Digital Signal Processing Reference
In-Depth Information
process of evolution. These algorithms are nevertheless extremely efficient, and
are used in fields as diverse as stock exchange, production scheduling or pro-
graming of assembly robots in the automotive industry.
Below the selected aspects of genetic and evolutionary algorithms are outlined.
Special attention is paid to their application for protection and control in power
systems. The theory is then followed by selected examples of genetic optimization
of ANNs as well as genetic search for optimal settings of over-current relays.
13.1 Basics of Evolution and Genetics
for Technical Problems
Evolutionary and genetic algorithms are a result of emulation (numerical imple-
mentation) of the evolution principles observed in the nature, with particular
concern given to natural selection appearing in the population of living beings.
Evolution is a natural process leading to the maintenance of a population's ability
to survive and reproduce in a specific environment. This ability (quality) is called
evolutionary fitness that can also be viewed as a measure of the organism's ability
to anticipate changes in its environment. Emulation of natural evolution is an
iterative process involving creation a population of individuals, evaluating their
fitness, generating a new population through genetic changes and repeating the
steps a number of times.
Generally speaking, among the stochastic search methods that mimic the
metaphor of natural biological evolution one can distinguish:
• Genetic Algorithms (GAs), the most general optimization approach with binary
coding and full range of genetic operators) [
11
].
• Evolutionary strategies (ESs, use natural problem-dependent representations,
and primarily mutation and selection as search operators) [
4
].
• Evolutionary programing (EPs, the solutions are in the form of computer pro-
grams, and their fitness is determined by their ability to solve a computational
problem) [
8
,
16
].
In this chapter the GAs are only described; however, the examples provided
later belong both to GA and ES groups of methods.
The GAs are mainly applied for solving various optimization problems. It can
be said that genetic optimization iteratively improves the quality of solutions until
an optimal, or at least feasible, solution is found. From the other traditional
optimization approaches the GAs are singled out with the following features:
• The GA does not process the problem parameters directly but in a coded form.
• Searching for an optimum is performed commencing not from a single starting
point but from a certain population of initial guesses.
• The genetic optimization process is controlled by suitably defined goal function.
• Probabilistic selection rules are applied.
Search WWH ::
Custom Search