Research On Improved Pso And Its Application On Container Terminal Scheduling

Intelligent algorithm is a kind of calculation method which has the characteristics of self-adaptation, self-replication and the capability of evolution. And it is inspired from the natural phenomenon or the principles and mechanism of the biology itself or groups in Nature. It has the characteristics of parallel, simple and robustness, many scholars have researched and analyzed it and have also put forward various improved algorithm."No Free Lunch" theorem points out that there is no algorithm with both computing performance and time complexity algorithm, but many scholars’ researches show that there is still much room to improve the algorithm. Therefore, how to improve the performance of algorithm and enlarge the field of the application is an important content of modern optimizing algorithm. The major contributions and research achievements of this thesis include:(1) A literature review for Particle swarm optimization algorithm research is made. Firstly, the basic mechanism of the algorithm, implementation and processes of solving problem is introduced, and then a variety of algorithms to improve the classification methods and strategies are introduced, including the form of improvement, control of the particles, the topology and discrete versions.(2) In order to improve the diversity of the swarm and ability to communication of particles, an improved particle swarm optimization with dynamic topology structure is proposed by using k-means clustering method and Ring structure(KPSO). For communication two kinds of patterns of updating position and speed of every particle are given. Through the tests of Benchmark Functions, the performance of KSPO is compared with PSO. Then an experiment is carried out for the choice of parameters of KPSO:clustering number and population size.(3)Based on KPSO, a new algorithm (adaptive PSO based on clustering method, called AKPSO-C) is proposed that every particle adjusts its parameters automatically according to the state in the process of searching. Then the convergence of the algorithm is analyzed theoretically. Through the tests of Benchmark Functions, contrasts and comparisons are made in many aspects(including distribution of swarm, searching ability, convergence performance and sensibility analysis of initial values) by using different algorithms (including APSO-C, classic PSO, KPSO)(4)The best location of ship berthing and the shortest time in container terminal are researched in this paper. Under the situation of continuous berth, considering the berth allocation and quay crane scheduling problem at the same time, a new berth and crane allocation strategy when ship is berthing and crane allocation strategy for ships in container terminal are proposed. At last above improved algorithm is used to solve the problem.