The Research Of Particle Swarm Optimization Algorithm Improvement

In many practical fields, such as economics, management, military,scientific and engineering fields, they are related to optimization problems. And practically the problems in those fields are very complicated, multipolarization, nonlinear, strong constraints, and hard to model. Therefore, the requirements about optimization are increasing and urgent.Particle swarm optimization algorithm originate from the research on a flock of birds behavior, and then develops an intelligent optimization algorithm, this method is especially suitable to the handing of traditional search methods which an not solve the complex and non-linear problems. And genetic algorithm is similar to the particle swarm algorithm which is a population-based optimization tool, the system is initialized as a set of random solutions, through the iterative search for the optimal value. But the particle swarm optimization which has no crossover and mutation as genetic algorithm, the particle to follow the optimal in the building to search the solution. Therefore, when the particle swarm algorithm has been found, scholars from various countries give their attention, and form a research hotspot.In this paper, consulted a lot of literature about particle swarm optimization algorithm, and did research and analysis on the standard particle swarm optimization. Through the research on the predecessors, the paper found the flaw of particle swarm optimization algorithm.The flaw is premature and easy to convergence, for that the paper give the solution to the two issues. Through this research, the paper give two modified particle swarm algorithms.First, the thesis introduces the research on the topological model of particle swarm. And it proposes the core group of particle swarm optimization algorithm. and two-tier local model algorithm, which is applied to the core group of particle swarm optimization.It increases the particle variability.Second, the thesis improves the speed of updating formula of the particle swarm algorithm, including the local best partice as the reference learning mechanism, making the learning factor of particles from 2 to 3. And it mixs the core of group particle swarm algorithm with improved formula algorithm. Finally, the thesis showed the programming using multi-threading JAVA. To a certain extent, the improvement of the time is efficiency. To verify the effectiveness of the algorithm, the improved algorithm is applied to solve nonlinear equations. The result demonstrates that the improvement is very effective.