Professional Documents
Culture Documents
3, JULY 2010
Abstract—This paper presents an optimization methodology by modeling may possibly be developed, jointly with the measured
using a genetic algorithm (GA) to obtain the parameters of a soil resistances, considering the soil as a set of horizontal layers,
that can be represented in a multilayer structure. The method uses overlaid in parallel planes, and with distinct resistivities to
a curve of experimental apparent resistivity obtained from mea-
surements made in the soil. This experimental curve is compared each other [1] (i.e., a model of the soil stratified in horizontal
with another curve of apparent resistivity of the soil, theoretical, multilayers).
produced by the GA. The theoretical curve is based on Sunde’s Al- The process to obtain the thickness of each soil’s layer, with
gorithm and is exactly the inverse process used for the horizontal its respective resistivities, is called, in this paper, a soil stratifica-
stratification of the soil in multilayers. With both curves in hand, tion method. And for the accomplishment of a soil stratification
the error produced in the process of soil stratification can be esti-
mated. From the estimated errors, the parameters are optimized. process, it is necessary to find the values of the soil apparent re-
The main difference of this method, comparing the already devel- sistivity in function of the distance between electrodes .
oped methods is just that, besides optimizing the resistivities and However, the well-known processes to produce the horizontal
the thicknesses of each soil layer, the proposed method also op- stratification of the soil, in its majority, stratify the soil only in
timizes the layers quantity, seeking the best soil stratification. The
two or three layers [3], [4]. In other words, they fix the number
obtained results of this paper are compared with other soil strati-
fication methods. of soil layers, not considering that the same can present a
different quantity of layers. A process that considers the real
Index Terms—Apparent resistivity, genetic algorithm (GA), soil
stratification modeling. number of layers that the soil presents would be able to achieve
more accurate stratification results. However, a process that con-
siders the actual number of layers presented by the soil, would
I. INTRODUCTION improve the stratification results accuracy. Nevertheless, the tra-
ditional methods of soil stratification into several layers, both
Then, isolating the in (5), the expression for the soil elec-
trical resistivity calculation is obtained
(6)
Pirson’s Method [24] for stratifying soils into multihorizontal the soil [15]. Wenner’s Method for four electrodes provides the
layers is used. The result is used as the initial population for the apparent resistivity as a function of the distance between elec-
GA. trodes . Therefore, it is possible to relate
in (4) with the function . Thus, with some algebraic
IV. PROPOSED METHODOLOGY manipulation
With the values of , , and in hand, the inverse process
(14)
can be accomplished by using Sunde’s algorithm, assuming the
earth isotropic and, thus, the propagation constant remaining un-
is achieved, where , and the other terms in-
changed at any direction [17]. The theoretical apparent resis-
volved in calculating are presented
tivity curve is obtained through Sunde’s algorithm and the
GA is used to compare the theoretical and the experimental (15)
apparent resistivity curve. First, it is necessary to calculate
the stratification factors of Sunde’s algorithm through
(16)
(7)
(17)
Then, an iterative calculation process is applied, relating two
new functions, such as Having obtained the function in (14), it can be written
considering pairs of values for , i.e., pairs of values for and
(8)
(18)
where , is an independent auxiliary variable
which assumes values of zero at , and . Using (19)
some algebraic manipulation, it is possible to write
Dividing and subtracting (18) and (19), results in (20) and
(9) (21), respectively
(10) (20)
(21)
Making , one gets . It can be seen
from this equation that a variable changing from to can where and are defined
be made. As and , knowing the by
possible value of and estimating the values of and with
the GA, it is possible to obtain the function, through (22)
Sunde’s algorithm. The function can be decomposed in
an approximately 40 polynomial, defined by (11) [17] In (20) and (21), there is
(11) (23)
TABLE II
THEORETICAL RESISTIVITY VALUES— CASE STUDY 1
(26)
(27)
V. RESULTS
In this section, the results obtained from three different soil
horizontal stratification cases are presented. The first case study
compares results obtained through the GA and the other two
methods used to stratify the soil in multilayers.
One method used at the comparison employs an algorithm
of exact optimization, known as Quasi-Newton. This method
also utilizes the experimental curve of apparent resistivity as ref- Fig. 2. Experimental and theoretical curves of apparent resistivity of the
erence and obtains the soil horizontal stratification parameters soil—Study of Case 1.
through an optimized theoretical curve of apparent resistivity.
However, once it employs numerical differentiation and matrix
inversion during the optimization process, some restrictions are A. Case Study 1
inherent due to inconsistency and numeric indetermination.
The other method used in the results comparison is known For Case Study 1, measures in the field for obtaining ex-
as the complex images method, described by Chow et al. [25]. perimental apparent resistivities of the soil values were made.
This method is widely employed in the accomplished of works These values with the distances among electrodes are presented
in this field [5]–[7]. in Table I. The values of theoretical apparent resistivity of the
In the second case study, the parameters results of a soil’s soil computationally obtained from the proposed method can be
stratification process were taken from the work of Gonos analyzed in Table II. Both experimental and theoretical values
and Stathopulos [10]. The described values in this paper are are plotted in Fig. 2.
compared with the values obtained by the proposed GA. In the In Table III, the deviations values between experimental and
third case, a comparison is made between a soil’s stratification theoretical curves of apparent resistivity of the soil are listed.
process which imposes a predetermined number of layers These deviations were calculated by (27). In Table IV, the re-
and the proposed method that optimizes the number of layers sulting parameters from each method of soil horizontal stratifi-
of the stratified soil. cation used in Case Study 1 are presented.
1254 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 3, JULY 2010
TABLE IV
STRATIFICATION PARAMETERS RESULTS—CASE STUDY 1
TABLE V
EXPERIMENTAL RESISTIVITY—CASE STUDY 2
TABLE VII
DEVIATIONS BETWEEN EXPERIMENTAL AND
THEORETICAL CURVES—CASE STUDY 2
B. Case Study 2
In this case study, a comparison is made between the results
obtained by the soil stratification proposed method, and the re-
sults presented in Gonos and Stathopulos’s work [10]. For that,
the measured experimental values of apparent resistivity of the
soil and the parameters , , and computationally obtained
by the method used in the cited work are utilized for compar- with the theoretical curves of apparent resistivity of the soil ob-
ison with the same theoretical results computationally obtained tained by each method are plotted.
in this paper. The experimental values of apparent resistivity are Table VII presents the deviations calculated by (27), between
listed in Table V which is used as a reference for the theoretical the experimental curve of apparent resistivity of the soil and
curve of apparent resistivity, see Table VI, and the parameters the obtained theoretical one. Table VIII presents the comparison
, , and , which are obtained through the proposed GA. In between the parameters obtained with each soil’s stratification
Fig. 3, the experimental curve of apparent resistivity together method.
CALIXTO et al.: PARAMETERS ESTIMATION OF A HORIZONTAL MULTILAYER SOIL USING GA 1255
TABLE VIII
STRATIFICATION PARAMETERS RESULTS—CASE STUDY 2
TABLE IX
EXPERIMENTAL RESISTIVITY—CASE STUDY 3
Fig. 4. Experimental and theoretical curves of apparent resistivity of the soil
(two layers stratification)—Case Study 3.
TABLE XI
DEVIATIONS BETWEEN EXPERIMENTAL AND
THEORETICAL CURVES— CASE STUDY 3
TABLE X
THEORETICAL RESISTIVITY VALUES— CASE STUDY 3
TABLE XII
STRATIFICATION PARAMETERS RESULTS—CASE STUDY 3
C. Case Study 3
In this case, a comparison is made between a soil’s stratifi-
cation method where the number of layers is inferred to be
two, and the proposed method of stratification that estimates the TABLE XIII
number of layers. Optimizing the quantity of stratified layers, DEVIATIONS BETWEEN EXPERIMENTAL AND
the GA is able to obtain a theroretical curve of soil apparent re- THEORETICAL CURVES—CASE STUDY 3
sistivity closer to the experimental one. Table IX presents the
points of the experimental curve of apparent resistivity distance
function between electrodes. Table X contains the points of
the theoretical curves of apparent resistivity, calculated with
the two-layers soil stratification process through three different
methods: proposed GA, quasi-Newton method, and the com-
plex images method. The values from Tables IX and X are then
plotted in Fig. 4.
Table XI, presents the calculated deviations between the ex-
perimental soil apparent resistivity curve, and the theoretical
ones computationally obtained by the three methods. Table XII in Table XIII, and the stratification parameters obtained by the
lists the parameters values found from each one of the three proposed GA are in Table XIV.
stratification processes. Fig. 5 illustrates the soil stratification process for Case Study
Finally, Table XIII and Table XIV present the results of the 3 where the number of soil layers was not fixed, but instead,
stratification made by the proposed GA. The deviations between was optimized by the proposed GA, resulting in a three-layers
experimental and theoretical curves of apparent resistivity are stratification. The deviations values decreased, resulting in a
1256 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 3, JULY 2010
REFERENCES
[1] D. H. Griffiths and R. F. King, Applied Geophysics for Engineers and
Geologists: The Elements of Geophysical Prospecting. Oxford, U.K.:
Pergamon, 1981, GBR.
[2] E. Orellana, Prospeccion Geoelectrica por Campos Variables.
Madrid, Spain: Biblioteca Tecnica Philips: Paraninfo, 1974.
[3] J. L. del Alamo, “A comparison among eight different techniques to
achieve an optimum estimation of electrical grounding parameters
in two-layered earth,” IEEE Trans. Power Del., vol. 8, no. 4, pp.
1890–1899, Oct. 1993.
[4] H. Yang, J. Yuan, and W. Zong, “Determination of three-layer earth
model from wenner four-probe test data,” IEEE Trans. Magn., vol. 37,
no. 5, pp. 3684–3687, Sep. 2001.
[5] P. J. Lagace, J. Fortin, and E. D. Crainic, “Interpretation of resistivity
Fig. 5. Experimental and theoretical curves of apparent resistivity of the soil
sounding measurement in n-layer soil using electrostatic images,”
(three-layers stratification)—Case Study 3.
IEEE Trans. Power Del., vol. 11, no. 3, pp. 1349–1354, Jul. 1996.
[6] B. Zhang, X. Cui, L. Li, and J. He, “Parameter estimation of horizontal
multilayer earth by complex image method,” IEEE Trans. Power Del.,
vol. 20, no. 2, pt. 2, pp. 1394–1401, Apr. 2005.
theoretical curve of soil apparent resistivity closest to the ex- [7] P. J. Lagace, M. H. Vuong, M. Lefebvre, and J. Fortin, “Multilayer re-
perimental one. sistivity interpretation and error estimation using electrostatic images,”
IEEE Trans. Power Del., vol. 21, no. 4, pp. 1954–1960, Oct. 2006.
[8] Recommended Practice for Grounding of Industrial and Commercial
Power Systems, IEEE Std. 142-1991, 1991.
VI. COMMENTARY [9] Guide for Measuring Earth Resistivity, Ground Impedance and Earth
Surface Potentials of a Ground System, IEEE 81:1983, 1983.
In all cases, the GA proposed in this paper achieved the best [10] I. F. Gonos and I. A. Stathopulos, “Estimation of multilayer soil pa-
performance. However, there will be cases where the obtained rameters using genetic algorithms,” IEEE Trans. Power Del., vol. 20,
no. 1, pp. 100–106, Jan. 2005.
results will not be coherent, not having any physical meaning. [11] E. Orellana, Prospeccion Geoelectrica en Corriente Continua.
This fact raises suspicions that at the stratified location, an Madrid, Spain: Biblioteca Tecnica Philips: Paraninfo, 1974.
oblique crack, an immense rock, or some extremely alkaline [12] R. W. Ryder, Earth Principles and Practice. London, U.K.: Pitman
and Sons, 1969.
material is present, changing the soil’s layers behavior and, [13] Y. L. Chow, J. J. Yang, and K. D. Srivastava, “Grounding resistance of
thus, its resistivity values. buried electrodes in multi-layer earth predicted by simple voltage mea-
Another issue that usually occurs is concerned with the surements along earth surface: A theoretical discussion,” IEEE Trans.
Power Del., vol. 10, no. 2, pp. 707–715, Apr. 1995.
number of layers estimation. The best estimation for the [14] A. L. Kinyon, “Correlation of measured and calculated substation
number of layers , without taking into account an optimiza- ground grid resistance,” AIEE Trans., 1959, paper 59-806, Nov.
tion process, consists in the calculation of the numbers of [15] F. A. Wenner, “Method of measuring earth resistivity,” Bull. Nat. Bu-
reau Std., Washington D.C., vol. 12, 1916.
inflections at the curve of experimental apparent resistivity of [16] F. P. Dawalibi and D. Mukhedkar, “Ground electrode resistance mea-
the soil in . A good demonstration of this fact surements in nonuniform soils,” IEEE Trans. Power App. Syst., vol.
can be observed in Case Study 3, where a 2-layers stratification PAS-93, no. 1, pp. 109–115, Jan. 1974.
[17] E. D. Sunde, Earth Conduction Effects in Transmission Systems. New
was forced. Analyzing the curve of experimental resistivity of York: MacMillan, 1968.
Fig. 4, it is possible to observed two inflection points, causing [18] Z. Michalewicz and D. B. Fogel, How to Solve It: Modern Heuristics.
a stratification in three layers. Developing a stratification in Berlin, Germany: Springer, 1999.
[19] W. P. Calixto, “Application of conformal mapping to the calculus of
three layers for Case Study 3, the deviations values are reduced Carter’s factor,” (in Portuguese) M.Sc. dissertation, Elect. Comput.
considerably, see Fig. 5. However, not always does the exper- Eng. School, Federal University of Goias, Goiania, Brazil, 2008.
imental curve of apparent resistivity really represent the soil [20] D. E. Goldberge, Genetic Algorithms in Search, Optimization, and Ma-
chine Learning. Reading, MA: Addison-Wesley, 1989.
with its actual layers quantity. This occurs because the soil does [21] M. Mitchell, An Introduction to Genetic Algorithms. Cambridge,
not have homogeneous behavior. MA: MIT Press, 1996.
CALIXTO et al.: PARAMETERS ESTIMATION OF A HORIZONTAL MULTILAYER SOIL USING GA 1257
[22] T. Back, Evolutionary Algorithms in Theory and Practice: Evolution Marcel Wu (S’10) received the B.Sc. and M.Sc. de-
Strategies Evolutionary Programming Genetic Algorithms. Oxford, grees in electrical engineering from the Federal Uni-
U.K.: Oxford Univ. Press, 1996, GBR.. versity of Uberlandia-Electromagnetism and Electric
[23] J. H. Holland, Adaptation in Natural and Artificial Systems. Cam- Grounding Systems Nucleus Research and Develop-
bridge, U.K.: MIT Press, 1992. ment, Uberlandia, Minas Gerais, Brazil, in 2007 and
[24] S. J. Pirson, Geologic Well Log Analysis. Houston, TX: Gulf, 1963. 2009, respectively, where he is currently pursuing the
[25] Y. L. Chow, J. J. Yang, and K. D. Srivastava, “Complex images of a Ph.D. degree.
ground electrode in layered soil,” J. Appl. Phys., vol. 71, pp. 569–574, He is Professor at the Northwest College of Minas,
1992. Minas Gerais. His fields of interest are electrical ma-
chines and electrical grounding.
Wesley Pacheco Calixto (M’09) received the B.Sc.
degree in physics and the M.Sc. degree in electrical
engineering from the Federal University of Goias,
Goiania, Brazil, in 2008.
He is currently a graduate student in the Depart- Keiji Yamanaka received the Ph.D. degree in arti-
ment of Electrical Engineering, Federal University ficial neural networks from the Nagoya Institute of
of Uberlandia-Electromagnetism and Electric Technology, Nagoya, Japan, in 1999.
Grounding Systems Nucleus Research and De- Currently, he is Professor in the Department
velopment, Uberlandia, Minas Gerais, Brazil, and Electrical Engineering, Federal University of Uber-
Lecturer at the Catholic University of Goias and the landia-Electromagnetism and Electric Grounding
Unified Institute of Higher Education. He works on Systems Nucleus Research and Development,
projects for the development and research into electrical grounding systems Uberlandia, Minas Gerais, Brazil. His areas of
at CELG—the Goias Electricity Supply Company. Interests: Electrical Ma- interest include computational intelligence, pattern
chinery, Electric Grounding Systems and Optimization. recognition, optimization, and high-performance
computation.