Sung In Cho

On-Line PD (Partial Discharge) Monitoring of Power System Components

School of Electrical Engineering Thesis submitted for examination for the degree of Master of Science in Technology. Espoo 09.09. 2011

Thesis supervisor: Prof. Matti Lehtonen Thesis instructor: D.Sc. (Tech.) Petri Hyvnen

Abstract

AALTO UNIVERSITY SCHOOL OF ELECTRICAL ENGINEERING

ABSTRACT OF THE MASTERS THESIS

Author: Sung In Cho Title: On-Line PD (Partial Discharge) Monitoring of Power system Components Date: 09.09.2011 Language: English Number of pages: 13+135

Department of Electrical Engineering Professorship: Power systems and High voltage Engineering Supervisor: Prof. Matti Lehtonen Instructor: D.Sc. (Tech.) Petri Hyvnen Condition based maintenance has emerged as a priority issue in modern power systems, and has reminded so for last several decades. Appropriate monitoring and diagnosis before severe faults occur make it possible to control and operate power systems in a more reliable, effective, and sustainable way. Compared other monitoring techniques, Partial Discharge (PD) monitoring seems the most promising methodology for detecting possible dielectric breakdown, aging and ultimately faults in power system components. In order to maximize the benefits of PD monitoring, proper sensing, de-noising and interpretation of its PD signal is of importance. On-line PD monitoring of power system apparatus is a very promising technique that has arisen in the field of condition based maintenance to assist in robust monitoring system which reduces outage time of the power system. For that reason, it is a priority to organize and make a holistic review of current on-line PD monitoring techniques of power system components in order to understand recent developments and trends in theory and in practice. Therefore this thesis is an intensive literature review of current on-line PD monitoring technology. Keywords: Partial Discharge, On-line PD monitoring, IEC 60270, IEC 62478, Pattern recognition, Feature extraction, Conventional method, Unconventional method Code: S-18

Acknowledgements
This thesis was done in the department of Electrical Engineering in Aalto University School of Electrical Engineering in Espoo, Finland in collaboration with Doble Lemke in Dresden, Germany. To begin with, I truly appreciate to my supervisor, Prof. Matti Lehtonen, with his guide and supports during this thesis work. In addition, I also would like to express my gratitude to D.Sc. (Tech.) Petri Hyvnen, instructor, for his guide, advice and encourage. Thanks to Dr. Stefan Kornhuber, engineering manager in Doble Lemke, I can finalize my thesis work in a more fruitful, valuable, and reliable way with his precise comments and critical advice. Moreover it is very important to express my appreciation to the Service team in Doble Lemke and other kind staffs especially for the one who took me to the city centre when I lost my last bus at the first day in the Kesselsdorf. I certainly appreciate my friends so-called Otaniemi Family in Finland who has the same family name, ByungJin, KyungHyun, and EunAh. I am deeply thankful to all brothers, sisters and KOSAFI members who support me all the time and I strongly believe they will be a great designer, engineer, and CEO in the very near future. Lovely Seyoung, without your supports and encourages, this thesis even cannot come into existence! Lovely Thanks to you! Lastly thanks to universal absolute unlimited encouragement from my family during my life in Finland!

Espoo, Finland 09.09.2011

Sung In Cho

ii

List of Abbreviations

PD IEC CBM HF VHF UHF AE UPS HVE GIS TEAM SNR UV OHTL LED DGA SF6 SVM CCD HFCT CNT PRPD 3-PARD

Partial Discharge International Electrotechnical Commission Condition Based Maintenance High Frequency Very High Frequency Ultra High Frequency Acoustic Emission Uninterruptible Power Supply High Voltage Equipment Gas Insulated System Thermal, Electrical Ambient, and Mechanical Signal to Noise Ratio Ultra Violet Over Head Transmission Line Light Emitting Diode Dissolved Gas Analysis Sulfur Hexafluoride Support Vector Machine Charge Coupled Device High Frequency Current Transformer Carbon Nano Tube Phase Resolved Partial Discharge 3 Phase Amplitude Relation Diagram iii

AM/FM k-NN NN BNN PNN PSA DP FDS PDC FRA C&PF C&DF IRA SRA FRA OLTC SCADA RFCT DC AC DAC VLF EHV HV MV LV

Amplitude Modulation/ Frequency Modulation K Nearest Neighbour Neural Network Back propagation Neural Network Probabilistic Neural Network Pulse Sequence Analysis Degree of Polymerization Frequency Domain Spectrum Polarization/Depolarization Current analysis Frequency Response Analysis Capacitance and Power Factor Capacitance and Dissipation Factor Impulse Response Analysis Step Response Analysis Frequency Response Analysis On Load Tap Changer Supervisory Control And Data Acquisition Radio Frequency Current Transformer Direct Current Alternating Current Damped Alternating Current Very Low Frequency Extra High Voltage High Voltage Medium Voltage Low Voltage iv

XLPE PVC EPR TDR FTRC ITRC HVDC PDIV GPS TCP/IP RTU PILC MIND EPR CLX FMC TEV RM UMP MCSA CT SSC WAN VT SA TEM

Cross-linked Polyethylene Poly Vinyl Chloride Ethylene Propylene Rubber Time Domain Reflectometry Frequency Turned Resonant Circuit Inductively Turned Resonant Circuit High Voltage Direct Current Partial Discharge Inception Voltage Global Positioning System Transmission Control Protocol/Internet Protocol Ring Main Unit Paper Insulated Lead Cable Mass-Impregnated Non-Draining paper insulated cable Ethylene Propylene Rubber Continuously Metal Clad Armored Flexible Magnetic Coupler Transient Earth Voltage Rotating Machine Unbalanced Magnetic Pull Motor Current Spectral Analysis Current Transformer Stator Slot Coupler Wide Area Network Voltage Transformer Surge Arrestors Transverse Electromagnetic Wave v

TE TM UI PC RF TF map

Transverse Electric Wave Transverse Magnetic Wave User Interface Personal Computer Radio Frequency Time/Frequency map

vi

List of symbol
Ca Cb
Capacitance of the test object Stray capacitance of the PD source Internal capacitance of PD source Measuring Capacitor Applied test voltage Voltage drop across the PD source Voltage drop across the Rm Measuring capacitor Measuring resistor Grounding switch Lower frequency limit Upper frequency limit Frequency band width Mid-band frequency which can be continuously tuned Noise filter High voltage supply Input impedance of the coupling device Coupling Device Measuring Instrument Series capacitance of the calibrator Calibration factor

Cc Ck U1
U2
U3

Cm Rm
Gs

f1 f2
f

fm
Z U

Z mi
CD MI

C0
k

vii

R0

Reading of the PD instrument Known calibrating charge The probability of appearance for that value xi in the i-th phase The mean value
2

q0

Pi

The variance
1 1

dyi dxi

The differential coefficient before and after the peak of the distribution The average discharge magnitude of the positive half cycle The average discharge magnitude of the negative half cycle The sum value of discharges of the mean pulse height distribution in the negative cycle

xi yi
Qs

Qs

The sum value of discharges of the mean pulse height distribution in the positive cycle

The number of discharges of the mean pulse height distribution in the negative voltage cycle

The number of discharges of the mean pulse height distribution in the positive voltage cycle

inc

The inception phase in the positive or negative voltage cycle The voltage difference between two consecutive pulses The phase difference between two consecutive pulses

T R E

Temperature in Celsius The gas constant (= 8.314 J/mole/K) The activation energy (= 113kJ/mole) Final DP

DPfinal

viii

DP initial
e

Initial DP Eulers number The HV capacitance of the bushing The LV capacitance of the bushing The test inductance (or external inductor) The cable capacitance The cut-off wave length The outer radius of the conductor The inner radius of the conductor The propagation velocity of the signal (30cm/ns)

C1
C2
L C
c

a b

c0

ix

Table of Contents
Abstract........................................................................................................................... i Acknowledgements .......................................................................................................ii List of Abbreviations ................................................................................................... iii List of symbol...............................................................................................................vii Table of Contents .......................................................................................................... x CHAPTER 1 ................................................................................................................. 1 1 Introduction .......................................................................................................... 1 1.1 1.2 1.3 1.4 1.5 2 Motivation................................................................................................. 1 Condition Based Maintenance on Power System ..................................... 3 PD monitoring in power system ............................................................... 5 Thesis Overview ....................................................................................... 6 The aim of the Thesis ............................................................................... 7

CHAPTER 2 ................................................................................................................. 8 PD measurement System ..................................................................................... 8 2.1 2.1.1 2.1.2 2.2 2.3 2.4 3 PD monitoring system configuration ........................................................ 8 Conventional PD monitoring system (IEC 60270) ................................... 9 Unconventional PD monitoring system .................................................. 18 Correlation of conventional and unconventional method ....................... 22 On-line VS Off-line PD measurement system........................................ 25 PD monitoring system in power system ................................................. 26

CHAPTER 3 ............................................................................................................... 28 Sensing and Processing ...................................................................................... 28 3.1 3.1.1 3.1.2 3.1.3 3.1.4 3.2 3.2.1 3.2.2 3.3 3.3.1 Detectable PD signals ............................................................................. 29 Electrical signal ...................................................................................... 29 Acoustic signal........................................................................................ 30 Chemical signal ...................................................................................... 30 Optical signal .......................................................................................... 30 Sensors .................................................................................................... 31 Electric sensors ....................................................................................... 31 Non-electric sensors................................................................................ 34 PD monitoring visualization ................................................................... 35 Phase-Resolved Partial Discharge (PRPD) ............................................ 35 x

3.3.2 3.3.3 3.4 3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.4.6 3.4.7 3.5 3.5.1 3.5.2 3.5.3 3.5.4 3.6 4

Time resolved method ............................................................................ 37 3-Phase Amplitude Relation Diagram (3 PARD) ................................... 38 PD feature extraction and de-noising ..................................................... 38 Noises in PD ........................................................................................... 39 Gating and Windowing ........................................................................... 39 Pulse arrival time difference ................................................................... 40 Digital filter method ............................................................................... 41 Signal processing method ....................................................................... 41 Statistical method.................................................................................... 42 PD pulse shape method ........................................................................... 44 PD pattern classification ......................................................................... 44 Distance classifier ................................................................................... 44 Neural Network (NN) ............................................................................. 45 Support Vector Machine (SVM) ............................................................ 46 Pulse Sequence Analysis (PSA) ............................................................. 47 Signal processing of PD signal ............................................................... 48

CHAPTER 4 ............................................................................................................... 49 PD Monitoring on Power System Components ............................................... 49 4.1 4.1.1 4.1.2 4.1.3 4.1.4 4.1.5 4.1.6 4.2 4.2.1 4.2.2 4.2.3 4.2.4 4.2.5 4.2.6 4.3 4.3.1 4.3.2 Transformer ............................................................................................ 49 Transformer in power system ................................................................. 50 PD types in Transformer ......................................................................... 51 Different diagnosis and monitoring techniques on transformer ............. 52 On-line PD monitoring on transformer................................................... 55 Available products for on-line PD monitoring of transformer ............... 58 Summary and Conclusion ....................................................................... 60 Cable ....................................................................................................... 60 Cable system in power system ................................................................ 61 PD types in cable system ........................................................................ 63 Different diagnosis and monitoring techniques on cables ...................... 64 On-line PD monitoring on cable ............................................................. 68 Available products for on-line PD monitoring of cable ......................... 69 Summary and Conclusion ....................................................................... 71 Rotating Machine.................................................................................... 72 Rotating machine in power system ......................................................... 72 PD types in rotating machines ................................................................ 74 xi

4.3.3 4.3.4 4.3.5 4.3.6 4.4 4.4.1 4.4.2 4.4.3 4.4.4 4.4.5 4.4.6 4.5 5

Different diagnosis and monitoring techniques on rotating machines ... 75 On-line PD monitoring on rotating machines ......................................... 77 Available products for on-line PD monitoring of RM ........................... 79 Summary and Conclusion ....................................................................... 80 GIS (Gas Insulated System).................................................................... 81 GIS in power system ............................................................................... 81 PD types in GIS ...................................................................................... 83 Different diagnosis and monitoring techniques on GIS ......................... 84 On-line PD monitoring on GIS ............................................................... 85 Available products on-line PD monitoring of GIS ................................. 87 Summary and Conclusion ....................................................................... 89 On-line PD monitoring on power system components ........................... 90

CHAPTER 5 ............................................................................................................... 91 Conclusion and Future work ............................................................................ 91 References ................................................................................................................... 93 Appendix 1: CASE STUDY 1 ....................................................................................... 116 Appendix 2: CASE STUDY 2 ....................................................................................... 120 Appendix 3: CASE STUDY 3 ....................................................................................... 125 Appendix 4: Comparison of on-line PD monitoring products for Transformer ...... 129 Appendix 5: Comparison of on-line PD monitoring products Cable ....................... 130 Appendix 6: Comparison of on-line PD monitoring products for RM ..................... 131 Appendix 7: Comparison of on-line PD monitoring products for GIS ..................... 132 Appendix 8: Commercial Sensors .......................................................................... 133

xii

CHAPTER

1 Introduction

1.1

Motivation

In modern society, electricity is regarded as the most important energy source enabling many electric facilities to operate correctly. In order to maintain and sustain those facilities, the quality of power from the grid should be as stable as possible to meet the requirements of electric equipment. Especially nowadays there are many factories and buildings that need a constant power supply to function, and the costs when the electricity fails can be great. In this sense, the appropriate monitoring and protection of power system is one of the significant issues in power system development and monitoring. Even though there has been growing concern about this issue, power systems have remained fairly similar for the last several decades. This has led to catastrophic cascading blackouts occurring several times all over the world in the recent years. These events illustrate the importance of protecting and 1

monitoring power systems which are the most intricate system humans have ever made in history. Compared to many protection methods in power system, Partial Discharge (PD) is considered as one of the most promising solutions for monitoring and detecting possible faults in the system before they occur. Thanks to the development of other engineering areas such as radio communication, computer science and signal processing, protection systems are becoming cheaper and more robust, also high sensitivity. PD is able to find possible symptoms of faults in the system in the most fundamental and simplest way. With IEC 60270 and other standards regarding PD monitoring, PD measurement techniques and calibration had been established with detailed explanations for monitoring purposes. Since direct detection of PD is not possible, conventionally technicians have been using so-called apparent change detection. Whilst traditional methods are detected after failure or discrete periodic interval monitoring, modem techniques are largely dependent on the relative changes of important parameters in time or frequency domain. As a result, Condition Based Maintenance (CBM) has been considered a powerful tool for real-time monitoring on power system components. In order to conduct on-line PD monitoring, the noise to signal ratio is the key variable to determine whether there is PD activity or not. That is the reason unconventional methods for detecting electromagnet PD phenomena using High Frequency (HF), Very high Frequency (VHF), or Ultra High Frequency (UHF) detection and Acoustic Emission (AE) detection have been developed for on-site and on-line PD monitoring being supported by IEC 62478 in near future. Nevertheless, whilst the theory behind of PD monitoring system is the same for different components, the application on power system apparatus such as transformer, switch gear, cable, or rotating machines differ from each other. Therefore, in order to understand PD monitoring from the theoretical to practical, well-organized survey reference will be required. Thus, intensive literature survey of PD monitoring of power system components will be presented as a big picture in the field of on-line monitoring.

1.2 Condition Based Maintenance on Power System

The most significant issue for industrial utilities is the protection of possible faults which usually incur tremendous repair cost and inconvenience to the customer. Even though Uninterruptible Power Supply (UPS) makes it possible to operate electrical equipment in hospital or factories that require a stable and continuous power supply, unexpected power interruption increases the possibility of large scale disaster and cascade blackout. Therefore utilities have been developing proper monitoring system for power system in order to predict and prevent electrical faults before they occur. Largely, there are two considerable reasons for CBM. 1. Maintenance of good operating condition has become a priority for preventing penalty cost and protecting expensive electric High Voltage Equipment (HVE). 2. With technological progress in computer science, signal processing, and radio communication, CBM operating with reasonable price and reliable accuracy has arisen [2]. In order to operate CBM in the power system, the monitoring equipment with proper features should provide adequate information for scheduling of repairs or replacement of HVE. Moreover, this information can be used for life prediction of power system components. In this sense, PD monitoring has gained its reputation as continuous monitoring for condition assessment purpose [3]. In particular, PD monitoring has been widely used in monitoring and the protection of HVE with the following purpose. 1. Design Test: for evaluating and checking a PD-free (or lower than some specific level) system of new insulation system 2. Quality Assurance Test: for confirming that there are no voids or cracks created during manufacturing and processing of the insulation system 3. Diagnostic Test: To determine if the high voltage electrical insulation such as rotating machines, transformers, Gas Insulated System (GIS) and cables have weakened because of any kind of electrical, mechanical or environmental stresses during operation. 3

Figure. 1.1 TEAM stress on Power System Components [7, 8]

Regarding the monitoring insulation system of power system components, there are four main influencing factors affecting the lifetime of the insulation system, known as the TEAM approach; Thermal, Electrical, Ambient, and Mechanical. Indeed, all different types of power system components are influenced by these factors intensively. Therefore for CBM based insulation monitoring, all factors, shown in Figure 1.1, should be taken into account. For example, a rotating machine would be monitored by vibration monitoring, temperature monitoring, electrical monitoring (e.g. partial discharge, dissipation factor, motor current spectrum analysis etc) and chemical monitoring. This kind of approach, of course, is possible for other power system components as well such as transformer, cable, and GIS. Since PD monitoring on power system components has already proven its efficiency compared to other monitoring techniques, continuous PD monitoring will ensure a safer power system with CBM based operation for the following reasons 1. On-line PD monitoring can be used on almost all HVE such as transformer, cable, rotating machine, and GIS with a very similar procedure for each. PD pattern is the only one and universal characteristic parameter in order to evaluate all HVE [4] 4

2. Appropriate signal processing makes it possible for on-line PD monitoring without noises around the monitored equipment. 3. Sensors and tools for on-line PD monitoring are widely available at a relatively reasonable price. 4. Continuous PD information with analysis facilitates possible life prediction modelling of HVE [5]. 5. On-line PD monitoring is possible while the system components are in operation otherwise they need to be disconnected and tested in the laboratory, entailing expensive costs for conducting off-line tests [6]. Since measuring electromagnetic field change is effective even in a noisy environmental and while power components are in operation, on-line PD monitoring on power system components will provide enough information for CBM, operating the power system in a safe, reliable and, predictable way.

1.3 PD monitoring in power system

The application of PD monitoring on different power system components differs in terms of other regarding sensor coupling, detecting method, and so forth. However, in general the PD monitoring behind application is as simple as in Figure 1.2.

Figure. 1.2 General PD monitoring scheme

If there are any PD similar activities on certain HVE, sensing of PD is performed in order to determine the existence or non-existence of PD activity by using sensor 5

placement internal or external of HVE. During sensing, back ground noise signal from different system components can be mixed with PD signals from the examined component. Therefore, noise deduction obtained from the sensors signal generates important PD features in order for a more precise diagnosis. These features have its distinct characteristics so that it is possible to classify them by comparing with prior data from the laboratory or on-site. This process is known as pattern recognition or pattern classification. By doing so, the PD monitoring system finally estimates the possible fault type. Finally, all of this process can be used for life prediction modelling of the HVE. Based on all of the information from PD sensing to life prediction, a more precise PD monitoring system is possible. Moreover on-line PD monitoring based on the above diagram makes it possible for real-time monitoring data analysis, resulting in a robust CBM operation.

1.4 Thesis Overview

The thesis consists of 5 chapters. The first chapter will explain motivations and a general overview of the thesis. Chapter 1 also describes the PD scheme from a wide perspective with a brief explanation about what PD monitoring is. Chapter 2 covers PD monitoring configuration, categorizing the conventional and unconventional methods which have been widely used by experts. In addition, on-line and off-line monitoring scheme with different techniques are covered. Chapter 3 deals with signal sensing and processing. Due to high Signal to Noise Ratio (SNR), promising or currently used feature extraction techniques and classification methods will be introduced. In chapter 4, the application of on-line the PD monitoring system on power system components such as transformer, cable, rotating machine, and GIS is examined. According to each system components, on-line PD monitoring configuration can be different. Lastly, chapter 5 conclude the whole thesis and illustrates the usefulness of on-line PD monitoring systems for life prediction modelling.

1.5 The aim of the Thesis

The objective of the thesis is be a holistic review of existent PD measurement, interpretation algorithms and applications for on-line monitoring of high voltage power system components from a theoretical and practical perspective. The aim is to not only collect the data related to PD monitoring system, but also categorization and discuss of each application and PD monitoring system is presented. In addition, the thesis clarifies the following questions regarding PD monitoring systems: 1. What kinds of methods are currently used to monitor a PD signal in power system components? 2. What kinds of sensors are currently used on different power system components to detect PD and its location? 3. How can the PD signal extracted from a noisy environment for on-line PD monitoring? 4. Based on different sensors, how can fault situation be defined according to the PD signal pattern? 5. In order to make more advanced on-line monitoring of PD, what kind of algorithms and de-nosing techniques are used to the interpret PD signal? 6. What kind of commercially available solutions exist for on-line PD monitoring on different power system components?

CHAPTER

2 PD measurement System

2.1 PD monitoring system configuration

In this chapter, the general PD detection system will be covered. The very beginning of PD detection traces back to nineteenth century by G. Ch Lichtenberg [9]. However in the 1970s with IEC 60270 standard, a lot of practical information was generated and implemented on different power system components such as cables, transformers, and switchgears. Especially the third version of 60270 describes the precise equivalent circuit and calibration method in order to apply PD detection system in a real case. On top of those, many papers have suggested knowledge rule on PD detection due to the fact that only expert engineers can interpret the meaning of PD signal in the right way at the moment.

Along with the growth of condition based monitoring system on power systems, effects have been made to apply PD detection systems in-real time while power system components are in operation. In this sense, it was pointed out that biggest problem of the conventional method with IEC 60270 is the high ratio of noise level per PD signal. Therefore recently different PD detection schemes such as ultra high frequency method, acoustic, optical and chemical detection have been developed to overcome the high level of noise without intricate signal processing. Moreover a new standard of PD detection using electromagnetic and acoustical methods have arisen named IEC 62478 in the near future. For this reason, this chapter will describe general system configuration from conventional PD detection system for apparent charge measurement and unconventional PD monitoring systems.

2.1.1 Conventional PD monitoring system (IEC 60270)

Conventional PD monitoring refers to PD measurement method according to the IEC 60270 standard [10], measuring induced apparent charge in the detection circuit. Since direct detection of partial discharge is impossible physically, this detection method uses recommended test circuits. Even though apparent charge measured by measuring impedance is hard to have a strict constant relationship with real discharge inside of the test object, the linear increase of apparent charge means higher partial discharge taking place. In [10], specific test measuring circuit, quantities, and calibration procedure are covered. This method has been widely used for on-site, laboratory, and commissioning PD measurement due to its accumulated knowledge in terms of measuring configuration and interpretation. In order to apply this method for on-line cases, noises from background or power system grid make it difficult especially on cable and rotating machine. In this section, overall system configuration, theoretical background, and calibration will be presented.

System layout [11-13] In order to comprehend the measuring mechanism according to the IEC 60270, the basic physical rationale behind the measuring system should be understood beforehand. As mentioned already, direct measurement of PD charge value is 9

impossible owing to inaccessibility to the PD spot inside of the test object. The simple equivalent capacitor arrangement of system layout so-called a-b-c model and measuring system is shown in Figure2.1.

Figure. 2.1 Simple capacitive a-b-c model and measuring mechanism [9]

Ca = Capacitance of the test object which is not affected by any PD Cb = Stray capacitance of the PD source

Cc = Internal capacitance of PD source

As we can see, three capacitance values represent capacitance of the insulation system, capacitance in series of PD occurrence, and capacitance of PD respectively. Usually the condition of the capacitances is Cb << Cc << Ca . For calculating indirect charge induced at Ck (measuring capacitor), a-b-c capacitive model with measuring capacitance can be used. This a-b-c model implies the fact that the induced or measurable apparent charge is a small fraction of the real discharge at the spot of PD occurrence by the ratio of the capacitance characteristic, Cb / Ca . Since usually Ca is much higher than Cb , Cb / Ca is much less than 1. Therefore measured discharge with measuring impedance is less than the actual discharge due to attenuation along to the unknown propagation path depending upon the test object itself and insulation structure. This lumped capacitors model, however, can be much more complicated in the case of GIS or a high voltage cable due to the fact that electromagnetic waves from PD propagate through the test object which should be regarded as a transmission line. A more detailed mathematical frame work and explanation regarding the relationship between induced and real discharge is in [14], and [15].

10

Apparent charge [9, 10]

Figure. 2.2 Apparent charge measurement equivalent circuit [9]

U1 = Applied test voltage

U 2 =Voltage drop across the PD source

U 3 =Voltage drop across the Rm

Ca =Virtual test object capacitance Cb =Stray capacitance of the PD source Cc =Internal capacitance of the PD source

Cm =Measuring capacitor

Rm =Measuring resistor
Gs =Grounding switch
The apparent charge measurement can be achieved by connecting measuring impedance on the test object according to IEC 60270. The equivalent circuit of apparent charge measuring is shown in Figure 2.2. The simple mathematical frame work is below in order to calculate qc (the charge created by PD at internal capacitance ( Cc )). Firstly the transient voltage U 3 across measuring device can be obtained

11

U3

U1

Ca (Ca Cm )

(2.1)

Simplification of the equation with the consideration that Cm is much higher than Ca

U3 Cm U1 Ca

qa

(2.2)

Taking into account that Cb Ca , the equation can also be expressed as

qa U1 Ca U 2 Cb
By multiply Ca / Ca , the final equation will be

(2.3)

qa

U 2 Ca Cb Ca

qc

Cb Ca

(2.4)

In other words, above equation describes that the discharge occurred at Cc will causes a voltage drop as U1 which will be transmitted through Cb to the capacitance Ca by the ratio as Cb / Ca .Therefore the measureable charge ( qa ) is a certain portion of actual charge ( qc ) at the PD site due to the fact that Cb / Ca 1. We should note that the measureable charge is proportional to the virtual test object capacitance. Thus, the apparent charge measured by the test coupling device cannot be a direct measure of true PD magnitude, rather it can provide one piece of the important information with regard to the condition assessment of the test object.

Quasi-integration of PD pulse and recommended frequency band [9] The assumption of measuring apparent charge in frequency domain is the linear integration with in measuring frequency band shown in Figure 2.3.

Figure. 2.3 Quasi-integration of PD measurement in frequency domain [9]

12

According to the IEC 60270, the relationship of frequency spectrum of PD and measuring frequency band was covered. Firstly, the integrated part of PD should be assumed as constant within measured frequency band width. Secondly, the upper and lower frequency band cut-off ( f1 and f 2 ) should be lower than measured constant part of PD value. Lastly the recommended gain gap between frequency spectrum of PD and measuring frequency band should be less than 6dB. Recommended frequency band widths in IEC 60270 standard can be categorized wide and narrow band measurement shown below. Wide band measurement Lower limit frequency: 30 kHz< f1 <100 kHz Upper limit frequency: f 2 <500 kHz Frequency band-width: 100 kHz <

f2

f1 <400 kHz

Narrow band measurement Frequency band-width: 9 kHz < f <30 kHz Mid-band frequency: 50 kHz < f m < 1 MHz

f1 =Lower frequency limit f 2 = Upper frequency limit

f = Frequency band width

f m = Mid-band frequency which can be continuously tuned

Coupling mode and device [9- 11] According to the IEC standard, there are two basic coupling modes depending on measuring impedance connection condition. In Figure 2.4, the coupling mode with measuring impedance is connected in series with the coupling capacitor. The Z, noise filter, is used in order to prevent noise coming from the HV side of the test transformer.

13

Figure. 2.4 Basic coupling mode in series with the coupling capacitor [11]

Figure. 2.5 Basic coupling mode in series with the test object capacitor [11]

Z= noise filter U= High voltage supply

Ck = Coupling Capacitance Ca = Test object capacitance Z mi = Input impedance of the coupling device
CD= Coupling Device MI= Measuring Instrument

In Figure 2.5, a similar coupling mode is shown. However this method is slightly different to that of Figure 2.4 in the sense that it can increase the sensitivity of PD detection connected in series with the grounding of the test object which entails the risk for the damage of measuring impedance due to possible high current flow.

14

Figure. 2.6 Polarity discrimination coupling mode [11]

Figure. 2.7 Balanced coupling mode [10]

Additionally, this coupling requires interrupting the grounding connection of the test object that can be done in a special case from a practical point of view. Therefore mostly the measuring impedance connected in series with coupling capacitor has been widely used. IEC 60270 also suggests slightly different connection configuration in order to resist background noise and other purposes shown in Figure 2.6 and 2.7. Polarity discrimination coupling circuit was proposed in order to identify polarity of PD. The logic system performs a comparison of the pulses from two coupling devices (CD, CD1), and gate those signals for polarity correction of the pulses. The balanced coupling mode, in Figure 2.7, can eliminate external electromagnetic noises by adjusting impedance of Z mi , and (Zmi )1 with an amplifier. Even though the balanced circuit can reduce certain amounts of noise, practically the coupling mode in Figure 2.4 is the most popular in which the measuring impedance and coupling capacitance are connected in series.

15

Figure. 2.8 Coupling device described by IEC 60270 [9]

Ck = Coupling Capacitance

Ca = Test object virtual capacitance

Rm = Measuring resistor Lm = Shunt inductor

Cm = Measuring capacitor

Conventional coupling device, according to the IEC 60270, consists of measuring impedance, signal filtering, high voltage protection part and so forth. Measuring impedance is the main components of a coupling device to deliver output voltage pulse converted from input PD signal. The signal filter can screen interferences caused by test voltage. The high voltage protection part is for suppressing damage from over voltage which can possibly occur in the case of a breakdown of the test object. A possible detailed circuit configuration is presented in Figure 2.8.

Calibration [9, 10] The calibration procedure circuit recommended at IEC 60270 is shown in Figure 2.9. The basic idea for calibration of a PD measurement system is the injection of a known pulse which can be detected by a coupling device, then a scaling of the measurement system for estimating real PD magnitude and finally a calculation of the calibration factor (k) which is the ratio between measurable apparent charge (qa ) and the reading 16

of the PD instrument ( R0 ) . The relationship of the series capacitance of the calibrator ( C0 ), test object ( Ca ), and coupling capacitor ( Ck ) can be expressed according to the IEC 60270 as shown in below.

C0

0.1 (Ca Ck )

(2.5)

Commercially available calibrators inject a known pulse ( q0

C0 U 0 ) with certain

time intervals connected near the coupling device shown in Figure 2.10. This can also ensure the connection of the whole measurement system. The following equation can simply explain how to calculate the calibration factor.

qa

q0

k R0

(2.6)

Figure. 2.9 Calibration circuit recommended by IEC 60270 [9]

Figure. 2.10 Calibration graph with 2000pC calibrator (LDJ-5)

Conventional PD measurement system IEC 60270 is regarded as a proven technique for performing PD measurements by many experts, utilities, and in the academic field. Since conventional PD monitoring with IEC standard 60270 has its strength by cumulated knowledge, references and knowhow to detect PD in power system apparatus, it can be applied to all kinds of 17

power system components. The third version of IEC 60270 presents detailed information from a coupling device to the calibration method as seen above. Even though this method is vulnerable to noise and other interferences, the biggest advantage over unconventional PD measurement system is the availability of the estimated magnitude of PD. A recent paper [16] pointed out some fundamental limitations of the conventional method with three points; integration error in case of non-linear, possible superposition error, calibration limits, and unknown attenuation of PD signal from PD spots to sensors. Those challenges tackle the advantages of the conventional method in terms of accuracy of the measurement system. Nevertheless, IEC 60270 has been widely used as an application for new power system components testing and commissioning, on-site measurement, and laboratory tests for periodic examination. For on-line application, calibration procedure and high signal to noise ratio makes it difficult to apply the IEC 60270 method. However transformer application such as multi-terminal measurements and GIS application for sensitivity verification have sometimes been combined with the unconventional method which will be covered in the upcoming section.

2.1.2 Unconventional PD monitoring system

Unconventional PD measurement was developed for GIS application several decades ago as a form of UHF PD detection system [17]. Since then, several other PD measurement techniques have been introduced beside the conventional PD monitoring method by using other indirect indicators of PD occurrence, which includes electrical (HF/VHF/UHF), acoustic, optical, and, chemical measurements [18, 19]. The unconventional method, in particular, has better characteristics regarding signal to noise ratio for on-site or on-line measurement of power system apparatus. Some of those techniques (Electrical/ Acoustic) will be standardized in the near future by IEC 62478. In this section, so-called unconventional PD measurement technique will be covered.

18

Unconventional PD detection methods [20] Electrical detection [21]: Electromagnetic measurement of PD consists of coupling devices and data acquisition unit. The most suitable frequency band for application regarding each power system components are shown in Table 2.1.
Cable HF (3 - 30MHz) VHF (30 300MHz) UHF (300M 3GHz) Transformer GIS Rotating Machine

O O

O O

O O -

Table 2.1 Suitable frequency band according to system components (O=Good, =OK, -=NO)

Appropriate sensors and its placement on test object detect electromagnetic signal. Detection of electromagnetic transient signal from PD occurrence is usually performed by capacitive or inductive sensors. More detailed information regarding system configuration, sensor type, and placement according to each system components is covered in chapter 4. The main advantage of this method is its accuracy and accessibility of the information about intensity, source, and possible fault type. However electrical interference during measurement is the main disadvantage.

Acoustic detection [22, 23]: Some measurement installations of PD are affected by severe electrical interferences that are difficult to control. However acoustic signal from a PD source is immune from electromagnetic noise. An acoustic signal from mechanical vibration of PD can be detected by piezoelectric transducers, fibre optic acoustic sensors, accelerometers, condenser microphones and sound-resonance sensors usual using frequency band as between 10 kHz and 300 kHz. AE detection has been successfully used in order to localize 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. Detailed acoustic wave propagation characteristics are shown in [24]. This method is very efficient for localizing PD source because of its immunity against electromagnetic noise. 19

Optical detection [25, 26]: Optical emission from PD can be detected by optical sensors. Unlike electrical signals from PD, optical signals largely depend on different factors such as insulation material, temperature, PD intensity and pressure. The spectrum of hydrogen or nitrogen depending on the surrounding material is the most dominant concerning the spectrum of PD. There are roughly two kind of optical PD detection techniques as a result of different kind of ionization, excitation and recombination processes during the discharge; direct detection of optical PD signal and detect of change of an optical beam. Detection of optical signal includes surface detection and the detection inside of the test object such as GIS and transformer. For cable application, corona emits the spectrum range around 280nm to 410nm at high voltage transmission line which can be detected by a UV-visible camera during the daytime. The rationale behind this is the ultra violet radiation ranging from 240nm to 280nm tends to be absorbed by the ozone layer. The optical sensors transferring signal to the outside at photomultiplier, also can be placed inside the test object which is efficient for a light-tight GIS impulse voltage test. This impulse voltage test is not suitable for an electrical PD detection system. Another method called opto-acoustic measurement catches sonic or ultrasonic range acoustic emission caused by PD which results in deformation of the optical fibre. One recent paper [27] describes optical PD detection on Over Head Transmission Line (OHTL) using fibre optic sensors with Light Emitting Diodes (LEDs), resulting in meaningful PD detection capability. The main advantages of this method are the immunity from electromagnetic interferences and high sensitivity compared to conventional electrical techniques.

Chemical detection [28-30]: Chemical PD detection on HVE is one of the most popular and simplest methods. In particular, PD activities in oil or gas insulated object can react chemically, emitting a by-product of the chemical reaction. For transformers, the most relatively used method is Dissolved Gas Analysis (DGA) with Duval Triangle diagnosis. For GIS, SF6 gas analysis with detecting its compound has been used. Even though this method indicates an abnormal condition of HVE, it provides only a rough condition assessment without any specific data regarding its intensity, source or location.

20

Electrical Applicative for all kinds of HVE Intensity, source, type, location of PD is assessable The most suitable for continuous online PD monitoring High electromagnetic interference Relative expensive cost

Acoustical High sensitivity Immunity against electrical noise Very efficient for localization of PD Relatively low cost

Optical Immunity against electrical noise High sensitivity Location of PD is assessable(in some case) Test is possible for impulse voltage condition

Chemical Immunity against electrical noise Easy to measure Provide critical information for Go/No Go decision

Advantage

Disadvantage

Low signal intensity Not good for continuous PD measurement Piezo-electric transducers Condenser microphones

No information about magnitude of PD

No information about location, source, intensity, and type of PD

Possible Sensors

Capacitive Inductive

Optical fibre UV detector photomultiplier tube

DGA Sensors SF6 Sensors

Main applicative area

All HVE

Transformer GIS

Cable, GIS Transformer

Transformer GIS Cable

Table 2.2 Different unconventional PD detection methods

Sensitivity and performance check Unconventional PD detection systems cannot verify apparent charge by calibration procedure. However there are sensitivity checks and performance checks in order to ensure validity of nonconventional PD measurement. By doing the sensitivity check, the measurement system obtains required sensitivity in a worst-case way with IEC 60270. Meanwhile performance check evaluates the function of whole measuring path including sensors and the acquisition system which can also be used to find the appropriate frequency band in on-site measurements [31]. Theoretical approach of sensitivity check on cable and transformer is covered in [32] and [31, 33, 34] respectively. In the GIS case, the sensitivity check is possible to have enough sensitivity that is equivalent as that of IEC 60270 shown in [35, 36].

21

Unconventional PD measurement system Unconventional PD measurement is much more suitable for on-site and on-line PD measurement in which the external interferences largely influence the measured signal. Especially electromagnetic wave and acoustic detection has been widely used in the field since these two methods simply provide sufficient information concerning the existence of PD and its possible location covering almost all kinds of power systems components. As seen below in the Table 2.3, possible on-line application of different system components can be realized by nonconventional PD measurement systems.
Cable Acoustic Electromagnetic Optical Chemical Transformer GIS Rotating Machine

O -

O O O

O O -

O O -

Table 2.3 Possible on-line PD detection techniques on power system components [37]

Since the most interference for on-site or on-line PD monitoring is in the lower frequency band, higher frequency monitoring within HF/VHF/UHF band has a good signal to noise ratio. Supported by IEC 62478 in the near future, nonconventional PD measurement will be used widely within a better frame work. The main disadvantage of the unconventional method is that the measuring method depending upon test object differs from each other. Because of that, the monitoring system covering all HVE will be expensive compared to the conventional method. Moreover most unconventional methods are not possible for calibration providing magnitude of PD which might introduce mistakes in terms of decision making.

2.2 Correlation of conventional and unconventional method

Even though the conventional and unconventional method measure different physical quantities, there has been some research regarding comparison and correlation of their measurement results. Those studies include the PD pattern, linearity of measuring quantity [38, 39, 40]. However so far, finding solid correlation between the two methods seems to be very difficult due to the fact that the results from both methods largely depend on the condition, sensor type, sensor location, manufacturer of test 22

object, test engineer and so on. In particular the nonconventional methods have not been supported by standard, resulting many different test set ups regarding higher frequency and other energy detection from PD occurrence. Standardization will bolster the analysis of the correlation of the measured quantities from both methods. On the other hand efforts have been made to combine the two techniques in order to overcome each drawback. In particular a combined solution is effectively applicative on transformer and GIS. This kind of integrated approach can detect PD occurrence with accuracy and scalable quantity in a low noise environment. In this section, correlation of the two measuring systems and its combining approach will be covered.

Conventional versus nonconventional PD monitoring Fundamentally, PD measurement systems according to IEC 60270 and

nonconventional methods are measuring different quantities, apparent charge and electromagnetic waves or others, even if it comes from the same source. Some questions have arisen regarding the correlation between the two different methods and interpretation of results [21]. The general comparison is shown below in Table2. 4.

Conventional Main Standard IEC 60270 Measuring impedance (the sensor for conventional method can be capacitive, inductive-HFCT or Rogowski coil)

Unconventional IEC 62478 (standard draft) Electric sensors Acoustic sensors Optical Sensors Chemical Sensors HF (3MHz-30MHz)* VHF(30MHz-300MHz)** UHF(300MHz-3GHz)*** AE (20 kHz to 250 kHz, and 100 Hz to 3 kHz) Sensitivity check Performance check Amps, mV, V/mm or dB

Sensor type

Frequency band

Wide (30-500kHz)/ f=100-400kHz Narrow(50kHz- 1MHz)/ f=9-30kHz

Calibration Measuring unit

Must be calibrated Usually pC,

23

Measuring quantity

Apparent charge

Transient earth voltage or current pulse ( Electromagnetic wave) Acoustic, Chemical by products, Optical spectrum Sensing components, transmission path, data acquisition unit Relatively low Off-line and on-line On-line (Electrical, Chemical)

Measuring system

Coupling device, transmission system, measuring instrument Relatively high Mostly Off-line (Laboratory, Onsite) On-line (Transformer)

Noise Level Application type

Table 2.4 Comparison of conventional and nonconventional method

*typical narrow band width for HF/VHF is 2MHz **Typical wide band range is 50MHz or higher ***Zero span mode for individual frequencies or for specific frequency range between 4 and 6MHz or higher

Combining of conventional and nonconventional method Several recent papers such as [41] have described the possibility of combining the two PD measuring methods whilst taking advantages from each. Successful applications have been conducted on transformer and GIS. In Figure 2.12, combined measurement configuration on a transformer is presented. On the one hand, injected UHF sensors at the oil valve can detect electromagnetic waves inside the transformer. On the other hand, measuring impedance on three bushings measures apparent charge. If there is no detection on UHF with high level of apparent charge from the IEC 60270 method, the PD source can be considered to be external. On the contrary, meaningful electromagnetic detection with high level of apparent charge measurement result can imply the fact that PD occurs inside the transformer [42]. On top of that, this multiterminal measurement on each bushing ensures possible PD location by analyzing each bushings apparent charge magnitude [43]. For GIS, the conventional IEC 60270 method can be used for sensitivity verification of the UHF/AE method in order to have enough sensitivity for instance 5 pC.

24

Figure. 2.11 Example of combining PD measurement methods on Transformer [29]

2.3 On-line VS Off-line PD measurement system

In this thesis, on-line PD monitoring means the system with following requirements PD measurement while the test object is in normal operation Continuous PD monitoring (Trendable) Permanent installation of PD coupling device Without any other voltage sources except for operating power from the grid Under the same circumstance as the normal operating condition such as temperature, pressure, humidity and so on

On the other hand off-line measurement has the characteristic:

PD measurement while the test object is out of connection from power grid Installation assessment and new high voltage equipment test Test voltage should be applied Inception and extinction voltage can be found

Seemingly both on-line and off-line PD measurement can be quite similar each other. However it is fundamentally different from its configuration to measurement results. The main disadvantage of off-line PD measurement is that problematic PD occurrence 25

sometimes cannot be detected using the off-line method because it is carried out in different circumstance to that of real cases such as load condition, vibration, temperature, humidity and so on. That means the test object which passes for off-line PD test can have potential failure in the power grid. This method, moreover, is expensive due to outage during PD measurement. However off-line PD measurement usually has high sensitivity and accuracy because of relatively low back ground noise and is very suitable for new equipment quality control. For on-line PD measurement, on the contrary, the measurement is very realistic because it performed under the real circumstances. The cost is relatively less expensive and it is possible to have trendable data for the test object, meaning that the life cycle management can be possible with on-line PD monitoring. The main research ongoing in the on-line PD monitoring field concerns signal processing due to high noise combined with a true PD signal. However recent papers and commercially available on-line PD measurement systems ensures effective on-line PD measurement with appropriate signal processing techniques.

2.4 PD monitoring system in power system

PD detection system has proven its efficiency as the most promising monitoring tool. Commercially available PD measurement systems either conventional or

unconventional can provide appropriate information according to its application on different test objects. Even though different PD measurement systems concern their specific signal generated by PD, they surely indicate any abnormal condition of power system components. The IEC 60270 method is already a proven technique in laboratory or on-site measurement. On the other hand, the recent trend of PD measurement systems is towards on-line monitoring which should provide PD occurrence in real time while the machine is in operation. Therefore unconventional PD monitoring techniques has been introduced and widely used. Since the conventional method is the only one measuring the magnitude of PD relatively accurately, correlation between unconventional methods with IEC 60270 can ensure linearity of PD measurement with true PD occurrence in the test object. Thus, solid correlation which can possibly be achieved by sensitivity verification for 26

UHF/AE technique can determine the concrete status of the test object. However different measuring configurations of UHF/AE make it difficult to have strict linearity and correlation. In this sense, draft IEC 62478 can assist to clarify the promising UHF/AE detection configuration and make it more robust in the near future.

27

CHAPTER

3 Sensing and Processing

In this chapter, signal processing analysis will be covered. The raw signal from the sensor needs to be processed in order to erase back ground noise and take meaningful features. This process can be called Feature extraction. Possible methods such as statistical, pulse shape and digital signal processing technique will be presented in this chapter. After being extracted, those features will be used for classification to find out the PD type based on prior knowledge from expert data analysis. Among classification techniques the most promising solutions such as distance classifier, neural network (NN), Support Vector Machine (SVM) and Pulse Sequence Analysis (PSA) will be shown.

28

3.1 Detectable PD signals

Partial discharge is detectable in a different way due to the fact that it generates certain reactions according to the insulation materials in the system components. Generated signals from PD are usually detectable in an electric, acoustic, chemical, and optical way [44]. Electrical and chemical signals are referred to for finding out PD occurrence in high voltage equipment, and acoustic signals are used to localize the spot where PD takes place. Depending on the characteristic of the power system components, appropriate signal detecting can differ. Nowadays, combining of the methods guarantees more accurate PD detection. In [45], a more physical approach to the PD mechanism is presented.

3.1.1 Electrical signal

PD occurrence in the power system equipment makes the electrical signal. That is because partial discharge brings about electron transfer in a short current impulse within nanoseconds [1]. As described in the Chapter 2, in order to detect electrical signal, there are two different types of measurement setup required. The first one is so-called apparent charge measurement that detects induced charge in the test circuit. The second one is to detect electromagnetic radiation using radio a frequency antenna or probe [1]. The biggest drawback of electric signal is the high noise to signal ratio due to electrical radiations from other equipment. Because of this, the electrical signal detection method needs more complicated signal processing techniques compared to other methods. On top of that, owing to the complicated structure of power system components, the signal can be attenuated or modified. However the electrical signal method is popular for PD detection thanks to highly sensitive sensor, analyzer and digital oscilloscope. Especially in the laboratory where there is relatively low noise compared to on-site, electrical signal detection is more advantageous for PD monitoring. Therefore this method is widely used for qualifying new power system components before installing, and with signal processing tools after installed on-site. For on-line PD monitoring, the electromagnetic wave (HF to UHF range) detection is a promising method compared to apparent charge measurement because of the possible immunity characteristic against background noise.

29

3.1.2 Acoustic signal

Even though electrical signals are the obvious evidence of PD occurrence, acoustic signal generated from the mechanical wave of a small explosion around the spot where the PD takes place is widely used for PD monitoring [46]. The biggest benefit of acoustic signal is the immunity from electromagnetic interferences [47, 18]. Moreover acoustic detection is not an intrusive method compared to other measurement types [46]. In addition, the acoustic signal detection method is favoured for localizing PD in the test object. By using several acoustic sensors on the object which have PD occurrences inside, the computation of travelling time difference from each sensor provide geometric information of PD location [48]. However even though acoustic signals represent is against electrical interferences, acoustic noise or mechanical vibration from other high voltage equipment can affect acoustic signal strength.

3.1.3 Chemical signal

Partial discharge also creates a chemical reaction with the insulation material. One common usage of chemical signal detection is on the oil-insulated transformer. PD in oil-insulated equipment caused chemical reactions releases, for example, carbon oxide, hydrogen, or methane. DGA has been widely used for detection of possible faults initiated by partial discharge. According to the gas generated by chemical reaction with PD, information about the type of PD or involving insulation material is available [49]. Therefore, this method is not appropriate for on-line or on-site PD monitoring. This method thus has been performed in the laboratory periodically. Chemical signal detection, moreover, does not determine the location of PD [47].

3.1.4 Optical signal

PD activities emit radiation in ultraviolet, visible, and infrared optical signals. The spectrum of light emission depends on the surrounding insulation material such as oil or gas. In particular, optical spectral diagnostics of the electromagnetic radiation for wavelengths between 10 nm to 30 mm is of interest. Photomultiplier or Charge-

30

Coupled Device (CCD) Cameras can detect optical signals with relatively higher sensitivity in air tight test objects such as GIS.

3.2 Sensors
In this section, the sensors used regarding PD detection are covered. Currently there are many sensors which have been used depending upon the measuring method and test object. Since the sensor plays an essential role in PD measuring configuration, appropriate selection and its location can affect the measurement result significantly. The basic requirements of PD sensors are below [52] Be able to sense and record measuring quantities from PD source for a set of defined frequency bands Can differentiate between PD signal and background noises Small enough in order to be attached to the test object

The sensors traditionally detect PD below 500kHz due to technical limits and lack of standardization. However for over ten years, higher frequency detection using a variety of sensors which can be internal or external according to the application has become attractive and applicative for all kinds of power system equipment [53]. Detailed applications on power system components of each sensor will be presented in chapter 4. Here a general specification of widely used sensors will be included.

3.2.1 Electric sensors

HFCT (High Frequency Current Transformer): This sensor is one of the most popular inductive sensors for all kind of applications on power system equipment due to its portable, cost effective, non-intruding characteristic and the independency of the frequency of the measured signal [54]. Using a ring type of ferrite core, the basic structure of HFCT consists of six or seven turns of copper wire over the ring core. Ferrites being ferromagnetic Ceramics with very high resistivity and permeability are most attractive materials for high frequency applications [55]. HFCT has especially been used in order to couple for ground rod or cable. The closed core or split core

31

version of HFCT is commercially available as shown in Figure 3.1. The HFCT detect PD up to several hundred MHz.

Figure. 3.1 Commercially available core closed and split type of HFCTs

Rogowski coil [56, 57]: The Rogowski coil is a proper sensor for PD working on the inductive principle with frequency bandwidth between 1 to 4 MHz. The Rogowski coil has a structure of a circular plastic mold with a winding mounted for a uniformly distributed density of turn with frequency dependant characteristic. By mounting around conductor, Rogowski coil generates induced voltage signal as an output [58]. The followings advantages of Rogowski coil are: Very high band width. Capability of measuring large current Non-saturation due to air cored structure Ease of use due to possible thin and flexible clipped around a measured conductor-Non-intrusive Very good linearity due to absence of magnetic materials.

Epoxy-mica encapsulated couplers [59-61]: This type of coupler is the most popular sensor especially for transformer and rotating machines. The epoxy-mica encapsulated coupler contains the capacity against the conductor and a stray capacity. Commercially 80pF up to 2nF epoxy-mica coupler has been widely used. The desirable frequency band can be achieved by the PD and noise ratio and winding frequency characteristic for rotating machine case. The main short coming is that the 32

capacitors has to be designed in order to withstand 60 Hz high-voltages, and it 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 that the pulse signals are usually much larger because they can be placed closer to PD spots. The PD activity in each phase, moreover, can be determined.

Figure. 3.2 Commercially available Epoxy-mica encapsulated couplers

Antenna type coupler: For unconventional PD detection in the higher frequency range (HF/VHF/UHF), antenna type sensors are widely used in a different shape shown in Figure 3.3. Since there are many practical constraints for sensor installation, practical antenna design can differ depending on the application. The external mounted and internal oil-valve type, for instance, on transformer has proven to be proper UHF PD detector. In particular, internal oil valve type sensors with as conical shape have the most sensitivity [62].

Figure. 3.3 Commercially available UHF type sensors

33

Currently Doble lemke (DN 50/80), and Omicron (UVS 610) uses this kind of sensors on their UHF PD measurement for power transformer. Externally mounted UHF sensors first developed as a GIS application and it has been widely used as transformer UHF detection as well [63]. Detailed information about various UHF antennas such as horn, loop, and, dipole type for GIS applications is described in [54]. Directional coupler [21, 60, 64-65]: The directional coupler is a combination of a capacitive with an inductive sensor. It is possible to use two directional in a cable joint. By doing so, it is possible to distinguish PD impulses coming from outside (left or right side) or inside the joint. In other cases, depending upon the direction of pulse, energy can be coupled to a different output port in case of special sensors with two outputs. The main application of this type of sensor is using a cable joint. For cable joint application, a directional coupler can achieve high sensitivity. Typical operating frequencies are usually from several MHz up to GHz.

3.2.2 Non-electric sensors

Fibre optic sensor: Detection of acoustic signal from a PD source can also be done using a fibre optic sensor. In other words, ultrasonic signal generated by PD in high voltage equipment can be successfully identified by fibre optic sensors which have been proven by some papers [66]. This also detects optical signal covered in chapter 2 in detail. Ultra or ultrasonic signal produces the pressure on the optical fibre which can be sent as light in the fibre [67]. Dissolved Gas Analysis (DGA) sensor [37, 68-69]: Usually the DGA technique is used for periodic sampling of oil from a transformer. However recent DGA sensors detecting chemical by-products such as hydrogen, carbon monoxide and so on have been used as an indicator of PD in the transformer. The gas sensors can be applicative as on-line monitoring of oil based insulated high voltage equipment as shown in some papers. However the sensors only provide information regarding Go/No go decisions. SF6 carbon nanotube sensor [70, 71]: Carbon nanotube (CNT) is a new material which has unique physical and chemical characteristic. Since the conductivity of CNT depends on the atomic structure and chemical absorption, unidentified oxidative decomposition products generated by PD can change electrical conductance of CNT 34

by increasing positive hole density in a p-type semiconductor. This sensor is currently only used in an academic field but its use has been shown in some papers. However based on this sensor, off-line PD detection in GIS is possible. Piezoelectric transducers [72]: The sensor is typically operating in the frequency band in the 120160 kHz range. In order to minimize the varying response according to the electromagnetic fields, the transducer can be either a differential type utilizing two crystals or a shielded single crystal transducer with an integral pre-amplifier circuit. Usually an integral pre-amplifier circuit type is the more common configuration due to high amplitude and low impedance output. Since the acoustic impedance of a sensing crystal differs from as that of the steel transformer wall, an efficient hard-epoxy resin material is used with thermal and electrical isolation characteristic. Commercially available acoustic detection for PD localization which is applicative for transformer has been successfully used.

3.3 PD monitoring visualization

In order to analyze the PD signal, visualization of the signal is of importance due to the fact that appropriate display pattern visualization of PD signal has great advantages. The trend of PD analysis is based on computer aided solutions [73]. In this section, the base of PD pattern visualization is covered

3.3.1 Phase-Resolved Partial Discharge (PRPD)

PRPD proposed in the late 1970s. This method is the most popular among almost all commercial PD measurement systems and has proven to be one of the most powerful tools to interpret PD signal [73]. As the name implies, PD signal is shown with respect to the test voltage as its phase resolved spot in Figure 3.4

35

Figure. 3.3 PRPD patterns as pulses and pattern (PD-Smart)

The most relevant information shown in PRPD is the measured PD signal with pulse magnitude, the phase angle at which PD occur, and the number density [74]. Because PRPD simply shows the most relevant quantities of PD, PRPD analysis of each measurement has played an important role to identify possible fault types on specific measured test objects [75]. The most commonly used distributions are below [41]: Number of PD pulses detected in each window plotted with respect to the phase position Average discharge magnitude in each window plotted with respect to the phase position Peak discharge in each window plotted with respect to the phase position Average discharge current in each window plotted with respect to the phase position Therefore the distributions and relationship of peak and average PD magnitude, phase angle and the number of repeated rate enhance simple PD pattern recognition. 36

However, this PRPD pattern of each measurement cannot entail complete identification of fault type because it depends on the PD measurement unit, sensor, frequency band, test object and multiple causes of overlapping faults. Because of that, there are some cases the typical patterns of PRPD do not match the true cause of PD [76]. In order to increase accuracy of PRPD match with true fault causes, the same measuring configuration and reference of each test object are required. A more sophisticated display of PRPD in 3D in terms of PD amplitude, cycle number, and phase position is shown in [77]. Pattern analysis and recognition based on PRPD will be introduced in the on feature extraction and classification section.

3.3.2 Time resolved method

PD display based on measuring time can be called time resolved PD data shown in Figure 3.5. Since this visualization focuses on more on the timing of PD occurrence, time resolved data can provide information on the location of PD with several sensors placed at different spots rather than PD magnitude. In [78], time of flight calculation based on time resolved PD pattern at GIS is presented in detail in chapter 4. Other applications of time resolved data is a Q-T diagram which uses the time between two consecutive discharges shown in [79]. Time versus frequency analysis (TF map) conducted in [80] is the analysis methods on time based PD measurement clustered by a fuzzy logic classifier which has been realized by Techimp.

Figure. 3.4 On-line Time resolved PD pattern with terminal voltage of generator (PD-Smart)

37

3.3.3 3-Phase Amplitude Relation Diagram (3 PARD)

3-PARD, or a star diagram, is cross talk between more than one phase on each measurement [43, 81-83]. So called multi-terminal measurement, measuring 3 phases with three couplers, can acquire synchronous PD data for all three phases of the test object such as three phase transformer or GIS. This method make it possible to compare the magnitude of PD occurrence on each phases, helping locate PD source occurring in perhaps one of the three phases and eliminating external noise shown in the display. The 3-PARD is a plot with a 120 phase shift of the three phase axis shown in Figure 3.6. This method has been developed by the Technical University of Berlin.

Figure. 3.5 3-PARD comparing PD magnitude on each phase [29]

3.4 PD feature extraction and de-noising

The biggest problem for PD measurement is noise. In the case of on-line PD monitoring especially, there is a lot of different noise which can cover a true PD signal by a high noise signal level. Therefore, features of true PD from the measured signal with noise is very critical for identifying PD occurrence and further classification of the fault type. Several noise types can be successfully caught and reduced by signal processing or other methods. Thus, PD feature extraction is the process which detects true PD data in order to obtain the characteristic of PD activities possibly classified as different faults type by a classification process. In other words, the purpose of feature extraction is to reduce dimensionality of true PD pattern with calculation of certain features or properties of the pattern [84]. In this section, de-noising methods which provide a feature vector without considerable noise will be covered.

38

3.4.1 Noises in PD
Detecting of a true PD signal from measured results is a matter of in-depth knowledge and incremented experience on measured signal and noise characteristic in different situations and test objects. Since PD activities in power equipment occur within less than a few hundred nanoseconds as fast rising time which is low level pulse depending on faults type of the test objects, the de-noising process can be achieved by understanding the noise characteristic and eliminating them from the true PD signal. Typical noise during PD measurement can be categorized [85, 86]. Sinusoidal noise: This type of noise is the narrow band noise signal such as communication carrier signal from AM/FM modulation which can be removed by applying, for instance, a digital filter. Pulse type (repetitive or random) noise: This type of noise possibly comes from power electronics, other switching operations or, Radio Frequency (RF) emissions from power equipment. Even though repetitive noise can be rejected by a gating circuit and other method which can detect periodic noise against PD signal, random pulse type noise is hardly eliminated. White noise: white noise can be referred to some random signal with flat spectrum density. This type of noise can be detected and removed by several signal processing techniques which will be covered in this section.

3.4.2 Gating and Windowing

Those de-noising techniques filter suspicious signals using gate antenna or manual windowing control. The basic principal of the gating method is detecting noise directly using gate antenna and those signals can be reduced by a gating signal shown in Figure 3.7. The most advantage of this method is its simplicity of installation for on-site PD measurement using gate antenna which are available from some commercial PD measurement products. Windowing is the simplest method for noise elimination. By applying phase window on the repetitive or suspicious PRPD patterns, the signal located on the designated phase window will be removed as shown in Figure 3.8 [87].

39

Figure. 3.6 Principal of gating method for noise reduction

Figure. 3.7 Windowing applied for PRPD pattern (PD-smart)

3.4.3 Pulse arrival time difference

This method applicative for on-line PD measurement uses two sensors for one measurement located a certain distance at least 2 meters away.

Figure. 3.8 Configuration of two coupler installation for noise elimination

40

Those two couplers detect signals at different spots with time difference for the same PD signal which can be from the test object side not from a grid. Thus, by comparison of pulse arrival time on two couplers, one can distinguish noise from the grid side. The basic scheme is shown in Figure 3.9 [1, 88].

3.4.4 Digital filter method

When PD is corrupted by noise caused by radio communication, a matched filter is a very well-suited as a solution. First of all, a matched filter can make it possible to maximize SNR of PD by suppressing noise. In addition, it can make accurate estimation on the time of arrival and magnitude of maximum PD pulse. The time of arrival of PD pulse and SNR are deeply related as those two variables are inversely proportional. Simply a matched filter uses a template which is a prediction of the shape and amplitude of a PD pulse. The coefficients should be determined in order to construct a specific matched filter for a specific measurement. One solution for this, for example for the cable case, is the injection of a known pulse and measuring its pulse propagation characteristic as impedance. By calculating time-of-arrival, the PD localization in cable can also be achieved which is realized by KEMA for cable online PD monitoring [89-92].

3.4.5 Signal processing method

Even though the simplest signal processing method is Fourier transform of PD signal, it has the main drawback that it loses time resolved information. Therefore the wavelet transform which might be the most popular signal processing method applied for PD measurement has been successfully used [73, 85, 93-95]. The wavelet transformer, moreover, allows one to obtain the information regarding time domain and frequency domain with its amplitude at the same time. Lots of research and papers have described wavelet technique as continuous or discrete with multi resolution . For on-line application of wave transform which shows non-stationary characteristic of the data, an adaptive wave transform can be applied. Despite the fact that on-line wavelet application is challenging because of difficulties in terms of selection of mother wavelet and resolution, threshold level, it can be used for on-line 41

PD measurement, reducing noises and extracting a very small amount of data from actual measurement [96]. The basic steps of wavelet transform applied for noise reduction are described below. Decomposition: set a mother wavelet and a maximum decomposition level, computing the wavelet decomposition coefficients at each level from 1 to N. Thresholding: Compute threshold coefficient for each and apply threshold to the coefficients at each level Reconstruction: Reconstruct the signal with the modified coefficients from 1 to N

3.4.6 Statistical method

Statistical methods for extracting PD features are based on PRPD pattern [73, 97-98]. By applying statistical computation on PRPD patterns, different distributions can be characterized as statistical parameters. The following distribution functions are used. Skewness: shows the asymmetry or degree of tilt of the data of the distribution compared to a normal distribution.

SK

( xi u )3 Pi
3

(3.1)

where xi is the measured value, Pi ( f ( xi ) /

i 1

f ( xi )) is the probability of appearance

for that value xi in the i-th phase window, u (

2

xi Pi ) is the mean value, and

( xi u ) 2 Pi ) is the variance.

If the measured PRPD pattern is symmetrical skewness will be close to zero. For the asymmetrical distribution to the left, skewess will be higher than zero, otherwise it will be less than zero. Kurtosis: shows sharpness of the distribution compared to a normal one.

KU

( xi u )4 Pi
4

(3.2)

If the measured PRPD pattern is shaper than the normal distribution, kurtosis will be higher than zero, in a flatter case, it will be less than zero. 42

Number of peaks: represents the distribution with single peak or more. The peak of the distribution can be defined as:

dyi dxi
Where

1 1

0,

dyi dxi

1 1

(3.3)

dyi dxi

1 1

is the differential coefficient before and after the peak of the distribution.

Cross-correlation factor: shows correlation of the distribution shape between positive and negative cycles of the distribution.

cc [

xi yi xi2 ( xi ) 2 / n]

xi [

yi / n yi2 ( yi ) 2 / n]

(3.4)

where xi is the average discharge magnitude of positive half cycle and yi is the that of negative cycle. When cc is close to zero, it means the shape of positive and negative cycles are the same, otherwise it will be asymmetrical. Asymmetry: shows the comparison of the mean level of the positive and negative half of the voltage cycle.

Qs / N Qs / N

(3.5)

Where Qs and Qs are the sum value of discharges of the mean pulse height distribution in the negative and positive voltage cycle, and N and N indicate the number of discharges of the mean pulse height distribution in each cycle. If the asymmetry is very close to zero, it means the mean level of each distribution is the same size. When the mean level of distribution on positive cycle is more than that of negative cycle, the asymmetry is close to -1, otherwise it will be close to 1. Phase factor: defines the difference in the inception voltage in the negative and positive half of the voltage cycle which can be expressed as:
inc inc

(3.6)

Where

inc

indicates the inception phase in the positive or negative voltage cycle. 43

3.4.7 PD pulse shape method

This method is based on time resolved PD data for instance, apparent charge and voltage magnitude within a certain time interval due to the fact that different PD source can generate different PD pulse shape [73, 99]. The features extracted on an one to one basis using single discharge source. The Following parameters can be used Pulse rise time: time required to rise from 10% to 90% levels of the peak value. Pulse decay time: time required to decay from 90% to 10% levels of the peak value. Pulse width: time interval between 50% levels on both sides of the peak value. Area under pulse: area enclosed by the q-t curve in the time interval for 10% levels in the rising and falling segments.

3.5 PD pattern classification

Many researchers and theses have studied the pattern classification of PD. Therefore many different methods have been introduced in order to understand and trace of certain PD pattern such as artificial neural network, fuzzy logic, genetic algorism, and support vector machine [100]. Most of them require prior knowledge with respect to feature vectors of PD measurement. Based on analysis of reference and history, new PD patterns can be classified into one of the typical PD types which might suggest the source and reason of the PD signal. This classification can help decision makers of the system in order to determine go or no go for certain power system components [97]. In [101], promising techniques and algorithms of computer science, neural computation, information theory and statistics which can be used as classifier for PD patter is introduced. Among many, some proven classifier here is presented

3.5.1 Distance classifier (k-NN)

A distance classifier is an efficient and simple method for classification [73, 102]. The basic idea of the distance classification is based on the fact that similarity between the measurement features presented as points in the Euclidean space is determination of their closeness. k-NN (k-Nearest Neighbour) based on minimum distance is one of the most promising classification technique by determining the number of neighbours (k). 44

The optimal number of neighbours depends on the data. Thus if there is new data coming to the feature space, it is classified by major voting of k-number closest neighbours of the new data spot. This also can be a drawback because certain types of classes with the more frequent examples tend to dominate and are highly possible to be selected. In order to overcome this problem, the class should be weighted by experts or based on experience. The mathematical explanation is in [103]. The advantage of this classification is easy to update new data to reference and, it is simple to implement because it do not require training. However if redundant features concerning the classification are included, possible errors can occur [104]. Therefore careful selection of the feature is of importance.

3.5.2 Neural Network (NN)

Artificial neural network has been applied for PD classification [73, 97, 105-108]. The basic idea of NN is based on biological neural functions taken from brain-like problem solving. The basic structure of NN consists of three mutually connected different types of layer, an input layer, hidden layers, and output layer shown in Figure 3.10.

.
Figure. 3.90 Structure of Neural Network [108]

The input layer has several input neurons fed by different values of features extracted from PD patterns such as statistical features of PRPD or time resolved PD pulse pattern. The hidden layer is to extract classification information from the data and the output later is defined according to user expectation showing final classification of a PD pattern. Among different NN types, back propagation neural network (BNN) and probabilistic neural network (PNN) seem to have good characteristics for PD 45

classification. Details of both methods are presented in the above references. Although NN is a very efficient tool for PD pattern classification especially due to the fact that it does not requires any assumption the PD data structure, it has several drawbacks including; dependence of convergence criteria upon learning coefficient such as the number of layers, learning time; and it is also difficult to include new features which requires retraining.

3.5.3 Support Vector Machine (SVM)

Support vector machine is one of the most promising techniques works by using outstanding learning algorithms especially in power systems such as load forecasting, power stability, and fault location detection [100, 109-110]. The main idea of SVM is to calculate the optimal hyperplane separating two classes. SVM uses the so-called non-linear kernel trick. SVM can find the solution of non-linearly separable condition using an implicit mapping technique into a high dimensional dot-product space called the feature space through the use of the kernel trick. A detailed explanation of the kernel method is shown in the above references. Despite the sophisticated procedure for calculation of kernel function, the advantage of this SVM is its proven efficiency, accuracy, and acceptable processing speed as classification. Some recent theses consider SVM as the best tool for classification of PRPD pattern at the moment [81, 109].

Figure. 3.101 Basic idea of SVM describing optimal kernel function to separate class and mapping of input to high dimensional feature space

46

3.5.4 Pulse Sequence Analysis (PSA)

Pulse Sequence Analysis (PSA) proposed by Martin Hoof and Rainer Patsch in 1990s is one of the most popular techniques for visualization of PD pattern classification [111-113]. The idea of this method is that two consecutive pulses caused by PD activities have a strong relationship. This means that the previous PD pulse has an impact on the condition of next pulse. Therefore analysis of the relationship of continuous pulses of voltage change due to the corresponding change of the local electric field at the PD spot is an important factor which can investigate correlations between consecutive pulses as shown in Figure 3.12.

Figure. 3.112 Basic principle of PSA

u = The voltage difference between two consecutive pulses

= The phase difference between two consecutive pulses

The following quantities are interesting in PSA Discharge amplitude Pulse position (related to the phase of the sine wave) Absolute cycle number (related to measurement activation) Instantaneous voltage at each pulse

47

Figure. 3.123 Example of PSA in GIS; surface and corona discharge [113]

The advantage of PSA is its clear differences between certain PD patterns due to the physical characteristics of PD activities according to the source of PD. However if the voltage differences of continuous PD activities cannot be defined from measurement, PSA is hard to apply

3.6 Signal processing of PD signal

Since measured PD signal has a very low magnitude happening with nano-second duration, it needs appropriate signal processing which can reveal the true PD pulse and its characteristics that can be used in order to interpret PD measurement in the right way. Technically speaking, de-noising and extraction of a meaningful feature vector from measured data is of importance for further classification. For the sake of a hidden indication of PD, applied signal processing should yield diagnostic information describing PD intensity, source and location for deciding optimal operation and repair schedule of power system components.

48

CHAPTER

4 PD Monitoring on Power System Components

4.1 Transformer
This chapter is to provide all relevant information regarding on-line PD monitoring on power transformer including other methods used for diagnosis and monitoring of transformer briefly. The first section covers transformer insulation characteristics, different PD types and detection methods, containing other possible off-line protection and monitoring techniques such as DGA. The second part will show PD detection systems which have been used for the last decades from conventional method to the latest detection systems concerning sensors specifications as well as recommended location, coupling methods, possible calibration techniques and data signal processing for each system. Finally some commercially available on-line PD detection systems and up-to-date trends will be covered

49

4.1.1 Transformer in power system

The transformer is one of the most complicated structured components in the power system. Normally most transformers operate efficiently for between 20-35 years, which can be extended with proper maintenance [114]. Moreover, even though the failure rate is quite low about 0.2-2% a year [115], it usually causes cascading faults on different system components. Therefore, appropriate maintenance based monitoring while in operation is the key point for preventing transformer failure. Transformer insulations and its characteristics are also a bit complicated compared to that of other components. The most common insulation material in transformer is mineral oil which is being replaced by environmentally friendly oil and cellulose [116]. In [33], failure rates according to the transformer parts are tap changer (41%), windings (19%), tank and oil (13%), terminal (12%) and so forth. Another statistical survey for transformer rate is shown in [117]

Transformer structure and failure rate

Components Core Tank Dielectric material The expansion tank Description Path for magnetic permeability between primary and secondary winding. Case of the transformer including the dielectric material, the core, and the windings or casing Fluid oils, gases, or dry solids which have poor conductibility and good characteristics for electrostatic fields Container for dry air or dry inert gas to maintain the fluid level. an insulating structure to insulate unexpected electric path from the grids or other electric transmission devices from the tank of the transformer insulator between the coils and between the coils and core connection point along a transformer winding allowing voltage regulation by selecting desired the number of turn ratio dissipation device for the internal heat generated in the transformer a protection device for the tank against excessive pressure release inside a transformer tank. Table 4.1 Main components of the Transformer

Bushing

Pressboard paper barriers The tap changer The radiator and fan The pressure relief

50

Continuous PD monitoring on a transformer According to the CIGRE data [117, 118], the tap changer has the highest possibility of failure and then leakage, winding etc. That means appropriate healthy monitoring of transformer can prevent possible failure beforehand. One recent paper demonstrates on load tap changer monitoring using a continuous DGA method [119]. When it comes to PD monitoring on transformer, it can be categorized as electrical, acoustical, and chemical detection [44]. Regarding the electrical signal detection method, both IEC 60270 and UHF detection is widely used. Due to the complexity in transformer and its bulky volume, electrical sensors can be mounted outside on a bushing (IEC 60270) or in side of the transformer using the oil drain valve (UHF). For locating PD course inside of the transformer, the acoustic emission method is used to calculate the time difference between different sensor placements [120, 121]. Chemical detection has been widely used in a periodical way with techniques such as DGA or Furan analysis.

4.1.2 PD types in Transformer

In some papers [4, 23, 122], there are different types of PD in the transformer which are distinguishable and largely categorized as void, floating part, surface and corona discharge. Void: If there are any an air bubbles within any hardware equipment or crack in the solid part of the transformer such as between windings or insulation paper and oil, avoid type defect can occur. Surface discharge: The surface discharge is the discharge between two parallel dielectric surfaces. In the transformer, this kind of discharge can happen because of the bubbles on the insulation surface or delaminating layers of the pressboard. Corona discharge: Corona discharge is the discharge between a sharp point and plane surface. Any particles from the manufacturing stage can generate corona discharge. Also, the sensor coupling part on the transformer bushing can generate corona type discharge which in this case can be classified as noise Floating part: Basically in the transformer, two different conducting parts that have different potential can generate a floating part type discharge due to capacitive 51

coupling. The main reason behind this discharge is a bad earth connection in a transformer

4.1.3 Different diagnosis and monitoring techniques on transformer

Among many monitoring techniques, this section only includes the on-line applicative monitoring method and compatible with continuous PD monitoring on power transformer. In this section, monitoring techniques are categorized as oil testing, electrical, mechanical, and thermal monitoring of transformer. In [123-125], more detailed transformer diagnosis and monitoring techniques are covered

Oil testing Oil is one of the widely used insulation materials for transformer. An Oil test is carried out by analyzing gases produced by local thermal stress or partial discharge taking place in the insulation liquid during abnormal operation. Therefore, gas analysis as part of the oil method is widely used for detecting electrical thermal insulation problems in transformer.

Dissolved Gas Analysis (DGA) DGA has been proven to be the most powerful and reliable tool to find out any incipient faults in oil-immersed transformer by detecting concentrations of gases that tend to be dissolved in oil. Although there are different gases depending upon different fault and insulation liquid, roughly Nitrogen (N2), Oxygen (O2), Hydrogen (H2), Carbon dioxide (CO2), Carbon monoxide (CO), Methane (CH4), are the main gases of faults [49, 126]. Moreover, those gases can indicate the source of a fault such as corona, overheating, and arcing in the oil [114]. Duvals triangle for DGA analysis is an efficient tool, which works by comparing the ratio of key gases and confirming the source [127]. Usually this procedure has been widely used periodically in the laboratory by sampling oil from the transformer with an occasional time interval. Recently utilities have started to use on-line monitoring of DGA using sensors [126, 119]. According to [116], on-line DGA is too expensive to use except for very high 52

MVA rating transformers and a portable detector is not so precise compared to that of one in a laboratory. However there have been many studies on this method including combining artificial neural network and expert knowledge [126].

Furan Analysis and Degree of Polymerization (DP) When the paper insulation in transformer lose the insulation strength, furanic compounds that are by-products from paper insulation material appear in the oil, which can be analyzed and used for paper aging prediction and DP. 2-furaldehyde is considered the main product of aging, initiated by 5- furaldehyde in the early stages [128]. Furan analysis is applied in the case of high level of thermal stress,

overloading, detection of high levels Carboxide, or sudden changes in oil color and moisture content rates in the oil [114]. Life estimation of transformer according to the DP is shown in [129]. The Constant K is defined as
E R (T 237)

Ae

(4.1)

T = temperature in Celsius, R = the gas constant = 8.314 J/mole/K E = the activation energy = 113kJ/mole A=the coefficient is obtained depending on operating conditions With constant K, life of transformer can be calculated as

1 DPfinal

1 DPinitial

k life

(4.2)

DPfinal = Final DP

DP initial = Initial DP
DPfinal , DP initial can be substituted as 200 and 1000 respectively [130]. Therefore, the
life of transformer can be obtained by combining (4.1) and (4.2)

53

Estimated life of the transformer =

0.004 A

13600

( T 237 )

(4.3)

Thermal monitoring Thermal monitoring is a widely used method and has possible on-line applications. High temperature means abnormal condition in any parts of a transformer losing electrical dielectric strength if the thermal continues without any maintenance or appropriate remedy actions. Usually thermal spots indicate possible faults and insulation failures caused by overloading or local overheating which can accelerate insulation aging rapidly. Because the transformer is complex equipment which has non-linear characteristics with different components such as winding, load tap changer, and core, thermal monitoring are not so precise to pinpoint the exact failure spots which may be inaccessible to an external probe [116]. Infrared scanning check of the external temperature on the transformer is now available [114]. One of the disadvantages is that this method costs a lot in order to sense temperature directly using fibre optic [116, 2]

Electrical Monitoring Electrical monitoring techniques of transformer have been widely used such as Frequency Domain Spectrum (FDS)&Polarization/Depolarization Current Analysis (PDC), loss factor, resistance of winding or insulation, FRA, Transfer function, Partial Discharge, Response analysis, Leakage Reactance, Capacitance and Power Factor (C&PF) and so on. The techniques inspect the dielectric characteristic of the insulation material which is usually oil and cellulose in the case of transformer. C&PF which is known also as C&DF (Capacitance & Dissipation Factor) have been used for measuring capacitance distribution in the transformer which can be a barometer of dielectric constant in the transformer [131]. FDS/PDC determines insulation humidity, tangent delta, and the polarization index. Response analysis use different excitation function such as Impulse Response Analysis (IRA), Step Response Analysis (SRA), and Frequency Response Analysis (FRA). This method monitors transformer behaviour depending on the input signal which it is possible to use for on-site measurement. PD monitoring of transformer is also popular monitoring method. 54

Mechanical Monitoring Mechanically, On Load Tap Changer (OLTC) is the part where many faults occur. Moreover, winding and core vibration can be detected by vibration sensors on the transformer wall. This vibration signature can be analyzed by Fourier or Wavelet transform. For visual inspection, checking of the pump isolation valve and oil flowing indicator should be performed in order to confirm oil circulation. Plus, the conservator breather also should be checked for the correct oil level. Fan and radiators should be kept clean in order to cool the transformer down.

4.1.4 On-line PD monitoring on transformer

On-line PD monitoring on transformer mostly uses the electrical detection method to decide PD occurrence, and the acoustic detection method to locate the PD source inside the transformer. Especially before installation, PD monitoring can be used for new transformers in order to find any possible manufacturing problems [132]. In this section, promising PD monitoring techniques using different methods, sensors, recommended sensor coupling methods and signal processing will be covered.

Conventional method [4, 9, 43, 133] PD monitoring using IEC 60270 is an already proven method and has been used widely for several decades. The sensors used in this method are capacitive coupling devices attached on bushing like the below Figure 4.1. Wide or narrow band pass coupling devices can be installed on a all accessible terminals such as HV or LV bushing, grounded neutral, grounded core clamps. However, bushing with capacitive couplers on the HV bushing or HFCT on a ground lead are the most common methods. Especially multi terminal measurement method can generate good results regarding each phase of PD activities. The results from the multi terminal method can also be used for pattern recognition of PD on transformer by using a 3-PARD diagram, then reducing external noises and finally comparing three terminal PD measurement results with each other. Since the signal from this method usually includes a high level of

55

noise, appropriate signal processing techniques will be required for continuous on-site PD monitoring.

Figure. 4.1 IEC 60270 recommendation for PD monitoring system on bushing

Ca =The test object capacitance

C1 =The HV capacitance of the bushing

C2 = The LV capacitance of the bushing

MI=Measuring Instrument Unconventional method [21, 37, 43, 133, 135] One obvious advantage of UHF PD monitoring on transformer is its strength despite noise. There are two different coupling methods widely used for UHF PD monitoring on transformer. The first method is to use inner sensing type sensors in the oil drain valve which is a non-destructive coupling accessing the PD location more closely seen in Figure 4.2. However this sensor type is not applicable for transformer which does not have a straight-through oil valve. The second method is to install external sensors against dielectric windows at different places on the transformer surface as shown in Figure 4.3. Both UHF detection techniques are currently used and have proven the efficiency as highlighted in research. In some papers, multi terminal IEC 60270 with UHF seems to be the most promising techniques in order to have a sufficient PD signal above noise and to Figure out PD activities at each phase which can assist in the localization of PD.

56

Figure. 4.2 Drain valve type UHF sensor [43, 136]

Figure. 4.3 UHF dielectric window type [137]

AE Method [121, 138-139] Locating the PD source in the transformer is possible using acoustic emission detection on the transformer. There are two different methods; using an electromagnetic PD signal analyzed as a PD pulse shape and amplitude, and; acoustic detection using known as triangulation. Acoustic detection for localization by using piezoelectric sensors on the transformer wall and calculates the arrival time from different sensor placements.

Figure. 4.4 Acoustic detection for localization of PD in Transformer [121]

57

In Figure 4.4, possible sensor place using Cartesian coordinates is shown. The biggest problem of the AE detection method for localizing the PD source in the transformer is its signal sensitivity. This method should measure acoustic signal at the same time with at least 3 or 4 different sensors in different positions. In [121], detailed mathematical explanations and possible signal processing techniques are covered.

4.1.5 Available products for on-line PD monitoring of transformer

Doble Lemke Doble Lemke GmbH uses a conventional and unconventional method for on-line PD monitoring of the transformer. In conventional PD monitoring, tap bushing coupling with a low voltage capacitor is used as a sensor. This method is also applicative for multi terminal measurement analyzed by a 3-PARD diagram eliminating noises and comparing each phase. For the noise reduction, gating from a gate sensor and the winding technique for phase-locked noise is used. In unconventional PD monitoring, oil drain valve type sensor in the UHF (300 MHz-1 GHz) band or a UHF tap hatch sensor is used. Furthermore, they use acoustic emission detector to localize the PD source inside of the transformer by using piezoelectric, acoustic sensors attached on the transformer wall which provide up to 8 sensors

Dynamic Ratings Dynamic Ratings provides a combined solution for transformer monitoring such as on-line DGA, and temperature monitoring. Regarding on-line PD monitoring, they use external sensors such as Radio Frequency Current Transformer, bushing sensors and a Rogowski coil. An AE sensor is compatible if it is necessary.

IPEC Limited IPEC Limited provides on-line PD monitoring equipment applicable to transformers. Sensors are HFCT, Capacitive coupler, and Airborne Acoustic Transducer. Total monitoring solutions can be combined with temperature and humidity detection. 58

Power diagnostix System GmbH Power diagnostix System GmbH uses the conventional method using capacitive tap on the bushing and the PRPD visualization method.

PowerPD, Inc. PowerPD, Inc uses electrical sensors (clamp-on type HFCT) and acoustical sensors on the transformer wall for on-line PD transformer monitoring. The sensitivity of the sensors is 5pC and 20pC respectively. This system is fully compatible with SCADA and remote accessibility.

Qualitrol Company LLC Qualitrol Company LLC uses an unconventional PD monitoring method by attaching rod type, window type, or drain valve type and hatch installation UHF sensors from three to six around the transformer orthogonally on the side and top wall. This method provides digital and analog output for web based or SCADA (supervisory control and data acquisition) monitoring applications. This method can be applied in many transformers at the same time with a Master/Slave connection for each signal collection box.

Techimp Energy Srl Techimp Energy Srl uses a HFCT, inductive sensor, usually clamped to a ground connection of transformer. In order to make it a non-invasive way of coupling, they use TEV (Transient Earth Voltage) and Horn antenna as sensors. This company uses fuzzy logic based PD identification technique and Time-Frequency map for noise reduction and identification of PD type.

59

4.1.6

Summary and Conclusion

Since transformer is the most intricate power system component, there are many different ways or monitoring techniques for preventing possible faults. As some companies have already provided on-line PD monitoring system on power transformer for a couple of years, one can infer the fact that on-line PD monitoring of transformer will be widely used in the very near future. Especially transformer application PD monitoring techniques can be combined with other chemical, mechanical or thermal monitoring with the other methods mentioned in this section. On-line PD monitoring on the transformer focuses preliminary on PD magnitude (peak value) and source location. Regardless of the apparent charge or UHF measurement, changing or increasing of PD magnitude inside the transformer means the fact that the transformer needs a more specific inspection or to be repaired. However, for the purpose of on-line monitoring, the UHF method is more reliable due to the strong resistance to back ground noise. For the localizing of the PD source, acoustic emission detection technique has the key solution of locating PD source inside the transformer as highlighted in many papers. From a practical point of view, the on-line monitoring method uses capacitive sensors, UHF sensors (oil drain valve type or dielectric window), or HFCT as the correct sensor type. Capacitive sensor application is compatible with multi terminal sensing which makes it possible to compare PD signals generated from each of the three phases. It can also be used for further signal processing and, reducing phased-locked noise. Nevertheless IEC 60270 has played an important role in guiding PD monitoring on Transformer. The upcoming standard for UHF/AE, IEC 62478 will be the most important standard especially for on-line PD monitoring of power transformer.

4.2 Cable
This chapter presents On-line PD monitoring for cable applications. There are different reasons for aging of cable including thermal, electrical, mechanical, and environmental. Based on these reasons, there are many different techniques used to monitor, and diagnose the faults [140]. Regarding electrical aging monitoring, PD has 60

been widely used in the laboratory, on-site as the form of on-line or off-line monitoring. Especially after installation of the cable system in the power system, detecting faulty connection by different PD monitoring methods such as Damped AC (DAC), Very Low Frequency (VLF) for example have been gaining its reputation. Therefore, in this section, all kinds of PD monitoring techniques in cable will be covered with detailed information regarding on-line PD monitoring in the cable system as well as its available products in the market

4.2.1 Cable system in power system

Cable network systems in the power system are one of the most important part but also the part most vulnerable to failure. Cable network can be categorized as Extra High Voltage (EHV), High voltage (HV), Medium Voltage (MV) and Low Voltage (LV) networks. The failure rate of the cable system is more frequent for lower voltage networks, meaning LV networks have the greatest outage time of all network. More than half of cable failure stems from electrical reason and the rest of them are due to external non-electrical inference [132]. In particular, in MV networks, the causes of outage time are the cable (81.1%), Switchgear (6.8%), transformer (3.8%) and others (8.3%).

Cable network structure and insulation characteristics The cable system in the power system consisted of cable joint, termination, and line. For insulation purposes, Cross-linked polyethylene (XLPE) is commonly used for HV and MV cables. Poly Vinyl Chloride (PVC) and ethylene propylene rubber (EPR) are suitable for LV cables [141]. XLPE, in particular, is popular due to its low dielectric losses. XLPE nowadays has been replacing oil-immerged insulated cables. Regardless of operation voltage or frequency, the cable usually has the same structure as described below in Table 4.2.In Figure 4.5, the typical XLPE cable structure is shown.

61

Figure. 4.5 XLPE cable structure [142]

Components Conductor

Description Transfer current at lowest loss, usually made by Cupper (Cu) or Aluminium (Al) Contribute smooth and homogeneous boundary surface, and for preventing any gaps or voids occurrence Endure electrical stress, single layer construction Electrical shielding, creates return path for capacitive charging current, provides mechanical stress when the cable is bending Provide mechanical strength, and low moisture penetration Table 4.2 Cable structure and its function [143]

The inner and outer semi-conductive screens Insulation layer

Earthed metallic screen

Protection sheath

Cable accessories are one of the main reasons for possible faults in the network system. According to [132], cable joint and terminations are the biggest reason for cable network failure. Because of this most tests are focused on these parts. In [144], more detailed defect types and causes are demonstrated.

62

Continuous PD monitoring on cable network Traditionally, PD monitoring on cables has been widely used due to its effectiveness in CBM based monitoring and in localizing of the faults area. Especially for PD monitoring of cables, standard procedures such as Very Low Frequency (VLF), Damped AC (DAC), Alternative Current (AC) or Direct Current (DC) testing are popular due to the fact that they can verify possible faults areas of joint and termination after assembling by detecting and localizing PD in the cable. However, for the purpose of on-line monitoring, high attenuation of the PD signal along the long cable line makes it difficult to pick the exact PD and its location. Nowadays on-line PD monitoring of the cable has been used by sensing the PD signal with HFCT, capacitive coupling sensors, and so on. More detail will be covered up in the coming section.

4.2.2 PD types in cable system

PD occurrence in the cable system can be divided into an internal, surface, and electrical tree [143, 145]. Internal PD: PD occurs in the air gaps or voids surrounded by solid material it depends on the size and the location. This PD occurrence in cavity can pit and erode the cable surface [146]. Surface PD: This type of PD can occur on the surface of solid-solid and solid-liquid material parallel with the surface of insulation. This can be happen as consequence of field enhancement on the area of missing outer semi-conductive screen or an incompletely removed outer semi conductive screen in the cable Electrical Tree PD: This PD type represents the PD generating tree-like shape on the insulation or dielectric body where the PD occurs. The growth of electrical tree can ultimately bridge undesired electrical paths potentially leading to complete breakdown. Moreover, if there is a high level of moisture, a water tree can be generated resulting in an electric tree at the same place. Usually electrical tree have many branches, and produce a higher PD magnitude than in the cavities which can grow until final breakdown occurs. 63

Corona: PD occurs in open air around the cable.

4.2.3 Different diagnosis and monitoring techniques on cables

There are many ways to monitor cables in a laboratory, on-site, or with on/off-line methods. After a brief explanation of different monitoring techniques used on cable networks, this section focuses on electrical, especially partial discharge method. As well as methods presented here, there are also destructive methods such as Cable sampling, lead sheath analysis, and paper analysis [147].

Tangent Delta (Loss angle, or Dissipation Factor testing) Measurement This method provides information regarding the aging of a cable by determining the loss factor due to the tangent delta value which is related to the composition of the connection, the trajectory, and the actual cable temperature. In perfect conditions, a cable has capacitive characteristics maintaining the phase difference between voltage and current at 90 degree. However, if there are defects in cable, the angle between voltage and current is no longer 90 degrees; rather it will be less than usual. However, this method is not used for XLPE cables owing to their low Tangent Delta value [147]

Leakage current monitoring This monitoring method measures leakage current from a high voltage cable to the ground by the surface of the insulator. The measured value is able to demonstrate pollution issues of the cable and its accessories. For example, washing cable accessories on the cable tower can influence leakage current which can be measured with sensors and optical fibre. This method has also been commercially available in a form of on-line [148, 149]

Temperature monitoring [150-152] Temperature monitoring on the power cable is an efficient tool for detecting an abnormal condition which is also applicative as an on-line monitoring tool by using 64

appropriate temperature sensors. Semiconductor type sensor or optical fibre is popular for continuous temperature monitoring of cables. The advantage of this monitoring is to have real-time thermal behaviour of the cable that it is possible to use for thermal rating re-assessment. However almost all cable systems are in operation practically at low load condition for most of their service time, making it impossible to calculate effective thermal resistivity of the cable. Therefore, temperature monitoring on a cable should focus on a particular time and section of the cable.

Partial discharge monitoring Even though thermal stress has a significant impact on the aging mechanism of the cable, electrical stress is prominent cause of aging. PD monitoring of the cable is the most effective method that is able to monitor electrical aging [153]. For localization of PD, Time Domain Reflectometry (TDR) which uses the reflection of pulse signal at the cable termination [154, 155] has been used. PD monitoring of the cable system can be clearly categorized into the off-line and on-line method. Regarding the off-line method, it has been widely used with an extensive voltage withstand test in order to validate the acceptance test for cables from the factory. It can also be used for on-site PD measurement. Before describing those techniques, a comparison of those the on and off line method are described.

a. On-line versus off-line cable monitoring Because of the structure characteristic of cables, on-line monitoring of the cable network is demanding due to signal attenuation depending on the cable length. Therefore, offline PD monitoring on-site or in the laboratory has been widely used. Especially multiple PD source in long cable line with relatively high background noise from different power system components makes it extremely difficult for carrying out on-line PD monitoring in cables. In many practical fields, both on-line and off-line is widely used for on-site or laboratory tests [156]. The Table below provides general comparison about on-line and off-line PD monitoring for the cable.

65

On-line monitoring

Off-line monitoring

Advantage

Can be performed while cable is in operation Economical Real operation condition can be taken into account.

Proven technology for on-site, laboratory test, and commissioning High sensitivity Calibration possible

Disadvantage

Low sensitivity Complicated data analysis is required Insulated earthing ground is required

Out of connection is required Relatively bulky equipment required Outage cost PD occurrence can be differ compared to its operation at service voltage Overall condition during testing (Temperature, humidity, vibration) can differ from operation condition

Table 4.3 On-line versus Off-line PD monitoring on Cable [157, 158]

b. On-line PD monitoring An on-line cable PD monitoring technique has proven its efficiency recently. As mentioned above, on-line PD monitoring on the cable network has many advantages over the off-line method. However monitoring long cable lines while they are in operation has too much noise and attenuation compared to other application cases. Usually HFCT around cables or the earth connection, inductive couplers, capacitive coupling sensors, and acoustic emissions have been used [159]. Due to the fact that high frequency from a PD signal is significantly attenuated in the cable, the sensors measure the HF/VHF range rather than UHF.

c. Off-line PD monitoring [160-162] The off-line method usually energizes the cable network and monitors PD occurrence by using a different voltage input. Roughly five methods have been used shown as below according to the applicative situations. Turned resonant test voltage method: There are two ways of carrying out this test using different test circuits for resonant testing; Frequency Turned Resonant Circuit

66

(FTRC) and, Inductively Turned Resonant Circuit (ITRC) with the following resonant equation (4.4) in cable.

1 (2 LC )

(4.4)

L= test inductance (or external inductor) C= cable capacitance The FTRC method uses power electronics converter generating harmonics and noises in the test system. Therefore appropriate signal processing techniques are required. However, there is no moving part included in this method. On the other hand, because ITRC usually use auto transformer, there are no such electronic pulse noises. Moreover voltage can increase smoothly which makes it easier to reach the PD inception voltage. The drawback of ITRC is its moving components which should be maintained periodically. Damped AC (DAC) voltage method: This method consists of a direct voltage source, switch, and inductance. After charging the direct voltage source enough for required peak value, the switch connects inductance with the test cable so that the capacitance of the cable and external inductance are able to oscillate. The frequency range of this method can differ depending on the cable length and PD inception voltage and occurrence at the cable. Advantages of Damped AC methods over others are relatively less power demand, lighter, and they is also applicable for all types of MV and HV cables. Very Low Frequency (VLF) voltage method: Since capacitive power of the cable is proportional to the frequency, if frequency decreases, the demand of capacitive power also decreases. Therefore VLF method turns down the frequency as from 0.01Hz to 0.1Hz for extruded-dielectric cable. Advantages of this method are also its light weight and low power demand. DC voltage method: Even though the DC voltage method cannot represent ac voltage related insulation stresses, this method has been used for thermal, conductivity problems. HVDC can be applied for cable acceptance test for recommended duration. 67

Advantages of this method are its simplicity, lightweight and cost effectiveness with low power required Impulse voltage method: Impulse voltage with a very fast rate of rise and decay rate similar to power frequency can be applied for on-site tests. This method has its strength owing to lightweight equipment. Disadvantages of this method are hard to determine the inception voltage of PD, high attenuation along the long cable length, distance dependent test results, and difficulty to find correlation between routine factory and on-site test regarding partial discharge values.

4.2.4 On-line PD monitoring on cable

IEC 60270 method is not appropriately applicative for on-line PD monitoring on cables. Usually HF or UHF detection for gaining high signal to noise ratio (SNR) is an attractive method for this purpose [163-169]. Since cable terminal and joint is the part of cable most vulnerable to failure, on-line PD monitoring on cable accessories is important for cable monitoring.

Figure. 4.6 Capacitive coupling method near cable joint and terminal [166]

Regarding sensor type, capacitive coupler, inductive sensor (e.g. HFCT, Rogowski coil), and directional coupler sensor near cable joint, terminal or cable earth have been widely applied. In Figure 4.6, a possible capacitive coupling method is shown. Since capacitive sensors have good sensitivity for nearby PD occurrence usually near terminal and joint, capacitive sensors are located near cable accessories.

68

Figure. 4.7 HFCT coupling application at the cable termination [168]

In Figure 4.7 HFCT in a slightly different location at the cable terminal is shown. According to the availability, the coupling spot can be adjustable. In order to localize the PD source in the cable system, dual sensor techniques (installing two sensors at each end of cable or cable joint) are required. Because of strong attenuation along the cable, PD localization requires more engineering techniques such as the pulse injection method, GPS application, or TDR. The pulse injection method injects periodic pulses from one side and the sensor located in other side detects the pulse which synchronizes two sensors at each end of the cable system. Therefore the propagation time and transfer impedance can be calculated. Another technique uses TDR due to the symmetrical characteristic of the cable system the pulse can propagate toward both ends of the cable with different magnitude and time. Therefore the direct pulse and reflected pulse can be detected by sensors which can be synchronized with GPS signals

4.2.5 Available products for on-line PD monitoring of cable

Doble Lemke Doble Lemke monitors cable terminations, and joint using UHF sensor directly attached to the sensing place such as GIS- cable termination. This data shown as PRPD can be analyzed and transferred through a TCP/IP network. 69

Emerson Their approach for online cable PD monitoring combines Tangent Delta testing, offline method with VLF PD monitoring, and the ultrasonic method. RF embedded noise reduction can eliminate noise from PD with RFCT as a sensor. Regarding localization of the PD source, they can make it possible to have about 1 % accuracy in up to 3 miles of cable length, which is an application for XLPE, EPR, PILC and CLX Armored cable types

HVPD HVPD uses HFCT attached around the earth connections and TEV attached magnetically to the outside of metal-clad switchgear sensor which is applicable for Polymeric (XLPE, PVC), Paper (PILC, MIND), Rubber (EPR), both 3-Core and Single-Core Cables, and 'Mixed' cables with transition joints. Two cable ends attached sensors are monitored for PD localization using a pulse injection method which is successfully performed for up to 5 km on MV cable

IPEC IPECs method monitors two ends of the cable terminal which provide PD source location with TEV sensors. Basically they use HFCT, capacitive coupling, and airborne acoustic sensors. This is applicative for MV, HV, and EHV cable networks.

KEMA KEMA uses inductive sensors at two ends of cable termination, avoiding significant signal attenuation at RTU (Ring Main Unit) or substation and monitoring only for the cable itself. This method also localizes the source of PD by injecting periodic pulse and measuring propagation time. By doing so, two different sensors located at each end of cable can communicate and get time synchronization with each other. All the information from each sensor is transferred to a control centre. Maximum cable length of this application method is 8 km (for XLPE), 4km (for PILC, MIND), and 2km (for

70

EPR). 3D visualization of PD in the cable moreover makes it possible to check PD occurrence according to the cable length, time and intensity.

Power PD Power PD uses HFCT as a sensor on shield ground cables which can be shown as a PRPD or 3D graph.

Techimp Techimp uses HFCT sensors, and FMC (Flexible Magnetic Coupler) sensors directly at the two terminations of the cable. In long cables, the installations can be performed at the middle of cable. For localization of a PD source, they analyze Amplitude/ Frequency characteristics of PD, TDM method, and Arrival Time Analysis with GPS (Global Positioning System). Moreover this can be connected to a Ethernet network, and controlled from a remote location.

4.2.6 Summary and Conclusion

In particular, cable PD monitoring with an on-site (off-line) method with different voltage levels and frequencies has been used. The after laying test is a proper example of on-site PD monitoring. Even though there are different methods of monitoring on cables, PD monitoring seems the most promising technique providing the possible faults including their location. For on-line PD monitoring on cables, IEC 60270 is not an appropriate detecting method because of its low frequency cover range in which high noise level makes it hard to achieve accurate PD measurement. Plus, long cables will attenuate propagating signal, making impossible to calibrate according to the IEC 60270 [164]. Recent research shows that for on-line monitoring on cables, the monitoring frequency range should be up to 100 MHz because of lower noise compared to low frequency band measurement. On the other hand, the cable acts as a low-pass filter,

71

thus the higher frequency pulses related to PD activity is only detected near the PD source. The most appropriate sensor selection for cable case is capacitive coupling and HFCT according to the application. Since cable accessories, joint and terminal, are the biggest cause of possible faults, on-line PD monitoring near joint or terminal of cable has been widely used. However, using two HFCT at each end of cable with PD localizing techniques by TDR or pulse injection method has been proven its efficiency on on-line PD monitoring for long length cable.

4.3 Rotating Machine

Rotating machine such as synchronous generator, induction motors and DC or AC machines is one of the most important parts of the power system. The main reasons of faults in rotating machines are thermal, electrical and, mechanical stress. Continuous PD monitoring of rotating machine has been considered as efficient diagnostic tool for several decades [170, 171]. In this section, on-line PD monitoring of rotating machine will be covered, including their insulation structure, different monitoring techniques and ultimately its configuration with specific Figures.

4.3.1 Rotating machine in power system

Rotating machine structure and material Rotating Machines (RM) have the most intricate structure among power system components similar to transformer. On top of that, they vary according to the types such as synchronous, induction, or permanent magnetic machines. In Table 4.4, typical electrical machines materials are shown.

72

Subassembly

Component Enclosure Heat exchanger electrical connections Bushings Bearings Frame Core Core clamp

Materials

Enclosure

Fabricated structural steel Steel, copper or brass tube Cast epoxy resin Steel Babbitt, high tensile steel rolling elements or soft bearing alloy on bearing shells

Stator body

Structural steel Electrical steel laminations Structural steel or non-magnetic, low-conductivity alloy

Stator winding

Conductors insulation End Winding support

Hard drawn copper or copper wire Mica-paper, glass or film impregnated with resin Glass fibre structural materials and impregnated insulation felt, ropes and board

Rotor winding

Conductors Insulation End winding support

Hard drawn copper or copper wire Mica-paper, glass, or film impregnated with resin Impregnated glass fibre rope

Rotor body

Shaft Core Core clamp slip rings Brushgear

Structural steel or forging Electrical steel laminations or steel Forging integral with shaft Structural steel or non-magnetic, low conductivity alloy steel, brass or copper Carbon or copper brushes in brass brush holders

Table 4.4 Materials used on general electrical machine in power system [172]

As we can see in the above Table, materials used in different part of rotating machine consist of a wide range of different components and its common structure is laminated or impregnated. Because of that PD attenuation and distortion occurs all over the rotating machine [172, 173]. Faults rates and Continuous PD monitoring on RM PD monitoring with appropriate sensors in North American utilities has been adapted on more than 50% of large generators [174]. In [172], there is detailed information for possible faults in rotating machines. The biggest cause of faults in rotating machines is mechanical stress. Electrical failure on RM which is one-third of the total failure 73

rate comes from persistent overloading (4.2%), and normal deterioration (26.4%). The main failed components on RM are stator ground insulation (23%), turn insulation (4%), and others (8%) [175,176]. More detailed information regarding RM failure is in [177]. Therefore PD monitoring stator windings has normally been performed in many industries and utilities. Continuous PD monitoring provide several advantages for rotating machines; (i) provides warning for personnel, and (ii) solves the problem of difficulty for RM testing under the same condition by supplying continuous trendable data [174]. Moreover, other stress such as thermal or mechanical vibration on RM can create a void or cracks which are detectable in the form of PD, expressed as a symptom of stator winding failure [178].

4.3.2 PD types in rotating machines

The most popular sensing place for PD monitoring on RM is at the machine terminal. However PD can occur inside of RM usually from stator winding which can be attenuated or distorted during propagation from the PD source to the measuring place. Therefore analysis of the magnitude and wave form of PD sometime provides inaccurate information regarding the location and type of PD [179]. Moreover, PD measurement can differ according to the loading, temperature, manufacturer, size and so on. Because of that PD monitoring and appropriate pattern recognition of RM is difficult compared to other part of the system. Here typical PD types and locations is shown below Slot-Discharge: The most harmful PD, slot-discharge happens between a magnetic core and bar or coil of the stator winding. In detail, so-called slot discharge takes place between the iron core and bar coil inside the slot [180, 181]. Slot-discharge erodes gradually semiconductor coating of coil and bar if it occurs continuously. It is load dependent and usually has a much larger magnitude in a negative cycle [182]. Internal Discharge (Voids or Delaminations): The voids in the ground wall insulation or delaminations inside of the coil can lead to internal discharge. This stems from bad quality of impregnation processes which are durable compared to

74

delaminated at the copper conductor due to the thermal overstressing [181]. This depends on the thermal condition of RM [182] Endwinding Discharge: This usually occurs in the overhang region when a contamination of the conductor takes place owing to mechanical corrosion or for particular RM, where bar coils belonging to different phases locate in the same slot [183]. Therefore the reason of this discharge results from phase to phase voltage with not enough room between coils of different phases or partly conductive contamination [76]. According to [182], this type of PD usually has a high magnitude in negative cycle and it is temperature dependant.

4.3.3 Different diagnosis and monitoring techniques on rotating machines

In [184], an intensive review of almost all possible monitoring techniques with regard to RM is covered in detail. Largely, there are thermal, chemical, mechanical and, electrical monitoring techniques have been widely used.

Mechanical monitoring Due to the high mechanical stress of RM compared to other power system components, mechanical monitoring on RM is of high importance. These include vibration monitoring, shock pulse method and examinations of Unbalanced Magnetic Pull (UMP) in the air gap. Regarding vibration monitoring, precise selection of sensor placement is of importance. The shock pulse method provides rotor bearing wear level and UMP calculation in air gaps deliver the information regarding the static eccentricity of the rotor with respect to the stator. Thermal monitoring According to [184], there are three different approaches for temperature monitoring on RM as shown below. Estimate the local point temperature with an embedded temperature detector, or resistance temperature detector. The placement of the detector is of importance. 75

Use thermal imaging, to find a hot spot in the RM, which has been used on a lot of other HVE.

Evaluate distributed temperatures of RM or bulk temperatures of the coolant fluid.

Chemical monitoring High thermal stresses in RM generate chemical reactions in the insulation material, usually starting from 120 Celsius by emitting hydrocarbons and ethylene. However, this method tends to be expensive to perform and is limited by its accuracy.

Electrical monitoring [185] With regard to electrical monitoring on RM, methods include the insulation resistance and polarization index, partial discharge, Capacitance and Dissipation Factor, Motor Current Spectral Analysis (MCSA), High Voltage DC Ramp and Power Monitoring shown below. Insulation resistance and polarization index uses moderate PD voltage (500V10000V) as an input across the ground wall insulation of the stator or rotor winding, and they measure resultant current. Capacitance and Dissipation Factor investigates the present of a void inside the stator insulation. Also known as tangent delta the measurements are performed by using bridges usually based on the transformer ratio-arm or Schering principle. High Voltage DC Ramp supplies ramped input and measure the current as a function of voltage. The results curve can indicate any abnormal condition or poor insulation status. Power Monitoring keeps track of power output at the terminal with the equation in [184]. This method ensures electrical health monitoring of RM by measuring instantaneous power and calculating the sum of each phase, which should be close to

76

DC due to the cancellation of AC components from different phases if the flux and torque is in a normal condition [186]. Motor Current Spectrum Analysis (MCSA) monitors stator current and its spectrum. This can be easily implemented with Current Transformer (CT) around supply cables. Because its accurate analysis and easy installation, this method has been widely used. Partial Discharge can be applied in two different ways, on-line and off-line. In the case of off-line, just like off-line PD monitoring after laying the cable case, high AC test voltage is fed into the cable and PD occurrences are recorded. Off- line PD monitoring on RM which are not in operation are analysed without any operating stress such as thermal or mechanical vibration, and other possible stresses while the machine is in the grid. This information can mislead or failure to notice possible faults in RM during operational condition. However, on-line PD monitoring on RM can provide realistic data under the same circumstances of real conditions and situations of load variation. In particular, on-line PD monitoring on RM largely depends on operation temperature and load condition. One limitation of PD monitoring on RM is that this cannot provide any information regarding the pulse-less discharge phenomenon [181].The reason for this discharge results from phase to phase voltage with limited room between coils of different phases or partly conductive contamination [76]. According to [182], this type of PD usually has a high magnitude in negative cycles and is temperature dependant.

4.3.4 On-line PD monitoring on rotating machines

VHF method [187-189] For on-line PD monitoring on rotating machines, capacitive sensors, Rogowski coils, and HFCT at the end of the voltage terminal have been widely used. In particular, 80 pF capacitive coupling installed at the generator bus bar or stator winding on each phase has mostly been used for permanent on-line PD monitoring as shown in Figure 4.8. Since PD occurrence on RM depends on the load and temperature condition,

77

measurement sometimes accompanies its temperature and load record, for example, a full load at a moderate temperature.

Figure. 4.8 Capacitive coupling method on RM [110]

For noise reduction, two sensors installed at different spots in one phase terminal can be used. The basis of this method is the arrival time difference between two sensors. By doing so, sensors can recognize the PD signal source from an external or internal spot.

Stator Slot coupler method (SSC) [190-191] The SSC coupler method uses special sensors located on stator winding wedges as name implied. The sensor in this application is a directional electromagnetic coupler, which can be permanently installed at the stator slot. Nevertheless the exact shape can vary according to the slot size of the generator.

Figure. 4.9 Stator Slot Coupler [220]

78

SSC is about 50 cm long and 1.7mm thick as shown in Figure 4.9. SSC has two coaxial cable outputs at one end which can be connected to a signal collector located on the outside of the generator. Typically six to nine SSC can be installed on one generator. The advantages of this type of connection are noise immunity from stator winding ends or other external sources, and a PD pulse detection ability from 1 to 5 nanosecond in stator winding.

4.3.5 Available products for on-line PD monitoring of RM

Doble Lemke Doble Lemke installs capacitive coupling at the generators bus bar of each phase. The sensors cover the frequency range according to IEC 60270 and VHF. In order to eliminate noises, gating antenna detecting noise signals are attached, for instance, grounding of the machine enclosure is used. By using PRPD analysis, the signal is interpreted and identified in terms of each phase

HVPD HVPD utilizes HFCT (capacitive coupling sensor in case of above 1000 amps on the supply cable) as a sensor attached around the supply cable, which should be capable of high amps conducting through the supply cable. The software program will automatically identify the PD types categorizing the end-winding and slot type.

IrisPower (Qualitrol Company LLC) Qualitrol-Iris power uses an epoxy mica capacitive coupler for hydro generator stator winding monitoring and a small turbo generator bus output rated 6kV and above. Plus, regarding turbo generator application, they use SSC installed under the line end stator winding wedges (in existing machines), or between the top and bottom bars (in new or rewound machines). Noise separation techniques use 40MHz high pass filters, time-of-arrival, and pulse shape characteristics. With Wide Area Networks (WAN), data collection, change configuration and any kind of control activities are possible from a remote location. 79

Power Diagnostix GmbH Power Diagnostix GmbH installs a capacitive coupler close to windings such as a bus bar at each phase if necessary especially for large machines. Global intranet access and visualization of the monitoring data can be connected to this installation.

PDtech (Qualitrol Company LLC) PDtech uses capacitive couplings near the generator terminal and HFCT around the cable which is available for all HV-machines rated current and voltage. This application provides an alarm and is compatible with SCADA systems.

PowerPD They capture PD signals from generators and motors by coupling in each phase using a Capacitor Coupler. This application has an early warning system and scans between 200 KHz-300 MHz.

Techimp Techimp uses capacitive coupling (1000 pF high voltage dry-type (mica/epoxy) capacitors) at each phase around machine terminal, but if it is not appropriate to install capacitive coupler, HFCT can be a substitute around ground connections. For noise reduction and classification of the PD signal, they use a TF map and fuzzy logic based classifier. After connecting to the monitoring device, the data can be transferred and interconnected with SCADA systems, which are also compatible with Ethernet networks with a static IP address.

4.3.6 Summary and Conclusion

PD occurrence on rotating machine such as generators are typically at a higher level even in normal state due to mechanical dynamic and rapid electrical field changes inside for RM. Therefore compared to other parts of the power system components, there is always a certain level of PD on RM while it is in operation. A large variety of electrical monitoring techniques can also provide adequate information in terms of abnormal operation conditions. Even though stator winding, in particular, has s higher 80

vulnerability to faults, the PD signal from stator winding can be distorted or attenuated at the sensing spot. For online PD applications, capacitive sensors at the generator terminal, Rogowski coils and HFCT at the end cable connection spot, or directional electromagnetic coupler at stator winding slot wedge has proven efficient in research and available products in the market. VHF is the most appropriate monitoring frequency range. In order to reduce noise and locate the PD source, two sensors at one phase different spots, pulse shape analysis with 3 PARD diagram can be used.

4.4 GIS (Gas Insulated System)

GIS has been widely used for HV insulation since 1960. Due to its high insulation characteristic and break down voltage with injected gas- usually SF_6 compared to air, GIS makes it possible to construct the substation in a more compact and reliable way [192]. PD detection techniques in GIS are conventional, unconventional or combined both covered in a recent paper [193]. Usually the sensor should be located within an appropriate distance so that the sensor can detect a PD signal from the GIS. UHF method in GIS was used for the first time in the 1980s [194]. In this section, on-line PD monitoring applications for GIS will be presented with their basic structure, failure type, and other possible monitoring methods.

4.4.1 GIS in power system

Structure and insulation components of GIS GIS has a different design, type, and size depending on the rated voltage, manufacturer, and so on. However the basic components and their materials are in Table 4.5.
Components Description Electrically integrated, grounded casing Single phase and Three phase types according to the application. Aluminum and steel is commonly used

GIS Enclosure

81

Gas

Usually Use as gas insulation material or Mixture with High pressure (4 bar) and low pressure (1.2bar) High pressure has better dielectric characteristics

[195]

Conductor system

Aluminum tubes according to the rated voltage and current, its thickness and diameter can be specified Silver plated contact surface Physical support of high-voltage conductors and mechanical operation of switchgear Cause electrical field distortion within GIS

Solid Spacer

Table 4.5 GIS's insulation and enclosure components and material [141]

The inside components of the GIS usually vary between types, such as the circuit breaker, disconnection switch, current transformer, voltage transformer, bus bar and so forth [141]. Therefore, different components can cause different failures inside the GIS. In [196], different analysis of GIS failures was conducted based on thirty years failure history from five German utilities and 7 companies. Depending on the location of GIS, the most common failures occurred in the switching compartment (40.4%), Voltage Transformer (VT), Surge Arrestors (SA) and bushing compartment (17.3%) and other compartment (42.2%). Failure in terms of the type of defects were particles and foreign bodies (20%), shields & bad electrostatic contacts (18%), load current flowing through poor contact (11%), poor dielectric withstand during capacitive switching operation (10%), spacer defects (10%) and so on. When it comes to the voltage state of failure, failure occurred at nominal voltage state (61%) and overvoltage state during switching operation (39%). Lastly the reasons in terms of failure origin were on-site installation or transportation (35%), poor design (32%), manufacturing defects (24%) and unknown (9%).

PD propagation in GIS [197-199] The PD occurrence in GIS can make electromagnetic transient up to 2 GHz, which can propagate inside of the GIS among its coaxial and symmetrical structure. The wave propagation at lower frequencies compared to the diameter of the structure has the characteristic that the electric and magnetic field of the waves are totally transverse to the direction of wave propagation, known as a Transverse 82

Electromagnetic Wave (TEM). This wave cannot pass through the opened contact switching compartment. At higher power frequencies, however, electric and magnetic fields of the transmitting waves are not entirely transverse to the direction of wave propagation, known as Transverse Electric (TE), Transverse Magnetic (TM). The waves have short wave lengths compared to TEM, and either electric or magnetic fields can have the components transmitting at the same direction toward wave propagation direction, which can pass though the opened contact in GIS. The cut-off frequency between TEM and TE, TM can be defined by the following equation
c

(a b)

(4.5)

Where

= cut-off wave length, a, b= outer and inner radius of the conductor and
c

chamber respectively. For example in the case of 420kV,

is about 1.2m so that the

cut-off frequency can be approximately 250MHz. This TEM and TE, TM can affect the detectable frequency range for on-line PD monitoring on GIS. Appendix A in [200], detail mathematical frame work of TEM, TE, and TM is covered.

4.4.2 PD types in GIS

In GIS, there are four distinguishable PD patterns according to the faults types such as fixed protrusions, floating electrodes, free moving particle, and particles fixed on the spacer or insulation surface. According to the situation, identifying all of the faults can be difficult [197, 200-202]. Fixed protrusions: This can occur during a poor manufacturing process or possible contacts of any part inside GIS during operation time. Protrusions in GIS can be dangerous due to the fact that it can distort the electrical distribution field strength which might cause breakdown under abnormal conditions such as lightening, and switching impulses. Floating electrodes: Floating components is relatively large discharge between a floating and an adjacent electrode which is decided by the relationship between its capacitance to the conductor and as to the ground. This type of PD in GIS can be easily detected by acoustic sensor because it produces acoustic pressure waves which contain much greater energy than corona discharge. 83

Free moving particle: This is a particle freely moving inside the GIS. Poor manufacture processes and contact between different electrodes inside GIS can cause free moving particles. Even though free moving particles have been found to be harmless compared, for example, to floating electrodes, continuous PD occurrence from free moving particle can result in SF_6 decomposition, and eventually breakdown in the GIS [203]. This was pointed out as the main cause of failure in a GIS [204] Particles fixed on the spacer or insulation surface: This case indicates that certain free moving particles can be fixed on the resin spacer or other insulation surface inside the GIS. Fixed particles can generate corona type PD or breakdown in the gassolid interface by accelerating electric field distribution strength at that location [205].

4.4.3 Different diagnosis and monitoring techniques on GIS

According to [206], indicative methods to find possible faults are PD diagnostic (22.4%), Visual Inspection of switchgear enclosure and surrounding area (18.1%), and Thermo-graphic Inspection (12%). Therefore PD monitoring and visual inspection will be reliable methods for continuous monitoring on GIS. SF6 Gas leakage detection [207, 208]: Since keeping the SF6 pressure in GIS is critical in order to maintain proper insulation characteristics, gas leakage detection is of importance. The tightness of insulation gas in the GIS is critical. Thus there is usually a gas density transmitter compensated temperature that analyzes gas density providing information regarding gas pressure and internal arc possibility. The SF6 gas is also a potent green house gas which should not leak in an inappropriate way. Widely used techniques to monitor gas leakage in GIS use either electric detectors which are hand held type alarming to detect gas leakage, and simple snoop detection where suspicious spot of leakage. Recently, the SF6 laser image system has become available with a video display pointing out the source of the leakage. Partial discharge [209]: Partial discharge is a proven technique for condition monitoring and commissioning of GIS. A PD-free test given the applied voltage over normal operating voltage is a very efficient tool before new GIS installation. PD detection, moreover, can be applicative for on-line monitoring enhancing safe 84

operation of GIS. The UHF/AE PD detection method has been regarded as the most promising PD monitoring technique for GIS. Detailed techniques in terms of on-line monitoring are covered in this chapter. High frequency current detection [209]: Current pulses caused by discharge in GIS can be detected using by capacitor or charged conductor and intermediary insulator equipped with a receiver electrode immersed in resin. This method has been used and is a possible application for permanent monitoring. SF6 quality assessment [209]: Impurities of GIS cause a significant impact on insulation failure. Therefore, appropriate monitoring for SF6 quality has been used periodically in order to decide dielectric strength. Gas analysis performed by sampling for gas-chromatography or infer-red spectrograph is a proven method. Air contents measurement with a portable oxygen detector can give an immediate indication of air contents in the GIS. Lastly continuous or periodic moisture measurement is efficient as well. Because this method is relatively expensive and redundant, periodic detection for the first month of operation is sufficient in order to ensure SF6 filling condition in the GIS.

4.4.4 On-line PD monitoring on GIS

Conventional method for sensitivity verification [36, 210-211] Conventional method on GIS PD detection is not appropriate for on-line application. However synchronous measurement of UHF/AE methods and IEC 60270 enables desired sensitivity in order to estimate apparent charge which is usually less than 5 pC for the optimum case. Usually there are two steps for sensitivity verification proposed by CIGRE. The first step is a laboratory test in order to Figure out the magnitude of the artificial low voltage pulse related to real PD by using IEC 60270 and UHF method at the same time which can be used later during an on-site test. The second step is to inject artificial pulses corresponding to the values established during the laboratory test. Finally the on-site sensitivity verification can be done by detecting those signals using by UHF sensors.

85

Unconventional method [196, 210] Unconventional on-line PD monitoring on GIS uses VHF/UHF detection and the AE detection method. As mentioned above, since TEM is highly attenuated at higher frequencies, TEM mode in GIS is strong in VHF (40 to 300 MHz). Even though the VHF method has many similarities with UHF in the sense that they can provide the location of PD, the most significant disadvantage of VHF is interference similar to IEC method. However, VHF method sometimes makes it possible to be calibrated by injecting a known pulse. On the other hand, the UHF (300 to 3 GHz) method can detect TE and TM which are generated by a very fast rising time PD current within ten picoseconds. In addition UHF has a similar sensitivity to the IEC method due to its good immunity to noise. The rule of thumb regarding distance between UHF sensors should be within 20m. The most widely used sensor types are in Figure 4.10.

Figure. 4.10 different types of VHF/UHF sensors for GIS application

Acoustic method [196, 210] By mounting AE sensors externally that usually cover the frequency band between 20 to 100 kHz on GIS, acoustic PD signal can be detected with high sensitivity in the case of, for instance, moving particles. This method can also be calibrated by IEC 60270 as the same procedure for the UHF PD detection method described above. AE detection concerns amplitude and flight time in order to identify the defect type and carry out a risk assessment. Since the intensity of an acoustic signal is lower than an electrical one, the preferred distance between sensors is several meters. Detailed characteristic of acoustic emission are covered in [212].

86

PD localization [135] PD source localization inside the GIS simply uses time-of-flight measurement with two different sensors. Electric PD signal inside the GIS can propagate two UHF PD sensors at different time intervals according to the sensor placement.

Figure. 4.11 Time of flight method for PD localization in GIS [135]

In Figure 4.11, the simple scheme of PD localization is shown. If there are two different UHF sensors and the PD occurs between them, the time domain PD location can be calculated as below in a situation when the time of flight at two different UHF sensors is known.
x1 x ( x2 x1 ) 2 x c0 t 2

(4.6)

c0 =the propagation velocity of the signal (30cm/ns)

which is proportional to insulator permittivity.

4.4.5 Available products on-line PD monitoring of GIS

Doble Lemke Doble Lemke uses unconventional UHF techniques with an inductive sensor near the GIS termination connected to the cable or GIS grounding bar. Possible noise can be eliminated by gating and windowing technique. The PD signal can be analyzed using PRPD.

Doble TransiNor AIA 87

Doble TransiNor AIA provide acoustic solution for GIS during normal service operation using piezoelectric type monitoring the acoustic signal from the PD source such as bouncing particles, protrusions and loose shields. The sensitivity of PD measurement will be higher than conventional method (IEC 60270).

DMS (Qualitrol company Ltd) DMS uses internal UHF coupler on the inside of the hatch cover plates. The signals from up to 3 UHF sensors are collected using optical convertor unit in which appropriate noise elimination takes place. Filtered signals from optical the convertor unit are sent to equipment cabinet usually in a control room as an optical data stream.

HVPD HVPD uses the unconventional way with Transient Earth Voltage (TEV) and external capacitive coupler sensors attached to the outer casing of the switchgear in the correct position. In addition, a HFCT clipped around the cable earth strap at the bottom of the switchgear is also used. The noise can be reduced by using frequency analysis for switch noise and wave shape analysis for RF and sinusoidal noise. Two or more TEV sensors with the time-of-flight method enable PD localization. For detecting PD from corona or the surface of the cable termination, sealing end, and air-insulated switchgear, Airborne AE probes can be used, which also can be combined with TEV/HFCT in order to localize PD source.

Power Diagnostix systems They use internal flange type sensors at a spare flange, shielded ring antenna type UHF sensors at isolated spacers, and external window sensors similar to internal flange type sensors. Especially external window sensor size can vary according to the GIS design. Ethernet cable connects the sensors to an acquisition unit. Power Diagnostix system uses a particular calibrator injecting very steep voltage output into the GIS which can cover up to 1.5GHz. By doing so, matrix of the attenuation between sensors at each bar or bus bar section can be obtained. Power Doagnostix systems also provides acoustic solution with the AIA compact which detect acoustic signal by using piezo-electric acoustic sensors. In addition, the sensitivity of acoustic 88

method is mostly comparable to the conventional detection according to IEC 60270. This method is superior in some defect type detection such as hopping or bouncing particles.

PSD Tech PSD Tech uses external or internal open barrier or metal-closed barrier type UHF sensors with noise sensors. The signal is first collected to a data acquisition unit which is connected to the main unit for signal analysis using neural network.

PowerPD PowerPD uses 4 acoustic detection sensors for GIS ductwork, HV GIS circuit breakers and cubicle type GIS (CGIS). Remote monitoring is possible via a PC or the internet.

Techimp Techimp uses VHF/UHF sensors such as a window coupler on the dielectric inspection window, tem antenna near GIS bushing, spacer coupler at dielectric spacer, and bushing coupler around a metal ring below the GIS bushing. Most of the sensors have a sensitivity of 5pC. They use TF analysis and fuzzy logic based pattern recognition as well as noise rejection.

4.4.6 Summary and Conclusion

Due to the different parts inside of the GIS and its complex structure, PD occurrence in GIS occurs in a variety of types and possible locations. By the early 1990s, the UHF method on GIS had developed which was the first unconventional electrical PD detection application on power system components. PD characteristic in GIS is a bit different compared to other power system components in the sense that there is an open contact space inside the GIS. Therefore TEM or TE, TM waves can propagate through GIS which can be detected using the VHF/UHF detection method.

89

For on-line PD monitoring, the VHF/UHF method has usually been widely used with internal or external sensors located on the GIS surface. Localization of PD source can be realized by using time of flight with two or more sensors in different places. From a practical point of view, many companies use similar internal or external sensors in different spots. On-line PD monitoring on GIS by companies using the UHF method is very common world-wide.

4.5 On-line PD monitoring on power system components

As describeed above, on-line PD monitoring on power system components has been widely used. Slightly different solutions are currently available depending on a variety of commercial products. In order to apply appropriate on-line PD monitoring techniques, a proper understanding of insulation characteristics and distinctive operation mechanisms on each power system apparatus should to be a priority. Combining with other monitoring techniques compensating for each methods drawbacks can also be taken into account. UHF/AE shows very promising potential for on-line PD monitoring.

90

CHAPTER

5 Conclusion and Future work

PD measurement and analysis can be considered very powerful tool to assess insulation condition monitoring of high voltage apparatuses in power systems as highlighted in this thesis. In the same was as PD has been widely used for commissioning and new equipment installation testing for several decades, on-line monitoring on power system components by means of PD measurement will enhance condition based effective high voltage equipment monitoring. The most significant benefits provided by continuous on-line PD monitoring are; Trend of insulation condition in real time HV equipment monitoring while the system components are in operation The monitoring can be done in the real operating condition Location specific information regarding the insulation condition and possible 91

fault types PD monitoring can be applied to all kinds of HVE

On-line PD monitoring is still a developing area which needs more research and experience. That is why PD measurements are still being developed by experts despite the fact that PD measurements have been performed for several decades. The author is sure that on-line PD monitoring is the most promising techniques for condition assessments of high voltage equipment. Life prediction modelling and life cycle management: Since on-line PD monitoring can provide trendable data for testing, based on continuous PD monitoring, the power system operator can simulate possible life prediction and cycle management of power system equipment. In order to complete this task, the following studies are required: Appropriate feature extraction: Since on-line PD monitoring generates huge amount of data, appropriate feature selection and storage is essential for data storage and analysis. However at the moment there is no dominant technique for this matter. Built-in PD monitoring system: Power system components from different manufactures have different structures, insulation materials, and life cycles even for the same purpose. Therefore PD monitoring for important system components from initial use is significant in order to make the right decisions from trend data. The most preferred solution for this is a built-in PD monitoring system, for example GIS or transformer in the manufacturing process. Some manufacturers actually have embedded their own PD monitoring systems including Japan AE, ABB, Mitsubishi Electric, Toshiba and so on. Integrating as part of a smart grid [213]: Smart grids are considered as the hottest issue in modern power systems. By using all possible infrastructure and recent techniques, optimum, reliable, economical, and efficient operation of power system is the goal of smart grid. Therefore possible integration of on-line high voltage equipment monitoring with smart grid infrastructure has high potential to make a leap forward in the provision of robust and reliable power systems operation.

92

References
THESIS & BOOK [1] Stone, G.C, Partial Discharge Diagnostics and Electrical Equipment Insulation Condition Assessment, dielectrics and Electrical Insulation, IEEE Transactions on Volume: 12, Issue: 5. Page(s): 891 904, 2005 [2] Han, Y.; Song, Y. H, Condition monitoring techniques for electrical equipment: a literature survey, Power Engineering Review, IEEE Volume: 22, Issue: 9, Page(s): 59, 2002 [3] Montanari, G.C., Cavallini, A, Insulation condition assessment of power equipments in electrical assets based on on-line monitoring of partial discharges, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 7 12. 2008 [4] Jitka Fuhr, Procedure for identification and localization of dangerous PD sources in power transformers, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 12, Issue: 5, Page(s): 1005 1014, 2005 [5] Blokhintsev, I.; Patterson, C.L.; Cassidy, B.J.; Loesch, A.H, Advantage of On Line Partial Discharge continuous monitoring of medium voltage substation, Electrical Insulation Conference, 2009. EIC 2009. IEEE, Page(s): 153 158, 2009 [6] Ortega, D.F.; Castelli-Dezza, F, On line partial discharges test on rotating machines supplied by IFDs, Electrical Machines (ICEM), 2010 XIX International Conference on, Page(s): 1 4, 2010 [7] C. Sumereder, M. Muhr, Estimation of Residual Lifetime Theory and Practical Problems, 8th Hflers Days Portoroz, 6. 8, November 2005 [8] Senn F., Muhr M, Ladsttter W., Grubelnik W., Complexity of Determining factors for the thermal evaluation of high voltage insulation systems on the example of rotating machines, DISEE 2008 Demnovsk Dolina, Slovakia, 2008 [9] WG D1.33 Guide for Electrical Partial Discharge Measurements in compliance to IEC 60270, Technical Brochure, Electra No.241, Page(s): 61-67, DEC 2008

93

[10] IEC 60270 High-voltage test techniques- Partial discharge measurements Edition 3, 2000 [11] E. Kuffel, W. S. Zaengl, J. Kuffel, High voltage engineering: fundamentals, Book, Newnes, 2000 [12] Xin Liu, Partial Discharge Detection and Analysis in Low Pressure Environments, Doctoral Thesis, The ohio state university, USA, 2006 [13] Pemen, A.J.M.; van der Laan, P.C.T.; Wout de Leeuw, Propagation of partial discharge signals in stator windings of turbine generators, Energy Conversion, IEEE Transactions on Volume: 21 , Issue: 1, Page(s): 155 161, 2006 [14] Pedersen, A.; Crichton, G.C.; McAllister, I.W., The theory and measurement of partial discharge transients, Electrical Insulation, IEEE Transactions on Volume: 26 , Issue: 3, Page(s): 487 497, 1991 [15] Fenger, M., Test procedure and acceptance criteria for PD commissioning testing of transmission class cables, Electrical Insulation (ISEI), Conference Record of the 2010 IEEE International Symposium on, Page(s): 1 5, 2010 [16] Cavallini, A.; Montanari, G.C.; Tozzi, M., PD apparent charge estimation and calibration: A critical review, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 17 , Issue: 1, Page(s): 198 205, 2010 [17] Hampton, B.F.; Meats, R.J., Diagnostic measurements at UHF in gas insulated substations, Generation, Transmission and Distribution, IEE Proceedings, Page(s): 137 145, 1988 [18] M. Muhr; R. Schwarz, Partial discharge measurement as a Diagnostic Tool for HV-Equipments, Properties and applications of Dielectric Materials, 2006. 8th International Conference on, Page(s): 195 198, 2006 [19] Muhr, M.; Schwarz, R.; Pack, S.; Koerbler, B., Unconventional partial discharge measurement [electrical insulation evaluation], Electrical Insulation and Dielectric Phenomena, 2004. CEIDP '04. 2004 Annual Report Conference on, Page(s): 430 433, 2004 [20] Kemp, I.J., Developments in partial discharge plant-monitoring technology, Partial Discharge, 1993., International Conference on, Page(s): 52 55, 1993 94

[21] CIGRE Working group D1.33 Guidelines for Unconventional Partial Discharge Measurements, Cigre Technical Brochure, DEC, 2010 [22] Lundgaard, L.E, Partial discharge. XIV. Acoustic partial discharge detection practical application, Electrical Insulation Magazine, IEEE, Volume: 8, Issue: 5, Page(s): 34 43, 1992 [23] Lundgaard, L.E., Partial discharge. XIII. Acoustic partial discharge detectionfundamental considerations, Electrical Insulation Magazine, IEEE, Volume: 8, Issue: 4, Page(s): 25 31, 1992 [24] Kozako, M.; Yamada, K.; Morita, A.; Ohtsuka, S.; Hikita, M.; Kashine, K.; Nakamura, I.; Koide, H., Fundamental study on partial discharge induced acoustic wave propagation in simulated transformer composite insulation system, Properties and Applications of Dielectric Materials, 2009. ICPADM 2009. IEEE 9th International Conference on the, Page(s): 477 480, 2009 [25] Schwarz, R.; Muhr, M. Modern technologies in optical partial discharge detection, Electrical Insulation and Dielectric Phenomena, 2007. CEIDP 2007. Annual Report - Conference on, Page(s): 163 166, 2007 [26] R. Schwarz, M. Muhr, S. Pack, Evaluation of partial discharge impulses with optical and conventional detection systems, Proceedings of the XIVth International Symposium on High Voltage Engineering, Tsinghua University, Beijing, China, August 25-29, 2005 [27] Oliveira, S.C.; Fontana, E., Optical Detection of Partial Discharges on Insulator Strings of High-Voltage Transmission Lines, Instrumentation and Measurement, IEEE Transactions on Volume: 58, Issue: 7, Page(s): 2328 2334, 2009 [28] Tominaga, S.; Kuwahara, H.; Hirooka, K.; Yoshioka, T., SF6 Gas Analysis Technique and its Application for Evaluation of Internal Conditions in SF6 Gas Equipment, Power Apparatus and Systems, IEEE Transactions on Volume: PAS100, Issue: 9, Page(s): 4196 4206, 1981 [29] Muhamad, N.A.; Phung, B.T.; Blackburn, T.R., Dissolved gas analysis (DGA) of partial discharge fault in bio-degradable transformer insulation oil, Power Engineering Conference, 2007. AUPEC 2007. Australasian Universities, Page(s): 1 6, 2007

95

[30] Kaufhold, M.; Bamji, S.S.; Bulinski, A.T., Optical detection of partial discharges in gas-insulated systems, Electrical Insulation and Dielectric Phenomena, 1996. IEEE 1996 Annual Report of the Conference on Volume: 2, Page(s): 618 - 623 vol.2, 1996 [31] Coenen, S.; Tenbohlen, S.; Markalous, S.M.; Strehl, T., Sensitivity of UHF PD measurements in power transformers, Dielectrics and Electrical Insulation, IEEE Transactions on, Volume: 15 , Issue: 6 , Page(s): 1553 1558, 2008 [32] P.D. Agoris; S. Meijer; E. Gulski; J.J. Smit; T.J.W. Hermans; L. Lamballais, Sensitivity check for on-line VHF/UHF PD detection on Transmission cables, Properties and applications of Dielectric Materials, 2006. 8th International Conference on, Page(s): 204 207, 2006 [33] Meijer, S.; Judd, M.D.; Tenbohlen, S., Sensitivity check for radio frequency partial discharge detection for power transformers, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 1031 1035, 2008 [34] Agoris, P.; Meijer, S.; Smit, J.J., Sensitivity Check of an Internal VHF/UHF Sensor for Transformer Partial Discharge Measurements, Power Tech, 2007 IEEE Lausanne, Page(s): 2065 2069, 2007 [35] Hoshino, T.; Maruyama, S.; Nojima, K.; Hanai, M., A unique sensitivity verification combined with real-time partial-discharge identification method, Power Delivery, IEEE Transactions on Volume: 20, Issue: 3, Page(s): 1890 1896, 2005 [36] Knapp, M.; Feger, R.; Feser, K.; Breuer, A., Application of the CIGRE sensitivity verification for UHF PD detection in three-phase GIS, High Voltage Engineering, 1999. Eleventh International Symposium on (Conf. Publ. No. 467) Volume: 5, Page(s): 82 - 85 vol.5, 1999 [37] CIGRE Working Group D1.02 Sensors and Sensing Used For Non-conventional PD Detection, CIGRE 2006 [38] Petchphung, P.; Leelajindakrairerk, M.; Pattanadech, N.; Yutthagowith, P.; Aunchaleevarapan, K., The comparison study of PD measurement with conventional method and unconventional method, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 32 36, 2008

96

[39] Reid, A.; Judd, M.; Stewart, B.; Fouracre, R., Comparing IEC60270 and RF partial discharge patterns, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 89 92, 2008 [40] Reid, A.J.; Judd, M.D.; Fouracre, R.A.; Stewart, B.G.; Hepburn, D.M., Simultaneous measurement of partial discharges using IEC60270 and radiofrequency techniques, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 18 , Issue: 2, Page(s): 444 455, 2011 [41] Stewart, B.G.; Judd, M.D.; Reid, A.J.; Fouracre, R.A., Suggestions to augment the IEC60270 partial discharge standard in relation to radiated electromagnetic energy, Electrical Insulation Conference and Electrical Manufacturing Expo, 2007, Page(s): 175 178, 2007 ger, M.; Winter, P., Advanced possibilities of synchronous conventional and UHF PD measurements for effective noise suppression, Electrical Insulation (ISEI), Conference Record of the 2010 IEEE International Symposium on, Page(s): 1 4, 2010 [43] Tenbohlen, S.; Pfeffer, A.; Coenen, S, On-site experiences with multi-terminal IEC PD measurements, UHF PD measurements and acoustic PD localization, Electrical Insulation (ISEI), Conference Record of the 2010 IEEE International Symposium on , Page(s): 1 5, 2010 [44] Xiaodong Wang; Baoqing Li; H.T. Roman; O.L. Russo; K. Chin; K.R. Farmer, Acousto-optical PD detection for transformers, Power Delivery, IEEE Transactions on Volume: 21, Issue: 3, Page(s): 1068 1073, 2006 [45] Morshuis, P.H.F, Partial discharge mechanism leading to breakdown, analyzed by fast electrical and optical measurements, Master thesis, Delft University of Technology, 1993 [46] Jouseau, E.; Vivien, G.; Odic, E.; Maroni, C.-S, Exhibition of an ageing criterion based on partial detection for Medium Voltage equipment, Diagnostics for Electric Machines, Power Electronics and Drives, 2005. SDEMPED 2005. 5th IEEE International Symposium on, Page(s): 1 5, 2005 [47] Alison 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

97

[48] S.M. Markalous and K. Feser, All- Acoustic PD Measurements Of Oil/Paperinsulated Transformers For PD-Location, Proc. of the Second International Conference on Advances in Processing, Testing and Application of Dielectric Materials (APTADM),September 2004. [49] Imad-U-Khan; Zhongdong Wang; Cotton, I.; Northcote, S, Dissolved gas analysis of alternative fluids for power transformers, Electrical Insulation Magazine, IEEE Volume: 23, Issue: 5, Page(s): 5 14, 2007 [50] Boczar, T.; Zmarzly, D., Optical spectra of surface discharges in oil, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 13 , Issue: 3, Page(s): 632 639, 2006 [51] Schwarz, R.; Muhr, M.; Pack, S., Partial discharge detection in oil with optical methods, Dielectric Liquids, 2005. ICDL 2005. 2005 IEEE International Conference on, Page(s): 245 248, 2005 [52] Baker, P.C.; Judd, M.D.; Mcarthur, S.D.J, A frequency-based RF partial discharge detector for low-power wireless sensing, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 17 , Issue: 1, Page(s): 133 140, 2010 [53] V.R Garcia-Colon, J. A. Estrada Garca, E. Betancourt Ramirez Development of ultra wide band partial discharge sensors for power transformer winding insulation, Contribution D1-211, Cigr. Session 2008 [54] Kaneko, S.; Okabe, S.; Yoshimura, M.; Muto, H.; Nishida, C.; Kamei, M., Detecting characteristics of various type antennas on partial discharge electromagnetic wave radiating through insulating spacer in gas insulated switchgear, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 16 , Issue: 5, Page(s): 1462 1472, 2009 [55] Mallikarjunappa, K.; Ratra, M.C., Detection of partial discharges in power capacitors using high frequency current transformers, Electrical Insulation and Dielectric Phenomena, 1990. Annual Report., Conference on, Page(s): 379 384, 1990 [56] Ray, W.F.; Hewson, C.R., High performance Rogowski current transducers, Industry Applications Conference, 2000. Conference Record of the 2000 IEEE Volume: 5, Page(s): 3083 - 3090 vol.5, 2000

98

[57] Argueso, Marta; Robles, Guillermo; Sanz, Javier, Implementation of a Rogowski coil for the measurement of partial discharges, Review of Scientific Instruments Volume: 76, Issue: 6, Page(s): 065107 - 065107-7, 2005 [58] Veloso, G.F.C.; da Silva, L.E.B.; Noronha, I.; Torres, G.L.; Haddad, J.; Pereira, R.R.; Se Un Ahn, Detection of partial discharge in power transformers using Rogowski coil and multi resolution analysis, Power Electronics Conference, 2009. COBEP '09. Brazilian, Page(s): 1006 1010, 2009 [59] Zhu, H.; Green, V.; Sasic, M.; Halliburton, S., Increased sensitivity of capacitive couplers for in-service PD measurement in rotating machines, Energy Conversion, IEEE Transactions on Volume: 14 , Issue: 4, Page(s): 1184 1192, 1999 [60] Muhr, M.; Woschitz, R., Partial discharge diagnostic, Properties and Applications of Dielectric Materials, 2000. Proceedings of the 6th International Conference on Volume: 1, Page(s): 223 - 226 vol.1, 2000 [61] Goodeve, T.E.; Stone, G.C.; Macomber, L., Experience with compact epoxymica capacitors for rotating machine partial discharge detection, Electrical Electronics Insulation Conference, 1995, and Electrical Manufacturing & Coil Winding Conference. Proceedings, Page(s): 685 689, 1995 [62] Lopez-Roldan, J.; Tang, T.; Gaskin, M., Optimisation of a sensor for onsite detection of partial discharges in power transformers by the UHF method, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 15, Issue: 6, Page(s): 1634 1639, 2008 [63] Judd, M.D.; Li Yang; Hunter, I.B.B., Partial discharge monitoring of power transformers using UHF sensors. Part I: sensors and signal interpretation, Electrical Insulation Magazine, IEEE Volume: 21, Issue: 2, Page(s): 5 14, 2005 [64] Pommerenke, D.; Strehl, T.; Heinrich, R.; Kalkner, W.; Schmidt, F.; Weissenberg, W., Discrimination between internal PD and other pulses using directional coupling sensors on HV cable systems, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 6 , Issue: 6, Page(s): 814 824, 1999 [65] Craatz, P.; Plath, R.; Heinrich, R.; Kalkner, W., Sensitive on-site PD measurement and location using directional coupler sensors in 110 kV prefabricated joints, High Voltage Engineering, 1999. Eleventh International Symposium on (Conf. Publ. No. 467) Volume: 5, Page(s): 317 - 321 vol.5, 1999

99

[66] Lima, S.E.U.; Frazao, O.; Farias, R.G.; Araujo, F.M.; Ferreira, L.A.; Santos, J.L.; Miranda, V., Mandrel-Based Fiber-Optic Sensors for Acoustic Detection of Partial Dischargesa Proof of Concept, Power Delivery, IEEE Transactions on Volume: 25 , Issue: 4, Page(s): 2526 2534, 2010 [67] Zargari, A.; Blackburn, T.R., Application of optical fibre sensor for partial discharge detection in high-voltage power equipment, Electrical Insulation and Dielectric Phenomena, 1996. IEEE 1996 Annual Report of the Conference on Volume: 2, Page(s): 541 - 544 vol.2, 1996 [68] Varl, A., On-line diagnostics of oil-filled transformers, Dielectric Liquids, 2002. ICDL 2002. Proceedings of 2002 IEEE 14th International Conference on, Page(s): 253 257, 2002 [69] Xuezeng Zhao; Yangliu Li, An On-Line Monitoring System for Gases Dissolved in Transformer Oil Using Wireless Data Acquisition, Power and Energy Engineering Conference, 2009. APPEEC 2009. Asia-Pacific, Page(s): 1 4, 2009 [70] Xiao-xing Zhang; Wang-ting Liu; Ju Tang; Peng Xiao, Study on PD detection in SF6 using multi-wall carbon nanotube films sensor, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 17 , Issue: 3, Page(s): 833 838, 2010 [71] Ding, W.; Hayashi, R.; Suehiro, J.; Imasaka, K.; Hara, M., Observation of dynamic behavior of PD-generated SF6 decompositions using carbon nanotube gas sensor, Electrical Insulation and Dielectric Phenomena, 2005. CEIDP '05. 2005 Annual Report Conference on, Page(s): 561 564, 2005 [72] IEEE Std C57.127-2000, IEEE Trial-Use Guide for the Detection of Acoustic Emissions From Partial Discharges in Oil-Immersed Power Transformers,, 2000 [73] Sahoo, N.C.; Salama, M.M.A.; Bartnikas, R., Trends in partial discharge pattern classification: a survey, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 12, Issue: 2, Page(s): 248 264, 2005 [74] Barkat, B.; Al-Marzouqi, H., A correlation based approach to interference suppression from phase resolved partial discharge patterns, Solid Dielectrics (ICSD), 2010 10th IEEE International Conference on, Page(s): 1 4, 2010 [75] Hudon, C.; Belec, M., Partial discharge signal interpretation for generator diagnostics, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 12 , Issue: 2, Page(s): 297 319, 2005 100

[76] Stone, G.C., Relevance of phase resolved PD analysis to insulation diagnosis in industrial equipment, Solid Dielectrics (ICSD), 2010 10th IEEE International Conference on, Page(s): 1 5, 2010 [77] Strachan, S.M.; Rudd, S.; McArthur, S.D.J.; Judd, M.D.; Meijer, S.; Gulski, E., Knowledge-based diagnosis of partial discharges in power transformers, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 15 , Issue: 1, Page(s): 259 268, 2008 [78] Hu, X.; Judd, M.D.; Siew, W.H., A study of PD location issues in GIS using FDTD simulation, Universities Power Engineering Conference (UPEC), 2010 45th International, Page(s): 1 5, 2010 [79] Hoogenraad, G.; Beyer, J., Digital HVDC partial discharge testing, Electric al Insulation, 2000. Conference Record of the 2000 IEEE International Symposium on, Page(s): 448 451, 2000 [80] Cavallini, A.; Montanari, G.C.; Puletti, F.; Contin, A., A new methodology for the identification of PD in electrical apparatus: properties and applications, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 12, Issue: 2, Page(s): 203 215, 2005 [81] Belkov, A.; Koltunowicz, W.; Obralic, A.; Plath, R., Advanced approach for automatic PRPD pattern recognition in monitoring of HV assets, Electrical Insulation (ISEI), Conference Record of the 2010 IEEE International Symposium on, Page(s): 1 5, 2010 [82] Rethmeier, K.; Kruger, M.; Kraetge, A.; Plath, R.; Koltunowicz, W.; Obralic, A.; Kalkner, W., Experiences in on-site partial discharge measurements and prospects for PD monitoring, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 1279 1283, 2008 [83] Koltunowicz, W.; Plath, R., Synchronous multi-channel PD measurements, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 15, Issue: 6, Page(s): 1715 1723, 2008 [84] Krivda, A., Automated recognition of partial discharges, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 2 , Issue: 5, Page(s): 796 821, 1995 [85] Hao Zhang; T.R. Blackburn; B.T. Phung; D. Sen, A novel wavelet transform technique for on-line partial discharge measurements Part 1: WT de-noising 101

algorithm, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 14 , Issue: 1, Page(s): 3 14, 2007 [86] Nagesh, V.; Gururaj, B.I., Evaluation of digital filters for rejecting discrete spectral interference in on-site PD measurements, Electrical Insulation, IEEE Transactions on Volume: 28 , Issue: 1, Page(s): 73 85, 1993 [87] Q. Su, Partial Discharge Measurements on Generators Using a Noise Gating System, Australasian Universities Power Engineering Conference, September, 26-29 1999, Darwin, Australia, 1999 [88] Stone, G.C.; Sedding, H.G., In-service evaluation of motor and generator stator windings using partial discharge tests, Industry Applications, IEEE Transactions on Volume: 31 , Issue: 2, Page(s): 299 303, 1995 [89] Veen, J.; van der Wiellen, P.C.J.M., The application of matched filters to PD detection and localization, Electrical Insulation Magazine, IEEE Volume: 19 , Issue: 5, Page(s): 20 26, 2003 [90] Sher Zaman Khan; Zhu Deheng; Jin Xianhe; Tan Kexiong, A new adaptive technique for on-line partial discharge monitoring, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 2 , Issue: 4, Page(s): 700 707, 1995 [91] Wagenaars, P.; Wouters, P.A.A.F.; van der Wielen, P.C.J.M.; Steennis, E.F., Adaptive matched filter bank for noise reduction in online partial discharge monitoring, Electrical Insulation and Dielectric Phenomena, 2009. CEIDP '09. IEEE Conference on, Page(s): 356 359, 2009 [92] Kopf, U.; Feser, K., Rejection of narrow-band noise and repetitive pulses in onsite PD measurements, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 2 , Issue: 6, Page(s): 1180 1191, 1995 [93] Shuzhen Xu; Middleton, R.; Xiaodong liu; Chenchen, A new method based on wavelet packet transformation to reduce narrow band noise in on-site PD measurement, Power Engineering Society Summer Meeting, 2002 IEEE Volume: 2, Page(s): 898 - 903 vol.2, 2002 [94] Mortazavi, S.H.; Shahrtash, S.M., Comparing denoising performance of DWT,WPT, SWT and DT-CWT for Partial Discharge signals, Universities Power Engineering Conference, 2008. UPEC 2008. 43rd International, Page(s): 1 6, 2008

102

[95] Ma, X.; Zhou, C.; Kemp, I.J., Interpretation of wavelet analysis and its application in partial discharge detection, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 9, Issue: 3, Page(s): 446 457, 2002 [96] Gaouda, A.M.; El-Hag, A.; Abdel-Galil, T.K.; Salama, M.M.A.; Bartnikas, R., On-line detection and measurement of partial discharge signals in a noisy environment, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 15, Issue: 4, Page(s): 1162 1173, 2008 [97] Andrej Krivda, Recognition of discharges Discrimination and classification, Doctoral Thesis, Delft University of Technology, 1995 [98] James, R.E.; Phung, B.T., Development of computer-based measurements and their application to PD pattern analysis, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 2, Issue: 5, Page(s): 838 856, 1995 [99] Mazroua, A.A.; Bartnikas, R.; Salama, M.M.A., Discrimination between PD pulse shapes using different neural network paradigms, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 1 , Issue: 6, Page(s): 1119 1131, 1994 [100] Ab Aziz, N.F.; Hao, L.; Lewin, P.L., Analysis of Partial Discharge Measurement Data Using a Support Vector Machine, Research and Development, 2007. SCOReD 2007. 5th Student Conference on, Page(s): 1 6, 2007 [101] Ethem ALPAYDIN, Introduction to Machine Learning, The MIT Press, October 2004, ISBN 0-262-01211-1, 2004 [102] Kranz, H.-G., Diagnosis of partial discharge signals using neural networks and minimum distance classification, Electrical Insulation, IEEE Transactions on Volume: 28 , Issue: 6, Page(s): 1016 1024, 1993 [103] Cover, T.; Hart, P., Nearest neighbor pattern classification, Information Theory, IEEE Transactions on Volume: 13, Issue: 1, Page(s): 21 27, 1967 [104] Poyhonen, S.; Conti, M.; Cavallini, A.; Montanari, G.C.; Filippetti, F., Insulation defect localization through partial discharge measurements and numerical classification, Industrial Electronics, 2004 IEEE International Symposium on Volume: 1, Page(s): 417 - 422 vol. 1, 2004 [105] Balasundaram Karthikeyan, Srinivasan Gopal, Srinivasan Venkatesh, Subramanian Saravanan, Pnn and Its Adeptive Version an ingenious Approach To 103

PD Pattern Classification Compared with BPA Network,, Journal of ELECTRICAL ENGINEERING, VOL. 57, NO. 3, 2006, 138145, 2006 [106] Hamilton-Wright, A. and Stashuk, D.W., Comparing "pattern discovery" and back-propagation classifiers, Neural Networks, 2005. IJCNN '05. Proceedings. 2005 IEEE International Joint Conference on Volume: 2, Page(s): 1286 - 1291 vol. 2, 2005 [107] Salama, M.M.A.; Bartnikas, R., Determination of neural-network topology for partial discharge pulse pattern recognition, Neural Networks, IEEE Transactions on Volume: 13 , Issue: 2, Page(s): 446 456, 2002 [108] Po-Hung Chen; Hung-Cheng Chen; An Liu; Li-Ming Chen, Pattern recognition for partial discharge diagnosis of power transformer, Machine Learning and Cybernetics (ICMLC), 2010 International Conference on Volume: 6, Page(s): 2996 3001, 2010 [109] Lai, K.X.; Phung, B.T.; Blackburn, T.R., Application of data mining on partial discharge part I: predictive modelling classification, Dielectrics and Electric al Insulation, IEEE Transactions on Volume: 17, Issue: 3, Page(s): 846 854, 2010 [110] Hao, L.; Lewin, P.L., Partial discharge source discrimination using a support vector machine, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 17 , Issue: 1, Page(s): 189 197, 2010 [111] Hoof, M.; Patsch, R., Analyzing partial discharge pulse sequences-a new approach to investigate degradation phenomena, Electrical Insulation, 1994., Conference Record of the 1994 IEEE International Symposium on Digital Object Identifier: 10.1109/ELINSL.1994.401501, Page(s): 327 331, 1994 [112] Hoof, M.; Patsch, R., Pulse-sequence analysis: a new method for investigating the physics of PD-induced ageing, Science, Measurement and Technology, IEE Proceedings - Volume: 142 , Issue: 1, Page(s): 95 101, 1995 [113] H. R. Mirzaei, A. Akbari and M. Allahbakhshi, Partial Discharge Identification Using Pulse Sequence Analysis and Neural Network, Proceedings of the 16th International Symposium on High Voltage Engineering SAIEE, Innes House, Johannesburg, 2009 [114] Wang, M., Vandermaar, A.J., Srivastava, K.D., Review of condition assessment of power transformers in service, Electrical Insulation Magazine, IEEE, Volume: 18 Issue:6, page(s): 12 25, Jan 2003 104

[115] Bengtsson, C, Status and trends in transformer monitoring, Power Delivery, IEEE Transactions on Volume: 11 , Issue: 3, Page(s): 1379 1384, 1996 [116] Stewart R. Hardie, B.E. (Hons), A Prototype Transformer Partial Discharge Detection System,, Ph.D Thesis, University of Canterbury, Christchurch, New Zealand, January 2006 [117] Jongen R et al,, A Statistical Approach To Processing Power Transformer Failure Data, 19th. CIRED (International Conference on Electricity Distribution) Vienna, Paper 0546, 21-24 May, 2007 [118] A. Bossi, J.E. Dind, J.M. Frisson, U. Khoudiakov, H.F. Light, et. al., An International Survey of Failures in Large Power Transformers in Service, Final Report of the CIGRE Working Group 12.05, Electra, no. 88, pp. 21-48, 1983 [119] Mackenzie, E.A.; Crossey, J.; de Pablo, A.; Ferguson, W, On-line monitoring and diagnostics for power transformers, Electrical Insulation (ISEI), Conference Record of the 2010 IEEE International Symposium on, Page(s): 1 5, 2010 [120] Sebastian Coenen, Stefan Tenbohlen, Falk-Rdiger Werner, Sacha Markalous, Localization of PD Sources inside Transformers by Acoustic Sensor Array and UHF Measurements, Conference Proceedings of CMD2010, 2010 [121] Markalous, S.; Tenbohlen, S.; Feser, K., Detection and location of partial discharges in power transformers using acoustic and electromagnetic signals, Dielectrics and Electrical Insulation, IEEE Transactions on, Volume: 15, Issue: 6, Page(s): 1576 1583, 2009 [122] Jongen, R.A.; Morshuis, P.; Meijer, S.; Smit, J.J., Identification of partial discharge defects in transformer oil, Electrical Insulation and Dielectric Phenomena, 2005. CEIDP '05. 2005 Annual Report Conference on, Page(s): 565 568, 2005 [123] Agoris, Pantelis D.; Meijer, Sander; Gulski, Edward; Smit, Johan J.; Kanters, A. Jos L. M, On-line partial discharge detection on transformers, Electricity Distribution, 2005. CIRED 2005. 18th International Conference and Exhibition on, Page(s): 1 5, 2005 [124] V.V. Smekalov et al., Condition assessment and life time extension of power transformer, CIGRE Session 2002, pp. 12102. 2002

105

[125] S. Tenbohlen, D. Uhde, J. Poittevin, H. Borsi, P. Werle, U. Sundermann, H. Matthes, Enhanced Diagnosis of Power Transformers using On- and Off-line Methods: Results, Examples and Future Trends, CIGRE Session, paper 12-204. 2000 [126] Yishan Liang; Zhenyuan Wang; Yilu Liu, Power transformer DGA data processing and alarming tool for on-line monitors, Power Systems Conference and Exposition, 2009. PSCE '09. IEEE/PES, Page(s): 1 8, 2009 [127] Duval, M., Dissolved gas analysis: It can save your transformer, Electrical Insulation Magazine, IEEE, Volume:5, Issue:6, page(s): 22, Nov.-Dec. 1989 [128] Emsley, A.M.; Xiao, X.; Heywood, R.J.; Ali, M, Degradation of cellulosic insulation in power transformers. Part 2: formation of furan products in insulating oil, Science, Measurement and Technology, IEE Proceedings Volume: 147, Issue: 3, Page(s): 110 114, 2000 [129] Aminian, M.; Kadir, M.; Ahmad, W.; Mahmood, S., Investigation of paper aging assessment by furan concentration monitoring Research and Development (SCOReD), 2009 IEEE Student Conference on, Page(s): 360 362, 2009 [130] Emsley, A.M., Stevens, G.C., A Reassessment of the Low Temperature Thermal Degradation of Cellulose, IEEE 6th International Conference on, pp: 229232, 1992 [131] Rojas, A., Power Factor Testing in Transformer Condition Assessment -Is There a Better Way?, Transmission & Distribution Conference and Exposition: Latin America, 2006. TDC '06. IEEE/PES, Page(s): 1 4, 2006 [132] CIGRE Working Group D1. Knowledge Rules for Partial Discharge Diagnosis In Service, TECHNICAL BROCHURE NO. 226, Electra , Pages: 62-22, 2003 [133] A. Pfeffer, S. Coenen, S. Tenbohlen, S. M. Markalous, T. Strehl. Onsite experiences with multi-terminal IEC PD measurements and UHF PD measurements, Proceedings 17th ISH, No: C-51, Cape Town, South Africa, 2009 [134] Reid, A.; Judd, M.; Johnstone, C, Development of UHF transformer probe sensors for on-line partial discharge measurement, Universities Power Engineering Conference (UPEC), 2009 Proceedings of the 44th International, Page(s): 1 4, 2009 [135] Tenbohlen, S.; Denissov, D.; Hoek, S.; Markalous, S.M., Partial dischar ge measurement in the ultra high frequency (UHF) range, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 15 , Issue: 6, Page(s): 1544 1552, 2008 106

[136] S. Coenen, S. Tenbohlen, S. M. Markalous, T. Strehl: Fundamental characteristics of UHF PD probes and the radiation behaviour of PD sources in power transformer, Proceedings 17th ISH, Cape Town, South Africa, 2009 [137] M. D. Judd, L. Yang, C. J. Bennoch, and I. B. B. Hunter, UHF diagnostic monitoring techniques for power transformers, in EPRI Substation Equipment Diagnostics Conf. XII, New Orleans, LA, February1518, 2004 [138] S. Tenbohlen, S.M. Hoek, D. Denissov, R. Huber, U. Riechert, S.M. Markalous, T. Strehl, T. Klein Electromagnetic (UHF) PD Diagnosis of GIS, Cable Accessories and Oil-paper Insulated Power Transformers for Improved PD Detection and Localization, CIGRE 2006 [139] S.Coenen,M. Reuter, S. Tenbohlen, S. Markalous Influence of PD Location in Transformer windings on IEC60270- and UHF-Measurements, Conference Proceedings of CMD2010, Tokio, 2010 [140] Densley, J., Ageing mechanisms and diagnostics for power cables - an overview, Electrical Insulation Magazine, IEEE Volume: 17, Issue: 1, Page(s): 14 22, 2001 [141] Hussein Anis, Roshdy Radwan, Mazen Abdel-Salam, Ahdab El Morshedy, High-Voltage Engineering: Theory and Practice, Second Edition, Revised and Expanded,, Book, Marcel Dekker, 2000 [142] Luton M-H., Anders G.J., Braun J.M., Fujimoto N., Rizzetto S., Downes J.A., Real Time Monitoring of Power Cables by Fibre Optic Technologies Tests, Applications and Outlook, JICABLE ' 03 - International Conference on Insulated Power Cables, paper A.1.6, Paris 2003 [143] Jarot Setyawan, Investigation of Partial Discharge Occurrence and Detect ability in High Voltage Power Cable Accessories, Master thesis, Delft University of technology, Nov 2009 [144] F.J. Wester, Condition Assessment of Power Cables Using PD Diagnosis at Damped AC Voltages, ISBN 90-8559-019-1, Ph.D. thesis TU Delft, 2004. [145] Montanari, G.C.; Cavallini, A.; Puletti, F., A new approach to partial discharge testing of HV cable systems, Electrical Insulation Magazine, IEEE Volume: 22, Issue: 1, Page(s): 14 23, 2006

107

[146] Boggs, S.; Densley, J., Fundamentals of partial discharge in the context of field cable testing, Electrical Insulation Magazine IEEE Volume: 16, Issue: 5, Page(s): 13 18, 2000 [147] Smit, J.J.; Gulski, E., Advanced condition assessment of high voltage power cables, Electrical Insulating Materials, 2005. (ISEIM 2005). Proceedings of 2005 International Symposium on, Volume: 3, Page(s): 869 - 872 Vol. 3, 2005 [148] Kheng Jern KHOR, Partial Discharge Propagation and Sensing in Overhead Power Distribution Lines, Master Thesis, School of Electrical and Computer Engineering College of Science, Engineering & Health RMIT University, March 2010 [149] Oliveira, S.C.; Fontana, E.; Cavalcanti, F.J.M., Real time monitoring of the leakage current of 230 kV insulator strings under washing Transmission and Distribution Conference and Exposition, 2008. T&D. IEEE/PES, Page(s): 1 4, 2008 [150] Bradley, R., Cable temperature monitoring, Distribution and Transmission Systems (Digest No. 1997/050), IEE Colloquium on Operational Monitoring of, Page(s): 3/1 - 3/2, 1997 [151] Nitz, S.; Iwainsky, A., Design and utilization of fiber networks for temperature monitoring along cable lines, Industrial Electronics Society, 1998. IECON '98. Proceedings of the 24th Annual Conference of the IEEE, Volume: 3, Page(s): 1824 1828 vol.3, 1998 [152] Kortschinski, J.; Leslie, J.R., A Power-Cable Temperature Monitoring System, Power Apparatus and Systems, IEEE Transactions on Volume: PAS-89 , Issue: 7, Page(s): 1429 1433, 1997 [153] Densley, J., An overview of aging mechanisms and diagnostics for extruded power cables Power Engineering Society Winter Meeting, 2000. IEEE Volume: 3, Page(s): 1587 - 1592 vol.3, 2000 [154] Puletti, F.; Olivieri, M.; Cavallini, A.; Montanari, G.C., Localization of partial discharge sources along HV and MV cable routes, Power Engineering Conference, 2005. IPEC 2005. The 7th International, Page(s): 1 199, 2005 [155] Mashikian, M.S.; Bansal, R.; Northrop, R.B., Location and characterization of partial discharge sites in shielded power cables Power Delivery, IEEE Transactions on Volume: 5 , Issue: 2, Page(s): 833 839, 1990

108

[156] Chengke Zhou; Hepburn, D.M.; Michel, M.; Xiaodi Song; Guobin Zhang, Partial discharge monitoring in medium voltage cables, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 1021 1024, 2008 [157] Michel, Matthieu, Comparison of off-line and on-line partial discharge MV cable mapping techniques, Electricity Distribution, 2005. CIRED 18th International Conference and Exhibition on, Page(s): 1 6, 2005 [158] Cavallini, A.; Montanari, G.C.; Puletti, F., Partial Discharge Analysis and Asset Management: Experiences on Monitoring of Power Apparatus, Transmission & Distribution Conference and Exposition: Latin America, 2006. TDC '06. IEEE/PES, Page(s): 1 6, 2006 [159] Tian, Y.; Lewin, P.L.; Wilkinson, J.S.; Sutton, S.J.; Swingler, S.G., Continuous on-line monitoring of partial discharges in high voltage cables, Electrical Insulation, 2004. Conference Record of the 2004 IEEE International Symposium on, Page(s): 454 457, 2004 [160] Gulski, E.; Cichecki, P.; Wester, F.; Smit, J.; Bodega, R.; Hermans, T.; Seitz, P.; Quak, B.; Vries, F., On-site testing and PD diagnosis of high voltage power cables Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 15 , Issue: 6, Page(s): 1691 1700, 2008 [161] Working Group D1.33 Task Force 05, Experiences In Partial Discharge Detection, Technical brochure, ELECTRA, No. 208, pp: 34-43, Jun 2003 [162] IEEE Guide for Partial Discharge Testing of Shielded Power Cable Systems in a Field Environment, IEEE Std 400.3-2006, Page(s): c1 36, 2007 [163] Phung, B.T.; Liu, Z.; Blackburn, T.R.; McMullan, P.; Burgess, G, Practical experience in on-line partial discharge monitoring of power cables, Power Engineering Conference, 2008. AUPEC '08. Australasian Universities, Page(s): 1 5, 2008 [164] B.T. Phung, Z. Liu, T.R. Blackburn, H. Zhang, P., McMullan, G. Burgess, and A. Zargari, On-line monitoring of MV and HV distribution cables using VHF partial discharge detection, paper D1-202, CIGR, Paris, 2008. [165] L Renforth, R Mackinlay, M Seltzer-Grant, R Shuttleworth, On-line Partial Discharge (PD) Spot Testing and Monitoring of High Voltage Cable Sealing Ends, Session 2008: CIGRE 2008; PARIS. Cigre; 2008. 109

[166] Tian, Y.; Lewin, P.L.; Wilkinson, J.S.; Schroeder, G.; Sutton, S.J.; Swingler, S.G., An improved optically based PD detection system for continuous on-line monitoring of HV cables, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 12 , Issue: 6, Page(s): 1222 1234, 2005 [167] Wu, R.-N.; Chang, C.-K., The Use of Partial Discharges as an Online Monitoring System for Underground Cable Joints Power Delivery, IEEE Transactions on Volume: PP , Issue: 99, Page(s): 1, 2011 [168] Van der Wielen, P.C.J.M.; Steennis, E.F., On-line PD monitoring system for MV cable connections with weak spot location, Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century, 2008 IEEE, Page(s): 1 8, 2008 [169] Van der Wielen, P.C.J.M.; Veen, J.; Wouters, P.A.A.F.; Steennis, E.F., Sensors for on-line PD detection in MV power cables and their locations in substations Properties and Applications of Dielectric Materials, 2003. Proceedings of the 7th International Conference on Volume: 1, Page(s): 215 - 219 vol.1, 2003 [170] Stone, G.C., Partial discharge measurements to assess rotating machine insulation condition: a survey Electrical Insulation, 1996, Conference Record of the 1996 IEEE International Symposium on, Page(s): 19 - 23 vol.1, 1996 [171] Fenger, M.; Stone, G.C.; Lloyd, B.A., Experience with continuous partial discharge monitoring of stator windings, Electrical Insulation Conference and Electrical Manufacturing & Coil Winding Conference, 2001. Proceedings, Page(s): 417 421, 2001 [172] P. Tavner, Peter. T, Li Ran, Jim Penman and Howard Sedding, Condition Monitoring of Rotating Electrical Machines, book, The Institution of Engineering and Technology, London, UK 2008 [173] Smeeton, P.; Bousbaine, A., Fault diagnostic testing using Partial Discharge measurements on High Voltage rotating machines, Universities Power Engineering Conference (UPEC), 2009 Proceedings of the 44th International, Page(s): 1 5, 2009 [174] Stone, G.C.; Lloyd, B.; Sasic, M., Experience with continuous on -line partial discharge monitoring of generators and motors, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 212 216, 2008 [175] Paolitti, G.; Golubev, A., Partial discharge theory and technologies related to traditional testing methods of large rotating apparatus, Industry Applications 110

Conference, 1999. Thirty-Fourth IAS Annual Meeting. Conference Record of the 1999 IEEE, Volume: 2, Page(s): 967 - 981 vol.2, 1999 [176] Paoletti, G.J.; Rose, A., Improving existing motor protection for medium voltage motors, Industry Applications, IEEE Transactions on Volume: 25 , Issue: 3, Page(s): 456 464,1989 [177] IEEE Committee Report, Report of Large Motor Reliability Survey of Industrial and Commercial Installations, Part I, Industry Applications, IEEE Transactions on Volume: IA-21, Issue: 4, Page(s): 853 864, 1985 [178] Tetrault, S.M.; Stone, G.C.; Sedding, H.G., Monitoring partial discharges on 4kV motor windings, Industry Applications, IEEE Transactions on Volume: 35, Issue: 3, Page(s): 682 688, 1999 [179] Zhu, H.; Green, V.; Sasic, M., Identification of stator insulation deterioration using on-line partial discharge testing, Electrical Insulation Magazine, IEEE, Volume: 17, Issue: 6, Page(s): 21 26, 2001 [180] Farahani, M.; Gockenbach, E.; Borsi, H.; Kaufhold, M., A method for the evaluation of insulation systems for high voltage rotating machines, Properties and Applications of Dielectric Materials, 2003. Proceedings of the 7th International Conference on Volume: 3, Page(s): 1108 - 1111 vol.3, 2003 [181] Hoof, M.; Lanz, S., PD diagnostics on rotating machines. Possibilities and limitations, Electrical Insulation Conference and Electrical Manufacturing & Coil Winding Conference, 1999. Proceedings, Page(s): 195 200, 1999 [182] August Johannes Marie Pemen, Detection of Partial Discharge in Stator Windings of Turbine Generators, Doctoral Thesis, Eindhoven University of Technology, 2000 [183] Cavallini, A.; Montanari, G.C.; Ciani, F.; Folesani, M., Artificial intelligence algorithms for the recognition of PD-generating defects in rotating machines. Electrical Insulating Materials, 2005. (ISEIM 2005). Proceedings of 2005 International Symposium on Volume: 2, Page(s): 451 - 454 Vol. 2, 2005 [184] Tavner, P.J., Review of condition monitoring of rotating electrical machines, Electric Power Applications, IET Volume: 2, Issue: 4, Page(s): 215 247, 2008

111

[185] Sedding, H.; Brown, A., Review of effective diagnostic and condition monitoring methods for rotating machines, Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century, 2008 IEEE, Page(s): 1 3, 2008 [186] Trzynadlowski, A.M.; Ghassemzadeh, M.; Legowski, S.F., Diagnostics of mechanical abnormalities in induction motors using instantaneous electric power, Energy Conversion, IEEE Transactions on Volume: 14, Issue: 4, Page(s): 1417 1423, 1999 [187] Bong-Keun Oh; Hyun-Il Kim; Dong-Sik Kang; Kee-Joe Lim, Comparison of on-line and off-line PD measuring system for hydro generator stator winding, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 204 207, 2008 [188] Green, V.; Sasic, M.; Wright, B., Summary of industry practices, guides and standards related to on-line partial discharge monitoring of stator insulation condition Petroleum and Chemical Industry Conference, 2005. Industry Applications Society 52nd Annual, Page(s): 359 365, 2005 [189] IEEE Trial-Use Guide to the Measurement of Partial Discharges in Rotating Machinery, IEEE Std 1434-2000, 2000 [190] Sedding, H.G.; Campbell, S.R.; Stone, G.C.; Klempner, G.S., A new sensor for detecting partial discharges in operating turbine generators, Energy Conversion, IEEE Transactions on, Volume: 6 , Issue: 4, Page(s): 700 706, 1991 [191] Stone, G.C.; Lloyd, B.A.; Campbell, S.R., On-line monitoring for condition assessment of motor and generator stator windings, Pulp and Paper Industry Technical Conference, 1994., Conference Record of 1994 Annual, Page(s): 94 103, 1994 [192] Okubo, H., Recent activity and future trend on ageing characterist ics of electrical insulation in GIS from manufacturer's view point, Properties and Applications of Dielectric Materials, 1994., Proceedings of the 4th International Conference on Volume: 2, Page(s): 837 - 840 vol.2, 1994 [193] Reid, A.J.; Judd, M.D.; Fouracre, R.A.; Stewart, B.G.; Hepburn, D.M., Simultaneous measurement of partial discharges using IEC60270 and radiofrequency techniques, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 18 , Issue: 2, Page(s): 444 455, 2011

112

[194] Hampton, B., UHF diagnostics for gas insulated substations, High Voltage Engineering, 1999. Eleventh International Symposium on (Conf. Publ. No. 467) Volume: 5, Page(s): 6 - 16 vol.5, 1999 [195] Singha, S.; Thomas, M.J., Toepler's spark law in a GIS with compressed SF6N2 mixture, Dielectrics and Electrical Insulation, IEEE Transactions on Volume: 10, Issue: 3, Page(s): 498 505, 2003 [196] Joint Working Group 33/23.12 Insulation co-ordination of GIS: Return of experience, on site tests and diagnostic techniques, Electra No. 176, Page(s):67 97, Feb 1998 [197] Meijer, S.,Partial Discharge Diagnosis of High-Voltage Gas-Insulated Systems, Doctoral Thesis, Delft University of Technology, 2001 [198] Judd, M.D.; Farish, O.; Hampton, B.F., Broadband couplers for UHF detection of partial discharge in gas-insulated substations, Science, Measurement and Technology, IEE Proceedings - , Volume: 142 , Issue: 3, Page(s): 237 243, 1995 [199] Hampton, B.F.; Meats, R.J., Diagnostic measurements at UHF in gas insulated substations, Generation, Transmission and Distribution, IEE Proceedings C, Volume: 135 , Issue: 2, Page(s): 137 145, 1988 [200] Baumgartner, R.; Fruth, B.; Lanz, W.; Pettersson, K., Partial discharge. IX. PD in gas-insulated substations-fundamental considerations Electrical Insulation Magazine, IEEE Volume: 7, Issue: 6, Page(s): 5 13, 1991 [201] Kurrer, R.; Klunzinger, K.; Feser, K.; de Kock, N.; Sologuren, D., Sensitivity of the UHF-method for defects in GIS with regard to on-line partial discharge detection, Electrical Insulation, 1996., Conference Record of the 1996 IEEE International Symposium on Volume: 1, Page(s): 95 - 98 vol.1, 1996 [202] Meijer, S.; Smit, J.J.; Gulski, E.; Girodet, A., Partial discharge early warning system in gas-insulated switchgear Transmission and Distribution Conference and Exhibition 2002: Asia Pacific. IEEE/PES Volume: 2, Page(s): 931 - 936 vol.2, 2002 [203] Baumgartner, R.; Fruth, B.; Lanz, W.; Pettersson, K., Partial discharge. X. PD in gas-insulated substations-measurement and practical considerations Electrical Insulation Magazine, IEEE Volume: 8 , Issue: 1, Page(s): 16 27,1992

113

[204] Hampton, B.F., Diagnostics for gas insulated substations, Advances in Power System Control, Operation and Management, 1993. APSCOM-93., 2nd International Conference on, Page(s): 17 - 23 vol.1, 1993 [205] Meijer, S.; Smit, J.J.; Kanters, A.J.L.M., Experience with PD measurements in GIS maintenance concept, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 551 554, 2008 [206] Golubev, A.; Blokhintsev, I.; Paoletti, G.; Modrowski, J., On-line partial discharge applications to MV electrical switchgear Electrical Insulation Conference and Electrical Manufacturing & Coil Winding Conference, 2001. Proceedings, Page(s): 225 232, 2001 [207] Fung, K.Y.; Sit, K.L.; Chan, L.K.; So, K.L., Condition monitoring on HV equipment in CLP Power, Condition Monitoring and Diagnosis, 2008. CMD 2008. International Conference on, Page(s): 679 683, 2008 [208] Kim, J.B.; Kim, M.S.; Kim, T.W.; Song, W.P.; Kim, D.S.; Park, W.Z., Development of smart diagnosis system for 800 kV gas -insulated switchgear, Transmission and Distribution Conference and Exposition, 2003 IEEE PES Volume: 2, Page(s): 486 - 490 vol.2, 2003 [209] J. Vigreux (on behalf of WG 23.03), Application of Condition Monitoring Technoques in Gas Insulated Substations, CIGRE ELECTRA, Feb, 1991 [210] Working group 15.03, Partial discharge detection system for GIS: sensitivity verification for the UHF method and the acoustic method, Electra No. 183, Page(s): 75-87, APRIL 1999 [211] Hoshino, T.; Koyama, H.; Maruyama, S.; Hanai, M., Comparison of sensitivity between UHF method and IEC 60270 for onsite calibration in various GIS, Power Delivery, IEEE Transactions on, Volume: 21, Issue: 4, Page(s): 1948 1953, 2006 [212] Lundgaard, L.E.; Runde, M.; Skyberg, B., Acoustic diagnosis of gas insulated substations: a theoretical and experimental basis, Power Delivery, IEEE Transactions on Volume: 5, Issue: 4, Page(s): 1751 1759, 1990 [213] Menno M.P. de Haas, Online Diagnostics in Smart Grids- An approach to the Future Power Grid-, Master Thesis, Delft University of Technology, June 2011.

114

WEBSITE

[214] www.doble-lemke.eu [215] www.omicron.at [216] www.techimp.com [217] www.ipec.co.uk [218] www.pd-systems.com [219] www.kema.com [220]www.irispower.com [221] www.qualitrolcorp.com [222] www.psdtech.com [223] www.nexigmaint.com [224] www.dynamicratings.com [225] www.emersonnetworkpower.com [226] www.hvpd.co.uk

115

Appendix 1: CASE STUDY 1

On-line PD monitoring on Rotating Machine On-line periodic PD measurement was carried out at Leuna Saalekreis, SaxonyAnhalt, Germany. The generator specification is shown below Product name and manufacturer: GEC Alstorm T600C Rated power: 56471 MVA Rated terminal voltage: 10.5kV

System Configuration An on-line PD monitoring system roughly consists of three capacitive sensors, data acquisition unit, and personal computer shown below in Figure 1 and 2

Figure 1 On-line PD monitoring system configuration

System description

1. Capacitive coupler at U (Bandwidth: 20 MHz, Rated capacitance: 2 nF) 2. Capacitive coupler at V (Bandwidth: 20 MHz, Rated capacitance: 2 nF) 116

3. Capacitive coupler at W (Bandwidth: 20 MHz, Rated capacitance: 2 nF) 4. Generator terminal 5. Data acquisition unit 6. Sensor signal collector 7. Power supply (50Hz, 230V) 8. Optic fibre (Yellow cable) to personal computer

Three capacitive couplers are permanently installed at each generator terminal phase (U, V, W) which are connected to a sensor signal collector installed outside the generator. The data acquisition unit can be connected to a sensor directly via a sensor signal collector. The data acquisition unit also has a voltage reference input for PRPD analysis and possible gate noise reduction input if it is needed. Optical fibre (yellow cable) directly connects the data acquisition unit and signal convertor in order to send a signal using an Ethernet connection so that it can be connected to a personal computer in which analysis and recording of the signal takes place using a User Interface (UI).

Calibration The calibration procedure was performed

according to IEC 60270. The Calibrator supplies a pulse periodically to the sensors monitored by PC on each phase. This can check also connectivity between all equipment by detecting exact pulse signal from the sensors. Plus, it contains the signal attenuation characteristics from sensor to PD. In this case, 2000nC pulses for each measuring frequency band (100500kHz, 560-3000kHz, and 9500- 10500kHz) were injected at each phase to the sensor. The calibration system inside the generator is shown
Figure 2 Calibration procedure

117

in Figure 2, highlighting the pulse injector. The arrows indicate connection spots between the pulse injector and sensor. The recommended calibration acceptable error is within 5% of the reference calibration pulse value.

Measurement The measurement was carried out while the generator was in operation in different frequency ranges; 100 500 kHz (IEC recommended), 560 3000 kHz and 950010500 kHz based on previous measurement record. In this thesis, 100-500 kHz measuring graphs are covered. In PRPD pattern graph, a reddish color indicates higher repetition rates compared to a blue or dark color which means low a repetition rate.

118

Figure 3 PD occurrence on each phase visualized as PRPD pattern and pulse

The graph shows a clear higher peak value of PD on the first phase compared to other phases. One noticeable thing here is the noise signal. Especially at the third phase, there are phase-lock noises, periodic red dot pattern from the voltage reference. If the noise is severe, then the gating technique can eliminate those noises. The pattern in this measurement should be compared to a PD pattern reference library for more accurate decision making.

119

Appendix 2: CASE STUDY 2

AE detection on transformer The acoustic emission detection was performed on the transformer at Siemens Transformer Factory, Dresden, Germany The transformer specification is shown below Product name and manufacturer: Dewa D417371 Rated voltage: 145kV/12kV

System Configuration An AE detection system roughly consists of acoustic sensors, data acquisition unit, and personal computer shown in Figure 4.

Figure 4 AE detection system on power transformer

System description

1. Acoustic sensor (total 8 sensors used) 2. Data acquisition unit (amplifier, 8 sensor signal input) 120

3. Personal Computer (Data analysis) The transformer was being inspected due to a high level of PD. In the measuring system, there are 8 acoustic sensors on the transformer applied test voltage, data acquisition unit, and PC which was specially designed for the purpose of AE measurement. There are amplifiers on each input in the data acquisition unit which improve the measurement signal.

Measurement The end user of the transformer reported abnormal conditions based on a high level of PD. However, there was no indication of the location of the PD source inside the transformer. Based on their experience and similar cases, the most likely parts of the transformer for fault are the cable connection box and transformer terminal for the cable shown in the white box in Figure 5. Especially the cable connection box inside the red rectangle is the most likely. Therefore the measurement started by placing sensors around the red box. Even though there was no significant signal detected by the sensor in this area, the signal detected by lower sensors was stronger than the upper sensors.

Figure 5 Sensor placements in the suspicious part of the transformer

121

While moving the sensors downward to find more AE signals, the strongest signals were emitted inside the blue box shown in the Figure 2. The signal from each sensor is shown in Figure 4. The graph shows eight different sensors signals simultaneously in two different frames (1 to 4 in the upper graph and 5 to 8 in the lower graph). The strong signals come from the first and fifth sensor. As indicated on the x-axis, there are clear time differences between sensors according to their placement on the transformer cable connection box. Nevertheless the magnitude can vary according to the propagation path and insulation material structure inside of the test object, the pulse arrival time clearly depends on the sensor placement.
1.2 1 0.8 0.6 0.4
Voltage [V]

g b c d e f b c d e f g b c d e f g b c d e f g

Sensor 1 Sensor 2 Sensor 3 Sensor 4

0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6 6.2 6.4 6.6 6.8 7 7.2 7.4 7.6 7.8 8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10 Time [ms]

1.2 1 0.8 0.6 0.4

Voltage [V]

g b c d e f b c d e f g b c d e f g b c d e f g

Sensor 5 Sensor 6 Sensor 7 Sensor 8

0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6 6.2 6.4 6.6 6.8 7 7.2 7.4 7.6 7.8 8 8.2 8.4 8.6 8.8 9 9.2 9.4 9.6 9.8 10 Time [ms]

Figure 6 Signals from 8 sensors located different spot around suspicious PD source

PD location and Signal time frame After gaining enough signal strength from each sensor, the coordinate of each sensor was recorded and analysed regarding the location of the PD signal according to the time difference of pulse propagation. In order to calculate the PD source, the signal can be fixed at the PD occurrence time.

122

Figure 7 Time frame of the first sensor signal

Figure 8 Time frame of the fifth sensor signal

In Figure 7 and 8, time frame work was when the PD magnitude and pulse shape are the most likely right time. The same process was done for all eight sensors for setting the exact propagation time from possible PD sources to each sensor placement.

123

Figure 9 Coordinate in the cable connection box

As shown in Figure 9, it was assumed the 3 axis runs along to the contour of the cable and connection box and pick the coordinates of each sensor placed inside of the scope (e.g. (x, y, z)). The software in the computer analyzes possible fault locations inside the rectangular space. As we can see in Figure 10, possible PD source can be calculated according to the AE detection techniques in the transformer, which is in detail at 4.1.4. On-line transformer PD monitoring on transformer. The options for this calculation are acoustic speed in oil dependent on temperature and the selectivity of sensors.

Figure 10 localization PD source according to the pulse propagation time difference from 8 sensors

124

Appendix 3: CASE STUDY 3

On-site switchgear PD monitoring On-site PD monitoring according to the IEC 60270 on switchgear in Altenberg substation, Germany. Especially the voltage transformer in switchgear is the measurement target. The switchgear specification is shown below Product name: GMB 24 16 06/ SF6 Rated voltage and current: 24 kV 630A (50 Hz) Test object: 2 block 14 Cubicles

System Configuration An on-site PD monitoring system on switchgear roughly consists of Measuring impedance, voltage supplier and auto transformer, data acquisition unit, and personal computer shown below in Figure 11

Figure 11 On- site PD monitoring system on switchgear

125

Figure 12 Measuring impedance and calibrator

System description 1. Module reactor 2. Inductance (40mH) 3. Capacitance (0.3nF) 4. Switchgear terminal 5, 6. Low pass filter 7. Data acquisition unit 8. Autotransformer and controller 9. Personal Computer 10. Calibrator 11. Measuring Impedance The on-site measurement was performed by applying AC test voltage to the switchgear. The transformer and module reactor supply voltage from the tested object to the switchgear terminal. If there is high noise, a low pass filter which is adjustable in different frequency bands can be applied. A data acquisition unit collects data by measuring the impedance which is connected to a the personal computer for recording and analyzing. In order to avoid possible noise discharge at the other terminal of the 126

switchgear, there are an anti PD caps on other terminals which are grounded appropriately to make sure PD such as corona does not occur at other terminals shown in Figure12 in the white box.

Measurement Before the measurement, there was a calibration according to the IEC 60270 which injects a known pulse magnitude to the measuring impedance. The two different measurements performed for switchgear were; High voltage withstanding test and a PD test. According to the standard HV, testing was at a 40kV voltage level, which is 80% of the HV test voltage of 24kV, for 60 seconds according to the HV standard test. This test was only for the purpose of withstanding.

127

Figure 13 Measurement results

The upper Figure shows a PRPD graph of the PD and the lower Figure shows the applied voltage level. Since the rated voltage of the switch gear terminal is 24 kV, the test voltage was about 26kV which is 1.1 X U (nominal) voltage. The measurement result clearly shows that there are certain amounts of PD occurrence at a and b. Since the PD magnitude in the first and second Figure above is within the dangerous threshold (50pC), the switchgears needed further action to fix, for example, to localize PD source. In the last Figure, periodic noises are shown at regular intervals.

128

Appendix 4: Comparison of on-line PD monitoring products for Transformer

Features\ products Specific features System type Sensor General Feature Compatible Standard Connection to PC Other issues PD-guard Electrical/ UHF (Doble Lemke) Conventional/ Unconventional Capacitive, UHF sensor IEC 60270 Optical Fibre SCADA, AE detection compatible 10kHz -2000khz (Capacitive) 100MHz1000MHz (UHF) 50 ohm Up to100,000 pC Up to 700mV <0.3 degree 12 bit bipolar 4 inputs for PD& synchronization signal Gating Windowing Multi terminal Measurement Auto alarming PRPD analysis PDcheck (Techimp Energy Srl) Conventional/ Unconventional Capacitive, HFCT, TEV, Horn sensor IEC 60270 Optical Fibre DGA, tan, VIB, DTS integrations option 16kHz -30MHz Up to 2 GHz(with Frequency shifter) 50 ohm Up to 4000mVpp <1 degree 10 bits 3 channels for PD & synchronization channel TF map (Time/ Frequency Map), Fuzzy logic based noise elimination Icmmonitor (Powerdiagnostix System) Conventional/ Unconventional Capacitive, UHF Sensor IEC 60270 Optical Fibre 2-20MHz (Standard) 40-800kHz (optional) 300-2000MHz (UHF) 50 ohm DMS PDMT (Qualitrol LLC) Unconventional UHF sensor Master/Slave connection PD-TM series (PowerPD, Inc.) Unconventional AE sensor, HFCT ANSI/TIA/EIA-422-B (for communication) Optical Fibre SCADA compatible PD-EYE (IPEC Ltd) Conventional/ Unconventional HFCT, AA transducer, Capacitive 50kHz-20MHz (HFCT) 20MHz-800MHz (Capacitive) 40kHz1kHz(AE) Roc. 50 ohm 12 bit DTM (Dynamic Ratings) Conventional/ Unconventional HFCT, Bushing sensor, AE (optinal) IEC 61850 (for communication) Optical Fibre SCADA compatible can be combined with bushing monitoring

PD acquisition Frequency Range Input impedance PD detection Range Phase Accuracy PD signal resolution Input channel

100-3000MHz (UHF)

80KHz~300KHz (AE) 100KHz~30MHz (HFCT) 5 channel (AE-4ch,HFCT-1ch) 10 channel (AE-8ch, HFCT-2ch)

3-6 channel (up to 250 channel with Master/Slave connection) Neural Network Genetic Algorithm Fuzzy logic SMS/E-mail alarm

14 channels for PD 4 channels for AE

Electrical Feature

8 inputs for PD

8 channels

Software Feature

De-noising Techniques

Other issues

Data integrated web, E-mail/SMS alarm

129

Appendix 5: Comparison of on-line PD monitoring products Cable

Features\ products Specific features System type Sensor PD-guard Electrical/ UHF (Doble Lemke) Unconventional Capacitive, UHF sensor Smart Cable Guard (KEMA) Unconventional HFCT Emerson Unconventional HFCT PDcheck (Techimp Energy Srl) Unconventional HFCT, Flexible Magnetic Coupler TDR with GPS, Arrival time Analysis, amplitude& frequency analysis HVPD Mini (HVPD) Unconventional HFCT, TEV Two HFCTs and Pulse injection method PD-EYE and PrecisePD (IPEC Ltd) Unconventional HFCT, AA transducer, Capacitive Distinguish arrival time of TEV signal from two ends PD-MCC&G400A (PowerPD) Unconventional 4 HFCTs

General Feature

Locating PD source

Two HFCTs and Pulse injection ~4km (PILC, MIND) ~8km (XLPE) ~2km (EPR) 200 kHz - 20MHz (HFCT) 6kV - 36kV 3 input for sensors (control unit) O

Time of flight analysis

Maximum PD localizing length

~4.8km

~2km

50kHz-20MHz (HFCT) 20MHz-800MHz (Capacitive) 40kHz1kHz(AE) Roc. 50 ohm 12 bit 8 channels O

PD acquisition Frequency Range Input impedance PD detection Range Voltage class PD signal resolution Input channel Two ends method De-noising Techniques

10kHz -2000khz (Capacitive) 100MHz1000MHz (UHF) 50 ohm Up to100,000 pC Up to 700mV 12 bit bipolar 4 inputs for PD& synchronization signal Gating Windowing Multi terminal Measurement Auto alarming PRPD analysis

10khz - 300Mhz (HFCT) 4kV - 345kV. Built-in RF noise reduction

2 MHz - 100 MHz (HFCT) 50 ohm Up to 4000mVpp 10 bits 3 channels for PD & synchronization channel TF map (Time/ Frequency Map), Fuzzy logic based noise elimination Web MSG service

100KHz~10, 12, 20MHz (HFCT) according to the type -50 ohm 3.3 kV to 45 kV 4 channel O PD pulse shape analysis based on previous record Web based application

4 channels for HFCT -

Electrical Feature

Matched filter bank

Software Feature

Other issues

Early warning and Alerts

Data integrated iSM web, E-mail/SMS alarm

PRPD analysis

130

Appendix 6: Comparison of on-line PD monitoring products for RM

Features\ products

Specific features System type Sensor

PD-guard Electrical/ UHF (Doble Lemke) Unconventional Capacitive

Iris Power Unconventional Capacitive (80pF) SSC (6 to 9 as a set, 6kV or higher) Load, Temperature Active/Reactive power Coaxial Cable 10-1000Mhz (SSC) 40-350Mhz (Capacitive) 50ohm (SSC) -

Longshot (HVPD) Unconventional Capacitive HFCT TEV 0 - 400Mhz (HFCT) 4MHz-100MHz (TEV) 4 channel

PDcheck (Techimp Energy Srl) Unconventional Capacitive coupler HFCT -

Icmmonitor (Powerdiagnostix System) Conventional/ Unconventional Capacitive Load or Temperature possible Optical Fibre 2-20MHz (Capacitive)

MICAMAXXpda (PDtech) Unconventional Capacitive HFCT -

PD-MCC&G400A (PowerPD) Unconventional 4 HFCTs

General Feature

Combined measurement Connection to PC PD acquisition Frequency Range Input impedance

Optical Fibre 10kHz -2000khz (Capacitive) 100MHz1000MHz (UHF) 50 ohm Up to100,000 pC Up to 700mV 12 bit bipolar 4 inputs for PD& synchronization signal Gating Windowing Multi terminal Measurement Auto alarming PRPD analysis

2 MHz - 100 MHz

50 ohm Up to 4000mVpp 10 bits 3 channels for PD & synchronization channel TF map (Time/ Frequency Map), Fuzzy logic based noise elimination SCADA compatible PRPD

50 ohm 8 inputs for PD

Roc. 50 ohm 12 bit 3 channels

4 channels for HFCT

Electrical Feature

PD detection Range PD signal resolution Input channel

Software Feature

De-noising Techniques

Pulse shape analysis Time of arrival High pass filter Alarm PRPD

Wavelet denoising Time of arrival

Other issues

PRPD

SCADA compatible PRPD Warning

PRPD analysis

131

Appendix 7: Comparison of on-line PD monitoring products for GIS

Features\ products

Specific features System type

PD-guard Electrical/ UHF (Doble Lemke) Unconventional Inductive

Longshot (HVPD) Unconventional HFCT, TEV, Airborne AE, external capacitive Time-of-flight 4MHz to 100MHz (TEV) 100KHz to 50MHz (HFCT) 3.3-36kV -

Sensor General Feature Localization Method Connection to PC PD acquisition Frequency Range Input impedance Electrical Feature PD detection Range Voltage class PD signal resolution

Icmmonitor (Powerdiagnostix System) Unconventional Flange type, shielded ring antenna, external window type Ethernet cable

PDMG-R (DMS) Unconventional Internal or external window sensor Optical fibre

Amos 3.0 (PSD tech) Unconventional Internal or external of oper barrier, metalclosed barrier type -

AIA (Doble TransNor ) Unconventional Acoustic sensors(piezoelectric) Searching along the GIS Data cable 10-100kHz (recommended setting) 2-50pC(sensitivity) 145-800kV 1 BNC input and 1 preamplifier for sensor 1 input for external sync. Signal Filtering

PD-MAT400A (PowerPD) Unconventional

4 AE sonsors

Optical Fibre

100MHz to 1GHz

300- 2000MHz (window type) 50 ohm 8 inputs for PD Coving one or two bay

500-1500MHz (window type) 3 channel for OCU & Max. 300 Channel for EC

0.5-1.5GHz (for all sensors) -

100-300kHz

50 ohm Up to100,000 pC Up to 700mV 12 bit bipolar 4 inputs for PD& synchronization signal Gating Windowing Multi terminal Measurement Auto alarming PRPD analysis

Input channel

16 channel for PD 2 channel for noise

4 channels for AE

Software Feature

De-noising Techniques

Pulse shape analysis High pass filter Alarm PRPD

ANN, fuzzy, genetic algorithm SCADA/SCS Alarm as E-mail and SMS

Gating, Neural Network, Filtering Web-based PRPD, PRPS

Other issues

Threshold & Envelop Alarms

132

Appendix 8: Commercial Sensors

Electrical sensor Name (Company) PDDC-17/24 (Doble lemke) 1 nF Coupler (Techimp) Type A (Doble lemke) UHF Sensor drain valve (DN50/80) (doble lemke) HFCT-140 (Techimp) Flexible Magnetic Coupler (Techimp) TEV (Techimp) Horn Antenna (Techimp) TEM (Techimp)

Rated voltage Frequency bandwidth Sensor type Main application area Physical Figure

17.5kV/24kV

20kV

20MHz

100MHz~ 1GHz inductive

200MHz~ 1GHz UHF

2MHz~ 100MHz Inductive (HFCT) Cable Grounding rod bar 310x320x40 mm/6kg

30kHz~ 200MHz Inductive

0.1MHz~ 300MHz Capacitive

500MHz~ 3GHz Electromagnetic

100MHz~ 3GHz

Capacitive

Capacitive

Electromagnetic

Rotating Machine

Rotating Machine

Cable Termination

Transformer

Cable joint /terminal

Switchgear GIS

GIS/GIL Transformer

Switchgear Induction motor

2,5kg/7.5kg 139(H*)140(D**) 350(H)180(D) -

295 (H) 187(D)

85 (H) 103(D)

3kg

120 x 480 x 9 mm/330g

130x 70 x 25 mm/80g

70 x 100 x50 mm/260g

80 x 150 x 50 mm/ 250g

Load impedance Other issues

50 ohm

50 ohm

50 ohm

50 ohm

2nF (Capacitance)

1nF

2nF (With integrated capacitance)

10mV/mA (Sensitivity)

N-type

N-type

133

Electrical sensor Name (Company) MCC112/124/ 205/210 (Omicron) MCT100/110 (Omicron) UVS 610 (DN50/80) (Omicron) HFCT 100/50 (IPEC Ltd) Capacitive Coupled TEV Sensor (IPEC Ltd) Sensor/ Injector Unit SUI405 (KEMA) Rated voltage Frequency bandwidth Sensor type 12/24/50/100kV 80 ~ 5 MHz 150~1 GHz 50kHz~ 20MHz Inductive (HFCT) Earth conductor 20~ 800MHz 6-36kV 200kHz20MHz Capacitive TEV Switchgear Inductive (HFCT) Cable Transformer Reactor GIS Hydro generator stator windings 100 ~ 3000 MHz UHF 500~ 1500MHz UHF 25/16/6.9kV DMS PDMT UHF (Qualitrol) DMS PDMG-R UHF (Qualitrol) Epoxy-mica PDA coupler (Irispower)

Capacitive

Inductive (HFCT) Ground connection

UHF

Capacitive

Main application area Physical Figure

Generator MV network

Transformer

140x200x140mm/ 150x250x150mm/ 450x575x450mm/ 450x755x450mm/ 3/3.5/8.2/10.5kg

115 x 120 x 65 mm/ 110 x 120 x 55 mm/

3.1kg

50(internal D)110 (external D)mm/ 0.4kg

60 (D) mm 30mm/80g

1.4kg Dimensions 250x115x50 mm (L, W, D)

2.3/1.6/1.1kg

Load impedance Other issues

About 1nF (Capacitance)

Split/ one piece

50 ohm Split core

50 ohm Magnetic coupling

Maximum cable circuit length (PILC, MIND= 4km XLPE=8km EPR=2km)

Hatch cover Drain valve type Internal or external

80pF3pF

134

Electrical sensor Name (Company) Stator Slot Couplers (Irispower) UHF/internal window type (psdtech) UHF/external window type (psdtech) Epoxy-mica capacitive coupling DREMC(dynamic ratings) Rated voltage Frequency bandwidth Sensor type Inductive UHF UHF 10-1000 MHz 0.5~1.5 GHz 0.5~1.5 GHz 0.5 ~ 500 MHz Capacitive 8/16/28kV Max current: 300A 100kHz 20MHz Inductive (HFCT) Main application area Physical Figure gas or steam turbine generators. 2.0 mm thick length trimmable to 53 cm GIS GIS Motors ,gener ators and switch gear 86/126/185(H) 0.95/1.4/2.1 (kg) 0.4kg 45mm(Internal) 107mm(Externa l diameter) Load impedance Other issues Max. sensitivity < 2pC Max. sensitivity < 5pC 80 pF+ 3 pF (Capacitance) Sensitivity:1p C Different size of split-core type is available (140/100,220/1 50) Different size of split-core type is available (100/50) Time of flight is possible by using more than 2 TEV sensors 50 ohm 50 ohm 0.5kg 40x25mm(Internal) 90x120mm(Extern al dimension) 50 ohm 50 ohm 60 x 50 x 25mm, 0.2kg HV cable HV cable Inductive (HFCT) Capacitive TEV sensor Switchgear GIS Max current: 300A 100kHz -25MHz 1MHz - 50MHz HFCT(100/50) (HVPD), portable HFCT(75/35) (HVPD), permanent HVPD TEV Sensor (HVPD)

Acoustic sensor PAC D9241 AE sensor (Doble TransNor) Airborne Acoustic Transducer (IPEC Ltd)

20-60 kHz

40KHz 1KHz Acoustic

Acoustic

Switchgear

24x20mm/56 mg

120x 40mm /75g

Resonance frequency: 30kHz

10pC (Sensitivity)

*H= Height **D=Diameter

135