840

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 3, JULY 2002

Thin Wire Representation in Finite Difference

Time Domain Surge Simulation
Taku Noda, Member, IEEE, and Shigeru Yokoyama, Fellow, IEEE

AbstractSimulation of very fast surge phenomena in a

three-dimensional (3-D) structure requires a method based on
Maxwells equations, such as the finite difference time domain
(FDTD) method or the method of moments (MoM), because circuit-equation-based methods cannot handle the phenomena. This
paper presents a method of thin wire representation for the FDTD
method which is suitable for the 3-D surge simulation. The thin
wire representation is indispensable to simulate electromagnetic
surges on wires or steel flames of which the radius is smaller than a
discretized space step used in the FDTD simulation. Comparisons
between calculated and laboratory test results are presented to
show the accuracy of the proposed thin wire representation. The
development of a general surge analysis program based on the
FDTD method is also described in the present paper.
Index TermsElectromagnetic transient analysis, finite
difference time domain (FDTD) methods, Maxwell equations,
simulation, surges, wire.

I. INTRODUCTION

ONVENTIONAL surge problems have successfully

been solved by the circuit theory, where transmission
lines consisting of wires parallel to the earth surface are modeled by distributed-parameter circuit elements and the other
components by lumped-parameter circuit elements [1]. The
distributed-parameter circuit theory assumes the plane-wave
propagation that is a reasonable and accurate approximation
for the transmission lines, and this assumption enables the
handling of the electromagnetic wave propagation within the
circuit theory. On the other hand, very fast surge phenomena in
a three-dimensional (3-D) structure, which includes surge propagation in a transmission tower and in a tall building, cannot
be approximated by the plane-wave propagation. Thus, those
phenomena cannot be dealt with by the circuit theory but need
to be solved by Maxwells equations as an electromagnetic field
problem. Nowadays, the surge propagation in a transmission
tower needs to be analyzed for economical insulation design.
Furthermore, in a tall building, it is also important to assess
the interference of lightning surges with information devices
inside the building.
In order to solve the very fast surge phenomena in a 3-D
structure as an electromagnetic field problem, the finite difference time domain (FDTD) method [2], [3] and the method
of moments (MoM) [4], [5] are currently available as practical

choices. In surge simulations, accurate modeling of a thin wire

is necessary to represent transmission wires and steel frames of
a building. Furthermore, an imperfectly conducting medium is
required to be accurately modeled to represent currents in the
earth. Comparing the theories of FDTD and MoM, the former
is more advantageous to handle 3-D currents in an imperfectly
conducting medium such as earth soil without any difficulty,
even if the medium is nonhomogeneous. This is reported in detail in [6]. On the other hand, the latter is more advantageous to
accurately represent the thin wire.
This paper proposes a method to accurately represent the
thin wire in the FDTD simulation (thin wire is defined
as a conductive wire of which the radius is smaller than
the size of a discretized cell used in FDTD). The proposed
method corrects adjacent electric and magnetic fields along
a thin wire according to its radius taking into account the
discretization error of FDTD. This correction gives accurate
surge impedance of the thin wire, which is very important
for surge simulations. In this regard, the proposed method is
different from a method by Umashankar et al. [7] which is
used for the calculation of fields scattered by a thin wire.
The Umashankar method corrects only magnetic fields without
considering the discretization error. By utilizing the proposed
thin wire representation, FDTD is able to model both the
thin wire and the earth currents accurately, although MoM
cannot handle the earth currents except in simple configurations
[8]. This paper first describes the proposed method, and then
comparisons between calculated and laboratory test results are
shown to validate the method.
II. REVIEW OF THE FDTD METHOD
A. Formulation
There exist several different formulations of FDTD method.
In order to precisely describe the proposed method of thin wire
representation, the formulation used in this paper is briefly
reviewed here. Assuming neither anisotropic nor dispersive
medium in the space of interest, the Maxwell equations in the
Cartesian coordinates are
(1)
and

Manuscript received August 15, 2001.

T. Noda is with the Department of Electrical Insulation, Central Research
Institute of Electric Power Industry (CRIEPI), Tokyo 201-8511, Japan (e-mail:
takunoda@criepi.denken.or.jp).
S. Yokoyama is with Central Research Institute of Electric Power
Industry (CRIEPI), Komae-shi, Tokyo 201-8511, Japan (e-mail:
yokoyama@criepi.denken.or.jp).
Publisher Item Identifier S 0885-8977(02)05920-4.

where
electric field;
magnetic field;
charge density;
permittivity;

0885-8977/02$17.00 2002 IEEE

(2)

NODA AND YOKOYAMA: THIN WIRE REPRESENTATION IN FDTD SURGE SIMULATION

permeability;
conductivity.
The space of interest is a rectangular-parallelepiped, and it is
(referred to as the space step
discretized by a small length
hereafter) in all the directions. As a result, the space is filled
, and each cube is called a
with cubes of which the sides are
cell. Fig. 1 shows the cell with the configuration of electric and
magnetic fields that are considered to be constant within the cell.
In (1), the derivatives with respect to , , and are replaced by
a central difference formula
(3)

841

and the derivatives with respect to time are replaced by

(4)
denotes a component of
or
. Assuming
where
that the electric fields are calculated at time steps
and the magnetic fields at
by turns, we finally
obtain (5)(10), shown at the bottom of the page (an approxis employed in the
imation
denotes
component electric
derivation).
,
,
, and at
field at position

(5)

(6)

(7)

(8)

(9)

(10)

842

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 3, JULY 2002

Fig. 1. Configuration of electric and magnetic fields in cell.

time
, and the other components are expressed in the
,
, and
are given by
same manner. Coefficients
(11)
Equations (5)(11) are the FDTD formulas of the Maxwell
equations [2], [3]. Although (2) is not explicitly formulated, it
is proven that (5)(11) automatically satisfies (2) [3].
B. Time Step and Space Step
Equations (5)(10) are considered as numerical integration,
and stable integration is performed if the following condition is
satisfied (Courants condition) [3]
Fig. 2. Thin wire and configuration of adjacent electric and magnetic fields.

(12)
On the other hand, the grid dispersion error is minimized when
the above relation is an equality. Thus, the following formula is
used in all calculations in this paper to determine time step
by user defined space step
:
(13)
is a small positive value specified by a user in order to prevent
instability of the numerical integration due to round-off error in
(5)(10).
III. PROPOSED THIN WIRE REPRESENTATION
If the space step were chosen to be small enough to represent
the shape of wires cross section, an accurate representation
would be possible. However, it requires impractical computational resources at this moment. The thin wire is defined
as a conductive wire of which the radius is smaller than the
size of a cell in the FDTD simulation. In antenna simulations, the thin wire is mainly used to represent an antenna
elementthe most important part. In surge simulations, it is
also important to represent wires (phase and ground wires of
a transmission/distribution line) and steel frames of a building
along which surges propagate. Umashankar et al. proposed
a method of thin wire representation by correcting the adjacent magnetic fields of the wire according to its radius [7],
and [9] reports that the method is valid for the calculation
of radiated fields by an antenna. However, the Umashankar
method cannot give accurate surge impedance, as shown in
Section V-B of this paper.

A. Modification of Permittivity and Permeability

The proposed method of thin wire representation that corrects both the adjacent electric and magnetic fields of the wire
according to its radius gives accurate surge impedance. The correction of the fields is carried out by equivalently modifying
the permittivity and permeability of the adjacent cells. Fig. 2(a)
shows a wire with radius placed in the direction, and the
permittivity and permeability of the space are and . Fig. 2(b)
shows the cross section of the wire with the adjacent electric
fields, and Fig. 2(c) with the adjacent magnetic fields. In the
FDTD method, a wire is, in principle, represented by forcing
the electric fields along the center line of the wire to be zero,
s are forced to be zero in this case. Calculated electric
and
and magnetic fields around the wire are in a certain distribution
(including the effects of space and time discretization), and the
distribution coincides with a real one around a wire with radius
. In other words,
is considered to be the radius of which
the real distribution of electric and magnetic fields around the
wire is the same as one obtained by the FDTD method by simply
forcing electric fields along a line to be zero, and is called the
intrinsic radius in this paper (the value of is evaluated later).
Therefore, in order to represent the desired radius , permittivity
to calculate the adjacent electric fields
,
,
,
[see Fig. 2(d)] is multiplied by a correction factor , and also
,
permeability to calculate the adjacent magnetic fields
,
,
[see Fig. 2(e)] is divided by the same factor ,
based on the fact that forcing zero electric fields along the wire
automatically gives the intrinsic radius . The correction factor
is, of course, a function of .

NODA AND YOKOYAMA: THIN WIRE REPRESENTATION IN FDTD SURGE SIMULATION

Fig. 3.

843

Electric field around thin wire.

B. Correction Factor and Intrinsic Radius

The correction factor is determined so that the four adjacent
electric fields of Fig. 2(b) are equal to those of Fig. 2(d),
and also so that the four adjacent magnetic fields of Fig. 2(c)
are equal to those of Fig. 2(e). Because the distance between
is
, the distance is theoretically
the wire and boundary
short enough to regard the electric field perpendicular to the
wire surface as inversely proportional to the distance from the
center of the wire in the region between the wire surface and
. Therefore, is an approximate equipotential surface with
can be determined by equating the
respect to the wire, and
capacitance (between the wire and ) of Fig. 2(b) and (d).
is a cylinder with radius
for simplicity,
Assuming that
the following equation must be satisfied
(14)
Thus, the correction factor

the vicinity electric field perpendicular to the thin wire is analytically given as in inverse proportion to distance from the
center of the wire
(16)
. Fig. 3(c) shows the curve
This is also normalized as
,
, and
calculated by the
of (16) and electric fields
FDTD calculation as shown by circles. The circles farther than
agree well with the curve (even in farther region which is not
represents the electric field in
shown in the figure). Because
and
(the origin of is at the
the range between
center of the thin wire), the potential difference between
and
obtained by the FDTD calculation is
that
corresponds to the area enclosed by a broken line in Fig. 3(c). On
the other hand, the analytical expression (16) gives the potential
difference in the following form:

is obtained as

(17)
(15)

Equating the above expression to

gives
(18)

of Fig. 2(d) gives the same capacThe modified permittivity

itance value as Fig. 2(b) with desired radius and with original
permittivity . In the same manner, it can also be derived that the
of Fig. 2(e) gives the same inducmodified permeability
tance value as Fig. 2(c) with desired radius and with original
permeability . The above theory is based on that the electric
and magnetic fields are electrostatic and magnetostatic respectively in the vicinity of the wire (within boundary ).
Next, we evaluate the value of the intrinsic radius . Fig. 3(a)
shows a thin wire in an FDTD calculation, and the electric fields
along the thin wire are simply forced to be zero without the
permittivity and permeability corrections described previously.
Fig. 3(b) is a current waveform converging into a constant value
, and this current is flowed in the thin wire in the FDTD caland
culation. We take notice of electric field strength
of which the direction is perpendicular to the thin wire. The
after the current
actual FDTD calculation gives
sufficiently, in the case that
is normalized to
reaches to
unity, as given in Appendix A. It should be noted that the value
also takes into account the effects of the discretization, i.e., the finite difference formulation, with respect to time
and space. Because the current reaches the constant value ,

This is the value of the intrinsic radius of the FDTD thin wire
representation. Substituting (18) into (15) gives the final form of
the correction factor
(19)
The proposed thin wire representation is summarized as follows.
1) Preceding an FDTD calculation itself, the correction
factor of each thin wire is calculated due to (19).
2) In the FDTD calculation due to (5)(11), electric fields
around each thin wire are calculated using the modified
.
permittivity
3) In the same manner, magnetic fields around each thin
wire are calculated using the modified permeability
.
C. Theoretical Comparison with the Umashankar Method
The Umashankar method is based on the concept of the
subcell [3], [7]. According to the subcell concept, the
intrinsic radius of the FDTD thin wire representation is
which is different from what we

844

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 3, JULY 2002

obtained in (18). The subcell concept applies Faradays elec, where

tromagnetic induction law to a region smaller than
the effects of the discretization with respect to time and space
are significant. However, the subcell concept does not take
into account the discretization effects. On the other hand, the
proposed value of the intrinsic radius given in (18) takes into
account the discretization effects as mentioned above.
Another difference is that the Umashankar method modifies
magnetic fields only. This difference may come from the difference of purposes. The main purpose of the Umashankar method
is to accurately simulate electromagnetic fields scattered by thin
wires, where the surge impedance is unimportant [7], [9]. On
the other hand, the proposed method modifies both the electric and magnetic fields in order to precisely simulate the surge
impedance.
IV. GENERAL SURGE ANALYSIS PROGRAM
The development of a general surge analysis program, named
Virtual Surge Test Lab. (VSTL), which is based on the FDTD
method and the proposed thin wire representation method, is
briefly described in this section. Sections IV-AE feature the
developed program.
A. Treatment of Boundaries
Each boundary of the space of interest can independently
be defined as a perfectly conducting plane or an absorbing
plane. The perfectly conducting plane can easily be represented
by forcing the tangential components of electric fields at the
boundary to be zero. The second-order Liaos method is used
to represent the absorbing plane, because it is more accurate
and requires less memory compared with other methods [10].
An open space can be assumed by applying the absorbing plane
to all the boundaries of the space of interest.
B. Imperfectly Conducting Earth
The goal of surge analysis is usually to find the solution of
surge propagation in a 3-D skeleton structure above an imperfectly conducting earth. In the FDTD calculation, the representation of the imperfectly conducting earth with resistivity can
in
be achieved by simply setting the value of in (11) to
the region defined as the earth soil.
C. Rectangular-Parallelepiped Conductors
The geometrical shape of most power equipments can be
represented by a combination of several rectangular-parallelepiped objects. The rectangular-parallelepiped conductor is
simply modeled by forcing the tangential electric fields on its
surface to be zero.
D. Localized Voltage and Current Sources
Unlike the static electric fields, the transient electric fields
. Thus, in the analysis of transient
do not satisfy
fields, the voltage or the voltage difference do not make sense
in general. However, if we take notice of an electric field
component of a cell, the voltage difference across a side of
, because
the cell can reasonably be defined as
waves of which the wave length is shorter than
do not present in the FDTD calculation due to the bandwidth

Fig. 4. Calculation procedure of developed program.

limitation of
. Based on this fact, a localized voltage source
with and without its internal resistance can be modeled in the
FDTD calculation as in [3].
In the case of a current source, because current itself is a
general quantity even in the transient fields, a localized current
source with and without its internal resistance can also be modeled as in [3].
E. Calculation Procedure and Output
The flowchart of the calculation procedure of the developed
program is shown in Fig. 4. The output of the program includes
the waveform of localized voltage differences and current
intensities at a specified position in a specified direction, and
an animation of electric or magnetic field distribution in an
arbitrary section is also included. The visualization of the
animation is carried out with the help of MATLAB.
V. SIMULATION RESULTS
A. Horizontal Conductor System
Fig. 5 shows a horizontal conductor system, one of the
most fundamental elements of the surge analysis, where a thin
m is placed
wire conductor with radius and length

NODA AND YOKOYAMA: THIN WIRE REPRESENTATION IN FDTD SURGE SIMULATION

Fig. 5.

845

Conductor arrangement (horizontal conductor system).

= 20
(1 = 5 )

Fig. 7. Electric field strength at t

ns (horizontal conductor system);
electric field strength in [V/m] corresponding to gray scale at the bottom; unit
of vertical and horizontal axes is in cells
s
cm .

Fig. 6. Calculated and measured waveforms (horizontal conductor system):

(a) waveform of voltage source used in the simulation was obtained as piecewise
linear approximation of measured open voltage; (b) and (c) comparison between
measured and calculated waveforms.

above a copper plate at height . The horizontal conductor

is excited by a pulse generator (PG) of which the internal
resistance is 50 , and connected via a vertical lead wire. In
this configuration, voltage and current waveforms at PG were
measured, and an FDTD simulation was also carried out. In
the simulation, the dimensions of the analysis space were 2 m,
6 m, and 2 m in the , , and directions, respectively, and the
space step was 5 cm. The time step was determined by (13)
. All the six boundaries were treated as the
with
second-order Liaos absorbing boundary, and the resistivity of
two-cell layers at the bottom of the analysis space was set to
simulating the copper plate. PG was modeled
as a -direction voltage source with its internal resistance 50
in series, of which the waveform was given by a piecewise
linear approximation of its measured open voltage as shown
in Fig. 6(a). The radii of the horizontal conductor and the
vertical lead wire were taken into account by the proposed
method. Fig. 6(b) and (c) show the measured and calculated
cm and
cm, and the
waveforms in the case of
calculated waveforms agree well with measured ones. Fig. 7 is
ns on the conductor
the electric field distribution at
plane (a snapshot of its animation visualized by MATLAB).

Fig. 8. Comparison between calculated and measured surge impedance;

different density (gray scale) indicates different conductor height.

This shows an instance that the reflected wave is about to go

back toward the sending end.
B. Accuracy of Surge Impedance
Fig. 8 shows a comparison of surge impedance between
measured and calculated values with varying and . The
calculated values are obtained both by the proposed method and
by Umashankars method. The surge impedance was defined
ns. It is obvious that
as the average value between
the proposed method is far more accurate than Umashankars
method.
C. Vertical Conductor System
The modeling of a vertical conductor is important as a basis of
transmission tower modeling. Fig. 9 shows a vertical conductor
system consisting of four cylindrical pipes each of which the radius is 16.5 mm. This is the same configuration in which a measurement was carried out in [11]. The vertical conductors are
excited by a PG through a current lead wire, and the tower-top
voltage is defined as the voltage between the tower top and a
voltage measuring wire. In the simulation, the dimensions of

846

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 3, JULY 2002

Fig. 9. Conductor arrangement (vertical conductor system).

= 35 5
=
(1 = 10 1 )

Fig. 11. Magnetic field strength at t

: ns (vertical conductor system);
expressed by gray scale gradation from dark black
0 A/m to clear white
0.1 A/m; unit of axes is in cells
s
: cm .

and memory is considerable, and even a personal computer

can be used for the FDTD calculations. In fact, the simulations
presented in this paper were performed by a personal computer
with Pentium III 600 MHz CPU and 256 MB RAM. The
computation time for the horizontal conductor case is 2 min
and 53 s, and that for the vertical conductor case is 8 min
and 35 s.
VI. CONCLUSIONS

Fig. 10.

Calculated and measured waveforms (vertical conductor system).

the analysis space were 9.09 m all in the , , and directions

cm. The time step was determined
with space step
, and all the six boundaries were treated as the
with
second-order Liaos absorbing boundary. The thickness and the
m.
resistivity of the earth were set to 3.03 m and 1.69 10
PG was modeled by a current source, with internal resistance
5 k , of which the waveform was given in Fig. 10(a). Fig. 10(b)
and (c) show waveforms of tower-top voltage and current, and
the calculated results agree well with measured ones. Under
a different condition that four 1.01-m vertical grounding electrodes are attached to the tower feet in the earth and the earth
resistivity is set to 100 m, another simulation was carried out.
ns on
Fig. 11 shows the magnetic field distribution at
the tower plane, when a part of the incoming wave reflects
at the earths surface and the remaining part penetrates into the
earth.
D. Computation Time
It may be believed that the FDTD method is a time-consuming
method. However, the progress of computers in terms of speed

In this paper, a method of thin wire representation in the

FDTD calculation was developed, and it was shown by a
comparison with a laboratory test result that the new method
gives more accurate surge impedance than previously proposed
Umashankars method. This paper also described the development of a general surge analysis program using the FDTD
method incorporating the new thin wire representation method.
Two conductor systems, a horizontal conductor system and
a vertical conductor system, were analyzed by the developed
program, and its accuracy was validated by comparisons between the simulation results and corresponding laboratory test
results.
APPENDIX
GENERAL VALIDITY OF
Although
was obtained with a particular value
,
is generally valid regardless of values of
of
as long as (13) is used to determine
because of the following
. In such
reason: A thin wire is placed in the space where
and
always appear in the form of
in the
space,
is always fixed
FDTD formulas (5)(11), and the ratio
by (13). Thus, any practical value of
gives
to be
as long as
is determined by (13).
ACKNOWLEDGMENT
The authors wish to thank R. Yonezawa, Tokyo University
of Agriculture and Technology, and H. Arai, CRIEPI, for their
contributions and Drs. T. Shindo, Y. Sunaga, and K. Tanabe,
CRIEPI, for their valuable discussions.

NODA AND YOKOYAMA: THIN WIRE REPRESENTATION IN FDTD SURGE SIMULATION

REFERENCES
[1] H. W. Dommel, Digital computer solution of electromagnetic transients
in single- and multi-phase networks, IEEE Trans. Power App. Syst., vol.
PAS-88, pp. 388399, Apr. 1969.
[2] K. S. Yee, Numerical solution of initial boundary value problems involving Maxwells equations in isotropic media, IEEE Trans. Antennas
Propagat., vol. AP-14, pp. 302307, May 1966.
[3] K. S. Kunz and R. J. Luebbers, The Finite Difference Time Domain
Method for Electromagnetics. Boca Raton, FL: CRC, 1993.
[4] R. F. Harrington, Field Computation by Moment Methods. New York:
Macmillan, 1968.
[5] G. J. Burke and A. J. Poggio, Numerical Electromagnetics Code
(NEC)Method of Moments: Lawrence Livermore Laboratory, 1981.
[6] K. Tanabe, Novel method for analyzing the transient behavior of
grounding systems based on the finite-difference time-domain method,
in Proc. IEEE Power Engineering Society Winter Meeting, vol. 3, 2001,
pp. 11281132.
[7] K. R. Umashankar et al., Calculation and experimental validation of
induced currents on coupled wires in an arbitrary shaped cavity, IEEE
Trans. Antennas Propagat., vol. AP-35, pp. 12481248, Nov. 1987.
[8] G. J. Burke and E. K. Miller, Modeling antennas near to and penetrating a lossy interface, IEEE Trans. Antennas Propagat., vol. AP-32,
pp. 10401049, Oct. 1984.
[9] T. Kashiwa, S. Tanaka, and I. Fukai, Time domain analysis of
Yagi-Uda antennas using the FDTD method, IEICE Trans. Commun.,
vol. J76-B-II, pp. 872872, Nov. 1993.
[10] Z. P. Liao, H. L. Wong, B.-P. Yang, and Y.-F. Yuan, A transmitting
boundary for transient wave analysis, Science Sinica, Series A, vol. 27,
no. 10, pp. 10631063, 1984.
[11] T. Hara et al., Transmission tower model for surge analysis, in Proc.
H3 IEE Japan Power and Energy Conf., 1991, Paper no. II-270.

847

Taku Noda (S94M97) was born in Osaka,

Japan, on July 4, 1969. He received the B.S., M.S.,
and Ph.D. degrees in engineering from Doshisha
University, Kyoto, Japan, in 1992, 1994, and 1997,
respectively.
He was with DEI Simulation Software, Neskowin,
OR, in 1994, and was a Consultant at the Bonneville
Power Administration (BPA), Portland, OR, in 1995.
In 1997, he joined the Central Research Institute of
Electric Power Industry (CRIEPI), Tokyo, Japan,
where he holds the position of Research Scientist.
Since January 2001, he has been a Visiting Scientist at the University of
Toronto, Toronto, ON, Canada. His research interests include transient analysis
of power systems.
Dr. Noda is a Member of IEE of Japan.

Shigeru Yokoyama (M83SM91F96) was born

in Miyagi, Japan, on March 5, 1947. He received the
B.S. and Ph.D. degrees in engineering from the University of Tokyo, Tokyo, Japan, in 1969 and 1986,
respectively.
In 1969, he joined the Central Research Institute of
Electric Power Industry (CRIPEPI), Tokyo, where he
currently holds the position of Associate Vice President. His research interests include lightning protection and the insulation coordination of transmission
and distribution lines.
Dr. Yokoyama is one of the Vice Presidents of IEE of Japan.