pd in cables

8/12/2019 PD in Cables

1/5

Partial Discharges in Low-Voltage Cables

J. P. Steiner

Biddle InstrumentsBlue Bell PA

Franois D. Martzloff

National Institute of Standards and TechnologyGaithersburg [email protected]

Reprinted, with permission, fromConference record of the IEEE International Symposium on Electrical Insulation, Toronto, 1990

SignificancePart 4 Propagation and coupling of surgesThis paper is only indirectly related to the topic of surge propagation in the context of surge-protective devices, but is included here as it might be a gateway to further information on pulsepropagation the basic technique used in the investigation summarized by this paper


2/5

Conference Record of the 1990 IEEE International Symposium on Electrical Insulation, Toronto, Canada, June 3-6 1990

P A R T I A L D I S C H A R G E S I N L O W - V O L T A G E C A B L E SJ.P. SteinerBiddle Inst rumentsBlue Bell PA 19422

ABSTRACT-Testing of high voltage apparatus for partial dis-charges has long been recognized as an import ant part of quali tycontrol for these devices. Recently, interest has been focused onmethods for testing low voltage cables to determi ne their integrityunder adverse operating conditions such as a loss of coolant ac-cident. A new method, utilizing partial discharges, is presentedwhich has the potential for locating breaches in the insulat,ion ofin situ, low voltage, multi-conductor cables.

I N T R O D U C T I O NCables used in power plants are selected on the basis of quali-fication tests l]which provide a sufficient degree of confidencein the abili ty of the cables to mainta in operational readiness fora well-defined life dura tion. As the cables age under the var-ious environmental stresses prevailing in a power plant, nuclearplants in particular, the question arises of how much residuallife exists in the cables as they approach their rat ed, qualifiedlife. There is considerable interest in assessing this readiness, anda number of investigators have proposed various test methods [2]In particular, a method that could allow nondestructive, in situassessment would be extremely valuable. The detec tion of par-tial discharges at identifiable sites along a cable offers a methodfor assessing the likelihood of a fault t o occur during a loss ofcoolant accident on a cable containing an incipient defect. Suchan incipient defect might be a breach of the insulation that wouldremain undetected under normal operating conditions.Low-voltage cables, in contrast with cables designed for high-voltage service, arc not provided with an insulation structureaimed at making them free from partial discharges under mo del-ate overvoltage conditions such as the application of a test vo lt-age. Therefore, the occurrence of a partial discharge at an incipi-ent defect site has to be differentiated from tlie background of theexpected partial discharges that will occur over the length of thecable in the absence of any significant defcct. The test met hoddescribed in this paper offers the opportuni ty to make this differ-entiation and locate th e discharge sites associated with insulationbreaches. Three representat ive types of cables in which artificialdefects had been created were subjected to the test; every defectwas successfully located.

P R IN C IP L E O F T H E M E T H O DPar tia l discharges will occur along the enti re length of a cablewhen voltages greater than the partial discharge inception voltageare applied. To develop a strategy for analyzing experimentaldata from a cable that also contains defects resulting in partialdischarges at specific sites, several assumptions need to be made.The first is that there are no preferred partial discharge sites inthe cable under test. Therefore, it would be expected tha t, ideallyand in the absence of any defect, the par tia l discharges would beuniformly distr ibuted along the length of the cable.

F.D. MartzloffNational Insti tuteof Standards and TechnologyGai thersburg M D 20899

The second assumption is that, in the absence of a defect, thepartial discharge pulses occur a s random events in time, i.e ., theobserved process is a marked random point process [3]. Usingthese assumptions causes the observed partial discharge processto resemble, asymptotically, a Poisson shot noise process [3 4]with pulses emanat ing from all positions along the cable. If asample of da ta werc analyzed by calculating a histogram of thepositions from which each pulse in the sample emanated, then,using this model, the histogram would correspond to a uniformdistribution.Th e third assumption is that amplitudes of the pulses, expressedin units of charge at each part ial discharge site, are gamma-distributed [4] with the parameters of the probability densitybeing independently and identically distributed. The observedprobabil ity densi ty of the charge will then have the form

where p(q) is the probability density of the observed charge inthe entire cable, p(q/z,w) is a conditional probability density ofthe charge at a single site, p(z) is the probability density of thelocations of th e partial discharge sites along the cable and p(w)is the probability density of th e parameter w such as the de-gree of imperfection in the insulation (local field enhancements)that control the most probable discharge amplitude q at thatsite. By assu~ npt ion, here is no preferred site and experimen-tal evidence indicates that the p hys ic ~l ehavior at each site issimilar. This suggests that the superposition of these infinitesi-mal processes will yield a seemingly well behaved, predominantlyunimodal probability dcnsity for the charge and this has beenobserved experimentally.Now consider the modcl for a defect. The goal of this measure-ment technique is to detect an insulation flaw which is defined,for these purposes, as a conlplete crack through the insulation t othe conductor. Partia l discharge in electrode arrangements th atconsist of a conductor-conductor interface or conductor-dielectricinterface is often more active than when the electrodes are adielectric-dielectric interface. The implication is tha t tlie partial-discharge sites at the flaws will be more active than the sites inthe rest of the cable and consequently cause significant deviationsfrom the ideal model.Th e basic idea of the nleasurement technique is to identify whetherthe data significantly deviate from t,he behavior defined by cqua-tion 1). In this case, equation 1) is modifird by adding a termin which p(w) is replaced by p(wlz)

where w depends on z and is highly localized to particular pointsz . .4dmittcdly, thc model is a gross idcalization of the cxprc ted

U.S. Government w ork not protected b y U . S . copyright


3/5

igure 1 Block diagram of partial discharge detection andrecording.behavior, however, the general concept has merit if th e defects dobehave differently from the normal background partial dischargein the cable. In practice, there will be naturally occurring, pre-ferred partial discharge sites that are not flaws but cause varia-tions in the expected background count rates. In fact two suchpreferred locations have already been identified, the ends of thecable.

E X P E R I M E N T A L A P P A R A T U SThe experimental apparatus implemented is a general purposesystem designed for acquiring broadband partial discharge wave-forms and processing the acquired signals as shown in the blockdiagram of Figure 1. Th e key elements are the high-voltage sys-tem, the coupling circuit and the data acquisition and analysissystem. The high-voltage system used t o apply the necessaryvoltages to the cables under test consists of a 60 Hz high-voltagetransformer with appr opriate noise filtering. Measurements onthe cables are made at voltage levels exceeding the partial dis-charge inception voltage by 10, 20, and 40 percent. Th e inceptionlevel is defined as that voltage which causes at least one partialdischarge pulse above a specified charge level q per second.The partial discharge signals are coupled from the cable undertest to the measurement system through a coaxially mounted,

nF capacitor. The capacitor is the first element in a high-passfilter having a lower cutoff frequency of 30 kHz and a 1000 in-put impedance. Initial testing with sensitivities ranging down to0.05pC did not produce useful results because the partial dis-charge magnitudes at inception are in excess of 10 pC. Since thepredominant partial discharge levels are greater than 10pC, anattenua tor and input protection are included in the circuit whichlimits the input sensitivity to 1 pC. Th e capacitance of t he inputprotection limits the bandwidth of the system to approximately10 MHz. This input bandwidth is adeq uate for these experimentssince the cables under test are lossy and limit the usable band-width to within this range. A variable attenuator is includedbetween the inpu t preamplifier and th e gain block of the systemto provide ampli tude scaling which prevents saturat ion of theamplifier for larger partial discharge amplitudes.Par tia l discharge pulses a re recorded using a LeCroy 9400A1 digi-tal oscilloscope which has an 8-bit digitizer t hat can sample 32 000

The identification of comm ercial materia ls and their sources is ma de todescr ibe the exper iment adequate ly . In no case does th is ident i f ica t ion implyrcomm endat ion by the Nat ional Inst i tu te of Standard s and Technology nordoes i t imply tha t the in st rumen t is the best avai lab le.

points at a 100 MHz rate. Th e advantage of using this oscilloscopeis its ability to segment its 3 kB acquisition memory so that itcan acquire up to 250 separate waveforms at its maximum sam-pling rate before having to transfer them to a computer over anIEE E 488 interface. Th e computer is a PC-AT class computer.Th e ASYST scientific software package is used for acquisitioncontrol and da ta analysis. Up to 20000 waveforms can be col-lected before the available disk space limits the computer s abil-ity to process the data; typically 5000 waveforms are collected.Active da ta collection occurrs only during a few seconds of thevoltage application to the cable under test, however, voltage istypically applied for 8 minutes since the high voltage system isnot under control of th e computer and t ime is needed to archivethe dat a. Since the sys tem uses a computer with only moderatecomputational ability, all processing has to be performed off line.

M E A S U R E M E N T STwo fundamenta l estimates are made from each recorded wave-form: an e stimate of t he charge, and an estimate of the locationof t he partial discharge site in the cable. It would be desirable tocalculate estimates for all detectable pulses; however, since themeasurement system is general purpose, trade-offs for cost, speed,and d at a handling capabilities have been made. With these con-straints, only a limited number of waveforms can be collected andanalyzed.The meth od used to calculate the charge in the partia l dischargewaveform is the same as th at used in [4] and has been shown to bemore accurate than the usual measurement of the peak value ofthe pulse. The technique uses the pulse energy derived from thewaveform to develop an est imate of the charge and assumes tha tthe measured partial discharge waveform is similar to calibrationpulses injected into t he cable. Th e calibration pulses are gener-ated by applying a fast-rise step of known amplitude through aknown capacitance into th e cable under test. Further correctionsin the estimates to account for the cable attenuation were notdeemed necessary.If the flaw is detectable, then the histogram of the charge willno longer be unimodal. Using this fact, the measurement can beenhanced by sett ing a lower discrimination level so tha t the in-strument measures only those partial discharges which appear tobe anomalous . The lower discrimination level is also referred to asthe trigger level, the level which the partial discharge pulse mustexceed to init iate th e recording cycle of the digital oscilloscope.The basic principle used to determine the location of the partialdischarge can be understood by referring to Figure 2. When apartial discharge pulse originates at some point interior t o a cableit splits evenly into two pulses which travel in opposite directions.One pulse, referred to as the direct pulse, proceeds toward thedetector end arriving at t ime t . The other pulse proceeds towardthe open end of t he cable and is reflected from the open circuitback toward the detector end, arriving at the detector at timet 2 Knowing the velocity of propagation, v for the cable, thelocation, e is given by

where the location is referenced from the open circuit end of thecable.To save time in processing, a simple but accurate method is usedto calculate the locations. T he technique applies an approxima tematched filter to the recorded p l s e s whose shape resembles a


4/5

DISCHARGEDETECTOR SITEEN0 A OPENEND

F i g u r e 2. Position versus time diagram of direct pulse and re-flected pulse propagation.par tial discharge pulse. In effect, this assumed pulse is cross-correlated with the pulses in the measured dat a. Th e time delaybetween th e two largest pulses in the waveform is determinedand converted into a location using a m easured propagation ve-locity. M ore accurate delay esti mator s are not justified since theincreased computational effort does not provide enough increasein accuracy. W hen t he two largest pulses in the d ata record arefrom two distinct discharge sites , the estim ate is referred to as aphanto m delay. Th e probability of calculating a pha ntom delayis small at the voltage levels used, with less than 1 percent of theestimate s being phantoms. These phantoms are not a problemsince the t im e delay between th e pulses is random and does notlead to a count rate that favors any particular s ite. A limitationof this technique is the existence of a measurement blind spot inthe cable: when th e arrival times of the dire ct pulse and reflectedpulse are nearly the same, it becomes difficult to differentiatebetween a single pulse and two closely spaced pulses. Wi tho utspecial, computationally intensive processing, these sites all ap-pear t o be a t a distance range near the end of the cable, in effectcreating a blind spot.

T ST R SULTSIn this case, the measurement technique is intended for use insitu on 600V class nuclear power plant cables and so represen-tative cable types were used in the experiments. These cablescan be grouped into three categories: shielded cables, unshieldedmulticonductor cables, and unshielded single conductor cables.T he following samples were used in a series of expe rimen ts:

An unaged, unshielded, two conductor, #14 AWG, 600Vclass, neoprene jacketed cable of the same type used forcontrol cabling in nuclear power plants.An unaged, 50o hm , coaxial cable of th e same typ e used forinstrum entation cabling in nuclear power plants .An aged, unshielded, 15 conductor, 18 AWG cable of thesam e type used for instrum entation cabling in nuclear powerplants .

No experiments were conducted on unshielded single conductorcables since the techniques to be discussed require th at the cablehave some transmission line properties; a s ingle conductor, in theabsence of g round plane, does not.

Vertical:Number of pairs 250counted per bin ofthe histogram150.

Horizontal:Distance in feet 58.8from the open endNote the scattering of counts over the length of the cable

F i g u r e 3 Histogram of delay times between the two majorpulses in each event of a 5000-element record containing truepairs and random pairs .

Vertical:Positive PD

Numberof 2r.o Cpulses perbin of the 8 . 9 9histogram 7 9 . 0 2 1 9 . 3 5 9 . 4 9 0 . 639

Horizontal:Pulse chargein pC

F i g u r e 4. Histogram of the distribution of partial dischargepulses ( PD ) as a function of their charge transfer magnitudepulse height ).Tests were performed on 100-m (330ft) length of unaged, two-conductor cable with artificial defects introduced at three loca-tions. A small portion of the jacket was removed at t hre e loca-tions and the insulation was carefully pierced through to t he con-duc tor using a #60 twist drill in two of th e locations; a sm all razorknife was used t o slit the wire insulation in the third . In neithercase was any of th e insulation remo ved; only a break was createdfrom the conductor surface to the surrounding atmosphere. Thepartial discharge inception voltage of the cable was 4500V andmeasurements were made at voltages levels 10, 20, and 40 per-cent above inception. W ith a small lower discrimination level,the low-level background partial discharges have dominant countrates indicating that the partial discharges are distributed alongthe len gth of the cable as shown in Figure 3. If th e lower dis-crimination level is increased, then th e count rates of the p artialdischarges from the defect s ites dom inate the background countrate as indicated by the charge histograms of figure 4; note th atthe histograms are not unimodal. Th e locations of th e indi-vidua l dam age sites were found using those delay values whichcorresponded to data near the peaks of the charge histograms.Each of the damage sites was correctly identified as can be seenin figures 5 6 and 7. Other measurements were made on a20 m (66 ft) length of coaxial cable with artificial damag e inflictedby slightly abrading th e shield. Th e partial discharge inceptionvoltage was 5200 V. In this case, the defect was imm ediately


5/5

Vertical:Number of pairs 2 0 0 Icounted per bin ofthe histogram

1 t IIHorizontal:Distance in feet 48 8 --from the open end 4 d e i i a . 2 d 0 ? a hNote the emergence of a defect signal at 70.2 feet location of anactual defect.

Figure 5. Histogram of time delays for 5000 V and pruned data.

Vertical:Number of pairscounted per bin ofthe histogram lZasHorizontal:Distance in feetfrom the open end

Note the emergence of a defect signal at 4 feet actual defect),with minor indications at 70 feet already detected in Figure 5 ,and at 178 feet no known defect at that point).Figure 6. Histogram of time delays at 5200V and pruned data.

Note the emergence of a defect signal at 240 feet actual defect),with substantial signal at 70 feet already detected in Figure 6 ,small signal at34 f t as in Figure5 ,and new signal at 280 feet.

Vertical: 12s.Number of pairscounted per bin ofthe histogramHorizontal:Distance in feet afrom the open end

Figure 7. Histogram of time delays at 6000 V and pruned data.

C--

--

--

identified because the partial discharge activity at the damagesite was much greater than the background partial discharges.Similar results were also obtained while measuring a 12-m 40 ft)length of aged multiconductor cable containing an artificial de-fect. The artificial defect was created by piercing the insulationon one of the 15 conductors with a small razor knife, similar tothe slit described above. The ~ a r t i a l ischarge inception voltagewas 2 3 0 0 V . Similar to the previous case, the partial dischargeactivity at the damage site was greater than the background ofthe rest of the cable allowing easy identification of the flaw.

+ * ,*,a ,*.e Sb . Z l b . 2 7, 3 8 8 8 -

CONCLUSIONSThe t est method based on partial discharge detection using signalprocessing makes possible the identification and location of in-sulation breaches, undetectable under normal service conditions,th at may become a fault under additional environmental stresses,or may progress into a fault under further aging.Tests performed on three representative types of low-voltage ca-bles have demonstrated that the method can differentiate be-tween signals emanating at incipient defect sites and the back-ground of partia l discharges occurring over the length of th e cable.In order t o induce partial discharges, th e test voltage must clearlybe raised above the inception voltage. Depending on the cablestructure, this test voltage will be on the order of a few thousandvolts, a subject of possible concern, but not a fatal limitationfor a test method offering the reward of definite identification ofdefects that could become fault locations.The demonstration was obtained with a general-purpose PC-typecomputer, using a commercial software package augmented bymoderately complex custom-designed software. In a more dedi-cated sys tem, with furthe r software enhancement, additional ben-efits are possible, such as the detection of incipient defects withonly brief exposure of the insulation to the test voltage, auto-mated scanning of th e dat a for most significant content, and op-timization of the test procedure into an expert system approach.

ACKNOWLEDGEMENTSThe authors acknowledge the contributions of S. Clevenger andD.R. Flach in the development and implementation of the menu-driven software used in the applica tion of th e principle of th emethod. Th e signal processing concepts were developed by J.P.Steiner as described in Reference 4 and implemented prior tohis new position with Biddle Instruments. Th e implementationand demonstration t ests were performed by J.P. Steiner and F.D .Martzloff with funding provided by the USNRC Nuclear PlantAging Research Program.

REFERENCES[I] IEEE Std 383-1974 Standard for Typ e Test of Class 1E Electrical Ca-

bles, Field Splices, and Connection for Nuclear Power Generating Sta-tions.21 Proceedings: Workshop on Power Plant Cable Condition Monitoring

EPRI EL/NP/CS-5914-5R, July 1988.[3] Snyder, D.L., Random Point Processes John Wiley and Sons,

New York NY, 1975.[4] Stein er, J.P. Digital Measurement of Partial Dischargen, Ph .D . The sis,Purdue University, W. Lafayette IN, May 1988.[5] Steiner, J.P ., Weeks, W.L . and Furgason, E.S ., UDigital Estimation ofPartial Discha rgen, 1987 Annual Report, Conference on Electrical Insu-lation and Dielectric Phenomena, Gaithersburg MD, October 1987.

pd in cables

Documents