Anda di halaman 1dari 10

IEEE Transactions on Power Delivery, Vol. 8, No.

3,July 1993 1367

FEM ANALYSIS OF DC SATURATION TO ASSESS


TRANSFORMER SUSCEPTIBILITY TO GEOMAGNETICALLY INDUCED CURRENTS

SHU LU (Student Member, IEEE) YILU LIU (Member, IEEE)


The Bradley Department of Electrical Engineering
Virginia PolytechnicInstitute and State University
Blacksburg, VA 24061

Abstract: The paper presents results of a systematic H, is in the order of 10 A-thch (about 400 A-t/m) along
finite element (FEM) simulation of transformer
susceptibility to geomagnetically induced currents (GIC).
DC saturation patterns for five different transformer core
!
the core lamination [3 Leakage or stray flux is wthm
tolerance level as far as osses are concerned.
designs are examined usin FEM; magnetic fields along Under GIC bias, the saturated core becomes a much
various traverses are provifed. Impedance matrix entries higher reluctance path which requires more ampere-turns
of a single phase transformer are compared for normal to produce the same amount of flux in order to sustain the
and dc saturation operations, The study aims to s stem sinusoidal voltage level. Also, the core nonlinearity
contribute in the understanding of fundamental Jemands extremely large harmonic components in the
transformer magnetization mechanism under GIC. exciting currents. Reduced core permeability and
increased exciting magnetic gradient result in a large
Ke words: Power transformers, Solar Magnetic amount of stray flux bemg generated. As a consequence,
Dis*ransformer Magnetization, Finite Element stray flux heating could occur in windings, structural
Analysis, Geomagnetically Induced Currents. metallic parts, winding supports, and the tank plate. Local
hot spots could develop which could damage paper
insulation and cause gassing and transformer oil
INTRODUCTION combustion.
Atmosphere Solar Magnetic Disturbance (SMD) will Many aspects of the effect of eomagneticall induced
induce voltage on the earth's surface [1,2]. The potential currents on power systems have %een studied c ruring the
difference between different locations will cause last 25 ears by Albertson, Kappenman, Aspnes, et. al..
geomagnetically induced currents (GIC) to flow through GIC efkcts on power transformers, in particular, were
power transformers via grounded wye neutrals. GIC are studied by Gattens, Girgis, and Nevins [4], Ringlee and
Steward [3], Walling and Khan 51, Bolduc and Aubm [lo],
tJcpicaUy in the range of 0.001 to 0.1 Hz, and could be as
hgh as lOOA and last from several minutes to hours. As
these quasi-dc currents flow through transformer
I
and McNutt [SI. To study trans ormer heating due to GIC,
measurements on scale models by Pasco, Norton, Nilsson
a series of reactions will occur associated with [6], Tay and Swift [ll] as well as field tests on actual
~ndF
the h cycle saturation of the transformer core. The most
profound effects are the extremely large harmonic
transformers by Kappenman [I, were erformed. Very
P
few reports on transformer flux ana ysis usin finite
contents in excitation currents, increase in transformer elements method were found [4 5 and they often focused
reactive power consumption, false operation of protective on one particular design. O v e r 2 a comprehensive FEM
relays caused by harmonic currents, and transformer analysis covering most of the transformer core designs was
heating caused by stra flux that were "pushed out" of the not available.
magnetic core due to l e a saturation. Different degrees
of heating symptoms in ei& EHV autotransformers were In this paper, a systematic finite element analysis of
observed during the severe geomagnetic disturbance in transformer GIC magnetization is presented. DC
March 1989. saturation of five simpllfied generic core configurations
are examined. The studies include the analysis of profdes
A transformer core is made of high permeabilit of magnetic field intensity (H) and transformer impedance
ferromagnetic material. Its role is to provide a we8 matrix (Z) parameters. A finite element computer
defined low reluctance path for magnetic flux. During software [9] is used for this study. The objective of this
normal o eration, the majority of the flux will stay in the initial study is to understand the trends of dc flux
core. O& about 1percent of the load current is required distribution under different levels of core saturation for
to maintain the magnetizing force. The magnetic gradient, various transformer designs and core configurations.
Each design is a unique case. Although completely
92 SM 510-8 PWRD A paper recommended and approved general laws are not practical based on the generic cases
by the IEEE Transmission and Distribution Committee analyzed at the present time, typical cases can be used to
of the IEEE Power Engineering Society for presenta- infer similar designs.
tion at the IEEE/PES 1992 Summer Meeting, Seattle,
WA, July 12-116, 1992. Manuscript submitted February
3, 1992; made available for printing May 15, 1992.
DC MAGNETIZING FLUX SIMULATION
DC flux distributions of simplified core geometries,
winding and tank structures are studied first. The
objective is to study the GIC effect on core saturation for
0885-8977/93$03.000 1992 IEEE
1368

various transformer designs and core configurations. This Cpc = 2Qr (1)
study also considers the effect of extreme GIC levels that The reluctance or each segment of the core can be
may have not been observed in the past, and cases which expressed as,
would be difficult to achieve in full scale tests due to either
limitations of adequate dc test sources [6] or the risk of
damaging a healthy transformer [7]. The following
transformer core configurations are included in this study.
where p = po pr,L is the length of the segment, and A is
the cross sectional area. The core can be modeled by a
lumped magnetic equivalent circuit shown in Figure 2.
Since the cross sectlonal area of the side legs and the
yokes is half that of the main leg, the dimensions of the
transformer core are,
Ar = 1/2Ac, Lr = 5ft, Ly = 4.5ft (3)
Fundamental GIC Excitation Analysis of
a Single Phase Transformer Design where Lr and Ly are the effective length of the side legs
and the yokes respectively. The reluctance values of the
The single phase designs can always provide a return elements in the equivalent circuit are calculated as,
path or zero sequence magnetic flux. Flux distribution is
well confined inside the core during normal operation. Rr = 2 Rc, Ry = (915) Rc (4)
Under GIC excitation, the dc current will bias the core at The mmf is,
different saturation regions depending on the magnitude
of the current. As a result, part of the flux is being pushed 2Ry + Rr
out of the saturated core, causing leakage flux, or stray F = Qc(Rc + 5
) = K7Q c R c (5)
flux, surrounding the core. Flux distribution of a single
phase transformer under heavy saturation is shown in Thus,
Figure 1. Because of its tendency to seek a low reluctance
path, the stray flux is going to concentrate in regions where
magnetic materials exist. Therefore, heating problem due
to eddy loss is likely to occur in these regions under ac
operation.
Br = Q c/2
1 F
When a transformer saturates, the whole core usually
does not saturate on the same level. That is, part of the
-
Ar
= --
3.8 RcAc
core may saturate more than other parts, while some part
may not saturate at all. The core saturation patterns vary where Rc, Rr and Ry are the reluctances of the main leg,
with, not only core configurations, but also the core leg side leg, and the yoke. F is the magnetomotive force
dimensions. (mmf). @.c, Bc, and Ac are the flux, flux density and the
cross sechonal area of the main leg, respectively. Br is the
flux density of each side leg or yoke.

Figure 2. A lumped magnetic equivalent circuit of the


Figure 1. Single phase transformer core configuration and single phase core
its dc flux distribution under heavy saturation
Based on the equivalent circuit, two additional cases
Figure 1shows a slngle phase transformer core design. with respect to Figure 1 are examined. First, the side leg
Since only dc excitation current is applied for the cross sectional area Ar is increased such that Ar is equal
simulation, onl the primary winding is drawn on the to Ac. Then,
figure. The folzwing parqraphs describe the impact of
changing side leg cross section on overall core saturation. A r = A c , Lr = 6 f t , L y = S f t (8)
During normal operation without GIC, assuming no Hence the flux density can be calculated,
leakage flux leaving the core, the operating point lies in the
linear region of the B-H curve and the core possesses a
somewhat uniform permeability p If the cross sectional
Bc = 1 -- F
2.8 RcAc (9)
area of the side legs and yokes is LaI of that of the main
leg, the relationship between the main leg flux, Qc, and the 1
Br = -- F
side leg flux,Qr, as shown in Figure 1,is given by, 5.6 RcAc
1369

I .. I
I
00 50e4 lie5 15b5 00 50e4 1 Oe5 15e5 00 5Oe4 1 Oe5 1 5 e 5
H [Ampslmefer) H [Ampsimeler) H [Ampslmefer)
a) nrde leg area I S equal fo he m a n leg area b) r,de leg area IS twice of b.n main leg area C)I side leg area #aa qwner o i he main leg area

Figure 3. Biasing levels for each core segment under the same low dc excitation for Single phase transformers
Second, the side leg cross sectional area Ar is may not be the same depending on core configurations.
decreased such that Ar is equal to a quarter of Ac. Then,
H profiles along A and B lines (Figure 1) under heavy
Ar = 1/4 Ac, Lr = 4.5 ft, Ly = 4.25 ft (11) saturation is shown in Figure 4. It is found that the inside
corner regions have the highest H if at least one leg
The flux density is calculated as, adjacent to the comer is saturated. The stray flux shows
more concentration in those regions. Also, two peaks are
Bc = 1 --
6.1 RcAc
F seen at the edges of the windings.

150%
Br = -1 - F I . . . . . . . +,
3.05 RcAc . . . . . . . . . .
+ w

It can be seen from the flux density results that for the
same dc excitation level, the larger cross sectional area Ar
results in the lower dc flux density in the side leg while
higher dc flux density in the main leg. Therefore, one can
conclude that:
. . . . . . .
a) Comparing the main leg saturation level of the three . . . . . .
cases, the one with the largest side leg will saturate first.
b) Comparing the side leg saturation level of the three
cases, the one with the smallest side le will saturate first. 0.0 2.0 4.0 0.0 8.0
c) For the case shown in Figure 1t%atthe side leg cross M
sectional area is half of that of the main leg, the entire core Figure 4. H rofdes along transverses of a single phase
will reach saturate at the same time. transformer
Resulted from finite element solution, the biasing levels
for each core segment under low saturation dc excitation GIC Effects of Three Phase Core Configurations
are shown in Figure 3. Note that once the core reaches
saturation, calculations based on the above e uivalent Under balanced GIC excitations in the three phases, it
circuit are no longer valid because the permeabity pr is seems logical that the following magnetic designs are more
not uniform. Also, equation (1) does not hold due to the susceptible to half cycle saturation since they allow a flux
existence of stray flux. path for zero sequence excitation "f.Those include
three phase conventional as well as three phase five leg
The results obtained from the above analysis can be and seven leg core designs.
applied to different transformer core configurations. The
core saturation atterns for a particular transformer can
be used to roughy predict the patterns for those having
the same core configuration but different core leg
dimensions.

H Fields Along Transverses


of A Single Phase Transformer
Examining the field intensity H outside the core
regions can help understand otential stray flux heating
effect in those regions. Since 8ux density B is proportional
to H, knowing the magnitude of the field intensity H in the
air region can lead to rough prediction of eddy losses if
conducting material exists. Areas having higher H are
like1 to be more sensitive to heating problem, although H Figure 5. Three phase three leg core form transformer
will Be redistributed in the presence of magnetic materials. configuration and flux distribution under
It is important to point out that ac and dc flux patterns balanced dc excitation
1370

5
0-. 5*-0.0
H (Ampolmeter) H (Amps/meter)

a) thrm phase firm leg c m l0rm b) Uwm phue five leg C) thm, *S. reven leg

Figure 6. Biasing points of different three phase core designs under the same balanced dc excitation currents
(shown by the square markers)
Other designs where a closed path through the iron It is verv imDortant to notice that a three phase
core for zero sequence mmf is not possible are expected to threeleB core form transformer will eventually reach
be much less susceptible to GIC. The three phase three saturation if the dc bias is increased further. In this
leg core form is an example. Its configuration including simulation, when the dc excitation current is doubled, the
tank is illustrated in Figure 5. When the three hases are core will operate at the knee of the saturation curve, as
biased under balanced dc excitation currents, &ere exists shown by the cross markers in Figure 6.a.
no return path for the dc flux. DC flux tends to leave the
core. Since the air exhibits a very high reluctance path for The same GIC bias as that used in the three phase
the magnetic flux, the dc flux bias IS low. Consequently, three leg core form transformer is applied to a three phase
the three phase three leg core type is less susceptible to five leg and a seven leg transformers. Saturationoccurs as
GIC if the dc excitation 1s balanced. This is illustrated by expected in both cases. A comparison of the saturation
showing its operating points in the linear region under very magnetizing flux between these two cases is illustrated by
high excitation currents in Figure 6.a. Figure 7 and Figure 8. The results indicate that art of the
seven leg core has higher saturation level than %e five leg
core. Their bias points are shown in Figure 6.b and 6.c.
A three phase conventional core design is shown in
Figure 9. As we increase the dc excitation, it is apparent
that region 1 and 2 will saturate first since they provide
paths for two branches of flux. The same dc bias as that
used in the three phase three leg core form is applied.
Region 1and 2 are now under heavy saturation.

Figure 7. Three phase five leg core transformer


configuration and flux distribution under
balanced dc excitation

Figure 9. Three phase conventional core transformer and


flux distnbution under balanced dc excitation
Based on the simulation with the specified three hase
core dimensions shown from the figures, an or er of cf
increasing susceptibilityto GIC can be given as following:
- The three phase five le core
- The three phase seven feg core
- The three phase conventional core
This order may change if the proportion of the dc flux
return area (or side leg dimensions) varies. Because a
Figure 8. Three phase seven leg core transformer transformer saturation level depends on both the core
configuration and flux distribution under codiguration and the core dimension, there is no absolute
balanced dc excitation susceptibilityorder for the above configurations.
1371

Effect of Unbalanced GIC Bias


on A Three Phase Three Leg Core Design a 5c.
' A ' * I
How neutral GIC current divides among transformer
windm is determined by the dc resistances along the . . .
three prases as well as the system resistances. Usually it is *b5i . . . .
assumed that the three phase windings have approximately
the same resistance and the rest of the power system has
less effect due to smaller resistances. Experiments (3,q on
actual transformers tend to suggest that the GIC may not
be evenly distributed among the three phases. Some
degree of flux cancellation may result due to this
unbalanced situation.
Flux distribution of the three phase three leg core form
transformer under unbalanced dc excitations was
examined for two setups. First, the unbalanced excitation
is obtained by decreasmg the dc excitation current in one 0.0 9.0 0.0 ea
of the three phases. For example, current in winding 2 is F d
decreased. Second, the unbalanced excitation is obtained
by increasing winding 3 current. The resultant flux
dtstribution for the second setup is shown in Figure 10. A I I I

return path is provided for the net magnetic flux due to the
unbalanced flux in the three phases. As a result, part of
the core will saturate, as illustrated by the flux density in
Figure 10.

Figure 11. H profiles along the transverses of the three


phase three leg core form transformer

2. H profiles along transverses outside the tank


Figure 10. Three phase three leg core form transformer
flux distribution under unbalanced dc excitation H along lines D and F for all configurations decreases
uniformly towards the direction leaving the tank, except at
the comers where there are small peak values, as shown in
Figures 11and 12.
H Field Profiles of Three Phase Transformers
3. H profiles along transverses across the main leg
1. H profiles along transverses outside the core
H values along lines A and B in Figure 11are low in
Considering the three phase three leg core form and the corners since the dc bias inside all the main leg cores
the five le transformer mentioned above, the magnitude are low. They peak in the air outside of the corners and in
of the fielcfintensity H along lines A ~rF (shown in Figure regions between windings a and b, b and c. In Figure 12,
6 and 7) are lotted in F' es 11 and 12, respectively. It the H value alonF line A has almost the same shape as that
can be seen tgat the H v x s along lines C and E outside in Figure 11, whde the H along line B shows two peaks in
the core are similar in these configurations. Since high H the side leg corners due to the side leg saturation. Since
is found outside the cores, high dc flux density near the yokes 1 and 2 are under heavy saturation in the
tanks is expected. conventional shell form configuration, it is understandable
that H near these two yokes is very high.
H outside the three phase conventional shell form core
is also examined. An interesting phenomenon occurs that The three phase seven leg transformer has similar H
H along line C and line E, as shown in Figure 13, is much profiles as the five leg configuration except that along B
lower compared with those in the previously described line there is a minimum appearingbetween windings b and
three phase configurations. Flux density inside the tank is c as well as windings a and b. This is resulted from the
therefore very low. extra core legs between the two adjacent windings.
7 * A ' .

+ m
. . . . I .
I
TRANSFORMER IMPEDANCE MATRIX
AND CIRCUIT MODEL UNDER GIC
2.8.5'
' * c
* D ' ' . . It is usually convenient to model magnetic circuits
. . . . . . .
using their electric analogies. A typical transformer model
for low frequency is the equivalent Tee circuit (Figure 14).
It consists of leakage inductances Llkl and Llk2,
magnetizing inductance M, and ac wind resistances,
R1 and R 2 The shunt resistance R is u s 3 to represent
core losses. Transformer two po3network impedance
matrix is defined as:

0.0 3.0 6.0 0.0 12


Fen

v1 v2

Figure 14. Tee equivalent circuit model of a single phase


transformer
The impedance matrix of a single hase transformer
Figure 12. H rofiles along the transverses of the three without including tank is calculatef using the finite
phase &e leg core transformer element method. Figure 15 shows the single phase
transformer 2D model with finite element meshes. Both
3.k5
primary and secondary windings are included in the
I . . I . , . : . . ' . . I model. In this example, the core conductivity is set at
l e + 6 4 s . The windin conductivity is 5.8e+7 m/S
(copper). Since the impefance matrix calculation involves
ac operation, the periodically varying flux inside any
conducting materials tends to be crowded toward the
surface. This phenomenon is known as skin effect. The
numerical calculation requires that the largest mesh size of
any conducting material must be smaller than half of its
skln depth. In order to meet this requirement without
drastically increasing the computation time, the model has
been scaled down to millimeter range and a frequency of
0.0 3.0 6.0 0.0 12 10 Hz is used. Modeling of core lamination becomes less
Foot critical with the smaller core size. Three cases are used to
' 3.5.5 1 simulate the different GIC bias levels.
I

= Ii'.;
a
5"
2.1.5

Xl.4.5

:+; 1
7.64
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
.

. .
" ' " " ' 1
. . . . . .
:
0.0 2.0 4.0 0.0 8.0

Foot
Figure 13. H profiles along transverses of the three phase
conventional core transformer Figure 15. Single phase transformer with FEM mesh
1373

Case 1 simulates the normal operation assuming an indicates that the mutual inductance MT is nearly in
average relative permeability of 50,000 for the core proportion to the core permeability, which is true when
material. The parameters as defined in equation (14) is Rm is much greater then w M.
given by Z I . Case 2 is under saturation. Since the relative
permeability pr is no longer uniform in the core as Leakage Inductances
discussed earlier, piece-wise p is assigned for different
parts of the core, as shown in Figure 16. gives The leakage inductance Llk of the primary winding
parameters for this case. Case 3 simulates the extreme for this particular design can be iound from the difference
case by assuming an air core (pr = 1). Its parameters are between L11 and L12
given by %.
Normal operation:
L k l * L11- L12 = 3.6205e-3 - 3.6202e-3 = 2.6e-7 MKS

-
Under core saturation:
Lkl Lll - L12 = 2.5730e-5 - 2.5650e-5 = 0.8e-7 MKS
Air core case:
L k l * L11- L12 = 6.1629e-7 - 4.6033e-7 = 1.5e-7 MKS
These numbers show that leakage inductance varies a
Figure 16. A piecewise model of a single phase little under dc bias. This indicates that net change of the
transformer core under saturation, the numbers leakage flux introduced by the load current is almost
indicate values of pr in each region negligible. The major leaka e flux under saturation is
resulted from the drasticaiy increased magnetizing
All units in the impedance matrixes are in MKS. Note current.
that the 2 matrix entries are for an conceptual transformer
design. Only the changes of those parameters are used to Resistances
illustrate the change due to GIC bias:
The difference between R11 and R 1in the 2 matrix is
the winding loss RI in the equivalen? circuit. Here R1
(1.8771e-1,3.6202e-3) (2.0304e-1,3.6206e-3)1
[ (2.0303e-1,3.6205e-3) (1.8771e-1,3.6202e-3) represents copper loss as well as eddy current loss in the
Z1= windmg. Rm represents the core loss which was
k y s e c t e d in the expressions of all the 2 matrix entries.
(1.5326e-2,25730e-5) (6.2434e-6,2.5650e-5)
6.2434e-6,2.5650e-5) (1.5326e-2,2.6240e-5) I Based on the above discussion, if the basic parameters
in the transformer equivalent circuit are to be reserved,
they will have to be modified accordingly. TBis merits
(1.5326e-2,6.1629e-7) (4.3721e-9,4.6033e-7)
4.3721e-9,4.6033e-7) (1.5326e-2, 1.006Oe-6) I further study since saturation occurs only during part of
each half c cle and the model needs to reflect the dynamic
nature of t i e transformer operation under GIC.
The following changes can be observed:
Simplifications
Mutual Coupling
In our study, some 2D simulations have been compared
The tight coupling introduced by the iron core is less with 3D simulations. On the symmetric lanes, 2D and 3D
effective at saturation. The coupling coefficient K is used
to indicate the coupling efficiency.. It is given as:
simulations show very similar results. T%Sproves that 2D
simulation can also provide useful informahon. Windings
are all modeled as concentric cylindrical arrangements.
Core laminations and structural parts are not included in
K = MT / ( L + ~b d~u 2 ' (15) the model. In the impedance matrix simulation, skin
where, effects requires extreme1 small mesh size. In order to
MT = L12 or h1 make computation possibre, the core dimension is scaled
down to millimeter range. This approach will, in the
FEM analysis of the above single hase shell form meantime, reduced the error of not modeling laminations
transformer shows that coupling coefgcients decrease in the core.
when the saturation level increases. Here, the coupling
coefficient is 0.9999 under the normal operation and
0.9872 during core saturation. The air core coupling CONCLUSION
coefficient is 0.6521.
The FEM analyses described in this paper have
Mutual inductances between the primary and provided new findings regarding GIC susceptibility of
secondary winding decrease from MT = 3.620e-3 MKS to different transformer core conf'iiations, both confirmed
MT = 4.603e-7 MKS corresponding to a change of and modified some early predictions based on analytical
average pr from 50,000 to 1, respectively. The simulation approaches.
1374

The study concludes that All single phase and three phase [4] P. R. Gattens, R. Girgis, R. Nevins, "Investigation of
designs are susceptible to GIC! A transformer core Transformer Overheating Due to Solar Magnetic
saturation pattern is determined by both the core Disturbances," IEEE Special Panel Session Report,
configuration and the relative core leg dimensions. A three 9OTHO291-5 PWR.
phase three leg core form design could also reach saturation [SI R. A. Walling, A. H. Khan, "Solar-Magnetic Disturbance on
and become susceptible if the dc bias is increased further. It Power System Performance and Security," EPRI GIC
will also be susceptible if unbalanced GIC currents in the Workshop Report, 1990.
three phases become significant. The three phase [6] W. Pasco, E. T. Norton, S. L. Nilsson, "High Voltage Direct
conventional core design in our study seems to have less Current Converter Transformer Magnetics," EPRI Report
heating risk in the tank since H (dc) outside the core is low. EL-4340, Research Project 1424-3, December 1985.
AC analysis is needed in order to confirm this. [7] J. G. Kappenman, "Transformer DC Excitation Field Test &
Results," IEEE Special Panel Session Report, 9OTHO291-5
Under GIC bias, the major source of possible stray flux PWR.
heating is very likely the magnetizing current which [8] W. J. McNutt, "The Effect of GIC on Power Transformers,"
contains higher harmomcs and has an increased total effect IEEE Special Panel Session Report, July 1990.
during saturation. 91 Ansoft Corporation, << ANSOFT n, University Technology
Development Center, 4516 Henry Street, Pittsburgh, PA
Including the transformer tank in the simulation proves to 15213, (412)683-4846, January, 1989.
be necessary. The results show that the magnetic field 101 L. Bolduc, J. Aubin, "Effects of Direct Currents In Power
intensity outside the core with tank is much higher than Transformers,Part I. A General Approach, Part 11, Simplified
without the tank. Obviously, winding location also affects Calculations for Large Transformers," Electric Power
the magnetic field distribution during saturation. Under Systems Research, Vol. 1, 1978, pp. 291-304.
normal circumstance, the effect is much less significant. 111 H. C. Tay, G. S. Swift, '*Onthe Problem of Transformer
Overheating due to Geomagnetically Induced Currents",
The work presented in this paper is the first phase of IEEE Trans., Vol. PAS-104, No.1, Jan. 1985, pp.212-219.
investigation in order to assess the possible stray flux
heating problems in a GIC affected transformer. Future
work will include eddy loss analysis using actual
transformer design data. Better methods of combined
application of dc and large ac excitations in FEM study will Shu Lu was born in Guangzhou, China in December 1962. She
be investigated. More structural details and core received the B.S. degree from South China Institute of
laminations may be considered in the problem. Thermal Technology, Guangzhou, China in 1984. She -eceived the
study will be performed based on knowledge of eddy losses M.S.E.E. degree from Virginia Polytechnic Institule and State
analysis. University in 1990 and is working towards her Ph.D. degree at the
same school.
It is hoped that the study will help to reveal and confirm
the fundamental transformer stray flux heating mechanism was born in Chengdu, China, in 1959. She received
and provide information for predicting future transformer the B.S.E.E. degree from Xian Jiaotong University, Xian, China
heating possibilities with respect to different designs and in 1982, and the M.S.E.E. and Ph.D. degrees from The Ohio State
core configurations; classification of the GIC subjected University in 1985 and 1989. Dr.Liu is an Assistant Professor of
transformers in the igneous rock zone by their potential risk Electrical Engineering at Virginia Polytechnic Institute and State
factors may become possible. University. Her current research interests include: power system
transients and harmonics; transformer modeling and finite
element analysis. She is a member of IEEEPES T&D working
Acknowledgment group on geomagnetically induced currents and their effects to
power systems.
This study is being supported by the National Science
Foundation under Grant Number ECS-9018443 and the
Virginia Polytechnic Institute and State University
equipment grant. The authors would like to thank Dr.Jaime
De La Ree at Virginia Tech for providing essential
computing facilities.

References
(11 V. D. Albertson, J. A. Van Bellen, "Electric and Magnetic
Fields at the Earth's Surface Due to Auroral Currents," IEEE
Trans. on Power Apparatus and Systems Vol. PAS-89, No. 4,
April 1970, pp. 578-584
[2] J. G. Kappenman, V. D. Albertson, "Bracing for the
Geomagnetic Storms," IEEE Spectrum, March 1990, pp.
27-33.
[3] R. J. Ringlee, J. R. Steward, "Geomagnetic Effects on Power
Transformers," IEEE Special Panel Session Report,
90TH0291-5 PWR.
1375
Discussion and maximum saturation can occur in different locations in
the core structure when ac and dc are considered together.
R. A. Walling ( G E I n d u s t r i a l and Power Systems) - Can the conclusions regarding the relative susceptibility of
Finite-element analysis is a powerful method for revealing the
magnetic behavior of transformers. The discusser is seriously different transformer types to GIC be considered generally
concerned, however, that the limitation of the analysis to valid when they were made using dc-only analysis and the
dc-only excitation could lead to conclusions which are not flux offset values do not necessarily represent that which
necessarily valid under the realistic condition of simultaneous can be expected in practical circumstances?
ac voltage and direct (GIC) current excitation. Under such
combined ac and dc excitation, the flux atterns must b e Why do the authors consider drastically unbalanced GIC
considered instant-by-instant over the full Rndamental cycle. flow in the different transformer phases? Considering that
At no point in the cle are the conditions the same as for dc transmission systems have resistances which are very well
excitation alone, s u x as discussed in this paper. balanced in practice, and further considering that a
transformer cannot develop a dc volta e except by direct
For the dc-only excitation analysis described in this paper to current flow through its (well balanced! resistance, how do
have practical value, one must assume that the dc and ac flux the authors hypothesize that such a GIC phase imbalance
effects a r e superimposable. To ex lain this further, could ever occur? Would not the findings of Reference [SI
considering dc excitation only, a very smalramount of GIC will regarding the different flux offsets appearing in center and
bias the flux to a level more than the rated value and nearly to outer phases of three-phase transformers be a better
complete saturation. This is illustrated by the flux bias shown explanation of the differences in phase saturation observed
in the pa er, such as Figure 3a. If one then superimposes a 1 in the field?
p.u. funfamental-frequency flux wave on the main legs,
centered on this bias point, nearly half of the ac flux wave is in X. Chen (Department of Electrical Engineering, Seattle University,
the air-core (fully-saturated) slope of the magnetizing Seattle, WA): The authors are to be congratulated for their efforts in
characteristic and the transformer exciting current would be exploring the mechanism of transformer magnetization under GIC. I
many times rated current. This is obviously not true as would like the authors to clarify or comment on the following:
transformer failure would occur very quickly from such high
exciting currents, and such failures have not generally been (a) The paper stated that the mutual inductance MT is nearly in
observed des ite exposure of many power transformers to proportion to the core permeability. When relative permeability
significant G
&
.I' changes from 50,000 to 1, the mutual inductance should also
change 50,000 times. But in the paper, MT changes from 3.62e-3
It is obvious that superposition cannot be applied to the to 4.603e-7 for a change of relative permeability from 50,000 to 1,
extremely nonlinear magnetic circuits of a transformer when which is only 7864 times.
subjected to combined ac and dc excitation. Under such (b) The leakage inductance together with the winding resistance will
combined excitation, the flux bias point is substantially less determine the short circuit characteristics of a transformer. The
than that determined by dc-only analysis. This is because the paper stated that leakage inductance changes from 2.6e-7 MKS
flux offset is biased on1 to the point where the average value, under normal conditions (unsaturated core) to 0.8e-7 MKS under
or dc component, of d e exciting current is identical t o the core saturation. This is an extremely large variation of leakage
applied direct current. Also, the ac component of the flux inductance (3.25 times!). The self and mutual inductances have
l i n k a g e f o r e a c h p h a s e winding must b e t h e 1 .U. very large unsaturated values. The subtraction of two very large
fundamental-frequency value of the appropriate relative p!ase numbers results in large errors. Computing the leakage inductance
angle. These are the boundary conditions for combined ac and directly, if possible, might solve the problem.
dc excitation, and are described at length in Reference [SI for (c) Under nonlinear conditions, inductance has two different defini-
both single-phase and three-phase transformers. When the tions: apparent inductance ( h / i ) and incremental inductance
boundary conditions are met, the core is saturated for only a ( d h / d i ) . While incremental inductance for a grain-oriented steel
small portion of each cycle. core modern power transformer is almost constant, apparent in-
ductance is a function of the flux in the core. What definition of
Figure 2 of Reference [SI shows how the flux bias versus direct the inductance did the authors use in their paper?
current in a single-phase transformer is dependent on the ac (d) It is intuitively not right that the leakage inductance under core
'
voltage level when the boundary conditions are properly saturation (0.8e-7 MKS) is smaller than that of the air core (Me-7
considered. Althou h the bias versus dc characteristic is MKS).
identical to the transformer magnetization curve when the ac (e) The tee-equivalent circuit model of a single-phase transformer
voltage is zero, the bias is substantially reduced when (Figure 14) is only valid for linear-core transformer. When core
simultaneous ac voltage excitation is applied. In practice, GIC saturation is considered, especially when the core is severely
does not offset the flux bias more than 30% - 50% of rated. saturated, no equivalent circuit such as Figure 14 can be derived
for the transformer.
In three-phase core structures, the bias on the various phase
flux linkages will differ due to the asymmetry of the magnetic
circuit when the boundary conditions are considered. This was S. Lu and Y. Liu: We would like to thank Dr. Chen and Mr. Walling
shown in Reference [SI. When only dc excitation is considered, for their discussions and interest in this paper. These discussions will
the center and outer legs of a five-leg transformer have equal help us to better understand the very complex transformer saturation
offset such as shown in Figure 6b of the paper. Combined ac phenomenon under GIC.
and dc excitation can also create maximum saturation in We would like to respond first to Dr. Chen's questions. In the
different core members than is found by considering dc alone. impedance matrix simulation, the definition of apparent inductance
For example, the discusser has noted significant saturation in ( A / i ) is used. We pointed out that the mutual inductance MT is nearly
the yokes of a five-le transformer when combined ac and dc in proportion to the core permeability. This is especially true if the
analysis is performet but the authors show the saturation mutual flux path is mostly within the core material. However when
concentrated in the outer legs when only dc is considered. saturation occurs, the mutual flux path will also include a large portion
of air. The combined flux path will obviously determine the mutual
The following questions are directed to the authors: coupling. Under such circumstance, the mutual inductance is not
linearly proportional to the core permeability.
1. What value is seen by the authors in performing analysis of We agree with Dr. Chen that the subtraction of the very large self
dc excitation alone, as compared to the realistic condition and mutual inductance values could result in errors in the calculated
of a c voltage imposed on the transformer windings leakage inductances. The main purpose of including this example in
simultaneously with GIC flow throu h the transformer? the paper is to illustrate that the changes in leakage inductances are
This question is raised in view of the facts that core offsets much less compared with the changes in mutual inductances during
differ substantially in magnitude and relative phase value, saturation.
1376
The T-equivalent circuit of transformer with constunt parameters is drawn based on analytical studies and predictions. FEM analysis is
not accurate when the core reaches heavy saturation. However, due to able to produce a systematic survey. The results had explained some
lack of any sophisticated model for a saturated transformer at present thought provoking field test results. One example is about GIC suscep-
time, it may become necessary to use the T-equivalent circuit by tibility of the three phase three leg core form transformers. The FEM
introducing uuriublepummeters as a reasonably good approximation. results reveal that even for three phase three leg core form design,
Mr. Walling correctly pointed out that the practical problems are saturation will occur if dc bias is high enough. It is dangerous to
constrained by dc and ac boundaq conditions together. It would be assume they are immune to GIC as have been understood before. The
ideal if the problems can be analyzed with a FEM solver that consid- dc flux distribution is also a valuable parameter which provides infor-
ers both ac and dc simultaneously. Unfortunately, there is no such mation on saturation levels at different parts of transformer under
software available, at least to the extent for practical use. Even if there GIC alone and GIC susceptibility of different core designs.
is one, the computation procedure would be extremely tedious and As we have pointed out in the paper, field measurements [7] tend to
inefficient because step by step solution is required. The simple and suggest that GIC may not be evenly distributed among the three
imperfect dc analysis, on the other hand, can be valuable since a phases. It is certainly possible that the dc offsets could be different in
solution for a particular dc input corresponds to a point on the center and outer phases of a three phase transformer even with
saturation curve. Within this context, the conclusions reached under balanced bias currents.
dc analysis can be applied when dc and ac are considered together. Again, we appreciate the discussers valuable comments.
We also believe there is a need to first totally understand dc
magnetization for different transformer core configurations. Before
this effort, many conclusions about how dc flux will distribute were Manuscript received September 28, 1992.

Anda mungkin juga menyukai