1544 S. Tenbohlen et al.

: Partial Discharge Measurement in the Ultra High Frequency (UHF) Range

Partial Discharge Measurement in the Ultra High

Frequency (UHF) Range
S. Tenbohlen, D. Denissov and S. M. Hoek
Universität Stuttgart
Pfaffenwaldring 47
70569 Stuttgart, Germany

and S. M. Markalous
LDIC Diagnostic GmbH
Zschoner Ring 9
01723 Kesselsdorf, Germany

ABSTRACT
The paper provides essentially a summary of PD measurements applying the UHF
range in order to increase the detection threshold, to improve the localization accuracy
and to perform on-line measurements of Partial Discharge (PD) in noisy environments.
The electromagnetic UHF technique offers good signal to noise ratios, because external
PD signals and disturbances can be shielded effectively. A new developed method
allows the localization of PD in gas-insulated substations (GIS) by frequency domain
measurements. The basic idea is the displacement law of Fourier transformation. The
interference phenomena of superposed signals from two sensors give information about
the time delay of the sensor signals. On-site PD measurements are made at cable
connectors by means of monopole antennas housed in a barrel sleave, while the cable is
in service. Thus a sensitive PD measurement even in noisy environment is possible. PD-
measurements on several 72 kV cable connectors were performed in an unshielded
laboratory. On-site measurements during operation showed the great potential for
condition assessment. For decoupling sensitive UHF PD signals from the inner of a
power transformer tank UHF sensors applied through drain/oil valves are used.
Experimental studies indicate that all relevant types of PD possibly occurring within a
transformer emit high frequency spectra to be detected with UHF sensors.
Furthermore in laboratory experiments and on-site measurements very moderate UHF
signal attenuations have been observed.
Index Terms — Gas insulated switchgear, cable accessories, power transformers,
on-site partial discharge measurement, UHF-sensor, condition assessment.

installed equipment in service by means of PD

1 INTRODUCTION measurements. Unfortunately conventional methods show
PARTIAL discharges (PD) can constitute a high risk for some drawbacks and limitations if performed on-site/on-
the dielectric stability of an insulation system. Therefore PD line e.g. regarding the applicability of sensors and being
measurement is of a tremendous importance to assess the receptive for several disturbances. For offline
condition of high voltage equipment. Different PD measuring measurements there are possibilities to reduce external
techniques are using different physical peculiarities of the PD noise e.g. use of spectrum analyzer and external voltage
phenomenon e.g. electric currents (according to IEC 60270), source with variable frequency.
gas formation (dissolved gas analysis), electromagnetic In general three information are important regarding a
(UHF-range) or acoustic radiation. Well-known and PD activity – its level, its type and its location. The
approved partial discharge (PD) measurements according to information of the PD origin together with the knowledge
IEC60270 and measurement of μV according to IEEE of the surrounding insulation material is essential to assess
.57.12.90-1999 are the basis for e.g. acceptance tests of the e.g. the risk potential of the fault, because there are PD-
insulation system of high voltage equipment. As a powerful resistant materials for example or different types of PD-
diagnostic tool there is an increasing demand to evaluate also sources which may disappear during the test. It is of great
importance to know about the PD origin to plan and start
Manuscript received on 10 April 2008, in final form 18 August 2008. maintenance / repair actions cost and time efficiently.

1070-9878/08/$25.00 © 2008 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 6; December 2008 1545

The advantages of the unconventional PD measuring an expensive fast digital oscilloscope is more economical.
methods are their (i) wide immunity against external The interference phenomena of two sensor signals, which
disturbing signals on-site, (ii) the fact that the sensors need are added, give information about the time delay (Δt)
no electrical connection to the high voltage circuit and (iii) between the signals. The idea is based on the displacement
the inherent possibility of determining the failure location law of a Fourier-transformation (equation 2) of the received
(localization of the PD) using arrival times of these signals. signals.
Corona as strongly interfering electrical process on-site, has
only a reduced influence on UHF measurements (contrary [
FFT f (t − Δt )] = FFT [ f (t )]⋅ e − jωΔt
(2)
to electric PD measurement according to IEC 60270). The
electromagnetic (UHF) technique offers very low noise To visualize the interference phenomena, three power
levels in GIS and transformers since the housings act as spectra are needed. The power spectrum is the absolute
Faraday cage shielding external noises successfully. An value of the complex Fast Fourier Transformation (FFT).
often mentioned drawback of these methods is that no The three measured power spectra are obtained from
assured apparent charge information (electric level in pC) is Sensor 1 (equation 4), Sensor 2 (equation 5) and the added
delivered. This issue is increasingly addressed with so- signal of Sensor 1 and 2 (equation 6) with a conventional
called Performance and Sensitivity Checks [1 - 3]. But it spectrum analyzer. The last signal is obtained by means of
has to be considered that exact measurement of apparent a Radio Frequency power combiner. These three signals are
charge is not really needed for measurement on complex combined in equation 3 in order to get the theoretical
insulating system, for the investigation of the risk the cosine function. The time difference (Δt) can be calculated
behavior of the PD-source during the test is much more with the resulting cosine function in case the approximation
important. f(t) ~ g(t) is possible. This cosine function has equidistant
The different application aspects of the UHF PD minima (Figure 2). The distance between these minima
measuring methods and their interpretation will be represents the interference frequency Δf = 1/Δt , which can
presented in this contribution for GIS, cable accessories and be interpreted as interference phenomena [3].
power transformers.
H (ω ) ⎛ ω ⋅ Δt ⎞
2 GAS-INSULATED SWITCHGEAR = K (ω ) ≈ K t (ω ) = cos⎜ ⎟ (3)
F (ω ) + G (ω ) ⎝ 2 ⎠
Beside acoustic impulses, which also can be used for
localization, very fast electric pulses with rise times below F (ω ) = FFT [ f (t )] (4)
1 ns, are emitted by a PD source and propagate in all
directions along the GIS duct. A simple and obvious way of G (ω ) = FFT [g (t )] (5)
locating PD in GIS is the measurement with the time-of-
H (ω ) = FFT [g (t ) + f (t − Δt )] (6)
flight method. By the time-of-flight technique the time
difference between the wave fronts arriving at two UHF-
Two similar signals f(t) and g(t) are required to obtain
PD-sensors indicates the location of the PD source. The
useful results from equation (3). To keep the characteristics
time difference (Δt) is usually in tens of 1 ns, so that a fast of both signals similar, the effect of dispersion should be
digital acquisition has to be applied for measurements. kept as small as possible [4]. The group velocity (vg) of the
Transversal Electric- / Transversal Magnetic-wave modes is
frequency dependent, which is a precondition for
dispersion. Below the lowest critical frequency of all modes
(in GIS the fc of TE11), only TEM-modes are able to
propagate [5]. A requirement for a successful result (Kt ≈
K) is a sensitive measurement in this frequency range.
2.2 Interpretation of Frequency Domain
Measurements
Figure 1. Cross section of a GIS with PD-sensors. With different GIS types and set-ups the interference
phenomena are not always clearly visible and the distance
The distance X1 can be calculated with the equation (1) between the minimum (interference frequency) Δf is not
in case the time difference (Δt) is known. manually estimable. An objective method is necessary in
order to fit the combined measurement K(ω) with
X − ( X 2 − X 1 ) X − c0 ⋅ Δt theoretical cosine functions Kt(ω) of different Δt. The best
X1 = = (1)
2 2 correlation in a manually selected frequency range of the
measured combined signal with the theoretical cosine
2.1 THEORETICAL BACKGROUND function is determined by the calculation of the maximum
Another method to determine the time difference (Δt) cross-correlation. The theoretical function with Δf = 1 / Δt
and to localize PD in GIS is to use the frequency domain. A possesses the largest correlation and thus the value of the
measurement procedure with a spectrum analyzer instead of cross-correlation is maximal [4]. A disadvantage is, that the
1546 S. Tenbohlen et al.: Partial Discharge Measurement in the Ultra High Frequency (UHF) Range

interesting frequency range of the Cross-Correlation x 10

7

Method (X-Corr. Method) and the Minimum Method must

be determined manually (Figure 2). 1

Interference frequency [Hz]

To determine the distance Δf of the minima even with
more complex measurement set-ups, an analysis with 2
wavelet transformation can be applied. The wavelet family
3
is chosen very similar to the theoretical cosine function
Kt(ω). Thus a large selectivity to the searched interference 4
is possible in relation to resonances and disturbances in
complex test set-ups. The result of this Wavelet Method 5
shows the similarity of the measured spectrum K(ω) over 6
the complete frequency range ω and theoretical cosine
Kt(ω) with a certain interference frequency Δf. The 100 150 200 250 300 350 400 450 500
differentiation between an interference, which belongs to a Frequency [M Hz]
corresponding delay time, and another, which belongs to
disturbances or reflections, takes place over maximum Figure 3. Wavelet Method applied to measurement at 300 kV GIS bay.
values and plausibility (Fig 4). A good result is visible by max = 121.3297 , f = 2.25333e+008
7
increased signal energy over a concrete interference x 10 Intf = 2.2e+007 , t = 4.54545e-008

frequency Δf and over a wide frequency range. The 1.8

absolute value of the time delay |Δt| can be estimated by the

Interference frequency [Hz]

interference frequency Δf. 2.0
2.3 EXAMPLE
Figure 2 shows the measurement of the interference 2.2
phenomena in a bay of a 300 kV GIS. The two sensors are
capacitive UHF-PD-sensors, placed at the bus bar and at
2.4
the termination of the GIS (circuit breaker: open). This
distance corresponds to the typical distance of sensors. The
PD-source was a pulse generator with an antenna. 2.6
150 200 250 300
5 Frequency [M H z]
Figure 4. Detailed view of the Wavelet Method (Fig 3) with maximum
0
evaluation.

3 CABLE ACCESSORIES
-5
3.1 PRINCIPLE OF DIAGNOSIS AND LAB TESTS
[dB]

-10
Utility experience shows that poor termination and
jointing is the major cause of cable failures [6]. This is due
to the fact that, in contrast to the cable itself, these
-15
components are assembled and installed under on-site
theoretical cosine
conditions and thus exposed to the higher risk of defects
100 150 200 250 300 350
Frequency [MHz] and contaminations [7].
Figure 2. Calculated combination of the power spectra for the Modern plug-in cable connectors (terminations) for GIS
measurement at the 300 kV GIS. and transformers are made from silicone rubber. The
The time difference is measured as Δt = 45 ns by using electrical life span of this high polymeric material normally
the oscilloscope. The theoretical cosine function matches exceeds 40 years, but only in absence of PD activity that
best at Δf = 21.2 MHz with the X-Corr Method in a inevitably causes material’s degradation. Although there
manually chosen frequency range (Figure 2). The Δt is are several well known off-line test techniques (i.e. dc,
calculated as 47 ns. damped oscillating wave, very low frequency voltage tests),
In this example the Wavelet Method evaluates a good result. which are successfully applied to diagnose power cables
The maximum value is at Δf = 22.5 MHz. The time difference including their accessories, they all need a line
Δt can be calculated as Δt = 45.4 ns (Figure 3 and 4). disconnection, load flow redispatching and a separate
Because of the more complex arrangement the evaluation voltage source to energize the cable line apart from the
of the measurement at 300 kV GIS bay (Figure 2) is not as network [8]. The online UHF test approach overcomes
simple as the measurement at a bus bar or a GIL. The these difficulties allowing sensitive measurement on the
reasons are the additional reflections in the GIS. More terminations, while the cable is in normal operation [9].
complex methods like the Wavelet Method are able to Figure 5 demonstrates the principle of UHF diagnosis on
interpret these signals. the plug-in cable connectors. A portable metallic housing is
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 6; December 2008 1547

clamped on the cable behind the connector and fulfils two

functions: firstly, as a housing for field couplers (antennas)
1
and secondly, as a grounded screen against the disturbances
from outside.
Bushing Connector Inductive
UHF sensor
BNC jack
2

PD defect Figure 6. On-site test set-up: portable screening sleeve with the UHF
sensors (2) mounted on each termination (1) in turn.
Portable Capacitive
0 ,6
screening sleeve UHF sensor

Figure 5. Plug-in cable termination and principle of the UHF PD 0 ,4

diagnostics.

Amplitude (V)
0 ,2
The transient electromagnetic field caused by PD can be
detected by coupling its electric (capacitive sensor) or 0 ,0
magnetic (inductive sensor) components. The capacitive
sensor represents a copper disc with the diameter of 2 cm, - 0 ,2

soldered to the copper pin in the middle. The inductive

sensor is a two-winding coil made from an insulated wire. - 0 ,4

One end of the coil is grounded; another one is connected -1 0 0 0 100 200 300 400 500
to a measuring coaxial cable RG-214. T im e (n s )
Nanosecond pulses emitted at the PD fault location and Figure 7. Typical fast pulse emitted by the faulty termination on-site.
captured with different sensors are pre-amplified, filtered
and processed with a computer-based oscilloscope. As Range of
diagnosis parameters the pulse’s shape, its spectral amplification
characteristics and repetition rate are considered. Phase
resolved PD patterns were taken to figure out if the pulse
activity is power cycle related [10].
The calibration of the UHF method in terms of amount
of charge is not possible. In order to perform a sensitivity Reproducible
check comparative PD measurements were performed with frequency
built-in artificial defect using both the UHF method and the components of the
measured fast
conventional method according to IEC 60270. An apparent impulse
charge below 5 pC turned out to be detectable by both type
of UHF sensors. Lab tests also demonstrated that one could
easily discriminate the internal PD defects in the
termination insulations from corona noise originated in
neighboring equipment insulated with air or SF6 [10].
3.2 ON-SITE EXPERIENCE
Figure 8. Frequency spectrum of PD-pulse at cable connector.
On-site PD measurements are made on cable
terminations in the manhole of a GIS, while the cable is in
operation as shown in Figure 6. 4 POWER TRANSFORMERS
Figure 7 shows such a typical fast pulse picked up on a For decoupling UHF PD signals from the inner of the
connector on-site. The frequency spectrum of this pulse is transformer tank UHF sensors (acting mainly capacitive) can
plotted in ffigure 8. Besides some broadcast (DVB-T) and be installed through a drain/oil valve. Different sensor designs
GSM frequency spikes, there are several other high - monopole-formed, disc-shaped and cone-shaped – can be
frequency components that indicate presence of PD distinguished. While for laboratory test both types have been
activity. The frequencies around 700 MHz and 1.5 GHz used, only the disc-shaped and cone-shaped sensors were
were detected with this specific connector only, thereby applied in on-line measurements (Figure 9). The non-
indicating a critical condition. 10 days after the destructive application of the UHF sensors can be managed
measurement the connector failed and thus the indication while the transformer stays in full service, since there is no
was proved. galvanic connection needed to the high voltage circuit. Very
1548 S. Tenbohlen et al.: Partial Discharge Measurement in the Ultra High Frequency (UHF) Range

low noise levels (due to shielding effects of the transformer

tank) and low signal attenuation in solid insulation materials, 0.35
oil and complete structures within transformers enable 0.30
sensitive measurements even under noisy on-site condition.
Furthermore aspects as (i) verification of hydrogen and 0.25

amplitude (Vs)
electrical PD measurements, (ii) alternative robust and easy 0.20
PD decoupling, (iii) distinguishing between internal and
external PD signals and finally (iv) locating of PD with 0.15
additional acoustic measurements (as e.g. reported on a 0.10
200 MVA, 380/220 kV-single-phase Transformer [11])
underline the practical meaning of UHF measurements on 0.05
transformers. 0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0
frequency (GHz)
Figure 11. UHF PD spectrum of a surface discharge PD-source in oil up to
3 GHz (apparent charge of 356 pC).

As internal PD sources a needle-sphere source and a

Figure 9. Active UHF drain/oil valve sensors for oil-insulated surface discharge source were used. The source for surface
transformers (right: during installation). discharges consisted of an electrode sharpened on one side,
a ground plate and a pressboard between them.
4.1 SPECTRA OF UHF PD PULSES
Figure 10 shows an UHF PD spectrum of a needle-
The first question to answer when measuring PD in the sphere source (apparent charge 11.6 pC). Significant
UHF band on a power transformer might be whether PD of frequency components started around 385 MHz and went
all types are sufficiently fast to radiate UHF pulses in oil. In up to 2 GHz. The rise time of the UHF pulses was typically
the following a brief discussion of UHF spectra of different about 0.5 ns for that source.
PD sources are presented. The UHF PD pulses generated by the surface discharge
All experimental measurements were conducted in such source featured as well pulse rise times down to about
a way, that an occurring UHF PD signal triggered a 0.3 ns but sometimes also slightly slower pulses (rise times
simultaneous electric PD measurement, to get the apparent up to 2-3 ns). The UHF pulses remained detectable at all
charge value according to IEC 60270. The relation to the times. In Figure 11 an UHF PD spectrum of a surface
standard PD-measurement method was hence obtained for discharge (apparent charge of 356 pC) with frequency
every UHF PD signal. To capture the un-amplified UHF components up to around 1.6 GHz is shown.
signals, a transient recorder with analogue bandwidth of To generate external corona, the needle-sphere source
3 GHz was used. The experimental set-up for the was put outside of the tank in free air and the sensor on a
investigation of “internal” PD sources consisted basically position with 1.1 m distance and an inclination of 45° with
of a metal test tank with dimensions (1.0 x 0.5 x 0.5) m3 respect to the source. Figure 12 gives the resulting UHF PD
without metal cover, the respective PD source and one disc- spectrum of the needle-sphere source in air with a quite
shaped UHF sensor [12]. The complete set-up was located high apparent charge of 2580 pC. The relevant frequency
in a shielded laboratory. The open tank should offer a quick components decayed rapidly above 250 MHz. Thus
decay of the UHF PD signals without strong resonances distinction between external and internal PD is possible by
and hence limit the impact of the metal housing on the the analysing the frequency content of the PD signal.
appearance of the electromagnetic signal.

0.6 10

0.5
8
amplitude (Vs)

0.4
amplitude (Vs)

6
0.3
4
0.2

0.1 2

0.0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0.00 0.25 0.50 0.75 1.00
frequency (GHz) frequency (GHz)
Figure 10. UHF PD spectrum of a needle-sphere PD-source in oil up to Figure 12. UHF PD spectrum of the same needle-sphere PD-source in air
3 GHz (apparent charge of 11.6 pC). up to 1 GHz (apparent charge of 2580 pC).
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 6; December 2008 1549

PD measurements of voids and particles on floating numbers and at least two of them have to be non-zero. In
potential are well reported. A pressboard void arrangement Figure 13 the UHF PD spectrum of a rod-plane PD source
showed peak frequencies between 500-800 MHz in [13] is shown together with some calculated cavity resonances
while floating particle discharge had spectra up to 600 MHz (values of m, n, p are given in the boxes). The measured
[14] or up to 1 GHz [13] respectively. cavity resonances show fairly good agreement with the
The characteristics of the stated UHF spectra of different calculated ones.
PD types are in general in good agreement with reported
studies on similar PD defects [13-18]. This is indicating
4.3 UHF PD SIGNAL ATTENUATION
that all relevant types of PD, possibly occurring within a
CHARACTERISTICS OF DIFFERENT MATERIALS
transformer, emit sufficiently high frequency spectra to be
AND STRUCTURES
detected with UHF sensors.
The subject of this investigation is the basic impact of
4.2 CAVITY RESONANCE INVESTIGATION damping and alteration of UHF PD signals, traveling
The measurable UHF spectrum of a PD in a transformer through different insulating materials and whole structures,
depends, beside the natural radiation from the PD, strongly which make up large parts of power transformers. The goal
on different transfer functions included in the propagation was to get an integral indication on how strong and in
path from source to measuring system. In addition to the which way the UHF PD pulses are affected through these
sensor and cable characteristics, the surrounding materials generic structures. With the experiments dominantly
and objects with their properties of transmission, scattering existent materials and structures within transformers were
and reflection have a strong impact on the appearance of modeled. These materials and structures respectively have
the recorded UHF PD signal. The transformer tank, as an been 8.4 cm pressboard, a 0.5 cm gap in a metal shielding
almost closed metal enclosure, is highly reflective for and a sector of a disc winding.
incoming electromagnetic waves. To emphasize that An important aspect of the tests was that always one and
scattered nature of internal UHF signals inside of metal the same UHF PD pulse was measured with two
enclosures, a consideration of possibly present cavity independent disc-sensors. The whole propagation paths,
resonances was made in a laboratory investigation. disc-sensors and connecting cables were as similar and
The experimental measurements were again arranged in symmetric as possible. The impact of a structure put in one
a small test tank with the dimensions of 1.0 m x 0.5 m x propagation path can consequently be analyzed directly
0.5 m. A monopole antenna was used for decoupling the through comparison with the second measured signal. On
UHF PD pulses of a rod-plane PD source [12]. PD of one hand the measured UHF PD time signals and their
apparent charges down to 50 pC was measured with a spectra and, on the other hand, the signal energy of the
transient recorder (analogue bandwidth 1 GHz) without any pulses was used for comparisons.
amplification. The laboratory setup basically contained a half-closed
spectrum of 285 pC UHF PD signal metallic test tank (1.0 m x 0.5 m x 0.5 m), a needle-sphere
cavity resonances PD source, two identical disc-sensors (see Figure 13 for
1.0
011 111 112 012 131 312 233 schematics) and a transient recorder with an analogue
210 311 132 323
bandwidth of 3 GHz. No additional amplification was used.
0.8 A metal wall between them electromagnetically decoupled
amplitude (Vs)

the two sensors. This metal wall divided the closed half of
0.6 the tank in two sections, with one sensor in every part
(Figure 14).
0.4

0.2 Side 1

PD

0.0
Side 2
εr1 *
0 100 200 300 400 500 600 700 800 900 1000
frequency (MHz)
a) b)
Figure 13. UHF PD spectrum of a rod-plane PD source (apparent charge
Side 1
285 pC) with some dominant cavity resonances.
Quelle
PD
Regarding the tank as cavity bounded by conducting *
walls the following equation Side 2

2 2 2
c0 ⎛m⎞ ⎛n⎞ ⎛ p⎞ (7) c) d)
f nmp = ⎜ ⎟ +⎜ ⎟ +⎜ ⎟
2 εr ⎝ a ⎠ ⎝b⎠ ⎝ c ⎠ Figure 14. Schematics of the 2-path test setup: a) top view of the two-
defines the frequencies of cavity resonances, which can be sided experimental setup with half-closed test tank (1.0x0.5x0.5) m3, two
calculated analytically. The symbol c0 stands for the speed disc-sensors and needle-sphere PD source; b) front view of the setup with
8.4 cm pressboard in 'side 2'; c) top view of the setup with a 0.5 cm gap in
of light, εr can be filled with 2.2 for oil and a, b, c are the
a metal plate in 'side 2'; d) front view of the setup with stacked parts of a
geometric dimensions of the tank. For a proper computation disc-winding in 'side 2' (with gaps of 0.5 cm width within the disc-
of the cavity resonances m, n, p should be filled with whole winding).
1550 S. Tenbohlen et al.: Partial Discharge Measurement in the Ultra High Frequency (UHF) Range

This so called '2-path-arrangement' offers the great 2.70 dB) in a series of fourteen measurements (ranging
advantage to independently measure one and the same from 1.1 – 6.0 pC) with an increased standard deviation
UHF PD pulse simultaneously on two sensors. of 8.30 %.
Prior to the analysis of different materials and
structures, reference measurements with the solely oil- 6
filled '2-path-arrangement' were carried out. In this case 'side 1'
it could be examined to which degree the two measured 4
signals resemble each other, when they should ideally be 'side 2'

amplitude (mV)
identical, due to the fact of two identical propagation 2
paths. The energy content of both pulses was calculated
and compared. Since throughout the experiments 'side 1' 0
remained unchanged (see Figure 14) – the calculated
signal energy of 'side 1' acted as the reference value. The -2
mean value for the pulse energy attenuation in the
reference `'oil-oil' test without any barrier on ‘side 2’ -4
was 4.27 % (-0.19dB) with a standard deviation of
approximately 4.36 % in a series of seven measurements -6
(ranging from 1.8 – 3.7 pC). -50 -25 0 25 50 75 100
To simulate a huge amount of solid insulation material
time (ns)
in the propagation path of an UHF PD signal two pieces
of thick pressboard, with dimensions of Figure 15. UHF PD signals of a needle-sphere PD-source in oil up to 3
(500 x 247 x 42) mm3, were used and put into 'side 2', as GHz (bold line: 'side 1' - oil path; thin red line: 'side 2' - path included
shown in Figure 13 b). Investigation of a two-channel pressboard; apparent charge of 2.7 pC).
measurement of a 2.7 pC UHF PD pulse revealed light
attenuation. Comparing the amplitudes of the un- The last experimental setup included a part of a disc
attenuated and attenuated signal in time domain winding in 'side 2' (Figure 14 d)). The disc-windings
(Figure 15) a difference in the maximum pulse interaction with UHF PD signals, which are propagating
amplitudes (here from around 5 mV to 4 mV for the inclined through it, were studied. Due to the fact that the
positive peek) is visible. Concerning the signal energy of winding comprises metal (copper), pressboard spacers and
the pulses, an average attenuation of 34.46 % (-1.83 dB) paper a combination of the earlier observed effects of
was recognized for the 8.4 cm thick pressboard. The frequency-dependent attenuation and delayed detection
standard deviation was 4.28 % in seven measurements were recognized (time delay again approximately 200 –
(ranging from 2.7 – 6.9 pC). Due to the pressboard the 300 ps). The average attenuation of the described disc-
signals featured a slight time shift of about 200 – 300 ps. winding investigation was 38.25 % (-2.09 dB) in eleven
This effect on the UHF signal was expected, since the measurements (ranging from 2.1 – 7.1 pC) with a standard
permittivity of the pressboard (εr around 3.5) is higher deviation of 8.87 %.
than that of oil (εr around 2.2) resulting in a reduced Summarizing the three experiments leads to the
propagation velocity for the electromagnetic waves assumption that UHF PD signals can propagate within the
within the pressboard. whole transformer with comparatively low attenuation
The investigation with a 0.5 cm gap in a metal plate factors. For PD occurring within windings either a
was conducted to examine the impact of metallic propagation path in the gap of e.g. high and low voltage
structures with a certain aperture as e.g. the gap between winding or even through a disc winding is possible. The
high and low voltage winding (which is certainly in the approximate average values of the observed attenuation
range of several centimetres for power transformers) or factors were:
passages through grounded components within a
transformer, when no line-of-sight between source and - 34 % (or 1.8 dB) for the pressboard (thickness 8.4 cm)
sensor is provided. The electromagnetic waves radiated - 46 % (or 2.7 dB) for the 0.5 cm gap in a metal plate
from the PD cannot propagate through the metal plate, - 38 % (or 2.1 dB) for the sector of a disc winding.
but have to reach the sensor of 'side 2' by mechanisms of
reflection or diffraction. Figure 14 c) illustrates the Recent UHF attenuation measurements on 220 MVA
experimental configuration as well as a possible reflected single-phase autotransformers showed attenuation
path. The experimental results showed an attenuation and magnitudes (for e.g. a propagation path through complete
delay. The sensor behind the metal plate was clearly HV/LV-windings etc.) [19] comparable to expected
responding later (delays ranging from 2.9 -5 ns). The attenuation factors of the investigated generic structures.
corresponding spectra showed stronger differences in the The attenuation of UHF signals within the transformer
higher frequency ranges (600 – 840 MHz and 1.35 – varied there between 1 and 6 dB/m, depending on the
1.55 GHz) than observed before with pressboard. The location and therewith the propagation path of the UHF
average attenuation of the pulse energies was 46.31 % (- signals.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 15, No. 6; December 2008 1551

4.4 CASE STUDIES Judging from the high frequency components up to

4.4.1 PD ACCEPTANCE TEST IN SHIELDED 750 MHz – 1 GHz the UHF pulses were regarded as true
LABORATORY internal PD pulses. Additionally the UHF PD pulses
Performing the PD acceptance test in a shielded could be used to enhance acoustic PD measurements
laboratory on a 3-phase autotransformer (450 MVA, with an UHF-triggered acoustic averaging to locate PD
400 kV; dimensions greater than (10m x 3.5mx 4m)) [11]. Thus the hydrogen and electric PD measurements
simultaneous electric PD measurements according to IEC have been verified with sensitive UHF PD measurement
60270 and UHF PD measurements have been carried out on-site. Finally the acoustic location result was verified
[13]. The UHF sensor was inserted in an oil valve at the from the transformer manufacturer to be a “known
bottom of the tank and connected to a transient recorder suspicious position”.
with an analog bandwidth of 3 GHz without any amplifiers.
The electric PD measurements showed PD activity on low-
voltage bushings of two phases up to 70 pC. Additionally 5 CONCLUSION
UHF PD pulses with frequency components up to 600 MHz Unconventional electromagnetic (UHF) PD
have been recorded. A further proving aspect has been that measurement applied on-site offers the great advantage
UHF signals have been detectable for UHF sensor insertion to be more immune against disturbances than PD-
depth from 10 mm inside the tank to 20 mm outside of it measurement according to IEC 60270. Examples for
(with regard to the inner tank wall) before fading away GIS, cable accessories and oil-paper insulated power
when pulling out the sensor further. transformers show the improved PD detection and
The geometrical distance from the low-voltage bushings localization. Localization in GIS can be done both in
to the inserted UHF probe was approximately 10 meters. the time and frequency-domain. The latter allows a
Nevertheless this low-level PD was detectable with the very cost-efficient measurement.
UHF sensor; hence the resulting UHF attenuation within Also, condition assessment of cable connectors can
the active part of the transformer is sufficiently low. The be performed on-site and on-line by means of UHF PD
final internal PD was not investigated further, since the PD measurement. A metallic sleeve equipped with
was below acceptance level and therefore the transformer monopole antennas is clamped around the plug-in
did pass the Factory Acceptance Test. connector to detect electromagnetic waves emitted from
PD within the connector. Thus, a sensitive PD
4.4.2 ON-SITE AND ON-LINE UHF PD measurement even in noisy environment is possible.
MEASUREMENTS UHF PD signals can be decoupled from the inner of
After indication of PD by means of dissolved gas transformer tanks by means of sensitive UHF sensors,
analysis on-site PD measurements were performed on a which could be applied in service through a drain
200 MVA single-phase transformer over a period of valve. The spectra of these signals reflect cavity
several months. An electric PD measurement revealed resonances of the tank. Experimental results regarding
levels up to 600 pC during an offline applied voltage test. the sensitivity furthermore demonstrate a moderate
For the carried out electromagnetic on-line UHF attenuation of the UHF PD pulses within the presence
measurements, a disc-shaped UHF sensor was applied of solid insulation or structures inside transformers.
[13]. Again a transient recorder with an analogue UHF PD pulse propagation even through disc-windings
bandwidth of 3 GHz and no amplification was used for seems possible. On-line UHF measurements and
recording of the UHF signals. The measurements featured laboratory experiments on power transformers featured
very low noise levels. Figure 16 shows an on-line very low noise levels and confirmed the high UHF PD
recorded UHF signal with its corresponding spectrum. detection sensitivity.

0.6

0.5 REFERENCES
[1] S. Coenen, S. Tenbohlen, S. M. Markalous and T. Strehl,
0.4 “Performance Check and Sensitivity Verification for UHF PD
amplitude (Vs)

Measurements on Power Transformers”, 15th International

0.3 Symposium on High Voltage Engineering, Paper No. T7-100
Ljubljana, Slovenia, 2007.
[2] CIGRE TF 15/33.03.05, “PD Detection System for GIS:
0.2
Sensitivity Verification for the Method and the Acoustic
Method”, Electra No 183, pp. 74 - 87, 1999.
0.1 [3] S. Meijer, P. D. Agoris and J. J. Smit, “UHF PD Sensitivity
Check on Power Transformers”, 14th International Symposium
0.0 on High Voltage Engineering, Beijing, China, 2005.
0.00 0.25 0.50 0.75 1.00 1.25 1.50 [4] S. M. Hoek, U. Riechert, T. Strehl, S. Tenbohlen and K. Feser,
frequency (GHz) “A New Procedure for Partial Discharge Localization in Gas-
insulated Switchgear in Frequency Domain”, 14th Intern.
Figure 16. UHF PD spectrum of an on-line recorded UHF PD signal on a Sympos. High Voltage Engineering, Paper No. G-05, Beijing,
200 MVA single-phase transformer. China, 2005.
1552 S. Tenbohlen et al.: Partial Discharge Measurement in the Ultra High Frequency (UHF) Range
[5] R. Kurrer and K. Feser, “The Application of Ultra-High- Stefan Tenbohlen (M’07) received the Diploma
Frequency Partial Discharge Measurements to Gas-Insulated and Dr.-Ing. degrees from the Technical
Substations”, IEEE Trans. Power Delivery, Vol. 13, pp. 777- University of Aachen, Germany, in 1992 and
782, 1998. 1997, respectively. In 1997 he joined AREVA
[6] J. J. Smit, E. Gulski and F. Wester, “Economical and technical Schorch Transformatoren GmbH, Mönchen-
aspects of advanced PD diagnostics to support condition based gladbach, Germany, where he was responsible for
maintenance of HV assets“, Transmission and Distribution basic research and product development and in
Conference and Exhibition: Asia Pacific. IEEE/PES, Vol. 2, this function working in the field of on-line
pp.1110 – 1115, 2002. monitoring of power transformers. From 2002 to
[7] C.-H. Lee, Y.-C. Lin, M.-Y. Chiu, C.-H. Huang, S.-S. Yen and 2004 he was the head of the electrical and
C. Haeng, “The study on diagnostics for aging trend of cable mechanical design department. 2004 he was
termination”, Intern. Conf. Condition Monitoring and appointed to a professorship and head of the
Diagnosis, CMD, Paper No. C2-17, Beijing, China, 2008. institute of Power Transmission and High Voltage Technology of the
[8] K. Uchida et al. “Study on detection for the defects of XLPE Universität Stuttgart, Germany. In this position his main research fields are
cable lines”- IEEE Trans. on Power Delivery, Vol. 11, April diagnostic of equipment of electrical networks, development of high
1996, pp. 663-668. voltage measurement technique and different aspects of electromagnetic
[9] E. Lemke, H. Elze and W. Weissenberg “Experience in PD compatibility (EMC). Prof. Tenbohlen holds several patents and published
diagnosis tests of HV cable terminations in service using the more than 100 papers. He is member of the IEEE German committees of
ultra-wide band PD probing”, 13th Intern. Sympos. High CIGRE A2 (Power Transformers), D1 (Emerging Technologies), C4
Voltage Engineering, Paper No. 404, Netherlands, 2003 (System Technical Performance), several international working groups.
[10] T. Klein, D. Denissov, W. Köhler, S. Tenbohlen “Application Since 2007 he is member of the board of the German Power Engineering
of UHF method for on-line PD diagnostics of cable Society VDE-ETG.
terminations”, CIGRE 2008, Paris, France, Paper No. B1-203,
2008, Stefan M. Hoek was born in Stuttgart-Bad
[11] S. M. Markalous, S. Tenbohlen and K. Feser, “Improvement of Cannstatt, Germany in 1976. He received the
acoustic detection and localization accuracy by sensitive Dipl.-Ing. degree in electrical engineering from
electromagnetic PD measurements under oil in the UHF range”, the Universität Stuttgart, Germany, in 2002. His
14th Intern. Sympos. High Voltage Engineering, Paper No.G- experience includes the field of electrical energy
039, Beijing, China, 2005. systems and measuring systems. As a lecturer, he
[12] S. M. Markalous, Detection and Location of Partial Discharges worked during his research, at the faculty of
in Power Transformers using Acoustic and Electromagnetic electrical engineering. At several international
Signal”, Ph.D. thesis, Universität Stuttgart, Germany, 2006. and national symposia he presented his work to
[13] Q. Chen, X. Gong, W. Gao, K. Tan and M. R. Raghuveer, the public.
“Experimental study on UHF measurement for Partial
Discharge in oil-impregnated insulation”, 13th Intern. Sympos. Denis Denissov received the B.Sc. and M.Sc.
High Voltage Engineering, Delft, Netherlands, ISBN 90-77017- degrees in electrical power engineering from
79-8, Millpress, Rotterdam, Paper No. 159, 2003. Novosibirsk State Technical University, Russia in
[14] T. Babnik, R. K. Aggarwal, P. J. Moore and Z. D. Wang, 2002 and from the Technical University of
“Radio Frequency Measurements of Different Discharges”, Aachen, Germany in 2004, respectively. At
IEEE Bologna Power Tech Conf., Vol. 3, Bologna, Italy, 2003. present, he is a Ph.D. student in the Institute of
[15] W. R. Rutgers and Y. H. Fu, “UHF PD-Detection in a Power Power Transmission and High Voltage
Transformer”, 10th Intern. Sympos. High Voltage Engineering, Technology at the Universität Stuttgart, Germany.
Vol. 4, pp. 219-222, Québec, Canada, 1997. His main research topic is in the field of the ultra
[16] M. D. Judd, B. M. Pryor, S. C. Kelly, B. F. Hampton, high frequency partial discharge detection in high
”Transformer Monitoring Using the UHF Technique”, High voltage cable accessories.
Voltage Engineering Symposium, London, United Kingdom,
Conf. Pub. No. 467, 1999. Sacha M. Markalous (M’07) received the Dipl.-
[17] R. Kuppuswamy, F. Devaux and S. Lelaidier, "Recognition of Ing. degree in electrical engineering from the
Discharge Sources Using UHF PD Signatures”, IEEE Electr. University of Stuttgart, Germany, in 2002. In
Insul. Mag. Vol. 18, No. 5, pp. 8-14, 2002. 2006 he received the doctorate degree at the
[18] F. Marangoni, J. P. Reynders and P. J. de Klerk, ”Investigation Institute of Power Transmission and High Voltage
into the Effects of Different Antenna Dimensions for UHF Technology, Universität Stuttgart in the field of
Detection of Partial Discharges in Power Transformers”, IEEE Ultra-High Frequency Partial Discharge
Bologna Power Tech Conference, Vol. 3, Bologna, Italy, 2003. measurement and acoustic location in power
[19] S. Coenen, S. Tenbohlen, S. M. Markalous and T. Strehl, transformers. In 2006 he joined LDIC GmbH,
“Sensitivity of UHF PD Measurements in Power Dresden, Germany as technical director where he
Transformers”, IEEE Trans. Dielectr. Electr. Insul., Vol. 15, is coordinating the departments of product
pp. , 2008. (in this issue), research & development, product management, service activities and
production. Here his work comprises diagnostics, monitoring and high-
voltage measurement techniques of power cables/accessories, transformers,
gas-insulated systems and overhead-lines. Dr.-Ing. Markalous published
more than 40 papers and is member of the German Power Engineering
Society VDE-ETG.