Study On The Swarm Intelligent Algorithm And Its Application  Posted on:20110227  Degree:Doctor  Type:Dissertation  Country:China  Candidate:J Zhao  Full Text:PDF  GTID:1118360302987727  Subject:Light Industry Information Technology and Engineering  Abstract/Summary:  PDF Full Text Request  Swarm Intelligent (SI) algorithm is an algorithmic approach, which has gradually attracted more attention. To achieve the purpose of optimizing, SI simulated social behavior of various groups of animals and the individuals in the groups exchange information and cooperate each other. Compared with other optimization algorithms SI is easier to performe and more efficient.Particle swarm optimization (PSO) is an evolutionary computation technique developed by Dr. Kennedy and Dr. Eberhart in1995, inspired by social behavior of bird flocking or fish schooling. PSO is simple in concept, few in parameters, and easy in implementation. It was proved to be an efficient method to solve optimization problems. Based on the deep study of PSO algorithm and inspired by quantum physics, Quantumbehaved Particle Swarm Optimization (QPSO) algorithm is proposed. QPSO algorithm has much less parameters and much stronger global search ability than the PSO algorithm.Theoretical analyses and algorithm improving on PSO algorithm and QPSO algorithm are mainly discussed in our work and the application of QPSO algorithm are also studied in this work. The main contents of this dissertation are as follows:1. The concept of optimization problem and its solution are introduced to explain the research background of the swarm intelligence algorithm and several common intelligent optimization algorithms are described in detail. The basis of our study is illustrated by the no free lunch theorem. Against the defects of PSO algorithm, the research objectives, research content, research ideas and methods in the work are proposed.2. After the principle and procedure of PSO algorithm is presented, two important versions, PSO with inertia weight and PSO with contraction coefficient, are discussed. Some improved PSO methods are also mentioned for reference. The thought of QPSO algorithm is discussed. Convergence criteria of random search algorithms are introduced, including global convergence criteria and local convergence criteria. Based on these two convergence criteria, QPSO algorithm is proven to be a global search stochastic algorithm. By comparing QPSO algorithm and PSO algorithm, the characteristics of QPSO are indicated. Finally try to introduce a new mutation mechanism in the QPSO algorithm and the quantumbehaved particle swarm optimization based on cloud model mutation (QPSONCM) is proposed to increase the diversity of the population and improve the ability of the algorithm to fall into local optimization so as to enhance the global search capability.3. Premature convergence is also appeared in QPSO algorithm when solving multimodal problems. The reason for premature convergence lies in the collections of swarm which makes the swarm diversity decline and the particles lose the ability of searching in a wide space. An improved Quantumbehaved Particle Swarm Optimization using the notion of species for solving multipeaks functions optimization problems is proposed. In the proposed Speciesbased QPSO (SQPSO), the swarm population is divided into paralleled species subpopulations based on their similarity and each peaks are ensure to be searched equally, regardless if they are global or local optima. Our experiments for static and dynamic multipeaks environments demonstrate that global search ability and local search capabilities of the improved algorithm have been greatly enhanced.4. In order to overcome the difficulty that the leastsquares method cannot deal with timedelayline identification, QPSO algorithm combined with the single neuron is proposed to improve the local search capabilities and identification accuracy. Then the improved QPSO is applied to online identify parameters of a system described by differential equations. The improve QPSO algorithm has faster convergence speed and smaller length of the identification window, so it is more suitable for realtime online identification in practice. Timedelay and parameter changes for the simulation experiment illustrates the stability and tracking capability of improved QPSO algorithm are better. By introducing a sessionbased admission timeratio feedback control mechanism an adaptive control of Web QoS based on system model online identification using QPSO algorithm is designed and implemented which dynamically adjust parameters of proportionalintegral (PI) controller according to the changes of system model.5. QPSO algorithm is used to identify parameters of chaotic systems, periodic systems and stability systems. The simulation results of QPSO compared with PSO and GA demonstrate that in the system parameter identification QPSO algorithm has best performance. For the existence of noise in chaotic systems, the online parameter identification based on QPSO is proposed and the effectiveness of the method is proved.6. The study of QPSO algorithm in fault diagnosis research. Intelligent fault diagnosis technology is a combination of artificial intelligence and fault diagnosis, which use a computer to simulate human expert through artificial methods so as to diagnosis complex systems. A single radial basis function neural network (RBF NN) is a good performance feed forward network which not only has biological context, but also match with the function approximation theory and is suitable for multivariable function approximation. It is validity to use GA to optimize the structure and weight parameters of RBF neural network. However the complexity genetic manipulation (such as selection, crossover, and mutation) of GA causes training time of the neural network increasing exponentially with the increase of the scale and complexity of the problem. To solve these problems, a RBF network algorithm based on QPSO is presented to effectively improve faults diagnosis.The main contributions in this work are summarized at last and further research considerations are put forward.
 Keywords/Search Tags:  Swarm intelligence, optimization technique, particle swarm optimization, quantumbehaved particle swarm optimization, multipeaks optimization, system identification, faults diagnosis  PDF Full Text Request  Related items 
 
