C4

Power system technical

performance

Electromagnetic computation
methods for lightning surge studies
with emphasis on the FDTD method

Reference: 785

December 2019
Electromagnetic computation
methods for lightning surge studies
with emphasis on the FDTD method
WG C4.37

WG C4.37

Members

Y. BABA, Convenor JP A. TATEMATSU, Secretary JP

G. ALA IT J. HE CN
F. HEIDLER DE K. R. HUBBARD ZA
B. KORDI CA U. KUMAR IN
G. MASLOWSKI PL A. MIMOUNI DZ
V. A. RAKOV US F. SILVEIRA BR
E. SOTO CO T. H. TRAN VN
K. YAMAMOTO JP F. YAPRAKDAL TR
P. YUTTHAGOWITH TH

Corresponding Members
A. AMEDEO IT D. DE MACEDO CORREIA BR
F. RACHIDI CH T. TSUBOI JP

Reviewers
W. A. CHISHOLM CA G. LIETZ DE

Copyright © 2019
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any
form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their
internal intranet or other company network provided access is restricted to their own employees. No part of this publication may
be reproduced or utilized without permission from CIGRE”.

Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted
by law”.

WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.

ISBN : 978-2-85873-487-0
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Executive summary
This Technical Brochure aims at introducing electromagnetic computation methods with particular
emphasis on the finite-difference time-domain (FDTD) method for solving Maxwell’s equations, and
showing representative applications of the FDTD method to lightning electromagnetic field and surge
studies. The FDTD method is based on a simple procedure for full-wave solutions in the time domain,
and does not require the assumption of a transverse electromagnetic (TEM) field structure, which
circuit simulation methods are based on. Also, it is capable of treating complex geometries and
inhomogeneities, as well as incorporating nonlinear effects and components. Therefore, the method
can be applied to studying lightning electromagnetic fields and surges of a complex three-dimensional
(3D) conductor system such as a power transmission/distribution line struck by lightning or illuminated
by lightning electromagnetic pulses, and can yield relatively accurate results.
The first peer-reviewed paper, in which the FDTD method was used in surge simulation, was
published in 2001. More than 100 journal papers and many conference papers, in which the FDTD
method is used in lightning electromagnetic field and surge simulations, have been published during
the past fifteen years. Interest in using the FDTD method continues to grow because of the availability
of both commercial and non-commercial FDTD software implementations and increased
computational capabilities.
The first attempt in CIGRE to review electromagnetic computation methods applicable to transient
simulations was done by WG C4.501: Numerical Electromagnetic Analysis and its Application to
Surge Phenomena, convened by Professor Akihiro Ametani. In 2013, they produced a useful technical
brochure (Guide for Numerical Electromagnetic Methods: Application to Surge Phenomena and
Comparison with Circuit Theory-based Approach, CIGRE Technical Brochure, no. TB543) (CIGRE
WG4.501). The present technical brochure is more focused on the FDTD method, and includes its
state-of-the art implementations and applications. It is suitable for electric power engineers and
researchers, who are concerned with lightning surge protection studies.
This Technical Brochure is composed of six chapters. In Chapter 1, we overview electromagnetic
computation methods that have been applied to lightning surge studies to date. In Chapter 2, we brifly
describe lightning discharge phenomenon. Lightning surges in electric power systems are also
described, as well as measures for protecting these systems against lightning surges. In Chapter 3,
we present update equations for electric and magnetic fields used in FDTD computations in the 3D
Cartesian coordinate system, and we describe representation of lumped sources, lumped linear circuit
elements, such as resistors, inductors and capacitors, and lumped nonlinear elements. Also, we
discuss representation of thin wires, lightning channel, twisted-wire pairs, coaxial cables, and surge
arresters. Furthermore, we provide useful information about the methods based on electric field
integral equations, which have also been applied to lightning electromagnetic field and surge studies.
These include the method of moments, the partial-element equivalent-circuit method, and the hybrid
electromagnetic model. In Chapter 4, we review representative applications of the FDTD method to
lightning electromagnetic field and surge simulations. These include calculations of lightning surges on
grounding electrodes, lightning surges on overhead power transmission lines and towers, lightning
surges on overhead power distribution lines, lightning surges on wind turbines, lightning surges and
electromagnetic environment in buildings, lightning electromagnetic pulses, and others. In Chapter 5,
we discuss an application of the FDTD method to study switching surges in substations. In Chapter 6,
Chapters 1 to 5 are summarized.

3
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Contents
Executive summary ............................................................................................................. 3

1. Overview................................................................................................................... 12
1.1 Electromagnetic computation methods for lightning surge studies .................................................. 12

2. Lightning .................................................................................................................. 13
2.1 Lightning discharges .............................................................................................................................. 13
2.1.1 Introduction ........................................................................................................................................ 13
2.1.2 Downward negative lightning discharges to ground ........................................................................... 13
2.1.3 Positive lightning discharges to ground.............................................................................................. 15
2.1.4 Upward lightning discharges .............................................................................................................. 17
2.2 Lightning surges...................................................................................................................................... 18
2.2.1 Surges due to direct lightning strike ................................................................................................... 18
2.2.2 Surges induced by a nearby lightning strike ...................................................................................... 20
2.2.3 Surges coming from grounding due to its potential rise ..................................................................... 20
2.3 Summary .................................................................................................................................................. 20

3. Fundamental theories of electromagnetic computation methods ....................... 22

3.1 Finite-difference time-domain (FDTD) method ...................................................................................... 22
3.1.1 Introduction ........................................................................................................................................ 22
3.1.2 FDTD expressions of Maxwell’s equations ........................................................................................ 22
3.1.3 Absorbing boundary conditions .......................................................................................................... 25
3.1.4 Representation of lumped sources and lumped circuit elements ....................................................... 26
3.1.5 Representation of thin wire ................................................................................................................ 28
3.1.6 Representation of lightning return-stroke channel ............................................................................. 29
3.1.7 Representation of surge arresters...................................................................................................... 33
3.1.8 Representation of corona discharge .................................................................................................. 34
3.1.9 Representation of breakdown characteristics of arcing horns ............................................................ 36
3.1.10 Representation of breakdown characteristics of transmission line surge arresters with air gaps ....... 40
3.1.11 Representation of effects of AC voltages ........................................................................................... 41
3.1.12 Representation of twisted-wire pairs .................................................................................................. 42
3.1.13 Representation of lossy thin wire ....................................................................................................... 44
3.1.14 Representation of lossy coaxial cable ................................................................................................ 47
3.1.15 Advantages and disadvantages ......................................................................................................... 51
3.1.16 Other techniques................................................................................................................................ 51
3.2 Electric field integral equation (EFIE) based methods for lightning studies ...................................... 57
3.2.1 Introduction ........................................................................................................................................ 57
3.2.2 Numerical solution of the equation ..................................................................................................... 58
3.2.3 Modelling of the soil ........................................................................................................................... 62
3.2.4 Other issues ....................................................................................................................................... 63
3.2.5 Advantages and disadvantages ......................................................................................................... 63
3.3 Partial element equivalent circuit (PEEC) method ................................................................................ 64
3.3.1 Formulation of partial element equivalent circuit in the frequency domain ......................................... 66
3.3.2 Formulation of partial element equivalent circuit in the time domain .................................................. 67
3.3.3 Advantages and disadvantages ......................................................................................................... 67
3.4 Hybrid electromagnetic model (HEM) .................................................................................................... 68
3.4.1 Introduction ........................................................................................................................................ 68
3.4.2 Brief history of the model evolution .................................................................................................... 68
3.4.3 Fundamental theory ........................................................................................................................... 68
3.4.4 Electromagnetic coupling expressions ............................................................................................... 70
3.4.5 Integrating the physical system .......................................................................................................... 71
3.4.6 Computational aspects ...................................................................................................................... 72
3.4.7 Advantages and disadvantages ......................................................................................................... 72
3.4.8 Comparison with experimental data ................................................................................................... 73
3.5 Summary .................................................................................................................................................. 75

4
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4. Applications to lightning surge studies ................................................................. 76

4.1 Introduction .............................................................................................................................................. 76
4.2 Surges on grounding electrodes............................................................................................................ 76
4.2.1 FDTD simulation of a grounding grid with an aerial wire .................................................................... 76
4.2.2 FDTD simulation of grounding electrodes considering soil ionization ................................................ 81
4.2.3 PEEC-based simulations of grounding electrodes ............................................................................. 87
4.3 Overhead power transmission lines and towers .................................................................................. 91
4.3.1 FDTD simulation of lightning surges on an overhead conductor in the presence of ground-wire corona
discharge ........................................................................................................................................... 91
4.3.2 FDTD lightning surge simulation of an HV air-Insulated substation with back-flashover phenomena 94
4.3.3 FDTD simulation considering AC operating voltage for an air-insulated substation ........................... 99
4.3.4 HEM simulation of lightning surges on overhead transmission lines and towers ............................. 104
4.4 Overhead distribution lines .................................................................................................................. 112
4.4.1 FDTD simulation of lightning-induced surges .................................................................................. 112
4.4.2 HEM simulation of lightning-induced surges .................................................................................... 122
4.5 Lightning surges in wind turbine generator towers ........................................................................... 129
4.5.1 Transient characteristics of a wind turbine grounding system .......................................................... 130
4.5.2 Common-mode lightning current from a structure struck by lightning to a wind turbine ................... 133
4.6 Lightning surges in microwave relay stations .................................................................................... 137
4.7 Lightning electromagnetic pulses........................................................................................................ 143
4.7.1 FDTD study of errors in magnetic direction finding of lightning due to the presence of conducting
structure near the field measuring station ........................................................................................ 143
4.7.2 FDTD modeling of LEMP propagation in the earth ionosphere waveguide ...................................... 151
4.8 Others ..................................................................................................................................................... 155
4.8.1 FDTD simulation of lightning current in a multilayer CFRP panel with triangular-prism cells ........... 155
4.8.2 FDTD analysis of the electric field of a substation arrester under a lightning overvoltage ............... 162
4.9 Summary ................................................................................................................................................ 169

5. Switching surges in substations .......................................................................... 172

6. Conclusions ........................................................................................................... 178

References ....................................................................................................................... 180

Illustrations & Figures

Figure 2.1 Diagram showing the luminosity of a downward negative lightning flash to ground containing three
strokes and the corresponding current at the channel base: (a) still camera image, (b) streak-camera image, and
(c) channel-base current. ....................................................................................................................................... 14
Figure 2.2 Average negative first- and subsequent-stroke currents each shown on two time scales, A and B, as
reported by Berger et al. [31]. The lower time scales correspond to the solid-line curves (A), while the upper time
scales correspond to the broken-line curves (B). The vertical scale is in relative units, the peak values being equal
to negative unity. ................................................................................................................................................... 16
Figure 2.3 Schematic diagram showing the luminosity of an upward negative flash and the corresponding current
at the channel base. (a) Still-camera image, (b) streak-camera image, (c) current record. ................................... 18
Figure 2.4 Schematic diagram of a lightning strike to a phase conductor of a two-circuit overhead power
transmission line (only one circuit is shown in this diagram). Since lightning hits a phase conductor without being
intercepted by an overhead ground wire (also referred to as shield wire) installed above the phase conductors,
this phenomenon is called shielding failure. .......................................................................................................... 19
Figure 2.5 Schematic diagram of a lightning strike to the top of an overhead power transmission-line tower or to
an overhead ground wire near the tower. If the resultant surge voltage at the tower is higher than the withstand
voltage of insulator strings, a flashover occurs and potentially results in ground fault. This phenomenon is called
back-flashover since it is due to the voltage rise at the tower, which is normally at ground potential. ................... 19

5
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 2.6 Schematic diagram of a lightning strike to a tree located near an overhead power distribution line. In
such a case, transient electromagnetic fields generated by lightning illuminate line conductors and induce surge
currents and voltages on them. ............................................................................................................................. 20
Figure 2.7 Schematic diagram of a lightning strike to a man-made metallic object, such as a radio-relay-antenna
tower, located near an overhead power distribution line, both installed on poorly-conducting ground. In such a
case, the potential rise of the antenna tower foot may result in surge voltages that enter the overhead distribution
line via its grounding. ............................................................................................................................................. 21
Figure 3.1 Placement of electric field and magnetic field components in or on a cell. ........................................... 22
Figure 3.2 (a) Electric field component in the x direction Exn at a location (i+1/2, j, k) and the circulating magnetic
field components closest to it and (b) magnetic field component in the x direction Hxn+1/2 at a location (i, j+1/2,
k+1/2) and the circulating electric field components closest to it. .......................................................................... 24
Figure 3.3 (a) Conceptual picture of a z-directed electric field Ez propagating in the negative x direction with the
speed of light c and entering the absorbing boundary located at x=x1 and (b) electric-field computation points near
the absorbing boundary. ........................................................................................................................................ 26
Figure 3.4 Cross-sectional views of a z-directed wire (central solid circle) having the radius a, and the
configuration of electric and magnetic field components closest to the wire: (a) a < a0 (m < 1) and (b) a > a0 (m >
1). .......................................................................................................................................................................... 29
Figure 3.5 Schematic representations of lightning return-stroke channel. ............................................................. 30
Figure 3.6 Piecewise-linear representation of voltage vs. current characteristic of a surge arrester. .................... 33
Figure 3.7 FDTD representations. (a) Inception of corona discharge at the wire surface and (b) Radial expansion
of corona discharge. .............................................................................................................................................. 36
Figure 3.8 Normal and stabilized regions in the analysis space. ........................................................................... 38
Figure 3.9 Time line of the electric and magnetic fields in the case of m = 3. ........................................................ 38
Figure 3.10 Arrangement of the electric and magnetic fields at the boundary in the first step (m = 3). ................. 39
Figure 3.11 Normalized coordinate in the time axis in the first step....................................................................... 39
Figure 3.12 Arrangement of the electric and magnetic fields at the boundary in the second step (m = 3)............. 40
Figure 3.13 Transmission line surge arrester. (a) Configuration and (b) Proposed model. ................................... 40
Figure 3.14 Voltage sources inserted in phase conductors to simulate the effect of AC voltages. ........................ 41
Figure 3.15 Configuration of a twisted-wire pair above the ground plane. ............................................................. 42
Figure 3.16 Arrangement of the electric fields to calculate Ex by an interpolation technique. ................................ 44
Figure 3.17 Local coordinate system for interpolation functions. (a) Two-dimension and (b) One-dimension. ...... 44
Figure 3.18 Series of lumped-parameter impedances to simulate the conductor internal impedance of the thin
wire. ....................................................................................................................................................................... 45
Figure 3.19 Electric and magnetic fields surrounding a thin wire and a coaxial cable. .......................................... 48
Figure 3.20 Illustration of transmission line with a lumped branch. ........................................................................ 52
Figure 3.21 Profile of the applied voltage on a 5-cm rod-rod air gap. The wavefront time is 0.12μs. .................... 53
Figure 3.22 The spatial partitioning of a transmission line with both fine and coarse grids, Nfg=3. Note that the
branch line is also horizontal. ................................................................................................................................ 54
Figure 3.23 Example using surge arrester with 5-cm series gap. .......................................................................... 55
Figure 3.24 Profiles of the residual voltage on the arresters with and without air gap; time step is 0.03 μs. ......... 56
Figure 3.25 Comparison of the residual voltage profiles, three calculation modes are considered: coarse grid
mode, fine grid mode and subgridding mode......................................................................................................... 57
Figure 3.26 Flow chart for transient analysis using NEC-2 and Fast Fourier Transform (FFT). ............................ 60
Figure 3.27 Arbitrary path in wire. .......................................................................................................................... 60
Figure 3.28 Procedures in the simulation of PEEC models [134]. ......................................................................... 65
Figure 3.29 Discretization of structure in the PEEC models. ................................................................................. 65
Figure 3.30 Pulse basis functions (bn and bm) on current cells and potential cells. ............................................... 66
Figure 3.31 Equivalent circuit model of a PEEC cell. ............................................................................................. 66
Figure 3.32 Particular system subjected to lightning currents. ............................................................................... 69
Figure 3.33 Current sources IT (transversal) and IL (longitudinal) associated with each segment that composes the
evaluated system................................................................................................................................................... 69
Figure 3.34 Mathematical approach to accomplish coupling expressions. ............................................................ 71
Figure 3.35 Distribution of IT and IL in a single segment. ....................................................................................... 71
Figure 3.36 Equation corresponding to the n nodes of system organized as a matrix equation. ........................... 72
Figure 3.37 Flowchart of HEM model operation. ................................................................................................... 72
Figure 3.38 Measured and calculated voltages across upper, middle, and lower insulator strings of a 500-kV–
and–62.8-m-high transmission line tower. Applied current wave: 3-µs-time to crest, 3.2-A crest value. ............... 73
Figure 3.39 Measured and calculated electric field waveforms over a 16.67 Ωm ground, at a distance of 3 m from
the return-stroke channel. (a) Vertical field and (b) Horizontal field. ...................................................................... 74
Figure 4.1 Experimental setup for measuring the ground potential rises of a grounding grid. (a) Top view and (b)
Side view. .............................................................................................................................................................. 77
Figure 4.2 Experimental setup for measuring the induced voltages on an aerial wire. .......................................... 78
Figure 4.3 Resistivities of the soil simulated in the FDTD method. ........................................................................ 78
Figure 4.4 Waveforms of the current injected into the grounding grid. .................................................................. 79
Figure 4.5 Local coordinate system determined on the grounding grid. ................................................................ 79
Figure 4.6 Calculated and measured GPRs of the grounding grid. (a) (20, 15), (b) (20, 30), and (c) (20, 60). ...... 79
Figure 4.7 Calculated and measured GPRs outside the grid. ................................................................................ 80

6
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.8 Calculated and measured currents through the grounding grid. (a) Currents flowing from (20, 0) to (20,
60) and (b) Currents flowing from (20, 30) to (30, 30). .......................................................................................... 80
Figure 4.9 Calculated and measured GPRs of the aerial wire and the voltages induced on the aerial wire. (a)
GPRs of the aerial wire and (b) Induced voltages. ................................................................................................ 81
Figure 4.10 Resistivity profile of Ala et al.’s soil-ionization and de-ionization model (Ala et al. 2008), based on the
dynamic model proposed by Liew and Darveniza (1974), used for FDTD computations. ..................................... 82
Figure 4.11 A single horizontal grounding conductor having a length of 8.1 m in a homogeneous soil of ρ0=200
Ωm, energized by a lumped current source that is connected to an auxiliary vertical grounding conductor of length
1 m, analyzed using the FDTD method. ................................................................................................................ 83
Figure 4.12 FDTD-computed waveform of voltage at the current-injection point into the 8.1-m horizontal grounding
conductor (with the soil-ionization model, Ec=200 kV/m, τ1=3.5 μs, and τ2=4.5 μs), and the corresponding
measured waveform (Sekioka et al. 1998). Also shown is the FDTD-computed voltage waveform without the soil-
ionization model..................................................................................................................................................... 83
Figure 4.13 Resistivity distributions of the soil around the 8.1-m horizontal grounding conductor at time (a) t=0 μs,
(b) t=0.8 μs, and (c) t=3.9 μs. ................................................................................................................................ 84
Figure 4.14 A single vertical grounding conductor of length 0.6 m in a homogeneous soil of ρ0 = 50 Ωm, energized
by a lumped current source that is connected to four auxiliary vertical conductors, to be analyzed using the FDTD
method................................................................................................................................................................... 85
Figure 4.15 FDTD-computed waveforms of voltage at the top of the vertical grounding conductor of length 0.6 m
with the soil-ionization model (Ec = 200 kV/m, τ1 = 1.3 μs, andτ2 = 4.5 μs), and the corresponding measured
waveform (Liew and Darveniza 1974). Also shown is the FDTD-computed voltage waveform without the soil-
ionization model..................................................................................................................................................... 85
Figure 4.16 Resistivity distributions of the soil in the vicinity of the 0.6-m vertical grounding conductor at time (a) t
= 0 μs, (b) 2.6 μs, and (c) 5.2 μs. The resistivity before the soil ionization is ρ0 = 50 Ωm...................................... 86
Figure 4.17 Four vertical grounding conductors of length 3.05 m in a homogeneous soil of ρ0=80 Ωm, energized
by a lumped current source that is connected to an auxiliary grounding conductor, analyzed using the FDTD
method................................................................................................................................................................... 86
Figure 4.18 FDTD-computed waveform of voltage at the top of the four vertical grounding conductors of length
3.05 m with the soil-ionization model (Ec=150 kV/m, τ1=2 μs, and τ2=4.5 μs), and the corresponding measured
waveform (Bellaschi et al. 1942). Also shown is the FDTD-computed voltage waveform without the soil-ionization
model. .................................................................................................................................................................... 87
Figure 4.19 The structures of the grids 1x1, 2x2, and 6x6 the simulation (Grcev et al. 1996). .............................. 88
Figure 4.20 Applied current in the simulation......................................................................................................... 88
Figure 4.21 Comparison of PEEC-calculated waveforms (Yutthagowith et al. 2011) and MoM-calculated
waveforms (Grcev et al. 1996) at point A of the grid grounding systems. .............................................................. 89
Figure 4.22 Comparison of measured and PEEC-computed transient potentials at point A of the square ground
grid. ....................................................................................................................................................................... 89
Figure 4.23 Comparison of measured and PEEC-computed transient potentials at point B of the square ground
grid. ....................................................................................................................................................................... 90
Figure 4.24 Configuration of the T-shape grounding electrode and soil parameters. ............................................ 90
Figure 4.25 Comparison of the FDTD and PEEC-computed transient potentials and the injected current at the
current injected point of the T-shape grounding electrode. .................................................................................... 91
Figure 4.26 (a) 3D and (b) cross-sectional (xz-plane) views of a horizontal single wire and a four-conductor bundle
of length 1.4 km located 14-m above the ground and horizontally 2-m away from the single wire. ....................... 92
Figure 4.27 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of surge voltage
at d = 0, 350, 700, and 1050 m from the energized end of the horizontal single wire above ground of conductivity
10 mS/m. The applied voltage is positive and Ecp = 0.5 MV/m. ............................................................................. 93
Figure 4.28 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of surge voltage
at d = 0, 350, 700, and 1050 m from the energized end of the horizontal single wire above ground of conductivity
10 mS/m. The applied voltage is negative and Ecn = 1.5 MV/m. ............................................................................ 93
Figure 4.29 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of voltage
induced on the nearby four-conductor bundle at d = 0, 350, 700, and 1050 m, located 14-m above ground of
conductivity 10 mS/m. The applied voltage is positive and Ecp = 0.5 MV/m. ......................................................... 94
Figure 4.30 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of voltage
induced on the nearby four-conductor bundle at d = 0, 350, 700, and 1050 m, located 14-m above ground of
conductivity 10 mS/m. The applied voltage is negative and Ecn = 1.5 MV/m. ........................................................ 95
Figure 4.31 Calculation arrangement used to simulate the 77-kV transmission line and substation. .................... 96
Figure 4.32 77-kV transmission line tower model. ................................................................................................. 97
Figure 4.33 Lightning voltages at the entrance of the substation and the lightning currents through the shield wire
and the main leg at the transmission line tower struck by lightning. (a) Measured results and (b) Calculated
results. ................................................................................................................................................................... 98
Figure 4.34 Incoming line model of a 500-kV air-insulated substation. (a) Tower, (b) Gantry, and (c) Overall view.
............................................................................................................................................................................ 100
Figure 4.35 FDTD simulation model of a 500-kV air-insulated substation. .......................................................... 101
Figure 4.36 Equipment layout of a 500-kV air-insulated substation model. ......................................................... 101
Figure 4.37 Equipment layout of a 500-kV air-insulated substation model. (a) Side view and (b) Layout of voltage
source. ................................................................................................................................................................. 102

7
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.38 Voltage waveforms at the circuit breaker terminal acquired by FDTD analysis. (a) Waveform when the
lightning stroke hits the first tower and (b) Waveform when the lightning stroke hits the second tower. .............. 103
Figure 4.39 Voltage waveforms at the circuit breaker terminal acquired by FDTD analysis. (a) Waveform when the
lightning stroke hits the first tower and (b) Waveform when the lightning stroke hits the second tower. .............. 104
Figure 4.40 Representation of the simulated event. (a) 138-kV Transmission line and (b) Tower configuration.
Adapted from [153]. ............................................................................................................................................. 105
Figure 4.41 Simulated grounding electrode arrangements. (a) Configuration applied for 100 and 300 Ωm soil
resistivity and (b) Configuration applied for 600, 1000, 2000, and 4000 Ωm soil resistivity. Adapted from [153]. 105
Figure 4.42 Representative current waveforms of first (a) and subsequent and (b) strokes considering median
parameters measured at Mount San Salvatore station [189]. .............................................................................. 106
Figure 4.43 Simulated overvoltage across upper insulator string of the 138-kV line for constant and frequency-
dependent soil parameters for different values of soil resistivity ρ0. (Left column: First stroke; Right column:
Subsequent stroke). (a),(b) 100 Ω.m, (c),(d) 600 Ω.m, (e),(f) 1000 Ω.m, and (g),(h) 4000 Ω.m. ......................... 107
Figure 4.44 First-stroke critical currents Ic of 230- and 138-kV lines as a function of soil resistivity ρ under the
assumption of constant and frequency dependent soil parameters. .................................................................... 108
Figure 4.45 Response of counterpoise wires 70-m long buried 0.5 m deep in a 2000-Ωm soil subjected to a
representative first-stroke lightning current, along with their concise representations ZP and RLF. Calculated ZP,
RLF and RIP: 14.2 Ω, 18.9 Ω, and 13.7 Ω. Adapted from [194]. ............................................................................ 110
Figure 4.46 GPR of the tower-footing electrodes subjected to the median first-stroke current. Counterpoise wires
70-m long buried 0.5 m deep in a 2000-Ωm soil. Adapted from [194].................................................................. 111
Figure 4.47 Overvoltage across insulator string of a 138-kV line subjected to the median first-stroke current.
Counterpoise wires 70-m long buried 0.5 m deep in a 2000-Ωm soil. Adapted from [194]. ................................. 111
Figure 4.48 Probability of backflashover occurrence as a function of grounding representation: physical
representation of electrodes and concise representations by RLF and ZP .(a) 100 Ωm, (b) 300 Ωm, (c) 600 Ωm, (d)
1000 Ωm, (e) 2000 Ωm, and (f) 4000 Ωm. ........................................................................................................... 112
Figure 4.49 Calculation arrangement of the lightning-induced voltages on a three-phase distribution line above a
perfectly-conducting ground plane. ...................................................................................................................... 114
Figure 4.50 Side view of the calculation arrangement used in case 4 above the perfectly conducting ground plane.
(a) Side view (ZX) and (b) Side view (YZ). .......................................................................................................... 114
Figure 4.51 V-I characteristics of the surge arresters. ......................................................................................... 115
Figure 4.52 Calculated results of the lightning-induced voltages on phase A in case 1 without the shielding wire or
the lightning arresters above the perfectly conducting ground plane. (a) FDTD method and (b) Rusck model. .. 116
Figure 4.53 Calculated results of the lightning-induced voltages on phase A in case 4 without the shielding wire or
the lightning arresters above the perfectly conducting ground plane. (a) FDTD method and (b) Rusck model. .. 116
Figure 4.54 Experiment configuration of 1-cm-radius and 1.4-km-long four-conductor line: (a) cross-sectional (yz-
plane) view of a four-conductor line and (b) plan (xy-plane) view of the distribution network A. .......................... 117
Figure 4.55 Circuit representation of different system elements in the model: (a) surge arrester equivalent circuit
and voltage–current curve of the nonlinear resistor Rpr, (b) surge arrester and transformer point, (c) grounding
point, and (d) transformer and measuring point. .................................................................................................. 118
Figure 4.56 Three-dimensional view of the lightning channel and the four-conductor line distribution network,
analyzed using the 3-D FDTD method. ............................................................................................................... 119
Figure 4.57 Initial (8-μs) part of the simulated lightning current waveform with a peak of 34 kA, front time of 2 μs,
and half-peak width of 85 μs. ............................................................................................................................... 119
Figure 4.58 FDTD-computed waveform of lightning-induced voltage and the corresponding measured waveform
[at point M1]. ........................................................................................................................................................ 120
Figure 4.59 Experimental configuration 1: Plane view of 1-cm-radius and 1.4-km-long four-conductor line with
nearby buildings: (a) he = 0 m (no buildings), (b) he = 5 m, and (c) he = 15 m. The distances between the
measuring point and the closest set of surge arresters located on its left and right sides are labeled se and sd.
Indicated dimensions refer to the full-scale system. ............................................................................................ 121
Figure 4.60 Three-dimensional view of the lightning channel and the four-conductor line distribution network,
analyzed using the 3-D FDTD method. ............................................................................................................... 122
Figure 4.61 FDTD-computed and corresponding measured lightning-induced voltage waveforms for a triangular
lightning current pulse with peak of 34 kA, risetime of 2 μs, and time to half-peak value of 85 μs and with buildings
having heights of (a) 0, (b) 5, and (c) 15 m. The distances between the measuring point (M) and the closest set of
surge arresters were se = sd = 75 m. ................................................................................................................... 123
Figure 4.62 FDTD-computed and corresponding measured lightning-induced voltage waveforms for a triangular
lightning current pulse with peak of 34 kA, risetime of 2 μs, and time to half-peak value of 85 μs and with buildings
having heights of (a) 0, (b) 5, and (c) 15 m. The distances between the measuring point (M) and the closest set of
surge arresters were se = 148 m and sd = 174 m................................................................................................. 124
Figure 4.63 Lightning stroke locations A and B. .................................................................................................. 126
Figure 4.64 Simulated first (a) and subsequent (b) return stroke current waveforms. Median peak currents and
front times measured at MSS station: (a) Ip = 31.1 kA, Td30=3.83 µs and (b) Ip = 11.8 kA, Td30 = 0.67 µs. ..... 126
Figure 4.65 Lightning-induced voltages at the center of the overhead line considering stroke location A for
constant and frequency-dependent soil parameters for different values of soil resistivity ρ0. (Left column: First
stroke; Right column: Subsequent stroke). (a),(b) 100 Ω.m, (c),(d) 1000 Ω.m, a€(e),(f) 10,000 Ω.m. (v=0.3c). .. 127
Figure 4.66 Lightning-induced voltages at the extremity of the overhead line considering stroke location A for
constant and frequency-dependent soil parameters for different values of soil resistivity ρ0. (Left column: First
stroke; Right column: Subsequent stroke). (a),(b) 100 Ω.m, (c),(d) 1000 Ω.m€nd (e),(f) 10,000 Ω.m. (v=0.3c). . 128

8
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.67 Lightning-induced voltages at the center and both close and far terminations of the overhead line
considering stroke location B for constant and frequency-dependent soil parameters for different values of soil
resistivity ρ0. (Left column: First stroke; Right column: Subsequent stroke). (a),(b) 1000 Ω.m and (c),(d) 10,000
Ω.m. (v=0.3c). ...................................................................................................................................................... 129
Figure 4.68 Grounding system of a wind turbine. (a) Top and side views and (b) Incidental grounding electrodes.
............................................................................................................................................................................ 130
Figure 4.69 Experimental set-up.......................................................................................................................... 131
Figure 4.70 Measured and simulated results of the injected current and the potential rise. (a) Injected current
waveform and (b) Potential rise waveform........................................................................................................... 131
Figure 4.71 Layers of ground resistivity at the test site. ....................................................................................... 132
Figure 4.72 FDTD simulation model. ................................................................................................................... 132
Figure 4.73 Potential rises around the foundation. (a) Result at 6 m from the foundation and (b) Result at 20 m
from the foundation. ............................................................................................................................................. 133
Figure 4.74 Potential rises around the foundation. (a) Comparisons of the peak values and (b) Comparisons of the
wave-tail values. .................................................................................................................................................. 133
Figure 4.75 Equalization of the wind turbine and the tall tower. (a) Tower and wind turbine and (b) Equalization
using grounding wires. ......................................................................................................................................... 134
Figure 4.76 Detail drawings of the foundations. (a) Wind turbine and (b) Tall tower. .......................................... 135
Figure 4.77 Grounding resistivity. ........................................................................................................................ 135
Figure 4.78 Experimental setup. (a) Overall view and (b) Top view. ................................................................... 136
Figure 4.79 FDTD simulation model. ................................................................................................................... 137
Figure 4.80 Injected current. ................................................................................................................................ 137
Figure 4.81 Comparisons of the potential rises at the foundation of a wind turbine. (a) Connected and (b)
Disconnected. ...................................................................................................................................................... 138
Figure 4.82 Reduced-scale model of a microwave relay station. (a) Microwave tower on a building (type I-a) and
(b) Microwave tower next to a building (type I-b). ................................................................................................ 139
Figure 4.83 Arrangement inside the building. ...................................................................................................... 139
Figure 4.84 Calculation model of the microwave relay station model. ................................................................. 140
Figure 4.85 Analysis space simulated by the FDTD method. .............................................................................. 140
Figure 4.86 Comparison of the measured and calculated peak values of the currents in case A-1..................... 141
Figure 4.87 Waveforms of the currents in case A-1. (a) Injected current and (b) Current flowing into metal box A.
............................................................................................................................................................................ 142
Figure 4.88 Route of the ground wire of the waveguide outside the building. (a) Case C-1 and (b) Case C-2.... 143
Figure 4.89 FDTD simulation model of a vertical lightning channel above flat, perfectly-conducting ground. (a)
Single quadrant model with two magnetic walls and (b) its equivalent four-quadrant model. .............................. 144
Figure 4.90 Plan view of the model for studying the influence of grounded structure (not to scale) on azimuth
measured with respect to West for the case of the true lightning azimuth ϕ = 45°. ............................................. 145
Figure 4.91 Illustration of lightning direction finding from Hx and Hy for the case of the true lightning azimuth ϕ =
45°. (a) Ideal case of no azimuth error (direction found from Hx and Hy is the same as the true direction to
lightning) and (b) The case of the presence of 60-m tall grounded structure at (7080, 7040) (see Table 4.12 and
Figure 4.92), which causes an error Δϕ =16°. ..................................................................................................... 146
Figure 4.92 Color-coded azimuth error (in degrees) due to the presence of a nearby grounded structure for the
case of true azimuth ϕ equal to 45°. Each small square, except for the one in the center, represents the position
of grounded structure, with the total number of such positions being 24. ‘+’ indicates the location of observation
point, and the arrow indicates the true direction to lightning. ............................................................................... 147
Figure 4.93 Color-coded azimuth error (in degrees) due to the presence of a nearby grounded structure having a
height of 60 m and a square cross-section of 40 m × 40 m for the case of true azimuth ϕ equal to 45° and current
risetimes equal to (a) 0.5 μs and (b) 3 μs. ........................................................................................................... 147
Figure 4.94 Color-coded azimuth error due to the presence of a nearby grounded structure having a height of 60
m and a square cross-section of 40 m × 40 m for the case of true azimuth ϕ equal to 45° and ground conductivity
values equal to (a) 0.1 mS/m, (b) 5 mS/m, and (c) 1000 mS/m. .......................................................................... 148
Figure 4.95 Azimuth errors as a function of ground conductivity when the grounded structure is located at (7000
m, 7000 m), (7000 m, 7080 m), (7040 m, 7080 m), (7120 m, 7080 m), and (7160 m, 7080 m). .......................... 149
Figure 4.96 Color-coded azimuth error due to the presence of a nearby grounded structure having a height of 60
m and a square cross-section of 40 m × 40 m for true azimuth ϕ equal to 45° and different values of structure
conductivity equal to (a) 0.1 mS/m, (b) 1 mS/m, (c) 10 mS/m, and (d) 1000 mS/m. ............................................ 149
Figure 4.97 Color-coded azimuth error due to the presence of a nearby grounded structure having a height of 60
m and a square cross-section of 40 m × 40 m for the case of true azimuth ϕ equal to 26.6° (tan-1ϕ= 0.5). ......... 150
Figure 4.98 Illustration of locations of lightning channel, grounded structure, and observation point when the
center of the structure is located on the straight line that connects the lightning channel with the observation point.
(a) true azimuth ϕ is 26.6°, (b) true azimuth ϕ is 45°, and (c) true azimuth ϕ is 0°. .............................................. 151
Figure 4.99 Configuration (not to scale) used in the 2D-FDTD simulations. ........................................................ 152
Figure 4.100 Waveform of lightning return-stroke current with a peak of 30 kA, and risetime of 10 μs. .............. 153
Figure 4.101 FDTD-computed waveforms of vertical electric field Ez (for three cases: free space, daytime and
nighttime conducting atmosphere) at (a) d = 50 km, (b) d = 100 km, (c) d = 200 km, and (d) d = 500 km. The
computations were performed for perfectly-conducting ground, and 30-kA-peak, 10-μs-risetime lightning current.
............................................................................................................................................................................ 154

9
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.102 FDTD-computed waveforms of azimuthal magnetic field H (for three cases: free space, daytime and
nighttime conducting atmosphere) at (a) d = 50 km, (b) d = 100 km, (c) d = 200 km, and (d) d = 500 km. The
computations were performed for perfectly conducting ground, and 30-kA-peak, 10-μs-risetime lightning current.
At 50 km, 2nd skywaves are not seen in this Figure, but they are discernible in the computed waveforms. ........ 155
Figure 4.103 Measured thermographic picture (observed from above) at time 200 ms of a multilayer CFRP panel
in the vicinity of a fastener, in which a 3-kA current of Component A specified in (SAE 2013) is injected. .......... 157
Figure 4.104 Cross-sectional (x-y) views of electric-field grid and magnetic-field grid when triangular-prims
electric-field cells are used. ................................................................................................................................. 158
Figure 4.105 Plan view of the surface of a 300 mm × 300 mm CFRP panel and magnetic field sensors installed on
the CFRP panel. .................................................................................................................................................. 159
Figure 4.106 FDTD simulation model of the 300 mm × 300 mm CFRP panel with a thickness of 30, 46 or 62 mm,
represented using triangular-prism cells. ............................................................................................................. 159
Figure 4.107 FDTD-computed waveforms (dotted lines) of magnetic field in kA/m on the surface of the 16-layer
CFRP panel at a horizontal distance of 80 mm from the left-side edge (current injection side) and at a distance of
0, 40 and 80 mm from the straight line connecting the current injection and output points, and the corresponding
measured waveforms (solid lines). (a) 0 mm, (b) 40 mm, and (c) 80 mm. ........................................................... 160
Figure 4.108 Same as Figure 4.7.5 but for the 24-layer CFRP panel. (a) 0 mm, (b) 40 mm, and (c) 80 mm. ..... 160
Figure 4.109 Same as Figure 4.7.5 but for the 32-layer CFRP panel. (a) 0 mm, (b) 40 mm, and (c) 80 mm. ..... 160
Figure 4.110 FDTD-computed peak values of magnetic field on the surface of the 16-, 24- and 32-layer CFRP
panels at a distance of 80 mm from the left-side edge and at a distance of 0 to 100 mm from the straight line
connecting the current injection and output points, and the corresponding measured values. (a) 16-layer CFRP
panel, (b) 24-layer CFRP panel, and (c) 32-layer CFRP panel............................................................................ 161
Figure 4.111 Distributions of time integration of the square of FDTD-computed current of the top four layers for
the first 10 μs. (a) 45-degree layer (top), (b) 90-degree layer (second), (c) -45-degree layer (third), and (d) 0-
degree layer (fourth). ........................................................................................................................................... 163
Figure 4.112 Superimposed distribution of time integration of the square of FDTD-computed current of the top
four layers for the first 10 μs. ............................................................................................................................... 164
Figure 4.113 Measured characteristics of a ZnO element having a thickness of 36 mm and a radius of 21 mm. (a)
V-I characteristic of the element and (b) ρ-E characteristic. ................................................................................ 165
Figure 4.114 FDTD simulation model of a 2.6-m long substation arrester. (a) Side view and (b) Plan view (shield
wire is not shown). ............................................................................................................................................... 166
Figure 4.115 FDTD-computed waveforms. (a) Source output voltage and the FDTD-computed voltage across the
2.6-m long arrester and (b) Current flowing in the top of the arrester. ................................................................. 166
Figure 4.116 FDTD-computed waveforms of electric fields at a height of 3.5 m from the ground (2.5 m from the
arrester bottom) and radial distances of 0 (center) and 40 mm (arrester side surface) for models with and without
a 0.3-m-radius shield ring suspended 0.6 m from the arrester top. (a) Vertical electric field and (b) Radial electric
field. ..................................................................................................................................................................... 167
Figure 4.117 Same as Figure 4.7.14 but for electric fields at a height of 1.1 m from the ground (0.1 m from the
arrester bottom) (a) Vertical electric field and (b) Radial electric field. ................................................................. 168
Figure 4.118 FDTD-computed distributions of electric fields along the surface of the 2.6-m high and 40-mm-radius
arrester core for models with and without shield ring suspended 0.6 m below the arrester top at time 2 μs. (a)
Vertical electric field and (b) Radial electric field. ................................................................................................ 169
Figure 4.119 FDTD-computed distributions of electric fields along the surface of the 2.6-m high and 40-mm-radius
arrester core for models with and without shield ring suspended 0.3 m below the arrester top at time 2 μs. (a)
Vertical electric field and (b) Radial electric field. ................................................................................................ 170
Figure 5.1 (a) Equally spaced subconductors of a bundled conductor and (b) Unequally spaced subconductors of
a bundled conductor. ........................................................................................................................................... 172
Figure 5.2 Arrangement of voltage sources and switches. .................................................................................. 173
Figure 5.3 Representation of switch B to simulate the occurrence of a restrike. ................................................. 173
Figure 5.4 Analysis space used for calculating switching surges. ....................................................................... 174
Figure 5.5 Calculation model of the 500-kV air-insulated substation. .................................................................. 175
Figure 5.6 (a) Grounding of busbar #2 near the bus section in case A and (b) Grounding of busbar #2 near the
bus section in case B. .......................................................................................................................................... 176
Figure 5.7 (a) Calculated results of the induced voltages on busbar #2 at the top of the bushing at the bus section
in case B and (b) measured results at the same position., .................................................................................. 176

Tables
Table 1.1 Number of papers on electromagnetic computation methods published in IEEE journals between the
years 2000-2019 (by IEEE Xplore ®). ................................................................................................................... 12
Table 2.1 Parameters of downward negative lightning derived from channel-base current measurements, as
reported by Berger et al. [31]. ................................................................................................................................ 17
Table 4.1 Conditions of lightning overvoltage analysis using EMTP. ..................................................................... 99
Table 4.2 Conditions of lightning overvoltage analysis using FDTD. ................................................................... 101
Table 4.3 Lightning overvoltage values for a circuit breaker based on FDTD and EMTP analyses..................... 104
Table 4.4 Simulated electrode length L as function of low-frequency soil resistivity ρ0. The low frequency
grounding resistance (Rg) for each case is also included.................................................................................... 105

10
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Table 4.5 Peak overvoltages across upper insulator string. ................................................................................ 107
Table 4.6 Influence of frequency dependence of soil parameters on critical first-stroke peak currents. .............. 108
Table 4.7 Scenarios for soil resistivity distribution. .............................................................................................. 109
Table 4.8 Impact of frequency dependence of soil parameters on the expected outage rate of a 138-kV line. ... 109
Table 4.9 Induced voltage peak values at the center of the overhead line considering stroke location A. .......... 127
Table 4.10 Decrease of Induced voltage amplitude at line terminations and line center due to the frequency
dependence of soil parameters for stroke location B (ρ0 = 1000, 5000, and 10,000 Ωm).................................... 129
Table 4.11 Peak values of the currents in cases B-4, C-1, and C-2. ................................................................... 143
Table 4.12 FDTD-computed peak values of Hx and Hy, and azimuth errors estimated from Equation (4.10) for the
case of true azimuth ϕ = 45° and current risetime equal to 1 μs. ......................................................................... 146
Table 4.13 Electric field ionospheric reflection heights. ....................................................................................... 155
Table 4.14 Magnetic Field Ionospheric Reflection Heights. ................................................................................. 156
Table 4.15 Three sets of conductivity of each layer in the fiber direction, horizontally perpendicular direction (to
the fiber), and in the vertical direction of the CFRP panel. .................................................................................. 162
Table 5.1 Peak values of induced voltages in cases A and B.............................................................................. 176

11
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

1. Overview
1.1 Electromagnetic computation methods for lightning surge studies
Lightning return-stroke electromagnetic fields have been calculated using analytical expressions,
derived for a vertical lightning channel above perfectly conducting ground (e.g., [1]). Effects of finite
ground conductivity on lightning electromagnetic fields have also been studied using analytical
expressions (e.g., [2]). These analytical expressions are still in use. Lightning-induced voltages on an
overhead power distribution lines or telecommunication lines have been calculated using an
engineering model of the lightning return stroke (e.g., [1]) and a field-to-conductor coupling model
(e.g., [3]-[6]). Horizontal electric fields above a finitely conducting ground, which are needed for
calculating lightning-induced voltages, have been evaluated using approximate expressions such as
the Cooray-Rubinstein formula [7]. Note that the Cooray-Rubinstein formula is given in the frequency
domain. Lightning surges due to a direct lightning strike to an overhead power transmission or
distribution line have been analyzed using distributed-circuit simulation methods such as the electro-
magnetic transients program (EMTP) [8]. EMTP and other similar programs are still widely used in
lightning surge simulations.
Around 1990, electromagnetic computation methods were first applied to lightning electromagnetic
and surge simulations. One of the advantages of electromagnetic computation methods, in
comparison with circuit simulation methods, is that they allow a self-consistent full-wave solution for
both the transient current distribution in a three-dimensional (3D) conductor system and resultant
electromagnetic fields, although they are computationally expensive. In [9], Podgorski and Landt
applied the method of moments (MoM) in the time domain [10], [11] to analyzing the lightning current
along a tall object struck by lightning. In [12], Grcev and Dawalibi applied the MoM in the frequency
domain [13] to analyze the surge characteristics of a grounding electrode. Since then, the MoM in the
frequency domain has been frequently used in lightning surge simulations (e.g., [14]).
In 2001 [15], Tanabe applied the finite-difference time-domain (FDTD) method [16], which is one of the
electromagnetic computation methods, to studying surge characteristics of a grounding electrode. In
2003 [17], Baba and Rakov used the FDTD method to compute lightning electromagnetic fields.
Between the years from 2000-2019, approximately 320 papers whose index term was FDTD or the
finite-difference time-domain method were published in the IEEE Transactions on Electromagnetic
Compatibility, which is one of the journals most closely related to studies on lightning electromagnetic
fields and surges (see Table 1.1). Interest in using the FDTD method continues to grow. The FDTD
method is the most widely used electromagnetic computation method in lightning electromagnetic field
and surge simulations (e.g., [18], [19]).
Other electromagnetic computation methods such as the partial-element equivalent-circuit (PEEC)
method [20], the hybrid electromagnetic model (HEM) [21], the transmission-line modeling method
(TLM) [22], the finite-element method (FEM) [23] have been recently applied to the analysis of
lightning electromagnetic field and surge simulations (e.g., [24]-[27]).
In this brochure, fundamental theories of electromagnetic computation methods are explained, and
their applications to lightning electromagnetic field and surge simulations are shown with emphasis on
the FDTD method.

Table 1.1 Number of papers on electromagnetic computation methods published in IEEE journals
between the years 2000-2019 (by IEEE Xplore ®).
FDTD MoM TLM FEM PEEC HEM
All IEEE journals 4200 16000 3100 11000 190 30
IEEE Trans. EMC 320 260 260 90 50 20

12
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

2. Lightning
2.1 Lightning discharges
2.1.1 Introduction
Lightning discharges can be classified into two categories: cloud-to-ground discharges and cloud
discharges. Cloud-to-ground discharges are of primary interest in studying and designing lightning
protection of man-made structures located on the ground surface. Note that cloud discharges are of
interest in studying the interaction of lightning with airborne vehicles and their protection against
lightning.
Cloud-to-ground discharges are classified, on the basis of the polarity of the charge effectively
transferred to ground and the direction of the initial leader that creates a conducting path between the
thundercloud and ground prior to high-current return strokes, into four types: downward negative
lightning, upward negative lightning, downward positive lightning, and upward positive lightning. It is
believed that about 90% of all cloud-to-ground lightning are downward negative lightning discharges
and about 10% are downward positive lightning discharges. It is thought that upward lightning
discharges occur only from tall objects that are higher than about 100 m or from objects of moderate
height located on mountain tops.
Time-domain waveshapes of electric and magnetic fields generated by a lightning return stroke
depend on the distance of the observation point from the lightning return-stroke channel, the ground
characteristics such as ground conductivity, the spatial and temporal distribution of current along the
channel, and so on.
Transient voltages or currents generated by lightning in man-made systems such as an electric power
system and a telecommunication system are called lightning surges. Lightning surges on a system are
generated by a direct lightning strike to the system, by a nearby lightning strike through
electromagnetic coupling with conductors of the system, by the ground potential rise due to a nearby
lightning strike, or by lightning surges entering the system on electrical connections from other
systems.

2.1.2 Downward negative lightning discharges to ground

In this subsection, a general explanation of downward negative lightning flashes, which account for
about 90% of all cloud-to-ground lightning, is presented. Figures 2.1 (a) and (b) schematically show
still and time-resolved images of a downward negative lightning flash containing three strokes,
respectively. Figure 2.1 (c) shows the corresponding current at the channel base. In Figures 2.1 (b)
and (c), time advances from left to right. Each of the three strokes is composed of a downward-moving
process termed “leader” and an upward-moving process termed “return stroke”.
The leader creates a conducting path between the cloud negative charge region and ground and
deposits negative charge along this path. The return stroke traverses the leader path upward from
ground to the cloud charge region and neutralizes the negative leader charge. Thus, both leader and
return-stroke processes contribute to transporting negative charge from the cloud to ground. The
leader initiating the first return stroke develops in virgin air, and appears to be an optically intermittent
process. Therefore, it is termed “stepped leader”. The stepped-leader branches are directed
downward, which indicates that the stepped leader (and the flash) is initiated in the cloud and
develops downward. The stepped leader extends toward ground at an average speed of 2 × 105 m/s
in a series of discrete steps, with each step being typically 1 μs in duration and tens of meters in
length, with the interval between steps being 20 to 50 μs (e.g., [28]). The peak value of the current
pulse associated with an individual step is inferred to be 1 kA or greater. Several coulombs of negative
charge are distributed along the stepped-leader channel. The stepped-leader duration is typically
some tens of milliseconds, and the average leader current is some hundreds of amperes. The
stepped-leader channel is likely to consist of a thin highly-conducting (plasma) core that carries the
longitudinal channel current, surrounded by a low-conductivity corona sheath whose diameter is
typically several meters and that contains the bulk of the leader charge. The process by which the
extending plasma channels of the upward and downward leaders makes contact, via forming a
common streamer zone, is called the break-through phase or final jump. The break-through phase can
be viewed as a switch-closing operation that serves to launch two waves from the junction point. The

13
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 2.1 Diagram showing the luminosity of a downward negative lightning flash to ground containing
three strokes and the corresponding current at the channel base: (a) still camera image, (b) streak-
camera image, and (c) channel-base current.1

length of an upward connecting leader involved in a first stroke is some tens of meters if that leader is
launched from the ground, and it can be several hundred meters long if it is initiated from a tall object.
The return stroke serves to neutralize the leader charge, although it may not neutralize all the leader
charge or may deposit some excess positive charge onto the leader channel and into the cloud charge
source region. The speed of the return stroke, averaged over the visible channel, is typically between
one-third and one-half of the speed of light (e.g., [29]). The speed decreases with increasing height,
dropping abruptly after passing each major branch. The first return-stroke current measured at ground
typically rises to an initial peak of about 30 kA in some microseconds and decays to half-peak value in
some tens of microseconds while exhibiting a number of subsidiary peaks, probably associated with
branches (e.g., [28]). This impulsive component of current may be followed by a current of some
hundreds of amperes lasting for some milliseconds. The return stroke effectively lowers to ground the
several coulombs of charge originally deposited on the stepped-leader channel, including that residing
on all the branches.
When the first return stroke current ceases, the flash may end. In this case, the lightning is called a
single-stroke flash. However, more often the residual first-stroke channel is traversed by a downward
leader that appears to move continuously, as a downward-moving dart, along the pre-conditioned path
of the preceding stroke or strokes. Hence, these leaders are termed “dart leaders”. The dart leader
progresses downward at a typical speed of 107 m/s, typically ignores the first stroke branches, and
deposits along the channel a total charge of the order of 1 coulomb (e.g., [28]). The dart-leader current
peak is about 1 kA. Some leaders exhibit stepping near ground while propagating along the path
traversed by the preceding return stroke, these leaders being termed “dart-stepped leaders”.
Additionally, some dart or dart-stepped leaders deflect from the previous return-stroke path, become
stepped leaders, and form a new termination on the ground.
When a dart leader or a dart-stepped leader approaches the ground, an attachment process similar to
that described for the first stroke takes place, although it probably occurs over a shorter distance and
consequently takes less time, the upward connecting-leader length being of the order of some meters.
Once the bottom of the dart leader or dart-stepped leader channel is connected to the ground, the
second (or any subsequent) return-stroke wave is launched upward and serves to neutralize the
leader charge. The subsequent return-stroke current at ground typically rises to a peak value of 10 to
15 kA in less than a microsecond and decays to half-peak value in a few tens of microseconds. The
upward propagation speed of such a subsequent return stroke is similar to or slightly higher than that
of the first return stroke (e.g., [29]). Note that due to the absence of branches the speed variation
along the channel for subsequent return strokes does not exhibit abrupt drops.

1Reprinted with permission from V. A. Rakov and M. A. Uman, Lightning: Physics and Effects, Cambridge University Press,
Cambridge, UK, p. 110, Figure 4.2, 2003.

14
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The impulsive component of the current in a subsequent return stroke is often followed by a continuing
current that has a magnitude of tens to hundreds of amperes and a duration up to hundreds of
milliseconds. Continuing currents with a duration in excess of 40 ms are traditionally termed “long
continuing currents”. Between 30 and 50% of all negative cloud-to-ground flashes contain long
continuing currents. The source for continuing current is the cloud charge, as opposed to the charge
distributed along the leader channel, the latter charge contributing to at least the initial few hundred
microseconds of the return-stroke current observed at ground. Continuing current typically exhibits a
number of superimposed surges that rise to a peak in some tens to hundreds of microseconds, the
peak being generally in the hundreds of amperes range but occasionally in the kiloamperes range.
These current surges are associated with enhancements in the relatively faint luminosity of the
continuing-current channel and are called “M-components”.
The time interval between successive return strokes in a flash is usually several tens of milliseconds,
although it can be as large as many hundreds of milliseconds if a long continuing current is involved
and as small as one millisecond or less. The total duration of a flash is typically some hundreds of
milliseconds, and the total charge lowered to ground is some tens of coulombs. The overwhelming
majority of negative cloud-to-ground flashes contain more than one stroke. Although the first stroke is
usually a factor of 2 to 3 larger than a subsequent stroke, about one-third of multiple-stroke flashes
have at least one subsequent stroke that is larger than the first stroke in the flash (e.g., [30]). Note that
terms, “lightning”, “lightning discharge”, and “lightning flash” are used interchangeably to refer to the
overall lightning discharge process.
Figure 2.2 shows, on two time scales, A and B, the average impulsive current waveforms for
downward negative first and subsequent strokes. The rising portion of the first-stroke waveform has a
characteristic concave shape. Table 2.1 shows parameters of downward negative lightning derived
from channel-base current measurements, as reported by Berger et al. [31]. It appears from Table 2.1
that the median return-stroke current peak for first strokes is two to three times higher than that for
subsequent strokes. Also, negative first strokes transfer about a factor of four larger total charge than
do negative subsequent strokes. On the other hand, subsequent return strokes are characterized by
three to four times higher current maximum rate of rise. Note that the smallest measurable time in
Berger et al.’s current oscillograms was 0.5 μs versus the 95% value of 0.22 μs for the front duration
for subsequent strokes in Table 2.1, which is a prediction of the log-normal approximation. The
maximum dI/dt in Table 2.1 is likely to be an underestimate: 50% value for subsequent strokes is 40
kA/μs vs. 100 kA/μs obtained using modern instrumentation for triggered-lightning strokes. Only a few
percent of negative first strokes are expected to exceed 100 kA. The action integral in Table 2.1
represents the energy that would be dissipated in a 1-Ω resistor if the lightning current were to flow
through it.
Typical values of return-stroke wavefront speed (based on optical measurements) are in the range of
one-third to one-half of the speed of light [29], as stated before. The equivalent impedance of the
lightning return-stroke channel is expected to be in the range from 0.6 to 2.5 kΩ [32], as estimated
from measurements of lightning current at different points along the 530-m-high Ostankino Tower in
Moscow, the radius of the lightning return-stroke channel is expected to be about 3 cm (e.g., [33]), and
the resistance per unit length of a lightning channel is estimated to be about 0.035 Ω/m behind the
return-stroke front and about 3.5 Ω/m ahead of the return-stroke front [33].
Upward lightning flashes can be also induced (triggered) by other lightning discharges occurring near
the tall object.

2.1.3 Positive lightning discharges to ground

Positive lightning discharges, defined as those transferring positive charge from cloud to ground,
account for only about 10% of all lightning discharges taking place between cloud and ground, but
they have lately attracted considerable attention of scientists and engineers. This is because positive
lightning discharges, which more often than their negative counterparts have higher currents and
larger charge transfers to ground, can cause more severe damage to various objects and systems
than negative lightning discharges. It is thought that positive lightning discharges tend to occur in the
following five situations [28]: (a) the dissipating stage of an individual thunderstorm, (b) winter
thunderstorms, (c) shallow clouds such as the trailing stratiform regions of mesoscale convective
systems, (d) severe storms, and (e) thunderclouds formed over forest fires or contaminated by smoke.

15
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 2.2 Average negative first- and subsequent-stroke currents each shown on two time scales, A and
B, as reported by Berger et al. [31]. The lower time scales correspond to the solid-line curves (A), while
the upper time scales correspond to the broken-line curves (B). The vertical scale is in relative units, the
peak values being equal to negative unity.2

According to the parameters reported from direct current measurements by Berger et al. [31] for
positive and negative lightning discharges, the 5% peak current for positive discharges is significantly
greater than that for negative first return strokes (250 kA vs. 80 kA), while the median peak current for
positive discharges is not much different from that for negative first return strokes (35 kA vs. 30 kA).
Also, the median charge transfer by positive discharges is about an order of magnitude greater than
that by negative discharges. All current waveforms observed by Berger et al. [31] for positive lightning
can be divided into two types. The first type includes microsecond-scale waveforms similar to those for
negative lightning, and the second type includes millisecond-scale waveforms with risetimes up to
hundreds of microseconds. While microsecond-scale waveforms are probably formed in a manner
similar to that in downward negative lightning, millisecond-scale waveforms are likely to be a result of
the M-component mode of charge transfer to ground [34]. Indeed, if a downward current wave
originates at a height of 1 to 2 km as a result of connection of the upward connecting leader to a
charged in-cloud channel, the charge transfer to ground associated with this wave is likely to be a
process of M-component type, which is characterized by a relatively slow current front at ground.
It is thought that positive discharges have the following characteristics [28]: (a) Positive flashes are
usually composed of a single stroke, whereas about 80% of negative flashes contain two or more
strokes. (b) Positive return strokes tend to be followed by continuing currents that typically last for tens
to hundreds of milliseconds. (c) Positive return strokes often appear to be preceded by significant in-
cloud discharge activity. (d) Positive lightning discharges often involve long horizontal channels, up to
tens of kilometers in extent. (e) Positive leaders can move either continuously or intermittently (as
seen in time-resolved optical records), while negative leaders are always stepped when they progress
in virgin air.

2 Reprinted with permission from K. Berger, R. B. Anderson, and H. Kroninger, Parameters of lightning flashes, Electra, vol. 80,
p.36, Figures 12 and 13, 1975.

16
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Table 2.1 Parameters of downward negative lightning derived from channel-base current measurements,
as reported by Berger et al. [31].3
Percentage exceeding
Sample tabulated value
Parameters Units
size
95% 50% 5%
Peak current (minimum 2 kA) kA
First stroke 101 14 30 80
Subsequent stroke 135 4.6 12 30
Charge (total charge) C
First stroke 93 1.1 5.2 24
Subsequent strokes 122 0.2 1.4 11
Complete flash 94 1.3 7.5 40
Impulse charge (excluding continuing current) C
First strokes 90 1.1 4.5 20
Subsequent strokes 117 0.22 0.95 4
Front duration (2 kA to peak) μs
First strokes 89 1.8 5.5 18
Subsequent strokes 118 0.22 1.1 4.5
Maximum dI/dt kA/μs
First strokes 92 5.5 12 32
Subsequent strokes 122 12 40 120
Stroke duration (2 kA to half peak value on the μs
tail)
First strokes 90 30 75 200
Subsequent strokes 115 6.5 32 140
Action integral A2s
First strokes 91 6.0×103 5.5×104 5.5×105
Subsequent strokes 88 5.5×102 6.0×103 5.2×104
Time interval between strokes ms 133 7 33 150
Flash duration ms
All flashes 94 0.15 13 1100
Excluding single-stroke flashes 39 31 180 900

2.1.4 Upward lightning discharges

Upward lightning, as opposed to downward lightning, would not occur if the grounded strike object
were not present. Hence, it can be considered to be initiated by the object. Objects with heights
ranging from approximately 100 to 500 m experience both downward and upward lightning flashes.
The fraction of upward flashes increases with the height of the object. Structures having heights less
than 100 m or so are usually assumed to be struck only by downward lightning, and structures with
heights greater than 500 m or so are usually assumed to experience only upward flashes. If a
structure is located on the top of a mountain, then an effective height that is greater than the structure

3 Reprinted with permission from K. Berger, R. B. Anderson, and H. Kroninger, Parameters of lightning flashes, Electra, vol. 80,
p. 27, Figure 1, 1975.

17
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

physical height is often assigned to the structure in order to account for the additional field
enhancement due to the presence of the mountain on which the structure is located. For example, the
two towers used by Berger in his lightning studies on Monte San Salvatore in Switzerland each had a
physical height of about 70 m while their effective height was estimated to be 350 m by Eriksson [35].
Eriksson’s estimate is based on the observed percentage of upward flashes initiated from towers of
different height. Note that upward flashes more often transport negative than positive charge to
ground.
Figure 2.3 shows schematic diagrams that illustrate upward lightning discharges with still and time-
resolved photographic records along with the corresponding current record at the channel base.
Upward negative discharges are initiated by upward positive leaders from the tops of grounded
objects. The upward positive leader bridges the gap between the object and the negative charge
region in the cloud and serves to establish an initial continuous current, typically lasting for some
hundreds of milliseconds. The upward positive leader and initial continuous current constitute the
initial stage of an upward flash. The initial stage can be followed, after a no-current interval, by one or
more downward-leader-upward-return sequences, as illustrated in Figure 2.3. Downward-leader-
upward-return-stroke sequences in upward lightning are similar to the subsequent leader-return-stroke
sequences in downward lightning.

Figure 2.3 Schematic diagram showing the luminosity of an upward negative flash and the corresponding
current at the channel base. (a) Still-camera image, (b) streak-camera image, (c) current record.4

2.2 Lightning surges

2.2.1 Surges due to direct lightning strike
The term “surges” denotes transient voltages or currents. Figure 2.4 schematically illustrates a
lightning strike to a phase conductor of a two-circuit transmission line (only one vertically-arranged
three-phase circuit is shown in this diagram). When lightning hits a phase conductor without being
intercepted by a shield or ground wire installed above the phase conductors, this phenomenon is
called shielding failure. In such a case, lightning surge currents propagate in both directions along the
phase conductor from the lightning attachment point. If a 10-kA lightning current is injected to a phase
conductor with characteristic impedance of 350 Ω, as a result of shielding failure, a 5-kA lightning
surge current and its associated 1750-kV (=350 Ω × 5kA) surge voltage propagates in each direction.
If this lightning surge voltage is higher than the withstand voltage between the phase conductor and a
nearby phase conductor or a ground wire or (most likely) the withstand voltage of insulator string at
the nearby transmission-line tower, a flashover would occur between them and likely result in ground
fault.

4Reprinted with permission from V. A. Rakov and M. A. Uman, Lightning: Physics and Effects, Cambridge University Press,
Cambridge, UK, p. 243, Figure 6.1, 2003.

18
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 2.5 illustrates a lightning strike to the top of an overhead power transmission line tower or to a
ground wire. In this case, lightning surge currents flow in the ground wire in both directions and in the
tower. When the tower is relatively small, the voltage at the tower is mainly determined by the product

Figure 2.4 Schematic diagram of a lightning strike to a phase conductor of a two-circuit overhead power
transmission line (only one circuit is shown in this diagram). Since lightning hits a phase conductor
without being intercepted by an overhead ground wire (also referred to as shield wire) installed above the
phase conductors, this phenomenon is called shielding failure.

Figure 2.5 Schematic diagram of a lightning strike to the top of an overhead power transmission-line
tower or to an overhead ground wire near the tower. If the resultant surge voltage at the tower is higher
than the withstand voltage of insulator strings, a flashover occurs and potentially results in ground fault.
This phenomenon is called back-flashover since it is due to the voltage rise at the tower, which is
normally at ground potential.

of its grounding impedance and the tower current. When the tower is relatively high, the voltage at the
tower top is mainly determined by the product of its characteristic impedance and the tower current. If
this lightning surge voltage generated at the tower is higher than the withstand voltage between arcing
horns, a flashover occurs and potentially results in a ground fault. This phenomenon is called “back-
flashover” since it is due to the voltage rise at a normally unenergized element: tower.

19
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Note that lightning surges due to shielding failure and back-flashover are similarly generated on
overhead power distribution lines. Lightning surge voltages due to shielding failure and back-flashover
are of primary interest in estimating lightning performance of power transmission and distribution lines
and in developing optimal design of the lines. Thus, it is essential to analyze lightning surges on power
transmission and distribution lines with sufficient accuracy.

2.2.2 Surges induced by a nearby lightning strike

Figure 2.6 illustrates a lightning strike to an object located close to an overhead power distribution line.
In such a case, transient electromagnetic fields generated by lightning illuminate a nearby overhead
wire, and induce surge currents and voltages on the wire. Lightning-induced surge voltages are often
a threat to overhead distribution power and metallic-conductor telecommunication lines, since the
withstand voltages of insulation of these lines are low compared with transmission lines. Note that in
the case of nearby lightning strike, a portion of lightning current flowing in the ground may enter the
power line via its grounding, in addition to electromagnetic coupling to an overhead conductor.

Figure 2.6 Schematic diagram of a lightning strike to a tree located near an overhead power distribution
line. In such a case, transient electromagnetic fields generated by lightning illuminate line conductors
and induce surge currents and voltages on them.

2.2.3 Surges coming from grounding due to its potential rise

Figure 2.7 schematically illustrates a lightning strike to a man-made metallic object, such as a radio-
relay-antenna tower, located close to an overhead power distribution line, both installed on poorly-
conducting ground. In such a case, the potential of the antenna tower foot may be high and surge
voltages enter the overhead distribution line via its grounded neutral (e.g., [36]). Note that in this case,
the resulting overvoltages will be a combination of the surges coming from the grounding, and induced
overvoltages due to the transient lightning electromagnetic field.

2.3 Summary
In this chapter, cloud-to-ground lightning discharges, which are of primary interest in evaluating and
designing lightning protection of man-made structures located on the ground surface, have been
described. Then, it has been explained how lightning surges are generated or invaded electric power
and telecommunication systems.

20
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 2.7 Schematic diagram of a lightning strike to a man-made metallic object, such as a radio-relay-
antenna tower, located near an overhead power distribution line, both installed on poorly-conducting
ground. In such a case, the potential rise of the antenna tower foot may result in surge voltages that enter
the overhead distribution line via its grounding.

21
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3. Fundamental theories of electromagnetic

computation methods
3.1 Finite-difference time-domain (FDTD) method
3.1.1 Introduction
The finite-difference time-domain (FDTD) method [16] is one of the most widely-used electromagnetic
computation methods for a variety of electromagnetic problems. The FDTD method uses the central
difference approximation to Maxwell’s curl equations, which are Faraday’s law and Ampere’s law, in
the time domain. Gauss’ law is also satisfied. The method solves the resultant update equations for
electric and magnetic fields at each time step and at each discretized space point in the working
volume using the leapfrog method. For the analysis of the electromagnetic response of a structure in
an unbounded space, an absorbing boundary condition such as Liao’s condition [37] or perfectly
matched layers [38], which suppresses unwanted reflections, needs to be applied.
In this section, update equations for electric and magnetic fields used in the three-dimensional (3D)
FDTD computation are given. An absorbing boundary condition, which is needed for the analysis of
electromagnetic fields in an unbounded space, is explained. Representations of lumped sources and
lumped circuit elements such as a resistor, an inductor and a capacitor are described. Also,
representations of a thin-wire conductor and the lightning return-stroke channel are discussed.
Further, representation of nonlinear elements such as a surge arrester, corona discharge on a high-
voltage overhead conductor, and back-flashover phenomena at an arcing horn of a transmission line,
twisted-wire pairs, and coaxial cables is explained.

3.1.2 FDTD expressions of Maxwell’s equations

The FDTD method in the 3D Cartesian coordinate system requires the whole working space, which
accommodates a conductor system to be analyzed, to be divided into cubic or rectangular
parallelepiped cells with side lengths Δx, Δy, and Δz, as shown in Figure 3.1. The electric field
components are placed at the midpoints of the sides of cells: Ex components are placed at the
midpoints of sides oriented in the x-direction, Ey components are placed at the midpoints of y-directed
sides, and Ez components are placed at the midpoints of z-directed sides. The magnetic field
components are placed at the centers of the faces of the cubic or rectangular parallelepiped cells, and
are oriented normal to the faces: Hx components are placed at the center points on yz-faces, Ey
components are placed at the center points on zx-faces, and Ez components are placed at the center
points on xy-faces. The electric field components are computed at integer time steps nΔt, where n is
an integer number and Δt is the time increment, and the magnetic field components are computed at
half-integer time steps (n+1/2) Δt.

Figure 3.1 Placement of electric field and magnetic field components in or on a cell.

Time-update equations for electric field components in x-, y-, and z-directions, Ex, Ey and Ez, are
derived from Ampere’s law, and those for magnetic field components, Hx, Hy, and Hz, are derived from
Faraday’s law. These are shown below.
Ampere’s law is given as follows:

22
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

1 1
n n
n
1
E 2 n
1
E 2 n
1
 H 2
 J 2
  E 2 Equation 3.1
t t
where H is the magnetic field vector, E is the electric field vector, J (=σE) is the conduction-current-
density vector, ε is the electric permittivity, σ is the electric conductivity, t is the time, and n-1/2 is the
time step number for the magnetic field computations. ε ∂E/∂t is the displacement-current-density
vector (due to time variation of the electric field vector). Ampere’s law states that the conduction
current and/or time-variation of electric field create magnetic field in the direction of right-hand curl. If
the time-derivative term in Equation (3.1) is approximated by its central finite difference, Equation (3.1)
is expressed as follows:
1
n
E 2 n
1
E n  E n 1 E n  E n 1 n
1
  E 2
    H 2 Equation 3.2
t t 2
Note that E n-1/2 in the second-term of Equation (3.2) is approximated by its average value, (E n +E n-
1)/2. If Equation (3.2) is rearranged, the update equation for the electric field vector at a time step

number n, E n, from its one time-step previous value, E n-1 ,and the half time-step previous magnetic-
field curl value   H
n1 / 2
is obtained as follows:
 t   t 
 1  2  n 1    n
1
E 
n
 E   t    H 2
Equation 3.3
 1  t   1 
 2   2 
From Equation (3.3), the update equation for Exn at a location (i+1/2, j, k) (see Figure 3.2 (a)), for
example, is expressed as follows:
  i  1/ 2, j , k  t t  n 
1
1  n 
1
1 
1  H z  i  , j , k  H y  i  , j , k  
2 2
 1 
Ex n  i  , j , k  
2  i  1/ 2, j , k   1 
Ex n 1  i  , j , k  
  i  1/ 2, j , k 
  2    2 
 2 
1
  i  1/ 2, j , k   t  2 
1
  i  1/ 2, j , k  t  y z 
2  i  1/ 2, j , k  2  i  1/ 2, j , k   
 

Equation 3.4

  i  1/ 2, j , k  t
1
2  i  1/ 2, j , k   1 
 E n 1 i  , j , k 
  i  1/ 2, j , k  t x  2 
1
2  i  1/ 2, j , k 
t  n  12  1 1  n 
1
1 1  
 H z  i  , j  , k  z  H z  i  , j  , k  z 
2
  i  1/ 2, j , k  1   2 2   2 2  

  i  1/ 2, j , k  t z y  n 
1
1 1 n 
1
1 1 
1   H y  i  , j , k   y  H y  i  , j , k   y 
2 2
2  i  1/ 2, j , k    2 2  2 2 

where the spatial derivative terms in Equation (3.4) are approximated by their central finite differences.
Update equations for Exn and Eyn are derived in the same manner, which are given below:
  i, j  1/ 2, k  t
1
 1  2  i, j  1/ 2, k   1 
E y n  i, j  , k   E n 1 i, j  , k 
 2    i, j  1/ 2, k  t y  2 
1
2  i, j  1/ 2, k 
Equation 3.5
t  n  12  1 1 n 
1
1 1 
 H x  i, j  , k   x  H x  i, j  , k   x 
2
  i, j  1/ 2, k  1   2 2  2 2 

  i, j  1/ 2, k  t z x  n 
1
1 1  n 
1
1 1  
1   H z  i  , j  , k  z  H z  i  , j  , k  z 
2 2
2  i, j  1/ 2, k    2 2   2 2  

23
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(a) (b)
n
Figure 3.2 (a) Electric field component in the x direction Ex at a location (i+1/2, j, k) and the circulating
magnetic field components closest to it and (b) magnetic field component in the x direction Hxn+1/2 at a
location (i, j+1/2, k+1/2) and the circulating electric field components closest to it.

  i, j , k  1/ 2  t
1
 1 2  i, j , k  1/ 2   1
Ez n  i, j , k    E n 1 i, j , k  
 2   i, j , k  1/ 2  t z  2
1
2  i, j , k  1/ 2 
Equation 3.6
t  n  12  1 1 n 
1
1 1 
 H y  i  , j , k   y  H y  i  , j , k   y 
2
  i, j , k  1/ 2  1   2 2   2 2  

  i, j , k  1/ 2  t xy  n 
1
1 1 n 
1
1 1 
1   H x  i, j  , k   x  H x  i, j  , k   x 
2 2
2  i, j , k  1/ 2    2 2  2 2 

Faraday’s law is given as follows:

H n
  E n   Equation 3.7
t
where μ is the permeability. Faraday’s law states that the time variation of magnetic field creates
electric field in the negative direction of right-hand curl. If the time-derivative term in Equation (3.7) is
approximated by its central finite difference, Equation (3.7) is expressed as follows:
1 1
n n
H n H 2
H 2
     E n Equation 3.8
t t
If Equation (3.8) is rearranged, the update equation for magnetic field at a time step number n+1/2 is
obtained from its one-time-step previous value H n-1/2 and the half time-step previous electric-field curl
value   E as follows:
n

t
H n 1/2  H n 1/2   En Equation 3.9

From Equation (3.9), the update equation for Hxn+1/2 at a location (i, j+1/2, k+1/2) (see Figure 3.2 (b)),
for example, is expressed as follows:

24
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

 1 n  1
1 1
n 1 1
 i, j  , k    H x  i, j  , k  
2 2
Hx
 2 2   2 2
 n 1 1 n 1 1 
t  Ez  i, j  2 , k  2  E y  i, j  2 , k  2  
      
  i, j  1/ 2, k  1/ 2   y z 
 

Equation 3.10
 1
1
n 1
 Hx 2
 i, j  , k  
 2 2
 n 1 n 1 
 Ez  i, j  1, k  2  z  Ez  i, j , k  2  z 
t 1      

  i, j  1/ 2, k  1/ 2  yz  n 1  n 1  
  E y  i, j  , k  1 y  E y  i, j  , k  y 
  2   2  

where the spatial derivative terms in Equation (3.10) are approximated by their central finite
differences. Update equations for Hyn+1/2 and Hzn+1/2 are derived in the same manner and given below:
 1 1 n  1
1 1
n 1
 i  , j, k    H y  i  , j, k  
2 2
Hy
 2 2   2 2

 n 1  n 1   Equation 3.11
 Ex  i  2 , j , k  1 x  Ex  i  2 , j , k  x 
t 1      

  i  1/ 2, j, k  1/ 2  z x  n 1 n 1 

 z 
E i  1, j , k   z  E z  i , j , k  
 
z
  2  2 

 1 1  n  1 
1 1
n 1
i  , j  ,k   Hz i  , j  ,k 
2 2
Hz
 2 2   2 2 

 n 1  n 1   Equation 3.12
 E y  i  1, j  2 , k  y  E y  i, j  2 , k  y 
t 1      

  i  1/ 2, j  1/ 2, k  xy  n 1  n 1  
  Ex  i  , j  1, k  x  Ex  i  , j , k  x 
  2   2  

By updating Exn, Eyn, Ezn, Hxn+1/2, Hyn+1/2, and Hzn+1/2 at every point in the working volume, transient
electric and magnetic fields throughout the working volume are obtained.
For the FDTD solution to be stable, the time increment Δt needs to be set to fulfill the Courant stability
condition [39] given as follows:
1
t 
1 1 1 Equation 3.13
c  
 x   y   z 
2 2 2

where c is the speed of light.

3.1.3 Absorbing boundary conditions

For the analysis of the electromagnetic response of a structure in an unbounded space, an absorbing
boundary condition, which suppresses unwanted reflections, needs to be applied to planes that
truncate the open space and accommodate the working volume. There are two types of absorbing
boundary conditions. One is a differential-based absorbing boundary condition such as Liao’s
condition [37], and the other is a material-based absorbing boundary condition such as perfectly-
matched layers (PMLs) [38]. PMLs are composed of thin layers whose conductivity starts with a very
small value at the free space-PML interface and gradually increases until it reaches its maximum
value at the last layer of the PML region. The PML is known to effectively absorb all the waves
propagating towards the absorbing boundaries. Here, Liao’s absorbing boundary condition is
explained since it is often used in lightning surge simulations with the FDTD method.
Figure 3.3 (a) shows the conceptual picture of a z-directed electric field Ez, which propagates in the
negative x direction with the speed of light c, and enters the absorbing boundary located at x=x1. The

25
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(a) (b)
Figure 3.3 (a) Conceptual picture of a z-directed electric field Ez propagating in the negative x direction
with the speed of light c and entering the absorbing boundary located at x=x1 and (b) electric-field
computation points near the absorbing boundary. 5

z-directed electric field at x1 at time step number n, Ezn(x1), could be estimated from Ezn-2(x1+2cΔt) and
Ezn-1(x1+cΔt) using a linear approximation, which is given below:
Ez n  x1   2 Ez n 1  x1  ct   Ez n  2  x1  2ct  Equation 3.14

Since locations of x1+2cΔt and x1+cΔt do not coincide with the electric-field computation points:
x1+2Δx, x1+Δx, and so on, as shown in Figure 3.3 (b), Ezn-2(x1+2cΔt) and Ezn-1(x1+cΔt) are estimated
using a quadratic interpolation, which is given below:
Ez n  x1   2T11 Ez n 1  x1   2T12 Ez n 1  x1  x   2T13 E z n 1  x1  2x 
 T112 Ez n  2  x1   2T11T12 Ez n  2  x1  x 
Equation 3.15
  2T11T13  T12 2  Ez n  2  x1  2x   2T12T13 Ez n  2  x1  3x 
 T132 Ez n  2  x1  4x 

where

T11 
 2  s 1  s  , T12  s  2  s  , T13 
s  s  1
s
ct
, Equation 3.16
2 2 x

Equation (3.15) is Liao’s second-order absorbing boundary condition. Note that the use of Equation
(3.15) in a single-precision floating-point computation often causes numerical instability (e.g., [40]). In
order to avoid this numerical instability, the following T11 is used.

T11 
 2  2d  s 1  s 
Equation 3.17
2

Uno suggests that d=0.0075 would be effective to supress numerical instability [41].

3.1.4 Representation of lumped sources and lumped circuit elements

3.1.4.1 Lumped voltage source
A lumped voltage source Vsn, in the z-direction, at a location (i, j, k+1/2) is represented by specifying
vertical electric field Ezn at the source location as follows:
 1
Vs n  i, j , k  
 1  2 Equation 3.18
Ez n  i, j , k    
 2 z

5 Reprinted with permission from T. Asada, Y. Baba, N. Nagaoka, and A. Ametani, A study of absorbing boundary condition for
surge simulations with the FDTD method, IEEJ Trans. Power and Energy, vol. 135, p. 414, app. Figure 1, 2015.

26
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3.1.4.2 Lumped current source

A lumped current source Isn-1/2, in the z-direction, at location (i, j, k+1/2) is represented by specifying
the z-component Jzn-1/2 of conduction-current density J n-1/2 in Equation (3.1) at the source location as
follows:

 1 1 n 12  1
1
n
Jz 2
 i , j , k    I s  i, j , k   Equation 3.19
 2  xy  2
Therefore, the update equation for Ez at (i, j, k+1/2) is given by
  i, j , k  1/ 2  t
1
 1 2  i, j , k  1/ 2   1
Ez n  i, j , k    E n 1 i, j , k  
 2   i, j , k  1/ 2  t z  2
1
2  i, j , k  1/ 2 
t  n  12  1 1 n 
1
1 1 
 y 
H i  , j , k   y  H 2
 i  , j , k   y 
  i, j , k  1/ 2  1   2 2
y
 2 2 
 Equation 3.20
  i, j , k  1/ 2  t xy  n  1 1 
2 
1 1
1 n  1
1   H x  i, j  , k   x  H x  i, j  , k   x 
2
2  i, j , k  1/ 2    2 2  2 2 
t
  i, j , k  1/ 2  1 n 
1
1
 I s 2  i, j , k  
  i, j , k  1/ 2  t xy  2
1
2  i, j , k  1/ 2 
Note that a lumped current source Isn-1/2, in the z-direction, at location (i, j, k+1/2) can be represented
in a simpler way by specifying the four circulating magnetic fields closest to the source, if Δx=Δy, as
follows [17]:
 1 n  1  1 n  1
1 1 1 1
n 1 1 n 1 1
 i, j  , k     I s 2  i, j , k   ,  i, j  , k    I s 2  i, j , k  
2 2
Hx Hx
 2 2 4x  2  2 2  4x  2
Equation 3.21
 1 1 n  1  1 1 n  1
1 1 1 1
n 1 n 1
 i  , j, k    I s 2  i, j , k   ,  i  , j, k     I s 2  i, j , k  
2 2
Hy Hy
 2 2  4y  2  2 2 4y  2

A lumped current source in the x- or y-direction is represented similarly to Equation (3.20) or (3.21).

3.1.4.3 Lumped resistance

A lumped resistance R, in the z-direction, at location (i, j, k+1/2) in a lossless medium (σ = 0) is
represented by specifying the z-component Jzn-1/2 of conduction-current density J n-1/2 in Equation (3.1)
at the lumped-resistance location as follows:
 1
1
n
 i, j , k   z
2
Ez
 1  1
1 1
n 1 n 1  2
 i, j , k     i, j , k   
2 2
Jz Iz
 2  xy  2  xy R
Equation 3.22
 1  1
E n i, j , k    Ez n 1  i, j , k  
z 1 z  2  2

xy R 2

Therefore, the update equation for Ez at (i, j, k+1/2) is given by

t z
1
 1 2 R  i, j , k  1/ 2  xy n 1  1
Ez n  i, j, k    Ez  i, j , k  
 2  1  t z  2
2 R  i, j , k  1/ 2  xy
t  n  12  1 1 n 
1
1 1 
 H y  i  , j , k   y  H y  i  , j , k   y 
2
  i, j , k  1/ 2  1   2 2   2 2  

t z x y  n 
1
1 n 
1
1 
1 1 1
  H x  i, j  , k   x  H x  i, j  , k   x 
2 R  i, j , k  1/ 2  xy
2 2

  2 2  2 2 
Equation 3.23
A lumped resistance in the x- or y-direction is represented in the same manner.

27
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3.1.4.4 Lumped Inductance

A lumped inductance L, in the z-direction, at location (i, j, k+1/2) in a lossless medium (σ = 0) is
represented by specifying the z-component Jzn-1/2 of conduction-current density J n-1/2 in Equation (3.1)
at the lumped-inductance location as follows:
n
1
1 n 
1
1 1 1  n  12  t  1
2  xy L 0
Jz 2
 I z 2  i, j , k    Ez  i, j , k   z dt
xy   2
Equation 3.24
1 z t n 1 m  1

xy L m 1
 Ez  i, j, k  2 
Therefore, the update equation for Ez at (i, j, k +1/2) is given by
 1  1
Ez n  i, j , k    Ez n 1  i, j , k  
 2  2
 n  12  1 1 n 
1
1 1 
 H y  i  , j , k   y  H y  i  , j , k   y 
2
t 1   2 2  2 2 

  i, j , k  1/ 2  xy  n 
1
1 1 n 
2 
1
1 1 
Equation 3.25
  H x  i, j  , k   x  H x  i, j  , k   x 
2

  2 2  2 2 
z  t 
2
n 1
 1

L   i, j , k  1/ 2  xy
E m 1
z
m
 i, j , k  
 2

A lumped inductance in the x- or y-direction is represented in the same manner.

3.1.4.5 Lumped capacitance

A lumped capacitance C, in the z-direction, at location (i, j, k+1/2) in a lossless medium (σ = 0) is
represented by specifying the z-component Jzn-1/2 of conduction-current density J n-1/2 in Equation (3.1)
at the lumped-capacitance location as follows:
 1
1
n

 i, j , k   z
2
dEz
 1
1 1
n

1 n 1  2
 i, j , k   
2 2
Jz Iz C
xy  2  x y dt Equation 3.26
1 C z  n  1  1 
 Ez  i, j , k    Ez n 1  i, j , k   
xy t   2  2 

Therefore, the update equation for Ez at (i, j, k +1/2) is given by

 1  1
Ez n  i, j , k    Ez n 1  i, j , k  
 2   2
t  n  12  1 1 n 
1
1 1 
 H y  i  , j , k   y  H y  i  , j , k   y 
2
  i, j , k  1/ 2  1   2 2   2 2  

C z x y  n 
1
1 n 
1
1 
1 1 1
  H x  i, j  , k   x  H x  i, j  , k   x 
  i, j , k  1/ 2  xy
2 2

  2 2  2 2 
Equation 3.27
A lumped capacitance in the x- or y-direction is represented in the same manner.

3.1.5 Representation of thin wire

Several representations of a thin wire for a 3D FDTD simulation have been proposed (e.g., [42]-[48]).
Here, the thin wire representation proposed by Noda and Yokoyama [44], which has been most
frequently used in surge simulations, is explained.
In [44], Noda and Yokoyama have shown that a straight, perfectly conducting wire in a lossless
medium, represented by forcing the tangential components of electric field along the wire axis to zero
in 3D FDTD simulations, has an equivalent radius a0 = 0.23Δs, where Δs is the lateral side length of
cells employed. Further, they have represented a wire having a radius a other than a0 by embedding
the wire of a0 = 0.23Δs in an artificial medium parallelepiped. In order to represent a thinner wire than
the wire having the corresponding equivalent radius, the relative permeability for calculating the
circulating magnetic field components closest to the wire needs to be increased and the relative
permittivity for calculating the radial electric field components closest to the wire decreased. In a lossy

28
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

medium, the conductivity also needs to be modified, similarly to the relative permittivity [45]. The
modified conductivity σ’, modified relative permittivity εr’, and modified relative permeability μr’ are
given as follows:
r
 '  m ,  r '  m r , r ' 
m
 s 
ln   Equation 3.28
m  0 ,
a
a0  0.23s
 s 
ln  
 a 
where σ, εr, and μr are the conductivity, relative permittivity, and relative permeability of the original
medium, and m is the modification coefficient.
Note that in representing a wire whose radius a is smaller than the equivalent radius a0, the modified
relative permeability μr’ is also employed in computing axial magnetic field components closest to the
wire in addition to the closest circulating magnetic field components in order to avoid numerical
instability [47], as shown (for a z-directed wire) in Figure 3.4 (a). Also, in representing a wire whose
radius a is larger than the equivalent radius a0, the modified relative permittivity εr’ is employed in
computing axial electric field components closest to the wire, in addition to the closest radial electric
field components [47], as shown in Figure 3.4 (b).

Figure 3.4 Cross-sectional views of a z-directed wire (central solid circle) having the radius a, and the
configuration of electric and magnetic field components closest to the wire: (a) a < a0 (m < 1) and (b) a >
a0 (m > 1).6

3.1.6 Representation of lightning return-stroke channel

3.1.6.1 Lightning return-stroke channel
There are eight types of representation of lightning return-stroke channel used in electromagnetic
pulse and surge computations [14], [18], [49], [50]:
(1) a perfectly conducting/resistive wire in air above ground;
(2) a wire loaded by additional distributed series inductance in air above ground;
(3) a wire surrounded by a dielectric medium (other than air) that occupies the entire half space above
ground (this fictitious configuration is used only for finding current distribution, which is then applied to
a vertical wire in air above ground for calculating electromagnetic fields);

6Reprinted with permission from Y. Taniguchi, Y. Baba, N. Nagaoka, and A. Ametani, An improved thin wire representation for
FDTD computations. IEEE Trans. AP, vol. 56, p. 3251, Figure 3, 2008.

29
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(4) a wire coated by a dielectric material in air above ground;

(5) a wire coated by a fictitious material having high relative permittivity and high relative permeability
in air above ground;
(6) two parallel wires having additional distributed shunt capacitance in air (this fictitious configuration
is used only for finding current distribution, which is then applied to a vertical wire in air above ground
for calculating electromagnetic fields);
(7) a phased-current-source array in air above ground, each current source being activated
successively by the arrival of lightning return-stroke wavefront propagating upward at a specified
speed; and
(8) a wire with many small radial branches in air above ground, which is called “bottle brush” model.
These eight channel representations are illustrated in Figure 3.5.

Figure 3.5 Schematic representations of lightning return-stroke channel.7

The return-stroke speed, along with the current peak, largely determines the radiation field initial peak
(e.g., [51]). The characteristic impedance of the lightning return-stroke channel influences the
magnitude of lightning current and/or the current reflection coefficient at the top of the strike object
when a lumped voltage source is employed. It is therefore desirable that the return-stroke speed and
the characteristic impedance of simulated lightning channel agree with observations that can be
summarized as follows:
(i) typical values of return-stroke speed are in the range from c/3 to c/2 [29], as observed using optical
techniques, where c is the speed of light;
(ii) the equivalent impedance of the lightning return-stroke channel is expected to be in the range from
0.6 to 2.5 kΩ [32].
Type (1) was used, for example, by Baba and Rakov in their FDTD simulation of electromagnetic
fields due to a lightning strike to flat ground [17]. Note that this lightning-channel representation was

7 Adapted from Y. Baba, and V. A. Rakov, Applications of the FDTD method to lightning electromagnetic pulse and surge
simulations, IEEE Trans. EMC, vol. 56, p. 1508, Figure 2, 2014.

30
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

first used by Podgorski and Landt [9] in their simulation of lightning current in a tall structure with the
method of moments (MoM) in the time domain [11]. The speed of the current wave propagating along
a vertical perfectly conducting/resistive wire is nearly equal to the speed of light, which is 2 to 3 times
larger than typical measured values of return-stroke wavefront speed (c/3 to c/2). This discrepancy is
the main deficiency of this representation. The characteristic impedance of the channel-representing
vertical wire varies with height above ground and for a radius of 3 cm is estimated to be around 0.6 kΩ
at a height of 500 m. This is right at the lower bound of its expected range of variation (0.6 to 2.5 kΩ).
Note that a current wave suffers attenuation (distortion) as it propagates along a vertical wire even if it
has no ohmic losses [52]. Further attenuation can be achieved by loading the wire by distributed
series resistance.
Type (2) was used, for example, by Baba and Rakov in their FDTD simulation of current along a
vertical lightning channel [49]. Note that this lightning-channel representation was first used by Kato et
al. in their simulation of lightning current in a tall structure and its associated electromagnetic fields
with the MoM in the time domain [53]. The speed of the current wave propagating along a vertical wire
loaded by additional distributed series inductance of 17 and 6.3 μH/m in air is c/3 and c/2,
respectively, if the natural inductance of vertical wire is assumed to be L0 = 2.1 μH/m (as estimated by
Rakov for a 3-cm-radius wire at a height of 500 m above ground [33]). The corresponding
characteristic impedance ranges from 1.2 to 1.8 kΩ (0.6 kΩ × [(17+2.1)/2.1]1/2=1.8 kΩ, and 0.6 kΩ ×
[(6.3+2.1)/2.1]1/2=1.2 kΩ) for the speed ranging from c/3 to c/2. The characteristic impedance of the
inductance-loaded wire is within the range of values of the expected equivalent impedance of the
lightning return-stroke channel. Note that additional inductance has no physical meaning and is
invoked only to reduce the speed of current wave propagating along the wire to a value lower than the
speed of light. The use of this representation allows one to calculate both the distribution of current
along the channel-representing wire and remote electromagnetic fields in a single, self-consistent
procedure. In [54], Bonyadi-ram et al. have incorporated additional distributed series inductance that
increases with increasing height in order to simulate the optically-observed reduction in return-stroke
speed with increasing height (e.g., [55]).
Type (3) was used, for example, by Baba and Rakov in their FDTD simulation of current along a
vertical lightning channel [49]. Note that this lightning-channel representation was first used by Moini
et al. in their simulation on lightning electromagnetic fields with the MoM in the time domain [56]. The
artificial dielectric medium was used only for finding current distribution along the lightning channel,
which was then removed for calculating electromagnetic fields in air. When the relative permittivity is 9
or 4, the speed is c/3 or c/2, respectively. The corresponding characteristic impedance ranges from
0.2 to 0.3 kΩ (0.6 kΩ/√9=0.2 kΩ, and 0.6 kΩ/√4=0.3 kΩ) for the speed ranging from c/3 to c/2. These
characteristic impedance values are smaller than the expected ones (0.6 to 2.5 kΩ).
Type (4) was used, for example, by Baba and Rakov in their FDTD simulation of current along a
vertical lightning channel [49]. Note that this lightning-channel representation was first used by Kato et
al. [57] in their simulation of lightning electromagnetic fields with the MoM in the frequency domain
[13]. In [49], Baba and Rakov represented the lightning channel by a vertical perfectly conducting wire,
which had a radius of 0.23 m and was placed along the axis of a dielectric rectangular parallelepiped
of relative permittivity 9 and cross-section 4 m × 4 m. This dielectric parallelepiped was surrounded by
air. The speed of the current wave propagating along the wire was about 0.74c. Such a representation
allows one to calculate both the distribution of current along the wire and the remote electromagnetic
fields in a single, self-consistent procedure, while that of a vertical wire surrounded by an artificial
dielectric medium occupying the entire half space (type (3) described above) requires two steps to
achieve the same objective. However, the electromagnetic fields produced by a dielectric-coated wire
in air might be influenced by the presence of coating.
Type (5) was first used by Miyazaki and Ishii in their FDTD simulation of electromagnetic fields due to
a lightning strike to a tall structure [58]. Although the exact values of relative permittivity and relative
permeability of the coating were not given by Miyazaki and Ishii [58], the speed of the current wave
propagating along the wire was about 0.5c. Similar to type (4), this representation allows one to
calculate both the distribution of current along the wire and the remote electromagnetic fields in a
single, self-consistent procedure. For the same speed of current wave, the characteristic impedance
value for this channel representation is higher than that for type (4), since both relative permittivity and
permeability are set at higher values in the type (5) representation.
Type (6) has not been used in lightning electromagnetic pulse (LEMP) and surge simulations with the
FDTD method to date. It was, however, used by Bonyadi-Ram et al. in their simulation with the MoM in
the time domain [59]. The speed of the current wave propagating along two parallel wires having

31
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

additional distributed shunt capacitance in air is 0.43c when the additional capacitance is 50 pF/m. In
this model, each of the wires has a radius of 2 cm, and the separation between the wires is 30 m.
Similar to type (3) described above, this representation employs a fictitious configuration for finding a
reasonable distribution of current along the lightning channel, and then this current distribution is
applied to the actual configuration (vertical wire in air above ground).
Type (7) was used by Baba and Rakov in their FDTD calculations of lightning electromagnetic fields
[17]. This representation can be employed for simulation of “engineering” lightning return-stroke
models. Each current source of the phased-current-source array is activated successively by the
arrival of lightning return-stroke wavefront that progresses upward at a specified speed. Although the
impedance of this channel model is infinity, appropriate reflection coefficients at the top and bottom of
the structure and at the lightning attachment point can be specified to account for the presence of a
tall strike object and upward-connecting leader (e.g., [60], [61]). Note that the MTLL and MTLE models
were implemented by Maslowski and Ziemba as Type (7) in their simulation of induced voltages inside
a conductor loop situated above flat lossy ground using the MoM in the frequency domain [62].
Type (8) has not been used in FDTD calculations but has a theoretical background including return
stroke models. The presence of many small radial branches effectively decreases the average speed
of current propagating upward [63].
Among the above seven types, types (2) and (5) appear to be best in terms of the resultant return-
stroke wavefront speed, the characteristic impedance, and the procedure for current and field
computations. Type (7) is also useful since the return-stroke wavefront speed and the current
attenuation with height are controlled easily with a simple mathematical expression: “engineering”
return-stroke models such as the transmission-line model [1], the traveling-current-source (TCS)
model [64], or the modified transmission line model with linear current decay (MTLL) [51]. Note that
numerical oscillations and numerical instability would occur when the TCS model is employed in the
FDTD simulations owing to its inherent discontinuity at the propagating current front. Practical aspects
of the implementation of various electromagentic models of lightning return storkes are discussed in
[65].

3.1.6.2 Excitations
Methods of excitation of the lightning channel used in electromagnetic pulse and surge computations
include
(1) closing a charged vertical wire at its bottom end with a specified impedance (or circuit);
(2) a lumped voltage source (same as a delta-gap electric-field source);
(3) a lumped current source; and
(4) a phased-current-source array.
Type (1) was used, for example, by in their FDTD simulation of currents along a vertical lightning
channel [49]. Note that this representation was first used by Podgorski and Landt in their simulation of
lightning currents with the MoM in the time domain [9]. Baba and Rakov represented a leader/return-
stroke sequence by a pre-charged vertical perfectly conducting wire connected via a nonlinear resistor
to flat ground [49]. In their model, closing a charged vertical wire in a specified circuit simulates the
lightning return-stroke process.
Type (2) representation was used, for example, by Baba and Rakov in their FDTD simulation of
lightning currents [49], but it was first used by Moini et al. in their simulation of lightning-induced
voltages with the MoM in the time domain [66]. This type of source generates a specified electric field,
which is independent of current flowing through the source. Since it has zero internal impedance, its
presence in series with the lightning channel and strike object does not disturb any transient
processes in them. If necessary, one could insert a lumped resistor in series with the voltage source to
adjust the impedance seen by waves entering the channel from the strike object to a value consistent
with the expected equivalent impedance of the lightning channel.
Type (3) was used, for example, by Noda [67]. However, in contrast with a lumped voltage source, a
lumped current source inserted at the attachment point is justified only when reflected waves returning
to the source are negligible. This is the case for a branchless subsequent lightning stroke terminating
on flat ground, in which case an upward connecting leader is usually neglected and the return-stroke
current wave is assumed to propagate upward from the ground surface. The primary reason for the
use of a lumped current source at the channel base is a desire to use directly the channel-base

32
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

current, known from measurements for both natural and triggered lightning, as an input parameter of
the model. When one employs a lumped ideal current source at the attachment point in analyzing
lightning strikes to a tall grounded object, the lightning channel, owing to the infinitely large impedance
of the ideal current source, is electrically isolated from the strike object, so that current waves reflected
from ground cannot be directly transmitted to the lightning channel (only electromagnetic coupling is
possible). Since this is physically unreasonable, a series ideal current source is not suitable for
modeling of lightning strikes to tall grounded objects [60].
Features of type (4) excitation are described in Section 3.1.6.1 for type (7) representation of lightning
return-stroke channel.

3.1.7 Representation of surge arresters

Tatematsu and Noda have proposed a technique to represent a surge arrester, the physical size of
which is much smaller than the wavelength of interest, by a lumped nonlinear resistor [68]. The
voltage versus current (V-I) characteristics of nonlinear resistors are represented by piecewise linear
curves as shown in Figure 3.6. The specific points on the characteristic are obtained from measured
voltage versus current curve. In Figure 3.6, Im and Vm represent the current and voltage at the mth
specified point, respectively, and the total number of points is denoted by M. The voltage versus
current characteristic of the nonlinear resistor shown in Figure 3.6 is approximated as follows:
n
1
 n 1  n
1
V  R0  I 2  I 0   V0
2
for V 2
 V1
 
n
1
 n
1
 n
1
V 2  Rm  I 2  I m   Vm for Vm  V 2
 Vm 1 1  m  M  3
 
Equation 3.29
n
1
 n 
1
 n
1
V 2  RM  2  I 2  I M  2   VM  2 for VM  2  V 2

 
Vm 1  Vm
Rm 
I m 1  I m

The voltage versus current characteristic for voltages smaller than V0 and larger than VM-1 are
represented by a linear extrapolation of (I0, V0) and (I1, V1), and (IM-2, VM-2) and (IM-1, VM-1), respectively.

Figure 3.6 Piecewise-linear representation of voltage vs. current characteristic of a surge arrester.8

The current through the nonlinear resistor is obtained from Equation (3.29) as follows:

8 Reprinted with permission from A. Tatematsu, and T. Noda, Three-dimensional FDTD calculation of lightning-induced voltages
on a multiphase distribution line with the lightning arresters and an overhead shielding wire, IEEE Trans. EMC, vol. 56, p. 160,
Figure 1, 2014.

33
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

1
n
n
1
V 2  Vm 
I 2
   Im  Equation 3.30
Rm  Rm 
When the nonlinear resistor is along the z-axis and located at point (i, j, k+1/2) in a lossless medium (σ
= 0), Equation (3.30) becomes:
 1 n  1
1 1
n
 i, j , k    J z  i, j , k   xy
2 2
Iz
 2   2
 1
1
n
 i, j , k    V
2
Vz
 2 
   m  Im  Equation 3.31
Rm  m
R 
 n 1  1 n 1 
 Ez  i, j , k  2   Ez  i, j , k  2   z 1  V 
   
   m  Im 
2 Rm  Rm 
From Equations (3.1) and (3.31), the update equation for Ez at (i, j, k +1/2) is given by
t
1
 1 2 Rm  i, j , k  1/ 2  z n 1  1
Ez  i, j , k   
n
Ez  i, j , k  
 2  1 t  2
2 Rm  i, j , k  1/ 2  z
t  n  12  1 1 n 
1
1 1 
 H y  i  , j , k   y  H y  i  , j , k   y 
2
  i, j , k  1/ 2  1   2 2   2 2  

t x y  n 
1
 n 
1
 
1 1 1 1 1
  H x 2  i, j  , k   x  H x 2  i, j  , k   x 
2 Rm  i, j , k  1/ 2  z  2  2 
 2 2
t
  i, j , k  1/ 2  1  Vm 
   Im 
t x y  m
R 
1
2 Rm  i, j , k  1/ 2  z
Equation 3.32
Since the voltage versus current relation of the lumped nonlinear resistor is given by a piecewise
linear function, the electric field along the nonlinear resistor is updated using Equation (3.32) in the
following simple procedure:
(i) Step 1: Update each electric field using Equation (3.32) with m=0 with the assumption that V n+1/2
satisfies the condition V n+1/2 < V1, then go to Step 2.
(ii) Step 2: If the computed V n+1/2 satisfies the assumption in Step 1, the computed electric field is
correct. Otherwise, go to Step 3 with m =1.
(iii) Step 3: If m = M-2, go to Step 5. Otherwise, update each electric field using Equation (3.32), with
m from Step 2 or from Step 4 and with the assumption that V n+1/2 satisfies the condition Vm ≤ V n+1/2 <
Vm+1, then go to Step 4.
(iv) Step 4: If V n+1/2 satisfies the assumption in Step 3, the computed electric field is correct.
Otherwise, add one to m, and go back to Step 3.
(v) Step 5: Update each electric field using Equation (3.32) with m = M-2 with the assumption that V
M-2 ≤ V
n+1/2 satisfies the condition V n+1/2, then go to Step 6.

(vi) Step 6: If V n+1/2 satisfies the assumption in Step 5, the computed electric field is correct.
In this procedure, the electric field along the surge arrester represented by a nonlinear resistor can be
obtained using Equation (3.32) M-1 times at the most.

3.1.8 Representation of corona discharge

Thang et al. proposed a simplified corona-discharge model for computing surges propagating on an
overhead wires using the FDTD method [69]. The radial progression of corona streamers from
overhead wire was represented as the radial expansion of a cylindrical conducting region whose
conductivity is several tens of microsiemens per meter [70].

34
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The critical electric field E0 on the surface of a cylindrical wire of radius r0 for initiation of corona
discharge is given by the following expression [71].

 0.1269 
E0  m  2.594 106 1  0.4346  [V/m] Equation 3.33
 r 
 0 
where m is a coefficient depending on the wire surface conditions.
The critical background electric field necessary for streamer propagation [72] (which determines the
maximum extent of the radially expanding corona region) for positive, Ecp, and negative, Ecn, polarity is
set as follows [73]:

 Ecp  0.5　　[MV/ m]
 Equation 3.34
 Ecn  1.5　　
[MV/ m]
It is shown in [74] that the statistical inception delay, streamer development process, and ionization
process, all of which are microsecond-scale phenomena, should be considered in developing a
corona-discharge model for lightning surge computations. In the FDTD computations, the ionization
process is roughly approximated by increasing the conductivity of the corona-discharge region from
zero to σcor=20 or 40 μS/m, and the statistical inception delay and streamer development process are
simply ignored. The time constant, CR=ε0/σcor(C and R are the capacitance and resistance of
cylindrical corona discharge region, respectively), is equal to about 0.5 or 0.25 μs, respectively. The
corona radius rc was obtained, using analytical expression (3.35), based on Ec (0.5 or 1.5 MV/m,
depending on polarity; see Equation (3.34)) and the FDTD-computed charge per unit length (q). Then,
the conductivity of the cells located within rc was set to σcor =20 or 40 μS/m.
q q
Ec   [V/m] Equation 3.35
2 0 rc 2 0 (2h  rc )

Equation (3.35), which is an approximation valid for rc  2h, gives the electric field at distance rc
below an infinitely long, horizontal uniform line charge, +q [C/m], located at height h above flat
perfectly conducting ground. A more general equation, not requiring that rc  2h, but assuming that
corona sheath is a good conductor, yields similar results.
Simulation of corona discharge implemented in the FDTD procedure is summarized below.
(a) If the FDTD-computed electric-field, Ezbn, at time step n and at a point located below and closest to
the wire (at 0.5 Δz from the wire axis shown in Figure 3.7 (a)), exceeds 0.46E0, the conductivity of σcor
=20 or 40 μS/m is assigned to x- and z-directed sides of the four cells closest to the wire. Note that
Ezbn is almost the same as Exl n and Exrn at points located on the left- and right-hand sides of the wire,
respectively, and closest to the wire (at 0.5Δx from the wire axis) (the difference is less than 1%).
Therefore, Ezbn is monitored only for determining initiation of corona discharge. Also note that neither
computed radial current nor q-V curves change even if the same conductivity is also assigned to y-
directed (axial direction) sides of the four cells. Also note that E0 is given by Equation (3.33).
(b) The radial current In per unit length of the wire at y=jΔy from the excitation point at time step n is
evaluated by numerically integrating radial conduction and displacement current densities as follows.


I n ( j y )    Exln  Exr

n
 
z  Ezan
 Ezb
n

x  y

 Exln  Exln 1 E xnr  Exr
n 1  n 1
 E n  E za E n  E znb1  
Equation 3.36
 0    z   za  zb  x  y
 t t 


 t t  

whereExl, Exr, Eza, and Ezb are radial electric fields closest to the wire shown in Figure 3.7 (b). The
total charge (charge deposited on the wire and emanated corona charge) per unit length of the wire at
y=jΔy from the excitation point at time step n is calculated as follows:

I n 1 ( j y )  I n ( j y )
q n ( j y )  q n 1 ( j y )  t Equation 3.37
2
From qn yielded by Equation (3.37) and Ec given by Equation (3.34), the corona radius rcn+1 at time
step n + 1 is calculated using Equation (3.35). The conductivity of σcor =20 or 40 μS/m is assigned to
the x- and z-directed sides of all cells located within rcn+1.

35
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 3.7 FDTD representations. (a) Inception of corona discharge at the wire surface and (b) Radial
expansion of corona discharge.9

3.1.9 Representation of breakdown characteristics of arcing horns

Several models to simulate the breakdown characteristics of a long air gap such as arcing horns
installed in high-voltage transmission lines have been proposed for circuit-theory-based simulations
(e.g., [75]-[77]). These models can take into account the effect of predischarge currents to decay
voltages across an air gap by simulating leader development in the gap. Compared with other models,
the flashover model proposed by Motoyama [75] has the following advantages: (i) This model was
developed on the basis of the measured breakdown characteristics of a long air gap with a gap length
of 1 to 3 m for a short-tail lightning impulse, which occurs across an arcing horn at a transmission line
tower struck by lightning; (ii) This model can be easily incorporated into time-domain transient
electromagnetic simulation programs such as EMTP.
In [78], Tatematsu et al. proposed a technique to apply the above flashover model to FDTD-based
surge simulations on the basis of the assumption that the gap length is sufficiently small compared
with the minimum wavelength, and, in general, this assumption is satisfied in FDTD-based lightning
surge simulations. In this technique, the flashover model is treated as a lumped-parameter element,
and it is placed on the section of a cell in a lossless medium. For simplicity, here it is assumed that the
analysis space is divided uniformly into cubic cells with a section of Δs and the flashover model is
placed in the z-direction. To incorporate the flashover model into the FDTD method, equations to
simulate the leader development process and so forth are discretized as follows [75], [78]-[80]:
(i) Leader onset condition:
V n  500  D Equation 3.38

where V [kV], D [m], and the superscript n denote the voltage across a gap, the gap length, the time
nΔt (Δt: time discretization), respectively.
(ii) Leader development process:
I L n 1/ 2  2 K 0 L n 1/ 2 Equation 3.39

9 Reprinted with permission from T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov, A
simplified model of corona discharge on overhead wire for FDTD computations, IEEE Trans. EMC, vol. 54, p. 587, Figure 3,
2012.

36
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

  V n 1/ 2 
 K '1   E0  　　　　(0  X L n  D / 4)
  D  2 X n

 L n 1/ 2
L
  Equation 3.40
  V n 1/ 2

 ( D / 4  X L n  D / 2)
 K ''1  D  2 X n  E '0    L 　　
  L 

X L n 1  X L n   L n 1/ 2 t Equation 3.41

where IL [A], vL [m/s], and XL [m] represent the leader current, leader development speed, and average
leader length (the average of the upper and lower leaders), respectively. The constants in Equation
(3.39) and Equation (3.40) are defined on the basis of the measured results as follows: K0 = 410
[μC/m], K’1 = 2.5 [m2/Vs], K’’1 = 0.42 [m2/Vs], E0 = 750 [kV/m]. E’0 and υ’L in Equation (3.40) are
obtained by the following procedure. In the case of 0 ≤ XL ≤ D/4, V/(D−2XL) and υL are calculated at
each time; the values of V/(D−2XL) and υL immediately before XL exceeds D/4 are used as E’0 and υ’L,
respectively, in the case of D/4 < XL ≤ D/2, which are treated as constant values.
(iii) Criterion of breakdown:
X Ln  D / 2 Equation 3.42

(iv) Criterion to stop leader development process:

Vn
 E0 Equation 3.43
D  2 X Ln

The flashover model can be represented in the FDTD method by taking into account the above
equations in updating the electric field across the flashover model. For example, in the case of 0 ≤ XLn
≤ D/4, substituting the first equation in Equation (3.40) and Vn+1/2 = (Ezn+1 + Ezn)Δs/2 into Equation
(3.39) gives:
 ( E n 1  Ez n )s 
I L n 1/ 2  2 K 0 K '1  z  E0  Equation 3.44
 2( D  2 X L )
n


The electric field Ez in a lossless medium is updated by the following equation:

t t
Ez n 1  Ez n    H n 1/ 2  J z n 1/ 2 Equation 3.45
 
where H, Jz, and ε denote the magnetic fields, current density, and permittivity, respectively.
Substituting Equation (3.44) and Jzn+1/2=ILn+1/2/Δs2 into Equation (3.45), the following equation to
update the electric field across the flashover model can be obtained:
1 A t B
Ez
n 1
 Ez 
n
 H
n 1/ 2
 Equation 3.46
1 A (1  A) 1 A

tK 0 K '1 2tK0 K1E0

A ,B  Equation 3.47
s( D  2 X L )
n
s 2
The equations used to update the electric fields across the flashover models in the other directions
can be readily derived in the same manner.
The incorporation of the flashover model into the FDTD method in the aforementioned manner may
cause numerical instability due to its dynamic nonlinear characteristics. Although one of the
countermeasures to avoid such numerical instability is to set the time discretization to a value much
smaller than the limit of Courant’s condition, this method results in a much longer calculation
time.Tatematsu et al. proposed a technique to avoid both numerical instability and an increase in the
calculation time [78], [79]. In this technique, as shown in Figure 3.8, the analysis space simulated by
the FDTD method is divided into two regions, that is, a small region encompassing a source of
numerical instability such as the flashover model and the remainder of the analysis space. The former
and latter regions are called the stabilized region and normal region, respectively, hereafter. The time
discretization Δt in the normal region is set to a value close to the limit of Courant’s condition Δtc, for
example, Δt =0.99Δtc, while the time discretization in the stabilized region is reduced to Δt/m (m:
integer) to suppress numerical instability. Therefore, in this technique, the electric and magnetic fields
in the normal and stabilized regions are alternately placed at intervals of Δt/2 and Δt/(2m),

37
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

respectively. This means that the number of time steps in the stabilized region is m times larger than
that in the

Normal region： Δ t 　 Δ t c Analysis space

simulated
Stabilized region : Δ t /m by the FDTD method
： models that may cause
numerical instability such as
flashover models

Figure 3.8 Normal and stabilized regions in the analysis space.10

normal region. However, when the FDTD method is applied to lightning surge analysis, in general, the
volume of the stabilized region is much smaller than that of the whole analysis space, and so the
application of this technique does not increase the calculation time significantly. Note that multiple
stabilized regions in the analysis space simultaneously can be applied, not a single stabilized region.
In the stabilized and normal regions, the electric and magnetic fields are updated from (n−1/2)Δt to
(n+1/2)Δt by the following two steps: the first step is from t = (n−1/2)Δt to nΔt and the second step is
from t = nΔt to (n+1/2)Δt. As described later, to simplify the synchronization of the electric and
magnetic fields in the two regions at an interval of Δt/2, the integer m is assumed to be an odd positive
number. In what follows, as an example, the procedure to update the electric and magnetic fields in
the two regions in the case of m = 3 is described, and the time line used to update the electric and
magnetic fields is shown in Figure 3.9.

1
n 1 n n n
Normal H 2 E E H 2

region
1 1 1 1 1 1
n n n
Stabilized H n 2
E
n 3
H
n 6
E
n
E
n
H 6
E 3
H 2

region
First step Second step
Figure 3.9 Time line of the electric and magnetic fields in the case of m = 3.11

(i) First step from (n−1/2)Δt to nΔt

The electric fields in the normal region and on the boundary between the two regions are updated
using the normal expression with a time discretization of Δt. On the other hand, the electric and
magnetic fields in the stabilized region are updated by the following equations:
t / 3
E n 1 / 3  E n  2 / 3    H n 1 / 2 Equation 3.48

t / 3
H n 1 / 6  H n 1 / 2    E n 1 / 3 Equation 3.49

Although the electric fields on the boundary are calculated in the normal region at an interval of Δt, the
values at a time of (n−1/3)Δt are necessary to update the magnetic fields adjacent to the boundary
between the two regions. For example, as shown in Figure 3.10, the electric field on the boundary,
Ezn−1/3(i, j+1, k+1/2), which is calculated at an interval of Δt in the normal region, is required to
calculate the magnetic field Hxn−1/6(i, j+1/2, k+1/2). In this technique, Ezn−1/3(i, j+1, k+1/2) is obtained by
interpolating Ezn−2(i, j+1, k+1/2), Ezn−1(i, j+1, k+1/2), and Ezn(i, j+1, k+1/2) calculated in the normal

10Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1552, Figure 1, 2016.
11Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1552, Figure 2, 2016.

38
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

region based on the second-order function. As shown in Figure 3.11, the normalized coordinate ξ is
defined on the time axis, and E(ξ) is expressed by Equation (3.50).

Stabilized n 13 Normal
Ey (i, j+ 12 , k+1)
region region
n 16
Hx (i, j+ 12 , k+ 12 )
1 n 13
E nz 3 (i, j, k+ 12 ) Ez (i, j+1, k+ 12 )

n 13
Ey (i, j+ 12 , k)

Figure 3.10 Arrangement of the electric and magnetic fields at the boundary in the first step (m = 3).12

Normalized
Time
coordinate
n 2
E t = (n 2)Δt ξ= 1

n 1
E t = (n 1)Δt ξ= 0
1
n 1)Δt
E 3 t = (n 3 ξ= 23
n
E t = nΔt ξ= 1

Figure 3.11 Normalized coordinate in the time axis in the first step.13

E(  )  a E n  2  b E n 1  c E n Equation 3.50

a  0.5 (  1)

b  (1   )(1   ) Equation 3.51
  0.5 (  1)
 c
Substituting ξ = 2/3, which corresponds to t = (n−1/3)Δt, in the above equations gives Ezn−1/3(i, j+1,
k+1/2).
(ii) Second step from nΔt to (n+1/2)Δt
The magnetic fields in the normal region and on the boundary are updated using the normal
expression at an interval of Δt. The electric and magnetic fields in the stabilized region and the electric
fields on the boundary are updated by Equations (3.48) and (3.49) at an interval of Δt/3. Similar to the
first step, to update the electric fields on the boundary, the magnetic fields at a time of (n+1/6)Δt are
necessary, but some of them are calculated in the normal region. For example, as shown in Figure
3.12, the calculation of Exn+1/3(i+1/2, j, k) in the stabilized region requires Hyn+1/6(i+1/2, j, k−1/2),
Hyn+1/6(i+1/2, j, k+1/2), and Hzn+1/6(i+1/2, j+1/2, k), which are updated at an interval of Δt in the normal
region. In a similar way to the first step, these magnetic fields at a time of (n+1/6)Δt are obtained by
interpolating the magnetic fields at times of (n−3/2)Δt, (n−1/2)Δt, and (n+1/2)Δt, which are calculated
in the normal region. Using the normalized coordinate ξ on the time axis, which respectively
corresponds to −1 and 1 at times of (n−3/2)Δt and (n+1/2)Δt, H (ξ) is expressed by:
H(  )  a H n 3 / 2  b H n 1 / 2  c H n 1 / 2 Equation 3.52

12Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1552, Figure 3, 2016.
13Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1552, Figure 4, 2016.

39
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

where ϕa, ϕb, and ϕc are given by Equation (3.51). From the above equation, the magnetic fields at a
time of (n+1/6)Δt are calculated at ξ equal to 2/3.

Hy
n 1
6 (i+ 12 , j, k+ 12 )
Normal
region
n 1
Ex 3 (i+ 12 , j, k)

n 16 n 16
Hz (i+ 1
2 ,j 1
2 , k) Hz (i+ 12 , j+ 12 , k)

Stabilized 1
n
region Hy 6 (i+ 12 , j, k 1
2)

Figure 3.12 Arrangement of the electric and magnetic fields at the boundary in the second step (m = 3).14

Note that the above-described technique can be readily extended to other odd integers m larger than
3.
The long-air-gap flashover model described above was developed on the basis of the measured
breakdown characteristics of a rod-rod gap with a gap length of 1 to 3 m. In [79], Tatematsu and Ueda
extended it to a short air gap with a gap length of 0.65 m to simulate the breakdown characteristics of
short-air-gap (0.65 m) arcing horns installed in 77-kV transmission lines. As mentioned above, the
long-air-gap flashover model has the constants K’1 and K’’1, which represent the relationship between
the leader development speed and the electric field at the gap. Comparing the calcualted results with
measured results, Tatematsu and Ueda confirmed that the breakdown characteristics of the short air
gap with a gap length of 0.65 m can be simulated by setting K’1 and K’’1 to 0.70 and 0.42, respectively,
in the flashover model [79].

3.1.10 Representation of breakdown characteristics of transmission line surge

arresters with air gaps
In [79], Tatematsu and Ueda proposed a technique for representing transmission line surge arresters,
which are installed in 77-kV transmission lines and comprise surge arresters and short air gaps (see
Figure 3.13 (a)). As shown in Figure 3.13 (b), the proposed model is composed of two models, that is,
the surge arrester model described in Section 3.1.7 and the flashover model described in Section
3.1.9. Using the surge arrester and the flashover model, Tatematsu and Ueda presented the
calculated breakdown characteristics of the transmission line surge arrester with an air gap of 0.35 m
when the constants K’1 and K’’1 representing the relationship between the leader development speed
and the electric field at the gap in the flashover model are set to 0.30 and 0.42, respectively, are in
good agreement with results measured by Ueda et al. [81].

Air gap

Surge arrester Surge arrester Flashover model

(a) (b)
Figure 3.13 Transmission line surge arrester. (a) Configuration and (b) Proposed model.15

14Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1552, Figure 5, 2016.
15Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1553, Figure 8, 2016.

40
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3.1.11 Representation of effects of AC voltages

It is well known that AC voltages applied to phase conductors influence the breakdown characteristics
of arcing horns installed at transmission line towers. In the case of lightning surge analysis using
circuit-theory-based simulation programs such as EMTP, the effect of the AC voltages is easily
simulated by obtaining steady-state solutions before the lightning current is injected. Tatematsu and
Ueda proposed a technique for taking account of the effect of AC voltages in FDTD-based surge
simulations [79].
The frequency range of lightning surge analysis is much higher than that of AC, and so the effect of
AC can be treated as that of DC. On the basis of this concept, as shown in Figure 3.14, a voltage
source is inserted into each phase conductor to apply a voltage to simulate the effect of AC voltages,
and the waveform of the voltage produced by the source has a quasi-steady state such as that of step
waves. Here, the output voltage of each source Vi(t) (i=1–3) in a quasi-steady state is denoted by Vpi,
respectively. In the arrangement in Figure 3.14, one of the two phase conductors connected to the
voltage source is attached to an absorbing boundary to avoid the occurrence of a reflected traveling
wave.

Voltage source
Vp1
Phase conductor

Vp2 V11
Absorbing V12
boundary V13
Vp3 V21
V22
V31 V23
V32
V33

Ground

Figure 3.14 Voltage sources inserted in phase conductors to simulate the effect of AC voltages. 16

The procedure to simulate the effect of AC voltages on each phase conductor is as follows:
(i) Step 1: Set Vp1, Vp2, and Vp3 to 1, 0, and 0, respectively, and obtain the voltage of each phase
conductor (V11, V21, and V31) in a quasi-steady state.
(ii) Step 2: Set Vp1, Vp2, and Vp3 to 0, 1, and 0, respectively, and obtain the voltage of each phase
conductor (V12, V22, and V32) in a quasi-steady state.
(iii) Step 3: Set Vp1, Vp2, and Vp3 to 0, 0, and 1, respectively, and obtain the voltage of each phase
conductor (V13, V23, and V33) in a quasi-steady state.
(iv) Step 4: Solve the following linear equation, and obtain Vp1, Vp2, and Vp3 to simulate the specified
voltage of each phase conductor (Vs1, Vs2, and Vs3) in a quasi-steady state:
1
 V p1   V11 V12 V13   Vs1 
     
 V p 2    V21 V22 V23   Vs 2  Equation 3.53
V  V V V33   Vs 3 
 p 3   31 32

(v) Step 5: Set Vp1, Vp2, and Vp3 obtained in step 4, and carry out an FDTD-based surge simulation
until the voltage of each phase conductor leads to a quasi-steady solution, and then inject a lightning
current into a calculation model.
Note that although the technique in the case of a three-phase transmission line conductor is decribed,
it can be readily extended to other conductor arrangements such as a double-circuit transmission line
conductor.

16Reprinted with permission from A. Tatematsu and T. Ueda, FDTD-based lightning surge simulation of an HV air-insulated
substation with back-flashover phenomena, IEEE Trans. EMC, vol. 58, p.1554, Figure 12, 2016.

41
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3.1.12 Representation of twisted-wire pairs

Tatematsu et al. proposed a hybrid method for evaluating the induced voltages on a twisted-wire pair
illuminated by a nonuniform incident electromagnetic field [82]. The exciting electromagnetic fields are
first obtained by a 3D FDTD method with an orthogonal coordinate system and then the induced
voltages are calculated using the Agrawal et al. model [4], where the exciting electric field along each
wire of the twisted pair is obtained by interpolating electric fields calculated by the 3D FDTD method.
In the Agrawal et al. model, assuming lossless lines for simplicity, voltages and currents induced on an
N-conductor transmission line along the line are calculated by solving the following equations:

V s (l , t ) I (l , t )
 L '(l )  Ete (l , t ) Equation 3.54
x t

I (l , t ) V S (l , t )
 C '(l ) 0 Equation 3.55
x t
where Vs(l,t) and I (l,t) are the vectors of scattered voltages and currents along the conductors,
respectively. L '(l ) and C '(l ) , respectively, denote the per-unit-length line inductance matrix (N × N)
e
and capacitance matrix (N × N). Et (l , t ) is the vector of exciting electric fields tangential to the
conductors and evaluated in their absence. Finally, the vector of the total induced voltages V(l,t) is
obtained by the following equations:
hi
V (l , t )  V s (l , t )   href
E ze (l , t )dz Equation 3.56

where E ze (l , t ) is the vector of the exciting vertical electric fields, which is integrated between the
reference conductor (ground plane) and each conductor of the twisted pair. Note that the above
equations can be extended for lossy lines and lines surrounded by lossy media [83]. In this study, the
Agrawal et al. model is applied to the calculation of voltages induced on a twisted-wire pair (N = 2) and
solve Equations (3.54) and (3.55) in the time domain, where the length of the conductors corresponds
to the total length of each wire of the twisted-wire pair Ltwp, and not the horizontal distance between
both ends of the twisted-wire pair Dtwp, similar to other studies (e.g., [84], [85]). The tangential electric
fields Ete (l , t ) on each wire of the twisted-wire pair and the vertical fields E ze (l , t ) under each wire are
obtained from results calculated by the 3D FDTD method as explained in what follows.
First, it will be described how to specify positions on each wire of the twisted-wire pair, and secondly
will be explained how to calculate electric fields on and under each wire from the FDTD results. Here,
the local Cartesian coordinate system (p, q, r) centered at the left end of the twisted-wire pair placed
above the ground plane is considered, as shown in Figure 3.15. With the assumption that the twisted-
wire pair is an ideal bifilar helix, the position coordinates (p1, q1, r1) and (p2, q2, r2) on each wire of the
twisted-wire pair with an arc length l from the end are specified by the following expressions [84], [86]:

p (p1, q11, r1 )
l
q d r
(p2, q12, r2 )
g (= twist pitch)
Ground plane

Figure 3.15 Configuration of a twisted-wire pair above the ground plane.17

17 Reprinted with permission from A. Tatematsu, F. Rachidi, and M. Rubinstein, A technique for calculating voltages induced on
twisted-wire pairs using the FDTD method, IEEE Trans. EMC, vol. 59, p. 302, Figure 1, 2017.

42
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

 p 1  (d / 2) cos( l )
 2

q 12  (d / 2) sin( l ) Equation 3.57

 r   gl / (2 )
 12
where d and g are the distance between the two wires and the twist pitch, respectively, and α is given
by:
1/ 2
   d / 2    g / (2 ) 
2 2
Equation 3.58
 

In this case, the horizontal distance Dtwp between both ends of the twisted-wire pair is expressed by:

Dtwp   gLtwp /  2  Equation 3.59

The tangential unit vectors along the two wires at position l, (dp1/dl, dq1/dl, dr1/dl) and (dp2/dl, dq2/dl,
dr2/dl), are obtained from Equation (3.57) as follows:

 dp 12
  (d / 2)sin( l )
 dl
 dq 1

2
 (d / 2) cos( l ) Equation 3.60
 dl
 dr1
 2   g / (2 )
 dl

The tangential electric fields El1 and El2 on the two wires at position l along the wire are obtained by
multiplying the tangential unit vectors in Equation (3.60) by the electric fields at position l, (Ep1, Eq1,
Er1) and (Ep2, Eq2, Er2), which are obtained from the FDTD results as described below.
The electric fields are evaluated in the center of the sections of cells in the FDTD method using an
orthogonal coordinate system. To calculate electric fields at arbitrary positions (namely along or under
the wires) from FDTD results, interpolation functions are applied. As shown in Figure 3.16, to calculate
Ex at an arbitrary point from the electric fields Exi (i = 1–8), which are placed on the centers of the cell
sections surrounding Ex, first, ExA is calcaulted by interpolating Ex1–Ex4 with the basis function used for
the first-order element in the finite element method [87]. A local coordinate system in the rectangle
whose corners correspond to the positions of Ex1–Ex4 is designated, as shown in Figure 3.17 (a), and
then ExA is obtained by the following interpolation function:
ExA  14 (1   )(1   ) E x1  1
4
(1   )(1   ) E x 2  1
4
(1   )(1   ) E x 3  1
4
(1   )(1   ) E x 4 Equation 3.61

In the same manner, ExB can be obtained by applying the basis function to Ex5–Ex8. Finally, as shown
in Figure 3.17 (b), a one-dimensional local coordinate system in the straight line through the positions
of ExA, ExB, and Ex is designated, and Ex can be obtained by the following interpolation function
Ex  1
2
(1  s ) E xA  1
2
(1  s ) E xB Equation 3.62

43
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Ex8
Ex4 Ex3 Ex7
Ex ExB

Ex1 Ex5
z Ex6
y x ExA Ex2
Figure 3.16 Arrangement of the electric fields to calculate Ex by an interpolation technique.18

The other electric field components Ey and Ez at an arbitrary point can be calculated in a similar
manner. In addition to the above-described spatial interpolation, electric fields at an arbitrary time t can
be calculated using a similar interpolation technique. Higher-order interpolation techniques can be also
applied.

η
η=1
Ex3 Ex4
ExA Ex ExB
ξ s
ExA s = −1 s=1
Ex2 Ex1
η = −1

ξ = −1 ξ=1
(a) (b)
Figure 3.17 Local coordinate system for interpolation functions. (a) Two-dimension and (b) One-
dimension.19

Once the exciting electric field components are determined using the above-described approach, the
Agrawal et al. Equations (3.54) and (3.55) are solved using a one-dimensional FDTD technique,
where partial time and space derivatives are approximated by first-order functions.

3.1.13 Representation of lossy thin wire

It is not straightforward to take into account frequency-dependent lossy effects in the FDTD method,
and Holland and Simpson proposed a techniuqe for modeling a lossy thin wire simply by a constant
resistance [88]. Bingle et al. proposed a technique for simulating the effect of the conductor internal
impedance of a thin wire using an analytical expression [89], but this technique was developed for
microwave (S-band) simulations, not for lightning or switching surge simulations. Recently, Du et al.
proposed a technique to simulate the effect of the conductor internal impedance of a thin wire [90]. In
this technique, it is necessary to solve the current distribution in the cross-section of the thin wire at
each time step by the difference method in addition to updating electromagnetic fields in an analysis
space.
In [91], Tatematsu proposed a technique for taking into account the effect of the conductor internal
impedances of a thin wire in a straightforward manner by using a series of lumped-parameter
impedances placed along the lossy conductor. The frequency-dependent parameters of the
impedances are given by analytical expressions.

18 Reprinted with permission from A. Tatematsu, F. Rachidi, and M. Rubinstein, A technique for calculating voltages induced on
twisted-wire pairs using the FDTD method, IEEE Trans. EMC, vol. 59, p. 302, Figure 2, 2017.
19 Reprinted with permission from A. Tatematsu, F. Rachidi, and M. Rubinstein, A technique for calculating voltages induced on
twisted-wire pairs using the FDTD method, IEEE Trans. EMC, vol. 59, p. 303, Figure 3, 2017.

44
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The internal impedance Ztw in a solid conductor in the frequency domain is expressed as follows [92]:
 I 0 ( tw atw1 )
Ztw  2 atwtw1 tw I1 ( tw atw1 ) Equation 3.63

 tw  jtw tw Equation 3.64

where atw1, μtw, and σtw denote the radius, permeability, and conductivity of the conductor, respectively.
I0(∙) and I1(∙) are respectively the modified Bessel functions of the zeroth and first order of the first kind.
On the other hand, the internal impedance in a tubular conductor is obtained using the following
approximate expression [93]:
 T
Ztw  2 atwtw3twtwTtw coth  twTtw  Wtw Equation 3.65

Ttw  atw3  atw 2 Equation 3.66

Wtw  2 a 1
 atw 2  atw 3  tw Equation 3.67
tw 3

where atw2 and atw3 denote the inner and outer radii of the tubular conductor, respectively.
Assuming that a lossy conductor can be simulated by placing a series of lumped-parameter
impedances along the conductor, as shown in Figure 3.18, the relationship V(ω) = ZtwΔsI(ω) in the
Lumped-parameter impedances to simulate the
effect of the conductor internal impedance

Thin wire Δs
Δs

Figure 3.18 Series of lumped-parameter impedances to simulate the conductor internal impedance of the
thin wire.20

frequency domain is obtained, where V and I denote the voltage across the cell and the current
flowing through the lossy conductor, respectively.
Using the above expressions, Tatematsu derived the equation to update the electric field on the lossy
thin wire by taking into account the effect of the frequency-dependent conductor internal impedance as
follows [91]:

E n 1  11 AA E n  11A t   H n 1/ 2   1BA Equation 3.68

In the case of a solid conductor, A and B in Equation (3.68) are given by the following equations:

20Reprinted with permission from A. Tatematsu, A technique for representing lossy thin wires and coaxial cables for FDTD-
based surge simulations, IEEE Trans. EMC, vol. 60, no. 3, p. 707, Figure 4, 2018.

45
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

A t
A ', B  2 t
A' B ' Equation 3.69
s 2 s 2

  
1
A '  Asc   tw  t / 2   Asc  p 1 exp  j p2
 t
2 sc
Equation 3.70

B '  K 'nsc11   p 1 K 'nsc21 ( p)


Equation 3.71

K 'nsc11  K scn 11  12  Asc I n 3/2   tw  t / 2  I n 1/2  Equation 3.72

K scn 1  K snc11 
Asc
2 I n 1/2
 I n 3/2  Equation 3.73

K 'nsc21 ( p)  exp   j p2 t /  sc  K scn 21 ( p) 

Asc
2 
exp  j p2 t
2 sc I n 1/ 2

Asc
2 
exp  j p2 3 t
2 sc I n 1/ 2
 I n 3/ 2  Equation 3.74

K scn 2 ( p) 
Asc
2 
exp  j p2 t
2 sc I n 1/ 2
 K 'nsc21 ( p) Equation 3.75

Asc   tw atw2 1 

1
Equation 3.76

 sc  tw tw atw
2
1
Equation 3.77

jp denotes the pth positive zero of the Bessel function of the first order of the first kind. jp can be
calculated approximately using the following expression [94]:

j p    81  4(  3(8
1)(7   31)
 32(  1)(8315(8982   3779)
 64(  1)(6949  153855  2 1585743   6277237)
2 3

 )3 )5 105(8  )7
Equation 3.78

where β = (p+1/4)π and μ = 4.

In the case of a tubular conductor, A’ and B’ in Equation (3.69) are given by the following equations:

   2W 
1
A '   Atc   tw  t / 2  2 Atc  p 1 exp  2 p 2
 t
2 tc tw
Equation 3.79

B '  K 'tcn 11   p 1 K 'tcn 21 ( p)


Equation 3.80

K 'tcn 11  Ktcn11  12  Atc I n 3/2   tw  t / 2  I n 1/ 2  Equation 3.81

Ktcn1  Ktnc11 
Atc
2 I n 1/ 2
 I n 3/ 2  Equation 3.82

K 'tcn 21 ( p)  exp   2 p 2 t /  tc  K tnc21 ( p )  Atc exp  2 p 2  t

2 tc I n 1/ 2

 I
Equation 3.83
 Atc exp  2 p 2 3 t
2 tc
n 1/ 2
 I n 3/ 2 

Ktcn 2 ( p)  Atc exp  2 p 2  t

2 tc I n 1/ 2
 K 'tnc21 ( p) Equation 3.84

Atc   2 tw atw3Ttw 

1
Equation 3.85

Btw   2 /  tc Equation 3.86

 tc  tw twTtw2 Equation 3.87

The effect of the conductor internal impedance of the thin wire can be taken into account in FDTD-
based surge simulations by employing Equation (3.68) to update the electric fields along the thin wire
instead of simply setting them to zero, while the equivalent radius of the thin wire is simulated using
the technique described in Section 3.1.5.

46
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3.1.14 Representation of lossy coaxial cable

In [19], Tatematsu proposed a technique for representing a coaxial cable for FDTD-based surge
simulations employing coarse meshes. In this technique, not only thin wires connected to the internal
conductor of a coaxial cable but also the metal sheath of the coaxial cable is directly modeled in the
FDTD simulation by applying the thin-wire representation technique [44]. On the other hand, the
behavior of the internal electromagnetic field of the metal sheath is solved by applying transmission
line theory on the assumption of the transverse electromagnetic (TEM) mode.
Figure 3.19 shows a coaxial cable and a wire connected to its internal conductor, both of which are
placed in the direction of the z-axis. In Figure 3.19, (i, j, k) corresponds to the position (iΔs, jΔs, kΔs).
The metal sheath of the coaxial cable, the wire, and the end of the coaxial cable, where the thin wire is
connected to the internal conductor, are treated in the FDTD method using the following procedure.
The radii of the metal sheath of the coaxial cable and the thin wire are represented by the thin-wire
representation technique described in Section 3.1.5.
To take into account the interaction between the external electromagnetic field of the coaxial cable
simulated in the FDTD method and the internal electromagnetic field of the coaxial cable solved by
applying the transmission line model at the point connecting the wire and the coaxial cable, Tatematsu
derived expressions to update the magnetic fields at the end of the coaxial cable [19] on the basis of
Faraday’s law [95]. For example, the expression to update the magnetic field at the end of the coaxial
cable, Hy(i+1/2, j, k+1/2), is obtained as follows:
H yn  3/ 2 (i  1/ 2, j , k  1/ 2)  H yn 1/ 2 (i  1/ 2, j, k  1/ 2)
t
2  C
 V n 1  sEx n 1 (i  1/ 2, j , k )  sEz n 1 (i  1, j, k  1/ 2) Equation 3.88
s
sExn 1 (i  1/ 2, j, k  1)  sEz n 1 (i, j, k  1/ 2)

where Vc denotes the voltage between the internal conductor and the metal sheath at the end of the
coaxial cable. Expressions to update the other magnetic fields at the end of the coaxial cable can be
easily derived in the same way. The transient voltages and currents propagating inside the metal
sheath are obtained on the basis of transmission line theory, although, in this study, Tatematsu
assume that the coaxial cable is a lossless transmission line and that the shielding effectiveness of the
metal sheath is high [19]. In this technique, to update the voltages at the ends of the coaxial cable
where the internal conductor is connected to the FDTD region, the currents outside the coaxial cable

47
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

rw
Thin wire
Ex (i+1/2, j, k+2)

Hy (i+1/2, j, k+3/2)
Ez (i+1, j, k+3/2)
Ex (i+1/2, j, k+1)

Connecting point Hy (i+1/2, j, k+1/2)

of thin wire and Ez (i+1, j, k+1/2)
internal conductor
(i, j, k) E x (i+1/2, j, k)

Hy (i+1/2, j, k− 1/2)
Δs Ez (i+1, j, k− 1/2)
z Ex (i+1/2, j, k− 1)

y x Internal Metal sheath

conductor rs
ri

Coaxial cable

Figure 3.19 Electric and magnetic fields surrounding a thin wire and a coaxial cable.21

are necessary, and they can be calculated from the magnetic fields using Ampere’s law in the FDTD
method.
On the basis of the above technique, in [91], Tatematsu proposed a technique for simulating the effect
of the surface transfer impedance of the coaxial cable, i.e., the voltage induced inside the coaxial
cable by an external sheath current (e.g., [96]), where the metal sheath is handled by using the
proposed technique in the FDTD method, while solving surge phenomena inside the coaxial cable on
the basis of transmission line theory.
The surface transfer impedance Zt in a coaxial cable with a tubular shield is expressed by the following
equation in the frequency domain [93]:
 s Tc
Z t    ac 2  ac 3 Tc
1
sinh  s Tc  Equation 3.89
s

Tc  ac 3  ac 2 Equation 3.90

s  j s s Equation 3.91

where ac2, ac3, σs, and μs denote the inner radius, outer radius, conductivity, and permeability of the
metallic sheath, respectively.
Similar to the above technique, in this technique for representing a lossy coaxial cable in FDTD-based
surge simulations, the propagation of the voltage V and current I in the cable is solved using the
telegrapher’s equations based on 1D transmission line theory on the assumption of TEM mode
propagation. To take into account the effect of the surface transfer impedance (i.e., the voltage
induced in the coaxial cable by the current Is flowing through the metal sheath, whose return path is an
external conductor), the telegrapher’s equations taking into account the surface transfer impedance
are expressed as follows [96]:

21 Reprinted with permission from A. Tatematsu, A technique for representing coaxial cables for FDTD-based surge
simulations, IEEE Trans. EMC, vol. 57, p. 489, Figure 3, 2015.

48
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

V ( )
l
  j L I ( )  ( Z c  Z si ) I ( )  Zt I s ( ) Equation 3.92

I ( )
l
  jCV ( ) Equation 3.93

where L and C respectively denote the per-unit-length inductance and capacitance between the core
and sheath. Zc and Zsi denote the conductor internal impedances of the core and the metal sheath for
the internal current, respectively. l denotes a position on the coaxial cable. Note that the current
induced inside the coaxial cable by the potential of the sheath and the effect of the internal current
coupling back to the outside voltage and current are ignored in the above equations on the condition
that the cable has a good shielding, which is generally satisfied. Zc and Zsi are expressed by [92], [93]
 I0 ( c ac1 )
Zc  2 acc1 c I1 ( c ac1 ) Equation 3.94

 c  jc c Equation 3.95

T
Z si  2 acs2c sTc coth  sTc  Wsi Equation 3.96

Wsi  2 a 1
 ac 2  ac 3  s Equation 3.97
c2

where ac1, σc, and μc denote the radius, conductivity, and permeability of the core, respectively. I0(∙)
and I1(∙) are respectively the modified Bessel functions of the zeroth and first order of the first kind.
Note that Equation (3.96) is an approximate expression.
From the above equations, the following equation to update the current by taking into account the
effects of Zc, Zsi, and Zt can be obtained [97]:

I kn1/1/22  U  L
t

Wsi
2 I n 1/ 2
k 1/ 2 U 1
s V n
k 1  Vkn   U  Fkn1/ 2  G4 k 1/ 2  H 4 k 1/ 2  Equation 3.98

Fkn1/ 2  F '1n   p 1 F '2n ( p)


Equation 3.99

F '1n  F1n 1 
At
2 3I n 1/2
s k 1/2 1/2  I s k 1/2 
 2 I s nk 1/2 n  3/2
Equation 3.100

F1n  F1n 1 
At
2 I n 1/2
s k 1/2  I s nk 1/2
3/2
 Equation 3.101

F 'n2 ( p)  exp   Bs p 2 t  F2n 1 ( p)  At  1 exp   Bs p 2 t  3I s kn 1/1/ 22  2 I s nk 1/1/ 22  I s kn 1/3/22  Equation 3.102
p

F2n ( p)  exp   Bs p 2 t  F2n 1 ( p)  At  1  I s nk 1/1/ 22  I s nk 1/1/ 22 

p

Equation 3.103
 At  1 exp   Bs p 2 t  I s nk 1/1/ 22  I s kn 1/3/22 
p

G3 
Ac
2 1   
p 1
G3 ( p)  Equation 3.104

G4 k 1/ 2  G1kn 11/ 2 

Ac
2
I kn1/3/22   p 1 exp   j p2 t /  c G2 kn 11/ 2 ( p)  2c  I kn1/1/22  I kn1/3/22 
 A

Equation 3.105
 G3 ( p) I
Ac n 1/ 2

2 k 1/ 2 

H 3  3 Asi  1
2
  p 1 exp   Bs p 2 t 

 Equation 3.106

H 4 k 1/ 2  H1kn 11/ 2  Asi  I kn1/1/22  12 I kn1/3/22 

Equation 3.107
  p 1 exp   Bs p 2 t  H 2 nk 11/ 2 ( p )  Asi  2 I kn1/1/22  I kn1/3/22  


 
 
1
U  G3  H 3 
L Wsi
t 2
Equation 3.108

49
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

G1nk 1/2  G1nk 11/2 

Ac
2 I n 1/2
k 1/2  I kn1/2
3/2
 Equation 3.109

G2 nk 1/2 ( p)  exp   j p2 t /  c  G2 kn 11/2 ( p)  I n 1/2

 I kn1/2   2c exp   j p2 t /  c  Ikn11/2/2  Ikn1/23/2 
Ac 1/2 A
2 k 1/2 Equation 3.110

G3 ( p ) 
2 c
t  1
j 2p
 1
j 2p 
exp   j p2 t /  c   exp   j p2 t /  c  Equation 3.111

H1nk 1/2  H1nk 11/2 

Asi
2 I n 1/2
k 1/2  I kn1/2
3/2
 Equation 3.112

H 2 nk 1/2 ( p)  exp   Bs p 2 t  H 2 nk 11/2 ( p)  Asi  I kn1/2

1/2
 I kn1/2
1/2
  Asi exp  Bs p2 t  Ikn1/21/2  I kn13/2/ 2  Equation 3.113

At   s  ac 2  ac 3  Tc 
1
Equation 3.114

Ac   c ac21
1
Equation 3.115

 c  c c ac21 Equation 3.116

Asi  2 s ac 2Tc 

1
Equation 3.117

Bs   2 /  s Equation 3.118

 s   s sTc2 Equation 3.119

In the above, Δs is equivalent to that used in the FDTD method to solve electromagnetic transient
phenomena outside the coaxial cable. The current flowing through the outside metal sheath Is, which
has an external return path, is obtained using the rotational magnetic fields around the metal sheath
on the basis of Ampere’s law.
On the other hand, the expressions used to update the voltage inside the coaxial cable is obtained as
follows:

Vkn 1  Vkn  Cts  I kn1/2

1/2
 I kn1/2
1/2
 Equation 3.120

In case of a braided shield, it is well known that it is necessary to take into account the effect of the
shield structure such as woven bands of wire filaments, and there have been several techniques for
modeling the surface transfer impedance of a braided shield (e.g., [96]). For example, in the model
proposed by Vance for modeling the transfer impedance of a single braided shield [96], Zt in Equation
(3.89) is replaced with the following equation:
 sd
Zt  4  j M quation 3.121
 d 2 NW  s cos( ) sinh  s d 

where d, N, W, and α denote the diameter of the wire filament, the number of filaments in each carrier,
the number of carriers in the shield, and the weave angle of the shield, respectively. The first term on
the right hand side of Equation (3.121) represents the effect of the diffusion of the electromagnetic
field through the shield, while the second term represents the effect of the inductance due to the
magnetic field penetrating through shield holes, which is specified by the structure of the braided
shield. In this case, the expression to update the current inside the coaxial cable is given by:

I kn1/1/22  U  L
t

Wsi
2 I n 1/ 2
k 1/ 2 U 1
s V n
k 1  Vkn   U  Fkn1/ 2  Qkn1/ 2  G4 k 1/ 2  H 4 k 1/ 2  Equation 3.122

Qkn1/2  Mt  I s nk 1/2

1/2  I s k 1/2 
n 1/2
Equation 3.123

Here, it is assumed that, similar to a tubular shield, the electrostatic coupling through the shield is less
important also in case of a braided shield, and the surface transfer impedance is more dominant
compared with the surface transfer admittance, which is satisfied at low frequencies and for a highly
shielded cable [83]. However, for example, in case of SGEMP (system generated electromagnetic
pulse), where electric fields near a cable sheath can be highly intensified due to electron emission, it is
more important to model the surface transfer admittance for obtaining accurate results of induced
voltages inside a cable (e.g., [98]).

50
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The procedure to update the electric and magnetic fields in an analysis space and the voltage and
current in a coaxial cable by taking account of (i) the radius of the metal sheath using the thin-wire
representation technique, (ii) the effect of the conductor internal impedance of the metal sheath, (iii)
the effect of a thin wire connected to the core of the coaxial cable, and (iv) surge phenomena inside
the coaxial cable incorporating the effect of the surface transfer impedance is summarized as follows:
(i) Step 1: Update the radial electric fields adjacent to the thin wire and the metal sheath of the coaxial
cable at time (n+1)Δt using the thin-wire representation technique in Section 3.1.5.
(ii) Step 2: Update the electric fields on the thin wire and the metal sheath using Equation (3.68) to
simulate the effect of the conductor internal impedance.
(iii) Step 3: Update the other electric fields in the analysis space at time (n+1)Δt using the normal
expression to update the electric field.
(iv) Step 4: Update the voltage in the coaxial cable at time (n+1)Δt using Equation (3.120).
(v) Step 5: Update the rotational magnetic fields adjacent to the thin wire and the metal sheath of the
coaxial cable at time (n+3/2)Δt using the thin-wire representation technique in Section 3.1.5, except for
the magnetic fields at the end of the cable.
(vi) Step 6: Update the magnetic fields at the end of the coaxial cable at time (n+3/2)Δt using Equation
(3.88)
(vii) Step 7: Update the other magnetic fields in the analysis space at time (n+3/2)Δt using the normal
expression to update the magnetic field.
(viii) Step 8: Calculate the current flowing through the metal sheath of the coaxial cable, treated as a
thin wire, at (n+3/2)Δt on the basis of Ampere’s law.
(ix) Step 9: Update the current inside the coaxial cable at (n+3/2)Δt using Equation (3.98) or (3.122) to
simulate the effect of the surface transfer impedance.
(x) Step 10: Increment the time step and return to Step 1.

3.1.15 Advantages and disadvantages

Advantages of the FDTD method in comparison with other electromagnetic computation methods can
be summarized as follows:
(1) It is based on a simple procedure, and therefore its programming is relatively easy;
(2) It is capable of treating complex geometries and inhomogeneities;
(3) It is capable of incorporating nonlinear effects and components; and
(4) It can handle wideband quantities from one run with a time-to-frequency transforming tool.
Its disadvantages are:
(1) It is computationally expensive compared to other methods such as the MoM;
(2) It cannot deal with oblique boundaries that are not aligned with the Cartesian grid when the
standard orthogonal grid is employed, and needs a staircase approximation for oblique boundaries
(although a non-orthogonal grid that can avoid staircase approximation and will be shown in Section
4.7.1 is available); and
(3) It would require a complex procedure for incorporating dispersive materials/media.

3.1.16 Other techniques

In this section, some techniques which were developed for applying the FDTD method to solving the
telegrapher’s equations numerically are explained.

3.1.16.1 Using subgridding FDTD method to incorporate air gap breakdown

In FDTD simulation, in order to maintain a large time step on the transmission line, while using a small
time step on the simulation of air gap breakdown, the sub-gridding [99] of the transmission line is

51
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

developed by He et al. [100]. At the vicinities of node where air gap component is connected, the line
is partitioned into a fine grid, while the other part of the line is partitioned into a coarse grid. Based on
the wave-equation representation of transmission line, an interface method between coarse and fine
grids is proposed. As an example, the protection effectiveness of surge arrester with series gap is
studied by the proposed method.

3.1.16.1.1 Analysis of transmission line with air gap branch

With the presence of a branch on a transmission line, the Agrawal equations under the condition of
perfectly conducting ground are [4]:
V s  x, t  I  x, t 
 RI  x, t   L  E e L  x, t  Equation 3.124
x t

I  x, t  V s  x, t 
C  I avr .B Equation 3.125
x t
where I (x,t) is the current along the line, and V s(x,t) is the scattered voltage of the line.The total line
voltage Vtotal is given by:

Vtotal  x, t   V s  x, t   0 E eT  x, t  dz
h
Equation 3.126

EeT and EeL are the components of the incident electric field that are transverse to the line and parallel
to the line, respectively. Iavr.B is the spatial averaged current injected into the line, which is equal to
zero. Denoting the branch current flowing out of the line as IB gives:
IB
I avr .B   Equation 3.127
x
where Δx is the spatial step used in the discretization.
Suppose the branch is a lumped one, as shown in Figure 3.20 and the V-I relation of the branch is
known, a recursive formula can be obtained easily by discretizing Equation (3.125) using second order
central difference:

Figure 3.20 Illustration of transmission line with a lumped branch.22

Vkn 1  Vkn 
t 1 n  12
x

n 1 n 1
C I k  I k 12  I B 2  Equation 3.128

Given the V-I relation of the branch, Vn+1 can be obtained from Equation (3.128). One of the most
commonly used methods in the simulation of time-lag characteristics of short air gap breakdown is the
disruptive effect method, which models the breakdown as a cumulative process. In this section, this
method is adopted and the volt-time characteristic of a gap is determined from the following integration
equation [101]:

DE   V  t   K1 
tb K2
dt Equation 3.129
t0

where DE, K1 and K2 are constants for a particular air gap, t0 is the time when V(t0)=K1, and K1 can be

22 Reprinted with permission from J. L. He, S. C. Wang, R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol. 4, p. 46, Figure 1, 2012.

52
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

determined experimentally. K2 is always assumed to be 1-2 and tb is the time when breakdown
happens.
In the simulation, Equation (3.129) is evaluated numerically using the trapezoidal integration scheme.
Before the breakdown happens, the pre-discharge current is neglected, and the air gap is assumed to
be a 1-MΩ resistance. Once the breakdown happens, the air gap is turned into a perfect conductor
with zero resistance. The 5-cm rod-rod breakdown data [101] is employed to fit the parameters in
Equation (3.129). The simulation of a 5-cm air breakdown using Equation (3.129) with different time
steps are presented in Figure 3.21. The breakdown happens before the voltage reaches its peak
value. The time step has a notable influence on the breakdown time and the peak value of applied
voltage.

Figure 3.21 Profile of the applied voltage on a 5-cm rod-rod air gap. The wavefront time is 0.12μs.23

In the simulation, a symmetrical oscillation phenomenon is observed when the switching occurs.
Inspired by Marti et al. [102], Equation (3.128) is modified to the following formula at the time step of
switching:

Vkn 1  Vkn 
x

t 1 n  12 n 1
C I k  I k 12  I Bn 1  Equation 3.130

Using the formulas presented above, the numerical simulation of a transmission line with air gap
branch can be implemented. It can be seen from Figure 3.21 that a small time step is needed to
produce an accurate result. However, a small time step means a high computational expense in the
lightning electromagnetic field calculation. In the next section, a subgridding method is proposed to
solve this problem.

3.1.16.1.2 Subgridding technique

As discussed in the previous section, the necessity of a fine grid arises from the need for accurate
simulation of the fast transient of the component connected to the transmission line. The branch,
either a lumped component or a transmission line, might need a temporal or spatial step smaller than
that used on the main line. The uniform small step on all parts means higher computational cost. In
this section, a subgridding method is proposed, which enables the use of a small time step on the part
of the transmission line system requiring a fine grid, and large time step on other parts.
The transmission line is partitioned in such a manner that the branch is connected to one of the
voltage nodes. The sections adjacent to the branch are partitioned into a fine grid. The geometry of
the subgridding method is illustrated in Figure 3.22 (a).
Taking the partial derivative of Equation (3.124) with respect to x and the partial derivative of Equation
(3.125) with respect to t and combining the two equations to eliminate the current term gives:

23 Reprinted with permission from J. L. He, S. C. Wang, R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol. 4, p. 46, Figure 2, 2012.

53
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 3.22 The spatial partitioning of a transmission line with both fine and coarse grids, Nfg=3. Note
that the branch line is also horizontal.24

 2V V  2V    I
 RC  LC  EL   L  R  B Equation 3.131
x 2
t t 2
x  t  x
Discretizing Equation (3.131) by the second order central difference method at the kth voltage node
gives:

 2Vkn  Vkn  V n  E 
  LC  RC k  L 
t 
1
 RC LC  
2
2t x k ,  n ,  
Vkn       Equation 3.132
 2t t    Vkn  2Vkn  Vkn  L t  R  I B n
2 

  
x  2x
2
 
where α=1/Nfg,…, Nfg/ Nfg, and Nfg is the subgridding ratio. IB is the branch current flowing out of the
n n
node. Vk  and Vk  can be obtained through spatial interpolation:
Vkn  Vkn   Vkn  Vkn1 
Equation 3.133
Vkn  Vkn   Vkn1  Vkn 

When the branch is a lumped component, the V-I relation of the branch can be represented as:


I B  f B Vk  0
h
ET , k dz  Equation 3.134

then, Equations (3.134) and (3.132) can be combined to obtain Vkn+α. However, when the branch is a
transmission line instead of a lumped component, the procedure is not straightforward. In the following
section, a method of handling this scenario is proposed.

3.1.16.1.3 Subgridding technique in FDTD solution of a transmission line with different grids
Another problem associated with the FDTD simulation of transmission lines is the high computational
cost associated with handling frequency-dependent parameters, as the convolution should be carried
out on each discretized segment. The morespatial steps there are, the higher the simulation cost.
Despite the acceleration brought by recursive convolution, the FDTD method is still not suitable for
long line simulation. This problem can also be solved by the subgridding method. As illustrated in
Figure 3.22(b), the important part of the transmission line is simulated with a fine grid, while the other
part is simulated with a coarse grid. This reduces the number of spatial segments, and, therefore the
computation time. For the convenience of the deduction, the frequency-dependent parameters are
neglected in the interface section.
In order to illustrate the deduction,a more general case can be considered: a transmission line is
connected with a transmission line branch, as shown in Figure 3.22(b). The horizontal line is referred

24 Reprinted with permission from J. L. He, S. C. Wang, R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol. 4, p. 47, Figure 3, 2012.

54
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

to as the main line, and the vertical line as the branch line. If the section of main line on the right side
of the branch is cut down, the system becomes a single transmission line. Although Equation (3.128)
is still correct for this condition, it cannot be evaluated directly, because the relation between Vk and IB
at each fine time step is unknown. The FDTD representation of Equation (3.125) at the terminal of the
branch is

1  n  12 I Bn 1  I Bn  VBn,11  VBn,1
 B ,1
I  
 B C 0 Equation 3.135
x / 2  2  t
Combining Equation (3.135) and (3.128) gives:

VBn,11  Vkn 1  Vkn  2

t
x
 1 1


 2C1  CB  I kn12  I kn  2  I Bn,1 2
1 1

 Equation 3.136

With the help of Equation (3.136), the voltage recursion on the main line and the branch line can
proceed forward.
A subgriding formulation similar to Equation (3.132) can be obtained:

 2Vkn  Vkn  V n  
  LB CB  RB CB k 
t 
1
R C LC  
2
2t 
Vkn    B B  B B2    Equation 3.137
 2t t    VB ,1  VB ,1 EL
n n
 LB t  RB  I B 
    n

x  x k , n ,  2x
2
 
Combining Equations (3.132) and (3.137), the term which contains the branch current IB is cancelled
out, and Vkn+αcan be obtained.
In case of a multi-conductor transmission line, the formulas developed above can also be used by
simply replacing the capacitance and inductance with capacitance and inductance matrices,
respectively.

3.1.16.1.4 Simulation results

In practice, the installation of a surge arrester with a series gap is a promising method to prevent the
flashover caused by lightning induced overvoltage [103]. The purpose of the air gap is to prevent
aging of the surge arrester caused by the power frequency voltage. In this section, a single-conductor
overhead line excited by a lightning stroke is studied. One surge arrester with a 5-cm series gap is
installed to examine the protection effectiveness of the arrester against fast front overvoltage. The
grounding resistance of the pole grounding device is assumed to be 30 Ω. The geometry of the
problem is given in Figure 3.23. There is usually a grounding wire of about 10-m long between the
arrester and the grounding device, which was not considered in this simulation.

Figure 3.23 Example using surge arrester with 5-cm series gap.25

Since the 10-kV surge arrester is typically very short, it can be modeled as a lumped nonlinear
resistance. The Newton-Raphson method is employed to handle the nonlinearity of the surge arrester.
Parameters for the 5 cm gap are: DE=6.45, K1=40.30 and K2=2.0. Heidler’s formula is used to

25 Reprinted with permission from J. L. He, S. C. Wang, R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol. 4, p. 47, Figure 4, 2012.

55
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

represent the lightning current formula and a fast front lightning current is adopted in this section
(τ1=0.1 μs, τ2=25 μs and Ip=42 kA).
The overvoltage profiles at the point where the surge arrester is installed are observed. The residual
voltage on surge arresters with and without a series air gap is compared in Figure 3.24. Due to the
time-lag characteristic of the breakdown process, there exists a voltage spike before the discharge of
surge arrester. This spike might undermine the protection effectiveness of the surge arrester.

Figure 3.24 Profiles of the residual voltage on the arresters with and without air gap; time step is 0.03
μs.26
According to experience [100], the largest stable subgridding ratio is 3:1 (Nfg=3) for the method
proposed. In the simulation, the coarse time step is adopted as 0.03 μs and the fine time step is 0.01
μs. The corresponding spatial steps are 10 m and 10/3 m, respectively. Under this lightning surge, the
breakdown process lasts about 0.1 μs as shown in Figure 3.24. It takes approximately 3 time steps for
the coarse grid and 10 time steps for the fine grid. At the vicinities of the voltage node where the surge
arrester is installed, the transmission line is partitioned into a fine grid. Inside the fine grid section,
Equation (3.132) is used to evaluate the voltage value at the two adjacent voltage nodes. Equation
(3.130) is used to take the branch into account in the recursive solution.
The simulation results are presented in Figure 3.25. While the subgridding method is in good
agreement with the fine grid method, the result given by the uniform coarse grid presents a notable
difference when compared with them. As can be seen, for the highest voltage value the difference is
about 20%.

26 Reprinted with permission from J. L. He, S. C. Wang, R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol. 4, p. 48, Figure 5, 2012.

56
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 3.25 Comparison of the residual voltage profiles, three calculation modes are considered: coarse
grid mode, fine grid mode and subgridding mode.27

3.1.16.2 Numerical method for nonlinear element branch

Surge arresters can be simulated by a nonlinear resistor, and supposing the V-I characteristic of the
surge arrester can be divided into several segments, then every segment can be approximately
presented by
i  pv q Equation 3.138

where, i and ν are the current through the arrester and the voltage on the arrester, respectively, and
constants p and q can be obtained from the discrete points. For the case in Figure 3.22(a), the branch
current can be presented by

I 
B ,k i  fi Vk i  Vg  Equation 3.139

where, fi presents the V-I characteristics of i-th surge arrester, Vg is the voltage on grounding device of
the tower connected with the surge arresters, Rg is the groundingresistance of the tower grounding
device. {Vk}i presents the vector Vk of i-th element.
So, for the branch point of surge arrester in Figure 3.22(b), the relationship between the voltage and
current satisfies the following relationship:

  f V 
Vg
i i k ,i  Vg   Equation 3.140
Rg

3.2 Electric field integral equation (EFIE) based methods for lightning
studies
3.2.1 Introduction
3.2.1.1 Genesis of the electric field integral equation
The Electric Field Integral Equation (EFIE) relates the electromagnetic potentials or fields in space to
their sources, which are basically formed by charges and currents. These sources could be either
physical involving true charges and currents, or mathematically equivalent ones, which are employed
to represent different media.
For a better description, consider a conducting surface placed in a linear, homogeneous and isotropic
media occupying the whole space. Then the EFIE relating source to field anywhere can be written as
[104].

27 Reprinted with permission from J. L. He, S. C. Wang, R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol. 4, p. 48, Figure 6, 2012.

57
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

𝜌(𝒓′ ,𝑡 ′ ) 𝑹 ∇∙𝑱(𝒓′ ,𝑡 ′ ) 𝑹 𝜕𝑱(𝒓′ ,𝑡 ′ )

] 𝑑𝑠 ′
1 1
𝑬𝑡 (𝒓, 𝑡) = − ∫𝑆 [ − − Equation 3.141
4𝜋𝜖 |𝑹|3 𝑐|𝑹|2 𝑐 2 |𝑹| 𝜕𝑡 ′

Where the left hand side is the total electric field at the observation point at given global time t, primed
quantities on the right hand side correspond to the source quantities subjected to time retardation, ρ
and 𝐽⃗ represent the surface charge and current densities, 𝑅⃗⃗ is the distance vector between any source
point (with position vector 𝒓′) to the observation point (with position vector 𝒓). Note that the
unknowns appear only in the integrand and hence Equation (3.141) is a Fredholm integral equation of
first kind [104]. It can be seen from Equation (3.141) that the source terms (apart from permittivity and
permeability of the media) are scaled by the geometry terms, which form the kernel of the integral
equation. With the knowledge of impressed sources such as lumped voltage/current sources and
incident fields, along with the condition to be satisfied on the conducting surfaces (and dielectric
interfaces), the integral equation can in principle be solved for the unknown source quantities.
In practice, the solution of the integral equation is carried out using the MoM, which is a boundary
based methodology involving the discretisation of only the problem boundary. The unknown source
quantities are discretized into a piece-wise defined trial or basis functions. In general, the numerical
solution requires a relatively large amount of analytical preparation and implementation of
sophisticated numerical procedures. Further, these tend to be highly inefficient for inhomogeneous
media and not readily applicable to nonlinear media [105]. Unlike its counterpart, the Magnetic Field
Integral Equation (MFIE), the EFIE is not limited to closed surfaces and it can be applied to even open
and thin objects [106]. However, the MFIE is known to possess better low-frequency stability and is
hence generally suggested for modelling closed surfaces. In the majority of the lightning literature to
date, only EFIE is extensively employed and in view of the same, the MFIE will not be dealt here.
3.2.1.2 EFIE in lightning-related work
In lightning studies, EFIE is primarily applied to two areas: (i) the study of surge response of isolated
and interconnected towers, transmission lines and other ground-based structures; and (ii) modelling of
the return stroke evolution and for evaluation of return stroke generated electromagnetic fields.
The use of thin-wire EFIE for calculating the transient performance of the network of grounding
conductors has been reported by Grecv et al. [12]. The frequency-domain approach is employed for
the solution of EFIE, which is then transformed into the time domain using the Inverse Fast Fourier
Transform (IFFT).
The introduction of Numerical Electromagnetic Code (NEC-2), which solves EFIE in the frequency
domain, for lightning related work can be attributed to Ishii and Baba (e.g., [107]-[110]). In these
works, NEC-2 along with Fourier Transform techniques have been employed to quantify the surge
response of towers and transmission lines, as well as electromagnetic field produced by the lightning
channel during a strike to ground or tall object. Subsequently, the above approach has been employed
for a variety of lightning-related problems, including analysis of surge response of lightning protection
towers with guy wires [111], interconnected lightning protection towers [112], insulated mast design
[113] and currents in interconnected lightning protection systems [114].
Limited efforts towards a direct time domain field solution approach can be found for modelling of the
return stroke [56], [115], as well as for evaluating the lightning surge response of down-conductors
[116].
Incidentally, the modelling of the specific structures in lightning-related studies can be found with wind
generators and aircraft. These are generally carried out with specific/commercial codes using different
numerical methods including those based on integral equations. Owing to the complex nature of these
structures, a highly simplified representation of the lightning channel/lightning is rightly employed and
for brevity they will not be discussed here.

3.2.2 Numerical solution of the equation

For the normal range of lightning stroke current rise times, which span between fractions of a
microsecond to several tens of microseconds, the associated range of significant frequency
components are within 10 MHz. The corresponding wavelengths in free space can be reckoned to be
above 30 m. In view of this, the conducting core of the lightning channel, ground wires and phase
conductors of power lines, lattice elements of the tower and even the tower cross-section as a whole
can all be considered as electrically-thin objects (i.e. the cross-section is much less than the

58
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

wavelengths of significant frequency components). Incidentally, the above-listed entities are also
geometrically thin. Based on these observations, thin-wire modelling using the EFIE was adopted for
lightning-related work. Therefore, all further discussions will be limited to Thin-Wire (TW) EFIE.
Basically, the numerical solution of EFIE for lightning-related studies can be categorised into the
frequency-domain solution-based approach and the direct time-domain solution based approach. They
will be dealt in the next sections.

3.2.2.1 General aspects of the frequency-domain based approach

The frequency-domain based approach for aerial conductors has been extensively carried out using
NEC-2. It is a well-established public domain code, which solves for the currents and fields in the
frequency domain. A detailed description of the theoretical formulation of the numerical solution of
EFIE is deemed unnecessary here as it can be readily found [117]. Nevertheless, some of the salient
aspects of its usage will be briefly described here.
NEC-2 combines the integral equation for smooth surfaces with one specialised for wires and provides
accurate modelling for a wide range of frequencies. As mentioned in Section 3.2.2, most of the
lightning related works employed only the thin-wire part of the code. It handles non-radiating networks
and transmission lines, perfect and imperfect conductors and lumped loading with circuit elements (R,
L, and C). Ground can be modelled either as a perfect or as an imperfect conductor. While both
lumped voltage source and incident plane wave can form excitation, lightning studies are usually
concerned only with the former. The output of the code can be currents and charges in the discretised
geometry, near or far electric and magnetic fields and the power loss. It can handle multiple wires and
their junction. Also, an extended kernel can be employed when the conductor cross-section is non-
negligible compared to the separation distance between the conductors.
A conductor system to be analyzed by NEC-2 needs to be modelled as a composition of short
cylindrical segments. The coordinates of two end points of segments and the radius define a
cylindrical segment. In order to have adequate resolution, there is a restriction on the length of the
discretized segment Δl. The upper frequency limit arises from resolution of the minimum wavelength
and it requires Δl < 0.1 λmin, which is a commonly found constraint across different numerical schemes.
With regard to the lower frequency limit, it is due to the constant term and cosine term in the
interpolation function tending to become equal to each other. In order to restrict the resulting error, Δl
> 10-3 λmax is suggested [117]. Also, in order to be well within the purview of thin-wire formulation, the
ratio of the length of the segment to its radius should not be less than 8 to 10 [108], [117]. When the
model includes segments with this ratio less than about 2, the extended thin wire kernel option must
be utilized [117].
The NEC-2 field solution yields currents and fields in the frequency domain for the discretised
geometry. Using the same, time-domain currents can in principle be obtained for a range of excitation,
provided their frequency spectrum is within the range of frequencies considered for the field solution.
The following aspects hold true for any approach employing frequency-domain field solution for the
transient analysis.
For the conversion from one domain to the other, the Fourier Transform (FT) and Inverse Fourier
Transform (IFT) need to be employed. A flowchart for the solution procedure is provided in Figure
3.26, which is adopted from [108]. Despite tedious efforts towards frequency-domain resolution, the
time-domain inversion could pose problems [118]. Based on the procedure adopted in related areas,
some measures can be adopted. The source waveform needs to be modelled in a larger window with
an initial and end zero padding and the image of the source is to be appended after it reaches zero.
These exercises will be very useful in eliminating the offset and low frequency problems, which are
encountered in many cases.

59
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 3.26 Flow chart for transient analysis using NEC-2 and Fast Fourier Transform (FFT).28

The improved version of the NEC called NEC-4 can handle conductors buried in soil (homogeneous),
in addition to several other improvements [119].
As mentioned previously, the NEC-2 based approach has been extensively employed for various
kinds of lightning problems.
In the next section, the time-domain thin-wire formulations will be discussed.

3.2.2.2 General aspects of time domain based approach

The electric field in the time domain can be expressed in terms of retarded electric scalar potential and
magnetic vector potential by the well-known relation:

   
E r , t   r , t 
t
 
A r, t Equation 3.142


where, r ' is the location of measurement point from the origin and t denotes the global time. As
mentioned earlier, the problem involves electrically thin objects and hence thin-wire formulation can be
 
sought. For a filamentary current I( r ', t') flowing through an arbitrary path C( r ') shown in Figure 3.27,
the resulting electric field can be written as [11]:

Figure 3.27 Arbitrary path in wire.

28Reprinted with permission from M. Ishii, and Y. Baba, Advanced computational methods in lightning performance – the
numerical electromagnetic code (NEC-2), Proc. IEEE PES Winter Meeting, Singapore, Figure 1. 2000.

60
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method


 q r ', t '  
 ds ' 0 s' 
 
E r, t  
4 0 C
1
 r ' 
 
 R  4 C  r '  R t '

I r ', t ' ds ' 
 
Equation 3.143

 
 q r ', t '   
  s '  I r ', t '  ds '
  0  c 2 
4 C  r '   R  R t '   
   

where, r' and t'= (t - R / c) denotes the location of source point and time respectively and the time is
 
c is velocity of light in vacuum, q (r ' , t ' ) and
 
reckoned from the measurement point, R = | R | = | r - r '|,

I (r ' , t ' ) are charge per unit length and current respectively. The measurement is carried out on the
filament (wire) surface, while, the source is assumed to be concentrated along the filament axis. The
first term on RHS can be expanded as:

    1 q r ', t '  q r ', t '   1 

 q r ', t '

 R  R
     R 
 
Equation 3.144

1 q r ', t '     R   q r ', t ' R
 
R  t '
 
cR  
 R 3
  

The linear charge density q can be obtained from the current using the continuity equation:
 
s '
 
I r ', t '   q r ', t '
t '
  Equation 3.145

For calculations along the filamentary path (wire), it will be convenient to express the spatial
 
dependency of current and charge in terms of parametric variable s, as: -I( r ', t')  I(s',t') and q( r ', t') 
q(s',t') [11]. Under such conditions, the electric field due to the filamentary current can be expressed
as:

0 s'  R  
 
E r, t  
4 C r'

  R t '
I  s ', t '  c 2
R s ' R
R
I  s ', t '   c 2 3 q  s ', t '   ds '

Equation 3.146

The electric field along the path of the filamentary current (wire) needs to satisfy:


s  E  E A  IZ s  Equation 3.147

where, the left-hand side involves the component of incident (EA) and reaction (E) field along the
filamentary path and right-hand side indicates the impedance (ZS) drop per unit length along the same
path. This impedance drop contains the voltage drop across the resistive and inductive loading.
In the literature, the numerical solution of the time-domain equation has been carried out with different
interpolation and test functions. It may be worth noting here that as compared to the frequency-domain
based approach for the lightning related transient analysis, the literature on the use of the time-domain
approach is rather limited.
In the numerical solution of Equation (3.147) using the MoM the following steps are involved: (i) the
structure is divided into a number of segments NS, which is dictated by the highest frequency content
of the excitation or the resulting currents (ii) the unknown segment current is expanded on each
segment using suitable interpolation function in two dimensions i.e., in space and time (ii) the
boundary condition given by Equation (3.147) is enforced in either a pointwise manner at the center of
every segment or in the weighted average sense over the whole segment. As the surface
currents/charges on the segments are expressed in terms of line source along the axis of the
segments, the computational efforts are greatly simplified. However, in order to avoid any singularity,
the evaluation needs to be restricted to the surface or in other words, the source and measurement
are radially displaced by the wire radius. The implementation of the boundary conditions involves
assessment of the interaction of all the segments with one another leading to a system of Ns linear
algebraic equations, which can be solved for the segment currents or the amplitudes of the basis
functions.

61
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The interpolation function usually considered in the general thin-wire formulation are spatial pulse
function, quadratic function and the RWG basis function. There is also an isolated attempt to employ
the spatial interpolation function provided in NEC-2 for the frequency-domain approach, in time
domain studies [120].
With regard to the testing function in MoM, the sub-sectional point collocation has been extensively
employed. While the Galerkin’s testing function can provide a better accuracy, it involves lot more
calculation as integration over whole of the segment under consideration is necessary.

3.2.2.3 Matrix formulation, solution and long-term stability issues

Generally, the filling up of the matrix element demands considerable time and so is the accounting of
the contribution from other segments required for the time marching. Whenever problem involves long
straight elements (or symmetric or repetitive geometry), it would be possible to prepare a lookup table
with representative segments and use it efficiently. In the time-domain analysis, the resulting left-hand
side matrix is relatively sparse for typical geometries encountered in lightning studies and requires
inversion only one time provided there are no nonlinear elements in the system.
The time-domain solution possesses an inherent late time instability problem, which may become
evident when runtimes are extended to long tails. This instability depends on the spatial discretisation
of wires, which is dictated by the required spatial resolution. As the ratio of length to radius decreases,
onset of instability tends to be quicker over time. Further, when two closely spaced conductors are
involved, it would be better to sample over the circumference for the evaluation of the mutual
interaction between them rather than one-point sampling. This helps in improving the stability.

3.2.2.4 Inclusion of circuit elements and nonlinear devices

Inclusion of linear circuit elements in the formulation in principle requires a modification of the RHS
term of Equation (3.147). The resistive and inductive loading has been employed to for validating with
the NEC-2 based results, as well as, to realise the reduced velocity of the propagation for the current
along the channel [118]. When the Miller et al’s [11], [121] quadratic space-time interpolation function
is employed along with the point collocation, only the current of the segment under consideration gets
into the calculation.
The resistive loading of the segment is scaled to resistance per unit length and employed in the RHS
of Equation (3.147). For modelling the inductive loading of the segment, the quadratic interpolation
function for the current is differentiated to get the corresponding coefficient for the present and history
currents (note that in the above said approach involving quadratic time interpolation, two history terms
will contribute to the present time step).
The nonlinearity in the channel core has been successfully modelled in [115]. Handling of nonlinearity
with inserted components requires more attention and needs to be established.

3.2.2.5 Modelling of the proximity effect

Proximity of another conductor (i.e. wire) causes the current to be non-uniformly distributed across the
cross section of the wire and hence the coupling between the closed spaced conductors is affected. It
would then be not appropriate to employ the common approximation that the source is distributed
along the axis of the conductor in the form of a line source. Fortunately, for the frequencies of interest
and for geometries such as parallel running down conductors or cables, both the cross section of the
conductors and their separation distance can be considered to be electrically small. A general method
has been proposed in [116], [122] which has been claimed to work for set of arbitrary shaped
conductors running parallel to each other without any restriction on their current or charge.

3.2.3 Modelling of the soil

For modelling of the soil, which is usually considered as a semi-infinite (linear) conducting media
(without any frequency-dependent absorption/dispersion), the kernel of the integral equation becomes
more complicated. In the case of a frequency-domain approach, Sommerfeld’s integral needs to be
employed [12].

62
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

On the other hand, direct time-domain simulation complicates the matter further by imposing severe
computational requirements. A detailed discussion on modelling of the aerial conductor over lossy
ground is given by Pantoja et al. [123] and others. In any case, in lightning research efforts towards
modelling of lossy soil directly in time domain for the thin-wire formulation is rather limited and hence
further work is necessary in this direction.

3.2.4 Other issues

In this sub-section, some of the related issues which require further attention will be briefly discussed.

3.2.4.1 Modelling of corona

Modelling of corona in principle is feasible, with certain modelling approximation. For corona model,
which is radially symmetric about the segment, the charge decay into the corona can be modelled
either as mild dissipating media or as the slowly expanding charge wave. In the both the cases, due to
the radial symmetry of the corona envelope, it would not lead to any dynamic (i.e. magnetic and
radiation) fields.

3.2.4.2 Modelling of insulated cables

As such, the presence of insulation over the conductor (either insulating jacket over the
armour/metallic sheath or the insulation over the single core cables without armours) and further the
multi-conductor cables lead to complicated propagation characteristics and field distributions, which
cannot be easily implemented in the thin-wire or any boundary based methodologies. Some quasi-
static modelling approach can be attempted; however, it is rather difficult to find any serious effort
made in this direction.

3.2.4.3 Modelling of the wire junctions

Modelling of the junction requires overlapping spatial interpolation function like what has been
employed in NEC-2. In other words, the current in the segment forming the junction would be a
combination of the current interpolated from all the connected segments. Some effort seems to be
underway in this direction for even the direct time-domain formulation.

3.2.4.4 Modelling of secondary arcs

This can be attempted in an efficient manner by evaluating the possible location and trajectories of the
flashover arc. Then the geometric model is discretized and a high resistive loading is provided for
these segments such that till the instant of flashover, which may be dynamically determined through
the results of previous time steps, the resistance loading employed can be altered as per the spark/arc
law that is intended to be employed for the modelling.

3.2.5 Advantages and disadvantages

The use of EFIE for the lightning related studies has been quite successful for both modelling of
lightning return stroke, as well as, for the evaluation of surge response of lightning protection systems.
The EFIE generally solved using the MoM and it involves only the discretisation of the boundary,
whichis highly advantageous.
Both frequency domain and time-domain based approaches could be found in the literature, with their
own advantages and disadvantages. The frequency-domian based approach can be readily adopted
using well-known public domain code NEC-2 with forward and inverse discrete Fourier transforms.
However, any nonlinearity is rather difficult to incorporate.
The time-domain based approach requires elaborate code development and validation. Nevertheless,
it would provide flexibility in incorporating corona and nonlinear devices. One of the other major
drawbacks of the time-domain approach could be the long-term instability, which is inherent to such

63
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

numerical methods. However, usually this problem becomes evident after the critical time period of
interest.
In summary the integral equation based approach can be very useful in modelling both the lightning
return stroke, as well as lightning surge response of protection system.

3.3 Partial element equivalent circuit (PEEC) method

The partial element equivalent circuit (PEEC) method is introduced by A. Ruehli in 1974 (Ruehli 1974).
The PEEC method is also an integral equation based technique. It is based on the mixed potential
integral equation (MPIE) from which an equivalent circuit is extracted. Despite the PEEC method
provides full-wave solution, it was not primary developed for calculation of electromagnetic fields. The
PEEC method was strongly developed in the 1990s when retardation, excited fields, and treatment of
dielectric material are investigated and included [124], [125]. Unlike differential equation based
methods, the PEEC method gives fewer unknowns than differential equation based method since it
does not require discretization of the whole domain space. Though the resulting matrices are dense,
recent fast solvers have greatly improved the solution time for PEEC simulations [126]. During the last
twenty years, an interest, research efforts, and development of the PEEC method have increased
significantly in many areas such as electromagnetic (EM) radiation from printed circuit boards [127],
transmission line modeling [128], scattering problems [125], power electronic circuits [129], antenna
analysis [130], lightning effects in electronic devices [131] and in power systems [24], electromagnetic
coupling effect on a measuring system [132], ground potential rise due to lightning strike in power
systems [133].
The procedures of the PEEC simulation are expressed in Figure 3.28. The first step for the PEEC
models starts from discretizing geometry structure into small cells which are composed of current cells
and charge or potential cells. The current and potential cells are interleaved with each other as shown
in Figures 3.29 and 3.30. The rectangular pulse is employed both charge and current basis functions.
Then Galerkin method is applied to enforce the MPIE, which is written in the time frequency domain
for the sake of simplified explanation as given in Equations (3.148) and (3.149):
J
E i  r,    j0  g  r , r'  J  r' ,   dV '   r ,   Equation 3.148
 e  j 0 ( r  1) V'

1
  r' ,     g  r , r'    r' ,   dV ' Equation 3.149
4 0 V' v

64
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Starting

Segmentation of structure
and circuit extraction

Formulation of matrix
equations

Solving equations

Yes
Post process
Post process calculation
requirement

No

End

Figure 3.28 Procedures in the simulation of PEEC models [134].29

Cell length ln

Potential cell and

potential node
Current cell and
current node J(r)

Figure 3.29 Discretization of structure in the PEEC models.30

where Ei is an incident electric field, J is an current density, g is Green’s function associated with the
boundary conditions,  is a scalar potential, σe is conductivity of a conductor, and εr is relative
permittivity of a conductor.
The MPIE in the time domain can be directly derived from Equations (3.148) and (3.149) and the fine
detail can be found in [135].

29 Reprinted with permission from P. Yutthagowith, Development of the PEEC method and its application to a lightning surge
study, Ph.D. thesis, p.3-3, Figure 3.1, Doshisha University, Kyoto, Japan, 2010.
30 Reprinted with permission from P. Yutthagowith, Development of the PEEC method and its application to a lightning surge
study, Ph.D. thesis, p.3-7, Figure 3.2, Doshisha University, Kyoto, Japan, 2010.

65
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Inbnc(r)

Current cell

qmbmP(r)

Potential cell

Figure 3.30 Pulse basis functions (bn and bm) on current cells and potential cells.31

The term on the left hand side of Equation (3.148) can be represented as a part of a voltage source,
Vij which is affected from an external field and/or an excited source. The first term on the right hand
side is represented as a resistor Rij, the second term on the right hand side is represented as an
inductor Lij and a part of Vij affected by magnetic coupling from other elements, the third term on the
right hand side is represented as a capacitor Cmij affecting from dielectric material, and the last term on
the right hand side in Equation (3.149) is represented a voltage across the current cell and can
interpret as a capacitor Cij and dependent current sources Ii and Ij affected by electric coupling from
self and other elements through the continuity equation. The equivalent circuit of the PEEC model
from the interpretation is illustrated in Figure 3.31.

Cmij

i Rij Lij Vij j

-+

Ii Ci Cj Ij

Figure 3.31 Equivalent circuit model of a PEEC cell.32

From 3D geometries of a considered structure, the equivalent circuit can be extracted. Applying
Kirchhoff's voltage law to all current cells and applying the continuity equation to all potential cells, the
formulation of the modified nodal analysis (MNA) is composed. The resulting of the MNA formulation
in a matrix form can be solved by an appropriate network simulator in both time and frequency
domains.

3.3.1 Formulation of partial element equivalent circuit in the frequency domain

From interpretation of the mixed potential integral equation, Kirchhoff's voltage law applied to current
cells, and the continuity equation or the charge conservation equation is applied via Kirchhoff's current
law to potential cells. Whole system equations from all potential and current cells in the frequency

31 Reprinted with permission from P. Yutthagowith, Development of the PEEC method and its application to a lightning surge
study, Ph.D. thesis, p.3-7, Figure 3.3, Doshisha University, Kyoto, Japan, 2010.
32 Reprinted with permission from P. Yutthagowith, Development of the PEEC method and its application to a lightning surge
study, Ph.D. thesis, p.2-26, Figure 2.5, Doshisha University, Kyoto, Japan, 2010.

66
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

domain can be written in a matrix form corresponding to a MNA formulation as shown in Equation
(3.150).

 j P 1  Ya AT     I S 
      Equation 3.150
 A R  j L   I  U S 
where A is an incident matrix which expresses the cell connectivity, R is a matrix of series resistances
of current cells, L is a matrix of partial inductances of current cells including the retardation effect, P is
a matrix of partial potential coefficients of potential cells including the retardation effect,  is a vector
of potentials on potential cells, I is a vector of currents along current cells, US is a vector of voltage
sources, IS is a vector of external current sources, and Ya is an additional admittance matrix of linear
and nonlinear elements.
From Equation (3.150), the potentials on the potential cells and currents along the current cells can be
solved at each interested frequency. The time responses of the potential and current can be
calculated by inverse Fourier or Laplace transform.

3.3.2 Formulation of partial element equivalent circuit in the time domain

For full-wave time-domain PEEC models, accounting for the individual retardation between the current
cell and the potential cell, potentials and currents at many time steps to be stored, and it requires
heavy storage and computational time resources. For the time-domain full-wave case, the PEEC
models are formulated in the form of neutral delay differential equations (NDDEs) [136]. The full wave
formulation requires the separation of the self (instantaneous) and mutual (retarded) couplings in the
previous matrix formulations.
An effective way to increase the efficiency of the time domain PEEC method in terms of execution
time is to neglect time retardation in the PEEC method, which is called the quasi-static PEEC method.
The method is still effective in cases when the structure has to be electrically small compared with the
minimum wave length of an applied source. The formulation of the quasi-static PEEC method in the
time domain can be derived by using Equation (3.150) and is given in Equation (3.151).

 P 1 dtd  Ya AT     I S 
   
R  L dtd   I  U S 
Equation 3.151
 A
where L is a matrix of partial inductances of current cells neglecting the retardation effect, and P is a
matrix of partial potential coefficients of potential cells neglecting the retardation effect.
The matrices, P and L, express electromagnetic coupling among the cells, and are dense. The
solution of Equation (3.151) based on quasi-static assumption can be obtained by employing an
appropriate integration scheme. For simplicity, the backward Euler scheme is applied to Equation
(3.151) by discretizing in time. Equation (3.152) is obtained, which is given below.

 1t P 1  Ya AT   n   I Sn 1  P 1 1t  n 1 
      Equation 3.152
 A R  1t L   I n   U Sn 1  L 1t I n 1 
It has been clear that the electromagnetic retardation can be neglected in most cases of lighting surge
analyses in the transmission systems [137]. The method is greatly reduced the execution time in a
numerical simulation.

3.3.3 Advantages and disadvantages

The main difference of the PEEC from the MoM is the possibility to extract equivalent circuits from the
integral equations. Unlike differential equation-based methods, the PEEC method provides fewer
unknowns, since it does not require discretization of the whole space of interests. Though the resultant
matrices are dense, recent fast solvers have greatly improved the solution time for PEEC simulations
[126]. Like the other IE techniques, spurious resonances may occur [130], [138], most likely resulting
from poor geometric meshing. The formulations of the PEEC methods can be carried out either in the
frequency or in the time domain.

67
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

In the case of frequency-domain formulation, the time response is obtained by using the inverse
Fourier or Laplace transform. In the case of the time-domain formulation, nonlinear elements are
handled easily. However, the solution is quite sensitive, i.e. it may result in an instability problem.
Potentials and currents are required to be memorizes every time step. Therefore, computation time
efficiency is not much different from that in the frequency-domain formulation.
The first advantage of this method is that it can incorporate electrical components based on a circuit
theory, such as resistors, inductors, capacitors, transmission lines, cables, transformers, switches, and
so on. The second one is that in the frequency-domain formulation the composition matrix in this
method depends on the configuration of a considered system and a medium, and does not depend on
sources. The third one is the main advantage of the PEEC method. A potential on a conductor is
calculated directly from a node potential. Therefore, post processing for calculating scalar potentials
and currents is not required.

3.4 Hybrid electromagnetic model (HEM)

3.4.1 Introduction
The hybrid electromagnetic model (HEM) is an electromagnetic computational method developed for
numerical solution of lightning problems [21], [139], [140]. Its acronym HEM reflects the hybrid
electromagnetic-circuit approach of the model, since its formulation is based on EM theory to compute
the coupling among system elements from a numerical implementation of basic EM equations (including
propagation effects) and its results are expressed in terms of circuital quantities such as voltages and
currents.
The HEM model is a frequency domain model that adopts the inverse Fourier transform to provide
results in the time domain. It has been efficiently applied to several lightning-related studies involving
return-stroke current distribution, lightning electromagnetic fields, transient grounding behavior,
transmission line response to direct lightning strikes, lightning-induced voltages, transient behavior of
elevated structures, and so on.

3.4.2 Brief history of the model evolution

The history of the HEM model starts in ninties when Prof.Visacro published the basic ideas of the model
and implemented its first simplified version [139], [141], [142]. At that time, model application was only
dedicated to evaluations of the transient behavior of grounding electrodes subjected to lightning
currents.
Over the years, model formulation was improved as well as its solution algorithm, contributing to extend
the possibilities of HEM model application. From 1996 to 1998, model formulation was extended to allow
applications involving aerial conductors, such as those related to lightning direct strikes on high-voltage
transmission lines [143], [144]. From 2002 to 2007, additional implementations allowed the model to
calculate lightning-induced voltages [145]-[148]. In 2005, a paper dedicated to explaining model
formulations and the general application of the HEM was published in the IEEE Transactions on Power
Delivery [21]. In 2006, an extension of Norton's approximation [149] to solve Sommerfeld’s integrals was
developed and implemented in the HEM model allowing it to represent lossy ground on lightning-induced
voltage calculations [150]. In 2012, the effect of frequency-dependent soil parameters [151], were
included in the HEM model by means of Visacro-Alipio expressions [152]. Calculations of overvoltages
across transmission line insulators and lightning-induced voltages on single wire considering the
frequency dependence of soil parameters were first published by Visacro et al [153] and Silveira et al.
[154]. More elaborated calculations of such effects using the HEM model are presented by Visacro and
Silveira [155], [156], Silveira et al. [157], Silveira and Visacro [158].
Until 2019, more than 40 papers related to the HEM application for calculations related to lightning
problems were published in peer-reviewed journals and renowned international conferences.

3.4.3 Fundamental theory

In order to detail the fundamental theory of the HEM model, Figure 3.32 depicts a particular system
subjected to lightning currents, composed by return-stroke channel, line conductors of a transmission
line and grounding.

68
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 3.32 Particular system subjected to lightning currents.33

The HEM model is based on the discretization of all elements that compose the system under analysis
into several fillamentary segments, and the association of two different types of electromagnetic field
sources to each segment. Such electromagnetic field sources are described in terms of current
sources named (i) transversal current IT and (ii) longitudinal current IL, whose effects are considered
separately. Figure 3.33 illustrates such current sources.

Each segment: two different current sources

(field sources)

Transversal current IT

IT

... +
Longitudinal current IL

IL

Figure 3.33 Current sources IT (transversal) and IL (longitudinal) associated with each segment that
composes the evaluated system.34

The transversal current IT is a leakage current that crosses the element surface and is spread out to
the surrounding medium. It is related to a divergent electric field that establishes a potential rise on the
segment itself and on all other segments in relation to remote earth. Taking a pair of elements as
reference, the transversal current IT allows representing capacitive and conductive coupling derived
from the average potential established along one of the elements by the transversal current that flows
from the other one.
The longitudinal current IL that flows along the element is related to a non-conservative electric field
that establishes a voltage drop along the segment itself and contributes to the voltage drop along all
the other segments. Taking a pair of elements as reference, the longitudinal current IL allows
representing resistive and inductive coupling derived from the voltage drop established along one of
the elements.
The electromagnetic couplings including propagation effects are calculated by expressions derived
from the electric scalar potential and the magnetic vector potential as described in the next section.

33 Reprinted with permission from S. Visacro, and A. Soares Jr., HEM: A model for simulation of lightning-related engineering
problems, IEEE Trans. PWRD, vol. 20, no.2, p. 1206, Figure 1, 2005.
34 Reprinted with permission from S. Visacro, and A. Soares Jr., HEM: A model for simulation of lightning-related engineering
problems, IEEE Trans. PWRD, vol. 20, no.2, p. 1206, Figure 1, 2005.

69
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Considering such electromagnetic coupling for each pair of elements and to the element itself, two
independent matrices representing two linear systems are constituted, each one related to one
electromagnetic field source (divergent and non-conservative electric fields; transversal and
longitudinal current). Such coupling effects are represented by means of impedances (transversal
impedance and longitudinal impedance) taking into account propagation effects.
Since a single element simultaneoulsy causes potential rise (due to the transversal current IT) and
voltage drop (due to the longitudinal current IL), the developed coupling expressions in the form of
transversal and longitudinal impedance matrices may be accomplished in only one linear system,
considering both voltage drop and average potential of each segment described in terms of the
potentials at its extremity nodes. As a result, the system is represented by a matrix equation (Ax = b)
relating injected currents and the average potential developed in all system elements. The solution of
such linear system provides results in terms of currents and voltages.

3.4.4 Electromagnetic coupling expressions

(a) Transversal electromagnetic coupling
The transversal electromagnetic coupling between a pair of elements is described in terms of a
transversal impedance ZTij, one of them considered as the ‘current emitter or source element’ (j) and the
other one as the ‘receiving-effect element’ (i).
Such transversal impedance ZTij is based on the expression of the electric scalar potential at a point P
due to the linear transversal current density of the current emitter element ‘j’, including propagation
effects:

1 ITj e  r
 
4     j   L L j r
dl j Equation 3.153
j

The average potential established in element ‘i' due to ITj is calculated as follows:

1 e  r

ITj
Vij   dl i  Vij 
Li Li 4  [  ( )  j   ( ) ] L j L i  
Li Lj
r
dl j dli Equation 3.154

σ, ε, μ: medium conductivity, electric permittivity, magnetic permeability

r: distance between i and j
k: plane-wave propagation constant
L: element length
The transversal impedance ZTij is defined by the relation between the average potential developed at
any element ‘i’ due to the linear transversal current density (ITj/Lj, where LJ is the length of element j) of
element ‘j’, as indicated in Equation (3.155).

Vij e  r
Z Tij   ZT  1
ITj ij
4  [  ( )  j   ( ) ] L j L i  
Li Lj
r
dl j dli Equation 3.155

(b) Longitudinal electromagnetic coupling

The longitudinal electromagnetic coupling between a pair of elements is described in terms of a
longitudinal impedance ZLij, one of them considered as the ‘current emitter or source element’ (j) and
the other one as the ‘receiving-effect element’ (i).
Such longitudinal impedance ZTij is based on the expression of the magnetic vector potential at a point
P due to the average longitudinal current (ILj) of the current emitter element ‘j’, including propagation
effects:
  e  r 
A
4 
Lj
ILj
r
d lj Equation 3.156

The resulting voltage drop along the receiving segment ‘i' is calculated as follows:

70
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

    ILj e  r  
Vij   j   A  dl
Li
i  Vij   j
4  
Li Lj
r
d l j  d li Equation 3.157

The longitudinal impedance ZLij defined by the relation between the voltage drop developed at any
element ‘i’ due to the average longitudinal current (ILj), is indicated in Equation (3.158).

Vij e r  
ZLij   ZL   j  
ILj ij
4  
Li Lj
r
dl j  d l i Equation 3.158

3.4.5 Integrating the physical system

Based on Equations (3.155) and (3.158) (ZTij and ZLij), two sets of linear equations are described in
terms of the following matrix systems:
V  Z T IT Equation 3.159

ΔV  ZL IL Equation 3.160

Since they refer to the same physical system and a single element simultaneously causes potential
rise (due to the transversal current IT) and voltage drop (due to the longitudinal current IL), the
developed coupling expressions may be accomplished in only one linear system.
The electric potential of each node (VN) is expressed as function of average potential (Vi) and voltage
drop (ΔVi) for the corresponding element as follows:
1. Voltage drop along any segment equals the difference of potentials at the nodes positioned at
element extremities
2. Average potential of any segment equals the arithmetic mean of the potential at its extremity
nodes
Figure 3.34 illustrates the describedmathematical approach.

Figure 3.34 Mathematical approach to accomplish coupling expressions.35

Also, considering the principle of current continuity and that IT is derived from the nodes at the
segment extremities as illustrated in Figure 3.35, currents IT and IL are expressed as a function of
potentials at system nodes VN (for more details see [21], [139], [140]. Figure 3.36 details the system
final equation.

Figure 3.35 Distribution of IT and IL in a single segment.36

35 Reprinted with permission from S. Visacro, and A. Soares Jr., HEM: A model for simulation of lightning-related engineering
problems, IEEE Trans. PWRD, vol. 20, no.2, p. 1206, Figure 1, 2005.
36 Reprinted with permission from S. Visacro, and A. Soares Jr., HEM: A model for simulation of lightning-related engineering
problems, IEEE Trans. PWRD, vol. 20, no.2, p. 1206, Figure 1, 2005.

71
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

“information” external injected current

of ZT and ZL
potential at the nodes

Figure 3.36 Equation corresponding to the n nodes of system organized as a matrix equation.

In summary, the HEM model is based on three steps:

1. Establishment of electromagnetic coupling among all the elements of the partitioned system based
on coupling expressions ZT and ZL;
2. Coupling expressions comprised into an integral system representation based on the current
continuity principle applied to the elements, following a distributed circuit approach;
3. Representation of the resulting system by a matrix equation involving injected currents and average
potential developed in all elements of the system.

3.4.6 Computational aspects

Figure 3.37 shows a flowchart illustration of the main operations of the HEM model.

Definitions:
Frequency range, frequency step
Electrical properties of the surrounding media
System configuration

Electromagnetic coupling among segments, including propagation effects

(Scalar Electric Potential and Vector Magnetic Potential expressions)
Current (Field) sources: IT and IL (each segment)

Composition of two linear systems

Matrices ZT and ZL

Distributed circuit approach:

Integrate coupling expressions into an integral system representation
(Current continuity principle + Currents IT and IL expressed as function
of potentials at system nodes VN)

: “Information” of ZT and ZL
: Potential at the nodes
: External injected current

Application of Inverse Fourier Transform:

Determination of all physical quantities in time domain

Figure 3.37 Flowchart of HEM model operation.

3.4.7 Advantages and disadvantages

(1) Its hybrid approach allows accurate and efficient calculations:
- Field approach allow accuracy on quantifying electromagnetic coupling relations;
- Distributed circuit approach results on efficiency on performing the connecting relations and
composing the final matrix expression for the problem under evaluation;
(2) Relatively simple formulation;
(3) Accurate results in comparison with experimental data;
(4) Validity for general application in lightning problems is proved;

72
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(5) Frequency-domain effects (e.g. frequency dependence of soil parameters) are straightforward
representing without adjustments or simplifications;
(6) Time-domain effects (e.g. soil ionization and corona effect) are not directly represented;
(7) Similar to the FDTD method, the representation of large systems composed of several segments
reduces the computational efficiency of the model.

3.4.8 Comparison with experimental data

3.4.8.1 Transmission line response to direct strikes: Comparison with measured
waveforms
In [21], voltages across insulator strings of a 500-kV double-circuit transmission line due to the
application of an impulsive current to the top of a tower were compared by Visacro and Soares with
experimental results presented by Ishii et al. [159]. Figure 3.38 shows the obtained results and
denotes the good agreement between measured and calculated results, indicating HEM model
consistency.

Figure 3.38 Measured and calculated voltages across upper, middle, and lower insulator strings of a 500-
kV–and–62.8-m-high transmission line tower. Applied current wave: 3-µs-time to crest, 3.2-A crest value.37

3.4.8.2 Lightning-induced voltages and electromagnetic fields: Comparison with Ishii

et al. (1999)
Electric fields and lightning-induced voltages calculated by HEM model were compared by Silveira et
al. [150], with measured data obtained by Ishii et al. [160] in a reduced-scale experiment that
consisted of a 28-m-long and 0.25-mm-radius coiled wire representing a return-stroke channel
positioned close to a 25-m-long 0.5-m-high single wire with 0.25 mm radius. In the experiment, an
impulsive current was injected at the channel bottom by a voltage pulse generator and propagated
upward the channel at a 125 m/µs-speed. The near end of the line was either terminated with a 430-Ω
resistor or left open. The far end was terminated in a 430-Ω resistor. The induced voltages were
measured at both ends of the line using voltage probes with an input capacitance of 20 pF. The
resistivity (ρ) and the relative permittivity (εr) of the ground were estimated to be 16.67 Ωm and 10,
respectively. The same conditions were simulated using the HEM model with frequencies in the 200-
kHz-to-50 MHz range with a 200 kHz frequency step, corresponding to a 10 ns time step and a time
interval from 0 to 5 µs.
Figure 3.39 presents measured and calculated waveforms of vertical and horizontal electric fields at a
distance of 3 m from the return-stroke channel. The results agree quite well. Figure 3.40 indicated

37 Reprinted with permission from S. Visacro, and A. Soares Jr., HEM: A model for simulation of lightning-related engineering
problems, IEEE Trans. PWRD, vol. 20, no.2, p. 1207, Figure 3, 2005.

73
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(a) (b)
Figure 3.39 Measured and calculated electric field waveforms over a 16.67 Ωm ground, at a distance of 3
m from the return-stroke channel. (a) Vertical field and (b) Horizontal field.38

3 6
near end Measured 5 Measured
near end
2
Voltage (V)

HEM

Voltage (V)
4 HEM
1 3
2
0 1
-1 0
far end -1 far end
-2 -2
0 100 200 300 0 100 200 300
Time (ns) Time (ns)
(a) (b)
Figure 3.40 Measured and calculated induced-voltage waveforms at the close and distant ends of the line.
(a) Both ends of the line are terminated in a parallel circuit of 430-Ω resistance and 20-pF capacitance. (b)
The close end is terminated in a 20-pF capacitance and the distant end is terminated in a parallel circuit of
430-Ω resistance and 20-pF capacitance.39

simulated and measured induced voltage waveforms at the near and far ends of the overhead line.
Good agreement is achieved.

38Reprinted with permission from F. H. Silveira, S. Visacro, J. Herrera, and H. Torres, Evaluation of lightning-induced voltages
over lossy ground by the hybrid electromagnetic model, IEEE Trans. EMC, vol.51, no.1, p.158, Figure 4, 2009.
39Reprinted with permission from F. H. Silveira, S. Visacro, J. Herrera, and H. Torres, Evaluation of lightning-induced voltages
over lossy ground by the hybrid electromagnetic model, IEEE Trans. EMC, vol.51, no.1, p.158, Figure 4, 2009.

74
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

3.5 Summary
In this chapter, fundamental theories of four electromagnetic computation methods have been
explained. Advantages and disadvantages of each of these four methods are summarized in Table
3.1.

Table 3.1 Advantages and disadvantages of four electromagnetic computation methods presented in this
chapter.

Method Advantages Disadvantages

FDTD  It is based on a simple procedure,  It is computationally expensive
and therefore its programming is compared to other methods;
relatively easy;  It cannot deal with oblique
 It is capable of treating complex boundaries that are not aligned with
geometries and inhomogeneities; the Cartesian grid when the standard
 It is capable of incorporating orthogonal grid is employed, and
nonlinear effects and components; needs a staircase approximation for
and oblique boundaries; and
 It can handle wideband quantities  It would require a complex procedure
from one run with a time-to-frequency for incorporating dispersive
transforming tool. materials/media.
 It can be accerelarated efficiently
using parallel computing technqiues.
MoM  It is quite efficient in the analysis of  The frequency-domain method
thin-wire structures; cannot incorporate nonlinear effects
 Both frequency- and time-domain and components directly;
methods are available;  The time-domain method cannot
 The frequency-domain method can consider lossy ground;
consider lossy ground and wires in  The time-domain method would
lossy ground; and require a complex procedure for
 Also, the frequency-domain method incorporating dispersive
can consider frequency-dependent materials/media; and
effects and components.  The time-domain method has
sometimes long-term stability
problems.
PEEC  It is quite efficient in the analysis of  The frequency-domain method
thin-wire structures; cannot incorporate nonlinear effects
 Both frequency- and time-domain and components directly;
methods are available;  The time-domain method cannot
 It can incorporate lumped and consider lossy ground;
distributed circuit components; and  The time-domain method would
 It deals with and yields scalar require a complex procedure for
potential directly. incorporating dispersive
materials/media; and
 The time-domain method has
sometimes long-term stability
problems.
HEM  It is quite efficient in the analysis of  Only the frequency-domain
thin-wire structures; formulation is presently available;
 It can incorporate circuit components; and
and  It cannot incorporate nonlinear
 It deals with and yields scalar effects and components directly.
potential directly.

75
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4. Applications to lightning surge studies

4.1 Introduction
In this chapter, some examples of the application of the finite-difference time-domain (FDTD) method,
the partial-element equivalent-circuit (PEEC) method, and the hybrid electromagnetic model (HEM) to
lightning electromagnetic transient analysis are presented.
In Section 4.2, the FDTD method and the PEEC method are used for simulating the transient
response of a grounding structure, and in some simulations, the presence of an aerial wire to simulate
control cables and a soil ionization effect have also been considered.
In Section 4.3, the FDTD method is used for simulating the distortion characteristics of surge voltages
propagating on an overhead conductor and lightning overvoltages entering an air-insulated substation
when a transmission line tower is struck by lightning. In addition, the HEM is applied to calculating
lightning surge phenomena of a 138-kV transmission line, and the critical lightning currents to cause
back flashover and the probability of back flashover are estimated.
In Section 4.4, the FDTD method is applied to the simulation of lightning-induced voltages on a
distribution line with surge arresters, an overhead shielding wire, and the effect of transformers, and
the shielding effects of the surrounding building models are also studied. In addition, lightning-induced
voltages are calculated using the HEM to analyze the effect of the frequency-dependent electrical
parameters of the soil.
In Section 4.5, the transient response of the grounding system of a wind turbine is simulated using the
FDTD method, and the effect of the common grounding between a wind turbine and its nearby
lightning protection tower is also studied.
In Section 4.6, the FDTD method was applied to the lightning surge simulations of a microwave relay
station and the effect of the layout of the microwave relay station and the reinforcing bars of the
building and foundation on the current distribution was studied.
In Section 4.7, the effect of conducting structures near a lightning electromagnetic field measuring
station on errors of a magnetic direction finding of lightning was analyzed through FDTD simulations.
In addition, the FDTD simulation of a lightning electromagnetic pulse propagation was carried out
while considering the effect of a conducting atmosphere, and the effective ionospheric reflection height
was estimated.
In Section 4.8, the FDTD method was applied to the simulation of lightning surge phenomena of a
multilayer carbon fiber reinforced plastic (CFRP) by taking into account anisotropic conductivities. In
addition, the electric field distributions in and on a lightning surge arrester model including the effect of
a nonlinear material were simulated to study the effect of a shield ring using the FDTD method.
In Section 4.9, the applications of the FDTD method, the PEEC method, and the HEM to lightning
electromagnetic transient analysis presented in this Chapter are summarized.

4.2 Surges on grounding electrodes

4.2.1 FDTD simulation of a grounding grid with an aerial wire
4.2.1.1 Introduction
The operating voltages of low-voltage control circuits in power plants and substations have become
lower owing to the installation of digital-control equipment. This significantly increases the risk of faults
and malfunctions of such circuits due to transient voltages induced by electromagnetic interference.
The transient voltages arise mainly from a lightning current flowing through a grounding grid and a
main power circuit [161]. The control cables and wires are actually located in a three-dimensional
arrangement, which cannot be rigorously treated by simulation methods such as the electro-magnetic
transients program (EMTP) based on circuit theory. Recently, the finite-difference time-domain (FDTD)
method [16] is considered useful for this kind of surge analysis. In [162], Tatematsu et al. calculated
the ground potential rises (GPRs) of a grounding grid and induced voltages on a control wire above
the grounding grid using the FDTD method, when a surge current flows into the grounding grid, and
validated the calculated results with the associated measured values.

76
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.2.1.2 Experimental setup

Figure 4.1 shows a test system of the grounding grid in Akagi Testing Center of Cenral Research
Institute of Electric Power Industry (CRIEPI). Using the test system without the control wire, Tatematsu
et al. measured (i) the GPRs of the grounding grid with respect to a voltage reference wire and (ii) the
currents flowing through the grounding grid [162]. The grounding grid whose size is 30 m × 60 m with
10 m × 10 m meshes is located 0.5 m below the ground surface. The grounding grid is composed of a
bare copper wire with a cross-section of 60 mm2. The pulse generator is placed 0.5 m above the
ground surface to inject a current into the grounding grid, and is connected to the grounding grid by an
IV (Vinyl-Insulated) wire with a cross-section of 8 mm2. The current injection and voltage reference
wires are situated parallel to the ground surface, at heights of 0.5 m and 1.5 m, respectively. Both the
current injection and voltage reference wires are represented by IV wires with a cross-section of 2
mm2. The length of the current injection wire is 155 m, and its far end is grounded with a matching
resistance. The far end of the voltage reference wire is also grounded with a matching resistance.

matching resistance

voltage reference wire

30 m

grounding grid
pulse generator
current voltage reference wire
120 m injection wire
60 m
1m matching
resistance
45.5 m 0.5 m ground surface
matching 0.5 m
155 m resistance grounding grid
pulse generator
matching resistance
current injection wire

(a) (b)
Figure 4.1 Experimental setup for measuring the ground potential rises of a grounding grid. (a) Top view
and (b) Side view.40

In addition, Tatematsu et al. measured the induced voltages on the control wire above the grounding
grid [162]. Although the control wires used in the low-voltage control circuit has many core wires or
metal sheathes, the control wire is simply represented by an open-end IV wire with a cross-section of
8 mm2 and a length of 30 m. As shown in Figure 4.2, the wire is located parallel to the ground surface,
and its height is 0.5 m.
In this work, the following items were measured: (i) GPRs of the grounding grid, (ii) GPRs of the
control wire, (iii) voltage differences between the grounding grid and the control wire, and (iv) currents
flowing through the grounding grid.

4.2.1.3 FDTD simulation

The calculation arrangement used in the FDTD method is similar to the experimental one shown in
Figure 4.1. The dimensions of the analysis space is 330 m × 360 m × 300 m, and is divided into 356 ×
451 × 220 cells. All surfaces of the analysis space are treated as absorbing boundaries using Liao’s
formulation of the second order to assume an open space [37]. The bottom space with a thickness of
150 m corresponds to the ground soil. The grounding grid, current injection wire, and voltage
reference wire are simulated using the thin-wire representation techniques [44], [45]. The sizes of the
six cells adjacent to each thin wire are set to 0.25 m, while the sizes of the other cells become
gradually larger further away from the thin wires and the sizes of the other cells range from 0.3 m to 2

40Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1246, Figure 1, 2009.

77
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

IV wire

30 m

15 m

pulse generator

Figure 4.2 Experimental setup for measuring the induced voltages on an aerial wire.41

m. On the other hand, the time discretization is set to 0.2 ns based on Courant's condition. Since the
far end of the current injection wire has little influence on the calculated results, it is connected to the
absorbing boundary to simulate the experimental arrangement. On the other hand, the position of the
far end of the voltage reference wire is closer to the grounding grid than that of the current injection
wire, and its position affects the calculated results. For this reason, the far end of the voltage reference
wire is located at the same position with that of the experimental arrangement and it is grounded with
its matching resistance.
Although the electric parameters of the soil at over 40 m deep are unknown, the fourth layer has little
influence on the calculated results. Here, the ground soil is assumed to have three layers and the
resistivity of each layer is estimated on the basis of the measured results using 4-Point Wenner Test,
as shown in Figure 4.3. The relative permittivity of the ground soil is unclear, and thus the relative
permittivities of all the layers are assumed to be the same. The relative permittivity affect the
propagation speed of the current, and the relative permittivity is set to 30 by comparing the calculated
results of the current propagating through the grounding grid with the measured results.

ground surface
250 Ωm
depth : 2.5 m
2100 Ω m
depth : 18 m

500 Ω m

bottom of the analysis space

Figure 4.3 Resistivities of the soil simulated in the FDTD method.42

The pulse generator is represented by a voltage source, of which the output waveform simulates a
measured result of the output voltage of the pulse generator with an open-end.
The control wire is simulated using the thin wire representation technique similarly to the grounding
grid. To take into account the effect of the input capacitance of a differential voltage probe used for
measuring the voltage differences between the aerial wire and the grounding grid, a lumped-
parameter capacitance of 3.5 pF is inserted between them for calculating the voltage differences.

41Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1247, Figure 3a, 2009.
42Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1248, Figure 5, 2009.

78
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.2.1.4 Comparison between the measured and calculated results

Figure 4.4 shows the waveform of the current flowing into the grounding grid obtained by the
calculation and the experiment.

Figure 4.4 Waveforms of the current injected into the grounding grid.43

As shown in Figure 4.5, a local-coordinate system is determined on the grounding grid to designate
positions, where the current injection point corresponds to (20, 0). The arrows in Figure 4.5 will be
referred to in Figure 4.8.
(0, 60) (30, 60)

x
(0, 0) (30, 0)

current-injection point

Figure 4.5 Local coordinate system determined on the grounding grid.44

Using the aforementioned calculation arrangement, Tatematsu et al. calculated the GPRs of the
grounding grid in the absence of the aerial wire in the first step. As an example of the results, Figure
4.6 shows the GPRs of the grounding grid at (20, 15), (20, 30), and (20, 60).
potential rise [V/A]

potential rise [V/A]

potential rise [V/A]

calculation calculation calculation

measurement measurement measurement
10 10 10

0 0 0

0 2 4 6 8 0 2 4 6 8 0 2 4 6 8
time [  s] time [  s] time [  s]

(a) (b) (c)

Figure 4.6 Calculated and measured GPRs of the grounding grid. (a) (20, 15), (b) (20, 30), and (c) (20, 60).45

43Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1248, Figure 8, 2009.
44Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1248, Figure 9, 2009.
45Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1249, Figures 10a, 10b, and 10d, 2009.

79
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The GPRs get deformed with increasing distance from the current injection point (20, 0). The
deformations of the waveforms are reproduced well in the calculation and the differences between the
average calculated potential rises at 6-8 μs and the measured potential rises are within 5%. The
calculated results agree well with the measured values. Figure 4.7 shows the average ground-
potential rises at 6-8 μs at the positions x = 30 to 60 m and y = 30 m outside the grounding grid.

10

potential rise [V/A]

calculation
measurement

0
30 40 50 60
x [m]

Figure 4.7 Calculated and measured GPRs outside the grid.46

The calculated GPRs outside the grounding grid also correspond to the measured GPRs.
Figures 4.8(a) and (b) show the waveforms of the currents flowing through the grounding grid at the
positions x = 20 m and y = 0 to 60 m, and x = 20 to 30 m and y = 30 m, respectively. The positions are
designated as arrows in Figure 4.5, and the directions of the arrows correspond to those of the flowing
currents. The currents in Figures 4.8(a) and (b) propagate in the directions of the y-axis and x-axis,
respectively. The average calculated currents at 6-8 μs are similar to the measured values within a
difference of 0.008 A. Close agreement is also found between the calculated and measured currents.
As shown in Figure 4.8(b), the sign of the currents flowing through the grounding grid from (20, 30) to
(30, 30) inverts on the way, and such a phenomenon is reproduced in the calculation. In addition,
similar phenomena occur at the positions between (0, 60) and (10, 60), (10, 60) and (20, 60), and (20,
60) and (30, 60) in the calculated and measured results, while such a phenomenon isn't observed at
the other positions. Although the reasons for the phenomenon remain unclear, the electromagnetic
coupling between the currents flowing through the grounding grid is considered to be one of the
reasons.

(a) (b)
Figure 4.8 Calculated and measured currents through the grounding grid. (a) Currents flowing from (20, 0)
to (20, 60) and (b) Currents flowing from (20, 30) to (30, 30).47

Next, Tatematsu et al. calculated the GPRs of the grounding grid and the aerial wire, the induced
voltages on the aerial wire above the grounding grid, and the currents flowing through the grounding
grid when the wire was placed over the grounding grid, and compared the calculated results with the

46Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1249, Figure 11, 2009.
47Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1249, Figure 12, 2009.

80
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

measured ones [162]. Figure 4.9 shows the calculated and measured results. Figures 4.9 (a) and (b)
show the GPRs of the aerial wire and the voltage differences between the grounding grid and the
aerial wire at the positions (20, 15) and (20, 45), respectively. The GPRs of the grounding grids and
aerial wires, and the voltage differences obtained by the calculation and the measurement agree well
with each other. The validity of the application of the FDTD method to induced voltage analysis in the
aerial wire above the grounding grid to simulate control cables is confirmed.

voltage difference [V/A]

10 calculation
potential rise [V/A]

10 calculation
(20, 15) measurement measurement

5 5
(20, 15)
(20, 45)
0 0
(20, 45)
0 2 4 6 8 0 2 4 6 8
time [  s] time [  s]

(a) (b)
Figure 4.9 Calculated and measured GPRs of the aerial wire and the voltages induced on the aerial wire.
(a) GPRs of the aerial wire and (b) Induced voltages.48

4.2.2 FDTD simulation of grounding electrodes considering soil ionization

4.2.2.1 Introduction
The role of grounding electrodes is to drain system fault and lightning currents effectively into the soil,
and thereby to mitigate damage of equipment of telecommunication systems and electrical power
systems. Thus, the performance of such systems is influenced by proper functioning of grounding
electrodes. If a high surge current flows in a grounding electrode, the soil in the vicinity of the
grounding electrode would be ionized and the voltage generated at the top of the grounding electrode
would be reduced.
Recently, using electromagnetic computation methods such as the method of moments (MoM) [13]
and the FDTD method [16], surge characteristics of grounding electrodes have been studied (e.g.,
[12], [15]). The FDTD method appears to be more suitable for analyzing surges with considering
nonlinear effects or phenomena such as soil ionization (e.g., [163]-[165]), although useful techniques
for considering soil ionization in MoM-based computations (e.g., [166]-[168]) have been proposed.
Note that soil ionization models have also been proposed for circuit-theory-based simulations (e.g.,
[169]-[175]).
In [163], Ala et al. have proposed a soil ionization model, on the basis of the dynamic soil-resistivity
model of Liew and Darveniza [169], for FDTD computations. In the model, the resistivity of each soil-
representing cell is controlled by the instantaneous value of the electric field there and time. Ala et al.
have tested its validity against impulse high-current experiments on a 0.61-m single vertical grounding
conductor buried in a 50-Ωm soil (for a single-peak unipolar current having a magnitude of 3.5 kA and
a risetime of 5 μs) and a 3.05-m single vertical grounding conductor buried in a 87-Ωm soil (for a
similarly-shaped current of 13 kA and 7.5 μs) [163].
In this section, in order to test the validity of this model more thoroughly, the model is applied to
analyzing the surge response of a horizontal grounding conductor having a length of 8.1 m buried in a
soil of resistivity 200 Ωm to a 22.2-kA current, and that of a vertical grounding conductor having a
length 0.6 m buried in a soil of resistivity 50 Ωm to a 3.5-kA current. These FDTD-computed
responses are compared with the corresponding responses measured by Sekioka et al. [170] and
Liew and Darveniza [169]. Also, in order to show the validity of the model for more complex
configurations, the surge response of four parallel vertical grounding conductors of length 3.05 m to a
12-kA current are analyzed, and the FDTD-computed response is compared with the corresponding
response measured by Bellaschi et al. [176].

48Reprinted with permission from A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on
an aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy Society, vol. 129, no. 10, p.
1250, Figures 13b and 13c, 2009.

81
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.2.2.2 Representation of soil Ionization and de-Ionization

In this section, the soil ionization and de-ionization model proposed by Ala et al. [163] is briefly
described, based on the dynamic model of Liew and Darveniza [169], for FDTD surge computations.
Figure 4.10 shows the resistivity profile of Ala et al.’s model. In the model, the resistivity of each soil-
representing cell is controlled by the instantaneous value of the electric field there and time.
When the instantaneous value of the electric field E at a soil-representing cell is lower than the critical
electric field Ec, the resistivity ρ is equal to its steady-state value ρ0.
 (t )  0 Equation 4.1

When E at a soil-representing cell exceeds Ec, ρ begins to decrease with time as follows:

 t 
 (t )  0 exp    Equation 4.2
 1 
where t is time defined so that t = 0 at the instant when E attains Ec, and τ1 is the ionization time
constant. This resistivity decreasing with time represents the soil-ionization process.
When E at a cell in the ionized-soil region falls below Ec, ρ begins to increase with time as follows:

  t  
2
E
 (t )  i   0  i  1  exp     1   Equation 4.3
   2    Ec 
where ρi is the minimal resistivity value reached by the ionization process, t is time defined so that t =0
at the instant when E = Ec, and τ2 is the de-ionization time constant. This resistivity increasing with
time from ρi to ρ0 represents the de-ionization process of soil.
It follows from the above that this engineering model has three adjustable constants: critical electric
field Ec, and ionization time constant τ1 and τ2.

Figure 4.10 Resistivity profile of Ala et al.’s soil-ionization and de-ionization model (Ala et al. 2008), based
on the dynamic model proposed by Liew and Darveniza (1974), used for FDTD computations. 49

4.2.2.3 Analysis and results

4.2.2.3.1 Response of a single horizontal grounding conductor to a 22.2-kA current
Figure 4.11 shows a single horizontal grounding conductor having a length of 8.1 m buried at a depth
of 1 m in a homogeneous soil of resistivity ρ0 =200 Ωm, to be analyzed using the FDTD method. The
horizontal grounding conductor is energized by a lumped current source that is connected to an
auxiliary vertical grounding conctor of length 1 m via a 20 m long horizontal overhead wire. This

49 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
163, Figure 1, 2015.

82
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

configuration represents an experiment carried out by Sekioka et al. [170]. Note that the employed soil
resistivity, 200 Ωm, was the value measured at the experimental site.

Figure 4.11 A single horizontal grounding conductor having a length of 8.1 m in a homogeneous soil of
ρ0=200 Ωm, energized by a lumped current source that is connected to an auxiliary vertical grounding
conductor of length 1 m, analyzed using the FDTD method.50

The working volume of 200 m × 680 m × 50 m is divided nonuniformly and surrounded by six planes of
Liao’s second-order absorbing boundary condition [37] to minimize unwanted reflections there. Cell
sizes are not constant: 20 mm in the vicinity of the horizontal grounding conductor, and increasing
gradually. The thickness of the homogeneous soil layer (distance between the ground surface and the
bottom absorbing boundary) is set to 25 m, while the height of the working volume is 50 m. The time
increment is set to 0.039 ns.
The equivalent radius of the horizontal grounding conductor is 4.6 mm (≈ 0.23 Δs =0.23 × 20 mm,
where Δs is the minimum cell dimension) [44], [45], as needed for representing the corresponding
grounding conductor of radius equal to 4.4 mm of Sekioka et al.’s experiment [170].
Figure 4.12 shows the waveform of voltage at the current-injection point into the 8.1-m horizontal
grounding conductor, computed using the FDTD method with the soil-ionization model, for a current
having a positive magnitude of 22.2 kA and a risetime of 2.3 μs, and the corresponding measured
waveform [170]. In the present FDTD computations, the voltage is evaluated integrating tangential
electric field on the ground surface from a point 340 m away from the current-injection point in a
direction perpendicular to the horizontal grounding conductor.

1000
Measured
800 Computed

600 Without Soil Ionization

Voltage [kV]

400
With Soil Ionization
200

-200
-5 0 5 10 15 20 25 30
Time [s]

Figure 4.12 FDTD-computed waveform of voltage at the current-injection point into the 8.1-m horizontal
grounding conductor (with the soil-ionization model, Ec=200 kV/m, τ1=3.5 μs, and τ2=4.5 μs), and the

50 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
165, Figure 6, 2015.

83
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

corresponding measured waveform (Sekioka et al. 1998). Also shown is the FDTD-computed voltage
waveform without the soil-ionization model.51

In the computation, Ec=200 kV/m, τ1=3.5 μs, and τ2=4.5 μs are employed. Since in Liew and Darveniza
[169], Ec, τ1, and τ2 have been selected within relatively narrow ranges: Ec=50 to 200 kV/m, τ1=1.5 to 2
μs, andτ2=0.5 to 4.5 μs for soil resistivity within 50 to 310 Ωm, values of these three parameters were
selected by a trial-and-error process considering these ranges. Also shown in Figure 4.12 is the
voltage waveform computed without the soil-ionization model.
It appears from Figure 4.12 that the FDTD-computed voltage waveform with the soil ionization for the
8.1-m horizontal grounding conductor agrees reasonably well with the corresponding measured
waveform. The peak voltage is reduced by 34%.
Figures 4.13 (a), (b), and (c) shows the resistivity distributions of the soil around the 8.1 m horizontal
grounding conductor at time t=0 μs, 0.8 μs, and 3.9 μs, respectively, computed with the soil-ionization
model. It is apparent from Figure 4.13 that the region of soil ionization expands with increasing
current, and the soil ionization is more significant below the horizontal grounding conductor.

11 m

2.2 m
= 0 [Ωm]

(a) = 200 [Ωm]

11 m

2.2 m
= 0 [Ωm]

(b) = 200 [Ωm]

11 m

2.2 m
= 0 [Ωm]

(c) = 200 [Ωm]

Figure 4.13 Resistivity distributions of the soil around the 8.1-m horizontal grounding conductor at time
(a) t=0 μs, (b) t=0.8 μs, and (c) t=3.9 μs.52

4.2.2.3.2 Response of a single vertical grounding conductor to a 3.5-kA current

Figure 4.14 shows a single vertical grounding conductor of length 0.6 m in a homogeneous soil of
resistivity ρ0 = 50 Ωm, to be analyzed using the FDTD method. The vertical grounding conductor is
energized by a lumped current source, which generates a current having a magnitude of 3.5 kA and a
risetime of 5 μs. Four auxiliary vertical conductors are used in order to provide the current return path

51 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
165, Figure 7, 2015.
52 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
165, Figure 8, 2015.

84
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

in the soil. This configuration represents an experiment carried out by Liew and Darveniza [169]. Note
that the employed soil resistivity, 50 Ωm, was the value measured at the experimental site.
The working volume of 65 m × 65 m × 20 m is divided nonuniformly and surrounded by six planes of
Liao’s second-order absorbing boundary condition [37] to minimize unwanted reflections there. Cell
sizes are not constant: 30 mm in the vicinity of the vertical grounding conductor, and increasing
gradually. The thickness of the soil layer (distance between the ground surface and the bottom
absorbing boundary) is set to 2.5 m, while the height of the working volume is 20 m. The time
increment is set to 0.067 ns.

Figure 4.14 A single vertical grounding conductor of length 0.6 m in a homogeneous soil of ρ0 = 50 Ωm,
energized by a lumped current source that is connected to four auxiliary vertical conductors, to be
analyzed using the FDTD method.53

Figure 4.15 shows the waveform of voltage at the top of the 0.6-m vertical grounding conductor,
computed using the FDTD method with the soil-ionization model, and the corresponding measured
waveform. Note that Ec = 200 kV/m, τ1 =1.3 μs, and τ2 = 4.5 μs are employed for the soil ionization
model. Also shown in Figure 4.15 is the voltage waveform computed without the soil-ionization model.
The voltage is evaluated integrating tangential electric field on the ground surface from a point 32.5 m
away from the current-injection point. The FDTD-computed voltage waveform with the soil-ionization
agrees well with the corresponding measured waveform: the FDTD-computed peak voltage (87 kV) is
only 0.6 % higher than the measured peak voltage (87.5 kV). The peak voltage computed is reduced
from 174 kV to 87 kV by the soil ionization (about 50% reduction).

180
Measured
160
Computed
140

120 Without Soil Ionization

Voltage [kV]

100

80

60

40 With Soil Ionization

20

0
0 5 10 15 20
Time [s]

Figure 4.15 FDTD-computed waveforms of voltage at the top of the vertical grounding conductor of length
0.6 m with the soil-ionization model (Ec = 200 kV/m, τ1 = 1.3 μs, andτ2 = 4.5 μs), and the corresponding
measured waveform (Liew and Darveniza 1974). Also shown is the FDTD-computed voltage waveform
without the soil-ionization model.54

53 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
163, Figure 2, 2015.
54 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
164, Figure 4, 2015.

85
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.16 (a), (b), and (c) shows the resistivity distributions of the soil in the vicinity of the 0.6-m
vertical grounding conductor under test at time t = 0, 2.6, and 5.2 μs, respectively, computed with the
soil-ionization model. It appears from Figure 4.16 that the region of soil ionization expands with
increasing current and the soil ionization is significant near the bottom tip of the grounding conductor.

0.96 m 0.96 m 0.96 m

0.96 m 0.96 m 0.96 m

= 0 [Ωm] = 0 [Ωm] = 0 [Ωm]
Vertical
grounding
conductor
= 50 [Ωm] = 50 [Ωm] = 50 [Ωm]

(a) (b) (c)

Figure 4.16 Resistivity distributions of the soil in the vicinity of the 0.6-m vertical grounding conductor at
time (a) t = 0 μs, (b) 2.6 μs, and (c) 5.2 μs. The resistivity before the soil ionization is ρ0 = 50 Ωm.55

4.2.2.3.3 Response of four vertical grounding conductors to a 12-kA current

Figure 4.17 shows four vertical grounding conductors of length 3.05 m in a soil of resistivity ρ0=63 Ωm,
to be analyzed using the FDTD method. The four vertical grounding conductors are energized at their
center point by a lumped current source whose other terminal is connected to an auxiliary grounding
conductor via an overhead wire. This configuration represents an experiment carried out by Bellaschi
et al. [176]. The current source generates a current having a magnitude of 12 kA and a risetime of 20
μs, which also simulates a current waveform injected by Bellaschi et al. in their experiment.

Figure 4.17 Four vertical grounding conductors of length 3.05 m in a homogeneous soil of ρ0=80 Ωm,
energized by a lumped current source that is connected to an auxiliary grounding conductor, analyzed
using the FDTD method.56

The working volume of 150 m × 75 m × 25 m is divided nonuniformly, and surrounded by six planes of
Liao’s second-order absorbing boundary condition [37]. Cell sizes are not constant: 55 mm in the
vicinity of the vertical conductor, and increasing gradually. The thickness of the soil layer is set to 15
m, while the height of the working volume is 25 m. The time increment is set to 0.105 ns.
The equivalent radius of each vertical grounding conductor is 12.7 mm (≈0.23 Δs =0.23 × 55 mm) [44],
[45], as needed for approximately representing the corresponding grounding rod of radius equal to 25
mm of Bellaschi et al.’s experiment [176].
Figure 4.18 shows the waveform of voltage at the top of the four vertical grounding conductors,
computed using the FDTD method with the soil-ionization model, and the corresponding measured

55 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
164, Figure 5, 2015.
56 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
168, Figure 13, 2015.

86
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

waveform [176]. Note that, following Liew and Darveniza (Table 4 in [169]), Ec=150 kV/m, τ1=2 μs, and
τ2=4.5 μs were employed for the soil ionization model. Also shown in Figure 4.18 is the voltage
waveform computed without the soil-ionization model. The voltage is evaluated by integrating
tangential electric field on the ground surface from a point 60 m away from the current injection point
in the direction perpendicular to horizontal current lead wire.

90
Without Measured
80 Soil Ionization Computed
70

Voltage [kV] 60

50
With
40 Soil Ionization

30

20

10

0
0 5 10 15 20 25 30 35 40
Time [s]

Figure 4.18 FDTD-computed waveform of voltage at the top of the four vertical grounding conductors of
length 3.05 m with the soil-ionization model (Ec=150 kV/m, τ1=2 μs, and τ2=4.5 μs), and the corresponding
measured waveform (Bellaschi et al. 1942). Also shown is the FDTD-computed voltage waveform without
the soil-ionization model.57

The FDTD-computed voltage waveform with the soil-ionization agrees well with the corresponding
measured waveform. The computed peak voltage is reduced from 88 kV to 75 kV (15%) by the soil
ionization.

4.2.2.4 Summary
In this section, in order to test the validity of the soil-ionization model proposed by Ala et al. [163] for
FDTD simulations, it has been applied to analysis of the surge responses of grounding electrodes: a
horizontal grounding conductor of length 8.1 m, a vertical grounding conductor of length 0.6 m, and
four parallel vertical conductors of length 3.05 m, buried in soils with resistivities of 50 to 200 Ωm. The
FDTD-computed responses with the soil-ionization model for these grounding electrodes agree
reasonably well with the corresponding measured surge responses. The values of three adjustable
constants, critical electric field Ec, ionization constant τ1, and de-ionization constant τ2, employed in
this model lie within relatively narrow ranges: Ec = 150 to 200 kV/m, τ1 = 1.3 to 3.5 μs, and τ2 = 4.5 μs.
For a fast-rising current, a smaller value of τ1 within this range needs to be employed.

4.2.3 PEEC-based simulations of grounding electrodes

In this subsection, the partial element equivalent circuit (PEEC) methods either including or neglecting
retardation effect are utilized for analyzing transient behaviors of grounding electrodes. In [177], the
PEEC methods have been validated with simple structures of grounding electrodes. Close agreement
was observed. For confirming validity of the PEEC methods in more complicated cases, in the first case,
ground potential rises of a ground mesh and lightning effective area of a grounding mesh are
investigated. Calculated results by the PEEC method are compared with the MoM-calculated results
[178], [179]. Also, the experiment on a ground grid carried out by Stojkovic et al. [180] was selected as
the second case. In the third case, a T-shape electrode [163] was also selected as the other test case
for comparison of the FDTD and PEEC computed results. In this case, the soil ionization model is
included in the PEEC method.

57 Reprinted with permission from H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, “FDTD surge simulation of grounding
electrodes considering soil ionization,” in Proc. the 9th Asia-Pacific International Conference on Lightning (APL), no. TC4.1-2, p.
168, Figure 14, 2015.

87
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

In the first case, the mesh is a grounding grid composed of horizontal electrodes. The mesh sizes of
1x1, 2x2, and 6x6 grids in Figure 4.19 are investigated. Each mesh is composed of a copper wire of 7-
mm radius, and buried at a depth of 0.5 m. The soil conductivity is set to be 1 mS/m, and the relative
permittivity is set to be 9. An impulse current with a peak value of 1 A [178] is injected to point A of the
ground mesh. The voltage at point A is calculated.

Figure 4.19 The structures of the grids 1x1, 2x2, and 6x6 the simulation (Grcev et al. 1996).58

The same conditions are simulated by the full-wave PEEC methods in a frequency range from 0 to 10
MHz with a frequency step of 39.06 kHz. The observation time is up to 20 µs for a comparison with the
simulated results [178]. An injected current waveform is taken from [178] by piece-wise linear
approximation as shown in Figure 4.20 for a comparison with the MoM- calculated results [178].

Figure 4.20 Applied current in the simulation.59

Figure 4.21 shows the calculated results by the PEEC method [133] and by the MoM [178]. The
PEEC-calculated results agree well with the MoM-calculated results. The maximums of transient
grounding potential rises are equal for all grids with sizes from 20 x 20 m 2 to 60 x 60 m2, which implies
that the effective area of all analyzed grids is less than 20 x 20 m 2.
The configuration of the experiment of transient performance of a square ground grid carried out by
Stojkovic et al [180] is illustrated in Figure 4.22. The square ground grid has 10-m sides and four
meshes that are constructed of copper conductors with 4-mm radii and buried at a depth of 0.5 m. The
soil resistivity of the equivalent uniform model is estimated to be 35 Ωm. An impulse current was

58 Reprinted with permission from P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, "Application of the partial element
equivalent circuit method to analysis of transient potential rises in grounding systems," IEEE Trans. EMC, vol. 53, no. 3, p. 730,
Figure 11, 2011.
59 Reprinted with permission from P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, "Application of the partial element
equivalent circuit method to analysis of transient potential rises in grounding systems," IEEE Trans. EMC, vol. 53, no. 3, p. 730,
Figure 12, 2011.

88
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

applied to the point denoted by “A”. The voltages were recorded at the injection point and the point
denoted by “B” of the ground grid as shown in Figure 4.22. The injected current amplitudes being
relatively low, soil ionization effect was unlikely to appear and was not considered in the computation.

60

50

40
Voltage [V]

30

20

10

0
0 2 4 6 8 10 12 14 16 18 20

Figure 4.21 Comparison of PEEC-calculated waveforms (Yutthagowith et al. 2011) and MoM-calculated
waveforms (Grcev et al. 1996) at point A of the grid grounding systems.60

20
Voltage [V] or Current [A]

18 A B
16
14 10 m
12
10 10 m

8
6
Injected current
4
2
0
0 20 40 60 80 100 120 140 160 180 200
Time [s]
Measurement [32]
Computed by the full-wave PEEC method
Computed by the PEEC method neglecting retardation effect

Figure 4.22 Comparison of measured and PEEC-computed transient potentials at point A of the square
ground grid.61

The same conditions were simulated by the PEEC methods in the time domain neglecting the
retardation effect with a 500 ns time step and the full-wave PEEC method in the frequency range from
7.8125 kHz to 2 MHz with a frequency step 7.8125 kHz. The relative permittivity of soil was set to be
10.
Figure 4.22 and Figure 4.23 show comparison results between experimental results and those
calculated by the PEEC method neglecting retardation effects in the time domain and the full-wave
PEEC method in the frequency domain.

In the third case, the results computed by the PEEC method neglecting the retardation effect are
compared with the FDTD-computed result in [163]. The T-shape electrode under consideration is
composed of perfect conductors with 14-mm radii as shown in Figure 4.24. The conductivity and

60 Reprinted with permission from P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, "Application of the partial element
equivalent circuit method to analysis of transient potential rises in grounding systems," IEEE Trans. EMC, vol. 53, no. 3, p. 731,
Figure 13a, 2011.
61 Reprinted with permission from P. Yutthagowith, A. Ametani, F. Rachidi, N. Nagaoka and Y. Baba, "Application of a partial
element equivalent circuit method to lightning surge analyses," Electric Power Systems Research, vol. 94, p. 36, Figure 10,
2013.

89
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

relative permittivity of soil are set to be 0.02 S/m and 10, respectively. The same configuration and
conditions were simulated by the PEEC method neglecting retardation in the time domain with a 200
ns time step. The current was injected at the top end of the electrode. In this case, the injected current
is high enough to initiate soil ionization. Therefore, the soil ionization model was included in the
simulation.
20

Voltage [V] or Current [A]

18
16
14
12
10
8
6
Injected current
4
2
0
0 20 40 60 80 100 120 140 160 180 200
Time [s]
Measurement [32]
Computed by the full-wave PEEC method
Computed by the PEEC method neglecting retardation effect

Figure 4.23 Comparison of measured and PEEC-computed transient potentials at point B of the square
ground grid.62

0.61 m

0.61 m
Soil: 0.02S/m, r = 10, Ec = 300 kV/m

Figure 4.24 Configuration of the T-shape grounding electrode and soil parameters.63

During the soil ionization process a longitudinal current along the elements still flows inside the
conductors, and thus it is estimated that the series-, self-, and mutual impedances of the elements are
not changed and only the shunt admittances of the elements are affected by the soil ionization. During
the soil ionization process, the radii of the elements are varied when the transverse electric field on the
surface of the element is over the critical electric field strength of soil (Ec). In the simulation Ec was set
to be 300 kV/m. The transverse electric field on the boundary of the ionized zone can be calculated by
the following equation [171].
ITi E
 c Equation 4.4
2 ri li 
where ITi, ri, and li is the transverse current, the radius, and length of the i-th element, respectively. ρ is
the resistivity of soil.
Figure 4.25 shows a comparison of the FDTD-computed results [163] and the PEEC computed results
of transient potential of a T-shape electrode when soil ionization is neglected and included in the
simulation. The figure shows a satisfactory agreement of the calculated results by the PEEC method
with the FDTD-computed results [163].

62 Reprinted with permission from P. Yutthagowith, A. Ametani, F. Rachidi, N. Nagaoka and Y. Baba, "Application of a partial
element equivalent circuit method to lightning surge analyses," Electric Power Systems Research, vol. 94, p. 36, Figure 11,
2013.
63 Reprinted with permission from P. Yutthagowith, A. Ametani, F. Rachidi, N. Nagaoka and Y. Baba, "Application of a partial
element equivalent circuit method to lightning surge analyses," Electric Power Systems Research, vol. 94, p. 36, Figure 13,
2013.

90
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

In fact, the PEEC method neglecting the retardation effect assumes quasi-static electromagnetic
fields. The dimensions of the considered system should be smaller than the minimum wavelength of
the applied source. A satisfactory agreement is still observed not only for the amplitudes but also for
the waveshapes in all cases.

120
Solid line: computed by the PEEC method
100 Dash line: computed by the FDTD method
No ionization
Voltage [kV]

80
Ionization
60

40
30 kA
20 Injected current
0
0 5 10 15 20 25 30
Time [s]
Figure 4.25 Comparison of the FDTD and PEEC-computed transient potentials and the injected current at
the current injected point of the T-shape grounding electrode.64

4.3 Overhead power transmission lines and towers

4.3.1 FDTD simulation of lightning surges on an overhead conductor in the presence
of ground-wire corona discharge
Thang et al. have applied the simplified corona-discharge model [69] to analyzing lightning surges
propagating along overhead wires with corona discharge [181]. The FDTD-computed waveforms
(including wavefront distortion and attenuation at later times) of surge voltages at three distances from
the energized end of the wire agree reasonably well with the corresponding measured waveforms.
Also, the FDTD-computed waveforms of surge voltages induced on a nearby parallel bundled
conductor agree fairly well with the corresponding measured waveforms.
Figure 4.26 (a) shows 3-D view of a 12.65-mm radius, 1.4-km-long overhead horizontal single
perfectly conducting wire located 22.2-m above ground of conductivity 10 mS/m and a 1.4-km-long
bundled perfect conductor (four conductors in the bundle) located 14-m above the same ground with a
horizontal separation of 2-m from the single wire. The radius of each conductor of the bundle is 11.5
mm and the distance between conductors is 0.4 m. One end of the single wire is energized by a
lumped voltage source, and the other end is connected to the ground via a 490-Ω (matching) resistor.
The bundled conductor is not grounded in order to reproduce the configuration of Inoue’s experiment
[182]. The four conductors in the bundle are electrically connected at the sending and receiving ends.
Corona discharge is assumed to occur only on the energized single wire.
For FDTD computations, this conductor system is accommodated in a working volume of 60 m × 1460
m × 80 m, which is divided nonuniformly into rectangular cells and is surrounded by six planes of
Liao’s second-order absorbing boundary condition [37] to minimize reflections there. At each ground
connection point, a perfectly conducting grounding electrode of 20 m × 20 m × 10 m is employed
(although no information on the geometry or grounding resistance value of grounding electrodes used
in the experiment is available in [182]). The side length in the y-direction of all cells is 1 m (constant).
Cell sides along the x- and z-axes are not constant: 5.5 cm in the vicinity (220 cm × 220 cm) of the
horizontal single wire, increasing gradually (to 10, 20 and 100 cm) beyond that region, except for a
region around the bundled conductor, and 5 cm in the vicinity (80 cm × 80 cm) of the bundled
conductor, except for a region around the horizontal single wire, increasing gradually (to 10, 20, and
100 cm) beyond that region, as shown in Figure 4.26 (b).

64 Reprinted with permission from P. Yutthagowith, A. Ametani, F. Rachidi, N. Nagaoka and Y. Baba, "Application of a partial
element equivalent circuit method to lightning surge analyses," Electric Power Systems Research, vol. 94, p. 36, Figure 14,
2013.

91
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

The critical electric field E0 on the surface of cylindrical wire of radius r0 for initiation of corona
discharge is 2.4 MV/m for m = 0.5 [71]. The critical background electric field [72] necessary for
streamer propagation is set to 0.5 MV/m for positive, and 1.5 MV/m for negative polarity, respectively
[73]. The ionization process is roughly approximated by increasing the conductivity of the corona-
discharge region from zero to σcor = 40 μS/m. The statistical inception delay and streamer
development process are simply ignored.

Figure 4.26 (a) 3D and (b) cross-sectional (xz-plane) views of a horizontal single wire and a four-
conductor bundle of length 1.4 km located 14-m above the ground and horizontally 2-m away from the
single wire.65

Figures 4.27 and 4.28 show, for positive and negative applied voltages, respectively, waveforms of
surge voltage at d = 0, 350, 700, and 1050 m from the energized end of the horizontal single wire
above ground whose conductivity is 10 mS/m, computed using the FDTD method for corona-region
conductivity σcor = 40 μS/m. The critical electric field for corona onset on the wire surface was set to E0
= 2.4 MV/m (for m = 0.5). The corresponding measured waveforms [182] are also shown in these
figures. The peak voltages are 1580 kV for positive polarity (see Figure 4.27), and 1670 kV for
negative polarity (see Figure 4.28). FDTD-computed waveforms agree reasonably well with the
corresponding measured waveforms. Both FDTD-computed and measured waveforms of surge
voltage suffer distortion, which becomes more significant with increasing the applied voltage peak and
the propagation distance. Note that plateaus in FDTD-computed voltage waveforms seen in these

65 Reprinted with permission from T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov,
“FDTD simulation of lightning surges on overhead wires in the presence of corona discharge,” IEEE Trans. EMC, vol. 54, no. 6,
p. 1235, Figure 1, 2012.

92
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

figures are associated with the simplifying assumption that a uniformly conducting region expands
instantaneously after the FDTD-computed radial electric field at 0.5Δx (and 0.5Δz) from the wire axis
exceeds 0.46 E0. Also note that the wavefront distortion of surge voltage is due to an abrupt increase
of radial current, and the attenuation at later times is associated with corona losses as well as
reduction of the equivalent characteristic impedance of the wire.

Figure 4.27 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of surge
voltage at d = 0, 350, 700, and 1050 m from the energized end of the horizontal single wire above ground
of conductivity 10 mS/m. The applied voltage is positive and Ecp = 0.5 MV/m.66

Figure 4.28 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of surge
voltage at d = 0, 350, 700, and 1050 m from the energized end of the horizontal single wire above ground
of conductivity 10 mS/m. The applied voltage is negative and Ecn = 1.5 MV/m.67

Maximum corona radii for positive voltage peaks of 1580, 1130, and 847 kV are 66, 44, and 27.5 cm,
respectively, and those for negative voltage peaks of 1670, 1200, and 901 kV are 16.5, 11, and 5.5
cm, respectively. The maximum applied voltage in the experiment [182] was 1580 kV (positive) or
1670 kV (negative), which is somewhat lower than expected voltages due to direct lightning strikes to
overhead wires. For example, 6 MV (= 30 kA × 400 Ω/2) will be generated when a 30-kA lightning

66 Reprinted with permission from T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov,
“FDTD simulation of lightning surges on overhead wires in the presence of corona discharge,” IEEE Trans. EMC, vol. 54, no. 6,
p. 1235, Figure 2a, 2012.
67 Reprinted with permission from T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov,
“FDTD simulation of lightning surges on overhead wires in the presence of corona discharge,” IEEE Trans. EMC, vol. 54, no. 6,
p. 1236, Figure 3a, 2012.

93
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

current (typical for first return strokes) is injected into an overhead transmission-line conductor having
a characteristic impedance of 400 Ω. Maximum corona radii for positive and negative voltage peaks of
6 MV are 5.1 and 0.94 m, respectively.
Figures 4.29 and 4.30 show, for positive and negative applied voltages, respectively, waveforms of
induced voltage at d = 0, 350, 700, and 1050 m on the 1.4-km-long horizontal four-conductor bundle
(located horizontally 2-m away from the energized horizontal wire and 14-m above flat ground. The

Figure 4.29 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of voltage
induced on the nearby four-conductor bundle at d = 0, 350, 700, and 1050 m, located 14-m above ground
of conductivity 10 mS/m. The applied voltage is positive and Ecp = 0.5 MV/m.68

corresponding measured waveforms [182] are also shown in these figures. The peak voltages, applied
to the nearby single wire, are 1580 kV for positive polarity (see Figure 4.29), and 1670 kV for negative
polarity (see Figure 4.30). It is clear from these figures that the computed waveforms of voltages
induced on the bundled conductor agree fairly well with the corresponding measured waveforms,
although computed waveforms of induced voltage at d = 0 are somewhat different from the
corresponding measured waveforms. The discrepancies at d = 0 m might be caused by possible
differences in setup configurations of the voltage measuring system at d = 0 m in the simulation and
experiment.

4.3.2 FDTD lightning surge simulation of an HV air-Insulated substation with back-

flashover phenomena
Using the techniques for simulating the breakdown characteristics of short-air-gap arcing horns and
transmission line surge arresters with an air gap and for taking into account the effects of AC voltages,
Tatematsu and Ueda calculated lightning overvoltages invading a 77-kV air-insulated substation in
Japan in the case of a direct lightning strike to its nearby transmission line tower [79]. Lightning-

68 Reprinted with permission from T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov,
“FDTD simulation of lightning surges on overhead wires in the presence of corona discharge,” IEEE Trans. EMC, vol. 54, no. 6,
p. 1241, Figures 11a, 11b, 11c, and 11d, 2012.

94
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

induced voltages arising from the lightning current flowing through the lightning channel and
transmission line tower were considered (which are commonly ignored in conventional circuit-theory-
based simulations) as were multi-phase back-flashover phenomena, and the effect of applied AC
voltages. A comparison of the calculated results was made with values measured by Ueda [183] for
validation.

Figure 4.30 FDTD-computed (for σcor = 40 μS/m and E0 = 2.4 MV/m) and measured waveforms of voltage
induced on the nearby four-conductor bundle at d = 0, 350, 700, and 1050 m, located 14-m above ground
of conductivity 10 mS/m. The applied voltage is negative and Ecn = 1.5 MV/m.69

Figure 4.31 shows the calculation arrangement used to simulate the 77-kV transmission line and
substation. The volume of the analysis space is 2030 m × 400 m × 3688 m. The bottom space with a
thickness of 100 m is treated as the soil, which has a conductivity of 0.0143 S/m and a relative
permittivity of 1. The conductivity of the soil is estimated by comparing the calculated grounding
resistance of the transmission line tower with the measured result, and this conductivity also agrees
with the value at the corresponding area in a ground-conductivity distribution map of Japan measured
in the past. Absorbing boundaries based on Liao’s formulation of the second order [37] are applied to
all the external surfaces of the analysis space to assume an open space. In the analysis space, five
77-kV transmission line towers, a shield wire, a double-circuit transmission line conductor, and a 77-
kV air-insulated substation are simulated.
Figure 4.32 shows a 77-kV transmission line tower model, which is composed of the main legs,
bracings, and cross arms represented by thin wires based on its blueprint in as much detail as
possible. As mentioned later, in this study, a lightning current flowing through the main leg is
calculated which requires modelling of the tower, and in particular, its bracings in detail. As shown in
Figure 4.32, each transmission line tower model has three short-air-gap flashover models (See
Section 3.1.9) and three transmission line arrester models (See Section 3.1.10). The foundation of the
transmission line tower is modeled by some cuboid conductors to simulate reinforcing bars of the
foundation. The shield wire and phase conductors, which have radii of 5.25 mm and 9.1 mm,

69 Reprinted with permission from T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov,
“FDTD simulation of lightning surges on overhead wires in the presence of corona discharge,” IEEE Trans. EMC, vol. 54, no. 6,
p. 1241, Figures 12a, 12b, 12c, and 12d, 2012.

95
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

respectively, are represented by the thin wire representation technique [44]. Both wires are connected
to the absorbing boundaries to avoid reflection at the far ends. At the near ends, the shield wire and
double-circuit phase conductors are respectively connected to the gantry of the substation and to
incoming conductors to the substation, which will be described below. To represent the effect of AC
voltages, a voltage source is inserted into each phase conductor. The output waveforms of the voltage

Shield wire
Upper-phase conductor (Phase R)
Absorbing
boundary Middle-phase conductor (Phase W)
Lower-phase conductor (Phase B)
Voltage sources to
100 m simulate the effect
of AC voltages

200 m Ground

530 m Lightning channel

3688 m

470 m Voltage source to inject

Voltage sources a lightning current
to simulate the
390 m
effect of AC Tower #5
voltages 240 m
Tower #4

100 m
Tower #3
Circuit #2
Tower #2 Circuit #1

2030 m Tower #1

77 kV substation
Ground
Conductivity : 0.0143 S/m
400 m
Relative permittivity : 1

Figure 4.31 Calculation arrangement used to simulate the 77-kV transmission line and substation.70

sources are step waves with a rise time of 0.1 µs, and the peak values of the voltage sources are
obtained using the technique described in Section 3.1.11 based on the measured results. As shown in
Figures 4.31 and 4.32, the upper, middle, and lower phases are called “phase R”, “phase W”, and
“phase B” and the double circuits are called “circuit #1” and “circuit #2”, respectively.
As shown in Figure 4.31, based on the measured lightning current flowing through the shield wire, the
striking point is assumed to be the third transmission line tower from the substation (tower #3).
Although an ideal current source is used to represent a return stroke current in some studies, in these
cases, the impedance of the current source is treated as an infinite value. As mentioned in [49], the
equivalent impedance of a lightning return-stroke channel is expected to range from 0.6 to 2.5 kΩ on
the basis of measured results, and the use of an ideal current source is not appropriate in the case of
a direct lightning strike to a grounded tall structure since it cannot take into account currents
propagating from the structure to the lightning channel. Therefore, in this study, instead of an ideal
current source, a voltage source, which was employed in other studies on a direct lightning strike to a
structure or building, is attached to the top of the transmission line tower (tower #3), and a vertical and
straight inductor-based lightning channel model is connected to the other end of the voltage source.
The inductor-based lightning channel model is represented by a series of lumped-parameter
inductances. Since the return-stroke speed was not measured, the per-unit-length inductance is set to
7.55 μH/m, which reduces the return-stroke speed to about 0.36c (c: speed of light). Note that the
return-stroke speed typically ranges from 0.33c to 0.67c in the bottom few tens of meters to 100 m of
lightning channels based on measured results [29]. Here, the return-stroke speed is calculated from
the virtual origin when the extrapolated leading edge passing through the 10% and 90% amplitude
points of the peak current at a height of 100 m from the channel bottom intersects zero in the case that
the lightning channel is attached to a perfectly conducting ground plane. Moreover, the characteristic

70Reprinted with permission from A. Tatematsu and T. Ueda, “FDTD-Based Lightning Surge Simulation of an HV Air-Insulated
Substation with Back-Flashover Phenomena,” IEEE Trans. EMC, vol. 58, vo. 5, p. 1555, Figure 13, 2016.

96
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

impedance of the lightning channel, represented by an inductance of 7.55 μH/m, is estimated to be in

the range of 0.9 to 1.5 kΩ by the following expression given by

Z l  (2.1  Ll ) / 2.1  Z 0 　 Equation 4.5

where Zl [kΩ] and Ll [μH/m] denote the characteristic impedance of the lightning channel and the per-

Circuit #2 Circuit #1

Upper phase
(Phase R)

Middle phase
(Phase W)
39.5 m
Lower phase
(Phase B)

Flashover model Transmission line

with a gap of 0.65 m surge arrester model
with a gap of 0.35 m

Current probe to obtain a

lightning current flowing
through the main leg

Figure 4.32 77-kV transmission line tower model.71

unit-length inductance, respectively [14]. Z0 is the characteristic impedance of a vertical wire without
additional inductances and is estimated to be 0.4 to 0.7 kΩ [14]. The estimated characteristic
impedance of the inductor-based lightning channel is within the range of the abovementioned
expected equivalent impedance of 0.6 to 2.5 kΩ. Since the lightning current through the lightning
channel was not measured, the output waveform of the voltage source on the top of tower #3 is
adjusted to reproduce the current flowing through the shield wire at tower #3, which was measured
close to the striking point.
In the FDTD simulations, to reduce the calculation time and the required memory capacity, the
analysis space shown in Figure 4.31 is divided into 1081 × 377 × 666 nonuniform cells. The size of the
cells around the transmission line towers, shield wire, phase conductors, and substation is set to 0.25
m, while the size of the other cells is increased gradually further away from them. As a result, the size
of the cells ranges from 0.25 to 10 m, and the time discretization is set to 0.14 ns.
Using the calculation model of the transmission line and substation, Tatematsu and Ueda calculate the
lightning overvoltages invading the substation. The lightning current is injected into the tower at a time
of 20 μs by the voltage source placed at the top of tower #3, when the voltages applied to the phase
conductors are in a quasi-steady state to simulate the effect of AC voltages. Figure 4.33 shows the
calculated and measured waveforms of the lightning current flowing through the shield wire at tower
#3 struck by the lightning, the lightning current flowing through the main leg of tower #3, and the
voltages at the surge arresters installed at the entrance of the substation. The position to obtain the
lightning current flowing through the main leg is designated in Figure 4.32. Note that there is no

71Reprinted with permission from A. Tatematsu and T. Ueda, “FDTD-Based Lightning Surge Simulation of an HV Air-Insulated
Substation with Back-Flashover Phenomena,” IEEE Trans. EMC, vol. 58, vo. 5, p. 1555, Figure 14, 2016.

97
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

measured result of the lightning voltage at the surge arrester at phase R in circuit #1. The calculated
lightning current through the channel bottom has a peak value of 40.2 kA. As shown in Figure 4.33 (a),
the lightning voltages positively deviate in the first part due to the voltages induced in the phase
conductors by the return stroke current and lightning current flowing through the transmission line
tower. The induced voltages result in back flashover at phases R, W, and B in circuit #1 and phases R
and B in circuit #2. The lightning overvoltages due to back flashover are suppressed by the surge
Circuit #2
250 250 0
Circuit #1

Voltage [kV]

Voltage [kV]

Current [kA]
Phase W (Circuit #1) Phase B (Circuit #1)
0 0 −4 Main leg

Shield wire
–250 –250 −8
20 30 40 50 20 30 40 50 20 30 40 50
Time [  s] Time [  s] Time [  s]
250 250 250
Voltage [kV]

Voltage [kV]

Voltage [kV]
Phase R (Circuit #2) Phase W (Circuit #2) Phase B (Circuit #2) Shield wire Current
0 0 0
Substation
–250 –250 –250
20 30 40 50 20 30 40 50 20 30 40 50 Current
Time [  s] Time [  s] Time [  s] Tower #3

Lightning currents through the shield wire and

Lightning voltages at the substation the main leg at tower #3 struck by lightning

(a)

250 250 250 0

Voltage [kV]

Voltage [kV]

Voltage [kV]

Current [kA]
Phase R (Circuit #1) Phase W (Circuit #1) Phase B (Circuit #1) Main leg
0 0 0 –4

Shield wire
–250 –250 –250 –8
20 30 40 50 20 30 40 50 20 30 40 50 20 30 40 50
Time [  s] Time [  s] Time [  s] Time [  s]
250 250 250
Voltage [kV]

Voltage [kV]

Voltage [kV]

Phase R (Circuit #2) Phase W (Circuit #2) Phase B (Circuit #2) Shield wire Current
0 0 0
Substation
–250 –250 –250
20 30 40 50 20 30 40 50 20 30 40 50 Current
Time [  s] Time [  s] Time [  s] Tower #3

Lightning currents through the shield wire and

Lightning voltages at the substation the main leg at tower #3 struck by lightning

(b)
Figure 4.33 Lightning voltages at the entrance of the substation and the lightning currents through the
shield wire and the main leg at the transmission line tower struck by lightning. (a) Measured results and
(b) Calculated results.72

arresters installed at the entrance of the substation. The wave tails of the overvoltages in circuit #1 are
smaller in comparison with those in circuit #2 due to the effect of the transmission line surge arresters.
Compared with circuit-theory-based simulation techniques, it is more straightforward to consider
lightning-induced effects in full-wave numerical approaches such as the FDTD method. In
conventional circuit-theory simulation techniques, induced voltages are obtained by the following two
steps. In the first step, conductor lines are ignored, and lightning electromagnetic fields are calculated
using analytical expressions or simplified full-wave numerical approaches. In the second step, the
voltages induced in the conductor lines are obtained by applying field-to-transmission line coupling
techniques, where the effect of the lightning electromagnetic fields is incorporated into Telegrapher’s
equations. Unlike circuit-theory-based simulation techniques, full-wave numerical approaches can
handle conductor lines and lightning channels at the same time and calculate the voltages induced in
the conductor lines in a step. There are several papers on the application of the FDTD method to the
calculation of lightning-induced voltages in a distribution line as shown in Section 4.4.1. In the FDTD
method, the effect of the lightning electromagnetic fields can be considered simply by installing
lightning channel models in the analysis space. As shown in Figure 4.33 (b), the voltages induced in
the phase conductors by the return stroke current and the current flowing through the transmission line
tower are represented in the FDTD-based surge simulations, and the following back-flashover
phenomena are reproduced by the short-air-gap flashover models and transmission line surge arrester
models. The phases of the back flashover in the FDTD method are identical to those of the measured
results. In addition, similar to the measured results, the lightning overvoltages invading the substation

72Reprinted with permission from A. Tatematsu and T. Ueda, “FDTD-Based Lightning Surge Simulation of an HV Air-Insulated
Substation with Back-Flashover Phenomena,” IEEE Trans. EMC, vol. 58, vo. 5, p. 1557, Figure 16, 2016.

98
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

are suppressed by the surge arrester models installed at the entrance of the substation and the
overvoltages in circuit #1 have the smaller wave tails compared with those in circuit #2. Note that it is
presented in [183] that the measured results with back-flashover phenomena cannot be well
reproduced without the lightning-induced effects in conventional EMTP simulations.
In the abovementioned results, to calculate the fraction of the lightning current that flows through the
main leg of the 77-kV transmission line tower model, and to compare the calculated result with the
measured current, a transmission line tower is modeled in detail to take into account the configuration
of the bracings. Note that Tatematsu and Ueda confirmed that lightning surge phenomena similar to
those presented in Figure 4.33 can be obtained by employing a simplified transmission line tower
model and dividing the analysis space in Figure 4.31 into 791 × 233 × 579 nonuniform cells [79]. The
size of these cells ranges from 0.5 to 10 m, and thus the calculation time and the required memory
capacity can be significantly reduced in comparison to those in the above results.

4.3.3 FDTD simulation considering AC operating voltage for an air-insulated

substation
In [185], J. Takami et al. calculated the lightning surge at the open end of the circuit breaker of a 500-
kV air-insulated substation under the standard lightning stroke condition using the FDTD method. The
lightning overvoltage value at the time of back-flashover is analyzed, particularly considering the
superimposition of the AC operating voltage of the power line. Conversely, circuit analysis of the
lightning overvoltage value under the same condition is also performed as a means of comparison.
The subjects of analysis are the vertically suspended double-circuit lines in a 500-kV system, which
are typically used in Japan and the air-insulated substation. Circuit analysis using EMTP has normally
been used for lightning protection design, the main conditions of which are listed in Table 4.1.

Table 4.1 Conditions of lightning overvoltage analysis using EMTP.73

Lightning stroke position The first tower
lightning stroke

Aspect of surge intrusion Upper phase back-flashover

Aspects of

150 kA (500 kV power line)

Lightning stroke current waveform
1 / 70μs ramp wave
Lightning channel impedance 400 Ω
Alternating current phase The superimposed AC voltage is considered.
Flashover Flashover model based on the leader model
Arcing horn Horn spacing 3.3 m
Transmission line

8 phase J.Marti, 2 overhead ground wires,

Ground wire: TASR 150mm2 x 1 wire,
Transmission line
power line 3 phases x 2 lines,
Power line: TACSR 610 mm2 x 6 wires
Multi-story transmission tower model (The constant is
Transmission tower determined based on the actual measurement), grounding
resistance 10 Ω
Anchor structure 2-level model, grounding resistance 1 Ω
Circuit configuration Circuit breaker at the incoming line entrance open
Single-phase distributed constant line, lossless route
Substation
equipment

Air bus bar

Surge impedance 350 Ω
Simulation of π-type capacitance
Open circuit breaker
(Equivalent ground capacitance 366 pF)
Bushing Simulation of capacitance 200 pF
Surge arrester Simulation of nonlinear resistance, V10kA = 1,098 kV

The FDTD-based surge simulation code Virtual Surge Test Lab. Restructured and Extended Version
(VSTL REV) [186] is used for the FDTD analysis. The analysis model of the incoming line to the
substation is shown in Figure 4.34, and the input image of the FDTD analysis model in Figure 4.35.
The arrangement of the double-circuit 6-phase power lines is changed from a vertical arrangement to
a horizontal arrangement between the first transmission tower and gantry, and there are 3 overhead
ground wires between the first tower and gantry. The transmission tower is simulated up to the second

73 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 807, Table
1, 2015.

99
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

tower, and the transmission line is connected to the absorbing boundary at a position 150 m from the
second tower. All boundary surfaces in the analysis space are simulated to be Liao’s second-order
absorbing boundary surfaces [37]. The lightning return-stroke channel is represented by a vertical
conductor of radius 0.23 m. A lumped source having an internal resistance of 400 Ω is inserted
between the bottom of the channel-representing vertical conductor and the tower top.

30000

14000 8000
4000
36000 22000

4000
14000

24000
81000

4000
15000

6000
24000
45000
15000

2000
33000
45000

27000
7000 7000 7000 7000 7000 7000 7000 7000
19000

Middle phase

Middle phase
Lower phase

Lower phase
Upper phase

Upper phase
15000

(a) (b)

81000
33000
19000

45000 150000 350000 150000

(c)

Figure 4.34 Incoming line model of a 500-kV air-insulated substation. (a) Tower, (b) Gantry, and (c) Overall
view.74

74 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 808,
Figure 1, 2015.

100
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.35 FDTD simulation model of a 500-kV air-insulated substation.75

The equipment layout inside the air-insulated substation is illustrated in Figure 4.36. The capacitance
to ground for representing the bushing (200 pF) is arranged directly under the gantry. The open circuit
breaker was simulated with equivalent ground capacitance (366 pF). The lumped element model
attached to the FDTD-based surge simulation code is used for the flashover of the surge arrester and

35m 10m

Incoming
引き込み線
line
Circuit
遮断器
breaker Lightning
避雷器 ブッシング
(開放) arrester
(Open) Bushing

Figure 4.36 Equipment layout of a 500-kV air-insulated substation model.76

arcing horn, and the individual nonlinear characteristics are considered in the analysis. The discharge
voltage level contained in the V-I characteristics of the surge arrester is 1,098 kV_10 kA. Since the
computational resource (memory used and computation time) is one of the issues for FDTD analysis,
the computational conditions and computation time are listed in Table 4.2.

Table 4.2 Conditions of lightning overvoltage analysis using FDTD.77

Calculation program FDTD-based surge simulation code VSTL REV [186]
Space size 500 × 750 × 650 m
Nonuniform cells: 1 m around the transmission line route
Space cell
(height: 150 m, width: 100 m), 10 m elsewhere
Divergence preventing coefficient α = 0.3 (∆t = 1.35 ns)
Result output frequency ∆t = 5.39 ns
20 μs: Approx. 1 hour, 100 μs: Approx. 5 hours
Computation time
(Corei7-3280 3.6 GHz，Memory 64 GB)
Spatial boundary surface Liao’s second-order absorbing boundary

In the case of analyzing a lightning overvoltage with back-flashover, there are cases where the
overvoltage value increases if the AC operation voltage is superimposed on the power line side with
polarity opposite to the lightning stroke current. Hence, this is taken into consideration in this study. To
simulate the AC operation voltage, a voltage source is connected to each phase of the power line to

75 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 808,
Figure 2, 2015.
76 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 808,
Figure 3, 2015.
77 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 808, Table
2, 2015.

101
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

ensure that the voltage converges to a value equivalent to the peak AC voltage value. The circuit
diagram of the superimposed ac model is illustrated in Figure 4.37. The voltage source generating the
AC peak value voltage is installed within the span of the second tower and the absorbing boundary
surface. Here, the superimposed AC voltage can be regarded as DC bias in a lightning overvoltage
analysis whose time domain is several tens of μs. In order to suppress oscillation caused by changes
of output voltage, a stepped wave with a front duration of 1 μs is adopted. The internal impedance of
the voltage source to simulate the superimposed AC voltage is set to 1 MΩ so that the presence of
vertical conductors and the sources would not influence the surge phenomena. If voltage is applied to
a certain phase, an induced voltage is generated on other phases, so the peak value of each voltage
source must be determined considering the induced voltage for applying the prescribed voltage. In this
study, the method devised by Tatematsu and Ueda [79] was used.
In the FDTD analysis here, the lightning stroke model was composed of a single current source and a
lightning channel represented by a thin wire conductor. Analysis is performed with lightning stroke
currents of 150 kA and 200 kA under a condition with superimposed AC voltage. In addition, the
analysis is performed considering the lightning stroke at the first or second tower. Note that the peak
value of the opposite polarity of the lightning current is set for the superimposed AC voltage.
The surge voltage waveform of the circuit breaker terminal when the lightning strikes the first tower
computed with the FDTD method is indicated in Figure 4.38. Since no flashover occurs under
conditions of the lightning strike to the first tower with a lightning stroke current of 150 kA, which is the
standard for lightning protection design, the waveform is not drawn in Figure 4.38a. A flashover occurs
under the condition of a lightning stroke current of 200 kA, but since the overvoltage at the circuit
breaker terminal is less than the discharge voltage of the surge arrester until around 4 μs, voltage
builds up due to repeated reflections of the lightning surge, and peaks at 1,371 kV. Note that the part
Voltage Transmission line
source V2W V1B

Thin
wire
V2R V1R
Absorbing
No.1 tower No.2 tower boundary 2L 1L

Transmission line Voltage

source
Circuit Gantry
Thin V2B V1W
breaker Bushing
(Open) wire

45m 150m 350m 100m Earth

50m

(a) (b)
Figure 4.37 Equipment layout of a 500-kV air-insulated substation model. (a) Side view and (b) Layout of
voltage source.78

which shows negative polarity at the beginning of the waveform represents the state of the developing
leader in the arcing horn of the tower. This waveform is due to electromagnetic induction from the
lightning current which flows into the tower adding to the lightning current which flows in the overhead
ground wire.
The surge voltage waveform of the circuit breaker terminal at the time of the lightning strike to the
second tower is indicated in Figure 4.38b. Although a flashover occurs with the lightning stroke current
of 150 kA, since it occurs at the wave tail, the overvoltage waveform of the substation rises relatively
slowly, and is limited almost to within the discharge voltage of the surge arrester. The maximum
overvoltage value that appeared after repeated surge reflections was 1,366 kV. With the lightning
stroke current of 200 kA, the initial rise is steep compared to the time of the 150-kA lightning stroke
and an overvoltage which exceeds the discharge voltage of the surge arrester is generated at the
beginning of the waveform. The maximum value was 1,461 kV. The oscillation of approximately 1 μs

78 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 809,
Figure 4, 2015.

102
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

period, which appears on both waveforms of the first and second towers, is generated by the round-
trip reflections of surge between the surge arrester and circuit breaker.
For comparison, the surge voltage waveforms acquired by circuit analysis are shown in Figures 4.39.
Under the condition of a lightning stroke current of 150 kA, flashover occurs at the wavefront with the
lightning on the first and second towers, and the initial rise in lightning overvoltage is steep. The
lightning overvoltage peaks in the first wave of the surge waveform, regardless of which tower is hit by
lightning. The value is 1,999 kV when lightning strikes the first tower, and 2,041 kV when it strikes the
second; hence the values are almost the same. As summarized in Table 4.3, the overvoltage value
after their waveform evaluation is below 1,800 kV, and lightning impulse withstand voltage (LIWV) of
1,800 kV is considered reasonable. After the second wave, it oscillates, centering on the residual
voltage level of the surge arrester. The oscillation has a period of approximately 1 μs, which appears
on both waveforms, and is the same as the FDTD analysis result described above. With the lightning
strike to the second tower, an oscillation of approximately 3 μs is superimposed on the oscillation of
approximately 1 μs. It corresponds to the surge propagation period between the second tower, which
is the point of the lightning stroke, and the surge arrester of the substation. In addition, since the
transmission tower and transmission line models are independent in the circuit analysis, the effect of
electromagnetic induction by lightning current that shows negative polarity pending completion of a
flashover in the FDTD analysis is not considered.

<FDTD> Voltage at the CB terminal <FDTD> Voltage at the CB terminal

3000 3000
200kA (Max:1371kV) 200kA (Max:1461)
200 kA (Max: 1461 kV) 150kA (Max:1366)
2000 2000
200 kA (Max: 1371 kV) 150 kA (Max: 1366 kV)
V [kV]
V [kV]

1000 1000

0 0

-1000 -1000

-2000 -2000
0 5 10 15 20 0 5 10 15 20
Time [μs] Time [μs]

(a) (b)
Figure 4.38 Voltage waveforms at the circuit breaker terminal acquired by FDTD analysis. (a) Waveform
when the lightning stroke hits the first tower and (b) Waveform when the lightning stroke hits the second
tower.79

<EMTP> Voltage at the CB terminal <EMTP> Voltage at the CB terminal

3000 3000
200 kA (Max: 2090 kV) 200kA (Max:2090kv) 200 kA (Max: 2186 kV) 200kA (Max:2186kv)
150kA (Max:1999kv) 150kA (Max:2041kv)
2000 2000
150 kA (Max: 1999 kV) 150 kA (Max: 2041 kV)
1000 1000
V [kV]

V [kV]

0 0

-1000 -1000

-2000 -2000
0 5 10 15 20 0 5 10 15 20
Time [μs] Time [μs]

(a) (b)

79 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 812,
Figure 9, 2015.

103
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.39 Voltage waveforms at the circuit breaker terminal acquired by FDTD analysis. (a) Waveform
when the lightning stroke hits the first tower and (b) Waveform when the lightning stroke hits the second
tower.80

Table 4.3 Lightning overvoltage values for a circuit breaker based on FDTD and EMTP analyses. 81
Lightning strike to the first tower 150 kA 200 kA
Overvoltage peak value NFO 1,371 kV
FDTD
Waveform evaluation － 1,205 kV
Overvoltage peak value 1,999 kV 2,090 kV
EMTP
Waveform evaluation 1,654 kV 1,735 kV

Lightning strike to the second tower 150 kA 200 kA

Overvoltage peak value 1,366 kV 1,461 kV
FDTD
Waveform evaluation 1,185 kV 1,337 kV
Overvoltage peak value 2,041 kV 2,186 kV
EMTP
Waveform evaluation 1,747 kV 1,865 kV

4.3.4 HEM simulation of lightning surges on overhead transmission lines and towers
4.3.4.1 Introduction
The lightning performance of high voltage transmission lines in terms of backflashover is governed by
the balance between the resulting overvoltage across line insulators and the capability of such
insulators to withstand it. In this context, the transient behavior of tower-footing grounding has a
relevant influence on the amplitude of lightning overvoltage, and, consequently, on the lightning
performance of transmission lines.
This section presents results of the application of the hybrid electromagnetic model (HEM) [21], [139],
[140] to simulate transmission line response to direct lightning strikes in terms of overvoltages across
line insulator strings. The analysis includes the assessment of the frequency-dependent effect of soil
parameters [155] and of the use of concise representations of tower-footing electrodes [187] on such
overvoltages taking a 138-kV transmission line as example. The lightning performance of transmission
lines in terms of backflashover occurrence is also estimated.

4.3.4.2 Methodology
4.3.4.2.1 Simulated transmission line
A self-sustained and 400-m span transmission line with voltage level of 138 kV was considered in the
simulations. The associated tower configuration is depicted in Figure 4.40. It has an average height of
30 m and supports a single shield wire (0.4-cm radius) and 3 phase conductors (1.13-cm radius)
following a non-uniform arrangement. In simulations, the main metallic parts of the tower were
represented observing tower geometry, including cross arms.

80 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 812,
Figure 10, 2015.
81 Reprinted with permission from J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level,” IEEE Trans. DEI, vol. 22, no.2, p. 813, Table
4, 2015.

104
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

400 m 2.9 m
3.03 m
400 m 1.86 m
1.65 m 3.72 m
0.8 m

30 m
23.25 m
6.0 m

(a) (b)
Figure 4.40 Representation of the simulated event. (a) 138-kV Transmission line and (b) Tower
configuration. Adapted from [153].

4.3.4.2.2 Representation of tower-footing grounding

Figure 4.41 illustrates the simulated grounding electrode arrangements. It consists of counterpoise
wires of 0.6-cm radius, buried 0.5 m deep in the soil, connected to four vertical 2.5-m long rods of 5-
cm radius, which represent the metallic parts of the tower that are buried in the soil. The low-frequency
soil resistivity was varied from 100 to 4000 Ωm. For each soil condition, the length L of counterpoise
wires were varied as indicated in Table 4.4.

L L

6m 20 m
6m
6m
6m

(a) (b)
Figure 4.41 Simulated grounding electrode arrangements. (a) Configuration applied for 100 and 300 Ωm
soil resistivity and (b) Configuration applied for 600, 1000, 2000, and 4000 Ωm soil resistivity. Adapted
from [153].

Table 4.4 Simulated electrode length L as function of low-frequency soil resistivity ρ0. The low frequency
grounding resistance (Rg) for each case is also included. 82
ρ0 (Ωm) 100 300 600 1000 2000 4000
L (m) 5 10 30 50 70 110
Rq 5 11 11 12 19 26

Two assumptions were explored: constant parameters of soil (ρ=ρ0; εr=10) and those given by the so-
called Visacro-Alipio expressions [152] defined as follows.
   0 {1  [1.2  106   0 0.73 ][( f  100)0.65 ]}1 Equation 4.6
 r  7.6 103 f 0.4  1.3 Equation 4.7

where ρ, ρ0 and εr stand for the soil resistivity at frequency f in Hz, the soil resistivity at 100 Hz, and the
relative permittivity at frequency f in Hz, respectively. Such expressions are valid on the 100-Hz to 4-
MHz frequency range, as denoted in [152].

4.3.4.2.3 Return stroke current waveform

Figure 4.42 shows the return stroke current representations adopted in simulations to reproduce
typical lightning current waveforms of first and subsequent strokes [188]. Such current waveforms are
assumed to be impressed at the top of the tower illustrated in Figure 4.40.

82 Reprinted with permission from S. Visacro, F.H. Silveira, "The Impact of the Frequency Dependence of Soil Parameters on
the Lightning Performance of Transmission Lines," IEEE Trans. EMC, vol. 57, no.3, p. 436, Figure 5c, 2015.

105
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

14
35
12

Current (kA)
Current (kA) 30
25 10
20 8
15 6
10 4
5 2
0 0
0 10 20 30 40 0 2 4 6 8 10
Time (s) Time (s)
(a) (b)
Figure 4.42 Representative current waveforms of first (a) and subsequent and (b) strokes considering
median parameters measured at Mount San Salvatore station [189].83

4.3.4.2.4 Estimation of critical current and probability of backflashover occurrence

Based on the resulting overvoltages across line insulator strings, the disruptive effect (DE) model [190]
was applied to determine the critical peak current that lead the 138-kV line to backflashover,
considering a 650-kV line Critical Flashover Overvoltage (CFO). Taking the IEEE cumulative peak-
current distribution of first-strokes as reference [191], the percentage of peak currents exceeding such
a critical value and the backflashover probability of occurrence are estimated.

4.3.4.3 Effect of frequency dependence of soil parameters

4.3.4.3.1 Simulated overvoltages
Figure 4.43 illustrates voltages developed across the upper insulator string of the simulated 138-kV
line due to a direct strike to the tower top considering constant and frequency-dependent soil
parameters for low-frequency soil resistivity values varying from 100 to 4000 Ωm.

500 580
= 0,r = 10 = 0,r = 10
400 ,
480 ,
Voltage (kV)
Voltage (kV)

380
300 280
200 180
80
100
-20
0 -120
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

(a) (b)

600 600
= 0,r = 10 = 0,r = 10
500 500
, ,
Voltage (kV)

Voltage (kV)

400 400
300
300
200
200 100
100 0
0 -100
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

(c) (d)

83 Reprinted with permission from S. Visacro, F.H. Silveira, "The Impact of the Frequency Dependence of Soil Parameters on
the Lightning Performance of Transmission Lines," IEEE Trans. EMC, vol. 57, no.3, p. 435, Figure 3, 2015.

106
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

600 700
= 0,r = 10 600 = 0,r = 10
500 ,

Voltage (kV)
,

Voltage (kV)
500
400 400
300 300
200
200
100
100 0
0 -100
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)
(e) (f)

800 650
= 0,r = 10 = 0,r = 10
700 550
, ,

Voltage (kV)
Voltage (kV)

600 450
500 350
400
250
300
200 150
100 50
0 -50
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

(g) (h)
Figure 4.43 Simulated overvoltage across upper insulator string of the 138-kV line for constant and
frequency-dependent soil parameters for different values of soil resistivity ρ0. (Left column: First stroke;
Right column: Subsequent stroke). (a),(b) 100 Ω.m, (c),(d) 600 Ω.m, (e),(f) 1000 Ω.m, and (g),(h) 4000 Ω.m.

The results show that the effect of frequency dependence of soil parameters does not significantly affect
the amplitude of overvoltages across line insulators for soils of resistivity up to 600 Ωm. The influence
of this effect becomes significant for increasing values of soil resistivity, causing peak overvoltage to
diminish. To complement the analysis, Table 4.5 indicates the influence of the frequency-dependent
effect in terms of peak overvoltage.

Table 4.5 Peak overvoltages across upper insulator string.

Overvoltage (kV)
ρ0 L First stroke current Subsequent stroke current
(Ω.m) (m) ρ=ρ0, ρ=ρ(ω), ρ=ρ0, ρ=ρ(ω),
Variation % Variation %
εr=10 ε(ω) εr=10 ε(ω)
100 5 457.7 455.0 -0.6 589.3 588.2 -0.2
600 30 533.0 504.9 -5.3 604.6 595.8 -1.5
1000 50 564.5 519.3 -8.0 608.9 597.3 -1.9
4000 110 727.3 567.2 -22.0 615.0 595.4 -3.2

As can be noted, the reduction promoted by the frequency-dependent effect on overvoltage related to
first stroke current is more significant, from around 0.6% to 22% for soils with ρ0 varying from 100 to
4000 Ωm against reductions from around 0.2% to 3% of subsequent-stroke overvoltages [155].

4.3.4.3.2 Estimating the probability of backflashover occurrence

To quantify the frequency-dependent effect of soil parameters on the probability of backflashover
occurrence, the DE model was applied to the resulting overvoltages across line insulator strings
obtained for each soil-resistivity condition to determine the critical peak current Ic under the
assumption of constant and frequency-dependent soil parameters.
Table 4.6 indicates the calculated critical currents along with the percentage of peak currents
exceeding the critical value taking the IEEE cumulative peak-current distribution of first-strokes [191].

107
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Table 4.6 Influence of frequency dependence of soil parameters on critical first-stroke peak currents.84
138-kV line
ρ0 ρ = ρ0, εr = 10 ρ(ω), ε(ω)
(Ω.m)
IP > IPC IP > IPC Δ* (%)
Ip (kA) Ip (kA)
A(%) B(%)
100 121 2.8 122 2.8 3
300 85 6.8 91 5.8 15
600 86 6.6 93 5.5 16
1000 78 8.4 87 6.4 24
2000 58 16.2 69 10.9 33
4000 47 25.8 61 14.8 43
* Relative decrease of the percentage of critical currents due to the frequency dependence of soil parameters.

In all cases, the amplitude of critical currents is increased when considering the frequency-dependent
effect of soil parameters, and, accordingly, the percentage of first-stroke peak-currents exceeding Ic is
decreased. Such behavior is more significant with increasing soil resistivity. The impact of the effect is
significant even for low resistivity soils (decrease of 7% in the percentage of currents leading to
backflashover for 300 Ωm soils). Decreases from about 24% to 43% are achieved for values of ρ0
varying from 1000 to 4000 Ωm. Similar behavior is shown for a 230-kV line, as demonstrated in [155].
To complement the analysis, Figure 4.44 depicts the aforementioned behavior in terms of critical
currents for 230- and 138-kV lines [155]. A significant increase of critical currents due to the frequency
dependence of soil parameters is shown. The relative increase of Ic is approximately the same for the
230- and 138-kV lines. Such results suggest that frequency dependence of soil parameters cause
significant decrease of backflashover rates, which becomes more pronounced with increasing soil
resistivity.

250
= ,
200 0 ,r = 10

150 230 kV
Ic (kA)

100

50
138 kV
0
0 1000 2000 3000 4000
 (W.m)

Figure 4.44 First-stroke critical currents Ic of 230- and 138-kV lines as a function of soil resistivity ρ under
the assumption of constant and frequency dependent soil parameters.85

4.3.4.3.3 Assessing the impact on the outage rates of lines

A study of the impact of frequency dependence of soil parameters in terms of the resulting
backflashover outage rate is conducted in [155] considering six scenarios in terms of the distribution of
soil resistivity along the line route. The analysis considering three of these scenarios (related to non-
uniform distribution of soil resistivity) is reproduced herein according to Table 4.7.

84 Reprinted with permission from S. Visacro, F.H. Silveira, "The Impact of the Frequency Dependence of Soil Parameters on
the Lightning Performance of Transmission Lines," IEEE Trans. EMC, vol. 57, no.3, p. 439, Table 3, 2015.
85 Reprinted with permission from S. Visacro, F.H. Silveira, "The Impact of the Frequency Dependence of Soil Parameters on
the Lightning Performance of Transmission Lines," IEEE Trans. EMC, vol. 57, no.3, p. 438, Figure 8, 2015.

108
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Table 4.7 Scenarios for soil resistivity distribution.86

Hypotheses for the distribution of ρ (%)
100 Ω.m 300 Ω.m 600 Ω.m 1000 Ω.m 2000 Ω.m 4000 Ω.m
15 25 30 30 0 0
10 10 30 30 10 10
0 0 30 30 25 15

(i) only moderate and low resistivity soils (no samples of 2000 and 4000 Ωm included)
(ii) moderate resistivity (samples of all values of resistivity included, larger frequency of 600 and 1000
Ωm soils)
(iii) moderate and high resistivity soils (no samples of 100 and 300 Ωm included)
The analyses assumed a number of 30 strikes per 100 km per year to the 138-kV. Table 4.8
summarizes the results.

Table 4.8 Impact of frequency dependence of soil parameters on the expected outage rate of a 138-kV
line.87
Outages/100 km/year
Hypotheses for the distribution of ρ (%)
(138-kV line) Δ
100 300 600 1000 2000 4000 (ρ=ρ0, (%)
ρ(ω), ε(ω)
(Ω.m) (Ω.m) (Ω.m) (Ω.m) (Ω.m) (Ω.m) εr=10)
15 25 30 30 0 0 2.0 1.6 18
10 10 30 30 10 10 2.9 2.1 28
0 0 30 30 25 15 3.7 2.6 31

The estimated outage rates (per 100 km per year) under the assumption of constant soil parameters
range from 2 to 3.7, rising with increasing soil resistivity. When considering the frequency-dependent
effect of soil parameters, the outage rate ranges from 1.6 to 2.6. These results denote the significant
improvement of the lightning performance of transmission lines due to the frequency dependence of
soil parameters, with the exception of lines installed on very low-resistivity soils. The impact of this
effect is even more relevant with increasing soil resistivity.

4.3.4.4 Effect of concise representations of tower-footing electrodes

4.3.4.4.1 Introduction
The characteristics of the grounding system are a fundamental aspect to be considered in the
evaluation of backflashover occurrence in high voltage transmisison lines subjected to direct lightning
strikes.
An adequate configuration of tower-footing grounding electrodes can guarantee the reduction of
overvoltage across insulator strings, diminishing the probability of backflashover occurrence.
In computational evaluations of the lightning performance of transmission lines, the representation of
tower grounding may influence the obtained results. The application of electromagnetic models such
as the HEM and the FDTD method allows the physical representation of grounding electrodes.
However, when using analytical or circuit-based tools, the physical representation of grounding
electrodes is not possible and simplified representations are needed, the most common being the use
of low-frequency grounding resistance. Despite the practical interest in using such simplified
representations, their adoption requires assessing the impact of their use in the calculated lightning
performance in relation to that obtained under the physical representation of the electrodes.

86 Reprinted with permission from S. Visacro, F.H. Silveira, "The Impact of the Frequency Dependence of Soil Parameters on
the Lightning Performance of Transmission Lines," IEEE Trans. EMC, vol. 57, no.3, p. 439, Table 4, 2015.
87 Reprinted with permission from S. Visacro, F.H. Silveira, "The Impact of the Frequency Dependence of Soil Parameters on
the Lightning Performance of Transmission Lines," IEEE Trans. EMC, vol. 57, no.3, p. 440, Table 7, 2015.

109
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

In [187], a detailed investigation of the effect of concise representations of grounding is presented by

means of systematic simulations using the HEM model. The effect of representing grounding by its low
frequency grounding resistance RLF or by its impulse impedance Zp in comparion to the response
provided by the physical representation of grounding electrodes in terms of backflashover occurrence
is assessed. The main results are presented as follows.

4.3.4.4.2 Response of grounding electrodes subjected to lightning currents

GPR is the main response of grounding electrodes subject to impressed currents. An in-depth
discussion on the main physical aspects related to the behavior of grounding response for fast time-
varying currents, such as those of lightning return strokes and low-frequency currents are presented in
[192], [193].
Figure 4.45 illustrates the resulting GPR due to a typical first return stroke lightning current impressed
to an arrangement of electrode consisting of 70-m-long counterpoise wires buried 0.5 m deep in a
2000-Ωm ground resistivity soil. Concise representations of the response of grounding electrodes,
namely the low-frequency resistance RLF, the impulse resistance RIP, and the impulse impedance ZP
are also shown. Such quantities are defined as follows:
RLF: Ratio of instantaneous values v(t)/i(t) at the tail of the waves;
RIP: Ratio of instantaneous values v(t)/i(t) at the time of the peak current;
Zp: Ratio of peaks of the GPR and impressed current, ZP=VP/IP;
As can be seen, the waveforms of GPR and impressed current are very similar, but the peak values of
GPR and current are not simultaneous. It is justified by the reactive and propagation effects in the soil.
For lightning protection applications, the impulse impedance Zp is a very useful parameter since it
allows determining the maximum GPR as function of peak current. Also, as demonstrated in [195],
once defining a representative waveform of first-stroke current to be applied on grounding, the
resulting impulse impedance does not change in the range of typical front times of first strokes.
Figure 4.46 compares the effect of three different representations of tower-footing grounding in terms
of GPR for the same conditions of Figure 4.45: physical representation of electrodes, and concise
representation by RLF and ZP.

500 35
V(t_Ip) 30
400
Current (kA)
Voltage (kV)

v(t) 25
300 20
Ip Vp
200 15
i(t)
RLF=v(t)/i(t) 10
100 ZP=Vp/Ip
5
RIp=V(t_Ip)/Ip
0 0
0 5 10 15 20 25 30 35 40
Time (s)
Figure 4.45 Response of counterpoise wires 70-m long buried 0.5 m deep in a 2000-Ωm soil subjected to a
representative first-stroke lightning current, along with their concise representations ZP and RLF.
Calculated ZP, RLF and RIP: 14.2 Ω, 18.9 Ω, and 13.7 Ω. Adapted from [194].

110
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method
GPR of the tower-footing electrodes
subjected to the median FST current

700
V
600
I*Rg

Voltage (kV)
500
400
I*Zp
300
200
100
0
0 5 10 15 20 25 30 35 40
Time (s)
Figure 4.46 GPR of the tower-footing electrodes subjected to the median first-stroke current.
Counterpoise wires 70-m long buried 0.5 m deep in a 2000-Ωm soil. Adapted from [194].

The response of the counterpoise wires is approximately governed by ZP in the first µs (at the
wavefront). At the wavetail such response becomes governed by RLF. It is important to note that the
use of RLF as a concise representation of grounding leads to much higher values of GPR.

4.3.4.4.3 Overvoltage across transmission line insulator strings

Figure 4.47 illustrates the resulting overvoltage across the upper insulator of a 138-kV line as function
of the three different representations of tower-footing grounding considered in Figure 4.46.

700
600
Voltage (kV)

500
400
300 RLF

200 Zp
100
0
0 10 20 30
Time (s)
Figure 4.47 Overvoltage across insulator string of a 138-kV line subjected to the median first-stroke
current. Counterpoise wires 70-m long buried 0.5 m deep in a 2000-Ωm soil. Adapted from [194].

The wavefront of the overvoltage calculated under the physical representation of the counterpoise
wires is very close to that obtained under their representation as ZP. For these representations, the
overvoltage peak values are nearly the same (difference about 3%). Using RLF representation leads to
much higher overvoltage amplitude (including peak value) in relation to those obtained under physical
representation of grounding electrodes.

4.3.4.4.4 Probability of backflashover occurrence

Figure 4.48 presents a comparison of the lightning performance (in terms of the percentage of
currents exceeding the critical current in a first-return-stroke peak-current distribution [191] of a 138-kV
tower for several soil resistivity values as function of the representations of tower-footing grounding. In
the figure, CW, ZP, and RLF stand for counterpoise wire (physical representation), impulse impedance,
and low-frequency ground resistance, respectively. The simulations considered different counterpoise
lengths and soils of resistivity varying from 100 to 4000 Ωm, as used in [187]. The results were
obtained from the application of the DE model on the overvoltages across insulator strings of the line
calculated by the HEM model. The percentage of currents exceeding the critical current means the
probability of backflashover occurrence.

111
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4 10
CW CW
Zp Zp
8
3 RLF RLF

% I > Icr
% I > Icr
6
2
4
1 2

0 0
5 10 15 20 25 5 10 15 20 25 30 35 40 45
Length (m) Length (m)

(a) (b)
18 30
CW CW
16 Zp 25 Zp
14 RLF RLF
12 20
% I > Icr

% I > Icr
10
15
8
6 10
4
5
2
0 0
10 20 30 40 50 60 70 20 30 40 50 60 70 80 90
Length (m) Length (m)

(c) (d)
45 70
CW CW
40 Zp 60 Zp
35 RLF RLF
30 50
% I > Icr

% I > Icr
25 40
20 30
15 20
10
5 10
0 0
40 50 60 70 80 90 100 110 120 70 90 110 130 150 170
Length (m) Length (m)

(e) (f)
Figure 4.48 Probability of backflashover occurrence as a function of grounding representation: physical
representation of electrodes and concise representations by RLF and ZP .(a) 100 Ωm, (b) 300 Ωm, (c) 600
Ωm, (d) 1000 Ωm, (e) 2000 Ωm, and (f) 4000 Ωm.88

The results show that the lightning performance obtained under ZP representation practically match
those obtained under the physical representation of tower-footing electrodes. Conversely, the results
obtained under RLF representation are very conservative, with a much higher probability of
backflashover occurrence (> 20% to 50% for soils of 300 Ωm and above).
As remarked in [187], the quality of these results holds true for other transmission line voltage levels,
the presence of adjacent towers and distinct configurations of grounding electrodes.
In summary, using a simple real number, corresponding to the impulse grounding impedance Zp, to
represent the tower-footing electrodes is enough to guarantee the correct assessment of the lightning
performance of transmision lines in terms of backflashover occurrence.

4.4 Overhead distribution lines

4.4.1 FDTD simulation of lightning-induced surges
4.4.1.1 Lightning-induced voltages on a three-phase distribution line with a
multipoint-grounded shielding wire and surge arresters
4.4.1.1.1 Introduction
When a lightning strike occurs near a distribution line, the electromagnetic field generated by the
return-stroke current may induce overvoltages on the distribution line (e.g., [196], [197]). To suppress
such overvoltages, lightning arresters and/or shielding wire are installed on a distribution line, and the
effectiveness of these countermeasures is usually studied by simulations. For accurate simulation
results, taking into account the lossy ground is one of the most important factors.
Traditionally, the lightning-induced voltages are calculated based on distributed-parameter circuit
theory using the following two steps (e.g., [198]). In the first step, the incident electromagnetic fields
radiated by the return-stroke current are calculated in the absence of the distribution line. Then, In the
second step, the telegrapher’s equations of the distribution line with embedded electromagnetic

88Reprinted with permission from S. Visacro, F.H. Silveira, "Lightning Performance of Transmission Lines: Requirements of
Tower-Footing Electrodes Consisting of Long Counterpoise Wires," IEEE Trans. EMC, vol. 31, no.4, p. 1530, Figure 13, 2016.

112
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

excitation sources obtained at the first step are solved to find lightning-induced voltages. In this
second step, use is made of the so-called field-to-transmission coupling models [3]-[5].
Recently, numerical electromagnetic-field calculation methods have been applied to the calculation of
lightning-induced voltages. The application of these methods, in which Maxwell’s equations are
numerically solved, is quite straightforward compared with the field-to-transmission coupling technique
mentioned above, since the return-stroke path and the distribution line can be modeled simultaneously
in the analysis space. Moreover, these numerical electromagnetic field calculation methods have the
advantages of directly modeling non-straight lightning channels, three-dimensional structures,
grounding systems, and non-horizontal wires. Among these numerical electromagnetic-field
calculation methods, the FDTD method is known to require a longer computation time and a larger
memory capacity, but these requirements are no longer a big issue due to recent progress of high-
performance computers, in particular, the development of the GPGPU (general-purpose computing on
graphics processing units). In addition, the FDTD method is advantageous, since it can take into
account inhomogeneous electrical parameters of the ground, the finite conductivity and the relative
permittivity, and the shapes of the ground surface such as mountains and rivers. Currently, however,
the FDTD method has only been applied to simple simulation cases of a single-phase distribution line
without overhead shielding wires or lightning arresters, or to a single-phase distribution line with only a
shielding wire. Thus, applying the FDTD method to realistic simulation cases of the lightning-induced
voltages is important research work.
In [68], Tatematsu and Noda applied the FDTD method to the calculation of the lightning-induced
voltages under realistic conditions. In this work, lightning-induced voltages on a three-phase
distribution line with surge arresters and a multipoint-grounded overhead shielding wire were
calculated using the FDTD method, where the nonlinear voltage-current relationship of the surge
arresters was simulated by a piecewise linear function defined by some discrete points in detail using
the technique described in Section 3.1.7. For validation purposes, the calculated results were
compared with those obtained by a traditional method [199], [200] based on the field-to-transmission
coupling technique, that is, the Rusck model [201].

4.4.1.1.2 Calculation arrangement

Figure 4.49 shows the calculation arrangement used here. The bottom surface of the analysis space
was assumed to be a perfectly conducting ground plane and the other surfaces were treated as
absorbing boundaries using Liao’s formulation of the second order [37]. A return-stroke model and the
three-phase distribution line were set up in the analysis space. In this study, the following four cases of
the distribution-line arrangement were considered:
(1) case 1: the distribution line without the shielding or the lightning arresters;
(2) case 2: the distribution line with the multipoint-grounded overhead shielding wire and without the
lightning arresters;
(3) case 3: the distribution line without the shielding wire and with the lightning arresters;
(4) case 4: the distribution line with both the multipoint-grounded overhead shielding wire and the
lightning arresters.
Figure 4.50 represents side views of the calculation arrangement used in case 4.

113
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Return-stroke model

600 m
1400 m
50 m Distribution
line
10 m
Z
300 m
Y

X Perfectly-conducting 650 m
ground plane

Figure 4.49 Calculation arrangement of the lightning-induced voltages on a three-phase distribution line
above a perfectly-conducting ground plane.89

Z 0.6 m 0.6 m Z
Shielding
X Return-stroke model wire X Y
Y
Distribution C B A 1m
1400 m line
Shielding wire Lightning
Striking
Lightning arrester
0.2 m point
arrester 50 m
200 m Distribution
Grounding wire
Grounding line
Grounding
resistance
resistance
Perfectly-conducting ground plane
l Perfectly-conducting
l=0m l = 1400 m ground plane

(a) (b)
Figure 4.50 Side view of the calculation arrangement used in case 4 above the perfectly conducting
ground plane. (a) Side view (ZX) and (b) Side view (YZ).90

The return stroke was located at a position 50 m away from the center of the distribution line. The
return-stroke model was vertical against the ground surface and represented by the transmission line
model composed of a phased array of current sources [17], and its propagation speed was set to 120
m/μs. The return-stroke current had a triangular waveform of which wavefront time, wavetail time, and
peak value are 1 μs, 50 μs, and 100 kA, respectively. The three-phase distribution line was simulated
by three thin wires parallel to the ground plane. The distance between the neighboring phase lines
was 0.6 m, and the radius and height of the distribution line were 4 mm and 10 m, respectively. The
length of the distribution line was 1400 m, and both ends of the distribution line were connected to the
absorbing boundaries to assume no reflection at both ends. The three-phase lines were called
“Phases A, B, and C”, in ascending order of the distance from the return-stroke model. As shown in
Figure 4.50, a local-coordinate l was determined on the distribution line, where l = 0 m and 1400 m
correspond to both ends of the distribution line, respectively. In cases 2 and 4, the multipoint-grounded
overhead shielding wire with a radius of 2.5 mm was located at a position 1 m higher than the
distribution line conductors. The overhead shielding wire was grounded via a lumped-element resistor
of 20 Ω at the interval of 200 m, that is, at the positions of l = 100 m, 300 m, 500 m, … , 1300 m. In
cases 3 and 4, the surge arresters were installed on all the phase lines of the distribution line at the
positions of l = 100 m, 300 m, 500 m, …, 1300 m. The total number of the installed lightning arresters

89 Reprinted with permission from A. Tatematsu and T. Noda, “Three-dimensional FDTD calculation of lightning-induced
voltages on a multiphase distribution line with the lightning arresters and an overhead shielding wire,” IEEE Trans. EMC, vol. 56,
no. 1, p. 162, Figure 3, 2014.
90 Reprinted with permission from A. Tatematsu and T. Noda, “Three-dimensional FDTD calculation of lightning-induced
voltages on a multiphase distribution line with the lightning arresters and an overhead shielding wire,” IEEE Trans. EMC, vol. 56,
no. 1, p. 163, Figure 5, 2014.

114
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

was 21. The V-I characteristics of the lightning arresters were specified in detail using 16 discrete
points as shown in Figure 4.51.

40

Voltage [kV]
20

−20

−40
−500 0 500
Current [A]

Figure 4.51 V-I characteristics of the surge arresters.91

The analysis space was divided into 938 × 461 × 505 nonuniform cells. Although the thin-wire-
representation technique has been proposed for the uniform grid type [44], the sizes of the three cells
adjacent to the thin wire were kept uniform to apply the thin-wire representation technique to the
nonuniform type [202]. In this calculation, the sizes of the three cells adjacent to the thin wires were all
set to 0.2 m and those of the other cells became gradually larger with distance from the thin wires. The
sizes of the other cells ranged from 0.3 m to 2 m. The time discretization was 0.2 ns, which was set to
half of Courant’s condition to avoid numerical instability due to the thin wire representation technique.
Note that, in the FDTD calculation, a GPU-based parallel code was applied to a 16-GPU(Tesla
C2070)-based computer. For example, the calculation time in case 4 was 0.31 hours.

4.4.1.1.3 Calculated results

Under the above described condition, Tatematsu and Noda calculated the lightning-induced voltages
on the distribution line for cases 1-4 by the FDTD method and the conventional circuit-theory based
calculation method [199]. In the FDTD method, the lightning-induced voltages were obtained by
integrating the vertical electric fields between the distribution line and the ground surface. As an
example of the calculated results, Figures 4.52 and 4.53 show the lightning-induced voltages on
phase A of the three-phase distribution line at the positions of l = 600 m, 400m, and 200 m in cases 1
and 4, respectively. The positions of l = 600 m, 400 m, and 200 m are 100 m, 300 m, 500 m away from
the midpoint of the distribution line closest to the striking point, respectively. Although the calculated
results of phases B and C are not illustrated here, the waveforms of the lightning-induced voltages on
phases B and C are similar to the ones on phase A. The peak values of the lightning-induced voltages
on the three phases A, B, and C calculated by both methods in the cases 1-4 are summarized in [68].
In case 1, the differences between the peak values of the voltages obtained by the two methods are
within 0.8 %. In case 2, as is well known, the lightning-induced voltages are suppressed by the
overhead shielding wire in comparison to those in case 1. The differences between the peak values
calculated by the two methods at the position of l = 600 m are within 1.1 %. In cases 3 and 4, the
lightning-induced voltages are further reduced by the lightning arresters, and the peak values
calculated by the FDTD method at the position of l = 600 m agree well with those calculated by the
conventional method within the difference of 5.8 %, but the differences between the two methods in
cases 3 and 4 are larger than those in cases 1 and 2. Unlike the FDTD method, in the circuit-theory-
based calculation method, the vertical grounding wire between the grounding resistance and the
lightning arresters is modeled by a lumped-element inductor. Therefore, the main reasons for these
larger differences in cases 3 and 4 are considered to be as follows. The voltage differences occur due
to the different model of the vertical wire and such voltage differences propagate through the
distribution line. In addition, the different model of the vertical wire has influence on the surge
propagation, that is, the reflection at the vertical wire.

91 Reprinted with permission from A. Tatematsu and T. Noda, “Three-dimensional FDTD calculation of lightning-induced
voltages on a multiphase distribution line with the lightning arresters and an overhead shielding wire,” IEEE Trans. EMC, vol. 56,
no. 1, p. 163, Figure 6, 2014.

115
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(a) (b)
Figure 4.52 Calculated results of the lightning-induced voltages on phase A in case 1 without the
shielding wire or the lightning arresters above the perfectly conducting ground plane. (a) FDTD method
and (b) Rusck model.92

(a) (b)
Figure 4.53 Calculated results of the lightning-induced voltages on phase A in case 4 without the
shielding wire or the lightning arresters above the perfectly conducting ground plane. (a) FDTD method
and (b) Rusck model.93

From these results, the validity of the technique for representing the lightning arresters and the three-
phase distribution line with the multipoint-grounded overhead shielding wire using the nonuniform grid
in the FDTD method is confirmed.
In addition, Tatematsu and Noda calculated the lightning-induced voltages on the three-phase
distribution line with the multi-point grounded shielding wire and the surge arresters while taking into
account the finite conductivity of the ground to confirm the numerical stability of the FDTD method with
the nonlinear elements applied to the practical arrangement of the distribution line [68].

4.4.1.2 Lightning-induced voltages on multiconductor lines with surge arresters,

shielding wires, and pole Transformers
The main protective measures against short interruptions and voltage sags originated by lightning-
induced voltages in medium voltage networks can be identified as (i) use of shielding wires and/or (ii)
use of surge arresters, and (iii) increasing of the line insulation level.
Early works on the effect of shielding wires have considered them to be at zero potential at any point
along their length and at any time. Later, imporved approaches have been adopted in which the
shielding wire is considered as one of the conductors of the multi-conductor line (see e.g., [199], [203]-
[205]).
In [205], Paolone et al. investigated the effect of periodically-grounded shielding wires and surge
arresters on the attenuation of lightning-induced voltages. They compared the mitigation effect of
shielding wires with that achievable by the insertion of surge arresters along the line. The computation
results were first validated by means of experimental results obtained using a reduced-scale line

92 Reprinted with permission from A. Tatematsu and T. Noda, “Three-dimensional FDTD calculation of lightning-induced
voltages on a multiphase distribution line with the lightning arresters and an overhead shielding wire,” IEEE Trans. EMC, vol. 56,
no. 1, p. 164, Figure 7, 2014.
93 Reprinted with permission from A. Tatematsu and T. Noda, “Three-dimensional FDTD calculation of lightning-induced
voltages on a multiphase distribution line with the lightning arresters and an overhead shielding wire,” IEEE Trans. EMC, vol. 56,
no. 1, p. 164, Figure 10, 2014.

116
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

model illuminated by a nuclear electromagnetic pulse (NEMP) simulator. One of the main conclusions
of their study was that the effectiveness of shielding wires and surge arresters depends mostly on the
spacing between two adjacent grounding points or surge arresters.
In [206], Thang et al. have computed lightning-induced voltages on multiconductor lines with surge
arresters and pole transformers using the 3D-FDTD method. The FDTD method employs a subgrid
model, the same as the one used by Sumitani et al. [207], in which spatial discretization is fine (cell
side length is 0.5 m) in the vicinity of overhead wires (1455m × 320m × 30m) and coarse (cell side
length is 5 m) in the rest of the computational domain.
Figure 4.54 shows the experimental configuration of 1-cm-radius and 1.4-km-long three-phase
conductors and neutral conductor located above perfectly conducting ground, which represents a 15-
kV nonenergized distribution line [208]. Figure 4.54 (a) shows the cross-sectional (yz-plane) view of
conductors system, and Figure 4.54 (b) shows the plan (xy-plane) view of the distribution network.
Three-phase conductors and one neutral conductor are located 10 and 8 m above ground,
respectively. The distance between adjacent upper-phase conductors is 0.75 m, and each end of each
phase conductor is connected to ground via a 455-Ω matching resistor. The distance between the
lightning strike point and the voltage measuring point, M1, is about 82 m (20 m in x direction and 80 m
in y direction).

Figure 4.54 Experiment configuration of 1-cm-radius and 1.4-km-long four-conductor line: (a) cross-
sectional (yz-plane) view of a four-conductor line and (b) plan (xy-plane) view of the distribution network
A.94

Figure 4.55 shows the circuit representation of each component of the distribution network presented
in Figure 4.54 (b). The lightning-induced voltages were measured at node M1. The V–I characteristic
of the surge arrester model [see Figure 4.55 (a)] is fairly similar to that of an actual ZnO distribution
arrester with rated voltage and current of 12 kV and 5 kA, respectively [208]. Each element such as a
nonlinear or linear resistor, inductor or capacitor is represented by one side of the cell in the present
FDTD simulation. Although in [208] the grounding downconductors were represented by 10-μH

94 Reprinted with permission from T. H. Thang, Y. Baba, V. A. Rakov, and A. Piantini, “FDTD computation of lightning-induced
voltages on multi-conductor lines with surge arresters and pole transformers,” IEEE Trans. EMC, vol. 57, no. 3, p. 443, Figure 2,
2015.

117
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.55 Circuit representation of different system elements in the model: (a) surge arrester equivalent
circuit and voltage–current curve of the nonlinear resistor Rpr, (b) surge arrester and transformer point,
(c) grounding point, and (d) transformer and measuring point.95

inductors, in the FDTD simulation they are represented by vertical perfectly conducting wires of radius
1 cm.
Figure 4.56 shows the 3D view of the lightning channel, represented by a 900-m long vertical phased-
current-source-array model [17], simulating the transmission line model [209], and the distribution
network for the four-conductor line system. The upward propagation speed of current along the
simulated lightning channel is set to 0.11c, where c is the speed of light, following the experimental
condition. Note that the length of the simulated lightning channel in the experiment was 600 m (12 m
in the 1/50 small-scale model), which will not affect the lightning-induced voltages within the first 18 μs
or so.
For FDTD computations, this system is accommodated in a working volume of 1480m × 500m ×
1000m, which is divided into cubic cells of 5m × 5m × 5m, except for a space in the vicinity of the
distribution network (1455m × 320m × 30m), where cubic cells of 0.5m × 0.5m × 0.5m are employed.
The total number of cells in the working volume is about 11.8 × 107 (≈ 1480/5 × 500/5 × 1000/5 +
1455/0.5 × 320/0.5 × 30/0.5). Liao’s second-order absorbing boundary condition [37] is applied to five
planes (the top plane and four side planes) to minimize unwanted reflections there. The bottom plane
is set to be a perfect conductor. Although the distance between the phase conductors was 0.75 m in
the measurement, it is set to 1 m in the FDTD simulation. This difference has little influence on the
lightning-induced voltages on phase conductors.
The current injected into the simulated lightning channel was represented by a triangular waveform
with peak value of 34 kA, front time of 2 μs and half-peak width of 85 μs, which was used in [208]. The
initial part of the current is shown in Figure 4.57. Figure 4.58 shows the lightning-induced voltage
waveform at node M1 of the network presented in Figure 4.54 (b), computed using the FDTD method
for the triangular lightning current with 34-kA peak. Also shown in this figure is the corresponding
measured voltage waveform. Note that the starting time (t = 0 in the plot) of the measured waveform

95 Reprinted with permission from T. H. Thang, Y. Baba, V. A. Rakov, and A. Piantini, “FDTD computation of lightning-induced
voltages on multi-conductor lines with surge arresters and pole transformers,” IEEE Trans. EMC, vol. 57, no. 3, p. 444, Figure 3,
2015.

118
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.56 Three-dimensional view of the lightning channel and the four-conductor line distribution
network, analyzed using the 3-D FDTD method.96

Figure 4.57 Initial (8-μs) part of the simulated lightning current waveform with a peak of 34 kA, front time
of 2 μs, and half-peak width of 85 μs.97

was adjusted since sometimes, due to noise, the exact time of t = 0 was unknown in the measured
results. It follows from Figure 4.58 that the FDTD-computed lightning-induced voltage waveform
agrees reasonably well with the corresponding measured waveform. Magnitudes of current flowing
through the surge arresters located closest and second closest to the lightning strike point [see Figure
4.54(b)] are about 800 A and about 100 A, respectively. Magnitudes of current of other arresters are
smaller than 50 A. It follows that arresters located closer to the lightning strike point are more involved
in suppressing the lightning-induced voltages.

96 Reprinted with permission from T. H. Thang, Y. Baba, V. A. Rakov, and A. Piantini, “FDTD computation of lightning-induced
voltages on multi-conductor lines with surge arresters and pole transformers,” IEEE Trans. EMC, vol. 57, no. 3, p. 444, Figure 4,
2015.
97 Reprinted with permission from T. H. Thang, Y. Baba, V. A. Rakov, and A. Piantini, “FDTD computation of lightning-induced
voltages on multi-conductor lines with surge arresters and pole transformers,” IEEE Trans. EMC, vol. 57, no. 3, p. 445, Figure 6,
2015.

119
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.58 FDTD-computed waveform of lightning-induced voltage and the corresponding measured
waveform [at point M1].98

4.4.1.3 Lightning-induced voltages in the presence of nearby buildings

In [210], Thang et al. have computed lightning-induced voltages on overhead distribution lines in the
presence of nearby buildings using the 3D FDTD method. In the simulations, four-conductor lines with
surge arresters and pole transformers are also considered. It appears that the presence of nearby
buildings causes the reduction of lightning-induced voltages, as expected. The observed trend is in
general agreement with that reported from the small-scale experiment by Piantini et al. [208], [211].
Figure 4.59 shows the experimental configurations used in [208], [211], where the dimensions are
referred to the full-scale system by taking into account the scale factor for lengths (1:50). It includes 1-
cm-radius and 1.4-km-long three-phase conductors and neutral conductor located above perfectly
conducting ground, which represent a 15-kV unenergized distribution line. Figure 4.59 (a) shows the
case of no buildings (building height he = 0 m), Figure 4.59 (b) is for he = 5 m, and Figure 4.59 (c) is for
he = 15 m. The three horizontally arranged phase conductors and one neutral conductor are located
10 and 8 m above ground, respectively. The distance between adjacent phase conductors is 0.75 m,
and each end of each phase conductor is connected to ground via a 455-Ω matching resistor. The
distance between the lightning strike point and the voltage measuring point, M, is 20 m. The distances
between the measuring point and the closest set of surge arresters located on its left and right sides
are labeled se and sd, respectively. Two cases with different values of se and sd were considered: case
1 with se = 75 m and sd = 75 m, and case 2 with se = 148 m and sd = 174 m.
Figure 4.60 shows the 3-D view of the lightning channel and the distribution network for the four-
conductor line system. For FDTD computations, this system is accommodated in a working volume of
1480 m × 500 m × 1000 m, which is divided into cubic cells of 5 m × 5 m × 5 m, except for the space
in the vicinity of the distribution network (1455 m × 320 m × 30 m), where cubic cells of 0.5 m × 0.5 m
× 0.5 m are employed. Liao’s second-order absorbing boundary condition [37] is applied to five planes.
Note that each element such as a nonlinear or linear resistor, inductor or capacitor is represented by
one side of the cell in the present FDTD simulation.
Figure 4.61 (a)–(c) shows lightning-induced voltage waveforms at node M of the network shown in
Figure 4.59, computed using the FDTD method for buildings having heights of 0, 5, and 15 m. In the
simulations, the distances between the measuring point (M) and the closest set of surge arresters
were set to se = 75 m and sd = 75 m. Also shown in these figures are the corresponding measured
voltage waveforms. Figure 4.62 (a)–(c) is the same as Figure 4.61 (a)–(c), except for se = 148 m and
sd = 174 m. From these figures, the ratios of the calculated voltage peaks corresponding to building
heights he = 5 and 15 m with respect to he = 0 m are 0.85 and 0.53 for Figure 4.61 and 0.84 and 0.50
for Figure 4.59, whereas for the measured voltages the ratios are 0.75 and 0.59 for Figure 4.61 and
0.73 and 0.34 for Figure 4.62, respectively.

98 Reprinted with permission from T. H. Thang, Y. Baba, V. A. Rakov, and A. Piantini, “FDTD computation of lightning-induced
voltages on multi-conductor lines with surge arresters and pole transformers,” IEEE Trans. EMC, vol. 57, no. 3, p. 445, Figure 7,
2015.

120
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

90 m 210 m 210 m 300 m 210 m 210 m 170 m

150 m M
se sd

20 m

150 m
Lightning stroke location

Surge arrester Transformer Grounding point M Measuring point

(a)

90 m 210 m 210 m 300 m 210 m 210 m 170 m

22 m
150 m M

20 m

150 m
Lightning stroke location

30 m 40 m
80 m 22 m
80 m 80 m
80 m

Surge arrester Transformer Grounding point M Measuring point

(b)

90 m 210 m 210 m 300 m 210 m 210 m 170 m

22 m
150 m M

20 m

150 m
Lightning stroke location

30 m 40 m
80 m 22 m
80 m 80 m
80 m

Surge arrester Transformer Grounding point M Measuring point

(c)
Figure 4.59 Experimental configuration 1: Plane view of 1-cm-radius and 1.4-km-long four-conductor line
with nearby buildings: (a) he = 0 m (no buildings), (b) he = 5 m, and (c) he = 15 m. The distances between

121
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

the measuring point and the closest set of surge arresters located on its left and right sides are labeled se
and sd. Indicated dimensions refer to the full-scale system.99

Figure 4.60 Three-dimensional view of the lightning channel and the four-conductor line distribution
network, analyzed using the 3-D FDTD method.100

It follows from these figures that the presence of nearby buildings causes the reduction of lightning-
induced voltages, as expected. The observed trend is in general agreement with that measured in
[208], [211]. This indicates that despite the differences in magnitude and phase, measured and
calculated waveforms are generally similar and exhibit the same trend: decrease as the height of
building increases.

4.4.2 HEM simulation of lightning-induced surges

4.4.2.1 Introduction
The relevance of lightning-induced voltages for the performance of medium- and low-voltage
distribution lines was proved a long time ago. Several studies of their impact on electric systems were
presented in the literature since the last century. Traditionally, the modeling of the electromagnetic
interaction between lightning channel and nearby overhead lines follows two distinct approaches, the
so-called ‘coupling models’ and ‘electromagnetic models’. The latter approach, in which the HEM is
included, the current distribution along the lightning channel and the coupling effects between channel
and line conductors are calculated following an integrated approach, avoiding the previous definition of
a return-stroke model. Except for the FDTD method [16], [212], most of these kinds of models are
based on a frequency-domain approach, the results being obtained in the time domain using the
Fourier transform (e.g, Numerical Electromagnetics Code - NEC [213], [214], HEM [21], [157],
COMSOL Multiphysics software [215], [216].
Several papers considering HEM application for lightning-induced voltages evaluations are presented
in the literature [25], [150], [217]-[220]. The details of model application for this kind of study are
summarized as follows.

4.4.2.2 Calculation of lightning-induced voltage

The induced voltage on an overhead line due to nearby lightning is determined by the HEM model
following two procedures, both leading to the same results [147]:

99 Reprinted with permission from T. H. Thang, Y. Baba, A. Piantini, and V. A. Rakov, ”Lightning-induced voltages in the
presence of nearby buildings: FDTD simulation versus small-scale experiment,” IEEE Trans. EMC, vol. 57, no. 6, p. 1602,
Figure 1, 2015.
100 Reprinted with permission from T. H. Thang, Y. Baba, A. Piantini, and V. A. Rakov, ”Lightning-induced voltages in the
presence of nearby buildings: FDTD simulation versus small-scale experiment,” IEEE Trans. EMC, vol. 57, no. 6, p. 1604,
Figure 3, 2015.

122
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

(i) Induced voltage calculated by integrating the resulting electric field from soil surface to line
conductor, considering both the conservative (electrostatic) and non-conservative (magnetic)
components of electric field [219];
(ii) Induced voltage calculated as the voltage across a defined-impedance-value element inserted
between the line and ground [150], in a procedure similar to that adopted in NEC-4 [214].
The application of the HEM model for evaluations of lightning-induced voltage on overhead lines

150
Computed
Measured
100
Induced voltage [kV]

50

-50
0 2 4 6 8 10

Time [s]
(a)

150
Computed
Measured
100
Induced voltage [kV]

50

-50
0 2 4 6 8 10

Time [s]
(b)

150
Computed
Measured
100
Induced voltage [kV]

50

-50
0 2 4 6 8 10

Time [s]
(c)
Figure 4.61 FDTD-computed and corresponding measured lightning-induced voltage waveforms for a
triangular lightning current pulse with peak of 34 kA, risetime of 2 μs, and time to half-peak value of 85 μs

123
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

and with buildings having heights of (a) 0, (b) 5, and (c) 15 m. The distances between the measuring point
(M) and the closest set of surge arresters were se = sd = 75 m.101

250
200 Computed
Measured
150

Induced voltage [kV]

100
50
0
-50
-100
-150
-200
0 2 4 6 8 10

Time [s]
(a)
250
200 Computed
Measured
150
Induced voltage [kV]

100
50
0
-50

-100
-150
-200
0 2 4 6 8 10

Time [s]
(b)
250
200 Computed
Measured
150
Induced voltage [kV]

100
50
0
-50
-100
-150
-200
0 2 4 6 8 10

Time [s]
(c)
Figure 4.62 FDTD-computed and corresponding measured lightning-induced voltage waveforms for a
triangular lightning current pulse with peak of 34 kA, risetime of 2 μs, and time to half-peak value of 85 μs

101 Reprinted with permission from T. H. Thang, Y. Baba, A. Piantini, and V. A. Rakov, ”Lightning-induced voltages in the
presence of nearby buildings: FDTD simulation versus small-scale experiment,” IEEE Trans. EMC, vol. 57, no. 6, p. 1604,
Figure 5, 2015.

124
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

and with buildings having heights of (a) 0, (b) 5, and (c) 15 m. The distances between the measuring point
(M) and the closest set of surge arresters were se = 148 m and sd = 174 m.102

considers the representation of relevant features of lightning-return stroke channel that are capable of
affecting the return-stroke current distribution, such as the losses along channel core, corona radius,
and the variation of return stroke current speed and of the height of the attachment between upward
and downward leaders.
Channel core losses are represented by attributing defined values to the resistance of unit length of
channel core. Corona effect is represented by an equivalent increase on channel radius, aiming to
enlarge the equipotential surface, which is established around channel, just before the attachment and
flow of return-stroke current. This kind of approach, computed from “qV” curves, is verified to be
consistent in experimental tests involving high-voltage electrodes. The radius amplification is applied
only for transversal current in the model. This means that, for the longitudinal current that flows along
the core, the radius remains the same of original core. However, it is important to note that this kind of
representation does not consider the time needed to the establishment of the corona sheath before
attachment. Such radius is instantaneously represented along all the channel. Sensitivity analysis of
the effect of such parameters on the resulting lightning return stroke distribution are presented in
[221].
The lightning return stroke current speed is controlled by an artificial increase of both the permittivity
and permeability values in the HEM expression used to calculate those self-elements in matrix ZL
corresponding to the channel representation. Such a procedure applies only to these specific self-
elements, maintaining the electromagnetic coupling between channel and line considering air as the
involving medium [150].
Also, the HEM model allows simulating the upward leader of a lightning return-stroke channel. The
effect of the attachment height of downward and upward leaders, that typically occurs at some tens of
meters above soil and more above tall structures, is investigated in [219].
The effect of lossy ground on lightning-induced voltages is considered by means of a modification of
Norton’s approximation [149] implemented in the HEM model [150]. More specifically, such effect is
attained by modifying those mutual elements of matrix ZL that represent the coupling between the
return-stroke channel and the line. Such elements are calculated assuming that the voltage drop
across the line is the integral of the horizontal electric field given by Norton, dropping the two last
terms that represent the effect of the electrostatic field. This effect is already computed by the
transversal coupling between the return-stroke channel and the line in matrix ZT [150].
As a step forward, the frequency-dependent effect of soil parameters was implemented in the HEM by
means of considering Visacro-Alipio and Alipio-Visacro models [152], [222].
In [157], the impact of the frequency dependence of soil resistivity and permittivity on lightning
overvoltages induced on overhead lines over lossy ground is investigated considering the
implementation of Visacro-Alipio expressions on the HEM. The most relevant results are presented as
follows.

4.4.2.3 Lightning-induced voltage considering the frequency-dependent effect of soil

parameters
The results presented below are related to lightning-induced voltages on a 1-km-long, 10-m-high
single overhead line, impedance matched at both line extremities due to stroke locations A and B
illustrated in Figure 4.63.

102 Reprinted with permission from T. H. Thang, Y. Baba, A. Piantini, and V. A. Rakov, ”Lightning-induced voltages in the
presence of nearby buildings: FDTD simulation versus small-scale experiment,” IEEE Trans. EMC, vol. 57, no. 6, p. 1605,
Figure 6, 2015.

125
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

1 km
50 m
50 m 50 m

B A
Stroke location

Figure 4.63 Lightning stroke locations A and B.103

Simulations assumed the representative current waveforms of first and subsequent strokes illustrated
in Figure 4.64. Such waveforms are analytically obtained by the sum of Heidler functions [188], [223]
and closely reproduce the median current parameters measured at Mount San Salvatore station
(MSS) [189].

35 14
30 12

Current (kA)
Current (kA)

25 10
20 8
15 6
10 4
5 2
0 0
0 10 20 30 40 0 2 4 6 8 10
Time (ms) Time (ms)

(a) (b)
Figure 4.64 Simulated first (a) and subsequent (b) return stroke current waveforms. Median peak currents
and front times measured at MSS station: (a) Ip = 31.1 kA, Td30=3.83 µs and (b) Ip = 11.8 kA, Td30 = 0.67
µs.104

The lightning return-stroke channel was simulated as an 1800-m-long vertical conductor impedance
matched at its top. The assumed current propagation speed for both first and subsequent stroke
events was 0.3c.
For first stroke events, simulations assumed the attachment of the upward and downward leaders
occurring at 50-m height. The earth termination was represented by connecting a low-value
impedance between the channel bottom and remote earth [224]. The simulation of subsequent stroke
events assumed current departing from soil level.

4.4.2.3.1 Lightning-induced voltages: stroke location A

Figure 4.65 illustrates the induced voltages waveforms at the center of the simulated overhead line
considering stroke location A, for low-frequency soil resistivity ρ0 from 100 to 10,000 Ωm and constant
and frequency-dependent electrical parameters of soil. Table 4.9 summarizes the peak voltages for
both conditions.
The results indicate that the frequency dependence effect of soil parameters contributes to reduce the
amplitude of induced voltages at the center of the simulated overhead line. Such reduction is even
more relevant for increasing values of soil resistivity. Also, the decreases experienced by voltages
induced by first and subsequent strokes are very similar: for soil resistivity of 1000, 2500, 5000 and
10,000 Ω.m the observed decreases are about 7, 15, 24, and 34%, and about 8, 17, 24 and 33% for
first and subsequent strokes, respectively.
Figure 4.66 illustrated results of induced voltage at line extremity of the overhead line considering
stroke location A. The obtained results are quite different from those related to line center and
depicted in Figure 4.65.
Differently from the results of Figure 4.65, the slight impact of the frequency dependence effect on the
amplitude of voltages induced by first stroke at line is noted. Also, the waveforms are quite different

103 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1130,
Figure 2, 2014.
104 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1130,
Figure 1, 2014.

126
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

from those of Figure 4.65. Such behaviors are justified by the dynamics resulting from the 50-m-high
attachment.
For subsequent strokes, the frequency-dependent soil parameters increase the amplitude of induced
voltages for high-resistivity soils. The explanation of such behaviors are presented in [157] and are
related to the effect of the transition from the condition of a perfectly conducting soil to a highly
resistive one and the behavior of the horizontal electric field at far distances from the stroke location.
To complement, the frequency dependence of soil parameters results in the reduction of soil resistivity
with increasing frequency, leading to a condition closer to that of a perfectly conducting soil. This may
result in amplitudes of induced voltages larger than the ones developed under the assumption of
constant soil parameters [157].
140 100
r = r 0 ,er = 10 r = r 0 ,er = 10
120
r = r w,w 80 r = r w,w
Voltage (kV)

Voltage (kV)
100
80 60
60 40
40
20
20
0 0
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(a) (b)

250 120
r = r 0 ,er = 10 r = r 0 ,er = 10
200 r = r w,w 100 r = r w,w
Voltage (kV)

Voltage (kV)
80
150
60
100
40
50 20
0 0
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(c) (d)
500 200
r = r 0 ,er = 10 r = r 0 ,er = 10
400 r = r w,w r = r w,w
Voltage (kV)

150
Voltage (kV)

300
100
200
50
100
0 0
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(e) (f)
Figure 4.65 Lightning-induced voltages at the center of the overhead line considering stroke location A
for constant and frequency-dependent soil parameters for different values of soil resistivity ρ0. (Left
column: First stroke; Right column: Subsequent stroke). (a),(b) 100 Ω.m, (c),(d) 1000 Ω.m, a€(e),(f) 10,000
Ω.m. (v=0.3c).105

Table 4.9 Induced voltage peak values at the center of the overhead line considering stroke location A.106
ρ0 First stroke Subsequent stroke
(Ω.m) ρ=ρ0, εr=10 ρ=ρ(ω), ε(ω) % ρ=ρ0, εr=10 ρ=ρ(ω), ε(ω) %
100 139.0 138.0 -0.7 85.8 85.1 -0.8
1000 213.8 198.0 -7.4 111.6 102.3 -8.3
10,000 406.4 267.9 -34.1 172.1 116.2 -32.5

105 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1131,
Figures 3a, 3b, 3c, 3d, 3i, and 3j, 2014.
106 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1131, Table
1, 2014.

127
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.4.2.3.2 Lightning-induced voltages: stroke location B

Figure 4.67 illustrates lightning-induced voltages at the center and at both close and far line
terminations, considering stroke location B, for 1000 and 10,000 Ω.m soil resistivity. Table 4.10
summarizes the overvoltage decrease promoted by frequency-dependent soil parameters.
The resulting induced voltage profile is quite different from those related to stroke location A, notably
the negative polarity of the voltages observed at the center and far extremity. However, the frequency
dependence effect of soil parameters follows the same trend. In addition to peak voltage decrease, it
is also observed the significant reduction of the slope in the wavefront, contributing to smooth the
waveform.
This behavior influences the decrease of peak voltage differently. Minimum reductions about 7% are
observed for 1000 Ω.m soil resistivity at the close extremity while decreases about 50% are observed
at the far extremity for the 10,000 Ω.m soil.
100 60
r = r 0 ,er = 10 r = r 0 ,er = 10
80 50
r = r w,w r = r w,w
Voltage (kV)

Voltage (kV)
40
60
30
40 20
20 10
0 0
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(a) (b)

80 50
70 r = r 0 ,er = 10 r = r 0 ,er = 10
r = r w,w 40 r = r w,w
Voltage (kV)
Voltage (kV)

60
50 30
40
30 20
20 10
10
0 0
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(c) (d)

60 40
50 r = r 0 ,er = 10
r = r w,w 35 r = r 0 ,er = 10
Voltage (kV)

40 30 r = r w,w
Voltage (kV)

30 25
20 20
10 15
0 10
-10 5
-20 0
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(e) (f)
Figure 4.66 Lightning-induced voltages at the extremity of the overhead line considering stroke location A
for constant and frequency-dependent soil parameters for different values of soil resistivity ρ0. (Left
column: First stroke; Right column: Subsequent stroke). (a),(b) 100 Ω.m, (c),(d) 1000 Ω.m€nd (e),(f) 10,000
Ω.m. (v=0.3c).107

4.4.2.4 Summary
In general, the effects of the frequency dependence on the induced voltage consist of the reduction of
the voltage amplitude and the distortion of its waveform. The reductions are very similar for voltages
induced by first and subsequent strokes. It is negligible for 100 Ω.m soils and reaches about 33% for
10,000 Ω.m soils.
The voltage induced at line extremities for strokes nearby the line center (position A) is an exception to
the behavior described above. Though the effect of soil-parameter frequency dependence remains

107 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1132,
Figures 4a, 4b, 4c, 4d, 4i, and 4j, 2014.

128
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

negligible for low resistivity soils, specifically for high resistivity soils, it is responsible for increasing the
voltage induced by lightning currents departing from ground level.
These results have a relevant impact on lightning protection of medium and low-voltage electric
systems installed in high-resistivity soils, contributing to decrease the expected flashover rate.

4.5 Lightning surges in wind turbine generator towers

In recent years, accidents associated with the use of a large number of wind turbines have increased
in number. In particular, lightning surge causes extensive and serious damage to electrical and/or
electronics systems [225]-[230].
For the lightning protection of wind turbines, the FDTD method has been widely used. One of the
research fields to which the FDTD simulations apply is transient characteristics of wind turbine
150 40
r = r 0 ,er = 10 r = r 0 ,er = 10
r r w,w 30

Voltage (kV)
r r w,w
Voltage (kV)

100 close
20
close
50 10
center center
0
0
-10 far
far
-50 -20
0 5 10 15 20 0 5 10 15 20
Time (ms) Time (ms)

(a) (b)

250 80
200 60
Voltage (kV)

150
Voltage (kV)

40 close
100 r = r 0 ,er = 10
20
50 close r r w,w 0
center
0 -20
-50 center
-40 far r = r 0 ,er = 10
-100 -60
-150
far r r w,w
-80
0 5 10 15 20
0 5 10 15 20
Time (ms) Time (ms)

(c) (d)
Figure 4.67 Lightning-induced voltages at the center and both close and far terminations of the overhead
line considering stroke location B for constant and frequency-dependent soil parameters for different
values of soil resistivity ρ0. (Left column: First stroke; Right column: Subsequent stroke). (a),(b) 1000 Ω.m
and (c),(d) 10,000 Ω.m. (v=0.3c).108

Table 4.10 Decrease of Induced voltage amplitude at line terminations and line center due to the
frequency dependence of soil parameters for stroke location B (ρ0 = 1000, 5000, and 10,000 Ωm).109
Overvoltage decrease (%)
ρ0 First stroke Subsequent stroke
(Ω.m) Close Center Far Close Center Far
1000 7.6 32.3 17.4 9.2 33.3 26.5
5000 26.6 41.2 27.6 26.8 46.3 42.1
10,000 30.6 53.5 40.9 33.7 52.7 49.4

grounding systems. In this field, measurements of the transient grounding characteristics at several
wind turbines are used as the basis for FDTD simulation models. The FDTD method is also applied to
surge propagation on a grounding wire between two wind turbines. An EMTP model that is equivalent
to the FDTD simulation model has also been constructed. In this section, these applications of the
FDTD methods are introduced.

108 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1132,
Figures 5a, 5b, 5e, and 5f, 2014.
109 Reprinted with permission from F.H. Silveira, S. Visacro, R. Alipio, A. De Conti, “Lightning-Induced Voltages Over Lossy
Ground: The Effect of Frequency Dependence of Electrical Parameters of Soil,” IEEE Trans. EMC, vol. 56, no. 5, p. 1133, Table
2, 2014.

129
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.5.1 Transient characteristics of a wind turbine grounding system

Measurements of transient grounding characteristics have been carried out at several wind turbines
[231]-[238]. Some of the measured grounding characteristics have already been verified [239]-[242]
using the FDTD method [16]. From the measured results, it has been obvious that the grounding
characteristics strongly depend on the stratiform ground resistivity at each wind turbine site, and
shape of the grounding system including the foundation made of reinforced concrete.
The grounding systems of wind turbines consist of the foundation, ring earth electrodes, grounding
meshes, foundation feet, grounding copper plates, counterpoises and so on. Insulating coating or bare
wire is used for the counterpoise, which is sometime utilized for junctions between several grounding
systems.
The grounding systems of wind turbines can be diverse, as shown in Figure 4.68 [243]. The shapes of
the foundations and foundation feet are designed considering structural stability, not lightning
protection for electrical and electronics equipment. However, the grounding characteristic of a wind
turbine depends considerably on these shapes. There are few grounding systems which are designed
taking into consideration both transient and steady-state grounding characteristics efficiently. Design
methodologies could be improved in order to realize better, cost-effective grounding characteristics of
wind turbines. The FDTD simulation is suitable for such designs.

Top view 12 m

12 m

Grounding wire
2m
4m Insulated wire Foundation of
4.3 m
Vertical grounding rod substation
Side view facilities
14 mm × 1.5 m
4m

Grounding plate 1.5 Insulated wire

m
1.6m
3.3m Grounding plate
1.9m 0.9 m×0.9m ×1.5mm
Grounding rod 16 m

(a) (b)
Figure 4.68 Grounding system of a wind turbine. (a) Top and side views and (b) Incidental grounding
electrodes.110

Figure 4.69 shows one of the layouts used in the experiments. In the experiment, IV wire was laid over
100 m from the impulse generator at a height of 1 m from the ground surface to an area near the point
of current injection at the tower foot of the wind turbine generator system, and was connected to the
foundation through a resistor of 400-500 Ω that acted as a matching resistance. The impulse
generator is regarded as a source of current because the value of the resistor is much smaller than
the grounding impedance of the wind turbine grounding system. Therefore, the steep wavefront
current can be injected.

110 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 79, Figure 1, 2012.

130
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

42 m
IV cable
58 m I
Voltage measurinng wire
500W
　
V

I
Current read wire

75 m

I.G
25 m
Figure 4.69 Experimental set-up.111

To recognize the wind turbine grounding characteristic, the voltage between the voltage measuring
wire grounded at a remote point and the foundation is usually measured as the potential rise of the
grounding system with the injected current. Figure 4.70 shows an example of measured and simulated
results of the injected current and the potential rise.

50 60

50
40
Simulation 40
voltage[V]
current[A]

30
Experiment 30 Experiment
20
20
10 10

0 0
Simulation
-10 -10
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30
time[s] time[s}

(a) (b)
Figure 4.70 Measured and simulated results of the injected current and the potential rise. (a) Injected
current waveform and (b) Potential rise waveform.112

The stratiform ground resistivity at the site of the wind turbine generator system is also important to
simulate lightning surge phenomena. The ground resistivity is usually measured by the Wenner
method. The typical resistivity is shown in Figure 4.71 at the site of Figures 4.69 and 4.70. Incidentally,
the steady-state grounding resistance of the grounding system of the wind turbine generator system
was 0.57 Ω.

111 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 80, Figure 3a, 2012.
112 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 81, Figure 5, 2012.

131
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

40 Wm 0.5 m

20 Wm ∞

Figure 4.71 Layers of ground resistivity at the test site.113

The measurements were reproduced by using electromagnetic field analysis through the FDTD
method. The simulation model is shown in Figure 4.72. The dimensions of the analysis space were
130.5 m × 180 m × 146 m, and it was divided into cube cells with a side length of 0.5 m. The
absorbing boundary condition was set as 2nd order Liao. The ground level was 24.5 m from the
bottom of the analysis space; the resistivity of the ground was the same as that shown in Figure 4.71.
Thin-wire models to model the current lead wire, voltage measuring wire, and grounding mesh were
used [44], [244]. The foundation of the wind turbine and the substation facilities were modeled as
rectangular parallelepiped conductors of 2 m × 12 m × 12 m and 0.5 m × 2 m × 4 m, respectively. The
current source parallel with the resistance of 500 Ω was connected between the foundation and
current lead wire.

Voltage measuring wire

V 146 m

Current read wire I

V soil
180 m
40 Wm 0.5 m
20 Wm 24 m

130.5 m

Figure 4.72 FDTD simulation model.114

Figure 4.70(a) shows the comparison of the simulated and measured injected currents. The injected
current showed a ramp wave which includes a wide frequency component, and its peak value and rise
time were approximately 45 A and 0.2 μs, respectively. The simulated injected current agreed well
with the measured current.
The potential rise at the top of the foundation as shown in Figure 4.70 (b) exhibited inductive
characteristics. The transient grounding impedance is greater than that of steady-state. The voltage
waveform oscillated at the wavefront. The potential rise at the wave tail converges to that correspond
to the steady-state grounding resistance of 0.57 Ω. At the wavefront, there were a few differences.
However, the simulated potential rise agreed well with the measured values.

113 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 80, Figure 2, 2012.
114 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 81, Figure 4, 2012.

132
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.73 shows comparisons of the potential rise around the foundation at several points. As seen
in the potential rise of the foundation, the inductive effect was seen in the potential rises around the
foundation. The wave shape shown in Fig. 6 was almost analogous to the potential rise of the wind
turbine grounding system shown in Figure 4.70 (b).

60 60

50 50

40 40
voltage[V]

voltage[V]
30 30

20 20
Simulation
10 10 Simulation
0 Experiment 0
Experiment
-10 -10
0 5 10 15 20 25 30 0 5 10 15 20 25 30
time[s] time[s]

(a) (b)
Figure 4.73 Potential rises around the foundation. (a) Result at 6 m from the foundation and (b) Result at
20 m from the foundation.115

Figure 4.74 shows the comparisons of the simulated and measured values of the grounding potential
rise at the wavefront and wave tail around the foundation. The simulated results agreed well with the
measured results, as shown in Figure 4.74.
50 14

12
40

10 Experiment
Experiment Simulation
voltage[V]

Simulation
voltage[V]

30
8

6
20

10
2

0 0
0 5 10 15 20 0 5 10 15 20

distance[m] distance[m]

(a) (b)
Figure 4.74 Potential rises around the foundation. (a) Comparisons of the peak values and (b)
Comparisons of the wave-tail values.116

4.5.2 Common-mode lightning current from a structure struck by lightning to a wind

turbine
In Taikoyama wind farm constructed in the north part of Kyoto, Japan and near the Sea of Japan, wind
turbines have often been damaged by lightning in winter. To solve the problem and build lightning
protection methodologies, a meeting of the committee had been convened from March, 2003; many
professionals of lightning protection and wind turbines had joined. As a result of the discussions at the
meeting, a tall tower was built near one of wind turbines in December, 2003. Following construction of

115 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 82, Figures 6d and 6g, 2012.
116 Reprinted with permission from K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic
of a 600 kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, p. 82, Figure 7, 2012.

133
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

the tower, many lightning discharges have been caught by the tower and the efficiency of the tower has
been recognized.
In the winter from 2011 to 2012, two typical accidents occurred. In one of the accidents, a fire occurred
inside the wind turbine near the tall tower on January 28th, 2012. This accident was caused by a lightning
strike to the wind turbine itself; remarkably, the BTB (Back-To-Back) system had burned out. Another
accident was the breakdown of the anemometers placed on the tall tower on March 24th, 2012.
Both accidents were caused by the potential differences between the grounding systems of the wind
turbine and the neighboring tall tower. To equalize the potentials, these grounding systems have been
connected as shown in Figure 4.75 [245].
In this section, the comparisons between the measured and the FDTD simulation results of the synthetic
grounding system are introduced.
In Taikoyama wind farm, a steel tower had been built near one of wind turbines, and these grounding
foundations had been connected with insulating covered grounding wires. The grounding systems of
the steel tower had been connected to the grounding systems of neighboring wind turbines in the
Taikoyama wind farm.
Figures 4.75 and 4.76 show the grounding wires between the foundation of the wind turbine and the
steel tower, and the details of the wind turbine grounding system.
The grounding resistivity was measured by using the Wenner method. Figure 4.77 shows the layer of
grounding resistivity around the test site.
Figure 4.78 shows the experimental setup. The current was led to the foundation from the impulse
generator (IG) by using insulated copper wire (length: 100 m, cross-section: 5.5 mm2) as the current
lead wire. The height of the current lead wire was approximately 1 m. The fast-front current generated
by the IG was injected into the foundation through a resistance of 500 Ω from the current lead wire.

Tower

Wind turbine
Wind turbine
0.6 m
Soil

Lightning Wind turbine

protection tower Grounding
wires
32.7m

Wire 1 (32.7 m)
Soil
Wire 2 (41.4 m)
Insulating covered wire Wire 3 (32.7 m)

(a) (b)
Figure 4.75 Equalization of the wind turbine and the tall tower. (a) Tower and wind turbine and (b)
Equalization using grounding wires.117

117 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 55,
Figure 1, 2013.

134
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Grounding plate
Ring earth Grounding wire
electrode

Grounding plate

9.0 m

7.0 m
Connect with another 0.9 m × 0.9 m
wind turbine
Underground beam
width 0.7 m
16 m height 2.0 m

Mesh
grounding wire

4.25 m 5.0 m 4.25 m

Ring earth electrode

Grounding plate 2.5 m
0.9 m × 0.9 m 4.0 m Mesh grounding wire
1.0 m 1.0 m
1.6 m 2.5 m
2.0 m
0.8m 1.4 m
13.5 m 7.0 m

(a) (b)
Figure 4.76 Detail drawings of the foundations. (a) Wind turbine and (b) Tall tower.118

The peak value of current was 16 A and the wavefront was approximately 0.2 μs. The injected current
was measured at the end of the current lead wire near the foundation by using a current probe as
shown in Figure 4.78. The potential rise at the foundation of the wind turbine was measured as the
voltage difference between the top of the foundation and the voltage measurement wire. The height of
the wire was 1 m, and it was grounded at the remote end.
In this research, measurements were taken of the grounding characteristics in the cases of the wind
turbine and tower being connected and disconnected.

427 Wm 4.97 m

17.8 Wm ∞

Figure 4.77 Grounding resistivity.119

118 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 56,
Figure 2, 2013.
119 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 56,
Figure 3, 2013.

135
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Wind turbine

0.6 m Voltage measuring wire

Soil
76 m Foundation
of lightning protection tower Foundation
V of wind turbine
Lightning protection tower 76 m Grounding wires
IV cable V 76 m

I I
I
Voltage measuring wire
50 m I 76 m
　
500W
500W
　
Grounding wires 50 m
Current lead wire
100 m
50 m
I.G Current read wire
I.G
100 m 50 m

(a) (b)
Figure 4.78 Experimental setup. (a) Overall view and (b) Top view.120

In the FDTD analysis, the experimental layout was expressed as faithfully as possible by analysis
spaces as shown in Figure 4.79. The size of the analysis space was 166 m in the x direction, 141 m in
the y direction, and 147 m in the z direction, and grid spacing was all set to 0.5 m. Six faces that
surround the analytical space of the rectangular parallelepiped modeled as open spaces simulated
using the Liao’s second-order absorbing boundary condition. The ground in the experiment was
expressed by filling the space with the substance of resistivity ρ = 166 Ωm and relative permittivity 10
to 18 m high from the bottom of the analysis space, then, over the top of it, filling the space with the
substance of resistivity ρ = 427 Ωm and relative permittivity 10 to 5 m high. I-V lines used for current
lead and the voltage measuring wires were expressed by the thin wire model. Foundations were
simulated as faithfully as possible using perfect conductors. The injected current in the simulation was
set to be the same as that of the experiment result, which was a comparatively steep rising current of
16 A peak value and 0.2 µs wave-front and was injected into the top of the foundation via a 500-Ω
resistance. In order to identify the transient grounding characteristics of the wind turbine generator
system, current that flows into the foundation and potential rise of the foundation were simulated.
Potential rise of the foundation was simulated as a potential difference between the anchor top and
the voltage measuring wire.
Figure 4.80 shows a comparison of the measured and simulated injection current waveform. The
calculated injection current showed a step wave, and its peak and rise time were approximately 16A
and 0.2 μs, respectively. The waveform showed agreement with the experimental result.

120 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 56,
Figure 4, 2013.

136
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Voltage measurement wire 124 m

V
Current lead wire I Grounding wire
(48m)

Soil
5m
141 m 17.8 Wm 427 Wm 18 m

166 m
Grounding system of another
wind turbine
(100m× 20m× 3m)

Figure 4.79 FDTD simulation model.121

20
15
current [A]

Experiment
10 Simulation
5
0
0 5 10 15 20 25 30 35
time [s]

Figure 4.80 Injected current.122

Figure 4.81 shows a comparison of the experimental and analytical results of potential rise at the
foundation. In each case, the waveform exhibited inductive characteristics and analytical results
agreed roughly well with experimental results. As shown in Fig. 4.81, although the common grounding
wire does not affect the transient performance of the wind turbine primarily due to the inductance of
the wire, it can reduce the potential difference in the wave-tail, which is effective for reducing the total
energy following into or out of the wind turbine, causing the disturbances of electronic devices due to
the potential difference.
From these results, it was proven that FDTD simulation is useful when considering grounding
characteristics of the wind turbine grounding system.

4.6 Lightning surges in microwave relay stations

Microwave relay stations play an important role in transmitting information to control power grids and
maintain their stability. However, lightning strikes to a microwave relay station may result in faults,
malfunctions, or even physical damage to microwave radio equipment. To investigate the
effectiveness of lightning protection methodologies, the prediction of surge phenomena in a
microwave relay station is required. In [246], Tatematsu et al. first set up a reduced-scale model of a
microwave relay station composed of a microwave tower, a building with double-layered reinforcing

121 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 58,
Figure 7, 2013.
122 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 58,
Figure 8, 2013.

137
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

bars, a waveguide, a ring earth electrode, a deep earth electrode, and so forth, to simulate an actual
microwave relay station and measured the distribution of the transient currents flowing through the
100 100
80 80

voltage [V]
voltage [V] 60 60 Simulation
40 Simulation 40 Experiment
Experiment
20 20
0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
time [s] time [s]

(a) (b)
Figure 4.81 Comparisons of the potential rises at the foundation of a wind turbine. (a) Connected and (b)
Disconnected.123

microwave relay station model. Then, for validation purposes, they compared the measured results
with those obtained by three-dimensional FDTD-based calculations to simulate the above reduced-
scale model in detail by taking into account the configuration of the reinforcing bars, which is an
important factor for designing lightning protection methodologies for microwave relay stations.
Secondly, using the validated FDTD simulations, Tatematsu et al. analyzed the effects of the
reinforcing bars of the building, the ground wire of the waveguide, and the deep earth electrode on the
lightning current distribution in the case of a direct lightning strike to the microwave tower [246].
Figure 4.82 shows a reduced-scale model of a microwave relay station set up at the Akagi Testing
Center of CRIEPI. Two representative types of microwave relay stations, that with the microwave
tower on the building (type I-a) and that with the microwave tower on the ground next to the building
(type I-b), as respectively shown in Figures 4.82 (a) and (b), can be simulated using this reduced-
scale model. Both the building and the foundation for the microwave tower are made of reinforced
concrete with double-layered reinforcing bars. Hereafter, as shown in Figure 4.82, the foundation of
the building and the foundation for the microwave tower are referred to as foundations A and B,
respectively. The waveguide is fixed along a ladder mounted inside the microwave tower at some
points, where insulating rubber is inserted between the waveguide and the ladder. In practice, a
waveguide is electrically connected to a microwave tower through a metal frame of a parabolic
antenna. To simulate this condition, the polyethylene sheath of the waveguide is removed at its end
near the top of the microwave tower, and the copper tube of the waveguide is electrically connected to
the microwave tower. The deep earth electrode is located 2 m away from the ring earth electrode, and
it has an earth-resistance-reducing material around the cylindrical earth conductor as shown in Figure
4.82 (b).
Figure 4.83 shows the arrangement inside the building. Two rectangular metal boxes, a ground bus,
and some ground wires are installed in the building. One metal box (metal box A) represents the metal
frame of microwave radio equipment, and the waveguide drawn into the building is connected to the
top of metal box A. The other metal box (metal box B) represents the metal frame of a dehydrator,
which is installed to maintain a dry condition inside a waveguide. Metal box B is connected to the
waveguide via a copper tube with a diameter of 10 mm. As shown in Figure 4.83, inside the building,
the ground bus is placed at a height of 1.7 m. Metal boxes A and B are connected to the ground bus
via the same type of IV wire. An additional IV wire is placed between metal box A and the ring earth
electrode outside the building, and this wire simulates the ground wire of a DC power source cable
(grounded DC wire). When a lightning impulse current was injected into the top of the microwave
tower with an impulse generator, the current distribution in the microwave relay station model was
measured.
Figure 4.84 shows an overview of the microwave relay station model of type I-b simulated by the
FDTD method in detail based on the blueprint of the aforementioned model used for the
measurement. The microwave tower and the ladder mounted inside the microwave tower are
simulated by combining numerous small conductor plates. The circular structure at the top of the
microwave tower, waveguide, earth electrode, and ground wires are represented by thin wires. Each

123 Reprinted with permission from N. Yoshikawa, A. Ametani, and K. Yamamoto, “A Study on Grounding Characteristics of a
Wind Turbine System”, in Proc. the 6th International Symposium on EMC and Transients in Infrastructures, no. ISS-20, p. 58,
Figure 9, 2013.

138
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

of the double-layered reinforcing bars of foundation B, the columns, the walls, the ceiling, and
foundation A is simulated by a thin wire as exactly as possible in accordance with the blueprint. The
deep earth electrode is represented by a rectangular conductor. Inside the building, metal boxes A
and B are simulated by
Current injection wire
Image

Waveguide Current injection

wire
10 m
Earth-resistance-
reducing material
Microwave tower
Depth : 0.4 m Waveguide
Reinforced concrete 0.5 m Microwave tower
building Bare copper
wire Reinforced concrete
building
2m 65 m
Ground level 3.5 m 3.5 m Ring earth electrode Ring earth electrode
： depth : 0.4 m) electrode

4.5 m
9m
Reinforced concrete φ 66 mm Reinforced concrete foundation
Deep earth Reinforced concrete foundation Deep earth of the building (foundation A)
foundation for the electrode of the building (foundation A)
microwave tower electrode Reinforced concrete foundation for
(foundation B) the microwave tower (foundation B)

(a) (b)
Figure 4.82 Reduced-scale model of a microwave relay station. (a) Microwave tower on a building (type I-
a) and (b) Microwave tower next to a building (type I-b).124

Double-layered reinforcing bar Ground bus Waveguide About

10 cm
Microwave radio
1.7 m Layout of the
equipment (metal P-3 P-2 P-1 copper tube
2m box A)
Ground wire to Ground wire of
represent a DC power waveguide
source cable Dehydrator
0.4 m Foundation (metal box B)
Ring earth electrode
Figure 4.83 Arrangement inside the building.125

rectangular conductors. The ground bus, the grounded DC wire, and the ground wires between the
metal boxes and the ground bus are modeled by thin wires. The thin wires are represented using thin-
wire representation techniques developed to apply the FDTD method to surge analyses [44], [45].
Figure 4.85 shows the analysis space simulated by the FDTD method. The resistivity of the soil is set
to values estimated by comparing measured results of the ground potential rises of the ground
structure with results calculated by the FDTD method based on the resistivity distribution estimated by
the four-point Wenner test. The relative permittivity of the soil is set uniformly to 30, as obtained in a
previous study [162]. To assume an open space, all the external surfaces of the analysis space are
treated as the absorbing boundaries of Liao’s formulation of the second order [37]. The impulse
generator is simulated by a lumped voltage source, and a thin wire representing the current injection
wire is connected to the top of the microwave tower and the voltage source. The other end of the
voltage source is connected to a ground structure inside the soil.

124 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1617, Figure 1, 2015.
125Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a
microwave relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1617, Figure 2, 2015.

139
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

To reduce the required calculation time and memory capacity, the analysis space is divided into
nonuniform cells, and the cell sizes are varied arbitrarily. The analysis space is divided into 800 × 564
× 837 cells, and in the rectangular area that encompasses the microwave relay station model except
Inside the building
Waveguide

Microwave
Ring ground bus Copper tower
tube
Metal box A

Metal box B

Grounded DC wire

Reinforcing
bar
Ladder

Deep earth Reinforcing

electrode bar
Waveguide
Ring earth
electrode

Figure 4.84 Calculation model of the microwave relay station model.126

406.5 m 675 m
212 m

Microwave relay
station model 272.5 m Current injection wire

200 m Ground surface

V
200 m 140 Ω m
Voltage source
266.3 m

200 m Depth: 2.2 m

Deep earth electrode 1144 Ω m
Ground surface
Depth: 14.3 m
263 Ω m
z y
x Bottom of the
analysis space

Figure 4.85 Analysis space simulated by the FDTD method.127

for the deep earth electrode and has surfaces 0.25 m away from it, the size of the cells is kept at 0.025
m. The size of the ten cells adjacent to the current injection wire in its radial direction is also set to
0.025 m. On the other hand, the sizes of the other cells are gradually increased further away from the

126 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1620, Figure 8, 2015.
127 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1620, Figure 9, 2015.

140
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

calculation model. As a result, in the analysis space, the size of the cells ranges from 0.025 m to 2 m.
The time discretization is set to 0.024 ns, which is half of the limit given by Courant’s condition, to
avoid the numerical instability involved in the thin-wire representation technique.
Using the above-described measurement and calculation arrangements, Tatematsu et al. measured
and calculated the distribution of the currents flowing through the microwave relay station model for
three cases, which are referred to as cases A-1, A-2, and A-3, respectively [246]. In case A, the
microwave tower is mounted on the roof of the building, and the reinforcing bars are employed for
grounding. In cases A-2 and A-3, the microwave tower is mounted on foundation B next to the building
in type I-b, and the reinforcing bars are and are not employed for grounding in cases A-2 and A-3,
respectively. As an example of the calculated results, Figure 4.86 shows the distribution of the peak
values of the currents in case A-1, where the microwave tower is mounted on the roof of the building.
Figure 4.86 shows the peak values of the currents when the peak value of the current injected into the
microwave tower is normalized to 100 %. Figures 4.87 (a) and (b) respectively show the waveforms of
the current injected into the microwave tower and the current flowing into metal box A, representing
the metal frame of microwave radio equipment, through the waveguide in case A-1.

(Side view) Calculated Measured

100.0 / 100.0
Microwave result [%] result [%]
tower －：No measured results

2.7 / 2.9

1.4 / 2.6 3.7 / 3.9

2.7 / 2.9 Current flowing into

1.8 / － metal box A, which
Ring earth 0.05 / − 0.4 represents microwave
electrode radio equipment,
through the waveguide

(Top view) Ring earth electrode

9.2 / 8.2 14.8 / 12.6

0.08 / －
0.63 / －
Ring ground bus
Metal box A

4.1 / 5.3

0.06 / － 0.58 / －

11.1 / 10.7 46.6 / 43.6 71.3 / 70.6

Deep earth electrode

Figure 4.86 Comparison of the measured and calculated peak values of the currents in case A-1.128

(a) (b)

128 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1621, Figure 10, 2015.

141
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.87 Waveforms of the currents in case A-1. (a) Injected current and (b) Current flowing into metal
box A.129

Although the current flowing into metal box A, representing microwave radio equipment, is below 10 %
of the injected current, the calculated results of the current there agree well with the measured results
with differences of less than 1 %. In addition, Tatematsu et al. found good agreement between the
calculated and measured distributions of the currents in the microwave relay station model in cases A-
2, and A-3 [246]. From these results, the applicability of the FDTD method to the calculation of the
distribution of a lightning current through a microwave relay station was confirmed. The reason for
some disagreement between the calculated and measured results is considered to be that the
electrical parameters of the soil simulated in the FDTD calculations are not completely identical to the
actual ones. However, the effects of the arrangement of the microwave tower, the application of the
reinforcing bars of the building to the grounding, the ground wire of the waveguide, and the deep earth
electrode on the distribution of the lighting current in the FDTD calculations have similar tendencies to
those observed in the measured results.
The above-calculated and measured results indicate the following; (i) In cases A-1 through A-3, the
current flowing into metal box A decreases as a result of the copper tube connected to metal box B,
representing a grounded dehydrator. This confirms the effectiveness of grounding a dehydrator; (ii)
Compared with case A-3, the current flowing into the building is significantly reduced by grounding the
waveguide with the reinforcing bars of the wall in case A-2. This confirms the effectiveness of using
the reinforcing bars for grounding; (iii) In contrast to cases A-1 and A-2, in case A-3 the ground wire of
the waveguide outside the building has a smaller effect on the current flowing into the building through
the waveguide. This point is studied further in [246], as mentioned below.
Further, using the FDTD method, Tatematsu et al. simulated a microwave relay station struck by
lightning and analyzed the effect of the reinforcing bars, the ground wire of the waveguide outside the
building when the reinforcing bars of the building are not employed for grounding, and the deep earth
electrode, for both first- and subsequent-stroke currents [246]. To simulate a direct lightning strike to a
microwave tower, the vertical transmission line model [17] is attached to the top of the microwave
tower, while the upper end of the transmission line model is connected to the upper surface of the
analysis space. Here, the current waveform at the bottom of the transmission line model is expressed
by Heidler’s equation as follows:
i (0, t )  ( I max / k )(t /  1 )10 / {1  (t /  1 )10 }exp(t /  2 ) Equation 4.8

which is specified in IEC Standard 62305, part 1, regarding the lightning protection. The constants in
Equation (4.8) are set to represent first and subsequent stroke currents, where the lightning protection
level is assumed to be 1. The return stroke current along the lightning channel is given by i(z, t)=i{0, t –
(z – h)/v}, where h and v are the height of the channel base and the return stroke speed, respectively.
It is well known that the propagation speed of the return stroke current is slower than the speed of
light, and it is set to 100 m/μs, which is a typical value.
Although the effect of the reinforcing bars, the ground wire of the waveguide, and the deep earth
electrode is analyzed [246], an overview of the analysis of the ground wire of the waveguide is
mentioned here.
Although it is well known that a ground wire should be kept as short as possible to reduce an adverse
effect due to its inductance, as shown in [246], when the microwave tower is mounted on foundation B
next to the building, most of the current injected into the microwave tower flows into the ring earth
electrode through the tower feet and foundation B, and a small fraction of the current flows into the
building through the waveguide and into the ring earth electrode through the ground wires inside the
building. As a result, on account of the unbalance of the transient ground potential rises in the ring
earth electrode, the current through the waveguide cannot significantly flow into the ground wire of the
waveguide, which is connected to the earth electrode close to the microwave tower.
To confirm this, the arrangement of the ground wire of the waveguide is changed as shown in Figure
4.88. In cases C-1 and C-2, contradictory to the basic concept, the ground wire of the waveguide is
not connected to the earth electrode under it and its length is greater than case B-4 where the
waveguide is connected to the ring earth electrode by a shorter wire below the waveguide. Table 4.11

129 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1621, Figure 11, 2015.

142
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

shows the peak values of the calculated currents flowing through positions P-1 through P-3. In case B-
4, about 30 % of the current through position P-1 flows into the ground wire of the waveguide for both
the first and subsequent strokes. On the other hand, as shown in Table 4.11, although the ground wire
is longer than that in case B-4, the currents flowing into the building through the waveguide are
reduced by the ground wire of the waveguide by 51 % and 58 % in cases C-1 and C-2, respectively.
These results indicate that the basic concept of keeping ground wires as short as possible is not
necessarily valid and that the prediction of surge phenomena is required to design grounding systems
effectively.
Waveguide Waveguide

Ring earth Ground wire of Ring earth Ground wire of

electrode the waveguide electrode the waveguide

(a) (b)
Figure 4.88 Route of the ground wire of the waveguide outside the building. (a) Case C-1 and (b) Case C-
2.130

Table 4.11 Peak values of the currents in cases B-4, C-1, and C-2.131
Peak values of the currents [%]
Position First stroke Subsequent stroke
Case B-4 Case C-1 Case C-2 Case B-4 Case C-1 Case C-2
P-1 7.6 8.2 8.6 7.5 8.1 8.5
P-2 5.3 4.0 3.6 5.3 4.0 3.6
P-3 4.0 3.0 2.8 3.9 3.0 2.8

4.7 Lightning electromagnetic pulses

4.7.1 FDTD study of errors in magnetic direction finding of lightning due to the
presence of conducting structure near the field measuring station
4.7.1.1 Introduction
Two vertical and orthogonal loops, each measuring the magnetic field from a given vertical radiator,
can be used to obtain the direction to the source; that is, as a magnetic field direction finder (DF) (see,
for example, Ch. 17 of [28]). This is the case because the output voltage of a given loop is proportional
to the cosine of the angle between the magnetic field vector and the normal vector to the plane of the
loop. Therefore, the ratio of the two signals from the loops is proportional to the tangent of the angle
between the normal vector to the plane of one of the loops and direction to the source.
In multiple-station lightning locating systems (LLSs), such as the U.S. National Lightning Detection
Network (NLDN), the direction to lightning is estimated on the basis of the ratio of the peaks of
magnetic fields measured by two orthogonal loop antennas. A combination of magnetic direction
finding and time-of-arrival (TOA) techniques is used to obtain lightning location.
There are often mountains and tall grounded structures near magnetic direction finding stations in
countries like Japan. As a result, the direction (azimuth) estimated from two orthogonal magnetic fields
may be influenced by the presence of mountains and/or grounded structures. Errors in azimuth due to
the presence of conducting structures near the direction finding station are referred to as site errors.

130 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1623, Figure 17, 2015.
131 Reprinted with permission from A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave
relay station using the FDTD method,” IEEE Trans. EMC, vol. 57, no. 6, p. 1624, Table 6, 2015.

143
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

In [247], lightning electromagnetic fields in the presence of conducting (grounded) structure having a
height of 60 m and a square cross-section of 40 m × 40 m within about 100 m of the field measuring
point are calculated using the FDTD method [16]. Influence of the presence of grounded structure on
lightning magnetic fields at the observation point is studied, and the direction (azimuth) errors are
discussed. Influences of ground conductivity and lightning current waveshape parameters are also
studied.

4.7.1.2 Methodology
Figure 4.89 (a) shows the FDTD simulation model of a vertical lightning channel above flat ground.
The working volume of 12 km × 12 km × 11 km is divided uniformly into 20 m × 20 m × 20 m cubic
cells. The time increment is set to 38.1 ns, which fulfills the Courant stability condition. The ground
thickness (between the ground surface and the bottom absorbing boundary condition plane) is set to 1
km, and the ground conductivity is set to a value in the range from 0.1 to 1000 mS/m or to ∞ (perfect
conductor). The xz plane and the yz plane passing through the source (see Figure 4.89) are set to be
magnetic walls, and other boundaries are each represented by Liao’s second-order absorbing
boundary condition [37]. Magnetic wall in the FDTD simulation is represented by forcing the tangential
magnetic field components on the wall surface to zero. This makes the working volume shown in
Figure 4.89 (a) be equivalent to that shown in Figure 4.89 (b). The lightning channel is represented by
a vertical phased-current-source array [17], which is located at (x, y)=(0, 0) of the working volume. The
array is activated so as to simulate the current distribution predicted by the transmission-line model [1].
The return-stroke current waveform is represented using the sum of the Heidler function [64] and a
double-exponential function, given as follows:

 t 1  exp   t   I exp   t   exp   t  

2
I 01
I  0, t     02      Equation 4.9
  t  1 2  1  2    3    4 

Magnetic walls

12 km
12 km

11 km
B A

A C D
y y
z z
Observation point
x x

(a) (b)
Figure 4.89 FDTD simulation model of a vertical lightning channel above flat, perfectly-conducting
ground. (a) Single quadrant model with two magnetic walls and (b) its equivalent four-quadrant model.132

This function reproduces the observed concave rising portion of a typical lightning current waveform.
In the present simulations, I01=11 kA, I02 = 7.5 kA, τ2 = 5.0 μs, τ3 = 100 μs, and τ4 = 6.0 μs. For
representing 10-to-90% risetime of 0.5, 1, and 3 μs, τ1 = 0.305, 0.75, and 4.026 μs, and η = 0.836,
0.717, and 0.34585, respectively. The return-stroke speed is set to 1.5 ×108 m/s.

132 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 3, Figure 1, 2016.

144
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.90 shows a plan view of the configuration used for studying fields in the presence of a
conducting (grounded) structure having a cross-section of 40 m × 40 m. The point at which the
magnetic field is observed is located at (x,y)=(7080 m, 7080 m), which is 10012 m (10.012 km) away
from the lightning channel and the azimuth angle is 45° with respect to the x-axis (and y-axis) in the xy
plane. Also, computations were carried out for observation points at (x,y)=(8960 m, 4480 m) and
(10000 m, 0 m), which are 10017 m and 10000 m, respectively, away from the lightning channel and
the azimuth angles are 26.6° (tan-1ϕ = 0.5) and 0° with respect to the x-axis and in the xy plane. Note
that the azimuth vector passes through the grounded structure diagonally for ϕ = 45°, parallel to its
walls for ϕ = 0° while it passes through the structure neither diagonally nor in parallel to its walls for ϕ =
26.6°.

Cell size
Observation point
Δx= Δy=Δz=20 m
West Grounded structure
45° Height : 60 m
12 km
Plan-view area : 40 m×40 m
10012 m
Grounded structure
Ground conductivity
Lightning σ=∞
channel
Return-stroke speed
y v=1.5×108 m/s
12 km
x

Figure 4.90 Plan view of the model for studying the influence of grounded structure (not to scale) on
azimuth measured with respect to West for the case of the true lightning azimuth ϕ = 45°.133

The grounded structure is represented by a perfectly-conducting rectangular parallelepiped of 40 m ×

40 m × 60 m, which is centered at different points near the magnetic-field observation point in the
quadrant A. Note that computations for structure conductivities of 0.1, 1, 10, 100 and 1000 mS/m were
also performed. The total number of positions of the grounded structure was 24, located within a
square area of 200 m × 200 m, at the center of which, (7080 m, 7080 m), the magnetic-field
observation point was located for the case of azimuth angle equal to 45°. Identical grounded
structures are present in quadrants B, C, and D, because of the symmetry due to the employed
magnetic walls, as shown in Figure 4.89 (b). Since the structures in quadrants B, C, and D are far
away from those in quadrant A or the observation point, their influence on fields calculated in quadrant
A is small.
The direction error Δϕ in degrees for the true azimuth angle ϕ equal to 45° is calculated from Equation
(4.10) given below:
H 
 []  tan -1  x   45
H
 y  Equation 4.10
where Hx and Hy are peak values of magnetic field in two orthogonal directions (East-West and South-
North, respectively) at the observation point. The first term is the azimuth found from the ratio of Hx to
Hy, and the second term is the true azimuth for the configuration shown in Figure 4.90. Figures 4.91
(a) and (b) illustrate two cases with ϕ = 0°and ϕ = 16°, respectively, for ϕ = 45°. Note that magnetic
fields in the West direction (negative x direction) and in the North direction (positive y direction) are
each defined as positive.

133 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 3, Figure 2, 2016.

145
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.7.1.3 Analysis and results

4.7.1.3.1 Influence of the presence of grounded structure
Table 4.12 shows FDTD-computed peak values of Hx and Hy and the azimuth errors in degrees due to
the presence of a nearby grounded structure having a height of 60 m and a square cross-section of 40
m × 40 m estimated from the ratio of Hx to Hy peaks. The results are computed for a current risetime
equal to 1 μs. In this computation, the ground conductivity is set to ∞. Error in the counterclockwise
direction relative to the correct azimuth of ϕ = 45° (see Figure 4.90) is defined as positive. Figure 4.92
shows the color-coded plot of azimuth error due to the presence of nearby grounded structure. In
Figure 4.92, the darker the shading, the larger the error.
When the grounded structure is symmetrically located on the straight line that connects the lightning
channel with the observation point (in other words, the azimuth vector passes through the grounded
structure diagonally), its presence equally influences Hx and Hy. Therefore, no azimuth error occurs.
H
Hy
90° 90°
 H Hy

West Hx West Hx 
Observation Observation
 point 45°  point

True direction Direction

Direction to lightning found from
to lightning Hx and Hy

Lightning channel Lightning channel
H  H 
  tan  x   45
1
  tan 1  x   45
 Hy   Hy 
  0   0

(a) (b)
Figure 4.91 Illustration of lightning direction finding from Hx and Hy for the case of the true lightning
azimuth ϕ = 45°. (a) Ideal case of no azimuth error (direction found from Hx and Hy is the same as the true
direction to lightning) and (b) The case of the presence of 60-m tall grounded structure at (7080, 7040)
(see Table 4.12 and Figure 4.92), which causes an error Δϕ =16°.134

Table 4.12 FDTD-computed peak values of Hx and Hy, and azimuth errors estimated from Equation (4.10)
for the case of true azimuth ϕ = 45° and current risetime equal to 1 μs.135
Position of structure center [m] H x [mA/m] H y [mA/m]  [degrees] Position of structure center [m] H x [mA/m] H y [mA/m]  [degrees]
(7000, 7000) 66 66 0 (7080, 7120) 80 38 20
(7000, 7040) 64 69 -2 (7080, 7160) 69 58 5
(7000, 7080) 58 68 -5 (7120, 7000) 62 59 2
(7000, 7120) 59 62 -2 (7120, 7040) 53 54 -1
(7000, 7160) 62 62 0 (7120, 7080) 38 80 -20
(7040, 7000) 69 64 2 (7120, 7120) 74 74 0
(7040, 7040) 74 74 0 (7120, 7160) 70 64 2
(7040, 7080) 38 76 -18 (7160, 7000) 62 62 0
(7040, 7120) 54 53 1 (7160, 7040) 59 63 -2
(7040, 7160) 63 59 2 (7160, 7080) 58 69 -5
(7080, 7000) 68 58 5 (7160, 7120) 64 70 -2
(7080, 7040) 76 38 18 (7160, 7160) 66 66 0

When the grounded structure is not located on that line, its influence is greater when it is located
closer to the observation point. If the distance from the structure to the observation point is the same,

134 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 4, Figure 3, 2016.
135 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 6, Table 1, 2016.

146
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

the structure behind the observation point is more significant than that in front of it. Note that the
azimuth error is zero when two identical structures are located symmetrically with respect to the
direction to lightning, for example at (7040, 7080) and at (7080, 7040), since positive and negative
errors compensate each other.

4.7.1.3.2 Influence of lightning current risetime

Figures 4.93 (a) and (b) show FDTD-computed plots of color-coded azimuth error for true azimuth ϕ
equal to 45° and current risetimes of 0.5 and 3 μs, respectively. In these computations, the ground
conductivity is set to ∞. It appears from comparison of Figures 4.92 and 4.93 (b) that the change from
1 to 3 μs in current risetime only slightly influences the azimuth error. When the current risetime is 0.5
μs, the azimuth error due to the presence of nearby grounded structure is larger, except for the cases
that the structure is located at (7040 m, 7120 m) or (7120 m, 7040 m).
[m]
7160

7120
20゜

7080 0

7040
-20゜
7000
y
7000 7040 7080 7120 7160 [m]
x

Figure 4.92 Color-coded azimuth error (in degrees) due to the presence of a nearby grounded structure
for the case of true azimuth ϕ equal to 45°. Each small square, except for the one in the center,
represents the position of grounded structure, with the total number of such positions being 24. ‘+’
indicates the location of observation point, and the arrow indicates the true direction to lightning. 136

Figure 4.93 Color-coded azimuth error (in degrees) due to the presence of a nearby grounded structure
having a height of 60 m and a square cross-section of 40 m × 40 m for the case of true azimuth ϕ equal to
45° and current risetimes equal to (a) 0.5 μs and (b) 3 μs.137

4.7.1.3.3 Influence of ground conductivity

Figure 4.94 shows FDTD-computed plots of color-coded azimuth error due to the presence of nearby
grounded structure for true azimuth ϕ equal to 45° and ground conductivity values equal to 0.1, 5 and
1000 mS/m. The structure conductivity is set to ∞. The current risetime is set to 1 μs. Figure 4.95
shows azimuth errors for the true azimuth ϕ equal to 45° as a function of ground conductivity when the

136 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 6, Figure 6, 2016.
137 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 8, Figure 7, 2016.

147
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

grounded structure is located at (7000 m, 7000 m), (7000 m, 7080 m), (7040 m, 7080 m), (7120 m,
7080 m), and (7160 m, 7080 m).
It appears from these results that the azimuth error (when it is non-zero) due to the presence of
nearby grounded structure decreases with decreasing ground conductivity.

4.7.1.3.4 Influence of grounded structure conductivity

Figure 4.96 shows FDTD-computed color-coded azimuth error due to the presence of nearby
grounded structure for true azimuth ϕ equal to 45° and different values of structure conductivity equal
to 0.1, 1, 10, 1000 mS/m. Note that the ground conductivity is set to ∞ and the current risetime is set
to 1 μs.
It appears from these results that the azimuth error decreases with decreasing structure conductivity.
This is because the induced current in the grounded structure and the resultant scattered field
decrease with decreasing structure conductivity.
[m] [m]
7160 7160

7120 7120

7080 7080

7040 7040

7000 7000
y y
7000 7040 7080 7120 7160 [m] 7000 7040 7080 7120 7160 [m]
x x
(a) (b)

[m]
7160

7120 20゜

7080
0

7040
-20゜
7000
y
7000 7040 7080 7120 7160 [m]
x (c)

Figure 4.94 Color-coded azimuth error due to the presence of a nearby grounded structure having a
height of 60 m and a square cross-section of 40 m × 40 m for the case of true azimuth ϕ equal to 45° and
ground conductivity values equal to (a) 0.1 mS/m, (b) 5 mS/m, and (c) 1000 mS/m.138

138 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 9, Figure 8, 2016.

148
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

0.5 1.5 2.5 3.5 4.5 5.5

5
◆(7000 m, 7000 m)
0.1 1 5 10 1000
0
＊(7000 m, 7080 m)

Azimuth error [゜]

-5
■(7160 m, 7080 m)

-10

▲(7040 m, 7080 m)
-15

-20
×(7120 m, 7080 m)

-25
Ground conductivity [mS/m]

Figure 4.95 Azimuth errors as a function of ground conductivity when the grounded structure is located
at (7000 m, 7000 m), (7000 m, 7080 m), (7040 m, 7080 m), (7120 m, 7080 m), and (7160 m, 7080 m).139

[m] [m]
7160 7160

7120 7120

7080 7080

7040 7040

7000 7000
y y
7000 7040 7080 7120 7160 [m] 7000 7040 7080 7120 7160 [m]
x (a) x (b)
[m] [m]
7160 7160

7120 7120

7080 7080

7040 7040

20゜
7000 7000

y y
0
7000 7040 7080 7120 7160 [m] 7000 7040 7080 7120 7160 [m]
x x (d)
(c) -20゜

Figure 4.96 Color-coded azimuth error due to the presence of a nearby grounded structure having a
height of 60 m and a square cross-section of 40 m × 40 m for true azimuth ϕ equal to 45° and different
values of structure conductivity equal to (a) 0.1 mS/m, (b) 1 mS/m, (c) 10 mS/m, and (d) 1000 mS/m.140

139 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 9, Figure 9, 2016.
140 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 10, Figure 10, 2016.

149
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.7.1.3.5 Influence of structure location relative to the azimuth vector

Figure 4.97 shows FDTD-computed color-coded azimuth error due to the presence of nearby
grounded structure for true azimuth ϕ equal to 26.6°. Note that the observation point at (x,y)=(8960 m,
4480 m), which is 10017 m away from the lightning channel and the azimuth angle is 26.6° (tan -1 ϕ =
0.5) with respect to the x-axis (-x direction) in the xy plane. The conductivity of both the structure and
ground is set to ∞. The current risetime is set to 1 μs.
It appears from these results that an azimuth error occurs even when the center of the grounded
structure is located on the straight line that connects the lightning channel and the observation point.
This is because the grounded structure having a square cross-section is not symmetrically located on
that line (in other words, the azimuth vector does not pass through the grounded structure diagonally)
as shown in Figure 4.98 (a). Note that for ϕ = 45° no azimuth error occurs since the azimuth vector
passes through the grounded structure diagonally, as shown in Figure 4.98 (b), so that the structure
equally influences Hx and Hy and Hx/Hy is the same as in the absence of structure. Also note that for ϕ
= 0° no azimuth error occurs since the azimuth vector is parallel to the walls of the structure as shown
in Figure 4.98 (c). This is confirmed by the FDTD simulation, although the results for are not shown
here.

4.7.1.4 Summary
Lightning electromagnetic fields in the presence of grounded structure having a height of 60 m and a
square cross-section of 40 m × 40 m within approximately 100 m of the field point have been analyzed
using the 3D FDTD method. Influence of the structure on the two orthogonal components of magnetic
field, Hx and Hy, has been analyzed, and the resultant errors in the estimated lightning azimuth (site
errors) has been studied. When the azimuth vector passes through the center of grounded structure
diagonally (azimuth angle is 45°) or parallel to its walls (azimuth angle is 0°), the presence of structure
equally influences Hx and Hy, so that Hx/Hy is the same as in the absence of structure. As a result, no
[m]
4560
4540

4520
25゜
4500
4480 0
4460
4440
-12゜
4420
4400
y

x 8880 8920 8960 9000 9040 [m]

Figure 4.97 Color-coded azimuth error due to the presence of a nearby grounded structure having a
height of 60 m and a square cross-section of 40 m × 40 m for the case of true azimuth ϕ equal to 26.6°
(tan-1ϕ= 0.5).141

141 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 11, Figure 11, 2016.

150
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Observation point
Observation  = 45°
point
 = 26.6°
Grounded
Grounded structure structure
Observation
y Grounded structure y y point
Lighting channel Lighting channel =0
Lighting channel

x x x

(a) (b) (c)

Figure 4.98 Illustration of locations of lightning channel, grounded structure, and observation point when
the center of the structure is located on the straight line that connects the lightning channel with the
observation point. (a) true azimuth ϕ is 26.6°, (b) true azimuth ϕ is 45°, and (c) true azimuth ϕ is 0°.142

azimuth error occurs in these conditions. When the structure is not located on that line, its influence is
greater when it is located closer to the observation point. If the distance from the structure to the
observation point is the same, the influence of structure behind the observation point is more
significant than that in front of it. The azimuth error due to the presence of structure depends on the
risetime of lightning current. When the current risetime is 0.5 μs, the azimuth error is generally larger
than those for risetimes equal to 1 or 3 μs, except for some specific locations. The azimuth error
decreases with decreasing ground conductivity and with decreasing conductivity of the structure.

4.7.2 FDTD modeling of LEMP propagation in the earth ionosphere waveguide

Lightning electromagnetic pulse (LEMP) propagation over flat finitely conducting ground have been
studied by Aoki et al. [248] using the FDTD method [16] in the 2D cylindrical coordinate system. The
influences of return-stroke wavefront speed, current risetime, and ground conductivity on electric and
magnetic fields were considered in that study. However, the presence of conducting atmosphere was
neglected.
Recently, ionospheric reflections in measured lightning electric field waveforms have been examined
by Somu et al. [249]. It has been found that the apparent ionospheric reflection height tends to
decrease with stroke order within the same flash. In order to explain these observations, modeling of
ionospheric reflections is needed.
In [250], vertical electric and azimuthal magnetic fields at distances of 50 to 500 km from the vertical
lightning return-stroke channel on flat perfectly conducting ground are computed using the 2D-FDTD
method. In the simulations, the conductivity of atmosphere up to 110 km above ground level is
assumed to increase exponentially with height. Both daytime and nighttime atmospheric conditions
were considered.
Figure 4.99 shows the configuration, including a vertical lightning channel on flat, perfectly conducting
ground, analyzed using the 2D-FDTD method. The thickness of ground is set to 10 km. The thickness
of atmosphere is set to 110 km, and its conductivity is assumed to increase exponentially with height.

142 Reprinted with permission from Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of
errors in magnetic direction finding of lightning due to the presence of conducting structure near the field measuring station,”
Atmosphere, 7, 92, p. 12, Figure 12, 2016.

151
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

1000 km

Conducting atmosphere

110 km

Lightning return-stroke
(MTLL model)

7 km

z
Flat ground 10 km
φ
r
d

Figure 4.99 Configuration (not to scale) used in the 2D-FDTD simulations.143

The lightning return stroke is represented by the modified transmission-line model with linear current
decay with height (MTLL model) [51], which is expressed by the Equation (4.11):

 z   z 
I ( z',t )  1   I  0, t   Equation 4.11
 H   
where I(z’,t) is a current at height z’ along the channel and time t, H is the channel length, and v is the
return-stroke wavefront speed.
The MTLL model is simulated in the FDTD computations by a vertical phased-current-source array
[17]. In the simulation, H is set to 7000 m, and ν is set to c/2. The channel-base current is represented
using Heidler's function [64].
The conductivity of atmosphere as a function of height h above ground is expressed as Equation
(4.12) [251], [252]:

 (h)   0 exp    h  h ' Equation 4.12

where σ0 is set to 2.22 μS/m, β is set to 0.67 (1/km), and reference height h’ is set to 84 km for
nighttime conditions, while for daytime conditions β is set to 0.40 (1/km) and h’ is set to 73 km.
For FDTD simulations, the 2D working space of 1000 km × 120 km is divided uniformly into 50 m ×
100 m rectangular cells. Liao’s second-order absorbing boundary condition [37] is applied to the
upper, bottom, and right-side boundaries to minimize unwanted reflections there. The vertical electric
field and azimuthal magnetic field are computed on the ground surface at distances ranging from 50 to
500 km.
Figure 4.100 shows the waveform of lightning return-stroke current at the channel base that was used
in the calculations. The peak of the current is 30 kA, and its zero-to-peak risetime is 10 μs. As noted
earlier, the lightning channel length is set to 7 km, and the return-stroke wavefront speed is set to c/2.

143 Reprinted with permission from T. H. Tran, Y. Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling of LEMP propagation in
the earth-ionosphere waveguide with emphasis on realistic representation of lightning source,” Journal of Geophysical
Research: Atmospheres, vol. 122, p. 12920, Figure 1, 2017.

152
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

35

30

25
I [kA]
20

15

10

0
0 10 20 30 40 50 60 70 80 90 100
Time [s]

Figure 4.100 Waveform of lightning return-stroke current with a peak of 30 kA, and risetime of 10 μs.144

Figure 4.101 shows waveforms of vertical electric field Ez at different distances from the lightning
channel, computed using the 2D-FDTD. Free space, daytime and nighttime atmospheric conditions
were considered in the simulations. The fields were computed at d ranging from 50 to 500 km. Figure
4.102 shows corresponding waveforms of azimuthal magnetic field H.
Ground waves and sky waves (when discernible) are labeled in Figure 4.101 and Figure 4.102. The
number of discernible skywaves is increasing with distance. As expected, the time delays of skywaves
relative to the ground wave are larger at night, which is indicative of a larger effective ionospheric
reflection height under nighttime conditions.
Table 4.13 and Table 4.14 summarize peak-to-peak time intervals between the ground wave and the
first two skywaves and corresponding ionospheric reflection heights computed following the approach
used in [249]. In the Tables, subscripts 1 and 2 correspond to the first and second skywaves,
respectively. It appears that within 200 km the estimated reflection heights are not far from the
reference height h’ (84 km for nighttime conditions and 73 km for daytime conditions) for both first and
second skywaves.

18 10

16 Ground wave Free space Ground wave Free space

Nighttime atmosphere 8 Nighttime atmosphere
14
Daytime atmosphere Daytime atmosphere
12
6
10
Ez [V/m]
Ez [V/m]

8 4

6 1st skywaves
1st skywaves
2
4

2
0
0
50 km 100 km
-2 -2
0.0001 0.0003 0.0005 0.0007 0.0009 0.0011 0.0003 0.0005 0.0007 0.0009 0.0011 0.0013
100 300 500 700 900 1100 300 500 700 900 1100 1300
Time [s] Time [s]
Time [s] Time [s]
(a)
(a) (b)
(b)

144 Reprinted with permission from T. H. Tran, Y. Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling of LEMP propagation in
the earth-ionosphere waveguide with emphasis on realistic representation of lightning source,” Journal of Geophysical
Research: Atmospheres, vol. 122, p. 12921, Figure 2, 2017.

153
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

6 2
Ground wave
Free space Free space
Ground wave Nighttime atmosphere 1st skywaves Nighttime atmosphere
4 Daytime atmosphere 1 Daytime atmosphere
2nd skywaves
3rd skywaves

Ez [V/m]
Ez [V/m]

1st skywaves
2 0
2nd skywaves

0 -1

200 km 500 km
-2 -2
0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0016 0.0018 0.002 0.0022 0.0024 0.0026
600 800 1000 1200 1400 1600 1600 1800 2000 2200 2400 2600
Time [s] Time [s]
Time [s] Time [s]
(c) (d)
(c) (d)
Figure 4.101 FDTD-computed waveforms of vertical electric field Ez (for three cases: free space, daytime
and nighttime conducting atmosphere) at (a) d = 50 km, (b) d = 100 km, (c) d = 200 km, and (d) d = 500 km.
The computations were performed for perfectly-conducting ground, and 30-kA-peak, 10-μs-risetime
lightning current.145

The vertical electric field Ez and azimuthal magnetic field H at different distances (50 to 500 km) from
the lightning channel have been computed using the 2D-FDTD method in the cylindrical coordinate
system. In the simulations, the presence of daytime and0.05
nighttime conducting atmosphere has been
considered. One-hop and two-hop skywaves have been identified in computed waveforms and used
Ground wave Free space
for the estimation of apparent ionospheric reflection heights.
0.04 It appears that within 200 km the
Nighttime atmosphere
estimated reflection heights are not far from the reference height h', 84 km for nighttime
Daytimeconditions
atmosphere and
73 km for daytime conditions, for both first and second0.03skywaves.

A full-wave, FDTD analysis of the electric field propagation including the effect of the ionospheric
H [A/m]

0.02
reflections was presented in [253]. The FDTD simulation results were compared with the measured
fields associated with upward flashes initiated from the Säntis Tower. The data consists of current and
1st skywaves
0.01
wideband electric field waveforms at 380 km associated with upward lightning flashes initiated from
the Säntis Tower, Switzerland. The dataset presented in their study includes the first simultaneous
0
records of lightning currents and associated fields featuring ionospheric reflections for natural upward
flashes, and the longest distance at which natural upward
-0.01
lightning fields have been measured50 km
simultaneously with their causative currents. It was found0.0001that the
100 300
model0.0005
0.0003 reproduces
500
0.0007 reasonably
700
0.0009
900
well
0.0011
1100
the measured waveforms and the times of arrival of the one-hop and two-hop skywaves
Time
Time [s]
[s] relative to the
groundwave. (a)
0.05 0.03

Ground wave Free space Ground wave Free space

0.04 Nighttime atmosphere Nighttime atmosphere
Daytime atmosphere 0.02 Daytime atmosphere

0.03
H [A/m]
H [A/m]

0.02 0.01 1st skywaves 2nd skywaves

1st skywaves
0.01
0
0

50 km 100 km
-0.01 -0.01
0.0001 0.0003 0.0005 0.0007 0.0009 0.0011 0.0003 0.0005 0.0007 0.0009 0.0011 0.0013
100 300 500 700 900 1100 300 500 700 900 1100 1300
Time [s]
[s] Time [s]
Time Time [s]

(a) (b)
0.03 0.015

Free space Ground wave Free space

Ground wave
145 Reprinted with permission from T. H. Tran, Y.atmosphere
Nighttime Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling ofNighttime
LEMPatmosphere
propagation in
the0.02
earth-ionosphere waveguide with emphasis on realistic representation
Daytime atmosphere 0.01 of lightning source,” Journal ofDaytime
Geophysical
atmosphere
Research: Atmospheres, vol. 122, p. 12922, Figures 3a, 3b, 3c, and 3f, 2017.
H [A/m]

1st skywaves
H [A/m]

0.01 1st skywaves 2nd skywaves 0.005

2nd skywaves

0 154 0
 TB
0 785 - Electromagnetic computation methods for lightning
0 surge studies with emphasis on the FDTD method

100 km 200 km
-0.01 -0.005
0.0003 0.0005 0.0007 0.0006 0.0008 0.001 0.0012 0.0014 0.0016
300 500 700 (a) 0.0009
900
0.0011
1100
0.0013
1300 600 800 (b) 1200
1000 Time 1400 1600
Time [s] [s]
Time [s] Time [s]
(b) (c)
0.015 0.006

Ground wave Free space Ground wave Free space

Nighttime atmosphere Nighttime atmosphere
0.004
0.01 Daytime atmosphere Daytime atmosphere
1st skywaves
2nd skywaves 3rd skywaves
0.002

H [A/m]
H [A/m]

1st skywaves
0.005
2nd skywaves
0

0
-0.002

200 km 500 km
-0.005 -0.004
0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0016 0.0018 0.002 0.0022 0.0024 0.0026
600 800 1000 Time [s] 1200 1400 1600 1600 1800 2000 2200 2400 2600
Time [s]
Time [s] Time [s]
(c) (d)
(c) (d)
0.006
Figure 4.102
Ground FDTD-computed
wave waveforms
Free space of azimuthal magnetic field H (for three cases: free space,
daytime and nighttime conducting atmosphere)
Nighttime atmosphere at (a) d = 50 km, (b) d = 100 km, (c) d = 200 km, and (d) d =
0.004
500 km. The1stcomputations
skywaves were performed
Daytime atmospherefor perfectly conducting ground, and 30-kA-peak, 10-μs-
risetime lightning current. At 50 km, 2nd skywaves are not seen in this Figure, but they are discernible in
2nd skywaves 3rd skywaves
0.002 the computed waveforms.146
H [A/m]

4.8 0 Others
4.8.1
-0.002 FDTD simulation of lightning current in a multilayer CFRP panel with triangular-
prism cells
500 km
4.8.1.1
-0.004
0.0016
Introduction
0.0018 0.002 0.0022 0.0024 0.0026
1600 1800 2000 2200 2400 2600
Time [s]
CFRP has been recentlyTime used
[s] as the main material of aircraft bodies and wings since it has low weight
and high strength compared (d) to aluminum alloy or other metallic material. In [254], Apra et al., using
the FDTD method [16], have analyzed transient magnetic fields and voltages induced on cables in an
aircraft struck by lightning. The aircraft, which is partially made of CFRP of conductivity 1.5 × 10 4 S/m,
is 22 m in length, 30 m in wing span, and 9 m in height. It is represented with cubic cells of 0.1 m × 0.1
m × 0.1 m. Perrin et al. have carried out a similar FDTD computation for estimating currents induced
on cables in an aircraft struck by lightning [255]. The aircraft, which is partially made of CFRP of
conductivity 1000 S/m, is represented with cubic cells of 50 mm × 50 mm × 50 mm. In analyzing
lightning-induced effects on electrical or electronic systems in an aircraft made of CFRP, it would be
sufficient to represent the aircraft body as a uniform bulk material as done in [254] and [255].
However,
Table 4.13 Electric field ionospheric reflection heights.147
Nighttime conditions

d (km) t1 (μs) h1 (km) t2 (μs) h2 (km)

50 402 82 936 82

100 323 85 826 83

200 219 87 651 85

500 128 100 398 91

146 Reprinted with permission from T. H. Tran, Y. Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling of LEMP propagation in
the earth-ionosphere waveguide with emphasis on realistic representation of lightning source,” Journal of Geophysical
Research: Atmospheres, vol. 122, p. 12923, Figures 4a, 4b, 4c, and 4f, 2017.
147 Reprinted with permission from T. H. Tran, Y. Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling of LEMP propagation in
the earth-ionosphere waveguide with emphasis on realistic representation of lightning source,” Journal of Geophysical
Research: Atmospheres, vol. 122, p. 12924, Table 1, 2017.

155
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Daytime conditions

d (km) t1 (μs) h1 (km) t2 (μs) h2 (km)

50 335 71 771 69

100 250 72 671 71

200 175 77 525 74

500 108 91 341 84

Table 4.14 Magnetic Field Ionospheric Reflection Heights.148

Nighttime conditions

d (km) t1 (μs) h1 (km) t2 (μs) h2 (km)

50 425 85 958 83

100 323 85 833 84

200 221 88 654 85

500 127 99 398 91

Daytime conditions

d (km) t1 (μs) h1 (km) t2 (μs) h2 (km)

50 338 71 801 71

100 252 72 679 72

200 177 78 523 74

500 109 91 337 83

in studying thermal sparks, which would occur on or in the CFRP material of the aircraft struck by
lightning and might cause fuel explosion, the layered structure of CFRP needs to be considered. This
is because thermal sparks might be generated by local concentration of lightning current near metallic
fasteners that connect aircraft body and/or wing components.
In [256], Hirano et al. have carried out an experiment for studying fracture of a multilayer CFRP panel
of 150 mm × 100 mm × 4.7 mm, to the surface of which a lightning impulse current of magnitude of 10
to 40 kA is injected. Each layer of the CFRP panel has a fiber direction of 45, 90, -45 or 0 degrees,
and therefore, the conductivity is high in fiber axial direction and low in other directions. Feraboli et al.
[257] and Zhou et al. [258] also performed similar experiments.
It is quite difficult to study the transient distribution of lightning current in the vicinity of a fastener of
multilayer CFRP panel using the FDTD method with common cubic or rectangular-parallelepiped cells
since they cannot allow us to represent a multilayer CFRP panel, each layer of which has a fiber
direction of 45, 90, -45 or 0 degrees. In this section, transient distribution of lightning current on and in
a multilayer CFRP panel is studied using the FDTD method with triangular-prism cells [259], [260].

4.8.1.2 Heat distribution over a multilayer CFRP panel

Figure 4.103 shows the color-indicated thermographic picture taken from above at time 200 ms
around a fastener of a multilayer CFRP panel in a given region when a 3-kA lightning impulse current,

148 Reprinted with permission from T. H. Tran, Y. Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling of LEMP propagation in
the earth-ionosphere waveguide with emphasis on realistic representation of lightning source,” Journal of Geophysical
Research: Atmospheres, vol. 122, p. 12925, Table 2, 2017.

156
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

specified as Component A in [261], is injected in the CFRP panel through the fastener. This
thermographic picture would probably indicate the total heat distribution (not only due to the surface
layer current). It appears from Figure 4.103 that most current flows in directions of carbon fibers (-45,
0, 45 and 90 degrees) connected directly or touched to the fastener since the higher-heated area
would indicate higher density of current. Also it appears that some injected current flows in lower or
deeper layers of the CFRP panel via the fastener. Since thermal sparks might be generated by local
concentration of lightning current near fasteners that connect aircraft body and/or wing components, it
is necessary to analyze the current distribution in the vicinity of fasteners of a multilayer CFRP panel.

Figure 4.103 Measured thermographic picture (observed from above) at time 200 ms of a multilayer CFRP
panel in the vicinity of a fastener, in which a 3-kA current of Component A specified in (SAE 2013) is
injected.149

4.8.1.3 FDTD modeling with triangular-prism cells

Since it is impossible to define 45- and -45-degree-directed (relative to the x or y axis) conductivity in a
common 3D-FDTD simulation with cubic or rectangular-parallelepiped (electric- and magnetic-field)
cells, triangular-prism electric-field cells (e.g., [262]) are employed for representing a multilayer CFRP
panel, each layer of which has a fiber direction of -45, 0, 45 or 90 degrees and therefore has an
anisotropic conductivity. Note that triangular-prism electric-field cells are used only for representing a
horizontally-placed multilayer CFRP panel and the spaces right above and below the CFRP panel.
The rest of the working volume is divided into rectangular-parallelepiped cells. This allows the use of a
common absorbing boundary condition, which truncates the working space but suppresses unwanted
reflections there for simulating an unbounded space, without any modification. Higdon second-order
absorbing boundary condition [263] is employed in the present simulation.
Figure 4.104 shows cross-sectional view (x-y plane) of electric-field and magnetic-field grids. When a
triangular-prism electric-field cell is employed, magnetic-field cells have square and octagonal face
shapes. Note that at the boundary between triangular-prism electric-field cells and rectangular-
parallelepiped electric-field cells, hexagonal and pentagonal integral paths are used.

149 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 327, Figure 1, 2016.

157
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.104 Cross-sectional (x-y) views of electric-field grid and magnetic-field grid when triangular-
prims electric-field cells are used.150

Update equations for magnetic field H and electric field E are given below.
t Ne
H n 1 / 2  H n 1 / 2 
  Se
E
i 1
i
n
 lei Equation 4.13

1    t / 2 n t /  1 Nh
E n 1 
1    t / 2
E  
1    t / 2 S h
H
i 1
i
n 1/ 2
 lhi Equation 4.14

where n is the time step number, Δt is the time increment, μ is the permeability, σ is the conductivity, ε
is the permittivity, Ne is the side number of an electric-field polygonal loop, Δlei is a side length of the
electric-field polygon, ΔSe is the area of the electric-field polygon, Nh is the side number of a magnetic-
field polygon, Δlhi is a side length of the magnetic-field polygon, ΔSh is the area of the magnetic-field
polygon.

4.8.1.4 Comparison with an experiment

A CFRP panel of 300 mm × 300 mm, composed of 16, 24 or 32 thin plies (layers), was used for testing
the validity of the FDTD simulation. The thickness of each layer was about 0.2 mm. A metallic fastener
(like a peg or nail) was used for the current injection point and another metallic fastener was used for
output terminal point. The waveform of the injected current was Component A specified in [261], which
had a magnitude of 3 kA. Figure 4.105 shows the location of magnetic-field sensors (Lightning
Technologies, NTS Company) installed 5 mm above the surface of the CFRP panel.
Figure 4.106 shows the model for the FDTD simulation. The working volume is set to 600 mm × 600
mm × 200 mm with triangular-prism cells. The CFRP panel and the spaces right above and below the
panel are divided into triangular-prism cells, each of which has an isosceles triangle cross-section of
5√2 mm in equilateral side and 10 mm in base and has a height of 2 mm. The rest of the space is
divided into rectangular-prism cells of 10 mm × 10 mm × 2 mm. The conductivity of each layer of the
CFRP panel in the fiber direction is set to 1000 S/m, the conductivity in the horizontally perpendicular
directions is set to 0.2 S/m, and the conductivity in the vertical direction is set to 0.2 S/m. Although the
actual thickness of each layer is 0.2 mm, the thickness of each layer is set to 2 mm (10 times larger).
This is because the use of smaller cell, for example, a 1 mm × 1 mm × 0.2 mm cell, for representing
0.2-mm-thick layers would require a greater amount of total cells, and correspondingly a greater

150 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 328, Figure 2, 2016.

158
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

amount of computation time and memory. The total thickness is 30, 46 or 62 mm for 16, 24 or 32-ply
CFRP panel in the FDTD simulations. To compensate this vertically expanded FDTD model, the

Figure 4.105 Plan view of the surface of a 300 mm × 300 mm CFRP panel and magnetic field sensors
installed on the CFRP panel.151

Figure 4.106 FDTD simulation model of the 300 mm × 300 mm CFRP panel with a thickness of 30, 46 or 62
mm, represented using triangular-prism cells.152

conductivities of each layer in horizontal directions are set to one-tenth (1000 S/m, 0.2 S/m) of the
corresponding values (10000 S/m, 2 S/m) of the actual CFRP panel used in the experiment.
Figures 4.107 (a), (b) and (c) show FDTD-computed waveforms of magnetic field on the surface of the
16-layer CFRP panel at a horizontal distance of 80 mm from the left-side edge (current injection side)
and at a distance of 0, 40 and 80 mm, respectively, from the straight line connecting the current
injection and output points, and the corresponding measured waveforms. Figures 4.108 and 4.109
show those for 24- and 32-layer CFRP panels, respectively. It appears from Figures 4.107 to 4.109
that the peak of FDTD-computed magnetic field decreases with an increasing number of CFRP layers
(or the total thickness), while the measured results are less affected by the thickness. The increase of
the CFRP panel thickness due to the increase of the total number of layers from 16 to 32 in the FDTD
model is 32 mm, while it is 3.2 mm in the experiment. The former is about 6 times larger than the
distance (5 mm) between the magnetic-field observation point and the CFRP panel surface, while the

151 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 328, Figure 3, 2016.
152 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 328, Figure 4, 2016.

159
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

latter is only about 3/5 of it. This difference would be one of the main reasons for the above difference
in the tendency.

(a) (b) (c)

Figure 4.107 FDTD-computed waveforms (dotted lines) of magnetic field in kA/m on the surface of the 16-
layer CFRP panel at a horizontal distance of 80 mm from the left-side edge (current injection side) and at
a distance of 0, 40 and 80 mm from the straight line connecting the current injection and output points,
and the corresponding measured waveforms (solid lines). (a) 0 mm, (b) 40 mm, and (c) 80 mm.153

(a) (b) (c)

Figure 4.108 Same as Figure 4.7.5 but for the 24-layer CFRP panel. (a) 0 mm, (b) 40 mm, and (c) 80 mm.154

(a) (b) (c)

Figure 4.109 Same as Figure 4.7.5 but for the 32-layer CFRP panel. (a) 0 mm, (b) 40 mm, and (c) 80 mm.155

Figure 4.110 (a) shows FDTD-computed peak values of magnetic field on the surface of the 16-layer
CFRP panel at a distance of 80 mm from the left-side edge (current injection side) and at a distance of
0 to 100 mm from the straight line connecting the current injection and output points, and the
corresponding measured values. FDTD simulations are carried out for three sets of conductivity
shown in Table 4.15. Note that Figures 4.107 to 4.109 are computed for Case 1. Figures 4.110 (b) and
(c) show those for the 24- and 32-layer CFRP panels. It appears from Figures 4.107 to 4.110 that
FDTD-computed magnetic fields for Case 1 agree with the corresponding measured values better
than those computed for Cases 2 or 3 do. Also, it follows from these results that the magnetic field
intensity (or current density) decreases with increasing distance from the center line.
Note that, for Case 3 (isotropic conductivity), the FDTD computation has also been done with
rectangular cells of 10 mm × 10 mm × 2mm. The results computed with triangular-prism cells, each of
which has an isosceles triangle cross-section of 5√2 mm in equilateral side and 10 mm in base and

153 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 329, Figure 5, 2016.
154 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 329, Figure 6, 2016.
155 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 329, Figure 7, 2016.

160
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

has a height of 2 mm, agree reasonably well with those computed with the rectangular cells (not
shown here). The memory capacity and computation time required for one simulation with triangular-
prism cells (with a time increment of 5.21 ps and the maximum observation time of 50 μs) are about

(a)

(b)

(c)
Figure 4.110 FDTD-computed peak values of magnetic field on the surface of the 16-, 24- and 32-layer
CFRP panels at a distance of 80 mm from the left-side edge and at a distance of 0 to 100 mm from the

161
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

straight line connecting the current injection and output points, and the corresponding measured values.
(a) 16-layer CFRP panel, (b) 24-layer CFRP panel, and (c) 32-layer CFRP panel.156

Table 4.15 Three sets of conductivity of each layer in the fiber direction, horizontally perpendicular
direction (to the fiber), and in the vertical direction of the CFRP panel.157

0.9 GB and 60 hours (when a 3-GHz PC is used), respectively, while those with rectangular cells (with
a time increment of 6.36 ps) are about 0.8 GB and about 20 hours, respectively.
Figure 4.111 shows distributions of time integration of the square of FDTD-computed current of the top
four layers for the first 10 μs. Figure 4.112 shows the superimposed distribution of time integration of
the square of FDTD-computed current of the top four layers for the first 10 μs. Since the heat
conductivity is low in the vertical direction, only the heat generated in the top several layers would
appear on the surface layer. The FDTD-computed distribution (near the fastener) of the time
integration of the square of current shown in Figure 4.112 agrees qualitatively with the measured heat
distribution shown in Figure 4.103.

4.8.1.5 Summary
Transient distributions of magnetic field on the 16-, 24- or 32- layer CFRP panel, each layer of which
has anisotropic conductivity, have been studied using the FDTD method with triangular-prism cells.
The computed magnetic-field distributions agree well with the corresponding measured values. This
shows that the present FDTD simulation is valid and effective in analyzing lighting current in a CFRP
aircraft.

4.8.2 FDTD analysis of the electric field of a substation arrester under a lightning
overvoltage
4.8.2.1 Introduction
Zinc-oxide (ZnO) arresters are installed in power substations to protect their high-voltage equipment
against overvoltages. Since a substation arrester usually has a length of meters, the transient electric
field distribution in and around the arrester might be complex when a lightning overvoltage is imposed.
The occurrence of locally intense electric fields might damage the ZnO material or insulation material.
The FDTD method [16] for solving Maxwell’s equations has been used frequently in analyzing
lightning electromagnetic fields and surges (e.g., [18]). In FDTD simulations of surges in power
distribution lines, a small surge arrester has been modeled as one cell or one side of a cell [68], [206],
which has a piecewise-linear approximated characteristic of voltage versus current in a specified
direction.
In [264], the electric field in and around a 2.6-m long ZnO arrester, which simulates a 200- to 300-kV
class substation arrester with a 1000-kV lightning impulse imposed, is analyzed using the FDTD
method. The substation arrester is represented as a combination of 10 mm × 10 mm × 10 mm cubic
cells, each of which has a nonlinear resistivity (or conductivity) in each of the x, y, and z directions,
dependent on the electric field in each direction.

156 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 330, Figure 8, 2016.
157 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 329, Table 1, 2016.

162
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

4.8.2.2 Model: Representation of the nonlinear characteristic of a cell

In order to obtain the nonlinear voltage-current characteristic (V-I characteristic) of surge-arrester
material, an experiment has been carried out with a ZnO element having a thickness of 36 mm and a
radius of 21 mm.

(a) (b)

(c) (d)

Figure 4.111 Distributions of time integration of the square of FDTD-computed current of the top four
layers for the first 10 μs. (a) 45-degree layer (top), (b) 90-degree layer (second), (c) -45-degree layer (third),
and (d) 0-degree layer (fourth).158

Figure 4.113 (a) shows the measured V-I characteristics of this element (black diamonds). Figure
4.113 (b) shows the resistivity versus electric field (ρ-E characteristic) based on the measured V-I
characteristics shown in Figure 4.113 (a). The ρ-E characteristic is approximated using the following
simple expression [264]:
  E   4.8087×(3.0×106 E )28.16  6.0 103×(3.6 105×E ) 4.255  3.541×109 ×E Wm  . Equation 4.15

The thin solid line in Figure 4.113 (b) is drawn using Equation (4.13). The approximation using
Equation (4.13) is fine even beyond 250 kV/m: 360 kV/m (approximated) vs. 400 Ωm (measured) for
286 kV/m, 43 vs. 43 Ωm for 308 kV/m, 0.23 vs. 0.21 Ωm for 378 kV/m, 0.12 vs. 0.11 Ωm for 394 kV/m,
0.062 vs. 0.076 Ωm for 425 kV/m, 0.036 vs. 0.033 Ωm for 475 kV/m, 0.010 vs. 0.0094 Ωm for 681
kV/m. The V-I characteristic shown in Figure 4.113 (a) as gray diamonds is drawn on the basis of

158 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 330, Figure 9, 2016.

163
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Equation (4.15). Note that a different material will yield a ρ-E characteristic different from Figure 4.113
(b), but it could be well approximated by Equation (4.13) with different values of the constants.

Figure 4.112 Superimposed distribution of time integration of the square of FDTD-computed current of
the top four layers for the first 10 μs.159

30

25

20
Voltage [kV]

15

10

0
0.00001 0.001 0.1 10 1000 100000
C urrent [A ]

(a)

(b)

159 Reprinted with permission from M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, “FDTD simulation of lightning
current in a multilayer CFRP panel with triangular-prism cells,” IEEE Trans. EMC, vol. 58, no. 1, p. 330, Figure 10, 2016.

164
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.113 Measured characteristics of a ZnO element having a thickness of 36 mm and a radius of 21
mm. (a) V-I characteristic of the element and (b) ρ-E characteristic.160

Since this simple representation requires no iterative procedure, it is suitable and useful for simulating
an arrester. In the FDTD calculation, the resistivity ρ(E) given by Equation (4.13) is incorporated in
time-update equations for the electric fields. The time-update equation for the electric field in the x-
direction, Ex, is given as follows [265]:
t
 r 0 
Ex n 1 
 E
2  Ex n

t  n 
1
Equation 4.16
 H 2 
n
t x
t  x
 r 0   r 0 
 
2  Ex n  
2  Ex n

where εr is the relative permittivity, ε0 is the permittivity of vacuum, Δt is the time increment, H is the
magnetic-field vector, and n is the time step. Electric fields in the y- and z-directions are given similarly
to Equation (4.16).

4.8.2.3 Model: Representation of a substation arrester

Figure 4.114 shows the model of a substation arrester, represented with 10 mm × 10 mm × 10 mm
cubic cells, for the FDTD simulations. It is mounted on a 1-m high perfectly conducting rack on a flat,
perfectly conducting ground. The ZnO core has a length of 2.6 m and a radius of 40 mm. Each of the
cells representing the ZnO core has the nonlinear ρ-E characteristic shown in Figure 4.113 (b) in the x,
y, and z directions. The relative permittivity is set to 2. The porcelain insulator, coaxial with the ZnO
core, has an outer radius of 150 mm, and a thickness of 30 mm. The relative permittivity is set to 6.
There is a 20-mm air gap between the ZnO core and the porcelain insulator. A perfectly conducting
disc having a radius of 90 mm is placed on the top and bottom of the arrester. A shield ring having a
radius of 0.3 m is suspended 0.6 m below the top of the arrester.

2.6m 2 . 6mm
20 m
2 . 6 m 30 mm
60 mm 40 mm
30 mm 0.75 m

0.6 m
30 mm

2.6 m
0.3 m
ZnO

90-mm-radius disc
150 mm

1-m-high rack
1.0 m

0.5aam 4 cm
40 mm
6 cm
60 mm 2 20
cmmm
330
cmmm

(a) (b)

160Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 615, Figure 1, 2016.

165
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.114 FDTD simulation model of a 2.6-m long substation arrester. (a) Side view and (b) Plan view
(shield wire is not shown).161

A voltage having a magnitude of 1 MV and a risetime of 1 μs, shown in Figure 4.115 (a) (thin solid
line), is applied to this arrester via a 0.75-m long horizontal lead wire and a 100-Ω lumped resistor.

(a)

(b)
Figure 4.115 FDTD-computed waveforms. (a) Source output voltage and the FDTD-computed voltage
across the 2.6-m long arrester and (b) Current flowing in the top of the arrester.162

The conductor system described above is accommodated in a working volume of 1.2 m × 1.8 m × 4 m.
The five planes, except for the bottom perfectly conducting plane, which surround this working volume,
are represented by Liao’s second-order absorbing boundary [37] in order to suppress unwanted
reflections.

4.8.2.4 Analysis and results

The thick solid line in Figure 4.115 (a) is the FDTD-computed waveform for the voltage across the 2.6-
m long arrester. Note that about 80-kV (=100 Ω × 800 A) of the voltage drop from the applied voltage
of 1 MV is due to the 100-Ω resistor connected in series. Figure 4.115 (b) shows the FDTD-computed
waveform of the current flowing in the top of the arrester. It appears from Figure 4.115 (b) that the
arrester starts working at about 0.8 μs. Note that the small oscillation observed in the beginning is due
to successive reflections in the lead wire.

161Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 616, Figure 2, 2016.
162Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 616, Figure 3, 2016.

166
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.116 (a) shows FDTD-computed waveforms of vertical electric fields at a height of 3.5 m from
the ground (2.5 m from the arrester bottom) and radial distances of 0 (center) and 40 mm (arrester
side surface). Also shown in Figure 4.116 (a) are waveforms computed for the model without shield
ring. Figure 4.116 (b) shows the radial (or horizontal) electric fields.
Figure 4.117 shows the electric fields at a height of 1.1 m from the ground (0.1 m from the arrester
bottom) and radial distances of 0 and 40 mm.

(a)

(b)
Figure 4.116 FDTD-computed waveforms of electric fields at a height of 3.5 m from the ground (2.5 m from
the arrester bottom) and radial distances of 0 (center) and 40 mm (arrester side surface) for models with
and without a 0.3-m-radius shield ring suspended 0.6 m from the arrester top. (a) Vertical electric field and
(b) Radial electric field.163

It is clear from Figures 4.116 and 4.117 that the vertical electric field distribution is nonuniform and the
arrester starts working from its top. It is also clear that the radial electric field on the ZnO surface near
the top of the arrester is more intense than the vertical electric field, while the radial electric field near
the bottom is smaller than the vertical electric field. The intense radial electric field near the arrester
top is reduced from 56 kV/cm to 26 kV/cm (about 50% reduction) by a 0.3-m-radius shield ring
suspended 0.6 m below from the arrester top (see Figure 4.116 (b)). These electric fields are high
enough to initiate corona discharges and leaders in air. The vertical electric field on the side surface of
the arrester top is reduced from 5.0 kV/cm to 3.8 kV/cm (about a 25% reduction) by the shield ring
(see Figure 4.116 (a)), while the vertical electric fields at the center are not influenced by the shield
ring. Note that the vertical electric field would be 3.85 kV/cm (=1000 kV/260 cm) if the field is assumed
to be distributed vertically uniformly along the arrester.

163Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 617, Figure 4, 2016.

167
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 4.118 (a) shows FDTD-computed distributions of the vertical electric field along the surface of
the 40-mm-radius arrester core for models with and without the shield ring. Figure 4.118 (b) shows
those of the radial electric field. Note that both spatial distributions are obtained at time 2 μs. It is clear
from Figure 4.118 that both vertical (tangential) and radial electric fields have nonuniform distributions
and are high near the top of the arrester. They are mitigated by the 0.3-m-radius shield ring: the
maximum vertical electric field is reduced from 15 kV/cm to 9.0 kV/cm (about 40% reduction) for a 1
MV lightning voltage, and the maximum radial electric field is reduced from 58 kV/cm to 27 kV/cm
(about 50% reduction).

(a)

(b)
Figure 4.117 Same as Figure 4.7.14 but for electric fields at a height of 1.1 m from the ground (0.1 m from
the arrester bottom) (a) Vertical electric field and (b) Radial electric field.164

Figure 4.119 shows the case of the shield ring being suspended at 0.3 m from the arrester top.
Although the distributions are changed, the maximum electric fields are not influenced much: 8.0
kV/cm in the vertical direction and 30 kV/cm in the radial direction.

4.8.2.5 Summary
The electric field in and around a 2.6-m long substation arrester, on which a 1 MV lightning impulse is
imposed, is analyzed using the FDTD method. The arrester is represented as a combination of 10 mm
× 10 mm × 10 mm cubic cells, each of which has a nonlinear resistivity in each of the x, y, and z
directions, dependent on the electric field in each direction. This simple representation requires no

164Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 617, Figure 5, 2016.

168
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

iterative procedure and causes no numerical instability. The transient electric field in and around the
arrester with and without a shield ring, and the spatial distributions of vertical and radial electric fields
along the surface of the arrester are studied. Both vertical and radial electric fields have nonuniform
distributions and are higher near the top of the arrester. Near the top of the arrester, the radial electric
field is more intense than the vertical electric field. The former is reduced about 50% by a 0.3-m-radius
shield ring, and the latter is reduced about 25%.

(a)

(b)
Figure 4.118 FDTD-computed distributions of electric fields along the surface of the 2.6-m high and 40-
mm-radius arrester core for models with and without shield ring suspended 0.6 m below the arrester top
at time 2 μs. (a) Vertical electric field and (b) Radial electric field.165

4.9 Summary
In this chapter, representative examples of the application of the FDTD method, the PEEC method,
and the HEM to lightning electromagnetic transient analysis have been presented.
In Section 4.2, the FDTD method has been applied to simulation of the transient response of a
grounding grid with an aerial wire to simulate control cables and earth electrodes with consideration
given to the effect of soil ionization. The calculated results have been validated through comparison
with measured results. In addition, the PEEC method has been used for calculating the response of a
grounding grid and an earth electrode and the calculated results have been compared with measured
results and validated using other simulation techniques.

165Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 618, Figure 6, 2016.

169
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

In Section 4.3.1, the FDTD method has been applied to simulate the distortion characteristics of surge
voltages propagating on an overhead conductor and the calculated results have been compared with
measured results for validation. In addition, voltages induced on a nearby bundled conductor have
been presented. In Section 4.3.2, an actual 77-kV air-insulated substation, in close proximity to
transmission line towers, and a transmission line have been modeled using the FDTD method.
Lightning overvoltages entering the substation in the case of a direct lightning strike to its nearby
transmission line tower with simulation of back-flashover phenomena have been calculated and it has
been confirmed that the calculated results agree with those measured for an actual lightning strike. In
Section 4.3.3, the lightning overvoltage at the open end of the circuit breaker of a 500-kV air-insulated

(a)

(b)
Figure 4.119 FDTD-computed distributions of electric fields along the surface of the 2.6-m high and 40-
mm-radius arrester core for models with and without shield ring suspended 0.3 m below the arrester top
at time 2 μs. (a) Vertical electric field and (b) Radial electric field.166

substation has been simulated by the FDTD method while taking into account the AC operating
voltage on the power line and back-flashover phenomena at the first transmission line tower, and the
calculated results have been compared with those obtained by a circuit-theory-based simulation
program. In Section 4.3.4, the HEM has been applied to lightning surge simulations of a 138-kV
transmission line. The critical lightning currents to cause back flashover and the probability of back
flashover have been estimated while taking into account the frequency-dependent electrical
parameters of the soil. The effect of concise representations of tower-footing electrodes has also been
studied.
In Section 4.4, first, the FDTD method has been applied to calculate lightning-induced voltages on a
distribution line with surge arresters and a multipoint-grounded overhead shielding wire, and the
applicability of the FDTD method to the simulations of a practical configuration has been validated by
comparing the calculated results with those obtained by a field-to-transmission line coupling

166Reprinted with permission from S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, “FDTD analysis of the electric field of a
substation arrester under a lightning overvoltage,” IEEE Trans. EMC, vol. 58, no. 2, p. 618, Figure 7, 2016.

170
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

technique. Secondly, a reduced-scale model of a distribution line with surge arresters and
transformers represented by capacitances has been modeled by the FDTD method, and a good
agreement between the calculated and measured results has been found. Furthermore, some building
models simulated by conductor plates in the reduced-scale model have been incorporated into the
FDTD simulations, and the lightning-induced voltages have been calculated by taking into account the
shielding effects of the surrounding building models and the calculated results have been compared
with the measured results for validation. Finally, it has been explained how to apply the HEM to the
calculation of lightning-induced voltages, and the lightning-induced voltages have been calculated to
analyze the effect of the frequency-dependent electrical parameters of the soil on the induced
voltages.
In Section 4.5, the transient response of the grounding system of a wind turbine has been calculated
using the FDTD method while simulating its practical configuration. In addition, the effect of the
common grounding between a wind turbine and its nearby lightning protection tower has been
simulated. It has been confirmed that the FDTD results agree well with the measured results.
In Section 4.6, a reduced-scale model of a microwave relay station has been simulated by the FDTD
method while taking into account the detailed configuration of the reinforcing bars of the building and
foundation. The current distribution in the relay station model has been calculated when a lightning
impulse current is injected into the top of the microwave tower, and it has been presented that the
calculated current distribution and waveforms are in good agreement with the measured results.
In Section 4.7, the effect of conducting structures near an electromagnetic field measuring station on
errors of a magnetic direction finding of lightning has been analyzed using the FDTD method to study
the influence of the rise time of a lightning current, a ground conductivity, a grounded structure
conductivity, and the location of a structure. In addition, the FDTD modeling of a lightning
electromagnetic pulse propagation has been presented. In this study, a lightning return-stroke channel
and a conducting atmosphere in the daytime and in the nighttime have been simulated using the two-
dimensional FDTD method, and the lightning electromagnetic pulses have been calculated in the
distance of 50 km to 500 km. From the calculated results, the effective ionospheric reflection height
has been estimated.
In Section 4.8, a CFRP, each layer of which has a fiber direction of -45, 0, 45 or 90 degrees, has been
modeled using the FDTD method with triangular-prism cells by taking into account its anisotropic
conductivity. The magnetic fields on the surface of the CFRP model for different layers, when a
lightning impulse current is injected, and the calculated results have been compared with measured
results for validation. In addition, a lightning surge arrester for a substation has been modeled using
the two-dimensional FDTD method, where the nonlinear characteristics of the surge arrester materials
are represented by incorporating the resistivity-electric field relationship on the basis of the measured
results. The electric field distributions in and on the lightning surge arrester model have been
simulated to study the effect of a shield ring.

171
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

5. Switching surges in substations

Switching overvoltages on busbars occur during the operation of circuit breakers and disconnectors in
power plants and substations, and the transient electromagnetic interferences arising from the
switching surges may cause disturbances in sensitive electronic devices in low-voltage control circuits
(e.g., [266], [267]). Also, the transient electromagnetic fields may induce voltages on other busbars
grounded for maintenance [268]. However, it is not easy to properly evaluate the effectiveness of
countermeasures against electromagnetic interferences such as grounding wires in the case of high-
frequency surge phenomena, since the grounding wires may be so long and the configuration of
busbars is complex (particularly in extra high voltage (EHV) substations), and it is useful to predict
surge phenomena in busbars and accurately evaluate the effectiveness of protection measures.
Musa et al. applied the FDTD method to the analysis of switching transient phenomena [269]. The
frequency characteristics of electric and magnetic fields around a bus section were analyzed in a
simple configuration composed of a three-phase bus section in air and a current source to inject a
damped sinusoidal wave to simulate switching surges.
With the aim of applying the FDTD method to switching surge analysis in a practical configuration,
Tatematsu et al. proposed the following techniques for modeling components (busbars, remaining
voltages on busbars, and the occurrence of restrikes during the operation of disconnectors) in an EHV
air-insulated substation [270].
In an EHV substation, some incoming power lines and busbars comprise bundled conductors. In
FDTD-based surge simulations, it is generally quite difficult to divide an analysis space into sufficiently
small cells to handle each subconductor directly using the thin-wire representation technique, so in
this study, bundled-conductors (Figure 5.1) are approximated by an equivalent single conductor as
explained in [270].

rs
d
d

Subconductor
Subconductor

(a) (b)
Figure 5.1 (a) Equally spaced subconductors of a bundled conductor and (b) Unequally spaced
subconductors of a bundled conductor.167

In [270], Tatematsu et al. calculate switching surges arising from a restrike during the operation of a
disconnector, which may occur when the voltage difference between the two contacts of the
disconnector (that is, the difference between the remaining voltage at one contact and the AC voltage
at the other contact), exceeds the insulation strength in air. Although multiple restrikes (up to 5000 or
more [271]) may occur in the case of operating disconnectors on account of the slow movement of the
contacts, in this study, switching surge phenomena are predicted by simulating the occurrence of a
restrike at a disconnector, not multiple restrikes, since the interval between restrikes is sufficiently
large to handle each restrike individually [267].
In FDTD-based surge simulations, to simulate the effect of remaining voltages on busbars after
operating circuit breakers and disconnectors and before the occurrence of a restrike, voltage sources
A are connected to busbars via switches A as shown in Figure 5.2. After the voltages on the busbars
applied by voltage sources A, which produce a step wave, become a steady state, voltage sources A
are disconnected from the busbars by opening switches A at time t0. In the FDTD method, the
switches are placed on cell sections, and the electric fields across the switches are forced to zero by
time t0. After time t0, the electric fields across the switches are updated by the normal expression. In a
quasi-steady state, the voltages of the sources and busbars are the same, and thus the opening of the
switches does not significantly change the electromagnetic fields around the switches and does not

167Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2325, Figure 3, 2018.

172
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Voltage To simulate effect of

sources B AC voltages
Vp3
Vp2
Vp1

Switches A
Voltage
sources A

To simulate effect of
remaining voltages Switch B

Figure 5.2 Arrangement of voltage sources and switches.168

result in severe numerical oscillation, even though the switches are represented in a very simple
manner here.
To simulate the effect of AC voltages on busbars using the technique explained in Section 3.1.11, as
shown in Figure 5.2, voltage sources B are inserted in incoming power lines.
In this study, the remaining voltage is set to −1 p.u. and the voltage on the incoming power line is set
to +1 p.u at the occurrence of a restrike, which produces the maximum switching surge.
In FDTD-based switching surge simulations, to simulate the occurrence of a restrike at a disconnector,
as shown in Figure 5.2, switch B is inserted between the two contacts of the disconnector. When
voltages applied on busbars to simulate the effect of AC voltages and remaining voltages are in a
quasi-steady state, the closing of switch B simulates the occurrence of a restrike during the operation
of the disconnector and generates switching surges. In contrast to switch A mentioned above, the
closing of switch B significantly changes the electromagnetic field around the switch, and when the
electric field across the switch is simply set to zero, corresponding to closing the switch, it leads to
numerical oscillation. In this study, to avoid such numerical oscillation, the switch is simulated by the
following steps, as shown in Figure 5.3. Before time t1, the electric field across switch B is calculated
using the normal expression in the FDTD method to simulate the opened switch. At time t1, to avoid
numerical oscillation due to the abrupt change in the status of the switch, switch B is replaced with a
time-dependent lumped resistor in a transition phase (time duration: td).

Time t1 t 1 + td

Status of Open Transition Close

switch

Representation
of switch E is calculated R (t) E=0
using (1)

Figure 5.3 Representation of switch B to simulate the occurrence of a restrike.169

The lumped-parameter resistance in the FDTD method is represented using May et al.’s technique
[272]. After time t1 + td, the electric field across switch B is set to zero to represent the closed switch. In
this study, the time duration of the transition phase is set to be 0.5 μs. Through FDTD simulations to
validate the proposed technique, it is confirmed that the numerical oscillation due to the closing of the
switch can be suppressed significantly in the case of a time duration of 0.5 μs. In this case, it takes
about 0.1–0.2 μs until the current reaches a quasi-steady state through the transition phase. This time

168Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2326, Figure 4, 2018.
169Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2326, Figure 5, 2018.

173
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

delay corresponds to the voltage collapse time (several tens or some hundreds of nanoseconds)
between the contacts of equipment in air-insulated substations [271].
Using these techniques, Tatematsu et al. model a 500-kV air-insulated substation based on its
blueprint. Figure 5.4 shows the analysis space, whose volume is 405 m × 248 m × 135 m. The bottom
space with a thickness of 50 m corresponds to the soil. All external surfaces are treated as absorbing
boundaries using Liao’s second-order formulation to assume an open space. The resistivity of the soil
in the corresponding area ranges from 33 Ωm to 100 Ωm according to a previously measured ground-
conductivity distribution map of Japan [273], and thus it is set to 50 Ωm in this study. On the other
hand, the relative permittivity of the soil is unknown, and it is set to 10, which is used in some surge
analyses, based on typical values of 5 to 15 [96].

Voltage sources to
simulate effect of AC
voltages (VS #1)

85 m
Circuit #1

Circuit #2

50 m

Grounding grid
z
y Soil Main transformers

405 m
248 m

Figure 5.4 Analysis space used for calculating switching surges.170

The analysis space is divided uniformly into cubic cells of 0.5 m, and the time discretization is set to
0.481 ns. Figure 5.5 shows the calculation model of the 500-kV air-insulated substation, which is
composed of incoming power lines, busbars, gas-insulated equipment, steel structures, grounding
wires, main transformers, potential dividers, and a grounding grid. Although the size of the grounding
grid modeled in this study is 282 m × 148 m, the purpose of the FDTD simulations is not the
calculation of the ground potential rise of the grounding grid with reference to a remote zero-potential
point, and it is not required to set a larger analysis space. Tatematsu at el. confirmed that the volume
of the analysis space is large enough for obtaining accurate induced voltages by comparing the results
with those calculated using an expanded analysis space with a volume of 1105 × 948 m × 835 m
[270].
Gas-insulated equipment, such as gas-insulated circuit breakers (GCBs), gas-insulated composite
switchgear (GCS), and gas-insulated switchgear (GIS), has a coaxial structure. In this study, the
analysis space is large and the cells (0.5 m) are insufficiently small to model the circular shape of such
equipment in a straightforward manner, for example, by using a staircase approximation. Here, the
cross-section of the sheath of gas-insulated equipment is modeled by a square conductor of side 1 m,
and the center conductor is simulated by a thin wire. The internal surge impedance of the gas-
insulated equipment is assumed to be uniform and 70 Ω, which is recommended in the guide for
designing lightning protection in power plants and substations that is widely applied in Japan [184].
The radius of the internal thin wire is set to 167.5 mm based on the results of FDTD-based surge
simulations. The main transformers and potential dividers (PDs) are simulated by lumped-parameter
capacitances of 3000 pF and 2000 pF, respectively [184]. The steel structures are simulated by
combining square conductors with a cross-section of 2 m × 2 m.

170Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2328, Figure 8, 2018.

174
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Using the substation model described above, Tatematsu et al. calculate voltages on a grounded
busbar induced by a restrike at a disconnector. As shown in [270], one of the three busbars (busbar
Incoming power lines
Circuit #1 Bus sections
Grounding grid

Steel structures
Circuit #2

Bus ties
Potential dividers
Main transformers

Figure 5.5 Calculation model of the 500-kV air-insulated substation.171

#2) is grounded, while the other two busbars (busbars #1 and #3) are in operation. The incoming
power lines of one of the two circuits connected to this substation are disconnected from the
substation by GCS. Under these conditions, it is assumed that switching surges arise from the
occurrence of a restrike at each phase of the disconnector during its operation. As explained above, to
simulate the effect of the remaining voltage on each phase of the disconnected busbar and the
occurrence of a restrike at each phase of the disconnector, a voltage source and switches are
connected to each phase of the busbars. Connecting an additional wire to the sheath of the bus
section (GIS) at each phase, as shown in Figure 5.6, the induced voltages are calculated by the
voltage difference between the busbar and the additional wire at the top of the bushing and the
difference between the center conductor and the sheath of the bus section (GIS) at each phase.
One end of busbar #2 is grounded inside the bus tie. At the other end, it is grounded in two cases,
which are referred to as cases A and B. As shown in Figure 5.6 (a), in case A, each phase of busbar
#2 is grounded via the lower busbar with a height of 8.5 m, which is grounded using a grounding wire
with a cross-section of 100 mm2. As shown in Figure 5.6 (b), in case B, each phase of busbar #2 is
grounded directly using the same type of grounding wire near the top of the bushing of the bus
section.

Busbar Top of bushing

： height: 8.5 m： ： height: 11 m：
Grounding wire
Grounding wire
： length: 9.6 m：
Grounding grid Grounding grid ： length: 13 m：
： depth: 1 m： ： depth: 1 m：

Grounding wire
Grounding wire

Grounding wires Grounding wires

Additional wires Additional wires

for obtaining for obtaining
induced voltages induced voltages
Bus section Bus section

Phase B Phase W Phase R Phase B Phase W Phase R

(a) (b)

171Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2328, Figure 9, 2018.

175
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Figure 5.6 (a) Grounding of busbar #2 near the bus section in case A and (b) Grounding of busbar #2 near
the bus section in case B.172

The induced voltages on busbar #2 are calculated for the case of closing the switch to simulate the
occurrence of a restrike at each phase.
Figure 5.7 (a) shows the calculated voltages induced on one of the three phases of busbar #2 at the
top of the bushing in case B when the restrike occurs at each phase individually. Although the effect of
conductor internal impedances is ignored in the FDTD simulations, the induced voltages are
calculated when the conductor internal impedances of the main busbars were simulated using the
technique [91], and it is confirmed that the effect of conductor internal impedances is not an important
factor for obtaining accurate results for the induced voltages in this study. Figure 5.7 (b) shows the
measured voltage induced on busbar #2 at the same position (point A) [268] when the corresponding
disconnector was operated at the actual 500-kV air-insulated substation. Table 5.1 gives the peak
values of the induced voltages at the three phases of busbar #2 in cases A and B when the restrike
occurs at each phase of the disconnector. In Table 5.1, the values in parentheses are the peak values
of the induced voltages inside the bus sections (GIS). Table 5.1 shows that the peak values of the
voltages at the top of the bushings are similar to those inside the bus sections in cases A and B. For
comparison, the measured results of the induced voltages on busbar #2 in cases A and B are also
given in Table 5.1. The measured results I and II, respectively, show the averaged values of the peak
induced voltages when the disconnectors were closed and opened five times (Note that discharges
occurred many times during each operation).

10 μs

(a) (b)
Figure 5.7 (a) Calculated results of the induced voltages on busbar #2 at the top of the bushing at the bus
section in case B and (b) measured results at the same position.173,174

Table 5.1 Peak values of induced voltages in cases A and B.175

Peak value of induced voltage [kV]
Case Phase of restrike
Phase B Phase W Phase R
B 32 (33) 26 (26) 25 (29)
W 38 (37) 30 (32) 41 (45)
A R 38 (46) 37 (36) 33 (33)
Average value at each phase 36 (39) 31 (31) 33 (36)
Measured results I 16.3 19.0 17.9
Measured results II 17.8 21.8 21.1
B 22 (22) 20 (22) 28 (29)
B W 31 (30) 30 (32) 50 (47)
R 28 (30) 46 (44) 36 (36)

172Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2330, Figure 16, 2018.
173Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2331, Figure 17, 2018.
174Reprinted with permission from T. Ueda, “Measurement of disconnecting surges at 500 kV substation,” in Proc. Technical
Meeting on High Voltage Engineering, IEE Japan, HV-17-49, p. 164, Figure 10. 2017.
175Reprinted with permission from A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, p. 2331, Table 3, 2018.

176
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

Average value at each phase 27 (27) 32 (33) 38 (37)

Measured results I 14.3 20.5 21.4
Measured results II 17.8 22.0 22.0

Although in the measured result the phase at which the restrike occurred is unknown (since the
disconnectors at the three phases were operated simultaneously), the calculated induced voltages
oscillate with an interval of approximately 1 μs and the waveforms of the induced voltages are in good
agreement with the measured result. It should be noted, as previously mentioned, in this study the
collapse voltage between the two contacts of the disconnector is set to 2 p.u. to generate the
maximum switching surge, and the peak values of the calculated induced voltages in case A are on
average about 1.8 times larger than the measured results. In case A, as shown in [270], the grounding
wires are not connected directly to busbar #2 but are connected to the other busbars located below
busbar #2, and thus the lengths of the lines from the grounding point to point A, where the induced
voltages are evaluated, increase in the order of phases R, W, and B. However, as shown in Table 5.1,
despite the difference in the line lengths, the induced voltages at the three phases are similar, mainly
due to the effects of the capacitances of the potential dividers installed in busbar #2. Moreover,
although in case B the grounding wires are connected to busbar #2 close to the bus section, the
induced voltages are not suppressed by the grounding wires because of the long total length of the
grounding wire (11 m) above the soil and the grounding wire connecting the grounding point on the
ground surface to the grounding grid inside the soil. The long wire has a large inductance to
deteriorate the effectiveness of the grounding wire. The measured results also show that the peak
values of the induced voltages in case A are almost the same as those in case B and that the
grounding wires are not effective in reducing the induced voltages even in case B [268]. From the
above results, the FDTD simulations can reproduce the waveform of the induced voltage and the
ineffectiveness of the grounding wires installed to suppress the induced voltages, and thus the
applicability of the FDTD method to switching surge analysis in an EHV air-insulated substation in a
practical configuration is confirmed.
Furthermore, Tatematsu et al. evaluated the effectiveness of other grounding wires such as multiple
grounding wires to suppress the induced voltages [270].

177
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

6. Conclusions
Full-wave numerical approaches such as the finite-difference time-domain (FDTD) method, the
method of moments (MoM), partial element equivalent circuit (PEEC) method, and hybrid
electromagnetic model (HEM), do not require an assumption of the transverse electromagnetic (TEM)
mode and can solve electromagnetic transient phenomena in three-dimensional structures and
grounding systems accurately. These approaches have been widely applied to electromagnetic
transient analysis, and in particular, to lightning phenomena. In this technical brochure, the
fundamental theories of the numerical electromagnetic computation methods and several examples of
the application of the computation methods to the analysis of lightning electromagnetic transient
problems are provided.
In Chapter 1, an overview of electromagnetic computation methods for lightning surge studies was
presented, and the history of the application of the FDTD method to surge analyses was summarized.
In Chapter 2, the classification of lightning discharges in terms of the polarity of the charge and the
direction of the initial leader was presented, and the diagrams of the negative and positive discharges,
the parameters of the relevant lightning currents, and the transient phenomena of direct and indirect
lightning strikes were explained.
In Chapter 3, the fundamental theories and characteristics of the FDTD method, EFIE-based methods,
the PEEC method, and the HEM were explained. In addition, the following numerical techniques
developed for applying the FDTD method to electromagnetic transient analysis were presented, with
representation of: (i) thin wires with arbitrary radii; (ii) a lightning return-stroke channel and resulting
lightning electromagnetic fields; (iii) lightning surge arresters with detail-simulated voltage-current
relationships; (iv) corona discharges on high-voltage overhead conductors; (v) arcing-horn flashover
phenomena at transmission line towers; (vi) transmission line surge arresters; (vii) the effect of AC
voltages on power lines; (viii) twisted-wire pairs using transmission line theory; (ix) lossy thin wires;
and (x) lossy coaxial cables.
In Chapter 4, some examples of the application of the FDTD method, the PEEC method, and the HEM
to lightning electromagnetic transient analysis were presented.
In Section 4.1, the examples of the application in Chapter 4 were introduced.
In Section 4.2, the FDTD method was applied to simulation of the transient response of a grounding
grid with an aerial wire to simulate control cables and earth electrodes with consideration given to the
effect of soil ionization. The calculated results were validated through comparison with measured
results. In addition, the PEEC method was used for calculating the response of a grounding grid and
an earth electrode and the calculated results were compared with measured results and validated
using other simulation techniques.
In Section 4.3.1, the FDTD method was applied to simulate the distortion characteristics of surge
voltages propagating on an overhead conductor and the calculated results were compared with
measured results for validation. In addition, voltages induced on a nearby bundled conductor were
presented. In Section 4.3.2, an actual 77-kV air-insulated substation, in close proximity to transmission
line towers, and a transmission line were modeled using the FDTD method. Lightning overvoltages
entering the substation in the case of a direct lightning strike to its nearby transmission line tower with
simulation of back-flashover phenomena were calculated and it was confirmed that the calculated
results agreed with those measured for an actual lightning strike. In Section 4.3.3, the lightning
overvoltage at the open end of the circuit breaker of a 500-kV air-insulated substation was simulated
by the FDTD method while taking into account the AC operating voltage on the power line and back-
flashover phenomena at the first transmission line tower, and the calculated results were compared
with those obtained by a circuit-theory-based simulation program. In Section 4.3.4, the HEM was
applied to lightning surge simulations of a 138-kV transmission line. The critical lightning currents to
cause back flashover and the probability of back flashover were estimated while taking into account
the frequency-dependent electrical parameters of the soil. The effect of concise representations of
tower-footing electrodes was also studied.
In Section 4.4, first, the FDTD method was applied to calculate lightning-induced voltages on a
distribution line with surge arresters and a multipoint-grounded overhead shielding wire, and the
applicability of the FDTD method to the simulations of a practical configuration was validated by
comparing the calculated results with those obtained by a field-to-transmission line coupling
technique. Secondly, a reduced-scale model of a distribution line with surge arresters and
transformers represented by capacitances was modeled by the FDTD method, and a good agreement

178
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

between the calculated and measured results was found. Furthermore, some building models
simulated by conductor plates in the reduced-scale model were incorporated into the FDTD
simulations, and the lightning-induced voltages were calculated by taking into account the shielding
effects of the surrounding building models and the calculated results were compared with the
measured results for validation. Finally, it was explained how to apply the HEM to the calculation of
lightning-induced voltages, and the lightning-induced voltages were calculated to analyze the effect of
the frequency-dependent electrical parameters of the soil on the induced voltages.
In Section 4.5, the transient response of the grounding system of a wind turbine was calculated using
the FDTD method while simulating its practical configuration. In addition, the effect of the common
grounding between a wind turbine and its nearby lightning protection tower was simulated. It was
confirmed that the FDTD results agreed well with the measured results.
In Section 4.6, a reduced-scale model of a microwave relay station was simulated by the FDTD
method while taking into account the detailed configuration of the reinforcing bars of the building and
foundation. The current distribution in the relay station model was calculated when a lightning impulse
current was injected into the top of the microwave tower, and it was presented that the calculated
current distribution and waveforms were in good agreement with the measured results.
In Section 4.7, the effect of conducting structures near an electromagnetic field measuring station on
errors of a magnetic direction finding of lightning was analyzed using the FDTD method to study the
influence of the rise time of a lightning current, a ground conductivity, a grounded structure
conductivity, and the location of a structure. In addition, the FDTD modeling of a lightning
electromagnetic pulse propagation was presented. In this study, a lightning return-stroke channel and
a conducting atmosphere in the daytime and in the nighttime were simulated using the two-
dimensional FDTD method, and the lightning electromagnetic pulses were calculated in the distance
of 50 km to 500 km. From the calculated results, the effective ionospheric reflection height was
estimated.
In Section 4.8, a multilayer carbon fiber reinforced plastic (CFRP), each layer of which has a fiber
direction of -45, 0, 45 or 90 degrees, was modeled using the FDTD method with triangular-prism cells
by taking into account its anisotropic conductivity. The magnetic fields on the surface of the CFRP
model for different layers, when a lightning impulse current was injected, and the calculated results
were compared with measured results for validation. In addition, a lightning surge arrester for a
substation was modeled using the two-dimensional FDTD method, where the nonlinear characteristics
of the surge arrester materials were represented by incorporating the resistivity-electric field
relationship on the basis of the measured results. The electric field distributions in and on the lightning
surge arrester model were simulated to study the effect of a shield ring.
In Section 4.9, the applications of the FDTD method, the PEEC method, and the HEM to lightning
electromagnetic transient analysis shown in Chapter 4 were summarized.
In Chapter 5, unlike the aforementioned simulations, an example of the application of the FDTD
method to switching surge analysis was presented. In this study, to apply the FDTD method to the
switching surge analysis of an extra-high-voltage substation, some techniques for modeling substation
components and the occurrence of a restrike at a disconnector were explained. In addition, the
calculated induced voltages on a grounded busbar due to a restrike at a disconnector during its
operation at a 500-kV air-insulated substation were presented to study the effect of the grounding
wire, and the calculated results were compared with the measured results obtained at the
corresponding actual substation.

179
 TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

References
[1] M. A. Uman, D. K. McLain, and E. P. Krider, The electromagnetic radiation from a finite antenna,
American Journal of Physics, vol. 43, pp. 33-38, 1975.
[2] F. Rachidi, C. A. Nucci, M. Ianoz, and C. Mazzetti, Influence of a lossy ground on lightning-induced
voltages on overhead lines, IEEE Trans. EMC, vol. 38, no. 3, pp. 250-264, 1996.
[3] C. Taylor, R. Satterwhite, and C. Harrison, The response of a terminated two-wire transmission line
excited by a nonuniform electromagnetic field, IEEE Trans. AP, vol. 13, no. 6, pp. 987-989, 1965.
[4] A. K. Agrawal, H. J. Price, and S. H. Gurbaxani, Transient response of multiconductor transmission
lines excited by a nonuniform electromagnetic field, IEEE Trans. EMC, vol. 22, no. 2, pp. 119-129,
1980.
[5] F. Rachidi, “Formulation of the Field-to-Transmission Line Coupling Equations in Terms of Magnetic
Excitation Field,” IEEE Trans. Electromagnetic Compatibility, vol. 35, no. 3, pp. 404-407, Aug. 1993.
[6] V. Cooray, F. Rachidi, and M. Rubinstein, Formulation of the field-to-transmission line coupling
equations in terms of scalar and vector potentials, IEEE Trans. EMC, vol. 59, no. 5, pp. 1586-1591,
2017.
[7] M. Rubinstein, An approximate formula for calculation of the horizontal electric field from lightning at
close, intermediate, and long ranges, IEEE Trans. EMC, vol. 38, no. 3, pp. 531- 535, 1996.
[8] H. W. Dommel, Digital computer solution of electromagnetic transients in single and multiphase
networks, IEEE Trans. PAS, vol. 88, no. 4, pp. 388-399, 1969.
[9] A. S. Podgorski, and J. A. Landt, Three dimensional time domain modeling of lightning, IEEE Trans.
PWRD, vol. 2, no. 3, pp 931-938, 1987.
[10] M. Van Baricum, and E. K. Miller, E. K., TWTD --- A Computer Program for Time-Domain Analysis
of Thin-Wire Structures, UCRL-51-277, Lawrence Livermore Laboratory, California, 1972.
[11] E. K. Miller, A. J. Poggio, and G. J. Burke, An integro-differential equation technique for the time
domain analysis of thin wire structure: Part I. The numerical method, Journal of Computational
Physics, vol. 12, pp. 24-28, 1973.
[12] L. Grcev, and F. Dawalibi, An electromagnetic model for transients in grounding systems, IEEE
Transactions on Power Delivery, vol. 5 (4), pp. 1773–1781, 1990.
[13] R. F. Harrington, Field Computation by Moment Methods, Macmillan Co., New York, 1968.
[14] Y. Baba and V. A. Rakov, “Applications of electromagnetic models of the lightning return stroke,”
IEEE Trans. Power Delivery, vol. 23, no. 2, pp. 800-811, Apr. 2008.
[15] K. Tanabe, Novel method for analyzing dynamic behavior of grounding systems based on the finite-
difference time-domain method, IEEE Power Engineering Review, vol. 21, no. 9, pp. 55-57, 2001.
[16] K. S. Yee, “Numerical solution of initial boundary value problems involving Maxwell’s equations in
isotropic media,” IEEE Trans. Antennas Propagat., vol. AP-14, pp. 302–307, May 1966.
[17] Y. Baba, and V. A. Rakov, On the transmission line model for lightning return stroke representation.
Geophysical Research Letters, vol. 30, no. 24, 4 pages, 2003.
[18] Y. Baba, and V. A. Rakov, Applications of the FDTD method to lightning electromagnetic pulse and
surge simulations, IEEE Trans. EMC, vol. 56, no. 6, pp. 1506-1521, 2014.
[19] A. Tatematsu, A technique for representing coaxial cables for FDTD-based surge simulations, IEEE
Trans. EMC, vol. 57, no. 3, pp. 488-495, 2015.
[20] A. E. Ruehli, Equivalent circuit models for three-dimensional multiconductor systems, IEEE Trans.
MTT, vol. 22, no. 3, 216-221, 1974.
[21] S. Visacro, and A. Soares Jr., HEM: A model for simulation of lightning related engineering
problems, IEEE Trans. PWRD, vol. 20, no. 2, pp. 1206-1207, 2005.
[22] P. B. Johns, and R. L. Beurle, Numerical solutions of 2-dimensional scattering problems using a
transmission-line matrix, Proceedings of IEE, vol. 118, no. 9, pp. 1203-1208, 1971.
[23] M. N. O. Sadiku, A simple introduction to finite element analysis of electromagnetic problems, IEEE
Trans. Education, vol. 32, no. 2, pp.85-93, 1989
[24] P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, Lightning-induced voltage over lossy
ground by a hybrid electromagnetic circuit model method with Cooray-Rubinstein formula, IEEE
Trans. EMC, vol. 51, no. 4, pp. 975-985, 2009.

180
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[25] F. H. Silveira, A. De Conti, and S. Visacro, Lightning Overvoltage Due to First Strokes Considering a
Realistic Current Representation. IEEE Transactions on Electromagnetic Compatibility, v. 52, p.
929-935, 2010.
[26] A. Shoory, F. Vega, P. Yutthagowith, F. Rachidi, M. Rubinstein, Y. Baba, V. A. Rakov, K.
Sheshyekani, and A. Ametani, On the mechanism of current pulse propagation along conical
structures: Application to tall towers struck by lightning, IEEE Trans. EMC, vol. 54, no. 2, pp.332-
342, 2012.
[27] S. Yuda, S. Sawaki, Y. Baba, N. Nagaoka, and A. Ametani, Application of the TLM method to
transient simulations of a conductor system with a lossy ground: grounding electrodes and an
overhead wire, IEEE Trans. EMC, vol. 55, no. 1, pp. 175-182, 2013.
[28] V. A. Rakov, and M. A. Uman, Lightning: Physics and Effects, Cambridge University Press, UK, pp.
1-687, 2003.
[29] V. A. Rakov, Lightning return stroke speed, Journal of Lightning Research, vol. 1, pp. 80-89, 2007.
[30] V. A. Rakov, M. A. Uman, and R. Thottappillil, Review of lightning properties from electric field and
TV observations, Journal of Geophysical Research, vol. 99, pp. 10745-10750, 1994.
[31] K. Berger, R. B. Anderson, and H. Kroninger, Parameters of lightning flashes, Electra, vol. 80, pp.
23-37, 1975.
[32] B. N. Gorin, and A. V. Shkilev, Measurements of lightning currents at the Ostankino tower,
Electrichestrvo, vol. 8, pp. 64-65, 1984 (in Russian).
[33] V. A. Rakov, Some inferences on the propagation mechanisms of dart leaders and return strokes,
Journal of Geophysical Research, vol. 103, no. D2, pp. 1879-1887, 1998.
[34] V. A. Rakov, D. E., Crawford, K. J., Rambo, G. H., Schnetzer, M. A., Uman, and R. Thottappillil, M-
component mode of charge transfer to ground in lightning discharges, Journal of Geophysical
Research, vol. 106, pp. 22817-22831, 2001
[35] A. J. Eriksson, Lightning and tall structures, Transaction of South African IEE, vol. 69, pp. 238-252,
1978.
[36] K. Nakada, T. Yokota, S., Yokoyama, A. Asakawa, and T. Kawabata, Distribution arrester outages
caused by lightning backflow current flowing from customer's facility into power distribution lines,
Electrical Engineering in Japan, Wiley, vol. 126, no. 3, pp. 9-20, 1999.
[37] Z. P. Liao, H. L. Wong, B. P. Yang, and Y. F. Yuan, “A transmitting boundary for transient wave
analysis,” Scientia Sinica, Series A, vol. 27, no. 10, pp. 1063–1076, 1984.
[38] J. P. Berenger, A perfectly matched layer for the absorption of electromagnetic waves, Journal of
Computational Physics, vol. 114, pp. 185-200, 1994.
[39] R. Courant, K. Friedrichs, and H. Lewy, Über die partiellen differenzengleichungen der
mathematischen, Physik. Mathematiche Annalen, vol. 100, no. 1, pp. 32-74, 1928 (in German).
[40] T. Asada, Y. Baba, N. Nagaoka, and A. Ametani, A study of absorbing boundary condition for surge
simulations with the FDTD method, IEEJ Trans. Power and Energy, vol. 135, no. 6, 408-416, 2015.
(in Japanese).
[41] T. Uno, Finite Difference Time Domain Method for Electromagnetic Field and Antennas, Corona
Publishing Co., Ltd., Tokyo, Japan, 1998 (in Japanese).
[42] R. Holland and L. Simpson, "Finite-difference analysis of EMP coupling to thin struts and wires,"
IEEE Trans. Electromagn. Compat., vol. EMC-23, no. 2, pp. 88-97, May 1981.
[43] K. R. Umashankar, A. Taflov, and B. Beker, Calculation and experimental validation of induced
currents on coupled wires in an arbitrary shaped cavity, IEEE Trans. AP, vol. 35, no. 11, pp. 1248-
1257, 1987.
[44] T. Noda and S. Yokoyama, “Thin wire representation in finite difference time domain surge
simulation,” IEEE Trans. Power Del., vol. 17, no. 3, pp. 840–847, Jul. 2002.
[45] Y. Baba, N. Nagaoka, and A. Ametani, Modeling of thin wires in a lossy medium for FDTD
simulations, IEEE Trans. EMC, vol. 47, no. 1, pp. 54-60, 2005.
[46] C. J. Railton, D. L. Paul, and S. Dumanli, The treatment of thin wire and coaxial structures in
lossless and lossy media in FDTD by the modification of assigned material parameters, IEEE Trans.
EMC, vol. 48, no. 4, pp. 654-660, 2006.
[47] Y. Taniguchi, Y. Baba, N. Nagaoka, and A. Ametani, An improved thin wire representation for FDTD
computations. IEEE Trans. AP, vol. 56, no. 10, pp. 3248-3252, 2008.

181
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[48] T. Asada, Y. Baba, N. Nagaoka, and A. Ametani, An improved thin wire representation for FDTD
transient simulations, IEEE Trans. EMC, vol. 57, no. 3, pp. 484-487, 2015.
[49] Y. Baba and V. A. Rakov, “Electromagnetic models of the lightning return stroke,” J. Geophysical
Research, vol. 112, D04102, doi:10.1029/2006JD007222, 2007.
[50] Y. Baba, and V. A. Rakov, Electric and magnetic fields predicted by different electromagnetic
models of the lightning return stroke versus measured fields, IEEE Trans. EMC, vol. 51, no. 3, pp.
479-483, 2009.
[51] V. A. Rakov, and A. A. Dulzon, Calculated electromagnetic fields of lightning return stroke, Tekh.
Elektrodinam., vol. 1, pp. 87-89, 1987 (in Russian).
[52] Y. Baba, and V. A. Rakov, On the mechanism of attenuation of current waves propagating along a
vertical perfectly conducting wire above ground: application to lightning, IEEE Trans. EMC, vol. 47,
no. 3, pp. 521-532, 2005
[53] S. Kato, T., Narita, T. Yamada, and E. Zaima, Simulation of electromagnetic field in lightning to tall
tower, Paper presented at 11th International Symposium on High Voltage Engineering, no. 467,
London, UK, 1999.
[54] S. Bonyadi-Ram, R. Moini, S. H. H. Sadeghi, and V. A. Rakov, On representation of lightning return
stroke as a lossy monopole antenna with inductive loading, IEEE Trans. EMC, vol. 50, no. 1, 118-
127, 2008.
[55] V. P. Idone, and R. E. Orville, Lightning return stroke velocities in the Thunderstorm Research
International Program (TRIP), Journal of Geophysical Research, vol. 87, pp. 4903-4915, 1982.
[56] R. Moini, B. Kordi, G. Z. Rafi, and V. A. Rakov, A new lightning return stroke model based on
antenna theory, Journal of Geophysical Research, vol. 105, no. D24, pp. 29,693-29,702, 2000.
[57] S. Kato, T. Takinami, T. Hirai, and S. Okabe, A study of lightning channel model in numerical
electromagnetic field computation, Paper presented at 2001 IEEJ National Convention, no. 7-140,
Nagoya, Japan, 2001 (in Japanese).
[58] S. Miyazaki, and M. Ishii, Reproduction of electromagnetic fields associated with lightning return
stroke to a high structure using FDTD method, Paper presented at 2004 IEEJ National Convention,
no. 7-065, p. 98, Kanagawa, Japan, 2004 (in Japanese).
[59] S. Bonyadi-Ram, R. Moini, S. H. H. Sadeghi, and V. A. Rakov, Incorporation of distributed capacitive
loads in the antenna theory model of lightning return stroke, Paper presented at 16th International
Zurich Symposium on EMC, pp. 213-218, Zurich, 2005.
[60] Y. Baba, and V. A. Rakov, On the use of lumped sources in lightning return stroke models, Journal
of Geophysical Research, vol. 110, no. D03101, doi:10.1029/2004JD005202, 2005.
[61] Y. Baba, and V. A. Rakov, Influences of the presence of a tall grounded strike object and an upward
connecting leader on lightning currents and electromagnetic fields, IEEE Trans. EMC, vol. 49, no. 4,
pp. 886-892, 2007.
[62] Masłowski G., and R. Ziemba, Calculation of lightning-induced voltages inside the structure using
engineering return-stroke models, Proc. 28th International Conference on Lightning Protection,
Kanazawa, Japan, pp. 1132-1137, 2006.
[63] J.G. Anderson to W.A. Chisholm, private communication, 1984
[64] F. Heidler, Traveling current source model for LEMp calculation, Proc. of 6th Int. Zurich Symp. on
EMC, Zurich, Switzerland, pp. 157-162, 1985.
[65] H. Karami, F. Rachidi, and M. Rubinstein, On practical implementation of electromagnetic models of
lightning return-strokes, Atmosphere, vol. 7, no. 135, doi:10.3390/atmos7100135, 2016.
[66] R. Moini, B. Kordi, and M. Abedi, Evaluation of LEMP effects on complex wire structures located
above a perfectly conducting ground using electric field integral equation in time domain, IEEE
Trans. EMC, vol. 40, no. 2, pp. 154-162, 1998.
[67] T. Noda, A tower model for lightning overvoltage studies based on the result of an FDTD simulation.
IEEJ Trans. Power and Energy, vol. 127, no. 2, 379-388, 2007 (in Japanese).
[68] A. Tatematsu and T. Noda, “Three-dimensional FDTD calculation of lightning-induced voltages on a
multiphase distribution line with the lightning arresters and an overhead shielding wire,” IEEE Trans.
Electromagn. Compat., vol. 56, no. 1, pp. 159-167, Feb. 2014.
[69] T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov, A simplified
model of corona discharge on overhead wire for FDTD computations, IEEE Trans. EMC, vol. 54, no.
3, pp. 585–593, 2012.

182
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[70] Masłowski G., and V. A. Rakov, A study of the lightning channel corona sheath, J. Geophys. Res.,
vol. 111, no. D14110, doi:10.1029/2005JD006858, 2006.
[71] G. Hartmann, Theoretical evaluation of Peek’s law, IEEE Trans. Industry Applications, vol. 20, no. 6,
pp. 1647-1651, 1984.
[72] V. Cooray, The Lightning flash, p. 79, The Institution of Electrical Engineers, UK, 2003.
[73] R. T. Waters, D. M. German, A. E. Davies, N. Harid, and H. S. B. Eloyyan, Twin conductor surge
corona, Paper presented at the 5th Int. Symp. High Voltage Engineering, Braunschweig, Federal
Republic of Germany, 1987.
[74] T. Noda, Development of a transmission-line model considering the skin and corona effects for
power systems transient analysis, Ph.D. Thesis, Doshisha University, 1996.
[75] H. Motoyama, Experimental study and analysis of breakdown characteristics of long air gaps with
short tail lightning impulse, IEEE Trans. PWRD, vol. 11, no. 2, pp. 972-979, 1996.
[76] T. Shindo and T. Suzuki, A new calculation method of breakdown voltage-time characteristics of
long air gaps, IEEE Trans. PWRD, vol. 104, no. 6, pp. 1556-1563, 1985.
[77] N. Nagaoka, A flashover model using a nonlinear inductance, Trans. IEE of Japan, vol. 111, no. 5,
pp. 529-534, 1991 (in Japanese).
[78] A. Tatematsu, T. Noda, and H. Motoyama, Development of simulation techniques for super-thin
wires and nonlinear elements in the FDTD Method, CRIEPI Report, no. H06006, 2007 (in
Japanese).
[79] A. Tatematsu and T. Ueda, “FDTD-Based Lightning Surge Simulation of an HV Air-Insulated
Substation with Back-Flashover Phenomena,” IEEE Trans. Electromagn. Compat., vol. Vol. 58, No.
5, pp. 1549-1560, Oct. 2016.
[80] H. Motoyama, K. Shinjo, Y. Matsumoto, and N. Itamoto, Observation and analysis of multiphase
back flashover on the Okushishiku test transmission line caused by winter lightning, IEEE Trans.
PWRD, vol. 13, no. 4, pp. 1391-1398, 1998.
[81] T. Ueda, M. Morita, A. Ametani, T. Funabashi, T. Hagiwara, and H. Watanabe, Flashover model for
arcing horns and transmission line arresters, Trans. IEE of Japan, vol. 112, no. 12, pp. 1085-1091,
1992 (in Japanese).
[82] A. Tatematsu, F. Rachidi, and M. Rubinstein, A technique for calculating voltages induced on
twisted-wire pairs using the FDTD method, IEEE Trans. EMC, vol. 59, no. 1, pp. 301-304, 2017.
[83] F. M. Tesche, M. V. Ianoz, and T. Karlsson, EMC Analysis Methods and Computational Models,
John Wiley & Sons, Inc., 1997.
[84] R. B. Armenta and C. D. Sarris, Efficient evaluation of the terminal response of a twisted-wire pair
excited by a plane-wave electromagnetic field, IEEE Trans. EMC, vol. 49, no. 3, pp. 698-707, 2007.
[85] S. A. Pignari, and G. Spadacini, Plane-wave coupling to a twisted-wire pair above ground, IEEE
Trans. EMC, vol. 53, no. 2, pp. 508-523, 2011.
[86] C. D. Tayor, and J. P. Castillo, On the response of a terminated twisted-wire cable excited by a
plane-wave electromagnetic field, IEEE Trans. EMC, vol. 22, no. 1, pp. 16-19, 1980.
[87] O. C. Zienkiewicz, Finite Element Method, third edition, McGraw-Hill Book Co., 1983.
[88] R. Holland, and L. Simpson, Implementation and optimization of the thin-strut formalism in THREDE,
IEEE Trans. Nucl. Sci., vol. 27, no. 6, pp. 1625-1630, 1980.
[89] M. Bingle, D. B. Davidson, and J. H. Cloete, Scattering and absorption by thin metal wires in
rectangular waveguide – FDTD simulation and physical experiments, IEEE Trans. MTT, vol. 50, no.
6, pp. 1621-1627, 2002.
[90] Y. Du, B. Li, and M. Chen, The extended thin wire model of lossy round wire structures for FDTD
simulations, IEEE Trans. PWRD, vol. 32, no. 6, pp. 2472–2480, 2017.
[91] A. Tatematsu, “A technique for representing lossy thin wires and coaxial cables for FDTD-based
surge simulations,” IEEE Trans. Electromagn. Compat., vol. 60, no. 3, pp. 705-715, Jun. 2018.
[92] S. A. Schelkunoff, The electromagnetic theory of coaxial transmission line and cylindrical shields,
Bell System Technical Journal, vol. 13, pp. 532-579, 1934.
[93] L. M. Wedepohl, and D. J. Wilcox, Transient analysis of underground power-transmission systems.
System-model and wave-propagation characteristics, Proc. IEE, vol. 120, no. 2, pp. 253-260, 1973.
[94] M. Abramowitz, and I. A. Stegun, Handbook of Mathematical Functions with Formulas, Graphs, and
Mathematical Tables, Dover Publications, Inc., New York, 1970.

183
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[95] J. G. Maloney, K. L. Shlager, and G. S. Smith, A simple FDTD model for transient excitation of
antennas by transmission lines, IEEE Trans. AP, vol. 42, no. 2, pp. 289-292, 1994.
[96] E. F. Vance, “Coupling to shielded cables,” John Wiley and Sons, 1978.
[97] M. Feliziani and F. Maradei, Full-wave analysis of shielded cable configurations by the FDTD
method, IEEE Trans. Magn., vol. 38, no. 2, pp. 761-764, 2002.
[98] D. F. Higgins, Time-domain calculation of the leakage of SGEMP transients through braided cable
shields, IEEE Trans. Nucl. Sci., vol. 36, no. 6, pp. 2042-2049, 1989.
[99] S. S. Zivanovic, K. S. Yee, and K. K. Mei, A subgridding method of time-domain finite-difference
method to solve Maxwell’s equations, IEEE Trans. MTT, vol. 39, no. 3, pp. 471-479, 1991.
[100] J. L. He, S. C. Wang, and R. Zeng, Using subgridding FDTD method to incorporate air gap
breakdown in the analysis of lightning induced voltage, Journal of Lightning Research, vol.4, Suppl
1, pp. 45-48, 2012.
[101] P. Chowdhuri, A. K. Mishra, and B. W. McConnell, Volt-time characteristics of short air gaps under
nonstandard lightning voltage waves, IEEE Trans. PWRD, vol. 12, no. 1, pp. 470-476, 1997.
[102] J. R. Marti, and J. Lin, Suppression of numerical oscillations in the EMTP power systems, IEEE
Trans. Power Systems, vol. 4, no.2, pp.739-747, 1989.
[103] S. Q. Gu, W. J. Chen, J. L. He, H. Shen, and S. Zhang, Monitoring results of the effectiveness of
surge arrester spacings on distribution line protection, IEEE Trans. PWRD, vol. 22, no. 4, pp. 2191-
2198, 2007.
[104] G. R. Martin, A. Salinas, and A. R. Bretones, Time-domain integral equation methods for transient
analysis, IEEE Antenna and Propagation Magazine, vol. 34. no. 3, pp. 15 – 22, 1992.
[105] T. P. Sarkar, A. R. Djordjevic, and B. M. Kolundzija, Method of moments applied to antennas,
http://mtt.etf.bg.ac.rs/Mikrotalasna.Tehnika/Clanci/ANTMOM.pdf, 2000.
[106] C. W. Gibson, The Method of Moments in Electromagnetics, Chapman and Hall/CRC, Taylor
Francis Group, 2007.
[107] M. Ishii, and Y. Baba, Numerical electromagnetic field analysis of tower surge response, IEEE
Trans. PWRD, vol. 12, no.1, pp. 483-488, 1997.
[108] M. Ishii, and Y. Baba, Advanced computational methods in lightning performance – the numerical
electromagnetic code (NEC-2), in Proc. IEEE PES Winter Meeting, Singapore, pp. 2419-2424,
2000.
[109] Y. Baba, and M. Ishii, Numerical electromagnetic field analysis of lightning currents in tall structures,
IEEE Trans. PWRD, vol. 16, no. 2, pp. 324-328, 2001.
[110] Y. Baba, and M. Ishii, Characteristics of electromagnetic return stroke models, IEEE Trans. EMC,
vol. 45, no.1, pp. 129 – 134, 2003.
[111] U. Kumar, and P. K. Nayak, Investigations on the post-stroke voltages and currents in lightning
protection schemes involving single tower, IEEE Trans. EMC, vol. 47, no. 3, pp. 543–551, 2005.
[112] U. Kumar, V. Hegde, and P. B. and Darji, Investigations on voltage and currents in the lightning
protection system of the Indian satellite lauch pad-I during stroke interception, IET Sci,
Measurement, Techol , vol. 5, pp. 225-231, 2007.
[113] U. Kumar, Investigation on the potential rise and currents in insulated mast scheme with single
tower during stroke interception, Journal of Lightning Research, vol. 1, pp. 28-38, 2009.
[114] V. Hegde, and U. Kumar, Studies on characteristics of lightning generated currents in an
interconnected lightning protection system, Journal of Electrostatics, vol. 67, pp. 590-596, 2009.
[115] R. U. Raysaha, U. Kumar, and R. Thottappillil, A macroscopic model for first return stroke of
lightning, IEEE Trans. EMC, vol. 53, no. 3, pp. 782-791, 2011.
[116] U. Kumar, and R. K. Govinda, Estimation of electrical stress in insulated down conductors and
cables, Asia-Pacific International Conference on Lightning, Seoul, Korea, 2013.
[117] G. J. Burke, and A. J. Poggio, Numerical Electromagnetic Code (NEC)-method of moments part I &
III, Technical document 116, Naval Ocean Systems Center, San Diego, 1980.
[118] U. Kumar, V. Hegde, and V. Shivanand, Preliminary studies on the characteristics of the induced
currents in simple down conductors due to nearby strike, IEEE Trans. EMC, vol. 48., no. 4, pp. 805-
816, 2006.

184
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[119] G. J. Burke, Recent advances to NEC: Applications and validation, modern antenna design using
computers and measurement, application to antenna problems of military interest, Ankara, Turkey,
October 1989; UCRL -100651.
[120] J. K. Santosh, and U. Kumar, Time domain approach for modelling transformer windings for
propagation of fast discharge pulses, Paper presented at the 14th International Conference on
Electromagnetic Interference and Compatibility (INCEMIC 2016), Bangalore, India, 2016.
[121] A. J. Poggio, E. K. Miller, and G. J. Burke, An integro-differential equation technique for the time-
domain analysis of thin wire structures: II. Numerical results, Journal of Computational Physics, vol.
12, pp. 210 - 233, 1973.
[122] U. Kumar, and V. Sreeram, Modelling proximity effects in transverse magnetic mode for lightning
studies, International Conference on Lightning Protection (ICLP), Shanghai, China, no. 474, 2014.
[123] M. F. Pantoja, A. G. Yarovoy, A. R. Bretones, and and S. G. Garcia, Time domain analysis of thin-
wire antennas over lossy ground using the reflection-coefficient approximation, Radio Science, vol.
44, RS6009, pp. 1–14, 2009.
[124] A. E. Ruehli, and H. Heeb, Circuit models for three-dimensional geometry including dielectrics, IEEE
Trans. MTT, vol. 40, no. 7, 1507-1516, 1992.
[125] A. E. Ruehli, J. Garrett, and C. R. Paul, Circuit models for 3D structures with incident fields, in Proc.
of the IEEE Int. Symposium on EMC, pp. 28-31, Dallas, Texas, USA, 1993.
[126] G. Antonini, Fast multipole method for time domain PEEC analysis, IEEE Trans. Mobile Computing,
vol. 2 no. 4, pp. 275-287, 2003.
[127] A. E. Ruelhi, and A. C. Cangellaris, Application of the partial element equivalent circuit (PEEC)
method to realistic printed circuit board problem, in Proc. of the IEEE Int. Symposium on EMC, pp.
182-197, Denver, CO, USA, 1998.
[128] A. E. Ruehli et al., Comparison of differential and common mode response for short transmission
line using PEEC models, in Proc. of the IEEE Topical Meeting on Electrical Performance and
Electronic Packaging, pp. 169-171, Napa, CA, USA, 1996.
[129] G. Antonini, and S. Cristina, EMI design of power drive system cages using the PEEC method and
genetic optimization techniques, in Proc. of the IEEE Int. Symposium on EMC, pp. 156-160,
Montreal, Canada, 2001.
[130] A. E. Ruehli, U. Miekkala, and H. Heeb, Stability of discretized partial element equivalent EFIE
circuit models, IEEE Trans. AP, vol. 43, no. 6, 553-559, 1995.
[131] G. Antonini, S. Cristina, and A. Orlandi, PEEC modeling of lightning protection systems and coupling
to coaxial cables, IEEE Trans.EMC, vol. 40, no. 4, 481- 491, 1998.
[132] P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, Influence of a measuring system to a
transient voltage on a vertical Conductor, IEEJ. Trans. EEE., vol. 5, no. 2, 221-228, 2010.
[133] P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, Application of the partial element equivalent
circuit method to analysis of transient potential rises in grounding systems, IEEE Trans. EMC, vol.
53, no. 3, 726-736, Aug. 2011.
[134] P. Yutthagowith, Development of the PEEC Method and Its Application to a Lightning Surge Study,
Ph.D. thesis, Doshisha University, Kyoto, Japan, 2010.
[135] J. Nitsch, F. Gronwald, and G. Wollenberg, Radiating Non-Uniform Transmission Line Systems and
the Partial Element Equivalent Circuit Method, John Wiley and Sons, Ltd. Pub., 2009.
[136] A. Bellen, N. Guglielmi, and A. Ruehli, Methods for linear systems of circuit delay differential
equations of neutral type, IEEE Trans. Circuits Sys., vol. 46, no. 1, 212-216, 1999.
[137] P. Yutthagowith, A. Ametani, N. Nagaoka, and Y. Baba, Application of the partial element equivalent
circuit method to tower surge response calculations, IEEJ Trans. EEE., vol. 6, no. 4, 324-330, 2011.
[138] S. Kochetov, and G.Wollenberg, Stable and effective full-wave PEEC models by full-spectrum
convolution macromodeling, IEEE Trans. EMC, vol.49, no. 1, 25-34, 2007.
[139] S. Visacro, Modelagem de Aterramentos Elétricos, PhD Dissertation, Federal University of Rio de
Janeiro (COPPE/UFRJ), 1992 (in Portuguese).
[140] S. Visacro, A. Soares Jr., and M. A. Schroeder, An interactive computational code for simulation of
transient behavior of electric system components for lightning currents, in Proc. 2002 Int. Conf.
Lightning Protection (ICLP), pp. 732–737, 2002.

185
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[141] S. Visacro, and C. Portela, Investigation of earthing systems behavior on the incidence of
atmospheric discharges at electrical systems, in Proc. of the 20th International Conference on
Lightning Protection, Interlaken, Switzerland, 1990.
[142] S.Visacro, and C. Portela, Modelling of earthing systems for lightning protection applications
including propagation effects, in Proc. of the 21st International Conference on Lightning Protection,
Berlin, Germany, 1992.
[143] S. Visacro, and A. Soares Jr., Simplified models for transmission line tower-footing, In Proc. of the
23rd International Conference on Lightning Protection, Firenze, Italy, 1996.
[144] S. Visacro, Evaluation of communication tower grounding behavior on the incidence of lightning, in
Proc. of the 24th International Conference on Lightning Protection, Birmingham, England, 1998.
[145] F. H. Silveira, S. Visacro, and C. R. Mesquita, Evaluation of the influence of lightning channel and
return current characteristics on induced overvoltages, in Proc. of ther 26th International Conference
on Lightning Protection (ICLP), pp. 191-196, Krakow, Poland, 2002.
[146] F. H. Silveira, S. Visacro, A. Soares Jr., M. A. O Schroeder, A. De Conti, and M. H. M. Vale, The
influence of transmission line configuration on the amplitude of lightning induced overvoltages, in
Proc. of 26th International Conference on Lightning Protection (ICLP), pp. 197-201, Krakow, Poland,
2002.
[147] F. H. Silveira, Modelagem para Cálculo de Tensões Induzidas por Descargas Atmosféricas, PhD
Dissertation, LRC/PPGEE – Federal University of Minas Gerais, Belo Horizonte, Brazil, 2006 (in
Portuguese).
[148] F. H. Silveira, and S. Visacro, Lightning effects in the vicinity of elevated structures, J. Electrostatics,
vol. 65, no. 5/6, pp. 342-349, 2007.
[149] K. A. Norton, The propagation of radio waves over the surface of the earth and in the upper
atmosphere, part II – The propagation from vertical, horizontal, and loop antennas over a plane
earth of finite conductivity, Proceedings of the Institute of Radio Engineers – IRE, vol. 25, no. 9, pp.
1203-1236, 1937.
[150] F. H. Silveira, S. Visacro, J. Herrera, and H. Torres, Evaluation of lightning-induced voltages over
lossy ground by the hybrid electromagnetic model, IEEE Trans. EMC, vol. 51, no. 1, pp.156–160,
2009.
[151] S. Visacro, R. Alipio, M. H. Murta Vale, and C. Pereira, The response of grounding electrodes to
lightning currents: the effect of frequency-dependent soil resistivity and permittivity, IEEE Trans.
EMC, vol. 53, no. 2, pp. 401–406, 2011.
[152] S. Visacro and R. Alipio, "Frequency dependence of soil parameters: experimental results,
predicting formula and influence on the lightning response of grounding electrodes," IEEE Trans.
Power Del., vol. 27, no. 2, pp. 927-935, Apr. 2012.
[153] S. Visacro, F. H. Silveira, S. Xavier, and H. B. Ferreira, “Frequency Dependence of Soil Parameters:
The Influence on the Lightning Performance of Transmission Lines,” in Proc. International
conference on lightning protection (ICLP) 2012, paper 116, 2012.
[154] F. H. Silveira, S. Visacro, H. B. Ferreira, A. De Conti, Lightning-induced voltages calculated with the
hybrid electromagnetic model considering frequency-dependent soil parameters, in Proc. of 2013
International Symposium on Lightning Protection (XII SIPDA), pp. 308-312, 2013.
[155] S. Visacro, and F. H. Silveira, The impact of the frequency dependence of soil parameters on the
lightning performance of transmission lines, IEEE Trans. EMC, vol. 57, no.3, pp. 434-441, 2015.
[156] S. Visacro, and F. H. Silveira, Lightning performance of transmission lines: Methodology to design
grounding electrodes to ensure an expected outage rate, IEEE Trans. PWRD, vol. 30, p. 237-245,
2015.
[157] F. H. Silveira, S. Visacro, R. Alipio, and A. De Conti, Lightning-induced voltages over lossy ground:
The effect of frequency dependence of electrical parameters of soil, IEEE Trans. EMC, vol. 56, p.
1129-1136, 2014.
[158] F. H. Silveira, and S. Visacro, Lightning performance of transmission lines: Impact of current
waveform and front time on backflashover occurrence, IEEE Trans. PWRD, doi:
10.1109/TPWRD.2019.2897892 (in press).
[159] M. Ishii, T. Kawamura, T. Kouno, E. Ohsaki, K. Shiokawa, K.Murotani, and T. Higuchi, Multistory
transmission tower model for lightning surge analysis, IEEE Trans. PWRD, vol. 6, no. 3, pp. 1327–
1335, 1991.
[160] M. Ishii, K. Michishita, Y. Hongo, and S. Ogume, Experimental study of lightning-induced voltage on
an overhead wire over lossy ground, IEEE Trans. EMC, vol. 41, no. 1, pp. 39–45, Feb. 1999.

186
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[161] A. Ametani, H. Motoyama, K. Ohkawara, H. Yamakawa, and N. Suga, “Electromagnetic

disturbances of control circuits in power stations and substations experienced in Japan,” IET
Generation, Transmission & Distribution, vol. 3, no. 9, pp. 801-815, 2009
[162] A. Tatematsu, K. Yamazaki, K. Miyajima, and H. Motoyama, “A study on induced voltages on an
aerial wire due to a current flowing through a grounding grid,” IEEJ Trans. Power and Energy
Society, vol. 129, no. 10, pp. 1245-1251, 2009.
[163] G. Ala, P. L. Buccheri, P. Romano, and F. Viola, F., Finite difference time domain simulation of earth
electrodes soil ionisation under lightning surge condition, IET Science, Measurement and
Technology, vol. 2 (3), pp. 134-145, 2008.
[164] K. Otani, Y. Shiraki, Y. Baba, N. Nagaoka, A. Ametani, and N. Itamoto, “FDTD surge analysis of
grounding electrodes considering soil ionization,” Electric Power Systems Research, vol. 113, pp.
171-179, 2014.
[165] H. Tanaka, D. Tanahashi, Y. Baba, and N. Nagaoka, FDTD surge simulation of grounding
electrodes considering soil ionization, Paper presented at the 9th Asia-Pacific International
Conference on Lightning, No. 1136, 7 pages, Nagoya, Japan, 2015.
[166] G. Ala and M.L. Di Silvestre, "A simulation model for electromagnetic transients inlightning
protection systems," IEEE Trans. Electromagn. Compat. vol. 44, no. 4, pp. 539–554, 2002.
[167] K. Sheshyekani S. H. H. Sadeghi, and R. Moini, A combined MoM-AOM approach for frequency
domain analysis of nonlinearly loaded antennas in the presence of a lossy ground, IEEE
Transactions on Antennas and Propagation, vol. 56 (6), pp. 1717-1724, 2008.
[168] L. Grcev, Time- and frequency-dependent lightning surge characteristics of grounding electrodes,
IEEE Transactions on Power Delivery, vol. 24 (4), pp. 2186-2196, 2009.
[169] A. C. Liew, and M. Darveniza, Dynamic model of impulse characteristics of concentrated earth,
Proceedings of IEE, vol. 121 (2), pp. 123-135, 1974.
[170] S. Sekioka, H. Hayashida, T. Hara, and A. Ametani, A., Development of a nonlinear grounding
resistance model for high impulse currents (in Japanese), Transactions of IEE Japan, vol. 118 (1),
pp. 92-101, 1998.
[171] A. Geri, “Behaviour of grounding systems excited by high impulse currents: The model and its
validation,” IEEE Trans. Power Del., vol. 14(3), pp. 1008–1017, Jul. 1999.
[172] Y. Asaoka, H. Motoyama, H. Matsubara, and H. Sugimoto, Development of grounding resistance
analysis model of rod electrode considering the effect of large-current characteristic for distribution
lines (in Japanese), IEEJ Transactions on Power and Energy, vol. 125 (10), pp, 979-987, 2005.
[173] S. Sekioka, M. I. Lorentzou, M. P. Philippakou, and J. M. Prousalidis, Current-dependent grounding
resistance model based on energy balance of soil ionization, IEEE Transactions on Power Delivery,
vol. 21 (1), pp. 194-201, 2006.
[174] R. Zeng, X. Gong, J. He, B. Zhang, and Y. Gao, Lightning impulse performances of grounding grids
for substations considering soil ionization, IEEE Transactions on Power Delivery, vol. 23 (2), pp.
667-675, 2008.
[175] Z. Mazloom, N. Theethayi, and R. Thottappillil, Modeling indirect lightning strikes for railway systems
with lumped components and nonlinear effects, IEEE Transactions on Electromagnetic
Compatibility, vol. 53 (1), pp. 250-252, 2011.
[176] P. L. Bellaschi, R. E. Armington, and A. E. Snowden, Impulse and 60-cycle characteristics of driven
grounds – II, Transactions on American Institute of Electrical Engineers, vol. 61 (6), pp. 349 – 363,
1942.
[177] P. Yutthagowith, A. Ametani, F. Rachidi, N. Nagaoka and Y. Baba, " Application of a partial element
equivalent circuit method to lightning surge analyses," Electric Power System Research Journal, vol.
94(4), pp. 30-37, Jan. 2013.
[178] L. Grcev, "Computer analysis of transient voltages in large grounding systems," IEEE Trans. Power
Del., vol. 11(2), pp. 815-823, Apr. 1996.
[179] L. Grcev and M. Heimbach, "Frequency dependent and transient characteristics of substation
grounding systems," IEEE Trans. Power Del., vol. 12(1), pp. 172-178, Jan. 1997.
[180] Z. Stojkovic, M. S. Savic, J. M. Nahman, D. Salamon, and B. Bukorovic, “Sensitivity analysis of
experimentally determined grounding grid impulse characteristics,” IEEE Trans. Power Del., vol.
13(4), pp. 1136–1142, Oct. 1998.

187
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[181] T. H. Thang, Y. Baba, N. Nagaoka, A. Ametani, J. Takami, S. Okabe, and V. A. Rakov, FDTD
simulation of lightning surges on overhead wires in the presence of corona discharge, IEEE Trans.
EMC, vol. 54 (6), pp. 1234–1243, 2012.
[182] A. Inoue, Study on propagation characteristics of high-voltage traveling waves with corona
discharge, (in Japanese) CRIEPI Rep. 114, p. 123, 1983.
[183] T. Ueda, “Investigation of effect of transmission line arresters verified by lightning surges at
substations,” T. IEE Japan, vol. 177-B, no. 10, pp. 1389-1396, 1997. (in Japanese)
[184] Subcommittee for Power stations and Substations, Study Committee on Lightning Risk, “Guide to
Lightning Protection Design of Power Stations, Substations and Underground Transmission Lines
(rev.2011),” CRIEPI Report, no. H06, 2012.
[185] J. Takami, T. Tsuboi, K. Yamamoto, S. Okabe, Y. Baba, “FDTD simulation considering an AC
operating voltage for air-insulation substation in terms of lightning protective level”, IEEE
Transactions on Dielectrics and Electrical Insulation, Vol. 22, No.2, pp. 806-814, April 2015.
[186] A. Tatematsu, “Development of a Surge Simulation Code VSTL REV Based on the 3D FDTD
Method,” in Proc. Joint IEEE Int. Symp. EMC and EMC Europe, pp. 1111-1116, 2015.
[187] S. Visacro, F.H. Silveira, "Lightning Performance of Transmission Lines: Requirements of Tower-
Footing Electrodes Consisting of Long Counterpoise Wires", IEEE Trans. Power Del. vol. 31, no. 4,
pp. 1524-1532, Aug. 2016.
[188] A. De Conti and S. Visacro, “Analytical representation of single- and double-peaked lightning current
waveforms,” IEEE Trans. Electromagn. Compat., vol.49, No.2, May 2007.
[189] R. B. Anderson and A. J. Eriksson, “Lightning parameters for engineering application,” Electra,
vol.69, pp. 65-102, 1980.
[190] A. H. Hileman, "Insulation Coordination for Power Systems". Boca Raton, FL: CRC, 1999, pp. 627-
640.
[191] IEEE, Guide for Improving the Lightning Performance of Transmission Lines, IEEE Std. 1243, 1997.
[192] S. Visacro, "What Engineers in Industry Should Know about the Response of Grounding Electrodes
Subject to Lightning Currents"; IEEE Transactions on Industry Applications, vol. 51, no. 6, p.4943-
4951, Nov./Dec. 2015.
[193] S. Visacro, "A comprehensive approach to the grounding response to lightning currents," IEEE
Trans. Power Del., vol. 22, no. 1, pp. 381-386, Jan. 2007
[194] S. Visacro, F. H. Silveira, and T. V. Magalhães, “On Concise Representations of Grounding
Electrodes for Assessment of the Lightning Performance of Transmission Lines,” in Proc.
International conference on lightning protection (ICLP) 2016, paper 230, 2016.
[195] S. Visacro, “The use of the impulse impedance as a concise representation of grounding electrodes
in lightning protection application”, IEEE Transaction on Electromagn. Compatibility, Vol. 60 , no. 5,
pp. 1602 – 1605, Oct. 2018.
[196] S. Yokoyama, K. Miyake, H. Mitani, N. Yamazaki, “Advanced Observations of Lightning Induced
Voltage on Power Distribution Lines,” IEEE Trans. Power Systems, vol. PWRD-1, no. 2, pp. 129-
139, Apr. 1986.
[197] S. Yokoyama, K. Miyake, S. Fukui, “Advanced Observations of Lightning Induced Voltage on Power
Distribution Lines (II),” IEEE Trans. Power Delivery, vol. 4, no. 4, pp. 2196-2203, Oct. 1989.
[198] F. Rachidi, “A Review of Field-to-Transmission Line Coupling Models With Special Emphasis to
Lightning-Induced Voltages on Overhead Lines,” IEEE Trans. Electromagnetic Compatibility, vol. 54,
no. 4, pp. 898-911, Aug. 2012.
[199] S. Yokoyama, “Calculation of Lightning-Induced Voltages on Overhead Multiconductor Systems,”
IEEE Trans. Power Apparatus and Systems, vol. PAS-103, no. 1, pp. 100-108, Jan. 1984.
[200] S. Yokoyama, “Distribution Surge Arrester Behavior due to Lightning Induced Voltages,” IEEE
Trans. Power Delivery, vol. PWRD-1, no. 1, pp. 171-178, Jan. 1986.
[201] S. Rusck, “Induced lightning overvoltages on power-transmission lines with special reference to the
overvoltage protection of low voltage networks,’’ Trans. Roy. Inst. Technol. Stockholm, vol. 120,
1958, pp. 1-118.
[202] A. Tatematsu, T. Noda, and H. Motoyama, “Simulation of induced voltages on an aerial wire due to
a current through a buried bare wire using the FDTD method,” in Proc. of the International
Conference on Lightning Protection (ICLP), vol. 1, pp. 459-464, 2006.

188
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[203] F. Rachidi, C.A. Nucci, M. Ianoz, C. Mazzetti, “Response of multiconductor power lines to nearby
lightning return stroke electromagnetic Fields”, IEEE Trans. on Power Delivery, 12(3), pp. 1404-
1411, 1997.
[204] F. Rachidi, C.A. Nucci, M. Ianoz, “Transient analysis of multiconductor lines above a lossy Ground”,
IEEE Trans. on Power Delivery, 14(1), pp. 294-302, 1999.
[205] M. Paolone, C.A. Nucci, E. Petrache and F. Rachidi, "Mitigation of Lightning-Induced Overvoltages
in Medium Voltage Distribution Lines by Means of Periodical Grounding of Shielding Wires and of
Surge Arresters: Modelling and Experimental Validation", IEEE Trans. on Power Delivery, Vol. 19,
No. 1, pp. 423-431, Jan. 2004.
[206] T. H. Thang, Y. Baba, V. A. Rakov, and A. Piantini, FDTD computation of lightning-induced voltages
on multi-conductor lines with surge arresters and pole transformers, IEEE Trans. EMC, vol. 57 (3),
pp. 442–447, 2015.
[207] H. Sumitani, T. Takeshima, Y. Baba, N. Nagaoka, A. Ametani, S. Okabe, and V. A. Rakov, 3-D
FDTD computation of ligh voltages on an overhead two-wire distribution line, IEEE Trans. EMC, vol.
54 (5), pp. 1161–1168, 2012.
[208] A. Piantini, J. M. Janiszewski, A. Borghetti, C. A. Nucci, and M. Paolone, A scale model for the study
of the LEMP response of complex power distribution networks, IEEE Trans. PWRD., vol. 22 (1), pp.
710–720, 2007.
[209] M. A. Uman and D. K. McLain, The magnetic field of the lightning return stroke, J. Geophys. Res.,
vol. 74, pp. 6899–6910, 1969.
[210] T. H. Thang, Y. Baba, A. Piantini, and V. A. Rakov, Lightning-induced voltages in the presence of
nearby buildings: FDTD simulation versus small-scale experiment, IEEE Trans. EMC, vol. 57 (6), pp.
1601–1607, 2015.
[211] A. Piantini and J. M. Janiszewski, Lightning induced voltages on distribution lines close to buildings,
in Proc. 25th Int. Conf. Lightning Protection, pp. 558–563, 2000.
[212] Y. Baba and V. A. Rakov, “Voltages induced on an overhead wire by lightning strikes to a nearby tall
grounded object,” IEEE Trans. Electromagn. Compat., vol. 48, no. 1, pp. 212–224, Feb. 2006.
[213] G. J. Burke, and A. J. Poggio “Numerical electromagnetics code (NEC) – method of moment,”
Lawrence Livermore Laboratory, Jan. 1981.
[214] R. K. Pokharel, M. Ishii, and Y. Baba, “Numerical electromagnetic analysis of lightning-induced
voltage over ground of finite conductivity,” IEEE Trans. Electromagn. Compat., vol. 45, no. 4, Nov.
2003.
[215] COMSOL Multiphysics Software Package. (2011), http://www.comsol.com.
[216] M. Akbari, K. Sheshyekani, A. Pirayesh, F. Rachidi, M. Paolone, and C.A. Nucci, “Evaluation of
Lightning Electromagnetic Fields and Their Induced Voltages on Overhead Lines Considering the
Frequency-Dependence of Soil Electromagnetic Parameters” IEEE Trans. Electromagn. Compat.,
vol. 55, no. 6, pp. 1210–1219, Dec. 2013.
[217] F. H. Silveira, A. De Conti, S. Visacro, Voltages induced in single-phase overhead lines by first and
subsequent negative lightning strokes: influence of the periodically grounded neutral conductor and
the ground resistivity. IEEE Transactions on Electromagnetic Compatibility, v. 53, p. 414-420, 2011.
[218] F. H. Silveira, S. Visacro, On the Lightning-Induced Voltage Amplitude: First Versus Subsequent
Negative Strokes. IEEE Transactions on Electromagnetic Compatibility, v. 51, p. 741-747, 2009.
[219] F. H. Silveira, S. Visacro, The Influence of Attachment Height on Lightning-Induced Voltages. IEEE
Transactions on Electromagnetic Compatibility, v. 50, p. 743-747, 2008.
[220] F. H. Silveira, S. Visacro, Lightning effects in the vicinity of elevated structures. Journal of
Electrostatics, v. 65, p. 342-349, 2006.
[221] S. Visacro, F.H. Silveira, Evaluation of current distribution along the lightning discharge channel by a
hybrid electromagnetic model. Journal of Electrostatics, v. 60, p. 111-120, 2004.
[222] R. Alipio, S. Visacro, Modeling the Frequency Dependence of Electrical Parameters of Soil, IEEE
Trans. Electromagn. Compat. vol. 56, no. 5, pp. 1163 – 1171, Oct. 2014.
[223] F. Heidler, “Analytische blitzstromfunktion zur LEMP-berechnung,” in in Proc. 18th Int. Conf. Lightn.
Protec., Munich, Germany, Sep. 1985, pp. 63–66.
[224] V. A. Rakov, M. A. Uman, K. J. Rambo, M. I. Fernandez, R. J. Fisher, G. H. Schnetzer, R.
Thottappillil, A. Eybert-Berard, J. P. Berlandis, P. Lalande, A. Bonamy, P. Laroche, and A. Bondiou-
Clergerie, “New insights into lightning processes gained from triggered-lightning experiments in
Florida and Alabama,” J. Geophys. Res., vol. 103, no. D12, pp. 14117– 14139, 1998.

189
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[225] I. Cotton, B. Mcniff, T. Soerenson, W. Zischank, P. Christiansen, M. Hoppe-Kilpper, S. Ramakers, P.

Pettersson and E. Muljadi: “Lightning Protection for Wind Turbines”, in Proc. 25th International
Conference on Lightning Protection, pp. 848–853, Rhodes, Greece (2000)
[226] IEC 61400-24, Wind Turbine Generator Systems-Part 24: Lightning protection (2010)
[227] NEDO, “Wind Turbine Failures and Troubles Investigating Committee Annual Report”, (2006) (in
Japanese).
[228] NEDO, “Wind Turbine Failures and Troubles Investigating Committee Annual Report”, (2007) (in
Japanese).
[229] J. Ribrant et al.: “Survey of Failures in Wind Power systems With Focus on Swedish Wind Power
Plants During 1997-2005”, IEEE Transactions on Energy Conversion, Vol. 22, No. 1, pp 167-173
(2007-3)
[230] F. Rachidi et al.: “A Review of Current Issues in Lightning Protection of New-Generation Wind-
Turbine Blades”, IEEE Transactions on Industrial Electronics, Vol. 55, No. 6, pp 2489-2496 (2008-6)
[231] K. Yamamoto, T. Noda, S. Yokoyama and A. Ametani: “Experimental and Analytical Studies of
Lightning Overvoltages in Wind Turbine Generator Systems”, Electric Power Systems Research Vol.
79, Issue 3, pp. 436-442, ISSN:0378-7796 (2009-3)
[232] K. Yamamoto, S. Yanagawa, S. Sekioka and S. Yokoyama, “Transient Grounding Characteristics of
an Actual Wind Turbine Generator System at a Low Resistivity Site,” IEEJ Transactions on Electrical
and Electronic Engineering, Vol. 5, No. 1, pp 21-26, 2010.
[233] N. Ueda, A. Ametani, S. Yanagawa and K. Yamamoto, “Transient Grounding Characteristics of a
Wind Turbine Generator System at a high resistivity site,” Institute of Electrical Engineers of Japan,
Research Meeting of High Voltage Engneering, HV-10-086, pp 23-28, 2010. (in Japanese)
[234] K. Yamamoto and S. Yanagawa, “Measurements of Transient Grounding Characteristics at a Wind
Turbine under Construction”, ISWL2011: 3rd International Symposium on Winter Lightning, No. 3-4,
pp. 97-101, 2011.
[235] K. Yamamoto, S. Yanagawa and T. Ueda, “Field Tests of Grounding at an Actual Wind Turbine
Generator System,” International Conference on Lightning Protection (ICLP), No. 1307, Cagliari,
Italy, 2010.
[236] K. Yamamoto and S. Yanagawa, “Experimental Studies of Transient Grounding Characteristics of a
Wind Turbine - Effectiveness of Counterpoises and a Ring Earth Electrode -,” Institute of Electrical
Engineers of Japan, Research Meeting of High Voltage Engneering, HV-10-087, pp 29-34, 2010. (in
Japanese)
[237] S. Yanagawa, D. Natsuno and K. Yamamoto, “A Measurement of Transient Grounding
Characteristics of a Wind Turbine Generator System and its Considerations”, Asia-Pacific
International Conference on Lightning (APL), pp 401-404, Chengdu, Chinea (2011-11)
[238] J. Niihara, A. Ametani and K. Yamamoto, “Transient Grounding Characteristics at a Wind Turbine
with a Counterpoise”, Asia-Pacific Symposium on Electromagnetic Compatibility (APEMC),
Singapore (2012-5)
[239] K. Yamamoto, S. Yanagawa, K. Yamabuki, S. Sekioka, S. Yokoyama, “Analytical Surveys of
Transient and Frequency Dependent Grounding Characteristics of a Wind Turbine Generator
System on the Basis of Field Tests”, IEEE Transactions on Power Delivery, Vol. 25, Issue 4, pp
3034-3043, 2010.
[240] K. Yamamoto and S. Yanagawa, “Grounding Characteristics of a Wind turbine Measured
Immediately after its undergrounding”, Asia-Pacific Symposium on Electromagnetic Compatibility
(APEMC), Singapore, 2012.
[241] K. Yamamoto, S. Yanagawa, T. Ueda, “Verifications of transient grounding impedance
measurements of a wind turbine generator system using the FDTD method”, International
Symposium on Lightning Protection (SIPDA), pp. 255-260, Brazil, 2011.
[242] S. Yanagawa, D. Natsuno and K. Yamamoto, “An Analytical Surveys of Transient Grounding
Characteristics of a Wind Turbine Generator System and its Considerations”, International
Conference on Lightning Protection (ICLP), Vienna, Austria (2012-9)
[243] K. Yamamoto, S. Sumi, S. Yanagawa and D. Natsuno, “Transient Grounding Characteristic of a 600
kW Class Wind Turbine”, 2012 CIGRE SC C4 Colloquium in Japan, No. III-4, pp.79-84, 2012.
[244] M. Tsumura, Y. Baba, N. Nagaoka and A. Ametani, “FDTD simulation of a horizontal grounding
electrode and modeling of its equivalent circuit,” IEEE Trans. Electromagnetic Compatibility, vol. 48,
no. 4, pp. 817–825 (2006)

190
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[245] N. Yoshikawa, A. Ametani and K. Yamamoto, “A Study on Grounding Characteristics of a Wind

Turbine System”, The 6th International Symposium on EMC and Transients in Infrastructures, pp.
55-58, Samui, Thailand (2013-11).
[246] A. Tatematsu, K. Yamazaki, and H. Matsumoto, “Lightning surge analysis of a microwave relay
station using the FDTD method,” IEEE Trans. Electromagn. Compat., vol. 57, no. 6, pp. 1616-1626,
Dec. 2015.
[247] Y. Suzuki, S. Araki, Y. Baba, T. Tsuboi, S. Okabe, and V. A. Rakov, “An FDTD study of errors in
magnetic direction finding of lightning due to the presence of conducting structure near the field
measuring station,” Atmosphere, 7, 92, 2016; doi:10.3390/atmos7070092.
[248] M. Aoki, Y. Baba, and V. A. Rakov, FDTD simulation of LEMP propagation over lossy ground:
Influence of distance, ground conductivity, and source parameters, J. Geophys. Res. Atmos., vol.
120, no. 16, pp. 8043-8051, 2015
[249] V. B. Somu, V. A. Rakov, M. A. Haddad, and S. A. Cummer, A study of changes in apparent
ionospheric reflection height within individual lightning flashes, Journal of Atmospheric and Solar-
Terrestrial Physics, vol. 136, pp. 66-79, 2015
[250] T. H. Tran, Y. Baba, V. B. Somu, and V. A. Rakov, “FDTD modeling of LEMP propagation in the
earth-ionosphere waveguide with emphasis on realistic representation of lightning source,” Journal
of Geophysical Research: Atmospheres, vol. 122, pp. 12918-12937, 2017.
[251] J. R. Wait, and K. P. Spies, Characteristics of the Earth-ionosphere waveguide for VLF radio waves,
National Bureau of Standards, Boulder, Colorado, NBS Tech. Note, 300, 1964
[252] E. D. Schmitter, “Remote sensing and modeling of lightning caused long recovery events within the
lower ionosphere using VLF/LF radio wave propagation, Adv. in Radio Sci., vol. 12, pp. 241–250,
2014
[253] M. Azadifar, D. Li, F. Rachidi, M. Rubinstein, G. Diendorfer, W. Schulz, H. Pichler, V. A. Rakov, M.
Paolone, D. Pavanello, "Analysis of lightning-ionosphere interaction using simultaneous records of
source current and 380 km d”stant electric field", Journal of Atmospheric and Solar-Terrestrial
Physics, Vol. 159, pp. 48-56, 2017.
[254] M. Apra, M. D’Amore, K. Gigliotti, M. S. Sarto, and V. Volpi, Lightning indirect effects certification of
a transport aircraft by numerical simulation, IEEE Trans. EMC, vol. 50 (3), pp. 513-523, 2008.
[255] E. Perrin, C. Guiffaut, A. Reineix, and F. Tristant, “Using a design-of-experiment technique to
consider the wire harness load impedances in the FDTD model of an aircraft struck by lightning,”
IEEE Trans. EMC, vol. 55 (4), pp. 747-753, 2013.
[256] Y. Hirano, S. Katsumata, Y. Iwahori, and A. Todoroki, Fracture behavior of CFRP laminate under
artificial lightning strike, Journal of the Japan Society for Composite Materials, vol. 35 (4), pp.165-
174, 2009.
[257] P. Feraboli, and M. Miller, “Damage resistance and tolerance of carbon/epoxy composite coupons
subjected to simulated lightning strike,” Composites Part A: Applied Science and Manufacturing, vol.
40 (6–7), pp.954–967, 2009.
[258] Y. Zhou, S. Fu, L.Shi, Q. Si, and Z. Huang, Experiment research of CFRP destroyed by lightning
current, Paper presented at 2014 Int. Conf. on Lightning Protection (ICLP), Shanghai, China, pp.
1303-1306, 2014.
[259] M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, FDTD simulation of lightning current
in a multilayer CFRP panel with triangular-prism cells, IEEE Trans. EMC, vol. 58 (1), pp. 327-330,
2016.
[260] M. Nakagawa, Y. Baba, H. Tsubata, T. Nishi, and H. Fujisawa, FDTD simulation of lightning current
in a CFRP panel: Comparison of the use of conductivity matrix approach with that of triangular prism
cells, IEEE Trans. EMC, vol. 58 (5), pp. 1674-1677, 2016.
[261] SAE, Aircraft lightning environment and related test waveforms, SAE-5412B, 2013.
[262] N. Tanabe, Y. Baba, N. Nagaoka, and A. Ametani, High-accuracy analysis of surges on a slanting
conductor and a cylindrical conductor by an FDTD method (in Japanese), IEEJ Trans. Power and
Energy, vol. 123 (6), pp. 725-733, 2003.
[263] R. L. Higdon, Absorbing boundary condition for difference approximations to the multi-dimensional
wave equation, Mathematics of Computation, vol. 47 (176), pp.437-459, 1986
[264] S. Imato, Y. Baba, N. Nagaoka, and N. Itamoto, FDTD analysis of the electric field of a substation
arrester under a lightning overvoltage, IEEE Trans. EMC, vol. 58 (2), pp. 615-618, 2016.

191
TB 785 - Electromagnetic computation methods for lightning surge studies with emphasis on the FDTD method

[265] A. Taflov, and S. C. Hagness, Computational Electrodynamics: The Finite-Difference Time-Domain

Method, 3rd edition, Artech House, MA, USA, 2005.
[266] B. D. Russell, S. M. Harvey, and S. L. Nilsson, “Substation electromagnetic interference part 1:
characterization and description of the transient EMI problem,” IEEE Trans. Power App. Syst., vol.
PAS-103, no. 7, pp. 1863–1870, Jul. 1984.
[267] C. M. Wiggins and S. E. Wright, “Switching transient fields in substations,” IEEE Trans. Power Del.,
vol. 6, no. 2, pp. 591–600, Apr. 1991.
[268] T. Ueda, “Measurement of disconnecting surges at 500 kV substation,” in Proc. Technical Meeting
on High Voltage Engineering, IEE Japan, HV-17-49, pp. 159–164, 2017.
[269] B. U. Musa, W. H. Siew, and M. D. Judd, “Computation of transient electromagnetic fields due to
switching in high-voltage substations,” IEEE Trans. Power Del., vol. 25, no. 2, pp. 1154–1161, Apr.
2010.
[270] A. Tatematsu, S. Moriguchi, and T. Ueda, “Switching surge analysis of an EHV air-insulated
substation using the 3-D FDTD method,” IEEE Trans. Power Del., vol. 33, no. 5, pp. 2324-2334,
Oct. 2018.
[271] CIGRE Working Group 36.04, “Guide on EMC in power plants and substations,” CIGRE Technical
Brochure, no. 124, Dec. 1997.
[272] M. P.-May, A. Taflove, and J. Baron, “FD-TD modeling of digital signal propagation in 3-D circuits
with passive and active loads,” IEEE Trans. Microw. Theory. Techn., vol. 42, no. 8, pp. 1514–1523,
Aug. 1994.
[273] Subcommittee of ground conductivity, “Ground conductivity in Japan,” T. IEE Japan and IEICE,
1969.

192