混合粒子群优化算法研究及其应用

发布时间：2021-12-30 01:23

　　粒子群优化算法（PSO）来源于鸟类的群体行为,是一种基于种群的随机搜索算法。粒子群优化算法的性能在很大程度上取决于参数的选择。其中,惯性权重是该算法的参数之一,用于平衡粒子群优化算法的探测和开发能力。为了提升算法的性能,PSO应较好地兼顾全局搜索能力与局部搜索能力。PSO算法具有收敛速度快、易于实现、参数调整少等优点,其有效的搜索策略使之具有潜力在各种应用中解决不同优化问题的方法。PSO在电力系统优化、过程控制、动态优化、自适应控制和电磁优化等领域有着广泛的应用。虽然PSO在求解许多优化问题时表现出良好的性能,但与大多数随机搜索方法一样,它也存在早熟收敛问题,尤其是在多模态优化问题中。虽然取得了显著的进展和丰硕的成果,但如何较好地平衡粒子群优化算法的探测和开发能力以确定复杂优化问题的高质量解决方案,仍然具有一定的挑战性。本文将局部搜索与粒子群算法相结合,以平衡算法的探测与开发能力,并将改进后的方法应用于许多优化问题和实际问题。本文的主要工作如下:（1）为了解决这一问题,本文提出了一种基于梯度局部搜索策略的自适应惯性权重粒子群优化算法（SIW-APSO-LS）。该算法旨在平衡全局搜索算法... 

【文章来源】：江苏大学江苏省

【文章页数】：116 页

【学位级别】：博士

【文章目录】：
DEDICATION
Abstract
中文摘要
Chapter1 Introduction
    1.1 Background of particle swarm optimization
        1.1.1 Research significance of PSO
        1.1.2 Review and analysis of hybrid PSO
    1.2 Analysis and significance of study
    1.3 Main work and novelty
    1.4 The organization of thesis
Chapter2 Preliminaries
    2.1 Inertia weight in PSO
        2.1.1 Constant inertia weight
        2.1.2 Time-varying inertia weight strategies
        2.1.3 Adaptive inertia weight
    2.2 Dynamic multi swarm particle swarm optimization
    2.3 Gravitation search algorithm
    2.4 Conclusions
Chapter3 An improved self-hybrid inertia weight adaptive particle swarm optimization with local search
    3.1 Introduction
        3.1.1 Gradient based local search(BFGS)algorithm
    3.2 The proposed hybrid algorithm
        3.2.1 The self-hybrid inertia weight adaptive particle swarm optimization
        3.2.2 Framework of SIW-APSO-LS
    3.3 Experimental results and discussion
        3.3.1 Static test functions
        3.3.2 Moving peaks benchmark
        3.3.3 CEC13 test suit
        3.3.4 Comparison with inertia weight adjusting techniques
        3.3.5 Comparison of different methods through moving peaks benchmark(MPB)
        3.3.6 CEC13 test functions
        3.3.7 Comparison of the proposed algorithm(SIW-APSO-LS)with other PSO variants
        3.3.8 Computational cost
        3.3.9 Real world optimization problem
    3.4 Conclusions
Chapter4 Feature selection based SIW-APSO-LS and C4.5 decision tree classifier
    4.1 Introduction
        4.1.1 Preliminary feature selection methodology
        4.1.2 C4.5 decision tree classifier
    4.2 The proposed feature selection method
    4.3 Experimental results and discussions
        4.3.1 Datasets
        4.3.2 Parameters setting
        4.3.3 Evaluation criteria
        4.3.4 Comparison of proposed algorithm to other algorithms
    4.4 Conclusions
Chapter5 Hybrid gravitational search algorithm with dynamic multi swarm particle swarm optimization
    5.1 Introduction
    5.2 The proposed hybrid PSO based on GSA
    5.3 Experimental study and discussion
        5.3.1 Static test functions
        5.3.2 CEC13 test suit
        5.3.3 Moving peaks benchmark
        5.3.4 Comparison with other techniques
        5.3.5 Computational cost of different algorithms
        5.3.6 Comparison through CEC13 test functions
        5.3.7 Comparison of different methods through MPB
        5.3.8 Tension and compression spring design
    5.4 Conclusions
Chapter6 Training a feedforward neural networks using GSADMSPSO
    6.1 Introduction
        6.1.1 Feed-forward neural network and multi-layer perceptron
    6.2 The proposed GSADMSPSO algorithm
        6.2.1 GSADMSPSO method for training FNNs
    6.3 Results and discussions
        6.3.1 The N bits parity(XOR)problem
        6.3.2 Comparison through three bits parity problem(3-bit XOR)
        6.3.3 Comparison with other techniques through standard classification datasets
    6.4 Conclusions
Chapter7 Conclusions and future Work
    7.1 Conclusions
    7.2 Future work
References
Acknowledgements
Publications and participating research fundings



本文编号：3557200