Genetic and Evolutionary Algorithms for PSP - Digital Signal Processing in Power System Protection and Control - page 274

CREATING
INTERMEDIATE
POPULATION

Program
parameters

(SELECTION)

No SELECTION

CRITERION
FULFILLED?

Ye s

GENETIC
MODIFICATIONS
-NEXT
GENERATION

INITIAL
POPULATION

No MAX. NUMBER

OF GENERATIONS
REACHED?

Ye s

FITNESS
EVALUATION

FITNESS
EVALUATION BEST

SOLUTION

Fig. 13.2 GA optimization loop

interruption can occur when preset level of the goal function is reached, if need
be—with given accuracy. The GA procedure can also be interrupted when its
further operation does not improve the optimization outcomes (no better results
can be expected). If the GA interruption conditions are satisfied the best chro-
mosome is accepted as final solution. Otherwise the algorithm operation is con-
tinued until certain time has passed or the prescribed number of iterations is
reached (Fig. 13.2 ).

13.1.1 Selection Versions

After the individuals of the population have been assessed the best of them are
selected for further processing (genetic changes). The intermediate population is
formed from the individuals selected according to one of the following methods
(not exclusive list):

• Wheel of roulette (individuals are ordered according to their fitness; the better
their chromosomes are, the more chances to be selected they have), Fig. 13.3 .
• Stochastic universal sampling (a number of NP parents are selected from the
ranked set of individuals), Fig. 13.4 .
• Truncation selection (first x% of individuals with the highest fitness are selected
for further breeding).
• Tournament selection (randomly selected pairs of individuals compete, the ones
with higher fitness win).

The idea of the wheel of roulette method is illustrated in Fig. 13.3 . In this
example the population consists of ten individuals represented by their chromo-
somes (Chr1-Chr10). The sizes of wheel sections correspond to the quality of
individuals, thus one may expect that the individuals with higher fitness would