GXS UZLETAG

Evaluation of high-voltage circuit breaker performance with a validated arc model

RPRSmeets and VKortsz Abstr9ct: A digital current-zero mcasurrng system was used to monitor about 200 zero crossings of short-line fault currents in three designs of commercial high-voltage circuit breakers. The breakers were subjected to standard JEC 60056 tests in a high-power laboratory. With these results, an empirical arc model was dcrivcd, based on the series connection of thrcc classical modified Mayr models. It was found that only three frec parameters, extracted from the arc vokage and current during the prezero period can describe the state of the breaker after each test. With this model, thc margin of the interrupting capability could be directly obtained. A strong relation between this margin and the arc conductivity at current zero can lx cstabhshed. Independent of arcing time, arcing current and state of the breaker, it was found that the current zero conductivity must decrease to l2ms (only depending on the design of the breaker) to make a successful interruption possible.

M&nkirtxja

lntroduution

High-voltage circuit breakers (CBs) have a vital function in isolating faulted sections of power networks. At slioti-cir cuit current zero (CL). the instantaneous energy input into the arc is minimal, enabling the arc to extinguish. immedi ately after the extinction of the arc, the power network macis with a transient reeoay voltage (TRy) that stresses the gap. The ability of the CU to interrupt is verified in highpower laboratories. Although the main aim of high-power tests is to prove the ability of a Cli to pertdrm a gisen switching duty. information on the quality of that per formance is usually not available. An international research program was set up by KEMA and sponsored by the European Commission, aimed at extracting the miviruumn possible amount of information on the interruption process during high-power tests. This information is presented and ucrformance margins of three different CBs are tnesti ated. Measuringsystem

flexible visualisation of arc current, voltage, conductivity, is possible [2]. The digitising part of the system, placed on floating potential in the close vicinity of the test object, stores the data locally. Immediately after the test, data is sent through an optxDal fibre to the control centre. Current levels down to 5OmA could be measured.
etc.

Measurements

A program of high-power tests set up a database of CZ information of a number of commercially available modern high-voltage SE Clis of various technology (puffer-type, double nozzle type and sellblast type). Tests were per formed under 9(1:4 short-line fault (SIT) conditions (IEQ 64)056) at 6(1Hz. In Table 1, an overview is giwn of the number of tests, number of current zeros for each CB, amid the number of resultant interruptions (mt.) and reignitions (Reign.). All breakers were in new condition before the
tests.

2 itt

Table?: Overview of tests

Ratin of CR
72.SkV@31.5kA 123kV@315kA I45kV@31.f,tA 145kV@31,skA

A ncw measuring system of high ceuracy and high resolu lion wes developed, based on KEMAs long experience [tJ. With the aid of an intelligent triggering module, with this system, current and voltage during up to four consecutive CZs having a total window of 6m are automaticeily recorded with a sampling frequency of 40MHz combined with a 12 bit resolution, A specially designed wideband (> 10 MHz) Rogowski coil is uted as a current sensor: the arc voltage is measured with wideband ( 5MHz) voltage dividers. With custom-made signal processing software, a
0 1FF, i4l lEE Iir c4e m. 3300)233 noz lOi&4grd2n333238

Short
CR72 CR123 CR145 CR145

CircuIt
d:rcl thetic direct aynthetic

Tests
28 33 2? 27

1 tnt. CZ
64 33 50 27 24 24 26 21

ReIgn,
40 9 24 6

Papx issi rrxxwcd rkh Jth zind n rc. teen 00 Nastrnbc iJ)
h s5h KbMA tti-ti Lthxtee, Unsc 310. 0112. ,R Ara01ra the
IFS ?-Gci. Sie

An example of a tea: result is the arc current and voltage during three consecutive CZs in one teat shown in Fig. 1. Two reignitions labelled I and 2 (after arcing times of 7.9 and 16 3ms) and one interruption labelled 3 (after 24.6rns) arc showiL In this example, a p0513cc current of several amps distorted the TRV heavily. CBs in a good condition normally produce postare currents below lOOmA or so. In the event of unusually large values of postare current (aS in hg .; was the cxpDrierec that the breaker chamber is very near to the cod of itS ability to clear thc SLF.

D,.,ra l,L 14:0

35

25-

I
-

known. Because of the extreme rapidity of prrxces in SF 6 gas, presumably oni energy exchange takes place during the fast period. Flow and other complex 3D processes are not likely to play an essential role. Therefore a lumped model could suflice, accomplished by a set of ordinary dif ferential equations (ODEs) [6]. Aiv-dicuk interaction is obviously strong and the time constants are very smalL Consequently, in the arc-circuit interaction only those fac tors play a role that are related to fast changes, like para sitic elements close to the breaker, the rate of rise of the current close to (2, and the initial rate of rise of the lRV. In our opinion higher-order ODE modelling is necessary because the fast period itself has quite diWcrent time scales oving to varying conditions along the arc channeL 4.2 Outline of new model Making extensive use of the new C7. data base, it was po sible to dcrivc a new (fast) black-box model, based on a combination of the classical proven approaches by Mayr [7] and Cassie [8]. Tne parameters of this model have been extracted from esperimcntal data creating the engineering part of the model. The new model consists of three modified Mayr models connected in series representing three series arc sections each with three parameters: the time constant 7 [s]. the quantity fl [AV ) related to power loss and the dimen 55 sionless model parameter A, (1 index for each submodel)

5 E4ma from CZ, s

4W O4rrrnf

Fin. 1 I,ir,cra 3) rc ctii zri

Afl. viLre
ir WTtfli

v,d

6u&g
RV

erc

vfrir

mu-

oxg nom.i TUV

tzt (m n

o1d17
np4Io w.r.h

1, 2,

Noi ThV. b Arc curit

New arc model

4 -G

1,2 and 3 (1)

4.1 Model philosophy The new model is foundcd on classical results. A great deal of work on arc modelling was earned out by CIGRE WG 13.01, resulting in the publication of a number of surveys of the state of the art [3 5]. Considering the very dilTcrcnt tune scales of the physical processes in the chamber, the modelling concept is based on the division of the whole interruption process into a period where slow arid a period where fast phenomena prevail. In the slow period the arc-circuit interaction is csscntiaUy weak, and the current is imposed by the circuit. Ibe fast period starts at some ten.s of trs bcft)rc (2. In the slow period (starting at the contact separation) the plivsic.al conditions (pressure butld up. arc temperature, etc.) arc set that determine the initial conditions for the fast period. The characteristic time constants of the slow processes are prob ably some hundreds of s or more, and for the fast proc esscs some p-s or less. A sophisticated slow model must contain a large number of parameters depending on the chamber design and internal condition, the gas pressure. and the stress factors [ike current history and arcing time. It is most likely that the slow model is strongly (B depend ent. It can lx assumed that the fast processes arc detemitncd by the evolution of the relevant quantities during the stow rcriod and the properties of the SF gas, Consequently, the fast model is probably Ofliy weakly dependent on the CLI. and is determined mainly by the SF 6 properties, ha param eters should reflect the initial conditions determined by the slow period. In this sense. although only a tiny portion of the entire arcing time is isiken into account during the fast period model, the whole arcing history has a ftmdarncntal unpact on the phenomena in the last period. Based on this philosophy, the fast modal can be used in its own right, completely separated from the Iow modeL provtded the narameters reflecting the Initial conditions are

where (1, [S] is the conductivity, L4 [V] the arc voltage and r ]sJ is the time. [I A 2 cqn. I rcdiics to the classical Mayr model [7] with Ti P 0 [WI the usual power loss constant. If A 1 cqn. I is the Cassic model [8] and Ti , with U 2 L,, [VJ is the classical steady-state arc voltage. The following equations describe the dinxtly measurable current I and voltage U and the conductivity G:
-

J=UG

(2)

This model contains nine poruneters itt prinplc, however. Six of them proved to be empirical constants having the fol lowing values and relationships:
1 A
-

1.41
IfI
Il.

1.9;

An =2O

--

113

2 f 3 k 1

LIased on the measuring results, the first submodel was cho sen to be a MayTCassic type arc, the second one almost a pure Mavr type arc, and the third one is a pure Mayr arc. The alucs of remain the same throughout all the tests, The bicaker paramct.cts k, k and k 3 depend on the actual CLI design, and keep their validity during all tests on one breaker. see Tahie 2. T , fl, and LI,, are considered then to 1 Table 2: Breaker parameters Model parameters (constant in ati rests)
cruit brenkor jYsrrna:ers tvary from CB

A,, A ,A 2 3
k, 2 k 3 k

toCBt

CB72 C81Z3

5.7
4.

5 5 10

100 100 100

COlAS State
prrr1ecs tvar from test to test)

57

7), it,, Ii,

be the free parametem, deseribiiig the slate of the CU cliii ing each interruption attempt. These parameters have a large range, and depend on the actual tcst Jitions (rnaixthr arcing time, slope of arc Current thJdg denoted by 5) and the statc of degradation of the CR, icth&ss dear dilicreuces can be xcn reflecting the diffctent CU designs. in Fig. 2. the cumulative distribution of the meas 1 and the cooling power ured values of the time constant T fT of the third (pure Mayr) series arc have > 1 constant F been visunlised.
100

the result of which is predicted correctly. Fig. 3 illustrates the course of the described validation procedure.
,cu

syribeUc ls with r&gnOn

VOll2O

a,

Current zero
curr1i

e.4taie

I
loLl
lime oorsla.1i. r,s a

8Oj

1000

L pa(amaLE1
E aiernoZj

I
*_-_-_

C.

1od pr.dd:

rt
,-J I !i . 4
L-vi
rOQfl4CrI

sc
40

L__

0_ic 1.00 moo

olbi powwr loss. iW
C)

Fftj.3 sfoJwr

,,f dshrg rnethsiL Ixrc wt:h snthia 8121

riai

dukg

>

FIg.2
c 1 z c,xar.i T I crg XWt toss

nafrp

thdscr&s

i- ui

C1172 0 CB123
A

ii; (Ii,

K CB145

4.3 Model validation

The main idea is to make distinction betwixti the stochastic and the deterministic part of the interruption process. The actual shape of th arc channel in the critical time interval several i5) before CZ is a result of a stochastic process. (tI)However. once a specific channel has been developed, it is likely that the intemiption pws from that point is es.miiallv deterministic. Therefore the actual measured traces of cach CZ are taken to extract arc parameters that are exclusively related to that specflc case. Due to the fact that the very snrnll value of postaic current of healthy breakers cannot be measured accurately, only arc informa lion up to and including (7. s used. From the prezero rcon, the arc parameters of each indivalual test were extracted unng a special multidimensional suxcasive gradi cut method. Next, with these an: parameters, the model of eqn. I with eqn. 2 was used to calculate the maximum value f power frequency current slope 5 L that the arc would interrupt. Thcn this value is compared to the actual S during the test. lius if S < S the model predicts a reignilion. if S > S an interruption is prcdictci This prediction is ultimately con fronted with the real measured asuk of the interruption a tempt after some eprimisation stspcs resuitirie a art over ui modet perfortaanca ci 9 fEs 01 the rntcrrcptton at:cmpts,
JLS

I
Flg. 4
Essvsq*

of vc rwrent c&
CBJ,23

uliagc

iesvr#oXs aK QJ71lt

xro a

rniirsi (5?, ii(reinicri ,, ri.jts nCtr,k ..nip p.,rnis

observation is failure, a

prediction is interruption and so-ca&d mixed type of failure was observed (Fig. 4). The part of the arc channel that is deserihal by the third purely Mayr) equation of eqo. I having the smallest time constant is mainly responsible for the interrupt;eu and the (immediate) thermal reignition. Because of this small rime constant (30 200ns). this section uf the an: uoois down much faster than the other parts. it seems that rho failure s a dieieetnc reignition from the point of view of the Mar arc part, and a thermal failure if the remainder and still hot parts of the arc art considered, Another check of thc validity of the present model is its ability to simulate the correct wave traces. It was observed that from several hundreds of tis before C? up to (2 the. fitting is very good. The model is correct also in piedicting small post arc currents in accordance with the measure rncnLs. A typical example of the correct wave tracc simula fort is shown in Fig. 5 (Figs. fai and h show the same (2 with diflcrcnt timac scale.

In some cases, where

. 2. .4J.h Oi2 0 7wr Durth.. 5t )4, ,V

a C

I
5.0 voitagr

>

on

.7

-6

-5

-4

-3
b

.2

.1

progress of tests on every pole, reflecting the degradation of the breaker by the tests. (b) In all but two cases (indicated with wrong model pre diction) the new model predicts from it.s pceto input suc cessfully an interruption or reignition (c) The rate of degradation is clearly depending on the amount of powcr-frcquer&y current stress (pole A 24.4, pole 13 211, pole C l7,4k%?.& From these observations, the conecpt of interpreting the margin quantity M as a valid measure of CB degradation is clear. As another cxample, the margin quantits of the inter ruption attempts plotted in Fig 1 arc gin: Atz.ernpt 1: Al 0.17, attempt 2: Al -0.083 and attempt 3 (interrup tion) M 0.015. The latter (rv small margin) represents the deteriorated state of the chamber only marginally being able to clear the fault. In Fig. 7, the margin quantity for each interruption attempt is plotted against the arcing time. A steady incrcase of Al can be noted up to the minimum arcing time (around 8ms). followed by an optimum near 2ms, sliowmg the breakers best perforrnano.
-

a cwr,n: ,MS ivh

Fig.5 C 1r. iCBl2S,

(,z)

nsz,d ni

0.4
. 0.0
K a a

ms
scutuct wsehpes

ta, IOs

04

Interruption perfonnance evaluation

Since the new model can theoretically predict the maximum persible dud, (Si) thc arc is able to interrupt at the specific circuit conditions and at the specific state of the CB cham ber, the margin of interruption can be estimated when this quantity is compared to the actual 1il1, (5). A suitable parameter to quantily this margin is

10
drcl1g time, m gkwr prof

15

Fig.?
obsrvi

/tht M ,r:0
in irupim

53 olexvoi rdutim

asst arurg rrla7fir CR145

ci s
S

+5

(3t

The value of Mis between l and 1. Negalrve M indicates rerguition, positive value implies interruption. With the quantity M. it is possible to map the degradation of the breaker during the succession of test. An example (C1172) is given in Fig. (. Three poles (A, 1 and C) ci the same CII we tested using the same circuit.
A A

rnoda4 pector

A good candidate of a physical parameter that is indica for the actual ability of the breaker to clear the SLF is th conductivity of the breaker at CZ (G ). In Fig. 8, the 0 relationship between this residual conductivity is plotted against the margin M for the breakers under test, thus establishing a direct link between G 0 and the quality of the interruption. The dependenec of G 0 is only marginally on the stress factors. As a general rule, however, it can be con cluded from the actual measurements that interruption becomes possible when arc conductivity has dropped below about I -2mS (see 1-1g. 9. the enlargement around the inter rupt ion limit 0.05 e M c 0.05 of Fig. 8).
tive

0,25
*

6
j.A A

Summary and conciusion

5
test

Fig6 i-aiii ng.Sf ifcwair ixrakr CB?2fs- aKh ;s4- Lt5 i/ SL.E use the A used with 24.4 B cstnd th 221 C Tcst,J ,ih 74 LA)
A nbscivest niiiir,s

A abcrnd

The following sed

observations

from Fig. 6

must

be

cmplta

can ne mapped
24

(e 145kV) has shown that the pross of CB degradation by the present method. Further, it was made pLausible that the arc eonductisity at (2 (relatively
tAt Fru-Gt Tri Dw.th,. init Ut. N. 2. ntth

(varyng from test to test). With the aid of the model, margins of individual inter ruptions can be spciicd. Its application to connttcialJy available power CBs of various tedinologics and rating

with a tailor-made highfrequency high-resolution current zero data-aquismuon systent Data from approximately 180 tests were used for the parameter definition and evaluatictn of the new model, a series connection of three modiiicd Mayr e.uations, yielding only three free state parantcrs

A new empiric-al arc model has been presented that has been shown to predict correct results of standard (IEC 60056) SLF interruption tests based on preiern informa tion. Arc voltage and arc current wave tra were recorded

00 0 4

I0
U, N U

AA

1
.

rgviion prcckcd
10
.2

pc1
0 0.3

0.6

.0.9

-00

-0.3
m.ri iii

IntrrupaU,n M
with nsLcrnni:

Fig. 8
A.

Cz4123 S (I3i45
2000

CaT2

Crehi&s br.cani

cz,7rn-:Li

amni

aai ikrrnfsi

tn&sjzn

.,tjor a Irs

errand

Iotricd ksmqti

a
0

1
.000
a 0

50s0
F.9 FjsLzmen ona& irnatqvion A (tin
(-niI .ioio

o
OE leerruprion

0,025
M

0.050

in

cin:

015145

is a reasonable indicator for the margin of the interruption. Knowledge of the mar gin can be applied in the following fields:

easy accsssible for measurement)

development of CBs by manufacturing companies application of CBs in non-standard circuits and for non standard ratings estimation of the state of a breaker after caended service
development of future standards dcdcatcd high powcr testing and comparison ol tct-C1r cuils

(lborclcc cv.) and Dr. P. Scat-pa (Liege Univcr.sity) for the opportunity to do various types of cperimcnts in their lab oratories and to Prof L van der Sluis (tlfi University) for discussion. 8
I

Reterences

Given the wide range of the arc parameters describing thc

actual state of the GB, it is our opinion that a generally applicable arc model such as often offered in network analysis software does not, exist, not even for a single GB

desic,

Acknovvledgrnent

DAMSIRA. G.C., and KERTtSZ. V.: Dcdopncnt and applica on of a 10 MHr digital vstcni for Cura,it.rcro TncsuraT)cnW, lEE Pr-&i Mnay. Tcrhnr,L 1995, 142,(2). pp. l25132 2 SMEETh. R.P.P., E1NlNK. A.H., and KERTESZ, V.: A new high resolution high.frq.-y current-zero nxasuxirrg astc1n. Presented at the ERA coofcrcncc on 1 rrnei9 and calibration in high-idragr tc.rreit, 1995, Paper [2 CIGRE WOR.K.JNG GROUP 13.01,: Prtcal aliestioa of arc pbys:ca rn icuit breakcr. Surwy of cakuiat:on zrthods and zpplia ton guidc, Eiccrra. May i9S, ii: pp.66-79 (IGRE WORKING GROUP 13.0L: Apphcthara of btick bos rnodcflin to circuft brcakrs, Eiecrrrs. Augu..L 1994, 149, pp. 41lI 5 CIGRE WQR.jCJN(j GROUt 13.01, Sny on andticssJ and graphicil tooR for circuit beikrr behaviour description 6 HABEDANK, Applknkm of a new arc rondd for the evaluation of shcit-cutuo b.-cak:n tests. iEEE Trrs, 1993, PWRf4, (4), pp.
.

Thanks arc due to Dr. H. Knobloch (Siemens AG) and to Dr. C. Neumann (RWE Energic AG; for supplying some of the test objects. Thanks are also due to Mr. A. Even

9211925

hung, ,4rc*ir/ii, retcchnik, 1943. pp. 588608 CASSIL, AM.: Arc rupture and cixnat tiit: GRE report. 1939

MAYR. 0,: tr5

des iocna und ancr new tbeory.Cl

SEE

Trm

rs,ro.

Vsi. Sr N,, 2, t.ht 20C5)

125