Document [original]

DEVELOPMENT OF A MULTI-SENSOR

DIAGNOSTIC METHOD FOR THE EARLY

DETECTION OF PARTIAL DISCHARGES IN CABLE

CONNECTORS OF MEDIUM-VOLTAGE

SWITCHGEARS

ALI SINAI

DISSERTATION

The results presented in this Ph.D. thesis were obtained from the research project NikMonET at

HTW Berlin, supported by Germany’s Federal Ministry of Education and Research. The project

supervisors were Professor Thomas Gräf, Professor Mathais Menge, and Professor Thomas

Hücker.

Berlin 2024

Ali Sinai: Development of a Multi-Sensor Diagnostic Method for The Early Detection of

Partial Discharges in Cable Connectors of Medium-Voltage Switchgears, , © August 2024

Development Of a Multi-Sensor Diagnostic

Method For The Early Detection Of Partial

Discharges In Cable Connectors Of

Medium-Voltage Switchgears

vorgelegt von

Ali Sinai, M.Sc.

an der Fakultät IV - Elektrotechnik und Informatik

der Technischen Universität Berlin

zur Erlangung des akademischen Grades

Doktor der Ingenieurwissenschaften

- Dr.-Ing. -

genehmigte Dissertation

Promotionsausschuss:

Vorsitzender: Prof. Dr. Ing. Julia Kowal

Gutachter: Prof. Dr. Ing. Ronald Plath

Gutachter: Prof. Dr. Ing. Frank Jenau

Gutachter: Prof. Dr. Ing. Thomas Hücker

Gutachter: Prof. Dr. Ing. Daniel Pepper

Tag der wissenschaftlichen Aussprache: 27. Mai 2024

Berlin 2024

For my parents, Daniel†and Shokooh

ABSTRACT

The ever-increasing demand for electrical energy, along with the need for sus-

tainable energy sources, have imposed new challenges for power grid opera-

tors. Namely, the need to extend their grids and partially overuse the existing

assets. The lack of qualified staff and high equipment costs are additional

problems. However, a stable electrical energy supply must be guaranteed, and

the network operators should be able to solve upcoming problems quickly.

For this reason, the continuous monitoring of electrical networks is gaining

more importance. Monitoring systems must detect various issues in high- and

medium-voltage grid assets and inform the grid operator at an early stage, in

case of a problem. One of the problems that can lead to equipment breakdown

and operational restrictions is partial discharges (PDs), generating a failure in

the insulation materials.

This work presents a novel frequency-selective multi-sensor PD diagnostic

method for monitoring medium-voltage (MV) switchgear cable connectors dur-

ing installation and operation. The developed measurement system makes it

possible to detect faulty components reliably, even with massive background

noise. The monitoring system can be installed during the switchgear’s first

commissioning, or later maintenance. The developed prototype was success-

fully tested under various circumstances in the high-power test laboratory CESI /

IPH, Berlin.

The multi-sensor approach enables the simultaneous acquisition of PD sig-

nals in different frequency ranges and over various propagation paths. Hence,

redundant decision-making is possible, increasing the PD diagnosis’s reliabil-

ity. To this end, nine different sensing methods are investigated to find the

most suitable sensors.

In the next step, the disturbance signal propagation over the medium-voltage

cable is studied. The frequency-selective measurement approach is developed

based on the sensor’s and MV cable’s properties. The effectivity of the frequency-

selective PD measurement to reject external interferences is also discussed.

The measurement system architecture and the corresponding analog signal

processing are discussed. The PD diagnostic method is verified with various

faulty cable connectors and external discharges while background noise signals

disturb the measurements. Finally, the measurement system’s performance is

validated in a high-power test laboratory with harsh noise conditions.

iii

ZUSAMMENFASSUNG

Die ständig steigende Nachfrage nach elektrischer Energie und der Bedarf an

nachhaltigen Energiequellen stellen die Stromnetzbetreiber vor neue Heraus-

forderungen. Sie müssen ihre Netze ausbauen und die vorhandenen Anla-

gen teilweise über die jeweils geplante Lebensdauer und Belastungsgrenzwerte

hinaus nutzen. Der Mangel an qualifiziertem Personal und hohe Investition-

skosten sind zusätzliche Probleme. Eine stabile Versorgung mit elektrischer

Energie muss jederzeit gewährleistet sein. Netzbetreiber müssen daher auftre-

tende Probleme in kürzester Zeit erkennen und beheben. Aus diesem Grund

sollten elektrische Netze permanent überwacht werden. Die Überwachungssys-

teme müssen verschiedene Probleme in den Anlagen des Hoch- und Mittelspan-

nungsnetzes erkennen und den Netzbetreiber im Falle eines Fehlers frühzeitig

informieren. Eine Störung, die zum Ausfall von Anlagen und zu den Betrieb-

seinschränkungen führen kann, ist Teilentladung (TE), die Isolationsversagen

der Anlagenkomponenten hervorrufen kann.

In dieser Arbeit wird ein neuartiges frequenzselektives Multisensor-TE-

Diagnoseverfahren zur Inbetriebnahme und Überwachung von Kabelsteckern

in Mittelspannungsschaltanlagen vorgestellt. Das entwickelte Messsystem er-

möglicht es, die fehlerhaften Komponenten zuverlässig zu erkennen, selbst

wenn starke Störsignale existieren. Der entwickelte Prototyp wurde unter ver-

schiedenen Bedingungen im Labor und im CESI / IPH Institut für elektrische

Hochleistungstechnik erfolgreich getestet.

Der Multisensoransatz ermöglicht die gleichzeitige Erfassung verschiedener

TE-Signalemissionen in unterschiedlichen Frequenzbereichen und auf zum Teil

unterschiedlichen Ausbreitungswegen. Damit ist eine redundante Entschei-

dungsfindung möglich, was die Zuverlässigkeit der TE-Diagnose erhöht. In

diesem Zusammenhang werden neun verschiedene Erfassungsmethoden

untersucht um die effektivsten Sensoren auszuwählen.

In einem weiteren Schritt wird die Ausbreitung von Störsignale über das

Mittelspannungskabel untersucht. Anhand der Eigenschaften der Sensoren

und des Mittelspannungskabels wird die frequenzselektive Messung erläutert.

Die Effektivität der frequenzselektiven TE-Messung bei der Unterdrückung

externer Störsignale wird diskutiert.

Die Messsystemarchitektur und die hierzu zum Einsatz kommende analoge

Signalverarbeitung wird vorgestellt und diskutiert. Schließlich wird die TE-

Diagnosemethode mit verschiedenen fehlerhaften Kabelsteckern verifiziert,

wobei verschiedene Störsignale und externe Entladungen die Messgüte bee-

influssen.

ACKNOWLEDGEMENTS

I would like to extend my sincere thanks to my esteemed Ph.D. supervisor,

professor Ronald Plath, who has supported me with his immense knowledge

and wealth of experience during my research.

My special thanks go to my Ph.D. supervisor, professor Thomas Hücker,

who has supported me over the past few years in all areas. His expertise was

invaluable for my research, with his insightful help and feedback bringing my

work to a higher level all round.

I would like to express my sincere gratitude to my further Ph.D. supervisors,

professor Matthias Menge and professor Thomas Gräf, who encouraged me to

start my research on their project. They have supported me with all possibili-

ties to reach my goal.

I am deeply grateful to professor Frank Jenau and professor Daniel Pepper

for their invaluable support in reviewing my Ph.D. thesis. Their insights and

guidance have played a crucial role in refining my thesis.

Björn Böttcher, as a colleague and friend, has always supported me during

my professional work and private life. He is the person who always helped me

overcome problems and worries. Without him, the project goals would have

been impossible to fulfill. Thank you, Björn!

I also wish to express my gratitude for all the support I received from profes-

sor Michael Schmidt, professor Petra Bittrich, Henning Müller, and professor

Stefanie Molthagen-Schnöring who gave me the opportunity to complete my

dissertation. In addition, I would like to thank my colleagues in the department

of engineering – energy and informationat the university of applied sciences HTW

Berlin.

Finally, I would like to express my appreciation to my family, especially my

mother, sisters, and uncle S.M. Emami, for their encouragement and support

throughout my studies.

PUBLICATIONS

This thesis is based on the following peer-reviewed publications:

• A. Sinai; B. Böttcher; M. Menge; T. Gräf; T. Hücker: Big Data approach

to monitoring of energy systems and partial discharge (PD) detection. In:

Proceedings of International Conference on Condition Monitoring, Diag-

nosis and Maintenance, CMDM 2017, S. 194-201, Cigre, Bucharest, 2017,

ISSN 2344-245x.

• A. Sinai; B. Böttcher; M. Menge; T. Gräf; R. Plath; T. Hücker: Multi-

Physical Sensor Fusion Approach For Partial Discharge Detection On

Medium Voltage Cable Connectors. In: 2019 2nd International Conference

on High Voltage Engineering and Power Systems (ICHVEPS), S. 202-207,

-, 2020, ISBN 978-1-7281-2669-2.

• A. Sinai; B. Böttcher; M. Menge; T. Gräf; R. Plath; T. Hücker: Frequency

Selective Multi-Sensor System for Partial Discharge Detection on Medium

Voltage Cable Connectors. In: Tagungsband ETG-Fachbericht 162: VDE

Hochspannungstechnik, S. 294-299, Frankfurt, 2020, ISBN 978-3-8007-5353-

• B. Böttcher; A. Sinai; M. Menge; R. Plath; T. Gräf; T. Hücker: Operational

Risk Evaluation Of Cable Plugs Using An Automated Multisensor Classi-

fication System. In: VDE-Hochspannungstechnik 2018; Berlin, S. 714-720,

VDE, Frankfurt, 2018, ISBN 978-3-8007-4807-5.

• B. Böttcher; A. Sinai; M. Menge; T. Gräf; R. Plath; T. Hücker; P. True: Algo-

rithms for Partial Discharge Monitoring of Medium Voltage Cable Plugs,

Using a Multi-Sensor Expert System. In: Tagungsband ETG-Fachbericht

162: VDE Hochspannungstechnik, S. 85-90, Frankfurt, 2020, ISBN 978-3-

8007-5353-6.

• B. Böttcher; A. Sinai; M. Menge; T. Gräf; R. Plath; T. Hücker: Algorithms

for a Multi Sensor Partial Discharge Expert System Applied to Medium

Voltage Cable Connectors. In: 2019 2nd International Conference on High

Voltage Engineering and Power Systems (ICHVEPS), S. 190-195, -, 2020,

ISBN 978-1-7281-2669-2.

vii

Statement of Contributions:

The author of this thesis and Bjoern Boettcher collaborated on the NikMonET

project at the University of Applied Sciences HTW-Berlin. The author primar-

ily focused on studying and developing the sensors, filtering and amplification

hardware, and assembling the entire measurement system. Additionally, the

author was responsible for developing a MATLAB user interface for partial dis-

charge (PD) signal processing and visualization. Bjoern Boettcher was respon-

sible for developing the software for the LPC-Link2boards and Raspberry Pi,

as well as the PC software for analyzing PD signal quantities and classifying

PD signals.

IEEE Copyright:

In reference to IEEE copyrighted material which is used with permission in

this thesis, the IEEE does not endorse any of TU Berlin’s products or services.

Internal or personal use of this material is permitted. If interested in reprint-

ing republishing IEEE copyrighted material for advertising or promotional pur-

poses or for creating new collective works for resale or redistribution, please

go to http://www.ieee.org/publications_standards/publications/rights/

rights_link.html to learn how to obtain a License from Rights Link.

viii

CO NTENTS

Acronyms xi

Symbols xiii

1 introduction 1

2 theoretical foundations 5

2.1Theory of Partial Discharge . . . . . . . . . . . . . . . . . . . . . . . 5

2.2PDSignalQuantities .......................... 12

2.3Conventional PD Measurement Method . . . . . . . . . . . . . . . 15

2.4Non-Conventional PD Measurement Methods . . . . . . . . . . . . 16

2.4.1HFCTSensor........................... 18

2.4.2CapacitiveSensor ........................ 20

2.4.3AcousticSensor.......................... 22

2.5Medium Voltage Electrical Equipment . . . . . . . . . . . . . . . . 23

2.5.1Medium Voltage Switchgear . . . . . . . . . . . . . . . . . . 24

2.5.2Medium Voltage XLPE Cable . . . . . . . . . . . . . . . . . . 25

2.5.3Medium Voltage Cable Connector . . . . . . . . . . . . . . . 25

2.6State-of-the-art PD Detection Techniques on

MVCableConnectors.......................... 27

2.6.1IEC 60270 PDDetection..................... 27

2.6.2HFCTSensor........................... 28

2.6.3Electromagnetic PD Detection . . . . . . . . . . . . . . . . . 29

2.6.4Acoustic PD Detection . . . . . . . . . . . . . . . . . . . . . . 29

2.6.5Capacitive PD Detection . . . . . . . . . . . . . . . . . . . . . 29

2.6.6OhmicDetection......................... 30

2.6.7Summary ............................. 30

2.7TermsandDefinitions.......................... 31

2.7.1CAL2B – Pulse Generator . . . . . . . . . . . . . . . . . . . . 31

2.7.2S-Parameters ........................... 32

2.7.3SNR ................................ 34

2.7.4PRPD ............................... 34

3 evaluation of sensors 36

3.1DescriptionofSensors.......................... 36

3.1.1HFCTSensor........................... 36

3.1.2Electromagnetic PD Detection Using Antennas . . . . . . . 43

3.1.3CoaxialSensor .......................... 49

3.1.4Switchgear’s Bushing . . . . . . . . . . . . . . . . . . . . . . 53

3.1.5OhmicSensor........................... 55

3.1.6AcousticSensor.......................... 56

xcontents

3.2Measurement Results (of All Sensors) . . . . . . . . . . . . . . . . . 62

3.2.1PD-Afflicted Cable Connector and Measurement Setup . . 63

3.2.2Sensitivity of the Sensors . . . . . . . . . . . . . . . . . . . . 65

3.2.3Noise Sensitivity of the Sensors - SNR . . . . . . . . . . . . 68

3.2.4Cross-Coupling.......................... 69

3.2.5Price and Installation Effort . . . . . . . . . . . . . . . . . . . 71

3.3Discussion of Measurement Results and Conclusion of Chapter . 72

4 high-frequency characteristics of the mv-cable 75

4.1Connection interface between MV cable and measurement device 75

4.1.1Characterization of the developed adapters . . . . . . . . . 77

4.2Cut-off frequency of 10 m MV XLPE cable . . . . . . . . . . . . . . 78

4.3ConclusionofChapter ......................... 79

5 pd monitoring system 80

5.1The Principle of the Frequency-Selective

Multi-Sensor PD Measurement . . . . . . . . . . . . . . . . . . . . 80

5.2Effect of Signal Filtering . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.3The Principle of the Logarithmic Peak Detector . . . . . . . . . . . 83

5.4The Principle of the Developed PD Measurement System . . . . . 84

5.5Practical Investigations . . . . . . . . . . . . . . . . . . . . . . . . . 86

5.5.1Measurement Results of HFCT . . . . . . . . . . . . . . . . . 87

5.5.2Measurement Results of the Coaxial Sensor . . . . . . . . . 88

5.5.3Measurement Results of the Ohmic Sensor . . . . . . . . . . 90

5.6ConclusionofChapter ......................... 91

6 multi-sensor signal analysis 92

6.1MeasurementSetup........................... 92

6.1.1First Noise Source:

Frequency-Converter-Controlled Motor . . . . . . . . . . . . 93

6.1.2Second Noise Source: Rectangular Burst Signal . . . . . . . 94

6.1.3Third Noise Source: Corona Discharges . . . . . . . . . . . 95

6.1.4SpectralOverview ........................ 96

6.2LaboratoryTests............................. 97

6.2.1Multi-Sensor Measurements of

PD-afflicted Cable Connectors . . . . . . . . . . . . . . . . . 97

6.2.2Multi-Sensor Measurements of External Faults . . . . . . . 109

6.3System Test in High Power Test Laboratory CESI / IPH . . . . . . 120

6.4ConclusionofChapter .........................125

7 conclusion and outlook 127

Appendix 132

a design and measurement of ar-msa 134

Bibliography 140

ACRONYMS

Acronym Description

AC Alternating current

ADC Analog to digital converter

AE Acoustic emission

AR-MSA Annular-ring microstrip antenna

CD Coupling device

DC Direct current

DUT Device under test

EMI Electromagnetic interference

FCM Frequency-converter-controlled motor

FOS Fiber-acoustic optic sensors

GIL Gas-insulated line

GIS Gas-insulated switchgear

GTEM Gigahertz transverse electromagnetic

HF High-frequency

HFCT High-frequency current transformer

HV High-voltage

MV Medium-voltage

pC Pico coulomb

PDs Partial discharges

PE Polyethylene

PRPD Phase-resolved partial discharge

PRPS Phase-resolved pulse sequence

RBS Rectangular burst signal

RMS Root mean square

S-parameters Scattering parameters

SF6Sulfur hexafluoride

SMA Surface-mount assembly

xii Acronyms

Acronym Description

SNR Signal-to-noise ratio

TEV Transient earth voltage

UHF Ultrahigh-frequency

VHF Very high frequency

VNA Vector network analyzer

XLPE Cross-linked polyethylene

SYMBOLS

Symbol Description

CCapacitance

−→

EElectric field strength

∇Nabla operator

heff(f)Effective antenna height

AcThe cross-sectional area of the HFCT’ core

lsThe mean path of the HFCT’s core

ωAngular frequency

RResistance

tan δ Tan delta or dissipation factor

xiii

INTRODUCTION

The increasing demand for electrical power confronts grid operators world-

wide with new challenges. The decarbonized electricity generation leads to a

decentralized grid power feeding that increases the number of power network

branches. In addition to developing the assets, the grid operators have to use

the already existing (old) assets that have been in use for many years and are ex-

posed to demanding influences such as temperature, humidity, vibrations, etc.

In order to ensure a reliable power supply and prevent unpredictable break-

downs, the failures of the grid components have to be detected and controlled

at an early stage. Defects within the insulation system of medium-voltage (MV)

and high-voltage (HV)1equipment can, for example, lead to partial discharges

(PDs) before breakdown.

The installation of cable connectors, an essential interface between MV cable

and switchgear, is usually performed on-site. Time pressure or other undesir-

able circumstances during an installation can negatively impact the quality of

the cable connector assembly. Installation errors on MV cable connectors often

lead to PDs, resulting in possible equipment failure.

Partial discharges, caused by assembly errors, often appear immediately after

installation or in the early operating phase. A monitoring system that observes

a cable connector in its first operation days or weeks helps to identify faulty

connectors and prevent consequential damage. The conventional method for

PD detection – based on IEC 60270 standard [1] - requires an additional cou-

pling capacitor in the high-voltage setup, and is therefore costly. Under on-

site conditions, the low-frequency range, according to IEC 60270, leads to a

low attenuation of interfering signals even over long distances, which usually

reduces the PD measurement sensitivity considerably. These disadvantages

make the IEC method unattractive for PD monitoring and on-site PD measure-

ments in general. Alternative sensor-based approaches enable cost-effective

and sensitive monitoring. Moreover, most PD detection methods are affected

by interfering surrounding external PD signals, which prevent a correct diag-

nosis.

1MV is an electric voltage level in a range of 1–60 kV, while HV supports a voltage level higher than 60

kV.

2 introduction

This work aims to find a reliable monitoring approach for detecting PDs on

cable connectors under on-site conditions. This monitoring method fits the

cost structure of medium-voltage systems and does not impose changes to the

MV setup. Interference sources should only marginally influence the diagnos-

tic reliability of the approach.

Possible approaches include:

1.IEC 60270-based method: Monitoring of the cable connectors can be per-

formed with IEC 60270-compliant measuring devices. The disadvantages

have already been mentioned: Additional components in the primary cir-

cuit, cost, and sensitivity to interference.

2.High-frequency current transformer sensors: The high-frequency cur-

rent transformer (HFCT) sensor is often used for PD detection. Depend-

ing on the corresponding pass-band cut-off frequencies, PD signal mea-

surement with the HFCT sensor can be evaluated with IEC 60270 standard-

compliant measuring devices. The sensor is cost-effective and can be used

flexibly. Applying the HFCT sensor does not usually require any change

in the primary setup. However, electrical interference signals, for exam-

ple, from frequency converters or PD of other medium-voltage compo-

nents, can affect the result of the PD measurement on cable connectors

using HFCT. For example, the HFCT sensor can detect partial discharge

signals from adjacent cable connectors and joints. Thus, the localization

of a PD source is associated with additional efforts. Moreover, IEC 60270-

compliant measuring devices are often associated with high costs.

3.The use of HFCT or IEC 60270 compatible measurement equipment, in-

cluding additional interference signal suppression (i.e., analog and/or

digital filtering): As described, interference signals from external sources

– away from the cable connector – can affect reliable PD detection. The

sensor signals can be filtered with analog electronic circuits or digital fil-

tering methods to suppress the interfering signals. The filter’s frequency

band should be adapted to the sensor’s frequency range and the interfer-

ing signal’s frequency. The disadvantage is that the interference source is

often unknown in actual operation, and the frequency of the interfering

signals might be in the same range as that of the sensor. Contrary to ana-

log filter circuits, the filtering frequency of digital filters can be changed

during operation. Digital filtering methods, such as wavelet algorithms,

have been the subject of many research projects. Implementing such algo-

rithms requires considerable computing power and high sampling rates,

which generally contradict the aforementioned cost factor.

introduction 3

4.The use of sensors that are not affected by sources of electrical interfer-

ence: Sensors that detect electromagnetic emissions of partial discharges

can be disturbed by external electromagnetic interference sources, for ex-

ample, rotating machines, frequency converters, external corona, etc.

The use of, for example, acoustic emission (AE) or fiber optic sensors,

which are not influenced by electromagnetic interference signals, may im-

prove the noise sensitivity of the measurement. For example, acoustic PD

measurements are performed in transformers or gas-insulated switchgear

(GIS). Acoustic sensor signals in an industrial medium-voltage environ-

ment may be affected by acoustic interferences, for example, noise emis-

sion from wind turbines. To the author’s knowledge, the acoustic PD

detection on MV cable connectors has not been sufficiently investigated.

Acoustic PD measurement with fiber-acoustic optic sensors is a relatively

new approach. These sensors are not affected by electromagnetic interfer-

ences. The fiber-acoustic optic sensors can also be used in places with a

high electric field intensity because these sensors are made from insulat-

ing materials. Besides the cost factor, the disadvantage of acoustic fiber

optic sensors is sometimes their extremely high sensitivity to acoustic

emissions not related to internal PD from the test object.

5.The use of multiple sensors (multi-sensor approach): Simultaneous mea-

surement with different sensor types is advantageous because the interfer-

ence sources do not often disturb all the sensors in the same way. There-

fore, partial discharges may be better distinguished from interfering sig-

nals. A combination of different sensor types that respond to various PD

emissions, for example, electromagnetic and acoustic signals, can improve

the diagnostic reliability of the monitoring system and enable redundant

decision-making.

This work’s investigations

It has been detailed that the intended monitoring approach should enable

the realization of a cost-effective system. The PD diagnostics should not be

affected by interferences. In addition, changes in the primary MV setup are

not desired.

Different possible approaches are explored in this work, and their suitability,

considering the mentioned objectives, is investigated. Single-sensor and IEC

4 introduction

60270-based PD measurement methods are compared. In addition, the multi-

sensor PD detection approach, which seems to be promising, is examined.

In a multi-sensor PD measurement system, if different sensors confirm the ex-

istence of the PDs, a more reliable statement about the local PD’s existence can

be made. If different sensors measure in various frequency bands (frequency-

selective), the immunity of measurements against narrowband disturbances

will be increased.

To the author’s knowledge, frequency-selective multi-sensor PD detection on

cable connectors has not been yet adequately investigated. This approach com-

prises sensors, analog pre-processing, sampling and data digitalization, digital

pre-processing, storage, and data analysis. In this work, the measurement sys-

tem components are discussed and explained.

After presenting the theoretical backgrounds in Chapter 2, nine different

electromagnetic and acoustic sensors, including two self-developed sensors,

are investigated in Chapter 3, and suitable sensor combinations are evaluated.

Implementing the frequency-selective approach using bandpass filters as-

sumes the knowledge of the (noise) signal propagation over the MV cable.

Hence, in Chapter 4, the MV cable’s high-frequency (HF) signal analysis is pre-

sented.

In Chapter 5, the PD analog signal processing is explained, and the system

architecture of the self-developed PD measurement system is discussed.

Chapter 6deals with the signal analysis of the PD-afflicted cable connec-

tors and external PD sources. Comparative measurements are performed with

different faulty cable connectors, external discharges, and in the presence of in-

terferences. Employing the performed analyses, the suitability of the proposed

approach for identifying partial discharges in cable connectors is finally dis-

cussed. The performance of the developed PD measurement system is tested

in the high-power test laboratory CESI / IPH Berlin.

The conclusion of this work and the outlook for future works are presented

in Chapter 7.

THEORETICAL FOUNDATIONS

This chapter presents the background of PDs and the associated sensing meth-

ods. Furthermore, the MV switchgear, cable, and cable connector are reviewed.

At the end of this chapter, the state of the knowledge for PD detection on the

cable connectors is discussed.

2.1 theory of partial discharge

Insulating materials support conductors mechanically and insulate the conduc-

tors electrically from each other and the ground potential. The voltage differ-

ence between various conductors creates an electric field in the corresponding

insulation materials, which could be gaseous, liquid, or solid. To ensure the

longtime device operation, the electric field strength must be less than a specific

limit, i.e., the long-term insulation material’s breakdown electric field strength.

For example, gas-filled voids or impurities in solid insulation material can

lead to a higher electric field strength than desired over these stress regions [2].

The electric field applies a force on atoms and their electrons. The electrons

may leave the atom structure if the critical electric field strength is reached. In

this case, the atom is ionized, and a free electron is available in the related ma-

terial structure. The electric field accelerates the free electron that may collide

with other atoms, i.e., pre-discharge, which creates more ions and free elec-

trons. The free electrons may generate further pre-discharges and an avalanche

effect inside the ionized area.

The high electric field generates pre-discharges occurring in the stress re-

gion partially and not between the conductors or ground lines. Hence, these

discharges are called partial discharges. A high electric field intensity also

appears on strongly curved edges of the conductors and may cause partial dis-

charges and an insulation breakdown.

A discharge releases energy in different forms, like electromagnetic waves

– from low to high-frequency ranges – and chemicals. Furthermore, transient

electric waves are generated over the conductors. The mentioned released en-

ergies and waves are explained in detail in the following subsections.

6 theoretical foundations

The generation of electrical transients

The creation of the electrical transients can be explained by considering two

parallel plates with different charges and an insulation material with an air-

filled void (Figure 2-1a). In a pre-discharge equilibrium, the ionization process

in the void has still not started. In this case, the positive and negative charges of

the two plates are spread along the plates creating the voltage V0. As described

earlier, the established electric field has a higher intensity in the stress region,

whereas the ionization process begins by reaching critical field strength values.

In the post-discharge equilibrium state (Figure 2-1b), the created positive ions

and the free electrons are spread on the void surface so that the negative void

charges are closer to the positively charged plate. The positive void charges

are located on the opposite side.

In the post-discharge equilibrium, the positive and negative plate charges are

not spread homogeneously when compared to the pre-discharge equilibrium.

The non-homogeneous charge distribution is due to the void’s space charges

that attract the opposite charges of the plates. An avalanche breakdown in

the void creates many space charges and leads to regions with higher charge

densities on the plates. It seems that partial discharges induce charges into

the plates and in a specific area. As the result of non-homogeneous charge

distribution on the plates, a potential difference, i.e., transient voltage, is gen-

erated in each plate [3]. The voltage source acts as a charge source and tries to

compensate and homogenize the charge distribution by pumping new charges

into the plates – the so-called apparent charges.

(a) (b)

Figure 2-1: (a) The pre- (b) post-discharge equilibrium on two parallel plates with a solid insu-

lator having an air-filled void [3].

The measurement of the electrical transients is a well-known approach

for detecting partial discharges in a high-voltage component. Mathematical

considerations regarding creating electrical transients due to PD can be found

in [4,5].

2.1 theory of partial discharge 7

The generation of light

The collision of accelerated free electrons with atoms can create more ions

and free electrons inside the ionized area. It could also be that the collision

leads to an excited atom. This state occurs when the atom electrons receive

the external energy and jump from the lower atom orbits to higher ones. If

the electrons of the excited atoms or ions jump back to their original states

(lower orbits), a photon of light is emitted. Further information regarding the

creation of photons can be found in the quantum physics literature or chapter

6of [4].

The generation of sound

The generation of partial discharges leads to the release of energy, which

heats the surrounding material. The generated heat causes an expansion of

surrounding material, for example, gas [5]. As a result of the expansion, a

sudden volume change creates sound waves [6]. In solid materials, thermal

stress can cause cracks which can also be acoustically detected. Further

fundamental details regarding acoustic PD detection are well presented in [7].

The generation of electromagnetic waves

Around 1888, Heinrich Hertz proved Maxwell’s electromagnetic theory

using a so-called Hertz’s spark transmitter [8]. Hertz has shown that electro-

magnetic waves are generated due to electric discharges. Figure 2-2illustrates

the measurement setup used by Hertz. The transmitter includes an induction

coil and a capacitor, which generate an alternating high voltage. The AC

voltage charges the sphere capacitor and creates a spark (electrical discharge).

The electrical discharge creates electromagnetic waves, which travel to the

receiving wire loop. The receiving wire loop also concludes a gap. The

electromagnetic wave couples to the wire loop and induces a current loop,

leading to discharges in the spark gap.

Different types of partial discharges

As described, the insulation material of HV components could be gaseous,

liquid, or solid. If the electric field strength is larger than the dielectric strength

of the insulation material in a specific area, partial discharges can be observed

in this region. Such PDs can be categorized as internal, external, or surface

discharges [10].

8 theoretical foundations

Figure 2-2: The Overview of the measurement setup of Heinrich Hertz for the detection of

electromagnetic waves generated by electrical discharges [8,9].

Internal PDs occur within a solid or liquid insulating material and typically

at stress regions, concluding voids or material impurities. External discharges

happen in the air and outside an HV device or component containing gaseous,

liquid, or solid insulation material [11]. Surface discharges occur at the

interface of two insulating materials with different aggregation states, for

example, on the surface of a solid insulator and the surrounding gas or liquid

[12]. The mentioned PD categories are described in the following subsections.

Internal discharges

An insulation material, including a void, is illustrated in Figure 2-3a. The

insulation is placed between two conducting plates, while the upper and lower

conductors have a positive and negative voltage, respectively. The electric field

strength within the insulation material and void is simulated, and the result

is shown in Figure 2-3a. The electric field strength reaches its highest value

inside the void, leading to discharges.

The discussed example in Figure 2-3acan also be explained from the equiva-

lent circuit point of view (Figure 2-3b). The capacitor C0with the voltage v(t)

represents the capacitance between the upper and lower electrodes. The capac-

itances between the void and electrodes are illustrated as C21 and C22. The

2.1 theory of partial discharge 9

capacitance C2with the voltage v2(t)is the equivalent series capacitance of C21

and C22. The capacitance of the void – with the voltage v1(t)– is illustrated

as C1. The electric discharge current flows through the resistor R1- plasma

conductivity - and a spark gap, S[10].

(a) (b) (c)

Figure 2-3: (a) A simulated example of electric field strength distribution in an insulation ma-

terial, including a void leading to internal discharges. (b) The corresponding equiv-

alent circuit diagram. (c) The voltage breakdown over the void caused by the PD

[10].

The blue sine wave in Figure 2-3cshows the voltage v1(t)in case that no

discharge happens. The black line illustrates the case where discharges occur.

The increase of sine voltage v(t)leads to a rise in voltage v1(t)and a rise in

charge of capacitor C1. When the voltage reaches a critical value – the so-called

inception voltage – a discharge happens inside the void, and the capacitor C1

discharges. Hence, a voltage breakdown occurs over the capacitor C1. The

capacitor C0recharges the C1, and C2branch, leading to a voltage drop over

C0. The power supply charges all the capacitors, and the described procedure

is repeated continuously and as a function of v1(t).

The electric field of the void is equal to the negative gradient of the void’s

voltage, represented by −→

E= −∇v1(t). Therefore, the absolute maximum of

the electric field is around 0°, 180°, and 360°, where the discharges appear.

External corona discharges

External discharges happen in the air and outside of HV devices or com-

ponents. For example, corona discharges occur at strongly curved or sharp

metallic edges of HV- or associated ground conductors.

10 theoretical foundations

As illustrated in Figure 2-4a, a sharp edge with HV potential leads to an

inhomogeneous electric field with a high electric field strength around the

corresponding tip. After the breakdown electric field strength of the gas has

been reached, the ionization process starts leading to free space charges. As

a result of the alternating voltage, the time-variant electric field accelerates

free space charges periodically with alternating direction. A periodic current

flow is created, which heats the affected area [12]. Further voltage increase

creates a conductive channel between the electrodes and causes an insulation

breakdown.

The corresponding equivalent circuit diagram of the peak on HV is shown in

Figure 2-4b.C1represents the capacitance of the discharge path. A conductive

path is built between the electrodes when the discharge occurs. Current flows

through the resistor R1, the plasma conductivity, and the discharge gap S.C2

represents the capacitance between the HV and ground electrodes [10].

Generally, it can be assumed that R1≫1/(ωC1)and the current i1=v(t)/R1

is almost just defined by the resistor R1. Hence, as shown in Figure 2-4c, the

voltage v1(t)over the capacitance C1lags behind the proof voltage v(t)by 90°

[12]. According to the equation −→

E= −∇v1(t), the maximum of the electric

field strength over C1is around the zero crossing points of v1(t)corresponding

90° and 270° of v(t).

(a) (b) (c)

Figure 2-4: (a) A simulated example of electric field strength distribution around a peak on HV

potential leading to external discharges. (b) The corresponding equivalent circuit

diagram. (c) The voltage breakdown caused by the PD [12].

If the sharp edge is connected to the HV, more electrons can additionally

join the ionized region, plasma, during the negative half cycle of the AC

voltage. As described, around the sine angle of 270°, the electric field strength

reaches its peak and starts the discharge process. Further voltage increase

2.1 theory of partial discharge 11

leads to the appearance of additional corona discharges around the maximum

positive voltage amplitude, sine angle of 90°.

If the sharp edge is on the ground side, additional electrons can join

the plasma from the ground side during the positive half cycle of the AC

voltage. The electric field strength reaches its maximum around 90°, where

the discharge activity starts. A further voltage increase generates additional

discharges around the sine angle of 270°.

Surface discharges

The surface discharges occur at the interface of two insulating materials with

different aggregation states, for example, on the surface of solid insulators

in gases or liquids [12]. This kind of discharge can mainly be found on the

polluted surface of HV equipment, for example, insulators, bushings, cable

terminations, and insulating housings [11]. A detailed study considering the

creation of surface discharges and corresponding prevention methods can be

found in [11,12].

Figure 2-5shows the cross-section view of an HV cable, where the insulation

material and the screen are partially removed, for example, for installing a

cable connector. The cross-linked polyethylene (XLPE) insulation material and

the surrounding air isolate the cable’s inner conductor from the cable screen.

If the cable is operated in this condition without a proper cable termination,

discharges on the surface of the HV cable insulating material may happen.

Figure 2-5: The cross-section view model of an HV-cable (where the insulation material and

the screen are partially removed) and the corresponding simulated electric field

strength. The associated AC equivalent circuit model considers external particles

on the insulation material [12].

12 theoretical foundations

In Figure 2-5, the simulated electric field strength of the prepared HV cable is

also shown. The electric field strength is higher on the sharp edges of the cable

screen. The tangential component of the electric field parallel to the insulation

surface applies forces on the free electrons and the positive ions on the insula-

tion surface. Hence, the discharges usually begin from the sharp edge of the

screen, which contacts the insulation material. The dielectric strength of the

insulation’s surface is less than the surrounding air causing surface discharges.

Figure 2-5also details the equivalent circuit parameter of the illustrated cable

model with an AC voltage [11]. This circuit represents the case when external

substances, for example, water due to humidity, are on the surface of the in-

sulation material, increasing the conductivity and the discharge current flow.

The equivalent circuit assumes that the insulation surface can be divided into

infinitesimal distances with the length of △x. The △Crepresent the capaci-

tance of the insulation material between the electrodes, while the resistors △R

demonstrate the surface conductivity. The capacitors △Csillustrate the stray

capacitances, which are negligible in this example [11].

2.2 pd signal quantities

It has been shown that PDs depend on the applied voltage and can occur at

different phase positions. Partial discharges can be detected using different

measurement techniques. For example, the transient signals generated at the

terminals of an HV device can be evaluated using a current transformer sensor.

The corresponding electromagnetic emissions can be detected with antennas,

while acoustic or optical emissions can be captured using related sensors.

Depending on the measuring method, applied voltage, and PD fault type,

different signal characteristics can be observed in the time- and frequency

domain with various statistical behavior. These signal characteristics are used

to detect and analyze partial discharges. Some of the most important PD

signal quantities used in this thesis are explained in the following subsections.

Discharge amplitude

PDs lead to minor electric potential differences on the terminals of an HV

device. Consequently, a transient current can be observed that flows over the

connected cables or lines, recharging the HV device. The so-called apparent

charge can be calculated by integrating the transient current over time.

2.2 pd signal quantities 13

The apparent charge is defined in the international standard IEC 60270

[1] as: “. . . that charge which, if injected within a very short time between the

terminals of the test object in a specified test circuit, would give the same reading on

the measuring instrument as the PD current pulse itself”.

The apparent charge is proportional, but generally not equal to the charge

amount displaced in the stress zone [13].

Figure 2-6exemplary illustrates a PD sequence over time. The amplitude of

the apparent charge is shown with “qi”, for which the unit pico coulomb (pC)

is commonly used.

Figure 2-6: The exemplary illustration of PD signals with their apparent charge (qi), the cor-

responding phase (φi) and time of occurrence (ti), and the instantaneous voltage

value (vi) [14].

Time of occurrence

The occurrence of PDs in a stress zone depends on the electric field strength

modulated by the applied sine voltage in an AC system. PDs occur in a time

frame where the electric field strength reaches a critical value. Therefore,

measuring PD amplitude together with the associated time of occurrence can

be helpful for PD evaluation. As the electric field strength value changes peri-

odically, the sine function’s phase angle (φi) is often used as a correspondent

to the equivalent time of occurrence (ti) - Figure 2-6.

14 theoretical foundations

Instantaneous voltage value

In Figure 2-6, the parameter v(ti)represents the instantaneous voltage value

of the reference sine wave voltage in which the PDs happen.

Repetition rate

The repetition rate is the number of acquired PD signal pulses within a

specific time range. Repetition rate and time of PD occurrence are especially

interesting parameters for partial discharge characterization in HV direct

current (DC) applications, where the phase angle of the PD signal cannot be

concluded from the reference voltage [15].

Inception and extinction voltage

The inception and extinction voltages (Viand Ve) represent the root mean

square (RMS) values of the applied sine voltage on the HV device terminals

where the measurable PD signals start or end. For HV DC applications, the

voltage amplitude is considered.

Frequency range

Figure 2-7illustrates a PD signal of a faulty MV cable connector and the

corresponding spectrum measured with a high-frequency current transformer

(HFCT). As shown in Figure 2-7a, this PD signal reaches a maximum absolute

value of around 100 mV, and the PD signal duration is about 5µs. The

corresponding frequency spectrum of the PD signal is shown in Figure 2-7b.

The spectrum has the highest amplitudes at 1.9MHz and 4MHz. Higher

frequency components, up to 40 MHz, have smaller amplitudes. It should be

noted that the illustrated PD signal shape and the corresponding spectrum are

just representative examples.

Depending on the HV equipment, insulation medium, fault type, measure-

ment setup, etc., the PD signals may have frequency components, even in the

GHz range. For example, the partial discharges in gas-insulated switchgear

(GIS) or a power transformer possess frequency components up to 3GHz [16].

The measured PD signal amplitudes vary depending on the measurement

method, propagation path, etc., which can attenuate the signal amplitudes or

even amplify them by possible resonances.

2.3 conventional pd measurement method 15

(a) (b)

Figure 2-7: (a) A partial discharge signal of a PD-afflicted MV cable connector measured with

HFCT and (b) the corresponding spectrum.

Considering the discharge spectrum can be beneficial. For example, when

selecting sensitive sensors, distinguishing certain discharge types, evaluating

interference signals, etc.

The definition of additional PD signal quantities, i.e., average discharge cur-

rent, discharge power, and the quadratic rate, is not in the scope of this work

and can be found in [1,11,13].

2.3 conventional pd measurement method

The standard method for measuring partial discharges is described in IEC

60270. In most cases, this measurement method is used and is therefore called:

the conventional PD measurement technique.

The basic setup for the measurement of the apparent charge, according

to IEC 60270, is illustrated in Figure 2-8[1]. The test object having the

capacitance Cais connected via a current limiting impedance Zto a voltage

source. A coupling capacitor Ckis connected in parallel to the device under

test (DUT). In the case of PD occurrence, the coupling capacitor can recharge

the DUT rapidly. This recharging current can consequently be measured with

a coupling device (CD) connected to the ground path of Ckor DUT’s ground

path.

According to IEC 60270, the conventional PD measurement can be per-

formed by integrating the discharge current in the time domain. Detailed

explanations of the mentioned measurement principle are presented in [1,11,

13].

16 theoretical foundations

Figure 2-8: Basic PD measurement setup according to IEC 60270 with a coupling device con-

nected to the coupling capacitor [1].

Apparent charge measurements can be performed wideband or narrowband.

In a wideband measurement, the lower cut-off frequency can be chosen from

30 kHz to 100 kHz, whereas the upper cut-off frequency should be less than

or equal to 500 kHz. The bandwidth should be between 100 kHz and 400 kHz.

For narrowband measurements, a bandwidth between 9kHz and 30 kHz is

required. The center frequency shall be between 50 kHz and 1MHz [1].

The PD signals of the DUT are not attenuated significantly by the measure-

ment setup in the discussed kilohertz frequency ranges. However, the draw-

back is that the PD signals emitted from surrounding devices or overhead

lines within a radius of a few 100 meters are also not damped significantly

and cannot be distinguished from the DUT’s PD signals. This indicates that

the conventional PD measurement method is more suitable for laboratory or

factory applications than for permanent onsite monitoring in an unshielded

environment with a harsh noise situation.

2.4 non-conventional pd measurement methods

Non-conventional PD detection methods try to overcome the noise sensitivity

issue or the need for a coupling capacitor of the IEC 60270-based measurement

method by using antennas, acoustic, optical, or other sensors to detect the

PD signals. The non-conventional PD measurements can not be calibrated in

pC according to IEC 60270 standard but may detect PD more sensitively or

with less effort in a given environment. Depending on the device under test,

different sensing methods can be implemented for PD detection. Recommen-

dations considering acoustic and electromagnetic PD detection techniques are

provided in IEC 62478 [17].

2.4 non-conventional pd measurement methods 17

Acoustic PD signals can be best detected in a frequency range between

10 Hz to 300 kHz with acoustic emission sensors [17]. Acoustic PD detection

is frequently used for GIS and transformers. This measurement method can

also be used to localize the PD source.

The electromagnetic PD detection in transformers, GIS, and gas-insulated

lines (GIL) in the ultrahigh-frequency (UHF) range is common practice. The

frequency range utilized for UHF measurements is mainly between 300 MHz

and 3GHz [17].

As described earlier, the electrical discharges lead to the emission of

photons, and the generated light waves can be measured with the associated

sensors. Light waves with wavelengths from infrared to ultraviolet are emitted

depending on the insulation material around the PD stress zone [13]. Using

infrared and ultraviolet cameras, and having a free measurement view, the hot

spots and corona discharges of the HV assets can be localized, respectively

[18]. Furthermore, the PD light emissions can be detected with optical

fibers. As the fibers do not contain metals, they can be embedded into the

HV equipment, for example, transformers [18]. One of the most important

advantages of this PD detection principle is its immunity to electromagnetic

interference (EMI), which makes this sensor suitable for harsh electromagnetic

environments. Furthermore, the test devices and the operators are safe against

possible discharges over measurement cables, which can be hazardous to them.

The generation of heat and ultraviolet radiation resulting from discharges

leads to chemical changes in the surrounding insulation material. Further-

more, creating gases like ozone, hydrogen, nitrogen oxides, etc., indicates the

PD’s existence in the corresponding insulation material providing a possibility

for chemical PD detection [13].

PD detection using non-conventional methods has been the research subject

for many decades. A detailed description of all these measurement principles

is not in the scope of this work. In this thesis, several PD detection methods are

investigated so that a cost-effective PD monitoring approach on MV cable con-

nectors can be realized. The primary studies of this work have demonstrated

that integrating optical fibers in the cable connector is highly challenging and

cannot fulfill the objectives of a cost-effective monitoring device. Due to the

compact and closed structure of the cable connector, chemical PD detection

is not promising. Other non-conventional PD sensing methods, for example,

HFCT, capacitive, acoustic, and electromagnetic PD detection methods, are

studied in detail.

18 theoretical foundations

2.4.1HFCT Sensor

HFCT sensors are current transformers, usually having a broadband signal re-

sponse (tens of MHz). Figure 2-9illustrates the structure of the HFCT sensor. It

consists of a ferrite core, windings, and usually a BNC connector. The primary

current-carrying conductor with current I1generates the magnetic flux (ΦB)

in the ferrite core. The magnetic flux induces the current I2in the secondary

windings causing the voltage drop (V) over the measurement impedance (R).

Figure 2-9: The HFCT sensor’s structure, the concentrated magnetic flux in the ferrite core, and

the induced current in the sensor’s secondary windings.

The functionality of a current transformer, the corresponding sensitivity and

frequency response, and the associated signal losses depend on the sensor’s

structure and the material of the magnetic core. The discussion about the

losses requires a short review of the magnetic materials, which can be divided

into two categories, i.e., soft and hard magnetic materials.

Soft magnetic materials are used in various applications, such as power

electronics, high-frequency applications, EMI filters, etc. [19]. Soft magnetic

materials can be made from different materials and alloys. Ferrites are soft

magnetic materials that are suitable for high-frequency applications. Mn-Zn

and Ni-Zn are the most common type of ferrite materials. The main difference

between these two materials is caused by the corresponding frequency-

dependent losses [20]. The Mn-Zn material is used for applications with a

frequency range of up to around 30 MHz. In contrast, the Ni-Zn material is

suitable for higher frequencies up to some hundred megahertz [20].

The current transformer losses can be divided into core and winding losses

described in the following [19].

2.4 non-conventional pd measurement methods 19

Winding losses

• The windings wire has a DC resistance, which should be considered. The

DC resistance loss of the wire leads to heat generation in the wires [19].

• With increasing frequency, the current in the wire tends to flow near the

surface of the wire. This effect is known as the skin effect [19]. The skin

effect introduces further wire losses in higher frequency ranges.

• The magnetic coupling between adjacent windings, which carry AC,

causes a further loss factor known as the proximity effect. The AC of one

wire produces a circulating current in the neighboring wire, also known

as the eddy current. The eddy current leads to an increase in losses [19].

• The parasitic capacitance between the windings and the core leads to

frequency-dependent losses. Furthermore, the capacitances between the

turns of the windings should be considered. The mutual capacitance

is the third parasitic capacitance between the core’s two windings that

should also be considered [19].

Core losses

• The magnetic field of the primary windings generates a magnetic flux in

the core. The formation of the magnetic flux requires the orientation of

the tiny magnetic dipoles of the core. Depending on the core material, the

formation of the dipoles requires different magnetic moments. Hence, var-

ious core materials have different core losses, known as hysteresis losses

[21].

• The magnetic field produces eddy currents in the magnet core, as well.

These currents heat the ferrite core and cause further losses. The amount

of the produced eddy current in the core depends on the core material’s

electrical resistivity [19].

The described current transformer’s losses are frequency-dependent, mak-

ing designing and verifying the transformer in higher frequencies challenging.

Numerous literature has been published, dealing with the current transform-

ers’ characteristics and functionality. Some of these interesting publications

are addressed in the following paragraph.

A detailed insight into the properties of the transformer and the corre-

sponding physical and mathematical descriptions are given in the book of

W. G. Hurley and W. H. Wölfle [19]. In [22], the HFCT simulation model

20 theoretical foundations

is studied, and different sensor structures are simulated. It is shown that

a higher number of sensor windings leads to a smaller output amplitude

and bandwidth. Similar results based on practical experiments are gained

in [23], where wires with 1mm and 0.8mm diameter and various numbers

of windings (35 and 40) in the structure of HFCT are studied. Defining

the transfer function of the HFCT based on time- and frequency-domain

measurements are proposed in [24]. A numerical calculation based on a 3D

FEM simulation method for a shielded HFCT is presented and validated by

measurements in [25]. In [26], it is shown that using an HFCT sensor for PD

detection in higher frequency ranges, i.e., up to 1GHz, helps suppress the

noise signals and localize the PD detection. In [27] and [28], the effect of an

air gap within the sensor’s ferrite core is studied. It is detailed that an air gap

can improve the saturation characteristics of the sensor if HFCT is mounted

over MV and HV cables carrying currents up to hundreds of amperes. The

calibration of the HFCT and the associated PD apparent charge calculation are

investigated in [29].

The presented HFCT background reveals that the sensor’s structure signif-

icantly impacts the sensor’s output signal. The design and calibration of an

HFCT sensor with which the actual PD transient wave shape can be captured

are very challenging. Nevertheless, the PD’s existence in a monitoring system

can be reliably approved using this sensor. The existing commercial sensors,

however, cannot satisfy the cost expectations in the MV sector. Based on the

discussed publications results, a complementary study to find a straightfor-

ward sensor design that is simple to build and cost-effective is presented in

Chapter 3.

2.4.2Capacitive Sensor

The capacitive sensor enables the measurement of PD signals without direct

galvanic contact with the HV conductor. An example of a capacitive PD

measurement method in cable joints is proposed in [30]. A copper electrode

is placed upon the HV cable’s semiconducting layer (Figure 2-10a), and the

exterior insulating material surrounds the copper electrode. The capacitances

C1(between the HV-cable inner conductor and the copper electrode) and C2

(between the copper electrode and the measurement connector) enable the

capacitive PD measurement.

The basic equivalent circuit diagram of the capacitive sensor [30] is illustrated

in Figure 2-10b. The parallel circuit of the resistor Rand the capacitance C2

represents the electric characteristics of the exterior insulating material. In

2.4 non-conventional pd measurement methods 21

[30], the basic equivalent circuit is developed further by considering the joint

geometry and using the lossy transmission line model. For simplicity, the

complementary circuit is not discussed here. The capacitive PD measurement

on cable joints is presented in other research results, for example, in [31–34].

(a) (b)

Figure 2-10: (a) The proposed structure of a capacitive sensor for PD detection in cable joints

and (b) the associated equivalent circuit model [30].

The capacitive PD measurement approach is also applied in a transient earth

voltage (TEV) sensor [35]. This sensor is used for PD detection in metal-

enclosed equipment. In [36], the physical principle of a TEV sensor is well

presented. As illustrated in Figure 2-11 the sensor consists of a metal electrode

placed on a dielectric material, creating a capacitance to the metal wall of the

device. The electromagnetic PD signals within the equipment, for example,

switchgear, generate surface currents in its metal wall. The described capac-

itance between the metal plates, together with the capacitance of the coaxial

cable, enables the measurement of the PD current excited in the metal wall.

Figure 2-11: The general principle of a TEV sensor [36].

22 theoretical foundations

2.4.3Acoustic Sensor

The advantages of acoustic PD measurement are the relative immunity to EMI

and the possibility of locating PD in transformers and GIS. However, like

the other non-conventional measurement methods, calibrating the acoustic

sensors to the PD apparent charge is not possible [37].

For the investigation of acoustic PD measurements in the thesis, AE sensors

are used, described briefly in the following paragraphs.

Figure 2-12aillustrates the piezoelectric effect utilized in the used AE sensor.

A cylindrical piezoelectric material is placed between two metallic electrodes.

The piezoelectric material could be quartz or a piezoelectric ceramic [38]. The

mechanical stress over a piezoelectric material leads to the displacement of its

internal electrical charges. Hence, the material is no longer electrically neutral,

and the two electrodes have an electrical potential difference. The potential

difference is an electric voltage that can be measured.

The general structure of a commercial AE sensor is shown in Figure 2-12b.

The detection face of the sensor is connected to the metallic electrode of the

piezoelectric element. Depending on the sensor type, a damper material can be

placed on top of the piezoelectric material. The use of a damper material leads

to the broadband frequency response of the sensor. In contrast to the broad-

band AE sensor, the resonance-type sensors do not have any damper material

and can detect acoustic waves at a specific frequency. Further details regarding

the differences between the two mentioned sensor types can be found in [38],

page 37.

(a) (b)

Figure 2-12: (a) Principle of the piezoelectric effect and the generation of electric voltage due

to mechanical stress on a piezoelectric material. (b) The structure of a resonant

and broadband-type AE sensor [38].

2.5 medium voltage electrical equipment 23

The acoustic PD signals generated from internal discharges usually have

very small amplitudes. Hence, there is a need for an amplifier to boost the

sensed signals. The amplifier can be connected externally to the AE sensor or

integrated into the package.

Figure 2-13 illustrates one acoustic PD event versus the corresponding elec-

tric signal detected by an HFCT sensor. The used acoustic sensor has an inte-

grated amplifier with 46 dB gain - VALLEN VS30 [39]. The HFCT detects the

PD signal with an amplitude of around 20 mV, and the corresponding time

duration is about 10 µs. The acoustic sensor detects the corresponding PD

acoustic signal with a time difference of about 10 µs. The amplitude of the

amplified acoustic PD signal reaches absolute values of around 250 mV and

has a time duration of approximately 500 µs.

Figure 2-13: The measurement of a PD signal with an acoustic sensor and an HFCT.

In this section, three non-conventional PD detection methods were explained

that are the focus of this thesis. The introduction of all other existing PD-

capturing approaches is not within the scope of this work.

2.5 medium voltage electrical equipment

In this work, a PD monitoring approach, including different sensing methods,

is studied with which the cable connectors of medium voltage switchgear can

be monitored. To better understand this application, the MV switchgear, cable,

and cable connectors are briefly explained in the following sections.

24 theoretical foundations

2.5.1Medium Voltage Switchgear

The MV switchgear is an essential asset in the MV power grids for switching,

controlling, metering, etc., the electrical power. Power transmission and distri-

bution systems and medium to large commercial and industrial buildings use

medium-voltage switchgear. Such switchgear is a centralized system of circuit

breakers, fuses, and switches used to safeguard, regulate, and isolate electrical

equipment.

An example of medium voltage switchgear is shown in Figure 2-14. The

illustrated 24 kV/630 A switchgear is MINEX from Driescher Wegberg GmbH

[40]. In this indoor switchgear, sulfur hexafluoride (SF6) gas is used as

the insulating material. The switchgear possesses three cable connection

compartments. In a typical application, the electric power flows from the

transformer connection compartment (field) and fuses over the switchgear’s

busbars. The power is distributed toward loads over MV cables connected to

the switchgear on the cable connection fields. The cable connectors are used

to connect the MV cable to the switchgear.

Figure 2-14: The MINEX 24 kV/630 A switchgear from Fritz Driescher KG1.

1Fritz Driescher KG allowed the author to reuse this figure in this work.

2.5 medium voltage electrical equipment 25

2.5.2Medium Voltage XLPE Cable

An example of a XLPE MV cable structure is illustrated in Figure 2-15 [41]. The

conductor is made of aluminum wires twisted together and covered with the

inner semiconducting layer. The XLPE insulation material has high dielectric

strength and is covered with an outer semiconducting layer on the ground

side. The semiconducting layers ensure a uniform radial symmetric electric

field in the insulation material. The conductive tape contributes to shaping the

copper wire screen. The cable shield is surrounded by a water-blocking tape

and a polyethylene (PE) outer sheath.

The shown 12/20 kV MV cable is Nexans NA2XS(F)2Y with a cross-section

of 240 mm2[41]. According to the datasheet, the capacitance and inductance

of this MV cable are 0.30 µF/km and 0.50 mH/km, respectively.

Figure 2-15: The structure of an XLPE MV cable [41].

2.5.3Medium Voltage Cable Connector

This work aims to implement a multi-sensor PD detection approach for the

MV cable connectors. To better understand this component, the structure of

an MV cable connector is explained in the following paragraphs.

The MV cable connector is the interface between the cable and switchgear,

transformer, etc. Figure 2-16 illustrates an MV cable connector and its related

components. To connect the MV cable with the bushing of the device, a cable

lug is installed on the cable’s inner conductor. The cable lug is screwed to the

bushing. To prevent direct contact between the MV cable’s inner and outer

conductor, the cable’s semi-conductive is removed, and copper screens are

bent back onto a specific distance given in the corresponding datasheet. The

copper screen and the ground lead are connected to the ground potential.

26 theoretical foundations

The semiconducting layers on the inner and outer surface of the insulation

material assure a radial symmetric electric field distribution along the cable.

However, the cable connection with the bushing assumes the partial removal

of the insulation’s outer semiconducting layer (Figure 2-16a), creating a high

electric field density at the edge of the shorted semiconducting outer layer.

There is also an uncontrolled electric field over the cable lug and the peeled

insulation material.

In order to overcome the problem mentioned above, two semi-conductive

field grading elements are included in the structure of the cable connector

and the corresponding stress cone adapter. These control the electric field

and prevent PDs. The cable connector’s inner screen directly contacts the MV

cable’s lug and has the same electric potential.

(a) (b)

Figure 2-16: (a) Different MV cable and cable connector components. (b) The cross-section

view of the installed cable connector.

The adapter’s conductive layer has a field grading function, homogenizing

the electric field on stressed regions with a high electric field density. The

stress cone adapter is connected to ground potential via the MV cable’s semi-

conductive layer. Some manufacturers create a zero-voltage outer surface by

covering and grounding the cable connector body with a thin semi-conductive

external screen.

To better understand the electrical stress control inside a cable connector,

the corresponding electric field has been simulated, and the result is shown

in Figure 2-17. Figure 2-17avisualizes the electric field of an MV cable with

the installed lug without the cable connector and the stress cone adapter.

As shown, the high electric field density should be controlled to avoid PD,

especially on the shortened edge of the cable’s screen. Figure 2-17billustrates

2.6 state-of-the-art pd detection techniques onmv cable connectors 27

the effectiveness of the implemented semi-conductive elements that control

the electric field density in the stressed regions.

(a) (b)

Figure 2-17: The simulation of the electric field strength at cable connector (a) without and (b)

with conductive field grading elements.

The installation of the cable connectors is usually performed on-site under

time pressure. Hence, failures may happen during the installation, leading

to possible changes in the electric field distribution and PDs. If the PDs are

not detected early, severe damage in the connected device, i.e., switchgear,

transformer, etc., might be caused by an electric arc.

2.6 state-of-the-art pd detection techniques on mv cable con-

nectors

The PD detection on MV and HV cable connectors and terminations has been

the subject of many research activities and publications. In this section, some

state-of-the-art approaches and results are reviewed. According to the utilized

PD detection method, the following publications relevant to the subjective of

this thesis have been selected:

2.6.1IEC 60270 PD Detection

In [42], overvoltage consequences tests on MV XLPE insulated cables and their

corresponding cable terminations are investigated. The overvoltage tests are

the so-called accelerated aging measurements. The tan δ and conventional

PD measurement techniques are compared by analyzing the dielectric and

28 theoretical foundations

insulation changes over time. It is shown that a tan δ measurement is not

a practical method for cable termination faults detection. PD measurements

deliver more reliable information considering the insulation failures.

Different PD types in the HV cable terminations, based on the IEC 60270

measurement method, are presented and discussed in [43]. The examined

fault types are corona, a cavity in insulation, surface discharges, loose metal

contacts, and floating particles. The measurement results are compared using

the phase-resolved partial discharge (PRPD) patterns 2. It is discussed that

the cable’s PD and background noise signals cannot be distinguished from the

cable termination’s PD signal using the conventional measurement method.

2.6.2HFCT Sensor

The inductive PD detection in different HV and MV applications using the

HFCT sensor has been a well-known approach for a long time. It is a common

PD detection method in many publications, for example, in [44–46]. The

onsite application of the HFCT sensor is straightforward, as the cable acts

as a coupling capacitor, and the HFCT measures the transient current in the

ground wire.

In [47], PD detection on a 35 kV cable termination using HFCT sensors

is investigated. The sensor is connected with the corresponding developed

measurement instrument having a bandwidth of around 1GHz. The devel-

oped measurement system analyses the recorded signals. A fault classification

is performed based on the signal amplitude and frequency range. The

algorithms classify the signals into the internal, surface, corona discharges,

noise, and invalid data.

Eight different cable termination faults are prepared and investigated in

[48]. The measurements are performed using a commercial HFCT sensor and

the conventional PD detection method. The measurements are compared by

considering the related inception voltage and the PD signal amplitude. The

PD faults and associated breakdown probability are estimated based on the

measurements. It is shown that the used HFCT is not sensitive enough to

measure the PD signals at the inception voltage.

In [49], it is shown that the HFCT sensor is sensitive to background noise.

Hence, correct PD detection assumes the implementation of corresponding

noise rejection methods.

2PRPD pattern is explained in Section 2.7.4

2.6 state-of-the-art pd detection techniques onmv cable connectors 29

The accelerated ageing measurements on MV cable terminations using the

HFCT sensor are presented in [50]. The creation of electrical trees due to

overvoltage and the development of PD signals over 13 weeks are illustrated

and discussed.

In conclusion, it can be noted that the HFCT sensor is an appropriate sen-

sor for PD monitoring applications. However, this sensor is sensitive to back-

ground noise signals. The common commercial HFCT sensors are expensive

for MV monitoring approaches.

2.6.3Electromagnetic PD Detection

In [51,52], an electromagnetic PD detection approach for 300 kV GIS cable

connectors is introduced. For this, a monopole antenna is placed in a barrel

sleeve around the cable connector, and the PD detection is performed and

described in time- and frequency-domain measurements. Implementing the

presented monopole antenna in the MV cable connectors seems challenging

and is beyond the scope of the MV assets considering costs.

2.6.4Acoustic PD Detection

The acoustic PD detection on the MV cable connector using fiber-acoustic

optic sensors (FOS) and a Sagnac interferometer is presented in [53]. The

fiber-acoustic optic application has the advantage of good immunity to

electromagnetic noise. However, these sensors are susceptible to acoustic

noise. In the mentioned study, the acoustic PD signals of the cable connector

are analyzed and discussed in the time domain. Using fiber-acoustic optic

sensors, including the associated measurement hardware, is expensive. It

cannot fulfill the cost requirements of MV equipment monitoring (at the time

of writing this document).

The acoustic PD detection on a cable termination using an AE sensor is pre-

sented in [32]. An AE sensor with a frequency range of 60 kHz – 150 kHz with

a corresponding amplifier is used in the mentioned study. The measurements

are performed in the time domain using an oscilloscope. The use of AE sensors

is interesting but must be cost-optimized for PD monitoring of MV equipment.

2.6.5Capacitive PD Detection

The use of a capacitive voltage divider device connected parallel to the HV

cable connector is suggested in [44]. However, the proposed method is suitable

for offline PD measurements and is not cost-effective.

30 theoretical foundations

A TEV sensor for PD detection on MV cable connectors is proposed in

[54]. The sensor is made of a flexible polymer material, which enables the

installation of the capacitive TEV sensor around the back plug of the connector.

The proposed dual capacitance structure enables higher sensitivities than the

other TEV structures.

The switchgear bushing might have an integrated capacitive voltage di-

vider with which the corresponding applied voltage is measurable. The men-

tioned interface can be used for PD detection in an MV cable termination and

switchgear [55]. It is shown that the bushing’s capacitive interface has a lower

sensitivity for PD detection than HFCT and TEV sensors.

2.6.6Ohmic Detection

In [32], PD detection in cable connectors is realized by connecting a series

resistor to the associated ground lead. The measurements are performed in

the time domain using an oscilloscope. The measurement results show the

possibility of a sensitive PD acquisition method in higher frequencies.

The principle of the ohmic measuring resistor is illustrated in Figure 2-

18 [13]. The current i(t), for example, generated due to partial discharges,

flows through a low-ohmic resistance R1. The voltage drop over R1can be

measured with a high-ohmic impedance Z. The voltage drop can be calcu-

lated by i(t)·R1. Note that the corresponding resistance of the measurement

impedance Zshould be much higher than the measuring resistors R1.

Figure 2-18: Principles of the measurement with a low-ohmic resistor.

2.6.7Summary

Different international publications have shown that PD measurement is an

essential method when evaluating the insulation quality of MV cable termi-

nations. The conventional PD detection method is not suitable for online PD

monitoring. The inductive PD measurement using the HFCT sensor is suit-

2.7 terms and definitions 31

able for monitoring approaches. However, the HFCT method is sensitive to

external noise signals and external PD sources. Electromagnetic and acous-

tic PD detection using antennas and AE sensors has not been the subject of

much research for MV cable terminations yet. The capacitive PD measurement

methods seem promising; however, the TEV sensor is more suitable for detect-

ing PD signals originating from the switchgear. Switchgear’s bushing and the

ohmic measurement principle can enable cost-effective and straightforward PD

monitoring applications with limited sensitivity.

2.7 terms and definitions

The analysis of various PD sensors and the corresponding measurement results

are presented in the further course of this work. The current section provides

some of the essential definitions required for implementing and evaluating the

measurements of this thesis.

2.7.1CAL2B – Pulse Generator

According to IEC 60270 [1], a charge calibrator should be used to calibrate the

conventional PD measurement setup. For non-conventional sensors, which

do not evaluate charge but mostly only the amplitude of the measured signal,

pulse generators have been used to test the sensitivity and functionality of the

sensors.

CAL2B pulse generator, shown in Figure 2-19a, is an example of a broad-

band pulse generator [56]. The pulse amplitude can be varied from 2V to

50 V. The related pulse shape, with an amplitude of 2V, is measured with a

50 Ohm input impedance and shown in Figure 2-19b. The captured pulse has

a rise time of around 200 ps and a signal duration of approximately 350 ns.

The associated frequency spectrum of the CAL2B pulse and the corresponding

reference level is illustrated in Figure 2-19c. The pulse has high-frequency

components up to 3GHz, making the pulse generator suitable for very high

frequency (VHF) and UHF applications.

The CAL2B is used in this work for multiple purposes. In addition to defin-

ing the measurement setup sensitivity, the broadband frequency spectrum of

the pulse generator is used to determine the frequency responses of different

sensors.

32 theoretical foundations

(a) (b)

(c)

Figure 2-19: (a) The CAL2B pulse generator. (b) The corresponding time-domain signal shape.

2.7.2S-Parameters

In the following chapters, the frequency-dependent characterization of sensors

and the MV cable is performed by the scattering parameters (S-parameters).

Hence, a brief review of the S-parameters is presented in this section.

Measuring voltages and currents at higher frequencies at the input and

output ports of the DUT is challenging. A more practical approach is the

consideration of incident, reflected, and transmitted power waves at the DUT

ports [57]. The S-parameters describe the relationship between the mentioned

waves. The S-parameters can be measured directly with a vector network

analyzer (VNA).

In Figure 2-20, a two-port network, as well as the related incident, transmit-

ted, and reflected power waves, are illustrated. A portion of the incident wave

toward the DUT (a1(t),a2(t)) is forward transmitted through the DUT, and

the rest is reflected on the corresponding port. The reflected waves on each

port add to the reverse transmitted waves and build the output waves from the

ports, i.e., b1(t),b2(t).

2.7 terms and definitions 33

Figure 2-20: A two-port network and the related incident, transmitted and reflected waves.

As shown in Equation 1, the S-parameter matrix defines the relation between

the incident, transmitted, and reflected wave.

[︄b1(t)

b2(t)]︄=[︄S11 S12

S21 S22 ]︄[︄a1(t)

a2(t)]︄(1)

The equation system can also be written as:

b1(t) = S11a1(t) + S12a2(t)(2)

and

b2(t) = S21a1(t) + S22a2(t)(3)

Using Equation 2and Equation 3, the elements of the s-parameter matrix can

be defined separately. The parameters S11 and S21 can be written as:

S11 =b1(t) − S12a2(t)

a1(t)(4)

S21 =b2(t) − S22a2(t)

a1(t)(5)

Assuming that port 2is matched and no reflection takes place, a2(t)is zero,

and the latter two equations can be simplified to:

S11 =b1(t)

a1(t)|a2(t)=0(6)

S21 =b2(t)

a1(t)|a2(t)=0(7)

Equation 6shows that the S11(t)is the ratio of the reflected- to the inci-

dent wave on port 1.S11(t)is called the input reflection coefficient. The

34 theoretical foundations

measurement of S11(t), using a VNA, enables the definition of the reflected

power from the input port of a DUT. The VNA ports and measurement cables

usually have an impedance of 50 Ohm. Hence, the reflection coefficient is a

parameter with which the mismatch of the DUT port to 50 Ohm can be defined.

Parameter S21(t)is the ratio of the transmitted wave through the DUT to

the incident wave. S21(t)is the so-called forward transmission coefficient. The

S21(t)can be measured with a VNA or a spectrum analyzer in the network

mode. The measured S21(t)illustrates the signal attenuation or amplification

of the DUT to the incident waves.

The remaining two parameters, i.e., S22(t)and S12(t), are the output reflec-

tion and reverse transmission coefficient. Further detailed information regard-

ing the S-parameter can be studied in Chapter 4.3of [57]. In this work, the

S-parameters are used to characterize the sensors and MV cable in the HF

range.

2.7.3SNR

One of the main challenges in the PD diagnostic is the presence of noise signals

originating from different sources, for example, rotating machines, frequency

converters, telecommunication devices and radar signals, etc. In such a harsh

electromagnetic environment, the noise signals superpose to the PD signals

and couple to the measurement sensors and devices. The signal-to-noise ratio

(SNR) can be defined as the ratio of the PD signal power and the background

noise power.

SNR =PSignal

PNoise

(8)

Equation 8can be expressed in dB using the logarithmic function:

SNRdB =PSignal,dB −PNoise,dB (9)

2.7.4PRPD

Section 2.2explained that the corresponding discharges occur around specific

phase angles of the reference sine voltage depending on the PD failure. Fig-

ure 2-21aillustrates an example with three periods of the sine voltage and

the associated PD signals, mainly occurring around 90° and 270°. For bet-

ter visualization, the PD events are consecutively numbered in Figure 2-21a.

Figure 2-21bshows the same events ignoring the different sine period num-

2.7 terms and definitions 35

bers illustrated. The phase-resolved partial discharge pattern (PRPD) is the

2-dimensional statistics of the PD repetition rate. The phase angle is ordered

on the horizontal axis, and the pulse amplitude is shown on the vertical axis.

The repetition rate is illustrated by color coding for a given phase and ampli-

tude.

(a)

(b)

Figure 2-21: (a) Partial discharges over three following sine periods. (b) The corresponding

PRPD pattern of the discharges. For better visualization, the PD events are con-

secutively numbered.

3EVALUATION OF SENSORS

The existing commercial PD measurement sensors are expensive when com-

pared to the costs of the MV equipment, and the IEC 60270-based detection

method requires changes in the test setup. Furthermore, PD measurements

are usually performed using sensors of the same kind and with the same

frequency operating range. Hence, a specific noise source disturbs all sensors

in the same manner.

This chapter presents the evaluation of different sensing methods for a

multi-sensor approach used to monitor MV cable connectors. The goal is

to select sensors that are cost-effective, easy to install and have different

operating principles, for example, non-identical frequency operating ranges.

To this end, nine different PD sensors are examined. The sensors detect the

PD signals in different ways. The HFCT measures the PD transient current

flowing toward the ground. The electromagnetic PD emission is studied

utilizing two antennas. The capacitive PD measurement is also applied to

the MV cable connector by introducing two novel measurement techniques.

A further studied PD measurement approach is the use of an ohmic sensor.

Finally, the acoustic PD emissions are investigated using three AE sensors.

The mentioned sensors are examined in the following sections. The sensors

are compared using PD measurements. Based on the gained measurement

results, the best sensors concerning sensitivity, cost, and practical usability are

selected.

3.1 description of sensors

3.1.1HFCT Sensor

To the author’s knowledge, the existing commercial HFCT sensors usually

have a bandwidth of up to tens of megahertz and cost in the region of many

hundreds of euros. In such a case, the sensor price is higher than the cable

connector, plus the installation cost. Therefore, other cost-effective HFCTs are

required.

3.1 description of sensors 37

The goal is to develop an HFCT sensor that is cost-effective, simple to install,

and measures the PD signals reliably. Furthermore, the HFCT should have a

broadband frequency response supporting PD source localization.

To this end, a proper test setup for the characterization of the HFCT sensor

is described. Then, the wire diameter, the windings number, and the core type

are studied based on the performed measurements. Finally, PD measurements

characterize the sensor functionality.

After a PD event in a cable connector, the discharged capacity of the stress

zone is recharged by the MV cable’s capacitance. This transient current I1can

be measured with an HFCT installed over the cable screen connected to the

ground (Figure 3-1). The current I1induces the current I2in the HFCT sensor

secondary winding. A PD measurement device can capture the voltage drop

V over a connected resistor R.

Figure 3-1: The placement of the HFCT around the MV cable screen.

The HFCT test setup

Based on the literature reviewed in Section 2.4.1, no straightforward test

setup for the characterization of the HFCT is known. Therefore, a suitable

measurement setup should be applied. For this, different standards are

studied, and a proper and easy-to-implement test setup is chosen in ISO

11452-4[58], describing the automotive current probe’s test and calibration

methods. The test setup is adopted to characterize the HFCT PD application

of this thesis.

38 evaluation of sensors

ISO 11452-4suggests using a two-port network analyzer and a broadband

50 Ωmatching and measurement fixture (Figure 3-2). The current transformer

is placed around the inner conductor of the coaxial structure of the measure-

ment fixture. Port 1of the network analyzer generates the incident wave a1

transmitted toward the 50 Ω-load. The incident wave induces a voltage in the

transformer’s secondary windings measured in port 2of the network analyzer

(b2). The ratio of b2to a1is the transmission coefficient (Section 2.7.2), with

which the frequency-dependent transfer function of the current transformer

can be defined.

Figure 3-2: The suggested measurement of ISO 11452-4[58] for characterization or calibration

of the current transformer.

A similar measurement setup is used in this work (Figure 3-3a). The main

difference is the lack of the broadband 50 Ω-measurement fixture. Here a test

setup is chosen closer to a real MV cable ground connection used for actual

PD measurements, where a simple bypass cable of 10 cm length is used in the

ground path.

The described test setup of Figure 3-3aintroduces a significant impedance

discontinuity to the 50 Ωstructure of the used coaxial cable and generates

a considerable signal reflection and attenuation. If the network analyzer’s

built-in generator with a signal amplitude of 0dBm (around 316 mV peak

value) is used, the test setup attenuates the signals significantly, and the

input port cannot detect them. For this reason, the CAL2B pulse generator is

used, which generates a PD-like pulse with a higher signal amplitude over a

broadband frequency range, i.e., 2V maximum amplitude.

The frequency components of the calibrator’s signal do not have a constant

amplitude over the whole frequency range, and they are attenuated by the

3.1 description of sensors 39

(a) (b)

Figure 3-3: The measurement setup for the characterization of the HFCT sensor. (b) The test

setup’s reference frequency spectrum was measured with the reformed coaxial ca-

ble and CAL2B pulse.

measurement setup differently. Therefore, the frequency response of the test

setup to the pulse signals should be measured. The CAL2B’s pulse is therefore

injected into the reformed coaxial cable on one side and measured with the

spectrum analyzer on the other side. The measured spectrum (Figure 3-3b)

is the reference spectrum injected into the HFCT sensor and includes the

measurement setup’s discontinuities, reflections, and attenuations.

Different HFCT sensor designs can be compared using the described setup.

One example sensor is tested in Figure 3-4a, where the inner conductor of

the coaxial cable goes through the sensor’s ferrite core, and the end of the

reformed coaxial cable is 50 Ωterminated. The sensor has two windings with

a laminated copper wire diameter of 0.8mm around the NiZn ferrite core

Würth 74270191. A short coaxial cable connects the windings of the sensor

to the spectrum analyzer’s input. The corresponding measurement results

are shown in Figure 3-4b. The illustrated sensor’s frequency response is

post-calculated by subtracting the reference spectrum of the test setup (Figure

3-3b) from the measured input spectrum of the HFCT. It is shown that up to

around 150 MHz, the sensor transmission loss is less than -10 dB. Note that

this measurement setup is suitable for comparing the functionality of different

structures and not for the sensor’s calibration.

To visualize and ensure the functionality of the test method, a time-domain

measurement is also performed (Figure 3-4c). The pulse of the CAL2B is

illustrated with the blue line. The induced voltage into the sensor (the

orange line) has no significant signal shape changes – for example, rise time

differences.

40 evaluation of sensors

(a) (b)

(c)

Figure 3-4: (a) Measurement setup for the characterization of an HFCT. (b) The measured fre-

quency response of the HFCT. (c) Time-domain response of the HFCT.

The optimum sensor structure

After establishing a test setup for comparing the different HFCTs, the most

suitable sensor structure is found, considering sensor sensitivity over a wide

frequency range and sensor cost-effectivity. For this, the proper core material

and size are examined in the first step. Then, the appropriate wire thickness

and the number of turns are defined.

Core - In the first step, eleven different cores with the same wire thickness

(0.8mm) and two windings are compared. With this, the detected time-domain

pulse amplitude of each sensor structure is measured and noted in Table 3.1.

3.1 description of sensors 41

Table 3.1: The characteristics of the investigated ferrite cores and the corresponding detected

pulse amplitude with two windings and a wire thickness of 0.8mm.

No. Core AC[mm2]1IC[mm]2Material Detected pulse

amplitude [mV]3

1Würth 74271358 270.0 229.3Ni-Zn 189

2Würth 742701112 76.5 191.6Ni-Zn 194

3Würth 74270104 101.6 159.2Ni-Zn 197

4Laird 28B2400-000 322.6 303.2Ni-Zn 202

5Kemet ESD-R-25S120.0 125.6Mn-Zn 214

6Kemet ESD-R-47 360.0 230.9Mn-Zn 215

7Laird 28B1417-200 165.1 185.3Ni-Zn 256

8Würth 74270115 200.0 158.6Ni-Zn 308

9Kemet ESD-R-28C156.0 138.2Ni-Zn 332

10 Würth 74271378 614.3 227.9Ni-Zn 369

11 Würth 74270191 510.0 303.1Ni-Zn 480

1Cross-sectional area of the core.

2The mean path of the core.

3Core with two windings and laminated copper wires with a diameter of 0.8mm.

The noted cores in Table 3.1are of different materials and sizes. The core

size is determined as the mean path of the core (ls) and the corresponding

cross-sectional area Ac. The primary measurements have shown that the

HFCT with Würth 74270191 has the highest detected pulse amplitude of 480

mV and has a transmission loss of -10 dB around 150 MHz. Hence, this core

is selected for further measurements.

Coil - The number of turns and wire thickness are investigated in the second

step. Laminated copper wires with diameters of 0.2mm, 0.8mm, and 1.5mm

are used for the sensor’s windings. Three different windings are realized

with each wire: 2,5, and 10 turns. Overall, nine variations are measured and

compared using the prescribed measurement method.

Figure 3-5indicates the measurement results of the HFCT sensitivity as a

function of turns and wire diameter derived from the maximum induced sig-

nal amplitude. It is illustrated that a lower number of turns leads to a higher

sensitivity. As explained in Section 2.4.1, a lower number of windings corre-

sponds to a shorter wire and a lower ohmic resistance, attenuating the signal

amplitude. A thicker wire with a diameter of 1.5mm has smaller resistance,

increasing the sensor’s sensitivity.

42 evaluation of sensors

Figure 3-5: The maximum signal amplitude in the time domain measurement of the HFCT

(sensitivity) as a function of the wire’s diameter and the number of turns.

The average of the sensor’s transmission losses is examined separately in

the frequency ranges of 1kHz – 20 MHz and 20 MHz – 150 MHz using the

spectrum analyzer (Figure 3-6). As shown, the average transmission loss of

the sensor decreases in all frequency ranges using fewer turns with a thicker

wire. More windings lead to a higher DC resistance, increased proximity effect,

and parasitic capacitance. However, the wire’s diameter does not impact the

transmission loss as much as the number of turns. In higher frequencies, the

windings losses increase.

(a) (b)

Figure 3-6: Transmission loss of the HFCT vs. the wire diameter for a frequency range of (a) 1

kHz – 20 MHz, (b) 20 MHz – 150 MHz.

Based on the presented measurement results in Figure 3-5and Figure 3-6,

the sensor structure with two turns and a wire diameter of 1.5mm is selected

for further investigations of this work. This sensor has a higher sensitivity up

3.1 description of sensors 43

to a frequency of around 150 MHz. In Section 3.2, the PD measurements with

the HFCT are discussed, and the results are compared with the other sensors.

3.1.2Electromagnetic PD Detection Using Antennas

As described in Chapter 2, using antennas for electromagnetic PD detection of

HV equipment, for example, GIS, is a well-known approach [59]. Nevertheless,

to the author’s knowledge, the PD detection on the MV cable connectors using

antennas is not well studied. To this end, the investigations in this section aim

to find a suitable antenna for PD detection of MV cable connectors.

The frequency range of PD signals emitted by different PD-afflicted cable

connectors needs to be specified in order to find and develop an appropriate

antenna structure. Therefore, in the first step, the corresponding PD spectrum

is studied. Based on this information, the required antenna size can be derived.

Two different antennas for PD detection are investigated, and the advantages

and drawbacks of the proposed approaches are discussed.

Frequency range of PD signals emitted from the PD-afflicted cable connec-

tors

Depending on the application field, the PD measurement spectrum varies.

For example, PD detection in GIS or power transformers is carried out between

300 MHz and 3GHz [60], whereas a measurement frequency range of 0.5–

1.3GHz is proposed in [61] to detect PDs in inverter-fed motors. Furthermore,

the antenna’s size and shape are crucial for HV/MV use. The antennas should

generally be small and not have sharp edges to prevent PD generation.

To measure the frequency range of emitted electromagnetic PD signals from

the PD-afflicted cable connectors, a broadband biconical antenna [62] is used

(Figure 3-7). The illustrated antenna has a -6dB bandwidth from 20 MHz to

300 MHz. Nevertheless, the antenna can also measure electromagnetic signals

in higher frequency ranges that are not calibrated. The antenna is large and

cannot be used for monitoring approaches and only verifies the required

bandwidth.

The measurements are performed inside a shielded metallic chamber, where

external electromagnetic disturbances are limited. The biconical antenna at a

distance of 1m to the PD location is connected to an amplifier and a spectrum

analyzer with a stop measurement frequency of 700 MHz.

44 evaluation of sensors

Figure 3-7: Measurement setup with a broadband biconical antenna to evaluate cable connec-

tor’s PD spectrum.

Three PD-afflicted cable connectors are connected to an MV switchgear, and

a test voltage of up to 18 kV AC is applied to the cable connectors. The built-in

PD defects in the cable connector are created by:

• a notch in the cable insulation,

• a longer conductive layer, as defined in the installation guide of the con-

nector, and

• a cable screening wire placed on the cable insulation.

Additionally, the spectrums of three external PD sources are measured. The

external discharges are:

• corona discharges - needle on HV potential,

• corona discharges - needle on the ground potential,

• discharges due to a loose contact nut in the bushing’s holder case.

Further details regarding the implemented faults are presented in Chapter 6.

The corresponding measurement results are shown in Figure 3-8.

Recorded spectra for the PD-afflicted cable connectors (Figure 3-8a) have

shown electromagnetic emissions between 30 MHz and 250 MHz. The spectra

of external PD sources (Figure 3-8b) show a broader frequency range of up to

approximately 650 MHz. The measurements indicate that the antenna used

for detecting PD-afflicted cable connectors should have a bandwidth of up to

250 MHz. An even higher bandwidth, up to 650 MHz, could be helpful to

recognize and discriminate external discharges like corona discharges.

3.1 description of sensors 45

(a) (b)

Figure 3-8: The measured spectrum of discharges generated by (a) three different PD-afflicted

cable connectors and (b) three different external faults. The biconical antenna is

calibrated just to 300 MHz.

As illustrated in Figure 3-7, the biconical antenna’s size cannot satisfy a

compact online monitor system’s requirement. Therefore, antennas with

smaller sizes should be used. In the following sections, two different antennas

are investigated with which the PDs of cable connectors in the investigated

frequency ranges can be detected. The antennas are chosen based on their

compact structures, enabling an easy installation in the switchgear’s cable

compartment.

The first investigated antenna is an annular-ring microstrip antenna (AR-

MSA), which is realized with an FR4substrate and has a small, flat structure.

The leaky feeder antenna is the second investigated approach, studied in Sec-

tion 3.2.

3.1.2.1Annular-Ring Microstrip Antenna (AR-MSA)

Figure 3-9shows the general overview of the AR-MSA. This antenna is simple

to produce, install, is cost-effective, and has a large bandwidth, for example,

from 2.5GHz to 12 GHz in [63]. A detailed study regarding radiation prop-

erties, input impedance, and analytic analysis is performed in [64–66]. The

antenna consists of an annular ring, a feeding microstrip line, and a ground

plane under the feeding line. An FR4substrate with a thickness of 1.5mm and

a copper thickness of 35 µm is used.

46 evaluation of sensors

Figure 3-9: The overview of the annular-ring microstrip antenna (AR-MSA) on FR4substrate.

The antenna is so designed that the corresponding bandwidth covers the

signal frequencies of PD-afflicted cable connectors, i.e., a minimum of 30 MHz

–250 MHz. The following two steps are performed using field simulation

software for the AR-MSA design. First, the appropriate size of the annular

ring is found. Then, the feeding microstrip line is designed. The simulations

have led to an annular structure with an outer radius of 100 mm and an inner

radius of 40 mm. The feeding line has a width of 2.8mm on the surface-mount

assembly (SMA) socket side and 0.7mm on the annular side. The detailed

development process is presented in Appendix A.

The bandwidth of the fabricated antenna is defined by the measurement

of return loss (S11) with a vector network analyzer (Figure 3-10). Consider-

ing –6dB matching, a wide bandwidth of around 1.9GHz (from 150 MHz to

2GHz) is measured with this antenna. This measurement does not demon-

strate the desired simulated matching from 30 MHz, due to manufacturing

tolerances, measurement inaccuracies, and differences in material parameters

compared to the simulation parameters used. However, the antenna can mea-

sure the HF signal components of the local PD faults. The antenna’s PD mea-

surements are presented in Section 3.2, where all the sensors are compared.

3.1.2.2Leaky Feeder Antenna

The leaky feeder antennas have found extensive data communication usage

in harsh environments such as on aircraft, trains, skyscrapers, and in tunnels,

mines, etc. [67]. Besides the flexibility, robustness, and ease of installation,

the leaky feeder antennas can operate in different frequency bands and have

3.1 description of sensors 47

Figure 3-10: The return loss measurement and the simulation results of the annular-ring mi-

crostrip antenna.

a broadband frequency response. The mentioned advantages could be an

attractive approach in the HV industry for contactless monitoring of HV

cables, joints, etc. They can be laid parallel to the HV components without

direct contact and capture the emitted PD electromagnetic waves.

Figure 3-11 shows the general structure of the leaky feeder antenna. The

antenna is like a coaxial cable, including an inner and outer conductor and in-

sulating material, with the difference that the outer conductor is slotted. These

slots lead to the radiation of the cable. The currents flow over the inner and

outer conductors, and wave propagation occurs through the insulating dielec-

tric. The waves can leak through the outer conductor at the slots’ position, and

the coaxial cable radiates [67].

Figure 3-11: The structure of the leaky feeder antenna.

48 evaluation of sensors

According to the antennas’ reciprocity theorem, external waves, like PD

emissions, can enter and propagate along the cable. The leaky feeder an-

tenna’s emission characteristics can be changed using different slot shapes or

variations in the slot positions. Further reading can be found in [67,68].

The leaky feeder is investigated because it can be installed in the switchgear’s

cable compartment without contact with the MV cable connectors. This is

an interesting approach as this antenna can be installed while maintaining

switchgear.

A leaky feeder antenna from Radio Frequency Systems GmbH is used,

i.e., the RLK12-50JFNA cable [69]. According to the antenna’s datasheet, the

antenna has a diameter of 1.27 cm, a low weight (0.23 kg/m), and is suitable

for applications in tunnels and buildings. The leaky feeder cable’s insertion

loss for the high frequencies is 2.17 dB/100 m and 10.51 dB/100 m at 75 MHz

and 960 MHz, respectively.

The leaky feeder cable used for this work is prepared by the company Radio

Frequency Systems GmbH with a length of 1m, and has two 50 ΩN-type

male connectors on both ends. In this work, the far end is 50 Ωmatched, and

the measurements are performed over the near end. The associated bandwidth

measured with a VNA considering the return loss (S11) parameter is illustrated

in Figure 3-12. Considering -6dB matching, the antenna can be operated over

the measured frequency range of 0.01-1GHz.

Figure 3-12: The return loss measurement results of the leaky feeder antenna.

3.1 description of sensors 49

3.1.3Coaxial Sensor

The capacitive PD detection of the MV cable connector is not well studied,

and the existing methods assume galvanic coupling with the cable connector.

Furthermore, the lack of space in the cable compartment of an MV switchgear

and the sensors’ cost-effectivity are important for the MV-cable connectors’ PD

detection. To the author’s knowledge, the existing capacitive sensors cannot

satisfy these requirements.

This section investigates the development of a cost-effective capacitive

approach with high sensitivity, easy installation, and a broadband frequency

response. The theoretical backgrounds of existing PD capacitive sensing

methods are explained in Section 2.4.2. In this section, the proper sensor

positioning is studied in the first step. Then, a coaxial sensor is developed and

characterized. A second PD measurement method, i.e., the capacitive voltage

divider of the switchgear’s bushing, is discussed and analyzed in the next

section. In Section 3.2, the performance of the capacitive approaches for PD

detection is compared to the other sensors.

Proper sensor position

The early investigations of this work have shown that any sensor with a

metallic structure placed on the surface of the cable connector, where the

electric field strength has a high value, produces discharges. It was, therefore,

concluded that the proper capacitive sensor should not have any direct

contact with the cable connector’s surface at a point with high electric field

density. For this, and to define the electric field in different positions, the

corresponding electric field density of the cable connector is studied using a

3D-field simulation program.

The simulated electric field density of the cable connector is illustrated in Fig-

ure 3-13a. The shield of the MV cable has ground potential. The cable’s inner

conductor, lug, and connector’s inner screen are under high voltage. The max-

imum field density is between the cable’s inner conductor and the screen. Fur-

thermore, there is a high field density between the stress cone adapter - with a

ground potential - and the cable connector’s inner screen – with a high voltage

potential. The placement of a sensor with a metallic housing and ground po-

tential in the upper region of the cable connector (Figure 3-13b) leads to a high

electric field density and the generation of discharges. As illustrated in Figure

3-13c, the proper position for installing the sensor is in the lower cable connec-

tor’s region and close to the stress cone, where the electric field is controlled.

50 evaluation of sensors

(a) (b) (c)

Figure 3-13: (a) The simulated electric field density of the cable connector at 10 kV. (b) The

generation of high electric field density by placing a sensor with grounded metal

housing on the cable connector’s body. (c) The proper sensor position in the lower

cable connector’s region.

Development of the coaxial sensor

A coaxial cable is used to develop a flexible and small-sized capacitive sen-

sor. However, the coax cable is so modified that its screen is partially removed,

and a “window” is created (Figure 3-14). The created window generates a

capacitance between the coaxial cable’s and the MV cable’s inner conductors.

Furthermore, a second capacitance exists between the coaxial sensor’s inner

conductor and the corresponding screen. These two mentioned capacitances

enable capacitive electric field measurement. Additionally, the screen of the

coaxial sensor shields the inner conductor from the electric field of the neigh-

boring cable connectors and improves the signal-to-noise ratio.

The sensor is called a coaxial sensor, as the MV cable’s inner connector and

the sensor’s inner conductor build a coaxial-like structure.

The electric field of the MV cable connector generates an electric potential

difference between the sensor’s inner and outer conductors. The wave propaga-

tion takes place along the sensor length and in two directions. To measure the

waves in the near-end port, the sensor’s far-end port can be electrically short-

connected, floated, or 50 Ωmatched. The three mentioned possible scenarios

are tested, and the most appropriate structure, i.e., the short-circuit far-end,

is considered for further measurements (Figure 3-15). The short-circuited sen-

sor provides measurement signals with higher amplitudes and prevents 50 Hz

reference voltage signal coupling to the sensor signal.

3.1 description of sensors 51

(a) (b)

Figure 3-14: (a) The 3D model of the developed coaxial sensor around the cable connector and

(b) the corresponding sensor’s structure.

The described coaxial sensor does not have a 50 Ωimpedance due to the

performed changes in the coaxial cable structure. Due to the impedance vari-

ations, impedance matching is inappropriate for this sensor’s measurements.

Therefore, impedance bridging with a high load impedance is implemented.

The used amplifier for this sensor has an input impedance of 500 kΩ. The

amplifier has a broadband frequency response of up to 500 MHz, which is

required for PD detection in this work.

(a) (b)

Figure 3-15: (a) The developed coaxial sensor. (b) Mounting of the coaxial sensor with a hous-

ing around the MV cable connector.

For the sensor’s positioning around the cable connector and over the stress

cone, a corresponding 3D-printed housing is developed (Figure 3-15b). The

illustrated housing ensures a specific distance between the sensor and the cable

connector. The coaxial sensor’s opening can also be precisely adjusted toward

the cable connector using the housing.

52 evaluation of sensors

Frequency response of the coaxial sensor

A possible measurement setup for determining the frequency response

of the coaxial sensor is the use of a gigahertz transverse electro magnetic

(GTEM) cell (Figure 3-16a), in which an electric field over a broadband

frequency range of up to 2GHz can be generated. The principle of wave

propagation in a GTEM cell is presented in Appendix A. The coaxial sensor

with the corresponding housing is so placed that the electric field lines are

perpendicular to the sensor’s opening (Figure 3-16b). This measurement setup

can specify the frequency ranges in which the sensor is more sensitive.

(a) (b)

(c)

Figure 3-16: (a) The GTEM cell was used to measure the coaxial sensor’s frequency response.

(b) The sensor is placed in the GTEM cell. (c) The frequency response of the

coaxial sensor.

For the measurement, the spectrum analyzer is utilized in the network mode.

The integrated signal generator generates the frequency sweep and feeds a

broadband amplifier and the GTEM cell. The generated electric field couples

to the sensor connected with the associated amplifier. The amplified sensor’s

signals are transferred to the input port of the spectrum analyzer, which

3.1 description of sensors 53

calculates the ratio of the incoming to outgoing signals (the S21 transmission

coefficient explained in Section 2.7.2). Before the measurement, the setup is

calibrated, and the impact of the measurement cables, adapters, amplifiers,

etc., is considered.

The frequency response of the coaxial sensor is illustrated in Figure 3-16c.

The shown spectrum is calculated by subtracting the idle measurement from

the measured transmission coefficient. The coaxial sensor has a broadband

frequency. However, this sensor has a higher sensitivity (more than 30 dBm) in

the 40 MHz – 250 MHz frequency range.

3.1.4Switchgear’s Bushing

During the investigations, the question is raised as to whether PD detection

could be possible using the integrated capacitive voltage divider of the

bushing. If this approach fulfills the requirements of this work, two significant

advantages can be gained. Firstly, this measurement interface is already

installed in the switchgear to measure the applied voltage for safety reasons.

Secondly, the capacitive divider enables the simultaneous measurement of the

PD and the reference voltage.

The structure of the used switchgear’s bushing is shown in Figure 3-17 [70].

However, the illustrated figure is estimated and does not correspond to the

exact original dimensions and structure. The bushing is made of an inner

conductor, which carries the current. A capacitive voltage divider in two

parallel cylindrical forms is placed around the inner conductor. One of the

cylinders is grounded, while the other is connected to the measurement point.

The voltage divider ratio of this bushing contact is around 1/52.

(a) (b)

Figure 3-17: (a) The estimated 3D structure of the switchgear’s bushing with an integrated

capacitive divider. (b) The estimated cross-section view of the bushing with the

corresponding capacitances.

54 evaluation of sensors

Frequency response of the bushing’s capacitive divider

The CAL2B pulse generator with a broadband signal (Section 2.7.1) is used

to measure the bushing’s capacitive divider’s frequency response. To achieve

this, one special housing is fabricated (Figure 3-18a), with which the MV

cable connector is connected to the bushing easily. The inner conductor of

the bushing can be reached from the backside. The CAL2B pulse is injected

into the bushing’s conductor using a coax splitter adapter. On the MV cable

connector’s side, the signal is detected over the measurement port of the

bushing. A second coax splitter adapter transfers the received signals to the

spectrum analyzer. The ground sockets of both adapters are connected to the

housing.

Both used coax splitter adapters introduce a distortion in the original

spectrum of the CAL2B pulse in the higher frequencies. Therefore, the CAL2B

pulse spectrum is also measured with both adapters, and the corresponding

used cables. The frequency response of the bushing’s capacitive divider is

calculated by subtracting the CAL2B pulse spectrum as a reference from the

measured spectrum. The result is illustrated in Figure 3-18b. As shown, in the

lower frequency range, the divider factor of 1/52 approximately corresponds

to the measured value of -19 dBm. The high-frequency signals traveling over

the bushing’s inner conductor can be measured with the capacitive divider

with a higher intensity. As illustrated, at around 150 MHz, 280 MHz, and

500 MHz, the measured signal amplitude is about – 2dBm.

(a) (b)

Figure 3-18: (a) The test setup for measuring (b) the frequency response of the bushing’s ca-

pacitive divider as a PD sensor.

3.1 description of sensors 55

3.1.5Ohmic Sensor

As discussed in Section 2.5.3, the MV cable connector body should mostly be

equipped with a ground bead. It is examined whether the PD signals can be

detected using the connector body’s grounding wire. Hence, this section’s

investigations aim to develop an additional cost-effective sensor to detect the

PD signals of the MV cable connector.

The cross-section view of the stress cone adapter is illustrated in Figure 3-19.

A conductive layer is built into the stress cone adapter. This layer is placed

above the MV-cable semi-conductive screen’s edge to control the electric field

strength. Theoretically, a direct electric contact exists between the cable’s outer

screen and the adapter’s semi-conductive layer. However, assembly lubricant

is used during the installation and mounting, which changes the electric

contact situation by creating an insulating layer. This layer may be subject to

breakdowns under operation conditions.

Partial discharges inside the cable connector couple to the stress cone, forcing

a voltage drop at the impurity layer. These PD-induced voltage variations can

be measured with an ohmic sensor acting as a near-field antenna. The sensor

consists of an ohmic resistor (RL) connected in series to the cable connector

body’s grounding wire. The measurement device with a high input impedance

can capture the voltage drop (VL) over the resistor.

Figure 3-19: The cross-section view of the MV cable connector and the stress cone adapter with

the ohmic sensor.

Frequency response of the ohmic sensor

The frequency spectrum of the ohmic sensor is measured using a multi-

channel digital oscilloscope and the CAL2B pulse generator. A coax splitter

adapter injects the CAL2B pulse into a short MV cable (Figure 3-20a). The

spectrum of the CAL2B pulse, containing measurement setup signal losses, is

56 evaluation of sensors

(a) (b)

Figure 3-20: (a) The measurement setup to examine (b) the frequency spectrum of the ohmic

sensor.

captured with an oscilloscope on the bushing side. The coupled signal spec-

trum to the sensor is measured with a second oscilloscope channel. The latter-

mentioned spectrum is subtracted by the CAL2B pulse reference spectrum.

Figure 3-20bshows the resulting spectrum of the ohmic sensor. The sensor has

a broadband frequency response and a higher sensitivity than -6dBm in the

80 MHz – 150 MHz frequency range.

3.1.6Acoustic Sensor

Acoustic PD detection is a well-known approach for monitoring HV equipment

such as transformers, cables, and cable joints [16,71,72]. However, to the au-

thors’ knowledge, using acoustic PD measurements for MV cable connectors

is not well studied. In this work, the suitability of the acoustic measurement

method for the PD detection of MV cable connectors, considering sensor sensi-

tivity, frequency, and cost-effectivity is studied. To this end, three AE sensors

with different frequency ranges are chosen and described. Then, the acoustic

attenuation of the silicon rubber material is measured and discussed.

The chosen AE sensors

Three different AE sensors (Table 3.2) with various frequency ranges are

used to examine the acoustic properties of the cable connector material over

a wide frequency range from 25 kHz to 900 kHz [39,73,74]. The VALLEN

sensors have an integrated single-ended amplifier. The signal amplification of

the sensor VS30-SIC is 46 dB, while the amplification of the other two sensors

3.1 description of sensors 57

is 34 dB. The sensors’ cost are high compared to the MV cable connector price

and even the MV switchgear price.

Table 3.2: The used AE sensors for PD detection of MV cable connector.

Sensor name Vendor Frequency range Amplification Price1

VS30-SIC-46dB [39] Vallen 25 kHz - 80 kHz 46 dB High

VS150-RIC [73] Vallen 100 kHz - 450 kHz 34 dB High

VS900-RIC [74] Vallen 100 kHz – 900 kHz 34 dB High

1Relative to the MV cable connector price

Problems of acoustic PD measurements on MV cable connectors

The placement of the acoustic sensors with metal housing in higher regions

of the MV cable connector leads to surface discharges. These discharges

generate acoustic wave signals that can be measured with the acoustic sensors

themselves. To avoid the described faulty measurement, the sensors should

be placed in the lower part of the cable connector, where the electric field

density has a small value. Several measurements with different AE sensors

have shown that the internal PD acoustic signals cannot be measured with

these sensors.

There are various reasons that lead to a decrease in the amplitude of acoustic

waves [5]. Firstly, the acoustic waves are propagated in different directions.

Secondly, discontinuities in a medium or the transmission between different

materials cause reflections and the attenuation of the acoustic signals. Finally,

the materials absorb acoustical waves. Further investigations were conducted

to prove the proposed theory, presented in the following section.

Measurement setup to determine the acoustic attenuation of MV cable

connectors

As illustrated in Figure 3-21a, an ultrasonic sound generator is poured

into self-mixed cylindrical silicon. A rectangle sweep signal, which has a

broadband frequency spectrum, feeds the sound generator. The mentioned

AE sensors are placed at a distance of 5cm on the surface of the silicon and

measure the generated acoustic signal. Through this, the amplitude of the

signals over the silicon (transfer function T1) can be measured.

In the second step (Figure 3-21b), the cylindrical silicon form, with the

poured sound generator, is inserted into the MV cable connector. The AE

sensors are placed on the surface of the cable connector, and the generated

58 evaluation of sensors

(a) (b)

Figure 3-21: Measurement concept of the transfer function for defining the frequency-

dependent acoustic signal amplitude over (a) ordinary self-mixed silicon and (b)

MV cable connector.

acoustic signal is measured with the sensors. This measurement includes

the transfer function T1and the acoustic attenuation of the cable connector

material (transfer function T2). To calculate T2, T1should be subtracted from

the measurement results of the second step.

The associated measurement setup for determining the signal amplitudes

on the cylindrical silicon is shown in Figure 3-22a. According to the datasheet,

the ultrasonic sound generator (KPUS-40FS-18T-447) [75] has a resonance

frequency of 40 kHz. However, investigations have shown that the sound

generator can also be operated at higher frequencies. The sound generator is

fed by a sweep rectangle signal with a start and stop frequency of 10 Hz and

1MHz, with an amplitude of 10 V with a sweep time set as 10 s. The sensors

VS30, VS150, and VS900 are used as receivers with a corresponding charge

amplifier.

Figure 3-22billustrates the second measurement step, in which the cylindri-

cal silicon containing the sound generator is inserted into the cable connector.

The acoustic sensors are placed on the MV cable connector with the same hori-

zontal distance to the sound generator as in the first step (Figure 3-22a). In the

following section, the corresponding measurement results of Figure 3-22aand

Figure 3-22bare indicated as ”measurement on silicon” and ”measurement on

the cable connector”, respectively.

Measurement results of cable connector’s acoustic signal attenuation

The results of the measurements on the silicon are presented in Figure

3-23a. The illustrated blue lines correspond to the T1transfer function. The

VS30 sensor, with an operating frequency range of 25 kHz - 80 kHz, detects

the generated acoustic signals up to 400 kHz. The highest detected signal

3.1 description of sensors 59

(a) (b)

Figure 3-22: Measurement setup for defining the acoustic signal amplitudes on (a) the cylindri-

cal silicon and (b) the MV cable connector’s surface.

amplitude of 55 dBm is around 60 kHz. The frequencies less than 14 kHz

and higher than 400 kHz are not considered, as no signal is captured in these

ranges. The VS150 sensor detects acoustic signals in a frequency range of

110 kHz – 380 kHz with a maximum signal amplitude of 39 dBm around

160 kHz. The highest detected signal amplitude by the VS900 sensor is 23

dBm at 295 kHz. The VS900 measuring frequency range is 130 kHz – 410 kHz.

The measurement results on the cable connector are shown with the red

lines in Figure 3-23a. Due to the acoustic signal attenuation of the cable

connector material, the associated signal amplitudes are smaller than the

measurements on silicon.

The subtraction of the cable connector measurements from the on-silicon

measurements determines the acoustic attenuation of the cable connector

material (Figure 3-23b). The signal attenuation measured with VS30 reaches

its maximum value of ca. 30 dB around 60 kHz. The calculated signal

attenuation by the sensor VS150 reveals a high attenuation value of ca. 30 dB

around 170 kHz. The highest calculated signal attenuation value by the sensor

VS900 is ca. 20 dB, around 330 kHz.

The measurement results in Figure 3-23 should be completed in future works

with different sound generators, sensors, and transmitted signal amplitudes.

The discussed conclusions give a rough estimation of the acoustic attenuation

of the cable connector material.

60 evaluation of sensors

(a) (b)

Figure 3-23: (a) Measurement results of different AE sensors to define signal amplitudes on

the silicon and the cable connector. (b) the calculated acoustic attenuation of the

cable connector material.

PD acoustic signal amplitude on the cylindrical silicon

The spectrum of the acoustic PD signals on the silicon is measured in

the next step. Comparing the corresponding PD spectrum with the defined

acoustic signal attenuation (Figure 3-23b) justifies the unsuccessful acoustic

PD measurements on the MV cable connector, as described at the beginning

of this section.

A copper wire is connected to a cable lug, and the whole structure is poured

with silicone (Figure 3-24). The cable lug is connected to an MV cable and the

HV transformer. As shown in Figure 3-24, the ground plane is connected to the

cylindrical silicon opposite the cable lug. The explained measurement setup

generates PDs with an inception voltage of around 5kV. An IEC 60270-based

measurement has shown that the placement of the AE sensor on the silicon

does not change the PD pattern shape and inception voltage. Hence, the

sensor does not generate additional discharges in the shown measurement

setup, and it was assured that the acoustic PD signals of the copper wire are

measured.

The question might here arise as to why the shown silicon test specimen is

not inserted into the cable connector for the HV measurements. The reason is

that the structure of the cable connector, including the inner screen, does not

allow HV measurements with the cylindrical silicon and the associated ground

plane.

3.1 description of sensors 61

Figure 3-24: Measurement setup for determining the amplitudes of acoustic PD signals.

The upper diagram of Figure 3-25aillustrates the PD spectrum, measured

with the VS30 sensor. The main frequency components of acoustic PD signals

are measured around 60 kHz and 117 kHz. The previously discussed acoustic

attenuation of the cable connector is shown in the middle diagram of Figure

3-25a. The expected acoustic PD signal amplitude on the cable connector

is calculated by subtracting the acoustic attenuation of the cable connector

material from the PD signal amplitude over the silicon (Figure 3-25a, lower

diagram). The predicted signal amplitude has negative values, almost over the

entire measurement spectrum. The negative values show that measuring the

acoustic PD signals on the cable connector material is impossible.

(a) (b)

Figure 3-25: Acoustic PD spectrum, attenuation, and calculated signal amplitude on the cable

connector (upper, middle, and lower diagrams, respectively) measured with (a)

the VS30 sensor and (b) the VS150 sensor.

62 evaluation of sensors

A similar observation is made with the VS150 sensor. The corresponding

measurement results are shown in Figure 3-25b. The measured PD signal

amplitude and the calculated acoustic signal attenuation are shown in the

upper and middle diagrams, respectively. The predicted signal amplitude

on the cable connector material is illustrated in the lower graph, including

negative values. Again, the negative values show that the acoustic signal

attenuation of the cable connector material is higher than the PD signal

amplitudes.

The VS900 sensor was unable to detect the acoustic PD signals. Therefore,

the measurement results of the latter-mentioned sensor are not presented.

The discussed measurement results are evidence of the high acoustic atten-

uation of the cable connector’s material. However, the acoustic signal attenu-

ation of the cable connector in its actual configuration should be higher than

the amplitudes presented in Figure 3-25. The cable connector consists of two

main parts, i.e., the screened body and the stress cone adapter. Each of the

mentioned components and the used assembly lubricants increases acoustic at-

tenuation and decreases PD acoustic signal amplitudes. Furthermore, acoustic

signal reflections should also be considered at the interfaces between the com-

ponents. Based on the results presented, acoustic PD detection is no longer

considered for the upcoming PD measurements.

3.2 measurement results (of all sensors)

In the previous sections of this chapter, nine different sensors were charac-

terized. It should be discussed which sensors measure the local PD signals

with high sensitivity, are more immune to the noise signal, and fulfill the cost

requirements of an MV PD monitoring system.

It was discussed that the three AE sensors are unsuitable for the PD detection

of MV cable connectors. A consistent measurement and evaluation method

is used to compare the performance of the remaining sensors. This method

considers the following sensor properties:

•Sensitivity: All the sensors are installed around a PD-afflicted cable con-

nector, and the PD activity is measured in the time domain over five min-

utes with all the sensors simultaneously. A threshold voltage is defined

for each sensor, set at 0.7times the maximum detected signal amplitude,

corresponding to half of the maximum signal power. The sensor’s sensi-

tivity is calculated as the mean value of the detected amplitudes higher

than the threshold voltage.

3.2 measurement results (of all sensors)63

•SNR: The SNR calculation (Section 2.7.3) is performed using the measure-

ments with a spectrum analyzer. Two measurements with each sensor are

performed to calculate the required parameters for SNR. First, the mean

value of the PD signal’s frequency components without background noise

is calculated (PPD). Second, the mean value of an external noise source (a

frequency-converter-controlled motor) in the absence of PD is calculated

(Pnoise). The difference between PPD and Pnoise gives the corresponding

SNR value.

•Cross-coupling: The sensitivity of the sensor to the PD signals of the

neighboring cable connector is defined as the PD cross-coupling. For this,

the PD spectrum of the sensor installed on a PD-free cable connector is

captured using a spectrum analyzer, while the PD source is in the neigh-

boring cable connector.

•Price of the sensor: The cost of sensors is determined by the author.

•Installation effort: The author has rated the installation effort.

3.2.1PD-Afflicted Cable Connector and Measurement Setup

For the sensitivity, SNR, and cross-coupling measurements, the implemented

fault of Figure 3-26 is used. This fault is called a cable screening wire, defined

as a local fault. The cable connector fault is generated by lying a cable

screening wire over the conductive layer of the MV cable. The stress cone

adapter is then installed over the screening wire, and the cable connector is

assembled completely.

The cable screening wire has a ground potential and leads to a high electric

field concentration and PDs, particularly on its tip. This fault type generates

PDs even in lower test voltages, around 3kV, leading to a quick breakdown.

Figure 3-26: The implementation of the local fault “cable screening wire”.

64 evaluation of sensors

Figure 3-27: Measurement setup for the characterization of the different sensors regarding their

effectivity for PD measurements.

The general overview of the measurement setup used for determining the

sensitivity of different sensors is illustrated in Figure 3-27. Several sensors

are installed for simultaneous PD measurement around the PD-afflicted cable

connector. The sensors are connected to two synchronized four-channel oscil-

loscopes, which capture the cable connector’s PD activities over five minutes.

To compare the non-conventional with the conventional PD measurement

method, the PD signals are also captured with an IEC 60270 compatible

coupling device. To compare the functionality of the self-developed HFCT

sensor, a commercial HFCT is also installed for the PD measurement.

According to the small size of the AR-MSA and the leaky feeder antenna,

the installation could be performed easily. As shown, the antennas are fixed

beside the cable connectors in the cable compartment. The coaxial sensor is

mounted around the cable connector using the related sensor housing. In this

measurement setup, the measurement port of the bushing is connected to the

inner conductor of the used coaxial measurement cable, and the coax screen is

connected to the bushing’s ground. The ohmic sensor is also connected to the

measurement device.

The used 12/20 kV MV cable is Nexans NA2XS(F)2Y with a cross-section

of 240 mm2[41]. The installed cable connector is Raychem RSTI-58xx [76],

suitable for the used MV cable. The cable connector is installed on the

24 kV/630 A SF6insulated medium voltage switchgear type MINEX of Fritz

Driescher KG [40].

3.2 measurement results (of all sensors)65

3.2.2Sensitivity of the Sensors

The sensitivities of the sensors are determined based on time-domain mea-

surements. For this, one PD event is measured with the sensors, and the

measurement results are compared in the first step. Then, based on 5min

time-domain measurements, a threshold voltage corresponding to 0.7times

the maximum detected signal amplitude is calculated for each sensor. The

mean value of the signal amplitudes higher than the threshold voltage is

calculated, and defined as the sensitivity of each sensor.

Figure 3-28ashows the PD signal captured with the self-developed HFCT

and the commercial one. The sensitivity of the self-made HFCT is around four

times higher than the commercial sensor. The commercial sensor has a higher

number of windings with a thinner wire leading to lower signal sensitivity.

The measured signal time duration is around 5µs. The rapid rising and falling

edges indicate that the self-made HFCT measures HF components of the PD

signal. The sensor’s signal also contains edges with lower rise times showing

the sensor’s ability to measure low-frequency PD signal components.

Figure 3-28band Figure 3-28crepresent the corresponding measurement

results of AR-MSA and the leaky feeder antenna, respectively. As illustrated,

both antennas detect the HF components of the PD signal reliably but with

small amplitudes. The duration times of antenna signals are around 500 µs.

The detected PD signal by the coaxial sensor is shown in Figure 3-28d. As

described earlier in Section 3.1.3, the corresponding signals are boosted with

a30 dB amplifier. Hence, the PD signal has a high amplitude. The PD signal

shape, with a duration of around 0.5µs, indicates that the sensor measures

the PD signals’ HF components.

Figure 3-28eillustrates the ability of the bushing’s capacitive divider for

PD measurement. The detected signal also has a duration of around 0.5µs

containing high-frequency components. As shown in Section 3.1.5, the ohmic

sensor operates in the low and high frequency ranges. Figure 3-28freveals the

described sensor characteristics. The PD signal has a duration of around 2.5

µs and an amplitude of ca. 50 mV.

To compare the conventional and non-conventional measured PD signals, the

signal of an IEC 60270-based coupling device is also measured. The detected

signal has an amplitude of around 60 mV (Figure 3-28g) and a duration of

approximately 3µs.

66 evaluation of sensors

(a) (b)

(e) (f)

(g)

Figure 3-28: The measurement of one PD event. (a) The commercial HFCT and the self-made

sensor signals. (b) The annular ring antenna signal. (c) The leaky feeder antenna

signal. (d) The coaxial sensor signal. (e) The switchgear’s bushing signal. (f) The

ohmic sensor signal. (g) The IEC 60270-based coupling device signal.

3.2 measurement results (of all sensors)67

The presented measurement results in Figure 3-28 reveal that using various

sensors leads to different measurement signal shapes. The frequency range of

the sensor, the used measurement cable, the impedance of the measurement

device, etc., can change the PD signal shape and amplitude.

The results of the long-time PD measurement with the HFCT sensor are

illustrated in Figure 3-29. The maximum absolute signal amplitude is 160 mV.

The corresponding defined threshold voltage, which corresponds to 0.7times

the maximum amplitude, is 112 mV. Using MATLAB, the mean value of the

detected signal amplitudes larger than the threshold voltage is calculated and

defined as the sensor sensitivity, i.e., 130 mV.

Figure 3-29: PD measurement over five minutes to define the sensitivity of self-developed

HFCT.

For simplicity reasons, the remaining sensors’ long-time measurement dia-

grams are not presented. The corresponding calculated sensor sensitivities are

summarized and noted in Table 3.3.

Table 3.3: The sensitivity of the sensors based on a PD measurement over five minutes.

Sensor name Sensor sensitivity Sensor name Sensor sensitivity

HFCT 130 mV Coaxial 2.2V

Annular-ring antenna 100 mV Switchgear’s bushing 110 mV

Leaky-feeder antenna 12 mV Ohmic 92 mV

As noticed in Table 3.3, the PD measurements over five minutes with the

antennas indicate a sensitivity of 100 mV for AR-MSA and 12 mV for the

leaky feeder antenna. The main reason for the leaky feeder antenna’s low

sensitivity is the antenna’s position being a relatively large distance from the

cable connector (Figure 3-27).

68 evaluation of sensors

The overall sensitivity of the coaxial sensor in long-time measurement is

2.2V. This high sensor sensitivity is due to the used amplifier. The sensitivity

of the bushing’s capacitive divider and the ohmic sensor is 110 mV and 92 mV,

respectively.

3.2.3Noise Sensitivity of the Sensors - SNR

In addition to the PD signal sensitivity, the immunity of the sensors to the

external noise signals is an important characteristic. The noise signals can

couple to the sensors over the MV cable (conductive noise) or in the form of

electromagnetic wave coupling. The conductive noise can be generated by, for

example, a frequency converter, rotating machines, etc. The electromagnetic

noise sources could be, among others, radio-, radar-, or telecommunication-

signals coupling to the sensors. However, due to the use of the sensors in a

closed metallic cable compartment, the direct electromagnetic coupling of the

noise signals is improbable. Hence, this work focuses on the conductive noise

and its coupling to the sensors.

The noise sensitivity measurements are performed using the cable connector

with PD fault caused by the “cable screening wire” (Figure 3-26). For this,

the PD spectrum was measured with a spectrum analyzer while the test

voltage was 10 kV. Then, the test voltage is decreased to a PD-free level,

and a frequency-converter-controlled motor placed in the same measurement

room is turned on. The generated noise signals couple to the transformer

controlling device, transformer, and MV cable. The noise signals travel to

the MV cable connector and reach the sensors. The PD and noise spectrum’s

mean values – up to the maximum measured PD signal frequency component

– are subtracted using MATLAB to get the SNR values.

The ohmic sensor’s noise sensitivity is shown in Figure 3-30. The maximum

detected PD frequency component is about 330 MHz. The blue line shows the

spectrum of PD signals caused by the “cable screen wire”. The mean value

of the PD spectrum is -57.9dBm, as illustrated by the yellow line. The red

line represents the spectrum of the external noise signal with a mean value of

-65.91 dBm (purple line). To calculate the SNR of the ohmic sensor, the mean

value of external noise is subtracted from the mean value of the PD spectrum,

i.e., 8.01 dBm.

3.2 measurement results (of all sensors)69

Figure 3-30: PD vs. noise spectrum to calculate the noise sensitivity of the ohmic sensor.

For simplicity, the measurements of the other sensors are not illustrated. The

corresponding noise sensitivity values are noted in Table 3.4. The coaxial sen-

sor has the highest noise sensitivity, followed by the AR-MSA, HFCT, and

ohmic sensor. The switchgear’s bushing and the leaky feeder antenna have

poor noise sensitivity.

Table 3.4: The noise sensitivity (SNR) of the sensors.

Sensor name Noise sensitivity Sensor name Noise sensitivity

HFCT 8.62 dBm Coaxial 14.45 dBm

Annular-ring antenna 4.93 dBm Switchgear’s bushing 5.33 dBm

Leaky-feeder antenna 1.72 dBm Ohmic 8.01 dBm

3.2.4Cross-Coupling

The PD signal cross-coupling between adjacent cable connectors is studied

in the next step. For this, two similar sensors are used simultaneously. The

first sensor is placed on the PD-afflicted cable connector, and the second is

on the neighboring PD-free cable connector (Figure 3-31a). The faulty cable

connector is under high voltage, and the PD-free connector is open-circuited.

Two similar spectrum analyzers are used to measure the sensor signals

simultaneously. The coupled signal spectrum is calculated as the percentage

of the reference PD spectrum in each frequency sample. The mean value of

the percentages over the entire measured frequency range reveals the final

signal cross-coupling value.

Figure 3-31bshows the measurement results captured with the HFCT sensor

installed on the faulty cable connector (blue line) and the sensor installed on

the neighboring cable connector (orange line).

70 evaluation of sensors

(a) (b)

Figure 3-31: (a) Measurement setup to determine PD signal coupling from the faulty cable con-

nector to the sensors installed on the neighboring PD-free cable connector. (b) The

PD spectrum measured with HFCT installed on the PD-afflicted cable connector

vs. the spectrum of the cross-coupled signal to the HFCT installed on the neigh-

boring PD-free cable connector.

The cross-coupling of PD signals to the HFCT installed on the adjacent cable

connector is 86.7%, calculated with MATLAB. For simplicity, the correspond-

ing measurement results of the other sensors are noted in Table 3.5and are

not illustrated.

The definition of the signal coupling for the used antennas is impossible,

as the antennas receive the electromagnetic emissions of all cable connectors.

Hence, a critical drawback of the antennas for PD measurement in the cable

compartment is that a faulty cable connector’s localization is not easily

possible. The antenna would detect the PD electromagnetic emissions of all

cable connectors if multiple faults were presented. This means that the PD

signal coupling is equal to 100%.

The PD signals couple to other bus bars of the switchgear and can be de-

tected on the adjacent bushing. Hence, the switchgear bushing’s capacitive

divider shows a high signal-coupling value. The measurements with the re-

maining sensors – HFCT, coaxial, and ohmic, indicated signal-coupling values

of around 90%.

Table 3.5: The PD signal cross-coupling ratio between adjacent MV cable connector sensors.

Sensor name Signal coupling Sensor name Signal coupling

HFCT 86.7% Coaxial 89.08 %

Annular-ring antenna 100 % Switchgear’s bushing 98.13 %

Leaky-feeder antenna 100 % Ohmic 90.50 %

3.2 measurement results (of all sensors)71

The discussed values in Table 3.5indicate a high cross-coupling between

sensors installed on various cable connectors. Therefore, the localization

of faulty cable connectors might be challenging. Further investigations are

necessary for future work to reduce the cross-coupling values.

3.2.5Price and Installation Effort

The author estimates each sensor’s price and installation effort based on the

information gathered during the research. The self-made HFCT can be made

with materials at a cost of less than fifteen euros. The installation of the HFCT

assumes placing the sensor around the screening cable of the MV cable, which

can be performed within a matter of minutes.

The AR-MSA requires around 0.08 m2FR-4substrate and an SMA connector.

Considering the etching process, the antenna fabrication costs less than ten

euros. Installation in the cable compartment of the MV switchgear can be

performed in a few minutes. The setup of the leaky feeder antenna in the

cable compartment can also be performed in minutes using plastic antenna

holders. The leaky feeder antenna is to be fabricated by the manufacturer and

as an individual production process. This antenna’s cost is somewhere in the

range of fifty to one hundred euros.

The coaxial sensor can be produced with less than ten euros of material,

including the required coaxial cable and the BNC connector. The 3D-printed

housing needs materials to the value of around four euros. The installation of

the coaxial sensor takes less than five minutes to complete. The switchgear

bushing is pre-installed and connected to the voltage monitoring module

of the switchgear having measuring sockets. Hence, the resulting cost and

installation effort for PD monitoring are low.

The ohmic sensor is made of the grounding wire of the MV connector, a low

ohmic resistance, a BNC connector, and a housing. Overall, the sensor has

a material cost of less than eight euros. The installation of the ohmic sensor

takes less than five minutes.

Each of the AE sensors studied in this section costs hundreds of euros.

Furthermore, the sensors have integrated amplifiers and must be supplied

with an external power source. This increases the installation effort.

72 evaluation of sensors

The summary of the presented estimations regarding the price and installa-

tion effort of each sensor is noted in Table 3.6.

Table 3.6: The author’s estimation regarding each sensor’s price and installation effort.

Sensor name Price Installation effort

HFCT Low Low

Annular-ring antenna Low Low

Leaky-feeder antenna Medium Low

Coaxial Low Low

Switchgear’s bushing Low Low

Ohmic Low Low

AE High High

3.3 discussion of measurement results and conclusion of

chapter

It was discussed that the sensors should detect the PD signals of the local

faults reliably, be cost-effective for PD monitoring, be immune to external

noise, and be easy to install. With this in mind, the obtained results of the

investigated sensors are compared by considering various characteristics, i.e.,

PD sensitivity, SNR, cross-coupling, operating frequency range, cost, and

installation effort.

The summary of the characteristics of the studied sensors is noted in

Table 3.7. The sensitivity was defined in a long-time PD measurement over

five minutes. The operating frequency range of the sensors was determined

using the CAL2B pulse generator, the GTEM cell measurements, or based

on the VNA measurements. The SNR was defined by subtracting the mean

noise amplitude from the mean PD power amplitude. The cross-coupling

is expressed as the coupling of PD signals to the sensors installed on the

neighboring PD-free cable connectors. The author has estimated each sensor’s

price and installation effort.

In Section 3.1.1, it was shown that the HFCT sensor with Ni-Zn core material

(Würth 74270191) with two turns and 1.5mm wire diameter could fulfill the

requirements of MV cable connector condition monitoring. A wide-band

frequency range from 10 kHz to 150 MHz allows the measurement of HF

components of the local PD faults with high sensitivity. Considering its SNR,

the sensitivity of the HFCT to the external noise is lower than that of other

sensors. Additionally, this sensor can be produced cost-effectively and is easy

to install.

3.3 discussion of measurement results and conclusion of chapter 73

Table 3.7: The characteristics of the studied sensors.

Sensor Sensitivity to

local faults

Cross-

coupling

Signal-to-

noise ratio

(SNR)

Frequency-

range

Price Installation ef-

fort

HFCT 130 mV 86.7%8.62 dBm 10 kHz – 150

MHz

Low Low

AR-MSA 100 mV 100%4.93 dBm 150 MHz -

2000 MHz

Low Low

Leaky

feeder

12 mV 100%1.72 dBm 30 MHz - 980

MHz

Medium Low

Coaxial 2.2V89.08%14.45 dBm 40 MHz - 300

MHz

Low Low

Bushing 110 mV 98.13%5.33 dBm 150 MHz –

500 MHz

Low Low

Ohmic 92 mV 90.05%8.01 dBm 80 MHz – 150

MHz

Low Low

VS30-SIC Not able to

detect

- - 25 kHz - 80

kHz

High High

VS150-

RIC

Not able to

detect

- - 100 kHz – 450

kHz

High High

VS900-

RIC

Not able to

detect

- - 100 kHz - 900

kHz

High High

The annular-ring microstrip and leaky feeder antenna can detect PD

measurements of cable connectors with different sensitivities. They are cost-

effective and can be installed easily. However, the most significant drawback

of the antennas is their cross-coupling issues. The PD signals of three cable

connectors in a cable compartment can be detected with one antenna, and the

faulty cable connector localization is impossible.

The coaxial sensor fulfills the requirements for the PD detection of cable

connectors. The sensor has a broadband frequency response, is easy to

install, and is not expensive. The coaxial sensor possesses the second-best

SNR value and shows adequate immunity to external noise. Because the

opening window of the sensor can be adjusted to the cable connector, the

cross-coupling to the adjacent cable connectors is low. This increases the

ability of the sensor to localize the PD-afflicted cable connector. The coaxial

sensor is cost-effective and can be installed quickly using the designed housing.

The capacitive divider of the bushing enables a sensitive PD measurement

that can be realized with low costs and installation effort. However, this inter-

face shows a low SNR value, as the conductive noise can couple to the sensor.

A further drawback of this PD sensing method is the corresponding high cross-

coupling value.

74 evaluation of sensors

The ohmic sensor is a cost-effective approach for PD detection of the cable

connector. Simple reconfiguration in the ground lead of the connector’s body

enables the production of this easy-to-install sensor. However, comparing the

ohmic to the HFCT and coaxial sensors shows the ohmic sensor has a poor

cross-coupling characteristic. Because the ground lead is not shielded and acts

as a small antenna, the PD signals of the adjacent cable connectors can also be

measured with this sensor.

The measurement results of this section have shown that acoustic PD

detection is not an appropriate approach for MV cable connectors. The sensors

VS30, VS150, and VS900 cannot detect the PD signals as the acoustic signal

attenuation of the cable connector material is high. These sensors cannot fulfill

the requirements of a low-cost PD monitoring system.

Based on the presented results, the suitable sensors for the objectives of this

work are the HFCT, coaxial-, and ohmic sensors.

HIGH-FREQUENCY CHARACTERISTICS OF THE

MV-CABLE

External noise signals and discharges can lead to a wrong PD diagnosis,

complicating the localization of the PD-afflicted cable connector. Hence, the

noise signal characteristics that might reach the sensors should be defined.

The direct coupling of electromagnetic noise waves to the sensors is im-

probable, as the sensors are placed in a nearly closed metallic enclosure, i.e.,

the cable compartment of the switchgear. Therefore, the conductive noise

reaching the sensors over the MV cable is the primary noise propagation path

that should be investigated.

The MV cable has a low-pass filter characteristic for noise signals. Hence,

the 3dB cut-off frequency of the MV cable is examined in this chapter. As a

result, it can be defined which signal frequencies can reach the sensors and

falsify the PD diagnosis.

In [77,78], the high-frequency characteristics of the XLPE cable were

discussed based on time-domain measurements. The frequency-domain

measurements of the XLPE cable were presented in [79,80]. To the author’s

knowledge, the main problem in the cited studies is the implementation of

HF-HV connection, with which the measurement devices are connected to the

XLPE cable. The used connection interfaces introduce significant impedance

mismatches in the higher frequency ranges.

In this chapter, an HF-HV connection method is developed and tested up to

4GHz. The 3dB cut-off frequency of the MV XLPE cable is measured in the

frequency domain using a VNA.

4.1 connection interface between mv cable and measurement

device

The main challenge for the characterization of MV cable, considering its HF

properties, is connecting the MV cable to a measurement device, for example,

a VNA, with N-type coaxial connectors.

76 high-frequency characteristics of the mv-cable

Figure 4-1: The cross-section views of the MV cable and an N-type coaxial connector.

As shown in Figure 4-1, the inner conductor of the N-type connector has

a diameter of 2.4mm, while the inner diameter of the MV cable is 18 mm.

Furthermore, the distance between the inner- and the outer conductor is

4.8mm for the N-type connector and 5.5mm for the MV cable. The con-

nection of the discussed geometries introduces a challenge because the inner

and outer conductors (ground) should be connected, and discontinuities in

the connection interface should be avoided. The discontinuities generate

impedance mismatch, signal reflections, and emissions in higher frequencies

and prevent a precise measurement.

An adapter, illustrated in Figure 4-2, is designed to realize the precise HF

measurements of the MV connector. The adapter is connected to the MV cable

on one side and an N-type connector on the other. The adapter consists of two

cone-shaped aluminum parts – the red and blue parts in Figure 4-2. The inner

part of the adapter connects the MV cable’s aluminum conductor to the N con-

nector’s inner conductor. The outer part of the adapter enables the connection

of the MV-cable shield with the ground of the N connector.

Figure 4-2: The cross-section view of the designed adapter.

4.1 connection interface between mv cable and measurement device 77

The idea of the proposed adapter is based on the Klopfenstein taper pre-

sented in [81] that is simulated and optimized with 3D field simulation soft-

ware. The realized taper enables the impedance matching between the MV ca-

ble and the N connector. One of the main parameters considering the 50 Ohm

matching of the MV cable is the length of the adapter. Generally, the longer

the taper, the better the matching considering signal reflection and attenuation.

However, fabricating long tapers is more challenging, and a trade-off should

be made between the taper length and the manufacturability. The produced

inner and outer tapers have a length of 12 cm and 15 cm, respectively.

4.1.1Characterization of the developed adapters

For the measurements of this section, two adapters are manufactured (Figure

4-3a), which are connected by a short MV cable with a length of 5cm. The

adapters are measured with a VNA up to a frequency of 4GHz. The VNA is

calibrated with the corresponding measurement coaxial cables.

The measurement result of the two adapters and the 5cm MV cable is

presented in Figure 4-3b. The insertion loss measurement indicates a 3dB sig-

nal amplitude loss at approximately 900 MHz. The discussed measurements

demonstrate the suitability of the fabricated adapter for the HF measurements

of the MV cables at least up to 900 MHz.

(a) (b)

Figure 4-3: The measurement setup for testing the manufactured adapters regarding their (a)

Calibration Adapter and (b) Insertion Loss (S21).

For further measurements, the discussed signal attenuation losses of the pro-

posed adapters should be considered. Therefore, the measurement setup, in-

cluding the adapters, is calibrated once more.

78 high-frequency characteristics of the mv-cable

4.2 cut-off frequency of 10 m mv xlpe cable

For the measurement of the MV cable cut-off frequency, it is supposed that, in

practice, the cable has a minimum length of 10 m. The longer the cable, the

lower the 3dB cut-off frequency [77].

Considering this, an MV cable with a length of 10.05 m is prepared. The

additional 5cm length is because of the used MV cable in the calibration setup

(Figure 4-3a), which should be considered in the measurements. The adapters

are installed on both ends of the MV cable and are connected to the VNA ports

(Figure 4-4). The measurement frequency range begins from 9kHz and has

a stop frequency of 900 MHz, which is the adapters’ supported –6dB matching.

The associated insertion loss measurement of the 10 m cable is illustrated

in Figure 4-4b. The corresponding insertion loss at the lowest measured

frequency of 9kHz is around 0.2dB. At the highest measured frequency, the

insertion loss is around -50 dB. The corresponding 3dB signal loss of the

calculated insertion loss envelope for 10 m MV XLPE cables is at ca. 52 MHz1.

(a) (b)

Figure 4-4: (a) The measurement setup with the 10 m MV cable. (b) The corresponding inser-

tion loss measurement.

The peaks and dips in the diagrams of the insertion loss refer to the

impedance transformation along the measured cables. The peaks show the

frequencies with lower signal attenuation and good matching, while the dips

occur at higher impedances with increased signal losses. This behavior fol-

lows the quarter-wavelength theory explaining that one quarter-wavelength

1Based on the presented measurement results and test method, further investigations in future works can

be done to extract the mathematical function of frequency-dependent MV cable attenuation. A similar

function for an RG58 coaxial cable is discussed in [82].

4.3 conclusion of chapter 79

transforms the load to a mismatched impedance, whereas the next quarter-

wavelength transfers the impedance back to 50 Ohm. According to [83], the

peaks occur if the cable length is an odd number of quarter-wavelengths. The

corresponding cable length can also be calculated based on the frequency dif-

ference between the peaks. For simplicity, the mentioned calculations are not

presented here2.

4.3 conclusion of chapter

External disturbances can reach the sensors over the MV cable, complicating

the localization of the PD-afflicted cable connector. The insertion loss mea-

surement (S21) indicates a 3dB cut-off frequency of 52 MHz for an MV cable

with a length of 10 m. Based on this result, it can be expected that sensors

working in the frequency ranges above 52 MHz will be less influenced by dis-

turbances transmitted over the MV cable compared to sensors operating in

lower-frequency ranges than 52 MHz.

2Detailed explanation and visualization regarding impedance transformation are given in

“www.microwaves101.com/encyclopedias/cable-length-rule-of-thumb” (last visited in 03.2022).

PD MONITORING SYSTEM

The existing monitoring systems for PD data acquisition are expensive. Such

costly devices cannot fulfill the expected costs for PD monitoring in the

MV sector. Hence, a low-cost monitoring system prototype was developed.

This section proposes the corresponding principles of analog hardware for

a multi-sensor data acquisition system. The digital and software of the

developed prototype are being documented by B. Böttcher, the author of [84],

in his research work.

Based on the previously achieved results, the amplification and filtering

of the sensor signals are discussed first. The advantages of an analog peak

detector for minimizing the system expenses are then explained. Further,

the structure of the developed PD detection system is discussed. Finally, the

system performance is verified by the PD signal measurements.

5.1 the principle of the frequency-selective

multi-sensor pd measurement

The frequency-selective approach contains analog filters that attenuate the

external noise signals. Based on the sensors and MV cable characteristics, the

corner filter frequencies are defined and explained in the following paragraphs.

In Chapter 3, it was shown that the three nominated sensors operate in

different operating frequency ranges (Figure 5-1a– lower diagram). It was

discussed that the HFCT sensor covers the frequency range of 10 kHz –

150 MHz, while the coaxial sensor detects the HF frequency components

between 40 MHz and 300 MHz. The ohmic sensor has a higher sensitivity in

the 80 MHz – 150 MHz frequency range.

In Chapter 4, the HF (noise-) signal propagation over the MV cable connector

was studied. It was discussed that an MV cable has a low-pass filter character-

istic. It was shown that an MV cable with a length of 10 m has a 3dB cut-off

frequency at around 50 MHz, i.e., the highest noise signal frequency that may

reach the sensors. The mentioned frequency ranges are illustrated in the upper

diagram of Figure 5-1a.

5.1 the principle of the frequency-selectivemulti-sensor pd measurement 81

(a) (b)

Figure 5-1: (a) The frequency range of the sensors, PD, and noise signals. (b) The implemented

filtering frequencies for the frequency-selective multi-sensor PD measurement.

To suppress the conducted noise signals with frequency components less

than 50 MHz, the coaxial and ohmic sensor signals are bandpass filtered with

corner frequencies 50 MHz – 80 MHz and 80 MHz – 100 MHz, respectively

(Figure 5-1b– lower diagram). The filtering corner frequencies are so chosen

that the two mentioned sensor signals do not overlap. Different measurement

spectrums increase the immunity to a specific noise signal and enable a

redundant decision-making approach.

The bandpass filters also suppress the PD’s low-frequency signal com-

ponents originating from the PD-afflicted cable connector. However, the

associated high-frequency signal components pass through the filter. Hence,

the filter outputs contain the HF PD signal components of the faulty cable

connector and no external noise signals (Figure 5-1b– upper diagram).

It could be the case that an HF noise signal disturbs the coaxial and ohmic

sensors and affects their signals. Hence, the HFCT sensor should cover the

low-frequency ranges, enabling a redundant measurement. Furthermore,

HFCT measurements in the LF range allow the monitoring system to detect

possible PD signals from the surrounding MV devices and components, for

example, cable joints. Therefore, as shown in the lower diagram of Figure 5-1b,

a bandpass filter in the LF range of 10 kHz -10 MHz is applied to the HFCT

measurement channel.

82 pd monitoring system

5.2 effect of signal filtering

The presented filtering strategy is an important principle of the applied

frequency-selective multi-sensor measurement. The combination of HFCT-

coaxial or HFCT-ohmic sensors measures the different PD frequency ranges

and is less sensitive to external noise signals. The simultaneous appearance of

two signals, captured with different sensors and frequency ranges, indicates a

local broadband emission source, i.e., PD. If the HFCT detects PD-like signals

and the other sensor does not show any signal activity, PDs in the adjacent

MV components are possible. The proposed redundant measurement method

enables a more reliable decision-making PD diagnostic.

To illustrate the effect of the proposed signal filtering, a PD-like signal, i.e.,

the CAL2B pulse (Figure 5-2a), is bandpass filtered in MATLAB. A Butterworth

filter 3rd order filters the signal in three frequency ranges, i.e., 10 kHz – 10

MHz, 50 MHz – 80 MHz, and 80 MHz – 100 MHz. As shown in Figure 5-

2b, the low-frequency components (up to 10 MHz) of the pulse contain the

most signal energy. Figure 5-2cand Figure 5-2dillustrate that the higher the

filtering frequency range, the lower the signal energy, amplitude, and time

duration. Hence, the HF-filtered signals should be amplified.

(a) (b)

Figure 5-2: (a) The signal of the CAL2B pulse generator. The pulse is bandpass filtered, and the

corresponding time-domain signals are illustrated separately for frequency ranges

of (b)10 kHz – 10 MHz, (c) 50 MHz – 80 MHz, and (d) 80 MHz – 100 MHz.

5.3 the principle of the logarithmic peak detector 83

5.3 the principle of the logarithmic peak detector

The digitalization of HF signals with short time durations assumes high-speed

analog to digital converters (ADCs), increasing the monitoring system’s costs.

To avoid the use of high-performance ADCs, a logarithmic peak detector can

be used. The output of such analog chips is proportional to the logarithm

of the input signal peak amplitude. The advantage of a logarithmic peak

detector is its high dynamic range. Low and high input signal amplitudes

will be logarithmically scaled in a specific configurable amplitude range. The

decay time of the output peak amplitude can be adjusted using an external

RC circuit. More detailed information considering the internal circuits of the

logarithm peak detectors can be found in [85].

The general functionality of a logarithmic peak detector is illustrated in

Figure 5-3. For example, a PD signal with a duration of around 2ms, a small

amplitude of ca. 100 mV, and high-frequency signal components is the input

of the peak detector. The associated output is the logarithmic envelope of the

input signal with a duration of around 12 ms, an amplitude of ca. 500 mV,

and a DC offset of about 200 mV.

Comparing the input and output signal of the logarithmic peak detector of

Figure 5-3reveals that the digitalization of the output signal requires lower

sampling rates than the digitalization of the input signal. Hence, an ADC with

a lower sampling rate can be used, and the monitoring system costs can be

reduced.

Figure 5-3: The principle of a logarithmic peak detector.

84 pd monitoring system

5.4 the principle of the developed pd measurement system

A measurement device with four PD input channels is developed based on

the described filtering and peak detection method (Figure 5-4). Moreover, one

input port is designed for the 50 Hz reference voltage synchronization.

Deriving two parameters from the reference voltage – the effective value and

the sine wave phase – is essential for further AC PD signal processing. Hence,

as presented in Figure 5-4, the sine voltage is captured using a capacitive

voltage divider. The corresponding output is connected to a voltage follower

circuit, which transmits the 50 Hz wave to the ADC.

Figure 5-4: Block diagram of the developed PD measurement system.

As shown in Figure 5-4, the sensor signals are bandpass filtered and

amplified in the first stage. Next, the signals are transferred to the mea-

surement device, where the sensor signals are measured utilizing a peak

detector with logarithmic amplification functionality. The fifth input channel

of the measurement device is connected to the PD output of an IEC 60270

compatible coupling device (CD), enabling the comparison of the multi-sensor

measurements to the conventional PD measurement method.

The analog-to-digital conversion and the phase-resolved pulse sequence

(PRPS) data extraction are performed using two LPC-Link2boards of NXP

semiconductors. The integrated 12-bit ADCs with sampling frequencies of

10 MHz digitize the generated signal envelopes. The LPC-Link2boards have

5.4 the principle of the developed pd measurement system 85

ARM Cortex-M4processors with which the PRPS is extracted from the peak

values of the sampled envelope signals and the correlated sine reference

voltage. The PRPS data stream is transferred to the second single-board

computer (Raspberry PI) to combine the stream of the LPC-Link2boards.

The single-board computer can evaluate the signals directly or transfer the

data sequence to a computer. Further information will be presented in the

dissertation of B. Böttcher, the author of [84].

Figure 5-5visualizes the different signal processing steps. The sensors detect

the original PD signals in the first step (Figure 5-5a). Next, the peak detectors

generate the envelopes of filtered and amplified PD signals (Figure 5-5b). In the

third step, the PRPS extraction is performed. From each test voltage-dependent

measurement period, the RMS value of the sine voltage (ui) and up to 1000 PD

intensity values (qi) and their correlated phase angles (ϕi) are extracted from

the peak values of the PD envelopes. The PRPS data stream qi(ti,ϕi,ui)is

transferred to the Raspberry PI.

(a) (b)

(c)

Figure 5-5: (a) The PD signals and the corresponding. (b) PD envelopes. (c) PRPS measurement

of PD envelope.

The developed measurement system contains PD analysis software that re-

ceives the data stream of the measurement device. In addition, multiple visual-

ization diagrams, for example, PRPD, ∆u/∆ϕ [14], etc., enable the user to per-

86 pd monitoring system

form a PD live interpretation of the measurements. The measurement can be

stored on the local hard drive. Further analysis tools are implemented to create

a database for the data classification. The software also enables an on-the-fly

classification and risk assessment. A detailed study of various visualization

and classification possibilities of the developed software will be presented in

the dissertation of B. Böttcher, the author of [84]. Furthermore, the software

enables the export of data to MATLAB for advanced scientific investigations1.

5.5 practical investigations

The proposed frequency-selective method’s effectivity is examined using a PD-

afflicted cable connector and external discharges. For this, two measurements

are performed successively.

First, an MV cable with a length of around 2m and a PD-afflicted cable

connector is connected to the MV switchgear and the transformer. The im-

plemented PD fault is the cable screening wire described earlier in Section 3.2.

Partial discharges of the cable connector are measured with HFCT, coaxial, and

ohmic sensors (Figure 5-6a).

The second measurement step includes the measurements of the external

discharges with the mentioned sensors and using an MV cable with a length

of 2m and a PD-free cable connector. The external PD source, i.e., a needle

on the ground plate (Figure 5-6b), is implemented at the interface, which

connects the MV cable to the transformer.

As shown in Figure 5-6c, sensor signals are divided into two equal parts

using Wilkinson power splitters. One output of each Wilkinson power splitter

transfers the original sensor signal to the oscilloscope. The second output of

the power divider is connected to the PD measurement hardware, including

the bandpass filters, amplifiers, and logarithmic peak detectors. The oscillo-

scope also captures the output signals of the peak detectors. The DC offset

of the envelope signals is filtered for better visualization of the sensor signals.

Two synchronized oscilloscopes are used to capture the corresponding sensor

signals simultaneously.

1The author of this thesis primarily focused on studying and developing the sensors, designing the filter-

ing and amplification hardware, assembling the complete measurement system, and creating a MATLAB

user interface for PD signal processing and visualization. Bjoern Boettcher was responsible for develop-

ing the software for the LPC-Link2boards and Raspberry Pi, as well as the PC software for analyzing

and classifying PD signals.

5.5 practical investigations 87

(a) (b)

(c)

Figure 5-6: (a) The measurement setup consisting of a PD-free cable connector and the cable

connector with the “cable screening wire” fault connected to the switchgear. (b)

The connection of the MV cable to the HV transformer and the generation of exter-

nal discharges using a needle on the ground plane. (c) Schematic diagram of the

measurement setup to capture the PD signals with/without filtering and amplifi-

cation.

5.5.1Measurement Results of HFCT

Figure 5-7illustrates the HFCT measurements. The cable connector’s PD

signal, in Figure 5-7a, has an amplitude of around 90 mV and a duration of

about 1µs. Figure 5-7bshows the envelope of the bandpass-filtered PD signal

with an amplitude of about 600 mV and a time duration of approximately 12

µs.

The external discharge of the needle on the ground plane captured with

HFCT is illustrated in Figure 5-7c. The PD signal amplitude is around 10 mV

with a duration of less than 1µs. The logarithmic amp generates an envelope

amplitude of roughly 400 mV with a duration time of ca. 8µs. As shown,

88 pd monitoring system

(a) (b)

Figure 5-7: Measurement results of the HFCT sensor. (a) One PD signal of cable connector

with the fault “cable screening wire”. (b) The envelope of the bandpass filtered

and amplified PD signal. (c) One external discharge signal is in its original form,

caused by a needle on the ground plane. (d) The envelope of the bandpass filtered

and amplified discharge signal caused by a needle on the ground plane.

the HFCT signal bandpass filtering (10 kHz - 10 MHz) does not affect the

measurement of the external discharge signal.

The advantage of the envelope detector is visible in Figure 5-7. For the

digitalization of the signal envelope, it is sufficient to use a cost-effective, slow,

and low-resolution ADC rather than a high-performance ADC needed to

acquire the original PD signal.

5.5.2Measurement Results of the Coaxial Sensor

Figure 5-8illustrates the measurement results with the coaxial sensor. The

cable connector’s PD signal with an amplitude of around 30 mV and a

duration time of ca. 1.5µs is shown in Figure 5-8a. The PD signal is bandpass

5.5 practical investigations 89

(a) (b)

Figure 5-8: Measurement results of the coaxial sensor. (a) One PD signal of the faulty cable

connector “cable screening wire”. (b) The envelope of the bandpass filtered and

amplified PD signal. (c) One external discharge signal is in its original form, caused

by a needle on the ground plane. (d) The envelope of the bandpass filtered and

amplified discharge signal caused by a needle on the ground plane.

filtered in the 50 MHz – 80 MHz range and transferred to the envelope

detector. As illustrated in Figure 5-8b, the associated signal amplitude and

duration time are around 400 mV and 10 µs, respectively.

The measurement of the external discharge signal with the coaxial sensor

is shown in Figure 5-8c. The time duration of the discharge signal is around

1.5µs with an amplitude of approximately 9mV. The MV cable attenuates

the HF components, and mainly the LF discharge signal components reach

the sensor. The bandpass filter suppresses the LF components. Therefore, the

corresponding envelope (shown in Figure 5-8d) has a low amplitude, around

30 mV.

90 pd monitoring system

5.5.3Measurement Results of the Ohmic Sensor

The measurement results of the ohmic sensor are shown in Figure 5-9. The

cable connector’s PD signal with an amplitude of ca. 80 mV and time duration

of approximately 1.5µs is illustrated in Figure 5-9a. The envelope of the

bandpass-filtered PD signal of Figure 5-9bhas an amplitude of around 100 mV

and a time duration of about 6µs.

The sensor can detect the LF signal components of the external discharge

(Figure 5-9c). The bandpass filter attenuates the external discharge signal.

Therefore, the corresponding envelope (Figure 5-9d) has almost an amplitude

of around zero volts.

(a) (b)

Figure 5-9: Measurement results of the ohmic sensor. (a) One PD signal of the faulty cable

connector “cable screening wire”. (b) The envelope of the bandpass filtered and

amplified PD signal. (c) One external discharge signal is in its original form, caused

by a needle on the ground plane. (d) The envelope of the bandpass filtered and

amplified discharge signal caused by a needle on the ground plane.

5.6 conclusion of chapter 91

5.6 conclusion of chapter

This chapter detailed the principles of the developed PD monitoring system.

It was shown that the coaxial and ohmic sensors detect the PD signals of the

PD-afflicted cable connector. The applied frequency-selective approach leads

to the low sensitivity of these sensors to external discharges. The HFCT sensor

captures the cable connector’s PD signals and external discharges with high

sensitivity.

Utilizing frequency-selective filtering and peak detection enables the imple-

mentation of a low-cost PD detection system. This system can detect MV cable

connectors’ PD with a much higher sensitivity than external discharges or dis-

turbances that couple to the sensor over the MV cable.

MULTI-SENSOR SIGNAL ANALYSIS

PD-afflicted cable connectors should be maintained immediately, otherwise,

they can damage the switchgear and the energy system equipment. The

detection of PDs assumes the use of a reliable monitoring system. External

PDs or noise signals should not trigger a false alarm by the monitoring system

and cause additional maintenance costs. Considering this, a simple method

for monitoring and distinguishing cable connectors’ PD signals and external

discharges and interferences is discussed in Chapter 5.

The proposed multi-sensor frequency-selective approach is tested on differ-

ent PD-afflicted cable connectors and external discharges in the university’s

laboratory as well as in the industrial test laboratory CESI / IPH. The suitability

of the mentioned PD detection approach in the presence of external noise is

also studied.

This section first explains the laboratory test setup, including the noise

sources. Next, different PD-afflicted cable connectors and external PD faults

are described and examined with the proposed measurement system. Based

on the results, the functionality of the monitoring approach is discussed. In

the next step, the verification measurements in CESI / IPH laboratory are pre-

sented. Finally, the summary of all measurements and a conclusion of this

chapter are presented.

6.1 measurement setup

The laboratory measurement setup is illustrated in Figure 6-1. A controlling

device sets the voltage of the single-phase 100 kV transformer. A series resistor

(Zn) limits the transformer current. The MV cable is connected to Znon one

side and a cable connector on the other. The cable connector is plugged into

the MV switchgear. In this work, a 24 kV SF6insulated Minex switchgear of

the company Fritz Driescher KG is used [40].

6.1 measurement setup 93

Different sensors are installed on and over the MV cable connector, as ex-

plained in Chapter 2. The sensor signals (HFCT, coaxial, and ohmic) are band-

pass filtered and amplified in a primary stage. Next, the associated signals

are transferred to the multi-channel PD measurement system. The coupling

device (CD), connected to a 1nF HV capacitor, enables the caption of 50 Hz

reference voltage (Uref). Furthermore, CD allows the IEC 60270-based PD de-

tection needed to compare and evaluate the performance of the developed

measurement system.

Figure 6-1: The applied measurement setup for the multi-sensor PD measurements.

The measurement box captures the signals of the different sensors and trans-

fers the data to an expert analysis tool, as explained in the previous chapter.

This analysis tool possesses a MATLAB interface for further analytic investi-

gations. This interface is used for the evaluation of measurement results in

this work. The noise immunity of the developed PD measurement approach

is evaluated with two different noise sources explained in the following. Fur-

thermore, external corona discharges in the measurement setup are also used

as an additional disturbance source.

6.1.1First Noise Source: Frequency-Converter-Controlled Motor

The first noise source is a frequency-converter-controlled motor (FCM) work-

ing beside the HV transformer’s controlling device. The generated noise

couples over the supply voltage cable to the controlling device’s supply

voltage cable and the pre-described measurement setup (Figure 6-2).

94 multi-sensor signal analysis

(a) (b)

(c)

Figure 6-2: The generated noise signal by the frequency-converter-controlled motor measured

with HFCT in (a) the time domain, (b) the frequency domain, and (c) the corre-

sponding PRPD.

Figure 6-2aillustrates the signal shape of the FCM noise in the time domain,

captured with a current transformer installed around the supply voltage ca-

ble of the FCM. Using a spectrum analyzer, the corresponding spectrum is

also measured (Figure 6-2b). The FCM noise has frequency components up to

10 MHz. Figure 6-2cillustrates the PRPD of FCM noise measured with HFCT

installed on a PD-free cable connector. The noise signal is spread over the

entire sine period and reaches amplitudes of around -38 dBm.

6.1.2Second Noise Source: Rectangular Burst Signal

The second noise source generates noise signals with frequency components

up to 110 MHz. This noise signal is implemented because the coaxial and

ohmic sensors measure the PD signals in higher frequency ranges, i.e., 50 MHz

-100 MHz. Hence, the corresponding noise sensitivity of the PD monitoring

system in higher frequency ranges can be studied in a worst-case noise

situation.

6.1 measurement setup 95

(a) (b)

(c)

Figure 6-3: The burst signal induced in the measurement setup as a noise signal measured

with HFCT in (a) the time domain and (b) the frequency domain, and (c) the corre-

sponding PRPD.

As illustrated in Figure 6-3a, a rectangular burst signal (RBS) is generated

with a function generator. The burst signals are induced to the measurement

setup using a current transformer installed around the HV transformer’s pri-

mary windings cable. The associated spectrum of the RBS noise is shown in

Figure 6-3bhaving frequency components up to around 110 MHz. The PRPD

of the RBS noise is shown in Figure 6-3c, including signals spread over the

whole sine period.

6.1.3Third Noise Source: Corona Discharges

Corona discharges cause the third external disturbance signal. As shown in

Figure 6-4a, a metallic object, equipped with a needle on top, is placed on the

connection interface that joins the MV cable to the transformer. The MV cable

length between the needle and the sensors is around 2m.

96 multi-sensor signal analysis

(a) (b)

(c)

Figure 6-4: (a) Measurement setup for implementing external corona discharges. (b) The corre-

sponding frequency spectrum and (c) the related PRPD are measured with HFCT.

The corresponding PRPD of the corona discharges measured with HFCT

is shown in Figure 6-4c. As discussed in Section 2.1, the PRPD of corona

discharges contains pulses around the sine angle of 270°, and further voltage

increase leads to discharges around 90°.

6.1.4Spectral Overview

Figure 6-5shows a spectral overview of the measurement setup. In Section

3.1.2, it was demonstrated that the PD-afflicted cable connectors and external

discharges have signal frequency components of up to 250 MHz and 650 MHz,

respectively. Furthermore, the FCM and RBS noise signals possess frequency

components of up to 10 MHz and 110 MHz, respectively. The MV cable with

low-pass characteristic filters the signal frequency components of the external

PDs and RBS noise higher than 50 MHz. The HFCT channel measures the

external PD, noise, and cable connector’s PD signals in a frequency range of

10 kHz – 10 MHz. The coaxial and ohmic sensor channels detect the cable

connector’s PDs in 50 MHz – 80 MHz and 80 MHz – 100 MHz, respectively.

6.2 laboratory tests 97

Figure 6-5: The lower diagram demonstrates the frequency spectrum of external PDs and noise

signals. These external signals can reach the sensors with maximum frequency

components of up to 50 MHz because the MV cable acts as a lowpass filter. The

upper diagram represents the bandpass filtering frequencies applied to the sensor

signals. It is shown that the filtering enables the detection of local PD signals with

all sensors. The external noise- and PD signals cannot be detected by coaxial and

ohmic channels.

6.2 laboratory tests

As described previously, the investigated PD faults are categorized into local

and external. Each PD fault is acquired with the multi-sensor measurement

system. To examine the noise immunity of the proposed frequency-selective

approach, the described noise sources are used separately or in combination

with the measurements.

6.2.1Multi-Sensor Measurements of PD-afflicted Cable Connectors

Four local faults are examined in this section. Three of them are the im-

plemented faults during the cable connector assembly. The last one is an

untightened nut in the bushings fixture of the switchgear. These faults are

labeled as local because the sensors and PD locations are all located in the

same area, i.e., around the MV cable connectors.

Notch in the insulation

A notch in the MV cable insulation (Figure 6-6) changes the electric field

distribution in the location of the created notch. The investigations of this work

have shown that the closer the notch to the cable’s conductive layer, the higher

the PD occurrence possibility. Such faults can occur if, during the assembly,

sharp knives are used.

98 multi-sensor signal analysis

Figure 6-6: The implementation of the local fault “Notch in the insulation”.

The PD inception voltage of the implemented notch in the insulation fault

is around 6kV. The corresponding PRPD measured with the coupling device

and without any noise signals is illustrated in Figure 6-7.

Figure 6-7: PRPD of the PD-afflicted cable connector with the fault “notch in the insulation”

without external noise signals measured with CD.

In a second measurement step, the RBS noise and the external corona

discharges are applied simultaneously to examine the noise immunity of the

monitoring system.

The PRPDs of different sensor signals are illustrated in Figure 6-8. The

PRPD of HFCT signals (Figure 6-8a) contains the PD signals and external

disturbances. The local PD signals reach amplitudes of up to 0dBm. They

are visible around 60° and 250°. The RBS noise is distributed over the whole

sine period, while the external corona discharges have higher amplitudes and

are located between 220° and 330°. The HFCT signals are bandpass filtered

in a frequency range of 10 kHz – 10 MHz, where the spectra of the external

disturbances are present and overlap with the local PD signals.

6.2 laboratory tests 99

(a)

(b)

(c)

(d)

Figure 6-8: PRPD of the PD-afflicted cable connector with the fault “notch in the insulation”

overlapped with corona discharges and the RBS noise, captured with (a) HFCT, (b)

ohmic sensor, (c) coaxial sensor, (d) IEC 60270-based coupling device.

100 multi-sensor signal analysis

The ohmic sensor only detects local PD signals (Figure 6-8b). The corre-

sponding signal amplitudes are lower than the ones seen in HFCT PRPD.

This is because the ohmic signals are bandpass filtered (80 MHz – 100 MHz),

and the associated signal amplitudes are smaller in higher frequencies, as

explained in Figure 5-2. Although the wideband noise has corresponding

spectra in the mentioned frequency range, the external disturbance signals are

not detected. This results from the low-pass characteristic of the MV cable

discussed in Chapter 4.

The local PD signals of the notch in the insulation are also detected by the

coaxial sensor (Figure 6-8c) around 60° and 250°. The coaxial sensor signals

are bandpass filtered in the 50 MHz – 80 MHz frequency range. The coaxial

sensor detects the PD signal with higher amplitudes and intensity than the

ohmic sensor. This is because the lower the measuring frequency range, the

higher the signal amplitude, as discussed in Chapter 5. The suppression of the

external noise signals follows the same argumentation as the ohmic sensor.

Finally, the measurement results of the frequency selective approach are

compared with the IEC 60270-based measurement method, illustrated in

Figure 6-8d. The conventional measurement method detects the PD signals of

the local fault, RBS noise, and external corona discharges.

The summary of the measurement results in Figure 6-8is presented in Table

6.1. All four measurement channels detect the local PD signals, while the

ohmic and coaxial channels suppress the external disturbances.

Table 6.1: Summary of the measurements with the PD-afflicted cable connector with the fault

“notch in the insulation”.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Notch in

the insulation Local ✓ ✓ ✓ ✓

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

Longer conductive layer

The second local PD fault is caused by a longer conductive layer than

the defined length suggested in the assembly instruction of the MV cable

connector [76]. According to [76], the MV cable connector’s conductive layer

length should be 40 mm. This length is set to ensure that the edge of the

6.2 laboratory tests 101

conductive layer is under the stress cone adapter, which controls the electric

field strength at the conductive layer’s edge. A longer conductive layer (Figure

6-9) leads to PD creation at its edge with no stress control. Investigations of

this work have shown that the shorter conductive layers, compared to the

defined 40 mm length, are not as critical as the longer conductive layers.

Figure 6-9: The implementation of the local fault “Longer conductive layer”.

The PD inception voltage is around 10 kV for the PD-afflicted cable connec-

tor with a longer conductive layer. The associated PRPD measured with the

CD and without external noise signals is illustrated in Figure 6-10.

Figure 6-10: PRPD of the PD-afflicted cable connector with the fault “Longer conductive layer”

without external noise signals measured with CD.

The frequency-selective approach is applied in the second measurement step.

The implementation of noise sources in the measurements varies to examine

the system performance under different conditions. Hence, the measurements

of the cable connector with the longer conductive layer are disturbed just by

the RBS noise. The corresponding PRPD patterns of the sensor signals are

presented in Figure 6-11.

102 multi-sensor signal analysis

(a)

(b)

(c)

(d)

Figure 6-11: PRPD of the PD-afflicted cable connector with the fault “Longer conductive layer”

overlapped with the RBS noise, captured with (a) HFCT, (b) ohmic sensor, (c)

coaxial sensor, (d) IEC 60270-based coupling device.

6.2 laboratory tests 103

The HFCT detects the local PD signals and the external interferences with

high amplitudes and intensities (Figure 6-11a). The local PD faults create

wave-like patterns between 0° to 130° and 180° to 310°, while the RBS noise is

spread over the whole sine period.

The PRPDs of the ohmic and coaxial sensor show the local PD patterns and

the suppression of the external RBS noise (Figure 6-11band Figure 6-11c). Both

sensors detect the local PD signals with high intensities of around –30 dBm.

Nevertheless, the corresponding signal amplitudes are less than the HFCT

signals. This amplitude difference is because of the measurement of the high

and low frequency components of the PD signals. The IEC 60270-based mea-

surement (Figure 6-11d) shows the local PD signals and the external RBS noise.

The information extracted from the measurement results, as shown in Figure

6-11, is detailed in Table 6.2. In this summary, it is observed that each of the

four measurement channels successfully identifies the local PD signals. Addi-

tionally, notable distinctions arise as the ohmic and coaxial channels suppress

external disturbances, showcasing their insensitivity to external noise signals.

Table 6.2: Summary of the measurements with the PD-afflicted cable connector with the fault

“Longer conductive layer”.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Longer

conductive layer Local ✓ ✓ ✓ ✓

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

Cable screening wire

The third implemented local PD fault is created by lying an MV cable screen-

ing wire over the conductive layer of the MV cable (Figure 6-12). The cable

screening wire has the ground potential, leading to high electric field intensity

and PDs, especially on its tip. However, this fault type is interesting for labo-

ratory measurements and seems to not happen in actual implementation due

to assembly failure.

104 multi-sensor signal analysis

Figure 6-12: The implementation of the local fault “Cable screening wire”.

This fault type generates PDs even in lower test voltages, around 3kV, and

can lead to fast breakdowns compared to the other investigated local faults.

The associated PRPD measured with the CD and without external noise

signals is illustrated in Figure 6-13.

Figure 6-13: PRPD of the PD-afflicted cable connector with the fault “Cable screening wire”

without external noise signals measured with CD.

To examine the functionality of the monitoring system in the presence of

external disturbances, the RBS noise is applied to the measurement setup. The

corresponding PRPDs of different sensing channels are shown in Figure 6-14.

The PRPD of HFCT (Figure 6-14a) shows the captured local PD signals and

the RBS noise simultaneously. The discharges of the “cable screening wire”

fault start from 0° and 180° with increasing amplitudes of up to ca. 180° and

340°, respectively. The low-frequency components of the RBS interferences

reach the HFCT and are captured with the corresponding measurement

channel.

6.2 laboratory tests 105

(a)

(b)

(c)

(d)

Figure 6-14: PRPD of the PD-afflicted cable connector with the fault “Cable screening wire”

overlapped with the RBS noise, captured with (a) HFCT, (b) ohmic sensor, (c)

coaxial sensor, (d) IEC 60270-based coupling device.

106 multi-sensor signal analysis

The PRPDs of the ohmic and coaxial sensors are shown in Figure 6-14band

Figure 6-14c, respectively. In both diagrams, the HF components of the RBS

noise are suppressed by the MV cable low-pass characteristic. In contrast, the

HF components of the local PD signals reach the sensors and pass through the

bandpass filters.

The PRPD of the IEC 60270-based measurement (Figure 6-14d) shows

local PD signals and broadband noise. The same pattern shape as HFCT

is observed, where the discharges start from 0° and 180° and reach their

maximum amplitudes around 60° and 250°, respectively.

The summary of the measurement results of the “cable screening wire” fault

is presented in Table 6.3. It is noted that all the measurement channels detect

the local PD signals. The frequency-selective approach suppresses the external

disturbances in the ohmic and coaxial sensor measurement channels.

Table 6.3: Summary of the measurements with the PD-afflicted cable connector with the fault

“Cable screening wire”.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Cable

screening wire Local ✓ ✓ ✓ ✓

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

Not tightened bushing’s nut

As previously mentioned in Section 3.1.4, the switchgear bushing’s capaci-

tive divider is grounded on its lower side. This grounding plays a critical role

in ensuring the safe and stable operation of the switchgear. The grounding

is realized using a nut (Figure 6-15), which securely connects the capacitive

divider with the grounded switchgear’s metal body, forming a robust conduc-

tive path. However, it could be the case that during the installation process,

the mentioned nut was not tightened correctly, potentially due to human error.

A loose nut may result in fluctuating grounding, creating electrical discharges

next to the bushing, which can be detected by the installed sensors. This sce-

nario is used to realize the fourth local discharge fault, with the fault occurring

at a distance of less than 10 cm from the sensors.

6.2 laboratory tests 107

Figure 6-15: The implementation of the local fault “Not tightened bushing’s nut”.

The discharge inception voltage of the implemented fault in Figure 6-15 is

around 2kV. The corresponding PRPD measured with the CD is shown in

Figure 6-16. The local discharges have high amplitudes that exceed the ADC

measuring range, illustrated as continuous lines around 5dBm.

Figure 6-16: PRPD of the local fault “Not tightened bushing’s nut” without external noise

signals measured with CD.

To examine the functionality of the frequency-selective approach in the

presence of background noise, the FCM noise is implemented and induced

into the measurement setup. The corresponding PRPDs captured with

different measurement channels are illustrated in Figure 6-17.

The PRPD of the HFCT (Figure 6-17a) shows discharges with very high am-

plitudes in 10°-130° and 190°-310° ranges. Around 5dBm, the corresponding

ADC’s dynamic range is saturated, and higher amplitudes are not captured

correctly. The FCM noise can be observed around -50 dBm level, distributed

over the whole sine period.

108 multi-sensor signal analysis

(a)

(b)

(c)

(d)

Figure 6-17: PRPD of the local fault “not tightened bushing’s nut” overlapped with the FCM

noise, captured with (a) HFCT, (b) ohmic sensor, (c) coaxial sensor, (d) IEC 60270-

based coupling device.

6.2 laboratory tests 109

The ohmic and coaxial sensors detect the discharges of the “not tightened

bushing’s nut” with high intensities, as shown in Figure 6-17band Figure

6-17c, respectively. The discharges have high energies and are measured with

signal amplitudes of up to 0dBm. The external disturbances are not captured

with these two measurement channels.

The IEC 60270-based measurement of Figure 6-17dcontains the FCM noise.

Again, the high PD signal amplitudes saturate the associated ADC dynamic

range.

Table 6.4summarizes the presented measurement results of the “not tight-

ened bushing’s nut”. The disturbances are suppressed by the ohmic and coax-

ial measurement channels. The local discharges are captured with all the mea-

surement channels.

Table 6.4: Summary of the measurements with the local fault “not tightened bushing’s nut”.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Not tightened

bushing’s nut Local ✓ ✓ ✓ ✓

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

6.2.2Multi-Sensor Measurements of External Faults

In this section, the measurement results of three external faults are discussed,

and the functionality of the proposed frequency-selective approach is further

tested. Two external PD faults are located on the MV cable side and have a

distance of ca. 2m from the sensors. The third external PD fault is realized in

the switchgear.

Corona discharges

The corona discharges were used in the previous section in the mea-

surements of the local fault ”notch in the insulation” to generate external

disturbances. In this section, the corona discharges are measured alone using

the shown measurement setup in Figure 6-18.

As shown in Figure 6-18, a metallic object containing a needle on the

transformer-MV cable interface generates the corona discharges. In this mea-

surement setup a PD-free MV cable connector is connected to the switchgear.

110 multi-sensor signal analysis

Figure 6-18: The implementation of external fault “Corona discharges”.

The PD inception voltage of the described measurement setup is approx-

imately 7kV. The corresponding PRPD pattern of these corona discharges,

measured using the coupling device, is shown in Figure 6-19. The discharges

are predominantly visible within the phase range of 200°-300°, where the volt-

age approaches its peak negative value, as well as around 90°, corresponding

to the positive peak. These phase intervals indicate the conditions under which

the electric field strength is sufficient to ionize the surrounding air, leading to

the generation of corona discharges.

Figure 6-19: PRPD of corona discharges without external noise signals measured with CD.

To test the noise sensitivity of the measurement system, the RBS noise is ap-

plied to the measurement setup. As shown in Figure 6-20a, the HFCT detects

the corresponding corona discharges around 270° with signal amplitudes up

to around -40 dBm. The RBS noise is distributed over the entire sine period

and overlaps the corona discharges.

6.2 laboratory tests 111

(a)

(b)

(c)

(d)

Figure 6-20: PRPD of corona discharges overlapped with the RBS noise, captured with (a)

HFCT, (b) ohmic sensor, (c) coaxial sensor, (d) IEC 60270-based coupling device.

112 multi-sensor signal analysis

As expected, the HF components of the corona discharges and the broad-

band noise do not reach the ohmic and coaxial sensor measurement channels.

Therefore, the PRPDs of Figure 6-20band Figure 6-20cdo not contain specific

discharges. The PRPDs include some points distributed arbitrarily and are

probably due to the white noise of the measurement device.

The corona discharges captured with the IEC 60270-based measurement

method are illustrated in Figure 6-20d. The discharges in the negative

half-wave of the sine function – around 270° - with amplitudes up to -45 dBm

show the presence of the needle with HV potential in the measurement setup.

As with the previous measurements, the RBS noise is also present over the

whole sine period.

The summary of the discussed measurement results of the corona discharges

overlapped with the RBS noise is noted in Table 6.5. It is shown that the ohmic

and coaxial sensor measurement channels are immune to external corona

discharges and RBS noise. At the same time, these mentioned external signals

are detected by the HFCT and the IEC 60270-based measurement method.

Table 6.5: Summary of the measurements with the external corona discharges.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Corona

discharges Local ✓ ✓ X X

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

The needle on the ground

As illustrated in Figure 6-21, the connection interface between the trans-

former and the MV cable is mounted on two insulators that are installed on

a grounded copper plate. This copper plate not only provides a stable me-

chanical base but also serves as a component for ensuring proper electrical

grounding. By placing a needle on the ground plate, controlled discharges

are generated, which are intentionally used as the second external discharge

source for the purpose of testing. The distance between this artificially created

discharge source and the sensors, which are installed on a PD-free MV cable

connector, is approximately 2meters.

6.2 laboratory tests 113

Figure 6-21: The implementation of external discharges caused by the needle on the ground

plate.

The discharge inception voltage of the described measurement setup is

around 5kV. The associated PRPD (Figure 6-22) captured with the coupling

device shows discharges around the sine phase angle of 90°.

Figure 6-22: PRPD of external fault “Needle on the Ground“ without noise signals measured

with CD.

The RBS noise is applied to the measurement setup, and the functionality

of the frequency-selective approach in the presence of external discharges and

background noise is illustrated in Figure 6-23.

The PRPD of the HFCT signals is illustrated in Figure 6-23a. The corre-

sponding pattern includes discharges around 90° with high signal intensities

and amplitudes up to ca. -25 dBm. The RBS noise is distributed over the

whole sine period.

114 multi-sensor signal analysis

(a)

(b)

(c)

(d)

Figure 6-23: PRPD of the external fault “needle on the ground” overlapped with the RBS noise,

captured with (a) HFCT, (b) ohmic sensor, (c) coaxial sensor, (d) IEC 60270-based

coupling device.

6.2 laboratory tests 115

The ohmic and coaxial sensor measurement channels do not detect the

discharge signals of the needle on the ground and the RBS noise (Figure 6-23b

and Figure 6-23c). The PRPDs of Figure 6-23band Figure 6-23ccontain some

points spread arbitrarily over the sine wave due to the white noise of the

measurement device.

The PRPD of the IEC 60270-based measurements of the needle on the ground

(Figure 6-23d) includes external discharges and RBS noise with high intensities.

The summary of the discussed measurements of the needle on the ground

is noted in Table 6.6. This table shows the ability of ohmic and coaxial sensor

channels to reject external discharge signals and broadband noise. The HFCT

channel and the IEC 60270-based measurement method detect the external

discharge signals and the disturbances.

Table 6.6: Summary of the measurements with the external fault “needle on the ground”.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

The needle on

the ground Local ✓ ✓ X X

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

The sharp edges of the switchgear’s fuse

The third external fault is generated by creating sharp edges in the

switchgear’s fuse structure (Figure 6-24). The generated discharge signals tran-

sit over the SF6insulated switchgear busbar [40], reaching the sensors installed

around the corresponding cable connector through the bushing. The imple-

mented failure might not often occur in the actual operation mode. This failure

helps to explain the transmission of PD signals from the switchgear side.

Figure 6-24: The external discharge implementation in the switchgear is caused by creating

sharp edges in the fuse structure.

116 multi-sensor signal analysis

The PD inception voltage resulting from the sharp edges of the switchgear’s

fuse is around 9kV. The corresponding PRPD of the IEC 60270-based mea-

surement method with no background noise is shown in Figure 6-25.

Figure 6-25: PRPD of discharges caused by the “sharp edges of the switchgear’s fuse” without

noise signals measured with CD.

For the frequency-selective measurements, the FCM noise is applied. The

corresponding PRPD of the HFCT (Figure 6-26a) shows discharges with

high intensities and signal amplitudes up to -10 dBm. The discharges are

distributed mainly in the 20°-130° and 200°-310° phase ranges. Furthermore,

the FCM noise appears over the whole sine period. The same pattern shape

can also be observed in the PRPD of the IEC 60270-based measurements, as

shown in Figure 6-26d.

Although the discharge source has a distance of more than 1.5m from

the sensor positions, the ohmic and coaxial sensors detected the discharges

(Figure 6-26band Figure 6-26c). The ohmic measurement channel detects the

discharges with a low signal intensity and amplitudes up to around -50 dBm.

The coaxial sensor sensing channel detects PD signals with higher signal

intensities than the ohmic sensor. The external FCM noise does not appear in

the PRPDs of the ohmic and coaxial sensors.

The measurements show that the MV cable suppresses the HF components

of the external signals from the MV cable side. However, the discharge signals

originating from the switchgear side do not experience significant signal

attenuation caused by the SF6insulated switchgear construction.

6.2 laboratory tests 117

(a)

(b)

(c)

(d)

Figure 6-26: PRPD of discharges caused by the “sharp edges of the switchgear’s fuse” over-

lapped with the FCM noise, captured with (a) HFCT, (b) ohmic sensor, (c) coaxial

sensor, (d) IEC 60270-based coupling device.

118 multi-sensor signal analysis

To further examine the discussed observation, the coaxial sensor output is

connected to the spectrum analyzer. As shown in Figure 6-27, the discharge

spectrum is measured from lower frequency ranges of up to 550 MHz. This

measurement result shows that the switchgear’s structure does not attenuate

the HF components of the discharge signals.

Figure 6-27: Frequency spectrum of the discharges caused by the sharp edges of the

switchgear’s fuse.

Table 6.7summarizes the measurement results of the external discharges

generated by the “sharp edges of the switchgear’s fuse” in the presence

of FCM noise. It is noted that the discharges from the switchgear side are

measured with all channels. The ohmic and coaxial sensor channels do not

measure the FCM noise, while the other two measurement channels detect the

external noise signals.

Table 6.7: Summary of the measurements with the external discharges generated by the sharp

edges of the switchgear’s fuse.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Sharp edges of the

switchgear’s fuse Local ✓ ✓ ✓ ✓

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

6.2 laboratory tests 119

Summary of the multi-sensor signal analysis

The measurement results of this section are summarized in Table 6.8, in

which the PD faults are categorized into two groups – local and external. The

applied noise sources are the FCM and RBS interferences. In addition, corona

discharges are also used to generate external disturbances.

Table 6.8: Summary of frequency-selective multi-sensor measurements.

Signal source Signal type IEC 60270 HFCT Ohmic sensor Coaxial sensor

Notch in the

insulation Local ✓ ✓ ✓ ✓

Longer conductive

layer Local ✓ ✓ ✓ ✓

Cable screening

wire Local ✓ ✓ ✓ ✓

Not tightened

bushing’s nut Local ✓ ✓ ✓ ✓

Corona discharges External ✓ ✓ X X

The needle on

the ground External ✓ ✓ X X

Sharp edges of the

switchgear’s fuse External ✓ ✓ ✓ ✓

Disturbances External ✓ ✓ X X

✓Measurement channel detects the signals.

XMeasurement channel does not detect the signals.

The local PD faults are generated by a notch in the insulation, a longer

conductive layer, cable screening wire, and an untightened bushing’s nut. The

external discharges are caused by a needle on the HV connection interface

(corona discharges), the needle on the ground, and the sharp edges of the

switchgear’s fuse.

The applied frequency-selective approach includes the bandpass filtering

of the HFCT signals in a frequency range of 10 kHz – 10 MHz. The coaxial

and ohmic sensor signals undergo bandpass filters with corner frequencies

50 MHz - 80 MHz and 80 MHz – 100 MHz respectively.

The local PD signals reach the sensors over a distance of less than 10 cm.

The LF signal components are captured with the HFCT measurement channel,

while the ohmic and coaxial channels capture the HF signal components.

120 multi-sensor signal analysis

In contrast to the local faults, the HF signal components of the external

discharges originating from the MV cable side are attenuated by the MV cable.

Nevertheless, the corresponding LF signal components reach the sensors and

are measurable by the HFCT channel. The bandpass filters of the ohmic

and the coaxial sensor channels reject the mentioned LF signals. The same

argumentation can be applied to the external noise sources suppressed by

the ohmic and coaxial measurement channels and passed through the HFCT

measurement channel.

The presented results of Table 6.8can be summarized thusly:

• The IEC 60270-based measurement method and HFCT sensing channel

detect local and external discharges as well as background noise signals.

• The ohmic and coaxial sensing channels detect local PD discharges. The

two mentioned sensing methods reject external discharges and back-

ground noise signals.

• The ohmic or coaxial sensor is sufficient to localize the PD-afflicted cable

connector.

• With the ohmic or coaxial sensor, it is not possible to distinguish the PD-

afflicted cable connector’s discharges and the internal switchgear’s PDs,

for example, originating from the switchgear’s fuse.

• The switchgear is usually PD-tested after fabrication and before onsite

installation. Hence, the existence of PDs in the switchgear in the first

weeks of operation is improbable.

6.3 system test in high power test laboratory cesi /iph

The system’s development was performed under laboratory conditions and

tested in an environment without industrial noise conditions. Therefore, the

frequency-selective multi-sensor PD approach is tested by further measure-

ments in the industrial laboratory CESI / IPH in Berlin. The Marzahn HV

substation (380/220/10 kV) next to the laboratory increases the background

noise activity in CESI / IPH laboratory. Furthermore, rotating machines,

different MV transformers, switchgears, etc., in different laboratory sections

lead to a harsh noise environment at CESI / IPH.

6.3 system test in high power test laboratory cesi /iph 121

Measurement setup

The test object is an MV cable termination installed on a 30 m MV cable

that connects the transformer with a PD-free MV vacuum circuit breaker. The

cable termination generates PD signals with an inception voltage of around

7kV. The PD fault of the existing cable termination is not known.

The IEC 60270-based measurement captured with Omicron MPD 600 is

shown in Figure 6-28. It is illustrated that the noise level is around 10 pC, and

the PD discharges reach amplitudes of up to ca. 300 pC.

Figure 6-28: The IEC 60270-based measurement of the PD-afflicted cabler termination with the

Omicron MPD 600.

Following the conventional PD measurement, the frequency-selective

multi-sensor approach is applied. The previous chapter shows that the ohmic

and coaxial sensors are specially designed for the MV cable connectors, and

installing the sensors on the cable termination of Figure 6-29ais challenging.

In the measurement setup of Figure 6-29a, the HFCT sensor is installed

around the screen of the MV cable. The connection cable of the ohmic sen-

sor is installed on the lowest part of the cable termination, where the electric

field is controlled. The coaxial sensor and associated holder are placed around

the MV cable termination. Note that the coaxial sensor holder is not developed

for this type of cable termination.

122 multi-sensor signal analysis

(a) (b)

Figure 6-29: (a) Multi-sensor measurement setup in test laboratory CESI / IPH for PD detection

on an MV cable termination using (b) the developed measurement device.

As explained in Chapter 5, the sensors are connected to the primary filtering

and amplification hardware. Then the sensor signals are transferred to the PD

measurement device (Figure 6-29b). The measurement device has five input

channels. The first channel detects the 50 Hz reference voltage, measured with

a capacitive voltage divider on the transformer side. The second input channel

is for detecting the HFCT signals. The third and fourth channels process

the coaxial and the ohmic sensor signals, respectively. The fifth channel

was idle. The processed data is transferred with a fiber-optic channel to the

measurement PC in the control room.

Multi-sensor measurement results

The PRPD of HFCT signals is illustrated in Figure 6-30a. The captured

HFCT signals have high intensities and amplitudes of up to 0dBm. The PD

pattern is located in the 30°-130° and 210°-340° range. The background noise is

measured with the HFCT over around -40 dBm and -60 dBm amplitude lines.

As shown, the bandpass filter of the HFCT does not suppress the background

noise with low-frequency signal components.

The PRPD of the ohmic and coaxial sensor channels (Figure 6-30band

Figure 6-30c) shows the local cable termination PD signals with relatively high

signal amplitudes up to -20 dBm. The measurement results reveal the presence

of a local PD fault as HF signal components pass through the corresponding

analog bandpass filters. The background noise signals are also not captured.

6.3 system test in high power test laboratory cesi /iph 123

(a)

(b)

(c)

Figure 6-30: PRPD of faulty cable termination overlapped with background noise, captured

with (a) HFCT, (b) ohmic sensor, (c) coaxial sensor.

To ensure that the PD signals originate from the faulty cable termination

connected to the circuit breaker, a needle is placed on the other end of the MV

cable at the connection point to the transformer.

The PRPD of the HFCT measurement channel (Figure 6-31a) detected the

cable termination and corona discharges. However, the corona discharges are

not captured with ohmic and coaxial sensor channels (Figure 6-31band Figure

124 multi-sensor signal analysis

6-31c) because the MV cable suppresses the high-frequent signal components

of the corona discharges. Hence, the PD fault originates from the cable termi-

nation close to the sensors.

(a)

(b)

(c)

Figure 6-31: PRPD of faulty cable termination and corona discharges on the transformer side,

captured with (a) HFCT, (b) ohmic sensor, (c) coaxial sensor.

6.4 conclusion of chapter 125

6.4 conclusion of chapter

In this chapter, the functionality of the introduced frequency-selective multi-

sensor PD measurement approach was tested. The goal was to examine if the

discharge signals of a PD-afflicted cable connector can be distinguished from

external discharges and background noise.

In the first step, the laboratory measurement setup was explained. Three

external noise generation scenarios were implemented, disturbing the PD

diagnosis. The three sensors, HFCT, coaxial, and ohmic, were installed around

the MV cable connector. The sensors were connected to the measurement

hardware containing the bandpass filters. The HFCT signals were filtered in

an LF range of 10 kHz – 10 MHz. The bandpass corner frequencies of the

coaxial channel were 50 MHz and 80 MHz, while the ohmic sensor signals

underwent the filter range of 80 MHz – 100 MHz.

The four implemented local faults were caused by a notch in the insulation,

a longer conductive layer as proposed in the cable connector installation

guide, a cable screen wire lying over the cable insulation, and a loose ground

contact of the switchgear’s bushing holder. The PRPDs of the sensor signals

show that the HFCT measurement channel detects the local discharges and

the background noise signals. On the contrary, the ohmic and coaxial sensing

channels reject the noise signals and detect the local discharges. The reason

for this is that the MV cable filters the HF components of the noise signals.

Just the associated LF signal components reach the sensors. Hence, the HFCT

channel lets the LF noise signals pass through, and the other two measurement

channels filter the LF signals. In the case of local discharges, the HF signals

can pass through the ohmic and coaxial channel filters.

The external discharges are generated by a needle on the interface joining

the transformer to the MV cable (corona discharges), a needle on the ground

of the mentioned connection interface, and the sharp edges of the switchgear’s

fuse. The first two mentioned external charges illustrate the interferences

from the MV cable side, while the last one demonstrates discharges from

the switchgear’s side. Using the PRPDs, it was shown that the proposed

frequency-selective approach rejects the interferences from the MV cable side,

while the switchgear’s discharges can interfere with the local cable connector

discharges. The reason is because the corresponding HF signal components

are not attenuated and reach the sensors over the related bushing.

126 multi-sensor signal analysis

Based on the measurement results, a prototype was realized and tested in

CESI / IPH laboratory in Berlin, Germany. The measurement setup included

a30 m MV cable, which connects the HV transformer to a PD-free circuit

breaker. The corresponding cable termination on the circuit-breaker side was

PD-afflicted. The measurements show that the HFCT measurement channel

detects the local PD faults overlaid with the background noise. The PRPDs of

ohmic and coaxial measurement channels show the local PD fault without the

background noise.

CONCLUSION AND OUTLOOK

The power grid’s proper functionality assumes that the corresponding

assets operate stably. An unexpected component failure can cause severe

damage. For this reason, monitoring individual network components using

autonomous monitoring systems is becoming increasingly important. A

monitoring system can alert the network operator and notify of possible

problems. One of these problems could be the improper functionality of the

insulation materials of the MV and HV assets, leading to the creation of partial

discharges.

Partial discharges can be detected using different methods. One of the

most common detection approaches is based on the IEC 60270 standard – the

so-called ‘conventional measurement’ method. Equipping a power grid with

monitoring systems that work with the conventional test method requires

large investments. The recommended low-frequency measurement ranges

of IEC 60270 make the PD measurements sensitive to disturbances and PD

signals originating from different sources.

PDs can also be detected by non-conventional measurement methods, which

use sensors to capture the PD signals. The non-conventional PD detection

principles can decrease the costs of the monitoring system when compared to

the IEC 60270-based approach. The associated sensors can usually be installed

without changing the HV/MV setup. However, the sensor signals can be

distorted by interferences with frequencies in the sensor’s operating range or

by PD signals originating from various sources.

The objective of the present work was to implement a PD monitoring ap-

proach for detecting partial discharges in cable connectors of MV switchgear

based on the non-conventional PD detection principles. The system can be

realized with simple and cost-efficient sensors and measurement hardware.

Furthermore, the system sensitivity can be minimized to the interfering noise

signals.

The implementation of the mentioned goals assumes the review of the

theoretical background of PDs, associated measurement methods, sensors

and their principles, the structure of the MV cable connector, and the MV

127

128 conclusion and outlook

switchgear. Hence, these theoretical backgrounds were discussed in the first

two chapters of this thesis. In the later course of this work, the development

of the monitoring system, including sensors, and measurement hardware, was

explained.

Chapter 3examined different simple-to-install and low-cost sensors regard-

ing their suitability for PD detection of MV cable connectors. It was shown

that:

• The HFCT sensor with Ni-Zn core material (Würth 74270191) with two

turns and a 1.5mm wire diameter has better functionality than the other

examined sensor structures, considering signal sensitivity and frequency

response (bandwidth 10 kHz - 150 MHz).

• The spectrum of electromagnetic PD emissions of the faulty cable con-

nectors is in the 30 MHz – 250 MHz frequency range. The annular-ring

microstrip antenna (AR-MSA) and the leaky feeder antenna are low-cost

and compact-size antennas that can detect the PD emissions of the MV

cable connectors. However, the investigated antennas had the disadvan-

tage of receiving the PD signals of all the cable connectors, making phase

localization impossible.

• The metal housings of the sensors might produce surface discharges if

placed in regions with a high electric field density. The proper sensor

position is the lower region around the stress cone adapter, where the

electric field is controlled.

• Using a coaxial sensor (bandwidth 40 MHz - 300 MHz), the PD signals of

the MV cable connector can be detected, while no direct contact with the

cable connector is needed.

• The integrated capacitive voltage divider of the bushing enables the PD

measurement. The main drawback of the use of bushing’s capacitive di-

vider was the high PD cross-coupling ratio to the adjacent bushings.

• Using the cable connector’s grounding wire, an ohmic sensor was devel-

oped to detect MV cable connector PDs. This sensor’s bandwidth was in

the frequency range of 80 MHz – 150 MHz.

• Due to the high acoustic attenuation of the cable connector’s material, the

acoustic PD signal detection on the MV cable connector is very challeng-

ing.

• Considering the objective of this work, the three nominated sensors with

the best functionalities were the HFCT, coaxial and ohmic sensors.

conclusion and outlook 129

In the further course of this thesis, it was discussed that the background noise

signals might reach the sensors installed around the cable connector, mainly

over the MV cable. The electromagnetic noise coupling, for example, from

telecommunication and radar signals, is improbable because the sensors are

placed in the metal closed cable compartment of the switchgear. Hence, the

(noise) signal transition over the MV cable was studied in Chapter 4. From

this, an adapter for the HF characterization of the MV cable connector was

developed. It was shown that:

• The MV cable has a low-pass filter characteristic for HF (noise) signals.

The 3dB cut-off frequency of a 10 m cable was measured at around 50

MHz. It was concluded that the noise signals with frequency components

less than the mentioned corner frequency could reach the sensors.

Once the sensors and the MV cable signal transitions were studied, the prin-

ciples of the developed PD monitoring system’s hardware were discussed in

Chapter 5. From this, the multi-sensor frequency-selective approach was intro-

duced and investigated. The following points were discussed:

• To prevent the external interferences with frequency components less than

the MV cable corner frequencies from reaching the ohmic and coaxial

measurement channels, the associated sensor signals could be bandpass

filtered in 50 MHz – 80 MHz and 80 MHz – 100 MHz frequency ranges,

respectively.

• Filtering the sensor signals in the mentioned frequency range assures that

the ohmic and coaxial measurement channels capture almost no exter-

nal noise signals from the MV cable side. If high-frequency signals pass

through the filters, they are most likely from a local wideband signal

source in the cable compartment, i.e., PDs.

• It could be the case that an HF noise signal disturbs the HF measuring

channels. Therefore, the HFCT sensor measuring channel was equipped

with a bandpass filter in the LF range, i.e., 10 kHz – 10 MHz, enabling a

redundant measurement.

• After filtering the sensor signals, using a logarithmic peak detector allows

the detection of PD peak amplitude and the use of cost-effective, low-

resolution ADCs.

130 conclusion and outlook

• The general structure of the developed PD monitoring system, consisting

of analog and digital hardware, was described1.

• After data digitalization, the PRPS digital signal processing enables the

detection of the signal peaks and correlating them to the corresponding

sine wave phase angle and amplitude. The PRPS data extraction method

allows the implementation of different PD visualization and interpreta-

tion algorithms, for example, for PD classification.

• The functionality of the proposed frequency-selective approach with mul-

tiple sensors was successfully tested using PD signals of a faulty MV cable

connector and external corona discharges.

The complementary measurements in the university laboratory and indus-

trial test laboratory CESI / IPH Berlin for validating the developed PD detec-

tion approach were presented in Chapter 6. From this, different PD-afflicted

cable connectors were measured. Furthermore, external discharges were gen-

erated in the measurement setup, including a PD-free MV cable connector.

Additional noise signals were induced to the measurement setup to test the

system under a harsh background noise environment. It was shown that:

• All measurement channels capture the MV cable connector’s partial dis-

charges reliably.

• In contrast to HFCT, the ohmic and coaxial sensor measuring channels do

not detect external discharges. However, if the external discharge source

is in the switchgear, all the sensors capture them.

• Further measurements have proved that the internal structure of the

switchgear enables HF signal propagation, and these HF signal compo-

nents can reach the sensors and pass through the corresponding filter

circuits.

• The measurements in the CESI / IPH laboratory in Berlin, under indus-

trial noise conditions, have proved the functionality of the PD monitoring

system. It was illustrated that the frequency-selective multi-sensor ap-

proach had detected the local cable termination fault PD signals reliably

without capturing the background noise.

1The author of this thesis primarily focused on studying and developing the sensors, filtering and ampli-

fication hardware, assembling the entire measurement system, developing a MATLAB user interface for

PD signal processing and visualization. Bjoern Boettcher was responsible for developing the software for

the LPC-Link2boards and Raspberry Pi, as well as the PC software for analyzing PD signal quantities

and classifying PD signals.

conclusion and outlook 131

Outlook

Further investigations can be conducted considering the effect of assembly

lubricants and their electrical properties. For example, these lubricants may

introduce resistances in the kilo-ohm range between the conductive layers of

the MV cable and the stress cone adapter.

The presented frequency-selective PD monitoring with multiple sensors was

tested mainly on the example of MV cable terminations. In future work, the

suitability of the proposed approach should be tested on further MV and HV

components like cable joints and HV cable connectors. For this, the sensors

might be adjusted considering their size.

The measurement system should be further developed, and all cable connec-

tors of the switchgear should be equipped with two sensors. The proposed

frequency-selective approach should be applied to each cable connector. It

should be investigated if a faulty cable connector is recognizable, for example,

between nine cable connectors attached to the switchgear.

Furthermore, the functionality of the whole measurement system should be

tested in a long-time measurement in switchgear carrying electrical power. In

addition, the calibration of the measurement hardware should be investigated.

The presented prototype of the PD monitoring approach can be further de-

veloped and commercialized. The measurement box can be adapted to acquire

the signals of different sensors not discussed in this work. For example, ad-

ditional parameters like voltage, current, temperature, humidity, etc., can be

measured. This cost-effective measurement box can be further designed as a

monitoring hub that transfers the observed assets live state to a computing

cloud. The use of cloud-based artificial intelligence algorithms facilitates data

analysis and processing. Therefore, the introduced monitoring system can en-

able the cost-effective implementation of condition or predictive-based asset

management. Hence, the power grid operators can save unnecessary mainte-

nance costs on the assets equipped with monitoring hubs.

APPENDIX

132

ADE SIGN AND MEASUREMENT OF AR-MSA

Section 3.1.2.1explained the annular-ring microstrip antenna (AR-MSA)

for PD detection. This annex presents a detailed description of the design

procedure and the following measurement of the antenna height.

Figure A-1shows the general overview of the AR-MSA. This antenna has a

broadband bandwidth, is simple to produce and install, and is cost-effective. A

detailed study regarding radiation properties, input impedance, and analytic

analysis can be obtained in [64–66]. The antenna consists of an annular ring,

a feeding microstrip line, and a ground plane under the feeding line. An FR-

4substrate with a thickness of 1.5mm and a copper thickness of 35 µm is used.

For the AR-MSA design, the following two steps are performed. First, the

appropriate size of the annular ring is found. Second, the feeding microstrip

line is designed.

Figure A-1: General overview of an annular-ring microstrip antenna on an FR4-substrate.

134

design and measurement of ar-msa 135

In [65], as a general rule, the center operating frequency of AR-MSA is

presented as a function of the antenna’s outer radius (R). It is mentioned that

R is equal to half of the wavelength corresponding to the antenna’s center

frequency. Field simulation software is used to find the proper size of the

AR-MSA. The optimum annular structure has an outer radius of 100 mm and

an inner radius of 40 mm.

The feeding microstrip has an impedance of 50 Ωon one side and the an-

nular’s input impedance on the other. In other words, the feeding microstrip

transfers the 50 ΩSMA socket impedance to the impedance of the annular

and prevents signal reflections.

Hence, the first step would be the calculation of the annular’s input

impedance. Therefore, the annular is simulated with a concise feeding line

(around 0.5mm) with which the input impedance of the annular can be

defined. The impedance is calculated using the S11 simulation results of the

annular and Equation 10.

ZA=Z0·1+S11

1−S11

(10)

Where ZAis the required input impedance of the annular. Z0is the trans-

mission impedance (50 Ω) of the used short microstrip line, which feeds the

antenna in this simulation. The value of S11 is obtained from the simulation

results of the annular at 100 MHz. This frequency point is chosen as the

antenna should detect PD signals in the lower frequency ranges, i.e., 30 MHz

–250 MHz. The value of ZAis 96.69 Ωat the mentioned frequency point of

100 MHz. To match the annular input impedance to the 50 Ωimpedance of

the SMA socket, a microstrip line with variable width is designed - a so-called

Klopfenstein taper [81]. As illustrated in Figure A-1, the taper has a larger

width (2.8mm) on the SMA socket side, corresponding to 50 Ω. A width of

0.7mm on the annular side realizes the calculated 96.69 Ω. The change in

the taper’s width enables the impedance variation over the taper length while

minimizing the mismatch reflection.

Measurement of return loss

Figure A-2illustrates the return loss measurement and simulation results of

the AR-MSA. The measurement is performed using a vector network analyzer,

and the antenna’s bandwidth is defined. Considering -6dB matching, a wide

bandwidth of around 1.9GHz (from 150 MHz to 2GHz) is measured with this

136 design and measurement of ar-msa

antenna. This measurement does not prove the desired simulated matching

from 30 MHz.

Figure A-2: The return loss measurement and simulation of the annular-ring microstrip an-

tenna.

Measurement of effective antenna height

A common method for characterizing VHF/UHF antennas, suitable for PD

detection in transformers or GIS, is the corresponding antenna factor (AF)

[60,86]. The developed AR-MSA antenna is also characterized using the

measurement of AF. The antenna factor (Equation 11) is the electric field

strength ratio to the injected voltage on the antenna’s 50 Ωterminal [87].

AF(f) = Eincident(f)

Vinjected(f)(11)

A graphical illustration of the incident electric field strength to the antenna

and the associated voltage drop across the antenna’s terminals is presented in

Figure A-3.

According to Equation 11, a sensitive antenna with a high Vinjected value

leads to a small value of AF. For the characterization of the VHF/UHF sensors,

design and measurement of ar-msa 137

Figure A-3: Graphical illustration of the incident electric wave to an antenna and the corre-

sponding generated voltage across the antenna terminals [87].

the effective antenna height - heff(f)- is also a commonly used parameter in

the literature [88] and is defined thusly:

heff(f) = 1

AF(f)(12)

The use of the gigahertz transverse electromagnetic cell (GTEM) for measur-

ing the antenna parameter was suggested in 1992 by Bronaugh [89]. The advan-

tages of the antenna measurements in a GTEM cell are immunity to external

noise signals, the measurement place’s flexibility, and the constant electromag-

netic field impedance, i.e., 377 Ω(citeBronaugh.. The schematic of a GTEM

cell is presented in Figure A-4[90]. The RF signal generator – Port 1of the

VNA - feeds the GTEM cell at the RFin input port. The electric field (E-field)

is built between the septum and the cell’s metallic housing, and the magnetic

field is perpendicular to the E-field. To avoid reflection, absorber materials and

matching resistors are installed at the end of the cell.

Figure A-4: Schematic of a GTEM cell and the corresponding measurement setup [90])

138 design and measurement of ar-msa

As illustrated in Figure A-4, a vector network analyzer is used for the

antenna factor measurements. Port 1of the VNA generates the signals and

feeds the input of the GTEM cell. The received voltage of the antenna is

measured with port 2of the VNA. Using the S-parameter S21, the forward

voltage gain, i.e., the received power on port 2over the transmitted power

from port 1, can be calculated. The use of S21 for the definition of AFis

suggested in [91].

AF(f)dB(1

m)=|K

h·S21

|(13)

In Equation 13, h is the height of the septum in the position of DUT, and

K is a constant correction factor that considers mismatches of the GTEM cell.

The linear form of AF(f) can be expressed as [88]:

AF(f)lin =10|K

20·h·S21 |(14)

Further reading considering the proposed mathematical approach using

S-parameters can be found in [92–94].

To define the constant K in Equation 14, the antenna factor of a monopole

antenna is calculated. The monopole antenna is widely used, and its

associated AF(f) value is also known in literature [95]. According to the

definitions in [95], the antenna used in this work is a short monopole antenna

with a physical length smaller than λ/8. The heff of the short monopole is

half the physical length. Based on the information and the S21 measurement

results of the short monopole, the constant K is defined for the used GTEM cell.

After defining the constant correction factor K, the effective antenna height

of the developed AR-MSA is calculated and illustrated in Figure A-5c. In this

figure, the related measurement setup is also shown.

It can be concluded from Figure A-5cthat the antenna height in frequency

ranges 90 MHz – 180 MHz, and 550 MHz upward has a value larger than 5cm.

In the mentioned frequency bands, the antenna is more sensitive. The mean

value of heff over the measured frequency range is 5.67 cm.

design and measurement of ar-msa 139

(a) (b)

(c)

Figure A-5: (a) The used GTEM cell. (b) The position of the AR-MSA in the GTEM cell. (c) The

corresponding effective antenna height of AR-MSA.

BIBLIOGRAPHY

[1] IEC 60270:2000, High-voltage test techniques – Partial discharge measurements. Stan-

dard, International Electrotechnical Commission, Geneva, Switzerland, 2000.

[2] K. Rajagopal, Textbook of engineering physics (Part II), 1st ed. New Delhi:

Prentice-Hall Of India Pvt. Limited, 2009.

[3] S. P. Fuchs, Partial discharge and corona theory and measurement. Master Thesis, Uni-

versity of Dayton, Dayton, USA, 1993. [Online]. Available: https://ecommons.

udayton.edu/cgi/viewcontent.cgi?article=3776&context=graduate_theses.

[4] C. G. Herbert, Mass spectrometry basics, 1st ed. Boca Raton, Florida, USA: CRC

Press, 2002.

[5] L. E. Lundgaard, “Partial discharge. XIV. acoustic partial discharge detection-

practical application,” IEEE Electrical Insulation Magazine, no. Vol. 8. No.5, pp. 34–

43, 1992. doi: 10.1109/57.156943.

[6] C. Auerswald, Mikromechanischer Körperschallsensor zur

trukturüberwachung

[Micro-mechanical structure-borne sound sensor for structural monitoring]. Disser-

tation, Technische Universität Chemnitz, Chemnitz, Germany, 2016. [Online].

Available: https://nbn-resolving.org/urn:nbn:de:bsz:ch1-qucosa-205864.

[7] L. E. Lundgaard, “Partial discharge. XIII. acoustic partial discharge detection-

fundamental considerations,” IEEE Electrical Insulation Magazine, no. Vol. 8. No.

4, pp. 25–31, 1992. doi: 10.1109/57.145095.

[8] L. Pharaoh, C. Bishop, and C. Gidzewicz, Physics student book : AQA A-level Year

2, 1st ed. London: Collins, 2016.

[9] P. Dhande, “Antennas and its applications,” DRDO Science Spectrum, pp. 66–78,

2009. [Online]. Available: https://www.academia.edu/9647737/Antennas_and_

its_Applications.

[10] R. Schwarz, Messtechnik und Diagnostik an elektrischen Betriebsmitteln [Measurement

and diagnostics of electrical equipment]. Postdoctoral thesis, Technische Universität

Graz, Graz, Austria, 2009. [Online]. Available: https://online.tugraz.at/tug_

online/vak.wbdownloadvadoc?pVaNr=52663&pVaDocType=PDF.

140

design and measurement of ar-msa 141

[11] A. Küchler, High voltage engineering fundamentals - technology - applications

(Springer eBook Collection Engineering), 1st illustrated edition. Berlin, Heidel-

berg: Springer Vieweg, 2017. doi: 10.1007/978-3-642-11993-4.

[12] G. Hilgarth, Hochspannungstechnik [High voltage engineering] (Leitfaden der Elek-

trotechnik), 2nd ed. Wiesbaden: Vieweg+Teubner Verlag, 1992. doi: 10 . 1007 /

978-3-663-10316-5.

[13] K. Schon, High voltage measurement techniques: Fundamentals, Measuring Instru-

ments, and Measuring Methods (Power Systems), 1st ed. Cham, Switzerland:

Springer Nature Switzerland AG, 2019. doi: 10.1007/978-3-030-21770-9.

[14] H.-G. Kranz, “PD pulse sequence analysis and its relevance for on-site PD de-

fect identification and evaluation,” IEEE Transactions on Dielectrics and Electrical

Insulation, no. vol. 12, no. 2, pp. 276–284, 2005. doi: 10.1109/TDEI.2005.1430397.

[15] A. Pirker and U. Schichler, “Partial discharge measurement at DC voltage —

Evaluation and characterization by NoDi∗ p attern" IEEE Transactions on Dielectrics

and Electrical Insulation, vol. 25, no. 3, pp. 883–891, 2018. doi: 10.1109/

TDEI.2018. 006742.

[16] S. Markalous, S. Tenbohlen, and K. Feser, “Detection and location of partial

discharges in power transformers using acoustic and electromagnetic signals,”

IEEE Transactions on Dielectrics and Electrical Insulation, vol. 15, no. 6, pp. 1576–

1583, 2008. doi: 10.1109/TDEI.2008.4712660.

[17] IEC/TS 62478, High voltage test techniques - Measurement of partial discharges by elec-

tromagnetic and acoustic methods. Standard, British Standards Institution, London,

UK, 2016.

[18] W. R. Habel, U. Buchholz, G. Heidmann, M. Höhse, and C. Lothongkam, “Fibre-

optic sensors for early damage detection in plastic insulations of high voltage

facilities,” ISH 2011, 7th International Symposium on High Voltage Engineering,

Hanover, Germany, 2011. [Online]. Available: https://www.researchgate.net/

publication/260032947_Fibre-optic_Sensors_for_Early_Damage_Detection_

in_Plastic_Insulations_of_High_Voltage_Facilities.

[19] W.G. Hurley and W. H. Wölfle, Transformers and inductors for power electronics:

theory, design and applications. Chichester: Wiley, 2013.

[20] T. Brander, A. Gerfer, B. Rall, and H. Zenkner, Eds., Trilogie der induktiven Bauele-

mente [Trilogy of inductive components]: Applikationshandbuch für EMV-Filter, getak-

tete Stromversorgungen und HF-Schaltungen, 5. überarbeitete und erweiterte Au-

flage. Künzelsau, Germany: Swiridoff and Swiridoff Verlag, 2013.

142 design and measurement of ar-msa

[21] Pulse Electronics, Introduction to transformer magnetics, Online document, San Diego,

USA, 2.2019. [Online]. Available: https://www.pulseelectronics.com/wp-

content/uploads/2020/12/Introduction-Transformer-Magnetics.pdf.

[22] B. M. Amna and U. Khayam, “Design and simulation of high frequency cur-

rent transformer as partial discharge detector,” 3rd Conference on Power Engineer-

ing and Renewable Energy (ICPERE), 29B2–4, 2016. doi: 10.1109/ICPERE.2016.

7904854.

[23] S. Birlasekaran and w. H. Leong, “Comparison of known PD signals with the

developed and commercial HFCT sensors,” IEEE Transactions on Power Delivery,

no. 3, pp. 1581–1590, 2007. doi: 10.1109/TPWRD.2007.899795.

[24] X. Hu, W. H. Siew, M. D. Judd, and Peng, “Transfer function characterization

for HFCTs used in partial discharge detection,” IEEE Transactions on Dielectrics and

Electrical Insulation, no. 2, pp. 1088–1096, 2017. doi: 10.1109/TDEI.2017.006115.

[25] C. Zachariades, R. Shuttleworth, R. Giussani, and MacKinlay, “Optimization of a

high-frequency current transformer sensor for partial discharge detection using

finite-element analysis,” IEEE Sensors Journal, no. 20, pp. 7526–7533, 2016. doi:

10.1109/JSEN.2016.2600272.

[26] D. Goetz, H. Putter, F. Petzold, and R. Puig, “Performance of dedicated HFCT

(high current transformer) at UHF (ultra high frequency) for partial discharge

detection,”

ISH 2017,

The 20th International Symposium on High Voltage

Engineering, Buenos Aires, Argentina, 2017. [Online]. Available: https : / / www .

researchgate . net /publication / 314440484 _ Performance _ of _ dedicated _ HFCT

_ high _ current _ transformer _ at _ UHF _ ultra _ high _ frequency _ for _

partial _ discharge _ detection.

[27] M.J. Foxa, A.P. Duffy, J. Gow, M. Seltzer-Grant, L. Renforth, “Development of a

new high current, hybrid ‘ferrite-rogowski’, high frequency current transformer

for partial discharge sensing in medium and high voltage cabling,” 59th Inter-

national Wire & Cable Symposium - Rhode Island Convention Centre, Providence, RI,

USA., 2010.

[28] B. A. Siddiqui, P. Pakonen, and P. Verho, “Novel inductive sensor solutions for

on-line partial discharge and power quality monitoring,” IEEE Transactions on

Dielectrics and Electrical Insulation, no. 1, pp. 209–216, 2017. doi: 10.1109/TDEI.

2016.005908.

[29] A. Rodrigo, P. Llovera, V. Fuster, and A. Quijano, “Influence of high fre-

quency current transformers bandwidth on charge evaluation in partial dis-

design and measurement of ar-msa 143

charge measurements,” IEEE Transactions on Dielectrics and Electrical Insulation,

no. 5, pp. 1798–1802, 2011. doi: 10.1109/TDEI.2011.6032852.

[30] D. Passow, M. Beltle, S. Tenbohlen, J. Hohloch, and R. Grund, “Online and on-

site partial discharge measurement of long length power cables by using joints

with integrated PD sensors,” Proceedings of the 21st International Symposium on

High Voltage Engineering, Budapest, Hungary, pp. 219–231, 2019. doi: 10.1007/978-

3-030-31676-1{\textunderscore}21.

[31] W. z. Chang, Y. p. Gong, J. g. Bi, G. m. Ma, and Y.-c. Zhuang, “High frequency

partial discharge measurement of straight power cable joint,” IEEE Electrical In-

sulation Conference (EIC), Baltimore, MD, USA, pp. 229–232, 2017. doi: 10.1109/

EIC.2017.8004711.

[32] J.-C. Hsieh, C.-C. Tai, C.-C. Su, C.-Y. Chen, T.-C. Huang, and Y.-H. Lin, “Evalua-

tion and comparison of on-line PD detection methods for high-voltage power ca-

ble,” 12th A-PCNDT 2006 – Asia-Pacific Conference on NDT, Auckland, New Zealand,

2006. [Online]. Available: https://www.ndt.net/article/apcndt2006/papers/

72.pdf.

[33] Y. Tian, P. L. Lewin, and A. E. Davies, “Comparison of on-line partial discharge

detection methods for HV cable joints,” IEEE Transactions on Dielectrics and

Electrical Insulation, no. 4, pp. 604–615, 2002. doi: 10.1109/TDEI.2002.1024439.

[34] Y. Tian, P. L. Lewin, A. E. Davies, S. J. Sutton, and S. G. Swingler, “Partial dis-

charge detection in cables using VHF capacitive couplers,” IEEE Transactions

on Dielectrics and Electrical Insulation, no. 2, pp. 343–353, 2003. doi: 10.1109/

TDEI. 2003.1194121.

[35] C. Zhang et al., “Partial discharge monitoring on metal-enclosed switchgear with

distributed non-contact sensors,” Sensors, Basel, Switzerland, vol. 18, no. 2, 2018.

doi: 10.3390/s18020551.

[36] H. Prasetia, U. Khayam, Suwarno, A. Itose, M. Kozako, and Hikita, “PD pattern

of various defects measured by TEV sensor,” International Conference on High

Voltage Engineering and Power Systems (ICHVEPS), pp. 23–28, 2017. doi: 10 .

1109 /ICHVEPS.2017.8225861.

[37] S. Chakravorti, D. Dey, and B. Chatterjee, Recent Trends in the Condition Monitor-

ing of Transformers: Theory, Implementation and Analysis (Power Systems). London:

Springer London, 2013. doi: 10.1007/978-1-4471-5550-8.

[38] The japanese society for non-destructive, Practical acoustic emission testing. Tokyo:

Springer Japan, 2016.

144 design and measurement of ar-msa

[39] Vallen Systeme GmbH, V

30-

SIC

-46d

–sensor data sheet, Vallen Systeme

GmbH, Online document, Wolfratshausen, Germany, 12.2015. [Online].

Available: https://www.vallen.de/wp-content/uploads/2019/03/VS30-

SIC-46dB_1512. pdf.

[40] DRIESCHER WEGBERG, Medium voltage switchgear type

MINEX

G.I.S.E.L.A

SF6

insulated - rated voltage up to 24 k

- rated current 630 A, DRIESCHER

WEGBERG GmbH, Online document, Wegberg, Germany, 2017. [Online]. Avail-

able: https : / / www . driescher - wegberg . de / upload / download / PDF / 03 .%

20Montageanleitung_MINEX%20- %20GISELA%2012- 24kV%20deu-

englisch_03-2017.pdf.

[41] NKT GmbH, N

(

)

Y 12/20 - medium voltage cables with

XLPE

insulation,

NKT GmbH, Ed., Online document, Köln, Germany, 2019. [Online]. Available:

https: //nkt.widen.net/content/saz9vycnul/pdf/

NA2XS_F_2Y-12_20_DSDEEN.pdf?

u = gj0n1yhttps : / / nkt . widen . net / content / saz9vycnul / pdf / NA2XS _ F _ 2Y

-12_20_DSDEEN.pdf?u=gj0n1y.

[42] C. C. Uydur and O. Arikan, “Use of tan delta and partial discharge for evaluating

the cable termination assembly,” Energies, vol. 13, no. 20, p. 5299, 2020. doi:

10.3390/en13205299.

[43] A. Eigner and K. Rethmeier, “An overview on the current status of partial dis-

charge measurements on AC high voltage cable accessories,” IEEE Electrical

Insulation Magazine, no. 2, pp. 48–55, 2016. doi: 10.1109/MEI.2016.7414231.

[44] N. Ahmed, O. Morel, and N. Srinivas, “Partial discharge measurement in

transmission-class cable terminations,” IEEE Transmission and Distribution Con-

ference, Vol. 1, pp. 2–7, 1999. doi: 10.1109/TDC.1999.755303.

[45] L. Renforth, M. Seltzer-Grant, R. Mackinlay, S. Goodfellow, D. Clark, and R.

Shuttleworth, “Experiences from over 15 years of on-line partial discharge

(OLPD) testing of in-service MV and HV cables, switchgear, transformers and

rotating machines,” IX Latin American Robotics Symposium and IEEE Colombian

Conference on Automatic Control, Bogota, Colombia, pp. 1–7, 2011. doi: 10.1109/

LARC.2011. 6086862.

[46] C. Cheng, C.-L. Fan, H.-C. Hsiao, and W.-M. Wang, “On-site partial discharge

measurement of underground cable system,” 7th Asia-Pacific International Confer-

ence on Lightning, Chengdu, China, pp. 575–580, 2011. doi: 10 . 1109 / apl . 2011 .

6110192.

[47] Y. Zhao, Z.-c. Zhou, H.-c. Li, W.-x. Zhang, and P. Chen, “Application of PD mea-

surement system for high voltage XLPE cable,” International Conference on Electric-

design and measurement of ar-msa 145

ity Distribution (CICED), Guangzhou, China, pp. 1–4,2008.doi:10.1109/CICED.

2008.5211664.

[48] C. Suwanasi, T. Suwanasri, P. Fuangpian, and S. Ruankon, “Investigation on

partial discharge of power cable termination defects using high frequency cur-

rent transformer,” 10th International Conference on Electrical Engineering/Electron-

ics, Computer, Telecommunications and Information Technology (ECTI-CON 2013):

Krabi, Thailand, pp. 1–4, 2013. doi: 10.1109/ECTICon.2013.6559509.

[49] G. C. Montanari, “Partial discharge detection in medium voltage and high volt-

age cables: Maximum distance for detection, length of cable, and some answers,”

IEEE Electrical Insulation Magazine, vol. 32, no. 5, pp. 41–46, 2016. doi: 10.1109/

MEI.2016.7552375.

[50] M. Knenicky, R. Prochazka, and J. Hlavacek, “Partial discharge patterns during

accelerated aging of medium voltage cable system,” IEEE International Conference

on High Voltage Engineering and Application : Athens, Greece, pp. 1–4, 2018. doi:

10.1109/ICHVE.2018.8641847.

[51] S. Tenbohlen, D. Denissov, S. Hoek, and S. M. Markalous, “Partial discharge

measurement in the ultra high frequency (UHF) range,” IEEE Transactions on

Dielectrics and Electrical Insulation, vol. 15, no. 6, pp. 1544–1552, 2008. doi:

10.1109/TDEI.2008.4712656.

[52] D. Denissov, R. Grund, T. Klein, W. Köhler, and S. Tenbohlen, “UHF partial

discharge diagnosis of plug-in cable terminations,” 7th International Conference

on Insulated Power Cables, Jicable., no. 7, pp. 24–28, 2007. [Online]. Available:

https: //www.jicable.org/2007/Actes/Session_C72/JIC07_C728.pdf.

[53] M. Koelling, T. Meinl, Y. H. Malik, T. Gräf, and M. Menge, “Acoustic partial

discharge measurements on medium voltage cable connectors using fiber optic

sensors,” VDE High Voltage Technology 2018, ETG-Symposium, Berlin, Germany,

pp. 1–5, 2018. [Online]. Available: https://www.vde- verlag.de/proceedings-

en/454807143.html.

[54] M. Zhao, X. Cao, K. Zhou, Y. Fu, X. Li, and L. Wan, “Flexible sensor array based

on transient earth voltage for online partial discharge monitoring of cable termi-

nation,” Sensors, vol. 20, no. 22, p. 6646, 2020. doi: 10.3390/s20226646.

[55] G. C. Montanari, L. Cirioni, G. Galvagno, and S. Mastroeni, “Condition assess-

ment of MV switchboards: Performance of a new sensors for partial discharge

monitoring,” 2020 IEEE/PES Transmission and Distribution Conference and Exposi-

tion (T&D), Chicago, IL, USA, pp. 1–5, 2020. doi: 10.1109/TD39804.2020.9300023.

146 design and measurement of ar-msa

[56] Power Diagnostix, Partial discharge testing from power diagnostix - product cata-

logue, Power Diagnostix, Online document, Aachen, Germany, 5.202. [On-line].

Available: https://media.megger.com/mediacontainer/medialibraries/

meggerse/images/produktkataloger/powerdiagnostix_catalogue_2020_en_

digital.pdf.

[57] D. M. Pozar, Microwave engineering, 3. ed. Hoboken, NJ, USA: Wiley, 2005.

[58] ISO 11452-4:2011, Road vehicles — Component test methods for electrical disturbances

from narrowband radiated electromagnetic energy — Part 4: Harness excitation methods,

4. ed. Standard, ISO - International Organization for Standardization, Geneva,

Switzerland, 2011.

[59] Q. Khan, S. S. Refaat, H. Abu-Rub, and H. A. Toliyat, “Partial discharge detec-

tion and diagnosis in gas insulated switchgear: State of the art,” IEEE Electrical

Insulation Magazine, no. 4, pp. 16–33, 2019. doi: 10.1109/mei.2019.8735667.

[60] M. Siegel, M. Beltle, and S. Tenbohlen, “Characterization of UHF PD sensors

for power transformers using an oil-filled GTEM cell,” IEEE Transactions on

Dielectrics and Electrical Insulation, vol. 23, no. 3, pp. 1580–1588, 2016. doi:

10.1109/TDEI. 2016.005562.

[61] P. Wang et al., “Design of an effective antenna for partial discharge detection in

insulation systems of inverter-fed motors,” IEEE Transactions on Industrial Elec-

tronics, vol. 69, no. 12, p. 1, 2021. doi: 10.1109/TIE.2021.3130335.

[62] Electro-Metrics Corporation, Instruction manual - biconical antenna - model bia-30s

-20

MHz

– 300

MHz

, Electro-Metrics Corporation, Online document, New York,

2017. [Online]. Available: https://em- metalworks.com/wp- content/uploads/

2022/11/TYP-EM-6912A-INSTRUCTION-MANUAL-TYPICAL-DATA-2017.pdf.

[63] R. Azim, M. T. Islam, and N. Misran, “Printed circular ring antenna for UWB

application,” International Conference on Electrical & Computer Engineering (ICECE

2010), Dhaka, Bangladesh, pp. 361–363, 2010. doi: 10.1109/ICELCE.2010.5700703.

[64] J. Dahele, K. Lee, and D. Wong, “Dual-frequency stacked annular-ring microstrip

antenna,” IEEE Transactions on Antennas and Propagation, vol. 35, no. 11, pp. 1281–

1285, 1987. doi: 10.1109/TAP.1987.1143997.

[65] W. Chew, “A broad-band annular-ring microstrip antenna,” IEEE Transactions on

Antennas and Propagation, vol. 30, no. 5, pp. 918–922, 1982. doi: 10 . 1109 / TAP .

1982.1142913.

design and measurement of ar-msa 147

[66] S. E. El-khamy, R. M. El-Awadi, and E.-B. El-Sharrawy, “Simple analysis and de-

sign of annular ring microstrip antennas,” IEE Proceedings H Microwaves, Anten-

nas and Propagation, vol. 133, no. 3, p. 198, 1986. doi: 10.1049/ip-h-2.1986.0035.

[67] F. Hirtenfelder and S. Murray, “Leaky cables make fine broadband anten-

nas,” Microwave Journal, pp. 6–8, 2015. [Online]. Available: https : / / www .

signalintegrityjournal . com / articles / 163 - leaky - cables - make - fine -

broadband - antennas# : ~ : text = Editor ’ s % 20Note % 3A % 20A % 20cable % 20that ,

approach%20are%20called%20leaky%20feeders..

[68] J. H. Wang and K. K. Mei, “Theory and analysis of leaky coaxial cables with

periodic slots,” IEEE Transactions on Antennas and Propagation, vol. 49, no. 12,

pp. 1723–1732, 2001. doi: 10.1109/8.982452.

[69] Radio Frequency Systems, 1/2 radiaflex

RLK

cable,

-series - data sheet, Radio

Frequency Systems, Online document, Hannover, Germany, 5.2017. [Online].

Available: https://www.rfsworld.com/pim/product/html/RLK12-50JFNA.

[70] Martin Grote,

MHHPA

: Anhydrid-

Härter

für

poxidharzanwendungen [

MHHPA

Anhydride hardener for epoxy resin applications], DRIESCHER WEGBERG GmbH,

Online document, Energiehalle in der DASA , Dortmund, 2015. [Online]. Avail-

able: https://www.baua.de/DE/Angebote/Veranstaltungen/Dokumentationen/

Gefahrstoffe/pdf/Vortrag-REACH-2015-07.pdf?__blob=publicationFile&v= 2.

[71] U. B. Buchholz, M. Jaunich, W. S. Stark, W. H. Habel, and B. A. T. Petersson,

“Acoustic data of cross linked polyethylene (XLPE) and cured liquid silicone

rubber (LSR) by means of ultrasonic and low frequency DMTA,” IEEE

Transactions on Dielectrics and Electrical Insulation, vol. 19, no. 2, pp. 558–566, 2012.

doi: 10.1109/TDEI.2012.6180250.

[72] M. M. Yaacob, M. A. Alsaedi, J. R. Rashed, A. M. Dakhil, and S. F. Atyah, “Re-

view on partial discharge detection techniques related to high voltage power

equipment using different sensors,” Photonic sensors, vol. 4, no. 4, pp. 325–337,

2014. doi: 10.1007/s13320-014-0146-7.

[73] Vallen Systeme GmbH, V

S150

RIC

–

Sensor

data sheet, Vallen Systeme GmbH,

Online document, Wolfratshausen, Germany, 2015. [Online]. Available:

https: //www.vallen.de/wp-content/uploads/2019/03/VS150-RIC_1512.pdf.

[74] Vallen Systeme GmbH, Acoustic emission sensors, specification, Vallen Systeme

GmbH, Online document, Wolfratshausen, Germany, 2017. [Online]. Available:

https://www.vallen.de/wp-content/uploads/2019/03/sov.pdf.

148 design and measurement of ar-msa

[75] Conrad Electronic SE, Waterproof type ultrasonic sensor -

KPUS-40FS-18T

-447,

Conrad Electronic SE, Online document, Hirschau, Germany, 2011. [Online].

Available: https : / / asset . conrad . com / media10 / add / 160267 / c1/ - /en /

000505680DS01 / datasheet - 505680 - kpus - 40fs - 18t - 447 - ultrasonic -

transmitter-ultrasonic-transmitter-ultrasonic-transmitter.pdf.

[76] TE connectivity, Installation instruction

EPP

-0982-1/11, TE connectivity, Online

document, Munich, Germany, 2013. [Online]. Available: https://www.te.

com / commerce / DocumentDelivery / DDEController ? Action = showdoc & DocId =

Specification + Or + Standard % 7FEPP - 0982 % 7F1101 % 7Fpdf % 7FEnglish % 7FENG _

SS_EPP-0982_1101_EPP-0982.pdf%7FCM0012-005.

[77] S. Boggs, A. Pathak, and P. Walker, “Partial discharge. XXII. high frequency

attenuation in shielded solid dielectric power cable and implications thereof for

PD location,” IEEE Electrical Insulation Magazine, vol. 12, no. 1, pp. 9–16, 1996.

doi: 10.1109/57.484104.

[78] M. Fenger and J. P. Levine, “Sensitivity assessment for HV & EHV field

partial discharge measurements via laboratory testing,” 2012 IEEE International

Symposium on Electrical Insulation, San Juan, PR, USA, pp. 613–618, 2012. doi:

10.1109/ELINSL.2012.6251544.

[79] R. Papazyan and H. Edin, “Design and application of a coupling connection for

high frequency and high voltage diagnostic measurements on power cables,”

18th Nordic Insulation Symposium (NORD-IS), pp. 183–190, 2003. [Online]. Avail-

able: https://www.researchgate.net/publication/239571642_Design_and_

Application _ of _ a _ Coupling _ Connection _ for _ High _ Frequency _ and _ High _

Voltage_Diagnostic_Measurements_on_Power_Cables.

[80] R. Papazyan, P. Pettersson, H. Edin, R. Eriksson, and U. Gafvert, “Extraction

of high frequency power cable characteristics from S-parameter measurements,”

IEEE Transactions on Dielectrics and Electrical Insulation, vol. 11, no. 3, pp. 261–270,

2004. doi: 10.1109/tdei.2004.1306724.

[81] R. Klopfenstein, “A transmission line taper of improved design,” Proceedings of

the IRE, vol. 44, no. 1, pp. 31–35, 1956. doi: 10.1109/JRPROC.1956.274847.

[82] D. Large and J. Farmer, Broadband cable access networks: The HFC plant, 1st ed.

Burlington, MA: Elsevier and Morgan Kaufmann, 2009, vol. 156.

[83] F. Di Paolo and F. Di Paolo, Networks and devices using planar transmission lines.

Boca Raton, Florida, USA: CRC Press, 2018.

[84] B. Böttcher et al., “Algorithms for partial discharge monitoring of medium

voltage cable plugs, using a multi-sensor expert system,” VDE High Voltage

design and measurement of ar-msa 149

Technology 2020; ETG-Symposium, pp. 1–6,2020. [Online]. Available: https : / /

ieeexplore.ieee.org/document/9275424.

[85] E. Nash, Ask the applications engineer—28: Logarithmic amplifiers explained, Analog

Devices Inc., Online document, Norwood, Massachusetts, USA, 1999. [On-line].

Available: https : / / www . analog . com / en / analog - dialogue / articles /

logarithmic-amplifiers-explained.html.

[86] D. Gautschi and P. Bertholet, “Calibration of UHF sensors for GIS:

Comparison of different methods and testing of a calibration system based on

a conical antenna,” 2010 International Conference on High Voltage Engineering and

Application, New Orleans, LA, USA, pp. 28–31, 2010. doi: 10.1109/

ICHVE.2010.5640874.

[87] A. Mathur, R. Trivedi, and G. B. Mathur, Antenna and wave propagation, 1st ed.

Scientific Publishers, 2011.

[88] M. Siegel et al., “Calibration proposal for UHF partial discharge

measurements at power transformers,” Energies, vol. 12, no. 16, p. 3058, 2019.

doi: 10 . 3390 /en12163058.

[89] E. L. Bronaugh and J. D. Osburn, “Measuring antenna parameters in a GHz

transverse electromagnetic (GTEM) cell,” IEEE Antennas and Propagation Society

International Symposium 1992 Digest, Chicago, IL, USA, no. 4, pp. 2064–2066,

1992. doi: 10.1109/APS.1992.221461.

[90] AMETEK CTS GmbH,

GTEM

cells - emissions and immunity testing in a

single, shielded environment, AMETEK CTS GmbH, Ed., Online document,

Reinach, Switzerland, 2017. [Online]. Available: https : / / device . report /

manual / 3794933.

[91] B. Kordi, J. Lo Vetri, G. Bridges, and I. Jeffrey, “Calibration methods for elec-

tric field probes and GTEM cells,” 2004 International Symposium on

Electromagnetic Compatibility (IEEE Cat. No.04CH37559), Silicon Valley, CA, USA,

no. 3, pp. 908–912, 2004. doi: 10.1109/ISEMC.2004.1349946.

[92] M. Judd, M. Siegel, and S. Coenen, “UHF PD sensor characterisation using

GTEM cells,” VDE-Hochspannungstechnik 2018 - ETG-Fachtagung, Berlin, Germany,

pp. 1–6, 2018. [Online]. Available: https : / / www . vde - verlag . de / proceedings

- en / 454807118.html.

[93] F. Pythoud et al., Antenna factor for loop antennas, EURAMET e.V., Online

document, EURAMET e.V., Braunschweig, Germany, 2014. [Online]. Available:

https : / / www . euramet . org / Media / docs / projects / EURAMET - P1171 _ ELECT _

Final_Report.pdf.

150 design and measurement of ar-msa

[94] M. J. Alexander, M. J. Salter, D. G. Gentle, D. A. Knight, B. G. Loader, and

K. P. Holland, Calibration and use of antennas, focusing on

EMC

applications,

National Physics Laboratory, Online document, Teddington, UK, 2004. [Online].

Available: https://eprintspublications.npl.co.uk/3520/1/mgpg73.pdf.

[95] V. Trainotti and G. Figueroa, “Vertically polarized dipoles and monopoles, di-

rectivity, effective height and antenna factor,” IEEE Transactions on Broadcasting,

vol. 56, no. 3, pp. 379–409, 2010. doi: 10.1109/TBC.2010.2050627.