ABSTRACT A unified approach for calculatzng the low and Meliopoulos[l3],Velazquez[25] ) proposed timedomain solu
high frequency, and the transient performance of a network of tions to transient problems based on transmission line theory.
grounding conductors is given. Results which are derived first in This approach, however, is strictly valid for long horizontal con
the frequency domain, are converted to the time domain by Fast ductors and is not suited for the analysis of vertical conductors
Fourier Transform algorithms. A grounding system is modeled and arbitrary interconnected conductors. Refer to the work of
by an ensemble of thinwire conductor segments. The computer Pluniey[l9] for an interesting discussion of this topic. Another
model used to determine the electric fields and current distribu publication by Takashima[24] accounted for the effects of soil
tion in and around a grounding system is described and its main permittivity but ignored induction effects. A recent publica
limitations discussed. tion by Papalexopoulos[l8] examined the transient problem
KEYWORDS: Electromagnetic Fields, Transient perfor using a timedomain approach but restricted the analysis to fre
mance, Time Harmonic Performance, Computer Modeling, quencies low enough to satisfy the condition that the integration
Grounding. path dependent component of the electrical field is negligible (in
practice for frequencies on the order of 0.1 to 0.5 MHz depend
ing on the size of the substation grounding). Research work
1 Introduction by Dawalibi[G,i]analyzed the response of practical grounding
There is a massive body of literature describing the performance grids using electromagnetic field theory in the frequency domain.
of wire conductors energized at low and high frequencies in space, Although the solution was formulated for low and high frequen
air or soil. It is impossible to list all major research contributions cies, the analytical expressions used in the computer model were
in this research field. T h e reader is referred to the national derived based on the quasistatic assumption to avoid direct nu
and international journals devoted to the subject of antennas merical integration of the general solution. Later Grcev devel
oped in his P h D. thesis [8] the methodology for the analysis of
and propagation, geophysical prospecting and electromagnetic
compatibility for an overview of the recent advances in this field. arbitrary grounding conductors at all frequencies based on com
putational techniques involving the method of moments, numer
However useful materials pertinent to this research paper can be
found in the References section. ical integration of Soinmerfeld integrals and Fast Fourier Trans
forms ( F F T ) in a way similar t o the approach used by Burke
Most of the cited work however has concentrated on t,he
[14,5]. T h e initial computer model which resulted from this work
response of metallic structures excited by a voltage applied
was specialized to horizontal linear conductors (Grcev[S]).
to two closely spaced (but isolated) feed point terminals lo
This paper summarizes the followon research work which ex
cated on the energized structures or by an incident electromag
tends the computer model to a network of interconnected ground
netic field illuminating the structures (Adams[2], Burke[5],
conductors and discusses the advantages and limitations of the
Miller[l4],Mitra[l5,16], Olsen[l7]). Another important por
model. Detailed numerical results for practical grounding sys
tion of the work specializes in electromagnetic geophysical
tems will be presented in a future publication.
prospecting where some methods use short rods immersed in
a dissipative medium (Wait[26]).
There are only a few published research works involving high 2 Frequency Domain Analysis
frequency performance of power system grounding structures.
The transient problem is first solved by a formulation in the fre
Pioneering work was conducted by Sunde in the late thirties. His
quency domain. T h e timedomain response is then obtained by
work culminated in the publication of his well known reference
application of a suitable Fourier inversion technique. T h e re
book (Sunde[23]). Recently, several papers (Mazzettie[l2],
sponse to a steadystate, time harmonic excitation is computed
for a wide range of frequencies starting at OHz. From this fre
90 lili.I 1305 PiRD A paper recommended and approved quency response, a transfer function is constructed for every fre
by t h e IEEE S u b s t a t i o n s Committee of t h e IEEE Power quency considered. T h e transfer function is dependent only on
E n g i n e e r i n g S o c i e t y f o r p r e s e n t a t i o n a t t h e IEEE/PES the geometric and electromagnetic properties of the grounding
1990 Winter Meeting, A t l a n t a , Georgia, F e b r u a r y system and its environment.

4 8, 1990. Manuscript s u b m i t t e d August 27, 1989;
made a v a i l a b l e f o r p r i n t i n g December 27, 1989.
If i ( t )represents the injected current at a point in the ground
ing system, and r ( t ) denotes an observed response, then
z ( t ) = 3 { W ( j W ) . F[Z(t)]}
CO, PO Air
Grounding System
General Coordinate
System
where F and Fare the Fourier and inverse Fourier transforms, a scat,tered electric field E. See Harrington[lO] for a. detailed
respectively, and w is the angular frequency and where j = R. description of this analytical approa.ch. The socalled sca.t tered
Since the Fourier transform is straightforwardly computed field E is the results of the currents a.nd charges induced in the
using F F T algorithms [16], the major part of t,he problem is to grounding system by the impressed electric field E
compute the transfer function V V ( j w ) . The expression for the induced currents is derived by stat
The physical model is based on the following assumptions. ing the boundary conditions for t,he tangential component of
the electric field at the surface of the conductor segment,s. The
The earth and the air media. are homogeneous and occupy
boundary condition at the surface of a lossless conductor seg
halfspaces with a common horizontal plane boundary be
ment imposes the following equation. Note that because of t,hr
tween them.
thinwire approximation, the equation requires only t,he axial
The earth and the grounding electrodes exhibit linear and component of the electric field at the conductor surfa.ce.
isotropic, arbitrary characteristics.
t . (E + E ) = 0 (2)
The grounding system is assumed to be made of cylindrical
metallic conductors with arbitrary orienta.tion. However, The sca.ttered electric field E produced by the axial elec
they are assumed to be subject to the thinwire approxi tric currents l e and cha.rges g t in t,he grounding system can be
mation, i.e., the ratio of the length of the conductor seg expressed in terms of ret,arded pot,ential integrals as follows. Ap
ment to its radius is ,> 1 . In practice a ratio of about 10 is pendix A provides a brief description of the steps involved in the
satisfactory. See Adams[l] for a detailed discussion of the derivation (Mitra[l5]):
implications of the thinwire approximation.
E
Energization occurs by the injection of a current impulse
of arbitrary sha.pe produced by an ideal current generator
with one terminal connected to the grounding system, and
the other to the ground at, infinity. The influence of t,he
connecting leads is ignored.
The subscripts 0 and 1 in Figure 1 and in the following text refer
to air and soil cha.racteristics, respectively.


1
(VV .  7 ; )
4*J.<, 1 t . It(r)gl(r,r)dt
The first step in the development of the mathematical model is 7: = w2p1c1 and c1 = + $.
to determine the current distribution in the grounding system
for a timeharmonic excitation for a number of appropriate fre Here f is the frequency measured in Hz. The conductance
quencies. This task is conducted in two steps. First the solution (S/m), permittivity ( F / m ) and permeabilit,y ( H / m ) of the
for the unbounded conducting medium case is obtained, then medium are denoted by U ] , 1 and respect,ively. T h e various
the effects of the airsoil interface is accounted for in the final geometric quantities are defined in Figure 2. Not,e t,hat g l ( r , r)
expression as explained hereafter. stands for the Greens function for an unbounded medium
(Miller[l4]). Also note that t,he timevariation exp(Jb1f) has
been suppressed throughout this paper.
3.1 Unbounded Conducting Medium The required mathematical model follows from the substitu
In this Section, we start first with lossless conductors. Then the tion of ( 3 ) into ( 2 ) . The result of this substitut,ion[l4] yields the
lossy case is introduced (See equations ( 5 ) and (6)). electric field integral equation
Figure 2 illustrates two conductor segments of the grounding
systems under consideration. T h e position of a. point on t,he a.xis
of a segment is defined by a vector r and that of a point on the
surface S of a segment is defined as vector r. The vect.or t rep where G1 ( r .r) is the dyadic Greens function for the electric
resents the unit vector along the segment axis and t represents field at r due to a current element at r. Here
the axial unit vector tangential to the segment surface.
In t,he general case the total electric field E in t,he ground can
be considered a.s the sum of an impressed electric field Et and
1775
The thinwire approximation does not imply that the finite
thickness of the conductor segments is neglected since the bound
ary conditions are applied a t the surface of each segment. Fur
ther, although the radial currents in the conductors are ignored,
Zt
the model does not neglect the radial leakage currents t o the
ground along each segment.
In the case of lossy conductors, the boundary conditions (2)
need to be modified. In this case the tangential component of the
total electric field at the surface of a conductor segment equals
the internal voltage drop per unit length along the segment.
T h a t is,
t . (Ez+ E) = 1 . Z, (5)
on the conductor surface S. Here Z, is the surface internal
impedance. Let u2, 2 and pz represent the coininon conduc X
tance, permittivity and permeability of the conductors. We have
(Stratton[l]): General Coordinate
System
I ( l )= Ce .Pt(4 (9)
where RF . . . R; are given in Appendix B, and where x,y, z and
,=1
p, #I and z are unit coordinate vectors for the rectangular and for some coefficients c,. In this work, the M = N piecewise
cylindrical coordinate systems, respectively. T h e superscripts H
consta.nt basis functions
and V indicate the contributions of the aforementioned horizon
tal and vertical dipoles, respectively. 1 on the ith segment; a,nd
P,(4 = 0 otherwise
T h e correction field components and also the complete dipole
fields can be derived in a rather straightforward manner from
are selected. They give a piecewise constant representatioll of
Maxwells equations subject to the boundary conditions at t h e
the current distribution in which
surface of the ground (Sommerfeld[22]). T h e correction field
components are given in Appendix B.
Equations (4), ( 7 ) and (8) yield the mathematical model.
c,= { I,
0
on the ith segment; and
otherwise.
1776
When the lossy nature of the conductors is taken into account,
only the main diagonal elements, which represent the general
ized selfimpedances of the segments, have t,o be modified. T h e
I(L)  Stepwise Approximation diagonal elements become:
t
I, =II
k/ R e d Current
Distribution
zh, = z,,, + Z,AE, (13)
path of integration. Therefore, a t the higher frequencies range, 0.05 0.01 0. I t@s)
it is extremely important to define the test circuit used for the
voltage measurement.
Figure 5: Influence of F, o n Transient Impedance Conipii
tat ions.
6 TimeDomain Solutions
Once the transfer function M ; ( g w ) have been determined for This leads to a significant.ly higher frequency than t,lia.t.ba.sed
each calculated quantity, for example the electric field or current on the usual approach. T h e maximum frequency F,, is usn
at specified points, the timedomain solutions can be obtained ally determined as the frequency above which the amplitudes
by direct application of (I). For example, in the case of the in the recorded spectrum are negligible (Mazzettie[l2]). For
impedance, the timedomain response of the voltage a t the feed instance, the maximum frequency F , may be specified by the
point is evaluated using (1) and the transient impedance follows requirement that 99% of a wave energy lies in the frequency
using band below F,. This is clearly illustrated in Figure 5. which
shows a lossless cylindrical conductor in an unbounded conduct
ing medium. T h e conductor is energized by the injection of
The Fourier transform of the excitation current z ( t ) can be a doubleexponentialcurrent wave (Grcev[9], Mazzettie[121,
Meliopoulos[l3]). T h e value of Fm which captures 99%#of the
obtained analytically for simple cases, but in general numerical
energy of I ( J w )is GOO KHz a.nd 7 MHz for V ( J W(Grcev[8]).
)
integration is required.
However, even F, = 10 MHz (curve 1) does not allow the evalu
The calculation of the inverse Fourier transform is carried out
ation of the maximum of the impulse impedmce. This value or
by a Fast Fourier Transform algorithm (Brigham[4]) which is
curs at At % 0.04ps and therefore F, % 25MHz. Here F,, = 20
wellsuited for the evaluation of the timedomain responses. This
MHz (curve 2) leads to a.pproxima.te results for early times while
technique however is approximate because it involves bandlim
F , = 40 M H z (curve 3 ) yields the most accurate results for early
iting (windowing) of functions and requires regular sampling of
times. All curves give accurate results for t > 0 . 0 3 ~ s .
continuous functions which are generally of rather long dura
tions and infinite spectrums. It is thus necessary t o determine
the proper density of sampling and the Nyquist (maximum) fre 7 Computer Model
quency in all calculations of a transfer function W ( j w ) . T h e
number of samples is minimized by application of an adaptive The flowchart for the computer model is shown in Figure 6.
interpolation algorithm (Grcev[S]). Essentially, the transfer function is approximated by frequency
sampling in a loop around Modules 1, 2 and 3. The frequency
and then the timedomain electromagnetic field generated by
6.1 Nyquist Frequency single or multiple injections of current impulses into the ground
ing network is computed in Module 4. Further details are now
T h e Nyquist frequency and the number of samples is determined given.
by the accuracy required for the representation of the tiine
domain response at early times of the current excitation (for
example during the calculation of responses t o current waves 7.1 Structure of the Computer Model
with steep fronts). The Nyquist frequency range is given by
To approxima.te the tra.nsfer function, the frequencydomain re
(Brigham[4])
sponse of the network to single frequency, timeharmonic input
1
Fm = current,s is found. T h e spacing A bet,ween t,hese sa.mpling frr
quencies is implicitly defined by Nyquist relation F, = &
where At denotes the earliest time for which accurate informa where F , is the frequency bound discussed in Section 6.1. For
tion on a response is desirable (assuming that the excitation each single sa.mpling frequency, the loop computa.tions are as
current is also known at that time). follows.
If the time step At is specified in advance, the Nyquist fre
quency (18) yields the maximum frequency which has to be ac Module 1 . Preliminaries. T h e Soinmerfeld integrals a.nd t,he
counted for in the frequency spectrum. correction terms given in Appendix B are computed for
1778
7.2 Computation Time
T h e ma.jor disadvantage of this computer model lies in the sig
nificant computation time required to numerically evaluate the
va.rious integrals involved and t o solve the simultaneous 1inea.r
equations (11) for the current distribution. For insta.nce, this
evalua.tion for a grounding grid broken down into N z 300 seg
inent,s requires over 30 hours of CPU on a.n a.rra.y processor at,
tached to the QBus of a VAX (with a MAP4000 army proces
sor running a t 40 MFLOPS peak) for a timedoma.in solution
Som..rt.ld Intaqrals corresponding to a 30 MHz frequency spectrum.
A major reduction of this computation time can be achieved
by applying whenever possible, for instance a t low frequency and
E".l".tion
El.6tromaqn.tIc
Of
Field
!I
I
1
for short distances between a segment and observation points.
formal solutions of the Sommerfeld integrals (Dawalibi[G,7]).
Furthermore, code optimization and use of asympt,ot,icexpan
sions valid for particu1a.r combinations of media characteristics
can be also employed to reduce computation t,imes (Wait[26]).
This work is presently in progress and significant achievements
a t Prmscribed (i.e., reductions of comput,er time by a factor of 5 to 10) will be
0bs.rv.tIon Points
report,ed in fut,ure publications.
8 Numerical Results
Detailed numerical analysis for a variety of practical ground
ing systems will be the subject of a forthcoming paper. Note
that various numerical results for simple electrodes (in both the
frequency and time domains) can be found in the previous pub
lications by the first author (Grcev[S,S]).
9 Conclusions
The development and application of a. computer model for a.n
alyzing the transient performance of grounding systems ba.sed
on electromagnetic field theory has been described. T h e use
Figure 6: Computation Flowchart. of a combination of numerical int,egra.tion techniques, met,hod
of moments, adaptive interpolation and Fast Fourier Tra.nsforln
constitutes the basis for the computation of various physical
points on a grid in the Cartesian coordinates system. All quantities such as the electric fields in the ground, longitud
points of interest in the problem are within the confines nal and leakage currents in the ground conductors and ground
of the grid. The computed values are stored for later use. impedances. It is seen that t,he analysis of conductforsenergized
The mesh size of the grid is chosen sufficiently small, so by currents waves can require computations a t frequencies higher
that further values of computed quantities can be found by than the significant frequencies in the spectrum of the excitat,ion
interpolation between the grid points. signal while simpler models may fail to predict accurately the
transient performance.
Module 2. Current Dzstrzbutzon. First the values of correction
terms are interpolated in order to evaluate the integrals in T h e main limitation of t,his computer model is attributed to
the computer time required for the analysis of large or complex
(12). This yields the generalized impedances from which
grounding syst,ems. However, research work by t,he authors is
the current distribution for the singlefrequency input is
presently conducted in order t o reduce computer time sufficiently
determined by (11).
to enable its use on general purpose computers.
Module 3. Electroaagnetzc Fzeld. The electromagnetic field
generated by the single input frequency is computed a t
points (field profiles) selected by the user and also along an Acknowledgments
appropriate path when the grounding system impedance is
requested. T h e research work and the development of the extended coin
puter model was supported by Safe Engineering Services RL tech
Repetition of the loop across the various frequencies completes nologies Itd (SES). The authors would like t,o t,hank Dr. Alan
the computation, or rather more precisely, the sampling of the Selby of SES for his constructive comments a.nd hlr. Walt,rr
transfer function. Tavcar of SES for his assista.nce in t,he final software develop
Finally in Module 4, the frequency domain representation of ment, and in the prepara.t,ion of the manuscript. Finally, t.he
the computed electromagnetic field responses are then converted first author would like t o thank Professor Z. Ha.2nada.r of the
to the timedomain by application of the Fast Fourier Transform Universit,y of Zagreb, Yugoshvia, for his continuous support.
to each component of interest.
1779
Appendix A in terms of the fields of a vertical and a horizontal current ele
ment. These are shown in Figure 3, along the zaxis and xaxis,
The scattered electric field E(r) a t a point r produced by a
respectively. T h e components of the electric field are as follows:
current t.I(r)and a charge ue(r)can be expressed in terms of
the magnetic vector potential A(r) and electric scalar potential
4(r) as follows:
E(r)= jwA(r)  Vd(r) (19)
The magnetic vector potential A(r) and the electric scalar
potential 4(r) are given by the retarded potential int,egrals
(29)
1 (34)
Es(r) = /It . I(r) . V] [Vgl(r,r)
4*.iwC, e
(35)
de
 y?tI(r)gI(r,r)] (26)
Appendix B
Sommerfelds solution for the electromagnetic fields due t o an
infinitesimal current element radiating in the presence of a con where CI = &
ducting halfspace is based on Maxwells equations subject t o 1