This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY 1

Evaluation of the Direct-Stroke Shielding

Performance of Substations Through Stochastic
Modeling of Lightning Attachment
Alexios I. Ioannidis , Student Member, IEEE, and Thomas E. Tsovilis , Senior Member, IEEE

Abstract—This article introduces a methodology for assessing inception and progression. Recently, lightning attachment mod-
the direct-stroke shielding performance of high-voltage substa- els have been combined with advanced simulation techniques to
tions. The proposed methodology considers both negative and
assess the lightning performance of substations [16], [17], [18].
positive lightning polarity and can be employed by any light-
ning attachment model that follows a probabilistic approach. A Rapid advancements in computing resources have enabled
stochastic lightning attachment model is used, which considers the the development of lightning attachment models based on the
lightning discharges probabilistic progression. Stochastic model concept of fractal structures. These are applied for the simulation
results on lightning discharge propagation, lightning attachment, of laboratory-scale electrical discharges [19], [20] as well as
and induced electric fields are in satisfactory agreement with field
lightning discharges [21], [22]. In addition, fractal models were
observations; comparisons are also made and discussed with re-
spect to other well-established lightning attachment models. An employed for the lightning performance assessment of common
application of the proposed methodology is performed to estimate structures [23], [24], power system components such as over-
the lightning incidence rate and shielding failure flashover rate of head transmission lines [25], [26], [27], [28] and substations
a 69 kV substation; the effect of lightning activity at the substation [29], [30], [31], and more recently to lightning incidence esti-
location is elucidated.
mation to wind turbines [32], [33].
Index Terms—Fractal structures, lightning performance, The complexity of shielding analysis and difficulties in for-
positive lightning, shielding failure, substations. mulating lightning performance assessment to substations into
a generalized methodology are mainly associated with:
1) The intricacy of the actual 3-D geometry and multiple
I. INTRODUCTION basic insulation levels and insulation types involved;
IGHTNING strikes impose a high risk of power system 2) the lack of up-to-date and reliable field data on substa-
L outages [1], [2], [3]. Substations, critical part of a power
system, shall operate with high reliability and experience a
tion failures and lightning parameters with seasonal and
geographical variation;
practically zero outage rate [4]. Thus, lightning-related studies 3) the inherent stochastic factors characterizing lightning
regarding the direct-stroke shielding system design to substa- discharge propagation.
tions and analysis of the arising surge overvoltages at substation Thus, this work contributes to the development of a methodol-
have been a subject of extensive research [5], [6], [7], [8], [9], ogy for assessing the direct-stroke shielding performance of sub-
[10]. stations through stochastic modeling of lightning attachment.
Direct-stroke shielding of substations is performed based on The proposed approach considers both negative and positive
the IEEE Std. 998 [11] adopting empirical and electrogeometric lightning polarity and can be employed by any lightning at-
(EGM) methods. Besides, several alternative physical-based tachment model that follows a probabilistic approach. In this
models of lightning attachment [12], [13], [14], [15] are de- study, a stochastic (fractal-based) lightning attachment model
scribed in the IEEE Std. 998, accounting for upward leader’s is employed, which considers the physical processes involved
in lightning attachment. The adopted stochastic model can con-
sider:
Manuscript received 21 April 2023; revised 23 June 2023; accepted 16 1) The actual 3-D geometry of substations and substation
October 2023. The work of Alexios I. Ioannidis was supported by Greece and buses operating voltage;
the European Union (European Social Fund-ESF) through the Operational Pro- 2) the stochastic progression of lightning discharges in a 3-D
gramme “Human Resources Development, Education and Lifelong Learning”
in the context of the Act “Enhancing Human Resources Research Potential by domain;
undertaking a Doctoral Research” Sub-action 2: “IKY Scholarship Programme 3) lightning incidence to substations against direct lightning
for PhD candidates in the Greek Universities.” (Corresponding author: Thomas strikes of positive polarity;
E. Tsovilis.)
The authors are with the High Voltage Laboratory, School of Electrical and 4) the dynamic variation of downward to upward leader
Computer Engineering, Aristotle University of Thessaloniki, 541 24 Thessa- velocities supported by recent field observations.
loniki, Greece (e-mail: [email protected]; [email protected]). Thus, the effects of lightning polarity, object height, and sta-
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TEMC.2023.3325994. tistical dispersion of lightning interception parameters are elu-
Digital Object Identifier 10.1109/TEMC.2023.3325994 cidated. The stochastic model validity is demonstrated against

© 2023 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
https://creativecommons.org/licenses/by/4.0/
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

field observations on lightning discharge propagation, lightning as [37]

attachment, and induced electric fields. Comparisons are also   2 
1 ln I − ln I¯
made and discussed with respect to other well-established light- f (I) = √ exp − 2 (3)
ning attachment models from literature. 2πσln I 2σln
The proposed methodology is applied to a practical 69 kV
with Ī and σ ln denoting the median value and the standard
substation scheme to estimate its lightning incidence rate and
deviation of the natural logarithm, respectively.
shielding failure flashover rate; the effects of ratio of negative to
5) IS (kA) is the minimum (critical) current causing flashover
positive lightning flashes, peak current distribution parameters,
of substation insulation, which can be roughly estimated
and ground flash density are quantified and discussed stressing
as [11], [38]
the need for acquisition of up-to-date lightning data based on
lightning location systems (LLSs) [34] and lightning strike 1.1 · BIL 0.94 · CF O · 1.1
IS = = (4)
measurements [35]. (1/2) · ZS (1/2) · ZS
where 1.1 is a factor to account for the reduction of stroke
current terminating on a conductor as compared to zero
II. METHODOLOGY FOR ASSESSING THE LIGHTNING impedance earth [11], BIL (kV) is the basic insulation level
PERFORMANCE OF SUBSTATIONS that substation buses insulators are designed to withstand,
In this section, the methodology for assessing the direct-stroke obtained based on their electrical characteristics, CFO
shielding performance of substations is formulated; it is noted (kV) is the negative polarity critical flashover voltage
that the proposed methodology can be applied by employing any of the insulation (0.94 factor is used to represent a 6%
lightning attachment model that follows a probabilistic approach reduction in CFO values in order to produce a withstand
on lightning incidence estimation. The direct lightning incidence level roughly equivalent to the BIL rating), and ZS (Ω) is
rate, ND (flashes/yr), and the shielding failure flashover rate, the surge impedance of substation buses (1/2 is used to
SFFOR (flashovers/yr), on an annual basis to a substation can account for the assumed equal split of the stroke current),
be given as which can be calculated according to the procedure de-
scribed in (11, Annex C) based on the conductor’s height
 ∞ and radius and the equivalent corona radius. Accurate
N D = NG · A (I) f (I)dI (1) estimation of IS requires electromagnetic transient (EMT)
0 simulations due to the nonstandard waveform of lightning
 ∞
overvoltage surges [39] and reflections at substation equip-
SF F OR = NG · pSF (I) A (I) f (I)dI (2)
IS
ment of high surge impedance [11].
6) pSF (p.u.) is the probability of shielding failure computed
from stochastic model simulations; pSF is defined as the
where
ratio of direct lightning strikes to substation buses to
1) NG (flashes/km2 /yr) is the ground flash density at the
the total number of direct lightning strikes to the sub-
substation location.
station’s protected and protective equipment. pSF is ob-
2) A (km2 ) is the lightning collection area of the substation.
tained through the proposed stochastic model results (see
3) A(I) of complex structures such as substations can be
Section V-A) or employing another probabilistic lightning
calculated by employing the following procedure in any
attachment model.
CAD software; in this work, the AutoCAD software was
Equations (1) and (2) can be generalized for both lightning
employed.
polarities, considering the ratio, r (p.u.), of the number of nega-
a) Lightning interception zones with radius equal to the
tive to total lightning flashes in a region. Thus, the total lightning
interception radius R are drawn from substation equip-
incidence, NDt , and shielding failure flashover, SFFORt , rates
ment and air terminals for each lightning peak current I.
are defined as
The interception radius is defined as the maximum lat-
eral distance between the air terminal and the lightning NDt = r · ND− + (1 − r) · ND+ (5)
interception point [36] and is commonly computed as a
SF F ORt = r · SF F OR− + (1 − r) · SF F OR+ (6)
function of I and object height, H; R values in this work
were obtained through stochastic lightning attachment where ND− , SFFOR− and ND+ , SFFOR+ are the corresponding
model simulations (see Section V-A). computed rates for negative and positive lightning. A typical
b) The lightning collection area, A, is equal to the ground value of r = 0.9 [40] is commonly adopted; nevertheless, r may
surface area defined by the contour of the union of the vary depending on seasonal and geographical characteristics.
drawn circles of step 1. ND and SFFOR values depend significantly on the lightning
c) Collection area values are estimated as a function of activity characteristics at the substation location [NG , f(I), r]
the lightning peak current; A(I) in (1) and (2) is the as well as on the shielding effectiveness of the adopted air-
fitting curve of the obtained values per I. termination system (pSF ). Thus, the need for acquisition of
4) f(I) is the lightning peak current distribution with a proba- accurate lightning data based on reliable LLSs is important. A
bility density function, which is log-normally distributed flowchart of the methodology is presented in Fig. 1.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IOANNIDIS AND TSOVILIS: EVALUATION OF THE DIRECT-STROKE SHIELDING PERFORMANCE OF SUBSTATIONS 3

over-relaxation (SOR) iterative procedure [41], [42]

[n+1] [n] ∗
Vi,j,k = (1 − ω) · Vi,j,k + ω · Vi,j,k (7a)
∗ [n] [n+1] [n]
Vi,j,k = 1/6 · (Vi+1,j,k + Vi−1,j,k + Vi,j+1,k
and [n+1] [n] [n+1] (7b)
+Vi,j−1,k + Vi,j,k+1 + Vi,j,k−1 )
where ω is the over-relaxation parameter (1 ≤ ω < 2, 1.85 in
this study). ω can be determined with the aid of theoretical equa-
tions [41] or through dedicated studies. Appropriate boundary
conditions were selected comprising the following:
1) Dirichlet type (V = const.) for the cloud, ground, air-
termination system, and substation equipment;
2) Neumann type (∂V /∂x = 0, ∂V /∂y = 0) for the lateral
boundaries.
Electric potential calculation comprises a significant task in
the employed model as it drives all the physical calculations
involved; thus, accurate and fast computations are important.
Accurate estimation of the electric potential requires the use of
a very fine grid spacing for the domain discretization as well as a
low converging tolerance in FDM and SOR; this may lead to an
excessively large number of grid points, and thus, computational
cost as also reported in [20] and [22]. Hence, in this work, code
optimization techniques, developed in [42], were employed to
significantly reduce the total simulation time (approximately
1–3 hours per simulation run) retaining at the same time the
Fig. 1. Flowchart of the proposed methodology.
accuracy in electric field computations.

At each simulation step, a single node L is added to the
lightning discharge path provided that the average electric field
with the parent node L, is higher than a critical electric field, Ecr .
The proposed methodology, following a stochastic approach, The next leader discharge node is selected from the following:
may assist in overcoming simplifications of the methods adopted ⎧ ⎫
⎪ (ELL )η ⎪
by the IEEE Std. 998 [11] associated with: 
⎨  n
N M
, for E LL  ≥ Ecr ⎬

p(L → L ) = (ELL )η (8)

1) Estimation of the substation lightning collection area em- ⎪
⎩ n=1 m=1 ⎪
⎭
ploying geometrical approaches; 0, for ELL < Ecr
2) estimation of the substation lightning incidence and where η is a model parameter affecting the random progression
flashover rate neglecting possible occurrence of positive of the simulated lightning discharges, N is the number of the
lightning flashes; lightning discharge nodes (n = 1 …N), and Mn (m = 1 …Mn )
3) determination of the shielding failure rate of substations is the number of the potential leader nodes that satisfy ELL
based on semiempirical purely deterministic methodolo- ≥ Ecr . The denominator in (8) refers to the summation of all
gies. possible candidate points Mn from all discharge nodes N and
as a result normalizes the propagation probability distribution.
III. STOCHASTIC MODELING OF LIGHTNING ATTACHMENT Thus, (8) confers a probabilistic character on the proposed
model to account for the several random factors during natural
A description of the employed stochastic model algorithm, lightning discharge propagation. Fig. 2(a) and (b) depicts a 2-D
developed in MATLAB, is presented in Section III-A. Model schematic representation of the next point selection procedure
parameters for simulation of lightning discharges of negative for the downward leader (DL) at two successive simulation steps
and positive polarity are given in Section III-B. y and y+1 based on the cumulative probability density function
P derived from (8). Based on Fig. 2(a), N in (8) is equal to 2
(points L1 and L2 ), thus, the new DL discharge point is selected
A. Model Description among the eligible points (empty dots) surrounding both L1 and

The 3-D simulation domain is initially discretized employing L2 ; hence, for n = 1 (point L1 ), eligible leader nodes are L 11

a uniform grid spacing. The electric potential, V, is computed and L 12 (M1 = 2), whereas for n = 2 (point L2 ), eligible leader
    
at all points of the simulation domain (i, j, k), at each algo- nodes are L 21 , L 22 , L 23 , L 24 , and L 25 (M2 = 5). Then, a
rithm step, solving the Laplace equation. For that purpose, a cumulative probability density function P is created, as shown
computational electromagnetics method is employed combining in Fig. 2(b), where P1 = p11 , P2 = p11 + p12 … P7 = 1,
the finite-difference method (FDM) solved by the successive and based on a random number m, the new leader discharge
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

TABLE I
EMPLOYED PARAMETERS FOR DOWNWARD AND UPWARD LEADERS

Fig. 2. (a) and (b) 2-D schematic representation of the next point selection
procedure for the DL at two successive simulation steps y and y +1 based on the
cumulative probability density function P derived from (8). (c) Corresponding
part of a 3-D simulation domain. Algorithm step y; L1d and L1u (full dots)
 
and L 1d and L 1u (empty dots) represent the current and possible next nodes
for discharge propagation of the downward (subscript d) and upward (subscript
u) leaders, respectively. (d) Corresponding part of a 3-D simulation domain.
Algorithm step y+1; discharge evolution to leader nodes L2d and L2u .
path per algorithm step; this approach was adopted to consider
the interaction of all leader discharge branches based on the

point is selected (for example point L 24 in Fig. 2(a), equal to dynamic recalculation of the electric potential at all points of
the new leader node L3 at simulation step y+1, Fig. 2(b); the the simulation domain after new point addition.
same algorithm steps are then continued for N = 3 to select
the next leader discharge point. It is noted that an eligible point B. Parameter Selection

L may be associated with two different probabilities from two The employed stochastic model parameters are listed in
 
neighboring leader points (as for example nodes L 11 and L 21 Table I for both lightning polarities. These parameters are based
from L1 and L2 , respectively) allowing, thus, the creation of on physical grounds of lightning discharges as well as on field
leader branches. The same procedure is also followed for the case measurements and observations.
of a 3-D simulation domain and for upward leaders’ propagation, The DL tip potential was selected according to the formula
as depicted in Fig. 2(c) and (d), illustrating a downward and an proposed by E. M. Bazelyan [44]; as an approximation, this
upward leader at two successive simulation steps (y and y+1) expression can also be used for positive lightning [45]. The
and the corresponding neighboring and eligible nodes for leader potential of the DL tips in the simulated tortuous and branched
propagation. channel was considered by means of the average discharge
After leader node selection, the potential at the newly added gradient, Ech , as Bazelyan’s formula refers to a straight leader
node is updated considering an average discharge gradient, Ech , channel. Considering the selected value of Ech , tips that deviate
based on the following formula [43]: from the primary DL channel are characterized by a reduced
VL = V0 − s · dtotal · Ech (9) leader tip potential. Thus, the differences among the charge dis-
tribution of the main channel and branches have been implicitly
where V0 is the leader potential at the starting point of the considered employing a voltage drop along the leader channel.
simulation, s is the sign of the charge carried by the leader The average critical electric field, Ecr (kV/m), was selected as
(“−” for leaders of negative polarity and “+” for leaders of the electric field at the streamer zone of the downward and
positive polarity), and dtotal is the total lightning discharge path upward leader tips, respectively, required for stable streamer
computed as a sum of all the discrete steps from the starting propagation. The average discharge gradient was chosen Ech =

point till the newly added node L . Then, the electric potentials 10 kV/m, as this value is considered representative of a typical
at the remaining points of the domain are recalculated; thus, the leader channel internal field [48].
competition among upward leaders and their interaction with The upward leader inception criterion proposed in [46] and
the DL is considered. These algorithm steps apply for both the [47] was selected as it can integrate the atmospheric conditions
downward and each one of the multiple upward leaders. and altitude variation into the proposed expression and is val-
The proposed stochastic model belongs to the single-element idated against field observations [51]. Actually, this criterion
category, i.e., only one point is added to the lightning discharge dictates the uniform grid spacing used for the 3-D domain
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IOANNIDIS AND TSOVILIS: EVALUATION OF THE DIRECT-STROKE SHIELDING PERFORMANCE OF SUBSTATIONS 5

discretization. Values of 1.4 and 2.9 m were selected for negative

[29] and positive [30] lightning, respectively, in line with the
critical streamer length for stable upward leader inception given
by Petrov and Waters [46], [47]; it is noted that as an approxima-
tion the same spatial step was adopted for the DL and upward
leaders although the descending stepped leader propagates in
significantly larger discrete steps in natural lightning [47]. These
values are small compared to the total kilometer-range lightning
discharge path, thus, the employed code acceleration techniques
[42] are vital for the required multiple stochastic simulation runs.
The velocities of the downward and upward leaders were
considered in the employed model based on their corresponding
ratio; this ratio is a parameter that considerably affects lightning
incidence results. A velocity ratio was adopted in this work and
not absolute velocity values as an engineering approach; it is
noted that most lightning attachment models from literature
consider a constant speed ratio. However, recent high-speed
video records [52], [53], [54], [55], [56] have shown a dynamic
variation of downward and upward leader velocities; actually,
velocity ratios even lower than unity were recorded [52], [54],
[55]. In this work, a dynamic velocity ratio formula is employed
based on the expression proposed in [12] and modified in [49] to
account for recent field observations on downward and upward Fig. 3. Interception cases for negative lightning. Tip-to-tip interception:
leader velocity ratios (allowing also for values lower than unity) (a) simulation result and (b) field observation; adapted from [24]. Tip-to-side
and consider the effect of atmospheric conditions. As an approxi- interception: (c) simulation result and (d) field observation; adapted from [24].
Side flash: (e) Simulation result and (f) field observation; adapted from [32].
mation, the same dynamic velocity ratio formula was employed
for both negative and positive lightning as shown in Table I.
Nevertheless, it is important to note that the adopted velocity
ratio formula can consider the effect of lightning polarity. This
below its highest tip [see Fig. 3(c)], as recently observed in nat-
is due to the fact that in case of negative lightning, upward leaders
ural lightning discharges [24], [58], [59] [see Fig. 3(d)] as well as
incept earlier (at higher altitudes of the DL) when compared to
scale-model experiments [60]; stochastic model simulations can
positive lightning [47], due to the larger upward leader inception
also predict attachment cases, although rarely recorded in field
field required in the latter case; hence, the velocity ratio variation
observations, where the UCL intercepts to the lateral surface
from upward leader inception up to the final jump phase is higher
of the DL above its tip [58], as in the lightning attachment
for negative than positive lightning, since for the former ΔE
simulation model of [61]. In addition, it can reproduce cases
(see Table I) takes generally lower values. This observation is
of side flashes [see Fig. 3(e)] as the ones recorded in recent field
in line with the modeling analysis of Dellera and Garbagnati
observations [32] [see Fig. 3(f)].
[12]; however, definitely more and reliable field data on positive
lightning attachment are needed emphasized on records of DL
growth through advanced monitoring systems [57].
B. Fractal Dimension of Lightning Discharges
The fractal dimension is a parameter that is used to quan-
IV. MODEL RESULTS AND COMPARISON WITH FIELD DATA
tify the complexity of a fractal pattern characterized by self-
In this section, the employed stochastic lightning attachment similarity. The fractal dimension of the simulated lightning dis-
model is validated against field observations and data. In addi- charges was estimated based on the box-counting and sandbox
tion, results of the stochastic model regarding basic shielding algorithms [62], [63]. The employed fractal estimation methods
design parameters (striking distance and interception radius) were used to compute the 2-D fractal dimension of the simulated
are compared with well-established lightning attachment models 3-D lightning discharges as the mean value from different projec-
from literature. tions [63]. Thus, comparison with photographic records of natu-
ral lightning is enabled. Fractal growth of the simulated lightning
discharges depends strongly on the propagation parameter, η,
A. Field Observations of Lightning Attachment
in (8). Thus, to allow for proper parameter selection, Fig. 4
The stochastic model predicts predominantly lightning attach- depicts the variation of fractal dimension with the propagation
ment through a tip-to-tip connection of downward and upward parameter; based on Fig. 4, η was selected equal to 8 (dotted
leaders as shown in Fig. 3(a) and (b). It must be noted that there frame) as for this value the associated mean fractal dimension
are simulation results indicating cases, where the DL intercepts and its statistical dispersion lie within the limits (1.1–1.4) [64]
to the lateral surface of the upward connecting leader (UCL) reported in field observations of natural lightning.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

Fig. 6. Adopted definitions of striking distance. (a) Inception, Sinc . (b) Inter-
ception, Sint . R is the interception radius and H the object height.

Fig. 4. 2-D fractal dimension, Df , as a function of the propagation parameter

η; negative lightning, cloud-to-ground (CG). Solid and dashed vertical bars
denote the maximum and minimum limits for the box-counting and sandbox
method, respectively. Horizontal lines indicate the recorded fractal dimension
values (1.1–1.4) based on field observations [64]; 100 simulations runs per η.

Fig. 7. Striking distance results based on the stochastic model and field data;
negative lightning, 500 runs per current level. Empty points: Simulation results,
and full points: Field data. Results with green color refer to the inception
definition.

D. Striking Distance Field Data

In this work, the following two interpretations of the striking
distance parameter were taken into account:
1) Sinc , which represents the distance of the DL and the air-
terminal at the moment when the UCL begins [as depicted
in Fig. 6(a)];
Fig. 5. Lightning interception points of a 60 m tower for positive lightning. 2) Sint , which is the distance between the point where the
Blue dots denote stochastic model simulation results and red crosses field data
from Nakamura et al. [66] derived from rocket-triggered lightning to a 60 m DL and UCL intercept and the air-terminal [as shown in
transmission tower (1000 runs per lightning current level). Fig. 6(b)].
Fig. 7 depicts stochastic model results on striking distance
together with field data on natural lightning (first strokes) to
60 m towers from Eriksson [67] and reproduced by Uman [68]
C. Lightning Interception Points and Visacro et al. [69], referring to inception definition; this
Stochastic model predicts different lightning interception figure also shows striking distance data of Saba et al. [53]
points in every simulation run. Interception points for the case derived from natural lightning (subsequent strokes) to a 52 m
of a single lightning mast create an interception area above the building derived based on both striking distance definitions. It
air-terminal that increases with increasing lightning peak current should be mentioned that although striking distance field data
and has a statistical notion [65], as shown in Fig. 5. Dots in are associated with a huge dispersion even for similar lightning
Fig. 5 depict the various interception points for the case of a peak currents, field data lie within the range of striking distance
60 m tower for positive lightning polarity; there are few cases values obtained from stochastic modeling; error bars denote
that interception occurs lower than the air-terminal tip, in line the range of striking distance values based on the total number
with field observations (side flashes) [24], [32]. It is important of simulations. Stochastic model results of Fig. 7 may explain
to note that field data from Nakamura et al. [66] (red crosses) the weak relationship between striking distance and lightning
derived from rocket-triggered lightning experiments to a 60 m return-stroke peak current also found from field observations of
transmission tower are located within the “cloud” of interception Mazur and Ruhnke [70] indicating that striking distance cannot
points (blue dots) obtained from stochastic model simulation be described by purely deterministic expressions; this nondeter-
results. ministic striking distance behavior may also be associated with
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IOANNIDIS AND TSOVILIS: EVALUATION OF THE DIRECT-STROKE SHIELDING PERFORMANCE OF SUBSTATIONS 7

Fig. 8. Electric field amplitude at ground level prior to return stroke; compar- Fig. 9. Striking distance (inception definition) according to the stochastic and
ison of stochastic model results (100 simulation runs) with field data of Jerauld other lightning attachment models from literature; mast height 21.3 m, negative
et al. [71] for different distances from the lightning termination point. Dots for lightning, and 1000 runs per current level. With green solid line the fitting
the stochastic model represent average values. Inset figure (enclosed blue dotted curve of stochastic model results (triangles); and vertical error bars denote the
frame) depicts the variation of the electric field with time. maximum and minimum striking distance values obtained through the stochastic
model.

the probabilistic variation of the downward leader velocity as

shown in [14].

E. Electric Field at Ground Level

Stochastic model results regarding the electric field amplitude
at ground level prior to return-stroke are depicted in Fig. 8 for
different distances from the lightning termination point against
field observations from Jerauld et al. [71] for a 51.9 kA negative
CG lightning. Error bars denote the minimum and maximum
values derived from stochastic model simulation runs since
the electric field at the ground level depends on the stochastic
variation of the downward leader path prior to lightning termi-
nation to the ground; the derived statistical dispersion of the Fig. 10. Maximum interception radius, R, according to the stochastic and
electric field amplitude at the ground level was also shown in other lightning attachment models from literature; mast height 21.3 m, negative
lightning, and 1000 runs per current level.
another simulation study [21]. Thus, points on the ground level
may be associated with higher electric field amplitude values if
the lightning discharge propagates predominantly above these
points during the final steps prior to lightning termination to current-dependent striking distance expression may underes-
the ground. The inset figure in Fig. 8 (enclosed blue dotted timate striking distance values especially for higher objects.
frame) depicts the time variation of the electric field, measured Nevertheless, results based on the considered models lie in the
at a distance 260 m from the lightning termination point. A statistical dispersion obtained from the stochastic model; an
satisfactory agreement was found for the average values of excellent agreement was found between stochastic model results
stochastic model results (see Fig. 8) with respect to field data, as and SLIM [73].
well as for the electric field time variation at 260 m (inset graph); The interception radius, R, represents the largest lateral dis-
this enhances the validity of the employed stochastic model in tance from the air terminals at which a lightning strike can occur
terms of the adopted electric field computational method and [as illustrated in Fig. 6(b)]. A comparison of R obtained through
model parameter selection. the stochastic model and other models from the literature is
displayed in Fig. 10; a good agreement with Rizk’s model
[15] exists. From Fig. 10, it can be observed that results on
F. Comparison With Other Lightning Attachment Models interception radius obtained through stochastic modeling are
Striking distance and interception radius values (see Fig. 6) lower than the widely adopted EGM of IEEE Std. 998 in the
obtained through stochastic modeling are compared with other low lightning peak current region (5–10 kA), and then, become
lightning attachment models from literature. Fig. 9 shows the higher for lightning peak currents in the region of 20–30 kA. A
striking distance, for a mast height of ∼20 m, based on the possible explanation for this trend is that stochastic modeling
EGM of the IEEE Std. 998 [11] as well as other well-established considering the inception and stochastic propagation of upward
lightning attachment models. The EGM employing a solely leaders, the object height, and the operating voltage of substation
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

Fig. 11. (a) Top view of the 69 kV substation under study [11]. (b) 3-D model
of the substation in MATLAB; squares denoted upward leader inception points.

Fig. 12. Shielding failure probability, pSF (p.u.), of the 69 kV substation for
buses predicts generally higher interception values than the both lightning polarities; 20 000 simulation runs in total.
EGM of IEEE Std. 998 [11]; the latter is purely deterministic and
neglects the aforementioned parameters. Nevertheless, results
from these two models are in good agreement in the low lightning
peak current region (5–10 kA) where the shielding system is
inception criterion, hence resulting in potentially larger number
designed for the majority of the substations; in this region, the
of shielding failures.
effect of upward leaders’ propagation is minimal and does not
It is noteworthy that shielding failure cases do exist for cur-
significantly influence the lightning attachment phenomenon.
rents larger than the maximum shielding failure current, IMSF ,
estimated as 7.2 kA according to the EGM of IEEE Std. 998 [75].
V. LIGHTNING PERFORMANCE ASSESSMENT OF A 69 KV Practically all shielding failure currents would cause flashover
SUBSTATION since the critical current causing flashover of substation insu-
The lightning performance of a 69 kV substation against direct lation, IS , is 2.6 kA according to (4) (BIL = 350 kV, ZS =
lightning strikes has been evaluated based on the introduced 300 Ω [11]). pSF was found significantly lower for negative
methodology; it comprises a typical substation of sectional- lightning since stable positive upward leaders can incept earlier
ized bus arrangement, commonly found in practical substation from masts when an average electric field is satisfied equal
schemes, presented in IEEE Std. 998 [11]. to the electric field required for positive streamer propagation,
500 kV/m, within a critical positive streamer length according to
[46]. For the case of positive lightning, inception of stable neg-
A. Stochastic Modeling of Lightning Attachment and Results ative upward leaders can be achieved when an average electric
The 69 kV substation under study comprises three buses of field is satisfied within a critical negative streamer length, which
different heights. As a case study, an air-termination system is more than double the corresponding one of positive upward
designed based on the fixed angle method (45° protection an- leaders, with typical electric field values required for negative
gle) was adopted; the relative position and dimensions of the streamer propagation ranging between 1000 and 1600 kV/m
lightning masts were employed based on Annex B of [11]. [76]. As a result, the positive downward leader needs to approach
Modeling of the substation was performed in MATLAB as points very closely to a mast to enhance sufficiently the electric field
of fixed electric potential considering the substation layout and in its vicinity, leading, thus, to potentially higher probability of
dimensions as well as the grid spacing employed for 3-D do- shielding failure. Thus, for the case of positive lightning, masts
main discretization (see Table I). A top view of the substation initiating late and short upward leaders exhibit shorter striking
layout and the MATLAB model employed for simulations are distance and interception radius when compared to negative
depicted in Fig. 11(a) and (b), respectively. Fig. 11(b) also shows lightning [30], [47].
eligible upward leader inception points, from the substation In addition, Fig. 12 indicates that IMSF has a probabilistic
layout, associated with an elevated electric field; points at the notion and shall be treated as a statistical parameter, rather
lateral surfaces of masts and substation components were also than determining a unique value corresponding to zero shielding
considered allowing, thus, for side flashes [see Fig. 3(e)]. failure probability based on a deterministic approach. Thus, the
Fig. 12 shows the probability of shielding failure, pSF (p.u.) employed stochastic model considering the effect of lightning
of the 69 kV substation for negative and positive lightning. polarity, object height, and random characteristics associated
Lightning discharges associated with low prospective return with lightning attachment, could predict cases of shielding fail-
stroke peak currents are less likely to initiate an UCL from an ures documented by utilities in IEEE surveys [11], [77]; these
air-terminal (masts and/ or catenary wires) due to the reduced refer to recorded unexpected cases in substations with a shielding
electric field induced to its region; thus, the downward leader system designed based on IEEE Std. 998 procedures. Simulation
need to approach closely to the direct-stroke shielding system of results of shielding failures based on stochastic modeling are
air terminals in order the upward leaders to be able to satisfy the shown in Fig. 13.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IOANNIDIS AND TSOVILIS: EVALUATION OF THE DIRECT-STROKE SHIELDING PERFORMANCE OF SUBSTATIONS 9

Fig. 13. Shielding failure cases for (a) negative and (b) positive lightning.

TABLE II
LIGHTNING ACTIVITY PARAMETERS FOR DIFFERENT REGIONS

Fig. 15. SFFOR (flashovers/yr) for the distributions of Table II and a ratio of
negative to total lightning flashes r between 0.5 and 1; NG = 10 flashes/km2 /yr.
Dotted horizontal line denotes a typical acceptable SFFOR due to direct lightning
strikes (0.05% [11]).

Fig. 14. ND (flashes/yr) for the distributions of Table II and a ratio of negative
to total lightning flashes r between 0.5 and 1; NG = 10 flashes/km2 /yr.
Fig. 16. (a) ND (flashes/yr) and (b) SFFOR (flashovers/yr) of the 69 kV
substation for different regions (see Table II). Dotted horizontal line denotes
a typical acceptable SFFOR due to direct lightning strikes (0.05% [11]).
B. Shielding Performance of the 69 kV Substation
The direct lightning incidence rate, ND (flashes/yr), and Fig. 15 demonstrates the SFFOR of the 69 kV substation
shielding failure flashover rate, SFFOR (flashovers/yr), on an an- for the distributions of Table II. SFFOR for negative lightning
nual basis to the 69 kV substation were computed [see (1)–(6)]; (r = 1) varies up to ∼70 times, among the distributions; the
the effect of lightning activity at the substation location was also highest flashover rate is estimated for distribution no. 1, which
investigated. Recorded lightning peak current distributions and exhibits the lowest median value, commonly found in temperate
lightning activity parameters for different regions around the zones. Thus, although a larger number of direct lightning flashes
world are listed in Table II. (ND ) is expected for substations in tropical regions, the SFFOR
Fig. 14 shows ND values for the distributions of Table II may be bigger in temperate zones, with the same NG , since the
and a ratio of negative to total lightning flashes, r, between probability of low-intensity currents that may cause shielding
0.5 and 1. ND for negative lightning (r = 1) varies consider- failures is higher. For the case of the higher percentage of positive
ably, up to ∼2.7 times, among the distributions; the highest lightning, thus for decreasing r, SFFOR increases drastically
incidence rate is estimated for distribution no. 4, which ex- due to the high shielding failure probability (pSF ) associated
hibits the highest median value, commonly found in tropical with positive lightning (see Fig. 12). Besides, the importance of
regions. For the case of higher percentage of positive lightning accurate knowledge of the lightning parameters at the substation
(1−r), thus for decreasing r, lightning incidence rate decreases location is stressed by the fact that the same substation config-
due to the reduced interception radius associated with positive uration for a ratio of negative to total lightning flashes r = 0.5
lightning [30], [47]. exhibits an acceptable SFFOR of 0.05% (1 per 2000 years [11])
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

for distributions no. 3 and no. 4, whereas SFFOR is higher for accuracy estimates of the SFFOR and the implementa-
the other regions with lightning peak current distributions of a tion of a fixed shielding design system to a substation
lower median value (see Table II). does not necessarily guarantee an acceptable shielding
To further demonstrate the prominent effect of lightning activ- performance at a global level. This stresses the need for
ity parameters on the substation lightning performance, Fig. 16 the acquisition of reliable and up-to-date lightning data
shows ND [see Fig. 16(a)] and SFFOR [see Fig. 16(b)] values considering the seasonal and geographical variation of
for the 69 kV substation based on lightning activity data from lightning activity parameters.
Table II (typical values per region). It is noted that the 69 kV 4) Substations in tropical regions with commonly intense
substation when located at a region such as Colombia with lightning activity and a high percentage of positive light-
intense lightning activity (NG ≥ 20 flashes/km2 /yr and r ≤ 0.8) ning need an advanced shielding system to achieve an
would exhibit an unacceptable SFFOR (> 0.05%) in contrast to acceptable SFFOR. Substations located in temperate
other locations, which exhibit a practically perfect direct-stroke zones, commonly associated with a relatively low median
shielding performance (< 1 flashover per 10 000 years). value, may experience a high SFFOR due to the higher
Nevertheless, lightning activity parameters may vary signif- probability of low-intensity currents that cause shielding
icantly even for the same region depending on seasonal and failures.
geographical factors; thus, up-to-date lightning activity studies 5) Stochastic analysis on direct-stroke shielding of substa-
are of importance. It is also noteworthy that the total lightning- tions may be adopted by international standards espe-
related failure rate of substations is not only attributed to shield- cially for mission-critical applications, such as substations
ing failures due to direct lightning strikes to the substation. feeding hospitals, data centers, and government and mil-
Incoming lightning overvoltages from the interconnected over- itary infrastructure. This is more so considering that the
head lines are the main cause of lightning-related failures, and available computing power will be significantly expanded
thus, may substantially affect the performance of substations within the next years allowing for more sophisticated
against lightning [82], [83]. modeling of lightning attachment through advanced com-
putational techniques.

APPENDIX
VI. CONCLUSION
In this Section, the methodology for the estimation of the
In this work, a methodology for assessing the direct-stroke coefficient u in the critical propagation electric field formula of
shielding performance of high-voltage substations has been Table I is presented. The methodology considers the dependence
introduced; the methodology can be utilized employing any of the negative and positive leaders on altitude and atmospheric
lightning attachment model following a probabilistic approach. conditions. The following analysis is based on a combination of
A stochastic lightning attachment model has been employed, theoretical analysis and experimental or/ and simulation results
considering the probabilistic progression of lightning discharges from literature studies [46], [50], [76].
in a 3-D domain as well as lightning polarity. The stochastic Based on [46] the dependence of pressure, p, temperature, T,
model has been validated against field observations on lightning and absolute humidity, γ, on altitude, z, can be given as
discharge propagation, lightning attachment, as well as induced
electric fields. An application to a practical 69 kV substation p(z) = p(0) · exp(−z/z0 ) (A1)
configuration has been made. Results of this study have revealed where p(0) = 760 mmHg is the pressure at ground level and z0
the following: a constant equal to 8 km
1) Striking distance shall not be treated as a deterministic
but as a statistical quantity depending on object height T (z) = T (0) − k · z (A2)
and lightning polarity. Thus, the statistical dispersion of by assuming a linear temperature change where T(0) = 293 K
striking distance field records can be explained. is the temperature at ground level and k a constant equal to 6 K
2) Shielding failure flashover events are rare but of high km−1 ,
impact; the employed model, following a stochastic ap-
proach for lightning discharge propagation, can predict γ(z) = γ(0) · exp(−z/zH ) (A3)
shielding failure cases (rare events) as documented in -3
where γ(0) = 11 gm is the absolute air humidity at ground level
surveys conducted by the IEEE that cannot be explained and zH a constant equal to 3 km. In addition, the electric field
by the conventional shielding analysis. Reliable field data required for stable positive streamer propagation can be given
regarding shielding failures to substations are required as [50]
 
1.37  γ
for further validation of lightning attachment models and
shielding design methods. Est+ (δ, γ) = 464δ · 1 + − 11 (A4)
3) SFFOR of substations of the same configuration and 100 δ
shielding design exhibit significant variation when located where δ is the relative air density as a function of pressure and
in different regions due to the various lightning activity temperature and standard sea-level conditions given as [46]
characteristics. Thus, the use of a universal distribution as p 293
proposed by international standards may result in reduced δ= · . (A5)
760 T
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IOANNIDIS AND TSOVILIS: EVALUATION OF THE DIRECT-STROKE SHIELDING PERFORMANCE OF SUBSTATIONS 11

[8] S. Okabe, “Lightning interaction with power substations,” in Lightning

Interaction With Power Systems-Volume2: Applications, A. Piantini, Ed.
London, U.K: Inst. of Eng.and Technol., 2020.
[9] A. Tatematsu, S. Terakuchi, T. Yanagi, T. Kamibayashi, and R. Mori,
“Lightning current simulation of 66-kV substation with power cables
using the three-dimensional FDTD method,” IEEE Trans. Electromagn.
Compat., vol. 63, no. 3, pp. 819–829, Jun. 2021.
[10] M. Alexandru, L. Czumbil, A. Polycarpou, H. Nouri, R. Andolfato, and
D. D. Micu, “Mitigation of transient ground potential rise in gas insulated
substations during very fast transient overvoltage,” Electric Power Syst.
Res., vol. 207, Jun. 2022, Art. no. 107824.
[11] IEEE Guide for Direct Lightning Stroke Shielding of Substations, IEEE
Standard 998, 2012.
[12] L. Dellera and E. Garbagnati, “Lightning stroke simulation by means of the
leader progression model. Part I. Description of the model and evaluation
of exposure of free-standing structures,” IEEE Trans. Power Del., vol. 5,
no. 4, pp. 2009–2022, Oct. 1990.
[13] F. D’Alessandro and J. R. Gumley, “A ‘collection volume method’
Fig. 17. Dependence of the critical propagation electric field of negative (red for the placement of air terminals for the protection of structures
curve) and positive (blue curve) streamers on the altitude above ground level. against lightning,” J. Electrostatics, vol. 50, no. 4, pp. 279–302,
Mar. 2001.
[14] M. Becerra and V. Cooray, “Time dependent evaluation of the lightning
By substituting (A1)–(A3) and (A5) into (A4), the coefficient upward connecting leader inception,” J. Phys. D, Appl. Phys., vol. 39,
u can be roughly estimated as u = 7.5 km and Ecr0 = 464 kV/m, pp. 4695–4702, Oct. 2006.
[15] F. A. M. Rizk, “Modeling of substation shielding against direct lightning
for streamers of positive polarity as can be deduced from the strikes,” IEEE Trans. Electromagn. Compat., vol. 52, no. 3, pp. 664–675,
parameters of the fitting curve of Fig. 17 (blue curve). Aug. 2010.
Following the same analysis also for streamers of negative [16] A. Rahiminejad and B. Vahidi, “LPM-based shielding performance anal-
ysis of high-voltage substations against direct lightning strokes,” IEEE
polarity and by considering that the dependence of the average Trans. Power Del., vol. 32, no. 5, pp. 2218–2227, Oct. 2017.
stable negative streamer propagation electric field on relative air [17] P. Sarajcev, D. Jakus, J. Vasilj, and S. Vodopija, “Application of genetic al-
density is given as [76] gorithm in designing high-voltage open-air substation lightning protection
system,” J. Electrostatics, vol. 93, pp. 43–51, Jun. 2018.
Est− (δ) = 1100 · δ (A6) [18] F. D’Alessandro, “New approach for lightning protection of substations,”
in Proc. 2nd Int. Conf. High Voltage Eng. Power Syst., 2019, pp. 1–7.
the coefficient u can be roughly estimated as u = 10 km and Ecr0 [19] A. I. Ioannidis, P. N. Mikropoulos, T. E. Tsovilis, and N. Karanikiotis,
“Fractal-based approach for modeling electric breakdown of air gaps: An
= 1100 kV/m, for negative streamer polarity as can be deduced application to a 75 cm positive rod-plane gap,” in Proc. 21st Int. Symp.
from the parameters of the fitting curve of Fig. 17 (red curve). High Voltage Eng., 2019, pp. 1–6.
[20] J. Guo et al., “Validation of the lightning fractal model based on the
attachment probability experiments using rod-rod air gaps,” IEEE Trans.
ACKNOWLEDGMENT Power Del., vol. 37, no. 6, pp. 5133–5144, Dec. 2022.
[21] W. Shi, Q. Li, and L. Zhang, “A stepped leader model for lightning in-
Results presented in this work have been produced using the cluding charge distribution in branched channels,” J. Appl. Phys., vol. 116,
Aristotle University of Thessaloniki High-Performance Com- no. 10, Sep. 2014, Art. no. 103303.
puting Infrastructure. The authors would like to thank Prof. [22] S. Xie, F. D’Alessandro, and X. Zhao, “A three-dimensional downward
leader model incorporating geometric and physical characteristics,” Elec-
N. Kantartzis for the fruitful discussions and assistance with tric Power Syst. Res., vol. 163, pp. 10–17, Oct. 2018.
the employed computational electromagnetics methods. The [23] N. I. Petrov, G. N. Petrova, and F. D’Alessandro, “Quantification of
publication of the article in OA mode was financially supported the probability of lightning strikes to structures using a fractal ap-
proach,” IEEE Trans. Dielectrics Elect. Insul., vol. 10, no. 4, pp. 641–654,
by HEAL-Link. Aug. 2003.
[24] R. Jiang et al., “Simulation of cloud-to-ground lightning strikes to struc-
REFERENCES tures based on an improved stochastic lightning model,” J. Atmospheric
Sol. Terr. Phys., vol. 203, Jul. 2020, Art. no. 105274.
[1] E. Volpov and E. Katz, “Characterization of local environmental data and [25] J. He, X. Zhang, L. Dong, R. Zeng, and Z. Liu, “Fractal model of lightning
lightning-caused outages in the IECo transmission-line network,” IEEE channel for simulating lightning strikes to transmission lines,” Sci. China
Trans. Power Del., vol. 31, no. 2, pp. 640–647, Apr. 2016. E, Technol. Sci., vol. 52, pp. 3135–3141, Dec. 2009.
[2] J. Bao, X. Wang, Y. Zheng, F. Zhang, X. Huang, and P. Sun, “Lightning [26] J. Li, Q. Yang, W. Sima, C. Sun, T. Yuan, and M. Zahn, “A new estimation
performance evaluation of transmission line based on data-driven lightning model of the lightning shielding performance of transmission lines using
identification, tracking, and analysis,” IEEE Trans. Electromagn. Compat., a fractal approach,” IEEE Trans. Dielectrics Elect. Insul., vol. 18, no. 5,
vol. 63, no. 1, pp. 160–171, Feb. 2021. pp. 1712–1723, Oct. 2011.
[3] S. Ekisheva, M. Papic, M. G. Lauby, and M. Elkins, “Assessment of impact [27] A. Rahiminejad, B. Vahidi, and J. He, “A fractal-based stepped down-
of AC circuit attributes to outage frequency in the WECC system,” IEEE ward leader model including branched channel charge distribution
Trans. Power Del., vol. 36, no. 2, pp. 1159–1172, Apr. 2021. and branch fading,” Electric Power Syst. Res., vol. 176, Nov. 2019,
[4] A. R. Hileman, “Shielding of substations,” in Insulation Coordination for Art. no. 105940.
Power Systems. Boca Raton, FL, USA: CRC Press, 1999, pp. 275–312. [28] Z. G. Datsios, A. I. Ioannidis, T. A. Papadopoulos, and T. E. Tsovilis, “A
[5] A. P. S. Meliopoulos and G. J. Cokkinides, “Substation lightning shielding stochastic model for evaluating the lightning performance of a -400kV
and risk assessment,” Eur. Trans. Elect. Power, vol. 13, no. 6, pp. 407–412, HVDC overhead line,” IEEE Trans. Electromagn. Compat., vol. 63, no. 5,
Nov. 2003. pp. 1433–1443, Oct. 2021.
[6] W. Nowak and R. Tarko, “Computer modelling and analysis of lightning [29] A. I. Ioannidis and T. E. Tsovilis, “Fractal-based approach for evaluating
surges in HV substations due to shielding failure,” IEEE Trans. Power the shielding design of high voltage substations against direct lightning
Del., vol. 25, no. 2, pp. 1138–1145, Apr. 2010. strikes,” in Proc. IEEE Ind. Appl. Annu. Meeting, 2020, pp. 1–8.
[7] P. N. Mikropoulos, T. E. Tsovilis, and S. G. Koutoula, “Lightning perfor- [30] A. I. Ioannidis and T. E. Tsovilis, “Shielding failure of high voltage
mance of distribution transformer feeding GSM base station,” IEEE Trans. substations: A fractal-based approach for negative and positive lightning,”
Power Del., vol. 29, no. 6, pp. 2570–2579, Dec. 2014. IEEE Trans. Ind. Appl., vol. 57, no. 3, pp. 2317–2325, May/Jun. 2021.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY

[31] J. Guo, X. Zhang, B. Wang, X. Hao, S. Zheng, and Y.-Z. Xie, “A three- [55] Q. Qi et al., “Two-dimensional striking distance of lightning flashes to a
dimensional direct lightning strike model for lightning protection of the cluster of tall buildings in Guangzhou,” J. Geophys. Res. Atmos., vol. 126,
substation,” Inst. Eng. Technol. Gener. Transmiss. Distrib., vol. 15, no. 19, no. 22, Nov. 2021, Art. no. e2021JD034613.
pp. 2760–2772, Oct. 2021. [56] M. M. F. Saba, P. B. Lauria, C. Schumann, J. C. de O. Silva, and F. D. L.
[32] X. Bian et al., “Simulation of cloud-to-ground lightning strikes to wind Mantovani, “Upward leaders initiated from instrumented lightning rods
turbines considering polarity effect based on an improved stochastic light- during the approach of a downward leader in a cloud-to-cround flash,” J.
ning model,” Atmosphere, vol. 14, no. 1, Jan. 2023, Art. no. 108. Geophys. Res. Atmos., vol. 128, no. 8, pp. 1–11, Apr. 2023.
[33] X. Bian et al., “Quantitative characteristics of the striking distance to wind [57] A. I. Ioannidis and T. E. Tsovilis, “Introducing the concept of the vol-
turbine blades based on an improved stochastic lightning model,” Inst. ume lightning strike density,” IEEE Trans. Power Del., vol. 38, no. 4,
Eng. Technol. Gener. Transmiss. Distrib., vol. 17, no. 10, pp. 2317–2330, pp. 2973–2976, Aug. 2023.
May 2023. [58] W. Lu et al., “Two basic leader connection scenarios observed in negative
[34] J. Li, L. Cai, J. Wang, M. Zhou, Y. Fan, and Q. Li, “Electrical field lightning attachment process,” High Voltage, vol. 1, no. 1, pp. 11–17,
parameters of natural return strokes at different distances,” IEEE Trans. Jul. 2016.
Electromagn. Compat., vol. 64, no. 3, pp. 786–794, Jun. 2022. [59] T. Shindo, “Lightning striking characteristics to tall structures,” IEEJ
[35] F. H. Silveira and S. Visacro, “Lightning parameters of a tropical region Trans. Elect. Electron. Eng., vol. 13, no. 7, pp. 938–947, Jul. 2018.
for engineering application: Statistics of 51 flashes measured at Morro do [60] A. Y. Kostinskiy et al., “Observations of the connection of positive and
Cachimbo and expressions for peak current distributions,” IEEE Trans. negative leaders in meter-scale electric discharges generated by clouds
Electromagn. Compat., vol. 62, no. 4, pp. 1186–1191, Aug. 2020. of negatively charged water droplets,” J. Geophys. Res. Atmos., vol. 121,
[36] P. N. Mikropoulos, J. He, and M. Bernardi, “Lightning attachment to over- no. 16, pp. 9756–9766, Aug. 2016.
head power lines,” in Lightning Interaction With Power Systems—Volume [61] A. Borghetti, F. Napolitano, C. A. Nucci, M. Paolone, and M. Bernardi,
1: Fundamentals and Modelling, A. Piantini, Ed., London, U.K.: Inst. of “Numerical solution of the leader progression model by means of the finite
Eng. and Technol., 2020. element method,” in Proc. 30th Int. Conf. Lightning Protection, 2010,
[37] CIGRE Working Group C4.407, Lightning Parameters for Engineering Art. no. 1498.
Applications, Technical Brochure 549, Aug. 2013. [62] K. Kudo, “Fractal analysis of electrical trees,” IEEE Trans. Dielectrics
[38] D. W. Gilman and E. R. Whitehead, “The mechanism of lightning flashover Elect. Insul., vol. 5, no. 5, pp. 713–727, Oct. 1998.
on high voltage and extra-high voltage transmission lines,” Electra, vol. 27, [63] A. I. Ioannidis, Z. G. Datsios, G. A. Tsaousakis, and T. E. Tsovilis,
pp. 65–96, Mar. 1973. “Analysis of the fractal dimension of lightning discharges based on a
[39] Z. G. Datsios, P. N. Mikropoulos, and T. E. Tsovilis, “Estimation of stochastic lightning attachment simulation model,” in Proc. 36th Int. Conf.
the minimum shielding failure flashover current for first and subsequent Lightning Protection, 2022, pp. 1–6.
lightning strokes to overhead transmission lines,” Electric Power Syst. [64] Z. Kawasaki and K. Matsuura, “Does a lightning channel show a fractal?,”
Res., vol. 113, pp. 141–150, Aug. 2014. Appl. Energy, vol. 67, no. 1/2, pp. 147–158, Sep. 2000.
[40] M. Nicora et al., “Estimation of the lightning performance of over- [65] P. N. Mikropoulos and T. E. Tsovilis, “Interception probability and shield-
head lines accounting for different types of strokes and multiple strike ing against lightning,” IEEE Trans. Power Del., vol. 24, no. 2, pp. 863–873,
points,” IEEE Trans. Electromagn. Compat., vol. 63, no. 6, pp. 2015–2023, Apr. 2009.
Dec. 2021. [66] K. Nakamura et al., “Lightning discharge parameters by rocket triggered
[41] R. J. LeVeque, Finite Difference Methods for Ordinary and Partial Differ- lightning to transmission line in winter,” in Proc. 10th Int. Symp. High
ential Equations. Philadelphia, PA, USA: SIAM, 2007, pp. 76–77. Voltage Eng., Montreal, QC, Canada, 1997, pp. 1–4.
[42] A. I. Ioannidis, Z. G. Datsios, A. K. Gerodimos, and T. E. Tsovilis, “Accel- [67] A. J. Eriksson, “Research paper no 4: Lightning and tall structures,” Trans.
erating stochastic lightning attachment simulations for the estimation of South Afr. Inst. Elect. Eng., vol. 69, no. 8, pp. 238–252, Aug. 1978.
lightning incidence to overhead lines,” IEEE Trans. Electromagn. Compat., [68] M. A. Uman, “Attachment Process,” in The Lightning Discharge. New
vol. 65, no. 3, pp. 839–849, Jun. 2023. York, NY, USA: Academic, 2001, p. 101.
[43] E. R. Mansell, D. R. MacGorman, C. L. Ziegler, and J. M. Straka, “Sim- [69] S. Visacro, M. Guimaraes, and M. H. M. Vale, “Striking distance de-
ulated three-dimensional branched lightning in a numerical thunderstorm termined from high-speed videos and measured currents in negative
model,” J. Geophys. Res., vol. 107, no. 9, pp. 1–12, May 2002. cloud-to-ground lightning,” J. Geophys. Res. Atmos., vol. 122, no. 24,
[44] E. M. Bazelyan, “Mechanism of orientation and parameters of lightning in pp. 13356–13369, Dec. 2017.
context of lightning protection,” Plasma Phys. Rep., vol. 45, pp. 252–263, [70] V. Mazur and L. H. Ruhnke, “Determining the striking distance of lightning
Apr. 2019. through its relationship to leader potential,” J. Geophys. Res. Atmos.,
[45] E. M. Bazelyan, personal communication, Sep. 8, 2020. vol. 108, no. D14, pp. 1–7, Jul. 2003.
[46] N. I. Petrov and R. T. Waters, “Determination of the striking distance of [71] J. Jerauld, M. A. Uman, V. A. Rakov, K. J. Rambo, D. M. Jordan, and
lightning to earthed structures,” Proc. Math. Phys. Eng., vol. 450, no. 1940, G. H. Schnetzer, “Electric and magnetic fields and field derivatives from
pp. 589–601, Sep. 1995. lightning stepped leaders and first return strokes measured at distances
[47] R. T. Waters, “Lightning phenomena and protection systems,” in Advances from 100 to 1000 m,” J. Geophys. Res. Atmos., vol. 113, no. D17, pp. 1–15,
in High Voltage Engineering, M. Haddad and D. Warne, Eds., vol. 40, Sep. 2008.
Piscataway, NJ, USA: IEE, 2004, pp. 107–114. [72] P. N. Mikropoulos and T. E. Tsovilis, “Striking distance and interception
[48] A. A. Syssoev, D. I. Iudin, A. A. Bulatov, and V. A. Rakov, “Numerical probability,” IEEE Trans. Power Del., vol. 23, no. 3, pp. 1571–1580,
simulation of stepping and branching processes in negative lightning Jul. 2008.
leaders,” J. Geophys. Res. Atmos., vol. 125, no. 7, pp. 1–28, Mar. 2020. [73] V. Cooray, U. Kumar, F. Rachidi, and C. A. Nucci, “On the possible
[49] A. I. Ioannidis and T. E. Tsovilis, “Stochastic modeling of shielding failures variation of the lightning striking distance as assumed in the IEC lightning
to substations: Implications for the energy absorption requirements of protection standard as a function of structure height,” Electric Power Syst.
surge arresters,” in Proc. IEEE Ind. Appl. Annu. Meeting, 2022, pp. 1–16. Res., vol. 113, pp. 79–87, Aug. 2014.
[50] P. N. Mikropoulos, C. A. Stassinopoulos, and B. C. Sarigiannidou, “Posi- [74] P. N. Mikropoulos and T. E. Tsovilis, “Estimation of lightning incidence to
tive streamer propagation and breakdown in air: The influence of humid- telecommunication towers,” IEEE Trans. Electromagn. Compat., vol. 61,
ity,” IEEE Trans. Dielectrics Elect. Insul., vol. 15, no. 2, pp. 416–425, no. 6, pp. 1793–1802, Dec. 2019.
Apr. 2008. [75] P. N. Mikropoulos and T. E. Tsovilis, “Lightning attachment models
[51] N. I. Petrov and R. T. Waters, “Lightning to earthed structures: Striking and maximum shielding failure current of overhead transmission lines:
distance variation with stroke polarity, structure geometry and altitude Implications in insulation coordination of substations,” Inst. Eng. Technol.
based on a theoretical approach,” J. Electrostatics, vol. 112, Jul. 2021, Gener. Transmiss. Distrib., vol. 4, no. 12, pp. 1299–1313, Dec. 2010.
Art. no. 103599. [76] E. M. Bazelyan and Y. P. Raizer, “Long streamers,” in Spark Discharge,
[52] W. Lu et al., “Lightning attachment process involving connection of the E. M. Bazelyan and Y. P. Raizer, Eds. Boca Raton, FL, USA: CRC Press,
downward negative leader to the lateral surface of the upward connecting 1998, p. 157.
leader,” Geophys. Res. Lett., vol. 40, no. 20, pp. 5531–5535, Oct. 2013. [77] A. M. Mousa and R. J. Wehling, “A survey of industry practices regarding
[53] M. M. F. Saba et al., “Lightning attachment process to common buildings,” shielding of substations against direct lightning strokes,” IEEE Trans.
Geophys. Res. Lett., vol. 44, no. 9, pp. 4368–4375, May 2017. Power Del., vol. 8, no. 1, pp. 38–47, Jan. 1993.
[54] Q. Qi et al., “High-speed video observations of natural lightning attach- [78] M. Nicora, D. Mestriner, M. Brignone, M. Bernardi, R. Procopio, and E.
ment process with framing rates up to half a million frames per second,” Fiori, “A 10-year study on the lightning activity in Italy using data from
Geophys. Res. Lett., vol. 46, no. 21, pp. 12580–12587, Nov. 2019. the SIRF network,” Atmos. Res., vol. 256, Jul. 2021, Art. no. 105552.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IOANNIDIS AND TSOVILIS: EVALUATION OF THE DIRECT-STROKE SHIELDING PERFORMANCE OF SUBSTATIONS 13

[79] J. Herrera, C. Younes, and L. Porras, “Cloud-to-ground lightning activity in Thomas E. Tsovilis (Senior Member, IEEE) was
Colombia: A 14-year study using lightning location system data,” Atmos. born in Piraeus, Greece, in 1983. He received the
Res., vol. 203, pp. 164–174, May 2018. M.Eng. and Ph.D. degrees in electrical and computer
[80] O. Pinto Jr., “An overview of cloud-to-ground lightning research in Brazil engineering from the Aristotle University of Thes-
in the last two decades,” in Proc. 20th Int. Lightning Detection Conf. 2nd saloniki (AUTh), Thessaloniki, Greece, in 2005 and
Int. Lightning Meeting Conf., 2008, pp. 1–9. 2010, respectively.
[81] M. M. F. Saba et al., “High-speed video observations of positive lightning He held various managerial positions with the
flashes to ground,” J. Geophys. Res. Atmos., vol. 115, no. D24, pp. 1–9, R&D Department, Raycap Corporation, leading inno-
Dec. 2010. vation in surge protective devices. He was the Director
[82] M. S. Savic and A. M. Savic, “Substation lightning performance estimation of the High Current Labs of Raycap in Drama, Greece
due to strikes into connected overhead lines,” IEEE Trans. Power Del., from 2012 to 2015 and in Ljubljana, Slovenia, from
vol. 30, no. 4, pp. 1752–1760, Aug. 2015. 2016 to 2018. In 2018, he joined the AUTh, where he is currently an Associate
[83] S. Bedoui and A. Bayadi, “Probabilistic evaluation of the substation Professor. His research interests include the broad area of high-voltage engineer-
performance under incoming lightning surges,” Electric Power Syst. Res., ing with emphasis given to electrical discharges, lightning and surge protection,
vol. 162, pp. 125–133, Sep. 2018. and insulation coordination for power systems. He has authored more than 100
scientific papers in these fields, and he is the inventor of eight granted U.S.
patents on surge protective devices and testing techniques.

Alexios I. Ioannidis (Student Member, IEEE) was

born in Ptolemaida, Greece, in 1993. He received the
M.Eng. and Ph.D. degrees in electrical and computer
engineering from the Aristotle University of Thes-
saloniki (AUTh), Thessaloniki, Greece, in 2017 and
2023, respectively.
He is currently a Research Associate with the High
Voltage Laboratory, AUTh. His research focuses on
computational electromagnetics for high-voltage en-
gineering applications and lightning protection; em-
phasis is given to the development of a stochastic
lightning attachment model for assessing the lightning performance of power
systems.