Particle Swarm Optimization Algorithm For Many-Objective Programming And Its Application Research

The many-objective programming refers to the optimization problem with more than four objective functions.Due to the increase of the number of objectives,the traditional evolutionary algorithm based on Pareto domination faces many challenges,such as dominant impedance,computational complexity and visualization.At present,there are four kinds of methods to deal with many-objective programming,including the method based on weak Pareto domination,the method based on decomposition,the method based on indicator and the method based on reference point.They mainly improve the fitness allocation and diversity protection of the algorithm,but they still have shortcomings in the convergence of the algorithm.In order to solve the problem of dominant impedance,a reference point based method is used in this paper and the aggregation function is used to balance the convergence and diversity.The main work of this paper is as follows:Aiming at the problem that multi-objective programming particle swarm optimization algorithm is easy to fall into premature,this paper proposes a multi-objective particle swarm optimization algorithm based on health degree(HMOPSO).First of all,the HMOPSO algorithm computes the times of particle oscillation and optimization stagnation respectively in each iteration.Then update the health value of the population particles use the records.When the health value of the particles is lower than the minimum limit,it is considered unhealthy.Next,a special guiding factor is used to perform mutation on unhealthy particles so as to avoid invalid search of unhealthy particles.Finally,the dynamic maintenance based on crowding sorting is used.Every time the crowding distance of an individual is calculated,one redundant individual is eliminated,which is conducive to the improvement of diversity.The experimental results show that the HMOPSO algorithm can effectively solve the double target test problem,even in the discontinuous or multi-modal problems,it has good convergence and distribution.Aiming at the dominant impedance phenomenon of many-objective evolutionary algorithms,this paper proposes a reference point based particle swarm optimization algorithm(RMa OPSO).Firstly,the RMa OPSO algorithm divides the target space into several subspaces using evenly distributed reference points.Then,according to the distance between the population particles and the reference vector,the particles are associated with different subspaces.Next,calculate the aggregate function value of each particle in the corresponding subspace.Finally,the aggregate value is used to update the individual extremum and local extremum of particles.Guided by individual extremum and local extremum,population particles search the target space.In RMa OPSO algorithm,the selection of aggregation function has a certain impact on the search performance.In this paper,Tchebycheff aggregation function and PBI aggregation function are used,and the influence of aggregation function on Pareto optimal solution set is tested on the three objective test problem.The feasibility of the algorithm is verified by comparing the approximate Pareto front obtained by RMa OPSO and MOEA/D on many-objective problems with 5,10 and 15 objectives.