Abstract:
Ant colony optimization (ACO) is a nature inspired population-based strategy widely used to solve optimization problems. ACO algorithms have been applied to solve many real-life problems from a wide area of applications. The performance of ACO algorithms shows their competitiveness and applicability in solving real-world optimization problems. ACO algorithms were originally developed to solve combinatorial optimization problems. Combinatorial optimization problems belong to the category of problems where each decision
variables assume its state from a nite set of states. In the last decade, many attempts
have been made to apply ACO algorithms for solving continuous optimization problems.
Continuous optimization problems are problems where the decision variables assume a
real value within the predefined limits. However, all the available ACO algorithms have
a tendency to converge at suboptimal points. Also, minimal effort has been made to explore the ability of ACO algorithms in solving constrained optimization problems, that is, optimization problems where the feasible region is reduced and even may be broken
into feasible subregions due to certain restrictions are known as constraints. In this work a novel strategy is proposed to extend ACO algorithm, so that it can be applied to solve large scale continuous optimization problems, constrained optimization problems, and
complex real-life problems. Firstly, we demonstrate the extension of ACO to solve unconstrained
continuous optimization problems. We explain the proposed algorithm, discuss
and analyze the different design components. Along with that, we show various testing and simulations results. We check the performance of the proposed algorithm by comparing its results on standard benchmark problems with other state-of-the-art algorithms
available in the literature applied to solve considered benchmark problems. Finally, we investigate how our algorithm performs on eight unconstrained real-life problems, coming from the various fields of applications. On the basis of the outcome of performance
analysis of proposed ACO on unconstrained continuous optimization problems, we discuss the design choices made to make the proposed algorithm capable of tackling constrained
optimization problems. We incorporate various constrained handling techniques available
in the literature with the proposed ACO and analyze and compare their performances.
Finally, the performance of the proposed algorithm is investigated on complex constrained
real-life problems coming from the engineering design. The first real-life problem is model
order reduction problem in which the complexity of the complex dynamical system is
reduced in such a way that their input-output behavior is as much preserved as possible.
The next application is the problem of most economical generation of demanded power,
known as Economic Load Dispatch (ELD)" problem. The third and last application is
the problem of secure power distribution in the presence of distributed generators (DGs)
using overcurrent relay. In this application directional overcurrent relays (DOCRs) are
coordinated in such a way that the time taken by relays to isolate the fault line from the
rest of the system as soon as possible.