INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

Methods And Solvers Used For Solving Mixed

Integer Linear Programming And Mixed
Nonlinear Programming Problems: A Review
Pujari. Harish Kumar, Dr. R. Mageshvaran

Abstract: This paper presents a complete review of the significance of deterministic mixed -integer linear program (MILP) and mixed-integer nonlinear
program (MINLP) solution methods for problems involving linear, nonlinear, convex and nonconvex functions. The mathematical d escription of
methodologies, algorithms, software, and solvers to solve this problem are presented. Mixed-integer program (MIP) problem is one where some of the
decision variables are constrained to be integer values. MILP involves problems in which only some variables are constrained to be integers, while other
variables can be non-integers. It deals with linear objective function subjected to linear constraints. Minimisation and maximisation of a nonlinear
objective function subject to nonlinear constraints with continuous and discrete variables are performed by MINLP solution methods. In this paper, the
recent MILP and MINLP solution methods, algorithms along with solvers and software, introduced by various researchers has bee n elaborated
profoundly.

Index Terms: Mixed-integer linear programming, Mixed-integer nonlinear programming, optimization solvers, Convex and Nonconvex, Combinatorial
optimization, Global optimization.
——————————  ——————————

1 INTRODUCTION Here, belotti also describes about the tree search method
In general, all the optimization problems that are available in which is one of the widely used method to solve MINLP
the current scenario, necessitates the basic need for modelling problems. The tree search method has been classified into
the continuous variables and discrete variables using the single tree and multi-tree method. These two classes of
Mixed Integer Linear Program (MILP) and Mixed Integer Non- methods solve a problem involving function like the convex
Linear Program (MINLP) methods. Both the objective type. The classical single tree method uses nonlinear branch
functions and constraints in MILP problems should be and cut method, cutting plane method, and branch and bound
considered as linear, whereas, in the case of MINLP, they method for solving MINLP problems. The conventional multi-
should be considered as a non-linear variable. Furthermore, tree approach comprises of both outer approximation and
the optimization problems related to MINLP are classified benders decomposition methods, in order to obtain global
mainly into two groups. The first one is a convex MINLP which optimal solution for the considered problem. By combining the
deals with the minimization of convex objective function with a above approaches, a new hybrid method has been proposed
feasible convex region. The second group is a non-convex in the literature, for solving convex MINLP problems more
MINLP which indeed deals with the non-convex objective effectively, in terms of computation time and the quality of
function. At the same time, it also deals with the values that optimal solutions. In [11], Bonami et al. have provided a
have convex nature, but unidentified in the feasible area. Refs. summary of various convex MINLP algorithms and software
[1-4] describes MILP techniques elaborately and utilised the used to solve multiple MINLP problems. It is too challenging to
same for solving their optimization problems. To solve the solve nonconvex MINLP problems because it consists of
MILP problems, different advanced methods had been devised nonconvex functions and nonlinear constraints in the objective
experimentally in [5]. However, many researchers have function. Methods like Spatial branch and bound Piecewise
developed a few commercial and non-commercial solvers to linear approximation, and Generic Relaxation Strategies are
solve both the MILP and MINLP problems. CPLEX [6], Gurobi used to solve nonconvex MINLP problems. In [12],
[7], and XPRESS-MP [8] are the some of the solvers that are Tawarmalani and Sahinidis N.V have explained about global
developed and utilised in the literature. In [9] John K.Karlof optimization theory for solving MINLP problems using different
have discussed the integer programming theory for solving algorithms and solvers.
both MILP and MINLP problems. The methods used to solve
MINLP problems has significant advantages when compared 2 MIXED INTEGER LINEAR PROGRAM (MILP)
to the methods that are used to solve MILP and Nonlinear An integer programming (IP) problem denoted as a
programming (NLP) problems. In [10], Belotti et al. have mathematical optimization process and a viability program in
provided a comprehensive review of MINLP problems and also which few or all of the variables are limited to be integers. In
about the methods to solve them. [13], if decision variables are not discrete, then the problem is
considered as a MIP problem.

———————————————— 2.1 Linear Programming (LP)

• P. Harish Kumar, Research Scholar, School of Electrical Linear Programming (LP) technique has applied for optimizing
Engineering, Vellore Institute of Technology, Vellore, Tamilnadu, a linear objective function, which is subject to both the linear
India, E-mail: pujari.harishkumar2017@vitstudent.ac.in.
equality and inequality constraints.
• Dr.R. Mageshvaran, Associate Professor, School of Electrical
Engineering, Vellore Institute of Technology, Vellore, Tamilnadu, The canonical representation of a linear integer program is
India, E-mail: rmageshvaran@vit.ac.in. (Corresponding author stated as in [14]:
Maximize 𝑐 𝑥 , Subject to 𝐴𝑥 ≤ 𝑏, 𝑥 ≥ 0, and 𝑥 ∈ ℤ (1)

1872
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

 The standard representation for an Integer linear value and optimal solution.
programming (ILP) has stated as 2.3.3 Branch and Price Algorithm
Maximize 𝑐 𝑥 , Subject to 𝐴𝑥 + 𝑠 = 𝑏, 𝑠 ≥ 0, 𝑥 ≥ A hybrid algorithm developed with the combination of
0, and 𝑥 ∈ ℤ methods: branch and bound method along with the column
(2) generation known as Branch and price algorithm. This
Whereas 𝑥 is the vector of variables to be computed. 𝑐 and 𝑏 algorithm has utilized for solving the ILP problems with many
are the known coefficient vectors. 𝐴 represents a known variables associated with the problem. For obtaining the
constant matrix in which all the entries are integers. optimal solution, this algorithm uses the current set of
variables that has preserved, and the column generation is
2.2 Mathematical representation of MILP done according to the requirement of linear programming
A Mixed Integer Programming (MIP) problem performs both problem. In [15], the column generation procedure has used
minimize and maximize of problems with a linear objective for a given specific problem, and this can be again used to
function which is subject to linear constraints. But MILP interrelate with branching decisions to obtain the lower bound
involves problems with variables that are constrained to be value.
only integers and allow all to be non-integers. The general
representation of MILP is express as 2.4 Basic components of MILP solvers
Min { 𝑐 𝑥 𝐴𝑥 ≥ 𝑏, 𝑥 ≥ 0, 𝑥 ∈ ℤ }. (3) The main essential parts of MILP solvers are Pre-solving,
For example, if variables are said to be rational, then the set is Cutting plane generation, Sophisticated branching strategies,
said to be null, which represents an LP problem solved by Primal Heuristics in MILP, Variable selection, and Parallel
using the polynomial-time method. If variables are considered implementation.
to be integers, then it denotes a MIP problem.
2.4.1 Pre-solving
2.3 Algorithms used for solving ILP problems In the pre-solving phase of MILP, solvers make an effort to
Mainly, the underlying algorithms implemented for solving the identify individual variations in the input, which helps in
ILP problems are listed as follows: achieving the improved execution of the solution strategy,
 Branch and Bound algorithm. without changing the optimal solution for the problem. Pre-
 Branch and Cut algorithm. solving denotes to a group of problem reductions that have
 Branch and Price algorithm. used before the start of the branch and bound procedure, and
this collection makes to construct its formulation and reduce
2.3.1 Branch and Bound Algorithm the size of the problem. The MILP models are converted into
This methodology is used to find the solution for ILP problems. problem formulation, effectively by removing excess data (like
And the subproblems are formed according to the limiting variables and constraints) and by reinforcing variable integrity.
range of integer variables in the program. Only two values are One main advantage of this pre-solving method is to provide a
assigned to binary variables, one variable is 0, and the other is smaller LP relaxation process that helps in solving the
1. The lower bound value represents as L and an upper bound problems quickly and for improving the approximating of the
value as U, which are considered to be variable that is convex integer solution. Finally, more sophisticated pre-solve
separated into two problems having a range from ‗L to m‘ and mechanisms are available to find the necessary inferences
‗m+1 to U‘. In general, the LP relaxation method is being and substructures for the computation of branching
used to obtain lower bound values for a given problem, to determination and the cutting plane generation. Martin in [16]
satisfy both the objective function and limitations of constraints and Savenlsbergh in [17] has clearly explained pre-solving
for an optimal solution. If the obtained optimal solution is methods to solve MILP problems.
integral, then it is represented as a solution to a subproblem.
The search for the subproblem whose lower bound is higher 2.4.2 Cutting plane algorithm
can be terminated using these values [15]. Ralph E.Gomory developed this algorithm in 1950 and used
for solving both ILP and MIP problems. The basic idea of the
2.3.2 Branch and Cut Algorithm algorithm is to cut off some part of the feasible regions in the
In this algorithm, the LP relaxation process is being utilized to solution obtained by the LP relaxation process, in such way
estimate the lower bound values. The optimal solution for the optimal integer solution becomes an extreme point and can be
ILP problem denotes the values in the feasible region, which found by using the simplex method [18]. The cutting plane
satisfies the constraints. The step by step process of the algorithms include different groups such as Chvatal Gomory
branch and cut algorithm is: cuts, mixed-integer rounding cuts, lift-and-project cuts,
Step 1: If obtained optimal solution by LP relaxation process is Gomory mixed-integer cuts, and split cuts [19].
not integral, then the algorithm search for the constraint, which The basic form of formulating the cutting plane algorithm is as
is violated by this optimal solution and not violated by optimal follow:
integer solutions. The constraint is known as cutting plane. Step 1: If the solution of the LP problem is optimal, and if
Step 2: If the old obtained optimal solution by the LP relaxation the feasible region of the solution doesn‘t contain a
method is not valid, then the cutting plane constraint values line, then easily an extreme point is obtained, which is
are being added to the LP problem. The new optimal solution optimal.
will be dissimilar, and this provides a better lower bound value. Step 2: If an obtained optimum is not an integer solution,
Step 3: The cutting plane algorithm runs continuously until an a linear inequality exists which separates the optimum
optimal integral solution obtained; else, it is difficult to get the from the convex hull of the exact feasible set.
cutting plane for the problem. In [15], another technique known
as the traditional branch method is used to obtain lower bound
1873
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

Step 3: Finding such inequality is called a separation methods construct the solution for the IFS by using partial
problem and is considered as cutting line in the solutions. The transformation of the solution obtained from the
feasible region. LP relaxation method into IFS has achieved by applying the
Step 4: Then, the obtained current non-integer solution is heuristics methods at the internal nodes and root node of the
no longer feasible for the relaxation process, and the search tree in MILP solver.
above steps continuously repeated until an optimal
integer solution is obtained. 2.4.5 Parallel Implementation
When the process of search tree node independent,
2.4.3 Sophisticated Branching Strategies parallelization is a must for the search tree. The branch and
At any step of processing, the branching mechanism requires bound algorithm are used widely for these purposes [24]. The
two independent and significant decisions, one is node different categories of parallel integer programming
selection, and the second is variable selection. The branching investigation are sorted depending on the type of parallel
strategies involves choosing a variable 𝑌 , with a fractional computing architecture that existed. The Message transferring
value 𝑌 , in the optimal solution to the current LP relaxation to communicate solution of the algorithm depends on
process. Then constraints for the variable are formulated as [ distributed memory architecture. Shared memory computers
𝑌 ≤ 𝑌 ] and [ 𝑌 ≥ 𝑌 ]. move the data between the Central Processing Unit (CPU) by
 Node Selection: Doran and Michie proposed the node reading from, and writing to a common memory pool. In the
selection process in 1966 based on the evaluation parallel branch and bound method, the order in which node
function. The different procedures used by Rafael calculations performed may cause an impact on the execution
Pastor, Albert Corominas to the node selection are as of the program and sometimes lead to abnormal behaviour
follow [20]: [25].
 British Museum Procedure.
 Random Search [Rich and knight in 1991]. 2.5 MILP Optimization software packages
 Depth First Search (Liao in 1994) or Single Branch A complete review of the features of both commercial and non-
Search (Ibaraki in 1976). commercial MILP software packages has presented in [26].
 Breadth-First Search (Dijkstra and Moore in 1959). The history of this MILP software package development and
specifications like SCICONIC, UMPIRE, and MPSX/370, etc.,
 Best Bound Search (Ibaraki 1976/1987).
are evaluated by Forrest and Tomlin in [27]. With the help of
 Best Few Search (Agusti and Lopez in 1987).
these package set up, several successful methods have
 Parallel Depth First Search (Imai in 1982). formulated for handling the problems related with the
 Jump Backtracking (Fisher and Hochbaum,1980). branching and node selection, out of which the significant
 Locally Best Bound Rule (Liao in 1994). techniques are still in use for solving MILP. In [19] & [20] some
 Weighted -A* algorithm (Korf in 1993). fundamental changes as having been advanced in MILP
 Depth m-Search (Ibaraki in 1978). techniques to solve many major problems in the engineering
 Variable Selection: This selection problem decides field. The advanced methods used like the pre-processing
how to perform the partition of the current node method and structure-specific cutting plane method are used
consequently, which variable to branch on it, such that even today for solving many of the engineering problems.
to create two children. The variable selection mainly Some authors in [28,29] represented the inequalities which are
done based on branching rules which have presented also utilised efficiently as a branch and cut method, that can
in [21] are listed as the most fractional variable, active be accompanied in the cutting plane procedure. In [30]
branching variable, pseudo cost branching variable, detailed history of MILP software over a past period is
and branching priorities. Basically, in the variable explained clearly by R.Bixby.
selection problem, a traditional choice is to branch on
a variable with the fractional part closest to 0.5. In 2.6 User Interface
[22], Pseudocost branching was another sophisticated The user interface is an essential feature for designing
technique used for the variable selection problem software for solving MILP problems. As the extent of purposes
implemented by Benichou. This technique keeps track for MILP programming software is generally enormous, as a
of success variables which are already branched on. result of this user interface types number is correspondingly
becoming large.
Another method developed by Linderoth in [23], for the
variable selection problem based on the strong branching 2.7 Commercial software packages
technique. In this technique at any node, each fractional The following commercial software packages used to solve
variable has to branch on for checking progress, before actual MILP problems are available for academic users at free (no
branching is done. cost) and limited cost licensing options. Han Mittelmann has
executed independently many MILP software to get optimal
2.4.3 Primal Heuristics in MILP solutions and had published results for the past many years.
This method provides solutions and improvement for finding Table .1 provides information about the MILP Optimization
the Integer Feasible Solutions (IFS) in the search tree and the Software Packages.
upper bound value for a minimization problem. There are two
heuristic methodologies implemented in MILP solver, which
are rounding and iterative rounding heuristic. Heuristic

1874
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

TABLE 1
MILP OPTIMIZATION SOFTWARE PACKAGES
Types of
MILP Interfaces,
Software Algorithms Parameters
Optimizatio Founders Features Specifications modelling
name utilized included
n Software languages
Packages.
Capable of
Bixby the Branch and calculating
CPLEX [31] founder of cut multiple Version: 12.8.0, Website: C, C++, Java,
(IBM ILOG CPLEX, algorithm optimal http:// www.ibm.com/ .NET, MATLAB,
Mipemphasis
CPLEX retained and and solutions and analytics/cplex-optimizer. Python, Microsoft
meta parameter
Optimization provided by Dynamic the solutions License: proprietary. Excel.
Studio) IBM. search have stored
algorithm. in a solution
pool.
New MILP
solver that is
Include designed
Zonghao Gu, cutting with modern Object-oriented
Version: 3.0, Website:
Edward planes multicore interfaces
MIP-Focus meta www.gurobi.com.
GUROBI Rothberg, algorithm, processing for C++, Java, .NE
parameter License: proprietary.
and Robert heuristics technology to T, and Python.
Bixby. and search obtain an
Commercial techniques. optimal
Software solution.
Packages
It offers
Significantly
different
LINDO also Faster on
forms of
LINDO [32] comprises a Large
cutting
(Linear, mipemphasismet Quadratic Version:10.0, Website:
LINDO planes C, Visual Basic,
Interactive, a parameter that Models. www.lindo.com.
SYSTEMS algorithms MATLAB,
and has used for Improved License: proprietary.
INC. and Microsoft Excel.
Discrete adjusting Handling of
different
Optimizer) algorithm Models with
node
parameters. Discontinuou
selection
s Functions.
rules.
Parameters MOSEK C, C++, Java and
Branch and
include optimizer interior-point Python languages.
bound,
choice for optimizer can Mosek is
branch and
solving linear reliably Version:9 beta, Website: accessible for use
Mosek ApS, a cut, and
problems, detect a www.mosek.com. License: by customers
MOSEK Danish state-of-the-
turning pre-solve primal and proprietary. through a GAMS
company. art interior-
parameter value dual interface on the
point
and infeasible NEOS Server.
optimizer
feasibility of status of
algorithm.
tolerances value. solutions.
COIN-OR
linear
programmin
g solver,
cutting
Open source planes It controls the
solver algorithm Parameters to suitable
BLIS(
developed by and parallel solve problems methods to Version:0.91, Website:
BiCePS
the COIN-OR tree search are taken from progress a https://projects.coin.or.org/ C++ library is
Linear
project to algorithms the COIN-OR state-of-the- CHiPPS. similar to
Integer
Non- solve MILP are used to linear art parallel License: Common Public SYMPHONY.
Solver),
Commercial problems. solve MILP programming algorithm for License.
Software (COR@L problems, solver and CGL. a particular
Packages Lab). available problem set.
[33] from the
Coin Cut
Generation
Library
(CGL).
John Forrest, COIN-OR CBC can be
Parameters
an open- linear parallelized
taken to solve Version:2.5, Website:
CBC source solver, programmin using
problems are https://projects.coin.or.org/CB
(COIN-OR developed by g solver, shared-
taken from the C. C++
Branch and COIN-OR cutting memory
COIN-OR linear License: Common Public
Cut Solver) (Computation planes parallelism to
programming License.
al algorithm solve the
solver and CGL.
Infrastructure and parallel problem for
1875
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

for tree search finding an

Operations algorithms optimal
Research.) are used to solution.
project to solve MILP
solve MILP problems,
problems. available
(COR@L from CGL.
Lab).
GLPK
utilizes the
simplex
method and
the primal-
Problems
dual
have
interior-
modelled in
point
the GNU
method to
GLPK [34] A set of language, Version: 4.44, Website:
solve non-
(GNU Andrew O. subroutines with MathProg http:// www.gnu.org/
integer GNU, MathProg,
Linear Makhorin, a callable library shares the software/glpk/,
problems. C.
Programmin GNU Project. and black box syntax with License: GNU General Public
The branch
g Kit) solver has used. AMPL, and is License (GPL).
and bound
solved with
algorithm,
standalone
including
solver
Gomory
GLPSOL.
mixed-
integer, cuts
to solve
integer
problems.
It offers
constraint
classification
s, pre-
processing,
primal
heuristics
and
Branch and Global and constraints
Savelsbergh Version:3.1, Website: C, MINTO 1.4 is
MINTO bound local constraints generation
and http://coral.ie.lehigh.edu/mint accessible on
(Mixed algorithm are valid at any automatically
Nemhauser, o/, SUN SPARC
Integer with LP node and .
1993. License: Provided by the station with
Optimizer) relaxations constraints are MINTO is
library only. CPLEX.
method. generated. accessible
for users
through an
AMPL
language
interface, on
the NEOS
Server.

Branch and
cut It supports about
The strategy
SCIP [35] Konrad Zuse- algorithm, 20 constraint
of SCIP has
(Solving Zentrum fur and Branch types for MILP, Version: 1.2, Website: C, C++, GAMS,
constructed
Constraint Information and price MINLP, and http://scip.zib.de/, MATLAB, JAVA
on the
Integer stechnik algorithm mixed integer all ZIB Academic License. PYTHON.
concept of
Programs) Berlin (ZIB). used to quadratic
constraints.
solve programming
problems.
Sequential
Ability to
and parallel Version:5.6,5.6.17,
The user of the Solve C, AMPL, GMPL,
types of the Website: https://projects.coin-
library can bi-objective GAMS (General
branch, cut, or.org/SYMPHONY/,
SYMPHON Developed change the MILPs and Algebraic
and price www.coin-or.org,
Y [36] by COIN-OR. algorithm using perform Modelling
algorithms License: Common Public
custom data basic System).
are used to License (CPL), open-source
files. sensitivity
solve MILPs solver.
studies.
problems.
continuous variables and nonlinear functions in the objective
3 MINLP INTRODUCTION AND MATHEMATICAL function and constraints. MINLP includes non-linear
REPRESENTATION OF MINLP programming (NLP) and MILP as subproblems while modelling
MINLP optimization problems contain both the discrete, a system for obtaining an optimal solution. The general
1876
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

formula of an MINLP model representation has expressed as These lower bound problems are known as a master
follow: problem. For solving the subproblem solution, first the
= 𝑓(𝑥, ), 𝑔(𝑥, ) ≤ 0, 𝑥 ∈ , ∈ 𝑌 ( ) (4) binary variables are fixed, and the next 𝑓𝑥 has
Where 𝑓( ) 𝑑 𝑔( ) are differentiable functions, 𝑥 is a computed. If the solution obtained for 𝑓𝑥 is a
continuous variable and are discrete variables. feasible one, then it gives an upper bound value
( 𝑒𝑟 𝑒).If the obtained solution is not feasible,
MINLP has expressed as: and then the feasible-MINLP method is used to solve the
𝑥 𝑓( ) = ( 𝑓( )) subproblem to get upper bound value. This process has
(5) performed until tolerance exists between lower and upper
Generally, the integer variables are limited to 0, and 1 bound values.
values and constraint function are naturally linear [37]. There  Generalized Benders Decomposition (GBD): GBD is
are two nonlinear programming subproblems obtained from almost similar to the outer approximation process; a slight
MINLP, used in many solving methods. difference exists in the linear master problem formulation
 Continuous relaxation of MINLP(r-MINLP): It is stated [43]. The master problem formulation considers discrete
as variables and inequality variables.
= 𝑓(𝑥, ), 𝑔(𝑥, ) ≤ 0, 𝑥 ∈ , ∈ 𝑌_(𝑅) (𝑟 )  Extended Cutting Plane (ECP): The ECP method also
(6) follows the same concept as an outer approximation, but
Where 𝑌_(𝑅) denotes the continuous relaxation of 𝑌, this method does not solve the nonlinear programming
integer variables has considered as continuous. subproblems [44].In this process solution obtained from a
 Fixed MINLP (𝑓𝑥 ): The following expression master MILP, ( ), ( ) contains all the constraints
is used when nonlinear programming subproblems having that are linearized, and then linearized constraints are
fixed , violated have been added to the master problem. When
= 𝑓 (𝑥, ) , 𝑔(𝑥, ) ≤ 0, 𝑟 ∈ 𝑅 (𝑓𝑥 ) (7) violation lies is almost of the specified tolerance value,
when the solution is feasible, the upper bound and lower convergence has achieved. After each iteration algorithm
bound for MINLP are obtained by (𝑓𝑥 ) and gives a lower bound value. The main disadvantages of
(𝑟 ) respectively. this algorithm are convergence is slow, and it doesn‘t
provide an upper bound value, feasible solution until it
3.1 Classifications of MINLP achieves required convergences value [45].
MINLP generally classified into two types, such as convex
MINLP and nonconvex MINLP. 3.2 Convex MINLP optimization solvers
Many convex MINLP solvers are used to optimization problem
3.1.1 Convex MINLP solving in [46,47]. Some essential convex MINLP solvers have
Convex MINLP includes a minimization problem with the discussed below.
objective function around a feasible convex region. The  Basic open source nonlinear mixed integer
different methods used for solving Convex MINLP has programming (BONMIN) [48]: This is a free solver to the
described below. users for solving MINLP problems, using methods like
 Nonlinear Branch and Bound Method. outer approximation, branch and bound method,
 Outer Approximation Method. extended cutting plane, NLP branch and bound method.
 Generalized Benders Decomposition Method and  Discrete and Continuous Optimizer (DICOPT) [49]: This
 Extended Cutting Plane Method. is a commercial solver to the users for solving the
 Nonlinear Branch and Bound Method: It might have MINLP problems by using an outer approximation
considered as the advanced method to the linear algorithm that has interfaced with GAMS.
branch and bound method. This method uses an  Mixed Integer Linear and Nonlinear Optimization [50]:
extensive tree search process for obtaining an optimal This is an open-source solver used to solve nonlinear
solution over an integer variable [38-40]. programming (NLP) problems by using nonlinear branch
and bound, and outer approximation algorithms in
Step 1: Initially, it solves the continuous relaxation method on MATLAB software.
MINLP problems.  Mixed-Integer Nonlinear Programming Branch and
Step 2: If the solution assigns integer values to all integer Bound (MINLPBB) [51]: MINLPBB is a commercial
variables using the linear relaxation method, then it is said to solver to users for solving the MINLP problems by using
be an optimal solution, and then the algorithm stops working. generalized bender decomposition and outer
Step 3: And if integer variable values have not assigned at approximation algorithms.
the current node, then it is chosen as [𝑉 = (𝑉 (0))].  Mixed-Integer Nonlinear Optimization Toolkit Algorithms
Step 4: Next branching method is applied to these integer Under Estimators and Relaxation (MINOTAUR): This is
variables, forms a new NLP problem having bound value also an open-source toolkit used by users for solving
represented as [𝑉 ≤ (𝑉 (0))] 𝑑 [𝑉 ≥ (𝑉 (0))] . convex MINLP problems by using nonlinear branch and
Step 5: The above steps have repeated until an optimal bound algorithm interfaced with C++ [52].
integer feasible solution is obtained.  Alpha Extended cutting plane(α-ECP) [53]: This is a
 Outer Approximation: This method employs two problems commercial based solver used for solving MINLPs by
such as Master-MIP (M-MIP) and Fixed-MINLP (𝑓𝑥 using the extended cutting plane method.
) [41,42]. The M-MIP have used for finding a lower  Simple Branch and Bound (SBB): This solver has
bound value ( 𝑒𝑟 𝑒) and also used to find out executed in GAMS, which allows the users to select an
integer solution for the given estimated problem ( )
1877
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

NLP solver for solving MINLP problems by using the method [70]. MILP problem can be solved quickly, extremely
NLP branch and bound algorithm. fast and effectively using commercially existing solvers such
 Branch and Reduce Optimization Navigator as CPLEX, Gurobi, Xpress and MATLAB [71].
(commercial solver) [54].
 Algorithms for Continuous Global Optimization of 3.4 MINLP and MILP applications
nonlinear equations [55]. MINLPs has widely used in the field of engineering and
 Convex Over and Under Envelopes for Nonlinear scientific applications. MINLP models of electrical engineering
Estimation (open sources solver) [56]. applications comprise the efficient management of electricity
 Solving constraint integer programs (non-commercial transmission in [72], transmission expansion in [73],
solver) [57]. transmission switching and contingency analysis in [74] and
blackout prevention of electric power systems in [75].
3.3 Non-Convex MINLP Chemical engineering applications of MINLP models comprise
The most complex and dynamic fields in optimization are non- the design of water in [76], gas distribution networks in [77],
convex MINLP. For the exact modelling of the real-time and for the optimal design of reliable chemical plants in [78].
problems, the nonconvex constraints are used for obtaining Other applications MINLP model comprise like operational
the optimal solution is necessary [58,59]. The spatial branch reloading of nuclear reactors in [79] and electric distribution
and bound algorithm are the deterministic methods used for networks in [80]. In [81], MINLP models have used for
solving the non-convex MINLP problems. The main concept stochastic service system design problems. MILP has used in
related to the nonconvex solving process involves relaxations various fields of engineering and scientific applications.
of factorable formulations [60,61] and bounds tightening [62- Electrical engineering applications of MILP include optimizing
65]. In [66,67], a complete explanation about deterministic and distributed generation locations and sizes, considering
heuristic techniques has presented. different DGs modes of connection to the distribution grid in
[82]. In [83] S-MILP operational model is proposed for
3.3.1 Spatial branch and bound algorithm minimizing the sum of the most associated cost terms
The spatial branch and bound algorithm is a deterministic (switching costs, emissions, expected costs of operation, and
method used for solving non-convex MINLP problems and by unserved power in the system), by satisfying the number of
using a convex relaxation process to find the optimal solution. model constraints. In [84] MILP based technique is
implemented for smart self-healing grids. In [85],
Step1: This algorithm calculates convex relaxation of the NLP reconfigurable computing, architectures and design methods
problem which affords the lower bound value. are developed based on MILP models. In [86], approximated
Step 2: Next, a local or open-source solver is used to solve the MILP models are used to AC transmission expansion planning.
NLP problem. If the NLP problem has a feasible solution, then In [87] MILP model is used to reliability optimization in active
it gives an upper bound value. distribution networks. In [88] MILP branch flow model is used
Step 3: if the solution is not feasible, then the upper bound in synchronize AC multistage transmission expansion and
value is assumed as infinity. reactive power planning with security constraints. In [89] MILP
Step 4: If the differences between the upper bound and lower model is proposed for integrating expansion planning of multi-
bound values are within the tolerance limits, the algorithm carrier active energy systems. In [90] MILP model is
stops executing and provides an optimal solution for the implemented in the charging coordination problem of electric
problem. vehicles in unbalanced distribution systems.
Step 5: Otherwise, the two new NLP problems have formed by
using branches of variables. Later, each NLP problem 4 CONCLUSION
computed using the solver to obtain both upper bound and In this paper an overview of the deterministic MILP and MINLP
lower bound value by using previous steps. The upper bound solution methods with the mathematical description of
has taken from the new region, which is having the lowest methodologies, algorithms, software, and solvers had been
upper bound value. presented. MIP problem is one of the linear programs where
Step 6: This process is repetitive until the upper bound and the variables are constrained as integer. However, MILP
lowers bound is within the tolerance limits. involves problems with some variables, constrained to be
integers and other as non-integers. Whereas, MINLP solution
3.3.2 Non-Convex MINLP solvers method for problem performs both minimizing and maximizing
Many MINLP solvers have been built to solve the nonconvex a nonlinear function subject to nonlinear constraints. Here,
problems by using the branch and bound algorithm. The various MILP and MINLP methods and algorithms introduced
solvers used to solve the nonconvex MINLP are α-branch and by the researchers along with solvers and software that are
bound solver (α-BB), Convex over and under envelopes for used to solve the real time problems in the field of engineering
nonlinear estimation (COUENNE), Branch and reduce and science had been discussed in detail.
optimization navigator (BARON) and Lipschitz global optimizer
(LGO). The main advantage of the MILP over MINLP is that 5 ACKNOWLEDGMENT
convergence of the solution is definite for the given problem The authors are extremely grateful to the Chancellor, Vice-
[68]. Otherwise, if the obtained solution is optimal, then it is Chancellor and Vice-Presidents, of VIT University, Vellore, for
easy to determine the optimum local value for the MILP providing the excellent infrastructure facilities and
problems. In case if the constraints oppose each other or if the encouragement which have made this research work possible.
problem is abundant in the way of the objective function, then
the optimal solution cannot be determined [69]. The solution
accuracy can be specified using the branch and bound gap
1878
IJSTR©2020
www.ijstr.org
 INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

6 REFERENCES (PDF),‖ ChE 4G03.Optimization in Chemical

[1] Harold. Greenberg, ―8 Branch and Bound Engineering, pp.1-7,2018.
Procedures,‖ Mathematics in Science and [14] C.H. Papadimitriou, K.Stieglitz,
Engineering Elsevier, vol. 76, pp. 177-191, ―Combinatorial optimization: algorithms and
1971. complexity,‖ Mineola.Publisher: Dover
[2] Alexander. Schrijver, ―Theory of Linear and Publications.Unabridged edition,pp.1-
Integer Programming,‖ A Wiley-Interscience 496,1998.
Publication, John Wiley & Sons, Inc., 605 [15] L.N.George ,et al., . ―Integer and
Third Avenue, New York, NY 10158-0012, Combinatorial Optimization,‖ Publication: John
USA;1999. Wiley and Sons, Wiley-Interscience New York,
[3] Nemhauser. George, et al. ―Integer and NY, USA.ISBN:0-471-82819-X, pp.31-
Combinatorial Optimization,‖ publisher: Wiley- 32,1998.
Interscience, Wiley, ISBN:0-471-82819-X, [16] A.Martin ,‖ General mixed integer
New York, NY, USA;1988. programming: Computational issues for
[4] Robert J. Vanderbei. ―An Interior Point Code branch-and-cut algorithms,‖ Computational
for Quadratic Programming,‖ Journal Combinatorial Optimization,2241 of Lecture
Optimization Methods and Software, vol.11, Notes in Computer Science,pp.1–25,2001.
no.1-4, pp.451-484, 1999. [17] M.P.W. Savelsbergh, ―Pre-processing
[5] R. Bixby and E. Rothberg, ―Annals of and probing techniques for mixed integer
Operations Research,‖ Kluwer Academic programming problems,‖ ORSA Journal on
Publishers-Plenum Publishers, vol.149, no.1, Computing, vol.6, pp.445-454, 1994.
pp.37-41,2007. [18] R.E. Gomory, ―Outline of an algorithm
[6] IBM ILOG CPLEX Optimization Studio CPLEX for integer solutions to linear programs,‖
User‘s Manual, Version 12 Release 6, tech. Bulletin of the American Mathematical Society,
rep.,2013, vol.64, pp.275–278,1954.
https://www.ibm.com/support/knowledgecenter [19] G. Cornuejols, ―Valid inequalities for
/SSSA5P_12. mixed integer linear programs,‖ Mathematical
[7] Gurobi solver.Gurobi Optimizer Reference Programming, vol.112, pp.3–44, 2008.
Manual 2018.Version 8.1.www.gurobi.com. [20] Rafael Pastor, Albert Corominas.
[8] FICO XPRESS-MP Reference Manual.tech. ―Strategies of Node Selection in Search
rep., 2009. Procedures for Solving Combinatorial
[9] John K.Karlof. ―Integer programming theory Optimization Problems: A Survey and a
and practice. eBook‖, CRC Press Taylor and General Formalization,‖ Sociedad de
francis group, pp.1–305,2006. Estadistica e Investigacidn Operativa, vol.8,
[10] J.Kronqvist and Bernal, E. David and no.1, pp.111-134,2000.
Lundell, Andreas and I.E. Grossmann, ―A [21] T.Achterberg, T.Koch, et al.,
review and comparison of solvers for ―Branching roles revisited,‖ Operations
convex MINLP,‖ Optimization and Research Letters, vol.33,pp.42–54,2005.
Engineering, vol.20, no.2, pp.397-455, 2019. [22] M. Benichou,J.M Gauthier,et al.,
[11] P.Bonami, M. Kilinc, et al. , ―Algorithms ―Experiments in mixed-integer programming,‖
and Software for Solving Convex Mixed Mathematical Programming,vol.1, pp.76–
Integer Nonlinear Programs,‖ Chapter in Jon 94,1971. doi: 10.1007/BF01584074.
Lee et Sven Leyffer editors,Mixed Integer [23] J.T. Linderoth, M.P.W. Savelsbergh, ―A
Nonlinear Programming, IMA Volumes in computational study of search strategies for
Mathematics and its mixed integer programming,‖ INFORMS
Applications,vol.154,pp.1–39,2012. Journal on Computing, vol.11, pp.173–
[12] M.Tawarmalani, N.V. Sahinidis, 187,1999.
―Convexification and Global Optimization in [24] T.G. Crainic, B. Gendron, ―Parallel
Continuous and Mixed-Integer Nonlinear branch and bound algorithms: Survey and
Programming: Theory, Algorithms, Software, synthesis,‖ Operations Research, vol.42,
and Applications,‖ Kluwer Academic pp.1042–1066,1994.
Publishers, vol.65, pp.1-478,2002. [25] T.H.Lai, S. Sahni, ―Anomalies in
[13] C.Benoit, ―Mixed Integer Linear parallel branch and bound algorithms,‖ In
Programming (MILP): Model Formulation Proceedings of the 1983 International

1879
IJSTR©2020
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

Conference on Parallel Processing ,pp.183– Mixed Integer Nonlinear Programming,‖

190,1983. Computers and Operations Research, vol.21,
[26] T.L. Jeffrey, et al., ―MILP software,‖ pp.359–367,1994.
citeseerx.ist.psu.edu/view, pp.1-17,2001. [40] S. Mehrotra, R.A. Stubbs, ―A branch-
[27] J. J. H. Forrest,J.A. Tomlin, ―Branch and-cut method for 0-1 mixed convex
and bound, integer and non-integer programming,‖ Mathematical Programming,
programming,‖ Annals of Operations Research vol.86, pp.515–532,1999. doi:
,vol.149,pp.81–87,2007. 10.1007/s101070050103.
[28] E.Balas, S.Ceria et al., ―Gomory cuts [41] P. Kesavan , R.J. Allgor and
revisited,‖ Operations Research Letters, P.I.Barton, ―Outer approximation algorithms
vol.19,pp.1–9,1996. for separable non-convex mixed-integer
[29] R.E. Gomory, ―An algorithm for the nonlinear programs,‖ Mathematical
mixed integer problem,‖ Technical Report RM- Programming ,vol.100,pp.517–535,2004.
2597, The Rand Corporation 1960. [42] R. Fletcher,S. Leyffer, ―Solving mixed
[30] R. Bixby, E. Rothberg, ―Progress in integer nonlinear programs by Outer
computational mixed integer programming – a Approximation,‖ Mathematical
look back from the other side of the tipping Programming,vol.66,pp.149-327,1994. doi:
point,‖ Annals of Operations Research, 10.1007/BF01581153.
vol.149, pp.37–41,2007. [43] O.E. Flippo,A.H.G.R Kan,
[31] H.D Mittelmann, ―Recent Benchmarks ―Decomposition in general mathematical
of Optimization Software,‖ 22nd European programming,‖ Mathematical
Conference on Operational Research, vol.4, Programming,vol.60,pp.361–382,1993.
pp.1-314,2014. [44] T.Westerlund , F.Pettersson, ―An
[32] C.Nash, John, ―Optimizing Add-Ins: extended cutting plane method for solving
The Educated Guess. What's Best! PC convex MINLP problems,‖ Computers and
Magazine,‖ vol.10, pp.130-132,1991. Chemical Engineering,vol.19,pp.131–
[33] J.T Linderoth, T.K Ralphs, ―Non- 136,1995.doi:10.1016/0098-354(95)87027-X.
commercial software for mixed integer linear [45] J.E. Kelley, ―The cutting plane method
programming. In Integer Programming: Theory for solving convex programs,‖ Journal of the
and Practice,‖ CRC Press Operations SIAM, vol.8, pp.703–712,1960.
Research Series 2005, pp.253–303,2005. doi: [46] M.R. Bussieck, S. Vigerske, ―MINLP
10.1201/9781420039597.ch10. solver software,‖ Wiley Encyclopedia of
[34] Eiji Oki, ―Linear Programming and Operations Research and Management
Algorithms for Communication Networks: A Science 2010.
Practical Guide to Network Design Control and [47] C.D. Ambrosio, A.Lodi, ―Mixed integer
Management,‖ CRC Press 2012. nonlinear programming tools: a practical
[35] T. Achterberg, ―SCIP: solving overview,‖ 4OR ,vol.9,pp.329–349,2011. doi:
constraint integer program,‖ Mathematical 10.1007/s10288-011-0181-9.
Programming Computation, vol.1, no.1, pp.1– [48] P.Bonami , L.Biegler , et al., ―An
41,2009. doi: 10.1007/s12532-008-0001-1. algorithmic framework for convex mixed
[36] M.Guzelsoy, T.K Ralphs, ―The integer nonlinear programs,‖ Discrete
SYMPHONY callable library for mixed integer Optimization journal, vol.5,pp.186–204,2008.
programming,‖ In the Proceedings of the Ninth doi: 10.1016/j.disopt.2006.10.011.
Conference of the INFORMS Computing [49] J.Viswanathan , I.E. Grossmann, ―A
Society 2005, pp.61–71,2005. combined penalty function and outer-
[37] I.E. Grossmann, ―Review of nonlinear approximation method for MINLP
mixed-integer and disjunctive programming optimization,‖ Computers and Chemical
techniques,‖ Optimization and Engineering, Engineering, vol.14,pp.769–782,1990. doi:
vol.3, pp.227–252,2002. 10.1016/0098-1354(90)87085-4.
[38] S. Nabar, L. Schrage, ―Modelling and [50] H.Y. Benson, D.F. Shanno, ―An exact
Solving Nonlinear Integer Programming primal-dual penalty method approach to warm
Problems,‖ Presented at Annual AIChE starting interior point methods for linear
Meeting 1991. programming,‖ Computational Optimization
[39] B. Borchers, J.E. Mitchell, ―An and Applications, vol.38, pp. 371–399,2007.
Improved Branch and Bound Algorithm for doi: 10.1007/s10589- 007-9048-6.
1880
IJSTR©2020
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

[51] H.Y. Benson, D.F. Shanno, ―Interior- [63] M.W. Savelsbergh, ―Pre-processing
point methods for nonconvex nonlinear and probing techniques for mixed integer
programming: regularization and warm starts,‖ programming problems,‖ ORSA Journal on
Computational Optimization and Applications, Computing, vol.6, pp.445–454,1994.
vol.40, pp.143–189,2008. doi: [64] D. Erling,K. Andersen, ―Pre-solving in
10.1007/s10589-007-9089-x. linear programming,‖ Mathematical
[52] P.Belotti , Christian Kirches, et al., Programming ,vol.71,pp.221–245,1995.
―Mixed integer nonlinear optimization,‖ Acta [65] F.Messine, ―Deterministic global
Numerica journal,vol.22,pp.1-131,2013. optimization using interval constraint
[53] T.Westerlund , K. Lundqvist, ―Alpha- propagation techniques,‖ RAIRO Operations
ECP.Version 5.01 An interactive MINLP-solver Research, vol.38,pp.277–293,2004. doi:
based on the Extended Cutting Plane 10.1051/ro:2004026.
Method,‖ Process design laboratory, Abo [66] M. Gaviano,D.Lera, et.al., ―A Parallel
Akademie university 2001. Algorithm for Global Optimization Problems in
[54] Sahinidis, N.V. Baron, ―A general- a Distributed Computing Environment,‖
purpose global optimization software Applied Mathematics,vol.10A,pp.1380-
package,‖ Journal of Global Optimization, 1387,2012.doi:10.4236/am.2012.330194
vol.8, pp.201–205,1996. doi: [67] L. Liberti, ―Introduction to global
10.1007/BF00138693. optimization,‖ Technical report. LIX. Ecole
[55] R.Misener,C. Floudas, ―ANTIGONE: Polytechnique, Palaiseau F-91128, France,
Algorithms for continuous/integer global liberti @lix.polytechnique.fr 2008.
optimization of nonlinear equations,‖ journal of [68] Luca Urbanucci, ―Limits and potentials
global optimization, vol.59,pp.503– of Mixed Integer Linear Programming methods
526,2014.doi: 10.1007/s10898-014-0166-2. for optimization of poly-generation energy
[56] P.Belotti , J.Lee, ―Branching and systems,‖ Energy Procedia, vol.148, pp.1199-
bounds tightening technique for non-convex 1205,2018. doi:
MINLP,‖ Optimization methods and software, 10.1016/j.egypro.2018.08.021.
vol.24,no.4,pp.597-634,2009. doi: [69] Ortega Jordi, ―Modelling environment
10.1080/10556780903087124. for the design and optimization of energy poly-
[57] Berthold Timo, Stefan Heinz and Marc generation systems,‖ Ph.D. Thesis 2010.
E. P fetsch, ―Solving Pseudo-Boolean [70] Chapin, E. Jonathan, ―Review of
Problems with SCIP,‖ ZIB report 2008, vol.8, Integer Programming,‖ by A. Laurence,
no.12, pp.1–15,2008. Wolsey, Wiley and Sons 1998, journal:
[58] C.A. Floudas, ―Deterministic global SIGACT News 2005, vol.36, no.3, pp.10-
optimization: theory, methods, and 12,2005. doi:
applications,‖ eBook Springer, vol.37, pp.1- acm.org/10.1145/1086649.1086655
698,2000. [71] I.E. Grossmann, ―Optimization and
[59] I.E Grossmann, T.Francisco, ―Review Engineering,‖ Publisher: Kluwer Academic
of Mixed‐Integer Nonlinear and Generalized Publishers, vol.3, pp.227-252,2002.
Disjunctive Programming Methods,‖ Chemie [72] R.Bacher, ―The Optimal Power Flow
Ingenieur Technik journal ,vol.86,pp.991- (OPF) and its solution by the interior point
1012,2014.doi: 10.1002/cite.201400037. approach,‖ EES-UETP Madrid. Short Course,
[60] G.P. McCormick, ―Computability of pp.10-12,1997.
global solutions to factorable nonconvex [73] L.L. Garver,―Transmission network
programs: Part I Convex underestimating estimation using linear programming,‖ IEEE
problems,‖ Mathematical programming, vol.10, Transactions on Power Apparatus
pp.147–175,1976. Systems,vol.89,pp.1688–1697,1997.
[61] E.Smith , C.C Pantelides, ―Global [74] E.F. Bartholomew, R.P. O‘Neill, el.al.,
optimization of nonconvex MINLPs,‖ ―Optimal transmission switching,‖ IEEE
Computers and Chemical Transactions on Power Systems, vol.23,
Engineering,vol.21,pp.791–796,1997. doi: pp.1346–1355,2008.
10.1016/S0098-1354(97)87599-0. [75] D.Bienstock, S.Mattia , ―Using mixed-
[62] E.Davis, ―Constraint propagation with integer programming to solve power grid
interval labels,‖ Artificial intelligence,vol.32,no. blackout problems,‖ Discrete
3,pp.281-331,1987. Optimization,vol.4,pp.115–141,2007.
1881
IJSTR©2020
www.ijstr.org
INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 01, JANUARY 2020 ISSN 2277-8616

[76] R.Karuppiah , I.E. Grossmann, ―Global [84] T.Akbari , Mohammad, ―Approximated

optimization for the synthesis of integrated MILP model for AC transmission expansion
water systems in chemical processes,‖ planning: global solutions versus local
Computers & Chemical Engineering solutions,‖ IET Generation, Transmission
,vol.30,pp.650–673,2006. &amp Distribution,vol.10,no.7,pp.1563-
A. Martin,M.Moller, ―Mixed integer models 1569,2016,IET Digital Library.
for the stationary case of gas network [85] Aljohani Nader, Cody Ruben, ―MILP
optimization,‖ Mathematical Model for Reliability Optimization In Active
Programming, vol.105,pp.563– Distribution Networks,‖ 2018 North American
582,2009. Power Symposium (NAPS),2018.
[77] Yixin Ye, I.E. Grossmann, et.al., [86] Macedo, V. Christian, et.al., ―MILP
―Mixed-integer nonlinear programming models branch flow model for concurrent AC
for the optimal design of reliable chemical multistage transmission expansion and
plants,‖ Computers and Chemical reactive power planning with security
Engineering, pp.1-14,2018. constraints,‖ IET Generation, Transmission
A. J Quist, T. Terlaky, et.al., ―Application & Distribution,vol.10,no. 12,pp.3023-
of nonlinear optimization to reactor 3032,2016,IET Digital Library.
core fuel reloading,‖ Annals of Nuclear [87] Ghasemi, Hosein,et.al., ―A MILP Model
Energy, vol.26, pp.423–448,1998. for Integrated Expansion Planning of Multi-
[78] S.Lakhera , U.V Shanbhag U.V, et.al., Carrier Active Energy Systems,‖ IET
―Approximating electrical distribution networks Generation Transmission & Distribution,
via mixed-integer nonlinear programming,‖ IET Digital Library 2018, vol.13,no.7,pp. 1177 –
International Journal of Electric Power and 1189,2019.
Energy Systems, vol.33,no.2,pp.245– [88] F.John, Franco, Marcos, et al., ―A
257,2011. Mixed-Integer Linear Programming Model for
[79] S.Elhedhli, ―Service System Design the Electric Vehicle Charging Coordination
with Immobile Servers, Stochastic Demand, Problem in Unbalanced Electrical Distribution
and Congestion,‖ Manufacturing & Service Systems,‖ IEEE Transactions on Smart Grid
Operations Management, vol.8,no.1,pp.92– ,vol.6,no. 5,pp.2200-2210,2015.
97,2006.
[80] Teshome, F.Dawit, ―The 11 LIST OF ABBREVIATIONS:
comprehensive mixed-integer linear CPU Central Processing Unit
programming model for distribution system IFS Integer Feasible Solutions
reconfiguration considering DGs,‖ IET ILP Integer Linear Programming
Generation Transmission &amp IP Integer Programming
Distribution,vol.12,no. 20,pp.4515- LP Linear Programming
MILP Mixed Integer Linear Programming
4523,2018,IET Digital Library.
MINLP Mixed-Integer Nonlinear Programming
[81] F.V, Dantas, S.F Santos, et.al., MIP Mixed Integer Programming
―Dynamic Reconfiguration of Distribution NLP Nonlinear Programming
Network Systems: A Key Flexibility Option for S-MILP Stochastic Mixed Integer Linear Programming
RES Integration,‖2017 IEEE International
Conference on Environment and Electrical
Engineering and 2017 IEEE Industrial and
Commercial Power Systems Europe (EEEIC /
I&CPS Europe), 13 July 2017.
[82] Al-Owaifeer, Al-Muhaini, Mohammad,
―MILP-based technique for smart self-healing
grids,‖ IET Generation Transmission &amp
Distribution, vol.12, no. 10, pp. 2307-
2316,2018, IET Digital Library.
[83] T.J Todman,G.A Constantinides, et.al.,
―Reconfigurable computing: architectures and
design methods,‖ IEE Proceedings -
Computers and Digital Techniques,vol.152,no.
2,pp.193-207,2005, IET Digital Library.
1882
IJSTR©2020
www.ijstr.org