Linear generators for direct-drive wave energy conversion

H. Polinder B.C. Mecrow, A.G. J ack, P. Dickinson M.A. Mueller

Delft University of Technology University of Newcastle University of Durham
Mekelweg 4,2628 CD Delft Men Court, Newcastle upon Tyne NE1 7RU
The Netherlands United Kingdom United Kingdom
South Road, DurhamDHl3LE
Abstrnct- The Archimedes Wave Swi ng (AWS) is a system that
Converts ocean waveenergy into electric energy. The goal of the
research described in this paper is to identify the most suitable
generator type for this application. Of the more conventional
of the AWS, energy yield calculations and the generator design
are described in [ 5] . Some generator lest results are given in [6].
B. Goal of the paper
generator type% the three-phase pe-ent-magnet synchronous
ne linear PM cuendy applied in the AWS has a
generator with iron in both stator and translator is most suitable,
because it is cheaper and more eflicient than the induction
generator, the switched reluctance generator and the permanent- The force and translator
maenet eenerator with air-gar, winding. The vauer al so vrowses
number of disadvantages:
big problems to the bearings, which should be made
.. .. .. . . . . .
a newtrnmierse-flux prmanent-magnet genrratur topolog). that maintenance-free. Even \r,hen the generator is made
could heauitahlu for this npdimtion. Thb new duuhledouhlr-sided double-sided to balance the attractive forces (as is
TFPM generator has flux-cincentrators, magnets, and conductors
on the stator, while the translator only consists of iron.
Inder'Terms-wave energy, linear directdrive generators,
permanent-magnet generators, transverse flux generators
1. IN'IRODUCI~ON
A. The Archimedes Wave Swing
There are different ways of converting Oceau wave energy
into electric energy [l-31. Several systems use linear generator
systems [4]. One of them is the Archimedes Wave Swing
(AWS) b61.
The AWS basically is an air-filled chamber, the lid of which,
called the floater, can move in vertical direction. The principle
of operation is illustrated in Fig. 1. When a wave is above the
AWS, its volume is reduced by the weight of the water above.
When a wave trough is above the AWS, its volume increases
because of the air pressure inside. From this linear motion,
energy can be extracted and converted into electrical energy. In
principle, this can be done by converting the linear motion into
rotating motion and by using a rotating generator. However, it
appears to beextremely difficult to build a robust maintenance-
free gear for such a conversion. Therefore, a linear generator is
used, which converls the energy fromthe linear motion directly
into electrical energy. A convener is used to connect the
generator to the grid.
To test the complete system a pilot plant bas been built on a
pontoon as depicted in Fig. 2. The system has not yet been
submerged. For this pilot plant, a linear pennanent-magnet
(PM) generator has been designed. The principle of operation
?he the financial suppon of Netherlands Organization for Scientific
Research ("0) is greatly acknowledged.
currently done), deviations from the ideal air gap result in
severe bearing loads.
The generator is very large and therefore expensive. To
obtain a commercially viable system the generator should
be made as cheap as possible.
Losses in the generator have to bekept at an acceptable
level to reduce cooling problems and to increase the
annual energy yield.
2)
3)
. .
Rg. 1. Sketch of the AWS illustrating the operation principle
Rg 2 Photo of theAWS pnlot plmt bsfwt \uhmcrgng
0-7803-7817-2/03/$17.C0 82C03 IEEE 798
Therefore, the goal of the research described in this paper is
to identify the most suitable generator type for application in
direct-drive wave energy conversion. This is done in two steps.
First, section I1 describes the comparison of some more
conventional generator types. This comparison is limited to
generator types without moving connections, because moving
connection are considered to be not maintenance-free.
Generatortypes withoutironin eitherthe statororthe translator
are considered, because they eliminate the bearing problems
due to attractive forces between stator and translator. The
criterion for comparison takes cost and losses into account.
Next, section I11 discusses whether a transverse-flux
permanentmagnet (TF'PM) generator could be cheaper and
more efficient than the currently applied PM synchronous
generator. This is done because in literature, it is claimed that
TFPM machines have very high force densities and low losses
[4,7,81. For the current pilot plant, TF'PM generators have not
been considered because they were estimated to be too difficult
to analyse and build.
Although the paper is about the generator for the AWS, it is
not only of importance for the AWS, but also for other
applications with low-speed high-force linear motion.
n. COMPARISON OF CONVENTIONAL GENERATOR TYPES
This section describes the comparison of five more or less
conventional generator types for application in the AWS. The
section starts with the introduction of a criterion for generator
oplimisation and comparison. Next, the characteristics of the
AWS necessary to evaluate the criterion are discussed.
Subsequently, the following generator types are modelled,
roughly optimized and compared
-
-
- the switched reluctance generator,
-
-
The section concludes with a discussion of the results
A. Crirerion for optimization and comparison
includes generator cost and losses:
the induction generator without iron in the secondary,
the induction generator with iron in the secondary,
the PM synchronous generator with air-gap winding, and
the PM synchronous generator with stator iron.
The criterion to compare the different generator types
(1) C =Cgen +P EdCkwh
where
CZen is the active generator material cost,
P is a period of 5 years,
Ed is the annual energy dissipated in the generator, and
C,,is the price of energy.
When this criterion is used to optimize machine designs, it
means that an additional investment in the generator efficiency
must be earned back in 5 years.
- iron costs 3 euroikg,
- copper costs 10 eurokg,
As basis for our calculations, wefurther assume:
- magnets cost 30 euroikg.
- energy costs 0.10 EurokWh.
is necessary, which is the subject of the next section.
B. Characterisrics of the A WS
The principle of operation of the AWS is discussed in 151. It
is shown that under ideal circumstances, the floater motion is
sinusoidal. The rated pilot-plant floater velocity amplitude is
2.2 d s , while the rated floater motion amplitude is 3.5 m. It is
further shown that the generator has 10provide a damping force
(a force proportional lo the floater velocity), depending on the
wave height and the wave period.
From the combination of force and speed. the energy
extracted from a wave can be calculated. The annual energy
yield can be calculated if this is combined with the wave
distribution 151, the annual number of waves as a function of
wave height and wave period. For each generator type, this is
also used to calculate the annual energy dissipation in the
generator which is used in the criterion (I ).
To extract energy from the floater motion, the maximum
force the generator has to provide is 1 MN. This sounds as very
large, hut it is in the order of the gravity force on one and a half
metres of water above the floater with its 9 m diameter. This
huge force mainly determines the size and cost of the generator.
C. Oprimization and assumptions
For each generator type, a set of basic dimensions for a few
pole pitches is laken. These basic dimensions are chosen so that
when they are combined with the material properties and the
allowed current densities and flux densities, the performance of
a few pole-pitches of the machine can be calculated. The
number of poles is made large enough to generate the maximum
required force, 1 MN. Then the optimization program varies the
basic dimensions to optimize the machines to the criterion ( I ) .
In this way, roughly optimized generators are compared.
The following assumptions are used in the calculation.
- Thetranslatorismade 12m2 largerthanthestatorto keep
overlap when the translator moves (as in the current
generator).
The air-gap length is g =5 mm because it could not be
made smaller for mechanical reasons.
The stator slot fill factor is k,fi, =0.585, which means that
58.5% of the stator slot is filled with copper.
For the induction and pennanent-magnet generators, the
number of phases is 3 because most available converters
have 3 phases.
- As in the current generator, the stack length is I s =I m.
- The current density in the stator copper is limited to
5 A/m2 to prevent excessive losses.
~ The iron losses in the stator are calculated as
To evaluate this criterion. a better understanding of the AWS
-
-
-
199
where PFe,,pec '=4.9 Wkg at 50 Hz and 1.5 T.
D. Induction generafor modelling
Fig. 3 depicts a cross section of two pole pitches of a linear
induction machine with some dimensions. In this figure, the
number of slots per pole per phase is 2. The translator is the
secondary placed between two stator sides. Table I gives some
important dimensions resulting from the optimization.
Forthe induction generator without iron in the secondary, the
translatorisa24mmthickcopperplate betweenadouble-sided
stator, so that the translator fill factor is kfi,,, =1. For the
induction generator with iron in the secondary, the 100 mm
thick translator consists of iron teeth and copper bars of equal
width, so that the translator fill factor is kFfl" =0.5.
The amplitude of the flux density wave in the air gap of a 3-
phase machine if there is no current in the secondary is given by
(3)
where
kw is the winding factor,
Ns is the number of stator turns per phase in two pole pitches
per machine side,
i _ is the stator phase magnetizing current amplitude, and
g8 is the effective air-gap length, including the secondary
thickness for the induction machine without iron in the
secondary, but-excluding the secondary thickness for the
induction machine with iron in the secondary.
It is assumed that this flux density remains constant when
there is current in the secondary. To keep this flux density
constant, it is necessary that the current in the secondary is
compensated by an additional current in the two stator sides.
We assume a phase shift of 90" between the stator magnetizing
current and the stator current compensating the secondary
current. These assumptions are too optimistic because they
neglecttheadditional statorcumentnecessary tocompensate for
leakage fields and they neglect end effects. However, they are
used here for a fnst approximation.
With these assumptions, the current density J. in the
- .
secondary can be calculated as
b. b.
, ,
Fig. 3. Crass section o f a linear induction generator.
(4)
. .
Pc.
where
pcu is the resistivity of the secondary, and
is the velocity difference between the secondaq'and the
air-gap flux density wave.
With the applied assumptions, the maxima of the current
density and the flux density in the secondary coincide.
Therefore, the force density or shear stress can be calculated as
( 5)
1 - -
Fd = ?B,J,6'kfiil
To limit saturation in the stator, the amplitude of the air-gap
flux density is limited to 0.9 T. At low forces, flux weakening
may be applied to reduce losses.
E. Switched reluctance generator modelling
A part of the considered linear switched reluctance generator
is depicted in Fig. 4. The msl ator is placed between two stator
sides.~smachinehas4statorteethper6translatorteeth.This
4-6 machine doesn't function properly as a motor, hut that is
not necessary. We prefer this 4-6 machine. to the (more
common) 6-4 machine because the 4-6 machineleaves more
space for copper in the stator. Table I gives some imporlant
dimensions resulting from the optimization.
It is assumed that the current steps from zero to the desired
value in the aligned position, and steps hack to zem in the
unaligned position. The flux @ in the aligned position is limited
by saturation; the flux density in satnration is assumed to be
1.8 T. p e flux in the unaligned position is assumed to he 20%
of the flux in the aligned position. The average force is then
calculated from the surface in the flux-magnetomotive force
plane [9]:
where
F,,, is the magnetomotive force, and
Wm is the magnetic energy
Ax is the displacement during one cycle
Fig. 4. Cmss section of a switched reluctance generator
800
F. Permanent-magnet generafor modelling
Fig. 5 depicts a permanent-magnet generator with a three-
phase air-gap winding in the stator and a translator with
surface-mounted magnets on both sides of the stator. The
permanentmagnet synchronous generator as currently applied
in the AWS pilot plant has surface-mounted permanent magnets
and a three-phase stator as depicted in Fig. 6. Table I gives
some important dimensions resulting from the optimization.
The fundamental of the air-gap flux density due to the
magnets is calculated as
t t t
(7)
+ + +
where
B, is the remanent flux density of the magnets,
I ,
p, is the recoil permeability of the magnets,
h, is the slot height,
b, is the magnet width,
zP
is the pole pitch, and
grf f is the effective air-gap length, which is 2g +h, including
theslotheightforgeneratorwiththeair-gapwinding, and 2g
excluding the slot height forthe generator with stator iron.
The force is calculated using this flux density in the
is the magnet length in the direction of magnetization,
expression for the hrenz force.
G. Results
Table I summarizes the results. We conclude that the PM
synchronous generator with stator iron is the most suitable
generator. Alternatives are al l more expensive and less efficient.
Besides, the magnets add to the air gap, so that deviations in the
ai r gap result in smaller unbalanced forces than for induction
machines with iron in the secondary or switched reluctance
machines. The induction machine without iron in the secondary
and the PM machine with air-gap winding give less bearing
problems because the attractive force between stator and
translator is eliminated. However, both are so much more
expensive and less efficient that they are not considered to be
serious alternatives for the PM synchronous generator.
Probably, the large air gap and the low speed are the main
reasons that induction generators give had results.
Fig. 5. Cross section of a permanent-magnet synchronous generator with
air-gap winding.
I t t t I I t t +I I 4 t t I I t t t I
It t tI I + t t I I t 4 +I I t t tlp,
Fig. 6. Cross section of a permanent-magnet synchronous generator.
80 I
In. NEW DOUBLE DOUBLE-SIDED TFPM MACHINE
This section SW with the introduction of a new TFFM
generator topology that might besuitable for application in the
AWS. Next, t hi s topology is analysed: the generated force is
calculated and additional eddy-current losses due to three-
dimensional fields are estimated. The section ends with a
discussion of the results and further work.
A. New TFPM generator topology
In literature, it is often claimed that TFPM machines can
have much higher force densities than other machine types [4,
7, 81. Therefore, it is investigated whether a TFPM generator
could be cheaper and more efficient than the current generator.
TFPM machines with surface-mounted magnets tend tohave
a very large leakage from the poles that are not covered by U-
cores into the U-cores, resulting in a disappointing performance
[IO]. Topologies with flux concenisation generally perform
better [I 1-13].
In the currently applied PM generator, the coils are in the
stator and the permanent magnets are on the translator. The
translator is longer than the stator, which means that a part of
the (rather expensive) pennanent magnets is not used.
Therefore, it would be advantageous to have generator gpology
with both the coils and the permanent magnets in the &tor and
a translator consisting of only iron. Kang [14] describes a
TFPM topology where this is the case. The form of the coils in
thedouble-sidedTFPMmachinedescribedin 112, 13) canalso
be changed in such a way that both the magnets and the coils
are on the stator, while the translator consists of only iron.
These ideas form the basis for the new TFPM topology
proposed here. It is called the double double-sided TFPM
machine because it is more or less double the double-sided
TFPM machine described in [12, 131. The easiest way to
introduce t hi s topology is by going back to one of the original
ideas of Weh [SI, depicted in Fig. 7. In this topology, the stator
consists of coils and U-cores on both sides of the rotor. The
rotor consists of two rows of permanent magnets and flux
concentraton with space for construction material in between.
To come to the new proposed topology depicted in Fig. 8, the
material between me two rows of magnets and flux
concentrators is replaced by the conductors. Both the
conductors and the magnets are kept stationary. The U-cores get
a simpler shape because they do not leave space for coils, and
these cores or yokes form the translator.
As for the more conventional generator types, the air gap of
the TFPM generator is 5 mm. Therefore, the pole pitch is made
50 mm A smaller. pole pitch results in a much lower flux
linkage because leakage increases. A larger pole pitch results in
a lower frequency, which results in a lower induced voltage.
\v
Fig. 7. W M machine with flux concentration and moving magncls
Pig. 8. W M machine with flux commaation and moving mp e s
B. FEM analysis ofpower factor and force
One pole pitch of the double double-sided TFPM machine
has been analysed in a three-dimensional FEMpackage
including magnetic saturation. Fig. 9 and Fig. 10 depict some
no-load results.
Fig. 11 depicts the magnetomotive force (mmf) of the stator
current as a function of the translator position. The current is
maximum in the unaligned position, which means that the
current is in phase with the no-load voltage. Fig. 11 also depicts
the resulting flux calculated with FEMand the derivative of the
flux to the position x , which has the same form as the voltage.
This voltage is far from sinusoidal due to heavy saturation in
the translator yokes. There is space to increase the width of
these yokes to reduce saturation. From themagnetomotive
force, and the derivative of the flux to the position, the power
factor can be calculated as 0.31. This is rather low.
Fig. 12 gives the flux-magnetomotive force characteristic,
from which the average force can be calculated with (6) [ l l -
131. The surface area of this graph is 115 I. According to this
calculation, 870 poles are necessary to make the required force
of 1 MN.
Rg. 9. Three-dimensional FEM
resu11 in the aligned position.
Rg. IO. Rough twwdimensional E M
result in the unaligned position.
sn,
-
I
20
-50'
0
*
~
& 0.5
z
20 40 60 80 100
Fig. 11. Magnetomotive force, mulling flux and the derivative of the fl ux to
the position as a function of the position.
-0.5
0
position (m)
I
0 IO 20 30
magnetomotiveforce (kA)
Fig. 12. Flux-magnetomotive fome characteristics in the two opposite
aligned pi t i ons
C. Eddycurrent losses in rhe cores
To make the translator as simple as possible, the translator
yokes shouldbe made of laminations, where the laminations are
perpendicular to the direction of the current. A disadvantage of
this way of building the machine is that there are additional iron
losses due to the three-dimensional fields.
Fig. 9 and Fig. IO depict the results of FEM calculations in
the aligned and the unaligned position. The fields entering the
translator yokes horizontally enter the laminations
perpendicular to the lamination direction and cause additional
eddy-current losses. According to these FEM calculations, the
maximum of the horizontal component of the flux density in the
iron is between 0.7 T and I T depending on the permeability of
the iron in the horizontal direction. Within 15 mm. this reduces
to half this value.
To obtain a fust, very rough indication of the eddy-current
losses in the translator yokes, we use the well-known
expression for eddy-current losses per unit of volume in
laminations:
where
w is theangular frequency,
B
bFe is the width of the lamination, and
pFc is the resistivity of theused material.
1)
is the flux density, uniform over the lamination width (the
effect of eddy-currents on the flux density is neglected),
To use t hi s equation, weuse a number of assumptions.
The flux density entering the yoke horizontally has an
amplitude of 0.5 T over a height of 30 mm, so that we can
use b, =30 mm.
This flux causes eddy currents in one direction over the
fust 15 mmand in the other direction over the rest of the
yoke.
The eddy-current losses in the rest of the yokeare
negligible compared to the eddy-current losses in the fmt
15 nnn
The resistivity of iron is pF, =0.5 pRm.
2)
3)
4)
The resulting average additional eddy-current loss for the
complete generator is 21 kW. This is enough to worry abut,
because it results in an additional reduction of the annual
energy yield with 184 MWh.
Because this result is worrying and the calculation method is
very rough, we also did some FEM calculations to determine
the additional eddy-current losses in the laminations. The
resulting average additional eddy-current loss for the complete
generator is 13 kW, which is still too much.
There are several ways of reducing theseeddy-current losses
1) It is possible to use powdered iron (soft magnetic
composites, SMC) instead of laminated iron. However, this
reduces eddy-current losses, but increases hysteresis losses and
decreases the magnetic permeability and the saturation flux
803
density.
2) It is possible to change the lamination direction in the
lower part of a translator yoke. However, this complicates the
construction considerably.
3) It is possible to make narrow slots in the laminations, as
proposed by Pistoyes [IS] to increase the path length for the
eddy currents. By making a slot with a depth of a few
centimetres every centimetre, the average additional eddy-
current loss should decrease with a factor 5 to 10 to something
in the order of 2.5 kW, which probably is acceptable.
D. Results and discussion
The results of the analysis F e summarized in table II. The
double double-sided TFPM machine appears to beonly a little
cheaper than the current PMSM, but more efficient. A number
of items need further investigation.
I ) The preliminary calculations on eddy-current losses
presented above, have to be refined and verified.
2) The power factor is calculated as 0.3 1. This power factor
means that the convener must be overrated by a factor 3. This
is not yet a reason to stop research in this direction, because $e
low-speed direct-drive generator is much more expensive than
the convener, so that it might be sensible to use a more
expensive convener if thegenerator cost can bereduced.
3) Generally, mannfacturing TFPM machines is rather
difficult. For the proposed topology, this might be worse,
because the proposed stator is a rather thin and very long
construction, consisting of rows of magnets and flux
concentrators with coils in between these rows.
4) In [12, 131 it is shown that the use of claw-pole struchms
improves the performance of TFPM machines. Probably,
increasing the width of the magnets and flux concentrators
close to the air gap will improve the performance of the
proposed topology.
These results make it worthwhile to investigate further. A
thoroughoptimizationof thetopology mi&timprovetheresnlts
considerably. However, it is expected that the impmvement
compared to the currently applied PM synchronous generator is
not so dramatic as hoped.
TABLE Il
CHnnnCIErUSlllJ OFME DO&LE~DOUBLE-S~DED W M GENERATOR
[Generator twe
DuLblcdoublc.~idcdIl~'M 221 125 194 I I9
IV. CONCLUSIONS
Of the more conventional generator types, the three-phase
permanent-magnet synchronous generator with iron in both
stator and translator is most suitable for application in the
Archimedes Wave Swing, because it is cheaper and more
efficient than induction generators, switched reluctance
generators and permanent-magnet synchronous generators with
air-gap windings.
The new double double-sided transverse-flux permanent-
magnet machine with flux concentration has both magnets and
conductors are on the stator, while the wanslator only consists
of iron. Although this TFPM generator is probably difficult to
build, it appears to be a little cheaper and much more efficient
than the currently applied linear PM synchronous generator.
Additional eddy-current losses in.laminations due to three-
dimensional fluxes can probably be limited to acceptable values
by using Pistoyes slots. Therefore, it is worthwhile to
investigate this topology further.
REFERENCES
[I ]
N.I. Meyer. K. Nielsen. 'The Danish wave energy programme second
year status'. in Proceedings of the fourth wave energy conference,
Mborg, 2000,pp. 10-18.
T.W. Thorpe, 'The Waveenergyprogammein theUKandthe European
wave energy network'. in Proceedings of the fourth waw energy
conference, 2000, pp. 19-27.
D. Scam. Wove energy: rechnology tromfer dr R&D recommendations,
Report of Ove Arup &Partners htemationd Ltd, London. 2000
M.A. Mueller. 'Electrical generators for direct dri ve wave energy
conveners'. in IEE Proceedings - genernrion, rrnnrmission and
distribution, 2W2. vol. 149. pp. 446456.
H. Polinder, F. Gardner. B. Vdesema. 'Linear PM generamfar wave
energy conversion in the AWS, in Proceedings of the Eighth
Int emt i onol Conference on Electrical Maehincs, Helsinki, August
2000, pp. 309-313.
H. Polinder, M.E.C. Damen. F. Gardner, 'Madelling and test results of
the AWS linear PM generator system'. in Proceedings of rhe
Intemrionnl Confcrcncc on Electric Machines, Bngs. August 2OM.
M. R. Dubois, H. Polinder, and 1. A Femira. 'Comparison of generator
topologies for direct-drive wind turbines'. ! n Pmc. of the Nordic
Countria Power and Industrial Electmnics Conference (NORPIE).
[2]
[3]
141
(51
161
171
Adbag. June 2000, pp. 22-26.
H. Wch, H. Hoffman, 1. hdrath. 'New permanent m-et excited
synchmnous machine with high efficiency at low speeds', in
[SI
Proceedings of the Seventh lnremationnl Conference on Elecrrical
Machi no, Pisa, September 1988, pp. 35-40,
191 T.J.E. Miller, Switched reluctance motom nnd their contml, Oxford:
Magna Physics Publishing and Clarendm Pnss, 1993.
[I O] M.R. Harris, G.H. Pajmman, S.M. Abu Sharkh, 'Performanceand
design aptimisation of electric motors with hempolarsulface magnets
and homopolar windings', in IEE Proceedings - Electric Power
Applications, 1996, vol. 143. pp.429436.
[ I l l C.P. Maddison.B.C.Mecrow.A.G.Jack.'Clawpolegmmeuiesfor high
perfmanee ~ TMS VCR~ flux machines', in Proceedings of the Sevennth
Int emt i onal Confewnee on Electrical Machi nrs. Istanbul. August
1998, pp. 340-345.
1121 B.C.Mecrow.A.G.Jack.C.P. Maddison. 'Permanent mamet machines . . -
for high torque. l ow sped applications'. in Proceedings of the Sixth
lntemrionnl Conference on Electrical Machines, Vigo. August 1996.
1131 C.P. Maddison, Trmmerreflurmockinrs for high torque applicntions.
R D thesis Univmity of Newcastle-upon-Tyne. 1999.
1141 D.H. Kang. T~~nrversolflurrmnrchinen met permonenter Erregung nls
Li nmrmt ri ebe imrehirnmgebundenen Vetkhehr, PhDTksisTechnische
Universitaet Camlo-Wilhelmina, Braunrhweig, 1996.
1151 H. de Pistoyes, 'Eddy c u m1 losses in L e stator ends of machines of
large pole pitch and means of reducing these losses', in Revue Gcncmle
de I'Eiect"citie, Paris. 1927, "01 21.