Metaheuristics for Intelligent Electrical Networks

2Population-based Methods

2.1. Introduction

Metaheuristics, also known as stochastic algorithms, are characterized by the introduction of the notion of randomness. Contrary to exact methods that explore the search space by making use of different techniques such as enumeration and tree-based methods, stochastic algorithms visit only part of the search space in a random fashion, therefore reaching a satisfactory approximated solution in a reasonable time. Exact methods guarantee, under certain conditions, the optimality of the solution obtained, which is not the case for metaheuristics.

The term “solution population-based metaheuristics” refers to algorithms that manipulate a set of solutions whose size is predefined. Indeed, after initializing the starting population, the different solutions communicate with each other and evolve throughout iterations, generating a new population until the stopping criterion is satisfied. The search for the optimum is thereby achieved by manipulating a number of solutions that travel through the search space according to their own information as well as the information that other solutions transmit with a very specific mechanism. Moreover, metaheuristics based on populations of solutions are characterized by the fact that these favor exploration, that is that they broaden their search over several regions of the search space.

In this chapter, we describe a selection of algorithms based on populations of solutions by separating them into two ...