The Learning Strategy Of Particle Swarm Optimization In Complex Function Optimization And Its Improvement

Particle Swarm Optimization algorithm(PSO)is an intelligent optimization which has been widely concerned and studied in recent years.With few parameters,simple process and good search results,it has gained more recognition and application.However,some existing PSO algorithms are often difficult to balance the population diversity and convergence accuracy,and easily trapped in local optimum,especially in complex multimodal function and large scale function optimization.There are a lot of problems to improve,when it is applied in real world problems.On the basis of previous research results,this paper propose some improved methods for the problem of premature convergence and easy to get into local optimization.The main work are as follows:1.This paper propose a self-adaptive multi-swarm particle swarm optimization algorithm(SMPSO).In this SMPSO,a self-adaptive regrouping operator is proposed to reinforce the population diversity and to void premature convergence.In addition,particles' historical information are used to direct the best solution to carry out a detecting operator and to increase the ability to escape a local optimal.To accelerate convergence speed and improve solutions' accuracy of PSO,two local search strategies are proposed.The comparisons of SMPSO with other five PSO algorithms on some benchmark functions and an engineering application indicate that the proposed strategies can effectively enhance ability of escaping local optimal solution,speed up the convergence and raised solutions' accuracy.2.We propose a Sophisticated PSO(SopPSO)based on multi-level adaptation and purposeful detection.In SopPSO,a particle not only updates its learning model according to its fitness landscape,but also periodically re-selects target dimensions that the particle learns from its neighbors.The adaptive strategy applied in multi-level endows population with a more favorable diversity.In addition,a tabu detecting and a local searching strategies base on some historical information are proposed to help the population to jump out of local optima and improve the accuracy of solutions,respectively.The extensive experimental results illustrate the effectiveness and efficiency of the proposed strategies.Furthermore,the comparison results between SopPSO and other per algorithms on different problems verify its favorable performance on unimodal,multimodal and large-scale problems as well as some real applications.