3

rd
International Symposium on Electrical Engineering and Energy Converters September 24-25, 2009, Suceava

33

Abstract This paper deals with the additional losses
estimation in the Roebel bar windings of high power
synchronous generators. In order to solve this problem, in the
paper is presented a strips-method that provides a high
accuracy estimation of the elementary currents through the
strand conductors. According to this method, the cross section
of each strand in N parallel-strips with the slot basis was
divided, so that the strip-current density could be considered as
a constant. The aim of the paper is to present the basic
elements of this approach.

Index Terms electromagnetic analysis, induction
generators, loss measurement, magnetic losses, resistance
measurement.
I. INTRODUCTION
Stator windings in large hydrogenerators use Roebel bars
in which the strands (elementary conductors or
subconductors) are complete transposed in the slot portion
by a variety of arrangements. In a normal bar, that has a 2
transposition, each strand has two crossovers, and, after
passing through the slot portion, is in the same position
relative to the other strands. Thus, each strand has the same
mean depth in the slot, the same leakage reactance and the
same voltage induced by the both main and leakage fluxes.
However, in the end-regions of the generator, leakage flux
flowing through the bars produces voltage differences
between the strands. Currents circulate between the strands
in addition to the normal load current, and additional copper
losses result.
Minimization of these additional losses is an important
objective having as effect the increase of the efficiency of
high power synchronous generator. The Roebel bar
optimization problem supposes to find the number of the
strands and the transposition angle in the slot portion so that
the bar losses have the minimum value. In order to solve this
optimization problem it is necessary to know all bar
currents.
The calculation methods of losses in Roebel bars used at
present are based largely on mutual impedances [1, 2] or
finite elements method [3, 4, 5] which involve the most
precise knowledge possible of the currents in the elementary
conductors and of the distribution of these currents on the
cross-sections.
The aim of this paper is to present an analytical model
that provides a high accuracy estimation of the strand
currents, taking into account all the magnetic fields that
produce some copper losses.
II. THE ROEBEL BAR MODEL
The structure of a Roebel bar results from Fig.1. In this
figure one can see the arrangement of the elementary
conductors in two columns, numbered from 1 to m on a
column.


Fig. 1 The Roebel bar structure

In Fig. 2.a.) the case of a bar composed of 8 elementary
conductors (m=4) placed on 2 columns is shown.
For calculation, this actual structure of the bar is replaced
with an idealized bar whose elementary conductors is made
of many longitudinal sections with axes which are parallel to
the slot basis and to the machine axis, as seen in Fig. 2.b).

Fig. 2 The Roebel bar model
a) Actual structure; b) Idealized structure

One considers an elementary section defined by the
vertical plans Z and Z+1 from Fig. 2. According to the
proposed method (strip method), the elementary section of
each strand in N parallel-strips with the slot basis was
divided, so that the strip-current density could be considered
as a constant. A family of Nstrips is formed, of strip
thickness h (Fig. 3). They are generally numbered from
ni to nf for the section of level . The levels are numbered
as one can see in Fig. 2.b). The number N of strips is chosen
so as to be able to consider that on the thickness h the
density of the electric current is constant, with an admissible
error.
This is the strips model that allows the analytical
computation of strip currents and strand currents of the bar
Roebel Bar Model for Additional Losses
Estimation in High Power Hydrogenerators
Toma DORDEA
1
, Victor PROCA
2
, Gheorghe MADESCU
1
, Marius BIRIESCU
3
,
Marian GRECONICI
3
, Marian MO
1
, Ileana TORAC
1
, Lucian OCOLIAN
1

1
Romanian Academy, Timioara Branch, Timioara, Romania, [email protected]
2
ICMET, Craiova, Romania, [email protected]
3
Politehnica University, Timioara, Romania, [email protected]
34
[6, 7] in order to estimate the additional copper losses.

Fig. 3 An elementary section of layer
III. THE SLOT MAGNETIC FIELD
In the slot of an electric machine there are two kinds of
magnetic fields. The two fields can generate additional
losses in the copper conductors located in the slot.

Synchronous generators are generally high and very high
power electric machines, with windings made of copper bars
located in slots. If an alternating current flows through such
bar, it generates a leakage field with lines closing
transversely through the slot (Fig. 4).

Fig. 4 The slot magnetic fields:
a) of inner current; b) and of external current

This leakage field, time-variable, induces in the bar eddy
currents which determine an unevenly current distribution
on the bar surface. Actually, it is found that the current
density is lower close to the slot base (Fig. 5) and higher
close to the slot opening (towards the air gap).


Fig. 5 Current density distribution in the bar

If three or many bar layers are in the slot, then similar
distribution curves are got.
Because of this effect, the electrical resistance of the bars
in alternating current (R
a
) is several times higher than the
electric resistance of the same bar in direct current (R
0
):
R
a
= k
ra
R
0
where kra > 1 (1)
Consequently, in the alternating current generator
winding, additional losses appear unavoidably.
For the computation of energetic parameters the lines of
the leakage field from the slot are considered to be parallel
to the slot base, i.e. perpendicular to the slot walls.
This field model is the basis for the computation of the
current displacement effect, namely for the coefficient of
increase of the bar electric resistance in alternating current
(k
ra
).
A slot where m-layers of rectangular bars are located is
considered (Fig. 6).
Each layer has certain coefficient of electric resistance
increase. By example, for the bar from the layer p, the
coefficient of resistance increase is [8]:
( )
( )
( )
2
u u p
p
p
I I I
k
I
+
= +

(2)
where the current I
p
is the current from the considered layer,
and I
u
is the sum of the currents from all the layers located
below the considered layer.


Fig. 6 Open slot with M- bars

If the same current flows through all the m bars, an
average value per slot of the coefficient (2) results [8]:
( ) ( )
2
1
3
ra
m
k

= +

(3)
where the functions ) ( and ) ( have the well know
expressions [8].
The problem of current displacement effect could be
solved in this way also in the case when the bar is formed by
many subconductors, connected in parallel.
The excitation field of a synchronous generator is
generated by the current from the excitation winding,
generally placed on the generator rotor. This magnetic field
passes from rotor through air-gap into stator teeth, but enters
also into the slot space, where the bars of stator winding are
located (Fig. 7). It generates inner circulating currents
through the bar subconductors, which are short-circuited at
the ends, forming short-circuit loops. These currents cause
additional losses in the stator winding, leading to the
decrease of the electric energy conversion efficiency.
This magnetic field from the stator slot area was
calculated, with view to determining the currents and
additional losses in bars. For this purpose, in a previous
paper [9] the conformal mapping method was used, in order
to benefit from the advantages of the analytical methods for
field computation.
The main results of the paper [9] in this paragraph are
presented.
It is considered the numerical example in which the slot
width b
c
= 24 mm, the air-gap =5 mm and flux density
B
max
=1 T.
The magnetic field in the slot was calculated by analytical
method (conformal mapping method) and by finite element
method (FEM), at different depths (y=0.... -10 mm). Some
computation results are presented in Fig. 8. Due to the
3
rd
International Symposium on Electrical Engineering and Energy Converters September 24-25, 2009, Suceava

35
symmetry of the magnetic field in the slot, the distribution
curves of the magnetic field radial components were
represented only on a half of the slot width (from the slot
wall up to the slot axis)..


Fig. 8 Some curves of the flux density radial component
across slot width direction: analytical method

On the basis of many computation examples, the
following analytical relations of the radial magnetic field
distribution in the slot width direction are proposed:
( )
|
|

\
|
=
c
rm r
b
x
B x B sin , for 2 y
( )
8 , 0
sin
|
|

\
|
|
|

\
|
=
c
rm r
b
x
B x B for
2
3
y
( )
5 , 0
sin
|
|

\
|
|
|

\
|
=
c
rm r
b
x
B x B , for y .
(4)

Close to the air gap, when < y , the distribution can be
no more approximated by a sine curve; within these zones,
the curves are like those ones calculated in Fig. 8 for y = -2
mm, y = -1 mm, y = -0.5 mm.
In order to develop as general as possible analytical
solutions, the radial component of the magnetic flux density
(B
rm
) in the slot axis and its distribution in the depth
direction was analyzed further on.
By using many families of curves got by computation,
also the MATHLAB facilities, an analytical relation for
approximating the values of the magnetic field density in the
slot axis was built under the following form, valid in any
point from the slot axis:
max
e


= B B
rm
(5)
where:

n m
n
y
m +

=
(6)
and:
1246 2654 4 , 909
1061 6507 1 , 665 39 , 38
1 2 3
1 2 3
|

\
|

+ |

\
|

+ |

\
|

+ |

\
|

+ |

\
|

\
|

=
c c c
c c c
m
b b b
b b b
m
3786 , 0 0857 , 0
1
+
|
|

\
|

=
c
n
b
n .

By using the approximate relation (5) the value of
induction in the slot axis can be determined, avoiding the
difficulties in using the conformal mapping relations. The
relations (4) and (5) in the computing program for additional
loses estimation was included.
IV. THE CALCULUS OF THE CURRENTS
One considers the elementary section belonging to the
layer (Fig. 3). One chooses a curve traced through the
upper parts of two consecutive strips with the order numbers
-1 and . One performs the line integral along this curve of
electric field strength and obtains a relation between the
currents of the two considered strips. The currents through
the strips and -1 are i

and i
-1
respectively. The currents
vary in a sinusoidal shape in time. In consequence, one can
use complex quantities. One can calculate the I
c
current of
the strip depending on the currents of the elementary
conductors under the form:

+ =
u c c
I T I S I (7)
Here I
c
is the sum of the two currents in the neighbor
conductors in the same layer , and I
u
is the sum of all the
currents in the conductors placed between the slot base and
the -layer. For S

and T

one considers the following

=
cu
cu
b
b
K ;
c cu
cu
b
b h
j B

2
2
;
C
1
= 1 E
1
=0;

=
B
K
A
3
3
;

=
B
B
D
3
3

( )


=

=
2
1
i
n
C D C D A C
( )
|
|

\
|
=

2
1
1
i
n
E D E D A E

=

=
N
f
C C
1
;

=

=
N
f
E E
1

f
C
C
S

= ;

= E E S T
f
.
(8)
In these relations: is the magnetic permeability of the
conductor material (virtually equal with the magnetic
permeability of the vacuum environment
0
); b
cu
the
width of the conducting cross section of an elementary
conductor in the strip; b
c
the width of the rectangular
slot; j is the imaginary unity.
36
The bar ends are treated similarly as the slot portions.
One then chooses another closed curve
1
traced through
the first strips of the two successive subconductors and
+1. This curve is closed at the two bar ends, where all the
2m-subconductors are short-circuited. A line integral of the
electric field intensity along
1
-curve was calculated, by
means of which a relation between the currents of the
subconductors was established. Thus are obtained (2m-1)
relations between the strand currents. One also takes into
account the fact that the sum of all the strand currents is
always equal with the current of the bar. Consequently:
( ) ( ) =

=
erc
m
U I G ,
1 2
1
, =1, 2,, 2m-1
I I
m
=

2
1

(9)
The G coefficients can all be calculated depending on
geometrical dimensions and of material constants, all
known, and U is the electromotive force from the strand
conductor induced by the radial magnetic fields. The system
contains (2m) equations, with (2m) unknown variables and
therefore the currents in the strand conductors can be
determined.
Now one can calculate the currents of every strip for
any strand conductor and one obtain for the strip of the
elementary conductor :
,
2
1
,
(
(

(
|
|

\
|
+
+

|
|

\
|
+
+ =

u N
N r
r
c N
N r
r
N r
r
I V
P
P
V
I U
P
P
U
B
I
P
P
I

(10)
where, generally K

= b
cu,
/ b
cu,-1
. At hollow strand
conductors of rectangular form there are only two
dimensions (b
cu1
and b
cu2
). Considering K= b
cu2
/b
cu1
the
following relations for the coefficients are obtained:

= K K K P
r
...
2 1
;

=

=
N
r N r
P P
1

=

=
1
S S ;

=

=
1
T T

=

=
1
S S ;

=

=
1
T T

=

=
1
S S ;

=

=
1
T T
(11)
|

\
|
=

S S K U
3
2

|

\
|
+ =

T T K V
3
2
1

=

=
N
N
U U
2
;

=

=
N
N
V V
2

and in a similar way for the ends winding.
Considering a constant current density on the cross
section of the strip, one determines the losses in the strip.
Adding the losses from all the strips, for all the elementary
sections, one obtains the losses in the entire bar.
V. CONCLUSION
Based on the previously presented model, using such kind
of relations, a computing program for both currents and
additional loses estimations was developed. The influence of
any input data (strands number, transposition angle,
geometrical dimensions, material data) can be considered.
The program allows a faster analysis of a bar structure and,
if is necessary, can provide an optimal bar structure. The
research work will be continued with an experimental model
in order to validate this approach.
ACKNOWLEDGMENT
Part of these researches was supported by the Romanian
Government PNCDI 2, with the collaboration of CNMP
Project 21028/2007.
REFERENCES
[1] Haldemann, J., "Transpositions in Stator Bars of Large
Turbogenerators", IEEE Transactions on Energy Conversion, Vol.19,
Nr.3, pp.553-560, Sept. 2004.
[2] MacDonald, D.C., "Circulating-current loss within Roebel-bar stator
windings in hydroelectric alternators", Proc. IEE, Vol.118, Nr.5,
pp.689-697, May, 1971.
[3] Chari, M.V.K., Csendes, Z.J., "Finite element analysis of the skin
effect in current carrying conductors", IEEE Trans. on Magnetics,
vol. 13, no. 5, pp. 1125-1127, sept. 1977.
[4] Preis, K., Bardi, I., Biro, O., Richter, K.R., "Nonlinear periodic eddy
currents in single and multiconductor systems", IEEE Trans. on
Magnetics, vol. 32, no. 3, pp. 780-783, may 1996.
[5] Arshad, W.M., Lendenmann, H., Persson, H., "End-winding
inductances of MVA machines through FEM computations and IEC-
specified measurements", IEEE Trans. on Industry Applications, vol.
44, no. 6, pp. 1693-1700, nov./dec. 2008.
[6] Dordea T., Biriescu M., Madescu G., Torac I., Mo M., Ocolian L.,
"Calcul des courents electriques par les conducteurs elementaires
dune barre Roebel. Part I: Fondements de calcul", Rev. Roum. Sci.
Techn. lectrotechn. et. nerg., vol. 48, nr.2-3, pp. 359-368, Bucarest,
2003.
[7] Dordea T., Biriescu M., Madescu G., Torac I., Mo M., Ocolian L,
"Calcul des courents electriques par les conducteurs elementaires
dune barre Roebel. Part II: Dtermination des courents", Rev. Roum.
Sci. Techn. - lectrotechn. et. nerg., vol. 49, nr. 1, Bucarest, pp. 3-
29, 2004.
[8] Richter R., Elektrische Maschinen. Erster Band, Verlag Birkhauser
Basel, s. 243, 1951.
[9] Dordea, T., Proca, V., Madescu, G., Greconici, M., Mo, M.,
"Magnetic field analysis in the open stator-slots of high power
hydrogenerators", Proc. of ECOS 2008, Cracow Gliwice, Poland,
pp. 1185-1192, Jun. 2008.