Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Surface charge accumulation on spacers under switching impulses in sulphur hexafluoride gas Cherukupalli, Sudhakar Ellapragada 1987

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1987_A1 C47_8.pdf [ 10.87MB ]
Metadata
JSON: 831-1.0097073.json
JSON-LD: 831-1.0097073-ld.json
RDF/XML (Pretty): 831-1.0097073-rdf.xml
RDF/JSON: 831-1.0097073-rdf.json
Turtle: 831-1.0097073-turtle.txt
N-Triples: 831-1.0097073-rdf-ntriples.txt
Original Record: 831-1.0097073-source.json
Full Text
831-1.0097073-fulltext.txt
Citation
831-1.0097073.ris

Full Text

Surface Charge Accumulation on Spacers Under Switching Impulses in Sulphur Hexafluoride Gas by S u d h a k a r E l l a p r a g a d a C h e r u k u p a l l i B . E . , B.M.S. College of Engineering, Bangalore, INDIA, 1974 M . E . , Indian Institute of Science, Bangalore, 1976 A THESIS S U B M I T T E D IN PARTIAL F U L F I L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF D O C T O R O F PHILOSOPHY in T H E F A C U L T Y OF G R A D U A T E STUDIES Department of Electrical Engineering We accept this thesis as conforming to the required standard T H E UNIVERSITY O F BRITISH C O L U M B I A July, 1987 © Sudhakar Ellapragada Cherukupalli, 1987 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 Date /SOCT 1$2>? Abstract Metal encapsulation with compressed sulphur hexafluoride gas insulation has re-duced the size of power substations and improved their reliability. Due to the superior insulating properties of this gas, its application in Gas Insulated Trans-mission lines (GITL) and Gas Insulated Substations (GIS) is increasing. Such systems invariably require solid support spacers for supporting high voltage con-ductors and for compartmentalizing sections of the systems. It has been found that, although the dielectric strength of sulphur hexafluoride is high compared to other gases, in the presence of a solid spacer, the dielectric integrity of the sys-tem is impaired. For DC GIS and GITL, free of any particulate contamination, anomalous fiashover during a polarity reversal of the applied voltage has been at-tributed to charge accumulation on the spacer surface. The present work examines the effect of switching impulse voltages on the charge accumulation on support spacers in sulphur hexafluoride gas and the effect of AC precharging on impulse fiashover. The charge accumulation on the spacer has been measured using a ca-pacitive probe. A three-dimensional surface charge simulation technique has been developed to convert the probe potential measurements to an equivalent charge distribution. Electric field variation on the spacer surface due to this computed charge can be then obtained with this program. The results indicate that spacers acquire charges even under impulse voltage application in non-uniform field gaps. These charges affect the prebreakdown phe-nomenon and the breakdown behavior of spacer gaps. Under AC voltages, spacers in uniform fields do not acquire charges as has been reported by previous inves-tigators. Under non-uniform field conditions however, AC precharging of spacers i i iii does affect the impulse fiashover. The variation however, seems to fall within the statistical scatter of pure impulse fiashover. Contents A B S T R A C T i i L I S T O F T A B L E S vi1 L I S T O F F I G U R E S vm A C K N O W L E D G E M E N T S x v 1 I N T R O D U C T I O N 1 2 B R E A K D O W N B E H A V I O R O F S P A C E R S I N G I S 10 2.1 Spacer Material 10 2.2 Behavior of a Spacer in the Presence of Contamination 13 2.2.1 Free Metallic Particles , 13 2.2.2 Fixed Metallic Particles 16 2.2.3 Non conducting Particles 21 2.2.4 Water Vapor 22 2.2.5 Decomposed Gases 22 2.3 Spacer Profiles 22 2.3.1 Post Spacers 22 2.3.2 Conical Spacers 25 2.3.3 Disc Spacers 26 2.3.4 Multiblade Spacers 27 2.4 Effect of Voltage Waveshape 27 2.5 Models for Surface Charge Accumulation 31 iv V 2.5.1 DC Voltages 31 2.5.2 Impulse and AC Voltages 36 2.6 Surface Discharge Initiation and Propagation 37 2.7 Present Work 3 9 3 T E C H N I Q U E S F O R S U R F A C E C H A R G E M E A S U R E M E N T 44 3.1 Introduction 44 3.2 Electro Optic Methods 44 3.2.1 The Kerr Effect 45 3.2.2 The Pockel Effect 45 3.3 Electric Field Mill 46 3.4 Capacitance Probes 48 3.4.1 Design of the Capacitance Probes 52 3.5 Dust Figure Technique 58 4 E X P E R I M E N T A L S E T U P 60 4.1 Introduction 60 4.2 Pressure Vessels 60 4.3 Electrode Assembly 66 4.4 Electrode Drive Assembly 66 4.5 Capacitance Probe Assembly 68 4.6 Power Supplies 75 4.7 Insulating Spacers 75 4.7.1 Selection of a method to neutralize surface charges 77 5 E L E C T R I C F I E L D C O M P U T A T I O N 81 5.1 Introduction 81 5.2 Principle of Charge Simulation Method 83 vi 5.2.1 CSM Applied to Two Dielectric Arrangement 85 5.3 Principle of Surface Charge Simulation 85 5.4 Present Problem 89 6 R E S U L T S A N D D I S C U S S I O N S 96 6.1 Introduction 96 6.2 Pre-breakdown phenomenon 97 6.2.1 Pre - breakdown currents 98 6.3 Probability Distribution of Corona Inception Time 103 6.4 Von-Laue Plots 103 6.4.1 Discussion 106 6.5 Surface Potential Measurements 119 6.5.1 Experimental procedure 119 6.6 Results with switching impulses 120 6.6.1 Results with lightning impulses 142 6.6.2 Discussion 146 6.6.3 Mechanism of charging 153 6.7 Effect of AC Pre-Charging 158 6.8 Breakdown Behavior of the Spacer Gap 166 6.8.1 Plane-Parallel Electrodes 166 6.8.2 Rod-plane Electrodes 167 7 C O N C L U S I O N S A N D S U G G E S T I O N S F O R F U T U R E W O R K 1 7 2 7.1 Pre-breakdown phenomena 172 7.2 Spacer Charge Accumulation 173 7.3 Scope for future work 175 B I B L I O G R A P H Y 177 List o f Tables 2.1 Effects of various kinds of particles on the fiashover voltage of prac-tical spacers at 310 kPa SF 6 [17] vii List of Figures 1.1 Section through a typical GIS [2] 3 1.2 Various cast-epoxy spacer designs [3] 5 1.3 Effect on spacer fiashover in uniform field of a gap at an electrode interface or in middle of spacer[8] 8 2.1 DC fiashover voltages for four different types of post spacers with 6.4 mm wire particle in SF 6 [18] 15 2.2 Effect of amount of fine copper powder (30//m) on the AC breakdown voltages of a coaxial system with cone type spacer [17] 17 2.3 Lift-off and crossing voltages for free particles and breakdown volt-ages for free and fixed particles in a SF 6 filled 76/250 mm coaxial system with post type spacer[43] 18 2.4 Influence of the length of copper wire on the fiashover voltage of a disc type spacer [30] 19 2.5 Arrangement of a cylindrical spacer between plane parallel elec-trodes, 1 - Electrodes, 2 - Insulating spacer. Gap d\ is exaggerated to expose the triple junction 24 2.6 Residual fields on post spacers after test at +600kV DC stress [38]. . 28 2.7 The uniform field fiashover voltages of a 10 mm cylindrical spacer under AC, DC, and Impulse voltages [39] 30 2.8 Residual potential distribution on both untreated and roughened spacers and theoretical calculation of charge accumulation. Gap spacing between metal inserts is 20mm [50] 34 viii ix 2.9 Radial and tangential field components of post spacer in a coaxial bus and dust figure of spacer energized with negative DC of 100kV for 5 hours [50] 35 2.10 A toroid test chamber for surface fiashover studies[54] 40 3.1 A schematic diagram of the field mill 47 3.2 A schematic diagram of a capacitance probe showing its different capacitances 49 3.3 Cross sectional view of the various probes 53 3.4 Experimental setup to determine probe resolution 55 3.5 Variation of probe sensitivity with probe to surface spacing for var-ious probes 56 4.1 Test arrangement at the BC Hydro Laboratory 61 4.2 Schematic of the gas cart 63 4.3 Arrangement of the electrode drive and probe in the high pressure test chamber 64 4.4 Test arrangement at the UBC laboratory 65 4.5 Electrode drive arrangement in the small pressure vessel 67 4.6 Electrode drive arrangement in the high pressure vessel 69 4.7 Cross-sectional view of the electrode drive used in the high pressure vessel 70 4.8 Cross-section of the probe cable 72 4.9 Levels at which the residual charge measurements were made. Dis-tance x is the measure of the charge scan radius from the axis along the spacer surface 73 4.10 Schematic of the charge recording setup 74 4.11 Photographs of the insulating spacers. . .' 76 X 4.12 Experimental setup to establish a suitable residual charge neutral-izing technique 78 4.13 Effect of using various techniques to neutralize residual charge on a small disc spacer 80 5.1 Representation of conical and spherical surface charges. . 88 5.2 Charge and contour point location for the divergent field geometry. . 90 5.3 Surface field distribution on unenergised insulating spacer 94 5.4 Surface field distribution on spacer following impulse voltage appli-cation 95 6.1 Surface electric field distribution in a rod-plane gap with an cylin-drical insulating spacer. In case of a gas gap the field calculation is along the imaginary boundary of the spacer 99 6.2 Current and voltage oscillogram (a) lightning impulse voltage (+52kV) and current (48.2mA):(b) switching impulse displacement current and corona current (rise time 1.5/us) current:(c) same as in (b) but with spacer (rise time 0.9/xs) 100 6.3 Variation of the mean corona pulse current with applied voltage for different materials 102 6.4 Variation of the mean corona inception time with applied voltage for different materials 104 6.5 Computed log-normal distributions of the corona pulses with and without epoxy spacer at different voltages with 5mm rod 105 6.6 Von-Laue plots for different materials at +94kV (Rod electrode di-ameter: 5mm) 107 6.7 Von-Laue plots for acrylic spacer at different voltages (Rod electrode diameter: 5mm) 108 6.8 Von-Laue plots for epoxy spacer at different voltages (Rod electrode diameter: 5mm) 109 xi 6.9 Von-Laue plots for epoxy spacer at +77kV (Rod electrode diameter: 5mm and 10mm) 110 6.10 Von-Laue plots for 5/xm epoxy spacer at +108kV for 2 different pressures. (Rod electrode diameter: 6mm) I l l 6.11 Von-Laue plots for 5/im epoxy spacer at +124kV for 2 different pressures. (Rod electrode diameter: 6mm) 112 6.12 Surface potential variation for epoxy spacer placed between plane-parallel electrodes prior to voltage application 122 6.13 Surface potential variation for acrylic spacer placed between rod-plane electrodes prior to voltage application 123 6.14 Surface potential variation for epoxy spacer placed between plane-parallel electrodes after subjecting it to positive switching impulses upto 190kV 124 6.15 Surface potential variation for 5/zm surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) OkV b) +48kV 127 6.16 Surface potential variation for 5/j,m surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) +66kV b) +77kV (on the spacer top) c) +77kV (entire spacer) . 128 6.17 Surface potential variation for 5p,m surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) OkV b) +48kV 129 6.18 Surface potential variation for a 5/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diam-eter 5mm a) +63kV b) +77kV 130 6.19 Surface potential variation for a 20/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diam-eter 5mm a) +77kV (top surface only) b) +77kV (entire spacer) . . 131 xi i 6.20 Surface potential variation for 20/xm surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) OkV b) +48kV 133 6.21 Surface potential variation for 20/im surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) +62kV b) +62kV (entire spacer) c) +77kV (entire spacer) . . . . 134 6.22 Surface potential variation for a 5/zm surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diam-eter 10mm a) OkV b) +48kV 135 6.23 Surface potential variation for 5/xm surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +77kV b) +116kV (insulator top) c) +116kV (entire spacer) . . 136 6.24 Surface potential variation for 4/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) OkV b) +49kV 137 6.25 Surface potential variation for 4/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +63kV b) +77kV i 138 6.26 Surface potential variation for 20/zm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diam-eter 10mm a) +77kV (top surface) b) -f 77kV (entire spacer) . . . . 139 6.27 Surface potential variation for 20/zm surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) OkV b) +49kV 140 6.28 Surface potential variation for 20/zm surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +77kV b) +95kV 141 6.29 Shot-to-shot variation spacer placed between rod-plane electrodes with repeated positive switching impulse application at 62kV [rod diameter 5mm]: a) Shot 1 b) Shot 2 . 142 xiii 6.30 Surface potential variation for a epoxy spacer placed between rod-plane electrodes following positive lightning impulse, [rod diameter 10mm]: a) 62kV b) 77kV c) 95kV 144 6.31 Surface potential variation for a 20/j,m acrylic spacer placed between rod-plane electrodes following negative 62kV lightning impulse, [rod diameter 10mm] 145 6.32 Shot-to-shot variation spacer placed between rod-plane electrodes with repeated negative lightning impulse application at 62kV [rod diameter 5mm]: a) Shot 1 b) Shot 2 147 6.33 Surface potential distribution on an acrylic spacer placed between a 3.2mm rod-plane gap in air at O.lMPa subjected to positive DC voltages for 20 minutes, a) lOkV b)20kV c)30kV d)Maller et al., results[82] 150 6.34 Surface potential distribution on a ptfe spacer placed under a 10 mm rod-plane gap in air at O.lMPa subjected to negative DC voltages for 20 minutes, a) 20kV b)40kV 154 6.35 Surface potential distribution on a ptfe spacer placed under a 10 mm rod-plane gap in N2 at O.lMPa subjected to negative DC voltages for 20 minutes, a) OkV b)20kV c)40kV d) +40kV 155 6.36 Surface potential distribution on a ptfe spacer placed under a 10 mm rod-plane gap in SF6 at O.lMPa subjected to negative DC voltages for 20 minutes, a) 20kV b)55kV c)80kV 156 6.37 a) Negative ion density distribution along a rod-plane gap for vari-ous DC voltages; b) Negative ion density in a rod-plane gap at var-ious times during a negative impulse application. Voltage=124kV, p=0.1MPa and T=293K[84] 159 6.38 Surface potential variation for a spacer placed between plane-parallel electrodes subjected to AC voltage of 70kVrms for 20 minutes . . . . 161 6.39 Surface potential variation for a spacer placed between rod-plane electrodes subjected to AC pre-charging, rod diameter 5mm . . . . . 162 xiv 6.40 Surface potential variation for a spacer placed between rod-plane electrodes subjected to AC pre-charging, rod diameter 10mm a) 30kV 5min b) 30kV 30min c) 30kV 60min d)43kV 30min 163 6.41 Shot-to-shot variation for a spacer between rod-plane electrodes with no AC pre-charging (a) positive impulse fiashover voltage (b) fiashover time 168 6.42 Shot-to-shot variation for a spacer between rod-plane electrodes, with AC pre-charging at 50kV for 30 minutes (a) positive impulse fiashover voltage (b) fiashover time 169 X V Acknowledgements I wish to express my gratitude to my parents, who inspired me to undertake this venture. I would like to thank all the staff in Electrical Engineering workshop for all the help rendered during the fabrication of the various components necessary for this investigation. My thanks to Mr. A. Reed and Dr. J.B. Neilson of BC Hy-dro Research Laboratory for providing me the facility to undertake a part of this investigation. I am grateful to my friend Subroto for spending the long hours developing the Surface Charge Simulation program. My appreciation to Dr. S.R. Naidu for offering the many suggestions during my thesis writing. My special thanks to my office-mates Nick and William for the lively and inspiring debates during the coffee breaks. I am thankful to my colleagues for making my stay at UBC memorable one. The financial assistance of the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged. I wish to express my gratitude to my teacher and supervisor, Prof. K.D.Srivastava, who despite his very busy schedule always found time to help me out with the prob-lems I faced during this work. On the purely personal side, I thank my wife, Kamala, for the patience and forbearance during my doctoral program. Dedicated TO My PAJUNTS Geeta V-16 r x v i Chapter 1 INTRODUCTION Sulphur Hexafluoride (SF6 ) gas is a non-toxic, non-flammable gas. Its dielectric strength is substantially greater than that of traditional dielectric gases like Nitro-gen and. Helium. Its dielectric strength is attributed to the electron attachment property of the SF 6 molecules. By forming relatively heavy and sluggish negative ions, higher electric fields are required to cause cumulative ionization in the gas leading to a sparkover, as compared with the case when only free electrons are present. Further, its molecules when dissociated due to sparkovers at high electric fields, recombine rapidly. Thus, when the source of spark energy is removed, the gas recovers its strength quite rapidly. This makes it uniquely effective in quench-ing of high energy arcs. As a result, it has found widespread application in modern high power circuit breakers. Metal encapsulation with compressed SF 6 gas insulation has reduced the size of power substations and improved their reliability over conventional air insulated substations. Due to the superior insulating properties of compressed sulphur hex-afluoride gas, its application in Gas Insulated Substations (GIS) and Gas Insulated Transmission Lines (GITL) is increasing. Both GIS and GITL provide inherent compactness, improved aesthetics and simultaneously provide immunity to pol-luted environments as compared to conventional equipment both at extra high voltage (EHV) and ultra high voltages (UHV) used for transmission. High Voltage AC (HVAC) compressed gas insulated systems are now under operation world-wide at voltages ranging from 145 kV to 550 kV [l]. More recently, there have been pro-totype installations at voltage levels of 800 kV and 1200 kV which are under test. 1 2 Most of these systems have been of the isolated phase rigid bus design. The cen-tral coaxial conductor in these designs is supported by cast epoxy spacers located in an aluminum enclosure. This enclosure can either be an extrusion or a spiral welded aluminum pipe. Typical gas pressures at which these systems operate lie between 0.4 and 0.5 MPa. Recent research has been directed to building both 3 phase enclosed systems and semi- flexible systems. This concerted effort in devel-oping GITL has been motivated by the ability to build sealed inert systems with low losses. In comparison to conventional air insulated lines, GITL offer improved reliability, reduced fire hazards and improved aesthetics. High voltage direct current (HVDC) transmission is becoming a more significant factor in power systems because of its advantages of system isolation, stability and efficient power transfer over long distances. Experimental installations of DC GITL have been used to explore the possibility of their use for bulk power transmission. The use of DC GITL would also mitigate the pollution problems associated with conventional air insulated DC transmission lines. From the design point of view a DC GITL is identical in construction to an AC system. Typically, it is made in modules, 18m in length, and shipped to site for assembly. Although DC GITL construction is similar to AC GITL, they differ in me-chanical and electrical considerations in their design. For example, in a DC GITL there are no hysteresis and eddy current losses. Consequently, the outer conductor can be made of plain steel. This results in a reduction in the overall system costs (both fabrication and material). On the other hand, however, with an aluminum enclosure, for the DC GITL, power can still be delivered using ground as return in the event of a pole fault. Electrically a DC GITL is subjected to different types of stresses as compared to AC GITL. Particularly, an operation involving a reversal of polarity in the DC system can cause system overvoltages to reach levels of upto 2.5 p.u. This results in a severe stress on spacers and can cause a drastic reduction in the fiashover voltage. There are no commercial DC GITL systems in service, but a few experimental facilities do exist. More recently, attempts have also been made to build DC converter stations by encapsulating them in an SF 6 environment for upto 330 kV. Another area where SF6 gas has found application is in Gas Insulated Substa-3 1. Busbar 6. Cable isolator 2. Busbar isolator 7. Make-proof earthing switch 3. Current Transformer 8. Vokage transformer 4. Work-in progress earthing switch 9. Cable termination 5. Circuit Breaker 10. Gas-tight, arc proof bushing Figure 1.1: Section through a typical GIS [2j. 4 tions (GIS). A metal enclosed substation (GIS) is an integrated system comprising high voltage bus, circuit breakers, isolators, disconnect switches, current transform-ers, potential transformers and surge arresters. Figure 1.1 shows a cross-sectional view of a typical GIS system. Several utilities worldwide have installed GIS par-ticularly in regions where land costs are high. This is because the use of SF 6 gas as an insulating medium has reduced the station sizes by a third. The advantages of a maintenance free and reliable system has spurred their extensive use in power systems. Typically these systems operate in an SF6 gas pressure range of 0.24 - 0.4 MPa. A critical part in the development of GIS has been the support spacer. Improved insulating materials, electric stress analysis and contamination control techniques have played significant roles in developing practical spacers which are able to with-stand high electrical stresses. Spacer geometries change with the type of applica-tion. Figure 1.2 illustrates some of the typical designs that have been developed and the following section describes the advantages and disadvantages of these designs from the flashover performance point of view. Spacers in GIS are principally manufactured in two ways, namely : • the spacer is cast directly on to the central conductor • it is separately cast (usually of a polymeric material, polysulfone, polypropy-lene etc.) and fitted into the system. Disc spacers have been frequently used in GITL systems. They have the ad-vantage that they are of simple geometry and occupy less space. It was probably around 1973 that post spacers were employed at UHV for GITL systems. Due to the relatively large system sizes involved at these voltages post spacers offered considerable economic savings, as compared to the disc spacers. Simple post spacers are cast with metallic inserts to make the surface field on the spacer nearly uniform in order to reduce the stress at the gas-electrode-insulator junction. Improper design of these inserts can result in substantial stresses inside the spacer which might exceed the breakdown strength. Tripost spacers were developed in an attempt to reduce both the internal and surface electric fields. 5 Cutout Disc Tripost SINGLE CONDUCTOR DESIGNS Dipost MULTI CONDUCTOR DESIGNS DISC SPACERS Inner conductor ^ 1 V 7 P Shielding electrodes Outer conductor POST SPACERS Inner conductor CONICAL SPACERS • Metal Inserts Outer conductor Shielding electrodes Metal Inserts Figure 1.2: Various cast-epoxy spacer designs [3]. 6 These spacers essentially consist of three posts distributed symmetrically around the conductor (see figure 1.2) moulded with a hub on the central conductor of the coaxial system. A complex metal enclosed substation is shipped to the site in sections. Indi-vidually tested sections are then assembled at site. Conical spacers (see figure 1.1) are normally used to assemble these individual units. There are several practical advantages in using compartmentalized units, ease of maintenance, being one of the important reasons. Conical spacers, however, tend to occupy more space. With developments pertaining to flexible EHV GITL systems, new types of spacer designs have evolved. The thrust in the design has been to reduce the triple junction stresses and find low-cost alternatives to replace the cast epoxy spacers. The flexible GITL has 2 or 3 separate subconductors which fit around the star shaped thermoset moulded spacer (see figure 1.2). Typically, in these designs the maximum field on the insulator surface is 40% lower than a conventional coaxial spacer [4],[6]. This results in a higher fiashover voltage. The disadvantage however, is that the maximum conductor field for the bundled system is higher by 30% in comparison to a coaxial design so care is needed with the conductor. Cylindrical post spacers are perhaps the simplest shapes that have been used to study the effects of spacer on a plain gas gap fiashover. For spacers with this kind of geometry, the fiashover performance is dependant on the degree of contact between the electrode and spacer, and the surface condition of the spacer. Corrugated cylindrical spacers have also been employed and the choice of profile depends on the type of application and material used. Quite recently, new types of ribbed post and conical spacers have been devel-oped for a 500kV 8000A GITL [7]. The authors claim that the performance of these spacers under particulate contamination shows a substantial improvement over the existing designs. It has been recognized that most often the presence of a spacer in GIS or GITL systems reduces the breakdown voltage in comparison to a plain gas gap. The fiashover voltage of a spacer in such a system, (measured as a function of increasing pressure), may be lower than in the gas gap without the spacer. Of course, this 7 will be a strong function of the spacer shape, spacer material, electrode, spacer gas interface, presence of contamination and the voltage waveform. Figure 1.3 illustrates this reduction. The reduction in the breakdown voltage may even be up to 60% for a poor spacer configuration. It has been found that often it is the gas-spacer-electrode junction, also known as the triple junction, which constitutes the weakest point in the system. Due to the high local electrical stress, microdischarges originating at the junction can lead to surface flashover. Studies have been conducted with narrow gas gaps 0.5mm [8],[9] underneath a spacer placed between plane parallel electrodes. These studies have shown that discharges occurring at the junction can lower the overall flashover voltages (FOV) by a factor of 2 to 3 when compared to the case with no gap. This deterioration was found to increase with applied voltage and gas pressure. There-fore, in order to get a better understanding of the physical phenomenon, attempts to isolate these triple junction effects have been made by several investigators. One of the measures adopted has been to metallize the insulators at the junction where they make contact with the electrodes. The presence of water-vapor, free and fixed conducting and non-conducting particles, decomposed gases resulting from arcing, can also alter the flashover char-acteristics of spacers. Particulate contamination, in the form of thin wire particles, has been found to drastically reduce the dielectric integrity of both GIS and GITL systems. As these systems are assembled on site, it is difficult to ensure the cleanliness of the system. Further, during the operation of the circuit breaker and disconnect switches, metal particles (dust) can be introduced into the system. Better contact designs with minimum wear have been found to minimize this ingress of particulate contamina-tion. Considerable amount of research has been directed towards the understanding of particulate behavior and methods to trap them. Test techniques together with particle traps have been able to minimize the problem. These particle traps es-sentially deactivate the particles by physically moving them into regions of zero or low electric field where they cannot acquire charge and levitate under the am-bient electric field. Careful designs, like application of adhesive to particle traps, have made it possible to prevent these particles from levitating. It has been re-8 500-1 Figure 1.3: Effect on spacer flashover in uniform field of a gap at an electrode interface or in middle of spacer[8]. 9 ported [1] that typically 40% of GITL systems in service, have these particle traps. Other measures like applying a dielectric coating to the outer electrode surface has been found to prevent a particle from acquiring charge and thus preventing it from levitating under the operating electrical stress. Despite all the above mentioned problems with GIS and GITL systems research efforts are still being made to improve the system reliability and at the same time provide economical designs. In a recent publication, Cookson [1] predicted that, "the major thrust for future research and development for new GITL systems will be driven by economics and reliability". Research is already underway to examine methods to automate factory and field assembly of GIS and GITL systems. There have been attempts to design semi-flexible and flexible SFe gas cables to reduce the number of field joints and reduce the associated problems, improving reliability. Present epoxy spacers have proven to be expensive and the operating stresses are limited, thus providing little to economize on the system costs. Future research attempts to use improved materials, particularly thermoplastics, have been suggested. Increasing the operating stresses in the system would reduce system size and hence costs. Efforts have also been directed to using alternate gases or gas mixtures of SF6 and N2 . This would result in a cost reduction. In summary, spacer flashover is a little understood phenomenon. Despite rigor-ous factory testing of both GIS and GITL systems, in-service failures are causing concern in the industry. There have been several studies undertaken to determine the cause for breakdown voltage reduction for a gas gap in the presence of a spacer. Particulate contamination, pressure and spacer geometry are some of the several factors that affect the breakdown behavior. In an effort to give a broad overview of the role of each of these parameters on the breakdown behavior (surface flashover) the following chapter presents a literature survey. Chapter 2 B R E A K D O W N BEHAVIOR OF SPACERS IN GIS For a gas insulated system, not only is the intrinsic quality of the gas important, but also the supporting spacer plays a significant role in determining the overall insulation strength. In addition, a practical gas insulated substation system is contaminated with non-conducting and conducting contaminating particles, water vapor and decomposition products due to arcing. Some of this contamination may be present on the spacer surface. All these factors significantly affect the dielectric integrity of the system. The following is a review of these factors. 1.1 Spacer Material In practical GIS and GITL systems, various types of spacer materials have been tested and used, which include porcelain, epoxy resins and thermoplastics. The choice of a specific material involves examining the following properties (this list is not made in any particular order of importance): • the dielectric constant • ability to withstand surface arcing • volume and surface resistivities • performance at elevated temperatures 10 11 • compatibility with SF 6 environment especially under the influence of arc byproducts of the gas • mechanical strength. • high internal electrical strength and long life endurance at the operating stress • surface secondary emission characteristics Some of these factors will now be discussed and their influence on spacer fiashover performance examined. Under impulse and AC voltages, the voltage distribution along the spacer sur-face and inside the spacer is dependent on the dielectric constant of the spacer. A spacer with a high dielectric constant having surface irregularities such as a depression may strongly distort the electric field acting along its surface. Cooke and Trump [10] have shown that a spacer with surface irregularities, significantly distorts the local electric field along the spacer surface. To examine the effect of a conducting particle adhering to a spacer surface, Andrias and Trump [11] con-ducted experiments in which it was shown that conducting particles on a high dielectric constant spacer could initiate microdischarges at relatively low voltages. The microdischarges could in turn trigger a premature fiashover. In cases where the spacers are premoulded and later fitted on the conductor, it is preferable to use a low dielectric constant material for the spacer. This would minimize the stress en-hancement at the spacer triple junction, the junction at the gas - electrode - spacer interface. In contrast, for DC operation of spacers, the situation is more complex. It has been recognized that the breakdown behavior of SF6 gas itself at DC stress does not differ significantly from AC stress (except for corona stabilization). In the presence of a spacer located in the electric field, the breakdown stress under DC is lower than that under AC stress. The initial voltage distribution under DC voltage application will be determined by the capacitive grading. Over prolonged periods of DC voltage application the stress distribution is determined by the resistivity of the spacer. Any surface charging will introduce additional changes in the voltage distribution. The bulk and surface resistivity of the spacer material influence this 12 charge accumulation and dissipation. A decrease in surface resistivity facilitates the decay of such trapped charges and reduces the enhancement of the electric field. Under high alternating voltage stresses, partial discharges can cause deteriora-tion of the spacers made of synthetic material. Although great care is taken during manufacture to eliminate cavities in the spacer material, differential expansion be-tween the conductor and spacer can lead to cavity formation in the spacer and/or at the electrode -gas - insulator junction and lead to partial discharges within the cavities. It is extremely important for a spacer material to be highly resistant to the effects of these discharges which can occur either within internal gas cavities of the spacer or on its surface. Dakin and Studianarz [12] have shown that even if spacer cavities in the bulk of the spacer go undetected at the primary testing stage, they can still be present resulting in an accelerated failure of the spacer under voltage stress. So the choice of the spacer material must be made on the basis of its resistance to these discharges. A spacer needs to have a good tracking resistance. In a practical gas insu-lated substation or gas insulated transmission line, arcing occurs during discon-nect switch operation. During high energy discharges involving a spacer surface, certain spacer materials erode resulting in damage to the spacer surface. This will deteriorate the insulation performance of the system. The self-restoring quality of the insulation system should be maintained. P T F E and unfilled epoxy resins tend to carbonize under a high energy discharge along the spacer surface[l3,14]. Alu-mina filled Cycloaliphatic and Hydantoin resin spacers have good resistance to arc damage under high energy discharges. By introducing fillers like Aluminum Trihy-drate into the spacer composition, surface damage problems have been mitigated. However, the relative dielectric permittivity of the spacer material is increased by the Aluminum Trihydrate fillers. Also, Aluminum Trihydrate filled spacers have a higher thermal expansion coefficient as compared to a silica or quartz filled epoxy spacer. For this reason sometimes, blends of fillers are used to obtain a spacer material having a low dielectric constant and also a low thermal coefficient of ex-pansion. Chenoweth et al. [15] have described the development of a high track resistant epoxy resin. 13 Although spacer materials mentioned above do not react with pure SF6 , de-composition byproducts of SF 6 gas attack and degrade the spacer. Therefore, the choice of a suitable resin/filler system should take into account the presence of decomposition by-products of arced SF$ gas. Due to its high resistance to arcing damage, porcelain has been used as a support spacer. But the combination of arcing byproducts deposited on the spacer together with moisture has proven to deteriorate the spacer performance. Also, porcelain is expensive, extremely brittle and has a high dielectric constant. On the other hand, thermoplastic spacers deform under elevated temperatures and hence are not suitable for GIS application. The most successful spacer material developed so far is a low-cost epoxy compound (Hydantoin) having a high resistance to arc damage. More recent installations of gas insulated transmission lines and gas insulated substations utilize epoxy resins as spacer material for post, conical and disc spacers. P T F E has been employed for multiblade spacers used in flexible GITL systems. 2.2 Behavior of a Spacer in the Presence of Con-tamination GIS and GITL are generally transported from the factory to the site, in sections. Metallic contaminants inadvertently get introduced into the system during final site assembly. Mechanical abrasion and arcing occurring during operation of the isolating switches and circuit breakers produce metallic particles. Further, the system may contain water vapor, nonconducting particles and byproducts of gas decomposition as a result of discharges. It is well known that any metallic particle present in the gas lowers the corona onset and breakdown voltages considerably. The effect of particle contamination on spacer fiashover is reviewed in the following subsections. 2.2.1 Free Metallic Particles A gas insulated substation or transmission line with spacers is sensitive to particles [17,40]. The influence of these particles on the breakdown behavior of a spacer 14 system is difficult to predict as it depends on the location of the particles, their sizes, their number, shape and the gas pressure. When a conducting particle lying on the outer conductor of a horizontal GIS/GITL section is exposed to a DC electrical field, it acquires charge and levitates and moves towards the inner high voltage electrode. If the particle is close to the high voltage electrode it may lose charge through a microdischarge. It has been suggested that the microdischarge makes the particle behave as an extended protrusion from the electrode into the inter-electrode gap. This results in a premature breakdown [41]. In the presence of a spacer however, particle contaminants may lift off the enclosure (low voltage electrode) and then get attached to the spacer. Under DC voltages, the spacer accumulates charge and the particles adhere strongly to the spacer. Cronin et al. [42] have found that under positive polarity of the inner conductor, the particles tend to accumulate on the spacer surface, whereas for negative polarity they strike the inner conductor. The reasons for this behavior are obscure. Figure 2.1 shows the maximum breakdown strength of 4 different types of spacers for rapidly applied DC voltages [18]. Flashover of the gaseous insulation and spacer flashover voltages without contamination are shown by solid and dashed lines respectively. The figure also shows the effect of metallic particles (6.4 mm long 0.45mm in diameter) on the surface flashover strength of spacers with and without corrugations. Spacer types a and b failed at lower voltages as compared to those of type c and d. These studies showed that the increase in creepage path does not necessarily increase the breakdown voltage and that insulator surfaces should avoid high stress. For alternating applied voltages free conducting particles in the GIS/GITL systems hover in dynamic equilibrium around the inner conductor accompanied by a certain luminosity. These particles are referred to as fireflies [18]. In the presence of a spacer, fireflies are frequently seen approaching the vicinity of the spacer and moving away without touching or hopping on to the spacer surface. Experimental evidence shows that the flashover characteristics of a spacer with metallic particle on its surface depends on the particle location. It has been shown for example, that, for a conical spacer, particles present on the convex side of the spacer result in lower flashover voltages as compared to when particles are present 15 I 1 1 1 0 500 1000 1500 PRESSURE [kPa] Figure 2.1: DC fiashover voltages for four different types of post spacers with 6.4 mm wire particle in SF 6 [18]. 16 on the concave side [42]. Nitta [17] found that a small amount of metallic powder (350 mg, size 30//inch) is harmful to the flashover performance of the spacer. Figure 2.2 shows the influence of copper powder on the AC breakdown of two systems, with and without spacers [17]. Figure 2.3 shows the liftoff, crossing and breakdown voltages for free particles present in a coaxial gap, as well as the breakdown voltages for particles fixed on a spacer. It can be seen that for SF$ gas at pressures below 500 kPa, the AC breakdown voltages are the lowest for free particles when compared to fixed particles on the spacer and on the conductor. 2.2.2 Fixed Metallic Particles Several researchers [30,36,39] have studied surface fiashovers with the particle ei-ther fixed on the spacer or on the electrode surface close to the spacer surface. With protrusions on an electrode surface, breakdown is usually caused by field enhance-ment. Figure 2.4 shows the effect on surface flashover voltage when a particle is located on a spacer surface. DC breakdown voltages of a post spacer in compressed SF 6 , with a particle fixed to the spacer surface, have been found to be upto 30% lower than the corresponding minimum AC voltages. The length of the particle fixed on the spacer has a profound effect on the flashover voltages of disc spacers under AC and impulse voltages. For DC voltages the values are substantially lower than the corresponding AC breakdown voltages. In order to get a better understanding of the factors controlling the discharge propagation on a spacer surface, Pfeiffer et al. [44] recorded the luminous phe-nomenon using a highly sensitive image intensifier in a spacer gap. The test elec-trode arrangement comprised of two hemispherically capped electrodes embedded in an insulating spacer so that the spacer was aligned in the region of the high electric field. The quasi-uniform field distribution was perturbed by two arrange-ments. In one case, a small hole on the spacer (0.8mm in diameter) was filled with conductive paint at regions close to the anode and cathode on 2 different samples. In the other small conducting particle protruded from the surface of the insulator (1 mm) at regions close to the anode and cathode on 2 different samples. 17 Figure 2.2: Effect of amount of fine copper powder (30/xm) on the AC breakdown voltages of a coaxial system with cone type spacer [17]. 18 4 0 0 - 1 E > o > o < L d o: C D 3 0 0 -2 0 0 -1 0 0 -o Free • particle [with spacer] Fixed O particle on spacer Fixed V particle on conductor [with spacer] 6.4 mm long wire particle Lift Off Breakdown Crossing _ O r 0 —I \ — 5 0 0 1 0 0 0 PRESSURE [kPa] 1 5 0 0 Figure 2.3: Lift-off and crossing voltages for free particles and breakdown voltages for free and fixed particles in a SF 6 filled 76/250 mm coaxial system with post type spacer[43]. Figure 2.4: Influence of the length of copper wire on the flashover voltage of a disc type spacer [30]. 20 It was found that in direct stress breakdown experiments, in samples with conduct-ing paint, the decrease in breakdown voltage as compared to an unperturbed field arrangement was 10% regardless of the position of the conducting spot. The pre-breakdown discharge altered with the location of the conducting paint spot on the insulator. In similar experiments with the latter samples, the breakdown voltages were lower than 30%. The image intensifier pictures showed that the discharge initiation and propagation did not involve the spacer surface but the discharge developed in the same manner as in a field distorted gas gap. Laghari and Qureshi [39] investigated protrusion initiated breakdown in SFe -N 2 mixtures, for a uniform field (Harrison profile plane-parallel electrodes). With a hemispherically capped protrusion on the electrode (cathode), the breakdown voltages were observed to be reduced as compared to a uniform field gap. However, with a spacer located in the vicinity of this protrusion the DC flashover voltage of this field distorted gas gap was found to increase. This increase in breakdown voltage was ascribed by the authors to the possible reduction in the electric field at the protrusion due to the presence of the spacer. Similar results were obtained for gas pressures of upto 500 kPa. Cronin et al. [33] concluded that, in the presence of contamination, corrugated disc spacers show a more consistent performance compared with that of a smooth disc spacer, particularly under impulse conditions. To examine the breakdown behavior of a disc spacer with particles, under impulse conditions, Eteiba et al. [36] located a particle oriented radially ie., in the direction of the surface electric field, on a disc spacer placed in a coaxial SF 6 bus. They determined that this orientation was the most effective way to reduce the flashover voltage of a disc spacer. So if particle contamination on a spacer aligns itself in the direction of the surface electric field it effectively shortcircuits the gap (proportional to its length) causing a reduction in the breakdown voltage. Under long term DC voltage application, volume charges in clean epoxy spacers accumulate at a moderate rate and distribute uniformly. The charge polarity is the same as that of the inner electrode of the coaxial system. However, with a particle fixed to the spacer, this homogeneity is lost. Not only have sharp peaks in the charge intensity been observed corresponding to the particle position but also their polarity are reversed. Steady pre-breakdown currents have also been observed in 21 Kinds of % Fiashover voltage (kV) particles AC Impulse no contamination 95-100 95-100 copper wires 0.2mm dia., 20mmlong 52-62 70-90 0.18mm dia. 5-10mm long 70-85 80-95 aluminum filings, 0.3mm, 50mg total 80-90 90-100 epoxy pieces 10-20mm 95-100 92-100 epoxy powder 95-100 92-100 cotton lint 10-20mm 95-100 95-100 Table 2.1: Effects of various kinds of particles on the fiashover voltage of practical spacers at 310 kPa SF 6 [17] the case of an elongated fixed particle on the spacer surface which eventually leads to a fiashover. For alternating voltages, experiments conducted with a particle fixed on a spacer have also revealed a drastic reduction in breakdown voltages. With increasing pres-sure it has been found that the corona inception and breakdown voltages coincide. In more recent experiments [69] with a particle contaminated system subjected to AC voltages, it has been found that wire particles lift and reach the spacer at voltages comparable to the system operating voltage. The particles adhere to the spacer surface for several hours but do not crawl towards the high voltage conduc-tor. If the particle is near the high voltage conductor and on the spacer, a dramatic reduction in breakdown voltage (down to 25% of an uncontaminated system) has been observed. 2.2.3 Non conducting Particles The effect of insulating particles in reducing the fiashover voltages of a coaxial system is less than that due to conducting particles [17]. Table 2.1 indicates the effect of various particulate contamination on spacer fiashover voltages. It has been observed that insulating particles in combination with conducting particles [18] could impair the insulation performance. The conductivity of a particle really influences the fiashover in a particle contaminated SF 6 gas gap. 22 2 . 2 . 4 Water Vapor Although there are reports available [47] which indicate a reduction in breakdown voltage for a plain gas gap in the presence of moisture, Nitta [17] reports that the presence of moisture hardly influences the flashover voltages as long as no condensation takes place on the spacer surface. Hogg et al. [48] have observed a similar behavior. 2.2.5 Decomposed Gases One of the major products of decomposition as a result of arcing in SF$ has been found to be SF 4 [48]. Under moist conditions, SF 4 is likely to react with the Si0 2 filler in the epoxy spacer to form SiOF 2 , HF etc. These compounds are deposited on the spacer surface and lead to reduction in the surface resistivity. The reduced surface resistivity could result in a non-uniform field distribution and partial breakdowns between resistive layers which can eventually lead to a flashover. Further, arcing on the spacer surface could release carbon byproducts which are also deposited on the spacer surface. Deterioration of the spacer material and premature flashover of the spacer follow rapidly. More recently, it has been found that silica filler and cycloaliphatic resin spacers are prone to react with the arc by-products of SF 6 gas. It is therefore important to select the appropriate filler for a spacer. 2.3 Spacer Profiles The mechanisms of surface flashover of spacers in compressed gases are not fully understood and there is a lack of agreement on the actual parameters controlling surface flashover. For this reason, several optimal designs of spacers have been proposed. The following sections review the spacer designs that have been proposed and their relative merits. 23 2.3.1 Post Spacers One of the simplest support spacers, the post spacer, was introduced by Cooke and Trump [10]. Metal inserts are generally used in post spacers to reduce the triple junction effects. Simple cylindrical post spacers have also been extensively employed by many researchers [8,16]. They have been used both in uniform and non-uniform field gaps to study the influence of a spacer on the breakdown of an SF 6 gas gap. The exposed surface area of the post spacer is small thereby reducing the prob-ability of contaminating particles getting attached to the spacer under operating conditions. Studies have been undertaken to examine the effect of corrugations on post spacers. It has been observed that under clean conditions corrugations on spacers reduce the fiashover voltages [16,17]. Under contaminated conditions, however, corrugations did not necessarily increase the fiashover voltage of a con-taminated spacer [18]. Experimental results reported in literature indicate that corrugations could either increase or even decrease the fiashover voltages [20],[21]. Trump [11] et al. have reported the advantages of these corrugations at higher pressures. They have suggested that corrugations provide shielding effects against microdischarges initiated at a triple junction. It appears that when a corrugated spacer is terminated at its depression on the electrode, the nearest corrugation may act as a shield for microdischarges initiated at the triple junction. Thus, studies carried out to improve the performance of these spacers in the presence of contamination have been inconclusive. The factors controlling breakdown behavior of post spacers when placed be-tween plane parallel electrodes are: • the degree of contact between the electrode and spacer • the condition of the spacer Figure 2.5 illustrates a typical arrangement of a cylindrical spacer placed between plane parallel electrodes. Manufacturing defects could give rise to a small gas gap between the electrode and the spacer, as shown in the figure. The electric field in this region would be the product of the dielectric constant and the average field Figure 2.5: Arrangement of a cylindrical spacer between plane parallel electrodes, ] - Electrodes, 2 - Insulating spacer. Gap dx is exaggerated to expose the triple junction. 25 [20]. As an example, consider a cylindrical spacer placed between a pair of plane parallel electrodes. If d\ is the gas gap and c^ is the spacer length, er the relative dielectric constant of the material, then the average stress in the gas gap is given by: Eg = — T — . 2.1 Corona inception starts at a relatively low voltage and is given by: d2 Vi = Vd 1 + (2.2) [di X 6r)_ where Vd is the gap sparking potential. In order to improve the corona inception voltage, the triple yurcc£icm(gas-metal-spacer junction) may be shielded by recessing the HV and ground electrodes as shown in figure 2.5. It has been found that such a measure improves the spacer fiashover performance. Thus, the inception voltage depends on gap length, spacer dielectric constant and pressure. Skipper and McNeal [20] indicate that for spacer gaps less than 200 micrometers, a perfect contact is not essential. The spacer surface also has significant influence on the discharge development. Surface defects, charge patches etc., could attract metallic particles and initiate microdischarges leading to a premature fiashover. These factors lower the fiashover voltages in uniform fields. 2.3.2 Conical Spacers Figure 1.2 shows the cross-section of typical conical spacers. Conical spacers occupy more space than other spacers. They are extensively used, however, as gas stop joints [24,28] for single phase GITL. Their contamination performance has also been observed to be better than other types [27]. The angle of inclination which the spacer surface makes with the high voltage conductor influences the overall field distribution. A smaller angle of inclination results in a reduction of the spacer surface field but the electric field at the triple junction near the high voltage electrode is increased. Increasing the angle of incli-nation causes higher stresses at the triple junction near the outer grounded sheath. The optimal angle of inclination for conical spacers for use in GIS and GITL spac-ers has been found to be 45° [27]. Shielding electrodes are inserted in the post and 26 disc spacers, both at the inner and outer electrodes, to reduce the triple junction stresses. In contrast, for conical spacers the inner conductor is moulded to the spacer and the outer boundary has an insert and is flanged to facilitate its assem-bly between two GITL sub-sections. With electric field analysis, Itaka [19] reports further reductions in the the triple junction stresses at the low voltage electrode for conical spacers with the optimum angle of inclination. The authors describe the modifications made to the shape of the flanged end of a conical spacer (where it is bolted between two flanges of the SF 6 gas insulated apparatus). This modification permitted dispensing with the normal metallic inserts which are used to reduce the triple junction stresses. Problems associated with the incompatibility in the coefficients of expansion of metal inserts and spacer material leading to cavity for-mation, partial discharge and eventual failure are therefore avoided. The successful installation of corrugated conical spacers in a 420kV bus is reported in [28]. 2.3.3 Disc Spacers Figure 1.2 shows the construction of disc spacers. Due to their compactness and simple geometry, these spacers have been very widely used [29,33] Split disc [25], doughnut shaped [26] disc and corrugated disc [33] are some of the variations employed in GIS and GITL. Takuma et al. [34] derived the optimum disc spacer profile by means of field computations for materials with different permittivities. A spacer profile angle was defined to be the angle between the tangent to the spacer surface and the electric field lines for a coaxial electrode gap with a disc spacer. Depending on this angle and the dielectric constant of the spacer, the electric field distribution on the spacer gets altered. Evidently, the optimum spacer profile angle is a function of the relative diameters of the conductors of the coaxial system. They also report that the spacer profile angle is larger for smaller inner to outer diameter ratios. However, the maximum breakdown electric field strength in the presence of the optimum profile spacer decreases to 70-80% of that of a coaxial gas gap without a spacer. Comparative studies undertaken by Vandermeeren [35] showed that an optimal disc spacer performs better than a conical spacer. For the disc spacer the discharge initiation at the triple junction can be delayed due to the uniformity in 27 the tangential field. However, once the discharge is initiated, discharge propagation cannot be arrested. The cone spacer on the other hand is very susceptible to triple junction effects but has reduced surface tangential field, criticalfor discharge propagation. Researchers at IREQ, based on field computation and experiments, developed a composite profile cone spacer. The profile was a combination of both the disc and cone spacers. This design combined the advantage of a long leakage path of a conical spacer (30° angle) with the near uniform field of a disc spacer. By introducing additional metal inserts at the outer end i.e., at the low voltage end, triple junction stresses were further reduced. They claimed that the flashover performance of this design was almost as good as that of the system without the spacer. 2.3.4 Multiblade Spacers Figure 1.2 shows a schematic of the Trefoil spacer which is particularly convenient to use in multiconductor SF 6 GITL. Conventional spacers are difficult to apply in a flexible cable because the spacers have to be fitted around the flexible inner conductor (eg. post spacers supporting the central conductor in a coaxial system). Therefore, a design in which the conductor is fitted around the spacer is more advantageous. This has been achieved by use of multi-blade spacers fitted around the conductors. The normal central conductor is replaced by several sub-conductors at the same potential with the spacer blades protruding between the conductors. It has been shown by Hampton [4] that the maximum surface stress for these 2, 3 and 4 blade spacers are approximately 50, 60 and 65% of that possible for a plain disc spacer, with the same outer to inner diameter ratios of 2.7. Also the flashover voltage has been observed to be independent of small gaps between the spacer and conductor. Banford [37] has reported that a 3 blade Trefoil spacer performed well under contamination conditions. 2.4 Effect of Voltage Waveshape The breakdown behavior of solid support spacers in GIS, depends on the wave-shape of the applied voltage. In the case of alternating and impulse voltages the 28 Figure 2.6: Residual fields on post spacers after test at +600kV DC stress [38 . 29 voltage distribution over the spacer is governed by the distributed capacitance. For direct voltages however, the spacer resistivity determines the field distribution. Further, the flashover performance of spacers for the various types of voltages is quite different in a particle contaminated environment because the dynamics of free particle motion is also dependant on the waveshape of the applied voltage. In the case of DC voltages it has been reported [38] that the spacer may accumulate both surface and volume charges when the voltage is applied over long periods of time. Figure 2.6 shows the maximum residual field at the inner conductor of a post spacer in SF6 as a function of duration of the applied voltage. It may be ob-served that the amount of surface charging depends on the length of time the field is maintained and the magnitude of the applied voltage [38]. The resulting field distribution due to the charges could result in the breakdown of the surrounding gas: Further, charge patches of this kind can attract contaminating free metallic particles present in the system, resulting in excessive local stresses and promoting long-term failure by stress enhancement. Figure 2.7 shows the variation of flashover voltage with pressure for different voltage waveshapes [39]. The results have been obtained with rapid application of the voltages. From the figure it is apparent that the impulse flashover voltages are higher than alternating or DC voltages. In a coaxial system, the impulse breakdown voltages depend on the spacer shape, type, size and location of particulate contamination. For example, a Tre-foil spacer [4] has a higher positive impulse flashover voltage as compared to the negative impulse flashover voltage; the opposite when there is a metallic particle on the spacer. Thus, the prediction of spacer performance subjected to different waveshapes is difficult. Inspite of the volume of experimental data that have been collected so far on the surface flashover in SFQ , the following questions still remain unanswered: • What is the nature of the accumulated charges ? • What are the energy levels of these trapped charges! • Are the measured charges only surface charges or are they in the bulk of the solid? 30 240-1 > Uj 160 O O > cr > o < 80 H 0 - T • IMPULSE [1.5/50 s] 1.5 2 2.5 3 3.5 PRESSURE [aim,abs] Figure 2.7: The uniform field fiashover voltages of a 10 mm cylindrical spacer under AC, DC, and Impulse voltages [39]. 31 • What is the interaction between a charged spacer surface and an adjacent electron avalanche? 9 What is the physical mechanism of a surface fiashover and how does the surface influence the discharge propagation? The following section reviews some of the work reported so far, and the models that have been advanced to explain spacer surface charging and fiashover phenom-ena. 2.5 Models for Surface Charge Accumulation Most of the investigations on surface charge accumulation of spacers have been carried out with direct voltages. Over extended periods of DC voltage application, the bulk conductivity of the spacer together with its surface conductivity plays a significant role in determining the overall charge distribution. By using mate-rials with higher surface conductivity as compared to volume conductivity, some investigators [50] maintain that the bulk conductivity is no longer the dominant mechanism, for the formation of surface charges. 2.5.1 D C Voltages Charge accumulation on the spacer surface has been studied using post and coni-cal spacers by several authors [38,50]. Mangelsdorf and Cooke [38], have identified two distinct charge patterns for post spacers stressed at high DC voltages in an SF6 environment. They found that the post spacers accumulated a smooth con-tinuous charge. The magnitude of these charges increases with time and voltage. This smooth charging has been found to be of the same polarity as of the central conductor of the coaxial system of electrodes. The charge distributions have been measured by removing the sample from the SF 6 environment. The maximum resid-ual field appears to level off with increased duration of the applied voltage, at levels of +100kV. However, the residual field shows a pronounced non-linearity at voltage levels of +600kV. It is suggested by the authors that this non-linearity in charging 32 may be due to the non-linear conductivity of the material or perhaps due to the on-set of another charging mechanism whose threshold is around +600kV. The author reported analogous results when the polarity of the applied voltage was reversed. On the other hand, with a particle deliberately fixed to the spacer, at the center of the spacer and located parallel to the spacer axis, the spacer surface around the particle was flooded with charges opposite to that of the inner electrode, thereby disturbing the homogeneity of the uniform charge pattern seen earlier. Due to the time dependent nature of homogenous charging, Mangelsdorf and Cooke suggest that the bulk conductivity of the spacer was responsible for charge accumulation. The second charge pattern that has been observed by Mangelsdorf and Cook is termed as streak charging. This pattern has been attributed to the possible ionization of the gas surrounding the particle. It has been concluded that this static charge accumulation would lead to a high electric field which could result in an electron avalanche in the gas adjacent to the spacer. Knecht [51] examined the surface charge accumulation under both AC and DC voltages in N 2 and SF 6 environments using cylindrical epoxy spacers 40 mm in diameter and 40 mm in length. The spacers, with suitably placed metal inserts have been located between plane parallel electrodes in the gas environment. The gas pressures used by Knecht are 0.1-0.4 MPa for SFQ and 0.4MPa for nitrogen. At constant voltage no clear dependence of the charge accumulation on either the gas pressure or gas type has been observed. From the measured charge patterns, it has been suggested that the charge accumulation is determined by the concentration of carriers and their mobility in the direction of the electric field normal to the spacer surface. The source of these charges has not been discussed. Knecht's hypothesis has been supported by the fact that the volume and surface conductivities were low for the spacer materials employed, thereby minimizing their influence on the charge accumulation. These experiments suggest that spacer charge accumulation is a consequence of gaseous conduction. Nakanishi et al. [50] have proposed that the nonlinear surface conduction of the spacer is the principal cause for heterogeneous charge accumulation. By using cylindrical samples cast with metal inserts to reduce the triple junction effects and placing them in a uniform field, it has been observed that the spacer surface con-33 dition strongly influences the charge pattern. It has therefore been suggested that the charging is due to surface conduction only. Theoretical considerations based on surface and volume conductivity measurements show that the bulk charging mech-anism would require more than 80 hours of charging at high voltages to produce the observed charge distributions. On the other hand, the experimental charge patterns appeared within five hours of voltage application. Therefore bulk con-duction could not have been the dominant charging mechanism. Nakanishi et al. attribute the charging to a mechanism solely involving the spacer surface. Intro-ducing the concept that the surface conductivity has a strong dependence on the local electric field, the possible charge distribution on the spacer surface has been computed and it matches the measured charge distribution. An example of the calculated and measured charge distributions are shown in figure 2.8. It can be seen that, in a field dependent surface conduction model, the computed charge dis-tribution matches both the dust figures and the probe measurements along the line A - A'. Figure 2.9(a) and (c) indicate the type of spacer used between the coaxial electrodes. Ez is the component of the electric field tangential to the spacer sur-face and En is the electric field normal to the spacer surface. Figure 2.9(b) shows the distribution of the tangential and normal components Ez and En as shown in figure 2.9a of the electric field on the spacer. Despite the asymmetric distribution of the normal component of the electric field, the measured charge pattern has a symmetric distribution. There is less agreement between the observed charge patterns and the theo-retical charge distributions predicted by Knecht's model. In particular, Knecht's model does not yield a symmetric charge distribution even though the normal component of the electric field is asymmetrically distributed. The field-dependant surface conduction model provides a better explanation for the observed charge patterns. Nakanishi et al. have further observed that when the the inner electrode of the coaxial system is negative and has a surface finish of approximately 20 to 30 /zm, the inner surface of the conical spacer accumulates charge. On the contrary, with the inner electrode positive, no charge accumulation has been observed. Further, when the inner electrode was polished to a finish of 5/zm, there is no charge accumulation 34 Figure 2.8: Residual potential distribution on both untreated and roughened spac-ers and theoretical calculation of charge accumulation. Gap spacing between metal inserts is 20mm [50]. 35 c Dust fig u r e pattern following voltage a p p l i c a t ion Figure 2.9: Radial and tangential field components of post spacer in a coaxial bus and dust figure of spacer energized with negative D C of 100kV for 5 hours [50]. 36 observed on the spacer irrespective of the polarity of the inner electrode, giving some credence to the field-emission theory. However, more recent studies [52] with -500kV applied to the central conductor of a DC GIS with a conical spacer show that the application of DC voltages for 30 minutes at 0.4 MPa SF 6 pressure does result in a negative charge distribution on the concave side and a positive charge accumulation on the convex side respectively. When the the central conductor is made positive, a weak positive charge distribution is observed on the concave side and a negative charge distribution on the convex side. The mechanism of positive charge generation is not well understood at the present time. It has been suggested that field emission is probably not the only mechanism causing this charge accumulation [52]. From the experimental evidence presented it may be recognized that spacers get charged under DC voltage application. Studies on scaled models and also on actual spacers used in GIS seem to indicate that these charges have large decay times (i.e, they reside on the spacer surface over several hours [61]). Under such circumstances, these charges can distort the normal electric field which can result in a premature spacer flashover. 2.5.2 Impulse and A C Voltages Few systematic investigations have been reported on charge accumulation on a spacer and its influence on the surface flashover voltage with impulse and alternat-ing applied voltages. Cun-Yi Yu et al. [53] have examined the influence of charge accumulation due to DC voltages on the impulse flashover of spacers in SF6 . In this study, the spacers have been subjected initially to direct voltages and par-ticulate contamination introduced in the test arrangement. The influence of the spacer profile on the impulse flashover voltage has been subsequently examined. They found that corrugated spacers tested at 0.2MPa SFe pressure have almost the same first flashover as a spacer with smooth surface. However, when the spac-ers are exposed for a long time to direct voltages or spacers are contaminated by metal particles, the performance of the corrugated spacer is superior. There has been very little reported on charge accumulation on spacer surfaces subjected to impulse and AC voltages [46]. The evidence of spacer charging under AC voltages 37 is conflicting. Knecht [51] observed a very small charge accumulation apart from a local peak under AC stress. Other experiments with hollow spacers in vacuum and N 2 gas indicate that spacers do get charged under alternating stress. But no pre - breakdown currents were recorded. The authors attribute this anomaly to bandwidth limitations of the equipment. It has been recognized that spacers in GIS have a significant influence on the V - t characteristics of the insulation. Experimental work conducted at IREQ [55] indicates that spacers subjected to repeated impulse voltages of a particular polarity cause sudden reductions in the predicted 50% probability of breakdown voltages (V50) It has been suggested that surface charging is the probable reason for this sudden reduction in the V 5 0 voltages. No systematic investigation to deter-mine the cause for this abnormal behaviour is reported. In summary, it has been recognized that with DC stress, insulating spacers in GIS accumulate charge. The nature of charge accumulation is one of the principle causes for sudden reductions in the fiashover voltages. Systematic investigations of this nature with alternating and impulse voltages have not been reported so far. The following section discusses studies undertaken to determine the interaction between a spacer and the adjacent gas gap. 2.6 Surface Discharge Initiation and Propaga-tion In a gas gap without space charge, the highest electric stress according to Laplace equation is always at an electrode surface and in a direction normal to the electrode surface. As mentioned earlier, a dielectric introduced in the same field distorts the homogeneity of the field resulting in both normal and tangential components on the spacer surface. The magnitudes of these components will get altered with sur-face and bulk charges. In an attempt to determine if solid dielectric surfaces are inherently much more limited in their ability to sustain electric fields, Cooke [54] compared published data on the electric stress for fields, mainly normal and tan-gential to a spacer surface. Cooke [54] suggests that an epoxy surface is capable of sustaining electric stress comparable to that of a gas gap and that for a clean uncon-38 taminated insulation system with a spacer, gas ionization is an important process and provides the upper bound for improved insulation. Since surface flashover voltages are sensitive to external perturbations e.g., triple junction effects, charges and surface irregularities etc., these factors have to be isolated and/or controlled for a systematic study of surface flashovers. Surface flashover process is typically a two step event: first, the discharge initi-ation and then propagation. The initiation of the discharge results from local field disturbances of sufficient magnitude. Propagation results when there is sufficient field along the propagation path. Cooke [54] describes an experimental model to explain the observation of surface flashover voltages. In these studies, a cylindrical spacer, with a piece of conducting wire attached to its surface, has been placed between toroidal electrodes. Triple junction effects have therefore been eliminated by shielding. The growth of the number of electrons in the vicinity of the wire have been considered, with a critical avalanche size of 108 electrons being taken as the criterion for discharge inception. The spacer by itself was not considered to be active in the discharge development process, except as a means to influence the electric field pattern because of its geometry and relative permittivity. It is presumed that flashover results at every instant the critical field is reached. In very divergent geometries, this criterion for the critical avalanche size does not necessarily result in a flashover. This condition is also known as the streamer criterion and results in a corona discharge. Fig-ure 2.10 gives the details of the experimental setup used to verify the predicted discharge inception stress. A good correlation between the calculated discharge in-ception stress and the flashover voltage was found. It is also reported that, a direct stress application followed by a polarity reversal resulted in a very low breakdown voltage which deviated considerably from the predicted voltage using the streamer criterion. It is possible that surface or space charges have an influencing role on breakdown. The described criterion does not account for these charges in the pre-diction of the breakdown voltages. The possible role of the spacer on the ionization process, the role of the spacer on charging and its subsequent influence on break-down have not been accounted for, by his model. Cooke's experiments represent one of the earliest attempts to explain surface flashover in SF 6 gas with direct 39 voltages. In order to study spatial and temporal pre-discharge development, Pfeiffer et al. [58] used a special electrode/spacer configuration. The geometry is described on page 16. A high speed image converter system has been used to record the pre-discharge phenomenon occurring between the hemispherical electrodes subjected to impulse voltage excitation. Due to the high framing speed and short exposure time of the camera, only two frames with a framing interval of 1 nanosecond (ns) could be recorded at each impulse voltage application. The impulses have been applied repeatedly to obtain different time frames prior to final breakdown. Finally, several records have been put together to reconstruct the discharge development process. It has been found that in SF6 , the first luminous spot may be observed at random points along the spacer surface. The number of luminous spots grow with time eventually combining and leading to breakdown. It has also been observed that the discharge propagation velocity on the spacer is higher than that in a plain gas gap. In an attempt to determine the role of surface charges on the corona discharge phenomenon, Mailer and Srivastava [56] have conducted experiments using a cylin-drical spacer placed in a non-uniform field (rod plane gap). They suggested that the charges on the spacer surface may have a significant influence on the corona discharge phenomenon. The study has been aimed at determining the role of ini-tiatory electrons in a developing discharge. Stoop et al. [61] describe the role of surface charges on the discharge devel-opment. By using a laser beam to trigger a fiashover across a cylindrical spacer placed between plane parallel electrodes subjected to DC voltages, they have ob-served that the fiashover times are smaller than that of a plain gas gap fiashover. It has been suggested that photon induced detachment of electrons from the spacer surface is the cause of reduced fiashover times with spacers. 2.7 Present Work Surface fiashover of spacers in compressed SFQ is a much studied but poorly un-derstood phenomenon. Much of the modelling of spacer fiashover has been based 40 R o d I n s u l a t o r studies Toroids: 7in OD, 3in s e p a r a t i o n TOROID Rod Diameter: 2.0inch Rod dielectric constant: 2.5 Rod length: 12inch Figure 2.10: A test chamber for surface flashover studies[54]. 41 on the more mature understanding of breakdown in gases. Breakdown in gases has been modelled by considering the Townsends's ionization coefficients namely: a the rate of free electron and positive ion formation by electron impact ionization, n the rate coefficient for formation of negative ions by attachment, 7 , the rate of secondary electron formation at the cathode, 6 the electron formation rate from an unstable negative ion and (3 the conversion of an unstable negative ion into a stable one. All these coefficients for each gas are in turn a function of the gas pressure and the electric field. Breakdown, at or near a spacer involves a modification of the above ionization coefficients in some yet unknown manner. Variations in the ion-ization coefficients are, in turn, affected by a host of other parameters. Although considerable data has been collected on the parameters affecting spacer fiashover, little progress has been made in building a theoretical model. Surface charging has been recognized as an important parameter influencing spacer fiashover. The present work therefore deals mainlv with the phenomena of surface charging. The fundamental parameter of interest, related to surface fiashover, has been surface charge densities. It has been mentioned earlier that most of the investi-gations on spacer surface charging have been conducted with DC voltages. The present work is concerned with the study of charge accumulation on spacer fiashover with impulse voltages primarily because it has been widely recognized from work with direct voltages that charge accumulation on an insulator surface modifies the electric field distribution in the gap which, in turn, influences the initiation and propagation of a surface discharge. Impulse voltages have been chosen in order to minimize the role of bulk conductivity in charge accumulation. The choice of impulse voltages also assists in a better interpretation of the electric field probe data, since it has been recognized that bulk charges complicate the extraction of surface charge densities from probe measurements. A rod-plane and a parallel-plane gap have been used in the experiments. This choice of gap geometries has been made for the following reasons: • They are representative of the spectrum of uniform and highly non-uniform fields in service • The degree of non-uniformity may be easily controlled by varying the rod 42 diameter • The axial symmetry built into the system provides easier electric field com-putation. Therefore theoretical models may be tested against experimental data. • Triple junction effects may be shielded by having a recess (2 mm deep 45 mm dia.) cut into the electrodes for the plane-parallel electrode configuration. For a systematic study of the complex surface flashover phenomena, it is essential that the investigation be clearly focussed on a few limited aspects of the phenom-ena. The selected aspects however, should be aimed at elucidating the phenomena and be of practical engineering interest. It is therefore of interest to study the role of surface charging on impulse flashover of spacers in the absence of particle contamination. It has been recognized that, to be able to obtain an understanding of the nature and growth of charges under electric stress in an SF6 environment, it is important to monitor these charges in-situ. Exposure of the spacer to air following voltage application could significantly alter the charge pattern. By removing the spacer out of the SF 6 gas environment, the dielectric strength of the gas (air in this case) surrounding the spacer has altered (reduced). Thus local discharges can occur resulting in a modification of the original charge formation. Many researchers in the past, while monitoring surface charge accumulation on spacers in SF 6 , have removed the specimen from the SF 6 environment to measure the residual charge. Conclusions based on charge magnitudes and fields under these conditions could yield misleading results. There are practical difficulties associated with in-situ measurements of surface charges using a field mill or a capacitive probe. The complexities of building an in-situ charge measuring system increase with the size of the SF 6 system. For the geometry used in this study, a in-situ charge measuring system is designed to provide for charge scanning of a cylindrical spacer approximately 40mm in diameter and 30 mm long placed on a recessed plane (low voltage) electrode. To reduce the probe measurement errors due to probe vibration, the insulating spacer should be moved past the probe. This requires an electrode drive arrangement by which both 43 the top face of the spacer (the region close to the high voltage conductor) and the cylindrical side could be scanned. The drives have been designed to provide both an axial and a rotary motion of the bottom electrode thereby enabling the spacer scans in-situ. Appropriate sealing is essential, since the drives are being located external to the pressure vessel. Since the capacitive probe heads were connected to the electrometer through the flanges and had to be withdrawn during the voltage application, special seals were made at these joints. The details of the charge measuring system and electrode drive, are included in chapters 3 and 4. To summarize, the main objective of this research is to examine the role that surface charges may have on surface fiashover of spacers in SF 6 gas. This study examines the nature of charging of spacers subjected to impulse voltages. The particular choice of impulse voltages was made to minimize the role of bulk con-ductivity of spacers in charge accumulation. Although precise in - situ charge measurements are difficult to make, it is essential to have at least an approximate knowledge of the charge patterns and their behaviour for improving the understand-ing of the surface fiashover phenomenon. This study would entail the development of proper measurement and interpretation techniques. Hitherto, such work with impulse voltages has not been reported in the literature. Another aspect of the work focuses on the behaviour of an SF 6 spacer system when subjected to composite stresses for example, AC and impulse. Both the long term and short term effects of 60Hz AC pre-charging on subsequent impulse fiashover is studied. Typically, GIS systems in service are subject to these com-posite stresses and their insulation performance has been unpredictable. All the above investigations have been conducted in both quasi-uniform and non-uniform field configurations. Chapter 3 examines the techniques available for monitoring charge accumula-tion on spacers and the design of the capacitive probes used in this work. Chapter 4 describes the experimental setup used in this investigation. Chapter 5 describes the numerical method used to compute the electric field distribution on the spacer surface from the measured probe data. Chapter 6 presents the results and conclu-sions with the Chapter 7 describing the scope for future work. Chapter 3 T E C H N I Q U E S FOR S U R F A C E C H A R G E M E A S U R E M E N T 3.1 Introduction Several techniques for measuring the charge distributions on spacer surfaces have evolved. Passive and active capacitance probes, electrostatic flux meters, free body probes, vibrating and biased vibrating probes and electro-optic devices are some examples. The discussions in the subsequent section will, however, be confined to the capacitance probes, field mill and electro-optic device since they are the more commonly used systems. 3.2 Electro Optic Methods The electro-optic methods exploit the change in the index of refraction of a material caused by an electric field. The material on which the charge is to be measured, should either be optically active (i.e., a material whose refractive index changes with change in the electric field), or, such a material should be embedded in the dielectric at locations where the charges have to be measured. Electro-optic methods rely on the Kerr, Faraday or Pockel effects. The relation between the index of refraction and the electric field is given by [62] An = a + j3E + ^E2 + . . . , (3.1) 44 45 where a, f3,7 . . . are constants and An is the unspecified change in the refractive index, E is the electric field. 3.2.1 The Kerr Effect The electro-optic Kerr effect may be defined by the expression: An = npar - nper = XBE2 , (3.2) where npar and nper are the indices of refraction parallel and perpendicular to the applied field E. A is the wavelength of light transmitted through the Kerr Cell and B is the Kerr coefficient. Nitrobenzene is a liquid which exhibits the Kerr Effect. A plane polarized light beam is passed through a test cell to ensure that the components are oriented parallel and perpendicular to the applied field. Upon voltage application, a phase difference is introduced between these two orthogonal components by the Kerr Effect. A polarization analyser is used to detect the phase difference. The measurement of the phase difference enables the calculation of the electric field. It has been shown [46] that the electric field may be calculated from the undisturbed background fringe spacing Ay and the fringe displacement 6y by the expression: g - e = w k ' (3'3) where 1 is the length of the light path in the cell. 3.2.2 The Pockel Effect The Pockel effect may be described by the expression: An = nlfijE , (3.4) where UQ is the ordinary index of refraction, rij is the Pockel's coefficient, and E is the applied field. 46 The Pockel effect may only be observed in solids and the direction of the ori-enting field may be transverse or longitudinal with respect to the incident beam. All pockel crystals are piezoelectric, i.e., the crystal dimensions alter with applied field [62]. This has been a drawback in constructing accurate charge measuring systems based on Pockel Effect. Electromagnetic interference, stray capacitance effects and grounding problems are eliminated in electro-optic methods and the frequency response of the measuring system is high (typically GHz). The light beam processing equipment of the electro-optic device may be located at considerable distance from the high voltage source thus requiring very little insulation from high voltages. Kerr cells have been built to measure electric fields with an accuracy of ±0.3% . They have been used for field measurements in liquid dielectrics. In contrast, electric field distributions may be measured with an accuracy of ±4.5% with measuring systems based on the Pockel The application of electro-optic techniques to measure electric fields is quite complex. The instrumentation is comprised of lasers, optical sensors, lenses, po-larizers etc. The advantage is that the charge/field measurements may be made during voltage application. On the other hand capacitive probes are intrusive and they cause a local perturbation of the fields to be measured. The following sec-tion presents the capacitive probe technique for measuring charge distributions on dielectric spacers. Figure 3.1 shows the general construction of a field mill. It essentially consists of a multivane rotor, rotating in front of a multivane sensor plate. The rotor alternately exposes each sensor blade to the ambient electric field and shields the sensor blade by the grounded rotor blade [63]. When each sensor blade is shielded, the accumulated charge on the sensor plate is discharged. The induced charge from the displacement current is given by: Effect. 3.3 Electric Field Mi l l (3.5) Sensor Plate Figure 3.1: A schematic diagram of the field mill. 48 substituting for D = Eeo in equation 3.5, q = £0 • A • e0 (3.6) The measured voltage across an impedance is related to the electric field by [63]: V = ± * f p , (3.7) and the maximum current can be expressed in terms of capacitance as: i = / • V(Cmax - Cmin) , (3.8) where A is the area exposed, Z is the impedance, f is the frequency of exposure and shielding and EQ is the static electric field. The impedance Z is typically a capacitor of about 100 pF, and an electrometer with an input impedance of lOMfi is used to measure the voltage across the impedance. It has been found that the accuracy of these probes is ±1%. In the above system, the polarity of the electric field cannot be obtained. By employing a phase sensitive circuit these instruments have been made polarity sensitive. Recently [51] this device has been used to monitor surface charge on spacers placed between plane parallel electrodes subjected to DC voltages in N 2 and SF6 environments. 3.4 Capacitance Probes A capacitance probe consists essentially of an electrode surface insulated from a surrounding grounded screen (coaxial electrode system). The central electrode is usually mounted flush with the screen. The central electrode is then connected to a high impedance electrostatic voltmeter. This device measures the charge induced on the probe surface due to electric flux density emanating from the charges on the spacer surface. Figure 3.2 shows the construction of a capacitance probe. The probe is placed a distance d from the dielectric with surface charges (density o over an effective are Ae. The dielectric has a thickness t and a relative permittivity of er. Figure 3.2 shows the various capacitances involved. Csg represents the capacitance of the dielectric to ground immediately below the probe head and Cps represents BNC Connector Probe body • Probe tip (a) (b) Coaxial cable Screen T ps Z Csg Sample Figure 3.2: A schematic diagram of a capacitance probe showing its different pacitances. 50 the capacitance between the probe and the dielectric surface. Since the spacing between the probe head and dielectric film is kept constant and is small compared to the screen dimensions i.e., d/D is small, the electric field may be assumed to be uniform. The charge induced on the probe is proportional to the voltage across Cm. It has been shown by Foord [64] that the voltage drop across Cm is related to the charge density by the expression: v m = ; c—5 r • 3.9 Cm(l + ^ + Csg) If Cm » Csg then a x Ae , It may be seen from equation 3.10 that Vm is proportional to the induced charge. An initial calibration of the probe is required to determine the ratio of Ae/Cm. The accuracy of this measurement depends on the uniformity of the sample thickness t, the uniformity of the probe to surface spacing d and on the probe scan time. It has been suggested that if the time constant RCm is made approximately 100 times larger than the scan time, accuracies of 1% could be achieved. Usually, an electrometer with an input impedance of the order of lOMfi is used for measuring the probe voltage. Due to the dominating influence of the cable capacitance on the measurement, movement of the cable can lead to large errors in measurements. The following factors have to be considered when using capacitance probes: • For good sensitivity the probe-head to dielectric distance must be kept as small as possible. A minimum spacing is, however, required to avoid dis-charges between probe head and the charged surface. These discharges could give rise to spikes on the measuring signal leading to a misinterpretation of the actual charge distribution. • The probe signal is considerably influenced by the area of charge Ae on the dielectric surface. If the probe tip dimension (d) becomes comparable to the charge area Ae, then the probe may underestimate the charge density on the small area. Reducing the diameter of the central conductor in the probe head improves spatial resolution at the loss of probe sensitivity. Such a resolution is desired when scanning narrow channels of charge produced, for example, by surface streamers. Probe tip diameters as low as 80/Lim have been used. 51 • The probe signal is dependent on the distance between probe tip and charged surface as well as on the rigidity of the probe structure and its associated cable. As a consequence, it is always preferable to keep the probe fixed and move the charged area past the probe tip. • The presence of surrounding metal surfaces close to the charged area could significantly affect the probe measurements. Therefore, the errors in mea-surements caused by the presence of metallic surfaces should be corrected for by some numerical techniques. • Finally, probe measurements do not provide us with a means to distinguish between charges on the dielectric surface and those in the bulk (volume charges) and therefore assumptions regarding the location of charges are un-avoidable. Though capacitive probes are easy to apply and require simple instrumentation, they have become the subject of serious controversy [65]. The main objection to the interpretations of capacitive probe measurements has been the procedure adopted for their calibration. Typically, a capacitance probe is calibrated using a conducting surface at a known potential. This condition implies that the surface being scanned has a uniform charge distribution. It is likely, however, that narrower lines of charge may exist on the dielectric surface to be scanned. Further, surface charges of both polarities can coexist on the spacer surface. Under such circumstances, severe field distortions at the probe tip could occur. The field distortions would cause deviations in the assumptions of a plane-parallel capacitor geometry. The equations 3.9 and 3.10 are based on the assumption that the capacitances indicated in figure 3.2 are plane - parallel capacitors and therefore the surface charge on the dielectric spacer is easily calculated from the surface potential measurement. In practice, however, these assumptions are incorrect and they lead to erroneous charge magnitudes. Further, in the presence of bulk charges in the insulating material, there is no simple relationship between the voltage measured by the probe and the surface charge on the dielectric spacer. Despite these drawbacks, there has been a noticeable increase in the use of both capacitive and field mill probes for monitoring surface charges on spacers. Several researchers [50,52], have 52 used numerical techniques to reconstruct the charge distribution from the observed probe potentials, thereby overcoming these drawbacks to a certain extent. In the present work, several capacitive probes have been described. The choice of the probe has been made after several experiments on probe sensitivity and res-olution. A probe with a shield diameter of approximately 12mm and tip diameter of 2mm has been used in the final measurements. 3.4.1 Design of the Capacitance Probes It may be recalled from Chapter 2, that the spacer geometry used in the present work is cylindrical. Since the charge growth needs to be monitored both on the top face of the cylindrical spacer (close to the high voltage electrode) and the side of the spacer, two probes have been used. One of the probes has been used to monitor the charges on the cylindrical surface of the spacer and the other probe has been used for measuring charges on the plane surface. For measurements in the plane parallel electrode configuration, the charge has been monitored only on the cylindrical side of the spacer. Figure 3.3 shows the cross-sectional view of the various probes constructed. Following voltage application, both the high voltage and low voltage electrodes were connected to ground. The gas pressure was increased by 0.04 MPa above system pressure. The maximum charge density which can be measured by the probe is limited by breakdown of the gas between the probe and the charged surface. For thin films in air at atmospheric pressure, the limit is a few 100(xC / cm2, while for thick insulators the maximum density that can be measured in SF6 at O.lMPa is approximately 80//C/cm2. By increasing the gas pressure, therefore, the dielectric strength between the charged surface and probe is increased. The spacer was lowered by operating the electrode drive and each probe in turn was brought within a distance of 2mm from the spacer surface. The rotary drive was used to move the spacer past the probe face and the probe voltage recorded. The drive enabled both updown and rotary motions about the spacer axis. The probe cable was made rigid to ensure that the cable remained immobile during the probe potential measurement. 53 A Figure 3.3: Cross sectional view of the various probes. 54 To obtain reasonable information about the charge distribution on the spacer surface using capacitive probes it is important to establish: a) the minimum de-tectable charge (sensitivity) and b) accuracy of reproduction (precision) of the spatial variation of the charge. It is apparent that the geometrical configuration of the probe sets the limit for reproducing the spatial variation of the charge dis-tribution. To establish the sensitivity and precision of the probes, the following experiments have been conducted: • The experimental arrangement shown in figure 3.4 has been used to obtain the relationship between the probe sensitivity and probe spacing. The probes are held at several positions from a charged metallic plate which is charged to 5kV by a stabilized DC power supply (stability ±1%). The probe voltage is then measured with a Keithley 610C electrometer. Figure 3.5 shows the variation of probe sensitivity with spacing for various probes. It can be seen from the results that probes numbered 5 and 6 (in figure 3.3) give the best results. The larger the spacing between the probe and charged surface the lower is the sensitivity of the probe. Reducing the probe to charged surface spacing, improves the probe sensitivity. Reducing this spacing, however, can cause sparkovers to occur between the charged surface and the probe. It is for this reason that a tradeoff between the probe sensitivity and the risk of unwanted flashover is to be made and a suitable probe - dielectric spacing selected. • The probe resolution is established using the arrangements shown in figure 3.4 a and b. Evenly spaced ridges are cut on a piece of metal as shown in fig-ure 3.4a. The width of the ridges are progressively reduced, ranging between 5mm and 0.5mm. This experiment establishes the capability of each of the probes in identifying a minimum line charge width of 0.5mm. Below this width it was seen that the probe voltage was negligible indicating that this is the minimum width of charge which the probe could resolve. A similar experiment has been conducted to determine whether two line charges of opposite polarity could be discriminated by the probe. For this test, several brass rods of the same diameter (approximately 2mm) are placed parallel to each other 55 - C A P A C I T A N C E PROBE i n fl i BRASS WIRES PLEXIGLAS Figure 3.4: Experimental setup to determine probe resolution. 56 SPACING = 2 m m SPACING = 3 m m SPACING = 4 m m 0 2 4 CHARGE WIDTH [mm] Figure 3.5: Variation of probe sensitivity with probe to surface spacing for various probes. 57 on a dielectric sheet of acrylic. The spacing between the rods is progressively reduced with alternate rods held at positive and negative voltages. It is found that when the spacing between the rods is less than 0.25mm, the probe is unable to distinguish between the positive and negative lines of charge. Thus, if line charges (such as in a streamer channel on a spacer) of both polarities coexisted within 0.5mm, the probe will be incapable of resolving this difference. From these experiments it has been established that for the present application, probes 5 and 6 are best suited to monitor spacer charge on the side and top face respectively. A probe - to - spacer spacing of 2mm has been considered to be the optimum value. This spacing is adequate from the sensitivity point of view of the test arrangement. The probes are provided with a coaxial connector to enable easy hookup to the electrometer. With this scheme, interchanging of probes is considerably simplified. Pedersen [65] has shown that the relationship between the voltage recorded by the probe and the charge density distribution on the dielectric really requires a prior knowledge of the charge distribution on the dielectric. The presence of other metal objects in the region of the field introduce additional errors in the derived charge distribution. It is therefore necessary to adopt some analytical or numerical technique by which the measurement errors could be corrected. Ootera et al. [70] have reported the development of a computer program to account for the inac-curacies mentioned above and transform the probe voltage into an equivalent 3 dimensional surface charge distribution. Using this technique, the surface charge distribution on a conical spacer in a ±500kV DC GIS has been reconstructed from the probe data. Their results show reasonable agreement with dust figures. It has been assumed that no bulk charges exist in the dielectric. In the present investiga-tion, a similar attempt has been made to reconstruct the charge distribution from probe measurements and to compute the surface electric fields on the spacer. The surface charge simulation (SSM) technique has been used in the computations. In the technique described by Ootera, the conductors and spacers are segmented into triangular patches of charge. The charge density is assumed to be uniformly dis-tributed across each triangular surface. A set of points on both electrodes, spacer and probe, equal in number to the number of charge elements are selected and the 58 Dirichlet and Neumann boundary conditions imposed. The probe potential at the ith measurement location P, is expressed as follows: where C, is the potential on the probe caused by the metal boundaries around the spacer, Oj is the charge density at jth small triangle and is a coefficient dependant on the geometry of the system. The surface charge distribution is evaluated by substituting the probe data into the equation 3.11. The method used in this thesis replaces the various boundaries by axisymmetric parabolic and conical charges instead, to simulate the geometry. The details of the technique are discussed in Chapter 5. The advantage of using this method was that for the geometry used, programming was simpler. Details of this numerical technique are discussed in Chapter 5. This technique enables a qualitative assessment of the surface charge distribution. Typically red and black powders used in electrostatic copiers are used to dust the charged insulators. The obtained pattern (with regions of red indicative of negative charges and the black regions indicative of positive charge) may be used to get a qualitative picture. This method suffers from the disadvantage that the charged spacer needs to be taken out of the SF 6 environment for dusting. Once dusted the spacer needs to be cleaned before being used again. This is a very awkward and laborious technique to merely establish the polarity of the charges. The charge distribution could be disturbed by local discharges between regions of high charge concentrations, during the process of removing the spacer from the SF 6 environment to dust it in air. This technique has been used as a secondary method to confirm probe interpretation of the charge distributions [51,50]. In summary, capacitive probes are simple to use and require no elaborate in-strumentation. They have been quite extensively used in the industry. With care, it is possible to obtain reasonably accurate estimates of the original surface charge distribution. For these reasons, capacitive probes have been used in the present n (3.11) 3.5 Dust Figure Technique work for surface charge measurements. Chapter 4 E X P E R I M E N T A L S E T U P 4.1 Introduction This chapter discusses the details of the experimental apparatus for the study of the charging of spacer surfaces in SF6 gas. The details of the electrode geometries, types of spacers, electrode and spacer polishing techniques, electric field probe assemblies and the design of the electrode drive are given. Some of the probe measurements have been obtained with a high pressure vessel in the British Columbia Hydro (BC Hydro) Research and Development Laboratory in Surrey. The major part of the measurements, however, have been obtained with the small pressure vessel in the Electrical Engineering Department of The University of British Columbia. The designs of the electrode drive mechanisms and the capacitance probe holders are different for the two pressure vessels due to the differences in the two pressure vessels. 4.2 Pressure Vessels Figure 4.1 shows the high pressure vessel (HPV) in the BC Hydro Research and Development Laboratory. It is a coaxial SF 6 GITL approximately 13m long. The inner conductor has an external diameter of 0.12m and the inner diameter of the outer enclosure is 0.36m. The HPV has two end-T sections and two 6m long straight bus sections. Conical spacers are located at each of the joints. A vertical high voltage gas insulated bushing rated for 2S0kVrms is bolted to one of the T -sections. The second T - section is used as the test chamber which is sealed 60 61 c A Vacuum pump and compressor station B Coaxial SF 6 GITL C High voltage bushing D Front flange of the test chamber E SF 6 storage tank Figure 4.1: Test arrangement at the BC Hydro Laboratory. 62 from the rest of the bus. Since the test chamber is isolated from the main S F 6 G I T L , it can be evacuated and pressurized independent of the main system. The test chamber is evacuated and pressurized using a rotary vacuum pump (Edwards Hivac Model E l M40, 40m 3 and a compressor (Corken oil less single stage D90) respectively. The vacuum pump, compressor and an SFg gas cylinder are connected to a manifold with appropriate valving. This arrangement enables: • evacuation of the test chamber to a vacuum of 8Pa; • extraction of the SF6 gas from the test chamber and compressing it into the spare gas tank (capacity 1.57m3); • replenishment of the gas in the test chamber from the S F 6 gas bottle in case of need; • storage of the used gas from the test chamber into the spare gas tank. Figure 4.2 gives the schematic of the gas cart. A thermocouple gauge [Varian 531] calibrated against a Mcleod gauge has been used to monitor the in-line vacuum at the pressure vessel. High pressure dial gauges indicate the gas pressure in the test chamber. Three ports of varying sizes and two additional observation windows, approximately 75mm in diameter, are provided on the test chamber. The flange at the front end (D in figure 4.1) is used for mounting the electrode drive assembly. The test chamber can be evacuated to a vacuum of 8Pa. and pressurized to a maximum gas pressure of 0.4MPa. Figure 4.3 shows the arrangement of the electrode drive and the electrodes mounted on the high pressure test chamber. Figure 4.4 shows the small pressure vessel which is provided with four port holes. Two of the port holes are utilized to manipulate the low voltage electrode from the outside. A flexible Neoprene rubber hose connects the third port to a 250 lit/min two-stage Sargent Welch rotary vacuum pump. A high pressure dial gauge and a thermocouple gauge head are also connected to this flange in order to monitor both the S F 6 gas pressure and the vacuum during pressure cycling. A Tygon tubing connection to the same flange with a needle valve allows S F 6 gas from the gas bottle into the chamber. The test chamber is fitted with a bushing rated 150kVrTn3. 63 To lest chamber Ail vent To storage tank To SF f c bottle Figure 4.2: Schematic of the gas cart 64 1. High Voltage electrode 2. Capacitive probes 3. Low voltage electrode 4. Rotary and linear drives Figure 4.3: Arrangement of the electrode drive and probe in the high pressure test chamber. 65 2 4 1. Pressure vessel 2. Rotary drive 3. Linear drive 4. Observation window 5. Multi-test set Figure 4.4: Test arrangement at the UBC laboratory. 66 4.3 Electrode Assembly The plane parallel electrodes were made of aluminum and were approximately 150mm in diameter and 30mm thick. One pair of electrodes had a recess approx-imately 45mm in diameter and 2mm deep. The recess was provided to shield the triple junction. The plane parallel electrodes were 150mm in diameter and 30mm thick. The hemispherically capped rod electrodes were made of brass with tip diameters of 5mm and 10mm. The HV electrode was firmly fixed to the central conductor in both the pressure vessels and the low voltage (LV) electrode was supported by an electrode drive mechanism. After each test series, breakdown tracks on the electrode surfaces were removed by the following procedure: • The electrodes were remachined. • The electrodes were then wet polished with 600 grit paper. • Subsequently they were polished with 1/z alumina powder. If the electrode damage was extensive, it was necessary to commence the polishing with a 400 grit paper. 4.4 Electrode Drive Assembly The electrode drive system enabled both an up - down motion as well as a rotation of the LV electrode about its own axis. It was operated by reversible DC motors located external to the test chamber. By using nylon shafts, insulation was provided between the bottom electrode and ground. The LV electrode was insulated from ground for 100 volts so that, by connecting a resistor between the LV electrode and ground, the gap currents can be monitored. Dynamic oil seals incorporated in the flanges provided the necessary sealing both under vacuum as well as at high pressure. Linear potentiometers (Piher 10kf2) were connected to the drive shaft which facilitated the recording of the angular and linear movements of the Figure 4.5: Electrode drive arrangement in the small pressure vessel. 68 electrode. Limit switches restricted the total angular travel (360° ±2° ) and the linear up - down motion (20mm ± 1 mm). Figure 4.6 shows the drive arrangement for use in the high pressure vessel. Figure 4.7 shows the cross-sectional view of the flange and the drive mechanism. Both the rotary and linear drives have been located on the same flange. Once again, dynamic oil seals have been used to seal the drive shaft feed- through in the flange. These seals proved to be adequate for a vacuum of 8Pa. and gas pressure of 0.4MPa. A linear bearing incorporated in the central section of the flange converts the rotary motion of the motor shaft to a linear up - down motion of the LV electrode. Limit switches have been set to give the desired travel. A linear potentiometer (Piher 10kf7) provides information on the linear spatial movement. The rotary motion has been provided by a gear drive. In this case however, the information on rotary movement has been provided by a three and a half turn rotary potentiometer. The LV electrode is supported by a shaft which is aligned by the linear bearing in the electrode drive (see figure 4.7). After each test series, the linear drive was used to lower the LV electrode (i.e. spacer was moved away from the high voltage electrode) to permit the charge scan of the top face of the insulating spacer. At each position, the rotary drive was operated to rotate the spacer in a clockwise direction (looking down on the spacer) past the probe tip. After completing the scan of both the spacer top and the sides the linear drive was operated to move the spacer back towards the high voltage electrode (as it was during the voltage application). The output of the potentiometer was fed to the X-channel of a YEW X-Y plotter. Power for the for the motor operation and the potentiometers were provided by adjustable DC power supplies. 4.5 Capacitance Probe Assembly Figure 4.8 shows the cross-sectional view of the probe cable and the seal. The probe cable has been drawn through a 10mm diameter, 300mm long stainless steel tube. Adequate dynamic sealing has been provided to prevent gas leaks during the movement of the probe assembly. Silicone sealant provides sealing against leaks 69 Figure 4.6: Electrode drive arrangement in the high pressure vessel. 70 Figure 4.7: Cross-sectional view of the electrode drive used in the high pressure vessel. 71 from the core and sheath of the RG58 cable used. The experimental procedure adopted for the measurements consisted of the following steps: • The insulating spacer is placed on the low voltage electrode after cleaning. • The pressure vessel is then evacuated to a vacuum of 8Pa and then filled with SFQ gas to the desired pressure. • The lower electrode is moved away from the high voltage electrode until it reaches the farthest (lowermost) position from the high voltage electrode. The capacitive probe (6 in figure 3.3) is then moved towards the spacer and set to scan the flat face of the spacer at 5 levels as shown in Fig 4.9. At each position (levels 1-5), the rotary drive is operated to rotate the spacer past the probe, thereby obtaining a 360° scan. The probe voltage is measured using a 610C Kiethley Electrometer and plotted on a YEW X-Y plotter at each level. • This probe is then withdrawn and the second probe (5 in figure 3.3) to scan the cylindrical side of the spacer is moved inwards. A similar scan is con-ducted. • The lower electrode is then raised (moved towards the high voltage electrode) by the drive so that a 360° scan can be performed at 4 different levels (levels 6-9) as shown in figure 4.9. • After the scans, the probes are withdrawn fully. • The bottom electrode is raised so that the insulating spacer is firmly fixed between the two electrodes. • The voltage is then applied. • Following the voltage application, the gas pressure is raised slightly and the LV electrodes is lowered. Probe voltage scans are then performed as described above. Silicone sealant Adapted BNC connector(female) Stainless steel tube Figure 4.8: Cross-section of the probe cable. 73 QUASI-UNIFORM FIELD NON-UNIFORM FIELD CONFIGURATION CONFIGURATION Level Number Distance x in mm Rod-plane gap Parallel-plane gap 1 0.5 3.5 2 5.0 8.5 3 10.0 13.5 A 15.0 18.5 5 19.5 6 23.5 7 28.5 8 33.5 9 38.5 Figure 4.9: Levels at which the residual charge measurements were made. Distance x is the measure of the charge scan radius from the axis along the spacer surface. Figure 4.10: Schematic of the charge recording setup. 75 4.6 Power Supplies The Haefely Multi-Test system has been used as the voltage source for the ex-periments in the small pressure vessel. The system generates 190kV lightning or 180kV switching impulses or 200kV DC or 80fcVrma voltages. A capacitive voltage divider (CVD, 500pF) has been used to measure the high voltage. The voltage signal across the secondary of the CVD has been connected to an impulse peak voltmeter (Haefely Type 65) and also to a Biomation waveform digitizer (Type 6500) for recording the signal. The trigger signal for the Biomation unit has been derived from the trigger spark gap of the impulse generator. A fiber - optic link has been used to lead the light of the spark gap discharge into a light - to - volt-age conversion unit. This unit provides a clean trigger signal for the Biomation unit. Improved shielding of the laboratory, use of triax cables, proper grounding are some of the steps taken to overcome any jitter in the triggering of the Bioma-tion unit. A flexible aluminium lead has been used as the connector between the impulse generator and the test chamber. The high voltage transformer provided in the multi-test has been rated for 80 K^msj 5kVA . The supply to the low voltage winding is regulated by a variac and the high voltage output recorded by the peak voltmeter connected to the secondary of the CVD. For the measurements at BC Hydro, a 500kV five-stage impulse generator has been used. Various combinations of front and tail resistors are available to generate different waveforms. The charge measurements at BC Hydro have been obtained with the 1.2/50 /us and 100/2500 fis waveforms. The impulse voltage has been measured with a 500pF SF 6 gas insulated CVD having a division ratio of 1:9958. 4.7 Insulating Spacers Figure 4.11 shows the cylindrical spacers used in this investigation. All samples have been carefully machined to obtain surface finishes of 4/mi and 40/xm to exam-ine the influence of surface finish on the charging phenomenon. The spacers have been made of PTFE, acrylic and unfilled epoxy, and were 40mm in diameter and 76 Figure 4.11: Photographs of the insulating spacers 77 30mm long. The surface finish has been confirmed by SEM scans performed on small samples machined under identical speed-feed ratios. 4.7.1 Selection of a method to neutralize surface charges With voltage, insulator surfaces in gases collect charge when the rate of charge arrival exceeds the rate of removal from the surface by conduction or other neu-tralizing mechanisms. Such accumulation is dependent upon the source of free charges, transport of charges to the surface and the charge neutralization processes. When drifting charges accumulate on the spacer surface, they alter the electric field thus affecting the flashover behavior. It is important to have a relatively charge free spacer before the commencement of voltage application to obtain the exact nature of charging. Therefore, it is necessary to find means of neutralizing the charges on spacer surfaces at the beginning of each test series. Some experiments were conducted to obtain a quick, simple and an effective method for minimizing residual charges on the spacer before introduction into the pressure vessel for the high voltage tests. Neutralization of surface charges by infra - red radiation and thermal radiation have been investigated with thin insulating disc samples before establishing a procedure for charge neutralization between tests. Figure 4.12 shows the experimental arrangement. All these experiments have been conducted in air. A needle electrode has been used to charge disc specimens of acrylic, P T F E and ceramic placed on a rotating metallic platform. The samples were cleaned and washed in distilled water and dried in a vacuum oven at 60°C for 4 hours, then wiped with alcohol and placed on a metallic platform. The point electrode was then connected to a DC supply and maintained at 3 kV for 1 min. This caused the formation of a corona discharge at the tip, resulting in charge deposition on the spacer surface. The surface was then scanned by the field probe at the same radius as that of the corona source. Surface scans were then obtained following an infrared lamp exposure, using a hot air blower and a 1000W lamp. It may be seen from the charge profiles, shown in figure 4.13 that the 1000 W lamp and the hot blower give the best results. These measurements establish 78 Electrode drive 1. High Voltage DC supply 2. Electrometer 3. X-Y plotter 4. Corona source 5. Capacitive probe 6. Turntable (LV electrode) 7. Support stand 8. Blow drier Figure 4.12: Experimental setup to establish a suitable residual charge neutralizing technique. 79 that charge neutralization is effective when the surface temperatures of the spacer reach 70°C approximately. Although the blower is equally effective, the lamp ar-rangement has been preferred. Before introduction of the spacers into the pressure vessel, they are cleaned and discharged by adopting the following procedure: • Ultrasonic cleaning with Alconox detergent for 20 minutes. • Rinsing in plain tap water. • Thorough rinsing in deionized water. • The samples are then heated in an oven at 65°C for 4 hours and stored in an evacuated dessicator. • Prior to their introduction into the pressure vessel, they are washed with isopropanol and any trace of dust is wiped clean with Kleenex tissue dipped in isopropanol. • The sample is finally placed on a stand and exposed to a high intensity (1000W) lamp for 5-8 minutes to minimize any residual charge on the spacer. Spacers and electrodes are always handled with disposable plastic gloves. The electrodes are wiped clean with Kleenex tissue dipped in isopropanol before being introduced into the pressure vessel. Subsequently, the spacer is introduced into the test chamber and the test chamber is immediately evacuated. Charge scans performed under vacuum following the above treatment have indicated that the residual charge on the insulator is of the order 0.1 nC/cra 2 . S F 6 gas is introduced into the test chamber following its evacuation to 8Pa. At times, traces of residual charge greater than 1 n C / c m 2 has been neutralized by the gas. This has been demonstrated by the charge scans performed after the introduction of the S F 6 gas. 80 Blow drier 6-4° 0.16 pC/m* Infra-red lamp -0J6 p/m1 , a B p l e 1000 W lamp •0.03 Charged aample -3.73 ^Cjrrf-1000 H lamp Charged sample -0.36 pC./^ CERAMIC 55.0 aim d ia . 9 .0 an thk. PLEXIGLAS 77.0 an d ia . 3.6 am thk. TEFLON 77.0 an d i a . 3.3 an thk. Figure 4.13: Effect of using various techniques to neutralize residual charge on a small disc spacer. Chapter 5 E L E C T R I C FIELD C O M P U T A T I O N 5.1 Introduction The surface charge distribution on the spacer surface should be derived from the ca-pacitive probe measurements which are in the form of the probe potential variation as a function of its position. Either a computational technique, like for example a surface charge simulation technique, or calibration against a known charge distribu-tion is used for the transformation of the probe potential variation to an equivalent surface charge charge distribution. The surface charge simulation technique is used for the conversion of probe data to surface charge distribution. The principle of the technique and the approximations used in the analysis, have been described. It was mentioned in chapter 3 that capacitance probes are normally calibrated against a charged metallic surface held at a known potential V volts. If Vpr is the the voltage indicated by the electrometer connected to the probe held at a distance d from a charged metallic surface held at V volts, the electric field between the probe and the charged object is E = ^ = f^. Hence the charge density on the metal object is a = l u y ^ , where e0 is the permittivity of free space. The probe potential is related to the charge density as Vpr = aa, where a is the probe transformation function. Substituting the value of a in the expression for the probe potential Vpr, the probe transformation function a is obtained. Using this probe transformation function a, the surface charge distribution on an insulating spacer is determined. The electric field in the gas near a charged spacer (whose relative permittivity is er) is modified when a metal object is brought near the surface charge. This is 81 82 because the electric field is concentrated in the gas layer between the spacer and the metal. The charge density due to a local charge AQ, on an area AAi is defined as a = The electric field between the charged area and the metal object(the probe in this case) is given as E = 7 ^ - where ej of the gas is unity. It is assumed that this charged patch resides on the surface of the dielectric. When the probe is withdrawn from the charged area the electric field is reduced to E = — T T X — • The above interpretation implies the following assumptions: • a plane parallel electrode configuration, • the absence of other metal objects, • the absence of any bulk charges • the absence of both positive and negative charges coexisting on the spacer surface. • the effects of the material can be neglected (specifically the relative dielectric permittivity) Pedersen [73] has shown that the probe response function for probe measurements with a simple geometry (similar to figure 3.3) depends on the relative permittivity of the insulating spacer, the relative dimensions of the entire system and the probe position. Further, meaningful interpretation of the surface charge distribution on the insulating spacer from probe measurements are possible if these factors are considered. The surface charge simulation technique takes into consideration the presence of the insulating spacer and the electrodes. It is more accurate than the calibra-tion of the probe against a charged metallic surface because a systematic error is introduced in such a calibration by the internal molecular polarization which is present in the dielectrics with residual charges. The principal assumption made in this analysis is that no charges exist in the bulk of the specimen following the impulse voltage application. This is a reasonable assumption since the volume con-ductivities of the spacer samples are of the order of 6 x 10 - 1 6f2 - 1/cm and the time constants for charge migration are of the order of 60 hours. 83 Computational techniques may be broadly classified into differential and integral methods. The Finite Difference (FDM) and the Finite Element Methods (FEM) are differential methods and charge simulation methods are integral methods. The F E M and FDM are also referred to as the Domain methods and CSM and SSM as the Boundary methods. The SSM technique has been chosen as the computational procedure for the following reasons: • The F E M technique requires the generation of isoparametric elements, due to the 3-dimensional nature of the problem, thereby, increasing the complexity in programming; • The axisymmetry in the model enables the use of conventional ring-like sim-ulation charges for which the potential and field coefficients could be easily derived. Any asymmetry due to the spacer surface charges may be accounted for by selecting a charge distribution which periodically varies with the angle of rotation around the axis of symmetry; • The electric field distribution on the electrode surface may be obtained di-rectly from the surface charge density using the Guass theorem: = f . (5.1) Although a finite difference technique has been developed to asses the influence of the charge on an insulating spacer from probe measurements on the overall elec-tric field distribution, the study is restricted to a 2-dimensional problem [74]. The authors have directly used the charge density distributions obtained by previous investigators to compute the electric field distribution of the system under the in-fluence of the charges. However, the accuracy of the original charge distribution magnitudes and the technique used to interpret the probe data remains uncertain. 5.2 Principle of Charge Simulation Method The distributed surface charges on the conductors, are substituted by a set of dis-crete charges which are placed inside the region in which the field is to be computed (for example inside the conductors). The choice of the type of charges (i.e., line, 84 point or ring charge) depends on the electrode configuration. The magnitude of these charges are then determined such that the boundary conditions are satisfied exactly at a selected number of points on the boundary. The accuracy of the CSM improves with the number of simulation charges. The computational procedure involves the following steps for a geometry not involving any solid dielectric: • Locate n (point, line or ring) charges inside the electrodes, • Selecting a set number of n contour points on the boundaries. It is required that at the contour points, the potential resulting from all the the charges should be exactly equal to the electrode potential. Application of this prin-ciple to the ith contour point on an electrode results in an equation of the form given by: Efli9y = *<, (5-2) where Pij is the potential coefficient at the ith contour point due to all n charges (i.e., j = 1,2,3 ri). The potential coefficient is typically a metric coefficient involving distance terms. Similar equations are obtained at each contour point, leading to a system of n linear equations: = [P] [Q] = [•] • (M t'=l j=l The P matrix in equation 5.3 is often referred to as the potential coefficient matrix. The vector Q is the array of unknown discrete charges and $ is the vector of prescribed potentials (Dirichlet condition) • Solve the system of equations 5.3 to obtain the magnitudes of the discrete simulation charges. • Having obtained the magnitudes of the simulation charges, it is necessary to check the potential at points on the boundaries other than those used in step2. As a measure of accuracy, Singer [68]et al. define the difference between the known electrode potential and the value computed by CSM as the potential error. It has been shown [68] that in order to obtain a 1 % 85 accuracy for the field magnitude in the region of interest, it is necessary to limit the potential error to less than 1%. • With the potential error checked, the field at any other point can be computed in the x and y directions for a 2-dimensional problem. For example the electric field in the x direction at any point i due to all the charges j = 1,2,3...n can be written as: and in the y direction Equations 5.4 and 5.5 can be expressed as E = [£,niQi+ (5-6) 3 = 1 3=1 Ffj and F^ are the called the field coefficients in the x and y directions. 5.2.1 C S M Applied to Two Dielectric Arrangement Due to the realignment of the dipoles in a dielectric exposed to an electric field, the charges internal to the dielectric cancel each other. A net surface charge however, appears on the dielectric surface. The dielectric may therefore be simulated by locating discrete charges on either side of the dielectric boundary [68]. It should be noted that: • The dielectric boundary does not represent an equipotential surface; • The electric field should be calculated on both sides of the boundary. The usual boundary conditions for the tangential and normal components of the electric field should be satisfied. 5.3 Principle of Surface Charge Simulation In the SSM, the distributed surface charges on the conductor and dielectric bound-aries are substituted by sheet-like charges which are positioned directly on these 86 boundaries. The magnitude of these charges are such that the boundary condi-tions are exactly satisfied at a selected number of points. The Dirichlet condition is used for the electrode boundaries and the Neumann condition at the dielectric interface. For fields which do not have a rotational symmetry, ring charges whose density varies with angle of rotation a (a Fourier-type surface charge) are used as the simulation charges. A charge distribution given by [71]; n n A = ^ A ^ c o s ^ a ) + ^AMsin()ua) , (5.7) may be used. In equation 5.7, n represents the order of the harmonic charge distribution and A is the charge density (charge per unit length). The expression for the potential at any point P(rp,tp, zp) (in cylindrical coordinates) due to the mth harmonic charge whose coordinates are Q(rc,0,zc) is «Mr„ zp) = £-£fipQ*-i t ( ' P ' 2 c ) 2 r t c r c 2 + r p 2 ] C O S ^ ' ( 5 " 8 ) considering only the cosine function in the charge representation. ...j is com-monly referred to as the torus function. Several methods have been used to solve this function but a matrix method described by Sato [72] is used in this work. In reference [71] it was shown that 3-dimensional axi-symmetric fields like a sphere-plane geometry could be computed by using charges in the form of cones placed directly on the sphere surface. The effect of the ground plane was simulated by considering the image charges. The charge density around the axis of symmetry is similar to the equation 5.7 except that charge density variation along the conical surfaces could be expressed as a function of a non-dimensional parameter u which adequately describes the contour. This choice of a charge element essentially con-stitutes a surface charge representation. Both linear and parabolic expressions have been employed for defining the charge density variation along the u direction (see figure5.1). For contours with circular arcs, r c and zc are trigonometric functions of an angle. Thus a surface charge element can be represented as n n Hu) = HA^(u)cos(^) + EM")8"1^) , (5-9) where A(u) indicates that the charge density varies along the u direction. In order to generalize the application, the charge density was considered variable in the xj) 87 direction as well. The potential at any point F(rp,ip,zp) (see figure5.l) due to the mth harmonic surface charge element, described by equation 5.9 is given by: i t , -v cos(m?/;) f \FC (zp- zc)2 + rc2 + rp2 <Pm{rp,ip,zp) = x / A m ( u ) J - g m _ i [ ^ '-— p-\du , u * i s i n( mV0 w /" w x f ( z p - z c ) 2 + r e 2 + r p 2 + - ^ x / A - ( u ) V ^ g m " [ — ^ — 1 ( } u V P c The potential at a point i due to all the harmonic charges is given by; n = £ $ m • (5.11) In a similar manner the electric field in the (r, ip, z) directions can be computed [72]. In equation 5.10, the coefficients obtained after excluding the charge term is referred to as the potential coefficient as in charge simulation. It is quite apparent that this expression for potential coefficient is a metric term just involving the various distances between the charge location and the point at which the potential is being computed. In a similar manner, the field coefficients are obtained by differentiating the expression for potential in the {r,tb,z). Contours may be described by polynomials; for example a sphere is replaced by parabolic elements [71] and other axi-symmetric 3D surfaces are represented by conical, cylindrical and toroidal segments [72]. It is apparent from the above that for a 3-dimensional axi-symmetric field problem with a CSM technique a very large number of charge rings will have to be used to solve for the field distribution. In the case of the SSM, however, by using relatively few area charges as given by the equation 5.9, the field distribution can be computed. Depending on the complexity of the problem, the use of SSM will speed up the computation. Sato and Zaengl [72] have used cylindrical and toroidal segments to compute the electrostatic field distribution for a 3D problem consisting of 3 hemispherically capped metal electrodes and a dielectric sphere placed above a ground plane. For this problem 3 harmonics and 54 area charges were used. The ground plane was simulated by using image charges. Therefore, a charged axi-symmetric surface c:[rc, 0, zc] has been defined by means of the angular parameter 0(0 < 0 < 2n) and a nondimensional unit parameter u(0 < u < 1). By this representation, cylindrical, conical, spherical and toroidal charged surfaces are easily described. Figure 5.1 88 Figure 5.1: Representation of conical and spherical surface charges. 89 illustrates these surface charge elements. The coordinates for the charged conical and toroidal surface are expressed by equations 5.12 and 5.13, respectively. r c = i2i + u[R2 - Ri] , zc = Zi + u[Z2 -Zt] 0 < u < 1 , (5.12) and rc(u) — pcos0(u) , zc(u) = ps'm0(u) , 0{u) = 6X + u[02 -h] 0 < 0 < 2TT . (5.13) Spherical and hemispherical charged surfaces may also be expressed by substituting the following identities into equation 5.13: p = 0 0i = —TV 02 = 7T 0 = 0 0! = O 02 = 7T A variety of functions may be used to describe the surface charge as a function of u. It is reported that a Chebyshev polynomial series gives good results [72]. Depending on the choice of the value for n in equation 5.9 the number of unknowns and the complexity increases. The choice of n also determines the number of evaluations of the toroidal function. These functions are quite involved. A simplified matrix method for evaluating the toroidal functions has been described by Sato and Zaengl. The programming effort is considerably reduced by this technique. 5.4 Present Problem In this study, the high voltage electrode of the rod-plane gap has been represented by 3 circular segments to simulate the hemispherical tip and by 4 straight cylindri-cal segments to simulate the straight section of the electrode. The bottom electrode has been simulated by 4 disc type charges for the fiat section and by 2 parabolic segments to simulate the curved edges. The value of n in equation 5.9 is chosen to be 3. The vertical part of the cylindrical insulating spacer is simulated by three segments and the top fiat section of the insulating spacer by 4 disc charges. Thus a total of 18 surface charges with a value of n = 3 would lead to 162 unknown magnitudes of harmonic charges (neglecting the sine terms due to the symmetry of II. W. 73-99 PS High Voltage Electrode 16 It l< 13. II. 196-162 • 9. • • - e, —| • Low Voltage Electrode 1-45 It. 17. f i . *• I. •A Figure 5.2: Charge and contour point location for the divergent field geometry. 91 the problem) Therefore, 162 contour points are selected. Figure 5.2 illustrates the schematic of the surface charges and contour point locations. Contour point 1-45 are located on the low voltage electrode, points 46-99 on the high voltage electrode, points 100-135 on the insulating spacer top and points 136-162 on the cylindrical side of the spacer. The potential of the points located on the electrode surfaces are known that is, the Dirichlet boundary condition is applicable. On the other hand the Neumann conditions are applicable to contour points on the spacer surface. If -E'i(P) and JE^P ) denote the fields in regions 1 and 2 with relative dielectric constants ei and e2 on either side of the point P located on the dielectric surface, then applying the Neumann condition results in: o-r = {z2-el)t Fdu+ (V e 2 )% J Z€o (5.14) F is the field coefficient for the field normal to the surface defined as the electric field at point P due to a charge j, ar is the polarization charge and 6tJ is called the Kronecker's delta. The potential <p\ at any point F(rp,ip,zp), due to all the surface charge elements (1-162, which includes all the harmonics) using only the coefficient of charge in the cosine terms in equation 5.10 is (5.15) where qi,q2,--- a r e the unknown charges and P l i i , P l i 2 are the potential coef-ficients at points 1,2 due to charges 1,2 respectively. For contour points located on the dielectric surface the criterion described in equation 5.14 is applied. Since points 100 — 136 are located on the top surface of the cylindrical insulating spacer the field in the z direction is used to define the field coefficients and for points 137 — 162 the field in the r direction is used to define the field coefficient. Applying these criteria to all the contour points would lead to a system of equations, (<t>\\ Ph.l Pll,2 • Ph,2 • -Pll,162 ^ ^12,162 ( ^ \ ?2 0\ -F^lOO.l ^100,2 - FZ100IIQ2 0~37 F 7*136,2 -^ 7*136,162 \0-63 J . . . •P1r162,1162 / V ?162 / (5.16) 92 The formulation of Fz and Fr are similar to those described in [72]. It is apparent that voltage vector on the left hand side of equation 5.16 is not completely known, because o~\ through <763 charges on the dielectric are unknown, (i.e., for points 100-162 on the dielectric surface the surface charge density is not known). Using the equation 5.16 the charges <fr (q\ through <7i62) can be expressed by 99 63 3=1 k=l In equation 5.17 the first summation term may be calculated whereas the second summation term is unknown. To determine these unknown surface charges on the dielectric surface a new set of contour points (63 in this case) are then chosen at which the probe potentials are measured after voltage application. The probe potential at any new point P due to all the charges can be expressed as: <t>\ = E p*n*j • 3 = 1 (5.18) Substituting equation5.17 into equation 5.18 will give a system of equations which describe the known probe potentials as a function of the unknown charge distribu-tions both the dielectric and the electrode surfaces which are expressed as 2 2 ( P*i,i Wis) P2l,2 P 2 2 2 \P2, 63,1 -P21162 -P22 162 -P263 X62 / X Pl2,l PU.2 • PU.2 • Pll,162 ^ -P12.162 -1 (*\\ <t>\ P^ioo.i FzWo,2 F0ioo,162 ^136,1 0-37 • -^162,1162 J V<763 J (5.19) Once again the terms P 2 i i etc., are the potential coefficients due to all the un-known simulation charges including the desired residual charges on the dielectric surface, at the new points at which the probe potentials have been measured. Us-ing this equation 5.19 the dielectric charges Oy to <763 are calculated. These values are then backsubstituted into equation 5.16 and the overall charge distribution is 93 obtained. Finally, the voltage and the electric fields in the three directions (r,V>,z) are computed. Figures 5.3 and 5.4 show the overall field distribution on the insu-lator surface prior to the voltage application and subsequent to the impulse voltage application, respectively. It is important to select the contour points such that the boundary conditions can be applied. It is also important to select these points to prevent the P matrix from becoming singular. For the geometry described by Sato et al., [79] to keep the field errors low, the first set of 163 contour points should be selected such that they are symmetrically distributed around the electrodes and spacer in the ip direction and located at the zeros of the Chebyshev polynomial in the u direction. In summary, a numerical method which takes into account the presence of the dielectric and metallic objects has been developed. The technique also enables the computation of the surface electric field on the spacer surface from which com-parisons could be made between the fields magnitudes due to the applied voltages and the corresponding magnitudes due to surface charges acquired by the spacer following impulse voltage application. Such comparisons should establish whether the electric fields due to the surface charges significantly distort the the main field (during voltage application) and cause anomalous fiashover. Figure 5.3: Initial tangential field distribution on a spacer during impulse voltage application prior to the charge accumulation on the spacer. 95 Figure 5.4: Surface field distribution on spacer following impulse voltage applica-tion. Chapter 6 RESULTS A N D DISCUSSIONS 6.1 Introduction In this chapter, the experiments to investigate the nature of charge growth on spacers in S F 6 gas under impulse and A C voltages are presented. The repeated application of impulse voltages to a spacer gap in S F 6 gas and its effect on spacer fiashover voltage and time has been studied with the aid of capacitive probes. The effects of A C pre-charging of spacers on the subsequent impulse fiashover voltages are also included. As mentioned before, both plane-parallel and rod-plane geometries have been used in these experiments. This particular choice of electrode geometry served two purposes: (a) the axis of symmetry permitted easier measurement of surface charge on the spacer and also made it amenable for simpler field calculation, and (b) they were representative of both quasi-uniform and non-uniform fields in service. By altering the rod diameter (5mm and 10mm), the degree of non-uniformity could be selected. Three different insulating spacer materials were used namely, unfilled epoxy, ptfe and acrylic. These materials have different surface and bulk resistivities and the effects of these parameters on the charge accumulation and dissipation have been studied. Insulating spacers were also made with both smooth(4/zm) and rough(40//ra) surface finish to examine the influence of surface finish on the charge growth. Two different capacitive probes were used to measure the surface charge dis-96 97 tribution on the spacer. In case of the rod-plane spacer gap, both the top and cylindrical sides of the spacer were scanned. Physical limitations prevented charge measurements of the spacer from being performed in the region close to the low voltage electrode (approximately upto 12mm above the low voltage electrode). In case of the plane-parallel gap however, the top face of the spacer (facing the high voltage electrode) was not scanned. As described in chapter 4 the probe registers the potential at its tip due to surface charges on the spacer. Most of the results on charge accumulation are presented as 3D plots. The X and the Y axes indicate the radial distance from the tip of the high voltage electrode at which the probe potential was measured. The Z axis indicates the magnitude of the probe potential as a result of these charges on the spacers. Switching, lightning and AC voltages were used in the experiments. The re-spective voltage waveforms were 100/2000/us, 1.2/50/xs and 60Hz. Most of the experiments presented in the following sections were conducted at room temperature (20° C ) and an SF 6 gas pressure of O.lMPa. Limited exper-iments were done at higher gas pressures of 0.2MPa and 0.3MPa, under impulse voltages only. The measured currents and the distributions of the inception time are presented first. The effect of the spacer material, surface finish and gas pressure on the corona inception times are discussed. This is followed by an examination of the nature of charge growth on spacers in SF 6 subjected to higher voltages. The influence of the material, its surface finish and electrode geometry is also discussed. Finally, the effect of these accumulated charges and the associated electric field on impulse flashover voltage and time are1 presented and discussed. 6.2 Pre-breakdown phenomenon Figure 6.1 shows the tangential surface electric field distribution for a rod-plane gap with a cylindrical insulating spacer placed in the gap. The X-axis shows the distance (in cm) from the tip of the high voltage electrode along the insulator sur-face and the Y-axis is the tangential surface electric field in kV/cm, normalized with respect to the applied voltage. As expected when a dielectric spacer is placed 98 in the gap, the electric field distribution in the region close to the high voltage electrode gets altered depending on the dielectric constant of the spacer and the diameter of the rod electrode, keeping the other dimensions unaltered. For com-parison, the electric field distribution for a plain rod-plane gap along the imaginary boundary of the spacer is also included. These calculations were performed using a charge simulation program developed on the theory described adequately by [68]. It is apparent that the presence of an insulating spacer with no surface charges in the gas gap alters (increases) the electric field distribution when compared to a plain gas gap. It is of interest to examine the perturbation of this symmetric field distribution due to surface charges accumulated by the spacer under different voltage application. The following section describes the effect of the spacer on pre-breakdown cur-rents in a rod-plane spacer gap under impulse voltages. The nature of pre-breakdown currents (corona pulses) and the distribution of their inception times, should aid in understanding the role of the spacer in the pre - breakdown current growth. 6.2.1 Pre - breakdown currents Pre - breakdown currents have been measured both with and without spacers in a rod-plane gap under positive switching impulse voltage application. Figure 6.2b shows a typical oscillogram of the current in the gap following a positive switching impulse voltage application to a plain gas gap. The pre - breakdown current was measured as described in section 4.4. Figure 6.2a shows the oscillogram of the applied lightning impulse voltage pulse and the corresponding gap current. Figure 6.2c shows the gap current for a spacer gap. For a plain gas gap, at low voltages, a single corona pulse was observed. This signals the formation of the streamer in the region near the rod electrode. With increasing amplitude of the applied voltage, the number of pulses increased and the duration between the subsequent pulses decreased. The corona pulses then decay due to the cessation of the ionization activity. This is due to the decrease in the space charges left behind by the previous avalanches. If the voltage is increased it clears the space charge quickly, restoring the original electric field distribution. Fresh avalanche activity is initiated. 99 ELECTRODE RADII = 0.8 and 5mm Surface field with spacer e r = 2.0 Rod radius = 0.8mm Surface field with spacer e r = 8.0 Surface field with spacer e r = 2.0 Field along imaginary boundary Plain gas gap 1 2 3 4 DISTANCE FROM T H E HV E L E C T R O D E [cm] NOTE: Surface f i e ld is tangential to the spacer surface Figure 6.1: Surface electric field distribution in a rod-plane gap with an cylindrical insulating spacer. In case of a gas gap the field calculation is along the imaginary boundary of the spacer 100 -. - f \ - l U i i u > a c. Corona pulse rise time 0.9 us Figure 6.2: Current and voltage oscillogram (a) lightning impulse voltage (+52kV) and current (48.2mA) :(b) switching impulse displacement current and corona cur-rent (rise time 1.5/xs) current:(c) same as in (b) but with spacer (rise time 0.9/is) 101 In case of a spacer gap, however, the corona pulse had a faster rise time. With increasing voltage the number of corona pulses increased and the time interval be-tween the pulses reduced. The successive pulse heights either increased or decreased in magnitude in relation to the first pulse perhaps indicating that the local fields setup by the charges deposited on the spacer influenced the streamer propagation. With higher voltages, the subsequent pulses increased in magnitude in relation to the first pulse ultimately leading to breakdown. Figure 6.3 shows the mean value of the peak corona pulse current (in mA) measured as a function of the applied voltage for two different materials at a pressure of O.lMPa for a 5mm diameter rod. It can be seen that the current gradually increases at low voltages, but at higher voltages however, the increase is relatively steep. This may be attributed to the surface charges deposited on the spacer which are perturbing the local field distribution. Further, in case of ptfe, the currents are the lowest in comparison to the other specimen suggesting lower charge deposition on the spacer surface. This will be discussed in light of the surface potential measurements made at these voltages. Another observation that has been made with regard to the corona pulses is that their amplitudes reduce with increasing gas pressure. This is attributed to the higher charge magnitudes deposited on the spacer near the rod electrode which is diminishing the electric field and thus inhibiting the current flow. No significant differences were observed for acrylic spacers with different surface finish (smooth and rough finish) with regard to the current growth as a function of the applied voltage. This suggests that for the choice of the surface finish of spacers made in these experiments its effect on current growth as a function of the applied voltage is minimal. Typically, insulating spacers in GIS have a surface roughness of 4 — 5/tm. The mean value of the corona inception times from inception voltages upto those close to breakdown are shown in Fig. 6.4 for epoxy, acrylic and ptfe spacer gaps. The corona inception time is defined as the time interval from the virtual zero of the applied impulse wave to the appearance of the first corona spike on the current wave. In case of the acrylic spacers, results are included for two different grades of surface finish namely smooth (4jum) and rough finish (20pm). It may be seen that both the spacer material and the spacer surface finish have a considerable influence 102 Figure 6.3: Variation of the mean corona pulse current with applied voltage for different materials. At each voltage 15 impulses were applied and the mean current of the first pulse measured was obtained. The Y-axis is a plot of this value. 103 on the mean corona inception time which is in agreement with the observations reported earlier by [56]. 6.3 Probability Distribution of Corona Inception Time The corona time delay shows a considerable variation near the inception voltage even when impulses of the same peak magnitude are applied. It has been shown that the most suitable distribution which fits the experimental values of corona inception time t is the log-normal distribution given by the equation 6.1 1.0 -(lnt - Int)2 , . . , x p{t) = — = x exp K ' - 0 < ln(t < oo 6.1 OlyjTX la* where t is the mean value of time t and a is the log standard deviation The delay of the first corona pulse was recorded irrespective of whether it was followed by other pulses or not. The dispersion in the discharge inception time is linked with the rate of production of initiatory electrons. The measured data was analysed and the computed distributions have been plotted. Figure 6.5 shows the computed distributions using equation 6.1 for an epoxy spacer at various voltages placed in a rod-plane gap (with a rod diameter of 5mm). For comparison, the results for the same rod-plane gap without the spacer have been included. It is apparent from fig. 6.5 that with increasing voltages, the average value of the time delay and standard deviation decrease. The presence of the spacer causes a considerable decrease in the time delay. As an example at a pressure of O.lMPa and +77kV, the mean value of the corona inception time is about 40/xs, but this value is without the spacer. It is apparent that the spacer plays an important role in the avalanche development. This fact becomes obvious in the following section. 6.4 Von-Laue Plots The Von-Laue plot has been used to determine the initiating electron density in the critical volume of the gas in the region near the positive rod electrode. The plot is essentially a probability plot of — log(l — p(t)) versus time to appearance 104 140-i 40 H 1 1 1 ! 0 50 100 150 200 TIME IN MICROSECONDS Figure 6.4: Variation of the mean corona inception time with applied voltage for different materials 105 3 - 1 TIME IN MICROSECONDS Figure 6.5: Computed log-normal distributions of the corona pulses with and with-out epoxy spacer at different voltages with 5mm rod 106 of the first corona, where, p(t) represents the probability of a corona discharge occurring before time t. A large slope in this plot is indicative of a high electron production rate. Figures 6.6-6.9 show the Von-Laue plots for various cylindrical spacers in a rod-plane gap. The X-axis is the value for the corona inception time (/AS) and the Y-axis is the value ln[l — Ep(t)], where p(t) is the probability of occurrence of the corona pulse at time t. Figure 6.6 shows that at a given voltage the slope of the Von-Laue plot is very much a function of the spacer material and its surface finish. A large slope in case of the epoxy spacer was observed whereas, in case of the acrylic spacer the slope was small. A large slope in the Von-Laue data in this case is indicative of the large influence of the insulating spacer on the corona activity. The other noticeable feature of Figs. 6.6-6.8 is that at low voltages the slope of the Von-Laue plot is small. The slope seems to increase towards a saturation level with increasing voltages. It can also be seen from Fig. 6.7, that the slope of the Von-Laue plot is smaller for a plain gas gap compared to that of a spacer gap. This suggests the participation of the insulator in the electron production process. Also, at low voltages, the surface charges setup local fields which increases the initiatory electrons probably by some detrapping of surface electrons/ions on the spacer. It is well known that the binding energies (work function) of the surface layers is dependent on the material and its surface condition. This effect can probably explain the influence of surface finish on the Von-Laue plots seen in Fig. 6.6. Figure 6.9 shows the Von-Laue plots for 2 different rod diameters with an epoxy spacer placed in the gap. This plot demonstrates the field dependency of the initiating electron density. Figure 6.10 shows the effect of gas pressure on the initiating electron density. The results indicate that the initiating electron density is high at low pressure but reduces at high pressure. 6.4.1 Discussion Breakdown in a gas gap is initiated by a critical avalanche, the starting point of which is a primary electron. The primary electrons are created in the gas by ionization due to cosmic radiation, UV radiation etc., Under the influence of the electric field, the primary electrons are accelerated towards the positive electrode. During their drift towards the anode, primary electrons collide with the neutral 4/urn EPOXY  20 fxm ACRYLIC  Plain gas gap  4 /zm ACRYLIC 20 / u m PTFE 100 200 300 TIME IN MICROSECONDS 400 Figure 6.6: Von-Laue plots for different materials at 94 kV (Rod electrode diameter 5mm) 108 94 kV 109 kV 115 kV 124 kV 0 50 100 150 TIME IN MICROSECONDS 200 Figure 6.7: Von-Laue plots for acrylic spacer at different voltages (Rod electrode diameter: 5mm) 109 o . H 82.8 kV _77_kV_ JK3.1JcV 108 kV 116 kV 01-1 L 1 " | 0 20 4 0 60 80 TIME IN MICROSECONDS 100 Figure 6.8: Von-Laue plots for epoxy spacer at different voltages (Rod electrode diameter: 5mm) 110 100 TIME IN MICROSECONDS Figure 6.9: Von-Laue plots for epoxy spacer at +77kV (Rod electrode diameter: 5mm and 10mm) Figure 6.10: Von-Laue plots for 5/xm epoxy spacer at +108kV for 2 different pres-sures. (Rod electrode diameter: 6mm) 112 Figure 6.11: Von-Laue plots for 5/xm epoxy spacer at +124kV for 2 different pres-sures. (Rod electrode diameter: 6mm) 113 molecules, some of the collisions resulting in ionization. The secondary electrons move rapidly to the head of the avalanche, whereas, the heavy ions are left behind in the tail of the avalanche. Let a be the number of electrons produced per unit length per unit time by ionizing collisions. Since SF6 is an electronegative gas, some of the electrons are attached to the SF6 molecule to form relatively sluggish negative ions. The process of attachment results in a depletion of the electrons in the region of the avalanche formation and let n be the attachment probability of an electron per unit volume. The well known streamer criterion states that breakdown takes place when the number of electrons at the head of the avalanche reaches the critical size of 108 electrons. In other words, when the number of electrons n(x) at any point distance x from the cathode becomes: where no is the number of electrons released from cathode at x = 0, breakdown ensues. Thus, in order that breakdown may occur two conditions must be satisfied: (a) there must be at least one primary electron in a suitable location in the gap in order to initiate the first avalanche and (b) the electric field must be of sufficient strength and adequate distribution to produce a sequence of avalanches thereby satisfying equation 6.2. This criterion [57] has been found to give breakdown voltages close to the measured values in uniform fields in SF6 gas. In non-uniform field in SF 6 however, the above process describes the onset of corona. With DC voltages there is usually adequate time for the initiating electrons created by natural ionizing events to form an avalanche. In case of impulse voltage application, unless the presence of the initiating electron is ensured for example by artificial irradiation, breakdown may not occur even at voltage levels beyond the DC levels. The time lag defined as the time elapsed between voltage level sufficient to cause breakdown and the time at which breakdown actually occurs can be decomposed into a statistical time lag and a formative time lag. The statistical time lag corresponds to the creation of an initial electron inside the critical volume around the stressed electrode. The formative time lag is associated with the formation of an avalanche of a critical size. So, in the case of impulse (6.2) 114 voltage application both these times affect the breakdown voltages. The critical volume or the initiation volume is the region associated with the surface area around the point electrode (in case of a rod-plane gap geometry) for negative corona and for a positive corona corresponds to a volume in the gas gap. In other words, these are regions where the release of an electron could give rise to an avalanche containing at least 108 electrons. For a positive point plane geometry the critical volume is bounded by two boundaries. The outer boundary is the contour of all points at which the net ionization coefficient (a — n) > 0 and the boundary closest to the rod electrode is determined by the minimum length x needed for the avalanche formation of a critical size. In an SF 6 non-uniform field gap, the formation of an avalanche of a critical size signals the appearance of a corona pulse under impulse voltages. Therefore, in the rapidly time varying field it becomes important for the initiatory electron to be present in the critical volume so as to produces the ionization required for the avalanche formation. It can be seen from equation 6.2 that the timing of the entire process clearly depends on n0 the number of the initiatory electrons and consequently the delay of streamer formation. Therefore, both the density of initiatory electrons in the gap and their position in the gap determine the pro-cess of an avalanche formation. Under impulse voltages, this quantity no becomes extremely important, since the applied voltage is continuously changing during the production and acceleration of electrons. Also, the position of the initiatory electron would determine the likelihood of a streamer formation. In SF 6 , the predominant species of negative ions is SFg. Negative ions may loose their electrons by several processes which include field detachment, photo detachment and collisional detachment. It has been shown that amongst the above processes the predominant mechanism resulting in electron generation is collisional detachment ie., SFe + SFe => 2SF6 + e (6.3) In a non-uniform field gap, negative corona is initiated by electrons released from the cathode by field emission and for positive corona by field-enhanced collisional detachment. In case of gaps exposed to artificial irradiation, corona activity is 115 enhanced by photoelectric emission from the cathode surface in case of negative corona and photodetachment of negative ions for positive corona. The probability per unit time for the formation of a detectable electron avalanche can be expressed as: Pt = x / P\(r)pi{r)pz{r), Pi(r) = probability per unit time that an electron exists at position r, P2(r) = probability of release of an electron from the electron site, P3( r) = probability that an electron released from the site yields an avalanche <pi = the critical volume In case of a positive corona in S F 6 , the probability pi (r) of electrons being available at a suitable position in the gas gap per unit time is mainly given by the rate at which negative ions become available for detachment at any location in the critical volume (assuming that the electron production is by negative ion detachment). The probability of an electron release from a site is determined basically by the mechanism of electron release. Obviously, the factor P2(r) is a function of the electric field, gas pressure and ambient radiation. It has been suggested [76] that electron detachment from negative ions is the most likely contributing factor to the factor p 2(r). The third probability factor Pz{?) depends on the electric field to number density ratio at the point of the initial electron release. In summary, therefore the dispersion in the discharge pulse delay is a function of the production rate of initiatory electrons, which in turn is dependent on the gas pressure, the impulse wave shape, the electrode shape and hence the electric field distribution. It has been shown by Anis [60] that the number of electrons detached per unit gas volume per unit time rp(E,p) can be expressed as shown below: Negative ion density ' collisional detachment lifetime The product of the electron production rate and the critical volume indicate the number of initiatory electrons expected to contribute to the streamer formation. The initiation volume is a function of the field distribution around the point electrode and thus varies with time under impulse voltage application. 116 In the presence of an insulating spacer in a rod-plane, the critical volume calcu-lations are difficult to perform because of the perturbation of the ambient electric field distribution caused by the presence of surface charges. It can be seen from Figures 6.4 and 6.5 that the mean value of the corona inception time decreases with increasing applied voltages. Also, the dispersion in the corona inception times is reduced with increasing voltage. From figure 6.5 it is also apparent that the dispersion in the corona inception times for a plain gas gap is larger than the dispersion in the case of an insulator gap. These results suggest that the initial corona activity close to the rod electrode is influenced by the presence of the insulating spacer. The presence of the insulator in the high field region can result in a secondary feedback mechanism, with the insulating surface playing an important role in the corona discharge mechanism. Jaksts and Cross [81] made preliminary measurements to examine the influence of a spacer on the electron and ion current components of prebreakdown current pulses. The measurements were made by bridging a uniform field gap with a ptfe spacer in air and iV2 gas. An UV pulse of approximately 10ns duration from a iV2 laser was used to photoemit electrons from the cathode surface. Measurement of avalanche currents with and without the insulating spacer present in the gap indicated an alteration in the electron component of the gas in the presence of an insulating spacer. The ion component remained unchanged. More recently, Mahajan et al.,[80] reported the differences in the prebreakdown current measurements made near acrylic and nylon spacer surfaces placed in a plane parallel field configuration in iV2 gas at 101.3kPa. Whereas, the avalanches near an acrylic spacer were very similar to that of a plain gas gap, increased avalanche current (the electron component) was observed in case of the nylon spacer. No mechanism was proposed to explain the increased avalanche current but it was postulated that the photoemission yield of the spacer was a dominant factor in altering the ionization coefficients and increasing the prebreakdown currents. It is evident that the decrease in the dispersion of time delays in case of the spacer gap as compared to a plain gas gap is a direct result of the contribution of the spacer to the initial electron density and hence streamer propagation. The Von-Laue plot can give some clues as to both the type and number of possible mechanisms. 117 The Von-Laue plot gives an image of the initiating electron density in the critical volume of the gas to initiate a streamer. From fig. 6.6 it is apparent that the probability of finding an electron inside the critical volume of the gas to initiate a streamer is very much dependent on the insulating material. Further, since no sharp discontinuities are seen in the Von-Laue plots it can be said that there is only one kind of possible mechanism for the electron generation. The larger slope in the Von-Laue plot at low voltages for a spacer gap, as compared to that of a plain gas gap suggests that the spacer has a dominant effect on electron production. This larger slope (hence more initiatory electrons) in a spacer gap may be due to the removal of trapped electrons from the spacer surface by secondary processes, like photoemission. It may also be due to the increase in the local field due to the deposited charges thereby supporting electron detachment in the gas surrounding the electrode. With increased voltages, since the ambient fields are very high, the influence of the spacer on the measured corona inception times is reduced. Since the phenomenon of electron/ion detachment from the insulating spacer surface is an extremely complex phenomenon, one can only postulate a possible mechanism leading to electron detachment from the spacer surface. It is postulated here that photoemission yield from the spacer surface is a dominant factor contributing to electron/ion detachment from the spacer surface. To a first approximation, it can be said that the photoemissive yield of an insulator can be affected both by the material and its surface condition. It is well known that polymer solids are not 100% crystalline. The non-crystalline or amorphous regions are structurally irregular and are potential sites for impurities. Also the surface of the polymer indicates the irregularity of the crystal structure. These surface states are located in the band gap. Further, the surface can hold impurities which have been absorbed and the degree of absorption depends on the surface condition of the spacer. The photoemissive yield therefore, is a function of both the material and its surface condition. The differences in the prebreakdown currents measured for different materials (figure 6.3) and the observations by Mahajan et al., and those of Jaksts and Cross, discussed earlier lend some credence to this hypothesis. The results of figure 6.6 also support this hypothesis. In case of the 5mm rod the electric fields are higher in the region close to the rod electrode as compared to a 10mm 118 rod. The higher fields associated with the 5mm rod imply a higher photon efficiency providing for a larger number of charge carriers from the spacer resulting in a higher slope of the Von-Laue plot. In summary, it can be said that the high electric field results in photon generated charge carriers from the spacer surface which contribute to the streamer formation. Although considerable work has been reported about photoconduction in solids, similar work on the photoemissive yield from surface layers of insulating spacers placed in an SF6 environment when exposed to UV light is not available. Figure 6.10 indicates that the slope of the Von-Laue plot decreases with increas-ing pressure at low voltages. Anis and Srivastava [60] by performing calculations of the critical volume at different voltages and gas pressures (O.lMPa - 0.3MPa) have shown that at voltages close to the inception voltages, the critical volume re-duces considerably with pressure, whereas, this change is smaller at higher applied voltages. Further, at a given pressure, the critical volume increases sharply with voltage but then saturates at higher voltages. These results, discussed in light of the slope of the Von-Laue plots imply that: • At a given pressure of the SF 6 gas, the slope of the Von-Laue plot is small at voltages close to the inception levels and increases sharply with voltage higher than the inception levels. The difference in the slope at these high voltages is insignificant. This is apparent from figures 6.7 and 6.8. • At voltages close to inception, the the slope of the Von-Laue plots should be high at lower pressures for a given field configuration as compared to the slopes for higher pressures. At higher applied voltages, however, the effect of SF 6 gas pressure on the Von-Laue slopes should be small. This is demonstrated by the results shown in figures 6.10 and 6.11. This observation has been consistent with that of previously reported figures for spacer gaps [56]. Quantitative analysis of the critical volume for spacer gaps is rendered difficult because of the perturbation of the symmetric field distribution around the rod electrode due to the asymmetric fields caused by the surface charges on the spacer. 119 The above results suggest the following for the pre-breakdown behavior of rod-plane spacer gaps under positive switching impulses: • The insulating spacer significantly reduces the corona inception time of a spacer gap • Both the material and its surface roughness have some influence on the pre -breakdown activity • The rate of electron production is very much field-dependent • Due to the presence of the spacer in the high field region, photoemissivity from the spacer surface has been suggested to be a possible mechanism for the charge generation contributing to the streamer growth 6.5 Surface Potential Measurements 6.5.1 Experimental procedure The surface potential has been measured in both quasi-uniform (plane parallel electrodes) and non-uniform fields with various insulating spacers. For experiments with the rod-plane gaps, capacitive probes scanned the top surface of the spacer (this probe will be referred to as the top probe ) as well as the cylindrical surface of the spacer (this probe will be referred to as the side side probe ). In case of the plane-parallel configuration only the cylindrical side of the spacer was scanned using the side probe . The experimental procedure adopted in obtaining the surface potential distri-bution is given below: • Following the cleaning procedure for both the electrodes and the spacer, as described in Chapter 4, the vessel was evacuated to approximately 8 Pa. For experiments with the plane parallel electrodes the side probe was brought to a distance of 2mm from the spacer and the probe potential recorded by a 610C Keithley electrometer. The X-Y plotter recorded the potential as a function of the angular position of the spacer. 120 • SF& gas was then introduced into the chamber and the scans described above were repeated. • The probes were then withdrawn and the voltage was applied. Switching, lightning and 60Hz AC voltages were all used in different experiments. • The gas pressure in the chamber was then increased slightly (by approxi-mately 0.03MPa) and the probes reintroduced into the vicinity of the spacer for fresh surface potential scans. This increase in gas pressure avoided the possibility of spurious fiashovers to the probe shield from occurring due to possible large charge magnitudes on the spacer. • The probes were then withdrawn and the gas pressure was restored. • This procedure was repeated for different voltages. No attempt was made to discharge the spacer before the voltage level was increased. In case of the rod-plane spacer gap, the top probe was used to scan the spacer first, followed by a scan of the cylindrical surface by the side probe . The measured potentials have been digitized and replotted as a three dimensional distribution. In the following text, both surface potential and surface charge distribution have been used. This is based on the premise that the relationship between the measured potential distribution and the equivalent charge distribution can be obtained using the charge simulation program described in chapter 5. 6.6 Results with switching impulses Figures 6.12 and 6.13 show the surface potential distribution for two different spacers prior to the application of voltage in experiments with plane-parallel and rod-plane spacer gaps respectively. It may be seen that the measured surface potentials are very small. The circular base shown in the figure represents the concentric scans made both on the top and cylindrical sides of the insulating spacer. If the distances along the X and Y axes are in the range 0 < i < 20 the plots correspond to the scans only on the top surface of the spacer. If the distances are in the range 0 < x < 40 the scans include both the top and cylindrical surfaces of the spacer. The surface potential distribution due to the residual charge on the spacer 121 is very low (typically < 0.05 volts), which corresponds to approximately 0.5pCm~2 and a maximum surface electric field of about 0.05kV/mm. On an average, at the commencement of the experiments all the insulating spacers showed similar levels of residual charge. For an insulating spacer placed between plane-parallel electrodes the surface potential distribution was monitored following 20 impulse voltage applications at each pre-determined level of voltage. The voltage application was started from approximately +45kV upto +190kV (the limit of the impulse generator). Surface potential scans made at these levels did not show any noticeable change in the surface potential distribution and the maximum voltage was still less than 0.1 volt, corresponding to a charge magnitude of about lpCm~2. This is illustrated by the Fig. 6.14. These experiments with plane-parallel electrodes were done with a recess on both the high and low voltage electrode surfaces, thereby shielding the triple-junction at both the high and low voltage electrodes. Experiments with flat (i.e. unrecessed) electrodes showed a similar behavior. In order to examine the effect of perturbing the uniform field, spacers shaped in the form of frustrum of a cone (see figure 4.11) were also used. The surface potential distributions showed similar low levels. The gap current monitored during the experiments did not indicate any corona activity even upto average electric field magnitudes of 6.3 kV/mm under positive switching impulse voltages. This figure of the electric field is obtained by assuming a uniform field distribution for a 30mm spacer gap subjected to a switching impulse of +190kV. Due to the low levels of charging observed with switching impulses for both polarities of the top electrode, in case of the plane-parallel electrode configuration, no further experiments were conducted. Higher voltages could not be achieved due to voltage limitation of the high voltage source. For the voltage levels used in this study in case of the plane parallel electrode spacer gaps it may be said that no significant deposition of charge on the spacer occurs. In an experiment conducted with an imperfect spacer (which had its edge damaged accidentally) higher levels of charge were measured on the spacer surface at the voltage levels described earlier. This may be attributed to the field distor-tion created by this imperfection. It is well known that a spacer introduced in a 122 Figure 6.12: Surface potential variation for epoxy spacer placed between plane-parallel electrodes prior to voltage application 123 Figure 6.13: Surface potential variation for acrylic spacer placed between rod-plane electrodes prior to voltage application 124 Figure 6.14: Surface potential variation for epoxy spacer placed between plane-parallel electrodes after subjecting it to positive switching impulses upto 190kV 125 plane parallel gap with an imperfect contact with the electrode can lead to partial discharges due to the resulting high stress in the gap between the spacer and the electrode even at low voltages. These partial discharges can explain the higher charge levels observed in case of the damaged spacer. It has been established [50,51] that under DC voltage application, cylindrical spacers placed between plane parallel electrodes in SFQ gas accumulate significant levels of charge at high voltages and extended periods of time. There is no agree-ment on a mechanism to explain this charge accumulation. Gaseous conduction, bulk conduction and non-linear surface conduction have been some of the mech-anisms proposed to explain this charge accumulation under DC voltages. These charges, however, have been identified to be one of the principal causes for spacer flashovers during a polarity reversal. The use of impulses (both lightning and switching) provide very little time for either bulk or surface conductivities to con-tribute to the charge accumulation process. This is because the mobility of charge carrier in the bulk of the material is in the range 10 - 1 1 to 10 _ 1 5cm 2jV • s. So, gas conduction process appears to be the only mechanism which can explain any charge accumulation under impulse voltages. More detailed discussions regarding the possible charging mechanism under impulse application are presented in section 6.6.2. Using the procedure described earlier, surface potential scans were made on cylindrical spacers placed in a rod-plane gap subjected to switching impulses. Fig-ures 6.15-6.16, show the surface potential variation of a 5pm surface finish epoxy spacer placed in a 5mm rod-plane configuration. The plots show the surface po-tential variation of the top surface of the spacers, i.e., the region close to- the high voltage electrode. Scans were made initially for a cleaned spacer to exam-ine the residual charge levels and were followed by 20 impulse applications at pre-determined voltage levels. Figures 6.16 b and c show the surface potential variation both on the top surface and the entire surface of the cylindrical spacer respectively. Under this level of charging, impulse fiashover occurred when the volt-age level was raised to +124kV. Surface potential scans made following a fiashover show considerable amounts of residual charges. From figures 6.15-6.16, it is quite obvious that at low voltage levels, the surface potentials are quite low. Further, in 126 the immediate vicinity of the rod electrode, for positive impulses, negative charge deposition is seen and positive charge appears to have been deposited towards the spacer edge, where there is an abrupt change in the direction of the ambient field. At voltages close to fiashover levels, significant positive charge is deposited on the spacer surface. The above results were reproducible when the experiments were re-peated on the same spacer which was cleaned by the procedure described to reduce the residual charge. The results were not identical but the same general trend was apparent. Figures 6.17-6.18 show the results for a 4/j,m surface finish acrylic spacer. The residual charge in this case showed similar charge patterns as observed for the epoxy spacer. Upto a voltage level of +77kV, significant negative charges were observed on the top surface of the spacer in the vicinity of the rod electrode with the positive charges towards the spacer edge. At higher voltages of +95kV and + 108kV, the charge patterns reversed with the positive charges being deposited on the spacer. Figure 6.19 shows the results obtained for a 20//m surface finish spacer and shows that the top surface of the spacer has significant positive charges, including the cylindrical surface, with localized peaks of negative charge appearing on the cylindrical surface. Although bipolar charging is seen at voltage levels of +77kV and +H6kV, significant positive charging has been observed. Different samples of the spacer, subjected to similar voltage levels, showed similar charge patterns with minor variations in the location of the peaks. Experimental results with a 20/xm surface finish acrylic spacer showed similar charging behavior. No significant differences were observed in charge magnitudes for the different surface roughness of the selected acrylic spacer. Figures 6.20-6.21 show the results for a 20/xm surface finish ptfe spacer placed in a rod-plane gap. The distinct feature in the charge distribution at various voltages was that the magnitudes were lower in comparison to those obtained for the epoxy and acrylic spacers at similar voltage levels. A further increase in the voltage level from +62kV to +77kV only increased the magnitudes of the respective positive and negative peaks with minimal changes in the overall charge distributions. Once again, at low voltages (+62kV) regions close to the high voltage electrode had negative charge deposition, and those further away from the high voltage electrode 127 The rings show c i r c u l a r scans done on the top of the spacer. When the radius of the rings exceed 20mm the 3D plo t shows r e s u l t s of measurements on both the top surface and the c y l i n d r i c a l surface of the i n s u l a t i n g spacer. Figure 6.15: Surface potential variation for h\im surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) OkV b) +48kV 128 Figure 6.16: Surface potential variation for 5p,m surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) +66kV b) +77kV (on the spacer top) c) +77kV (entire spacer) 129 Figure 6.17: Surface potential variation for bum surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) OkV b) +48kV 130 Figure 6.18: Surface potential variation for a 5/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) +63kV b) +77kV 131 a b Figure 6.19: Surface potential variation for a 20/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) +77kV (top surface only) b) +77kV (entire spacer) 132 had positive charge deposition. Figures 6.22-6.23 show the corresponding results obtained for a 5/xm surface finish epoxy spacer placed in a 10mm rod-plane gap. Once again, at low voltage, negative charges appear in the vicinity of the rod electrode and positive charges appear towards the spacer edge. At high voltages, significant positive charges are seen on the spacer surface. In this case, the charge levels are lower. Figures 6.24-6.25 show the results obtained for a 4/ura surface finish acrylic spacer placed in a 10mm rod-plane gap. Figure 6.26 shows the results for a 20/j,m surface finish acrylic spacer. Figures 6.27-6.28 show the results for a 20/zm surface finish ptfe spacer. It is apparent from these results that despite the very random nature of charg-ing, and coexistence of bipolar charges there are some noticeable trends: • At low voltages, the spacer is negatively charged in the vicinity of the rod electrode with the positive charges depositing towards the spacer edge. • As the voltage is increased say between +62kV and +77kV in the case of epoxy, significant negative charges are deposited on the spacer surface with some positive charges appearing towards the spacer edge. • At higher voltages, close to the breakdown levels, the spacer is mainly charged positively. Figure 6.29 shows the shot-to-shot variation in surface charge on a spacer placed between rod-plane electrodes subjected to repeated switching impulses. Initially the spacer, as seen in figure 6.15a, was with low levels of residual charges. The impulse voltage was then applied to the system and the procedure described in section 6.5.1 was used to scan the spacer surface. Following the scan, another impulse was applied to the spacer. The charge scan was performed once again. In this manner the charge distribution was monitored with each application of the impulse voltage. It can be seen from the figure that minor changes occur in the charge distributions. These results were obtained for a 20/xm ptfe spacer. All the above results were obtained for an SF 6 gas pressure of O.lMPa. First order approximations of the charge magnitudes (i.e., uniform field distribution 133 Figure 6.20: Surface potential variation for 20/xm surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) OkV b) +48kV 134 Figure 6.21: Surface potential variation for 20/zm surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 5mm a) +62kV b) +62kV (entire spacer) c) +77kV (entire spacer) 135 Figure 6.22: Surface potential variation for a 5/xm surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) OkV b) +48kV 136 Figure 6.23: Surface potential variation for 5/xm surface finish epoxy spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +77kV b) +116kV (insulator top) c) +116kV (entire spacer) 137 Figure 6.24: Surface potential variation for 4pm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) OkV b) +49kV 138 Figure 6.25: Surface potential variation for 4/iro surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +63kV b) +77kV 139 Figure 6.26: Surface potential variation for 20/xm surface finish acrylic spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +77kV (top surface) b) +77kV (entire spacer) 140 Figure 6.27: Surface potential variation for 20/zm surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) OkV b) +49kV 141 Figure 6.28: Surface potential variation for 20/im surface finish ptfe spacer placed between rod-plane electrodes at different voltages, rod diameter 10mm a) +77kV b) +95kV 142 between the probe and the charged patch) estimated from the probe potentials are 20fj,Cm~2, which corresponds to surface fields of the order of 1.5kV/mm. In comparison to the applied field for these corresponding figures, these values of the field due to the surface charges are approximately ±20%. This suggests that the ambient field i.e., the electric field distribution on the spacer surface without the surface charges may be perturbed by upto ±20% due to the presence of charges. In summary, under uniform field conditions, insulating spacers subjected to im-pulse voltages at atmospheric pressure SF 6 do not acquire significant charge. On the contrary, under non-uniform field conditions, spacers can get charged signifi-cantly. The nature of charging is very random and depends on many parameters such as the material, its surface finish and the applied voltage. In order to examine the effect of increasing the rate of rise of the impulse voltage on charging, lightning impulse voltages (1.2/AS rise time) were used. The following section presents the results obtained. 6.6.1 Results with lightning impulses Figure 6.30 shows the surface potential distribution for a spacer subjected to pos-itive lightning impulse voltage. The insulating spacer was of surface finish epoxy placed in a 10mm rod-plane gap. The charge patterns were similar to those with positive switching impulse voltage applications. At low voltages, significant negative charges can be seen in the region close to the rod electrode, with positive charges deposited at the edge of the insulating spacer. At higher voltages however, significant positive charging is seen on the entire spacer. Figure 6.31 shows the results of application of a negative polarity to the rod electrode with a 20/j.m surface finish ptfe spacer placed in the rod-plane gap. Sig-nificant positive charges are now deposited on the spacer in the vicinity of the rod electrode. At low voltages positive charges appear towards the edge of the insu-lating spacer. This pattern of charging is opposite to the results obtained with the positive polarity. At higher voltages the insulating spacer acquires significant negative charges. In conclusion, therefore, both lightning and switching impulse voltages show similar behavior. 143 Figure 6.29: Shot-to-shot variation spacer placed between rod-plane electrodes with repeated positive switching impulse application at 62kV [rod diameter 5mm]: a) Shot 1 b) Shot 2 144 Figure 6.30: Surface potential variation for a epoxy spacer placed between rod-plane electrodes following positive lightning impulse, [rod diameter 10mm]: a) 62kV b) 77kV c) 95kV 145 Figure 6.31: Surface potential variation for a 20/xm acrylic spacer placed be-tween rod-plane electrodes following negative 62kV lightning impulse, [rod diame-ter 10mm] 146 To examine the shot-to-shot variation in the charge pattern experiments were conducted to monitor surface potential distributions following each impulse voltage application. Figure 6.32 shows the results obtained after two consecutive negative lightning impulse voltage applications. The results are analogous to those obtained for the case with positive switching impulse voltage application i.e., at low voltages the region on the spacer close to the high voltage electrode had significant positive charges with little or some negative charge towards the spacer edge. These conclu-sion may be drawn by comparing the results presented in the figures 6.29 and 6.32. In conclusion, comparing the results of positive switching and lightning impulse voltage application the trend in the charge deposition are matched. The charge patterns under positive impulse applications are complementary to the results of the negative impulse application i.e., reflect a mirror image behavior. 6.6.2 Discussion Comparing the results obtained with SF6 spacer gaps in quasi-uniform and non-uniform fields, the following principal conclusions may be drawn: • Under uniform field conditions, insulating spacers subjected to impulse volt-ages in SF 6 at atmospheric pressure do not acquire significant surface charge. Defects present on the spacer or the electrode surface principally at the triple junction can cause charge deposition. The magnitude and the nature of the charges are dependent entirely on the severity and location of the defect and the applied voltage. • Under non-uniform field conditions, even under impulse voltage application significant charge deposition on the spacer surface can occur. The charge patterns show the existence of bipolar charges. At low voltages, in the vicinity of the high voltage electrode charges of a polarity opposite to the applied voltage are deposited with homopolar charges deposited further away from the high voltage electrode. However, at very high voltages close to breakdown the spacer is principally charged to the same polarity as that of the applied voltage in the vicinity of the high voltage electrode. 147 Figure 6.32: Shot-to-shot variation spacer placed between rod-plane electrodes with repeated negative lightning impulse application at 62kV [rod diameter 5mm]: a) Shot 1 b) Shot 2 148 Mailer et al.,[56] have also shown the coexistence of both positive and negative charges on spacers subjected to positive impulse voltages in SF6 gas at O.lMPa. Their results, however, were obtained following the depressurization of SF6 and making the measurements in air. Further, the surface potential measurements were made only across one diameter of the cylindrical spacer thereby not giving adequate information as to the overall charge distribution. Since their measure-ments were not made in-situ their validity is questionable. This is because the dielectric strength of the gas between the capacitive probe and the charged surface has been reduced by depressurization which can result in back discharges to oc-cur between the charged patches following the introduction of the probe into the vicinity of the charge patch. This can affect the original charge distribution. This fact was confirmed by some experiments during this investigation. The residual charge distribution following DC voltage application to a spacer gap in SF 6 gas yielded different results when the chamber was depressurized and measurements were made in air. Further, back discharges occurred between the spacer and the probe tip consequently altering the original charge distribution. Mailer's [82] ex-periments conducted in air under DC conditions indicated a trend similar to those obtained in this investigation, i.e., the charge distribution in the vicinity of the high voltage electrode at high voltages reverses polarity as compared to the distribu-tions observed at low voltages. Of course, in their case the results, show that: with experiments in air at O.lMPa that the charge distribution on the top surface of the spacer has the same polarity as that of the applied voltage at low voltages. For example the charge distribution is principally positive, for positive polarity of the rod electrode at low voltages. At high voltages on the rod electrode, however, the charge distribution is reversed in the region close to the electrode. Once again these observations were made in their experiments after surface potential measurements made across one diameter. In order to confirm their observations the experiments reported by Mailer et al.,[82] were duplicated to the extent possible. A cylindrical acrylic spacer 40mm in diameter and 30mm long was placed in a 3.2mm rod-plane gap. In the experimental results reported by Mailer, the rod diameter was 1.6mm. Ensuring very low levels of residual charges on the spacer prior to voltage appli-cation, the insulator was subjected to positive DC voltages of 10, 20 and 35kV 149 in air at O.lMPa. The voltage was maintained for approximately 20 minutes at each voltage level and the surface potential scans were done. Figure 6.33 show the results obtained. The X and Y axes refer to the distance from the tip of the high voltage electrode and the Z axis is the surface potential distribution. At a volt-age of 40kV the spacer flashed over. The experiments were discontinued following the occurrence of flashover. It can be seen that even in this case bipolar charging is dominant. Further, reconstructing the distribution across a suitable diameter a match was obtained with the results reported by Mailer et al. This however, does not give the composite picture of the behavior of the charge distribution on the spacer in air subjected to DC voltages. The explanation given for their ob-servations was that the corona generated light at higher voltages induced charge injection into the bulk of the material. This is supposed to have caused electrons to appear at the spacer surface and neutralize the homopolar charge distribution and result in a heterogenous charge distribution. The argument here is that the overall charge distribution is extremely heterogenous even under low voltage and not ho-mopolar as suggested by them. Previous investigators [51] have also confirmed the existence of bipolar charges on spacers subjected to DC voltages in SF 6 gas. No at-tempt however, was made by them to explain the observed bipolar (heterogenous) charging. Cooke [85] has stated that charges arriving from a surrounding gaseous atmo-sphere may accumulate on an insulating surface. The accumulation of charges are dependent on their source, for example, a protrusion on an electrode surface, and on their transport in the gas to the insulating surface. Once on the surface, the charges can either migrate on the surface, which depends on the surface condition of the spacer or migrate into the bulk. These charges can also get neutralized via the gas. Therefore, the net charge on the insulator surface can be expressed by the equation 6.5: = charge arrival rate — charge loss rate. (6-5) The charges transported to the insulator surface may employ a gas ionization mech-anism or other field dependent processes. A remote ionization emitter in the form of a pin 0.25cm to 2cm in length and placed at varying distances from a post insulator in SF6 gas was used. The insulator was placed between plane parallel 150 Figure 6.33: Surface potential distribution on an acrylic spacer placed between a 3.2mm rod-plane gap in air at O.lMPa subjected to positive DC voltages for 20 minutes, a) lOkV b)20kV c)30kV d)Maller et al., results[82] 151 electrodes with the pin located on the ground plane. Cooke confirmed that the observed charging of the insulator could be explained using the basic properties of charge transport in the surrounding gas even at elevated gas pressures and volt-ages (+600kV DC voltage). It was found that the charges followed drift patterns and were deflected according to the space charge field. It is important to mention here that the observed charge patterns were obtained after the chamber was de-pressurized and exposed to atmospheric air. Further, the capacitive probe had a probe tip diameter of 6mm and was placed at a distance of 10mm from the charged surface. As is well known, under these conditions it is quite likely that integration of the local charge distribution could have possibly masked the possible existence of bipolar charges. No definite conclusions can be drawn as to the possibility of heterogenous charging of the spacer under corona charging in SFQ gas. The results mentioned above indicate that corona charging in SF© leads to bipolar charging. There seems to be some contradiction in the results obtained by Mailer et al., and the present investigation with regards to the observations made with DC voltages in air at O.lMPa. There seems to be agreement with their results obtained in SF6 gas under impulse voltage application. In order to get a better understanding of the phenomenon of corona charging and consequent charge deposition on the spacer surface simple experiments were conducted to examine the influence of the gas on the gas conduction and transport phenomenon. Insulating spacers 40mm in diameter and 20mm long were placed in a rod-plane gap in the pressure vessel. Negative DC voltage was applied to the rod electrode. The spacers used were cleaned to ensure relatively low levels of residual charges. The spacers were subjected to DC voltages for 20 minutes. At each voltage level, the surface potential scan was carried out. Since the electric field lines intercept the spacer surface at right angles to the surface, with the shortened spacer, negative charge carriers generated in the gap in the vicinity of the rod electrode are expected to follow the field lines and deposit on the spacer surface. Figures 6.34-6.36 show the results obtained with DC corona charging on spacers with air [87], N 2 and SF 6 gas at O.lMPa respectively. In case of air and SF 6 the results showed bipolar charging. At low voltages, immediately below the rod electrode, positive charge accumulates. At higher voltages, however, the overall 152 charge distribution is principally negative in the regions immediately below the electrode with pockets of positive charges appearing at the spacer edge. In case of N2, for voltages upto — 40kV the insulator surface is charged negative. When the polarity was reversed on the same insulating spacer previously subjected to negative voltages the charge deposition on the spacer was principally positive. These results suggest that the intervening gas has an important role to play in the charge accumulation process. Both air and SF 6 are electronegative gases. The major ionic species in a neg-ative corona in air at O.lMPa is the CO3 ion. The negative ions generated by the corona should then follow the field lines (moving away from the negative rod electrode) and land on the spacer surface. This ion may then be neutralized by giving up an electron to the conduction band near the surface. The electron is subsequently trapped in a surface state and needs some activation energy to fa-cilitate its movement into the bulk. As stated by Baum [83] this energy can be supplied thermally or by excited molecules. The energy supplied by the excited molecules may take place through the emission of a photon of sufficient energy by the excited molecule, with subsequent photoinjection of the trapped charge. A similar phenomenon may occur for SF 6 gas gap in corona. Data on the ion species for corona in low fields in SF6 gas, has suggested the existence of SFQ ion together with heavy cluster negative ions [84]. It has been suggested that in GIS, even at electric fields under normal service conditions, SF$ ions or associated clusters exist. With increased ^ ratios, i.e., the ratio of the electric field to the gas pres-sure, these clusters will be destroyed by collisions, so that the dominant species of negative ions under pre-breakdown conditions will be SF$ . There is no published information as to the behavior of these ions or the clustered species when they are incident on the spacer surface, with regards to their mobility, charge transfers etc. The results however, show similarities as to the heterogenous nature of charging of spacers subjected to DC corona in air and SFg . Similar experiments conducted with iV2 gas seems to indicate the absence of bipolar charging and the charging is predominantly homopolar. Unlike in SF 6 and air, in the case of N2 no negative ion formation occurs. The electrons released from the cathode in these experiments, follow the field lines and deposit on the spacer 153 resulting in a homopolar distribution consistent with the present understanding of corona charging of spacers. The above results showed that the spacer charge accumulation in N2 gas is consistent with the present understanding of gaseous conduction and significant differences in the charge accumulation are observed if the surrounding gas is electronegative in nature. Thus, in case of the SFe gas significant bipolar charging of the spacer occurs. 6.6.3 Mechanism of charging In the experimental investigations conducted with a 30mm spacer placed in a rod plane gap there is a component of the electric field tangential to the spacer surface. The experiments described earlier suggest that an insulating spacer in SF 6 gas at O.lMPa acquires charge under both lightning and switching impulses. From the design and operations standpoint, this is an important finding and is cause for concern. Although, the nature and phenomenon of charging is quite complex, a designer is interested in the following: • How do these charges perturb the original ambient field distribution? • Do these charges really affect the overall insulation performance of the sys-tem? Before one can attempt to answer these questions, accepting the fact that the spacer does get charged, it is of interest to know: • What are the sources of charges on spacer? • Why does the spacer gets charged in the manner illustrated in Figures 6.15-6.28 at various voltages? • Can these charges explain the observed changes in the Von-Laue plots, i.e., small slopes at low voltages and large slopes with increasing voltages? The differences in these slopes at low voltages for plain gas gaps and spacer gaps? The following discussion focuses on the above mentioned aspects, in an attempt to explain the mechanics of charge accumulation. The phenomenon of spacer charging 154 Figure 6.34: Surface potential distribution on a ptfe spacer placed under a 10 mm rod-plane gap in air at O.lMPa subjected to negative DC voltages for 20 minutes, a) 20kV b)40kV Figure 6.35: Surface potential distribution on a ptfe spacer placed under a 10 mm rod-plane gap in N 2 at O.lMPa subjected to negative DC voltages for 20 minutes, a) OkV b)20kV c)40kV d) +40kV 156 Figure 6.36: Surface potential distribution on a ptfe spacer placed under a 10 mm rod-plane gap in SF 6 at O.lMPa subjected to negative DC voltages for 20 minutes, a) 20kV b)55kV c)80kV 157 is a complex process. A clearer understanding of this phenomenon may be possi-ble by drawing an analogy with the present understanding of the phenomenon of positive impulse flashover of gas gaps. A quantitative model for the prediction of breakdown voltages of highly inhomogenous electric field gas gaps (as an example a rod plane gap) has been developed by Wiegart et al [84]. This breakdown model was restricted to positive polarity because of its greater practical importance. The present studies show that negative breakdown voltages for spacer gaps are much higher than the positive impulse breakdown voltages. The probability of failure of an SF 6 gas gap subjected to impulse voltages depends on the appearance of a free electron in the region of high electric field of sufficient magnitude to cause an avalanche. For positive impulse voltages the electron is detached from the negative ions in the gas due to field-induced collisional detachment. The negative ions in the SFe gas are also generated by cosmic rays. Some of these ions are lost through recombination or by collision with the walls of the test apparatus. An equilibrium zero field ion density distribution is then established. Computations have been made to determine this equilibrium distribution and comparisons have been made with measurements. Wiegart et al., [84] have shown that the initial electron re-lease would have to result from an ionization of the SF 6 molecule by cosmic rays or from negative ion detachment. For DC voltages, it has been suggested that both these processes may be contributing factors for the electron release. Under impulse voltage application, however, only electron detachment from the negative ion is re-sponsible for the availability of the initial electron for the avalanche development. Further, it has also been shown that the statistical time lag is directly proportional to the critical volume which in turn is proportional to the ion concentration. The electron production rate is a field dependent process. In other words, the detachment coefficient is sensitive to the electric field. It has been shown that the equilibrium zero-field ion density distribution is sensitive to small voltage changes. The distributions have been shown to alter dramatically even for small bias voltages. This is as a consequence of the high field-induced mobility of the negative ions. When an impulse is applied to an electrode arrangement, as a result of the field-induced ion drift, the ion density has been shown to be altered appreciably. Thus each impulse can either reduce the 158 ion concentration or increase the equilibrium distribution. In summary, the ion density distribution in a gas gap is sensitive to low fields. The results of the surface charge distributions show that at low voltages the spacer has negative charges in the vicinity of the rod electrode for positive im-pulses. From the previous discussion it may be said that the presence of the negative charges on the spacer can result in the depletion of negative ions in the gas gap between the rod electrode and spacer. So, the depletion of negative ions imply the depletion of the initiatory electrons in the critical volume, resulting in a reduced slope of the Von-Laue plot. This argument seems to be consistent with the observations. In summary, it may be said that the presence of the charges on the spacers in the present investigations are due to gas conduction. The influence of the spacer material on the charge accumulation/dissipation on the surface will also influence the final charge magnitude. For example, the differences observed between epoxy and ptfe spacers. The spacer acquires charges and modifies the ambient field distribution. The results obtained are consistent with those obtained bv previous investigators. 6.7 Effect of A C Pre-Charging These experiments were carried out to investigate the effect of 60Hz AC voltage application on spacers placed in both quasi-uniform and non-uniform fields. Fol-lowing the initial cleaning procedure and the measurement of the residual charge on the spacer (i.e. the measurement of the surface potential by the probe), 60Hz AC voltage was applied at different voltages for different durations. In case of the quasi-uniform field configuration, the effect of duration of voltage application and the magnitude of the applied voltage on the spacer charging was minimal. Figure 6.38 shows the surface potential variation for a spacer placed between plane-parallel electrodes subjected to AC voltages of 70kVrmAoT 20 minutes. The same spacer was subjected to varying durations of voltage at the same magnitude. This did not increase the surface potential magnitudes. Since the high voltage test gener-ator could generate a maximum of only 75/cVrmsthis was applied to the system. No fiashover was observed at these voltage levels. These results were obtained for electrodes which shielded the triple junction. Similar results were obtained when unrecessed electrodes were used in the tests. It is quite apparent in these experi-159 4000 2 4 ELECTRODE SPACING [cm] 4 0 0 0 -I 2 4 6 ELECTRODE SPACING [cm] Figure 6.37: a) Negative ion density distribution along a rod-plane gap for various D C voltages; b) Negative ion density in a rod-plane gap at various times during a negative impulse application. Voltage=124kV, p=0.lMPa and T=293K[84] 160 ments that the electric field on the insulator surface was uniform and in a direction only tangential to its surface. Experiments conducted with conical spacers which resulted in both normal and tangential components of electric fields on the spacer surface produced very similar results. Assuming that charges were generated in the gas or from the electrode surface, they would follow the field lines and de-posit on the spacer surface. This is because of a component of the electric field normal to the spacer surface. The results with conical spacers also did not show any appreciable charging. This suggests that under uniform field conditions with reasonably good electrode surfaces charge accumulation of spacers in SFQ gas at O.lMPa is minimal. This observation is independent of the relative proportions of the normal and tangential components of the electric field on the spacer surface and the duration of the applied voltage. Also for the range of the applied voltage no significant charging of the spacer occurs. Figure 6.39 shows the surface potential distribution with AC pre-charging on epoxy spacers placed in a rod-plane gap, with rod diameter of 5mm. Figure 6.40 shows the surface potential distribution on a spacer placed in a rod-plane gap subjected to various magnitudes of DC voltage for different times. The rod diameter was 10mm. In the region close to the high voltage electrode some charging is observed. Regions away from the high voltage electrode, however, show very low levels of charge. With the duration of the applied voltage increased from 5 minutes to 60 minutes some increase in the charge levels were observed. Once again the region close to the high voltage electrode has a higher charge deposition. When the voltage level was increased to 43kV, charge magnitudes close to the high voltage electrode increased. It can be seen from these figures that charging increases with the AC voltage. Both positive and negative charges co-exist and no single polarity charge dominates. These patterns become more complex with increasing level of the AC voltage and duration. The results of charge distributions obtained with AC and impulse voltage ap-plication to spacers suggests that spacers in non-uniform fields acquire charges. The charge patterns are heterogenous in nature in both air and SFQ gas. In case of N 2 , however, the results indicate homopolar charging under corona conditions. Negligible charging was observed on the spacers placed in quasi-uniform fields. 161 Figure 6.38: Surface potential variation for a spacer placed between plane-parallel electrodes subjected to AC voltage of 70fcVrmg for 20 minutes 162 Figure 6.39: Surface potential variation for a spacer placed between rod-plane electrodes subjected to AC pre-charging, rod diameter 5mm 163 o 6 Figure 6.40: Surface potential variation for a spacer placed between rod-plane electrodes subjected to AC pre-charging, rod diameter 10mm a) 30kV 5min b) 30kV 30min c) 30kV 60min d)43kV 30min 164 Slight perturbation of the uniform field with the use of spacers in the form of conical frustra also showed very low levels of charge accumulation on the spacer. This suggests that under quasi-uniform fields charge accumulation on the spacer is negligible both under AC and impulse voltage applications. The sources of charges in case of the rod-plane spacer gap are due to gaseous conduction and the secondary emission yield from the spacer surface. The observed charge patterns are most likely due to the interaction between the charge carriers in the gas and the spacer surface. The bulk conductivities for the spacers used in these experiments are very low. So, for any significant charging to occur via the bulk the voltage must be applied over several hours. Under impulse voltage application the duration of the applied voltage is very small when compared to the time required for the charge carriers in the bulk of the spacer to move. So, if the bulk has to play any role in the charge accumulation process the bulk conductivity must be highly non-linear or non-uniform. Nakanishi et al., [50] have shown that in epoxy spacers, bulk conductivity is almost constant and independent of the electric stress. Therefore, bulk conduction is not considered to play an influencing role in the observed charging phenomenon. The phenomenon of spacer charging under non-uniform fields is quite complex. Having examined the possible sources of charges, and their nature of distribution on the spacer it is important to establish the significance of these charges on the breakdown behavior of the spacer gap. The questions that need to be addressed are: • Do these charge patches alter the breakdown behavior of a spacer gap? • Do they influence the time-to-breakdown under impulse application? • Does AC pre-charging of the spacer affect the spacer impulse breakdown behavior? As far as a gas gap is concerned, the statistical time lag plays an important role in impulse stressing of SF 6 insulated equipment. This determines both the temporal and spatial distributions of breakdowns. As mentioned before, most of the primary electrons are produced by collisional detachment processes from negative ions and it has been shown both by measurements and calculations that bias voltages may 165 effectively influence the ion density distribution in the gap and thus the statistical time-lag. A lack of primary electrons will lead to a greater variation in the time-to-breakdown. On the other hand, an excess of primary electrons will reduce the statistical scatter of the formative time-lag. A typical GIS system is stressed with AC voltages at the rated voltage of the system. Under abnormal conditions, either a switching or a lightning transient can appear on the system resulting in a composite stress of the insulation system. It is of importance to know how the GIS would behave under the application of this composite stress. It is well known that for plain gas gaps, subject to impulse voltages in SF6 gas, the necessary and sufficient conditions for breakdown are: • The presence of a negative ion in the critical volume from which an electron can be detached, • A field distribution along the gap which promotes the growth of an avalanche formation to a critical size. These two conditions are considered adequate to describe streamer breakdown in a gas gap. At higher gas pressures and higher voltages these conditions become only necessary conditions but still do not qualify to signal the onset of breakdown and a third condition is necessitated namely; • A field distribution in the gap such that the maximum field at the leader tip remains sufficiently large such that a critical charge is injected into the critical volume [84]. It has been suggested [84] that normal AC voltage application can cause a peri-odic lack of negative ions in the critical volume close to the high voltage electrode. Further, as mentioned earlier, low DC bias voltages can also cause a redistribution in the equilibrium negative ion density profiles. Extending this observation to a spacer gap, it may be said that the presence of charges on the spacer can affect the negative ion density in the region close to the high voltage electrode. As a consequence of this, it may be said that the presence of surface charges on the spacer may influence the impulse breakdown probability of the spacer gap. This hypothesis is a direct extension from the observations made for plain gas gaps. 166 For plain gas gaps it has been observed that higher breakdown probabilities are obtained when impulse voltages are applied sequentially without adequate wait-ing time between voltage applications. In other words, it has been proposed [84] that to obtain statistical independence from the time between voltage applications, measurements should be made with a minimum waiting time of 6 minutes between voltage applications. The following section presents the results obtained for breakdown voltages of spacer gaps under both sequential impulse voltage application and those with AC precharging. 6.8 Breakdown Behavior of the Spacer Gap It has been shown that large reductions in the dielectric strength of a spacer gap are observed when pre-existing stress of DC voltage and the applied impulse voltage are of opposite polarities. Further, for fast transients the breakdown behavior of spacer gaps is different when compared to plain gas gaps. Although work with DC pre-charging of a spacer and its influence on impulse behavior have been reported quite extensively, very little has been reported on the composite stress of preexisting AC voltage on the impulse fiashover behavior of the system. The breakdown of quasi-uniform and non-uniform fields spacer gaps due to impulses with different levels of AC pre-stress are presented. 6.8.1 Plane-Parallel Electrodes In a plane-parallel electrode configuration, very low levels of charging have been observed with impulse voltage application. Due to the limitation in the available voltage source voltages beyond ±190kV could not be obtained. No flashovers, however, were observed upto +190kV impulse levels. Moreover, even with AC pre-charging at 70kV for 20 minutes, no premature flashovers of the spacer gap have been observed. Similar results have been obtained when the AC pre-charging time is increased to 6 hours. It is known that long-term DC pre-stressing of spac-ers in SF 6 gas leads to increased charge accumulation [50]. It is evident from fig. 6.38, however, that no noticeable charging occurs over extended periods of AC 167 pre-charging in the uniform field configuration used. In a limited number of exper-iments, charges of both polarity have been observed with AC pre-charging. This is probably due to imperfections at the triple junction. 6.8.2 Rod-plane Electrodes In an attempt to investigate the effect of pre-charging in non-uniform fields ex-periments, similar to the one described above have been conducted with epoxy and acrylic spacers between a 10mm hemispherically capped rod and a plane elec-trode. In case of recording the impulse breakdown voltage of the spacer with AC pre-charging it was important to obtain the breakdown voltage values free of any history of the residual charge on the spacer. Consequently, spacers were charged to an AC voltage at a desired level, initially ensuring low levels of residual charge. The AC voltage was then removed and the generator was setup to generate impulse voltages. The charging voltage on the generator was adjusted to voltage levels close to breakdown. The impulse voltage was applied and the breakdown voltage was recorded. If the flashover occurred on the tail of the impulse wave, the peak value of the impulse was recorded as the breakdown voltage. If the flashover, however, occurred on the impulse front, then the voltage at breakdown was recorded as the breakdown voltage. This was consistent with industry practice. The spacer was replaced following each flashover by depressurizing the system and replacing it with a freshly cleaned spacer. Thus, each data point in figure 6.42 corresponds to measurements with a fresh sample. Figure 6.41 shows the shot-to-shot variation in positive impulse flashover volt-age for the epoxy and acrylic spacers. It may be observed that there is a wide variation in the flashover voltage. It is also apparent that the flashover voltage does not recover following repeated flashover for the epoxy spacer. In fact, attempts to restore the strength by applying impulses of the opposite polarity (negative) so as to discharge the surface have not been successful. Figure 6.42 shows the effect of AC pre-charging on both the first flashover voltage and the time-to-flashover for impulses. The results show that the scatter in the pure impulse flashover voltages can be high. It may be seen that although the AC pre-charging influences the first flashover (both voltage and time) the variation falls well within the statistical scat-168 u < St ma O > DC U > o X < b. W 3 Cu S o z z H as o 80-60-40 20-A l l A \ ^ V A ACRYLIC X x _EP0XY 10 SHOT NUMBER 15 20 u S3 r— a: w > o as < w D CU 1.5 1.2 o.g 0.6 0.3 0 -A ACRYLIC x EPOXY 10 SHOT NUMBER — r -15 20 Figure 6.41: Shot-toshot variation for a spacer between rod-plane electrodes with no AC pre-charging (a) positive impulse fiashover voltage (b) fiashover time 169 110 70-80- * ACRYLIC x EPOXY SO- 10 20 30 40 50 AC PRECHARGING VOLTAGE [KV rms] 60 2 Figure 6.42: Shot-to-shot variation for a spacer between rod-plane electrodes, with AC pre-charging at 50kV for 30 minutes (a) positive impulse flashover voltage (b) flashover time 170 ter for positive lightning impulse flashover voltages with no A C pre-charging. In some experiments, it has been observed that following the first flashover the sub-sequent flashovers occurred at lower voltages. With repeated flashovers, however, this downward trend is reversed in some cases. Giesselmann et al., [86] examined the role of an insulating spacer on the D C flashover of spacer in SF6 and JV2 gas. The pre-discharge development along the spacer surface was recorded using a sensitive high speed image intensifier. The pre-discharge development figures showed that, in case of SF6 gas, multiple discharges initiate along the spacer surface. The results showed that even under D C voltage application, there is a considerable scatter in the breakdown voltage of the spacer gap. Trinh et al., [55] reported that the presence of a spacer in a coaxial gap in SFe gas reduced the withstand voltage of the system. Also, under sequential appli-cation of impulse voltages, (about 100 shots) a considerable scatter was observed in the breakdown voltages. For example, under positive impulses, the scatter was approximately +20% to —35%. During this application, they also observed certain flashovers to occur at unexpectedly low voltages. This was attributed to possible surface charging of the spacer. The recovery in the spacer flashover levels was attributed to a partial or a complete neutralization of the trapped charges during subsequent breakdowns. In their experiments, no attempt was made to measure the surface charge distributions following impulse voltage application. In the present investigation, a similar scatter in the breakdown voltages under impulse voltage application was observed. These results are in agreement with the results reported by Trinh et al. In case of D C GIS it is well known that spacers subjected to one polarity voltage cause a dramatic reduction in the breakdown voltage of the spacer when the polarity of the D C voltage is reversed. In the industry, this type of reduction in the spacer flashover voltage is referred to as a polarity reversal type of breakdown. From the measured surface charge distributions following impulse voltage application, it is apparent that the spacer acquires bipolar charges. The charge distributions vary depending on the applied voltage, spacer material etc. These charges perturb the ambient electric field distribution in the spacer gap. The presence of bipolar charges can also explain the possibility of a polarity reversal type breakdown behavior. 171 Following flashover, a redistribution of the charges on the spacer can occur. Thus charge accumulation on the spacer can occur even under impulse voltages and explains the anomalous breakdown behavior reported by Trinh et al. following sequential impulse application on a GIS spacer in SF6 gas. The presence of these charges can also explain the statistical scatter observed during sequential impulse tests. In the present study it was observed that the scatter in the pure impulse flashover was ±35% in case of the acrylic spacer. In the case of epoxy spacer, repeated flashover of the spacer resulted in damage to the spacer. This explains the deterioration in the impulse flashover voltage seen in figure 6.42a with the shot number. Chapter 7 CONCLUSIONS AND SUGGESTIONS FOR F U T U R E WORK This thesis examined the role of charge accumulation on spacers in uniform and non-uniform field gaps in SF$ gas in both the pre-breakdown and breakdown stages of the discharge phenomena. Charge accumulation under AC voltages and the effect of AC pre-charging on impulse fiashover were also studied. Section 7.3 makes suggestions for future research. 7.1 Pre-breakdown phenomena From the results obtained for a point-plane geometry it is apparent that the pres-ence of a spacer in the high field region affects the pre-breakdown behavior of a spacer gap as compared to that of a plain gas gap. This effect is demonstrated by comparing the results of the Von-Laue plots for spacer gaps and plain gas gaps in Chapter 6. Since, no sharp discontinuities in the Von-Laue plots exist, it may be said that there is only one mechanism active for initial electron generation. Also, at a given pressure of the SF6 gas, the slope in the Von-Laue plot increases sharply with voltages above the inception level. Moreover, the slope of the Von-Laue plot for a spacer gap is larger compared to that of a plain gas gap. This larger slope in the Von-Laue plot indicates an increase in the initiatory electron density for a spacer gap, and may be due to the enhanced local electric field from trapped surface charges and the corona-light induced photoemission yield from the spacer surface. It is concluded that the insulating spacer significantly reduces the corona 172 173 inception time of a spacer gap. Both the spacer material and its surface roughness influence the mean corona inception times. 7.2 Spacer Charge Accumulation Under uniform field conditions, spacers subjected to impulse voltages in SF6 gas at atmospheric pressure do not acquire significant surface charge. Results of this investigation have shown that any defects present either on the spacer or on the electrode surface can lead to charge deposition on the spacer for both AC and impulse voltage application. The magnitude and the nature of charging under these conditions are dependent on the severity and location of the defect and the applied voltage. On the other hand, under non-uniform field conditions, spacers acquire significant charge under AC and impulse voltage application. The charge distribution is highly non-uniform with both positive and negative charges present in close proximity. At low voltages, the charges near the rod electrode have opposite polarity to that of the electrode, with the same polarity charges deposited towards the spacer edge with peak values of about 2 — 5/iCm"2 corresponding to fields of about 13-19kV/cm. This is about 25-30% of the ambient field. When the voltage nears the the breakdown levels, the charge on the spacer surface close to the high voltage rod electrode has the same polarity as that of the rod electrode. The peak charge magnitudes at these levels range between 15 — 20fxCm~2. Also, the charge magnitudes increase with the applied voltage. The cylindrically symmetric field distribution in case of the spacer with no charges is destroyed following this charge accumulation. It is necessary to make in-situ charge measurements in order to draw some meaningful conclusions as to the nature of charging, polarity and magnitudes of the accumulated charge. By opening the test chamber the dielectric strength of the gas surrounding the spacer is reduced. Thus local back discharges can occur between the spacer and the capacitive probe resulting in a modification of the original charge dsitribution. The spacer retains charges over considerable periods of time. Charge measurements made several times following a voltage application, showed very little change in the overall pattern, indicating the tenacity of the charges and their poor mobility on the spacer surface. Similar experiments were 174 conducted to examine the influence of the probes on the charge mobility. Once again, the results demonstrated the low mobility of these charges on the spacer surface. Experimental results for different gases, show that there is a distinct difference in the charge patterns obtained on the spacers when exposed to N2 gas as compared to the results obtained in air and SF6 . For example, in experiments with a rod-plane spacer gap, with the rod electrode held at a certain distance from the spacer, for a negative voltage application to the rod, the spacer top was charged negatively in case of N2 gas. In case of SFe and air the spacer had both positive and negative charges. Similarly, for a positive voltage application with N2, the spacer top was charged positively but bipolar charges were present in case of SF6 and air. The electronegative nature of the gas seems to affect these charge patterns. The sources of charges in the case of the rod-plane spacer gap are due to gaseous conduction and the secondary emission yield from the spacer surface. The observed charge patterns are most likely due to a complex interaction between the charge carriers the spacer surface and the local electric field. Since the decay time constant of the trapped surface charges are large and their surface mobility poor, despite the high local fields, brought about by the introduction of the probes, it may be concluded that surface conductivity has a negligible role to play in the charge accumulation process. It is well known that the mobility of charge carriers in the bulk of the spacer is very low so it would require considerable time of voltage application to observe any significant accumulation of charges on the spacer. The present investigations reveal that even under impulse voltages (where voltage application lasts for only 2 ms) the spacer acquires charges, suggesting the insignificant role bulk conductivity has in the charge accumulation process. In conclusion it may be said that under impulse voltages, both bulk and surface conductivity have a minimal role to play in the charge accumulation on spacers and it is principally due to ionization and conduction in the surrounding gas. Results of the spacer fiashover under sequential application of impulse voltages show a considerable scatter, amounting up to ±30%. The presence of charges 175 on the spacer results in a perturbation of the symmetric ambient field distribution which may account for this statistical scatter in the breakdown voltage values. The spatial distribution of bipolar charges changes after each fiashover and may trigger a reversal of polarity type of breakdown observed in DC GIS breakdown studies. Although AC pre-charging influences the first fiashover and time-to-fiashover for impulse voltages, the variation falls well within the statistical scatter for positive impulses with no AC pre-charging. 7.3 Scope for future work The present work has shown that spacer charging may explain the anomalous fiashover under impulse voltages that has been reported in literature. The observed surface charge accumulation takes place for a variety of reasons. Some of the more important ones being: • emission sites on the electrode surface giving rise to charge carriers that follow field lines and deposit themselves on the spacer surface. • charge carriers generated by ambient irradiation of the surrounding gas in the spacer gap causing charge deposition on the spacer surface. Extensive investigations have revealed that the time constants for charge decay are very large (can be several hours or days, depending on the ambient gas pressure). In order to examine the effect of the spacer charging on spacer fiashover, in greater de-tail, it is important to know the influence these charges and their fields have on the transport and ionization properties of the charge carriers involved in breakdown. From high speed photographs of surface fiashover, there is evidence to suggest that local fields anywhere on the spacer can initiate breakdown of spacer gaps. Due to charge patches of both polarities such local field concentrations exist and may explain their influence on the breakdown behavior. From a practical standpoint, therefore, it is important to examine how the ionization coefficients are affected and how one can include surface charging to build a reasonable model to predict spacer fiashover in SF 6 . Field calculations with strategically located known charge distributions can be used to compare the predicted pre-breakdown current growth 176 and the measured values of the pre-breakdown corona current. By improving the bandwidth for the fast risetime corona current measurements, better correlation between the deposited charge and the measured current is possible. This work can then be extended to examine the influence of charges (both magnitude and location) on the breakdown voltages. This may lead to a better understanding of the reasons as to why local discharges occur and conditions under which this could result in spacer fiashover. It is also important to conduct these studies using higher gas pressures since in a practical system the typical operating SF 6 gas pressure is about 0.3-0.6MPa. These experiments have to be complemented with high speed photographic techniques using image intensifiers. Application of the model for cur-rent growth to a practical GIS system with spacers operating at higher pressure may provide a better understanding of the phenomenon and ultimately help in improvement of the fiashover performance of these systems. Although it has been established in this work that spacer pre-charging under AC does not influence the impulse fiashover voltage significantly, spacer charging is undesirable since the presence of charges on a spacer can attract metallic con-tamination. It is well known that the presence of such contamination on a spacer causes a dramatic reduction in the breakdown voltage. The practical implications of this are: • spacer charge accumulation is to be minimized and • even if the spacer accumulates charge it is important to explore means to dissipate these charges quickly. A solution to the first problem could be to ensure by suitable design, for both the spacer and the electrode system,that charge accumulation on the spacer surface is minimized. A solution to the second problem could be to suitably coat the spacer with a material whose photoemissive yield is low, thus ensuring that charge accumulation is minimized. Another aspect of the study could examine spacer coatings which provide quick dissipation of any accumulated charge. Bibliography [l] Cookson A. H., Gas-Insulated Cables, IEEE Transactions on Electrical Insu-lation, Vol EI-20, No. 5, Oct 1985, pp.859-890 [2] Siemens Catalog No. E-129/1420-101 [3] Cookson A. H., Review of high voltage gas breakdown and insulators in com-pressed gases, Proceedings of the IEE, Vol 128, Part A, No. 4, 1981, pp.303-312 [4] Hampton, B. F. Outline of a flexible SFQ insulated cable, Proceedings of the IEE, Vol. 123, 1976, pp.159-164 [5] Giao-Trinh, H., Vincent, C. Bundled conductors for EHV Transmission, IEEE Power Engg. Society, Summer Meeting, Paper No. 78-192-7, 1977. [6] Lautenschlager, G., Cooke, C. M. Bundled conductors for gas insulated high voltage power transmission cables, IEEE Transactions on Power Apparatus and Systems, Vol PAS-100, 1981, pp.394 [7] Kobayashi, T., Mori, S., Koshiishi, M. , Minomiya, K. Development of compact 500 kV 8000 A gas insulated transmission line - Study on insulator design, IEEE Transactions on Power Apparatus and Systems, Vol PAS-103, No. 11, 1984, pp.3154-3164 [8] Takuma, T., Watanabe, T., Kuono, T. Effect of a narrow gas gap on the sur-face flashover of spacers in, International Symp. on High Voltage Engineering, Zurich, 1975, pp.440-450 [9] Laghari, V.R., Quereshi, A. H. Flashover studies of cylindrical spacers in SFG and N2 mixtures, IEEE Power Engg. Society Winter Meeting Paper No. A80-055-4, New York, 1980. 177 178 [10] Cooke, C. M., Trump, J. G. Post type spacers for compressed gas insulated cables, IEEE Transactions on Power Apparatus and Systems, Vol PAS-92, No. 5, 1973, pp.1441-1447 [11] Trump, J. G., Andrias, J. High voltage DC flashover of solid spacers in com-pressed nitrogen, Trans. AIEE No. 41, 1941, pp.986-990 [12] Dakin. T. W., Studianarz, S. A. The voltage endurance of epoxy resins, Con-ference Record of the 1978 IEEE International Symposium on Elec. Ins., 1978, pp.216-221 [13] Wooton, R. E. , Emery, F. T., Cookson, A. H. A test for the effect of high energy arcs on the flashover strength of insulation in compressed SF$ gas, IEEE Transactions on Electrical Insulation, Vol EI-75, No. 2, 1980, pp.74-80 [14] Greenwood, P. Surface flashover in compressed SFQ , Proc. II Int. Symp. on Gas Discharges and their applications, 1983, pp.233-235 [15] Chenoweth, T. E. , Meoman, F. A., Cookson, A. H. Epoxy castings for UHV service in compressed SFQ gas insulation equipment, Proc. 13th Elec-trical/Electronics Insulation Conference, 1977, pp.322-326 [16] Cookson, A. H., Electrical breakdown for uniform fields in compressed gases, Proceedings of the IEE, Vol 117, No. 1, 1970, pp.269-280 [17] Nitta, T. Factors controlling surface flashover in SFQ gas insulated systems, IEEE Transactions on Power Apparatus and Systems, Vol-97, No. 3, 1978, pp.959-965 [18] Cooke, C. M. Particle contamination in compressed gas insulation, limitations and control, International Symp. on Gaseous Dielectrics, 1978, pp.162-189 [19] Itaka, K., Hara, T., Misaki, T., Tsuboi, H. Improved structure avoiding lo-cal field intensification on spacers in SF$ gas, IEEE Transactions on Power Apparatus and Systems, Vol PAS-102, No. 1, 1983, pp.250-255 179 [20] Skipper, D. J., McNeal, P. I. Impulse fiashover measurements on compressed gas spacers for EHV power cables, Proceedings of the IEE, Vol-102, 1965, pp.102-108. [21] Hampton, B. F. IEE Conf, Publication 44, 1968, pp.61-64. [22] Laghari, J. R., Quereshi, A. H. Surface fiashover in SF$ - N2 mixtures under uniform and non-uniform DC fields, Proc. Conf. on Electrical Insulation and Dielectric Phenomenon, 1979, pp.268-276 [23] Laghari, J. R., Quereshi, A. H., Proc. Canadian Conf. on Communication and Power, 1980. [24] Cookson, A. H., Garrity, T. E. Research and development in the U.S. on a 3 conductor compressed gas insulated transmission lines for heavy load trans-mission lines, Proc. Int. Conf. on High Voltage Systems, CIGRE Paper 23-12, 1978, pp.1-12 [25] Cookson, A. H. Compressed gas insulated transmission systems : the present and the future, International Symp. on Gaseous Dielectrics, 1978, pp.268-313 [26] Mizukami, T. Gas insulated transmission with semi- fabricated unit, IEEE Transactions on Power Apparatus and Systems, Vol PAS-98, 1979, pp.1709-1716 [27] Menju, S. Electrical potential and field of conical spacers used for metalclad switchgear, IEEE Transactions on Power Apparatus and Systems, Vol PAS-92, No. 2, 1973, pp.390-398 [28] Baer, G. R., Diessner, A., Luxa, G. F., J^20kV SFQ insulated tubular bus for the Wehr pumped storage plant : electric tests, IEEE Transactions on Power Apparatus and Systems, Vol PAS-95, No. 2, 1976, pp.469-477 [29] Itaka, K., Ikeda, G. Dielectric characteristics of compressed gas insulated ca-bles, IEEE Transactions on Power Apparatus and Systems, Vol PAS-89, No. 8, 1970, pp.1966-1970 180 [30] Nitta, T., Fujiwara, Y., Endo, F., Ozawa, J. Effects of electrode and solid insulation on flashover in compressed SFQ , Proc. Int. Conf. on Large High Voltage Systems, 1976, Paper 15-04, pp.1-12 [31] Itaka, K., Influence of local field concentration on surface flashover character-istics of spacers in SF$ gas, IEEE Transactions on Electrical Insulation, Vol EI-80, 1980, pp.56-60 [32] Trinh, N. G., Vincent, C. Bundled conductors for EHV Transmission systems •with compressed SFG insulation, IEEE Transactions on Power Apparatus and Systems, Vol PAS-97, No. 6, 1978, pp.2198-2206. [33] Cronin, J. C , Perry, E. R. Optimization of spacers for gas-insulated systems, IEEE Transactions on Power Apparatus and Systems, Vol PAS-92, No. 2, 1973, pp.558-564. [34] Takuma, T., Watanabe, T. Optimum profiles of disc type spacers for gas in-sulation, Proceedings of the IEE, Vol 122, No. 2, 1975, pp.183-188. [35] Vandermeeren, L. Discussion in the paper published in IEEE Transactions on Power Apparatus and Systems, Vol PAS-97, No. 3, 1978, pp.919-925. [36] Eteiba, M. , Rizk, F. A. M. , Trinh, N. G., Vincent, C. Influence of a con-ducting particle attached to an epoxy resin spacer on the breakdown voltage of compressed gas insulation, Proc. II Int. Symp. on Gas Discharges and their applications, 1980, pp.250-255 [37] Banford, H. M. Particles and breakdown in SFQ insulated apparatus, Proceed-ings of the IEE, Vol-122, No. 2, 1976, pp.159-162. [38] Mangelsdorf, C. W., Cooke, C. H. Static charge accumulated by epoxy post in-sulation stressed at high DC voltages, Conf. on Elect. Insulation and Dielectric Phenomenon, 1981, pp.220-226 [39] Laghari, J. R., Quereshi, A. H., it Surface flashover of spacers in compressed gas insulated systems, IEEE Transactions on Electrical Insulation, Vol EI-16, No. 5, 1981, pp.373-387 181 [40] Kuwahara, H., Imamura, S. , Watanabe, T., Arahata, Y. Effect of solid im-purities on breakdown in compressed SFQ gas, IEEE Transactions on Power Apparatus and Systems, Vol PAS-93, No. 5, 1974, pp.1546-1555 [41] Rizk, F. A. M., Masetti, C , Comsa, R. P. Particle initiated breakdown in SF$ insulated systems under high voltages, IEEE Transactions on Power Apparatus and Systems, Vol PAS-98, No. 3, 1979, pp.825-836. [42] Cronin , J.C., International Symp. on Gaseous Dielectrics, 1978, pp.116-137 [43] Report No. DOE(ERDA) E(49-18)-2125, March 1979 [44] Pfeiffer, W., Volker, P. Influence of conducting particles on the DC voltage strength of spacers in compressed SFQ gas, International Symp. on Gaseous Dielectrics, 1980, pp.243-249 [45] Laghari, J. R., Quereshi, A. H. Fiashover studies in SF6 - N2 mixtures us-ing direct applied voltages, International Symp. on Gaseous Dielectrics, 1980, pp.227-235 [46] Lee, R., Rhinehart, H., Thompson, H., Sudarshan, T. S. Pre-discharge current measurements and optical surface field measurements associated with insulator surfaces fiashover, International Symp. on Gaseous Dielectrics, 1982, pp.349-355 [47] Watson, P. K. The electric strength of N2 at elevated pressure and small gap spacing, J. of App. Phys., Vol 40, 1969, pp.328-334. [48] Hogg, P., Schmidt, W., Strasser, H. Dimensioning of SFQ metalclad switchgear to ensure high reliability, Proc. Int. Conf. on Large High Voltage Systems, 1972, paper 23-10. [49] Cookson, A. H., Eaton, E. R. Discussions in International Symp. on Gaseous Dielectrics, 1984, pp.545 [50] Nakanishi, K., Yoshioka, A., Arahata, Y., Shibuya, Y. Surface charging on epoxy spacer stressed at DC stresses in compressed SF6 gas, IEEE Summer Power meeting, 1983, Paper No. 83 SM 306-8 182 [51] Knecht, A. Development of surface charges on epoxy resin spacers stressed with Direct Applied voltages, International Symp. on Gaseous Dielectrics, 1982, pp.356 [52] Ootera, H., Nakanishi, K., Shibuya, Y., Arahata, Y., Nitta, T. Measurement of charge accumulation on conical spacer for 500 kV DC GIS, International Symp. on High Voltage Engineering, Athens 1983, Paper No. 52 [53] Cun-Yi Y., Rusck, S., Vlastos, A. E. The influence of charge accumulation on impulse fiashover of spacer in compressed SFQ , International Symp. on High Voltage Engineering, , 1983, Paper No. 34-06 [54] Cooke, C. M. Surface fiashover of gas/solid interfaces, International Symp. on Gaseous Dielectrics, 1982, pp.331-347 [55] Giao Trinh, G., Mitchell, G., Vincent, C , Influence of an insulating spacer on the V - t characteristics of a coaxial gas-insulated cable, Part I: Study on a reduced scale coaxial conductor, IEEE PES Summer Power Meeting, Mexico, 1986 [56] Mailer, V. N., Srivastava, K. D. Corona discharge studies in 5F 6 with in-sulators under positive switching impulses, IEEE Transactions on Electrical Insulation, Vol EI-18, No. 5, Oct 1983, pp.528-535 [57] Pedersen, A. Criteria for spark breakdown in sulphur hexafluoride, IEEE Transactions on Power Apparatus and Systems, Vol PAS-89, No. 8, 1970, pp.2043-2048 [58] Pfeiffer, W., Gieselmann, M., Volker, P., Spatial and temporal pre - discharge development at insulator surfaces in compressed gases, IEE Gas Discharges, 1980, pp.277-279 [59] Berger, G., Gayet, P., Hutzler, B., Jouaire, J., Senouci, B. Time-lags in SFQ under impulse conditions, International Symp. on Gaseous Dielectrics, 1982, pp.219-226 183 [60] Anis, H., Srivastava, K. D. Prebreakdown discharges in rod-plane gaps in SF$ under positive switching impulses, IEEE Transactions on Electrical Insulation, Vol EI-16, No. 6, 1981, pp.552-563 [61] Stoop, Th., Tom, J., Verhaart, H. F. A., Verhage, A. J. L. The influence of surface charges on flashover of spacers in Gas Insulated Systems, International Symp. on High Voltage Engineering, Athens 1983, PaperNo. 32-10 [62] Hebner, R. E. , Malewski, R. A., Cassidy, E. C , Optical methods of electrical measurements at high voltage levels, Proceedings of IEEE, Vol 65, Nov 1977, pp.1524-1548 [63] Seeker, P. E. The design of simple instruments for measurement of charge on insulating surfaces, J. of Electrostatics, Vol. 1, 1975, pp.27 [64] Foord, T. R. Measurement of distribution of surface electric charge by use of a capacitive probe, J. Phys. E - J. Sci. Inst., Vol 2, No. 2, 1969,pp.411-413 [65] Pedersen, A. On the electrostatic field near a charged surface of an insula-tor with special reference to surface charge probe measurements, International Symp. on Gaseous Dielectrics, 1984, pp.414-421 1975, pp.27-36 [66] Davies, D. K. The examination of electric properties of spacers by surface charge measurements, J. Phys. E - J. Sci. Inst., Vol. 44, 1967, pp.521-524. [67] Diessner, A., Trump, J. G. Free conducting particles in a coaxial GIS system, IEEE Transactions on Power Apparatus and Systems, Vol PAS-89, No. 8, 1970, pp. 1970-1978 [68] Singer, H., Steinbigler, H., Weiss, P. A charge simulation for the calculation of high voltage fields, IEEE Transactions on Power Apparatus and Systems, Vol PAS-93, 1974, pp.1660-1668 [69] Srivastava, K. D., Van HeesWijk, R. G., Particle initiated breakdown at spacer surface in Gas-Insulated bus, CEA Report 154 - T343, 1985 [70] Ootera, H., Nakanishi, K., Analytical method for evaluating surface charge distribution on a dielectric from capacitive probe measurements - Application 184 to a cone type spacer in ±500 kV DC-GIS, Paper No. 86 SM 451-9, IEEE-PES Summer meeting, Mexico City, July 20-25 1986 [71] Singer, H. Flachenladungen zur Felberechnung von Hochspannung system, Bull SEV/VSE 65, 10, 1974, pp.739-746 [72] Sato, S. Zaengl, W. S., Effective 3-dimensional electric field calculation by surface charge simulation method, Proceedings of the IEE, Vol. 133, Pt. A, March 1986, pp.77-83 [73] Pedersen, A., On the electrostatics of probe measurements of surface charge densities, International Symp. on Gaseous Dielectrics, 1987 [74] Bektas, S. I., Farish, O., Hizal, M. , Computation of the electric field at a solid/gas interface in the presence of volume and surface charges, Proceedings of the IEE, , Vol 133, Pt A, No. 9, Dec 1986, pp.577-586 [75] Boeck, W., SFe - insulation breakdown behaviour under impulse stress, Surges in high voltage networks, Ed. K. Ragaller, Plenum Press, NY, 1980 pp.207-226 [76] Van Brunt, R. J., Misakian, M. , Mechanism for inception of DC, 60Hz AC corona in SFe , IEEE Transactions on Electrical Insulation, Vol EI-A, 1982, pp.106-120 [77] Van Brunt, R. J., Misakian, M. , Role of photo-detachment in initiation of electric discharges in SFQ and 02 , J. App. Phy. 54, 1983, pp.3074-3079 [78] Wiegart, N., Pinnekamp, F., Final Report of the CEA contract 153 T310, 1986 [79] Sato, S., Private communications [80] Mahajan, S. M. , Sudarshan, T. S., Dougal, R. A., Avalanches near a solid insulator in Nitrogen gas, presented at the Fifth Int. Symp. on Gaseous Di-electrics, Knoxville, Tennessee, USA, May 3-7 1987. [81] Jaksts, A., Cross, J. D., The influence of a solid dielectric spacer on electron avalanches in Nitrogen gas at atmospheric pressure, Can. Elec. Eng. Jour. Vol 6, No. 2, 1981, pp.149-152 185 [82] Mailer, V. N., Srivastava, K. D., Corona inception phenomena in solid-air composite systems IEEE Transactions on Electrical Insulation, Vol EI-18, No. 4, 1983, pp.402-408 [83] Baum, E. A., Lewis, T. J. , Toomer, R. Decay of electrical charge on polyethy-lene films J. Phys. D:Appl. Phys., Vol. 10, 1977, pp.487-497 [84] Wiegart, N. Corona stabilization to the testing of Gas-insulated switchgear CEA Report # 153 T310. [85] Cooke, C. M. Charging of insulator surfaces by ionization and transport in gases IEEE Transactions on Electrical Insulation, Vol. EI-17, No. 2, 1982, pp.172-177. [86] Giesselmann, M. , Pfeiffer, W., Influence of solid dielectrics upon breakdown voltage and predischarge development in compressed gases International Symp. on Gaseous Dielectrics, ,1984, pp.431-436 [87] Tsuruta, K., Private communications 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0097073/manifest

Comment

Related Items