Improvement And Research Of Multi-objective Partic Le Swarm Optimization Algorithm

With the widespread application in real life and the driving of the actual production problems,the multi-objective optimization problems have been received more and more attention in the engineering practical field and academic field,and the methods of solving the problem are also increasing.Nowadays the intelligent optimization algor ithm is one of the hottest researches,which is successfully applied to the production management,power control and other practical problems.Particle Swarm Optimization(PSO),as a typical group intelligent optimization algorithm,is proposed on the basis of the predation of birds.PSO algorithm is widely used in solving the optimization problems and the actual engineering production problems because of its advantages,such as less parameter,easy to implement,fast convergence and so on.However,Particle Swarm Optimization(PSO)algorithm still has the problem of low stability and poor diversity of solutions when solving multi-objective optimization problems.In order to overcome these shortcomings,three improvement strategies are proposed to solve the multi-objective optimization problems.(1)Dynamic acceleration coefficients based on sine transformation: according to the characteristics of the time-domain dynamic transformation of the sine function,the algorithm has a large c1 in the early search stage and the self-consciousness of the particles is enhanced to increase the diversity of the individuals.In the later per iod,the large c2 enhance the social cognition ability so that the population can move to the real Pareto optimal front as soon as possible;(2)Dynamic displacement fluctuation operator: inspired by the movement and the dynamic cyclotron which may occur in the actual flight of the birds,a depth optimization of the personal best position is carried out to improve the approximation degree of the solutions and the real Pareto optimal front;(3)Mutation operator generated by modified Lévy flight: by increasing the randomness of the algor ithm based on the modified Lévy flight,the population is subjected to non-uniform mutation operation with random probability to increase the diversity of Pareto solution set,which leads to more possible solutions and avoid falling into local optimum.Then the validity of these three improved strategies is verified by simulation experiments.Finally,in terms of the combination of the above three improvement strategies,an improved multi-objective particle swarm optimization(IMOPSO)algorithm is proposed.Then seven classical test functions are validated and compared with the results of MOPSO.The simulation results illustrate that the performances of IMOPSO algorithm have been improved in the perspective of convergence,diversity and uniformity.