Research On Particle Swarm Optimization Algorithm Based On Swarm Diversity

The optimization problem refers to finding the best feasible solution that makes the optimization goal achieve the maximum or minimum value under certain constraints.Since many optimization problems that exist today are large-scale combinatorial optimization problems,conventional algorithms are difficult to handle due to the high dimensionality and multiple poles of the problems.As one of the algorithms to solve such problems,the PSO has fewer control parameters,faster convergence speed and better search accuracy.Therefore,it has attracted the attention of lots of scholars,which in turn promotes its development in the field of theoretical and practical engineering optimization.Based on a large number of related documents,the author proposes two novel PSO algorithms.In summary,the main content of this paper includes:1.Analyze the background and significance of the PSO algorithm,summarize the recent or classic PSO variants,and point out that the core problem in the PSO algorithm is the premature convergence caused by the loss of swarm diversity.2.Propose a single-swarm PSO based on a novel diversity evaluation strategy.By encoding each subspace in the search space and with the help of hash tables,the diversity of the swarm can be effectively obtained.In addition,this paper presents the concept of exploratory degree based on swarm diversity and evolutionary state to reflects the degree of the demand for dispersion.In order to reduce the waste of the evaluation function,the perturbation update mode(replace the position of the poorer particles with the new positions generated by the best position perturbation)is proposed to increase the efficiency of the swarm search.3.Propose Artificial Multi-Swarm PSO composed of three swarms(exploration swarm,exploitation swarm and convergence swarm)which have different functions by adopting different evolution mechanisms.based on particle position and fitness,this paper also proposes a diversity evaluation mechanism without loss of information to provide a basis for adaptive update of parameters.Finally,in order to accelerate the exploitation swarm to converge to the local optimal solution and assist the convergent swarm to jump out of the local trap in time,this paper proposes the partial swarm reconstruction and the complete swarm reconstruction technology.