New Method for Measuring Statistical Distributions of Partial Discharge ...

e>the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content.
New Method for Measuring Statistical Distributions of Partial Discharge Pulses
Volume 102, Number 5, SeptemberOctober 1997
Journal of Research of the National Institute of Standards and Technology
[J. Res. Natl. Inst. Stand. Technol. 102, 569 (1997)]
New Method for Measuring Statistical
Distributions of Partial Discharge Pulses
Volume 102
Number 5
SeptemberOctober 1997
Yicheng Wang
National Institute of Standards and
Technology,
Gaithersburg, MD 20899-0001
A new digital detection system is described
for measuring pulsating partial discharges
(PDs). The PD detection system can contin-
uously record all PD pulses that occur over
extended periods of time, with a minimum
inter-pulse time separation of 6
s and a
vertical amplitude resolution of 12 bits.
Earlier PD detection systems detected PD
pulse amplitude and time using complex
custom-designed hardware while the present
system continuously records the complete
electrical waveform that carries the PD
pulses using a commercial data acquisition
board and extracts, in real time, the time
and amplitude information of all PD pulses
in software. The current approach consider-
ably reduces the development and mainte-
nance cost of the PD detection system, sig-
nificantly increases the system portability,
and may prove to be a crucial step for
transferring the digital PD detection and
analysis technology developed in laborato-
ries to industry. The features of the new
system are illustrated by the study of dc-
excited PD pulses occurring in a point-to-
plane gap in air. A new surface-mediated
burst mode of PDs is discovered in which a
PD pulse has a certain probability to in-
duce another pulse. The probability is de-
termined for several gap voltages and is
found to vary strongly with the applied
voltage.
Key words
: amplitude distributions; log-
normal distribution; partial discharges;
random point process; time separation
distributions.
Accepted:
June 11, 1997
1.
Introduction
Partial discharge (PD) measurements are widely
employed in industry for the evaluation of electrical
insulation and HV apparatus [13]. In recent years,
digital techniques for PD measurements have allowed
researchers to gather a tremendous amount and variety
of PD data and to envision new applications of PD
measurements [4, 5]. This paper addresses a new digital
method for PD detection and analysis. To better illustrate
the features of the new system, let us consider dc-excited
PDs that occur in a point-to-plane gap in air as an
example, as shown schematically in Fig. 1.
When the voltage across the gap is sufficiently high,
a pulsating PD current appears through the gap and can
be detected by an external circuit. The intrinsic current
pulses of the PD waveform are very narrow ( 1 ns)
compared to the time separation (usually > 1
s) be-
tween two adjacent pulses and can be attributed to the
motion of the electrons created in the discharges [6].
The PD pulses form because once a discharge is ini-
tiated, the electrons are quickly depleted in the gap,
either by striking the point electrode in the case of a
positive PD (the point is positive with respect to the
plane), or by attachment to gas-phase molecules for a
negative PD. The space charge so created reduces the
electric field near the point, quenches the discharge, and
inhibits any new discharge until the space charge dissi-
pates. The discharge magnitude q , as measured by the
integrated charge of the pulse, varies from one PD pulse
to another because the size of a discharge depends, in
part, on the gas composition in the gap which is altered
by the presence of ions, metastables, and radicals
produced by the previous pulse. The size of this previous
pulse, in turn, depends on the influence of its previous
pulse, and so on, indefinitely. In general, the time
separation from the previous pulse t also changes from
one PD pulse to another in a non-deterministic fashion,
569 Volume 102, Number 5, SeptemberOctober 1997
Journal of Research of the National Institute of Standards and Technology
parameters and cannot acquire a complete record of the
PD pulse train in a given time interval. Multichannel
analyzers have been used often for PD analysis, mainly
for the measurement of pulse-height distribution and
corresponding pulse rates [10, 11]. However, the current
trend in digital PD recording and analysis is clearly to
adopt computer-based measurement systems. For exam-
ple, Okamoto and Tanaka [12] described a custom data
acquisition system which allows for the simultaneous
measurement of height and phase angle of ac-excited
PD pulses. Recognizing that PD is, in general, a non-
Markovian, random point process in which memory ef-
fects play an important role, Van Brunt and coworkers
[13, 14] developed and reported in this journal a real-
time analog PD stochastic analyzer that allows direct
measurement of various conditional distributions. The
analog analyzer, however, filters the data as received
with a preset circuit and thus does not provide a com-
plete record of all PD events during the time of observa-
tion. To remedy these limitations, von Glahn and Van
Brunt [15] later designed a custom PD digitizer inter-
faced to a personal computer. Their digitizing system
generated a complete record of all PD pulses in an
indefinitely long period and subsequently allowed a
complete stochastic analysis of the recorded data. Their
method has revealed quantitatively, for the first time,
how PD can become non-stationary even under rela-
tively simple, well defined discharge gap conditions and
has brought attention to the inadequacy of conventional
PD testing standards which rely only on the apparent
charge. The potential of their approach for further con-
tributing to the understanding of the partial discharge
process has been well recognized [4]. However, their PD
digitizer exhibits the shortcomings of typical custom-
designed fast digital circuitry, such as high development
and maintenance cost, sensitivity to noise, and difficulty
in keeping pace with changing personal computers.
These limitations of the digitizer hinder its application
in the field as a diagnostic tool for the evaluation of HV
apparatus and cables.
In this paper, a different, much simpler digitizing
system is described for PD measurements and analysis.
The present approach digitizes the entire PD current
waveform (Fig. 1) and relies on software to sort out PD
pulse information {q
i
, t
i
}, while the instrument of von
Glahn and Van Brunt detects each PD pulse using
custom-designed hardware, with the computer serving
chiefly as a storage device. The much faster personal
computers
and
better-designed
commercial
data-
acquisition boards now available make fast digitizing
and concurrent sorting possible without any dead time.
The design of the instrument is described in the next
section and the capability of the new PD measurement
system is illustrated with the results obtained for
Fig. 1.
(a) Schematic diagram of the partial discharge detection
system. (b) A portion of a typical long trace acquired with the system.
depending on how the initiatory electrons are created
and how fast the space charge is cleared.
While a PD process can be viewed as a random point
process [7] and can be simply specified by a finite set of
two random variables {q
i
, t
i
}, i = 1, 2, 3, . . . , the
process not only defies a complete theoretical modeling
but also continues to challenge experimenters for better
recording systems so that the stochastic properties of PD
processes can be unraveled [8]. Theoretical difficulties
in dealing with partial discharges stem in part from the
lack of knowledge about how a PD modifies the gas
composition in the discharge gap and how the altered
gas in turn influences the next PD [9]. Experimentally,
there is a lack of commercial instruments that can be
used to record all PD pulses in a sufficiently long period
to yield statistical properties of the PD pulses.
Most digital techniques used for partial discharge
measurements are limited to analyzing a few chosen
570 Volume 102, Number 5, SeptemberOctober 1997
Journal of Research of the National Institute of Standards and Technology
dc-excited partial discharges in a point-to-plane air gap
in Sec. 3.
2.
Measurement System
A simplified block diagram of the measurement sys-
tem is shown in Fig. 1(a). The system was tested with
partial discharge pulses occurring in a point-to-plane air
gap. For such a discharge configuration, PD pulse
current can be sensed and conditioned with a simple RC
circuit in front of a high-impedance preamplifier. The
values of R and C are selected so that the RC circuit
functions as an integrator during a PD pulse, i.e., the
current leaked through R during the PD pulse (typically
< 10 ns) is negligible. For example, C = 1 nF and
R = 40 k
were chosen for gap separations ranging
from 0.5 cm to 4 cm. A typical PD waveform which
appears at test point A is also shown in Fig. 1(b). Each
vertical step of the waveform corresponds to a PD