The Modified PSO Algorithm And Its Application In Multi-objective Optimization Problem

In modern scientific research, multi-objective optimization is one of the most important research areas in optimization problem and most problems in real-world are characterized by multi-objective, so it is usually not easy to handle them. Therefore, solving multi-objective optimization problem is a topic which has practical significance and scientific research value. Formerly, there were many methods which belong to operational research and decision making to solve multi-objective optimization problem. With the development of modern science and technology and penetration of different subjects, and the appearance of new cross subjects, new thinking ways and new computation methods energize the study and development of multi-objective optimization technology and offer more space for study.With the development of the computational intelligence technology, the evolution algorithm was applied to solving multi-objective optimization problem in the middle 1980s.And many multi-objective evolution algorithms have emerged in large numbers, such as SPEA,PESA,NPGA and so on. Making use of evolution algorithm to solve multi-objective problems have gradually become a hot issue and the most important research area. It breaks the limitation of multi-objective optimization in the operational research and possesses the characteristics which distinguish itself from the traditional single objective evolution algorithm and has great practical value in industrial engineering, science and national defense military.Particle Swarm Optimization was an optimization algorithm which was proposed on the basis of swarm intelligence in the 1990s and it directs the optimization search by the swarm intelligence which was produced by the cooperation and the competition of the swarm particles. And this new algorithm was inspired by the swarm food-hunting behavior of birds, insects and fish groups. Because of its simplicity and efficiency, it immediately drew extensive attention and at the same time its advantages which was showed when solving the single objective problem was quite applicable for solving multi-objective problems.The thesis focuses on the study of particle swarm optimization algorithm in the swarm intelligence and further applies it to the multi-objective field. The main works are as follows:1. The brief introduction of theory basis of particle swarm optimization and multi-objective optimization.The brief introduction of the concepts, the basic procedure and the development of particle swarm optimization; the basic theory of multi-objective optimization and common classical algorithms, and the exploration of the key theories of multi-objective evolution algorithms and presentation of some common multi-objective evolution algorithms.2. Propose a cauchy particle swarm optimization based on dynamic probability mutation.In view of the disadvantages of particle swarm optimization, through the 0/1 knapsack problem, the performance of Fuzzy PSO was verified. Using the basic thoughts of Gaussian Swarm and Fuzzy PSO, cauchy particle swarm optimization was proposed and further the mutation operation of genetic algorithm was introduced and formed dynamic probability mutation cauchy particle swarm optimization algorithm.The experiments proved that improved algorithm was better than the traditional algorithm.3. The proposition of a model of multi-PSO co-evolution.Drawing lessons from the co-evolution thoughts of swarm intelligence and the basic theory of genetic algorithm, an algorithm of multi-PSO co-evolution was proposed and verified by experiments. This algorithm was improved by combining the common characteristics of swarm evolution algorithms and proposed a model of multi-PSO co-evolution and the main structure of the model was presented.4. The improvement and proving of the constructing method of Pareto, the realization of improving PSO in solving multi-object knapsack problem.At the end of this thesis, the constructing method of Pareto was improved on the basis of Arena's Principle and the salvation of multi-knapsack problem of multi-objective optimization was realized through the improved PSO. The model program was developed and completed on the platform of Windows XP with VC++.NET 2005. Three examination samples were testified and satisfactory results was produced.