Research On Particle Swarm Optimization Algorithm And Its Application

With the increasing requirements of scientific research and application, practical optimization problems become more and more complex. Traditional optimization methods often have some limitations when solving these problems. Now, along with the development of computer technology, it has become a new hotspot by using intelligent optimization methods for these complex problems.As a kind of heuristic optimization algorithm, particle swarm optimization (PSO) algorithm outperforms traditional optimization methods in terms of structural complexity, control parameters, implementation cost and the ability to find the best solution. However, since the theoretical basis of PSO is still far from mature, the problems with the premature convergence remain for exploiting, leading further to more spaces to improve while been applied to practical engineering. Based on studying PSO theory, this work proposes a strategy for choosing the parameters to enhance its dynamic characteristic. Several improved PSO algorithms have been presented to improve the global searching ability and the convergence rate. Then, some modified PSO algorithms are applied to the construction project optimization and antenna array synthesis. The main work can be summarized as follows:In PSO study, there are a few control parameters affecting its performance. Generally, these parameters are derived from experience. This leads to the difficulty in obtaining the optimal combination of the parameters and affects the use of PSO. In view of this, the effects of the control parameters are systematically investigated by employing benchmark functions. A guideline of choosing these parameters is proposed to improve the performance of PSO.Based on the study of PSO theory and information sharing mechanism, inspired by some effective ideas of human sociology and combined with the advantages of the chaotic optimization algorithm, several improved PSO algorithms are proposed, which include:(1) By simulating the act of human randomized searching behaviors, a novel stochastic focusing search PSO (SFSPSO) is proposed. With dynamic neighborhood topology and subpopulation strategy, SFSPSO can improve the global searching ability while keeping the diversity and avoiding the astringency of a local extremum. The simulation results show that SFSPSO is competitive to solve various benchmark problems.(2) A hierarchical subpopulation PSO (HSPSO) is presented to improve the convergence speed and accuracy by using the strategy of subpopulation hierarchy. The information is gradually flowed from the upper level to the lower level of the hierarchy, and the communication in the particle swarm is moderate and suitable for a proper balance between the global and local searching ability. (3) A hierarchical chaos PSO (HCPSO) is proposed, which is based on the structure of the hierarchical multi-subpopulations. The novel algorithm adopts chaotic mutation in the nonlinear and decreasing inertia weight. The new global best position is the average position of several individuals that are picked out as exemplars when the new global best positions are updated in each dimension. The radius of the chaotic searching region can be adaptively adjusted. The simulation results show that HCPSO is more effective to overcome the slow convergence and prematurity.The HSPSO has been applied in time-cost-quality synthesis optimization of the construction project. Satisfied results can be quickly obtained by using HSPSO with a smaller swarm. The exhaustive enumeration is given to verify the effectiveness of HSPSO.Novel PSO algorithms are also applied to the antenna array synthesis, which include:(1) The multiple subpopulation PSO (MSPSO) is presented in the pattern synthesis of equally spaced linear arrays. MSPSO is built by employing the strategy of hierarchy and subpopulation with the neighbor structure of von Neumann. A modified objective function model, which utilizes different fitness functions according to the character of the top and bottom layer, is proposed to balance the local and global searching ability. The simulation results show that it achieves relatively high performance and can avoid the premature and easily trapping in local optima.(2) A chaotic search PSO (CSPSO) is proposed by fusing the advantages of both chaotic optimization algorithm and PSO. The novel algorithm utilizes chaotic searching strategy when the swarm is trapped into stagnancy. To enhance the diversity of samples, several individuals are picked out as exemplars when the new global best position is updated in each dimension. The simulation results show that it achieves relatively high performance by applying CSPSO in the pattern synthesis of antenna arrays with sidelobe reduction and null control.(3) To deal with the pattern synthesis of the equally spaced linear array, unequally spaced linear array and conformal array, a chaotic PSO (OPSO) is presented. Experimental results present its high performance in the pattern synthesis with low sidelobe level, multi-nulls and shaped beam.(4) The chaotic binary PSO (CBPSO) is presented as a useful alternative for the synthesis of thinned arrays. CBPSO is improved by nonlinear inertia weight with chaotic mutation to increase the diversity of particles. An extensive numerical analysis has been performed by addressing thinned linear and planar arrays with sidelobe suppression. Simulation results are proposed to compare with published results to verify the effectiveness of the proposed method.