6
1 Panel Session on Data for Modeling System Transients Insulated Cables Bjørn Gustavsen SINTEF Energy Research N-7465 Trondheim, Norway [email protected] Abstract: The available EMTP-type programs have dedicated support routines (Cable Constants) for calculating an electric representation of cable systems in terms of a series impedance matrix Z and a shunt admittance matrix Y, based on cable data defined by geometry and material properties. Z and Y are then used as the basic input for the various cable models applied in time domain transient simulations. This paper describes necessary procedures for converting the available cable data into a new set of data which can be used as input for Cable Constants. In particular, the paper shows how to handle the semiconducting screens of single core coaxial type cables. In situations where the cable plays an important role in the transient simulation, the user should also consider obtaining a specimen of the cable in order to verify the geometrical data provided by the manufacturer. The recommendations in this paper are supported by field test results. Keywords: Electromagnetic Transients, Insulated Cables, Modeling, EMTP. I. INTRODUCTION The modeling of insulated cables for the simulation of electromagnetic transients requires 1) Calculation of cable parameters from geometrical data and material properties [1],[2]. 2) Conversion of the cable parameters into a new set of parameters for usage by the transmission line/cable model. This paper deals with the first step in the procedure, namely the calculation of cable parameters. All the commonly used programs for simulation of electromagnetic transients (EMTP/ATP/EMTDC) have dedicated support routines for this task. The routine(s) have very similar features and will in this presentation be given the common generic name “Cable Constants” (CC). Data conversion is often needed by the user in order to bring the available cable data into a form which can be used as input by CC. This conversion is needed because 1) The data can have alternative representations with CC only supporting one of the representations. 2) The CC routine does not consider certain cable features, such as semiconducting screens and wire screens. The situation is made further complicated by the fact that the nominal thickness of the various layers (insulation, semiconducting screens) as stated by manufacturers can be smaller than the actual (design) thickness of the layers. Therefore, the information on geometrical data from the manufacturer can be inaccurate from the viewpoint of cable parameter calculations. This paper demonstrates the needed conversions for one real case of a single core coaxial cable system, and proposes how to best use the available data to produce a reliable cable model. The effect of inaccurate data on a time domain simulation is also shown. The paper further discusses the shortcoming of CC in taking into account possible attenuation effects caused by the semiconducting screens. II. CABLE PARAMETERS The basic parameters used by transmission line/cable models are the following: ) ( ) ( ) ( w w w w L j R Z = (1) ) ( ) ( ) ( w w w w C j G Y = (2) where R,L,G,C are the series resistance, series inductance, shunt conductance and shunt capacitance per unit length of the cable system. These quantities are n by n matrices where n is the number of (parallel) conductors of the cable system. The variable ϖ reflects that these quantities are calculated as function of frequency. Z and Y are calculated using CC based on the geometry and material properties of the system [1],[2]. III. ACTUAL CABLE VS. CABLE CONSTANTS REPRESENTATION A. Geometry In the following we consider CC applied to systems of parallel single core coaxial type cables (SC cables). The user must specify the following input data: The location of each cable ( x-y coordinates). The geometry of each SC cable. In general, CC represents each SC cable by a set of concentrically located homogenous pipes, separated by insulating layers. Figure 1 shows the representation which would be used for a SC cable without armour.

Panel Session on Data for Modeling System Transients

Embed Size (px)

DESCRIPTION

Panel Session on Data for Modeling System Transients

Citation preview

  • 1Panel Session on Data for Modeling System TransientsInsulated Cables

    Bjrn GustavsenSINTEF Energy Research

    N-7465 Trondheim, [email protected]

    Abstract: The available EMTP-type programs have dedicatedsupport routines (Cable Constants) for calculating an electricrepresentation of cable systems in terms of a seriesimpedance matrix Z and a shunt admittance matrix Y, basedon cable data defined by geometry and material properties. Zand Y are then used as the basic input for the various cablemodels applied in time domain transient simulations. Thispaper describes necessary procedures for converting theavailable cable data into a new set of data which can be usedas input for Cable Constants. In particular, the paper showshow to handle the semiconducting screens of single corecoaxial type cables. In situations where the cable plays animportant role in the transient simulation, the user shouldalso consider obtaining a specimen of the cable in order toverify the geometrical data provided by the manufacturer. Therecommendations in this paper are supported by field testresults.

    Keywords: Electromagnetic Transients, Insulated Cables,Modeling, EMTP.

    I. INTRODUCTION

    The modeling of insulated cables for the simulation of electromagnetictransients requires1) Calculation of cable parameters from geometrical data and material

    properties [1],[2].2) Conversion of the cable parameters into a new set of parameters

    for usage by the transmission line/cable model.

    This paper deals with the first step in the procedure, namely thecalculation of cable parameters. All the commonly used programs forsimulation of electromagnetic transients (EMTP/ATP/EMTDC) havededicated support routines for this task. The routine(s) have verysimilar features and will in this presentation be given the commongeneric name Cable Constants (CC).

    Data conversion is often needed by the user in order to bring theavailable cable data into a form which can be used as input by CC. Thisconversion is needed because1) The data can have alternative representations with CC only

    supporting one of the representations.2) The CC routine does not consider certain cable features, such as

    semiconducting screens and wire screens.

    The situation is made further complicated by the fact that thenominal thickness of the various layers (insulation, semiconductingscreens) as stated by manufacturers can be smaller than the actual(design) thickness of the layers. Therefore, the information ongeometrical data from the manufacturer can be inaccurate from theviewpoint of cable parameter calculations.

    This paper demonstrates the needed conversions for one real caseof a single core coaxial cable system, and proposes how to best use theavailable data to produce a reliable cable model. The effect of inaccuratedata on a time domain simulation is also shown. The paper furtherdiscusses the shortcoming of CC in taking into account possibleattenuation effects caused by the semiconducting screens.

    II. CABLE PARAMETERS

    The basic parameters used by transmission line/cable models are thefollowing:

    )()()( wwww LjRZ += (1)

    )()()( wwww CjGY += (2)

    where R,L,G,C are the series resistance, series inductance, shuntconductance and shunt capacitance per unit length of the cable system.These quantities are n by n matrices where n is the number of(parallel) conductors of the cable system. The variable w reflects thatthese quantities are calculated as function of frequency. Z and Y arecalculated using CC based on the geometry and material properties ofthe system [1],[2].

    III. ACTUAL CABLE VS. CABLE CONSTANTSREPRESENTATION

    A. Geometry

    In the following we consider CC applied to systems of parallel singlecore coaxial type cables (SC cables). The user must specify thefollowing input data: The location of each cable (x-y coordinates). The geometry of each SC cable.

    In general, CC represents each SC cable by a set of concentricallylocated homogenous pipes, separated by insulating layers. Figure 1shows the representation which would be used for a SC cable withoutarmour.

  • 2 Fig. 1 CC representation of system of 3 SC cables

    Figure 2 shows an actual XLPE single core coaxial cable. Clearly,this cable design is different from the simple configuration assumed inFigure 1. In particular, the user needs to decide how to represent The core stranding The inner semiconducting screen The outer semiconducting screen The wire screen (sheath)

    Fig. 2 SC XLPE cable

    B. Material properties

    The user must specify the following material constants: The soil resistivity and relative permeability rg , mg The core resistivity and relative permeability rc , mc The sheath resistivity and relative permeability rs , ms The insulation relative permittivity er(In non-magnetic materials the relative permeability equals 1.0.)

    The CC-routine assumes the relative permittivity er of each insulatinglayer to be real )0( =e and frequency independent, therebyneglecting any relaxation phenomena in the insulation. This implies :

    )()()( wwww LjRZ += (3)

    CjY ww =)( (4)

    C. Eddy current effects

    The CC-routine takes into account the frequency dependent skin effectin the conductors, but neglects the proximity effect between parallel

    cables. This means that CC assumes a cylindrically symmetricalcurrent distribution in all conductors. The assumed cylindricaldistribution also means that the helical winding effect of the wirescreen is not taken into account.

    IV. MODELING REQUIREMENTS VS. PHENOMENON

    For situations with straight sheaths (i.e. no crossbondings), highfrequency transients propagate mainly as uncoupled coaxial waveswithin each SC cable. The earth characteristics have in this situationonly a mild effect on the resulting phase voltages and phase currents.In the following we shall therefore focus on the representation of thecable within the protective jacket (oversheath).

    V. CONVERSION PROCEDURES

    A. Core

    The CC-routine requires the core data to be given by the resistivity rcand the radius r1. However, the core conductor is often of the strandeddesign (Figure 2), whereas CC assumes a homogenous (solid)conductor. This makes it necessary to increase the resistivity cr ofthe core material to take into account the space between strands:

    ccc A

    r 21prr = (5)

    where Ac is the efficient (nominal) cross sectional area of the core. Theresistivity cr for to be used for annealed copper and hard drawnaluminum at 20C is according to IEC 28 and IEC 889: Copper: 1.7241E-8 Wm

    Aluminum 2.8264E-8 Wm

    If the manufacturer provides the DC resistance for the core, thesought resistivity can alternatively be calculated as

    l

    rRDCc

    21pr = (6)

    B. Insulation and semiconducting screens

    ProcedureThe semiconducting screens can have a substantial effect on thepropagation characteristics of a cable in terms of velocity, surgeimpedance and possibly the attenuation [3],[4]. Unfortunately, CCdoes not allow explicit representation of the semiconducting screens,so an approximate data conversion procedure must be applied :

    1) Calculate r2 as r1 plus the sum of the thickness of thesemiconducting screens and the main insulation.

    2) Calculate the relative permittivity er1 as

    0

    121 2

    )/ln(

    pee

    rrCr = (7)

    where C is the cable capacitance stated by the manufacturer ande0 = 8.854E-12. If C is unknown, er1 can instead be calculated based onthe relative permittivity erins of the main insulation:

    )/ln(

    )/ln( 12ins1 ab

    rrrr ee = (8)

    where a and b are the insulation inner and outer radius, respectively.For XLPE erins equals 2.3.

    Inner semiconductorOuter semiconductor

    Wire screen

    CoreInsulation

    rg , m g

    corer1

    r2

    r3

    r4

    sheath

    e1e2

    rc, mc

    rs, ms

    Insulation

    x

    y

    air

    soil

  • 3JustificationThe inner and outer semiconducting screens have a relative permittivityof the order of 1000, due to the high carbon content used in thesemiconducting screens. This implies that the capacitance of thescreens is much higher than that of the insulation and will tend to act asa short circuit when calculating the shunt admittance between core andsheath. A similar effect is caused by the ohmic conductivity of thesemiconducting screens, which is required by norm to be higher than1E-3 S/m.

    At the same time the conductivity of the semiconducting screens ismuch lower than that of the core and the sheath conductors, implyingthat the semiconducting screens do not contribute to the longitudinalcurrent conduction.

    This implies that when entering the geometrical data in CC, theuser should let the XPLE insulation extend to the surface of the coreconductor and the sheath conductor, and increase the relativepermittivity to leave the capacitance unaltered. Note that this modelingneglects the possible attenuation caused by the semiconductingscreens. The attenuation could have a strong impact on very highfrequency transients. This is discussed in Section X.

    C. Wire screen

    When the sheath conductor consists of a wire screen, the mostpractical procedure is to replace the screen with a tubular conductorhaving a cross sectional area equal to the total wire area As. With aninner sheath radius of r2, the outer radius r3 becomes

    223 r

    Ar s +=

    p(9)

    VI. APPLICATION TO 66 kV CABLE

    A. Manufacturers data

    The procedures outlined in the previous sections will be demonstratedfor a 66 kV cable similar to the one shown in Figure 3. For this cable(manufactured in the 1980s), the following data were provided by themanufacturer:

    2mm1000=cAnF/m24.0=C

    /m59.2DC -= ERmm5.191 =r

    Thickness of inner insulation screen: mm8.0Thickness of insulation: mm14Thickness of outer insulation screen: mm4.0Wire screen: 2mm50=sA

    B. Data consistency

    In Section VB it was justified that the insulation screens can berepresented by short circuit when calculating the shunt admittance.This is equivalent to a capacitance between two cylindrical shells withradius :

    mm3.20mm)8.05.19( =+=amm3.34mm14 =+= ab

    )/ln(

    2 0ab

    C repe

    = (10)

    With a relative permittivity of 2.3 for XLPE, this defines acapacitance of 0.244 nF/m which is in agreement with the capacitanceof 0.24 nF/m stated by the manufacturer.

    C. Data conversion

    Core

    From the manufacturer:

    mm5.191 =rThe resistivity is calculated by (6) :

    /m104643.3 8-=cr

    Insulation and insulation screens

    mm7.34)4.0148.0(12 =+++= rr486.21 =re (by (7))

    Wire screen

    The outer radius is calculated using (9):

    mm93.343 =r

    /m8718.1 -= Esr (copper)

    VII. INACCURACY IN DATA FROM MANUFACTURER

    The relevant cable norms (e.g. IEC 840, IEC 60502) puts limitations onthe minimum thickness of each cable layer (in relation to the nominalthickness), but not on the maximum thickness. Therefore, themanufacturer is free to use thicker layers than the nominal ones, e.g. toaccount for dispersity in production and ageig effects. This situation isprevalent for both the main insulation, the oversheath, and thesemiconducting screens.

    By measurement on a specimen of the 66 kV cable it was foundthat the insulation and in particular the semiconducting screens werethicker than stated in the data sheets :

    Thickness of inner insulation screen: 1.5 mm

    Thickness of insulation: 14.7 mm

    Thickness of outer insulation screen: 1.1 mmSeparation between outer insulation screen and centre ofeach conductor in wire screen: 1 mm

    This gives a modified model :

    mm5.191 =rmm8.372 =r

    856.22 =re (by (7))

  • 4VIII. SENSITIVITY

    At high frequencies, the asymptotic (lossless) propagation velocityand surge impedance are given as

    CLv 0/1= (11)

    CLZ c /0= (12)

    where

    )/ln(2 12

    00 rrL p

    m= (13)

    with 740 -= EpmWe will now compare the asymptotic propagation characteristics

    as calculated by the following procedures:

    Case #1:Neglecting the semiconducting screens. Capacitance and inductancecalculated using (10) and (13) with a=r1=19.5 mm, b=r2=33.5 mm, ander1=2.3.

    Case #2:Taking the semiconducting screens into account. Capacitance andgeometrical data from the manufacturer: r1=19.5 mm, r2=34.7 mm, ander1=2.486.

    Case #3:Taking the semiconducting screens into account. Capacitance from themanufacturer, geometrical data from cable specimen: r1=19.5 mm,r2=37.8 mm, and er1=2.856.

    Using the inductance calculated from (12), the velocity andcharacteristic impedance are calculated as:

    Table 1. Sensitivity of cable propagation characteristics

    case #1 case #2 case #3

    v [m/s] 197.7 190.1 (-3.8%) 177.4 (-10.3%))

    Zc [W] 21.39 21.91 (+2.4%) 23.49 (+9.8%)

    Thus, the cable propagation characteristics are highly sensitive tothe representation of the core-sheath layers.

    IX. FIELD TEST AND TIME DOMAIN SIMULATION

    A field test was carried out on a 6.05 km length of the cable. One coreconductor was charged up to a 5 kV DC voltage and then shorted toground. Thus, a negative step voltage was in effect applied to the cableend (see Figure 3).

    Fig. 3 Cable test setup

    Figure 4 shows the measured initial inrush current flowing into the coreconductor in p.u. of the DC-voltage. The initial current corresponds to

    the surge admittance of the cable core-sheath loop, which is the inverseof the surge impedance.

    The inrush current was also simulated using EMTDC v3 with aphase domain cable model [5],[6]. The CC routine was applied for thethree different cases defined in Section VIII. It is seen that using thecable representation in case #3 gives a calculated response which is infairly close agreement with the measured response. The two otherrepresentations have a much larger discrepancy. (The spike occurringat about 50 s resulted because of long leads connecting the two cablesections).

    Fig. 4 Measured and simulated inrush current

    X. IMPROVED MODELING OF SEMICONDUCTINGSCREENS

    Reference [3] suggests to model the admittance between the core andthe sheath using the circuit in Figure 5, in which each semiconductingscreen is modeled by a conductance in parallel with a capacitor. Withcomponent values obtained from measurements, they obtained a goodagreement between measured attenuation and calculated attenuation inthe range 1 MHz125 MHz. The attenuation effect of thesemiconducting screens was strong.

    Reference [4] gives a systematic investigation of the effects ofsemiconducting screens on propagation characteristics.

    Fig. 5 Improved model of insulation screens [3]

    The conductivity and permittivity of the semiconducting screensdepends very much on the amount of carbon added, the structure of

    core

    Inner semiconducting screen

    Outer semiconducting screen

    Main insulation

    G1

    G2

    C1

    C2

    C

    sheath

    Y

    3.85 km 2.2 km15 m

    core

    sheath

    Negativestep voltage

  • 5the carbon, and the type of base polymer. Very high carbonconcentrations are used (e.g. 35%). IEC 840 requires the resistivity tobe lower than 1000 Wm for the inner screen, and below 500 Wm forthe outer screen. One manufacturer stated that they use a much lowerresistivity, typically 0.1 Wm10 Wm. The relative permittivity isvery high, typically of the order of 1000. The permittivity andconductivity can be strongly frequency dependent.

    In order to investigate the possible attenuation effects of theinsulation screens of the cable considered in this paper, arepresentation as in Figure 5 was employed assuming frequencyindependent conductances and capacitances. The component valueswere calculated as follows:

    m/nF24.0=C (from manufacturer))/ln(/2 201 brC repe=)/ln(/2 102 raC repe=

    )/ln(/2 21 brG ps=)/ln(/2 12 raG ps=

    wherea: Outer radius of inner semiconducting screenb: Inner radius of outer semiconducting screener : Relative permittivity of semiconducting screenss : Conductivity of semiconducting screens

    Figure 6 shows the attenuation per km, for a few combinations ofs and er. The curves define to which peak value a sinusoidal voltage of1 p.u. peak value decays to over a distance of 1 km. (The signal decaysexponentially as function of length). The model predicts a significantcontribution from the semiconducting screens for a low value of boththe relative permittivity (10, 100) and the conductivity (0.001). Withthe high permittivity (1000), the capacitance tends to short out theconductance, and no appreciable increase of the attenuation is seen.The lowest value for the permittivity (10) is probably unrealistic.

    Fig. 6 Effect of semiconducting screens on attenuation

    XI. DISCUSSION

    This paper has focused on the importance of correctly modeling thesemiconducting screens of single core coaxial type cables. It is shownthat a careless modeling tends to produce a model with a too low surgeimpedance and a too high propagation velocity. The importance ofaccurate modeling is strongly dependent on the type of transientstudy. If the cable is part of a resonant overvoltage phenomenon, theaccurate representation of the cable the surge impedance andpropagation velocity is crucial.

    XII. CONCLUSIONS

    This paper describes necessary conversion procedures for the availablecable data for usage by Cable Constantstype routines (CC), withfocus on single core (SC) coaxial type cables. The main conclusions arethe following:

    CC does not directly apply to SC cables with semiconductingscreens, so a conversion procedure is needed before entering thecable data into CC. This paper describes the needed conversionsand also describes the conversions needed for handling the corestranding and wire screens.

    The nominal thickness of the various insulation andsemiconducting cable screens as stated by manufacturers can besmaller than those found in actual cables. This can result in asignificant error for the propagation characteristics of the cablemodel.

    CC has no means for taking into account any additional attenuationat very high frequencies resulting from the semiconducting screens.

    XIII. REFERENCES

    [1] L.M. Wedepohl and D.J. Wilcox, Transient Analysis ofUnderground Power Transmission System ; System-Model andWave Propagation Characteristics, Proc. IEE, vol. 120, No. 2,February 1973, pp. 252-259.

    [2] A. Ametani, A General Formulation of Impedance andAdmittance of Cables, IEEE Trans. PAS, Vol. 99, No. 3,May/June 1980, pp. 902-909.

    [3] G.C. Stone and S.A. Boggs, "Propagation of Partial DischargePulses in Shielded Power Cable, Proceedings of Conference onElectrical Insulation and Dielectric Phenomena, IEEE 82CH1773-1, October 1982, pp. 275-280.

    [4] W.L. Weeks and Yi Min Diao, Wave Propagation inUnderground Power Cable, IEEE Trans. PAS, Vol. 103, No. 10,October 1984, pp. 2816-2826.

    [5] A. Morched, B. Gustavsen, and M. Tartibi, A Universal LineModel for Accurate Calculation of Electromagnetic Transients onOverhead Lines and Cables, IEEE trans. PWRD, vol. 14, no. 3,July 1999, pp. 1032-1038.

    [6] B.Gustavsen, G. Irwin, R. Mangelrd, D. Brandt, and K. Kent,"Transmission Line Models for the Simulation of InteractionPhenomena between Parallel AC and DC Overhead Lines",IPST'99 International Conference on Power System Transients,Budapest, 1999, pp. 61-67.

    XIV. BIOGRAPHY Bjrn Gustavsen was born in Norway in 1965. He received theM.Sc. degree in 1989 and the Dr.-Ing. degree in 1993, both from the

  • 6Norwegian Institute of Technology in Trondheim. Since 1994 hehas been working at SINTEF Energy Research (former EFI). Hisinterests include simulation of electromagnetic transients andmodeling of frequency dependent effects. He spent 1996 as aVisiting Researcher at the University of Toronto, and the summerof 1998 at the Manitoba HVDC Research Centre, Winnipeg,Canada.

    APPENDIX DATA CONVERSION

    The following Matlab code does the recommended data conversionfor the case described in Section VI. All geometrical quantities are inmeters.

    INPUT:

    C =0.24e-9; %capacitance stated by manufacturer [F/m]Acore =1000e-6; %core nominal cross sectional areaAsheath=50e-6; %sheath nominal cros sectional areatins =14e-3; %thickness: main insulationtins1 =0.8e-3; %thickness: inner insulation screentins2 =0.4e-3; %thickness: outer insulation screenr1 =19.5e-3; %core radiusRDC =2.9e-5; %core DC resistance [ohm/m]eps0 =8.854e-12; %vacuum permittivity

    OUTPUT:

    rhoc=RDC*pi*r1^2 %core resistivityr2=r1+tins1+tins+tins2; %sheath inner radiusr3=sqrt(Asheath/pi+r2^2); %sheath outer radiusepsr1=C*log(r2/r1)/(2*pi*eps0); %effective rel. permittivity %of core sheath layer