You are on page 1of 9

1592 IEEE Transactions on Power Delivery. Vol. 7. No. 3.

July 1992

Full Scale Lightning Surge Tests of Distribution Transformers and Secondary


Systems
Gary L. Goedde,Member, IEEE Roger C. Dugan, Sr.Member, IEEE Lawrence D. Rowe

Cooper Power Systems


Thomas A. W o n Technical Center
Franksville, WI 53126

Keywords: Low-sidesurges, DistributionTransformers,Secondary appears to be subjected to a sudden current surge injectedintoits


Surges. unprotected low voltage (secondary) terminals. This current
may induce a high voltage in the primary winding causing,
U Low&& : surges ale known to CsUIse failulw of disaibution typically, a layer-to-layer insulation failure. The Task Force on
transformers. They also subject load devices to overvoltages. A Low-side Surge Requirements for Distribution Transformers of
full-scale model of a residential service has been set U in a the IEEE Transformers Committee was commissioned to study
laboratory and subjected to impulses approximating ligitning this problem and a report has: been recently prepmd[ IO]
strokes. ’Ibe tests were made to determine the impulse
characteristics of the secondarysystem and to test the validrty of Low-side surges are actually driven by the voltage developed
previous analyses. Among the variables investigated were across the service cable in re onse to lightning currents flowing
stroke location, thebalance of the surges in the service cable, and through the service cable. %us, any element connected in the
the effectiveness of arrester protection. Low-side surges wee secondary system is exposed to surgesas well as the transformer.
found to consist of two basic components: the natural frequency There continues to be considerable interest in this subject from
of the system and the inductive response of the system to the theend user viewpoint due to the increased attention being given
stroke current The latter component is m p s i b l e for transformer to power quality. (111 Sensitive electronics in modern load
failms while the formermay be responsible for discharge spots equipment are more subject to failure and misoperation due to
often found around secondary bushings. Arresters at the service transient “spikes” on the secondary thanolder electromechanical
entrance are effective in diverting most of the energy from a equipment. An@ystsworking in this area typically assume that
lightning strike, but may not protect sensitive loads. Additional lightning surges are cou led to the secondary through the
local protection is also needed. The tests affirmed previous interwindingcapacitanceopthe distributionm f o m r . ~lthough
simulations and uncovered additional phenomena as well. it may not matter to the end user how the spikes originate, one
thing that will be shown in thispaper is that the low-side surges
seen by the transformer and the spikes appearing at the load are
closely related and are likely due to surge currents flowing on
There has been considerable interest in the subject of lightning- service cable neutral.
induced surges on 240/120-volt secondary circuits within the
last decade. Electric utilities have generally not provided This paper describes recent full-scale tests of a residential 240/
overvoltage rotection for this class of service beyond that 120-voltsewice. The urpose ofthe tests was to more accurately
providedby J$spark gaps in the revenue meter. The expense for identify and d e b alfthe phenomena associated with low-side
providing protection has not been consideredeconomical except surges that had been observed previously and to evaluate the
in special cases with persistent failure problems. Thus, both the effectiveness of protection options for secondary systems as a
distribution transformer and load devices are exposed to surge whole.
voltages that arise on the secondary.
DESCRIPTION OF TESTS
Martzloff [11 was one of the first to analyze surges on the
secondary and propose test waves for sensitive load devices. Figure 1shows aschematicdiagram of the test circuit used in the
This work was mostly done from the perspective of the load full-scale tests described in this paper. It consists of a IS-kVA
looking back into the utility system. More recently, there has distribution transformer with a tank-mounted surge arrester and
been considerable analysis of these surges by transformer a mock-up of at i d residential service. For the tests described
manufacturers and interested utilities in lightning-proneregions here, the transgmer was a core-type design with interlaced
withhigh ftansformerfailure rates. [2-101 Thisgrouphascoined secondaries. The primary of the transformer was terminated
the phrase “low-voltage-side cumnt surges” or, simply, “low- with a 500-ohm resistor to represent the surge impedance of the
side surges” to describe the phenomena that cause dstribution overhead line. The pole ground was represented by a 16-
I transformer failures. From the perspective of the transformer, it microhemy inductorin serieswith aresistorthat was varied from
3 to 25 ohms, depending on the case bein represented. The
inductor represents the inductance of the pofe ground lead from
This paper was presented at the Transmission the transformerto the ground,estimated at 0.4 to 0.5 microhenry
per foot. The service entrance ground was represented by a
and Distribution Conference in Dallas, Texas resistance, with about 6 ft. of leads attached No other intentional
from September 22-27, 1991 at the Dallas Con- inductance was added.
vention Center. Sponsored by the BEE Power
The transformerwas connected tothe service entmnce using 150
Engineering Society. ft. of standard No.6 AWG triplex cable. A standard meter base
cabinet was used along with an actual meter, which contains
spark gaps as indicated on the figure. Two 120-volt feeders, one
1593

work. For voltage measurements, a resistor was placed across


about 60 ft. and the other about 66 ft. in length, consisting of 12/ the point to be measured and the CI'was placed around the
2 Type NM cable, were run from the service entrance to load resistor lead, which passed through an insulating tube. This
locatrons. Various loads were attached to the ends of these techoique permits the meassuremat of voltage difkencesbetween
cables, including standard lamp sockets, 120-volt outlets, and two points in the circuit while providing sufficient insulation
simple resistors. between the metering instruments and the test specimen. The
mistor-CT combination was c a l i i by floating an o s c i l l v
With the test circuit in place, the surge generator was connected to line potential to measure the actualdifferentialvoltages. This
to various places as indicated to represent the effects of strokes techni ue is more accurate than using two conventionally-
to different locations. The surge generator consists of a large groungd dividers and subtracting the two measurements: the
capacitor bank tbat was charged sufficiently to deliver voltages to ground are very high at some points in the circuit and
a proximately5000amperesina4xIOmicrosecondwaveshape. it is difficult to retain sufficient accuracy by obtaining the
As was decreased at times when surge voltages became too
high and increased at other times when attempting to produce
difference between two nearly equal voltages.
, failures of devices. The surge generator was placed in three OPEN-CIRCUITVOLTAGES
different locations:
1. Primary phase conductor Figure 2 shows the typical waveforms of the open-circuit
2. Primary neutral conductor voltage appearing across the load (black wire-to-white wire).
3. Neutral conductor at the service entrance The two shotsshown were taken at reduced surge current levels
(2.6kA) tokee thevoltagesinthe 10-15kVrange. Therewere
These arebelieved to be the most common locations for lightning two different c L e s of waveforms observed: one for strokes to
strokes. The fust two repwent strokes to the overhead line the primary side of the circuit and another for strokes to the
structures. The third re resents either a direct stroke to a service entrance side. Surges to either the primary phase or the
P
grounded prticWr of tbe oad structure or simply a stroke to
ground in the vicinity of the service entrance ground.
primary neubal produced essentially the same waveform, although
strokes to the phase wire produced slightly higher magnitudes.
Surges to the service entrance produced a waveform containing
lower natural frequencies.
Since the low-side surge phenomena result from cumnt flowing
011 the triplex neutral conductor between the load and the
transfomer,the grounding resistances were ad'ustedto pull most The open-circuit voltage waveforms consist of two basic
of (he surge current toward the end of the triplex cable opposite components:
the surge generator. The grounding resistance on the end being
surgedwas set to 25 ohms, which is the typical assumed value for 1. The higher frequency oscillations due to the natural
a single driven ground. The other end was set to 3 ohms, which, frequencies of the circuit, and
in the case of a residentialload, might represent 811 exceptionally
good ground achieved by connection to a water system. For 2. A lower-frequency component that is essentially the
strokes to the serviceentrance, one justification for using 3 ohnis inductive and resistive voltage drop along the neutral of
at the distribution transformer would be to assume that the the triplex.
transformerwas cable-fed with the neutral of the cable providing
an excellent ground. This is, in fact, the suspected mechanism The main higher frequency component is about 2.5-3.0 M H z for
for many padmounted transformers that have been observed suokes to the primary side and about 800 kHz for strokes to the
having the signature of a low-side surge failure. This procedure service entrance. These frequencies have been attributed to the
resultedin apeak surge current in the cable of approximately 80 parameters of the lines and cables in the circuit. The role of the
percent of the peak of the surge generator current. (The peaks transformer capacitance has not been fully investigated, but is
generally do not coincide.) believed to be minimal.
Both voltage and current measurements were made using special The lower-frequency component is the root of the low-side
iostrument current transformers designed for laboratory surge surge failure problem in transformers and has been the subject
1594

of much of the previous work, which typically neglected the tank leaving small pits in the paint. The result can be readily
higher -eocy components. [5,6,9] Early test circuits were observed in many transformers by removing the cover and
intent only on &tennining transformer withstand cvability pnd inspecting the low-voltage bushings.
it was not uatil recently that measFment eqwpment and
~ l s ~ o f a c a u a t e y lcneuqeryfw
*-s d e ~ There is no evidence at present that the voltages causing these
componeat were applied. discharges a damagingby themselves,although there is concern
that they couldprecipitatearm-to-twnMumwithin the secondary
These voltage waveforms taken at the load show the natural if repeatedly applied. However, it Seems likely that they are
ikquency oscillations to be very high in magnitude. The accompanied by low-side surges, indicating that the transformer
zr tuciewasverysensitive totheadditionofsmallindu-
capacitances to the circuit. Note that there is significiitit
difference between the magnitudes on the two sides of the
has been in a location that is particularly susceptible to low-side
srirge phenomena. This would suggest that secoodary arrester
protection would be prudent for transformers in this location.
circuit. This is attributed to the fact that one run of the 12./2 Type Variables that might make aparticularlocation more susceptible
NM cable was slightly longer (about 6 ft.) than the other and was include high lightning exposure, long triplex runs, use of o n-
laid in a different route to the termination poitit. The cable
lengths chosen for this test result in reinforcement of some
wire secondary, or a much better ground on one side
service cable.
04elhe
frequencies present in the natural oscillations of the secondary I

system. Thrs may or may not be representative of an actual The high frequency component has, perhaps, more potential for
installation where there are many more cables of much more
varied lengths. Nevertheless, it serves to indicate the type of
voltages present under lightning surge conditions.
damage to sensitive electronic load equipment. Its
Tis
such that it requires a fast-acting surge su pressor to c amp the
voltage. A suppression technique that u d z e s a series inductor
The lower-frequency component is balanced in both sides, as
would be expected for a system containing an interlaced core-
to block high
2 encies should be more successful than a
simple MOV. AI o u p not shown, cases have been recorded
where thepeak of the tugh-frequencytransient voltage to remote
type transformer. This couldbeSeen if one were to draw a curve (true) ground to be in excess of 20 kV. Thus, any electronic
through the average of the higher-frequency oscillations. equipment referenced to another, remote ground point, such I
through TV cable, is particularly vulnerable to experiencing
The high-frequency component of low-side surge phenomena insulation failures.
was not investigated until discharge spots were noticed around
the secondary bushings inside of transformer tanks (see Figure
3). Tesu&m of failed transformers have suggested a correlation
with the appearance of these spots and failure of transformers
due to low-side surges. Also, our laboratory tests had reviously
ikerminectimt a very large low-si& current w e , wit! suecient
rateof change, could cause flashover of the secondarybushings.
However, there was considerable burning observed with these
tests. It is our hypothesisthat the discharges are due to the high
!hquencycompooentof the voltage, alargepercentageof which
IS dropgd across the inductance of the transformer secondary
when re is a path for current to flow. There is little ener
involved and the distance across the bushing is too great for 1%
volts to suslain a power follow arc. Therefore, low energy
spitting occurs off sharp edges on the secondary leads inside the

P
In
STROKE TO
SEWCE ENTRANCE

I I I I I I I I I
TlME z MICROSEC/DIV

Figure 3: Typical disclmrge spots (discolored area above


Figure 2: Open-circuitvoltages appearing at the load for and to the I& of the bushing on the lep) often found around
surges to the primary and surges to the service entrance. secondary bushings in distribution tran~omersin lightning-
vrone areas. Spots are aligned with tlie sharp points on the
secondary lead connectors.
c

1595

The interaction of the secondary system elements involvedin the 2


high fie ency ringing has not yet been thoroughly studied. e
Besides lengths of the cable, the type of service cable also 8
F a t l y influences the amount of ringing obsemd.When Type
SEU cable, which has more capacitance, is used instead of
triplex, the high-ffaquency componentis evenmoreproaouaced. pl
’I
SHORT-CIRCUJT CURRENTS
8
Figure 4 shows the currents resulting in the Circuit with a short
circuit across the load In the particular case shown, lamp $
sockets containing 25-Watt incandescent bulbs flashed over. 8
Thus, the cumnt increases at a relatively low rate for about 1 a
microsecond until the flashover occurred ’ h i s explains the
apparent discontinuity at the beginning of the waveforms. The
surge was applied on the primary phase conductor. Therefore
there is a minor difference in the cumnts in the 120-volt sides of 5
f
the circuit due to the bias caused by the arrester discharge
voltage.

Note that the currents are nearly void of the high-frequency Figure 5. Surge voltage and current waveformsfor a 5-kA
nts observedin the open circuit voltage. The induaances stroke to the primary with a 400-watt (balanced) load.
FE= crrcuit elements suppress the flow of current at those
frequencies and sigai6icantly reduce the rise time of the lower W E
frequency component of the current as well, as notedearlyin the
investigation of low-side surge phenomena [2]. Thereby, the Figure 6 shows several waveforms throughout the circuit for the
often-measured “low-sidecurrentsurge”is created. In the short- typical case with MOV-typearresters at the service entrance.
circuited load case, most of the high-frequency voltages appear For the usual 5-kA test wave applied, the surge currents in the
near the end of the triplex where the surge was applied two 120-volt sides of the circuit reach 880 and 816 amps,
respectively. They differ slighrly due to the bias of the primary
arrester discharge. For strokes to the primary neutral, they
would be expected to be almost exactly equal, assuming the
service entrance arrester is sound

With no load present for damping, the voltages at the load (6b)
are dominatedbythe higher-frequency component. The seFice
8 entrance arrester clamps the lower fiequency component to the
w
expected value, but is not able to control the higher-frequency
component that appears at the load. This is due to both the
characteristic of the arrester and the response of the system. It
is, perhaps, a classical case of the arrester not being close enough
to the element to be protected for the frequency involved to be
effective. This voltage transient did not cause a flashover in this
test, but robably would have if a conventional light socket or
electricloutlet had been installed at the load. Such events did
occurin subsequent tests after such a transient persisted for about
1-2 microseconds. A voltage of this magnitude is certainly of
concern for sensitive electronics, which, at the least, might
TIME 5 MICROSEC/DIV inteTret this as an interferiog signal. Therefore, there i s tbe
contmuing need for localized transient suppression even if a
Figure 4 . Currents tlirough the triplexfor a short circuit at serviceentrancearresterispresent. Thepresence of asignificant
the load for lightning stroke to the primary. resistive load (light bulbs, 120-volt heaters, etc.) can damp the
high ffaquency component appearing at the load; however,
WITH RESISTIVE LOAD inductive loads (motors, lam s with ballasts) will probably not
be helpful because of the hi$: frequency.
1 Figure 5 shows voltage and current waveforms measured at the
indicated locations in the circuit for a “normal”case where there The voltages across the transformer low side (6a) in this case,
is a resistive lighting load of 400 W on each 120-volt side of the measured Xl-X2and X3-X2,also exhibit a significant amount
circuit. For a 5-kA surge, 4.5 kA goes to ground through the of the higher-frequency component. In fact, there is a much
triplex neutral and load circuits. The currents in each half of the higher Srepuency present at the bansformer for which the metering
triplex aks at approximately 600 Am s, differing by about 20 was inadequate to capture properly. The lower frequency
Amps. R e load voltage reaches 9.4 kgand displays the higher- component is of the niagnitude and waveshape that would be
fkquency component noted in the open-circuit voltage, but ex cted for an interlaced transformer subjected to a nearly
significantl7 more dam d If this voltage were able to persist bacedcurrent surge. The separationofthevoltage waveforms
longer at thls level, the g p socket is likely to flashover, which is again due to the primary arrester discharge being reflected
it did in subsequent tests.
1596

ALANCED SURGES
The preceeding has described tests in whichthe surge currentsin
both 120-volt sides of the circuit were nearly balanced The
inductances in the circuit, articularly in the transformer, tend to
force this condition. &n the surge current through the
transformer is balanced, interlaced secondaries offer natural
protectioaforthe~,essreaialEypy0fferioglessindudance
to the surge current than would a non-mterlaced design. It can
be showninthelabonrtoIythatsut#cting anioterlaaxidormer
to an unbalanced surge, on only one half of the secondary
wioding, will result in a signature low-side surge faiIure,[7]
Tear-downs of some failed transformershave shown breakdown
patterns cone ondine to the volta e distributionanticipatedfor
an unbalance?low-slde surge. k e is also the additional
r
evideooe that non-interlaced @amformers with secondary amster
protection have lower failure rates than unprotected interlaced
transformers. This has led to the hypothesis that the existence of
unbalancedsurges in the SeCoMbry system explaiosthe umpeded
lightning failures of transformers with interlaced secondaries.

( a )Transformer Voltages The laboratory tests have shown that unbalanced surges do
generate some very high voltages, but effortsto induce a failure
in the interlaced transformer were thwarted by laboratory
limitations. On two OccasioLls the circuit fhhed over to h a d u c e m
causing damage to expensive digital oscilloscopes and other
metering. The t of damage would suggest the existence of
some very high& ency transients, ossibly due to arcing
phenomena, but this is conjecture since waveforms were not
captured because of the fadures. Subsequent unbalanced surge
shots were performed with conservatively low values of surge
current to prevent further instrumentation damage.

The worst cases of unbalanced surges seem to occur when one


I- -I 120-volt leg of the circuit is protected by a transient surge
suppressor (outlet protector) while there is neither load nor
protection on the other side. Figure 7 shows the voltages
appearing at the transformer terminals and the current in the
triplex for one such case. A 5-kA lightning surge was applied at
the service entraoce in this case. The, 120-volt outlet protector
allows the current to build in one side for about 2 microseconds

- 1 and then the meter gap on the other si& flashes over. The
secondaryvoltages at the transformer initially swing in opposite
- polarities, as expected, until after the flashover when the surge
1 1 1 1 1 1 1 1 1 1 1 1 1 1 current returns to a more balanced equilibrium. Then the
voltages track more closely together, yet not as they would For a
TIME 2 WK#OSEC/ON perfectly balanced surge. There is still some imbalance in the

(b)Load Voltagesand Triplex Currents surge current in the transformer because the lightning current in
one side goes straight across the meter gap while it must travel
a circuitous route through the outlet protector in the other side
Figure 6. Voltage and current wavefomsfor a 5-kAstroke to (see Figure 7).
the primary with a MOV-type arrester at the service entrance.
Load is open-circuited. Although the investigation of unbalanced surge phenomena
could not be completed as planaed, one interesting findin was
through the turns ratio. There is ooe interesting anomaly with that ~nbalaa~ed surges genedy cause m u d higher overvoftages
these waveforms: the sudden increase that occurs at about 12 thao balanced surges. Besides the oscillographic evidence, lab
microsecondsinto the surge. This hasbeen traced to the turn-off personnel also observed that flashovers frequently occurred in
of the service entrance arrester and the resulting change of multiple locations in the system during unbalanced shots. Perhaps,
current in an inductive circuit. There is some concern that this one flashover would precipitate the others. Such pheoomena
mi t cause additional voltage stresses inside the transformer seemed to be worse when the outlet protector failed, typically by
a$tfithereb, precipitate turn-to-turn failures in the secondary puncturing an MOV disk. Therefore, it would seem prudent to
winding. Simulationshave indicated that the voltage magnitude try to force the surge current to be more balanced between the
of the turn-off transient stays relatively constant with varying two 120-volt sidesofthesecondatysystem. Heavy-duty arresters
lighming stroke magnitudes and is primarily a function of the in the service entrance would be one way to help ensure a
arrester discharge voltage and the ductances of the system. balanced condition.
There have been no failures attributed to this transient, but
investigations are continuing into this phenomenon.
1
1597

Table I
sharin of SurgeCurrents
Between48OVfkrviceEntrance Arrester
and l50V Outlet Protector
for a nominal 5-M,4x 10Surge
SeWb ouikt
Entrance Protector
Length of S.E.Arrester Outlet Protector

TAANSFOWER

3 .O

well when used in conjunctionwith a heavy-duty service entrance


arrester. There must be enough cable se arating the two devices
so that the cable's inductance forces the & u k of the surge energy
through the serviceentrance arrester before the current has time
to build to a damaging level in the outlet arrester. Table I shows
the results of current-sharing tests performed for two shorler
lengths of cable than the 60 ft. used m our usual tests. It should
be noted that this was for one type of surge current waveform (4
x 10 microseconds) to the service entry anester. Other waveforms
TIME 5 HICROSEC/DIV will give much differentresults because of the circuit constants.

For even as little as 10 feet with this surge, most of the surge
't x1
TRIPLEX
cu- 1 current is forced into the service entrance iurestet and h e outlet
protector was able to survive this test. In practice, however,there
w should be more separation between the service entrance and the
outlet protector to be more certain of safe coordination. Many
low-side current surges, while usually less than 5 kA, have a

I slower wavefront than the test wave, requiring more inductance


to limit the current.

Some outlet protectors were able withstand a higher current


level under these conditions than when there were no service
entrance arresters. The difference appears to be related to the
current waveform. Once the service entrance arrester clamps the
voltage, it appears that the outlet protector is being charged by
adcvoltagesourcethroughaninductance.Thecunentwaveform
resembles a ramp function. Thus, with the service entrance
arresterpresent, the outlet protectorsees neafiy the same waveform
Figure7:Voltageandcurrentwavefofonnsresultin~oman for every low-side surge, differing in magnitude only due to the
unhhncedsurgecurrentcreatedby astroke totlieservice duration of the surge. Without the arrester, the current rises more
entranceandonel20-voltoutletprotectoratthe load. steeply, sometimes causing puncture failures of MOVs in the
outlet protectors.
ARRESTER APPLICATION
Another interesting observation is that serviceentrance arresters
These laboratory tests have served to c o n h the low-side surge virtually guarantee that the lower-frequency component of the
phenomena previously identified by analysts through simulation low-side surge will be balanced Thus, they will contribute to
and some testing. [2,5,6,7] Therefore, the arrester rating the ability of interlaced transformers to survive low-side current
recommendations of reference 7 still apply. For the 150 fi. of surges. They would also significantly reduce the possibility of
triplex cable used in this test, it was found that approximately the secondary circuit being subjected to the extremely high
115 of the current that flows toward the opposite end of the cable voltages that were encounted in the laboratory when attempting
would flow through the arresters. As the triplex cable length to analyze unbalanced surges. Non-interlaced transfoFers
increases, the current through the arrester at the transformer and would require coordinated protection.
at the service entrance can approach 113 of the lightning stroke
ament. To properly coordinate with a 1oo-kA plimary distribution Switching-surgeandpower frequency arrester coordination were
arrester requires a device capable of 33-kA surge current. not specifically tested, however, several of the lower-rated
arresters began conducting well below the switching surge
Regarding other issues in the coordination of arresters in the protective level of the primary arrester reflected through the
secondary system, we have found that some outlet protectors, transformer. Since there are many common switching events on
when used alone, are basically one-shot protectors, fading for as power systems that cause voltages to swell well above normal for
little as a l-kA surge. In some cases, the varistors appear to several cycles, it would seem likely that some of these devices
survive the surge but the fuses in series operate. Other types, will fail thermally. Therefore, we favor a higher-rated arrester
particularly those with multiple varistors appear to be more than the standard 175V secoodary anester for the seMce entrance.
robust. However, even the weaker ones survived our tests quite A 480V arrester developed for low-side surge protection of
1598

distribution transformers performs acceptably and has a lower Power Delivery, Vol. 3,No. 2, April 1988.
discharge voltage than a 650V mster. If Figure 6b is typical of 5. Roger C. Dugan, Stephen D.Smith. “Low-
the voltages reaching the loads, the dischar voltage of the Voltage-Side Current-Surge Phenomena in Sin e-
service entrance amster is not very critica ras long as it is Phase Distribution Transformers Systems”. ZEfE
maintained in the 1-2kV range. Transactions on Power Delivery, Vol. 3,No. 2,
April 1988,p ~637-647.
.
CONCLUSIONS 6. J. L. Pun, et al. ‘‘Ligbtninf Induced Failures In
Distribution Transformers’ . IEEE Transactionson
Power Delivery, Vol. 3,No.4,October 1988.
Low-side surges fyp‘callyconsist of two major components: a 7. Roger C. Dugan, Stanley S.Kershaw, Jr., Stephen
component ranging m fhquency from 0.8to 3 M H z anda slower- D. Smith. “Protecting Distribution Transformers
changing corn nent that represents the inductive response of from Low-Side Current Surges”. ZEEE
the system to g r a t e of change of the lightning stroke current in Transactions on Power Delivery, Oct 1990, pp.
the triplex cable. The tests confirmed earlier analyses of low- 1892-1901.
side surge phenomena in transformers and verified that the 8. M. Darveniza, D. R. Mercer. “Lightning Protection
phenomena affecting the transformer can be analyzed by simply of Pole Mounted Transformers”. IEEE
considering the inductances in the secondary system, neglecmg Transactionson Power Delivery, April 1989,pp.
the cable capacitances. The 1087-1095.
9. D. R. Smith, J. L.Puri. “A Simplified Lumped
Parameter Model for Fmdhg Distribution
its greatest impact is Transformer and Secondary System Responses to
Lightning”,IEEE Transactions on Power Delivery.
Arresters at the service entrance are useful for forcing low-side July 1989,p . 1927-1936.
surges to be balanced and for diverting most of the surge energy 10. IEEE TransPormers Comminee, “ScondafY (LOW-
away from outlet protectors withlessewrgy-handlingcapability. side) Surges in Distribution Transformers, Task
Unbalanced surges were found to generate much higher surge Force Report, submitted for presentation at the
voltages in the secondary system and are,therefore, undesirable. lEBE T&D Exposition, Dallas, TX, Sept. 1991.
Balaocing the surges between the two 120-volt side also aids in
the ptectim of tramfoxmm with interlaced secondaty windings.
11. -
J. Burke, et. al., ‘Tower Quality Two Different
Perspectives,” ZEEE Transactionson Power
Non-interlaced transformers subject to low-side surge failure Delivery, Vol5., No.3,July 1990,pp 1501ff.
sbould be protected with arresters in accordance with previous
a=b=. mJ.O~eceivedhisBSdegreeinElectricalEngioeering
Technology at the Milwaukee School of Engineering in 1980.
The service entrance arrester in all ratings studied (175V to He has been with Cooper Power Systems, Franksville, WI,
since 1969.Hisexperience has beeninsemiconductordevices,
insulation and transformer core evaluations. He has been
involved in the development of ceramic-boded varistors,
electrical porcelain and has been responsible for projecls that
include mechanical, electrical, and ultrasonic testing. He is
currently theproject leader of an extensive capacitor fuse study
roject and is involved in electricaldistribution system testing
practical lengths of secondary circuit feeder cable would be for devising improved protection techniques. Gary has co-
sufficient to force most of the energy from a low-side sur e authored four ublications and has five U.S. patents. He is a
through the service entrance arrester and, thus, help prevent tie
failure of the local protective device. One area that remains open
P
Registered pro essional engineer in Wisconsin.
to investigation is the effect of the turnoff transient of MOV-t
arresters in load circuits on the transfomler. Another is t
switchingsurge andpowerfrequency coordinationof arresters in
T C. -(M ’73,SM ’82)received his BSEE degree
from OhioUniversityio1972andhisMas~rofEngrinElectric
Power Engineering from RPI, Troy, NY,in 1973. Since then
the secoadacy Circult. he has been with the Systems Engineering department of
Cooper Power Systems, first in Canonsburg, PA, and recently
REFERENCES transfenin with the dtpartment to Thomas A. Edison Technical
Center in kaaksville, WI. Recently, his main interests have
1. F. D. Martzloff, T h e Propagation and Attenuation of been in the computer simulation of power systems, transient
Surge Voltages and Surge Currents in Low-Voltage response of transformers, and power quality issues such as
AC Circuits,” ZEEE Transactionson Power harmonics and low-side surges. He is a member of both the
. Apparatus and Systems, Vol PAS-99, Jan/Feb 1980, Power Engineering Society and the Computer Society. He
. 129-133. currently serves on the Computer and Analytical Methods
2. F“Susceptibility
J. McMillen, C. W. Schoendube, D.W. Caverly.
of Distribution Transformers to Low
Subcommittee of the Power System Engineering Committee.
Voltage Side Lightning Surge Failure”. ZEEE ace D.Rowe received his A.A.S Degree in Electrical
Transactionon Power Apparatus and Systems, Vol. Engineering Technology from the Milwaukee School of
PAS-101,No.9,Sept. 1982,pp. 3457-3470. Engineeringin 1989.He hasbeen with CooperPowerSystems,
3. Surge Cltaracteristicsand Protection of Distribution Franksville, WI,since 1980.His ex rience has been in testing
Transformers. Electrical Power Research Institute and data acquisition in the High Vogge Labs. Thistesting has
Final Re rt EL-3385,Jan. 1984. beenperformedinsupportofvariousproductlinesaodnsearch
4. C. J. M c k l e n , C. W. Schoendube, D. W.Caverly. projects. He developed the data aquisition roceduze for cbe
“Scaled Low-Voltage Side Surge Current Tests on a secondary surge testing performed since 19&.
Model Distribution System”. IEEE Transactionson
1599
Discussion arresters to interlaced transformers for the purpose
of protecting the transformer.
R. Hartana, J. Skliutas,R. Walling, (GE Industrial and Power Systems, 1. The authors state that tear-downs of failed
Schenectady, NY): The authors have presented an interesting study of
secondary-side surges. We do not disagree that there may be benefit to transformers have shown evidence of unbalanced
certain customers of applying arresters at the service entrance, although secondary surges, and they also state that
there may be controversy in the industry as to whether this arrester non-interlaced transformers with secondary
installation should be the responsibility of the utility or the customer. arresters have lower failure rates than unprotected
Numerous tests and studies have shown the susceptibility of non-inter- interlaced transformers. The authors cite no
laced transformers to secondary surges, and the discussers agree that reference nor do they offer any data to support
application of transformer secondary arresters is necessary for these this contention. The discussers do not believe
transformers. We do have concern that some of the statements in the paper such statements are appropriate for publication
could be interpreted as justification for applying secondary arresters to without substantiation.
interlaced transformers for the purpose of protecting the transformer. 2. The magnitudes of the unbalanced surge currents
1. The authors state that tear-downs of failed transformers have shown produced by the authors’ tests (Figure 7 ) are
evidence of unbalanced secondary surges, and they also state that inconsequential in that they are far below that
non-interlaced transformers with secondary arresters have lower failure required to fail an interlaced transformer. The
rates than unprotected interlaced transformers. The authors cite no authors note their difficulty in testing at higher
reference nor do they offer any data to support this contention. The discharge currents due to uncontrolled flashovers
discussers do not believe such statements are appropriate for publica- in the test circuit. We have performed extensive
tion without substantiation. study of secondary surges with profound load
2. The magnitudes of the unbalanced surge currents produced by the imbalance, both simulation using a sophisticated
authors’ tests (Figure 7) are inconsequential in that they are far below wide-bandwidth transformer model and full-scale
that required to fail an interlaced transformer. The authors note their power laboratory testing, and conclude that
difficulty in testing at higher discharge currents due to uncontrolled secondary flashovers inherently balance unbalanced
flashovers in the test circuit. We have performed extensive study of surges before they approach the magnitude
secondary surges with profound load imbalance, both simulation using endangering primary-winding layer-to-layer
a sophisticated wide-bandwidth transformer model and full-scale power insulation. Our results indicate that current
laboratory testing, and conclude that secondary flashovers inherently unbalance does not exceed 1 M , which is less than
balance unbalanced surges before they approach the magnitude endan- 50% of the critical magnitude. The authors observe
gering primary-winding layer-to-layer insulation. Our results indicate that unbalanced surges create high voltages, and
that current unbalance does not exceed 1 kA,which is less than 50% of this is the underlying reason why the flashovers
the critical magnitude. The authors observe that unbalanced surges result.
create high voltages, and this is the underlying reason why the flashovers
result. 3 . For an arrester application to be prudent, the
3. For an arrester application to be prudent, the present worth of savings present worth of savings in equipment failures must
in equipment failures must exceed the investment required for the exceed the investment required for the arrester.
arrester. Because the failure rates of interlaced transformers are very Because the failure rates of interlaced
low, it is very questionable if any incremental reduction achieved by a transformers are very low, it is very questionable
secondary arrester could justify its installation. Using the economic if any incremental reduction achieved by a
factors shown in Table 1 of this discussion and applying the procedures secondary arrester could justify its installation.
described in Reference [I], the failure rate of an interlaced transformer Using the economic factors shown in Table 1 of this
would need to be reduced from the typical .2%/yr level to minus dlscussion and applying the procedures described in
.OS%/yr to justify a $25 investment in the secondary arrester. Further- Reference [l], the failure rate of an interlaced
more, there are a number of distribution transformer failure modes, transformer would need to be reduced from the
such as excessive overload, which are not affected by secondary typical .2%/yr level to minus .08%/yr to justify a
arrester installation. Although the economic factors may vary between $25 investment in the secondary arrester.
utilities, we have seen no evidence that justifies the installation of Furthermore, there are a number of distribution
arresters on the secondaries of interlaced transformers for the purpose transformer failure modes, such as excessive
of protecting the transformers. overload, which are not affected by secondary
arrester installation. Although the economic
Reference factors may vary between utilities, we have seen no
evidence that justifies the installation of
[I] D.J. Ward, “Evaluating Product Reliability Costs”, IEEE Transac- arresters on the secondaries of interlaced
tions on Power Delivery, April, 1990, pp. 724-729. transformers for the purpose of protecting the
transformers.

Reference

R. Hartana, J. Skliutas. R. Walling (GE Industrial and 1. D. J. Ward, ”Evaluating Product Reliability Costs”,
Power Systems, Schenectady, NY) - The authors have IEEE Transactions on Power Delivery, April, 1990,
presented an interesting study of secondary-side pp. 724-729.
surges. We do not disagree that there may be benefit
to certain customers of applying arresters at the
I service entrance, although there may be controversy in Table 1
the industry as to whether this arrester installation Economic Factors Used in Failure Rate Evaluation
should be the responsibility of the utility or the
customer. Transformer Cost $600
Failure Changeout Cost $400
Numerous tests and studies have shown the Fixed Charge Rate .17
susceptibility of non-interlaced transformers to Discount Rate .12
secondary surges, and the discussers agree that Escalation (Inflation) 4X/yr
application of transformer secondary arresters is Economic Life 30 yr
necessary for these transformers. We do have concern
that some of the statements in the paper could be
interpreted as justification for applying secondary Manuscript received October 16, 1991.
Gary L. Goedde, Roger C. Dugan, Lawrence D. Rowe: We wish to tial because the surge generator charge was reduced to levels the system
thank the discussers for their excellent comments and questions on this would tolerate. However, when operated at higher levels, numerous
paper. We will address the three major points individually. flashover events occurred. Shortly before this paper was presented if was
Point I: We have performed tear-downs on several batches of failed learned that the 15-WA interlaced transformer used in the test incurred a
distribution transformers from utilities in Florida, Wisconsin, and Ohio. A non-typical extended-creep failure in the secondary. We believe that this
typical batch consisted of 10 transformers, of which 5 to 7 were found to occurred during our attempts to generate an unbalanced surge since those
have indication of low-side surge failure. We agree that the lightning-re- were the only events in our entire test series that generated any significant
lated failure rate of interlaced transformers is lower than that of non-inter- voltages at the transformer secondary terminals. It is possible that one
laced designs. If our samples were a valid indication, the rate is approxi- cannot rely on absolute differences in peak current values to define the
mately one-third as much. severity of an “unbalanced surge”. Rates of rise exert an important
Failures of non-interlaced transformers due to low-side surges are easy influence and the timing of the peaks may be another factor: in the
to detect: they have a “signature failure” due to an inherent weakness to example we present, the peaks do not occur at the same time.
low-side surges. They are also more prone to showing “spotting” around Point 3: First, there seems to be no question that it is economically
the secondary bushings. Interlaced transformers have been found to ex- justifiable to apply arrester protection to non-interlaced transformers in
hibit a variety of failure modes compatible with theoretical modes of lightning-prone regions. The question is concerning the justification of
failure for low-side surges. We have seen units with failures between interlaced transformers.
secondaries, turn-to-turn in the secondary, layer-to-layer in the primary, Our calculations using the economic factors suggested by the discussers
and with discharges from the edge of the coil to the core. The latter two in their Table 1 yield a cost justification of about $8.50 for each 0.1
primary winding failure modes occur in locations where unbalanced percent of improvement in the failure rate. The discussers suggest that the
low-side surges would be expected to produce high voltage stresses. We failure rate for interlaced transformers is 0.2 X . Of the published data on
have at least one clear example of each failure mode from our relatively transformer failure rates from utilities with significant lightning, that is the
small sample and we believe that other investigators should easily be able lowest rate we have seen. Most of the typical failure rates are closer to
to verify that independently. 0.4%. Assuming that a failure rate of 0.05% is achievable with secondary
The point is that there is a measurable failure rate of interlaced arrester protection, protection costing less than $30 should be justifiable
transformers due to low-side surge phenomena; they are not immune as for utilities in lightning-prone regions. For non-interlaced transformers,
some would contend. Our comments concerning the relative failure rates the economic value of adding secondary arrester protection would be
of arrester-protected non-interlacedtransformers and interlaced transform- expected to be 2-3 times greater.
ers come from our internal data that has been reported in various IEEE It is sound engineering practice to install low voltage distribution
Transformer Committee forums for several years. The best failure rates arresters at the service entrance just as it is at the transformer. It is
we have seen for interlaced transformers installed in lightning-prone particularly important to do so when interlaced transformers or secondary
regions vary from 0.2%/yr to 0.4%/yr. From 1986 to 1988, our company arrester-protected transformers are used.
produced nearly 25 OOO non-interlaced transformer protected with MOV-
type secondary arresters. The annual failure rate of these units as of 1989 Reference
was 0.05%. Interestingly, this is very similar to the failure rate reported
by Horton, et. al., [l] for a region of the country with low lightning 111 W. F. Horton, S. Goldberg, and C. A. Volkmann, “The Failure
incidence and, sometimes, no arrester protection at all. It is possible that Rates of Overhead Distribution System Components,” proceedings
this is the achievable failure rate if the transformer can be completely of the 1991 IEEE Power Engineering Society Transmission and
pnHeaed from lighting. Distribution Conference, Dallas, TX, 1991, (91CH3070-0).
Point 2: The magnitudes reported in Figure 7 are indeed inconsequen- Manuscript received January 21, 1992.

You might also like