【作者】 孙冲；

【导师】 李士勇；

【作者基本信息】 哈尔滨工业大学 ， 控制科学与工程， 2011， 硕士

【摘要】 2000年提出的混合蛙跳算法是一种新的智能优化算法,它的基本思想来源于文化基因传承,其显著特点是具有局部搜索与全局信息混合的协同搜索策略。经过大量仿真测试表明,解决高维、病态、多局部极值等函数问题混合蛙跳算法具有优越性,是一种行之有效的优化技术。本文从基本混合蛙跳算法的原理、步骤以及参数选择对算法的影响等方面出发,通过多个具有典型特征的无约束基准测试函数对混合蛙跳算法进行仿真测试,并与粒子群算法、遗传算法两经典算法对比,来考查算法的优化性能。混合蛙跳算法虽然具有易理解,参数少等优点,但是存在不易跳出局部最优,收敛速度慢等问题。针对这些不足,本文对算法提出三种改进思想:首先设计了两种基于算法局部搜索策略的改进算法。一种是利用种群最优个体和局部最优个体二者以一定的随机比例同时影响文化基因体中的最差青蛙个体,建立新的个体进化公式,来更新最差个体位置。另一种针对局部搜索的每次独立进化同时更新几个适应度较差的个体,使文化基因体整体快速的到达最优位置。这两种算法均针对局部搜索部分的单一进化给予改进,仿真结果表明,两种改进策略提高了算法的求解速度,有效性也有所加强。然后将量子算法与混合蛙跳算法相结合,提出了量子蛙跳算法。它用量子位的概率幅构造青蛙个体,量子旋转门改变量子比特的相位以更新个体,量子非门对种群最优个体进行变异,最终实现寻优。经过测试认为它的收敛速度快,求解成功率高,有效性好。为了验证量子蛙跳算法和基本混合蛙跳算法在控制工程领域的优化性能,用它们解决PID控制器参数整定和模糊控制器参数优化问题。通过与粒子群算法、遗传算法的仿真结果对比,认为量子蛙跳算法在控制领域的应用具有让人满意的表现。更多还原

【Abstract】 Shuffled Frog Leaping Algorithm which proposed on the 2000-year, is a new kindof intelligent optimization algorithm. Its basic idea comes from the cultural geneticinheritance and its notable feature is a collaborative search strategy that is a mixture oflocal search and global information. Lots of simulation tests show that, Shuffled FrogLeaping Algorithm is a superlative and effective optimization technology when thefunction is high-dimension, sick, and more local optimum.The paper describes the principles of the algorithm, steps, and key parameters,verifies the algorithm by use of the typical characteristics of unconstrained testfunction, and compares advantageously with particle swarm algorithm and geneticalgorithms.Shuffled frog leaping algorithm, while having the advantages of easy tounderstand and fewer parameters, is not easy to jump out of local optimum, andconverges slowly. So three improved methods are proposed:Firstly, two improved algorithms based on the local search strategy are designedin the paper. One is using of the optimal crop individual and local population, bothaffect the worst frog genome individual at a certain proportion, the individualevolution formula is established and updates the worst individual location. The other,some poor fitness individuals are updated by use of independent evolution of localsearch. Two algorithms both improve against the single evolutionary of the localsearch part. The simulation results show that the two strategies improve the algorithmsolution speed. Certainly, the effectiveness has also been strengthened.Then, combined with the quantum algorithm and shuffled frog leaping algorithm,the quantum leapfrog algorithm is proposed. It uses the probability of the quantum bitto structure the frog individual, the revolving door changes quantum bit phase toupdate the individual, the quantum NOT gate to verify the optimal individual, andultimately achieve the optimization. It has a high convergence speed, a high solvingsuccess rate and good effectiveness.To verify the capability of the quantum leapfrog algorithm and the basic mixedleapfrog algorithm to optimize the control engineering problems, the paper use them to tuning PID controller parameter and optimize fuzzy controller parameters. Comparedwith particle swarm optimization and genetic algorithms simulation results, thequantum leapfrog algorithm has satisfactory performance in control engineering field.更多还原