Electrical Power and Energy Systems 61 (2014) 219–228

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier .com/locate / i jepes

Development of non-intrusive monitoring for reactive switching of highvoltage circuit breaker

http://dx.doi.org/10.1016/j.ijepes.2014.03.0480142-0615/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +61 738661213.E-mail address: jlopez-roldan@powerlink.com.au (J. Lopez-Roldan).

Jose Lopez-Roldan c,⇑, Ryszard Pater a, Sébastien Poirier a, David Birtwhistle e, Tee Tang d, René Doche b,Mark Blundell c

a Hydro-Québec Research Institute (IREQ), Montreal, Canadab Hydro-Québec TransÉnergie, Montreal, Canadac Powerlink Queensland, Virginia, Australiad Queensland University of Technology, Brisbane, Australiae Private Consultant, Brisbane, Australia

a r t i c l e i n f o

Article history:Received 1 November 2012Received in revised form 20 March 2014Accepted 24 March 2014

Keywords:Circuit breaker (CB)Reactive switchingControlled switchingRestrikesRe-ignitionsIn-service diagnosis

a b s t r a c t

High-voltage circuit breakers are among the most important equipments for ensuring the efficient andsafe operation of an electric power system. On occasion, circuit breaker operators may wish to checkwhether equipment is performing satisfactorily and whether controlled switching systems are producingreliable and repeatable stress control. Monitoring of voltage and current waveforms during switchingusing established methods will provide information about the magnitude and frequency of voltage tran-sients as a result of re-ignitions and restrikes. However, high frequency waveform measurement requiresshutdown of circuit breaker and use of specialized equipment. Two utilities, Hydro-Québec in Canada andPowerlink Queensland in Australia, have been working on the development and application of a non-intrusive, cost-effective and flexible diagnostic system for monitoring high-voltage circuit breakers forreactive switching. The proposed diagnostic approach relies on the non-intrusive assessment of keyparameters such as operating times, prestrike characteristics, re-ignition and restrike detection. Transientelectromagnetic emissions have been identified as a promising means to evaluate the abovementionedparameters non-intrusively.

This paper describes two complimentary methods developed concurrently by Powerlink and Hydro-Québec. Also, return of experiences on the application to capacitor bank and shunt reactor switching ispresented.

� 2014 Elsevier Ltd. All rights reserved.

Introduction

Field experience has shown that capacitor and shunt reactorswitching are among the most severe duties of high-voltage circuitbreakers (CBs) because of frequent dielectric and mechanical stres-ses that may lead to failures compromising personnel and equip-ment safety. Shunt reactors and capacitor banks are generallyswitched frequently, often daily. Overvoltages may occur due toinrush current or restrikes in the case of capacitor banks and cur-rent chopping or re-ignitions in the case of shunt reactors [1,2].Significant oscillatory transient overvoltages due to internal volt-age amplification in EHV transformers during capacitive bankswitching have been reported recently [3,4]. It has also beenreported that current-limiting reactors for fault current protection

can also generate transient recovery voltages (TRV) exceeding theCB’s ratings [5]. Therefore, it is of interest to monitor the conditionof these stressed CBs.

Monitoring of system voltage and current waveforms duringswitching can provide, if acquisition is performed in sufficientlyhigh resolution, valuable information about the magnitude and fre-quency of re-ignitions or restrikes that could be a precursor tointerrupter failure. It is possible to undertake full voltage and cur-rent site tests to record voltages during switching. While tests yieldsignificant information they require expensive equipment and con-siderable shutdown time. De-engergization of lines or substationequipment is more and more difficult in today’s generally loadedHV networks.

Moreover, the decision to install an on-line monitoring systemon a high-voltage CB is not straightforward, either, as it is notalways economically effective. In fact, typical monitoring systemsare not well suited for CBs already in service since they involve

220 J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228

expensive retrofitting in order to add intrusive sensors such aspressure and contact travel sensors.

Therefore, there is a need for a simple and cost effective CBmonitoring system that will provide information about interrupterperformance at high-voltage levels, such as re-ignitions, and notlimited to operating mechanism timing measurements.

Experiences of failures of high-voltage CBs during shunt reac-tors and capacitor bank switching, combined with increasingrestriction on the availability of high-voltage equipment for inves-tigation and inspection have led two utilities, Hydro-Québec inCanada and Powerlink Queensland in Australia, to explore newnon-intrusive methods to assess CB condition [6,7].

Failures of HV SF6 circuit breakers during reactive switching

Issues for capacitor switching

CBs installed on capacitor banks are frequently subjected tosevere dielectric stresses during current interruption. The TRV peakcan reach twice the nominal voltage across each pole in the case ofa grounded neutral capacitor bank and two times and a half acrossthe first-pole-to-clear in the case of a non-solidly groundedcapacitor bank. Most modern CBs are designed with a very lowprobability of restrike (class C2) [8]. However, degradation withage and wear may lead to a deterioration of dielectric withstandof the CB and may increase the restrike probability up to an unac-ceptable level over the years. On some occasions, restrikes lead toovervoltage which could potentially damage neighboring equip-ment [3,9].

Over the last several years, some critical failures of SF6 live tankCBs switching capacitor banks have occurred on Hydro-Québec’stransmission network in Canada. These critical failures were theresult of an incomplete closure [6] or restrike of the CB causingan internal arc fault. For example, Hydro-Québec has experienceda case of a catastrophic failure of an SF6 CB which exploded afterswitching off a 120-kV shunt capacitor bank. The investigationsbased on digital fault recorder data and visual inspections of inter-rupter fragments made clear that multiple restrikes in two poleshad occurred. These repetitive restrikes led to overvoltages slightlyhigher than 4 p.u. Such restrikes may cause important damage onCB’s main contacts and nozzle.

Issues for reactor switching

Bachiller et al. [10] conducted a series of field tests to examinethe performance of a high-voltage CB during shunt reactor switch-ing. Switching tests were made on a 110-Mvar, 420-kV shunt reac-tor and re-ignitions were observed in 25% of 68 three-phaseswitching operations. Tests on a 150-Mvar, 420-kV shunt reactorresulted in six re-ignitions for 26 switching operations. In thereview of experience from the UK National Grid Company [11] Reidet al. commented that they have observed very fast re-ignitiontransients of up to 5 p.u. in the 400-kV system. They suggest thatin the calculation of re-ignition currents and voltages, it may benecessary to use distributed circuit parameters to represent thereactor and the busbar-side connections.

Work by Spencer et al. [12] identifies causes of CB failures underreactor switching applications has shown that modern single-interrupter SF6 CBs may be affected over time by high frequency-arcing produced by repeated re-ignitions. The tendency of arcs tooccur outside the nozzle was shown to be affected by the abilityof the polytetrafluoroethylene (PTFE) nozzle material to absorband store small quantities of negative charge. The stored chargedistorts the field around the nozzle and causes re-ignition arcs to

occur external to the nozzle. Spencer found that charge trappedin PTFE nozzles could be stored for up to 2 years.

Powerlink has as well experienced several catastrophic failuresof high-voltage SF6 CBs used in shunt reactor switching operationson its 275 kV network.

A 300-kV dead tank SF6 CB catastrophically failed in the openposition several hours after switching off of a 30-Mvar, 275-kV shuntreactor. The CB had completed about 1800 switching operationswhen the fault occurred. One pole exploded and was completelydestroyed by arcing while the two other poles remained in theirentirety. An internal view of one of the surviving interrupter nozzlesshowed that the inside of the interrupter nozzle was severelydegraded by arcing. There was also clear evidence of arc damageon the moving contacts outside the nozzle and puncture of the noz-zle appeared to have occurred at a point close to the contacts.

In another case [13], failure to ground occurred within an inter-rupter of a 275-kV dead-tank CB switching a 5-limb, star-con-nected, line shunt reactor earthed via a neutral earthing reactor.During a routine opening operation, the dead tank CB failed to clearon phase A and subsequently faulted internally to ground. This CBhad been in service for over four years and had been operatedalmost daily.

During the fault investigation and breaker disassembly, clearindications of severe arcing puncture were found in the nozzlesof the interrupters in A and C phases. This nozzle damage appearedto have occurred prior to failure possibly due to re-ignitions duringopening operations. It was deduced that failure of the last pole toopen was due to a puncture of the PTFE insulating nozzle betweenthe moving main contact and the fixed arcing contact of the inter-rupter. The 50 Hz current within the nozzle was extinguished butionized gases were considered to have been forced though thepuncture by pressure rises due to the puffer. It seemed that subse-quently a power-frequency arc established between the main con-tacts outside the nozzle and out of the effective area of arc coolingand control and the catastrophic failure followed.

Catastrophic failure modes

In case of an unsuccessful opening or an incomplete closure of alive tank CB, the arc current could remain across a gap in the inter-rupter limited to normal load level and could last until the ruptureof the enclosure due to an overpressure or a thermal shock to theporcelain insulator. The presence of an internal arc could remainundetected by the grid protective relays because there is no signif-icant current disparity between the normal and fault condition.Generally, live tank CBs do not have any pressure relief device.An internal arc in an interrupter of high-voltage live tank CB withporcelain insulators can cause major consequences: risk of com-promising personnel safety, system outage and damage to sur-rounding equipment.

For dead tank CBs the risk of major consequences is reduced asany internal arc fault will rapidly develop to a line-to-ground faultand will be cleared by the grid protection. In case of overpressurethe gases will be released through a pressure relief device. Never-theless, the catastrophic failures are possible because of high arcenergy due to fault current which could be released inside a CBenclosure.

Transients electromagnetic emissions due CB switching orelectric discharges

Previous work

Transient electromagnetic emissions (TEE) due to CB switchingin substations were studied in the 1980’s and the 1990’s. This

Fig. 1. Schematic of equivalent circuit of CB interrupter.

J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228 221

research was essentially motivated by the need to assess electro-magnetic interference that could affect the reliability of wirelesscommunication and electronic equipment newly installed inhigh-voltage substations [14,15].

More recently, this field of research has been reactivatedbecause TEEs have been shown to be useful for the assessment ofCB operations. For example, Furlong et al. [16] suggest that instantsof contact separation on each of the three phases could be identi-fied by a time–frequency analysis of RF emissions produced duringa CB opening operation. Chapman [17] proposes a method todirectly measure the arc duration of an opening operation bymeans of an electric field sensor mounted in the vicinity of theCB. According to the author, the effectiveness of this method relieson an optimal coupling of the electric field sensor and the RF signalemitted by the electric arc.

Moore [18] proposes assessing the closing times for each pole ofa live tank CB by means of four antennas spaced around the periph-ery of the CB. The closing sequences between the three poles wereperformed by means of location of the emission source using aTime Difference of Arrival (TDOA) technique. Portugues et al. [19]propose a method based on TEE for partial discharges monitoringof substations. Results of two study cases are presented for400 kV and 69 kV substations. The above mentioned studies high-lighted a major benefit of TEE-based diagnostics, namely its abilityof non-intrusive operation. Thus, it offers the possibility of retriev-ing relevant diagnostic information about a CB while it remains inservice.

Description of physical phenomena

UHF transient electromagnetic emissionsSeveral transient phenomena occur in the CB interrupter: dis-

ruptive discharge (prestrike, restrike and re-ignition), sustainedarc instabilities (arcs during load current interruptions and arcsduring faults) and partial or corona discharges (generated inresponse to overvoltage). In this paper only disruptive dischargesare analyzed.

The disruptive discharge is characterized by the fact that ini-tially no current flows through the circuit and voltage is presentacross the contacts. On closing, the capacitance and the electricfield increase between the two approaching contacts. When thedielectric breakdown occurs, the accumulated charges cause ahigh-frequency transient current and a wideband TEE is generatedindicating the exact instant of prestrike. A similar phenomenonmay occur after opening if the TRV becomes higher than the gapwithstand voltage. In this case, the corresponding TEE will indicatethe instant of restrike or re-ignition.

Following the Toepler model [20], the gap between the CB con-tacts in an interrupter during a dielectric breakdown can be mod-eled as a time-dependent resistance (R(t)) which changes its valuefrom infinity to the low resistance of pre-arc in a few nanoseconds.Looe et al. [21] has modeled the interrupter as an equivalentlumped circuit RLC which is depicted schematically in Fig. 1. Thebreakdown in the interrupter, i.e. change of the gap resistance, isequivalent of applying voltage ramp across the capacitance C. Looeshowed that the response of the circuit on perturbing voltage rampof rise time 1 ns has the form of voltage oscillation at a frequencyabout 400 MHz and a decay time of 10 ns. Moreover, the frequencyand the decay time are similar for fully opened and half-openedinterrupter contacts. Hundreds of measurements on differentlive-tank CBs on the Hydro-Québec grid confirm the presence ofTEEs with dominant frequency around 400 MHz during prestrikes,re-ignitions and restrikes.

For live-tank CB, the TEEs generated by these high-frequencyoscillations are little affected by interrupter housing because itswalls are made from dielectric material, generally porcelain or

polymers, which let the electromagnetic (EM) wave pass through.As the antennas, placed at the distance of several meters from CB,are in far field of this radiation, the usual EM radiation propagationlaws can be applied including the location of the source.

Switching electric phenomenaConcurrently to phenomena generating TEEs, the results of

slower switching phenomena, such as inrush current and TRVcould be monitored. Especially for reactor and capacitor bankswitching, lower-frequency transients contain a great deal of infor-mation about the sequence of events which could help determinethe time of occurrence of restrikes with respect to other electricphenomena such as current interruptions. The frequency rangespreads in this case from industrial frequency (50 Hz or 60 Hz) tohundreds of kilohertz or a few megahertz. These phenomena couldbe measured by capacitive coupling. In high-voltage substation theelectric and magnetic fields are generated by many sources andprecise modeling is very complex [22]. In steady state analysis, itwould be difficult to distinguish the contribution of each polebecause of the crosstalk. However, in switching transient phenom-ena the contribution of each pole follows the CB operation timingand parameters such as making instant and TRV can be observedfor each pole using capacitive sensors installed below the busbars.

Measurement techniques

Capturing and analyzing high-frequency TEEs and lower-fre-quency transients can be a very effective non-intrusive diagnostictool for detection and location of re-ignitions and restrikes in theinterrupter. Therefore the simultaneous use of high frequencyantennas, low frequency capacitive sensors with large bandwidthand measurement of line current can provide valuable informationfor CB monitoring. Different antenna solutions could be appliedaccording to the CB type: dead tank or live tank.

Two complimentary monitoring methods, developed concur-rently by Powerlink and Hydro-Québec, are described. The Power-link method uses an active high-frequency active antenna (HFAA)and three low-frequency passive antennas (LFPA), which could beconsidered as capacitive sensors of electric field. Its astuteness isthe analysis of three phase voltage image waveforms allowingper pole investigation. The Hydro-Québec method uses four pas-sive UHF antennas (HFPA) and is based on location of the sourceallowing the per interrupter analysis of live-tank breakers.

Monitoring using high-frequency active antenna and low-frequencypassive antenna

The active antenna designed by Powerlink for this applicationfor capturing radiated TEE consists of a passive dipole antennaand a broadband RF amplifier. The passive dipole is 150-mm long,with a diameter of 1.6 mm (Fig. 2a). The RF amplifier is designed tohave a frequency response from 1.5 MHz to 1.5 GHz, and a signalgain of 12 dB.

In operation, the RF amplifier acts as an impedance transformer,converting the high output impedance of the electrically shortdipole at the low frequency range to a low-impedance output. At

Vs

C1

CrR Vo

Measuring Instrument

Fig. 3. Equivalent circuit of passive antenna.

222 J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228

the high frequency end, the RF amplifier merely amplifies thereceived signal.

HFAA is particularly useful for measuring the signals of lowamplitude relatively to background emissions such as noise andtelecom. This is generally the case of dead tank CB where the TEEis generated inside the metal grounded enclosure which attenuatesthe emissions.

The passive antenna, or capacitive electric field sensor, con-structed for this application consists of a metallic cylinder enclosedat both ends mounted on a plastic insulating tube (Fig. 2b). At thepower line frequency, the passive antenna head is effectivelycapacitively coupled to adjacent HV power lines and busbars(Fig. 3). Thus the optimal configuration is three LFPAs installed infront of each pole.

Referring to Fig. 3, if C1 is the coupling capacitance from a pas-sive antenna to a HV line, the measured voltage V0 across the inputof an oscilloscope is given by:

VO

VS¼ C1

C1þ Cr

� �jxRðC1þ CrÞ

jxRðC1þ CrÞ þ 1

� �ð1Þ

where Vs = HV line voltage, R = input resistance of oscilloscope, andCr = combined cable capacitance and input capacitance ofoscilloscope.

The voltage transfer function of (1) acts like a high-pass filterwith a 3-dB corner frequency at

f ¼ x2p¼ 1

2pRðC1þ CrÞ ð2Þ

In the trials described in this paper, the voltage between theantenna head and ground was measured across a parallel R-C ter-mination, values of which were chosen to restrict the measuredvoltage to not more than about 10 V and to produce higher gainfor restriking transients than for 50 Hz. As a capacitive coupler,the passive antenna has a high-pass frequency response. Its 3-dBcorner frequency was approximately 500 Hz when a 30 m dou-ble-shielded coaxial cable was used. Due to cable losses, the effec-tive 3-dB bandwidth of this passive antenna is approximately50 MHz. However, the LFPA is capable of operating at higher fre-quencies if higher losses are permissible, allowing also the captureof some TEEs.

The monitoring consists of synchronized recording of both sig-nal from one HFAA and three LFPAs.

Fig. 2. Active antenna (left), passive antenna (right).

Diagnostic system using high frequency passive antennasAlthough capacitive sensors (LFPA) can also be used for a

live tank CB, the real advantage of this configuration withoverhead separated interrupters would be taken if one can dis-tinguish which interrupter is actually generating the TEE.Hydro-Québec explored such a method using passive UHFdipole antennas.

Four low-cost HFPAs are installed at least 1 m above theground at a maximum distance of 25 m from the CB and con-nected to an oscilloscope by 50-Ohms coaxial cables (Fig. 4).Two configurations are used: antennas placed around the CB oron beside the CB. The latter is appropriate to monitor severalCBs simultaneously. The location is performed by precise steep-front detection (Fig. 5) and evaluation of the TDOAs, similar tothat described in [18]. To locate the interrupter, the TEE shouldbe recorded with high sampling rate of 1 GS/s with adequatefiltering.

In order to perform a comprehensive analysis of CB operation,the TEEs recordings are completed by measurements of line cur-rent and opening/closing coil current. Detailed interpretation ofthe coil current itself can provide relevant information on the CBcondition [23,24]. The rising edge of the coil current is used asthe trigger signal of the acquisition system. The currents are cap-tured by passive AC current microclamps and recorded at 16 kS/s. The coil current is not a sinusoidal signal thus recording it withan AC sensor requires a special interpretation (Fig. 6).

Each recorded TEE is transformed into a discrete temporal sig-nal associated with a specific interrupter of the monitored CB.These discrete signals are superposed on current measurementsand are interpreted in the context of CB operation (Fig. 7).

Application to reactor switching monitoring

Following is a description of the setup and results of severalswitching tests performed on a 275-kV reactor on Powerlink’s highvoltage network.

Site arrangements

The CB shown in Fig. 8b is a 275-kV dead tank type switching a30-Mvar directly grounded 5 limb shunt reactor. Controlledswitching was employed for both closing and opening. The voltagemeasuring system shown in Fig. 8a consists of three LFPAs, oneHFAA, one digital oscilloscope and a portable power supply.

Signals from the three LFPAs and the HFAA were connected to afour-channel 100-MHz 2.5-GS/s oscilloscope. The HFAA channelinput provided the oscilloscope trigger.

One passive antenna was located under each phase of thebreaker bushings on the load (reactor) side as shown in Fig. 8band the active antenna was located below the centre phase.

C2

C1

B2

B1

A2

A1

Ch 4

Ch 3

Ch 2

Ch 1

-9

-7

-5

-3

-1

1

3

5

7

9

-9 -7 -5 -3 -1 1 3 5 7 9

[m]

Fig. 4. Measurement setup with four antennas installed around a 230-kV SF6 CB; the positions of antennas are marked by Ch 1, Ch 2, Ch 3 and Ch 4 and of interrupters by A1,A2, B1, B2, C1 and C2.

5.6 5.8 6 6.2 6.4

x 10-7t (s)

0 0.5 1 1.5

x 10-6t (s)

Fig. 5. Example of TDOA evaluation.

Fig. 6. Interpretation of coil current recording.

J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228 223

Closing time phase A

Fig. 7. Occurrence of TEE (lower graph) and currents (upper graph). Coil and line currents have different scales.

PA

a b

AACB

Fig. 8. Reactor switching field tests: general test arrangement (a) and positioning of HFAA and LFPAs under the CB terminals (b).

224 J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228

Shunt reactor closing

Fig. 9 shows the closing signals detected by each LFPA duringclosing of the shunt reactor CB. The three high-frequency pulsesindicate the instants at which current starts in each phase. Fromthe Fig. 9 it can be seen that the peaks of each of the high-fre-quency signals are clipped because of the strength of the HFAA sig-nal. The time differences X1 and X2 in Fig. 9 are 3.7 ms and 3 msrespectively, whereas the controlled switching system was set togive equal time separations of 3.3 ms between each contact make.The different values shown in Fig. 9 were attributed to prestrike.

Shunt reactor opening

Fig. 10 shows the waveforms recorded by the LFPAs duringopening of the CB controlling the shunt reactor. The opening pro-cess can be divided into three periods. The first period is beforeany current interruption which is before 26 ms. Transients beforethis time are due to arc instability oscillations prior to currentinterruption.

The second period from 26 ms to about 33 ms is the time fromthe interruption of current in the first pole to clear to current inter-ruption in the last pole to clear.

The 1.03 kHz oscillation that can be seen in Fig. 10 is typical ofthe reactor oscillation due to current chopping. The large differ-ence between 50 Hz and transient peaks is due to the characteris-tics of the measuring circuit.

Beating voltage oscillations with slightly different frequenciescan be clearly seen in Fig. 10. The beating may be due to mutualcoupling between individual phases.

Fig. 10 shows an expanded part of the waveforms from 18 msto 38 ms. It shows more detailed information on the current inter-ruption region. The transient after the first current interruptioncan be clearly seen at 26 ms. At this time the largest transient isobserved on the LFPA adjacent B phase, establishing this as thefirst phase to open. This is followed by an oscillation which startsat 29.3 ms and has the largest value on the LFPA closest to Cphase; this indicates that C phase is the second pole to clear.The third pole then is interrupted at 32.7 ms, the largest transientvoltage being in this case on A phase. The opening sequence forthis test is determined from Fig. 10 to be B–A–C which is thesequence set by the controlled switching system. The time differ-ence between the current interruption points can be easily deter-mined: they are indicated as X1 and X2 in Fig. 10. X1 wasmeasured to be approximately 3.3 ms and X2 was measured tobe approximately 3.4 ms.

Fig. 9. Voltage waveforms recorded by LFPAs during shunt reactor closing from 26–44 ms.

J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228 225

Detection of restrikes/re-ignitions

A controlled switching system was used to reduce the arcingtimes of the test CB in order to generate re-ignitions, care beingtaken to ensure the arcing times were chosen to avoid largere-ignitions or restrikes that might be harmful to the CB. Theinitial setting gave an arcing time of 2 ms and this was increasedin 1 ms steps up to a maximum of 4 ms. Fig. 11 shows a re-ignition record on phase A during an opening with a 3 ms arcingtime.

The two points in Fig. 11 in which sudden voltage collapsesprovides evidence of the occurrence of re-ignitions. Due to thelow value of current and the contact separation, the dischargesproduced by the re-ignitions are immediately extinguished andvoltage increases after each re-ignition. Each re-ignition producesa step voltage coupled in each of the phases (shown in the threeLFPA records), allowing phase A to be identified as the re-ignitionsource.

Application to capacitor bank switching monitoring

A series of measurements was performed by Hydro-Québec onfour models of live tank CB used for switching capacitor banks at120 kV, 230 kV and 315 kV. Up to 500 switching operations atnominal current were measured. The collected data reveals thatseveral indicators for CB diagnostic can be derived. Two sets ofindicators are presented: timing of CB operation and restrike andre-ignition detection.

Timing of CB operation

Prestrike characteristicsAs explained in [6] and in Fig. 12, there are two important

events at closing: prestrike and contacts touch. The times of thoseevents are denoted by tp and tm respectively. While the secondevent is purely mechanical, instant of prestrike depends on the

Fig. 10. Voltage waveforms recorded by LFPAs during shunt reactor opening from 0–100 ms (left) and detail (right).

Fig. 11. Double re-ignition in phase A. Recorded by LFPA and by the HFAA.

-0.0

01

-0.0

006

-0.0

002

0.00

02

0.00

06

0.00

1

0.00

14

0.00

18

0.00

22

0.00

26

0.00

3

[ms

t 0

t p

t m

RDDS

Fig. 12. Evaluation of evaluation

226 J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228

voltage across the contacts. Therefore the pre-arcing time variesaccording to the voltage angle at the initiation of closing. Fig. 12shows how to evaluate prestrike times. It is assumed in the modelthat the rate of decrease of dielectric strength (RDDS) due to theapproaching contacts and the contact touch time are constants.In Fig. 12 the curve represents the voltage across the CB open con-tacts and the sloped straight line is the RDDS. It could be observedthat tp varies between t0 and tm, where t0 is assumed constant andrepresents the time from the closing coil current initiation to theprestrike instant at the voltage peak (tp = t0 if breakdown occursat voltage peak and tp = tm if breakdown occurs at voltage zerocrossing). Pre-arcing time is equal to tm � tp.

The value of tp is function of the voltage angle at closing coilcurrent initiation which is random. For a CB without point-on-wave closing, the maximal value of tp over many CB operations isthe estimator of tm: max(tp) ? tm while the minimal value of tp isthe estimator of t0: min(tp) ? t0. For a CB with point-on-wave clos-ing the range of tp distribution will be narrowed to the values closeto the target value which, for a capacitive current, is voltage zerocrossing (tp � tm).

0.00

34

0.00

38

0.00

42

0.00

46

0.00

5

0.00

54

0.00

58

0.00

62

0.00

66

0.00

7

]

prestrike time distribution.

Fig. 13. Closing operation with resistors.

Fig. 14. Breaking arc and restrike related TEE observed on 120 kV SF6 CB.

J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228 227

TEE measurements allow the precise evaluation of the prestriketime (tp) for each interrupter. By monitoring the distribution of tp

over many CB operations, the RDDS, which reflects the contactvelocity, can be estimated. For a CB with point-on-wave closing,the proper operation of the control system can be validated. Thestatistical analysis required for this method can be easily per-formed on CBs switching capacitor bank as well as shunt reactorsas these operate frequently.

It should be noted that, for CBs with multiple interrupters inseries, the model in Fig. 1 applies only for the first making inter-rupter. The remaining series interrupters will breakdown withina few microseconds after the first breakdown because the voltageover the contacts will increase within a few hundreds of nanosec-ond. The method using TEE is very suitable for this analysis since itallows a very precise measuring of the prestrike delay between theinterrupters. The model in Fig. 1 is more complex for some partic-ular cases, such as capacitor banks with floating neutral, where thevoltage on the second and third making poles is affected by thebreakdown event on the first making pole.

CB analyzers evaluate CB operation timing in reference tomechanical contact touch which could not be directly evaluatedusing TEEs. Instead we measure, on closing, the moment of pre-strike occurring just earlier. The time difference between TEEsand contact touch is variable but smaller than a few milliseconds.Statistical analysis over several measurements allows for evalua-tion and validation of basic timing parameters on closing.

Capacitor bank switching with closing resistorsSome CBs are equipped with closing resistors in order to limit

transient overvoltages. The closing sequence has two phases:insertion of the resistors by closing of the auxiliary contacts inevery interrupter and short-circuiting of the resistors by closingthe main contacts. Such a sequence is presented in Fig. 13 wherethe prestrikes times are denoted by diamond shapes. The firstgroup of prestrikes occurring around 45 ms is due to the closingof the auxiliary contacts and the second group occurring around53 ms is due to the closing of the main contacts. The resultsobtained with TEE methods corresponds tightly to the mechanicalclosing operation times specified by the manufacturer [6].

Restrikes monitoring

The TEE method is very suitable for detection and monitoringrestrikes and re-ignitions. The frequency of occurrence can be pro-vided for each interrupter with exact timing within the operationsequence. Moreover, all types of restrikes and re-ignitions areobserved including non-sustained disruptive discharges (NSDD).Fig. 14 illustrates a case of restrike on a 120 kV SF6 live tank CBwith one interrupter per pole. Another example of restrike for a315-kV CB is given in [6].

Conclusion

Experience has shown that re-ignitions and restrikes producedduring reactor and capacitor bank switching of high voltage CBscan cause important damages to the breaker itself and to neighbor-ing equipment.

Two non-intrusive techniques for monitoring CB operation dur-ing reactor and capacitor bank switching, based on detection andanalysis of transient electromagnetic emissions, have been experi-mented. A first method uses three capacitive sensors of electricfield and a single high-frequency active antenna. It has a greatadvantage for dead tank CBs. The second method uses four UHFantennas. It is most suitable for live tank CBs and allows per inter-rupter analysis.

228 J. Lopez-Roldan et al. / Electrical Power and Energy Systems 61 (2014) 219–228

The following parameters can be determined:

� Occurrence of restrikes and re-ignitions.� Interrupter and pole making and breaking times.� Interrupter and pole closing resistors insertion times.� RDDS characteristic.� Controlled switching performance evaluation.

The significant advantages of those methods are: that equip-ment can be set up and measurements taken while the high volt-age equipment remains in service, that no physical connection tothe HV part of the equipment is required, and the measurementsystem is portable and can be readily moved between differentswitchgears and substations. It is also not restricted to any partic-ular model of CB.

The measurement system required for those methods is quitefeasible with today’s technologies. In fact, the performance of thedata recorder compares to midrange, high-speed digital oscillo-scope commonly available on the market. On the other hand, theaccessories needed to make measurements are simple and easyto fabricate (e.g. passive antennas). The two methods could be eas-ily merged into one system combining the features of each method.

References

[1] IEC 62271-110. High voltage switchgear and control gear – Part 110: Inductiveload switching; 2009.

[2] CIGRE working group 13.02. Interruption of small inductive currents. ElectraNo. 101. p. 13–38.

[3] Mitra P, De A, Chakrabarti A. Resonant behaviour of EHV transformer windingsunder system originated oscillatory transient over voltages. Int J Electr PowerEnergy Syst 2011;33(10):1760–6.

[4] Debnath D, De A, Chakrabarti A, Kothan DP. Studies on the impact of capacitorbank switching on grid connected transformers. Int J Electr Power Energy Syst2012;43(1):126–30.

[5] Chen S, Yang P, Xu W, Bian X, Wang W. Side effects of current-limiting reactorson power systems. Int J Electr Power Energy Syst 2013;45(1):340–5.

[6] Poirier S, Pater R, Doche R. A non-intrusive diagnostic approach to preventcircuit-breaker critical failures. In: Proceedings of the 2010 InternationalConference on Condition Monitoring and Diagnostics, Tokyo; 2010.

[7] Lopez-Roldan J, Birthwhistle D, Blundell M, Tang T. A non-invasive method fordetecting restriking: application to the switching of HV shunt reactors. IEEETrans Power Delivery 2012;27(2):541–7.

[8] IEC 62271-100. High-voltage switchgear and controlgear – Part 100:Alternating-current circuit-breakers; 2007.

[9] Ramli S. Investigation of circuit-breaker switching transients for shunt reactorsand shunt capacitor banks. M Eng thesis, Queensland University ofTechnology; 2008.

[10] Bachiller JA, Cavero E, Salamanca F, Rodriguez J. The operation of shuntreactors in the Spanish 400 kV network – study of the suitability of differentcircuit-breakers and possible solutions to observed problems. Paris: CIGRE;1994 [Report 23–106].

[11] Reid JF, Tong YK, Waldron MA. Controlled switching issues and the NationalGrid Company’s experiences of switching shunt capacitor banks and shuntreactors. Paris: CIGRE; 1998 [Paper 13–112].

[12] Spencer JW, Leclerc JL, Jones GR. High frequency discharges and their effectson insulation in SF6 filled circuit-breakers. IEE Colloquium Surf Phenom AffectInsulator Perform 1998;235:1–3.

[13] Lopez-Roldan J, Blundell M, Wing D, Birtwhistle D, Rangli M, Tang T. Analysisof modern high voltage circuit-breakers failures during shunt reactorswitching operations and corrective measures. In: Proc Conf MATPOST,Lyon; 2007.

[14] Ari N, Blumer W. Transient electromagnetic fields due to switching operationsin electric power systems. IEEE Trans Electromagn Compat 1987;29:233–7.

[15] Wiggins CM, Wright SE. Switching transient fields in substations. IEEE TransPower Delivery 1991;6(2):591–600.

[16] Furlong SA, Jones GR, Humphries J, Gibson JR, Lewis RG. Radio frequencyemissions from arcing in high voltage circuit-breakers. In: Proc XIII Int ConfGas Discharges and their Appl; 2000.

[17] Chapman M. Radio-frequency-based determination of arcing duration in HVACcircuit breakers. In: 18th Int Conf Electr Distribution (CIRED), Turin; 2005.

[18] Moore P. Radiometric measurement of circuit-breaker interpole switchingtimes. IEEE Trans Power Delivery 2004;19(3):987–92.

[19] Portugues IE, Moore PJ, Glover IA, Johnstone C, McKosky RH, Goff MB, et al. RF-based partial discharge early warning system for air-insulated substations.IEEE Trans Power Delivery 2009;24(1):20–9.

[20] Singha S, Thomas MJ. Toepler’s spark law in a GIS with compressed SF6-N2mixture. IEEE Trans Dielectr Electr Insul 2003;10(3):498–505.

[21] Looe HM, Brazier KJ, Huang Y, Coventry PF, Jones GR. High-frequency effects inSF6 circuit breakers. IEEE Trans Power Delivery 2004;19(3):1095–104.

[22] Ramli S, Birtwhistle D, Lopez-Roldan J, Tang T. EHV circuit-breaker conditionmonitoring using antenna-based coupling methods. In: Proc Second Int ConfCondition Monitoring Diagnostics, Beijing; 2008.

[23] Stephen B, Strachan S, McArthur S, McDonald J, Hamilton K. Design of tripcurrent monitoring system for circuit breaker condition assessment. IET GenerTransm Distrib 2007;1:89–95.

[24] Johal H, Mousavi MJ. Coil current analysis method for predictive maintenanceof circuit breakers. In: IEEE/PES Trans Distribution Conf Expos; 2008.