摘要
许多仿真实验的结果说明混合遗传算法在求解约束优化问题时不仅实用而且具有较高的效率。因此提出一种构造特殊罚函数以及改进选择算子、交叉算子、变异算子以及融合随机方向法、修复策略的求解约束问题的混合遗传算法框架。绝大多数文献中,用遗传算法求解约束优化问题时很少用有关理论去研究所改进的算法是否收敛,仅根据仿真实验得到的数据去下结论。为此,用概率论和数列极限有关理论研究了所提出的混合遗传算法收敛的几个条件,最后证明了所给的改进遗传算法将以概率1收敛到全局最优解,并且与初始种群无关。
Many simulation experiment results show that the hybrid genetic algorithm in solving constrained optimization problems not only practical but also has higher efficiency. Therefore a kind of special penalty function in construction is puts formard, it improves the selection operator, crossover operator, mutation operator and random direction method, repairs strategy framework of hybrid genetic algorithm for solving constraint problems. Most of the papers when using genetic algorithm to solve constrained optimization problem seldom use related theory to research whether the improved algorithm is convergent or not, only according to the data obtained from the simulation experiments to decide the conclusions. Therefore, using probability theory and the related theory of sequence limit to study several conditions of convergence which is proposed hybrid genetic algorithm. Finally, it is proved that the improved genetic algorithm can convergence to the global optimal solution with probability 1, and has no relation with the initial population.
出处
《自动化技术与应用》
2015年第9期6-9,共4页
Techniques of Automation and Applications
基金
广东省教育研究院研究课题(GDJY-2014-B-b085)
关键词
约束优化
遗传算法
收敛
全局最优解
constrained optimization
genetic algorithm
convergence
global optimal solution
