Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Procedia Computer Science 00 (2017) 000–000
Available online at www.sciencedirect.com
Procedia Computer Science 00 (2017) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Procedia Computer Science 125 (2018) 382–388

6th International Conference on Smart Computing and Communications, ICSCC 2017, 7-8
6th International Conference on Smart2017,
December Computing and Communications,
Kurukshetra, India ICSCC 2017, 7-8
December 2017, Kurukshetra, India
Optimal Distributed Generation Placement with Multiple Objectives
Optimal Distributed Generation Placement with Multiple Objectives
Considering Probabilistic Load
Considering Probabilistic Load
Ankit Uniyalaa, Ashwani Kumarb*
Ankit Uniyal , Ashwani Kumarb*
a
National Institute ofTechnology, Kurukshetra, 136119
b
a
National Institute ofTechnology, Kurukshetra, 136119
b
National Institute ofTechnology, Kurukshetra, 136119

Abstract
Abstract
In this paper, optimal DG sizing and siting has been obtained considering three objectives i.e. active power loss index (APLI),
linethis
In loading
paper,index
optimal(LLI)
DGand voltage
sizing deviation
and siting indexobtained
has been (VDI). Multiple objective
considering NSGA-II i.e.
three objectives method along
active with
power lossfuzzy
indexsatisfying
(APLI),
line
methodloading
(FSM) index
has (LLI) and to
been used voltage
solve deviation index (VDI).
multiple objective Multipleproblem.
optimization objectiveProbabilistic
NSGA-II method alonghas
load model with fuzzy
been satisfying
considered in
this study.
method A two
(FSM) haspoint
beenestimate
used to (2PEM) methodobjective
solve multiple has beenoptimization
used carry out o obtain
problem. probabilistic
Probabilistic load
load flowhas
model solution. The analysis
been considered in
has been
this study.carried
A twoout for estimate
point IEEE 33 (2PEM)
radial busmethod
test system. Theused
has been analysis
carryhas
outbeen carried
o obtain out using coding
probabilistic in MATLAB.
load flow solution. The analysis
has been carried out for IEEE 33 radial bus test system. The analysis has been carried out using coding in MATLAB.
© 2017 The Authors. Published by Elsevier Ltd.
© 2018 The Authors. Published by Elsevier B.V.
Peer-review
© under responsibility
2017 The Authors. Published byofthe scientific
Elsevier Ltd. committee of the 4th International Conference on Power and Energy Systems
Peer-review under responsibility of the scientific committee of the 6th International Conference on Smart Computing and
Peer-review
Engineering.
Communications under responsibility ofthe scientific committee of the 4th International Conference on Power and Energy Systems
Engineering.
Keywords: Distributed generation; indices; two point estimation method; probabilistic load flow.
Keywords: Distributed generation; indices; two point estimation method; probabilistic load flow.

1. Introduction
1. Introduction
Distributed Generation (DG) system is a part of smart grid concept which is presently forming the backbone of
Distributed
modern Generation
era power (DG)networks
distribution system is[1].
a part
TheofDG
smart grid concept
systems which isinto
are demarcated presently forming
two broad the(i)
parts: backbone of
renewable
modern era power
energy sources distribution
(RES) networks
and (ii) fossil [1]. The
fuel based DG systems
sources. are demarcated
RES include into two broad
DGs like photovoltaic, windparts: (i) renewable
turbines, biomass,
energy sources (RES) and (ii) fossil fuel based sources. RES include DGs like photovoltaic, wind turbines, biomass,

*
Corresponding author. Tel.: +911744233389; fax: +911744238050.
E-mail
* address: ashwani.k.sharma@nitkkr.ac.in
Corresponding author. Tel.: +911744233389; fax: +911744238050.
E-mail address: ashwani.k.sharma@nitkkr.ac.in

--------------© 2017 The Authors. Published by Elsevier Ltd.

Peer-review under
--------------© 2017responsibility
The Authors. ofthe scientific
Published committee
by Elsevier Ltd.of the 4th International Conference on Power and Energy Systems Engineering.
Peer-review under responsibility ofthe scientific committee of the 4th International Conference on Power and Energy Systems Engineering.

1877-0509 © 2018 The Authors. Published by Elsevier B.V.

Peer-review under responsibility of the scientific committee of the 6th International Conference on Smart Computing and Communications
10.1016/j.procs.2017.12.050
 Ankit Uniyal et al. / Procedia Computer Science 125 (2018) 382–388 383
2 Ankit Uniyal et al. / Procedia Computer Science 00 (2017) 000–000

small hydro, geothermal etc. Fossil fuel DGs include internal combustion engines (IC), combustion turbines and fuel
cells [2-3]. The DGs should be connected in distribution system in such a way that it avoids degradation of power
quality and reliability, which makes its placement a complex combinatorial optimization issue requiring concurrent
optimization of multiple objectives [4]. DG can be optimally allocated depending on various criteria; also there is a
variety in the number and type of DGs that have been used in various works. Ahmadi, et.al. [5] aimed at the optimal
power flow to maximize DG’s real power output by reducing network losses for radial and mesh distribution
networks. A multi-objective optimal power flow (OPF) problem is formulated which incorporates a direct Pure
Primal Dual Interior Point Method as the solution methodology. S. Taira, etal. [6] introduced a decision technique
for optimal scheduling of Battery Energy Storage, plug in EV’s and DG’s using Particle Swarm Optimization(PSO)
An optimization problem is formulated with references of DGs, PEVs and tap transformer positions in order to
reduce distribution losses. T. Tayjasanant, et. al. [7] calculated optimal PV-DG size, its position in distribution
system considering minimization of total system losses using the loss sensitivity factor and taking environmental
uncertainties, voltage and harmonic level into account.. Atwa, et. al. [8] calculated actual energy losses in
distribution system with embedded DG’s using MINLP approach by taking wind DG as varying power source
considering wind speed and load profile variations. Khatod,et al. [9] used an evolutionary programming (EP)
technique for optimal placement of wind and solar DG units in radial distribution system for loss reduction and
voltage profile improvement. The probabilistic modeling of DG and objective functions was done using Monte Carlo
approach. Grisales, et. al. [10] utilized three DG techniques: Wind, PV and Small scale hydro (SSH). They used
Chu-Beasely Genetic Algorithm to determine candidate nodes and PSO to determine optimum level of power
injection. Kansal, et. al. [11] utilized PSO technique to find out optimal placement of wind DG in radial distribution
test systems minimizing the real power losses and improvement of voltage profile. The results have been verified
with analytical method which employs use of wind turbine characteristics for proper deployment of the DG unit.
Kumar and Murthy [12] compared optimal DG allocation methods in radial distribution system based on sensitivity
based approaches. DG size was decided based on minimum real power loss in the system. In this paper, optimal DG
sizing and siting has been done considering four objectives i.e. active power loss index (APLI), reactive power loss
index (RPLI), line loading index (LLI) and voltage deviation index (VDI) taken three at a time. Multiple objective
NSGA-II method along with fuzzy satisfying method (FSM) has been used to solve multiple objective optimization
problem. A two point estimate (2PEM) method has been used do probabilistic load flow [13]. The analysis has been
carried out for IEEE 33 radial and IEEE 69 radial bus test system in MATLAB [14].

Nomenclature

ܲ‫ܮ‬௜ and ܳ‫ܮ‬௜ are active and reactive loads on ݅ ௧௛ bus respectively
ܲ‫ ் ܮ‬and ܳ‫் ܮ‬ are the total active and reactive load on the system
‫ܧ‬௣ଵ and ‫ܧ‬௣ଶ the values of two probabilistic points for active power
‫ܧ‬௤ଵ and ‫ܧ‬௤ଶ the values of two probabilistic points for reactive power
APLI active power loss index
VDI voltage deviation index
LLI line loss index
FSM fuzzy satisfying method
MCS Monte Carlo simulation

2. Probabilistic load model

The loads on the distribution system vary instantaneously owing to the uncertainty of the power demand at the
user end thus loads are assumed as random variables and are modelled using probability distribution functions
(PDF’s). Although there are lots of PDF’s like Weibull, beta, uniform etc. but normal distribution or Gaussian
distribution is widely used owing to its accuracy in modelling of probabilistic loads [15]. There are several methods
to perform PLF in the power system. The mostly used methods are sampling methods like MCS, LHS etc., but these
methods require a great deal of time. It is due to the fact in such methods that large number of samples (about 15000-
384 Ankit Uniyal et al. / Procedia Computer Science 125 (2018) 382–388
"Ankit Uniyala" / Energy Procedia00 (2017) 000–000 3

20000) are required to be generated and then DLF is implemented at each of these samples, thus DLF would run for
15000 to 20000 times depending on the number of samples taken which of course takes a lot of time. This drawback
is removed using point estimate method (PEM) as it takes as minimum time as taken to run DLF for four samples. In
case of PEM, �� and �� � � points are estimated for n input variables for�� estimate method and �� � � estimate
method respectively, where � varies from 1 to 2. For instance in case of 2PEM, 2 points is estimated for � input
variables. So, DLF would run for�� � � number of times. In the PLF using 2PEM, the input variables are only real
and reactive power loads on the system, thus DLF would run only 4 times. Hence time taken in the process is greatly
reduced.
The algorithm for 2PEM is shown as follows:
1.) Read line data, load data in the given distribution system.
2.) Calculate ∝�� and∝�� using equations (1) and (2)
∝�� � � ��� ��� � (1)
∝�� � � ��� ��� � (2)
Where,��� and ��� are active and reactive loads on � �� bus respectively and �� � and �� � are total active and
reactive loads on the system
3.) Take mean value of active power ′�� ′ and reactive power ′�� ′ equal to ��� and �� � respectively.
4.) Take standard deviation of active power ′�� ′ and reactive power ′�� ′ equal to 10% of �� and �� respectively.
5.) Calculate the values of two probabilistic points for active power i.e. ��� and ��� using equation (3)
6.) Calculate the values of two probabilistic points for reactive power i.e. ��� and ��� using equations (4)
��� � � �� � �� , ��� � � �� � �� (3)
��� � � �� � �� , ��� � � �� � �� (4)
7.) Calculate two standard probabilistic locations for active powers and reactive powers on each bus using equations
(5) and (6)
��� � � ��� � ��∝�� , ��� � � ��� � ��∝�� (5)
��� � � ��� � ��∝�� , ��� � � ��� � ��∝�� (6)
Where ��� and ��� are the two standard probabilistic locations for active powers on � �� bus and ��� and ��� are the
two standard probabilistic locations for reactive powers on � �� bus
8.) Run DLF using ��� and ��� as new active and reactive loads on each bus, where ��� and ��� are the active power
load and reactive power load on � �� bus for � �� standard location, here � equals 1 and 2 for two standard locations.
9.) Calculate the losses, voltage and other output parameters using equation (7)
�� � ��� � ∑���� ��� (7)
Where,�� is the output parameter and ��� is the output parameter for � �� standard location.

3. Performance indices with DGs

In this work performance indices are used to analyze the operation of the distribution system with distributed
generation systems. In this four performance indices are taken into account with DG integration in the distribution
system. They are given as follows [80]:
(i) Active Power Loss Index (APLI)

APLI is the measure of how much the active power has varied with the integration of DG.
�����
���� � (8)
�������

Where ������� and ����� are the total active power loss without DG and with DG respectively.
(ii) Voltage Deviation Index (VDI)

VDI is the measure of how much the voltage of the system deviates from the nominal substation voltage. For an
ideal DG interconnection, the ��� value should decrease with DG placement.
4 Ankit Uniyal
Ankit Uniyal et al. / Procedia
et al. / Procedia Computer
Computer ScienceScience 125000–000
00 (2017) (2018) 382–388 385

���� ��� �
��� � ∑�
��� � � (9)
����
Where,���� is the nominal substation voltage (1.00 p.u.), �� is the voltage in p.u. at ��� bus of the system and � is
the total number of nodes in the system.
(iii) Line Loading Index (LLI)

LLI is the ratio of the maximum line loading in the distribution system without and with the wind turbine.

��
��� � ��� � � ���� � � ��� �� � �� (10)
�������

Where,��� and ������� are the complex power flows in line �with and without DG respectively. �� is the total
number of branches/lines in the radial distribution system.

4. Methodology

The objective in this paper is to find optimal location and size of DG with minimization of three performance
indices APLI, LLI and VDI. The analysis is performed on IEEE 33 bus radial distribution system. The equality
constraints are load flow constraints. The backward forward sweep method is used to obtain the load flow solution.
The inequality constraints are:
Voltage constraint:
|���� | � |�| � |���� | (11)
DG power constraint constraints:
������� � ��� � ������� (12)
Where, |�| is the magnitude of voltage at any bus, |���� | and |���� | are the magnitudes of the minimum and
maximum voltage limits taken as 0.9 per unit and 1.05 per unit respectively. ��� is the DG size, ������� and �������
are the minimum and maximum value of DG size, taken as 1 KW and 3 MW respectively.
The multiple objective optimization approach used in this work is combination of NSGA-II and fuzzy satisfaction
method (FSM). The pareto optimal solutions have been obtained using NSGA-II. The best solutions out of all
feasible pareto optimal solutions have been obtained using FSM. Firstly, the minimum and maximum limits for all
the performance indices have been decided by observing the area of the pareto optimal front which contains more
solutions satisfying the minimization of all three indices equally. Using the maximum and minimum limits, the
membership functions of the three indices for all the pareto optimal solutions have been calculated. The membership
function for a performance index of a pareto optimal solution shows the degree of satisfaction of that solution for the
very performance index. Now, the solutions having higher membership function for all three performance indices
have been selected as the best solutions. In this case only two best solutions have been considered as the others have
not been able to satisfy the minimization of the three indices at a time. These solutions have been categorized as
Solution1 and Solution 2.

4.1 Optimal DG allocation considering minimization of VDI, LLI and APLI for 33 bus system

In this study, VDI, LLI and APLI are taken as indices to be minimized. The NSGA-II method has been applied to
obtain the pareto optimal solutions as is shown in Fig. 1 Table 1 shows maximum and minimum values of VDI, LLI
and APLI taken to calculate their individual membership value using FSM. As shown in Table 2, the optimal DG
locations are at the same bus 8 and optimal size is 2.4183 MW for solution 1 and 2.3384 for solution 2. After placing
the optimally sized DG at the optimal bus, various other parameters like total real power loss (TPL), total reactive
power loss (TQL), total real power taken from the substation (TPS), total reactive power taken from the substation
386 Ankit Uniyal et al. / Procedia Computer Science 125 (2018) 382–388
"Ankit Uniyala" / Energy Procedia00 (2017) 000–000 5

(TQS), values of all three indices i.e. VDI, LLI and APLI are calculated which are shown in Table 2.The value of
VDI is coming less in case of solution 1 while values of LLI, APLI and RPLI are coming less in case of solution 2.
DG size is higher in case of solution 1 than solution 2. TPL and TQL are higher in case of solution 1. Figure 2 shows
the variation of voltages at each bus of 33 bus system, where solution 1 gives somewhat better voltage profile than
solution 2. However voltage profiles from both the solutions are better than the without DG case. Figure 3 shows the
variation of real power loss at each branch, where solution 2 is bit better than solution 2; while both the solutions are
better than the without DG case. Figure 4 shows the variation of reactive power loss profile at each branch, where
solution 2 is slightly better than solution 1; while both the solutions are better than the without DG case

0.8
AP L I

0.6 0.04

0.4 0.03
5
4
3 0.02
2
1
0 0.01 VDI
LLI
Fig. 1. Pareto optimal solutions for 33 bus system

1.02
1 Without
0.98 DG
0.96
Voltage (pu)

0.94 Solution 1
0.92
0.9
0.88 Solution 2
0.86
0.84
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33
Bus Number

Fig. 2. Voltage profile for IEEE 33 bus system

Table 1 Maximum and Minimum values of objective functions for 33 bus system

Performance index Performance index Performance index

Max VDI Min VDI Max LLI Min LLI Max APLI Min APLI
0.03 0.015 3.5 1.5 0.66 0.54
 Ankit Uniyal et al. / Procedia Computer Science 125 (2018) 382–388 387
6 Ankit Uniyal et al. / Procedia Computer Science 00 (2017) 000–000

60
Without
50
DG
Real Power Loss (KW) 40
Solution 1
30

20
Solution 2
10

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Branch Number

Fig. 3. Real power loss profile case 3 for 33 bus system

40
35 Without
Reactive Power Loss (KVAR)

DG
30
25
Solution 1
20
15
10
Solution 2
5
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Branch Number

Fig. 4. Reactive power loss profile for case 3 for 33 bus system

Table 2 Most preferred Pareto Optimal solutions using NSGA-II for IEEE 33 bus system
Solution 1 Solution 2
DG bus location 8 8
DG size (MW) 2.4183 2.3384
TPL (KW) 130.3500 128.0113
TQL (KW) 96.7509 94.3335
TPS (MW) 1.4271 1.5046
TQS (MVAr) 2.3968 2.3943
VDI 0.0195 0.0212
LLI 2.2583 2.0534
388 Ankit Uniyal et al. / Procedia Computer Science 125 (2018) 382–388
"Ankit Uniyala" / Energy Procedia00 (2017) 000–000 7

APLI 0.6094 0.5985

RPLI 0.6672 0.6505
Membership function µ(VDI) 0.7003 0.5857
Membership function µ(LLI) 0.6208 0.7233
Membership function µ(APLI) 0.4216 0.5128

4. Conclusions

In this work, optimal DG allocation has been obtained taking multi-objective criteria of minimizing Active Power
Loss Index, Line Loading Index and Voltage Deviation Index. Based on the comparison of the results, it is observed
that: there is voltage profile improvement with DG. The real and reactive power loses reduces with DG. The results
obtained for solution 1 and 2 are almost similar with DG. For solution 2, the DG size obtained is slightly lower and
real and reactive power losses are also observed slightly lower. The losses are observed lower with DG installation.
DG technology in the future distribution systems will be beneficial from technical as well as economically and is
gaining importance due to their environmental friendly nature, less gestation time and modular nature. This study
will provide insight to the distribution system operator for better planning of the distribution network.

References

1. R. E. Brown, electric power distribution reliability, CRC press, 2008.

2. Y. A. Katsigiannis and P. S. Georgilakis, “Effect of customer worth of interrupted supply on the optimal design of small isolated power
systems with increased renewable energy penetration,” IET Gener. Transm. Distrib., 7, (3), pp. 265–275, 2013.
3. J. A. Peças Lopes, N. Hatziargyriou, J. Mutale, P. Djapic, and N. Jenkins, “Integrating distributed generation into electric power systems: a
review of drivers, challenges and opportunities,” Elect. Power Syst. Res., 77, (9), pp. 1189–1203., 2007.
4. A. Alarcon-Rodriguez, G. Ault, S. Galloway, “Multi-objective planning of distributed energy resources: A review of the state-of-the-art,”
Renewable and Sustainable EnergyReviews, vol. 14, pp. 1353–1366, 2010.
5. A.R. Ahmadi and T.C. Green , “Optimal Power Flow for Autonomous Regional Active Network Management System”, IEEE/ PES general
meeting, pp. 1-7, 2009.
6. S. Taira; Z. Zaidi and T.Funabashi, “Assessment of impact of distributed generators, plug-in electric vehicle and Battery Energy Storage
System on power distribution losses”, IEEE/ IFEEC 1st International Conference, pp. 675-680, 2013.
7. Tayjasanant and V. Hengsritavat “Comparative Evaluation of DG and PV-DG Capacity Allocation in a Distribution System”, IEEE/ICHQP
15th International Conference, pp. 293-298, 2012.
8. Y. M. Atwa; EI- Saadany; R. Seethapathi and M. Salama, “Effect of wind-based DG seasonality and uncertainty on distribution system
losses”, Power Symposium NAPS’08 40th North American Conference, pp.1-6, 2008.
9. D. K. Khatod; V. Pant and J. Sharma,“Evolutionary Programming Based Optimal Placement of Renewable Distributed Generators”, IEEE
Transactions on Power Systems, vol.28, pp. 683-795, 2013.
10. L. F. Grisales; A. Grajales; O.D. Montoya; A. Hincapie and M. Granada, “Optimal Location and Sizing of Distributed Generators Using a
Hybrid Methodology and Considering Different Technologies”, pp.1-4, 2015.
11. S. Kansal; B.B.R Sai; V. Tyagi and V. Kumar, “Optimal location and sizing of Distributed Generators using a hybrid methodology and
considering different technologies”,IEEE 6th Latin American Symposium on Circuits & Systems (LASCAS), pp.1-4, 2015.
12. V.V.S.N. Murthy and A. Kumar A, "Comparison of optimal DG allocation methods in radial distribution systems based on sensitivity
approaches", Electrical Power and Energy Systems 53, pp. 450–467, 2013.
13. Chun-Lein Su,“Probabilistic Load-Flow Computation Using Point Estimate Method”,IEEE Transactions on Power Systems, Vol. 20, No 4,
pp.1843-1851, 2005.
14. MATLAB R2013a © 1994-2013 The Math Works, Inc.
15. Sooraj Narayan K, Ashwani Kumar, “Probabilistic Analysis of Radial Distribution System Performance with Varying Wind Speed Levels”, in
Proc. 6th IEEE Power India Conference (PIICON), pp. 1-6, New Delhi, December 7-9, 2014.