www.ietdl.

org
Published in IET Generation, Transmission & Distribution
Received on 19th June 2013
Revised on 1st November 2013
Accepted on 14th November 2013
doi: 10.1049/iet-gtd.2013.0423

ISSN 1751-8687

Optimal distributed generation placement in shunt

capacitor compensated distribution systems
considering voltage sag and harmonics distortions
Soma Biswas1, Swapan Kumar Goswami2, Amitava Chatterjee2
1
Department of Electrical Engineering, JIS College of Engineering, Kalyani, West Bengal 741235, India
2
Department of Electrical Engineering, Jadavpur University, Kolkata, West Bengal 700032, India
E-mail: [email protected]

Abstract: The present study proposes a method of solving the distributed generations (DGs) placement problem by considering
multiple aspects of a power system operation. In addition to the commonly considered objectives of reduction of the loss and
improvement of the voltage profile, this study has optimised other power quality related objectives such as minimisation of
the voltage sag and harmonic distortion. A new formulation of a composite, constrained objective function is put forward by
considering objectives such as the cost of the power losses, the cost of the DGs and the cost of loss of load because of the
voltage sag and the constraints such as line flow limits, number/size of the installed DGs and the power quality limits of the
standard IEEE-519. The system under consideration is a complex one consisting of both linear and non-linear loads as well as
the power factor correcting capacitors. The effect of the non-linear harmonic generating loads and the compensating
capacitors on the penetration of the DGs in the distribution system is investigated. This optimisation problem is solved for
several distribution systems by using several metaheuristic optimisation techniques. However, detailed results are presented on
a benchmark IEEE 33 bus radial distribution system using genetic algorithm to demonstrate the effectiveness of the proposed
method. A comparative performance analysis of various metaheuristic optimisation techniques is also presented to show the
applicability of different optimisation techniques in solving the proposed optimisation problem.

1 Introduction The aim of DG planning is to find the feasible types, sizes,

locations and configurations of the DGs by considering the
Distributed generation (DG) is generally defined as power network constraints and the economic concerns. Proper
generation through relatively small units (from a few kW up optimisation tools are capable of discovering the optimal
to 10 MW) and is normally placed near the load centre. Gas allocation of the DG units, reducing the marginal costs and
turbines, reciprocating engines, fuel cells, solar cells, wind satisfying the technical constraints.
turbines and microturbines are some of the practical DG Several researchers have been addressing this problem in
technologies [1, 31]. Apart from lowering the environmental recent times in different corners of the world. A
effect, when compared with the traditional fossil and nuclear multi-objective index-based approach to optimally determine
sources of power, the DGs have multiple different usefulness the size and the location of the multi-DG units in the
from the perspectives of the endusers, the utilities distribution distribution system with the non-unity power factor
companies and the commercial power producers. Despite considering different load models has been proposed in [2].
these advantages there are still several disadvantages of DG Similarly, determination of the DG penetration considering
planning. Stability problems, complex protection strategies, the voltage limits have been discussed in [3, 33]. In recent
power quality, cost of operation and maintenance, long-term years, non-gradient-based evolutionary optimisation
reliability of the units, interconnection and the islanding techniques have also found popular applications in addressing
problem are some of the issues to be studied carefully while similar problems. Several optimisation techniques such as
placing new DG units. The issue of congestion management genetic algorithm (GA), artificial bee colony (ABC) and
in the distribution networks by considering a methodology particle swarm optimiser (PSO) with passive congregation
for generation curtailment because of the voltage constraints (PSOPC) and so on have been proposed for the optimal
is addressed in [32]. In recent times, one problem that has placement of the DG in [4–6]. It has been demonstrated in
gained paramount importance is that of the power quality. [7] that the DG can also reduce the voltage sag problem. To
This is a power problem manifested in the magnitude and address another important issue, the placement of the DG
the waveform of the voltage and the current deviations from satisfying the harmonic limits has been discussed in [8].
the prescribed ones that may ultimately lead to the failure A comprehensive review of the state-of-the-art approaches
or the maloperation of the consumer equipments. employed in multi-objective distributed energy resources

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 783
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org
(DER) planning has been presented in [26] which discuss the the harmonics which cause a variety of undesirable
challenges, the trends and the latest developments in this field, phenomena on the cables, generators, transformers and
at length. In [27], the problem of optimal sizing and siting of other loads. Hence, these two interrelated problems are
the embedded generations (EG) has been formulated as a considered together in the present paper, mainly while
constrained, multi-objective and non-differentiable placing the DGs, and suitable methods of analysing these
optimisation problem by applying the double trade-off PQ problems have been discussed in the next subsection.
method. Here, the authors have included the harmonic
distortion issues in the objective functions and have also
2.1 Voltage sag analysis
considered the problem of maintaining the voltage profile.
However, achieving the solutions for these problems by Owing to the rapid growth in the applications of the power
taking the major power quality issues of voltage sag and electronics in the commercial and industrial sectors, voltage
harmonic distortions simultaneously into account remains sags have become one of the major power quality concerns
an open challenge. Thus, the present researchers intend to in recent years. The sensitive power electronics loads such
address this issue with a new approach for problem as adjustable speed drives, process control equipments and
formulation. In [17, 25], attempts were made to reduce the computers can be easily interrupted by voltage sag and the
voltage sag with simultaneous reduction of the loss and the resulting significant production and financial losses. The
cost of the DG. In that paper, harmonic producing loads, individual load equipment sensitivity to the voltage sags
that is, non-linear loads and power factor correcting can be obtained from the manufacturer, the standards
capacitors were not considered. available or through laboratory tests [9]. Generally, the
In this paper, a new formulation of an optimisation problem information technology industry council and the computer
is proposed for simultaneous power quality improvement with business equipment manufacturers association curves are
optimal placement of the DG in the radial distribution well known for providing the voltage-tolerance
networks, comprising of non-linear loads and a power characteristics of the loads [10, 11].
factor correction capacitor. The technical factors such as The main causes of the voltage dip are faults, overloads,
reduction of the loss, improvement of the voltage, reduction short circuits, transformer switching, starting of large
of the voltage sag problem and reduction of the harmonics, motors and so on. The parameters that decide the
along with reduction of the installation cost, are considered characteristics of the voltage sag are magnitude, duration,
as various important aspects of the DG placement problem. unbalance and phase angle jump. These parameters again
An IEEE 33 bus system has been considered as the test depend on the location as well as the type of fault in the
system. The entire algorithm is implemented in MATLAB system.
environment. Several stochastic prediction methods such as method of
The rest of the paper has been structured as follows; critical distance, method of fault position, Monte Carlo
Section 2 describes the analysis method for two key power method and analytical approach [12–14] have been
quality issues namely voltage sag and harmonic distortion. proposed in the literature to compute the sag. The voltage
In Section 3, the problem formulation is discussed in great sag can be determined either as the voltage drop or as the
detail that leads to the construction of the objective remaining voltage at the PCC (point of common coupling).
function. Section 4 discusses how the GA can be suitably In this paper, the voltage sag magnitude refers to the
utilised to solve this problem. The method has been remaining voltage at the PCC.
implemented on a radial distribution system and the results How often the customers experience sags, that is, the sag
are discussed in Section 5. Section 6 shows the applicability frequency is also necessary to be estimated. A probabilistic
of the proposed model to the other optimisation techniques approach, coupled with the network reliability data [15] are
and their detailed comparative study. Finally, Section 7 often used for this purpose. The network reliability data
shows the conclusion. considers permanent, active and temporary failure rates
merged into a single index and thus presents the component
2 System modelling for the power quality failure rate.
issues In the present paper, the sags because of the short-circuit
faults are mainly considered. The system fault statistics
Voltage sag and harmonic distortion are very important power comprising of the bus fault and the line fault rates are
quality problems that lead to the magnitude distortion and the shown in Table 1. A total number of 56 faults including
waveform distortion phenomenon. The modern power bus and line faults have been considered. The number of
electronic devices have evolved as one of the most SLGFs, LLGFs, LLLGFs and LLGs are computed as
important power quality problems in a distribution network. shown in Table 1.
On the one side these are very sensitive to the voltage To calculate the voltage dip for the balanced and the
disturbances such as voltage sag or dip and on the other unbalanced type of faults simple short-circuit analysis or
hand these cause significant waveform distortion, that is, symmetrical component analysis are performed [16].

Table 1 System fault rates

Type of Fault type Bus fault rate, Average fault rate, Line fault rate, Average fault rate, Exact number
fault distribution, % events/year events/year events/100 km/year events/100 km/year of faults

LG 70 3 2.1 25 17.5 39
LLG 17 0.51 4.25 10
LL 7 0.21 1.75 4
LLLG 6 0.18 1.5 3
Total 56

784 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org
Equations (1)–(4) show the general expressions for the (4) Repeat the same process for the other faults and add all
during-fault phase voltages (sag magnitude) at a random the individual loads disturbed to obtain the total load
bus m, because of the four types of faults at location K on a disturbed.
line. The derivation has been shown in Appendix 1. (5) Finally, compute the total cost of the load disturbed
associated with the faults.
(a) Three-phase faults(3PF):

1
ZmK 2.2 Harmonic power flow (HPF) analysis
Vmfault = Vmpref − 1
VKpref (1)
ZKK To determine the effect of the harmonics on the distribution
network, the magnitude of the harmonic currents flowing in
(b) Single line-to-ground faults (SLGF): all the elements of the power network must be calculated
which needs the HPF solution.
0
ZmK + bZmK
1
+ gZmK
2 To perform the HPF some assumptions are made as
,m = bVw,m −
Vwfault pref
VKpref (2) follows: (i) positive, negative and zero sequence harmonic
0
ZKK + ZKK
1 + Z2
KK effects can be considered separately; the distribution lines
are modelled by using appropriate sequence network
(c) Line-to-line faults (LLF): circuits, (ii) the linear loads can be modelled by using
simple equivalents, (iii) the non-linear loads can be
1
ZmK − gZmK
2 modelled by using current injections at appropriate
,m = bVw,m −
Vwfault pref
1 + Z2
VKpref (3) harmonics, (iv) the series impedance of the feeder and the
ZKK KK substation at the harmonic frequencies can be determined
from the fundamental frequency values of the resistance and
(d) Double line-to-ground faults (LLGF): the reactance, (v) the skin effect and the line charging
capacitance can be neglected, (vi) the effects of the phase
angle diversity between the distributed harmonic sources
, m = Vw, m
Vwfault pref


1  2
1  0  can be ignored, (vii) the load apparent volt-amperes at the

bZmK − ZmK
0
ZKK + bZmK − gZmK
2
ZKK pref non-linear load buses will be known and (viii) the type of
− 1 + Z1 Z2 + Z2 Z0
VK
0
ZKK ZKK KK KK KK KK
non-linearity is known.
The first step in the development of the HPF program is to
(4)
establish appropriate mathematical models for all the power
system components. In a multi-branch interconnected
where β = 1 and γ = 1 for the j = A-phase; β = α 2 and γ = α network, an admittance matrix Y is formed from the
for the j = B-phase; β = α and γ = α 2 for the j = C-phase; individual elements, for each particular harmonic frequency h.
,m is the pre-fault voltage of the mth bus at phase j
and Vwpref Hence, a harmonic study for the system is implanted and
(j is A, B or C ). consequently the bus harmonic voltages, the line harmonic
currents and the voltage and the current distortion factors
012
ZmK are the sequence transfer impedances corresponding to are calculated and compared with the corresponding limits
012
buses m and K, ZKK are the sequence driving point set by the standards. The limits of the allowable voltage
impedances at bus K and α is the complex number operator, and the current harmonic distortions are set by the IEEE,
0
e j120 . the IEC, the EN and the NORSOK standards. Many
The economic consequences are assessed by multiplying countries also have their own standards in this regard.
the sag frequency and the cost by the number of customers
in [15]. The threshold voltage of the equipment in this 2.2.1 Current source model: To model the effect of one
paper is considered as 75%, that is, below this voltage the or more harmonic sources, the most common modelling
equipment suffers from the voltage dip problem. For each technique is to model all the harmonic generating loads or
event of the fault, the loads are computed in terms of MVA/ non-linear loads as constant-current sources. The non-linear
kVA which have voltage below the threshold value. The loads which are assumed as harmonic generating loads are
cost of the outage is considered as 4 US$/kVA as per Jussi considered as the 6 pulse converter, the 12 pulse converter
et al. [24]. and the fluorescent type. The harmonic current spectrums
Thus, the method of analysing the voltage sag can be for these types of loads are presented in Table 2 [18].
summarised in the following steps:
2.2.2 Modelling of the transmission lines: It is
(1) Fix a time period. Compute the length of each line of the assumed that the three-phase radial distribution system is
radial system under consideration. Calculate the total number balanced and can be represented by an equivalent
of faults and the number of each individual fault from Table 1. single-phase system. Thus, in the HPF, three-phase
Choose their locations. Set the fault impedance to zero and transmission lines are modelled by a single-phase
the pre-fault voltages of all the buses at 1 pu. pi-equivalent with the correct phase sequence, for both the
(2) Form ZBus of the radial distribution system from the line fundamental and the harmonic frequencies [18]. This model
and the bus data, as given in Table 3 in the next section. is also used for the conventional power flow (CPF). The
(3) Calculate the bus voltages for a fault, by using (1)–(4). If harmonic representation of the transmission line is given by
the fault occurs inside a line then increase the number of
buses by one considering the fault point as a bus. Then, ( f)
identify the buses where the voltages are below the Z
L = (R( f ) + jvL( f ) )dohm
threshold value. Add the loads connected at those buses to ( f)
compute the load disturbed for this particular fault. Y
L = jvC ( f ) dmho

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 785
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org
Table 2 Different types of non-linear loads and their injected harmonic currents (in %)
harmonics order 1st 5th 7th 11th 13th 17th 19th 23rd 25th 29th 35th 37th 41st
6 pulse converter 100 20.0 14.3 9.1 7.7 5.9 5.3 4.35 4 3.45 2.86 2.70 2.4
12 pulse converter 100 1.8 1.6 6.6 5.4 0.33 0.3 1.5 1.3 0.25 0.20 0.8 0.4
1st 3rd 5th 7th 9th
fluorescent lamp 100 10.0 7.0 0 0

where all the parameters are represented by their normally corresponding slips. In the presence of the harmonics, the
used convention. harmonics of the multiple of three represents the zero
sequence harmonics and has infinite impedance. h = 3n + 1
(1, 4, 7, 10, 13, 16 and 19) represents the positive sequence
2.2.3 Load model: Static load: By using the specified harmonics and h = 3n − 1 (2, 5, 8, 11, 14, 17, 20 and 23)
values of the real and the reactive power demands at 50 Hz, represents the negative sequence harmonics. The motor
P50 and Q50, the load can be represented by a series impedances for modelling the induction motor loads are
connection or a parallel connection of the resistive and the calculated by using the model given in [18].
reactive elements based on the model presented in [18].
Then, the resistance and the reactance of the series load and 2.2.4 Power factor correction capacitor model: The
the parallel load models can be evaluated for each power factor correction capacitor is modelled by a shunt
frequency and are, respectively, given as variable admittance and can be placed either at the terminal
bus of a transmission line or in the middle of a long line.
1 1 Considering the capacitor without losses, the admittance
ZS (h) = RS + jhXS ; YP (h) = − jh
RP XP only has its imaginary component and it can take values in
a specified range (usually between 0 and the maximum
where R, X are calculated knowing the voltage and the power capacity studied). The equivalent reactance is calculated by
rating of the buses and h is the order of the harmonics. using the base voltage and the VAr rating of the capacitors
Induction motor type load: A Δ- or Y-connected motor at the fundamental frequency first. Then, the reactance
has infinite zero-sequence impedance. The positive and value at different frequency is modified by multiplying it
the negative sequence impedances are the functions of the with the order of the harmonics.

Table 3 Load data, line data and power generation of the capacitors in the 33-bus IEEE test system
Bus Linear load Non linear Device Capacitor, Line Positive/negative sequence Zero sequence
no. data load data type kVAr impedance impedance
(P, Q) kW/ (P, Q) kW/kVAr From To (R, X ) Ω (R, X ) Ω
kVAr bus bus

1 0, 0 — — — — —
2 100, 60 — — — 1 2 0.0922, 0.047 0.1844, 0.094
3 90, 40 (IM) — — — 2 3 0.4930, 0.2511 0.986, 0.5022
4 — 120, 80 6 pulse — 3 4 0.366, 0.1864 0.732, 0.3728
5 60, 30 — — — 4 5 0.3811, 0.1941 0.7622, 0.3882
6 60, 20 — — — 5 6 0.819, 0.707 1.638, 1.414
7 200, 100 — — — 6 7 0.1872, 0.6188 0.3744, 1.2376
8 — 200, 100 6 pulse 7 8 0.7114, 0.2351 1.4228, 0.4702
9 60, 20 — — 250 8 9 1.03, 0.74 2.06, 1.48
10 60, 20 — — — 9 10 1.044, 0.74 2.088, 1.48
11 45, 30 — — — 10 11 0.1966, 0.065 0.3932, 0.13
12 60, 35 (IM) — — — 11 12 0.3744, 0.1238 0.7488, 0.2476
13 60, 35 — — — 12 13 1.468, 1.155 2.936, 2.31
14 — 120, 80 6 pulse — 13 14 0.5416, 0.7129 1.0832, 1.4258
15 60, 10 — — — 14 15 0.591, 0.526 1.182, 1.052
16 60, 20 — — — 15 16 0.7463, 0.545 1.4926, 1.09
17 60, 20 (IM) — — — 16 17 1.289, 1.721 2.578, 3.442
18 — 90, 40 6 pulse 250 17 18 0.732, 0.574 1.464, 1.148
19 90, 40 — — — 2 19 0.164, 0.1565 0.328, 0.313
20 90, 40 — — — 19 20 1.5042, 1.3554 3.0084, 2.7108
21 — 90, 40 6 pulse — 20 21 0.4095, 0.4784 0.819, 0.9568
22 90, 40 — — 250 21 22 0.7089, 0.9373 1.4178, 1.8746
23 90, 50 — — — 3 23 0.4512, 0.3083 0.9024, 0.6166
24 420, 200 — — — 23 24 0.898, 0.7091 1.796, 1.4182
(IM)
25 — 420, 200 6 pulse 250 24 25 0.896, 0.7011 1.792, 1.4022
26 60, 25 — — — 6 26 0.203, 0.1034 0.406, 0.2068
27 60, 25 — — — 26 27 0.2842, 0.1447 0.5684, 0.2894
28 60, 20 — — — 27 28 1.059, 0.9337 2.118, 1.8674
29 — 120, 70 6 pulse — 28 29 0.8042, 0.7006 1.6084, 1.4012
30 200, 600 — — — 29 30 0.5075, 0.2585 1.015, 0.517
31 150, 70 (IM) — — — 30 31 0.9744, 0.963 1.9488, 1.926
32 210, 100 — — — 31 32 0.3105, 0.3619 0.621, 0.7238
33 — 60, 40 6 pulse 250 32 33 0.341, 0.5302 0.682, 1.0604

786 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org
After completion of the component modelling the YBus loss to compute the cost of the energy losses within this
matrix is formed for different harmonic frequencies. Next, time period. The loss because of the harmonics is calculated
the current source modelling is carried out to establish the to be 0.056184 W which is very less compared with the
IBus matrix at different harmonic frequencies. Thus, once loss because of the fundamental frequency (283.22 kW).
the injected harmonic currents and the bus admittance Hence, it is neglected.
matrix are calculated, the bus harmonic voltages are
calculated from
3.1.2 Minimisation of the voltage sag problem: The
second objective of the DG placement problem is the
[VBus ]h = [Y Bus ]−1
h [I Bus ]h minimisation of the effect of the voltage sag measured in
terms of the total load disturbed (SDIST) because of the
The total harmonic distortion (THD) or the distortion factor
voltage sag.
(DF), in terms of the current and the voltage, is a popular
Mathematically, it can be expressed as
measure employed to indicate the harmonic distortion. Once
the harmonic voltages are calculated for different buses the
THD at different buses are computed based on the voltages 
NF
Minimise SDIST = LDISTi (8)
at different buses at normal condition by using the i=1
following formula
 ⎛ ⎞ where LDISTi is the load disturbed for the ith fault and NF is the
1  1
Vrms 2 total number of faults within a specified time duration. To
%THDV = Vh2 × 100 = ⎝ −1 ⎠ × 100 compare the effect of the sag performance with the loss
V1 h=2 V1 rms
effect, here also the equivalent financial loss is found as the
(5) product of the total load disturbed and the cost of the unit
load disturbed.
Similarly, the current distortion factor CDF can be measured
by using the performance index 3.1.3 Minimisation of the total cost of the DGs: The
 ⎛ ⎞ cost of the DG depends upon its type. The total cost of the
1  1
Irms 2 DGs will vary as the number and the size of the DG
%THDI = Ih2 × 100 = ⎝ −1 ⎠ × 100 changes. Thus, the third objective of the minimisation of
I1 h=2 I1 rms
the cost of the DG can be formulated as
(6)

NDG
Minimise CDG = Ksize PDGi (9)
3 Problem formulation in the presence of the i=1
DG and the non-linear loads
where CDG is the total cost associated with the DGs, PDGi is
3.1 Individual objective functions the size of the ith DG, NDG is the total number of the DG
connected and Ksize is the cost of the DG per kW.
A general multi-objective optimisation approach for the
hybrid electricity markets considering the cost 3.2 Operational constraints
characteristics of the DGs together with the minimisation of
the voltage sag problem, the harmonics and the line loss 3.2.1 Power flow constraints: At the fundamental
has been formulated to find the optimal location, the size frequency, the power flow constraints are represented by the
and the number of DGs. nodal active and reactive power equations as
Thus, the problem consists of three objective functions as
given follows. 
n

PG, i − PL, i − Vi Vm Yi,bm cos di − dm − ui, m = 0 (10)
m=1
3.1.1 Active power loss minimisation: One of the
significant utilisations of the DG installation in the 
n

distribution systems is to minimise the total active power QG, i − QL, i − b
Vi Vm Yi,m cos di − dm − ui,m = 0 (11)
loss. This objective is set as m=1

b
Minimise PLoss where Yi,m is the mutual element of the fundamental
frequency bus admittance matrix, θi, m is the angle of Yi,mb
,
where PG, i and QG, I are the generated active and reactive powers
at bus i and PL, i and QL, i are the active and reactive load
  powers at bus i.

l
PLoss = gk Vi2 + Vj2 − 2Vi Vj cos (di − dj ) (7) At the harmonic frequencies, the power flow equations are
k=1
represented by Baghzouz [37]
    
where l is the number of lines. The kth line connects buses i Y Bus h V Bus h = I Bus h (12)
and j. gk is the conductance of the kth line. Vi and Vj are the
voltages at buses i and j, respectively, and δi and δj are the where [YBus]h is the nodal admittance matrix of the network at
corresponding phase angles. The total energy loss is found the hth harmonic, [IBus]h is the injected node current vector at
by multiplying the power loss with the total time of the hth harmonic and [VBus]h is the node voltage vector at the
operation. Then, it is multiplied by the cost per unit energy hth harmonic.

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 787
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org
3.2.2 Line flow limits: The line flow limit specifies the and the size of each DG (size) is configured as
maximum power that a given transmission line is capable of
transmitting under certain given conditions. The limit can cost = f (number, location, size)
be based either on the thermal consideration for shorter
lines or on the stability considerations. The power flow 
NSC 
NF 
NDG

limit constraints of either side can be expressed as = Kloss LossK + KMVA LDISTi + Ksize PDGi
k=1 i=1 i=1

Sij ≤ Sij max 

NSC
 
(13) + Kflow d1K SK − SK max 
S ji ≤ S ji max k=1


Nb  
 
3.2.3 Bus voltage limit: The bus voltages remain within + Kvolt d2K Vk − Vk min / max 
an allowable narrow range of levels and they are expressed k=1
as an inequality constraint given in (13)  

Nb
 
+ Kharm d3K THDVi − THDmaxi 
Vi min ≤ Vi ≤ Vi max (14) i=1

Vimin and Vimax denote the lower and the upper limits of the 
Nb   
 
voltage at bus i. + Kl harm i d4K Vhi − Vhmax 
i=1 h=3, 5, 7

3.2.4 DG capacity limits: The DGs have a maximum and (18)

a minimum generating capacity beyond which it is not
feasible to generate electricity because of technical or where Kloss denotes the cost of the energy losses which is
economic reasons. The generating limits are specified as the considered as 0.08 US$/kWh, KMVA is the cost of the load
upper and the lower limits for the real and the reactive disturbed which is 4 US$/MVA, Kflow is the penalty
power outputs. In this paper, the DGs are considered to multiplier for the power flow limit violation, Kvolt is the
feed the real power only. The real power generation limits penalty multiplier for the voltage limit violation, Kharm is
the penalty multiplier for the voltage THD limit violation
PDG min ≤ PDG ≤ PDG max (15) and KL_harm_i is the penalty multiplier for the individual
voltage harmonics limit violation. δ1K, δ2K, δ3K and δ4K are
3.2.5 Harmonics limits: Two types of harmonic the binary flags, each reset to zero when the constraint is
constraints are considered; voltage total harmonic distortion satisfied and set to 1 when the constraint is violated.
and individual voltage harmonic limits. The distortion of It may be mentioned here that the modified objective
the voltage should be bounded by the maximum total function multipliers (i.e. Kloss, KMVA and Ksize) are all
harmonic distortion of the voltages (THDV) related to the actual cost values and thus the relevant costs
for the system are used for them.
THDV , THDmax (16) On the other hand, the penalty multipliers, that is, Kflow,
Kvolt, Kharm and KL_harm_i are used to penalise the objective
and THDmax is the standard value of the THD which is 5% function such that infeasible solutions are not selected by
here as per the IEEE 519 standards. Again, the optimisation procedure. Thus, their values are selected
to be much higher than the cost multiplier such that
Vh ≤ Vh max (17) whenever any set limit is violated the corresponding
objective function value becomes very high and hence a
Each individual harmonic voltage should not exceed the feasible solution is always selected as the optimal solution.
standard limits which is 3% here. The actual values of the constants penalty multipliers are
1012 each. The unit of each such penalty multiplier is so
chosen that the corresponding contribution made by its
3.3 Overall objective function associated term in the cost function is converted to the
The problem as formulated above is essentially a equivalent US$. For example, the unit of Kvolt is chosen as
multi-objective one with constraints. However, all the US$/pu volt.
objective functions may be converted in terms of the actual
costs and thus the overall problem may be represented as a 4 Problem solution employing GA
single objective optimisation problem expressing all the
objective functions in terms of the actual cost functions. For The problem formulated in Section 3 and the cost function
example, for the loss component, the cost of the losses may formulated in (18) is a mixed-variable problem comprising
be calculated in terms of the generation cost of the actual of some integer variables and some floating point variables.
power that is lost. Similarly, the cost for the voltage sag It is well known that solving such mixed-variable,
may be calculated by following [24] in terms of the cost of hyper-dimensional, non-linear optimisation problems may
interruption. become extremely difficult and, very often, almost
To form the objective function, the ideas of a well-known impossible by using a conventional optimisation technique.
weighting method for each individual objective and each In recent years, several metaheuristic optimisation
constraint violation and the penalty function approach have algorithms have been proposed to overcome such
been employed. Finally, the overall objective function (or limitations of the conventional optimisation techniques and
the cost function) designed to obtain the optimum number they have been increasingly employed in practice to solve
of the DGs (number), the location of each DG (location) such difficult optimisation problems. Our problem

788 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org
Each solution string-fitness is computed by using (18). The
higher fitness solution string has more probability to have
more copies. This copying procedure is called
Fig. 1 Chromosome encoding for the n number of DG
‘reproduction’. The ‘crossover’ is used for innovating the
solution strings. A mutation can help the solution strings to
formulated in (18) can be potentially solved by using many have a wider area of feasible solutions. After these three
such candidate algorithms and, in this paper, we have genetic operations, namely, reproduction, crossover and
chosen the GA, traditionally known as the oldest popular mutation, the new generation solution strings exist. These
variant of the evolutionary algorithms, to solve this new generation solution strings repeat the genetic operations
problem. The GA is a biologically inspired optimisation again and again till the feasible solution is reached.
algorithm that works with a population of solutions and In this paper, roulette wheel selection, which chooses
creates new generations (solutions) by appropriate genetic parents by simulating a roulette wheel with different sized
operators. The objective to be met is represented by the slots, each proportional to the fitness of each individual, is
appropriate fitness function. The GA employs parallel chosen. For the crossover operation, the one point and the
searching, robust searching and searching mechanisms scattered crossover mechanisms were tested in this paper.
based on the principle of natural evolution and has found The mutation mechanism used in this paper implied
applications in many areas [19]. generating a random gene number and flipping the bit
When applying the GAs to optimise the DG allocation and found at that position. Fig. 2 shows the entire flowchart of
sizing problem, an important aspect is the coding of the the proposed method.
potential solutions. The coded variables are the number of
the DG units for installation and their locations and sizes. 5 Implementation of the method for a radial
Fig. 1 shows a candidate solution vector SV that comprises distribution system
of two sub vectors L and S such that SV = [L T S T]T. The
sub vector L is called the location sub vector that contains An IEEE 33-bus radial distribution system [20] is considered
the bus numbers where the probable DGs are located and as a test system for implementation of the proposed method.
can be given as L = [l1, l2⋯li⋯ln]T. The size sub vector S The main feeder originates from the substation and passes
contains the sizes of the individual DGs and si denotes through different consumer loads. The laterals are
the size of the ith DG placed in location i. S is given as: connected to individual loads as shown in Fig. 3. The load
S = [s1, s2⋯si⋯sn]T. Hence, a composite candidate solution data and the line data are given in Table 3. Some changes
or chromosome is given as SV = [l1, l2⋯li⋯ln s1, s2⋯si⋯sn]T. have been incorporated in the original system for the
purpose of analysis. To perform the harmonic power flow
and to test the effect of the capacitor, a certain percentage
of loads are considered as non-linear load and a number of
capacitors, each with a fixed capacity, have been
considered, respectively. Besides, a few number of loads is
considered as induction motor type, instead of considering
all the loads as static ones.

5.1 System description

5.1.1 Static, induction and non-linear type load

model: Total peak power demand of the system having
static, induction and non-linear type of load is 4.5485
MVA, and the average power factor is 0.81. To demonstrate
the effect of the harmonics, 40% of the total load is
considered as of non-linear type in this problem. Among

Fig. 2 Flowchart for the proposed method Fig. 3 IEEE 33bus test feeder

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 789
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org
these 6 pulses, 12 pulse converter type and fluorescent type of values obtained from the fault statistics data. In these
loads are considered. The distribution of the different types of situations, these fault locations are assumed to be uniformly
loads has been shown in Table 3. During the peak loading distributed along the feeders. The exact fault locations
condition, the load diversity factor is considered as unity in considered are shown in Appendix 2.
the subsequent analysis. The time invariant loads are
considered while placing single or multiple DGs. 5.2 Performance evaluation
5.1.2 Summary of the network: A radial distribution To solve the DG placement problem, the determination of the
system having several lateral branches is considered as the DG location is considered as the main problem and the
test network having 33 buses, connected by 32 lines as determination of the sizes and the number of DGs is solved
shown in Fig. 3. The transmission lines are considered to as the slave problem. The harmonics generated by the large
be at equal voltage levels. The details about the line induction machines are limited by the existing standards
impedances and the load data have been given in [20]. All and are negligible. The harmonics offered by the converter
the data have been used on the per unit basis where the connected DGs are also significantly small because of the
base values of the power and the voltage are chosen as 100 utilisation of the advanced PWM technique. While solving
MVA and 11 kV, respectively. The sequence impedances this problem, the maximum DG size is limited to 50% of
for the short-circuit analysis are presented in the last two the peak demand. Hence, each DG size is varied from 500
columns of Table 3. The positive sequence and the negative kW to 2 MW, in steps of 100 kW. The maximum number
sequence impedances are considered as same. of DGs is limited to 10. The location of each DG may vary
from bus 2 to bus 33. The problem is subdivided into two
5.1.3 Distributed generator: Depending upon the input
cases.
energy sources, the DGs may be of different types. Hence,
their characteristics are also different. In this paper, the
diesel generator is used. Each DG source has a maximum 5.2.1 Without power factor correcting capacitor: In
capacity of about 50% of the annual peak load. The buses this case, the systems with static loads, induction motor and
having DGs are modelled as PQ bus in the load flow non-linear loads are considered. To compare the
analysis. As the present problem is very much concerned performance of the system with a capacitor less system the
with the presence of the harmonics, the non-linear capacitors are not included first. The effect of the DG on
characteristics of the DG technology cannot be avoided. As the system loss and the bus voltage during the normal
we are considering the diesel generators as the DG sources, condition is very well known to us. The present formulation
it is fair to assume that their harmonic content will be introduces two more aspects named sag and harmonics.
negligible. Also, it can be remarked that for the renewal Figs. 4 and 5 show the sag and the harmonic performance
type of DG sources, the use of advanced pulse width of the above mentioned distribution system under the
modulation (PWM) technique in DG installation, which is environment of the DG. Both the figures prove that these
very common now-a-days, typically reduces the harmonics power quality phenomena are very much influenced by the
contributed by it to a very low value [28]. The installation
and the maintenance cost of the DG and other related data
are given in [21]. The cost of the DG in the present paper is
considered as 300 US$/kW of DG.
5.1.4 Power factor correction capacitor: The role of
the capacitors in the transmission as well as in the
distribution system is very important. These capacitors are
used not only to improve the power factor but also to
improve the system voltage. Five capacitors having a
capacity of 250 kVAr have been considered in the present
formulation. The locations of these capacitors are kept fixed
at the buses as given in Table 2.
5.1.5 Fault data: A total duration of 10 years has been Fig. 4 Variation of the load disturbed with the variation of the
considered for the sag analysis. The number of faults within location of the DG
this period is calculated to be 56, as reported in Table 1.
Four different types of fault (3PF, LLGF, SLGF and LLF)
are assumed with a relative share of 7%:6%:70%:17%.
Different types of faults and their corresponding frequencies
of occurrences have been reported in Table 1. As it is
impossible to predict the exact locations of the faults, the
candidate fault locations are assumed by maintaining some
simple uniformity. For the SLGF fault, in case of lines
having short lengths, only one fault is considered for each
line and for lines having long lengths, two faults are
considered for each line. For each short line, the fault
location is considered at 50% of the line length, whereas,
for the long lines, the two faults in each line are considered
located at 25 and 75% of the line length. For the other
types of faults, this philosophy could not be followed as the Fig. 5 %THD of bus 9 for the DG less system and a system with a
number of fault occurrences would have exceeded the single DG located at different buses

790 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org
Table 4 Result of the N DG placement
Size, kW Location (bus number) and cost in million $

N=1 N=2 N=3 N=4 N=5

500 (10) 2.214 (30, 13) 1.668 (26, 14, 31) 1.502* (7, 14, 24, 30) 1.462* (6, 9, 14, 24, 30) 1.511*
600 (10) 2.176 (30, 13) 1.632 (13, 6, 30) 1.507 (8, 24, 30, 15) 1.508 (7, 14, 24, 26, 32) 1.622
700 (10) 2.146 (30, 12) 1.619 (24, 13, 30) 1.562 (13, 24, 6, 30) 1.585 (3, 7, 14, 24, 31) 1.766
800 (10) 2.123 (30, 11) 1.616* (14, 24, 30) 1.607 (24, 8, 15, 30) 1.7303 (6, 12, 19, 25, 30) 1.945
900 (9) 2.106 (30, 11) 1.638 (24, 11, 30) 1.657 (7, 13, 24, 30) 1.850 (13, 31, 6, 24, 2) 2.125
1000 (9) 2.093 (30, 10) 1.676 (24, 12, 30) 1.742 (2, 12, 24, 30) 2.022 (12, 24, 28, 4, 2) 2.353
1100 (9) 2.086 (30, 10) 1.733 (24, 30, 11) 1.8407 (3, 12, 24, 29) 2.184 (2, 10, 19, 24, 30) 2.498
1200 (9) 2.084* (30, 11) 1.812 (24, 29, 10) 1.972 (2, 11, 24, 30) 2.313 (2, 4, 10, 24, 30) 2.740
1300 (9) 2.088 (30, 8) 1.882 (24, 11, 29) 2.116 (24, 28, 9, 2) 2.512 (9,19,20,24,30) 3.031
1400 (9) 2.097 (30, 8) 1.972 (24, 9, 28) 2.272 (23, 30, 9, 2) 2.703 (2, 4, 9, 23, 30) 3.306
1500 (9) 2.111 (29, 8) 2.074 (24, 9, 29) 2.429 (2, 9, 24, 29) 2.886 (2, 9, 19, 24, 30) 3.391

Table 5 Result of the DG placement when both the number and the sizes are varied
No. of DG Size, kW Location (bus number) Cost in million $ %Savings Loss, kW Sag, MVA

No of DG — — 2.805 — 283.22 212

1 1200 9 2.084 25.75 170.45 136.63
2 800 30, 11 1.616 42.37 125.56 67.072
3 500 26, 14, 31 1.502 46.45 126.22 44.271
4 500 7, 14, 24, 30 1.462* 47.88 111.27 23.13
5 500 6, 9, 14, 24, 3 0 1.511 46.13 98.25 20.51

presence of the DG and are very much affected by their solution. In all the cases, the %THD remains within the
locations. permissible limit.
In Fig. 5, it is observed that for some particular positions, the
DG is sufficient to reduce the %THD of the bus below 5%. 5.2.2 With power factor correction capacitor: Several
Hence, one may comment on this result that a single DG is research papers have been reported so far on the optimal
sufficient to reduce the %THD to an acceptable value, if placement of the capacitors [22, 23]. The capacitors can
properly placed. The results of the N (varies from 1 to 5) DG aggravate the existing harmonic distortion. Although these
placements have been shown in Table 4. The numbers devices do not generate the harmonics, they can cause
marked by (*) in Tables 4–7 show the optimum cost of resonance conditions (both series and parallel) that can
operation for the corresponding DG placement problem. magnify the harmonic levels produced by the non-linear
From Table 4, the optimum among the optimum solutions loads. Fig. 8 compares the %THD at all the buses for the
of the DG placement can be obtained. It is observed that the system with the capacitor and the system without the
minimum cost of the operation is $1 461 700 which is capacitor situations. It has been observed that the harmonic
obtained when the number of DGs is four and each DG is distortions because of the presence of the capacitors are
of size 500 kW. The corresponding optimum locations are much higher than a capacitor less system. This in turn
bus numbers 7, 14, 24 and 30. At this instance, the system affects the optimal locations of the DGs. Therefore
loss is 111.27 kW and the total load disturbed is 23.13 traditional DG placement methodologies in the modern
MVA, shown in Table 5. distribution systems need to be reformulated in order to
Fig. 6 shows that as the number of DGs is increased, the include the harmonic resonance restrictions.
percent savings increases and becomes maximum when the In the present paper, a fixed number of capacitors having
number of the DGs is four and then it gradually decreases. fixed capacity, have been considered as the power factor
Fig. 7 shows the harmonic performance of the optimum correcting devices. It has been observed that at least three

Fig. 7 Variation of the %THD of different buses for different

Fig. 6 Variation of savings with the number of DGs number of DGs having optimium locations and sizes

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 791
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org

Fig. 9 Variation of the % THD at different buses (considering the

harmonic limits)
Fig. 8 Variation of the % THD at different buses for both the cases
with capacitor and without capacitor situation
in proper selection of the maximum size of each DG. At
any other load, the optimum operating condition may be
DGs are required for reducing the harmonic distortion of all obtained by solving the problem of scheduling the DGs
the buses below the specified limit. No solutions are found along with the power to be drawn from the main grid so as
where one or two DGs are sufficient to reduce the %THD to minimise the operating cost, the harmonic distortion and
below the specified limits. The results of the optimal the voltage sag issues.
placement problem considering the harmonic limits are
presented in Table 6.
Table 7 shows the optimum number, the size and the
locations of the DGs and the corresponding cost of the 6 Comparative evaluation of the various
operation, the percent savings, the loss and the load optimisation techniques in solving the problem
disturbed. The optimum cost for this system is less than the
capacitor less system as the loss and the sag both are The formulation of the problem is generic in nature and thus
improved because of the presence of the capacitor. may be solved by using many robust (general purpose)
Figs. 9 and 10 show the harmonic performance in terms of optimisation techniques and using evolutionary or swarm
the %THD and the individual harmonics magnitude. For each intelligence-based algorithms that have been proposed in
case, the standard limits are mantained. Fig. 11 shows the sag recent years. To demonstrate the applicability of other
performance of the optimum locations for different number of similar such algorithms from the same genre, we have also
DGs. It shows the pu voltage of each bus when a single line to solved the same problem by using some contemporary,
ground fault occurs at the middle of the line connecting bus 7 powerful metaheuristic optimisation techniques that have
and bus 8. been proved to be very useful in solving difficult,
The load in any system has both daily and seasonal multi-dimensional optimisation problems. The optimisation
variations. However, in our case study, for the planning techniques selected for the purpose of comparative study are
purpose, a peak load has been selected which will help us ABC [29,30] algorithm, PSO [34], differential evolution

Table 6 Result of the N (3, 4 and 5) DG placement for the system with the capacitors
Size, kW Three DG placement Four DG placement Five DG placement

Location Cost in million $ Location Cost in million $ Location Cost in million $

(bus number) (bus number) (bus number)

500 25, 15, 30 1.162 27, 33, 15, 25 1.1067* 26, 25, 15, 27, 31 1.144*
600 25, 15, 30 1.159* 27, 31, 15, 25 1.1372 7, 25, 16, 20, 32 1.301
700 25, 15, 30 1.185 31, 6, 15, 25 1.2094 26, 25, 32, 15, 5 1.419
800 25, 15, 30 1.24 31, 6, 15, 25 1.3413 15, 29, 25, 23, 20 1.547
900 25, 15, 30 1.322 31, 6, 15, 25 1.5129 25, 27, 21, 15, 31 1.856
1000 25, 15, 30 1.339 30, 19, 25, 15 1.7183 30, 25, 23, 20, 16 2.111
1100 25, 15, 30 1.382 5, 15, 25, 29 1.9169 19, 8, 31, 25, 15 2.395
1200 25, 15, 30 1.449 30, 25, 15, 21 2.1669 26, 25, 15, 19, 30 2.616
1300 25, 15, 30 1.536 28, 25, 6, 18 2.7376 16, 25, 28, 19, 3 2.813
1400 25, 15, 30 1.644 19, 15, 25, 32 2.6619 16, 25, 21, 28, 19 3.218
1500 25, 15, 30 1.772 19, 28, 16, 25 2.8501 12, 30, 2, 21, 19 3.345

Table 7 Result of the DG placement when both the number and the sizes are varied
No. of DG Size, kW Location (bus number) Cost in million $ %Savings Loss, kW Sag, MVA

No of DG — — 2.287 — 212.79 204.07

3 600 25, 15, 30 1.159 49.32 59.30 52.395
4 500 8, 15, 25,30 1.066* 51.61 50.09 29.94
5 500 6, 7, 15, 25, 30 1.125 50.81 39.64 25.28

792 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org

Fig. 10 Variation of the individual harmonics at different buses with the 4DGs at their optimal locations (considering the harmonic limits)

Fig. 11 Bus voltages in p.u. for an SLGF fault and the DGs are at their optimal locations (considering the harmonic limits)

Table 8 Parameters used in the different optimisation techniques

Maximum no. of generation = 100
Population size = 10

BBO PSO ABC DE

habitat modification probability = 1 elitism parameter = 2 colony size = 16 weighting factor

mutation probability = 0.005 Inertial constant = 0.3 no of food = 0.5
elitism parameter = 2 cognitive constant = 1 source = 8 crossover
lower bound of immigration probability = 0 social constant of swarm interaction = 1 limit = 5 constant = 0.5
upper bound of immigration probability = 1 social constant of neighbourhood interaction = 1
step size = 1 size of particle neighbourhood = 0
maximum immigration rate = 1
maximum emigration rate = 1

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 793
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org
(DE) [35] and biogeography-based optimisation (BBO) has been found that all these methods are almost equally
algorithm [36]. capable of solving the problems and the results found are
The implementations of all these methods are quite similar quite comparable. For lack of space, the detailed results are
to that reported for the GA except for the tuning of different not presented here and only the summary of the results is
method specific free parameters. The values of these presented here. For the single DG problem, all the methods
parameters for different methods are reported in Table 8. It produced identical results, which are 1200 kW DG at bus 9,

Table 9a Summary of the best results obtained from the different methods for without the capacitor case
No. of Optimum size, Location (bus Minimum cost in % Loss, Sag, Method
DG kW number) million $ savings kW MVA

No of – – 2.805 – 283.22 212

DG
1 1200 9 2.084 25.75 170.45 136.63 GA, BBO, PSO, ABC,
DE
2 800 30, 11 1.616 42.37 125.56 67.072 GA, BBO, PSO, ABC
3 500 6, 14, 32 1.495 46.7 126.58 42.512 BBO, DE
4 500 7, 14, 24, 30 1.462 47.88 111.27 23.13 DE, BBO,GA
5 500 6, 9, 14, 24, 30 1.511 46.13 98.25 20.51 GA, BBO, DE

Table 9b Summary of the best results obtained from the different methods for the with capacitor case
No. of DG Size, kW Location (bus number) Cost in million $ %savings Loss, kW Sag, MVA Optimisation techniques

No of DG – – 2.287 – 212.79 204.07

3 600 25, 15, 30 1.159 49.32 59.301 52.395 GA, BBO, PSO, ABC, DE
4 500 8, 15, 25 30 1.066 51.61 50.09 29.94 BBO, GA
5 500 6, 7, 15, 25 30 1.125 50.81 39.64 25.28 DE, GA

Table 10 Results of the various methods for the four DG optimal placement problem without the capacitor
Size, kW Location (bus number) and cost in million $

No. of DGs = 4

BBO PSO ABC DE

500 (7, 14, 24, 30) 1.462 (13, 25, 26, 30) 1.487 (14, 24, 26, 31) 1.472 (7, 14, 24, 30) 1.462
600 (7, 14, 24, 30) 1.495 (6, 12, 24, 30) 1.5301 (7, 13, 24, 30) 1.503 (7, 14, 24, 30) 1.495
700 (14, 25, 26, 31) 1.600 (6, 17, 25, 30) 1.6487 (6, 14, 24, 30) 1.580 (6, 13, 24, 31) 1.5902
800 (13, 24, 26, 30) 1.705 (7, 12, 25, 31) 1.7375 (7, 14, 24, 32) 1.716 (7, 13, 24, 32) 1.7193
900 (6, 14, 24, 30) 1.850 (6, 10, 25, 29) 1.8984 (6, 14, 25, 33) 1.923 (3, 12, 24, 30) 1.8792
1000 (6, 11, 23, 32) 2.107 (14, 19, 24, 30) 2.0619 (2, 9, 24, 31) 2.072 (4, 10, 24, 29) 2.0228
1100 (11, 19, 23, 30) 2.202 (2, 12, 24, 31) 2.2106 (2, 13, 24, 30) 2.187 (12, 19, 24, 30) 2.1717
1200 (10, 19, 24, 30) 2.323 (2, 12, 25, 30) 2.3562 (2, 10, 25, 28) 2.379 (2, 10, 24, 30) 2.3113
1300 (3, 12, 20, 30) 2.643 (2, 12, 23, 28) 2.556 (12, 24, 19, 29) 2.524 (2, 9, 24, 29) 2.4922
1400 (2, 4, 9, 27) 2.912 (11, 19, 23, 29) 2.7338 (12, 23, 2, 29) 2.736 (2, 9, 24, 29) 2.6794
1500 (11, 19, 23, 27) 3.041 (12, 19, 23, 30) 2.9722 (2, 32, 11, 24) 3.051 (2, 9, 23, 29) 2.9037

Table 11 Results of the various methods for the 4 DG optimal placement problem with the capacitor
Size (kW) Location (bus number) and cost in million $

No of DG = 4

BBO PSO ABC DE

500 (8, 15, 25, 30) 1.066 (7, 15, 25, 33) 1.0959 (30, 28, 15, 25) 1.128 (7, 16, 25, 30) 1.0729
600 (6, 15, 25, 31) 1.1194 (16, 25, 27, 33) 1.1781 (15, 25, 33, 26) 1.149 (8, 16, 25, 29) 1.153
700 (6, 16, 25, 32) 1.2289 (8, 16, 25, 31) 1.251 (25, 6, 16, 31) 1.226 (7, 15, 25, 29) 1.2219
800 (7, 15, 25, 30) 1.3281 (6, 16, 25, 31) 1.3621 (15, 25, 27, 30) 1.366 (7, 16, 25, 31) 1.3657
900 (6, 15, 25, 32) 1.5194 (5, 15, 25, 31) 1.5393 (25, 28, 3, 15) 1.567 (15, 25, 27, 31) 1.5531
1000 (4, 15, 25, 30) 1.6822 (15, 19, 25, 30) 1.7183 (15, 25, 30, 21) 1.772 (7, 16, 25, 30) 1.735
1100 (6, 17, 25, 29) 2.0584 (16, 22, 25, 31) 2.0973 (15, 25, 31, 6) 1.968 (6, 18, 25, 29) 2.1018
1200 (2, 16, 25, 29) 2.1108 (16, 22, 25, 29) 2.2691 (15, 25, 33, 2) 2.226 (2, 15, 25, 29) 2.0622
1300 (15, 24, 25, 29) 2.5575 (15, 24, 25, 33) 2.7442 (15, 25, 21, 30) 2.404 (6, 15, 25, 29) 2.530
1400 (3, 15, 25, 28) 2.6044 (4, 15, 25, 33) 2.8122 (15, 25, 22, 31) 2.810 (2, 16, 25, 29) 2.5684
1500 (2, 15, 25, 28) 2.7606 (15, 21, 25, 33) 3.1762 (15, 25, 19, 31) 2.890 (15, 19, 25, 31) 2.8851

794 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org
correction capacitors has less distortion of the harmonics.
However, the presence of the power factor correction
capacitors greatly enhances the harmonic distortion because
of the local resonance. The final conclusion that can be
drawn is that, the reduction in the effect of the harmonics in
the presence of the capacitors can be achieved by placing
the DGs at suitable locations. The future scope of this paper
will focus on simultaneous, suitable placement of the DGs
and the capacitor in the unbalanced distribution system in
such a fashion that the maximisation of the net savings is
achieved by mainitaining the improvements in the technical
performances stated above.

8 References
1 Ackermann, T., Anderson, G., Söder, L.: ‘Distributed generation: a
Fig. 12 Convergence characteristics of the different optimization definition’, Electr. Power Syst. Res., 2001, 57, (3), pp. 195–204
techniques 2 El-Zonkoly, A.M.: ‘Optimal placement of multi-distributed generation
units including different load models using particle swarm
optimisation’, IET Gener. Transm. Distrib., 2011, 5, (7), pp. 760–771
3 Chen, P.-C., Salcedo, R., Zhu, Q., et al.: ‘Analysis of voltage profile
without the capacitor. For the multiple DG problems, either problems due to the penetration of distributed generation in
very close or identical results were obtained, in each case. low-voltage secondary distribution networks’, IEEE Trans. Power
A summary of the best results obtained from the different del., 2012, 27, (4), pp. 2020–2028
methods are shown in Tables 9(a) and 9(b), for the cases 4 Singh, R.K., Goswami, S.K.: ‘A genetic algorithm based approach for
optimal allocation of distributed generations in power systems for
without and with the capacitors, respectively. Although the voltage sensitive loads’, ARPN J. Eng. Appl. Sci., 2009, 4, (2), pp. 78–87
detailed results for all the methods for different number of 5 Abu-Mouti, F.S., El-Hawary, M. E.: ‘Optimal distributed generation
DGs could not be presented, for the four DG problem, the allocation and sizing in distribution systems via artificial bee colony
corresponding results are presented in Tables 10 and 11, as algorithm’, IEEE Trans. Power Deliv., 2011, 26, (4), pp. 2090–2101
6 Aref, A., Davoudi, M., Seifi, A., Ganjkhany, I., Davoudi, M.: ‘Particle
the four DGs have been found to be the optimum number swarm optimization based method for optimal placement and
of DGs for the network under consideration. The sample estimation of DG capacity in distribution networks’, Int. J. Sci. Tech.,
convergence characteristics of the different optimisation 2012, 2, (7), pp. 486–491
techniques are shown in Fig. 12, for the problem of without 7 Macken, K.J.P., Bollen, M.H.J., Belmans, R.J.M.: ‘Mitigation of voltage
the capacitors, in the four DG placement problem. From the dips through distributed generation systems’, IEEE Tran. Ind. Appl.,
2004, 40, (6), pp. 1686–1693
results presented in Tables 9–11, a closer inspection reveals 8 Bhowmik, A., Maitra, A., Halpin, S.M., Schatz Joe, J.E.: ‘Determination
that, in many situations, more than one optimisation of allowable penetration levels of distributed generation resources based
technique evolves as the superior solution. However, in on harmonic limit considerations’, IEEE Trans. Power Deliv., 2003, 18,
Table 9, the GA could achieve the superior solution in four (2), pp. 619–624
9 Milanovic, J.V., Gupta, C.P.: ‘Probabilistic assessment of financial
out of the five cases, which aptly justifies the suitability of losses due to interruptions and voltage sags – Part I: the
application of this method in solving the problem under methodology’, IEEE Trans. Power Deliv., 2006, 21, (2), pp. 918–924
consideration. 10 Heydt, G.T., Ayyanar, R., Thallam, R.: ‘Power acceptability’, IEEE
Power Eng. Rev., 2001, 21, (9), pp. 12–15
11 Park, C.-H., Hong, J.-H., Jang, G.: ‘Assessment of system voltage sag
7 Conclusions performance based on the concept of area of severity’, IET Gener.
Transm. Distrib., 2010, 4, (6), pp. 683–693
A novel method is proposed in this paper for DG placement 12 Bollen, M.H.J.: ‘Understanding power quality problems; voltage sags
and interruptions’ (IEEE Press Series on Power Engineering,
and sizing in the presence of the harmonics generating Piscataway, NJ, 2000)
loads. A new, composite, constrained objective function is 13 Juarez, E.E., Hernandez, A.: ‘An analytical approach for stochastic
formulated by using the power quality issues under assessment of balanced and unbalanced voltage sags in large system’,
consideration where the objectives are presented in the form IEEE Trans. Power Deliv., 2006, 21, (3), pp. 1493–1500
14 Moschakis, M.N., Hatziargyriou, N.C.: ‘Analytical calculation and
of system loss, voltage sag and the cost of the DGs, and the stochastic assessment of voltage sags’, IEEE Trans. Power Deliv.,
constraints are formulated on the basis of factors such as 2006, 21, (7), pp. 1727–1734
the power quality limits proposed in the IEEE-519 standard 15 Heine, P., Pohjanherimo, P., Lehtonen, M., Lakervi, E.: ‘A method for
(i.e. maximum and minimum rms bus voltages, maximum estimating the frequency and cost of voltage sags’, IEEE Trans.
individual harmonic distortion and maximum THD), power Power Syst., 2002, 17, pp. 290–296
16 Park, C.H., Jang, G.: ‘Stochastic estimation of voltage sags in a large
flow limits and the allowable number of DGs (at each bus meshed network’, IEEE Trans. Power Deliv., 2007, 22, pp. 1655–1664
and for the entire feeder). The proposed method has been 17 Biswas, S., Goswami, S.K., Chatterjee, A.: ‘Optimum distributed
extensively simulated for an IEEE 33 bus radial distribution generation placement with voltage sag effect minimization’, Energy
system. Among all the constraint factors introduced in the Convers. Manage., 2012, 53, (1), pp. 163–174
proposed fitness function and the inclusion of the power 18 Wakileh, G.J.: ‘Power system harmonics-fundamentals, analysis and
filter design’ (Springer, 2001)
quality constraints in the fitness functions prevented the 19 Goldberg, E.: ‘Genetic algorithms in search, optimization & machine
occurrence of the harmonic parallel resonance. Our learning’ (Pearson Education, 2009)
extensive experimentations proved that the proposed 20 Acharya, N., Mahat, P., Mithulananthan, N.: ‘An analytical approach for
method applying DG lowers the voltage THD, lowers the DG allocation in primary distribution network’, Electr. Power Energy
Syst., 2006, 28, pp. 669–678
system losses, improves the voltage profile, improves the 21 Pipattanasomporn, M., Willingham, M., Rahman, S.: ‘Implications of
power quality indices and achieves greater annual benefits. on-site distributed generation for commercial/industrial facilities’,
It is observed that the system without the power factor IEEE Trans. Power Syst., 2005, 20, (1), pp. 206–212

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 795
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014
www.ietdl.org
22 Bhattacharya, S.K., Goswami, S.K.: ‘A new fuzzy based solution of the 37 Baghzouz, Y.: ‘Effects of nonlinear loads on optimal capacitor
capacitor placement problem in radial distribution system’, Expert Syst. placement in radial feeders’, IEEE Trans. Power Deliv., 1991, 6, (1),
Appl., 2009, 36, (3), pp. 4207–4212 pp. 245–250
23 Goswami, S.K., Ghose, T., Basu, S.K.: ‘An approximate method for
capacitor placement in distribution system using heuristics and greedy 9 Appendix 1
search technique’, Electr. Power Syst. Res., 1999, 51, (3), pp. 143–151
24 Jussi, A., Sami, R., Pekka, V., Pertti, J.: ‘Possibilities to improve
reliability of distribution network by intended island operation’, In [16], the three phase voltages at any bus m because of a
Int. J. Innovat. Energy Syst. Power, 2009, 4, pp. 22–43 single line ground fault at bus K have been given as
25 Biswas, S., Goswami, S.K.: ‘Genetic algorithm based optimal placement
of distributed generation reducing loss and improving voltage sag pref
0
ZmK + ZmK
1
+ ZmK
2
performance’. Proc. of Int. Conf. on Advances in Electrical & m = VA, m −
VA,fault VKpref (19)
Electronics, 2010, pp. 49–51
0
ZKK + ZKK
1 + Z2
KK
26 Alarcon-Rodriguez, A., Ault, G., Galloway, S.: ‘Multi-objective
planning of distributed energy resources: a review of the
state-of-the-art’, Renew. Sustain. Energy Rev., 2010, 14, (5), 2 pref
0
ZmK + a2 ZmK
1
+ aZmK
2
m = a VA, m −
VB,fault VKpref (20)
pp. 1353–1366 0
ZKK + ZKK1 + Z2
KK
27 Carpinelli, G., Celli, G., Mocci, S., Pilo, F., Russo, A.: ‘Optimization of
embedded generation sizing and siting by using a double trade-off
method’, IEE Proc. Gener. Transm. Distrib., 2005, 152, (4), pref
0
ZmK + aZmK
1
+ a2 ZmK
2
pp. 503–513 VC, m = aVA, m −
fault
VKpref (21)
28 Bracale, A., Caramia, P., Carpinelli, G., Russo, A., Verde, P.: ‘Site and ZKK + ZKK + ZKK
0 1 2

system indices for power-quality characterization of distribution

networks with distributed generation’, IEEE Trans. Power Deliv., where VA,pref
B, C, m is the pre-fault voltage of the mth bus at
2011, 26, (3), pp. 1304–1316 012
29 Karaboga, D., Basturk, B.: ‘On the performance of artificial bee colony
different phases, ZmK are the sequence transfer impedances
012
(ABC) algorithm’, Appl. Soft Comput., 2008, 8, pp. 687–697 corresponding to buses m and K, ZKK are the sequence
30 Biswas, S., Chatterjee, A., Goswami, S.K.: ‘An artificial bee driving point impedances at bus K and α is the complex
colony-least square algorithm for solving harmonic estimation 0
problems’, Appl. Soft Comput., 2013, 13, pp. 2343–2355 number operator, e j120 .
31 Harrison, G.P., Wallace, A.R.: ‘OPF evaluation of distribution network The above equation has been simplified as the following
capacity for the connection of distributed generation’, IEE Proc. Gener. equation
Transm. Distrib., 2005, 152, (1), pp. 115–122
32 Zhou, Q., Bialek, J.K.: ‘Generation curtailment to manage voltage
0
ZmK + bZmK
1
+ gZmK
2
constraints in distribution networks’, IET Gener. Transm. Distrib., , m = bVw, m −
Vwfault pref
VKpref (22)
2007, 1, (3), pp. 492–508
ZKK + ZKK + ZKK
0 1 2

33 Ayres, H.M., Freitas, W., De Almeida, M.C., Da Silva, L.C.P.: ‘Method

for determining the maximum allowable penetration level of distributed
where β = 1 and γ = 1 for the j = A-phase; β = α 2 and γ = α for
generation without steady-state voltage violations’, IET Gener. Transm. the j = B-phase; β = α and γ = α 2 for the j = C-phase; Vwpref
, m is
Distrib., 2010, 4, (4), pp. 495–508 the pre-fault voltage of the mth bus at phase j (j is A, B or
34 Niknam, T., Doagou-Mojarrad, H.: ‘Multiobjective economic/emission C), ZmK 012
are the sequence transfer impedances corresponding
dispatch by multiobjective-particle swarm optimisation’, IET Gener. 012
Transm. Distrib., 2012, 6, (5), pp. 363–377
to buses m and K, ZKK are the sequence driving point
35 Attia, A., EI-Fergany: ‘Optimal capacitor allocations using evolutionary impedances at bus K and α is the complex number operator,
0
algorithms’, IET Gener. Transm. Distrib., 2013, 6, (7), pp. 593–601 e j120 .
36 Bhattacharya, A., Chattopadhyay, P.K., ‘Application of biogeography
based optimization to solve different optimal power flow problems’,
IET Gener. Transm. Distrib., 2011, 1, (5), pp. 70–80 10 Appendix 2: see Table 12

796 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0423
 www.ietdl.org
Table 12 Number of different fault types and their locations
LG fault location LLG Fault location LL Fault position LLLG Fault position

Line/bus Distance Line information Distance Line information Distance Line information Distance
information w.r.t w.r.t. w.r.t. w.r.t.
from bus from bus from bus from bus
From bus To bus in % From bus To bus in % From bus To bus in % From bus To bus in %

1 2 25 2 3 25 11 12 50 4 5 50
1 2 75 5 6 50 15 16 50 11 12 50
2 3 25 8 9 50 24 25 50 24 25 50
2 3 75 11 12 50 32 33 50 Total 3
at bus 3 – – 14 15 50 Total 4
3 4 50 17 18 50
4 5 50 21 22 50
5 6 50 24 25 50
6 7 25 28 29 50
6 7 75 32 33 50
7 8 50 Total 10
8 9 25
8 9 75
9 10 50
10 11 50
11 12 50
at bus 12 – –
12 13 50
13 14 50
14 15 50
15 16 50
16 17 50
17 18 50
2 19 50
19 20 50
20 21 50
21 22 50
3 23 50
23 24 50
24 25 50
25 26 50
26 27 50
at bus 27 – –
27 28 50
28 29 50
29 30 50
30 31 50
31 32 50
32 33 50
Total 39

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 5, pp. 783–797 797
doi: 10.1049/iet-gtd.2013.0423 & The Institution of Engineering and Technology 2014