You are on page 1of 8

IEEE Transactions o n Power Delivery, Vol. 3, No.

2, April 1988 599

REVIEW OF GENERATOR SURGE COORDINATION


INCLUDING GENERATOR BREAKERS

E.P. Dick ( M )
H . Dhirani
B.K. Gupta (SM) P.W. Powell (M) R.A. Shinn (M) A. Greenwood (F) J.W. Porter (SM)

Ontario Hydro Virginia Power Tennessee Valley Rensselaer Electric Power


800 Kipling Ave 1 James R Plaza Authority Polytechnic Inst Research Institute
Toronto, Ont Richmond, VA Chattanooga, T N Troy, NY Washington, DC

Abstract - The surge protection practices for large generators were In this paper, the term generator breaker is reserved for switchgear
reviewed through a utility survey. The surge environment of large on the L V side of the step-up transformer. Figure 1 shows the
generators, their 60 Hz withstand test value, their designated typical configuration and the names used for other apparatus. For
i m p u l s e w i t h s t a n d , a r r e s t e r p r o t e c t i v e levels, a n d i n s u l a t i o n . .
directlv-connected generators the eauimnent -
shown in the Figure
breakdown tests on two machines were then compared. Several between the machine and the auxiliary service transformer is
topics are identified for further study including arrester protective oraitted.
margins, zinc oxide arrester voltage endurance, stuck breaker poles
on fault clearing, breaker out-of-phase closing surges, step-up
transformer winding resonance, and resonance within the machine
windings.
H V SYSTEM
HV CAPACITOR BANK
I. INTRODUCTION
H V S T A T I O N BUS
The Electric Power Research Institute, through the award of H V T I E LINE, BREAKER
contract 2594- 1 “Surge Protection of Generators”, has initiated a
H V ARRESTER
review of generator protection practices. This review was prompted
by the failure of one large generator and a changing technological S T E P - U P T R A N S F O R M ER
environment including the use of generator breakers, zinc oxide
L V ARRESTER
arresters, and higher transmission voltages.
L V CAPACITOR
This paper is a status report on the progress to date. As such it
aims to invite comments and input from interested parties. The L V SYSTEM P T
project is scheduled for completion by the end of 1987 and AUXILIARY SERVICE
feedback at this time from utilities and consultants could assist in
GENERATOR BREAKER
guiding the balance of the studies. More detailed technical
reporting of the results achieved and a workshop are planned for GENERATOR ARRESTER
the end of the project.
GENERATOR CAPACITOR
The scope includes a utility survey on the ranges of configurations, GENERATOR P T
the design of protection and the occurence of any failures possibly
GENERATOR SERVICE
related to surges. In addition, the paper covers the calculation of
generator terminal voltages for many significant surge sources and GENERATOR
the testing of insulation withstand in two machines using 60 Hz,
dc, and impulse waveforms. The voltage levels are compared to the NEUTRAL DISCONNECT
machine rating, arrester protective levels, and arrester vulnerability N E U T R A L TRANSFORMER
to temporary overvoltages. Areas needing further study are
identified. N E U T R A L RESISTOR

2. DESCRIPTION O F SYSTEM CONFIGURATIONS

A questionaire was sent to about SO utilities asking for information Figure I Configuration of G e n e r a t o r a n d Associated
on their present practices for generator overvoltage protection and Apparatus
typical system configurations. Detailed information was also
requested for one recently designed unit with responses received
for 32 thermal and two hydraulic stations. Thirteen of these
included single line diagrams. Turbogenerator ratings covered the
range from 270 to 1450 MVA, 18 to 26 kV. Nine are configured
with generator breakers. Generator breakers have made a significant penetration into the
United States utility market - 84 three phase units being supplied
by two manufacturers since 1972. Their voltage ratings cover 19 to
27 kV with interrupting current ratings from 18 kA in load
switches to 275 kA in circuit breakers. The main interrupter
normally uses an axial air blast, with a shunt resistor of about one
a 7 CFI 216-5 A p a p e r recommended a n d a p p r o v e d ohm at the higher current ratings to reduce transient recovery
by t h e I E E E S u r g e P r o t e c t i v e D e v i c e s Committee of voltages. A smaller interrupter is used to break resistor current.
t h e I E E E Power E n g i n e e r i n g S o c i e t y € o r p r e s e n t a - This may be shunted by high ohmic resistors with delayed switches
t i o n a t t h e IEEE/PES 1987 W i n t e r M e e t i n q , New O r l e a n s , or by silicon carbide elements, possibly to control voltages due to
L o u i s i a n a , F e b r u a r y 1 - 6, 1987. M a n u s c r i p t s u b m i t t e d chopping of step-up transformer magnetizing current. Recently,
August 29, 1986; made a v a i l a b l e f o r p r i n t i n g SF, puffer breakers have become available, rated to interrupt 100
December 2 , 1986. kA faults.

0885-8977/88/W599$01.WO1988 IEEE
600

The units equipped with LV arresters (26 of 34) usually employ The following design methods were found to be practiced by
gapped (22), rather than zinc oxide types (4). They are usually different utilities:
applied at the generator (20), and less often at the transformer (4)
or both (2). Their voltage rating (line-to-ground) varies from 1.0 a) the majority use guidelines and past experience,
to 1.13 times that of the generator (line-to-line). About half of the
units surveyed use LV surge capacitors (17). These are located at b) a few use analytical calculations of the surges transferred to
the generator (15), the transformer ( I ) , or both (1). The value the generator based on the method of inductive and
where indicated on single line diagrams is usually 0.125 pF and in capacitive components[ I],
one case 0.5 pF. The isolated phase bus ranges in length from 20 to
120 m, with a typical capacitance of 80 pF/m. c) one utility uses a detailed Electromagnetic Transients
Program simulation of these components, supplemented by
All single line diagrams showed neutral resistance grounding, calculation of generator breaker transient recovery voltages.
although reactance grounding is also mentioned in the literature. This approach was used following a generator insulation
Two diagrams showed neutral disconnect switches, which could for failure caused by HV to LV flashover of the step-up
whatever reason be left open. This may lead to ferroresonance due transformer.
to PT saturation, as will be discussed later.
Figure 2 illustrates on a log-time scale the 60 Hz rating, factory
The survey indicated that most utilities have transmission voltages tested 60 Hz withstand, impulse designated withstand, and
of 345 kV (16), other levels being 138 ( I ) , 230 (7), 500 (9),and 765 characteristics of protection for a typical 24 kV generator. These
kV (1). Three phase step-up transformers are most common (22 of are compared with the breakdown voltages measured for two
33), other options being single phase units (9) and parallel banks machines under this project. All voltages are shown as peak line-
(2). The HV arrester is usually silicon carbide (24 of 32), and less to-ground values.
often zinc oxide (8) with a voltage rating (line-to-ground) ranging
from 74 to 87 % of the system voltage (line-to-line). The HV tie
line to the switchyard ranges in length from 15 to 1800 m with a V
typical capacitance of 12 pF/m. For units dedicated to HVDC
systems, large shunt capacitor banks usually are present. These will
affect both switching surges and the possibility of unit self 340 4
excitation. - 320J
14

cl
3. UTILITY PROTECTION PRACTICES 5 150
Table 1 highlights the variations in protection practiced by the
2 140 t
utilities surveyed. Several prefer to rely on HV arresters for A13
limiting external surges. Reasons given during a telephone follow-
up for not using LV arresters include good past experience and the Ill 4
Y
fear of arrester failure. Most of these utilities (5) also indicated
that the surge transfer ratio of the step-up transformer is unknown.
Thus explicit calculations of the generator surge environment were
n
>
60 4; 6
not done.

TABLE 1

SURVEY O F GENERATOR PROTECTION


(Responses total 34)

1. Design done: -in house 14 1 -nameplate rating 2 -60 Hz factory test


-by consultants 20 3 -impulse withstand 4 -protective level, gapped unit
5 -protective level, zinc oxide 6 -60 Hz sparkover, " "

2. Surge transfer ratio of step-up 7 -60 Hz reseal, gapped unit 8 -voltage endurance, zinc oxide
transformer: -measured 3 9,lO -ac breakdown voltage 11,12 -dc breakdown voltage
-estimated 7 13,14 -impulse breakdown voltage
-unknown 23
Figure 2 Comparison of Designated Withstand, Breakdown
3. LV arresters are specified: Voltages, and Characteristics of Surge Protection
-usually 23
-sometimes 5
-never 2 Point 1 identifies the rated 60 Hz terminal voltage while 2 shows
-rarely 4 the test voltage of twice rated (line-to-line) plus 1 kV applied to
the new machine. An impulse withstand level widely used is 125 %
4. Surge capacitors are applied: of the peak 60 Hz test value [2] as shown by a line at 3 without a
-with arresters 14 precise range specified for the surge risetime. In Figure 4 of that
-alone 2 reference, a line is also used covering the range 0.2 to 10 p s . Note
that this is a designated withstand. The breakdown voltage would
usually be much higher, and might have a turn-up for the fastest
About half of the utilities surveyed use surge capacitors. In large fronts as is normal in solid insulation. On the other hand, a report
generators with single turn coils, it would seem unnecessary to by an IEEE Working Group [3] shows a reduced withstand for fast
slope back steep-fronted surges as required with multi-turn coils. fronts without clearly limiting the scope to machines with multi-
However these surges could conceivably be amplified internally in term coils. This probably explains the widespread use of surge
the winding due to reflections. Also gapped arresters tend to have capacitors found in the survey.
a higher protective level for steep-fronted surges, and capacitors
may be helpful in sloping back the surge risetimes. The follow-up
identified a move to eliminate PCB-filled capacitors, with some Arrester ratings are normally specified to just exceed the machine
utilities questioning the need to replace these units. line-to-line voltage. With rating increments of 3 kV one might
60 1

consider the limiting case to be a 27 kV arrester applied with a The inductive component is based on the HV arrester protective
24 k V machine. The associated protective levels on impulse for level times the transformer turns ratio, modified by the particular
gapped silicon carbide and gapless zinc oxide arresters are shown winding connection. A step source is assumed which produces an
by lines 4 and 5 respectively and are typically twice the crest oscillatory overshoot due to transformer leakage inductance and
arrester rating for a discharge current of IO kA. Particular capacitance to ground of the LV winding and of the isolated phase
arresters may have lower protective levels while any distance bus. Damping is associated with a shunt resistance taken as the
between the generator and arrester increases the level for fast generator surge impedance. Internal transformer winding
fronts. Also studies of the surge environment may show that lower oscillations at a higher frequency are thought to increase this
discharge currents can be assumed. component by up to 20 O/o. Finally the 60 Hz potential including
phase shifts due to the transformer connections is added.
T h e a r r e s t e r s a r e themselves v u l n e r a b l e t o 60 H z d y n a m i c
overvoltages. A gapped unit may spark over at 150 Yo of rating Figure 3 shows the results of low voltage measurements made at
(line 6) and reseal at 100 O/o of rating (line 7), which is only slightly Atikokan to check the validity of this procedure. The generator
above levels associated with single phase faults. A voltage breaker was closed and transformer HV phases B and C shorted to
endurance curve for zinc oxide is shown by line 8 indicating that ground. A mercury-wetted relay and battery were used to apply a
these arresters may fail on dynamic overvoltages which are lower 12 V step to HV phase A. The input or upper trace in the Figure
than those for gapped units. The high temperatures found in shows a spurious transient on the step front which can be ignored.
arrester cubicles may further depress curve 8 which is based on an The transformer phase A output or lower trace had an average
ambient temperature of 20' C. value of 0.325 V, damped frequency of 25 kHz and envelope time
constant of 250 ps.
Figure 2 also shows the measured breakdown voltage for coils of a
542 MW, 24 kV generator under 60 Hz ( 9 ) , dc ( 1 1 ) and impulse
(13) conditions. Points I O , 12 and 14 repeat these for the
groundwall insulation of a modern 11,000 HP, 6.6 kV induction
motor which was included in this project due to the small number
of machines available for destructive testing. It was assumed that
rescaling the breakdown voltage by 24/6.6 would give some
indication of the strength available in a 24 kV rated machine.
Note the large margin between the designated withstand and tested
breakdown voltage.

A significant number of utilities rely on the HV arrester at the


step-up transformer for surge protection of the generator. Since
HV systems are effectively grounded, this arrester can be given a
lower per unit rating than an L V arrester applied at the generator.
This permits an additional margin in protective level assuming that
the transfer ratio of transformers is not much higher for surges
than at 60 Hz. However, large transfers can occur at some
frequencies due to internal winding resonance as shown in section
4.2. Also for systems equipped with generator breakers, LV
arresters may be needed to limit surges associated with a stuck pole
and bus fault condition, as described in section 4.7.

The operating experience with respect to utility protection practices


has been good, according to the survey. Although several
respondents indicated generator insulation failures, only three were
for large turbo generators. Two of these involved the deterioration Figure 3 Step-up Transformer HV to LV Time Response,
of the insulation into a powder associated with partial discharges Loaded by Generator
with only a remote possibility that their failure was caused by
surges. The other failure was associated with HV to LV flashover
of the step-up transformer.

The following modeling parameters are available. The transformer


ratio was 239 Y/17.1 A kV, leakage inductance 390 pH, isolated
phase bus and transformer capacitance 17 nF, generator capacitance
4. GENERATOR SURGE ENVIRONMENT 0.38 pF, and subtransient reactance 460 pH, all quantities being per
phase. Frequency response measurements showed an average
The project was planned with an emphasis on modeling, rather generator surge impedance of 40 R at 20 to 30 kHz.
than surge monitoring since significant surges due to lightning,
switching or faults are relatively rare events. Also full scale staged The calculation method in [ l ] would predict an output transient of
testing would be impractical due to hazards to power equipment average value 0.492 V, which is overdamped (non oscillatory) with
and undesirable effects on the operating system. Ontario Hydro's a time constant of 3 ps. This is obviously not in good agreement
first unit to be equipped with a generator breaker, at Atikokan with Figure 3. At the measured frequency of 25 kHz, the modeling
TGS, was being commissioned a t this time and was available f o r of the impedance by a fixed resistor appears to warrant further
extensive low voltage measurements to validate modeling. The unit study.
is rated 270 MVA, 18 kV and supplies the 230 kV transmission
sytem in Northwestern Ontario near the Minnesota border. The M u c h b e t t e r a g r e e m e n t is a c h i e v e d if t h e g e n e r a t o r s u r g e
following sources of surges are under study. impedance is replaced by a model using subtransient inductance in
shunt with winding capacitance as recommended elsewhere[4].
4.1 Lightning The damping is now associated with the high frequency winding
resistance of the transformer and generator. Based on impedance
A steep-fronted surge arrives at one HV terminal of the step-up tests at frequencies between 20 and 30 kHz at Atikokan, this resis-
transformer and is limited to the arrester protective level. A tance in ohms was typically I O times the inductance in millihenries.
widely-used reference[ I ] suggests calculating the surge transferred This alternative model predicts an output transient of average value
to t h e g e n e r a t o r using i n d u c t i v e l y a n d c a p a c i t i v e l y c o u p l e d 0.334 V, damped frequency 28 kHz and time constant 200 ps, which
components as follows. is in good agreement with Figure 3.
602

Returning to the capacitively coupled component, this has usually If an oscillatory fault transient excites this resonance, the response
been based on the voltage divider formed by transformer HV to can be calculated using methods developed f o r the internal
LV capacitance and LV to ground capacitance including the oscillations of transformers[6]. From the width of the resonance at
isolated phase bus. The loading effect of the generator removes the 1.5 MHz in Figure 4, the ratio 6 of two consecutive peaks (same
capacitive component before the magnetic component becomes polarity) of the transformer's natural response is about 0.9. Now
significant. assume a ratio of 0.8 for consecutive peaks of the applied fault
transient. Calculations show that the generator would be stressed
Several further characteristics are under study. This includes the by a 1.5 MHz sinewave increasing to a maximum amplitude of
d i f f e r i n g loading effects of single and three phase banks, 140 % of that given by the turns ratio alone.
propagation and reflections on the isolated phase bus, and sloping
of very steep surges in the transformer. 4.3 HV Switching

4.2 &I& On opening, HV live tank breakers could flash over across a pole at
phase opposition, due to icing of the bushings or contamination.
Single phase faults on the isolated phase bus increase the 60 Hz The transient seen at the generator should resemble that of Figure 3
voltage on the other phases by 73 96. The associated transient but further studies are needed to resolve the effect of the 60 Hz
overshoot has been measured at Atikokan to be 80 % of the fault point on wave and HV system oscillations.
prestrike voltage, giving a peak exceeding 2.3 per unit. Studies will
be extended to cover self excitation including harmonics and faults 4.4 Ferroresonance
under overvoltage conditions.
In Ontario Hydro, ferroresonance may have been responsible for an
Voltage escalation, caused by multiple interruptions of an arcing arrester failure which occurred after the generator links were
fault at the 60 Hz current zeros and restrikes at the peak recovery opened during construction to allow energizing the unit service
voltages, is a classic scenario[5]. This appears to be the rationale transformer from the HV side. Generally ferroresonance can occur
for choosing the generator neutral resistor as low as the 60 Hz where the zero sequence capacitive reactance Xco and potential
capacitive reactance, in order to discharge the recovery voltage. To transformer magnetizing reactance X form the sole connection to
be consistent, a similar neutral resistance should be installed on the ground. This could also occur if t?;e neutral grounding resistor
transformer side of generator breakers, since this system is also was inadvertently left out of service. For typical configurations,
ungrounded before closing the breaker and with surge capacitors Xc,/X, may range from 0.001 to 0.01 . Figure 5, modified from
this system has a similar capacitive reactance. Of course the [7], shows how the boundary of ferroresonance is affected by this
transformer delta winding connection makes this impractical. Such ratio and the operating voltage. Electromagnetic Transients
a design intention may in any case be unnecessary due to the Program simulations are being conducted to define the overvoltage,
modest interrupting capability of an arc in air and the discharging and the resulting stress on LV arresters.
effect of the PTs on trapped charge due to saturation.

Faults within several spans of the step-up transformer create 2.01


travelling waves characterized by a decaying square wave of high
frequency. Winding resonances were measured at Atikokan using
an HP model 3570A network analyzer with a swept frequency
output of several volts. The transformer was loaded by the
generator. The ratio of LV to HV Phase A potential was recorded
with HV phase B and C left floating. As shown in Figure 4, an
amplification of six on the LV base was observed at the generator
terminals of the Atikokan step-up transformer at about 1.5 MHz.
This could be associated with the inductance of transformer
internal leads and winding capacitances. Similar resonances have
been noted on other transformers tested for the project.

0 180
0
-10 Figure 5 Regions of Potential Transformer Ferroresonance on
-180
Ungrounded Systems
-20

-30
4.5 Closing Surges
-40 The remaining surge sources discussed in section 4 are found only
on systems configured with generator breakers. On synchronization
such breakers usually close across insignificant pole voltages.
-50
However on black startup, energizing the step-up transformer
would cause a maximum of one per unit closing voltage on the first
-60 pole. Out-of-phase paralleling, for whatever reason, would double
this voltage.
-70
0 2M 4M 6M I 1 1 M Figure 6 illustrates the equivalent circuit which is used to find the
FREQUENCY ( Hz) closing surge on the second and third pole. Assume black start-up
with the first pole prestriking near peak generator voltage. This
results in a unit surge travelling to the machine and step-up
Figure 4 Step-up' Transformer HV to LV Frequency transformer, transferring through to the other two phases, and
Response, Loaded by Generator giving a second pole closing voltage of 1.5 pu. In addition, a
603

transient overswing due to L , C , C, and L, could be expected. due to 60 Hz overvoltages associated with self excitation, exciter
However, the overswing wa! f o i n d at Atikokan to be greatly lags and unit overspeed. In a related scenario, the system side bus
reduced due to the difference in transient frequencies on the might fault first, followed by a stuck pole on fault clearing.
generator and transformer side.
a C

t t
I +VI-

Figure 6 Equivalent Circuit for Generator Breaker Closing


Surge on the Second Pole
t
According to the survey, at least one unit uses surge capacitors on
the system side of the breaker. This has been recommended by
breaker manufacturers to control transient recovery voltages on Figure 7 Generator Phase A Surge Due to Stuck Pole B and
fault clearing. As a result, the frequencies on the system and System Fault on Phase C
generator side may be almost the same, causing the transients to
overlap. An overshoot of 1.8 per unit on the closing of pole one is
therefore added to the existing 60 Hz voltage on pole two of 0.5 4.8 Svstem Side Ferroresonance
per unit. A prestrike on pole two at the peak closing voltage would
result in a 2.3 per unit steep-fronted surge. The surge capacitors With the generator breaker open, the system side is not grounded
on the system side would direct most of this towards the generator. by the generator neutral resistor. Based on Figure 5 and the
There the impedance mismatch with the bus could be 3 to 1, comments in section 4.4, this part of the system could be
further increasing the level to 3.5 per unit. Studies show that the vulnerable to ferroresonance, especially if the relatively small
actual peak voltage reaches only 2.6 per unit due to the 60 Hz ground capacitance of the transformer and isolated phase bus are
p o i n t - o n - w a v e . F o r t h e remote possibility of o u t - o f - p h a s e augmented by surge capacitors.
switching, the surge could reach 5.2 per unit.

O n generator breaker closing, t h e switching transients may 5. INSULATION WITHSTAND TESTING


themselves resonate within the generator winding. Unusual system
conditions, such as black startup with line or cable capacitance on The first machine tested was a four-pole, 24 kV, 653.3 MVA, 542
the HV side of the step-up transformer will be reviewed as part of MW machine with 72 slots and two parallel paths per phase@].
the project. Each parallel has two groups of six coils. A coil is made up of two
Roebel bars with 19 hollow copper tubes insulated with one half-
4.6 Opening Surges lapped layer of 0.127 mm bituminous varnished glass tape. The
groundwall is mainly 6.1 mm-thick shellac-bonded micafolium in
Fault clearing transients are inherently limited to less than twice the slot section and bituminous mica tape in the end winding
peak line voltage, even if shunting resistors are not used. For section.
normal load switching, modest current chopping in the resistor
switch could also occur. Low voltage measurements were made at This generator had been in service at a nuclear plant as a base load
Atikokan GS in which currents of about 20 mA were chopped unit from 1972 to 1985, with no known insulation problem. There
using a small vacuum switch. Assuming a maximum breaker had however been ftequent leaks in the stator cooling system,
chopping level of 20 A, these results show the transient voltage is which required replacement of the entire winding.
limited to only 2.5 kV.
A 49-kV one minute ac hipot was applied to each of five parallels.
For unit rejection with a long line or H V shunt capacitor bank left An internal breakdown occurred in parallel Cla-C2a at 44 kV.
connected, self excitation tends to increase the 60 Hz potential. Higher ac test voltages were impractical due to intense partial
Tripping of the generator breaker could lead to large capacitive discharge activity.
recovery voltages and the possibility of restrikes. Electromagnetic
Transients Program simulations are in progress. Twenty coils were tested for dc breakdown ( 1 kV/s rate of rise)
and for six of these coils external flashover at the ends could not be
4.7 Breaker Failure prevented. Results appear in Table 2 with the coil numbering
beginning at the neutral end. Little correlation is evident between
A stuck pole on opening could leave the generator and system tied breakdown voltage and the variations in 60 Hz stress normally
on one phase as they slip out of synchronism. The voltages to present on different coils. Also note the consistency of the results,
ground would remain balanced on the generator side due to the with a mean of 116 kV and standard deviation 8 kV. Three
neutral resistor and large winding capacitance. Accordingly, the complete groups were also tested with breakdowns at 110, 112, and
system side voltage would reach about 2.6 per unit on two phases as 115 kV.
shown in Figure 7. If this should result in a fault, the generator
terminal A would then be stressed to 3.46 per unit at 60 Hz. If The impulse strength of the groundwall for 1/50 p s impulses was
LV arresters were used, they might fail explosively. Otherwise, a measured for 18 intact coils (with ends floating). Aside from two
surge of 5.5 per unit would stress the generator, assuming an external flashovers at 100 kV, all the coils in the three groups
overswing factor of 1.8. These voltages would be further increased tested withstood 125 kV, which was the limit of the instrument.
604

TABLE 2 The ratio of internal winding to terminal voltages was also


measured for the 542 MW generator over the frequency range of
GENERATOR COIL BREAKDOWN VOLTAGES (BDV) IN K V 100 Hz to IO MHz at low voltage. As shown in Figure 8, the
largest ratio (12:l) was found at 30 kHz at the neutral end of the
Coil BDV Coil BDV Coil BDV Coil BDV winding when this was open-circuited. For the neutral end
grounded, the maximum ratio (2.5:l) occurred at the winding
Ala-I 120 A2-8 135 B1-1 120 B1-6 110 midpoint at 60 kHz. These connections simulate ground and
Ala-3 125 A2-10 110 BI-2 110 C2-11 110 differential mode surges respectively. From [6] and the width of
Ala-5 120 A2-ll 110 BI-3 110 the resonance at 30 kHz in Figure 8, rhe ratio 6 of consecutive
Aia-6 115 A2-12 105 BI-5 120 peaks for the natural response of the generator is about 0.7.
Assuming an applied ground mode transient of 30 kHz with a ratio
of consecutive peaks also of 0.7, gives an amplification of over 4
The bars were examined after removal from the core. The semi- per unit. Thus the effective withstand of this generator is
conducting outer surfaces and micafolium groundwall were in proportionately reduced at this frequency.
excellent condition, with no significant delamination or other signs
of age. The conductor bundle was well consolidated. Only very
minor and scattered signs of partial discharge activity (small m-- 180
particles of white powder) were found at the bundle/groundwall
interface.

The second machine tested was a primary heat transport pump '*P-1'
10
0
-180

'H
motor rated 6.6 kV, 1 1 000 HP from a nuclear station. Each phase
of this four-pole delta wound motor has four parallels with six
coils per parallel and twelve two-strand turns per coil. The motor
was in service for 56 000 hours and had not failed although the
reliability of the inter-turn insulation was suspect. It was assumed
that the groundwall insulation would be similar to that of a
generator.

Eleven coils were subjected to a ramp ac voltage (5 kV/s) until the


groundwall insulation broke down. External flashover could not be
prevented in one case. The results appear in Table 3 with a mean
of 37 kV rms, and standard deviation 4 kV. Again the results are
rather consistent with no apparent correlation with normal 60 Hz 0 200 400 600 800 1000
stress distribution along the winding. FREQUENCY ( k H z )

Figure 8 Voltage Amplification at Generator Neutral End


TABLE 3 for Sinewave Excitation

MOTOR AC BREAKDOWN IN KV. RMS

Coil BDV Coil BDV Coil BDV Coil BDV


6. DISCUSSION OF RESULTS
a3-5 41 a4-3 29 b3-5 37 c3-3 31
a3-6 37 a4-4 45 b3-6 37 From the survey, failure of generator insulation appears to be rare.
a4-2 39 b3-4 36 c3-2 35 It is to be hoped that utilities will remain vigilant, although the
lack of standardization is not comforting. The dependence on HV
arresters, where the surge transfer ratio of the transformer is
Nine coils were dc ramp tested (2 kV/s) with results shown in unknown, is a cause for concern. The relatively wide acceptance
Table 4. The mean is 66 kV and standard deviation 11 kV. Finally of surge capacitors for machines with single turn coils is puzzling.
17 coils were impulse tested with 1/40 ps surges applied to one end
of the floating coils. The results shown in Table 5 have a mean of A relatively fast transition from silicon carbide to zinc oxide
112 kV and standard deviation of only 8 kV. arresters is in progress, and these exhibit different characteristics
with respect to protective level and dynamic overvoltage capability.
Protection must also be expanded to cover detuning against internal
TABLE 4 winding resonances.

MOTOR DC BREAKDOWN IN K V T h e g e n e r a t o r s u r g e e n v i r o n m e n t w i t h respect to l i g h t n i n g ,


switching, faults and ferroresonance has been studied using models
Coil BDV Coil BDV Coil BDV validated by low voltage testing at a recently-commissioned unit
equipped with a generator breaker. Full scale testing of most of
a4-1 80 b4-5 59 c4-4 72 these surges would be impractical due to hazards to equipment or
b4-1 45 b4-6 65 ~4-5 58 to a low probability of their occurance.
b4-2 15 ~4-2 75 ~4-6 65
Generator breakers have made a fairly large penetration into the
utility industry in the United States. Where LV arresters are not
TABLE 5 used, a stuck pole and bus fault can lead to voltages exceeding 5.5
per unit.
MOTOR IMPULSE BREAKDOWN IN K V
Steep-fronted surges due to breaker closing under out-of-phase
Coil BDV Coil BDV Coil BDV Coil BDV conditions could exceed 5 per unit. Due to propagation effects,
arresters would have to be installed within several metres of the
a2-1 110 a4-6 120 b4-4 120 c4-3 110 generator to be effective.
a2-2 120 b3-1 110 c2-1 110
a2-3 120 b3-2 110 c3-1 100 The network on the system side of a n open breaker appears to be
a4-5 90 b4-3 110 c4-1 110 vulnerable to PT ferro-resonance, especially where surge capacitors
605

are used. Generator breaker restriking under conditions of unit related to the failure o f ari-esters or capacitoi-s would be especially
harmonic resonance requires checking, as do multiple re-ignitions helpful.
on dropping an unloaded step-up transformer.

Lightning surges inductively transferred through the step-up REFERENCES


transformer should he calculated using a shunt L. C model for the
generator. Further study is needed to show if the distance between I. Jackson, D.W. "Surge Protection of Rotating Machines",
the arrester and generator is significant for the capacitive l E E E T u t o r i a l 'Surge Protection i n Power Systems',
component. Measurements on the Atikokan transformer show an 79-90- I 1 I p 90.
internal resonance, possibly related to the inductance of connecting
leads, which increases the capacitive component and can amplify 2. AlEE Committee Report. "lmpulce Testing of Rotating A C
HV oscillatory surges. Other tests not detailed here support the Machines", AIEE Trans V 79 (June 1960). p 182.
existance of this mechanism.
3. IEEE \\'orking Group "Impulse Voltage Strength of A C
The 60 Hz breakdown voltage of one parallel of the 542 MW Rotating Machines", Report on IEEE P792. I E E E PAS
generator was slightly below the factory test level. There was some Vol 100 (1981) N o 8 (Aug) p 4041.
evidence that the failed coil had been previously repaired. The 60
Hz breakdown voltage for the motor groundwall was over twice the 4. I\hire, E . L . "Surge T r a n s f e r e n c e Chni-acteristics o f
factory test level. Large margins between these breakdown Generator - Transformer Installation". IEE Proc V o l I 16
voltages a n d withstand levels a r e probably normal in most (1969) No 4 (Apr).
machines.
5. Westing house "Tra n s ni iss io 11 a nd D is t rib u t io 11 IHa n d boo k "
The impulse to 60 Hz withstand ratio of the generator and motor Pittsburgh, 1964, Chapter 14. Section 10. Subsection
exceeded 180 Yn and 210 O h respectively while the widely used value "Analog Computer Studies". p 5 19..
for coordination is only 125 ?IO.This and the above margin result
in a f a i r l y conservative coordination between L V a r r e s t e r 6. ClGRE Working Group 12.07 "Resonance Behaviour o f
protective levels and generator breakdown voltage. However the High-Voltage Transformzrs". Paper 12- 14, I 9 8 4 Session.
voltage endurance capability of arresters does not appear to have Figure 2 and 3.
the same conservative margin with respect to the temporary 60 Hz
overvoltages. This increases the risk to the generator of two or 7. Peterson, "Transients in Power Systems", IViley, N e w
three arresters failing simultclneously and faulting the machine. 'fork, 1951, Figure 10.3, p 268..

The impulse to dc breakdown voltage ratio for the motor was 8. Gupta, B.K. et a1 "Destructibe Tests on a 542 M W
170 Yo. Thus a dc test to give an indication of the impulse strength, G e n e r a t o r Winding", I E E E Symposium on Electrical
as previously recommended[9], would be overly conservative. J ns Ula t ion, Wash in g to n I 986, 86CH 2 1 96 - 4 - DE I , p2 85.

Readers are invited to comment on the above results and proposed 9. Moses, G.L. "lmpulse Levels of Large IIigh Voltage
study areas before the end of the project. Practical experiences Generators", IEEE PAS Vol 84 (1965) N o I I (Nov) p 1007.

Discussion equal and opposite 2.6 pu surge would he injected at the ground end of the
generator neutral grounding equipmcnt. Depending upon the characteristics
of this equipment, thcrc may or may not be significant delay before the
David W. Jackson (R. W . Beck and Associates. Waltham, MA): Section surge reaches the generator neutral. When it docs. it will divide, and one-
4.1 deals primarily with the inductively coupled component of lightning third will pass toward generator phase A terminal. Whcn thc surge reachea
surge transfer through the stepup transformer. The paper demonstrate,s the the phase A open terminal, it would tend to double. The voltage between
inaccuracy of the conventional modeling of the generator as a surge system grounded phase c and generator open terminal A will bc the 3.46 pu
impedance, and suggests a model including inductances and capacitances of 60-Hz voltage plus and minus the instantaneous surges within the generator
bus, transformer, and generator using leakage and subtransient reactances. winding reflecting between ground, neutral, and phase A . Without internal
The paper shows a good correlation between measurements and this winding rcsnnances, the maximum these could reach would bc twice one-
mathematical model for the Atikokan test generator. Are the authors able to third of 2.6 pu, and likely somewhat less. In such case, the stuck breaker
cite any other correlation studies which would support the adoption of this hazard should expose a generator winding to no more than 5.2 pu voltage.
model in general use for generator protection studies? and probably lesa than 5 . 0 pu. This is equivalent to 6. I pu on the common
Section 4.2 points out the variability of atepup transformer surge voltage per unit voltage base for \urge stress. A 24-kV machine has a 4.4 pu surgc
transfer ratio with frequency. An important possibility is surge amplitica- withstand strength per references [2] and 131 using a I .25 factor. and 5.3 pu
tion when the system is disturbed by surge fronts which excite the system at withstand if a 1.5 lactor is recognized. On the other hand, if internal
resonant frequencies. This i5 a phenomenon which is largely unknown and winding resonances were excited and were to result in a four-fold
unrecognized among surge protection cngineers. The authors are to be amplification as the authors suggest, then terminal A insulation could he
commended for publicizing this concern. and for noting the appropriate stressed to 3.46 plus 413 of 2.6 pu, or 6.9 pu on the authors' base and 8.5 pu
measurement and calculation techniques to quantify the hazard. on the common surge strew base.
In Sections 4.2, 4.5, and 4.7, the paper speaks generally of voltage The presence of capacitors on single-turn machineb may not be logical.
magnitudes in per unit. and it may be deduced that the base voltage in all but it is not purzling!. In thc course of a survey on failures of machine
cases is the rms line-line machine rated voltage. It is common in protective capacitors and arresters among utility users for PES Surge
considering surge voltage stress on machine insulation to use a base voltage Protective Devices Committee. the most cniiinion reason given for the
which is the crest of the machine line-neutral voltage. or 0.8165 of the rms presence of protective devices was that they were recommended by thc
line-line voltage. It would be helpful if the authors would confirm the manufacturer. Published instances of manufacturers' guidelines exist which
voltage base used. This is particularly important in view of the fact that the when paraphrased say-if i n doubt. put them on. A commonly held view is
current recommendations for machine impulse strength, reference 131 of the that capacitors may help. cannot hurt. and they are relatively cheap
paper, are stated on this crest voltage base. compared to a winding failure. Ignorancc and fear frequcntly rcsult in
In Section 4.7 and the "discussion of results,'' thc paper descrihes a 5.5 illogical expenditureh.
pu overvoltage stress due to stuck breaker pole and subsequent phase fault The survey of utility practices by the authors and their measurements of
to ground. The description of this possibility appears to be overly simplistic insulation breakdown strength and surge withstand strength lead to twn
and may be unduly optimistic or pessimistic. At the instant of the system conclusions which were not stated. The first is that the successful operation
side phase c fault, generator phase A would be 3.46 pu above ground. An of niost utility generators is primarily the result nt' good strong machine
606
insulation with plenty of impulse withstand margin, and cannot be attributed require more careful study than in the past, with an example being a fault on
to a strong body of knowledge or carefully applied protection practices. The an isolated phase bus leading to a unit load rejection and removal of
second conclusion is that the measurements on the two machines tested so excitation. Depending on the type of unit (hydraulic or thermal), load power
far support the thesis presented by Jackson in discussion on reference [3], factor, exciter characteristics, overspeed and capacitance left connected
that machine insulation surge withstand strength is higher than the 1.25 (say on HV cables or tie lines), the terminal voltage may rise by 115 or 120
guideline factor specifies, and a 1.5 factor may be sufficiently conservative percent above the root three neutral shift for several seconds. Arresters are
for general application. often applied with a rating equal to line-to-line voltage. They may be
located in an ambient of 45°C and have an initial element temperature of
Manuscript received February 25, 1987.
60°C (not overly conservative). Application guides for station class zinc-
oxide arresters are likely to show these conditions result in thermal
overloading. In fact, this may be acceptable for the arrester since system
Albert Mayer (BBC Brown Boveri & Company, Ltd., P.O. Box 8242, voltage has been removed. Note that gapped units with a power frequency
8050 Zurich, Switzerland) and Alfred Leibold (BBC Brown Boveri, Inc., sparkover of 150 percent rating have a comfortable margin to spare. Should
North Brunswick, NJ 08902): The review of generator surge protection is the contingency of this fault and breaker failure (leading to a long line left
highly interesting and meets fully European experience. Damage of large on the unit) be considered? These questions are not intended to discourage
generators and motors due to lightning or switching surges happens rarely the application of zinc-oxide units, but to highlight their differences.
but is reported from time to time. Each case was accompanied by Mr. Jackson’s discussion questioned whether the two equivalent circuits
considerable damage and very high cost. In the past, when only active gap of Section 4.1 had been compared in tests other than at Atikokan GS. No
arresters have been used, it was common practice to protect hydropower further direct results are available but parameter values for generator
generators by special designed arresters, while turbo generators remained capacitance C,, subtransient inductance L;, transformer and bus capaci-
on the generator-side often unprotected. This was due to the fact that tance Cb, and transformer leakage inductance L, are generally known.
hydropower generators, running at lower revolutions, have multilayer Calculations can be made if one assumes that the machine surge impedance
windings and are sensitive to interturn failures. Turbo generators have been Z, is given bym8, which was supported by measurements on several
protected only by surge capacitors and in some cases at the neutral point by other machines.
arresters. For large thermal units, the damping coefficient d from equation (8.6) of
This situation changes rapidly, due to the advantages of modern, gapless paper reference [ l ] tums out to be greater than one (overdamped) for any
metal-oxide arresters. There are especially three effects making this new situation which was checked. Smaller units show an overswing. This
approach very attractive. appears related to the behavior of C,, in the sense that it hardly changes
1) The protection level of MO arresters can be kept very low providing a with unit size, while the other parameters scale with base impedance. Now
good surge protection, compared with the insulation level of rotating the step responses of power system elements normally show less damping
machinery. This is possible because of the temporary overvoltage capability with an increase in power rating. On this basis, one might argue quite
of MO arresters. While spark gap arresters are limited in this respect and it generally that the equivalent circuit in that reference requires modification
was never allowed that the sparkover voltage was passed-not even for one- for use with large units. Ironically, while damping and frequency differ, the
half wave-the MO arresters can withstand for seconds a considerable two methods predict about the same peak voltage for large units. The lack of
temporary overvoltage and with MCOV protection level can be kept low. an overswing in the circuit of reference [ l ] is compensated by the lack of
2) The protection level of conventional spark gap arresters for very steep voltage divider action through L, and L[ .
fronted waves was about 30 percent higher than for lightning impulse The paper should have explicity stated that crest line-neutral voltage is
surges. Modem arresters have reduced this percentage to 10-15 percent. In the per unit base, which would be consistent with the sentence preceding
most cases, additiond surge capacitors can be avoided. They still can be Fig. 1.
helpful to reduce transferred surges on generator-transformer. Many power With respect to the stuck breaker pole and subsequent phase fault study,
stations use surge capacitors on the 1-V transformer terminals and MO Mr. Jackson’s discussion appears to be suggesting a lattice diagram
arresters at the generator terminals. approach, while the paper uses Pi models as in Fig. 6. This is similar to the
3) Modem MO arresters have a very stable protective characteristic. above question of whether the transformer loading should be modeled using
With the modem technology of MO it is possible to manufacture MO L;, C,, or Z,. Tests of the terminal-to-terminal step response of large
resistors, which show no ageing at all. In contrast to gapped arresters, they generators conducted for this project are dominated by a damped, single-
will show no change of protective characteristic even after frequent frequency transient in the 5- to 15-kHz range. There is little evidence of a
switching surges. step component which has propagated through the machine, presumably
Including these properties in the application of arresters to rotating because high-frequency losses in the laminations are significant. In fact, the
machinery remains a challenging engineering task. Arresters and their initial response seen on the far end of the winding includes a small negative
MCOV must be carefully selected. The protection scheme as well as load component which has been related to magnetic coupling between phases.
shedding factor of the generator musf be known. Internal winding resonances are being considered in the project, but are
Special high-duty arrester has been used in a large number of stations usually significant only when the surge source itself is a damped sinusoid,
with excellent field experience. It is’recommended that the Surge Protection which is not the case in this study.
Devices Committee investigate this further for revision of the recommenda- The presence of capacitors on single-tum machines may in part be
tion for generator and large motor protection. logical, if the turn-up in protective level of arresters at fast rise times is
considered. Perhaps what we found puzzling was that none of the
Manuscript received March 2, 1987. respondents to the follow-up mentioned this rationale.
Although this paper does not include conclusions, the discussion of
results did note the measured impulse to 60 Hz withstand ratios of 180 and
E. P. Dick, H. Dhirani, B. K. Gupta, P. W. Powell, R. A . Shinn, A. 210 percent. More recent measurements on 48 coils in an 80-MW, 13,8-kV
Greenwood, and J. W. Porter: The authors thank the discussers for their hydraulic machine showed a ratio of only 1.31, which would be further
comments. The discussion by Messrs. Mayer and Leibold points out several reduced if several anomalously low 60-Hz withstands were discarded from
areas where metal-oxide arresters should have superior performance to the averaging. Thus caution in conclusion would be prudent!
gapped units. As a word of caution, modest temporary overvoltages may Manuscript received April 28, 1987.

You might also like