Measured Partial Discharge Inception Voltage for a by drg59916


									       Measured Partial Discharge Inception Voltage for a Cavity at Different
                              Applied Frequencies

                                           Cecilia Forssén and Hans Edin
                                            KTH Electrical Engineering

Abstract                                                      The present paper studies experimentally how the
Partial discharges (PD) can be measured at variable           frequency of the applied voltage influences the PDIV.
frequency of the applied voltage. The interpretation of       The PDIV is measured with applied frequency in the
such measurements requires a physical understanding of        range 0.1 to 100 Hz for disc-shaped cavities insulated in
the observed PD frequency dependence. This paper              polycarbonate. Measurements are done with different
studies the influence of the applied frequency on the         pre-excitation of the specimen to study how this affects
partial discharge inception voltage (PDIV). The PDIV          the PDIV. It is also briefly investigated how the cavity
is measured for insulated disc-shaped cavities at applied     ages due to PD, with the purpose of avoiding aging
frequency in the range 0.1 to 100 Hz. It is found that the    effects in the PDIV measurements.
PDIV increases with increasing applied frequency,
probably as an effect of the statistical time lag. In         2. Experimental methods
addition, the measured PDIV is lower than that
predicted by Paschens’s law. The effect of pre-               2.1. Test object
excitation on PDIV is also studied. A period without          The specimen used for this study is a disc-shaped cavity
voltage supply prior to each PDIV measurement is seen         embedded in polycarbonate (relative permittivity 3), see
to cause a faster increase in the PDIV with frequency.        Figure 1. It is made by pressing together three 1 mm
                                                              thick polycarbonate plates with a drilled hole centered
                                                              in the middle plate. The cavity has a diameter of 10 mm
1. Introduction                                               and a thickness of 1 mm. The polycarbonate plates
                                                              come from the supplier with a protection plastic that is
Partial discharges (PD) are usually measured at a single      removed just before the measurements. The plates are
frequency of the applied voltage, in general 50/60 Hz.        inspected for irregularities in the drilled hole. The plates
There are however reasons for measuring PD also at            are not cleaned.
other applied frequencies. A lower applied frequency
(commonly used is 0.1 Hz) reduces the power and size
needed for the voltage supply equipment. Another
option is to measure PD at variable applied frequency
(i.e. at more than one frequency) [1]. This method
provides more information than measurements at a
single frequency since the variation in applied
frequency changes the local state at defects in the
insulation. These changes can be used to better
characterize the defects, provided that the frequency
dependence of PD is interpreted physically.
                                                              Fig. 1 – Schematic picture of specimen (rotational symmetry,
One important property of PD is the partial discharge         measures given in millimeter).
inception voltage (PDIV). This is the lowest applied
voltage where discharges appear. Another property is
                                                              2.2. Measurement system
the statistical time lag. This is the time from that the
                                                              The phase resolved PD measurement system is
electric field is above the critical level for discharge
                                                              described in [1] and is based on an ICM (Insulation
until the discharge occurs. The statistical time lag is
                                                              Condition Monitoring) system [3].
caused by a lack of free electrons to start a discharge. If
                                                              For the PDIV measurements a function generator is
the applied voltage changes significantly over a period
                                                              used to generate a sinusoidal signal with amplitude that
comparable to the statistical time lag, this influences
                                                              increases linearly from zero at ramp rate 0.1 kV/s. The
measurements of the PDIV. Therefore these
                                                              first discharge is detected with an oscilloscope (LeCroy
measurements are sensitive to the rate of change of the
                                                              9310AM, 400MHz, 100MS/s).
applied voltage and to changes in the statistical time lag.
Experiments [2] show that light irradiation affects PDIV
measurements in cavities, presumably since the electron       2.3. Aging tests
generation rate in the cavity changes with the irradiation    The purpose of the aging tests is to find a time interval
conditions.                                                   where the PD activity in the cavity only changes
                                                              slightly due to aging phenomena. It is then assumed that
PDIV measurements within this time interval are not so
much affected by aging.
Aging tests are done on two equal specimens. A
sinusoidal voltage with amplitude 10 kV and frequency
50 Hz is applied for 7 h with interruptions (about 40 s
long) for phase resolved PD measurements at 50 Hz.

2.4. PDIV measurements
Partial discharge inception voltage measurements are
done on two equal specimens. Prior to the PDIV
measurements the specimens are conditioned (aged) for
1.5 h at 10 kV and 50 Hz. The PDIV is measured with
an applied sinusoidal voltage with frequency in the
range 0.1 to 100 Hz and with amplitude linearly
increasing from zero at a ramp rate of 0.1 kV/s. The
PDIV is defined as the amplitude of the applied voltage
when the first discharge is detected. It is calculated as
the product of the ramp rate and the time to the first
discharge. In addition to the influence of the frequency
on the PDIV, also the effect of pre-excitation of the
specimen is investigated. With “pre-excitation” is meant
voltage applied to the specimen prior to the PDIV
measurement. Three cases are studied:
     1. Pre-excitation at variable frequency
         The PDIV is measured ten times at each
         frequency. Prior to the first measurement at
         each new frequency, a voltage with that new
         frequency and with amplitude 10 kV is applied
         for 30 min.
     2. No pre-excitation
         The specimen is left without voltage supply for
         14 h. Then PDIV is measured ten times at each
         frequency. Prior to the first measurement at
         each new frequency, the specimen is left
         without voltage supply for 30 min.
     3. No pre-excitation (10 min pause)
         Same as 2 but the specimen is left without
         voltage supply for 10 min between each
         measurement and only five measurements are
         done at each frequency.
The effect of ramp rate on the PDIV measurements is
also investigated by varying the ramp rate in the range
0.1 to 0.4 kV/s. Ten PDIV measurements at 100 Hz are
done at each ramp rate. Since this is the highest
frequency used in this study, the statistical time lag
should have the largest impact on the PDIV
measurements here. Prior to the first measurement at
each new ramp rate, a voltage with frequency 50 Hz and
amplitude 10 kV is applied for 30 min.

3. Results and discussion

3.1. Aging
The aging tests show similar results for the two equal      Fig. 2 – PD pattern before aging (a) and after 1.5 h (b), 4 h (c)
specimens. Figure 2 shows the PD pattern before aging       and 7 h (d) aging. Mind the differences in y-axis scaling!
and after 1.5, 4 and 7 h aging, respectively.
                                                            The PD pattern changes greatly during the first 5 – 30
                                                            min of aging; the maximum apparent charge roughly
halves. After 1.5 h and up to 7 h aging there are no
distinct changes in the PD pattern and the maximum
apparent charge is about constant. The change in PD
pattern with aging is probably due to decreasing
statistical time lag for discharges and increasing
conductivity of the cavity surface, as discussed in [4].
This causes lower apparent charge since discharges
occur at weaker field in the cavity.
In the present study, an aged specimen that is left
without voltage supply for some time (30 min or longer)
is seen to resume its PD pattern from the early stages of
aging (1 to 5 min aging). If voltage is applied for
another 30 min (after the interruption), the PD pattern
gets the same shape as after the first 90 min of aging.
This is somewhat similar to observations in [5] where
                                                             Fig. 3 – Average PDIV at 100 Hz measured with different
an aging test (on a spherical cavity in epoxy) is seen to    ramp rates. Error bars show 90 % confidence intervals.
give anomalous PD patterns during the first minutes of
voltage supply after a period with voltage removed.          Figure 4 shows PDIV measured at different applied
As conclusion from the aging tests, measurements done        frequencies, and with different pre-excitation (as
after 1.5 h (and up to 7 h) are not significantly affected   described in Section 2.4). The measured PDIV is clearly
by aging. Therefore all specimens are conditioned for        frequency dependent in the cases of pre-excitation at
1.5 h prior to the PDIV measurements. In addition,           variable frequency (a), and no pre-excitation (b). The
leaving a specimen without voltage supply changes the        PDIV increases as the frequency goes up. The
PD activity, at least temporarily. As far as possible this   explanation is probably a larger influence of the
should be avoided, if not intended.                          statistical time lag at higher frequencies, i.e. at shorter
                                                             period times. As the applied voltage amplitude has
3.2. PDIV                                                    exceeded the critical level for discharge, the time spent
The PDIV measurements show similar results for two           above this level during each voltage cycle is shorter at
equal specimens. There is a considerable scatter in some     high frequency than at low frequency. Therefore PDs
of the PDIV measurement results. Certainly part of this      may come later at high frequency resulting in a higher
scatter is due to measurement errors but the most is         PDIV.
probably caused by the statistical nature of PD.             The case of no pre-excitation and 10 min pause between
Therefore, throughout this paper, the results of PDIV        each measurement (c) is somewhat different from the
measurements are shown as average values with error          other two cases. The PDIV at the lowest frequency (0.1
bars indicating 90 % confidence intervals (normal            Hz) is similar to the other cases but the increase in
distribution is assumed).                                    PDIV with frequency is faster. The explanation may be
Figure 3 shows the effect of the ramp rate on the            that the statistical time lag is longer in case (c) than in
measured PDIV. No drastic influence of the ramp rate is      (a) and (b), since there has been a pause of 10 min with
seen for the ramp rates tested here. There is a tendency     no applied voltage prior to each measurement. A longer
to a decrease in PDIV at the lowest ramp rate (0.1 kV/s)     statistical time lag makes the effect of delay of
although it is not pronounced. The data is also less         discharges visible at lower frequencies.
scattered at lower ramp rate. Both these observations        In [7] PDIV for an insulated disc-shaped cavity in
are expected since at lower ramp rate there is less          polyethylene (placed in a dark box) is measured at
influence from the statistical time lag. They are also in    different applied frequencies. If compared to this study,
accordance with results from [2] where a systematic          the most similar case is (c). The results in [7] however
decrease in measured PDIV was seen as the ramp rate          show no clear frequency behavior of the PDIV.
was reduced. It was not possible to lower the ramp rate      The scatter in data is smallest in case (a), larger in (b)
below 0.1 kV/s in this study, otherwise that would have      and largest in (c). This probably comes since the
been instructive.                                            statistical time lag is shorter in the case of pre-excitation
If Paschen’s law (for gaseous breakdown between plane        (a) than in the other two cases, and since measurements
metal electrodes) is used to estimate the PDIV of the        were only repeated five times in (c). Moreover there is a
current specimen, a value of 7.7 kV is obtained. As seen     large scatter in the 100 Hz measurements for both (b)
from Figure 3, the measured PDIV is lower than this for      and (c). This is caused by a longer statistical time lag
all ramp rates. This is in contradiction to [6] where the    here since these are the first measurements done after
measured PDIV of disc-shaped cavities was seen to            the 14 h without voltage supply.
obey Paschen’s law. The disagreement may be due to           Also Figure 4 shows that the measured PDIV is lower
differences in the measurement methods but no detailed       (up to 40 %) than that predicted by Paschens’law. The
description of this is given in [6].
difference is largest for the cases with smallest           4. Conclusions
influence of the statistical time lag.
                                                            Partial discharge inception voltage (PDIV) is measured
                                                            in disc-shaped cavities at different frequencies of the
                                                            applied voltage. The measured PDIV is lower than that
                                                            predicted by Paschen’s law.
                                                            The PDIV is seen to increase with increasing frequency.
                                                            This may be explained by a larger influence from the
                                                            statistical time lag at higher frequencies, causing
                                                            discharges to be delayed and occur at higher applied
                                                            The effect of pre-excitation (voltage applied prior to
                                                            PDIV measurements) on the PDIV is also investigated.
                                                            If the specimen is left without voltage supply for 10 min
                                                            prior to each PDIV measurement, the PDIV is seen to
                                                            increase faster with frequency than if the PDIV
                                                            measurements are done without pauses. A probable
                                                            explanation is that leaving the specimen without voltage
                                                            supply increases the statistical time lag. This causes the
                                                            effect of delay of discharges to appear at lower

                                                            5. Acknowledgements
                                                            This study was sponsored by the Centre of Competence
                                                            in Electric Power Engineering (EKC2), Sweden.

                                                            6. References
                                                            [1] H. Edin, Partial Discharges Studied with Variable
                                                                Frequency of the Applied Voltage, Ph. D. Thesis,
                                                                KTH, Stockholm, Sweden, 2001.
                                                            [2] A. Cavallini, F. Ciani, G. Mazzanti, and G.C.
                                                                Montanari, “First Electron Availability and Partial
                                                                Discharge Generation in Insulation Cavities: Effect
                                                                of Light Irradiation”, IEEE Trans. on Dielectrics
                                                                and Electrical Insulation, 2005, pp. 387-394.
                                                            [3] Power Diagnostix Systems GmbH, Bruesseler
                                                                Ring      95a,    52074,     Aachen,      Germany,
                                                                (, last visited March
                                                            [4] P.H.F. Morshuis, Partial Discharge Mechanisms,
                                                                Ph. D. Thesis, Delft University of Technology,
Fig. 4 – Average PDIV measured with pre-excitation at           Delft, the Netherlands, 1993.
variable frequency (a), no pre-excitation (b) and no pre-   [5] P. Morshuis, A. Cavallini, G.C. Montanari, F.
excitation and 10 min pause between each measurement (c).
                                                                Puletti, and A. Contin, “The Behavior of Physical
Error bars show 90 % confidence intervals.
                                                                and Stochastic Parameters from Partial Discharges
                                                                in Spherical Voids”, in Proc. IEEE Conf. on
                                                                Properties and Applications of Dielectric
                                                                Materials, Xian, China, 2000, pp. 304-309.
                                                            [6] H.C. Hall and R.M. Russek, “Discharge inception
                                                                and extinction in dielectric voids”, IEE Proc.
                                                                Power Engineering, 1954, vol. 101, pp. 47-55.
                                                            [7] W. Hauschild, A. Cavallini, and G.C. Montanari,
                                                                “Effect of Supply Voltage Frequency on Testing
                                                                of Insulation System”, IEEE Trans. on Dielectrics
                                                                and Electrical Insulation, 2006, vol. 13, pp. 1189-

To top