International Journal of Electrical and Electronics Engineering 1:2 2008

Effect of Non Uniformity Factors and

Assignment Factors on Errors in Charge
Simulation Method with Point Charge Model
Gururaj S Punekar, N K Kishore Senior Member IEEE, H S Y Shastry
[P] [Q] = [V]
(1)
Where
[P] is the potential coefficient matrix
[Q] is the column vector of unknown charges
[V] is the column vector of known potentials at the contour
points.
Resulting simulation accuracy strongly depends on the type
and number of charges, locations of contour points and
complexity of electrode geometry [1-2]. As CSM accuracy
depends on the choice of type of simulating charge, their
number, location of these charges and the contour points, the
CSM programs for particular application become case specific
and depend on the programmer (or developer). Hence, to set
up an accurate CSM model, familiarity and understanding of
the programmer (or developer) with the CSM plays a key role.
In order to help the user, the empirical relations relating
location of charges with those of the contour points become
useful. One such widely used parameter is the assignment
factor fa [1]. The attempts have also been made to locate the
charges using optimized charge simulation methods [5-9]
instead of assignment factor as the guiding parameter. With
GA as a tool with number of charges pre-decided (by the
programmer or developer) using point charges (also predecided by the user) automatic allocation of these charges and
contour points is attempted, relatively recently [10]. All these
efforts are to reduce the need of users experience in setting up
accurate CSM models. Even with all these efforts the CSM
programs have remained user and case specific as regards to
the choice of type and number of charges. The CSM being a
semi analytical technique makes use of potential and field
coefficients of simulating charge configurations [1-2], user
interference and knowledge can be an advantage. Based on
the symmetries of the simulating charges and those of the
geometry simulated, user with his/her experience can guide
the charge arrangement in relation with the contour point
locations. Hence, it is felt that empirical guiding parameter
like assignment factor and its impact on the simulation errors
needs further understanding. The present work is such an
effort, with symmetrically arranged six point charges in
simulating electric fields associated with the sphere-plane
geometry.
The earlier investigations [1, 11] of CSM errors involving
assignment factors are not relating to electric field non

AbstractCharge Simulation Method (CSM) is one of the very

widely used numerical field computation technique in High Voltage
(HV) engineering. The high voltage fields of varying non
uniformities are encountered in practice. CSM programs being case
specific, the simulation accuracies heavily depend on the user
(programmers) experience. Here is an effort to understand CSM
errors and evolve some guidelines to setup accurate CSM models,
relating non uniformities with assignment factors. The results are for
the six-point-charge model of sphere-plane gap geometry. Using
genetic algorithm (GA) as tool, optimum assignment factors at
different non uniformity factors for this model have been evaluated
and analyzed.
It is shown that the symmetrically placed six-point-charge models
can be good enough to set up CSM programs with potential errors
less than 0.1% when the field non uniformity factor is greater than
2.64 (field utilization factor less than 52.76%).

KeywordsAssignment factor, Charge Simulation Method, High

Voltage, Numerical field computation, Non uniformity factor,
Simulation errors.

I. INTRODUCTION

HE charge simulation method (CSM) is one of the widely

used numeric field computation technique ideally suited
for simulating open boundary problems [1-3]. The high
voltage engineering makes use of this technique extensively
[4]. The method in its simplest form (conventional CSM)
computes the charge magnitudes by satisfying the boundary
conditions at the selected number of contour points. The
locations of the charges and the boundary conditions are
predetermined and supplied based on the experience [1] of the
researcher. The unknown charges are computed from the
relation (1) by setting up simultaneous equations

Gururaj. S. Punekar is with the with the Department of Electrical &

Electronics Engineering, National Institute of Technology, Karnataka 575025
INDIA (phone:+91-(0)824-2474000Ext3495; fax: +91-(0)824-2474033; email: gururaj.punekar@ gmail.com).
N. K. Kishore, is with the Department of Electrical Engineering, Indian
Institute
of
Technology,
Kharagpur
721302
INDIA
(e-mail:
[email protected]).
H. S. Y. Shastry is with the Department of Electrical & Electronics
Engineering, National Institute of Technology, Karnataka 575025 INDIA (email: [email protected]).
.

115

International Journal of Electrical and Electronics Engineering 1:2 2008

specific to a model based on the type of charges and their

number. For sphere-plane model this can vary in the range of
0 to2/ (=0.63) as explained below, in the model details.

uniformities. Also, the efforts give guidelines with ring

charges as the fictitious charges and indicate that assignment
factors in the region of 1 to 2 can achieve acceptable
accuracies.
The effort in this work has been to relate the assignment
factor with the field non uniformity factor [12] which
quantifies the degree of non uniformity associated with
geometry. This has been attempted with six-point-charge
model of sphere-plane gap. Using GA as tool best assignment
factors at which errors are minimum have been computed at
selected set of non uniformity factors, covering the wide range
(near uniform to highly non uniform) of non uniformities.
These results are believed to be unique and are being reported,
perhaps, for the first time. They should help the CSM
programmers, in furthering the understanding of CSM errors
and aid in setting up accurate CSM programs.

B. Model details
1) Geometric details of sphere-plane model
The geometric model of the sphere-plane gap with the
image sphere is as shown in figure 2. Image sphere is used to
simulate the infinite ground plane.
Z

HV Sphere
Potential V=1 p.u.
Radius r=1 p.u.
-X

II. PARAMETERS OF STUDY AND MODEL

A. Assignment Factor fa
It is defined as the ratio of the distance between a contour
point and the corresponding charge a2 to the distance
between two successive contour points a1, as given in
relation (2). The schematic showing the charges and the
contour points given in figure 1 depicts distances a1 and
a2.

h=gap
-Y

Image Sphere
X
-Z
Fig. 2 The sphere-plane gap simulated (shown along with the
Cartesian frame of reference and image electrode).

The sphere electrode radius, r, is considered as 1 per unit.

And with respect to this, the gap separation is, h per unit. It
is the dimension of h in relation with r that decides the
electric field non uniformity (and hence the electric field non
uniformity factor) of the geometry [12].

Fig. 1 Schematics (a) and (b) to explain assignment factor.

fa =

a2
a1

(2)

2) CSM Model details

The sphere-plane geometry (figure 2) is simulated using the
six point charges arranged symmetrically as shown in figure 3.
The charges are placed inside the sphere with corresponding
image charges inside the image sphere. These charges are on a
concentric sphere of radius rc. This radius rc can assume
values form 0 to r(=1); where r is the radius of the sphere
electrode. This forms a2 of the assignment factor (equation
2). The contour points in the model are chosen such that, they
are on the electrode surface, along the line joining the
corresponding charge and the center of the sphere. Then the
distance between the two successive contour points is r/2
units. This distance forms a1 of the assignment factor
(equation 2). Hence, with these ranges for a1 and a2, the
assignment factor can assume value between 0 to 2/ (i.e. 0

In setting up a CSM model, in simulating particular gap

geometry, user decides on the type of simulating charges
based on the profile of the geometry. S/he also decides the
number and general arrangement taking in to account the
accuracy requirements and symmetry. Further, it is the exact
location of these charges in relation with the contour points
that needs to be decided, to maximize the accuracy. This
aspect is quantified by the parameter, assignment factor.
Literature, states that it should be in the region of 1 to 2 for
low error [1, 11]. This is true, with ring charges, particularly
with large number of charges (number of charges 10 or more).
In actuality, the range over which this parameter can vary is
116

International Journal of Electrical and Electronics Engineering 1:2 2008

The maximum error in potential in this case is seen to be of

the

to 0.63).
q5
q2

r=1 p.u

O
q4

q3

q1
q6

Fig. 3 General charge arrangement within the HV sphere shown in

figure 1. (O is the center of the sphere. Charges q1, q2, q3, q4,
q5 and q6 are the six charges placed on the coordinate axes
within the sphere)

C. Field non uniformity factor f

The electric field non uniformity is defined as the maximum
electric field intensity in the gap (which occurs near the tip of
the high voltage electrode to the average electric field
intensity (V/h; figure 2). The reciprocal of this, field non
uniformity factor, is called the electric field utilization factor.
The field utilization factor can assume value from 0 to 100%
and can be used to interpret how best the insulation in the gap
is being utilized. These factors have been extensively used in
the literature in interpreting the electric field dependency,
breakdown and corona inception behaviors of electrical
insulation. For this purpose these factors are also computed
and reported in the literature for the simple geometric forms
[12, 13]. These parameters are used to understand the error
variation in simulating the CSM model along with charge and
contour point arrangement. The field utilization factors (also
the field non uniformity factor) which depend on the gap
spacing h and the sphere radius r are reported for the
sphere-plane gap in reference [12]. This data is used to
compute the corresponding CSM errors by successive
simulations.

Fig.. 4 Plot of percentage potential error on the sphere surface for

sphere-plane model. (Assignment factor fa=0.6, field non uniformity
factor f=9.55).

order of 1.3e-3 percent. The field deviation angle error is the

angular difference between the normal component of electric
field intensity at the electrode surface and surface normal
drawn at that point. The field deviation angle error for this
particular case is observed to be less than 0.14. These errors
are acceptable, as generally the CSM errors less than 0.1% in
the potential is considered reasonable [2]. The similar test
simulation runs indicated that the errors less than 0.1% are
achievable with the field utilization factors being less than
52.76% (non uniformity factor greater than 2.64).
The variation of maximum potential error as a function of
assignment factor obtained by numerical experimentation in
the range 0.56 to 0.63 is given in figure 5. As seen from these
simulation results the maximum potential error varied in the
range of 113% to 1.14e-006%. Due to this wide variation, in
this error plot, z-axis is with logarithmic (base 10) values.

III. RESULTS AND DISCUSSION

The numerical experiments are conducted on the six-pointcharge CSM model of the sphere plane gap with the
assignment factor and the field utilization factors as the
parameters. The results for maximum potential error on the
surface of spherical electrode taking 100 test points, computed
for selected values of field utilization factor, chosen from the
entire range are reported. The useful range of the assignment
factors varying from 0.2 to 0.63 are considered in analyzing
the CSM simulation errors.
The plots of percentage potential error on the sphere
electrode surface for the field non uniformity factor of 9.55
(electric field utilization factor of 10.5%), with typical
assignment factor (with fa=0.6) is as given in the figure 4. The
x and y axes in these plot are the angle phi and theta
coordinates of the points on the sphere (polar co-ordinates).

Fig. 5 Percentage maximum potential error on the sphere surface as

a function of assignment factor fa for selected values of field
utilization factor (over its entire range).
117

International Journal of Electrical and Electronics Engineering 1:2 2008

The plot shows dependency of CSM maximum potential

error on both assignment factor and the field utilization factor.
At higher field utilization factors the influence of the
assignment factors is relatively low with errors being higher.
But in the lower range of field utilization factors (<40%), the
assignment factors show a higher influence. There is definite
unique value of the assignment factor (specific to non
uniformity factor) at which the simulation errors are a
minimum. The observation made is, as the utilization factor
decreases the assignment factor at which best results are
obtained shifts towards the higher value.
The maximum field deviation angle error, also termed as
Beta error (max), varied in the range of 65.3 to 1.47e-005
(over the entire range of field utilization factor and useful
range of assignment factor studied).
At lower values of field utilization factor the assignment
factor shows a great degree of influence. On the contrary as
the field utilization factor decreases (from 100%), the
simulation accuracies increase (for any chosen assignment
factor). It is to say that, high accuracy simulations are
possible, when the fields are non uniform.
The summation of the simulating charges in CSM should
remain constant irrespective of their locations (also of contour
point) for a fixed non uniformity. The variation in this can be
a possible error criterion, which is investigated. Figure 6
shows the variation in summation of simulating charges as a
function of assignment factor and the field utilization factor.
As expected, system capacitance decreases with decrease in
field utilization factor. But the summation of simulating
charges shows little variation with the changes in the
assignment factor.

Optimal assignment factors using Genetic Algorithm (GA)

The optimal value of the assignment factor is obtained for
few typical values of the utilization factors chosen from its
entire range (listed in [12]), using GA as the optimization tool
[14]. The root mean square (rms) potential error on the surface
formed the objective function in error minimization. The rms
value is obtained by evaluating errors at 100 regularly spaced
points on the surface of sphere. The GA as the optimization
tool used randomly generated initial population of size 40.
The algorithm used 25 number of generation as the
termination criteria. The bound for assignment factor used is
0.2 to 0.63.
The evaluated optimal values of the assignment factors for
the corresponding field non uniformity factors are given in
table-I. Using these optimal values of assignment factors
(obtained using GA-CSM program), the maximum potential
error, maximum deviation angle error, rms potential error and
rms deviation angle error are calculated for different field non
uniformity factors listed in table-I. These results are given in
figure 7 and 8. Plots in figure 7 indicates that, at higher values
of assignment factor the simulation errors are the lowest; but
higher optimal values are possible only at higher non uniform
field factors (as indicated by figure 8 and table-I).

Fig. 7 CSM simulation errors as a function of optimal assignment

factor (corresponding to the each non uniformity factor) for sphere
plane gap (with six-point-charge model).

These results indicate that even with the optimal assignment

factor there is a definite limit to achieving minimum error for
a particular non uniformity. Further reduction in errors, is
possible only with the different set (shape) of charges and/or
increased number of charges. In this sense, CSM models have
remained case specific and user specific as regards to the
choice of type and number of charges. It is felt, that under
such circumstances this effort to analyze the error variations
with few (six in this case) symmetrically arranged point
charges (being the most basic and simplest charge
configuration) and the results reported for different non

Fig. 6 Plot of summation of simulating charges of the sphere as a

function of assignment factor fa for selected values of field
utilization factor over its entire range (six-point-charge model).

118

International Journal of Electrical and Electronics Engineering 1:2 2008

[3]

TABLE I
FIELD NON UNIFORMITY FACTOR, OPTIMAL ASSIGNMENT FACTORS AND
CSM ERRORS
Root Mean Square
Non Uniformity
Optimal Assignment
Potential Error in
Factor
Factor
Percentage
1.0336
0.2417
7.5446
1.1020

0.3154

3.4553

1.6846

0.4561

0.4087

2.0342

0.4896

0.1752

3.1516

0.5222

0.0690

9.5511

0.5830

0.0031

49.5049

0.6188

3.5794e-005

100.0000

0.6269

3.2093e-006

1000.0000

0.6286

1.4768e-006

[4]

[5]

[6]
[7]
[8]

[9]

uniformity factors will be useful.

[10]

[11]

[12]
[13]
[14]

Gururaj S Punekar received his B.E (Electrical Engg.) from Karnataka

University Dharwad, Karnataka, India. He received the MSc (Engg.) degree
from High Voltage Engineering department, IISc, Bangalore in 1991. He is
presently pursuing his doctoral research having registered with Department of
Electrical Engineering, IIT Kharagpur, West Benga, India.
Currently, he is with department of Electrical Engineering, National Institute
of Technology, Karnataka, India. He was with Tata Electric Co., Mumbai, in
the year 1991-92. From 1992 he is in academics. His areas of interest include
HV testing, electric field computation, optimization and GIS.
Mr. Punekar is a member of Institution of Engineers (India), Life member of
Indian Society for Technical Education, Life Member of Systems Society.
N K Kishore (SM96) obtained B.E.(Electrical Engg.) from Osmania
University, Hyderabad in 1983, M.E. ( Electrical Engg.) from IISc Bangalore
in 1985 and Ph. D from IISc. Bangalore in 1991. He worked as a Scientific
Officer with IISc Bangalore from 1987 to 1991. He joined on the faculty of
Electrical Engg. at IIT Kharagpur from 1991.
Currently, he is a Full Professor there. His areas of interest include High
Voltage Engineering, Power Systems, Lightning, EMI/EMC, Condition
Monitoring of Power Apparatus and Industrial Applications of High Voltages.
Dr Kishore is a Senior Member of IEEE, a Fellow of Institution of Engineers
(India) and a life member of System Society of India.
H. S. Y. Shastry received BE degree in electrical engineering from Bangalore
University, Karnataka, India, (1969), M.E. degree in power systems from
Bangalore University, Karnataka, India, (1971), and the Ph.D degree from IIT
Kanpur (1984).
He joined department of electrical engineering K R E C Surathkal (presently
NITK), Karnataka , India as lecturer in the year 1972, became Assitant
Professor in the year 1984. He became full Professor in the year 1994. His
areas of interests include power system dynamics and numerical computations.
Dr Shastry is a member of Indian Society for Technical Education (ISTE).

Fig. 8 CSM simulation errors as a function of non uniformity factor

(with optimal assignment factor based charge locations) for sphere
plane gap (with six-point-charge model).

IV. CONCLUSION
The CSM error dependency on charge-contour point
arrangements and field non uniformity is reported for sphereplane model using most elementary charges, namely, point
charges.
Generalized interpretation is, its simpler to setup highly
accurate CSM models when the electric field non uniformity
of the geometry being simulated is high. In such situation the
charges are clustered far away from the electrode surface
(contour points).
REFERENCES
[1]
[2]

S. Chakravorty, Charge Simulation Method: a critical overview, Ie (I)

journal-el, Vol. 78, March 1998, pp. 210-214.
E. Kuffel, W.S.Zeangle & J.Kuffel, High Voltage Engineering
fundamentals, Newnes, An imprint of Butterworth-Heinemann, A
division of Reed Educational and Publishing Ltd, Woburn, MA. 2000,
pp.254-269.
H Anis, A Zeitoun, M El-Ragheb m and El-Desouki, Field calculations
Around Non-Standard Electrodes Using Regression and Their Spherical
Equivalence, IEEE Trans on PAS, Vol.PAS-96, no.6, NovemberDecember 1977, pp.1721-1730.
A Yializis, E.Kuffel, P H Alexander, An Optimized Charge Simulation
Method for the Calculation of High Voltage fields, IEEE Transaction
on PAS-97, 1978, pp. 2434-40.
Y L Chow and C Charalambous, Static-field computation by method of
optimized simulated images, Proc. IEE, Vol.126, No.1, 1979, pp.123125.
M R Iravani, M R Raghuveer, Accurate Field Solution in the Entire
Inter electrode Space of A Pod-Plane Gap Using Optimized Charge
Simulation, IEEE Transactions on EI Vol.EI-17 No.4, August 1982, pp.
333-337.
M M Abouelsaad, M M El Bahy: Accurate Field Computation of
Needle-Plane Gaps using an Optimized Charge Simulation Method,
Conference on Electrical Insulation and Dielectric Phenomena, 2000,
pp506-509.
Ryo Nishimura, Katsumi Nishimori, Naganori Ishihara, Automatic
arrangement of fictitious charges and contour points in charge
simulation method for two spherical electrodes, Journal of electrostatics
57, 2003, pp. 337-346.
M.Th.El-Mohandes, H Okubo, Error analysis based on the interaction
between simulating charges in the CSM for the Electric-Field
Calculation of HV Apparatus, European Transactions on Electric
Power, ETEP, Vol.4, No.6, November/December 1994, pp565-570.
H. McL Ryan and C.A. Welley, Field Auxiliary Factors for simple
electrode geometries. Proc IEE Vol 114, no.10, Oct, 1967, pp 15291536.
Y. Qui, Simple expression of field nonuniformity factor for
hemisphereically capped rod-plane gap, IEEE Trans. Electrical
Insulation, Vol.21, No.4, 1986, pp. 673-675.
Christopher R Houck, Jeffery A. Joines and Michael G.Kay, A genetic
algorithm for function optimization: A Matlab implementation, North
Carolina
State
University,
available
at
http://www.ie.ncsu.edu/mirage/GAToolBox/gaot//papers/gaotv5.ps.

H Singer, H. Steinbigler P. Weiss, A change simulation method for the

calculation of high voltage fields, IEEE trans, PAS vol.93, 1974,
pp1660-68.
Nazar H Malik, A review of the charge simulation method and its
application, IEEE Transaction on electrical insulation, Vol.24, No.1,
1989, pp. 1-20.

119