Research On Trust-Tech Methodology Based On Particle Swarm Optimization And Its Application

Because of the growing use of optimization in science,engineering,economics,and industry,our goal in this thesis is to develop an understanding of optimization algorithms for practitioners.The novel optimization algorithms for solving the unconstrained optimization problem,the optimization problem with the equality constraint and the black box optimization problem are proposed in these theses.These algorithms are applied for solveing the the short term load forecasting,the multiple local solutions of optimal power flow problems,the economic dispatch problems and the optimal selection of SVM parameters problems.A novel three-stage methodology is presented,termed the “Consensus-Based Particle Swarm Optimization-Assisted Trust-Tech Methodology”,for finding global optimal solutions of nonlinear unconstrained optimization problems.It is composed of Trust-Tech methods,consensus-based particle swarm optimization,and local optimization methods integrated to compute a set of high-quality local optimal solutions that can contain the global optimal solution.The proposed methodology compares very favorably with several recently developed optimization algorithms on 20 small-dimension benchmark optimization problems and 20 large-dimension test functions from the CEC 2010 competition.Experimental results demonstrate that the proposed methodology can rapidly obtain high-quality optimal solutions that can contain the global optimal solution.This algorithm is applied to the short term load forecasting and multiple local solutions of optimal power flow problems.A novel theoretical and deterministic methodology,Trajectory-Unify Methodology for systematically solving nonlinear equality-constraint optimization problem with simple bounds is proposed.We present a dynamic system derived from the model of the optimization problem,Augmented Lagrangian Gradient System,and aim at finding its stable equilibrium points,which are proved to be local optimal solutions of the original optimization problem.There are two main phase for solving multiple solutions: approach a local minimum and escape from a local minimum.To improve the performance,the first phase consists of three stage: exact integration of the dynamic system,Pseudo-Transient and local solver.Then,we can escape from the obtained solution,move towards nearby local optimal solutions so as to realize locating multiple local optimal solutions which lie within tier-by-tier stability regions.The methodology is illustrated with several numerical examples with promising results.This algorithm is applied to the economic dispatch problems.A novel Consensus-Based Particle Swarm Optimization-Guided Trust-Tech methodology is developed to overcome the challenges faced by black-box optimization problems involving the following issues:(i)premature converging to a local optimum,(ii)extensive computation efforts,and(iii)low quality solutions.The methodology consists of three stages: consensus-based particle swarm optimization,construction of local surrogate models,and computing multiple high-quality local optimal solutions using the Trust-Tech method.Unlike other meta-heuristic methods or response surface-based approaches,which are stochastic methods and extensive computation efforts,the proposed multi-stage method is deterministic,flexible,and high performance.To show its effectiveness,the proposed methodology is illustrated on several test examples with promising results.This algorithm is applied to the optimal selection of SVM parameters problems.