Ocean Engineering 143 (2017) 113–124

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Gyroscopic power take-off wave energy point absorber in irregular sea states MARK
a,⁎ b c d
Zili Zhang , Bei Chen , Søren R.K. Nielsen , Jan Olsen
a
Department of Engineering, Aarhus University, 8000 Aarhus, Denmark
b
Key Laboratory for Wind and Bridge Engineering, Hunan University, 410082 Changsha, China
c
Department of Civil Engineering, Aalborg University, 9000 Aalborg, Denmark
d
JOLTECH, Langbrogade 3, 6400 Sønderborg, Denmark

A R T I C L E I N F O A BS T RAC T

Keywords: The Gyroscopic power take-off (GyroPTO) wave energy point absorber has the operational principle somewhat
Gyroscopic power take-off similar to the so-called gyroscopic hand wrist exerciser. Inside the float of the GyroPTO, there is a mechanical
Rigid body dynamics system made up of a spinning flywheel with its spin axis in rolling contact to a ring. At certain conditions, the
Magnetic coupling ring starts to rotate at a frequency equal to the peak angular frequency of the wave excitation. In this
Irregular sea states
synchronized state, the flywheel is running at almost constant speed, so the generated power from the generator
Linear wave theory
becomes constant as well. In this paper, the performance of GyroPTO in irregular sea waves is investigated. To
improve the stability (synchronization) margin of the device, a magnetic coupling mechanism has been added
between the spin axis and the flywheel, which also makes the semi-active control of the device possible.
Theoretical modeling of the GyroPTO is carried out using analytical rigid body dynamics, and a 4-DOF
nonlinear model is established. Further, linear wave theory has been applied to calculate the hydrodynamic
moments acting on the float. Rational approximation is performed on the frequency response function of the
radiation damping moments, leading to an extended state vector formulation of the coupled structure- wave
system. Simulation results show that magnetic coupling successfully improves the stability of the flywheel in
irregular sea states. With given significant wave height and peak frequency, it is shown that the synchronization
of the device is more easily obtained in narrow-banded wave than in broad-banded wave. As a result, larger
values of generator gain can be chosen in narrow-banded waves, leading to larger power output. Making use of
both the generator and the magnetic coupling, semi-active control algorithm might further improve the
performance of the GyroPTO in real sea.

1. Introduction
As an important renewable energy source, wave energy has recently
received significant attention in energy and policy agendas. A wave
energy converter (WEC) is defined as a dynamic system for converting
the energy in waves into mechanical energy stored in the oscillating
system. Several types of WEC devices have been proposed Falnes
(2002), leading to some commercial WEC projects that use different
buoy concepts, such as the Oscillating-water-column (OWC) plants, the
Pelamis WEC Pelamis, 2004, overtopping WEC type like the Wave
Dragon Wave Dragon, 2010, the SEAREV device Ruellan et al. (2010),
and point absorber approaches as used for the Wavestar device Wave
Star, 2003; Sichani et al. (2014).
Among others, point absorbers are the most widely investigated
WEC devices. A point absorber is a WEC that is capable of absorbing
energy from waves propagating in any direction, and with horizontal
dimensions much smaller than the dominating wave length Nielsen
et al. (2013). Different types of point absorber have been proposed,
such as the Wavestar point absorber and heave absorber Nielsen et al.
(2013). The device is typically equipped with an electric power
generator via a hydraulic force system. The reaction forces from the
hydraulic system are usually used to actively control the motion of the
point absorber in such a way that a maximum mechanical energy is
supplied to the absorber. With a certain loss due to friction in the
actuator, the control forces are then transferred to the generator, where
they are converted into electric energy. For almost all point absorbers,
the instantaneous absorbed energy varies significantly with time,
making the expensive additional power electronics mandatory before
the power can be supplied to the grid. This motivates a search for an
alternative device which is able to deliver a more constant power to the
grid without introducing power electronics.
The Gyroscopic power take-off (GyroPTO) wave energy point
absorber is a possible solution Nielsen et al. (2015). The general idea
is to use the gyroscopic moment on a spinning flywheel inside a floating

⁎
Corresponding author.
E-mail addresses: zili_zhang@eng.au.dk (Z. Zhang), chamber@hnu.edu.cn (B. Chen), srkn@civil.aau.dk (S.R.K. Nielsen), JAO@joltech.dk (J. Olsen).

http://dx.doi.org/10.1016/j.oceaneng.2017.07.050
Received 28 February 2017; Received in revised form 21 June 2017; Accepted 17 July 2017
0029-8018/ © 2017 Elsevier Ltd. All rights reserved.
Z. Zhang et al.
a b
Spin axis
Fig. 1. The GyroPTO point absorber. a) Schematic view. b) Scaled model of the GyroPTO device in the lab.
structure, so that almost constant power output from the generator can
be obtained from the absorbed wave energy, as proposed by Kanki
et al., (2009, 2006) and Bracco et al., (2011, 2010).
The operational principle of the GyroPTO device is somewhat
similar to the so-called gyroscopic hand wrist exerciser Powerball
Gulick and O'Reilly (2000). As shown in Fig. 1a, it consists of a float
rigidly connected to a lever. In the other end the lever is supported by a
hinge, which allows for rotations around a horizontal axis and a vertical
axis. Inside the float is a mechanical system made up of a ring, a
spinning flywheel and a generator. The ring is free to rotate in a plane
orthogonal to the lever, while the spin axis of the flywheel is supported
by a track in the ring with a width slightly larger than the diameter of
the axis. This track forms a guidance for the precession of the spin axis,
which is assumed to roll on the inner side of the track during rotations
of the ring without slip. To start up the device, the generator of the
device is used as an electric motor, with which the flywheel can be
accelerated to a prescribed spinning speed. In operational conditions,
the wave induced pitch and roll motions of the float produce a time-
varying rotation of the ring, which combined with the spinning velocity
of the flywheel creates a gyroscopic moment. This moment produces
the necessary contact force between the spin axis and the inner side of
the track, to provide the friction force making the rolling of the spin
axis possible. Therefore, the gyroscopic moment enforces a kinematical
constrain between the rotational velocities of the spin axis and the ring.
When synchronization of the angular frequency of the ring to the peak
frequency of the wave loading takes place, the responses of the ring and
the flywheel become almost harmonic. This phenomenon is the basic
reason for the functioning of the system. At synchronization, this
means that the generated electric power becomes almost constant in
time, making the need for additional power electronics unnecessary
before the power can be supplied to the grid. In our previous study
Nielsen et al. (2015), it is shown that the GyroPTO device performs well
in monochromatic waves, and at synchronization the rotational speed
of the ring is equal to the wave angular frequency. However, synchro-
nization of the device is easily lost in irregular sea waves.
In this paper, the performance of the GyroPTO point absorber in
irregular sea waves is investigated. In order to improve the stability
(synchronization) margin of the device, a magnetic coupling mechan-
ism is added between the spin axis and the flywheel, which acts as a
linear viscous damping mechanism on the flywheel. Theoretical
modeling of the GyroPTO with magnetic coupling has been carried
out by means of analytical rigid body dynamics, and 4-degree-of-
freedom (4-DOF) nonlinear model is established. The damping con-
stant of the magnetic coupling and the generator gain turn out to be the
two design parameters that influence the stability of the device, which
in principle make the semi-active control of the device possible.
Furthermore, based on linear wave theory, the hydrodynamic moments
Ocean Engineering 143 (2017) 113–124
Lever
Float
Ring
Flywheel
acting on the float are calculated as superpositions from the quasi-
static buoyancy moment, the radiation moment and the external wave
excitation moment. Rational approximation is performed on the
frequency response function of the radiation damping moments,
resulting in an extended state vector formulation of the GyroPTO-wave
system. Simulation results show that the introduced magnetic coupling
mechanism successfully improves the stability of the GyroPTO in
irregular sea waves. With given significant wave height and peak
frequency, synchronization of the device is more easily maintained in
narrow-banded waves than in broad-banded waves. This means larger
values of the generator gain can be chosen in narrow-banded waves,
leading to larger power output. Finally, influence of the generator gain
and the damping constant of the magnetic coupling on the performance
of the GyroPTO has been evaluated, considering different values of the
significant wave height. This study also enables further investigation of
semi-active control of the GyroPTO device for optimal performance in
irregular sea waves.
2. Modeling of the GyroPTO point absorber with magnetic
coupling
Based on rigid body dynamics, a 4-DOF nonlinear model is to be
established for the GyroPTO device with magnetic coupling. The DOFs
are φ11(t ), φ31(t ), ψ (t ) and ω(t ), where φ11(t ) and φ31(t ) represent the
rotations of the lever and float (the external structure), ψ (t ) represents
the rotational angle of the ring and ω(t ) represents the rotational angle
of the flywheel.
2.1. Kinematic constrain due to the gyroscopic moment
As show in Fig. 1, the GyroPTO point absorber consists of a lever
rigidly connected to a float with a ring and a flywheel inside. In the
other end the lever is simply supported at a point O. The float of the
GyroPTO device is made up of two semi-spheres (diameter d)
connected with a cylindrical part (height c), see Fig. 2. The spin axis
of the flywheel has the radius r1, and the inner radius of the ring is r2. If
the ring is set in motion with an angular frequency (angular velocity)
ψ˙ (t ), a gyroscopic moment J14ω˙ (t )ψ˙ (t ) on the spinning flywheel is
generated according to the law of moment of momentum, where J14
is the polar mass moment of inertia and ω˙ (t ) is the angular spin
frequency of the flywheel. This moment produces the necessary contact
force between the spin axis and the inner side of the track, providing
the friction force to make the rolling of spin axis possible. With the sign
definitions given in Fig. 3, the no-slip rolling of the spin axis on the
track implies the following non-holonomic constrain between the
angular frequencies of the spin axis and the ring:
114
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

Fig. 2. Geometry of the GyroPTO point absorber and definition of the (x11, x 21, x31) -coordinate system fixed to the float.

Fig. 3. Rolling contact between the precessing ring and the spinning flywheel due to the gyroscopic moment.

θ˙(t ) = Nψ˙ (t ) (1) always orthogonal to the plane of the ring. Furthermore, the mass of
r2 the lever and float is denoted m1, and the principal axis mass moment
where N = ≫ 1 is the gear ratio, θ˙(t ) is the angular frequency of the
r1 of inertia at rotations around the xj1-axis is denoted Jj1. Obviously,
spin axis. J11 = J31 due to the rotational symmetry of the device.
Due to the constraint in Eq. (1), the internal dynamics of the device Under sea waves, the lever and float is undergoing rotational
can be described by two degrees of freedom ψ (t ) and ω(t ), rather than motion with the angular velocity vector φ˙ (t ). The components of φ˙ (t )
three. in the (x11, x 21, x31)-coordinate system are denoted φ˙11(t ), φ˙21(t ), φ˙31(t ),
respectively. No rotations take place around the x21-axis, so φ˙21(t ) ≡ 0 .
2.2. Rigid body dynamics of the lever and float Hence, φ11(t ) and φ31(t ) represent the degrees of freedom of the external
structure (the lever and float), as illustrated in Fig. 2.
Four different principal axes coordinate systems attached to four Correspondingly, the translational velocity vector v̇(t ) of G1 has the
rigid bodies moving relative to each other are to be introduced for following components in the (x11, x 21, x31)-coordinate system:
deriving the equations of motion of the device. The components of a
⎡ v˙1(t )⎤ ⎡ 1 ⎤
vector defined with regard to the kth principal axes coordinate system ⎢ 1 ⎥ ⎢−aφ˙3 (t )⎥
(x1k , x 2k , x3k ) are indicated with an upper k, where k = 1, 2, 3, 4 . ⎢ v˙21(t ) ⎥ = ⎢ 0 ⎥
⎢ 1 ⎥ ⎢ 1 ⎥
Therefore, the jth component of a vector v in the kth principal axes ⎢⎣ v˙3 (t ) ⎥⎦ ⎣ aφ˙1 (t ) ⎦ (2)
coordinate system is denoted as vjk. Furthermore, the mass of the kth
rigid body is denoted mk, and the mass moment of inertia of the kth Then, the kinetic energy of the lever and float T1 becomes:
rigid body around the axis xjk is denoted Jjk.
1 1 1 2 2 1 2 1 2
First, an (x11, x 21, x31)- principal axes coordinate system is introduced T1 = m ((v1̇ (t )) + (v31̇ (t )) ) + J11(φ1̇ 1(t )) + J31(φ3̇ 1(t ))
2 2 2
to be fixed to the float with origin at G1, where G1 is the mass center of 1 2 2
gravity of the lever and float as shown in Fig. 2. Due to the rotational = (J11 + m1a 2 )((φ1̇ 1(t )) + (φ3̇ 1(t )) )
2 (3)
symmetry of the device, the centerline of the lever forms a principal
axis chosen as x21, which is orientated in the direction away from O. G1
is placed on this centerline a distance a from O. In the referential 2.3. Rigid body dynamics of the precessing ring
position, the x31− axis is horizontal and orientated outward (Fig. 2). The
ring is precessing around the x21-axis, so the plane of the flywheel is Next, an (x12 , x 22, x32 )- principal axes coordinate system is introduced

115
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

Fig. 4. Definition of the (x12, x 22, x32 )-coordinate system fixed to the ring.

to be fixed to the precessing ring with origin at G2, as shown in Fig. 4. ⎡ ψ˙ 2(t )⎤ ⎡ 1 1 ⎤
⎢ 1 ⎥ ⎢ cosψφ˙1 (t ) − sinψφ˙3 (t )⎥
G2 is the mass center of gravity of the ring and is placed along the x21- ⎢ ψ˙ 2(t )⎥ = ⎢ ψ˙ (t ) ⎥
axis at a distance b from G1. G2 is also the mass center of gravity of the ⎢ 2 ⎥ ⎢ 1 1 ⎥
sub-structure made up of the spin axis and the flywheel. The x22-axis is ⎢⎣ ψ˙32(t )⎥⎦ ⎣ sinψφ˙1 (t ) + cosψφ˙3 (t ) ⎦ (6)
co-directional to the x21-axis, and the x12-axis is placed along the spin
axis which is rotated from the x11-axis by an angle ψ (t ) around the x22- Then, the kinetic energy of the ring T2 becomes:
axis. Therefore, ψ˙ (t ) denotes the angular velocity of the precessing ring. 1 2 1 2 2 1 2
T2 = m ((v1̇ (t )) + (v31̇ (t )) ) + J12(ψ1̇ 2(t ))
In a similar way, the mass of the ring is denoted m2, and the 2 2
principal axis mass moment of inertia at rotations around the xj2-axis is 1 2 1 2
1 + J22(ψ2̇ 2(t )) + J32(ψ3̇ 2(t ))
denoted Jj2. It is seen that J12 = J32 = 2 J22 , with J22 signifying the polar 2 2
moment of inertia of the ring. 1⎛1 ⎞ 2 2
= ⎜ J22 + m 2(a + b )2⎟((φ1̇ 1(t )) + (φ3̇ 1(t )) )
The translational velocity vector v̇(t ) of G2 has the following 2⎝2 ⎠
components in the (x11, x 21, x31)-coordinate system: 1
+ J22(ψ̇ (t ))2
2 (7)
⎡ v˙1(t )⎤ ⎡ 1 ⎤
⎢ 1 ⎥ ⎢−(a + b )φ˙3 (t )⎥
⎢ v˙2 (t ) ⎥ = ⎢
1
0 ⎥
⎢ 1 ⎥ ⎢ 1 ⎥ 2.4. Modeling of the magnetic coupling and the flywheel
⎢⎣ v˙3 (t ) ⎥⎦ ⎣ (a + b )φ˙1 (t ) ⎦ (4)
In stead of a rigid connection, the spin axis is connected to the
Using coordinate transformation, the components of the angular flywheel via a magnetic coupling mechanism, which allows for a
velocity vector φ˙ (t ) of the float in the (x12 , x 22, x32 )-coordinate system are relative rotation between the spin axis and the flywheel. The idea of
given by: introducing this mechanism is to reduce the fluctuation of ω˙ (t ), and
hence stabilize the synchronization of the GyroPTO device under
⎡ φ˙ 2(t )⎤ ⎡ φ˙ 1(t )⎤ ⎡ 1 ⎤
⎢ 1 ⎥ ⎡ cosψ 0 −sinψ ⎤⎢ 1 ⎥ ⎢ cosψφ˙1 (t ) − sinψφ˙3 (t )⎥
1 irregular sea waves. Fig. 5 shows the details of the coupling. The upper
⎢ φ˙ 2(t ) ⎥ = ⎢ 0 1
⎥⎢ 1 ⎥ =
0 ⎥ φ˙2 (t ) ⎢ 0 ⎥ circular plate (grey) is fixed to the spin axis, and the lower circular plate
⎢ 2 ⎥ ⎢ ⎢ ⎥ ⎢ 1 ⎥ (red) is fixed to the flywheel. The flywheel and the lower plate are not in
⎢⎣ φ˙32(t ) ⎥⎦ ⎣ sinψ 0 cosψ ⎦⎢⎣ φ˙31(t ) ⎥⎦ ⎣ sinψφ˙1 (t ) + cosψφ˙3 (t ) ⎦
1
(5) contact with the spin axis, so the spin axis is free to rotate relative to
the flywheel. A magnetic field is present between the plates, which
Correspondingly, the components (ψ˙12, ψ˙22, ψ˙32 ) of the angular velo- produces a torque Mc(t) on the spin axis and the flywheel. The
city vector ψ˙ (t ) of the ring in the (x12 , x 22 , x32 )-coordinate system magnitude of Mc(t) is proportional to the relative angular velocity of
become: the flywheel with respect to the spin axis, or the so-called slip
ω˙ (t ) − θ˙(t ):

a b

Fig. 5. The magnetic coupling. a) Technical details: the upper plate (grey) is fixed to the spin axis and the lower plate (red) is fixed to the flywheel. b) Schematic principle.

116
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

Fig. 7. Definition of the (x14, x 24, x34 ) -coordinate system fixed to the flywheel (including
Fig. 6. Definition of the (x13, x 23, x33) -coordinate system fixed to the upper plate of the the lower plate).
magnetic coupling.
1 3 1 2 2 1 3 2
T3 = m ((v1̇ (t )) + (v31̇ (t )) ) + J13(θ1̇ (t ))
2 2
1 2 1 2
Mc(t ) = cc(ω˙ (t ) − θ˙(t )) = cc(ω˙ (t ) − Nψ˙ (t )) (8) 3
+ J23(θ2̇ (t )) + J33(θ3̇ (t ))
3
2 2
1 2 2 1 2
As shown in Fig. 5, Mc is considered positive when acting in = m3(a + b )2((φ1̇ 1(t )) + (φ3̇ 1(t )) ) + J13(ψ1̇ 2(t ) + θ (̇ t ))
2 2
opposite direction of ω˙ (t ). Hence, this coupling acts as a linear viscous 1 2 2
+ J13((ψ2̇ 2(t )) + (ψ3̇ 2(t )) )
damping mechanism on the flywheel, with the damping constant cc 4
depending on the strength of the magnetic field. 1⎛1 ⎞ 2 2
= ⎜ J13 + m3(a + b )2⎟((φ1̇ 1(t )) + (φ3̇ 1(t )) )
In Fig. 6, an (x13, x 23, x33)- principal axes coordinate system is 2⎝2 ⎠
1
introduced to be fixed to the spin axis, with its origin placed at G2.
2 2
+ J13((ψ̇ (t ))2 + (cosψφ1̇ 1(t ) − sinψφ3̇ 1(t )) +4(cosψφ1̇ 1(t ) − sinψφ3̇ 1(t ))θ (̇ t ) + 2(θ (̇ t )) )
4
Along with the spin axis, the x13-axis is co-directional to the x12-axis.
(11)
The x22- and x32- axes are rotated into the x23- and x33- axes by the
rotational angle θ (t ) around the x12-axis. θ˙(t ) is the angular velocity of Finally, an (x14 , x 24 , x34 )- principal axes coordinate system is intro-
the spin axis, which follows the kinematic constrain in Eq. (1). The total duced to be fixed to the spinning flywheel with origin at G2, as shown in
mass of the spin axis and the upper plate is denoted m3, and the Fig. 7. The x14-axis is co-directional to the x12- and x13- axes. The x22-
principal axis mass moment of inertia around the xj3-axis is denoted and x32- axes are rotated into the x24- and x34- axes by the rotational
1
Jj3. Due to the symmetry property J23 = J33 = 2 J13, where J13 signifies the angle ω(t ) around the x12-axis. ω˙ (t ) is the angular velocity of the
polar moment of inertia of the spin axis and the upper plate. flywheel.
The translational velocity vector of the spin axis and upper plate is The mass of the flywheel (with the lower plate fixed to it) is denoted
identical to that of the ring and is given by Eq. (4). m4, and the principal axis mass moment of inertia around the xj4-axis is
Using coordinate transformation, the components of the angular 1
denoted Jj4. It is seen that J24 = J34 = 2 J14 , where J14 signifies the polar
velocity vector ψ˙ (t ) of the ring in the (x12 , x 22 , x32 )- and (x13, x 23, x33)- moment of inertia of the flywheel.
coordinate systems are related as: The translational velocity vector of the flywheel is identical to that
of the ring as given by Eq. (4).
⎡ ψ˙ 3(t )⎤ ⎡ ψ˙ 2(t )⎤ ⎡ ψ˙12(t ) ⎤
⎢ 1 ⎥ ⎡1 0 0 ⎤⎢ 1 ⎥ ⎢ ⎥ The components of ψ˙ (t ) in the (x12 , x 22, x32 )- and (x14 , x 24 , x34 )-coordi-
⎢ ψ˙ (t )⎥ = ⎢ 0 cosθ sinθ ⎥⎢ ψ˙ (t )⎥ = ⎢ cosθψ˙ (t ) + sinθψ˙ (t ) ⎥
3 2 2 2 nate systems are related in a similar way as in Eq. (9):
⎢ 2 ⎢
⎥ ⎣ ⎥ ⎢ 2 ⎥ ⎢ 2 3 ⎥
0 −sinθ cosθ ⎦⎢ 2 ⎥ ⎢
⎢⎣ ψ˙33(t )⎥⎦ ⎣ ψ˙3 (t )⎦ ⎣−sinθψ˙2 (t ) + cosθψ˙3 (t )⎥⎦
2 2
(9) ⎡ ψ˙ 4(t )⎤ ⎡ ψ˙ 2(t )⎤ ⎡ ψ˙12(t ) ⎤
⎢ 1 ⎥ ⎡1 0 0 ⎤⎢ 1 ⎥ ⎢ ⎥
3 3 3 ⎢ ψ˙ 4(t )⎥ = ⎢ 0 cosω sinω ⎥⎢ ψ˙ 2(t )⎥ = ⎢ cosωψ˙ 2(t ) + sinωψ˙ 2(t ) ⎥
Hence, the components (θ˙1 , θ˙2 , θ˙3 ) of the angular velocity vector ⎢ 2 ⎥ ⎢⎣ ⎥⎢ 2 ⎥ ⎢ 2 3 ⎥
0 −sinω cosω ⎦⎢ 2 ⎥ ⎢
⎢⎣ ψ˙34(t )⎥⎦ ⎣ ψ˙3 (t )⎦ ⎣−sinωψ˙2 (t ) + cosωψ˙3 (t )⎥⎦
2 2
θ˙(t ) of the spin axis and upper plate in the (x13, x 23, x33)-coordinate (12)
system become:
Correspondingly, the components (ω˙ 14 , ω˙ 24 , ω˙ 34 ) of the angular
⎡ ˙3 ⎤ ⎡ ψ˙12(t ) + θ˙(t ) ⎤
⎢ θ1 (t ) ⎥ ⎢ ⎥ ˙ (t ) of the flywheel in the (x14, x 24, x34 )-coordinate system
velocity vector ω
⎢ θ˙ 3(t )⎥ = ⎢ cosθψ˙ 2(t ) + sinθψ˙ 2(t ) ⎥ become:
⎢ 2 ⎥ ⎢ 2 3 ⎥
⎢⎣ θ˙ 3(t ) ⎥⎦ ⎢⎣−sinθψ˙ 2(t ) + cosθψ˙ 2(t )⎥⎦
3 2 3 (10) ⎡ ω˙ 4(t )⎤ ⎡ ψ˙12(t ) + ω˙ (t ) ⎤
⎢ 1 ⎥ ⎢ ⎥
⎢ ω˙ 24(t ) ⎥ = ⎢ cosωψ˙2 (t ) + sinωψ˙3 (t ) ⎥
2 2
Then, the kinetic energy of the spin axis and upper plate T3 ⎢ 4 ⎥ ⎢ ⎥
becomes: ⎢⎣ ω˙ 3 (t ) ⎥⎦ ⎢⎣−sinωψ˙22(t ) + cosωψ˙32(t )⎥⎦
(13)

Then, the kinetic energy of the flywheel T4 becomes:

117
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

T4 =
1 4 1 2 2 1
m ((v1̇ (t )) + (v31̇ (t )) ) + J14(ω̇14(t ))
2
Qψ (t ) = NMc(t ) = Ncc(ω˙ (t ) − Nψ˙ (t )) ⎫ ⎪

2 2 ⎬
1 2 1 2 Qω(t ) = − Mc(t ) − Mg(t ) = − cc(ω˙ (t ) − Nψ˙ (t )) − Mg(t )⎭ ⎪

(19)
+ J24(ω̇ 24(t )) + J34(ω̇ 34(t ))
2 2
1 2 2 1 2 Further, the reaction of the generator torque on the stator, and
= m 4(a + b )2((φ1̇ 1(t )) + (φ3̇ 1(t )) ) + J14(ψ1̇ 2(t ) + ω̇ (t ))
2 2 hence on the float, is positive, when acting in the positive x12-direction.
1 2 2
+ J14((ψ1̇ 2(t )) + (ψ2̇ 2(t )) ) The components in the (x11, x 21, x31)-coordinate system follow from the
4
inverse coordinate transformation given in Eq. (5):
1⎛1 ⎞ 2 2
= ⎜ J14 + m 4(a + b )2⎟((φ1̇ 1(t )) + (φ3̇ 1(t )) )
2⎝2 ⎠ ⎡Q φ˙ 1(t )⎤ ⎡ ⎤
⎢ 1 ⎥ ⎡ cosψ 0 sinψ ⎤⎡ Mg(t )⎤ ⎢ cosψMg(t ) ⎥
⎢Q φ˙ 1(t ) ⎥ = ⎢ 0 ⎥⎢ ⎥=
1 2
+ J14((ψ̇ (t ))2 + (cosψφ1̇ 1(t ) − sinψφ3̇ 1(t )) +4(cosψφ1̇ 1(t ) − sinψφ3̇ 1(t ))ω̇ (t ) + 2(ω̇ (t ))2) 1 0 0 ⎥ ⎢ 0 ⎥
4 ⎢ 2 ⎥ ⎢−sinψ 0 cosψ ⎥⎢
⎢⎣Q φ˙ 31(t ) ⎥⎦ ⎣ ⎦⎣ 0 ⎦ ⎢⎣−sinψMg(t )⎥⎦
(14) (20)
In the present paper, an asynchronous generator with the following
characteristic is used Simoes and Farret (2015):
2.5. Equations of motion
Mg(t ) = cgω˙ (t ) (21)
The total kinetic energy of the GyroPTO device is obtained by
where cg is a gain factor, the value of which depends on the electronic
adding the contributions defined in Eqs. (3), (7), (11) and (14):
field between the rotor and the stator of the generator.
T = T1 + T2 + T3 + T4 Finally, inverting the kinetic energy given by Eq. (15) into the left-
1 2 2 1 hand side and the work-conjugated moments into the right-hand side
= J1[(φ1̇ 1(t )) + (φ3̇ 1(t )) ] + J2(ψ̇ (t ))2
2 2 of Eq. (17), the equations of motion of the 4-DOF model are obtained:
1 2
+ J3[(cosψφ1̇ (t ) − sinψφ3̇ (t )) + 4Nψ̇ (t )(cosψφ1̇ 1(t ) − sinψφ3̇ 1(t ))]
1 1
⎛ ⎛ 1 ⎞ ⎞ ⎛ 1 ⎞
2 ⎜J1 + ⎜J3 + J4⎟cos2ψ ⎟φ¨11 − ⎜J3 + J4⎟cosψ sinψφ¨31
1 2 ⎝ ⎝ 2 ⎠ ⎠ ⎝ 2 ⎠
+ J4[(cosψφ1̇ 1(t ) − sinψφ3̇ 1(t )) + 4ω̇ (t )(cosψφ1̇ 1(t ) − sinψφ3̇ 1(t ))
4 ⎛ 1 ⎞
+J32N cosψψ¨ + J4cosψω¨ − ⎜J3 + J4⎟(sin(2ψ )φ1̇ 1 + cos(2ψ )φ3̇ 1)ψ̇
+2(ω̇ (t ))2 ] ⎝ 2 ⎠
(15) −J32N sinψ (ψ̇ )2 − J4sinψωψ
̇ ̇ = cosψcg(t )ω̇ (t ) + Mφ 1(t ) (22)
1
where the kinematic constrain in Eq. (1) has been used. Further, the
following quantities have been introduced: ⎛ 1 ⎞ ⎛ ⎛ 1 ⎞ ⎞
−⎜J3 + J4⎟sinψ cosψφ¨11 + ⎜J1 + ⎜J3 + J4⎟sin2ψ ⎟φ¨31
⎝ 2 ⎠ ⎝ ⎝ 2 ⎠ ⎠
J1 = J11 + 2 J22 + 2 J13 + 2 J14 + m1a 2 + (m 2 + m3 + m 4 )(a + b )2 ⎫
1 1 1
⎪ ⎛ 1 ⎞
1 1 ⎪ −J32N sinψψ¨ − J4sinψω¨ −⎜J3 + J4⎟(cos(2ψ )φ1̇ − sin(2ψ )φ3̇ 1)ψ̇
1
J2 = J22 + (N 2 + 2 )J13 + 2 J14 ⎪ ⎝ 2 ⎠
⎬
1 3
J3 = 2 J1 ⎪ −J32N cosψ (ψ̇ )2 − J4cosψωψ
̇ ̇ = − sinψcg(t )ω̇ (t ) + Mφ 1(t ) (23)
⎪ 3
4 ⎪
J4 = J1 ⎭ (16)
1⎛ 1 ⎞ 2 2
J32N cosψφ¨11 − J32N sinψφ¨31 + J2ψ¨ + ⎜J3 + J4⎟[sin(2ψ )((φ˙11) − (φ˙31) )
As mentioned, φ11(t ), φ31(t ),
ψ (t ) and ω(t ) make up the four degrees of 2 ⎝ 2 ⎠
freedom of the system. +2cos(2ψ )φ˙11φ˙31] + J4ω˙ (sinψφ˙11 + cosψφ˙31) = Ncc (ω˙ (t ) − Nψ˙ (t ))
The equations of motion follow from the Lagrange equation Pars
(24)
and Treatise (1979):
⎫ J4cosψφ¨11 − J4sinψφ¨31 + J4ω¨ − J4(sinψφ˙11 + cosψφ˙31)ψ˙
d ⎛ ∂T ⎞
⎜ ⎟ = Q φ 1(t ) ⎪ = − cc(ω˙ (t ) − Nψ˙ (t )) − cgω˙ (t )
dt ⎝ ∂φ˙ 1 ⎠
1
1
⎪ (25)
⎛ ⎞ ⎪
d ∂T ⎪ Mφ 1(t ) and Mφ 1(t ) are the external hydrodynamic moments work
⎜ ⎟ = Q φ 3(t )
dt ⎝ ∂φ˙ 3 ⎠ ⎪ 1 3
1
1
⎬ conjugated to φ11(t ) and φ31(t ), which will be described in the next
⎛ ⎞ ⎪ section.
d ∂T
⎜ ⎟ −
∂T
= Q ( t ) ⎪
dt ⎝ ∂ψ˙ ⎠ ∂ψ ψ
⎪
⎪ 3. Hydrodynamic moments in irregular sea states
d ∂T
( ) = Q ω (t ) ⎪
dt ∂ω˙ ⎭ (17)
Due to the symmetry property of the device, no self-induced
where Q φ 1(t ), Q φ 3(t ), Qψ (t ), Qω(t ) denote the external conservative and
1 1 couplings take place between the hydrodynamic moments Mφ 1(t ) and
non-conservative moments work-conjugated to the degrees of freedom 1
1 1
φ1 (t ), φ3 (t ), ψ (t ) and ω(t ), respectively. Mφ 1(t ), i.e., Mφ 1(t ) only depends on the degree of freedom φ11(t ) and
3 1
The generator has a stator fixed to the ring and a rotor fixed to the Mφ 1(t ) only depends on the degree of freedom φ31(t ). Assuming linear
3
flywheel. Hence, the power output depends on the angular spin velocity wave theory, Mφ 1(t ) and Mφ 1(t ) may be written as superpositions from
ω˙ (t ) of the flywheel relative to the ring. Moreover, the generator torque
1 3
the quasi-static buoyancy moment, the radiation moment and the
Mg(t) must be acting in the opposite direction of the spin, i.e. in the external wave excitation moment:
negative x14-direction. Assuming virtual rotations δψ (so δθ = Nδψ )
and δω , the virtual work (done by the non-conservative moments) Mφ 1(t ) = − k1φ11(t ) − Mr1(t ) + Me1(t ) (26)
1
related to δψ and δω can be written as:
Mφ 1(t ) = − k 3φ31(t ) − Mr3(t ) + Me3(t ) (27)
3
δW = Mcδθ − Mcδω − Mgδω
= NMcδψ − (Mc + Mg )δω (18) where k3 denotes the buoyancy stiffness coefficient at small rotation
φ31(t ) around the static equilibrium state. k1 is an artificial mechanical
Therefore, the work-conjugated moments Qψ (t ) and Qω(t ) in Eq. spring stiffness introduced to the device in order to prevent drift in the
(17) become: unsupported degree of freedom φ11(t ). Mr1(t ) and Mr3(t ) are the radiation

118
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

moments generated by the motions of the float in still water, which ⎡ y (t ) ⎤

⎢ d ⎥
remove mechanical energy by generating a wave train propagating ⎢ dt y(t ) ⎥ ⎡0⎤
away from the float. Me1(t ) and Me3(t ) are the wave excitation moments ⎢ 2 ⎥ ⎢0⎥
caused by the wave action. ⎢ d y (t ) ⎥ ⎢ ⎥
⎢ dt 2
⎥
z r j (t ) = ⎢ ⋮ ⎥ , br = ⎢ 0 ⎥
⎢⋮⎥
⎢ d n −2 ⎥ ⎢0⎥
⎢ n −2 y(t )⎥ ⎢⎣ 1 ⎥⎦
⎢ dt ⎥
3.1. Radiation moments and related rational approximations ⎢ d n −1 y(t )⎥
⎣ n −1
dt ⎦ (35)
The radiation moments Mr1(t ) and Mr3(t ) may be written in terms of
⎡ 0 1 0 ⋯ 0 0 ⎤
the differential-integro form as follows Faltinsen (1990): ⎢ 0 0 1 ⋯ 0 0 ⎥
⎢ ⎥
Ar j = ⎢ ⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⎥
Mrj (t ) = −Jhjφ¨ 1j (t ) − Mrj,0(t )
⎢ 0 0 0 ⋯ 0 1 ⎥
t ⎢⎣−qn −qn −1 −qn −2 ⋯ −q2 −q1⎥⎦ (36)
= −Jhjφ¨ 1j (t ) − ∫ hrjφ1̇ (t − τ )φ1j̇ (τ )dτ , j = 1, 3
−∞ j (28)
The output differential equation, Eq. (32), can also be written in the
where Jhj is the added mass moment of inertia at infinite high frequency vector form:
with respect to the degree of freedom φj1(t ), j=1,3. Mrj,0(t ) in the Mrj,0(t ) = pr zrj(t ), j = 1, 3
j (37)
convolution integral form is the radiation damping moment, where
hrjφ̇1(t ) is a causal impulse response function for the radiation moment where
j
brought forward by the angular velocity φ˙ 1j (t ), j = 1, 3. The correspond- pr = [pm pm−1 ⋯ p1 p0 0 ⋯ 0]
j (38)
ing frequency response function (FRF) is given by:
∞
Hrjφ˙1(ω) =
j
∫0 e−iωt hrjφ˙1(t )dt ,
j
j = 1, 3
(29) 3.2. Wave excitation moments

The convolution integral Mrj,0(t ) (the radiation damping moment) The sea-surface elevation process {η(t ), t ∈ R} is modeled as a zero-
can be replaced by an equivalent system of coupled first-order mean, stationary Gaussian process defined by the following one-sided
differential equations, which is solved along with the equations of JONSWAP spectrum Hasselmann et al. (1973):
motion of the 4-DOF GyroPTO model. This method is based on a
⎛ −4 ⎞
Hs2 β ⎛ ω ⎞ 5⎛ ω ⎞
−5
replacement of the actual FRF Hrjφ̇1(ω) by an approximating rational
j Sη(ω) = α γ ⎜⎜ ⎟⎟ exp⎜ − ⎜⎜ ⎟⎟ ⎟
function H͠ r φ̇1(ω), given in the form: ωp ⎝ ωp ⎠ ⎜ 4⎝ω ⎠ ⎟
j j ⎝ p ⎠ (39)
P (s ) where
H͠ rjφ̇1(s ) = , s = iω, j = 1, 3
j Q (s ) (30)
α= ⎫
0.0624
⎪
0.230 + 0.0336γ −
0.185
⎪ 1.9 + γ
P(s ) = p0 s m + p1s m −1 + ⋯ + pm −1s + pm ⎫
⎪
⎛ ⎛ ω − ω ⎞2 ⎞ ⎬
⎬ ⎪
β = exp⎜⎜ − 2 ⎜ σω ⎟ ⎟⎟ ⎪
n n −1 1 p
Q(s ) = s + q1s + ⋯ + qn −1,1s + qn ⎪ ⎭ (31) ⎝ ⎝ p ⎠
⎠⎭ (40)
Such a replacement is always possible, because the impulse where ω is the angular frequency in rad/s. σ = 0.07 for ω ≤ ωp and
response function is causal. The parameters p0 , p1 , …, pm −1 , pm and 2π
σ = 0.09 for ω > ωp . ωp = is the peak angular frequency with Tp being
q1, …, qn −1, qn of the polynomials P(s) and Q(s) are all real. The order of Tp

the rational function as given by (m, n ) can be chosen freely as long as the peak period. Hs is the significant wave height. γ is the so-called
m ≤ n − 1. Further, all poles must have negative real parts to ensure the peak enhancement parameter which controls the bandwidth of the
rational filter to be asymptotic stable. Then, Mrj,0(t ) may be obtained as spectrum.
output of the following system of differential equations: From linear random vibration theory, the two-sided auto-spectral
density functions of the wave excitation moments SMe Me (ω) and
1 1
d my d m −1y dy SMe Me (ω) can be written as Nielsen and Zhang (2017):
Mrj,0(t ) = p0 + p1 m −1 + ⋯ + pm −1 + pm y, j = 1, 3 3 3
dt m dt dt (32) 1
SMe Me (ω) = |Hejη(ω)|2 Sηη(ω) = |He η(ω)|2 Sη(ω), j = 1, 3
j j 2 j (41)
n n −1
d y d y dy where Hejη(ω), j = 1, 3, is the FRF relating the wave excitation moment
+ q1 n −1 + ⋯ + qn −1 + qny = φ˙ 1j (t ), j = 1, 3
dt n dt dt (33) Mej (t ) to the sea-surface elevation η(t ).
The hydrodynamic parameters and functions k3, Jh1, Jh3 , Hr1φ̇ 1(ω),
where y(t) is an auxiliary function, which cannot be related with any 3
Hr3φ̇ 1(ω), He1η(ω) and He3η(ω) have all been calculated by the program
physical interpretation. The filter differential equation, Eq. (33), may 3

be written in the following state vector form: WAMIT, which is based on the boundary element method WAMIT,
2011.
d From Eq. (41) and considering first-order wave theory, the realiza-
zr (t ) = Arj zrj(t ) + br φ˙ 1j (t ), j = 1, 3
dt j (34) tions of the stationary wave excitation moments Me1(t ) and Me3(t ) may
be obtained by the following random phase model Nielsen and Zhang
where (2017):

119
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

K
In the present paper, rational approximation of the order
Mej (t ) = ∑ 2 aj, k cos(ωk t + Φk ), ωk = k Δω, j = 1, 3 (m, n ) = (2, 3) is performed on both Hr3φ̇ 1(ω) and Hr1φ̇ 1(ω) using
k =1 (42) 3 1
Matlab toolbox Mathworks (2011), and the results for Hr3φ̇ 1(ω) are
3
where illustrated in Fig. 8. It is seen that even a rational filter with low order
(m, n ) = (2, 3) can lead to a good agreement. Similar results have been
aj , k = 2SMe Me (ωk )Δω = |Hejη(ωk )|2 Sη(ωk )Δω , j = 1, 3
j j (43) obtained for the rational approximation of Hr1φ̇ 1(ω).
1
Three different values of the peak enhancement parameter γ in the
K denotes the total number of harmonic components in the spectral
JONSWAP spectrum have been considered, and the corresponding
decomposition. a1, k and a3, k are the standard deviations of the
one-sided auto-spectral density functions are shown in Fig. 9. It is seen
harmonic components with angular frequency in the interval
that as γ increases from 1 to 50, the sea wave changes from broad-
]ωj , ωj + Δω]. Φk are the random phases, which are mutually indepen-
dent and uniformly distributed in [0, 2π ]. banded process to be narrow-banded process. For all three curves, the
areas below are the same, which correspond to the variance of the sea-
1
surface elevation and are equal to 16 Hs2 .
3.3. Extended state vector equation of the structure-wave system
Fig. 10 shows the generated realizations of the wave excitation
moments Me1(t ) and Me3(t ) using Eq. (42) with K=4000, when the peak
Combining Eqs. (22), (23), (24), (25), (26), (27), (34) and (37), the
enhancement parameter γ = 5. As expected, the wave excitation in this
following extended state vector equation of dimension Nst = 8 + 2n is
case is neither very narrow-banded nor very broad-banded. It is also
obtained:
seen that Me3(t ) is slightly more narrow-banded than Me1(t ), owing to
d the properties of the FRFs He1η(ω) and He3η(ω) that have been calculated
z(t ) = g(z(t ), Me1(t ), Me3(t ), t )
dt (44) from WAMIT. Similarly, realizations of the wave excitation moments
corresponding to γ = 1 and γ = 50 can be generated in a same manner
where
(not shown here), which turn out to be broad-banded and narrow-
⎡ y( t ) ⎤ banded, respectively.
⎢ ⎥
⎢ y˙(t ) ⎥ T
z(t ) = ⎢ z (t ) ⎥ , y(t ) = [φ11(t ), φ31(t ), ψ (t ), ω(t )]
r1 4.2. Performance of the GyroPTO device under irregular sea states
⎢ ⎥
⎣ zr3(t )⎦ (45)
With fixed geometry of the scaled model, there are two design
⎡ f(z(t ), Me (t ), Me (t ), t )⎤ parameters that can be varied, namely the damping constant cc of the
⎢ 1 3
⎥
magnetic coupling and the generator gain cg. Both parameters influ-
g(z(t ), Me1(t ), Me3(t ), t ) = ⎢ Ar1zr1(t ) + br1φ˙1 (t ) ⎥
1
⎢ ⎥ ence the response and stability of the system. Synchronization of the
⎢⎣ Ar3zr3(t ) + br3φ˙31(t ) ⎥⎦ flywheel to the peak angular frequency of the irregular wave can only be
(46)
achieved within certain ranges of cc and cg values.
y(t ) is the 4 degrees of freedom of the GyroPTO device. zr1(t ) and zr3(t )
Fig. 11 shows the responses φ11(t ) and φ31(t ) of the float under wave
are the state vectors from the rational approximation of the radiation excitation moments Me1(t ) and Me3(t ) as presented in Fig. 10, i.e.
damping moments. g is an Nst dimensional nonlinear vector function, Hs=0.283 m, γ = 5. The values of cc and cg are chosen as cc=0.052
and f is an 8 dimensional nonlinear vector function. Nms and cg = 3.5 × 10−4 Nms. It is seen that the horizontal motion φ11(t )
of the float has a slightly larger amplitude than the vertical motion
4. Numerical example φ31(t ). In the frequency domain, a clear spectrum peak around 1.3 rad/s
is observed for φ11(t ) as shown in Fig. 11b, corresponding to the
4.1. The scaled model and the irregular wave excitation eigenfrequency of the float in the horizontal direction due to the
artificial spring. Actually, the eigenfrequency of the float in the
A scaled model of the GyroPTO device as shown in Fig. 1b is horizontal direction can be approximately calculated as
considered in the numerical simulation. This scaled model was ω1 ≈
k1
= 1.29 rad/s, where k1 is the spring stiffness, J1 is the
J1 + Jh1
constructed and tested in the wave basin (under regular waves) at
Aalborg University with a fully equipped device which included the equivalent mass moment of inertia as given in Eq. (16), Jh1 is the added
generator and power electronics Kramer et al. (2015). An “off the shelf” mass moment of inertia at infinite high frequency. It should be noted
generator was used and it was shown to perform as expected. The that the indicated angular frequency is merely a rough estimation of the
eigenfrequency for small horizontal vibration of the float. For both φ11(t )
structural and hydrodynamic parameters have been indicated in
Table 1. The numerical time-integration of the differential equations and φ31(t ), peaks around 3.14 rad/s are observed in the frequency
in Eq. (44) has been performed by a standard 4th order Runge-Kutta domain, resulting from the wave excitation, since the peak angular
scheme. frequency ωp of the irregular wave is 3.14 rad/s in this case.
The corresponding responses of the ring, the flywheel as well as the
Table 1
power output are shown in Fig. 12. In this case (cc=0.052 Nms,
Structural and hydrodynamic parameters of the scaled GyroPTO device. cg = 3.5 × 10−4 Nms), synchronization of the GyroPTO device is
obtained with respect to the peak angular frequency ωp of the wave.
Parameter Value Unit Parameter Value Unit The ring is rotating with an angular velocity around 3.14 rad/s equal to
ωp, although this angular velocity is oscillating between 2 rad/s to
a 0.9891 m m4 20.75 kg
b 0.0761 m J11 21.68 4 rad/s. This means the proposed magnetic coupling mechanism
kg m2
c 0.080 m J22 0.0952 successfully enables the synchronization of the ring to the wave
kg m2
d 0.550 m J13 0.0009208
excitation. Due to the kinematic constrain Eq. (1), the flywheel is
kg m2
r1 0.009 m J14 0.3671
spinning at almost constant angular velocity 3.14 × N = 71 rad/s,
kg m2
r2 0.203 m k1 146.7 Nm/rad
although oscillation also takes place from 60 rad/s to 75 rad/s, as
m1 20.27 kg k3 2945 Nm/rad shown in Fig. 12b. The resulting power output P(t) from the generator,
m2 5.270 kg Jh1 14.39 kg m2 which is calculated as P(t ) = cgω˙ (t ), is shown in Fig. 12c. It is clearly
m3 1.963 kg Jh3 32.16 kg m2 seen that the generated power output becomes almost constant over
time, with the mean value about 2.0 W. This is indeed the biggest

120
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

a b

Fig. 8. Rational approximation of order (m, n ) = (2, 3) to the radiation moment FRF Hr 3φ˙ 1(ω) . a) Real part of Hr 3φ˙ 1(ω) . b) Imaginary part of Hr 3φ˙ 1(ω) .
3 3 3

16.7 W, which is plotted as a red line in Fig. 12c. Comparison of this

Fig. 9. One-sided auto-spectral density function of the sea-surface elevation process for
γ = 1, 5, 50 . Hs=0.283 m, ωp=3.14 rad/s.
advantage of the GyroPTO device as mentioned in Section 1, that
constant power can be delivered to the grid without additional power
electronics. Furthermore, it should be mentioned that the instanta-
neous power input to the float is completely related to the motion of the
float, and can be calculated as Pin(t ) = Me1(t )φ˙11(t ) + Me3(t )φ˙31(t ), which is
oscillating dramatically around zero axis from −200 to 400W (not
shown here). The mean value Pin of the instantaneous power input is
red line with the power output P(t) implies that the efficiency of the
GyroPTO device is not very high. However, the clear advantage of
constant positive power output from the GyroPTO makes it a compe-
titive solution. It might be possible to increase the efficiency by semi-
active control of the flywheel rotational speed using changing values of
cg and cc. Moreover, a significant improvement may be obtained by an
active control of the motion of the float using an extra actuator, similar
to the one used for the Wavestar device Sichani et al. (2014).
Fig. 13 shows the corresponding results, when the generator gain cg
is increased to be cg = 1 × 10−3 Nms, with all other conditions un-
changed as in Fig. 12. In this case, synchronization of the device is
maintained for only 400 s and then lost, i.e. the angular velocities of the
ring and the flywheel quickly drop to zero, so is the power output. On
one hand, as long as synchronization is maintained, the angular
velocity of the flywheel is always around 71 rad/s. Therefore, increasing
the value of cg is basically beneficial because the generated power
output is proportional to cg. However on the other hand, synchroniza-
tion of the device will be lost when cg is increased to a certain value. As
shown in Fig. 13c, during first 400 s of the time-series, the generated
power output is around 4 W, much larger than that in Fig. 12c because
cg is increased. But this large value of cg does not guarantee
synchronization (stability) of the device for a sufficiently long time

a b

c d

Fig. 10. Realizations of the wave excitation moments Me1(t ) and Me3(t ), Hs=0.283 m, ωp=3.14 rad/s, γ = 5. a) Time-series of Me1(t ) . b) Fourier amplitude of Me1(t ). c) Time-series of
Me3(t ) . d) Fourier amplitude of Me3(t ).

121
Z. Zhang et al. Ocean Engineering 143 (2017) 113–124

a b

c d

Fig. 11. Responses φ11(t ) and φ31(t ) of the float, Hs=0.283 m, ωp=3.14 rad/s, γ = 5, cc=0.052 Nms, cg = 3.5 × 10−4 Nms. a) Time-series of φ11(t ) . b) Fourier amplitude of φ11(t ) . c) Time-
series of φ31(t ) . d) Fourier amplitude of φ31(t ) .

duration, making the increased power output during the first 400 s
meaningless. Nevertheless, by properly changing the value of cg in real
time using a certain semi-active control law, synchronization might be
maintained. This will be investigated in the future research.
4.3. Parametric study
Parametric study is carried out in this subsection, where different
values of cc and cg are considered to evaluate the performance and
stability of the GyroPTO device in irregular sea waves with three
different bandwidths (γ = 1, 5, 50 ).
Fig. 14 shows the mean power output from the GyroPTO device as a
function of cc and cg. The significant wave height Hs and peak angular
frequency of the wave ωp are fixed. Each simulation has been carried
out with a time length of 2000 s, from which the mean power output P
is calculated. The mean power output is calculated and plotted in
Fig. 14 on condition that synchronization is maintained during this
2000 s. It is noticed that when cg is increased to certain large values,
synchronization is lost and P is not plotted anymore.
In Fig. 14a, the results for γ = 1 corresponding to very broad-
banded wave is illustrated. In this case, synchronization of the device is
hardly maintained. For each fixed value of cc (cc = 2 × 0.052 and
cc = 3 × 0.052 ), P increases monotonously with cg until cg reaches the
value of 1.5 × 10−4 , where synchronization is lost. The maximum mean

a b c

Fig. 12. Responses of ring and of the flywheel, and the power output. Hs=0.283 m, ωp=3.14 rad/s, γ = 5, cc=0.052 Nms, cg = 3.5 × 10−4 Nms. Synchronization maintained. a) Time-
series of the angular velocity of the ring ψ˙ (t ) . b) Time-series of the angular velocity of the ring ω˙ (t ) . c) Time-series of the power output, together with the mean value of the power input Pin
to the float (red line).

122
Z. Zhang et al.
a b
Fig. 13. Responses of ring and of the flywheel, and the power output. Hs=0.283 m, ωp=3.14 rad/s, γ = 5, cc=0.052 Nms, cg = 1 × 10−3 Nms. Synchronization lost. a) Time-series of the
angular velocity of the ring ψ˙ (t ) . b) Time-series of the angular velocity of the ring ω˙ (t ) . c) Time-series of the power output, together with the mean value of the power input Pin to the float
(red line).
power output can be obtained is about 0.62 W, which is very small
comparing with the mean power input (about 16 W, not shown here). It
should also be noted that synchronization is always lost when cc=0.052
and therefore the mean power output is not plotted. Fig. 14b and c
present the corresponding results when γ = 5 and γ = 50 , respectively.
Similarly, for a given value of cc, P increased almost linearly with
increased cg, but the stability margin is larger comparing with Fig. 14a.
On the contrary to Fig. 14a, for both γ = 5 and γ = 50 , synchronization
can be maintained when cc=0.052 and is more easily lost when
cc = 3 × 0.052 . The maximum mean power output is obtained at 1.6
W (when cc=0.052, cg = 3.5 × 10−4 ) for γ = 5, and at 3.5 W (when
cc=0.052, cg = 7.5 × 10−4 ) for γ = 50 , respectively. Therefore, both cc
and cg influence the stability and performance of the device, and the
value of them need to be optimized in different sea states. Further, the
general conclusion is that synchronization (stability) of the GyroPTO
device is more easily maintained in narrow-banded sea waves than
broad-banded case, leading to larger value of cg and thus larger power
output. The reason for this can be revealed from the auto-spectral
density functions in Fig. 9, where more energy of the sea wave is
concentrated around the peak wave frequency ωp for the narrow-
banded case. Only the wave energy around this angular frequency can
be absorbed by the flywheel since it's operating at the same angular
velocity. For broad-banded waves, energies are more evenly distributed
to other frequencies, which can not be absorbed by the GyroPTO
device.
a b
Fig. 14. Mean power output as a function of cc and cg. Hs=0.283 m, ωp=3.14 rad/s. a) γ = 1. b) γ = 5. c) γ = 50 .
123
Ocean Engineering 143 (2017) 113–124
c
Finally, four different values of the significant wave height Hs are
considered. Fig. 15 shows the maximum mean power output Pmax as a
function of Hs and cc, where the maximum possible cg for synchroniza-
tion is used. As expected, the power output from the GyroPTO device
increases monotonously as Hs increases, since the variance and thus
power of the sea wave is proportional to Hs2. Again, the device
performs much better in narrow-banded waves (γ = 50 ), where the
maximum mean power output of 18 W can be obtained when
Hs=1.132 m and cc=0.052 Nms.
5. Conclusions
The performance of the GyroPTO point absorber in irregular sea
waves is investigated in this paper. Inside the float of the GyroPTO
device, in stead of a rigid connection, the spin axis is connected to the
flywheel via a magnetic coupling mechanism. The idea of this mechan-
ism is to reduce fluctuation of the spinning velocity of the flywheel, and
hence stabilize the synchronization of the GyroPTO device. Nonlinear
equations of motion of the GyroPTO have been derived based on a 4-
DOF rigid body model, and the wave loading on the float is determined
from the first-order wave theory. The frequency response functions of
the two radiation damping moments are approximated by rational
functions, resulting in an extended state vector formulation of the
GyroPTO-wave system. The wave excitation moments are then gener-
ated by random phase model using the JONSWAP spectrum.
c
Z. Zhang et al.
a b
Fig. 15. Maximum mean power output as a function of Hs and cc, where the maximum possible cg is used. ωp=3.14 rad/s. a) γ = 1. b) γ = 5. c) γ = 50 .
It is shown from the simulation results that the proposed magnetic
coupling improves the performance of the GyroPTO in irregular sea waves.
The stability of the device is influenced by the generator gain cg and the
damping constant cc of the magnetic coupling, and synchronization is only
achieved within certain ranges of cg and cc. Synchronization of the
GyroPTO to the peak frequency of the wave is more easily maintained in
narrow-banded sea waves than in broad-banded sea waves. This implies
that larger values of cg can used in narrow-banded sea waves, leading to
larger power output. Future investigations will aim at developing semi-
active control algorithms based on cg and cc, to achieve optimal perfor-
mance of the GyroPTO point absorber in irregular sea waves. Moreover, in
the present case the motion of the float is completely induced by the
incoming wave. A significant improvement of the efficiency of the device
may be obtained by an active control of the motion of the float using an
extra actuator. This will on the other hand lead to larger energy consump-
tion for the actuation, and an optimal solution needs to be investigated.
Acknowledgements
The authors acknowledge the support from the Danish Energy
Technological Development and Demonstration Program (EUDP) for
financing the Project 2015-I 64015-0007 Gyro electric energy con-
verter theory and analysis. The support from the Chinese Scholarship
Council under the State Scholarship Fund, the National Science
Foundation of China (No. 51278189, No. 51422806) and the State's
Key Project of Research and Development Plan (No.
2016YFE0127900) are also greatly acknowledged.
References
Bracco, G., Giorcelli, E., Mattiazzo, G., 2011. Iswec: a gyroscopic mechanism for wave
power exploitation. Mech. Mach. Theory 46 (10), 1411–1424.
124
Ocean Engineering 143 (2017) 113–124
c
Bracco, G., Giorcelli, E., Mattiazzo, G., Poggi, D., Taylor, J., 2010. Iswec: experimental
tests on a small scale prototype model. In: Proceedings of the 3rd International
Conference on Ocean Energy (ICOE), Bilbao, Spain, 6 October.
Falnes, J., 2002. Ocean waves and oscillating systems: linear interactions including wave-
energy extraction. Cambridge University Press.
Faltinsen, O., 1990. Sea loads on ships and offshore structures. Cambridge University
Press.
Gulick, D.W., O'Reilly, O.M., 2000. On the dynamics of the dynabee. J. Appl. Mech. 67
(2), 321–325.
Hasselmann, K., et al., 1973. Measurements of Wind Wave Growth and Swell Decay
During the Joint North Sea Project (JONSWAP), Erganzungsheft zur Deutschen
Hydrograph. Z., Reihe A, No. 12, Hamburg.
Kanki, H., Arii, S., Furusawa, T., Otoyo, T., 2009. Development of advanced wave power
generation system by applying gyroscopic moment. In: Proceedings of the 8th
European Wave and Tidal Energy Conference, Uppsala, Sweden, pp. 280–283.
Kanki, H., Arii, S., Hata, T., 2006. Gyro dynamics for new wave power generation system.
In: Proceedings of the 7th IFToMM-Conference on Rotor Dynamics, 25-28
September 2006, Vienna, Austria.
Kramer, M.M., Pecher, A.F.S., Guaraldi, I., Andersen, M.T., Kofoed, J., 2015. Hydraulic
evaluation of joltech’s gyropto for wave energy applications, Tech. rep., Aalborg
University, DCE Technical Reports; No. 178.
Mathworks, Control System Toolbox 9.1, Mathworks, Inc., (2011).
Nielsen, S.R.K., Zhang, Z., 2017. Stochastic Dynamics. Aarhus University Press.
Nielsen, S.R.K., Zhou, Q., Kramer, M.M., Basu, B., Zhang, Z., 2013. Optimal control of
nonlinear wave energy point converters. Ocean Eng. 72, 176–187.
Nielsen, S.R.K., Zhang, Z., Kramer, M.M., Olsen, J., 2015. Stability analysis of the
gyroscopic power take-off wave energy point absorber. J. Sound Vib. 355, 418–433.
Pars, L.A., 1979. A Treatise on Analytical Dynamics. Ox Bow Press.
Pelamis Wave, 2004. 〈http://www.pelamiswave.com/〉.
Ruellan, M., BenAhmed, H., Multon, B., Josset, C., Babarit, A., Clement, A., 2010. Design
methodology for a searev wave energy converter. IEEE Trans. Energy Convers. 25
(3), 760–767.
Sichani, M.T., Chen, J.B., Kramer, M.M., Nielsen, S.R.K., 2014. Constained optimal
stochastic control of non-linear wave energy point converters. Appl. Ocean Res. 47,
255–269.
Simoes, M.G., Farret, F., 2015. Modeling and Analysis with Induction Generators Third
edition. CRC Press, Taylor and Francis Group.
WAMIT, Inc., 2011. Wamit user manual, version 7.0, 〈http://www.wamit.com/〉.
Wave Dragon, 2010. 〈http://www.wavedragon.net/〉.
Wave Star A/S, 2003. 〈http://www.wavestarenergy.com/〉.