IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Grounding grid design in electrical power substation using optimization

methods
To cite this article: V K Voon et al 2019 IOP Conf. Ser.: Mater. Sci. Eng. 495 012037

View the article online for updates and enhancements.

This content was downloaded from IP address 103.219.236.4 on 02/09/2020 at 06:38

CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

Grounding grid design in electrical power substation using

optimization methods

V K Voon, K I Wong, T C Tiong, A Mansour and K H Law

Department of Electrical and Computer Engineering, Faculty of Engineering and
Science, Curtin University, Malaysia, CDT 250, Miri 98009, Sarawak, Malaysia

E-mail : [email protected]

Abstract. A good substation grounding grid is essential to ensure safety of personnel working
in the substation, and also to protect the equipment to ensure continuous operation of the
substation. The design of the substation grounding grid is often carried out to meet the IEEE
Std 80-2000 requirements of grounding resistance, touch voltage and step voltage at an
economical cost. However, the conventional grounding grid design is normally over calculated
to meets these IEEE requirements without considering the total cost. The grounding grid is
often costly to implement due to high number of grounding conductors and rods, and large area
size of the interconnected bare conductors. In this study, Gravitational Search Algorithm
(GSA) and Particle Swarm Algorithm (PSO) are proposed to achieve an effective grounding
grid calculation. The reliability of these algorithms has been validated using a typical medium
voltage substation grounding grid parameters. It was found both that GSA and PSO were
capable to achieve a lower cost with of grounding grid with 29% cost saving as compared to
the conventional calculation technique, without compromising the safety requirements of
grounding grid.

1. Introduction
The purpose of substation grounding system is to provide a safe neutral point for all the conductive
part of the substation equipment, including transformer, capacitor and reactor, and switchboard. It also
provides the discharge path for the lighting impulses, fault current, or any high voltage switching
related surges into the earth via the protective conductors so that it can ensure the safety of the
personnel by limiting the hazardous potential gradients that can exist anywhere within a substation.
Besides that, worker always disconnect and ground live equipment in substation before they proceed
with the maintenance works. This is to ensure any electrical current in the conductors is discharged
into the ground. Thus a good grounding system is playing an essential role in term of protection for
personnel and equipment.
The grounding grid that is comprised of interconnected horizontal bare conductors and vertical bare
conductor (or rod), is often buried under a substation to provide sufficient low resistance path to
ground to minimize rise in ground potential with respect to remote ground [1]. It is necessary to ensure
that the entire ground potential rise is as low as possible so that personnel are not exposed to the
danger of hazardous electrical shock and equipment are protected from any circumstances of short
circuit current or over voltage transient during an abnormal or fault incident. The calculated touch
voltage and step voltage that are computed as defined in IEEE Std 80-2000 during the abnormal
condition must be within the safe values [1]. Safe values are defined as the values that are not large

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

enough to induce a current that can cause ventricular fibrillation in human body [1].Without the safe
and effectively grounded system, the safety and fitness of the personnel and equipment will be at
higher risk getting electric shock, due to the failure of protection relaying system and the high rise of
the ground system potential that is unsafe for human contact.
Design of a proper substation grounding system based on ANSI/IEEE Std 80-2000 are related to
how to achieve the thresholds of a number of variables, such as touch voltage, mesh voltage, step
voltage, tolerable touch voltage, tolerable step voltage, ground potential rise and grid resistance. These
safety aspects of variables have been widely revised over the past years for safety improvements. This
paper describes about applying gravitational search algorithm (GSA) and particle swarm optimization
(PSO) for modelling and analysis optimization for ground grid system.
The contribution of this paper is total grounding conductor length reduction which results in 29%
cost saving compared to the existing un-optimized grounding grid.

2. Literature Review
The grounding grid design parameters, which are consists of step voltage, touch voltage as well as
grounding resistance, are the most important criteria to ensure safety in a substation grounding. The
method proposed by ANSI/IEEE Std 80-2000 is not enough to provide an accurate solution. Therefore,
the related computer program and numerical computing methods have been developed since 1970.
Dividing the grounding conductors and grounding rods using the numerical simulations and
superposition was first proposed in [2]. In addition, the unbalance current distribution could be
handled by multi step analysis of interconnected earthing electrodes.
Prior computing work on grounding resistance considering the effects of leakage current magnitude
due to adjacent cross conductors, parallel conductor and end effect was done in [3]. This approach
could also verified touch and step potentials. Several heuristic techniques have been developed over
these years incorporated optimization process for the grounding grid system [3]. The system
optimization objective was to minimize the cost function with optimized parameters for the grounding
system [3]. These included conductor size, spacing between conductors, total excavation area, grid
depth for the grounding grid. The optimized design should produce actual step and touch voltages
below the tolerable limits.
Evolutionary algorithm was used to optimize the design of 115/13kV substation with a size of 70x70m
grounding grid (without the rod). The computed earth surface was compared with the corresponding
formulae of ANSI/IEEE Standard 80-2000 and the results were in good agreement.
PSO was used to optimize the grounding grid design [5]. The authors proposed an cost
minimization objective with constraints including material, installation, and excavation cost. PSO
was applied in [6] to optimize the substation grounding grid with 200 x 150m size and fault current
21kA. The author proposed a cost function which included the diameter of the rod, length of the rod,
depth of grid, excavation by taking into account the safety tolerance of ground potential rise (GPR),
step potential and touch potential as defined in ANSI/IEEE Standard 80-2000.

3. Methodology

3.1. Gravitational Search Algorithm

GSA is an algorithm which is based on Newton’s law gravitation and is proposed by an Iranian
professor, Esmat Rashedi, in 2009 [7]. Based on this algorithm, it stated that any two particles in the
universe will be attracted to each other by a force that in the direction of the centre line. From the
algorithm, agents are normally the objects which its performance is determined by means of masses.
The whole agents will be attracted to each other by the gravitational force and subsequently the
movement for all agents are induced by the force globally move towards the agents with heavier
masses. In GSA, there are 4 variables which are inertial mass, position, active gravitational mass and
passive gravitational mass. Based on the algorithm, every population will have two capabilities which
are exploitation and exploration. From the exploration capability, it prevents local optimum issue at

2
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

the beginning. Other particles are attracted by a time function which called as Kbest particle agent.
Nevertheless, the performance of GSA is essentially enhanced by controlling the exploration and
exploitation will decrease linearly according to the time and there left only the agent with the heavy
mass which represents the final solution. The complete steps of the GSA are described below:
a) Identification of search space
b) Produce the Initial population
c) Assess the fitness function for each particle of the population
d) Update the gravitational constant value accordingly
GSA can be concluded as the following steps:
Step 1: Initialization of Agents – Within the given search interval, initialize the position of N
number of agents randomly by using (1)

 
 xi  xi1..., xid ..., xin for i

where xid is the position of i th mass in d th dimension.

Step 2: Compute best fitness and fitness evolution for each required agent at each iteration (t). From
each iteration, it should locate the best and worst fitness.

best (t )  min fit j (t ) 

j{1,...,N ma}

worst (t )  max fit j (t ) 

j{1,...,N mass}

where fit (t) is defined as the fitness value of particle of each iteration, best(t) and worst (t) are best
which is minimum and worst which is maximum fitness of all agent.
Step 3: Compute gravitational G at iteration t as (4)

( )
( ) 

where : initial gravitational constant

α : constant
T : total number of iteration of algorithm
Step 4: Calculate the mass of agents at each iteration (t) from (5) to (7)

M ai  M pi  M ii  M i 

fiti (t )  worst(t )
mi (t )  
best (t )  worst(t )

where mi is the inertial mass of particle i

mi (t )
M i (t )  

Nmass
m j (t )
j1

Step 5: Calculate each agent’s acceleration at each iteration 

Fi d (t )
aid (t )  
M ii (t )

3
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

where

N mass
Fi d (t )   rand F
jkbest, j i
d
j ij (t ) 

kbest : set of first K agents with the best fitness value and biggest mass.
Step 6: Update velocity and position of the agents at the next iteration (t + 1)

vid (t  1)  randi  vid (t )  aid (t ) 

Massid (t  1)  Massid (t )  vid (t  1) 



where rand : random number from [0, 1]

: speed of particle i at dimension d at time t.
: position of particle i at dimension d at time t.
Step 7: Repeat the steps from step 2 to step 6 until the iteration reach the criteria. The algorithm
will return the value of positions of the corresponding agents at specified dimension at final iteration.

3.2. Particle Swarm Optimization

PSO was introduced by Kennedy and Eberhart that adopted the idea of swarm behaviour of birds in
1995 [6]. The basic ideas of PSO are:
a. Each particle is searching for optimum and at the same time each particle is moving and hence
has a velocity.
b. Along the moving progress, each particle will remember the position it was which develops the
best result.
c. By cooperation among the particles in the swarm, they exchange information of the locations
they have visited.
d. To exchange information, it depends on the neighbourhood associated. Then the particles will
know the fitness among these neighbourhood, and use the position which has the best fitness.
Positions are then used to update their velocities accordingly.

Step 1: Population Initialization

Initialize the population of the PSO with randomly generated npopu particles, population size, in the k-
dimensional searching space, and each particle include position Xi and velocity Vi where Xi represent
the position of ith particle, Xi =( Xi1…, Xij…, Xik) whereas Vi represent the velocity of ith particle,
Vi =( Vi1…, Vij…, Vik). The parameter of the search space represents the number of variable in the
application [6]. Position Xi and velocity Vi represent the solution of the problem and its displacement
in the searching space respectively.
Step 2: Fitness Function
Fitness function is determined for each particle. The higher the value of the particle, the better fit of
the particle. Otherwise the particle is considered less fit. The fitness function’s equation is shown
below.

( ) (12)
( ) ∑ ( )

where
Ob(.) : Objective function of optimization problem
constant that lets the denominator of Fit(.) be a positive

4
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

: Coefficient of p-th constraint

- are respectively the violations of step 2 to 5
Step3: Update Velocity and Position
Every particle keep hold of its optimal position value based on the characteristic of flying birds which
they accumulated the flying experience. Subsequently, the particle best local position, Pbesti, can be
selected respectively. The particles share their optimal values as well as their best local position
Subsequently the global best, Gbest can be selected according to the fitness value and position.
Therefore, the velocity of each particle will be guided by previous particle velocity (Pbesti and Gbest)
[6].

( ) ( ) (13)

(14)

Where
W = wmax - iter·(wmax - wmin)/itermax
Pbesti – (Pbesti1…, Pbestij…, Pbestik)
Gbesti – (Gbesti1…, Gbestij…, Gbestik)
Pbest is the optimal position of the ith particle
Gbest is the optimal position of entire particles
– acceleration coefficient
w – Coefficient of the inertia weight
wmin – minimum coefficient of the inertia weight
wmax – maximum coefficient of the inertia weight
iter – current iteration number
itermax – maximum iteration number

Step 4: Bound of velocity and Position

The particle will be updated accordingly with the bounds of the k-dimensional search space as shown
below

and (15)

, (16)

where ( )
– maximum velocity of j-th dimension
– minimum value of j-th dimension
– maximum value of j-th dimension

4. Simulation of grounding planning

The design is initiated for a substation rectangular grounding grid 132/11kV case study for a small
general industry with the transmission circuit and the step down delta-wye transformer. The
grounding grid size is given as 90m (length) and 50m (width). The average soil resistivity is
300 Ωm. Table 1 shows the substation parameters.

5
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

Table 1. Input parameters for substation.

Average Soil resistivity 300 .m
Fault Current 3100A
Resistivity of crushed Rock layer 3000 .m
uniform
Soil location type
Switch Yard operator >50kg
Soil location Type Uniform
Projection Factor 20%
Shock Duration 0.15s
Fault duration 0.15s
Surface layer derating factor 1.2
Crushed rock layer inside sub 0.10m
X/R ratio at the fault 15
Current Division Factor Sf 0.8
Length in X direction 90 m
Length in Y direction 50 m
Ambient Temperature 30°C
Grid Shape Rectangular

The objective function is to optimize 4 parameters for the grounding grid conductors, i.e. number of
horizontal rows, number of horizontal columns, vertical rod length and number of vertical rods. The
population size is set as 10n with n as the number of parameters to be optimized, thus 10 x 4 = 40.
The attempt of searching the optimized design is taken by 20 iterations to be completed during the
simulation. The upper and lower bound of these 4 defined parameters are 10 and 3 respectively. The
gravitational time constant G0 is set at 150 and the value of α is at 12. GSA input parameters are
shown in in Table 2.
Table 2. Input parameter setting for GSA.
Population Max Upper Lower Α G0
size iteration bound bound
40 20 10 3 12 150

The maximum tolerable voltages for step and touch scenarios can be calculated empirically from
IEEE Std Section 8.3 for operator body weights of 50 kg and 70 kg: Estep, 70  5664V ,
Estep, 50  4184V and touch voltage, Etouch, 70  1720V , Etouch, 50  1270V are obtained.
For setting up the optimization of GSA, there are methods used in objective function that will
minimize the cost of the construction for the rectangular grounding grid system. This could be
achieved through an optimization process which can find a minimal cost needed to construct a
grounding grid.
Based on the formula of ANSI/IEEE Std.80, the calculated maximum grid current IG = 1860A by
assuming that the ground conductor material are the copper-clad steel wire at ambient temperature
(Ta) of 300C . The ground potential rise (GPR) and grounding grid resistance with respect to remote
earth (Rg) are 5829V and 2.275 Ω respectively based on ANSI/IEEE Std.80[1].
GSA parameters have been initialized (such as x – number of horizontal rows, y – number of
horizontal columns, z – number of vertical rods, w – number of vertical rod length), with each agent
randomly selected while satisfying different equality and inequality constraints, upper bound and
lower bound as 10 & 3 respectively. There are 40 agents generated. Therefore each set in the agent
matrix X presents a solution for the 4 parameters. Then the grounding function program could
determine the different dependent variables such as tolerable touch and step voltages, GPR, grid

6
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

resistance and total length of conductor. The G(t), best(t), worst(t) and mi(t) are updated for each set
of agents. The fitness function or objective is calculated by using (9). The velocity and position of each
agent are also calculated by using (10) and (11) accordingly.
The dependent variables like total length, GPR etc will be iteratively calculated for the agents in the
matrix. In the end of process, the fitness value for each newly generated set of agent matrix X are
determined. This process is updating G(t), best(t), worst(t) and mi(t) to optimize the fitness value until
maximum iterations i.e. 20 is reached.
Figure 1 shows a screenshot for GSA simulation (one trial sample) with Matlab.

Figure 1. Screenshot for GSA simulation.

For setting up the optimization of PSO, there are also 4 parameters to be optimized which are number
of horizontal rows (x), number of horizontal columns (y), number of vertical rods (z), and vertical rod
length (w). For PSO optimization, the input parameters listed in Table 3. The population size is set as
10n with n as the number of parameters to be optimized, thus 10 x 4 = 40. Maximum iteration is set it
as 20. The upper bound and lower bound limits are set at 10 and 3 respectively, similar to the
parameter settings for GSA. An initial population is generated with randomly generated particles. Then
from each particle, the fitness value which is the total length of conductors will be determined for
number of horizontal rows, number of horizontal columns, number of vertical rods and each vertical
rod length. If the fitness value is better (lower) than its personal best, the current value of total length
will set as the new Pbesti. Therefore the particle with the best or lowest fitness value will be chosen as
Gbesti. For each particle, its position will be calculated according to (15). Subsequently the particle
velocity will be updated according to (16). Anyway, it is applicable to use the inertia weight which is
decreasing over time, typically from 0.9 to 0.4 as it has the effect of narrowing the search according to
the analysis.. So the process will keep running until the maximum iteration which is 20 in this case.

Table 3. Input parameter setting for PSO.

Population Max Upper Lower wmax wmin C1 C2
size iteration bound bound
40 20 10 3 0.9 0.4 0.7 0.7

7
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

Figure 2 shows a screenshot of PSO simulation (one trial sample) with Matlab.

Figure 2. Screenshot for PSO simulation.

5. Simulation results
Table 4 shows that increase in soil resistivity from 300 Ωm to 600 Ωm does not introduce a
remarkable change in tolerable step and touch voltage but it affects the grid resistance, Rg. When soil
resistivity increases from 300 Ωm to 600 Ωm, Rg increases from 2.297Ω to 4.410 Ω. Increase in Rg
will increase GPR, which in turn require more earthing rods. Hence it is desirable to keep soil
resistivity and GPR low.

Table 4. Safety limit with different soil resistivity.

Soil resistivity & Safety GSA PSO
crush rock Factor
resistivity (Ωm)
300, 3000 Estep70 (V) 5664 5664
Etouch70 (V) 1720 1720
Estep50 (V) 4186 4186
Etouch50 (V) 1271 1271
GPR (V) 6428.4 6154
Rg (Ω) 2.297 2.407
600, 3000 Estep70 (V) 5890.5 5890.5
Etouch70 (V) 1776.6 1776.6
Estep50 (V) 4353 4353
Etouch50 (V) 1313 1313
GPR (V) 11229 11941
Rg (Ω) 4.410 4.339

Table 5 shows that increase in crush rock resistivity from 3000 Ωm to 10000 Ωm increase the tolerable
step and touch voltage i.e. Estep70 increases from 5664 V to 17406 V and Etouch70 increases from
1720 V to 4656 V. Increase in tolerable step and touch voltage lower the safety requirement.

8
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

Table 5. Safety limit with different crush rock resistivity.

Soil resistivity & Safety GSA PSO
crush rock Factor
resistivity (Ωm)
300, 3000 Estep70 (V) 5664 5664
Etouch70 (V) 1720 1720
Estep50 (V) 4186 4186
Etouch50 (V) 1271 1271
GPR (V) 6428.4 6154
Rg (Ω) 2.279 2.407
300, 10000 Estep70 (V) 17406 17406
Etouch70 (V) 4656 4656
Estep50 (V) 12860 12860
Etouch50 (V) 3440 3440
GPR (V) 6768.7 6815.3
Rg (Ω) 2.324 2.660

Table 6 shows the GSA and PSO optimum parameters for earthing rods; number of horizontal
rows, number or horizontal columns, number of vertical rods and each vertical rod length. The table
shows that increase in soil resistivity from 300 Ωm to 600 Ωm increase the total length of rod
conductors i.e. (GSA 672m to 1425m, PSO 673m to 1440m). GSA and PSO produce optimum results
with negligible difference in total length of conductor (1.5%, 1.05%).

Table 6. Optimized parameters with different soil resistivity.

Soil resistivity & Optimized GSA PSO Difference
crush rock Parameter (%)
resistivity (Ωm)
300, 3000 Number of row 4 4 N/A
Number of 6 6 N/A
column
Number of rod 5 4 N/A
Rod length(m) 3 3 N/A
Total Length of 672 673 1.5
conductor
Grid resistance 2.297 2.407 5.2
Rg (Ω)
600, 3000 Number of row 10 10 N/A
Number of 10 9 N/A
column
Number of rod 5 10 N/A
Rod length(m) 5 9 N/A
Total Length of 1425 1440 1.05
conductor
Grid resistance 4.400 4.339 1.39
Rg (Ω)

Table 7 shows the GSA and PSO optimum parameters when crush rock resistivity increases from
3000 Ωm to 10000 Ωm. Increase in rock resistivity decreases the total length of conductors (GSA
672m to 447m, PSO 673m to 429m).

9
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

Table 7. Optimized parameters with different crush rock resistivity.

Soil resistivity & Optimized GSA PSO Difference
crush rock Parameter (%)
resistivity (Ωm)
300, 3000 Number of row 4 4 N/A
Number of 6 6 N/A
column
Number of rod 5 4 N/A
Rod length(m) 3 3 N/A
Total Length of 672 673 1.5
conductor
Grid resistance 2.297 2.407 5.2
Rg (Ω)
300, 10000 Number of row 3 3 N/A
Number of 3 3 N/A
column
Number of rod 9 3 N/A
Rod length(m) 3 3 N/A
Total Length of 447 429 4.02
conductor
Grid resistance 2.324 2.660 12.8
Rg (Ω)

Table 8 shows that total length of conductors required decrease from up-optimized 956m to GSA
optimum 672m, saving 29.7 % buried conductor length. This can be achieved at negligible increase
(1.0%) in grid resistance that results in negligible increase in GPR.

Table 8. Comparison between un-optimized grid and optimized grid by using GSA.
SOIL RESISTIVITY AND CRUSH ROCK RESISTIVITY (300 ΩM , 3000 ΩM)
Parameter Un-optimized Optimized with Difference(%)
grid GSA
Number of horizontal rows 6 4 N/A
Number of horizontal columns 7 6 N/A
No. of vertical rod 22 5 N/A
Vertical rod length (m) 3 3 N/A
Grid resistance Rg (Ω) 2.275 2.297 1.0
Total Length of conductor (m) 956 672 29.7

Table 9 shows that total length of conductors required decrease from up-optimized 956m to PSO
optimum 673m, saving 29.6 % buried conductor length. This can be achieved at small increase (6%)
in grid resistance that results in small increase in GPR.

Table 9. Comparison between un-optimized grid and optimized grid by using PSO.
SOIL RESISTIVITY AND CRUSH ROCK RESISTIVITY (300 ΩM , 3000 ΩM)
Parameter Un-optimized Optimized with Difference(%)
grid PSO
Number of horizontal rows 6 4 N/A
Number of horizontal columns 7 6 N/A
No. of vertical rod 22 4 N/A
Vertical rod length (m) 3 3 N/A
Grid resistance Rg (Ω) 2.275 2.407 6.0
Total Length of conductor (m) 956 673 29.6

6. Conclusion
Substation grounding grid design is an important part for substation design process. The design of the
grounding grid must be developed for safety and reliability. There are several steps to produce a safe
and effective grid design. Manual calculation will be a tedious and difficult process as it takes time and

10
CUTSE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 495 (2019) 012037 doi:10.1088/1757-899X/495/1/012037

the precision might be an issue. Hence, computer program is developed nowadays to greatly improve
the process of grounding grid design in terms of time saving, reliability and preciseness.
In this research an overview of substation grounding grid design based on IEEE Std. 80-2000 was
carried out. The grounding grid design equations, provided by the IEEE standard, were used to design
a typical substation grounding grid. The substation grounding grid was then designed using
optimization algorithms: GSA and PSO. For the optimization process, the necessary data including the
population size and iteration number, are required to key into the templates provided by GSA and PSO
respectively. These templates were further integrated with the preliminary grounding design code to
work out the most optimized grid parameters. The results showed that both GSA and PSO were
capable of optimizing the problem of grounding grid design and results in minimum total length of
conductor rods. Both GSA & PSO produced consistent results and were able to produce a more cost
effective grounding grid as compared to the equations provided by IEEE Std. 80-2000.

7. References
[1] IEEE Power Engineering Society Jan 2000 IEEE Guide for Safety in AC Substation Grounding
IEEE-SA Standards
[2] F. Dawalibi F and Mukhedkar D Apr 1975 Optimum design of substation grounding in two
layer earth structure analytical study IEEE Trans. Power Apparatus and Systems 94 no. 2
pp 252-261
[3] Heppe R J Dec 1979 Computation of potential at surface above an energized grid or other
electrode allowing for non-uniform current distribution IEEE Trans. Power Apparatus and
Systems 98 no. 6 pp 1978-89
[4] Ghoneim S, Hirsch H, Elmorshedy A and Amer R Optimum Jun. 24–28, 2007 Grounding grid
design by using an evolutionary algorithm Proc. IEEE Power Eng. Soc. Gen. Meeting pp 1–7
[5] Lee C and Shen Y 2009 Optimal planning of ground grid based on particle swam algorithm
Int. Journal of Electrical and Computer Engineering 3 no.12 pp 2235-42
[6] Nezhad N K, Fallahi M H and Dozein M G Jul 2013 An optimal design of substation grounding
grid considering economic aspects using particle swarm optimization Research. J. Appl. Sci.
Eng. Technol. 6 no. 12 pp 2159–65
[7] Tong C Mar 2014 Gravitational search algorithm based on simulated annealing
J. Convergence Information Technol. (JCIT) pp 231–7

11