Partial discharge oscillograms and pulse distribution

Document Sample
Partial discharge oscillograms and pulse distribution Powered By Docstoc
					                                                                                 aa
Proceedings of 131hInternational Conference on Dielectric Liquids ( ICDL '99 ), N r ,Japan, July 20-25, 1999




      Partial Discharge Oscillograms and Pulse Distribution in
               Transformer Oil-Pressboard Insulation
                     Y.P. Nerkar, S. Senthil Kumar, M.N. Narayanachar and R.S. Nema

                                   Department of High Voltage Engineering
                                         Indian Institute of Science
                                         Bangalore 560 012, INDIA




                     Abstract                            tem resulting in failure of the system. Failure oc-
                                                         curred due t o puncture in the pressboard sample.
   Partial discharges cause degradation of insulation
system and a study of their behaviour can give use-
ful information regarding the state of the insulation         Experimental Setup and Procedure
system. Partial discharge tests are therefore, car-
                                                            Straight detection method was used t o detect
ried out as quality checks for the insulation system.
                                                         PDs with 50Hz ac voltage applied t o the test sam-
A model transformer insulation system consisting
                                                         ple. The PD detector was a narrow band in-
of pressboard and transformer oil with uniform field
                                                         strument whose output was coupled t o a multi-
electrode arrangement was tested for its behaviour
                                                         channel analyser and a digital storage oscilloscope.
under partial discharges. The system was allowed
                                                         Records of P D distributions and oscillograms were
to age till failure of the sample occurred in the form
                                                         obtained a t regular intervals during ageing of the
of a puncture.
                                                         oil-pressboard system stressed at 13kV/mm. The
   Records of partial discharge distributions were
                                                         distribution records were acquired for a time inter-
acquired along with oscillograms. Oscillograms
                                                         val of 10 seconds and oscillograms were obtained
help in the traceability of the measurement when
                                                         for one ac cycle (20ms) period.
compared with distributions. A comparative study            The experiments consisted of two parts; (a) the
of the effect of amplifier setting and the pulse rep-
                                                         calibration experiments and (b) the ageing experi-
etition rate on the partial discharge distributions
                                                         ments.
and the oscillograms acquired during ageing of the
sample has been presented. It is shown that by ap-
propriate threshold on the distribution, distinctive                Calibration Experiments
features can be extracted.
                                                            Calibration experiments were performed to de-
                                                         termine the amplifier gain setting (range fixation)
                   Introduction                          to represent the PD distribution in the best pos-
                                                         sible way. The amplifier system had a maximum
   Oil-pressboard insulation forms a major insula-       output of 8V and this required setting of the am-
tion component in high voltage transformers. Par-        plifier gain corresponding t o the maximum magni-
tial discharge (PD) is one of the prime causes for       tude of the pulses. Two gain settings corresponding
the failure of insulation. Model systems are con-        t o 200pC and 1400pC were used in the calibration
venient in studying the P D behaviour of press-          experiments.
board insulation system. T.R. Blackburn et a1 [l]
have considered a number of examples simulating
the many practical cases where PDs develop in an      Amplifier gain setting
oil wedge formed by insulation interfaces. In the
present work, a model insulation consisting of an      Figure 1 shows an oscillogram and a pulse distri-
oil-impregnated pressboard placed between 7n/12      bution obtained with amplifier setting of 8V corre-
uniform field Rogowski electrodes is studied for its sponding to 200pC. The oscillogram (Figure l ( a ) ) is
PD behaviour. PDs occurred in the wedge shaped       an envelop of ten, one cycle records superimposed.
oil gap between the pressboard and the electrodes    When the setting of the amplifier gain was changed
causing degradation of the oil and pressboard sys- to 1400pC range, the resulting oscillogram and the



                                                  - 265 -
                                                                     200                                                                  1400


                                                                     150                                                                  1050


                                                                     100                                                                  700


                                                                     60                                                                       350




 I                                                              -I
           2 00   v                  M 2 OOms Llnex   1.6                                2 00   v                  M 2 OOms Line1   1.8
                           (a) Oscillogram                                                                (a) Oscillogram

     800,                                                   I                      400   I                                                I
     7nn   1
     500   1                      I




      "
       0               50       io0       150      200                                 -0                  500          1000
                      (b) Discharge magnitude (PC)                                                  (b)   Discharge magnitude (pC)



Figure 1: (a) Oscillogram and (b) distribution with                              Figure 2: (a) Oscillogram and (b) distribution with
200pC range                                                                      1400pC range


distribution are as shown in Figure 2. The two fig-                              time (60ps) of the narrow band system. This over-
ures (Figure 1-2) obtained a t two different amplifier                           lap due t o superposition of pulses reduce the pulse
gain settings depict contrasting features of the PD                              magnitudes of subsequent pulses.
behaviour.                                                                          That, errors due to superposition of pulses can
   In Figure 1, the distribution is negatively skewed                            occur is demonstrated with the help of Figure 3.
with very few pulses at the saturation level of 8V                               Figure 3(a) shows the calibrator pulse (CHl), the
(200pC) and many pulses in the mid range (150pC).                                measuring impedance output (CH2) and the re-
This observation is also indicated by the oscillogram                            sponse (CH3) of the narrow band amplifier. Fig-
(Figure l ( a ) ) . However, the distribution shows less                         ure 3(b) shows the response of the amplifier when
number of lower magnitude pulses in contrast t o                                 the calibrator pulse frequency is increased and over-
the observations made from the oscillogram.                                      lapping of the pulses occurs. It can be seen that the
   In Figure 2, the distribution is positively skewed.                           magnitude of the second pulse becomes less than
The pulse count is reduced and there are many                                    the magnitude of the first pulse due t o superposi-
large magnitude pulses ( 2 200pC) which are not                                  tion. Therefore, when the repetition rate of the
observed as saturated pulses of 8V when the gain                                 pulse is high, superposition of pulses reduce the
setting was 200pC (Figure l ( b ) ) and this effect is                           magnitudes of the subsequent pulses. This appar-
prominent at phase positions where the number in-                                ently could be the reason for obtaining fewer satu-
tensity of pulses are high (observation from oscillo-                            rated (8V) pulses at the amplifier setting of 200pC
grams, Figure l ( a ) and Figure 2(a)).                                          which could result in incorrect representation.
                                                                                     Overlapping of the pulses also reduces the dis-
                                                                                 charge count as many of the overlapping pulses will
 Discussion                                                                      be measured as a single pulse [2]. The smaller
                                                                                 pulses have much higher repetition rate compared
  Amplifier setting with the range of 200pC showed                               t o the larger pulses. Therefore, the error is likely to
the presence of large number of pulses. When                                     be higher for the lower discharge pulses than that
observed closely, many of the pulses were found                                  for the larger discharge pulses. Hence, the distri-
to be overlapping as pulses appeared with a very                                 bution could have low count for smaller discharge
high repetition rate and within the resolution                                   pulses (Figure l ( b ) ) . The error due t o limited res-



                                                                           -   266 -
1


                                                                                                                      IPCl




                                                                                                                  1050


                                                                                     I
                                                                                   . I                                ?OD




                                                               2    oov               IC12 OOms L l n e l   2 4
                                                                          (a) Oscillogram

                                                           5000 I                                                 I

1




    I                                     I




Figure 3: Pulse response (a) without overlap and         Figure 4: (a) Oscillogram and (b) distribution im-
(b) with overlap error                                   mediately on application of 13kV/mm stress


olution is reduced with the amplifier gain set to        ure 4(a) and (b) show the oscillogram and the PD
1400pC. At this gain setting, the smaller magni-         distribution obtained immediately after application
tude pulses of high repetition rate are automatically    of voltage stress. Figure 5(a) and (b) show the
eliminated t o a large extent due t o poor sensitivity   records obtained just before the failure of the sam-
of detection.                                            ple.
   Hence, one has t o be cautious about the ampli-          The oscillograms show that the number inten-
fier gain setting and the limitation of the detector     sity of discharge pulses increases with ageing. The
resolution before attempting interpretation of dis-      distributions show that significant changes occur
tribution using shape sensitive statistical and image    with larger magnitude pulses (maximum changes
processing tools.                                        in range 500pC t o 1400pC). In contrast t o the in-
                                                         crease in the number intensity as observed from the
                                                         oscillograms, the count of smaller discharge pulses
              Ageing Experiments                         reduces (Figure 4(b) and Figure 5(b)), just before
                                                         failure of the sample. This could apparently be due
   The ageing experiments were conducted a t a           to insufficient resolution of the detector. Therefore,
stress of 13kV/mm with the amplifier gain setting        the phenomenon can be safely studied by putting
a t the range of 1400pC. The experiment was termi-       a threshold on the distribution, thus avoiding the
nated when the sample failed by puncture. For the        ambiguity regarding recording of the smaller mag-
30 number of samples tested, the failure time var-       nitude pulses. The effect of putting a threshold on
ied from 37 min t o 194 min. It was observed that,       the distribution is shown in Figure 6 which clearly
the number of P D pulses, their phase span on the         shows the changes occurring in the pulse distribu-
ac cycle and the maximum discharge magnitude in-         tion with ageing. These changes can be analysed
creased with ageing. A maximum discharge value            with the help of appropriate statistical tools [3].
of lOOOpC was found to be critical, after which the
failure of the sample was imminent.
    Typical oscillograms and PD distributions                                   Conclusions
recorded a t two instants of time during the course
of ageing are given in Figure 4 and Figure 5. Fig-         With the help of PD oscillograms and their pulse



                                                   - 267 -
 I    " '       " '   " " " " '   " "   " - " " '   " "   " " " " ' I


                                                                        lPCl




                                                                        1400


                                                                        1050


                                                                        700


                                                                        350




                                                                                                 (a) Discharge magnitude (pC)




                1                                                                     300     ;
                                                                                     tI
     3000




     25001                                                                            200




     'oooI                                                                         =
                                                                                   <? 1 5 0



     '-L
     1500




                                                          ~




            0
                                  500         1000                                                   500           1000
                             (b) Discharge magnitude (pC)                                        (b) Discharge magnitude (pC)



Figure 5: (a) Oscillogram and (b) distribution just                                Figure 6: Distributions with threshold (a) immedi-
before failure                                                                     ately after application of 13kV/mm stress, (b) just
                                                                                   before failure
distributions, the importance of amplifier gain set-
ting and the errors that could creep in measure-                                     2. S. Senthil Kumar, M.N. Narayanachar and
ment and analyses are demonstrated. The age-                                            R.S. Nema, Response of Detectors in the
ing behaviour of the model insulation system of                                         Measurement of Partial Discharges- Proposal
pressboard-oil sample due t o P D is studied. The                                        for Partial Discharge Detectors with Extended
advantage of setting a suitable threshold and selec-                                     Resolution, Fourth Workshop and Conference
tively observing the changes in the larger magni-                                        on EHV Technology, Bangalore, July 17-18,
tude pulses is also demonstrated.                                                        1998, pp. 153-157.
                                                                                     3. Y.P. Nerkar, S. Senthil Kumar, M.N.
                                   References                                           Narayanachar and R.S. Nema, Partial Dis-
                                                                                        charge Oscillograms and Pulse Distribution in
 1. T.R. Blackburn, R.E. James et al, Advanced                                          Transformer Oil-Pressboard Insulation, Semi-
    Techniques for Characterisation of Partial Dis-                                     nar on Transformer Insulation System, Central
    charges an Oil-Impregnated and Gas Insulated                                        Power Research Institute, Bangalore, January
    systems, 15-102, CIGRE, 28 Aug - 3 Sept, 1994                                       28-29, 1999, pp. S VI (1-5).
    Session.




                                                                               - 268 -