TRAVELING WAVES ON ELECTRIC POWER SYSTEMS

L. V. BEWLEY

The problems of traveling waves on the transmission lines of a

power system 1 differ considerably from those of traveling waves on
telephone or telegraph circuits. The primary object in the case of the
former is to know how to protect the system from abnormal voltage
disturbances which might damage apparatus or cause discontinuity
of service ; whereas the object in the case of the latter is the transmission of signals. Attenuation, distortion, wave shape modification, and
successive reflections are deliberately sought after on the power system as a means of rendering the surges innocuous, but these effects
must be carefully avoided or nullified on the communication circuits
so as to preserve the wave shape and transmit the signal with strength,
fidelity, and without interference. On the power lines, the surges often
originate from unknown causes, or at the point of origin are of unknown magnitude and shape (except from a statistical point of view) ;
while on the communication circuits the initial shape and magnitude
of the wave train are known with exactness. External fields (due to
charged clouds), corona, flashovers, faults, and so on are of great
importance with respect to surges ; but are of no concern in the normal
functioning of a telephone or telegraph line. Thus on the power lines
surges originate from external or undesirable causes and every effort
must be made to withstand or control them; while on communication circuits the transients are the direct means to the end.
These differences have led to corresponding differences in the
mathematical approach. The power engineer is satisfied with approximations which would be intolerable to the communications
engineer, and he is willing to take a license with mathematical rigor
which would make any self-respecting mathematician groan. Higher
mathematics has found little or no application in the study of surges
on power systems. This has been due primarily to the fact that the
boundary conditions are not definite enough to justify purely mathematical refinements; particularly since engineering results must be
obtained in a short time by men who are not mathematicians. However, there are numerous aspects of the problem which lend themselves to mathematical excursions. It is my purpose in this lecture to
An address delivered before the meeting of the Society in Bethlehem, Pa., on
December 30,1941, by invitation of the Program Committee under the title The mathematical theory of traveling waves-, received by the editors January 10, 1942.
1
Traveling Waves on Transmission Systems, L. V. Bewley, New York, Wiley.

527

528

tAugust

L. V. BEWLEY

give you the engineering methods, and perhaps, here and there, to
suggest how advanced mathematics might be applied.
The general problem. Figure 1 indicates a multi-conductor transmission line, protected by a ground wire2 at the peak of the towers,
terminating at a power station containing a lightning arrester, a circuit breaker, a transformer, and a generator. A cloud, floating over
the line, has been charged by the action of the rising air currents on

tower ^^.<>
footing -""^ <
resistance "="

breaker

arrestor

transformer

FIG. 1

the falling raindrops, and this cloud charge has induced a stationary
counter-charge on the conductors of the transmission line. As the
potential of the cloud increases, local breakdowns occur throughout
its mass, thereby uniting some of its regions and making available,
through partially ionized paths, a reservoir of charge. Eventually the
field gradient reaches an intensity sufficient to initiate a leader stroke
or dart, which starts towards ground. The progress of this dart is not
continuous, but by jerks, each jerk depending upon the supply of
additional charge to the head of the dart. It is like the armored force
in a blitzkrieg break-throughit strikes to the limit of its capabilities
and must then wait for tactical reinforcements and logistical support
before renewing its attack. As the dart approaches earth, the field
2
The purpose of a ground wire is to intercept a lightning stroke and thereby protect the power conductors from destructive surges. Those surges which do come in to
the station are absorbed by the lightning arrestor.

1942]

TRAVELING WAVES

529

gradient at the transmission line increases and this causes a migration of charge up through the towers onto the ground wire, and from
the remote parts on the line conductors towards the region of field
concentration. The dart finally makes contactsay with the ground
wire at the towerand a lightning surge moves out in both directions
on the ground wire, inducing waves on the line conductors. But when
these waves reach the next tower, reflections occur, and very soon all
the neighboring spans are filled by numerous waves reflecting back
and forth, and perhaps flashovers have taken place to the line conductors. These waves are rushing, with the speed of light, towards
the power station, where they may enter the windings of transformers
and generators, causing steep gradients which may breakdown the
turn-to-turn insulation, and oscillations which may develop destructive voltages on the major insulation to ground. Perhaps a bushing
flashover, or an insulation failure will cause the circuit-breaker to
function, interrupting the normal 60 cycle power current, and this
operation will initiate a new transient which is called a "switching
surge/' And perhaps an insulator flashover out on the line will culminate in intermittent arcing which may result in a cumulative building-up of dangerous voltagesthe so-called "arcing ground."
The engineer does not tackle this problem in its entirety by attempting to include all the terminal apparatus as boundary conditions. Rather is he compelled to make a piecemeal attack, by
handling each part of the problem as a separate and independent
proposition, thereby isolating and defeating them in detail. To this
end, certain aspects of the problem will be considered under four
main headings.
Multi-velocity waves (tensor notation). Consider a system of n
overhead transmission line conductors with voltages er currents ir,
charges Qry and fluxes <f>r. Then in terms of Maxwell's electrostatic
potential coefficients prs
(1)

er = prsQ'

from which
(2)

Qr =

K"ea

where Krs is the inverse of prs in the matrix [prs]- These coefficients
are calculated for parallel cylindrical conductors in the presence of
ground by including the images of the conductors in the ground
surface.

530

L. V. BEWLEY

[August

The magnetic flux linkages are given in terms of the inductance

coefficients by
(3)

</>r =

Lr8i*.

There are leakage currents flowing to ground and between conductors of amount Grses. The Grs coefficients are supposed to include
the effects of both leakage and corona.
And finally, there are resistance drops Rrsi8 in the conductors due
to the flow of currents.
Herefrom, the differential equations for the multi-conductor transmission system become (putting pd/dt, in the Heaviside sense)
der
(4)

dx
dil
(5)

+ Rrti1

fa

= (Lrtp

Rrt)i*>

= pQl + Gtses = (Ktsp + Gts)es.

dx
Eliminating il there results
(6)

( Jr - r
) es = 0
\
dx2/

in which sr is the Kronecker delta and

fr = (Lrtp + Rrt)(KtSp + G*').

(7)

Now if the losses are ignored (Rra = 0, Grs 0), (6) is satisfied by the
traveling wave 3
es = asa (>(* >()0

(8)

which substituted in (6) gives

/

ts 2

s d2 \

() m

(9)

( LrtK p - r
) as f{a)(x - v(a)t) = 0.
\
dx2/
Since this equation must be satisfied for waves of the same velocity
(10)

(LrtK"v\a) - l)ala) = 0.

The velocities are given by the roots of the determinant

An enclosed index is used here to suspend the summation convention.

531

TRAVELING WAVES

1942]

(11)

I LrtKfl() r I = \Cr\ = 0 .

To each root of (11) there correspond n values of ala\ and any (n 1)

of them may be determined in terms of one value taken arbitrarily.
Let this one value be eft* = 1. Solving (10)
as

(12)

,f- (LrtK V(a))

where s ^ 1, r 5* 1,

in which | b'rs\ is | c'rs\ with the 5 = 1 row and r = 1 column deleted, and
Ars is the cofactor of bsr in | b'sr\
The complete solution then becomes
(13)

er =

ar[fa(x

d(a)t) +

Fa(x

V{a)t)].

By (5) the corresponding currents are

(14)

- p i

Krsesdx = KSaas{fa - Fa)via) = F m (/ - Fa).

Transition points. Consider the general case of Figure 2 in which

any number of incoming lines terminate a t a transition point consisting of an inter-connected network and any number of outgoing
incident
waves ^

transmitted
waves

reflected
waves

*'*XJ

(r,s)

-^

(j,k)

(u,v)

^ * outgoing lines

incoming lines,
transition
network
FIG.

lines. When the incident waves on the incoming lines reach the transition point, currents will flow into the network, transmitted waves
will move out on the outgoing lines, and reflected waves will start
back on the incoming lines. In tensor notation let:
zrs = surge impedances of incoming lines,4
4

The surge impedance of a line is the coefficient of proportionality between its

voltage and current {ezi). Its reciprocal is called surge admittance. For surges on
transmission lines, these parameters are essentially constant, and are an indication of
the current associated with a surge of given voltage.

532

L. V. BEWLEY

[August

yrs s u r g e admittances of incoming lines (inverse of zrs),

zuv =surge impedances of outgoing lines,
yuv = s u r g e impedances of outgoing lines (inverse of zUv),
Zjk= branch impedances of the network,
C' = transformation tensor specifying the total interconnections
of the network and outgoing lines.
Then
(1)
(2)

Za$ =(Zjk+Zuv) = impedance of network and outgoing lines before interconnection,

Za/0' = CZ>Cp>Zap = impedance after interconnection of the network and outgoing lines.

Now Za>p> may include branches other than those connected to the
incoming lines. The open circuit branches will have been eliminated
by C', but the branches other than those connected to the incoming
lines will have to be eliminated by the substitutions :
\OJ

Zja'fi'

\^/

J-^r

(5)

ZJTS

~\~ Zjrq

sL/rs*-

~|

"~T" ZJ pS ~t~
-*rql

pq,

0 = ZFSP + ZvqL\

from which
(6)

Er = (Zrs - ZrqYZp,)P
r

= Zr> .

Now let (er, i ) and (e'r, i' ) be the incident and reflected waves,
respectively, on the incoming line. Then at the transition point

(8)

i' + r = Ir,

so that
(9)

er + e'r = zUir

+ r) = ZrW%{fit ~ *')

from which
(10)

{K + Z'rsySt)et = ( - br + Z'rsySt)et.

This system of equations defines the reflected voltage waves er.

The total voltage at the transition point then follows by (7), the
total current by (8), the remaining network currents by (5) and the
network voltages by (4), and so on.
Successive reflections. The calculation of successive reflections is
oftentimes a long and involved process; particularly in those cases

1942]

TRAVELING WAVES

533

where reflections may occur from a whole series of neighboring junctions. An example is the case of lightning striking the ground wire at
midspan. The incident waves move out in opposite directions until
they reach the nearest towers, where they reflect as a consequence of
the surge impedance of the continuing ground wire in parallel with
that of the tower. Transmitted waves quickly reach the foot of the
tower from which they reflect as a result of the ground resistance.
Other waves reflect from the next tower, and from the next after
that, and so on. Thus within a few microseconds the system is alive
with a whole series of waves moving in different directions, arriving
at different times, of different magnitudes and polarity, and having
experienced different attenuations and distortions.
In order to keep track of all these components a lattice diagram has
been devised, such as shown in Figure 3 for the case of lightning
striking a ground wire at midspan. The progress of each wave component is easily followed as it slides downhill along its zig-zag path,
giving rise to reflections at each junction. Thus at any instant of time
the waves at all points on the line can be identified; or at.any point
on the line the time of arrival of each wave can be seen.
To construct such a reflection lattice it is first necessary to determined the reflection and refraction coefficients5 at each junction,
and to post these on the sketch of the system being studied, as has
been done on Figure 3. The coefficients are, in general, Heaviside
operators, such that when operating on an incident wave (regarded
as a time function reckoned from its instant of arrival at the particular
junction) they give the reflected or transmitted wave. Those shown
in Figure 3 are A, B, C, D, Bf, C', D'. The initial wave coming down
the lightning stroke of surge impedance 2Z0 (the 2 is occasioned by
the condition of symmetry which permits the amputation of the system to the left of the stroke) refracts onto the ground wire a wave
A-f(t) which moves on to the top of Tower 1, where it reflects a
portion AB' -f(t) back towards its origin, and transmits a portion
AB -f(t) to the next ground wire section and also down the tower.
When the wave reaches the foot of the tower it reflects a portion
ABD''(/) back up the tower. Likewise reflections return from the top
and foot of Tower 2 and from the towers beyond. Now each of these
reflections could be traced out independently on the lattice and all
waves fully accounted for. But the labor is great. The work can be
5
A "reflection coefficient" is the operator determined from equation (10) of the
previous section on transition points, which permits the calculation of a reflected wave
in terms of the incident wave. Likewise, the "refraction coefficient" gives the transmitted wave.

534

L. V. BEWLEY

[August

simplified by introducing the concept of "wave trains," and "retarder operators."

The system of waves reflected back on the ground wire due to the
arrival at a tower top of a wave (/') and the successive reflections
lightning stroke
B (transmits)

(reflects) B'

AB' f(t)

>A-f(t)

AB'-lit)
AB-f(t)

Tower 1

Tower 2

t>'
tower footing
resistance

~4r

ABD
ABD(DF)
ABD(D'F)2

ABD(D'F?

FIG.

up and down the tower is seen from the lattice to be a "wave train"
of Type I :
e' = '(/') + BD'Ff{t'

(1)

,r
m

= 5 WW

- 2 A) + BD'F(D'F')f(t'

BD'F _
+

~W~ W-W

-|

- 2Am)J

- 4A) +

535

TRAVELING WAVES

i 9 42]

in which <f>(m) is a "retarder operator" such that the time of arrival

of a wave to which it is applied is retarded by (2htn)> thus

(2)

*(*0 (*') = / ( * ' - 2 M .

The wave train transmitted on to the next tower by the incident

wave and its tower reflection is the Type II wave train
e" =

{D'F')-l-f{t' -

+ e' = B |~/(0 +

2hm)\

(3)

= BMW).
In terms of the wave trains of Types I and II the complete history
of the reflections can now be written down.
The initial wave transmitted by a lightning surge to midspan is
Af(t) which arrives at the first tower at time 0.5s and gives rise to the
Type I wave train of the first order.
(4)

ei = AB'amf(t-

0.5s).

This wave train arrives at midspan at time s and reflects therefrom as

(5)

e2 = AB'C'amf(t

- s).

When the reflected wave train arrives at the tower at time (1.5s)
it generates a new Type I wave train of the second order
ez = AB'Cfamanf{t

(6)

- 1.5s).

Continuing this process we find combinations of the form

in which products are to be interpreted as
CLmOLnOLp = [l + <*Z b~ ^(f) ] [l + dj^ J*-ty(n)] [l + <*
(7)

m l

= [1 + 3a J2 b ~ <t>{m) + 3a X) X ^

fl IEE^+ ^(m

m+w

"V(w + n)

+ n + p)].

Now in addition to those wave trains operating between midspan

and Tower 1, contributions eventually arrive from neighboring
towers. Thus at time (0.5s) there is transmitted beyond Tower 1,
owing to the initial wave Af(t), the Type II wave train of the first
order
(8)

Ei = ABfimf(t-

0.5s).

This wave train reaches Tower 2 at time (1.5s) where it generates

a Type I wave train of the second order

^'^(p)]

536

L. V. BEWLEY

[August

2 = ABBfpmanf(t-

(9)

l.Ss).

This wave train arrives at Tower 1 at time (2.5s) and generates a

third order wave train of Type I
(10)

3 = AB*B'(3manppf(t -

2.5s),

and this wave train, reaching midspan at (3s), reflects therefrom as

4 = AB2B'Cpmanprf(t

(11)

- 3s).

With the assistance of the lattice diagram and retarder operators,

the potential at any point can be written down. For example, at the
top of Tower 1
V = AB0mf(t - 0.5*) + ABB'C'ampnf(t
(12)

+ AB(BCy<xmanppf{t
+ AB*B'pmanppf(t
+

1.5s)

- 2.5s) +

- 2.5s) +

Induced lightning surges. Suppose a cloud bearing a charge o

is over a transmission line, Figure 4, and is discharging either to

Q(x)h

-G(x).
-j

bound
* charges

s W ^ W / V ^ ^ Nisv^Ov^v-A^
k^SSS^^^V^
FIG.

ground or to another cloud according to some function of time yp{t) ;

that is, the charge remaining in the cloud at any instant is Q0 [l \j/(t) ].
Depending upon the shape and size of the cloud charge, its height

194*3

TRAVELING WAVES

537

above ground, and its position with respect to the transmission line,
the line will experience a gradient
(1)

g(*,t)

=G(*)-[l-iK0]

in which G(x) represents the initial distribution of gradient (at the

beginning of cloud discharge) as a function of distance along the line.
Under the influence of this field, charges of opposite sign to that of
the cloud will leak over the insulators, or migrate from the remote
parts of the system, and collect on the line conductors as bound
charges. The density of bound charge at any point x will be proportional to the gradient and to the height h of the conductor above
ground (since the field is substantially uniform for a hundred feet
or so above ground). These bound charges nullify the potential due to
the external field, so that initially the line charges are given by
(2)

Vr = 0 - G(x) hr + prsQs,

r = 1, 2, , n.

Suppose then m of the n wires are ground wires perfectly grounded

throughout their length, and let these ideal ground wires be represented by (j, k) indices. The remaining (n m) wires are power conductors, and will be represented by (u, v) indices.
Now if the field gradient G(x) is suddenly removed, the bound
charges on the line wires will not change at the first instant, but those
on the ground wires are instantly replaced by new charges Q' since
the ground wires must remain at zero potential. Therefore
(3)
(4)

0 = PiiQ'h +
Vu = pukQ'k

Pi&\
puvQ\

From (2) and (3) all the charges Qs and Q'k may be found ; and from
(4) the potentials Vu may be determined. These potentials immediately move out as pairs of traveling waves (in opposite directions).
At the first instant, however, the forward and backward waves
fu(xvt) and Fu(x+vt) add up to the voltages given by (4), and the
resultant current flow must be zero in the isolated power conductors.
Thus
(5)
(6)

fu+Fu=Vu,
Y"(JV - Fv) = 0.

Hence

(7)

Fu = fu Vu/2,

538

L. V. BEWLEY

that is, the forward and backward waves on a conductor are the
same shape and magnitude.
However, the release of the bound charge is not instantaneous, but
according to the law of cloud discharge yp{t). The corresponding
traveling waves are then given by Duhamel's theorem

dMr)
\f[x + v(t - r)] + j[x - v(t - r)]} - dr
or
n

(9)

= lim X {fix + <n ~ k)At] + f[x - v(n At-+0

k)At]}tyk

where n-At=t, k-At=r and A\pk=\l/[(k + l)At] ^[ife-A/]. The application of the integral is limited to relatively simple expressions for and
\p, but the summation can be used for any functions whose graphs
are known or assumed. Ultimately, since both and \f/ derive from
experimental data, it is best to use the summation expression. Both
graphical and tabular methods have been devised for its ready application, and engineering solutions are quickly arrived at.
Equation (8) can also be derived by setting up the conditions in
terms of retarded potentials. Very few engineers deal with retarded
potentials, whereas a considerable number of them are familiar with
Duhamel's theorem through Heaviside's operational calculus.
L E H I G H UNIVERSITY