IET Generation, Transmission & Distribution

Research Article

ISSN 1751-8687
Optimal DG allocation and sizing in presence Received on 20th March 2015
Revised on 27th June 2015
of storage systems considering network Accepted on 3rd August 2015
doi: 10.1049/iet-gtd.2015.0367
configuration effects in distribution systems www.ietdl.org

Fazel Abbasi 1 ✉, Seyed Mehdi Hosseini 2

1
Department of Electrical and Computer Engineering, Shahrood University, Shahrood, Iran
2
Department of Electrical and Computer Engineering, Babol Noshirvani University of Technology, Babol, Iran
✉ E-mail: [email protected]

Abstract: Optimal performance of power distribution networks greatly depends on network configuration, location and
size of distributed generations (DGs) units and storage systems. That is, for different configurations, different optimal
locations and sizes of DGs can be found and vice versa. Therefore, the impact of both location and size of DG and
network configuration should be considered in the planning process, simultaneously. Also, the presence of storage
systems in the distribution system leads to some loads to be supplied even in fault conditions. In this study,
distribution system reconfiguration (DSR), for considering network configuration effect that runs in offline mode with
constant loads, and optimal DG allocation and sizing problems are studied simultaneously to find an optimal condition
for distribution network based on operational thresholds and reliability improvements. Non-dominated Sorting Genetic
Algorithm is used to solve these problems simultaneously. Power losses, energy not supplied (ENS) and the costs
associated with DG are the objectives that are studied. The method of calculating ENS in DSR problem in the presence
of DGs with storage systems is explained and the impact of protective equipment is considered, as well. The proposed
approach is applied on different test systems, and its effectiveness is shown in various conditions.

1 Introduction programming. Nevertheless, they have not exactly modelled the

network losses [2]. In [3], the losses are exactly modelled in a
This Today’s power companies are trying to increase their profits as Benders decomposition solution, but because of being non-convex,
well as improving the satisfaction level of their customers. Therefore, their work cannot ensure a global solution.
they have moved towards optimal operation of distribution systems. In this paper, the meta-heuristic methods are used to solve the
The necessities of this optimal performance are the application of multi-objective optimisation problem. Such methods are popular
distributed generation (DG) resources with storage systems and due to their wide application in black-box models [4]. However,
improvement of reliability in distribution systems. Location and the final solution in these methods may be a local optimum
size of DGs in a distribution network greatly influence their solution. References [5–8] are the first papers in which the DSR
effectiveness. On the other hand, finding the best configuration of problems have been studied via using the heuristic methods, and
network is another efficient and cheap way, which has direct and most of them have been derived from the work in [5]. In [9] the
indirect impacts on the system reliability. There are two types of optimal power flow and sensitivity analysis have been used to
switches in power distribution networks: sectionalised switches determine the next switch that should be opened. Gradually,
that are normally closed, and tie switches that are normally meta-heuristic and intelligent algorithms such as genetic algorithm
opened. Changing the status of these switches in a way that [10–12], ant colony optimisation [13], particle swarm optimisation
maintains the operational constraints and radial condition of [14], tabu search [15, 16] and so on, have been applied on DSR
distribution system is called distribution system reconfiguration problem.
(DSR). From the optimisation prospective, various methods have In [17], the micro-genetic multi-objective reconfiguration
been presented so far in the literature to solve the DSR problem. algorithm has been used to minimise the power loss as well as to
These methods can be classified into two main groups: improve the reliability indices, in which the solutions are finally
presented as pareto optimal front. Niknam et al. have proposed a
new modified Honey Bee Mating Optimisation (MHBMO)
† Classic or mathematical methods (deterministic), algorithm to simultaneously investigate the DSR problem and
† Heuristic and meta-heuristic methods (stochastic). optimal sizing of renewable DGs [18]. However, the location of
DGs was determined before optimisation. Similar to the previous
In the recently published mathematical methods, at first, DSR reference, optimisation in this approach is also based on
problem is converted into a convex problem, and then it is solved non-dominated pareto optimal solution. Minimisation of power
by Lagrangian method or some other methods. The main losses, cost, voltage deviation of buses and the total emission
advantage of these methods is their global optimum solution. produced are the goals of this reference. Recently in [19], the
However, there are some difficulties in these methods; for DSR problem with the purposes of loss reduction and
example, the radial constraint and other constraints such as the bus improvement of ENS has been investigated using Non-dominated
voltages or branches currents should be modelled as a Sorting Genetic Algorithm (NSGA)-II optimisation method.
mathematical function. From the viewpoint of optimisation, DSR NSGA II is one of the well-known algorithm that is used in many
is a mixed-binary non-linear optimisation problem where binary optimisation problems [20].
variables and continuous variables represent switches status and One of the important issues that should be studied here is the
power network parameters, respectively [1]. Ramos et al. at first, considerable mutual dependence between the optimal DG site and
have converted the DSR problem to a linear problem and then size determination and the configuration of system. In other words,
have solved it with the aim of loss minimisation using linear the optimal configuration of distribution system also depends on

IET Gener. Transm. Distrib., pp. 1–8

& The Institution of Engineering and Technology 2015 1
the site and size of DGs. The main contribution of this paper is to 2.1 Assumptions
investigate simultaneously the best configuration of the network
and the optimal DG allocation and sizing problem. Moreover, this In the proposed method, the following suppositions are made for the
paper discusses how to calculate the ENS by considering ENS calculation, as listed below:
continuous supplies for DGs (because of storage systems). The
problem objectives are reliability improvement and loss reduction. i. Failures are only related to lines and they are modelled like a
In this paper, the system reliability index is energy not supplied two-state model: normal state or repair state (connected or
(ENS). The NSGA-II method has been used for solving this disconnected). In addition, they are all independent from each other.
optimisation problem. The proposed multi-objective algorithm ii. Protective equipments work properly, such that any fault in the
does not require any weighting factors for converting several system with radial structure is cut by the nearest overcurrent
objective functions into an equivalent single one. Also the protective device (from the source side).
objectives are minimised separately, and contrary to the iii. When a fault occurs in a part of the system and there is a DG
conventional methods, it may result several solutions in the form with storage system in downstream of the fault, the DG should be
of pareto front. disconnected from the system, based on IEEE Std 1547.
This paper is organised as follows: Section 2 discusses the Nevertheless, it is assumed that the storage system still supplies
assumptions of reliability calculation and investigates the the demand at its own bus, although it cannot supply other buses.
condition of the nodes when a fault occurs. In Section 3, the Therefore, since now the storage system supplies its load locally,
ENS calculation in the presence of energy storages is explained the losses are negligible. In addition, it is assumed that the loads at
and calculated for different topologies of a six-bus test system, the bus to which the DG has been connected (DG bus), are
and then the effects of system configuration, sites, and sizes of controllable. In other words, there is a capability of connecting or
DGs on ENS value are shown. Multi-objective optimisation disconnecting these loads. In fact, in a fault condition, if storage
method is explained in Section 4. In addition, the numerical system generation exceeds the total load demand, all loads are
results of simulation on two test systems with 33 and 17 buses supplied by the storage system. Otherwise (i.e. when storage
are presented in Section 5. Finally, the conclusion is given in system generation is lower than the total load demand), load
Section 6 of paper. shedding should be performed, and consequently, some loads of
the DG bus will experience the interruption. In this paper, it is
supposed that the apparent power of the storage system is equal to
apparent power of DG.
2 Reliability considerations

This paper focuses on distribution systems with radial structures. 2.2 Bus classification in fault situation
Such systems are formed from a group of equipment such as
lines, cables, protective devices, buses and other equipments. In When a fault occurs in the distribution system, the system buses can
radial distribution systems, the electric power demands at each be classified in different groups. According to Fig. 1, this
bus are supplied through all the equipment between the bus and classification is explained as follows:
power source. Therefore, the basis of series system can be
applied to these systems [21]. The main indices of reliability for Group 1: The nodes that are not influenced by the fault occurrence
this kind of systems can be divided into two main categories: and still supply their connected loads (all nodes except 3, 9, 10,
customer-based indices, including system average interruption 11 and 12).
frequency index, system average interruption duration index, etc., Group 2: The nodes that experience the outage when a fault occurs,
and load-based indices including ENS, average customer but can be supplied by their own branches after switching actions
curtailment index, etc. [21]. Among them, ENS is of special (nodes 3, 9 and 11).
importance because it is related to both the network profit and Group 3: The nodes that experience the outage, but can be supplied
customer satisfaction. Therefore, the ENS index is used in this by closing one or more tie switches after some time (node 12).
paper to carry out the reliability assessment. Moreover, it should Group 4: The nodes that are disconnected from the source and
be noted that the analytical approach is used in the reliability should be remained interrupted until the faulted line would be
calculation in this paper. repaired (node 10).

Fig. 1 Civanlar system during fault with two supposed DGs and storage systems (SS)

IET Gener. Transm. Distrib., pp. 1–8

2 & The Institution of Engineering and Technology 2015
 2.3 Impact of protective equipments the active power demand of each consumer, SSS is the nominal
apparent power of the storage system (in this paper equal to
Reliability calculation severely depends on the kind of protective nominal apparent power of DG), cosj is the power factor of the
devices. These devices can be classified into two general load, nc is the number of consumers who can be supplied by the
categories: the first category includes the devices without the storage system at the DG-bus, and PSS-used is the total active
capability of cutting the fault current; however, they are able to cut power load that can be supplied by the storage system at DG-bus
the nominal current, such as sectionalisers. In contrast to the first (or the total active power generation of the storage system).
category, in the second category there are equipments with the Fig. 2 shows three different topologies of a six-bus distribution
capability of cutting the fault current such as breakers, fuses and system. The system data are given in Table 1. In addition, nominal
sectionaliser-fuses. In general, the higher number of the devices of apparent power of the DG is 500 KVA, and repair and manoeuvre
the second category in a distribution system leads to higher times are assumed equal to 1 and 0.5 h, respectively.
reliability level. Thus, the impact of kind and site of these devices The difference between topologies (a) and (b) is related to the
should be considered in reliability calculation algorithm. As the system configuration and between (b) and (c) is related to the
various configurations are checked in DSR problem, the used location of DG. The results of ENS calculation for these
algorithms in DSR problem should have generality. In other topologies are listed in Table 2. This table shows the DG location
words, the algorithms, at first, should be capable to recognise the affects the unsupplied power (Ppns) of the DG-bus, when this bus
nodes, branches and protective equipment of the spanning tree. is influenced by the fault. As well, the system configuration
influences the unavailability of load buses (U ). In fact, these
results not only imply the importance of the system configuration
3 ENS calculation and assessment and DG location on the reliability, but also they demonstrate that
for proper performance of the distribution system, the DSR
After recognising the spanning tree and considering the impact of problem and optimal DG allocation and sizing problem should be
protective devices corresponding to each configuration, ENS can considered simultaneously. Although the size of storage system
be calculated. First, for each line outage, the system nodes are (equal to size of DG) is similar in all the three topologies, it is
classified according to the previous explained section. The nodes obvious that the size of storage system is one of the most
of the groups # 2 and 3 experience the outage during the important parameters affecting the value of Ppns. Among the
manoeuvre time. As well, the nodes of group 4 should be topologies of Fig. 2, topology (c) has the smallest value for the
interrupted during the repair time. Then, the unavailability of ENS index. In Table 2, it should be noted that because the
nodes (Ui) for each contingency is obtained through multiplying capacity of storage system exceeds the load demand at bus 6, the
the outage time (r) by the outage rate of faulted line (l j), as loads at this bus are always supplied by the storage system, and
formulated in (1). therefore, Ppns is zero at this bus.


ENS = Ni=1 Pi . Ui

L (1)
Ui = j=1 lj . r 4 Multi-objective optimisation

The ENS corresponding to each bus is obtained via multiplying the 4.1 Problem formulation
unavailability of the bus (Ui) by the active power demand at the bus
(Pi). Finally, the total ENS is calculated by adding the ENS of all In this paper, the DSR problem and the DG sitting and sizing
buses together. It should be noted that the active power demand problem considering storage systems are performed to minimise
(Pi) represents the value of not supplied active load at each bus the following objectives
(this parameter at DG-bus is shown by Ppns). For the DG-buses at 
the downstream of fault, the Ppns is calculated by (2)–(5), since in Plosses = 3Ri Ii2 (6)
a faulted condition, the storage systems are still able to supply the
local loads at these buses. 
N
ENS = Pi .Ui (7)
P i=1
P= i (2)
Nc
Subject to the constraints:
 
S × cos w
nc = SS (3) Vi min ≤ Vi ≤ Vi max (8)
P

PSS-used = P × nc (4) Ij ≤ I j max (9)

Ppns = Pi − PSS-used (5) where R, I, Vmin, Vmax and Imax are respectively resistance of line,
current of line, minimum voltage of bus, maximum voltage of bus
where Nc is the number of the controllable loads (or consumers), P is and maximum current of line. It should be noted that the correct

Fig. 2 Three different topologies of six-bus test system

IET Gener. Transm. Distrib., pp. 1–8

& The Institution of Engineering and Technology 2015 3
Table 1 Data for six-bus test system identified, are temporarily removed from population, and are put in
the first pareto front (rank = 1). Then, the remaining chromosomes
Line Node l, f/year KW–KVAr a
No. of customers a
of the population are compared together again, and the
L1 1–2 0.2 800–300 28
non-dominated chromosomes are temporarily removed from
L2 2–3 0.3 600–250 20 population and are put in the second front (rank = 2). This process
L3 2–4 0.2 1000–400 45 is repeated until the total number of chromosomes in fronts
L4 3–5 0.1 750–300 25 exceeds the initial population size. Then, for each member of the
L6 5–6 0.2 400–150 22
L5 4–5 0.1
fronts a criterion called ‘crowding distance’ is assigned, that is the
normalised distance to the closest neighbours in the front and in
a
Values at the end of the line
the objective space [23]. In this stage, the parents are selected to
be used for producing the next generation. If the number of
parents exceeds the initial population size, some of the
solutions must have the radial structures. Therefore, holding the chromosomes in the latest front (front K) should be removed, so
radial structure can be considered as another constraint for the that the number of parents equals to the initial population size.
topologies of the problem. A multi-objective evolutionary This work is performed by using the concept of diversity
algorithm based on non-dominated optimal solution, called preservation of the solutions, so that the diversity of parents
NSGA-II, is proposed for solving the above problem. For the first grows. According to this concept, the solutions with higher
time, this method has been suggested by Deb et al. [22]. All crowding distances cover the larger domain of the objective space.
intelligent evolutionary algorithms are started with an initial Therefore, it is desirable to remove the solutions with lower
population, which is usually created randomly, and then, this crowding distances. Now, two required chromosomes (father and
population improves at each stage or iteration. Sometimes, this mother) for the crossover operation are selected from the parents
random creation may cause some problems. For example, the population (selected in previous stage) via using a tournament
randomly produced individuals may represent a looped system or a selection approach (with n = 2). Here, because there are two fitness
system with some islands, despite the radial constraint should be functions for each chromosome, the selection operator is different
satisfied. To avoid these difficulties, a new coding technique, from the single-objective genetic algorithm. Between two
called edge window decoder technique, is used in [12] to produce randomly selected individuals, the individual with lower rank is
the chromosomes with radial structures. The current paper also selected as the father or mother. If both individuals are from the
applies the edge window decoder technique for producing initial same front, then the individual with higher crowding distance is
population. The losses, voltage and current constraints are
obtained from power flow solution. After calculating the problem
objectives (losses and ENS index), the fitness functions are defined
for considering the effects of constraints. Thereby, the constrained
optimisation problem is converted to an unconstrained one. In
intelligent algorithms based on pareto optimal, each objective
function is separately optimised during the programme execution.
For this reason, all the constraints should be observed in each
fitness function. The fitness functions are formulated as follow
   
Ffit Plosses + l1 N (V ) + l2 N (I)
F= f1
= (10)
Ffit f2 ENS + l1 N (V ) + l2 N (I)

N (V ) = The number of nodes with out of range voltages

(11)
N (I) = The number of lines with out of range currents

where l1 and l2 are the penalty factors, and here they are supposed
equal to 1000.

4.2 NSGA-II method

After calculating the fitness functions in each iteration, all the parent
and offspring chromosomes are compared together using the
dominance concept and then they are classified in different pareto
fronts. Determination of fronts from the parent and offspring
chromosomes always guarantees the elitism. According to the
dominance concept, in a minimisation problem, chromosome A
dominates chromosome B, if and only if none of the fitness
functions yield higher (worse) value for A (than B) and at least for
one fitness function, the value of A is lower (better) than B. After
comparing all the chromosomes in the population with each other
(parents and offspring), the non-dominated chromosomes are

Table 2 Results of six-bus test system

U, h/year Ppns (KW) ENS (KWh/year)

Topology U6 U5 U4 U3 U2

(a) 0.6 0.4 0.4 0.55 0.2 300 1250

(b) 0.65 0.45 0.5 0.4 0.2 300 1295
(c) 0.65 0.45 0.5 0.4 0.2 0 1035
Fig. 3 Block diagram of the proposed method

IET Gener. Transm. Distrib., pp. 1–8

4 & The Institution of Engineering and Technology 2015
Fig. 4 Roll of crossover and mutation operators in the proposed method

selected. After determining the father and mother, the crossover is It should be mentioned that in the distribution systems, several
applied. In the final step, the mutation operator is applied on the lines usually may have no protective equipment and they cannot
offspring resulted from the crossover operation. This process is change their state (open/close). Therefore, these lines cannot take
repeated so that the number of offspring is equal to the number of part in the reconfiguration problem and they should always be in
initial population. After generating the new population, the losses the connected state (close). Hence, these lines should not be
and ENS are calculated again, and the above-mentioned processes selected for disconnecting (opening) in the mutation process.
are repeated. Fig. 3 shows the block diagram of this algorithm. The mutation operator for the first part of the chromosome is
Fig. 4 illustrates the crossover and mutation operators. In this performed as follows: A line from the offspring’s lines (which can
problem, the proposed chromosome has been constructed from take part in the DSR) is randomly selected and disconnected from
three parts. The first part represents the system configuration and it the system. In this state, the system is divided into two separated
is related to the reconfiguration problem. In this part, the gens are parts. Then another line (from those outside the offspring) is
the numbers of lines that are closed (connected) in the system. The randomly selected and added to the offspring. Like the crossover
second and third parts show the location and size of DGs with operator, that line should be chosen to satisfy the radial constraint.
storage systems, respectively, and they are related to the DG In addition, for the second and third parts of the chromosome, the
sitting and sizing problems. As of the structure of proposed mutation operator can be performed similar to the first part.
chromosome (its three different parts), and the constraint of radial According to the previous studies, this kind of mutation procedure
structures for the distribution system, the crossover operator is needs a high mutation rate so that to achieve a proper
separately performed for each part of the chromosome. In the first convergence [12].
part (that is related to the system configuration), at first, the lines
that are common in the both parents (P1 & P2) are selected to
initialise the offspring. Then, the uncommon lines are randomly Table 4 Number of controllable loads or customers at each bus
selected from P1 and P2 to complete the offspring’s gens. This Civanlar Baran
procedure may cause a closed loop configuration. Therefore, the
system structure should be checked. In other words, the lines Bus No. of Bus No. of Bus No. of
should be selected such that the offspring introduces a full customers customers customers
spanning tree, which satisfies the radial constraint. The crossover
1 0 1 0 18 15
process in the second and third parts of chromosome is simple and 2 0 2 11 19 15
it is from one point type. 3 0 3 8 20 13
4 0 4 10 21 9
5 8 5 8 22 9
6 9 6 9 23 5
Table 3 Parameters of simulation 7 15 7 15 24 10
8 15 8 15 25 50
Parameter Civanlar system Baran system 9 5 9 5 26 8
10 10 10 10 27 8
initial population size 15 30 11 9 11 9 28 10
crossover rate 0.8 0.8 12 12 12 12 29 14
mutation rate 0.8 0.8 13 12 13 12 30 4
number of generation 800 600 14 20 14 20 31 28
repair time, h 1 1 15 2 15 2 32 38
manoeuvre time, h 0.5 0.5 16 12 16 12 33 9
DGs size (or storage systems size), KW 2500–3500–5000 300–500–700 17 10 17 10

IET Gener. Transm. Distrib., pp. 1–8

& The Institution of Engineering and Technology 2015 5
Fig. 5 All of the obtained topologies during the simulation of Civanlar system

5 Simulation results These parameters reduce 81.42 and 59.53%, respectively,

compared to the original state of the system (original configuration
5.1 Case study 1 and without DG and storage system).

Civanlar system (100 MVA and 23 KV) is the first system on which
the proposed approach is applied, as illustrated in Fig. 1. This system 5.2 Case study 2
by default (original configuration and without DG and storage
Fig. 6 shows the Baran test system with 10 MVA and 12.66 KV,
system) has three tie lines with the losses and ENS equal to
along with the type and location of protective equipments. The
511.439 KW and 112.43 MWh/year, respectively. The lines L13
dashed lines denote the tie lines of system. The losses and ENS
and L12 cannot participate in DSR and always should be
are 202.6817 KW and 6865.4 KWh/year in the original system,
connected to the system. The simulation parameters and sizes of
respectively. Also in the original state, the voltages of 21 nodes
candidate DGs are listed in Table 3. In addition, the number of
are lower than the desirable limit (0.95 per unit). The lines that
controllable loads or customers at each bus, for the both systems
cannot participate in DSR are L4, L5, L21, L24, L26 and L27. In
under analysis (Civanlar and Baran), is brought in Table 4. More
this system, finding the best configuration and DG sitting and
data about the systems are available in [17]. The proposed method
sizing problems are simultaneously implemented supposing the
is implemented in MATLAB software environment, and is solved
installation of four DG with storage systems, according to Table 3.
via using a laptop computer with CPU-2.3 GHz Core i5 and 4 GB
For the Baran system, all the produced topologies during the
RAM.
simulation are shown in Fig. 7. The numerical results associated
Suppose we want to install three DG with the storage systems in
with the optimal solution are brought in Table 6. The smallest
the Civanlar test system with the sizes according to Table 3. In
values of losses and ENS are related to Topologies 1 and 10,
this regards, Fig. 5 shows all the topologies that are obtained
respectively. In these topologies, both the power losses and ENS
during the simulation. The triangles represent the results of
drop to 74.38 and 43.05%, respectively, as compared to the
NSGA-II. According to this figure, there are five optimal solutions
original system. Moreover, the nodal voltages in all the optimal
for the Civanlar system. Table 5 presents the numerical results of
topologies are within the allowable range.
these solutions. Each of them can be a solution for this problem
since there is no absolute superiority between them. The difference
between the topologies from the viewpoints of configuration, 5.3 Costs associated with DG
locations and sizes of DG with storage systems in each optimal
solution demonstrate their simultaneous effects on improving the When the size of DG is optimised, it is better to consider the costs of
system performance. According to Table 5, Topology 1 with associated with DG (capital and installation, operation and
95.01 KW presents the smallest loss value. However, the smallest maintenance (O&M) and fuel). As of this, in this section, the costs
value of ENS is related to the Topology 5 with 45,490 KWh/year. of DG are considered as objective function, instead of losses, and
the simulation is repeated for ENS and cost. The kind of DGs is
photovoltaic (PV) and the costs associated with PV is calculated
as follow [24]
Table 5 Numerical results of pareto front for the Civanlar system
Configuration (the Location (bus) & size Losses ENS C(p) = a + b × p (12)
tie line number) of DG and storage (KW) (KWh/  
system (KW) year) Capital cost $/KW × Capacity (KW) × Gr
a= (13)
1 11 – 17 – 18 9 & 5000 – 11 & 5000 95.01 51,660 Life time (year) × 365 × 24 × LF
– 14 & 5000
2 11 – 17 – 18 11 & 5000 – 14 & 5000 98.6368 51,435 b = Fuel cost ($/KWh) + O&Mcost($/KWh) (14)
– 18 & 5000
3 8 – 11 – 18 11 & 5000 – 14 & 5000 100.9580 49,935
– 18 & 5000 where LF and Gr are the load factor and the annual interest rate,
4 10 – 11 – 17 7 & 5000 – 11 & 5000 124.6759 47,390 respectively. The value of LF and Gr are 1 and 0.03, respectively.
– 14 & 5000 Also for rated capacity of 100 KW and for lifetime of 20 years,
5 10 – 17 – 19 7 & 5000 – 11 & 5000 130.6838 45,490
– 14 & 5000
the capital cost and O&M cost are 6675 $/KW and 0.005 $/KWh,
respectively [24]. It should be noted that, PV will not have a fuel

IET Gener. Transm. Distrib., pp. 1–8

6 & The Institution of Engineering and Technology 2015
Fig. 6 Baran test system

Fig. 7 All of the obtained topologies during the simulation of the Baran system

cost as no fuel is required. Like the previous section, all the 5.4 Results analysis
produced topologies during the simulation and the numerical
results of the Baran test system are brought in Fig. 8 and Table 7, Table 8 presents the values of the active power losses and the ENS
respectively. between three scenarios for the Civanlar test system. In the first
scenario, the DG with storage system allocation and sizing is
Table 6 Numerical results of pareto front for the Baran system performed and the lowest values of power losses and ENS are
obtained from pareto optimal front. Then for this system, only
Configuration (the Location (bus) & size Losses ENS reconfiguration is performed. In this scenario, the locations and
tie line number) of DG and storage (KW) (KWh/ sizes of DGs and storage systems are fixed and have been derived
system (KW) year)
from one of the solution of the pareto optimal front in the last
1 11 – 28 – 32 – 33 – 34 8 & 700 – 14 & 700 – 51.9078 5109.4
scenario. Finally, the DSR and DG with storage system allocation
25 & 700 – 32 & 700 and sizing are simultaneously performed on the system. It can be
2 11 – 27 – 32 – 33 – 34 8 & 700 – 14 & 700 – 52.4258 5010.4
25 & 700 – 32 & 700
3 11 – 17 – 27 – 33 – 34 8 & 700 – 14 & 500 – 55.1761 4860.9
25 & 700 – 32 & 700
4 7 – 8 – 14 – 17 – 27 10 & 700 – 24 & 700 56.2246 4675.1
–
25 & 700 – 32 & 700
5 7 – 11 – 14 – 17 – 27 8 & 700 – 24 & 700 – 56.6934 4222.9
25 & 700 – 32 & 700
6 11 – 14 – 16 – 27 – 33 8 & 700 – 24 & 700 – 57.908 4127.4
25 & 700 – 32 & 700
7 7 – 11 – 14 – 17 – 37 8 & 700 – 24 & 700 – 62.8877 4125.4
25 & 500 – 32 & 700
8 11 – 14 – 17 – 33 – 37 8 & 700 – 24 & 700 – 67.9177 4035.1
25 & 500 – 32 & 700
9 11 – 14 – 16 – 33 – 37 8 & 700 – 24 & 700 – 68.5638 4022.5
25 & 500 – 32 & 700
10 11 – 14 – 16 – 33 – 37 24 & 700 – 25 & 500 73.3432 3909.5
–
30 & 700 – 32 & 700 Fig. 8 All of the obtained topologies during the simulation of the Baran
system for the cost and ENS optimisation

IET Gener. Transm. Distrib., pp. 1–8

& The Institution of Engineering and Technology 2015 7
Table 7 Numerical results of pareto front for the Baran system for cost with the optimal DG allocation and sizing problem, which is
and ENS optimisation solved by the NSGA-II algorithm. This solution approach allows
Configuration (the tie Location (bus) & size of Cost ENS the losses, ENS and costs of each topology to be separately
line number) DG and storage system ($) (KWh/ optimised under specific loads and constraints. The proposed
(KW) year) algorithm is applied on two test systems and the effectiveness of
this algorithm is then verified by the analysis of the results. The
1 11 – 14 – 16–27 – 33 24 & 300 – 25 & 300 – 52,561 4719.2 best topologies are shown in the frame of a set (front). As well,
31 & 300 – 32 & 300
2 11 – 14 – 16–27 – 33 24 & 500 – 25 & 300 – 61,322 4368.9
the proposed algorithm can be extended to optimise other
31 & 300 – 32 & 300 objectives like either the system parameters or reliability indices.
3 11 – 14 – 16–27 – 33 24 & 500 – 25 & 500 – 70,082 4053.6 This feature can be useful for planners to improve the performance
31 & 300 – 32 & 300 of distribution systems.
4 11 – 14 – 16–33 – 37 24 & 500 – 25 & 500 – 87,602 3909.5
30 & 700 – 32 & 300

7 References

Table 8 Comparative analysis for the Civanlar system 1 Jabr, R.A., Singh, R., Pal, B.C.: ‘Minimum loss network reconfiguration using
mixed- integer convex programming’, IEEE Trans. Power Syst., 2012, 27, (2),
Scenario Losses ENS (KWh/ pp. 1106–1115
(KW) year) 2 Romero-Ramos, E., Exposito, A.G., Santos, J.R., et al.: ‘Path-based distribution
network modeling: application to reconfiguration for loss reduction’, IEEE
Trans. Power Syst., 2005, 20, (2), pp. 556–564
DG (along with the storage system) allocation 114.0516 53,130 3 Khodr, H.M., Martinez-Crespo, J., Matos, M.A., et al.: ‘Distribution systems
and sizing without reconfiguration reconfiguration based on opf using benders decomposition’, IEEE Trans. Power
reconfiguration without DG (along with the 114.2364 45,490 Deliv., 2009, 24, (4), pp. 2166–2176
storage system) allocation and sizing 4 Taylor, J.A., Hover, F.S.: ‘Convex models of distribution system reconfiguration’,
reconfiguration with DG (along with the 95.01 45,490 IEEE Trans. Power Syst., 2012, 27, (3), pp. 1407–1413
storage system) allocation and sizing 5 Merlin, A., Back, H.: ‘Search for a minimal-loss operating spanning tree
simultaneously configuration in urban power distribution systems’. Proc. 5th Power Syst. Comp.
Conf., Cambridge, U.K., September 1975, pp. 1–5
Losses and ENS are associated with the best values of losses and ENS in 6 Civanlar, S., Grainger, J.J., Yin, H., et al.: ‘Distribution feeder reconfiguration for
pareto optimal front loss reduction’, IEEE Trans. Power Deliv., 1988, 3, (3), pp. 1217–1223
7 Baran, M.E., Wu, F.F.: ‘Network reconfiguration in distribution systems for loss
reduction and load balancing’, IEEE Trans. Power Deliv., 1989, 4, (2),
pp. 1401–1407
seen in Table 8 that when these two problems are combined, the 8 Shirmohammadi, D., Hong, H.W.: ‘Reconfiguration of electric distribution
smaller values of losses and ENS are obtained. This table shows networks for resistive line losses reduction’, IEEE Trans. Power Deliv., 1989, 4,
the effect of simultaneous consideration of these two problems. (2), pp. 1492–1498
9 Gomes, F.V., Carneiro, S. Jr., Pereira, J.L.R., et al.: ‘A new distribution system
The analysis of simulation results for two test systems also shows reconfiguration approach using optimum power flow and sensitivity analysis for
that when DG and storage system are added to the system, the loss reduction’, IEEE Trans. Power Syst., 2006, 21, (4), pp. 1616–1623
voltage constraint is satisfied and power quality is improved. In 10 Nara, K., Shiose, A., Kitagawa, M., et al.: ‘Implementation of genetic algorithm for
the topologies that the DG and storage systems have been installed distribution systems loss minimum reconfiguration’, IEEE Trans. Power Syst.,
1992, 7, (3), pp. 1044–1051
to buses with high load demand, the ENS has smaller value. 11 Lucia, C., Borges, T., Ferreira, V.: ‘Multistage expansion planning for active
Moreover, the analysis of the optimal solutions resulted for either distribution networks under demand and distributed generation’, Int. J. Electr.
of the two systems demonstrates that the system configuration, and Power Energy Syst., 2012, 36, (1), pp. 107–116
location and size of DGs are related together and can influence on 12 Torres, J., Guardado, J.L., Rivas-Davalos, F., et al.: ‘A genetic algorithm based on
the edge window decoder technique to optimize power distribution systems
each other. It means that for different configurations, different reconfiguration’, Electr. Power Energy Syst., 2013, 45, pp. 28–34
optimal locations or sizes of DG with storage systems can be 13 Alshehri, A.M.: ‘Optimal reconfiguration of distribution networks using ant colony
found. Therefore, these problems should be simultaneously studied method’. Master thesis, College of Engineering Electrical Engineering Department,
for better operation of the system. King Saud University, 2007
14 Abdelaziz, A.Y., Mohammed, F.M., Mekhamer, S.F., et al.: ‘Distribution system
Generally, the number of solutions in the pareto optimal front reconfiguration using a modified tabu search’, Electr. Power Syst. Res., 2010,
depends on the kind of system, the number of DGs and storage 80, (8), pp. 943–953
systems, and the number of customers at DG-buses. Each point of 15 Olamaei, J., Niknam, T., Gharehpetian, G.: ‘Application of particle swarm
pareto optimal front can be a solution for the problem because optimization for distribution feeder reconfiguration considering distributed
generators’, Appl. Math. Comput., 2008, 201, (1–2), pp. 575–586
there is no absolute superiority between these points. Although 16 Rugthaicharoencheep, N., Sirisumrannukul, S.: ‘Feeder reconfiguration with
depends on the existing conditions, planners can choose some dispatchable distributed generators in distribution system by tabu search’.
answer of the pareto optimal front as the final topology. Generally Universities Power Engineering Conf. (UPEC), 2010, pp. 1–5
there are two ways for choosing the final solution, as follows: 17 Mendoza, J.E., López1, M.E., Coello Coello, C.A., et al.: ‘Microgenetic
multiobjective reconfiguration algorithm considering power losses and reliability
indices for medium voltage distribution network’, IET Gener. Transm. Distrib.,
† First, a range is specified on the axes as the desirable range. Then, 2009, 3, (9), pp. 825–840
after obtaining the optimal solutions from the pareto optimal front, 18 Niknam, T., Fard, A.K., Seifi, A.: ‘Distribution feeder reconfiguration considering
fuel cell/wind/photovoltaic power plants’, Renew. Energy, 2012, 37, (1),
the topology that is within the desirable range is chosen as the pp. 213–225
final topology. 19 Mazza, A., Chicco, G., Russo, A.: ‘Optimal multi-objective distribution system
† The topology that is located in the middle of the pareto optimal reconfiguration with multi criteria decision making-based solution ranking and
front is chosen as the final topology. For example in the Civanlar enhanced genetic operators’, Electr. Power Energy Syst., 2012, 54, pp. 255–267
20 Moravej, Z., Adelnia, F., Abbasi, F.: ‘Optimal coordination of directional
system, topology 3 can be selected as the final topology. overcurrent relays using NSGA-II’, Electr. Power Syst. Res., 2015, 119,
pp. 228–236
6 Conclusion 21 Billinton, R., Allan, R.N.: ‘Reliability evaluation of power systems’ (Plenum
Publishing, New York, 1996, 2nd edn.)
22 Deb, K., Pratap, A., Agarwal, S., et al.: ‘A fast and elitist algorithm: NSGA-II’,
Losses and ENS and costs associated with DGs are three important IEEE Trans. Evol. Multiobject. Genet. Comput., 2002, 6, (2), pp. 182–197
parameters in distribution systems that the planners are trying to 23 Jensen, M.T.: ‘Reducing the run-time complexity of multiobjective EAs: the
reduce them. In this paper, at first, an approach for ENS NSGA-ii and other algorithms’, IEEE Trans. Evol. Comput., 2003, 7, (5),
pp. 503–515
calculation in the presence of DGs and storage systems is 24 Niknam, T., Taheri, S.I., Aghaei, J., et al.: ‘A modified honey bee mating
proposed. Then, the problem of minimising of two of above optimization algorithm for multiobjective placement of renewable energy
parameters is established via simulating of DSR problem along resources’, Appl. Energy, 2011, 88, (12), pp. 4817–4830

IET Gener. Transm. Distrib., pp. 1–8

8 & The Institution of Engineering and Technology 2015