sensors

Article
Novel Simulation Technique of Electromagnetic Wave
Propagation in the Ultra High Frequency Range
within Power Transformers
Takahiro Umemoto 1, * and Stefan Tenbohlen 2
1 Advanced Technology R&D Center, Mitsubishi Electric Corporation, Amagasaki 6618661, Japan
2 Institute of Power Transmission and High Voltage Technology, University of Stuttgart, 70569 Stuttgart,
Germany; stefan.tenbohlen@ieh.uni-stuttgart.de
* Correspondence: Umemoto.Takahiro@df.MitsubishiElectric.co.jp




Received: 6 November 2018; Accepted: 27 November 2018; Published: 3 December 2018


Abstract: Diagnoses of power transformers by partial discharge (PD) measurement are effective
to prevent dielectric failures of the apparatus. Ultra-high frequency (UHF) method has recently
received attention due to its various advantages, such as the robustness against external noise and the
capability of PD localization. However, electromagnetic (EM) waves radiated from PD tend to suffer
attenuation before arriving at UHF sensors, because active part of the transformer disturbs the EM
wave propagation. In some cases, that results in poor detection sensitivity. To understand propagation
and attenuation characteristics of EM waves and to evaluate the detection sensitivity quantitatively,
a computational approach to simulate the EM wave propagation is important. Although many
previous researches have dealt with EM wave simulation for transformers, validations of those
simulations by comparing with the experimental ones have seldom been reported. In this paper,
cumulative energies, signal amplitudes and propagation times of EM waves were measured using a
630 kVA transformer. EM wave propagation was computed using the time-domain finite integration
technique and the results were compared with the experimentally obtained ones. These simulation
results showed good agreement with the experimental ones. The results can serve as guidelines to
improve the efficiency of UHF PD detection and offer the possibility to achieve optimal placement of
UHF sensors in power transformers.

Keywords: power transformers; partial discharges; electromagnetic wave simulation; UHF PD

measurement; UHF antennas

1. Introduction
Power transformers are key components in power systems and their dielectric failures severely
influence the system operation [1–3]. Continuous activity of partial discharge (PD), which might occur
within the transformers due to undesirable local electric field enhancement, is one of the main causes
of transformer failures, hence diagnoses based on PD measurement is a promising method to assess
the condition of the apparatus [4].
Although various PD measurement techniques have been proposed and developed over a long
period [5–8], the ultra-high frequency (UHF) method, that is, detecting electromagnetic (EM) waves
in the UHF range (300 MHz–3 GHz) radiated due to a short rise time of the PD current pulse, has
recently received much attention [9,10]. Attractive advantages are, for example, the robustness against
external noise [11,12] and the capability of PD localization by using time-difference of arrival (TDOA)
between multiple UHF sensors [13,14]. Due to these advantages, the UHF method is suitable for
factory acceptance tests (FAT) and site acceptance tests (SAT), as well as on-line diagnoses [15].

Sensors 2018, 18, 4236; doi:10.3390/s18124236 www.mdpi.com/journal/sensors

Sensors 2018, 18, 4236 2 of 16

However, the EM waves radiated from PD tend to suffer severe attenuation within the transformer
before arriving at UHF sensors. In some cases, this results in low detection sensitivity of the PD signals,
especially when the active part of the transformer (e.g., windings, core and leads) disturb the EM
wave propagation [16,17]. Furthermore, localization based on TDOAs also leads to large errors due
to the roundabout propagation path of the EM waves. In order to understand the propagation and
attenuation characteristics of EM waves within transformers and to evaluate PD detection sensitivity
as well as the propagation time quantitatively, a computational approach to simulate the EM wave
propagation is essential.
Simulation of the EM wave propagation in gas insulated switchgears (GIS) has been studied for
more than 15 years and their results were compared with the theoretical or experimental ones for
validation [18–20]. On the other hand, simulation for power transformers has also been investigated
by many researchers [13,15,21–23]. In Reference [21], influences of transformer windings and
insulation papers on amplitudes of the EM waves propagating through them were discussed based
on only numerical computation. In Reference [22], the propagation times of PD induced EM signals
within power transformer were computed in order to improve the accuracy of PD localization. In
Reference [23], the signal amplitudes of EM waves were computed as a function of UHF sensor
positions based on the simulation using an actual transformer model. However, the validity of the
EM wave simulation was not discussed, hence the appropriate computational conditions were still
unclarified. Considering the above fact, validations of simulations of EM wave propagation and those
simulated results by comparing with the experimental ones using actual transformer structures have
been seldom reported, therefore the validations are insufficient.
The objective of this paper is to propose the simulation of EM wave propagation, including
the detailed simulation conditions, which are validated by the experimental results using actual
transformers. First, validations of antenna modeling methods, an exciting signal as well as a model
of a transformer tank were evaluated by measurement with an empty transformer tank (i.e., without
active parts of a transformer). Second, cumulative energies of the EM waves, their signal amplitudes
and propagation times to each UHF sensors were investigated by simulation and measurement
using a distribution transformer for validating the transformer modeling. For both investigations,
the simulated results showed good agreement with the measured ones. Thus, the authors successfully
validated this novel simulation technique.
The rest of this paper is organized as follows: Section 2 presents the experimental setup and
measurement system of UHF signals, including a transformer structure. Detailed EM simulation
methods and 3-D modeling technique are described in Section 3. In Section 4, both simulated and
measured results are compared and the validity of the simulation is discussed, while conclusions and
future work suggestions are presented in Section 5.

2. Experimental Method

2.1. UHF Sensors and EM Wave Source

Figure 1 illustrates a schematic diagram of a steel tank of 1350 kVA transformer and positions of
four UHF drain valve sensors [11,12] and a monopole antenna in the first experiment. Inside dimension
of the transformer tank was 1720 mm in length, 760 mm in width and 1550 mm in height, respectively.
There was a hole with 100 mm in diameter on the top of the tank, through which a monopole antenna
was inserted and used as an EM wave source. Note that in this experiment, the transformer tank was
not filled with the insulating oil.
On the wall of the tank, there are two DN50 and two DN80 gate valves. Four UHF drain valve
sensors, named A, B, C and D, were mounted with each gate valve, as shown in Figure 2. Figure 3
shows an image of the UHF sensor [4]. A probe (top of the UHF sensor) has a truncated cone shape.
The detailed dimension of the probe will be described later in Section 3.2. The antenna factor (AF),
which indicates sensitivity of the sensor, was described in Reference [12]. The probes of the sensors
Sensors 2018, 18,
18, x4236
FOR PEER REVIEW 3 of3 of
16 16

from the monopole antenna and detected by these UHF sensors were digitized and recorded by an
(top portion(LeCroy,
oscilloscope of the sensors)
WaveProwere 7300,inserted
Chestnut100 mm NY,
Ridge, intoUSA,
the tank,
3 GHzwhich results20inGS/s)
bandwidth, a high sensor
without
sensitivity and is suitable
any analog filter and amplifier.for the comparisons with the simulated results. EM wave signals, radiated
fromAthe monopole
monopole antenna
antenna of 20andmmdetected
in lengthby
andthese UHFinsensors
1.3 mm diameter were
wasdigitized
used as anandEMrecorded by an
wave source,
oscilloscope
instead (LeCroy,
of a typical PDWavePro 7300,
source (e.g., Chestnut Ridge,
a needle-plane NY, USA,
electrode system)3 GHz
sincebandwidth, 20 GS/s)
it radiates stable without
EM waves
any analogamplitudes
regarding filter and amplifier.
and frequency spectra. The antenna was excited by a voltage pulse generator
(DobleA Engineering,
monopole antenna of 20 mmLDC-7/UHF,
UHF calibrator in length andWatertown,
1.3 mm in diameter
MA, USA) was used as
through an Ω
a 50 EM wave cable
coaxial source,
instead
of of a typical
approximately PDmm
2000 source (e.g.,The
length. a needle-plane electrode
output voltage system)
of the pulse since itwas
generator radiates
set to stable
60 V. EM waves
In theamplitudes
regarding first experiment, time-domain
and frequency signals
spectra. The and cumulative
antenna energies
was excited by a of the EM
voltage waveforms
pulse generator
detected
(Doble by the four UHF
Engineering, UHF sensors
calibratorwere evaluated toWatertown,
LDC-7/UHF, validate theMA,
modeling of antenna,
USA) through Ω coaxial
a 50exciting signal
cable
and the transformer
of approximately tank
2000 mmin length.
the EM The wave simulation.
output voltage of the pulse generator was set to 60 V.

1720

Sensor C Sensor B
DN80 gate valve

100
160 140
1270

760
Monopole antenna
(EM wave source)
115 115

DN50 gate valve

Sensor D Sensor A

(a) Top view

1720

Sensor A 130

Coaxial cable 200 115

1000 140
Sensor C
1550 Sensor B

160

115 200 Monopole antenna

(EM wave source)
90 Sensor D

(b) Side view

Figure 1.
Figure Schematicdiagram
1. Schematic diagramofofaatransformer
transformertank
tankand
andantenna
antennapositions
positions(for
(for the
the first
first experiment).
experiment).
Sensors 2018, 18, 4236 4 of 16

In the first experiment, time-domain signals and cumulative energies of the EM waveforms
detected by the four UHF sensors were evaluated to validate the modeling of antenna, exciting signal
and the
Sensors
Sensors transformer
2018,
2018, 18, xx FOR
18, tank
FOR PEER
PEER in the EM wave simulation.
REVIEW
REVIEW 44 of
of 16
16

DN80 gate
DN80 gate valve
valve

UHF sensor
UHF sensor

Figure 2.
Figure 2. UHF
UHF drain
drain valve
valve sensor
sensor mounted
mounted with
with aa DN80
DN80 gate
gate valve.
valve.

Figure 3.
Figure 3. An
An image
image of
of the
the UHF
UHF drain
drain valve
valve sensor
sensor [4].
[4].

2.2. Active
2.2. Active
2.2. Part
Active Part of
Part of the
of the Transformer
the Transformer
Transformer
The
The active
The active partof
active part
part ofaaathree-phase
of three-phase630
three-phase 630kVA
630 kVAdistribution
kVA distributiontransformer,
distribution transformer,which
transformer, which
which isis
is mainly
mainly
mainly composed
composed
composed of
of
of high
high and
high and
and low low
low voltagevoltage
voltage windings, windings,
windings, an an ironan iron
iron core core
core and and
and leads, leads,
leads, was was
was utilized utilized
utilized forfor thefor the
the second second
second experiment. experiment.
experiment. Note Note
Note
that thattransformer
that the
the the transformer
transformer tank tank
tank usedused
used in the
in theinexperiment
the experiment
experiment is larger
is larger
is larger thanthan
than normally
normally
normally used
used usedfor for
for thisthis
this active
active
active part,
part,
part, in
in
in order
order to
order to
to allowallow
allow the the
the UHF UHF
UHF sensors sensors
sensors to to
to be be inserted
be inserted
inserted deeply deeply
deeply into into
into the the tank,
the tank, resulting
tank, resulting
resulting in in improved
in improved sensitivities.
improved sensitivities.
sensitivities.
There
There are
There are four
are four high
four high voltage
high voltage (HV)
voltage (HV) and
(HV) and two
and two low
two low voltage
low voltage (LV)
voltage (LV) cylindrical
(LV) cylindrical windings
cylindrical windings in
windings in one
in one phase
one phase
phase
and
and each
each HV
HV and
and LV
LV winding
winding consists
consists of
of 156
156 and
and 57
57 layers,
layers,
and each HV and LV winding consists of 156 and 57 layers, respectively. However, the HV winding respectively.
respectively. However,
However, the
the HV
HV winding
winding
of
of one
of one phase
one phase
phase was was removed.
was removed.The
removed. Theinnermost
The innermostdiameter
innermost diameterof
diameter ofof the
the
the LV
LV LV windings,
windings,
windings, the
the
the outermost
outermost
outermost diameter
diameter
diameter of
of
of
thethe
the HV
HV HV windings
windings
windings and
and and their
their
their height
height
height areapproximately
are
are approximately207
approximately 207207mm,mm,350
mm, 350mm
350 mmand
mm and780
and 780mm,
780 mm,respectively.
mm, respectively.
respectively.
Three
Three identical
Three identical monopole antennas
identical monopole
monopole antennasof
antennas of20
of 20mm
20 mmin
mm ininlength,
length,described
length, describedin
described inin Section
Section
Section 2.1,
2.1,
2.1, were
werewere also
also also used
used
used as
as
as EM
EM wave
EM wave
wave sources sources
sources in in
in this this second
this second experiment.
second experiment.
experiment. These These
These antennas antennas
antennas were were
were set set around
set around
around the the windings
windings at
the windings at
at
different
different heights
heights and
and in
in different
different positions
positions before
before the
the active
active parts
parts
different heights and in different positions before the active parts were installed into the transformer were
were installed
installed into
into the
the transformer
transformer
tank. The
tank. The
tank. detailed
The detailed structure
detailed structure
structure of of this
of this transformer
this transformer
transformer and and
and the the positions
the positions
positions of of three
of three monopole
three monopole antennas
monopole antennasantennas will will be
will be
be
illustrated
illustrated in
illustrated in Section
in Section 3.1,
Section 3.1, as
3.1, as aaa 3-D
as 3-D computational
3-D computational model.
computational model.
model.
In
In this
In thissecond
this secondexperiment,
second experiment,propagation
experiment, propagationtimes
propagation times
times ofof the
of EM
the
the EM
EM waves
waves
waves from each
from
from monopole
each
each monopole
monopole antenna to the
antenna
antenna to
to
sensors
the sensors
the were
sensors were also measured
were also
also measured in
measured in addition to
in addition signal
addition to amplitudes
to signal
signal amplitudes and their
amplitudes and cumulative
and their energies.
their cumulative Figure
cumulative energies.
energies. 4
illustrates
Figure the measurement
Figure 44 illustrates
illustrates the setup for propagation
the measurement
measurement setup for
setup times of thetimes
for propagation
propagation EM waves,
times thein
of the
of EM
EMwhich the output
waves,
waves, in which
in which of the
the
the
pulse generator
output of
output of the and
the pulse the EM
pulse generator wave
generator and signals
and the the EMwere
EM wavesimultaneously
wave signals signals were measured.
were simultaneouslyThe propagation
simultaneously measured. times
measured. The Theof
the EM waves
propagation times
propagation were
times of calculated
of thethe EM as
EM waves the
waves were time difference
were calculated
calculated as of arrival
as the
the time between
time differenceboth
difference of signals,
of arrival considering
arrival between
between both the
both
signal
signals, propagation
considering time
the within
signal the coaxial
propagation cables.
time
signals, considering the signal propagation time within the coaxial cables. within the coaxial cables.
Sensors 2018, 18, 4236 5 of 16
Sensors 2018, 18, x FOR PEER REVIEW 5 of 16

Coaxial cable
Active parts of the
Pulse generator
transformer
(Doble, LDC-7/UHF)

Oscilloscope
(LeCroy, WavePro 7300)

Monopole antenna Coaxial cable

(EM wave source) The EM wave UHF drain valve sensor
propagation path
(it propagates around a center winding in this figure)

Figure 4. Measurement
Figure 4. Measurement setup
setup for
for propagation
propagationtimes
timesof
ofthe
theEM
EMwaves.
waves.
2.3. Denoising of the Cumulative Energies
2.3. Denoising of the Cumulative Energies
Cumulative energies of the EM wave signals are commonly used not only to evaluate the
Cumulative energies of the EM wave signals are commonly used not only to evaluate the PD
PD signal strength quantitatively but also to determine the arrival times of the PD signals for
signal strength quantitatively but also to determine the arrival times of the PD signals for the
the localization [21,24]. Cumulative energy E(t) of a discrete voltage waveform can normally be
localization [21,24]. Cumulative energy E(t) of a discrete voltage waveform can normally be
calculated as,
calculated as, t
∆t
V 2 (i · ∆t)
E(t) = ∑ t · ∆t (1)
Z
i =0t
V (i  t )
2

where V(i∆t) is a voltage of the EM signal E (at 

t ) t== i∆t. Z and ∆t 
i =0 Z
t an impedance of the measurement
are
(1)

circuit (usually 50 Ω) and a sampling period, respectively [21,25]. However, in this experiment, some
measured
where V(iΔt)signals showedoflow
is a voltage theSNR (signalattot =noise
EM signal iΔt. Zratio)
and Δtduearetoan
theimpedance
severe attenuation of the EM
of the measurement
waves
circuitby the deflation
(usually andareflection,
50 Ω) and resulting
sampling period, in calculation
respectively errors
[21,25]. of the cumulative
However, energies. some
in this experiment,
In thissignals
measured research, in order
showed lowtoSNR
evaluate thetocumulative
(signal noise ratio)energies
due to accurately
the severe even for the of
attenuation lowtheSNR
EM
waveforms,
waves by the background
deflation andnoise components
reflection, on the
resulting incumulative
calculation energies
errors of were deleted as,energies.
the cumulative
In this research, in order to evaluate the cumulative energies accurately even for the low SNR
t
waveforms, background noise components V (i · ∆t)
∆t on 2the cumulative energies were deleted as,
E(t) = ∑ · ∆t − A · t (2)
i =0 t
Z
t
V (i  t )
2
where A is a compensation factor of E = 
(t )background
the noise, which
t − A  tcan be obtained as a slope of the
Z at the sensor. Figure 5 shows an example of the
i = 0arrives
(2)
cumulative energy before the EM wave signal
measured
where A isEM waveform, its cumulative
a compensation factor of theenergy calculated
background noise,by which
(1) andcan
the be
denoised cumulative
obtained as a slopeenergy
of the
calculated by (2), respectively. Without this denoising procedure, the cumulative
cumulative energy before the EM wave signal arrives at the sensor. Figure 5 shows an example of the energy continues
to increase EM
measured evenwaveform,
before theits
EM wave signal
cumulative arrives
energy and after
calculated byit(1)
sufficiently attenuates,
and the denoised which leads
cumulative energyto
large errors in the total energy (i.e., the convergence energy). In a case of Figure
calculated by (2), respectively. Without this denoising procedure, the cumulative energy continues 4, the cumulative
energies at 100
to increase evennsbefore
with and
thewithout
EM wave thesignal
denoising
arrives areand
10.6after
fJ and 16.3 fJ, respectively.
it sufficiently Thewhich
attenuates, relative error
leads to
islarge
approximately 53.8%.
errors in the total energy (i.e., the convergence energy). In a case of Figure 4, the cumulative
energies at 100 ns with and without the denoising are 10.6 fJ and 16.3 fJ, respectively. The relative
error is approximately 53.8%.
Sensors 2018, 18, 4236 6 of 16
Sensors 2018, 18, x FOR PEER REVIEW 6 of 16

0.02 18
16

Cumulative energy (fJ)

Signal amplitude (V)
0.01 14
12
10
0
8
6
-0.01 EM wave signal
4
Cumulative energy before denoising
2
Denoised cumulative energy
-0.02 0
-40 -20 0 20 40 60 80 100 120 140 160

Time (ns)
Figure 5. Measured EM waveform and its cumulative energy before and after denoising procedure.
Figure 5. Measured EM waveform and its cumulative energy before and after denoising procedure.
3. Simulation Technique of the EM Wave Propagation
3. Simulation Technique of the EM Wave Propagation
3.1. 3-D Modeling of the Transformer
3.1. 3-D Modeling of the Transformer
The EM wave propagation within the transformer was simulated by using the CST Microwave
The EM wave
Studio software withpropagation within
transient solver. thecalculation
The transformer in was simulated
the software by using
is based the finite
on the CST Microwave
integration
Studio
theory software with transient
(FIT), in which the Maxell’ssolver. The calculation
equations in the software
are numerically solved, isnotbased on the finite
in differential integration
forms used in
theory (FIT), in which
the finite-difference the Maxell’s
time-domain equations
(FDTD) methodare numerically
[26] but integral solved,
forms not[23,27].
in differential forms used
in theFigure
finite-difference
6 shows a 3-D time-domain
computational(FDTD) method
model, which [26] but integral
simulates forms [23,27].
the transformer tank and the active
Figure
part of 6 shows
the 630 a 3-D computational
kVA transformer, which was model,
used inwhich simulates the
the experiment as transformer
illustrated intank and1.the
Figure active
Basically,
part of the 630 kVA transformer, which was used in the experiment as
the structure of the active part and its size in this model are the same as the actual ones. However, HVillustrated in Figure 1.
Basically, the structure
and LV leads, which connect of theboth
active part and
windings its size inon
to bushings thisthemodel
top ofare
thethe
tank,samewereasnot
themodeled
actual ones.
due
However, HV and LV
to their complicated leads, which
structures. connect both
Furthermore, eachwindings
winding was to bushings
modeledonasthe top of the cylinder
a conductive tank, were to
not modeled due to their complicated structures. Furthermore, each winding
make the model simple and reduce the computational time drastically. This simplification is possible, was modeled as a
conductive
because thecylinder
windings to ofmake
this the model simple
transformer used and reduce
in this studythe arecomputational
the cylindricaltime type,drastically.
hence there Thisis
simplification is possible, because the windings of this transformer
no oil gap between each layer. In fact, there are quite small gaps between conductors because of used in this study are the
cylindrical type,and
layer-insulation hence
the EM there is nocan
waves oiltheoretically
gap between each layer.
propagate In fact,
through themthere are quite
to some extent.small gaps
However,
between conductors
these propagating because
paths can beofignored
the layer-insulation
in this research and the EM
because thewaves can theoretically
EM waves propagate
attenuate severely and
through them to some extent. However, these propagating paths can be
cannot be detected experimentally. The validations of this transformer modeling will be discussed in ignored in this research
because the EM waves attenuate severely and cannot be detected experimentally. The validations of
Section 4.2.
this transformer
The positions modeling
of the three will monopole
be discussed in Section
antennas (EM4.2. wave sources) are also indicated in Figure 6.
Two of them, named positions 1 and 2, are located at the same sources)
The positions of the three monopole antennas (EM wave side around are the
alsocenter
indicated in Figure
windings 6.
but at
Two of them,
different named
heights. positions
Another 1 and 2,position
EM source, are located
3, is at the other
same side of around the center windings but
the windings.
at different heights. Another EM source, position 3, is at the other side of the windings.
 Sensors 2018, 18, 4236 7 of 16

Sensors 2018, 18, x FOR PEER REVIEW 7 of 16

Sensor B

Sensor A

Core

Metal poles
1550

Sensor C Windings

Sensor D
760 1720

(a) Perspective view (a part of the tank walls is set to be transparent)

Position 3
350 245

Position 2 Position 1

(b) Top view of the active parts

950

Position 2

780
620
385 Position 3
355 Position 1

(c) Side view of the active parts

3-D
Figure6.6.3-D
Figure computational
computational model
model of of
thethe transformer
transformer andand positions
positions of the
of the three
three monopole
monopole antennas.
antennas.
Sensors 2018, 18, 4236 8 of 16
Sensors 2018, 18, x FOR PEER REVIEW 8 of 16

3.2.
3.2.Antenna
AntennaModeling
Modeling
AAfeeding
feedingmethod
methodofofan anantenna
antennaforforthe
thesimulation
simulationhas hasnot
notbeen
beenestablished
establishedyet,
yet,although
althoughseveral
several
methods
methods have been proposed [28–32]. In this paper, the gap feeding method [33,34]was
have been proposed [28–32]. In this paper, the gap feeding method [33,34] wasapplied
appliedtoto
simulate
simulatebothboththethemonopole
monopoleantenna
antennafor forradiating
radiatingthetheEM EMwaves
wavesand andthe
theUHF
UHFsensors
sensorsforforreceiving
receiving
them, due to its simplicity. In this feeding method, a coaxial cable to feed the
them, due to its simplicity. In this feeding method, a coaxial cable to feed the antenna was notantenna was not modeled,
while the feeding
modeled, while port (e.g., voltage
the feeding source)
port (e.g., was introduced
voltage source) was at the gap between
introduced a probe
at the (e.g., monopole)
gap between a probe
and grounded
(e.g., monopole) conductor.
and grounded conductor.
Figure
Figure 77 illustrates models of
illustrates models of the
themonopole
monopoleand andUHFUHF antennas
antennas with
with the the
gapgap feeding.
feeding. The
The
monopole in the simulation model was 20 mm in length and 1.3 mm in diameter with aaplate
monopole in the simulation model was 20 mm in length and 1.3 mm in diameter with plate
conductor
conductorhaving
havingan anarea mm××10
areaofof1010mm 10mm.
mm. The
Theprobe
probeof ofthe
theUHF
UHFsensor
sensorwaswasaacircular
circulartruncated
truncated
cone
cone shape with a bottom diameter of 30 mm, a top diameter of 10 mm and a heightofof30
shape with a bottom diameter of 30 mm, a top diameter of 10 mm and a height 30mmmmand anditit
had
hadaaconductor
conductorwith witha adiameter
diameter ofof
3030mm.mm. TheThegap lengths
gap lengthsof the feeding
of the portport
feeding were set to
were set0.5
tomm for
0.5 mm
both the monopole and UHF sensors and their impedance was set to 50 Ω.
for both the monopole and UHF sensors and their impedance was set to 50 Ω. An exciting voltage An exciting voltage signal
was applied
signal to the feeding
was applied port of the
to the feeding portmonopole antennaantenna
of the monopole and the and
received voltage voltage
the received waveform across
waveform
the portthe
across of the
portUHF
of thesensors were analyzed.
UHF sensors were analyzed.

10 10 30 Feeding port
(0.5 mm gap)
Feeding port
(0.5 mm gap) 10

20
30

1.3

30
(a) Monopole antenna (b) UHF sensor

Figure7.7.Modeling
Figure Modelingofofthe
themonopole
monopoleand
andUHF
UHFsensors.
sensors.

3.3.
3.3.Other
OtherComputational
ComputationalConditions
Conditions
Table
Table11presents
presentsconductivity,
conductivity,relative
relativepermittivity
permittivityand
andpermeability
permeabilityofofthe
thematerials
materialsused
usedininthe
the
EM wave simulation. On the surface of the copper of the transformer windings, oil-impregnated
EM wave simulation. On the surface of the copper of the transformer windings, oil-impregnated paper
0.3 mm0.3
paper thick
mm was set was
thick as a set
coating materialmaterial
as a coating to represent the layerthe
to represent insulation of the windings.
layer insulation of the windings.
Table 1. Material properties used in the simulation.
Table 1. Material properties used in the simulation.
Material
Material Conductivity (S/m) Relative Permittivity Relative Permeability
(Model Component) Conductivity (S/m) Relative Permittivity Relative Permeability
(Model Component)
Air 0 1 1
Air
Copper
0 1 1
Copper 6.0 × 107 ∞ 1
(winding model) 6.0 × 107 ∞ 1
(winding model)
Paper
1.0 × 10−14 3.9 1
(layer-insulation)
Paper
Silicon steel 1.0 × 10−145 3.9 1
(layer-insulation) 1.0 × 10 ∞ 6000
(core)
Silicon steel
Steel 1.0
5.0××10
5
105 ∞∞ 6000
500
(core)
(tank)
Steel
5.0 × 105 ∞ 500
(tank)
When a voltage pulse is applied to an antenna, the voltage waveform at the antenna terminal
is determined both by the frequency-dependent input impedance of the antenna and characteristic
Whenof
impedance a voltage pulse
the coaxial is applied
cable to an antenna,
[35]. Generally, it is notthe voltage
easy waveform
to determine theatactual
the antenna
excitingterminal
signal ofis
determined
the both
antenna. In thisby the frequency-dependent
research, the exciting voltageinput impedance
waveform of the antenna
was determined basedand characteristic
on the reflected
impedance of the coaxial cable [35]. Generally, it is not easy to determine the actual exciting signal of
the antenna. In this research, the exciting voltage waveform was determined based on the reflected
 voltage waveform from the open-ended top of the antenna, as proposed in Reference [25]. The
amplitude and rise time (10–90%) of the exciting signal applied for the simulation were set to 60 V
and 0.8 ns, respectively.
The hexahedral mesh was used in this computation. The frequency range and cell numbers per
wavelength were set to 0–1500 MHz and 30, respectively. This results in approximately 76,000,000
Sensors 2018, 18, 4236 9 of 16
total mesh cells. The time durations of the simulation were set to 800 ns for the first experiment
without the active parts and 100 ns for the second experiment. At these times, the EM waves,
propagating within the tank, attenuated sufficiently to evaluate the convergence cumulative energies.
voltage waveform from the open-ended top of the antenna, as proposed in Reference [25]. The
amplitude and rise time
4. Evaluations of the(10–90%)
SimulatedofResults
the exciting signal applied for the simulation were set to 60 V and
and Discussions
0.8 ns, respectively.
The4.1. Validations ofmesh
hexahedral the Antenna and Transformer
was used Tank Modeling The frequency range and cell numbers per
in this computation.
wavelength Forwere thesetfirst
to 0–1500
step, theMHz
EM and
wave30, respectively.
propagation This
in the results
tank in approximately
without the active parts76,000,000
of the total
transformer was simulated and the results were compared with the measured ones in order
mesh cells. The time durations of the simulation were set to 800 ns for the first experiment without the to
remove the influence of the active parts and validate the modeling technique of the monopole and
active parts and 100 ns for the second experiment. At these times, the EM waves, propagating within
UHF sensors as well as the transformer tank.
the tank, attenuated sufficiently to evaluate the convergence cumulative energies.
Figures 8 shows an example of the simulated and measured EM waveforms by sensor A. In this
figure, the signal amplitudes of both the simulated and measured values were normalized by the
4. Evaluations
maximumofsignal
the Simulated Results
strength of each and Figure
waveform. Discussions
8a,b show the entire waveforms up to 800 ns and
the enlargements of the first 50 ns, respectively. From these figures, it can be seen that the attenuation
4.1. Validations
degrees of
of the
the Antenna and Transformer
signal amplitudes and the Tank ModelingEM waveforms by the simulation and
time-domain
Formeasurement
the first step, showed quite good agreement with each other, especially for the first 10 ns in Figure
the EM wave propagation in the tank without the active parts of the transformer
8b. Such a good agreement of the simulated waveform with the measured one has never been
was simulated and
reported before.
the results were compared with the measured ones in order to remove the influence
of the activeFigure
parts9and validate
compares the modeling
the simulated technique
and measured of the monopole
cumulative energies for and UHF
the four UHFsensors
sensors.as well as
the transformer tank.
In this figure, the cumulative energies were normalized by the values by sensor A. Both the simulated
and measured
Figure 8 showsresults showed of
an example a similar trend that the
the simulated andsensors A and BEM
measured showed the lowestby
waveforms and highestA. In this
sensor
sensitivities, respectively, although the maximum error between the simulated
figure, the signal amplitudes of both the simulated and measured values were normalized and measured results by the
was about 21% for sensor B. Figure 10 shows cumulative energies as a function of time for the four
maximum signal strength of each waveform. Figure 8a,b show the entire waveforms up to 800 ns
UHF sensors. Although convergence values of the cumulative energies by the simulation and
and themeasurement
enlargements of the
showed somefirst 50 ns, especially
differences, respectively. From
for sensor B asthese figures,
expected it can
from Figure be degree
9, the seen that the
attenuation degrees
of increase of the
in the signal amplitudes
cumulative energies in theand the time-domain
simulation EM waveforms
showed reasonable agreementbywith
the the
simulation
measured ones.
and measurement showed quite good agreement with each other, especially for the first 10 ns in
Baseda on
Figure 8b. Such goodthe agreement
results of theoffirst
theexperiment,
simulatedpresented
waveform in Figures 8–10,
with the it can be said
measured one that
has the
never been
modeling techniques of the monopole and the UHF sensors, the exciting voltage waveform as well
reported before.
as the tank modeling, including the material properties, are reasonable.

1
Signal amplitude (arb. unit)

0.5

-0.5 Simulation
Measurement
-1
0 100 200 300 400 500 600 700 800
Time (ns)
Sensors 2018, 18, x FOR PEER REVIEW 10 of 16
(a) Simulated and measured entire waveforms up to 800 ns.
1
Signal amplitude (arb. unit)

Simulation
Measurement
0.5

-0.5

-1
0 10 20 30 40 50

Time (ns)
(b) Enlargement of the waveforms at the first 50 ns.

FigureFigure 8. Examples
8. Examples of the
of the simulatedand
simulated and measured
measuredEM
EMwaveforms by sensor
waveforms A.
by sensor A.

2.5
umulative energies (p. u.)

1.5

0.5 Measurement
Sensors 2018, 18, x FOR PEER REVIEW 10 of 16
Sensors 2018, 18, x FOR PEER REVIEW 10 of 16
1

unit)
1 Simulation

unit)
Sensors 2018, 18, 4236 Measurement
Simulation 10 of 16

(arb.
0.5 Measurement

(arb.
0.5

amplitude
Figure 9 compares 0the simulated and measured cumulative energies for the four UHF sensors.

amplitude
0
In this figure, the cumulative energies were normalized by the values by sensor A. Both the simulated
and measured results -0.5showed a similar trend that the sensors A and B showed the lowest and highest

Signal
sensitivities, respectively,
-0.5 although the maximum error between the simulated and measured results

Signal
was about 21% for sensor -1 B. Figure 10 shows cumulative energies as a function of time for the four UHF
sensors. Although convergence
-1 0 values10of the cumulative
20 energies
30 by the40simulation50and measurement
showed some differences, 0especially 10 for sensor B20asTime 30 from Figure
expected
(ns) 40 9, the degree
50 of increase in
the cumulative energies in the simulation showed reasonable Time (ns) agreement with the measured ones.
Based on the results of(b) theEnlargement of the presented
first experiment, waveformsinatFigures
the first8–10,
50 ns.it can be said that the
(b) Enlargement of the waveforms at the first 50 ns.
modeling techniques ofExamples
Figure 8. the monopole and the UHF
of the simulated sensors, the
and measured EMexciting
waveformsvoltage waveform
by sensor A. as well as
the tank modeling, including
Figure 8. Examplestheofmaterial properties,
the simulated are reasonable.
and measured EM waveforms by sensor A.
2.5
u.) u.)

2.5
(p. (p.

2
energies

2
energies

1.5
1.5
Cumulative

1
Cumulative

1
0.5 Measurement
0.5 Measurement
Simulation
0 Simulation
0 A B C D
A B sensorsC
UHF D
UHF sensors
Figure 9. Simulated and measured cumulative energies as a function of the UHF sensor positions.
Figure 9. Simulated and measured cumulative energies as a function of the UHF sensor positions.
Figure 9. Simulated and measured cumulative energies as a function of the UHF sensor positions.
Dotted: measurement, solid: simulation
2.5
u.)u.)

Dotted: measurement, solid: simulation

2.5 B
(p.(p.

2 B
energy

2 C
energy

1.5 C
1.5
Cumulative

1
Cumulative

1
0.5
0.5 D A
0 D A
0 0 100 200 300 400 500
0 100 200 Time (ns)
300 400 500
Time (ns)
Figure 10. Cumulative energies as a function of time for the four UHF sensors.
Figure 10. Cumulative energies as a function of time for the four UHF sensors.
4.2. Propagation Figure 10.the
Times of Cumulative
EM Wavesenergies
within as a function of time for the four UHF sensors.
a Transformer
4.2. Propagation Times of the EM Waves within a Transformer
As the nextTimes
4.2. Propagation step, of
the
theactive partswithin
EM Waves of theatransformer
Transformer were installed into the tank. In order to
validate the size and positions of the transformer windingwere
As the next step, the active parts of the transformer modelinstalled
in the into the tank.propagation
simulation, In order to
As the size
nextand
step, the active parts of the transformer wereininstalled into thepropagation
tank. In order to
times of the EM waves from the EM wave sources to each sensor were computed and compared times
validate positions of the transformer winding model the simulation, with
validate
of the EM the size and
waves frompositions
the EM of the transformer
wave sources to winding
each sensormodel
were incomputed
the simulation,
and propagation
compared times
with the
the experimental and theoretically calculated ones. In the experiment, the propagation times were
of the EM waves
experimental andfrom the EM wave
theoretically sourcesones.
calculated to each
In sensor were computed
the signal
experiment, and compared with the
obtained by simultaneous measurement of the exciting and thethe propagation
resultant EM wavetimes were
signals,
experimental
obtained and theoretically
by simultaneous calculated
measurement ones. In the signal
ofsystematically
the exciting experiment,
and the the propagation
resultant times were
where arrival times of the EM waves were calculated based on theEM wave
Energy signals,
criterion
obtained by
where arrival simultaneous measurement of the exciting signal and the resultant EM wave signals,
method [8,21]. times of the EM waves were systematically calculated based on the Energy criterion
where arrival
method [8,21]. times of the EM waves were systematically calculated based on the Energy criterion
Theoretical propagation times were calculated, assuming the transformer windings as simple
method [8,21].
cylindrical obstacles [36]. Figure 11 illustrates geometrical model of the propagation path around a
Sensors 2018, 18, 4236 11 of 16

cylinder and its 2-D projection. In a 2-D projection, we assume coordinates of the EM wave source and
UHF sensor as (xm , ym ) and (xs , ys ). Then, θ 1 , θ 2 , θ 3 and θ 4 illustrated in Figure 11 can be expressed as,

θ1 = tan−1 (ym /xm ) (3)

q
θ2 = tan−1 ( xm 2 + ym 2 − r2 /r ) (4)
q
θ3 = tan−1 ( xs 2 + ys 2 − r2 /r ) (5)

θ4 = tan−1 (ys /xs ) (6)

where r is a radius of the cylindrical obstacle. Considering the height difference in 3-D, the propagation
distance from the source to the sensor, LAD , is expressed as,

1/2
( xm 2 + ym 2 − r2 )

L AD = + r · ( π − θ1 − θ2 − θ3 − θ4 )
1/2
i o1/2 (7)
2 2
+( xs + ys − r )2 + ( z m − z s )2

where zm and zs denote z coordinates of the EM wave source and the sensor, respectively [36].
Theoretical propagation time considering the obstacles can be calculated by LAD /c, where c is the light
Sensors 2018,
speed 18,case.
in this x FOR PEER REVIEW 12 of 16

EM wave source D
(xm, ym, zm) UHF sensor
(xs, ys, zs)
x

A
y
Diameter: r

θ1
θ4 x
A θ2 θ3 D

EM wave source UHF sensor

(xm, ym) B C (xs, ys)

Figure 11. Propagation

Figure 11. Propagation path
path of
of the
the EM
EM waves
waves around
around aa cylindrical
cylindrical obstacle
obstacle and
and its
its 2-D
2-D projection.
projection.

Figure 12 shows propagation times of the EM waves, obtained by the simulation, the experiment
9.5
and theoretical calculation described above as functions of positionsSimulation
of the UHF sensors and the EM
8.5 Theoretical
wave sources. For comparison, direct propagation times were also plotted, which were calculated
Direct propagation
Propagation time (ns)

based on the Euclidian 7.5 distance between an EM source and a UHF sensor divided by the light speed.
Measurement
It can be seen that the6.5simulated and theoretically calculated propagation times showed quite good
agreement for all UHF sensors and the source positions, while in some cases, there are some differences
5.5
of about 1 ns between the simulated results and those calculated assuming direct propagation. This fact
4.5
indicates that the influences of the active part on the propagation times were accurately simulated and
thus the modeling of3.5 the transformer active part, mainly the positions and size of the windings, was
successfully validated.2.5
It should be noted that some measurement results (e.g., sensor C in EM source position 1 or sensor
1.5
A in EM source 3) showed large
A Berrors
C from
D AtheBotherC results.
D A These
B C wereD caused by the roundabout
propagation path and resulting severe attenuations of the EM waves, which made the arrival of the
EM source: 1 EM source: 2 EM source: 3
UHF sensors (A-D) and EM source positions (1-3)

Figure 12. Propagation times of the EM waves by the simulation, experiment and theoretical
calculations.
 Diameter: r

θ1
θ4 x
A θ2 θ3 D
Sensors 2018, 18, 4236 12 of 16
EM wave source UHF sensor
(xm, ym) B C (xs, ys)
EM signals unclear. These large errors in the determination of the arrival time will lead to a critical PD
localization
Figureerror, so the influence
11. Propagation path of on
the the
EM location accuracy
waves around will be evaluated
a cylindrical obstacle andinits
future works.
2-D projection.

9.5
Simulation
8.5 Theoretical
Direct propagation
Propagation time (ns)
7.5
Measurement
6.5
5.5
4.5
3.5
2.5
1.5
A B C D A B C D A B C D
EM source: 1 EM source: 2 EM source: 3
UHF sensors (A-D) and EM source positions (1-3)

Figure 12. Propagation times of the EM waves by the simulation, experiment and theoretical calculations.
Figure 12. Propagation times of the EM waves by the simulation, experiment and theoretical
calculations.Energies and Signal Amplitudes as a Function of Sensor Positions
4.3. Cumulative
Finally, cumulative
4.3. Cumulative energies
Energies and Signaland signal amplitudes
Amplitudes of the
as a Function EM wave
of Sensor signals from the simulations
Positions
were evaluated by comparing with those from the experimental using the active parts of the transformer
Finally, cumulative energies and signal amplitudes of the EM wave signals from the simulations
in order to validate the newly developed simulation technique.
were evaluated by comparing with those from the experimental using the active parts of the
Figure 13 shows the entire and the first 30 ns of the time-domain EM waveforms, which were
transformer in order to validate the newly developed simulation technique.
obtained by sensor C at the EM wave source position 3, respectively. The signal amplitudes of both the
Figure 13 shows the entire and the first 30 ns of the time-domain EM waveforms, which were
simulated and measured values were normalized by the maximum signal strength of each waveform.
obtained by sensor C at the EM wave source position 3, respectively. The signal amplitudes of both
From Figure 13a, the attenuation degree of the EM waves as a function of time agreed well with
the simulated and measured values were normalized by the maximum signal strength of each
the measured one. Furthermore, the EM waves attenuated sufficiently within 100 ns, while in the
waveform. From Figure 13a, the attenuation degree of the EM waves as a function of time agreed
first experiment without the active parts, it took more than 500 ns as shown in Figure 8a. This rapid
well with the measured one. Furthermore, the EM waves attenuated sufficiently within 100 ns, while
attenuation was caused by increasing the reflection and diffraction of the EM waves due to the active
in the first experiment without the active parts, it took more than 500 ns as shown in Figure 8a. This
parts. Also, Figure 13b indicates that the waveforms both by the simulation and measurement were
rapid attenuation was caused by increasing the reflection and diffraction of the EM waves due to the
quite similar up to 50 ns. These agreements in Figure 13a,b suggest that the modeling of the active part
active parts. Also, Figure 13b indicates that the waveforms both by the simulation and measurement
of the transformer is reasonable.
were quite similar up to 50 ns. These agreements in Figure 13a,b suggest that the modeling of the
Figures 14 and 15 show the simulated and measured cumulative energies and signal amplitudes
active part of the transformer is reasonable.
as functions of the source and the sensor positions, respectively. In both figures, the vertical axes were
Figures 14 and 15 show the simulated and measured cumulative energies and signal amplitudes
normalized by the values from sensor A at the source position 1. For both cumulative energies and
as functions of the source and the sensor positions, respectively. In both figures, the vertical axes were
signal amplitudes, on the whole, the simulated results show a similar trend to the measured results.
However, the cumulative energies show better agreement, because the signal amplitudes tend to be
strongly affected by resonances of the EM waves, which are difficult to simulate accurately.
As shown in Figures 12–15, the propagation times of the EM waves, the time-domain EM
waveform and signal strength (i.e., cumulative energies and their amplitudes) as a function of the
sensor position by the simulation showed reasonable agreement with the measured ones. Thus,
the newly developed simulation technique for the EM wave propagation has been successfully
validated. Furthermore, it firstly enables us to investigate the sensitivities of PD measurement as a
function of UHF sensor positions for actual transformers by computation.
The authors believe that this simulation technique will contribute to further investigations for the
optimization of the UHF sensor positions, their numbers as well as the type of sensor, by applying the
antenna modeling technique described in Section 3.2 and the sensitivity investigation as a function of
sensor positions in Section 4.3.
 Sensors 2018, 18, x FOR PEER REVIEW 13 of 16

normalized by the values from sensor A at the source position 1. For both cumulative energies and
signal amplitudes, on the whole, the simulated results show a similar trend to the measured results.
However,
Sensors the
2018, 18, cumulative energies show better agreement, because the signal amplitudes tend
4236 13 to be
of 16
strongly affected by resonances of the EM waves, which are difficult to simulate accurately.

Signal amplitude (arb. unit)

Simulation
Measurement
0.5

-0.5

-1
0 20 40 60 80 100
Time (ns)
(a) Simulated and measured entire waveforms up to 100 ns.

1
Signal amplitude (arb. unit)

0.5

-0.5
Simulation
Measurement
-1
0 10 20 30
Time (ns)
(b) Enlargement of the waveforms at the first 30 ns.
Figure13.
Figure 13.Example
Exampleofofthe
thesimulated
simulatedand
andmeasured
measuredEM
EMwaveforms.
waveforms.(EM
(EMwave
wavesource
sourceposition
position3,3,UHF
UHF
Sensors 2018, 18, x FOR PEER REVIEW 14 of 16
sensorC).
sensor C).

As shown in Figures 2 12–15, the propagation times of the EM waves, the time-domain EM
Cumulative energies (arb. unit)

Simulation
waveform and signal strength (i.e., cumulative energies and their amplitudes) as a function of the
simulationMeasurement
sensor position by the1.5 showed reasonable agreement with the measured ones. Thus, the
newly developed simulation technique for the EM wave propagation has been successfully validated.
Furthermore, it firstly enables
1 us to investigate the sensitivities of PD measurement as a function of
UHF sensor positions for actual transformers by computation.
The authors believe
0.5 that this simulation technique will contribute to further investigations for
the optimization of the UHF sensor positions, their numbers as well as the type of sensor, by applying
the antenna modeling 0technique described in Section 3.2 and the sensitivity investigation as a
B C4.3.D A B C D A B C D
A Section
function of sensor positions in
EM source: 1 EM source: 2 EM source: 3
UHF sensors (A-D) and EM source positions (1-3)

Figure 14.Simulated
Figure14. Simulatedand
andmeasured
measuredcumulative
cumulativeenergies
energiesfor
foreach
eachUHF
UHFsensors and
sensors EM
and wave
EM source.
wave source.

1.4
Signal amplitudes (arb. unit)

1.2
1
0.8
0.6
0.4
0.2 Simulation
Measurement
0
A B C D A B C D A B C D
 0

C
A B C D A B C D A B C D
EM source: 1 EM source: 2 EM source: 3
UHF sensors (A-D) and EM source positions (1-3)
Sensors 2018, 18, 4236 14 of 16
Figure 14. Simulated and measured cumulative energies for each UHF sensors and EM wave source.

1.4

Signal amplitudes (arb. unit)

1.2
1
0.8
0.6
0.4
0.2 Simulation
Measurement
0
A B C D A B C D A B C D
EM source: 1 EM source: 2 EM source: 3
UHF sensors (A-D) and EM source positions (1-3)

Figure 15. Simulated and measured signal amplitudes for each UHF sensors and EM wave source.
Figure 15. Simulated and measured signal amplitudes for each UHF sensors and EM wave source.
5. Conclusions
5. Conclusions
The authors have proposed a simulation technique for EM wave propagation within transformers
The authors
and validated have proposed
the simulated a simulation
results by comparing withtechnique for EM wave
those experimentally propagation
obtained, using a 630within
kVA
transformers and validated
distribution transformer. the simulated results by comparing with those experimentally obtained,
using a 630
First, kVA distribution
validities of modeling transformer.
methods for a monopole antenna as an EM wave source, UHF
First, validities of modeling
sensors as well as the transformer tank methods
werefor a monopole
investigated by antenna
comparing as with
an EM wave source,
experimental UHF
results
sensors with
obtained as wellan as the transformer
empty transformer tank
tank.were investigated
Consequently, the by comparing
simulated with experimental
time-domain results
EM waveforms,
obtained with an empty transformer tank. Consequently, the simulated time-domain
the attenuation rate of EM wave strengths and cumulative energies as a function of UHF sensor EM waveforms,
the attenuation
position showed goodrate of EM wavewith
agreement strengths and cumulative
the measured energies
ones. Therefore, as amodeling
those function methods
of UHF sensor
were
position showed
successfully validated.good agreement with the measured ones. Therefore, those modeling methods were
successfully validated.
Second, propagation times, signal amplitudes and cumulative energies of the EM waves were
Second,
evaluated propagation
by simulation, times, signaland
measurement amplitudes
theoreticaland cumulativeby
consideration energies
using aof630
thekVA
EMdistribution
waves were
evaluated by simulation, measurement and theoretical consideration by using
transformer in order to confirm the validation of the modeling of the active parts of a transformer.a 630 kVA distribution
transformer in order to confirm the validation of the modeling of the active
As a result, the simulated EM waveforms, their propagation times, cumulative energies and signal parts of a transformer.
As a result,
amplitudes asthe simulated
a function EM sensor
of UHF waveforms, their
position propagation
showed reasonabletimes, cumulative
agreement with energies and signal
the experimentally
amplitudes as a function of UHF sensor position showed reasonable
and theoretically obtained ones. This suggests that the computational conditions, including agreement with the
the
experimentally
modeling and theoretically
of the transformer obtained
structure ones. This suggests that the computational conditions,
were appropriate.
including the modeling of the transformer structure
Based on these results, this newly developed simulation were appropriate.
technique, proposed in this paper, will
contribute to the optimization of the UHF sensor positions andtechnique,
Based on these results, this newly developed simulation proposed
their numbers in this
as well paper,
as the typewill
of
contribute to the optimization of the UHF sensor positions and
UHF sensors to obtain the desired PD detection sensitivity for power transformers. their numbers as well as the type of
UHF sensors to obtain the desired PD detection sensitivity for power transformers.
Author Contributions: T.U. conceived and performed the experiments, analyzed the data and wrote the paper;
S.T. offered valuable suggestions and guidance.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Godina, R.; Rodrigues, E.M.G.; Matias, J.C.O.; Catalao, J.P.S. Effect of loads and other key factors on
oil-transformer aging: Sustainability benefits and challenges. Energies 2015, 8, 12147–12186. [CrossRef]
2. Yang, Q.; Su, P.; Chen, Y. Comparison of impulse wave and sweep frequency response analysis methods for
diagnosis of transformer winding faults. Energies 2017, 10, 431. [CrossRef]
3. Cennamo, N.; Maria, L.D.; D’Agostino, G.; Zeni, L.; Pesavento, M. Monitoring of low levels of furfural in
power transformer oil witih a sensor system based on a POF-MIP platform. Sensors 2015, 15, 8499–8511.
[CrossRef] [PubMed]
4. Tenbohlen, S.; Denissov, D.; Hoek, S.M.; Markalous, S.M. Partial discharge measurement in the ultra-high
frequency (UHF) range. IEEE Trans. Dielectr. Electr. Insul. 2008, 15, 1544–1552. [CrossRef]
Sensors 2018, 18, 4236 15 of 16

5. International Electrotechnical Commission (IEC). IEC 60270, High Voltage Test Techniques—Partial Discharge
Measurements; IEC: Geneva, Switzerland, 2000.
6. Li, S.; Gao, G.; Hu, G.; Gao, B.; Yin, H.; Wei, W.; Wu, G. Influences of traction load shock on artificial
partial discharge faults within traction transformer—Experimental test for pattern recognition. Energies
2017, 10, 1556.
7. Duval, M. A review of faults detectable by gas-in-oil analysis in transformers. IEEE Electr. Insul. Mag.
2002, 18, 8–17. [CrossRef]
8. Markalous, S.; Tenbohlen, S.; Feser, K. Detection and location of partial discharges in power transformers
using acoustic and electromagnetic signals. IEEE Trans. Dielectr. Electr. Insul. 2008, 15, 1576–1583. [CrossRef]
9. Albarracin, R.; Ardila-Rey, J.A.; Mas’ud, A.A. On the use of monopole antennas for determining the effect
of the enclosure of a power transformer tank in partial discharges electromagnetic propagation. Sensors
2016, 16, 148. [CrossRef]
10. Zhang, X.; Zhang, G.; Li, Y.; Zhang, J.; Huang, R. On the feasibility of gap detection of power transformer
partial discharge UHF signals: Gap propagation characteristics of electromagnetic waves. Energies
2017, 10, 1531. [CrossRef]
11. Siegel, M.; Beltle, M.; Tenbohlen, S.; Coenen, S. Application of UHF sensors for PD measurement at power
transformers. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 331–339. [CrossRef]
12. Siegel, M.; Beltle, M.; Tenbohlen, S. Characteristics of UHF PD sensors for power transformers using an
oil-filled GTEM cell. IEEE Trans. Dielectr. Electr. Insul. 2017, 23, 1580–1588. [CrossRef]
13. Mirzaei, H.; Akbari, A.; Gockenbach, E.; Miralikhani, K. Advancing new techniques for UHF PD detection
and localization in the power transformers in the factory tests. IEEE Trans. Dielectr. Electr. Insul. 2015, 22,
448–455. [CrossRef]
14. Tang, Z.; Li, C.; Cheng, X.; Wang, W.; Ji, J.; Li, J. Partial discharge location in power transformers using
wideband RF detection. IEEE Trans. Dielectr. Electr. Insul. 2006, 13, 1193–1199. [CrossRef]
15. Judd, M.; Yang, L.; Hunter, I. Partial discharge monitoring for power transformers using UHF sensors part 1:
Sensors and signal interpretation. IEEE Electr. Insul. Mag. 2005, 21, 5–14. [CrossRef]
16. Tang, Z.; Li, C.; Wang, W.; Wang, H.; Wang, L.; Ding, Y. The propagation characteristics of electromagnetic
wave generated from partial discharges in power transformer by FDTD simulation. In Proceedings of the
Electrical Insulation and Dielectric Phenomena, Vancouver, BC, Canada, 17–20 October 2007; pp. 200–203.
17. Wang, P.; Li, J.; Zhao, L.; Jiang, T.; He, Z. Investigation on the propagation characteristic of electromagnetic
waves in converter transformers. In Proceedings of the IEEE International Conference on High Voltage
Engineering and Application, Shanghai, China, 17–20 September 2012; pp. 353–357.
18. Okabe, S.; Yuasa, S.; Kaneko, S.; Yoshimura, M.; Muto, H.; Yoshiyasu, H.; Nishida, C.; Kamei, M. Simulation
of propagation characteristics of higher order mode electromagnetic waves in GIS. IEEE Trans. Dielectr.
Electr. Insul. 2006, 13, 855–861. [CrossRef]
19. Yoshimura, M.; Muto, H.; Nishida, C.; Kamei, M.; Okabe, S.; Kaneko, S. Propagation properties of
electromagnetic wave through T-branch in GIS. IEEE Trans. Dielectr. Electr. Insul. 2007, 14, 328–333.
[CrossRef]
20. Behrmann, G.; Smajic, J. RF PD signal propagation in GIS: Comparing S-parameter measurements with an
RF transmission model for a short section of GIS. IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 1331–1337.
[CrossRef]
21. Mirzaei, H.R.; Akbari, A.; Gockenbach, E.; Zanjani, M.; Miralikhani, K. A novel method for
ultra-high-frequency partial discharge localization in power transformers using the particle swarm
optimization algorithm. IEEE Electr. Insul. Mag. 2013, 29, 26–39. [CrossRef]
22. Choi, W.; Hwangbo, S.; Park, C.; Park, J. Denoising of UHF Signals Based on RBPF and the Localization of PD
Sources Using FDTD Method in Power Transformer; No. A2-105; Cigre2016: Paris, France, 2016.
23. Du, J.; Chen, W.; Xie, B. Simulation analysis on the propagation characteristics of electromagnetic wave
generated by partial discharges in power transformer. In Proceedings of the 2016 IEEE Conference
on Electrical Insulation and Dielectric Phenomena (CEIDP), Toronto, ON, Canada, 16–19 October 2016;
pp. 179–182.
24. Sinaga, H.H.; Phung, B.T.; Blackburn, T.R. Partial discharge localization in transformers using UHF detection
method. IEEE Trans. Dielectr. Electr. Insul. 2012, 19, 1891–1900. [CrossRef]
Sensors 2018, 18, 4236 16 of 16

25. Judd, M.D.; Hampton, B.F.; Farish, O. Modeling partial discharge excitation of UHF signals in waveguide
structures using Green’s functions. IEE Proc. Sci. Meas. Technol. 1996, 143, 63–70. [CrossRef]
26. Yee, K.S. Numerical solution of initial boundary value problems involving Maxwell’s equations in isotropic
media. IEEE Trans. Antennas Propag. 1996, AP-14, 302–307.
27. Weiland, T. Time domain electromagnetic field computation with finite difference methods. Int. J. Numer.
Modell. Electr. Netw. Devices Fields 1996, 9, 295–319. [CrossRef]
28. Homsup, N.; Jariyanorawiss, T. An improved FDTD model for the feeding gap of a dipole antenna.
In Proceedings of the IEEE SoutheastCon 2010, Concord, NC, USA, 18–21 March 2010; pp. 475–478.
29. Viola, F.; Romano, P. Building Partial Discharge Signal Wireless Probes; Progress in Compact Antennas;
Laure Huitema, Intech: London, UK, 2014; pp. 177–203.
30. Liu, T.; Zhao, Y.; Zheng, Y.; Jia, M. Accurate computation of input admittance of the antennas on the platform
using the advanced delta-gap source model. In Proceedings of the Cross Strait Quad-Regional Radio Science
and Wireless Technology Conference, Taipei, Taiwan, 23–27 July 2012; pp. 30–34.
31. Liu, L.; Weng, Y.; Cheung, S.; Yuk, T.; Foged, L. Modeling of cable for measurements of small
monopole antennas. In Proceedings of the Antennas and Propagation Conference, Loughborough, UK,
14–15 November 2011.
32. Hyun, S.; Kim, S.; Kim, Y. Finite-difference time-domain model for the feeding gap of coaxial probe driven
antennas. IET Microw. Propag. 2009, 3, 501–506. [CrossRef]
33. Hertel, T.; Smith, G. On the convergence of common FDTD feed models for antennas. IEEE Trans.
Antennas Propag. 2003, 51, 1771–1779. [CrossRef]
34. Akbazadeh, A.R.; Shen, Z. On the gap source model for monopole antenna. IEEE Antennas Wirel. Propag. Lett.
2008, 7, 115–118. [CrossRef]
35. Schmitt, H.J.; Harrison, C.W., Jr.; Williams, C.S., Jr. Calculated and experimental response of thin cylindrical
antennas to pulse excitation. IEEE Trans. Antennas Propag. 1966, 14, 120–127. [CrossRef]
36. Yang, L.; Judd, M.D. Propagation characteristics of UHF signals in transformers for locating partial
discharge sources. In Proceedings of the 13rd International Symposium on High Voltage Engineering,
Delft, The Netherlands, 25–29 August 2003.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).