【作者】 杨海清；

【导师】 黄德才；

【作者基本信息】 浙江工业大学 ， 计算机应用技术， 2004， 硕士

【摘要】 遗传算法是模拟自然界生物进化过程与机制求解问题的一类自组织与自适应的人工智能技术，已广泛应用于计算机科学、人工智能、信息技术及工程实践，但在算法性能方面还有不足之处，因而对其进行改进研究具有理论意义和实用价值。 国内外对遗传算法已经作过许多改进研究。目前，遗传算法的主要问题有两个：早熟收敛和开采能力差。相应的解决策略是：维持种群个体多样性和增强对局部领域的搜索开采能力。 针对遗传算法存在的问题，在总体解决思路的指引下，本文在三个方面对遗传算法进行改进研究。 一、将自适应策略引入种间竞争遗传算法。该方法不仅运用交叉算子和变异算子的自适应调节技术协调种内进化过程，而且通过种间竞争频率的自适应调节促进最优个体的生成。 二、将小生境技术和单纯形方法融入遗传算法中提出一种新的基于小生境的混合遗传算法。该方法一方面运用小生境技术增强遗传算法“探测”能力，另一方面通过使用单纯形搜索方法提高遗传算法的“开采”能力，从而有效消除遗传算法的两大弱点。 三、针对多模态函数优化问题中的小生境半径变化差异悬殊的特点，给出一种小生境半径自动确定的设计方法。算例仿真表明了该方法的有效性。 最后，应用改进的遗传算法求解供水泵站效率优化问题，验证了改进算法的有效性。更多还原

【Abstract】 Genetic algorithm is an artificial intelligence technology of self-organization and adaptation, which imitates natural organisms’ evolutionary process and mechanism to solve problems. Although the algorithm has been widely applied to the computer science, artificial intelligence, information technology and engineering projects, but it still suffers from performance deficiency. So more studies should be carried out to improve its performance because of its large value in theory research and practical application.Presently, the algorithm still suffers from two drawbacks, premature convergence and weak exploitation capabilities. The effective ways to overcome such problems are to maintain the population diversity and enhance exploitation of local search domains.By using the idea mentioned above to solve such problems, three modified schemes of genetic algorithm are proposed in this paper from following aspects:1. A new method of connecting adaptive techniques with genetic algorithm based on the competition between populations is proposed. The method not only can coordinate evolutionary process inside each population by using an adaptive regulation of crossover and mutation operators, but also can advance the formation of-best design by using adaptive adjustment of the competition frequency between populations.2. A new niche hybrid genetic algorithm is proposed, which organically merges the niche technique and simplex method into genetic algorithm. The proposed method not only makes the exploration capabilities of genetic algorithm stronger through niche techniques, but also has more powerful exploitation capabilities by using simplex method.So it effectively alleviates the two major drawbacks of genetic algorithm.3. A new technique for calculating niche radius automatically is proposed, according to the character of the large difference of the niche radii in multi-modal function optimization problems. A set of benchmark functions is used to demonstrate the validity of the method.Finally, the first method mentioned above is applied to the efficiency optimization of a water-supply bumping station, and the simulation reveals its reliability.更多还原