analysis of leakage current wave forms for field-aged and new composite insulators

7/27/2019 Analysis of Leakage Current Wave Forms for Field-Aged and New Composite Insulators

1/4

Analysis of Leakage Current Wave Forms for Field-aged and New CompositeInsulatorsM.A.R.M. Fernando S.M. Gubanski

Chal mers University of TechnologyDepartment of Electric Power EngineeringDivision of High Voltage Technology41 2 96 Goteborg, Sweden

AbstractResults of measurements and analyses of leakage current(LC) patterns on field-aged and non-aged insulators arepresented. The field-aged insulators were exposed tomarine and industrial pollutions as well as to tropicalweather co nditions for one year. T he LC measurementswere performed on them under natural and artificialcontaminated cond itions. The non-aged insulators w erepreviously used for pollution testing. In most cases, theLC patterns on the field-aged insulators were sinusoidalin shape whereas they were highly deformed on the non-aged insulators. T hes e different LC patterns were quanti-tatively analysed by means of a circuit model represent-ing a polluted and wet insulator surface. A modificationof the previously developed neural network for evalua-tion of the third and fifth harmonic contents of the LCpatterns is also presented.

IntroductionThe good pollution performance of composite insulatorsis mainly provided by the water repellency of their sur-faces. Most of the insulators, however, show changes ofthe surfac e properties under difficult field and laboratoryconditions. They are no longer static but dynamic ob-jects indicating different flashover voltages when testedin the laboratory conditions [1, 21.Out of many parameters describing the state of a con-taminated insulator, a surface leakage current (LC) pro-vides much useful information. However, the LC waveforms can significantly vary not only at different stagesof the flashover development but also depending on thesurface hydrophobicity. Our previous report [3] pointedout that the low level LC wave forms on artificially pol-luted insulators are non-linear and contain harmoniccomponents.A comparison of LCs between artificially contaminatednon-aged and field-aged insulators may provide usefulinformation for a predication of insulator performance.

In this study, three different field sites located in tro picalenvironment were select ed. Th e insulators were in serv-ice for about one year in those places and they were re-spectively exposed to industrial (cement dust), marine aswell as clean environments. The LC measurements onfield-aged insulators were performed and compared withartificially contaminated ones. In addition the non-linearbehaviour i n LC wave forms was also analysed. Thepreviously developed neural network [3] was modifiedto evaluate the harmonic contents.

Experimental ProcedureSix 33 kV insulators, i.e., four composite insulators(three silicone rubber (SR), and one EPDM), one RTVcoated, and one porcelain, were selected for the testing.Parameters of the insulators are given in Table 1.Table 1. Insulator parameters.

The insulators were installed at three places in SriLanka. The description of the test sites is provided inTable 2. In the first two sites insulators were installed inthe existing 33 kV distribution system whereas at thethird site, they were hanged in an open area without be-ing energized.

(0-7803.3851-0) 1997 IEEE Annual Report - Conference on Electrical Insulation and Dielectric Phenomena, Minneapolis, October 19-22, 1997


2/4

Table 2. Details of the test sitesSite

II1

I11

Contaminant Average weathermainly cement dus t, annual rainfall (-1300sometimes saltmainly salt temperature -28 "Cno pollution , onlychanging weather (high mm)rainfall and solar radia-tion) humidity -80 %

humidity -81 %annual rainfall (-1755temperamre -27 "C

The LC measurements were performed by means of aspecial system which can register time variations of avoltage drop across a shunt resistor in laboratory condi-tions. The LCs were first recorded in a digital oscillo-scope and then were stored and analyses in a personalcomputer. A special circuit protected both instrumentsagains t over-voltages and over-currents.The LCs were measured on field-aged insulators in thehigh voltage laboratory inside a fog chamber (fog in-tensity of -50 g/m3/hour). The testing lasted for aboutten minutes and the L C wave forms at 0, 5 and 10 min-utes after applying the voltage were recorded. The sup-plied voltage was set for 19 kV. Afterwards the insula-tors were cleaned and the measurements were repeated.Then the insulators were artificially polluted in a slurrycontaining 10 g/1 salt and 40 gA kaolin. The resultingSD D level for this slurry was similar as those observedin the field. In parallel, a set of non-aged reference insu-lators was contaminated in the same slurry. The insula-tors were previously exposed to many contaminatingprocedures and their surface properties might havechanged. They remained in rest for seven months beforethe latest measurements. The time lapse between theartificial contamination and measurements was one hour.Leakage Current AnalysesAn attempt was made to analyse the non-linear behav-iour of the LCs. To represent it a circuit model was se-lected. It seems that the non-linear currents might becaused by a formation of wetted areas separated fromeach other by unwetted regions. Properties of the wetareas are represented by a voltage controlled non-linearresistance, as show n in Fig. 1. The dry areas between theneighbouring wet areas are characterised by a parallelcapacitance-resistance coupling. In addition another ca-pacitance-resistance com bination is used to represent therest of the insulator surface (hydrophobic or hydro-philic). The resultant cir cuit diagram is shown in Fig. 2.

0 . 2 ~ ~....._........... ..-............1 .0.05 _ _ _ .

V [ W1 13 19

Fig 1 , The non-linear resistor representing wet are as.Insulator.---__.-..E-_______ . . _ . . .

1 :I I I I

Fig 2 . Circuit model of the insulator surfaceNeural Network AnalysisA neural network (NN), being a m odification of a previ-ously developed NN [3], was trained to evaluate con-tents of third and fifth harmonics in the LC patterns withreference to the fundamental frequency (50 Hz). TwoNNs containing 2-25-4 neurons were used for each ofthe harmonics. The LC wave forms were first normal-ised. Then the two inputs were obtained from averagevalues of two positive quarter cycles of LC wave form(starting from LC=O and proceeding to the next 0 valuein the positive half cycle). The harmonic contents werealso obtained from the fast fourier transform (FFT)analysis and these were used as the outputs to train thenetwork. The assumed relationship between the outputof the NN and the harmonic contents is shown in Table3. The levels of the output classes were arbitrarily se-lected.Table 3. Relationship between output values of NN and thirdand fifth harmonic contents

Output value Third harmonic Fifth harmoniccontent< 3.7 %0 1 0 0 3.7 - 18 % 2 - 6 %

6 - 1 0%> 24 % > 10%

content

0 0 1 0 1 8 - 2 4 %

351


3/4

Results and Discussion

Insulator

LC Measurements

Filed aged Non-agedNaturally con- I Artificially

Table 4 show s the LC wave forms for naturally and arti-ficially polluted field-aged insulators from site 11. Theobserved wave forms were similar to those measured onthe insulators from s ite I.Table5. LC wave formsfor naturally and artificially contami-natedf ield aged insulators from site II.

The illustrated wave forms were recorded 5 minutes af-ter applying the voltage. During the testing, no surfacedischarges were observed.The LC levels on naturally contaminated SIR compositeand RTV coated insulators were low and their patternswere sinusoidal. On the other hand, the LC levels forinsulator #4 and #S were comparatively higher. The lat-ter LCs were also deformed. Their third harmonic con-

~

352

tent was about 10 %. The measured LC shapes weresimilar for both the naturally and artificially contami-nated field aged insulators.The L C wave forms for insulators from site I11 were allsinusoidal and the current levels were low. Site I11 hadno pollution, and the insulators remained clean one yearafter the installation. This possibly may explain the lowlevel of the currents.Table 5 shows the LC levels and third harmonic contentsfor the field-aged and non-aged insulators. When com-parison was made between the behaviour of the field-aged insulators in the naturally and artificially contami-nated states, similar LC patterns in terms of th e LC leveland the harmonic content were foun d.Table 5 . LC levels and harmonic contents on field aged andnon-aged insulators [L-LC level (d),-harmonic con-tent(%)]

The artificially contaminated non-aged insulators hadhigher LC levels and harmonic contents. The LCs onthese insulators were non-linear, probably because offormation of isolated wet areas which appeared as a re-sult of mo dification of surface properties caused by pre-vious contaminations. Such changes probably did nottake place on the field-aged insulators, where relativelyhigh hydrophobicity in the naturally contaminated statewas still maintained for the SIR and RTV coated insula-tors.LC AnalysesIn the model circuit (Fig. 2), arbitrary values for R, RI,R2 , C1 and C2 were selected. R2 and C2 values were var-ied in order to model different L C wave forms. The se-


4/4

lected R I and C1 values were respectively 50 M R a nd1000 pF. the variation of R' is shown in Fig. 1. Table 6shows the modelled LC s for different R2 and C 2 values.Similar experimentally measured LC wave forms arealso included fo r comparison.

~~

Wave form

Table 6. Theoretically modelled and expe rimentally observedLC patterns

Harmonic contentN N FFr3 r d 1 5 t h 3 rd 1 5 t h

came completely wet. In that case (lower Rz), the waveforms were sinusoidal. However, when the surface be-came more hydrophobic (higher R2), non-liner currentsappeared du e to the influence of R', R I , and C, . Duringthe non-linear process, the change in Cz also influencedLC shape. At the present stage it is rather difficult toassign a physical meaning to the circuit parameters.Additional model studies are necessary.

~~Theoretical I ~~~ Exnerimental

C2 = 100 p F, R2 = 5 MR II I 110-

n IS -,I 00C2 = IO p F , R2 = 200 MR I

Table 7. Third andfifth harmonic contents evaluated from th eN N and FFT

om-,PO

It was assumed that R2 varied from 200 M R - 5 MQ andC2 changed from 10 p F - 100 pF when the surface be-

NN AnalysisThe modified NNs worked successfully with al l themeasured LC patterns. Table 7shows some of the resultsof the N N analysis on the third and fifth harmonic con-tents of thi. mi.:isured LC wave forms. The results fromFET n i l y s i s ;ire ~ i l s ancluded for comparison.

ConclusionsThc iiisul.itors aged under marine and industrial pollu-tion i n tropical clinintc show similar leakage currentpatterns indcpcndently it' they are naturally or artificiallycor1tamin~itc.d. he LC s on the polluted R TV coated andSIR insulators are low in magnitude and sinusoidal inshape. The EPDM and porcelain insulators have higherLC lewls. On the other hand, the LC wave forms onnon-aged polymeric insulators, which were earlier mul-tiple contaminated, are highly deformed from sinusoidalshape. The deformation is probably caused by existingwet areas on the surfaces of the insulators. Th e effect canbe represented by a non-linear resistance in a modelequivalent circuit.The deformed patterns contains high amounts of thirdand fifth harmonics. The developed and trained neuralnetwork can be used to evaluate this content.

References[ I ] . M.A. Mbwana, M.A.R.M. Fernando, S.M. Guban-ski, "Variation of Surface Properties on ArtiJiciallyPolluted Composite Insulators", Proc. Nordic InsulationSymp. (NOR D-IS), Bergen, N orway, 1996, pp. 107-114 .[2]. R. Matsuoka, H. Shinokubo , K. Kondo, T. Fujimura,"Assessment of Basic Contamination Withstand VoltageCharacteristics of Polymer Insulators", IEEE Trans. onPD , Vol. 1, No. 4, October 1996, pp. 1985-1900.[3]. M.A.R:M. Fernando, S.M. Gubanski, "LeakageCurrent Patterns on Artificially Polluted CompositeInsulators", Annual Report on IEEE CEIDP, San Fran-cisco, USA, 1996, vol. 1 , pp. 394-397.

353

analysis of leakage current wave forms for field-aged and new composite insulators

Documents