You are on page 1of 14

718

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, VOL. PAS-88, NO. 5, MAY 1969

Analysis

of

Transmission Lines: 1-Unipolar Lines


DC
MEMBER, IEEE, AND

Corona Losses on

MARUVADA P. SARMA,

WASYL JANISCHEWSKYJ, MEMBER, IEEE

Abstract-Theoretical calculation of corona losses for practical unipolar dc transmission line configurations presents considerable difficulty because of the nonlinear nature of the equations describing the space-charge fields. The application of numerical methods to obtain solutions of practical interest is discussed. One of the difficulties in the analysis of space-charge fields is the determination of the actual charge distribution around the conductor in the corona. A numerical iterative method of computing this charge distribution is presented. The method is applicable to any general configuration for which the space-charge-free field can be calculated. The lineto-plane configuration is considered. A method of including the effect of conductor surface irregularities in the theoretical calculation of corona losses is outlined, and it is suggested that, by the same method, Popkov's formula may also be modified to make it applicable to lines with practical transmission-line conductors. Calculations by the method of analysis developed as well as by the modified Popkov equation are compared with experimental results.

The two features mentioned in the preceding paragraph enhance the necessity for theoretical analysis of corona losses on practical dc transmission lines. In the paper the case of unipolar dc corona is discussed. Existing methods for analysis of corona losses in certain conductor configurations are reviewed, and the underlying assumptions are discussed. A numerical iterative method for solving the equations of unipolar corona in the general case is presented. The method is applied to calculate corona losses in a line-to-plane geometry.
EQUATIONS OF UNIPOLAR CORONA Unipolar corona is characterized by the flow of ions of one polarity in the interelectrode space, the ion polarity being the same as the polarity of the conductor in corona. The electric field in the interelectrode region is governed by the space charge of the ions, while the ionic flow itself is a function of the electric field. The equations describing the electric field and the ionic flow are

INTRODUCTION

various problems associated with it are being studied extensively. One of these problems is the corona occurring on the transmission lines and the radio interference and power loss associated with it. Corona is a self-sustained partial breakdown of air in the nonuniform field around the conductor. At alternating voltages, the space charge created by corona is constrained to the vicinity of the conductor because of the periodic reversal of the electric field. Thus ac corona losses depend more on the conductor surface voltage gradient than on the actual line configuration, and it may be possible to predict with reasonable accuracy the corona losses on a proposed line design from experimental results obtained in a different configuration. On the other hand, the space charge created by dc corona fills the entire interelectrode region, resulting in a greater dependence of corona losses on the actual line geometry. Another important factor which affects corona losses is weather. Tests on ac lines show that corona losses become economically significant only during foul weather conditions. In contrast, because of smaller difference between fair and foul weather performance, fair weather corona losses are a more important consideration in the design of highvoltage dc lines.
Paper 68 TP 75-PWR, recommended and approved by the Transmission and Distribution Committee of the IEEE Power Group for presentation at the IEEE Winter Power Meeting, New York, N. Y., January 28-February 2, 1968. Manuscript submitted April 10, 1967; made available for printing December 11, 1967. The authors are with the University of Toronto, Toronto, Ont., Canada.

HIGH-VOLTAGE dc possesses many advantages over conventional ac for long-distance and bulk-power transmission. Recent advances in the development of mercury-arc and solidstate power conversion equipment have increased the technical and economic feasibility of dc transmission. Consequently, the

V.E= p

CO

(1)

j = kpE

(2)

(3) The first is the Poisson's equation for the electric field; the second is the equation for current density; and the third is the equation of current continuity. Thermal diffusion of Tons is neglected. The electric field is given in terms of the potential as E = -VD. (4) By combining (2) and (3) and substituting for p from (1), the following equation is obtained for the electric field: E.V (V-E) + (V.E)2 = 0 (5) or, in terms of the potential,
V. (VPvq) = 0.

Vj = 0.

(6)

Equations (5) and (6) are the general equations describing a unipolar ionized field. They are nonlinear partial differential equations, and there are no known methods available for solving them in a general case. An analytical solution may be obtained only for simple cases like concentric spheres or coaxial cylinders.

Solution for the Cylindrical Configuration Townsend [1] was the first to solve exactly for the ionized field in a coaxial cylindrical configuration, as shown in Fig. 1. For voltages above corona onset, ionization occurs in a narrow region close to the conductor (shown by the dashed circle) creating a unipolar ionized field in the interelectrode space. In the corona layer, charged particles of both polarities exist.

SARMA AND JANISCHEWSKYJ: CORONA LOSSES ON DC TRANSMISSION LINES

719

Fig. 1. Coaxial cylindrical configuration.

Analysis of Practical Configurations Apart from the simple configurations discussed, it is very difficult to obtain analytical solutions without making simplifying assumptions. Deutsch [3] made an approximate analysis of some of the more complicated cases such as the line-to-plane geometry. Following Deutsch's work, Popkov [4] in his analysis of the cylindrical and the line-to-plane geometry introduced certain improvements. His analytical solution for the cylindrical configuration is identical to (9). Popkov assumed that the line-to-plane configuration can be approximated by an equivalent cylindrical system and obtained the following equation for its voltage-current characteristic:

For the electric field in the unipolar region (5) simplifies to


d E-+ --+ dri dr2 r dr

0.41P
(7) with

In

vo

ro

(1 + Y)"12 - 1 + ln

+ ~~~1 (I1+ y)"2


(11)

d&E

3EdE

dE 2

O.

It can be verified by direct substitution that the following expression for E

2irEokEoro)

(12)

A2r2 2 E2 +

BJ2

where A and B are arbitrary constants, satisfies (7). The constants A and B are determined from appropriate boundary conditions. The first boundary condition required is the potential difference between the two electrodes. The second boundary condition used is the electric field at the boundary between the corona layer and the space-charge region. Following an approximate analysis by Kapzow [2], it is assumed that the electric field at the conductor surface remains constant at the onset value as the conductor voltage is increased and that the net space charge in the corona layer is zero. A rigorous theoretical analysis of the electrical field distribution in the ionization layer of a dc corona discharge under equivalent steady-state conditions has been carried out by the authors for both polarities of the conductor. The analysis, which is intended for a future publication, resulted in several conclusions pertinent to the present discussion. The most significant of these shows that indeed the change in the field intensity at the surface of the conductor from the onset value is negligible for the range of corona currents encountered on practical transmission lines. The analysis also indicates that the accuracy of corona loss calculations is not significantly affected if the unipolar space charge is assumed to start at the conductor surface itself rather than at the corona layer boundary shown in Fig. 1. The analysis shows, however, that for large corona currents, the field at the conductor surface decreases for positive polarity and increases for negative polarity. If desired, this change in the field at the conductor surface may be included into the accurate calculation of corona losses presented. Based on the above boundary conditions, the voltage-current relation for the cylindrical configuration is obtained in the following nondimensional form:
V-Vo

nonuniform current distribution plane electrode. ~~~~~~~~(8) and Morse [5] have along thediscussed the applicaSimpson recently

The empirical constant P is introduced to take into account the

bility of Popkov formulas to practical systems. Good agreement was observed between (9) and the experiment for corona on smooth conductors in a cylindrical configuration. Accurate measurements were also made of corona currents on an outdoor test line using stranded conductors, and the results were compared with Popkov's formula (11). Reasonable agreement [6] between theory and experiment could be obtained only when P was increased from 1.65 (originally suggested by Popkov) to about 5. Results also indicated that P, contrary to the assumption of Popkov, will vary with line parameters as well as with the intensity of corona.

NATURE OF THE GENERAL PROBLEM The analysis of unipolar dc corona consists basically of the analysis of the dc ionized electric field. A detailed investigation of the mathematical aspects of dc ionized fields has been recently presented by Felici [7]. With the availability of modern highspeed digital computers, the possibility of obtaining accurate numerical solutions for complex problems such as that of dc ionized fields has become a reality. The basis for the discussion of the general problem of dc ionized fields is formed by (1)- (3). Combining again equations (2) and (3) and substituting now for V-E from (1), the following equation relating the field and charge density distribution is obtained:
Co

-E.Vp.

(13)

ln-=

(1 +

Y)"J2-1 + ln
2FeM
I

+ (1+Y)2

(9)

E. Vp can be replaced by Edp/ds, where s denotes the distance measured along the corresponding flux line, so that dp/ds represents the charge density gradient along the flux line. Now (13) can be written as dp _ 1 ds
p2
eo E

(14)

with
yz=
=

Or

(10)

Integration of (14) yields 1 1 rds - = - I-+ constant. P coJ E

720

IEEE

TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, MAY

1969

The constant of integration depends on whether integration is started from the emitter (conductor in corona) or the collector (ground electrode). The analysis of Felici [7] has shown that the choice of the collector as the starting point in the integration is improper, as it may lead to some mathematical difficulties. Hence starting at the emitter the expression becomes 1 1 1 s ds
p
Pe

4 is a scalar point function of the space coordinates depending on charge distribution. From (2), (3), and (16),

E'*V(4p) = 0.
that
and

(17)

According to (17), 4p is a constant along any given flux line of the space-charge-free field. Substituting (16) in (1) and noting

EOE

(15)

V*E' = 0

Equation (15) together with (1) can be solved iteratively to obtain the charge and field distribution. First, the space-chargefree field is assumed, and a first approximation to the charge distribution is obtained from (15). This, of course, requires the value of Pe to be known. Substituting the newly calculated charge distribution in (1), a second approximation to the field distribution is obtained, and so on. However, there are two limitations to this iterative method of solution. First, it is very difficult and time consuming to solve Poisson's equation except in a one-dimensional problem. In order to reduce the complexity of the numerical solution, it is therefore necessary to reduce the given problem to an equivalent onedimensional problem. This can be done by assuming that the space charge affects only the magnitude but not the direction of the electric field. It is believed that such an assumption does not contribute to any serious error for normal charge densities. The second limitation is in the choice of Pe. Only the two limiting cases are discussed by Felici [7] in which p. is assumed to be equal to either very large or very small value. It can be seen from (15) that when Pe is very large, the charge distribution is practically independent of Pe, while for very small Pe the charge density is very nearly constant throughout the interelectrode space. In the former case saturated solutions are obtained, while the latter corresponds to the so-called weak current approximation. An inevitable consequence of such approximations is the assumption of a uniform charge distribution at the surface of the emitter. In spite of these limitations, however, such assumptions have been shown to give some useful approximate solutions to many practical problems. A different approach is proposed by the authors.

E'*V4= El
E'-=
ds

co

P.

(18)

corona,

Integrating (18) along any flux line starting at the conductor in


eOC- o E Eds
(19)

Substituting E' = - d4/ds and changing the limits of integration accordingly, eO [V p4

C=4~+1O14

(E' )2

(20)

where V is the conductor potential and 4 is the space-charge-free potential at the point where 4 is being evaluated. At the conductor surface, according to assumption 5),

(21) 4(0 = Vo/v with Vo being the corona onset voltage. Integrating both sides of (20) with respect to 4 between the limits 0 and V, using ij as the dummy integration variable,
d
=:

(e, dO +

From (16)

-0 I
ds

dO)

(22)

__d(
o

METHOD OF ANALYSIS In a practical problem, both assumptions regarding Pe may not be true. A method of evaluating Pc, and hence the charge and electric field distribution in the interelectrode region as well as the current distribution at the electrodes, is presented.
Assumptions 1) The entire interelectrode region is filled with unipolar space charge. 2) The mobility of ions is constant (independent of field intensity). 3) Thermal diffusion of ions is neglected. 4) The space charge affects only the magnitude but not the direction of the electric field. 5) The electric field at the surface of the conductor in corona remains constant at the onset value.

ds
1

d d4)X = _t

Substituting this and (21) in (22) and simplifying,


v

Jv JEv Jo dt7

)2.

(23)

At the same time (15) can be rewritten in terms of 4, E', and 4) as 1 1 J1 V 4


P
Pe c 0

(E' )2

(24)

For given values of V, Vo, and the distribution of E', (20), (23), and (24) yield the distributions of p and E along any flux line of the space-charge-free field. The resultant current distribution can be calculated subsequently from (2).
Iterative Solution Fig. 2 shows qualitatively the variation of charge density along any given flux line, obtained from (24), as a function of 4) for different values of p.. Very large p. corresponds to the saturated solution, and the curve for very small values of p. corresponds to the weak current approximation. The curves in between show the distribution of p for intermediate values of

Equations for Charge and Field Distributions If E is the electric field in the presence of space charge and E' the space-charge-free field, according to assumption 4),

E= (E'.

(16)

SARMA

AND JANISCHEWSKYJ: CORONA LOSSES ON DC TRANSMISSION LINES

721

ct
z

v.

Fig. 2. Charge density distribution along a flux line for different Pe,

value of Pe. This is a more difficult problem than the inverse problem discussed above since it is impossible to obtain a solution in closed form as in the previous case. It is only possible to arrive at a solution using a suitable iterative technique. The well-known secant method [8] of iteration is found to be particularly suited to the problem since the method does not involve the evaluation of any derivatives, while at the same time it is considerably faster than any linear iteration method. The iteration procedure is illustrated in Fig. 3. In the first step Pm is computed for the flux line under consideration using the given data in (26). Then two values of Pel and Pe2 are arbitrarily chosen such that Pe2 > Pel > Pm. For each of these values, the charge density distribution along the flux line and the corresponding mean charge densities Pmi, pm are evaluated using (25) and (27). Subsequent values Pei are calculated using the following iteration function:
Pei = Pei-1 +

(Pmp-mi 1) n
(Pmi-1
-

Pmi-2)

(pei-i

Pei-2), i

3,4,

z
0 w

fm a

4S

(29) The corresponding value of Pmi is then calculated before proceeding to the next iteration. The procedure is continued until Pm, approaches Pm to the required precision e. The speed of convergence of iteration depends on the choice of Pei and Pe2. If these two initial values are chosen such that
P4 P PO3 a EMITTER CHARGE DENSITY

Pei = fiPm
Pe.

Fig. 3. Iterative procedure for determining


Pe. For any

(30)

Pe2 = f2PN
pm

given value of Pe, a mean charge density defined such that,


PM [=

can

be

(E')V

(EJ)2Jdd
Pe.

(25)

From (24) it is obvious that at any point p < pm defined in (25) is always smaller than Pe. Using (23), Pm may be written as
Pm= eo(V- Vo)
d

Consequently,

then the factors fi and f2 must be chosen to achieve the fastest convergence. It has been observed that optimum values of fi and f2 vary for each flux line as well as with the value of p.. The choice of fi= 1.5 and f2= 3.0 resulted in a reasonably fast convergence rate on the average. In order to achieve an agreement of closer than 0.1 percent between pm, and pm, an average number of four iterations were necessary for each calculation.
APPLICATION TO PRACTICAL CONFIGURATIONS The method of analysis proposed is applicable to any general unipolar conductor configuration for which it is possible to determine the space-charge-free field either analytically or numerically. Because of the large amount of numerical computation involved, the calculations have to be done on a high-speed digital computer. The method of calculation has been applied to the well-known line-to-plane geometry shown in Fig. 4 with either smooth conductors or conductors having surface irregularities such as stranding, etc. Details of the calculation procedure are given in the Appendix. Beside the line-to-plane configuration discussed, the method can also be applied to lines with bundle conductors, homopolar lines, and lines with overhead ground wires. A general numerical method of calculating the space-charge-free field for such practical configurations has been developed, and the results of corona loss calculations for characteristic examples will be published at a later date. Lines with Smooth Conductors The first case considered is a hypothetical line comprised of a smooth conductor of 1-cm radius placed at different heights above the ground plane. For purposes of comparison, corona losses for this line have been calculated by the proposed method

do (26) (El(E)2 and pm may be calculated directly from the given data. Note that according to (17) along any flux line of the space-charge-free field, pt = Pete. Substituting this condition in (14) and (18), the intergration after change of the variable to O yields the following modified equations for the distributions of p and t:
1
1

2
EOPe,e

d
(27)

p2

Pe2

J, (E' )2
d4CP (E'

2+2pe4e = 02 2(2+ EPctJO

28 (28)

Pm

It can be seen from (25) and (27) that the mean charge density for any specific flux line is a function of the emitter charge density Pe. If Pe is known, (27) gives the distribution of charge density along the flux line, and the corresponding mean charge density may be calculated directly using (25). The variation of pm as a function of Pe is shown as the solid line in Fig. 3. In the present analysis, however, pm is known from the given data through (26), and it is required to obtain the corresponding

722

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, MAY 1969


a

FLUX

~1.
z
w

:3

Fig. 4. Line-to-plane configuration.


600

4
0 0
0

-I

ir

500

z
w

I~-

400

H=

100

200

250

300

350

400

450

500

APPLIED VOLTAGE

KV

xc 300
D

< 200
0

Fig.W8' Corona losses for experimental line with a Lark conductor [6]; ro-1.02 cm, H-9.217 meters. Positive polarity. Vertical lines show range of experimental values. A-Calculated by method presented; B-calculated by modified Popkov formula.
0o

250

300

350

400

450

500

55 $0

APPLIED VOLTAGE

KV
32 28
1. A

Fig. 5. Corona losses for a hypothetical line with a smooth conductor; re-1 cm, H-1 meter. A-Calculated by method presented; B-calculated by Popkov's formula.
41

,E
I-z

1.0

W-2

0'

It
0.6
04
z

o02

__X__ __
-.2000

Z
0
0

I:

_-10

150 650

200

250

300

350

400

450

500

300

350

400

450

500

550

600

APPLIED VOLTAGE

KV

APPLIED VOLTAGE

KV

Fig. 6. Corona losses for a hypothetical line with a smooth conductor; ra-L cm, H-20 meters. A-Calculated by method presented; B-calculated by Popkov's formula.

Fig. 9. Corona losses for experimental line with a Lark conductor [6]; ro-1.02 cm, H-5.36 meters. Positive polarity. Vertical lines show range of experimental values. A-Calculated by method presented; B-calculated by modified Popkov formula.

E 24

20
16

20

I-,
ILl

(:
z
0

12
8
4

0 o

_ 2g _a msA7-~~~~~10 -_**
36 38
40

z 0

42

44

46

250

300

350

400

APPLIED VOLTAGE

KV

APPLIED VOLTAGE

450 KV

Fig. 7. Corona losses for a model line [9] with a smooth conductor; ro-0.04 inch, H-10 inches. Vertical lines show range of experimental values. A-Calculated by method presented; B-calculated by Popkov's method.

Fig. 10. Corona losses for experimental line with a Lark conductor [6]; ro-1.02 cm, H-9.35 meters. Negative polarity. Vertical lines show the range of experimental values. A-Calculated by method presented; B-calculated by modified Popkov formula.

SAYIMA

AND JANISCHEWSKYJ: CORONA LOSSES ON DC TRANSMISSION LINES

723

as well as by Popkov's formula. The results for two extreme values of the conductor height, 1 and 20 meters, respectively, are shown in Figs. 5 and 6. As already stated, in the derivation of Popkov's formula the actual line-to-plane configuration is approximated by an equivalent coaxial cylindrical system, and an empirical constant P is introduced to take into account the difference between the current distribution along the plane in the actual configuration and the outer cylinder in the equivalent system. With a decrease in the H/ro ratio, the current distribution around the conductor becomes more nonuniform, which in turn changes the distribution of current on the plane, thus requiring a new value of P. In addition, the nonlinear nature of the equations suggests that the current distribution on the plane changes also with the intensity of corona. This shortcoming of Popkov's method is illustrated in Figs. 5 and 6. The deviation of the results calculated by Popkov's formula increases with a decrease in H/ro as well as with an increase in the intensity of corona. Corona currents measured on a laboratory model [9] consisting of a smooth conductor, 0.08 inch in diameter, placed 10 inches above the ground plane, are compared with theoretfcal results in Fig. 7. It can be seen from these results that there is negligible difference between the two methods of calculation and also that the experimental results agree quite well with theoretical calculations. The reason for the small difference observed is the fact that the range of voltages above corona onset for which the calculations are made is very small.

Lines with Stranded Conductors Simpson [6] compared the experimental results of corona currents, for a Lark conductor (1.02-cm radius) and for different heights above the ground plane, obtained on an outdoor experimental line, with prediction of Popkov's formula. He suggested that a reasonably good agreement may be obtained between the Popkov formula and experiment by increasing the value of the empirical constant P by a factor of about three. The effect of conductor stranding has been considered to correct computed corona currents only near the region of onset. Calculations for the lines with the smooth conductors discussed in the previous section indicate that, although the empirical constant P may vary with the line parameters as well as with corona intensity, the extent of this variation is not very large. Consequently, it seems reasonable to assume that the effect of conductor surface irregularities, rather than the value of P, is responsible for the observed differences between the experimental results and Popkov's formula. Practical transmission-line conductors are normally of stranded construction and often contain surface irregularities such as strand lift (birdcaging), nicks, scratches, grease, dirt, etc. Irregularities distort the field in the immediate vicinity of the conductor and reduce the corona onset voltage. It should be noted that the distortion is confined only to a very narrow region around the conductor and that the field in the remaining interelectrode space is not noticeably affected. In calculation of corona losses, conductor surface irregularities may be taken into account by replacing for the actual conductor an equivalent smooth conductor of the same diameter with a reduced corona onset voltage. The method of ncluding the effect of surface irregularities in the calculation of corona losses is discussed in the Appendix. The reduction in the corona onset voltages due to surface irregularities depends on the increase in the conductor surface gradient as well as on the actual field distribution near the con-

ductor. Although the field distortion caused by any one factor alone, such as stranding, may be calculated theoretically, it is difficult to relate the distortion to the reduction in the onset voltage. In practice, the influence of the various surface irregularities cannot be separated, and it is only possible to obtain an overall reduction factor by testing actual conductor samples. Stone [10] investigated the reduction in ac corona onset voltage due to stranding by testing aluminum tubes wrapped with a single layer of strands. The stranding factors obtained by Stone ranged from 0.88 to 0.98. Experiments conducted by the authors [11] on an actual sample of a new stranded conductor (0.502-inch diameter) in a cylindrical cage under both ac and dc excitation gave values for the stranding factor of about 0.8. It appears reasonable to expect still lower values of the reduction factor for conductors on practical transmission lines because of the combined effect of all existing surface irregularities. The experimental results for the Lark conductor [6] are compared with theoretical calculations using Popkov's formula as well as the method presented. Figs. 8 and 9 show the results for positive polarity and for conductor heights of 9.216 and 5.36 meters, respectively, while Fig. 10 shows the results for negative polarity and conductor height of 9.35 meters. The experimental data for these cases in(licate the value of the overall reduction factor for the corona onset voltage to be about 0.7. The theoretical calculations are made with this value for the reduction factor. The mobility of both positive and negative ions is taken as 1.5 X 10-4 m2/V.s. The agreement between the theoretical and experimental results is good for the larger conductor heights for both positive and negative corona. For the smaller height, the experimental results are larger than those predicted by the theory at higher voltages. The large differences observed may be attributed to the fact that at smaller conductor heights, the nonuniformities in corona generation on the conductor due to the proximity of the ground plane, conductor sag, etc., become increasingly significant. It should also be noted that the suggested modification to Popkov's formula results in a reasonably good agreement with experimental values without changing the empirical constant P. However, the calculations based on the method of analysis developed in the paper are closer to the experimental results. The modified Popkov equation again shows larger deviation for lower values of conductor height and for higher corona intensity, indicating the slightly variable nature of P. CONCLUSIONS On the basis of the theoretical analysis presented, the following conclusions may be drawn about the dc corona losses on unipolar conductor configurations. 1) The numerical iterative method of solving the equations of unipolar dc corona derived in the paper is a useful technique for an accurate determination of the field and charge distribution in the interelectrode space. Based on these, current distribution at the electrodes and the associated power loss may be established. 2) The method of analysis is applicable to any general unipolar conductor configuration for which the space-charge-free field can be calculated. 3) Corona losses calculated for a model line with a smooth conductor by the method proposed show a very good agreement with experimental results. 4) Popkov's formula for the line-to-plane geometry gives reasonably accurate results for smooth conductors; however, the

724

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, MAY 1969

empirical constant P is found to be a function of the line parameters and of the intensity of corona. 5) For practical lines using stranded conductors, the proposed method, modified to include the effect of conductor surface irregularities, gives good agreement with experimental results for practical line heights. For lower heights, the agreement is good only for a comparatively small range of voltages above corona onset. 6) The suggested method of accounting for the conductor surface irregularities has also been incorporated into Popkov's equation and was found to give reasonable agreement with experimental results. Nevertheless, results calculated using the modified Popkov equation are slightly lower than those using the proposed method, the deviation being larger for lower conductor heights and higher corona intensity as in the case of smooth conductors. APPENDIX Application of the method of calculation proposed in he paper is demonstrated by the calculation of corona currents from a conductor in a line-to-plane configuration shown in Fig. 4.

Space-Charge-Free Field
For this configuration, the space-charge-free electric field may be calculated analytically by any of the well-known techniques. Using the method of conformal transformations [4], the complex potential function is given by
V

Fig. 11. Flow chart for calculation of corona losses of line-to-plane configuration.

H+ z

ln (2H/ro) Hl-n

with w = q5 + i+, z = x + iy. The function transforms the actual field of the conductor-plane geometry in the z plane to the known field between a pair of infinite parallel plates in the w plane. Constant values of 4 and 4l' represent equipotential lines and flux lines, respectively, of the original field. The potential gradient at any point is given by
|d dz
=

For the calculation of current distribution around the conductor at each voltage, points along the conductor periphery corresponding to equal intervals of AKi/ from 0 to 7r/2 are chosen. In order to reduce the computer storage requirements, current densities from each flux line are calculated for the entire voltage range and stored before proceeding to the next flux line. Finally, the current distributions corresponding to each voltage are integrated to obtain the V-I characteristic. The following abbreviated notation has been used in the flow chart: G

(E')2 d)
dq

(33)

H In

(2H/ro)

[cosh 2KO + cos 2KVt]

Jv

with
2K= -ln-. V ro

(34)

(32)

Equation (32) gives, for constant values of 2K4I, the variation of E' with 4) along a specific flux line. 4) varies from zero on the ground plane to V on the conductor. K*&, defined in Fig. 4, varies from zero to wr/2 for the different flux lines.
Flow Chart for the Computer Program

p (4)) denotes the charge density as a function of 4) as obtained from (27). For purposes of calculating the functions I(+5) and p (4)), as well as for the various numerical integrations involved, the range 0-V of 4 is divided into a number of equal intervals, the actual number of divisions depending on the accuracy required. The calculations have been made on an IBM 7094 computer.

Effect of Conductor Surface Irregularities


The method of including the effect of conductor surface irregularities in the calculation of corona losses is discussed in this section. The first step is the calculation of the corona onset voltage, which is given by

ir/2.

The flow chart, based on which the computer program for the calculations is written, is shown in Fig. 11. Corona currents are calculated for voltages starting at Vs (> Vo) up to a maximum voltage VM at fixed increments of AV. For convenience, the calculations are made in the 4y4 coordinates instead of the x-y coordinates. Because of the symmetry about a vertical axis through the center of the conductor, it is sufficient if the current distribution is calculated only for the range 0 < K# <

Vo = mroEo ln (2H/ro)

(35)

where m is the factor by which the onset voltage is reduced due to surface irregularities. The onset gradient Eo at the conductor surface is assumed to be the same as that for a smooth conductor

SARMA AND JANISCHEWSKYJ: CORONA LOSSES ON DC TRANSMISSION LINES

725
ACKNOWLEDGMENT

of the same diameter. If the actual conductor is replaced by an equivalent smooth conductor, the gradient at its surface corresponding to the voltage V0 would be mEo. For evaluating the integrals shown in (33) and (34), the space-charge-free field distribution E' is calculated assuming the equivalent smooth conductor and using (32). It is presumed that the field distortion, which is confined only to a small region near the conductor, does not affect the values of these integrals appreciably. After determining pe for each flux line by the iterative procedure described, the corresponding current density at the conductor surface is calculated from

The authors wish to thank Dr. J. H. Simpson of the National Research Council of Canada for furnishing the experimental data and for his keen interest in this work.
REFERENCES
Berlin: VEB Deutscher Verlag der Wissenschaften, 1955, 488-494. [3] W. Deutsch, "tVber die Dichtverteilung unipolarer lonenstrome," Ann. Physik, vol. 5, pp. 589-613, 1933. [4] V. I. Popkov, "On the theory of unipolar D.C. corona, Elektrichestvo, no. 1, pp. 33-48, 1949; NRC Tech. Transl. TT-1093. [51 J. H. Simpson and A. R. Morse, "Corona on direct current transmission lines," Bull. Radio Elec. Engrg. Div. (Natl. Res. Council of Canada), vol. 14, pp. 18-30, 1964. [6] J. H. Simpson, "Theoretical and experimental studies of corona loss from D.C. lines " presented at the Corona Res. Meeting, Montreal, P. Q., Canada, March 17-18, 1966. [7] N. J. Felici, "Recent advances in the analysis of D.C. ionized electric fields," Direct Current, pt. I, pp. 252-260, September 1963; pt. II, pp. 278-287, October 1963. [81 S. D. Conte, Elementary Numerical Analysis. New York: McGraw-Hill, 1965. [9] M. M. Khalifa, "Study of overhead ground wires for dc transmission lines," IEEE Trans. Power Apparatus and Systems, vol. PAS-87, pp. 1648-1656, July 1968. [101 L. N. -Stone,, "EHV single and twin bundle conductorsInfluence of conductor diameter and strand diameter on radio influence voltage and corona initiation voltage," AIEE Trans. (Power Apparatus and Systems), vol. 78, pp. 1434-1443, December 1959. [11] W. Janischewskyj, "Use of artificial contamination for studies of corona in air, presented at the 1966 EHV Symp., Manitoba, Canada.

[1] J. S. Townsend, Phil. Mag., vol. 28, p. 83, 1914. [2] N. A. Kapzow, Elektrische Vorgdnge in Gasen und im Vacuum. pp.

je = kpeEo

(36)

and integrated over its surface to obtain the total corona current. It is suggested that in order to take into account the effect of conductor surface irregularities in Popkov's formula, his equations should be modified in exactly the same way. If Vo and Eo have the same significance as above, the modified Popkov equations will be

0.41P

Vo

ln ln-= (1 + y)12-1 + 1 ro1


PI
H
2

n2 y12
(37)

1Y= __P1

27reokm kEoro.(

H2(38)

NOMENCLATURE

The MKS system of units is used throughout. electric field in the presence of space charge space-charge-free electric field corona onset field at conductor surface height of conductor above ground plane H corona current per unit length of conductor I current density j,j, de constant K ionic mobility k surface irregularity factor m P Popkov's empirical constant radius of outer cylinder R radial coordinate of any point r radius of conductor ro s distance measured along a flux line V voltage applied to conductor corona onset voltage Vo complex plane w coordinates x, y Y nondimensional current z complex plane permittivity of free space 4. potential in the presence of space charge space-charge-free potential flux of the space-charge-free electric field charge density at any point p

Et E E', E' Eo

Discussion
Heine Martensson (ASEA, Ludvika, Sweden): We welcome the new approach taken by the authors to the theoretical approach on the question on corona losses on dc transmission lines. Even if considerable information is available from different test lines, it is essential that this information is backed up by theoretical approaches. It is easier to change a parameter in an equation than to make the same change on a big test line. The result of this new calculation method seems to agree very well with earlier results from Popkov's equation. Evidently, more work has to be done in this field to get a still better correlation between the theoretical and practical approaches. We have a detailed comment on the examples shown in Figs. 5-7 of the paper presented. The corona onset voltage, which we assume is the voltage at which corona can first be detected, seems to be extremely high. The electric field at the conductor surface seems to be between 40 and 50 kV/cm, which is higher than the withstand value in normal atmosphere. We may, however, have misunderstood the definition of corona onset voltage. ASEA published results earlier in [12] from investigations performed through 1964. We have further results from that time and have also made investigations on how to calculate dc corona, which we hope to publish within the next year. REFERENCES Madz[12] N. Hylt'n-Cavallius, S. Annestrand, H. Witt, and V. corona and arevic, "Insulation requirements, corona losses,IEEE Trans. radio interference for high-voltage d-c lines,"

Pei charge density at the emitter charge density at the collector Pc pill,p...*p,,,,i mean charge densities for a flux line

Pe, Pell,
i, (e X0

Power Apparatus and Systems, vol. 83, pp. 500-508, May 1964.

scalar function

dummy integration variable.

Manuscript received February 8, 1968.

726
J. B. Jordan and P. L. Pirotte (Laval University, Quebec, P. Q., Canada): This paper is an excellent example of new possibilities open in sciences with the application of computers. The authors succeeded in presenting an elegant method of calculating the corona current of a dc unipolar conductor on a completely theoretical basis, whereas the preceding formulas were at least semiempirical. With regard to the proposed method itself, we would like to ask the authors whether eventual application of Popkov's formula in calculating the charge density of emitter Pe would not have the advantage of arbitrary choice of coefficients f, and f2 in (30). Thus the current I could be calculated from (11) and (12) and the appropriate emitter charge density from
Pei =-

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, MAY

1969

2-xrokEoD
-

2w1E

(39)

Iterative values of pei could be found, for example, from the function
Pm l P ei = P ei-1) (i + m

Pm(i-l)0

Pm

m(1)

with i = 2, 3,

(40)

where the value of Pm is calculated from (26) and pmZ from (25) and (27). General assumptions accepted in the proposed method may have some influence upon the results of calculations and their relation to the experimental values. The first assumption omits automatically the electron component of corona current at a positive conductor. The effect of this component may be negligible for lower currents and larger distances, but it should be more pronounced in opposite conditions. The electron path is much shorter than the ion path because it is limited to the direct neighborhood of the conductor. On the other hand, however, the electron mobility is much higher than the mobility of ions. The second assumption concerning the constancy of ion mobility

seems to be justified, except in the high-field region around the conductor. However, the authors apply the same value of mobility for positive and negative ions, whereas the results of measurements met generally in technical literature attribute higher mobilities to negative ions. It is rather difficult to evaluate the impact of the fourth assumption without a deeper analysis. This assumption is completely justified in a coaxial field. The distribution of space charge around a real transmission line may be further modified by the wind and updrafts. Vonnegut et al. [13] detected the perturbations of a fair weather electric potential gradient caused by corona discharges from a dc line as far as 10 km from the line. Khalifa and Morris [14] found well pronounced effect of the wind to be well pronounced on the value of corona current itself.
REFERENCES

ground arrangement. This approach is evidently more accurate than the empirical constant P introduLced by Popkov to account for the nonuniform current density over the ground plane. Simpson [6] and the authors multiplied the constant P by three in order to obtain a reasonable agreement between Popkov's formulas (11) and (12) and field-test results. I agree with the authors that P depends upon the test-line geometry. This dependence may account for the fact that curves A and B are closer to each other in Fig. 6 than in Fig. 5 and in Fig. 7 than in Figs. 8 and 9. We found that multiplying P by four instead of three makes the curves A and B in Fig. 9 almost coincide. In using the authors' computer facilities, it may be another valuable contribution to the literature to give some charts indicating normalized corona losses of dc lines as a function of the line geometry and voltage relative to the corona inception level. An alternative may be to compute the variation with the line geometry of the factor P which makes Popkov's formula agree most closely with experimental results. This would be a big help to dc line designers, because a formula or set of charts is usually more easily available than elaborate high-speed computers. One of the authors' simplifying assumptions is that the space charge affects the magnitude but not the direction of the electric field. We realize that it is difficult to proceed in the calculations otherwise, and that it may be more difficult to check this assumption. In Fig. 10 the calculated curve A is significantly lower than the experimental results close to the inception voltage. This may be partly caused by the low value assigned to the mobility of negative ions, i.e., 1.5 X 10-4 was used by the authors instead of about 1.8 X 10-4 m2/V-s. Both of the calculated curves A and B sometimes do not fit the experimental results very well (Fig. 9). Would a better agreement be obtained by considering the effect of stranding on the field distribution at the conductor surface as previously done by Simpson [6]? An additional improvement may be achieved by considering the corona discharge to start on a limited part of the conductor circumference that extends tangentially as the conductor voltage increases. Perhaps the authors would comment on this in the

discussion.

[13] B. Vonnegut, C. B. Moore, G. E. Stout, D. W. Staggs, J. W. Bullock, and W. E. Bradley, "Artificial modification of atmospheric space charge," J. Geophys. Res., vol. 67, March 1962. [141 M. M. Khalifa and R. M. Morris, "A laboratory study of the effects of wind on dc corona," IEEE Trans. Power Apparatus
and Systems, vol. PAS-86, pp. 290-298, March 1967.

Manuscript received February 13, 1968.

M. Khalifa (University of Cairo, Giza, Egypt): The authors should be commended for their rigorous numerical analysis of dc corona on monopolar lines. Their paper is an improvement over Popkov's work [4]. Although the authors in their analysis use the same simplifying assumptions as Popkov's, they compute the point-topoint variation of the charge density in the field space between the high-voltage conductor and the ground plane of the line-to-

J. H. Simpson and A. R. Morse (National Research Council of Canada, Ottawa, Ont., Canada): It is clear from this paper that the dc unipolar corona problem for smooth lines, operating under standard clean conditions, has been solved. It is true that the accuracy is limited by the initial assumptions, but for the range of parameters used in practical unipolar lines the errors thus introduced are negligible. The powerful iterative method introduced by the authors should be very useful in the solution of other problems, and it is good news that they are now applying it to the important dc bipolar case. Another field in which it may prove especially useful is that of electric precipitation. With regard to the effects of stranding on the characteristic the picture is not so clear. Calculations of the field due to stranding performed in this laboratory indicate that stranding lowers the corona inception field strength by about 30 percent and that corona appears first at a single point on the strand nearest the ground plane. It then spreads over the conductor as the potential is raised. If field distortion due to the corona itself were negligible, the whole conductor would not be in corona until the potential reached a value some 16 percent above the corona inception voltage for a smooth conductor of the same diameter. There is thus a broad region of potential, which usually includes the operating potential, in which the line is partially in corona. The situation is aggravated by the fact that the height above the ground plane varies along the span. For this reason we now feel that a simple factor taking account of the lowering of the inception voltage due to stranding, cannot give accurate values of corona current in the normal operating region. This is especially true for large-diameter stranded conductors which tend to be operated below their theoretical corona inception

Manuscript received FebrLiary 13, 1968.

Manuscript received Februiary 16, 1968.

SARMA AND JANISCHEWSKYJ: CORONA LOSSES ON DC TRANSMISSION LINES

727

voltage. An attempt to take account of this "spreading" effect in an iterative calculation would form an interesting extension of the authors' work. However, from the point of view of application, the dividends would be greater in the case of the bipolar lines where the effect should be equally important. A final remark on the Popkov factor P is in order. We increased this factor empirically in an attempt to take into consideration air pollution and humidity effects at very high potentials. The introduction of the factor m, taking stranding effects at lower potentials into account, is a definite improvement although we are not clear on the theoretical justification for [37] and [38]. Perhaps Dr. Sarma and Prof. Janischewskyj wouild enlarge upon this in the discussion. In conclusion, we congratulate the authors of this paper for an exceptional piece of work which should assist materially in the development of a theoretical base for the design of dc transmission lines.

N. Hylten-Cavallius (Research Institute, Hydro-Quebec, Montreal, P. Q., Canada): May I compliment the authors for an extremely interesting and well-written paper. I am glad to note that this problem of strict theoretical predetermination of the unipolar dc corona is solved or on its way to being solved. But in this connection I would like to underline the comments by Mr. Martensson; the practical importance of dc corona losses is in the range of 15-25 kV/cm, which is lower than the range covered by the authors. I also think that a study of the losses during rain is of some practical importance and wonder whether the effect of rain can be mathematically interpreted as a decrease of the corona inception field strength Eo at the conductor. Some tests carried out during my time with ASEA in Sweden on a unipolar experimental line up to 1 MV may give a clue to the problem. The losses during heavy rain on a conductor with "practical" diameter seemed to increase to the same value as the losses under dry and wet conditions on the same conductor being provided with a number of small points along its entire length. They also seemed to reach roughly the same value as for a conductor of very small diameter. Another problem of practical importance is the bipolar dc corona, for which precalculation methods should be extremely valuable. It may suffice, however, with a determination of the derivative of the losses with reference to the most important parameters. This would permit the recalculation of experimental values obtained in one configuration to those valid for another. Such an analysis may be simpler than a determination of the losses themselves. Finally, I wish to draw attention to paper [15] put at my disposal by Dr. Popkov which should, according to him, also contain a rigourous solution of the unipolar corona.

In Electricite de France we have developed a rather different method for calculating corona losses on ae lines. Instead of integrating the differential equations linking field and current density in the space around the conductors, our program calculates the movement of space charges emitted by the conductor, taking into account the action of all other charges either on the conductor or in space. This method has been described in [16]. The comparison between the results of the calculation and those obtained in our EHV test station is very good, making us quite confident in this method of calculation as far as ac corona losses are concerned. Concerning dc, we have as yet no experience on our method. The principle of the method should still be valid, nevertheless. We have used it to calculate the corona current when a step of voltage is applied to a conductor. At the first moment this current is very high and subsequently decreases slowly to what the steadystate dc corona current should be. In one example treated, the step of voltage was 600 kV positive. The conductor diameter was 26.4 mm, with a stranding factor of 0.8. It was surrounded by a cylindrical collector with a radius of 7.65 m. The dimensions are rather similar to those on which Fig. 9 is based, the main difference apparently being in the conductor diameter. However, the steady-state current that we have calculated was 800,uA/m, which seems to be higher than the calculated values of Fig. 9. We would therefore like to ask the authors if they have experience, through testing or calculation, on the variation of corona current with the conductor diameter, provided the surface gradient without corona does not change and is very high. With ae voltage the capacitance of the conductor practically defines, with sufficient accuracy, the position of the earth, which the authors call the collector. This is due to the fact that the losses physically take place chiefly within 1 or 2 meters around the conductor. Is this approximation valid for dc? Our last question concerns the mobility of the ions. How was mobility chosen by the authors? Does the difference between positive and negative ion mobility justify the large difference between corona currents which are given in Figs. 9 and 10 and cannot be explained only by the difference in onset voltages?
REFERENCES

[16] J. J. Clade, C. H. Gary, and C. A. Lefevre, "Calculation of


corona losses beyond the critical gradient in alternating voltage," this issue, pp. 695-703.

(Bonneville Power Administration, Portland, Oreg.): The presentation includes practical suggestions which may be very important on the investigation of de corona. From the author's theoretical analysis, as well as from the result REFERENCES of investigation on testing lines [17], [18], it appears that the height influence rather [15] G. T. Usynin, "The calculation of the field and characteristics of a conductor has a large difference betweenthan the conductor the average surface of d.c. unipolar corona discharge (wire parallel to plane)," surface voltage gradient. The Izv. Akad. Nauk SSSR, Energ. i Transp., no. 4, pp. 56-70, voltage gradient and the maximum surface voltage gradient is 1966. too small to explain the large influence of height. Therefore, analysis based on the surface voltage gradient, which has been the most favorable method adopted for 60-Hz corona Manuscript received February 15, 1968. analysis, is not adequate in the case of dc corona (Fig. 12). Also, adoption of a corona cage or a reduced size model would bring different results as compared with actual overhead lines in the case of dc. Do the authors agree?
T. Udo

author's

REFERENCES

J. Clade and C. Gary (Electricit6 de France, Paris, France): In their very interesting paper, the authors give a mathematical method for calculating corona current due to dc voltage. The results of the calculation are confirmed by tests in a large range of voltage, and the authors are to be congratulated for this success.

"Corona losses, radio interference [17] N. Hylten-Cavallius et al., for and insulator requirements HVDC," Direct Current, vol. 9, August 1964. [18] R. Morris and B. Rakoshdas, "An investigation of corona loss and radio interference from transmission line conductor at high direct voltage," IEEE Trans. Power Apparatus and
Systems, vol. 83, pp. 5-16, January 1964.
1968.

Manuscript received February 19,

1968.

Manuscript received February 19,

728

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS,

MAY 1969

construct a method which would give an exact solution in the general case. The iterative method presented in the paper may be considered as greatly increasing the accuracy of determination of local quantities of the dc unipolar field. For instance, I am sure that, by using relation (15), this method considerably increases the knowledge of the charge density distribution in the interelectrode space. Nevertheless, because of the assumption that the space charge does not affect the direction of the elecAric field,< only an approximate solution is obtained. If it were the exact -solution, the quantity , defined by (16), would be a function of +

r = f(0) independently of the flux lines; however, in general case, this is


not so.

/L./

I/

VOLTAGE GRADIENT

Fig. 12. Relation between average conductor surface gradient (kV/cm) versus monopolar dc corona loss (kW/km). Solid lines are reproduced from author's calculation. Dotted lines are quoted from laboratory investigations [17], [18].
E. Schaefer (400 kV-Forschungsgemeinschaft e. V., Heidelberg, Germany): This paper is most valuable because it represents a first step to calculating the corona losses on a dc transmission line. Although the studies and calculations cover ranges of surface gradients which will not occur with a practical transmission line, a way is yet shown for calculating the corona losses. Also one of the aims of the 400 kV-Forschungsgemeinschaft is to predetermine corona losses and radio interference of any dc line in a similar way, as it is already possible for ac lines. Experimental tests to this purpose are still under way. However, main consideration is given to bipolar lines with surface gradients up to the order of 30 kV/cm, as have to be expected in practical operation, but the test program also includes unipolar lines. We try to determine experimentally the effects of the surface gradient, the pole distance, the conductor diameter, and the number of subconductors of a bundle conductor on the corona losses and radio interference. These tests are not only realized on a dc line but also by measurements in a corona cage. In this cage it will also be possible to produce surface gradients of similar magnitude as used in the investigation referred to in the paper so that a comparison of the results would be possible. The discussed method for a numerical evaluation of corona seems to become very difficult and extensive in the case of bipolar dc lines. Do you think it possible also for this problem to achieve a result of sufficient accuracy for practical application by simplifying assumptions? In such a case the 400 kV-Forschungsgemeinschaft would be glad to communicate results of their experimental investigations for comparative purposes.

Assumption 4) may have a meaning for some configurations in which the angular deviation of the flux lines between the ionized and the space-charge-free field is very small, but in the case of the line-to-plane configuration the angular deviation may take important values due to the very asymmetric geometrical aspect. I suppose that this would be seen if one plotted the values of the potential by the iterative method described by the authors and if one compared the flux lines of this potential with those of the spacecharge-free field. Moreover, we must keep in mind that, for a given point, the angular deviation depends on and increases with the ionization intensity, the maximum probably occurring for the saturated solution. The method presented by the authors for evaluating corona losses is the only valid one in principle. Other methods, such as that of Popkov, are mathematically unwarranted and give only rough approximations. In spite of the fact that the approximation due to assumption 4) is not warranted in some regions of the interelectrode space, the corona current is determined with a good accuracy; I think the compensation of the two terms p and E and the subsequent integration give a more accurate result for the current than for the local values of the field and the charge density. A very interesting point in the paper is the method of evaluating pe. In my work [19] I proved that the problem of the dc unipolar electric field is "properly posed," as mathematicians say, if together with (6) the following boundary conditions are given: 1) potential values on the two electrodes, and 2) charge density values on the emitter. As my work is essentially mathematical, I am not concerned with the determination of the charge density on the emitter from the experimental conditions. We are here confronted by a considerable difficulty and the authors' method of evaluating p. constitutes the first and an important contribution to its solution.
REFERENCES

[19] P. Atten, "D.C. ionized fields," Centre National de la Recherche Scientifique, Laboratoire d'Electrostatique et de Physique du Metal, Grenoble, France, Internal Rept.

Manuscript received February 15, 1968.

V. I. Popkov (Academy of Sciences, Moscow, USSR): It is both interesting and useful to find some suitable algorithm to calculate corona losses using computer technique. It may help to take into account the possible influence of some factors which are ignored or simplified in analytical descriptions of voltage-current characteristics of corona. Consider, for instance, the calculations using (1)-(4), without using a simplifying supposition of the kind

P. Atten (Centre National de la Recherche Scientifique, Grenoble, France): I have undertaken a theoretical analysis of dc ionized fields from a mathematical point of view, and my purpose is to

(41) where EP is the vector of the field intensity corresponding to the Poisson equation, EL is the vector of the field intensity corresponding to the Laplace equation, and 0 is some scalar function. Such a solution has already been devised on a sound mathematical basis [15]; of course, a great number of calculations are necessary to get closer and closer approximations, a task appropriate for a computer.
=

EP

OEL

Manuscript received February 19, 1968.

Manuscript received February 11, 1969.

SARMA AND JANISCHEWSKYJ: CORONA LOSSES ON DC TRANSMISSION LINES

729

It is also of interest to find the influence of ion mobility when K = variable (42) as it takes place in the actual field [20]. There are some other problems leading us to a more rigorous description and better understanding of the unipolar corona process. Unfortunately, what we actually find in the paper does not lead us any further. The approach is well known; in [4], solutions of (1)-(4) were obtained, based on (41) (which differs from (16) only in the symbols used). The final formulas describing field and charge distribution are obtained directly, within the limits of (41), as well as a more [4] or less [21] rough approximation of the voltagecurrent characteristics of unipolar corona. Although rather simply shaped electrodes were actually considered (a wire or live and ground wires parallel to ground), it is a matter of routine to find analogous solutions for other arrangements, such as bundle conductors, etc., using appropriate conformal transformations. That is essentially the same procedure that is now considered by the authors, who entrust to a computer the calculations which were already well adapted to slide-rule use. We regret to mention the mathematical error or negligence in the manipulations with vectorial equation (13). In dealing with a two-dimensional problem one naturally cannot substitute the normal derivative dp/ds for Vp. It may be shown that such a substitution means p is constant on the equipotential lines, which carries to an absurdity: p is constant from - co to + X on the ground plane. Rules of vector algebra would evidently lead to partial differential or integro-differential equations instead of what is written and used as (14), (15), and (24). Naturally it would be inconsiderate to draw any conclusions on the basis of such equations. It was interesting and complimentary to find in the paper references on several occasions to "Popkov's formula" which challenges

necessary. Eo in our formula, as well as in the authors' equations, is not a corona starting gradient for a smooth wire as it seems the authors supposed it to be. Eo is the boundary value, the field intensity at the surface of the wire under corona, which approximately may be equivalised as the corona starting gradient. Naturally, in the case of smooth wires with starting gradient Eo' one uses Eo = Eo', and for stranded wires as usual one uses Eo = mE0', where m is the surface irregularity factor. In the formula (38) there is a misprint. Evidently instead of m there should be M2.

1) To clear away the authors' speculations on this formula we would like to draw attention to the following facts. As shown in an English edition [22] rather good correlation of calculations was found with the results of experiments involving many variations in conditions: both polarities of the voltage, wire diameter varying from 0.11 to 3.71 cm, H/ro = 54 + 665, wires having smooth surfaces or ACSR, and dry surfaces as well as those wet by artificial rain. 2) We thank the authors for their efforts to modify "Popkov's formula" (see Figs. 8-10) although such a modification was un-

us to clarify some points.

The assumption of constant ionic mobility is generally valid excepting in the high-field intensity region. It is also generally true that the mobilities of both positive and negative ions are comparable although the presence of a few fast moving electrons will increase somewhat the apparent mobility of the negative ions. In a gaseous mixture such as air, it is rather difficult to assign a definite value for the mobility of ions, as in addition their mobility may be reduced by the formation of ion clusters. The values of ionic mobilities in air found in the literature vary from about 1.2 X 10-4 to 1.8 X 104 m2/V-s. Since all pertinent parameters affecting ionic mobility cannot be included in the analysis, a constant average value of 1.5 m2/V.s has been chosen for both positive and negative ions. Another important assumption made in the analysis implies that the space charge affects only the magnitude but not the direction of the electric field. Let us state first that without this assumption, the analysis of ionized fields would be extremely difficult as it would require numerical solution of nonlinear partial differential equations in complex geometries. The assumption is perfectly valid in symmetrical configurations such as the concentric cylindrical geometry, and it will be sufficiently close to reality also in other configurations if the space-charge density in the interelectrode region is not very large. Fortunately, in practical transmission-line configurations the field in the vicinity of the conductors, the region in which the space-charge density is a maximum, is very nearly cylindrical and therefore practically symmetrical. In the rest of the interelectrode region, where the field is not symmetrical, the spacecharge density is comparatively small. Consequently, in practical configurations the assumption that the geometric pattern of the field is unchanged by the space charge is reasonably justified. With reference to Mr. Martensson's question, the corona onset gradients for all the examples considered in the paper have been calculated using the well-known Whitehead's empirical formulas [23]. The withstand value of 30 kV/cm for normal atmospheric air applies only for uniform field gaps, while in the case of corona on conductors, the onset gradients will be higher because of the nonuniform field distribution. The higher values of corona onset gradients in the examples shown in Figs. 5-7 may be attributed to the smaller conductor sizes chosen and to the assumption of smooth conductors surfaces. In the case of practical transmission lines, corona onset gradients will be considerably lower due to the larger conductor sizes as well as due to the conductor surface irregularities. Dr. Jordan and Mr. Pirotte suggest calculation of p ei from Popkov's formula and propose an alternate iteration procedure.

periphery, being lower than the actual value of Pe for the flux lines in the vicinity of K4p = 0, and higher for those near K4, = 7r/2. In the method proposed by the authors the values of Pe1 and Pe2 for any particular flux line are chosen on the basis of Pm for that flux line and therefore are uniformly closer to the correct value of
Pe for all flux lines. Moreover, the method proposed in the paper can be applied to any general unipolar configuration, while that based on Popkov's formula can be applied only to the line-to-plane

Value of P,ei so calculated would be constant around the conductor

REFERENCES [20] V. I. Popkov, "On the ions mobilities in corona discharge," Dokl. Akad. Nauk SSSR, vol. 58, no. 6, pp. 1043-1046, 1947. [21] , "On the theory of d.c. corona discharge in the gas; General characteristics of unipolar corona and volt-ampere relation for the electrodes wire-ground," Izv. Akad. Nauk SSSR, Otd. Tekhn. Nauk, no. 5, pp. 664-674, 1953. , "Some special features of corona on high-voltage d.c. [22] transmission lines" in Gas Discharges and the Electricity Supply Industry. London: Butterworth, 1962.

M. P. Sarma and W. Janischewskyj: The authors are most gratified by the response to their paper and wish to thank all discussers for their contributions. Questions pertinent to several assumptions made in the paper have been raised by a number of discussers and will be treated first.

Manuscript received February 24, 1969.

The iterative procedure proposed by Dr. Jordan and Mr. Pirotte has only a linear rate of convergence while the secant method used in the paper has a very nearly quadratic rate of convergence. Fig. 13 shows the number of iterations necessary, in a typical case, to reduce the error to less than about 0.1 percent by the two methods. The secant method A requires only about five iterations, while the iteration procedure suggested by the discussers B requires about 14 iterations to achieve the same accuracy. Dr. Jordan and Mr. Pirotte further comment on some additional assumptions made in the paper. The electron component of the corona current is significant when establishing the transient behavior of corona discharges. For determination of the average corona loss, however, it is sufficient to consider only the continuous average drift of ions. The authors agree that the effect of wind will change the space-charge distribution, especially in the low-field intensity region. However, inclusion of wind effects, although possible, should considerably complicate the analysis. The authors concur with Dr. Khalifa that the proposed method of analysis should be employed to obtain data on dc corona losses in a form that can be easily applied to practical transmission-line design. The authors have adopted the proposed method to various practical transmission-line configurations and hope to publish the results soon. With reference to Fig. 10, indeed a higher value of

configuration.

730

IEEE

TRANSACTIONS ON

POWER APPARATUS AND

SYSTEMS;

MAY

1969

The total corona current from the conductor is given by I = 2r roje or, substituting from (36), I = 27rrokpEo. Now, Pe may be calculated from (1) as
Pe
=

(46)
(47)

EoV.E.

(48)
(49)

Substituting for E from (43) (r = ro for the equivalent smooth conductor), A2


P = 0

o/A'ro2
R2

+ B2

The constant A is obtained by substituting (45) and (49) in (47): I fRt2 mI m2EO2rO2

2ireo-

2r-eokm kEoror

(50)

putting
I R 27reokm \Eoro/ the constant A may be written as A2 = (m2Eo2ro2/R2) Y.

NUMBER OF ITERATnONS (i)

Fig. 13. ionic mobility should be needed to make the theoretical calculations agree with the experimental data. The authors appreciate the comments of Dr. Khalifa and those of Dr. Simpson and Mr. Morse on the effect of gradual spreading of corona around the conductor due to stranding upon corona losses. The authors have independently analysed the field distribution in the vicinity of a stranded conductor by solving numerically Laplace's equation, and their results agree very well with those of Dr. Simpson. However, as already pointed out in the paper, the increase in the conductor surface voltage gradient due to stranding, or any other surface irregularity, cannot be directly related to the reduction in the corona onset voltage, since it depends not only on the value of the surface voltage gradient but also on the field distribution in the vicinity of the conductor. Nevertheless, the nonuniform generation of corona due to stranding may be included in the proposed analysis, provided the actual variation of corona onset voltage for different points on the conductor periphery is known. Likewise, it is possible to take into account the effect of conductor sag by dividing the span into a number of segments, the height of each being assumed constant. Dr. Simpson and Mr. Morse ask about the theoretical justification for the introduction of the factor m in (37) and (38). The procedure consists of first deriving the voltage-current characteristic of a concentric cylindrical configuration with the conductor having surface irregularities such as stranding, etc., and subsequently introducing the Popkov's empirical constant P to make the result agree with the corresponding line-to-plane configuration. In the classical solution for the cylindrical geometry, the ionized field is assumed to be of the following form:

(51)
(52)

The constant B is obtained by substituting for A in (45):


B2
=

m2Eo2ro2[l

(ro2/R2)yJ.

Thus the ionized field is obtained as


EmEoro1

ro2

+r2 y]

(53)

Integrating (53) between the limits r = ro and r = R, and using (44), the equation for the voltage-current characteristic is obtained. The resulting equation is identical to (10) excepting that Vo and y are modified as described above to include the factor m. The corresponding equations for the line-to-plane geometry are obtained directly from the above analysis by introducing the empirical constant P and are given by (37) and (38). The authors agree with Mr. Hyltln-Cavallius that the effect of rain may have to be interpreted as a reduction in the corona onset voltage of the conductor. In such a case the effect of rain should be treated as a conductor surface irregularity. In reply to the discussion by Mr. Clade and Mr. Gary, the authors would like to state that the space-charge conditions in ac and dc are quite different, and therefore it would be rather difficult to obtain a theoretical correlation between corona losses present in the two cases. For the configuration cited by the discussers, the corona current calculated using Townsend's formtulas (9) and (10) is about 3.41 LA/m. Therefore, it seems that the large discrepancy observed by the discussers is probably due to the method of calculation rather than because of the small difference in conductor size. Finally, the difference between corona currents of Figs. 9 and 10 is mainly due to the difference in conductor height. The authors share with Dr. Udo the opinion that the dc corona losses depend more on the conductor height than on the conductor -/A2r2 + B2 that the line configuration has much E -= Ar'+B (43) diameter and performance than in the case ofa ac. larger influence on its corona r Mr. Schafer, Mr. Hylten-Cavallius, and Dr. Simpson and Mr. where arbitrary constants A and B are determined from appropriate Morse rightly point out the greater practical importance of biboundary conditions. The actual conductor with the surface ir- polar corona. The analysis of bipolar corona is further complicated regularities is replaced by an equivalent smooth conductor with by the mixing of space charges of opposite polarity and the concorona onset voltage reduced by the factor m sequent ion recombination in the interelectrode space. The authors have developed a similar analysis of bipolar corona also and hope Vo = mEoro hn (R/ro). (44) to make it public in the near future. The offer from the 400 kVForschungsgemeinschaft for exchange of information is hereby The space-charge-free field is calculated by considering only the gladly accepted. equivalent smooth conductor as discussed in the paper. According The authors appreciate Mr. Atten's comments on the theoretical to the boundary condition v, therefore, implications of the fourth assumption made in the paper. The authors nevertheless consider the assumption justified from an engineering VAro'2 + B'2 E (45) point of view when it is attempted to determine corona losses on practical dc transmission lines.

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS, VOL.

PAs-88, NO. 5, MAY 1969

731

The authors wish to thank Dr. Popkov for bringing to their attention the work of Mr. Usynin on unipolar corona. Mr. Usynin has obtained a rigorous analytical solution for the particular case of the line-to-plane geometry. It seems extremely diffictlt, however, to apply his method to the other practical configurations. In reply to Dr. Popkov's second question, the authors would like to point out that in (13) it is the vector dot product E.Vp which has been replaced by Edp/ds and not Vp by dp/ds, as was wrongly interpreted by Dr. Popkov. Since the direction of the electric field is assumed to be unchanged by the presence of the space charge, E.Vp will simply be the projection of Vp in the direction of E and, therefore, equal to Edp/ds. The subsequent conclusion drawn by Dr. Popkov, that the authors' analysis leads to the physical absurdity of constant charge density along the equipotential surfaces, is thus seen to be a direct consequence of his misinterpretation. The authors would like to emphasize that the paper's main contribution to the theory of unipolar corona is the method of calculating, unider the simplifying assumptions made in the paper which are also similar to those made by Dr. Popkov and others, the correct charge distribution on the conductor surfaces. Admittedly,

the analysis is still limited by the simplifying assumptions made, but it is definitely an improvement over the existing theories. Popkov's analysis, although starting from apparently the same set of assumptions, introduces several other simplifications, which are not necessary from a mathematical point of view, in calculating the current distribution on the electrodes. One of the reasons why Popkov's simplifications do not give rise to large errors in the calculation of corona currents is apparently due to the compensating nature of the errors in the calculation of charge and electric field distributions [7]. Agreement of Popkov's theory with experimental results does not, therefore, completely justify its theoretical validity. Finally, the authors' approach to include the effect of conductor surface irregularities in the calculation of corona losses is not based on a simple substitution of mE0 as was done by Popkov. The actual basis for the introduction of the factor m is described in the authors' reply to the discussion of Dr. Simpson and Mr. Morse.
REFERENCES

[23] J. B. Whitehead, "High voltage corona," in International Critical Tables. New York: McGraw-Hill, 1929.

Impedance of Nonmagnetic Overhead Power

Transmission Conductors
P. SILVESTER, MEMBER,
IEEE

Abstract-A high-accuracy method of calculating skin-effect resistance and reactance of nonmagnetic stranded overhead power transmission conductors is described. This method uses an equivalent network approach and takes into account the effects due to wire spiraling, as well as the proximity effect between strands and the nonuniform current distribution in each strand. It produces results of greater accuracy than manufacturing tolerances usually warrant for overhead transmission conductors and requires a very small investment in computation cost. A table of skin-effect resistance factors that covers a wide range of standard aluminum conductors has been computed.

INTRODUCTION

TRANDED aluminum, as well as aluminum alloy conductors, have in recent years assumed importance in transmission line design. Particularly in extrahigh-voltage transmission where heavy currents, radio interference, and corona problems all make large conductor diameters desirable, the allPaper 68 C 57-PWR, recommended and approved by the Transmission and Distribution Committee of the IEEE Power Group for presentation at the 1968 EHV Transmission Conference, Montreal, P.Q., Canada, September 30-October 2. Manuscript submitted May 20, 1968; made available for printing December 17, 1968. This work was suLpported in part by the Aluminium Laboratories Ltd., and in part by the National Research Council of Canada. The author is with the Department of Electrical Engineering, McGill University, Montreal, P. Q., Canada.

aluminum designs may show economic advantages over the more conventional steel-cored conductors [1]. At present, there exist few standards for large-diameter conductors, and most large conductors are in fact designed to satisfy the requirements of specific projects. However, for operation at 60 Hz, strandings are prescribed for conductors up to a 1.75-inch diameter by a Canadian standard [2], while British standards cover the smaller sizes of both aluminum and alloy conductors intended for 50-Hz service [3], [4]. The conductor power loss, the operating temperature, and, ultimately, the economic choice of the conductor, are significantly influenced by the conductor resistance, and it is therefore important for the designer to have a reasonably good assessment of the actual resistance at the operating frequency. The resistance rise due to skin effect is generally considered negligible in conductors up to about 1 cm in diameter, while for conductors of about 2-cm diameter sufficient accuracy is usually obtained by assuming that the resistance is equal to that of a solid cylinder of the same diameter and dc resistance. For large conductors, especially where mixed aluminum and alloy strandings are employed, better methods of resistance calculation are required. The nonuniform current distribution in the conductor may occasionally also have an effect on the line reactance, making it desirable to take account of its influence [51. The present paper furnishes a method of computing the actual ac resistance and reactance of nonmagnetic stranded conductors,

You might also like