Beruflich Dokumente
Kultur Dokumente
I. INTRODUCTION
Stator winding faults are of great importance in large
synchronous generators as they not only damage the
stator winding of the machine severely, but can also
destroy the whole stator core. Common reasons for these
faults are material fatigue of the insulation conditioned
by mechanical stress in the end winding zone as well as
thermal influences and moisture [1, 2 and 3].
The transient numerical field and network calculation appears to be an adequate and advantageous tool for
the quantitative prediction of effects of stator winding
faults since experimental studies are much too costly or
completely unfeasible. In opposition to a merely network based analysis the transient variation of mutual
and self inductances caused by changes in the level of
iron saturation is implicitly taken into account since the
discretised model can be considered a representation of
the whole magnetic circuit.
The output quantities can be used for further evaluations, e.g. the reconstruction of disturbances having
finally destroyed a generator. Forces having been exerted on the end-winding zone of a destroyed generator
for instance can be calculated from the resulting coil
currents.
1
(F )
( ) +
(1 ) h
(0)
(FN,1 )
(FN,2 )
(0)
(Mb )
(F ) (F ) ( A)
T
N,1
(Cw )
(1 ) h (B ) (D )
w
w
lFe
(1 ) h (B )
f
lFe
(0)
(rA )
(Cf ) (xb ) (rb )
(Df ) (if ) (rf )
N, 2
upper
layer
lower
layer
upper
layer
lower
layer
ws_I
slot numbers
7
21
20
19
18
25
26
27
28
41
42
positive
coil group
U
39
40
negative
coil group
ws_II
21
20
19
18
17
16
15
38
37
36
35
34
33
32
ps_I
(1)
The right hand side of (1) is calculated from initial
values or the results of the previous time step. It is emphasised that all network equations are explicitly taken
into account without the use of reductive methods like
node potential or loop current formulations. This approach for network analysis is known as the sparse
tableau approach as referred to in [9].
The number of network equations appears to be negligible in the presence of thousands of field equations.
Moreover, ill-conditioned systems, which would result
from practically cutting or short-circuiting branches by
extreme values of their resistances in reductive methods,
can be avoided by the sparse tableau approach.
The solution of the linear system of equations is carried out directly, where a total pivot search is applied
during the elimination of the lower network matrices.
The nonlinear magnetic characteristic of stator and rotor
core material makes the reluctance matrix () become
field dependent. This requires a few iterations per time
step, where the permeability is adapted by underrelaxation.
In the program used here the stator was formerly related to fixed topologies and winding arrangements, i.e.
36
37
38
39
40
41
42
11
12
13
14
15
16
17
ps_III
ps_II
35
34
33
32
31
30
29
10
ws_III
8
10
11
12
13
14
25
26
27
28
29
30
31
Furthermore, the voltage between the winding neutral point and ground becomes an accessible output
quantity as the branch voltage of the related network
element. The partial windings w1, w2 and w3 with one,
two and three turns can be shunted by a resistive branch
in parallel in order to simulate a turn to turn winding
fault within one coil group. The separating nodes between the partial windings w10 w4 and w11 w5 as well
as w6 w12 can be interconnected by corresponding
resistive branches allowing for the simulation of winding faults between neighbouring phases. All cases are
listed in Table 1. The undisturbed operation is modelled
by interconnecting resistances of 1012 .
case
ws_I
ws_II
ws_III
ps_I
phases
U
V
W
U/V
coil group
positive
positive
negative
neg. / neg.
ps_II
V/W
neg. / pos.
ps_III
W/U
pos. / neg.
affected turns
2nd from terminal U
5th, 6th from terminal V
3rd, 4th, 5th (central turns)
1st / 1st, 2nd from neutral
1st, 2nd from neutral
/ 1st-3rd from terminal W
1st-3rd from terminal W
/ 1st from neutral
Fig. 1d. Finite-difference polar grid of the machine cross-section Slot
numbering clockwise beginning with the first slot above the horizontal
line on the left side
ps_II
ws_I
ws_II
ws_III
ps_III
ps_I
Fig. 1c. Winding scheme with locations of short circuits within single coil groups ( ) and between different phases ( ). Shorted turns are marked
by for the shorts in a single coil group, and by for the shorts between different phases.
Case
ws_I
ws_II
ws_III
ps_I
ps_II
ps_III
|isc, peak|
kA
215.2
207.9
307.4
139.4
250.1
259.2
Isc, eff
kA
107.8
118.7
128.2
96.6
113.5
103.5
IV. RESULTS
A. Currents
In all investigated cases high short circuit currents
occur as listed in Table 2.
Obviously the transient behaviour after the occurrence of the short circuit is stronger the more windings
are shorted, as shown in Fig. 4c for case ws_III with
three shorted windings. The maximum effective value
occurs in the phase following the one with the shorted
turns, whereas the line current in the phase before is
decreased in cases ws_II and ws_III.
Winding faults between different phases are of much
higher effect regarding the terminal currents, as shown
in Fig. 5 a-c. The current peaks are much higher and a
stronger asymmetry occurs.
Case
ws_I
ws_II
ws_III
ps_I
ps_II
ps_III
G
F peak
G
F
MN
1.076
2.206
1.267
1.980
7.361
6.364
MN
0.741
1.718
0.582
1.894
4.157
3.275
max
G
F
max, disc
MN
0.750
1.604
0.502
1.850
3.541
2.753
G
F max,1.8s
MN
0.257
0.388
0.0847
0.496
0.525
0.558
G
The transients level of to a steady state value | F | max
until the machine is disconnected from the power grid.
Thereafter the force magnitudes decay like the shortcircuit currents. Case ws_I shows the lowest values in
Table 3, but is still more critical than rotor winding
shorts.
Figures 7 and 8 display the trajectories of the magnetic
force exerted on the rotor in the d-, q- reference frame.
Since all cases generate a similar pattern, only the least
and most critical cases are included here, namely ws_I
on the left side and ps_II on the right. The upper trajectories are shown during 0.2 s after the instant of the
short circuit and the lower ones display the time after
disconnection during de-excitation.
Fig. 7a. Trajectory of force vector on rotor in rotor reference frame for
a period of 0.2 s after winding fault in phase U (case ws_I).
Fig. 8a. Trajectory of force vector on rotor in rotor reference frame for
a period of 0.2 s after the fault between V and W (case ps_II).
C. Field map
The field map in Fig. 9 on the cross-section of the
investigated machine at an instant of time with maximum rotor forces at 3.693 s (case ps_II) shows obvious
irregularities. It illustrates the influence of the asymmetric field distribution on the unbalanced magnetic pull on
the rotor. The shorted coils are marked with the reference direction of the short-circuited loop.
Obviously the current in this loop strongly disturbs
the field. Knowing that normal components contribute
tensile stress to the fictitious Maxwell stress tensor,
whereas the tangential components result in compressive stress, the extreme high forces may be explained.
REFERENCES
[1] T. S. Kulig, Die innere Unsymmetrie von
Synchronmaschinen, PhD thesis AGH, Krakau
1974.
[2] T. S. Kulig, ber die Beeinflussung der Strme und
des Elektromagnetischen Drehmoments von
Turbogeneratoren
durch
Windungsund
Phasenschlsse, PhD thesis, Hannover 1979.
[3] M. Daneshnejad, Erfassung von Windungsschlssen
in
der
Erregerwicklung
eines
Turbogenerators, PhD thesis, University of
Dortmund, Institute of Electrical Machines, Drives
and Power Electronics, 2001.
[4] M. Klocke, Zur Berechnung dynamischer Vorgnge
bei von einem Drehstromsteller gespeisten
Antrieben mit Asynchronmaschinen und mehreren
gekoppelten Massen mittels Finite-DifferenzenZeitschrittrechnung, PhD thesis, University of
Dortmund, Institute of Electrical Machines, Drives
and Power Electronics, 1999.
[5] R. Gottkehaskamp, Nichtlineare Berechnung von
Asynchronmaschinen mit massiveisenem Rotor und
zustzlichem Dmpferkfig im transienten Zustand
mittels Finiter Differenzen und Zeitschrittrechnung,
PhD thesis, University of Dortmund, Institute of
Theoretical Electrical Engineering and Electrical
Machines, 1992.
[6] A. Krawczyk, J. A. Tegopoulos, Numerical Modelling of Eddy Currents, Oxford: Clarendon Press,
1993.
[7] T. S. Kulig Anwendung der numerischen
Feldberechnung zur Modellierung elektrischer
Drehstrom-Maschinen mit inneren Fehlern Bulletin
SEV/VSE(1990)7.
[8] M. Klocke, M. Daneschnejad: New Aspects of
Winding Faults in the Rotor of a Large
Synchronous Generator, Record of the 2001 IEEE
International Symposium on Diagnostics for
Electrical Machines, Power Electronics and Drives
(SDEMPED 2001), p. 161-166. Grado, Italy,
September 1-3, 2001.
[9] W. Mathis: Theorie nichtlinearer Netzwerke,
Springer-Verlag, Berlin, Heidelberg, New York,
1987.
V. CONCLUSION
Six different stator winding faults differing in the
number of shorted turns and affected phases have been
investigated by means of a combined transient networkfield computation. In all cases a strong impact on the
machine arises while the machine is still connected to
the power-grid. But even immediately disconnecting the
machine from the grid cannot be considered an effective
measure for reducing the occurring damage unless a
quick de-excitation is provided. The turn currents remain at a high level of about six to nine times their rated
values in the short circuit loop and also winding parts
not included in the short circuit path are exposed to
inadmissibly high currents, if the excitation remains at
its previous level.
As for the rotor, attention has been paid to the unbalanced magnetic pull which besides the damper currents
and eddy current losses not dealt with here appears to be
a quantity with problematic values. Further investigations concerning the mechanical consequences of such
strong impacts on the mechanical properties of bearings
and the shaft train seem to be indicated. However, practical observations of bearing damages or permanent
bending deformation of rotor shafts caused by stator
winding faults are not known to the authors.