7/21/2019 2015-D1-1 Partial Discharge Measurements in the Sub-VLF-Range

1/6

D1.1 9thInternational Conference on Insulated Power Cables D1.1

Jicable'15 - Versailles 21-25 June, 2015 1/

Partial Discharge Measurements in the Sub-VLF-Range

Kay RETHMEIER;Kiel University of Applied Sciences, (Germany),kay.rethmeier@fh-kiel.de

Rudi BLANK;b2electronic GmbH, (Austria),rudi@b2electronic.at

ABSTRACT

By reducing the test voltage frequency, the physical stresson a potential PD defect inside a cable insulation or atinterfaces in joints or terminations is consequentlydiffering from power frequency stress. Nevertheless, VLFtesting and VLF cable diagnosis are worldwide acceptedtools for dielectric quality control. With further increasingcable length, as HV export cables (of some 10 or evensome 100 kilometres) connecting offshore converters withthe grid on land, the lower limit frequencies of VLF testing

have to be discussed. When reducing the test voltagefrequency finally to DC, the electrical field control of thecable system may be defeated, leading finally to abreakdown of the cable even without a failure. On theother hand, diagnostic parameters, as the partialdischarge level or the PD inception voltage may also bedetermined with reduced VLF frequency. This paperdescribes the PD behaviour of typical cable defects at50 Hz in comparison to VLF test voltages of 0.1 Hz downto 0.01 Hz. All typical PD defects could be identified bytheir PRPD patterns for all test frequencies. The 50 Hzpatter was very comparable to the ones at 0.1 Hz andeven 0.05 Hz, 0.02 Hz and 0.01 Hz. For the PDIV, adecreasing tendency could be found for decreasing testvoltage frequencies. In particular, VLF 0.1 Hzmeasurements are sufficiently comparable to VLF

measurements with further decreased test voltagefrequency.

KEYWORDS

Partial Discharges, PD, Very Low Frequency, VLF, CableTesting, Cable Diagnostic, Super-long cables.

INTRODUCTION

Due to the high demand of capacitive power, high voltagecable testing is related to high effort, especially on site.With respect to this effort, the test voltage frequency canbe varied from power frequency. For resonance testsystems with fixed inductance the frequency range isextended from 20 Hz to 300 Hz [1]. But also powerelectronic operated test system with synthetically sinusshapes are admitted by testing standards. For mediumvoltage cables, VLF test voltage with 0.1 Hz is verycommon [2]. Nevertheless, with further increasing cablelength even these extended frequency ranges have theirlimits, as on a cable test of a 74 km HV cable of ca. 18 Fcapacitance, where a resonance test frequency of notmore than 17 Hz was reached [3]. Finally, customers andtest institute agreed to perform the test with this reducedfrequency.

But also for the mobile VLF test systems the testfrequency can be decreased in case of higher capacitiveloads. As the mandatory test voltage level and testingtime for VLF often is intensified compared to power

frequency tests [2], a further reduction of the testfrequency may consequently demand a further increase oftest voltage level or testing time, in order to compensatethe reduced number of high voltage test cycles applied tothe test object. With focus on diagnostic measurements,the PD behaviour of high voltage cables tested withreduced frequency may also be of interested. Therefore,this paper focuses on the effect of the test voltagefrequency of the phase resolved partial dischargediagrams (PRPD) and on other PD data, as the inception

voltage (PDIV) and extinction voltage (PDEV).

ALTERNATIVE TEST VOLTAGES

To bypass the problem of the large capacitive powerdemand of long cable systems, alternative test voltageshapes have been presented to the market. Damped AC(DAC) for instance, uses the resonance effect of thecapacitive test object combined with the inductance of thetest voltage generator. Here, the resulting test voltagefrequency can be close to power frequency. As theamplitude of the outcoming oscillation is decreasing dueto the internal losses of the test system, this voltage

shape cannot be used for cable testing. Nevertheless, itcan be used for PD tests, especially for PD location.

Combining a VLF rectangular test voltage with definedtransitions from minus to plus, a 50 Hz behaviour of thetest voltage cab be simulated with so-called CosRec testvoltages, or Slope 50 Hz test voltages. With this voltageshape, cable testing is as well possible as PD location.The generation of phase resolved PRPD patterns isdifficult here, as the 50 Hz slopes only covers a very shortphase window (

7/21/2019 2015-D1-1 Partial Discharge Measurements in the Sub-VLF-Range

2/6

D1.1 9thInternational Conference on Insulated Power Cables D1.1

Jicable'15 - Versailles 21-25 June, 2015 2/

As a significant drawback the MPD cannot indicate the PDlevel (pico coulombs) for VLF test voltage over a certainperiod of time. One period of 0.1 Hz for instance takes10 seconds of time, instead of 20 ms for 50 Hz. The IECPD standard 60270 [6] clearly dictates that the scalar

number for the charge level has to be evaluated with atime constant of not more than 440 ms. For powerfrequency test voltages this demand leads to a singlescalar PD level (for example of 10 pC) for ca. 20 periodsof the test voltage. In absolute contrast to this, for 0.1 HzVLF the scalar PD level has to be refreshed ca. 20 timeswithin a single test voltage period. So the PD level candrop down to zero even in constant PD patterns, if thereare only some phase angles not affected by these PD. Forinternal cable PD, typically the zero crossings of the testvoltage show PD, but not the area around the voltagepeaks at 90 and 270, respectively. Therefore, the IECPD level can by definition not be used in combination withVLF test voltage systems. Nevertheless, the charge

calculation for single PD pulses, as they are plotted anddisplayed in the PRPD diagrams is not affected by this atall. All single PD pulse can be correctly measured. Onlythe summary charge value (IEC-charge, with

7/21/2019 2015-D1-1 Partial Discharge Measurements in the Sub-VLF-Range

3/6

D1.1 9thInternational Conference on Insulated Power Cables D1.1

Jicable'15 - Versailles 21-25 June, 2015 !/

picture. If the corona pattern is not exactly located in theminimum of the test voltage sinus, the divider must not beused for VLF, or the phase error has to be corrected bysoftware.

For VLF, resistive voltage dividers can be used.Unfortunately, those resistive dividers do not worksatisfactory at 50 Hz, as their stray capacitances causefurther errors to the divider signal.

PD LOCATION IN THE SUB-VLF-RANGE

Once a PD pulse is ignited by sufficient electrical fieldstrength in the presence of an initial electron, theelectromagnetic wave of this pulse propagates from itsorigin in both direction of the cable, reflecting at thecables ends. With known propagation speed and/orknown cable length the PD fault position can be estimated

simply by measuring the time difference of direct PDsignal and reflected PD signal (TDR method, TimeDomain Reflectometry). For this process, the test voltagefrequency of the high voltage is not of interest at all. TDRPD location can be applied from DC test voltage up tohigh AC frequencies as 300 Hz for resonance cable testsor even some kilohertz in case of PD test under repetitiveimpulse voltage [8]. Therefore, PD location with sub-VLFtest voltage is possible in the same way as with 0.1 HzVLF or power frequency and therefore not subject of thiscontribution.

TESTING PROCEDURE

When recording PRPD diagrams, the test voltagefrequency is a significant factor. With only recording60 seconds of 50 Hz voltage, 3000 sinus cycles areconsidered for the PRPD. For 0.01 Hz, 60 seconds arenot even enough time to record a single completed sinuscycle. To reach an adequate and comparable dot densityin the PRPD diagram, 3000 cycles have to be recorded aswell. The estimated testing time would be 3.5 days,approximately. Due to this very practical reason, thetesting procedure was not aiming at a total comparability.Table 2 summarizes the relevant parameters of the testprocedure.

Frequency [Hz] Time per 0.5kVstep [minutes]

PRPDacquisition time

[minutes]

50 1 20

0.1 1 60

0.05 2 120

0.02 5 120

0.01 9 120

Table 2: Test voltage parameters

In the beginning of the test the test voltage of anyfrequency was applied with a level of 500 V to the testobject. After a certain step time, the level was increased

by additional 500 V, until PD inception could berecognized. After a certain PRPD acquisition time, the testvoltage was decreased in steps of 500 V to detect thecorrect level of PD extinction. Figure 3 shows a step testfor 50 Hz, figure 4 shows a test sequence for 0.01 Hz.

Fig. 3: Step test 50 Hz, >1 h

Fig. 4: Step test 0.01 Hz, >10 h

Only with time steps of sufficient duration, the PDIV canbe determined precisely and with a certain amount of

reproducibility. Of course, those time steps have to be inminimum as long as a full sinus period of the external testvoltage.

TEST OBJECTS (DUT) AND RESULTS

PD test were carried out on artificial reference defects aswell as on XLPE high voltage cable samples. Thereference defects, as corona and surface discharge, areneeded to confirm the reliability of the PD results.

Corona Reference

As already discussed above, the reliability of PDmeasurements strongly depends on the proper acquisitionof the test voltage level (in order to determine PDIV orPDEV) and the phase position of single PD events (inorder to generate phase resolved PD patterns). Byperforming a wire test, the correctness of the PD phaseposition can be confirmed easily. Figures 5 and 6 showthe PRPD of the corona wire at 50 Hz and 0.01 Hz,respectively, recorded directly at PDIV of ca. 5 kV for allselected test voltage frequencies. Due to the longrecording time, multiple disturbance pulses haven beenrecorded, beside the corona pulses. Those pulses wererandomly distributed over the 360 of the test voltagesphase and could clearly be distinguished from the PDpulses.

7/21/2019 2015-D1-1 Partial Discharge Measurements in the Sub-VLF-Range

4/6

D1.1 9thInternational Conference on Insulated Power Cables D1.1

Jicable'15 - Versailles 21-25 June, 2015 "/

Fig. 5: Corona @ 270(50 Hz), PDIV

Fig. 6: Corona @ 270(0.01 Hz), PDIV

With using a coupling capacitor and a voltage divider withconstant frequency response from 0.01 Hz up to someHertz, the correct phase position of the PD pulses couldbe demonstrated. As typical for corona discharges, thePD is concentrated at approximately 270 of the testvoltage, for all viewed test voltage frequencies. Also PDIV

and PDEV have been identical for 50 Hz down to 0.01 Hz,as well as the charge magnitude (pC level).

Surface Discharge

As a second reference test a surface discharge wasgenerated by a Toepler setup. Figures 7 to 9 show thePRPD diagrams at selected frequencies at a comparabletest voltage level of ca. 18 kV.

Fig. 7: Toepler discharge @0.1 Hz, ca. 18 kV

Fig. 8: Toepler discharge @0.05 Hz, ca. 18 kV

Fig. 9: Toepler discharge @0.01 Hz, ca. 18 kV (greencurve incorrectly displayed)

Besides randomly distributed disturbance pulses, the PDpattern are comparable for all selected test voltagefrequencies, starting at 0and 180with having maximumPD levels at 30 and 210, respectively. For 0.01 Hz, thesteep envelop of the front of the PD pattern is not thatdistinct anymore. Unfortunately, the PRPD pattern do notfit to the expectations, as there is no dominant half cyclewith significant PD magnitude as typical for surfacedischarges. The pattern has more similarity to slotdischarges, known from stator windings of electrical

7/21/2019 2015-D1-1 Partial Discharge Measurements in the Sub-VLF-Range

5/6

D1.1 9thInternational Conference on Insulated Power Cables D1.1

Jicable'15 - Versailles 21-25 June, 2015 5/

machines, indication delaminated insolation layers. It canbe assumed, that the used Toepler setup had anadditional PD source. Nevertheless, this resulting PDactivity can also be compared throughout the selected testfrequencies.

Figure 10 compares PDIV and PDEV for different testvoltage frequencies.

Fig. 10: Surface discharge, PDIV and PDEV

Here it can be seen, that PDIV and PDEV decrease withreduced test frequencies. A reduction of PDIV indicatesthat PD faults may be found with lower test voltage levels,if sub-VLF test voltage is used. Remarkable in this case isthe reduced PDIV for VLF in comparison to the 50 Hz test.

Defective Cable Terminations

Besides the model PD faults, more realistic PD defects

were realized by manipulating a high voltage cable (typeNA2XS(F)2Y 6/10 (12)kV). In a first step the mediumvoltage cable was tested with defective end terminations,causing heavy surface discharges. Figure 11 shows somePRPD diagrams of selected test voltage frequencies.

50 Hz 0.1 Hz

0.05 Hz 0.02 Hz

Fig. 11: Defective cable termination, PRPD @ 15 kV

It can be seen that the general shape and impression of

all PRPD diagrams are comparable, especially for theVLF frequencies. With 50 Hz test frequency, the highestlevel of partial discharges could be detected, leading to anexcellent signal to noise level (SNR). This can be a

benefit when lower absolute levels of partial dischargeshave to be measured. In general, the PRPD diagrams ofthe VLF and sub-VLF test frequencies show less denseclouds, finally caused by a significant lower number ofacquired PD pulses recorded in the defined measuring

time period.

When focussing on the PDIV, it can be seen that for verylow frequencies the inception voltage decreases, asalready found in the test series before. Figure 12 showsthe PDIV and PDEV for selected VLF and sub-VLF testvoltage frequencies.

Fig. 12: Defective cable termination, PDIV and PDEV

Defective Outer Semicon Layer

Figure 13 shows the defect of the outer semicon layer ofthe high voltage cable, causing partial discharge pulses.

Fig. 13: Defective semicon layer

Figure 14 shows the PDIV and PDEV for selected testvoltage frequencies.

Fig. 14: Defective semicon layer, PDIV and PDEV

7/21/2019 2015-D1-1 Partial Discharge Measurements in the Sub-VLF-Range

6/6

D1.1 9thInternational Conference on Insulated Power Cables D1.1

Jicable'15 - Versailles 21-25 June, 2015 /

It can be seen that the PDIV levels of all VLF tests areslightly higher compared to the power frequency test of50 Hz. Within the group of the VLF test voltages, thelevels of PDIV and PDEV show a decreasing tendency,which looks similar to the PD behaviour obeyed in the

tests before. Viewing the PRPD diagrams (fig. 15), againthe patterns are comparable.

50 Hz 0.1 Hz

0.05 Hz 0.02 Hz

Fig. 15: Defective semicon layer, PRPD @ 18 kV

It can be assumed, the those kind of PD defects can bedetected more easy by sub-VLF test voltage frequencies,as the necessary test voltage level to ignite those PD islower compared to 0.1 Hz VLF.

SUMMARY

It is a fact that high voltage tests on super-long cablesystems can only be realized by reduced test voltagefrequencies, even below 0.1 Hz, if small, light weighted,

cheap and mobile test systems shall be used. Althoughthe physical partial discharge process may be different forvarious test voltage frequencies, laboratory test at KielUniversity of Applied Sciences, Germany, showed, thatthe phase resolved PD patterns (PRPD) of all analysedPD faults had a similar look. For 0.1 Hz, as well as forfurther reduced test voltage frequencies of 0.05 Hz,0.02 Hz and 0.01 Hz, the PD clusters were located atsimilar phase positions and at comparable magnitudes inthe PRPD diagrams. Therefore, a PD identification byPRPD analysis can be done at all these differentfrequencies. PD location is totally independent from theexternal test voltage frequency at all.

Very important is the PDIV and the PDEV, respectively.

For VLF 0.1 Hz, those voltage levels are slightly highercompared to 50 Hz power frequency in most of the cases.This fact is already identified since years andconsequently implemented in international standards. VLF

test therefore have to be done with increased test voltagelevel, compared to power frequency. Within the group ofVLF test voltage frequencies, a decreasing trend could befound for the PDIV and PDEV, when reducing the VLFtest voltage frequency from 0.1 Hz down to 0.05 Hz,

0.02 Hz and 0.01 Hz. This indicates that the PD testvoltage level for super-long cables may be reduced, asthe PD defects analysed in this test series could have alsobe detected with lower test voltage levels. A reduced testvoltage level combined with a reduced test voltagefrequency leads to a further decreased capacitive powerdemand of the test object. Finally, longer cables can betested and measured with lower dielectric stress to thetest object.

Acknowledgments

The authors want to thank Mr. Riccardo Beck, graduate ofKiel University of Applied Sciences, for his patience while

performing sub-VLF PD measurements for hours, daysand weeks. Now we all know that 400 periods of sinus at0.01 Hz is the equivalent of a full working day.

REFERENCES

[1] IEC 62067 ed2.0 (2011-11), Power cables withextruded insulation and their accessories for ratedvoltages above 150 kV (Um = 170 kV) up to 500 kV(Um = 550 kV) - Test methods and requirements

[2] DIN VDE 0276-620 VDE 0276-620:2010-11, Powercables - Distribution cables with extruded insulation

for rated voltages from 3,6/6 (7,2) kV up to andincluding 20,8/36 (42) kV

[3] Schierig, S., Coors, P., Hauschild, W., HV ACTesting of super-long cables, Proceedings ofJicable07, Versailles, France, June 2007

[4] Datasheet HVA 120, b2 electronic, Austria, 2014

[5] Datasheet MPD 600, OMICRON electronics, Austria,2014

[6] IEC 60270, "High-voltage test techniques Partialdischarge measurements", 3

rdedition, 2000

[7] Datasheet MCC 210, OMICRON electronics, Austria,2014

[8] IEC/TS 61934 ed2.0 (2011-04), Electrical insulatingmaterials and systems - Electrical measurement ofpartial discharges (PD) under short rise time andrepetitive voltage impulses