先验驱动的多目标人工雨滴算法及其应用研究

发布时间：2018-12-14 07:39

【摘要】：演化算法是一类启发于自然现象或规律的智能搜索和优化技术的总称。由于其高效的优化性能和巨大的应用潜力,演化算法在过去的半个多世纪受到了国内外研究人员的广泛关注。有鉴于此,本文旨在近来提出的人工雨滴算法的基础上,对其在复杂连续优化问题的求解方面展开研究。主要工作如下:(1)为进一步理解人工雨滴算法的运行机理和计算效果,首先,利用相关的数学理论,证明人工雨滴算法在变量不相关的条件下是以概率1收敛到满意种群;其次,与三个演化算法在CEC2005测试平台上进行优化性能比较。实验结果证实了人工雨滴算法在解决复杂连续优化问题的有效性。(2)在利用人工雨滴算法求解多目标优化问题时,如何在算法设计过程中融合问题的特征是提高计算效率的重要方面。为此,提出一种先验驱动多目标人工雨滴算法。首先,通过结合非支配排序框架和人工雨滴算法搜索引擎,提出一种多目标人工雨滴算法;其次,为加快多目标人工雨滴算法的收敛速度,通过集成多目标优化的先验知识-搜索空间的中心点和二项交叉算子,来引导种群快速向理想Pareto前沿靠近;最后,为保持种群选择的有效性和非支配解的多样性,提出一种基于最近拥挤距离的非支配解修剪方法。为验证算法的优化性能,选取了 12个多目标测试函数进行验证,并与其它四个多目标优化算法进行对比。结果表明提出的算法比其它的优化算法能够更快地跳出Pareto局部最优解,并获得了更好的Pareto前沿的均匀性。(3)针对电力系统中的无功优化问题,首先,建立以电压偏差和有功网损为目标的多目标优化模型。其次,利用提出的多目标人工雨滴算法进行求解,详细描述了算法的编码和流程。最后,在IEEE-30节点系统进行测试,将优化前后的结果进行了对比,并将结果与刘佳的文献中的优化结果进行对比,实验结果表明所用算法实现了电力系统经济运行的同时,提高了电网的电压稳定性。
[Abstract]:Evolutionary algorithms are a class of intelligent search and optimization techniques inspired by natural phenomena or laws. Because of its high performance and great application potential, evolutionary algorithms have attracted much attention from researchers at home and abroad in the past half century. In view of this, this paper aims to study the solution of complex continuous optimization problems based on the recently proposed artificial raindrop algorithm. The main work is as follows: (1) in order to further understand the operation mechanism and calculation effect of artificial raindrop algorithm, it is proved that the artificial raindrop algorithm converges to a satisfactory population with probability 1 under the condition that the variables are not related to each other. Secondly, the optimization performance is compared with the three evolutionary algorithms on the CEC2005 test platform. Experimental results show that artificial raindrop algorithm is effective in solving complex continuous optimization problems. (2) when artificial raindrop algorithm is used to solve multi-objective optimization problem, How to fuse the characteristics of the problem in the process of algorithm design is an important aspect to improve the computational efficiency. Therefore, a priori driven multi-objective artificial raindrop algorithm is proposed. Firstly, a multi-objective artificial raindrop algorithm is proposed by combining the non-dominated sorting framework with the artificial raindrop algorithm. Secondly, in order to speed up the convergence of the multi-objective artificial raindrop algorithm, by integrating the priori knowledge of multi-objective optimization, the center of search space and the binomial crossover operator, the population is guided to the ideal Pareto front quickly. Finally, in order to maintain the effectiveness of population selection and the diversity of non-dominated solutions, a pruning method based on the nearest crowding distance is proposed. In order to verify the optimization performance of the algorithm, 12 multi-objective test functions are selected for verification, and compared with the other four multi-objective optimization algorithms. The results show that the proposed algorithm can jump out of the Pareto local optimal solution faster than other optimization algorithms, and obtain better uniformity of the Pareto frontier. (3) for reactive power optimization problems in power systems, firstly, A multi-objective optimization model with voltage deviation and active power loss as targets is established. Secondly, the multi-objective artificial raindrop algorithm is used to solve the problem, and the coding and flow of the algorithm are described in detail. Finally, the IEEE-30 node system is tested, the results before and after optimization are compared, and the results are compared with the optimized results in Liu Jia's literature. The experimental results show that the proposed algorithm realizes the economic operation of power system at the same time. The voltage stability of the power network is improved.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】