Study Accelerate The Convergene Of Operator Of Multi Objective Optimization

There are a variety of attributes about optimization problcems an the real worOfld.Generally the multiple targets were optimized simultaneously.In most cases,the objectives which were optimized are conflict with each other.To achieve the op timization of the overall objective,the conflicting objectives usu ally needs to be taken into account.That is the sub-objective must be compromised,Therefore,the mlti-objective optimization algorihm was p roposed for multi-objective optimization problem.There are several kinds of multi-objective optimizationo algorithms.The multi-objective particle swarm optimization is widely applied and relatively fast developed at present.Such the characteristics of multi-objective particle swarm algorithm was studied to acceler ate convergence of multi-ebjective particle swarm optimization in this paper.Some of these mehods can be used in any of a multi-o bjective optimizat ion algorithm.There are two most important factors that affect convergence speed in a multi-objective particle swarm algorithm.The first is mutation strategy.While using multi-obj ective particle swarm optimization to solve the multi-objective optimization problem,it is easy to fall into the wrung Pareto frontier and affect the convergenee speed to the true Pareto front.Now most scholars add mutation in the population space to improve the population diversity,such the algorithm is not easy to fall into the wrong,Pareto frontier.Howevevre,not all the mutation is effective.As a result,there is space to improve the convergence speed.The second is the structure of non-dominated solutions and the update of archiving sets.In the multi-objective optim i zation algorithm,a current population of non-do mi nated solutio n set must be constructed at the cevolution of each generation.For the multi-objective optimization algorithm thtat the external archive set is constructed,after get the current set of non-dominated solutions,the update of archinve set operations is a must.Large amount of ca lculation is needed.If the amount of calculation about the above two aspects can be reduced,the efficiency of the algorithm will greatly improved.In this paper,the above two aspects of multi-objective particle swarm optimization will be improved.The main research contents includes:(1)In the population space of multi-objective particle swarm optimization.According to the current value of the particles in each dimension,then make a decision whether it corresponds to the global optimum and individual particles in the corresponding optimal particle dimension is equal to accurately locate whether the mutation to improve the traditional random mutation strategy.To adapt to solve the multi-objective optimization problem,improve search efficiency of particle swarm optimization,overcome the shortcomings that traditional particle swarm optimization algorithm for solving multi-objective optimization problem easy to fall into local optima and improve the convergence speed(2)According to the variance in each dimension of all particles in archive,adaptive mutation is used,in the large variance dimensional with a large mutation probability while on a small variance dimensional with a small mutation probability.Thus the diversity of the particles in the archive was increased.As a result,the diversity of the Pareto frontier is further increased(3)Using a new set of non-dominated solutions constructor and the update of archiving set to reduce the time costs of traditional structure and updates.Besides,the efficiency of the algorithm improved.