Methods For Arc Flash Prediction in MV LV Switchgear

Department of Electrical Engineering and Automation
Arc-Flash is an unintentional discharge of

Methods for Arc-Flash
Aalto-DD 221/2015
Ghulam Amjad Hussain

electricity through air which produces very
high temperature (19,600°C) and a force
equivalent to being hit by a hand grenade.
Occurrence of such an electrical fault in Prediction in Medium
switchgear causes interruption of electricity
to the end users and damage to the other
substation equipment and to the personnel
Voltage and Low Voltage
working in its vicinity. This thesis provides
methods for pre-emptive detection of Switchgear
Methods for Arc-Flash Prediction in Medium Voltage and Low Voltage Switchgear
potential arc-flash in MV/ LV switchgear
using online condition monitoring
techniques. Three non-intrusive sensors i.e.
D-dot sensor, Rogowski coil and Thermal
Ionization Detector have been designed, Ghulam Amjad Hussain
implemented and tested in switchgear to
detect major causes of arc-flash.
Assessment of reliability, sensitivity and
operability of the mentioned sensors is done
based on the analysis of acquired data from
laboratory tests as well as on site in practical
switchgears. The integration of the proposed
solution in SCADA system will be a
breakthrough solution in minimizing arc-
flash accidents and saving precision lives.
ISBN 978-952-60-6600-4 (printed) BUSINESS +

9HSTFMG*aggaae+
ISBN 978-952-60-6599-1 (pdf) ECONOMY

ISSN-L 1799-4934
ISSN 1799-4934 (printed) ART +
ISSN 1799-4942 (pdf) DESIGN +
ARCHITECTURE
Aalto University
Aalto University
School of Electrical Engineering SCIENCE +
Department of Electrical Engineering and Automation TECHNOLOGY
www.aalto.fi
CROSSOVER
DOCTORAL
DOCTORAL DISSERTATIONS
DISSERTATIONS
2016
Aalto University publication series
DOCTORAL DISSERTATIONS 221/2015
Methods for Arc-Flash Prediction in

Medium Voltage and Low Voltage
Switchgear
A doctoral dissertation completed for the degree of Doctor of

Science (Technology) to be defended, with the permission of the
Aalto University School of Electrical Engineering, at a public
examination held at the lecture hall S1 of the school on 5th of
February 2016 at 12:00.
Aalto University
School of Electrical Engineering
Electrical Engineering & Automation
Power Systems and High Voltage Engineering
Supervising professor
Prof. Matti Lehtonen
Thesis advisor
Prof. Matti Lehtonen
Preliminary examiners
Professor Charles Kim, Howard University, USA.
Professor Hans Edin, KTH Royal Institute, Sweden.
Opponent
Professor Petr Toman, Brno University of Technology, Brno, Czech Republic
Aalto University publication series

DOCTORAL DISSERTATIONS 221/2015
© Ghulam Amjad Hussain
ISBN 978-952-60-6600-4 (printed)

ISBN 978-952-60-6599-1 (pdf)
ISSN-L 1799-4934
ISSN 1799-4934 (printed)
ISSN 1799-4942 (pdf)
http://urn.fi/URN:ISBN:978-952-60-6599-1
Unigrafia Oy
Helsinki 2015
Finland
Publication orders (printed book):

Espoo, Finland
Abstract
Aalto University, P.O. Box 11000, FI-00076 Aalto www.aalto.fi
Author
Name of the doctoral dissertation
Publisher School of Electrical Engineering
Unit Electrical Engineering and Automation
Series Aalto University publication series DOCTORAL DISSERTATIONS 221/2015
Field of research Power Systems and High Voltage Engineering
Manuscript submitted 7 September 2015 Date of the defence 5 February 2016
Permission to publish granted (date) 19 November 2015 Language English
Monograph Article dissertation (summary + original articles)
Abstract
Nowadays we are highly dependent on the electricity. It has become a need of daily life
including domestic, commercial, transformational, industrial, health care and
telecommunication purposes. Switchgear forms an integral part of the distribution network in
power system. Hence, the occurrence of an electrical fault in switchgear causes interruption of
electricity to the end users. Being highly dependent on electricity, such unexpected
interruptions are unbearable by the modern society. Arc-flash is the unintentional discharge
of electricity through air which produces very high temperature (19,600 ºC), which is hotter
than the surface of the sun and a force equivalent to being hit by a hand grenade. Occurrence of
an arc-flash in switchgear not only causes interruption of electric supply but damage to the
equipment and personnel working in the vicinity. In the worst cases, damage is extended to the
whole substation, hence maintenance duration can last upto few weeks. In USA alone, five to
ten explosions occur due to arc-flashes every day.
Due to the importance of the matter, arc-flash protection has been highly demanded for last
two decades. Several reactive protection techniques have been introduced in the literature and
some of them were implemented commercially. Continuous monitoring of the equipment and
earlier detection of potential failures can facilitate a more proactive and a comprehensive arc-
flash prevention system. This thesis mainly focuses on the pre-emptive detection of potential
arc-flash in medium voltage (MV) and low voltage (LV) switchgear using online condition
monitoring techniques. Three non-intrusive sensors i.e. D-dot sensor, Rogowski coil and
thermal ionization detector have been designed, implemented and tested in switchgear to
detect partial discharge, low power arcing and hotspots due to bad connections. Assessment of
reliability, sensitivity and operability of the mentioned sensors is done in the laboratory. D-dot
sensor is also tested on-site in operational MV switchgear for partial discharge. An effective de-
noising technique based on discrete wavelet transform (DWT) is presented in this thesis.
Moreover a novel idea for the detection of multiple faults in a panel based on correlation
between the cumulative energy and apparent charge of discharge events is also presented.
Hybrid low cost and non-intrusive solution integrated to the supervisory control and data
acquisition (SCADA) system capable of continuous monitoring of switchgear and indicating
any potential arc faults in switchgear before they lead to severe damage to the equipment and
vicinity will be a breakthrough solution in minimizing arc-flash accidents, emergency
interruptions and saving precious lives.
Keywords Medium voltage, low voltage, switchgear, arc fault, partial discharge, condition
monitoring, non-intrusive sensor
ISBN (printed) 978-952-60-6600-4 ISBN (pdf) 978-952-60-6599-1
ISSN-L 1799-4934 ISSN (printed) 1799-4934 ISSN (pdf) 1799-4942
Location of publisher Helsinki Location of printing Helsinki Year 2015
Pages 236 urn http://urn.fi/URN:ISBN:978-952-60-6599-1
Tiivistelmä
Aalto-yliopisto, PL 11000, 00076 Aalto www.aalto.fi
Tekijä
Väitöskirjan nimi
Valokaarivikojen ennakointimenetelmät keski- ja pienjännitekojeistoissa
Julkaisija Sähkötekniikan korkeakoulu
Yksikkö Sähkötekniikan ja automaation laitos
Sarja Aalto University publication series DOCTORAL DISSERTATIONS 221/2015
Tutkimusala Sähköverkot ja suurjännitetekniikka
Käsikirjoituksen pvm 07.09.2015 Väitöspäivä 05.02.2016
Julkaisuluvan myöntämispäivä 19.11.2015 Kieli Englanti
Monografia Yhdistelmäväitöskirja (yhteenveto-osa + erillisartikkelit)
Tiivistelmä
Nykyään olemme riippuvaisia sähköstä. Sitä tarvitaan jokapäiväisessä elämässä
kotitalouksissa, kaupallisessa toiminnassa, liikenteessä, teollisessa toiminnassa,
terveydenhoidossa ja tietoliikenteessä. Kojeistot muodostavat olennaisen osan
sähkönjakeluverkoissa. Sähkötekninen vika kojeistossa aiheuttaa keskeytyksen.
Odottamattomia keskeytyksiä ei juuri suvaita. Vikavalokaari on tahattomasti syntynyt
sähköpurkaus ilmassa, ja se synnyttää hyvin korkean lämpötilan (19600 ˚C), joka on kuumempi
kuin auringon pinta. Valokaari synnyttää myös voimavaikutuksen joka tuhoaa laitteita ja
aiheuttaa vaaran läheisyydessä työskenteleville ihmisille. Pahimmissa tapauksissa
tuhovaikutus ulottuu koko sähköasemaan, jolloin vaikutusten korjaaminen voi kestää useita
viikkoja. Yksin Yhdysvalloissa tapahtuu 5-10 valokaarivikaa joka päivä.
Koska asialla on suuri merkitys, valokaarisuojaukselle on syntynyt suuri kysyntä parin viime
vuosikymmenen aikana. Useita reaktiivisia suojausmenetelmiä on esitelty kirjallisuudessa, ja
joitakin niistä on otettu kaupalliseen käyttöön. Jatkuvatoiminen laitteistojen seuranta ja
vikaantumisten varhainen toteaminen voivat johtaa valokaarivikoja estävään järjestelmään.
Tämä järjestelmä tähtää pääasiassa valokaarivikojen ennaltaehkäisevään havaitsemiseen
keski- ja pienjännitekojeistoissa jatkuvatoimisten monitorointitekniikoiden avulla. Kolme
sensorityyppiä, jotka eivät edellytä suojattaviin laitteisiin tunkeutumista, D-Dot -sensori,
Rogowski-kela ja lämpöionisaatiosensori on suunniteltu, toteutettu ja testattu kojeistossa, jotta
osittaispurkaukset, pienitehoiset valokaaret ja huonoista liitoksista aiheutuvat kuumat pisteet
voidaan havaita. Em. sensorien luotettavuus, herkkyys ja toiminnallisuus on arvioitu
laboratoriossa. D-dot -sensori on testattu myös käytössä olevassa keskijännitekojeistossa.
Työssä on esitetty myös tehokas diskreettiin wavelet-muunnokseen perustuva
mittaussignaalin kohinanpoistotekniikka.
Edullinen, laitteisiin tunkeutumaton hybridiratkaisu, joka voidaan liittää
käytönvalvontajärjestelmään (SCADA) ja joka pystyy jatkuvatoimisesti monitoroimaan
kojeistoa ja indikoimaan kaikki mahdolliset valokaariviat, ennen kuin ne kehittyvät ja johtavat
vakavaan laitteiden ja lähiympäristön vaurioitumiseen, tulee olemaan läpimurto
valokaarionnettomuuksien ja niistä aiheutuvien vikakeskeytysten minimoinnissa ja
ihmishenkien pelastamisessa.
Avainsanat Keskijännite, pienjännite, kytkinkojeisto, valokaarivika, osittaispurkaus,

kunnonvalvonta, ainetta rikkomaton sensori.
ISBN (painettu) 978-952-60-6600-4 ISBN (pdf) 978-952-60-6599-1
ISSN-L 1799-4934 ISSN (painettu) 1799-4934 ISSN (pdf) 1799-4942
Julkaisupaikka Helsinki Painopaikka Helsinki Vuosi 2015
Sivumäärä 236 urn http://urn.fi/URN:ISBN:978-952-60-6599-1
Acknowledgement
All praise to The Almighty Allah, merciful and compassionate for providing me this
opportunity and granting me the capability to complete this doctoral dissertation.
Gratitude needs to be expressed to Prof. Matti Lehtonen for initially involving me in this project
by believing in my abilities and skills. He has always welcomed me for the valuable discussion
and new ideas in developing this thesis and ensuring all the necessary tools required for this
research. I truly thank him a lot for continuous encouragement, endless cooperation and strong
support. I am very grateful to Mr. Lauri Kumpulainen for proposing the need of this research
and arranging initial grants. He has always been open for the discussion and contributing to the
research ideas. I am thankful to Prof. Steffen Grossmann for arranging laboratory test up at TU
Dresden. Special thanks to Mr. John Kay for his valuable contribution in our PCIC-IAS articles.
Aalto University, The Finnish Foundation for Technology Promotion (TES), Finnish Cultural
Foundation, Association for Electrical Safety Promotion (Sähköturvallisuuden edistämiskeskus-
STEK), the Research Foundation of Aalto University and Vamp Oy are greatly acknowledged
for their financial support without which this research would not be possible.
I owe my gratitude to the current and former members of the Power Systems and HV laboratory
of Aalto University. First of all I want to thank Dr. M. Shafiq, Dr. Joni Klüss and Dr. Murtaza
Hashmi for providing initial guidance, technical support and countless ideas to start this
research work. Special thanks to Dr. Petri Hyvönen, Dr. Lauri Kütt and Mr. Tatu Nieminen who
have always been ready to understand, analyze, discuss, and suggest, for the realization and
execution of my proposed ideas in the HV Laboratory. At the same time, I want to thank Mr.
Veli-Matti, Mr. Jouni Mäkinen and Ari Haavisto for their laboratory instrumentation and IT
support.
I want to express my deep sense of gratefulness to all other members of our working group and
my friends who were helpful in various ways. My special thanks to Dr. Nagy Elkalashy, Dr.
John Millar, Mr. M. Humayun, Malik Farhan, Farhan Mahmood, Mehrnaz Farzamfar, Merkebu
Degefa, Mubbashir Ali, Bruno Sousa, Hammad Kabir, Ejaz Ahmed, Muhammad Sharif, Dr.
Usman Sheikh and Zoko Frank who were part of my social network or active discussions
throughout my doctoral research period.
Most importantly, my mother receives my veracious gratitude, who has dedicated her whole life
for me and my siblings and provided unconditional love and care. She has made me what I am
today. I would like to thank my sisters (Mrs. Sughra, Mrs. Kalsoom and Mrs. Sajida), brother-
in-laws and other family members who keep praying for my success. I am very grateful to my
uncle Mr. Khuda Bux Sabri, my teacher Mr. Abdul Rasheed Qureshi and my childhood friends
who have a great influence in building my personality which I possess today.

Espoo, Sep 2015
vii
viii
Dedicated to my Mother
For her unconditional and endless love, sincere prayers and tremendous care.
ix
x
List of Publications
This thesis consists of an overview of the following publications which are referred to in the
text by their roman numerical.
I. L. Kumpulainen, G. A. Hussain, M. Lehtonen, J. A. Kay, “Preemptive Arc Fault

Detection Techniques in Switchgear and Controlgear”, IEEE Transactions on Industry
Applications, vol.49, no.4, pp.1911-1919, July/Aug 2013.
II. L. Kumpulainen, G. A. Hussain, K. Kauhaniemi, M. Lehtonen, “Aspects of Arc-Flash

Protection and Prediction”, Electric Power Systems Research (Elsevier), vol. 116, no.
11, pp. 77-86, Nov 2014.
III. G. A. Hussain, L. Kumpulainen, M. Lehtonen, J. A. Kay, “Preemptive Arc Fault

Detection Techniques in Switchgear and Controlgear- Part II,” IEEE Transactions on
Industry Applications, vol. 50, no. 3, pp. 1649 – 1658, May/Jun 2014.
IV. G. A. Hussain, M. Shafiq, M. Lehtonen and M. Hashmi, “Online Condition Monitoring

of MV Switchgear by D-dot Sensor to Predict Arc-faults,” IEEE Sensors Journal, vol.
15, no. 12, pp. 7262-7272, Dec 2015.
V. M. Shafiq, G. A. Hussain, L. Kütt, M. Lehtonen, “Effect of Geometrical Parameters on

the High Frequency Performance of Rogowski Coil for Partial Discharge
Measurements”, Measurement (Elsevier), vol. 49, no. 1, pp. 126-137, Mar 2014.
VI. G. A. Hussain, M. Shafiq, M. Lehtonen, “Thermal Monitoring of LV Switchgear Using

Thermal Ionization Detector (TID)”, International Review of Electrical Engineering
(under review).
VII. G. A. Hussain, M. Shafiq, L. Kumpulainen, F. Mahmood, M. Lehtonen, “Performance

Evaluation of Noise Reduction during On-Line Monitoring of MV Switchgear for PD
Measurements by Non-intrusive Sensors”, International Journal of Electrical Power
and Energy Systems (Elsevier), Vol. 64, no. 1, pp. 596-607, January 2015.
VIII. G. A. Hussain, L. Kumpulainen, M. Lehtonen, M. Hashmi, M. Shafiq, “Signal

Processing of PD Measurements to Predict Arcing Faults in MV Switchgears”,
Proceedings of IEEE International Conference on Industrial Technology (ICIT), Cape
town- South Africa, pp. 916-921, 25-28 February 2013.
IX. G. A. Hussain, M. Shafiq, J. A. Kay, M. Lehtonen, “Pre-Emptive Arc Fault Detection

Techniques in Switchgear- Part III, From the Laboratory to Practical Installation” IEEE
Transactions on Industry Applications, vol. 51, no. 3, pp. 2615-2623, May/ Jun 2015.
X. G. A. Hussain, L. Kumpulainen, J. V. Klüss, J. A. Kay and M. Lehtonen, “The Smart

Solution for the Prediction of Slowly Developing Electrical Faults in MV Switchgear
Using Partial Discharge Measurement,” IEEE Transactions on Power Delivery, vol. 28,
no. 4, pp. 2309-2316, Oct 2013.
xi
xii
Author’s Contribution
Publication I:
Preemptive Arc Fault Detection Techniques in Switchgear and Controlgear
Lauri Kumpulainen and Amjad Hussain developed the main idea of the paper. The
paper writing was done by Lauri Kumpulainen and Amjad Hussain. John A. Kay and
Matti Lehtonen contributed through the comments, discussion and paper review.
Publication II:
Aspects of Arc-Flash Protection and Prediction
Lauri Kumpulainen and Amjad Hussain developed the main idea of the paper. The
paper writing was done by Lauri Kumpulainen and Amjad Hussain. Kimmo
Kauhaniemi contributed through the comments, discussion and paper review. Matti
Lehtonen supervised the work and contributed through comments and discussion.
Publication III:
Preemptive Arc Fault Detection Techniques in Switchgear and Controlgear- Part
II
Amjad Hussain developed the main idea of the paper. The experimental work, analysis
of results and paper writing was accomplished by Amjad Hussain. Lauri Kumpulainen
has contributed through discussion and comments. Matti Lehtonen has supervised the
work and contributed to improve the idea and contents. John A. Kay has been
contributing through the paper review.
Publication IV:
Online Condition Monitoring of MV Switchgear by D-dot Sensor to Predict Arc-
faults
of results and paper writing was accomplished by Amjad Hussain. Muhammad Shafiq
has contributed through discussion, comments and improvement of idea. Matti
Lehtonen has supervised the work and contributed to improve the idea and contents.
Murtaza Hashmi has been contributing through the analysis of results and paper review.
xiii
Publication V:
Effect of Geometrical Parameters on the High Frequency Performance of
Rogowski Coil for Partial Discharge Measurements
Muhammad Shafiq developed the main idea and was responsible for experimental
work, analysis of results, and writing of the paper. Amjad Hussain has contributed in the
main idea and writing of the paper. Lauri Kütt has contributed in the idea of the paper
and improved the organization of the contents. Matti Lehtonen has supervised the work
and contributed through comments and discussions.
Publication VI:
Thermal Monitoring of LV Switchgear Using Thermal Ionization Detector (TID)
of results and paper writing was done by Amjad Hussain. Muhammad Shafiq
contributed through the comments, discussion and improvement in the main idea. Matti
Lehtonen supervised the work and contributed through comments, discussion and
improvement of the contents.
Publication VII:
Performance Evaluation of Noise Reduction during On-Line Monitoring of MV
Switchgear for PD Measurements by Non-intrusive Sensors
The main idea of the paper was developed by Amjad Hussain. The experimental work,
analysis of results and paper writing was done by Amjad Hussain. Muhammad Shafiq
contributed through the comments, discussion and improvement in the main idea. Lauri
Kumpulainen and Farhan Mahmood contributed in discussion and paper review. Matti
Lehtonen supervised the work and contributed through comments, discussion and
improvement of the contents.
Publication VIII:
Signal Processing of PD Measurements to Predict Arcing Faults in MV
Switchgears
analysis of results and paper writing was done by Amjad Hussain. Lauri Kumpulainen
and Matti Lehtonen contributed through the comments, discussion and improvement in
the main idea. Murtaza Hashmi helped with the analysis of results and paper review.
Muhammad Shafiq contributed in discussion and paper review.
xiv
Publication IX:
Pre-Emptive Arc Fault Detection Techniques in Switchgear- Part III, From the
Laboratory to Practical Installation
analysis of results and paper writing was done by Amjad Hussain. Muhammad Shafiq
and John A. Kay contributed through the comments, discussion and improvement in the
main idea of the paper and its contents. Matti Lehtonen supervised the work and
contributed through comments, discussion and review of the manuscript.
Publication X:
The Smart Solution for the Prediction of Slowly Developing Electrical Faults in
MV Switchgear Using Partial Discharge Measurement
analysis of results and paper writing was done by Amjad Hussain. Lauri Kumpulainen
contributed through the discussion and improvement in the idea. Joni Klüss contributed
in the experimental work and writing alongwith discussion and comments. John A. Kay
and Matti Lehtonen contributed through comments, discussion and review of the
manuscript.
xv
xvi
Table of Contents
Abstract....................................................................................................................................... iii
Tiivistelmä ................................................................................................................................... v
Acknowledgement ..................................................................................................................... vii
List of Publications .................................................................................................................... ix
Author’s Contribution ............................................................................................................. xiii
List of Abbreviations ............................................................................................................... xx
List of Symbols ....................................................................................................................... xxii
CHAPTER 1 INTRODUCTION .............................................................................................. 1
1.1 Background ............................................................................................................................ 1
1.2 Aim of the Research............................................................................................................... 1
1.3 Research Methods .................................................................................................................. 2
1.4 Scientific Contribution ........................................................................................................... 2
1.5 Dissertation Outline ............................................................................................................... 3
CHAPTER 2 ARC FLASH IN SWITCHGEAR ..................................................................... 5
2.1 Arc-Flash................................................................................................................................. 5
2.1.1 Series Arc Faults............................................................................................................................. 5
2.1.2 Parallel Arc Fault ............................................................................................................................ 6
2.1.3 Energy of Arc Fault ........................................................................................................................ 6
2.1.4 Consequences of Arc Fault ............................................................................................................. 7
2.2 Sources of Arc Fault ............................................................................................................... 7
2.2.1 Mechanisms Gradually Leading to Arc Ignition ............................................................................ 7
a. Poor Connections ............................................................................................................................ 8
b. Degradation of Insulation ............................................................................................................. 10
i. Partial Discharge (PD) ........................................................................................................... 11
ii. Overvoltage ............................................................................................................................. 17
c. Poor Maintenance ......................................................................................................................... 18
2.2.2 Immediate Cause of Arc Ignition ................................................................................................. 18
a. Accidental Contacts ...................................................................................................................... 18
b. Foreign Objects............................................................................................................................. 18
CHAPTER 3 ARC-FLASH PROTECTION TECHNIQUES IN SWITCHGEAR ........... 19

3.1 A Holistic View of Existing Arc-flash Protection Technologies .......................................... 19
3.2 Arc-flash Prevention ............................................................................................................. 19
3.2.1 Education and Training ................................................................................................................ 20
3.2.2 Preventive Design ......................................................................................................................... 20
3.2.3 High Resistance Grounding .......................................................................................................... 20
3.2.4 Insulated vs. Bare Busbar ............................................................................................................. 20
3.2.5 Maintenance ................................................................................................................................. 21
3.3 Arc Flash Reactive Protection............................................................................................... 21
3.3.1 Current-Limiting Approach in Arc-Flash Protection .................................................................... 22
3.3.2 Limitation of Incident Energy by Reducing Arcing Time ............................................................ 23
3.3.3 Light and Overcurrent Based Protection ...................................................................................... 23
3.3.4 Arc Eliminators ............................................................................................................................ 24
3.4 Pre-emptive Arc Fault Protection ......................................................................................... 25
3.4.1 Online Condition Monitoring (Condition based maintenance) ..................................................... 25
3.4.2 Physical Emissions Indicating Developing Faults ........................................................................ 26
a. Electromagnetic (EM) Emissions ................................................................................................. 26
b. Acoustic (Ultrasonic) Emissions .................................................................................................. 26
c. Optical (Ultraviolet-UV) Emissions ............................................................................................. 27
xvii
d. HF Current Components ...............................................................................................................27
e. Thermal Emissions (IR emission) .................................................................................................27
f. Chemical Emissions .......................................................................................................................27
3.4.3 Measuring Sensors ........................................................................................................................28
a. Electromagnetic Sensor- UHF/ VHF Antenna ..............................................................................28
b. Coupling Capacitor (CC) ..............................................................................................................28
c. Directional Couplers......................................................................................................................29
d. High Frequency Current Transformer (HFCT) .............................................................................29
e. Rogowski Coil (RC) ......................................................................................................................29
f. Ultrasonic Sensor- Piezoelectric Probe ..........................................................................................29
g. Ultraviolet (UV) Sensors ...............................................................................................................30
h. Thermal Sensors ............................................................................................................................30
i. D-dot Sensor ..................................................................................................................................30
3.4.4 Comparison between Different Sensors ........................................................................................31
CHAPTER 4 DESIGN AND IMPLEMENTATION OF SENSORS ................................... 33

4.1 Online Condition Assessment System for Electrical Equipment .......................................... 33
4.2 Selection of Sensors for the Prediction of Arc- Flash ........................................................... 34
4.3 Rogowski Coil ....................................................................................................................... 34
4.3.1 Operating Principle and Construction of RC ................................................................................35
4.3.2 Electrical Model of RC .................................................................................................................36
4.3.4 Measurement of Incipient Faults using RC ...................................................................................37
4.4 D-dot Sensor ......................................................................................................................... 38
4.4.1 Operating Principle and Construction of D-dot Sensor ................................................................38
4.4.2 Electrical Model of D-dot Sensor ................................................................................................39
4.4.3 Measurement of Incipient Faults Using D-dot Sensor .................................................................41
4.5 Thermal Ionization Detector (TID) ...................................................................................... 41
4.5.1 Operating Principle and Construction of TID ..............................................................................42
4.5.2 Diagnosis of Bad Connections Using TID ...................................................................................43
CHAPTER 5 ON-LINE CONDITION ASSESSMENT OF SWITCHGEAR .................... 45

5.1 De-noising ............................................................................................................................ 45
5.1.1 External Noise During Online Measurements..............................................................................45
5.1.2 Discrete Wavelet Transform (DWT) Based De-noising ..............................................................46
a. Discrete Wavelet Transform.........................................................................................................46
b. De-noising Procedure ...................................................................................................................47
c. De-noising Results .......................................................................................................................48
d. Self-adaptive De-noising ..............................................................................................................53
5.2 Laboratory Measured Signals by the Sensors....................................................................... 55
5.2.1 Calibration and Performance of the Sensor ..................................................................................58
5.2.2 Time Domain Comparison ...........................................................................................................59
5.2.3 Frequency Domain Comparison ..................................................................................................60
5.3 On-site/ Practical Measured Signals ..................................................................................... 61
5.3.1 On-site Measured Noise ...............................................................................................................63
5.3.2 On-site PD Signals .......................................................................................................................64
5.3.3 Comparison Between On-site noise and On-site PDs ..................................................................65
5.4 Detection of Multiple Faults in a Switchgear ....................................................................... 66
5.4.1 Calculation of Apparent Charge and Cumulative Energy ............................................................66
5.4.2 Correlation Between Apparent Charge and Cumulative Energy ..................................................68
5.5 Thermal Monitoring of Switchgear using TID Sensor ......................................................... 70
5.5.1 Compartment at Room Temperature (Case 1) .............................................................................71
5.5.2 Compartment at an Elevated Temperature of 47 °C (Case 2) ......................................................73
5.5.3 Time Delay in TID Signal ............................................................................................................74
5.5.4 Challenges with Thermal Monitoring using TID .........................................................................74
5.5.5 Fault Location using IR Sensors ..................................................................................................75
5.6 Exact Location of Fault in a Switchgear .............................................................................. 75
xviii
5.7 Integration of On-line Condition Monitoring Sensors with Central Control System or
SCADA ....................................................................................................................................... 76
CHAPTER 6 CONCLUSIONS AND FUTURE WORK...................................................... 79
6.1 Conclusions .......................................................................................................................... 79
6.2 Future Work ......................................................................................................................... 80
REFERENCES .......................................................................................................................... 81
APPENDIX-I ............................................................................................................................. 89
Technical Details of the Thermal Ionization Detector (TID) ...................................................... 89
Errata ......................................................................................................................................... 91
PUBLICATIONS I-X ............................................................................................................... 93
xix
List of Abbreviations
Abbreviation Description
CB Circuit Breaker
CC Coupling Capacitor
CD Coupling Device
CL Current Limiting
CT Current Transformer
CWT Continuous Wavelet Transform
DCS Distributed Control Systems
DSI Discrete Spectral Interference
DSP Digital Signal Processing
DWT Discrete Wavelet Transform
EM Electromagnetic
EMI Electromagnetic Interference
FFT Fast Fourier Transform
FIR Finite Impulse Response
FPGA Field Programmable Gate Array
GIS Gas Insulated Switchgear
HF High Frequency
HFCT High Frequency Current Transformer
HV High Voltage
IEC International Electro-technical Commission
IEEE Institute of Electrical & Electronics Engineers
IIR Infinite Impulse Response
IR Infrared
LV Low Voltage
MSD Multi-resolution Signal Decomposition
MV Medium Voltage
PD Partial Discharge
PLC Power Line Carrier
PPE Personal Protective Equipment
PT Potential Transformer
PTFE Polytetrafluoroethylene (Teflon)
RF Radio Frequency
RFCT Radio Frequency Current Transformer
xx
RC Rogowski Coil
SAIC Science Applications International Corporation
SCADA Supervisory Control and Data Acquisition
SMA Sub-miniature Version A
TID Thermal Ionization Detector
UV Ultraviolet
UHF Ultra High Frequency
VHF Very High Frequency
WAN Wide Area Network
WT Wavelet Transform
xxi
List of Symbols
Symbol Description
A Peak amplitude of a PD pulse
Ca Capacitance of healthy part of an insulation in
parallel with defective part
Cb Capacitance of healthy part of an insulation in
series with defective part
Cc Capacitance of defective insulation
Cc Self capacitance of measuring cable
Cm Capacitance of measuring system
Crc Self capacitance of Rogowski coil
Cs Self capacitance of D-dot sensor
De or D Electric flux density
di Inner diameter of Rogowski coil
dm Mean diameter of Rogowski coil
do Outer diameter of Rogowski coil
drc Core diameter of Rogowski coil
dw Diameter of copper wire used for Rogowski coil’s
winding
Ea Electric field across an insulation
Ec Electric field inside the cavity of a defective
insulation
f Resonant frequency of sensor
fs Sampling frequency of data acquisition system
I”k Symmetric short circuit current
ip(t) Primary current to be measured by Rogowski coil
or HFCT
Lc Self inductance of measuring cable
Lrc Self inductance of Rogowski coil
Ls Self inductance of D-dot Sensor
ls Length of the sensor
Mc Mutual inductance of Rogowski Coil
N number of turns of Rogowski coil
Pi Power of a sample
qo Actual charge distributed within the PD cavity
xxii
ri Radius of sensing electrode of D-dot Sensor
ro Distance between the centre of sensing electrode
and the metallic case
Rc Self resistance of measuring cable
Rm Resistance of measuring system
Rs Self resistance of D-dot sensor
Rrc Self resistance of Rogowski coil
Ua Total voltage across the insulation
Uc Voltage stress across the defective part of
insulation
Ue Extinction voltage of a PD cavity
Ui Inception voltage of an insulation cavity
Un Nominal system voltage (RMS value)
Up Peak system voltage
Vp Peak value of a signal
Vpp Peak to peak value of a signal
Vi Voltage of a sample
vrc(t) Voltage induced in the Rogowski coil
Wi Energy of a sample
Zk Short circuit impedance of an electrical network
xxiii
Symbol Description
α1 Rate of rise-time of PD pulse
α2 Rate of fall-time of PD pulse
εd Permittivity of the insulating material
εk Relative permittivity of an insulating material
εo Permittivity of air
λ Wavelength of a wave
μo Permeability of air
μr Relative permeability of dielectric material of
coaxial cable
τ Time shift operator in a mathematical expression
xxiv
Chapter 1
INTRODUCTION
1.1 Background
Switchgear forms an integral part of the distribution network in an electric power

system. Hence, the occurrence of fault in switchgear results in the interruption of power supply
to the electricity users. Being highly dependent on electric supply, such emergency interruptions
are not affordable by the modern society. In addition, an arc-flash in the switchgear is the major
cause of serious injuries to the personnel working in the vicinity and damage to the switchgear
itself. It produces very high temperature (19,600 ºC), which is hotter than the surface of the sun
and a force equivalent to being hit by a hand grenade. It can cause memory and hearing loss to
human, in addition to severe burns [1]. Secondly, the damage to the equipment itself incurs
maintenance and interruption cost. Medium voltage (MV) switchgears are very costly and
sometimes takes long time to repair or replace them due to unavailability in local market, thus,
resulting longer downtime.
In USA alone, five to ten explosions occur due to arc-flashes every day [2]. Historical
data shows that major electrical fires occur on navy ships at a rate of 6 times per year [3]. The
near misses and unreported events are even more than this in number. In short, Arc-Flash
incident in the switchgear applications is one of the major causes of human loss or injury as well
as economic loss to the electrical utility company. Hence the solution to this problem is of very
high importance. Due to the importance of the matter, arc-flash protection has been highly
demanded for the last two decades. Several arc resistant switchgears and personal protective
equipment (PPE) have been employed. Arc protection relays are also available commercially
[4]. But all these applications focus on the reactive techniques (post arc ignition) i.e. minimising
the extent of damage after the occurrence of arc-flash. Hence, a pre-emptive and predictive
technology is demanded, to indicate the potential arc-flash events in near future in order to
avoid catastrophic damages due to arc-flash.
1.2 Aim of the Research
A proactive approach has been adopted in this research. The focus area is
comprehensive monitoring of MV and low voltage (LV) switchgear and detection of incipient
faults (causes of arc-flash) in real time, which may lead to an arc-flash with the progression of
time. The objectives mainly include evaluation of non-intrusive sensors for the early detection
of arc-flash in LV and MV switchgear and controlgear. Evaluation includes their cost,
sensitivity, and complexity analysis as well as studying the ease of application and availability
1
of not only sensors but the whole measuring systems. Objectives also include the
implementation of the sensors in practical switchgear, not only in laboratory environment but
also on-site, in order to assess the reliability of the sensors. Essential signal processing
techniques to de-noise the captured signals before further analysis for quantification is also one
part of this research. Implementation of cost effective and unique sensors rather than the
available commercial costly sensors is the idea. On the basis of reliability, sensitivity and
operability assessment of the sensors, a comprehensive solution for the continuous monitoring
of LV and MV switchgear is the final goal.
1.3 Research Methods
Conventional method of fault diagnostics relies on offline testing, time based

maintenance and hence, adopts a passive protection. This method has many drawbacks. For
example users may stay blind between the maintenance intervals. Interruption and maintenance
cost is incurred due to unnecessary maintenance, even when the equipment is quite healthy.
Unconventional methods are based on online condition monitoring techniques by using non-
intrusive sensors. The health of equipment can be monitored in real time. Unnecessary
shutdowns are actively avoided through these methods. Sensors used in this technology are very
cheap, small in size and light weight. Unlike the conventional methods, they don’t involve
heavy testing equipment to be transported to site for tests.
Due to the availability of advanced micro processing and non-intrusive sensors’

technology, it is now possible to monitor switchgear without disconnecting the power supply.
This is a major motivation to proceed with the real time monitoring of switchgear and prediction
of arc-flash. In short, by using unconventional methods, the prediction of arc-flash technology
in the MV and LV switchgear panels is implemented in this research work.
1.4 Scientific Contribution
Following are the major scientific contributions of this dissertation.
x The detection of incipient faults (PD, arcing, hotspots) by using new non-intrusive
sensors in the laboratory as well as on-site.
x Design, implementation and testing of non-intrusive sensors capable of predicting
arc-flash in switchgear. Sensors include D-dot Sensor, Rogowski coil (RC) and
thermal ionization detector (TID).
x Modelling and simulations of RC Sensor.
x Assessment of reliability, sensitivity and operability of the mentioned sensors for
online monitoring of switchgear.
x On-site testing of operational switchgear for partial discharge (PD) faults using
unique D-dot sensor and reliability comparison with high frequency current
transformer (HFCT).
x Performance evaluation of de-noising technique based on discrete wavelet
transform (DWT) for the online measured signals.
2
x A novel technique for the detection of multiple faults in a single switchgear panel
based on the correlation between the cumulative energy and apparent charge of a
discharge event.
1.5 Dissertation Outline
This thesis is organized as follows:
x Chapter 2 describes the arc-flash in switchgear along with its types, energy exposed due
to it and devastating consequences of arc-flash. Moreover, the major part of this chapter
covers various causes of arc-flash in MV and LV switchgear.
x Chapter 3 gives a big picture of various arc-flash protection techniques including
passive and active methods for arc-flash protection. It also includes pre-emptive arc-
flash protection techniques along with possible measuring sensors.
x Chapter 4 explains the design, construction and models of the sensors used in this
research.
x Chapter 5 covers results and useful discussion about the major contributions of this
research. It includes de-noising performance, laboratory and on-site test results and a
technique for the detection of multiple faults in switchgear. It also presents laboratory
test results related to thermal monitoring of switchgear using TID. At the end of this
chapter, an outline for connection of the non-intrusive sensors with the central control
or supervisory control and data acquisition (SCADA) is given.
x Chapter 6 presents the major conclusions derived from this research work and
possibility of extension of this work in future.
3
4
Chapter 2
ARC FLASH IN SWITCHGEAR
2.1 Arc-Flash
The dielectric strength of air, in standard conditions, is about 3x106 V/m (3 kV/mm).
Thus dry air is reasonably a good insulator, and the air-insulated switchgear is very common
globally. However, when the temperature of air rises to between 2000-3000 K, air becomes
conductive due to thermal ionization. At approximately 6000 K, the air becomes conductive
ionized plasma, consisting of nitrogen and oxygen molecules as well as other ionized atoms and
electrons. The ionized air along with ionized material from the electrodes form the conductive
plasma channel between the electrodes, and this is how an electric arc originates. [Publication I]
According to the approved IEEE draft standard “PC37.20.7, 12-2006”, an internal arc
fault or arc-flash is defined as “An unintentional discharge of electrical energy through air
within the confines of a switchgear enclosure.” [5]
The heating of surrounding materials due to an arc creates carbon remnant that acts as a
high impedance conductor. In most of the cases, an arc fault is developed between the faulty
connections and is soon extended into line-to-ground or line-to-line faults due to the extension
of carbon paths. Following the discussion in conferences, there is some confusion concerning
the difference between an arc fault and a bolted fault (short circuit fault). Arc fault is a resistive
fault between the conductors, whereas, the short circuit current is the highest possible current
supplied by the source [6], [Publication VI], [Publication X]. Types and characteristics of arcing
fault are further elaborated below.
An arc fault can be divided into two categories:
x Series Arc Faults

x Parallel Arc Faults
2.1.1 Series Arc Faults
If there is a connection failure, which lies in series to the connected load as shown in
Fig. 2.1, it is called a series arc fault. Major reason is bad connection or poor joint between the
conductors of the same phase, bad termination of the cables, loose cable lugs etc. It is explained
in more detail in the next section.
In series arc faults, fault current is limited by the load and is less than the current setting
of the circuit breaker, hence the conventional over current protection does not necessarily
5
operate [7], [8]. So, the arc keeps on burning for longer time duration until it creates explosion
in the worst scenario, in the switchgear compartment [9].
Arc Fault
AC Supply Load
Fig. 2.1. Typical series arc fault
2.1.2 Parallel Arc Fault
Arcing faults occurring between phase-to-ground or between two phases are called
parallel arc faults, where the load comes in parallel to the fault as shown in Fig. 2.2. In most of
the cases, bus bars are very close to each other in switchgear panel and parallel arc faults are
generated after a series arc-flash due to excessive plasma and burning of insulators. Other major
reasons of the parallel arc faults are accidental contacts and foreign objects. They are explained
in the section 2.2.4. Unlike the series faults, the fault current is not limited by the load in
parallel arc fault. So, the voltage drop across the fault impedance is very small and fault current
is higher compared to series faults. Arcing current levels are in the range of short circuit current
or bolted fault current [8].
Arc Fault
AC Supply
Load
Fig. 2.2. Typical parallel arc fault
2.1.3 Energy of Arc Fault
The released incident energy of an arc fault is defined as the amount of energy
impressed on the surface as a result of an arc event, at a certain distance from the source. An arc
is developed within milliseconds and leads to the discharge of enormous amount of energy. The
incident energy of an arc depends on the following factors [4],[10], [11]:
x Working distance
x Voltage
x Current
x Arcing Time
6
Considering the distance and voltage as constant parameters, empirically the discharged
energy during an arc is directly proportional to the square of the current and the arcing time (arc
burning duration), i.e. Arc Energy ~ I2t [12]-[17]. So, there are two arc mitigation approaches:
current limiting and arcing time limiting. The time limiting approach is usually preferred,
because the lower current may lead to longer arcing times and higher incident energy. Industry
is already using time limiting approach as an arc mitigation tool. It is based on light and current
detection and has developed an effective and proven solution for arc protection which is good
enough to avoid big damage caused by the arc [4],[10], [11], [18].
2.1.4 Consequences of Arc Fault
A high energy arc fault releases very high temperature to the surrounding, rapid rise in
pressure inside the enclosure and an intense light. Temperature can go upto 20,000 K [10], [11],
[12], [Publication IX]. This increase in temperature causes air volume expansion and hence high
pressure waves inside the enclosure. Conventional switchgear is not designed to withstand this
high pressure. So the high pressure forces the switchgear walls to blow open, ancillary
equipment to become airborne missiles and hot gases, flames and burning metals to escape.
High temperatures cause thermal stresses in the objects, melting and fire in the surrounding.
High pressure and temperature affect in many ways on the personnel working in the vicinity of
switchgear. Intense UV radiations may cause permanent human blindness. Shock waves can be
responsible for hearing loss. Flying debris can cause serious injuries to the personnel. High
temperature can cause severe burn. Ejection of flames, hot gases and smoke may harm the lungs
seriously [10]- [12], [19]. Even with the recent enhanced focus on electrical safety, the risk to
personnel related to arc flash incidents is still quite high, with an estimated 2000 serious arc-
flash burns to workers every year in the United States [20]. The switchgear is an integral part of
a distribution network, and hence, a power interruption is resulted if an electrical fault occurs in
a switchgear compartment. Thus, the protection against arc flash is of paramount importance.
2.2 Sources of Arc Fault
The statistical information on arcing faults is limitedly available. Hence, the estimation
of causes is based on either a single case study or practical experience of professionals.
Categorically, these are divided into two major groups [Publication I]:
x Mechanisms gradually leading to arc ignition

x Immediate causes of arc ignition
2.2.1 Mechanisms Gradually Leading to Arc Ignition
Every object has its certain operating life cycle. During its operation, an equipment
undergoes through various environmental and physical disturbances which result in gradual
degradation of the equipment. Depending upon various parts inside the equipment and
associated properties, degradation mechanisms which may lead to an arc, are divided into the
following sub-categories.
a. Poor Connections
7
b. Degradation of Insulation
c. Poor Maintenance
a. Poor Connections
An electrical contact is defined as “the interface between the current carrying parts of
electrical devices that assure the continuity of electric circuit, and the unit containing the
interface” [21]. The primary purpose of an electrical contact is to allow the uninterrupted flow
of electrical current through the contact interface. Electrical contacts are mainly divided into
two major categories, stationary and moving. Fig. 2.3 represents a general classification of
electrical contacts according to contact kinematics, functionality, and design features. In
stationary contacts, contact members are connected rigidly or elastically to the stationary unit of
a device to provide a permanent joint. Stationary contacts are divided into non-separable or all-
metal (welded, soldered and glued), and clamped (bolted, screwed or wrapped). In moving
contacts, at least one contact member is rigidly or elastically connected to the moving unit of a
device [21]. In this research work, the study of stationary joints is considered only, which
include cable termination joints or bus bar connections. Most of these connections are separable
(bolted or clamped).
Fig. 2.3. Classification of electrical contacts
An uninterrupted passage of electric current through an electrical contact can only be

achieved if a good conductor-to-conductor contact is established with a minimum resistance.
The processes occurring in the contact zone are complex and not fully explained within the
limits of currently available knowledge [21]. Although the nature of these processes may differ,
they all are governed by the same fundamental phenomena, the degradation of the contacting
interface and the associated changes in contact resistance, load, temperature, and other
parameters.
According to a case study performed for Science Applications International Corporation
(SAIC), by John Hopkins University Applied Physics Laboratory and US Navy, 60-80% of all
arcing events occurred in the switchgear were caused by the faulty connections [22]. A poor
connection or joint is defined as a connection which has a resistance higher than the conductors
entering or leaving the joint. The mechanism of degradation of electrical contacts is well
8
reported in many of the research articles [22]-[25], [Publication I], [Publication VI]. Failure, in
this context, is defined as the point at which the connection is no longer capable of supporting
the function for which it was designed, i.e. to provide uninterrupted flow of electric current. In
other words, it is the condition in which the bulk temperature of the joint, under normal current-
transfer conditions, rises significantly above the permissible connection operating temperature
and causes gradual degradation leading to melting of copper material. The complexity of failure
mechanisms in power connections is best described in the form of a cyclic process, as shown in
Fig. 2.4.
Fig. 2.4. Cyclic process of connection degradation
The resistance of a poor connection highly depends on the surface and cross sectional
area of the conductor as well as the pressure exerted between the joints. When nominal load
current flows through a faulty connection having elevated resistance, it causes the connection to
heat up. Localized heating increases the rate of oxidation of the conductors at the connection in
the presence of moisture. It also causes expansion of the conductors and thus stresses the
connection. When the power current reduces, the connection cools down and contracts.
Repeated expansion and contraction of the joint material tends to further loosen the joint.
Loosening of the joint and the oxidation further raises the joint resistance. Every load cycle of
the circuit tends to ratchet the resistance of the joint higher and thus increases the temperature of
the joint. Finally, the joint temperature rises to a level that causes melting of the connection
material, and eventually leads to an arc-flash. A corrosive atmosphere or vibration can increase
the rate of failure [Publication I], [Publication VI].
The above mechanism can be easily understood by having a look at the Figure 2.5. Note
that almost 5 inches of the insulation of conductor is damaged. Oxidation can also be easily
observed on the damaged part. [22]
9
Fig. 2.5. Damaged conductor due to overheating [22]
b. Degradation of Insulation
Each insulation system shows characteristic defects depending on various factors
including type of insulation, load profile, physical conditions etc. For instance, the aging of a
transformer is dominated by the thermal load on the solid insulating material consisting of oil
impregnated paper. At generators, the degradation of the insulation system is a multiple process
of electrical, mechanical and thermal load. Similarly, switchgear insulations involve degradation
of bushings, current transformers (CTs), potential transformers (PTs) and cable terminations etc.
Generally, the life time expectancy of an electrical insulation system depends on the following
TEAM stress [26]-[28],
x T = Thermal Stress
x E = Electrical Stress
x A = Ambient
x M = Mechanical
TEAM stresses are elaborated in Fig. 2.6. Thermal stresses in the switchgear applications result
into oxidation of aluminium or copper conductors, weakening of insulation, hydrolysis of the
insulator and thermo-mechanical effects. If the system is not supervised thermally, it ultimately
leads to an arc. Thermal stress is more prominent in LV equipment.
Fig. 2.6. TEAM stresses
10
Electrical stress is the major cause of insulation degradation in MV and HV electrical
insulations. Insulation material may experience excessive stresses caused by either partial
discharge (PD) or overvoltages. These phenomena are explained in detail in the following
section.
Ambient conditions such as moisture, gas and chemicals, present in the environment,
gradually deteriorate the chemical properties of the insulation and conductors. For instance,
moisture plays an important role in rusting the conducting material. Electrical joints and
contacts exposed to moisture get corrosion within a few days. Weathering radiations also
deteriorate the insulation strength of the material. Moisture gives rise to the electrical stresses in
the electrical components, which further accelerates the insulation damage process.
Mechanical stress also takes part in the process of insulation degradation. Mechanical
vibrations present in the surrounding environment cause cracks and mechanical breakdown of
the insulation material. Small cracks (less than 1 mm) cause electrical stresses across it.
i. Partial Discharge (PD)
The term partial discharge is defined by International Electro-technical Commission

(IEC) 60270, High Voltage Test Techniques- PD measurements, as “a localised electrical
discharge that only partially bridges the insulation between two conductors and which may or
may not occur adjacent to the conductor” [29]. PD can occur at any point in the insulation
system, where the electric field strength exceeds the breakdown strength of that portion of the
insulating material. It is confined in a limited area that doesn’t permit complete failure of the
system, until a complete breakdown occurs. Major causes of PD are [30]:
x Voids in solid insulation during manufacturing

x Mechanical breakdown or cracks in solid insulation
x Surface contamination and irregularities in solid insulation
x Particle contamination in the liquid insulation
PD occurs due to displacement of charge. This produces a rapid rate of change of

current that occurs when the dielectric begins to breakdown. This rate of change is highly
dependent on the environmental condition, nature of dielectric material and applied voltage
[30]. As PD is a small electrical avalanche caused by locally disrupted electric fields in
dielectric materials, it is a symptom of insulation weakness. Once PD starts, it causes
continuous damage to the insulation and ultimately an arc and fire [31]-[35]. PD can be divided
into following three types:
x Internal PD
o Cavity Discharges
o Treeing Discharges
x Surface PD
x Corona
11
The discharges from gas-filled voids, delaminations or cracks, etc. within a solid
insulation are called cavity discharges. The impurities in the insulation materials are shown in
Fig. 2.7. Voids may be present in the solid insulation during its manufacturing or air bubbles in
the liquid insulations. Cracks may occur due to vibration and thermal breakdown. A tree like
damage may start inside the solid insulation as a result of void or impurity in it. This is called an
electrical tree. Treeing discharges are mainly due to current pulses within an electrical tree. An
example of an electrical tree is shown in Fig. 2.7. Internal discharges are in many cases crucial
for the life of the insulation. Insulation life depends on several variables, the most important
ones being field strength, discharge magnitude, and material. [36]
a) gas bubbles
b) void between electrode and
insulation
c) void at insulation interface
d) foreign particles in insulation
e) electrical treeing
Fig. 2.7. Bubbles, void, foreign particle and treeing in solid insulation
Discharges occurring on the surface of an electrical insulation, where the tangential field
is high, are called surface discharges found in e.g. porcelain or polymeric housing of HV
devices, cable terminations or electrical joints. These discharges usually occur from the edges of
the electrode parallel to the surface of the dielectric. Hence they are also called “gliding
discharges” or “edge discharges” [35]-[37]. Surface discharges occurring at various insulation
interfaces are shown in Fig. 2.8. Majority of PDs in the switchgear panel are surface discharges
occurring on the surface of the bus bar insulator, bushing and cable terminations. Surface
discharges are less dangerous than the internal discharges because they are less concentrated
and mostly involve ionization of air and contaminants on the surface of an insulator, hence the
rate of damage is much slower than in the internal discharges. Hence, they are mostly visible
and can be tolerated for longer period if no corrective action can be taken.
a) electrode boundary
b) proximity of sharp electrode
c) series insulation interface
(boundary of air/solid insulation)
Fig. 2.8. Surface discharges
When the electric field around the surface of a conductor or an electrode exceeds the
breakdown limit of the gas molecules, it ionises the surrounding gas. This is called corona
discharge. It may occur around the electrodes of small radius of curvature. Corona is not much
dangerous for the bare conductors in open air insulations because ionized molecules are
constantly being replaced by the unionized molecules due to natural air flow and cooling
12
phenomenon. But in case of switchgear it may lead to heating, thermal ionization, chemical
emission and surface discharge which may ultimately cause faster degradation of insulation.
Byproducts such as ozone and nitric acids may chemically deteriorate nearby insulation
materials [30], [36], [37]. Corona at conductor’s sharp edges and on surface irregulatity of an
electrode is shown in Fig. 2.9.
(a) (b)
Fig. 2.9. Corona at (a) sharp edges, (b) surface irregularities of electrode
Figure 2.10 shows a well-known three capacitance model (a-b-c circuit) of PD source
due to an internal cavity. Any dielectric material can be represented by a capacitance. In Fig.
2.10, Cc represents the capacitance of the void as shown in the figure below, Cb shows the
capacitance of the dielectric material (healthy part) in series with Cc, whereas Ca is in parallel
with both Cc and Cb. Ua is the total voltage applied to the dielectric and Uc is the voltage stress
across the void. I shows the faulty part of dielectric and II shows the sound part of it in this
figure. Voltage stress across the void can be represented by the following equation [30],[31],
[36].
Cb (2.1)
Uc Ua
Cc Cb
Fig. 2.10. Equivalent circuit of a dielectric with cavity
Same representation can be shown for the surface discharges, which are more common
in the switchgear applications at the cable connection points, in Fig. 2.11(a). The capacitance
between the electrode and the dielectric surface having PD is shown by Cc, capacitance in series
13
with it, is shown by Cb and parallel capacitance is shown by Ca[30],[31]. Fig. 2.11(b) shows the
three capacitance model for corona. Capacitance of plasma material is shown by Cc, Cb is the air
capacitance in series to the Cc whereas Ca is the air capacitance in parallel to these two
capacitances.
(a) (b)
Fig. 2.11. Three capacitance model for; (a) surface discharges, (b) corona
The electric fields inside the conductor insulation (Ea) and in flat cavity (Ec) are given as,
De (2.2)
Ea
Hd
De (2.3)
Ec
Ho
where De is the electric flux density which is directly proportional to the applied voltage Va to
the dielectric material, εd is the permittivity of the insulating material and ε0 is the permittivity
of air. If the voltage is applied to such dielectric material, the electric field in the cavity becomes
greater than the field in the surrounding dielectric as a result of the lower permittivity of the air
or gas in the cavity, i.e., as εd ൐ε0 thus, Ec ൐Ea. At a certain instant, the electric field becomes
sufficiently high in the cavity and air can break down. As a result the cavity goes from non-
conducting to conducting, and the field in the cavity goes from very high to nearly zero
immediately after the discharge [30], [31].
The PD signal is the result of the change in the apparent charge on the electrodes which
is basically the transient change in the electric field distribution caused by the discharge. Such a
discharge generates a voltage PD signal between the system conductors as a result of the change
in the electric field configuration, which takes place when the discharge occurs. This
phenomenon can be well understood further by considering the basic relation between
capacitance, charge and voltage.
q (2.4)
C
V
Hence from the equation 2.4, it is clear that when the cavity goes from non-conducting to the
conducting phase, the voltage decreases and capacitance increases. In order to maintain constant
14
voltage across the dielectric material, a current (flow of charge) must flow down from the
supply. This current flow generates a voltage pulse, which propagates down the electric circuit.
It is further explained in detail with the help of Fig. 2.12, as follows.
Fig. 2.12. Electrical mechanism of PD activity
The applied voltage Ua =Up (sinωt + ϕ) varies periodically and completes its cycle as
shown in Fig. 2.12. When it increases to such a value that the cavity voltage Uc approaches to
the breakdown voltage Ui+(in the positive half cycle), cavity collapses and a rapid displacement
of charges, also called space charge, takes place inside the cavity. Due to transfer of the charge,
the potential difference across the cavity drops to a value Ue+ (extinction voltage, in the positive
half cycle) at which the electric field intensity becomes lower than that of required level in order
to maintain the discharge. The discharge ceases at this instant and the charge starts to
accumulate across the boundaries of the cavity. The voltage Uc starts to build-up due to the
increasing value of the applied voltage Ua. Again, the PD collapses at the Ui+ and another PD
event takes place. This activity goes on repeatedly. The PD activity occurs during both positive
and negative cycles of the applied voltage. During PD event the amount of space charge can be
expressed as [30], [38]
qo Cic Ui / U e / (2.5)
where qo is also called the actual charge which is distributed within the cavity. The PD event
occurs in the bulk of the insulation material and therefore, subsequently a proportional charge is
electro-statically induced on the electrode (cable conductor). The induced charge depends on the
location and magnitude of the actual charge qo. The relationship between actual charge and
induced charge can be obtained by using Poisson's equation implemented in [38]-[41].
15
Fig. 2.13. PD pulse in electrical power systems; mathematical (ideal) pulse vs. practically measured
typical PD pulse
The discharge phenomenon is extremely fast and accomplishes in a very small fraction
of the period of applied voltage cycle (having ns to μs duration). Respectively, the induced
charge is created rapidly during each PD event which results in a high frequency current pulse
induced on the electrode. A measured PD pulse is shown in Fig. 2.13. The mathematical model
of the pulse can be described as,
i pulse(t ) A eD2t eD1t (2.6)
where A is the peak value of the pulse, α1 is the rate of rise-time and α2 is the rate of fall-time.
Rise-time of the pulse is an important parameter of PD activity and also determines the highest
frequency components of the PD signal. The mathematical PD pulse is compared with the
practically measured PD pulse in Fig. 2.13. Comparison of both the pulses shows that practical
PD pulse contains high frequency components which are due to line’s inherent characteristics.
It is already explained in the PD characteristics, that primarily PD causes heat and

chemical emission. Main chemicals formed are ozone and nitrogen oxide. Heat and these
chemicals can easily deteriorate the insulation materials. Once PD begins, deterioration of
insulating materials continues ultimately leading to complete breakdown of insulation, which
may lead to an arc and fire. Repetitive discharge events cause irreversible mechanical and
chemical deterioration of the insulating material. Damage is caused by the energy dissipated by
high energy electrons or ions, ultraviolet (UV) light from the discharges, ozone attacking the
void walls, and cracking due to chemical breakdown. The chemical transformation of the
dielectric also tends to increase the electrical conductivity of the dielectric material surrounding
the voids. This increases the electrical stress in the unaffected region, accelerating the
breakdown process (tree formation). A number of inorganic dielectrics, including glass,
porcelain, and mica, are significantly more resistant to PD damage than organic and polymer
dielectrics. Consequently, PD detection and measurement is the most reliable source for the
estimation of life expectancy of the electrical insulation and electrical equipment.
16
ii. Overvoltage
Any level of voltage above the rated voltage level for which the equipment is designed
is called over voltage. When an electrical insulator is designed, it is tested and two voltage
levels are defined for it, i.e. puncture voltage and flashover voltage. Puncture voltage is the
level across the insulator which causes a breakdown and conduction through the interior of the
insulator. The heat resulting from the puncture arc usually damages the insulator irreparably.
Flashover voltage is the voltage which causes the air around or along the surface of the insulator
to break down and conduct, causing a 'flashover' arc along the outside of the insulator. There are
following major types of overvoltages occurring in power network [36]:
x Temporary (sustained) Overvoltage

x Slow-front (switching) Overvoltage
x Fast-front (lightning) Overvoltage
x Very-fast-front Overvoltage
Earth faults are the most common cause of temporary overvoltages. Magnitude of the
overvoltage depends on the type of earthing. During the earth fault in the unearthed and
compensated neutral system, healthy phases experience overvoltage between a single phase to
ground, approximately upto the level of line voltage. Another main reason of the temporary
overvoltages is load disconnection. Voltage level increases when a load is disconnected from
the grid. The level of overvoltage is more when load is inductive and grid’s short circuit power
is small.
Slow front overvoltages are mainly caused by the changes in the electrical circuit due to
switching operations, either by protection operation or load rearrangement. Magnitude and
shape of overvoltage depends on the switchgear used for current disruption and network
properties (resistance, inductance and capacitance). Connecting load to the network, fault
isolation or load disconnection are the major causes of such faults. Lightning is a well-known
phenomenon in HV engineering. It is a large discharge starting from clouds to ground or vice
versa. Overvoltages occurred due to lightning are mainly because of straight hit of lightning arc
to the conductor, back flashover through high impedance ground connection and induced
overvoltage. Very fast front overvoltages are caused by arc interruption and restriking action
while opening a disconnector. Tens of restrikings may occur during opening. Each restrike
generates high frequency oscillations into the network. Fig. 2.14 shows the voltage level
comparison between different over voltages stated above.
17
Fig. 2.14. Level comparison between different overvoltages
c. Poor Maintenance
Less skilled workers working on the maintenance of switchgear and other electrical
equipment may not ensure the quality of work. For example, they may not ensure the joints and
terminations of the cables to be clean and tight enough to avoid sparking between them. This
will lead to an unwanted electrical fault.
2.2.2 Immediate Cause of Arc Ignition
From the safety point of view, gradually developing electrical faults can be considered
less hazardous to personnel than those caused by direct human interaction, because, in gradually
developing faults, it is less likely that there are humans in the dangerous area. However, many
arcing faults ignite immediately, without giving signs (e.g. emissions), on which prediction
could be based.
a. Accidental Contacts
Slipped or forgotten tools and equipment malfunction during a maintenance work are
examples of direct human interaction potentially leading to an arc-flash. Sometimes, workers
forget tools inside the switchgear compartment. In the worst case they may keep the tools
between the two bus bars creating a conductive path between them and forget to remove them
after the maintenance is over. When the switchgear is energised in this case, arc occurs in the
compartment.
b. Foreign Objects
Foreign objects or animals can also lead to an immediate arc-flash. Lizards, snakes, rats
or other animals moving around in the switchgear compartment may bridge between the two
conductors or bus bars and may cause arc-flash. These types of faults are more challenging for
prediction by online monitoring system. One type of monitoring option might include motion
sensors that can automatically switch protection modes to an instantaneous tripping setting.
18
Chapter 3
ARC-FLASH PROTECTION TECHNIQUES IN SWITCHGEAR
3.1 A Holistic View of Existing Arc-flash Protection Technologies
Various approaches targeting the prevention and mitigation of arc-flash incidents can be
found in the literature and are summarised in [Publication II]. Arc-flash protection methods can
be categorized into various ways. Categorization can be based on division between passive and
active methods, and between proactive and reactive approaches as shown in Fig. 3.1. Passive
methods do not have any active protective component. Active methods include measurements
and either reactive response (after arc-flash has been detected) or proactive response (pre-arc
action when indication of a developing fault has been detected). Figure 3.1 presents the
categorization and a holistic view of arc-flash protection. In the following sections, the elements
of the holistic view are examined in more detail.
Fig. 3.1. A holistic view of arc fault mitigation principles
3.2 Arc-flash Prevention
Approaches aiming at prevention of arc-flash, like design of switchgear, education of

personnel, and maintenance practices, can be categorized into pre-ignition methods and passive
methods. These methods are discussed in this section.
19
3.2.1 Education and Training
Either direct or indirect human action has been a common cause in the reported arc
faults. Education of workers, safety culture and rules regarding operation in an environment
with possible arc-flash hazard should be a common practice. Incident energy calculations and
arc labelling of switchgear are a good example of safety culture. Education of personnel is
justified for safety and economic reasons. Arc faults cause direct damage costs and costs due to
process interruption, and when humans are involved, high medical or legal costs are added too.
3.2.2 Preventive Design
Prevention of internal arcs in switchgear starts naturally in design. Requirements have

been set in IEC and IEEE standards [42]-[45]. The IEC standard [42] presents a list of locations
where internal arc faults are most likely to occur in metal-enclosed switchgear and controlgear:
x Connection compartments
x Disconnectors, switches, grounding switches
x Bolted connections and contacts
x Instrument transformers
x Circuit breakers
The standard also lists possible causes of internal arc-flash and examples of measures to
decrease the probability of faults. These measures include both technical solutions and
personnel related recommendations. Some additional design options are evaluated below.
3.2.3 High Resistance Grounding
Conventional high resistance neutral grounding helps to limit single phase-to-ground

fault currents. It can be a useful technique as an arc-flash prevention. It reduces the dissipated
energy and sustained arcing faults in LV systems. This is why the proponents of this system
have proposed that it should become a standard for the industry [46]. However, high-resistance
grounding is only effective for ground faults and usually ground faults may lead to a phase-to-
phase fault inside switchgear [47]. Although majority of arc faults start as phase-to-ground
faults, other protection means are also necessary when high resistance grounding is applied.
Additionally, high-resistance grounding seems to have very limited window of application in
MV and HV systems for several reasons explained in [48].
3.2.4 Insulated vs. Bare Busbar
Apparently, insulated bus bars are believed to perform better than bare ones when it
comes to arc-flash, but there are different opinions about it. First, it is obvious that insulation
provides means to reduce the probability of arc-flash caused by falling objects or by vermin.
Another advantage is that insulation prevents single-phase faults from escalating to high power
multi-phase faults [49]. The third possible advantage is that the arc may travel to insulated area
and become self-extinguishing [50]. This is, however, controversial. In tests and real arc-flashes
reported in [6], the arc moved slowly and burned through the insulation, causing more damage.
The insulation does not necessarily extinguish the arc. The movement of arc has also been
20
noticed in the tests for partially insulated switchgears, which is fast with bare bus and stops at
the starting point of the insulation. Somewhat stable arc at the insulating barrier point causes
melting and vaporization of metal, leading to more significant damage compared to moving arc.
Fig. 3.2 illustrates the impact of a stationary arc at the point, where bus insulation starts.
The figure presents LV busbars after an arc-flash test with fast protection. It must be noted that
in this case the buses are only partially insulated. The insulation does not cover the joint area.
Although the duration of the arc in this test was only some tens of milliseconds, the arc melted
some of the busbar metal at the insulation barrier point. This observation is in line with the one
made in [6], i.e. if plasma is allowed to concentrate, the rate of damage will accelerate. Also
according to LV testing reported in [51], an insulating barrier prevents arc motion, and produces
higher arcing currents and higher incident energy.
Fig. 3.2. Partial melting of LV busbars after short time stationary arcing at the insulation barriers
(U = 726 V, Ik = 77 kA, tarc = 65 ms)
3.2.5 Maintenance
The probability of arc-flash can be decreased by appropriate preventive maintenance of

the equipment. Visual inspection, thermal imaging, PD testing, and time-based testing of
protection devices are some key examples of preventive maintenance actions. Preventive
maintenance is effective against aging, corrosion, pollution or vibration related faults. Condition
monitoring of equipment, often by on-line measurements and communication to upper level
information systems, is an advanced tool for active maintenance of the equipment to prevent
arc-flash.
3.3 Arc Flash Reactive Protection
Arc-flash due to immediate causes such as direct human interaction or equipment

malfunction are still challenging for preventive or predictive protection measures. This is why
reactive protection is justified. As the term defines, it is a protection which reacts within a small
fraction of time after the arc ignition. However, traditional over current protection is fairly
inefficient in arc faults, as explained in section 2.1.
21
Incident energy of an arc-flash is explained in [16]. Incident energy calculation helps to
evaluate the safe working distances from the potential arc ignition point as well as estimating
the potential damage to the equipment and need for PPE. The incident energy depends on
voltage, distance, current and arcing time [16]. In practice, the key factors are arcing current and
arcing time. Some protection approaches emphasize limitation of current while most solutions
aim at short arcing time.
3.3.1 Current-Limiting Approach in Arc-Flash Protection
Arc-flash is a random phenomenon which highly depends upon the arc resistance and
the involved electrical circuit. The length of the arc also varies in different cases. The magnitude
of arcing current is often derived from bolted fault current, but the bolted fault current is not
constant either. The highest incident energy can be caused by the highest or lowest bolted and
arcing fault current. This is because lower arcing fault current often leads to longer trip time in
the conventional protection methods (inverse time overcurrent protection) [14], [16], [52]. The
limitation of arcing current leads to reduction of incident energy, provided that the arcing time is
not significantly affected. The impact of the reduction of arcing current should not be analyzed
without examining the impact on operation time of the protection, the arcing time and the
released energy. According to [53], the initial symmetrical short-circuit current is calculated as:
cU n (3.1)
I k"
3Z k''
where c is voltage factor (0.95, 1.00, 1.05 or 1.10), Un is nominal system voltage (line-to-line),
and Z’’k is the sub-transient short-circuit impedance of the network. Figure 3.3 presents the
system diagram and the equivalent circuit diagram related to above equation. The equation
illustrates the importance of impedance of feeding network (consisting of RQt and XQt).
Fig. 3.3 System diagram and its equivalent circuit to calculate short circuit current [53]
The calculation of arc flash current from the available short circuit fault current has been
studied in [14], [54] and [55]. The available fault current data from the utility is not always
available. In calculations, the utility network is assumed to be an infinite bus, and this may lead
to an incorrect estimation of minimum fault current. The source impedance can also vary
because of changes in network connection or connected generation sources. The fault current
levels vary on the basis of type of generators, for example; conventional synchronous generators
provide short circuit current 5-10 times their nominal current, photovoltaic systems provide
fault currents only close to their nominal current [14], whereas, in Denmark, the high share of
wind power has led to reduction of short-circuit power [56]. The available short-circuit power
22
has major impact on short circuit current on MV level compared to LV (< 1000 V) level due to
high impedance of MV/LV transformers.
For high-voltage arcs, the arcing current is practically the same as the bolted fault
current, and it can be calculated such as in IEC standard 60909 [55], [57]. The arc-flash current
is lower than the bolted fault current for applications under 1000 V [16]. An arc-flash current
can be typically between 20-40% of the prospective or bolted fault current as reported in the
literature such as in [15], [57] and [58]. This lower value of the arc-flash current sets challenges
to the protection. The protection must be able to detect the fault, and able to operate extremely
fast in order to avoid severe damages. When inverse time over current protection is applied, the
faults with high fault currents will be cleared much quicker than the low current faults.
The behavior of current-limiting (CL) fuses depends on the current level as well. When
the current is in the current-limiting range, CL fuses are very effective in reducing the incident
energy level, the pressure wave and even the mechanical stress caused by the high fault current
[59], [60]. When CL fuses are used in combination with the switching devices, the maximum
arc energy may occur at current levels below the maximum interrupting rating [42]. The risk of
higher incident energy level because of lower current is also clearly illustrated in [16], where
test result figures for CL fuses are presented.
Another significant risk caused by low fault current is that the protection may not
properly respond or even trip at all [55], [59]. In conclusion, current-limiting approach to
mitigate the impact of arc-flash includes many uncertainties and risks related to the arcing time
and incident energy.
3.3.2 Limitation of Incident Energy by Reducing Arcing Time
Incident energy is directly proportional to the arcing time as illustrated in [Publication

II]. Reduction of arcing time can be a promising approach in arc-flash mitigation. Various
approaches for reducing arcing time have been presented in [61]. Compared to the traditional
over current protection, the protection time in bus bar faults can be improved by zone-selective
interlocking. Traditional bus bar differential protection provides better performance with an
operation time of 1 cycle only, however, the bus bar differential protection is not effective if the
fault occurs in cable compartment, typically in cable terminations. In order to implement fast
tripping, fast detection is required.
3.3.3 Light and Overcurrent Based Protection
A successfully proved, simple and reliable technology for arc flash mitigation is light
and overcurrent based protection. It is based on simultaneous detection of light and overcurrent
and provides an extremely fast and secure arc flash detection and mitigation [62]. Along with
detection of phase overcurrent, zero-sequence overcurrent detection can also be applied to
indicate phase-to-ground faults. Figure 3.4(a) illustrates the basic operation principle of light
and overcurrent based protection.
Optical fiber based point type and loop type sensors are commonly used in this
application. The sensitivity level of the sensor is not a critical issue because the light intensity
23
caused by an arc-flash is very high compared to ambient light intensity. However, special
attention has to be paid for the detection of light in the circuit breaker compartment of LV
systems, if the circuit breaker (CB) emits light. Figure 3.4(b) presents a scheme of a simple arc-
flash protection system which includes current measurement, light sensors, the input/ output
(I/O) unit for sensors, and the arc-flash relay. Existing current transformers (CTs) can be
utilized for the current measurement. Like the light, the current can also be detected within 1ms.
Thus, the arc detection time using light and current based detection is approximately 1ms and
the whole operation time of this protection is based on the switching device, such as CB. Thus it
is most important to ensure that the CBs are in working order.
When applying current and light based arc-flash protection, selective protection can be
achieved, especially with optical point sensors indicating the faulty section. This means that e.g.
in cable termination faults, only the breaker of the outgoing feeder will trip while the rest of the
systems remains energized.
(a) (b)
Fig. 3.4 (a) Basic principle of light and overcurrent based protection, (b) Arc protection system based on
light and overcurrent detection
3.3.4 Arc Eliminators
An arc eliminator is a device which eliminates an arc extremely fast by creating an

intentional parallel short circuit. The technology is also called crowbar technology, arc
quenching or high speed grounding. The operation principle is quite simple, i.e. when an arc-
flash is detected, the device creates an intentional high speed short circuit in the system so that
the voltage collapses and the arc is extinguished. So far, arc eliminators have not been widely
applied, although the technology is listed as an option to provide highest possible level of
protection in IEC standard [42]. Various manufacturers have different technologies related to
the methods for creation of short circuit, e.g. pyrotechnical pressure element, Thomson coil,
micro gas cartridges, or a spring mechanism assisted by an electromagnetic (EM) repulsion
system.
Arc eliminators can be used in combination with light and overcurrent based arc-flash
detection. The arc detection system is able to trip the eliminator within 1-2 ms, and the high
speed primary circuit device operates within a few milliseconds. Fast communication between
24
the arc-flash protection relay and the short-circuit device is necessary. A typical arcing time is
less than 5 ms. The arc-flash relay also sends a tripping command to the CB, and the CB
eliminates the short-circuit current within a few cycles. Since, the arcing time is minimal, the
hazards and damages caused by arc-flash are almost nonexistent. Obviously, a power system
interruption is caused when an arc eliminator operates. [63], [64]
This technology has been opposed due to the concerns on high currents. When
estimating the risks caused by the high current due to short circuit, two additional issues should
be kept in mind. Transformer failures caused by short-circuit events are relatively rare, and
according to IEEE and IEC standards, transformers are designed to withstand the EM forces and
the thermal stresses produced during the flow of a short-circuit current for a certain time
duration [65]. Additionally, since it is very difficult to find any evidence or experience of real
cases where arc eliminators had caused damage to equipment, the risk level related to high
current caused by arc eliminators can be estimated as acceptable.
3.4 Pre-emptive Arc Fault Protection
3.4.1 Online Condition Monitoring (Condition based maintenance)
The nature of incipient faults has been discussed in Chapter 2. There are various
methods applied in practice and reported in literature for the diagnostics of such faults. For
instance, the conventional methods for PD measurement are based on the measurement of
electrical quantities (current or voltage pulses) [30], [66], [67]. The IEC-60270 standard defines
the conventional measurement setup for the measurement of charge displacement observed at
the terminals of the test object (apparent charge) [68], [69]. Sensors used in this method can be
e.g. a pulse transformer, measuring impedance, HFCT or RC. A PD current pulse can be
measured which helps to quantify the extent of PD in terms of coulombs or amperes. There are
certain drawbacks of the conventional techniques. Measurements are done in shielded
laboratories or heavy measuring equipment need to be transported to the site. The mechanisms
and signal properties of discharge are not completely understood through the examination of the
PD pulse alone. The whole measuring system is quite expensive. The diagnostic measurements
are done on periodic basis and the equipment under test needs to be de-energised, which results
in interruption of power supply, even if the equipment is healthy. [70], [71]
Electrical discharges are accompanied by various physical effects such as emission of

UV light radiation, thermal radiation and ultrasound, as well as EM radiation. Unconventional
methods mostly rely on the measurement of these physical quantities instead of electrical
parameters. Hence, unconventional sensors do not require any connection to the HV equipment
(non-intrusive feature). These measurements are relatively new techniques for discharge
detection in various situations such as PD and poor connection. These methods have not yet
gained worldwide acceptance due to inability to quantify the discharge extent in terms of
coulombs or amperes. Another challenge is the reflection of the signals from enclosures and
ground connections. Multiple reflecting signals make the interpretation and analysis of recorded
data difficult. On the other hand, these unconventional measurements have various advantages.
Equipment can be monitored non-intrusively and the health condition of the equipment can be
25
assessed in real time. It also saves un-necessary shutdowns and man hours required for periodic
diagnostics. Detailed comparison between the conventional and unconventional techniques is
given in [69], [72]. Consequently, instead of time based maintenance, condition based
maintenance concept is evolved due to the recent developments in unconventional techniques.
While knowing the real time condition of the electrical equipment, any possibility of potential
fault can be predicted. Hence, necessary actions can be taken in order to avoid severe damage.
This is the main motive of this research.
3.4.2 Physical Emissions Indicating Developing Faults
Physical parameters resulted by the incipient faults and their measuring technologies have been
discussed in briefly this section. Same are presented in detail along with a comparison table in
[Publication I].
a. Electromagnetic (EM) Emissions

Currents flowing through the loose connection cause micro-sparks and ionization of the
air molecules in the surroundings (plasma creation) if the localized electric field exceeds the
dielectric strength of the air. Internal PDs cause surge of electrons (current) in the dielectric
material. Surface discharge and corona also ionise the insulating air around the electrode and
create plasma. All these phenomena are the sources of EM radiations. The propagation and
frequency of the EM radiations depend on the rise time of the current waveform which varies
with the nature of PD, location and material of insulation. The initial rise time of the current
waveform produced by PD is sufficiently small to extend the frequency spectrum to the radio
frequency (RF) region [73]. PD pulses are somewhat random and vary in terms of rise time.
Hence the spectrum of the RF emissions is in wide band spectrum. According to [67], the most
favourable frequency bands for proactive protection are high frequency (HF): 3-30 MHz, very
high frequency (VHF): 30-300 MHz and ultra high frequency (UHF): 300 MHz - 3 GHz.
b. Acoustic (Ultrasonic) Emissions

Mechanical vibrations are resulted in the electrical equipment, due to PD, which
produce acoustic signals in ultrasonic region. This technique has been widely used for PD
detection in various electrical equipments. Acoustic signal from a PD source is immune to EM
noise. Frequency of the signal highly depends on the type of PD. In the switchgear applications,
most of the PDs are surface discharges, so the range of acoustic signal is in the ultrasonic
region. They can be detected by piezoelectric transducers, fibre optic acoustic sensors,
accelerometers, and sound-resonance sensors usual using frequency band between 10 kHz and
300 kHz. Ultrasonic detection has been successfully used in order to locate the PD source inside
of the test object due to the fact that acoustic signal is strongly dependent upon the geometry of
the test object. Combined with electrical measurement techniques, acoustic measurement can
enhance its strength. This method is very efficient for localizing PD source because of its
immunity against EM noise. But the main issue is background sound signals which are very
common in industry. [74]-[79]
26
c. Optical (Ultraviolet-UV) Emissions
Optical or ultraviolet (UV) signals are produced as a result of various ionization,
excitation and recombination processes during an electrical discharge (PDs or arcing). The
intensity and wavelength of optical signals is dependent on various factors, such as insulation
material, temperature, discharge intensity and air pressure. Because PDs vary widely in its
nature, hence the resultant UV emissions lie in all the regions of optical spectrum i.e. UV,
visible and infrared (IR), however, the most dominant is UV. At the cable terminations inside
the switchgear, corona emits the spectrum range around 280 nm to 405 nm at the MV and HV
levels. The spectrum of strong flash is between 400 nm to 700 nm [80]. Roughly speaking, the
emitted light is proportional to the amount of charge transferred due to PD. There are roughly
two kinds of optical PD detection techniques, i.e. direct detection of optical PD signal and
detection of change of an optical beam. A method called optoacoustic measurement catches
sonic or ultrasonic range acoustic emission caused by PD which results in deformation of the
optical fibre. The main advantage of optical method is the immunity to electromagnetic
interferences (EMI) and high sensitivity compared to conventional electrical techniques. [81]-
[82]
d. HF Current Components
As it is discussed in detail in Chapter 2, PD signals are the current pulses or surges of
electrons. These current pulses are superimposed on the normal load current. These pulses have
very small rise time and high frequency. Power frequency (50/60 Hz) current measuring devices
such as CTs are not sensitive enough to record such high frequency pulses but high frequency
sensitive current measuring equipment is needed. This is quite mature technology nowadays and
has been applied for PD detection in various electrical equipments so far [83]-[85].
e. Thermal Emissions (IR emission)

One major cause of arc-flash is bad connections, which are explained in Chapter 2.
These mainly generate hot spots and high temperatures in both MV and LV switchgear. Arcing
also causes heating at loose contacts. In LV equipment (415 V) PD phenomenon is rare, hence
in LV equipment, thermal effect is more prominent than others. Heat produced due to these
phenomena lies in the range of infra-red spectrum of EM radiations. It can be measured by IR
thermometers, thermocouples, bimetallic sensors etc. One of the drawbacks of using
conventional thermal sensors is that a large number of sensors are required to provide an
adequate protection level for the whole equipment. This results in a complex and costly system.
High temperature cause thermal ionizations. Gases ionise thermally much before than the
metals. Some thermal ionization sensors have been implemented for switchgear applications but
they are not so common commercially [22], [6].
f. Chemical Emissions
During the process of ionization a neutral item or molecule loses or gains electrons
thereby acquiring a net charge. These ions combine with the ions of other atoms to produce a
molecule which is called byproduct of PD. Most common byproducts of PD in the air insulated
switchgears are Ozone and Nitrogen Oxide. When they react with the moisture and water
27
molecules, they form nitric acid [86]. Nitric Acid is very dangerous for most of the dielectrics
and insulators. It plays a major role in the decomposition of the chemical structure of the
insulating material. These gases can be detected by online analysers in transformers or gas
insulated switchgear (GIS) but the chemicals (like nitric acid) are difficult to detect through
online methods in the air insulated switchgears. [87], [88].
3.4.3 Measuring Sensors
A sensor is responsible for picking up the correct signal and hence, it is a backbone in a
measuring system. Its appropriate selection and location can affect the measurement results
significantly. In this section, the sensors which can be used for arc prediction are evaluated.
General specifications of various sensors, measuring above mentioned parameters, are
presented. The basic requirements of sensors for monitoring of switchgear are given below:
• The sensor must be able to sense and record measuring quantities from test object for a
set of defined frequency bands
• It must differentiate between the fault signal and the background noise
• It should be cost effective, small in size and easy to install.
a. Electromagnetic Sensor- UHF/ VHF Antenna

An antenna transforms an EM signal into electric signal (current or voltage). The
operating principle is based on EM induction. When an antenna is linked with the EM field, the
EM waves will pass by it and "sweep" it with their moving EM fields. This will induce a signal
in the antenna, proportional to the energy that the antenna captures. For unconventional PD
detection in the higher frequency range (HF/VHF/UHF), antenna type sensors are widely used
in different shapes [89], [90]. Since there are many practical constraints for sensor installation,
practical antenna design can differ depending on the application. Different types of antennas are
used for different frequency ranges (e.g. loop, bi-conical, Log-periodic and stick antennas).
Characteristics of various types of antenna for GIS applications have been studied and reported
in [91].
Advantages of this sensor include no physical or galvanic connection to the HV

equipment and small size. However, it works only near the PD source; if it is moved away from
the PD source, it captures other EM signals (signals from TV stations and mobile phone
networks; noise) instead of PD. Hence it needs proximity to the insulation surface.
b. Coupling Capacitor (CC)

Coupling capacitors are used to transfer PD energy from PD source to the measurement
setup. Epoxy-mica encapsulated capacitive couplers are the most popular sensors, especially for
transformers and rotating machines [92]-[94]. The main short coming is that the capacitors have
to be designed in order to withstand 60 Hz rated HV level of the equipment, and these should be
manufactured to have low inductance in order to have good high frequency response. These two
considerations are the reason for the relatively high price compared to for example radio
frequency current transformer (RFCT) type detector. On the other hand, the advantage is higher
sensitivity, because they can be placed closer to PD spots. Moreover, the PD activity in each
28
phase can be determined by using CC on each phase. Frequency range of this sensor is 1-500
MHz which is a great advantage. It has high sensitivity, up to 10 pC in many cases and can
detect surface as well as internal PDs.
c. Directional Couplers
The directional coupler is a combination of a capacitive and inductive sensor. It is
possible to use two directional couplers in a cable joint. By doing so, it is possible to distinguish
PD pulses coming from outside or inside the joint. The main application of this type of sensor is
on cable joints. This sensor has a higher sensitivity for cable joints or cable connections, upto 10
pC. Typical operating frequencies are usually from several MHz up to GHz [95].
d. High Frequency Current Transformer (HFCT)

The working principle of HFCT is same as power frequency (50/60 Hz) CT. The
magnetic field around a wire (e.g. ground connection or live wire) caused by the HF current
induces a voltage in the winding of the HFCT through EM induction. This sensor is one of the
most popular inductive sensors in the condition monitoring technologies for all kind of
applications on power system equipment due to its portable, non intrusive characteristic and the
independency of the frequency of the measured signals [96]. Using a ring type ferrite core, the
basic structure of HFCT consists of a certain number of turns of copper wire over the ring core.
Ferrites being ferromagnetic material having high permeability are the most attractive materials
for high frequency applications [84]. HFCTs are usually used to couple the ground connection
wire for discharge transient measurements. It is very sensitive and can detect up-to several
hundred MHz. However, heavy insulation and low inductance makes this sensor expensive
[84]- [85].
e. Rogowski Coil (RC)

Operating principle of RC is similar to the HFCT. It is designed with two wire loops
connected in electrically opposite directions in order to prevent the effect of external noise and
interference. The air cored coil is placed around the conductor, where current pulses produced
by PDs are to be measured. The changing current produces a magnetic field and the rate of
change of current induces voltage in the coil [96]-[98]. Advantages include no physical
connection to the HV equipment and high bandwidth. Further details about the construction of
this sensor are included in the next chapter.
f. Ultrasonic Sensor- Piezoelectric Probe

Piezoelectric sensors rely on the piezoelectric effect which was discovered by Jacques
Curie and Pierre Curie in the late 18th century. When a pressure (piezo is the Greek word for
pressure) is applied to a piezoelectric material, it causes a mechanical deformation and a
displacement of charges. These charges, proportional to the applied pressure, flow through the
connected circuit. These are directional sensors which can give reliable results only if installed
in the proper direction of PD source. The sensor typically operates in the frequency band of
100-130 kHz range, which is the optimum frequency range of PD acoustics in air insulated
switchgear [99]-[100]. It can detect surface PDs effectively, so it can be feasible for air
29
insulated switchgear, however internal PDs are challenging for this sensor. It is immune to the
EMI, however, it is not applicable to industrial substations due to large acoustic noise.
g. Ultraviolet (UV) Sensors
When light particles (photons) hit a certain semiconducting material, electrons are
emitted from the material or its electrical resistance is changed. This is the basic principle of UV
sensors. Such sensors are only sensitive to the lights of having wavelengths in the UV region.
PD phenomenon or low power arcing in the switchgear always emits light in the UV region. UV
sensing is specified to the UV spectrum i.e. 10 to 400 nm [100]-[101]. It is an inexpensive and
sensitive method and can be easily implemented without physical connection to the HV
electrical equipment. However, it is difficult to calibrate for pre-emptive detection.
h. Thermal Sensors
High temperature is quite significant because of bad connections due to rusting and
corrosion, especially at the joints and terminations. There is a wide variety of thermal sensors
available commercially. Their possibility of use in air insulated switchgear is discussed in detail
in [22], [6] and [Publication VI]. The major challenge for various sensors such as
thermocouples is sensitivity. Secondly, too many sensors would be required for the complete
monitoring of the switchgear. Conventionally, IR thermal sensors and thermographs were
widely used to locate the hot spots in electrical equipments. Such IR thermal cameras are very
expensive and not feasible for online condition monitoring. Nowadays there are miniature IR
sensors available commercially, which are small in size and easy to install in the switchgear
permanently. However, they are directional and multiple sensors are required for monitoring of
the whole equipment [102]. In this research, TID have been proposed for thermal monitoring of
switchgear, which can detect thermally created ions inside a switchgear compartment. However,
it can not determine the exact location of hot spot. So, further investigation will be required
using IR thermal cameras [Publication VI].
i. D-dot Sensor
The D-dot sensor is a coaxial sensor made from a standard straight bulkhead SMA jack
[103]. Burkhart [104] explains the basic principle of such probe as, “the normal electric field
generates a surface charge density on the centre conductor of the coaxial connector. At zero
frequency, the center conductor is held at zero potential through the 50 Ω termination. When the
coupled potential varies, the charge varies linearly, but the current path is through the
terminating resistance with the dissipated power representing the power that is coupled down
the coaxial transmission line.”
The output from a D-dot probe is proportional to the derivative of the electric field, the
change in electric field over change in time dE/dt, and is thus recorded by the oscilloscopes as
dV/dt (more accurately, the probe measures the change in electric flux density D, hence the
name D-dot). Consequently, this sensor can detect high frequency electric flux due to PD or low
power arcing. Compared to other sensors discussed in this chapter, it is very compact, has wide
bandwidth and cheap. It doesn’t need physical connection to the HV equipment. However, due
30
to high bandwidth, measurement results can be affected by the background noise [Publication
IV], [Publication X]. More discussion is presented on this sensor in the next chapter, along with
its construction.
3.4.4 Comparison between Different Sensors
Table 3.1 presents a comparison between different sensors discussed in the previous section.
TABLE 3.1 Comparison between different sensor technologies
Sensor type Advantages Disadvantages

RF Sensor x No connection to the HV equipment is required x Works only near the discharge source
x Can be used on-line x Highly sensitive for a wide range of signals
x No coupling device required x Sensitive to reflected signals
x Inexpensive and small in size x Direction sensitive
x Suitable for PD measurement
Coupling x Frequency range is 1-500 MHz (wide range) x Requires high insulation level
Capacitor x Very high sensitivity x High price
x Can detect surface and internal discharges x Each sensor is designed for a certain
x Availability application
Capacitive x No physical connection to the HV part is required x Requires direct line-of-sight
Sensor x No coupling device x Not very sensitive
x Very inexpensive x Sensitive to noise
x Easy to use
x Possibly portably
High Freq. x Robust to external noise x Usually insulation is required to protect
CT x No insulation is required if used around the from the high voltages if used on the live
ground wire. wire
x Can be used on-line x Costly
x Very sensitive
Rogowski x No physical connection to the HV part is required x Shielding makes it costly.
Coil x Very high band width x Less sensitive compared to HFCT due to air
x Can be used on-line core
x Ease of use, non-intrusive
Piezoelectric x Easy and inexpensive x Directional sensor
Ultrasonic x Can detect surface PDs x Sensitive to background sound noise
x Immune to electromagnetic interference. x Electrostatic forces may affect the
Sensor x Steel or fiber rods can be used to propagate the measurements
ultrasonic emissions from the discharge source to x Time delay
the transducer.
x Very sensitive
Ultraviolet x Very inexpensive and sensitive x Difficult to calibrate
Sensor x Easily implemented x External light may create problems
Thermal x Inexpensive, available, easy to implement x Calibration for variable weather conditions
Sensors x Temperature of individual phases can be x Placement often difficult
compared x Wiring unless wireless sensors
x High number of sensors required
Thermal x Reliability x Calibration
Ionization x Less number of sensors required x Not suitable for internal PDs
x Availability x Time delay
Detector x Useful for glowing contacts, arcing and hotspots
D-dot sensor x Very compact/ small in size and cheap x Sensitive to noise due to wide bandwidth
x Wideband spectrum (upto 18 GHz) x Advanced signal processing is required to
x Very sensitive for arcing and PDs de-noise the captured signal.
31
32
Chapter 4
DESIGN AND IMPLEMENTATION OF SENSORS
4.1 Online Condition Assessment System for Electrical Equipment
An online condition assessment system for any electrical equipment is a complete

system which is responsible for health monitoring of the equipment in real time, i.e. during its
normal operation. It must be reliable enough to indicate any abnormal condition in the
equipment and should not give any false alarm when the equipment is completely healthy, in
order to avoid unnecessary shut downs. Fig. 4.1 shows an online condition assessment system
[38], [71].
Initiation of Abnormal
Condition
Sensing System
Data Acquisition
Data Processing and

Analysis
Fault Detection Quantification Fault Location
Fig. 4.1. Online condition assessment system
The essential parts of an online condition monitoring system are explained as follows [71]:
1. Sensor is the backbone of a measurement system. It converts physical quantity to an

electrical signal. Various types of sensors are available commercially for the non-
intrusive monitoring of electrical equipments, as explained in Chaper 3. The selection of
the sensor relies on monitoring method, point of application and failure mechanisms.
Major factors for a good sensor can be its cost, sensitivity, invasiveness, ease of
application and calibration.
2. A data acquisition system is required for analogue to digital conversion, amplification
and pre-processing of the measured signal before sending it to central processor and
control. Amplification and pre-processing helps to reduce the burden on the central
processor and also reduces the risk of losing useful information during communication.
33
3. The main purpose of the data processing and analysis is to find out if there is an
incipient fault which may need alarm and immediate repair. Advanced features
extraction methods, artificial intelligence and algorithms help to identify, quantify and
locate the fault in equipment. On the basis of these final results, decision is taken for the
needed maintenance and its schedules.
4. Data transportation is done through proper communication channel. There are various
communication techniques being used in power systems nowadays. Power line carrier
(PLC), wide area network (WAN), and wireless communication are at the top of list.
Depending on the application and reliability requirement, a suitable communication
channel can be adopted.
4.2 Selection of Sensors for the Prediction of Arc- Flash
Various causes of arc-flash in switchgear have been discussed in Chapter 2. The major
slowly developing mechanisms in MV switchgear are PD in the insulation material and micro-
sparks across the loose connections. Whereas in LV switchgear, major causes are heating due to
poor connection at the cable termination or bus bar joints. Based on an extensive literature
survey done in [Publication I] and also re-written in Chapter 3, various sensors have been
evaluated for the possible application of online condition monitoring of switchgear. However, it
has been found that there is no single sensor which can cope with all these incipient faults, both
in MV and LV switchgear. During the selection of appropriate sensors, various factors have
been considered such as cost effectiveness, ease of use, previous experience and reliability, size
and compactness, ease of availability and manufacture, and the bandwidth and sensitivity.
Consequently, following sensors have been selected for further study during this research:
- Rogowski Coil (for MV application)

- D-dot Sensor (for MV application)
- Thermal Ionization Detector (for LV application)
- High Frequency Current Transformer (commercial HFCT used as a reference
sensor)
In the following sections, first three sensors are explained in detail along with their construction
and application w.r.t switchgear.
4.3 Rogowski Coil
PD current transients generated due to induced charge at defective insulation or low

power arcing current transients originated from loose connections propagate towards both ends
of the power line. As discussed in Chapter 3, various current sensing technologies are available
for such high frequency transient current measurements, including shunts, HFCT, RCs, Hall
effect sensors etc. Taking into account various factors such as costs, physical and operational
aspects, sensitivity, bandwidth, saturation and hysteresis, linearity, operating temperatures;
shunts and RCs are preferred over the others [96]. RCs have an advantage over the shunts due to
their better bandwidth and non-intrusive characteristics, whereas, shunts have to be in galvanic
contact with the live parts which makes it an expensive choice. RC current sensors are being
34
used in power electronics and power systems for measurement of current ranging from few milli
amperes to several thousand amperes with a frequency range from few hertz to several mega
hertz [105]-[107]. For high frequency transient measurements, RC is designed for high
frequencies up to some tens of MHz [108], [109].
4.3.1 Operating Principle and Construction of RC
Rogowski coil works on the principle of Faraday’s law of EM induction, i.e. when a
conducting loop is located in a time varying magnetic field (magnetic flux), an electromotive
force (emf) is induced in the conductor which creates a potential at its terminals. The conductor
carrying current ip(t) to be measured is defined as the primary winding and wire loop with N
number of turns is the secondary winding. Primary current is the source of EM flux. The output
of RC (vrc) is measured as voltage across the terminal of the secondary winding, which is given
by the following relation [Publication V];
di p (t ) (4.1)
vrc (t ) M c
d (t )
where Mc is the mutual induction between the primary conductor and the sensor’s winding. In
order to calculate the primary current, an integration of the above value is required, i.e.
1
i (t )
Mc ³v rc (t )dt (4.2)
Either electronic integrator can be implemented or digital integration is done during the
signal processing and analysis. A detailed discussion about integrators can be found in
[Publication V]. The coil’s winding is constructed on a non-magnetic material, in a toriodal
form, with two wire loops connected electrically in “opposite” directions. Using such a
construction, both terminals of coil winding are available at the same end of the coil. This
makes the coil a flexible open ended design, similar to clamp on CT, ready for fitting around a
power line or current carrying components. Fig. 4.2 shows the construction of an RC.
Fig. 4.2. Construction of Rogowski coil
Unlike the HFCT in which secondary winding is done on a ferromagnetic material, the
non-magnetic material is used as a core in the construction of RC. Due to higher permeability of
magnetic core than air, the sensitivity of HFCT is higher than RC, but due to non-saturable
35
nature of air core, the RC behaves linearly. Hence, higher current levels can be measured by RC
more accurately. In addition, power dissipation and losses are less in RC. Moreover, the return
loop provides cancellation of the effect of external magnetic fields.
The main geometrical parameters of RC are core diameter, internal and external
diameters of the coil, diameters of wire used for winding and number of turns. The internal and
external diameters determine the mean diameter of the coil, which depends on the application of
an RC. Effect of these parameters on the high frequency performance of the RC is studied in
[Publication V]. The physical model of the RC used in the measurements in this research work
and its geometrical parameters are given in Fig. 4.3 and Table 4.1 respectively.
Fig. 4.3. Geometrical model of Rogowski coil
Table 4.1 Geometrical parameters of the designed model of an RC
Model Parameters Symbol Specification

Number of Turns N 30
Outer diameter of coil do 18.25 cm
Inner diameter of coil di 14.1 cm
Core diameter drc 1.9 cm
Mean diameter of coil dm 16.15 cm
Diameter of wire dw 0.085 cm
4.3.2 Electrical Model of RC
Lumped parametric electrical model of RC is given in Fig. 4.4. Every piece of wire has
a resistance and inductance. There is a capacitance between every two loops of wires or between
wire and ground. Rc, Lc and Cc are the structure based inherent resistance, inductance and
capacitance of RC. These parameters direct the electrical behavior of the coil while responding
to the measured signals. The mathematical expressions to calculate these parameters are given
in [Publication V]. Rm is the terminating resistance at the terminals of RC, and Cm is the
capacitance of measurement system which comprises of co-axial cable. All the inductances and
capacitances involved in the whole measurement system determine the measureable frequency
36
range for incipient faults such as PDs and arcing. It is governed by resonant frequency, given in
(4.3), where L and C are the equivalent inductance and equivalent capacitance of the whole
measurement system respectively.
1 (4.3)
f
2S LC
Fig. 4.4. Electrical equivalent model of Rogowski coil
There is wide variety of formulas available in literatures [96], [110] and [111], to
calculate capacitance and inductance of an RC based on their geometry. However, it turns out
that formulas are not very precise when accurate results are needed. A more practical approach
to determine the values of these parameters are given in [97]. The advantage of this
measurement based methodology is that the coil measuring system is exposed to the
measurement environment and more precise coil parameters are found.
4.3.4 Measurement of Incipient Faults using RC
Figure 4.5 shows a transient current measuring system based on RC sensor. The first
element refers to the sensing of primary current by RC head, then a resistance is installed at the
terminals of the RC to damp the oscillations introduced by RC head followed by the
oscilloscope and data acquisition system to transform the obtained analogue signals into digital
samples. Integration of the obtained signal can be done by using numerical integrator during
signal processing and analysis on a computer. Digital integration makes the design of RC
system simple and provides stability to integration process for reconstruction of current
waveform. There is a broad family of numerical algorithms to compute the definite integral.
Method known as trapezoidal rule which in general is considered as having faster convergence
in particular cases of rougher signals is chosen in this work. This is calculated as [112]
1
i( N ) i ( N 1) >Vo ( N 1) Vo ( N )@ N>0 (4.4)
2 fs
where fs is sampling frequency and N is the order number of sample.
Fig. 4.5. Incipient fault detection system using RC
37
4.4 D-dot Sensor
Incipient faults are the sources of high frequency transients such as PD and arcing.
These phenomena are sources of air borne EM emissions in addition to electrical signals
superimposed on the power voltage and currents. The propagation of EM waves depends on the
frequency spectrum of the radiations which are usually in RF region [113], [67]. Hence, a
sensor, measuring such airborne signals, can be a good choice for online monitoring of
switchgear. The sensor should be small enough to be installed inside the panel, sensitive for
internal faults and insensitive to the external RF noise. Compared to other unconventional
sensors discussed in the previous chapter [Publication I], [114], [115], the D-dot sensor is an
inexpensive and robust solution for the measurement of PDs and arcing in switchgear. It is
extremely simple, miniature sized, and can be installed inside the switchgear compartment
easily without any issue. It has wide bandwidth (upto 18 GHz) which makes this sensor
sensitive enough for most major types of discharges inside the switchgear. It is most likely to be
immune to external discharges, if it is installed in a closed metallic compartment, i.e.
switchgear. The D-dot sensor has been used for research purposes, such as measuring
discharges in GIS applications [116], however, it is new idea to use it for air insulated
switchgear applications.
4.4.1 Operating Principle and Construction of D-dot Sensor
The D-dot sensor is a differential electric field probe. It is a co-axial sensor made from a
standard straight bulkhead subminiature version ‘A’ (SMA) connector/ jack. Figure 4.6 (a)
shows the sensor’s head. Outer case works as a ground electrode and the centre conductor
serves as a sensing electrode. The two electrodes are separated by Teflon (PTFE) insulation
material. The whole sensor thus makes a coaxial (cylindrical) assembly. The working principle,
elaborated in Fig. 4.6 (b), is similar to a short monopole antenna. The electric field, produced by
discharge transients, generates a surface charge density on the sensing electrode (centre
conductor) of the co-axial sensor. At zero frequency (direct current), the sensing electrode is
held at zero potential through the terminating resistance of the measurement system. When the
coupled electric field (E) varies, the charge varies linearly. D-dot probe effectively measures the
time derivative of the electric flux density at the surface of the sensing electrode [104]. This is
equivalent to the measurement of the time rate of change of the surface charge density or
displacement current density [117], i.e.
dD (4.5)
JD D
dt
The electric flux density D is related to the electric field intensity E by the expression D= ε0 E ,
where ε0 is the permittivity of free space.
38
(a) (b)
Fig. 4.6. D-dot Sensor, (a)- Sensor head, (b)- Working principle
As the output of the D-dot is the first derivative of the incident electric field, the
integration of this output produces the actual waveform of the electric field. When the length of
the probe approaches a quarter wavelength of the incident wave (l = λ/4), the sensor loses its
derivative properties and becomes a monopole antenna. Thus, the standard SMA connector with
a 4.4 mm center conductor (protruding above the ground plane) has a cut-off frequency of 18
GHz. Decreasing the length of the conductor will increase its bandwidth but at the same time
reduce its sensitivity [118]. The sensor was compiled by mounting a standard straight bulkhead
SMA connector onto a grounded plane, as shown in Fig. 4.7.
Fig. 4.7. D-dot sensor mounted on a grounded plane.
4.4.2 Electrical Model of D-dot Sensor
The output current of D-dot sensor can be represented as, [119]
dV (t ) (4.6)
i (t ) C
dt
39
where C is the cumulative capacitance between the discharge source and the centre conductor of
the sensor head along with the capacitance of the measuring system (coaxial signal transmission
cable) [120]. Electric potential and electric field are related to each other as V(t) = pE(t), where
p is the constant which depends on geometry of the sensor and surrounding, nature of fault and
dielectric medium. Hence, (4.6) can be written as,
dE (t ) (4.7)
i (t ) pC
dt
The electrical equivalent model of the D-dot sensor is shown in Fig. 4.8. The model is
valid for the frequency range of 1 kHZ to 200 MHz, which is the recommended frequency range
for arcing and PD measurements leading to arc-flash. It mainly consists of series and parallel
capacitances [118], [120]. Series capacitance (Ca) comprises of the air capacitance between the
source of electric field and sensing electrode. Parallel capacitance in the circuit (Cs), is the
capacitance of sensor’s head. It is the cumulative capacitance between the sensing electrode and
the grounded metallic casing of the sensor as shown in Fig. 4.6(b). It can be calculated by the
formula for the cylindrical capacitor given by,
Cs 2SH k H o
ls ªr º (4.8)
ln « o »
¬ ri ¼
where ls is the length of the sensor, ri is the radius of sensing electrode, ro is the distance
between the centre of sensing electrode and the metallic case and εk is the relative dielectric
constant of insulation material (PTFE). Both the radii ri and ro are marked in Fig. 4.6(b). The
resistance (Rs) and inductance (Ls) values of the sensor’s head can be negligible.
Terminating
Air Sensor Cable Resistance
Ca Rs Rc
Ls Lc i(t)
E(t) Cs Cc Vo(t)
Fig.4.8. Electrical equivalent model of the D-dot Sensor
A coaxial cable is used to transmit the captured signal to the digitizer (oscilloscope)
where measuring resistance Rm acts as a terminating resistance for the sensor’s output. Rc, Lc and
Cc, shown in the equivalent circuit, are the resistance, inductance and capacitance of the coaxial
cable in the measuring system. Resistance of a coaxial cable can be verified through
measurement. Capacitance can be found by the above formula for a cylindrical object, and is
usually given in the datasheet as well. Capacitance C in (4.6) and (4.7) is the equivalent
capacitance of Ca, Cs and Cc.
40
Inductance per unit length can be given by,
Lc Po P r ª ro º 4.9
ln « »
lc 2S ¬ ri ¼
where lc is the length of the cable, μo is the permeability of air and μr is the relative permeability
of dielectric material of coaxial cable.
4.4.3 Measurement of Incipient Faults Using D-dot Sensor
Fig. 4.9 shows a high frequency discharge transient measurement system based on D-dot
sensor. The sensor mounted on a metal plane can be installed inside the switchgear on the
interior of its wall. There is also a possibility to drill a hole in the wall of the panel and install
the sensor on the holes directly. The metal place can be grounded in order to ensure the sensor’s
outer casing to be grounded properly. A single sensor is sufficient for monitoring a single
compartment, such as cable connection compartment or bus bar section etc. The sensor is then
connected directly to the coaxial cable through an L-bend connector which transmits the
captured signal to the oscilloscope. The power dissipated at the terminating resistance of the
data acquisition device reflects the captured power by the sensor due to discharge event. The
oscilloscope converts the analogue data into digital samples at a high sampling rate (in GHz).
The data is then transferred to the PC for further analysis. Integration of the obtained signal can
be done by using numerical integrator during signal processing to obtain the voltage collapse.
However it is not necessary in the measurements aiming detection only. The recoded data can
be calibrated numerically.
Fig. 4.9. Incipient fault detection system using D-dot Sensor
4.5 Thermal Ionization Detector (TID)
One of the main causes of arc flash in switchgear is bad connection which leads to
excessive heating, hotspots and melting of conductor and ultimately resulting in fire and
explosion. This phenomenon is more prominent at high currents and LV levels. Various sensors
have been evaluated in the literature for the thermal monitoring of switchgear. Two main
sensing candidates in this regard are thermocouples and IR sensors. The major shortcoming
about using thermocouples for thermal monitoring in switchgear, can be its sensitivity. They
need to be installed in contact with the hotspot or in a close proximity of the hotspots which is
practically challenging for live equipment. Secondly, multiple sensors and wiring would be
required to provide monitoring of all the possible hot spots. Therefore the use of thermocouple
to detect incipient faults is impractical [6]. IR thermal imaging cameras are used widely in the
commercial power installations in order to identify over-heated connections before they create a
problem. These sensors measure heat radiations emitted from a hot object in the IR range of the
radiation spectrum.
41
Conventional thermographic cameras are impractical for online monitoring purpose.
However, small eye type (1 inch diameter) sensors can be used for online monitoring [102] but
the implementation of this sensor involves complicated mechanical assembly to help 180º rotary
movement of the sensor in both horizontal and vertical directions due to its directional
measuring feature. Unfortunately, switchgears don’t have enough space to install such a
complicated and reasonably large assembly inside them. It also incurs excessive wiring inside
the switchgear. Such limitations make the use of IR eye sensors unfavorable too [6]. To
overcome these problems, a new sensor which detects thermally created ions in the surrounding,
can be implemented in the switchgear for its online thermal monitoring. It is named as thermal
ionization detector (TID). It is reasonably small in size and can be installed on the ceiling of
cable compartment or bus bar connections. Further details about this sensor are given in the
following sub-sections.
4.5.1 Operating Principle and Construction of TID
Figure 4.10 (a) shows a thermal ionization detector (TID) while its operating principle
has been elaborated with the help of Fig. 4.10 (b). The ionization detector consists of an
ionization chamber also called reference chamber containing radioactive material and a smoke
chamber. The sensing or collector plate having a small hole in the middle divides these two
chambers. The radioactive source holder and the casing of the smoke chamber form the two
electrodes. The radioactive source mounted in the ionization chamber irradiates the air in the
both chambers and ionizes it by producing positive and negative ions. A voltage across the
electrodes produces an electric field. Ions are attracted to the electrodes of opposite polarity to
their own charge. Many recombine but still there is small current flow between the electrodes.
At the junction between the ionization and smoke chambers, the collector converts variations in
chamber current into voltage changes. When particles originated from hotspots enter the
detector via diffusion through holes in the outer cap, they disturb the current flow in the smoke
chamber more so than in the ionization chamber due to attachment of existing ions in the
chambers to these entering particles. This imbalance causes the sensing electrode to change its
potential which is measured and calibrated in terms of temperature [121]. The thermal
ionization detector used in this research work was the one commonly used in the heat detectors
used domestically. The detector was extracted from the thermal detector without making any
modification in its assembly or sensitivity, and is shown in Fig. 4.10(a). The datasheet of the
TID is given in appendix. An electronic amplifier is involved to amplify the measured signal.
(a) (b)
Fig. 4.10. Thermal ionization detector, (a)- Sensor head, (b)- Working principle
42
4.5.2 Diagnosis of Bad Connections Using TID
Fig. 4.11 shows a thermal monitoring system for switchgear application based on TID
sensor. The sensor can be installed on the ceiling of the switchgear compartment or section. It is
justified to install the sensor on the ceiling because thermally created ions move in the upward
direction and tend to accumulate near the ceiling. The sensor is shielded properly in order to
avoid any external noise. A single sensor is sufficient for monitoring a single compartment or
section, such as cable connection compartment or bus bar section etc. The captured signal need
to be amplified immediately before transmitting it to the data acquisition device in order to
avoid external interference. Hence, an electronic amplifier is required to connect immediately
after the sensor. The output of the amplifier is then connected directly to the data acquisition
device using shielded cables such as coaxial cables. The captured signal is DC, and non-variant
during the small time fraction. Hence, the data acquisition system doesn’t need to have high
sampling rate. The whole measuring system is very cheap as compared to the D-dot sensor
measuring system. The data acquisition device converts the analogue data into digital samples.
The data is then transferred to the micro processing or control device for further analysis. The
calibration of the obtained signal can be done by using numerical multipliers during signal
analysis to obtain the approximated temperature rise.
Fig. 4.11. Incipient fault detection system using TID
43
44
Chapter 5
ON-LINE CONDITION ASSESSMENT OF SWITCHGEAR
5.1 De-noising
5.1.1 External Noise During Online Measurements
In contrary to the periodic maintenance where the measurements were done in shielded
laboratories or offline, the on-line measurement faces a large amount of external noise in
measured signals. Extracting useful low-level (amplitude) signals from noisy backgrounds is a
major challenge for on-line condition monitoring. In some situations, the background noise
signals have very high amplitude compared to the PD signal. This directly affects the sensitivity
and reliability of the acquired PD data. Digital signal processing (DSP) techniques must be
applied to on-line PD measurements in order to recognize PD pulses within the EMI
background, which often swamps the PD signal on-site. The identification and separation of PD
signal from noise requires the knowledge of both the PDs and the noise. In general, noise
sources may be divided into the four major categories as shown in the Fig. 5.1. [Publication
VII], [122], [123]:
Fig. 5.1. Shapes of external noises
Discrete spectral interference (DSI) is a narrow-band interference caused by radio

(AM/FM) or TV broadcasts and other mobile communication networks. Periodical pulse shaped
disturbances are due to power electronic equipment. Random pulses are caused by lightning or
switching operations, whereas white noise is EMI caused by the measuring system or
surrounding equipment [122], [123]. DSI has a narrowband spectrum which is centred at certain
dominant frequencies. The DSI can be easily blocked by band stop filtering of such frequencies.
45
Periodical pulses can also be removed by implementing a gating circuit [115]. However, it is
challenging to eliminate the remaining two types of noise because they are more similar to PD
pulses [123]. Switchgear has an additional source of noise resulting from the reflections of PD
signals from the walls of switchgear which disturbs the air borne EM signal produced by the
initial signal.
Figure 5.2 presents an approximate spectrum distribution of various types of noise and
PD signal in the air insulated switchgear [67]. Periodic pulses are usually low frequency noise
(in kHz), whereas radio/TV broadband signals vary from few kHz to MHz. White noises are
usually very high frequency (VHF) or ultra high frequency (UHF) noises. DSI and periodic
pulses have higher amplitude than the white noise. From Fig. 5.2, it can be clearly observed that
PD signal and noise overlap and hence filtering techniques on the basis of frequency cannot be
effectively applied.
Fig. 5.2. Spectra of noise and PD signal in air insulated switchgear
With the advancements in DSP techniques, a variety of methods have been reported for
processing of PD signals such as finite impulse response (FIR) filters, infinite impulse response
(IIR) filters, fast Fourier transform (FFT) analysis, adaptive filtering and wavelet analysis. On
the basis of different advantages reported in [115], [124], Discrete Wavelet Transform (DWT)
technique for de-noising and PD pulse recognition has been implemented in this research. The
usefulness and effectiveness of DWT technique for the sensors have been evaluated and
discussed in the following sections.
5.1.2 Discrete Wavelet Transform (DWT) Based De-noising
a. Discrete Wavelet Transform

A wavelet is a small waveform with limited duration and a zero mean value. It starts
from zero, oscillates in amplitude and decays to zero again. It works similar to Fourier
transform (FT), which breaks up a signal into sine waves of various frequencies. The WT breaks
up a signal into shifted and scaled versions of the original (or mother) wavelet. If a mother
wavelet is defined as ψ(t), it has a property of (5.1), [125]
³ \ (t )dt
R
0 (5.1)
where R is the set of real numbers.
46
The continuous wavelet transform (CWT) of a continuous signal x(t), consisting of a
family of shifted and scaled wavelets associated with mother wavelet ψ(t) can be described as in
(5.2), [124], [125]
1 f (t W )
CWTx\ (W , s)
s
³ f
x(t )\ *
s
dt (5.2)
where τ is shift operator, s is the scaling function, and * stands for complex conjugation. The
value of CWT represents similarity extent between the examined section of x(t) and the scaled
and shifted wavelets. The greater the value of CWT, the more energetic it is, and greater the
similarity between wavelets and the original signal.
The DWT is an effective method of WT for discrete signals. It is faster and does not
generate redundant data unlike CWT. DWT is formulated as [123]- [125]
1 k nbR aRm
DWT\ f (m, k )
aRm
¦ x(n)\ (
n aRm
) (5.3)
The mother wavelet in (5.3) is discretely dilated by the scale parameter aom and
translated using the translation parameter nbοaοm, where aο and bo are fixed values with aο > 1
and bο > 0, while m and n are positive integer values. In case of dyadic transform, which can be
viewed as a special kind of the DWT spectral analyser, aο = 2 and bο = 1 are taken and the DWT
is realized by characterizing the scale variable s in s = 2m and assuming the time variable as τ =
n×2m.
b. De-noising Procedure
The steps for the implementation of DWT are given below:
x A suitable mother wavelet is chosen according to the signal and noise characteristics.
x A level is defined up to which the original signal is decomposed into approximation
and detail coefficients.
x The components corresponding to PD signals, interference, and random noise are
identified at each level based on its frequency spectrum and amplitude levels.
x Threshold level is defined for each level to filter the noise. Generally, noise is denser
with smaller magnitudes.
x Components corresponding to interference and random noise are discarded.
x De-noised signal is reconstructed by adding up all the useful signals after cutting the
low level noises through thresholding in the useful signals.
The analysis and synthesis in multi-resolution signal decomposition (MSD) by DWT is

depicted in Fig. 5.3. In this figure, X[n] is the signal to be de-noised. Signal is split up into high
frequency (detail coefficients) and low frequency components (approximation) through filters
(H0 and G0 respectively), followed by down sampling as depicted. The reconstruction of the
signal based on the selected components gives an interference-free signal. The selection of
“mother wavelet” is an important consideration in the above procedure. The mother wavelet,
47
having the closest resemblance to the original signal, is preferred for de-noising. Among the
available wavelets, the Daubechies wavelet family has almost all of the required properties such
as compactness, limited duration, orthogonality, and asymmetry for analysis of PD pulses [124].
Therefore, Daubechies wavelet family is selected for de-noising in this work.
(a)
(b)
Fig.5.3. (a)-Multi-resolution signal decomposition by DWT, (b)-Reconstruction from decomposition

coefficients
c. De-noising Results
All the signals measured and analysed in this research have been de-noised before
analysis. The measurement setups for various kinds of tests are explained in the next sections. In
this section, only de-noising methodology is demonstrated for the following two signals.
“Daubechies-10” has been selected as mother wavelet for these measurements.
x Signal-1: PD signal from D-dot Sensor

x Signal-2: PD signal from RC
For de-noising of Signal-1 using DWT, the signal is decomposed up to level 8. Figure
5.4(a) shows the detail coefficients (d1-d8) and approximation (a8) of the signal. It can be clearly
observed that detail components d1-d6 are high frequency noise and a8 is the resonance in
measurement system, triggered by PD event. These signals are of very small amplitudes as
compared to the captured signal s. Detail coefficient d7 has comparable amplitude and wave-
shape with s. Either d7 or both d7 and d8 can be combined to reconstruct the de-noised signal. In
Fig. 5.4(b) and (c), d7+d8 are considered as de-noised signal and rest of the components are
discarded. Thresholding techniques can also be used to eliminate low level noise in the selected
detail component (e.g. d7 and d8). In these measurements low level noise in d7 or d8 are
negligible, therefore thresholding has not been considered.
48
The de-noised and original signals are plotted on the same time scale in Fig. 5.4(b) and
5.4(c). The wave-shape of the de-noised signal is quite similar with the captured signals and
does not contain VHF and UHF oscillations as can be seen in original signals. Figure 5.5 shows
the FFT spectrum of original and de-noised signals. It is clear that the de-noised signal does not
contain the low frequency resonance and high frequency noise unlike their presence in the
original signal.
The results presented in Fig. 5.4 and 5.5 confirm that the DWT technique has
successfully de-noised the actual PD signal from interferences, electric and magnetic field
coupling and reflections in the measurement system while maintaining the wave-shape of the
PD signal. In order to get a clear idea about the nature of a fault, it is necessary to perform such
de-noising techniques before analysing the signals and finding fault characteristics.
Decomposition at level 8 : s = a8 + d8 + d7 + d6 + d5 + d4 + d3 + d2 + d1 .
1
s 0
-1
0.1
a8 0
-0.1
0.2
d8 0
-0.2
1
0.5
Amplitude (mV)
d7 0
-0.5
0.1
d6 0
-0.1
0.1
d5 0
-0.1
0.2
d4 0
-0.2
0.2
d3 0
-0.2
0.2
d2 0
-0.2
0.1
d1 0
-0.1
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Number of Sample (n)
(a)
49
(b)
1 Original Signal
Amplitude (mV)
0.5
-0.5
-1
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
1 De-noised Signal
Amplitude (mV)
0.5
-0.5
-1
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
(c)
Fig. 5.4. De-noising of Signal-1, (a) Decomposition and approximation coefficients of signal, (b) Original
and de-noised forms of signal with zoomed pulse, (c) Original and de-noised signal.
250 Original Signal

De-noised Signal
Harmonic Magnitude (mV)
200
150
100
50
0
0 1 2 3
10 10 10 10
Harmonic Frequency (MHz)
Fig. 5.5. FFT comparison between original and de-noised signals for Signal-1
50
Fig. 5.6(a) shows the de-noising process of Signal-2, whereas Fig. 5.6(b) shows the
original and de-noised forms of the signal. Signal-2 has been decomposed up to level 8 and
depending on the wave shape and amplitudes of the decomposed coefficients, d6, d7 and d8 are
considered as useful components and the de-noised signal is reconstructed from them. Rest of
the components are discarded. Moreover, the FFT comparison between original and de-noised
signals of Signal-2 (Fig. 5.7) confirm that the high frequency noise has been eliminated during
de-noising.
5
s -50
-10
1
a8 0
-1
4
d8 20
-2
5
d7 0
-5
Amplitude (mV)
-10
4
2
d6 0
-2
-4
1
d5 0
-1
5
d4 0
-5
2
d3 0
-2
0.5
d2 0
-0.5
0.1
d1 0
-0.1
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Number of Samples (n)
(a)
10
Original Signal
De-noised Signal
5
Amplitude (mV)
-5
-10
0 0.5 1 1.5 2 2.5 3

Time (μs)
(b)
Fig. 5.6. De-noising of Signal-2, (a) Decomposition and approximation coefficients of signal, (b) Original
and de-noised signal.
51
1.4
Original Signal
1.2 De-noised Signal
Harmonic Magnitude (mV) 1
0.8
0.6
0.4
0.2
0 -1 0 1 2 3
10 10 10 10 10
Fig. 5.7. FFT comparison between original and de-noised signals for Signal-2
De-noising of Modelled Signal having typical On-site Disturbances

During the on-site measurements, the major problem is EMI discussed in section 5.1.1,
which often subsumes the low level PD signals picked up by the sensors. Laboratory
measurements have been mixed with simulated disturbances and the resulting artificially
synthesized on-site signal is de-noised to evaluate the performance of DWT for on-site
measurements. This signal is simulated as follows,
x Signal-3: Signal-1 mixed with random noise and 1 MHz sinusoidal noise.
For the processing of Signal-3 using DWT technique, it is decomposed up to level 8

using “Daubechies-10” as a mother wavelet. The de-noised signal is obtained by selecting the
detail components at level 7 and 8 only and the rest are discarded. The actual and the de-noised
signals are shown in Fig. 5.8(b). It is clear that DWT technique has de-noised PD signal mixed
with random and sinusoidal noise of 1 MHz. The pulse shape of the de-noised signal and
Signal-1 closely match (or resemble) each other. This justifies the effectiveness of the technique
for on-site measurements as well, up to certain extent.
In [Publication VII], three sensors i.e. D-dot sensor, loop antenna and RCs are used to
measure the PD signals in the switchgear and DWT based de-noising is implemented in order to
evaluate the performance of the de-noising technique. Whereas in [Publication IV], D-dot
sensor is used to measure the PD signals and Arcing across a 0.2 mm gap and the signals are de-
noised by using DWT technique. Moreover, in this paper, DWT is compared with IIR filters
through the analysis and result comparison.
52
1
s 0
-1
0.5
a8 0
-0.5
0.4
0.2
d8 0
-0.2
1
Amplitude (mV)
0.5
d7 0
-0.5
0.05
d6 0
-0.05
0.1
d5 0.050
-0.05
0.1
d4 0
-0.1
0.2
d3 0
-0.2
0.2
d2 0
-0.2
d1 0.20
-0.2
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Number of Samples (n)
(a)
Simulated Signal
Amplitude (mV)
-1
0 1 2 3 4
Time (μs)
1
Amplitude (mV)
Denoised Signal
0.5
0
-0.5
-1
0 1 2 3 4
Time (μs)
(b)
Fig. 5.8. De-noising of Signal-3 (a) Decomposition and approximation coefficients of signal, (b)
Simulated and de-noised signal.
d. Self-adaptive De-noising
An automated online diagnostic system should be able to fetch data from on-line
sensors, analyze the measured data and provide appropriate information to upper level systems,
such as the protection relays, programmable logic control system, distributed control system
(DCS) or in some cases to the supervisory control and data acquisition (SCADA) system. Based
on the importance of signal processing requirements presented in this section, it seems justified
to have a separate processing unit, collecting information from the sensors, and sending the
processed (de-noised) information to either remote systems (local data analysis and control
53
centre) or central control station (e.g through SCADA). The presence of local control (local data
acquisition and processing) makes the system faster as compared to the central control only.
Moreover it reduces the amount of data transmitted to and from the supervisory control or
central control. The local control mainly focuses on the occurrence of sudden faults, whereas the
data sent to the central control can be used for monitoring and efficient asset management
purpose. A typical flow chart of automated online diagnosis is given in Fig. 5.9.
In the previous section, de-noising of measured signals has been explained, which are
already saved on computer or processing device and the de-noising was done manually by using
Matlab program. This section illustrates the programming flow chart of signal processing unit in
order to achieve adaptive de-noising without human intervention, as shown in Fig. 5.9.
Band Stop
Data
Filter for DSI
Remote Sensors Gating for

Periodic pulses
FFT Refined
Analysis Data
Data
Acquisition
Define No. Amplitude

Dominant
of Decomp. DWT of Signal
Frequency (f)
Local Levels (mV)
Processing Unit
Mother
Wavelet
Inspection of
Local Data Components (f, V)
Analysis &
Decision
Local Control
Centre (Relay, SCADA System Selection of
PLC etc) Components
Reconstruction of
de-noised signal
Fig. 5.9. Flow chart of adaptive de-noising procedure in automated on-line diagnosis
Major deciding factors for de-noising in DWT are: mother wavelet, number of
decomposition levels and inspection of decomposed coefficients on the basis of their amplitude
and frequency spectrum with respect to the amplitude and dominant frequency of the signal to
be de-noised. After the data acquisition, first of all band stop filtering and gating techniques are
implemented to filter the DSI and period pulses respectively, through known techniques [115].
Next, the amplitude of the refined signal is calculated on the basis of peak to peak value or
maximum value in terms of volts or milli-volts. FFT of the signal is calculated in order to find
the dominant frequency component of the signal. A suitable mother wavelet is preset. As stated
previously, ‘Daubechies’ mother wavelet provides satisfactory results [124].
Dominant frequency spectrum in the signal and pre-defined mother wavelet along with
the sampling frequency, determine the number of levels up to which the signal must be
decomposed through DWT. Frequency spectrum of each decomposition level for a certain
mother wavelet at a given sampling rate is defined by the low pass and high pass filters of a
54
mother wavelet. Number of levels is selected automatically in such a way that dominant
frequency of the captured signal is contained within the decomposed coefficients. Once the
signal is decomposed into different decomposition coefficients and approximation, their
amplitudes are compared with the amplitude of the captured signal. The components having
frequency spectra in the range of dominant frequency, and amplitudes comparable with the
captured signals are selected. The rest of the components are then discarded. Finally the de-
noised signal is reconstructed by adding the selected signals.
In order to have better understanding of the above methodology, de-noising of Signal-1

(already de-noised in the previous section) can be explained by this approach. If the signal was
captured online, FFT could indicate that the dominant frequency is at 10 MHz. For a signal
having 2 GS/s sampling rate, by selecting Daubechies-10 as mother wavelet, 10 MHz signal is
expected at decomposition level 7. Hence minimum of 7 levels are selected. After that,
frequency spectrum of each component is compared with the dominant frequency. Also the peak
to peak values are compared with the peak to peak values of the captured signal. During this
analysis, decomposition coefficient at level 7 and 8 (i.e. d7 and d8) proved to be the useful
components to reconstruct the de-noised signal.
5.2 Laboratory Measured Signals by the Sensors
Measurement Setup-1: PD Measurement:

The measuring setup used in the laboratory measurements is shown in the Fig. 5.10. The
figure shows the MV switchgear, whose one phase is powered by MV source. The circuit
breaker was put in the closed position and the outgoing side of powered phase was open
circuited while the other two phases were grounded. The energized phase was used to study
various PD conditions. The switchgear was placed on a wooden base and its case was grounded
through a single point. The D-dot sensor was fixed inside the upper portion of the switchgear
compartment whereas the RC and HFCT were installed across the ground connection.
HV Transformer
(230V/100kV) Switchgear
Compartment
Coup. Cap.
Capacitor (989 pF)
Discharge
Regulating (100pF) Location
Transformer
(0-230V)
Coup. Device CD D-dot Sensor
Voltage (Pulse Trans.)
Measurement
AC Source V Capacitor
(230V) (100nF) HFCT
Rogowski Coil
Digital Oscilloscopes
Ch1 Ch2 Ch3 Ch4 Ch1 Ch2 Ch3 Ch4
Coaxial Cables
Fig. 5.10. Laboratory measurement setup for PD measurement
55
For the conventional measurement setup, the coupling capacitor and coupling device
(pulse transformer) were installed after capacitive divider (before the switchgear). The coupling
capacitor transfers the discharge (PD) energy from the electrical network to the measuring
system by means of the capacitance between circuit nodes. The coupling device is usually an
active or passive four terminal network (quadrupole) and converts the input PD current pulse to
output voltage signals to be displayed on the digitiser (oscilloscope). Apparent charge is
calculated either by calibrating the peak value of the pulse or by calibrating the integral (area
under the curve) of the PD pulse. Two oscilloscopes were coupled together to measure 5
different signals as shown in Fig. 5.10.
Fig. 5.11 shows the switchgear and measuring equipment. PD pulses were measured by
conventional method (pulse transformer), HFCT, RC and D-dot sensor at certain voltage levels.
Measurements were carried out under the following PD conditions, also depicted in Fig. 5.12:
a) PD between pin to plate (corona).
b) PD in void (epoxy-resin solid insulation with air bubble).
c) Surface discharge at the insulator surface.
d) Corona due to multiple bare conductors.
Fig. 5.11. (left) - Switchgear compartment, (right) - AC supply and measuring equipment.
Fig. 5.12. Discharge sources: (a) PD between pin to plate (corona), (b) PD in void, (c) Surface discharge
at the insulator surface, (d) Corona due to multiple bare conductors.
56
Measurement Setup-2: Arcing Measurement
The behaviour of arcing across loose contacts is similar to the low energy arcs (sparks)
across a small arc gap. Both of them cause radio frequency electromagnetic emissions. In order
to study the response of various sensors under the arcing across loose contacts, a very small arc
gap (rod-to-sphere) of 0.2 mm was implemented. System was energized by an LV regulating
transformer and a 0.23/100 kV transformer. Figure 13 shows the circuit diagram and Fig. 14
represents the physical placement of different sensors (about 1 meter away from the arc source).
This measurement setup also includes capacitive divider for voltage measurement. Sensors were
connected to the digital oscilloscopes through 50 Ω coaxial cable to 50 Ω channel input of the
oscilloscope. Data was captured at 20 GHz sampling frequency, using a 16-bit digital
oscilloscope.
HV Transformer
(230V/100kV)
D-dot Sensor
Capacitor
(100pF) Arcing Source
Regulating
Transformer
(0-230V) Loop Antenna
Voltage
Measurement
AC Source V Capacitor
(230V) (100nF) HFCT
Rogowski Coil
Oscilloscope
Ch1 Ch2 Ch3Ch4
Coaxial Cables
Fig. 5.13. Laboratory measurement setup for arcing measurement
Fig. 5.14. Physical placement of sensors for arcing measurement in laboratory
57
5.2.1 Calibration and Performance of the Sensor
According to IEC60270 standard [68], the calibration of sensors is done in such a way
that a known short duration current pulse (similar to PD pulse) is injected into a test object and
output amplitude of sensor is scaled accordingly. Such calibration is valid for induction sensors
(HFCT, RC) or sensors in contact with HV part e.g. coupling capacitors (CC). It is possible to
calibrate D-dot sensor for certain conditions only due to the variable nature of air capacitance
(Ca). In our laboratory measurements, the distance between the PD source and the sensor was
fixed. D-dot sensor was calibrated w.r.t a pre-calibrated HFCT. The calibration scales can be
found in Table 5.1. In Fig. 5.15 and Table 5.1, the sensitivity of D-dot sensor has been
compared with HFCT for PD values of 52 pC and 100 pC. It is well known that the attenuation
of electric field through air is very high as compared to the conductors. As the D-dot sensor
measures air-borne differential electric field, the output of the sensor has lesser amplitude as
compared to HFCT. It has been reported in [115] that desired sensitivity range for PD
measurements in air insulated switchgear is 10 to 100 pC. D-dot sensor is capable of detecting
small level of discharges up to around 50 pC. However, the sensitivity of the sensor can further
be enhanced by extending the length of centre conductor outward. Moreover due to wide
bandwidth of D-dot sensor, it is more likely that it may capture other high frequency EM
disturbances. Such disturbance is prominent at lower PD levels (e.g. at 50 pC).
(a) (b)
Fig. 5.15. Output of D-dot Sensor (a) Signals at 52 pC, (b) Signals at 100 pC
TABLE 5.1
Output amplitude comparison between D-dot sensor and HFCT
52 pC 100 pC
Sensors VPP VP VPP VP
(mV) (mV) (mV) (mV)
HFCT 15.54 9.51 32.77 19.70
D-dot
3.01 1.47 6.34 3.17
Sensor
Pulse
- 24.3131 - 51.05
Transformer
*VPP stands for peak to peak values of output signal whereas VP is maximum value on either side.
58
5.2.2 Time Domain Comparison
Signals captured by the three sensor, i.e. D-dot Sensor, HFCT and RC in case of PDs in
void in an insulator are plotted in Fig. 5.16 Since the HFCT and RC were installed around the
ground connection, the signals captured by them have higher amplitude. The HFCT is the most
sensitive in this case followed by the RC, and the D-dot sensor respectively. The HFCT and RC
measure change in discharge current (di/dt) whereas the D-dot sensor measures the changing
electric field. Since all the sensors have a different bandwidth and a different sensitivity, a direct
proportionality between their measurements cannot be determined. Due to the difference in the
measured parameter and bandwidth, the measured signals slightly differ from each other.
However it is clear that, all the sensors responded to the same discharge. This proves the
reliability and usability of the sensors used. Sensors are considered good even though they
capture the signals with lower amplitude with a condition that the amplitude of the noise must
also be relatively smaller as is the case with the D-dot sensor.
Outputs of different sensors under surface discharge condition and arcing have been
plotted in Fig. 5.17 and Fig. 5.18 respectively. In the case of PD over the insulator surface, the
sensors behaved similar to PD in voids, but there is clearly different output of sensors in case of
arcing. Arcing causes high energy impulse in the system; hence the output of the sensors is
much higher. The HFCT is designed to measure the low frequency as well as high frequency
current pulses, this is the reason less oscillations are observed in HFCT signal due to the
dominancy of low frequency current pulse.
0.4
Amplitude (V)
0.2 HFCT
0
-0.2
-0.4
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
Amplitude (V)
0.05
Rogowski Coil
0
-0.05
-0.1
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
Amplitude (V)
0.01 D-dot Sensor
-0.01
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
Fig. 5.16. PD in voids in an insulator
59
-3
x 10
Amplitude (V) 5 HFCT
-5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

-3
x 10 Time (μs)
4 Rogowski Coil
Amplitude (V)
2
0
-2
-4
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
0.0145 D-dot Sensor

Amplitude (V)
0.014
0.0135
0.013
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
Fig. 5.17. PD over the insulator surface
Fig. 5.18. Arcing at 0.2mm arc gap (tod- Sphere)
5.2.3 Frequency Domain Comparison
The FFTs of the de-noised signals from each sensor under different types of discharges
were plotted in order to study the frequency response of the three sensors. Table 2 shows the
frequency ranges of the signals captured under each discharge condition. Due to the difference
in design parameters, measured variables and difference in sensitivity, the frequency spectrum
of various sensors slightly differ, e.g. the FFT or arcing signals captured by the HFCT and RC
have lower frequencies as compared to loop antenna and D-dot sensor due to their current
measuring capability, and hence the effect of lower frequency arcing pulse was dominating.
Frequencies of the surface discharge are the highest and the frequencies of the arc
signals are the lowest, even some are in the range of kHz. Generally it is understood and
believed that the frequency spectrum of PD is higher than the arcing due to their fast rise time,
60
so is observed in Table 2. It can be concluded that all the sensors are consistent and confirm the
reliability of each other to detect certain faults.
TABLE 5.2
Frequencies of signals captured by different sensors under various types of discharges
Type of Discharge HFCT RC D-dot
PDs in Epoxy Insulator 3-20 3-25 3-20

MHz MHz MHz
Surface Discharge 5-35 5-40 5-40
Over the Insulator MHz MHz MHz
Surface
Arcing at 0.2 mm Gap 1 kHz- 1 kHz- 1 kHz-
Distance 10 MHz 13 MHz 27 MHz
5.3 On-site/ Practical Measured Signals
The D-dot sensor and RC have already proved their performance and applicability for
air insulated switchgear in the laboratory, as reported in detail in [Publication I] and
[Publication III]. However, due to the wide bandwidth of the D-dot sensor, it is prone to other
disturbances such as radio and TV broadcast signals, power electronic noise and other EMIs. It
was still an open question, ‘if the sensor is capable of measuring discharge transients reliably
in the presence of huge amount of noise on site’. Hence, the major objective of on-site
measurements was to test the performance of the D-dot sensor under the real operating
conditions where the external noise is significant, unlike in the laboratory environment.
Furthermore, on-site measurements enable the user to capture and estimate the various noise
levels which can be used to make a better design of the sensor and to improve the signal de-
noising methodology. D-dot sensors and HFCT are used for on-site measurements. In these
measurements a clamp on HFCT is preferred over RC due to its ease of applicability under
practical situation. The on-site measurements were done in two phases as explained below.
During the first phase, the disturbances including any PD activity in the switchgear
system were measured without creating any artificial PD source in the switchgear. The
measurement setup is shown in Fig. 5.19 and Fig. 5.20. The measurements were done at three
different MV substations in Helsinki; one feeding an industrial load (a transformer
manufacturing factory using various types of electrical machines), the other two feeding
domestic loads. Two D-dot sensors were installed inside the compartment on the wall of the
switchgear as shown in Fig. 5.19, however results from one sensor have been reported only due
to their matching results. The HFCT was installed around the grounding screens/ strip of the
outgoing cable in the compartment. As the discharge transients emit high frequency signals, the
signals captured by the sensors were recorded by a digital storage oscilloscope (DSO) at high
sampling rate (few hundred MHz). Hundreds of samples were collected from each site.
61
During the second phase of measurements, an artificially created PD source was
installed in the switchgear as shown in Fig. 5.21. The PD source was created by wrapping
stranded bare conductors and insulated wires on bus bars. This switchgear was not feeding any
load and it was selected for PD creation due to safety reasons. The location of the sensors was
same as in the first phase (Fig. 5.19 and 5.20). There are certain challenges and risks involved in
creation of artificial PD signals in on-line switchgear. A PD source may cause flashover during
the nominal voltage and may trip the upstream circuit breakers, if it is not selected properly.
Moreover PD sources can not be installed too close to the grounded parts (metallic walls), in
order to avoid immediate flashover.
Bus Bars
Switchgear
Enclosure
i Ph-1 i Ph-2 i Ph-3
Circuit
Breaker
D-dot
Sensor
HFCT
Screen
Grounding
Wire
Cable
Cable
Conductor
Screen
(Copper)
Ch1 Ch2 Ch3 Ch4
Oscilloscope
Fig. 5.19. On-line PD Measurement Setup in MV Switchgear
Fig. 5.20. Placement of sensors in a switchgear compartment on site
62
Fig. 5.21. Creation of artificial PD source on a bus bar
5.3.1 On-site Measured Noise
The HFCT and D-dot sensor could measure various types of noise during the on-site
testing. Some of the measured signals are given in appendix (Fig. 5.22 to Fig. 5.23). Most of
these signals were recorded in a switchgear feeding an industrial load (transformer
manufacturing factory). Fig. 5.22 (left and right) represents the noise, possibly from some
electronic equipment which turns on for 15-20 μs to switch another equipment ON or OFF.
Fig.5.23 (left) represents typical power electronic repetitive noise. The repetitive pulses can be
from switching devices such as thyristor. Fig5.23 (right) shows the dominating white noise
which may be due to some machinery causing huge noise or due to some high frequency
broadband coupling.
Amplitude (mV)
HFCT Signal
Amplitude (mV)
4 HFCT Signal 5
2
0 0
-2 -5
-4
0 10 20 30 40 50 60 70 80 90 -10 0 10 20 30 40 50 60 70 80 90
Time (μs) Time (μs)
0.5 D-dot Signal
Amplitude (mV)
Amplitude (mV)
D-dot Signal
0.5
0
0
-0.5
-0.5
-10 0 10 20 30 40 50 60 70 80 90 -10 0 10 20 30 40 50 60 70 80 90
Fig. 5.22. (left figure)-Noise signal-A, (Peak to peak noise level: HFCT= 9.16 mV, D-dot = 1.20 mV);
(right figure)-Noise signal-B, (Peak to peak noise level: HFCT= 12.07 mV, D-dot = 1.02 mV)
HFCT Signal
Amplitude (mV)
HFCT Signal
Amplitude (mV)
5 5
0
0
-5
-5
-10
-10 0 10 20 30 40 50 60 70 80 90 -10 0 10 20 30 40 50 60 70 80 90
D-dot Signal
Amplitude (mV)
Amplitude (mV)
D-dot Signal
0.5 0.5
0 0
-0.5 -0.5
0 10 20 30 40 50 60 70 80 90 -10 0 10 20 30 40 50 60 70 80 90
Fig. 5.23. (left figure)-Noise signal-C, (Peak to peak noise level: HFCT= 11.62 mV, D-dot = 1.20 mV);
(right figure)-Noise signal-D, (Peak to peak noise level: HFCT= 11.45 mV, D-dot = 1.60 mV)
63
Few noise signals have been presented in Fig. 5.24 on a time scale of 5 μs. The pulse
shapes of the noise are not clear. The amplitude levels of the measured noise signals by the D-
dot sensor vary from 0.2 mV to 1.8 mV. It will be hard to detect and identify any PD signal
having an amplitude lower than this value. However, it can be noted from the calibration data in
section 5.2.1 that 3.01 mV represents a PD level of 50 pC which is negligible.
Amplitude (mV)
HFCT Signal
Amplitude (mV)
HFCT Signal 5
5
0
0
-5
-5
-10
0 1 2 3 4 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
D-dot Signal
Amplitude (mV)
Amplitude (mV)
1 D-dot Signal
0.5
0
0
-0.5
-1 -1
0 1 2 3 4 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Fig. 5.24. (left figure)- Noise signal-1, (Peak to peak noise level: HFCT= 9.52 mV, D-dot = 1.80 mV);
(right figure)- Noise signal-2, (Peak to peak noise level: HFCT= 11.53 mV, D-dot = 2.01 mV)
5.3.2 On-site PD Signals
Figure 5.25 shows the PD signals measured by the two sensors during the second phase
of the measurements. As compared to the measured noise shown in the previous section, there
are certain differences with respect to the noise. The pulse shape is clearer than the noise
signals. Secondly, the amplitudes are significantly higher than the noise levels. In Fig. 5.25(b),
the two signals measured by HFCT and D-dot sensor are plotted on the same axis. The
amplitudes of the two signals are scaled w.r.t each other, in order to make a clear comparison.
The figure shows that the two sensors have picked exactly the same signals. Similar results can
be observed for the PD signal shown in Fig. 5.26. The amplitude levels of the measured PD
signals by the D-dot sensor, during the second phase of the measurements, were from 3.6 mV to
15 mV and are plotted in a stem graph in Fig. 5.27. As per the laboratory calibration, 3.6 mV
refers to 60 pC and 15 mV refers to 250 pC.
HFCT Signal
Amplitude (mV)
10
-10
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4

Time (μs)
5
Amplitude (mV)
D-dot Signal
-5
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
(a)
64
(b)
Fig. 5.25(a) PD signal-1 (Peak to peak PD level: HFCT= 30.65 mV), D-dot = 5.28 mV (b) HFCT and D-
dot signals plotted on the same axis for pulse shape comparison
HFCT Signal
Amplitude (mV)
15
10
5
0
-5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time (μs)
4 D-dot Signal
Amplitude (mV)
2
0
-2
-4
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
Fig. 5.26. PD fault (Pk-Pk PD level: D-dot = 4.30 mV, HFCT= 23.20 mV)
individual PD pulses (mV)
14
Peak to peak vales of
12
10
0
20 40 60 80 100 120 140 160 180 200
No. of PD pulses
Fig. 5.27. Measured PD levels by D-dot sensor
5.3.3 Comparison Between On-site noise and On-site PDs
There are various factors which differentiate whether the on-site measured signal is noise or a
PD signal.
x The noise does not have a clean pulse shape whereas the PD signals have significantly
clear pulse shape as shown in Fig. 5.22 to Fig. 5.26.
65
x Noise levels are much smaller than the PD signal levels. The range of peak to peak
noise levels and PD levels are shown in Table 5.3. The maximum noise level captured
by D-dot sensor was 1.8 mV peak to peak whereas the minimum PD level was 3.6 mV
(representing approximately 60 pC).
x Through the comparison of FFT spectra of noise and PD signals, given in Fig. 5.28, it is
observed that the dominant frequency component, named as central band is about 3-4
times larger than the side band frequencies in a PD signal. Side bands represent the
noise content in the signal whereas the central band represents the useful PD signal. The
measured noise doesn’t have significant central band.
TABLE 5.3
Peak to peak levels of noise and PD signals measured by a D-dot sensor
Minimum level Maximum level
Noise Signals 0.2 mV 1.8 mV
PD Signals 3.6 mV 15 mV
Harmonic Magnitude(mV)
Central Band
0.15
Side Band Side Band 0.1
0.05
-1 0 1 2
10 10 10 10
Fig. 5.28 (left figure)-FFT of a typical PD signal (measured by D-dot sensor); (right figure)-FFT of a
typical noise signal (measured by D-dot sensor)
5.4 Detection of Multiple Faults in a Switchgear
5.4.1 Calculation of Apparent Charge and Cumulative Energy
Apparent charge of the PD pulse can not be measured using unconventional methods
alone but according to IEC60270 standard, it can be calculated by integrating the measured
current pulse (recorded by the conventional method, explained below) up to the first zero
crossing of the PD pulse measured by the coupling device (CD) and multiplying the calibration
factor with the result. Since the capacitance of a test object affects the circuit characteristics, the
standard specifies that re-calibration must be performed for each new test object.
The measurement setup discussed in section 5.2 is used for these measurements.
Applied voltage was varied between 3 kV-22 kV to get a variety of measurements having
different discharge levels. Fig. 5.29 shows a (de-noised) signal measured by the conventional
(pulse transformer) and unconventional method (D-dot) under fault condition (b) (PD in void in
66
Fig. 5.12). PD current pulses and radiated RF signals are captured simultaneously with the
oscilloscope, triggering on the conventional current pulse so that PD events that did not excite
RF signals would still be recorded by the conventional method. Amplitudes in this figure are not
to scale because the amplitude of D-dot sensor has been amplified numerically during the signal
processing, in order to get clear comparison.
D-dot Sensor Signal (De-noised)

Conventional Measurement (Pulse Transformer Signal)
20
Amplitude (mV) 15
10
-5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (μs)
Fig. 5.29. Signal captured by D-dot sensor and pulse transformer.
Peak value (VP) and peak-to-peak value (VPP) of the signal voltage give some idea about
the intensity of PD but the exact relationship between these values to the apparent charge is still
unknown. Another parameter, the cumulative energy of the measured signal, can provide more
insight concerning the extent of PD. Cumulative energy could be considered a more reliable
parameter than the peak-to-peak value because it takes the whole PD signal into consideration,
not only peak values. For this reason, cumulative energy is correlated to apparent charge in this
section. Following paragraph explains the calculation of cumulative energy of the measured
signal by D-dot sensor.
Voltage is induced in the sensor when a changing electric or magnetic field sweeps the
sensor. This induced voltage produces a current flow in the closed circuit. The terminating
resistance at the oscilloscope data acquisition port is a part of this circuit and the voltage drop
across the input resistor of oscilloscope is measured, displayed and recorded by the
oscilloscope. Power across the input resistor of the oscilloscope can be calculated as follows,
Vi 2 (5.4)
Pi
R
't 2 (5.5)
Wi Vi
R
where Δt is the sample duration, R is input resistance of oscilloscope and Vi is the sample
voltage. Equation (5.5) gives the energy (Wi) of one sample. If all the sample’s energies are
summed, as in equation (5.6), the total energy of the captured signal is obtained. N is the
number of samples in (5.6),
't N 2
WTot ¦Vi
R i1
(5.6)
67
5.4.2 Correlation Between Apparent Charge and Cumulative Energy
The correlation between the apparent charge and cumulative energy has been plotted in
Fig. 5.30. A distinct parabolic trend can be observed in most of the cases. This parabolic
behaviour shows that the rate of change of cumulative energy is faster than the change in
apparent charge. It is clear from Fig. 5.30 that the exact relationship between these two
quantities is different for each type of discharge. Thus, the positions of clusters vary depending
on the type and location of discharge (assuming that the position of the sensor is fixed).
Relationships between charge and cumulative energy can be determined by taking thousands of
such measurements at varying voltage levels and plotting their scatter. Through curve fitting,
the exact relationship between these two quantities for each type of fault can be found. These
relationships, together with further research investigating a wide range of other variable PD
sources, could be helpful in identifying the type and nature of faults.
Three different clusters are visible in Fig. 5.30(d) which implies that there are different
forms of discharges in this test object. Fig. 5.12(d) shows that PD can occur at different ends of
the conductors. Hence, it also identifies the presence of multiple discharges. The correlation
plots for all defects tested under the selected conditions (Fig. 5.12) are shown on the same scale
in Fig. 5.31. It is found that the order-of-magnitude relationship between the energy and
apparent charge levels varies significantly among various PD sources. Due to the large variation
in the cumulative energies in the same charge range, the PD sources can be clearly distinguished
and thereby identified.
Polynomial relations show that there is almost a linear relationship between the square
of charge and cumulative energy. However, the slope of the lines may vary from case to case. In
other words, the square root of the calculated cumulative energy is directly proportional to the
apparent charge dissipation, i.e. √WC ‫ ן‬q. From the calculation of cumulative energy, the extent
of charge dissipation in the test object can be estimated.
Energy of a single PD pulse (Wi) can be determined on the basis of an apparent charge
value measured for an object as follows [25],
Wi qiVs qiVm SinMi (5.7)
where qi is apparent charge of a single discharge, Vs is voltage value over object at the moment
of discharge inception, Vm is crest value (amplitude) of applied ac voltage and M i is the voltage
phase of discharge pulse. Once apparent charge is estimated, the actual energy dissipated due to
discharge can be calculated by equation (5.7). The relationship for energy dissipated and time is
given as (5.8), where P is the average power at which energy is being dissipated.
W Put (5.8)
68
(a)
(b)
(c)
(d)
Fig.5.30. Apparent charge correlated to cumulative energy: (a) PD between pin to plate (corona), (b) PD
in void, (c) Surface discharge, (d) Corona due to multiple bare conductors.
69
Fig. 5.31. Correlation clusters between apparent charge and cumulative energies for the investigated PD
sources.
5.5 Thermal Monitoring of Switchgear using TID Sensor
In order to study the performance of the TID, two types of measurements have been
carried out in this work. A layout diagram of the measurement setup used in the laboratory is
shown in Fig. 5.32. Bus-bars, having three bolted joints were installed in a closed metallic
chamber in order to replicate the practical switchgear having three bus bar joints (bus bar to
cable connection). Three thermocouples were installed at each joint, as shown. One
thermocouple was installed at the ceiling of the compartment. TID was also installed at the
ceiling of the chamber, next to thermocouple. It is justified to install the TID sensor on the
ceiling because the particles originated from the hotspots move upward towards the ceiling with
hot air. TC stands for thermocouple whereas abbreviations Rj1 and Tj1 are used for resistance
and temperatures for joint-1. Variables for joint-2 and joint-3 can also be seen in Fig. 5.32. Fig.
5.33 shows the placement of the bus bars and sensors physically.
The bus bars were fed high current (about 150 A) at 36 V, through step down heating
transformer (maximum current of 2000 A and minimum voltage of 2.74 V). The input current
was continuously measured by a current transformer (CT). Joints’ resistances were measured by
using micrometer. Measured values were 10 μΩ for the installed bus bar joints shown in Fig.
5.32. Temperatures of the joints were continuously monitored by thermocouples installed at
each joint. The time variant output of the sensor was amplified by a dedicated amplifier and
then fed to a digitizer (computer), along with the time variant outputs of thermocouples.
Following parameters were varied during the measurement.
- The temperature of the compartment (ambient temperature)
- Input current and consequent variation in temperature of joints
- Cooling of joint temperature by opening the door of compartment
70
TID (Sensor)
Metallic
TC
cabinet (Tc)
T0
2000A/ 2.74 V
To TC TC TC CT
Digitiser
Heating
(Rj1, Tj1) (Rj2, Tj2) (Rj3, Tj3)
Transformer
iin
Fig. 5.32 Layout diagram of test setup
Fig. 5.33. Test setup for thermal monitoring
Measurements were done in two phases. During the first measurement phase, the
compartment was kept at room temperature and the power supply was switched on. The
variation in joint temperature w.r.t time was measured. Also the compartment/ambinet
temperature and TID output was recorded. TID actually measured potential (in Volts) but it was
calibrated to read temperature for comparison purpose. In the second measurement, the
compartment was first heated up to a certain temperature and then bus bar supply was switched
ON. And then after a certain time elapse, the power was turned OFF. In order to record fast
cooling phenomenon, the doors were opened after a certain time of turning the power OFF. The
delay in picking up the signals by TID was also recorded in some measurements.
5.5.1 Compartment at Room Temperature (Case 1)
Fig. 5.34 shows the measured results for the first case when the compartment was at
room temperature. It can be observed that the output of all the sensors reads approximately 22ºC
which was the temperature of the compartment (ambient temperature). After duration of t=4
minutes, the power supply to the bus bars was turned ON and the temperature can be seen to be
rising immediately. Temperatures of bus bar joints initially rise exponentially, when the bus bar
joints just started to heat up. Somehow similar exponential behavior is observed from the output
of the TID. However the thermocouple installed at the ceiling of the compartment, next to the
71
TID, behaved almost linear.
After t = 100 minutes, the power supply to the bus bars is switched off and the bus
bars stated to cool down as shown from the output of the thermocouples measuring joint
temperatures. The output of the TID also behaved exponentially decaying, similar to the joint
temperatures. However, the compartment temperature (ambient temperature) acts linear.
At t = 107 minutes, the door of the compartment was opened and hence the
compartment/ ambient temperature fall down immediately, however the joint temperature and
TID output behaved exponential. As the TID measures thermal particles in the compartment and
the amount/ intensity of ions decreased due to opening of the door, the output of TID fell faster
than the joint temperatures.
Fig. 5.35 shows a better comparison between TID output, joint temperature and
compartment temperature (ambient). The rate of rise of TID output is slower than the rate of rise
of actual joint temperature but rate of fall is almost similar to the joint temperature. This is
justified because the TID measures the thermal ions and it is possible that some of the thermal
ions are dispersed. Also, it takes a bit of time to reach the ions to the TID.
Fig. 5.34. Measurements while compartment at room temperature
Fig. 5.35. Comparison between TID, Compartment temperature (ambient temperature) and Joint Temp
while Compartment at room temperature
72
5.5.2 Compartment at an Elevated Temperature of 47 °C (Case 2)
Fig. 5.36 shows the measured results for the second case when the compartment is first
heated up to 47 ºC before switching on the power supply to the bus bars. After the bus bars are
energized at a current of 100 A, at t = 5 minutes, the joint temperature rose exponentially and
TID also responded to the change in bus bar joint temperature. However the thermocouple
installed at the ceiling of the compartment didn’t respond to the change significantly. At t = 43
minutes, the input supply current is increased from 100 A to 150 A. It can be observed clearly
that TID output responded to the exponential increase of the joint temperature whereas the
thermocouple measuring the compartment temperature, didn’t respond it properly. At t = 66
minutes, the power supply is turned off and at t = 82 minutes, the cabinet is turned off and door
is opened too. Fig. 5.36 shows that TID behaved closely as compared to the joint temperature
whereas compartment temperature (ambient temperature) is totally different. To study this
difference more closely, TID output and ambient/compartment temperatures are plotted in Fig.
5.37.
Fig. 5.36. Measurements while compartment at T= 47 ºC
Fig. 5.37. TID and Compartment temperature while compartment at T= 47 ºC
73
5.5.3 Time Delay in TID Signal
Fig. 5.38 shows the time delay between the response of TID and thermocouple
measuring joint temperature. It can be clearly observed from the figure that TID starts to react
after a delay of 1.30 minutes, after switching the power supply ON. It can be justified because it
takes some time to the thermally created particles to reach the surface of ionization detector.
Fig. 5.38. Time delay in the output of TID
5.5.4 Challenges with Thermal Monitoring using TID
There are certain observations which cannot be ignored in this study. The major challenges
about this sensor are:
- Calibration
- Time delay in the output of TID
- Sensitivity
- Detection but not fault location
The precise calibration of the TID is a challenging task. The output of the TID sensor can be
influenced by many factors including ambient temperature, amplification factor of the signal,
moisture, smoke, distance from the hotspot etc. This is the reason why the TID is not matching
the actual hotspot temperature measured by thermocouple. Hence, calibration is still an open
question at this stage, which can probably be answered in the following research. Second
observation is the time delay between the actual hotspot temperature change and variation in
output signal of TID. The delay is approximated to be 1.3 minutes in the presented case. But
this can be more if the distance between the hotspot and the sensor is larger. Effect on the time
delay considering different location of hotspots (relationship between the distance and time
delay) is a very good idea to investigate, which will be considered in the next work by the
author. However, considering two main aspects of a monitoring system i.e., detection and
measurement, this research focuses on the detection of incipient faults while measurement
aspect has not been considered in this work due to many constraints discussed in this section.
Once the fault is detected, the efforts can be made to measure the extent of the fault more
74
precisely with some particular measurement solution discussed in the next sub-section. In spite
of this time delay, this sensor can be used for online monitoring because incipient faults take
hours or even days to get severe. Authors aim to focus on the design aspects of sensor,
parameter identification, modeling, and hence to calibrate its response in comparison with a
standard sensor in a future research and following research article.
However, TID presents a good approximation for incipient fault detection such as insulation
burn, joint excessive heating or arcing across the loose connection. Its performance can further
be increased by a better and customized design having higher sensitivity. The sensitivity can be
improved in various ways. The easiest way can be by decreasing the distance between the
positive and negative electrodes. This will strengthen the electric field and hence more ions will
be attracted from outside of the sensor through the openings of smoke chamber. This can be
done by decreasing the size of the smoke chamber. Also increasing the applied potential
between the ionization chamber and smoke chamber will improve the sensitivity of the sensor.
This sensor head doesn’t respond to momentary transient due to its slow response. However,
high frequency transients such as switching surges may cause interference in the auxiliary
electronics such as electronic amplifier. Hence, it is recommended to install the amplifier
outside the switchgear panel.
5.5.5 Fault Location using IR Sensors
The TID sensor detects the existence of thermally generated particles inside the closed
compartment, and hence it cannot pinpoint or identify the hotspot. For instance, if there are
multiple hotspots as there were in the investigated test setup in this research, this sensor cannot
locate which connection is hotter than others. Secondly, it is hard to determine the exact
temperature of the hotspot by using this technique. However, this sensor can provide an
indication of the presence of incipient faults and it can be found that something wrong is going
on in a certain compartment. Alarms can be triggered with a certain level of TID output. Once
we get some alarm signal from the online monitoring system using this sensor, further
investigation is required to locate the fault point. Such an investigation can be done by using
thermo-graphic camera. Hence, a coordinated technique will be a powerful tool in order to avoid
unnecessary shutdowns for periodic maintenance and testing. Rather, we may only need to test a
specific compartment or switchgear when we get the alarm signal only.
5.6 Exact Location of Fault in a Switchgear
The TID sensor measures the thermally generated ions inside the closed compartment,
and hence it cannot pinpoint or identify the hotspot. For instance, if we have multiple hotspots
as we had in our test setup, this sensor cannot locate which connection is hotter than others.
Secondly, it is hard to determine the exact temperature of the hotspot by using this technique.
However, this sensor can provide an indication of incipient faults and it can be found that
something wrong is going on in a certain compartment. Alarms can be triggered with a certain
level of TID output. Once we get some alarm signal from the online monitoring system using
this sensor, further investigation is required to locate the fault point. Such investigation can be
75
done by using thermo-graphic camera through glass window of the switchgear. Hence, a co-
ordinated technique will be a powerful tool in order to avoid unnecessary shutdowns for
periodic maintenance and testing. Rather, we may only need to test a specific compartment or
switchgear when we get the alarm signal only. Thermographic or IR cameras may also pinpoint
the incipient faults due to PDs. There are only few locations where possible faults can be
initiated, which are cable compartment, CTs, PTs, bushings or bus bar insulations and can be
located easily.
As explained in Chapter 2 and 3, PD in air insulated switchgear induces transient
voltages that propagate rapidly over the switchgear metal panels. Due to high frequency nature
of these signals, the energy travels across the panel surface and will easily move to panels
adjacent to the one with source of discharge activity which results in detection of PD energy in
multiple panels. In [126], an advanced technique based on field programmable gate array
(FPGA) is introduced in order to determine the faulty panel in a multiplanel switchboard.
5.7 Integration of On-line Condition Monitoring Sensors with Central

Control System or SCADA
Figure 5.39 shows a possible implementation of online monitoring system in a typical

MV distribution system where D-dot sensors are installed near breaker poles, in the bus-bar
compartment and in the outgoing cable termination boxes. The essential purpose of online
monitoring of switchgear is to prevent high energy faults. An online monitoring system should
be able to analyze the measured data, and provide appropriate information to upper level
systems, such as protection system, programmable logic control system, distributed control
system (DCS) or supervisory control and data acquisition (SCADA) system.
Based on the importance of signal processing requirement presented in this research

work, it is justified to have a separate processing unit, collecting information from the sensors,
and sending the processed (de-noised) data to either remote systems (local data analysis and
control centre) or to central control station (e.g through SCADA). Remote or local system
(relays) may have a decision capability to decide about sending a trip signal directly to a circuit
breaker before sending the information to the upper level, in order to avoid severe damage. This
may happen in case of sudden faults. The presence of local control makes the system faster as
compared to central control only. It also reduces the amount of data transmitted to SCADA. The
data sent to the central control can be used for monitoring and efficient asset management
purposes. The information to be sent to SCADA, can be an analogue signal, indicating e.g. the
level of PD activity, or it can be a digital information as well, e.g. alarm signal. An alarm signal
can be an early warning, giving opportunity to check other possible indications of developing
fault, and change power system switching state to isolate the zone of the developing fault.
76
D-dot Sensors
Data
Acquisition
Signal De-
noising
Fault Location Discharge

Local Control
Algorithm Quantification
Communication
over SCADA
Maintenance Central Control

Planning Room
(a) (b)
Fig. 5.39. Online monitoring system, (a)- Flow chart of online monitoring, (b)- Typical implementation in
MV substation
77
78
Chapter 6
CONCLUSIONS AND FUTURE WORK
6.1 Conclusions
An intensive literature survey on the existing and potential methods for prediction and
early detection of the arc-flash in MV and LV switchgear has been carried out in this research
work. It is found that there are two major slowly developing incipient faults leading to an arc-
flash in switchgear i.e. insulation degradation and poor connections. These phenomena result in
PD ignition, micro-sparks or low energy arcing and thermal impacts. A number of sensor
technologies, to detect these defects, have been discussed and compared.
Three sensors that have been tested in the laboratory are RC, D-dot Sensor and TID.
HFCT is a proved technology, which have been used since many decades for the high frequency
current transient measurements for PD diagnostics. In this research work, it is used to provide
reference signal for the other sensors. D-dot sensor, which measures differential electric field, is
a novel sensor for the online monitoring of switchgear which can detect PD and arcing signals
generated inside the switchgear over a wide bandwidth. The sensor is recommended due to its
low cost, smaller size, simple design and easy implementation as compared to other sensors. Its
sensitivity and performance evaluation with reference to HFCT proves its reliability for the
online monitoring of air insulated switchgear. It is also tested for on-site PD measurements in an
operational switchgear which further endorsed its applicability for the said application.
RC is a developing technology for the high frequency transient measurement and has
been reported to be found reliable for high frequency transient measurements in various
research studies. Hence, D-dot sensor and RC are mainly useful for PD and arcing measurement
in MV switchgear.
Poor connections lead to heating effects in the cable compartment or bus bar
connections. Heating phenomenon is more prominent in LV switchgear compared to MV. TID
is implemented and tested for the thermal monitoring of hot joints. TID is found having
satisfactory performance to detect abnormal heating phenomenon at joints. One TID is enough
to monitor a whole section such as cable connection compartment in switchgear. It is very cost
effective, small in size and the need of only one sensor per compartment makes the installation
quite simple. However, this sensor is unable to pinpoint the hotspot among multiple joints.
Therefore, a thermo-graphic or IR camera can be a suitable option for the fault localization.
Online or on-site measurements of high frequency transients are strongly influenced by

external interference or noise. DWT technique is used in this research to de-noise PD and low
79
power arcing signals captured by non-intrusive sensors. DWT is based on MSD both in time
and frequency domain. DWT is compared with IIR technique and found to be better than IIR.
An implementation of combined techniques based on conventional and unconventional

PD measurements helped to correlate the PD energy calculated from the D-dot sensor signal
with apparent charge quantity calculated from current pulse measured by IEC60270 standard
PD measurement setup. It is observed that when apparent charge is plotted against the
cumulative energy of the RF signal for a number of PD events, defects can be separated on the
basis of cluster positions within scatter plot. Hence, various types of discharges can also be
identified. It can be unique idea for the detection of multiple faults in a single switchgear panel
based on such correlation.
Advancement in non-intrusive sensor technologies, digital wireless communication and

modern microprocessors enable us detect incipient faults much earlier than they may lead to
severe damages. In spite of careful design, and recommended pre-emptive protection
technologies, the risk of arc flash may never be totally eliminated due to non-predictable causes
such as direct human interaction or insects. This is why fast, effective and reactive protection is
justified alongwith pro-active detection technologies. This combination of pre-emptive and
reactive technologies provides the best overall protection of the equipment as well as the
personnel working around the equipment and should be considered for broader level of
equipment protection and personnel safety.
6.2 Future Work
Possible future work will include the parameter identification modeling of the D-dot
sensor. Some parameters are directly related to the sensor’s head, which can be calculated from
the geometrical structure of the sensor, whereas others are related to the measurement setup and
can be measured only from the installed setup. A follow up research may address the
determination of the parameters along with the complete modeling and simulation of the sensor
through an electromagnetic simulation software. Once the precise model for the sensor along
with its measuring system is developed, the precise calibration of the sensor will also be
possible.
Integration of the new sensors to the existing protection relays and SCADA systems is
currently a major challenge in the way to practical installation of the online condition
monitoring system for the measurement of PD and low energy arcing signals. The reason is PD
and arcing are high frequency transients and data acquisition of such high frequency signals
makes the overall system quite expensive. It also requires bulk amount of data for analysis in
order to determine the need for maintenance. The size of data brings a challenge for the
communication as well. These issues need to be researched and developed in the future work.
More sensitive designs of D-dot sensor and TID are also required for more efficient and
reliable fault detection. Such designs can be put into practical prototypes with the help of
collaboration with the existing manufacturers.
80
References
[1]. M. Goodman, “Methods of inspection to determine the presence of potential arc-flash

incidents”, NETA World Magazine, pp. 1-6, Spring- 2007.
[2]. G. Gregory, “Preventing arc flash incidents in the workplace”, EC&M Magazine, June
2003.
[3]. H. Land III, C. Eddins, and J. Klimek, “Evolution of arc fault prediction technology”,
Johns Hopkins APL Tech. Dig., vol. 25, no. 2, pp. 140-153, 2004.
[4]. L. Kumpulainen, J.A. Kay and M. Aurangzeb, “Maximal protection: Lowering incident
energy and arc blast elements by minimizing arcing time”, 58th Annual IEEE IAS
Petroleum and Chemical Industry Conference (PCIC), Toronto, 19-21 Sep. 2011.
[5]. Draft guide for testing metal-enclosed switchgear rated up to 38kV for internal arcing
Faults - Corrigendum 1, IEEE Unapproved Draft Std. PC37.20.7-2007 Cor 1/D4, 2009.
[6]. H.B. Land III, “The behavior of arcing faults in low voltage switchboards”, IEEE
Transactions on Industry Applications, vol. 44, no. 2, pp. 437-444, March 2008.
[7]. J. L. Blackburn and T. J. Domin, “Protective relaying- principles and applications”, 3rd
Edition, CRC Press, 2006.
[8]. P. Muller, S. Tenbohlen, R. Maier and M. Anheuser, “Characteristics of serial and
parallel low current arc faults in time and frequency domain”, Proc. of the 56th IEEE
Holm Conf. on Electrical Contacts (HOLM), Charleston, 4-7 Oct. 2010.
[9]. H. Land III, “Sensing switchboard arc faults”, IEEE Power Eng. Review, vol. 22, no. 4,
pp. 18-27, April 2002.
[10]. J. A. Kay, J. Arvola and L. Kumpulainen, “Protecting at the speed of light”, IEEE Ind.
Appl. Mag., vol. 17, no. 3, pp.12-18, May-June 2011.
[11]. L. Kumpulainen and S. Dahl, “Minimizing hazard to personnel, damage to equipment,
and process outages by optical arc-flash protection” Proc. of IEEE/ PCIC Conf. Europe,
Oslo, 15-17 June 2010.
[12]. C. Inshaw and R.A. Wilson, “Arc flash hazard analysis and mitigation”, 58th Annu.
Conf. for Protective Relay Engineers, Texas, pp. 145-157, 5-7 April 2005.
[13]. T. Dugan, “Reducing the arc flash hazard”, IEEE Ind. Appl. Mag., pp. 51-58, May/Jun
2007.
[14]. P. Barkhordar, “How accurate are your arc flash hazard study results”, Proc. of 57th
Annu. PCIC/ IAS Conf., San Antonio, pp. 1-6, 20-22 Sep. 2010.
[15]. NFPA 70E Standard, “Electrical safety in the workplace”, 2000, 2004, and 2009
Editions.
[16]. IEEE Standard 1584, “Guide for performing arc-flash hazards calculations”, 2002 and
2004 amendments.
[17]. J. Buff and K. Zimmerman, “Reducing arc-flash hazards”, IEEE Ind. Appl. Mag., vol.
14, no. 3, pp. 40,47, May-June 2008.
[18]. G. Roscoe, M. E. Valdes and R. Luna, “Methods for arc-flash detection in electrical
equipment”, Proc. of 57th Annu. PCIC/ IAS Conf., San Antonio, 20-22 Sep. 2010.
[19]. R. H. Lee, “Pressures developed by arcs”, IEEE Trans. on Ind. Appl., vol. IA-23, no. 4,
pp. 760-763, July 1987.
81
[20]. H.L. Floyd, D.R. Doan, C.T. Wu, and S.L. Lovasic, “Arc flash hazards and electrical
safety program implementation”, 40th Annu. Ind. Appl. Conf., Hong Kong, vol. 3, pp.
1919-1923, 2-6 Oct. 2005.
[21]. M. Braunovic, V. V. Konchits, and N. K. Myshkin, “Electrical contacts: fundamental,
applications and technology”, Taylor & Francis Group LLC, New York, 2006.
[22]. H. B. Land III, C. L. Eddins, L. R. Gauthier Jr. and J. M. Klimek, “Design of a sensor to
predict arcing faults in nuclear switchgear”, IEEE Trans. on Nuclear Sci., vol. 50, no. 4,
pp. 1161-1165, Aug 2003.
[23]. D. Brechtken, “Preventive arc fault protection”, Proc. of IEEE/PES Conf. and Expo. on
Transmission and Distribution, Atlanta, 28 Oct- 02 Nov 2001.
[24]. X. Zhou and T. Schoepf, “Characteristics of overheated electrical joints due to loose
connection”, 57th IEEE Conf. on Elect. Contacts. Holms, San Antonio, 11-14 Sept.
2011.
[25]. Z.A.A. Zarim, A.B.A. Ghani and S.K. Tiong, “The measurement and temperature
profile of cable insulation failure due to loose connection at the cable termination”, Intl.
Sympo. on Elect. Insul. Materials (ISEIM), Mie, Japan, pp. 570-572, 7-11 Sept. 2008.
[26]. C. Sumereder and M. Muhr, “Estimation of residual lifetime- Theory and practical
problems”, 8th Holfer’’s day Pororoz, TU Graz, 6-8 November 2005.
[27]. M. Muhr, “Aging and degradation, their detection and monitoring & asset
management”, Intl. Sympo. on Elect. Insul. Materials (ISEIM), Mie, Japan, pp.183-186,
7-11 Sept. 2008.
[28]. F. Senn, M. Muhr, W. Ladstätter and W. Grubelnik, “Complexity of determining factors
for the thermal evaluation of high voltage insulation systems on the example of rotating
machines”, DISEE Demänovská Dolina, Slovakia, 2008.
[29]. High-voltage test techniques – Partial discharge measurements, IEC standard 2000-12,
Third ed., IEC, Switzerland.
[30]. G. M. Hashmi, “Partial discharge detection for condition monitoring of covered
conductor overhead distribution networks using rogowski coil”, PhD Thesis, Helsinki
University of Technology (TKK) Finland, TKK Publication Series, Ref. No. TKK-
DISS-2493, 22 August 2008.
[31]. F. H. Kreuger, “Partial discharge detection in high-voltage equipment”, Butterworths,
1989.
[32]. R. Bartnikas and E. J. McMahon (editors), “Corona measurement and interpretation”,
Engineering Dielectrics, vol. 1, ASTM STP 669, 1979.
[33]. E. Kuffel, W. S. Zaengl, and J. Kuffel, “High voltage engineering: Fundamentals”, 2nd
ed., Newnes, Butterworth-Heinemann, 2000.
[34]. H. Edin, “Partial discharge studied with variable frequency of the applied voltages”,
Kungl Tekniska Hogskolan (KTH), Doctoral Dissertation, Stockholm, Sweden, 2001.
[35]. Y. Qiu, A. Sun, and Z. Zhang, “Factors affecting partial discharges in SF6 gas
impregnated polymer film insulation”, 2nd International Conf. on Properties and
Applications of Dielectric Materials, Beijing, September 12-16, 1988.
[36]. M. Aro, J. Elovaara, M. Karttunen, K. Nousiainen, and V. Palva,
”Suurjännitetekniikka”, Otatieto 568, 2nd edition, 2003.
82
[37]. D. König and R. Y. Narayana, “Partial discharges in power apparatus”, Berlin: Vde-
Verlag GmbH, 1993.
[38]. M. Shafiq, “Design and implementation of partial discharge measurement sensors for
on-line condition assessment of power distribution system components”, Aalto
University Publication Series, Doctoral Dissertation, No. 152/2014, October 2014.
[39]. A. Pedersen, G. C. Crichton, and I. W. McAllister, “PD related field enhancement in the
bulk medium”, Journal of Gaseous Dielectrics VII, Springer US, pp. 223-229, 1994.
[40]. A. Pedersen, G. C. Crichton, and I.W. McAllister, “The theory and measurement of
partial discharge transients”, IEEE Trans. on Electr. Insul., vol. 26, no.3, pp. 487-497,
Jun. 1991.
[41]. C. G. Crichton, P. W. Karlsson, and A. Pedersen, “Partial discharges in ellipsoidal and
spheroidal voids”, IEEE Trans. on Electr. Insul., vol. 24, no. 2, pp. 335-342, 1989.
[42]. IEC 62271-200, High-voltage switchgear and controlgear – Part 200: AC metal-
enclosed switchgear and controlgear for rated voltages above 1 kV and up to and
including 52 kV, 2nd Ed., 2011.
[43]. IEC 60092-508 Electrical installations in ships – Part 508: Switchgear and controlgear
assemblies for rated voltages above 1kV and up to and including 15kV, 1st Ed., 4th Nov
2002.
[44]. IEC 60439-1 Low-voltage switchgear and controlgear assemblies – Part 1: Type-tested
and partially type-tested assemblies, Ed. 4.1, 2004.
[45]. IEEE C37.20.2, “IEEE standard for metal-clad switchgear”, July 2000.
[46]. J.P. Nelson and P.K. Sen, “High resistance grounding of low voltage systems: A
standard for the petroleum and chemical industry”, IEEE Trans. on Ind. Appl., vol. 35,
no. 4, pp. 941-948, July/August 1999.
[47]. D.C. Mohla, T. Driscoll, P.S. Hamer and S.A.R. Panetta, “Mitigating electric shock and
arc flash energy –A total system approach for personnel and equipment protection”,
Proc. of 57th Annu. PCIC/ IAS Conf., San Antonio, pp. 1-6, 20-22 Sep. 2010.
[48]. L.R. Kingrey, R.D. Painter and A.S. Locker, “Applying high resistance neutral
grounding in medium voltage systems”, IEEE Trans. on Ind. Appl., vol. 47, no. 3, pp.
1220-1231, May/June 2011.
[49]. J.R. Dunki-Jacobs, “The escalating arcing ground fault phenomenon”, IEEE Trans. on
Ind. Appl., vol. IA-22, no. 6, pp. 1156-1161, Nov/Dec 1986.
[50]. R.A. Jones, D.P. Liggett, M. Capelli-Schellpfeffer, T. Macalady, R.E. Saunders,
Downey, L.B. McClung, A. Smith, S. Jamil and V.J. Saporita, “Staged tests increase
awareness of arc-flash hazards in electrical equipment”, IEEE Trans. on Ind. Appl., vol.
36, no. 2, pp. 659-667, March/April 2000.
[51]. R. Wilkins, M. Lang and M. Allison, “Effect of insulating barriers in arc flash testing”,
IEEE Trans. Ind. Appl., vol. 44, no. 5, pp. 1354-1359, Sept.-Oct. 2008.
[52]. M. Lang, K. Jones and T. Neal, “Impact of arc flash events with outward convective
flows on worker protection strategies”, IEEE Trans. Ind. Appl., vol. 47, no. 4, pp. 1597-
1603, July/Aug 2011.
[53]. IEC Standard 909, “Short-circuit current calculation in three-phase A.C. systems”,
Second Impression, IEC, 1989.
83
[54]. I. Balasubramanian and A. Graham, “Impact of available fault current variations on arc-
flash calculations”, IEEE Trans. Ind. Appl., vol. 46, no. 5, pp. 1836-1842, Sep/Oct 10.
[55]. D. Sweeting, “Applying IEC 60909, short-circuit current calculations”, IEEE/ IAS
Annu. PCIC Conf., Toronto, Canada, September 19–21, 2011.
[56]. J. Eli Nielsen, “Analysis of how large scale DG affects network business”, CODGUNet
project report, 2003.
[57]. D. Sweeting, “Arcing faults in electrical equipment”, Paper No. PCIC-2009-1, IEEE/
IAS Annu. PCIC Conf., Anaheim, USA, September 14–16, 2009.
[58]. D. Shipp and D. Wood, “Mitigating arc-flash exposure”, IEEE Ind. Appl. Mag. pp. 28-
37, July/August 2011.
[59]. K. Malmedal and P. Sen “Arcing fault current and the criteria for setting ground fault
relays in solidly-grounded low voltage systems”, Conference Record of Ind. and
Commercial Power Systems Tech. Conf. Clearwater, USA, 7-11 May 2000.
[60]. S. Swencki, J. Smith, D. Roybal, D. Burns, G. Wetzel and D. Mohla, “Electrical safety,
arc flash hazards, and the “standards” a comprehensive overview”, IEEE/ IAS PCIC
Conference, Denver, USA, Paper no. PCIC-2005-39, Sept. 12–14 2005.
[61]. L. Kumpulainen, S. Dahl and J. Ma, “Mitigation of arc-flash hazards and reduction of
costs by selective arc-flash protection” Conf. Record of International Conference on
Electricity Distribution (CICED), Guangzhou, China, December 10-13, 2008.
[62]. J. Simms and G. Johnson, “Protective relaying methods for reducing arc flash energy”,
IEEE Trans. Ind. Appl., vol. 49, no. 2, pp. 803-813, March-April 2013.
[63]. R. Garzon, “The arc terminator”, IEEE Ind. Appl. Mag., pp. 51-55, May/June 2003.
[64]. J.A. Kay and L. Kumpulainen, “Maximizing protection by minimizing arcing times in
medium voltage systems”, IEEE Trans. Ind. Appl., vol. 49, no. 4, pp. 1920-1927,
July/August 2013.
[65]. M. Fortin, J. Gerth, R. McLaughlin and P. Riffon, “Short-circuit strength and short-
circuit testing of power and distribution transformers”, IEEE/PES Transformers
Committee, Spring 2008 Meeting, Charlotte, March 2008.
[66]. T.M. Bilodeau, J.J. Shea, G.J. FitzPatrick and W.J. Sarjeant, “A critical evaluation of
conventional partial discharge measurement techniques for discrete capacitors”, IEEE
Electr. Insul. Mag., vol. 3, no. 4, pp. 10-17, July 1987.
[67]. M. Muhr, R. Schwarz, S. Pack and B. Koerbler, “Unconventional partial discharge
measurement (electrical insulation evaluation)”, Annu. Conf. on Elect. Insul. and
Dielectr. Phenomena (CEIDP '04), Boulder, USA, pp. 430-433, 2004.
[68]. IEC standard 60270, High voltage test techniques- Partial discharge measurements, 3rd
Ed., 2000.
[69]. P. Petchphung, M. Leelajindakrairerk, N. Pattanadech, P. Yutthagowith and K.
Aunchaleevarapan, “The comparison study of PD measurement with conventional
method and unconventional method”, Proc. of Intl. Conf. on Cond. Monitoring and
Diagnostics (CMD’08), pp. 32-36, Beijing, 21-24 April 2008.
[70]. H. Okubo, “Recent progress and future perspective on condition monitoring and
diagnostic techniques for power equipment in Japan”, Proc. of Intl. Conf. on Cond.
Monitoring and Diagnostics (CMD’08), Beijing, pp. 13-18, 21-24 April 2008.
84
[71]. Y. Han and Y.H. Song, “Condition monitoring techniques for electrical equipment- A
literature survey”, IEEE Trans. on Power Del., vol. 18, no. 1, pp. 4-13, Jan 2003.
[72]. S.I. Cho, “On-line PD (Partial Discharge) monitoring of power system components”,
Master Thesis, Aalto University Publication Series, September 2011.
[73]. X. Sa, P.J. Moore, M.D. Judd and I. Portugues, “An investigation into electromagnetic
radiation due to partial discharges in high voltage equipment based on non-contact
measurements”, Conf. Record of IEEE Intl. Symposium on Electrical Insulation (ISEI
2008), Vancouver, pp. 567-570, 9-12 June 2008.
[74]. L.E. Lundgaard, “Partial discharge XIII acoustic partial discharge detection-
fundamental considerations”, IEEE Elect. Insul. Mag., vol. 8, no. 4, pp. 25-31, July/Aug
1992.
[75]. Y. Tian, P.L. Lewin, A.E. Davies and Z. Richardson, “Acoustic emission detection of
partial discharges in polymeric insulation”, 11th Intl. Symposium on High Voltage
Engineering, London, Publ. No. 467, pp. 82-85, 1999.
[76]. E. Jouseau, G. Vivien, E. Odic and C. Maroni, “Exhibition of an ageing criterion based
on partial discharge detection for medium voltage equipment”, 5th IEEE Intl.
Symposium on Diagnostics for Electric Machines, Power Electronics and Drives,
Vienna, 7-9 Sept. 2005.
[77]. S. Ahmadi, M.R. Naghashan and M. Shadmand, “Partial discharge detection during
electrical aging of generator bar using acoustic technique”, Conf. Record of IEEE Intl.
Symposium on Electrical Insulation (ISEI), San Juan, pp. 576-578, 10-13 June 2012.
[78]. A.K. Lazarevich, “Partial discharge detection and localization in high voltage
transformers using an optical acoustic sensor”, Master thesis, Virginia Polytechnic
Institute and State University, May 2003.
[79]. S.M. Markalous and K. Feser, “All- acoustic PD measurements of oil/paper- insulated
transformers For PD-location”, Proc. of the 2nd Intl. Conf. on Advances in Processing,
Testing and Application of Dielectric Materials (APTADM), Wroclaw, 15-17 Sep 2004.
[80]. M. Muhr and R. Schwarz, “Experience with optical partial discharge detection”,
Materials Science-Poland, vol. 27, no. 4/2, pp. 1139-1142, 2009.
[81]. R. Schwarz and M. Muhr, “Modern technologies in optical partial discharge detection”,
Conf. on Elect. Insul. and Dielectric Phenomena (CEIDP), Vancouver, pp. 163-166, 14-
17 Oct 07.
[82]. R. Schwarz and M. Muhr, “Evaluation of partial discharge impulses with optical and
conventional detection systems”, Proc. of the XIVth Intl. Symposium on High Voltage
Engineering, Tsinghua University, Beijing, China, August 25-29, 2005.
[83]. C. Suwanasi, T. Suwanasri, P. Fuangpian and S. Ruankon, “Investigation on partial
discharge of power cable termination defects using high frequency current
transformer”, 10th Intl. Conf. on Electrical Engineering/Electronics, Computer,
Telecomm. and Info. Tech. (ECTI-CON), Krabi, 15-17 May 2013.
[84]. K. Mallikarjunappa and M.C. Ratra, “Detection of partial discharges in power
capacitors using high frequency current transformers”, Proc. of Annu. Conf. on Elect.
Insul. and Dielectric Phenomena, Pocono Manor, pp. 379-384, 28-31 Oct 1990.
[85]. C. Suwanasi, T. Suwanasri, P. Fuangpian and S. Ruankon, “Investigation on partial
discharge of power cable termination defects using high frequency current
85
transformer”, 10th Intl. Conf. on Electrical Engineering/Electronics, Computer,
Telecomm. and Info. Tech. (ECTI-CON), Krabi, 15-17 May 2013.
[86]. S. Miller, “Inspection methods to determine potential arc flash”, IEEE/IAS Electrical
Safety Workshop (ESW), Toronto, pp. 1-3, 25-28 Jan. 2011.
[87]. P. Dejean, N. Foulon, M. Goldman and H. Dejean, “Chemical aspects of the ageing of
solid insulating materials submitted to partial discharges”, Proc. of the 4th Intl. Conf. on
Conduction and Breakdown in Solid Dielectrics, Sestri Levante, pp. 181-185, 22-25 Jun
1992.
[88]. M. Goldman, A. Goldman and J. Gatellet, “Physical and chemical aspects of partial
discharges and their effects on materials”, IEE Proc. on Science, Measurement and
Technology, vol. 142, no. 1, pp. 11-16, Jan 1995.
[89]. J. Lopez-Roldan, T. Tang and M. Gaskin, “Optimisation of a sensor for onsite detection
of partial discharges in power transformers by the UHF method”, IEE Trans. on
Dielectrics and Elect. Insul., vol. 15, no. 6, pp. 1634-1639, 2008.
[90]. M.D. Judd, Li Yang and I.B.B. Hunter, “Partial discharge monitoring of power
transformers using UHF sensors. Part I: sensors and signal interpretation”, IEEE Elect.
Insul. Mag., vol. 21, no. 2, pp. 5-14, 2005.
[91]. S. Kaneko, S. Okabe, M. Yoshimura, H. Muto, C. Nishida and M. Kamei, “Detecting
characteristics of various type of antennas on partial discharge electromagnetic wave
radiating through insulating spacer in gas insulated switchgear”, IEEE Trans. on
Dielectrics and Elect. Insul., vol. 16, no. 5, pp. 1462-1472, Oct 2009.
[92]. H. Zhu, V. Green, M. Sasic and S. Halliburton, “Increased sensitivity of capacitive
couplers for in-service PD measurement in rotating machines”, IEEE Trans. on Energy
Conversion, vol. 14, no. 4, pp. 1184-1192, 1999.
[93]. M. Muhr and R. Woschitz, “Partial discharge diagnostic”, Proc. of 6th Intl. Conf. on
Properties and Applications of Dielectric Materials, Xi'an, pp. 223-226, 2000.
[94]. T.E. Goodeve, G.C. Stone and L. Macomber, “Experience with compact epoxy-mica
capacitors for rotating machine partial discharge detection”, Electrical Manufacturing &
Coil Winding Conf., Rosemont, pp. 685-689, 1995.
[95]. D. Pommerenke, T. Strehl, R. Heinrich, W. Kalkner, F. Schmidt, and W. Weissenberg,
“Discrimination between internal PD and other pulses using directional coupling
sensors on HV cable systems”, IEEE Trans. on Dielectrics and Elect. Insul., vol. 6, no.
6, pp. 814-824, 1999.
[96]. X. Chucheng, Z. Lingyin, T. Asada, W.G. Odendaal and J.D. van Wyk, “An overview
of integratable current sensor technologies”, 38th IAS Annual Meeting and Ind. Appl.
Conf., Salt Lake City, vol. 2, pp. 1251-1258, 12-16 Oct. 2003.
[97]. M. Shafiq, L. Kutt, M. Lehtonen, T. Nieminen and M. Hashmi, “Parameters
identification and modeling of high-frequency current transducer for partial discharge
measurements”, IEEE Sensor Journal, vol. 13, no. 3, pp. 1081-1091, March 2013.
[98]. M. Shafiq, M. Lethonen, L. Kutt, G. A. Hussain and M. Hashmi, “Effect of terminating
resistance on high frequency behaviour of Rogowski coil for transient measurements,
Elektronika ir elektrotechnika, vol. 19, no. 7, pp. 22-28, 2013.
[99]. IEEE Std C57.127-2000, IEEE trial-use guide for the detection of acoustic emissions
from partial discharges in oil-immersed power transformers, 2000.
86
[100]. S.E.U. Lima, O. Frazao, R.G. Farias, F.M. Araujo, L.A. Ferreira, J.L. Santos and V.
Miranda, “Mandrel-based fiber-optic sensors for acoustic detection of partial
discharges- a proof of concept”, IEEE Trans. on Power Del., vol. 25 , no. 4, pp. 2526-
2534, 2010.
[101]. A. Zargari and T.R. Blackburn, “Application of optical fibre sensor for partial discharge
detection in high-voltage power equipment”, Annu. report of IEEE Conf. on Electrical
Insulation and Dielectric Phenomena, Millbrae, vol. 2, pp. 541-544, 1996.
[102]. Exertherm sensor product brochure, Exergen Ltd. Netherland. www.exergeneurope.com
[103]. Huber+Suhner SMA connector, Type 22 SMA-50-0-203/111NH .
[104]. S. Burkhart, “Coaxial E-field probe for high voltage microwave measurements”, IEEE
Trans. on Microwave Theory and Techniques, vol. 33, no. 3, pp. 262-265, March 1985.
[105]. D. A. Ward, and J. La. T. Exon, “Using Rogowski coil for transient current
measurements”, Engineering Science and Education Journal, vol. 2, no. 3, pp. 105-113,
1993.
[106]. C. R. Hewson, W. F. Ray, and R. M. Davis, “Verification of Rogowski current
transducer’s ability to measure fast switching transients”, 21st Annual IEEE Applied
Power Electronics Conference and Exposition (APEC), Dallas, March 2006.
[107]. P. Mähönen and V. Virtanen, “The Rogowski coil and the voltage divider in power
system protection and monitoring”, Internationale des Grandes Reseaux Electriques a
Haute Tension, CIGRE, 1996.
[108]. G. Robles, M. Argueso, J. Sanz, R. Giannetti, and B. Tellini, “Identification of
parameters in a Rogowski coil used for the measurement of partial discharges”, Proc. of
IEEE Conf. on Instr. and Meas. Tech. (IMTC), Warsaw, Poland, pp. 1-4, May 2007.
[109]. G. M. Hashmi, M. Lehtonen, and M. Nordman, “Modeling and experimental
verification of on-line PD detection in MV covered-conductor overhead networks”,
IEEE Trans. on Dielectr. and Elect. Insul., vol. 17, no. 1, pp. 167-180, 2010.
[110]. R. Hlavacek, R. Prochazka, K. Draxler and V. Kvasnicka, “The Rogowski coil
software”, 16th Symposium on Exploring New Frontiers of Instrumentation and
Methods for Electrical and Electronic Measurements, Florence, Italy, Sep. 2008.
[111]. A. Marta, G. Robles, and J. Sanz, “Implementation of a Rogowski coil for the
measurement of partial discharges”, Review of scientific instruments, vol. 76, no. 6,
June 2005.
[112]. L. D. Slifka, “An accelerometer based approach to measuring displacement of a vehicle
body”, Master Thesis, Dept. Electr. Comput. Eng., University of Michigan, Dearborn,
2004.
[113]. S. Xiao, P. J. Moore, M. D. Judd and I. E. Portugues, “An investigation into
electromagnetic radiation due to partial discharges in high voltage equipment”, IEEE
Power Engineering Society General Meeting, Tampa, pp. 1-7, 24-28 June 2007.
[114]. G. C. Stone, “Partial discharge diagnostics and electrical equipment insulation condition
assessment”, IEEE Trans. Dielectr. Elect. Insul., vol. 12, no. 5, pp. 891- 904, Oct. 2005.
[115]. M. Hikita, S. Okabe, H. Murase and H. Okubo, “Cross-equipment evaluation of partial
discharge measurement and diagnosis techniques in electric power apparatus for
transmission and distribution” IEEE Trans. Dielectr. Elect. Insul., vol. 15, no. 2, pp.
505-518, April 2008.
87
[116]. J. Klüss, “Measuring picoseconds flashover in pressurized sulfur hexafluoride (SF6)”,
Doctoral Dissertation, Aalto University Publication Series, Ref. no. 53/2011, July 2011.
[117]. I. A. Metwally, “ D-Dot probe for fast-front high-voltage measurement”, IEEE Trans.
Instr. and Meas., vol. 59, no. 8, pp. 2211-2219, Aug. 2010.
[118]. J. R. Andrews, “UWB signal sources, antennas & propagation,” Application Note AN-
14a, Picosecond Pulse Labs, USA, pp. 7-8, 2003.
[119]. Y. Gao and I. Wolff, “A simple electric near field probe for microwave circuit
diagnostics”, IEEE MTT-S International Microwave Symposium, San Francisco, Digest-
1996, vol. 3, pp. 1537-1540, 17-21 June, 1996.
[120]. Y. Wang, J. Li, and L.-X. Ran, “An equivalent circuit modeling method for ultra-
wideband antennas”, Progress in Electromagnetic Research, vol. 82, pp. 433-445, 2008.
[121]. B. A. Land III, J. M. Klimek, L. R. Gauthier Jr., C. L. Eddins, “Thermal Ionization
Detector”, United States Patent, Patent No. 6292105 B1, Sep 18, 2001.
[122]. V. Nagesh and B. LGururaj, “Evaluation of digital filters for rejecting discrete spectral
interference in on-site PD measurements”, IEEE Trans. on Elect. Insul., vo1. 28, no. 1,
pp. 73-85, Feb 1993.
[123]. M. Hashmi, M. Lehtonen, M. Nordman, R. A. Jabbar and S. A. Qureshi, “Wavelet-
based de-noising of on-line PD signals captured by Pearson coil in covered conductor
overhead distribution networks”, Intl. Journal of Elect. Power and Energy
Sys.(IJEPES), vol. 43, no. 1, pp. 1185-1192, Dec 2012.
[124]. L. Satish and B. Nazneen, “Wavelet-based de-noising of partial discharge signals buried
in excessive noise and interference”, IEEE Trans. on Dielectr. Elect. Insul., vol. 10, no.
2, pp. 354-367, April 2003.
[125]. C. S. Burrus, R. A. Gopinath and H. Guo, “Introduction to wavelets and wavelet
transforms: a primer”, Prentice-Hall Inc. 1998.
[126]. C. Walton, S. Carter, C. Smith and M. Michel, “The location of switchgear partial
discharge by panel and techniques to correlate switchgear and cable partial discharge
with load and substation environment”, Proc. of 21st Intl. Conf. on Electricity
Distribution, Frankfurt, 6-9 June, 2011.
88
Appendix-I
Technical Details of the Thermal Ionization Detector (TID)
The specifications are typical at 23ºC and 50% relative humidity unless otherwise stated.
The detector designed in UK and is subject to the requirements of Radioactive Substance Act
1993 and to the Ionising Radiations Regulations 1999 made under the provisions of the Health
and Safety at Work Act 1974.
The detector, independently tested by the Health Protection Agency (HPA), conform to
all the requirements specified in the ‘Recommendations for ionization smoke detectors in
implementation of radiation standards’ published by the Nuclear Energy Agency of the
Organization for Economic O-operation and Development (OECD) 1977. Technical details of
the detector are given in Table 1.
TABLE 1
Technical Details of the Thermal Ionization Detector (TID)

Detection Ionization Chamber
principle
Chamber Twin compensating chambers using one single
configuration sided ionizing radiation source
Radioactive Americium 241 (Am-241)
isotope
Quantity of Am- 1 micro curie
241
Activity 33.3 kBq (kilobecquerel), i.e. 33300 decays per
second
Supply wiring Two-wire supply; polarity insensitive;
Humidity 0 to 95% relative humidity (no condensation or
icing)
Dimensions Diameter: 3.4 cm
Internal chamber’s height: 2.2 cm
External chamber’s height: 1.2 cm
89
Errata
Article:
G. A. Hussain, M. Shafiq, M. Lehtonen, M. Hashmi, “Online Condition Monitoring of
MV Switchgear Using D-Dot Sensor to Predict Arc-Faults”, IEEE Sensors Journal, vol.
15, no. 12, pp. 7262-7272, Dec 2015.
Corrections:
1. Page: 7264, left column, first paragraph, first sentences should be read as:
“where p is the constant which depends on geometry of the sensor and
surrounding, nature of fault and dielectric medium.”
2. Equation (7) should be corrected as follows:
f C C1
f1 C
91
2316 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 4, OCTOBER 2013
G. Amjad Hussain (S’06) received the B.E. degree Matti Lehtonen (M’12) received both his Master’s
in electrical engineering from the University of Engi- and Licentiate degrees in electrical engineering from
neering and Technology (UET), Lahore, Pakistan, in Helsinki University of Technology, Helsinki, Fin-
2007 and the M.Sc. degree in electrical engineering land, in 1984 and 1989 respectively, and the Doctor
(power systems and high voltage engineering) from of Technology degree from Tampere University of
Aalto University, Espoo, Finland, in 2012, where Technology, Tampere, Finland, in 1992.
he is currently pursuing the Ph.D. degree in power He was with VTT Energy, Espoo, Finland, from
systems. 1987 to 2003, and since 1999, has been a Professor
He has authored and co-authored a number of at the Helsinki University of Technology, where he
IEEE publications. is now head of Power Systems and High Voltage En-
gineering. His main activities include power system
planning and asset management, power system protection including earth fault
problems, harmonic-related issues, and applications of information technology
in distribution systems.
Lauri Kumpulainen (M’06) received the M.Sc. and
Licentiate degrees in electrical power engineering
from Tampere University of Technology, Tampere,
Finland, in 1987 and 2003, respectively.
Currently he is Program Director with Oy Meri- John A. Kay (M’94–SM’98–F’12) received the de-
nova Ab, Vaasa, Finland. gree in electrical/electronic engineering technology
Mr. Kumpulainen is a member of CIGRE, and has from Conestoga College, Kitchener, ON, Canada, in
authored a number of conference papers, especially 1977.
related to arc Àash protection. He has authored a wide variety of award-winning
technical papers and other technical articles and
manuals related to medium-voltage electrical
control and protection systems, arc-resistant equip-
ment, and infrared technologies. Several of his
papers have been published in the IEEE Industrial
Joni V. Klüss received the B.Sc. degree from Applications Society TRANSACTIONS and IEEE
Lappeenranta University of Technology, Lappeen- Industry Applications Magazine.
ranta, Finland, in 2002, the M.Sc. degree in electrical Mr. Kay is a Fellow the IEEE Industry Application Society and is actively in-
engineering from the Helsinki University of Tech- volved with the IEEE Pulp and Paper Industry Committee, serving on its main
nology, Helsinki, Finland, in 2008, and the Ph.D. executive board—on the conference committee and on several subcommittees.
degree in electrical engineering from Aalto Univer- He serves on several other technical groups and participated on the local plan-
sity, Espoo, Finland, in 2011. ning committees for the 2011 IEEE-PCIC and 2002 IEEE-PPIC.
His research interests include high-voltage (HV)
engineering, ranging from gaseous insulation and di-
electric breakdown to HV measurement techniques
and EMP research.
Department of Electrical Engineering and Automation
Arc-Flash is an unintentional discharge of

Methods for Arc-Flash
Aalto-DD 221/2015

electricity through air which produces very
high temperature (19,600°C) and a force
equivalent to being hit by a hand grenade.
Occurrence of such an electrical fault in Prediction in Medium
switchgear causes interruption of electricity
to the end users and damage to the other
substation equipment and to the personnel
Voltage and Low Voltage
working in its vicinity. This thesis provides
methods for pre-emptive detection of Switchgear
potential arc-flash in MV/ LV switchgear
using online condition monitoring
techniques. Three non-intrusive sensors i.e.
D-dot sensor, Rogowski coil and Thermal
Ionization Detector have been designed, Ghulam Amjad Hussain
implemented and tested in switchgear to
detect major causes of arc-flash.
Assessment of reliability, sensitivity and
operability of the mentioned sensors is done
based on the analysis of acquired data from
laboratory tests as well as on site in practical
switchgears. The integration of the proposed
solution in SCADA system will be a
breakthrough solution in minimizing arc-
flash accidents and saving precision lives.
ISBN 978-952-60-6600-4 (printed) BUSINESS +

9HSTFMG*aggaae+
ISBN 978-952-60-6599-1 (pdf) ECONOMY

ISSN-L 1799-4934
ISSN 1799-4934 (printed) ART +
ISSN 1799-4942 (pdf) DESIGN +
ARCHITECTURE
Aalto University
Aalto University
School of Electrical Engineering SCIENCE +
Department of Electrical Engineering and Automation TECHNOLOGY
www.aalto.fi
CROSSOVER
DOCTORAL
DOCTORAL DISSERTATIONS
DISSERTATIONS
2016

Methods For Arc Flash Prediction in MV LV Switchgear

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Methods For Arc Flash Prediction in MV LV Switchgear

Uploaded by

Copyright:

Available Formats

Department of Electrical Engineering and Automation

Arc-Flash is an unintentional discharge of

Ghulam Amjad Hussain

ISBN 978-952-60-6600-4 (printed) BUSINESS +

ISBN 978-952-60-6599-1 (pdf) ECONOMY

Methods for Arc-Flash Prediction in

Ghulam Amjad Hussain

A doctoral dissertation completed for the degree of Doctor of

Aalto University publication series

© Ghulam Amjad Hussain

ISBN 978-952-60-6600-4 (printed)

Publication orders (printed book):

Avainsanat Keskijännite, pienjännite, kytkinkojeisto, valokaarivika, osittaispurkaus,

Ghulam Amjad Hussain

I. L. Kumpulainen, G. A. Hussain, M. Lehtonen, J. A. Kay, “Preemptive Arc Fault

II. L. Kumpulainen, G. A. Hussain, K. Kauhaniemi, M. Lehtonen, “Aspects of Arc-Flash

III. G. A. Hussain, L. Kumpulainen, M. Lehtonen, J. A. Kay, “Preemptive Arc Fault

IV. G. A. Hussain, M. Shafiq, M. Lehtonen and M. Hashmi, “Online Condition Monitoring

V. M. Shafiq, G. A. Hussain, L. Kütt, M. Lehtonen, “Effect of Geometrical Parameters on

VI. G. A. Hussain, M. Shafiq, M. Lehtonen, “Thermal Monitoring of LV Switchgear Using

VII. G. A. Hussain, M. Shafiq, L. Kumpulainen, F. Mahmood, M. Lehtonen, “Performance

VIII. G. A. Hussain, L. Kumpulainen, M. Lehtonen, M. Hashmi, M. Shafiq, “Signal

IX. G. A. Hussain, M. Shafiq, J. A. Kay, M. Lehtonen, “Pre-Emptive Arc Fault Detection

X. G. A. Hussain, L. Kumpulainen, J. V. Klüss, J. A. Kay and M. Lehtonen, “The Smart

CHAPTER 3 ARC-FLASH PROTECTION TECHNIQUES IN SWITCHGEAR ........... 19

CHAPTER 4 DESIGN AND IMPLEMENTATION OF SENSORS ................................... 33

CHAPTER 5 ON-LINE CONDITION ASSESSMENT OF SWITCHGEAR .................... 45

Switchgear forms an integral part of the distribution network in an electric power

1.2 Aim of the Research

1.3 Research Methods

Conventional method of fault diagnostics relies on offline testing, time based

Due to the availability of advanced micro processing and non-intrusive sensors’

1.4 Scientific Contribution

Following are the major scientific contributions of this dissertation.

1.5 Dissertation Outline

This thesis is organized as follows:

ARC FLASH IN SWITCHGEAR

An arc fault can be divided into two categories:

x Series Arc Faults

2.1.1 Series Arc Faults

Fig. 2.1. Typical series arc fault

2.1.2 Parallel Arc Fault

Fig. 2.2. Typical parallel arc fault

2.1.3 Energy of Arc Fault

2.1.4 Consequences of Arc Fault

2.2 Sources of Arc Fault

x Mechanisms gradually leading to arc ignition

2.2.1 Mechanisms Gradually Leading to Arc Ignition

Fig. 2.3. Classification of electrical contacts

An uninterrupted passage of electric current through an electrical contact can only be

Fig. 2.4. Cyclic process of connection degradation

Fig. 2.6. TEAM stresses

i. Partial Discharge (PD)

The term partial discharge is defined by International Electro-technical Commission

x Voids in solid insulation during manufacturing

PD occurs due to displacement of charge. This produces a rapid rate of change of

Fig. 2.8. Surface discharges

Fig. 2.10. Equivalent circuit of a dielectric with cavity

Fig. 2.12. Electrical mechanism of PD activity

i pulse(t ) A eD2t  eD1t (2.6)

It is already explained in the PD characteristics, that primarily PD causes heat and

x Temporary (sustained) Overvoltage

2.2.2 Immediate Cause of Arc Ignition

i pulse(t ) A eD2t eD1t (2.6)