Research On Improvement And Application Of Multi-objective Particle Swarm Optimization

Many problems in reality can be summarized as multi-objective optimization problems.Due to the characteristics of multi-objective optimization,such as diversification,nonlinear,high dimension,etc.,the traditional optimization method is not enough to solve the problem.At present,meta-heuristic intelligent algorithms are widely used in multi-objective optimization problems by virtue of their efficient search method.Among them,the particle swarm optimization is favored by the researchers with its advantages of simple implementation,fewer calling parameters and fast convergence speed.However,when solving the multi-objective optimization problems,there are still problems such as falling into the local optimal solution and poor diversity.Therefore,the algorithm also needs to improve its performance from the purpose of convergence,diversity,balanced convergence and diversity.This paper proposes the following methods for the above problems:(1)Multi-objective particle swarm optimization based on the hypercube and the distance(HDMOPSO)is proposed.The global optimal selection in the algorithm adopts the combination strategy of hypercube and distance,which enables the algorithm to accelerate convergence.In addition,to prevent the algorithm from falling into the local optima,a nonlinear decreasing back mutation strategy was used to enable the particles to explore more regions.Finally,a control strategy is used to perform archive maintenance and improve the convergence and diversity of the algorithm.(2)Multi-objective particle swarm optimization with dynamic population size(D-MOPSO)is proposed,so that the population size varies with the existing resources and is no longer an estimated population size.The population size in this paper increases or decreases according to the existing population resources,which can balance the convergence and diversity of the solutions.On the one hand,the particles are added according to the local perturbations to improve the population exploration capability.On the other hand,the non-dominated ranking values and population density information are used to control the population size to prevent excessive population size growth.This paper experiments the improved algorithms on three sets of benchmark test functions and compares their with seven algorithms such as MOPSO and NSGA-III.The experimental results show that the algorithm shows effectiveness in solving multi-objective optimization problems and is competitive with most algorithms.In addition,the improved multi-objective particle swarm optimization algorithm is applied to healthy diet.So that people from the diet intake of carbohydrates,protein,fat nutritional elements content to meet the health standard,and then the daily diet control,help to maintain health.