Electrical Power and Energy Systems 43 (2012) 607616

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Reliability evaluation and enhancement of distribution systems in the presence

of distributed generation based on standby mode
Rajesh Arya a,, S.C. Choube b, L.D. Arya c
a

Electrical Engg. Deptt., Medi-Caps Institute of Science & Technology, Indore, M.P., India
Electrical & Electronics Engg. Deptt., UIT, Rajiv Gandhi Technological University, Bhopal, M.P., India
c
Electrical Engg. Deptt., Shri G.S. Institute of Technology & Science, Indore, M.P., India
b

a r t i c l e

i n f o

Article history:
Received 13 April 2010
Received in revised form 8 April 2012
Accepted 24 May 2012
Available online 7 July 2012
Keywords:
Distribution system
Reliability indices
Failure rate
Repair time
Differential evolution
PSO

a b s t r a c t
This paper describes an analytical methodology for reliability evaluation and enhancement of distribution system having distributed generation (DG). Standby mode of operation of DG has been considered
for this purpose. Reliability of the composite distribution system has been optimized with respect to failure rate and repair time of each section of the distribution system subject to constraints on (i) failure
rates/ repair times and (ii) customer and energy based reliability indices i.e. SAFI, SAIDI, CAIDI and AFNS.
The objective function includes (i) cost of modication for failure rates/repair times (ii) additional cost of
expected energy supplied by DG. Differential evolution (DE), Particle swarm optimization (PSO) and coordinated aggregation based PSO (CAPSO) have been used to develop computational algorithms. Developed algorithms have been implemented on a sample distribution system.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Distribution systems were not originally designed to accommodate generation in itself. Hence increasing penetration level of DG
is causing changes in the planning, operation and maintenance of
distribution systems. Presence of DG in a distribution network affects network planning, operation and maintenance, auxiliary services, quality of service and regulatory aspects [14]. Ackerman et
al. [5] have analyzed the term DG in detail. Usually in reference to
this present paper, DG will be considered as electricity generation
systems connected to distribution networks, characterized by their
small rating and located near connection points. Distribution system operators (DSOs) may have their own DG or may encourage
large consumer to install DG which is owned and controlled by
these customers. Availability of such customer owned and controlled DG may not be high, but may be of great help in improving
the reliability indices of the distribution network.
The reliability of a distribution system may be increased by
modifying failure rate and repair time of each section of the network. Such modications may require additional investments
which in the presence of DG may be mitigated. This will result in
annual savings. On the other hand cost per unit energy obtained

Corresponding author.
E-mail address: aryarajesh@yahoo.com (R. Arya).
0142-0615/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2012.05.045

form DG may be high. Further the traditional reliability indices

covered sustained interruption durations. The time necessary to
start up the DG should be taken in to account for the reliability
evaluation of distribution system. If this time is sufciently short
the customers suffer a momentary interruption, while, if not, they
suffer a sustained interruption.
Various reliability optimization algorithms have been developed by reducing failure rate and repair time of each section of
the distribution systems. Sallam et al. [6] presented a methodology
for determination of optimal reliability indices for distribution system using gradient projection method. Chang and Wu [7] developed a methodology for obtaining optimal reliability design of
electrical distribution system using a primal dual interior point
algorithm. Chowdhury and Custer [8] developed a value based approach for designing urban distribution system. Bhowmik et al. [9]
described a distribution network planning strategy by considering
a nonlinear objective function with linear and non linear constraints for radial distribution system. Su and Lii [10] obtained
optimum failure rates and repair times for each section of distribution system using modied genetic algorithms. Particle swarm
optimization along with decomposition has been used for reliability optimization of radial distribution systems by Arya et al. [11].
Bae and Kim [12] developed methodology for evaluating reliability
of distribution system having DG in operation mode. Bae et al. [13]
developed an optimal operating strategy for distributed generation
considering hourly reliability worth. Louit et al. [14] presented a

608

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

simple technique for obtaining optimal interval for major preventive

maintenance for distribution system without DG. Costa and Matos
[15] described various modeling aspects for assessing the effect of
microgrids to the reliability of distribution network. Trebolle et al.
[1] proposed a market mechanism, referred to as reliability options
for distributed generation (RODG), which provides distribution
system operators with an alternative to the investment in new distribution facilities.
In view of the above discussion this paper presents a new methodology for obtaining optimal failure rate and repair time of each
section of a distribution system in the presence of distributed generation which is specically customer owned and controlled. It is
assumed that utility has to pay additional price for each unit
charged. An objective function which reects cost of modication
of failure rates, and repair rate and cost of energy supplied from
DG has been constructed and minimized subject to reliability
constraints.

s, rs
sw, s

Load Point

dg, rdg
Fig. 1b. Reliability network of the schematic diagram of Fig. 1a.

eq, req
Fig. 1c. A single component equivalent of reliability network of Fig. 1b.

LP - iA

2. Reliability modeling aspects with distributed generation

Distributed generation which is owned and controlled by customer or any other agency may be used as standby units for reliability enhancement at specic load points. This application is
similar to one used in transferring load from one substation to another by disconnecting feeders at their normal source (in the event
of failure of a section) end and closing the open point. Usually this
is done in the event of a failure causing loss of supply from the
source substation and it is not normal practice to parallel sources.
Now in the changed scenario of deregulated environment, the
transfer of load at a load point may be partial or total according
to the capacity available of DG. If DG capacity is sufcient to meet
the load at a load point in the event of failure of supply from substation then the situation at load point is represent as shown in
Fig. 1a.
Further it is to be understood that load is usually connected
from source via various distributor segments. By manual switching
the load may be shifted to DG in the event of trouble from source
side. The average time to carry out the switching operation is assessed as s hours. The equivalent reliability diagram for the system is shown in Fig. 1b in which source side, and DG are directly
connected in parallel but placed in series with important switching
element with failure frequency ksw and restoration time s. Reliability network of Fig. 1b may reduced by the application of seriesparallel approximate formulas [16]. The system of Fig. 1b is
reduced to an equivalent network (keq, req) shown in Fig. 1c using
following relations

keq ks kdg r s r dg ksw

r eq

ks  kdg  r s  rdg ksw  s

ks  kdg r s r dg ksw

where keq is the equivalent failure rate; req the equivalent interruption duration; ks the represents total failure rate up to load point
from substation supply; rs the represents average interruption duration from substation supply; kdg the failure rate of DG; rdg the average outage duration of DG; and ksw the failure rate of manual
switch; and s is the service restoration time from DG.

s, rs

dg, rdg

Main Source

DG
Load Point (LP)

Fig. 1a. Schematic diagram of a load point fed from main source and as stand by
supply from DG.

Distribution System

LP - iB

DG

Fig. 2a. Reliability modeling for ith load point having limited DG capacity available.

Service interruption rate and average interruption/restoration

time are calculated for each load point by successive applications
of series parallel reduction rules. This type of procedure is commonly employed in distribution systems reliability studies.
A different reliability modeling will be required if the DG capacity available is insufcient to fulll the maximum demand of all
the customers connected at a load point. In such situation the load
point is partitioned in two portions one of it is supplied from main
source as well as from DG as stand by and the other part is supplied
from main source only. This reliability modeling aspect is shown in
(see Fig. 2a). Partitioned load points are i-A and i-B. In this case
alternative supply is available for a fraction of customers connected at the load point. Figs. 2b and 2c shows reliability network
for evaluating the indices at load point LP  iA and LP  iB. Relations (1) and (2) along with series parallel reliability network
reduction formulas are used to evaluate the basic indices at LP  iA
or LP  iB.
3. Problem formulation
Reliability indices may be improved for a radial distribution system by modifying failure rate and average repair time of each segment of the system. Addition of distribution generation (DG) may
require additional cost per kWh to be paid to the private DG owners. In view of this the objective function is selected as follows

NC
X
ak
2
k1 kk

NC
X
bK
K1

rK

ADCOSTEENSO  EENSD

In above kK is the failure rate of Kth section; rK the average repair

time of kth section; aK and bK are the cost coefcients; EENSO the

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

s, rs

Source

Now to calculate USYS,i accounting DG, the equivalent failure

rate keq and equivalent repair time req are used as given by relations
(1) and (2) in Section 2. Hence USYS,i accounting DG is evaluated as
follows

LP - iA

Fig. 2b. Reliability network for evaluating indices at LP  iA.

U SYS;i keq  req

s, rs
sw, s

LP - iB

dg, rdg
Fig. 2c. Reliability network for evaluating indices at LP  iB.

expected energy not supplied without DG; EENSD the expected energy not supplied with DG and ADCOST is the additional cost per
kWh to be paid to DG owners.
The objective function constitutes of three parts i.e. (i) rst part
reects cost of modications of failure rates of each section. It includes cost of preventive measures for reducing failure rate. The
quadratic form is based on the reliability growth model of Duane
[17], (ii) second part gives cost of modications in average repair
time of each section which is again based on a growth model of repair time. Larger is the repair time lesser is the cost of repair and
vice versa, and (iii) third part represents additional cost to be paid
to the owners of DG. This term in fact is the expected energy supplied (EENSOEENSD) by DG multiplied by additional charge per
kW h (ADCOST). Thus the objective function provides a balance between additional cost incurred by the provision of DG and cost due
to modications in failure rates and repair times. It is to be noted
that the operation of DG has been considered in standby mode
and DG has been assumed to be privately owned. Hence power obtained from DG is costlier than that obtained from distribution
substation. Other side one has to pay price for modications in failure rates and repair times to achieve desired values of reliability
indices as given by relations (7)(10). Hence a compromise has
to be achieved between failure rate and repair time modications
and additive cost of energy obtained from DG sources. This, in fact,
is achieved by minimizing the objective function given by relation
(3). This framed objective function includes the cost of modications and additional cost of energy purchased from DSO. Further
note that as explained in Section 2, power cannot be taped immediately from DG, but with a time delay of s (service restoration
time from DG) hours and this has been accommodated in modeling
aspects. Another aspect which has been included that the demand
may be fullled even partially in emergency conditions. This has
been included in modeling aspects.
Expected energy not supplied (EENS) is calculated for all load
points as follows [18]

X
EENS
Li U SYS;i

U SYS;i

4a

k2S

S denotes set of distribution segment in series upto kth load

point; kk is failure rate of kth segment and rk is average repair time
for kth segment

4b

Whereas relation (4a) is used to evaluate USYS,i and in turn used to

evaluate EENSO (expected energy not supplied without DG).
Objective function (3) is minimized subject to following customer and energy based constraints [18]
(i) Bounds on the modication of failure rate and repair time of
each section

kk;min 6 kk 6 kk;max

r k;min 6 r k 6 r k;max

(ii) Inequality constraint on system average interruption frequency index (SAIFI)

SAIFI 6 SAIFId

System average frequency index (SAIFI) is dened as follows

X
SAIFI

kSYS;i Ni
X
Ni

(iii) Inequality constraint on system average interruption duration index (SAIDI)

SAIDI 6 SAIDId

SAIDI is dened as

X
SAIDI

U N
i SYS;i i
P
Ni

10

(iv) Inequality constraint on customer average interruption

duration index (CAIDI)

CAIDI 6 CAIDId

11

CAIDI is evaluated as follows

U SYS;i  Ni
CAIDI Xi
k N
i SYS;i i

12

(v) Constraint on average energy not supplied (AENS) per customer also known as average system curtailment index
(ASCI)

AENS 6 AENSd

13

Where AENS is given as follows

AENS

where Li is average load at ith load point; USYS,i is average annual

outage duration at ith load point.
EENS is calculated using relation (4) without DG is termed as
EENSO and that calculated with DG is termed as EENSD. Expected
energy not supplied for all load points is calculated using relation
(4). In this relation Li is load at ith load point and USYS,i is average
annual outage duration at ith load point and is calculated using following relation for radial systems.

kk r k

609

P
LU
Pi SYS;i
Ni

14

In all above Ni is the number of customers at ith load point; Li average load at ith load point; kSYS;i is average failure rate at ith load
point; rSYS,i is average interruption duration at ith load point; USYS,i
is average annual outage duration given as product of kSYS;i and
rSYS,i.
SAIFId, CAIDId, SAIDId and AENSd are threshold values of the SAIFI,
CAIDI, SAIDI and AENS respectively. In Eqs. (5) and (6) the upper
limits on failure rate (k) and repair time (r) are considered as current (present) values as provided by any system data. Lower limits
on failure rate (kk,min) and repair time (rk,min) are the values which
can be achieved in practice. In any case these values cannot be
made zero. These lower limits are obtained in practice using failure

610

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

and repair data analysis along with the associated costs. The procedure is similar to that of reliability growth model [17]. The threshold values of reliability indices i.e. SAIFId, SAIDId, CAIDId, and AENSd
are based on managerial/administrative decisions.
The formulated reliability optimization problem has been
solved using robust population based optimization algorithms
known as differential evolution (DE), particle swarm optimization
(PSO) and one of its variant known as co-ordinated aggregation
based particle swarm optimization (CAPSO). The formulated problem has been solved using these three optimization techniques so
as to get condence in the optimization results and applicability of
these methods for solving such problem may be veried and comparison of the result may be made. The main idea of the paper is to
obtain the optimum values of failure rates and repair times of each
distribution segment subject to satisfaction of threshold values of
reliability indices (SAIFI, SAIDI, CAIDI, and AENS) and incorporating
the effects of distributed generation available in standby mode.
The optimized values may be allocated as target values to preventive maintenance and corrective repair crew. Computational algorithms have been explained in next two sections.

k
Step-7. Select two random vectors X k
p and X q from the population, where p q i best.
Step-8. Generate a mutated vector using following relation

Vi

K
X best rX K
p  Xq
K

r is scale factor usually lies in the range [0, 1] and V i is mutant

vector.
K
Step-9. If any component of mutant vector i.e. V i violates the
bounds on decision variables then bounce back mechanism [20]
is applied. The bounce back method replaces violated element
by the new element whose value lies between the base parameters (selected as Xbest) value and bounds being violated as
follows

k
ij

X 0i k0i ; r0i ;

i 1; . . . ; M

k0i k0i1 ; k0i2 ; . . . ; k0i;NC 

r0i r0i1 ; r 0i2 ; . . . ; r 0i;NC 
Step-2. Evaluate r SYS;i ; kSYS;i and USYS,i at each load point with DG
and without DG using seriesparallel reduction formulas as
explained in Section 2.
Step-3. Calculate reliability indices SAIFI, SAIDI, CAIDI, AENS,
EENSO and EENSD as dened in the relations 8,10,12,14, and
4 respectively.
Step-4. Evaluate objective functions JX 0i for all vectors in the
population as given by relation (3).
Step-5. Evaluate inequality constraints (7),(9),(11) and (13) for
each vector of population. If a vector satises these constraints
call it F i.e feasible, otherwise ag it as NF i.e. not feasible.
Identify a vector X 0best i.e a vector feasible and is best from objective function view point.
Step-6. Select a target vector, i = 1.

if

: xbest;j randxj;max  xbest;j ;

if

v ijK 6 xj;min
v ijK P xj;max

rand is random digit between [0, 1].

k
Step-10. Trial vector t i is obtained using uniform cross over as
follows
k

Step-1. Initialization: Generate a population of size M. Each

vector consists of failure rates and repair times of each section
of distribution system. The values of each component of vector
are obtained by sampling uniformly between lower and upper
limits as given by relations (5) and (6).

8
< xbest;j randxj;min  xbest;j ;

16

4. Solution methodology using differential evolution (DE)

Storn and Price [19] developed a very simple population based
stochastic function optimizer known as deferential evolution (DE)
technique. DE has been applied to solve various natures of engineering problems [20]. DE solves the optimization problem by
sampling the objective function at multiple randomly chosen initial points. Bounds on decision variables dene the region from
which M vectors in the initial population are chosen. DE generates
new solution points in D dimensional space that are perturbations
of existing points. DE perturbs vectors with the scaled difference of
two randomly selected population vectors. Further DE adds the
scaled random vector difference to a third selected population vector (termed as base vector) to produce a mutated vector. A uniform
crossover is employed to produce a trial vector from target vector
and mutated vector. The DE based algorithm to solve the formulated problem is implemented in following steps.

15

tij

8
< v k ;
ij

if

randj 6 C r ;
k
xij ;

or

otherwise

j jrand

17

Cr is a crossover probability control parameter selected in the

range [0, 1], Cr is user dened value which controls the number
of decision variables which are copied from the mutant. randj
is random digit in the range [0, 1].
K
Step-11. performs reliability evaluation for trail vector t i and
obtain SAIFI, SAIDI, CAIDI, and AENS.
Step-12. If all the inequality constraints i.e. 7,9,11, and 13 are
satised than call the trial vector as F otherwise ag it as NF.
Step-13. Obtain EENSO and EENSD as given in relation (4).
k
Step-14. Evaluate objective function (3) for trail vector i.e. Jt i
Lampinens constraint handling technique is used to select or
k
reject trial vector in new population [21]. Trail vector t i is
selected in new population under following conditions.
K
K
K
(i) ti is feasible and Jti 6 JX i .
K
K
(ii) ti is F and X i is NF.
K
K
K
(iii) ti and X i both NF, but t i does not violate any conK
K
straint more than X i . Otherwise X i is retained in
new population.
Step-16. Select another target vector i = i + 1.
Step-17. If i 6 M, repeat from Step-7.
Step-18. Increase generation count, k = k + 1.
Step-19. If k 6 kmax, repeat from step-6, otherwise stop. Where
kmax is maximum number of generations specied.
It is stressed that DE algorithm may even be stopped before execution of maximum number of generation, if improvement in
objective function is not observed in a pre-specied generations.
5. Solution methodology using particle swarm optimization
Particle swarm optimization was originally proposed by Kennedy and Eberhart [22]. The methodology is population based
and applied to solve the formulated problem. The implementation
of PSO for solving the reliability optimization problem is in following steps.
Step-1. Population of size M is generated where each particle
contains failure rates and repair times of each section sampled
uniformly within bounds governed by relation (5) and (6).
Selected particles are assumed to be feasible i.e. these satisfy

611

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

inequality constraints 7,9,11, and 13. The generated particles

are represented as
0

Xi

ki1 ; ki2 ; . . . ; ki;NC ; r i1 ; r i2 . . . ; r iNC ;

i 1; . . . ; M

Step-2. This step consist of initialization of velocities of each

particle selected in step-1. The velocity of each element j of
particle i is generated at random within the boundary given
as follows

kj;min  k0ij 6 q0ij 6 kj;max  koij

18

r j;min  r 0ij 6 q0ij 6 rj;max  roij

Gbest at this current iteration is set as the best value in terms of

k
objective function among all Pbest;i .
Step-11. Increase iteration count k = k + 1
If k 6 kmax repeat from Step-5, otherwise stop. PSO is a heuristic
search procedure and as such no analytical convergence criterion
exists, except that PSO is terminated when a pre-specied maximum number of iterations are executed. The maximum number
of iteration may be specied based on computational experience
[23]. Further the convergence may be observed by plotting a graph
K
between JGbest and number of iterations.

Step-3. The initial best position, Pbest;i , of particle i is set as its

0
initial position and initial best position of group, Gbest , is determined as the position of an individual with minimum value of
objective function as obtained using relation (3).
Step-4. Set iteration count k = 1.
k
Step-5. Velocity of each individual qi is modied using following relation
k1
qik w  qik1 c1  rand1 P k1

best;i  X i
k1

k1

c2  rand2 Gbest  X i
k1
P best;i

19
th

k1
Gbest

where
is the best previous position i particle;
is the
global best position among all the participation in the swarm from
objective function view point; rand1, rand2 are random digits in the
range [0, 1]; w is inertia weight and c1, c2 are acceleration constants
A large inertia weight (w) facilitates global exploration and a smaller inertia weight tends to facilitate local exploration to ne tune
the current search area. Therefore the inertia weight w is an
important parameter for convergence behavior of PSO. A suitable
value of inertia weight usually provides balance between global
and local exploration abilities and consequently results in a better
optimum solution. In view of this it has been suggested that inertia
weight w is varied as follows [23]

wmax  wmin
w wmax 
k
kmax
where kmax is the maximum number of iteration specied and k denotes current iteration. wmax and wmin denotes maximum and minimum values of inertia weight.
K1
Step-6. position of each particle, X i
, is updated using following relation
k

k1

Xi Xi

qi

20

Step-7. If any component of updated particles violates the

bounds given by 5,6 then it is set to limiting value.
Step-8. Perform reliability evaluation for each updated particle
and evaluate SAIFI, SAIDI, CAIDI, AENS, EENSO and EENSD.
Step-9. Evaluate for each updated particle inequality constraints 7,9,11, and 13. If all these constraints are satised
the updated particle is retained in updated population. If
any of the constraints is not satised a y-back mechanism
[24] is adopted for handling the inequality constraints. In this
y back mechanism the idea is to maintain a feasible population by ying back such updated particle to its previous position. Since the population is initialized in the feasible region
ying back to its previous position will guarantee the solution
to be feasible. Flying back to its previous position when an
updated particle violates the constraints will allow a new
search closer to boundary.
k
k
Step-10. update P best;i and Gbest as follows
k

Pbest;i

8
< X k ifJX k < JPk1
i

best;i

k1
Pbest;i otherwise

21

6. Solution methodology using co-ordinated aggregation based

particle swarm optimization (CAPSO)
CAPSO is an improved variant of conventional PSO [25]. Each
particle moves considering only the positions of particles with better achievements than its own with the exception of the best particle which moves randomly. The coordinated aggregation may be
considered as a type of active aggregation where particles are attracted only by places with the most food. Hence CAPSO differs
from PSO described in previous section only in updating the velocity of each particle (Step-5 in previous section). Velocity of each
particle except the best in swarm is updated using following
relation.

qik w  qik1

k1

rj AF ij

k1

Xj

k1

 Xi

22

k1

rj denotes random digit between [0, 1] AF ij

is known as achievement factor, which is the ratio of difference between achievement
(objective function) of particle and better achievements by partik1
cle-j to the sum of all these differences. Expression for AF ij
is given as follows.
K1

k1

AF ij

JX i
X h

K1

 JX j

JX K1

i
l2T

K!
JX l

23

T represents the set of particles-j with better achievement in terms

if objective function.
Relation (22) is used to update the velocities of all particles except the best particle in the swarm. The velocity of best particle in
the swarm is updated using a random co-ordinator calculated between its position and the position of a randomly chosen particle in
the swarm as follows.

qpk w  qpk1 r X mk1  X k1

p

24

X k1
is randomly selected particle, X pk1 is the best particle in the
m
swarm and r is random digit between [0, 1]. The updating of best
particle of group behaves in a crazy way and helps the swarm to escape from local minima. Remaining steps are same as explained in
previous section. CAPSO has better convergence characteristics and
require only inertia weight parameter to start the updating
procedure.
7. Results and discussions
The algorithm developed in this paper has been implemented
on a sample radial distribution system as shown in Fig. 3 [11].
The system has seven load points in addition to a source sub-station and seven distributor segments. The system has been modied
due to presence of DG at load points LP-3, LP-6 and LP-8. In the absence of loss of supply at these nodes continuity is maintained with
the help of these distributed sources. Further it is assumed that
DG-6 present at LP-6 is sufcient to feed the need of only 50% consumers connected at the load point and distributed generator DG-3

612

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

DG-8

Table 2
Average load and number of customer at load points.

LP-8
#7

LP-7

Load point LP-K

Average load, LK, kW

Number of customer

1000
200

700
150

400
100

500
150

300
100

200
250

150
50

Table 3
Cost coefcients ak and bk for calculating objective function (J).

#6

Sr. No.

ak (Rs.)

246
100.00

40
72.90

81
72.00

250
200.00

32
180.00

4.5
38.40

5.0
93.60

bk (Rs.102)

LP-2
#1

#2

U SYS;4 k1 r 1 k2 r 2 k3 r 3

Source

#3

~
#4

rSYS4

U SYS;4
kSYS;4

DG-3

kSYS;5 k1 k4

LP-3
LP-5

U SYS;5 k1 r 1 k4 r 4
rSYS;5

#5

U SYS;5
kSYS;5

kSYS;7 k1 k2 k6
~

DG-6

LP-6

Fig. 3. Radial distribution system with DG at selected load points.

and DG-8 connected at load points LP-3 and LP-8 are having sufcient capacities to meet demand of all the consumers connected at
these load points. Table 1 gives current failure rates k0j , repair
time r 0j and bounds on these quantities for each section. Due to
presence of distributed energy resources modications on both
sides of decision variables have been considered. Records of
growth model of these variables allow to estimate bounds on these
variables. Table 2 provides average loads and number of customers
at each load point. Table 3 provides cost coefcient ak, bK as required to evaluate objective function (J). The ADCOST is selected
as Rs.1.5 per kW h.
Reliability modeling aspects are described in Section 2. kSYS ; rSYS
and USYS, for load points LP-2, LP-4, LP-5 and LP-7 (these do not
have DG) are evaluated using following relation.

U SYS;7 k1 r 1 k2 r 2 k6 r 6
rSYS;7

U SYS;7
kSYS;7

Since distribution generations are present at LP-3, LP-6, and LP-8,

the calculation of basic reliability indices will be different. The reliability network to evaluate basic indices for LP-3 is shown in Fig. 4
as per modeling aspects described in Section 2.

kS3 k1 k2
rS3

k1 r1 k2 r2
k1 k2

kSYS;3 and rSYS,3 with DG are obtained using relation (1) and (2) as
follows

kSYS;3 kS3 kd3 r s3 rd3 kSW3

ks3  kd3 r s3 r d3 kSW3  s3
kSYS;3

kSYS;2 k1

rSYS;3

r SYS;2 r 1

In above kd3 ; rd3 are failure rate and average down time of DG-3 kSW3
failure rate of manual switch of DG-3 s3 average service restoration
time of DG-3Similarly reliability diagram for LP-8 is shown in Fig. 5;
basic reliability indices kSYS;8 ; rSYS;8 and USYS,8 are evaluated using following relations.

U SYS;2 k1 r1
kSYS;4 k1 k2 k3
Table 1
System data for sample radial distribution system.
Distribution section

k0j 1=yr

kj;min 1=yr

kj;max 1=yr

r 0j hrs

rjmin(hrs)

rjmax(hrs)

1
2
3
4
5
6
7

0.4
0.2
0.3
0.5
0.2
0.1
0.1

0.2
0.05
0.1
0.1
0.15
0.05
0.05

0.6
0.3
0.5
0.8
0.3
0.15
0.15

10
9
12
20
15
8
12

6
6
4
8
7
6
6

15.0
14.0
18.0
30.0
22
12
18

613

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

Table 6
Initial population generated for PSO and CAPSO.

s3, rs,3
sw3, s3

LP - 3

d3, rd3
Fig. 4. Reliability network for LP-3.

s8, rs,8
sw8, s8

LP - 8

d8, rd8
Fig. 5. Reliability network for LP-8.

Decision variables

#1

#2

#3

#4

#5

k1
k2
k3
k4
k5
k6
k7
r1
r2
r3
r4
r5
r6
r7
Objective function,
J  104

0.3040
0.0719
0.1392
0.1522
0.1980
0.0941
0.0928
6.2027
7.4233
5.1253
22.0984
15.4493
9.5325
6.7674
2.7322

0.2253
0.0623
0.2257
0.2511
0.2519
0.1041
0.0943
7.4363
11.0731
7.5651
14.5564
8.9905
7.6173
8.3229
2.2322

0.2208
0.0674
0.2834
0.2043
0.2638
0.0944
0.0886
7.5631
8.5684
4.2047
14.1669
17.8474
8.6555
10.6841
2.2402

0.2248
0.1946
0.4197
0.1499
0.2388
0.1384
0.0654
6.0613
11.8895
6.1739
11.9885
8.7557
11.1945
6.3030
2.0557

0.2339
0.0539
0.2940
0.2087
0.2219
0.1231
0.0955
7.7577
10.0196
10.8947
8.4468
13.3241
6.2040
11.0793
2.6592

Table 4
Failure rate and average down time for distributed generation.
Sr.
No.

DG

Failure
Rate
kd 1=yr

Down
time rd
(h)

Failure rate of
manual switch ksw
(1/yr)

Service
restoration
time s (h)

DG3
DG6
DG8

1.50

15.78

0.5

1.5

1.75

18.68

0.5

1.5

2.00

20.61

0.5

1.5

2
3

Table 7
Control parameter for DE, PSO and CAPSO.

kS8 k1 k2 k6 k8
r S8

S. No.

Parameters

Values of parameters

1
2
3
4
5
6
7
8
9
10

Population size, (DE)

Step size (a)
Cross over rate (Cr)
Maximum generation specied kmax
c1
c2
wmax
wmin
Maximum iteration in PSO/CAPSO
Population size (PSO/CAPSO)

10, 20
0.7
0.3
300
1.0
1.0
1
0.1
300
5, 10, 20

k1 r 1 k2 r 2 k6 r 6 k8 r 8
kS8
distribution network or DG and remaining are supplied by DG-6
in case of loss of supply from source node. The basic reliability indices are evaluated for LP-6A using following formula.

kSYS;8 kS8  kd8 kd8 r S8 rd8 kSW8

r SYS;8

kS8  kd8 r S8 rd8 kSW8  s8

kSYS;8

kSYS;6A k1 k4 k5

kd8 ; r d8 are failure rate and average interruption duration of DG-8.

kSW8 ; s8 are failure rate of manual switch and service restoration
time for DG-8.
DG capacity available is sufcient to satisfy the requirement of
only 50% consumers connected at LP-6. The reliability network
for LP-6 is shown in Fig. 4. LP-6 is partitioned in two separate
nodes i.e. 6A and 6B. 50% consumer may be supplied by either

U SYS;6A k1 r 1 k4 r 4 k5 r 5
rSYS;6A

U SYS;6A
kSYS;6A

Basic reliability indices at LP-6B is evaluated using following relations as given by relations (1) and (2)

Table 5
Initial population generated for DE.
Decision variables

#1

#2

#3

#4

#5

#6

#7

#8

#9

#10

k1
k2
k3
k4
k5
k6
k7
r1
r2
r3
r4
r5
r6
r7
Objective function, J  104
F/NF*

0.2585
0.0636
0.2208
0.2217
0.2749
0.0767
0.1001
7.5019
11.5107
5.7112
8.8760
14.1625
11.8015
16.3618
2.2411
F

0.3914
0.1299
0.4862
0.5069
0.2705
0.0506
0.0890
10.4011
12.0494
11.9502
27.5932
19.2267
8.3456
8.8771
1.0170
NF

0.2320
0.2093
0.4127
0.7234
0.2323
0.0939
0.0711
8.6868
8.4746
14.4646
8.5542
7.2507
10.3021
8.4427
0.9647
NF

0.4874
0.1163
0.4514
0.5934
0.2777
0.1408
0.0996
6.7307
11.8140
5.4722
13.7090
17.1514
10.9863
14.5380
0.8031
NF

0.4799
0.1278
0.1356
0.3538
0.2767
0.1374
0.1354
6.4992
7.8369
11.7145
29.5521
12.6668
9.5162
12.5851
1.2098
NF

0.3470
0.1009
0.2651
0.6560
0.1829
0.1409
0.0945
12.8142
10.7832
6.5088
27.6815
20.6401
7.0240
6.1168
1.1411
NF

0.2565
0.0955
0.3874
0.3548
0.1969
0.1448
0.1019
13.3613
6.5648
7.8263
23.9893
13.8181
7.7196
11.6384
1.3185
NF

0.3955
0.2051
0.1332
0.4777
0.2419
0.0984
0.0840
12.2991
10.3990
8.4846
11.3934
16.8318
10.3243
12.6372
1.3519
NF

0.3186
0.0523
0.1849
0.4668
0.2273
0.0501
0.0543
11.7836
12.5356
17.3843
11.4329
13.2455
6.7416
10.2956
2.5975
NF

0.2985
0.0844
0.2336
0.4535
0.2082
0.0620
0.0992
12.6151
11.0427
14.8331
24.9299
21.6527
9.1818
14.4749
1.4889
NF

NF denotes not feasible solution.

F denotes feasible solution.

614

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

Source

#1

#4

#1

#5

#4

#5

Table 9
CPU Time for convergence required for DE, PSO, and CAPSO Techniques on Intel
core2duo processor, 2.9 GHz.

LP 6A

LP 6B

S. No

Techniques

Population size

CPU time (ms)

DE

10
20

125
250

PSO

05
10
20

78
141
281

CAPSO

05
10
20

62
141
250

DG - 6

Fig. 6. Reliability network for indices evaluation at LP-6.

Table 8
Optimized values of failure rates and repair times as obtained by DE, PSO and CAPSO.
Variables

Magnitudes as
obtained by DE

Magnitudes as
obtained by PSO

Magnitudes as
obtained by CAPSO

k1 (yr)
k2 (yr)
k3 (yr)
k4 (yr)
k5 (yr)
k6 (yr)
k7 (yr)
r1 (h)
r2 (h)
r3 (h)
r4 (h)
r5 (h)
r6 (h)
r7 (h)
Objective
function
(J)  104

0.2331
0.1789
0.3104
0.3430
0.2845
0.0920
0.1483
6.0033
6.0767
6.1595
8.0100
7.0643
9.2649
8.3064
1.1701

0.2471
0.1357
0.2932
0.3685
0.2998
0.0952
0.1487
6.0023
6.0012
13.2311
8.0029
7.0685
11.3771
6.0047
1.1836

0.2245
0.1743
0.3449
0.3731
0.2719
0.0901
0.1239
6.4414
6.7191
8.6474
8.1497
7.2750
6.1672
6.7933
1.1841

kSYS;6B kSYS;6A  kd6 r SYS;6A r d6 kSW6

r SYS;6B

kSYS;6A  kd6  r SYS;6A  r d6 kSW6  s6

kSYS;6B

In above kd6 ; r d6 are failure rate and average down time of DG-6. ksw;6
is failure rate of manual switch of DG-6. s6 denotes average service
restoration time of DG-6.
The basic indices evaluated as above are used to evaluate
SAIFI, SAIDI, CAIDI and AENS as given by relation 8,10,12, and 14

respectively. Similarly EENSO and EENSD are also evaluated using

relations 4,4a,4b.
Table 4 gives failure rate, average down time for distributed
generations present at load points 3, 6 and 8. It also provides failure rate (ksw) and service restoration time (s) for manual switch of
these distributed generations. In evaluating EENSO (Without DG)
the DG at location 3, 6, and 8 load points are not considered. The
problem as formulated in Section 3 has been solved using DE,
PSO and CAPSO based algorithm as explained in Sections 46.
Table 5 gives initial population generated for DE. Table 6 presents
the initial population generated and used in PSO and CAPSO algorithms. These populations were generated with an inbuilt random
number generator program. Table 7 gives the nal control parameter selected for DE, PSO and CAPSO (see Fig. 6).
Table 8 gives optimum failure rate and repair time of each section as obtained by these three methods, along with cost function.
It is observed that results obtained by these three methods are in
close agreement. Further in DE based optimization all the members
of population become feasible and cluster around the optimum
solution at the end of convergence. This has been an important
characteristics of the solutions obtained by DE and has been
pointed out in Ref. [20]. This has been also observed by us in
our computational procedure at the end. This has been put in
Table 8a. Table 8a shows all solutions (population) along with
one optimal solution with associated values of objective functions.
This table validates the mentioned characteristics. Table 9 gives
CPU time required to solve the optimization problem by three
methods with various population sizes. It was observed that PSO
converges with even ve population size whereas DE does not converge with ve population size. Solution with DE converges with
ten population size. But as population size is increased CPU time
with PSO increases from 78 ms to 281 ms with population size 5

Table 8a
Population as obtained at the converged iteration along with values of objective function.
Decision Variables

#1

#2

#3

#4

#5

#6

#7

#8

#9

#10

k1
k2
k3
k4
k5
k6
k7
r1
r2
r3
r4
r5
r6
r7
Objective function, J  104
F/NF*

0.2215
0.1501
0.3951
0.3893
0.2642
0.0992
0.1460
6.1616
6.0016
5.5218
8.0227
7.1219
6.9751
7.4315
1.1796
F

0.2195
0.1600
0.3716
0.3933
0.2672
0.0978
0.1472
6.4672
6.0016
5.3490
8.0153
7.0767
7.3910
7.1725
1.1786
F

0.2324
0.1616
0.3264
0.3632
0.3000
0.0936
0.1497
6.0052
6.0174
5.0580
8.0009
7.0000
6.0009
17.9415
1.1769
F

0.2231
0.1646
0.3733
0.3843
0.2658
0.0872
0.1369
6.6341
6.0015
6.0633
8.0070
7.1101
6.6675
10.2842
1.1836
F

0.2331
0.1789
0.3104
0.3430
0.2845
0.0920
0.1483
6.0033
6.0767
6.1595
8.0100
7.0643
9.2649
8.3064
1.1701
F

0.2312
0.1619
0.3286
0.3683
0.3000
0.0920
0.1495
6.0006
6.0013
6.8840
8.0004
7.0001
6.0009
17.9447
1.1768
F

0.2200
0.1639
0.3381
0.3928
0.2492
0.1065
0.1490
6.4915
6.2205
5.1022
8.0366
7.1857
6.9686
6.8025
1.1818
F

0.2247
0.1683
0.3320
0.3749
0.2922
0.0890
0.1448
6.0185
6.2610
5.0691
8.0115
7.0374
6.5093
8.1343
1.1757
F

0.2328
0.1576
0.3283
0.3702
0.3000
0.0914
0.1500
6.0033
6.0095
7.7476
8.0009
7.0000
6.0000
17.9345
1.1774
F

0.2333
0.1575
0.3296
0.3628
0.3000
0.0956
0.1500
6.0096
6.0004
5.8432
8.0012
7.0010
6.0014
17.9507
1.1772
F

NF denotes not feasible solution.

F denotes feasible solution.

615

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

Table 10
Current and optimized reliability indices for radial distribution system.
S. No.

Index

Current values

1
2
3
4

SAIFI interruptions/customer
SAIDI h/customer
CAIDI h/customer interruption
AENS kW h/customer

0.6610
7.2671
10.9933
22.2334

Fig. 7. Variation of best value of objective function (J) with population size 5 in PSO,
CAPSO and population size 10 in DE.

Optimized values

Threshold values

DE

PSO

CAPSO

0.4997
3.0731
6.1495
9.0454

0.5000
3.3297
6.6595
9.9999

0.5000
3.2324
6.4648
9.8024

0.5000
4.0000
8.0000
10.0000

Fig. 9. Variation of best value of objective function (J) with population size 20 in
PSO, CAPSO and DE.

8. Conclusions
This paper has presented a methodology for reliability design of
a radial distribution system in the presence of distributed generation. An optimization problem has been formulated which obtains
optimum failure rates and repair times of each section subject to
satisfaction of inequality constraints based on energy and customer based reliability indices. Operating philosophy of DG has
been assumed in standby mode and cost of purchased energy is
more than that supplied from distribution substation. Formulated
problem has been solved using two most robust and popular population based algorithm, i.e. DE and PSO. The solution has also been
obtained using CAPSO, one of the variant of PSO. The results obtained have been compared. PSO requires less CPU time (population size 5) even with less population size than used in DE
(population size 10). It was also observed that CPU time required
for convergence increases with population size for all the three
techniques. It was also observed that for population size twenty
CPU time required by DE is less than that required by PSO.
References
Fig. 8. Variation of best value of objective function (J) with population size 10 in
PSO, CAPSO and DE.

and 20 respectively. The same is observed with DE and CAPSO. For

population size 20 the CPU time required by DE is less than that required by PSO but equals to that required by CAPSO. In all the three
methods CPU time increases as population size increases. Table 10
gives reliability indices as obtained using current value of decision
variables as well as those obtained using optimized variables by
three methods. Figs. 79 shows convergence curves as obtained
by three methods for different population sizes.

[1] Trebolle D, Gomez T, Cossent R, Frias P. Distribution planning with reliability

options for distributed generation. Electr Power Syst Res 2010;80(2):2229.
[2] Mendez VH, de la Fuente JI, Gomez T, Arceluz J, Marin J, Madurga A. Impact of
distributed generation on distribution investment deferral. Int J Electr Power
Energy Syst 2006;28(4):24452.
[3] Mendez VH, Rivier J, Gomez T. Assessment of energy distribution losses for
increasing penetration of distributed generation. IEEE Trans Power Syst
2006;21:53340.
[4] Cossent R, Gomez T, Frias P. Towards a failure with large penetration of
distributed generation: is the current regulation of electricity distribution
ready ? regulatory recommendation under a European perspective. Energy
Policy 2009;37:114555.
[5] Ackerman T, Anderson G, Soder L. Distribution generation: a denition. Electr
Power Syst Res 2001;57:195204.

616

R. Arya et al. / Electrical Power and Energy Systems 43 (2012) 607616

[6] Sallam AA, Desouky M, Desouky H. Evaluation of optimal reliability indices for
electrical distribution systems. IEEE Trans Reliab 1990;39(1):25964.
[7] Chang WF, Wu YC. Optimal Reliability design in an electrical distribution
system via polynomial-time algorithm. Int J Electr Power Energy Syst
2003;25:65966.
[8] Chowdhury AA, Custer DE. A value based probabilistic approach to designing
urban distribution systems. Int J Electr Power Energy Syst 2005;27:64755.
[9] Bhowmik S, Goswami SK, Bhattacherjee PK. A new power distribution systems
planning through reliability evaluation technique. Electr Power Syst Res
2000;54:16979.
[10] Su CT, Lii GR. Reliability design of distribution systems using modied genetic
algorithms. Electr Power Syst Res 2002;60:2016.
[11] Arya R, Choube SC, Arya LD. A decomposed approach for optimum failure rate
and repair time allocation for radial distribution systems. J Inst Eng (India)
2009;Pt. El-Vol.89:38.
[12] Bae In-Su, Kim Jin-O. Reliability evaluation of distributed generation based on
operation mode. IEEE Trans Power Syst 2007;22(2):78590.
[13] Bae IS, Kim JO, Kim JC, Singh C. Optimal operating strategy for distributed
generation considering hourly reliability worth. IEEE Trans PS 2004;19(1):
192287.
[14] Louit D, Pascul R, Benjevic D. Optimal interval for major maintenance actions
in electric distribution networks. Int J Electr Power Energy Syst 2009;31(78):396407.
[15] Costa PM, Matos MA. Assessing the contribution of micro grids to the
reliability of distribution network. Electr Power Syst Res 2009;79:3829.

[16] Billinton R, Allan RN. Reliability evaluation of engineering systems: concepts

and techniques. Springer International Edition; 2007.
[17] Duane JT. Learning curve approach to reliability monitoring. IEEE Trans
Aerospace 1964(2):5636.
[18] Billinton R, Allan RN. Reliability evaluation of power system. Springer
International Edition; 1996.
[19] Rainer Storn, Price KV. Differential evolution a simple and efcient adaptive
scheme for global optimization over continuous spaces. Technical report TR95012, ICSI.
[20] Price Kenneth V, Storn Rainer M, Lampinen Jouni A. Differential evolution, a
practical approach to global optimization. Springer International Edition;
2005.
[21] Lampinen J. A constraint handling approach for differential evolution
algorithm. In: Proc of 2002 congress on evolutionary computation (CEC-02),
vol. 2. p. 146873.
[22] Kennedy J, Eberhart RC. Particle swarm optimization. Proc IEEE international
conference on neural networks, Piscataway, NJ; 1995. p. 19428.
[23] Park JB, Lee KS, Shin JR, Lee KY. A Particle Swarm optimization for economic
dispatch with non-smooth cost functions. IEEE Trans Power Syst
2005;20(1):3442.
[24] He S, Prempain E, Wu QH. An improved particle swarm optimization for
mechanical design problems. J Eng Optim 2004;36(5):585605.
[25] Vlachogiannis JG, Lee KY. A comparative study on particle swarm optimization
for optimal steady State performance of power systems. IEEE Trans Power Syst
2006;21(4):1318728.