PRE-BREAKDOWN AND BREAKDOWN STUDY OF TRANSFORMER …

PRE-BREAKDOWN AND BREAKDOWN STUDY OF

TRANSFORMER OIL UNDER DC AND IMPULSE

VOLTAGES

A thesis submitted to The University of Manchester for the degree of

PhD

in the Faculty of Science & Engineering

2017

JING XIANG

School of Electrical and Electronic Engineering

3

Contents

CONTENTS

CONTENTS ......................................................................................................................... 3

LIST OF FIGURES ..................................................................................................................... 7

LIST OF TABLES ..................................................................................................................... 15

ABSTRACT ....................................................................................................................... 17

DECLARATION ....................................................................................................................... 19

COPYRIGHT STATEMENT ................................................................................................... 21

ACKNOWLEDGEMENT ......................................................................................................... 23

CHAPTER 1. INTRODUCTION ....................................................................................... 27

1.1 Background ................................................................................................... 27

1.2 Research Objectives ...................................................................................... 29

1.3 Major Contributions ...................................................................................... 30

1.4 Outline of Thesis ........................................................................................... 31

CHAPTER 2. LITERATURE REVIEW ........................................................................... 33

2.1 Introduction ................................................................................................... 33

2.2 Streamer and Breakdown in Liquids under DC Voltage ............................... 33

2.2.1 Methodologies .............................................................................. 33

2.2.2 Current and emitted light .............................................................. 35

2.2.3 Photography measurement ............................................................ 37

2.2.4 Breakdown properties in Liquids.................................................. 40

2.2.5 Space Charge ................................................................................ 41

2.3 Streamer and Breakdown in Liquids under Impulse Voltage ....................... 42

2.3.1 General Streamer Characteristics ................................................. 42

2.3.2 Streamer and Breakdown under Different Impulse Waveforms .. 48

2.4 Gas Generation in Liquids under Electrical Faults ....................................... 56

2.4.1 Background Knowledge of Fault Gas Analysis ............................ 56

4

Contents

2.4.2 Fault Gas Studies under Electrical Faults ..................................... 59

2.5 Summary ....................................................................................................... 64

CHAPTER 3. EXPERIMENTAL DESCRIPTION ......................................................... 67

3.1 Liquids under Investigation ........................................................................... 67

3.2 Sample Preparation ....................................................................................... 67

3.3 Etched Tungsten Needles Based on Electrochemical Method ...................... 68

3.3.1 Basic Principle of Electrochemical Etching ................................. 68

3.3.2 Etching Procedure of Tungsten Needles ....................................... 69

3.4 DC Voltage Tests .......................................................................................... 71

3.4.1 Experimental Setup ....................................................................... 71

3.4.2 Experimental Procedures .............................................................. 73

3.5 Impulse Voltage Tests ................................................................................... 73


3.5.2 Impulse Waveforms with Different Tail times ............................. 75


3.6 Gassing Behaviour Tests ............................................................................... 77


3.6.2 Fault Control and Data Acquisition System Design ..................... 79

3.6.3 Oil-loop System Design ................................................................ 81


3.7 Summary ....................................................................................................... 84

CHAPTER 4. STREAMER AND BREAKDOWN PROPERTIES OF

TRANSFORMER LIQUIDS UNDER DC VOLTAGE ......................................................... 87

4.1 Introduction ................................................................................................... 87

4.2 Effect of Tip Radius on Streamer Initiation Voltage .................................... 87

4.3 Basic Characteristics of Streamers ................................................................ 91

4.3.1 Positive Streamer .......................................................................... 91

5

Contents

4.3.2 Negative Streamer ........................................................................ 95

4.4 Effect of Gap Distance on Breakdown Voltage ............................................ 99

4.4.1 Breakdown Phenomena ................................................................ 99

4.4.2 Breakdown Tests at Gap Distances from 2 mm to 30 mm ......... 100

4.5 Summary ..................................................................................................... 103

CHAPTER 5. STREAMER AND BREAKDOWN PHENOMENA OF

TRANSFORMER LIQUIDS UNDER DIFFERENT IMPULSE WAVEFORMS ............ 105

5.1 Introduction ................................................................................................. 105

5.2 Pre-breakdown Characteristics .................................................................... 105

5.2.1 Stopping Length .......................................................................... 105

5.2.2 Average Propagation Velocity .................................................... 106

5.3 Breakdown Voltage ..................................................................................... 108

5.3.1 Breakdown Tests in the Mineral Oil ........................................... 108

5.3.2 Prediction of Breakdown Voltage .............................................. 110

5.3.3 Verification in the Synthetic Ester Liquid .................................. 113

5.4 Effect of Impulse Waveform on Streamer Characteristics .......................... 115

5.5 Summary ..................................................................................................... 117

CHAPTER 6. CORRELATIONS BETWEEN GAS GENERATION AND SPARKING

FAULT IN TRANSFORMER LIQUIDS UNDER LIGHTNING IMPULSE VOLTAGE ....

..................................................................................................................... 119

6.1 Introduction ................................................................................................. 119

6.2 Data Processing ........................................................................................... 119

6.2.1 Calculation of Dissolved Gas Generation ................................... 119

6.2.2 Gas-in-total Calculation .............................................................. 121

6.2.3 Energy Calculation ..................................................................... 122

6.3 DGA Results and Analysis .......................................................................... 124

6.3.1 Comparison of Hydrogen Measurements ................................... 124

6.3.2 Effect of Spark Numbers ............................................................ 125

6

Contents

6.3.3 Effect of Gap Distance ................................................................ 128

6.3.4 Effect of Voltage Levels ............................................................. 131

6.3.5 Correlation between Fault Gas Generation and Fault Energy .... 133

6.4 Summary ..................................................................................................... 136

CHAPTER 7. CONCLUSIONS AND FUTURE WORK .............................................. 139

7.1 Conclusions ................................................................................................. 139

7.1.1 General ........................................................................................ 139

7.1.2 Summary of Results and Main Findings .................................... 140

7.2 Future Work ................................................................................................ 141

REFERENCES ..................................................................................................................... 143

APPENDIX I LIST OF PUBLICATIONS ................................................................... 149

7

List of Figures

LIST OF FIGURES

Figure 2-1 Typical discharge pulse in a needle to plane gap in a mineral oil under negative

polarity at room temperature, d = 5 mm, V = 14 kV [37]. ................................................... 35

Figure 2-2 Streamer current and emitted light signal in transformer oil under DC voltage,

positive polarity, d = 20 mm, V = 40 kV [27]. ...................................................................... 36

Figure 2-3 Streamer current waveforms in Hexane taken with a point cathode (Top –

Negative polarity) and with a point anode (Bottom – Positive polarity), point-sphere

electrode, d = 4.75 mm [29]. ................................................................................................. 37

Figure 2-4 Experimental setup of time delay system for streamer photograph capture under

DC voltage (re-produced) [24]. ............................................................................................. 38

Figure 2-5 The photograph of a streamer with the numbers indicating the sequence in which

the pictures were taken; time interval between each film 200 ns; (a) streamer photograph by

the camera; (b) streamer current signal; (c) streamer stopping length [29]. ......................... 39

Figure 2-6 Shadowgraph images of streamer expansion in transformer oil as a function of

time and pressure in a needle-to-plane geometry with a gap of 2.5 mm under negative

polarity [45]. ......................................................................................................................... 40

Figure 2-7 Breakdown voltage of oil under DC voltages of both polarities [48]. ................ 41

Figure 2-8 Typical images of positive streamers in a mineral transformer oil. (A) ‘1st mode’;

(B) ‘2nd mode’; (C) ‘2nd

mode’; (D) ‘3rd + 2nd modes’; (E) ‘4th mode’ [61]. ................... 43

Figure 2-9 Typical examples of positive streamer with emitted light (upper trace, arb. unit.),

transient currents (middle trace) and voltage (lower trace) in cyclohexane, d = 1.8 mm, rp =

1.2 μm, (A) 1st mode, V = 12 kV, (B) 2

nd mode V = 20 kV [71]. ......................................... 44

Figure 2-10 Propagation modes at high voltage in a natural ester liquid. d = 10 cm. Upper

oscilloscope trace: applied voltage, lower: streamer current. (A) 2nd

mode, V = 120 kV; (B)

3rd

+ 2nd

modes, V = 120 kV; (C) 3rd

mode, V = 142 kV; (D) 4th

mode, V = 152 kV [57]. .. 46

Figure 2-11 The propagation process of positive 3rd

+2nd

mode streamer at high voltage just

below Va; Diala S4 ZX-I, d = 50 mm; the ground electrode is at the bottom edge of the

streamer images [72]. ............................................................................................................ 46

Figure 2-12 The typical propagation process of 4th

+3rd

mode negative streamers in the GTL

oil, d=50 mm; the ground electrode is at the bottom edge of the streamer images from frame

3 to frame 6 [72]. .................................................................................................................. 47

8

List of Figures

Figure 2-13 Typical shapes of the positive 4th

mode streamers: (a) neat white oil, (b) neat

white oil at reduced pressure, and (c) DMA, all images taken just before breakdown, d = 80

mm [47]. ................................................................................................................................ 48

Figure 2-14 Typical shapes of the negative 4th

mode streamers: (a) neat white oil at reduced

pressure, (b) neat white oil, and (c) TCE, all images taken just before breakdown [47]. ..... 48

Figure 2-15 Characteristic voltage for oil breakdown on an impulse waveform, where L is

gap distance, ti and tb are the times corresponding to initiation and breakdown [73]. ......... 49

Figure 2-16 The effect of tail time on breakdown voltage in transformer oil at different gap

distance, positive point-plane electrode, gap distance: (o): 2.54 cm; (): 5.08 cm; (Δ):

10.16 cm; (V): 14.61 cm [73]. .............................................................................................. 49

Figure 2-17 Comparison of breakdown voltage and acceleration voltage in mineral oil

under lightning and step impulse, negative polarity [42]. .................................................... 50

Figure 2-18 Comparison of stopping length between step impulse and lightning impulse

under positive polarity, d = 50 [22]. ..................................................................................... 51

Figure 2-19 Comparison of average propagation velocity between lightning impulse and

step impulse under positive polarity, d = 50 mm [22]. ......................................................... 51

Figure 2-20 Comparison of (a) Breakdown voltage VbLI

and (b) time to breakdown tbLI

under lightning impulse versus gap distance in liquids of group IV (PMO, NMO, PB) [23,

42]. ........................................................................................................................................ 53

Figure 2-21 Average breakdown velocities vbLI

versus gap distance in insulating liquids, (a)

under lightning impulse; (b) under step impulse [23]. .......................................................... 54

Figure 2-22 Comparison of breakdown voltage in mineral oil PMO under step and lightning

voltages in point-sphere geometry [23]. ............................................................................... 55

Figure 2-23 Typical breakdown cases in mineral oil under lightning impulse voltage (full

lines) and step impulse voltage (dotted lines) [23]. .............................................................. 55

Figure 2-24 Comparison of breakdown voltage in natural ester NE under step and lightning

impulse voltage in point-plane geometry [23]. ..................................................................... 56

Figure 2-25 key gases generation versus number of breakdowns in mineral oil, d = 4 mm

[90] ........................................................................................................................................ 61

Figure 2-26 Comparison of fault gases generation in various liquids after 90 breakdowns

[90] ........................................................................................................................................ 61

Figure 2-27 Fault gases generation in mineral oil under different voltage levels, arc duration

= 15 mins [91]. ...................................................................................................................... 62

9

List of Figures

Figure 2-28 Fault gas generation in mineral oil under different arc durations, V = 20 kV

[91]. ....................................................................................................................................... 62

Figure 2-29 Fault gases generation in different insulation oils under sparking fault (re-

produced plot based on results in Table 2-7) [92]. ............................................................... 63

Figure 2-30 Fault gases generation per unit fault energy (µL/J) in Gemini X and FR3 under

sparking fault, averaged over a group of three tests. TCG = total combustible gases [92]. . 64

Figure 3-1 Principle of static etching (green A-C) and dynamic etching (orange A-D) [101].

............................................................................................................................................... 68

Figure 3-2 Measurement of the radius of curvature of needles [102]. .................................. 69

Figure 3-3 Failure results of etched tungsten needles without applying oscillating method.

............................................................................................................................................... 70

Figure 3-4 Ideal etched tungsten needle with the tip radius of 50 µm. ................................. 70

Figure 3-5 Sketch of the experimental setup used in streamer tests under DC voltage. ....... 71

Figure 3-6 Perspex made cubic test cell used for investigating streamer and breakdown

under DC voltage. ................................................................................................................. 72

Figure 3-7 Sketch of test setup of impulse tests with different tail-time. ............................. 74

Figure 3-8 The photo of the compact solid-state switch based impulse generator. .............. 75

Figure 3-9 The different impulse waveforms with tail time ranging from 8 µs to 3200 µs, V

= 24 kV. ................................................................................................................................ 76

Figure 3-10 The diagram of the experimental setup for gas generation tests under electrical

faults. ..................................................................................................................................... 78

Figure 3-11 The photos of the experimental setup for gas generation tests under electrical

faults. ..................................................................................................................................... 78

Figure 3-12 The photo of the cubic shaped stainless steel test cell used for gassing

behaviour test, (a) test cell; (b) electrode configuration with the gap distance of 10 mm. ... 79

Figure 3-13 The flow chart of the automatic control system with voltage output and data

recording. .............................................................................................................................. 80

Figure 3-14 Description of time domain in Labview setting. ............................................... 81

Figure 3-15 Sealing performance based on pressure reading in oil-loop system. ................ 81

Figure 3-16 The comparison of energy generation of individual breakdowns with different

time interval. ......................................................................................................................... 83

Figure 3-17 The flow chart of the experimental procedure for gassing behaviours tests ..... 84

10

List of Figures

Figure 4-1 The current and emitted light signals of streamer initiation under DC voltage (a).

Mineral oil – positive polarity (b). Mineral oil – negative polarity(c). Synthetic ester –

positive polarity (d). Synthetic ester – negative polarity; d = 10 mm, r = 10 µm. ............... 88

Figure 4-2 Weibull distribution plot of streamer initiation results with various tip radii (r =

5, 10, 20 and 50 µm) under DC voltage, d = 10 mm. ........................................................... 89

Figure 4-3 Effect of tip radius on streamer initiation voltage under DC voltage, d = 10 mm

(plot based on 50% initiation voltage). ................................................................................. 90

Figure 4-4 Initiation field versus tip radius in the mineral oil and the synthetic ester liquid

under positive and negative polarities. .................................................................................. 91

Figure 4-5 Typical positive streamer propagation in the synthetic ester liquid under DC

voltage, d = 10 mm, r = 10 µm, V = 28 kV; (a) voltage, current and monitor signals, (b)

streamer propagation, corresponding to the signals in (a). ................................................... 92

Figure 4-6 Stopping length of streamers in the mineral oil and the synthetic ester liquid

under positive polarity, d = 10 mm, r = 10 µm (error bars stand for one standard deviation).

............................................................................................................................................... 93

Figure 4-7 Average propagation velocity versus applied DC voltage in the mineral oil and

the synthetic ester liquid under positive polarity, d = 10 mm, r = 10 µm (error bars stand for

one standard deviation). ........................................................................................................ 94

Figure 4-8 Positive streamer stopping length as a function of maximum apparent charge in

the mineral oil and the synthetic ester liquid. ....................................................................... 95

Figure 4-9 Typical negative streamer propagation in the synthetic ester liquid under DC

voltage, d = 10 mm, r = 10 µm, V = -55 kV; (a) voltage, current and camera monitor signals,

(b) streamer propagation, corresponding to the signals in (a) .............................................. 96

Figure 4-10 Stopping length of streamers in the mineral oil and the synthetic ester liquid

under negative polarity, d = 10 mm, r = 10 µm (error bars stand for one standard deviation).

............................................................................................................................................... 97

Figure 4-11 Average propagation of streamers in the mineral oil and the synthetic ester

liquid under negative polarity, d = 10 mm, r = 10 µm (error bars stand for one standard

deviation). ............................................................................................................................. 98

Figure 4-12 Negative streamer stopping length as a function of maximum apparent charge

in the mineral oil and the synthetic ester liquid. ................................................................... 99

Figure 4-13 Breakdown in the synthetic ester liquid under DC voltage, positive polarity; d =

10 mm, r = 10 µm, exposure time 2 µs; (a) voltage, current and monitor signals, (b)

streamer propagation, corresponding to the signals in (a). ................................................. 100

11

List of Figures

Figure 4-14 Breakdown in the synthetic ester liquid under DC voltage, negative polarity; d

= 10 mm, r = 10 µm, exposure time 2 µs; (a) voltage, current and monitor signals, (b)

streamer propagation, corresponding to the signals in (a). ................................................. 100

Figure 4-15 Weibull distribution plot of breakdown results in the mineral oil and the

synthetic ester liquid with different gap distances under DC voltage, (a) Gemini X, (b)

MIDEL 7131, r = 10 µm ..................................................................................................... 101

Figure 4-16 Effect of gap distance on breakdown voltage in the mineral oil and the

synthetic ester liquid under DC voltage, r = 10 µm; based on 50% breakdown voltages .. 102

Figure 5-1 Stopping length of streamers in the mineral oil under positive polarity; d = 10

mm, r = 10 µm; error bars stand for one standard deviation. ............................................. 106

Figure 5-2 Stopping length of streamers in the synthetic ester liquid under positive polarity;

d = 10 mm, r = 10 µm; error bars stand for one standard deviation. .................................. 106

Figure 5-3 Average propagation velocity of streamers in the mineral oil under positive

polarity; d = 10 mm, r = 10 µm; error bars stand for one standard deviation. ................... 107

Figure 5-4 Average propagation velocity of streamers in the synthetic ester liquid under

positive polarity; d = 10 mm, r = 10 µm; error bars stand for one standard deviation. ...... 107

Figure 5-5 Weibull plot of breakdown voltages in the mineral oil under different impulse

waveforms, d = 10 mm, r = 10 µm, positive polarity. ........................................................ 108

Figure 5-6 Typical breakdowns in the mineral oil under different impulse waveforms, d =

10 mm, r = 10 µm, positive polarity. .................................................................................. 109

Figure 5-7 Time to breakdown in the mineral oil under different impulse waveforms, d = 10

mm, r = 10 µm, positive polarity. ....................................................................................... 109

Figure 5-8 Breakdown voltage and instantaneous breakdown voltages in the mineral oil

under different impulse waveforms; d = 10 mm, r = 10 µm, positive polarity. ................. 110

Figure 5-9 Flowchart of a mathematical model of breakdown voltage prediction. ............ 112

Figure 5-10 Simulated different impulse waveforms in the mineral oil based on the

mathematical model, positive polarity. ............................................................................... 113

Figure 5-11 Simulated different impulse waveforms in the synthetic ester liquid based on

the mathematical model, positive polarity. ......................................................................... 114

Figure 5-12 Typical breakdowns in the synthetic ester liquid under different impulse

waveforms, d = 10 mm, r = 10 µm, positive polarity. ........................................................ 114

Figure 5-13 Pre-breakdown in the mineral oil under different impulse waveforms, d = 10

mm, r = 10 µm, positive polarity, (a). 0.8/8 µs, V = 52 kV, lstoping = 7.62 mm (b). 0.8/14 µs,

12

List of Figures

V = 42 kV, lstoping = 7.71 mm (c). 0.8/30 µs, V = 34 kV, lstoping = 7.44 mm (d). 0.8/3200 µs,

V = 32 kV, lstoping = 7.85 mm............................................................................................... 115

Figure 5-14 Pre-breakdown in the synthetic ester liquid under different impulse waveforms,

d = 10 mm, r = 10 µm, positive polarity, (a). 0.8/8 µs, V = 40 kV, lstoping = 6.59 mm (b).

0.8/14 µs, V = 32 kV, lstoping = 6.42 mm (c). 0.8/30 µs, V = 30 kV, lstoping = 6.47 mm (d).

0.8/3200 µs, V = 26 kV, lstoping = 6.57 mm. ......................................................................... 115

Figure 5-15 Positive streamer area as a function of stopping length in the mineral oil under

different impulse waveforms, d = 10 mm, r = 10 µm. ........................................................ 116

Figure 5-16 Positive streamer area as a function of stopping length in the synthetic ester

liquid under different impulse waveforms, d = 10 mm, r = 10 µm. ................................... 117

Figure 6-1 Hydrogen concentration as gas in oil concentration from the TM1 hydrogen

monitor as a function of time during a 500 spark test, mineral oil, d = 10 mm, positive

polarity. ............................................................................................................................... 120

Figure 6-2 Hydrogen concentration as gas in oil concentration from the TM8 multi-gas

monitor as a function of time during a 500 spark test, mineral oil, d = 10 mm, positive

polarity. ............................................................................................................................... 120

Figure 6-3 The K factor based on Ostwald solubility coefficient under different temperature

in mineral oil ....................................................................................................................... 122

Figure 6-4 The voltage and current waveform in mineral oil on 99.9% breakdown voltage,

positive polarity, d = 10 mm, r = 50 µm. ............................................................................ 123

Figure 6-5 Typical spark in the mineral oil of positive polarity under lightning impulse, VB-

99.9%-positive = 31 kV, d = 5 mm, exposure time 0.5 µs. ......................................................... 125

Figure 6-6 Fault gas generation (GIT) in the mineral oil at different numbers of breakdowns,

positive polarity, d = 5 mm. ................................................................................................ 126

Figure 6-7 Fault gas generation (GIT) in the synthetic ester liquid at different numbers of

breakdowns, positive polarity, d = 5 mm. ........................................................................... 126

Figure 6-8 Comparison of hydrogen and acetylene generation (GIT) in the mineral oil and

the synthetic ester liquid under sparking fault at different numbers of sparks, positive

polarity, d = 5 mm. .............................................................................................................. 127

Figure 6-9 Individual fault gases as percentages of total fault gases in the mineral oil and

the synthetic ester liquid with a different number of sparks, d = 10 mm. .......................... 128

Figure 6-10 Fault gas generation (GIT) in the mineral oil as a function of the number of

breakdowns, Gemini X - Vb-99.9% = 39 kV, positive polarity, d = 10 mm. .......................... 129

13

List of Figures

Figure 6-11 Fault gas generation (GIT) in the synthetic ester liquid as a function of the

number of breakdowns, MIDEL 7131 - Vb-99.9% = 37 kV, positive polarity, d = 10 mm. .. 129

Figure 6-12 Comparison of hydrogen and acetylene generation (GIT) in the mineral oil and

the synthetic ester liquid at different gap distances under positive polarity. ...................... 130


the synthetic ester liquid with different spark numbers, Gemini X - Vb-99.9% = 39 kV,

MIDEL 7131 - Vb-99.9% = 37 kV, positive polarity, d = 10 mm. ......................................... 131

Figure 6-14 Fault gas generation in the mineral oil at different voltage levels after 200

sparks, Vb-99.9% = 39 kV, 1.5Vb-99.9% = 59 kV, d = 10 mm. ................................................. 132

Figure 6-15 Fault gas generation in the synthetic ester liquid at different voltage levels after

200 sparks, Vb-99.9% = 37 kV, 1.5Vb-99.9% = 56 kV, d = 10 mm. .......................................... 132


the synthetic ester liquid at different voltage levels with 200 sparks, d = 10 mm. ............. 133

Figure 6-17 Statistical analysis of fault energy for each spark in the mineral oil and the

synthetic ester liquid at the 10 mm gap distance, totally 1170 sparks. ............................... 134

Figure 6-18 Average energy per spark in the mineral oil and the synthetic ester liquid under

different test conditions. ...................................................................................................... 134

Figure 6-19 Fault gas volumes per unit fault energy (μL/J) of the mineral oil and the

synthetic ester liquid at the 5 and 10 mm gap distance with a different number of sparks.135

14

List of Figures

15

List of Tables

LIST OF TABLES

Table 2-1 Rate of rise corresponds to specific rising-voltage method [30]. ......................... 34

Table 2-2 Rate of rise corresponds to specific step-by-step method [30] ............................. 34

Table 2-3 Breakdown voltage and acceleration voltage obtained under step impulse and

lightning impulse [22]. .......................................................................................................... 52

Table 2-4 Bond Dissociation Energy [21]. .......................................................................... 57

Table 2-5 Indicator fault gases in mineral oil [76] ............................................................... 58

Table 2-6 Summary of experimental works of DGA analysis under electrical faults [90-99].

............................................................................................................................................... 60

Table 2-7 Fault gases volumes (µL) in Gemini X and FR3 for each of three groups of 15

breakdowns [92]. ................................................................................................................... 63

Table 3-1 Basic properties of testing liquids: MIDEL 7131 and Gemini X [42]. ................ 67

Table 3-2 The experimental conditions used to produce various tip radii of tungsten needles.

............................................................................................................................................... 71

Table 3-3 The parameters of the front resistor, tail resistor and charging capacitor used to

generate the four impulse waveforms. .................................................................................. 76

Table 3-4 The 99.9% breakdown voltages of the mineral oil and the synthetic ester liquid at

different gap distance under positive and negative lightning impulse. ................................. 82

Table 4-1 Weibull parameters of streamer initiation results with various tip radii at point-

plane electrode under negative and positive polarities ......................................................... 89

Table 4-2 Weibull parameters of breakdown results in the mineral oil and the synthetic

ester liquid with different gap distances at the point-plane electrode. ................................ 102

Table 5-1 Weibull parameters of breakdown results in the mineral oil obtained under

different impulse waveforms; d = 10 mm, r = 10 µm, positive polarity. ........................... 108

Table 5-2 Predicted breakdown voltage and experimental breakdown voltage in the mineral

oil obtained under different impulse waveforms; d = 10 mm, r = 10 µm, positive polarity.

............................................................................................................................................. 112

Table 5-3 Predicted breakdown voltage and experimental breakdown voltage in synthetic

ester liquid obtained under variable impulse voltages; d = 10 mm, r = 10 µm, positive

polarity ................................................................................................................................ 113

Table 5-4 Average charges injected into oil samples of the similar stopping length under

different impulse waveforms. ............................................................................................. 117

16

List of Tables

Table 6-1 Individual part energy as percentages of total energy generation based on 100

sparks in mineral oil, V = Vb-99.9%........................................................................................ 124

Table 6-2 Comparison of hydrogen measurements among the hydrogen monitor, the multi-

gas monitor and laboratory technique. ................................................................................ 125

Table 6-3 Fault gas volumes per unit fault energy (μL/J) of the mineral oil and the synthetic

ester liquid at different spark numbers, gap distances and voltage levels. ......................... 136

17

Abstract

ABSTRACT

Streamer characteristics, breakdown strengths and gassing behaviour of insulating liquids

under electric stresses are taken into account for a reliable design and safe operation of the

transformer. Ester liquids which are biodegradable and have high fire point have been

widely used in distribution transformers and some power transformers in recent years. It is

also interesting to introduce ester liquids into High Voltage Direct Current (HVDC)

converter transformers due to the fast development of HVDC transmission lines. Therefore,

this thesis aims to investigate the pre-breakdown, breakdown characteristics and gassing

behaviour of a synthetic ester liquid under DC and various impulse voltages where a

mineral oil is tested as the benchmark.

A comprehensive study of streamer characteristics and breakdown strength of the mineral

oil and the synthetic ester liquid under both positive and negative DC voltages was carried

out in the point-plane electric fields. Characteristics of streamer length, propagation

velocity and shape were analysed based on shadowgraph images obtained at a gap distance

of 10 mm, using a multi-channel ultra-high speed camera. Streamer inception voltages with

the tip radii of 5 µm, 10 µm, 20 µm and 50 µm and breakdown voltages at various gaps of 2

mm, 5 mm, 10 mm, 20 mm and 30 mm were also investigated. The results indicate that

there is no obvious streamer propagation (less than about 10% of the gap distance) under

negative polarity even when the applied voltage approaches breakdown voltage. At the

same applied voltage level, the streamer in the synthetic ester liquid propagates faster and

further than that in the mineral oil. As a result, the breakdown voltages of the synthetic

ester liquid are lower than those of the mineral oil at all the gap distances investigated

under both polarities.

Experimental and modelling studies of pre-breakdown and breakdown phenomena in the

mineral oil and the synthetic ester liquid under impulse waveforms with different tail-time

were carried out in the point-plane electric fields. A compact solid-state switch based

impulse generator was used to provide different impulse waveforms from short tail-time to

“step-like” tail-time: 0.8/8 μs, 0.8/14 μs, 0.8/30 μs and 0.8/3200 μs. A point-plane electrode

configuration with a small gap distance of 10 mm and a tip radius of 10 µm was used. The

results indicate that the shorter tail-time impulse waveform results in a shorter stopping

length and higher breakdown voltage; however it does not affect the instantaneous

breakdown voltage and time to breakdown. A mathematical model is therefore described to

predict the breakdown voltage under different impulse waveforms. In addition, with the

similar stopping length, higher energy injected from the short tail-time impulse caused the

streamers to have more branches than those under the long tail-time impulse.

The characteristics of fault gas generation in the mineral oil and the synthetic ester liquid

under various levels of electrical faults were studied. A test platform with functions of

automatic spark fault control and data acquisition was developed. The effects of spark

numbers (from 20 to 500), gap distance (5 mm and 10 mm) and voltage levels (Vb-99.9% and

1.5Vb-99.9%) on fault gas generation in liquids were studied. The key gases in the mineral oil

are H2 and C2H2, while the key gases in the synthetic ester liquid are H2, C2H2 and CO. The

amount of fault gas generation increases linearly with the number of sparks. However, the

number of sparks does not have an obvious effect on fault gas pattern and gas generation

per unit fault energy in µL/J. Spark at a larger gap distance or under a higher applied

breakdown voltage generates more fault gases due to higher injected fault energy.

18

19

Declaration

DECLARATION

I declare that no portion of the work referred to in the thesis has been submitted in support

of an application for another degree or qualification of this or any other university or other

institute of learning.

20

21

Copyright Statement

COPYRIGHT STATEMENT

(i). The author of this thesis (including any appendices and/or schedules to this thesis)

owns certain copyright or related rights in it (the “Copyright”) and s/he has given The

University of Manchester certain rights to use such Copyright, including for administrative

purposes.

(ii). Copies of this thesis, either in full or in extracts and whether in hard or electronic

copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988

(as amended) and regulations issued under it or, where appropriate, in accordance with

licensing agreements which the University has from time to time. This page must form part

of any such copies made.

(iii). The ownership of certain Copyright, patents, designs, trade marks and other

intellectual property (the “Intellectual Property”) and any reproductions of copyright works

in the thesis, for example graphs and tables (“Reproductions”), which may be described in

this thesis, may not be owned by the author and may be owned by third parties. Such

Intellectual Property and Reproductions cannot and must not be made available for use

without the prior written permission of the owner(s) of the relevant Intellectual Property

and/or Reproductions.

(iv). Further information on the conditions under which disclosure, publication and

commercialisation of this thesis, the Copyright and any Intellectual Property and/or

Reproductions described in it may take place is available in the University IP Policy (see

http://www.campus.manchester.ac.uk/medialibrary/policies/intellectual-

property.pdf), in any relevant Thesis restriction declarations deposited in the University

Library, The University Library’s regulations (see

http://www.manchester.ac.uk/library/aboutus/regulations) and in The University’s policy

on presentation of Theses.



http://www.manchester.ac.uk/library/aboutus/regulations)

22

23

Acknowledgement

ACKNOWLEDGEMENT

I would like to express my very great appreciation to my supervisor Dr. Qiang Liu for his

great supervision, guidance and support during my PhD studies in the University of

Manchester. He helped me build up concrete confidence bit by bit afterwards; any tiny

steps of my growth in the past three and half years are owing to his kind encouragement

and inspirations.

I would also like to thank my co-supervisor, Prof. Zhongdong Wang, since she has given

me valuable suggestions and comments on my research work and keeps me in the right

direction of research.

Great thanks are also given to the Qualitrol (Serveron) for the financial and technical

contributions to the project. In particular, he is Dr. John Hinshaw of Serveron Company.

Also, I would also like to thank Dr. Pascal Mavrommatis from TJ|H2b Analytical Services

for the lab measurements and technical contributions to the project. Mr. Adrian Walker, the

manager of EPSRC (Engineering and Physical Sciences Research Council) Engineering

Instrument Pool, His technical support and guidance for properly operating the Specialized

Imaging SIM 16 Camera guaranteed the success of this study and are greatly appreciated

here. In addition, many thanks to Prof. Yunpeng Liu from North China Electric Power

University (NCEPU), for the help of the impulse system design.

I wish to extend my thanks to all my colleagues in the transformer research group and all

my friends in the School of Electrical and Electronic Engineering. Thank you for offering

me an enjoyable working condition. Special thanks to Dr. Wu Lu, Dr. Zhao Liu and Mr.

Xiongfei Wang for the great help of technical supports during my experiments. In addition,

many thanks to Jiawen Xi, Xinyu Zhou, Zhepeng Lv, Harisanka K, Buyang Qi and Jiaxin

Li who I supervised during my PhD study, for the contribution of experimental studies.

Lastly but not least, I wish to give my heartfelt thanks to my family, to my parents for their

continuous support and encouragement.

24

25

List of Figures

27

Chapter 1 Introduction

CHAPTER 1. INTRODUCTION

1.1 Background

With the rapid development of society and technology, power systems are required to have

high stability and reliability, since they ensure our basic standard of living and daily

operation of industries. Transformers were introduced into the high voltage AC network in

the late nineteenth century [1]. Nowadays, a reliable power supply largely depends on the

fault-free operation of the networks, where transformers play an important role. The great

majority of power transformers in the world are still oil-immersed transformers, due to their

long lifetime expectancy.

In practice, the safe operation of high-voltage power transformers largely depends on the

insulation system design. The reason is that transformers in service are often exposed to

various types of overvoltage. One is the switching impulse voltage, which occurs when

apparatus is connected or disconnected from the power system, and is represented by

standard switching impulse voltage (250/2500μs) [2]. Another is due to lightning strikes to

overhead lines that are connected to the power transformer, represented by the standard

lightning impulse (1.2/50μs) [2]. Switching impulses and lightning impulses have been

recognised as one of the main causes of apparatus failure [1]. To verify the working

performance of insulation design, both switching and lightning impulse tests are required

for a transformer with a voltage rating higher than 170 kV [3].

Insulating liquids have been widely employed in high-voltage transformers for many

decades. They have several functions including electrical insulation, cooling medium and

information carriers. There are also some other expected properties for insulating liquids,

such as high fire point, environmentally friendly and cellulose solid insulation compatibility

etc. Nowadays, the dominated insulating liquid for power transformers is still mineral oil,

due to its successful and widespread usage over many decades. In recent years, the

preference for renewable, sustainable energy sources on an international level has meant

that ester liquids are now being considered as alternatives to mineral oils, due to their

improved fire safety and outstanding environment performance [4-6]. Ester liquids are

classified into synthetic esters synthesised from a combination of chemicals and natural

esters derived from sustainable and biodegradable resources [7]. For the last few decades,

28


natural esters have been successfully used in more than 45,000 low and medium power

installations [8]. Moreover, there is an increasing preference for ester liquids in large power

transformers. The synthetic ester liquid used in a 238 kV installation for Vattenfall in

Sweden has successfully been in operation since 2004, a 400 kV installation for National

Grid in the UK has been delivery in summer 2015 [9], and a 420 kV natural ester

transformer (operated at 380 kV) planted in Germany [10]. However, differences between

ester liquids and mineral oils in chemical and physical characteristics require further

investigation before they can be widely used in other types of transforms, e.g. HVDC

converter transformers.

Breakdown strength is one of the essential properties of insulating liquids. To understand

the breakdown mechanism in insulating liquids, the pre-breakdown phenomenon is the

critical aspect to be analysed [11]. Pre-breakdown in insulating liquids is a cause of further

deterioration of its insulation ability, which may eventually lead to failure of the electrical

equipment [1]. The discharge phenomenon in liquids before breakdown is usually called a

“streamer” [12]. Streamers are generally low-density conductive channels initiated in high

local electrical fields [13]. Generally, streamer characteristic analysis under impulse

voltages is one of the most effective ways to identify the intrinsic dielectric property of

insulating liquids [11]. However, due to a lack of coherent theory to describe streamers in

liquid, further research into the study of streamer characteristics, including length, velocity,

shape, area, current and light signals of the streamer, is required.

Streamer characteristics under standard lightning impulse and step-like impulse voltage

have been carried out for many decades, but there is a lack of understanding of the effects

of different impulse waveforms. In practice, an in-service transformer might be struck by

various impulse voltages, e.g. repetitive impulse voltage and voltage waveform with

different front and tail times. In recent years, more and more researchers have been

focusing on streamer and breakdown studies under non-standard impulse voltage [14-16],

e.g. chopped impulse. Therefore, the analysis of streamer characteristics and breakdown

strength under non-standard impulse voltage is worth further investigation.

In recent years, there has been an increasing interest in applying High Voltage Direct

Current (HVDC) transmission lines in some countries, e.g. European countries, Brazil and

China, due to their long distance or submarine bulk energy transmission [17]. In HVDC

transmission systems, converter transformers are the essential components. The insulation

systems of HVDC converter transformers could suffer both AC voltage and DC biased AC

29


voltage [18]. The streamer phenomena under AC and impulse voltage have been reported

for many years. In contrast to impulse voltage which has an extremely short duration within

microseconds, DC voltage has considerable time duration to work on insulating liquids.

However, there is a lack of streamer studies regarding liquids under DC voltage. To

understand the electrical performance of insulating liquids in HVDC converter transformers,

it is worth investigating the streamer characteristics and breakdown strength under DC

voltage.

Besides electrical breakdown, gassing behaviour due to discharge and breakdown has to be

considered for a reliable insulation design and safe operation of liquid insulated equipment

[19]. The gassing behaviour of mineral oil and ester liquids has been studied under

electrical fault of AC voltage for many years. In addition to the electrical fault of AC

voltage, transformers during operation could suffer various electrical faults, e.g. impulse

voltage. [20]. Moreover, past study [21] indicated that gas generation is highly associated

with fault energy. It is thus interesting to investigate the gassing behaviour of mineral oil

and ester liquids subjected to electrical faults under repetitive impulse with accurate energy

control.

1.2 Research Objectives

The aim of this PhD thesis is to evaluate the streamer, breakdown and gassing

characteristics of insulating liquids under various voltage stresses. Two types of insulating

liquids including a mineral oil and a synthetic ester liquid will be studied. Streamer

phenomena and breakdown strength under DC voltage were analysed under the various tip

radius and gap distances in a non-uniform electric field. Streamer characteristics and

breakdown properties of the two liquids under different impulse waveforms (from the

short-tail impulse to “step” like impulse voltage) will be investigated in a non-uniform

electric field. Moreover, to understand the gassing behaviour in the liquids after breakdown,

the correlation between gas generation and energy injection under electrical faults with

controlled fault energy were carried out. The following topics will be covered in detail in

this thesis:

(i) Streamer phenomenon and breakdown strength of insulating liquids under DC

voltages

Due to the application of converter transformers in HVDC transmission systems, the

investigation of dielectric properties of mineral oil and ester liquids in HVDC converter

30


transformers is required. The first attempt is to investigate the streamer and breakdown

phenomena under DC voltage in a non-uniform field. The effects of both tip radius and gap

distance on streamer initiation and breakdown strength are considered in this study.

(ii) Streamer characteristics and breakdown properties of insulating liquids under

different impulse waveforms

Lightning impulse is widely used in the power industry as one of the standard waveforms

for testing insulation performance. Step impulse has the advantage of fast rise-time that

mitigates the effect of space charge injection on streamer initiation and of long tail-time

that provides a stable voltage stress for streamer propagation. Previous studies [22, 23]

indicated that breakdown voltage highly correlates to instantaneous breakdown voltage and

time to breakdown under lightning and step impulse voltage. More valuable results, such as

breakdown voltage, can be obtained based on mathematical programs without practical

experiments. Therefore, to elaborate on the previous findings, the objective is to study

streamer and breakdown phenomena of a mineral oil and a synthetic ester liquid under

variable impulse voltages. A needle-to-plane electrode configuration and four different

impulse waveforms under positive polarity are considered in this study.

(iii) Gassing behaviour of breakdown in liquids

In recent decades, the gassing behaviour of mineral oil and ester liquids has been

investigated under electrical faults of AC and impulse voltage including discharge, sparking

and arcing. However, most previous studies lacked auto-controlled electrical fault

generation especially under AC voltage, which resulted in the inaccurate measurement of

fault energy injection. Therefore, the objective is to build up a testing system with an auto-

controlled spark generator and investigate the correlation between gassing behaviour and

electrical faults in both the mineral oil and the synthetic ester liquid.

1.3 Major Contributions

The major contributions of this thesis are given as follows:

(i) The effects of tip radius on streamer initiation voltage and the effects of gap

distance on breakdown voltage were studied under DC voltage. Although

initiation voltages of the synthetic ester liquid are comparable to those of the mineral oil,

the breakdown voltages of the synthetic ester liquid are lower than those of the

mineral oil at all the gap distances investigated under both polarities.

31


(ii) (ii) Pre-breakdown and breakdown phenomena in the mineral oil and the

synthetic ester liquid under impulse waveforms with different tail-time were

carried out in the point-plane electric fields. A mathematical model is described

to predict the breakdown voltage in the mineral oil and the synthetic ester liquid

under impulse waveforms with different tail times.

(iii) A relationship between fault gas generation and fault energy was investigated in

both the mineral oil and the synthetic ester liquid under lightning impulse. The

amount of fault gas generation increases linearly with the number of sparks.

However, the number of sparks does not have an obvious effect on fault gas

pattern and gas generation per unit fault energy in µL/J.

1.4 Outline of Thesis

The following is a summary of the chapters presented in this thesis:


This chapter introduces the background and motivation of the PhD study and also gives an

overview of the thesis.

Chapter 2 Literature Review

This chapter first provides background knowledge of discharge phenomena and

measurement techniques in liquids under DC. Then, the streamer modes and the effects of

impulse waveforms on streamers and the breakdown characteristics of insulating liquids are

discussed. Finally, recent research work related to fault gas generation in oil under

electrical faults is summarised in this chapter.

Chapter 3 Experimental Description

This chapter first discusses the liquids (a mineral oil and an ester liquid) under investigation

and the preparation procedures for oil samples. Then, the methodology of etched tungsten

needles based on electrochemical technique is explained. Finally, the experimental setups

for streamer and breakdown tests under DC and variable impulse voltage, and the setup of

fault gas generation under electrical faults are described in detail.

32


Chapter 4 Streamer and Breakdown Phenomena of Transformer Liquids under DC

Voltage

This chapter describes the streamer characteristics and breakdown strength of the mineral

oil and the ester liquid in a strongly non-uniform point-plane field under DC voltage. The

effects of tip radius on streamer initiation voltage and the effects of gap distance on

breakdown voltage are studied. At a fixed gap distance, streamer characteristics including

length, average propagation velocity and shape are analysed under both positive and

negative polarities.

Chapter 5 Streamer and Breakdown Phenomena of Transformer Liquids under

Various Impulse Voltages

This chapter reports the streamer phenomena and breakdown properties of the mineral oil

and the ester liquid under variable impulse voltages. Streamer length, average propagation

velocity, area and shape at a small gap distance are investigated. The relationship between

breakdown voltage, instantaneous breakdown voltage and time to breakdown is analysed,

and a feasible mathematical model for breakdown voltage estimation is proposed and

verified.

Chapter 6 Correlation between Gas Generation and Sparking Faults of Transformer

Liquids under Impulse Voltage

This chapter investigates fault gas generation in the mineral oil and the ester liquid under

sparking faults. A multi-gas on-line dissolved gas monitor was used to detect eight fault

gases. The effects of breakdown numbers, voltage polarities and voltage levels on gas

generations were studied. The relationship between energy generation and gas generation

was found.

Chapter 7 Conclusions and Future Work

This chapter summarises the main conclusions of this thesis and provides the suggestions

for future work no pre-breakdown and breakdown studies of transformer liquids.

33


CHAPTER 2. LITERATURE REVIEW

2.1 Introduction

This thesis focuses on streamer, breakdown and gassing behaviour of breakdowns in liquids

under variable electrical stresses. This chapter first discusses the fundamental knowledge

and recent findings of streamer and breakdown in liquids under DC voltage and then

summarises the most recent research into the streamer and breakdown in liquids under

impulse voltage. Finally, previous results of gassing behaviour of liquids under electrical

faults are summarised and discussed.

2.2 Streamer and Breakdown in Liquids under DC Voltage

Various measurement technologies including current measurement [24-27], acoustic

emission technique [18], ultra-high frequency technique [28] and image converter camera

technique [24, 29] were introduced to investigate streamer and breakdown phenomena

under DC voltage. A few studies on partial discharge (PD) of hexane under DC voltage

were published [24, 29], describing the correlation between current signal and streamer

shape.

2.2.1 Methodologies

It is important to define the method for applying DC stress, as it has an impact on the

results [30] e.g. streamer initiation, breakdown voltage etc. In the literature, there are three

main methods used: the rising-voltage method [18, 31-33], periodic step-by-step method

[34], which have been used for streamer and breakdown study under DC voltage over the

past decades. Each method has its own pros and cons, and also its own validity range.

When comparing the breakdown strength of oil samples, different testing methods might

lead to different results. Therefore, an applicable testing method is significant for

investigating breakdown strength under DC voltage.

Rising-voltage method

The rising-voltage method is normally used for breakdown tests under AC and DC voltage.

For a single test, the applied voltage should be initiated at 0 kV and increased at a constant

rate till breakdown occurs. This single test procedure is then repeated after a fixed time

interval, until a specific number of breakdowns is obtained [35]. In IEC 60243 [30], the

34


rising-voltage method is classified into a short-time test (breakdown in 10 to 20 seconds), a

slow rate-of-rise test (breakdown in 120 to 240 seconds) and a very slow rate-of-rise test

(breakdown in 300 to 600 seconds). The rate of rise shall be initially selected as defined in

Table 2-1 [30].

Table 2-1 Rate of rise corresponds to specific rising-voltage method [30].

Rising-voltage methods Rate of rise

short-time test 100 V/s, 200 V/s, 500 V/s, 1 000 V/s, 2 000 V/s,

5 000 V/s, etc.

slow rate-of-rise test 2 V/s, 5 V/s, 10 V/s, 20 V/s, 50 V/s, 100 V/s,

200 V/s, 500 V/s, 1 000 V/s, etc.

very slow rate-of-rise test 1 V/s, 2 V/s, 5 V/s, 10 V/s, 20 V/s, 50 V/s,

100 V/s, 200 V/s, etc.

Step-by-step method

The step-by-step method is normally used to observe streamer initiation, propagation or

breakdown test under impulse voltage due to the application of discontinuous voltage.

Applied voltage starts at a level of expected percentage (normally 40% - 70%) of initiation

voltage or breakdown voltage, and increases step by step at a constant step interval (e.g. 1

kV per step) till initiation or breakdown occurs. Between each step, a fixed amount of time

should be allowed to prevent the accumulative effect of space charge, particles, bubbles etc.

In IEC 60243 [30], a voltage at 40 % of the probable short-time breakdown voltage shall be

applied to the oil sample based on a 20-second step-by-step test. If the oil sample

withstands this voltage for 20 seconds without breakdown, the voltage should be increased

in incremental steps as defined in Table 2-2. Each increased voltage should be immediately

and successively applied for 20 seconds until breakdown occurs. Unless otherwise specified,

the test should be carried out with a time intervals of 60 seconds.

Table 2-2 Rate of rise corresponds to specific step-by-step method [30]

Starting voltage (kV) Increment (kV)

1.0 or less 10 % of start voltage

Over 1.0 to 2.0 0.1

Over 2.0 to 5.0 0.2

Over 5.0 to 10.0 0.5

Over 10 to 20 1.0

Over 20 to 50 2.0

Over 50 to 100 5.0

Over 100 to 200 10.0

Over 200 20.0

35


The step-by-step method under DC voltage was used in [34]. The experiment was carried

out using the periodic stress-grounding method, applying a DC voltage for 3 minutes and

then grounding for 2 seconds, and the same procedure was repeated 20 times. In addition,

different test methods were compared under impulse conditions, which indicates that 50%

breakdown voltage under 3 shots per step is lower than that under 1 shot per step [36].

2.2.2 Current and emitted light

The experiment of current measurement in a mineral oil under negative DC voltage was

carried out in [37]. The gap distance was between 3 and 25 mm with a needle-to-plane

geometry. A 50 Ω current shunt was placed at low voltage (LV) side of the test cell and was

connected to an oscilloscope through a 50 Ω coaxial cable. A coupling capacitor with 460

nF used for PD pulses detection was set in parallel with the test cell.

The negative streamer current in a mineral oil under DC voltage at a gap distance of 5 mm

in a needle-to-plane electrode is shown in Figure 2-1 [37]. When the applied voltage is

below 50% breakdown voltage, the oscillating discharge pulses might occur discreetly as

single pulses, in pairs, three pairs or more successive pulses with a stably increasing

magnitude [37]. For each single pulse, the rise time of about 1 ns indicates the PD

phenomena in mineral oil is as fast as that in air gap [37]. Based on the current

measurement, it is possible to understand the time domain from streamer initiation to

propagation.

(a) 100 ns/div (b) 5 ns/div

Figure 2-1 Typical discharge pulse in a needle to plane gap in a mineral oil under negative polarity at room

temperature, d = 5 mm, V = 14 kV [37].

The photomultiplier tube (PMT) is normally used to study streamers in optics, which

records the real-time light emission during streamer propagation. Previous studies of

streamers related to emitted light in oils were mostly under impulse voltage [38, 39],

36


showing that emitted light signals correspond to current signals. A study of emitted light of

streamers in transformer oil under DC voltage was carried out in a needle-to-plane

electrode of 20 mm gap distance as shown in Figure 2-2 [27]. A Hamamatsu made PMT

H5783P with a spectral response from 300 nm to 650 nm was used to monitor the emitted

light of a streamer. Compared to the current signal, the unipolar light signal has less noise

and signal oscillation, which enables researchers to have a deeper understanding of

streamer studies.

Figure 2-2 Streamer current and emitted light signal in transformer oil under DC voltage, positive polarity, d

= 20 mm, V = 40 kV [27].

In addition, the results from [24, 29] indicated that the streamer current waveform is

significantly different between positive and negative polarities under DC voltage, which

confirms the previous finding under impulse voltage [40, 41]. Figure 2-3 shows the typical

current waveforms of positive (Anode point) and negative (Cathode point) polarities under

DC voltage [29]. The streamer current under negative polarity is characterised by a period

of growth where the pulse amplitudes increase systematically followed by a longer period

of large pulse. Under positive polarity, both continue and pulsed current waveforms were

observed.

37


Figure 2-3 Streamer current waveforms in Hexane taken with a point cathode (Top – Negative polarity) and

with a point anode (Bottom – Positive polarity), point-sphere electrode, d = 4.75 mm [29].

2.2.3 Photography measurement

Besides the methods of electrical and optical measurements, photographing setup remains

the most widely used technique to study streamers. Recent studies of streamer images are

mainly obtained under impulse voltage [12, 38, 42, 43]. Research into streamer shapes was

normally limited to impulse studies due to difficulty in correlating the shuttering of the

camera with the initiation stage of streamer under DC voltage [29]. Under continuous

stressing of DC voltage, the randomness and uncertainty of a streamer initiation make

streamer image capture more difficult.

Figure 2-4 shows the experimental setup of streamer photograph investigation in Hexane

under DC voltage [24, 29]. A needle-to-plane electrode made of stainless steel at a gap

distance of 3.2 mm was set as electrode geometry. A 10 MΩ current limit resistor was

38


connected following high-voltage DC power supply to prevent a high current when the

breakdown occurs. A laser used for illumination when the streamer occurs was placed at the

left of the test cell. An image converter camera was placed at the right side of the test cell to

capture the streamer image. Lenses were placed between the test cell and camera to provide

high magnification. An image preserving optical delay was developed to preserve the image

long enough to shutter the camera. A Pockels cell shutter prevents the continuous

illumination generated by the laser from overexposure in image converter camera. At LV

side of the test cell, a current sensitive amplifier was directly connected to the needle

electrode to measure the current signal. The Pockels cell shutter and the camera are both

triggered simultaneously by a streamer current signal via the synchronising circuit and

current amplifier. The shutter opens after a streamer initiates, and after the camera is used

to photograph the growth of the discharge, a second pulse closes the shutter [29].

Figure 2-4 Experimental setup of time delay system for streamer photograph capture under DC voltage (re-

produced) [24].

The synchronised imaging technique, allows the features of streamer shape under DC

voltage to be investigated. Previous findings of streamers under impulse conditions

indicated that the intensity and occurrence time of streamers in photograph normally

correlate to those of current and light signals [12, 44]. A similar streamer feature

phenomenon was found under DC voltage with the applied voltage level of 15.5 kV [24,

29]. Figure 2-5 shows the streamer shape, current and stopping length in Hexane at a

needle-to-plane geometry of 3.2 mm gap distance under DC voltage [29]. The frames were

captured 200 ns apart and the exposure time was 40 ns. It was found that the isolated

current pulses were generated when the streamer propagated and the occurrence of a small

current pulse between 3 and 4 µs appears to be correlated with some slow development of

CW Argon

Laser 10 X

Image – Preserving

Optical Delay Pockels Cell

Shutter

Image

Converter

Camera

Trigger

Amp. Synchronizer

Digital Storage

Oscilloscope

DS

PS

10 M

250 pF

Mon.

100 K

39


the streamer. Once these current pulses ended, the streamer stopped growing and dissipated

slowly. Although the streamer can be captured by the synchronised imaging technique,

there is still an obvious time delay of the photograph at the initiation stage of the streamer

due to the difficulty of triggering a small current at an early stage.

Figure 2-5 The photograph of a streamer with the numbers indicating the sequence in which the pictures were

taken; time interval between each film 200 ns; (a) streamer photograph by the camera; (b) streamer current

signal; (c) streamer stopping length [29].

Figure 2-6 shows a comparison of streamer shapes leading to breakdown in transformer oil

under different external pressures in a needle-to-plane geometry with a gap distance of 2.5

mm under negative DC voltage [45]. A voltage amplifier was used to trigger the pulse

generator to initiate intensified charge coupled device (ICCD) camera. All pulse generators

were set to a minimum internal delay of 80 ns. This indicates that the structures of the

streamer formations shown for 300 torr (about 40 kPa) are identical to streamer formations

observed at atmosphere (about 101 kPa), presented as a uniform, luminous channel after

breakdown and no expanded regions [45]. When the environment pressure is at 30 torr

(about 4 kPa) or below, the streamer shapes begin to show some expansion and once

breakdown occurs, luminosity is seen throughout the expanded region. This finding

assumes that the breakdown mechanism in liquids of negative polarity under DC voltage is

based on bubble formation with subsequent carrier amplification in the gas phase, which

indicates pressure dependence, i.e. the expansion velocity decreases with increasing

pressure. These results support the bubble theory streamer formation and breakdown under

negative polarity. However, a similar experiment was investigated under positive polarity

(a)

(a)

(b)

(c)

Time

delay

40


and indicated no pressure dependence [46]. This finding confirms the previous results under

lightning impulse tests, which indicated that positive streamers rely more on electronic

processes, while negative streamers depend more on gaseous processes [47].

Figure 2-6 Shadowgraph images of streamer expansion in transformer oil as a function of time and pressure

in a needle-to-plane geometry with a gap of 2.5 mm under negative polarity [45].

2.2.4 Breakdown properties in Liquids

Breakdown strength is one of the parameters used to assess the performance of insulating

liquids. The breakdown studies under DC voltage, which are enhanced by adding semi-

conductive nanofluids (SNFs) in transformer oil are presented in [33]. The high voltage DC

source is capable of generating up to 200 kV. The point-sphere geometry with a gap

distance of 10 mm and a tip radius of 3 μm was chosen in the tests. The rising-voltage

method is applied to stress the oil sample with a ramping rate of 1 kV/s. The initial standing

time was 5 minutes, and the time interval between two successive shots was fixed at 1

minute. All experiments were performed at room temperature. After one breakdown

occurred, the oil sample and electrodes were changed; 10 samples in total were carried out

for each type of oil. Similar to the breakdown tests under impulse voltage [40], DC

breakdown voltages of positive polarity are almost half of those under negative polarity.

41


Figure 2-7 shows the breakdown tests in oil with two different gap distances in a needle-

plate electrode system under DC voltage of both polarities [48]. Similar polarity effects to

the results shown in [40] were observed; negative DC breakdown voltages are much higher

than positive DC breakdown voltage. Moreover, the polarity effect under oil alone

condition becomes significant with an increase of gap distance, and this is similar to what

was observed in past studies [38].

Figure 2-7 Breakdown voltage of oil under DC voltages of both polarities [48].

2.2.5 Space Charge

A lot of publications related to DC voltage focused on investigating space charge, as

insulating paper and oil are widely used as major insulation materials in HVDC equipment,

such as converter transformers, DC bushings and DC cables [49-52]. Most of the space

charge studies focused on oil paper insulation system, since the formation and dynamics of

space charge under high electric fields will change the distribution of electric fields [49].

In liquid only, the theory of space charge induced field distortion can be also used to

explain the polarity effect on the breakdown in liquids [11]. Under positive polarity, after

ionization occurs in the region near the needle electrode, electrons move quickly towards

the positive needle electrode, while positive ions move away slowly to the negative plane

electrode. The accumulated positive ions act as an extension of the positive needle electrode

and thus enhance the boundary electric field at the head of the positive streamer. This

promotes streamer propagation and consequently decreases breakdown voltage. Under

negative polarity, after ionization occurs in the region near the needle electrode, positive

ions move back slowly to the negative needle electrode, while electrons dissipate quickly

into the liquids. Therefore the diluted negative charges act as a shielding of the negative

42


needle electrode and reduce the electric field at the head of the negative streamer, which

slows down streamer propagation and consequently increases negative breakdown voltage.

2.3 Streamer and Breakdown in Liquids under Impulse Voltage

The transformer in service could be exposed to voltages in excess of the normal operating

voltage, such as transient overvoltage due to lightning strikes to earth near overhead lines

connected to the transformer (represented by standard lightning impulse 1.2/50 μs). In order

to understand the breakdown mechanism in liquids, pre-breakdown phenomena, also

known as streamer phenomena, shall be thoroughly investigated. Previous streamer and

breakdown measurements of insulating liquids under lightning impulse were studied in [12,

38, 42, 43, 53-55], and those under step impulse were published in [56-60]. The correlation

between breakdown voltage, instantaneous breakdown voltage and time to breakdown

under lightning impulse and step impulse at large gap distance was discussed in [22, 23],

and it was shown that more comprehensive data can be obtained without complex

experiments. In the following sections, general knowledge of streamer and breakdown

phenomena under various impulse voltages is discussed.

2.3.1 General Streamer Characteristics

The breakdown in liquids generally consists of the initiation and propagation of a streamer.

If the streamer stops propagating before bridging the electrodes, the phenomenon is treated

as pre-breakdown or streamer. Pre-breakdown phenomena in liquids normally behave as

luminous filaments propagating at a velocity from about 100 m/s up to more than 100 km/s

[61]. From early publications [62-65] of discharges in liquids, the terms ‘corona’, ‘leader’

and ‘streamer’ were used to define the phenomenon. As time progressed, the term ‘streamer’

with defined modes was widely accepted to describe all types of discharge phenomena in

liquids.

So far there are two main theories used to describe streamer mechanisms: ‘electronic

ionisation model’ and ‘gaseous bubble theory’. The electronic ionisation model is borrowed

from early studies of electrical breakdown in gases, which indicated breakdown occurs on

account of electron avalanches caused by continuous collision ionisation [66]. In the

gaseous bubble theory, a streamer initiates from local micro bubbles in liquids, which is

possibly formed by the vaporisation of liquid due to the charge injection induced heating

process or the natural existence due to the impurity of the liquid [67]. The bubble theory is

supported by a few previous studies [68, 69] under various pressure tests, such as the

43


streamer image shown in Figure 2-6 [45], in which both streamer initiation and propagation

are affected by the applied pressure. With more and more findings on the streamer, these

two theories are both feasible and reasonable to describe breakdown phenomena, probably

in the different stages of streamer, named as propagation ‘modes’ [11].

Streamer modes

The idea of streamer ‘modes’ was first proposed by Hebner [70] in 1988, who put forward a

classification of modes based mainly on streamer propagation velocities. The development

and the usage of the high-speed camera by Lesaint and others have led to the creation of

experimental phenomena (e.g. streamer shapes) and, as a result, the concept of ‘modes’ can

then be decided by such experimental phenomena. [12, 61]. In an identical nature of oil

sample, when the test conditions are changed (e.g. higher applied voltage), a rapid

transition in the streamer propagation velocity is observed. Not only does the streamer

velocity dramatically increase, but the streamer shape, current and emitted light signal are

also influenced. This phenomenon indicated that different physical processes do exist in

these streamer modes. The images of Figure 2-8 show the four different positive streamer

modes (1st, 2

nd, 3

rd and 4

th modes) induced in a mineral oil with a wide range of applied

impulse voltage (from 6 to 424 kV) and a needle-to-plane geometry of gap distances from 6

mm to 10 cm [61]. It was found that the streamer propagates towards the plane electrode

with an average propagation velocity from 100 m/s at small gap distance (A), to over 100

km/s at large gap distance under much higher voltage (E).

Figure 2-8 Typical images of positive streamers in a mineral transformer oil. (A) ‘1st mode’; (B) ‘2nd mode’;

(C) ‘2nd

mode’; (D) ‘3rd + 2nd modes’; (E) ‘4th mode’ [61].

Streamer images (A) and (B) presented as shadowgraphs indicate the streamer is

constituted by the low-density gas phase. All streamer modes are able to emit light but with

different intensity. The light intensity of 1st mode streamer in liquids is weakest, where the

44


light intensity is almost close to detection capability of PMT. It is difficult to capture these

kinds of streamers even by a sensitive high-speed camera due to the weak illumination of

the streamer. According to different literature, this kind of streamer is known by different

names such as ‘primary’, ‘bush-like’, ‘subsonic’ or ‘slow’ streamer. The 1st mode streamer

is normally observed only at the low voltage level and preferably in the pre-breakdown

stage. The typical streamer current signal in liquids is composed of a continuous current at

the early stage and a train of rapid pulses with nanosecond duration afterwards as shown in

Figure 2-9 (A) [71]. The emitted light signal has an extremely low intensity, which

indicates the irregular nature of the recorded signal. Therefore, the observation of 1st mode

streamer can only be achieved by a shadowgraph or the Schlieren technique.

When the applied voltage is increased, the shape of the streamer is suddenly transformed

into to the 2nd

mode streamer with the appearance of fine filaments (µm in diameter) as

shown in Figure 2-8 (B) [61]. Since the 2nd

mode streamer is faster, streamer currents in

Figure 2-9 (B) are made up of continuous components with significant intensity [71].

Moreover, the intensity of emitted light by 2nd

mode streamer is much stronger than that of

the 1st mode streamer. Therefore, it then becomes possible to record the 2

nd mode streamers

more clearly owing to more luminous branches and offshoots [61].

Figure 2-9 Typical examples of positive streamer with emitted light (upper trace, arb. unit.), transient currents

(middle trace) and voltage (lower trace) in cyclohexane, d = 1.8 mm, rp = 1.2 μm, (A) 1st mode, V = 12 kV, (B)

2nd

mode V = 20 kV [71].

Streamer pictures (C), (D) and (E) were obtained with a gated image intensifier showing the

emitted light integrated during propagation [61]. Streamer photographs (C), (D) and (E)

45


mainly show the most luminous filaments, while weakly luminous or dark filaments cannot

be seen in these images. The light emission of fast streamers at high voltage in 4th

mode (E)

is considerably greater than in (C). Due to strong light intensity in the image (E), the

aperture of the high-speed camera must be relatively reduced to avoid overexposure. By

contrast, weakly luminous filaments surrounding the main luminous channel cannot be

observed, due to the low light intensity of side streamer branches.

Figure 2-10 shows the typical transient currents of 2nd

, 3rd

and 4th

mode streamers in a

natural ester liquid at a gap distance of 10 cm [57]. As shown in Figure 2-10 (A), the 2nd

mode streamer was observed with a uniform velocity about 2 km/s and a stopping length ls

of 3 cm without breakdown. The associated current at the early stage is in the form of a

continuous current and afterwards is transformed into a train of large single pulses with

increasing amplitude, and has similar characteristics to streamers observed in Figure 2-9 but

different liquid nature. In mineral oil, the presence of aromatic molecules helps to provide a

stronger light emission, while the emitted light in ester liquids is weaker compared to

mineral oil [44].

When the applied voltage is increased to a certain level, a very different 3rd

mode streamer

can be observed at the early stage of streamer propagation in Figure 2-10 (B). Several

intense current pulses up to 1 A correspond to the bright streamer channel in picture (B).

The picture (B) also indicates that the 3rd

mode streamer does not propagate in continuous

form. The corresponding average propagation velocity of the 3rd

mode streamer is larger

than 10 km/s. Once the initial 3rd

mode streamer has ended, the streamer then continues to

propagate with the 2nd

mode at relatively lower velocity. In addition, when a voltage level is

repeated, either 2nd

mode or 3rd

mode streamer usually appears. For example in Figure 2-10

(A) and (B), both streamers were obtained at the same voltage – 120 kV, and both 2nd

and

3rd

mode streamers randomly showed up [61]. For the pure 3rd

mode streamer in Figure

2-10 (C), the time duration of current pulses (about 200 ns) is much longer than those of the

2nd

mode streamer, and the instantaneous velocity per step reaches above 100 km/s [61].

At a much higher voltage, streamer propagation from the needle-to-plane electrode is

completed in one step at an extremely fast velocity of about 120 km/s, which is called the

4th

mode streamer. The streamer current in the 4th

mode is completely in the form of

continuous components as shown in Figure 2-10 (D).

46


Figure 2-10 Propagation modes at high voltage in a natural ester liquid. d = 10 cm. Upper oscilloscope trace:

applied voltage, lower: streamer current. (A) 2nd

mode, V = 120 kV; (B) 3rd

+ 2nd

modes, V = 120 kV; (C) 3rd

mode, V = 142 kV; (D) 4th

mode, V = 152 kV [57].

Investigations of streamers in the 3rd

and 4th

modes were also carried out in [72]. Figure

2-11 shows the propagation process of positive 3rd

+2nd

mode streamer in Diala S4 at high

voltage just below acceleration voltage Va, which indicates more structure information of

the 3rd

mode streamer [72]. At the voltage just below Va, in the first stage during streamer

propagation, the 3rd

mode positive streamers appear, with speeds varying between 6 km/s to

11 km/s. These 3rd

mode streamers consist of one or two main branches surrounded by

numerous side branches. However, the 3rd mode streamers turn into 2nd mode streamers

afterwards with a speed of around 3 km/s during propagation.

Figure 2-11 The propagation process of positive 3rd

+2nd

mode streamer at high voltage just below Va; Diala

S4 ZX-I, d = 50 mm; the ground electrode is at the bottom edge of the streamer images [72].

47


Figure 2-12 shows the propagation process of negative 4th

and 3rd

mode streamers in Diala

S4 at high voltage above Va. When applied voltage is above Va, the propagation of negative

streamers starts with 4th

mode over a very short time (0.5 μs). The 4th

mode streamer has a

bright filamentary shape and a velocity of approximately 20 km/s. However, this bright

channel quickly dies out and streamer propagation falls to the 3rd

mode with a much lower

velocity of 5-6 km/s. The shape of the 3rd

mode streamer is similar to that of the 2nd

mode

streamer at lower voltage but has more branches. The 3rd

mode streamer then propagates for

about 3 μs before the streamer switches back to the 4th

mode again at the end of streamer

propagation, with a velocity of over 10 km/s.

Figure 2-12 The typical propagation process of 4th

+3rd

mode negative streamers in the GTL oil, d=50 mm; the

ground electrode is at the bottom edge of the streamer images from frame 3 to frame 6 [72].

Figure 2-13 and Figure 2-14 show the typical shapes of the positive and negative 4th

mode

streamers in neat white oil, neat white oil with reduced pressure (RP) and neat white oil

with dimethylaniline (DMA) under lightning impulse [47]. When applied voltage is above

Va, the 4th

mode streamers are observed. Under positive polarity, the streamers consist of

only a main bright channel with some lateral branches. Compared to the 2nd

and 3rd

mode

streamers, the 4th mode streamers have a much lower number of branches. This mode

streamer is as fast as 100 ~ 200 km/s [47]. Under negative polarity, streamers are composed

of some main channels surrounded by numerous lateral branches. The streamers are as fast

as 10 ~ 40 km/s, which are slower than those under positive polarity [47].

48


Figure 2-13 Typical shapes of the positive 4th

mode streamers: (a) neat white oil, (b) neat white oil at reduced

pressure, and (c) DMA, all images taken just before breakdown, d = 80 mm [47].

Figure 2-14 Typical shapes of the negative 4th

mode streamers: (a) neat white oil at reduced pressure, (b) neat

white oil, and (c) TCE, all images taken just before breakdown [47].

As well as investigating of streamer modes, paper [47] also indicated that both electronic

and gaseous processes exist during streamer propagation of both low and fast mode

streamers with a long gap. Fast positive streamers rely more on electronic processes, while

fast negative streamers depend on both electronic and gaseous processes [47].

Overall, it is feasible to observe transitions among streamer shapes corresponding to the

different streamer modes. The 1st mode streamer shows a ‘bubble-like’ shape, and the 2

nd

mode streamer shows a ‘tree-like’ shape composed of branched fine filaments. The 3rd

mode streamers consist of a bright filamentary channel and tree-like dark channels. The 4th

modes streamers are normally formed of filamentary channels with high light intensity [61,

72].

2.3.2 Streamer and Breakdown under Different Impulse Waveforms

Breakdown measurements in liquids under lightning impulse voltage are more

representative of conditions used to test real apparatus such as transformers. Step impulse

has the advantage of fast rise time and long tail time and is favourably used to interpret

breakdown phenomena since the applied voltage remains constant while the streamer

propagates [23]. An early study [73] in 1979 talked about the interactions between voltage

shape, streamer propagation, and breakdown. Figure 2-15 shows the characteristic voltage

49


for oil breakdown on an impulse waveform. In order for breakdown to occur a streamer

must initiate (Vi) and propagate across the gap before the voltage drops to zero. In fact,

breakdown occurs on the tail at a voltage which is considerably larger than zero, which may

confirm that a minimum voltage for breakdown Vm exists [73].

Figure 2-15 Characteristic voltage for oil breakdown on an impulse waveform, where L is gap distance, ti and

tb are the times corresponding to initiation and breakdown [73].

Figure 2-16 shows the effect of tail time of impulse on breakdown voltage at different gap

distances carried out with point-plane geometry in transformer oil [73]. It is clear that the

breakdown voltages increase with the decrease of tail time. When time constant is large

enough (step-like impulse), V0 is comparable to the minimum voltage for breakdown Vm.

This means the effects of parameters such as wave shape on impulse behaviour can be

predicted [73].

Figure 2-16 The effect of tail time on breakdown voltage in transformer oil at different gap distance, positive

point-plane electrode, gap distance: (o): 2.54 cm; (): 5.08 cm; (Δ): 10.16 cm; (V): 14.61 cm [73].

An interesting phenomenon related to breakdown voltage and acceleration voltage in

mineral oil under lightning and step impulse was also observed in [42] as shown in Figure

ti

50


2-17. At very small gaps, lightning breakdown voltage is identical to breakdown voltage

obtained under step voltage. With increasing gap distances, lightning breakdown voltage

becomes higher than the breakdown voltage under step voltage, probably due to the short

tail of lightning waveform, but is still lower than the acceleration voltage under step voltage.

When the gap distance reaches 55 mm, lightning breakdown voltage is almost the same as

the acceleration voltage obtained under step voltage [42].

Figure 2-17 Comparison of breakdown voltage and acceleration voltage in mineral oil under lightning and

step impulse, negative polarity [42].

Afterwards, two recent publications [22, 23] indicated that different impulse waveforms

have an influence on breakdown mechanism in both polarities, and came up with a theory

of breakdown voltage prediction. Experimental results of streamer under lightning (1.2/50

µs) and step impulse (0.4/1400 µs) were shared and compared [22]. Figure 2-18 shows the

comparison of the stopping lengths between lightning and step impulse voltage under

positive polarity [22]. Under positive polarity, the solid line indicates the growing trend of

stopping length with increasing voltage applied. It was found that the stopping length

increases linearly with applied voltage under both lightning and step impulse voltage.

Nevertheless, at the same applied voltage level, streamers under the step impulse propagate

further than those under the lightning impulse voltage. The measured stopping length under

lightning impulse is about 60% of that measured under step impulse. The shorter stopping

length under lightning impulse voltage is as a result of voltage decaying during streamer

propagation [22].

51


Figure 2-18 Comparison of stopping length between step impulse and lightning impulse under positive

polarity, d = 50 [22].

Average propagation velocity va is determined by the ratio of stopping length to

propagation time measured, based on the streamer current or the light signal. In the case of

a breakdown event, va is calculated by using gap distance d divided by time to breakdown tb.

Figure 2-19 shows the comparison of average propagation velocity between lightning and

step impulse voltage under positive polarity. Under positive polarity, the va under lightning

impulse follows exactly the same trend as that under step impulse, at about 2 km/s for the

2nd

mode streamer (responsible for Vb) [22].

Figure 2-19 Comparison of average propagation velocity between lightning impulse and step impulse under

positive polarity, d = 50 mm [22].

52


Based on the results in Figure 2-18 and Figure 2-19, different impulse voltage waveforms

have an influence on streamer and breakdown properties under positive polarity. Table 2-3

gives the comparison of 50% breakdown voltage and acceleration voltage under both

lightning and step impulse [22]. When the streamer propagates in the 2nd

mode (about 2

km/s), the time that the streamer takes to bridge the whole gap (50 mm) is 25 µs. During

this time, lightning impulse voltage drops from 126 kV to about 94 kV. The moment of

voltage drop (94 kV) presents the instantaneous breakdown voltage Vi. An interesting

phenomenon was observed when the instantaneous breakdown voltage under lightning

impulse is very close to the breakdown voltage (93 kV) measured under step impulse. In

terms of acceleration voltage, fast streamers bridge the whole gap in a short time, in which

the voltage drops under lightning impulse could be ignored [22].

Table 2-3 Breakdown voltage and acceleration voltage obtained under step impulse and lightning impulse

[22].

` Breakdown voltage (kV) Acceleration voltage (kV)

Positive Negative Positive Negative

Step impulse 93 174 260 240

Lightning impulse 126 228 260 250

At small gap distance (typically d < 1 cm) under positive polarity, the 50% breakdown

voltage under lightning and step impulse are almost the same due to the short time to

breakdown. At considerably large gap distances (typically d < 12 cm), the short tail time of

lightning impulse is not able to cause the streamer to bridge the whole gap in the 2nd

mode,

and so breakdown can only occur when the streamer is transformed into the fast streamer

[22, 42]. In addition, this phenomenon may not apply to the breakdown under negative

polarity, due to large variation and uncertainty [22].

This phenomenon was confirmed again based on a large number of experimental results in

[23]. The breakdown tests were carried out in both mineral oil and ester liquids under point-

plane and point-sphere electrode geometries with a large range of gap distance up to 35 cm

[23, 42]. Two different impulse waveforms were used, including the standard lightning

impulse (1.2/50 µs) and a specific step impulse. Figure 2-20 shows the breakdown voltage

and time to breakdown versus gap distance in point-sphere geometry in various insulating

liquids [23]. It was found that with the increase of gap distance, the increase of breakdown

voltages does not follow the exact same trend in terms of liquid nature. The different

transition points marked as a dashed line exist in various liquids as shown in Figure 2-20

53


(a). In addition, the transition points in breakdown voltage correspond to the reduction of

time to breakdown. The reduction of time to breakdown in a large gap distance indicates

the transformation from slow streamer to fast streamer, which leads to the conclusion in

[22].

Figure 2-20 Comparison of (a) Breakdown voltage VbLI

and (b) time to breakdown tbLI

under lightning

impulse versus gap distance in liquids of group IV (PMO, NMO, PB) [23, 42].

Figure 2-21 shows breakdown velocity versus gap distance in different liquids under

lightning impulse (a) and step impulse (b) [23]. Under lightning impulse, a uniform

streamer propagation velocity at about 2 km/s in a mineral oil was observed at small gap

distance (3 cm ≤ d ≤ 12 cm), which corresponds to the propagation of the 2nd

mode

streamer [44]. Under step impulse, the streamer propagation velocity at gap distance from 2

cm to 20 cm remains constant at 2 km/s. It is reasonable that under large gap distances, a

higher crest voltage of lightning impulse must be induced to allow the streamer to

54


accelerate (vp > 10 km/s) and achieve breakdown such behaviour strongly depends on the

liquid nature [23].

Figure 2-21 Average breakdown velocities vbLI

versus gap distance in insulating liquids, (a) under lightning

impulse; (b) under step impulse [23].

Figure 2-22 shows the comparison of breakdown voltage and instantaneous breakdown

voltage in mineral oil under lightning and step impulse. Firstly, breakdown voltage under

lightning VbLI

and step impulse VbST

are extremely close in small gaps (d ≤ 2 cm). Then, VbLI

becomes larger than VbST

with the increase of d (d ≥ 2cm). However, the instantaneous

breakdown voltage ViLI

(at the moment of breakdown) is identical to VbST

with the gap

distance d less than 12 cm. Finally, the difference between ViLI

and VbST

becomes larger

with d above 12 cm. Two cases (A and B) corresponding to different breakdown

mechanisms are able to explain the behaviour in mineral oil as shown in Figure 2-23. In

case A, to compensate for the voltage decay under lightning impulse, a crest voltage VbLI

higher than the minimum propagation voltage Vmin (similar to VbST

) must be applied to

induce breakdown. At this stage, the average propagation velocity in mineral oil still

remains constant, and VbLI

should not exceed the acceleration voltage Va. A past study [74],

indicated that breakdown can be induced only if Vi does not drop below Vmin before the

streamer propagates to the opposite electrode. Therefore, the breakdown criterion is able to

be written as [23]:

𝑽(𝒕 = 𝒕𝒃) = 𝑽𝒎𝒊𝒏 (2-1)

When time to breakdown corresponds to 2nd

mode streamers, an interesting phenomenon

was observed in which ViLI

highly corresponds to the criterion VbST

≈ ViLI

≈ Vmin. Moreover,

the average breakdown velocity under lightning impulse voltage should remain constant

(a) (b)

55


and similar to step voltage at about 2 km/s as shown in Figure 2-21. At a short gap distance

of less than 3 cm, the time to breakdown is relatively shorter, e.g. tb = 10 µs at d = 2 cm.

The lightning impulse voltage has less time duration for the voltage decay, about a 10%

drop. As a result, VbLI

is quite close to Vmin (VbST

≈ VbLI

≈ Vmin). In case B where the gap

distance is larger than 12 cm in mineral oil, VbLI

is potentially close to Va, which results in

the occurrence of a fast streamer as shown in Figure 2-20 and Figure 2-21.

Figure 2-22 Comparison of breakdown voltage in mineral oil PMO under step and lightning voltages in

point-sphere geometry [23].

Figure 2-23 Typical breakdown cases in mineral oil under lightning impulse voltage (full lines) and step

impulse voltage (dotted lines) [23].

56


Figure 2-24 compares the breakdown voltages Vb in a natural ester under lightning and step

impulse in point-plane geometry with the gap range from 2 cm to 20 cm [23]. The results

indicate that breakdown voltages under step voltage VbST

are nearly equal to breakdown

voltage under lightning impulse VbLI

, which is reasonable since breakdown is a result of the

fast streamer. Due to the high velocity of fast streamers, the decay of the lightning impulse

voltage is nearly negligible at this moment, which explains why VbST

≈ VbLI

in a natural

ester.

Figure 2-24 Comparison of breakdown voltage in natural ester NE under step and lightning impulse voltage

in point-plane geometry [23].

Overall, the breakdown characteristics of various liquids at various gap distances were

studied under lightning and step impulses. However, there is still a lack of breakdown

studies regarding other waveforms to develop the theory further. Therefore, a highly

flexible voltage generator will be built to investigate the breakdown characteristics in

liquids under impulse waveforms with different tail-time in this study.

2.4 Gas Generation in Liquids under Electrical Faults

2.4.1 Background Knowledge of Fault Gas Analysis

2.4.1.1 Theory

Discharges and breakdown cause chemical degradation of insulating liquids and generate

fault gases [75]. The insulating liquids are composed of different hydrocarbon atomic

groups, e.g. CH2, CH3 and CH. The molecular bonds used to link the molecular group

together (e.g. C-C and C-H bonds), will be broken when electrical or thermal energy is

57


stressed on the liquids. Newly formed unstable radical or ionic fragments will recombine

swiftly into gas molecules in the forms of ethylene (CH2=CH2), ethane (CH3-CH3),

methane (CH3-H), hydrogen (H-H), acetylene (CH≡CH), CO (C≡O) and CO2 (O=C=O).

Different energy levels are required to break different kinds of molecular bonds. Therefore,

different types and concentration of fault gases will be associated with the severity and

category of the transformer faults. The energy which is compulsory to crack the typical

molecular bond inside the insulating liquids is shown in Table 2-4 [21].

Table 2-4 Bond Dissociation Energy [21].

Bond C-C

(CH3-CH3) C-H (average)

C=C

(H2C=CH2)

C≡C

(HC≡CH)

Dissociation energy

(kJ/mol) 356 410 632 837

The IEC standard 60599 [76] classifies the electrical fault into three types: partial discharge

(PD), D1 (discharge of low energy) and D2 (discharge of high energy). Partial discharge

stands for the kind of streamer that only partially bridges the insulation gap between two

electrodes. Breakdown occurs after the streamer fully propagates through the gap of the

electrode, and it will act as small arcs known as sparking faults, called D1 and D2.

Compared to the PD fault, the sparking fault generates more amount of fault gases under

the same fault time and could have a critical effect on transformer operation.

Partial discharge, low energy sparking and arcing are some of the common faults that could

occur in oil-immersed transformers. When any of these faults occur, the insulating liquids

will be decomposed, and then a certain amount of combustible and non-combustible fault

gases will be generated [77-79].

2.4.1.2 Measurement

Dissolved gas analysis (DGA) is one of the most widely used diagnostic tools of oil-filled

transformers. It is known as the non-interrupt test method which has already functioned for

many decades. The function of DGA, similar to a human blood test, is to detect the health

problems of the transformer by analysing these fault gases dissolved in insulating liquids

[80]. These fault gases could be detected by DGA in a series of processes: collect oil

sample, extract dissolved gas, gas chromatograph and data interpretation.

58


According to IEC 60475 [81], the procedure of oil sampling involves carefully transferring

the oil sample from oil-filled equipment into a gas-sealed syringe. The international

standard IEC 60567 [79] describes four main methods for extracting gas, including Toepler

pump extraction, partial degassing extraction, stripping extraction and Headspace method.

After gases have been extracted from the dissolved oil, every single gas component can be

measured by means of various detection methods. IEC 60567 [79] describes the method of

gas chromatography (GC), which has been used to analyse dissolved gas in transformer oils

over 60 years. Furthermore, there is another method, known as photoacoustic spectroscopy

(PAS), which could also be used to measure gas components [82].

2.4.1.3 Data Interpretation

After the measurement of fault gas concentration, data interpretation is used to analyse the

type of fault based on the gas concentration and patterns in transformer oil. Based on the

international standard IEC 60599 [76], the main methods of data interpretation are the key

gas method, various gas ratio methods and the Duval triangle method.

The key gas method is a simple and acceptable method to monitor and determine the fault

types in transformer oil. The total dissolved combustible gases (TDCG) is calculated by the

percentage (%) of each combustible gas component in the total volume of evolved gases,

and normally used to determine the fault severity and its nature [83]. The key gases and

secondary gases that correspond to fault types are given in Table 2-5 [76].

Table 2-5 Indicator fault gases in mineral oil [76]

Fault types Key gases Secondary gases

PD in oil H2 CH4, C2H4 and C2H6

Breakdown in oil C2H2 H2, CH4 and C2H4

CO, CO2 (if involved cellulose)

Overheated oil C2H4 CH4, C2H2 and C2H6

Overheated cellulose CO CO2

There are several gas ratio methods, including the Dornenburg ratio method, Rogers ratio

method and IEC ratio method [83]. The most popular one is IEC 60599 ratio method which

calculates the three different ratios of the concentration of two specific fault gases,

including C2H2/C2H4, CH4/H2 and C2H4/C2H6. The Duval triangle method has been widely

used to detect the fault type in transformer oil since 1860 [84]. So far, there are seven

different Duval triangles used for different conditions, such as load tap changer, non-

mineral oil, specific temperature and thermal fault etc [85].

59


The on-line DGA monitor is a fully sealed system that can reflect the real-time operating

condition of the transformer. Compared with traditional DGA measurement methods, it

saves time transporting oil sample to a laboratory and avoids gas leakage in transit. An on-

line DGA monitor can be divided into single-gas monitor and multi-gas monitor.

The hydrogen monitor is a low-cost device mainly used in distribution transformers. It can

be directly installed into the valve of the transformer position between the cooling tank and

main tank for better oil circulation [86]. Unlike traditional DGA measurements, a hydrogen

monitor directly measures hydrogen in ppm level in the insulating oil. The hydrogen

monitor utilises a patented, solid-state hydrogen sensor that is immersed directly into the oil,

eliminating membranes and the potential for rupture [87]. It has potential benefit since all

types of electrical faults and some types of thermal faults involve a certain amount of

hydrogen generation.

The multi-gas monitor is a costly device and is mainly used in large power transformers.

Generally, it can measure multiple gases, including H2, CH4, C2H2, C2H4, C2H6, CO, CO2

and O2. Compared with the hydrogen monitor, a multi-gas monitor performs more accurate

measurements (the accuracy can be ± 5% of reading) [75]. Some multi-gas monitors are

similar to traditional lab based DGA measurements (e.g. Serveron TM8), since they have a

complete GC system of DGA measurements. Others might use photoacoustic spectroscopy

technique [82].

2.4.2 Fault Gas Studies under Electrical Faults

Most of the previous studies focused on electrical faults under AC voltage and very few

past studies carried out DGA analysis of electrical faults under impulse voltage. Statistics

show that the failure rate of HVDC converter transformers is approximately twice that of

AC transformers [88]. The valve windings ensure AC, DC, and repetitive impulse

combined voltages caused by inverters for an extended period of time [89]. The summary

of previous studies on fault gas generation under electrical faults is shown in Table 2-6 [90-

99]. In these studies, the applied number of breakdowns is up to 200. The point-plane

geometry with the non-uniform electric field could be successfully applied. The gap

distance is formed in a wide range from 1 to 45 mm. Both mineral oil and ester liquids were

considered.

60


Table 2-6 Summary of experimental works of DGA analysis under electrical faults [90-99].

Author Voltage

shape

Number of

breakdown

s

Electrode Gap

(mm) Oil type

Oil sample

(Liters)

Jovalekic

[90] Impulse 90 Point-Point 4

Mineral oil, Nature

Esters, Synthetic

Ester

1.62

Suwarno

[91] AC N/A Point-plane 1, 2, 3

Mineral oil, Nature

Esters N/A

Z. D. Wang

[92] AC 15 Point-plane 35

Mineral oil,

Synthetic Ester 2.57

Perrier

[93] AC 100

Spherical

electrodes 2.5

Mineral oil, Nature

Esters, Synthetic

Ester

0.4, not

sealed

Muhamad

[94] AC 200 Point-plane 45

Mineral oil,

Synthetic Ester N/A

Bandara

[95] AC

50, 75,

100

Three

electrodes 2

Mineral oil, Nature

Esters N/A

Gómez [96] AC 12 Plane-plane 2.5 Mineral oil, Nature

Esters N/A

Ghani [97] AC 6 Spherical

electrodes 2.5

Mineral oil,

Synthetic Ester

0.35 -

0.5

Eberhardt

[98] AC 10

U shaped

bow-plane 10, 20

Mineral oil, Nature

Esters, Synthetic

Ester

17

Khan [99] AC 3 Point-plane 10

Mineral oil, Nature

Esters, Synthetic

Ester

1.5

2.4.2.1 Electrical Fault under Impulse Voltage

A DGA test in a mineral oil and three ester liquids (two natural ester and one synthetic ester

liquids) under the sparking fault stressed by lightning impulse was carried out in [90]. A 4-

stage impulse generator is used as the voltage supply. A test cell with a volume of 1.618 L

contains a point-to-needle electrode of 4 mm gap distance. The applied impulse voltage

(1.2/50 µs) was set at 134 kV. Figure 2-25 shows the growing trend of key gases in mineral

oil with an increasing number of breakdowns [90]. The two dominated fault gases − H2 and

C2H2, increase linearly with the number of breakdowns in mineral oil. When the mineral oil

suffers fewer than 30 breakdowns, the generation of C2H2 is comparable to H2. And then,

the concentration of C2H2 exceeds the H2 afterwards and the difference becomes greater

with the increasing number of breakdowns. Figure 2-26 shows the concentration of fault

gases in four different liquids after 90 breakdowns [90]. It was found that both H2 and C2H2

are key gases in four liquids. Compared with ester liquids, the concentrations of H2 and

C2H2 in mineral oil are almost twice those in ester liquids, which is different from the

results under AC sparking fault in [92]. Therefore, it is worth investigating the

characteristics of gas generation under impulse voltage. In addition to H2 and C2H2, small

61


amounts of C2H4, CH4 in mineral oil and C2H4, CH4, CO in ester liquids were measured

after breakdown.

Figure 2-25 key gases generation versus number of breakdowns in mineral oil, d = 4 mm [90]

Figure 2-26 Comparison of fault gases generation in various liquids after 90 breakdowns [90]

2.4.2.2 Electrical Fault under AC Voltage

The fault gas generation under arcing fault in different voltage levels and arc durations was

carried out in [91]. The arcing fault was generated in a needle-to-plane electrode geometry

made of steel under AC voltage. Through the adjustment of the gap distance between the

point and plant electrode in 1, 2 and 3 mm, three types of arc energy under different applied

voltage (12 kV, 20 kV and 24 kV) were achieved. The duration of arc application was

chosen as 5, 10 and 15 minutes, respectively.

62


Figure 2-27 and Figure 2-28 show the comparison of fault gas generation under different

voltage levels and arc durations in mineral oil [91]. The results indicate that either under

different voltage levels or arc durations, C2H2 dominates in the combustible gases generated

by the arc. The concentration of all the combustible gases increases almost exponentially

with the applied AC voltage. In contrast to voltage levels, fault gas generation under

different arc duration increases almost linearly with the duration of arc application. Based

on the key gas method, oil samples under 12 and 20 kV indicate C2H2 as key gas and

correspond to arcing fault. However, when applied voltage increased to 24 kV, key gas of

C2H4 was also found indicating overheating. This phenomenon probably occurs because

high-voltage arcing releases a large amount of energy that heats the oil sample.

Figure 2-27 Fault gases generation in mineral oil under different voltage levels, arc duration = 15 mins [91].

Figure 2-28 Fault gas generation in mineral oil under different arc durations, V = 20 kV [91].

63


Table 2-7 and Figure 2-29 show fault gas generation in Gemini X and FR3 under sparking

fault [92]. In the experimental setup, a test cell with a volume of 2.57 L was used and a

needle-to-plane electrode geometry with 35 mm gap distance was installed in the test cell.

The results show the volumes of fault gases generated for each of the three groups of 15

breakdowns. It was found that the key gases were H2 and C2H2 in both liquids, with the

amount of H2 nearly twice that of C2H2, which is opposite to the results obtained in [91].

The higher concentration of H2 in this study is owing to the calculation of the total fault gas

generation, rather than dissolved gas only in the oil. Normally, due to the low solubility of

H2, a large amount of H2 generated under electrical fault easily escapes to the air or

headspace in a sealing system, which explains why a lesser amount of H2 was measured in

past studies [78, 91, 93]. In addition, compared with Gemini X, FR3 generates larger

amounts of CO, which leads to smaller percentages of hydrocarbon gases in FR3.

Table 2-7 Fault gases volumes (µL) in Gemini X and FR3 for each of three groups of 15 breakdowns [92].

Gas Gemini X FR3

Group 1 Group 2 Group 3 Group 4 Group 1 Group 2 Group 3 Group 4

H2 553.5 544.5 586.5 555.0 499.5 499.5 546.0 514.5

CH4 37.5 37.5 39.0 37.5 16.5 16.5 16.5 16.5

C2H6 4.5 4.5 0.0 3.0 1.5 0.0 1.5 1.5

C2H4 45.5 46.5 45.0 46.5 37.5 30.0 36.0 34.5

C2H2 258.0 265.5 261.0 261.0 237.0 217.5 259.5 238.5

CO 3.0 4.5 1.5 3.0 108.0 97.5 127.5 111.0

TCG1 880.5 903.0 933.0 906.0 900.0 861.0 987.0 916.5

1TCG = total combustible gases

Figure 2-29 Fault gases generation in different insulation oils under sparking fault (re-produced plot based on

results in Table 2-7) [92].

64


Figure 2-30 shows the fault gas generation per unit fault energy in Gemini X and FR3 [92].

It was found that the respective generation of H2 and hydrocarbons per unit fault energy

produced in FR3 were comparable to those produced in Gemini X. Nevertheless, a higher

amount of carbon monoxide was generated in FR3. Therefore, the total combustible gas

volume per unit fault energy (μL/J) in FR3 is slightly higher (< 20%) than that in Gemini X.

This finding suggests that the conventional gas methods for diagnosing sparking faults in

mineral oil are also applicable to natural ester liquids (FR3) [92].

Figure 2-30 Fault gases generation per unit fault energy (µL/J) in Gemini X and FR3 under sparking fault,

averaged over a group of three tests. TCG = total combustible gases [92].

In addition, some other papers [93-99] carried out similar electrical breakdown tests under

AC voltage. In summary, hydrogen and acetylene are key gases in a mineral oil, a synthetic

ester and five vegetable oils under low energy discharge [93]. Most of the fault gases

increase proportionally with moisture level [94]. The dissolved gas concentration increases

linearly with the number of sparks or arc duration and increases exponentially with applied

voltage in the mineral oil under electrical breakdown fault [95-98]. Compared to mineral oil,

FR3 generates significantly higher amounts of fault gases but less readily absorbs these

gases into the fluid as dissolved gases, and MIDEL 7131 generates the lowest amount of

fault gases [99].

2.5 Summary

An overview of streamer and breakdown in liquids under DC voltage was firstly given in

this chapter, mainly based on experimental methodologies and previous publications related

to streamers and breakdowns. It indicates that the basic feature of streamers in mineral oil

65


including current, emitted light and shape under DC voltage are comparable to previous

studies under impulse voltage. Moreover, the effect of background pressure on streamer

shapes and the effect of voltage polarity on breakdown voltage under DC voltage were

observed. However, to date there is a lack of study of dielectric performance in ester liquids

under DC voltage.

Then, the fundamental knowledge of streamer features and breakdown properties under

lightning and step impulse was reviewed. The publications have shown that, at a certain

range of gap distance, the relationship between breakdown voltage, instantaneous

breakdown voltage and time to breakdown follows a mathematical rule under different

impulse waveforms in liquids. Based on this theory, more comprehensive data without

complicated experiments can be obtained by simply measuring tb and Vi. And then, a

programme can be used to search the breakdown voltage from a large range of breakdown

values, which meets the criteria. However, the supporting publications only involved two

impulse waveforms. To support and promote this theory, there is a strong need to

investigate this phenomenon in different liquids under various impulse waveforms

Furthermore, the general knowledge of gassing behaviour in liquids and the past studies of

fault gas generation under electrical faults was explained. The publications have shown that

under spark fault, acetylene and hydrogen are the main gases, and the amount of hydrogen

and acetylene increases with the number of sparks and breakdown voltage levels. Most

studies have focused on electrical faults generated under AC voltage, which has a poor

record of breakdown number and fault energy control. It is worth creating a test system

with an auto-controlled spark generator in the attempt to find out the correlation between

gas generation and energy injection under electrical faults.

66


67


CHAPTER 3. EXPERIMENTAL DESCRIPTION

3.1 Liquids under Investigation

In this thesis, a mineral oil Gemini X and a synthetic ester liquid MIDEL 7131 were

investigated. The mineral oil with 3% aromatic content was tested in all the experiments as

the benchmark for cross comparison between the ester liquid and the mineral oil. The

synthetic ester liquid is a type of pentaerythritol ester, formulated synthetically by a

combination of chemicals. The basic properties of the investigated liquids are listed in

Table 3-1.

Table 3-1 Basic properties of testing liquids: MIDEL 7131 and Gemini X [42].

Property Unit Gemini X MIDEL 7131

Density g/cm3 0.882 0.97

Viscosity mm2/s 8.7 28

Pour point °C -60 -60

Flash point °C 144 275

Acidity mgKOH/g <0.01 <0.03

Aromatic content % 3

Water content mg/kg <20 50

Furan content mg/kg <0.1

Antioxidant, phenols Wt% 0.38

Dissipation factor, 90°C - <0.03 <0.03

Breakdown voltage kV

Before treatment - 40-60 (2.5 mm)

After treatment - > 70 (2.5 mm) > 75 (2.5 mm)

Permittivity - 2.2 3.2

3.2 Sample Preparation

For streamer studies under DC and impulse voltages, a pre-processing procedure including

filtering, degassing and dehydrating was performed on all liquid samples, in order to

minimise the effect of moisture and impurity on the test results. For gassing behaviour tests,

the oil samples were only dehydrated and degassed. The detailed procedure is described in

the following paragraphs.

First of all, the experimental oil samples were filtered by a Nalgene® MF 75 nylon

membrane filter with a pore size of 0.2 μm. The particle numbers in unfiltered and filtered

liquids were measured by using HIAC/ROYCO 8000A 8-channel particle counter with an

HIAC/ROYCO ABS2 automatic, which can detect particles in the diameter ranges of 1, 2,

5, 15, 25, 50, 100 and 200 μm. Based on previous results [11], the cumulative particle

68


number larger than 5 µm of filtered oil samples could reduce to about 500 per 100 mL oil

sample, which is close to the clean oil condition according to CIGRE Brochure [100].

Secondly, filtered oil samples were dehydrated and degassed in a vacuum oven at 85 ˚C and

500 Pa for over 48 hours in Gemini X and 72 hours in MIDEL 7131. After that, the oil

samples were given another 24 hours to cool down to ambient temperature under vacuum

conditions. The water content of the liquid samples after this processing was less than 10%

relative humidity.

3.3 Etched Tungsten Needles Based on Electrochemical Method

3.3.1 Basic Principle of Electrochemical Etching

Due to the demand of a large number of tungsten needles in the experiment, a feasible

method of producing the tungsten needle is required. Electrochemical etching is widely

accepted as an efficient and reliable method of producing sharp needles [101]. Figure 3-1

shows the principle of electrochemical etching in a static way and dynamic way [101]. In

the case of static etching, the meniscus shifts over time and the final probe becomes

irregular in shape. The dynamic electrochemical etching technique is optimised to produce

tungsten needles with controllable shape and radius of curvature, where the probe was

slowly lifted off from the liquid using a stepper motor [101]. The standard procedure is to

dip a tungsten wire into an electrolyte, and then bias DC voltage in the electrolyte as a

cathode and tungsten wire as an anode to start etching. The choice of DC voltage level and

electrolyte normally rely on the wire material. For the tungsten wire, the DC voltage level is

normally set at a lower voltage level, which ensures a much smoother tip. Generally,

potassium hydroxide (KOH) is used for tungsten wires.

Figure 3-1 Principle of static etching (green A-C) and dynamic etching (orange A-D) [101].

69


In IEC 60897 [102], it described a method for determining the radius of curvature of the

needle electrode by a metallographic microscope as shown in Figure 3-2. The radius R is

the distance from the tip circle of the needle to the centre of the circle along which the

needle curves. The simple way of measuring R without parameter a and b, is to draw a

circle overlapping the margin of the tip area as shown by the red dashed circle in Figure 3-2.

The radius of the circle is then identical to the radius of the tip.

Figure 3-2 Measurement of the radius of curvature of needles [102].

3.3.2 Etching Procedure of Tungsten Needles

In this thesis, tungsten needles with tip radii of 5 μm, 10 μm, 20 μm and 50 μm were

produced under well-controlled procedures by using an original 200 μm tungsten wire for

studying the effects of tip radius on streamer initiation. Before the process of etching

tungsten needles, a room temperature of 24 and humidity of 33.1% were measured. A 5

mol/L KOH was used as an electrolyte. A piece of tungsten wire was connected with the

cathode electrode in the KOH solutions. Another smooth and straight tungsten wire to be

etched was connected with the anode electrode. The selection of the radius of a tungsten

wire relies on the expected tip radius of etched tungsten needles. The radius of a tungsten

wire has to be twice as large as the expected tip radius of etched tungsten needles. When the

15 V DC was biased on the sample, it is necessary to apply a dynamic method to etch the

tungsten wire as shown in Figure 3-1 [101]. After electrochemical etching, all etched

tungsten needles were checked using a microscope to ensure that the tip surface is smooth.

70


The acceptable tip radius range is set as ± 10 % of the expected value. Without the

automatic system, the dynamic method can be achieved by manual operation at a certain

frequency, e.g. 1 second per shot. The time duration of tungsten wire immersed in a

solution should be as short as possible, in order to improve the accuracy of electrochemical

etching. Otherwise, the etched tungsten needles are likely to be ineffective as shown in

Figure 3-3.

Figure 3-3 Failure results of etched tungsten needles without applying oscillating method.

Figure 3-4 Ideal etched tungsten needle with the tip radius of 50 µm.

Table 3-2 summarises the experimental conditions used to etch tungsten needles with the

four tip radii in this thesis. The time durations per shot for 20 and 50 µm are both 1 second,

and the number of shots is 51 and 37 times respectively. However, for 5 and 10 µm, a pre-

processed procedure of 5 seconds per shot was applied to greatly reduce the tip radius.

Then, a well-controlled procedure of 1 second per shot was applied to etching accurately.

Figure 3-4 shows the ideal etched tungsten needle with a tip radius of 50 µm. A red circle

shows the exact tip radius based on the measurement of software in the microscope.

71


Table 3-2 The experimental conditions used to produce various tip radii of tungsten needles.

Tip Radius (µm) DC Voltage (V) Time Duration/Shot (second) Shots

5 15 5 38

10 15 5 27

20 15 1 51

50 15 1 37

3.4 DC Voltage Tests

3.4.1 Experimental Setup

The experimental setup used to investigate the streamer propagation and breakdown

strength of insulating liquids under DC voltage is shown in Figure 3-5. A bipolar high-

voltage DC source with a maximum voltage of 100 kV was used to deliver the continuous

DC voltage. A 2 MΩ resistor RL was placed in series with the DC source and the test cell to

limit the breakdown current as well as to protect the high-voltage DC source. A

compensated RC voltage divider (VD-100) was used to measure the DC voltage applied in

the test cell. A current shunt (10 Ω) was placed at the low voltage potential side of the test

cell to record the current signals. All the experiments were carried out at room temperature

and ambient pressure.

Figure 3-5 Sketch of the experimental setup used in streamer tests under DC voltage.

A 10 ×10 ×10 cm Perspex made test cell with a volume of 1 litre was used to hold the

electrodes and liquid sample as shown in Figure 3-6. The test cell contained a needle-to-

plane electrode system with an adjustable gap distance ranging from 0 to 50 mm. The

Oscilloscope

HV Divider

DC

RL

Current

Shunt

Test cell

Power Supply

Flood

Light SIM 16 Camera

72


needle electrode is held by a brass cylinder tube on the top of the test cell, and the length of

the needle that is exposed from the brass tube is fixed at 2.5 cm. The tungsten needles with

the tip radii of 5 μm, 10 μm, 20 μm and 50 μm were produced based on electrochemical

technique. The brass made plane electrode has a diameter of 70 mm.

Figure 3-6 Perspex made cubic test cell used for investigating streamer and breakdown under DC voltage.

A 16-channel ultra-high speed camera Specialized Imaging SIM16 with high resolution

intensified CCD sensors was used to study streamer characteristics. The high resolution of

1360 × 1024 pixels ensures streamer structure analysis, e.g. the stopping length resolution

is down to 0.02 mm. In total 16 frames of photos can be captured. With tuning of time

delay, explosion time aperture value and the distance to the flood light, the shadowgraph

image quality was optimised.

An oscilloscope with a bandwidth of 1GHz is not only used for recording voltage and

current signal but also used to synchronously trigger the high-speed camera. Once there is a

streamer current triggered in the oscilloscope, a 5 V TTL signal will be sent out from the

oscilloscope to fire the camera. However, due to the time delay of transferring TTL signal

from the oscilloscope to the camera (about 75 ns) and camera trigger time (75 ns), there is

an approximate 150 ns delay of the streamer image compared with the current signal.

73


3.4.2 Experimental Procedures

The probability of the streamer initiation was determined by using a rising-voltage method.

Firstly, the initial voltage applied was set at 60% of expected streamer initiation voltage.

Then, the applied voltage was increased in 1 kV steps. For each step of applied DC voltage,

60-second duration was applied to allow the capturing of streamer initiation, and a

minimum 60-second interval was used between each step. Finally, twenty initiation

voltages per sample were obtained to determine the 50% streamer initiation voltage with tip

radii of 5 μm, 10 μm, 20 μm and 50 μm.

Then, the streamer characteristics were studied using a rising-voltage method from

initiation to near breakdown level. The DC voltage was increased step by step with an

increment of 5 kV under both positive and negative polarities. At each step of voltage level,

a series of ten streamer phenomena were investigated. To minimise the cumulative effect of

DC voltage stress on the oil sample, a minimum 60-second interval was allowed between

each streamer investigation. The oil sample and needle electrode were renewed after one set

of tests under each polarity.

Finally, the breakdown voltage was measured by also using a rising-voltage method. The

initial voltage applied was set at 70% of expected breakdown voltage and increased in 2 kV

steps. For each step of applied DC voltage, 60-second duration was applied to allow the

capturing of breakdown, and a minimum of 120-second interval was used between each

step. After breakdown occurs, a minimum of 5 minutes was allowed before starting the next

breakdown test. Twenty breakdowns per sample were obtained to determine the 50%

breakdown voltage under gap distances of 2 mm, 5 mm, 10 mm, 20 mm and 30 mm. The

oil sample and needle electrode were renewed after the test of each tip radius or gap

distance.

3.5 Impulse Voltage Tests


The test setup used to study streamer and breakdown of insulating liquids under impulse

waveforms with different tail times is shown in Figure 3-7. A compact impulse generator

was used to generate different impulse waveforms. A compensated RC voltage divider

(VD-100) was used to measure the output voltage. A similar test cell used for DC voltage

tests was also used to study the streamer under impulse waveforms with different tail-time.

74


The needle electrodes of tungsten needles with a fixed tip radius of 10 µm were produced

based on an electrochemical technique. The etched tungsten needle was regularly changed

after each set of breakdown tests. The gap distance of the electrode was fixed at 10 mm.

The current was recorded by a 10 Ω non-inductive current shunt placed at the low voltage

potential side of the test cell. The same setup of ultra-high speed camera used in DC voltage

tests was also applied to investigate streamer phenomena under impulse voltage.

Figure 3-7 Sketch of test setup of impulse tests with different tail-time.

To simply generate the impulse waveform with the different tail times, a flexible and

compact solid-state switch (BEHLKE HTS-901-10-GSM) based impulse generator was

developed as shown in Figure 3-8. This compact impulse generator can be divided into two

parts: high voltage part (a) and control part (b). For the high voltage part, A bipolar high-

voltage DC source with a maximum voltage of 100 kV was used as a DC charging source.

A 500 MHz oscilloscope with the sampling rate of 1GS/s was used to monitor the impulse

voltage and current signals. A 2 MΩ resistor was placed in series with the DC source and

the impulse generator to limit the breakdown current as well as to protect the high-voltage

DC source as shown in Figure 3-7. Three parallel capacitors with a total capacitance of 1.8

µF were used as a charging capacitor C. Rs1 and Rs2 are 600 Ω protective resistors (grey

cylinder) used to prevent overcurrent damage to the solid-state switch. A set of the

replaceable front resistor Rf and tail resistor Rt (red cylinder), allows the generation of

positive impulse waveforms including 0.8/8 µs, 0.8/14 µs, 0.8/30 µs and 0.8/3200 µs. The

control part includes a shielding box and a control unit. The shielding box is used to

Oscilloscope

HV Divider

DC

Rf

Current

Shunt

SIM 16

Camera

Test cell

Power Supply

Flood

Light

A

B

Rt

Rs1

Rs2

RL

C

Control Unit

Impulse Generator

75


mitigate the effect of electromagnetic noise on the control signal. The separated control unit

is not only used to control the solid-state switch by a function generator, but also to monitor

the operation mode of the solid-state switch with LED indicators.

Figure 3-8 The photo of the compact solid-state switch based impulse generator.

3.5.2 Impulse Waveforms with Different Tail times

In this thesis, four impulse waveforms with different tail times (0.8/8 μs, 0.8/14 μs, 0.8/30

μs and 0.8/3200 μs) were used to investigate the streamer characteristics and breakdown

properties in liquids. Table 3-3 summarises the parameters of the front resistor, tail resistor

and charging capacitor used to generate the four impulse waveforms. The front time of four

impulse waveforms was fixed at 0.8 µs (Rf = 6.4 kΩ). The tail times of impulse waveforms

are determined by both tail resistor and charging capacitor. With the same charging

(a) High voltage unit

(b) Control unit

Solid-State

Switch

Charging

Capacitor

RS1

Protective

Resistors

Tail Resistor

Front

Resistor

76


capacitor, higher resistance of the tail resistor gives the longer tail-time impulse waveform.

With the same tail resistor, higher capacitance of the charging capacitor also gives the

longer tail-time impulse waveform.

Table 3-3 The parameters of the front resistor, tail resistor and charging capacitor used to generate the four

impulse waveforms.

Impulse

Waveforms

Front Resistor Rf

(kΩ)

Tail Resistor Rt

(kΩ)

Charging Capacitor C

(nF)

0.8/8 µs 6.4 3.2 1.8

0.8/14 µs 6.4 6.4 1.8

0.8/30 µs 6.4 40 1.2

0.8/3200 µs 6.4 54 1.8

Figure 3-9 shows the different impulse waveforms with tail time ranging from 8 µs to 3200

µs used in further streamer and breakdown studies. The voltage of the 0.8/3200 µs impulse

waveform does not decay for 10 µs, which is more than the maximum streamer propagation

time at the 10 mm gap distance. The 0.8/3200 µs impulse waveform is thus used as a ‘step-

like’ impulse.

Figure 3-9 The different impulse waveforms with tail time ranging from 8 µs to 3200 µs, V = 24 kV.


The streamer characteristics were studied by using a rising-voltage method from initiation

to near breakdown level. The impulse voltage was increased step by step with an increment

of 2 kV under positive polarity. At each step of voltage level, a series of ten streamer

-2 0 2 4 6 8 10 12 14 16 18 20

0

5

10

15

20

25

30

Vo

ltag

e (k

V)

Time (us)

-2 0 2 4 6 8 10 12 14 160

5

10

15

20

25

Vo

ltag

e (k

V)

Time (us)

0.8/8 us

0.8/14 us

0.8/30 us

0.8/3200 us

77


phenomena were investigated. To minimise the cumulative effect of impulse voltage

stressed on the oil sample, a minimum 60-second interval was allowed between each

impulse shot. The oil sample and needle electrode were renewed after the test of each

impulse waveform.

The breakdown voltages under different impulse waveforms were also measured by using a

rising-voltage method. The initial voltage applied was set at 70% of the expected

breakdown voltage. Voltage level was increased step by step with an increment of 1 kV.

Ten breakdowns per impulse waveform were obtained to determine the 50% breakdown

voltage.

3.6 Gassing Behaviour Tests


Figure 3-10 shows the sketch of the experimental setup used to investigate the gassing

behaviour of insulating liquids under electrical faults. Figure 3-11 shows the photos of the

testing area and the control area in the laboratory. In the testing area (Figure 3-11a), a

compact impulse generator used for previous impulse tests was also used to generate the

electrical sparks in gassing behaviour study. A 100 kV voltage divider was used to measure

the fault voltage level. A special designed gas-tight test cell was used to hold the electrodes

and liquid sample. A 10 Ω non-inductive current shunt used to measure the fault current

was placed at the low voltage potential side of the test cell. The on-line DGA monitors,

including a palladium sensor based Serveron TM1 hydrogen monitor and a gas

chromatography (GC) based Serveron TM8 multi-gas monitor, were employed to measure

fault gas generation. A stainless steel buffer chamber was built and connected to the TM1

hydrogen monitor to the oil circulation loop.

78


Figure 3-10 The diagram of the experimental setup for gas generation tests under electrical faults.

Figure 3-11 The photos of the experimental setup for gas generation tests under electrical faults.

In the testing area (Figure 3-11b), a national instrument (NI) data acquisition (DAQ) system

and a NI high-speed digitizer with the sampling rate of 5 GS/s were used to control and

acquire data. A high voltage DC source was used to charge the charging capacitor of the

compact impulse generator. A 5 V DC power supply supplies electric energy to the solid-

state switch of the impulse generator.

A cubic shaped stainless steel test cell with a volume of 1 litre used to hold a needle-to-

plane electrode system is shown in Figure 3-12. Two white bushings, capable of

withstanding up to 100 kV AC voltage (more than 100 kV for impulse voltage) are

connected at the each side of the cubic test cell. Two Perspex made windows are mounted

(a) Testing area (b) Control area

High-speed

digitizer

High voltage

DC source

500 MHz

Oscilloscope

5 V DC

source

Spark

generation

Voltage

divider

DGA test

system

79


on the front and back of the test cell, which allows the streamer/breakdown observation.

Two optic fibre connectors and one oil-in connector are fixed at the lid on the top of the test

cell. An oil-out connector is fixed at the bottom of the test cell. The gaskets were used to

fill the space between all the mating surfaces, which prevents leakage of oil sample from

the joined objects. Figure 3-12b shows the electrode configuration of the test cell. The

needle electrode with a tip radius of 50 µm was produced based on an electrochemical

technique. The brass plane electrode had a diameter of 20 mm. The gap distance of the

needle-to-plane electrode was fixed at 5 or 10 mm.

Figure 3-12 The photo of the cubic shaped stainless steel test cell used for gassing behaviour test, (a) test cell;

(b) electrode configuration with the gap distance of 10 mm.

3.6.2 Fault Control and Data Acquisition System Design

To generate a controlled number of breakdowns, an automatic control system of voltage

output and data recording was built up. The flow chart in Figure 3-13 shows the working

(a)

(b)

80


principle of the auto-controlled system. Since the spark generator is fully controlled by the

electronic trigger signal, a national instrument (NI) data acquisition (DAQ) system was

used to accurately control the breakdown numbers and the time interval of each breakdown.

In addition to the voltage output control, an NI high-speed digitizer was used to

automatically record the voltage and current signals. Once there is an electrical fault

generated in the liquid sample, the simultaneous fault voltage and current signal are stored

in the PC drive for further data analysis.

Figure 3-13 The flow chart of the automatic control system with voltage output and data recording.

The way of generating a certain number of breakdowns is achieved by using the time

domain in Labview setting as shown in Figure 3-14. The initial delay is starting time, which

is set as 0 second in this study. The high time is the time duration of triggering the impulse

generation, which should be larger than the time duration of impulse waveform. Insufficient

high time can chop the tail of impulse waveform. Low time is the time interval between

each triggering. The equation (3-1) shows the calculation of total time based on a certain

number of breakdowns.

81


Figure 3-14 Description of time domain in Labview setting.

𝐓 = 𝒕𝟏 × 𝒏 + 𝒕𝟐 × 𝒏 + 𝒕𝟎 (3-1)

Where t1 means high time; t2 means low time; t0 means initial delay; T means time out; n

means number of breakdowns.

3.6.3 Oil-loop System Design

With a volume of 1 litre in the test cell, a volume of 1.2 litres in the buffer chamber and an

extra volume of 0.5 litres in the TM8 multi-gas monitor, the total volume of the oil-loop

system is about 2.7 litres. To avoid leakage of dissolved gases during the experimental tests,

the sealing performance of the oil-loop system was tested before filling the oil sample. A

pressure gauge was connected within the oil-loop system. A syringe was connected at the

highest position of the oil loop and this was used to inject air into the oil-loop system. Once

the pressure of the oil-loop system reaches 100 mbar, all the valves should be switched off

and the pressure gauge monitored for 24 hours. Figure 3-15 shows the inner pressure

readings of the oil-loop system for 24 hours. The results indicate that there is only a 12%

drop of inner pressure after 24 hours. Considering the longest time duration of the

experiments is within 9 hours, the sealing performance of the test setup is acceptable.

Figure 3-15 Sealing performance based on pressure reading in oil-loop system.

82


Before filling the oil sample, the test cell and buffer chamber were washed and dried in an

oven at about 80˚C. Then, the processed oil sample was injected into the system by an oil

pump of the TM8 multi-gas monitor. At the end of the oil filling process, a syringe was

connected at the highest position of the oil loop to bleed the air. The test system is able to

carry out the experiment once the headspace existed in the test cell and the buffer chamber

is minimised. A fixed headspace of about 75 mL exists in the on-line DGA monitor and

another headspace of about 5 mL exists in the test cell. The oil sample is circulated by an

oil pump of the TM8 multi-gas monitor with an oil flow rate of 250 mL/min.


Before carrying out the gassing behaviour tests, the breakdown voltages in both liquids

were determined based on a rising-voltage method as described before. Table 3-4

summarises the 99.9% breakdown voltages of both liquids under lightning impulse with the

gap distances of 5 and 10 mm derived through the normal cumulative distribution.

Table 3-4 The 99.9% breakdown voltages of the mineral oil and the synthetic ester liquid at different gap

distance under positive and negative lightning impulse.

Oil

Types

Voltage

Polarities

Gap Distances

5 mm 10 mm

99.9% VB (kV) 99.9% VB (kV) 1.5 times 99.9% VB (kV)

Gemini

X Positive 31.3 39.1 58.6

MIDEL

7131 Positive 26.8 36.8 55.2

Moreover, to improve experimental efficiency and also to avoid the accumulative effect of

each breakdown on fault energy, an experiment of energy measurement with different time

intervals between each breakdown was carried out. Figure 3-16 shows the comparison of

fault generation with different time intervals of 1 minute, 2 minutes and 5 minutes. The data

is plotted for every 5 breakdowns with a total number of breakdowns of 200. The results

indicate that the energy generation remained stable in all tested time intervals. It shows

well-controlled output energy and no accumulative effect is observed even with the time

interval of 1 minute, probably due to the continuous circulation of the oil in the test system.

Therefore, 1-minute interval was used in the following study.

83


Figure 3-16 The comparison of energy generation of individual breakdowns with different time interval.

Figure 3-17 shows the flow chart of the gassing behaviour tests in both liquids from 20 to

500 breakdowns at either a 5 mm or 10 mm gap distance. Firstly, a 10-hour background

measurement by TM8 multi-gas monitor was allowed to measure the dissolved gases in the

oil sample. Then, after a certain number of breakdowns applied, another 8-hour

measurement by TM8 multi-gas monitor was allowed to measure the concentration of the

dissolved gases after the electrical fault. Finally, the tested oil of 50 mL was sampled from

the oil-out connector at the bottom of the chamber by using a full-sealed syringe. After

sampling, the buffer chamber is replenished by a new 50 mL oil sample. The difference

between the post-test reading and background reading is used to indicate gas generation due

to the breakdown.

84


Figure 3-17 The flow chart of the experimental procedure for gassing behaviours tests

At the 99.9% breakdown voltage level, the effect of voltage polarity was firstly investigated

with a different number of breakdowns (sparking) including 20, 50, 100, 200, 300 and 500

breakdowns under positive polarity at a gap distance of 5 mm. Then, the effect of gap

distance was investigated under positive polarity at a gap distance of 10 mm. Finally, an

additional set of 200 breakdowns test was investigated at the 10 mm gap distance but with a

higher voltage level i.e. 1.5 times of the 99.9% breakdown voltage. Each set of tests was

repeated twice under the same conditions to confirm the credibility of results. The time

interval between each breakdown was set at 1 minute, and the oil was continuously

circulated to help the dispersion of breakdown by-products e.g. gas bubbles, if there were

any.

3.7 Summary

Two types of liquid, a mineral oil – Gemini X and a synthetic ester liquid – MIDEL 7131,

are used in this study. The liquid samples were pre-processed through filtering, dehydrating

and degassing

An electrochemical method used to produce the etched tungsten wire was presented. The

etched tungsten wire with various tip radii of 5 μm, 10 μm, 20 μm and 50 μm were

produced under well-controlled procedures in this study.

85


For DC voltage tests, needle-to-plane electrodes with the tip radii of 5 μm, 10 μm, 20 μm

and 50 μm and adjustable gap distances up to 50 mm were contained in a cubic test cell. A

high-speed camera and a current shunt were used to observe the streamer characteristics.

An advanced oscilloscope was used to record the current signal and trigger the camera

synchronously.

For impulse voltage tests, a compact solid-state switch based impulse generator was built

up to deliver impulse waveforms with different tail times (0.8/8 µs, 0.8/15 µs, 0.8/30 µs and

0.8/3200 µs). Needle-to-plane electrodes with a fixed gap distance of 10 mm and a tip

radius of 10 µm are contained in a cubic test cell. A high-speed camera and a current shunt

were used to observe the streamer characteristics.

For gas generation tests, a DGA test platform based on a sealed circulating oil loop with

functions of automatic spark fault control and data acquisition was developed. Needle-to-

plane electrodes with a tip radius of 10 µm and a fixed gap distance of either 5 or 10 mm

were contained in a cubic shaped stainless steel test cell. A TM1 hydrogen monitor and a

TM8 multi-gas monitor were used to measure the gas generation throughout the experiment.

86


87

Chapter 4 Streamer and Breakdown Properties of Insulating Liquids under DC Voltage

CHAPTER 4. STREAMER AND BREAKDOWN PROPERTIES

OF TRANSFORMER LIQUIDS UNDER DC VOLTAGE

4.1 Introduction

This chapter presents experimental studies on streamer characteristics and breakdown

strengths of a mineral oil and a synthetic ester liquid under DC voltage. The effects of tip

radius and gap distance on streamer initiation and breakdown voltage are analysed under

both positive and negative polarities. Streamer shape, stopping length, average propagation

velocity and area are compared between the mineral oil and the synthetic ester liquid.

Finally, the dielectric strengths of the mineral oil and the synthetic ester liquid are defined.

4.2 Effect of Tip Radius on Streamer Initiation Voltage

The typical current and emitted light signals of initiated streamer in the mineral oil and the

synthetic ester liquid observed under both positive and negative polarities are shown in

Figure 4-1. Under positive polarity, the current and emitted light signals are more intensive

and consist of continuous current components with several discrete large single pulses.

Under negative polarity, only one single pulse is detected for both the current and light

signals. The time duration of negative pulses is extremely short (less than 100 ns). The

maximum pulse peak of current signals remains at 2 mA at the initiation stages for the

mineral oil and the synthetic ester liquid under both positive and negative polarities. Higher

magnitudes of emitted light signals were recorded under positive polarity, which explains

why streamer initiation under positive polarity is brighter than under negative polarity.

88


Figure 4-1 The current and emitted light signals of streamer initiation under DC voltage (a). Mineral oil –

positive polarity (b). Mineral oil – negative polarity(c). Synthetic ester – positive polarity (d). Synthetic ester

– negative polarity; d = 10 mm, r = 10 µm.

To statistically analyse the streamer initiation voltage, Weibull distribution was used to fit

the initiation results of 20 tests and calculate 50% streamer initiation voltage. Its cumulative

distribution function is given as shown in equation (4-1). Knowing the specific shape and

scale parameters of Weibull function, the initiation voltage at a specific initiation

probability can be easily deduced. The Weibull distribution plots of the streamer initiation

voltages obtained under various tip radii of the mineral oil and the synthetic ester liquid are

shown in Figure 4-2, and the parameters of shape and scale are summarised in Table 4-1.

𝐹𝑊𝑒𝑖𝑏𝑢𝑙𝑙(𝑥) = 1 − 𝑒−(

𝑥

𝛽)𝑎

(4-1)

where, α is shape parameter and β is scale parameter.

0 0.5 1 1.5 2-2

0

2

4

Time [us]

Str

eam

er C

urr

ent

[mA

]

0 0.5 1 1.5 2

-0.1

-0.05

0

0.05

Em

itte

d L

igh

t [A

rb]

Streamer current

Emitted light

0 0.5 1 1.5 2

-2

0

2

Time [us]

Str

eam

er C

urr

ent

[mA

]

0 0.5 1 1.5 2

-0.02

-0.01

0

0.01

Em

itte

d L

igh

t [A

rb]

Streamer current

Emitted light

0 0.5 1 1.5 2

0

5

10

Time [us]

Str

eam

er C

urr

ent

[mA

]

0 0.5 1 1.5 2

-0.1

-0.05

0

Em

itte

d L

igh

t [A

rb]

Streamer current

Emitted light

0 0.5 1 1.5 2

-2

0

2

Time [us]

Str

eam

er C

urr

ent

[mA

]

0 0.5 1 1.5 2

-0.02

-0.01

0

0.01

Em

itte

d L

igh

t [A

rb]

Streamer current

Emitted light

(a) (b)

(c) (d)

89


Figure 4-2 Weibull distribution plot of streamer initiation results with various tip radii (r = 5, 10, 20 and 50

µm) under DC voltage, d = 10 mm.

Table 4-1 Weibull parameters of streamer initiation results with various tip radii at point-plane electrode

under negative and positive polarities

Tip Radius

(µm)

Polarity Shape Scale

50 %

Breakdown

Voltage

5

Gemini X (+) 8.8 6.8 6.5

(-) 6.5 5.8 5.5

MIDEL 7131 (+) 8.6 6.8 6.5

(-) 6.7 6.1 5.7

10

Gemini X (+) 9.1 7.6 7.4

(-) 6.1 5.5 5.3

MIDEL 7131 (+) 8.7 7.1 6.9

(-) 7.7 6.2 6.0

20

Gemini X (+) 11.1 8.9 8.6

(-) 8.1 7.4 7.1

MIDEL 7131 (+) 7.8 8.3 7.9

(-) 6.9 7.2 6.9

50

Gemini X (+) 7.6 11.5 11.9

(-) 8.5 8.4 8.0

MIDEL 7131 (+) 10.1 9.9 9.5

(-) 11.0 8.3 8.0

r = 5 µm r = 10 µm

r = 20 µm r = 50 µm

90


Figure 4-3 shows the effect of tip radius on streamer initiation voltage of the mineral oil

and the synthetic ester liquid under DC voltage. With the increase of tip radius, streamer

initiation voltages of both positive and negative streamers increase with tip radius. The

polarity effect is observed which means that that initiation voltage under negative polarity

is slightly lower than that of positive polarity. This is because positive space charges that

accumulate ahead of the positive point weaken the near tip field and suppress streamer

initiation under positive polarity, while positive space charges that accumulate ahead of the

negative tip enhance the near tip field and then decrease the initiation voltage under

negative polarity. Under positive polarity, when tip radius r = 5 μm, the initiation voltages

of the mineral oil are the same as that of the synthetic ester liquid; when r > 5 μm, initiation

voltage of the mineral oil becomes noticeably higher than that of the synthetic ester liquid.

However, under negative polarity, the initiation voltages of the mineral oil are almost the

same as those of the synthetic ester liquid for all tested tip radii.

Figure 4-3 Effect of tip radius on streamer initiation voltage under DC voltage, d = 10 mm (plot based on 50%

initiation voltage).

From the initiation voltage presented in Figure 4-3, initiation fields Ei at the local needle

tips were calculated according to equation 𝑬𝒊 =𝟐𝑽𝒊

𝒓𝒑 𝐥𝐧(𝟒𝒅

𝒓𝒑)

(4-1). With the increase of tip radius, the initiation voltage is increasing, but the initiation

field, calculated based on the inception voltage and electrode geometry, actually is reducing

as shown in Figure 4-4 under both positive and negative polarities. This probably can be

91


explained by the area effect that large surface area of electrode increases the probability of

defects on the electrode and the stressed oil samples near the tip electrode.

𝑬𝒊 =𝟐𝑽𝒊

𝒓𝒑 𝐥𝐧(𝟒𝒅

𝒓𝒑) (4-1)

where Ei means initiation field, Vi means initiation voltage, rp means tip radius and d means

gap distance.

Figure 4-4 Initiation field versus tip radius in the mineral oil and the synthetic ester liquid under positive and

negative polarities.

4.3 Basic Characteristics of Streamers

4.3.1 Positive Streamer

An example of the current signal and shadowgraph of positive streamer propagation in the

synthetic ester liquid at a gap distance of 10 mm is shown in Figure 4-5. The current signal

is intensive and consists of continuous current components superimposed with large

discrete pulses. A camera monitor signal was used to correlate the streamer shadowgraph

and the current signal. Streamer propagation based on shadowgraphs is shown in Figure 4-5

(b), where only one main branch with small offshoots was observed. A clearly growing

path of the main branch was observed from Frame 1 to Frame 3. The streamer tip continued

propagation for a short distance even after the current pulses ended, as shown in Frame 4.

At this stage, it was observed that the initial channel, root, of the streamer, has dissipated

92


into the liquid where the streamer tip propagation continues due to either local ionisation or

to simply gaseous channel expansion. The final stopping length of streamer in Figure 4-5(b)

is 7.34 mm with average propagation velocity of 1.56 km/s.

Figure 4-5 Typical positive streamer propagation in the synthetic ester liquid under DC voltage, d = 10 mm, r

= 10 µm, V = 28 kV; (a) voltage, current and monitor signals, (b) streamer propagation, corresponding to the

signals in (a).

Stopping Length

Below the breakdown voltage, streamers can initiate and propagate, but finally stop at a

certain distance that is less than the full gap distance. The final length is usually called

stopping length ls, which is one of the most important parameters to characterise a streamer.

Stopping length means the straight-line distance from the farthest tip point of a streamer to

the point electrode, which is measured based on the streamer images.

The results of streamer stopping length in the mineral oil and the synthetic ester liquid at a

10 mm gap under positive polarity are shown in Figure 4-6, where 50% breakdown

voltages are also stated as reference (VB-Gemini X and VB-MIDEL 7131 stand for 50% breakdown

voltage of the mineral oil and the synthetic ester liquid respectively). At each voltage level,

the mean and standard deviation are given based on 10 measurements. For the shadowgraph,

each frame has a resolution of 1360×1024 pixels, and the length (mm) per pixel is 0.0212

mm/pixel.

Under positive polarity, the synthetic ester liquid behaves similarly to the mineral oil at

inception voltage level V = 15 kV; the stopping length increases with an increase of applied

voltage. When applied voltage is higher than 15 kV, the stopping length of the synthetic

0 2 4 6 8 10 12

0

5

10

Curr

ent (

mA

)

0 2 4 6 8 100

10

20

30

40M

onito

r Sig

nal (

V)

Time (us)

0 2 4 6 8 10

0

5

10

Curr

ent (

mA

)

0 2 4 6 8 10-5

0

5

10

Mon

itor S

igna

l (V

)

Time (us)

0 2 4 6 8 10

0

5

10

Curre

nt (mA

)

0 2 4 6 8 100

20

40

Monito

r Signa

l (V)

Time (us)

0 2 4 6 8 10

0

2

4

Monito

r Signa

l (V)

Time (us)

30 kV

2µs

5 mA

Frame 1 Frame 2 Frame 3 Frame 4

(a)

(b)


93


ester liquid becomes longer than that of the mineral oil. The higher increasing rate, mm/kV,

of the synthetic ester liquid, results in a lower breakdown voltage compared with the

mineral oil. Figure 4-6(b) shows the fully propagated streamers at the different voltage

levels in both liquids. It is clear that higher voltage level indeed promotes streamer growth

in terms of length, whereas it does not have much effect on branching in the investigated

range.

(a) Streamer stopping length versus applied DC voltage

(b) Corresponding streamer shape at different voltage levels

Figure 4-6 Stopping length of streamers in the mineral oil and the synthetic ester liquid under positive

polarity, d = 10 mm, r = 10 µm (error bars stand for one standard deviation).

Average Propagation Velocity

Average streamer propagation velocity va, is calculated by the ratio of stopping length l to

propagating time t, with the assumption that streamers in the 2nd

mode propagate at a

constant velocity. Once breakdown occurs, va is determined by using the gap distance d

A1 B1 C1 D1

A2 B2 C2

94


divided by time to breakdown tb.

Figure 4-7 shows the comparison of average streamer propagation velocity in mineral oil

and synthetic ester liquid at a 10 mm gap and 10 µm tip radius under positive polarity. It

was found that, at the same applied voltage level, average propagation velocity in synthetic

ester liquid is slightly higher than that in mineral oil. The velocity of both liquids remains in

the range from 1.5 km/s to 2 km/s, from initiated voltage to near 50% breakdown voltage.

Figure 4-7 Average propagation velocity versus applied DC voltage in the mineral oil and the synthetic ester

liquid under positive polarity, d = 10 mm, r = 10 µm (error bars stand for one standard deviation).

Streamer Charge

The apparent charge from PD measurement serves as a key indicator of insulation defect.

This raises the question of whether the apparent charge threshold adopted can suitably

reflect the presence of streamers in liquids under DC voltage.

Synchronised measurements of streamer shape and current allow the correlations between

stopping length (shown in Figure 4-6) and maximum apparent charge (calculated based on

the integration of current signals) to be established. Figure 4-8 shows the positive streamer

stopping length as a function of maximum apparent charge in both the mineral oil and the

synthetic ester liquid under positive DC voltage. The results indicate that the correlations

between streamer stopping lengths and apparent charges increase linearly in both the

mineral oil and the synthetic ester liquid. This is probably because no obvious side branches

were developed during the growth of streamer and the charge density within the streamer

channels seems uniform and constant. It also indicates that injected charge is driving the

95


streamer growth. The rate is about 0.31 mm per 100 pC in the mineral oil and 0.26 mm per

100 pC in the synthetic ester liquid.

Figure 4-8 Positive streamer stopping length as a function of maximum apparent charge in the mineral oil and

the synthetic ester liquid.

4.3.2 Negative Streamer

A similar experimental procedure was carried out under negative polarity. Figure 4-9 shows

an example of the current signal and shadowgraph of negative streamer propagation in the

synthetic ester liquid at a gap distance of 10 mm. A train of clearly recognised discrete

pulses with increasing amplitude is observed in the current signal, which is similar to the

previous finding of the negative streamer observed under ac and impulse voltages [38, 103].

A camera monitor signal was used to correlate the streamer shadowgraph and the current

signal. Streamer propagation based on shadowgraphs is shown in Figure 4-9 (b). Unlike the

streamer observed under positive polarity, the negative streamer slowly propagates in the

form of a cloud like from Frame 1 to Frame 5, and it dissipates slower than the positive

streamer after the end of the current signal. The final stopping length of the streamer in

Figure 4-9 (b) is 1.13 mm, with an average propagation velocity of 0.12 km/s.

96


Figure 4-9 Typical negative streamer propagation in the synthetic ester liquid under DC voltage, d = 10 mm,

r = 10 µm, V = -55 kV; (a) voltage, current and camera monitor signals, (b) streamer propagation,

corresponding to the signals in (a)

Stopping Length

Figure 4-10 shows the streamer stopping length in the mineral oil and the synthetic ester

liquid at a 10 mm gap under negative polarity, where 50% breakdown voltages are also

stated as a reference (VB-Gemini X and VB-MIDEL 7131 stand for 50% breakdown voltage of the

mineral oil and the synthetic ester liquid respectively). Under negative polarity, a

significant difference in the streamer stopping length characteristic exists in both liquids.

The streamers in both liquids increase extremely slowly from inception voltage to the

voltage before breakdown. A stopping length of only about 1 mm can be observed in the

synthetic ester liquid at the applied voltage of 50 kV (VB-MIDEL 7131 = 59.8 kV), and only

about 0.7 mm for the mineral oil at the applied voltage of 80 kV (VB-Gemini X = 92.9 kV).

Similar to the results of positive streamers, the stopping length of negative streamers of the

synthetic ester liquid is longer than that of the mineral oil. The higher increasing rate,

mm/kV, of the synthetic ester liquid, results in a lower breakdown voltage compared with

the mineral oil.

Figure 4-10(b) shows the fully propagated streamers at the different voltage levels in both

the liquids. At the inception voltage level (A1 and A2), only a tiny streamer tip around the

point electrode is observed in both the liquids. At the voltage level before breakdown (D1

and D2), the negative streamers barely propagate (stopping length < ~0.1d) and are in the

form of a cloud like with no clear branches and offshoots. This is obviously different from

0 5 10 15 20-20

0

20

Cu

rren

t (m

A)

0 5 10 15 20

-50

0

Mo

nit

or

Sig

nal

(V)

Time (us)

0 5 10 15 20

0

2

4

Mo

nit

or

Sig

nal

(V)

Time (us)

0 5 10 15 20-20

-10

0

10C

urre

nt (m

A)

0 5 10 15 20-5

0

5

10

Mon

itor S

igna

l (V

)

Time (us)

0 5 10 15 20-20

0

20

Curre

nt (m

A)

0 5 10 15 20

-50

0

Mon

itor S

ignal

(V)

Time (us)

0 5 10 15 20

0

2

4

Mon

itor S

ignal

(V)

Time (us)

10 mA

(a)

5 µs

- 55 kV

F 1 F 2 F 3 F 4 F 5 F 6 F 7

(b)

Frame 1 Frame 2 Frame 3 Frame 4 Frame 5 Frame 6

97


the negative streamer characteristics observed under impulse voltages, where the stopping

length increases exponentially with an increase in applied voltage [40, 53]. After ionisation

occurs near the streamer tip region, the dissipated negative space charges seem to form a

strong shielding effect and hence weaken the boundary field, which prevents further

propagation of the streamer.

(a) Streamer stopping length versus applied DC voltage

(b) Streamer area versus applied DC voltage

Figure 4-10 Stopping length of streamers in the mineral oil and the synthetic ester liquid under negative


Average Propagation Velocity

Figure 4-11 shows the comparison of average streamer propagation velocity in the mineral

oil and the synthetic ester liquid at a 10 mm gap under negative polarity. At the inception

voltage levels, there is only one single pulse of the current signal that was captured for the

negative streamer. The propagation time based on the current signal is thus extremely short,

A1 B1 C1 D1

A2 B2 C2 D2

98


from about 50 ns to 200 ns. In addition, the streamer length at this stage is extremely short,

so there is uncertainty about calculating the streamer velocity. With the increase of applied

voltage, the propagation time is measured based on multiple current pulses the average

propagation velocity can be reasonably calculated based on the ratio of stopping length and

propagation time. It is found that the negative streamer velocity in synthetic ester liquid is

comparable to that in mineral oil, almost remaining constant from 0.1 km/s to 0.3 km/s,

which corresponds to the 1st mode of streamer propagation [61].

Figure 4-11 Average propagation of streamers in the mineral oil and the synthetic ester liquid under negative


Streamer Charge

Figure 4-12 shows the negative streamer stopping length as a function of maximum

apparent charge in both the mineral oil and the synthetic ester liquid. Unlike the linear

results under positive polarity, the apparent charge firstly increases with the increase of

stopping length, and then the increase rate slow down when the apparent charge above 50

pC. This result indicates that although the streamer does not propagate much further as

shown in stopping length, the discharge becomes more intensive and the charge density

within the streamer channels become higher.

99


Figure 4-12 Negative streamer stopping length as a function of maximum apparent charge in the mineral oil

and the synthetic ester liquid.

4.4 Effect of Gap Distance on Breakdown Voltage

4.4.1 Breakdown Phenomena

Figure 4-13 shows an example of the streamer leading to the breakdown in the synthetic

ester liquid under positive polarity. A dramatic voltage drop from 30 kV to 0 kV was

observed when the breakdown occurred. A short train of consecutive pulses with increasing

amplitude is commonly observed for the positive streamer. Compared with the current

signal of pre-breakdown phenomena, the current signal becomes gradually stronger with

larger amplitude. At about 8 µs, the streamer propagated to the opposite plane electrode

with one main branch with strong luminance, indicating the breakdown event, as shown in

Figure 4-13 (b).

Figure 4-14 shows the streamer leading to the breakdown in the synthetic ester liquid under

negative polarity. Similar to positive polarity, a significant voltage drop from -60 kV to 0

kV was observed when the breakdown occurred. A long current train with discrete pulses is

observed before breakdown. A few small discrete current pulses with relatively lower

amplitude existed before Frame 1 as shown in the black dashed circle A in Figure 4-14 (a).

The amplitude of these small discrete pulses is nearly equal to the current amplitude (≈ 10

mA) as shown in Figure 4-9 (a). The average propagation velocity of the breakdown event

is about 1.01 km/s, which belongs to the 2nd

mode streamer. This result indicates that the

negative streamer changes from the 1st mode streamer in Figure 4-11 to the 2

nd mode

100


streamer in Figure 4-14 at the breakdown level. Comparing Figure 4-13 (b) and Figure 4-14

(b), the negative streamer is clearer and much thicker, and it has more branches with many

small offshoots than the positive streamer.

Figure 4-13 Breakdown in the synthetic ester liquid under DC voltage, positive polarity; d = 10 mm, r = 10

µm, exposure time 2 µs; (a) voltage, current and monitor signals, (b) streamer propagation, corresponding to

the signals in (a).

Figure 4-14 Breakdown in the synthetic ester liquid under DC voltage, negative polarity; d = 10 mm, r = 10

µm, exposure time 2 µs; (a) voltage, current and monitor signals, (b) streamer propagation, corresponding to

the signals in (a).

4.4.2 Breakdown Tests at Gap Distances from 2 mm to 30 mm

To statistically analyse the breakdown voltage, Weibull distribution was used to fit the

breakdown results and to calculate 50% streamer inception voltage. Figure 4-15 shows the

0 1 2 3 4 5 6 7 8 9 10-10000

-5000

0

5000

Cu

rren

t (m

A)

0 1.4 2.8 4.2 5.6-2

0

2

4M

on

ito

r S

ign

al

(V)

Time (us)

0 1 2 3 4 5 6 7 8 9 10-1

0

1

Mo

nit

or

Sig

nal

(V)

Time (us)

0 2 4 6 8 10-10

0

10

20

Cur

rent

(mA

)

0 2 4 6 8 10-5

0

5

10

Mon

itor S

igna

l (V

)

Time (us)

0 2 4 6 8 10-50

0

50

Curre

nt (m

A)

0 2 4 6 8 10-10

0

10

Mon

itor S

igna

l (V)

Time (us)

0 2 4 6 8 10

0

2

4

Mon

itor S

igna

l (V)

Time (us)

0 5 10 15 20 25 30-1

0

1x 10

4

Cu

rren

t (m

A)

0 8 16 32

-4

-2

0

2

Mo

nit

or

Sig

nal

(V)

Time (us)

0 5 10 15 20 25 30-1

0

1

Mo

nit

or

Sig

nal

(V)

Time (us)

20 25 30 35 40-4

-2

0

2

Cur

rent

(A

)

20 25 30 35 40-5

0

5

10

Mon

itor

Sig

nal (

V)

Time (us)

20 25 30 35 40-2000

0

2000

Curre

nt (m

A)

0 10 20 30 40 50 60-10

0

10

Monit

or Sig

nal (V

)

Time (us)

20 25 30 35 40

0

2

4

Monit

or Sig

nal (V

)

Time (us)

2 µs

10 mA

30 kV



(a)

(b)

5 µs 2 A

F 1 F 2 F 3 F 4 F 5 F 6 F 7

-60 kV

(b)

Frame 1 Frame 2 Frame 3 Frame 4 Frame 5 Frame 6

(a)

101


breakdown results of the mineral oil and the synthetic ester liquid under both positive and

negative polarities, and parameters are summarised in Table 4-2. Strong polarity effect was

observed and breakdown voltages of negative polarity are nearly 2.5 times those of positive

polarity. This conclusion is valid for both the mineral oil and the synthetic ester. The reason

is, under positive polarity, the concentration of positive charges enhances the boundary

field at the head of the charge cloud, which promotes streamer propagation and then

weakens breakdown voltages. However, the diluted negative space charges, like a shielding

of the negative tip, weaken the boundary field, which slows down the streamer propagation

and then increases the negative breakdown voltage.

Figure 4-15 Weibull distribution plot of breakdown results in the mineral oil and the synthetic ester liquid

with different gap distances under DC voltage, (a) Gemini X, (b) MIDEL 7131, r = 10 µm

(a)

(b)

102


Table 4-2 Weibull parameters of breakdown results in the mineral oil and the synthetic ester liquid with

different gap distances at the point-plane electrode.

Oil Type Polarity Gap distance

(mm) Shape Scale

50 % Breakdown

Voltage

Gemini X

Positive

(+)

5 22 18.9 21.5

10 33.9 37.0 36.9

20 39.1 59.5 58.9

30 44.7 92.1 91.4

Negative

(-)

2 32.3 33.9 33.5

5 61.4 55.9 61.0

10 49.3 93.5 92.8

MIDEL

7131

Positive

(+)

5 17.4 34.0 17.2

10 33.7 35.6 25.4

20 37.9 39.0 38.6

30 44.0 61.3 60.8

Negative

(-)

2 31.6 22.2 21.9

5 37.9 59.9 37.7

10 32.1 60.7 59.8

20 33.8 87.7 86.8

The effect of gap distance on breakdown voltages under both positive and negative DC

voltages are shown in Figure 4-16. It is clear that breakdown voltages of both positive and

negative polarities almost increase linearly with the gap distance in the investigated range.

In the mineral oil, the breakdown voltages of negative polarity are about 2.5 times higher

than those under positive polarity. In the synthetic ester liquid, the breakdown voltages of

negative polarity are about 2.2 times higher than those under positive polarity.

Figure 4-16 Effect of gap distance on breakdown voltage in the mineral oil and the synthetic ester liquid

under DC voltage, r = 10 µm; based on 50% breakdown voltages

Under both positive and negative polarities, the breakdown voltages of the mineral oil are

about 1.5 times higher than those of the synthetic ester, while the initiation voltages of both

103


the liquids are similar. It is because breakdown in the non-uniform electrical field is

controlled by propagation rather than initiation.

4.5 Summary

Electrical strengths of a mineral oil and a synthetic ester liquid in a non-uniform field under

DC voltage in both positive and negative polarities were examined in this chapter. Point-

plane electrodes with the various tip radii and gap distances were used throughout the

investigation.

It can be summarised that streamer initiation voltage in the synthetic ester liquid is

comparable to that of the mineral oil (Figure 4-3). Once a streamer is initiated, streamer

propagation is easier in the synthetic ester liquid than that in the mineral oil (Figure 4-6 and

Figure 4-10). Although initiation voltages of synthetic ester liquid are comparable to those

of mineral oil, breakdown voltages of the synthetic ester liquid are nearly 40% lower than

those of the mineral oil for the investigated divergent point-plane gaps. Only slow streamer

mode was observed in both the mineral oil and the synthetic ester liquid in this study.

In addition, an extremely strong polarity effect of streamer propagation in both the liquids

is observed. At pre-breakdown stage, under negative polarity, the streamer barely

propagates (lstopping < 0.5 mm in the mineral oil and lstopping < 1 mm in the synthetic ester

liquid) even with the applied voltage close to the breakdown voltage. The negative streamer

only changes propagation mode from 1st mode to 2

nd mode at the breakdown voltage level.

104


105

Chapter 5 Streamer and Breakdown Phenomena of Insulating Liquids under Various Impulse Voltages

CHAPTER 5. STREAMER AND BREAKDOWN PHENOMENA

OF TRANSFORMER LIQUIDS UNDER DIFFERENT IMPULSE

WAVEFORMS

5.1 Introduction

This chapter reports the pre-breakdown and breakdown properties of a mineral oil and a

synthetic ester liquid at a small gap distance of 10 mm under different positive impulse

waveforms. The different impulse waveforms from the short tail-time waveform with 8 µs

to “step-like” waveform with 3200 µs were stressed on oil samples. A 16-channel high-

speed camera was used to characterise the streamer stopping length, average propagation

velocity and shape. A comparison of breakdown voltages and instantaneous breakdown

voltages between the different impulse waveforms is given. A mathematical model of

breakdown voltage prediction is described in detail and verified by tests in another liquid.

5.2 Pre-breakdown Characteristics

5.2.1 Stopping Length

Figure 5-1 shows the comparisons of streamer stopping length in the mineral oil, where 50%

breakdown voltages are stated as the reference (Vb-8 µs, Vb-14 µs, Vb-30 µs and Vb-3200 µs stand

for 50% breakdown voltage with different tail times respectively). At each voltage level,

the average and standard deviation are given based on ten measurements. The results

indicate that, streamer stopping length increases gradually with voltage applied for all the

impulse waveforms investigated. At the same voltage level, streamers under longer-tail

impulses propagate much further than those under shorter tail-time impulses.

Figure 5-2 shows the comparisons of streamer stopping lengths in the synthetic ester liquid,

which follows similar trends as presented in the mineral oil. At lower voltage levels, a very

small difference in the stopping length can be observed. However, with the increase of the

applied voltage, the streamers propagate further under a longer tail-time impulse waveform

than those under a shorter tail-time impulse waveform at the same applied voltage level.

106

Chapter 5 Streamer and Breakdown Phenomena of Insulating Liquids under Various Impulse

Voltages

Figure 5-1 Stopping length of streamers in the mineral oil under positive polarity; d = 10 mm, r = 10 µm;

error bars stand for one standard deviation.

Figure 5-2 Stopping length of streamers in the synthetic ester liquid under positive polarity; d = 10 mm, r =

10 µm; error bars stand for one standard deviation.

5.2.2 Average Propagation Velocity

In order to characterise the streamer behaviours in the mineral oil and the synthetic ester

liquid with more details, the average streamer propagation velocity of the two testing

liquids under different impulse waveforms was also calculated at a wide range of voltage

levels, i.e. from the voltage where streamer propagates in the 2nd

mode to the voltage before

the 50% breakdown voltage.

107


Figure 5-3 shows the comparisons of streamer average propagation velocities in the mineral

oil under different positive impulse waveforms. The results indicate that the average

propagation velocities remain almost constant in the range from 1.5 km/s to 1.7 km/s with

the increase of applied voltage for all the impulse waveforms investigated. At the same

applied voltage level, there is no difference in the average propagation velocities among

different tail-time impulse waveforms. In addition, it confirms that all the streamers

observed during the tests were corresponding to the 2nd

mode streamer. The similar

phenomenon was observed in the synthetic ester liquid as shown in Figure 5-4.

Figure 5-3 Average propagation velocity of streamers in the mineral oil under positive polarity; d = 10 mm, r

= 10 µm; error bars stand for one standard deviation.

Figure 5-4 Average propagation velocity of streamers in the synthetic ester liquid under positive polarity; d =

10 mm, r = 10 µm; error bars stand for one standard deviation.

108


Voltages

5.3 Breakdown Voltage

5.3.1 Breakdown Tests in the Mineral Oil

To statistically analyse the breakdown voltages, Weibull distribution was employed to fit

the breakdown results and to estimate the breakdown voltages at 50% breakdown

probability. Figure 5-5 shows Weibull distribution plots of the breakdown results obtained

in the mineral oil under different impulse waveforms.

The 50% breakdown voltages of the mineral oil under different impulse waveforms are

summarised in Table 5-1. In the following discussions of this chapter, 50% breakdown

voltage is defined as the breakdown voltage Vb of the testing liquids. It is observed that the

50% breakdown voltage decreases with the increase of impulse tail-time.

Table 5-1 Weibull parameters of breakdown results in the mineral oil obtained under different impulse

waveforms; d = 10 mm, r = 10 µm, positive polarity.

Different Impulse

Waveforms Shape Scale

50% Breakdown

Voltage (kV)

0.8/8 µs 45.48 54.76 54.1

0.8/14 µs 39.17 43.46 42.9

0.8/30 µs 41.4 34.84 34.7

0.8/3200 µs 56.85 32.22 32

Figure 5-5 Weibull plot of breakdown voltages in the mineral oil under different impulse waveforms, d = 10

mm, r = 10 µm, positive polarity.

109


Figure 5-6 shows an example of typical breakdown voltage waveforms under each type of

impulse waveform. The breakdown voltage is normally defined as the highest voltage value

of impulse waveform as shown in the grey-dashed lines A1 ~ A4. There is another term

called instantaneous breakdown voltage Vi, which is the point before voltage drop down as

shown in the black-dash circle. The time to breakdown tb is measured based on the voltage

waveform from the voltage rising point to the point before the voltage drops.

Figure 5-6 Typical breakdowns in the mineral oil under different impulse waveforms, d = 10 mm, r = 10 µm,

positive polarity.

Figure 5-7 Time to breakdown in the mineral oil under different impulse waveforms, d = 10 mm, r = 10 µm,

positive polarity.

-2 0 2 4 6 8 10 120

10

20

30

40

50

60

Time (us)

Vo

ltag

e (k

V)

0.8/8 us

0.8/14 us

0.8/30 us

0.8/3200 us

A1

Time to breakdown

A2

A3

A4

Breakdown voltage

Instantaneous

breakdown voltage

110


Voltages

Figure 5-8 Breakdown voltage and instantaneous breakdown voltages in the mineral oil under different

impulse waveforms; d = 10 mm, r = 10 µm, positive polarity.

The time to breakdown is measured based on the start point of impulse waveform and the

end of the instantaneous breakdown point, and it was found to be almost the same for

different impulse waveforms. Further detailed results are summarised in Figure 5-7. This

can be simply explained by the similar streamer average propagation velocity shown in

Figure 5-3. The peak voltage of the impulse waveform is commonly referred as the

breakdown voltage whereas instantaneous breakdown voltage Vi, represents the instant

voltage when the breakdown occurs. 50% breakdown voltages and instantaneous

breakdown voltages obtained under different impulse waveforms are shown in Figure 5-8.

Although the 50% breakdown voltage decreases with the increase of tail-time, the

instantaneous breakdown voltage remains almost the same under different impulse

waveforms. In addition, the breakdown voltage and instantaneous breakdown voltage under

the long tail impulse waveform, i.e. 0.8/3200 µs are the same due to the negligible

reduction of instant voltage during the streamer propagation.

5.3.2 Prediction of Breakdown Voltage

This behaviour in the mineral oil indicates a correlation between breakdown voltage,

instantaneous breakdown voltage and time to breakdown. In this case, a mathematical

model of predicting the breakdown voltage under desired impulse waveforms can be

established. Equation (5-1) is used to simulate the desired impulse voltage waveform [104]:

111


𝒗(𝒕) = 𝑽𝟎(𝒆−𝜶𝒕 − 𝒆−𝜷𝒕) (5-1)

where a and β are used to define impulse voltage waveform; V0 means the voltage level of

impulse waveform.

Based on the results shown in Figure 5-6, Figure 5-7 and Figure 5-8, it is assumed that Vi

and tb remain stable under impulse waveforms with different tail times. In this case, Vi and

tb can be obtained by conducting breakdown tests under one impulse waveform to predict

the breakdown voltage of another impulse waveform.

Figure 5-9 shows the flowchart of the breakdown voltage prediction model. In this model,

four parameters α, β, Vi and tb are used as the input values to process breakdown voltage

prediction. Firstly, parameter α and β are used to determine the shape of the desired impulse

waveform. For the standard lightning impulse voltage (1.2/50 µs), normally α is 0.0143 and

β is 4.87. Then, the parameters Vi and tb are determined by one set of breakdown tests under

one type of impulse waveform. The prediction is controlled by tb. The applied voltage V is

initialised at the voltage level of Vi and increased by 1% of Vi per step. Finally, when the

time to breakdown of simulated impulse waveform meets the criterion tb, the breakdown

voltage prediction program stops and outputs the breakdown voltage Vb.

In the process of breakdown voltage prediction in the mineral oil, the experimental data of

0.8/8 µs are selected as the test impulse waveform. Another three impulse waveforms

(0.8/14 µs, 0.8/30 µs and 0.8/3200 µs) are treated as the desired impulse waveforms. Table

5-2 summarises the experimental results of 50% breakdown voltages and predicted

breakdown voltages under four impulse waveforms. The results indicate that the predicted

breakdown voltages are very close to the experimental results, and the relative difference is

within 6%. The three simulated breakdown voltage waveforms are shown in Figure 5-10,

where the shapes of the simulated waveforms are identical to the impulse waveforms

obtained by experiments shown in Figure 5-6.

112


Voltages

Figure 5-9 Flowchart of a mathematical model of breakdown voltage prediction.

Table 5-2 Predicted breakdown voltage and experimental breakdown voltage in the mineral oil obtained

under different impulse waveforms; d = 10 mm, r = 10 µm, positive polarity.

Impulse Voltage

Waveforms α β

50% Breakdown

Voltage (kV)

Predicted Breakdown

Voltage (kV)

Input 0.8/8 µs 0.1100 4.87 54.1 -

Output

0.8/14 µs 0.0543 4.87 42.9 40.4

0.8/30 µs 0.0243 4.87 34.7 35.8

0.8/3200 µs 0.0013 4.87 32 31.9

Desired impulse

waveform

Determine

α and β

Experimental

breakdown test

under test impulse

waveform

Determine

Vi and tb

Estimated time to

breakdown t ≥ tb

Increasing simulated

impulse voltage V by 1%

of Vi

Outputting

breakdown voltage V

≈ Vb

YES

NO

Calculating time to

breakdown t based on Vi

113


Figure 5-10 Simulated different impulse waveforms in the mineral oil based on the mathematical model,

positive polarity.

5.3.3 Verification in the Synthetic Ester Liquid

To verify the prediction model of breakdown voltage, other experimental breakdown tests

have been conducted in the synthetic ester liquid. The test conditions including gap distance,

tip radius, electrode configuration and voltage polarity are the same. First of all, the

experiment of breakdown measurements in the synthetic ester liquid has been investigated

under test impulse waveform 0.8/3200 µs. The instantaneous breakdown voltage Vi is

calculated as 26.6 kV and the time to breakdown tb is calculated as 5.1 µs. Then, parameters

α and β are obtained for the desired impulse waveforms of 0.8/8 µs, 0.8/14 µs and 0.8/30,

as given in Table 5-2. Finally, with the input values of α, β, Vi and tb, the breakdown

voltages can be predicted as given in Table 5-3. Based on these predicted breakdown

voltages, the simulated impulse voltage waveforms under breakdown conditions are plotted

in Figure 5-11.

Table 5-3 Predicted breakdown voltage and experimental breakdown voltage in synthetic ester liquid

obtained under variable impulse voltages; d = 10 mm, r = 10 µm, positive polarity

Impulse Voltage

Waveforms

Predicted Breakdown

Voltage (kV)

50% Breakdown

Voltage (kV)

Input 0.8/3200 µs - 26.6

Output

0.8/8 µs 42.9 42.0

0.8/14 µs 32.6 34.0

0.8/30 µs 29.4 29.9

-2 0 2 4 6 8 10 120

10

20

30

40

50

60

Time (us)

Vo

ltag

e (k

V)

0.8/8 us

0.8/14 us

0.8/30 us

0.8/3200 us

O

InstantaneousBreakdownPoint

Time to breakdown

114


Voltages

Figure 5-11 Simulated different impulse waveforms in the synthetic ester liquid based on the mathematical

model, positive polarity.

Figure 5-12 Typical breakdowns in the synthetic ester liquid under different impulse waveforms, d = 10 mm,

r = 10 µm, positive polarity.

To verify the accuracy of breakdown voltage prediction, experimental breakdown

measurements have been conducted under different impulse waveforms with 0.8/8 µs,

0.8/14 µs and 0.8/30 µs. The typical breakdown waveforms in the synthetic ester liquid

under the different impulse waveforms are shown in Figure 5-12. Both the waveform

shapes and voltage levels including ‘instantaneous breakdown voltage’ and ‘time to

breakdown’ are comparable to the simulated breakdown voltage waveforms as shown in

Figure 5-11. The 50% breakdown voltages are also given in Table 5-3, which are very close

-2 0 2 4 6 8 10 120

5

10

15

20

25

30

35

40

45

Time (us)

Vo

ltag

e (k

V)

0.8/8 us

0.8/14 us

0.8/30 us

0.8/3200 us

O

InstantaneousBreakdownPoint

-2 0 2 4 6 8 10 12-5

0

5

10

15

20

25

30

35

40

45

Time (us)

Voltage (

kV

)

0.8/8 us

0.8/14 us

0.8/30 us

0.8/3200 us

Time to breakdown

Time to breakdown

115


to the predicted values and the relative difference is within 5%. Therefore, it is confirmed

that this breakdown voltage prediction model is also applicable to the synthetic ester liquid.

5.4 Effect of Impulse Waveform on Streamer Characteristics

A series of streamer images have been captured using a high-speed camera to analyse

streamer propagation and eventual breakdown. During the process of streamer propagation,

with a similar stopping length the short tail-time impulse waveform needs a higher voltage

level to force streamers to propagate as shown in Figure 5-1 and Figure 5-2. This raises the

question of the impact of impulse waveform on steamer characteristics.

Figure 5-13 and Figure 5-14 show the typical images of positive streamers with a similar

stopping length obtained in the mineral oil and synthetic ester liquid, respectively, under

different impulse waveforms. The stopping lengths of the selected streamers in the mineral

oil are about 7.44 ~ 7.85 mm and in the synthetic ester liquid, they are about 6.42 ~ 6.59

mm. In both liquids, it is found that streamers under a short tail-time impulse waveform

have more branches with more small offshoots as shown in Figure 5-13 (a) and Figure 5-14

(a). However, only one or two main branches with a few small offshoots are observed under

a long tail-time impulse waveform as shown in Figure 5-13 (d) and Figure 5-14 (d).

Figure 5-13 Pre-breakdown in the mineral oil under different impulse waveforms, d = 10 mm, r = 10 µm,

positive polarity, (a). 0.8/8 µs, V = 52 kV, lstoping = 7.62 mm (b). 0.8/14 µs, V = 42 kV, lstoping = 7.71 mm (c).

0.8/30 µs, V = 34 kV, lstoping = 7.44 mm (d). 0.8/3200 µs, V = 32 kV, lstoping = 7.85 mm.

Figure 5-14 Pre-breakdown in the synthetic ester liquid under different impulse waveforms, d = 10 mm, r =

10 µm, positive polarity, (a). 0.8/8 µs, V = 40 kV, lstoping = 6.59 mm (b). 0.8/14 µs, V = 32 kV, lstoping = 6.42

mm (c). 0.8/30 µs, V = 30 kV, lstoping = 6.47 mm (d). 0.8/3200 µs, V = 26 kV, lstoping = 6.57 mm.

(a) (b) (c) (d)

(a) (b) (c) (d)

116


Voltages

Figure 5-15 and Figure 5-16 present the streamer areas corresponding to the stopping

lengths in the mineral oil and the synthetic ester liquid under different impulse waveforms.

The streamer area is defined as the apparent area occupied by all streamer branches in the

2D image, which is estimated using a programme [105]. Similar to the photographs

observed in Figure 5-13 and Figure 5-14, positive streamers under a short tail-time impulse

waveform generally have larger areas due to more branches when compared with those

under a long tail-time impulse waveform of a similar stopping length.

Due to the rapidly decaying voltage of short tail-time impulse waveforms, streamers need

minimum voltage (e.g. Vi) and time to breakdown from higher applied voltage to maintain

the streamer propagation and achieve breakdown. However, Figure 5-3 and Figure 5-4

indicate that all the positive streamers captured in the mineral oil and the synthetic ester

liquid are the 2nd

mode streamers [44, 61]. The average propagation velocities almost

remain constant at the pre-breakdown stage. Meanwhile, the statistical analyses of the

apparent charge and energy at the similar stopping length based on ten measurements are

given in Table 5-4. This indicates that the apparent energy acting on oil samples is higher

under a short tail-time impulse waveform with a similar stopping length. In this case, the

redundant energy injected into oil samples does not compel streamers to propagate further

but helps to create more branches.

Figure 5-15 Positive streamer area as a function of stopping length in the mineral oil under different impulse

waveforms, d = 10 mm, r = 10 µm.

117


Figure 5-16 Positive streamer area as a function of stopping length in the synthetic ester liquid under different

impulse waveforms, d = 10 mm, r = 10 µm.

Table 5-4 Average charges injected into oil samples of the similar stopping length under different impulse

waveforms.

Variable Impulse

Waveforms

Charge in

Gemini X (pC)

Charge in MIDEL

7131 (pC)

0.8/8 µs 6.52 13.97

0.8/14 µs 4.43 12.07

0.8/30 µs 4.16 6.97

0.8/3200 µs 2.46 3.94

5.5 Summary

Pre-breakdown and breakdown characteristics of stopping length, average propagation

velocity, streamer shape and breakdown voltage under positive impulse waveforms with

different tail times in a mineral oil and a synthetic ester liquid with a small gap distance of

10 mm were investigated in this chapter.

At the pre-breakdown stage, it was found that with similar stopping lengths, streamers

under a short tail-time impulse waveform have denser branches with more small offshoots,

while only one or two main branches with a few small offshoots were observed under a

long tail-time impulse waveform. This is due to higher energy injection into an oil sample

under a short tail-time impulse waveform, which encourages a streamer to grow densely.

Compared to the impulse voltage with a longer tail-time, the shorter tail-time impulse

waveform results in higher breakdown voltage, but does not have an obvious effect on the

118


Voltages

instantaneous breakdown voltage and the time to breakdown. In both tested liquids, the

breakdown voltage waveforms of different impulse waveforms almost intersect one point at

the moment of voltage drop, which results in the same instantaneous breakdown voltage.

Therefore, a mathematical model for breakdown voltage prediction under impulse voltage

with different tail time has been described in detail. At the same testing environment and

liquid nature, the breakdown voltage of insulating liquids under desired impulse waveform

characterised by α, β can be estimated based on Vi and tb from one set of breakdown tests

under a known impulse waveform.

119

Chapter 6 Correlations between Gas Generation and Breakdown in Liquids

CHAPTER 6. CORRELATIONS BETWEEN GAS

GENERATION AND SPARKING FAULT IN TRANSFORMER

LIQUIDS UNDER LIGHTNING IMPULSE VOLTAGE

6.1 Introduction

In this chapter, experimental studies that investigate the correlation between fault gas

generation and sparking faults in the mineral oil and the synthetic ester liquid were carried out

by using a hydrogen monitor and a multi-gas monitor. The well-controlled electrical fault with

different test conditions of gap distances, spark numbers and voltage levels were achieved using

a compact impulse spark generator. The sparking fault energy was derived from the recording

of voltage and current waveforms, hence the correlations between fault gas generation and

sparking fault energy were obtained.

6.2 Data Processing

6.2.1 Calculation of Dissolved Gas Generation

For the experiment, a single gas on-line DGA monitor – TM1 (hydrogen monitor) was used

to continuously monitor hydrogen generation and a multi gas on-line DGA monitor – TM8

was used to measure all eight fault gases and the oil temperature.

An example of dynamic hydrogen production measured by the TM1 hydrogen monitor

during the 500-spark test is shown in Figure 6-1. The reference cycle is an internal sensor

calibration that is typically performed every twelve hours. The TM1 hydrogen monitor

makes one measurement every 30 minutes [87]. The electrical fault stress in the oil sample

started at 0 minutes and ended at 500 minutes. The dissolved hydrogen concentration in the

oil increased gradually during the process of spark fault generation. It continued to increase

for a short period after the end of fault generation, and then reached a steady state upon

attaining gas–liquid equilibrium. In order to calculate the net hydrogen generation, the

dynamic hydrogen readings were divided into three stages: background detection, spark

fault period detection and post-test detection. Background detection measured the average

hydrogen concentration before the fault tests. Post-test detection measured the average

hydrogen concentration after cessation of spark fault generation. Therefore, hydrogen

120


generation due to the spark faults is the difference between the post-test detection and the

background detection, which is 142.5 ppm in the case shown in Figure 6-1.

Figure 6-1 Hydrogen concentration as gas in oil concentration from the TM1 hydrogen monitor as a function

of time during a 500 spark test, mineral oil, d = 10 mm, positive polarity.

Figure 6-2 Hydrogen concentration as gas in oil concentration from the TM8 multi-gas monitor as a function

of time during a 500 spark test, mineral oil, d = 10 mm, positive polarity.

Figure 6-2 shows an example of hydrogen measurement before and after 500 sparks

measured by the TM8 multi-gas monitor. The TM8 multi-gas monitor was not allowed to

measure fault gas concentration during the test period of sparking fault, but allowed the

pump to continuously circulate the oil sample. At the stage of background measurement,

121


the average hydrogen concentration in the oil sample maintains at about 85.8 ppm. After

500 sparks, hydrogen concentration increases to about 219.4 ppm. Thus, the hydrogen

generation due to the 500 sparks is 133.6 ppm, which is comparable to the hydrogen

generation (142.5 ppm) measured based on the TM1 hydrogen monitor as shown in Figure

6-1. The different background measurement of hydrogen between hydrogen monitor and

multi-gas monitor is probably due to the different hydrogen sensors in these two DGA

monitors.

6.2.2 Gas-in-total Calculation

Both the TM1 hydrogen monitor and the TM8 multi-gas monitor can only measure the fault

gas concentration in oil. However, when the oil system has a headspace volume, part of the

available fault gas migrates into the headspace and affects the measured fault gas

concentration in the total oil volume. Therefore, it is necessary to calculate the gas-in-total

(GIT) concentration using the gas-in-oil (GIO) concentration. The equations for calculating

GIT are shown in equation (6-1) to (6-4) [19, 92]. The equations are based on the

conversation of mass law and the relationship between GIO concentration and gas-in-gas

(GIG) concentration [19, 92].

𝑪𝑶 = 𝑪𝑮 ∗ 𝑲 (6-1)

𝑴𝑻 = 𝑴𝑶 + 𝑴𝑮 (6-2)

𝑪𝑻 ∗ 𝑽𝑶 = 𝑪𝑶 ∗ 𝑽𝑶 + 𝑪𝑮 ∗ 𝑽𝑮 (6-3)

𝑪𝑻 = 𝑪𝑶(𝟏 +𝟏

𝑲∗

𝑽𝑮

𝑽𝑶) (6-4)

where CO, CG and CT mean the gas concentration of oil, gas and total respectively; MO, MG

and MT mean the mass of gas dissolved in the oil, in the gas and the total mass respectively;

VO and VG mean the volume of oil and gas respectively; K means the solubility coefficient

of the gas.

The ineradicable headspace volume in the entire test system including the online DGA

monitors was estimated as 80 mL. The value of the K factor was based on the Ostwald

122


solubility coefficient, which is affected by oil temperature. For example, the Ostwald

solubility coefficient under different oil temperatures in mineral oil is shown in Figure 6-3.

The oil temperature measured by the TM8 multi-gas monitors was 21±2 . In addition, the

Ostwald solubility coefficients used in the synthetic ester liquid are provided by the oil

supplier, which have not been pulished.

Figure 6-3 The K factor based on Ostwald solubility coefficient under different temperature in mineral oil

The fault gas generation in ppm represents the ratio between total gas volume and oil

volume. It is more reasonable to correlate total volume of fault gas generation with fault

energy. The equation for converting fault gas concentration in ppm to generated fault gas

volume in μL is shown in equation (6-5).

𝑻𝒐𝒕𝒂𝒍 𝑮𝒂𝒔 𝑽𝒐𝒍𝒖𝒎𝒆 = 𝑪𝑻 × 𝑽𝑶 (6-5)

where CT means the gas concentration in total; VO means the oil volume.

6.2.3 Energy Calculation

The lightning impulse voltage is measured by a voltage divider and the fault current is

measured by a current shunt. Equations (6-6) and (6-7) show the product of the

instantaneous voltage V and current I resulting in the instantaneous fault energy. To obtain

a more quantitative comparison of fault gas generation in the mineral oil and the synthetic

ester liquid, the concentrations of fault gases were correlated with the sparking fault energy

E (joule) by integrating power (VI) with time.

123


𝑬 = ∫ 𝑽𝑰 𝒅𝒕 (6-6)

𝑬 = 𝑽𝟏𝑰𝟏∆𝒕 + 𝑽𝟐𝑰𝟐∆𝒕 + ⋯ ⋯ + 𝑽𝒏𝑰𝒏∆𝒕 (6-7)

Figure 6-4 shows an example of the voltage and the current waveform of a spark in the

mineral oil at 39 kV (99.9% breakdown voltage). The voltage and current waveform can be

divided into three parts. Part 1 is pre-breakdown stage, which interprets a process of the

streamer initiating from needle electrode and propagating to plane electrode with time

duration of 7 μs. Part 2 is a sparking fault stage. The voltage drops sharply and current

initially rushes to a peak and then decay gradually. The sparking arc lasts for about 25 μs

and then distinguished. Afterwards, the current becomes near zero and voltage recovers

slightly and slowly decays to zero, which is called Part 3. The reason is probably that the

breakdown conductive channel disappeared at about 42 μs, and the dielectric strength of the

oil sample was recovered, which leads to drop of the current.

Figure 6-4 The voltage and current waveform in mineral oil on 99.9% breakdown voltage, positive polarity, d

= 10 mm, r = 50 µm.

0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

Time (us)

Volt

age

(kV

)

0 10 20 30 40 50 60 70 80 90 100

0

2

4

6

8

Time (us)

Cu

rren

t (A

)

Part 1 Part 2 Part 3

124


Table 6-1 shows the average energy generation as a percentage for each part of voltage and

current waveform based on 100 sparks in the mineral oil. The pre-breakdown energy in part

1 only accounts for 1.45% at 99.9% breakdown voltage and 1.56% at 150% breakdown

voltage. Most of the energy is accumulated in part 2, since the sparking arc occurred at this

stage. The energy in part 3 occupied 7.98% on 99.9% breakdown voltage and 7.97% on 150%

breakdown voltage, which might be due to residual discharge or current signal distortion.

Overall the calculation is representative for sparking fault energy.

Table 6-1 Individual part energy as percentages of total energy generation based on 100 sparks in mineral oil,

V = Vb-99.9%.

Mean Value Standard Deviation

Vb-99.9%

Part 1 1.45 % 0.3114

Part 2 90.56 % 0.3631

Part 3 7.98 % 0.2063

Vb-150%

Part 1 1.56 % 0.3493

Part 2 90.48 % 0.6094

Part 3 7.97 % 0.2730

6.3 DGA Results and Analysis

6.3.1 Comparison of Hydrogen Measurements

Hydrogen is the lightest fault gas and has the highest mass transfer coefficient, which

means it is easier to escape from the oil compared to other fault gases. Therefore

comparison of hydrogen measurements between the TM8 multi-gas monitor, TM1

hydrogen monitor and laboratory gas chromatography (GC) system was carried out first.

When the TM1 hydrogen and TM8 multi-gas monitors had finished hydrogen

measurements after each group of spark tests, the oil was sampled with gas-tight syringes

for laboratory DGA measurement, where the fault gases were extracted by the headspace

method and analysed by GC system. Table 6-2 compares the hydrogen measurements by

the TM1 hydrogen monitor, the TM8 multi-gas monitor and the laboratory technique.

The results indicate that the hydrogen measured by the multi-gas monitor is comparable to

that measured by the hydrogen monitor and laboratory GC systems. The high relative

differences in sample 1 and sample 2 are due to the low concentration of hydrogen

measured. However the absolute differences were within a few parts-per-million, which are

negligible in terms of practical implication. When the hydrogen concentration is greater

than about 15 ppm, the reading from hydrogen monitor TM1 generally has higher values,

125


and the relative difference was within 20%. When the hydrogen is less than 15 ppm, the

relative difference was within a few parts-per-million. In the following, measurements from

the TM8 multi-gas monitor are used in the analyses.

Table 6-2 Comparison of hydrogen measurements among the hydrogen monitor, the multi-gas monitor and

laboratory technique.

Oil Sample

Hydrogen Concentration (ppm)

100%(TM1 -

TM8) /TM1

100%(TM1 - DGA)

/TM1 TM 8

Multi-gas

Monitor

TM1

Hydrogen

Monitor

Laboratory

DGA

Sample 1 3 3 5 0% - 67%

Sample 2 8 8 11 0% - 37%

Sample 3 15 18 16 17% 11%

Sample 4 52 57 48 9% 16%

Sample 5 108 135 109 20% 19%

Sample 6 211 227 192 7% 15%

6.3.2 Effect of Spark Numbers

To observe the occurrence of the spark in the liquids, a high-speed camera was used to

capture the image of the electrodes. Figure 6-5 shows a typical positive spark in the mineral

oil at the 5 mm gap distance under standard lightning impulse. It takes about 3 µs from the

streamer initiation to breakdown.

Figure 6-5 Typical spark in the mineral oil of positive polarity under lightning impulse, VB-99.9%-positive = 31 kV,

d = 5 mm, exposure time 0.5 µs.

Based on the equations (6-1) to (6-4), all the DGA results measured in GIO by TM8 on-line

DGA monitors were converted to GIT. The fault gas generation (GIT) based on the average

of two-group results in the mineral oil under positive sparking fault with the gap distance of

5 mm is shown in Figure 6-6. It is clear that fault gas generation increased gradually with

the number of sparks. When spark numbers were smaller than 50, only C2H2 was measured

with a small amount of 1.2 ppm. This phenomenon can be explained that a few sparks

generated very few other fault gases, which do not reach the measurement threshold of the

126


multi-gas monitor. When the spark number was increased to 100, a small amount of H2 was

measured with a value of 2.5 ppm. When the spark numbers were increased up to 500, H2

was generated with a value of 7.1 ppm and C2H2 of 26 ppm. It is also important to note that

very low amount of C2H4 normally created at the high energy or temperature was measured

[80]. The small amount of C2H4 was generated probably due to local high temperature

within the sparking arc. In addition, CH4, CO, and C2H6 were not measured during the tests.

Figure 6-6 Fault gas generation (GIT) in the mineral oil at different numbers of breakdowns, positive polarity,

d = 5 mm.

Figure 6-7 Fault gas generation (GIT) in the synthetic ester liquid at different numbers of breakdowns,

positive polarity, d = 5 mm.

127


Figure 6-7 shows the fault gas generation (GIT) in the synthetic ester liquid under positive

sparking fault with the gap distance of 5 mm. It is clear that fault gas generation increased

gradually with the number of sparks. When the spark numbers were smaller than 100, there

was no fault gas measured, due to the measurement threshold. When the spark number was

increased to 500, H2 with a value of 3.3 ppm, CO with a value of 8.8 ppm and C2H2 with a

value of 4.8 ppm were measured.

Figure 6-8 shows the comparison of H2 and C2H2 generation (GIT) in the mineral oil and

the synthetic ester liquid under sparking fault with different numbers of sparks. It is clear

that H2 and C2H2 generations increase linearly with the number of sparks. C2H2 generation

is higher than H2 generation, which implies the sparking fault at this small gap distance

tends to have high energy intensity. In addition, a number of fault gases in the mineral oil is

higher than that in the synthetic ester liquid, in particular for C2H2.

Figure 6-8 Comparison of hydrogen and acetylene generation (GIT) in the mineral oil and the synthetic ester

liquid under sparking fault at different numbers of sparks, positive polarity, d = 5 mm.

Figure 6-9 shows the individual fault gases as percentages of total fault gases in the mineral

oil and the synthetic ester liquid under positive sparking fault. Clearly key gases in the

mineral oil under sparking fault are C2H2 and H2 which accounted for more than 95% of the

overall gas concentration as shown in Figure 6-9 (a). However, H2, CO and C2H2 which

accounted for 20%, 30% and 50% respectively, are key gases in the synthetic ester liquid

under positive sparking fault as shown in Figure 6-9 (b). With the increase of breakdown

128


numbers, the fault gas patterns remain almost the same, which means there is no

accumulative effect induced secondary reactions during the sequential tests

200 300 500

(a). Gemini X – Positive

200 300 500

(b). MIDEL 7131 – Positive

Figure 6-9 Individual fault gases as percentages of total fault gases in the mineral oil and the synthetic ester

liquid with a different number of sparks, d = 10 mm.

6.3.3 Effect of Gap Distance

Normally, the larger gap distance results in higher breakdown voltage and longer

discharge/sparking channel. To investigate the effects of these factors on fault gas

generation, similar experiments were carried out at the gap distance of 10 mm.

Figure 6-10 shows the fault gas generation (GIT) in the mineral oil with the number of

sparks under positive sparking fault at the 10 mm gap distance. The fault gas generation

increased gradually with the number of sparks. After 20 sparks applied, a small amount of

H2 and C2H2 were measured with the values of 5.7 ppm and 5.2 ppm, respectively. When

the spark number was increased to 500, a large amount of H2 and C2H2 were measured with

the values of 191.4 ppm and 143.7 ppm, and a small amount of CH4, CO and C2H4 were

measured with the value less than 10 ppm.

Figure 6-11 shows the fault gas generation (GIT) in the synthetic ester liquid with the

129


number of sparks under positive sparking fault at the 10 mm gap distance. After 20

breakdowns applied, a small amount of H2, CO and C2H2 were measured with the values of

5.1 ppm, 3.2 ppm and 5.4 ppm, respectively. When the breakdown number was increased to

500, a large amount of H2, CO and C2H2 were measured with the values of 146.9 ppm, 65.2

ppm and 99.4 ppm, and a small amount of CH4 and C2H4 were generated with the values

less than 9 ppm.

Figure 6-10 Fault gas generation (GIT) in the mineral oil as a function of the number of breakdowns, Gemini

X - Vb-99.9% = 39 kV, positive polarity, d = 10 mm.

Figure 6-11 Fault gas generation (GIT) in the synthetic ester liquid as a function of the number of

breakdowns, MIDEL 7131 - Vb-99.9% = 37 kV, positive polarity, d = 10 mm.

130


Figure 6-12 shows the comparison of H2 and C2H2 in the mineral oil and the synthetic ester

liquid at 5 and 10 mm gap distance under positive sparking fault. It is clear that fault gas

generation in both liquids at the 10 mm gap distance is much higher than that at the 5 mm

gap distance. This can be explained that the longer discharge/sparking channels with high

total fault energy at the 10 mm gap distance induce much more fault gas generation

compared to the case of 5 mm gap distance. The amount of H2 generation is found to be

slightly higher than C2H2 generation, which implies the sparking fault at this large gap

distance tends to have low energy intensity. Similar to the results at the 5 mm gap distance,

fault gas generation in the mineral oil at the 10 mm gap distance is relatively higher than

that in the synthetic ester liquid.

Figure 6-12 Comparison of hydrogen and acetylene generation (GIT) in the mineral oil and the synthetic ester

liquid at different gap distances under positive polarity.

131


200 300 500

(a). Gemini X - Positive

200 300 500

(a). MIDEL 7131 - Positive


liquid with different spark numbers, Gemini X - Vb-99.9% = 39 kV, MIDEL 7131 - Vb-99.9% = 37 kV, positive

polarity, d = 10 mm.

Figure 6-13 shows the individual fault gases as percentages of total fault gases in the

mineral oil and the synthetic ester liquid with the spark numbers of 200, 300 and 500. It is

clear that H2 and C2H2 are both key gases in the mineral oil and the synthetic ester liquid. In

addition, it should be noted that CO is an additional key gas in the synthetic ester liquid

which accounts for around 20%. With the increase of spark numbers, the percentage

patterns of fault gas generation almost remain constant.

In previous studies which investigated the dissolved fault gases in the mineral oil and the

ester liquid, the volume of C2H2 is usually larger than that of H2, which is possibly due to

the lower solubility of H2 in the mineral oil and the ester liquid than that of C2H2. The

amounts of CH4, C2H4, and C2H6 were almost negligible: the energy dissipated from the

sparking fault was insufficient to generate those gases.

6.3.4 Effect of Voltage Levels

To investigate the effect of the higher voltage level on the fault gas generation, a similar

experiment was carried out at 1.5 times 99% breakdown voltage at the 10 mm gap distance.

132


Figure 6-15 shows fault gas generation (GIT) in the mineral oil at two different voltage

levels with 200 sparks.

Figure 6-14 Fault gas generation in the mineral oil at different voltage levels after 200 sparks, Vb-99.9% = 39

kV, 1.5Vb-99.9% = 59 kV, d = 10 mm.

Figure 6-15 Fault gas generation in the synthetic ester liquid at different voltage levels after 200 sparks, Vb-99.9%

= 37 kV, 1.5Vb-99.9% = 56 kV, d = 10 mm.

At the same number of sparks, higher breakdown voltage leads to more fault gas generation

including H2, C2H2 and C2H4. The concentration of H2 under 1.5Vb-99.9% is about 1.70 times

as large as that under Vb-99.9% and C2H2 is about 1.29 times as large as that under Vb-99.9%.

133


Figure 6-15 shows fault gas generation (GIT) in the synthetic ester liquid at different

voltage levels with 200 sparks. Similar to the phenomena in the mineral oil, higher

breakdown voltage results in more fault gas generation in the synthetic ester liquid

including H2, C2H2, C2H4 and CO. The concentration of H2 under 1.5Vb-99.9% is about 1.74

times as large as that under Vb-99.9%, 1.21 times for C2H2, 1.29 times for CO, and 1.19 times

for C2H4.

Figure 6-16 shows the individual fault gases as percentages of total fault gases in the

mineral oil and the synthetic ester liquid at different voltage levels with 200 sparks. It is

clear that H2 and C2H2 are both key gases in the mineral oil and the synthetic ester liquid at

both the tested voltage levels. CO is also an additional key gas in the synthetic ester liquid

which accounts for around 20%. With the increase of voltage levels, the percentage patterns

of fault gas generation almost remain constant.

99.9% VB 1.5 times 99.9% VB

(a). Gemini X – 200 sparks

99.9% VB 1.5 times 99.9% VB

(b). MIDEL 7131 – 200 sparks


liquid at different voltage levels with 200 sparks, d = 10 mm.

6.3.5 Correlation between Fault Gas Generation and Fault Energy

The fault gas generation is variable with different number of sparks, gap distances, voltage

134


levels and liquid natures. This raises the need to understand the correlation of fault energy

and fault gas generation.

The previous finding [106] indicated that energy is needed for the breaking of the chemical

bonds which comes from the energy of sparking faults. Figure 6-17 shows the statistical

analysis of fault energy for each spark of one group test with the total spark number of 1170.

It is clear that the fault energy for each spark almost remains constant in both the liquids.

The average energy per spark is about 0.121 J in the mineral oil and 0.111 J in the synthetic

ester liquid.

Figure 6-17 Statistical analysis of fault energy for each spark in the mineral oil and the synthetic ester liquid

at the 10 mm gap distance, totally 1170 sparks.

Figure 6-18 Average energy per spark in the mineral oil and the synthetic ester liquid under different test

conditions.

135


Figure 6-18 shows the average energy per spark in the mineral oil and the synthetic ester

liquid under different test conditions of gap distance, voltage level, and liquid nature. In

terms of different gap distances, larger gap distance results in higher fault energy, which

thus leads to more fault gas generation at the 10 mm gap distance than that at the 5 mm gap

distance. In terms of different voltage levels, higher voltage level also results in higher fault

energy, which thus leads to more fault gas generation at 1.5 times 99.9% VB than that at

99.9% VB. For the liquid nature, higher fault energy in the mineral oil than that in the

synthetic ester liquid was observed at the same test condition, which explains why sparking

fault in the mineral oil generated more fault gases than that in the synthetic ester liquid.

With the known fault energy of each breakdown, the fault gas volumes per unit fault energy

were calculated. Figure 6-19 shows the fault gas volumes per unit fault energy (μL/J) of the

mineral oil and the synthetic ester liquid at the 5 and 10 mm gap distance with different

number of sparks and voltage levels. When the applied spark numbers are between 20 and

100, the results are unstable due to the statistic uncertainty of measurement of low

concentration fault gases. When the applied spark numbers are equal to or above 200, the

gas generation rates almost remain stable. The results indicate that although more sparks

generate more fault gases, the fault gas volumes per unit fault energy almost remain

constant in both liquids. Therefore, in the investigated range of spark numbers, the different

number of sparks does not have an obvious effect on fault gas volumes per unit fault

energy.

Figure 6-19 Fault gas volumes per unit fault energy (μL/J) of the mineral oil and the synthetic ester liquid at

the 5 and 10 mm gap distance with a different number of sparks.

136


Based on the results from 200 sparks onward, fault gas volumes per unit fault energy in

μL/J of the mineral oil and the synthetic ester liquid at different spark numbers, gap

distances and voltage levels are summarised in Table 6-3. At the 10 mm gap distance, fault

gas generation rates at the high voltage level are almost identical to those at the low voltage

level. Fault gas generation rate of H2 and C2H2 in the mineral oil are also similar to those of

the synthetic ester liquid. In addition, it is clear that fault gas generation rates at the 10 mm

gap distance are higher than those at the 5 mm gap distance. It is possibly due to the

different discharge/sparking channel that results in different contact area of the spark in the

oil sample. Also, the limitation of fault energy calculation gives the difficulty of result

analysis, which needs further study of fault energy analysis.

Table 6-3 Fault gas volumes per unit fault energy (μL/J) of the mineral oil and the synthetic ester liquid at

different spark numbers, gap distances and voltage levels.

Gap Distance Voltage Level Gas Average Fault Gas Volumes per Unit Fault Energy (µL/J)

Gemini X MIDEL 7131

10 mm

99.9%

Breakdown

Voltage

H2 7.6812 6.2842

C2H2 6.3843 5.0476

CO - 3.2026

1.5 times

99.9%

Breakdown

Voltage

H2 7.7613 7.8992

C2H2 5.1288 4.7324

CO - 3.2106

5 mm

99.9%

Breakdown

Voltage

H2 0.6251 3.3651

C2H2 3.2435 1.0805

CO - 2.0967

6.4 Summary

This chapter discusses the gassing behaviour of sparking fault in the mineral oil and the

synthetic ester liquid. Two on-line DGA monitors: TM1 hydrogen monitor and TM8 multi-

gas monitor were used to measure dissolved gases in liquids after electrical faults. The three

main topics are a) effect of spark number (from 20 to 500); b) effect of gap distance (5 mm

and 10 mm); c) effect of voltage level (99.9% VB and 1.5 times 99.9% VB).

At the investigated range of sparking faults, the key gases in the mineral oil are H2 and

C2H2, while the key gases in the synthetic ester liquid are H2, C2H2 and CO. The amount of

fault gas generation increases linearly with the number of sparks. However, the number of

sparks does not have an obvious effect on fault gas generation per unit fault energy in µL/J.

At the same gap distance of 10 mm but with a higher applied breakdown voltage, more

fault gases were generated due to the higher injected fault energy, whereas the fault gas

137


generation per unit fault energy in µL/J remained stable. Fault gas generation rates at the 10

mm gap distance are higher than those at the 5 mm gap distance, which is possible due to

difficult underlying mechanisms. At the same test condition, sparking fault in the mineral

oil has more fault gas generation than that in the synthetic ester liquid.

138


139

Chapter 7 Conclusions and Future Work

CHAPTER 7. CONCLUSIONS AND FUTURE WORK

7.1 Conclusions

7.1.1 General

This thesis focuses on pre-breakdown and breakdown performance and gassing behaviour

of the mineral oil and the synthetic ester liquid under DC and impulse voltages by

considering the effects of tip radius, gap distance, voltage waveform, voltage polarity,

liquid nature and different electrical fault levels. Through experimental investigations and

data analyses, the research objectives have been achieved and thus some useful conclusions

and findings have been made.

Research topics covered in this thesis are:

Streamer phenomenon and breakdown properties of transformer liquids under DC

voltages

Effect of tip radius on streamer initiation voltage

Streamer characteristics including current, stopping length, velocity, charge and

shape

Effect of gap distance on breakdown voltage

Streamer characteristics and breakdown phenomena of transformer liquids under

different impulse waveforms

Effect of different impulse waveforms on streamer characteristics

Prediction of breakdown voltage using a mathematical model

Correlations between gas generation and sparking fault in transformer liquids under

lightning impulse voltage

Effect of spark number on fault gas generation

Effect of gap distance on fault gas generation

Effect of voltage level on fault gas generation

140


7.1.2 Summary of Results and Main Findings

In this thesis, the streamer characteristics and breakdown strength of the mineral oil and the

synthetic ester liquid under both positive and negative DC voltages under a divergent field

were studied. Streamer inception voltages with the tip radii of 5 µm, 10 µm, 20 µm and 50

µm and breakdown voltages at various gaps of 2 mm, 5 mm, 10 mm, 20 mm and 30 mm

were tested. Streamer initiation voltages increase with tip radius under both positive and

negative polarities. Streamer initiation voltage of the synthetic ester liquids is slightly lower

than that of the mineral oil. The stopping length of positive streamers gradually increases

with the applied voltage. However, the negative streamer remains in the 1st mode and

barely propagates with the increase of applied voltage, which is different from that was

observed in previous studies under impulse voltage [42, 107]. The apparent charge is

correlated to the streamer length. The correlations between streamer stopping lengths and

apparent charges increase linearly in both the liquids under positive polarity, while a 3-

stage power-law relationship is shown under negative polarity. In the investigated range,

there is no obvious difference in streamer shape between the mineral oil and the synthetic

ester liquid and hence the correlations between streamer stopping lengths and apparent

charges are identical in both liquids. At the same applied voltage level, the streamer in the

synthetic ester liquid propagates faster and further than that in the mineral oil. As a result,

the breakdown voltages of the synthetic ester liquid are lower than those of the mineral oil

at all the gap distances investigated under both polarities.

Pre-breakdown and breakdown characteristics of stopping length, average propagation

velocity, streamer shape and breakdown voltage under positive impulse waveforms with

different tail times (0.8/8 μs, 0.8/14 μs, 0.8/30 μs and 0.8/3200 μs) were investigated in the

mineral oil and the synthetic ester liquid at a gap distance of 10 mm. Compared to the

impulse waveform with longer tail times, the shorter tail-time impulse waveform results in

higher breakdown voltage, but does not have an obvious effect on instantaneous breakdown

voltage and time to breakdown. A mathematical model for breakdown voltage prediction

under impulse waveforms with different tail time has been described. At the same testing

environment and liquid nature, the breakdown voltage of both the mineral oil and the

synthetic ester liquid under any desired impulse waveform characterised by α, β can be

predicted based on Vi and tb from one set of breakdown tests under a known impulse

waveform. At the pre-breakdown stage, it was found that with the similar stopping length,

streamers under a short tail-time impulse waveform have denser branches, while only one

or two main branches are observed under a long tail-time impulse waveform. This is due to

141


the higher energy injected into oil sample under a short tail-time impulse waveform, which

encourages the streamer to grow densely.

Fault gas generation in the mineral oil and the synthetic ester liquid was investigated under

various levels of electrical stresses: different spark numbers (from 20 to 500), gap distances

(5 mm and 10 mm) and voltage levels (99.9% VB and 1.5 times 99.9% VB). The key gases in

the mineral oil are H2 and C2H2, while the key gases in the synthetic ester liquid are H2,

C2H2 and CO. The amount of fault gas generation increases linearly with the number of

sparks. However, the number of sparks does not have an obvious effect on fault gas

generation per unit fault energy in µL/J. Spark at a large gap distance has higher fault

energy, which results in more fault gas generation. At the same gap distance but with a

higher applied breakdown voltage, more fault gases were generated due to the higher

injected fault energy, whereas the fault gas generation per unit fault energy in µL/J

remained stable. At the same test condition, sparking fault in the mineral oil has more fault

gas generation than that in the synthetic ester liquid.

7.2 Future Work

In this thesis, some useful conclusions can be made about the dielectric performance and

gassing behaviour of the mineral oil and the synthetic ester liquid under DC and impulse

voltage. However, new questions have arisen and more research work could be done in the

future.

For investigations of streamer phenomena under DC voltage:

i. The work in this thesis focused mainly on streamer propagation and breakdown under

DC voltage. It might also be worth studying the streamers characteristics in mineral oil

and ester liquids under DC bias AC voltages.

ii. The work in this thesis focused mainly on streamer and breakdown study at the room

temperature. However, the oil temperature in transformers is practically higher than

room temperature. It is worth carrying similar experimental study at different oil

temperature.

iii. The results showed that the positive streamer gradually propagates with the applied

voltage, whereas negative streamers barely propagate with applied voltage in both

liquids. This kind of strong polarity effect was not observed in previous streamer

142


studies under AC and impulse voltage. Therefore, it might be worth carrying out

further experiments (e.g. gap distances, hydrostatic pressure) to investigate the

mechanism of negative streamers under DC voltage.

For investigations of streamer and breakdown phenomena under impulse voltage:

i. A mathematical model was developed to predict the breakdown voltage in the mineral

oil and the synthetic ester liquid based on limited experiments. It might be worth

applying this model to larger gap distance and other insulating liquids, e.g. natural ester

and Gas-to-Liquid (GTL).

ii. The work in this thesis focused mainly on positive polarity under impulse voltage. It

might also be worth investigating the relationship between breakdown voltage,

instantaneous breakdown voltage and time to breakdown under negative polarity.

For investigations of gassing behaviour under electrical faults:

i. This part of the study mainly focused on gassing behaviour in liquids under sparking

fault. Due to the insufficient energy of discharge fault generated by the current test

platform, an auto-controlled fault generator with higher voltage level (> 100 kV) is

expected to build up. It is worth investigating the gassing behaviour in liquids under

discharge faults.

ii. The work in this thesis focused mainly on fault gas generation under positive sparking

fault. As the previous finding indicated that the positive discharge relies on electronic

processes and the negative discharge depends more on gaseous processes. It is

interesting to find out the polarity effect on fault gas generation.

143

References

REFERENCES

[1] S. Chakravorti, D. Dey and B. Chatterjee, "Recent trends in the condition

monitoring of transformers", Power Systems Springer-Verlag: London, UK, 2013.

[2] IEC 60060-1,"High-Voltage Test Techniques—Part 1: General Definitions and Test

Requirements", 2010.

[3] IEC 60076-3,"Power Transformers-Part 3: Insulation levels, dielectric tests and

external clearances", 2000.

[4] R. Liu and C. Törnkvist, "Ester fluids as alternative for mineral oil: The difference

in streamer velocity and LI breakdown voltage", IEEE International Conference on

Electrical Insulation and Dielectric Phenomena, Virginia Beach, VA, USA, pp. 543-

548, 2009.

[5] C. T. Duy, O. Lesaint and A. Denat, "Streamer propagation and breakdown in rape-

seed oil at high voltage", IEEE International Conference on Dielectric Liquids,

Futuroscope-Chasseneuil, France , pp. 1-4, 2008.

[6] V.-H. Dang, A. Beroual and C. Perrier, "Comparative study of streamer phenomena

in mineral, synthetic and natural ester oils under lightning impulse voltage", IEEE

International Conference on High Voltage Engineering and Application (ICHVE),

New Orleans, LA, USA, pp. 560-563, 2010.

[7] L. Lewand, "Laboratory evaluation of several synthetic and agricultural-based

dielectric liquids", Doble International Client Conference, 2001.

[8] D. P. Stockton, J. R. Bland and T. McClanahan, "Natural ester transformer fluids:

safety, reliability & environmental performance", Petroleum and Chemical Industry

Technical Conference, Calgary, Alta., Canada, pp. 1-7, 2007.

[9] MIDEL. "First Midel 400kV Transformers for National Grid following Test

Project," Retrieved 06 August, 2017, from http://www.midel.com/news/first-midel-

400kv-transformers-for-national-grid-following-test-project.

[10] M. I. Materials. "Midel 7131 insulated 238kV power transformer," Retrieved

26.April, 2017; http://www.midel.com/applications/power.

[11] Q. Liu, "Electrical performance of ester liquids under impulse voltage for

application in power transformers", PhD Thesis, The University of Manchester,

2011.

[12] O. Lesaint, "Prebreakdown phenomena in liquids and their relation with breakdown

properties in high voltage applications", CIGRE A2-D1, 2007.

[13] O. Lesaint, "Streamers” in liquids: Relation with practical high voltage insulation

and testing of liquids", IEEE International Conference on Dielectric Liquids,

Futuroscope-Chasseneuil, France, pp. 1-6, 2008.

[14] S. Okabe, "Evaluation of breakdown characteristics of oil-immersed transformers

under non-standard lightning impulse waveforms-definition of non-standard

lightning impulse waveforms and insulation characteristics for waveforms including

pulses", IEEE Transactions on Dielectrics and Electrical Insulation, vol. 14, no. 1,

pp.147-155, 2007.

[15] S. Venkatesan and S. Usa, "Impulse strength of transformer insulation with

nonstandard waveshapes", IEEE Transactions on Power Delivery, vol. 22, no. 4, pp.

2214-2221, 2007.

[16] K. Bhuyan and S. Chatterjee, "Study of effects of standard and non-standard

impulse waves on power transformer", Joint International Conference on Power

Electronics, New Delhi, India, pp. 1-4, 2010.

http://www.midel.com/news/first-midel-400kv-transformers-for-national-grid-following-test-project

http://www.midel.com/news/first-midel-400kv-transformers-for-national-grid-following-test-project

http://www.midel.com/applications/power

144

References

[17] A. Lindroth, "The relationship between test and service stresses as a function of

resistivity ratio for HVDC converter transformer and smoothing reactors",

ELECTRA-CIGRE, pp. 33-33, 1994.

[18] R. Sarathi, P. D. Singh and M. G. Danikas, “Characterization of partial discharges

in transformer oil insulation under AC and DC voltage using acoustic emission

technique", Journal of Electrical Engineering Bratislava, vol. 58, no. 2, pp. 91, 2007.

[19] S. Tenbohlen, D. Vukovic and M. Jovalekic, "Dielectric performance and dissolved

gas analysis of natural esters for application in power transformers", CIGRE SC D1,

2009.

[20] Y. Wang, J. Li and Y. Zhao"Failure evaluation model of oil-paper insulation under

AC-DC combined voltages", IEEE International Conference on High Voltage

Engineering and Application (ICHVE), Shanghai, China, pp. 396-399, 2012.

[21] S. J. Blanksby and G. B. Ellison, "Bond dissociation energies of organic molecules",

Accounts of chemical research, vol. 36, no. 4, pp. 255-263, 2003.

[22] Q. Liu, Z. Wang and O. Lesaint, "Comparison of streamer propagation in mineral

oils under lightning and step impulse voltages", International Conference on

Dielectric Liquids (ICDL), Bled, Slovenia, pp. 1-4, 2014.

[23] A. Denat, O. Lesaint and F. Mc Cluskey, “Breakdown of liquids in long gaps:

influence of distance, impulse shape, liquid nature, and interpretation of

measurements,” IEEE Transactions on Dielectrics and Electrical Insulation, vol. 22,

no. 5, pp. 2581-2591, 2015.

[24] K. Stricklett, C. Fenimore and E. Kelly, "Observation of partial discharge in hexane

under high magnification", IEEE transactions on electrical insulation, vol. 26, no. 4,

pp. 692-698, 1991.

[25] J. Xiang, Q. Liu and Z. Wang, "Inception and breakdown voltages of insulating

liquids under DC stress", IEEE International Conference on High Voltage

Engineering and Application (ICHVE), Chengdu, China, pp. 1-4, 2016.

[26] B. Lehouidj, S. Sahra and A. Nacer, "Streamer currents, emitted lights and energies

into dielectric liquids submitted to DC stress", 18th International Conference on

Dielectric Liquids (ICDL), Bled, Slovenia, pp. 1-4, 2014.

[27] L. Shiyan, Z. Tao and L. Yunpeng, "Detections of ultrasonic sensor and PMT on

partial discharge in oil under DC voltage", International Conference on Dielectric

Liquids (ICDL), Bled, Slovenia, pp. 1-5, 2014.

[28] R. Sarathi, A. Reid and M. Judd, "Partial discharge study in transformer oil due to

particle movement under DC voltage using the UHF technique", Electric Power

Systems Research, vol. 78, no. 11, pp. 1819-1825, 2008.

[29] E. Kelley, M. Nehmadi and R. Hebner, "Measurement of partial discharges in

Hexane under dc voltage", IEEE transactions on electrical insulation, vol. 24, no. 6,

pp. 1109-1119, 1989.

[30] IEC 60243-1, "Methods of Test for Electric Strength of Solid Insulating Materials",

1998.

[31] A. I. Mohamed, R. Ozaki and K. Kadowaki, "Streak Observation of Dc Pre-

breakdown Phenomenon in Silicone Oil/Low Density Polyethylene (LDPE) with

Different Viscosities", International Journal on Electrical Engineering and

Informatics, vol. 5, no. 1, pp. 95-104, 2013.

[32] R. Sarathi and G. Koperundevi, "Investigation of partial discharge activity of single

conducting particle in transformer oil under DC voltages using UHF technique",

IET science, measurement & technology, vol. 3, no. 5, pp. 325-333, 2009.

[33] Y. Du, Y. Lv and C. Li, “Effect of semiconductive nanoparticles on insulating

performances of transformer oil", IEEE Transactions on Dielectrics and Electrical

Insulation, vol. 19, no. 3, 2012.

145

References

[34] Y. Wang, G. Li and Y. Yin, "The effect of nano-MGO addition on grounded DC

tree in cross-linked polyethylene", Conference on Properties and Applications of

Dielectric Materials (ICPADM), Sydney, NSW, Australia, pp. 285-288, 2015.

[35] W. Hauschild and W. Mosch, "Statistical techniques for high-voltage engineering:

IET", Stevenage, U.K.: Peregrinus, 1992.

[36] Q. Liu, Z. Wang and F. Perrot, "Impulse breakdown voltages of ester-based

transformer oils determined by using different test methods", IEEE Conference on

Electrical Insulation and Dielectric Phenomena, Virginia Beach, VA, USA, pp. 608-

612, 2009.

[37] M. Pompili, C. Mazzetti and R. Bartnikas, "Early stages of negative PD

development in dielectric liquids", IEEE Transactions on Dielectrics and Electrical

Insulation, vol. 2, no. 4, pp. 602-613, 1995.

[38] Q. Liu and Z. Wang, "Streamer characteristic and breakdown in synthetic and

natural ester transformer liquids with pressboard interface under lightning impulse

voltage", IEEE Transactions on Dielectrics and Electrical Insulation, vol. 18, no. 6,

2011.

[39] A. Denat, N. Bonifaci and M. Nur, “Spectral analysis of the light emitted by

streamers in hydrocarbon liquids,” IEEE Transactions on Dielectrics and Electrical

Insulation, vol. 5, no. 3, pp. 382-387, 1998.

[40] G. Massala and O. Lesaint, “A comparison of negative and positive streamers in

mineral oil at large gaps,” Journal of Physics D: Applied Physics, vol. 34, no. 10, pp.

1525-1532, 2001.

[41] R. Badent, K. Kist and A. Schwab, "Voltage dependence of prebreakdown

phenomena in insulating oil." IEEE International Symposium on Electrical

Insulation, Pittsburgh, PA, USA, USA, pp. 414-417, 1994.

[42] Q. Liu and Z. Wang, “Streamer characteristic and breakdown in synthetic and

natural ester transformer liquids under standard lightning impulse voltage,” IEEE

Transactions on Dielectrics and Electrical Insulation, vol. 18, no. 1, 2011.

[43] D. Linhjell, L. Lundgaard and G. Berg, “Streamer propagation under impulse

voltage in long point-plane oil gaps,” IEEE Transactions on Dielectrics and

Electrical Insulation, vol. 1, no. 3, pp. 447-458, 1994.

[44] O. Lesaint and G. Massala, “Positive streamer propagation in large oil gaps:

experimental characterization of propagation modes,” IEEE Transactions on

Dielectrics and Electrical Insulation, vol. 5, no. 3, pp. 360-370, 1998.

[45] M. D. Cevallos, M. Butcher and J. Dickens, “Imaging of negative polarity dc

breakdown streamer expansion in transformer oil due to variations in background

pressure,” IEEE Transactions on Plasma Science, vol. 33, no. 2, pp. 494-495, 2005.

[46] M. Butcher, A. A. Neuber and M. D. Cevallos, “Conduction and breakdown

mechanismsin transformer oil,” IEEE Transactions on Plasma Science, vol. 34, no.

2, pp. 467-475, 2006.

[47] N. Dung, H. Høidalen and D. Linhjell, “Effects of reduced pressure and additives on

streamers in white oil in long point-plane gap,” Journal of Physics D: Applied

Physics, vol. 46, no. 25, pp. 255501, 2013.

[48] M. Nagao, S. Jayaram and M. Sugio, "Breakdown characteristics of oil-polymer

composite insulation under AC and DC voltages." IEEE 13th International

Conference on Dielectric Liquids, Nara, Japan, pp. 140-143, 1999.

[49] C. Tang, G. Chen and M. Fu, "Effect of applied DC voltages and temperatures on

space charge behaviour of multi-layer oil-paper insulation." Journal of Physics D:

Applied Physics, pp. 012006, 2009.

146

References

[50] C. Tang, G. Chen and M. Fu, “Space charge behavior in multi-layer oil-paper

insulation under different DC voltages and temperatures,” IEEE Transactions on

Dielectrics and Electrical Insulation, vol. 17, no. 3, pp.775-784, 2010.

[51] C. Tang, R. Liao and G. Chen, “Research on the feature extraction of DC space

charge behavior of oil-paper insulation,” Science China Technological Sciences, vol.

54, no. 5, pp. 1315-1324, 2011.

[52] J. Hao, G. Chen and R. Liao, “Influence of moisture on space charge dynamics in

multilayer oil-paper insulation,” IEEE Transactions on Dielectrics and Electrical


[53] Q. Liu and Z. Wang, “Secondary reverse streamer observed in an ester insulating

liquid under negative impulse voltage,” Journal of Physics D: Applied Physics, vol.

44, no. 40, pp. 405203, 2011.

[54] M. Unge, S. Singha and N. Van Dung, “Enhancements in the lightning impulse

breakdown characteristics of natural ester dielectric liquids,” Applied Physics

Letters, vol. 102, no. 17, pp. 172905, 2013.

[55] A. Beroual and R. Tobazeon, “Prebreakdown phenomena in liquid dielectrics,”

IEEE transactions on electrical insulation, no. 4, pp. 613-627, 1986.

[56] A. Beroual, M. Zahn and A. Badent, “Propagation and structure of streamers in

liquid dielectrics,” IEEE Electrical Insulation Magazine, vol. 14, no. 2, pp. 6-17,

1998.

[57] C. T. Duy, O. Lesaint and A. Denat, “Streamer propagation and breakdown in

natural ester at high voltage,” IEEE Transactions on Dielectrics and Electrical


[58] A. Beroual, C. Marteau and R. Tobazeon, “Behavior of streamers in liquids under

step voltages in point-plane geometry,” IEEE transactions on electrical insulation,

vol. 23, no. 6, pp. 955-959, 1988.

[59] D. Linhjell, S. Ingebrigtsen and L. Lundgaard, "Streamers in long point-plane gaps

in cyclohexane with and without additives under step voltage", IEEE International

Conference on Dielectric Liquids (ICDL), Trondheim, Norway, pp. 1-5, 2011.

[60] P. Rain and O. Lesaint, “Prebreakdown phenomena in mineral oil under step and ac

voltage in large-gap divergent fields,” IEEE Transactions on Dielectrics and

Electrical Insulation, vol. 1, no. 4, pp. 692-701, 1994.

[61] O. Lesaint, “Prebreakdown phenomena in liquids: propagation ‘modes’ and basic

physical properties,” Journal of Physics D: Applied Physics, vol. 49, no. 14, pp.

144001, 2016.

[62] V. Komelkov, “Development of a pulse discharge in liquids,” Soviet Physics-

Technical Physics, vol. 6, no. 8, pp. 691-699, 1962.

[63] T. Liao and J. Anderson, “Propagation mechanism of impulse corona and

breakdown in oil,” Transactions of the American Institute of Electrical Engineers,

Part I: Communication and Electronics, vol. 72, no. 5, pp. 641-648, 1953.

[64] I. Stekolnikov and V. Y. Ushakov, “Discharge phenomena in liquids,” Sov. Phys.-

Technical Phys, vol. 10, no. 9, pp. 1307-1313, 1966.

[65] B. Farazmand, “Study of electric breakdown of liquid dielectrics using Schlieren

optical techniques,” British Journal of Applied Physics, vol. 12, no. 5, pp. 251-254,

1961.

[66] R. Bartnikas, "Electrical insulating liquids", ASTM Press, Philadelphia, pp. 262-307,

1994.

[67] A. Sharbaugh, J. Devins and S. Rzad, “Progress in the field of electric breakdown in

dielectric liquids,” IEEE transactions on electrical insulation, no. 4, pp. 249-276,

1978.

147

References

[68] P. K. Watson, W. Chadband and M. Sadeghzadeh-Araghi, “The role of electrostatic

and hydrodynamic forces in the negative-point breakdown of liquid dielectrics,”

IEEE transactions on electrical insulation, vol. 26, no. 4, pp. 543-559, 1991.

[69] O. Lesaint and P. Gournay, “On the gaseous nature of positive filamentary

streamers in hydrocarbon liquids. I: Influence of the hydrostatic pressure on the

propagation,” Journal of Physics D: Applied Physics, vol. 27, no. 10, pp. 2111-2116,

1994.

[70] R. E. Hebner, "Measurement of electrical breakdown in liquids," The liquid state

and its electrical properties, pp. 519-537: Springer, 1988.

[71] P. Gournay and O. Lesaint, “A study of the inception of positive streamers in

cyclohexane and pentane,” Journal of Physics D: Applied Physics, vol. 26, no. 11,

pp. 1966-1974, 1993.

[72] W. Lu and Q. Liu, “Pre-breakdown and Breakdown Mechanisms of an Inhibited

Gas to Liquid Hydrocarbon Transformer Oil under Positive Lightning Impulse

Voltage”, IEEE transactions on electrical insulation, vol. 23, no. 4, pp. 2450-2461,

2016.

[73] S. J. Rzad, J. C. Devins and R. J. Schwabe, “Transient Behavior in Transformer Oils:

Prebreakdown and Breakdown Phenomena,” IEEE transactions on electrical

insulation, vol. EI-14, no. 6, pp. 289-296, 1979.

[75] Qualitrol Company, "Severon TM8 on-line DGA monitor Product Brochure," 2015.

[76] IEC 60599, "Mineral oil-impregnated electrical equipment in service-Guide to the

interpretation of dissolved and free gases analysis," May, 2007.

[78] T. Committee, “IEEE Guide for the Interpretation of Gases Generated in Oil-

Immersed Transformers,” Institute of Electrical & Electronics Engineers, NY, 1992.

[79] IEC 60567, “Oil-filled electrical equipment-Sampling of gases and of oil for

analysis of free and dissolved gases–Guidance”," International Electrotechnical

Commission, Geneva, Switzerland, 2005.

[80] M. Duval, “Dissolved gas analysis: It can save your transformer,” IEEE Electrical

Insulation Magazine, vol. 5, no. 6, pp. 22-27, 1989.

[81] IEC 60475, "Method of sampling insulating liquids,” 2011.

[82] W. Chen, Y. YUN and C. PAN, “Gas-in-oil Analysis in Transformers Based on

Photoacoustic Spectrometry,” Automation of electric Power systems, vol. 15, pp.

940-946, 2007.

[83] H. C. Sun, Y. C. Huang and C. M. Huang, “A review of dissolved gas analysis in

power transformers,” Energy Procedia, vol. 14, pp. 1220-1225, 2012.

[84] S. Li, “Study of Dissolved Gas Analysis under Electrical and Thermal Stresses for

Natural Esters used in Power Transformers,” Mphil Thesis, The Univeristy of

Manchester, 2012.

[85] M. Duval, “The duval triangle for load tap changers, non-mineral oils and low

temperature faults in transformers,” IEEE Electrical Insulation Magazine, vol. 24,

no.6, pp.22-29, 2008.

[86] J. Dai, I. Khan and Z. Wang, "Comparison of HYDRAN and laboratory DGA

results for electric faults in ester transformer fluids." Conference on Electrical

Insulation and Dielectric Phenomena, Vancouver, BC, Canada, pp. 731-734, 2007.

[87] Qualitrol Company, “On-Line Transformer Monitor Operations and Maintenance

Guide—Model TM1,” 2014.

[88] U. Astrom, B. Westman and V. Lescale, "Power transmission with HVDC at

voltages above 600 kV." Power Engineering Society Inaugural Conference and

Exposition in Africa, Durban, South Africa, pp. 44-50, 2005.

[89] M.-h. Gao and Y. Leng, “Two direct components in converter transformer,”

Transformer, vol. 39, no. 6, pp. 15-18, 2002.

148

References

[90] M. Jovalekic, D. Vukovic and S. Tenbohlen, "Dissolved gas analysis of alternative

dielectric fluids under thermal and electrical stress", IEEE International Conference

on Dielectric Liquids (ICDL), Trondheim, Norway, pp. 1-4, 2011.

[91] J. Ganesha, "Dissolved Gas Analysis of Transformer Oils: Effects of electric arc",

Lisbon, Portugal: International Conference, September 22-24, 2006.

[92] Z. Wang, X. Wang and X. Yi, “Gas generation in natural ester and mineral oil under

partial discharge and sparking faults,” IEEE Electrical Insulation Magazine, vol. 29,

no. 5, pp. 62-70, 2013.

[93] C. Perrier, M. Marugan and A. Beroual, “DGA comparison between ester and

mineral oils,” IEEE Transactions on Dielectrics and Electrical Insulation, vol. 19, no.

5, 2012.

[94] N. A. Muhamad, B. Phung and T. Blackburn, “Dissolved gas analysis for common

transformer faults in soy seed-based oil,” IET electric power applications, vol. 5, no.

1, pp. 133-142, 2011.

[95] K. Bandara, C. Ekanayake and T. K. Saha, "Comparative study for understanding

the behaviour of natural ester with mineral oil as a transformer insulating liquid",

International Conference on Electrical Insulation and Dielectric Phenomena

(CEIDP), Des Moines, IA, USA pp. 792-795, 2014.

[96] H. Wilhelm, C. Santos and G. Stocco, “Dissolved gas analysis (DGA) of natural

ester insulating fluids with different chemical compositions,” IEEE Transactions on

Dielectrics and Electrical Insulation, vol. 21, no. 3, pp. 1071-1078, 2014.

[97] S. Ab Ghani and N. Muhamad, "Hydrocarbon gases dissolved in palm-based

insulation oil after experienced breakdown." Conference on Properties and

Applications of Dielectric Materials (ICPADM), Sydney, NSW, Australia, pp. 460-

463, 2015.

[98] R. Eberhardt, B. Wieser and W. Lick, "Dissolved gas analysis investigations on

ester liquids after breakdown", IEEE International Conference on Dielectric Liquids

(ICDL), Trondheim, Norway, pp. 1-4, 2011.

[99] I. Khan, Z. Wang and J. Dai, "Fault gas generation in ester based transformer fluids

and dissolved gas analysis (DGA)", International Conference on Condition

Monitoring and Diagnosis, Beijing, China, pp. 909-913, 2008.

[100] J. Aubin, “Effect of particles on transformer dielectric strength,” Cigré WG12, vol.

17, 2000.

[101] Y. Khan, H. Al-Falih and Y. Zhang, “Two-step controllable electrochemical etching

of tungsten scanning probe microscopy tips,” Review of Scientific Instruments, vol.

83, no. 6, pp. 063708, 2012.

[102] IEC 60897, "Methods for the determination of the lightning impulse breakdown

voltage of insulating liquids", 1987.

[103] J. Xiang and Q. Liu, “Current and emitted light characteristics of streamers in

insulating liquids under ac voltages,” The 19th International Symposium on High

Voltage Engineering (ISH), pp. 295, 2015.

[104] J. R. Lucas, “High voltage engineering,” Sry Lanka, 2001.

[105] Z. Liu, Q. Liu and Z. Wang, “Effect of electric field configuration on streamer and

partial discharge phenomena in a hydrocarbon insulating liquid under AC stress,”

Journal of Physics D: Applied Physics, vol. 49, no. 18, pp. 185501, 2016.

[106] R. Pandey and M. Deshpande, “Dissolved Gas Analysis (DGA) of Mineral Oil Used

in Transformer,” International Journal of Application or Innovation in Engineering

and Management, vol. 1, no.2, pp.208-211, 2012.

[107] O. Lesaint and R. Tobazeon, “Streamer generation and propagation in transformer

oil under AC divergent field conditions,” IEEE transactions on electrical insulation,

vol. 23, no. 6, pp. 941-954, 1988.

149

Appendix I List of Publications

APPENDIX I LIST OF PUBLICATIONS

Journal Papers:

[1]. J. Xiang, X.Y. Zhou, Q. Liu, Z.D. Wang, J. Hinshaw and P. Mavrommatis,

“Correlation between Hydrogen Generation and Electrical Faults in a Mineral

Transformer Oil”, Electrical Insulation Magazine, IEEE, 2017. (Under 2nd

Review)

[2]. J. Xiang, Q. Liu and Z.D. Wang, “Streamer Characteristic and Breakdown in a

Mineral Oil and a Synthetic Ester Liquid under DC Voltage”, IEEE Transaction on

Dielectric and Electrical Insulation. (Submitted)

[3]. Q. Liu, J. Xiang, Z.D. Wang and O. Lesaint, “Prediction of Breakdown Voltage of

Insulating Liquids under Different Impulse Waveforms”, IEEE Transaction on

Dielectric and Electrical Insulation. (To be Submitted)

International Conference Papers:

[4]. J. Xiang, Q. Liu and Z.D. Wang, “Current and emitted light characteristics of

streamers in insulating liquids under ac voltages”, The 19th International Symposium

on High Voltage Engineering (ISH), Pilsen, Czech Republic, pp. 295, 23-28 August,

2015.

[5]. J. Xiang, Q. Liu and Z.D. Wang, “Inception and breakdown voltages of insulating

liquids under DC stress”. High Voltage Engineering and Application (ICHVE),

Chengdu, China, pp. 1-4, September, 2016.

Documents

PRE-BREAKDOWN AND BREAKDOWN STUDY OF TRANSFORMER …