基于TS模糊推理的粒子群算法

发布时间：2018-10-25 06:53

【摘要】：粒子群优化算法(Particle Swarm Optimization PSO)是一种新兴的群体智能优化算法,具有分布式、协同合作性、自组织性和实现简单等特点,这使得该算法能够在全局信息缺乏时能够迅速地处理各种复杂问题,也为典型的复杂性问题的求解开辟了新的途径,但该算法在处理高维复杂问题时仍有相当大的可能陷入局部最优,如何通过保障Exploration和Exploitation之间的均衡来加强全局搜索能力,是该领域的研究热点和难点。从两个方面对PSO算法进行了改进,其一是基于孙俊等人的量子行为粒子群优化算法(Quantum-behaved Particle Swarm Optimization QPSO),提出了基于Takagi-Sugeno(TS)模糊推理的自适应量子行为粒子群优化算法(Adaptive Quantum-behaved Particle Swarm Optimization AQPSO),在惯性权重和种群多样性上对粒子群优化算法进行了改进。该算法利用群体分布和探索进程信息,由TS模糊推理动态地调整算法参数及其迭代方式,从而保证种群在更大的空间探索,减少陷入局部最优的概率。其二是基于Riget等人提出的attractive and repulsive PSO(ARPSO)算法,提出了动态地调整惯性权重的算法(Dynamic attractive and repulsive PSO DARPSO),该算法不是简单地用线性递减策略,而是根据粒子是收缩状态还是扩张状态而动态地调整惯性权重,同时根据TS模糊推理设计了一种新的粒子位置更新方式。若干标准测试函数仿真和威氏(Wilcoxon)符号秩检验的结果显示,AQPSO算法在处理多个局部最优解相差较小时效果较好,而DARPSO算法在处理全局最优解与局部最优解相差较大的问题时效果较好。同时,在处理复杂高维函数的优化问题上,本文提出的AQPSO算法、DARPSO算法,与QPSO算法、ARPSO算法以及PSO算法相比具有更好性能。
[Abstract]:Particle swarm optimization (Particle Swarm Optimization PSO) is a new swarm intelligence optimization algorithm, which has the characteristics of distributed, cooperative, self-organizing and simple implementation. This makes it possible for the algorithm to deal with all kinds of complex problems quickly when the global information is lacking, and also opens up a new way for solving typical complex problems. However, the algorithm is still likely to fall into local optimum when dealing with high dimensional complex problems. How to enhance the global search ability by ensuring the balance between Exploration and Exploitation is a hot and difficult point in this field. The PSO algorithm is improved from two aspects. One is the quantum behavior particle swarm optimization algorithm based on Sun Jun et al. (Quantum-behaved Particle Swarm Optimization QPSO),) an adaptive quantum behavior particle swarm optimization algorithm based on Takagi-Sugeno (TS) fuzzy reasoning (Adaptive Quantum-behaved Particle Swarm Optimization AQPSO),) is proposed. Particle swarm optimization algorithm is improved. Using the information of population distribution and exploration process, the algorithm dynamically adjusts the parameters of the algorithm and its iterative method by TS fuzzy reasoning, so as to ensure the population exploration in a larger space and reduce the probability of falling into local optimum. Secondly, based on the attractive and repulsive PSO (ARPSO) algorithm proposed by Riget et al., this paper proposes a dynamic algorithm to adjust the inertia weight, (Dynamic attractive and repulsive PSO DARPSO), which is not a simple linear decrement strategy. Instead, the inertia weight is adjusted dynamically according to whether the particle is contracted or expanded, and a new updating method of particle position is designed according to TS fuzzy reasoning. The simulation results of several standard test functions and the (Wilcoxon) sign rank test show that the AQPSO algorithm is effective in dealing with multiple local optimal solutions with small differences. The DARPSO algorithm is effective in solving the problem where the global optimal solution is different from the local optimal solution. At the same time, the AQPSO algorithm, DARPSO algorithm proposed in this paper have better performance than QPSO algorithm, ARPSO algorithm and PSO algorithm in dealing with the optimization problem of complex high-dimensional function.
【学位授予单位】：青岛大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】