机构地区: 广东工业大学应用数学学院
出 处: 《计算机应用》 2007年第1期216-218,共3页
摘 要: 提出了一种求解约束优化问题的混合遗传算法。它不是传统的在适应值函数中加一个惩罚项,而是在初始种群、交叉运算和变异运算过程中,把违反约束条件的个体用外点法处理设计出新的实数编码遗传算法。数值实验证明,新算法性能优于现有其他进化算法,是通用性强、高效稳健的方法。该方法兼顾了遗传算法和外点法的优点,既有较快的收敛速度,又能以非常大的概率求得约束优化问题全局最优解。 A new hybrid genetic algorithm was presented to handle constrained optimization. The traditional technique of a penalty term being added to the fitness function was not used, but the external point method was taken to keep those infeasible solutions created during the process of population initiation, crossover and mutation in the feasible region, and a new real-cede genetic algorithm was proposed. The new approach was compared against other evolutionary optimization techniques in several benchmark functions. The results obtained show the hybrid genetic algorithm is a general, effective and robust method. Its performance outperforms some other techniques. The new method has paid attention to both the advantages of external point method and genetic algorithms. It not only has a rather high eonvelgence speed, but also can locate the global optimum with a rather large probability.
领 域: [自动化与计算机技术] [自动化与计算机技术]