【作者】 孟庆莹；

【导师】 王联国；

【作者基本信息】 甘肃农业大学 ， 农业电气化与自动化， 2011， 硕士

【摘要】 混合蛙跳算法(SFLA)是由Eusuff和Lansey在2003年首次提出的,得到了国内外学者的广泛关注,对算法的研究应用已经渗透到多个领域,成为交叉学科中一个前沿性研究问题。混合蛙跳算法是一种新的基于全局协同搜索的智能优化方法,它结合了Memetic算法和粒子群优化算法两者的优点,是继粒子群优化算法之后的又一种新的群体智能优化算法。作为一种全新的仿生优化算法,具有概念简单、参数少、计算速度快、全局寻优能力强、易于实现等特点。但是SFLA还有很多不完善的地方,如:理论基础薄弱,研究成果及其应用较少,并且存在易收敛到局部最优、在求解部分函数优化问题时效果不够理想等。本文针对算法存在的缺陷提出一种基于邻域正交交叉算子的混合蛙跳算法(SFLA-OCO),并对该算法中主要参数对算法性能的影响、算法在电力系统负荷预测问题和TSP求解中的应用进行了分析,主要研究内容如下:1.简要介绍了最优化问题、智能优化算法以及混合蛙跳算法研究进展的相关内容,详细论述了混合蛙跳算法的基本原理、数学模型、算法流程、核心步骤以及算法特点。2.针对基本混合蛙跳算法在根据局部更新策略更新个体最差值Xw时带有一定的盲目性,寻优结果精度低收敛速度慢等缺点,对其进行了改进,提出了一种基于邻域正交交叉算子的混合蛙跳算法,改善新个体的部分分量值,从而提高种群的多样性,较好的平衡了算法的全局搜索能力和局部搜索能力,提高了算法的收敛速度和计算精度。3.通过仿真实验,分析了基于邻域正交交叉算子的混合蛙跳算法的主要参数对算法性能的影响,为进一步研究混合蛙跳算法提供了很好的参考依据。4.将基于邻域正交交叉算子的混合蛙跳算法用于支持向量机的参数优化,建立一种SFLA-OCO-SVM预测模型,并应用于电力系统短期负荷预测,取得了较好的效果。5.将基于邻域正交交叉算子的混合蛙跳算法用于求解TSP中,并取得了较好的效果,说明本文的改进算法能有效地解决大多数组合优化问题,具有良好的应用前景和实用价值。最后,对全文进行了总结并对混合蛙跳算法的研究进行了展望,由于该算法的理论分析还不够完善,许多问题有待于做进一步的研究。更多还原

【Abstract】 Shuffled Frog Leaping Algorithm (SFLA) was first proposed by Eusuff and Lansey in 2003, which have received wide attention from foreign scholars. At present, this algorithm research has already improved many other application and has already become extremely active from research question in the interdisciplinary studies. Shuffled frog leaping algorithm is a new intelligent optimization method based on global cooperative search, which combines the advantages of Memetic algorithm and Particle Swarm Optimization algorithm and it is a new swarm intelligence algorithm after the presence of PSO .As a new optimization algorithm, SFLA has been proved to have many advantages, such as simple concept、less parameters、fast convergence speed、better global search capability and easy implementation, and so on. However, its theoretical foundation is weak, less research and application, and there is easy to converge to a local optimum, some functions in solving optimization problems result is not satisfactory and other defects.Since SFLA has so much shortcomings, this paper proposes a Shuffled Frog Leaping Algorithm based on local orthogonal crossover operator(SFLA-OCO),and analyzed the influence of the main parameters to the algorithm. The improved algorithm is applied to power system load forecasting problem and solving TSP, the main contents are as follows:1.A brief introduction of intelligent optimization problems and intelligent optimization algorithm was made. The basic principles, mathematical model and algorithm flow of SFLA were discussed in detail. The research progress of AFSA was summarized. All of the above indicated the importance of researching the SFLA.2.Aiming at the low optimization precision、slow convergence speed and a certain degree of blindness when update the individual worst values, this paper proposes a Shuffled Frog Leaping Algorithm based on local orthogonal crossover operator(SFLA-OCO), the value of parts of the new individual was improved. Therefore, the new algorithm not only improved the convergence speed、the optimization accuracy and the diversity of population but also balanced the global search ability and local search ability successfully.3.Through experiments, parameters selection was researched in details, which provided the useful reference for the further study of the SFLA.4.The Shuffled Frog Leaping Algorithm based on local orthogonal crossover operato(rSFLA-OCO), which was given from reference, was used to be a solution for parameters optimization of support vector machines. The SFLA-OCO-SVM predict model established by this paper was applied into short-term load forecasting in power systems, which received preferable result.5.The improved SFLA was applied into solving the TSP, which received preferable result. The results show that the improved algorithm can effectively solve the most combinatorial optimization problems and has not only good prospects but also a wide range of practical value.In conclusion, this paper makes a summarization of the prospect of the SFLA. However, since the theoretical analysis of the algorithm is not perfect, many questions are needed to be further researched.更多还原