See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/241628702

Fault indicator allocation in power distribution network for improving

reliability and fault section estimation

Article · October 2011

DOI: 10.1109/APAP.2011.6180595

CITATIONS READS

10 244

2 authors, including:

Rahman Dashti
Persian Gulf University
44 PUBLICATIONS   225 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A new method presentation for locating fault in power distribution networks View project

fault location in compensated transmission line View project

All content following this page was uploaded by Rahman Dashti on 22 June 2016.

The user has requested enhancement of the downloaded file.

 2011The International Conference on Advanced Power System Automation and Protection

Fault Indicator Allocation in Power Distribution Network for

Improving Reliability and Fault Section Estimation

Rahman Dashti , Javad Sadeh

Electrical Engineering Department, Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, Iran

Abstract: Fault indicator (FI) is used in impedance-based methods for improving fault section estimation in power distribution
feeders. The location and number of FIs affects on the reliability indices and can extra charge the distribution companies and
consumers. In this paper, the optimal location and number of fault indicator (FI) are determined in Power Distribution Network
(PDN) with special economical combined objective function. In this objective function, four imposed costs on distribution com-
panies and consumers are taken into account. For solving this problem, a powerful method i.e. Genetic Algorithm ( GA) is used.
Then the proposed method is tested on the two case studies, which one of them is a real distribution feeder of Bushehr power dis-
tribution company in Iran. At the end the results are presented discussed.

Keywords: Distribution system, Fault location, Fault Indicator

1 Introduction Therefore, this is favorable for consumers, if location

of fault is found quickly and repaired them, conse-
Most of the researches presented so far in the field of quently time of restoration is reduced.
fault location are focused on the fault location on the Different methods are presented for fault location
transmission lines; however, these algorithms are not in PDN. These methods have two stages. In the first
suitable for locating the fault on the distribution net- stage, the fault distance is determined and in the
works. Since distribution systems are usually unba-
second stage, the fault section is estimated. Fault lo-
lanced nature such that mixed single phase and three
cation methods are classified into two categories:
phase laterals, non-homogeneous feeders,
non-homogeneous phase conductor and unbalanced a) Impedance based methods [2,3]
loading. These features can create some problems in b)Traveling wave based methods [5]
fault location algorithm and also in fault section esti- In the impedance based methods fundamental of
mation, from which the multiple-estimation problem is voltage and current signals at the beginning of feeder
the most important [1,2,3]. are used. Implementation of these methods is very
Based on a presented report, about 80% of inter- simple are cheap, but they have some problems such
ruptions are caused by faults in distribution networks as sensitivity to fault resistance and multiple-solution
[4]. PDN is susceptible to faults caused by different [1, 4]. Traveling-wave methods consider the voltage
reasons such as: and current waves, traveling at the speed of light from
the fault towards the line terminals [5]. These me-
x These networks are distributed in different re-
thods are considered as very accurate, however, also
gions (urban, residential, and concentrated in
as complex and costly for application, as requiring
urban marginal areas) which are available for high sampling frequency [6, 7], Multi response and
people. Detecting travelling wave in faults near to the fault
x weather conditions location devices and in zero domain of voltage or
x equipment failure current are difficult.
These faults cause to decrease power quality, de- With this fault location review, the most methods
stroyed reliability indices, decrease benefits, and sa- need to estimate Faulted section. Fault section estima-
tisfaction of consumers from distribution companies. tion can classified to three methods, using current

The authors are with the Electrical Engineering Department, Faculty of Engi-
neering, Ferdowsi University of Mashhad, Mashhad, Iran (e-mail: rahman-
[email protected] and [email protected]). APAP2011 www.apap2011.org
___________________________________
978-1-4244-9621-1/11/$26.00 ©2011 IEEE


 2011The International Conference on Advanced Power System Automation and Protection

pattern recognition, intelligent methods, different me- sented objective function which is presented in dif-
thods which have communications devices. These ferent Refs on this topic. So in continue, both of them
methods have different problems such as: are combined and effect of optimal FI allocation is
x Many indices are presented but are not shown and compared with different methods which
unique. are presented till now. Therefore, at the end, the pre-
x Some of methods need to communication sented method is tested on feeder of Bandargah of
device and addition equipments. Conse- Bushehr PDN in Iran.
quently, they are expensive and complex. The rest of this paper is organized as followsin
Section 2 the proposed method is described and in
x The most of methods need to data bank,
Section 3 the multi objective optimization is described
which make of it and update of it is very dif- Simulation results of proposed method in different
ficult. conditions are presented in Section 4. At the end con-
x … clusions are drawn.
With these notes, choice of method for locating
fault depends on both system topology and available 2 Proposed Method
instrumentation for system monitoring.
Fault Indicators (FI) allocation can improve im- The security and reliability and service continuity of
pedance based methods. Because of, it can speed up power distribution system is very important, but it is
finding correct position of fault section estimation or violating with occurring faults. Fast clearing and iso-
reduce multi section estimation, consequently restora- lation of different faults types are critical in maintain-
tion time is reduced [8, 9]. Important questions are ing a reliable power system operation and improve
mentioned in below for FI:
service continuity indexes.
x How many FIs does each feeder need?
The proposed method presents a new optimal FI allo-
x Which place do the designers prefer to install
cation method in PDN. In this method, anew com-
it?
bined economic objective function is assumed which
At the first FI is presented in [2] but this paper did must be optimized. Suggestion objective function is
not present any special method for finding number
combination of three main part of benefit and disad-
and position of FIs. In continue different methods are
vantage of finance such that show correct behavior of
presented for solving it such as GA (Genetic Algo-
rithm), CBGA (Chu-Beasley Genetic Algorithm) and mutual effect between consumer and distribution
IM (Immune Algorithm). GA method, which is pre- companies. This objective function is composed four
sented in [10], is used for solving the OFP problem. It cases as follows:
finds the optimal place of FIs. Ref [11] analyzed the x Energy not Supply (ENS) cost
effect of FIs on reliability indices of PDN and the x Operation cost and restoration cost
presented method in [11] is tested on real Iranian PDN. x Unsatisfied consumers cost
CBGA is used for optimal FI location in PDN [8].
x FI cost (buying & installing)
This method extends the presented method in [12] and
2-1 Energy Not Supply (ENS) cost
presents an objective function which is combined of
the number of suspected locations and the distance In each interruption, restoration time de-
among suspected locations. Also it uses the FI statuses pends to type of fault and its location. This cause
for helping to find faulted section. IM is used for FIs to be created addition cost which is forced on
optimal placement which is presented in [9]. It is pre- Power distribution companies. It is Energy not
sented a combined objective functions. The presented supply which is introduced by (1):
objective functions are combination cost of different
types of loads (residential, commercial, industrial and
‫ ͳܥ‬ൌ ܿ௜ σ௡௜ୀଵ ܲ௜ ‫ݐ‬௜
key customers). (1)
This paper presents optimal fault indictors alloca-
ܲ௜ : Amount of load ith
tion for fixed number and variable number of them in
‫ݐ‬௜ : Interruption time of load ith
PDN with multi objective genetic algorithm. In this
݊: Number of section
paper anew combined objective function is presented.
ܿ௜ : cost of each kWh
These objective function is formed to four necessary
function. In each objective function element, cost of 2-2 Operation cost and restoration cost
different level and types of loads has been assumed. Operation cost and restoration cost are divided to
The special feature of this presented objective func- two parts as follows:
tion is assuming forced cost on consumers and distri-
bution companies besides of assuming previous pre-


 2011The International Conference on Advanced Power System Automation and Protection

x The cost of engineer, employer and worker the solutions may be "seeded" in areas where optimal
who work to restore the feeder. solutions are likely to be found.
x Needed equipments for restoration. In this stage network data is inputted. Now n bits
These parts are introduced by (2): (0, 1) have assumed for allocating of FIs with atten-
tion to fitness function:
‫ ʹܥ‬ൌ σ௡௜ୀଵ ߣ௜ ݈௜ ݉௜ ܿ‫݋‬௜ ‫ݐ‬௜ ൅ σ௡௜ୀଵ ݉௜ ݇௜ ‫ݐ‬௜ ൅ ‫ܨ‬ (2) 0 0 0
1 1
ߣ௜ : Interruption rate per year.
݈௜ : Length of segment ith.
݉௜ : Employer number. n bits
ܿ‫݋‬௜ : Hourly cost of each worker. n: number of segments of PDN
‫ݐ‬௜ : Restoration time of segment ith. Figure 1. Genes of FI allocations
݇௜ : Hourly cost of operation car.
‫ܨ‬: cost of needed equipment for restoration. Genetic generate randomize genes. The network
݊: Number of segment. calculation is doing with assumed these genes. If the
value of bit ith is 1, this means segment ith has FI else
2-3 Unsatisfied consumers cost this segment does not have and it do not came to net-
When any interruption is occurred for each con- work calculation.
sumer, cause to they will be worry. Consequently the
electrical distribution company must be compensating 3-2- Fitness function
effect of this interruption for each consumer. It has a BGAs are essentially un constrained search proce-
standard fine cost for each hour for each consumer. dures within a given represented space where the
Unsatisfied cost is introduced in (3): search is guided according to a fitness function. In
this paper the fitness functions and the penalty func-
‫ ͵ܥ‬ൌ σ௠ ௝ୀଵ ߣ௜ ݈௜ ‫݌‬௜ ݂ܿ௜ ‫ݐ‬௜ (3) tion is:
ߣ௜ : Interruption rate
݈௜ : Length of segment ith
‫݌‬௜ : Active power consumption of each consumer ‫ ࢆ ܖܑܕ‬ൌ ࢝૚ ࡯૚ ൅ ࢝૛ ࡯૛ ൅ ࢝૜ ࡯૜ ൅ ࢝૝ ࡯૝ (6)
݂ܿ௜ : Hourly fine cost of each consumer
‫ݐ‬௜ : Restoration time of segment ith
݉: Consumer number That C1, C2, C3 and C4 are defined in section 2-1.
݊: Number of section The strings are stored according to their fitness
which is then ranked accordingly. The roulette wheel
2-4 FI cost (buying & installing) selection scheme is used for selecting the individuals
GA gives number and place of FIs in PDN as out- for reproduction.
put. Number of FIs has some cost such as purchasing
cost, installing cost and its effect on operation and 3-3 Genetic operators
repaired group. It is defined by (4): Genetic operators are the stochastic transition rules
applied to each chromosome during each generation
‫ܥ‬ସ ൌ ‫ ݐݏ݋ܿܫܨ‬ൌ procedure to generate a new improved population
ܾ‫ ݏܫܨ݂݋ݐݏ݋݈ܿ݃݊݅ܽݐݏ݊݅݀݊ܽ݃݊݅ݕݑ‬൅ (4)
from an old one. A genetic algorithm usually consists
‫ݐݏ݋ܿ݌ݑ݋ݎ݃݀݁ݎ݅ܽ݌݁ݎ‬ of reproduction, crossover and mutation operators.
• Reproduction
Now, these four objective functions must be opti- Reproduction is a probabilistic process for select-
mized by GA.
ing two parent strings from the population of strings
on the basis of “roulette-wheel” mechanism, using
their fitness values. This ensures that the expected
3- Genetic Algorithm (GA)
number of times a string is selected is proportional to
its fitness relative to the rest of the population. There-
3-1- Initial population
fore, strings with higher fitness values have a higher
Initially many individual solutions are randomly
probability of contributing offspring.
generated to form an initial population. The popula-
• Crossover
tion size depends on the nature of the problem, but
Crossover is the process of selecting a random po-
typically contains several hundreds or thousands of
sition in the string and swapping the characters either
possible solutions. Traditionally, the population is
left or right of this point with another similarly parti-
generated randomly, covering the entire range of
tioned string. This random position is called the cros-
possible solutions (the search space). Occasionally,
sover point. In this paper the characters to the right of


 2011The International Conference on Advanced Power System Automation and Protection

a crossover point are swapped. The probability of Population Size =20

parent-chromosomes crossover is assumed to be 0.7. Maximum number of generation =100
• Mutation Probability of mutation = 0.05
Mutation is the process of random modification of Probability of crossover = 0.7
a string position by changing “0” to “1” or vice versa,
with a small probability. It prevents complete loss of 4-1 Test system 1: IEEE 13-BUS system
genetic material through reproduction and crossover Single line of IEEE 13-bus is shown in Figure 3.
by ensuring that the probability of searching any re- Nine bits are selected for population length which is
gion in the problem space is never zero. In this paper candidate for FI position. With applying the proposed
the probability of mutation is assumed to be between method, the optimum number FI is two and the FI
0.01 and 0.1. position is selected 671-684, 632-645. In this task FI
location is done which its result is shown in Table1.
3-4 Region division
When Location of FIs is determined by GA then
different regions should be defined. Region is a sec-
tion of distribution feeder which is one of following
conditions:
a- Section between two FIs
b- Section between one FI and end buses
c- Section between the beginning of feeder and
the first FI
Figure 2 shows the different defined regions. In Figure 3. IEEE 13-bus test feeder
this figure three regions are defined R1 till R3.

R1
Table 1: Result of FI allocating in test feeder 13- bus IEEE
Optimal place
R2 1 671-684
2 671-684 & 632-645
Optimal number 2

R3
4-2 Test system 2: Bandargah feeder
fault indicator Bandargah feeder is one feeder of Bushehr distri-
Figure 2. Different defined regions
bution network in Iran. Its single line diagram is
With this FIs and Defined Regions restoration time shown in Figure 4. Length of this feeder is 25 km
of each region is decreased and is calculated by (7): with three laterals and five load taps. The additional
௟ information about this feeder is shown in Table 3 in
ܶ௜ ൌ ܶ଴௜ ሺσ೙ ೔ ሻ (7)
ೕసభ ௟ೕ Appendix. The result of optimal allocation with li-
Which: mited and unlimited number of FIs is shown in Table
݈௜ : Length of region ith 2.
ܶ௜ : Restoration time of region ith after FI allocation
ܶ଴௜ : Restoration time of region ith before FI alloca-
tion

4 Application and simulation results of

GA
Figure 4. Bandargah feeder
Proposed method is tested on two distribution
feeders which one of them is 13-bus IEEE test system
and other one is a real feeder in Bushehr distribution Table 2. Optimal location and number of FI
network which is named Bandargah feeder. The fol- Optimal place
lowing parameters are used in the present research 1 3-11
are: 2 2-9 & 3-11


 2011The International Conference on Advanced Power System Automation and Protection

3 2-9 & 3-11 & 5-6

Optimal number 3 FIs in 2-9 & 3-11& 5-6

Reliability indices (SAIFI, SAIDI, ENS) of this

test feeder is shown in Figure 5 (a,b,c). ). Figure 5-a
shows, FI does not have any influence on SAIFI, be-
cause of FI only affect on repair or restoration time
and fault location time consequently it affects on ENS
and SAIDI as shown in Figures 5-b and 5-c.

c
5 Conclusion
Figure 5. Reliability indices (SAIFI, SAIDI, ENS) of Bandargah
With attention to problem of fault section estimation feeder
in PDS and FI application for solving it, a new objec-
tive function is presented in this paper. The proposed 1 R. Dashti, and J. Sadeh, “A New Method for Fault
objective function is comprehensive because it is con- Section Estimation in Distribution Network”, Interna-
tional Conference on Power System Technology
sidered the four fundament necessary economical (POWERCON2010), 2010.
components (FI cost, unsatisfied consumer cost, ENS 2 R. Das, "Determining the Locations of Faults in Dis-
and operation and restoration cost). Solving this ob- tribution Systems", Ph.D. Thesis, College of Graduate
jective using GA, the optimum number and location Studies and Research, University of Saskatchewan,
of FIs are found. The results show this task can im- Saskatchewan, 1998.
3 S.-J. Lee, et al., “An intelligent and efficient fault loca-
prove the reliability indices and increase benefits. The
tion and diagnosis scheme for radial distribution sys-
proposed method is tasted on two test cases, on of tems,” IEEE Trans. Power Del., vol. 19, no. 2, pp.
them is a real and practical feeder of Iranian PDN. 524–532, Apr. 2004.
4 G. Morales-España, et al., “Elimination of Multiple
Estimation for Fault Location in Radial Power Systems
by Using Fundamental Single-End Measurements”
IEEE Trans. Power Del, vo;. 24, no. 3, July 2009.
5 Z.Q. Bo, G. Weller and M.A. Redfern, “Accurate fault
location technique for distribution system using
fault-generated high-frequency transient voltage sig-
nals” IEE Proc.-Gener. Transm. Distrib., vol. 146, no. 1,
pp. 73-79, Jan. 1999.
6 F.H. Magnago and A. Abur "Advanced techniques for
transmission and distribution system fault location"
Proc of CIGRE – Study Committee 34 Colloquium and
a Meeting, Florence,paper 215.
7 Y. Tang, et al., "Fault indicators in transmission and
distribution systems", Proc. of Int. Conf. on Electric
Utility Deregulation and Restructuring and Power
Technologies – DRPT, pp. 238–243, 2000.
8 D. Almeida, et al., “Optimal placement of faulted circuit indi-
cators in power distribution systems”, Elsevier, Electric Pow-
er Systems Research, vol.81, no.2, pp.669- 706, 2011.
9 H. Chin-Ying, et al., "Optimal Placement of Fault In-
dicators Using the Immune Algorithm”, IEEE Trans.
Power Syst., vol. 26, no. 1, pp. 38-45, 2011.
10 D.P. Cong, et al, “Optimization of fault indicators
placement with dispersed generation insertion,” Proc.
b 2005 IEEE Power Eng. Soc. General Meeting,
pp.355–362.
11 H. Falaghi, et al, “Fault indicators effects on distribu-
tion reliability indices,” Proc. 18th Int.Conf. Electricity
Distribution, 2005.
12 F.J.R. Santana, L.A.L. de Almeida and F.F. Costa, "Op-
timal position of geo-referenced short circuit sensors
for faster fault finding using genetic algorithm", Pro-
ceedings of the IEEE International Symposium on In-
dustrial Electronics, Cambridge, pp. 255–260,2008.


 2011The International Conference on Advanced Power System Automation and Protection

13 K. Deb, "Multi-Objective Optimization using Evolu- tier in Multiobjective Optimization", Journal of Engi-
tionary Algorithms", John Wiley & Sons, 2001. neering Optimization, no. 12, pp. 1–18, 2010.
14 A. Messac and C. A. Mattson, "Normal constraint me- 16 Deb, K. Multi-Objective Optimization using Evolutio-
thod with guarantee of even representation of complete nary Algorithms John Wiley & Sons, 2001.
Pareto frontier", AIAA journal, vol. 42, no10, pp. 17 D. Craft, et al., "Approximating convex Pareto surfaces
2101–2111, 2004. in multi objective radiotherapy planning". Medical
15 T. Erfani and S.V. Utyuzhnikov, "Directed Search Do- Physics, vol. 33, no. 9, pp. 3399–3407, 2006.
main: A Method for Even Generation of Pareto Fron-

Appendix:
Bandargah feeder information
Table 3. Bandargah Feeder information
Active
Sending Receiving Reactive
power Length(km) Z
Bus Bus power (kVAr)
(kW)
1 2 ---- ---- 18 5.004+j6.408
2 3 ---- ----- .5 0.139+j0.178
2 9 220 130 .7 04726+j0.6052
9 10 150 95 .3 0.17514+j0.22428
3 4 180 130 .7 0.1946+j0.2492
3 11 40 25 .6 0.5838+j0.7476
11 12 110 65 .5 0.1251+j0.1602
4 5 ---- ---- .65 0.1807+j0.2314
5 6 50 30 2.5 0.695+j0.890
5 13 50 30 .5 0.0695+j0.0890
6 7 150 80 .35 0.0973+j0.1246
7 8 110 75 .7 0.1946+j0.2492



View publication stats