| 英文摘要 | Throughout history, people have to confront the optimization problems, which exist widely in both the natural science and the social science. Myriad optimization algorithms have been proposed to solve the problems, and the swarm intelligence optimization algorithms are the latest, which were born in the 80s last century. They became famous accompanying the continuous development and successful application of the ant colony algorithm and the particle swarm optimization. Many original swarm intelligence optimization algorithms, such as bacteria foraging optimization algorithm and artificial fish swarm algorithm, have been proposed from the beginning of this century. At the same time, classical PSO is paid much attention and plays an important role in the optimization problems. Although swarm intelligence optimization algorithms have made marvelous achievement, their performance is still far away from the increasing demands. The feature of emergence is a great advantage of swarm intelligence optimization algorithms, therefore we should strengthen the learning ability of the particles and enhance the communication in the swarm. This thesis aims at these two parts, proposing three improved measures on two basic algorithms. PID controllers are the most widespread controllers in the modern industrial control by virtue of their several priorities such as the simplicity of structure, the convenience of debug, the robustness and the ease of hardware implementation. Because the performance of the PID controllers depends completely on the parameters, the tuning of parameters has been the core problem since PID controllers were proposed. Recent years, with the technology progressing rapidly, the traditional tuning methods expose much weakness, followed by many tuning methods based on optimization algorithms. The method based on swarm intelligence optimization algorithms is the up-to-date tuning method, and it has relatively broad prospects. The major contributions of the thesis are: Firstly, the strategy of double centers is put forward, in which the swarm is not searching toward the same point any more, and is applied to the fruit fly optimization algorithm. Secondly, the searching type of battle and the operator of cross on dimensions are presented. With the battle, the searching target of a particle is not only depends on the fitness, but also the distance, so that the swarm is not prone to fall into the local minima too early. Furthe... |
修改评论