并行多目标智能优化算法及其应用的研究
发布时间：2018-09-08 19:25
【摘要】：现实生活中最优化问题普遍存在,且往往涉及相互矛盾的多个目标。由于多项设计指标的引入,导致问题的搜索空间明显扩大,求解难度激增,因此,为求解计算量较大的多目标优化问题,充分利用智能优化算法的寻优能力和大规模高性能并行计算技术是切实可行的解决方案。本文基于非支配排序团队进步算法(NRTPA),引入自适应思想和并行化方案,提出了并行多目标团队进步算法。并行多目标团队进步算法将双群体演化机制和非支配排序的多目标策略相结合,以高效的寻优效率提升多目标解集的逼近性、均匀性和宽广性。根据测试函数集的直观验证和度量指标的定量对比,测试结果表明,并行多目标团队进步算法具备快速的收敛能力和良好的解集分布性,同时,算法稳定性也明显提高。选择直线阵为优化模型,将并行多目标算法应用于天线阵方向图的优化设计,对各阵元的激励幅值进行优化,以降低旁瓣电平和设计零陷位置。详细介绍了目标函数的建立过程,利用天线阵的对称结构和辐射特点,有效地缩减了可行域的搜索空间。在天线阵的应用算例中,算法能找到一系列优秀解,且在多个优化目标上表现良好,表明并行多目标团队进步算法具备解决电磁场优化等实际工程问题的能力。
[Abstract]:In real life, optimization problems are common and often involve conflicting goals. Because of the introduction of many design indexes, the search space of the problem is obviously enlarged, and the difficulty of solving the problem is greatly increased. Therefore, in order to solve the multi-objective optimization problem, which has a large amount of computation, It is a feasible solution to make full use of the optimization ability of intelligent optimization algorithm and large scale high performance parallel computing technology. This paper presents a parallel multi-objective team progress algorithm based on the adaptive idea and parallelization scheme based on the non-dominated ranking team progress algorithm (NRTPA),). The parallel multi-objective team progress algorithm combines the two-population evolution mechanism with the non-dominated sorting multi-objective strategy to improve the approximation uniformity and broadness of the multi-objective solution set with efficient optimization efficiency. According to the visual verification of test function set and the quantitative comparison of measurement index, the test results show that the parallel multi-objective team progress algorithm has fast convergence ability and good solution set distribution, and the stability of the algorithm is improved obviously. The parallel multi-objective algorithm is applied to the optimization design of antenna array pattern, and the excitation amplitude of each array element is optimized to reduce the sidelobe level and design zero trapping position. The establishment process of objective function is introduced in detail. The search space of feasible region is reduced effectively by using the symmetrical structure and radiation characteristics of antenna array. In the application example of antenna array, the algorithm can find a series of excellent solutions and perform well on multiple optimization objectives, which shows that the parallel multi-objective team progress algorithm has the ability to solve practical engineering problems such as electromagnetic field optimization.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18
本文编号：2231463
[Abstract]:In real life, optimization problems are common and often involve conflicting goals. Because of the introduction of many design indexes, the search space of the problem is obviously enlarged, and the difficulty of solving the problem is greatly increased. Therefore, in order to solve the multi-objective optimization problem, which has a large amount of computation, It is a feasible solution to make full use of the optimization ability of intelligent optimization algorithm and large scale high performance parallel computing technology. This paper presents a parallel multi-objective team progress algorithm based on the adaptive idea and parallelization scheme based on the non-dominated ranking team progress algorithm (NRTPA),). The parallel multi-objective team progress algorithm combines the two-population evolution mechanism with the non-dominated sorting multi-objective strategy to improve the approximation uniformity and broadness of the multi-objective solution set with efficient optimization efficiency. According to the visual verification of test function set and the quantitative comparison of measurement index, the test results show that the parallel multi-objective team progress algorithm has fast convergence ability and good solution set distribution, and the stability of the algorithm is improved obviously. The parallel multi-objective algorithm is applied to the optimization design of antenna array pattern, and the excitation amplitude of each array element is optimized to reduce the sidelobe level and design zero trapping position. The establishment process of objective function is introduced in detail. The search space of feasible region is reduced effectively by using the symmetrical structure and radiation characteristics of antenna array. In the application example of antenna array, the algorithm can find a series of excellent solutions and perform well on multiple optimization objectives, which shows that the parallel multi-objective team progress algorithm has the ability to solve practical engineering problems such as electromagnetic field optimization.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18
【参考文献】
相关期刊论文 前5条
1 王维博;冯全源;;粒子群优化算法在天线方向图综合中的应用[J];电子科技大学学报;2011年02期
2 公茂果;焦李成;杨咚咚;马文萍;;进化多目标优化算法研究[J];软件学报;2009年02期
3 薄亚明;;多维函数全局寻优的团队进步算法[J];南京邮电大学学报(自然科学版);2008年05期
4 于滨;杨忠振;程春田;;并行蚁群算法在公交线网优化中应用[J];大连理工大学学报;2007年02期
5 李昌隆,陈晓波,程鹏;并行多目标遗传算法在控制系统优化中的应用[J];微计算机信息;2005年12期
相关博士学位论文 前2条
1 王小根;粒子群优化算法的改进及其在图像中的应用研究[D];江南大学;2009年
2 莫鸿强;遗传算法搜索能力和编码方式研究[D];华南理工大学;2001年
相关硕士学位论文 前3条
1 赵博伟;多目标智能优化算法及其天线设计应用的研究[D];南京邮电大学;2016年
2 李东慧;分工搜索演化算法及其应用的研究[D];南京邮电大学;2011年
3 章春芳;自适应的并行蚁群算法及其应用[D];扬州大学;2006年
，本文编号：2231463
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2231463.html
