Study of airgap breakdowncharacteristics under ambient conditions

of reduced air densityA.J. Eriksson, Pr.Eng., B.Sc.(Eng.), Ph.D.. C.Eng., F.I.E.E., B.C. le Roux,

B.Sc.(Eng.), H.J. Geldenhuys, Pr.Eng., B.Sc.(Eng.), and D.V. Meal, Pr.Tech.

Indexing terms: Breakdown and gas discharges, Discharges, electric, Power system protection

Abstract: The paper summarises data and preliminary observations arising out of engineering and physicalstudies of airgap breakdown at a moderately high altitude (1425 m), outdoor high-voltage research laboratoryin South Africa. It is shown that in the case of positive switching impulses, the application of the presentstandard IEC atmospheric correction procedures can yield results which lie above the expected 'sea level' trends.The relative influences of humidity (and air density sensitivity) upon airgap breakdown characteristics are alsoexamined, both in engineering breakdown tests, as well as in studies of physical parameters of the discharge.

1 IntroductionThe projected power demand growth in South Africa up tothe year 2000 is estimated as 4-6% per annum. This highgrowth rate imposes major requirements for the provisionand transmission of electrical power throughout theregion. The fact that the regional geography includes largeareas which lie at altitudes in the range 1500-1800 m andwhich also experience a moderately high incidence oflightning (typically 5-10 ground flashes km"2yr"1 , or50-80 storm days per year) places great emphasis on theneed for background high-voltage research.

As a consequence, the National Electrical EngineeringResearch Institute (NEERI) of the South African Councilfor Scientific and Industrial Research (CSIR) recently com-missioned a new outdoor high-voltage research laboratory.A schematic layout of the laboratory is depicted in Fig. 1and a detailed description is given in Reference 1.

Being located at an altitude of 1425 m, this laboratoryconcentrates on the study of fundamental questionsregarding the effects of reduced atmospheric pressure andvarying meteorological influences (e.g. humidity) on theimpulse and AC breakdown characteristics of practicalairgaps. This knowledge is essential for optimising thedesign of modern high-voltage equipment, such as theconductor-window clearances in transmission-line towers,for example, as applied in the South African networks.

To facilitate the developing infrastructure, and todevelop local expertise in the fields of EHV and UHVengineering, the research approach adopted at the labor-atory is two-fold:

(a) Engineering study of statistical breakdown charac-teristics of practical airgaps, including meteorologicalinfluences, with a view to establishing a comprehensivedatabase which can facilitate the development of improvedatmospheric correction models.

(b) Study of physical aspects of breakdown in airgaps,with a view to clarifying observed atmospheric influencesand the role played by various physical parameters associ-ated with the breakdown process, as well as the implica-tions for low breakdown probability determinations.

Paper 4922A (S3), received 6th January 1986Mr. le Roux, Mr. Geldenhuys and Mr. Meal are, and Dr. Eriksson was formerly,with the Power Systems Technology Group, National Electrical EngineeringResearch Institute, Council for Scientific and Industrial Research, PO Box 395, Pre-toria 0001, South Africa. Dr. Eriksson is now with BBC Brown, Boveri & CompanyLimited, Department ATX-ST, PO Box 8242, CH 8050 Zurich, Switzerland

This paper summarises data and preliminary observationsarising out of the engineering studies undertaken thus far,

25600mm-

/^-—^safety barrier:

y* gates regulatorss/\ electricity

'substation

Fig. 1 Schematic layout of the outdoor high-voltage test laboratory

and also presents the first results emerging from thedeveloping study of physical aspects, including the possibleimplications for atmospheric correction procedures andlow breakdown probability studies.

2 Engineering study of airgap breakdowncharacteristics

The laboratory is currently engaged in establishing a com-prehensive database of test results across a variety ofairgap geometries, with a view to the primary study ofatmospheric influences. The initial phase of this work hasfocused on classic airgaps, to relate the local data (at

IEE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986 485

configuration

0 100mm

rod -plane conductor-plane conductor-window

fp 28mm

conductor-conductor

gap factor 1 . 0 1. 15 1.33 - 1.43

clearancesD . mS. m

3 - 5.516

2 - 45

Fig. 2 Main characteristics of the test configurations investigated to date

reduced air density) to results obtained in similar geomet-ries at lower altitudes.

2.1 Test configurationsThe main test configurations investigated to date aredepicted schematically in Fig. 2, together with the corre-sponding calculated gap factors [2] (defined as the ratio ofthe positive switching impulse strength at the critical time-to-crest to that of a rod-plane configuration having thesame clearance.)

2.2 Experimental procedureAs positive polarity impulses are of particular importancein the design of external insulation, these were used in themajority of the tests performed. The interval betweenimpulses was approximately 35 s and the impulse wave-shapes used ranged from lightning impulses (LI) to longswitching impulses (SI).

The traditional up-and-down testing procedure (basedon a nominal sample of 50 impulses, and using a step sizeof between 2 and 5%) was used to determine the 50% dis-ruptive discharge voltage of each airgap and impulsewaveshape investigated, using a computer based system ofimpulse generator control and data acquisition.

In a number of the test configurations, multiple leveltests were also performed to investigate low probabilityflashover and the standard deviation (a) of the flashovervoltage distribution. With the exception of the earlyconductor-window tests [3], in which atmospheric condi-tions were monitored at approximately 10-minute inter-vals, atmospheric variations, impulse peak voltage, impulsetime-to-breakdown and impulse generator chargingvoltage were all recorded on a shot-for-shot basis, using amicrocomputer based data acquisition system in the labor-atory [1]. Fig. 3 depicts the range of measured atmo-spheric parameters expressed in terms of relative airdensity and absolute humidity, as recorded during thetests.

2.3 Analysis of test resultsThe likelihood statistical analysis approach [4] was usedto analyse the binomial test data obtained in all the tests(generally assuming a normal distribution of flashovervoltages). As an example, Fig. 4 depicts the most likelyvalues of U50 and a, together with their 90% confidenceloci (likelihood ratio factor of 0.1 for two degrees offreedom [5]), as computed from the measured and IECcorrected voltages obtained in a 500-shot multiple leveltest (loci (i) and (ii)), as well as a 50-shot up-and-down test(loci (iii) and (iv)) in a 5.5 m conductor-window geometry;impulse waveshape: 230/2300 fis.

15r14131211

„ 10e 9D> 8

.? 7

i 61 5£ u

3

2

1

0.80 0.82 0.84 0.86 0.88 030air density _ p.u.

Fig. 3 Measured relative air density against absolute humidity recordedduring the impulse tests+ R/P

/* c/wx C/C

The analysis shows that the most likely values of U50and a computed from the two tests agree to within 1 % ofeach other, for both the measured and IEC corrected data.The analysis also shows that the up-and-down procedureis more suited to computing U50, in that the confidenceinterval for U50 is smaller than the corresponding intervalfor a. In this case also, it is seen that the multiple leveltechnique used was biased towards determining low prob-ability flashover, as evidenced by the skewed locus [6].

(iii)

1524 1548 1572 1596 1628 1644 1668 1692 1716 1740U50 voltage,kV

Fig. 4 Likelihood analysis ofbionomial test data

(i), (ii) 500 shots multiple level tests (measured and corrected)(iii), (iv) 50 shots up-and-down tests (measured and corrected)5.5 m conductor-window gap

IEE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986

2.4 Examples of test resultsFigs. 5 and 6 illustrate the dependence of 50% flashovervoltages (IEC corrected) on the time-to-crest values of theapplied impulse, in the rod-plane and conductor-windowconfigurations, whereas in Fig. 7, the resultant criticalflashover voltages (CFV) in the rod-plane configuration,

2.0

1.S

1.6

1.4

0.8

0.650 450150 250 350

time-to-crest, ps

Fig. 5 Influence of impulse time-to-crest on the positive polarity flash-over of rod-plane gaps (IEC corrected voltages) as measured at 1425 m• 5m+ 4mx 3mO 2 m

2200

2100

2000

1900

1800

1700

1600

1500

1400

1300

1200

1100

(iii)

100 400200 300time-to-crest, \ts

Fig. 6 Influence of impulse time-to-crest on the positive polarity flash-over of conductor-window gaps (IEC corrected voltages), as measured at1425 m(i) 5.5 m gap(ii) 4.0 m gap(iii) 3.05 m gap

IEE PROCEEDINGS, Vol. 133, Pi. A, No. 8, NOVEMBER 1986

both measured and IEC corrected, are plotted in relationto gap spacing.

1.2

i 1.0

0.8

0.6

0.41.5 5.52.5 3.5 4.5

rod-plane gap, m

Fig. 7 Critical flashover voltage in rod-plane gaps, as measured at1425 m

OIECx Pigini+ measured

2.5 Discussion of resultsAs is well known, the critical times-to-crest (Tc) in eachconfiguration (i.e. the impulse time-to-crest which gives theminimum flashover voltage) are proportional to the gapspacing. For standard atmospheric conditions (20°C,101.3 kPa and 11 g/m3), an accepted relation [2] is

Tc = (50 - 35(K - 1))D (1)

where

D = gap spacing, m

K = gap factor (K = 1 for rod-plane gaps)

Similarly, the critical flashover voltages (CFV) under stan-dard atmospheric conditions and for unity gap factor, canbe expressed as

CFV = 3400/(1 + 8/D) (2)

Eqns. 1 and 2 may be combined as

Tc = (280 (K - 1) - 400)/(l - 34OO/CFV) (3)

This expression is plotted in Figs. 5 and 6 (dashed lines)and agrees comparatively well with the measured data,and eqn. 2 is plotted as a solid curve in Fig. 7. On average,application of the current IEC procedures for atmosphericcorrection in the rod-plane configuration yields resultsapproximately 7.5% above the predicted trend (eqn. 2). Onthe other hand, as is evident from Fig. 7, application of analternative approach to atmospheric correction, as pro-posed by Pigini et al, [7], yields a closer consistency.

Eqn. 2, together with the calculated gap factor (K) ofeach geometry, can be used to calculate the value of CFVfor standard atmospheric conditions (CAL), i.e.

CAL = K x [eqn. 2] (4)

These values, together with the IEC corrected criticalflashover voltages, (IEC) (determined as the minima of the[/-curves in the rod-plane (R/P), rod-rod (R/R), conductor-window (C/W), conductor-plane (C/P), and orthogonal

487

Table 1 : Resultant critical flashover voltages in relation togeometry (IEC corrected) and the observed dispersionbetween measurements and predictions — as determined at1425 m

applied impulse, for the two groups of tests describedabove.

Configuration

R/PR/PR/PRIPR/RR/RR/RC/WC/WC/WC/PC/PC/PC/PC/Pc/cc/cc/c

Gapm

2.003.004.005.002.003.004.003.054.005.501.002.003.004.005.001.002.003.00

Gap factor,K

1.0001.0001.0001.0001.4291.3751.3331.2211.2211.2301.1501.1501.1501.1501.150———

CAL,(kV)

680927

11331308

97212751510114513831704434782

106613031503———

IEC,kV

755996

12081374

97913111592114113991698

470815

112513401510

55310771546

A, %

11.07.46.65.00.72.85.4

-0.31.2

-0.48.24.25.52.80.5———

conductor-conductor (C/C) geometries), are all sum-marised in Table 1. (Note that gap factors for the orthog-onal conductor-conductor configuration are not coveredin Reference 2.)

On average, the IEC corrected voltages lie some 7.5%and 3.0% above the calculated values for the rod-planeand rod-rod configurations, and 0.2% and 4.2% above thecalculated values for the conductor-window andconductor-plane configurations, respectively. With theexception of the rod-plane results, all the results thus liewithin the tolerance specified for CFV prediction in Refer-ence 2 (i.e. errors of less than 5%, with a confidence limitof 70%).

2.6 Influence of humidityAs an example, the range of atmospheric variationsencountered during testing in the 5 m conductor-planeconfiguration is depicted in Fig. 8.

The tests can roughly be grouped into two sets:(a) Group I: Average humidity (H) = 6.4 g/m3

(b) Group II: Average humidity = 10.9 g/m3

Fig. 9 illustrates the dependence of the 50% flashovervoltage (IEC corrected) on the time-to-crest value of the

OQ O

0.80 0.82 0.84 0.86 0.88air density, p.u.

0.90

Fig. 8 Measured relative air density versus absolute humidity recordedduring impulse tests in the 5 m C/P configuration

O group I+ group II

1600

150050 100 150 200 250 300

t ime- fo-cres t , (JS

Fig. 9 Influence of impulse time-to-crest on the positive polarity flash-over of a 5 m C/P gap (IEC corrected voltages)O H = 6.37 g/m3

+ H = 10.9 g/m3

300

Fig. 10 Normalised time-to-breakdown (TTB/TC) against time-to-crest(7^), as measured in a 5 m C/P configuration at 1425 m

O H = 6.37 g/m3

+ H = 10.9 g/m3

Even after application of current atmospheric correctionprocedures, the corrected data presented in Fig. 10 stillsuggest a remanent influence of humidity in thisgeometry. Although it is difficult to visualise the character-istic U-curves associated with these data due to the scatterand sparse distribution, mathematical fitting of a 3rd-orderpolynomial (using least-squares techniques) to the two setsof data, suggests that:

(i) the critical flashover voltage decreases with increas-ing humidity

(ii) the critical time-to-crest decreases with increasinghumidity.

The latter tendency is better indicated in Fig. 10 whichdepicts the normalised time-to-breakdown (TTB/TC)plotted against the impulse time-to-crest (7 )̂.

Similar tendencies in UHV gaps have been reported byother researchers [8, 9] and have been accounted for interms of the influence of humidity upon leader velocityduring the development of the discharge. The present(IEC) correction procedures [10], as well as the proceduresproposed by Pigini et al. [7], do not make sufficient provi-sion for correction of these effects. In Reference 9 an

IEE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986

attempt is made to introduce a secondary humidity correc-tion factor which corrects the wavefront (time-to-crest) ofthe applied voltage by a factor which is dependent onhumidity. This approach requires further investigation,however, and to understand better and compensate forthese effects in engineering atmospheric correction pro-cedures, it is necessary to investigate physical aspects oflong spark breakdown and the interdependencies of thevarious discharge parameters (notably in relation to atmo-spheric variations).

3 Study of physical aspects of airgap breakdown

Thus far, the preceding engineering studies and the avail-able database assembled under the local conditions ofreduced air density, indicate that the observed airgapbreakdown characteristics in a variety of classic geometriesare reasonably consistent with the trends of sea-level refer-ence studies [2]. In particular, the observed breakdowntrends are comparatively well accounted for in terms ofeither the existing IEC atmospheric correction procedures[10], or those recently proposed by Pigini et al. [7]. Inaddition, application of the gap factor approach yieldsconsistency between predicted and measured breakdowntrends. The largest dispersion in fact relates to the rod-plane geometry, which, of course, serves as the base for gapfactor estimation methods. The above remarks notwith-standing, dispersions of the order of 1-11% are evident inattempting to relate the measured data to current atmo-spheric correction procedures, implying some measure ofuncertainty, or sensitivity to the local circumstances ofreduced air density. In addition, as discussed in section 2.6above, and noted by several other authors, significanthumidity influences are certainly also present, especially inthe longer gaps.

In a complementary programme of work to the aboveengineering studies, therefore, the NEERI laboratory hasalso embarked upon a limited programme of research onphysical aspects, in an attempt to clarify some of theobserved sensitivities to atmospheric influences.

In implementing this work, and considering first posi-tive impulse breakdown, it must be recognised that the dis-charge mechanisms prevailing in the nonuniform fieldgeometries under study (i.e. gaps typically 0.5-5 m),involve the joint interaction and sensitivities of bothstreamer propagation and leader development processes[11, 12]: the former predominating in the shorter gaps (i.e.1-2 m), and the latter being increasingly important in thelarger gaps. This is illustrated by a sample photograph ofleader development in a 4 m rod-gap, shown in Fig. 11.

In the current phase of this research, two main projectsare currently in progress:

(a) a study of positive streamer voltage gradients (inrelation to atmospheric influences)

(b) a study of withstand leader physical parameters andtheir relationship to breakdown probability.

3.1 A study of positive streamer voltage gradientsAs noted earlier, during positive impulse breakdown ofnonuniform field airgaps, the positive streamer processpredominates in shorter gaps and still contributes signifi-cantly to the breakdown of the larger practical airgaps.The present project is concerned with a study of the effectsof air density and humidity on positive streamer voltagegradients, as measured during controlled tests within non-uniform field airgaps. The initial tests were carried out insmall sphere-plane gaps (gap sizes in the range 0.5-1 m)over the humidity range 3-30 g/m3. The typical param-eters measured included the apparent charge injected intothe gap during breakdown and the electric field at theplane, using a specially designed field filter probe [13]. Fig.12 illustrates a typical example of the measurement of the

Fig. 11 Discharge development in a 4 m rod-rod gap (outdoors) stressedwith a positive switching impulse which was chopped at the instant of thefinal jump

IEE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986

Fig. 12 Simultaneous measurement of applied voltage (upper trace) andelectric field (lower trace) at the plane in a 50 cm sphere-plane gap

Uncorrected peak voltage = 226 kVSphere radius was 1.25 cm, RAD = 0.817(i) first corona inception with a sudden increase in electric field(ii) arrival or the first streamers at the plane, defined as the minimum streamergradient E(iii) breakdown

electric field at the plane in a rod-plane gap, in thisinstance stressed by a positive switching impulse (455/2800 /is). The first corona is evident, as well as the arrivalof the first streamer at the plane, thereby defining theminimum streamer gradient £s

+. Repeating such measure-ments over a wide range of humidities and voltage impulseapplications, yielded the trend in observed minimumstreamer gradients in relation to humidity increase, asdepicted in Fig. 13.

489

A linear dependency is evident and this appears to berelatively insensitive to gap geometry in the range studied.

500

A 50

2 400

§ 350

30025 305 10 15 20

humidity _ g / m 3

Fig. 1 3 Minimum positive streamer gradient as a function of humidityThis analysis contains data from various arrangements!• D = 50 cm, R = 1.25 cmx D = 50 cm, R = 3.18 cm• D = 100 cm, R = 1.25 cmT = 30°C, RAD = 0.817, waveshape 455/2800 /zs

In general, the measured values of minimum streamer gra-dient were 15-20% less than the average voltage gradientsmeasured across the gaps. The observed linear humiditydependency can be expressed by the relation

£s+ = 379(1 + 1.49(/i - ll))/100 kV/m (5)

where h = absolute humidity in g/m3.The reference value of positive streamer gradient of

379 kV/m (i.e. for a reference humidity h = 11 g/m3), islower than the value of 500 kV/m adopted as the averagestreamer gradient in nonuniform gaps (Pigini et at). [7],but this may be accountable in terms of an air densitydependency [12], and the distinctions between theminimum streamer gradient (as measured) and the averageprevailing across the gap.

Positive streamer gradients in small airgaps and theinfluences of atmospheric parameters have also been inves-tigated by Phelps et al. [14], albeit using different measur-ing techniques. A consideration of their results can providesome guidance on the joint influences of air density andhumidity. In particular, regression through their datayields the following approximation for average positivestreamer gradient:

£„+ = 425(51-5 + (4 + 5«5)fe(kV/m) (6)

where <5 = relative air density.In contrast to the linear assumptions of the standard

atmospheric correction procedures, eqn. 6 implies thepossibility of a nonlinear dependency of breakdownvoltage upon air density in gaps in which the positivestreamer process predominates in the breakdown mecha-nism. In particular, application of eqn. 6 indicates thatbreakdown voltage dependencies in such gaps could beproportional to <5", where n varies from 1 to 1.3. (It may benoted that examination of data presented by Pigini et al.[15] relating breakdown voltages to air density in a 0.5 mrod-plane gap stressed by positive lightning impulses, alsosuggests a similar nonlinear dependency on air density, inthe range <5 = 0.3-1.2.)

Consideration of eqn. 6 further implies that humiditycorrection could itself be a function of air density, asapparent in the second term of this equation. Taking thecommonly expressed form of engineering humidity correc-tion factor [16]

and applying this in eqn. 6 over a range of air densities,yields the relative percentage dependencies in humiditycorrection factor, as given in Table 2.

Table 2: Relative percentage dependencies in humidity cor-rection factor

Relative air density (<5) % increase per g/m3 in relationto 11 g/m3(e)

1.00.80.6

1.722.052.57

The correction factor e appears to be approximatelyinversely proportional to air density. This was investigatedfurther during a limited series of laboratory tests in a 0.5 mrod-plane gap, as illustrated by the summarising data inFigs. 14 and 15. The resultant trends in average break-down voltage gradient in relation to humidity, yield

*„ 550

? 400

350300 10 20

humidity t g / m 3

Fig. 14 Measured average breakdown voltage gradients in a 50 cmsphere-plane gap (U50/d) as a function of humidity

Waveshape: 455/2800 usR = 1.25 cm, T = 30°C, RAD = 0.812Ui0/d = 439 (1 + 1.29(ft - 11)/100) kV/m

= s(h- / J O ) /100 (7)

10 20 30humidity, g /m 3

Fig. 15 Measured average breakdown voltage gradients in a 50 cmsphere-sphere gap (Uso/d) as a function of humidityWaveshape: 1.5/52 /JSR = 1.25 cm, t = 30°C, RAD = 0.812-0.822U50/d = 500(1 + 1.27(A - ll)/100) kV/m

humidity correction factors in the range 1.27-1.29 (as arelative percentage increase per g/m3).

Bearing in mind that this percentage increase is in factnot present in the application of the conventional standardcorrection procedures (i.e. as illustrated by Reference 16 inwhich e = 1 in positive rod-plane gaps), these observationssuggest that air density sensitivities in the humidity correc-tion factor may well have to be taken into account in thosegeometries where positive streamer gradients predominatein the breakdown mechanism.

A further series of laboratory tests under varying airdensities is being planned, as the next phase in this investi-gation.

1EE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986

3.2 Considerations on the role of leader charge andleader length at low breakdown probabilities

Insulation co-ordination of air-insulated electrical systemsrequires a knowledge of the switching surge breakdownprobability (P) of airgaps in the range 10~3-10~4 p.u. Inconventional breakdown testing, more than l/P impulses(shots) would thus be required to determine a specific Pwith some degree of confidence. As a possible alternativeapproach to conventional testing methods, the questionhas been asked [17] whether a knowledge of the physicalparameters associated with the breakdown process, (i.e.leader length, electrical charge, etc.) can be used to deter-mine low breakdown probabilities with more efficiency.Such an approach could lead to a reduction in the totalnumber of shots required to determine any specific P.

A project was undertaken early in 1985 in which thedependencies and relative interactions of peak charge,leader length (LL) and axial leader length (ZL) were inves-tigated in 4 m rod-plane and rod-rod gaps. LL and ZL aredefined in Fig. 16; both hemispherical and flat-tipped rodswere used.

HV rodhemisphere or flat tip

the spatial development of the leader plays an importantrole in determining either a withstand or a breakdown

plane

•100 mm

Fig. 16 Schematic diagram of the electrode set-up showing the variousleader parameters being measured

The charge measuring system used a fibre-optic link(analogue modulation with a — 3 dB bandwidth of 10 Hz-3 MHz) to transmit the measurement at the HV electrodeto earth potential. A digitiser (200 MHz, 8 bits) and amicrocomputer were used to record the peak charge. Thedimensional parameters of the leader were determinedfrom two orthogonal still photographs taken of each dis-charge. The cameras were equipped with glass optics witha lens aperture of/1.8 and 1000 ASA film.

Summarising results are reported in Reference 18.Examples of cumulative probability plots (normal prob-ability scale) of impulse peak charge are depicted in Figs.17 and 18.

As can be seen from these figures, there are significantdifferences, both in the shape of the cumulative charge dis-tribution (which is by no means normal), and in the peakcharge measured in the two electrode tip systems. Linearregressions were computed between peak charge and theleader dimensional parameters. A typical example is shownin Fig. 19 which depicts the correlation between peakcharge and LL. In all cases, the correlation between peakcharge and LL was better than the correlation betweenpeak charge and ZL.

This work further substantiates the earlier Les Renard-ieres studies [11], namely that a strong correlation existsbetween withstand leader length and the leader peakcharge, as shown in Fig. 19. These results also indicate that

50 60 70 80 90 100 110leader charge, pC

Fig. 17 Cumulative probability of measured leader peak charge, for a4 m rod-rod gap with flat-tipped rods• P = 8%x P = 10%• P = 23%

50 90 10060 70 80leader charge, ̂ ic

Fig. 18 Cumulative probability of measured leader peak charge for a4 m rod-rod gap with hemispherical!y tipped rods• P = 28%• P = 11.5%x P = 1.1%

100

90

80

<^ 70

o> 60o

S 50

40;

30 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0leader length,m

Fig. 19 Linear regression between measured leader length and peakcharge,for a 4 m rod-plane gap with aflat tipPeak voltage = 1205 kV (uncorrected)Q = 36 + 30.3 LLR2 = 0.808

IEE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986 491

condition, and suggests that if the discharge developsalong the gap axis, then a better chance exists that break-down will occur.

The critical leader charge was denned as the magnitudeof the injected charge just before the final jump in the caseof breakdown. A large dispersion was observed in the criti-cal leader charge, particularly with flat-tipped rods, asshown in Fig. 17. On the other hand, it was found that thelimiting leader peak charge could often exceed the meancritical leader charge, as illustrated in Figs. 17 and 18. As aconsequence, the strong correlation between real leaderlength and charge suggests that the dispersion in the criti-cal leader charge is due principally to the stochastic spatialbehaviour of the leader.

Thus far, therefore, this work generally supports theconcept of applying critical charge measurements, but clar-ifies their relative potential for determining low breakdownprobabilities. To be fully effective, analysis techniques willhave to be devised which acknowledge the large dispersionin critical charge, and take account of this in estimatingbreakdown probability. Our future work will be directedtowards consideration of such methods.

4 Future research activities

The laboratory plans to continue tests in phase-to-earthconfigurations to cover the complete spectrum of seasonalatmospheric variations, and to expand the database acrossa wide range of practical geometries. With the erection of a500 m long test span in the near future, tests will also beextended into phase-to-phase configurations. (The testspan will also facilitate studying the effects of corona onsurge propagation, and aspects of transmission-line shield-ing.)

The laboratory is also moving progressively into funda-mental studies of physical aspects of the long spark break-down mechanism in convergence with the engineeringtests. As these studies develop further, this facet of thework is likely to be an increasingly important adjunct tothe engineering scope of the high-voltage research pro-gramme, and is likely to provide the primary basis for the

implementation of optimised atmospheric correction pro-cedures.

5 References

1 LE ROUX, B.C., ERIKSSON, A.J., and KRONINGER, H.: 'Provi-sions of a new high-voltage research laboratory at conditions ofreduced air density', Trans. S. Afr. Insl. Electr. Eng., 1985, 76, (3)

2 THIONE, L.: 'Evaluation of switching impulse strength of externalinsulation', Electra 1983,(94)

3 ERIKSSON, A.J.: 'Results of prototype 800 kV tower window break-down studies'. CSIR Special Report ELEK 248, July 1982

4 BROWN, G.W.: 'Method of maximum likelihood applied to theanalysis of flashover data', IEEE Trans., 1969, PAS-88

5 WILKS,J.W.:'Mathematical statistics'(Wiley, 1962), Chap. 136 CARRARA, G., and YAKOV, S.: 'Statistical evaluation of dielectric

test methods', L'Energia Elettrica, 1983, Nl7 PIGINI, A., et al.: 'Influence of air density on the impulse strength of

external insulation'. IEEE PES Winter Meeting, February 19858 BUSCH, W.: 'Air humidity: An important factor for UHV design',

IEEE Trans., 1978, PAS-97, pp. 2086-20939 AIHARA, Y., WATANABE, Y., and KISHIZIMA, I.: 'Analysis of

new phenomenon regarding effects of humidity on flashover charac-teristics for long airgaps', ibid., 1983, PAS-102, (12)

10 'High-voltage test technique, Part 1. General definitions and testrequirements'. IEC Recommendation, Geneva, Switzerland, 1973

11 LES RENARDIERES GROUP: 'Researches on large airgaps at LesRenardieres', Electra, 1972, (23), 1975, (35); 1977, (53); 1981, (74)

12 WATERS, R.T.: 'Breakdown in nonuniform fields', IEE Proc. A, 1981,128,(4), pp. 319-325

13 WATERS, R.T.: 'Diagnostic techniques for discharges and plasmas.Electrical breakdown and discharges in gases' (Plenum Press, 1983),KUNHARDT and LUESSEN (Eds.), Series B: Physics, vol. 896

14 PHELPS, C.T. and GRIFFITHS, R.F.: 'Dependence of positivestreamer propagation on air pressure and water vapour content', J.Appl.Phys., 1976,47,(7)

15 PIGINI, A., RIZZI, G., BRAMBILLA, R., and GARBAGNATI, E.:'Influence of air density on the dielectric strength under impulse volt-ages'. Proceedings of the Fourth International Symposium on HighVoltage Engineering, Athens, Greece, 5th-9th September 1983, Vol. 2

16 JONES, B., and WATERS, R.T.: 'Air insulation at large spacings',Proc. IEE, 1978,125,(113), p. 1152

17 CARRARA, G., GALLUCCI, F., MANGANARO, S., and PIGINI,A.: Proceedings of the 3rd International Symposium on GaseousDielectrics, March 1982, pp. 199-202

18 GELDENHUYS, H.J., LE ROUX, B.C., and MEAL, D.V.: 'Criticalcharge and critical leader length as a means of determining low break-down probability in large air gaps'. Presented at the 8th InternationalConference on Gas Discharges and their Applications, Oxford, UK,16th-20th September 1985

IEE PROCEEDINGS, Vol. 133, Pt. A, No. 8, NOVEMBER 1986