oater will be considered. Thus,
the motion equation can be given as follows:
M A33 x3 B33 x_3 C33 x3 F3 t (9)
where M is the mass of the system; A33 , B33 and C33 are the added
mass matrix, damping matrix and hydrostatic matrix, respec-
tively, in the heave direction; and F3 is the external force on the
body in the heave direction. It should be noted that the external
forces may include, in addition to the wave excitation forces and
mooring, other forces, such as mechanical forces (e.g., power take-
off forces).
4.2. Power take-off (PTO)
A PTO system should be integrated in the WEC modeling to
calculate the wave energy conversion. A PTO applies a force on the
system in the opposite direction of the displacement to absorb its
kinetic energy and produces electricity or other desired form of
energy. Therefore, in the case of heaving point absorber, the
absence of PTO system results in a larger heave amplitude, specif-
ically in the resonance range. Although it is not completely realistic,
a simple pure damper model is used here to represent the power
take-off mechanism. The stiffness and control forces may be
included in later stages. The damping of the PTO system is assumed
to be frequency independent. The PTO force that is applied on the
buoy in the heave direction can be expressed as follows:
FPTO DPTO x_3 (10)
where DPTO is the PTO damping coefcient. After applying the PTO
force, the equation of motion for the system can be given as follows:
M A33 x3 B33 x_3 C33 x3 F3 t FPTO (11)
According to Falnes [32], a maximum amount of energy can be
captured if the power take-off mechanical damping is equal to the
radiation damping of the point absorber. Under these assumptions,
the PTO damping will be equal to that of the point absorber at
resonance. Falnes and Lillebekken [2] provided an expression for
the absorbed mean power (Ppto ) of a heaving wave energy con-
verter under harmonic motion as follows:
1
Ppto DPTO u2 jX3 j2 (12)
2
where u is the wave frequency; and X3 is the amplitude of the buoy
heave motion.
5. Immature determination
As mentioned in previous sections, the optimization of the buoy
starts by an immature determination, which includes the denition
of the upper and lower bounds for its diameter and draft, D and L,
respectively. These bounds should satisfy two design premises. The
rst one is associated with the maximum amount of power that the
buoy can absorb from incident waves, and the second one is related
to the maximum buoy response in heaving due to the incident
waves. Evidently, both premises are related to the waves' charac-
teristics, specically the range of predominant wave periods at the
WEC's installation site.
538 M. Shadman et al. / Renewable Energy 115 (2018) 533e546

Fig. 5. Average spectral density (m2/Hz) based on the hindcast from 2006 to 2010 for the nearshore region of Rio de Janeiro.

5.1. Diameter range (D)
The rst premise requires that the maximum possible energy
available from the incident wave is absorbed by the buoy. The
maximum energy that may be absorbed by a heaving axisymmetric
body equals the wave energy transported by the incident wave
front of width equal to the wavelength divided by 2p [32]. The
maximum capture width or absorption width, Lmax can be
dened as follows [32]:
Lmax l=2p
Then, the maximum wave power Pmax , absorbed by a heaving
axisymmetric body can be given as follows:
Pmax JLmax
where J is the energy ux per unit wave crest width. For linear
deep water waves, we can obtain
. according to Falnes' recommendation [2], would be
l gT 2 2p
. wave. However, there is a disagreement between the lower bound
J rg 2 TH2 32p
where r is the water density, T and H are the period and height of
the wave. If the capture width is divided by the device's width (e.g.,
the buoy diameter, D), we can obtain the non-dimensional or
relative capture width C wr , which, for successful devices, should
satisfy (recommended by Twidell and Weir) [35] the expression as
follows:
Lmax
Cwr 3 (17)
D
Based on Eqs. (13), (15) and (17) and knowing that the pre-
dominant wave periods for the nearshore region of Rio de Janeiro
(13) are between 7 s and 13 s, the upper bound of D can be established
for both limits: for the 7 s wave, D  4 m, and for the 13 s wave,
D  14 m. Additionally, there is a practical rule for a point absorber
used as a WEC that recommends that the point absorber diameter
(14) should preferably be in the range of 5%e10% of the prevailing
wavelength [2]. For the nearshore region of Rio de Janeiro, the
prevailing wavelength corresponds to the wave peak period
(Tp 9:7 s); thus, the recommended range for the buoy diameter,
(15) 7  D  14 m. The upper value is in good agreement with the one
obtained from the wave capture width consideration for the 13 s
(16) of the Falnes recommendation and the bound given by the capture
width for the 7 s wave. Although the capture width suggests buoy
diameters smaller than 4 m, the practical rule recommends buoy
diameters exceeding 7 m. For the optimization process, it is
 M. Shadman et al. / Renewable Energy 115 (2018) 533e546 539

Fig. 6. eCombined scatter and energy diagram: the colors denote the annual wave power level (MWh/m year), and the numbers indicate the probability occurrence per year (%) in
terms of signicant wave height and peak period. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

desirable to cover a wider range of buoy diameters; hence, the
bounds for D will be taken as 4 < D < 14 m.
5.2. Draft range (L)
Once the lower and upper bounds of D are dened, the draft of
the buoy may be determined. Now, the second design premise for
the buoy will be used, i.e., the range of heave natural periods of the
buoys should match the range of predominant wave periods of the
WEC installation site. Assuming a one degree-of-freedom free-
oating body, the buoy heave natural frequency u1n3 can be given as
follows:
s conclusions may be stated regarding the responses in waves for a
rgAwp
u1n3
M133 A133
where Awp is the buoy water plane area, which is a function of the
1 is the mass of the buoy based on the sub-
buoy diameter (D); M33
mersed volume (i.e., D and L); and A133 is the added mass coef-
cient, which can be calculated at this preliminary stage as follows
[36]:
A133 0:167rD3
To trigger the buoy resonance motions, Eq. (18) should be set
equal to the incident wave frequency. For the sea site's prevailing
wave periods (7e13 s) and the range of buoy diameters obtained in
Section 5.1 (4e14 m), a set of values for the buoy draft (L) that cause
the buoy to resonate within a range of prevailing waves was
determined, as indicated in Table 3.
Within each table cell, the value of the draft is listed for several
buoy diameters that satisfy the maximum capture width condition
for a given wave period to make the buoy resonate during heaving.
The cell colors are used to identify the magnitude of L (light gray for
smaller values and dark gray for the larger ones) so that the cor-
responding upper and lower bounds for L can be easily established
as 3 < L < 42 m. Nevertheless, because this range is excessively
wide, a few additional considerations should be noted to narrow
the limits for L.
Based on the typical linear responses of a mass-spring-damper
system under harmonic external excitation, a few preliminary
point absorber. In the resonance region, the body responses are
(18) dominated by damping, and the responses can achieve relatively
large amplitudes when the damping is small. In the low-frequency
region (frequencies lower than the resonance frequency), the body
responses are dominated by restoring. At very low frequencies, the
wavelength is large when compared to the horizontal length
(diameter) of the buoy, and it will follow the waves, upwards and
downwards. In the high frequency region (frequencies larger than
the resonance frequency), the buoy responses are dominated by
mass, which indicates that the waves lose their inuence on the
(19) behavior of the buoy; there are several wave crests and troughs
within the horizontal length (diameter) of the buoy [37]. From the
above statement, it can be concluded that to harness the wave
energy for a wider range of wave periods, it is desirable to design a
buoy whose natural period is tuned with the lowest prevailing
540 M. Shadman et al. / Renewable Energy 115 (2018) 533e546

Table 3
Resonant buoys.

Lmax Period Diameter,

(m) 14 13 12 11 10 9 8 7 6 5 4
(m) (s) D (m)
6.2 39.0 5.0 3.2 3.4 3.7 3.9 4.1 4.3 4.5 4.7 4.9 5.1 5.4
8.9 56.2 6.0 6.0 6.2 6.4 6.6 6.8 7.0 7.2 7.5 7.7 7.9 8.1
12.2 76.5 7.0 9.2 9.4 9.6 9.8 10.1 10.3 10.5 10.7 10.9 11.1 11.3
15.9 99.9 8.0 12.9 13.1 13.4 13.6 13.8 14.0 14.2 14.4 14.6 14.8 15.0
Dra , L
20.1 126.5 9.0 17.2 17.4 17.6 17.8 18.0 18.2 18.4 18.6 18.9 19.1 19.3
(m)
24.8 156.1 10.0 21.9 22.1 22.3 22.5 22.7 22.9 23.1 23.4 23.6 23.8 24.0
30.1 188.9 11.0 27.1 27.3 27.5 27.7 27.9 28.2 28.4 28.6 28.8 29.0 29.0
35.8 224.8 12.0 32.8 33.0 33.2 33.4 33.7 33.9 34.1 34.3 34.5 34.7 35.0
42.0 263.9 13.0 39.0 39.2 39.4 39.7 39.9 40.1 40.3 40.5 40.7 40.9 41.0

wave period (or perhaps, less) so that it works in the resonance The primary advantages of using the DOE can be stated as follows:
region for the shorter wave lengths (lower wave periods) and
follows the waves for the longer waves (higher wave periods). - Screening the factors to determine which are important for
Thus, the region where the waves lose their inuence is mini- explaining the process variation;
mized. In conclusion, to guarantee a wider operational wave sce- - Understanding how factors interact and drive the process;
nario for the WEC, shorter natural periods should be selected. For - Finding the factor settings that produce optimal process
example, from Table 3, we can select heaving natural periods be- performance;
tween 5 and 7 s such that the limits for L can be updated to - Decreasing the number of experiments or runs and time-saving;
3 < L < 12 m. Table 4 summarizes the upper and lower bounds for
D and L, as determined from the previous sections. There can be several experiments designed for a specied
number of design variables; however, the extreme designs that
encompass all of the others are called screening experiments and
6. Design of experiments methodology response surface experiments. Screening experiments are used
when the objective is the study of a large number of design vari-
The design of experiments (DOE) is a systematic technique for ables to determine the most important ones rather than the in-
studying any situation that involves a response that varies as a teractions between pairs of variables. The response surface
function of one or more independent variables. The DOE can experiments are capable of providing the primary effects of the
address complex problems where more than one variable may variables, factor interactions and terms to measure the curvature
affect a response (or a set of responses) and two or more variables induced in the response by each design variable [38]. Hence, the
may interact with each other. The technique provides answers to response surface method (RSM) is applied to the sequence of
specic questions on the behavior of a system requiring an opti- designed experiments to obtain an optimal response.
mum number of experimental observations. In this optimization study, the combination of the geometrical
So far a range of primary dimensions has been dened for the parameters (factors) of the buoy is sought to satisfy the desired
WEC, 4 < D < 14 m and 3 < L < 12 m. Theoretically, an innite resonance bandwidth, natural frequency and maximum power of
number of solutions is possible. The main task is to determine the the WEC (responses). By considering a certain natural frequency as
optimal solution without having to perform a large number of the goal, the response variable is dened as the combination of the
simulations. In fact, by applying the DOE method, it is possible to
study the performance of hundreds of buoys in the mentioned
range using just a few computer runs (13 buoys), as indicated in Table 5
Table 5. Designed experiments.
Each process has inputs that determine the behavior of a system
Factors Responses
as well as outputs that are produced by the process. The objective of
doing an experiment (or computer simulation in this study) is to D (m) L (m) Max mechanical Resonance Natural
power (kW) bandwidth (s) period (s)
determine how the inputs affect the outputs. The goal of engi-
neering experimentation is to learn how to control the process 1 5.46 10.68 323.80 0.23 7.08
2 14 7.5 293.73 1 6.87
inputs to generate the outputs. The process inputs are called vari-
3 9 7.5 233.58 0.56 6.30
ables or factors whereas the process outputs are called responses. 4 4 7.5 143.38 0.2 5.97
5 9 7.5 233.58 0.56 6.30
6 12.53 4.32 166.63 1.49 5.54
Table 4 7 5.46 4.32 107.87 0.41 4.84
Upper and lower bounds for the buoy's geometrical parameters. 8 9 7.5 233.58 0.56 6.30
9 9 12 577.91 0.4 7.80
Lower bound Upper bound
10 9 7.5 233.58 0.56 6.30
D (m) 4 14 11 9 7.5 233.58 0.56 6.30
L (m) 3 12 12 12.53 10.68 437.93 0.69 7.54
13 9 3 99.11 1.3 4.74
 M. Shadman et al. / Renewable Energy 115 (2018) 533e546
Fig. 7. 3D meshed geometry of a buoy.
maximum absorbed power and resonance bandwidth which pro-
vides a heaving point absorber that absorbs the maximum energy
within the broadest range of frequencies. The resonance bandwidth
is dened as the frequency interval in which the absorbed power is
more than half of its maximum value.
7. Results and discussion; mature determination
Thus far, the range of factors was determined based on the sea
site characteristics and the principle conditions applied using the
DOE method. In this step, the statistical analysis results are dis-
cussed, and the nal buoys are determined based on the resultant
contour and surface plots. The hydrodynamic analyses (experi-
ments) are performed in the frequency domain using ANSYS/
AQWA. Fig. 7 illustrates a typical geometrical model and its dis-
cretization as required by ANSYS/AQWA. An additional mechanical
damping in the heave direction, has been considered to simulate
the pure PTO system (the stiffness of PTO was assumed to be zero).
The PTO damping is considered to be equal to the hydrodynamic
damping of each buoy at its resonant frequency to maximize the
energy absorption at that frequency [32].
7.1. Diameter and draft interactions
After analyzing the designed experiments in Minitab, the results
are illustrated by the contour plots in Fig. 8.
The contour plots give us the ability of screening the in-
teractions of the geometrical parameters (factors) and their effects
on the resonance bandwidth (RB), maximum power (MP) and
natural period (NP) as the responses. In Fig. 8, the horizontal and
vertical axes represent the diameter (D) and draft (L) in the pre-
dened ranges that have been determined in the immature deter-
mination step. The solid and dashed lines show the responses. The
black represents the resonance bandwidth and the light and dark
gray show the maximum power and natural period, respectively. A
maximum and a minimum value of 1.4 s and 0.2 s are respectively
considered for the resonance bandwidth based on the frequency
domain analyses and DOE results. It means that, the buoys which
are located on the solid black line (RB 0:2 s) have the lowest
resonance bandwidth comparing to the others implying that its
performance capability is good close to the natural period but poor
elsewhere. Moving away from this zone toward the maximum
resonance bandwidth line (RB 1:4 s, dashed black line), the buoys
with the capability of absorbing wider range of wave periods are
541
Fig. 8. Contour plot of the geometrical parameters (draft and diameter) in meter and
the maximum power (MP) in kW, resonance bandwidth (RB) and natural period (NP)
in second.
achievable. The white area shows the buoys (D; L) bounded values,
which are in the ranges of 5e8 s and 100e570 kW for the natural
period and the maximum power respectively. For instance, the
buoy with the diameter of 13 m and draft of 11 m is located in the
white area and has a natural period close to 8 s with a maximum
power of approximately 570 kW at its resonance period. By
observing the contour plot, it can be inferred that the maximum
resonance bandwidth and the maximum power do not occur at the
same region. Fig. 8 illustrates that the larger diameters with rela-
tively small drafts (L < 6 m) provide a good resonance bandwidth
while the maximum energy absorption is achievable in the same
diameter range with larger drafts (L > 10 m). However, it should be
kept in mind that the maximum power in the present methodology
is the energy absorption at the buoy natural period. On the other
hand, the results show that the highest possible natural period
provided by a buoy in the geometrical parameter range is 8 s. It
implies that a very large buoy, outside the specied range of di-
ameters and drafts, is required to satisfy the resonant condition for
the sea site energetic waves, which are between 9 and 11 s (Fig. 6).
This is a challenge for the local seas with predominant wave pe-
riods beyond 6 or 7 s. Fig. 9 shows some frequency domain analyses
for different buoys selected from Fig. 8, to provide a better under-
standing of the interaction of the geometrical parameters and their
effects on the responses. For this reason, an arbitrarily selected
buoy of diameter 11 m and draft 5 m is considered. To observe the
effect of both diameter and draft, the simulations are performed for
the buoys with a constant diameter and varying drafts and vice-
versa. A constant PTO damping is applied to maximize the energy
absorption. The vertical and horizontal axes represent the
maximum mechanical power and wave periods, respectively. Fig. 9-
a illustrates the effect of changing diameters for a constant draft. As
can be seen, larger diameter improves the buoy performance by
increasing the maximum power, natural period and resonance
bandwidth. This is based on theoretical analyses for an uncon-
strained system. In practice, there are some limitations regarding
the cost, PTO forces, mooring etc. that restrict the buoy diameter
increase. Furthermore, increasing the diameter results in better
542 M. Shadman et al. / Renewable Energy 115 (2018) 533e546

Fig. 9. a) Different diameters with a xed draft; and b) different drafts with a xed diameter.
 M. Shadman et al. / Renewable Energy 115 (2018) 533e546
Fig. 10. Surface plot of the; a) the resonance bandwidth (s), and b) the maximum mechanical power (kW) vs. buoy diameter and draft.
absorption in low-frequency region (periods higher than the nat-
ural period) while letting the absorption in high-frequency region
unchanged. It may be a desirable effect for the sea regions that are
dominated by large wave periods. Fig. 9-b shows the inuence of
the draft in the performance of the buoy. Increasing the draft leads
to a higher maximum power and natural period but a lower reso-
nance bandwidth. As seen, as the buoy draft increases the power
curve narrowed leading to a short resonance bandwidth. It means
that the buoys with larger drafts provide high natural periods and
absorb a signicant amount of energy in its resonant condition.
Quantitatively, the effect of the draft is more signicant than that of
diameter. For instance, the maximum power increase of 13% is
reached for a diameter increase of 2 m (considering buoy (7; 5) and
buoy (9; 5) in Fig. 9-a), while this value is about 20% for the same
increment in draft (considering buoy (11; 7) and buoy (11; 9) in
Fig. 9-b). Therefore, it should be noted that modifying the draft
values can signicantly inuence the point absorber performance.
The contour plots enable us to understand the problem better and
identify the limits by providing a broad view of its domain.
7.2. Optimum diameter and draft
Surface plots can also illustrate the interaction of geometrical
parameters to analyze the statistical results. Fig. 10 shows the
surface plot of the maximum mechanical power and resonance
bandwidth versus buoy diameter and draft. As seen, the buoy with
the maximum mechanical power, buoy (14; 12), has a relatively
low resonance bandwidth whereas the buoy with the maximum
resonance bandwidth, buoy (14; 3), has a low mechanical power.
Therefore, considering Fig. 10-a, to guarantee a sufciently wide
resonance bandwidth, the values lower than half of the maximum
(maximum resonance bandwidth is equal to 2 s) are discarded. It
should be noted that, this is a relative resonance bandwidth, in
other words, the modied range provides the buoys that have a
larger resonance bandwidth values when comparing to the other
buoys in the main range. Then, the modied geometrical parameter
range is achieved, 8 < D < 14 m and 3 < L < 6 m.
The optimized buoy in this range is determined using the
Minitab response optimizer [20]. This feature helps identify the
combination of geometrical parameters that jointly optimize the
resonance bandwidth and maximum power. The joint optimiza-
tion must satisfy the requirements for all the responses in the set,
which here is the maximization of the combination of the reso-
nance bandwidth and maximum power. This is measured by the
composite desirability. The composite desirability assesses how
well a combination of input variables (D and L) satises the
543
targets dened for the responses. In this study, the objective is to
obtain a system that works as close as possible to the prevailing
wave periods while absorbing the maximum possible energy in a
relatively wide range of the sea dominant waves. By considering
the modied range of the factors resulted from Fig. 10, buoy
(14; 6) has the largest values of diameter and draft, providing the
highest natural period (6 s). This resonance period is selected as a
target for the response optimization process. The maximum
mechanical power and resonance bandwidth are set to satisfy the
maximum values. Eventually, the response optimizer mathe-
matically determines the buoy that satises the optimization
requirements. Under these conditions, 13:5 m diameter and 5 m
draft are determined for the optimized buoy. As indicated in
Fig. 11, to verify the mathematical optimization result, a set of
runs with buoys in the range of 8 < D < 14 m and 3 < L < 6 m are
performed in AQWA/ANSYS, and the results are compared to
those from the optimization process.
Fig. 11 illustrates the captured mechanical power for the buoy
over the range of period between 4 and 15 s. The vertical axis
represents the normalized mechanical power. The power values
are divided by Pc, which is the maximum mechanical power of
buoy (13:5; 5) and buoy (13:5; 6) in Fig. 9-a and b, respectively. It
can be seen that, as expected, decreasing the diameter in Fig. 9-a
results in a decrease in the maximum mechanical power, reso-
nance bandwidth and resonance period, which implies that a
diameter of 13.5 m is the best selection in the range. This result is
in accordance with the result of the numerical optimization. In
Fig. 9-b, buoy (13:5; 5) is compared with other buoys with
different drafts in the range of 3 to 6 m. The results are the same
as those discussed in Figs. 8 to 10. Increasing the draft leads to a
buoy with a higher maximum mechanical power, narrower
resonance bandwidth and a higher resonance period. Conversely,
decreasing the draft results in a lower maximum mechanical
power, wider resonance bandwidth and lower resonance period.
However, an important issue that must be considered during the
optimization of the buoy dimensions is the performance of the
buoy over the range of prevailing wave periods at the sea site. As
indicated in Fig. 11-b and considering buoy (13:5; 3), it can be seen
that although this buoy has the lowest maximum mechanical
power and resonance period, it captures more power over the sea
site prevailing wave periods (between 7 and 13 s) when
compared to the dimensions of other buoys. In Fig. 11-b, the dif-
ferences between the captured power graphs in the range of
predominant wave periods are not signicant; however, selecting
a buoy with a smaller draft results in a few benets, such as lower
cost and lower system loads due to a lighter weight system, which
544 M. Shadman et al. / Renewable Energy 115 (2018) 533e546

Fig. 11. Normalized mechanical power vs period for; a) different diameters with the same draft; and b) Different drafts with the same diameter.

consequently results in increasing the efciency of the system.
Thus, a buoy with a diameter of 13:5 m and a draft of 3 m is
selected as the optimized buoy for the mentioned nearshore sea
site of Rio de Janeiro.
It can be seen that for sea regions, such as the nearshore region
of Rio de Janeiro, with a predominant wave period greater than 7 s,
the challenge is to have a passive point absorber with a resonance
bandwidth tuned to the prevailing wave range. In our study, the
one-body point absorber that can meet this requirement is a large
buoy close to the geometrical parameters boundaries. Practically,
this type of system results in several challenges related to the cost,
slamming, water depth consideration, proper PTO system, etc. By
applying control methods, a smaller system that works tuned to the
prevailing wave range may be achievable.
 M. Shadman et al. / Renewable Energy 115 (2018) 533e546
8. Conclusion
A methodology for the geometric optimization of WECs based
on a series of frequency domain analyses and a statistical analysis
method known as Design of Experiments (DOE) was presented. The
optimization process is applied to the preliminary design of a one-
body point absorber with an axisymmetric oating cylinder for the
nearshore region of Rio de Janeiro. An ideal pure damper is
considered as power take-off (PTO) system, and the energy ab-
sorption is calculated for different wave frequencies. First, the local
sea characteristics has been described through a ve-year wave
hindcast (2006e2010) based on a third generation wind wave
model WAVEWATCH III. The results indicated a predominant wave
period range between 7 and 13 s with an energy period Te 9:7 s ,
as well as an average signicant height of Hs 1:33 m. Based on
these results, the calculated wave power level per unit width is
approximately 8:5 kW=m. Then, the optimization procedure is
applied in two principal steps, immature and mature determina-
tion. The lower and upper bounds of the geometrical parameters
are determined in the rst step based on the practical recom-
mendations regarding the ratio of the point absorber diameter to
the predominant wavelength and the non-dimensional capture
width. In the second step, after considering a set of frequency
domain analyses and the design of experiments method, the WEC's
geometrical parameters (diameter and draft) are determined to
achieve a system that absorbs the maximum energy over a wide
range of wave periods. In other words, a set of geometrical pa-
rameters (factors) are determined to maximize the response that is
a combination of the maximum absorbed power and the resonance
bandwidth of the buoy.
The results include the contour and surface plots of the factors
interactions. The effects of the different sets of diameter and draft
on the maximum mechanical power, resonance bandwidth and
natural period of the buoy were determined. The lower and upper
bounds of the geometrical parameters are further modied based
on the resonance bandwidth requirement. Considering the
modied domain of the diameter and draft, the maximum buoy
natural period is T 6s which corresponds to the buoy with a
diameter 14 m and draft 6 m. Consequently, this period is
considered as a satisfactory target natural period for the optimi-
zation process. The response optimizer, which is a feature of
Minitab, is used to mathematically determine the best set of
geometrical parameters that jointly maximize the maximum po-
wer and resonance bandwidth. The numerical results of the
response optimizer are then veried by a series of frequency
domain analyses and, eventually, the buoy with diameter
D 13:5 m and draft L 3 m is selected as the optimized buoy for
the design sea site. It can be seen that none of the buoys inside the
dened upper and lower bounds can provide a natural period and
resonance bandwidth tuned to the sea site predominant wave
range. This has been observed as the primary challenge for the
nearshore region of Rio de Janeiro that have prevailing wave pe-
riods beyond 7s.
Applying DOE method as a statistical analysis combined with
the frequency domain hydrodynamic analysis provide a wide
picture of the optimization problem domain, and the possibility of
observing the effects of different variables on the optimization
process objective. Furthermore, a considerable amount of CPU
time is saved by performing only a few runs for the optimization
process. Time domain simulations are recommended in more
advanced phases of the design to conrm the selected parameters
for the proposed WEC. The proposed methodology can be applied
to other types of wave converters with different geometrical
properties.
545
Acknowledgements
The authors acknowledge CAPES, Ministry of Education/Brazil,
for the D.Sc. Scholarship to the rst author, and CNPq (305338/
2013-7), Ministry of Science, Technology and Innovation/Brazil, for
supporting research activities of the second author. Special thanks
to FURNAS through ANEEL (contract number: 9000000692) (Bra-
zilian Electrical Energy Agency) P&D Program for the nancial
support of the research in progress at Subsea Technology Labora-
tory (COPPE) on wave energy.
References
[1] P.B. Garcia-Rosa, J.P.V.S. Cunha, F. Lizarralde, S.F. Estefen, I.R. Machado,
E.H. Watanabe, Wave-to-Wire model and energy storage analysis of an ocean
wave energy hyperbaric converter, IEEE J. Ocean. Eng. 39 (2013) 386e397,
http://dx.doi.org/10.1109/JOE.2013.2260916.
[2] J. Falnes, P.M. Lillebekken, Budal's latching-controlled-buoy type wave-power
plant, in: Proc. Fifth Eur. Wave Energy Conf, 2003, pp. 233e244.
[3] R. Waters, M. Stlberg, O. Danielsson, O. Svensson, S. Gustafsson,
mstedt, et al., Experimental results from sea trials of an offshore wave
E. Stro
energy system, Appl. Phys. Lett. 90 (2007), http://dx.doi.org/10.1063/
1.2432168.
[4] OPT powerbuoy Available from:http://www.oceanpowertechnologies.com/
n.d.
[5] G.F. Clauss, L. Birk, Hydrodynamic shape optimization of large offshore
structures, Appl. Ocean. Res. 18 (1996) 157e171, http://dx.doi.org/10.1016/
S0141-1187(96)00028-4.
[6] L. Brik, G. Clauss, Automated Hull optimisation of offshore structures based on
rational sea keeping criteria, in: Elev. Int. Offshore Polar Eng. Conf. Stavanger,
Norw, 2001.
[7] Brik L, Clauss G. Parametric hull design and automated optimization of
offshore structures. Tenth Congr. Int. Marit. Assoc. Mediterr. (IMAM 2002), n.d.
[8] L. Brik, Application of constrained multi-objective optimization to the design
of offshore structure hulls, J. Offshore Mech. Arct. Eng. 131 (2009) 11301,
http://dx.doi.org/10.1115/1.2957919.
[9] M. Vantorre, R. Banasiak, R. Verhoeven, Modelling of hydraulic performance
and wave energy extraction by a point absorber in heave, Appl. Ocean. Res. 26
(2004) 61e72, http://dx.doi.org/10.1016/j.apor.2004.08.002.
[10] T. Soulard, M. Alves, A. Sarmento, Force reacting principle applied to a heave
point absorber wave energy converter, in: Proc. nineteenth Int. Offshore Polar
Eng. Conf. Osaka, Japan, 2009, pp. 312e318.
kvist, R. Krishna, M. Rahm, V. Castellucci, A. Hagnestl, M. Leijon, On the
[11] L. Sjo
optimization of point absorber buoys, J. Mar. Sci. Eng. 2 (2014) 477e492,
http://dx.doi.org/10.3390/jmse2020477.
[12] J. Goggins, W. Finnegan, Shape optimisation of oating wave energy con-
verters for a specied wave energy spectrum, Renew. Energy 71 (2014)
208e220, http://dx.doi.org/10.1016/j.renene.2014.05.022.
[13] M.M. Kramer, P.B. Frigaard, Efcient wave energy amplication with wave
reectors, in: Proc. Twelfth Int. Offshore Polar Eng. Conf. Kitakyushu, Japan,
2002.
[14] A. Babarit, A.H. Cle ment, Shape optimisation of the SEAREV wave energy
converter, in: Proceedings of the 9th World Renewable Energy Congress,
Florence, Italy, August 2006, pp. 19e25.
[15] A.P. McCabe, Constrained optimization of the shape of a wave energy collector
by genetic algorithm, Renew. Energy 51 (2013) 274e284, http://dx.doi.org/
10.1016/j.renene.2012.09.054.
[16] Ruellan M, Ben Ahmed H, Multon B, Josset C, Babarit A, Cle ment AH. Design
methodology for a SEAREV wave energy converter. IEEE Trans. Energy
Convers. 2010;25:760e767. doi:10.1109/TEC.2010.2046808.
[17] A. Kurniawan, T. Moan, Optimal geometries for wave absorbers oscillating
about a xed axis, IEEE J. Ocean. Eng. (38) (2013) 117e130, http://dx.doi.org/
10.1109/JOE.2012.2208666.
[18] F. Mahnamfar, A. Altunkaynak, Comparison of numerical and experimental
analyses for optimizing the geometry of OWC systems, Ocean. Eng. 130 (2017)
10e24, http://dx.doi.org/10.1016/j.oceaneng.2016.11.054.
[19] D. Son, V. Belissen, R.W. Yeung, Performance validation and optimization of a
dual coaxial-cylinder ocean-wave energy extractor, Renew. Energy 92 (2016)
192e201, http://dx.doi.org/10.1016/j.renene.2016.01.032.
[20] HYPERLINK "http://support.minitab.com/en-us/minitab/17/" \o "http://sup-
port.minitab.com/en-us/minitab/17/"http://support.minitab.com/en-us/mini-
tab/17/(last Access on 09 Febraury 2017) (n.d).
[21] HYPERLINK "http://148.204.81.206/Ansys/150/Aqwa%20Users%20Man-
ual.pdf" \o "http://148.204.81.206/Ansys/150/Aqwa%20Users%20Manual.pdf"
http://148.204.81.206/Ansys/150/Aqwa%20Users%20Manual.pdf(last Access
on 09 Febraury 2017)(n.d).
[22] A. Chawla, D.M. Spindler, H.L. Tolman, Validation of a thirty year wave
hindcast using the Climate Forecast System Reanalysis winds, Ocean. Model
70 (2013) 189e206, http://dx.doi.org/10.1016/j.ocemod.2012.07.005.
546 M. Shadman et al. / Renewable Energy 115 (2018) 533e546
[23] R. Waters, J. Engstrom, J. Isberg, M. Leijon, Wave climate off the Swedish west
coast, Renew. Energy 34 (2009) 1600e1606, http://dx.doi.org/10.1016/
j.renene.2008.11.016.
[24] T. Ching-Piao, H. Ching-Her, H. Chien, C. Hao-Yuan, Study on the wave climate
variation to the renewable wave energy assessment, Renew. Energy 38 (2012)
50e61, http://dx.doi.org/10.1016/j.renene.2011.06.041.
[25] C.E. Hiles, On the Use of Computational Models for Wave Climate Assessment
in Support of the Wave Energy Industry (Master thesis), Department of Me-
chanical Engineering, University of Victoria, 2010.
[26] E.R. Beserra, A.L.T. Mendes, S.F. Estefen, C.E. Parente, Wave climate analysis for
a wave energy conversion application in Brazil, in: ASME 2007 26th Int. Conf.
Offshore Mech. Arct. Eng, 2007, pp. 897e902, http://dx.doi.org/10.1115/
OMAE2007-29597.
[27] P. Contestabile, V. Ferrante, D. Vicinanza, Wave energy resource along the
coast of Santa Catarina (Brazil), Energies 8 (2015) 14219e14243, http://
dx.doi.org/10.3390/en81212423.
[28] H.L. Tolman, User manual and system documentation of WAVEWATCH-IIITM
version 3.14, Tech. Note (2009), http://dx.doi.org/10.3390/ijerph2006030011.
[29] S. Saha, S. Moorthi, H.L. Pan, X. Wu, J. Wang, S. Nadiga, et al., The NCEP climate
forecast system reanalysis, Bull. Am. Meteorol. Soc. 91 (2010) 1015e1057,
http://dx.doi.org/10.1175/2010BAMS3001.1.
[30] J.H.G.M. Alves, E.O. Ribeiro, G.S. Grossmann Matheson, J.A. Moreira Lima,
C.E. Parente Ribeiro, Wave climate hindcast for Brazilian South-Southeast
between 1997 and 2005, Rev. Bras. Geos. 27 (2009) 427e445, http://
dx.doi.org/10.1590/S0102-261X2009000300010.
[31] L. Rusu, P. Pilar, C. Guedes Soares, Hindcast of the wave conditions along the
west Iberian coast, Coast Eng. 55 (2008) 906e919, http://dx.doi.org/10.1016/
j.coastaleng.2008.02.029.
[32] J. Falnes, Ocean waves and oscillating systems: Linear interactions including
wave-energy extraction, Appl Mech Rev 56 (2003), http://dx.doi.org/10.1115/
1.1523355. B3.
[33] G. Hagerman, Southern England wave energy resource potential, in: Proc.
Build. Energy 2001, New Engl. Sustain. Energy Assoc, Boston, Massachusetts,
2001.
[34] J.N. Newman, Marine Hydrodynamic, MIT press, Cambridge, Mass, 1977.
[35] J. Twidell, T. Weir, Renewable Energy Resources, third ed., 2015.
[36] J.P. Hooft, Oscillatory wave forces on small bodies, in: Int. Shipbuild. Prog, IOS
Press, 1970, pp. 127e135.
[37] J.M.J. Journee, W.W. Massie, Offshore Hydromechanics, 2001, p. 570, http://
dx.doi.org/10.1016/S0013-4686(01)00879-9.
[38] Paul G. Mathews, Design of experiments with MINITAB, Am. Stat. 60 (2006)
205, http://dx.doi.org/10.1198/tas.2006.s46.