This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

Design and Application of a Circular Printed

Monopole Antenna in Partial Discharge
Detection
George V. R. Xavier, Edson G. da Costa, Member, IEEE, Alexandre J. R. Serres, Member, IEEE,
Luiz. A. M. M. Nobrega, Adriano C. Oliveira, and Herbet F. S. Sousa

The most traditional method of PD detection was

Abstract—In this article, the design and application of a established by IEC 60270 [1]. This method consists in the
circular microstrip antenna for partial discharge detection is measurement of current pulses emitted during the occurrence
presented. For this purpose, the truncated ground plane of PDs. For this purpose, coupling capacitors are connected in
technique was applied for bandwidth enhancement. In addition,
aiming at an optimized structure, computational simulations with
parallel with the monitored equipment. In addition, the use of
several combinations of the antenna dimensional parameters coupling capacitors represents a practical limitation for the
were performed. In order to verify whether the developed method, since an electrical connection between the capacitor
antenna met the prerequisites (bandwidth and gain) for practical and the equipment becomes necessary. However, the coupling
application in partial discharge detection, the optimized antenna capacitor can assume large physical dimensions for
was subjected to tests in an anechoic chamber. Lastly, the applications at high voltage levels, characterizing this method
antenna sensitivity was compared with the IEC 60270 standard
method. For this purpose, simultaneous tests were carried out in
as highly invasive.
a partial discharge generator arrangement composed of an oil In order to overcome the practical difficulties faced in the
cell with point-to-plane and plane-to-plane electrode application of online monitoring by means of the IEC 60270
configurations. The developed antenna can detect apparent standard method, researchers have studied several
charge values higher than 30 pC and has an average gain greater methodologies for continuous monitoring of the PD activity in
than 2 dB along its bandwidth. The obtained results attested to high-voltage equipment [3–7]. One of these methods consists
the potential of the developed antenna for the development of
UHF sensors for partial discharge detection in high-voltage
in the detection of electromagnetic waves emitted by the PD
equipment. current pulses, which propagate through the insulation system
of the high-voltage equipment. These electromagnetic waves
Index Terms—Circular microstrip antenna, high-voltage have a frequency range of 300 MHz to 3 GHz in the UHF
equipment monitoring, partial discharge, UHF method operation range [8].
One of the main advantages offered by the UHF method is
the elimination of an electrical connection between the sensor
I. INTRODUCTION and the monitored equipment. Besides, the high-frequency
operation range isolates the detection system from most
P ARTIAL discharges (PDs) are low-magnitude electrical
discharges that partially short-circuit insulation material
subjected to intense electric fields [1]. The continuous action of
interference in substations, such as power electronics
switching and corona discharges, since the signals generated
the PDs on the insulation material results in its degradation, by these discharges have components with significant energy
which may lead to a complete dielectric breakdown and in frequencies between 200 and 300 MHz [9]. In addition, low
consequently the failure of high-voltage equipment. Therefore, frequency signals (< 300 MHz) are caused mainly by the
the continuous monitoring of PD activity in equipment can be interference and/or reflection phenomena and, therefore, not
used as a failure-prevention tool, indicating in advance the primarily related to the original charge of PD pulse. Therefore,
development of dielectric problems [2]. the UHF method is characterized as a non-invasive and
interference-immune technique, making it attractive for the
This work was supported by the Postgraduate Program in Electrical continuous monitoring of high-voltage equipment.
Engineering (PPgEE) of the Federal University of Campina Grande (UFCG), The first application of UHF sensors in PD detection is
the Coordination for the Improvement of Higher Level Education Personnel reported in [8], which presents the detection and location of
(CAPES), and the National Council for Technological and Scientific
Development (CNPq).
PDs in a 420-kV Gas Insulated Substation (GIS), attesting to
G. V. R. Xavier, L. A. M. M. Nobrega, A. C. Oliveira, and H. F. S. Sousa the efficiency of the UHF method. Over the years, other
are with PPgEE, UFCG, Campina Grande, Brazil (e-mails: george.xavier@ researches have proved the efficiency and reliability of the
ee.ufcg.edu.br, [email protected], [email protected] UHF method in GISs [10, 11], and its application has been
and herbet.sousa@ ee.ufcg.edu.br).
E. G. da Costa and A. J. R. Serres are with the Electrical Engineering
extended to other high-voltage pieces of equipment, as power
Department, UFCG, Campina Grande, Brazil (e-mails: transformers and high-voltage cable connections [12–14].
[email protected] and [email protected]).

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

In order to improve the sensitivity of UHF sensors, calculate the physical radius of the patch by means of the
researchers investigated the main frequency range in which following equations [17]:
PD electromagnetic pulses have a higher concentration of
 'mn c
irradiated energy for different insulating materials [9, 12]. In f mn = , (1)
general, more significant energy concentrations are reported in 2ae  r
the frequency range of 300 to 1500 MHz, aiding in the sensor where m and n are the number of half cycle variations along the x
sensitivity optimization process in the design stages. and y axes, respectively, c is the speed of light, εr is the substrate
The UHF sensors applied in PD detection can assume dielectric constant, and ae is the electrical radius of the patch,
several structures, such as loop electrode, disk electrode, whose value is higher than its physical radius due to the field
conical coupler, and antennas (monopole, dipole, microstrip, fringing effect. The values of χ’mn represent the Bessel’s function
and others) [15, 16]. Among these sensors, microstrip zeros and, according to (1), determine the order of operating
antennas are highlighted for having desirable characteristics frequency of the circular microstrip antenna. For example, the
for practical applications, such as low cost, attractive radiation first four values of χ’mn are:
patterns, and ease of installation/construction [17]. In addition,
microstrip antennas can be built into compact shapes, resulting  '11 = 1.8412
in minimally invasive sensors and lower structural impacts on  '21 = 3.0542
high-voltage equipment [18]. However, microstrip antennas  '01 = 3.8318
have a narrow bandwidth (1–2%), failing to give full coverage
of the frequency range of PD activity (300–1500 MHz).  '31 = 4.2012
In this way, to attest to the applicability of UHF microstrip
Finally, the patch’s physical radius can be approximated by:
antennas in PD detection, several researches have been
developed in the application of bandwidth enhancement F
a= , (2)
techniques, multi-resonance frequency antennas, dimensional  1/ 2
 2h   F  
reduction, gain improvement, and others [19–23]. However, 1 + ln   + 1.7726  
studies of microstrip antennas for PD detection application are   r F   2h  

still incipient and the previous cited works have either narrow
operating bandwidths or low gain antennas. where
Therefore, the main objective of this work is to design a
8.791 x10 9
circular microstrip antenna with a bandwidth wide enough for F = . (3)
fr r
PD detection in high voltage equipment. Hence, the truncated
ground plane technique was used to enhance the antenna’s In order to detect the signals of the incident electromagnetic
bandwidth. Adjustments of the antenna’s dimensional waves radiated over the patch, the antenna is connected to
parameters were performed in order to obtain an optimized measuring instruments by means of a coaxial cable. The
structure. The designed antenna was subjected to bandwidth connection between the coaxial cable and the antenna is usually
and gain measurement tests in an anechoic chamber. Finally, performed in three modes: microstrip line feeding, coaxial probe,
sensitivity tests of PD detection were performed through and coupling [17]. Regardless of the form of connection,
comparison of the results with the IEC 60270 method applied microstrip antennas have spurious radiation that intensifies their
in a PD generator arrangement composed of an oil cell with characteristic inductance, resulting in a narrow bandwidth with
point-to-plane and plane-to-plane electrode configurations. values of around 1 to 2% regarding the central operation
frequency [17]. Therefore, the main techniques for bandwidth
II. ANTENNA DESIGN enhancement consist in the insertion of capacitances into the
A microstrip antenna is a structure composed of a thin metallic structure to compensate for the characteristic inductance, resulting
layer (patch) which partially covers the area of a dielectric in a wideband antenna.
substrate, with thickness h, supported on a metallic ground plane. Differently from the probe and coupling feeding techniques,
The general structure of a circular microstrip patch is presented in microstrip line feeding allows the use of bandwidth enhancement
Fig. 1. techniques that result in bandwidths larger than 100% regarding
the central operation frequency [24]. For this type of technique,
Dielectric Substrate Patch the insertion of capacitive components occurs from modifications
of the antenna ground plane. In this work, the applied
modification consisted in ground plane truncation at the height of
h
the antenna feeding line, as presented in Fig. 2.
Ground Plane W

Fig. 1. General structure of a circular microstrip antenna.

L Patch
The design of a circular microstrip antenna is projected
according to the initial definition of three parameters: the Ground Plane
operating frequency, thickness, and dielectric constant of the
Connection
applied substrate. From these parameters, it is possible to Fig. 2. Circular microstrip antenna with truncated ground plane.

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

For the application of the truncated ground plane technique in a known gain (4.5 dBi), positioned at the far field distance (R)
circular microstrip antenna, the patch radius is usually from the constructed antenna as presented in Fig. 4.
approximated in an initial stage as λ/8 of the desired lower
operating frequency [24]. After the initial estimation, the antenna
R
dimensional parameters are subjected to adjustments until an
optimized structure is achieved.
Reference Antenna Constructed Antenna
Fig. 4. Schematic of the experimental arrangement applied in the gain
III. COMPUTATIONAL AND EXPERIMENTAL PROCEDURES measurement test.
For the simulation and design of the antenna, the software From the measured values of transmitted and received
High Frequency Structure Simulator (HFSS) from ANSYS power (PT and PR, respectively) and reference antenna gain
was used. (GR), the designed antenna gain (GD) was calculated according
The circular microstrip antenna was designed so that its to the Friis equation [25]:
bandwidth was approximately equal to the frequency band of
PD occurrence, that is, 300–1500 MHz. The substrate  4ππ 
G D (dB) = PR (dBm) - PT (dBm) - G R (dB) + 20log (dB). (4)
thickness and dielectric constant were defined as 1.52 mm and  λ 
4.4 (FR4), respectively. In order to reduce the effect of external interferences and
In order to achieve the desired bandwidth, the computational signal reflections during the tests, the return loss and gain
procedures were divided into two stages. The first stage measurements were carried out in an anechoic chamber.
consisted in the design of an initial model according to the To estimate the sensitivity of the antenna’s PD detection, the
approximations described in [24]. The initial model simulated experimental arrangement presented in Fig. 5 was applied.
by HFSS is presented in Fig. 3.
15 mH
Oil Cell
Designed Antenna
1000 pF

Electro des
0 - 100 kV
LDM-5

0 - 220 V

Coaxial Cable

Osciloscope

a) b)
Fig. 5. Experimental arrangement applied for PD measurement.
Fig. 3. Initial model simulated by HFSS for the circular microstrip antenna with
truncated ground plane: a) general view; b) details of the slot inserted in the ground The experimental procedure applied for the generation and
plane. detection of PD consisted in a gradual increase of voltage, by
The second stage consisted in the adjustment of the means of a regulating transformer, until PD activity was
dimensional parameters, such as the patch radius, ground generated in the test object (oil cell) and detected by the
plane height, and length/width of the ground plane slot. In the manufactured antenna and the IEC 60270 standard method.
adjustment stage, the impact exerted by each of the The standard method is represented by a coupling capacitor
dimensional parameters on the antenna performance was (1000 pF) and a resonant circuit (LDM–5). All the calibration
procedures established by the standard were executed before
evaluated. For this purpose, only one dimensional parameter
the voltage application tests.
was adjusted at a time, while the others kept their initial
In order to verify if the detected signals were from partial
values. In this way, the influences of the dimensional
discharges originated in the applied test object, previous
parameters on the antenna performance were individually
measurements with IEC 60270 standard method were made.
studied and later combined to attain an optimized structure.
The first previous measurement was regarding to the noise
The bandwidth of the simulated antenna was defined as the
level in the laboratory setup, which was made with the system
entire frequency range at which the return loss (S11) was below
de-energized. According to the IEC 60270 standard method,
the –10 dB threshold.
the background noise apparent charge values must be inferior
According to [22], for PD detection applications, microstrip
to 50% of the measured PD value [1]. In this measurement, the
antennas must present a mean gain higher than 2 dB for their
detected background noise was below 10 pC.
operating bandwidth. Therefore, the mean gain values from
The second measurement was made to ensure that the partial
the simulated antennas were compared to the 2-dB threshold
discharges were originated in the test object. For this, the test
recommended for the antenna approval. object was withdrawn from the laboratory setup. High voltage
After the simulation stages, the designed antenna was values (above 50 kV) were applied in the IEC 60270 standard
manufactured and subjected to return loss and gain method used and the signals detected remained in the
measurements. The return loss test used an N9923A spectrum background noise levels, i.e, there was no PD activity. Hence,
analyser. The gain test used an experimental arrangement any partial discharge signal detected for this high voltage
composed of a reference antenna (Hyperlog 30100X) with a application range (< 50 kV) was originated by the test object.

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

The distance between the antenna and the test object was increase is so significant that a single operation band is
defined according to the calculated far field limit regarding the achieved (251–1203 MHz), resulting in a bandwidth of
antenna’s central operation frequency. 130.94%.
In order to obtain a large range of apparent charge values
0
for the sensitivity test, two electrode configurations were
applied in the oil cell. For the generation of PDs with high −5
apparent charge magnitudes, a point-to-plane configuration
−10
with spacing of 2.0 cm was used. From this configuration,

S11 (dB)
apparent charges with maximum values of 7000 pC were −15
generated, resulting in a good representation of the PD levels
in high-voltage pieces of equipment with damaged insulating −20 27 cm
30 cm
systems. −25 32 cm
34 cm
For the PD generation with low magnitudes, a plane-to-plane 37 cm
configuration with spacing of 3.0 cm was used. From this −30
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Frequency (GHz)
configuration, apparent charges with minimum and maximum
values of 30 and 200 pC, respectively, were generated, Fig. 7. Simulated return loss for adjustments in the ground plane width.
resulting in a good representation of the PD inception levels in
0
insulating systems of high-voltage equipment.
−5
Additionally, the PD signals detected by the antenna were −10
analysed in the frequency domain for the verification of the −15
main frequency range of the generated discharges. −20

S11 (dB)
−25
IV. RESULTS AND ANALYSIS −30
−35 45 mm
The simulated return loss from the initial model (Fig. 3) is 47 mm
−40 50 mm
presented in Fig. 6. 53 mm
−45
0 55 mm
−50
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Frequency (GHz)
−5
Fig. 8. Simulated return loss for adjustments to the ground plane height.
−10
S 11 (dB)

After the evaluation of the impacts of return loss originated

−15
by the adjustments to the ground plane dimensions,
adjustments to the patch radius were made, as presented in
−20 Fig. 9.
0
−25
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 −5
Frequency (GHz)
−10
Fig. 6. Simulated return loss of the initial antenna model.
−15
S11 (dB)

The initial antenna model presented two different operation

−20
bands, with bandwidths equal to 52.43% (244–417.4 MHz)
−25
and 48.12% (697.8–1140 MHz), regarding their respective 105 mm
110 mm
central operation frequencies. Despite the significant −30 115 mm
120 mm
bandwidth increase compared to most microstrip antennas (1 −35 125 mm
130 mm
to 2%), adjustments to the dimensional parameters of the −40
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
initial model were still necessary to allow full coverage of the Frequency (GHz)
PD frequency range (300–1500 MHz). Fig. 9. Simulated return loss for adjustments to the patch radius.
The ground plane width was the first dimensional parameter
evaluated during the adjustment stage. The results of these The reduction of the patch radius results in higher return loss
changes are presented in Fig. 7. and bandwidth, mainly for the second operation band, as
The reduction of the ground plane width resulted in expected according to microstrip antenna theory and (1), (2),
increases of the lower and upper frequencies for both and (3).
operation bands, so that they approached the lower and upper Finally, adjustments to the length and width of the ground
PD frequency limits. Additionally, the verified return loss plane slot were evaluated, as presented in Figs. 10 and 11.
degradation promoted distancing between the two operating From the adjustments of the slot, it can be verified that the
bands. To improve the return loss, adjustments to the ground increase of the slot length results in higher return loss for the
plane height were made, as presented in Fig. 8. second operation band. It was also noticed that the inferior
From the results presented in Figure 8, it can be verified that frequency of the second band is displaced to lower values,
better return loss results are achieved with a higher ground while the superior frequency is displaced to higher values,
plane height. With a height of 53.0 mm, for example, the leading to the formation of a larger bandwidth. When the slot

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

width was adjusted, lower significant impacts on the return

loss were observed when compared to the slot length results.
However, as the width increased, displacements occur in the
lower frequency in the second band, providing a larger
bandwidth.
0

−10

−20
S11 (dB)

−30

−40 Without Slot Fig. 13. Comparison between measured and simulated return losses.
3x3 mm
3x5 mm
−50 3x10 mm For the gain analysis, radiation patterns, such as those shown
3x20 mm
3x30 mm in Figure 14, were simulated for the entire antenna operation
−60
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 bandwidth (305–1495 MHz) at 10-MHz intervals.
Frequency (GHz)
Fig. 10. Simulated return loss for adjustments in the slot length.
0

−10

−20
S11 (dB)

−30

−40 Without Slot

3x3 mm
5x3 mm
−50 8x3 mm
10x3 mm
−60
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Frequency (GHz)
Fig. 11. Simulated return loss for adjustments in the slot width.

After the combined analysis of the impacts on return loss for

each dimensional parameter, the best simulated antenna
presented the following final configuration: a patch radius of
110.0 mm, a ground plane height of 57.0 mm, and slot
dimensions of 6.0 mm wide and 10.0 mm long. The final
antenna size was 30.0 × 30.0 cm. The manufactured antenna
and the comparison between the measured and simulated
return loss are presented in Figs. 12 and 13, respectively.
The simulated (305–1495 MHz) and measured (312–1481
MHz) operating bandwidths were considered satisfactory,
since they covered most of the frequency range in which PDs
occur. The difference between the simulated and measured
return loss behaviors can be attributed to construction aspects
of the antenna, such as slight differences between the
simulated FR4 and the one used in practice, as well as the non-
consideration of the SMA connectors during the simulations.

a) b)
Fig. 12. Manufactured antenna: a) front view, b) back view.

Fig. 14. Radiation patterns for 350, 900, and 1500 MHz, respectively.

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

From the simulated radiation patterns, a mean value of method, resulting in a mean sensitivity rate (IEC/antenna) of
maximum gain of 4.58 dB was calculated. For the 6.2. This result was expected, since the standard method
manufactured antenna, the mean value of maximum gain, consists in a direct electrical connection with the monitored
calculated according to (4), was equal to 4.92 dB. The equipment, resulting in lower losses than in a radiometric
difference between the simulated and measured gains can also based method. In addition, differently from the standard
be attributed to the construction and simulation considerations method, there is no linear correlation between the antenna’s
mentioned before. measured voltage levels and their respective apparent charge
Despite the slight difference between the mean values of levels. The PD pulse numbers 4 and 5, for example, present
simulated and measured gain, both presented values higher very similar voltage levels (820 and 840 mV, respectively),
than the average limit established in [22] as satisfactory for although the same pulses present apparent charges with a
PD detection (2 dB). Therefore, the manufactured antenna was significant difference (2627 and 5932 pC). However, this
approved in the return loss and gain tests, being considered fit result is expected since other studies have already
for application in the PD sensitivity detection tests. demonstrated a nonlinear relationship between the results of
From the antenna’s final physical dimensions (30.0 × 30.0 UHF measurements and the IEC 60270 method [26, 27].
cm), the far field for the central frequency of 900 MHz was The non-linear correlation between the measured voltage
calculated as 1.0 m. Therefore, the antenna was positioned at and apparent charge can be attributed to the different
this distance for the first electrode configuration (point-to- frequencies of the PD pulse occurrences, resulting in particular
plane). The beginning of PD activity was detected upon the values of return loss and gain for each of the detected pulses,
application of 33.3 kV in the oil cell. Samples of the detected inducing different responses in the antenna terminals.
pulses are presented in Fig. 15. For the test of detection sensitivity regarding low-level PD
2000
(plane-to-plane electrode configuration), the beginning of PD
Reference Voltage
activity was detected on the application of 41.8 kV in the oil
Voltage (mV)

1000 PD − Antenna

0
cell. Fig. 16 presents a sample of the detected pulses.
−1000 Reference Voltage
200
Voltage (mV)
PD − Antenna
−2000 100
0 2 4 6 8 10 12 14 16 18 20
0
Time (ms)
−100
10000
Reference Voltage −200
Voltage (mV)

5000 PD − IEC 60270

0 2 4 6 8 10 12 14 16 18 20
Time (ms)
0

−5000 Reference Voltage

200
Voltage (mV)

PD − IEC 60270
−10000 100
0 2 4 6 8 10 12 14 16 18 20
0
Time (ms)
−100
Fig. 15. PD activity detected by the manufactured antenna and standard method for −200
the point-to-plane electrode configuration at 33.3 kV.
0 2 4 6 8 10 12 14 16 18 20
Time (ms)
The results presented in Fig. 15 show that the antenna was
Fig. 16. PD activity detected by the manufactured antenna and standard method for
able to detect all the PD pulses captured by the standard the plane-to-plane electrode configuration at 41.8 kV.
method for the point-to-plane configuration. For the
comparative analysis between the detection sensitivity of the For the PD samples presented in Fig. 16, the antenna was
antenna and the IEC 60270 standard method, the maximum able to detect all the 206 PD pulses detected by the standard
method with apparent charge between 30 and 100 pC. For
measured voltage and apparent charge results for each PD
better visualization of the antenna’s detection sensitivity, a
pulse shown in Fig. 15 are summarized in Table 1.
magnified view of the lowest apparent charge generated by the
Table 1. Magnitude of the PD pulses detected by the antenna and IEC 60270 plane-to-plane electrode configuration is presented in Fig. 17.
standard method for the point-to-plane electrodes.
20 Antenna
IEC 60270 Apparent Rate
Voltage (mV)

Pulse Antenna 10
Number (mV) (mV) Charge (pC) IEC/Antenna 0

1 220 2320 2457 10.55 -10

2 660 6320 6695 9.58 -20

3 1340 2080 2203 1.55 -30

4 820 2480 2627 3.02 13.151 13.152 13.153 13.154 13.155 13.156 13.157 13.158
Time (ms)
5 840 5600 5932 6.67
6 460 4720 5000 10.26 30 IEC 60270
Voltage (mV)

7 600 2080 2203 3.47 20

10
8 80 560 593 7.00 0
9 480 2400 2542 5.00 -10
10 140 640 678 4.57 -20
-30
13.151 13.152 13.153 13.154 13.155 13.156 13.157 13.158 13.159

The results presented in Table 1 show that, despite the Time (ms)
Fig. 17. Magnified view of a 30 pC PD pulse detected by the antenna and standard
effective capacity for PD detection, the antenna’s sensitivity is method for the plane-to-plane electrode configuration.
significantly lower than that of the IEC 60270 standard

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

The results presented in Figs. 15–17 attest to the antenna’s V. CONCLUSIONS

detection sensitivity, since it was able to detect PD pulses with The proposed antenna has the potential to be applied as a
apparent charge corresponding to insulating problems from UHF sensor for PD detection, since it meets the bandwidth
their origin (low-intensity pulses) until a critical stage (high- and gain requirements. The antenna design procedure proved
intensity pulses). itself effective, since the simulated and measured return loss
Samples of the PD pulses in the frequency domain for the and gain are quite similar. The manufactured antenna
point-to-plane and plane-to-plane electrode configurations are presented good detection sensitivity for the applied
shown in Figs. 18 and 19, respectively. experimental arrangement, being able to detect PD pulses with
−30
662 MHz Background Noise
apparent charge between 30 and 7000 pC for a 1.0 m range
333 MHz
Partial Discharges distance. Therefore, with proper adjustments, the proposed
−35
antenna can provide continuous, non-invasive, and relatively
1 .14 GHz
−40 low-cost monitoring of PD activity in high-voltage insulating
Power (dBm)

1 .02 GHz 1.21 GHz

1.08 GHz systems, from its inception until critical degradation levels.
−45
Studies in the frequency domain showed that the antenna’s
−50
483 .5 MHz sensitivity has been proven for the whole of the PD spectrum
(300–1500 MHz). In addition, the identification of different
−55 frequency patterns for each electrode configuration highlights
the future use of the antenna for pattern recognition and
−60
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 diagnosis of different PD defects.
Frequency (GHz)
Lastly, a nonlinear behavior between the measured voltage
Fig. 18. PD pulses frequency domain response for the tip-plane electrode
configuration. levels and their respective apparent charge was noticed,
representing a challenge for future calibration procedures.
−20
462.5 MHz Background Noise Therefore, in order to fully characterize the proposed
−25 Partial Discharges
prototype as a UHF sensor for PD detection, additional studies
−30 are still required to provide the PD classification and
Power (dBm)

−35 correlation between PD apparent charge and the values

−40 measured in millivolts.
833 .5
−45 MHz
REFERENCES
−50
[1] High Voltage Test Techniques: Partial Discharge Measurements. IEC
−55 60270, 2000.
−60 [2] F. H. Kreuger, E. Gulski, and A. Krivda, “Classification of partial
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 discharges,” IEEE Trans. Electr. Insul., vol. 28, pp. 917–931, 1993.
Frequency (GHz) [3] G. Stone, “Partial discharge diagnostics and electrical equipment
Fig. 19. PD pulses frequency domain response for the plane-to-plane electrode insulation condition assessment,” IEE Trans. Dielectr. Electr. Insul., vol.
configuration. 12, no. 5, pp. 891–904, 2005.
[4] M. Meira, C. R. Ruschetti, R. E. Álvarez, and C. J. Verucchi, “Power
In order to highlight the PD frequencies of occurrence, the transformers monitoring based on electrical measurements,” IET Gener.
background noise of the PD measurement setup was recorded Transm. Distrib., vol. 12, no. 12, pp. 2805–2815, 2018.
[5] G. C. Stone, “Condition monitoring and diagnostics of motor and stator
during 30 minutes before the high voltage application. Due to windings – a review,” IEEE Trans. Dielectr. Electr. Insul., vol. 20,
the absence of a shielded high voltage laboratory, some digital no. 6, pp. 2073–2080, 2013.
television UHF interference (532.5, 693.5, 775.5 and 882.5 [6] S. S. Refaat and M. A. Shams “A review of partial discharge detection,
MHz) was detected and considered as background noise. diagnosis techniques in high voltage power cables,” 2018 IEEE 12th Int.
Conf. Compatibility, Power Electronics and Power Engineering (CPE-
Then, any significant changes in the background noise curve POWERENG 2018), 2018.
was considered as partial discharges. [7] A. Sahoo, A. Subramaniam, S. Bhandari, and S. K. Panda “A review on
The frequency domain results show that the designed condition monitoring of GIS,” 2017 Int. Symp. Electrical Insulating
Materials (ISEIM), 2017.
antenna presented good detection sensitivity for almost all of [8] B. F. Hampton and R. J. Meats, “Diagnostic measurements at UHF in gas
the frequency range corresponding to PD activity (300–1500 insulated substations,” IEEE Proc., vol. 135, pp. 137–144, 1988.
MHz). In addition, it can be noted that different frequency [9] S. Tenbohlen, D. Denissov, S. M. Hoek, and S. M. Markalous, “Partial
discharge measurement in the ultra high frequency (UHF) range,” IEEE
patterns were originated for each electrode configuration. For Trans. Dielectr. Electr. Insul., vol. 15, no. 6, pp. 1544–1552, 2008.
the point-to-plane configuration, PD occurrence was mainly [10] K. Masaki, T. Sakakibara, H. Murase, M. Akazaki, K. Uehara, and S.
concentrated in the frequency ranges of 330–490, 620–670, Menju, “On-site measurement for the development of on-line partial
discharge monitoring system in GIS,” IEEE Trans. Power Del., vol. 9, no.
and 975–1400 MHz. For the plane-to-plane configuration, the 2, pp. 805–810, 1994.
pulses were concentrated in the ranges of 400–500 and 800– [11]M. Judd, O. Farish, and J. Pearson, “UHF couplers for gas-insulated
850 MHz. The difference between the frequency patterns substations: a calibration technique,” IEE Proc. Sci., Meas., Technol., vol.
144, no. 3, pp. 117–122, 1997.
originated can be attributed to the particularities attached to [12]M. Judd, B. Pryor, S. Kelly, and B. Hampton, “Transformer monitoring
the discharge formation processes for each electrode using the UHF technique,” in 11th Int. Symp. High Voltage Engineering
configuration applied [28]. (ISH), 1999.

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2896580, IEEE Sensors
Journal

[13]M. Judd, Li Yang, and I. Hunter “Partial discharge monitoring for power Alexandre J. R. Serres (M’13) received his B.S. in
transformer using UHF sensors. Part I: Sensors and signal interpretation,” Electronics Embedded Systems from INPG/ESISAR,
IEEE Electr. Insul. Mag., vol. 21, no. 2, pp. 5–14, 2005. France in 2005, his M.Sc. degree in Optical and Radio
[14] C. H. Lee, Y. C. Lin, M. Y. Chiu, C. H. Huang, S. S. Yen, and C. Haeng, Frequency from INPG/ENSERG, France in 2006, and his
“The study on diagnostics for aging trend of cable termination,” in Int. Ph.D. from UFCG, Brazil, in 2011. In 2011, he joined the
Conf. Condition Monitoring and Diagnosis, CMD, 2008. UFCG as Assistant Professor. His main researches
[15] T. Hoshino, K. Nojima, and M. Hanai, “Real-time PD identification in interests are antennas and RF circuits.
diagnosis of GIS using symmetric and asymmetric UHF sensors,” IEEE
Trans. Power Del., vol. 19, no. 3, pp. 1072–1077, 2004.
[16] T. Ju, X. Zhongrong, Z. Xiaoxing, and S. Caixin “GIS partial discharge
quantitative measurements using UHF microstrip antenna sensors,” in Luiz A. M. M. Nobrega was born in Patos, Paraíba,
Annual Report on Electrical Insulation and Dielectric Phenomena, 2007. Brazil, on September 30, 1991. He received the B.Sc.
[17] C. Balanis, Antenna Theory – Analysis and Design, New Jersey, John and M.Sc. degrees in electrical engineering from UFCG
Wiley, 2005. in 2014 and 2016, respectively. Since 2015 he has
[18] M. Judd, O. Farish, J. S. Pearson, B. F. Hampton, “Dielectric windows worked as a Lecturer with the Federal Institute of
for UHF partial discharge detection,” IEEE Trans. Dielectr. Electr. Insul., Paraíba. His research interests include monitoring of
vol. 8, no. 6, pp. 953–958, 2001. high-voltage equipment, PD localization, and antennas.
[19] J. R. Jung, K. R. Hwang, Y. M. Kim, E. T. Lyu, and H. J. Yang,
“Sensitivity verification and application of UHF sensor for partial
discharge measurement in high voltage power transformer,” in Int. Conf.
Condition Monitoring and Diagnosis, CMD, 2012. Adriano C. Oliveira was born in Campina Grande,
[20] J. Li, T. Jiang, C. Cheng, and C. Wang, “Hilbert fractal antenna for UHF Paraíba, Brazil, on September 13. He received the B.Sc.
detection of partial discharges in transformers,” IEEE Trans. Dielectr. and M.Sc. degrees in electrical engineering from UFCG,
Electr. Insul., vol. 20, no. 6, pp. 2017–2025, 2013. in 2013 and 2015, respectively. From 2015 to 2017 he
[21] B. Sarkar, D. K. Mishra, C. Koley, and N. K. Roy, “Microstrip patch worked as a Lecturer at Faculdade Maurício de Nassau.
antenna based UHF sensor for detection of partial discharge in high His research interests include high-voltage equipment and
voltage electrical equipment,” in Annual IEEE India Conf., 2014. antennas.
[22] H. Luo, P. Cheng, H. Liu, K. Kang, F. Yang, and K. Liu, “Research on
the UHF microstrip antenna for partial discharge detection in high voltage
switchgear,” in IEEE 11th Conf. Industrial Electronics and Applications
(ICIEA), 2016. Herbet F. S. Sousa was born in Pombal, Paraíba, Brazil,
[23] W. Yongqiang, W. Zhuang, and L. Jianfang, “UHF Moore fractal on September 13. He received the B.Sc. and M.Sc.
antennae for online GIS PD detection,” IEEE Antennas Wireless Propag. degrees in electrical engineering from UFCG, in 2015
Lett., 2017. and 2018, respectively. His research interests include
[24] J. Liang, C. C. Chiau, X. Chen, C. G. Parini, “Study of a printed circular high-voltage equipment and renewable energy.
disc monopole antenna for UWB systems,” IEEE Trans. Antennas
Propag., vol. 53, no. 11, pp. 3500–3504, 2005.
[25] M. I. Skolnik, Radar Systems, New York, McGraw-Hill, 1962.
[26] T. Hoshino, H. Koyama, S. Maruyama, and M. Hanai, "Comparison of
sensitivity between UHF method and IEC 60270 for onsite calibration in
various GIS", IEEE Trans. Power Del., vol. 21, pp. 1948–1953, 2006.
[27] A.J. Reid, M.D. Judd, R. A. Fouracre, B.G. Stewart, and D.M. Hepburn,
"Simultaneous measurement of PD using IEC60270 and radio-frequency
techniques", IEEE Trans. Dielectr. Electr. Insul., vol. 18, no. 2, pp. 444–
455, 2011.
[28] E. Kuffel, W. S. Zaengl and J. Kuffel, High Voltage Engineering,
Oxford, Newnes, 2000.

BIOGRAPHIES

George V. R. Xavier was born in Aracaju, Sergipe,

Brazil, on November 30, 1993. He received the B.Sc.
and M.Sc. degrees in electrical engineering from Federal
University of Sergipe and UFCG, in 2016 and 2018,
respectively. His research interests include high-voltage
equipment and antennas.

Edson G. da Costa (M’03) obtained the B.Sc, M.Sc and

D.Sc. degrees in electrical engineering, respectively, in
1978, 1981, and 1999 (Federal University of Paraíba).
Since 1978, he has been working as a professor at the
UFCG. His research interests include high-voltage
equipment, electric field mapping, partial discharges,
grounding, surge arresters, and insulation systems.

1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.