Design of a non-linear power take-off simulator for model

testing of rotating wave energy devices

M. F. P. Lopes 1 , J. C. C. Henriques1 , Miguel C. Lopes1 , L. M. C. Gato1 and António Dente2

1
IDMEC - Instituto de Engenharia Mecânica,
Instituto Superior Técnico,
Universidade Técnica de Lisboa,
Av. Rovisco Pais,
1049-001 Lisboa, Portugal
E-mails: [email protected]; [email protected]; [email protected];
[email protected]

2 CIE3 - Center for Innovation in Electrical and Energy Engineering,

Instituto Superior Técnico,
Universidade Técnica de Lisboa,
Av. Rovisco Pais,
1049-001 Lisboa, Portugal
E-mail: [email protected]

Abstract 1 Introduction
Eddy current brakes provide a versatile way of simu- The design of the power take-off system (PTO) simu-
lating the power take-off system (PTO) in the model test- lator is one of the most important challenges in the de-
ing of wave energy converters at small scale. These are sign of a wave energy converter small scale model. This
based on the principle that a conductive material moving is enhanced when the specified PTO damping character-
perpendicularly to a magnetic field generates a braking istic is non-linear and different characteristic curves are
force proportional to its velocity. to be tested. For small scale models, the extracted energy
This was applied in the design of the PTO simulator is more conveniently converted to heat, as the extracted
of a bottom-hinged flap wave energy converter model, power scales down with the power of 3.5 of the length.
at 1/16 scale. The efforts put into the accurate dynamic Throttling processes are often used to convert pneumatic
simulation of the device led to the development of a con- or hydraulic power into heat. These methods have how-
trollable PTO simulator, which can be applied to other ever several limitations as the damping level is difficult
small scale rotating wave energy device models. to control and calibrate, and undesirable non-linearities
A special power source was built to provide the re- tend to occur.
quired controllable current intensity to feed the magne- This paper presents the design and calibration of
tic field generating coils. Different non-linear damping a controllable PTO simulator based on eddy currents,
PTO characteristic curves can be simulated by basing the which was used for the testing of a 1/16 scale model
current control on real-time velocity measurement. of the WaveRoller1 wave energy converter. This wave
The calibration of the system was done by connecting energy device consists of a fully submerged bottom
the device to a constant rotating speed motor and mea- hinged plate that takes advantage of the large horizon-
suring the resistent torque produced by the PTO with tal velocity component associated with shallow water
a torquemeter for different values of current intensity waves.
through the coils. This type of geometry has been studied analytically
in 2D by several authors, using potential flow methods
Keywords: Wave energy, non-linear PTO, experimental in a variety theoretical cases. Evans [1] studied the wave
modelling, eddy current brake. diffraction by a submerged vertical plate with small os-
cillations, obtaining expressions for the forces and the
scattered waves away from the plate. Evans and Porter
c Proceedings of the 8th European Wave and Tidal Energy [2] addressed the hydrodynamic characteristics of this
Conference, Uppsala, Sweden, 2009 type of moving plate, obtaining a semi-analytical solu-
tion for the added inertia and radiation damping for dif-
1 http://www.aw-energy.com/

715
1
 Figure 2: Definition of the E-type magnetic core geometry: the
effective area A0 (shaded) is larger than the area of the central
and side leg faces.
(a) (b) static situation.
Figure 1: Two equivalent distributions of weight and buoyancy The rotation equilibrium equation may be established
forces on a rotating plate. as,
Im θ̈ = Md + Mr + Mhs + MPTO + Φ , (1)
ferent water depth to plate height ratios.
where Im is the plate moment of inertia, Md the excita-
Nearshore seabed mounted wave energy devices
tion (diffraction) moment caused by the incident waves,
based on this concept have been studied by the research
Mr is the radiation moment, Mhs is the the hydrostatic
group at Queen’s University Belfast. Several of papers
restoring moment, MPTO is the moment applied by the
document the modeling of this type of geometry, which
PTO and the term Φ is a force that accounts for the real
led to the development of the Oyster wave energy con-
fluid viscous effects in the water. The two last terms are
verter (see [3], [4] and [5]).
generally described as non-linear functions of the plate
Eddy current brakes are a well known application of
velocity.
induced magnetic fields. A simple application of this
The hydrostatic moment Mhs can be described as
type of brake is presented in [6]. An extended review of
electromagnetism can be found, for example, in [7]. Mhs = (−ρw gVb x̂b + mgx̂g ) | sin θ | . (2)
The application of eddy current brakes has been pre- Here ρw is the water density, Vb the plate volume, m the
viously used at the University of Edinburgh, in model plate mass, x̂b the distance from axis of the plate’s centre
scale study of the Sloped IPS Buoy heaving/surging de- of buoyancy and x̂g the distance from axis of the plate’s
vice. The basis for the design of the Sloped IPS Buoy centre of mass.
PTO model are presented in a paper by Taylor and The existence of the PTO simulator underneath the
Mackay [8], and results for the device performance at plate, but connected and rotating with it, will correspond
constant damping are reported in [9]. to a different mass distribution of the system, as com-
In the case of the bottom-hinged rolling plate, which pared to the prototype. The diffraction and radiation mo-
is the object of the current study, a low-speed rotating ments in Eq. (1) are not changed by the mass distribution
eddy current brake was designed, with the electroma- at model scale, provided that the geometric dimensions
gnets being located inside a watertight box. Six pairs of of the plate are scaled correctly. The scaling and model-
electromagnets create a magnetic field perpendicular to ling of the moment applied by the PTO is the central part
the rotating aluminum disc, which causes the disc to be of this study.
braked proportionally to its angular velocity. By means In the experimental model, the PTO simulator is made
of the variation of the current intensity on the electro- of an aluminium disc rotating around a vertical axis. The
magnets as a real time function of the measured angu- aluminium disc and its connections to the plate are ax-
lar velocity, the damping coefficient can be time variable isymmetric. As a consequence, they do not introduce
and follow any given function of the angular velocity. differences in the hydrostatic restoring moment term of
Eq. (1), but they increase the moment of inertia. This
2 Mathematical model needs to be compensated by a corresponding reduction
A mathematical model for the problem can be es- in the plate inertia, as to obtain similitude between the
tablished, considering a time-dependent rotation of the model and prototype according to the Froude criterium,
bottom-hinged plate, θ , induced by the incident water the moment of inertia must be accurately scaled. The
waves. From a static point of view, when the plate is dis- overall increase of the moment of inertia of the model
placed from its equilibrium position, a restoring moment could be only be compensated by a decrease in the added
occurs, as the plate density here is smaller than the water inertia of the radiation moment, which would be com-
density. Figure 1a presents a schematic representation plex and is not followed here.
of the geometry of the device and the applied forces in a Even though the restriction on the moment of inertia,
it is possible to change the variables in Eq. (2), as long as

716
2
Figure 3: Magnetic field lines of a middle cut diagram of two Figure 4: Preliminary tests with an aluminium pendulum.
iron-silicon-core electromagnets facing each other.
where N p is the number of electromagnet pairs, B̄ is the
the hydrostatic restoring moment Mhs remains constant. average absolute value of the magnetic field, r̄ and e are
This means that the plate gravity centre could even be lo- the rotating disc average radius and thickness, A0 = AC0 +
cated below bottom level, keeping similitude conditions, 2A0L , AC0 = kC AC , A0L = kL AL , AC and AL are the area of
as exemplified in Fig. 1b. As a consequence, it is pos- the central and side leg E-type cores faces (see Fig. 2), kC
sible that both the weight and buoyancy cause restoring and kL are correction factors to take into account the fact
moments, as long as their sum is equal to the original that the effective area is higher than the area of the core
restoring moment. faces due to the existence of round shaped flux density
Scaling the mass of the system, the plate volume and streamlines.
the mass and buoyancy centers is therefore not neces- Considering that the magnetic circuit is represented
sary, as long as Mhs is constant, but a very thin and light only by the air gap reluctances, we define the equivalent
plate may be required so to compensate the increased magnetic reluctance
contribution for the moment of inertia by the PTO simu-
lator. RL δ 2A0L + AC0
Req = RC + = , (8)
2 µ0 2A0L AC0
3 Design concept where
3.1 Electromagnets δ δ
RC = and RL = (9)
µ0 AC0 µ0 A0L
Moving a conductive material perpendicularly to a
magnetic field produces a force between them, which is are the central and side leg magnetic reluctance, and δ
proportional to the relative velocity. This electromagne- is the length of the air gap (Fig. 2) and µ0 is the mag-
tic braking force density, F, is given by the vector pro- netic permeability. As shown in Eq. (9), δ shall be as
duct of the current density J, and the flux density B, close as possible to e to minimize the magnetic reluc-
tance, but this is limited by the geometric accuracy of
F = J × B. (3) the PTO manufacture.
A current intensity I in the Ns turns winding produces
For an aluminium conductor it is considered that the cur- a flux φ = B̄A0 in the magnetic circuit central leg
rent density is proportional to the electric field E = v×B,
Ns I
as expressed by φ= , (10)
Req
J = σ E = σ (v × B), (4)
were Ns is the total number of wire turns.
where σ is the electric conductivity of the material mo- Current intensity I can be easily related with damping
ving at velocity v. In this way, factor C = MPTO /θ̇ by combining Eq.s (7), (8) and (10),
giving
F = σ (v × B) × B. (5) (2AC AL INs r̄µo )2 eσ
C = Np , (11)
(2A0L + AC0 )3 δ 2
Assuming B ortogonal to v, we obtain
where N p is the number of pairs of electromagnets and
F = σ vB , 2
(6) each pair corresponds to Ns turns. In the present design
is δ = 3.5 mm, e = 2 mm, r̄ = 180 mm, AC0 ' 2A0L =
where B = kBk. 0.0016 m2 , µ0 = 1.257 × 10−6 N·A−2 , Ns = 2000, N p =
For a rotating disc, at constant angular velocity θ̇ , the 6.
absolute value of the braking torque is Figure 3 shows the flux density streamlines in the
Z magnetic circuit, obtained as given by a 2D Finite Ele-
MPTO = N p σ r2 θ̇ B2 dV = σ r̄2 θ̇ B̄2 eA0 , (7) ment software [10]. As can be seen in Fig. 3 the majority

717
3
 Figure 5: Schematic representation of the PTO simulation system.

of the magnetic field lines are enclosed in the iron-silicon 4 Control Device
core and in the air gap region. Results plotted in Fig. 3
As described in Section 3, PTO model was designed
also show that the magnetic flux density streamlines are
to provide a specified
mostly straight lines in the air gap.
The reference design condition for the PTO simula- MPTO = C(θ̇ )θ̇ (12)
tion system was the specified maximum damping factor
Cmax for which a system with six pairs of electromagnets relationship, which might be easily modified within the
was chosen. Each pair of magnets is connected in serial, given working range. Here C is required to be depen-
such that the electric connection assures the magnetic dent on the plate velocity and this is achieved by properly
flux between the cores, and that the generated magnetic feeding current into coils (see Eq. (11)).
fields do not cancel out. A specific dedicated electronic device was developed
A first concept demonstration test with an aluminium to provide a controllable and reliable level of DC current.
pendulum passing a pair of electromagnets was per- This has three main components: (i) sensor part inclu-
formed, as represented in Fig. 4. In these tests, the pen- ding position, temperature and current sensor; (ii) power
dulum was started from a fixed point and the current in- control device comprising phase control integrated cir-
tensity in the coils was changed. cuit and pair of thyristor/diode modules; (iii) periph-
eral interface controller comprehending dsPIC 30F4013
3.2 Mechanical Components micro-controller and computer for data storage.
The plate is connected through two aluminium hinges The position sensor is attached to the PTO simulator
and a bevel “tee” gearbox to a 2 mm thickness rotating disc shaft to measure the angular position θ of the rota-
aluminium disc (see Fig. 5), located beneath the tank ting aluminium disc. The angular position information
bottom level (inside the tank pit) in a watertight box. In is acquired by the dsPIC.
the present case, the gear ratio of the bevel “tee” gearbox The dsPIC acquires signals, performs calculations,
is 1:1. If this was not the case it should be noted that the generates the control signal to feed current into the coils
moment of inertia of the disc should be multiplied by the and sends the logged data to the computer for storage
gear ratio. via a RS-232 connection. The main task of the dsPIC
The aluminium disc rotates about a vertical axis, its is to compute angular velocity θ̇ in real-time and adjust
angular rotation being the same predicted for the plate, the control output voltage accordingly. The calibration
i.e., maximum of 45 degrees to each side from the equi- described in section 5 enables a table to be build with
librium position. different values of current I as a function of the angular
The disc movement is damped by six pairs of elec- velocity θ̇ . Those values are then stored in the dsPIC me-
tromagnets. The cover watertight box is connected to mory. Based on 0 to 5 V DC input signal, the power con-
the atmosphere through two 100 mm diameter flexible trol device provides the required 0 to 10 A DC current to
tubes, used simultaneously for cabling and ventilation. the braking coils, which is measured by a LTS15NP Hall
effect current transducer to allow feedback control.

718
4
 Power Control Device

Cu
ge rr
ta en
ol ts
lv t nt up
ro en e pl
nt r r em y
Co Cu ur
dsPIC 30F4013 s
ea
m

Logged
data
TCA Temperature
785 and rotation
measurement

Computer Power Take-off Simulator

Figure 6: Relation between the peripheral interface controller, the power control device and the power take-off system.

Figure 7: Phase control TCA785 IC relevant pulses. Figure 8: Electric connection between the thyristor/diode
modules.
The power control device was built specifically for
this work and consists of a half-controlled single-phase voltage, V5 , the 100 Hz ramp voltage, V10 , which is com-
bridge circuit with two trigger pulse transformers for bined with the control voltage, V11 , as described above,
low-power thyristors, a current sensor and a 230 V AC to originating the thyristors trigger pulses V14 and V15 . A
15 V DC transformer and rectifier. The AC current feeds LM358 integrated circuit was also included to transform
the thyristor bridges and the transformer and provides dsPIC 0 to 5 V control voltage to TCA785 0 to 15 V
the synchronization signal to the half-controlled single- control voltage.
phase bridge circuit. The 230 V AC to 15 V DC trans- An ideal diode turns off, and becomes an open cir-
former supplies both the dsPIC micro-controller and the cuit, when its current decreases through zero. It turns
electronic circuits inside the power control device box. on, and becomes a short circuit, when its voltage in-
The latter controls a single-phase bridge circuit with two creases through zero. An ideal thyristor is the same as an
trigger pulse transformers for the low-power thyristors, ideal diode except that its turn on is inhibited unless the
which produces a rectified DC output current ranging thyristor firing pulse is on. Thyristors are mainly used
from 0 to 10 A. when high currents and voltages are involved, and are of-
A TCA785 phase control integrated circuit was used ten used to control alternating currents, when the change
to
Pulse Diagram
control the pair of thyristor/diode modules. This cir- of polarity of the current causes the device to automati-
cuit allows the recognition of zero passage, and provides cally switch off (zero cross operation). The device can
a large ramp 250 mA output current. The synchroniza- be said to operate synchronously as, once the device is
tion signal is obtained via a high-ohmic resistance from open, it conducts current in phase with the voltage ap-
the electrical grid. A zero voltage detector evaluates the plied over its cathode to anode junction with no further
zero
Semiconductor passages and transfers3 them to the synchronization
Group gate modulation being required. Present configuration
register, which controls a ramp generator. If the ramp adopted two pairs of SKKH27/16E thyristor modules in
voltage exceeds the control voltage, a signal is processed combination with a power diode, as shown in Fig. 8. A
to the TCA785 logic. Depending on the magnitude of resistor-capacitor (RC) snubber circuit was also included
the control voltage, the triggering angle can be shifted between thyristors anode and cathode terminals in order
within a phase angle between 0 and 180◦ . For every half to limit the dV /dt to prevent triggering by a high rate of
wave, a positive pulse of approximately 30 µs duration rise of off-state voltage.
appears at the outputs. The pulse duration can be pro-
longed up to 180◦ via a capacitor. Relevant pulse shapes
are shown in Fig. 7 including the electrical grid 50 Hz

719
5
 Current intensity, A
20

10
9
15
8
7

Torque, M, [Nm]
10 6

5
4

3
2
0 1
0.0 0.2 0.4 0.6 0.8 1.0
Angular velocity, θ̇, [rad/s]

Figure 9: Calibration with controllable constant rotation speed Figure 11: Schematic representation of the design space for
motor and torquemeter with identification of its components. the configuration presented in this paper. The current intensity
in the coils is limited to 10 A.
20.0

Calibration
1.0 1.0
15.0
Damping level [kg.m/s]

Dimensionless angular position,θ/θmax

Dimensionless torque, MPTO /MPTO,max

0.5 0.5

10.0

0.0 0.0

5.0
−0.5 −0.5

0.0 −1.0 −1.0

θ/θmax MPTO /MPTO,max
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
0 2 4 6 8 10
Current Intensity [A] Time, t [s]

Figure 10: Brake performance for magnet spacing δ = Figure 12: Example time series of plate rotation and corre-
3.5 mm. sponding torque for a non-linear power take-off law.

5 Model Calibration target values in the simulation, which is around 20 N·m.

The values obtained for damping coefficient, C, for
The experimental set-up for the calibration is pre- different values of current intensity, I, are represented
sented in Figure 9. in Fig. 10 obtained for a PTO simulator with six pairs
The PTO simulator was connected to a motor con- of electromagnets and δ = 3.5 mm. Although the tar-
trolled to produce fixed speed. As the desired range of get value for damping was achieved, the fact that the ex-
angular velocities to use in the calibration is under 1.0 pected parabolic shape of this curve is not verified for the
rad/s (9.55 rpm) and typical motors have a higher nom- higher current intensity suggests that a saturation phe-
inal rotation speed, a 1:30 reduction gear was used in nomenon may be happening, which will be further in-
order to use the motor within its normal working range. vestigated.
A torquemeter with a nominal torque of 50 N·m was Figure 11 represents the application of the calibration
installed between the motor and the PTO simulator. A function on a torque-angular velocity space, consider-
set of tests was performed for constant angular velocity, ing the current limited to 10 A, as given by inverting the
varying the current intensity through the coils and mea- curve fitted to the data plotted in Fig. 10.
suring the correspondent torque value for each case. Considering a specified relationship between the
During the calibration, the PTO simulator was con- torque MPTO and the angular velocity θ̇ , for a given wave
nected to the “tee” gearbox to which the plate is attached energy converter model, the values of the current inten-
during the tank tests. At the same time as having all sity needed for the corresponding damping and velocity
the friction sources connected during calibration, this are obtained from results plotted in Fig. 11.
allowed to have an estimation of the friction torque in A lookup table representing the final curve I(θ̇ ) is in-
the system when no current is fed into the coils. This cluded in the dsPIC real-time control program in order
torque was estimated to be 0.27 N·m at 10 rpm, which to perform the specified non-linear PTO damping char-
can be considered an acceptable value considering the acteristic curve (Eq. (12)).

720
6
 Figure 12 presents a time series of the wing rotation, technician team Norberto Marques, Carlos Antunes and
showing the correspondent torque applied to the system Pedro Alves for manufacturing the PTO simulator. Sup-
by the PTO simulator. This was obtained using a non- port from Prof. João Santana of CIE3 is also appreciated.
linear PTO curve within the design space of the device.

Conclusions References
A real time controllable eddy current brake PTO si- [1] D. V. Evans. Diffraction of water waves by a submerged
mulator for small-scale model testing of rotating wave vertical plate. Journal of Fluid Mechanics, 40:215–228,
energy converters was designed and tested at Instituto 1997.
Superior Técnico. Eddy current brakes are based on the
[2] D. V. Evans and R. Porter. Hydrodynamic characteristics
principle that a braking force proportional to the velo- of a thin rolling plate in finite depth of water. Applied
city is produced when a conductive material moves per- Ocean Research, 18:215–228, 1997.
pendicularly to a magnetic field. In the present device, [3] M. Folley, T. W. T. Whittaker, and J. van’t Hoff. The
the magnetic field is generated by electromagnets, being design of small seabed-mounted bottom hinged wave
the braking force function of the current intensity in the energy converters. In Proceedings of the 7th Euro-
coils. The rotor is an aluminium disc which is connected pean Wave and Tidal Energy Conference, Porto, Portugal,
to the rotating part of the wave energy model. 2007.
A special power source was built to provide the re-
[4] M. Folley, T. J. T. Whittaker, and A. Henry. The effect of
quired controllable current intensity level. Different non-
water depth on the performance of small surging wave
linear damping PTO characteristic curves can be simu- energy converter. Ocean Engineering, 34:1265–1274,
lated by basing the current control on real-time velocity 2007.
measurement.
The calibration of the system was done by connecting [5] Trevor Whittaker, David Collier, Matt Folley, Max Os-
the device to a constant rotating speed motor and mea- terried, Alan Henry, and Michael Crowley. The devel-
suring the resistent torque produced by the PTO with opment of oyster - a shallow water surging wave energy
a torquemeter for different values of current intensity converter. In Proceedings of the 7th European Wave and
through the coils. Tidal Energy Conference, Porto, Portugal, 2007.
This device was successfully used in the 1/16 model
[6] Manuel González. Experiments with eddy currents:
scale wave tank testing of the WaveRoller wave energy the eddy current brake. European Journal of Physics,
converter. In this application the PTO was located be- 25:463–468, 2004.
neath the tank bottom level (inside the tank pit) in a wa-
tertight box, which allowed realistic non-intrusive PTO [7] John R. Brauer. Magnetic Actuators and Sensors. John
simulation conditions. The described PTO simulator Wiley & Sons - New York, 2006.
constitutes a useful innovative device for experimental
modelling of wave energy converters that will be fur- [8] J. R. M. Taylor and I. Mackay. The design of an eddy
ther used in disclosed wave energy converter small scale current dynamometer for a free-floating sloped IPS buoy.
model testing. In Proc. International conference on Marine Renewable
Energies, Newcastle, 2001.

Acknowledgements [9] Grégory S. Payne, Jamie R.M. Taylor, Tom Bruce, and
Penny Parkin. Assessment of boundary-element method
This work has partially been supported by Wa- for modelling a free-floating sloped wave energy device.
veRoller contract PL375/08 and Portuguese Founda- part 2: Experimental validation. Ocean Engineering,
tion for Science and Technology (contract PTDC/EME- 35(3-4):342–357, 2008.
MFE/66999/2006, PhD grant SFRH/BD/29275/2006
and Ciência 2007 Initiative). [10] David Meeker. FEMM 4.2 Magnetics, Electrostatics,
The authors are thankful to Anı́bal Mota for his par- Heat Flow, and Current Flow - Reference Manual. Fi-
ticipation in the earlier stages of the work and the IST nite Element Method Magnetics, 2009.

721
7