Stored Information Recombination Based Particle Swarm Optimization Algorithm And Its Applications

Optimization problems are regularly encountered in various scientific researches and industrial applications.Consequently,research on optimization technology with high efficiency has great significance.Bionic algorithms are the important optimizers due to their features of less limitation on applications,fast convergence and easy implementation.As one of the remarkable bionic algorithms,Particle Swarm Optimization(PSO)algorithm has been used to solve optimization problems widely due to its simple structure and fast convergence ability.One of the most important features of PSO is the capability to store some valuable information of solutions in early iterations.However,most of the existing PSO algorithms only used the stored information directly to guide the search behavior of each particle in the swarm.Therefore,a potential approach to improve the performance of the PSO algorithm is to make good use of the stored information with selection and some recombination operations to explore the solution space.Some of the existing PSO algorithms attempted to make extra processing and utilization on the stored information of the population and have achieved better performance;however,there is lack of an in-depth study on such approach.This thesis focuses on the improved PSO algorithms based on the extra processing of stored information and their applications on engineering.The main works and contributions of the thesis are as follows:1)The elitist solutions produced in the evolutionary process of PSO have crucial influence on the search for the optimum solution.For single objective PSO algorithm,a stored information recombination method--elitist promotion strategy is proposed.When criteria are met,the personal best solutions of particles are used to reconstruct the new individuals through specified operators.In order to improve the global search ability of the algorithm,the new generated individuals with better fitness values are selected as the new personal best solutions and global best solution.Four operations are used as the elitist promotion operators in this study and their search characteristics are analyzed.After that,elitist promotion strategy is used to improve the standard PSO algorithm and the quantum-behaved PSO algorithm.A comprehensive simulation study is conducted on two sets of benchmark functions to compare the performances of the four elitist promotion operators.Furthermore,to verify the efficiency of the elitist promotion strategy,the improved quantum-behaved PSO algorithm based on elitist promotion with transposon operator(EPQPSO(TRANS))and the improved quantum-behaved PSO algorithm based on elitist promotion with DE operator(EPQPSO(DE))are compared with four existing state-of-the-art algorithms respectively.The proposed two algorithms perform more competitively in their corresponding benchmark test set in terms of global search capability,convergence rate and stability.2)Based on the framework of elitist promotion strategy,a novel double elitist promotion quantum-behaved PSO algorithm(DEP-QPSO-EDP)is proposed for power economic dispatch(ED)problem.The double elitist promotion strategy includes the elitist promotion strategy together with another strategy for improving the solutions qualities with the information of elitists directly.Moreover,to handle the heavy constraints of ED problem,a dedicated efficient heuristic handing technique for repairing the solutions is proposed to decrease the solutions' constraint violations together with a novel cooperative update method for pbest and gbest.The comparative results with other state-of-the-art approaches on three different widely used ED test systems confirm that the proposed DEP-QPSO-EDP algorithm is able to obtain better feasible optimal solutions efficiently and stably for practical ED problems.3)A stored information recombination method based on Bayesian theorem for quantum-behaved PSO algorithm was proposed.The fitness distance between each particle and average personal best position is stored and accumulated to build a memory vector to predict the substantial behavior of each particle.A corresponding coefficient was assigned based on the memory vector to each particle automatically and separately to perform individual guide to the members of the swarm so as to improve the search efficiency of algorithm.The quantum-behaved PSO algorithm based on this method called quantum-behaved PSO with mutation supported by memory(QPSO-MuM)is developed for multilevel thresholding for image segmentation.Simulation results on Berkeley dataset show that the proposed algorithm is superior to three existing PSO-based methods because it helps obtain more stable and clearer image segmentation results.4)The elitist seed repopulation strategy,a novel stored information recombination method for multi-objective optimization is proposed.In this strategy,the elitists in external archive are selected and used to generate the new population with transposon together with some other specified operators.In this way,the premature convergence led by traditional search approach of PSO is overcome.A PSO algorithm for multi-objective optimization,called elitist seed repopulation PSO algorithm with transposon(ESMOPSO-T)is proposed based on this strategy combined with a periodic time variant inertia weight update method which is beneficial for improving the algorithm searching efficiency and accuracy.The performance of ESMOPSO-T is verified on two sets of benchmark problems.Comparing with two state-of-the-art multi-objective genetic algorithms and a state-of-the-art multi-objective PSO algorithm,the simulation results indicate that ESMOPSO-T is highly competitive.Solutions obtained by ESMOPSO-T well approach the Pareto-optimal front and are evenly distributed over the front that cannot be achieved by the compared algorithms.5)According to the characteristics of the deployment problem of wireless sensor networks for real time oilfield monitoring,a multi-objective model for the problem is established which considers both the financial cost,and complexities of construction and maintenance.Based on the framework of ESMOPSO-T,a multi-objective discrete PSO algorithm(ESMOBPSO-T)is proposed to handle this problem.The highlights of the ESMOBPSO-T algorithm include the binary coding scheme catering the problem characteristics,constraint handling method without the needs of parametric control and hybrid search approach for enhancing the search efficiency of solutions.Comparing with three state-of-the-art algorithms,simulation results validate that the proposed ESMOBPSO-T algorithm is superior in locating the Pareto-optimal front and maintaining the diversity of solutions,which is beneficial for the reasonable deployment of the wireless sensor networks for real time oilfield monitoring.