中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
基于生物协同进化模式的群体智能优化算法研究

文献类型:学位论文

作者孙丽玲
学位类别博士
答辩日期2015-11-22
授予单位中国科学院沈阳自动化研究所
授予地点中国科学院沈阳自动化研究所
导师胡静涛 ; 陈翰宁
关键词群体智能 协同进化 层次拓扑 正交表策略 K均值聚类 生命周期
其他题名Research of Swarm Intelligent optimization algorithm Based on Biological coevolution model
学位专业检测技术与自动化装置
中文摘要本文以生物协同进化为背景,在查阅和分析国内外相关领域研究成果的基础上,围绕生物协同进化中个体、群体、群落的行为特性,以及协同演化过程中信息交流模式展开相关研究,并基于生物种群间协同进化关系设计了三种算法:1)基于仅存在主种群的共生模型,设计了基于正交策略的多种群PSO算法(OMSPSO);2)基于仅存在从种群的共进模型,设计基于K均值聚类的多目标ABC算法(CMOABC)及应用;3)基于主从种群都存在的共栖模型,设计了基于生命周期理论的多算法融合方法(LCMEAs)。本文的具体研究工作概括如下:基于正交表策略的多种群协同进化粒子群算法(OMSPSO):针对原始PSO算法易陷入局部最优的缺陷,利用多种群协同进化模型,将单种群独立进化改造成多种群协同进化模式;并借助层次拓扑的多种形式,丰富层次协同进化模型的信息交流模式,增进学习效果;为克服PSO算法在优化过程中存在的震荡现象,采用正交表策略,可根据个体位置信息、群内位置信息、群间位置信息,计算出潜在优化方向,服务于个体位置更新阶段,有效地提高了搜索效率;用标准函数测试算法性能,通过与其他经典算法比较,数值实验结果证明OMSPSO具有较高的寻优效率和搜索精度;并基于此款优化算法对一款蝶形贴片天线的结构参数加以优化,通过对实际天线进行测试证实此算法优化出的天线可覆盖无线通信和RFID通信频带,进一步证实了此算法的实用性。基于K均值聚类的多目标蜂群算法(CMOABC):针对原始ABC算法收敛速度慢的情况,对雇佣蜂阶段个体更新公式进行了改进,使其以一定概率向最优个体聚拢,增加其局部搜索能力;为平衡算法的“探索”与“开发”之间关系,通过引入K均值聚类策略,实现根据个体空间分布分组,避免个体过早的陷入局部最优,以提升算法全局搜索能力;并通过周期性改变子群(分组)数目实现子群间的信息交流;通过测试标准多目标函数,并比对多个经典多目标算法,证实算法在收敛精度和多样性上具有明显优势;在此处上,将该算法应用于电力系统优化调度中,实现了微型电网的有效配置,可有效平衡发电费用、污染排放与有功网损之间的关系,进一步证明该算法的实用性。基于生命周期理论的多算法融合协同进化方法(LCMEAs):根据没有免费午餐原理,不可能有一种算法适用于解决所有问题。为突破这一技术瓶颈,在该算法的设计中引入多种算法协同优化机制,将种群进行划分,每一个算法独立优化一个子群,此模式可以提升算法的可扩展性;并通过借鉴生命周期理论,构建了算法性能评价体系,以营养值为核心,使占据优良解空间的个体分裂(复制),而占据非优空间的个体消亡(移除),进而突出优势算法的核心地位,实现资源的自适应集中,提高了资源利用率;并通过采用相互学习策略,有效保持处于劣势算法亦然具有搜索效用;通过测试大量标准、CEC2005、CEC2014等测试函数,证实了该方法的通用性;综上所述,本文从机理建模、算法设计和工程应用等多个层面,针对生物启发式优化算法展开研究,论文所提出了三种层次协同演化算法,在具体的仿真和应用中体现了其有效性,具有一定的理论价值和实际应用价值。
英文摘要With the background of biological coevolution, after checking and analyzing of the domestic and foreign research results, based on the biological co-evolutionary behavior characteristics of the individual, group and community, and collaborative communication pattern in the evolution process, this paper designs three kinds of algorithms according to the the population co-evolution: (1) Based on the symbiosis mode with only the master swarms, A Orthogonal Learning based Multi-species Particle Swarms Optimization (OMSPSO) is designed; (2) Based on the co-evolution model with only slave swarms, an Artificial Bee Colony Algorithm based on K-means Clustering (CMOABC) is proposed; (3) Base on the commensalism model with both of master and slave swarms, a Lifecycle Configuration of Multiple Evolution Algorithms (LCMEAs) is structured. In this paper, the specific work is summarized as follows: Orthogonal Learning based Multi-species Particle Swarms Optimization (OMSPSO): For overcoming the original PSO’s drawbacks of easily trapping into local optimum, learning from the multiple population co-evolution model, the evolution mode with single population is promoted as the evolution mode with population co-evolution. And in accordance with the various forms of hierarchical topology, the information communication mode in the hierarchical co-evolution model is riched and the learning effec is enhanced. To overcome the oscillation phenomenon existing in the PSO optimization process, the orthogonal experiment strategy is introduced and the potential optimaztion direction is calculated from the individual location information, location information within a group and between groups, which can serve the individual update phase for effectively improving the search efficiency. With the test against the standard functions, compared with other classical algorithms, the numerical results prove that OMSPSO has high precision and efficiency; therefore, based on the propsed algorithm, the structure parameters of a butterfly-shaped patch antenna is optimized. The measured results prove the antenna optimized byOMSPSO can satisfy the frequency band of wireless communication and RFID communication, which can further confirms the practicability of the algorithm. Artificial Bee Colony Algorithm based on K-means Clustering (CMOABC): To fasten the convergence rate of the canonical MOABC algorithm, the way of information communication in the employed bees’ phase is modified. For keeping the population diversity, the multi-swarm technology based on k-means clustering is employed to decompose the population into many clusters. Due to each subcomponent evolving separately, after every specific iterations, the population will be re-clustered to facilitate information exchange among different clusters. Application of the new CMOABC on several multi-objective benchmark functions shows a marked improvement in performance over the the canonical multi-object algorithms. Finally, the CMOABC is applied to solve the real-world Optimal Power Flow (OPF) problem that considers the cost, loss, and emission impacts as the objective functions. The simulation results demonstrate that the proposed CMOABC is superior for solving OPF problem, in terms of optimization accuracy. All these studies revolve the practicability of CMOABC. Lifecycle Configuration of Multiple Evolution Algorithms (LCMEAs): according to “No Free Lunch” theory, there is no singal algorithm suitable for solving all problems. To break through the technical bottleneck, this configuration brings in collaborative evolution mechanism. The whole population is divided into some sub-populations and each involved algorithm independently optimizes a sub-population, which can improve the scalability of the configuration. For the reference of life cycle theory, a performance evaluation system is built. As the core of the configuration, nutritional value decides that the individuals located at good solution room will splite (copy) and the individuals located at poor solution room will die (remove). The core role of advantage algorithm is adaptively outstanding. By adopting a mutual learning strategies, and the disadvantage algorithms remain to search effectively; By testing a large number of standard, CEC2005, CEC2014 test function, the generality of the method is confirmed. In summary, this paper research on the mechanism modeling, algorithm design and engineering application. This paper proposes three hierachical co-evolution optimization algorithms, which achieves innovative and application value result. The proposed improved strategy and optimized methods have a certain significance meanings to expand research and to guide the practical application in related fields.
语种中文
公开日期2015-12-27
产权排序1
页码121页
源URL[http://ir.sia.ac.cn/handle/173321/17526]  
专题沈阳自动化研究所_信息服务与智能控制技术研究室
推荐引用方式
GB/T 7714
孙丽玲. 基于生物协同进化模式的群体智能优化算法研究[D]. 中国科学院沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2015.

入库方式: OAI收割

来源:沈阳自动化研究所

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。