Close and Return 726

F.G. Emission lumineuse et ini· F.G. Light emission and electrical tiation d'arborescences électriques tree initiation in highly·stressed dans le PRe sous fort champ élee­ XLPE cable insulation trique KABIR F., BRAUN JM., DENSLEY J., KABIR F., BRAUN JM., DENSLEY J., Ontario Hydra, Toronto, Canada Ontario Hydro, Toronto, Canada HAMPTON N., BICC Cables, Erith, UK HAMPTON N., BICC Cables, Erith, VERNE S., BICC Pic, Hemel Hempstead, Royaume-Uni UK VERNE S., BICC Pic, Hemel Hempstead, Royaume-Uni

Resumé Abstract

Les polymères extrudés utilisés comme isolants dans les Extruded polymers used as insulations for high-voltage câbles haute tension sont le siège de degradation causée cables are susceptible to electrical-tree degradation par la croissance d'arborescences électriques Ces which initiates at stress enhancements. Light is emitted dernières sont causées par un intense champs électrique during the tree- local. Durant la phase d'initiation, avant l'apparition de initiation phase, prior to the formation of channels and canaux et de décharges partielles, il y a émission de partial discharges. This paper presents data on the light photons par l'isolant. Nous présentons dans ce papier les emission characteristics for XLPE and shows that the résultats d'une étude sur l'émission lumineuse dans le emitted light is independent of temperature but is PRC. Nous montrons que la quantité de lumière émise affected by the gases dissolved in the polymer. est indépendante de la température mais sensible aux gaz dissous dans le polymère.

• Initiation phase, i.e. the gradual deterioration of the insolation which occurs prior to channel formation and Introduction partial discharges. • Propagation phase, when the channels initiated at the Polymers such as low density (LDPE) or cross­ stress enhancement progress through the insulation due to linked polyethylene (XLPE) are now beginning to be used as the action of partial discharges. insulation for high-voltage transmission-elass cables up to 500 • Final breakdown when one or more of the· tree channels kY. Although these materials have excellent electrical bridge the insulation and produce insulation failure. properties for cable applications, e.g., low permittitvity, low loss and a high breakdown strength, they are susceptible to The mechanisms responsible for the deterioration during tree electrical ageing when subjected to high electrical stresses [1]. initiation are not clearly understood. The following have been One form of electrical ageing is electrical treeing phenomena proposed: originating from local stress enhancements. • Joule heating [2]. Stress enhancements, which can be initiation sites for electrical • Impact ionization [3]. trees, are considered to be as important as gas-:filled cavities for • Damage due to hot electrons [4]. electrical ageing. Possible sources of stress enhancements in • Damage due to (uv) radiation [5]. extruded cables are protrusions on the semiconducting • Mechanical fatigue due to Maxwell stress [6]. shieldlinsulation interfaces or metallic or high permittivity contaminants occluded in the insolation or located at Laboratory studies of tree initiation have usually been interfaces. Thus, the quality of the interfaces is a very performed with metallic needle/plane or needle/needle important parameter in determining the electrical performance electrodes to model a stress enhancement. The electrostatic of extruded cables. field distribution can be calculated for such a geometry so that observed phenomena canbe direct1y related tothe applied Electrical treeing has 3 phases: Close and Return 727

field. In practice, however, space charge is present near the The tube had a dark noise of <12 eps when housed in a needle tip so that the actual field in this region is the cooling unit operating below -20°e. The 10 ns triangular combination ofthe applied and space charge fields. output pulses from the photomultiplier were measured by a At high applied AC electrical stresses, light is emitted; this specially built pulse height analyzer(PI-IA), which light can be caused by any of the above-mentioned mechanisms displayed the pulse amplitude distribution (15 channels) or by a chemical reaction. Previous studies have shown that and also the phase position of the pulses with respect to the [5,7-11]: AC waveform. The PRA also recorded the light intensity, defined as the product of the number of pulses per channel • The inception voltage of the light (electrohunineseence, times the channel number,each half cycle, the skewness EL) is closely related to the threshold voltage below which and kurtosis of the intensity distributions and the phase there is no electrical ageing. positions ofthe intensity peaks. • EL extends in wavelength from the visible into the uvand this uv rnay contribute to . The test procedure adopted was to increase the voltage in l kV • Gases dissolved in the insulation influence both the EL increments and to measure the EL for a period of 30 s at each and electrical tree initiation, e.g. the presence of oxygen voltage until the light intensity increased significantly above reduces the amount of light emitted at all wavelengths and the background noise level. The voltage was then removed decreases the tree inception voltage or time-to-tree by and the specimen heated for about 40 minutes to reach 70°C about 50% compared to nitrogen. and the voltage reapplied in similar steps to re-measure the • The phase relationship of the EL depends on the local light inception voltage (UV). The light intensity at zero fielcl i.e., the combination of the space.charge and applied voltage, i.e., the background level.was always greater at70°C fields. than at room temperature. The voltage was again removed and the specimen allowed to cool to room temperature. The UV Most of the previous studies have been limited to metallic was again measured and the voltage raised in l kV steps above needle electrodes. and performed at room temperature, while in the UV until an electrical tree was initiated or the maximum practice, the stress enhancements are more likely to occur at rated voltage of the test cell, 29 kV, was .reached; Tree the semiconducting shields and the insulation will he operating initiation was clearly detected by large increases in the number at temperatures up to or even greater than 90°e. The results and amplitude oflight pulses and theonset ofPDs. presented in this paper will include tests performed with semiconducting rnaterial electrodes at temperatures up to 70°e. Results and Discussion

Experimental Figure 1 shows how the light intensity varies with applied voltage for two XLPE specimens. The light intensity Each specimen consisted of a needle-shaped increased rapidly above light inception, with the light semiconducting electrode, made from a commercial-grade emission occurring in the increasing positive and negative compound, embedded in a cable-grade XLPE in such a way quadrants of the applied voltage. This is in agreement with the that the distance between the tip of the needle and the results ofother studies[9]. The peaks of the intensity occurred adjacent ground plane was 10 mm. Thetip radii of the at phase angles between 40 and 70° in the positive half cycle specimens varied between 1 and 15 um. Ail specimens and between.220 and 250° in the negativehalf cycle. .As·the were treated in a vacuum oven for 24 h at 60°C to remove voltage increased above light inception, the peak intensity the cross-linking by-products, occurred earlierin each half cycle and eventually the light couldoccur in the other two quadrants. The phase behaviour is Details of the light-tight test cell and light-collection due to the modification of the electricfield at the electrode tip system have been presented elsewhere[ Il]. An. unique due to the space charge set up by the mobile charges. The light feature of the test cell was that an elliptical mirror was intensity was approxîmately similarin .each half Cycle at used to increase the efficiency of the light collection. To inception but, at higher voltages, the intensity was greater in ensure a good response over a broad spectral range, the the negative half cycle, sometimes approaching twice that surface of the mirror was coatedfirstly with aluminum occurring in the positive halfcycle. and then with MgF2 .. Thetipof theneedle is located at the focal point of the mirror. To suppress partial The skewness, arneasure of the ·symmetry, of the intensity discharges, the specimens were either embedded in a distributionwas usually close to zero, i.e.,·a symmetrical transparent epoxy resin or tested with the test cell distribution, at light inception. The skewness increased to pressurised to two atmospheres of nitrogen. between 1and 2 withincreasing voltage above light inception, as can be seen in Figure 2. A positively skewed distribution The light collected and reflected by the mirror was focused has a greater number of large-amplitude than s1tla1l-amplitude onto the photocathode of a photomultiplier (Burle, model pulses. There was no difference in the behaviour of the C31034A02) which hada flat response from 200 - 900 nID. distributions in the positive or negative halfcycles. Close and Return 728

XLPE #1 XLPE#2

••••••••• ••

10 20 30 5 10 15 20 Applied Voltage (leV) Applied Voltage (leV)

-<>-90 10180 -- 180lb 270-- 27010360

Figure 1. Emitted Light Vs. Applied Voltage for 2 XLPE Specimens

XLPE# 1 XLPE#2

2 1.4 1.2 1.5 1 0.8 ilc: ilc: 0.6 ~ ~ e 0.4 .:JI. 0.5 .:JI. CI) CI) 0.2 0 0 -0.2 -0.5 -0.4 0 10 20 30 0 5 10 15 20 25 Applied Voltage (leV) Applied Voltage (kV) ---- Positive Ralf Cycle -0- Ne ative Ralf Cycle Figure 2. Skewness Vs. Applied Voltage for 2 XLPE Specimens

The variation in the kurtosis, a measure of the sharpness of a was about 25% lower than the initial value at room distribution, with applied voltage is shown in Figure 3. At temperature. However the second UV at room temperature light inception the kurtosis was usually negative, i.e., flat­ was similarto the 70°C value. The reduction in UV is due to topped, but became positive, i.e., more peaked, at higher the removal of oxygen out of the specimen by the increased voltages. There was no difference in the behaviour of the rate of diffusionwhen inunersed in nitrogen at 70°C. When distributions in the positive or negative halfcycles. The specimens wereembedded in epoxy resinthe diffusion of gases changesin skewness and kurtosisat highervoltages indicatea into or out of the specimen, even at 70°C was negligible with greaterincreasein the numberof large-amplitude light pulses. the resultthat the UV was independent of the test sequence as can be seen from Figure 4. The difference betweenthe LIV's Figure 4 shows the e:ffect of the test sequence on the light of the specimens inunersed in nitrogenand those embedded in inception voltage. The fust test was performed at room epoxy resinfor the initial roomtemperaturetest was due to the temperature, followed by a test at 70°C and, finally, by another smallertip radius used for the cast-in-epoxy specimens which at room temperature. The results for specimens testedin resulted in a lowerUv. nitrogen at 2 atm showthat the light inception voltage at 70°C Close and Return 729

XLPE# 1 XLPE#2

8 4 7 3.5 6 3 5 2.5 ... 4 ... 2 il! il! 1.5 oc0 3 oc0 :l :l 1 :M: 2 :M: 0.5 1 0 0 -0.5 ·1 -1 ·2 -1.5 0 10 20 30 0 5 10 15 20 25 Applled Voltage (kV) Applled Voltage (kV)

Figure 3. Kurtosis VsApplied Voltage for 2 XLPESpecimens.

NITROGEN EPOXY RESIN UV (kV) uv (kV) 22r--'----~-~---, 22.-'----'------, 20 ~+++

10

r +t 8'----'----'--~--'--' 8 23 1 70 232 231 70 232 TEMPERATlJRE OF TESTS(C) TEMPERATURE OF TESTS(C) 1 & 2 DENOTE FIRSTANDSECOND APPLICATIONS 1 & 2 DENOTE FIRSTANDSECOND APPlICATIONS

95%CONFIDENCE UMITS 95%CONFIDENCE UMITS Figure 4. Effect of Test Sequence on LIV.

Figure 5. Effect of Gas Diffusion on LIV (LDPE Specimens). Close and Return 730

Figure 5 shows the change in the UV for a LDPE specimen 2. Noto, F. and Yoshimura, N., 'Voltage and Frequency exposed to a controlled atmosphere. LDPE was preferred to Dependence of Tree Growth in Polyethylene', Annual XLPE to avoid the effects of the diffusion of cross-linking Report Conference Electrical Insulation and byproducts. The initial UV in nitrogen at 2 atm was about 25 Phenomena, pp. 207-217, 1974. kV After a vacuum treatment for 64 h, which removed the dissolved oxygen, the UV had decreased to 16.kV when fust 3. Kojima, K., Takai, Y and Ieda, M, Electronic Conduction immersed in nitrogen and to about 12 kV after a further 5 h in in Polyethylene Naphthalate at High Electric Fields'. Joum. nitrogen. Further evacuation for 19 h did not change the UV Appl. Phys., Vol. 59, pp. 860-4, April 1986. Ifthe emitted light was due to partial discharges there would have been a significant increase in the UV Thus, the light 4. Shibuya, Y, Zolededziowski, S. and Calderwood, J.H.. emitted is due to electroluminescence and not partial 'Light Emission and Deterioration in Epoxy Resin discharges. When the specimen was reimmersed in air, the Subjected to Power Frequency Electric Fields', Proc. IEE, UV increased to its original value after about 3 h as oxygen Vol. 125, pp. 352-4, 1978. diffused to the needle tip. 5. Bamji, S.S., Bulinski, A.T., and Densley, 1.. 'Evidence of Conclusions Near-Ultraviolet Emission during Electrical-Tree Initiation in Polyethylene', 1. Appl. Phys., Vol. 61(2), pp. 694-699. The experimental results show that: Jan. 1987.

• electroluminescence occurs at localized stress 6. Ieda, M., 'Dielectric Breakdown Process of Polymers', concentrations before the onset of partial discharges and IEEE Trans. Elec. Insul., Vol. El-15, pp. 206-224, June electrical treeing. 1980. • the light inception voltage is independent of temperature between room temperature and 70°C, provided the gases 7. Shimizu, N., Katsukawa, H., Miyauchi, M, Kosaki, M and dissolved in the polymer do not change. Horn, K.,.; 'The Space Charge. Behaviour and • the light inception voltage decreases when the oxygen Luminescence Phenomena in Polymers at 77 K', IEEE dissolved in the polymer is removed or replaced with Trans. Elec. lnsul., Vol. EI-14, pp. 256-263, 1979. nitrogen. • above the Iight inception voltage the. light intensity 8. Laurent, C, Mayeux, C. and Noel, S., 'Dielectric increases rapidly withvoltage, The increase in intensity is Breakdown of Polyethylene in Divergent Field: Role of made up of largeamplitude light pulses indicating more Dissolved Gases and Electroluminescence', J. Appl. Phys., energetic light pulses which could cause daIl1ageand lead Vol. 54, pp. 1532-1539, MarchJ983. to electrical tree initiation. 9. Champion, J.V, Dodd, SJ. and Stevens, o.c, Acknowledgements 'Quantitative Measurement of Light Emis~ion during the Early Stages of Electrical Breakdownin Epoxy and The authors would like to thank the Directors of BICC UnsaturatedPolyester.Résins', 1.Phys, D: Appl. Phys., Vol. Gables. for supporting the work described above and for 26, pp. 819-828, 1993. giving permission to publish the paper. 10. Stone, .G'C, van Heeswijk, KG. and Bartnikas, R, References 'Electrical Aging and Electroluminescence in Epoxy Under Repetitive Voltage Surges', IEEE Trans. Elec. Insul., Vol. 1. Densley, RJ. and Bartnikas, R and Bernstein, B.S., 27, pp. 233-244, 1992. 'Multiple-Stress Aging of Solid-Dielectric Extruded Dry­ Cured lnsulation Systems for Power Transmission Gables', . Il. Kabir, F., Braun, J-M, Densley, J. and Hampton, RN., IEEE Transactions on Power Delivery, Vol. 9(1), pp. 559­ 'Light Emissionfrom Highly-Stressed Polymerie Gable 571, Jan. 1994. Insulation', Conf Rec. of 1994 IEEE Inti. Symp. Elect. Insul., Pittsburgh" pp. 37-40, June1994.