Particle Swarm Optimization Algorithm And Applied Research  Posted on:20091128  Degree:Doctor  Type:Dissertation  Country:China  Candidate:H Mao  Full Text:PDF  GTID:1118360272457630  Subject:Mechanical Manufacturing and Automation  Abstract/Summary:  PDF Full Text Request  Kennedy and Eberhart established the Particle Swarm Optimization algorithm(PSO) in 1995. This algorithm simulates the action of the bird swarm looking for food by flying and get the optimization through the cooperation in the bird swarm. PSO obtain the evolution of all the swarm depending on changing the information between the singles. These little particles which are called swarm are nonvolume and nonquality. They can adjust the moving track of themselves and fly towards the best location which itself ever went previously and which the whole swarm ever went previously. In order to attain the aim, all the particles of the swarm have the ability of memory and they can adjust their locations between the current locations and the best locations they ever found, for example, in a minimum problem the socalled better location is a point corresponding to the lesser value of the aimfunction in the explain space. What is the superiority of PSO is that the algorithm can be realized easily and has the profound intelligence background. This algorithm is not only fit for scientific research but also fit for the engineering application. In very short time the PSO made the great progress and has been used in some fields.But the PSO also has its shortage. This algorithm is easy in premature convergence or become stagnant during the optimizing process. This is because every particle's speed can not be refreshed in the anaphase of the optimization, which cause the particles to collect together very tightly in some locations and can not make the local search more wide and more meticulous. To the variety of the swarm, in this time the swarm is went short of variety and the difference between each particles is very little, which can not urge the swarm to develop. The problem of premature convergence is not easy found in optimizing the single mode functions but once optimize the mulmode functions, which are nonlinear and have extensive searching space, plentiful local extremum and large bar, the algorithm will be easy in the premature convergence or become stagnant.The angle that this text is from the engineering applied physically set outs, with the simple principle is guide thought, cast aside to sophisticate to reason logically with tedious theories with compute, and apply the optimize of some classics the calculate way to discuss and study the design of the PSO in order to offering some can design method that draw lessons from. To carry out this task, this paper is presented the new ideas as follows:â˜†The acceleration c_{1} ,c_{2} are the very important coefficient to the PSO. They can improve the optimization quality of the PSO greatly. In the paper we do much experiment and analyse to elementarily study how the partnership about c_{1} ,c_{2} can affect the optimization quality of the PSO from three direction including the linear partnership relation of the c_{1} ,c_{2},the nonlinear partnership relation of the c_{1} ,c_{2} and the partnership relation with the inertia weightÏ‰.â˜†Applying the cloud model to realize the adjusting to the inertia weight by mulrule and uncertainty. This method has more quick convergence speed and more better optimization result using some testfunctions. On the other hand, the acceleration c 2is also a very important parameter of the PSO, so also to refer a method which adjust both the inertia weight and the acceleration c 2 using the mulrule and uncertainty. This method is proved more better than the method only adjusting the inertia weight.â˜†two subâ€”swarms substituting particle swarm optimization algorithm is proposed. This method is simple to use. It use two subswarms.One subswarm use the Globe PSO to search and find optimization, and the other use the Local PSO. Through the particles being replaced between the two swarm the variety of the swarm can be confirmed and also confirm the quick convergence speed. Through the examination this method can improve the efficiency of the optimization. Then use this improved PSO to optimize the parameters of the incomplete differential PID which is used in the industrial controlling universally. Through the simulation of the controlling to several objects, the results are excellent, at the same time also showing the excellent ability of the selfadaptive control when the object is changed or some disturbs are imported inside the system suddenly.â˜†a cooperative PSO which use the differential evolution algorithm(DE) and the standard PSO. The new algorithm bring the superiority of the DE to the PSO and confirm the variety of the swarm by applying the intersection and the variance to the particle. Use this improved PSO algorithm to optimize the NN and then use the NN to the system which detect and alarm to the fire in forepart. The system work very well.â˜†Referring to use the DE to variety the historical best location of each particle in order to confirm that these location would not change nothing or change very little in a long time, in the other hand, also variety the best location of all the swarm, so the particle's speed can be changed usually so that the search ability can be preserved and the ability of the avoiding the premature convergence will be improved, and has the more probability to attain the best value.
 Keywords/Search Tags:  the Particle Swarm Optimization algorithm(PSO), Optimal Design, Cloud Model, the differential evolution algorithm(DE), the Golden Section, Principle Component Analysis, NN, the Incomplete Differential PID  PDF Full Text Request  Related items 
 
