Вы находитесь на странице: 1из 11

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 56, NO.

1, FEBRUARY 2014 177

A Comparison of Frequency-Dependent Soil Models:


Application to the Analysis of Grounding Systems
Damir Cavka, Member, IEEE, Nicolas Mora, Member, IEEE, and Farhad Rachidi, Fellow, IEEE

Abstract—We present a review and comparison of different mod- I. INTRODUCTION


els representing the frequency dependence of the soil electrical
HE study of the frequency dependence of soil electrical
parameters (conductivity and permittivity). These models are ex-
pressed in terms of curve-fit expressions for the soil conductivity
and relative permittivity, which are based on experimental data.
T parameters is almost a century old. At low frequencies, the
earth conductivity can be estimated by the so-called four-point
Six available models/expressions are discussed and compared mak- probe resistance measurement. At higher frequencies, measure-
ing reference to two sets of experimental data. It is shown that the
soil models by Scott, Smith–Longmire, Messier, and Visacro–Alipio
ments of the electromagnetic wave attenuation can be used to
predict overall similar results, which are in reasonable agreement estimate the earth electrical parameters.
with both sets of experimental data. Differences between the soil One of the first comprehensive analyses was performed by
models are found to be more significant at high frequencies and Smith-Rose [1] who presented a complete set of measurements
for low-resistivity soils. The causality of the considered models is on the electrical conductivity and permittivity of a variety of soil
tested using the Kramers–Kronig relationships. It is shown that
samples with different moisture contents. The data were mea-
the models/expressions of Smith–Longmire, Messier, and Portela
satisfy the Kramers–Kronig relationships and thus provide causal sured at various frequencies ranging from 1 kHz to 10 MHz.
results. The soil models are applied to the analysis of grounding During the 1960s, 1970s, and 1980s, further measurement re-
systems subject to a lightning current. A full-wave computational sults were reported in different studies [2]–[12]. In three papers
model is adopted for the analysis. The analysis is performed consid- published in the 1960s, Scott [2]–[4] reported on measurements
ering two cases: 1) a simple horizontal grounding electrode, and 2)
performed to determine the electrical conductivity, permittivity
a realistic grounding system of a wind turbine. Two current wave-
forms associated with typical first and subsequent return strokes (dielectric constant), and magnetic permeability of rocks and
are adopted for the representation of the incident lightning cur- soils in areas of interest in studies of electromagnetic pulse
rent. In agreement with recent studies, simulations show that the propagation. In [5], Amant reported on a measurement dataset
frequency dependence of the soil parameters results in a decrease of dielectric constant and resistivity of some rocks as a function
of the potential of the grounding electrode, with respect to the
of frequency and temperature. Eberle and coworkers [6]–[8]
case where the parameters are assumed to be constant. It is found
that the models/expressions by Scott, Smith–Longmire, Messier, performed measurements of various soil samples in the fre-
and Visacro–Alipio predict similar levels of decrease, which vary quency range of 100 Hz to 100 MHz and proposed a graphical
from about 2% (ρL F = 20 Ω·m and first stroke) up to 45% (ρL F = interpolation method for predicting soil electrical parameters.
10 000 Ω·m and subsequent stroke). On the other hand, the models In [9], Hipp presented measurements on two different types of
of Portela and Visacro–Portela predict significantly larger levels of
the decrease, especially for very high resistivity soils. Furthermore,
soil as a function of soil density and moisture in the frequency
in the case of a high resistivity soil (10 000 Ω·m), the Visacro–Alipio range from 30 MHz to 4 GHz using standard transmission line
expression predicts a longer risetime for the grounding potential techniques. Soil parameters at UHF and microwave frequen-
rise, compared to the predictions of Scott, Smith–Longmire, and cies were measured as a function of moisture and temperature
Messier models. by Hoekstra and Delaney [10]. Mallon et al. [11] presented
Index Terms—Antenna theory, frequency dependence of soil, measurement data for soil properties in the frequency range of
GB-IBEM, grounding systems, input impedance, soil conductivity, 100 Hz to 10 MHz, which were in reasonable agreement with the
soil permittivity, transient impedance. data published by Scott [2]–[4]. Simmons and coworkers [12]
performed laboratory measurements on igneous and metamor-
phic rocks in the frequency range from 100 Hz to 50 kHz.
In [13], Portela reported laboratory measurement results for
68 different soil samples in the frequency range from 100 Hz
Manuscript received February 19, 2013; revised May 11, 2013; accepted June to 2 MHz. More recently, Visacro and Alipio reported a large
20, 2013. Date of publication July 16, 2013; date of current version January 27, number of measurements in field conditions [14], [15]. Also, He
2014. This work was supported in part by the Swiss Government under the et al. in [16] reported experimental data for sandy soil with dif-
program “Swiss Government Excellence Scholarships for Foreign Scholars”
and the BKW Ecology Fund. ferent humidity at various temperatures. Other recent laboratory
D. Cavka is with the Swiss Federal Institute of Technology, 1015 Lausanne, measurement reports can be found in [17] and [18]. Further con-
Switzerland, and also with the Faculty of Electrical Engineering, Mechanical tributions (see, e.g., [19]–[22]) refer mainly to measurements at
Engineering, and Naval Architecture , University of Split, 21000 Split, Croatia
(e-mail: dcavka@fesb.hr). microwave frequencies and are beyond the scope of this paper.
N. Mora and F. Rachidi are with the Swiss Federal Institute of Technology, The measured data on soil parameters as a function of fre-
1015 Lausanne, Switzerland (e-mail: nicolas.mora@epfl.ch; rachidi@epfl.ch). quency were used in some studies to develop curve-fit mod-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. els and expressions. The first expressions were proposed by
Digital Object Identifier 10.1109/TEMC.2013.2271913 Scott [1]–[3] and they were based on his own laboratory and field

0018-9375 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
178 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 56, NO. 1, FEBRUARY 2014

measurements. Based on the same measurement dataset, two A. Scott (S) Expressions
additional models were developed by Longmire and cowork-
Based on laboratory and field1 measurements from 100 Hz
ers [23], [24] and Messier [25], [26], respectively. In addition to to 1 MHz of various soil samples with different water content,
the mentioned models, we can mention some general curve-fit
Scott et al. [2] developed plots of permittivity, conductivity, and
models such as Debye’s model [27], the Cole–Cole model [28],
permeability as a function of frequency and percentage of wa-
and the Lorentz soil model [29]. These models are composed of ter content. Furthermore, Scott et al. [2] provided preliminary
functional terms that are usually used to fit measurement data
curve-fit expressions for the parameters σ and ε for an arbitrary
especially at high frequencies (see, e.g., [30] and [31]). Debye’s
frequency and water content. In a follow-up work [3], refined
model was used by Fuller and Wait [32], [33] who considered curve-fitting models were derived based on additional measure-
the frequency dependence of soil electrical parameters for the
ments. The proposed expressions for the relative permittivity
propagation of transient EM signals through the earth. (dielectric constant) and conductivity as a function of frequency
A useful discussion about various soil models for nuclear are given by [3]
electromagnetic pulse applications can be found in [34].
Very few authors considered the influence of the frequency- εr (f ) = 10D (1a)
dependent soil parameters on the performance of grounding
D = 5.491 + 0.946 log10 (σ100 Hz ) − 1.097 log10 (f )
systems. Visacro and Portela [35], [36] proposed a simple em-
pirical formulation for the modeling of electrical soil parame- + 0.069 log210 (σ100 Hz )−0.114 log10 (f )log10 (σ100 Hz )
ters variation with the frequency. Later, Portela proposed a new
+ 0.067log210 (f ) (1b)
model [13] for the frequency dependence of soil parameters
K
that is mostly applied to lightning protection. Recently, Visacro σ(f ) = 10 [mS/m] (2a)
and Alipio [15], [37] proposed other expressions for the fre-
K = 0.028 + 1.098 log10 (σ100 Hz ) − 0.068 log10 (f )
quency dependence of soil parameters based on their own field
measurements of grounding harmonic impedance. The expres- + 0.036 log210 (σ100 Hz )−0.046 log10 (f )log10 (σ100 Hz )
sions proposed by Visacro and Alipio [15] were used by Akbari
+ 0.018 log210 (f ) (2b)
et al. [38] in their analysis of grounding systems, taking into
account the frequency dependence of soil electrical parameters. where f is frequency in [Hz] and σ100 Hz is the conductivity at
Some other works discussing the frequency dependence of 100 Hz in [mS/m].
soil parameters were presented in the literature, e.g., [39] and
[40]. However, their analyses are limited to one soil type and B. Smith and Longmire (SL) Expressions
they will not be discussed further in this paper.
Using Scott’s experimental data, Longmire and Longley [23]
In this paper, we present a review and comparison of the
developed another analytical representation of the soil electri-
available models and expressions for the frequency dependence
cal parameters. In [23], curve-fit expressions were presented for
of soil electrical parameters, with special attention to the ap-
the frequency range 100 Hz to 1 MHz. Later, Smith and Long-
plication of the proposed models to the analysis of grounding
mire [24] presented a curve-fit model for frequencies ranging
systems subjected to lightning.
from 1 to 1012 Hz using the data by Wilkenfeld corresponding to
This paper is organized as follows. Section II gives a review
the measured conductivity and permittivity of samples of con-
of the considered models for the frequency dependence of soil
crete and grout, and derived what they called the universal soil
electrical parameters available in the literature. The models are
model. According to [24], the frequency dependence of relative
discussed and compared, making reference to 1) the Kramers–
permittivity and conductivity is as follows:
Kronig relationships to ensure causality, and 2) available mea-
surement data. In Section III, a short description of the com- 
13
ai
putational model for the analysis of grounding systems is pro- εr (f ) = ε∞ +  2 (3)
f
vided. It is based on a full-wave electromagnetic model which i=1 1+ Fi
is numerically treated using Galerkin–Bubnov variant of the in-  2
direct boundary element method (GB-IBEM) [41]. Section IV f

13
Fi
presents the results of numerical simulations and discussion on σ(f ) = σDC + 2πε0 ai Fi  2 [S/m] (4)
f
the results obtained using different frequency-dependent soil i=1 1+ Fi
models. Finally, concluding remarks are given in Section V.
where σDC is the dc soil conductivity, ε∞ is the high frequency
limit of the dielectric constant and is set to 5, and Fi is defined
II. FREQUENCY-DEPENDENT SOIL MODELS as
We consider in this analysis six models/expressions which Fi = F (σDC ) · 10i−1 [Hz] (5)
have been specifically proposed for the representation of soil
electrical parameters (the electrical conductivity and permittiv- F (σDC ) = (125σDC )0.8312 (6)
ity). General fitting models (Debye, Cole–Cole, and Lorentz)
and, finally, the coefficients ai are given in Table I.
were excluded from the analysis since they contain additional
fitting parameters that make them impractical from an engineer-
ing point of view. 1 only for the conductivity at low frequencies.
CAVKA et al.: COMPARISON OF FREQUENCY-DEPENDENT SOIL MODELS: APPLICATION TO THE ANALYSIS OF GROUNDING SYSTEMS 179

TABLE I TABLE II
COEFFICIENTS a i FOR UNIVERSAL SOIL MAIN CHARACTERISTICS OF EXPERIMENTAL DATASETS USED FOR THE
ELABORATION OF CURVE-FIT EXPRESSIONS CONSIDERED IN THIS STUDY

Note that the original expression in [24] is adapted here in


such a way that the input parameter is the dc conductivity instead
of the moisture content.

C. Messier (M) expressions


Another representation of the electrical soil parameters, also where σ100 Hz is the conductivity measured at 100 Hz, α is an
based on Scott’s data, was proposed by Messier [25]. The model adjustable parameter model of the soil, and Δi is the value of
was modified later in [26]. According to [26], the frequency ω ε at 1 MHz. Generally, the parameters α and Δi depend on
dependence of the soil parameters can be expressed using, as in the soil model. In this study, as done in [37], we adopted the
the previous model, the high frequency limit ε∞ of the dielectric median values α = 0.706 and Δi = 11.71 mS/m.
constant and the dc conductivity σDC :
  
ε∞ σDC F. Visacro and Alipio (VA) Expressions
εr (f ) = 1+ (7)
ε0 πf ε∞ Recently, Visacro and Alipio performed a large number of
  field measurements, comprising experiments in more than 30
4πf ε∞
σ(f ) = σDC 1 + [S/m] . (8) different locations to determine the frequency dependence of
σDC the electrical parameters of soils in their natural conditions [15].
In this model, the dc conductivity σDC is supposed to be equal Based on the obtained data, they proposed the following curve-
to the conductivity measured at low frequency, and ε∞ is set to fit expressions for the frequency dependence of the soil’s relative
8. The use of this model for grounding applications was first permittivity and conductivity:
discussed by Grcev in [42]. εr (f ) = 7.6 × 103 f −0.4 + 1.3 f ≥ 10 kHz (12)

D. Visacro and Portela (VP) Expressions σ (f ) = σ100 Hz


 0.73  
Visacro and Portela proposed empirical expressions repre- −6 1 0.65
× 1+ 1.2 × 10 (f −100) . (13)
senting the variation of soil conductivity and permittivity as a σ100 Hz
function of the frequency. As reported in [36], the frequency
dependence is expressed by According to [15], a conservative approach was followed to
 −0.535 develop (12) and (13) in an attempt to avoid an overestimation
1 of the grounding systems response, as a result of the frequency
εr (f ) = 2.34 × 106 · f −0.597 (9)
σ100 Hz dependence of the soil parameters. Note that (12) is valid for
 0.072 frequencies ranging from 10 kHz to 4 MHz, while (13) is valid
f
σ(f ) = σ100 Hz . (10) for frequencies ranging from 100 Hz to 4 MHz.
100 For expressing the relative permittivity below 10 kHz, Visacro
The expressions are based on the laboratory measurements of and Alipio [15] suggested using the value given by (12) at
various soil samples in the frequency range 40 Hz to 2 MHz. 10 kHz, namely 192.2 (as shown in [15, Fig. 10]).

E. Portela (P) Expressions G. On the Causality of the Models


Based on his own measurements using samples taken from The curve-fit expressions described previously are based on
a considerable number of soils obtained in the frequency range measurements of electrical conductivity and permittivity of soils
of 100 Hz to 2 MHz, Portela [13] proposed a model for the over a wide range of frequencies. A summary of the main char-
electrical parameters of five different types of soil. All the soil acteristics of the used datasets is given in Table II.
types follow the same law: As discussed in [34], it should be emphasized that many of

π   ω α these measurements do not take into account the requirement
σ(f ) ± jω ε (f ) = σ100 Hz + Δi cot α ±j of causality. As a consequence, time-domain results obtained
2 2π · 106
(11) by taking the inverse Fourier transform of a computed spectrum
180 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 56, NO. 1, FEBRUARY 2014

TABLE III
CAUSALITY TEST OF THE CONSIDERED MODELS USING
KRAMERS–KRONIG RELATIONSHIPS

may not be causal. To ensure causality, the Kramers–Kronig


relationship must be satisfied by the real and imaginary parts of
the effective permittivity [43] (see the Appendix). Noncausality
could lead to inaccuracies such as nonzero values at t = 0. It
is important to mention, however, that causality alone cannot
guarantee the accuracy of a model. Fig. 1. Percentage difference between the values obtained using Kramers–
We have tested the causality of the models discussed previ- Kronig’s relationships and the values provided by the models as a function of
ously using the Kramers–Kronig relationships. Causality tests frequency.
were performed by numerically integrating the Kramers–Kronig
expressions (using adaptive Gauss–Kronrod quadrature) and
finding whether the conductivity or permittivity could be recon-
structed one from the other. In order to quantify the causality of
a given soil model, the following index was defined:
   f2  f2 
 2 2 
 f 1 [XK K (ω)] dω− f 1 [X (ω)] dω 
Index = 1−   f 2   · 100%
 2 f2 2 
f 1 [XK K (ω)] dω+ f 1 [X (ω)] dω
(14)
where XK K (ω) is the frequency spectrum of the parameter
(conductivity or permittivity) obtained using Kramers–Kronig
expression, X (ω) is the frequency spectrum of the parameter
as given by the model, and f1 and f2 are the boundaries of
the frequency range which have been set, respectively, to the
validity limits of the proposed models/expressions (second col-
umn in Table II). Note that the defined index should be 100%
(within the numerical errors) for a causal function. The com-
puted values of the index for both permittivity and conductivity
are reported in Table III. The results in Table III were obtained Fig. 2. Comparison of different soil models (ρL F = 93.5 Ω · m).
for a low-frequency resistivity of 1000 Ω·m. The number of fre-
quencies was set to 500, spaced evenly in logarithmic scale in the
frequency range for which models/expressions were proposed.
Integrals in (14) were evaluated numerically using trapezoidal Messier, and Portela models, confirming the causality of these
rule since it was found sufficiently accurate due to nonoscilla- expressions.
tory nature of the integrants. It can be seen that, for the models
of Smith–Longmire, Messier, and Portela, the index is higher H. Comparison of the Models
than 99.9%, and therefore, these models can be considered as We present in this section a comparison between the six con-
causal. sidered soil models. Figs. 2–5 present the results for values of
Fig. 1 shows the percentage difference between the values low-frequency (100 Hz) resistivity (ρLF ) of 93.46, 343, 1806,
obtained using Kramers–Kronig relationships and the values and 4740 Ω·m, respectively. The adopted values correspond
provided by the models in the frequency range of 100 Hz to to the experimental data of Bigelow and Eberle [7] and He
1 MHz. In this figure, tags in the legends indicate the following: et al. [16], which are also plotted in the same figures. Note that,
S—Scott soil model, SL—Smith and Longmire soil model, M— in the case of the Smith–Longmire model, the value of the dc
Messier soil model, VA—Visacro and Alipio soil model, P— conductivity was obtained from the 100 Hz value by numerically
Portela soil model, VP—Portela and Visacro soil model.2 Again, solving (4).
the results show negligible difference for the Smith–Longmire, These two datasets have been used as reference because they
have been obtained independently from any of the discussed soil
2 The same notation stands also for all the figures in this paper. models in this paper.
CAVKA et al.: COMPARISON OF FREQUENCY-DEPENDENT SOIL MODELS: APPLICATION TO THE ANALYSIS OF GROUNDING SYSTEMS 181

Overall, it can be seen that:


1) The soil models by Scott (S), Smith–Longmire (SL),
Messier (M), and Visacro–Alipio (VA) predict overall sim-
ilar results, which are in reasonable agreement with the
independent experimental data of Bigelow–Eberle and He
et al.
2) Differences between the soil models are more significant
at high frequencies and for low-resistivity soils.
3) In general, predictions of the models by Portela and
Visacro–Portela differ significantly from those of other
considered models.

III. COMPUTATIONAL MODEL FOR THE ANALYSIS


OF GROUNDING SYSTEMS

The computational model for the analysis of a complex


grounding system embedded in a frequency-dependent soil is
Fig. 3. Comparison of different soil models (ρL F = 343 Ω · m). based on the full electromagnetic approach originally developed
by Grcev and Dawalibi [44] and the related set of Pocklington-
type equations for arbitrary oriented wires
NW 
  
In (s ) · ŝm ·ŝn · k 2 + ∇∇ Gn (sm , sn ) ds = 0
n =1 C n

m = 1, 2, . . . , NW (15)
where I(s ) is the induced current along the nth grounding sys-
tem wire and kernel Gn (x,x ) denotes the lossy medium Green’s
function. k is the wave number for the lossy medium and εeff is
the frequency-dependent complex permittivity of lossy ground
defined as
σ (f )
εeff (f ) = ε0 εr (f ) − j (16)
ω
where εr and σ are the corresponding frequency-dependent per-
mittivity and conductivity of the ground, respectively.
In this model, the excitation in terms of a current source
Fig. 4. Comparison of different soil models (ρL F = 1806 Ω · m). is incorporated into the formulation through the boundary
condition.
The governing equation (15) is numerically treated via the
GB-IBEM [41] in order to obtain the current distribution along
the grounding wire. After that, the input impedance of the
grounding system is calculated using the originally developed
IEMF-BEM method [45].
To obtain the grounding potential rise (GPR), the harmonic
grounding impedance is multiplied by the corresponding spec-
trum associated with the excitation current source to obtain
the frequency response of the grounding electrode. Finally, the
transient response is calculated by means of the hybrid inverse
Fourier transform [46].
More details on the computational methodology can be found
in other papers of the first author (see, e.g., [41], [45], and [46]).

IV. NUMERICAL EXAMPLES


Extensive numerical experiments have been undertaken in
order to compare the various soil models. The comparison was
Fig. 5. Comparison of different soil models (ρL F = 4740 Ω · m).
made first for a simple horizontal grounding electrode and then
for complex wind turbine grounding system. Different values
182 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 56, NO. 1, FEBRUARY 2014

Fig. 6. Input impedance for various soil models (L = 10 m; ρL F = 100 Ω·m). Fig. 8. Input impedance for various soil models (L = 10 m; ρL F =
10 000 Ω·m).

electrode for various soil models and for different values of the
low-frequency ground resistivity, namely 100, 1000, and 10 000
Ω·m.
The results presented in Figs. 6–8 show that the differences
between various soil models increase for higher values of the
low-frequency resistivity. Also, differences start to appear ear-
lier in the frequency spectrum with the increase of resistivity.
The variations appear to be more pronounced for the Portela and
Viasacro–Portela models/expressions.
The differences between the predictions of the frequency-
dependent models and those obtained assuming constant soil
parameters (labeled C in all the figures) increase with frequency
and ground resistivity. Generally, predictions obtained using
the models by Scott, Smith–Longmire, and Messier are very
similar. This is expected since these three models are based
Fig. 7. Input impedance for various soil models (L = 10 m; ρL F = on the same measurement dataset. Visacro–Alipio’s predictions
1000 Ω·m). are also, in general, similar to those of Scott, Smith–Longmire,
and Messier, except for the case of a low-resistivity soil (see
for the low-frequency ground resistivity were considered in the Fig. 8) where differences appear for frequencies higher than a
analysis (from 20 up to 10 000 Ω·m). few kilohertz. More significant differences are observed for the
Comparisons were made in both frequency and time domains. two other models (Portela and Visacro–Portela), especially for
Two waveforms were adopted for the representation of the in- very high values of ground resistivity (see Fig. 8).
cident lightning current, associated with typical first and subse- Figs. 9 and 10 show the corresponding transient response
quent return strokes [47]. The first return stroke current is char- in time domain (transient GPR), for first and subsequent re-
acterized by a peak value of 30 kA and a maximum steepness turn strokes, respectively. It can be seen, as already discussed
of 12 kA/μs, whereas the subsequent return stroke current has a in [14], [15], and [36]–[38], that the frequency dependence of
peak value of 12 kA and a maximum steepness of 40 kA/μs. The the soil parameters results in a decrease of the potential of
two waveforms are represented using Heidler’s function [48], the grounding electrode with respect to the case where the
the parameters of which are given in [49]. soil parameters are assumed to be constant. This decrease is
In the case of constant soil parameters, the relative permittiv- more significant for soils with very higher resistivity. While the
ity in all the cases was assumed to be εr = 5. models/expressions by Scott, Smith–Longmire, Messier, and
Visacro–Alipio predict similar levels of decrease, which vary
from about 2% (ρLF = 20 Ω·m and first stroke) up to 45%
A. Simple Horizontal Electrode
(ρLF = 10 000 Ω·m and subsequent stroke), the models of
Consider a simple horizontal electrode with a length L = Portela and Visacro–Portela predict significantly larger levels
10 m, a radius a = 5 mm, and buried at a depth d = 1 m. of the decrease, especially for very high resistivity soils. Fur-
Figs. 6–8 present the frequency spectrum (magnitude and thermore, in the case of a high resistivity soil (10 000 Ω·m),
phase) of the harmonic grounding impedance of the horizontal the Visacro–Alipio expression predicts a longer risetime for the
CAVKA et al.: COMPARISON OF FREQUENCY-DEPENDENT SOIL MODELS: APPLICATION TO THE ANALYSIS OF GROUNDING SYSTEMS 183

(a) (a)

(b) (b)

(c) (c)

Fig. 9. Transient GPR of a horizontal buried electrode for various soil mod- Fig. 10. Transient GPR of a horizontal buried electrode for various soil models.
els. First return stroke. (a) (ρL F = 100 Ω·m), (b) (ρL F = 1000 Ω·m), and Subsequent return stroke. (a) (ρL F = 100 Ω·m), (b) (ρL F = 1000 Ω·m), and
(c) (ρL F = 10 000 Ω·m). (c) (ρL F = 10 000 Ω·m).

GPR, compared to the predictions of Scott, Smith–Longmire,


and Messier models.
Figs. 11 and 12 present the relative decrease (with respect
to the case of a soil characterized by constant conductivity and
permittivity εr = 5) of the peak potential as a function of the
low-frequency resistivity ρLF , for first and subsequent return
strokes. It can be seen that all the considered soil models, ex-
cept for Portela and Visacro–Portela, predict very similar values.
Furthermore, it can be seen that according to these models (S,
SL, M, and VA), for a soil resistivity of 200 Ω·m or lower, the
frequency dependence of the soil parameters does not signifi-
cantly affect the potential (less than 5% for first strokes and less
than about 15% for subsequent strokes). For higher resistivity Fig 11. Relative decrease (with respect to the case of a soil characterized by
soils, however, a more significant decrease is predicted by these constant conductivity and permittivity εr = 5) in peak potential for various soil
models—First return stroke.
four models.
184 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 56, NO. 1, FEBRUARY 2014

Fig 12. Relative decrease (with respect to the case of a soil characterized by Fig. 15. Transient response for various soil models (WT, first return stroke
constant conductivity and permittivity εr = 5) in peak potential for various soil ρL F = 1200 Ω·m).
models—Subsequent return stroke.

Fig. 13. Initial values of the GPR (L = 10 m; ρL F = 1000 Ω·m, subsequent


stroke). Fig. 16. Transient response for various soil models (WT subsequent return
stroke, ρL F = 1200 Ω·m).

a low-frequency resistivity ρLF = 1200 Ω·m which is a typical


value of resistivity measured in the coastal part of Croatia.
Figs. 15 and 16 present the transient response of the ground-
ing system when subjected to a lightning current, using the
considered soil models, for first and subsequent return strokes,
respectively. Similar to the case of a horizontal electrode, the dif-
ferent soil models, except for the models of Portela and Visacro–
Portela, give very similar results for the ground potential. In the
case of a typical first return stroke, the frequency dependence of
Fig. 14. Geometry of the WT grounding system.
the soil parameters results in a decrease of the peak of the poten-
tial rise by about 10% for the models of Scott, Scott-Longmire,
It is worth noting that models/expressions that do not satisfy Messier, and Visacro–Alipio, and about 26% for the Visacro–
Kramers–Kronig relationships might suffer from a nonzero ini- Portela and Portela models, with respect to the case where the
tial value for the GPR. This is illustrated in Fig. 13, where initial soil parameters are assumed to be constant.
values of the GPR are presented for various models in the case On the other hand, for a typical subsequent stroke, the pre-
of a subsequent stroke and a soil resistivity of 1000 Ω·m. It can dicted decrease is more significant: about 20% for the models by
be seen that the expressions that do not satisfy causality result Scott, Scott-Longmire, Messier, and Visacro–Alipio, and about
in a nonzero initial value. This initial value, however, remains 40% for the models of Visacro–Portela and Portela. Note that
relatively small, and therefore, these models can also be applied the amounts of decrease predicted by the models are consistent
for engineering applications. with the case of a simple horizontal grounding wire.

B. Complex Wind Turbine Grounding Systems V. CONCLUSION


The analysis was also made considering a complex wind tur- We presented a review and comparison of different mod-
bine grounding system. The geometry of the grounding system els representing the frequency dependence of the soil electrical
is shown in Fig. 14. The grounding system is placed in a soil with parameters (conductivity and permittivity). These models are
CAVKA et al.: COMPARISON OF FREQUENCY-DEPENDENT SOIL MODELS: APPLICATION TO THE ANALYSIS OF GROUNDING SYSTEMS 185

 ∞
expressed in terms of curve-fit expressions for the soil conduc- 2 ω  Im [εeff (ω  )] 
Re [εeff (ω)] = ε0 + dω (17a)
tivity and relative permittivity which are based on experimental π 0 ω 2 − ω 2
data.  ∞
2ω Re [εeff (ω  )] 
Six available models were discussed and compared, making Im [εeff (ω)] = − dω (17b)
reference to two sets of experimental data obtained indepen- π 0 ω 2 − ω 2
dently by Bigelow and Eberle and by He et al. It was shown where the integrals must be evaluated as a Cauchy principal
that the soil models by Scott, Smith–Longmire, Messier, and values. Equations (19a) and (19b) can be rearranged and written
Visacro–Alipio predict overall similar results, which are in rea- as [34]
sonable agreement with both sets of experimental data. Differ- 
ences between the soil models were found to be more significant 2 ∞ ω  Im [εeff (ω  )]−ω Im [εeff (ω)] 
Re [εeff (ω)] = ε0 + dω
at high frequencies and for low-resistivity soils. π 0 ω 2 − ω 2
The causality of the considered models was tested using the (18a)
Kramers–Kronig relationships. It was shown that the models  ∞  
of Smith–Longmire, Messier, and Portela satisfy the Kramers– 2ω Re [εeff (ω )] − Re [εeff (ω )] 
Im [εeff (ω)] = − dω .
Kronig relationships and thus provide causal results. π 0 ω 2 − ω 2
In the second part of this paper, the soil models were applied to (18b)
the analysis of grounding systems subject to a lightning current.
A full-wave computational model was adopted for the analysis. Equations (18a) and (18b) are more convenient to be used com-
The analysis was performed considering two cases: 1) a simple pared to (17a) and (17b), since the denominator vanishes at the
horizontal grounding electrode, and 2) a realistic and complex singular point ω  = ω. Also, the integrals in (18) have the same
grounding system of a wind turbine. Two current waveforms Cauchy principal value as do their counterparts in (17).
associated with typical first and subsequent return strokes were As discussed in [34], any soil model based on the measure-
adopted for the representation of the incident lightning current. ment data should take into account the causality requirement.
In agreement with recent studies (Visacro and coworkers and
Akbari et al.), simulations showed that the frequency depen- ACKNOWLEDGMENT
dence of the soil parameters results in a decrease of the potential
The authors express their gratitude to Prof. M. Rubinstein,
of the grounding electrode, with respect to the case where the
Prof. S. Visacro, and three anonymous reviewers for their pre-
parameters are assumed to be constant. It was found that the
cious comments and suggestions on the manuscript.
models/expressions by Scott, Smith–Longmire, Messier, and
Visacro–Alipio predict similar levels of decrease, which vary
from about 2% (ρLF = 20 Ω·m and first stroke) up to 45% REFERENCES
(ρLF = 10000 Ω·m and subsequent stroke). On the other hand, [1] R. L. Smith-Rose, “Electrical measurements on soil with alternating cur-
the models of Portela and Visacro–Portela predict significantly rents,” Proc. IEE, vol. AP-75, pp. 221–237, Aug. 1934.
larger levels of the decrease, especially for very high resis- [2] J. H. Scott, R. D. Carroll, and D. R. Cunningham, “Dielectric constant
and electrical conductivity of moist rock from laboratory measurements,”
tivity soils. Furthermore, in the case of a high resistivity soil Sensor and Simulation Note 116, Kirtland AFB, NM, Aug. 1964.
(10 000 Ω·m), the Visacro–Alipio expression predicts a longer [3] J. H. Scott. (1966, May). “Electrical and magnetic properties of rock
risetime for the GPR, compared to the predictions of Scott, and soil,” Theoretical Notes, Note 18, U.S. Geological Survey [Online].
Available: https://www.ece.unm.edu/summa/notes/Theoretical.html
Smith–Longmire, and Messier models. [4] J. H. Scott, D. Carroll, and D. R. Cunningham, “Dielectric constant and
The results in this paper reveal differences, sometimes signifi- electrical conductivity measurements of moist rock: A new laboratory
cant, between the available models/expressions which are based method,” J. Geophys. Res., vol. 72, no. 20, pp. 5101–5115, 1967.
[5] M. S. Amant, “Frequency and temperature dependence of dielectric prop-
on different sets of experimental data. The observed inconsis- erties of some common rocks,” MSc thesis, Dept. Geol. Geophys., Mas-
tency can be partially attributed to the fact that the measurement sachusetts Inst. Technol., Cambridge, Boston, MA, USA, Aug. 1968.
of soil parameters is a challenging task. This calls for further [6] M. M. Judy and W. R. Eberle. (1969, Jun.) “A laboratory method for
the measurement of the dielectric constant of rock and soil samples
research in order to consolidate our knowledge on the soil elec- in the frequency range 102–108 Hz,” Sensor and Simulation Notes,
trical characterization. Note 88, U.S. Geological Survey [Online]. Available: https://www.
ece.unm.edu/summa/notes/Sensor.html
[7] R. C. Bigelow and W. R. Eberle, “Empirical predictive curves for resistivity
and dielectric constant of earth materials: 100 Hz to 100 MHz,” Open-File
Report 83-911, U.S. Geological Survey, 1972.
[8] W. R. Eberle “The effects of water content and water resistivity on the
APPENDIX dispersion of resistivity and dielectric constant in quartz sand in the fre-
quency range 102 to 108 Hz,” Open-File Report 83-914, US Geological
KRAMERS–KRONIG RELATIONSHIPS Survey, 1983.
Since any solution resulting from the Maxwell’s equations [9] J. E. Hipp, “Soil electromagnetic parameters as functions of frequency,
soil density and soil moisture,” Proc. IEEE, vol. 62, no. 1, pp. 98–103,
must be causal, the electrical parameters—permittivity and Jan. 1974.
conductivity—cannot be chosen in an arbitrary manner for [10] P. Hoekstra and A. Delaney, “Dielectric properties of soils at UHF and
any particular medium. The condition of causality leads to the microwave frequencies,” J. Geophys. Res., vol. 79, no. 11, pp. 1699–1708,
Apr. 10, 1974.
Kramers–Kronig relationships between the real and imaginary [11] C. Mallon, R. Denson, R. E. Leadon, and T. M. Flanagan.
parts of the effective dielectric constant (16) [34], [43]: (1981, Jan.). “Low-field electrical characteristics of soil,” Theoretical
186 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 56, NO. 1, FEBRUARY 2014

Notes, Note 315 [Online]. Available: https://www.ece.unm.edu/summa/ [35] F. S. Visacro and C. M. Portela, “Soil permittivity and conductivity behav-
notes/Theoretical.html ior on frequency range of transient phenomena in electric power systems,”
[12] G. Simmons, L. Caruso, and F. Miller, “Complex dielectric properties presented at the Symp. High Voltage Eng., Braunschweig, Germany, 1987.
of several igneous and metamorphic rocks,” Massachusetts Inst. Technol., [36] A. G. Pedrosa, R. S. Alı́pio, M. A. O. Schroeder, and M. M. Afonso, “Sim-
Cambridge, MA, USA, Tech. Rep., Jul. 1, 1975– Sep. 30, 1980. Available: ulation of behavior electrical grounding—Comparative analysis for per-
http://www.dtic.mil/docs/citations/ADA094301 formance in frequency-domain,” in Proc. Int. Workshop Appl. Model.
[13] C. M. Portela, “Measurement and modeling of soil electromagnetic be- Simul., Rio de Janeiro, May 5–7, 2010, vol. 1, Brazil, pp. 89–92.
havior,” in Proc. IEEE Int. Symp. Electromagn. Compat., 1999, vol. 2, [37] R. Alipio and S. Visacro, “Frequency dependence of soil parameters:
pp. 1004–1009. Effect on the lightning response of grounding electrodes,” IEEE Trans.
[14] S. Visacro, R. Alipio, M. H. Murta Vale, and C. Pereira, “The response Electromagn. Compat., vol. 55, no. 1, pp. 132–139, Feb. 2013.
of grounding electrodes to lightning currents: The effect of frequency- [38] M. Akbari, K. Sheshyekani, and M. R. Alemi, “The effect of frequency
dependent soil resistivity and permittivity,” IEEE Trans. Electromagn. dependence of soil electrical parameters on the lightning performance of
Compat., vol. 53, no. 2, pp. 401–406, May 2011. grounding systems,” IEEE Trans. Electromagn. Compat., vol. 55, no. 4,
[15] S. Visacro and R. Alipio, “Frequency dependence of soil parameters: doi: 10.1109/TEMC.2012.2222416.
Experimental results, predicting formula and influence on the lightning [39] J. H. Choi, B. H. Lee, and S. K. Paek, “Frequency-dependent grounding
response of grounding electrodes,” IEEE Trans. Power Del., vol. 27, no. 2, impedance of the counterpoise based on the dispersed currents,” J. Electr.
pp. 927–935, Apr. 2012. Eng. Technol., vol. 7, no. 4, pp. 589–595, 2012.
[16] J. He, R. Zeng, and B. Zhang, Methodology and Technology for Power [40] J. Montaña, J. Candelo, and O. Duarte, “Sand’s electrical parameters vary
System Grounding. Singapore: Wiley, 2013, p. 35. with frequency,” Ingenierı́a e Investigación, vol. 32, no. 2, pp. 34–39,
[17] G. R. Olhoeft and D. E. Capron, “Laboratory measurement of the radiofre- Aug. 2012.
quency electrical and magnetic properties of soils near Yuma, Arizona,” [41] D. Cavka, B. Harrat, D. Poljak, B. Nekhoul, K. Kerroum, and K. El Kham-
U.S. Dept. Interior, Washington, DC, USA, U.S. Geological Survey, Open lichi Drissi, “Wire antenna versus modified transmission line approach to
File Rep. 93-701, 1993. the transient analysis of grounding grid,” Eng. Anal. Boundary Elements,
[18] R. J. Sengwa and A. Soni, “Dielectric properties of some minerals of vol. 35, no. 10, pp. 1101–1108, 2011.
western Rajasthan,” Indian J. Radio Space Phys., vol. 37, pp. 57–63, Feb. [42] L. Grcev, “High frequency grounding,” in Lightning Protection,
2008. V. Cooray, Ed. London,, U.K.: IET, 2010, Ch. 10, pp. 503–529.
[19] N. R. Peplinski, F. T. Ulaby, and M. C. Dobson, “Dielectric properties [43] S. Ramo, J. R. Whinnery, and T. Van Duzer, Fields and Waves in Com-
of soils in the 0.3–1.3-GHz range,” IEEE Trans. Geosci. Remote Sens., munication Electronics. New York, NY, USA: Wiley, 1984.
vol. 33, no. 3, pp. 803–807, May 1995. [44] L. Grcev and F. Dawalibi, “An electromagnetic model for transients in
[20] N. R. Peplinski, F. T. Ulaby,, and M. C. Dobson, “Corrections to dielectric grounding systems,” IEEE Trans. Power Del., vol. 5, no. 4, pp. 1773–
properties of soils in the 0.3–1.3-GHz range,” IEEE Trans. Geosci. Remote 1781, Oct. 1990.
Sens., vol. 33, no. 6, p. 1340, Nov. 1995. [45] D. Cavka and D. Poljak, “On the calculation of input impedance and tran-
[21] V. L. Mironov, M. Craig Dobson, V. H. Kaupp, S. A. Komarov, and sient impedance for grounding electrodes using antenna theory,” COM-
V. N. Kleshchenko, “Generalized refractive mixing dielectric model for PEL: Int. J. Comput. Math. Elect. Electron. Eng., vol. 32, no. 6.
moist soils,” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 4, pp. 773– [46] D. Cavka, D. Poljak, V. Doric, and S. Antonijevic, “Some computational
785, Apr. 1995. aspects of using current and voltage sources in electromagnetic models of
[22] V. L. Mironov, L. G. Kosolapova, and S. V. Fomin., “Soil dielectric model lightning return strokes,” presented at the Int. Conf. Lightning Protection,
accounting for contribution of bound water spectra through clay content,” Vienna, Austria, Sep. 2012.
PIERS Online, vol. 4, no. 1, pp. 31–35, 2008. [47] K. Berger, R. B. Anderson, and H. Kroninger, “Parameters of lightning
[23] C. L. Longmire and H. J. Longley. (1973). “Time domain treatment of flashes,” Electra, vol. 41, pp. 23–37, 1975.
media with frequency dependent parameters,” Theoretical Notes 113, [48] F. Heidler, “Analytische Blitzstromfunktion zur LEMP- Berechnung,” in
Defense Nuclear Agency, Santa Barbara, CA, USA [Online]. Available: Proc. 18th Int. Conf. Lightning Protection, Munich, Germany, Sep. 16–20,
https://www.ece.unm.edu/summa/notes/Theoretical.html 1985, pp. 63–66, paper 1.9.
[24] K. S. Smith and C. L. Longmire, “A universal impedance for soils,” [49] F. Rachidi, W. Janischewskyj, A. M. Hussein, C. A. Nucci, S. Guerrieri,
Defense Nuclear Agency, Alexandria, VA, USA, Topical Report for Period B. Kordi, and J. S. Chang, “Current and electromagnetic field associated
Jul. 1 1975–Sep. 30 1975, 1975. with lightning return strokes to tall towers,” IEEE Trans. Electromagn.
[25] M. A. Messier, “The propagation of an electromagnetic impulse through Compat., vol. 43, no. 3, pp. 356–367, Aug. 2001.
soil: Influence of frequency dependent parameters,” Mission Res. Corp.,
Santa Barbara, CA, USA, Tech. Rep. MRC-N-415, 1980.
[26] M. Messier, “Another soil conductivity model,” internal rep., JAYCOR,
Santa Barbara, CA, 1985.
[27] P. Debye, Polar Moleculars. New York, NY, USA: Dover, 1929.
[28] K. S. Cole and R. H. Cole, “Dispersion and absorption in dielectrics I.
Alternating current characteristics,” J. Chem. Phys., vol. 9, pp. 341–351,
1941.
[29] K. S. Kunz and R. J. Luebbers, The Finite Difference Time Domain Method
for Electromagnetics. Boca Raton, FL, USA: CRC Press, 1993, pp. 139–
144.
[30] E. G. Farr and C. A. Frost. (1996, Oct.).“Time domain measure-
ment of the dielectric properties of water in a coaxial test fix- Damir Cavka (M’12) was born in Split, Croatia, in
ture,” Measurement Note 49, Kirtland AFB, NM [Online]. Available: 1982. He received the B.Sc. degree from the Univer-
https://www.ece.unm.edu/summa/notes/Measurement.html sity of Split, Split, Croatia, the M.Phil. degree from
[31] E. G. Farr and C. A. Frost, “Impulse propagation measurements of the the University of Wales, Cardiff, U.K., and the Ph.D.
dielectric properties of water, dry sand, moist sand, and concrete,” Mea- degree from the University of Split, in 2005, 2007,
surement Note 52, Kirtland AFB, NM, Nov. 1997. and 2011, respectively.
[32] J. A. Fuller and J. R. Wait, “Electromagnetic pulse transmission in ho- During 2012 and 2013, he was a Visiting Post-
mogeneous dispersive rock,” IEEE Trans. Antennas Propag., vol. AP-20, doctoral Fellow at the EMC Laboratory of the Swiss
no. 4, pp. 530–533, Jul. 1972. Federal Institute of Technology, Lausanne, Switzer-
[33] J. A. Fuller and J. R. Wait, “A pulsed dipole in the earth,” in Tran- land. He is currently a Postdoctoral Researcher Assis-
sient. Electromagnetic Fields, L. B. Felsen, Ed. New York, NY, USA: tant with the Department of Electronics, University of
Springer, 1976, ch. 5, pp. 238–270. Split. He has published 17 journal and conference papers in the area of computa-
[34] F. M. Tesche. (2002, Jul.). “On the modeling and representation tional electromagnetic compatibility. His research interests include frequency-
of a lossy earth for transient electromagnetic field calculations,” domain computational methods in electromagnetic compatibility, particularly
Theoretical Notes 367 [Online]. Available: https://www.ece.unm.edu/ in the numerical modeling of wire structures and human exposure to electro-
summa/notes/Theoretical.html magnetic fields.
CAVKA et al.: COMPARISON OF FREQUENCY-DEPENDENT SOIL MODELS: APPLICATION TO THE ANALYSIS OF GROUNDING SYSTEMS 187

Nicolas Mora (M’07) was born in Bogotá, Colom- Farhad Rachidi (M’93–SM’02–F’10) received the
bia, in 1985. He received the B.Sc.. degree in elec- M.S. degree in electrical engineering and the Ph.D.
tronics engineering and the M.Sc. degree in electrical degree from the Swiss Federal Institute of Technol-
engineering with a major in high voltage engineering ogy (EPFL), Lausanne, Switzerland, in 1986 and
from the National University of Colombia, Bogotá, in 1991, respectively.
2007 and 2009, respectively. Since early 2011, he has He worked at the Power Systems Laboratory of
been working toward the Ph.D. degree at the Swiss the EPFL until 1996. In 1997, he joined the Lightning
Federal Institute of Technology (EPFL), Lausanne, Research Laboratory, University of Toronto, Toronto,
Switzerland. ON, Canada. From April 1998 until September 1999,
He joined the Electromagnetic Compatibility Re- he was with Montena EMC, Switzerland. He is cur-
search Group of the National University of Colombia rently a Titular Professor and the Head of the EMC
under the tutelage of P. Francisco Roman in early 2007 . In September 2009, he Laboratory at the EPFL. He served as the Vice-Chair of the European COST
joined the EMC Lab at EPFL under the direction of Prof. F. Rachidi. Action on the Physics of Lightning Flash and its Effects (2005–2009) and the
Mr. Mora was the co-recipient of 2011 Frank Gunther Award of the Radio Chairman of the 2008 European Electromagnetics International Symposium.
Club of America and a recipient of Young Scientist Award from the International He is the author or coauthor of more than 350 scientific papers published in
Union of Radio Science. reviewed journals and presented at international conferences.
Dr. Rachidi is the Editor-in-Chief of the IEEE TRANSACTIONS ON ELEC-
TROMAGNETIC COMPATIBILITY, the President of the International Conference
on Lightning Protection, and the President of the Swiss National Committee
of the International Union of Radio Science. He received the IEEE Technical
Achievement Award and the CIGRE Technical Committee Award in 2005. In
2006, he was awarded the Blondel Medal from the French Association of Elec-
trical Engineering, Electronics, Information Technology and Communication.

Вам также может понравиться