You are on page 1of 15

Renewable Energy 96 (2016) 410e424

Contents lists available at ScienceDirect

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

Development of a nearshore oscillating surge wave energy converter


with variable geometry
N.M. Tom*, M.J. Lawson, Y.H. Yu, A.D. Wright
National Renewable Energy Laboratory, MS 3811, 15013 Denver West Parkway, Golden, CO 80401, USA

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

Article history: This paper presents an analysis of a novel wave energy converter concept that combines an oscillating
Received 8 October 2015 surge wave energy converter (OSWEC) with control surfaces. The control surfaces allow for a variable
Received in revised form device geometry that enables the hydrodynamic properties to be adapted with respect to structural
4 April 2016
loading, absorption range and power-take-off capability. The device geometry is adjusted on a sea state-
Accepted 7 April 2016
to-sea state time scale and combined with wave-to-wave manipulation of the power take-off (PTO) to
provide greater control over the capture efficiency, capacity factor, and design loads. This work begins
with a sensitivity study of the hydrodynamic coefficients with respect to device width, support structure
Keywords:
Oscillating surge wave energy converter
thickness, and geometry. A linear frequency domain analysis is used to evaluate device performance in
Variable structures terms of absorbed power, foundation loads, and PTO torque. Previous OSWEC studies included nonlinear
Structural load control hydrodynamics, in response a nonlinear model that includes a quadratic viscous damping torque that
Linearization was linearized via the Lorentz linearization. Inclusion of the quadratic viscous torque led to construction
of an optimization problem that incorporated motion and PTO constraints. Results from this study found
that, when transitioning from moderate-to-large sea states the novel OSWEC was capable of reducing
structural loads while providing a near constant power output.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction proposing a WEC concept that combines an oscillating surge wave


energy converter (OSWEC) with control surfaces. The concept for
Over the past decade, the global marine and hydrokinetic en- the control surfaces is similar to wind turbine airfoils and will
ergy sector has experienced a resurgence in the funding and incorporate the demonstrated success of wind turbine controls [3]
manpower allocated towards research and development, including developed at the National Renewable Energy Laboratory (NREL).
wave energy converter (WEC) conceptual research. As a whole, the The concept of controllable airfoils applied to wave energy con-
field of wave energy continues to host a wide diversity of tech- version has been recently pioneered by Atargis Energy Corpora-
nologies ranging from concept- to prototype-scale devices. How- tion's cycloidal device [4]. However, the device proposed in this
ever, the high cost of energy highlights the need for a structured paper suggests geometry control applied on a slower time scale
innovative approach to the development of WECs in hopes of similar to [5] which examined a bottom fixed pitching WEC where
achieving an optimal convergence in design and operation [1]. the main structure is comprised of a single large rotatable body. The
Recent studies on WEC system designs have shown that the resonance frequency of the pitching wave absorber in Ref. [5] was
development of advanced control methods that actively tune de- further tuned with ballast water; however, the design presented in
vice performance to maximize energy generation in operational this work looks to achieve this through additional geometric
conditions, while minimizing hydrodynamic loading in extreme changes. A preliminary investigation into the feasibility of the
sea states, is a necessary step in advancing wave energy technol- design [6] showed that rated power production levels could be
ogies towards commercial viability [2]. maintained over a wider range of sea conditions.
This paper attempts to address some of these issues by The development of nearshore OSWECs in recent years has been
led by Aquamarine Power's Oyster [7], AW-Energy Oy's Waveroller
[8], and Resolute Marine Energy's Surge WEC [9]. In addition,
* Corresponding author. Langlee Wave Power [10] and PolyGen Ltd are currently developing
E-mail addresses: nathan.tom@nrel.gov (N.M. Tom), Michael.Lawson@nrel.gov floating multi-flap OSWECs for deepwater deployment. However,
(M.J. Lawson), Yi-Hsiang.Yu@nrel.gov (Y.H. Yu), Alan.Wright@nrel.gov (A.D. Wright).

http://dx.doi.org/10.1016/j.renene.2016.04.016
0960-1481/© 2016 Elsevier Ltd. All rights reserved.
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 411

all of these designs have a fixed geometrical body that relies on and from results comparing the theoretical power performance
control of the power take-off (PTO) system to further optimize against rectangular flaps [6]. The major axis of each flap is 2 m and
power capture. Previous studies have shown that the general the minor axis 1/3 m. These dimensions were chosen in order to
OSWEC design experiences large hydrodynamic loads that drive obtain a streamlined shape when the flaps are rotated parallel to
the structural design and material costs [11]. The novelty of the the direction of wave propagation, 4 ¼ p/2. The streamlined shape
proposed design is the ability to alter the surfaces normal to the is expected to minimize viscous effects that decrease device effi-
motion in the principle degree of freedom thereby reducing hy- ciency as a percentage of the wave power is now dissipated in
drodynamic pressure and corresponding loads. This ability allows eddies rather than absorbed by the PTO. This is a difficult physical
for continued operation in larger sea states where structural process to model and limits the accuracy of linear hydrodynamic
loading constraints would force current designs to switch into analysis. The dimension and number of flaps used in this analysis
survival mode. However, resonant uncontrolled point absorbers are will remain constant.
generally narrow banded with high extraction efficiencies only The starting dimensions used in this study can be found in
around a small frequency range about the resonance period. Kur- Table 1. The base dimensions are significantly smaller than other
niawan and Flocard [5,12] investigated the ability to tune the listed designs. For example the width and thickness are one fourth
resonance period of a pitching WEC by altering the rotational of those described in Ref. [15]. The authors acknowledge that the
inertia by shifting the internal mass distribution. A device with a design may assist in tuning the device hydrodynamic properties
controllable geometry would not only be able to shed loads, but but will add loading and bending stress concentrations in the flaps.
would also tune the hydrodynamic coefficients to match the device This will likely require a reduction in the width of the device or the
resonance period with the peak excitation period of the sea spec- addition of cross bracing. The issues with structural loading are
trum, thus broadening its effective operating range. currently being investigated and will ultimately determine if the
Section 2 of this paper begins with a detailed introduction of the design is economical; however, this paper focuses on the theoret-
device concept. The device geometry is not assumed to be two ical hydrodynamic performance.
dimensional; therefore, the sensitivity of the hydrodynamic prop-
erties to device width and support thickness is explored to guide
current design choices. The device performance is then evaluated
3. Hydrodynamic modeling
with a traditional linear frequency domain analysis. The analysis is
extended to incorporate both motion and PTO constraints resulting
The first step in the study of the proposed WEC was the ex-
in nonlinear performance. However, current OSWEC designs
amination of the hydrodynamic coefficients for the geometric
require considerations for nonlinear hydrodynamics [13], and as a
configurations used in this study. In this work four geometries will
result, a nonlinear quadratic viscous drag moment will be added to
be explored, though the hydrodynamic coefficients for select ge-
the system dynamics. In order to continue use of spectral tech-
ometries will be presented in this section to describe the trend in
niques, the quadratic damping moment is linearized using the
width and thickness. The three geometries presented in this section
Lorentz theorem. The linearization allows for the construction of a
correspond to opening the top flap, the top two flaps, and the top
nonlinear optimization problem that can be solved while incor-
three flaps to 4 ¼ 90. The flaps will be opened in sequential order
porating the same motion and PTO constraints applied in the
and no other combination of open and closed flaps are considered.
inviscid analysis.
The hydrodynamic coefficients were obtained from WAMIT version
7.0 [16] while varying the device width and thickness of the support
2. Device description
structure.

The authors believe that pursuing controllable geometries will


lead to increased power absorption while shedding peak loads for
continued operation in larger sea states. A conceptual depiction of 3.1. Width variation
the device is shown in Fig. 1. The general shape of an OSWEC design
can be simplified to a flat plate, but the concept under development In order to size the device, the width of each flap configuration
aims to replace the main body by a set of identical actuated flaps. was increased from the base dimension listed in Table 1. The width
The flaps can be adjusted to any desired angle about their center of of the device was increased from half to four times the water depth
rotation with the flap pitch angle, 4, measured positive clockwise while maintaining the same height and thickness. The pitch added
from the radial axis of the body, see Fig. 2. However, for this study it moment of inertia, m55, pitch wave radiation damping, l55, and
was assumed that the flaps were placed either in the closed or fully pitch wave-excitation torque, X5, nondimensional coefficients are
open configuration, which correspond to 4 ¼ 0 and 4 ¼ p/2 plotted in Fig. 3 for three separate flap geometries. The purpose of
respectively. This binary control of the flap pitch angle will reduce this sensitivity study was to observe if results converged towards a
the complexity of the required actuators. The hydrodynamic co- two-dimensional approximation as the width of the plate becomes
efficients were also found to be sensitive to pitch angle [6], which much larger than the wavelength, i.e. lw/w ≪ 1. The results for a
may lead to unwanted modeling uncertainties. The flaps may be solid rectangular plate were reported in Ref. [17] and can be used as
controlled independently, but will only be allowed to open in guidance for the 0-flap geometry; however, the width variation is
descending order starting with the top flap, labeled as 1 in Fig. 1, completed here to evaluate the results for the load shedding
which is located closest to the free surface. Therefore, flap 3 can geometries.
only be opened when both flap 1 and flap 2 are already open. This The pitch mass moment of inertia and pitch hydrostatic
prevents configurations where there may be a closed flap between restoring coefficient, C55, for a solid rectangular plate can be ob-
two open flaps. The geometric shape of the OSWEC will be allowed tained from:
to change; however, the mass will be evenly distributed throughout
 ! 
the device, and the pitch mass moment of inertia, I55, will remain rwth3 t 2
constant. For this study the structural mass density, rm, was set to I55 ¼ 1þ (1)
6 2h
half the fluid density, r, to act as a baseline case.
The flaps have elliptical cross-sections as suggested in Ref. [14]
412 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

Fig. 1. Solidworks rendering of the OSWEC. Left: Front view of closed configuration (0-flap geometry), Middle: Side view of fully open configuration (4-flaps geometry), Right: Front
view of fully open configuration (4-flap geometry).

Fig. 2. Coordinate system for the device under investigation.

Table 1
Geometric parameters for hydrodynamic modeling.
rwth2 g
Water Depth d 10.00 m Flap Minor Axis tf 0.33 m C55 ¼ (2)
OSWEC Height h 10.00 m Flap Major Axis hf 2.00 m 4
OSWEC Thickness t 0.33 m Flap Width wf 4.50 m
OSWEC Width w 5.00 m Side Support Width ws 0.25 m where w is the width (along the y-axis), t is thickness (along the x-
OSWEC Volume c 14.10 m3 Radial Center of Buoyancy rb 4.80 m axis), h is the height (along the z-axis), and g is gravitational ac-
celeration, while the mass density is still assumed uniformly
distributed. The pitch mass moment of inertia and pitch hydrostatic
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 413

Fig. 3. Variation in hydrodynamic coefficients with device width: w/h ¼ [0.5,1.0,2.0,4.0]. The legend in the plots on the top row correspond to the ratio of the device width to water
depth and is constant for the remaining plots. Left and right columns correspond to the following nondimensional values: pitch added moment of inertia, m55 , and pitch wave
 
damping, l55 ; wave exciting torque magnitude, jX5j*, and phase, f5 . The nondimensional values are given by: m55 ¼ m55 =I55 , l55 ¼ l55 =uI55 , jX5j* ¼ jX5j/rgh2w, f5 ¼ f5 =p, u* ¼ uh/g.

restoring coefficient of the proposed device are approximately 78% as additional flaps are closed it is evident there is a stronger
of a solid rectangular plate with the same primary dimensions. The nonlinear relation between the hydrodynamic coefficients and
mass properties of the device were obtained from models con- device width. The largest change in the hydrodynamic coefficients
structed using the Solidworks CAD software. In this section, as the occurs between the 1- and 2-flap geometries while there is an order
device is lengthened only along the y-axis, the pitch mass moment of magnitude reduction in the radiation coefficients between the 1-
of inertia and pitch hydrostatic restoring coefficient will scale lin- and 3-flap geometries. This is to be expected as opening each flap
early with width. may be approximately compared to a solid rectangular plate where
In the 3-flap configuration, Fig. 3e and f, the plots of the the distance between the top of the device and the free surface
nondimensional hydrodynamic coefficients are nearly equivalent, increases, i.e. submerged plate. As each additional flap is closed and
which indicates that the dimensional hydrodynamic coefficients as the width is increased the pitch added moment of inertia in-
and the mass properties both scale linearly with device width and creases and its variation with frequency becomes larger, which is
three dimensional effects are minimal. The space between open consistent with the observations made in Ref. [17]. As the pitch
flaps allow the fluid flow to pass relatively undisturbed; however, mass moment of inertia and pitch hydrostatic restoring coefficient
414 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

remain constant, regardless of flap orientation, the natural period 4. Performance modeling
of the device will increase with each additional flap closed. The
natural period of the device also increases with the device width It is common practice to calculate the response amplitude
because of greater growth in the added moment of inertia rather operator (RAO) to access the performance of a WEC. For an incident
than changes in the mass properties, which follows the same trends wave described by:
in Ref. [27].
  n o
The hydrodynamic pressure generated by the incident wave is 1 vfI 
greatest near the free surface while the vertical variation in pres-
hðx; tÞ ¼ <  ¼ < AeiðutkxÞ (3)
g vt z¼h
sure is frequency dependent. In shallow water conditions the ver-
tical pressure variation is nearly uniform; however, as the wave where h is the wave elevation, < signifies the real component, fI is
frequency increases the vertical pressure variation quickly de- the incident wave potential, A is the wave amplitude, i ¼ 1 is the
creases converging towards an exponential decay under deep wa- imaginary unit, u is the wave angular frequency, and k is the wave
ter conditions. As a consequence, the lower flaps contribute number. The time-harmonic response of the floating body, in the j-
proportionately less to the overturning torque and the pitch-wave th degree-of-freedom, is given by:
exciting torque will decrease with each flap opened while the dif-
ference between flap geometries increases with the wave fre-  
quency. The greatest difference in the pitch wave-exciting torque zj ðtÞ ¼ R xj eiut ;
coefficients, between a pair of flap geometries, occurs at interme-  
diate wave frequencies just before reaching the deepwater limit, z_ j ðtÞ ¼ R iuxj eiut ; (4)
near u* ¼ 1.1. As the width of the device increases the peak in the
 
pitch wave-excitation torque grows and moves to a lower wave
z€j ðtÞ ¼ R  u2 xj eiut
frequency, the importance of which is highlighted in Sec. 4.1.1. The
results shown in Fig. 3 indicate that the design is successful at both
reducing the pitch wave-exciting torque and radiation coefficients where xj is the complex amplitude of motion for the j-th degree-of-
leading to reduced structural loads and tuning of the natural fre- freedom. The resulting harmonic motion when allowing for six
quency. The flap heights used in this study were chosen to be degrees of freedom can be described by the following coupled
uniform; however, it is reasonable to expect that nonuniform flap system of linear differential equations:
heights may allow for the hydrodynamic coefficients to be reduced
by a constant percentage with each flap opening. 6 h
X   i
Cij  u2 Iij þ mij þ iulij xj ¼ Fi (5)
j¼1

where Iij is the generalized inertia matrix, mij is the added mass
3.2. Thickness variation matrix, lij is the wave damping matrix, Cij is the restoring matrix,
and Fi is the complex wave-exciting force.
The thickness of the side and bottom support frame, the same as
t in Fig. 1, is small when compared to both width and height. The
dimensions were initially chosen so as to focus on the effect of the
flaps on the hydrodynamics; however, it is expected that the 4.1. Linear one degree of freedom motion
thickness of the support will need to be increased to withstand
structural loads as well as provide space for the PTO and flap ac- The OSWEC will pitch about the origin, O, which is fixed to the
tuators. A width of 20 m was selected to evaluate the effect of sea bed. The other five modes of motion will be fixed, and the one
increasing the thickness of the side and bottom support structure. degree of freedom pitch equation of motion is given by:
The thickness of the side and bottom support frame was increased
along the x-axis. The flap thickness was not increased in order to x5 X5
minimize the blockage coefficient when the flaps were rotated ¼ (6)
A C55  u2 ðI55 þ m55 Þ þ iu l55 þ Bg
parallel to the oncoming waves. It is evident that increasing the
thickness of the flaps will cause the flap dimensions to become where C55 is the pitch restoring coefficient, Bg is the linear PTO
more circular and reduce the geometrical change with pitch angle. damping coefficient required for power extraction and X5 is the
As the thickness increases the additional mass will collect near the complex wave exciting torque per unit wave amplitude (F5 ¼ AX5).
seabed which causes the radial center of gravity to drop by 17% and The OSWEC hydrostatic restoring moment is given by:
34% while the pitch hydrostatic restoring coefficient increases by
11% and 30% for thicknesses of 0.75 m and 1.50 m respectively.
mh ðtÞ ¼  rcrb  mrg g sinz5 ðtÞ (7)
The sensitivity of the hydrodynamic coefficients with respect to |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl}
the support structure thickness has been plotted in Fig. 4 for the 1- C55
flap and 2-flap geometries. The nondimensional radiation co-
efficients decrease slightly with thickness indicating the growth in where r is the fluid density, c is the WEC displaced volume in calm
mass properties exceeds the hydrodynamics. However, the reduc- water, rb is the radial distance from the origin to the center of
tion is proportional to the growth in the pitch mass moment of buoyancy, m is the OSWEC mass, rg is the radial distance from the
inertia which increases by 6% and 17% over the 0.33 m thickness origin to the center of gravity, and g is the gravitational accelera-
providing little to no change in the dimensional radiation co- tion. Eqn. (7) can be linearized by assuming small rotational motion
efficients. The pitch wave-excitation torque coefficients are also and making the following approximation sinz5 z z5. This is nor-
influenced minimally with thickness. Therefore, the dimensions of mally an appropriate assumption if the pitch amplitude of motion is
the support structure will be determined predominantly from constrained to less than thirty degrees.The instantaneous absorbed
structural analysis rather than because of favorable hydrodynamics. power by the PTO is calculated from:
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 415

Fig. 4. Variation in hydrodynamic coefficients with device thickness: t/h ¼ [0.033,0.075,0.150]. The legend in the plots on the top row correspond to the ratio of the device thickness
to water depth and is constant for the remaining plots. Left and right columns plot the following nondimensional values: pitch added moment of inertia, m55 , and pitch wave

damping, l55 ; wave exciting torque magnitude, jX5j*, and phase, f5 . The nondimensional values are calculated in the same manner as provided in Fig. 3.

 2 time-averaged power per unit width, Pw, contained within a


P Bg _ 2  
2 x5  propagating wave can be shown to equal:
2
¼ 2
z5 ¼ Bg u  A  sin ðut þ QÞ
2
(8)
A A rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
1 g 2kh
Pw ¼ rgA2 tanh kh 1 þ : (13)
where j,j denotes the magnitude and Q is the phase of pitch mo- 4 k sinh 2kh
tion. As the current analysis is in the frequency domain, it is more
appropriate to report the time-averaged power (TAP) absorbed by where h is the water depth.
the PTO which is calculated as follows:
4.1.1. Constrained one degree of freedom motion
ZT  2  2
PTAP 1 x  1 x  The results from Sec. 4.1 may require response amplitudes that
¼ Bg u2  5  sin2 ðut þ QÞdt ¼ Bg u2  5  (9) violate the small oscillation assumption especially when resonance
A2 T A 2 A
0 is tuned to long waves [18]. The maximum power absorption under
motion constraints was explored in Ref. [19] which led to the
where T is the wave period. Eqn. (6) can be inserted into Eqn. (9) following expression:
allowing for the optimal PTO damping, Bgjopt, at each wave fre-
quency to be calculated. This results in the following expressions: c
Pmax 1 jX5 j2 h 2
i
¼ 1  Hð1  dÞð1  dÞ (14)
 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A 2 8 l55
 2
Bg opt C55  u2 ðI55 þ m55 Þ
¼ 1þ ; (10) where H(x) is the Heaviside step function and d is the ratio between
l55 ul55
the constrained to optimal rotational velocity which is given by:

PTAP 1 jX5 j2 1 ujx5 jmax 2l55


2 opt ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi; (11) d¼ ; (15)
A 4 l55 C55 u2 ðI55 þm55 Þ 2 A jX5 j
1þ 1þ ul 55

and jx5jmax is the maximum pitch amplitude of motion. If a strongly


where at resonance Bg ¼ l55 leading to the maximum absorbed constrained case is considered the following approximation to Eqn.
power: (14) can be made:
   
PTAP 1 jX5 j2 c
Pmax 1 jX5 j2 1 1 jx5 jmax 1
max ¼ (12) ¼ d 1  d ¼ u jX5 j 1  d : (16)
A2 8 l55 A2 4 l55 2 2 A 2

which is consistent with the well known results from Ref. [18]. The In taking the limit of d / 0, the maximum constrained power can
416 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

be shown to be:      
x x
c Xr1 ¼  u2 m15 < 5  ul15 J 5  <fX1 g þi
Pmax 1 jx j A A
¼ ujX5 j 5 max ; (17) |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
A2 2 A A
     
x x
which implies that for heavily constrained designs, power is   u2 m15 J 5 þ ul15 < 5  JfX1 g : (22)
maximized at the frequency with the greatest wave exciting torque A A
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
rather than at the resonance frequency [20]. B

This allows the time domain corollary of Eqns. (20) and (21) to
4.1.2. Limits on PTO damping be written with the given frequency dependent coefficients. The
Not only is it possible for the motion of the WEC to be con- maximum reaction force can then be solved using basic calculus
strained, but it is also possible to limit the maximum PTO rotational and is left in Appendix A for derivation. The result of the derivation
damping. In order to understand how this constraint will affect is the maximum reaction force occurs at integer multiples of the
absorbed power, it will first be assumed the maximum achievable following time:
damping is less than the optimum damping as defined by Eqn. (10).  
1 2½A B þ <fX3 gJfX3 g
The following ratio will then be defined as: t ¼ tan1 : (23)
2u B2 þ JfX3 g2  A2  <fX3 g2
Bg
 ¼ g < 1: (18)
Bg opt

This ratio can be inserted into Eqn. (11) leading to: 4.3. Results from linear inviscid analysis

PTAP 1 jX5 j2 1 To evaluate the performance of the active control surfaces, a


2 const: ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi: (19)
A 4 l55 1þg2 C55 u2 ðI55 þm55 Þ 2 linear inviscid frequency domain analysis was completed with a
1þ 2g 1þ ul55 width of 20 m and a thickness of 0.75 m. The frequency domain
performance for each configuration was obtained for wave periods
As seen from Eqn. (19), there is now a multiplier next to the term
between 5 and 20 s and wave amplitudes between 0.1 and 2 m. The
in the denominator that incorporates the body dynamics. This term
wave amplitude range was chosen such that the shallow water
reaches a minimum of one as g / 1 and grows proportional to 1/g
wave breaking limit was not invalidated over the range of wave
as g / 0. The effect of g on the absorbed power will be greater near
periods. The performance quantities of interest are the TAP, the
resonance, when the reactance term, C55  u2(I55 þ m55), is
maximum foundation reaction force, and the power-take-off tor-
minimized.
que which are plotted in Figures 6, 7, and 8, respectively. Further-
more, as discussed in Sec. 4.1.1 and 4.1.2, constraints on the
4.2. Foundation reaction force maximum pitch motion amplitude and PTO damping were
imposed to evaluate how performance was affected.
The foundation must handle the reaction forces necessary to fix The first case places a 30-degree limit on pitch motion and a
the WEC to the seabed. This influences the structural loading that 20 MN, m, s limit on the allowable PTO damping, which is nearly
might be used in a control algorithm to shed extreme loads. If equivalent to an unconstrained PTO torque for the given sizing and
centrifugal forces are neglected and the body remains symmetric sea states. Under these constraints, the maximum TAP increases as
about the y-z plane, refer to Fig. 2, the reaction forces in the surge, each additional flap is closed, see the left column of Fig. 6. This is
Xr1, and heave, Xr3, directions, per unit wave amplitude are given consistent with the case of constrained motion as a larger wave
by: exciting force is proportional to the TAP. As the 3- and 4-flap ge-
h i ometries have lower wave exciting forces, these configurations
AðXr1 þ X1 Þ ¼  u2 m15 þ iul15 x5 (20) require larger amplitudes of motion for efficient power harvesting.
A more interesting scenario occurs when the PTO damping limit is
dropped to 2 MN, m, s, which corresponds to the right column of
AðXr3 þ X3 Þ ¼  ðrc  mÞg ; (21) Fig. 6. In this scenario, the maximum power production occurs for
|fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} the 3-flap geometry; however, there are operating conditions
fm
where the limit on pitch motion is exceeded. As there are no
where m15 is the surge-pitch radiation added mass, l15 is the additional control measures, the PTO torque is the only method to
surge-pitch wave radiation damping, X1 is the surge-wave exciting actively limit pitch motion. As a result of decreasing the PTO
force per unit wave amplitude, and X3 is the heave-wave exciting damping limit, the 1-, 2-, and 3-flap configurations have a range of
force per unit amplitude wave, see Fig. 5 for the hydrodynamic inoperable sea states and continued operation requires tran-
coefficients. The static heave reaction force, fm, associated with the sitioning to the 4-flap configuration.
difference between weight and buoyancy is significant when The maximum foundation force, Fig. 7, is found about the
calculating peak forces, but it is not a dynamic force and cannot be resonance frequency while the PTO reaction torque, Fig. 8, has a
controlled through either the PTO or changing geometry. The dy- local maximum due to the pitch-displacement amplitude
namic forces will drive fatigue damage accumulation in the foun- constraint yet both quantities increase with the closing of each
dation rather than the static mean load. In very small sea states the additional flap. The difference between the maximum foundation
static heave force will dominate, but as the wave amplitude grows force between the 1- and 4-flap geometries is roughly double when
the dynamic component of the reaction forces will have a larger constraints are set to 30 and 20 MN, m, s,. When the maximum
contribution in the magnitude of the reaction force. The following limit on the PTO damping is reduced the 1- and 2-flap geometries
analysis will not include the static reaction force in the presentation are affected the most with a 50e75% reduction in the foundation
of the results and focus on the dynamic forces. The above equations force magnitude. In terms of the PTO reaction torque, the difference
can be broken into real and imaginary components: between maximums for the 1- and 4-flap-open configurations is
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 417

Fig. 5. Surge and Pitch hydrodynamic coefficients. The number in the legends correspond to different flap geometries: 1 flap, 2-flap, 3-flap, and 4-flap geometry. The nondi-
mensional values are given by: l15 ¼ l15 =urh2 wt , m15 ¼ m15 =rh2 wt , jX1j* ¼ jX1j/rgwh, and jX3j* ¼ jX3j/rgwt.

nearly a factor of 5. Given the 1-flap geometry resonance frequency function of pitch angle and potentially of the radial position. Ap is
is outside of the wave regime, it will require a much greater PTO the projected area of the flap normal to the rotational velocity of the
torque for maximum energy harvesting. When the PTO damping OSWEC, j,j is the absolute value, and r is the position of the drag
limit is reduced, all flap configurations produce maximum torque element along the radial axis. The moment generated by taking a
values within 20% of each other. differential slice of the OSWEC along the xy-plane is given by:
The analysis performed in Ref. [21] found that the structural
costs drove the levelized cost of energy. As the structural costs do  
1  
not scale linearly with size, it is desirable to reduce the structural dMd ðt; 4Þ ¼  rCD wf z_ 5 ðtÞz_ 5 ðtÞr 3 dr; (25)
2
mass as cost reductions will be larger than the corresponding po-
wer loss. One of the benefits of the variable geometry design is the where Md is the drag induced torque and dr is the differential unit
ability to shed loads as the wave climate increases while continuing along the radial axis. In order to obtain the total drag torque, the
to optimize power. For example, if the foundation structure is above equation must be integrated over the height of the OSWEC:
designed with a maximum load of 0.5 MN, the 1-flap geometry
could produce 150 kW up to a wave amplitude of 1 m. At this point,   ZH
1  
the geometry could transition to 2 flaps, which can operate up to a Md ðt; 4Þ ¼  rCD wf z_ 5 ðtÞz_ 5 ðtÞ r 3 dr; (26)
wave height of approximately 1.4 m while continuing to produce 2
0
the same power output. This pattern can continue for the 3-flap
geometry up to 1.6 m and 4-flap geometry up to the 2 m where Eqn. (26) can be integrated directly if the flaps are closed;
survival conditions may need to be implemented. This flexibility however, as the flaps are pitched the projected area will need to be
allows one to control the loading on the device to minimize adjusted. The change in projected area with flap pitch angle was
structural costs while maintaining rated power. modeled as:
Though the PTO or power conversion chain is not a significant
cost driver based on the analysis in Ref. [21], the cyclic loading due Ap ð4Þ ¼ 2wf ðajcos 4j þ bjsin 4jÞ; (27)
to the wave periodicity can result in significant fatigue damage.
Based on the narrowbanded assumption [22], the fatigue damage with the bounds of integration in Eqn. (26) set to:
will be proportional to the cube of the torque amplitude. Therefore,
in larger sea states the 1-flap configuration may produce more than r ± ¼ rci ±ajcos 4j þ bjsin 4j; (28)
twice as much energy as the 4-flap configuration, but will accu-
mulate up to 125 times more fatigue damage. This is significant as where rci is the radial position of the center of rotation for the i-th
no power can be produced given a PTO failure, and the device may flap. The total drag torque is the summation over the total number
not be serviceable for several days. Therefore, it is worth consid- of flaps:
ering a reduction in rated power by a third to gain a significant
increase in fatigue life, which can be met with a variable geometry Md ðt; 4Þ ¼ SN
i¼1 Mdi ðtÞ; (29)
device.
  r þ
rwf CD ð4Þz_ 5 ðtÞz_ 5 ðtÞr4   :
1
5. Viscous drag considerations Mdi ðtÞ ¼ (30)
8 r

The drag coefficient was assumed to be constant over the foil.


The previous analysis was completed under the inviscid
The difficulty in modeling is the choice of CD(4) as the projected
assumption. This simplification often leads to significant over
area is reduced, at most by a factor of 1/6. For this analysis the
prediction in device motion and absorbed power. The first attempt
following drag coefficients were used: CD(0) ¼ 1.9 and CD(p/2) ¼ 0.1.
at modeling the viscous drag torque begins with Morison's equa-
Often the rotational velocity in Eqn. (26) is subtracted by the un-
tion [23]:
disturbed fluid orbital velocity, as described in Ref. [24] thus a
  nonlinear time-domain model was constructed to ascertain the
1  
fd ðt Þ ¼  rCD ðr; 4ÞAp ðr; fÞr z_ 5 ðtÞr z_ 5 ðtÞ; (24) effect of the modeling approximation. Results showed minor dif-
2
ferences and will not be discussed further in this preliminary
where fd is the drag force and CD is the drag coefficient that is a analysis.
418 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

Fig. 6. Time averaged power in kW for each geometric configuration with different applied constraints. The left and right columns apply the following constraints to the PTO
damping and pitch motion amplitude: 20 MN, m, s and p/6 and 2 MN, m, s and p/6. Each row plots the variation in performance for a given geometry beginning with 1 flap and
each row corresponds to one additional flap opening.

5.1. Lorentz linearization work done over one wave cycle is the same for both linear and non-
linear expressions and can be represented by:
The frequency domain analysis described in this work cannot be
used to model nonlinear systems. However, if the nonlinear terms      2
2  
can be linearized, then the use of spectral techniques are still Lz_ 5 ¼ Lnl z_ 5 z_ 5 ; (31)
possible. For sinusoidal waves, the torque generated by the
nonlinear drag term can be substituted by a linear term using the
Lorentz linearization [25]. This method consists of ensuring the where 〈,〉 stands for the average over one wave cycle, L is the
linearized viscous damping coefficient, and Lnl is the nonlinear
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 419

Fig. 7. Maximum foundation reaction force in MN for each geometric configuration with different applied constraints. The left and right columns apply the following constraints to
the PTO damping and pitch motion amplitude: 20 MN, m, s and p/6 and 2 MN, m, s and p/6. Each row plots the variation in performance for a given geometry beginning with 1
flap and each row corresponds to one additional flap opening.

viscous damping coefficient. The result from integrating both terms maintained. The linearized viscous damping coefficient can now be
in Eqn. (31) over one wave cycle provides the following expression inserted into Eqn. (6) leading to:
for the linearized damping coefficient:
x5 X5
¼  : (33)
8ujx5 j A
C55  u2 ðI55 þ m55 Þ þ iu l55 þ Bg þ 8uj3x5pjLnl
L¼ Lnl : (32)
3p
The expression for the linearized viscous drag coefficient is A result of the substitution is the presence of the device
wave amplitude dependent allowing the nonlinear nature to be amplitude of motion on both sides of the equation. An iterative-
420 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

Fig. 8. PTO reaction torque in MN, m for each geometric configuration with different applied constraints. The left and right columns apply the following constraints to the PTO
damping and pitch motion amplitude: 20 MN, m, s and p/6, and 2 MN, m, s and p/6. Each row plots the variation in performance for a given geometry beginning with 1 flap and
each row corresponds to one additional flap opening.

zero solver is required and the response is no longer invariant to the 5.2. Results from linearized viscous analysis
incoming wave amplitude. It is possible to combine Eqns. (9) and
(33), with slight modifications, to form a constrained nonlinear The optimization problem outlined at the end of Sec. 5.1 was
optimization problem. The solution can be obtained using fmincon solved for each flap configuration for several wave amplitudes and
in MATLAB [26] similar to the procedure described in Ref. [27]. As a wave periods to quantify the nonlinear viscous effects. Each flap
result, the optimum PTO damping for power absorption can be configuration required a different set of hydrodynamic radiation
quickly obtained allowing for faster design iterations rather than and excitation coefficients as well as a nonlinear viscous drag co-
running multiple time-domain simulations. efficient. For each sea state the optimizer solved for the optimal PTO
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 421

damping that maximized the time-averaged power while main- for the 20 MN, m, s limit the maximum foundation force is always
taining the same constraints on the pitch displacement amplitude greatest for the 1-flap configuration, which is roughly twice as large
and PTO damping as used in the linear inviscid analysis. The set of as the 4-flap configuration when producing its peak power. As the
results from the analysis have been plotted in Figs. 9 and 10. In wave amplitude increases, the pitch amplitude runs up against the
terms of power production, at the lowest wave amplitude the motion constraint, and power is maximized at all frequencies for
resonant peaks for each flap configuration provide the optimal the 1-flap configuration though it is followed by the largest foun-
power for a given frequency range. The peaks are present as the low dation forces. When the PTO damping limit is reduced, the resonant
wave amplitudes allow the device to oscillate freely, without peaks are apparent at all wave amplitudes highlighting the benefit
encountering the motion constraint, while the pitch velocity is of transitioning between geometries. At the largest wave amplitude
small reducing the influence of the linearized drag term. However, presented, it is observed that the 1- and 2-flap configurations have

Fig. 9. Performance metrics for each geometric configuration for a wave amplitude, A, of 0.25 m and wave angular frequencies, s, over the range of [0.31.2] rad/s. The numbers in
the legend correspond to the 1-flap, 2-flap, 3-flap, and 4-flap geometry. The left and right columns apply the following constraints to the PTO damping and pitch displacement
amplitude: 20 MN, m, s and p/6, and 2 MN, m, s and p/6.
422 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

a range of inoperable sea states that requires transitioning to the 3- power performance between the inviscid and viscous analysis has
and 4-flap geometry for continued operation. been reduced. This is due to the fact that the motion constraint in
Two metrics to quantify the capture efficiency are plotted in the large seas states forces the linear inviscid model to shed excess
middle row of Figs. 9 and 10. The first is the nondimensional cap- power it would normally absorb because of greater motion am-
ture width Cw ¼ PTAP =Pw  w of the device that provides an effi- plitudes. Even when including the viscous drag, at the largest waves
ciency metric based on the incident wave power across the width of the pitch amplitude continues to run against the motion constraint.
the device. This has an upper limit of 25% across all flap geometries Because the pitch amplitude is capped and the frequency is set, the
and is greatest at the lowest wave amplitude. Conversely, the ratio power difference is driven by the optimal damping value. As such,
between the absorbed power and the maximum constrained po- in the largest sea state the linearized viscous model is able to
wer, R, increases with wave amplitude implying the difference in absorb between 60% and 80% of the constrained maximum, while

Fig. 10. Performance metrics for each geometric configuration for a wave amplitude, A, of 1 m and wave angular frequencies, s, over the range of [0.31.2] rad/s. The numbers in the
legend correspond to the 1-flap, 2-flap, 3-flap, and 4-flap geometry. The left and right columns apply the following constraints to the PTO damping and pitch displacement
amplitude: 20 MN, m, s and p/6, and 2 MN, m, s and p/6.
N.M. Tom et al. / Renewable Energy 96 (2016) 410e424 423

having a nondimensional capture width between 10% and 20%. Appendix A. Maximum Foundation Reaction Force Derivation
The required PTO damping and corresponding torque to obtain
the optimum power production per the given constraints are The time domain corollary of Eqns. (20) and (21) can be written
plotted in the third row of Figs. 9 and 10. The optimum PTO as:
damping for each flap geometry follows the general form of Eqn.
(10) but increases with wave amplitude in order to offset the effect fr1 ðtÞ ¼ AðA cosut  B sinutÞ; (A.1)
of the increased drag moment on the pitch rotational velocity. Near
the resonance frequency the damping profile has a slight local fr3 ðtÞ ¼ Að<fX3 gcosut  JfX3 gsinutÞ; (A.2)
maximum because of the motion constraint. It is also still evident
that though the 1-flap geometry maximizes power for most wave
fr ðtÞ ¼ fr1bi þ fr3 b
k; (A.3)
climates it requires a torque that is approximately double the 2-flap
geometry though the power is not doubled. It also appears that qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
torque for the 20 MN, m, s constraint cases is linearly proportional kfr k ¼ f 2r1 þ f 2r3 : (A.4)
to wave amplitude, while for the 2 MN, m, s limit the 1- and 2-
þ 2 2
will maximize jjfrjj and
flap-open configurations experience a decrease in peak torque. Thus, maximizing the quantity fr1 fr3
can be analytically solved using calculus. This is achieved by setting
the time derivative of the reaction force magnitude equal to zero
and solving for time.

v 2 h  i
6. Conclusion fr1 ¼ A2 u  A2 þ B2 sin2ut  2A B cos2ut ; (A.5)
vt
This paper has proposed a new wave energy converter device h 
v 2
concept that combines an OSWEC with active control surfaces. The fr3 ¼ A2 u  <fX3 g2 þ JfX3 g2 sin2ut
control surfaces allow for a variable geometry device that was vt
i
shown to be effective at reducing the hydrodynamic and PTO loads  2<fX3 gJfX3 gcos2ut : (A.6)
while increasing the capacity factor. Furthermore, the design was
shown to be effective at tuning the hydrodynamic characteristics to Setting the sum of the two equations above and collecting like
match the resonant frequency of the device to the dominant wave- terms leads to:
excitation frequency thus allowing for optimum power extraction
in low-to-moderate sea states. As the sea state transitioned from sin2ut 2½A B þ <fX3 gJfX3 g
¼ ; (A.7)
moderate-to-large, the design was shown to be capable of both cos2ut B2 þ JfX3 g2  A2  <fX3 g2
reducing foundation loads and PTO torque while providing a near
constant TAP. In severe sea states the device is designed to give up which can be solved for t*; however, because differentiation only
wave power in favor of reduced structural loading. The reduced solves for time when the change in force is zero the minimum re-
structural mass is expected to provide greater savings in material action force may be found. In this case, the second derivative test
and construction costs compared to reductions in maximum can be used to determine if the solution is a local minimum or
absorbed power. The current analysis has been performed under maximum. If a local minimum is found the local maximum can be
regular wave excitation while assuming the PTO consists of a found by adding T/4 to t* due to the periodicity.
constant and continuous resistive linear damping coefficient;
however, it is known that active control of the PTO system can be
References
used to further optimize power absorption and assist in load
shedding. [1] J.W. Weber, D.L. Laird, Structured innovation of high performance wave en-
Analysis of the novel OSWEC began with an investigation into ergy converter technology, in: Proc. of the 11th European Wave and Tidal
Energy Conference; Sept. 6e11, 2015. Nantes, France, 2015.
the sensitivity of the hydrodynamic coefficients with respect to
[2] W. Musial, M. Lawson, S. Rooney, Marine hydrokinetic Technology (MHK)
device width, thickness of the support structure, and flap geometry. Instrumentation, Measurement, and Computer Modeling Workshop, National
Evaluation of device performance was completed using traditional Renewable Energy Laboratory, 2013. Technical report NREL/TP-5000e57605.
linear frequency domain techniques. Combinations of motion and [3] N. Wang, K.E. Johnson, A.D. Wright, Comparison of strategies for enhancing
energy capture and reducing loads using LIDAR and feedforward control, IEEE
PTO constraints were applied in an effort to understand how the Trans. Control Syst. Technol. 21 (3) (2013) 1129e1142.
power performance for each geometric configuration would be [4] S.G. Siegel, C. Fagley, M. Romer, T.E. McLaughlin, Experimental wave cancel-
effected. This highlighted the issues with solely maximizing power lation using a cycloidal wave energy converter, in: Proc. of the 9th European
Wave and Tidal Energy Conference; Sept. 5e9, 2011. Southampton, United
absorption as it corresponded to greater foundation loads and PTO Kingdom, 2011.
torque, which will require increased structural mass and a greater [5] A. Kurniawan, T. Moan, Characteristics of a pitching wave absorber with
levelized cost of energy. The analysis was further improved by rotatable flap, Energy Procedia 20 (2012) 134e147.
[6] N. Tom, M. Lawson, Y.H. Yu, A.D. Wright, Preliminary analysis of an oscillating
introducing nonlinear hydrodynamics namely a quadratic viscous surge wave energy converter with controlled geometry, in: Proc. of the 11th
drag torque. The nonlinear model was linearized using the Lorentz European Wave and Tidal Energy Conference; Sept. 6e11, 2015. Nantes,
theorem allowing for the construction of an optimization problem France, 2015.
[7] T. Whittaker, M. Folley, Nearshore oscillating wave surge converters and the
that incorporated both motion and PTO constraints. The results development of Oyster, Philos. Trans. R. Soc. A 370 (2012) 345e364.
highlighted the ability of the device to adapt to various wave cli- [8] J. Lucas, M. Livingstone, M. Vuorinen, J. Cruz, Development of a wave energy
mates in terms of power optimization, increase in capacity factor, converter (WEC) design tool - application to the WaveRoller WEC including
validation of numerical estimates, in: Proc. of the 4th International Conference
and shedding of hydrodynamic loads. Future work will be required
on ocean Energy; Oct. 17-19, 2012. Dublin, Ireland, 2012.
to evaluate the structural and flap actuator requirements in order to [9] E. Ramudu, Ocean wave energy-driven desalination systems for off-grid
determine final design feasibility. Furthermore, it will be necessary coastal communities in developing countries, in: Proc. of the IEEE Global
to combine nonlinear time-domain models with control of the PTO Humanitarian Technology Conference; Oct. 30-Nov. 1, 2011. Seattle, USA,
2011.
and device structure to investigate the potential for levelized cost of [10] J. Pecher, J. Kofoed, J. Espedal, S. Hagberg, Results of an experimental study of
energy improvement. the Langlee wave energy converter, in: Proc. of the 20th International offshore
424 N.M. Tom et al. / Renewable Energy 96 (2016) 410e424

and Polar Engineering Conference; June 20-25, 2010. Beijing, China, 2010. [19] D.V. Evans, Maximum wave-power absorption under motion constraints,
[11] P. Schmitt, S. Bourdier, D. Sarkar, F. Renzi, K. Dias, T. Doherty, T. Whittaker, Appl. Ocean Res. 3 (4) (1981) 200e203.
J. van’t Hoff, Hydrodynamic loading on a bottom hinged oscillating wave [20] M. Folley, T. Whittaker, A. Henry, The performance of a wave energy converter
surge converter, in: Proc. of the 22nd International offshore and Polar Engi- in shallow water, in: Proc. of the 6th European Wave and Tidal Energy Con-
neering; June 17-23, 2012. Rhodes, Greece, 2012. ference; Aug. 28-Sept. 2, 2005. Glasgow, United Kingdom, 2005.
[12] F. Flocard, T.D. Finnigan, Increasing power capture of a wave energy device by [21] Y.H. Yu, D.S. Jenne, R. Thresher, Reference Model 5 (RM5): Oscillating Surge
inertia adjustment, Appl. Ocean Res. 34 (2012) 126e134. Wave Energy Converter, National Renewable Energy Laboratory, 2015.
[13] A.J. Caska, T.D. Finnigan, Hydrodynamic characteristics of a cylindrical Technical report NREL/TP-5000e62861.
bottom-pivoted wave energy absorber, Ocean Eng. 35 (2008) 6e16. [22] Det Norske Veritas, Riser Fatigue, Det Norske Veritas, 2010. Technical report
[14] A. Kurniawan, T. Moan, Optimal geometries for wave absorbers oscillating DNV-RP-F204.
about a fixed axis, IEEE J. Ocean. Eng. 38 (1) (2013) 117e130. [23] J.R. Morison, J.W. Johnson, S.A. Schaaf, The force exerted by surface waves on
[15] D. Clabby, A. Henry, M. Folley, T. Whittaker, The effect of the spectral distri- piles, J. Petroleum Technol. 2 (5) (1950) 149e154.
bution of wave energy on the performance of a bottom hinged flap type ave [24] A. Babarit, J. Hals, M.J. Muliwan, T. Moan, J. Krokstad, Numerical bench-
energy converter, in: Proc. of the 31st international conference on ocean, marking study of a selection of wave energy converters, Renew. Energy 41
offshore, and arctic engineering; July 1-6, 2012. Rio de Janerio, Brazil, 2012. (2012) 44e63.
[16] WAMIT Version 7.0 User Manual, 2013. http://www.wamit.com. [25] J.T. Zimmerman, On the Lorentz linearization of a quadratically damped force
[17] R.P.F. Gomes, M.F.P. Lopes, J.C.C. Henriques, L.M.C. Gato, A.F.O. Falc~ ao, The oscillator, Phys. Lett. A 89 (1982) 123e124.
dynamics and power extraction of bottom-hinged plate wave energy con- [26] MATLAB Release 2014a User Manual, 2014. http://www.mathworks.com.
verters in regular and irregular waves, Ocean Eng. 96 (2015) 86e99. [27] M. Folley, T. Whittaker, J. van’t Hoff, The design of small seabed-mounted
[18] D.V. Evans, A theory for wave-power absorption by oscillating bodies, J. Fluid bottom-hinged wave energy converters, in: Proc. of the 7th European Wave
Mech. 77 (1) (1976) 1e25. and Tidal Energy Conference; Sept. 11e13, 2007. Porto, Portugal, 2007.

You might also like