IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO.

3, JULY 2008 1327

Ultrasaturation Phenomenon in Power

Transformers—Myths and Reality
Andrzej Wiszniewski, Waldemar Rebizant, Senior Member, IEEE, Daniel Bejmert, and Ludwig Schiel

Abstract—In this paper, the ultrasaturation phenomenon of

power transformers during their energization is studied. It is
shown that under special conditions, the currents observed after
transformer switching on do not contain enough restraining
information (e.g., second harmonic), which may lead to protection
maloperation. This paper concentrates on a thorough explanation
of the problem and possible causes of ultrasaturation. Theoret-
ical investigations are supported and illustrated with simulation
Fig. 1. Single-phase unloaded transformer equivalent circuit.
studies performed both with MATLAB and Electromagnetic Tran-
sients Program–Alternative Transients Program. The outcomes
of this research can further be used as hints for substation oper-
ation personnel as well as for the development of new protection
stabilization criteria, which is not discussed further in this paper.
Index Terms—Harmonic restraint, transformer protection,
transient analysis, ultrasaturation.

I. INTRODUCTION

S ATURATION of power transformer magnetic cores during

a sudden jump of the terminal voltages is a well-known and
thoroughly studied phenomenon. It occurs either during ener-
gization of the transformer or after clearing of a nearby short
circuit. In both cases, it may generate high and comparatively
slowly decaying inrush currents which may be several times
higher than the rated ones. In case of a single-phase transformer Fig. 2. Typical waveshapes of flux and magnetizing current during inrush after
(Fig. 1), the inrush current waveshape contains a large and de- transformer energization (transformer unloaded).
caying dc component; however, it is always smaller than the ac
component (Fig. 2). The waveshape shows flat regions—which
correspond to the time span when the core is not saturated. As
a result, there is always a substantial second harmonic in the
current which, even in cases of very heavy saturation, when the
residual flux in the core has the same polarity as the dc flux
caused by the voltage jump, does not fall below approximately
15% of the fundamental one. Therefore, the presence of the Fig. 3. Energization of a loaded transformer.
second harmonic became a restraining criterion in differential
relays. If the relay detects the second harmonic, which is higher However, there are situations—although very rare—of inad-
than 15–20% of the fundamental component, the operation of vertent operation of differential relays under inrush conditions
the relay is blocked. Today, this method became standard and is and, as a consequence, tripping of healthy transformers. There-
effective in most cases [1], [2]. fore, one may suspect that the restraining second harmonic was
smaller than the relay setting [3].
A study of this phenomenon has been presented in [3].
Manuscript received January 29, 2007; revised May 15, 2007. Paper no. The authors observed that energization of a loaded transformer
TPWRD-00033-2007.
A. Wiszniewski, W. Rebizant, and D. Bejmert are with the Wroclaw
(Fig. 3) may lead to the situation—called ultrasaturation—when
University of Technology, Wroclaw 50370, Poland (e-mail: andrzej. the dc flux in the core in the initial stage of the process increases
wiszniewski@pwr.wroc.pl; waldemar.rebizant@pwr.wroc.pl; daniel.bejmert@ instead of decaying (Fig. 4). As a result, the distortion of the
pwr.wroc.pl). current waveshape becomes smaller, and the percentage of
L. Schiel is with the Siemens AG, PTD EA, Berlin 13629, Germany (e-mail:
ludwig.schiel@siemens.com). the second harmonic falls below the relay restraining level. In
Digital Object Identifier 10.1109/TPWRD.2007.909211 fact, there may be extreme situations (which we call excessive
0885-8977/$25.00 © 2007 IEEE

.
1328
Fig. 4. Energization with flux increase after the transformer switching on.
Fig. 5. Excessive ultrasaturation after the transformer switching on.
ultrasaturation) when the magnetic flux in the core for several
periods is higher than the saturation level. As a consequence,
the current ac waveshape is almost undistorted and the level of
the second harmonic is negligible (Fig. 5).
This paper presents conditions which must be met to make
ultrasaturation and excessive ultrasaturation possible. They are
derived on the ground of a simplified equivalent circuit but af-
terwards they are confirmed by means of simulation of the non-
linear circuit of the transformer. The results prove that ultrasat-
uration is sometimes possible if there is a case of sudden voltage
jump on the terminals of a transformer which are either loaded
(Fig. 3) or connected to the primary impedance (Fig. 20). It
also shows that excessive ultrasaturation is a very rare but likely
phenomenon.
A simulation example of the ultrasaturation case for a three-
phase unloaded transformer is presented in Fig. 6. One can ob-
serve that the waveshapes of transformer terminal currents con-
sist of the purely sinusoidal component plus a dc component.
Their amplitudes and level of dc components initially rise during
the first five cycles and then begin to decay; however, they have
.
IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 3, JULY 2008
Fig. 6. Transformer energization with ultrasaturation (system scheme and con-
ditions as for Fig. 21): (a) HV terminal currents. (b) Second harmonic ratio.
the same polarity for a comparatively long time. In the case
shown in Fig. 6(b), the level of the Ih2/Ih1 ratio measured is
extremely low, falling down to even less than 5% for the current
in phase L1 (A). This would cause relay inadvertent operation
(overfunction) since the harmonic stabilization is not sufficient
enough to prevent tripping. The 100-Hz restraint ratio is low not
only for phase currents but also in the zero-sequence current,
thus also making this signal useless for transformer protection
stabilization.
The research performed was divided into three parts. The
theoretical approach presented in [3] was analyzed and verified
in Section II. The new theoretical approach to the problem
is presented in Section III. It is followed by both Matlab and
EMTP–ATP simulations (Section IV). Final conclusions are
proposed in Section V.
II. ANALYSIS OF THE CASES REPORTED IN THE LITERATURE
The ultrasaturation phenomenon is characterized by the fact
that after switching on the loaded transformer, the amplitudes of
WISZNIEWSKI et al.: ULTRASATURATION PHENOMENON IN POWER TRANSFORMERS 1329

Fig. 7. Simplified circuit of the transformer for ultrasaturation studies [3].

Fig. 8. Basic circuit of the loaded transformer for ultrasaturation studies.

the magnetic flux increase. In some cases, the fluxes oscillate in

the saturated region during a considerable period of time, which and
is called excessive ultrasaturation. The possibility of such an
event has already been reported in the literature [3]; however,
the explanation that is found is not sufficient.
In this paper [3], the authors suggest that the mechanism of
ultrasaturation phenomenon can be explained when the trans-
(6)
former under study, together with the primary feeding system
where
and load, is represented by the following simplified single-phase
circuit. The main assumption here was that the secondary side
of the transformer be fully resistive and that the magnetizing in- (7)
ductance be constant.
For the simplified transformer model from Fig. 7, the fol- The analysis of the previously cited equations (5) and (6) leads
lowing relationship for the main flux after switching on the to the conclusions that the excessive ultrasaturation is hardly
transformer with the sinusoidal supply with the amplitude can possible in real cases of power transformers with purely resis-
be derived: tive loads. One ought to bear in mind that the load resistance
must not be too small when compared with the source reactance.
(1) Otherwise, it would lead to the power system voltage instability.

where the amplitude and phase of the fundamental component III. NEW APPROACH TO THE ULTRASATURATION PROBLEM
of the flux are given by the formulae
Theoretical analyses of the possible transformer ultrasatura-
(2a) tion have been performed now for the system as shown in Fig. 8.
The most important differences with respect to the study pre-
sented in [3] are that the magnetizing current is negligible
compared with (i.e., , by with almost
(2b) infinite or at least much higher than the impedance ), and that
the secondary side reactance is considered.
the initial value of the component decaying with the time con- The extreme cases of inrush may be expected if the supplying
stant is equal to voltage is

(8)

which causes the current that can be expressed as

(3)
with and being the remanent flux level.
The initial value of the component decaying with time con- (9)
stant amounts to
where ,
, and
.
(4) Therefore, the voltage (Fig. 8) becomes
with .
The time constants of decaying components are defined as

(10)

(5) where and .

.
1330 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 3, JULY 2008

Bearing in mind that ,

, , and and sub-
stituting that into (14) for , one may write the formula,
which determines the maximum value of the dc flux increase

(16)

Therefore, the maximum value of the total flux in the core

becomes

(17)

At the time when the dc flux component is maximum, the total

flux oscillates between the and . The latter is

(18)
X = 1, X = 2, R = 0:1,
Fig. 9. Flux waveshape according to (11), for
or 0.5, R = 0:5 V=!Z
(flux values relative to ). Substituting (16) into (18) yields

(19)
The flux in the transformer core is then given as
The excessive ultrasaturation takes place if the flux ex-
(11a) ceeds the saturation level , that is, when

(20)
with
Substituting (19) into (20), one obtains the condition of ex-
cessive ultrasaturation

(11b) (21)

The flux increase consists of the ac component and There are two special cases which ought to be considered.
the dc component . The former is The first one, which represents the predominantly reactive load
of the transformer, is the following: , ,
(12) and . For these parameters, the condition of excessive
saturation becomes
while the dc flux increase becomes
(22)

The condition (21) in some rare cases may be met; therefore,

the excessive ultrasaturation is possible. The second case rep-
(13)
resents the purely active load of the transformer, that is, when
, , and . Then, the condition takes
or
the form

(14)
Analysis of (13) indicates that the dc flux component in-
creases with time only if
(15)
The formula (15) presents the condition of ultrasaturation.
Only when it is satisfied, does the aperiodic flux component start
to rise; otherwise, it decays, as presented in Fig. 9.
(23)
Since the left-hand side of (22) is very small, in this case, the
excessive ultrasaturation is practically impossible.
IV. SIMULATION STUDIES WITH EMTP–ATP
All the above considerations assumed that the magnetizing in-
ductance is much greater then the source inductance even
if the magnetic core is fully saturated. However, this is seldom
the case, and often the magnetizing inductance of the saturated
WISZNIEWSKI et al.: ULTRASATURATION PHENOMENON IN POWER TRANSFORMERS 1331

Fig. 10. Flux and magnetizing current waveshapes for the condition of exces- Fig. 12. Flux and magnetizing current waveshapes for the condition of exces-
X = 4:3
X = 10

sive ultrasaturation met tangentially ( , ) and X = X = 4:3

X = 10

sive ultrasaturation met tangentially ( , ) and X =

1000
. 5
.

Fig. 13. Maximum and minimum values of the steady-state flux as a function
Fig. 11. Flux and magnetizing current waveshapes for the condition of exces-
X = 4:3
X = 10
X = ofX X = 4:3
X = 10
R = 0:01
R = 1
9 = 0:79
( , , , , ,
sive ultrasaturation met tangentially (
50
.
, ) and
andV=! = 1:09 ).

transformer is comparable with the source inductance. This
affects the inrush phenomenon and conditions of ultrasaturation.
The simulation studies have been performed with the
EMTP–ATP package, first for a single-phase transformer with
the system structure as shown in Fig. 8. The transformer and
system parameters were first chosen in such a way that the
condition of excessive ultrasaturation was met just tangentially,
that is, for ( , , ,
, , ). It has been verified how the
value of saturation reactance may influence the possibility
of excessive ultrasaturation occurrence. In Figs. 10–12, the
obtained waveshapes of flux and magnetizing current are shown
for three chosen values of the transformer saturation reactance
(1000, 50, and 5 ), respectively. From Fig. 10, it can be seen
that for a high saturation reactance value , leading to distinct excessive ultrasaturation (Figs. 15 and 16, pa-
the magnetizing current is very small; however, the excessive
ultrasaturation takes place, confirming the analysis presented
in Section III.
If reactances of the saturated transformer are lower, what
is more realistic, the magnetizing current must not be neglected.
As a result, the magnetizing flux is not in the saturated region
of the curve all of the time. Since the waveshapes of the
magnetizing current show flat regions in every cycle (Fig. 12),
it means that with the saturation reactance of the transformer
decreasing, the probability of excessive ultrasaturation also
decreases.
However, even in cases of very small values of , the
amount of the second harmonic in the magnetizing currents is
very low (see Figs. 13 and 14).
Another set of simulations was performed for the conditions
rameters under the figures). Under such conditions, the second

.
1332 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 3, JULY 2008

Fig. 14. Maximum values of the magnetizing current and second harmonic Fig. 16. Maximum values of the magnetizing current and second harmonic
X X = 4:3
X = 10
R = 0:01
ratio as a function of X (X = 10
, X = 10
, R = 0:01
, R = 1
,
ratio as a function of ( , , ,
R = 1
9 = 0:79
, V=! = 1:09
, and ). 9 = 0:79 , V=! = 0:89 ).

Fig. 15. Maximum and minimum values of the steady-state flux as a function
ofX X = 10
X = 10
R = 0:01
R = 1
9 = 0:79
( , , , , ,
Fig. 17. Approximation error of minimum flux in relation to values obtained
X = 4:3
X = 10
R = 0:01
R = 1

andV=! = 0:89 ).
from simulation (
9 = 0:79 V=! = 1:09
, ).
, , , ,

harmonic ratio becomes negligible when the excessive ultrasat- This yields
uration takes place, which is the case for higher than some
25 (see Fig. 16).
By studying the plots presented in Figs. 13 and 15, one may
observe that a good correlation of analytical formulae and sim-
ulation results may be obtained, while calculating (19) in-
stead of and , one uses corrected values and (27)

(24)
A comparison of the minimum flux obtained from sim-
(25) ulation runs and calculated with the formula (27) is shown in
Figs. 17 and 18 for both cases presented in Figs. 13 and 15,
where respectively (i.e., with being equal to 4.3 and 10.0 ). The
values of corrective coefficient for both cases are given graph-
(26) ically in Fig. 19. One may observe that the flux approximation
quality is very good in the wide range of .
WISZNIEWSKI et al.: ULTRASATURATION PHENOMENON IN POWER TRANSFORMERS 1333

Fig. 20. Alternative system configuration for EMTP–ATP studies.

Fig. 18. Approximation error of minimum flux in relation to values obtained

X = 10
X = 10
R = 0:01
R = 1

from simulation ( , , , ,
9 = 0:79 V=! = 0:89
, ).

Fig. 21. Flux waveshape and transformer input current (primary), for X =
0:5X , X = 0:5X , R = 0:1X , X = 2X , V = V , and 9 =
0:79 .

Fig. 19. Corrective coefficient k according to (26) for two chosen values of the
reactance X .

Fig. 22. Flux waveshape and transformer input current (primary), for X =
0:5X , X = 0:5X , R = 0:1X , X = 2X , V = 1:2V , 9 =
The approximation formulae for fluxes , , and 0:79 .
can be easily derived, analogously as for . Analyses
have indicated that the accuracy of such equations is similar to
(27). Several cases of transformer energization were simulated for
Further simulation studies have been performed for a three- various impedances and , residual fluxes , saturated
phase power transformer ( , , transformer reactances , and system voltages . An ex-
YNd11 connection, , , five-leg core) ample is presented in Fig. 21 and it is a case of excessive ul-
and two structures of the system, namely: trasaturation. One may note that in the time span between 0.05
1) basic structure, as given in Fig. 8 (supplying system, trans- and 0.3 s, the transformer inrush current (dashed curve) has a
former, load); very low second-harmonic component. Therefore, the differen-
2) alternative structure, as presented in Fig. 20 (supplying tial relay operation is not blocked by the second harmonic re-
system, transformer—unloaded or loaded, additional line, straint. The situation is even worse if, for some reason (e.g.,
or load from the supplying side). emergency excitation buildup in the generator during a fault),
The system shown in Fig. 20 is highly unfavorable, since par- the supplying voltage is higher than nominal. In such a case
ticularly in such a configuration, the ultrasaturation may appear, (Fig. 22, increased by 20%), the flux induced after trans-
during voltage jumps caused by disconnection of a nearby short former energization is much higher and the resulting inrush cur-
circuit. rent contains almost no second harmonic at all.
1334
V. CONCLUSION
Ultrasaturation poses a great problem for protective relaying
of power transformers. The work presented here concentrates
on the phenomenon itself, trying to explain when excessive flux
without zero crossing may appear. Knowing why and when the
ultrasaturation may occur, one may better understand the pos-
sible cases of the differential protection maloperation. The out-
comes of this paper may be also treated as hints for substation
personnel, defining the system configurations, and parameters
that should be avoided in order to be on the safe side should un-
favorable energization occur.
It is clear that the explanation of the ultrasaturation and ex-
cessive ultrasaturation phenomena is only the first step toward
developing new ideas and criteria for more reliable transformer
protection that would better handle such unusual cases than cur-
rently employed relaying equipment. It seems that improvement
of the protection operation would be possible with the intro-
duction of new, modified, or extended criteria. The protection
Member of the Polish Institution of Electrical Engineers.
criteria should, on one hand, carry enough information on the
event to be distinguished and, on the other hand, ensure appro-
priate stabilization for other events for which protection opera-
tion is undesirable. Certain proposals of such new criteria can
be found in the literature, such as a complex second harmonic
restraint [5], flux restraint (estimated on basis of voltages) [6],
or current waveshape analysis [7]; others still wait for their in-
ventors. It is believed that a lot of improvement can be reached
with the introduction of adaptivity in the differential protection,
a simple example of which is to use adaptive thresholds as well
as adaptive measurement procedures. It has also been proved
that considerable improvement of the operation and quite simple
achievement of adaptive features of protection functions may
be obtained with the use of various artificial-intelligence tech-
niques. An example of using a fuzzy-logic technique for the re-
alization of fuzzy differential protection is given in [8]. There
is hope that an appropriate combination of classical and intelli-
gent techniques should bring additional benefits; therefore, fur-
ther investigations on the subject are desirable.
REFERENCES
[1] A. Wiszniewski, H. Ungrad, and W. Winkler, Protection Techniques in
Electrical Energy Systems. New York: Marcel Dekker, 1995.
[2] “Numerical Differential Protection Relay for Transformers, Genera-
tors, Motors and Mini Busbars,” SIEMENS AG, 2006, 7UT613/63x
V.4.06 Instruction Manual, Order . C53000-G1176-C160-2.
[3] X. N. Lin and P. Liu, “The ultra-saturation phenomenon of loaded
transformer energization and its impacts on differential protection,”
IEEE Trans. Power Del., vol. 20, no. 2, pt. 2, pp. 1265–1272, Apr. 2005.
[4] “EMTP-ATP Manuals,” EEUG, 2001.
[5] B. Kasztenny and A. Kulidjian, “An improved transformer inrush re-
straint algorithm increases security while maintaining fault response
performance,” presented at the 53rd Annu. Conf. Protective Relay En-
gineers, College Station, TX, Apr. 11–13, 2000.
IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 3, JULY 2008
[6] Y. C. Kang et al., “Transformer protection relay based on the increment
of the flux linkages,” presented at the CIGRE Study Committee B5
Colloq., Calgary, AB, Canada, Sep. 2005, paper 108.
[7] B. Kasztenny, E. Rosolowski, M. M. Saha, and B. Hillstrom, “A com-
parative analysis of protection principles for multi-criteria power trans-
former relaying,” in Proc. 12th Power Systems Computation Conf.,
Dresden, Germany, Aug. 19–23, 1996, pp. 107–113.
[8] B. Kasztenny, E. Rosolowski, M. M. Saha, and B. Hillstrom, “A self-
organizing fuzzy logic based protective relay—An application to power
transformer protection,” IEEE Trans. Power Del., vol. 12, no. 3, pp.
1119–1127, Jul. 1997.
Andrzej Wiszniewski received the Ph.D. and D.Sc.
degrees and the professorship in Electrical Engi-
neering from the Wroclaw University of Technology,
Wroclaw, Poland, in 1961, 1967, and 1972, respec-
tively. He has worked in the field of power apparatus
and systems as a specialist in the protection and
control of power systems. He is an author of nine
books and more than 130 scientific publications. He
was Rector of the university for two terms (from
1990 to 1996). He was Minister for Science for four
years, beginning in 1997.
Prof. Wiszniewski is a Distinguished Member of CIGRE and Honorary
Waldemar Rebizant (M’00–SM’05) was born in
Wroclaw, Poland, in 1966. He received the M.Sc.,
Ph.D., and D.Sc. degrees from the Wroclaw Univer-
sity of Technology (WUT), Wroclaw, in 1991, 1995,
and 2004, respectively.
Since 1991, he has been a faculty member with the
Electrical Engineering Faculty at the WUT and is cur-
rently Assistant Professor and Vice Dean for Faculty
Development and International Cooperation. His re-
search interests are digital signal processing and AI
techniques for power system protection purposes.
Daniel Bejmert was born in 1979 in Walbrzych,
Poland. He graduated from the Electrical En-
gineering Faculty of Wroclaw University of
Technology, Wroclaw, Poland, in 2004, where he is
currently pursuing the Ph.D. degree at the Institute
of Electrical Power Engineering.
His research interests include application of in-
telligent algorithms in digital protection and control
systems.
Ludwig Schiel was born in Weimar, Germany,
in 1957. He received the Dipl.-Ing. and Dr.-Ing.
degrees from the Institute of Technology Zittau,
Zittau, Germany, in 1984 and 1991, respectively.
In the 1991, he joined the Department of Power
Transmission and Distribution, Energy Automation,
Siemens AG, Berlin, Germany. He is Project Man-
ager of transformer differential protection systems.