Abstract: An improved Genetic Algorithm (GA) based on the virtual population technology was proposed to avoid the frequent occurrences of premature convergence,and this improved algorithm can efficiently jump out of the local optimum. The parameter value of the key genetic operators,such as the selecting,the crossing and the mutating,in the improved GA were unchanged with the increased generation. I}his can avoid the problem of the slow convergence, which was often caused by the improper strategy of changing the parameter value of the genetic operators. The diversity of the real population was increased implicitly by the informational exchange between the real population and the virtual one.Finally, the simulation result indicated that, with the same size population, the improved Virtual Population Genetic Algorithm (VPGA) can jump out of the local optimum within the least generations,and also,can convergent to the globaloptimum at the minimum generation.

Key words: Virtual population, Genetic algorithm, Premature convergence