Research On Balance Convergence And Distribution Particle Swarm Optimization Algorithms For Dynamic Multi-objective Problems

Dynamic multi-objective optimization problems are widely existed in many projects of scientific research and reality.Researching how to solve such dynamic target optimization problems has extremely important scientific and practical significance for promoting the development of dynamic multi-objective optimization.At present,the particle swarm optimization algorithm is applied to solve the static optimization problem with its unique fast convergence advantage.However,when the algorithm is applied to solve dynamic multiobjective optimization problems,it is difficult for particle swarm optimization to converge to the frontier because of the conflict between multiple targets,the variability of problems over time,and the inherent vulnerability of particle swarms.The characteristics of local optimization,the convergence and distribution of the equilibrium particle swarm optimization algorithm are urgently needed,so as to improve the description ability and tracking ability of the algorithm in the dynamic environment.However,today's strategies tend to emphasize distribution or convergence,only analyzing the target space,and lacking the ability to quantify the degree of environmental change.These problems have brought severe challenges to the optimization of dynamic multi-objective problems by particle swarm optimization algorithms.Based on the above facts,this thesis designs a strategy of balance convergence and distribution for the problem that the particle swarm optimization algorithm is easy to fall into local optimum,resulting in convergence and distribution imbalance.Characteristic,a new swarm prediction strategy and two-way selection strategy are designed.Finally,a dynamic multi-objective particle swarm optimization algorithm with balanced distribution and convergence is proposed.Specific work and innovations are as follows:(1)A convergence and distribution strategy for balance algorithm is proposed.Aiming at the problem that the particle swarm optimization algorithm is easy to fall into the local optimum in the evolution process,the decomposition strategy is first introduced into the algorithm,and then the algorithm is used to judge whether the population falls into the local optimal state according to the algorithm,and the opposite learning with enhanced convergence is designed.Strategy,if the population is trapped,the opposite learning strategy is used to make it jump out of the local optimum and converge on the forward side.At the same time,a weakly dominated archive set update strategy is designed to maintain the archive set from a global perspective,thus realizing the external archive set.Effective maintenance and update to effectively maintain the distribution of the population.Finally,experiments show that the algorithm based on opposite learning and weak dominance archive set is applied to solve 9 standard test functions.Compared with the four comparison algorithms,the convergence and distribution of the balance algorithm are better,and improve the performance of the set..(2)A dynamic multi-objective particle swarm optimization algorithm based on prediction strategy for balance convergence and distribution is proposed.Based on the previous chapter's equilibrium strategy,aiming at the characteristics of dynamic multi-objective optimization problems,a population prediction strategy is firstly designed to improve the adaptation and response of the algorithm to different dynamic problems by quantifying the severity of environmental changes.Secondly,by transforming the individual positions at different times in the same convergence direction into the time series in the target space,a time series prediction method is introduced,and the next moment position is predicted by the autoregressive model,thereby improving the diversity and effectiveness of the predicted population.Then,in order to avoid the mismatch between the individual solution and the subproblem after the change of the problem,a two-way selection strategy is designed to improve the rationality of the solution and sub-question selection,so as to effectively maintain the diversity of the population in the evolution process.The experimental results show that when the proposed strategy is applied to solve six standard dynamic multi-objective optimization problems,compared with the selected four comparison algorithms,it has obvious advantages in terms of convergence,distribution and stability.