Ain Shams Engineering Journal 11 (2020) 409–418

Contents lists available at ScienceDirect

Ain Shams Engineering Journal

journal homepage: www.sciencedirect.com

Electrical Engineering

An exact MINLP model for optimal location and sizing of DGs in

distribution networks: A general algebraic modeling system approach
Oscar Danilo Montoya a,⇑, Walter Gil-González b, L.F. Grisales-Noreña c
a
Programa de Ingeniería Eléctrica e Ingeniería Electrónica, Universidad Tecnológica de Bolívar, Km 1 vía Turbaco, Cartagena, Colombia
b
Universidad Tecnológica de Pereira, AA: 97, 660003 Pereira, Colombia
c
Instituto Tecnológico Metropolitano, 050012 Medellín, Colombia

a r t i c l e i n f o a b s t r a c t

Article history: This paper addresses the classical problem of optimal location and sizing of distributed generators (DGs)
Received 31 December 2017 in radial distribution networks by presenting a mixed-integer nonlinear programming (MINLP) model. To
Revised 6 May 2019 solve such model, we employ the General Algebraic Modeling System (GAMS) in conjunction with the
Accepted 19 August 2019
BONMIN solver, presenting its characteristics in a tutorial style. To operate all the DGs, we assume they
Available online 14 November 2019
are dispatched with a unity power factor. Test systems with 33 and 69 buses are employed to validate the
proposed solution methodology by comparing its results with multiple approaches previously reported in
Keywords:
the specialized literature. A 27-node test system is also used for locating photovoltaic (PV) sources con-
Distributed generation
Distribution systems
sidering the power capacity of the Caribbean region in Colombia during a typical sunny day. Numerical
General algebraic modeling system results confirm the efficiency and accuracy of the MINLP model and its solution is validated through the
Mixed-integer nonlinear programming GAMS package.
Optimal location and sizing of distributed Ó 2019 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Ain Shams University.
generation This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-
nd/4.0/).

1. Introduction positive and negative effects because they modify the behavior of
the state variables of the grid, which absolutely depend on their
1.1. General context location and sizing in the power system [10]. Advances in solid-
state electronics and software have boosted the high penetration
Nowadays, around the world, electricity is mainly produced by of renewable energy into electrical networks, mainly at distribu-
large-scale plants that operate using conventional sources of tion levels. Hence, strategies or methods that allow the correct
energy, such as hydraulic and thermal technologies. Electric plants integration of these emerging technologies are necessary [11,9].
are usually located far from final consumers and, therefore, energy In the last decade, different models, methods, and optimization
losses associated with transmission lines increase [1,2]. Addition- techniques for sizing and locating DGs in electric distribution net-
ally, the voltage profile can exceed its lower and upper bounds works have been proposed. They have allowed the integration of
[3,4]. For that reason, distributed generators (DGs) have become renewable energy sources (e.g., wind and photovoltaic (PV) gener-
a local solution for medium- and low-voltage power systems [5– ation), small-scale hydraulic generation, and biomass generation,
7]. DGs enable the injection of active and reactive power closer among others, in an appropriated way [12–14]. DGs enable an
to consumers, which can produce benefits in terms of quality of improvement of different technical aspects, such as voltage pro-
service [8,1,9]. Integrating DGs into the electric system has both files, the power capacity of the lines, and the reliability and quality
of service, as well as a reduction of active and reactive power losses
[15]. Said generation technologies also allow utility companies to
⇑ Corresponding author.
diversify their energy matrix and transform electric power grids
E-mail addresses: o.d.montoyagiraldo@ieee.org, omontoya@utb.edu.co (O.D.
into autonomous and smart systems [16].
Montoya), wjgil@utp.edu.co (W. Gil-González), luisgrisales@itm.edu.co (L.F. Gri-
sales-Noreña).
Peer review under responsibility of Ain Shams University.
1.2. Motivation

Advances in power electronics transform the possibility of hav-

ing electrical networks with a high penetration of distributed gen-
Production and hosting by Elsevier
eration at distribution levels into a reality, mainly with the

https://doi.org/10.1016/j.asej.2019.08.011
2090-4479/Ó 2019 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Ain Shams University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
410 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418

integration of multiple renewable energy sources. Therefore, math- distribution networks and power losses minimization via the inte-
ematical models and new solution methodologies should be con- gration of DGs. This work presents, in a numerical simulation, the
tinuously developed to address the problem of optimal location possibility of extending our proposed MINLP model for the optimal
and sizing of such sources in power distribution networks. For that integration of renewable energy resources in a typical electrical
reason, the motivation behind this study was providing the spe- distribution network in Colombia, which is not typically addressed
cialized literature with a powerful tool (a GAMS optimization pack- in metaheuristic or conventional MINLP models. Furthermore, this
age) for solving large-scale nonlinear discrete problems via paper contains a simple example with the implementation of the
mathematical interpretation. Such tool focuses on the mathemati- MINLP model, which will help researchers and students to use
cal modeling itself by concentrating the attention of the researches the GAMS package for evaluating future studies in this area and
in the correct mathematical modeling by using a compact and as a powerful comparative approach when emerging optimization
structured architecture. For that purpose, this paper presents a models are tested and validated.
simple implementation of the problem under study in order to
explain all the basic concepts for GAMS usage. 1.5. Document structure

1.3. Brief state-of-the-art The rest of this document is organized as follows. Section 2 pre-
sents the complete mathematical formulation of the problem by
The literature about the optimal location and sizing of DGs in describing and discussing all the equations along with their vari-
distribution networks is extensive and rich. This topic has a strong ables and meanings. In Section 3, we provide all the necessary ele-
background in terms of mathematical formulations and solution ments for using GAMS as an optimization package; in addition,
techniques. Regarding its mathematical formulation, this problem such section reports the complete mathematical implementation
corresponds to a nonlinear non-convex optimization model with of the MINLP model analyzed in this work as an opportunity to
discrete and continuous variables [17]; its mathematical structure identify all the concepts that compose the GAMS package. Section 4
is an extension of the classical distribution system power flow presents all the information related to the 33 and 69-node test
problem with discrete variables [18]. In terms of solution method- feeders. Section 5 details all the numerical results of the proposed
ologies, the approaches most commonly adopted are metaheuristic GAMS approach compared with approaches reported in the litera-
optimization techniques [19]. Such optimization approaches allow ture; in addition, we present the extension of the model for the
the separation of the location problem from the sizing problem by daily operation of distribution networks with PV integration in
adopting a master-slave methodologies [19,20]. the context of a Colombian electrical system located in the Carib-
In the case of master-slave approaches, multiple discrete opti- bean region. Section 6 draws the main conclusions derived from
mization methods have been proposed: genetic algorithms [21], this work as well as some possible future works, followed by the
ant lion optimizers [22], tabu search algorithms [23], simulated acknowledgments and the references.
annealing methods [24], krill herd algorithms [25–27],
population-based incremental learning [28], teaching-based learn- 2. Problem description
ing optimizers [29], bat and firefly algorithms [30–33], symbiotic
organism search algorithms [34], harmonic search algorithms 2.1. Mathematical formulation
[35], and imperialist competitive algorithms [19].
Regarding the methodology for solving the sizing problem, the The mathematical model of the optimal location and sizing of
most common approach is particle swarm optimization [6,28], DGs in RDN corresponds to a MINLP problem [17]. Here, integer
since it is easy to implement in any computational language and (binary) variables represent the decision variables associated with
its results are comparable with interior-point and convex opti- the location or not of a DG in the grid, while continuous variables
mization methods [36]. are associated to the classical power flow formulation, which is
The specialized literature has also proposed exact models for represented by magnitudes and angles of the voltage per node.
addressing the problem studied in this paper. In [17], a MINLP The following is the detailed mathematical model proposed in this
model for the problem of optimal location and sizing of DGs in dis- paper [17].
tribution systems was proposed by implementing a master-slave Objective function
approach in the decoupled form. That model combines sequential !
quadratic programming methods with a branch and bound X X  
min z ¼ Vi V j Y ij cos hi  hj  /ij ð1Þ
approach, which implies that, so far, a compact formulation has
i2XN j2XN
not been used as proposed in this paper. In [37], a MINLP model
was proposed to address the same problem, and the GAMS soft- where z is the value of the objective function, which corresponds to
ware was used for its solution. Nevertheless, its implementation the power losses in all the branches of the network under a load
has not yet been extended to daily operation with photovoltaic peak scenario of demand; XN , the set associated with the nodes of
(PV) sources, as proposed by us. the network; V i and V j , the voltages’ magnitudes at nodes i and j,
respectively; hi and hj , the voltages’ angles at nodes i and j, respec-
1.4. Contribution and scope tively; Y ij , the magnitude of the admittance associated with the line
connected between i and j nodes; and /ij , its angle.
Based on the review of the state-of-the-art above, this paper Constraints
presents a solution to the problem of optimal location and sizing X  
of DGs in distribution networks in a tutorial style by taking advan- PCG DG
i þ Pi ¼ Vi V j Y ij cos hi  hj  /ij
tage of the compact modeling available in the GAMS software and j2XN ð2Þ
its nonlinear optimization packages. Note that the main contribu- þPDi ; f8i 2 XN g
tion of our research is the possibility of implementing the exact
MINLP model of the problem using compact sets in GAMS without where P GC
i represents the active power generated at node i by a con-

adopting decoupling methods (e.g., master-slave algorithms), ventional generator; PDG

i , the active power generated by a DG
which allows us to focus on the mathematical formulation itself. located at node i; and P Di , the total active power demanded at node
In addition, the scope of our study is mainly defined by electrical i.
 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418 411

Eq. (2) represents the active power balance at each node in the 2.2. General comments
network.
X   The MINLP model described from (1) to (7) represents problem
Q CG D
i  Qi ¼ Vi V j Y ij sin hi  hj  /ij of optimal location and sizing of DGs in a RDN [17]. Such model
j2XN ð3Þ only focuses on the technical aspects related to active power losses
f8i 2 XN g in the branches of the network, respecting classical constraints of
the power flow problem [1]. Note that this model corresponds to
where Q GC
i denotes the reactive power generated at node i by a con- an adaptation of the optimal power flow problem reported by
ventional generator; Q DG
i , the reactive power generated by a DG
[38], in order to allow the location and sizing of DGs as a function
located at node i; and Q Di , the total reactive power demanded at of the total active power consumption.
node i. An adaptation for obtaining a power flow time-varying formu-
Eq. (3) represents the reactive power balance at each node in lation can be easily extracted for the model, as mentioned earlier,
the network. by adding some sub-indexes and sums [39]. Here, we used the
demand peak hour to define the optimal location and sizing of each
V min
i 6 V i 6 V max
i f8i 2 XN g ð4Þ distributed generator because it represents the worst operating
point in the RDN, with the highest power losses and voltage devi-
where V min
i and V max
i represent the minimum and maximum ations. In addition, we also extended this model to the daily oper-
allowed voltage values at each node. Note that (4) corresponds to ation of an electrical network in order to evaluate the possibility of
the voltage regulation constraint. sizing PV generators.
This mathematical formulation can be directly implemented in
0 6 PDG
i 6 xi PDG;max
i f8i 2 XN g ð5Þ the GAMS platform [40], which allowed us to obtain an adequate
solution with a low computational effort. Such solution can be
where PDG;max
i is the maximum allowed active power injection at local or global, depending on the characteristics of the problem
node i by a DG and xi represents the decision variable, which takes under analysis.
a value of 1 if the DG is located at node i and 0 otherwise. Eq. (5) The next section presents a possible GAMS implementation for
shows the possibility of locating and sizing a DG at any node in a small radial distribution network. Such implementation uses sets
the RDN. We considered only active power injection in the DGs, and a compact formulation [41].
which means that Q DG
i ¼ 0 in this paper.
X
xi 6 NDG
av a ð6Þ 3. General algebraic modeling system: GAMS
i2XN
The GAMS software is a powerful optimization package devel-
where NDGav a is the available number of DGs, which implies that (6) oped for interpreting and solving nonlinear large-scale optimiza-
limits the number of location possibilities for the distributed gener- tion problems based on a compact formulation [40,42]. Said
ation in the RDN. software works with a simple plain text structure, where the opti-
mization model is written using five essential components [43]:
xi 2 f0; 1g f8i 2 XN g ð7Þ
Finally, (7) expresses the binary nature of the decision variable. i. The sets where the variables make sense, e.g., set of nodes:
i 2 XN .
ii. All the scalars, parameters, and matrices involved in the
model, i.e., number of generators, matrices, and vectors.
iii. All the variables in the model, e.g., voltages, powers, angles,
etc.
iv. The equations’ names and their mathematical structures,
e.g., expressions (2) and (3) associated with the power bal-
ance constraints.
v. The nature of the model (i.e., MINLP) and displaying options.

Fig. 1 presents the GAMS interface and the words reserved for
implementing an optimization model.
Note that, at the bottom of Fig. 1, each reserved word is needed
to define all the particular components of the model under study.
In that sense, we present a simple example that can illustrate the
complete structure of an optimization model implemented in the
GAMS software [41]. Such example aims at guiding readers on
the easy utilization of this optimization toolbox for addressing
optimization problems in engineering. For that purpose, let us con-
Fig. 1. GAMS software environment.
sider the grid depicted in Fig. 2, an electrical network composed of
7 nodes and 6 lines operated at 23 kV as voltage output at the sub-

Fig. 2. Electrical configuration of the 7-node test system used in the GAMS implementation example.
412 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418

station (slack node). Its line parameters, as well as power con-

sumption, are reported in Table 1.
The 7-node test system was implemented as an example in From Algorithm 1, it can be seen that all the components in the
GAMS considering 23 kV and 1 MVA as voltage and base power, optimization model (1)–(7) were included. Hence, the following
respectively. We also considered the possibility of installing one are the most important features of this implementation:
distributed generator with unlimited capability.
i. The command ALIAS(N,NP), in line 5, allows the duplica-
Algorithm 1. GAMS implementation of the model in (1)–(7) for tion of the set of nodes N in the set NP to evaluate the power
the 7-node example balance equations and the objective function.
ii. All the parametric information of the model was defined
between lines 6 and 39.
iii. The set of variables was classified into continuous variables
(voltage, angles, and powers) and binary variables (optimal
location of the DG), as can be seen from lines 40 to 48.
iv. Lines 49 and 50 define the voltage constraint (4) and the typ-
ical behavior of the slack node in a radial distribution net-
work, i.e., plane voltage.
v. Lines 51 to 57 define the name of the equations, while lines
58, 59, and 61 are the objective function and the active and
reactive power balance constraints (i.e., the compact repre-
sentation of (1)–(3)).
v. Lines 63 to 65 represent the maximum number of DGs avail-
able as well as their minimum and maximum power outputs
(i.e., constraints (5) and (6)).
v. Lines 66 to 69 define the characteristics of the model and its
type (minimization), as well as its displaying features.

Note that, if we solve this model in GAMS by fixing the number

of available DGs at zero, then the base case of the network is
achieved. Figs. 3 and 4 present the GAMS outputs when zero and
one distributed generator are considered.
Note that, in Fig. 3, the total power losses without distributed
generation reach 128:058 kW, while in Fig. 4 the final losses
decrease to 56:9563 kW when one DG is installed. To reduce such
losses, GAMS determined that the distributed generator must be
located at node 3 with a total capacity of 6:3610 MVA.
It is important to mention that the model implemented in
GAMS is general for the problem analyzed in this paper, which
implies that, based on the parametric information of the test sys-
tem, it can find an optimal solution to the proposed MINLP model,
as will be confirmed in the next section. For detailed information
about GAMS and a complete description of its functionalities, refer
to [40,43,41].

4. Test systems and simulation cases

This section presents the electrical configuration, as well as the

test system information, of the radial distribution systems
employed in this work for validating the MINLP formulation and
its solution in the GAMS package. Two test system were used: a
33-node test system and a 69-node test feeder. The complete infor-
mation of these test systems is presented below.

4.1. 33-node test feeder

This test system is composed of 33 nodes and 32 branches with

12:66 kV of operating voltage. The slack node is located at node 1,
and its configuration is presented in Fig. 5. This feeder has
3715 kW and 2300 kVAr of total active and reactive power
demand. The initial active power losses of this system equal
210:9876 kW. For this test system, the possibility of installing 3
distributed generators was considered since that is the most com-
monly reported solution in the specialized literature [28]. Each dis-
tributed generator is limited from 0 kW to 2500 kW.1 In addition,
we considered voltage and power base values of 12:66 kV and
1000 kW, respectively.
 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418 413

Table 1
Electrical parameters of the 7-node test feeder used in the GAMS implementation example

Node i Node j Rij [X] X ij [X] P j [kW] Q j [kW]

1 2 0.5025 0.3025 1000 600

2 3 0.4020 0.2510 900 500
3 4 0.3660 0.1864 2500 1200
2 5 0.3840 0.1965 1200 950
5 6 0.8190 0.7050 1050 780
2 7 0.2872 0.4088 2000 1150

version of Windows 7 Professional. The implemented mathemati-

cal model is the same as the one presented in Section 3, except that
the information of each test feeder was modified.
To demonstrate the robustness and efficiency of the GAMS
package for locating and sizing DGs in distribution networks, we
compared our results with the solutions previously reported in
[34,24]. In addition, we considered that all the DGs were operated
with a unity power factor, as recommend in [28].
Fig. 3. GAMS output with zero DGs.

5.1. 33-node test feeder

Table 4 presents a list of solutions provided by [34] for the 33-

node test feeder with the corresponding location, size, and power
losses when 3 DGs are considered.
Note that the power losses results reported in Table 4 show that
the GAMS optimization package in conjunction with the BONMIN
Fig. 4. GAMS output with a unique DG. solver finds the best solution with respect to the all comparative
methods, i.e., 72:79 kW, followed by the REPSO method with
76:91 kW and the LSFSA approach with 82:03 kW, in the first three
positions. It is also important to highlight that the MINLP model we
proposed, solved through a GAMS implementation, finds an alter-
native set of nodes for locating all the distributed generators
(e.g., nodes 6, 18, and 30) with a total power injection of 2:9336
MW, while the REPSO and LSFSA approaches reach 2:5212 MW
and 2:4677 MW, respectively. Such values imply that the solutions
provided by REPSO and LSFSA can be stuck in local optima, while
Fig. 5. Electrical configuration of the 33-node test system. our approach allows the improvement of those solutions by
increasing the total power injection. In order to find the best
solution.
The information of all the branches, as well as the load con- Fig. 7 presents a comparison of the power losses reduction per-
sumption of the 33-node test feeder, is listed in Table 2. centages of all the approaches reported in Table 4, along with the
initial power losses. This figure confirms that the GAMS approach
4.2. 69-node test feeder allows the highest power losses reduction, 65:50 %, followed by
the REPSO and LSFSA approaches with 63:55 % and 61:12 %,
This test system consists of 69 nodes and 68 branches with respectively.
12:66 kV of operating voltage. The slack node is located at node
1, and its configuration is depicted in Fig. 6. This feeder has
3890:7 kW and 2693:6 kVAr of total active and reactive power
5.2. 69-node test feeder
demand. The initial active power losses of this system equal
225:0718 kW. For this test system, we also considered the possibil-
Table 5 presents a list of solutions provided by [34] for the 69-
ity of installing 3 distributed generators, and each of them limited
node test feeder with the corresponding location, size, and result-
from 0 kW to 2000 kW. In addition, we also considered 12:66 kV
ing power losses, when 3 DGs were considered.
and 1000 kW as voltage and power base values, respectively.
The behavior of the power losses presented in Table 5 proves
The information of all the branches, as well as the load con-
that the GAMS approach effectively converges to the best solution,
sumption of the 69-node test feeder, is presented in Table 3.
in contrast with the other methodologies. In this system, our
MINLP model, solved through the BONMIN solver, reaches final
5. Computational validation power losses of 72:09 kW, followed by the LSFSA and TLBO meth-
ods with 77:10 kW and 81:00 kW, respectively.
To solve the general MINLP model that represents the problem Fig. 8 shows the power losses reduction achieved by the pro-
of optimal location and sizing of DGs in radial distribution systems, posed GAMS approach as well as the other methods. Note that
we employed the GAMS optimization package with the solver the GAMS approach achieves the highest reduction in power losses,
BONMIM in a desktop computer with an INTEL(R) Core(TM) with 67:97 %, which confirms the efficiency and accuracy of the
i5  3550 3:5-GHz processor and 8 GB of RAM running a 64-bit proposed MINLP model and its solution by GAMS.
414 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418

Table 2
Electrical parameters of the 33-node test feeder.

Node i Node j Rij [X] X ij [X] P j [kW] Q j [kW]

1 2 0.0922 0.0477 100 60

2 3 0.4930 0.2511 90 40
3 4 0.3660 0.1864 120 80
4 5 0.3811 0.1941 60 30
5 6 0.8190 0.7070 60 20
6 7 0.1872 0.6188 200 100
7 8 1.7114 1.2351 200 100
8 9 1.0300 0.7400 60 20
9 10 1.0400 0.7400 60 20
10 11 0.1966 0.0650 45 30
11 12 0.3744 0.1238 60 35
12 13 1.4680 1.1550 60 35
13 14 0.5416 0.7129 120 80
14 15 0.5910 0.5260 60 10
15 16 0.7463 0.5450 60 20
16 17 1.2890 1.7210 60 20
17 18 0.7320 0.5740 90 40
2 19 0.1640 0.1565 90 40
19 20 1.5042 1.3554 90 40
20 21 0.4095 0.4784 90 40
21 22 0.7089 0.9373 90 40
3 23 0.4512 0.3083 90 50
23 24 0.8980 0.7091 420 200
24 25 0.8960 0.7011 420 200
6 26 0.2030 0.1034 60 25
26 27 0.2842 0.1447 60 25
27 28 1.0590 0.9337 60 20
28 29 0.8042 0.7006 120 70
29 30 0.5075 0.2585 200 600
30 31 0.9744 0.9630 150 70
31 32 0.3105 0.3619 210 100
32 33 0.3410 0.5302 60 40

them at nodes 10 and 16 with maximum capacities of 1:321 p.u

and 1:008 p.u, respectively2
Note that the GAMS package solves the problem for all the dif-
ferent options; in the case of 1 PV generator, it achieves a reduction
of 7:54 % in the total power losses per day, while with 2 and 3 gen-
erators, the reductions are 12:53 % and 17:32 %, respectively. These
results imply that, as the number of PV generators increases, power
losses decrease. Notwithstanding, these reductions tend to the sat-
uration due to the impossibility of generating power at night, as
Fig. 6. Electrical configuration of the 69-node test system.
depicted in Fig. 11, where the energy reduction in the system exhi-
bits an exponential decreasing asymptotic behavior approaching
5.3. Optimal location of renewable generators in a daily operational 1600 kWh/day.
environment
5.4. General comments
Here, we explore the possibility of using GAMS for locating
renewable generators (PV systems) in radial distribution systems The numerical validation presented in the section above shows
by considering the typical solar radiation performance in a Colom- that:
bian system in the Caribbean region. For that purpose, we employ
the 27-node test feeder reported in [42] with the branch and peak U The MINLP model and its implementation in GAMS can produce
load information reported in Table 6. The grid configuration of this excellent solutions in terms of power losses reduction for the
test feeder is illustrated in Fig. 9. To evaluate the daily operation of 33-node test feeder and the 69-node test feeder.
this system including PV systems, we employ the demand varia- U The 27-node test feeder, with a daily operation, revealed the
tion and the PV generation capacity in Fig. 10. In addition, we possibility of using the MINLP model with time-varying vari-
use 13:8 kV and 1000 kW as voltage and power bases, respectively; ables for solving the problem of optimal location and sizing of
during GAMS implementation. renewable generators (e.g., PV systems) through its GAMS
In this test system, we evaluate the possibility of installing from implementation.
1 to 3 PV generators with the curve of power generation reported U The integration of multiple PV modules for generating renew-
in Fig. 10. Note that, in this test system, the total power losses per able power in distribution grids allows the reduction of their
day are 2094:01 kWh/day when renewable power generation has total daily power losses. However, a massive integration of such
not yet been installed, while such losses are lower when different modules does not cause important reductions in said losses.
numbers of PV generators are installed, as reported in Table 7. In
addition, said Table shows the size of each PV generator, e.g., in
the case of 2 PV generators, the GAMS package suggests locating 2
To determine the total power output of each generator, the maximum capacities
reported in Table 7 should be multiplied by the typical generation provided in Fig. 10.
 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418 415

Table 3
Electrical parameters of the 69-node test feeder.

Node i Node j Rij [X] X ij [X] P j [kW] Q j [kW]

1 2 0.0005 0.0012 0 0
2 3 0.0005 0.0012 0 0
3 4 0.0015 0.0036 0 0
4 5 0.0251 0.0294 0 0
5 6 0.3660 0.1864 2.6 2.2
6 7 0.3811 0.1941 40.4 30
7 8 0.0922 0.0470 75 54
8 9 0.0493 0.0251 30 22
9 10 0.8190 0.2707 28 19
10 11 0.1872 0.0619 145 104
11 12 0.7114 0.2351 145 104
12 13 1.0300 0.3400 8 5
13 14 1.0440 0.3450 8 5
14 15 1.0580 0.3496 0 0
15 16 0.1966 0.0650 45 30
16 17 0.3744 0.1238 60 35
17 18 0.0047 0.0016 60 35
18 19 0.3276 0.1083 0 0
19 20 0.2106 0.0690 1 0.6
20 21 0.3416 0.1129 114 81
21 22 0.0140 0.0046 5 3.5
22 23 0.1591 0.0526 0 0
23 24 0.3463 0.1145 28 20
24 25 0.7488 0.2475 0 0
25 26 0.3089 0.1021 14 10
26 27 0.1732 0.0572 14 10
3 28 0.0044 0.0108 26 18.6
28 29 0.0640 0.1565 26 18.6
29 30 0.3978 0.1315 0 0
30 31 0.0702 0.0232 0 0
31 32 0.3510 0.1160 0 0
32 33 0.8390 0.2816 10 10
33 34 1.7080 0.5646 14 14
34 35 1.4740 0.4873 4 4
3 36 0.0044 0.0108 26 18.55
36 37 0.0640 0.1565 26 18.55
37 38 0.1053 0.1230 0 0
38 39 0.0304 0.0355 24 17
39 40 0.0018 0.0021 24 17
40 41 0.7283 0.8509 102 1
41 42 0.3100 0.3623 0 0
42 43 0.0410 0.0478 6 4.3
43 44 0.0092 0.0116 0 0
44 45 0.1089 0.1373 39.22 26.3
45 46 0.0009 0.0012 39.22 26.3
4 47 0.0034 0.0084 0 0
47 48 0.0851 0.2083 79 56.4
48 49 0.2898 0.7091 384.7 274.5
49 50 0.0822 0.2011 384.7 274.5
8 51 0.0928 0.0473 40.5 28.3
51 52 0.3319 0.1140 3.6 2.7
9 53 0.1740 0.0886 4.35 3.5
53 54 0.2030 0.1034 26.4 19
54 55 0.2842 0.1447 24 17.2
55 56 0.2813 0.1433 0 0
56 57 1.5900 0.5337 0 0
57 58 0.7837 0.2630 0 0
58 59 0.3042 0.1006 100 72
59 60 0.3861 0.1172 0 0
60 61 0.5075 0.2585 1244 888
61 62 0.0974 0.0496 32 23
62 63 0.1450 0.0738 0 0
63 64 0.7105 0.3619 227 162
64 65 1.0410 0.5302 59 42
11 66 0.2012 0.0611 18 13
66 67 0.0047 0.0014 18 13
12 68 0.7394 0.2444 28 20
68 69 0.0047 0.0016 28 20
416 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418

Table 4
Location and dispatch of the generators in the 33-node test feeder.

Method Power generation [p.u] (Node) Losses [kW]

GA [21] 1.5000 (11) 0.4228 (29) 1.0714 (30) 106.30
PSO [21] 1.1768 (8) 0.9816 (13) 0.9297 (32) 105.35
TLBO [29] 0.8847 (9) 0.8953 (18) 1.1958 (31) 104.00
REPSO [44] 1.2274 (6) 0.6068 (14) 0.6870 (31) 76.91
HSA [45] 0.5927 (16) 0.2133 (17) 0.1913 (18) 135.69
SOS [34] 2.2066 (6) 0.2000 (28) 0.7167 (29) 104.19
LSFSA [24] 1.1124(6) 0.4874 (18) 0.8679 (30) 82.03
GAMS 0.7709 (14) 1.0969 (24) 1.0658 (30) 72.79

Table 6
Electrical parameters of the 27-node test feeder.

Node i Node j Rij [X] X ij [X] P j [kW] Q j [kW]

1 2 0.15208 0.19855 0 0
2 3 0.65805 0.59745 0 0
3 4 0.19742 0.17924 297.5 184.4
4 5 0.43848 0.26038 0 0
5 6 0.48720 0.28931 255 158
6 7 0.48197 0.22732 0 0
7 8 0.87630 0.41330 212.5 131.7
8 9 1.09540 0.51663 0 0
9 10 0.87630 0.41330 266.1 164.9
2 11 0.87630 0.41330 85 52.7
Fig. 7. Power losses reduction of different methods in the 33-node test feeder.
11 12 1.07780 0.50836 340 210.7
12 13 0.65722 0.30998 297.5 184.4
13 14 0.49073 0.23145 191.3 118.5
14 15 0.87630 0.41330 106.3 65.8
Table 5 15 16 0.87630 0.41330 255 158
Location and dispatch of generators in the 69-node test feeder. 3 17 0.87630 0.41330 255 158
Method Power generation [p.u] (Node) Losses [kW] 17 18 0.52578 0.24798 127.5 79
18 19 0.78867 0.37197 297.5 184.4
GA [21] 0.9297 (21) 1.0752 (62) 0.9925 (64) 89.00 19 20 0.83248 0.39263 340 210.7
PSO [21] 0.9925 (17) 1.1998 (61) 0.7956 (63) 83.20 20 21 0.87630 0.41330 85 52.7
TLBO [29] 0.7574 (25) 1.0188 (60) 1.1784 (63) 81.00 4 22 0.87630 0.41330 106.3 65.8
HSA [45] 1.6283 (63) 0.1416 (64) 0.0149 (65) 86.66 5 23 0.87630 0.41330 55.3 34.2
SOS [34] 0.2588 (57) 0.2000 (58) 1.5247 (61) 82.08 6 24 0.35052 0.16532 69.7 43.2
LSFSA [24] 0.4962 (18) 0.3113 (60) 1.7354 (65) 77.10 8 25 0.52578 0.24798 255 158
GAMS 0.8131 (12) 1.4447 (61) 0.2896 (64) 72.09 8 26 0.52578 0.24798 63.8 39.5
26 27 0.70104 0.33064 170 105.4

Fig. 9. Electrical configuration of the 27-node test system.

Fig. 8. Power losses reduction of different methods in the 69-node test feeder.

This situation occurs because PV generators only inject power

during sunny hours, which makes them unusable during the
night period.

6. Conclusions

An exact mathematical model to represent the optimal location

and sizing of DGs in radial distribution networks, using a MINLP
representation, was presented in this paper. Such mathematical
model was solved by the GAMS optimization package, via compact Fig. 10. Percentage of power consumption and availability on a typical sunny day in
formulation through the BONMIM nonlinear large-scale discrete the Caribbean region of Colombia.
 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418 417

Table 7 [3] Rezaee Jordehi A. Allocation of distributed generation units in electric power
Power losses per day in the 27-node test feeder when different numbers of PV systems: a review. Renew Sustain Energy Rev 2016;56:893–905. doi: https://
generators are installed. doi.org/10.1016/j.rser.2015.11.086.
[4] Abdelaziz A, Ali E, Elazim SA. Optimal sizing and locations of capacitors in
Number of PV DGs Losses [kWh/day] Location radial distribution systems via flower pollination optimization algorithm and
power loss index. Eng Sci Technol, Int J 2016;19(1):610–8. doi: https://doi.org/
1 1936.20 1.520 (20)
10.1016/j.jestch.2015.09.002.
2 1831.58 1.321 (10) - 1.008 (16) [5] Iqbal F, Khan MT, Siddiqui AS. Optimal placement of DG and DSTATCOM for
3 1731.30 1.128 (10) - 0.975 (16) - 1.234 (20) loss reduction and voltage profile improvement. Alexandr Eng J. https://doi.
org/10.1016/j.aej.2017.03.002.
[6] Mahdad B, Srairi K. Adaptive differential search algorithm for optimal location
of distributed generation in the presence of SVC for power loss reduction in
distribution system. Eng Sci Technol, Int J 2016;19(3):1266–82. doi: https://
doi.org/10.1016/j.jestch.2016.03.002.
[7] Galiveeti HR, Goswami AK, Choudhury NBD. Impact of plug-in electric vehicles
and distributed generation on reliability of distribution systems. Eng Sci
Technol, Int J 2018;21(1):50–9. doi: https://doi.org/10.1016/
j.jestch.2018.01.005.
[8] Elmitwally A. A new algorithm for allocating multiple distributed generation
units based on load centroid concept. Alexandr Eng J 2013;52(4):655–63. doi:
https://doi.org/10.1016/j.aej.2013.08.011.
[9] Sadiq A, Adamu S, Buhari M. Optimal distributed generation planning in
distribution networks: a comparison of transmission network models with
FACTS. Eng Sci Technol, Int J. https://doi.org/10.1016/j.jestch.2018.09.013.
[10] Grisales Noreña LF, Restrepo Cuestas BJ, Jaramillo Ramirez FE. Location and
sizing of distribted generation: a review. Ciencia e Ingeniería Neogranadina
Fig. 11. Analysis of power losses reduction by increasing the number of the PV 2017;27(2):157–76. doi: https://doi.org/10.18359/rcin.2344.
systems. [11] Keane A, Ochoa LF, Borges CLT, Ault GW, Alarcon-Rodriguez AD, Currie RAF,
et al. State-of-the-art techniques and challenges ahead for distributed
generation planning and optimization. IEEE Trans Power Syst 2013;28
(2):1493–502. doi: https://doi.org/10.1109/TPWRS.2012.2214406.
solver, and compared with multiple methodologies reported in the [12] Theo WL, Lim JS, Ho WS, Hashim H, Lee CT. Review of distributed generation
specialized literature, which confirmed its efficiency and accuracy (DG) system planning and optimisation techniques: comparison of numerical
in terms of power losses reduction. Additionally, an illustrative and mathematical modelling methods. Renew Sustain Energy Rev
2017;67:531–73. doi: https://doi.org/10.1016/j.rser.2016.09.063.
example of the GAMS implementation in a small test feeder was [13] Pesaran M, A H, Huy PD, Ramachandaramurthy VK. A review of the optimal
presented to show its ease of use for solving nonlinear optimiza- allocation of distributed generation: objectives, constraints, methods, and
tion models. algorithms. Renew Sustain Energy Rev 2017;75(October):293–312. doi:
https://doi.org/10.1016/j.rser.2016.10.071.
Furthermore, an extension of the static MINLP model for opti- [14] Dixit M, Kundu P, Jariwala HR. Incorporation of distributed generation and
mal location and sizing of distributed generators in radial dis- shunt capacitor in radial distribution system for techno-economic benefits.
tributed networks was simulated to assess the possibility of Eng Sci Technol, Int J 2017;20(2):482–93. doi: https://doi.org/10.1016/
j.jestch.2017.01.003.
addressing, through GAMS implementations, daily operation prob-
[15] Ellabban O, Abu-Rub H, Blaabjerg F. Renewable energy resources: current
lems with renewable energy resources, as in the case of PV gener- status, future prospects and their enabling technology. Renew Sustain Energy
ators. This extension allowed an evaluation of the impact of PV Rev 2014;39:748–64. doi: https://doi.org/10.1016/j.rser.2014.07.113.
[16] Parhizi S, Lotfi H, Khodaei A, Bahramirad S. State of the art in research on
location and sizing on total energy losses during a typical sunny
microgrids: a review. IEEE Access 2015;3:890–925. doi: https://doi.org/
day in an electrical system in the Caribbean region in Colombia. 10.1109/ACCESS.2015.2443119.
In the future, the proposed MINLP model and its solution in [17] Kaur S, Kumbhar G, Sharma J. A MINLP technique for optimal placement of
GAMS can be used to size renewable generators (e.g., PV and wind multiple DG units in distribution systems. Int J Electr Power Energy Syst
2014;63(Supplement C):609–17. doi: https://doi.org/10.1016/j.
generators) with variable power factor capacities in radial test ijepes.2014.06.023.
feeders. In addition, such model will be modified to include capac- [18] Abido MA. Optimal power flow using tabu search algorithm. Electr Power
itors and battery energy storage systems for islanded microgrid Comp Syst 2002;30(5):469–83. doi: https://doi.org/10.1080/
15325000252888425.
applications. [19] HassanzadehFard H, Jalilian A. A novel objective function for optimal DG
allocation in distribution systems using meta-heuristic algorithms. Int J Green
Energy 2016;13(15):1615–25. doi: https://doi.org/10.1080/
7. Financial support 15435075.2016.1212355.
[20] Bohre AK, Agnihotri G, Dubey M. Optimal sizing and sitting of DG with load
models using soft computing techniques in practical distribution system. IET
This work was funded in part by the Administrative Department Gener Transm Distrib 2016;10(11):2606–21. doi: https://doi.org/10.1049/iet-
of Science, Technology, and Innovation of Colombia (COLCIENCIAS) gtd.2015.1034.
through its National Scholarship Program, under Grant 727-2015; [21] Moradi M, Abedini M. A combination of genetic algorithm and particle swarm
optimization for optimal DG location and sizing in distribution systems. Int J
in part by Instituto Tecnológico Metropolitano de Medellín, under Electr Power Energy Syst 2012;34(1):66–74. doi: https://doi.org/10.1016/j.
Project P17211; in part by Universidad Tecnológica de Bolívar, ijepes.2011.08.023.
under Projects C2018P020 and C2019P011; and in part by Univer- [22] D.P.R.P., V.R.V.C., G.M.T., Ant Lion optimization algorithm for optimal sizing of
renewable energy resources for loss reduction in distribution systems. J Electr
sidad Nacional de Colombia, under Proyect ”Estrategia de transfor- Syst Inf Technol 2018;5(3): 663–80. https://doi.org/10.1016/j.jesit.2017.06.
mación del sector energético Colombiano en el horizonte de 2030 - 001.
Energética 2030” - ”Generación distribuida de energía eléctrica en [23] Gandomkar M, Vakilian M, Ehsan M. A genetic based tabu search algorithm for
optimal dg allocation in distribution networks. Electr Power Comp Syst
Colombia a partir de energía solar y eólica” (Code: 58838, Hermes: 2005;33(12):1351–62. doi: https://doi.org/10.1080/15325000590964254.
38945). [24] Injeti SK, Kumar NP. A novel approach to identify optimal access point and
capacity of multiple DGs in a small, medium and large scale radial distribution
systems. Int J Electr Power Energy Syst 2013;45(1):142–51. doi: https://doi.
References org/10.1016/j.ijepes.2012.08.043.
[25] Sultana S, Roy PK. Krill herd algorithm for optimal location of distributed
generator in radial distribution system. Appl Soft Comput 2016;40:391–404.
[1] Prakash P, Khatod DK. Optimal sizing and siting techniques for distributed
doi: https://doi.org/10.1016/j.asoc.2015.11.036.
generation in distribution systems: a review. Renew Sustain Energy Rev
[26] Sultana S, Roy PK. Oppositional krill herd algorithm for optimal location of
2016;57:111–30. doi: https://doi.org/10.1016/j.rser.2015.12.099.
distributed generator in radial distribution system. Int J Electr Power Energy
[2] Bawan EK. Distributed generation impact on power system case study: losses
Syst 2015;73:182–91. doi: https://doi.org/10.1016/j.ijepes.2015.04.021.
and voltage profile. In: Power engineering society summer meeting, 2000.
IEEE, vol. 3; 2000. p. 1645–56.
418 O.D. Montoya et al. / Ain Shams Engineering Journal 11 (2020) 409–418

[27] ChithraDevi S, Lakshminarasimman L, Balamurugan R. Stud Krill herd [44] Jamian J, Mustafa M, Mokhlis H. Optimal multiple distributed generation
Algorithm for multiple DG placement and sizing in a radial distribution output through rank evolutionary particle swarm optimization.
system. Eng Sci Technol, Int J 2017;20(2):748–59. doi: https://doi.org/10.1016/ Neurocomputing 2015;152:190–8. doi: https://doi.org/10.1016/j.
j.jestch.2016.11.009. neucom.2014.11.001.
[28] Grisales-Noreña LF, Gonzalez-Montoya D, Ramos-Paja CA. Optimal sizing and [45] Kollu R, Rayapudi SR, Sadhu VLN. A novel method for optimal placement of
location of distributed generators based on PBIL and PSO Techniques. Energies distributed generation in distribution systems using HSDO. Int Trans Electr
2018;11(1018):1–27. Energy Syst 2014;24(4):547–61. doi: https://doi.org/10.1002/etep.1710.
[29] Mohanty B, Tripathy S. A teaching learning based optimization technique for
optimal location and size of DG in distribution network. J Electr Syst Inf
Technol 2016;3(1):33–44. doi: https://doi.org/10.1016/j.jesit.2015.11.007.
[30] Sudabattula SK, M K. Optimal allocation of solar based distributed generators Oscar D. Montoya received his BEE, M.Sc. and Ph.D
in distribution system using Bat algorithm. Perspect Sci 2016;8:270–2. doi: degrees in Electrical Engineering from Universidad
https://doi.org/10.1016/j.pisc.2016.04.048. recent Trends in Engineering and Tecnológica de Pereira, Colombia, in 2012 and 2014
Material Sciences. respectively. His research interests include mathemati-
[31] Yammani C, Maheswarapu S, Matam SK. A Multi-objective Shuffled Bat cal optimization, planning and control of power sys-
algorithm for optimal placement and sizing of multi distributed generations tems, renewable energies, energy storage, protective
with different load models. Int J Electr Power Energy Syst 2016;79:120–31. devices and smartgrids.
doi: https://doi.org/10.1016/j.ijepes.2016.01.003.
[32] Othman M, El-Khattam W, Hegazy Y, Abdelaziz AY. Optimal placement and
sizing of voltage controlled distributed generators in unbalanced distribution
networks using supervised firefly algorithm. Int J Electr Power Energy Syst
2016;82:105–13. doi: https://doi.org/10.1016/j.ijepes.2016.03.010.
[33] Behera SR, Dash SP, Panigrahi BK. Optimal placement and sizing of DGs in
radial distribution system (RDS) using Bat algorithm. In: 2015 International
conference on circuits, power and computing technologies [ICCPCT-2015]. p.
1–8. doi: https://doi.org/10.1109/ICCPCT.2015.7159295.
[34] Nguyen TP, Dieu VN, Vasant P. Symbiotic organism search algorithm for Walter Gil-González received his BEE and M.Sc. degrees
optimal size and siting of distributed generators in distribution systems. Int J in Electrical Engineering from Universidad Tecnológica
Energy Optim Eng 2017;6(3):1–28. doi: https://doi.org/10.4018/ de Pereira, Colombia, in 2011 and 2013 respectively. He
IJEOE.2017070101. is currently studying a Ph.D in Electrical Engineering at
[35] Nekooei K, Farsangi MM, Nezamabadi-Pour H, Lee KY. An improved multi- Universidad Tecnológica de Pereira, Colombia. His
objective harmony search for optimal placement of DGs in distribution research interests include mathematical optimization,
systems. IEEE Trans Smart Grid 2013;4(1):557–67. doi: https://doi.org/ planning and control of power systems, renewable
10.1109/TSG.2012.2237420. energies, energy storage, protective devices and smart-
[36] S.C., A.T., Optimal power flow using Moth Swarm Algorithm with Gravitational
grids.
Search Algorithm considering wind power. Future Gener Comput Syst.
https://doi.org/10.1016/j.future.2018.12.046.
[37] Nojavan S, Jalali M, Zare K. An MINLP approach for optimal dg unit’s allocation
in radial/mesh distribution systems take into account voltage stability index.
Trans Electr Eng 2015;39(E2):155–65. doi: https://doi.org/10.22099/
ijste.2015.3488.
[38] Engelmann A, Muhlpfordt T, Jiang Y, Houska B, Faulwasser T. Distributed AC
optimal power flow using ALADIN, IFAC-PapersOnLine 2017;50(1):5536–41. Luis F. Grisales received his BEE and M.Sc. degrees in
In: 20th IFAC World Congress.https://doi.org/10.1016/j.ifacol.2017.08.1095. Electrical Engineering from Universidad Tecnológica de
[39] Montoya OD, Grajales A, Garces A, Castro CA. Distribution systems operation Pereira, Colombia, in 2013 and 2015 respectively. He is
considering energy storage devices and distributed generation. IEEE Latin Am currently studying a Ph.D in Engineering at Universidad
Trans 2017;15(5):890–900. doi: https://doi.org/10.1109/TLA.2017.7910203. Nacional de Colombia. Actually, is professor in the
[40] GAMS Development Corp., General Algebraic Modeling System. <https://www. Instituto TecnolÓgico Metropolitano de Medellín,
gams.com/download/>.
attached to the Department of Electromechanics and
[41] Castillo E, Conejo A, Pedregal P, García R, Alguacil N. Building and solving
mechatronics, member of the research group MATyER.
mathematical programming models in engineering and science, pure and
applied mathematics: A Wiley Series of Texts, Monographs and Tracts. Wiley; His research interests include mathematical modelling,
2001. optimization techniques, planning and control of power
[42] Montoya OD, Garces A, Castro CA. Optimal conductor size selection in radial systems, renewable energies, energy storage, power
distribution networks using a mixed-integer non-linear programming electronic and smartgrids.
formulation. IEEE Latin Am Trans 2018;16(8):2213–20. doi: https://doi.org/
10.1109/TLA.2018.8528237.
[43] Montoya OD. Solving a classical optimization problem using GAMS optimizer
package: economic dispatch problem implementation. Ingenieria y ciencia
2017;13(26):39–63. doi: https://doi.org/10.17230/ingciencia.13.26.2.