IEEE Transactions on Power Apparatus and Systems, Vol. PAS-101, No. 6 June 1982 RANDOM PREBREAKDOWN DISCHARGES IN SF6 - A POSSIBLE DIAGNOSTIC CRITERION FOR PARTICLE-CONTAMINATED CONPRESSED GAS APPARATUS H. Anis University of Waterloo Canada M. M. A. Salama Ain Shams University Egypt K. D. Srivastava University of Waterloo Canada Abstract - The corona pulse voltage in SF6 gaps, under impulse voltages, is random in value; its random- ness is related to the electrode geometry and is influ- enced by the gas pressure and the rate of rise of the applied voltage. An analytical model is presented whereby the relationships among these quantities are derived. The model predicts for a given electrode ge- ometry at a certain gas pressure the distribution of the corona onset voltage. The results of testing rod- plane SF6 gaps under switching impulses are presented to verify the applicability of the analytical model. By establishing an electrostatic equivalence between rod-plane gaps and conducting particles in GIS, the above analysis could be extended to the latter problem. The possibility of using the present results and analy- sis to devise a diagnostic test procedure for particle- contaminated GIS is discussed. INTRODUCTION It is believed that partial discharges in gas-insulated switchgear (GIS) are indicative of a possible future breakdown in the systeml. A major cause of these par- tial discharges is the presence of conducting particles in the gas - a situation which could result in as much as 90% reduction in the system's breakdown strength. Two methods are currently in use to check for partial discharge activity; namely, ultra-sonic contact probes and electromagnetic coupling devices1. The two methods detect the presence of the discharge and its intensity under operating ac voltage. No clear indication how- ever may be gathered about the geometry of the conduct- ing particles from these measurements. It would be use- ful to obtain such an indication in view of the con- flicting roles of particle geometry in corona discharge and breakdown. These roles have been examined experi- mentally2 and modelled analytically elsewhere3. The "icritical" breakdown voltage for a given particle length may increase or decrease with the particle di- ameter depending on the particle's length - unlike the partial discharge voltage which decreases steadily with an increasing particle diameter. While long particles are more hazardous when their radii are small, short particles increase the chance of a breakdown when they are large in diameter3. One way of detecting the geometry of the conduct- ing particles is by making use of the randomness in the voltage at which a prebreakdown discharge current pulse occurs under impulse voltages. The present work analytically examines the link between electrode geome- On leave from Cairo University, Egypt 81 TD 601-4 A paper recommended and approved by the IEEE Insulated Conductors Committee of the IEEE Power Engineering Society for presentation at the IEEE PES 1981 Transmission and Distribution Conference and Ex- position, Minneapolis, Minnesota, September 20-25, 1981. Manuscript submitted March 26, 1981; made avail- able for printing June 17, 1981. try and random discharge pulse voltage. This is veri- fied against experimental results for point-plane gaps in SF6 under positive switching impulses. An electro- static analogy is worked out between conducting par- ticles in SF6 switchgear and laboratory point-plane gaps. Some recommendations are offered towards a pos- sible use of the studied phenomena in devising a diag- nostic test procedure. PHYSICAL BACKGROUND OF THE PROBLEM In a nonuniform field gap in SF the initial discharge, normally referred to as corona discharge, takes place when conditions for the formation of a streamer are fulfilled. In uniform or quasi-uniform fields the fulfillment of these conditions produces a complete breakdown in the gas. The calculation of the applied voltage at which a streamer formation occurs is a complex task in view of the many physical processes to be accounted for; e.g., space charge field, photoioniz- ation, attachment and ionic diffusion etc. For small gaps in air a simplified approach of the energy in a single critical avalanche has also been used.4 Under impulse voltages, the random time delay to discharge initiation, mwiy be expressed in terms of three separate components. Firstly, the voltage must reach a level sufficient to cause growth of electron avalanches. Secondly, an initiatory electron must be available, anc finally, the growth of an avalanche along any partic-- ular point will take a finite time. If it is assumed that free electrons are produced largely by collisional detachment, the availability of initiatory electrons would depend upon the spatial electric field distrib- ution around the anode, and hence it is a function of anode geometry and the impulse voltage waveform. Such considerations have led some researchers to propose the concept of "critical volume" near the anode for discharge initiation under positive impulse voltages.7 The critical volume vs. time criterion, however, does not take into account the probability distribution of discharge growth along different paths within the overall active volume. There are rigorous methods for calculating probab ility density function for discharge development in a gas. These, however, tend to be quite cumbersome to evaluate. In electronegative gases two major processes dominate, namely, the impact ionization (oc) and atta- chment (r). In SF6 net growth is possible only when a = (c-rI) is positive, and the necessary field and pressure for the net ionization to be positive is given by the limiting value (E/P)Z = 89 kV/cm. atm. Both an and rn are functions of pressure and the electric field. The variation of the surface electric field for a number of rod diameter-to-gap ratios (d/G) is shown in Fig. 1. The per unit field e is related to the actual field E by E = V . The algorithm for field e is based on a charge simulation technique. The rod electrode is assumed to be infinite in length and terminates in a hemispherical end.. 0018-9510/82/0600-1588$00.75 C 1982 IEEE 1588