Font Size: a A A

Research On Chaotic Particle Swarm Optimization Algorithm Ant Its Application

Posted on:2013-02-19Degree:DoctorType:Dissertation
Country:ChinaCandidate:W X XuFull Text:PDF
GTID:1118330374957384Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Due to the simplicity of particle swarm optimization algorithm (PSO) forimplementation, it has provides an effective means for solving complexoptimization problems and has driven attention of scholars of the area ofcontinuous optimization and intelligent computing. Though there has been afew successful application of this method, the achievements of its applicationin the field of chemical process are comparatively less. And there is rarely anyuniversal and effective mechanism for dealing with the constrained anddynamic optimization problems. This paper mainly works on the improvementof the existing PSO algorithms for application in different kinds ofoptimization problems involved in the production process of high densitypolyethylene (HDPE) and different types of optimization problems.To overcome the shortcomings of premature and low accuracy of PSOalgorithm, this paper proposes the strategy of hybridizing it with chaotic mapand traditional gradient based methods, and proposed chaotic PSO algorithmsbased on sequential quadratic programing (SQP), interior point method (IPM)and trust region reflective (TRR) for unconstrained (bound constrained), constrained and dynamic optimization problems respectively. The simulationresults for benchmark functions show that our new proposed algorithms havebetter accuracy, more probability of finding global optimum and faster speedof convergence than those reported in the literature. The feasibility of themethod is illustrated with the challenging ethylene piece yardage, productionquality modeling and grade transition optimization problem of a cascadeHDPE reaction. The main contributions of this work can be summarized asfollows:1. This paper proposes the strategy of hybridizing PSO with chaotic map andtraditional gradient based methods. Using logistic chaotic map and SQP localsearch, a chaotic PSO algorithm based on sequential quadratic optimizationalgorithm (CPSOSQP) is proposed. The simulation results for benchmarkfunctions show the feasibility of the hybridizing strategy.2. To further enhance the convergence speed and robustness of CPSOSQPalgorithm, several improvements on hybridizing strategies have been putforward which lead to two improved chaotic PSO algorithms based on SQPlocal search (CPSO-SQP and CPSO-SQPII), whose advantages have beenshown by the simulation results on unconstrained (bound constrained)benchmark function. CPSO-SQP has been applied in the challenging ethylenepiece yardage optimization problem of a cascade HDPE reaction course, withwhich its feasibility is illustrated.3. Making the best of ergodicity of piecewise linear chaotic map to help constraints handling mechanism based PSO with the global search whileemploying the IPM to accelerate the local search, a novel chaotic PSOalgorithm based on interior point method (CPSO-IPM) is proposed for solvingconstrained optimization problem. Simulation results on eighteen constrainedbenchmarks and three engineering optimization problems from the literaturedemonstrate the effectiveness and robustness of the algorithm. And thefeasibility of the method is illustrated with the challenging production qualitymodeling optimization problem of a cascade HDPE reaction course.4. Employing trust region reflective method based on preconditional conjugategradient to accelerate the local search and combining the chaotic PSO used inCPSO-SQPII, a novel chaotic PSO algorithm based on trust region reflective(CPSO-TRR) is proposed for solving large scale optimization problem. And itis combined with a simultaneous strategy to solve dynamic optimizationproblem. Its application in the slurry high density polyethylene gradetransition optimization demonstrated the effectiveness and practicability of theproposed method.
Keywords/Search Tags:particle swarm optimization, sequential quadraticprograming, interior point, trust region, preconditional conjugate gradient, dynamic optimization, constrainded optimization, unconstrained optimization, grade transition
PDF Full Text Request
Related items