一种基于多种群的退火DNA遗传算法

Multi- Population Simulated Annealing DNA Genetic Algorithm

研究多种群算法优化问题,针对传统单种群遗传算法易产生早熟收敛、局部搜索能力弱等问题,提出一种多种群的退火DNA遗传算法。首先,将DNA计算思想引入遗传算法的编码和遗传操作算子的设计中,采用主种群、辅助种群和由主种群、辅助种群每次迭代产生的最优个体组成的精英种群在不同的进化策略下协同进化,然后通过种群间的个体交叉来实现种群交流。并可引入模拟退火机制,防止算法陷入局部最优,加强算法的局部搜索能力。将改进后的算法应用于函数优化测试中,并与其它改进遗传算法进行比较。仿真结果表明,改进算法在全局寻优能力、算法稳定性方面具有明显的优越性。

Aiming at the issues such as premature convergence and poor local search ability of single population SGA, a multi-population simulated annealing DNA genetic algorithm was proposed. First of all, DNA computing was introduced into parameter coding and genetic operators designing. A main population, an adjuvant population as well as an elite population which is made up of the best individuals produced by main population and adjuvant population in every generation were adopted in this paper, evolving in coordination with each other in different evolution strategy. Then the communication of populations waaas realized by means of crossover of individuals. Besides, simulated an- nealing mechanism was introduced to avoid being trapped in local optimum and improve the local search ability. The improved algorithm was applied to function optimization test and compared with other improved genetic algorithms. The simulation results of function optimization show that the algorithm has obvious superiority in global optimization and stability of algorithm.