526 IEEE Transactions on Power Delivery, Vol. 10, No.

1, January 1995

SELF-GENERATED OVERVOLTAGES DUE TO OPEN-PHASING

OF UNGROUNDED-WYE DELTA TRANSFORMER BANKS
R. A. Walling, Member, IEEE R. K. Hartana, Member, IEEE W. J. Ros, Member, IEEE
GE Industrial and Power Systems
Power Systems Engineering Department
Schenectady, New York
Abstract -
Disconnecting one phase of an Surge arresters connected to the transformer side of the
ungrounded-wye delta transformer bank from the open primary cutout or fuse are exposed to these
source can result in severe overvoltages due to overvoltages. Particularly in the case of neutral-shift
neutral shift and ferroresonance. Neutral shift overvoltage, the arresters can fail rapidly [l]. Arresters
overvoltages are quantified and their impact on
metal-oxide surge arresters are evaluated. connected to the source side of the open point are not
Ferroresonance I s shown by test data to occur when subjected to the overvoltage but the transformer windings
low-loss banks, 15 kV and higher, are open phased. are left unprotected. Neutral shift, and particularly
Design and operating practices to eliminate or ferroresonance, can result in voltages applied to the
mitigate these self-generated overvoltages are transformer winding which are more severe than any of the
discussed. applied and induced voltage tests specified by ANSI
standards [2,3].
INTRODUCTION Self-generated overvoltages due to open-phased
ungrounded-wye delta transformer banks have been
The ungrounded-wye delta transformer connection is previously recognized by the industry [1,4,5,6]. The issue
commonly used in Nortb America to provide 240 Volt delta has received increased recent attention because of the
secondary service to small commercial and light industrial vulnerability of metal-oxide arresters to temporary
loads. With this connection, used primarily on overhead overvoltages, as well as the increased tendency of today's
transformer banks, 120/240 Volt service can be provided on higher efficiency distribution transformers to engage in
one phase for lighting and convenience-outlet loads, while ferroresonance. Previous studies of ferroresonance in these
also serving three-phase loads (four-wire delta service). banks found this phenomenon could only be sustained by
The primary neutral is not grounded to avoid zero-sequence the transformer's internal capacitance at the 25 and 35 kV
current flow into the bank due to primary feeder imbalance class [6]. Test results provided in this paper, however,
or faults. show that modern 15 kV-class banks are vulnerable to this
When a primary phase is disconnected from the source, phenomenon, and severe overvoltages can result.
the transformer bank can generate severe overvoltages An investigation has been performed to characterize
which may persist for as long as the open-phase condition overvoltages resulting from open-phasing of ungrounded-
is present. These overvoltages are the result of two wye delta transformer banks, and the results are
distinctly different phenomena: documented in this paper. Arrester duties resulting from
1. Shift of the neutral voltage due to load imbalance or neutral-shift and ferroresonant overvoltages are evaluated
2 Ferroresonance due to the interaction of the and compared. Design and operating practices are
transformers' core saturation characteristics with the suggested which can allow utilities to avoid or minimize
transformer windings' inherent capacitance. these overvoltages.
Neutral shift overvoltage occurs when the single-phase NEUTRAL-SHIFT OVERVOLTAGE
load is large compared to the three-phase load.
Ferroresonance, however, only occurs when the transformer
bank is unloaded and the Capacitance is sufficient to Phasor Analysis
sustain the interaction with the core. Figure 1 shows circuit and phasor diagrams for an
ungrounded-wye delta transformer bank with primary phase
C open and a load on one secondary, between phases a and
b. No other secondaries are loaded and all transformers are
assumed ideal in this illustration. Because the load
provides a finite impedance between A and N', the bank
neutral (N') voltage with respect to the system neutral
This paper was presented at the 1994 IEEE PES Transmission and (VN+N)shifts to become equal to VAN. The secondary delta
Distribution Conference and Exposition held in Chicago, Illinois, provides the constraint:
April l(L15, 1994. vBb+v b c t Vca= 0 (1)

0885-8977/95/$04.00 0 1994 IEGE

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 527

PRIMARY SECONDARY the bank is supplying loads on all phases, provided the
dominant load is on one of the two source-connected
C phases. For the general case, circuit analysis reveals the
c C’ open-phase voltage to ground (assuming the primary neutral

r- is at ground potential) is given by:

Y N
N = system neutral
1, where VA and V, are the phasor voltages of the two source
phases remaining connected to the bank and Yl and Y2 are
the complex admittance of the loads on these phases,
respectively. Phase C is assumed to be the phase which is
disconnected from the source and Y, is the admittance of
the load on this phase.
1 In the typical application, this type of transformer bank
serves a three-phase load which is divided equally between the
phases, plus an additional load on one phase only. Defining K
as the ratio of the single-phase load to the balanced three-phase
load, and assuming these loads are passive and have identical
power factor, (3) can be simplified to:
jl.73
.lnr2+K+1
IvopenI= K+2 (4)
Saturation Effects
It is widely assumed that transformer saturation limits
I the neutral-shift overvoltage to a magnitude much less
severe than the ideal phasor calculations. EMTP
simulations were performed by modeling a bank of three 25
kVA distribution transformers with typical leakage
impedance and saturation characteristics (1.5 Tesla design
flux density).
Figure 2 shows that transformer saturation results in
severe distortion of the open-phase voltage to ground, but
does not significantly reduce the crest voltage. The
simulation shown here represents 2.5 kW, or 10% load, on
one phase of the bank. Figure 3 compares the ideal open-
phase voltage, calculated by (4), with simulations for a
range of K. The single-phase component of the load was
held constant at 1 kW and 10 kW, and the three-phase load
was varied in the simulations. With a small load, the crest
neutral-shift voltage is reduced due to transformer core
losses and saturation. These effects are minimal with a
B large load and the crest voltage closely approaches the
ideal value, which is a function only of K.
Resistance Grounding
Fig. 1. Circuit and phasor diagrams for an ungrounded- Grounding wye-delta banks through a resistance is
wye delta transformer bank with a single-phase recommended by [41 as a means to limit neutral-shift
load and one primary phase open. overvoltages while also protecting the transformers from
Because the transformers are assumed ideal and V A N is excessive zero-sequence currents. For the most severe
condition of only one phase loaded, the open-phase to
zero: ground voltage can be limited to the normal phase-ground
VC’N’ = -vBN’ magnitude by grounding the primary neutral through an
(2)
The neutral shift causes VBN*to be equal to V A B and impedance 50% of the load impedance reflected to the
the transformer open-phase voltage with respect to the transformer primary. This neutral resistor, however, must
primary neutral, VC~N is elevated to 2.65 times normal. be able to dissipate power equal to 22% of the connected
load wattage in the event one secondary phase is loaded to
This neutral shift overvoltage occurs when the single-
rating and a primary phase is opened. The required power
phase load is on either of the remaining phases connected rating renders resistance grounding impractical in most
to the source. Less severe overvoltages can also occur if cases.

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 528

and become thermally unstable. The available fault current

[??.U,. 1 on the open phase is limited by the load impedance. Thus,
4 iVopen arrester failure may not be evident until the phase is
reclosed into the stiff source. Violent failure may result at
2
this time [l].
Manufacturer-supplied arrester TOV curves are based
on a stiff 60 Hz voltage source. In the neutral-shift
situation, however, the reflected load impedance and the
0
distortion of the voltage by transformer saturation reduce
the arrester duty. EMTP simulations, using a detailed
model of a 7200/12470Y 3x25 kVA transformer bank and
-2 the V-I characteristics of a typical 10 kV distribution
arrester, were used to equate arrester duty from neutral shift
to the duty resulting from a 60 Hz voltage source applied to
the same arrester model. Figure 4 plots the equivalent
undistorted TOV versus the resistive load applied to one
: phase of the bank. Using the example arrester TOV curves
0 20 40 60 published in [41, it can be seen that a 10 kV arrester may
survive the neutral shift, possibly long enough to complete
TIME IN MILLISECONDS switching of all three phases. At the 0.5 kW load level, 50
minutes can be withstood and one minute at 5 kW. With a
Figure 2. Open-phase to ground voltage with 2.5 kW load. similar 12 kV arrester, simulations show three minutes can
be withstood with a 5 kW load. Several manufacturers are
MOV Arrester Duty From Neutral Shift now offering gapless MOV arresters with even more
permissive TOV curves than used for this example.
Surge arresters connected to the open transformer bank The authors do not suggest that intentional switching of
phase, on the transformer side of the open fuse or cutout, ungrounded-wye delta banks with unbalanced loads is a
are exposed to the self-generated overvoltages of desirable practice. The simulation results, however,
ungrounded-wye delta banks. These overvoltages were of indicate gapless MOV arresters have greater tolerance of
less significance in the past, prior to the introduction of neutral shift than is implied by the voltage magnitude
MOV distribution arresters. The 60 Hz sparkover of present without the arrester.
conventional gapped silicon-carbide surge arresters is Gapped MOV arresters provide somewhat greater
approximately 2.34 p.u. of crest operating voltage [7]. This immunity from neutral shift voltages. To avoid gap sparkover
is sufficiently high to protect these arresters from damage due to neutral shift, an arrester rating 33% above the usual
due to neutral-shift overvoltages in most instances. arrester duty cycle voltage rating is necessary [8]. The gapped
Because the MOV blocks are continuously exposed to arresters have a lower protective level, relative to the arrester
the applied voltage in gapless arrester designs, neutral-shift rating, which offsets the impacts of overrating for current
overvoltages may cause gapless MOV arresters to overheat impulse conditions such as lightning. The gapped design
limits the voltage applied to the transformer only when the
gap sparks over, however, and msformers can be subjected to
VoDen I ~ . u . l
overvoltage stress which is not adequately addressed by present
transformer standards [2,3].
Vequiv Ip.u.1
1.58 I

1.56 -

1.54 -

1.52 -

1.5 -

1 10
K (- PI/P3)
100
1.48
a
' I

2 4
I I

6
I

8 10
Fig. 3. Comparison of crest open-phase voltage from kW
simulations modeling saturation to ideal Fig. 4. Equivalent 10 kV gapless MOV arrester TOV as a
calculations, as a function of load imbalance. function of single-phase load.

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 529

The transformer insulation breakdown mechanism for a Field Tests

long-duration overvoltage is decidedly different than that for a
single impulse; thus BIL is of questionable relevance. Applied Field tests were performed to measure ferroresonant
voltage test levels, specified by standards, are more appropriate overvoltages on two ungrounded-wye delta transformer banks.
for the overvoltages discussed here. Neutral shift can result in The first bank was rated 13.8/23.9Y kV and the second was
phase to ground voltage exceeding the applied voltage test 7.96/13.8Y kV. Both banks consisted of 3-25 kVA poletop
levels, specified by [2], for 14400/24940Y (125 BIL) and transformers. These transformers were all constructed on
19920/34500Y (150 BIL) banks. For 15 kV-class banks, the highly-efficient silicon-steel cores, and total three-phase core
maximum neutral shift voltage is less than the applied voltage losses were 164 W on the 23.9 kV bank and 154 W on
test level for 95 kV BIL units, but it should be considered that 13.8 kV bank.
the test standards require only that the transformer survive one Self-ferroresonancewas observed during switching of both
minute of overvoltage at this level [3]. banks, with crest voltages far exceeding that which can occur
from the neutral-shift phenomenon previously discussed.
FERRORESONANCE Figure 5 shows a typical voltage waveform recorded on the
open phase after the second phase of the 23.9 kV bank was
Ferroresonance can occur when a three-phase transformer energized. This chaotic waveform, with crests reaching 4 pu,
or bank is open-phased with a phase-ground capacitance left continued until the last phase was closed.
connected to the transformer. The phenomenon is usually Figure 6 shows the envelope of the open-phase voltage in
associated with underground distribution systems where cables one test of the 13.8 kV bank. Voltage crests exceeded 4 pu for
provide substantial capacitance. In overhead systems, prior to approximately 8 s, then dropped to 2.5 pu. Ungrounded-wye
the widespread adoption of core-loss economic evaluation by delta banks are widely applied to 15 kV-class distribution, and
the industry, ferroresonance was typically observed only where this test shows such banks as vulnerable to self-ferroresonance.
substantial lengths of overhead lateral on the transformer side In both tests, ferroresonant overvoltage magnitudes are much
of the switching point provide sufficient capacitance to support more severe than the 2.65 pu maximum which can occur from
ferroresonance [4,5,6]. neutral-shift in a bank with unbalanced load.
Transformers, however, also have significant intemal As a matter of interest, additional tests were performed on
capacitance which can support "self-ferroresonance" without the 23.9 kV bank to evaluate high-resistance neutral grounding
external capacitance. The largest component of intemal phase- as a ferroresonance mitigant. Ferroresonance was eliminated
ground capacitance responsible for self-ferroresonance in by a 55 kR resistance between the bank neutral and ground. A
ungrounded-primary banks is between the primary and resistance of this magnitude is more than adequate to protect
secondary windings (which are essentially at ground potential), the bank from zero-sequence currents due to primary
followed by the capacitances from the primary winding to the imbalance. The neutral resistor must, however, be capable of
tank and core. The equivalent capacitance from phase to withstanding phase voltage if there is any possibility of
ground of a typical 25 kVA 95 kV BIL transformer is on the neutral shift.
order of 2 nF.
With the less efficient distribution transformers used in Arrester Duty Due to Ferroresonance
past decades, self-ferroresonancewas limited to 25 and 35 kV-
class banks [61. There have been numerous recent field Although ferroresonance can create high overvoltage, the
reports, however, of self-ferroresonance in 15 kV-class circuit is an extremely weak source. Reference [lo] shows, in
ungrounded-wye delta banks. A recent investigation has
shown that the critical capacitance needed to support
ferroresonance in three-phase grounded-wye padmount
transformers is directly related to core loss [9]. It is reasonable
to postulate that the decrease in prevailing poletop transformer
core losses is responsible for the self-ferroresonance now
occurring at the lower distribution voltages.

-5 1 1 I I I I I
0 Time (ms) 1000
0 18
Time (seconds)
Fig. 6. Envelope of open-phase voltage maxima and minima
Fig. 5. Open-phase voltage due to self-ferroresonanceof the during self-ferroresonance of the 13.8 kV ungrounded-
23.9 kV ungrounded-wye delta poletop bank. wye delta poletop bank.

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 530

tests of gapless MOV arrester duty due to padmount poletop installation and would tend to be expensive if they
transformer ferroresonance, that the source weakness limits were available due to the voltage and power dissipation
arrester currents to a small value. These tests showed arresters requirements.
can withstand ferroresonance involving padmount transformers An alternative is to install gapless MOV arresters on
and substantial cable lengths for a significant period of time, transformer side of the fuse. If fuse operation results in
and liquid-immersed arresters can survive indefinite exposure. ferroresonance, these arresters provide the transformer with the
The capacitance involved in poletop bank self- necessary overvoltage protection. If unbalanced loading causes
ferroresonance is much smaller than in the underground severe neutral shift, the arresters may fail to short-circuited
situations considered in [IO]. As a result, arrester discharge condition but the transformer is protected from overvoltage.
currents are even smaller and gapless arresters can clamp Because the arrester could fail eventfully if reconnected to the
ferroresonant voltages to a tolerable value without excessive source, a reasonable practice is to always replace the arrester on
heating. Figure 7 shows open-phase voltage and 10 kV the open phase before reclosing. The much greater cost of a
arrester discharge current for a test similar to that shown in transformer relative to an arrester justifies considering the
Figure 6. The voltage is clamped to 1.5 pu by the arrester arrester as sacrifcial for infrequent open-phase incidents.
discharging 10 mA crest. The arrester dissipates only about Reference 141 recommends installing arresters on the
10 W in this case, and very minimal arrester heating results. source side of the cutout to avoid arrester failure due to neutral-
Tests on gapped MOV arresters show that they spark over shift overvoltages. The authors do not consider this practice
for ferroresonant waveforms at greater than 2.25 times the crest desirable as it leaves the transformer unprotected from
MCOV. If a 12 kV gapped arrester is applied to a 7.2112.47Y potentially destructive ferroresonant overvoltages. Likewise,
bank, as suggested by [8], ferroresonant voltages are only the use of overrated gapped arresters is considered by the
limited to 3.2 pu at best. Although below the applied-voltage authors to be an incomplete solution. At the higher
test level for a 95 kV BIL, analysis is required to insure that distribution voltages, the transformer is not protected from
the transformer insulation can withstand this degree of duties exceeding the applied-voltage test levels. For 15 kV-
overvoltage for a potentially long duration. class units, although the applied test level is not exceeded
when protected by a gapped MOV arrester, there remains some
RECOMMENDED PRACTICES question of the transformer insulation’s ability to withstand
voltages up to the gap sparkover level for durations far
Although the ungrounded-wye delta bank continues to be exceeding the one minute duration of the applied voltage test.
applied, its potential exposure to ferroresonant overvoltage,
coupled with concern over neutral shift and surge protection CONCLUSIONS
considerations suggest that utilities reexamine their application
criteria. Many utilities choose to not use the ungrounded-wye An open primary phase on an ungrounded-wye delta
delta connection [13. Where an ungrounded secondary service transformer bank can result in severe overvoltage between the
is necessary, the grounded-wye floating-wye connection is an open phase and ground. These overvoltages can be caused
alternative. Where four-wire delta service must be provided, either by unbalanced secondary loading, leading to neutral
the open-wye open-delta configuration is an alternative which shift, or by ferroresonance in banks that are virtually unloaded.
is free of self-ferroresonanceand neutral shift overvoltage. Neutral-shift overvoltages can lead to rapid failure of
When ungrounded-wye delta are used, different practices MOV surge arresters applied at the normal voltage rating. The
apply for controlling overvoltages during intentional switching arrester duty is somewhat reduced by the reflected load
and for fuse operations. Field switching is the most easily impedance and by transformer saturation.
solved situation. By far, the best solution is to temporarily Self-ferroresonance, without any external capacitance
ground the bank neutral prior to switching. This alternative connected to the transformer, can occur when ungrounded-wye
eliminates neutral shift and self-ferroresonance, and involves delta transformer banks rated 15 kV class and higher are
little risk. A primary fault during the short interval required is switched. Ferroresonant overvoltage magnitudes can exceed
highly improbable. Also, any overload of the transformer due 4 pu and protection of the transformer winding from these
to normal feeder voltage imbalance (zero sequence) is tolerable
for this duration.
Other alternatives which only control neutral shift
overvoltage include energizing the transformer serving the 1
single-phase load last and deenergizing it first, and p.u.
0
disconnecting the secondary load prior to switching. These
measures, recommended in [4] do not avoid self-ferroresonance. -1
Removing the load increases this risk because ferroresonance
will not occur in a transformer loaded to more than a few 20
percent of rating. These measures need to be augmented by an mA
additional measure to control the ferroresonance, such as 0
gapless MOV arresters connected to the transformer primary.
Avoiding self-generated overvoltages in ungrounded-wye -20
delta banks following a primary fuse operation is more 0 Time (ms) 20
difficult. The neutral could be grounded through an impedance
small enough to avoid neutral shift. Such resistors or Fig. 7. Arrester voltage and discharge current during self-
inductors are not commonly available in a form suitable for ferroresonance of the 13.8 kV bank.

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 53 1

voltages is recommended. Gapless MOV arrester duty from BIOGRAPHIES

poletop bank self-ferroresonance has been shown to be
Reigh A. Walling (M 78)was born i n New Jersey in 1952. He
minimal, due to the extreme source impedance provided by the received Bachelor’s and Master’s degrees in Electric Power Engineering
femmonant circuit. from Rensselaer Polytechnic Institute in 1974 and 1979, respectively.
Design and operating practices are recommended by this He joined the Power Systems Engineering Department of G E Industrial
paper which can eliminate or minimize overvoltages when and Power Systems in 1981, where he is presently a consulting application
engineer involved in a wide range of transmission and distribution areas.
these banks are open-phased. The preferred solution is to
Mr. Walling is a registered Professional Engineer and is a member of
temporarily ground the bank neutral during switching several IEEE and CIGRE working groups.
operations. Gapless arresters should be installed on the Rutisurhata K. Hartana (S’82-MS9) received the Electrical Engineer
transformer side of the cutout to protect against excessive degree from Trisakti University, Jakarta, Indonesia in 1982 and the MSEE
overvoltages in the event the transformer is open-phased by a and PhD degrees in electrical engineering from Louisiana State University in
1983 and 1989, respectively.
fuse operation. If a severe neutral shift accompanies the fuse
He is currently a senior application engineer i n the Power Systems
operation, however, the arrester may be sacrificial. Engineering Department of G E Industrial and Power Systems, involved in
the simulation and testing of distribution systems and a components.
ACKNOWLEDGMENT Walter J. Ros (M 70) is a graduate of the University of Massachusetts
with a BSEE. After working at the Hartford Electric Light Company as a
The support and assistance of Ron Reckard, Mahesh distribution engineer, he joined the General Electric Company 1969 as an
Sampat, and Ram Sinha of GE Distribution Transformers is application engineer, and has been involved in a wide variety of assignments
gratefully acknowledged. in GE’s Power Systems Business and Power Systems Engineering
Department.
REFERENCES He is currently a consulting engineer in the Power Systems Engineering
Department and is a member of the IEEE Distribution Subcommittee. He is
J. Gasal, “Prevent Overvoltage Failure of Arresters,’’ Electrical World, also the Chairman of the IEEE/PES Meetings Department and Vice
July, 1986. Chairman of ANSI C84.1.
“IEEE Standard General Requirements for Liquid-Immersed
Distribution, Power, and Regulating Transformers.” American National
Standard CS7.1 200-1987.
“IEEE Standard Test Code for Liquid-Immersed Distribution, Power,
and Regulating Transformers,’’ American National Standard CS7.12.90-
1987.
“Guide for the Application of Metal-Oxide Surge Arresters for
Alternating Current Systems,” American National Standard C62.22-
1991.
L.B. Crann. R.B. Flickinger. “Overvoltages on 14.4/24.9 kV Rural
Distribution Systems,” AIEE Transactions, Oct. 1954. pp. 1208-1212.
R.H. Hopkinson. ”Ferroresonance During Single-phase Switching of
Three-phase Distribution Transformer Banks.” IEEE Trans. on Power
Apparatus and Systems, Vol. 84, April 1965, pp. 289-293.
E.W. Boehne. R.E. Koch, G.L. Gaibros. H.W. Mikulecky, “Coordination
of Lightning Arresters and Current-Limiting Fuses,” IEEE Trans. on
Power Apparatus and Systems, Vol. 91, May/June 1972. pp. 1075-1078.
F. DeCesaro. D. Tupper, K. Leix. J . Tardugno, R. Reichert, “Arrester
Application Considerations on Floating Wye-Delta Transformer Banks,”
National Rural Electric Cooperative Association Tech Day, Dallas,
Feb. 14, 1993.
R.A. Walling, K.D. Barker, T.M. Compton. L.E. Zimmerman,
“Ferroresonant Overvoltages i n Grounded Wye-Wye Padmount
Transformers with Low-Loss Silicon-Steel Cores,” IEEE Transactions on
Power Delivery. Vol. 8, July 1993. pp. 1647-1660.
R.A. Walling, R.K. Hartana, R.M. Reckard. M.P. Sampat. T.R Balgie,
“Performance of Metal-Oxide Arresters Exposed to Ferroresonance in
Padmount Transformers,” IEEE Power Engineering Society 1993
Summer Meeting. Vancouver, Paper 93 SM388-9 PWRD.

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 532
Discussion [2] “IEEE Standard General Requirements for Liquid-Immersed
Distribution, Power, and Regulating Transformers,”
F. P. DECESARO, M. B. MARZ (Cooper Power Systems, American National Standard C57.12.00-;987.
Franksville, WI), C. W. PLUMMER (Cooper Power Systems,
Pewaukee, WI): This paper provides the industry with an excellent Manuscript received May 6, 1994.
description of self-generated overvoltages due to open-phasing of
ungrounded-wye delta transformer banks. The circuit and phasor
diagrams in Figure 1 of the paper concisely explain the neutral-
shift overvoltage phenomena under these conditions.

The authors’ comments on the effect of transformer saturation on R A. Walling, R K. Hartana, W. J. Ros (GE Industrial
neutral shift overvoltages are also appreciated. We have -
and Power Systems) We thank the discussers for the
previously modeled the effect of transformer saturation on open- opportunity to provide further clarification of our opinions
phased ungrounded-wye delta transformer banks using a Transient regarding overvoltage protection of ungrounded-wye delta
Network Analyzer. Our results confirm the authors’ statement that transformer banks.
transformer saturation causes severe distortion of the open phase
line-to-ground voltage, but does not significantly reduce the crest The discussers’ conclusions regarding the insensitivity of crest
voltage. line-to-ground voltage to transformer saturation are
particularly appreciated. The misconception that transformer
The authors mention test data which suggests that gapped MOV saturation generally limits arrester power-frequency TOV
surge arresters spark over at 2.25 times the peak of their MCOV duty is widespread in the utility industry. Saturation occurs
rating for ferroresonant waveforms (3.2 pu for a 12 kV gapped near the flux crest, and flux is ninety degrees out-of-phase
MOV distribution class surge arrester on a 12.47/7.2 kV base). with the voltage. As stated in the paper and supported by the
The tests performed to justify this statement are neither described discussers, saturation of wyeconnected transformers makes
or referenced in the paper. This high spark over level contradicts little change to the crest voltage magnitude applied to
our experience with existing resistively graded, gapped MOV
arresters. The arresters, of course, are sensitive only to the
distribution class surge arresters operating at ferroresonant
frequencies. Gapped 12 kV MOV distribution class surge arresters
crest voltage.
spark over at approximately 2 1 kV peak (2.1 pu on a 12.47/7.2 kV The discussers indicate that the low-frequency sparkover
base) for frequencies in the ferroresonant range. Once spark over voltage of a 12 kV duty-cycle rating resistively-graded gapped
occurs, the gap is shorted and the surge arrester provides protection MOV arrester is 21 kV crest. Published specifications for a
approximately equal to that of a gapless 10 kV MOV distribution popular make of resistively-graded gapped MOV distribution
class surge arrester. Can the authors provide more details on the
arrester, however, show a 21 kV rms guaranteed minimum
tests that determined the spark over level of 12 kV gapped MOV
sparkover voltage for a 12 kV arrester. The particular gapped
distribution class surge arresters to be 3.2 pu? Of particular
interest are the frequencies of the test waveforms, the ratings of the MOV arrester described here is of a design which should be
arresters tested, the types of gaps tested, and the number of tests very familiar to the discussers. The crest voltage of a
performed. sinusoidal 21 kV rms voltage is 29.7 kV, which is 2.06 times
the crest MCOV of this particular arrester. When applied to
Based on our testing and field experience[l], the spark over levels a 12.47Y/7.2 kV transformer bank, this is 3.0 p.u. of the crest
of gapped MOV distribution class surge arresters will protect nominal voltage.
transformers during both ferroresonant and severe neutral-shift
conditions without necessitating arrester replacement prior to re- Considering that the specification is for the minimum
energization. The 2.1 pu spark over level of 12 kV gapped MOV sparkover voltage, the 2.25 times crest MCOV (or 3.26 times
arresters is well below the 34 kV rms (4.7 pu on a 12.4717.2 kV nominal crest) sparkover results we obtained are quite
base) low frequency dielectric insulation requirement for 95 kV consistent. There are both manufacturing tolerances and
BIL transformers[2]. Even the 3.2 pu spark over level discussed in atmospheric differences which affect gap sparkover, and an
the authors’ paper is below this insulation requirement. average measured sparkover 10% greater than the guaranteed
Nevertheless, the paper’s authors recommend using gapless MOV minimum is reasonable. The discussers request information
arresters to protect transformers during severe neutral shift and on the number and types of arresters tested to reach our
ferroresonant overvoltage conditions even though the arresters are conclusions regarding resistively-graded gapped MOV
considered sacrificial and should be replaced following possible arrester sparkover voltages during ferroresonance. These
neutral-shift conditions. Would the authors comment on why the
details could be provided. However, it is not necessary to
use of sacrificial surge arresters would be advocated when an
effective, non-sacrificial alternative is already being successfully include this information in our closure as our results are
applied [l]? completely consistent with 60 Hz sparkover results published
by the arrester manufacturer.
REFERENCES The authors agree that a gapped 12 kV MOV arrester will
[l] F. DeCesaro, D. Tuper, K. Leix, J. Tardugno, R. Reichert. provide a discharge voltage comparable to a 10 kV gapless
“Arrester Application Considerations on Floating Wye-Delta arrester, once the gap has sparked over. For voltage levels up
Transformer Banks.” National Rural Electric Cooperative to the sparkover voltage, however, the arrester does not
Association Tech Day, Dallas, Feb. 14, 1993. mod@ the voltage applied to the transformer. For the

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.
 533

12.47Y/7.2 kV example with a 12 kV gapped arrester, the 21 a 24.94Yl14.4 kV ungrounded-wyc dclta bank to avoid
kV minimum sparkover is below the 34 kV low-frequency arrestcr failurc from ncutral shift ovcrvoltagc. Thc miniriiuni
insulation test magnifude for a 95 kV BIL transformer. As sparkover of a typical resistively-graded gappcd MOV
stated in the paper, ANSI (37.12.00-1987 only requires that arrestcr is 39 kV rms. With only a 10% variation in gap
thc transformcr insulation withstand onc minute of exposure performance, the sparkovcr can easily be 43 kV, which is in
to the applicd test voltagc. Ovcrvoltagcs during bank cutout csccss of tlic 40 kV low-frcqucncy voltagc withstand for 125
switching should normally bc lcss than a minutc in duration, kV BIL transformers. Ncutral shift will not causc voltagcs
provided thcre are no extcnded delays in complcting thc this high, but fcrroresonancc can. Espericncc has shown that
switching operation after the first phase is opened or the fcrrorcsonancc is quitc likely when an unloaded 25 kV or 35
second phasc closed. Overvoltages due to fuse operation, kV-class ungrounded-wyc dclta bank is open phased.
however, could easily persist for hours or cvcn days beforc
Thc choice bctwccn thc protection practice suggested in the
they arc discovered and corrected. There is no guarantcc that
a transformcr can withstand an overvoltage of nearly 3 p.u. papcr and that rccommcndcd by the discusscrs is
philosophical. The question for tlic distribution cnginccr is:
for a very long period of time, even if the ovcrvoltagc is less
than thc onc-minutc applied test magnitudc. should tlic arrester application be made to prolcct the arrcstcr
cvcn if the transformers arc left vulncrablc, or should the
At the highcr distribution voltagcs, thc ovcr-rated gappcd arrcstcr limit overvoltage dutics which cxcccd the durations
MOV arrcstcr application does not cvcn limit tcmporary or magnitude of voltages to which the prolcctcd transformers
ovcrvoltagcs to thc applicd tcst magnitudc for thc arc dcsigncd and tcstcd, cvcn if this results in infrequent
corrcsponding transformcr insulation levcls. For cxamplc, a arrcstcr failures?
24 kV rcsistivcly-gradcd gapped arrester could bc applied on

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on September 11,2024 at 13:26:11 UTC from IEEE Xplore. Restrictions apply.