Partial Discharge Imaging Correlated with Phase-Resolved Patterns in Non-Uniform Electric Fields with Various Dielectric Barrier Materials

Abstract

This paper describes a correlation of partial discharge phase-resolved patterns with an optical imaging performed in a non-uniform electric field configuration. The influence of different dielectric barrier materials, placed on the plane electrode, on the discharge propagation and surface landing was investigated. The investigations were focused on the corona at positive polarity of AC high voltage. It was found that the initial positive corona stage is similar for all cases whereas the discharge propagation and surface landing strongly depends on the barrier material properties. The observed streamer discharge modes have been described by the geometrical measures such as stem length, stretch of a discharge profile on the dielectric barrier surface and an hemispherical envelope of discharge filaments. Since various dielectrics reveal different properties of charge accumulation and surface neutralization, the charge memory effect may be visible and can be related to the ability to create and sustain of additional electric field component. It may refer to subsequent discharges as well as to conditions faced at the voltage polarity reversal. The correspondence between different forms of phase-resolved patterns have been associated with the modes of streamer discharges observed by optical imaging. Presented methodology poses huge potential for both scientific investigations on underlying discharge phenomena as well as on the application in future diagnostic systems of HV insulation.

1. Introduction

The background of this research is related to further understanding the discharge propagation mechanism. Air at atmospheric pressure is a very popular insulation medium in high voltage equipment, which is becoming more and more compact and operates on lower margins compared to the criteria in the past, due to our better understanding of the underlying mechanisms. This refers, for example, to both overhead lines, switchgears, dry-type transformers, insulators where usually long gaps are considered with a leader discharge mechanism as well as to micro distance or inclusions where Townsend or streamer to spark discharges are observed. The required high reliability of HV power equipment implies understanding of discharge inception, development and propagation in a strongly non-uniform electric field. Enormous research progress has been achieved in the last century since the first avalanche discharge theory (i.e., ionization process started by one electron) was proposed by Townsend in 1915 [1] and further enhanced by Loeb, Rather, Meek, Craggs and others [2,3,4]. The progress was related to introduced space charge and subsequent forms of discharges, such as streamer and leader. If a sufficiently high voltage is applied between a point and a plate in atmospheric air, a corona discharge is ignited, resulting in an ionic wind flowing between electrodes [5]. The corona effect and streamer propagation in the point to plane configuration are also subject of a numerical modeling (e.g., [6,7]). To simulate a streamer propagation in air several aspects should be considered, such as electron and ion transport as well as the effects of ionization, attachment, recombination and photoionization. The comparative phase-resolved analysis of AC corona discharges at very low frequency and power frequencies are shown in [8] and the effects under non-sinusoidal voltage excitation in [9].

This paper presents discharge imaging phases in a strongly non-uniform electric field, represented by a point-plane configuration. The arrangement is applicable to mimic both external discharges in an insulation system and the internal ones caused by microblades in voids in solid insulation. The partial discharges (PDs) have been acquired electrically in the phase-resolved mode and were simultaneously observed optically, recording the corresponding discharge phases and streamer development [10,11]. Hence, the new aspect is related to the correlation of optical imaging with PD forms appearing in a phase-resolved pattern. Electrons drifting to the anode under applied electrical field not only produce new electrons by ionizing collisions, but also excite radiation and those photons lead to light emission which can be externally recorded. Emission spectrum of partial discharges light is in the range 300–800 nm wavelength. The spectra of corona discharges show higher intensity in the 300–500 nm range, whereas the spectrum of spark-type discharges shows its main intensity in the 370–400 nm wave range [12,13,14].

The research was concentrated on the corona inception and development towards dielectric barriers on the plane electrode side at positive polarity of high voltage. The influence of different dielectric barrier materials on the discharge propagation and surface landing was investigated. Since various dielectrics reveal different charge accumulation and surface neutralization properties the corresponding set of materials was selected. The surface phenomena may have different origin such as ionic drift, conduction along the barrier surface and recombination, or trapping into dielectric or bulk conduction. In this context, the charge memory effect may visible and can be related to the ability to create and sustain of additional electric field component [15,16,17]. It may refer to subsequent discharges as well as to conditions faced at the voltage polarity reversal.

The correspondence between different forms of phase-resolved PD (PRPD) patterns have been associated with the modes of streamer discharges observed by optical imaging. Presented methodology poses huge potential for both scientific investigations on underlying discharge phenomena as well as on the application in future diagnostic systems of HV insulation.

2. Partial Discharge Mechanism in Strongly Non-Uniform Electric Field

Discharges in a non-uniform electric field develop depending on the mechanism of their initiation in the air or in the solid dielectrics. Thus, studies of the point-to-plane setup are related:

▪

in the first case, to various forms of discharge in the air at electrode distance from sub mm to several dozen cm, and at large distances to strong corona discharges,

▪

in the second case, to discharges in the solid dielectrics (e.g., in polymers, developing according to the electrical treeing mechanism).

Figure 1 presents model electrode arrangement with marked areas of strong electric field and ionization zone S_j in air.

The partial discharge mechanism in strongly non-homogenous electric field is usually represented by a point-plane electrode configuration, shown in Figure 1. In a strongly non-uniform electric field breakdown in gases is always preceded by pre-breakdown in form of discharges, called corona. The capacitance C_j represents the ionization zone, in the equivalent circuit in Figure 1, where the space charge is formed by electron and ion clouds.

The original avalanche model proposed and elaborated by Townsend [1] has been extended for streamer discharges, taking into account space charge, by Meek, Craggs and Rather [3,4]. The radius r_j of the ionization zone S_j is defined by electric field distribution so that the ionization coefficient is bigger than the attachment one: α ≥ η. Discharge phases in strongly non-homogenous electric fields in gas are initiated with a discharge around the tip of the electrode with high curvature, causing ionization and creating a region of plasma around the electrode. When the gas close to the tip becomes conductive, it has the effect of an apparent size increase of the conductor. The generated ions move in the electric field or recombine to form neutral gas molecules radiating photons.

One can distinguish several phases during the discharge development process: an initial one called glow, followed by a corona effect, referred to as single-electrode discharge as opposed to a two-electrode discharge when bridging of the electrodes by an electric spark occurs. In the case of a strong source supplying current, a spark will evolve into a continuous arc discharge. Corona discharges can be formed when the electric field at the surface of the tip exceeds a critical value. Usually the process is started when a neutral atom or molecule in the region of strong electric field (electrode tip) is ionized by a natural environmental event such as an ultraviolet photon or cosmic ray particle and in effect a positive ion and a free electron are created. Both particles receive kinetic energy in electric field and are accelerated in opposite directions i.e., electrons toward anode and positive ions towards cathode. The high electron versus ion mass asymmetry results in electron much higher velocity, i.e., gaining energy allowing for ionization while hitting another atom and causing chain reaction visualized as an avalanche process. Externally visible glow of the corona is caused by electrons recombining with positive ions to form neutral atoms and release of photons of light. In turn, the released photons may ionize other atoms, supporting the creation of electron avalanches. In case the ions reach the cathode, combine with electrons from the electrode, to become neutral atoms again.

The discharge mechanism, i.e., movement of the charges depends strongly on the tip electrode polarity, thus one can distinguish both positive or negative corona discharges. This effect and asymmetry results from the great difference in mass between electrons and positively charged ions and in fact only the electron is having the ability to undergo significant energetic collisions. The investigations presented in this paper are focused on the positive corona which occurs when the point electrode has a positive potential against a plane electrode and is initiated by an ionization event in a region with a high potential gradient. The electrons in a positive corona are concentrated close to the tip, in a region of the high electric field strength, resulting in high electron energy. In this way, all electrons are attracted inward toward the positive electrode and the ions are repelled outwards, towards the cathode. By undergoing collisions in the proximity of the tip electrode, further molecules are ionized, forming an electron avalanche. The electrons resulting from the ionization of an air molecule are then electrically attracted back towards the tip electrode. Distribution of the space charge during discharge at positive polarity of the corona electrode is shown in Figure 2. The dotted line represents the electric field without space charge across the gap while the solid line indicates the distorted electric field. The high field region is moving in time further into the gap extending the region for ionization. The field strength at the tip of the space charge may be high enough for the initiation of the cathode directed streamer which subsequently may lead to complete breakdown [18].

The difference between positive and negative coronas, in the matter of the generation of secondary electron avalanches, is that in a positive corona they are generated by the gas surrounding the plasma region—the new secondary electrons travelling inward—whereas in a negative corona they are generated at the tip of a point electrode itself—the new secondary electrons are moving outward. Thus, the rise time of current pulse is related to avalanche formation and electron current, whereas the falling slope of the pulse refers to the ionization current [3,18,19].

At AC voltage, in the configuration presented above i.e., when the point electrode is connected to a high voltage (HV) source and plane electrode is grounded, the corona starts usually first at negative half period (around the phase angle 270°) and while increasing the voltage the positive corona appears around the phase angle 90°, manifesting higher discharge magnitude, comparing to negative ones. In the first phase of corona development the inception voltage is determined by the curvature of the electrode and only in further stages the influence of the opposite electrode material will reveal. From the engineering point of view the corona inception electric field E₀ was determined experimentally by Frank William Peek in 1929 in the arrangement of concentric cylinders and can be derived from the following formula [20]:

$$E_0=21.6\left(1+\frac{0.301}{\sqrt{\mathsf{\delta }r}}\right)$$

(1)

where: r—radius of the corona electrode, δ—relative density of air.

The electric field strength at the distance x from tip in the point-plane configuration gives the following equation [21]:

$$E\left(x\right)=\frac{2U}{\ln \left(\frac{4a}{r}\right)}\times \frac{1}{2x+r-\frac{x^2}{a}}$$

(2)

where: U—applied voltage, a—distance between point and plane electrodes, r—radius of the point tip.

The relationship between the discharges’ repetition rate and voltage is influenced by the movement of negative ions in the interelectrode space. The discharge current is determined by the movement of negative ions toward the anode. In turn, the residence time of negative ions is the time from their formation in the area described by the distance x until they are neutralized at the anode. Thus, the residence time of negative ions t_r in the space between the electrodes is approximately equal to the following expression [21]:

$$t_r\cong \frac{a^2}{3}\frac{\ln \left(\frac{4a}{r}\right)}{U\times {\vartheta }_n}$$

(3)

where: ${\vartheta }_n$ is the negative ion mobility.

For example, oxygen ions formed from the connection of electrons have the mobility around 2 × 10⁻⁴ m²V⁻¹s⁻¹. The mobility of charge carriers is defined as the rate of flow of charged particles under the influence of electric field. Hence, for the electrode distance a = 20 mm, high voltage electrode radius r = 86 μm, the residence time at inception voltage U₀ will be equal to t_r = 500 μs, which is in very good proximity to the measured value t_r = 235 μs obtained for the same gap between electrodes at voltage U = 15 kV and electrode radius r = 40 μm.

3. Experimental Setup, Instrumentation and Specimens

Experiments have been performed using the point-plane configuration shown in Figure 3. The research was concentrated on the corona inception and development towards dielectric barrier at positive polarity of the point electrode. The influence of different dielectric materials on the plane electrode on the discharge propagation was investigated. The electrodes were made by stainless steel. The inter electrode air gap a was equal to 20 mm. The radius of a tip of a point electrode was 40 μm. The experiments were conducted at atmospheric pressure 0.1 MPa at room temperature.

The PD imaging and current pulses were measured in a setup presented in Figure 4. The partial discharge pulses were recorded using an ICM+ wideband detection system, connected to a host computer via a GPIB bus. The measurements were carried out in a phase-resolved mode resulting in D (φ, q, n) pattern acquisition (8 × 8 × 16 bit). The measurement time of PRPD pattern was 60 s. Synchronously to electrical acquisition the coupled optical observations by iCCD camera were performed. The imaging system consists of an iCCD iSTAR camera model DH334 (Andor, Oxford Instruments, Oxford, UK) equipped with a matrix CCD 1024 × 1024 pixels and an intensifier with spectral quantum efficiency in the range 200 ÷ 850 nm. The camera gain was set to 4000 and exposure time was varying up to 5 ms. In addition extension rings (adjustable tubes up to 10 mm) were installed on the lens in order to obtain clear focus at a discharge spot. The high voltage was obtained from a Model 20/20B HV amplifier (Trek Inc., New York, NY, USA) controlled by a programmable generator. The generator was also connected to a digital delay line (model SR DG645, Stanford Research, Sunnyvale, CA, USA), which triggered the camera. The predefined delay time, trimmed to the discharge capture, was adjusted with respect to the zero-crossing of high voltage. The partial discharges were acquired by high frequency current transformer CT1 terminated with 50 Ω.

The high voltage was obtained from the HV compensated divider (1:1000), denoted in Figure 4 as R₁ and R₂. In this way the PD clusters acquired electrically in a phase-resolved mode can be associated with a corresponding discharge image. The adjusted gate signal triggers camera recording. Exemplary waveforms are shown in Figure 5, where the yellow trace represents the high voltage, green the gate signal, blue the camera window and lilac the PD pulses. The snapshot obtained from the optical imaging refers to the individual discharge recorded within the shutter opening. It was observed that the recorded optical images are representative and repeatable for the distinctive clusters in the PRPD pattern, of course taking into account the stochastic character of discharges.

The experiments were performed at sinusoidal voltage at frequency 50 Hz up to a peak voltage of 20 kV. While increasing the voltage inception voltage occurs first at the negative polarity of the tip, upon further increasing the voltage the discharges also appear at the tip with positive polarity. The investigations presented in this paper are focused on positive corona and barrier discharges. The dielectric barrier was located on the plane electrode on the ground side. In the experiments the following dielectric barrier materials were investigated:

• EPX9—Epoxterm thermosetting insulation, polyester foil impregnated in epoxy,
• GLASS—used as a reference material,
• XLPE—cross linked polyethylene,
• PVC—polyvinyl chloride,
• EPR—ethylene propylene rubber.

The specimen specifications: electric permittivity ε_r, loss factor tgδ, surface resistivity ρ_s and thickness d measured by the authors are presented in

Table 1. Specification of the specimens.

Specimen		Ɛr 50 Hz	tgδ 50 Hz	ρs (Ω)	d (mm)
EPX9		1.95	0.0076	4.36 × 1012	1.64
GLASS		7.30	0.043	4.9 × 1011	3.16
XLPE		2.35	0.0012	>2 × 1013	0.87
PVC		2.24	0.0017	6 × 1012	1.65
EPR		3.30	0.0019	>2 × 1013	0.80

. The measurements of electric permittivity and loss factor were performed using a type 1260A impedance analyzer (Solartron, West Sussex, UK) with a 1296A dielectric interface system, whereas for the surface resistivity measurements a Trek 152-1 instrument was used, with a test voltage of 100 V.

4. Imaging of Corona Discharges and Correspondence with PD Phase-Resolved Patterns in Configuration with Dielectric Barrier

Electrical recording and fast imaging have been implemented to analyse the discharge stages in air and the correspondence between phase-resolved patterns obtained electrically with corresponding discharge images recorded optically. In the presented experiments, the main focus was on the observations and analysis of discharges occurring in the positive half period. PD pulses and optical images were synchronously recorded from inception phase up to breakdown (BD).

The discharge phenomena begins with the propagation of weak emissions from the HV point-tip in the form of tiny filaments followed by a glowing dot. An exemplary sequence of discharge modes in a point-plane setup with barrier is shown in Figure 6. Depending on the barrier dielectric material, the transition between the discharge modes is observed, especially the discharge landing profile and span on the surface.

The filaments start from the point tip and propagate in various directions towards plate. The morphology of discharges reveal various of those modes (Figure 6): Starting with a corona streamer at a tip, followed by tree-like streamer with branching. There is at certain stage characteristic formation of a stem. The stem acts as a prolongation of a tip and new streamer corona is observed at the stem forefront. [22]. In the following stage, while increasing the voltage, dielectric barrier discharges (DBD) occur. The role of dielectric surface material was analyzed observing interactions between filaments. When the streamer is reaching the surface of the barrier the filament channels are spreading over the dielectric. The discharge landing foot reflects above mentioned properties of dielectric materials. The bright, high conductive channel due to the temperature of plasma between tip and barrier was created.

The observed streamer discharge modes can be described and distinguished by the geometrical measures such as stem length l and stretch s of discharge profile on the dielectric barrier surface, as graphically depicted in Figure 7. The envelope of discharge filaments can be established (Figure 7a) in form of a semi hemispherical with defined radius r_e [23,24].

The synchronized camera and electrical observations confirm that more than one streamer may occur in the half period of AC high voltage cycle. The proceeding streamer may thus influence the subsequent ones, changing the ionization and discharge development environment. It can be in the form of residual charges or the gaseous channels, which may have not enough time to dissipate [25]. In this way the electrical field could be distorted and further influence the inception and propagation of subsequent discharges. In some cases, one may notice the streamer initiated not from the point tip but from the point-side. It results from the effect of space charge made by previous streamers which didn’t reach the opposite plane electrode [26,27]. The semi breakdown is observed when the discharge has reached the barrier on grounded plane electrode and high conductive channel is formed in the inter-electrode space. The barrier material limits also the short-circuit effect.

The real novelty of the presented methodology is in the correspondence between optically detected images coming from the radiation processes in a streamer and electrically acquired impulse discharge current. This relationship is illustrated in Figure 8, first with a discharge phase-resolved pattern with superimposed optical images of streamers with filaments, followed by the oscilloscope snapshot (Figure 8b) with captured individual discharge (blue trace) within gate window (green trace). In the phase resolved PD pattern there are denoted two discharge forms with corresponding times t₁ and t₂ elapsing from zero crossing of high voltage. The correspondence between electrical PRPD pattern and optical imaging observation is depicted as 1 and 2 in Figure 8a.

The different dielectric materials acting as a barrier were subjected to the high voltage in point-plane configuration. Depending on the material properties various discharge patterns recorded optically by fast imaging were observed. For certain barrier materials such as GLASS, XLPE the positive streamer channels develop as tip-branched while for the PVC, EPR the stem mode is observed. The streamer development and landing surface pattern strongly depends on the barrier material. This dependence can be attributed to the material electric permittivity, thickness and surface properties related to charge accumulation. The accumulated charges influence inception conditions for successive discharges as well as the availability of seed electrons and conditions for discharge inception after polarity reversal.

The observed evolution of a discharge pattern in point-to-plane arrangement for positive voltage polarity is graphically presented in Figure 9. At corona onset voltage (5 kV) the discharges are recorded at the crest of sinusoidal voltage waveform. While increasing the voltage to 10 kV, the discharge pattern is drifting toward earlier phase angle (zero crossing of high voltage), exhibiting simultaneously higher magnitudes. Further increase of applied voltage results in the appearance of second PD form distributed around the phase angle 90°, preserving the first form, which has drifted further towards HV zero crossing. The magnitude of the form 1 is usually much smaller than form 2, except the EPR barrier material, where the peak values of both forms where almost equal.

The coupled phase-resolved PD patterns and optical imaging obtained at 13 kV in point-plane configuration with various barrier materials: GLASS, XLPE, PCV, EPR are shown in Figure 10. Thanks to synchronization with phase resolved electrical recording the visible streamers can be associated with the discharge pattern in the positive AC half period. In this way two forms can be distinguished, denoted as form 1 and form 2, respectively.

Form 1 refers to the early corona streamers, while form 2 is activated around the crest voltage of the sinusoidal waveform. Form 2 corresponds also to the breakdown channel. The corona onset (form1) is present in all cases and is triggered at t₁ = 0.95 ÷ 1.0 ms after high voltage zero crossing, while the form 2 occurs at t₂ = 4.4 ÷ 5 ms, only for PCV and EPR barrier material at this voltage level. The form 1 streamer mode is tip-branched, except EPR material where individual channel is visible, reflecting origin of stem-branched mode, gradually changing into branch type at the barrier surface. In the next sections, the coupled phase-resolved PD patterns and optical images are shown, measured with various barrier materials in a function of applied voltage from inception to the breakdown. The first example presented in Figure 11 refers to a glass plate.

Form 1, corresponding to the onset of positive corona starts at voltage 5 kV at the peak (t₁ = 5 ms). At 10 kV the pattern has moved towards earlier phase angle (t₁ = 1.6 ms) and is present also at 15 kV and 20 kV, moving ahead with t₁ = 0.7 ms and t₁ = 0.38 ms, respectively. For the last two voltage values the individual streamer channels are reaching the plane electrode. Strong form 2 is visible at 15 kV and 20 kV. At 15 kV the discharge mode is a combination of long stem-branched and tip-branched, with broad stretch at the glass barrier surface. At 20 kV the single, compact breakdown channel can be noticed, with narrow surface foot. The form 2 mode is triggered at t₂ = 3.7 ms at 15 kV and t₂ = 3 ms at 20 kV.

The next example presented in Figure 12 refers to XLPE barrier material. Up to the voltage 20 kV only form 1 is visible. At 20 kV form 2 appears as a very strong tip-branched breakdown with a very strong main vertical plasma channel. The time is moving from t₁ = 5 ms for 5 kV, through 0.7 ms for 15 kV up to HV zero crossing at 20 kV. The form 2 at 20 kV occurs at time t₂ = 3.8 ms, moving towards the main cluster at positive voltage peak. The phase-resolved PD pattern is in this case vertically blurred and corresponds to the branched breakdown channels, whereas in the case of glass barrier the PRPD pattern was much broader in phase, however resulted in very narrow breakdown channel. The streamer mode is in general tip-branched. The breakdown mode has stretched and stable multipoint foot. The glow at the point tip is very strong and bright.

The following example presented in Figure 13 exhibits PVC barrier material. Form 1 has its inception as in other cases at 5 kV and is moving towards voltage zero crossing while increasing the voltage to 10, 15 and 20 kV with t₁ time 1.5, 0.70 and 0.55 ms respectively. The form 2 starts with PVC barrier already at 10 kV as a tiny glowing dot visible in imaging and blurred around peak voltage of the PRPD pattern. The mode 2 is first tip-branched at 15 kV and is switching to long stem-branched at 20 kV. The surface stretch of discharge foot is very broad. The timing t₂ of form 2 is equal to 4.7 ms, 4.4 ms and 3.5 ms for voltage 10, 15 and 20 kV, respectively.

The correspondence of discharge imaging and PRPD for EPR barrier is shown in Figure 14. The onset voltage for form 1 is at 5 kV with individual streamer. At 10 kV both forms 1 and 2 are active. While rising the applied voltage to 10, 15 and 20 kV, the form 1 is moving towards voltage zero crossing with reduced time t₁ in a following sequence 2.1, 0.95 and 0.22 ms. The form 2 starts already at 5 kV, however it was not captured optically. With the EPR barrier already at 10 kV strong tip-branched streamers are reaching the plane side.

At 15 kV the form 2 transforms into stem-branched discharges, with few plasma channels. In that context, interesting observation refers to form 1 triggered at t₁ = 0.95 ms at 15 kV and having scythe like shape, with high magnitude, reaching the level of form 2. In this way scissors-like tip-branched discharges are observed, revealed previously at 10 kV as form 2. At 20 kV the discharge form represents a mix of tip-branched mode, consisting of stem-branched filaments. The surface stretch of discharge foot is very intensive comparing to glass or PVC cases. The timing t₂ of form 2 is equal to 4.0 ms, 3.3 ms and 4.05 ms for voltages of 10, 15 and 20 kV.

The observation of the EPX9 barrier illustrated in Figure 15 reveals a different behavior. The corona onset voltage for a form 1 is equal to 5 kV (t₁ = 5 ms) and is moving at 10 kV to t₁ = 1.7 ms and further for 15 kV and 20 kV to t₁ = 0.7 ms and t₁ = 0.38 ms, respectively. The discharge intensity is relatively low compared to the previous dialectic materials and resembles the glass case also in respect to the streamer forms. The form 2 starts at 15 kV (t₂ = 4.2 ms) with a narrow plasma channel and tiny branches. This individual streamer channel, originating at time t₁ = 0.38 ms, is visible as form 1 at 20 kV and is transformed to form 2 starting before peak voltage at t₂ = 3.55 ms. The form 2 is recorded in this case as a tiny filament between point tip and EPX9 barrier. The stem like branches are released mostly just above the plasma foot towards dielectric surface.

Parameters of streamer modes recorded for various barrier materials are presented in

Table 2. Parameters of streamer modes recorded for various barrier materials.

Barrier Material	Stem Length l (a.u.)		Stretch of Discharge Profiles (a.u.)			Radius of Envelope re (a.u.)			Form 1 t1 (ms)		Form 2 t2 (ms)
	15 (kV)	20 (kV)	10 (kV)	15 (kV)	20 (kV)	5 (kV)	10 (kV)	15 (kV)	15 (kV)	20 (kV)	15 (kV)	20 (kV)
GLASS	0.39	-	-	F2 = 2.4	F2 = 0.35	0.92	1.28	2.0 BD	0.7	0.38	3.7	3.0
XLPE	-	0.43	-	F1 = 0.26	F1 = 0.37 F2 = 2.86	0.84	1.25	1.69	0.7	0.06	-	3.0
PVC	0.37	0.84	-	F2 = 2.55	F2 = 1.78	1.03	0.69	1.21	0.7	0.55	4.4	3.5
EPR	0.65	0.42	F2 1.63	F1 = 0.63 F2 = 1.44	F2 = 3.51	0.89	0.72	2.0 BD	0.95	0.40	3.3	4.1
EPX9	0.38	-	-	1.94	0.70	0.81	0.92	1.4	0.7	0.38	4.2	3.6

BD—breakdown, F1—form 1, F2—form 2.

. The classification has been performed with respect to the radius of an envelope r_e, stretch of discharge profile s measured on the plane barrier surface and a stem length l. All above parameters are expressed in arbitrary units. In addition, two time parameters are considered, i.e., the time from zero crossing to the appearance of the form 1 and form 2, measured in milliseconds and denoted as t₁ and t₂. Those parameters are measured at the voltage in the range 5 kV to 20 kV.

From the imaging profiles and results listed in Table 2 several conclusions can be drawn with regards to discharge development in the arrangement containing dielectric barriers placed on the plane electrode in a distance a = 20 mm from the high voltage point electrode. At the corona onset voltage equal to 5 kV the outreach of streamers, measured from the tip as a radius of an envelope r_e, is in a range from 0.81 (EPX9) to 1.03 (PVC). Rising the voltage to 10 kV results in the extension of that radius for GLASS, XLPE and EPX9 material, but also shortening in the case of PVC and EPR. Further voltage increase to 15 kV, yields the prolongation of the radius for form 1 for all investigated materials.

The stretch of the discharge profile measured on the barrier surface can be observed in fact at 15 kV. At 10 kV only in case of EPR the breakdown of the inter electrode gap occurs and the stretch could be measured. There is also substantial span between those values, i.e., rather broad for GLASS and PVC. At the 20 kV, when strong plasma channel is created GLASS, XLPE (form 1) and PVC manifest rather narrow footprint, whereas EPR, EPX9 and XLPE (form 2) exhibit broader spot. The low stretch value corresponds to the direct vertical channel without branching on the surface.

The discharge mode containing stem is present almost in all cases except XLPE. Most prominent are however the cases of PVC and EPR barrier materials. The stem is visible mainly below 20 kV, where mostly strong plasma channel is dominating.

Depending on the surface and dielectric properties the dielectric barrier has impact on the charge accumulation effect and in response modulation of the streamer development through the space charge. In this way one could observe the tip-branched discharges or formation of a stem. The surface conductivity determines also stretch of the discharges around the discharge footprint, leading to local electric field attenuation or enhancement. The charge deposition is likewise controlled by the polar properties of the dielectric material. Hence, the effects observed at positive corona can be modulated by the charges deposited at the negative half period, creating the modified initial conditions. In this way surface discharge behavior, i.e., landing and sliding and interplay of individual filaments is influenced.

From all above cases one can see that the corona onset voltage level is independent of the barrier material and is determined by the electric field and radius of the tip of the point electrode. The distance to the grounded electrode and barrier materials are influencing in turn the streamer development modes and breakdown phase.

5. Conclusions

This paper reports a correlation of partial discharge phase-resolved patterns with an optical imaging performed in a non-uniform electric field configuration with a dielectric barrier. The synchronous measuring technique between PRPD patterns and optical images was shown. The influence of different dielectric barrier materials, in terms of inherent polarity, electric permittivity and surface resistivity on discharge propagation and landing was compared. Depending on the surface charge accumulation properties the various vertical as well as surface landing streamer modes were observed, whereas the initial positive corona stage was similar for all cases. In order to compare and assess the streamer modes the geometrical factors were introduced describing the discharge profiles. The tip-branched and stem-branched profiles were introduced in a vertical space and in the horizontal one, the stretch of discharge footprint. The observations revealed clear influence of dielectric barrier materials on the discharge channel and surface charge dynamics, which can be attributed to the memory effects and dielectric repolarization, in the case of polar materials.

In this way, a novel approach to the association of the PD form or sub cluster in a phase-resolved pattern with corresponding optical mode was studied. The presented methodology and synchronous PD acquisition techniques may be used both for understanding underlying discharge physics and towards future diagnostics observations.