You are on page 1of 19

1 996 IEEE Transactions on Power Apparatus and Systems, Vol. PAS-98, No.6 Nov./Dec.

1979

A SURVEY OF METHODS FOR CALCULATING TRANSMISSION


LINE CONDUCTOR SURFACE VOLTAGE GRADIENTS
IEEE Corona and Field Effects Subcommittee Report
Radio Noise Working Group*
ABSTRACT ductors in the bundle increases. Ever since the advan-
tages of using bundle conductors for high voltage
This paper presents the results of a survey of transmission lines became evident, methods have been
methods for calculating transmission line conductor developed to calculate the electric field in the vicin-
surface voltage gradients. The main objective of this ity of such bundles. With the availability of digital
survey is to obtain comparative data on the accuracy computers in recent years, a large number of papers has
and the computational aspects of the different methods appeared in the technical literature on methods for the
presently in use for calculating conductor surface accurate determination of the conductor surface elec-
voltage gradients. A total of thirteen typical trans- tric fields of transmission lines using bundle conduc-
mission line configurations has been selected for the tors. For the practical user, however, this presents
survey which elicited worldwide response. In addition the problem of choosing between the different methods
to an analysis of the results, the paper also presents and particularly of choosing a reasonably simple meth-
a brief review of the existing methods of calculation od of acceptable accuracy.
and a discussion of the accuracy requirements from a
practical point of view. With the objective of obtaining comparative data
on the accuracy and possibly other information on the
INTRODUCTION computational aspects of the different methods of cal-
culation presently in use, the IEEE Radio Noise and Co-
One of the important considerations in the design rona Subcommittee (presently Corona and Field Effects
of high voltage ac and dc transmission lines is their Subcommittee) launched a worldwide survey of methods of
corona performance, which is usually defined in terms gradient calculation. A total of thirteen transmission
of corona loss (CL), radio interference (RI), audible line configurations, both ac and dc, have been select-
noise (AN), television interference (TVI), ozone, etc. ed, mostly from data on operating transmission lines,
generated by corona. From the point of view of trans- and included in a questionnaire prepared for the survey.
mission line design, the factor which most influences In addition to the members of the Subcommittee, the
the corona performance is the conductor surface voltage questionnaire has also been sent to persons all over
gradient. Thus an accurate knowledge of this parameter the world who have been known to be involved in the de-
is essential for estimating the corona performance of velopment of gradient calculation methods. An analysis
any proposed line. of the results of this survey is presented in this pa-
per. A brief review of the existing methods of calcu-
Up to a certain transmission voltage, increasing lation, and a discussion of the accuracy requirements
the conductor size is the simplest method of reducing from a practical point of view, are also included.
the conductor surface voltage gradient and thus of ob-
taining acceptable corona performance. At higher volt- It has also become evident in recent years that
ages (>200 kV), however, the use of single conductors different investigators have been using different val-
becomes impractical and multiple conductor bundles have ues of the conductor surface voltage gradient, such as
to be used for reducing the conductor surface voltage for instance, the average conductor gradient, average
gradient while at the same time keeping the conductor maximum bundle gradient, maximum bundle gradient, etc.,
sizes within practical limits. in corona calculations. Some further confusion has
been created in this direction due to the lack of pre-
The calculation of conductor surface voltage gra- cise definitions for the different gradient values used.
dient is fairly simple for transmission lines using An IEEE standard is presently in preparation, however,
single conductors, but becomes increasingly complex for on "Definitions of terms relating to overhead power
bundle conductors, especially as the number of subcon- line corona and radio noise", which includes defini-
tions of conductor surface voltage gradients. All the
*

The survey was initiated by the Working Group 3 of gradient terms used in this paper conform to the above
the Radio Noise and Corona Subcommittee. Following standard and definitions of the relevant terms are
recent reorganization, the survey paper has been com- given in the Appendix.
pleted by the Radio Noise Working Group of the Corona
and Field Effects Subcommittee. Members of the Work- REVIEW OF EXISTING METHODS OF CALCULATION
ing Group are: P.S. Maruvada, Chairman, R.J. Bacha,
A.C. Baker, W.E. Blair, M.E. Bulawka, V.L. Chartier, Certain basic assumptions are involved in all the
R. Cortina, L.B. Craine, G.R. Elder, C. Gary, J.F. existing methods for calculating the electric field in
hIall, W. Janischewskyj, N. Kolcio, T.J. McDermott, the vicinity of transmission line conductors:
R.M. M4orris, R.J. Nigbor, W.E. Pakala, A. Paldi, M.D. -
Perkins, J, Reichman, J.P. Reilly, R.J. Richeda, W.R. the ground is assumed to be an infinite horizontal
Schlinger, S.A. Sebo, P.D. Tuttle, F.W. Warburton, conducting plane surface;
B.F. Whitney and P.S. Wong.
- the conductors are assumed to be smooth infinitely
long circular cylinders parallel to each other and
to the ground plane;

- the conductors are assumed to be equipotential sur-


faces, with known potentials applied to them; the
F 79 257-7 A paper recommended and approved by the ground plane is assumed to be at zero potential;
IEEE Transmission and Distribution Commmittee of the IEEE
Power Engineering Society for presentation at the IEEE PES the influence of conductor support structures and of
Winter Meeting, New York, NY, February 4-9, 1979. any objects in the vicinity is neglected.
Manuscript submitted October 2, 1978; made available for
printing November 8, 1978.
0018-9510/79/1100-1996$00.75 © 1979 IEEE
1997

In addition to the above, it is also assumed that ferred to as Mangoldt's method instead of being right-
the horizontal spacing between the conductors remains fully referred to as Markt and Mengele's method, as
constant at a specified value and that the height above will be done here.
ground of each conductor is an average value equal to
H - 2 S where H is the height above ground at the sup- The method of calculation proposed by Markt and
port points and S is the conductor sag corresponding to Mengele may be divided into two stages: the first
the estimated mean annual conductor temperature. stage consists of replacing each bundle by an equiva-
lent (from the point of view of capacitance) single
By making all the assumptions described above, the conductor and determining the total charge on each of
problem of determining the conductor surface voltage them using the simple Maxwell potential coefficient
gradients of a transmission line is transformed to that method; the second stage consists of computing the elec-
of solving the two-dimensional electric field of a tric field of the bundle, by assuming that it is lo-
system of parallel cylindrical conductors above a zero- cated in free space, with the total charge obtained
potential ground plane with known potentials applied to from the first stage distributed uniformly among the
the cylinders. subconductors of the bundle. Further simplifying as-
sumptions are made in the second stage to obtain an an-
An underlying principle involved in almost all the alytical expression for the electric field distribution
methods of solving the electric field problem de- around the individual subconductors of the bundle.
scribed above is the representation of the unknown
charge distributions on the conductor surfaces in order Improved Methods for an Isolated Bundle
to satisfy the specified boundary conditions, namely
that the conductor surfaces are maintained as equipo- Markt and Mengele's method has been used exten-
tential surfaces. The influence of the ground plane is sively6-8 in one form or the other for calculating the
taken into account in all the methods by imaging all conductor surface voltage gradients of practical trans-
the conductors in the ground plane1. mission lines using bundle conductors. There have also
been many attempts at improving the accuracy of Markt
Maxwell Potential Coefficient Method and Mengele's second stage of calculation, namely that
of determining the electric field of an isolated tu-
An exact analytical solution exists for the case dle. Exact analytical solutions have been obtained '
of a single conductor above ground. By imaging in the for the case of an isolated two-conductor bundle. For
ground plane, the problem is transformed into that of the case of the isolated multiconductor bundle, King
solving the electric field of two parallel cylin- suggested an improved solution1l in which each subcon-
drical conductors in infinite space with equal and op- ductor is represented by a line charge located, not at
posite voltages applied to them. The analysis shows the centre of each conductor as in Markt and Mengele's
that the charge distribution on each conductor is rep- method, but at a small distance, determined by the bun-
resented exactly by means of a line charge located at a dle gdometry, from the centre of each conductor. In a
small distance away from the centre of the conductor. subsequent paperl2, King further improved the accuracy
The distance of the line charge location from the cen- of his solution by using two line charges, symmetrical-
tre of the conductor is a direct function of H/r, where ly displaced from the centre of the conductor. Tima-
H is the height of the conductor above ground and r is scheff used the method of conformal transformations13
the conductor radius. For large values of H/r (>100), to obtain an improved solution to the isolated bundle
as in the case of practical transmission line configu- problem. This method has subsequently been usedl4'15
rations, the line charge is located very nearly at the to obtain accurate solutions for practical transmission
centre of the conductor. This basic principle has been line configurations. Aleksandrov utilized the princi-
extended to determine the electric field of multicon- ple of imaging16 to improve the accuracy of the elec-
tric field of an isolated bundle.
ductor configurations where the heights of the conduc-
tors above ground as well as the distances between the
individual conductors are very large compared to the Accurate Methods for the Multiconductor Problem
radii of the conductors. The charge on each conductor
is then represented by a line charge located at its In all the methods mentioned above, the effects of
centre. This gives rise to the well-known Maxwell po- the ground plane and of the bundle conductor have been
tential coefficient method3 of calculating the charges treated as if they are independent of each other. In
and electric fields on the conductors of A-multi-con- practice, however, both effects are present simulta-
ductor system. neously, with the result that the total charge of a
bundle is not equally dividede among its subconductors,
The representation of a conductor charge by means as has been assumed in all the methods described above.
of a single line charge at its centre automatically im- The next stage of improvement therefore consists of
plies a uniform charge and electric field distribution taking into account the overall transmission line con-
around the conductor surface. Such an assumption be- figuration including the effects of the ground plane
comes inadequate in the case of transmission lines and of the bundle conductors.
using bundle conductors since the subconductor spacing
in a bundle is only of the order of 10-40 times the An extension of the method of conformal transfor-
subconductor radius. The charge on the individual sub- mations has been used17 to calculate the electric field
conductors is therefore nonuniformly distributed and of a multiconductor bundle above ground. The method is
cannot be represented by a single line charge at its not applicable, however, to practical three phase ac or
centre. bipolar dc transmission line configurations. Three
different approaches have been proposed in recent years
IMarkt and Mengele's Method for the accurate calculation of the conductor surface
electric fields of any general transmission line con-
Markt and Mengele4 were the first to suggest a figuration: method of successive images, method of mo-
method of calculating the conductor surface electric ments and charge simulation method.
field of bundle conductor transmission lines. Later on,
Mangoldt5 gave details of the same method in a compre- The method of successive images18 is based on the
hensive paper dealing with the electrical characteris- principle of imaging19 a line charge in a cylindrical
tics of bundle conductor transmission lines. For some conductor parallel to it in order to maintain the con-
unknown reason, however, this method has often been re- ductor surface as an equipotential. This principle has
1998

been applied to represent the charge distribution on of the conductors, which translates into a tolerance on
the individual conductors of a multiconductor system by the conductor diameter itself. Existing standards in-
a series of image line charges such that all the con- dicate that the tolerance on the conductor diameter
ductor surfaces are equipotentials. Since each succes- could be of the order of ±1 percent. Since the conduc-
sive imaging process restores the equipotential nature tor surface voltage gradient is almost inversely pro-
of the entire surface of a conductor in the system, the portional to the conductor diameter, this would result
process converges in the limit to the exact solution of in an uncertainty of 1 percent in the gradient values
the given multiconductor problem. In fact, an exact also.
analytical solution has recently been derived20 for the
multiconductor problem and it has been shown in this The average height of conductors above ground is
paper that the exact solution obtained is equivalent to also subject to variations for two important reasons.
the method of successive images. Firstly, the conductor sag is subject to variations due
to changes in the conductor temperature. Secondly, the
In the method of moments21'22, the charge distri- effective electrical ground plane below the conductors
bution on the surface of each conductor of a multicon- may vary due to changes in the moisture content of the
ductor system is expanded in terms of a harmonic series soil, in the level of the ground water table, in the
with unknown coefficients, which may be termed as snow cover above ground, etc. Similarly, the spacing
charge coefficients. The unknown charge coefficients between the conductors may vary due to wind conditions.
for all the conductors are determined to satisfy the It is not possible, however, to make a simple correla-
boundary condition that the surface of each conductor tion between variations in average conductor heights
is maintained at a given constant potential. By con- and conductor spacings and the corresponding changes in
sidering a sufficiently large number of charge coeffi- the gradient values.
cients for each conductor, the method of moments gives
the exact solution to the electric field of a mul- In addition to the above, the conductor surface
ticonductor system. In practice, the number of charge voltage gradient is also influenced by the proximity of
coefficients are chosen so as to obtain a desired accu- transmission towers. The gradient values may be in-
racy. creased by as much as 5 percent due to tower proximity.
Finally, it should be noted that the calculated gra-
The charge simulation method23 may be considered dients are nominal values, since the conductors are as-
as a simplified version of the method of moments to ob- sumed to be smooth cylinders. In practice, the conduc-
tain an approximate solution to the electric field tors are stranded and have other surface irregularities
of a multiconductor system. The charge on the conduc- such as nicks, scratches, organic and inorganic materi-
tors is represented in this case by a number of line al deposits, etc. Since the surface roughness factor
charges located uniformly around a circle inside the may change from conductor to conductor, a given nominal
conductor. The magnitudes of these line charges are gradient may not always correspond to the same true
determined to satisfy the constant potential boundary conductor surface gradient.
condition at a number of test points, equal to the num-
ber of the unknown line charges, on the surface of the All the factors discussed above contribute to some
conductor. The choice of the number as well as the lo- extent to the uncertainty in the value of the conductor
cation of the line charges is quite arbitrary in this surface voltage gradient calculated, irrespective of
method, and the accuracy of the method is ascertained the method of calculation used. Since no control can
through numerical experimentation. The method suffers be exercised over these factors, there is an inherent
from the drawback that the conductor is maintained as minimum uncertainty in the gradient values calculated.
an equipotential only at the test points and deviates Taking all the factors into consideration, an optimis-
from the equipotential at intermediate points. Some tic value for this minimum uncertainty will be about
numerical techniques have been proposed24 recently to 5 percent. A very accurate method of calculation will
improve the computational efficiency of the method for have the advantage of not adding any further errors to
a given accuracy. this inherent value. From a practical point of view,
however, it is preferable to set an upper limit to the
ACCURACY CONSIDERATIONS allowable maximum error for the method of calculation.
Thus, errors in the method of calculation of ±1 and ±2
Because of the importance of the conductor surface percent result in overall uncertainty of 5.1 and 5.39
voltage gradient as a basic parameter in corona perfor- percent respectively (rms addition of the two sources
mance calculations, much attention has been paid to the of error is assumed since they are independent of each
development of increasingly accurate methods of gra- other). In other words, even an error of ±2 percent in
dient calculations as discussed in the previous section. the method of calculation increases the overall uncer-
The overall accuracy of the calculated gradient values tainty in the values of conductor surface gradients
depends, however, not only on the accuracy of the meth- calculated by less than 10 percent (from 5 to 5.39),
od of calculation but also on the precision of the in- which seems reasonable from a practical point of view.
put data. The factors influencing the precision of the It may therefore be recommended that any method of cal-
input data are therefore examined below. culation which has an error of less than ±2 percent is
acceptable.
The principal input data for conductor surface
voltage gradient calculations consist of the physical ANALYSIS OF SURVEY RESULTS
parameters of the line, such as the conductor diameter,
the average height above the ground plane of the con- A total number of thirteen transmission line con-
ductors, the spacing between the conductors and, of figurations have been included in the survey question-
course, the line voltage. A degree of uncertainty al- naire. Of these, three configurations with overhead
ways exists, particularly in the values of the physical ground wires have been repeated without the ground
parameters used for the calculations. wires for purposes of comparison. Thus, there are ba-
sically ten configurations, eight of them of ac lines
The different types of aluminum conductors normal- and two of dc lines. These configurations are shown in
ly used on transmission lines conform to ASTM or simi- Figs. 1 to 10 and the corresponding line parameters are
lar standards. One of the aspects specified in these summarized in Table I. Configurations 3, 4 and 6 are
standards is a tolerance on the area of cross section repeated without the ground wires. In the survey ques-
1999
TABLE I

LINE LINE
LINE PARAMETERS
CONFIG- VOLT- H4 n
dg
URATNIGON |AGE- H1 H
|2 H 3: 1 D |2 3 4 |n| cc
(kV) (m) (m) (in) (m) (m) (m) (mM) '(m) - (cm) (cm) (cm)
I~~~~~_
1 345 11.18 - _ _ 7.92 - - - 1 4.475 _ _

2 345 13.61 - - _ 8.31 _ - 2 3.038 45.72 -

3** 345 I 26.31 18.85 12.29 33.93 I229 __..__


12.24 16. 81 12.85 7.32 2 3.165 45.72 1.463
4* 500 22.32 13.94 33.29 - 12.19 - 7.87 - 2 4.069 45.72 0.978 -
5
500 14.43 - - I - 12.19 - - - 3 2.959 45.72 -
6
7 I 1100
765 20.83
21.34
31.49

-
-

_
-

-
13.72
15.24
.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ i.
21.95
-
-

-
1 -

-
41 2.959
8 3.556
45.72 j 0.978
45.72 _

8 2000 45.00 _
- _ 35.00 - - 16[ 3.810 45.72 -

[
9 ]±375 13.91 - _ - 12.19 2 4.577 45.72
| [0j+±1000 18.29 _- - 16.76 8 4.572 45.72 _

Calculations have been made for these three cases with and without the ground wires.

NOTE: In the Table I above, n = number of subconduc-


tors in the bundle, dc = diameter of subconduc- Ground wires 0 0

tor, s = subconductor spacing and dg = diameter [-*- D, --p-


of ground wire.
0

|D l F
0 0 0 0

[2r -D- H4

0 0 0
0 0
H1j
H32

Hi /7l\'l,
Fig. 3 Line configuration 3

o o Ground wires

KD3-4

Fig. 1 Line configuration 1 0 0

0 0
0
H,3

Hi
Hfq

X/BE// D-
-
i

Fig . 2 Line conf iguration 2 Fig. 4 Line conf iguration 4


2000

0 000 00 000o
o o 0 0 0 00 0
0 0
0
0
0
0
0 0
0 0
0

0 0 0 0 0 0 0

o0oo0 00000 00000

Hi
,, 1
Fig. 5 Line configuration 5 Fig. 8 Line configuration 8

0 o Ground wires
0 0 00
1 D2-
0 0

HiI
0 0 O O O O
0

Hi WDi
Hi 'I

II,AN>V,1 k wr * Ao
-- B

Fig. 6 Line configuration 6 Fig. 9 Line configuration 9


0

Jo
0
0 0 O o
00 0 0 00 000 0 0
0 0 0
0 0
0 0 0 0 0
0 _ I0
J7A\XV t iDIX H
Di -1
Hi
o~~~~~~~~77TII o

-1
~A(<>Yf
Fig. 7 Line configuration 7 Fig. 10 Line configuration 10
tionnaire, 3a, 4a and 6a refer to configurations with
ground wires while 3b, 4b and 6b correspond to those The simplest, from a computational point of view,
without ground wires. All configurations, except 7, 8 are those based on the method of Markt and Mengele.
and 10, correspond to actual operating lines25. For most practical configurations, the method can be
adapted for use even on modern electronic calculators.
Seventeen usable the survey question-
responses to Some differences in detail exist, however, between ver-
naire were received*,
of which two were from the same sions of the method used by different participants.
For this reason, calculations have been made for all
person using two different techniques. Basically five
methods of calculation have been used by the different the line configurations using a simple version of Markt
participants: 1) methods based on the Markt and Men- and Mengele's method, the details of which are given
gele's technique; 2) method of successive images or below:
simplified versions of it; 3) charge simulation meth-
ods; 4) method of conformal transformations; 5) inte- Step 1: Each conductor bundle is replaced by an equiv-
gral equation method. Of the seventeen sets of results alent conductor having a radius req defined by
received, five belonged to the first category, six in 1
the second category, four in the third category and one
each in the fourth and fifth categories. r {n r Anl]n
eq

* where, n is the number of subconductors,


Participants in the survey included: M. Abdel-Salam r is the subconductor radius,
(Ain-Shams Univ.), R. Carberry (Northeast Utilities), A is the bundle radius.
V.L. Chartier (BPA), M.G. Comber (GE), R. Cortina
(ENEL), G.R. Elder (Dept. of Water & Power, LA), C. Step 2: With the bundles represented by the equivalent
Gary (EDF), F. Hirsch (FGH), S.Y. King (Univ. of Hong conductors, the total charge on each of them
Kong), P.S. Maruvada (IREQ), K.E. Ottosen (Consumers is calculated by the Maxwell potential coeffi-
Power Co.), H. Parekh (Univ. of Waterloo), P. Pirot- cient method assuming appropriate potentials
te (Univ. of Liege), J. Reichman (Ontario Hydro), J. on the different phases or poles. Any ground
P. Reilly (Johns Hopkins Univ.) and A.S. Timascheff wires are also taken into account.
(IREQ).
2001

Step 3: Knowing the total charge on the bundle qt from 4r


the preceding step, the average bundle gra-
dient is calculated as: 3
E
av
-
qt
27r 0
1
*n r is
2 I.
Step 4: The average maximum bundle conductor gradient
is then obtained as:
21
Em = Eav[l + (n-l)A 4,
CP
a-
- i .! 3
/A-
/..A--
~
l A---I-
~ ~ ~ .

5 77 '889 l/11 12 13 14 vW16


4
--

17
In the case of Markt and Mengele's method, since the -1
bundle is assumed to be isolated, the maximum bundle Participant no.
conductor gradient will also be equal to Em. -i
01 - Average bundle gradient, center phase
All the results, including those obtained using -2 0 - Average bundle gradient, outer phase
the simple method described above, are summarized in Average max. bundle gradient, center phase
Tables IIa and IIb. Participant number 1 refers to the A.A- Average max. bundle gradient, outer phase
simple method described above, while numbers 2 to 18 -3 Max. bundle gradient, center phase
refer to the seventeen responses received. Preliminary Max. bundle gradient, outer phase
versions of Tables IIa and IIb have been sent to all -4'
the participants for any possible corrections. Seven
of the participants sent back some corrections in their Fig. 12 Accuracy of different methods of calculation
results, and some of them took this opportunity to com- for line configuration 2.
plete their results. All these corrections and addi-
tions have been included in the tables presented here.
For purposes of evaluating the relative accuracy
° 1
of the different methods of calculation, the method 6
successive images, which has been theoretically shown
to converge to the exact solution, has been used as the
reference method. Computations have been made to a
high degree of precision using this method, and the re- Participant no.
sults are included in Tables IIa and IIb as those of a- I-I
-1
participant number 18.
Figures 11 to 23 show the percentage errors calcu- -2
lated on this basis of all the results included in
Tables IIa and IIb. The results of all the gradient
values calculated, and in the case of ac lines of both -31
outer and centre phases, corresponding to a given line
configuration are included in the same figure. The
identification of the individual curves shown in Figs. -41
11 and 12 apply also to all the remaining figures.
In addition to the gradient values, the different
participants have also included information on computa- Fig. 13 Accuracy of different methods of calculation
tion time, memory requirements of the computer, etc. for line configuration 3a.
The computers utilized were so different, however, that
no attempt was made to compare the computational as-
pects of the different methods of calculation. 2

3 r
Average bundle gradient, center phase cr~~~~~~~~~~~~~~~
2 -
*---O
-
* Ave"ge bundle gradient, outer phase
Average max. bundle gradient, center phase
C 6
b... Average max. bundle gradient, outer phase 4 d: Participant no. /'
o 1

aO

cr 1

-2

-3 -5

Fig. 11 Accuracy of different methods of calculation Fig. 14 Accuracy of different methods of calculation
for line conf iguration 1. for line configuration 3b.
2002

TABLE IIa

METhOD OF ~~~~~COMPUTED GRADIENT FOR THE DIFFERENT LINE CONFIGURATIONS- kV/cm


H CALCULATION <
H 1 ~~~~~2
3a 3b 4a 4b

P4 ~~~~~C.P 0.? C.P O.'P C.P O.P C.P O.P C.P O.P C.P 0.?
P4

Markt& A 15.51 14.62 14.69 13.52 14.49 14.17 14.57 14.16 15.36 15.26 14.97 15.33
1 Marktl& AM 15.51 14.62 15.66 14.42 15.50 15.15 15.58 15.14 16.73 16.62 16.31 16.70
Mengele ~MB 1-5.51 14.62 15.66 14.42 15.50 15.15 15.58 15.14 16.73 16.62 16.31 16.70
A 15.49 14.70 14.68 13.58 14.48 14.20 14.56 14.18 15.38 15.29 14.99 15.36
2 Images AM 15.49 14.70 15.63 14.45 15.44 15.15 15.53 .15.13 16.69 16.59 16.27 16.68
NB 15.49 14.65 15.63 14.66 15.55 15.16 15.64 15.14 16.69 16.78 16.27 16.86
Successive A 15.51 14.66 14.70 13.59 14.58 14.49 15.36 15.27 14.97 15.35
3 Images AM 15.51 14.72 15.65 14.47 15.55 15.46 16.67 16.57 16.25 16.66
(modified) NB 15.51 14.72 15.65 14.68 15.65 15.57 16.67 16.76 16.25 16.84
Modified
4 Markt & AM 15.51' 14.67 15.66 14.49 15.50 15.20 15.58 15.18 16.73 16.63 16.30 16.71
Mengele ___
Method A 15.50 14.67 14.70 13.59 14.49 14.22 14.56 15.37 15.27 14.98 15.35
5 of AM 15.51 14.73 15.68 14.49 15.49 15.20 15.58 16.74 16.63 16.31 16.72
Images NB 15.51 14.73 15.68 14.72 15.60 15.20 15.69 16.74 16.83 16.31 16.92
Conformal A 15.48 14.38 14.66 13.59
6 Tramsfor- A 54 43 56 44
ma tion AM154_1_8_5_3144
Charge ~ A 15.50 14.28 14.69 13.68 14.59 14.22 14.59 14.20 14.79 14.12 14.79 14.12
7 Chmuarge
Simulation
AM 15.64
NB 15.64
14.51
14.51
15.78
15.78
14.70
15.01
15.71
15.81
15.28
15.28
15.71
15.81
15.25
15.25
16.21
16.21
15.39
15.52
16.21
16.21
15.39
15.52
m,odified
8 Markt &
A
NB
14.69 13.58
14.49
14.50 14.58 15.37 15.27 14.97 15.35,
15.51 14.67 15.66 15.50 15.59 16.73 16.63 16.31 16.72
Mengele
9 Modified AM 15.67 14.49 15.49 15.58 16.73 16.63 16.31 16.71
__ Mengele NB 15.51 14.66 15.67 14.63 15.57 15.65 16.73 16.75 16.31 16.83
A 15.50 14.70 14.70 13.55 14.50 14.20 14.55 14.20 15.40 15.30 15.00 15.35
10 Images AM 15.50 14.70 15.60 14.45 15.45 15.15 15.50 15.10 16.70 16.60 16.30 16.70
NB 15.50 14.70 15.60 14.70 15.60 15.20 15.60 15.10 16.70 16.80 16.30 16.90

11
Charge
Simuation ~AAM 15.51 14.69
15.65
14.49
15.47
14.57
15.56
15.37
16.69
14.98
16.27
Simultion
NB 15.53 15.65 15.58 15.67 16.69 16.27
Integral A 15.53 14.71 14.53 14.61 15.29 15.36
12 Inutegra
Equation
AM
NB
15.53
15.53
15.67
15.67
15.51
15.62
15.59
15.70
16.59
16.78
16.67
16.86

Charge A 15.51 14.69


13 Simulation
Siuain
AM 15.51 15.64
B 15.51 15.64
Modified A 15.49 14.65 14.69 13.59 14.52 13.55 14.60 13.62 15.41 15.32 15.02 15.39
14 Markt & AM 15.50 14.71 15.64 14.46 15.48 14.44 15.57 14.52 16.70 16.60 16.28 16.69
Mengele NB 15.50 14.71 15.64 14.67 15.58 14.58 15.67 14.66 16.70 16.79 16.28 16.88

15
Charge
ChreAM ~A 15.51
15.51
14.72
14.72
14.69
15.62 14.44
14.49
15.43
14.57
15.52
15.27
16.54 16.66
15.35
16.63 16.24
Simulation NB 15.51 14.72 15.62 14.65 15.54 15.62 16.73 16.66 16.81 16.24
Method A 15.51 14.69 13.59 14.50 14.22 14.58 14.20 15.36 15.27 15.35 14.97
16

Imge NB 15.51 15.66 14.49 15.50 15.20 15.59 15.18 16.73 16.63 16.71 16.30
A 15.51 14.66 14.69 13.59 14.49 14.21 14.57 14.20 15.37 15.27 14.98 15.35
Successive A 55 47 56 44 54 52 55 52 66 65 62 66
17 Images NB 15.59 14.79 15.64 14.47 15.60 15.33 15.568 15.324 16.68 16.63 16.26 16.71
Succesive A 15.51 14.62 14.69 13.53 14.49 14.17 14.57 14.16 15.36 15.26 14.97 15.33
18 Successie AM 15.51 14.68 15.64 14.40 15.46 15.12 15.54 15.10 16.67 16.56 16.25 16.64
Images
N~B 15.51 14.68 15.64 14.59 15.56 15.12 15.65 15.10 16.67 16.75 16.25 16.83

A: Average bundle gradient; AM: Average maximum bundle gradient; NB: Maximum bundle gradient; C.P.: Center
Phase; O.P.: Outer Phase. In the cases of 3a and 3b, C.P. refers to middle phase and 0.P. to bottom phase.
2003

TABLE IIb

0
z COMPUTED GRADIENT FOR THE DIFFERENT LINE CONFIGURATIONS - kV/cm
M- ETHOD OF
A4 CALCULATION E-
H ~~~~Z 5 6a 6b 7 8 9 10

C.P O.P C.P O.P C.P O.P C.P O.P C.P O.P

Markt& A 15.22 14.15 18.53 17.15 18.53 16.95 13.69 12.31 11.41 10.32 17.46 16.15
1 Mrt&
Mengele
AM 16.93 15.74 21.07 19.50 21.07 '19.28 16.55 14.87
14.87
14.19
14.19
12.84
12.84
19.21
19.21
20.47
20.47
MB 16.93 15.74 21.07 19.50 21.07 19.28 16.55
A 15.24 14.22 18.54 17.24 18.55 17.06 13.72 12.40 11.38 10.35 17.46
2 Images AM 16.91 15.79 21.06 19.58 21.07 19.38 16.61 15.02 14.23 12.95 19.12 20.54
MB 16.92 15.96 21.07 19.85 21.07 19.67 16.68 15.68 14.28 13.48 19.40 21.58
Successive A 15.22 14.21 18.53 17.23 18.54 17.05 13.72 12.41 17.46 16.15
3 Images AM 16.89 15.77 21.06 19.57 21.06 19.36 16.61 15.02 19.13 20.53
(modified) NB 16.90 15.94 21.06 19.84 21.06 19.66 16.67 15.68. 19.40 21.57
Modified
4 Markt & AM 16.93 15.80 21.07 19.59 21.07 19.38 16.55 14.98 14.19 12.93 19.21 20.47
Method A 15. 23 14.21 18.54 17.23 18.54 17.06 13.,69 12.39 11.40 10.39 17.46 16.,15
S of AM 16.931 15.81 21.08 19.60 21.09J 19.39 16.54 14.97 14.18 12.92 19.22 20.48
Images MB 16.94 15.99 21.09 19.88 21.09 19.69 119.49 21.55
Conformal A 15.15 14.10 18.49 17.13 13.67 12.53 11.39 10.48 17.48 16.22
6 Transfor- AM 16.85 15.68 21.02 19.47 16.48 15.11 14.12 12.99 19.22 20.47
___ matiomn_ _ __ _ _ _

A 15.39 14.34 18.54 17.37 18.54 17.37 15.57 15.88


Chmuarge
Charge
7 Simultion
AM 17.15 15.95 20.79 19.51 20.79 19.51 19.32 20.46
MB 17.27 16.29 20.79 20.01 20.79 20.01 19.61 21.84

8
Modified
Markt &
A 15.22 14.22 17.39 16.25 17.39 16.08 '10.86 10.03 7.78 7.30
16.43
___Mengele
MB 16.93 15.80 20.77 19.40 20.77 19.20 16.77 15.50 17.50
Modified AM 17.13 15.97 21.08 19.59 21.08 19.38
9 Markt & NB 17.44 16.28 21.08 19.79 21.08 19.60
Mengele
A 15.20 14.27 18.50 17.20 17.55 17.05 13.73 12.40 11.38 10.35 17.45
10 Images AM 16.90 15.80 21.10 19.60 21.10 19.38 16.61 15.12 14.24 12.96 19.15
NB 16.90 16.00 21.10 19.90 21.10 19.70 16.70 15.70 14.30 13.50 19.40
Charge
Chre A 15.23 j18.53 18.54 13.70 17.47 16.16
11 AM 16.90 21.05 21.05 16.58 19.15 20.53
Simulation NB 16.90 21.05 21.06 16.60 19.43 21.59
A 15.25 18.56 18.57 13.74 11.49 17.49 16.18
12 Inutegra
Equation
AM
NB
16.93
16.94
21.09
21.10
21.10
21.10
16.66
16.69
14..41
14.43
19.17
19.44
20.59
20.6

Charge A 15.22 18.54 13.70 11.42 17.47 16.16


13 Chmuarge
Simultion n
AM
B
16.90
16.90
21.05
21.06
16.59
16.62
14.27
14.29
19.13
19.29
20.53
21.26
Modified A 15.24 14.22 18.57 17.26 18.57 17.08 13.78 12.46 11.86 11.24 17.47 16.27
14 Markt & AM 16.89 15.76 21.06 19.57 21.06 19.36 16.60 15.02 19.12 20.52
Mengele NB 16.90 15.94 21.06 19.84 21.06 19.66 16.66 15.66 19.39 21.55
A 15.23 18.54 18.54 13.72 11.,52 17.47 16.16
15 Charge AM 16.85 15.73 21.06 19.57 21.06 19.36 16.56 14.98 14.12 12.85 19.08 20.46
Simulation NB 16.85 15.91 21.06 19.84 21.07 19.66 16.62 15.58 14.48 13.58 19.35 21.40
Method A 15.22 14.21 18.53 17.23 18.53 17.04 17.46
16_ Imge NB 16.93 15.80 21.07 19.59 21.07 19.38 19.20
Succesive
A 15.23 14.21 18.54 17.23 18.54 17.05 13.72 12.41 11.43 10.40 17.46 16.16
17 Successie
Images
AM
NB
16.91 15.77
17.03 15.92
21.06
21.18
19.58
19.75
21.06
21.17
19.37
19.55
16.62
16.84
15.04
15.37
14.29
14.48
13.01
13.30
19.13
19.18
20.59
21.13
Succesive
A 15.22 14.15 18.53 17.15 18.53 16.95 13.70 12.31 11.41 10.33 17.46 16.15
18 Sucsie AM 16.89 15.71 21.05 19.48 21.05 19.26 16.59 14.91 14.28 12.92 19.13 20.55
Images NB 16.90 15.87 21.05 19.74 21.06 19.54 16.63 1S.SO 14.30 13.39 19.40 21.58'
A: Average bundle gradient; AM: Average maximum bundle gradient; N-B: Maximum bundle gradient; C.P.: Center
Phase; O.P.: Outer Phase.
2004

h-

c /12 3 4 6 7 10 11 12 13 14\15 ,'17 c 1 2 34 5 6 7 8, 12 13 t4 17


/ \1\*
\| !, Participant no
-1 / \\ |1 ,#, Participant no.
0--i~~~~~~~~~~~r
-2- -2_

-3
Fig. 18 Accuracy of different methods of calculation
-4- for line configuration 6a.
3_
-5 (707)
7 34)
747) 2-1
Fig. 15 Accuracy of different methods of calculation
for line configuration 4a.

0
2 12 3 4 7- I 1
31
f -1 P | *\\ $/Participant no.

-2
0
1
-3L
a)
0. Fig. 19 Accuracy of different methods of calculation
c for line configuration 6b.
a)
a. no.
-i DISCUSSION OF RESULTS
The most encouraging aspect of the results pre-
-21 sented above is the remarkable accuracy in most cases
of the straightforward application of Markt and Men-
gele's method. As can be seen from the step by step
description of the method, the use of large computers
may be necessary only for step 2, especially for double
circuit lines. For single circuit lines even with
ground wires, the matrix inversion involved in step 2
A, (-7.5i) can be performed on some of the modern electronic cal-
a- (-7.78) culators. Thus, except for double circuit lines or
*- (-7 89) other more complicated line configurations, all the
Fig. 16 Accuracy of different methods of calculation calculations can be performed on electronic calcula-
for line configuration 4b. tors. The results of Figs. 11-23 show that the simpli-
fied version of Markt and Mengele's method gives aver-
4_ age bundle gradients with less than 0.1 percent error,
and average maximum bundle gradients with less than
3-
o 1 percent error for all configurations. The method al-
so gives maximum bundle gradients with less than 2 per-
cent error for all line configurations having four sub-
2- conductors or less in the bundle. For configurations
with larger number of subconductors, the error may ap-
proach 5 percent or more.
It is inferred that five of the participants in
the survey have used a method of calculation based on
Markt and Mengele's. However, the exact details of
i)1 2 3 4 55 7 Q -7T--\ i
each method are not known, and it is possible that dif-
0- Participant no ferences exist between the individual methods. This is
probably the reason for the rather wide variation in
the accuracy of the results: the maximum individual
errors varied between 0.4 and 4.5 percent for the aver-
age bundle gradient, between 0.6 and 4.5 percent for
Fig. 17 Accuracy of different methods of calculation the average maximum bundle gradient and between 1.7 and
for line configuration 5. 3.6 percent for the maximum bundle gradient.
2005

0~~~~~~~~~~~~
A.1
0

-1 Participant no.
c 123 i10 12

-1 Nz, / Participant no. -2

-2 \ " -3

-3 _ -4-

-4 -

-5L - (-10.82)
Fig. 20 Accuracy of different methods of calculation Fig. 22 Accuracy of different methods of calculation
for line configuration 7. for line configuration 9.
0 (8.81) 3r

0
o4) -A I N.-,.~~~~
O -
I- . ~ = l6lm --; -- .L=F = .--

I-
- -

j
. .

--1
ni l!
0
0
aL)
0- -21
QoCP
0
c
a)
-3
Participant no.
-4 I I I
XIL
_5L
Fig. 23 Accuracy of different methods of calculation
for line configuration 10.

varied between 0.07 and 10.8 percent for the average


bundle gradients, between 0.1 and 7.5 percent for the
average maximum bundle gradient and between 0.2 and 7.8
Fig. 21 Accuracy of different methods of calculation percent for the maximum bundle gradient.
for line configuration 8.
The method of conformal transformations gave rea-
sonably accurate (less than 2 percent error) results
Methods based on the principle of successive im- for the average bundle and the average maximum bundle
ages have been used by six participants. The results gradients. The errors approached 4 to 5 percent, how-
obtained by the individual participants in this catego- ever, for the maximum bundle gradient for configura-
ry show the best consistency: the maximum individual tions having more than four subconductors in the bundle.
errors varied between 0.65 to 0.85 percent for the av- Similar comments apply also to the integral equation
erage bundle gradient, between 0.7 and 1.4 percent for method.
the average maximum bundle gradient and between 0.8 and
2.1 percent for the maximum bundle gradient. Finally, the results provide an evaluation of the
influence of ground wires on the calculated gradients.
The third important category of methods used is
Comparing the results obtained for configurations with
and without ground wires, it is observed that: in the
the charge simulation method. The results obtained by case of the double circuit configuration, the presence
the four participants using this method showed the most of ground wires decreases the middle phase gradients by
dispersion in accuracy: the maximum individual errors about 0.5 percent while scarcely affecting the bottom
2006
phase; the presence of ground wires has the maximum in- Minimum single conductor (or subconductor) gradient -
fluence on triangular configurations, increasing the The minimum value attained by the gradient E(O) as
centre phase gradient by up to 2.5 percent while de- given above as 0 varies over the range 0 to 27r.
creasing the outer phase gradient by about 0.5 percent;
for the flat configuration, the ground wires have neg- Average single conductor (or subconductor) gradient -
ligible effect on the centre phase gradients while the The value E obtained from
av
outer phase gradients are increased by about 1 percent. 27
CONCLUSIONS Eav | E (8) d(e)
The following main conclusions can be drawn from Approximately the average conductor gradient is given
the results of the survey presented in this paper: by
1. From a practical point of view, errors due to the E av
q
method of calculation of up to ±2% are acceptable 2Te0 r
for conductor surface voltage gradients.
where q = total charge on the conductor per unit length
2. A simple method of calculation, based on the Markt
and Mengele's method, gives sufficiently accurate o= permittivity of free space
r = radius of the conductor
results for transmission line configurations with
up to four subconductors in the bundle. Note: For practical cases the average conductor gra-
dient is approximately equal to the arithmetic
3. More complex methods such as the method of succes- mean of the maximum and minimum conductor gra-
sive images, the method of moments or the charge dients.
simulation method are required to obtain results of
acceptable accuracy for line configurations having Average bundle gradient - For a bundle of two or more
more than four subconductors in the bundle. subconductors, the arithmetic mean of the average gra-
All the methods based on the principle of succes-
dients of the individual subconductors.
4.
sive images give consistent results having errors Average maximum bundle gradient - For a bundle of two
of less than 2 percent. A large dispersion is ob- or more subconductors, the arithmetic mean of the maxi-
served in the errors in gradient values calculated mum gradients of the individual subconductors.
by methods based on the charge simulation technique.
Some methods gave errors less than 0.2 percent For example, for a 3-conductor bundle with individual
while others gave errors of more than 10 percent. maximum subconductor gradients of 16.5, 16.9, and
17.0 kV/cm,the average maximum bundle gradient would be
5. The influence of overhead ground wires on the con-
ductor surface voltage gradients depends on the 1/3 (16.5 + 16.9 + 17.0) = 16.8 kV/cm
line configuration, the highest being observed for
triangular configuration. Maximum bundle gradient - For a bundle of two or more
subconductors, the highest value amongst the maximum
APPENDIX gradients of the individual subconductors.
The terms relating to the conductor surface volt- For example, for a 3-conductor bundle with individual
age gradients used in this paper are taken from the maximum subconductor gradients of 16.5, 16.9, and
IEEE Standard presently in preparation (IEEE Standard 17.0 kV/cm, the maximum bundle gradient would be
Project P-539) on "Definition of terms relating to over- 17.0 kV/cm.
head power line corona and radio noise". Definitions
of the relevant terms are reproduced below:
Nominal conductor gradient - The gradient determined
for a smooth cylindrical conductor whose diameter is
Voltage gradient - A vector E equal to and in the di- equal to the outside diameter of the actual (stranded)
rection of the maximum space rate of change of the conductor.
voltage at the point specified. It is obtained as a
vector field by applying the operator V to the scalar REFERENCES
voltage function u. Thus if u = f (x,y,z)
1. J.C. Maxwell, "At
-E = Vu = grad u = i -aax + j aay + k aaz tism", London, Clarendon Press, 1892, Ch. XI.
2. E. Weber, "Electromagnetic theory", New York, Dover
Notes: 1) Voltage gradient is synonymous with poten- 1965, pp. 119-122.
tial gradient and is often referred to sim-
ply as "gradient". 3. G.E. Adams, "Voltage gradients on high voltage
transmission lines", AIEE Trans. (Power Apparatus
2) For alternating voltage, voltage gradient and Systems), Vol. 74, pp. 5-11, April 1955.
is expressed as the peak value divided by
the square root of two. For sinusoidal 4. G. Markt and B. Mengele, "Drehstromferntibertragung
voltages, this is the rms value. mit BUndelleitern", E und M, Heft 20, pp. 293-298,
May 1932.
Maximum single conductor (or subconductor) gradient -
The maximum value attained by the gradient E(e) as 0 5. W. von Mangoldt, "Electrical fundamentals of bun-
varies over the range 0 to 2vr, where E(0) is the gra- dle conductors" in BUndelleitungen, Berlin, Germa-
dient on the surface of the power line conductor (or ny, Siemens-Schuckert-WJerke AG, 1942, pp. 3-11.
subconductor) expressed as a function of angular posi-
tion 0. Unless otherwise stated, the gradient is nom- 6. C.J. Miller, Jr., "Mathematical prediction of ra-
inal gradient (defined below). dio and corona characteristics of smooth, bundled
2007

conductors", AIEE Trans. (Power Apparatus and Sys- 17. H. Singer and H. BYssler, "Berechnung des elektri-
tems), Vol. 75, pp. 1029-1037, October 1956. schen Feldes von BUndelleitern", Bull. ASE, Vol. 59,
No. 4, pp. 179-183, 1968.
7. W.E. Pakala and E.R. Taylor, Jr., "A method for
analysis of radio noise on high-voltage transmis- 18. Maruvada P. Sarma, W. Janischewskyj, "Electro-
sion lines", IEEE Trans. Power Apparatus and Sys- static field of a system of parallel cylindrical
tems, Vol. PAS-87, pp. 334-345, February 1968. conductors", IEEE Trans. on Power Apparatus and
Systems, Vol. PAS-88, pp. 1069-1079, July 1969.
8. N.N. Tikhodeev, "On the calculation of initial
voltages of total corona on dc transmission lines", 19. P. Hammond, "Electric and magnetic images", Proc.
Elektrichestvo, Vol. 10, pp. 12-19, Oct. 1957. IEE (London), pt. C, monogr. 379, pp. 306-313, May
1960.
9. A.I. Darevski, "The electrostatic field of a split-
phase", Elektrichestvo, No. 9, pp. 16-19, 1958. 20. V. Alessandrini, H. Fanchiotti, C.A. Garcia Canal
and H. Vucetich, "Exact solution of electrostatic
10. G. Quilico, "The electric field of twin conductors problem for a system of parallel cylindrical con-
(rigorous solution)", Elettrotecnica (Italy), No. ductors", Journal of Applied Physics, Vol. 45, No.
10, pp. 530-538, 1954. 8, pp. 3649-3661, 1974.
11. S.Y. King, "The electric field near bundled conduc- 21. Maruvada P. Sarma, "Application of moment methods
tors", Proc. IEE (London), pt. C, monogr. 338S, pp. to the computation of electrostatic fields. Part
200-206, June 1959. I: Parallel cylindrical conductor systems", IEEE
Conference paper No. C72 574-2, 1972.
12. S.Y. King, "Improved solution for the field near
bundle conductors", Proc. IEE (London), Vol. 110, 22. J.C. Clements, C.R. Paul and A.T. Adams, "Computa-
No. 6, pp. 1044-1050, June 1963. tion of the capacitance matrix for systems of di-
electric-coated cylindrical conductors", IEEE Trans.
13. A.S. Timascheff, "Field patterns of bundle conduc- on Electromagnetic Compatibility, Vol. EMC-17, No.
tors and their electrostatic properties", AIEE 4, pp. 238-248, 1975.
Trans. (Power Apparatus and Systems), Vol. 80, pp.
590-597, October 1961. 23. M.S. Abou-Seada and E. Nasser, "Digital computer
calculation of the potential and its gradient of a
14. A.S. Timascheff, "Fast calculation of gradients for twin cylindrical conductor", IEEE Trans. on Power
the center phase of a three-phase bundle conductor Apparatus and Systems, Vol. PAS-88, pp. 1802-1814,
line with any number of subconductors", IEEE Trans. December 1969.
on Power Apparatus and Systems, Vol. PAS-90, pp.
157-164, Jan./Feb. 1971. 24. M. Khalifa, M. Abdel Salam, F. Aly and M. Abou-
Seada, "Electric fields around conductor bundles of
15. A.S. Timascheff, "Fast calculation of gradients of EHV transmission lines", IEEE Conference paper No.
a three-phase bundle conductor line with any number A 75 563-7, 1975.
of subconductors. Part II: Gradient calculations
for the side phase", IEEE Trans. on Power Apparatus 25. Transmission Line Reference Book 345 kV and above,
and Systems, Vol. PAS-94, pp. 104-107, Jan.!Feb. Electric Power Research Institute, Palo Alto, 1975,
1975. Ch. 3.
16. G.N. Aleksandrov, "Corona discharge on transmission
lines" (in Russian), Moscow, Energia, 1964,- pp. 48-
61.
2008

Discussion h. Double circuit lines


Surface voltage gradient values of double circuit lines are in-
Stephen A. Sebo (The Ohio State University, Columbus, Ohio): The fluenced by their phase arrangement besides the factors men-
voltage gradient survey paper provides an excellent, clear and practical tioned previously. For example, for the "vertical" arrange-
way of comparison of several calculation methods. The paper is valuable ments lowest voltage gradients can be found in case of phase
since it provides an up-to-date survey of techniques that are available for sequence ABC/ABC from top to bottom (center-line sym-
investigations of the surface voltage gradients of transmission line con- metrical or super-bundle arrangement), and highest voltage
ductors. Two comments are offered here: gradients are produced by ABC/CBA arrangement (center-
1. It might be of some interest to briefly summarize the effects of point symmetrical or low-reactance arrangement).
factors influencing the conductor surface voltage gradient. i. Gradient distribution around subconductors
a. Operating voltage of the line Surface voltage gradients are not uniform around the periphery
Conductor surface voltage gradients are directly proportional of subconductors. The variation of surface voltage gradients
to the operating voltage of the line. is caused by the presence of two fields. One is a symmetrical
b. Spacing between phases radial (self) field, and the other is the field of other subcon-
Increasing phase spacing reduces surface voltage gradients. ductors. This variation can be approximated by the cosine law:
Table III summarizes some representative numerical values
to illustrate the extent of the effect in case of horizontal
single-circuit lines, if the phase spacing is increased by 10 per E(e) = Eav +-(n - cos
cent. The larger the number of subconductors of the bundle
the stronger is the effect. The cosine-law does not consider the effect of the other two
phases, ground wires, and ground. Maximum deviation from
Table III. Effect of 10% increase in phase spacingon surface the cosine curve is less than 0.5% of the maximum voltage
voltage gradient (horizontal single-circuit lines). gradient if n = 2, 3 or 4, and 1.0 to 1.7% if n = 6, 8, 10 or 12.
These figures indicate the deviations when 0 = 90°. At 0 = 00
Voltage level Voltage gradient reduction, per cent, for and 1800 deviations are zero.
kV center phase outside phase j. Gradient distribution around bundle
345, 500, 765 1.5 - 2.1 1.4 - 1.7 Surface voltage gradient variations around a bundle that con-
1200,1500 2.2 - 2.7 1.7 - 2.2 sist of a larger number of subconductors can be approximated
by a sinusoidal function.
c. Conductor height above ground k. Stranding
The effect of conductor average height above ground is small; Almost all computation methods and numerical results of
for practical purposes the effect is negligible for horizontal voltage gradient calculations are based on the assumption that
single-circuit lines (Table IV). Numerical values shown are the surface voltage gradients of stranded conductors used in
surface voltage gradient changes if the average height of con- practice are about 35-45 per cent greater than that of a
ductors is increased by 10 per cent. smooth conductor of the same outside radius and the same
charge. The ratio of these two voltage gradients is the strand
Table IV. Effect of 10%o increase in conductor average height factor, F. The corona inception voltage of stranded conductors
above ground on surface voltage gradient (horizontal is lower than that of smooth conductors.
single-circuit lines). The strand factor is a function of the number of strands in
the outermost layer of the conductor and does not depend on
Voltage level Voltage gradient changes, per cent, for the diameter of the conductor.
kV center phase outside phase 1. Surface and atmospheric conditions
Surface voltage gradient values are increased locally by factors
345 +0.05 -0.23 related to conductor surface and ambient atmospheric condi-
500,765 40.07... 0.17) -(0.11... 0.06) tions. The conductor surface is not perfect due to deficiencies
1200, 1500 -0.27 +0.07 (nicks, scratches, burrs, roughness) and solid particle con-
tamination (dust, solid particles, insects, bird droppings, etc.).
d. Number and diameter of subconductors in bundle Fair weather atmospheric conditions, i.e., pressure and tem-
Assuming the same tower geometry, voltage gradients decrease perature, do not change conductor surface voltage gradients
with increasing subconductor number and diameter, or in other but affect the visual corona inception (or critical) voltage
words, with total bundle cross-sectional area. It is a good ap- gradient. If 6 relative air density increases, the critical voltage
proximation to consider about 8% decrease of the voltage gradient also increases. Foul weather conditions change local
gradient for 10%o increase in conductor diameter, for 2-, 3-, surface voltage gradients greatly, due to the presence of rain-
and 4-subconductor bundles. drops, snowflakes, hoarfrost, and ice on the surface of the
e. Subconductor spacing conductor, Peek's critical voltage gradient also changes: during
Subconductor spacing is influenced by several factors: rain it is, e.g., one fourth the theoretical (fair weather) value.
mechanical motion (striking and sticking) of subconductors Peek's modified equation is a good approximation to deter-
due to wind and system faults, weather conditions (icing be- mine the Eo visual corona inception voltage for parallel strand-
tween subconductors), electrical parameters (inductance, ed conductors of overall radius r cm:
capacitance, surface voltage gradient), tower design, and 0.301-
cost. Subconductor spacing is usually 30-61 cm (12-24 in), kV peak/cm
the most frequently occurring spacing is 46 cm (18 in). Opti- Eo = 30 6m 1 +
mum subconductor spacing, with respect to surface voltage
gradients only, is about 8-14 times the subconductor diameter where m surface irregularity (or roughness) factor is usually
for bundles of 2, 3, or 4 subconductors. 0.75 for stranded and tension-strung, and 0.55 for stranded
f. Position of bundle conductor in single-circuit horizontal and dragged conductors. This factor reflects stranding and sur-
arrangement face conditions together.
For commonly used horizontal and vertical configurations, 2. The survey paper gives a very comprehensive list of papers which
maximum surface voltage gradients are found on the center describe and discuss particular voltage gradient computational methods.
phase. For horizontal arrangements voltage gradients of the In order to facilitate the reader's review of additional papers, a few more
outside phase conductors are 5 to 10 per cent less than those are listed here in the Reference Section (26-42).
of the center phase conductor.
g. Ground wires References
The effect of ground wires is practically negligible in case of
horizontal configurations. The presence of ground wires tends 26. M. Temoshok, "Relative surface voltage gradients of grouped con-
to change the charge distribution and surface voltage gradients. ductors," AIEE Trans. Power Apparatus and Systems, Vol. 67, pp.
This change is less than 0.25 percent for 735 and 765-kV lines. 1583-1 591, 1948.
2009

Reichman, "Bundled conductor voltage gradient calculations,"


27. J. After a thorough comparative analysis of different methods, the
AIEE Trans. Power Apparatus and Systems, Vol. 78, pp. 598-607, Subcommittee decided to consider as acceptable any method, whose
Aug. 1959. additionalerror (due to its imperfection) does not exceed 2%. This is
28. A. S. Timascheff, "Equigradient lines in the vicinity of bundled clearly expressed in the conclusion of the paper.
conductors," AIEE Trans. Power Apparatus and Systems, Vol. 82, Fortunately, a simple method by MARKT & MENGELE does not
pp. 104-1 10, April 1963. exist, which can be applied in most cases keeping the possibleerror of
29. R. H. Galloway, H. M. Ryan and M. F. Scott, "Calculation of elec- the calculation within the just mentioned limits.
IEE, Vol. 114, pp. 824-829,
tric fields by digital computer," Proc. To deal with cases where the mentioned method is not accurate
June 1967. enough (i.e. bundles with more than four subconductors) quite accurate
30. P. Thanassoulis and R. P. Comsa, "Calculation of maximum voltage methods are available (see i.e. Ref. 18).
gradient, PartI: Bundle conductors, Part II: Split bundle conduc- It should be commended that the simple method based on the ap-
tors," IEEE Trans. on Power Apparatus and Systems, Vol. PAS-90, proach of Markt & Mengele is accepted now by the IEEE under their
pp. 145-157,Jan./Feb. 1971. names (and not under the name of W.v.Mangoldt), as it was strongly
31. M. S. Abou-Seada and E. Nasser, "Calculation of the potential suggested in one of the discussions of a paper by P. S. Maruvada & R. M.
gradient of twin-cylindrical bipolar conductors with various geo- Morris (Trans IEEE. PAS, Nov/Dec. 1976, pp. 1934-1945).
metrical parameters," IEEE Trans. on Power Apparatus and Sys- There is no point at the present time to look for new methods be-
tems, Vol. PAS-90, pp. 1822-1829, July/Aug. 1971. cause both, simplified and "accurate" ways are readily available. In the
32. R. N. Allan and J. E. J. Cottrill, "Design and parameters affecting writers opinion, such efforts would be a waste of time and also of the
the surface stress of overhead-power-line conductor systems," Proc. rather restricted space in the IEEE-Publications.
IEE, Vol.118, pp. 1486-1492, Oct. 1971. (This point of view could possibly become of a certain value also
33. H. Steinbigler, H. Singer and S. Berger, "Berechnung der Randfeld- in other sectors of the IEEE-activity, where similar surveys would be
starken von Bundelleitem in Drehstromsystemen," ETZ-A, Vol. 92, from time to time of a definite value.)
pp. 612-617, Nov. 1971. A last remark has to be made concerning the real limit of acceptable
34. J. G. Andrews and A. J. Shrapnel, "Electric field distribution around inaccuracies.
an isolated stranded conductor," Proc.IEE, Vol. 119, pp. 1162- According to the present survey, a 2% error is being rightfully con-
1166, Aug. 1972. sidered as a limit. This criterion has been used, however, not only for
35. A. S. Timascheff, "Further studies on power transmission by multi- the average gradient and the average maximum bundle gradient, but also
conductor bundles and on field and gradient distribution around for the maximum gradient (maximum maximorum). To the writers
them," Conference Paper No. C73 161-7, New York: IEEE Power
Engineering Society, 1973.
knowledge, itis, however, presently still undecided, whether the "average
maximum" or the "maximum" is the limiting factor in terms of its in-
36. P. Thanassoulis and R. P. Comsa, "Distribution of voltage gradients fluence upon RI and AN.
in bundled transmission lines by the dipole method," IEEE Trans. If the average maximum bundle gradient should prove to be more
on Power Apparatus and Systems, Vol. PAS-92, pp. 769-774, important than the maximum gradient, then several approximate
March/April 1973. methods could be accepted as being within the desired limit of accuracy.
37. H. Parekh and E. Nasser, "Computation of the potential and fleld
of a stranded twin-bundle conductor above ground," Conference
Paper No. C 74 084-0, New York: IEEE Power Engineering Society,
Manuscript received March 2, 1979.
1974.
38. H. Parekh, M. S. Selim and E. Nasser, "Computation of electric R. M. Morris and A. R. Morse (National Research Council Canada,
field and potential around stranded conductor by analytical
method and comparison with charge-simulation method," Proc. Ottawa, Ontario, Canada) and P. S. Maruvada (Hydro Quebec, Varennes,
IEE, Vol. 122, pp. 547-5 50, May 1975. Quebec, Canada): The authors of this discussion published a series of
39. H. Parekh and E. Nasser, "The effect of geometric parameters on charts for convenient determination of gradients on conductors in bi-
electric field and capacitance of stranded conductors above polar D.C. line configurations (1) and recently have added graphs to
ground," Conference Paper No. A 75 540-5, New York: IEEE these charts to include bipoles of wider spacing and also monopolar
Power Engineering Society, 1975. configurations. Gradient factors were computed using the method of
40. P. Y. Foo and S. Y. King, "Bundle conductor electric field by inte- successive images (participant No. 18, Table II). These charts are now
gral equation method," Proc. IEE, Vol. 123, pp. 702-706, July applicable, with engineering accuracies, to the following "flat" con-
1976. figurations: D.C. bipolar; D.C. monopolar; A.C. three phase, both
41. A. Mohsen and M. Abdel-Salam, "A fast simulation technique for center and outer phases. The parameters are: conductor radius; con-
calculating electric fields," Conference Paper No. A 76 336-8, ductor bundle numbers, 1, 2, 3,4, 6, 8; pole spacing; line height.
New York: IEEE Power Engineering Society, 1976. Mr.M. G. Comber (2) gives the reason for the applicability of these
42. R. M. Morris and P. S. Maruvada, "Conductor surface voltage D.C. charts to the center phase of a three phase A.C.line: "The instan-
gradients on bipolar HVDC transmission lines," IEEE Trans. on taneous gradient on the center phase of a three phase line attains its
Power Apparatus and Systems, Vol. PAS-95, pp. 1934-1945, maximum value when the voltage on that phase is 1.0 per unit and the
Nov./Dec. 1976. voltage on the outer two- phases is -0.5 per unit; then the effect on the
center phase of two bundles at -0.5 per unit voltage and distance S on
Manuscript received March 8, 1979. either side of the center phase can be visualized to be approximately
equivalent to the effect of one bundle at -1.0 per unit voltage and
distance S from the center phase bundle - i.e. the equivalent of a two
A. S. Timascheff (Hydro-Quebec Inst. of Research, Varennes, Quebec, phase (or bipolar) line with the two bundles energized at +1.0 and -1.0
Canada): The Radio Noise Working Group of the IEEE Corona and per unit voltage, respectively". The approximations inherent in the
Field Effects Subcommittee has to be commended for the excellent chart application to A.C. lines are also indicated by Mr. Comber but
idea to prepare a survey of methods for calculating gradients. The effort these produce accuracies within the requirement developed in this
of the Group (efficiently chaired by Dr. Maruvada) resulted in the dis- paper.
cussed exhaustive report - just in time to limit the number of papers Application of the charts to the calculation of outer phase gradi-
dedicated to this topic. ents is similarly explained: The maximum outer phase gradient is ob-
Most engineering problems can be solved only with-a certain degree tained when the voltage on the outer phase is 1 p.u. and the voltages on
of accuracy, which can be called "engineering accuracy" (from this the center phase and the other outer phase are 0.5 p.u. The system acts
point of view, a small possible error has to be considered as acceptable). like two bipoles, one having twice the spacing of the other and each
This "inherent inaccuracy" is due to the imperfect knowledge of the having 0.5 p.u. voltage. The gradient factor of each bipole can be deter-
basic data entering in the calculation (see i.e. the Ref. 13, Oct. 1961). mined readily from the charts and the two gradient factors summed to
On top of this inherent inaccuracy come the possible errors due to obtain the gradient factor on the outer conductor. Note that the spac-
the more or less perfect method of calculation. "Exact" methods are ings between outer phases of three phase lines are beyond the range of
those, which deliver mathematically accurate results corresponding to the published charts (e.g. 90 ft. for case 6) but extrapolation of chart
the set of basic data accepted as exactly true. "Approximate" are those, values can be readily and accurately carried out, using the method of
which add to the "inherent inaccuracy" a certain error due to the im- Figure 7 ref (1), by extrapolating the curve of gradient factors as ob-
perfection of the corresponding calculation method. tained from individual charts for the conductor size in question when
2010

plotted against inverse pole-to-pole separation. This procedure is un- while the other circuit is operating at +450/-500 kV. Such "unequal
necessary in the new charts because wide pole spacings are included. bipole" systems can be analysed as two systems combined, one bipolar
From the charts it is possible to determine the maximum bundle and the other homopolar (e.g. ±475 kV and 1 25 kV). Hence, a set of
(MB) gradients for seven of the ten configurations studied in the work- homopolar charts, similar to the bipolar charts, would enable "unequal
ing group project. Table I illustrates the accuracy with which it is pos- bipole" system gradients to be determined easily within engineering
sible to "read" gradients on the charts, as compared with data from accuracy.
Table II, participant 18.
Figure 1 is an example of the new charts (1979) while Figure 2
shows gradients determined from this chart for A.C. and D.C. configura- References
tions of eight conductor bundles. Similar charts to Figure 1 are available
for 1, 2, 3, 4, and 6 bundles. (1) R. M. Morris, P. Sarma Maruvada, "Conductor Surface Voltage
An additional case which occurs fairly often involves HVDC that Gradients on Bipolar HV dc Transmission Lines". IEEE Trans. on
is not strictly bipolar, i.e. voltages on different poles may not neces- Power Apparatus and Systems, Vol. 95, No. 6, Nov/Dec. 1976, pp.
sarily be of the same magnitude. For example, a three stage bipole may 1934-1945.
be operating with three inverter stages on one pole and only two in- (2) Discussion by M. G. Comber on the above paper (page 1942-1943).
verter stages on the other pole; or, as in the eventual double-circuit
Nelson River system, one circuit may be operating at +500/-450 kV Manuscript received March 6, 1979.

Table I
Comparisons of Gradients Determined from Charts (ref 1)
Against Gradients from Table II (kV/cm)
Configuration 1 2 5 6 (a) 7 9 10
CP OP CP OP CP OP CP OP CP OP D.C. D.C.
From Charts 15.5 14.7 15.7 14.7 16.9 16.1 21 20 16.9 15.6 19.6 21.7
From Table II
Participant 18 15.5 14.68 15.64 14.59 16.9 15.87 21.05 19.74 16.63 15.5 19.4 21.58
(Successive Images)

CP = Center Phase OP = Outer Phase

LINE CONFIGURATION 7: IlOOkV AC 330


MBp 16.9kV/cm MBO; 15.5 kV/cm
oo a0 0 0 0 0 0o
0 0
0 0 0 0 00 O0
o 0 0 0 0
°o °S 0
0

k-50' 70'

I_I,If,IIf
,,, ,,i,f
I~?-

C
0
LINE CONFIGURATION IO: IOOOkV DC BIPOLAR
M B = 21.7 kV/cm
0 0
0 0 0 0
0 0 0 0T
-
0 0 0 0
)cx
a:
4 0~~~~6
0

CONDUCTOR OF CONFIGURATION 10 MONOPOLAR


MB = 16.6 kV/cm
0
0 0
IOOOkV-DC. 0 o0

60'

4.00 3.00 2.00 1.50 1.00


r, RADIUS (cm)

Fig. 1 Fig. 2
2011

V. L. Chartier (Bonneville Power Administration, Vancouver, WA): The why is the "restoration" of the equipotential nature not "final" and
Radio Noise Working Group and especially its Chairman, Dr. Maruvada, why it must be repeated in the next iteration sequence. The "missing"
are to be commended for putting together such an excellent paper. This portion of the explanation is related to the fact that at the same time as
paper verifies what many of us have suspected for a number of years, the strategic placement of a line charge restores the equipotential nature
that there are sufficient "accurate" methods for calculating conductor of the surface on the given conductor, it simultaneously disturbs, albeit
surface gradients; and that the Markt-Mengele method, which is over 40 to a very small extent, the potentials of all other conductors in the sys-
years old, is very accurate for normal symmetrical bundles. tem. This is the reason why this particular line charge must be imaged
There is one aspect of conductor gradient calculations that I believe into all other conductors, so as to reestablish the equipotential nature
is missing in this paper. When I began making conductor gradient calcu- of their surfaces, of the same surfaces that became disturbed by the very
lations in 1964, I was told by my colleagues that the Markt-Mengele introduction of the line charge in question. The imaging process is con-
method lost its accuracy when the ratio of the subconductor spacing to tinued until an arbitrarily assigned accuracy is achieved. Since this ac-
the subconductor diameter, S/dc, became very small. At that time, there curacy may be set to any desired level, there is no inherent limit to the
were no digital computers available that could handle the requirements accuracy achievable by the successive images approach. Reduction of
of the more accurate methods, such as the method of successive images. error will simply require a larger number of successive imagings and a
With the assistance of Dr. Maruvada, I have prepared the following larger computational effort.
Table which compares the conductor gradient calculations using the There are two other points that should be made with regard to the
method of successive images developed by Maruvada and Janischewskyj contents of the paper. The first of these refers to the criterion for accept-
and a method of images developed by Walker of BPA with the Markt- able accuracy required from any method adopted for determination of
Mengele method. The Markt-Mengele method has been programmed into surface voltage gradients. This requirement is governed by the end use
a Texas Instruments' TI-59 Programmable Calculator by R. Menke of of computations, which in our case is the assessment of corona perform-
BPA. The heights of the outside and center phases used for these calcu- ance for alternate transmission line designs. For such a purpose, the
lations were 24.38 and 42.67 meters respectively. The horizontal dis- absolute accuracy may not be as important as the consistency of the
tance from the center phase to the outside phase was 10.97 m. method from one alternate transmission line proposal to another. In
Table I shows that at even very small S/dc ratios, the Markt-Mengele order to assess this aspect of the accuracy, graphs should be rearranged
method is still within 3.8% of the more accurate methods for maximum and the relative error of each method should be plotted for each con-
bundle gradients and within 3.4% for the average maximum bundle figuration used. It is suggested that this be done in the closure of the
gradient. From the standpoint of calculating audible or radio noise, this paper and that obtained results be commented upon.
error is approximately 2 dB which is within the accuracy of radio and The second and final point of this discussion pertains to the influ-
audible noise measurements. At the more normal subconductor spacing ence of the surface voltage gradient upon the physical process of corona
of 40.64 cm, the accuracy is about 3% for maximum bundle gradient in the air volume adjacent to the surface of the conductor, and in turn
and 0.34% for average maximum bundle gradient. This shows that as its interrelation with the pertinent phenomenon under study, i.e. AN,
S/dc becomes smaller, the error increases more rapidly for the average RI, TVI, corona loss, etc. In the first place, the intensity of the corona
maximum bundle gradient than for maximum bundle gradient. process depends not only on the voltage gradient at the surface of the
There is one question that the Working Group should consider. conductor but also on its change with distance from the conductor, i.e.
Can the Markt-Mengele method be used to calculate conductor surface on the volume of air in which ionizing avalanches and other relaxation
gradients for asymmetrical bundles? processes take place. The corona process is govemed by the local strength
of the electric field, and for that reason, a larger volume of air becomes
Manuscript received March 1, 1979. involved when the electric field strength decreases more slowly with
distance from the conductor. While these questions lie clearly outside
W. Janischewskyj (University of Toronto, Toronto, Canada): Of necessity, the scope of the paper discussed, it is important to ascertain that a
in preparing a document of such an extent as the present paper, there separate study be undertaken in which the stated questions of corona
occurs a need for compression of explanations and abbreviation of behavior are investigated in detail. Since the physical corona process
statements with the expectation that references to original papers will may relate differently to each of its engineeringly significant manifesta-
help out. In some cases, however, such an approach may lead to situations tions (AN, RI, TVI, corona loss, etc.), it would be of interest which, if
where legitimate questions of the readership might not be sufficiently any, of the many available definitions of the conductor surface voltage
treated. It seems that the "congested" explanation of the successive gradient given in the Appendix to the paper, may apply in any given
imaging process experiences these difficulties. As explained in the text, instant.
each additional step "restores the equipotential nature of the entire sur-
face of a conductor in the system". The question may then be asked Manuscript received February 26, 1979.

Table I
Calculation of Conductor Surface Gradients - Comparison of Method of Images and Method of
Successive Images with Markt-Mengele Method for BPA Lyons 1200 kV 8-Chukar Conductor
Computed Gradient - kV/cm
Markt-Mengele Images Successive Images
Subconductor Spacing AM MB AM MB AM MB
cm (in.) C.P. O.P. C.P. O.P. C.P. O.P. C.P. O.P. C.P. O.P.
5.08 ( 2) 23.19 24.31 24.00 25.17 24.13 25.28 23.99 25.15 24.13 25.26
10.16 ( 4) 17.11 17.99 17.51 18.43 17.69 18.59 17.53 18.47 17.72 18.58
15.24 ( 6) 15.22 16.04 15.46 16.30 15.68 16.49 15.45 16.29 15.70 16.48
20.32 ( 8) 14.40 15.21 14.55 15.37 14.82 15.60 14.55 15.36 14.84 15.59
25.40 (10) 14.01 14.81 14.11 14.92 14.43 15.20 14.11 14.92 14.45 15.18
30.48 (12) 13.83 14.64 13.90 14.72 14.26 15.04 13.90 14.71 14.29 15.01
35.56 (14) 13.76 14.58 13.82 14.64 14.23 15.00 13.81 14.63 14.26 14.98
40.64 (16) 13.76 14.60 13.80 14.65 14.26 15.05 13.80 14.64 14.30 15.02
45.72 (18) 13.81 14.66 13.85 14.70 14.35 15.14 13.84 14.69 14.39 15.11
50.80 (20) 13.89 14.75 13.92 14.79 14.47 15.27 13.91 14.77 14.52 15.24
55.88 (22) 13.98 14.87 14.02 14.90 14.62 15.42 14.01 14.88 14.66 15.39
60.96 (24) 14.09 14.99 14.12 15.02 14.77 15.59 14.11 15.01 14.83 15.55
2012

H. Parekh (University of Waterloo, Waterloo, Ontario, Canada): The the paper. It has the advantage of reducing computer time in locating
Radio Noise Working Group is commended for writing a neat paper on the point of maximum gradient on the conductor surface.
the methods of calculating transmission line conductor surface voltage
gradients. I would like to make some comments on the charge simula- Method of Determining the Location of
tion method. Maximum Voltage Gradient on the Surface of a
The charge simulation method was first applied by Abou-Seada Cylindrical Conductor
and Nasser [Reference 23 of the paper] for the calculation of voltage
gradients of a two-subconductor bundle. The boundary conditions at The discussion uses the method of conductor images to maintain the
the conductor surface were matched in such a way that the resulting earth as a zero potential plane and the method of successive imaging to
equations became complicated. The most suitable locations of the line maintain the conductor surface at an equipotential.
charges were determined by trial and error. It was extremely difficult to Consider the conductor cylinder shown in Figure 1. The lines drawn
extend the simulation of the charges for a bundle with more than two from the centre of the cylinder terminate at the centre of the other con-
subconductors. ductors in the system. The location of the first order successive image
While working for my Ph.D. thesis [1I] under Dr. Nasser, it was charges are derived from the relation
found that the simulation can be made simple if one locates the line
charges symmetrically on a circle inside the conductor and selects equal
number of boundary points (also located symmetrically) on the con-
ductor surface. It was not necessary to take the first, second and the
fourth derivatives of the voltage on the conductor surface as was done
by Abou-Seada and Nasser. This made computer programming simple,
eliminated trial and error procedure, and the results were highly accurate
(error in voltage gradients less than 0.2%). This method was applied for
ac three-phase transmission lines with upto eight subconductors in a
bundle and unipolar and bipolar lines with upto twelve subconductors
in a bundle [21.
This method was also applied for stranded conductor lines [3]. In
this case, the method suffered from the drawback that considerable
error existed in the computed values of the voltage gradients in regions
where strands touched each other (discontinuity points). However, the
error in the maximum voltage gradient on the conductor surface or the
strand factor was not greater than 4% [4,5].
References
[1] H. Parekh, Computation of Electric Fields for EHV and UHV
Transmission Lines, Ph.D. Thesis, Iowa State University, Ames,
Iowa, 1974.
[2] H. Parekh, A User Oriented Computer Program to Calculate Electro-
static Fields and Capacitances for Extra High Voltage Transmission
Lines, Engineering Research Institute, Report ISU-ERI-AMES-
75190, Iowa State University, August 1975.
[3] H. Parekh and E. Nasser, "Computation of the Potential and field
of a stranded twin-bundle conductor above ground", IEEE PES
Winter Meeting, Paper No. C 74084-0, New York, January-February
1974. Fig. 1
[4] H. Parekh, M. S. Selim and E. Nasser, "Computation of electrical
field and potential around stranded conductor by analytical (1)
method and comparison with charge simulation method", Proc. ,-6 r
IEE, Vol. 122, No. 5, pp. 547-550, May 1975. 1)
[5] H. Parekh and E. Nasser, "The effect of geometric parameters on
electric field and capacitance of stranded conductors above where: r = radius of the cylinder
ground", IEEE PES Summer Meeting, Paper No. A 75 540-5, San D = distance between the cylinders
Francisco, July 1975. 6 = Distance along the line joining the centre of the
cylinders'.
Manuscript received February 6, 1979.
We choose an arbitrary point 'A' on the surface of the cylinder, making
J. S. Goodman (Manitoba Hydro, Winnipeg, Manitoba, Canada): This an angle 0 with the X axis and determine the voltage gradient of this
point due the charges +Q and -Q.
paper is an excellent presentation of the various existing methods used
in the calculation of conductor surface voltage gradients. Its the initial 1-. - -Q1t (a vector) (2)
step in providing some order in defining and calculating conductor gradi- (;1
-

(2IsI QltI
ents. There seems to be no reason why a standard technique cannot be
chosen from among the various existing method of calculating this very S t
1 1
important parameter.
With regard to the accuracy of the calculations, no reference has The tangential components of (2) must equal zero in order to maintain
been made to the type of computers and their inherent accuracy used the surface at an equipotential. The normal component or gradient is
by the participants in the survey. Some methods require the inversion of
Maxwell's coefficient matrix to arrive at the Q vector i.e. Cl = Q 1Cos -Q sille1 (3)
Q - CV wliere C -
s1 1t
In the case involving a large number of conductors, double precision is This can be expressed in terms of the angle as
required in the inversion to insure reasonable accuracy. Would the
authors comment on this aspect? C, - (Q (l) -r ) (4)
In determining RI and AN, the maximum voltage gradient on the
conductors is required. The following method for determining the loca- 1-)2 ( I+r -2_)r Cosry)
tion of the maximum gradient is suggested. The resulting accuracy seems 2 1)
adequate when compared with the acceptable error of ± 2 % suggested in
2013

The total gradient Gt due to all image changes within the jth cylinder is - 1

o1 S 1.1nI
L)
t Q1 ()1 -r )
2 9 Tanrg = 1
r P2 (1+ - 2r Cosy ) (5) (10)
I1rn2 I
I
D2I i
O
'1 (Co.s
Di
n = number of Image charges
rj = radius of jth cylinder The difference in results using (9) or (10) is about 0.002% and is insig-
We maximize the gradient by differentiating with respect to 0 i.e. nificant.
For lack of better terminology I have called this the Tangent Sum
Method.
n In order to check the validity of the 'Tangent Sum' method in
d
t \ Q i I (I) -r2) dy /(10
SinY itI
n0
determining the location of maximum conductor gradient, the follow-
ing two conductor configurations were investigated:
\ 3 2
dO DP ( +r2 -2r Cosy ) 1. Figure 2 shows a 3-phase delta system. Because of symmetry,
the maximum gradient vector on the centre phase is directed vertically
2D D
1
1
i 1~~~~~~I
As a first approximation, the denominator in (6) is nearly equal to unity,
hence
6. O[ -
1n
Q1 (D2-r2 )Stinyd y1l/(IlO Conductor 3

3) 22. 32m
.1 = I 12. I9mn

13.94 in
*y =0-( (I Ly 1
L. I
(10 (GROUND PLAN1E
Expanding (7)
n
< Q (D~2 -ri)2)SS inOCosf31
1(D1
D3
1 i
1=1
(8) 0 0
2 2

/l
(PD
3
)C osgSin~,l1
I - I
Din 0
or

Conductor Angle of Maximum Surface*


Q1 (I)7-r) ) SI1n Number Gradient in Degrees
1 24.33 1
2 269.997
3 152.565
T;anG l 11
1 I
(9) *Angle in degrees between a radial line to the point of maximum
gradient and the positive X-axis of the conductor.
Q ()2-r 2)Cos3
Fig. 2
1

downward making an angle of 2700 with a positive X axis. The proposed


Equation (9) can be further simplified if we assume method yields an angle of 269.9970.
2. Figure 3 shows an 8 conductor bundle bi-polar system which is
121 yielding the same as line configuration 10 in the paper. Table 3 compares the
L23 results of the maximum gradient angle as predicted from equation (9)
i and the maximum value as determined by evaluating equation (5) at
2014

2 0

0
(Condlctor 1 0 0 0

80 +ve Pole 04 0 -ve Pole 0

70 0
0
5
0
0
0

CIOUNI) PLANEI /
M_<~~~~~~~~~~~~~~~~~~~77 7/77r-

0 0
mrnage Conductors
0 0 0 0

0 0 0 0

0 0 0 0
0 0

Fig. 3

Table 3 sons. The accuracy of only the final results, rounded to the second
decimal, are compared in the paper. Mr. Goodman also proposes a
Conductor Angle of Maximum Surface Gradient Relative simplified method for locating the point on a subconductor surface
Number Predicted by Calculated by Maximum where the maximum gradient occurs. The "Tangent Sum" method
Equation (9) Equation (5) % Error* Gradient proposed by Mr. Goodman is a simple and elegant method for obtaining
the position of the maximum gradient on the subconductor and should
132.55 132.63 0.02 1.266 find application even in computer calculations.
2 86.08 86.21 0.04 1.273 Additional details of the method of successive images, provided by
3 41.98 42.03 0.01 1.280 Professor Janischewskyj, are very much appreciated. It has not been
4 359.67 359.70 0.01 1.284 possible, however, to rearrange the graphs as suggested by the discusser.
5 317.50 317.44 0.02 1.282 Although it is somewhat time consuming, the graphs in their present
6 273.87 273.71 0.04 1.275 form can be used to determine the relative error of the different methods.
7 227.98 227.88 0.03 1.268 We agree with Prof. Janischewskyj that more work is needed to deter-
8 180.41 180.41 0.00 1.264 mine which gradient value is best suited for calculations of AN, RI, etc.
Morris, Morse and Maruvada present some useful information on
* % Error based on the difference in angle as a percent of the 360 degrees. the use of charts for obtaining quickly the maximum bundle gradients
of different ac and dc transmission line configurations. Dr. Parekh
provides some additional details of the charge simulation method.
Professor Sebo provides a detailed discussion of the various factors which
0.01 degree intervals around the periphery of the conductor. The maxi- influence the conductor surface voltage gradients. It should be noted
mum gradient occurs on conductor 4 as expected. that the overall accuracy of gradient calculations depends on the ac-
curacy with which these factors are known. The number of references
Manuscript received February 6, 1979. given in the paper had to be restricted since it has already exceeded the
normal length of a paper. We therefore appreciate the additional bibliog-
P. S. Maruvada (Chairman, RNWG): I would like to thank, on behalf of raphy provided by Prof. Sebo. I would just like to add one more refer-
the Radio Noise Working Group, all the discussers for their excellent ence1 to this list.
contributions which greatly enhance the value of the paper. We appreciate the supportive comments of Dr. Timascheff on the
Mr. Chartier presents some interesting results of gradient calcula- question of acceptable accuracies. As mentioned by Dr. Timascheff,
tions which show that the Markt-Mengele method gives reasonably and also by Mr. Chartier, the accuracy of a method depends also on the
accurate results even for low values of the subconductor spacing to actual gradient parameter calculated. Thus most of the methods, includ-
subconductor diameter ratio. In response to Mr. Chartier's question as ing the Markt-Mengele method, give more accurate results for the average
to whether the Markt-Mengele method can be used for asymmetrical maximum gradient, which may also be the more relevant parameter for
bundles, it appears from the basic assumptions involved in Markt-Mengele RI, AN, etc., calculations, than for the maximum bundle gradient.
method that this may not be possible. One of the basic assumptions in-
volved in this method is that the total charge of the bundle is distributed Reference
equally among the subconductors, which is not true in the case of asym-
metrical bundles. [11 E. T. B. Gross, L. R. Stensland, "Characteristics of twin conductor
Mr. Goodman raises the question of the accuracy of numerical arrangements", AIEE Transactions on Power Apparatus and Sys-
calculations performed on the different computers. Although we asked tems, Oct. 1958, pp. 1-5.
some specific information related to the computer used by each partici-
pant, it was found almost impossible to make any meaningful compari- Manuscript receive& May 7, 1979.

You might also like