14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 1

RESONANCE EFFECTS DUE TO CONDUCTOR TRANSPOSITION

IN THREE-PHASE POWER LINES

J. A. Brandão Faria and M. V. Guerreiro das Neves

Centro de Electrotecnia Teórica e Medidas Eléctricas – Instituto Superior Técnico
Lisbon, Portugal
[email protected] [email protected]

Abstract — This paper is concerned with the effects The line structure in Fig. 1 includes three line sec-
produced by conductor transposition in long overhead tions A, B, and C, of length lA, lB, and lC respectively.
three-phase power lines. We conduct a frequency domain Assuming that each individual line section is uniform,
analysis that shows for the first time to our knowledge that the series-impedance Z(z) and shunt-admittance Y(z) in
the modal attenuation constants and modal phase veloci-
(1) are stepwise functions:
ties characterizing the transmission line structure exhibit a
repetitive resonant behavior for frequencies such that the
overall transposition cycle length gets close to an integer ZA for 0 < z < l A
multiple of one half wavelength. Consideration of these  −1
Z ( z ) =  Z B = PZ A P for l A < z < l A + l B
resonance phenomena is of major importance and should
be taken into account in a variety of situations, PLC com-
 Z = P −1 Z P for l + l < z < l + l + l
 C A A B A B C
munications, EMI, and line transient studies.
(2)
Keywords: Transposed conductors, multiconductor  YA for 0 < z < l A

transmission lines, modal analysis, nonuniform lines, Y ( z ) =  YB = PY A P −1 for l A < z < l A + l B
modeling and simulation, resonance  Y = P −1Y P for l + l < z < l + l + l
 C A A B A B C

1 INTRODUCTION where ZK and YK (with K = A, B, C) denote the series-

impedance and shunt-admittance matrices for the Kth
For balancing purposes many overhead power lines
line section. The permutation matrix operator P, in (2),
have their phase conductors periodically transposed
defined as
along the line length, a complete transposition cycle
consisting of three cascaded line sections differing on 0 1 0 
the conductor sequence order —see scheme in Fig. 1. P = 0 0 1  (3a)
The resulting global structure is thus to be classified as a 1 0 0
nonuniform line since its per-unit-length series-
impedance Z and shunt-admittance Y matrices are not
obeys the following special properties:
constant along the line longitudinal coordinate z.
The theme of nonuniform transmission lines is an
important actual research topic with applications rang- P 2 = P −1 = P t ; P 3 = E (3b)
ing from microwave systems [1-2] to power systems [3-
4]. This paper is a new contribution to the latter area. where E is the identity matrix, and superscript t denotes
The frequency-domain multiconductor transmission transposition.
line equations for nonuniform lines are
The standard approach to deal with the line structure
d d depicted in Fig. 1 employs a homogenization technique
V = −Z (z ) I ; I = −Y (z ) V (1)
dz dz leading to a uniform line model described by balanced
per-unit-length series-impedance <Z> and shunt-admit-
where V and I are complex column vectors gathering tance <Y> matrices:
the conductor voltages and conductor current phasors.
(A) (B) (C) Z p Zm Zm 
2 1  
1 3 < Z > = (Z A + Z B + Z C ) =  Z m Zp Zm 
1 3 3
2 Z Zm Z p 
3 2 1  m
lA lB lC (4)
z Y p Ym Ym 
 
0 L < Y > = 1 (Y A + YB + YC ) = Ym Yp Ym 
3
Figure 1: Transposition cycle for a single-circuit three-phase Y Ym Y p 
line structure.  m
14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 2
The preceding matrices being used for the evaluation of d
the propagation parameters characterizing the normal d 60º d
modes of the structure —one ground mode and two de-
generate aerial modes. h h h
As is well known, the modal attenuation constants α
and modal phase velocities v obtained with this ap-
proach are smoothly varying functions of the frequency,
(a) (b) (c)
Ground mode (g):
α g (ω ) + j ω = ( Z p + 2 Z m )(Y p + 2Ym ) (5a) Figure 2: Line configurations showing conductor arrangement
v g (ω ) and dimensions, h = 25 m, d = 10 m, conductor radii r=16 mm.
Aerial modes (a): (a) horizontal line, (b) vertical line, (c) triangular line.

α a (ω ) + j ω = ( Z p − Z m )(Y p − Ym ) (5b)
va (ω ) 2.1 Analysis of one Line Section
The entries of the complex symmetric matrices Z A
A first flaw with the standard approach can be im- and Y A depend on the frequency and on the geometri-
mediately pointed out by noting that equivalent uniform cal arrangement of the line conductors. Because the
overhead lines with Z = <Z> and Y = <Y> can never be commonly used procedure for computing the per-unit-
found [5]. Even, in the simplest case of losseless lines, length series-impedance and shunt-admittance matrices
the product <Z><Y> does not yield a diagonal matrix, has been established elsewhere [6-7] details will not be
modal phase velocities obtained from (5) being not gone into here and, thus, only a brief summary is pre-
equal to the speed of light in a vacuum, v ≠ c. sented. For section A, Z A and Y A are obtained via
The main problem with the standard approach is that
it is not based on a correct model, as the real structure is Z A = jωL + Z g + Z sk ; Y A = j ω2 L−1 (6)
not uniform along its entire length due to the presence c
of transposition interconnects. where L and Zg are, respectively, the external-induc-
The discontinuities associated with conductor trans- tance matrix and the ground return impedance matrix
positions generate forward and backward propagating (see Appendix I). Zsk, the phase conductor internal im-
waves that strongly interfere at certain critical frequen- pedance, is a frequency-dependent complex diagonal
cies, thus making the wave propagation parameters to matrix whose elements can be determined by using
vary not smoothly with the frequency. skin-effect theory results for cylindrical conductors, in-
In this paper we will take into account the nonuni- volving Bessel and Neumann functions.
formities originated by the presence of conductor trans-
positions. A typical single-circuit three-phase overhead The solution to the propagation equations in (1) can
line configurations with frequency dependent parame- be presented in various forms [8-9]. However, bearing
ters, consisting of three uniform line sections 100 km in mind the problem we have to solve, the solution in
long each, will be used as an example. the form of a transmission matrix is the most adequate
By cascading the line sections, the overall 6 × 6 - one.
transmission matrix characterizing the structure will be The transmission matrix for section A, TA, yields the
determined and from it the corresponding propagation relationship between line currents and voltages meas-
constants will be extracted. The modal attenuation con- ured at ports z = 0 and z = lA
stants and modal phase velocities will then be obtained
and plotted against the frequency, plots showing that V (0) V (l A ) A A BA
resonance manifestations repeatedly occur for frequen-  I (0)  = T A  I (l )  ; T A = C 
DA
(7)
cies that are a multiple of 500 Hz.    A   A
The magnitude and width of the resonance peaks are
influenced by several factors. In this paper we will dis- where the square complex sub-matrices AA, BA, CA, and
cuss the influence of line losses (wire losses and ground DA, can be obtained from ZA and YA as follows, [8-9]:
losses) as well as the influence of some line geometrical
parameters, e.g., section lengths and conductor ar- A A = D tA = COSH[S A l A ] (8a)
rangement (cross-sectional views of typical line con-
figurations showing conductor distances and heights are B A = B tA = SINH[S A l A ] S A Y A−1 (8b)
depicted in Fig. 2).
C A = C tA = Z −A1S A SINH [S A l A ] (8c)
2 MODAL ANALYSIS EQUATIONS
In (8), the matrix SA represents the square root of the
The multiconductor transmission line matrix equa- ZAYA product. The determination of
tions in (1) should be solved separately for each line
S A = SQRT(Z AYA )
section domain A, B, and C. However, since the solu-
tions regarding sections B and C can be obtained (using involves the solution of a matrix eigenvector-eigenvalue
the permutation operator) from the one for section A, it problem concerning the ZAYA matrix product (see Ap-
will suffice to pay attention to the latter. pendix II).
14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 3
2.2 The Transmission Matrix of the Structure 3.1 The Influence of Line Losses
The transmission matrix T characterizing the overall This subsection is concerned with the horizontal line
line structure is obtained by multiplying the individual configuration in Fig. 2a, attention being paid to the in-
transmission matrices pertaining to the three line sec- fluence of line losses on the computation of the modal
tions A, B, and C. propagation constants of the transposed line structure.
To take losses into account typical parameter values
A B  of 57.3 mΩ/km and of 100 Ωm were assigned to the
T = T A TB TC =   (9)
 C D wires dc resistance and ground resistivity, respectively.
Assuming a regular transposition scheme with lA = lB
Matrices TB and TC can be evaluated following the = lC = L/3 = 100 km we see from (13) that the first two
guidelines outlined in 2.1 for TA. However, such a job resonance events are expected to take place at about 500
can be greatly simplified by making use of the existing and 1000 Hz. Therefore, a frequency scan in the range
relationships (2) among ZA, ZB, ZC and YA, YB, YC, 400 to 1100 Hz will be suited for assessing the behavior
which also extend to the square-root matrix functions of the modal propagation constants as a function of the
-1 -1 frequency.
SA, SB, SC, that is, SB = PSAP and SC = P SAP.
Further simplification arises in the case when the The overall attenuation in dB, α (ω ) L × 20 log 10 e , as
line sections A, B, and C, have exactly the same length, well as the phase velocity in normalized units, v(ω)/c,
lA = lB = lC. For a strictly regular transposition scheme were evaluated for all the propagation modes.
we find To gain insight into the problem the strategy of pro-
−1 −1
gressively increasing the degree of complexity was
P 0   P 0  P 0  P 0  used. First, the modal propagation constants are evalu-
TB =  TA   ; TC =   TA  
 0 P  0 P  0 P  0 P ated considering an ideal lossless line. Then, wire losses
are added. Finally, the real scenario with wire and
thus yielding ground losses simultaneously present is examined. Re-
3 sults obtained are shown in Fig. 3. The following con-
A B  A A P B A P 
T= =  (10) clusions can be drawn:
 C D  C A P D A P  w Even when losses are absent structural attenuation
bands do indeed show up at around 500 Hz and 1 kHz.
The eigenvalues of T are [8] the positive and nega- In the lossless line case these bands are identified by
tive exponential propagation factors: regions of non-null attenuation with linearly increasing
phase velocity (for frequencies inside these bands the
+γ g L −γ g L
e ,e , e +γ a′L , e −γ a′L , e +γ a ''L , e −γ a '' L (11) net power flow in the line is zero —electric and mag-
netic fields have a phase lag of π/2).
where L = lA + lB + lC is the length of the transposition w The transitions between attenuation bands and
cycle and γk is the kth propagation constant of the overall passing bands, which are very sharp in the lossless case,
structure yielding the required information for comput- become smoother and smoother as losses are included.
ing the modal attenuation constants and modal phase w The resonant characteristics of the aerial modes a’
velocities and a”, although identical in magnitude, are frequency
γ k = α k (ω ) + j ω , for k = g, a’, a” (12) shifted. While mode a’ predominantly resonates at 500
vk (ω ) Hz, mode a” predominantly resonates at about 1 kHz.
where subscripts g, a’, and a” are used as remainders w The importance of the resonance peaks in the aer-
for ground and aerial modes. ial modes deserves special remark. In particular, it must
be emphasized that those resonance peaks are very
3 RESULTS weakly affected by line losses, an attenuation peak of
almost 2 dB being observed with or without losses con-
To get a clear understanding of the influence of the sidered. The value reached by the attenuation resonance
many factors that impact on the propagation properties peak is about 4 or 5 times greater than the one predicted
of the transposed-line structure a step by step procedure in (5b) by the standard approach.
will be used. First, we consider a commonly used over- w The ground mode shows identical resonance fea-
head line geometry —a horizontal line configuration— tures either at 500 Hz or 1 kHz. In the lossless case, the
and analyze the influence of line losses, i.e., wire and attenuation and phase velocity characteristics of the
ground losses. Afterwards, by changing the conductor ground mode cannot be distinguished from those be-
configuration and section lengths, we will examine the longing to one of the aerial modes —they are super-
influence of line geometrical parameters. posed. However, they immediately split as soon as wire
In this work attention will focus on certain critical losses are accounted for.
frequencies for which the structure period gets close to a w Contrary to the aerial modes the ground mode
multiple of one half wavelength L ≈ mλ / 2 , that is, for resonance characteristics are strongly affected by line
resonant frequencies given by losses. When only wire losses are considered the ground
mode attenuation peak still remains relevant, its value
f res ≈ mc /(2 L) (13)
being about three times greater than the one predicted in
14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 4
NO LOSSES INCLUDED WIRE LOSSES INCLUDED TOTAL LOSSES INCLUDED

a’ a“ a’ a“

a’
a“
a“, g a’, g a’ a“

a’
a’ a’

g g a“

a“
a“

Figure 3: Plot of the attenuation and phase velocity characteristics against frequency for the horizontal line configuration, assuming
a strictly regular transposition cycle. Resonance effects are displayed for three different situations: no losses, wire losses, and total
losses included. Dotted lines represent the attenuation and phase velocity characteristics obtained using equation (5) from the stan-
dard homogenization technique. The ground mode is identified by label g, whereas the aerial modes are identified by a’ and a”

(5a) using the standard approach. However, when w For strictly regular transposition cycles the aerial
ground losses are added, the resonance phenomenon is modes are seen to exhibit alternate resonance peaks.
seen to lose importance. Resonance marks although still The consideration of length irregularities makes both
visible almost disappear diluted with the high attenua- aerial modes to resonate at every critical frequency (al-
tion background. though with different magnitudes).
Lastly, the question about the influence of conduc-
3.2 The Influence of Geometrical Parameters tor’s geometrical arrangement on the modal propagation
The preceding results were obtained by considering constants of the transposed-line structure is addressed.
a strictly regular transposition cycle with L = 300 km Assuming a regular transposition scheme, with wire
and lA = lB = lC.. Now, we analyze the influence of trans- and ground losses accounted, we computed the modal
position cycle length irregularities. attenuation and modal phase velocity characteristics for
Using a random number generator we defined dif- the horizontal, vertical and triangular line configurations
ferent values for lA, lB , and lC, with relative deviations to depicted in Fig. 2. Results obtained, considering a fre-
their mean value L/3 as high as ± 15%. Results regard- quency sweep from f = 400 Hz to f = 1.1 kHz, are
ing the modal attenuation and modal phase velocity shown in Fig. 5. Comparison established leads to the
characteristics for regular and irregular transposition following conclusion:
cycles are compared in Fig. 4 (horizontal line configu- w While the horizontal and vertical line configura-
ration with global losses included). For analysis pur- tions have very identical characteristics, the triangular
poses a frequency scan comprising five resonant peaks line, with a naturally more balanced geometry, is seen to
from f = 400 Hz to f = 2.6 kHz was considered. be almost immune to the resonance effects due to con-
Results obtained lead to the following conclusions ductors transposition.
w For strictly regular transposition schemes, 3rd or-
der resonance peaks (1.5 kHz) are absent —indeed,
4 CONCLUSIONS
every mth order resonance with m a multiple of 3 van-
ishes. For balancing purposes many overhead power lines
w When length irregularities are taken into account have their phase conductors periodically transposed
resonance peaks of order 3, 6, 9, etc, show up. However, along the line length. The modal propagation constants
the magnitude of those peaks is rather small (while for of the overall line structure are ordinarily evaluated us-
the aerial modes a small disturbance at 1.5 kHz may still ing a simple homogenization technique based on aver-
be detected, for the ground mode the effect is absolutely age series-impedance matrices and average shunt-
negligible). admittance matrices.
14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 5
REGULAR TRANSPOSITION CYCLE IRREGULAR TRANSPOSITION CYCLE

g g

a“
a’ a“ a’ a“ a’ a“ a’ a’

a’ a’ a’ a’ a’
a“ a“ a“ a“

g g

Figure 4: Plot of the attenuation and phase velocity characteristics against frequency for the horizontal line configuration with total
losses included. Resonance effects are displayed for two cases: regular and irregular transposition cycles. Dotted lines represent re-
sults obtained using the standard homogenization technique. Labels g, a’ and a” have the same meaning as in Fig. 3.

HORIZONTAL LINE VERTICAL LINE TRIANGULAR LINE

g g g

a’ a“ a’ a“

a’ a“

a’ a’
a’ a“
a“ a“

g g
g

Figure 5: Plot of the attenuation and phase velocity characteristics against frequency for the horizontal, vertical and triangular line
configurations, assuming regular transposition cycles, and considering all losses included. Dotted lines represent results obtained
using the standard homogenization technique. The ground mode is identified by label g, whereas the aerial modes are identified by
a’ and a”.
14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 6
Albeit very appealing this procedure is not a rigor- [4] J. Faria, “High-Frequency Modal Analysis of
ous one as it cannot adequately account for the wave Lossy Nonuniform Three-Phase Overhead Lines
interference phenomena produced by conductor inter- Taking into Account the Catenary Effect”, Europ.
connections occurring at discrete transposition points. Transactions on Electrical Power, vol. 11, pp. 195-
Making use of matrix methods for nonuniform mul- 201, May/June 2001
ticonductor line structures, conductor transposition ef- [5] J. Faria, “On the Realizability of Balanced Un-
fects in single-circuit three-phase transmission lines transposed Overhead Line Configurations”, Euro-
have been analyzed in this work. Modal attenuation and
pean Trans. Electrical Power, vol. 8, pp. 451-454,
modal phase velocity characteristics against the fre-
Nov/Dec. 1998
quency have been determined for several transmission
line configurations, the influence of cycle irregularities [6] R. Galloway, W. Shorrocks and L. Wedepohl,
and the influence of line losses being examined as well. “Calculation of Electrical Parameters for Short and
The main conclusions of this research can be sum- Long Polyphase Transmission Lines”, Proc. IEE,
marized as follows. Results obtained show that con- vol. 111, pp. 2051-2059, Dec. 1964
ductor transposition actually gives rise to repetitive [7] C. Gary, “Approche Complète de la Propagation
resonance effects at certain critical frequencies —fre- Multifilaire en Haute Frequence par Utilisation des
quencies for which the overall transposition cycle length Matrices Complexes”, EDF Bulletin de la Direc-
approaches an integer multiple of one half wavelength tion des Études et Recherches, no. 3/4, pp. 5-20,
(e.g., for a cycle 300 km long, resonance peaks of differ- 1976
ent magnitude can be seen to occur at 0.5 kHz, 1 kHz,
1.5 kHz, etc). The ground and aerial modes of the trans- [8] J. B. Faria, “Multiconductor Transmission-Line
posed line structure exhibit peculiar resonant behaviors; Structures”, New York, Wiley, 1993, ISBN 0-471-
for real lossy transmission lines, the ground mode does 57443-0
not show significant resonance marks, however, the [9] C. Paul, “Analysis of Multiconductor Transmis-
aerial modes are highly affected (observed peak values sion Lines”, New York, Wiley, 1994, ISBN 0-471-
for the attenuation are about 4 or 5 times greater than 02080-X
predicted by the standard homogenization technique).
Resonance effects due to conductor transposition are not [10] C. Dubanton, “Calcul Approché des Paramètres
independent of the line geometry; while equally impor- Primaires et Secondaires d’une Ligne de Trans-
tant for horizontal or vertical lines, resonance effects port”, EDF Bulletin de la Direction des Études et
turn out to be practically negligible for triangular lines. Recherches, no. 1, pp. 53-62, 1969
Results herein obtained should find special applica- [11] J. Faria and J. Silva, “Irregular Eigenvalues in the
tion in all areas concerned with the propagation of Analysis of Multimodal Propagation”, 8th PSCC
waves along power lines, namely, PLC communica- Proceedings, pp. 760-764, Aug. 1984
tions, EMI, and line transients.
The evaluation of the resonance effects associated
with forward and backward surge impedances is a sub- APPENDIX I
ject that has not been covered here (a subject that will Evaluation of L and Zg matrices
involve the analysis of the eigenvectors of the global The per-unit-length external-inductance matrix L is
transmission matrix T). Further work on such a topic is a frequency-independent real symmetric non-singular
still required. matrix whose elements are evaluated according to

REFERENCES µ o 2hk
Lkk = ln
[1] N. Boulejfen, A. Kouki and F. Ghannouchi, “Fre- 2π r
quency- and Time-Domain Analyses of Nonuni- µ o (hi + h k )2 + ( y i − y k )2
form Lossy Coupled Transmission Lines with Lin- Lik = ln
4π (h − h )2 + ( y − y )2
ear and Nonlinear Terminations”, IEEE Trans. Mi- i k i k
crowave Theory Tech, vol. 48, pp. 367-379, March
2000 where r denotes phase conductor radius, and hk and yk
denote, respectively, the vertical and horizontal coordi-
[2] G. Xiao, K. Yashiro, N. Guan and S. Ohkawa, “A nates of conductor k.
New Numerical Method for Synthesis of Arbitrar- The per-unit-length ground return impedance matrix
ily Terminated Lossless Nonuniform Transmission
Zg, is a frequency-dependent complex symmetric matrix
Lines”, IEEE Trans. Microwave Theory Tech, vol.
49, pp. 369-376, Feb. 2001 whose elements can be determined using either Carson's
approach or Dubanton's complex ground plane ap-
[3] M. T. Barros and M. Almeida, “Computation of proach, the latter yielding [10]
Electromagnetic Transients on Nonuniform Trans-
mission Lines”, IEEE Trans. Power Delivery, vol.
11, pp. 1082-1091, April 1996 (Z g )kk = jω µ2πo ln1 + hp 
 k 
14th PSCC, Sevilla, 24-28 June 2002 Session 35, Paper 1, Page 7
(h + hk + 2 p )2 + ( y i − y k )2
( )
Zg
µ
= jω o ln i
ik 4π (hi + hk )2 + ( y i − y k )2
where p —the so-called complex depth— simultane-
ously depends on the ground resistivity ρ and on the
frequency.
ρ
p=
jωµ o

APPENDIX II
Evaluation of the square-root matrix S
Let Z and Y denote the series-impedance and shunt-
admittance matrices of a given uniform line section.
Their product ZY is a non-symmetric square complex
matrix.
If, as it usually happens, ZY is a simple structure
matrix [8] then ZY will be brought into diagonal form
by employing a non-singular similarity transformation
defined by the modal matrix M

M −1 ZY M = diag{γ 12 , γ 22 , γ 32 }

The eigenvectors of ZY are the columns of M, the

associated eigenvalues are the squared modal propaga-
tion constants.
The square-root matrix S is obtained as

S = SQRT (ZY ) = M diag{γ 1 , γ 2 , γ 3 } M −1

If, however, ZY happens to exhibit irregular eigen-

values [11], a non-singular modal transformation matrix
will never be found. In such a case the computation of S
will require a different more elaborated strategy making
use of Jordan normal forms [8].

ACKNOWLEDGMENT
This research was carried out within the framework
of the PRAXIS XXI Program under the sponsorship of
the FCT — Fundação para a Ciência e a Tecnologia
whose support the authors gratefully acknowledge.