A summary of streamer propagation in Exxsol-D140 and Marcol-52 under various experimental conditions is given. The results are compared to those of cyclohexane and mineral transformer oil. Streamer propagation is polarity and liquid dependent. The breakdown and acceleration voltages are also governed by liquid chemistry. Reduced pressure largely enhances streamer propagation of both polarities. A low ionization potential additive significantly affects positive streamers while an electron scavenger markedly impacts negative streamers. Adding a low ionization potential additive caused positive streamers to behave like those in mineral transformer oil while adding an electron scavenger had the same influence on negative streamers. Processes at the streamer heads are suggested to be responsible for streamer propagation.

The molecular ionization potential has a relatively strong electric-field dependence as compared to the excitation energies which has implications for electrical insulation since the excited states work as an energy sink emitting light in the UV/VIS region. At some threshold field, all the excited states of the molecule have vanished and the molecule is a two-state system with the ground state and the ionized state, which has been hypothesized as a possible origin of different streamer propagation modes. Constrained density-functional theory is used to calculate the field-dependent ionization potential of different types of molecules relevant for electrically insulating liquids. The low singlet-singlet excitation energies of each molecule have also been calculated using time-dependent density functional theory. It is shown that low-energy singlet-singlet excitation of the type n → π* (lone pair to unoccupied π* orbital) has the ability to survive at higher fields. This type of excitation can for example be found in esters, diketones and many color dyes. For alkanes (as for example n-tridecane and cyclohexane) on the other hand, all the excited states, in particular the σ → σ* excitations vanish in electric fields higher than 10 MV/cm. Further implications for the design of electrically insulating dielectric liquids based on the molecular ionization potential and excitation energies are discussed.

Streamer propagation guided within PTFE tubes were investigated in an 8 cm point-plane gap in white oil. Both positive and negative polarities were studied with a near step voltage of 50-400 kV. Streamers in a free volume were compared with those in tubes. Within tubes, branching of streamers is suppressed, and interaction between streamers and tube wall may occur. For positive streamers, the guiding tube slightly accelerates streamers and reduces acceleration voltage. However, for negative streamers the tube significantly increases the velocity and drastically decreases the acceleration voltage. Possibly, streamer propagation is governed by the streamers tip-field. When the interaction between streamers and the tube wall appears, the electric field at the streamer tip is increased. This leads to an increase in streamer velocity. The mechanism of streamer propagation is discussed.

A numerical 3D simulation of positive streamer propagation in a liquid, using cyclohexane as a reference case, in a needle-plane geometry is presented. The physical basis of this model is a version of the Townsend-Meek streamer criteria, known from gas discharges, adapted to a liquid containing an additive. The propagation of the streamer is modelled in the following way: Electrons located in an electric field above a given threshold (liquid dependent) create electron avalanches in the liquid. The electron avalanches leave behind positive ions that will move the high field region further into the liquid due to the field created by the high charge concentration. Subsequent avalanches in the liquid neutralize this region, but create a new highly ionized region, thus the ionization region (streamer head) moves further away from the needle, while still being connected to the needle by a weakly conducting plasma channel. The stochastic nature of streamer inception, propagation and branching is modeled through a random distribution of seed electrons in the liquid. Key results obtainable from the model includes streamer shape/degree of branching, speed (instantaneous and average), and inception delay of positive streamers as a function of applied voltage. The presented model is crude as it does not take into account the energy required for phase transition, and the dynamics of the plasma channel trailing the streamer head. In addition it is based on a rough estimate of the field in the liquid. This paper presents the model in detail.

Recent experiments show that modern dielectric liquids behave differently from traditional mineral oil, particularly with respect to breakdown voltages for lightning impulse. This paper describes an experimental investigation addressing underlying reasons for this. The influences of reduced pressure and additives on streamers in white oil were investigated under both positive and negative polarities using an 8 cm long point-plane gap. Reduced pressure significantly accelerates streamers, thus increasing stopping length and reducing both breakdown and acceleration voltages. With increasing applied voltage, different typical propagation modes of streamers were recorded for both polarities. A low ionization potential additive strongly affects positive streamers. It significantly changes streamer velocity and reduces the breakdown voltage but increases the acceleration voltage where breakdown streamer velocity increases drastically. Adding an electron scavenger influences streamers of both polarities, but it mainly increases the velocity of negative streamers and results in a reduction of both the breakdown and the acceleration voltages. The propagation mechanisms of streamers are also discussed.

The paper, prepared by CIGRE WG D1.03 (TF 09), presents the guidelines for risk assessment procedure on defects in GIS based on PD diagnostics. The procedure, described in detail in CIGRE Technical Brochure 525, starts with sensitive PD measurement to detect the critical defects and follows with identification of the type of the defect and its location inside the GIS. This information taken together with other essential data from laboratory measurements, manufacturer's experience, design aspects and trend analysis of the PD activity, are the base for the estimation of the criticality of the defects. Finally, the risk assessment is performed based on the estimated dielectric failure probability and failure consequences that can be different in case of on-site testing or in service activity.

This paper presents an experimental study of positive streamers in paraffinic model oil in a point-plane gap. The investigation was performed concerning the influence of voltage, impurities and additives. It was seen how increased voltage changes streamer shape and speed resulting in different propagation modes. Carbon particles markedly reduce inception, breakdown and acceleration voltages, while dissolved gases/air have little effect on these voltages. A low ionization potential additive decelerates the streamers, while an electron scavenger shows an increase in streamer velocity. The correlation between propagation velocity and streamer's branching is discussed.

The streamer channels in Marcol oil have been studied under the positive “step” voltages. Experiments were performed with a point-plane gap of 8 cm. Three kinds of streamer channels were observed around breakdown voltage. These are dark, partly illuminating, and fully illuminating (reillumination) channels. No current could be detected for the dark channels. The partly illuminating channels correlate with small current pulses in some hundreds of milli-amperes. However, the reilluminations are associated with much higher current pulses of some amperes. All kinds of channels occur simultaneously and propagate almost equally fast. Streamers with reilluminations almost always terminate with breakdown. On the other hand, streamers with dark or partly illuminating channels have much lower probability for breakdown. Reilluminations do not have significant effects on the propagation of non-breakdown and breakdown streamers as well as the time to cross the gap and breakdown. In the 2nd mode streamers, 30-50% of their branches are able to become illuminated, and the number of reilluminations as well as illuminating branches is proportional to the applied voltage. The effects of vacuum on streamers were also investigated. With vacuum, streamers travel with much lower voltage drop along the streamer channel. Therefore stopping length greatly increases, and the breakdown voltage drastically drops. All crossing streamers result in breakdown. The number of reilluminations in vacuum is lower than that in ambient pressure. The physics of the above streamer channels is discussed.

This is a study of the effect of the additives dimethylaniline (DMA) and trichloroethylene (TCE) on the pre-breakdown phenomena known as streamers, in a simple base liquid, cyclohexane. This has been done under a step-like impulse voltage in a long point-plane gap. The objective was mainly to find out which aspects of the chemistry of mineral transformer oils cause their good streamer-related dielectric properties, in particular the high voltage for the transition to a very fast mode of positive streamers, as pure cyclohexane does not have these good properties. It is found that DMA added indeed causes positive streamers to become very similar to positive streamers in mineral oil and has no effect on negative streamers. TCE causes the negative streamers to become more mineral-oil-like, and has only a little effect on positive streamers.

Pre-breakdown phenomena in pure n-tridecane in liquid and solid state have been investigated experimentally in a needle-plane geometry with impulse voltage. Light emission and charge injection from the high-field electrode were measured. Pre-inception currents in liquid and solid state were recorded and compared to finite element calculations to find plausible high-field conduction models for n-tridecane and cyclohexane. n-tridecane was chosen as a model system for polyethylene based on the known similarities in energy bands for polyethylene and alkanes. The results obtained in pure n-tridecane were compared to similar experiments on cyclohexane and polymer systems. High-field conduction in liquid n-tridecane was found to have a similar field dependence as previously reported for cross linked polyethylene (XLPE). Conduction currents and inception probabilities were found to be almost independent of polarity in frozen and liquid n-tridecane, which is also in line with results obtained for XLPE. Conduction currents were observed at lower voltages in n-tridecane than in cyclohexane, reflecting the lower space charge limited field (SCLF) in n-tridecane. Inception of streamers in liquid cyclohexane occurs at lower voltages than in liquid n-tridecane, and a stronger polarity dependence for all phenomena was observed in cyclohexane compared with n-tridecane. A possible explanation for the differences between n-tridecane and cyclohexane is given based on the field needed for electron avalanches and the SCLF in the two liquids. Inception voltages, light emission and charge injection were found to be similar in liquid and solid n-tridecane. This indicates that the same processes are responsible for inception and propagation of electrical trees in solids and streamers in liquids when the material is stressed by a fast transient. The similarity between the measured properties in solid and liquid phases leads to the conclusion that electrical treeing mainly takes place in the amorphous regions of the solid phase.

This paper focuses on studying characteristics of streamers in Exxsol D140. The experiments were carried out with 80 mm point-plane gap under step-voltage up to 540 kV. Inception delay time, stopping time and stopping length of streamers is voltage, polarity and impurity dependent. It is seen how increased voltage changes streamer shape and speed resulting in different propagation modes. Impurities markedly reduce inception, breakdown and acceleration voltage. A low ionization potential additive decelerates the positive streamers and accelerates the negative streamers. In fast mode, positive streamers are approximately 5 times faster than negative streamers, and their velocities can reach the value of 110 km/s and 25 km/s at 540 kV for positive and negative polarities respectively. Besides, positive streamers accelerate at 240 kV, while 420 kV is the value of negative streamers.

A fully enclosed oil circulation rig has been made for studying the temperature-dependent equilibrium distribution of contaminants between oil-impregnated cellulose materials and oil. This is relevant for the distribution of contaminants in the solid insulation in power transformers. Used with clean, new material the rig gave humidity equilibrium results reasonably close to results found in literature. With service-aged pressboard the humidity results departed from the values from literature for clean, new material. The results had some inner consistency indicating that they were reasonably trustworthy despite some uncertainty because of long diffusion times in pressboard and some measurement uncertainty. It was also attempted to look into temperature-dependent equilibrium distribution of acids, but the variation with temperature was too small compared to the measurement uncertainty. It was also found that the rig is suitable for studying the temperature-dependent humidity saturation in otherwise contaminated oil at high temperatures.

Prebreakdown phenomena in 0.1 M N,N-dimethylaniline (DMA) and 0.1 M trichloroethene (TCE) in n-tridecane have been investigated experimentally in liquid and solid phases in a needle-plane geometry with impulse voltage. Light emission and charge injection from the high-field electrode were measured. Results have been compared with results obtained in neat n-tridecane and in cyclohexane with the same additives. Adding TCE to n-tridecane increases the pre-inception current and decreases the inception voltage for negative and positive polarity. The increase in pre-inception current is probably caused by trap assisted conduction, and may be responsible for the decrease in inception voltage. TCE had no effect on propagation of streamers in n-tridecane. The main effect of DMA in n-tridecane is to enhance the propagation of streamers and electrical trees with positive polarity. This is in line with what has been reported for cyclohexane, and is explained by the low ionization potential of DMA. Charge injection for negative streamers is reduced by the addition of DMA in n-tridecane, while the emitted light is increased. This indicates that some of the energy otherwise used for propagation is emitted as light. A low electronic excitation energy for DMA at 3.8 eV supports this hypothesis. The additives have the same effect on prebreakdown phenomena in solid and liquid phases with positive polarity, and thus the same mechanism is suggested to be responsible for electrical treeing and streamer inception and propagation. The scatter is generally found to increase when going from liquid to solid phase, which is explained by the inhomogeneity in the solid phase. The transition to solid phase with negative polarity typically results in an even larger scatter than for positive polarity. The effect of additives seems to be secondary to the morphology at the point electrode with negative polarity.

An experimental setup for dielectric testing of insulating liquids in the temperature range from -60 °C to +250 °C has been built. The system uses point-plane geometry, a sensitive differential charge measurement technique, a computer controlled temperature regulator, a photomultiplier and a pulse generator with a fast rise time. This paper presents a detailed overview of the functionality of the system and results obtained in cyclohexane under positive and negative polarity below and above freezing point. Pre-inception current, inception probabilities, streamer charge and emitted light from streamers are reported. The results found in cyclohexane at room temperature compares well to results obtained in other similar setups. A temperature dependence on the injected streamer charge for liquid cyclohexane under positive polarity was observed. No polarity differences were observed for any of the measured quantities in frozen cyclohexane.

The condition assessment of the 190 MVA, OFWF, step-up generator transformer was performed by the mapping of degree of polymerization (DP). The transformer was taken out of service after 36 years and the active part, (transformer core and windings), was removed from the tank for close examination due to the major thermal fault at the transformer core. The transformer insulation consisted of two different materials, i.e. winding insulation was the thermally upgraded paper and spacer-pressboard system was Kraft paper. The temperature mapping is also shown, where the temperature estimation is based on the paper aging kinetics, transformer loading and insulation operating history. New equation for the relative aging rate is given considering all insulation conditions providing possibility of counting transformer loss-of-life more accurately. Copyright (C) 2010 John Wiley & Sons, Ltd.

The role of composite voltage waveforms on prebreakdown processes was studied at atmospheric conditions in a small point-plane gap with a synthetic ester Midel 7131 fluid, representing a model for the insulation in power electronic components. Shadowgraphic imaging and photomultipliers to record emitted light were used to measure streamer initiation voltages, streamer propagation lengths and velocities. The applied voltages were step voltages (100 ns rise time) with or without a dc bias. Results have demonstrated the presence of space charges in the vicinity of electrodes. With a step voltage with the same polarity as the dc prestress, initiation voltages increased and propagation lengths were reduced. For a polarity reversal for positive steps the initiation voltages were reduced and propagation lengths increased as expected, while for negative step the opposite was seen. The latter is explained by an electron "cleaning" effect of the positive dc prestress reducing possibilities for electron avalanches. It is shown that streamers may be initiated by a sudden turn off of a dc stress.

The effects of electron-attaching and electron-releasing additives in cyclohexane on the initiation and propagation of positive and negative streamers have been studied quantitatively. Fast impulses (<20 ns, <40 kV) were applied on a 10 mm point-to-plane gap and studied by shadowgraphic imaging and a differential charge measurement technique. The properties of both positive and negative streamers depend on the specific electronic characteristic of the additives. Electron-attaching additives facilitate the propagation of negative streamers, whereas the most effective electron-releasing additives reduce initiation voltages and facilitate the propagation of positive streamers. Depending on the reactivity and concentration of the additives, streamer filaments become thinner and fewer while propagating faster and further. The Townsend-Meek theory for streamer inception in gases has been adapted to a solution and applied to analyze the voltage dependence of the positive streamer propagation. Results show a quantitative dependency on the ionization potential and concentration in agreement with experimental trends.

The effect of dc prestress on positive streamer initiation and propagation was investigated at atmospheric conditions in a small liquid-filled point-plane gap, representing a model for the insulation in power electronic components. Shadowgraphic imaging and photomultipliers to record emitted light were used. The applied voltages were step voltages (100 ns rise time) with or without a dc bias. The results obtained under step voltage show that many features of positive streamers (shape, velocity, initiation voltage) in Midel 7131 fluid are the same as those previously observed in hydrocarbon and mineral oils in similar test conditions. The presence of space charges in the vicinity of electrodes was demonstrated. The superposition of a same polarity dc prestress on the applied step voltage results in the increase of streamer inception voltage and the decrease of propagation lengths. For a polarity reversal for positive steps the initiation voltages were reduced and propagation lengths increased.

• Uncertainty in effectiveness of methods available to assess a health of an equipment. In [5] e.g. is shown that in spite of low water content in the oil indicated by normal oil sampling procedures a large amount of water can be revealed within a transformer. How to rank the transformer, which needs drying? How to assess a critical level of water contamination to prevent failure or to determine permissible operating conditions, e.g. overloading?