Research Of Hybrid PSO And Applications In Stochastic Programming

With the economic and social development, people recognize the importance of the decision-making problems and optimization problems. A correct decision, a good optimization can bring enormous practical benefits often. So it provides a powerful support and a broad prospect for the study of programming model and optimization algorithm.Particle Swarm Optimization algorithm is a new swarm intelligent algorithm based on bionics. Because it has lots of advantage such as its simple principle, easy to achieve and better robustness, PSO algorithm has been widely applied in neural networks training, optimization problems, data clustering and many other fields. Therefore, PSO algorithm has become the hotspot of the intelligent algorithm research.Through learning and researching the basic theory of PSO, for overcoming the disadvantage of basic PSO such as premature convergence, this paper prevents two ideas to improve: reusing invalid particle and importing fitness-velocity. By reusing invalid particle, adjusting swarm structure timely and changing the formula for calculating velocity, the improved algorithm can avoid premature convergence and get more accurate and quick convergence than basic one. After simulation experiment, we can prove that the improved algorithm is desirable and effective.Now, the theory and application of stochastic programming has become increasingly sophisticated. In the same time, stochastic programming expands the scope of its application progressively. It has played an impotent role in many areas and won a lot of attention. Based on the improvements of PSO algorithm, this paper will present hybrid Intelligence Algorithm witch contain the improved PSO and Support Vector Regression algorithm by analyzing the characteristics of various stochastic programming models. In the hybrid algorithm, SVR is mainly used to learn and approach to the functions that contain random variables, and then achieve the quantitative assessment for the objective function and constraints. After that, PSO is used to find the optimal solution by some particles exploring widely in the feasible space meeting the constraints. Finally, we can prove that the hybrid algorithm is feasible and effective in solving uncertain programming problems, and has better accuracy and convergence rate than traditional hybrid intelligent algorithm.