Influencia de Altitud en Aislamiento de LT

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS VOL. PAS-86, NO. 8 AUGUST 1967

Influence of Air Density on Electrical Strengthof Transmission Line Insulation

T. A. PHILLIPS, SENIOR MEMBER, IEEE, LAWRENCE M. ROBERTSON, FELLOW, IEEE,ALBERT F. ROHLFS, FELLOW, IEEE, AND RICHARD L. THOMPSON, MEMBER, IEEE

Abstract-Flashover tests were conducted at the Leadville EHVTest Facility of the Public Service Company of Colorado and ProjectEHV at Pittsfield, Mass., to obtain comparative data to evaluate theinfluence of air density on switching surge and impulse flashoverstrength of transmission line insulation. The tests showed that nega-tive polarity flashover voltages are greater than those of positivepolarity. Therefore, positive polarity will be the critical requirementfor transmission line design for switching surge duty.

Air density corrections ranging from (RAD)0-6 to (RAD)1.0 werefound for positive polarity switching surge voltages as a function ofinsulation length and configuration. For impulse voltages, a full airdensity correction is required for both polarities under all conditions,with the exception of negative polarity where, with a massive prox-imity effect, no air density correction is required.

INTRODUCTIONHE ELECTRICAL strength of air and exposed porcelainT insulation systems is affected by variations in barometric

pressure, temperature, and humidity. To assure a desired with-stand voltage in service, the critical flashover voltage (CFO) ob-tained for standard sea level conditions must be increased bycorrection factors pertinent to the service conditions. The in-fluence of these factors exerts a profound effect on transmissionline economics, as has been evidenced on lines designed for high-altitude application.1E1The degree to which correction factors can influence design

requirements is illustrated by the following equation:CFOstandard = withstandservice X K1 X K2 X K3 X K4 X K5

conditions conditions

whereK1 ratio of CFO to withstand voltageK2 air density correctionK3 temperature correction (independent of air density)K4 humidity correctionK5 ratio of dry CFO to wet CFO.

A considerable range of uncertainty exists for these correctionfactors. In particular, the most significant correction for high-altitude applications is the product K1 X K2, which at 5000 feetmay be as great as 1.30. This product is uncertain for severalreasons.

1) K2 is calculated as 1/RAD-, where RAD is the relative airdensity.

RAD = 17.9 B (inches Hg)459 + T (OF)

Paper 31 TP 66-511, recommended and approved by the Trans-mission and Distribution Committee of the IEEE Power Group forpresentation at the IEEE Summer Power Meeting, New Orleans,La., July 10-15, 1966. Manuscript received June 15, 1966; madeavailable for publication March 21, 1967.

T. A. Phillips is with the Arizona Public Service Company,Phoenix, Ariz.

L. M. Robertson is with the Public Service Company of Colorado,Denver, Col.A. F. Rohlfs and R. L. Thompson are with the General Electric

Company, Pittsfield, Mass., and Schenectady, N. Y., respectively.

Fig. 1. Leadville, Colo.: configuration 1 shows transportableimpulse generator on left, suspension insulator hanging from phasewire in center, and V tower on right.

The exponent n is not certain and may vary with insulationlength,[1 tower configuration, and precipitation.

2) The standard deviation of the flashover voltage (and thusthe ratio of CFO to withstand) also may vary with insulationlength, tower configuration, precipitation, and altitude.

In 1965 General Electric Company undertook a research pro-gram for Arizona Public Service Company and Public Service ofColorado to determine the air density correction factor for trans-mission line insulation. Also participating were Southern Cali-fornia Edison Company, Pacific Gas and Electric Company, andthe Edison Electric Institute.

TEST PLANTest Sites

Consideration had been given to the use of pressure chambersto study the effect of air density, but the size and mechanicaldesign requirements of a chamber that would permit tests onstrings suitable for EHV systems made this approach impractical.

It was concluded that this question would best be answered bytesting identical insulation configurations at both low and highaltitudes. The Leadville Test Site of Public Service Company ofColorado was chosen as the high-altitude site (10 500 feet) sincemany of the necessary facilities were already in place; the EHVProject at Pittsfield was the low-altitude test site. Tests wereplanned for both warm and cold weather in Pittsfield in order toevaluate a possible independent effect of temperature.Test EquipmentAt Leadville, a transportable impulse generator, Fig. 1, built

by the General Electric Company High Voltage Laboratory, wasused to perform the tests. The generator is rated 3600 kV with200, 100, or 60 kilowatt-second connections. At Pittsfield thetests were performed at the EHV Project using the 3000-kV,250/125 kilowatt-second impulse generator.At both locations, the circuit constants were adjusted to

produce similar impulse and switching surge waveshapes.Oscillograms of the waveshapes are reproduced in Figs. 2 and 3.For the impulse wave the front time was measured according toASA C68.1 using twice the time between the 30 and 90 percentpoints. The impulse waveshape for the Leadville tests was 2.0 by

PHILLIPS ET AL.: AIR DENSITY AND TRANSMISSION LINE INSULATION

(a)

(b)

Fig. 2. Leadville, waveshapes: (a) impulse wave 2 X 45 js;(b) switching surge wave 260 X 3600 ,us.

45 and for the Pittsfield tests it was 2.5 by 48 microseconds. Thefront of the impulse wave was longer at Pittsfield because thegenerator was farther away from the test specimen than at Lead-ville. Consequently the circuit had greater inductance which re-quired greater series resistance to damp the oscillations at crest.This in turn produced the longer front. The difference in frontsdid not affect the CFO voltages, however, since the flashoversall occurred well out on the tail of the wave. There is no standardfor switching surge testing so in this case the time from zeroto the actual crest was used. The switching surge waveshape forthe Leadville tests was 260 by 3600 microseconds and for thePittsfield tests it was 240 by 3200 microseconds.

Scope of TestsThe flashover strength of transmission line insulation is sub-

stantially influenced by the mass of the tower and by the prox-imity of the insulation to the structural members of the tower.Because of these mass and proximity effects it would have beennecessary to examine a range of tower designs at both Leadvilleand Pittsfield. The expense and time involved in such a programled to a decision to test basic configurations representing bound-ary conditions which would encompass most tower designs.These tests were intended to determine the effect of air densityindependent of specific tower designs, that is, to define relation-ships which would permit the use of existing full-scale tower testdata for high-altitude applications.Tcst ConfigurationsThree basic test configurations were selected to include the

following variables:1) length of insulator assembly2) length of air gap3) influence of nearby ground plane4) wet and dry conditions5) switching surge and impulse waveshapes.Configuration 1, shown in Fig. 4, represents a suspension in-

sulator string with a minimum proximity effect. Configuration 3,shown in Fig. 6, represents a suspension insulator string with alarge proximity effect caused by the ground plane. Configuration2, shown in Fig. 5, provides a comparison of air vs. porcelainand represents a moderate proximity effect. In addition, cap andpin substation insulator assemblies were tested placed directly onthe ground (configuration 4) and on an 8-foot pedestal (configura-tion 5) shown in Fig. 7.A range of insulation lengths typical of 230- to 500-kV designs

was tested for configurations 1, 2, and 3 at Leadville and Pitts-

(b)Fig. 3. Pittsfield, waveshapes: (a) impulse wave 2.5 X 48us;(b) switching surge wave 240 X 3200,us.

field in the summer and fall of 1965. Winter tests of these sameconfigurations were conducted at Pittsfield in early 1966 to in-vestigate whether temperature has an effect in addition to thataccounted for by the relative air density. Tests of configurations4 and 5 were conducted at Leadville and during the winter testseries at Pittsfield.Leadville Test ArrangementThe Leadville test site is illustrated in Fig. 1. A plan view of the

area is given in Fig. 8. The test specimens, in the cases ofconfigurations 1, 2, and 3, were suspended from one ofthe outside phase conductors of the EHV test line so thatthe spacing from the test specimens to the V-tower was 55 feet.This spacing was selected to minimize any proximity effect thatthe tower might exert on the flashover voltage of the test speci-men. Earlier investigations have established that a clearanceratio (air strike to tower divided by insulation length) of two ormore will minimize the influence of the tower on the flashoverstrength of the insulation. The location of the impulse generatoris also shown.For the tests on configurations 3, 4, and 5 a one-inch wire mesh

ground plane (36 feet by 36 feet) was placed directly on theground under the test specimen. This ground plane was used atboth Leadville and Pittsfield in order to offset any difference inthe resistivity of the earth that might exist between the two testsites. The slope of the land under the test specimen was fivedegrees.Pittsfield Test ArrangementThe Pittsfield test site is shown in Fig. 9. A plan view of the

area is given in Fig. 10. Here a steel cable was strung betweenTower 2 of the EHV Project and the Quebec Hydro test tower.The test specimens were suspended from the midpoint of thecable so that they were 56 feet from either tower, thus minimizingthe proximity effect of the towers.At Pittsfield the land was fairly level; therefore, in the cases of

configurations 3, 4, and 5 the wire mesh ground plane was sup-ported on wooden frames to produce a sloping ground planesimilar to that which existed at Leadville.

TESTING PROCEDURESTest RunsEach test run consisted of a critical flashover determination

using 20 shots per level. The voltage levels were changed inapproximately four percent steps until at the lowest level therewere not more than five flashovers and at the highest level therewere not less than 15 flashovers.

949

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS AUGUST 1967

ENSION CONDUCTORCH CABLE BLOCK

-x 10 SUSPENSION INSULATORUNDER TEST

Fig. 4. Assembly details for configuration 1. Fig. 5. Assembly details for configuration 2. Fig. 6. Assembly details for configuration 3.

LOAD I

Fig. 8. Plan view of Leadville test site.

Fig. 7. Assembly details for configuration 5. (Configuration 4the same as configuration 5, except steel base was omitted and in-sulators set directly on ground.)

Fig. 9. Pittsfield, Mass., fall tests. Configuration 2 shows controltrailer in lower left and vertical rod gap at right. Bubble whichhouses impulse generator is in background. Tower 2 of EHVproject straddles the bubble. In lower portion is mock-up groundplane to represent slope of land at Leadville. Fig. 10. Plan view of Pittsfield EHV test site.

950


Z 1.

8

:r

-o 0.2 0. 0VAPOR PRESSURE 'HG

Fig. 11. Humidity correction factors:(A) switching surge, (B) impulse,negative, (C) impulse, positive.

The voltage of a level was determined by averaging the crestvalues of the full waves obtained at that level. The probability ofwithstand was calculated for each level and plotted on normalprobability paper. The CFO and the standard deviation were de-termined by drawing the best fitting straight line through thepoints and reading the voltage values at the 50 and 84 percentprobabilities. The voltage at 50 percent probability is the CFO.The difference between the two readings is the standard deviation.In fitting the straight line to the data points, the points nearerto the 50 percent probability were given more weight than thosefarther away. In order that the standard deviations for differentinsulation strengths could be compared, they were converted topercent of CFO.

Weather ObservationsWeather observations were made at the beginning and end of

each test run and in-between if conditions changed. The followingfactors were recorded:

a) barometerb) dry bulb temperaturec) wet bulb temperatured) sune) rain

Humidity Correction

f) snowg) fogh) windi) dewg) frost.

The test data were corrected for humidity in accordance withthe curves shown in Fig. 11. For impulse voltages, the standardcorrection curves from ASA C68.1, curves B and C were used.However, for switching-surge voltages these standard correctionshave been found to be too large. Curve A represents a humiditycorrection curve developed by General Electric High VoltageLaboratory from recent flashover test programs. This curve is thebest available at the moment but may require further refinement.

Wet Testing ProcedureThe wet tests were made using a precipitation rate of 0.2 inches

per minute. In the case of configuration 1 the rain spray headerwas supported in a horizontal position above the insulators andthe spray nozzles were directed so that the spray would be dis-tributed over the length of the insulator string. For configuration2, the header was placed in a vertical position and the nozzleswere directed so that the spray would cover the gap. Because ofbreezes the spray could not be controlled as well as in an indoortest facility and consequently the test results might be moreerratic.The water available at Leadville had low resistivity. There-

fore, it was processed through a demineralizer, brought up to ahigh resistivity, and then mixed with the local unprocessed waterto achieve a resistivity of 18 000 to 22 000 ohm centimeters. AtPittsfield the resistivity of the water was 22 000 ohm centimeters.

DATA ANALYSIS

Switching Surge ResultsThe switching surge data are plotted in Figs. 12 through 20.

The CFO is plotted as a function of the number of insulator unitsor the length of the air gap. RAD existing at the time of eachtest is recorded adjacent to each data point.In order to study the relationship between CFO and RAD the

data for configurations 1, 2, and 3 have been replotted in Figs.25 through 27. The CFO values uncorrected (raw) and correctedfor humidity are plotted as a function of RAD for selected in-sulator and gap lengths. For example, in Fig. 25, which is forconfiguration 1 under dry conditions, there are two sets of curves.The one on the left shows raw (uncorrected for humidity) CFOvalues and the one on the right shows CFO values corrected forhumidity. In each set of curves the CFO data from Leadville,Pittsfield fall, and Pittsfield winter are plotted for 10, 20, and 25insulator units. Adjacent to each test point the average air tem-perature existing during the test run is recorded. Straight linesare drawn to average the data points. Vertical lines indicate bytheir length the variation of the individual test points from theaverage curve. These variations may be due to several factors:

1) Consistency of voltage measurements: To check the consist-ency of the voltage measurements, the test voltages were plottedagainst the input charging voltage of the impulse generator.Such plots should produce straight lines with essentially no scat-ter of data points. These plots demonstrated that the data wereconsistent and that accurate comparisons could be made.

2) Statistical variation in CFO voltages: Based on probit anal-yses of selected tests, the 95 percent confidence limits for theCFO may typically encompass a 43 percent variation. Usingthe X2 statistical distribution, an analysis of the accuracy ob-tainable with 20 shots per voltage level verified this +3 percentrange. Therefore, the CFO obtained from a given test representsan estimate of the actual CFO, which with 95 percent certaintylies within a 6 percent band.

3) Humidity correction factor: Comparison of the raw sets ofcurves with the corrected sets will show whether correcting forhumidity improves the fit of the data for the plots of CFO volt-age vs. RAD. On positive polarity the variations are less forthe corrected data. On negative polarity, the corrected data andthe raw data produce about the same amount of variation and, ingeneral, the variations are considerably larger than for positivepolarity. Since all of the configurations have lower flashover volt-ages on positive polarity, it was concluded that the curves thathad been corrected for humidity should be used in the analysis ofthe effect of RAD.

4) Temperature effect (independent of RAD): As menitionedpreviously, the air temperatures existing during the test runs arelisted adjacent to the plotted points in Figs. 25 through 27.In general the test points obtained at the higher temperatures lieabove the

-average curves on positive polarity and below theaverage curves on negative polarity. Thus, it appears that theremay be a small temperature effect which is independent of RAD.However, this needs to be verified and evaluated by future testing.In most instances, particularly on positive polarity, the variationof the points from the average curve is less than i3 percent, thusfalling within the 95 percent confidence interval expected of thedata. (See item 2.) Therefore, it was decided to use the averagecurves to analyze the RAD vs. CFO relationship.The data were reduced to a common range of RADs so -that

one configuration could be compared to another configuration,one length to another length, etc., on a common basis. The rangeof RADs from 0.7 to 1.0 was chosen. The curves were extendedto RAD of 0.7. Voltage values were read off the curves atRADs of 0.7 and 1.0 and designated V0.7 and V1.0, respectively.Then the ratios V0.7/Vl.O were calculated. Curves, not repro-duced here, were plotted for configurations 4 and 5 (configura-tion 1. 2, 3 curves are shown in Figs. 25 through 27) in order to

1.16

1.12

951

'E


/.71cOO

___ - -,-172-

y99/.75

/ /.7 6

100 /

8o0o PITTSFIELD FALL TESTS/ + POS.

/.2 LEADVILLE TESTS40 7 PQs.

----WM2 04EG.

0S 1O 20 30 4.0NO. OF UNITS

Configuration 1: switchingsurge, dry.

2400 .

2001 - - - ___- .72

1600-- ///>--lOS'

1600 / .72-WET-50 oI!Pll 0T4 .71

.73

~~I2D- ___ f- ~~1.04 ~ .7-WL~~~.J99/ .~~~672-WET/~~ PITTS~FIELD FALL TESTS

600 .7 9DRY WET// -

~~~~+POS. 9 POsDx NEG.

6// 72 LEADVILLE TESTS400 / 0 / - ffl POS. A POS.l// ---0 MEG. NMEG.

0 4 6 10 16GAP LENGTH (FT)

Configuration 2: switchingsurge, dry and wet.

2400 / |___xNG

LOl III

1 0391600 - -

1000

PITTSFIELD WINTER TESTS-

400 -- N EG.

0 lo 20 30 41NO. OF UNITS


20-00 /Fig. 13.Cofiuraio1swt1.10;6Cc1--,surge Sooy.

/ PITTSFIELD WINTER TESTS+ POS.

400 NEG.

0 10 20 30 40NO. OF UNITS

Fig. 13. Configuration 1: switchingsurge, dry.

2000 - _ __ _ -

1#//J +1.1 111lW1e200

/ PITTSFIELD WINTER TESTS+ POD

MEG

4 8 12 1b

Fig. 16.GAP LENGTH (FT)


161.07 .7 I1600 , |2 THESE CURvtb wtKt/ EjDRAWN TO APPROXI-/ e -/ IMATE THE SHAPE OF-

I /|I Xl t THE CURVES FOR3 J / +1.08 CONFIGURATION 3

2~~~~ /'-~~~~~SEE FIGURES 176a10

>0 I/ PITTSFIELD WINTER TESTS

L_____ + POS./,

---~~~~~~N EG.

400- /4II1{st./ LEADVILLE TESTS

II'tZ j0 POS.| ^t |[email protected] 4 8 12 16

NG OF UNITS


I 1[1601

(I2C

o RoE

401

Fig. 14.

.73

A02

101 PITTSFIELD FALL TESTS1.01 r _ + POS.

/--_-_-___xMNEG.

LEADVILLE TESTSYPOS.

----0_ jNEG.

o0 20 30 40NO OF UNITS

Configuration 1: switchingsurge, wet.

1600.1 2 -- _1206

>160 .06 //I . 03-4 1.05/'1/ 1___x NEG.0412DC - - L LT

I06.060 .0 .72 NE

I r

0O. - NPITTSFIELD FALL TESTSL4

____+ POs&MEG.

LEADVILLE TESTS

-- N EG.

U, 0D 0 0 41,NO. OF UNITS


2400- - - - -

200--0 _ 7

1.03 /1600 ---- 1.071

- /->1 1.0

0) 60. __ 07,/.7 PITTSFIELD WINTER TESTS

+ PO0.// .,74 ' MEG.4 /

~~~LEADVILLE TESTS* POE.

--- MEG.

4 6 12 16NO. OF UNITS

20. Configuration 5: switchingslurge, dry.

952

Fig. 12.

Fig. 15.

b

Fig. 18. Fig.


1) .06

LLiA41:iv 74.74

A 11.04I100

o PITTSFIELD FALL TESTS

.2 ____ x NEG.0 LEADVILLE TESTS

& -i0POS.-

---Na NEG.

20 30NO. OF UNITS

1600 1I.'I-'--1.06

1200 - --

A72p PITTSFIELD FALL TESTS__ DEP- f POS.

400 / LEADVILLE TESTS* PUS.a NEG.

40

Fig. 21. Configuration 1: impulse, dry.

: 1.07+161 - --__ 1.086, ___ __

PITTSFIELD WINTER TESTS

400C-

0 I l20 30 40NO. OF UNITS


___0 __ RAW __

1600

40C12001 ___ + I_1000 ~ ~9,

RO 1 __+_

600. -- _ __

40 - -

8 12GAP LENGTH (FT)

953

I8

41.

40


I-09-1.D

l--

ooo ---~ - RAW

60,____25 UNITS

14001 - ___

1200 20N~1000

60

40 - - - - -

1.1 I.E 0.9 Uas 117RAD

Fig. 25. Configuration 1: switchingX negative).

1-

-

0

RAW

- t00 25 UNITSI6001400

1200

1063 -Boool

6006001 1

19-

I I- nq nR -1.1 R. uD uAS u.RAD

PITTSFIELD FALL TESTS72 ~+ p5S.

f, t/ D___ x NEE.10 3 - LEADVILLE TESTS

O, *_ t POS.A// ~----N DEE.X--.-20 30

NO OF UNITS


RAD

surge, dry (+ positive,

2131 154'

41 35-32'

25 -- 54!2T

45. 53.25e

26 4141E-

1.2 1.1 1.0 0.9RAD

8 5'a8 0.7

Fig. 26. Configuration 2: switching surge, dry(+ positive, X negative).

20

ID

IS

4-

12

.072/-10 1.04

1E00 - _ - 155 - - -_;; 12001 _ _ _ -L .2 _ _

hA

I

I

>

L5 CORR. FOR H25 UNITS

R.U U.90D 0.7RAD

20 O17 Im.72II

,:; I:

LD:x

l--iP-3

vu

I

I

I

k7 7

Fig. 27. Configuration 3: switching surge, dry(+ positive, X negative).


CONFIGURATION ISWITCHING SURGE

II - - - IMPULSE

0.1 RDDY - PUS. DR PS E

U.7 NEG. IIET POS. DRYINEG. DRY

O 10 20 30NO. OF UNITS

CONFIGURATION 2l.t _ SWITCHING SURGE

n e- r---IIMPULSENEG.DRY

,POS. DRY'POS. WET

V0.7 u.V1.0 0.0

IPOS. DRYv.v0 5 10

GAP LENGTH (FT)15

VO.7V1.

VO.7V I.0

Fig. 28. V0.7/Vl.o relationships.

CONFIGURATION 3 IMP NEG. DRY1.0

SS.. NEG. DRY

078

impOPot S P D0.7

IO 0 A0 30NO. OF UNITS

1.1CONFIGURATIONS 4 DiS

CONFIGURATION 40.8

S. S. POS. DRY

Po. DRYS. S. PO DRYICONFIGUR TION 50.7 _S.S. NEG. DRYJ

0.6 - 8NO. OF UNITS

Fig. 29. VO.7/Vl.o relationships.

determine V0.7/V1.0 ratios. The ratios for all configurations arelisted in Table I and are plotted in Figs. 28 and 29 to show howthe V0.7/V1.0 ratio varies with configuration, length, polarity,dry vs. wet, and switching surge and impulse waveshapes.

DISCUSSION OF SWITCHING SURGE RESULTSPositive Polarity: Configurations 1, 2, and 3

Referring to the positive polarity results in Table I, it is seenthat for the 10-unit insulation length, the V0.7/V1.0 ratio isapproximately equal to 0.70 for all except configuration 2 (fivefeet). The 0.59 ratio obtained for configuration 2 is not consistentwith the other data and appears to be unrealistic since it is un-reasonable for the ratio to fall below 0.70, corresponding to afull correction for air density. This low value is believed to be theresult of an unnoticed reduction in the air gap length at Leadvillecaused by increased sag in the span conductor due to temperaturechanges or to improper gap setting. This test was made at thebeginning of the test program and additional controls were im-posed during the remainder.As the insulation length increases, the ratio of V0.7/V1.0 in-

creases for all configurations. The ratio also increases with in-creased proximity. For configuration 1, small proximity, theratio at 25 units has increased only slightly to 0.72, but for con-figurations 2 and 3, moderate to large proximity, it is 0.76-0.80at 20 units and 0.81-0.83 at 25 units. The wet and dry compari-son obtained for configuration 1 shows that the VO.7/Vl.0 ratiois greater under wet conditions, but the difference is small and acheck point at the 25 unit length (12.5-foot air gap) for configura-tion 2 shows a slightly smaller ratio for wet conditions. Therefore,it would be conservative to assume no difference in wet and dryconditions.Negative Polarity: Configurations 1, 2, and 3The negative polarity data obtained for configurations 1 and 2

demonstrate a pattern similar to that observed for positivepolarity. The V0.7/Vl.o ratio increases for increasing insulationlength and proximity and is approximately the same forwet and dry conditions. However, for configuration 3 theVO.7/Vl.0 ratio is significantly greater for negative polarity, in-dicating almost no effect of air density on the longer insulationlengths. At 25 units, the V0.7/V1.0 ratio is 0.93.Configurations 4 and 5

Ratios of V0.7/V1.0 obtained for cap and pin insulation as-semblies are also shown in Table I. For positive polarity, these

results show good agreement with the results for line insulation.At an insulation length equivalent to 25 suspension units, theV0.7/Vl.0 ratio for configuration 4 is 0.83 compared to 0.81 forconfiguration 3. Configuration 5 and configuration 2 have similarratios up to an insulation length equivalent to 20 suspensionunits, but at a 25-unit length, air density has less influence on theair gap than on the cap and pin insulation. That is, the V0.7/V1.0ratio at a 25-unit length is 0.83 for configuration 2 and 0.77 forconfiguration 5. For negative polarity configuration 5 has anunusual pattern, with the influence of air density increasingwith increased length. The negative polarity VO.7/Vl.o ratio forconfiguration 5 is 0.79 for a 10-unit equivalent length and 0.71 fora 20-unit equivalent length.

Summary of Switching Surge ResultsFor positive polarity, there is a significant difference between

the air density effect for configuration 1 (small proximity)and configurations 2 and 3 (moderate to large proximity). A fullcorrection for air density would be appropriate for configuration 1for insulator string lengths up to 25 units. For configurations 2and 3, the air density effect is more a function of insulationlength with the ratio of V0.7/V1.0 ranging from 0.7 for short airgaps and insulator strings to 0.81-0.83 for a 12.5-foot gap or 25insulators. Configuration 4 is equivalent to configuration 3 foran equal insulation length, but configuration 5 is influenced moreby air density than configurations 2 and 3 at longer insulationlengths, with a V0.7/V1.0 ratio of only 0.77 at a 25-unit equivalentinsulation length (10 cap and pin units).For negative polarity, the effect of air density decreases signifi-

cantly with increasing proximity. With the exception of configura-tion 5 at negative polarity, the air density effect decreases as thelength of the insulation path increases.

DISCUSSION OF IMPULSE RESULTSTo study the relationship between CFO and RAD, the impulse

data were plotted in a manner similar to that used for the switch-ing surge data, as shown on Figs. 21-24. The resulting V0.7/V1.oratios are shown in Figs. 28 and 29 and summarized in Table II.The ratios of V0.7/V1.0 range from 0.68 to 0.76 for all configura-

tions and polarities except configuration 3 at negative polarity.In general, the air density effect decreases slightly with increasedinsulation length and is not influenced by proximity at positivepolarity. For practical purposes, a full correction for air densityshould be used for configurations 1 and 2 under all conditionsand for configuration 3 at positive polarity.

V0.7V1.0

954

.-V


TABLE IV0.7/V1.0 RATIOS-SWITCHING SURGE

Wet Positive Polarity Negative PolarityConfig- or Insulation Length, * units Insulation Length,* unitsuration Proximity Dry 10 15 20 25 10 15 20 25

1 small dry 0.68 0.68 0.68 0.72 0.65 0.67 0.68wet 0.70 0.72 0.74 0.76 0.70 0.71 0.72

2 moderate dry (0.59)t 0.76 0.83 0.72 0.75 0.78wet 0.81

3 large dry 0.69 0.74 0.80 0.81 0.88 0.89 0.91 0.934 large dry 0.835 moderate dry 0.73 0.74 0.75 0.77 0.79 0.75 0.71

* Insulation length is expressed in terms of an equivalent number of suspension insulator units for all configurations:Suspension units 10 15 20 25Air gap, ft 5 7.5 10 12.5Cap and pin units 4 6 8 10

t Gap spacing error suspected.

TABLE IIV0-7/V1l0 RATIOS-IMPULSE

Wet Positive Polarity, Negative Polarity,Config- or Insulation Length,* units Insulation Length,* unitsuration Proximity Dry 10 16 20 10 16 20 25

1 small dry 0.72 0.76 0.72 0.722 moderate dry 0.69 0.69 0.68 0.723 large dry 0.75 0.73 0.92 1.00 1.07

* Insulation length is expressed in terms of an equivalent number of suspension insulator units for all configurations:Suspension units 10 16 20 25Air gap, ft 5 8 10 12.5

For configuration 3 at negative polarity, the data indicate thatno correction for air density would be required; that is, the CFO atLeadville was approximately equal to the CFO at Pittsfield. Thevoltage meas urements for these tests have been investigatedand their accuracy confirmed. Furthermore, similar results wereobtained during both the fall and winter tests. The completeabsence of an air density effect in this case is most surprising andmerits further investigation.

STANDARD DEVIATION OF SWITCHING SURGEFLASHOVER DISTRIBUTION

The standard deviation was determined for each CFO tests asexplained earlier under testing procedures. Where several repeattests were made on one test condition, the standard deviationsometimes varied as much as 2 or 3 to 1. This indicates -(hat thestandard deviation for a single CFO should not be given muchweight, but that a number of them should be obtained and aver-aged in order to determine the proper standard deviation toemploy in comparing one test condition to another or in esti-mating the withstand voltage.Examination of the data obtained during this investigation

indicated that for any one configuration, the standard deviationwas essentially the same at all insulation lengths tested. There-fore, the standard deviations for all insulation lengths have beencombined for each configuration. Standard deviations are shownfor positive and negative polarity for Leadville and Pittsfieldtests in Table III.

TABLE IIISWITCHING SURGE STANDARD DEVIATIONS

Config- Number Standard Deviation*uration Location Polarity of Tests Average Range

I Leadville pos. 9 3.95 1.6-6.5neg. 10 1.91 0.6-3.0

1 Pittsfield pos. 7 4.48 2.5-6.0neg. 6 3.53 2.0-5.9

2 Leadville pos. 6 4.42 2.2-6.2neg. 4 3.85 2.3-5.8




4 Leadville pos. 1 3.14 Pittsfield pos. 1 4.3

neg. 1 3.65 Leadville pos. 3 3.06 1.5-4.0

neg. 3 2.30 1.7-2.85 Pittsfield pos. 3 5.63 4.0-6.7

neg. 4 3.67 2.1-4.8

* Expressed in percent of the CFO voltage.

955


The relationship of the standard deviations for various condi-tions were studied by means of the Student's t-tests. This is astatistical comparison method which determines for a specifiedlevel of confidence whether the means (in this case the averagestandard deviations) of two distributions are equlal or different.The results of these comparisons are summarized in Table IV.Where the averages were not found to be equal, the mean differ-ence and the 95 percent confidence interval estimate of the differ-ence in the averages are shown.

Referring to Table IV, it can be seen that no significant differ-ence was found in the standard deviations under wet and dryconditions. For this reason, wet and dry results were not shownseparately in Table III. The standard deviation is greater forpositive polarity than for negative polarity for all configurationsexcept configuration 3, in which the relationship is reversed.For configuration 3, the negative polarity standard deviation canbe assumed with 95 percent confidence to be at least two percentdifferent from the positive polarity standard deviation. It is ofinterest that for negative polarity this configuration displayed asignificantly smaller air density effect than for positive polarity orother configurations, thus indicating a basic difference in themechanism of flashover.The standard deviations at Leadville and Pittsfield were com-

pared for positive polarity and for negative polarity with andwithout configuration 3 data. In all cases the standard deviationat Leadville was smaller than at Pittsfield. For positive polarity,the mean difference between the Leadville and Pittsfield standarddeviations was 1.4 percent, and for negative polarity the meandifference was 2.6 percent. When the large negative polaritystandard deviations obtained for configuration 3 at Pittsfield areeliminated from the comparison, the mean difference for nega-

tive polarity is reduced to 1.1 percent.Comparing the three basic configurations, Table IV shows that

for positive polarity, the standard deviation is not a function ofproximity, since the standard deviations for configurations 1 and3 (small and large proximity effect) were found to be equal.However, the standard deviation for air gaps appears to beapproximately two percent above that for insulator strings, deter-mined by comparing configuration 2 (air gaps) with configura-tions 1 and 3 (insulators). For negative polarity, configuration 3has a significantly higher standard deviation than either con-figuration 1 or 2; based on the 95 percent confidence intervalestimates of the mean differences, the standard deviation forconfiguration 3 is at least three percent above that for configura-tion 1 and at least two percent above that for configuration 2. Thenegative polarity standard deviation for air gaps, configuration 2,is somewhat higher than for insulators without a large proximityeffect, configuration 1, with a mean difference of about one per-cent.

Standard Deviation of Impulse Flashover DistributionThe standard deviations of the impulse test data are summa-

rized by location and polarity for each configuration in Table V.The number of tests was not sufficient to permit a statisticalevaluation as was performed for the switching surge data.However, the variation is less and the standard deviations arereasonably consistent with the exception of configuration 3, nega-tive polarity, at Pittsfield. The average standard deviation forimpulse voltages was two percent and is independent of polarityand configuration, with the exception of configuration 3, negativepolarity, at Pittsfield, which had an average standard deviationof 5.8 percent.

TABLE IVt-TEST COMPARISONS OF SWITCHING SURGE STANDARD DEVIATIONS

Standard Deviations Result of t-test Mean 95% Confidence IntervalCompared Comparison Difference Estimate of Difference

Wet vs. dry-positive polarityConfiguration 1 and 2 o-wet = adry 0Wet vs. dry-negative polarityConfiguration 1 and 2 orwet = ordry 0Positive polarity vs. negative polarityConfiguration 1, 2, 4, 5 o-pos. > aneg. 1.9 1.2 - 2.6Positive polarity vs. negative polarityConfiguration 3 crpos.


Summary of Standard Deviation DataDifferences in the standard deviations for the Leadville and

Pittsfield data amounted to 1.4 percent on positive polarity and1.1 percent on negative polarity, Leadville always being lower.The difference for porcelain and air insulation is one to two per-cent. However, the lower limits of the 95 percent confidenceintervals of these differences are considerably less, and for prac-tical design purposes the use of the following standard deviationsis suggested:

Switching surge, positive polarityall configurations

Switching surge, negative polarityconfiguration 3all other configurations

Impulse, positive polarityall configurations

Impulse, negative polarityconfiguration 3all other configurations

5.0 percent

7.5 percent3.0 percent

2.0 percent

5.0 percent2.0 percent

In a recent paper concerning switching surge testing of theline insulation for the 500-kV Peace River project,181 the standarddeviation obtained for each test was reported. These have beenanalyzed and indicate that the standard deviation for trans-mission line insulation is five percent confirming the foregoingconclusions for positive polarity switching surges.

APPLICATION TO TRANSMISSION LINE DESIGNSwitching SurgeThe negative polarity flashover voltages were greater than the

positive polarity flashover voltages for all configurations tested.Therefore, for transmission line insulation design, positivepolarity flashover strength will be the critical requirement.The positive polarity switching surge CFO voltages, dry, ob-

tained for equivalent lengths of suspension insulators, air gap,and cap and pin insulators are compared in Table VI with dataobtained from full-scale tower tests at Project EHV in Pittsfield.The data shown are for 25 suspension insulators, ten cap and pininsulators, and a 12.5-foot air gap. The data have been correctedfor humidity.The data in Table VI indicate that with large proximity, sus-

TABLE VIMPULSE STANDARD DEVIATIONS

Config- Number Standard Deviation*uration Location Polarity of Tests Average Range

1 Leadvile pos. 2 2.0 1.7-2.2neg. 2 1.5 0.8-2.2






Total not includingConfiguration 3, Neg. Polarity,Pittsfield 25 2.0 0.8-3.9

* Expressed in percent of the CFO voltage.

pension insulators, and cap and pin insulators have approximatelyequal flashover strengths. With moderate proximity, the flashoverstrength of the cap and pin assembly is similar to that for the airgap. However, the air gap is affected somewhat less by air den-sity and thus has relatively greater strength at Leadville. It isapparent that the flashover voltage levels obtained at Pittsfieldfor configurations 2 (1550 kV) and 3 (1150 kV) bracket the rangeof flashover values for actual towers. Configuration 2 is at thehigh end of the range (consistent with moderate proximity) andconfiguration 3 is at the lower end of the range (consistent witha large proximity effect).

Table VI shows that for positive polarity, typical 500-kV in-sulation dimensions (25 insulators and 12.5-foot air gap) haveVO.7/Vl.o ratios of 0.81 or greater for both configurations 2 and 3.Therefore, it should be valid to use a V0.7/Vi.o ratio of 0.8 for thepractical range of 500-kV tower designs. A ratio of VG.7/V1.0 of 0.8is equivalent to an exponent n of 0.6. Tests of cap and pin insula-tion assemblies indicated that an air density correction of(RAD)0 7 would be required for a 10-unit assembly (equivalentto 25 suspension units). Therefore, a correction of (RAD)0.7should be conservative for 500-kV transmission line design andshould encompass substation applications. That is

T' = Vs(RADx)nwhere

VxRADXVsn

CFO at location xRelative air density at location xCFO at standard conditions (RAD = 1)0.7.

For an insulation length equivalent to 20 suspension units, thepositive polarity V0.7/V1.o ratios are 0.76 for configuration 2 and0.80 for configuration 3. Therefore, an air density correction of(RAD)0 8 would be appropriate for this length. A reliable V0.7/V1.0ratio for a 15-unit equivalent length was not obtained for con-figuration 2, but the ratio for configuration 3 is 0.74, indicatingthat an appropriate air density correction for this length wouldbe (RAD)0 9. As discussed earlier, a full air density correction,(RAD) 1 0, would be applicable for the 10-unit insulation length.In order to establish the relationship between the withstand

and CFO voltages, it is necessary to determine two things:1) standard deviation

TABLE VICOMPARATIVE POSITIVE POLARITY CFO VOLTAGESFOR TEST CONFIGURATIONS AND ACTUAL ToWERS*

Leadville, Pittsfield,Test Configuration kV kV

3 960 1165 fall1220 winter

4 1000 12505 1140 15002 1240 1520 fall

1550 winterFull-scale tower testsV-string (25-ft. window)

no corona rings 1280with corona rings 1160

Single stringtno corona rings 1360with corona rings 1300

Air gapconductor to leg or truss 1290conductor to guy wire 1550

* For 25 suspension insulators, 12.5-ft air gap, or 10 cap and pininsulators.

t Insulator string swung to a 300 angle. Clearance to towerequal to length of insulator string.

957


2) level of reliability required, and thus the number of standarddeviations that should be subtracted from the CFO voltage toestablish a realistic withstand voltage.The standard deviations were analyzed earlier. Limiting the

considerations to positive polarity, as was done here, it is appar-ent that a standard deviation of 5.0 percent should be used.

If the withstand voltage is considered to be two standarddeviations below the CFO voltage, the withstand voltage will be90 percent of the CFO. For three standard deviations below theCFO, the withstand voltage will be 85 percent of the CFO.Impulse

It is recommended that for impulse design consideration a fullair density correction, (RAD)1 0, should be used with a standarddeviation of 2.0 percent.

CONCLUSIONSSwitching Surge

1) The following air density correction factors are applicableto transmission line design:

Insulator orAir Gap Length, ft

5.07.510.012.5

CorrectionFactor

(RAD)' 0(RAD) 9(RAD)0.8(RAD)0.7

2) Air density correction is the same for wet and dry condi-tions.

3) Air density correction is approximately the same for air gapsand insulators.

4) Air density effect decreases with increasing length and in-creasing proximity of insulation to ground planes.

5) Negative polarity switching surges produce higher flashovervoltages than positive polarity. Therefore, positive polarityswitching surges determine the criteria to be followed in designingtransmission line insulation.

6) There seems to be a slight temperature effect that is in-dependent of RAD. However, this effect is small and will need tobe investigated further for verification.

7) The use of a humidity correction improved the correlationof the CFO data with RAD. Therefore, it is concluded that it isvalid to use a humidity correction.

8) A standard deviation of five percent is applicable to trans-mission line design.

9) The standard deviation of CFO voltages is essentially in-dependent of proximity on positive polarity. On negative polarityit increases with large proximity effects.

10) The standard deviation of CFO voltages is smaller at highaltitude than at low altitude.ImpulseThe air density correction that should be used for trans-

mission line design for impulse waves is (RAD) 1.0 and the stan-dard deviation is 2.0 percent of the CFO voltage.

ACKNOWLEDGMENTThe authors wish to acknowledge the contribution made by T.

Brownlee of the General Electric Company in supervising theLeadville portion of the program and the support he receivedfrom P. K. Jones of Public Service Company of Colorado and J.P. Benedict of the General Electric Company. The Pittsfield testswere supervised by F. J. Turner and performed by L. W. Grahamand T. W. Armstrong of the General Electric Company. Theingenuity of these people in coping with problems in the field andtheir close attention to details contributed much to the successof the program.

REFERENCES[1] W. R. Johnson, E. G. Lambert, J. B. Tice, and F. J. Turner,

"500-kY-line design: II-electrical strength of towers," IEEETrans. Power Apparatus and Systems, vol. 82, pp. 581-589, August1963.

[2] G. H. Aleksandrov, V. Y. Kizvetter, R. M. Rudakova, andA. N. Tushov, "The AC flashover voltages of long air-gaps andstrings of insulators," Elektrichestvo, pp. 27-32, May 1962.[3] H. S. Goff, D. G. McFarlane, and F. J. Turner, "Switchingsurge tests on Peace River transmission line insulation," IEEETrans. Power Apparatus and Systems, vol. PAS-85, pp. 601-613,June 1966.

Discussion

A. E. Kilgour (Allis-Chalmers, Milwaukee, Wis.): This paper holdsconsiderable interest and the authors are to be commended for theirfine presentation. It is encouraging that the comparative informationpresented is in generally close agreement. The explanations, pro-cedures, and graphic display will certainly be of value for futuredesign and operational purposes.

Several questions have been generated from reading this work.1) Regarding ground resistivity: we appreciate the elimination

of as many variables as possible for some testing, but were differencesfound in ground resistivity and what effect would the authors expectthe ground resistivity to have on the results obtained?

2) The details of factors effecting the variance between CFO andCWS seem to add to the complexity. Practical applications are thatthe relationship between CFO and CWS be weighed with a goodconfidence level but without excessive margins of safety so as toprevent an economic burden. Therefore, which factors do theauthors feel to be most significant?

3) The paper states the General Electric Company's method ofvariance was used in place of existing standards. What are themerits of the G.E. method? What efforts are being made to in-corporate them into the standards now under discussion for revisionand updating?

Manuscript received July 25, 1966.

M. K. Ramthun (Arizona Public Service Company, Phoenix, Ariz.):The authors are to be commended for casting a good deal of light onthe heretofore gray area of the effect of altitude on transmission linedesign.As a participant in the High Altitude Test Program at Leadville,

Colo., Arizona Public Service Company is particularly pleased withthe conclusions. These data were utilized by Arizona Public Servicein the design of the tower-insulation system for the Company's500-kV Four Corners-Colorado River Transmission Line.

After consideration of all factors, Arizona Public Service engineersadopted an exponent of 0.65 to the RAD. We emphasize, however,that this was done only after careful analysis of other factors whichwe felt indicated a rather conservative overall design.The following is an example of the difference these data made in

tower design.The tower-insulation system selected by Arizona Public Service,

using a RAD exponent of unity would have sufficed to an altitude ofabout 4200 feet. The exponent of 0.65 extended the altitude to about7000 feet. This permitted the use of a single type of tangent towerfor nearly 99 percent of the line. For those few miles above 7000feet, a few insulators were added to the tangent tower. For the fewtowers above 8000 feet, the light angle tower was utilized as atangent with additional insulation.Again it is emphasized that other considerations which indicated a

conservative line design encouraged the use of the 0.65 exponent.Among these considerations was a METIFOR analysis of theproposed line.

Manuscript received July 27, 1966.

958


John H. Moran (Lapp Insulator Company, Inc., LeRoy, N. Y.):The authors of this paper have done yeoman work in gathering andanalyzing the data which they present. It is axiomatic at this timethat the proper correctioni factors to be applied to the test valuesobtained on modern outdoor insulating structures are vitally needed.The influence of the air density is but one of a number of factorsknown to influence the strength of insulating structures, and itwill be only through the agency of such investigations as this thatthe entire truth will eventually become known. The authors deservethe thanks of the entire electrical insulating industry for the factspresented in this paper.The section entitled Scope of Tests contains the statement that

"the flashover strength of transmission line insulation is substantiallyinfluenced by the mass of the tower." I would like to suggest thatsome other term be used to describe this effect, such as, projectedtower area, since I do not believe that a change from one conduct-ing material in the tower to another conducting material, for in-stance, from aluminum to steel, which, although it would increasethe mass of the tower, would actually have any effect on the flashoverstrength.With respect to the wet tests conducted on configuration 2, which

utilized a rod gap, I would like to ask the authors if the flashoverpath was observed to follow consistently the area being wetted, orif the flashover path went beyond the wetted area.

Manutscript received July 28, 1966.

T. M. Parnell (Department of Electrical Engineering, Untiversity ofQueellsland, Australia): This paper is a most welcome contributionin a field where considerable effort is needed to establish sound designpractices for the higher transmission voltages. The authors are to becongratulated on their ingenuity in producing information from asituation where the cost of conventional experiments would probablybe prohibitive.The results do, however, raise a number of points where some

clarification seems necessary.1) In arriving at RAD corrections, humidity corrections had to be

applied in accordance with Fig. 11. Can the authors comment on themagnitude of the humidity corrections actually used, relative to theRAD corrections which they deduced from the corrected test resultsand on the probable errors in RAD corrections due to uncertaintyin the humidity corrections?

2) In their conclusions the authors suiggest that RAD correctiondecreases according to a power law as the length of the test objectincreases. Since both humidity and RAD may be presumed toaffect the breakdown process simultaneously and independently,can the authors suggest reasons for treating the humidity correctionsas a quantity independent of test object leingth? The standardquoted by the authors, ASA C68.1, suggests that this is not so in thecase of the shorter gaps.One aspect of the application of the authors' results appears to

deserve comment.In determining switching surge withstand levels, values of CFO-2ar

and CFO-3a are suggested-these corrresponding to withstandprobabilities of 0.977 and 0.999, respectively. Since relatively largenumbers of insulators and gaps will be stressed simultaneously by aswitching surge, the probability of flashover at a given withstandlevel can become quite high. If one assumes that 200 insulatorstrings and equivalent gaps may be simultaneously stressed by aswitching surge on a long line, the withstand probability for theline at a level equal to CFO-3a becomes 0.8, i.e., one would expect aflashover somewhere on the line for one in every five surges at thewithJstand level.

Is this figure likely to occur in practice or must we conclude thatadditional factors of safety, such as margins between the levels ofgenerated surges and the withstand level, will offset this effect?

T. Udo and M. Kawai (Central Research Institute of Electric PowerIndustry, Tokyo, Japan): The authors have presented a very in-teresting paper anid deserve our thanks for making it available to us.The influence of air density is an important subject in Japan,

too. In Tokyo and Chubu Electric Power Companies, the next 500-kV transmission lines are going to be planned beyond the moun-tainiouis area of about 5000 feet above sea level. This paper, then, istimely and valuable, indeed. As far as we are aware, previous datahave been limited to about 0.9 or so of RAD. Data have beenexpanded into the region of lower RAD, and the authors have shownthat the air density correctionfor impulse wave is (RAD)1'0 while thecorrection factor for switching surge wave is (RAD)07, tending toshow a saturation effect. This suggests that the insulation level inthe high altitude area might be installed more economically. Thestandard deviation for switching surge wave sometimes varied asmuch as two or three to one. We also have experienced similarresults on the standard deviation in our Shiobara Laboratory. Thisfact indicates that a number of tests should be necessary to deter-mine the withstand voltage. We use the wave of 150 - 200 by 3000gs in the test for determining the withstand voltage of line insulation,because this waveshape gives the minimum flashover voltage. Wewould like to ask why the authors use the longer wave front of240 ps.

Maniuscript received July 28, 1966.

A. R. Hileman (Westinghouse Electric Corporation, East Pittsburgh,Pa.): The effect of altitude on the switching surge critical flashovervoltage of air porcelain insulation has been an uncertain parameterin EHV insulation design. The results of this investigation are alarge contribution toward the solution of this problem.

Previously Aleksandrov et al. [',2 have suggested that for longgaps, data should be corrected by RAD to the nth power with nvarying from 1 for gaps of 1 meter to 0.7 for gaps of 7 meters. In arecent paper which presented results of tests on full-scale towerinsulation,[3] four test series using the same test object were madefor RAD ranging from 0.955 to 1.043. The data were compared whenuncorrected, corrected by the RAD to the first power, and cor-rected by the (RAD)0-5. Of these three methods of data correction,the use of n = 0.5 provides minimum data variation. Therefore thiscorrection factor was adopted. However, it was noted that RADvariations occurred principally because of temperature and there-fore it was suggested that full RAD corrections be employed forpressure variations.While the use of full RAD correction for low altitude designs[4]

does not incur a significant economic penalty, the use of full RAD incomparison to (RAD)0-6 as suggested by the authors would requireabout three additional insulators for 500-kV lines at 10 000 feet.Thus the application of these data will be of economic benefit toutilities with EHV lines in mountainous regions.

Using the authors' data in Table I, Fig. 30 has been prepared toillustrate the variation of n of the expression (RAD)n as a function ofthe equivalent number of suspensioin insulators. The dotted line showsthe authors' suggestions as given in the conclusioin and is a goodpractical estimate of the factor n. Do the authors have a theoreticalexplanation for the flatness of the curve for configuration 1?The standard deviation for recent simulated full-scale tower

switching surge tests[3] was 4.6 and 3.7 percent for test conditions ofpositive polarity, wet and dry, respectively, for insulators in V-string configuration. For insulators in vertical configuration thestandard deviation was approximately 3.7 for wet and 3.3 percentfor dry conditions. From tests on the Southern California Edison's500-kV tower[3] in which V-strings are employed, the standarddeviation was 4.8 and 3.2 percent for wet and dry conditions, re-spectively. The more recent tests performed on the Allegheny PowerSystems 500-kV tower[5] showed a standard deviation of 5.2 and4.0 percent for wet and dry conditions, respectively. Thus, for theseextensive tests it would be concluded that, 1) the standard deviationfor wet condition on V-string insulators, positive polarity, is over

Maniuscript received August 12, 1966.

959

Maiiuseript received July 28, 1966.


; FX30 ~I D

0.8~~~~~~~~~~20.8 I__, _0.6 __J _ ,

oil _I D70.4 5 10 1 2 25 31

0.2 _ -

0 5 10 15 20 25- 3CNO. OF EOUIVALENT SUSPENSION INSULATOR S

Fig. 30. Effect of number of equivalent suspension insulators on nof expression (RAD)n. Configuration number W stands forwet, D dry, noted on curves obtained from Table I.

one percent greater than for dry conditions, and 2) the standarddeviation for wet condition, V-string insulators, positive polarity, isfrom 4.6 to 5.2 percent; values which agree with the authors' averagevalue of five percent for these tests.The authors' results show a mean difference between low and high

altitude standard deviation of 1.4 percent, positive polarity, thehigh altitude being lower. This factor tends to offset the altitudecorrection factor of the critical flashover voltage since at highaltitude the withstand strength at three standard deviations belowcritical flashover is about 5 percent of the CFO greater than that atlow altitudes. Do the authors suggest using this factor in applicationat high altitude?The humidity correction factor curve of Fig. 11 is very interesting.

However, it is presented without full discussion of its derivation.It would be very helpful if the authors would present the analysis oftheir test data from which the curve was derived.The results of this investigation and the succinct and complete

manner of presenting data and results greatly adds to the solution ofthe problem of altitude correction factors.

REFERENCES1'] G. N. Aleksandrov, and V. L. Ivanov, "Electrical strength of

air gaps and insulator strings under the action of switching surges,"Elec. Tech. (USSR), vol. 3, pp. 460-473, 1962.

[2] G. N. Aleksandrov, V. Y. Kizeretter, V. M. Rudakova, and A. N.Tushnov, "The AC flashover voltage of long air gaps and strings ofinsulators," Elec. Tech. (USSR), vol. 2, p. 255, 1962.

[3] A. W. Atwood, Jr., A. R. Hileman, J. W. Skooglund, and J. F.Wittibschlager, "Switching surge tests on simulated and full-scaleEHV tower-insulator systems," IEEE Trans. Power Apparatus andSystems, vol. PAS-84, pp. 293-303, April 1965.

[4] W. C. Guyker, A. R. Hileman, and J. F. Wittibschlager, "Full-scale tower-insulation tests for APS 500-kV system," IEEE Trans.Power Apparatus and Systems, vol. PAS-85, pp. 614-623, June 1966.[5] A. R. Hileman, W. C. Guyker, H. M. Smith, and G. E. Grosser,Jr., "Line insulation design for APS 500-kV system," this issue,pp. 987-994.

E. Brasca and L. Zaffanella: (CESI-Centro Elettrotecnico Speri-mentale Italiano, Milan, Italy): We are happy to complimentthe authors of this very interesting paper on the extensive experi-mental work performed, which, for the first time, makes a sub-stantial contribution to the investigation of air density effects. Wecompletely agree that full-scale tests performed at different sea levels

are the best way to solve the problem of the economical design ofhigh altitude EHV transmission-lines insulation.We agree that the data shown in the paper allow the statement

that air density strongly affects insulation strength, but we thinkthat further work is required to establish definite values for theexponent n of the RAD correction factor. Our findings, as explainedin our paper,' show that when comparing results coming fromdifferent test plants, a confidence limit must always be taken intoaccount, as when test conditions are carefully duplicated. Thismeans that, in this case, it is possible to attribute t) air densitythe variations due to other factors. For instance, in the case ofnegative impulse tests of configuration 3, the discrepancies found bythe authors can be justified by the low consistency of the results.In fact, in our paper, the rod-to-plane configuration shows (TableI) the poorest consistency (a = 6 percent), so that differenceshigher than 34 percent between two results obtained in two differentlaboratories can occur with 5 percent of probability. Furthermore,large discrepancies in these results are reported by the authors (20units, Pittsfield tests).We could not completely agree with conclusion 5): "Negative

polarity switching surge produce higher flashover voltages thanpositive polarity." According to our test experience, which is con-firmed also by the literature, this statement is not generally valid,as wet negative can be, in quite a number of cases, more restrictivethan positive. On this point, we will be glad to have more detailedinformation on the spraying apparatus used both in Pittsfield andLeadville, particularly concerning nozzles, angle, and uniformity ofprecipitation.Apropos of the general remark on the consistency of the results,

we think that it is extremely difficult, if not impossible, to find atemperature effect independent of the RAD effect. We would alsolike to ask if any correlation was found between the results andconditions of sun, fog, snow, dew, wind, frost, and rain, recordedduring all the tests.We strongly support the opinion of the authors that "the standard

deviation for a single test, should not be given much weight," but anaverage value obtained from many tests must be used for each testcondition as the proper standard deviation. These values for in-sulator strings, calculated on the basis of very large number oftests, are given on Table VI of our paper. Table VII, given here,shows substantial agreement between our figures and those pro-posed by the authors.

TABLE VIIAVERAGE STANDARD DEVIATIONS OF FLASHOVER

DISTRIBUTION FOR INSULATOR STRINGS

LightningImpulse

pos- neg- Switching Impulseitive ative positive negative

Paper dry dry dry wet dry wet

This paper 2 2 5 3Brasca et al.1 1.1 1.4 4.2 4.3 2.2 3.3

1 E. Bracna, M. Tellarini, and L. Zaffanella, "The confidence limitof high-voltage dielectric test results," this issue, pp. 968-974.

T. A. Phillips, L. M. Robertson, A. F. Rohlfs, and R. L. Thompson:The authors wish to acknowledge the valuable contributions of thediscussers. We appreciate their thorough review of the data pre-sented and welcome the opportunity to respond to the questionsraised.To answer Mr. Kilgour, we would expect the ground resistivity to

Manuscript received September 8, 1966.

960

,o

Manuscript received August 15, 1966.

PHILLIPS ET AL.: AIR DENSITY AND TRANSMISSION LINEINSULATION

have a second order effect, if any, on the test results. However, toeliminate this variable, the wire mesh ground planes were incor-porated. One test was conducted at Leadville for configuration 3(10 suspension insulators, switching surge, positive polarity, dry)with and without the wire mesh ground plane, and the results agreedwithin 5 percent.

Referring to the last column of Table IV which shows the 95percent confidence interval estimate of the difference in standarddeviations, the factors that affect the standard deviation mostsignificantly are apparent by inspection of the lower limits of theconfidence intervals. For example, the effect of proximity on thenegative polaritys gma is pronounced. With 95 percent confidence,the sigma for configuration 3 was at least 3.2 percentage pointsgreater than the sigma for configuration 1 and 1.7 percentage pointsgreater than the sigma for configuration 2. While it is difficult todraw general conclusions from Table IV, the comparisons shownprovide useful information for application to specific design ques-tions and, as suggested in the text, appropriate values can be deter-mined for line design.

In determining the standard deviations, graphical methods wereapplied which are in general use throughout the industry. Linearregression methods and probit analysis could have been used, butit was felt that the graphical approach would be more convenient. Amodified linear regression method is included in the proposed stand-ard for disconnect switches.The discussion by Mr. Ramthun is particularl y gratifying in its

demonstration of the economies possible through practical applica-tion of the research data.Mr. Moran's point concerning use of the word mass is very well

taken. Certainly tower weight does not influence the flashoverstrength, though some measure which takes into account the depthof an adjacent tower member as well as the projected area may berequired.Although all flashovers were observed by test personnel, we do

not know if the flashover path went beyond the wetted area. We feelthat thorough investigation of this question would require pho-tographic techniques.We agree with the observation by Mr. Udo and Mr. Kawai that

substantial economies may be achieved in high altitude transmissionline insulation by incorporating realistic air density corrections, andthis is supported by Mr. Ramthun's discussion. It is interesting tolearn that results from the Shiobara Laboratory also show variationsof two or three to one in individual standard deviations. Our use ofthe 240-.os wave front was based on previous tests showing littledifference in flashover strength within the 100-300 ps wave-frontregion.

In response to Dr. Parnell, humidity corrections appliedat the two locations ranged from 1.0 to 1.08 at Pittsfield and 1.02 to1.06 at Leadville for switching surge, and 1.09 to 1.15 at Pittsfield and1.09 to 1.14 at Leadville for impulse. While some error may beintroduced by these corrections,any such errors would be in the samedirection for both locations and thus would, for the most part,balance out in the calculation of RAD corrections.

It is possible that the humidity correction varies with test objectlength. However, the ASA Standard C68.1 uses a full correctionfactor for all gaps with a flashover exceeding 140 kV. We are notaware of data suggesting other than a full correction for gaps ofthe lengths involved in the present investigation. We did not studythe effect of humidity and insofar as possible tested under similarhumidity conditions so as to eliminate this variable.

Based on Dr. Parnell's stated assumptions, one flashover in everyfive surges would, in fact, be expected to cause flashover. However,Transient Network Analyzer studies and field test programs havedemonstrated that switching surge magnitude is a statistical variablesuch that surges typically experienced on a given line will exceed90 percent of the maximum magnitude with a frequency of lessthan 0.10. Therefore, assuming that the maximum surge corre-sponds to a flashover probability of one in five, or 0.20, and thatsurges with magnitudes less than 90 percent of the maximum willnot cause flashover, the probability that a given surge will causeflashover will be less than 0.02, or one in 50 (0.2 X 0.1). In addition,atmospheric conditions will rarely combine to produce a minimuminsulation strength condition, and in fact the strength may beincreased as well as decreased by weather variables.

Comprehensive examination of flashover probability consid-ering the full range of statistical interaction of the variables in-fluencing inisulation performance, such as accomplished by the

METIFOR program,12 generally demonstrates a significantlylower flashover probability than that obtained by assumingcoincidence of a maximum surge withmminimum strenigth.We do not have a good theoretical explanation at present for the

apparent flatness of the curve for configuration 1 in Mr. Hileman'sFig. 30, but it is apparent that this is in some way related to the lackof tower proximity effect.Mr. Hileman has made a valuable contribution in summarizing

the standard deviations obtained in various tower test programs.These data provide welcome support for the selection of 5 percent asa conservative value for the standard deviation.We believe further analysis and correlation of standard deviation

data are required before differences on the order of one percent in thevalue of sigma are incorporated in designs. We would suggest atpresent that the indicated lower sigma at high altitude be consideredas further justification for applying the reduced air density correc-tions described in the paper.The humidity correction curve was developed from analysis of

limited tests conducted over a range of humidities and while notnecessarily exact, it does provide the best available estimate of thehumidity effect. Derivation of the curve required a fair amountof engineering judgment. It is expected that more complete support-ing data will become available.We agree with Mr. Brasca and Mr. Zaffanella that further refine-

ment of the air density correction is desirable, and we hope thattest work will also be undertaken at intermediate altitudes. How-ever, we do feel that the data presented in the paper provide a soundbasis for engineering application.While purely statistical variations may produce discrepancies in

test results, it is our feeling that the large differences that have beenobserved in results obtained by different laboratories are primarilydue to differences in test procedures, personnel, and philosophy.The test program reported in this paper, while conducted at twotest sites, was carried out with identical procedures, philosophy,and the same supervisory personnel at both locations. With thecareful controls employed and the large number of test points, webelieve variations in the data have been minimized. In addition,where inconsistencies were observed, test points were repeated.We do agree that the confidence limits for results obtained for

configuration 3 on negative polarity are of necessity broad, since theinherent natural behavior of that configuration produces a widescatter of data points.

General Electric Company has conducted a very large number ofswitching surge tests and only once or twice have situations beenobserved where the negative polarity CFO was lower than thepositive polarity. These were under wet conditions where there wasa large amount of water cascading over the insulators. Even underthese conditions, however, the flashover voltage at negative polaritywas only slightly below the positive polarity flashover voltage.The spray nozzles were of the type used on garden hoses. These

were used since they will throw a spray farther than the U.S.A.standard spray nozzles. We estimate that the angle of the precipita-tion was between 300 to 450 to the vertical. It was attempted tomake the spray as uniform as possible, but the ubiquitous breezesmade impossible as good uniformity as can be obtained indoors.We agree that a temperature effect independent of the RAD

effect is difficult to identify. Certain inferences in the literature ledus to include this variable in our investigation. We suspect thatthere is no such effect, but we cannot rule it out.We could find no correlation with sun, fog, dew, wind, and frost.

Generally the tests were stopped during rain, and there was verylittle snow. We did not include these parameters in the paper becauseof the complexity of presentation.Again we are pleased with the concurrence that "the standarddeviation for a single test should not be given much weight, but an

average value obtained from many tests must be used for each testcondition." The agreement between the discussers' and the authors'values for standard deviation is gratifying. Apparently, the valuesgiven in our paper are slightly conservative and should providereasonable measures for engineering application.

1 J. G. Anderson and L. 0. Barthold, "METIFOR, a statisticalmethod of insulation design of EHV lines," I'EEE Trans. PowerApparatus and Systems, vol. 83, pp. 271-280, March 1964.

2 J. G. Anderson and R. L. Thompson, "The statistical computa-tion of line performance using METIFOR," IEEE Trans. PowerApparatus and Systems, vol. PAS-85, pp. 677-686, June 1966.

961

Date post:	13-Sep-2015
Category:	Documents
Upload:	juan-david-ramirez
View:	213 times
Download:	0 times

Influencia de Altitud en Aislamiento de LT

Documents